1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2016 Intel Corporation.
5 
6   This program is free software; you can redistribute it and/or modify it
7   under the terms and conditions of the GNU General Public License,
8   version 2, as published by the Free Software Foundation.
9 
10   This program is distributed in the hope it will be useful, but WITHOUT
11   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13   more details.
14 
15   You should have received a copy of the GNU General Public License along with
16   this program; if not, write to the Free Software Foundation, Inc.,
17   51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 
19   The full GNU General Public License is included in this distribution in
20   the file called "COPYING".
21 
22   Contact Information:
23   Linux NICS <linux.nics@intel.com>
24   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 
27 *******************************************************************************/
28 
29 #include <linux/types.h>
30 #include <linux/module.h>
31 #include <linux/pci.h>
32 #include <linux/netdevice.h>
33 #include <linux/vmalloc.h>
34 #include <linux/string.h>
35 #include <linux/in.h>
36 #include <linux/interrupt.h>
37 #include <linux/ip.h>
38 #include <linux/tcp.h>
39 #include <linux/sctp.h>
40 #include <linux/pkt_sched.h>
41 #include <linux/ipv6.h>
42 #include <linux/slab.h>
43 #include <net/checksum.h>
44 #include <net/ip6_checksum.h>
45 #include <linux/etherdevice.h>
46 #include <linux/ethtool.h>
47 #include <linux/if.h>
48 #include <linux/if_vlan.h>
49 #include <linux/if_macvlan.h>
50 #include <linux/if_bridge.h>
51 #include <linux/prefetch.h>
52 #include <linux/bpf.h>
53 #include <linux/bpf_trace.h>
54 #include <linux/atomic.h>
55 #include <scsi/fc/fc_fcoe.h>
56 #include <net/udp_tunnel.h>
57 #include <net/pkt_cls.h>
58 #include <net/tc_act/tc_gact.h>
59 #include <net/tc_act/tc_mirred.h>
60 #include <net/vxlan.h>
61 #include <net/mpls.h>
62 
63 #include "ixgbe.h"
64 #include "ixgbe_common.h"
65 #include "ixgbe_dcb_82599.h"
66 #include "ixgbe_sriov.h"
67 #include "ixgbe_model.h"
68 
69 char ixgbe_driver_name[] = "ixgbe";
70 static const char ixgbe_driver_string[] =
71 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
72 #ifdef IXGBE_FCOE
73 char ixgbe_default_device_descr[] =
74 			      "Intel(R) 10 Gigabit Network Connection";
75 #else
76 static char ixgbe_default_device_descr[] =
77 			      "Intel(R) 10 Gigabit Network Connection";
78 #endif
79 #define DRV_VERSION "5.1.0-k"
80 const char ixgbe_driver_version[] = DRV_VERSION;
81 static const char ixgbe_copyright[] =
82 				"Copyright (c) 1999-2016 Intel Corporation.";
83 
84 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";
85 
86 static const struct ixgbe_info *ixgbe_info_tbl[] = {
87 	[board_82598]		= &ixgbe_82598_info,
88 	[board_82599]		= &ixgbe_82599_info,
89 	[board_X540]		= &ixgbe_X540_info,
90 	[board_X550]		= &ixgbe_X550_info,
91 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
92 	[board_x550em_x_fw]	= &ixgbe_x550em_x_fw_info,
93 	[board_x550em_a]	= &ixgbe_x550em_a_info,
94 	[board_x550em_a_fw]	= &ixgbe_x550em_a_fw_info,
95 };
96 
97 /* ixgbe_pci_tbl - PCI Device ID Table
98  *
99  * Wildcard entries (PCI_ANY_ID) should come last
100  * Last entry must be all 0s
101  *
102  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
103  *   Class, Class Mask, private data (not used) }
104  */
105 static const struct pci_device_id ixgbe_pci_tbl[] = {
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
133 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
134 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
135 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
136 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
137 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
138 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
139 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_XFI), board_X550EM_x},
140 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
141 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
142 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
143 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_1G_T), board_x550em_x_fw},
144 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
145 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
146 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
147 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
148 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
149 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
150 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
151 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
152 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
153 	/* required last entry */
154 	{0, }
155 };
156 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
157 
158 #ifdef CONFIG_IXGBE_DCA
159 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
160 			    void *p);
161 static struct notifier_block dca_notifier = {
162 	.notifier_call = ixgbe_notify_dca,
163 	.next          = NULL,
164 	.priority      = 0
165 };
166 #endif
167 
168 #ifdef CONFIG_PCI_IOV
169 static unsigned int max_vfs;
170 module_param(max_vfs, uint, 0);
171 MODULE_PARM_DESC(max_vfs,
172 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
173 #endif /* CONFIG_PCI_IOV */
174 
175 static unsigned int allow_unsupported_sfp;
176 module_param(allow_unsupported_sfp, uint, 0);
177 MODULE_PARM_DESC(allow_unsupported_sfp,
178 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
179 
180 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
181 static int debug = -1;
182 module_param(debug, int, 0);
183 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
184 
185 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
186 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
187 MODULE_LICENSE("GPL");
188 MODULE_VERSION(DRV_VERSION);
189 
190 static struct workqueue_struct *ixgbe_wq;
191 
192 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
193 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
194 
195 static const struct net_device_ops ixgbe_netdev_ops;
196 
197 static bool netif_is_ixgbe(struct net_device *dev)
198 {
199 	return dev && (dev->netdev_ops == &ixgbe_netdev_ops);
200 }
201 
202 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
203 					  u32 reg, u16 *value)
204 {
205 	struct pci_dev *parent_dev;
206 	struct pci_bus *parent_bus;
207 
208 	parent_bus = adapter->pdev->bus->parent;
209 	if (!parent_bus)
210 		return -1;
211 
212 	parent_dev = parent_bus->self;
213 	if (!parent_dev)
214 		return -1;
215 
216 	if (!pci_is_pcie(parent_dev))
217 		return -1;
218 
219 	pcie_capability_read_word(parent_dev, reg, value);
220 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
221 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
222 		return -1;
223 	return 0;
224 }
225 
226 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
227 {
228 	struct ixgbe_hw *hw = &adapter->hw;
229 	u16 link_status = 0;
230 	int err;
231 
232 	hw->bus.type = ixgbe_bus_type_pci_express;
233 
234 	/* Get the negotiated link width and speed from PCI config space of the
235 	 * parent, as this device is behind a switch
236 	 */
237 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
238 
239 	/* assume caller will handle error case */
240 	if (err)
241 		return err;
242 
243 	hw->bus.width = ixgbe_convert_bus_width(link_status);
244 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
245 
246 	return 0;
247 }
248 
249 /**
250  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
251  * @hw: hw specific details
252  *
253  * This function is used by probe to determine whether a device's PCI-Express
254  * bandwidth details should be gathered from the parent bus instead of from the
255  * device. Used to ensure that various locations all have the correct device ID
256  * checks.
257  */
258 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
259 {
260 	switch (hw->device_id) {
261 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
262 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
263 		return true;
264 	default:
265 		return false;
266 	}
267 }
268 
269 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
270 				     int expected_gts)
271 {
272 	struct ixgbe_hw *hw = &adapter->hw;
273 	int max_gts = 0;
274 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
275 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
276 	struct pci_dev *pdev;
277 
278 	/* Some devices are not connected over PCIe and thus do not negotiate
279 	 * speed. These devices do not have valid bus info, and thus any report
280 	 * we generate may not be correct.
281 	 */
282 	if (hw->bus.type == ixgbe_bus_type_internal)
283 		return;
284 
285 	/* determine whether to use the parent device */
286 	if (ixgbe_pcie_from_parent(&adapter->hw))
287 		pdev = adapter->pdev->bus->parent->self;
288 	else
289 		pdev = adapter->pdev;
290 
291 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
292 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
293 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
294 		return;
295 	}
296 
297 	switch (speed) {
298 	case PCIE_SPEED_2_5GT:
299 		/* 8b/10b encoding reduces max throughput by 20% */
300 		max_gts = 2 * width;
301 		break;
302 	case PCIE_SPEED_5_0GT:
303 		/* 8b/10b encoding reduces max throughput by 20% */
304 		max_gts = 4 * width;
305 		break;
306 	case PCIE_SPEED_8_0GT:
307 		/* 128b/130b encoding reduces throughput by less than 2% */
308 		max_gts = 8 * width;
309 		break;
310 	default:
311 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
312 		return;
313 	}
314 
315 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
316 		   max_gts);
317 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
318 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
319 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
320 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
321 		    "Unknown"),
322 		   width,
323 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
324 		    speed == PCIE_SPEED_5_0GT ? "20%" :
325 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
326 		    "Unknown"));
327 
328 	if (max_gts < expected_gts) {
329 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
330 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
331 			expected_gts);
332 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
333 	}
334 }
335 
336 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
337 {
338 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
339 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
340 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
341 		queue_work(ixgbe_wq, &adapter->service_task);
342 }
343 
344 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
345 {
346 	struct ixgbe_adapter *adapter = hw->back;
347 
348 	if (!hw->hw_addr)
349 		return;
350 	hw->hw_addr = NULL;
351 	e_dev_err("Adapter removed\n");
352 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
353 		ixgbe_service_event_schedule(adapter);
354 }
355 
356 static u32 ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
357 {
358 	u8 __iomem *reg_addr;
359 	u32 value;
360 	int i;
361 
362 	reg_addr = READ_ONCE(hw->hw_addr);
363 	if (ixgbe_removed(reg_addr))
364 		return IXGBE_FAILED_READ_REG;
365 
366 	/* Register read of 0xFFFFFFF can indicate the adapter has been removed,
367 	 * so perform several status register reads to determine if the adapter
368 	 * has been removed.
369 	 */
370 	for (i = 0; i < IXGBE_FAILED_READ_RETRIES; i++) {
371 		value = readl(reg_addr + IXGBE_STATUS);
372 		if (value != IXGBE_FAILED_READ_REG)
373 			break;
374 		mdelay(3);
375 	}
376 
377 	if (value == IXGBE_FAILED_READ_REG)
378 		ixgbe_remove_adapter(hw);
379 	else
380 		value = readl(reg_addr + reg);
381 	return value;
382 }
383 
384 /**
385  * ixgbe_read_reg - Read from device register
386  * @hw: hw specific details
387  * @reg: offset of register to read
388  *
389  * Returns : value read or IXGBE_FAILED_READ_REG if removed
390  *
391  * This function is used to read device registers. It checks for device
392  * removal by confirming any read that returns all ones by checking the
393  * status register value for all ones. This function avoids reading from
394  * the hardware if a removal was previously detected in which case it
395  * returns IXGBE_FAILED_READ_REG (all ones).
396  */
397 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
398 {
399 	u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
400 	u32 value;
401 
402 	if (ixgbe_removed(reg_addr))
403 		return IXGBE_FAILED_READ_REG;
404 	if (unlikely(hw->phy.nw_mng_if_sel &
405 		     IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
406 		struct ixgbe_adapter *adapter;
407 		int i;
408 
409 		for (i = 0; i < 200; ++i) {
410 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
411 			if (likely(!value))
412 				goto writes_completed;
413 			if (value == IXGBE_FAILED_READ_REG) {
414 				ixgbe_remove_adapter(hw);
415 				return IXGBE_FAILED_READ_REG;
416 			}
417 			udelay(5);
418 		}
419 
420 		adapter = hw->back;
421 		e_warn(hw, "register writes incomplete %08x\n", value);
422 	}
423 
424 writes_completed:
425 	value = readl(reg_addr + reg);
426 	if (unlikely(value == IXGBE_FAILED_READ_REG))
427 		value = ixgbe_check_remove(hw, reg);
428 	return value;
429 }
430 
431 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
432 {
433 	u16 value;
434 
435 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
436 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
437 		ixgbe_remove_adapter(hw);
438 		return true;
439 	}
440 	return false;
441 }
442 
443 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
444 {
445 	struct ixgbe_adapter *adapter = hw->back;
446 	u16 value;
447 
448 	if (ixgbe_removed(hw->hw_addr))
449 		return IXGBE_FAILED_READ_CFG_WORD;
450 	pci_read_config_word(adapter->pdev, reg, &value);
451 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
452 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
453 		return IXGBE_FAILED_READ_CFG_WORD;
454 	return value;
455 }
456 
457 #ifdef CONFIG_PCI_IOV
458 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
459 {
460 	struct ixgbe_adapter *adapter = hw->back;
461 	u32 value;
462 
463 	if (ixgbe_removed(hw->hw_addr))
464 		return IXGBE_FAILED_READ_CFG_DWORD;
465 	pci_read_config_dword(adapter->pdev, reg, &value);
466 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
467 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
468 		return IXGBE_FAILED_READ_CFG_DWORD;
469 	return value;
470 }
471 #endif /* CONFIG_PCI_IOV */
472 
473 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
474 {
475 	struct ixgbe_adapter *adapter = hw->back;
476 
477 	if (ixgbe_removed(hw->hw_addr))
478 		return;
479 	pci_write_config_word(adapter->pdev, reg, value);
480 }
481 
482 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
483 {
484 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
485 
486 	/* flush memory to make sure state is correct before next watchdog */
487 	smp_mb__before_atomic();
488 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
489 }
490 
491 struct ixgbe_reg_info {
492 	u32 ofs;
493 	char *name;
494 };
495 
496 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
497 
498 	/* General Registers */
499 	{IXGBE_CTRL, "CTRL"},
500 	{IXGBE_STATUS, "STATUS"},
501 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
502 
503 	/* Interrupt Registers */
504 	{IXGBE_EICR, "EICR"},
505 
506 	/* RX Registers */
507 	{IXGBE_SRRCTL(0), "SRRCTL"},
508 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
509 	{IXGBE_RDLEN(0), "RDLEN"},
510 	{IXGBE_RDH(0), "RDH"},
511 	{IXGBE_RDT(0), "RDT"},
512 	{IXGBE_RXDCTL(0), "RXDCTL"},
513 	{IXGBE_RDBAL(0), "RDBAL"},
514 	{IXGBE_RDBAH(0), "RDBAH"},
515 
516 	/* TX Registers */
517 	{IXGBE_TDBAL(0), "TDBAL"},
518 	{IXGBE_TDBAH(0), "TDBAH"},
519 	{IXGBE_TDLEN(0), "TDLEN"},
520 	{IXGBE_TDH(0), "TDH"},
521 	{IXGBE_TDT(0), "TDT"},
522 	{IXGBE_TXDCTL(0), "TXDCTL"},
523 
524 	/* List Terminator */
525 	{ .name = NULL }
526 };
527 
528 
529 /*
530  * ixgbe_regdump - register printout routine
531  */
532 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
533 {
534 	int i;
535 	char rname[16];
536 	u32 regs[64];
537 
538 	switch (reginfo->ofs) {
539 	case IXGBE_SRRCTL(0):
540 		for (i = 0; i < 64; i++)
541 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
542 		break;
543 	case IXGBE_DCA_RXCTRL(0):
544 		for (i = 0; i < 64; i++)
545 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
546 		break;
547 	case IXGBE_RDLEN(0):
548 		for (i = 0; i < 64; i++)
549 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
550 		break;
551 	case IXGBE_RDH(0):
552 		for (i = 0; i < 64; i++)
553 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
554 		break;
555 	case IXGBE_RDT(0):
556 		for (i = 0; i < 64; i++)
557 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
558 		break;
559 	case IXGBE_RXDCTL(0):
560 		for (i = 0; i < 64; i++)
561 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
562 		break;
563 	case IXGBE_RDBAL(0):
564 		for (i = 0; i < 64; i++)
565 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
566 		break;
567 	case IXGBE_RDBAH(0):
568 		for (i = 0; i < 64; i++)
569 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
570 		break;
571 	case IXGBE_TDBAL(0):
572 		for (i = 0; i < 64; i++)
573 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
574 		break;
575 	case IXGBE_TDBAH(0):
576 		for (i = 0; i < 64; i++)
577 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
578 		break;
579 	case IXGBE_TDLEN(0):
580 		for (i = 0; i < 64; i++)
581 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
582 		break;
583 	case IXGBE_TDH(0):
584 		for (i = 0; i < 64; i++)
585 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
586 		break;
587 	case IXGBE_TDT(0):
588 		for (i = 0; i < 64; i++)
589 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
590 		break;
591 	case IXGBE_TXDCTL(0):
592 		for (i = 0; i < 64; i++)
593 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
594 		break;
595 	default:
596 		pr_info("%-15s %08x\n",
597 			reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
598 		return;
599 	}
600 
601 	i = 0;
602 	while (i < 64) {
603 		int j;
604 		char buf[9 * 8 + 1];
605 		char *p = buf;
606 
607 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
608 		for (j = 0; j < 8; j++)
609 			p += sprintf(p, " %08x", regs[i++]);
610 		pr_err("%-15s%s\n", rname, buf);
611 	}
612 
613 }
614 
615 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
616 {
617 	struct ixgbe_tx_buffer *tx_buffer;
618 
619 	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
620 	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
621 		n, ring->next_to_use, ring->next_to_clean,
622 		(u64)dma_unmap_addr(tx_buffer, dma),
623 		dma_unmap_len(tx_buffer, len),
624 		tx_buffer->next_to_watch,
625 		(u64)tx_buffer->time_stamp);
626 }
627 
628 /*
629  * ixgbe_dump - Print registers, tx-rings and rx-rings
630  */
631 static void ixgbe_dump(struct ixgbe_adapter *adapter)
632 {
633 	struct net_device *netdev = adapter->netdev;
634 	struct ixgbe_hw *hw = &adapter->hw;
635 	struct ixgbe_reg_info *reginfo;
636 	int n = 0;
637 	struct ixgbe_ring *ring;
638 	struct ixgbe_tx_buffer *tx_buffer;
639 	union ixgbe_adv_tx_desc *tx_desc;
640 	struct my_u0 { u64 a; u64 b; } *u0;
641 	struct ixgbe_ring *rx_ring;
642 	union ixgbe_adv_rx_desc *rx_desc;
643 	struct ixgbe_rx_buffer *rx_buffer_info;
644 	int i = 0;
645 
646 	if (!netif_msg_hw(adapter))
647 		return;
648 
649 	/* Print netdevice Info */
650 	if (netdev) {
651 		dev_info(&adapter->pdev->dev, "Net device Info\n");
652 		pr_info("Device Name     state            "
653 			"trans_start\n");
654 		pr_info("%-15s %016lX %016lX\n",
655 			netdev->name,
656 			netdev->state,
657 			dev_trans_start(netdev));
658 	}
659 
660 	/* Print Registers */
661 	dev_info(&adapter->pdev->dev, "Register Dump\n");
662 	pr_info(" Register Name   Value\n");
663 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
664 	     reginfo->name; reginfo++) {
665 		ixgbe_regdump(hw, reginfo);
666 	}
667 
668 	/* Print TX Ring Summary */
669 	if (!netdev || !netif_running(netdev))
670 		return;
671 
672 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
673 	pr_info(" %s     %s              %s        %s\n",
674 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
675 		"leng", "ntw", "timestamp");
676 	for (n = 0; n < adapter->num_tx_queues; n++) {
677 		ring = adapter->tx_ring[n];
678 		ixgbe_print_buffer(ring, n);
679 	}
680 
681 	for (n = 0; n < adapter->num_xdp_queues; n++) {
682 		ring = adapter->xdp_ring[n];
683 		ixgbe_print_buffer(ring, n);
684 	}
685 
686 	/* Print TX Rings */
687 	if (!netif_msg_tx_done(adapter))
688 		goto rx_ring_summary;
689 
690 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
691 
692 	/* Transmit Descriptor Formats
693 	 *
694 	 * 82598 Advanced Transmit Descriptor
695 	 *   +--------------------------------------------------------------+
696 	 * 0 |         Buffer Address [63:0]                                |
697 	 *   +--------------------------------------------------------------+
698 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
699 	 *   +--------------------------------------------------------------+
700 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
701 	 *
702 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
703 	 *   +--------------------------------------------------------------+
704 	 * 0 |                          RSV [63:0]                          |
705 	 *   +--------------------------------------------------------------+
706 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
707 	 *   +--------------------------------------------------------------+
708 	 *   63                       36 35   32 31                         0
709 	 *
710 	 * 82599+ Advanced Transmit Descriptor
711 	 *   +--------------------------------------------------------------+
712 	 * 0 |         Buffer Address [63:0]                                |
713 	 *   +--------------------------------------------------------------+
714 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
715 	 *   +--------------------------------------------------------------+
716 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
717 	 *
718 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
719 	 *   +--------------------------------------------------------------+
720 	 * 0 |                          RSV [63:0]                          |
721 	 *   +--------------------------------------------------------------+
722 	 * 8 |            RSV           |  STA  |           RSV             |
723 	 *   +--------------------------------------------------------------+
724 	 *   63                       36 35   32 31                         0
725 	 */
726 
727 	for (n = 0; n < adapter->num_tx_queues; n++) {
728 		ring = adapter->tx_ring[n];
729 		pr_info("------------------------------------\n");
730 		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
731 		pr_info("------------------------------------\n");
732 		pr_info("%s%s    %s              %s        %s          %s\n",
733 			"T [desc]     [address 63:0  ] ",
734 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
735 			"leng", "ntw", "timestamp", "bi->skb");
736 
737 		for (i = 0; ring->desc && (i < ring->count); i++) {
738 			tx_desc = IXGBE_TX_DESC(ring, i);
739 			tx_buffer = &ring->tx_buffer_info[i];
740 			u0 = (struct my_u0 *)tx_desc;
741 			if (dma_unmap_len(tx_buffer, len) > 0) {
742 				const char *ring_desc;
743 
744 				if (i == ring->next_to_use &&
745 				    i == ring->next_to_clean)
746 					ring_desc = " NTC/U";
747 				else if (i == ring->next_to_use)
748 					ring_desc = " NTU";
749 				else if (i == ring->next_to_clean)
750 					ring_desc = " NTC";
751 				else
752 					ring_desc = "";
753 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
754 					i,
755 					le64_to_cpu(u0->a),
756 					le64_to_cpu(u0->b),
757 					(u64)dma_unmap_addr(tx_buffer, dma),
758 					dma_unmap_len(tx_buffer, len),
759 					tx_buffer->next_to_watch,
760 					(u64)tx_buffer->time_stamp,
761 					tx_buffer->skb,
762 					ring_desc);
763 
764 				if (netif_msg_pktdata(adapter) &&
765 				    tx_buffer->skb)
766 					print_hex_dump(KERN_INFO, "",
767 						DUMP_PREFIX_ADDRESS, 16, 1,
768 						tx_buffer->skb->data,
769 						dma_unmap_len(tx_buffer, len),
770 						true);
771 			}
772 		}
773 	}
774 
775 	/* Print RX Rings Summary */
776 rx_ring_summary:
777 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
778 	pr_info("Queue [NTU] [NTC]\n");
779 	for (n = 0; n < adapter->num_rx_queues; n++) {
780 		rx_ring = adapter->rx_ring[n];
781 		pr_info("%5d %5X %5X\n",
782 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
783 	}
784 
785 	/* Print RX Rings */
786 	if (!netif_msg_rx_status(adapter))
787 		return;
788 
789 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
790 
791 	/* Receive Descriptor Formats
792 	 *
793 	 * 82598 Advanced Receive Descriptor (Read) Format
794 	 *    63                                           1        0
795 	 *    +-----------------------------------------------------+
796 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
797 	 *    +----------------------------------------------+------+
798 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
799 	 *    +-----------------------------------------------------+
800 	 *
801 	 *
802 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
803 	 *
804 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
805 	 *   +------------------------------------------------------+
806 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
807 	 *   | Packet   | IP     |   |          |     | Type | Type |
808 	 *   | Checksum | Ident  |   |          |     |      |      |
809 	 *   +------------------------------------------------------+
810 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
811 	 *   +------------------------------------------------------+
812 	 *   63       48 47    32 31            20 19               0
813 	 *
814 	 * 82599+ Advanced Receive Descriptor (Read) Format
815 	 *    63                                           1        0
816 	 *    +-----------------------------------------------------+
817 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
818 	 *    +----------------------------------------------+------+
819 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
820 	 *    +-----------------------------------------------------+
821 	 *
822 	 *
823 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
824 	 *
825 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
826 	 *   +------------------------------------------------------+
827 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
828 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
829 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
830 	 *   +------------------------------------------------------+
831 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
832 	 *   +------------------------------------------------------+
833 	 *   63       48 47    32 31          20 19                 0
834 	 */
835 
836 	for (n = 0; n < adapter->num_rx_queues; n++) {
837 		rx_ring = adapter->rx_ring[n];
838 		pr_info("------------------------------------\n");
839 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
840 		pr_info("------------------------------------\n");
841 		pr_info("%s%s%s\n",
842 			"R  [desc]      [ PktBuf     A0] ",
843 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
844 			"<-- Adv Rx Read format");
845 		pr_info("%s%s%s\n",
846 			"RWB[desc]      [PcsmIpSHl PtRs] ",
847 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
848 			"<-- Adv Rx Write-Back format");
849 
850 		for (i = 0; i < rx_ring->count; i++) {
851 			const char *ring_desc;
852 
853 			if (i == rx_ring->next_to_use)
854 				ring_desc = " NTU";
855 			else if (i == rx_ring->next_to_clean)
856 				ring_desc = " NTC";
857 			else
858 				ring_desc = "";
859 
860 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
861 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
862 			u0 = (struct my_u0 *)rx_desc;
863 			if (rx_desc->wb.upper.length) {
864 				/* Descriptor Done */
865 				pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
866 					i,
867 					le64_to_cpu(u0->a),
868 					le64_to_cpu(u0->b),
869 					rx_buffer_info->skb,
870 					ring_desc);
871 			} else {
872 				pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
873 					i,
874 					le64_to_cpu(u0->a),
875 					le64_to_cpu(u0->b),
876 					(u64)rx_buffer_info->dma,
877 					rx_buffer_info->skb,
878 					ring_desc);
879 
880 				if (netif_msg_pktdata(adapter) &&
881 				    rx_buffer_info->dma) {
882 					print_hex_dump(KERN_INFO, "",
883 					   DUMP_PREFIX_ADDRESS, 16, 1,
884 					   page_address(rx_buffer_info->page) +
885 						    rx_buffer_info->page_offset,
886 					   ixgbe_rx_bufsz(rx_ring), true);
887 				}
888 			}
889 		}
890 	}
891 }
892 
893 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
894 {
895 	u32 ctrl_ext;
896 
897 	/* Let firmware take over control of h/w */
898 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
899 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
900 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
901 }
902 
903 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
904 {
905 	u32 ctrl_ext;
906 
907 	/* Let firmware know the driver has taken over */
908 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
909 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
910 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
911 }
912 
913 /**
914  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
915  * @adapter: pointer to adapter struct
916  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
917  * @queue: queue to map the corresponding interrupt to
918  * @msix_vector: the vector to map to the corresponding queue
919  *
920  */
921 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
922 			   u8 queue, u8 msix_vector)
923 {
924 	u32 ivar, index;
925 	struct ixgbe_hw *hw = &adapter->hw;
926 	switch (hw->mac.type) {
927 	case ixgbe_mac_82598EB:
928 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
929 		if (direction == -1)
930 			direction = 0;
931 		index = (((direction * 64) + queue) >> 2) & 0x1F;
932 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
933 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
934 		ivar |= (msix_vector << (8 * (queue & 0x3)));
935 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
936 		break;
937 	case ixgbe_mac_82599EB:
938 	case ixgbe_mac_X540:
939 	case ixgbe_mac_X550:
940 	case ixgbe_mac_X550EM_x:
941 	case ixgbe_mac_x550em_a:
942 		if (direction == -1) {
943 			/* other causes */
944 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
945 			index = ((queue & 1) * 8);
946 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
947 			ivar &= ~(0xFF << index);
948 			ivar |= (msix_vector << index);
949 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
950 			break;
951 		} else {
952 			/* tx or rx causes */
953 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
954 			index = ((16 * (queue & 1)) + (8 * direction));
955 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
956 			ivar &= ~(0xFF << index);
957 			ivar |= (msix_vector << index);
958 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
959 			break;
960 		}
961 	default:
962 		break;
963 	}
964 }
965 
966 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
967 					  u64 qmask)
968 {
969 	u32 mask;
970 
971 	switch (adapter->hw.mac.type) {
972 	case ixgbe_mac_82598EB:
973 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
974 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
975 		break;
976 	case ixgbe_mac_82599EB:
977 	case ixgbe_mac_X540:
978 	case ixgbe_mac_X550:
979 	case ixgbe_mac_X550EM_x:
980 	case ixgbe_mac_x550em_a:
981 		mask = (qmask & 0xFFFFFFFF);
982 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
983 		mask = (qmask >> 32);
984 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
985 		break;
986 	default:
987 		break;
988 	}
989 }
990 
991 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
992 {
993 	struct ixgbe_hw *hw = &adapter->hw;
994 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
995 	int i;
996 	u32 data;
997 
998 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
999 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
1000 		return;
1001 
1002 	switch (hw->mac.type) {
1003 	case ixgbe_mac_82598EB:
1004 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
1005 		break;
1006 	default:
1007 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
1008 	}
1009 	hwstats->lxoffrxc += data;
1010 
1011 	/* refill credits (no tx hang) if we received xoff */
1012 	if (!data)
1013 		return;
1014 
1015 	for (i = 0; i < adapter->num_tx_queues; i++)
1016 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
1017 			  &adapter->tx_ring[i]->state);
1018 
1019 	for (i = 0; i < adapter->num_xdp_queues; i++)
1020 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
1021 			  &adapter->xdp_ring[i]->state);
1022 }
1023 
1024 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
1025 {
1026 	struct ixgbe_hw *hw = &adapter->hw;
1027 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
1028 	u32 xoff[8] = {0};
1029 	u8 tc;
1030 	int i;
1031 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
1032 
1033 	if (adapter->ixgbe_ieee_pfc)
1034 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
1035 
1036 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
1037 		ixgbe_update_xoff_rx_lfc(adapter);
1038 		return;
1039 	}
1040 
1041 	/* update stats for each tc, only valid with PFC enabled */
1042 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
1043 		u32 pxoffrxc;
1044 
1045 		switch (hw->mac.type) {
1046 		case ixgbe_mac_82598EB:
1047 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1048 			break;
1049 		default:
1050 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1051 		}
1052 		hwstats->pxoffrxc[i] += pxoffrxc;
1053 		/* Get the TC for given UP */
1054 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1055 		xoff[tc] += pxoffrxc;
1056 	}
1057 
1058 	/* disarm tx queues that have received xoff frames */
1059 	for (i = 0; i < adapter->num_tx_queues; i++) {
1060 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1061 
1062 		tc = tx_ring->dcb_tc;
1063 		if (xoff[tc])
1064 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1065 	}
1066 
1067 	for (i = 0; i < adapter->num_xdp_queues; i++) {
1068 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1069 
1070 		tc = xdp_ring->dcb_tc;
1071 		if (xoff[tc])
1072 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1073 	}
1074 }
1075 
1076 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1077 {
1078 	return ring->stats.packets;
1079 }
1080 
1081 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1082 {
1083 	unsigned int head, tail;
1084 
1085 	head = ring->next_to_clean;
1086 	tail = ring->next_to_use;
1087 
1088 	return ((head <= tail) ? tail : tail + ring->count) - head;
1089 }
1090 
1091 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1092 {
1093 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1094 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1095 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1096 
1097 	clear_check_for_tx_hang(tx_ring);
1098 
1099 	/*
1100 	 * Check for a hung queue, but be thorough. This verifies
1101 	 * that a transmit has been completed since the previous
1102 	 * check AND there is at least one packet pending. The
1103 	 * ARMED bit is set to indicate a potential hang. The
1104 	 * bit is cleared if a pause frame is received to remove
1105 	 * false hang detection due to PFC or 802.3x frames. By
1106 	 * requiring this to fail twice we avoid races with
1107 	 * pfc clearing the ARMED bit and conditions where we
1108 	 * run the check_tx_hang logic with a transmit completion
1109 	 * pending but without time to complete it yet.
1110 	 */
1111 	if (tx_done_old == tx_done && tx_pending)
1112 		/* make sure it is true for two checks in a row */
1113 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1114 					&tx_ring->state);
1115 	/* update completed stats and continue */
1116 	tx_ring->tx_stats.tx_done_old = tx_done;
1117 	/* reset the countdown */
1118 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1119 
1120 	return false;
1121 }
1122 
1123 /**
1124  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1125  * @adapter: driver private struct
1126  **/
1127 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1128 {
1129 
1130 	/* Do the reset outside of interrupt context */
1131 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1132 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1133 		e_warn(drv, "initiating reset due to tx timeout\n");
1134 		ixgbe_service_event_schedule(adapter);
1135 	}
1136 }
1137 
1138 /**
1139  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1140  * @netdev: network interface device structure
1141  * @queue_index: Tx queue to set
1142  * @maxrate: desired maximum transmit bitrate
1143  **/
1144 static int ixgbe_tx_maxrate(struct net_device *netdev,
1145 			    int queue_index, u32 maxrate)
1146 {
1147 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1148 	struct ixgbe_hw *hw = &adapter->hw;
1149 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1150 
1151 	if (!maxrate)
1152 		return 0;
1153 
1154 	/* Calculate the rate factor values to set */
1155 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1156 	bcnrc_val /= maxrate;
1157 
1158 	/* clear everything but the rate factor */
1159 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1160 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1161 
1162 	/* enable the rate scheduler */
1163 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1164 
1165 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1166 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1167 
1168 	return 0;
1169 }
1170 
1171 /**
1172  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1173  * @q_vector: structure containing interrupt and ring information
1174  * @tx_ring: tx ring to clean
1175  * @napi_budget: Used to determine if we are in netpoll
1176  **/
1177 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1178 			       struct ixgbe_ring *tx_ring, int napi_budget)
1179 {
1180 	struct ixgbe_adapter *adapter = q_vector->adapter;
1181 	struct ixgbe_tx_buffer *tx_buffer;
1182 	union ixgbe_adv_tx_desc *tx_desc;
1183 	unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1184 	unsigned int budget = q_vector->tx.work_limit;
1185 	unsigned int i = tx_ring->next_to_clean;
1186 
1187 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1188 		return true;
1189 
1190 	tx_buffer = &tx_ring->tx_buffer_info[i];
1191 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1192 	i -= tx_ring->count;
1193 
1194 	do {
1195 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1196 
1197 		/* if next_to_watch is not set then there is no work pending */
1198 		if (!eop_desc)
1199 			break;
1200 
1201 		/* prevent any other reads prior to eop_desc */
1202 		smp_rmb();
1203 
1204 		/* if DD is not set pending work has not been completed */
1205 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1206 			break;
1207 
1208 		/* clear next_to_watch to prevent false hangs */
1209 		tx_buffer->next_to_watch = NULL;
1210 
1211 		/* update the statistics for this packet */
1212 		total_bytes += tx_buffer->bytecount;
1213 		total_packets += tx_buffer->gso_segs;
1214 		if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1215 			total_ipsec++;
1216 
1217 		/* free the skb */
1218 		if (ring_is_xdp(tx_ring))
1219 			page_frag_free(tx_buffer->data);
1220 		else
1221 			napi_consume_skb(tx_buffer->skb, napi_budget);
1222 
1223 		/* unmap skb header data */
1224 		dma_unmap_single(tx_ring->dev,
1225 				 dma_unmap_addr(tx_buffer, dma),
1226 				 dma_unmap_len(tx_buffer, len),
1227 				 DMA_TO_DEVICE);
1228 
1229 		/* clear tx_buffer data */
1230 		dma_unmap_len_set(tx_buffer, len, 0);
1231 
1232 		/* unmap remaining buffers */
1233 		while (tx_desc != eop_desc) {
1234 			tx_buffer++;
1235 			tx_desc++;
1236 			i++;
1237 			if (unlikely(!i)) {
1238 				i -= tx_ring->count;
1239 				tx_buffer = tx_ring->tx_buffer_info;
1240 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1241 			}
1242 
1243 			/* unmap any remaining paged data */
1244 			if (dma_unmap_len(tx_buffer, len)) {
1245 				dma_unmap_page(tx_ring->dev,
1246 					       dma_unmap_addr(tx_buffer, dma),
1247 					       dma_unmap_len(tx_buffer, len),
1248 					       DMA_TO_DEVICE);
1249 				dma_unmap_len_set(tx_buffer, len, 0);
1250 			}
1251 		}
1252 
1253 		/* move us one more past the eop_desc for start of next pkt */
1254 		tx_buffer++;
1255 		tx_desc++;
1256 		i++;
1257 		if (unlikely(!i)) {
1258 			i -= tx_ring->count;
1259 			tx_buffer = tx_ring->tx_buffer_info;
1260 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1261 		}
1262 
1263 		/* issue prefetch for next Tx descriptor */
1264 		prefetch(tx_desc);
1265 
1266 		/* update budget accounting */
1267 		budget--;
1268 	} while (likely(budget));
1269 
1270 	i += tx_ring->count;
1271 	tx_ring->next_to_clean = i;
1272 	u64_stats_update_begin(&tx_ring->syncp);
1273 	tx_ring->stats.bytes += total_bytes;
1274 	tx_ring->stats.packets += total_packets;
1275 	u64_stats_update_end(&tx_ring->syncp);
1276 	q_vector->tx.total_bytes += total_bytes;
1277 	q_vector->tx.total_packets += total_packets;
1278 	adapter->tx_ipsec += total_ipsec;
1279 
1280 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1281 		/* schedule immediate reset if we believe we hung */
1282 		struct ixgbe_hw *hw = &adapter->hw;
1283 		e_err(drv, "Detected Tx Unit Hang %s\n"
1284 			"  Tx Queue             <%d>\n"
1285 			"  TDH, TDT             <%x>, <%x>\n"
1286 			"  next_to_use          <%x>\n"
1287 			"  next_to_clean        <%x>\n"
1288 			"tx_buffer_info[next_to_clean]\n"
1289 			"  time_stamp           <%lx>\n"
1290 			"  jiffies              <%lx>\n",
1291 			ring_is_xdp(tx_ring) ? "(XDP)" : "",
1292 			tx_ring->queue_index,
1293 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1294 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1295 			tx_ring->next_to_use, i,
1296 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1297 
1298 		if (!ring_is_xdp(tx_ring))
1299 			netif_stop_subqueue(tx_ring->netdev,
1300 					    tx_ring->queue_index);
1301 
1302 		e_info(probe,
1303 		       "tx hang %d detected on queue %d, resetting adapter\n",
1304 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1305 
1306 		/* schedule immediate reset if we believe we hung */
1307 		ixgbe_tx_timeout_reset(adapter);
1308 
1309 		/* the adapter is about to reset, no point in enabling stuff */
1310 		return true;
1311 	}
1312 
1313 	if (ring_is_xdp(tx_ring))
1314 		return !!budget;
1315 
1316 	netdev_tx_completed_queue(txring_txq(tx_ring),
1317 				  total_packets, total_bytes);
1318 
1319 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1320 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1321 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1322 		/* Make sure that anybody stopping the queue after this
1323 		 * sees the new next_to_clean.
1324 		 */
1325 		smp_mb();
1326 		if (__netif_subqueue_stopped(tx_ring->netdev,
1327 					     tx_ring->queue_index)
1328 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1329 			netif_wake_subqueue(tx_ring->netdev,
1330 					    tx_ring->queue_index);
1331 			++tx_ring->tx_stats.restart_queue;
1332 		}
1333 	}
1334 
1335 	return !!budget;
1336 }
1337 
1338 #ifdef CONFIG_IXGBE_DCA
1339 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1340 				struct ixgbe_ring *tx_ring,
1341 				int cpu)
1342 {
1343 	struct ixgbe_hw *hw = &adapter->hw;
1344 	u32 txctrl = 0;
1345 	u16 reg_offset;
1346 
1347 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1348 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1349 
1350 	switch (hw->mac.type) {
1351 	case ixgbe_mac_82598EB:
1352 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1353 		break;
1354 	case ixgbe_mac_82599EB:
1355 	case ixgbe_mac_X540:
1356 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1357 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1358 		break;
1359 	default:
1360 		/* for unknown hardware do not write register */
1361 		return;
1362 	}
1363 
1364 	/*
1365 	 * We can enable relaxed ordering for reads, but not writes when
1366 	 * DCA is enabled.  This is due to a known issue in some chipsets
1367 	 * which will cause the DCA tag to be cleared.
1368 	 */
1369 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1370 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1371 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1372 
1373 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1374 }
1375 
1376 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1377 				struct ixgbe_ring *rx_ring,
1378 				int cpu)
1379 {
1380 	struct ixgbe_hw *hw = &adapter->hw;
1381 	u32 rxctrl = 0;
1382 	u8 reg_idx = rx_ring->reg_idx;
1383 
1384 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1385 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1386 
1387 	switch (hw->mac.type) {
1388 	case ixgbe_mac_82599EB:
1389 	case ixgbe_mac_X540:
1390 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1391 		break;
1392 	default:
1393 		break;
1394 	}
1395 
1396 	/*
1397 	 * We can enable relaxed ordering for reads, but not writes when
1398 	 * DCA is enabled.  This is due to a known issue in some chipsets
1399 	 * which will cause the DCA tag to be cleared.
1400 	 */
1401 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1402 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1403 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1404 
1405 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1406 }
1407 
1408 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1409 {
1410 	struct ixgbe_adapter *adapter = q_vector->adapter;
1411 	struct ixgbe_ring *ring;
1412 	int cpu = get_cpu();
1413 
1414 	if (q_vector->cpu == cpu)
1415 		goto out_no_update;
1416 
1417 	ixgbe_for_each_ring(ring, q_vector->tx)
1418 		ixgbe_update_tx_dca(adapter, ring, cpu);
1419 
1420 	ixgbe_for_each_ring(ring, q_vector->rx)
1421 		ixgbe_update_rx_dca(adapter, ring, cpu);
1422 
1423 	q_vector->cpu = cpu;
1424 out_no_update:
1425 	put_cpu();
1426 }
1427 
1428 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1429 {
1430 	int i;
1431 
1432 	/* always use CB2 mode, difference is masked in the CB driver */
1433 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1434 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1435 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1436 	else
1437 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1438 				IXGBE_DCA_CTRL_DCA_DISABLE);
1439 
1440 	for (i = 0; i < adapter->num_q_vectors; i++) {
1441 		adapter->q_vector[i]->cpu = -1;
1442 		ixgbe_update_dca(adapter->q_vector[i]);
1443 	}
1444 }
1445 
1446 static int __ixgbe_notify_dca(struct device *dev, void *data)
1447 {
1448 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1449 	unsigned long event = *(unsigned long *)data;
1450 
1451 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1452 		return 0;
1453 
1454 	switch (event) {
1455 	case DCA_PROVIDER_ADD:
1456 		/* if we're already enabled, don't do it again */
1457 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1458 			break;
1459 		if (dca_add_requester(dev) == 0) {
1460 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1461 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1462 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1463 			break;
1464 		}
1465 		/* fall through - DCA is disabled. */
1466 	case DCA_PROVIDER_REMOVE:
1467 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1468 			dca_remove_requester(dev);
1469 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1470 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1471 					IXGBE_DCA_CTRL_DCA_DISABLE);
1472 		}
1473 		break;
1474 	}
1475 
1476 	return 0;
1477 }
1478 
1479 #endif /* CONFIG_IXGBE_DCA */
1480 
1481 #define IXGBE_RSS_L4_TYPES_MASK \
1482 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1483 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1484 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1485 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1486 
1487 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1488 				 union ixgbe_adv_rx_desc *rx_desc,
1489 				 struct sk_buff *skb)
1490 {
1491 	u16 rss_type;
1492 
1493 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1494 		return;
1495 
1496 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1497 		   IXGBE_RXDADV_RSSTYPE_MASK;
1498 
1499 	if (!rss_type)
1500 		return;
1501 
1502 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1503 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1504 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1505 }
1506 
1507 #ifdef IXGBE_FCOE
1508 /**
1509  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1510  * @ring: structure containing ring specific data
1511  * @rx_desc: advanced rx descriptor
1512  *
1513  * Returns : true if it is FCoE pkt
1514  */
1515 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1516 				    union ixgbe_adv_rx_desc *rx_desc)
1517 {
1518 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1519 
1520 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1521 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1522 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1523 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1524 }
1525 
1526 #endif /* IXGBE_FCOE */
1527 /**
1528  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1529  * @ring: structure containing ring specific data
1530  * @rx_desc: current Rx descriptor being processed
1531  * @skb: skb currently being received and modified
1532  **/
1533 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1534 				     union ixgbe_adv_rx_desc *rx_desc,
1535 				     struct sk_buff *skb)
1536 {
1537 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1538 	bool encap_pkt = false;
1539 
1540 	skb_checksum_none_assert(skb);
1541 
1542 	/* Rx csum disabled */
1543 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1544 		return;
1545 
1546 	/* check for VXLAN and Geneve packets */
1547 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1548 		encap_pkt = true;
1549 		skb->encapsulation = 1;
1550 	}
1551 
1552 	/* if IP and error */
1553 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1554 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1555 		ring->rx_stats.csum_err++;
1556 		return;
1557 	}
1558 
1559 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1560 		return;
1561 
1562 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1563 		/*
1564 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1565 		 * checksum errors.
1566 		 */
1567 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1568 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1569 			return;
1570 
1571 		ring->rx_stats.csum_err++;
1572 		return;
1573 	}
1574 
1575 	/* It must be a TCP or UDP packet with a valid checksum */
1576 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1577 	if (encap_pkt) {
1578 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1579 			return;
1580 
1581 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1582 			skb->ip_summed = CHECKSUM_NONE;
1583 			return;
1584 		}
1585 		/* If we checked the outer header let the stack know */
1586 		skb->csum_level = 1;
1587 	}
1588 }
1589 
1590 static inline unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1591 {
1592 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1593 }
1594 
1595 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1596 				    struct ixgbe_rx_buffer *bi)
1597 {
1598 	struct page *page = bi->page;
1599 	dma_addr_t dma;
1600 
1601 	/* since we are recycling buffers we should seldom need to alloc */
1602 	if (likely(page))
1603 		return true;
1604 
1605 	/* alloc new page for storage */
1606 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1607 	if (unlikely(!page)) {
1608 		rx_ring->rx_stats.alloc_rx_page_failed++;
1609 		return false;
1610 	}
1611 
1612 	/* map page for use */
1613 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1614 				 ixgbe_rx_pg_size(rx_ring),
1615 				 DMA_FROM_DEVICE,
1616 				 IXGBE_RX_DMA_ATTR);
1617 
1618 	/*
1619 	 * if mapping failed free memory back to system since
1620 	 * there isn't much point in holding memory we can't use
1621 	 */
1622 	if (dma_mapping_error(rx_ring->dev, dma)) {
1623 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1624 
1625 		rx_ring->rx_stats.alloc_rx_page_failed++;
1626 		return false;
1627 	}
1628 
1629 	bi->dma = dma;
1630 	bi->page = page;
1631 	bi->page_offset = ixgbe_rx_offset(rx_ring);
1632 	page_ref_add(page, USHRT_MAX - 1);
1633 	bi->pagecnt_bias = USHRT_MAX;
1634 	rx_ring->rx_stats.alloc_rx_page++;
1635 
1636 	return true;
1637 }
1638 
1639 /**
1640  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1641  * @rx_ring: ring to place buffers on
1642  * @cleaned_count: number of buffers to replace
1643  **/
1644 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1645 {
1646 	union ixgbe_adv_rx_desc *rx_desc;
1647 	struct ixgbe_rx_buffer *bi;
1648 	u16 i = rx_ring->next_to_use;
1649 	u16 bufsz;
1650 
1651 	/* nothing to do */
1652 	if (!cleaned_count)
1653 		return;
1654 
1655 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1656 	bi = &rx_ring->rx_buffer_info[i];
1657 	i -= rx_ring->count;
1658 
1659 	bufsz = ixgbe_rx_bufsz(rx_ring);
1660 
1661 	do {
1662 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1663 			break;
1664 
1665 		/* sync the buffer for use by the device */
1666 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1667 						 bi->page_offset, bufsz,
1668 						 DMA_FROM_DEVICE);
1669 
1670 		/*
1671 		 * Refresh the desc even if buffer_addrs didn't change
1672 		 * because each write-back erases this info.
1673 		 */
1674 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1675 
1676 		rx_desc++;
1677 		bi++;
1678 		i++;
1679 		if (unlikely(!i)) {
1680 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1681 			bi = rx_ring->rx_buffer_info;
1682 			i -= rx_ring->count;
1683 		}
1684 
1685 		/* clear the length for the next_to_use descriptor */
1686 		rx_desc->wb.upper.length = 0;
1687 
1688 		cleaned_count--;
1689 	} while (cleaned_count);
1690 
1691 	i += rx_ring->count;
1692 
1693 	if (rx_ring->next_to_use != i) {
1694 		rx_ring->next_to_use = i;
1695 
1696 		/* update next to alloc since we have filled the ring */
1697 		rx_ring->next_to_alloc = i;
1698 
1699 		/* Force memory writes to complete before letting h/w
1700 		 * know there are new descriptors to fetch.  (Only
1701 		 * applicable for weak-ordered memory model archs,
1702 		 * such as IA-64).
1703 		 */
1704 		wmb();
1705 		writel(i, rx_ring->tail);
1706 	}
1707 }
1708 
1709 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1710 				   struct sk_buff *skb)
1711 {
1712 	u16 hdr_len = skb_headlen(skb);
1713 
1714 	/* set gso_size to avoid messing up TCP MSS */
1715 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1716 						 IXGBE_CB(skb)->append_cnt);
1717 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1718 }
1719 
1720 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1721 				   struct sk_buff *skb)
1722 {
1723 	/* if append_cnt is 0 then frame is not RSC */
1724 	if (!IXGBE_CB(skb)->append_cnt)
1725 		return;
1726 
1727 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1728 	rx_ring->rx_stats.rsc_flush++;
1729 
1730 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1731 
1732 	/* gso_size is computed using append_cnt so always clear it last */
1733 	IXGBE_CB(skb)->append_cnt = 0;
1734 }
1735 
1736 /**
1737  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1738  * @rx_ring: rx descriptor ring packet is being transacted on
1739  * @rx_desc: pointer to the EOP Rx descriptor
1740  * @skb: pointer to current skb being populated
1741  *
1742  * This function checks the ring, descriptor, and packet information in
1743  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1744  * other fields within the skb.
1745  **/
1746 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1747 				     union ixgbe_adv_rx_desc *rx_desc,
1748 				     struct sk_buff *skb)
1749 {
1750 	struct net_device *dev = rx_ring->netdev;
1751 	u32 flags = rx_ring->q_vector->adapter->flags;
1752 
1753 	ixgbe_update_rsc_stats(rx_ring, skb);
1754 
1755 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1756 
1757 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1758 
1759 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1760 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1761 
1762 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1763 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1764 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1765 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1766 	}
1767 
1768 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1769 		ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1770 
1771 	skb->protocol = eth_type_trans(skb, dev);
1772 
1773 	/* record Rx queue, or update MACVLAN statistics */
1774 	if (netif_is_ixgbe(dev))
1775 		skb_record_rx_queue(skb, rx_ring->queue_index);
1776 	else
1777 		macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1778 				 (skb->pkt_type == PACKET_BROADCAST) ||
1779 				 (skb->pkt_type == PACKET_MULTICAST));
1780 }
1781 
1782 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1783 			 struct sk_buff *skb)
1784 {
1785 	napi_gro_receive(&q_vector->napi, skb);
1786 }
1787 
1788 /**
1789  * ixgbe_is_non_eop - process handling of non-EOP buffers
1790  * @rx_ring: Rx ring being processed
1791  * @rx_desc: Rx descriptor for current buffer
1792  * @skb: Current socket buffer containing buffer in progress
1793  *
1794  * This function updates next to clean.  If the buffer is an EOP buffer
1795  * this function exits returning false, otherwise it will place the
1796  * sk_buff in the next buffer to be chained and return true indicating
1797  * that this is in fact a non-EOP buffer.
1798  **/
1799 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1800 			     union ixgbe_adv_rx_desc *rx_desc,
1801 			     struct sk_buff *skb)
1802 {
1803 	u32 ntc = rx_ring->next_to_clean + 1;
1804 
1805 	/* fetch, update, and store next to clean */
1806 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1807 	rx_ring->next_to_clean = ntc;
1808 
1809 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1810 
1811 	/* update RSC append count if present */
1812 	if (ring_is_rsc_enabled(rx_ring)) {
1813 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1814 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1815 
1816 		if (unlikely(rsc_enabled)) {
1817 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1818 
1819 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1820 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1821 
1822 			/* update ntc based on RSC value */
1823 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1824 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1825 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1826 		}
1827 	}
1828 
1829 	/* if we are the last buffer then there is nothing else to do */
1830 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1831 		return false;
1832 
1833 	/* place skb in next buffer to be received */
1834 	rx_ring->rx_buffer_info[ntc].skb = skb;
1835 	rx_ring->rx_stats.non_eop_descs++;
1836 
1837 	return true;
1838 }
1839 
1840 /**
1841  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1842  * @rx_ring: rx descriptor ring packet is being transacted on
1843  * @skb: pointer to current skb being adjusted
1844  *
1845  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1846  * main difference between this version and the original function is that
1847  * this function can make several assumptions about the state of things
1848  * that allow for significant optimizations versus the standard function.
1849  * As a result we can do things like drop a frag and maintain an accurate
1850  * truesize for the skb.
1851  */
1852 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1853 			    struct sk_buff *skb)
1854 {
1855 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1856 	unsigned char *va;
1857 	unsigned int pull_len;
1858 
1859 	/*
1860 	 * it is valid to use page_address instead of kmap since we are
1861 	 * working with pages allocated out of the lomem pool per
1862 	 * alloc_page(GFP_ATOMIC)
1863 	 */
1864 	va = skb_frag_address(frag);
1865 
1866 	/*
1867 	 * we need the header to contain the greater of either ETH_HLEN or
1868 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1869 	 */
1870 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1871 
1872 	/* align pull length to size of long to optimize memcpy performance */
1873 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1874 
1875 	/* update all of the pointers */
1876 	skb_frag_size_sub(frag, pull_len);
1877 	frag->page_offset += pull_len;
1878 	skb->data_len -= pull_len;
1879 	skb->tail += pull_len;
1880 }
1881 
1882 /**
1883  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1884  * @rx_ring: rx descriptor ring packet is being transacted on
1885  * @skb: pointer to current skb being updated
1886  *
1887  * This function provides a basic DMA sync up for the first fragment of an
1888  * skb.  The reason for doing this is that the first fragment cannot be
1889  * unmapped until we have reached the end of packet descriptor for a buffer
1890  * chain.
1891  */
1892 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1893 				struct sk_buff *skb)
1894 {
1895 	/* if the page was released unmap it, else just sync our portion */
1896 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1897 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1898 				     ixgbe_rx_pg_size(rx_ring),
1899 				     DMA_FROM_DEVICE,
1900 				     IXGBE_RX_DMA_ATTR);
1901 	} else if (ring_uses_build_skb(rx_ring)) {
1902 		unsigned long offset = (unsigned long)(skb->data) & ~PAGE_MASK;
1903 
1904 		dma_sync_single_range_for_cpu(rx_ring->dev,
1905 					      IXGBE_CB(skb)->dma,
1906 					      offset,
1907 					      skb_headlen(skb),
1908 					      DMA_FROM_DEVICE);
1909 	} else {
1910 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1911 
1912 		dma_sync_single_range_for_cpu(rx_ring->dev,
1913 					      IXGBE_CB(skb)->dma,
1914 					      frag->page_offset,
1915 					      skb_frag_size(frag),
1916 					      DMA_FROM_DEVICE);
1917 	}
1918 }
1919 
1920 /**
1921  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1922  * @rx_ring: rx descriptor ring packet is being transacted on
1923  * @rx_desc: pointer to the EOP Rx descriptor
1924  * @skb: pointer to current skb being fixed
1925  *
1926  * Check if the skb is valid in the XDP case it will be an error pointer.
1927  * Return true in this case to abort processing and advance to next
1928  * descriptor.
1929  *
1930  * Check for corrupted packet headers caused by senders on the local L2
1931  * embedded NIC switch not setting up their Tx Descriptors right.  These
1932  * should be very rare.
1933  *
1934  * Also address the case where we are pulling data in on pages only
1935  * and as such no data is present in the skb header.
1936  *
1937  * In addition if skb is not at least 60 bytes we need to pad it so that
1938  * it is large enough to qualify as a valid Ethernet frame.
1939  *
1940  * Returns true if an error was encountered and skb was freed.
1941  **/
1942 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1943 				  union ixgbe_adv_rx_desc *rx_desc,
1944 				  struct sk_buff *skb)
1945 {
1946 	struct net_device *netdev = rx_ring->netdev;
1947 
1948 	/* XDP packets use error pointer so abort at this point */
1949 	if (IS_ERR(skb))
1950 		return true;
1951 
1952 	/* Verify netdev is present, and that packet does not have any
1953 	 * errors that would be unacceptable to the netdev.
1954 	 */
1955 	if (!netdev ||
1956 	    (unlikely(ixgbe_test_staterr(rx_desc,
1957 					 IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1958 	     !(netdev->features & NETIF_F_RXALL)))) {
1959 		dev_kfree_skb_any(skb);
1960 		return true;
1961 	}
1962 
1963 	/* place header in linear portion of buffer */
1964 	if (!skb_headlen(skb))
1965 		ixgbe_pull_tail(rx_ring, skb);
1966 
1967 #ifdef IXGBE_FCOE
1968 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1969 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1970 		return false;
1971 
1972 #endif
1973 	/* if eth_skb_pad returns an error the skb was freed */
1974 	if (eth_skb_pad(skb))
1975 		return true;
1976 
1977 	return false;
1978 }
1979 
1980 /**
1981  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1982  * @rx_ring: rx descriptor ring to store buffers on
1983  * @old_buff: donor buffer to have page reused
1984  *
1985  * Synchronizes page for reuse by the adapter
1986  **/
1987 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1988 				struct ixgbe_rx_buffer *old_buff)
1989 {
1990 	struct ixgbe_rx_buffer *new_buff;
1991 	u16 nta = rx_ring->next_to_alloc;
1992 
1993 	new_buff = &rx_ring->rx_buffer_info[nta];
1994 
1995 	/* update, and store next to alloc */
1996 	nta++;
1997 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1998 
1999 	/* Transfer page from old buffer to new buffer.
2000 	 * Move each member individually to avoid possible store
2001 	 * forwarding stalls and unnecessary copy of skb.
2002 	 */
2003 	new_buff->dma		= old_buff->dma;
2004 	new_buff->page		= old_buff->page;
2005 	new_buff->page_offset	= old_buff->page_offset;
2006 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
2007 }
2008 
2009 static inline bool ixgbe_page_is_reserved(struct page *page)
2010 {
2011 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
2012 }
2013 
2014 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer)
2015 {
2016 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
2017 	struct page *page = rx_buffer->page;
2018 
2019 	/* avoid re-using remote pages */
2020 	if (unlikely(ixgbe_page_is_reserved(page)))
2021 		return false;
2022 
2023 #if (PAGE_SIZE < 8192)
2024 	/* if we are only owner of page we can reuse it */
2025 	if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
2026 		return false;
2027 #else
2028 	/* The last offset is a bit aggressive in that we assume the
2029 	 * worst case of FCoE being enabled and using a 3K buffer.
2030 	 * However this should have minimal impact as the 1K extra is
2031 	 * still less than one buffer in size.
2032 	 */
2033 #define IXGBE_LAST_OFFSET \
2034 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
2035 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
2036 		return false;
2037 #endif
2038 
2039 	/* If we have drained the page fragment pool we need to update
2040 	 * the pagecnt_bias and page count so that we fully restock the
2041 	 * number of references the driver holds.
2042 	 */
2043 	if (unlikely(pagecnt_bias == 1)) {
2044 		page_ref_add(page, USHRT_MAX - 1);
2045 		rx_buffer->pagecnt_bias = USHRT_MAX;
2046 	}
2047 
2048 	return true;
2049 }
2050 
2051 /**
2052  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
2053  * @rx_ring: rx descriptor ring to transact packets on
2054  * @rx_buffer: buffer containing page to add
2055  * @skb: sk_buff to place the data into
2056  * @size: size of data in rx_buffer
2057  *
2058  * This function will add the data contained in rx_buffer->page to the skb.
2059  * This is done either through a direct copy if the data in the buffer is
2060  * less than the skb header size, otherwise it will just attach the page as
2061  * a frag to the skb.
2062  *
2063  * The function will then update the page offset if necessary and return
2064  * true if the buffer can be reused by the adapter.
2065  **/
2066 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
2067 			      struct ixgbe_rx_buffer *rx_buffer,
2068 			      struct sk_buff *skb,
2069 			      unsigned int size)
2070 {
2071 #if (PAGE_SIZE < 8192)
2072 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2073 #else
2074 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2075 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2076 				SKB_DATA_ALIGN(size);
2077 #endif
2078 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2079 			rx_buffer->page_offset, size, truesize);
2080 #if (PAGE_SIZE < 8192)
2081 	rx_buffer->page_offset ^= truesize;
2082 #else
2083 	rx_buffer->page_offset += truesize;
2084 #endif
2085 }
2086 
2087 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2088 						   union ixgbe_adv_rx_desc *rx_desc,
2089 						   struct sk_buff **skb,
2090 						   const unsigned int size)
2091 {
2092 	struct ixgbe_rx_buffer *rx_buffer;
2093 
2094 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2095 	prefetchw(rx_buffer->page);
2096 	*skb = rx_buffer->skb;
2097 
2098 	/* Delay unmapping of the first packet. It carries the header
2099 	 * information, HW may still access the header after the writeback.
2100 	 * Only unmap it when EOP is reached
2101 	 */
2102 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2103 		if (!*skb)
2104 			goto skip_sync;
2105 	} else {
2106 		if (*skb)
2107 			ixgbe_dma_sync_frag(rx_ring, *skb);
2108 	}
2109 
2110 	/* we are reusing so sync this buffer for CPU use */
2111 	dma_sync_single_range_for_cpu(rx_ring->dev,
2112 				      rx_buffer->dma,
2113 				      rx_buffer->page_offset,
2114 				      size,
2115 				      DMA_FROM_DEVICE);
2116 skip_sync:
2117 	rx_buffer->pagecnt_bias--;
2118 
2119 	return rx_buffer;
2120 }
2121 
2122 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2123 				struct ixgbe_rx_buffer *rx_buffer,
2124 				struct sk_buff *skb)
2125 {
2126 	if (ixgbe_can_reuse_rx_page(rx_buffer)) {
2127 		/* hand second half of page back to the ring */
2128 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2129 	} else {
2130 		if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2131 			/* the page has been released from the ring */
2132 			IXGBE_CB(skb)->page_released = true;
2133 		} else {
2134 			/* we are not reusing the buffer so unmap it */
2135 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2136 					     ixgbe_rx_pg_size(rx_ring),
2137 					     DMA_FROM_DEVICE,
2138 					     IXGBE_RX_DMA_ATTR);
2139 		}
2140 		__page_frag_cache_drain(rx_buffer->page,
2141 					rx_buffer->pagecnt_bias);
2142 	}
2143 
2144 	/* clear contents of rx_buffer */
2145 	rx_buffer->page = NULL;
2146 	rx_buffer->skb = NULL;
2147 }
2148 
2149 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2150 					   struct ixgbe_rx_buffer *rx_buffer,
2151 					   struct xdp_buff *xdp,
2152 					   union ixgbe_adv_rx_desc *rx_desc)
2153 {
2154 	unsigned int size = xdp->data_end - xdp->data;
2155 #if (PAGE_SIZE < 8192)
2156 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2157 #else
2158 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2159 					       xdp->data_hard_start);
2160 #endif
2161 	struct sk_buff *skb;
2162 
2163 	/* prefetch first cache line of first page */
2164 	prefetch(xdp->data);
2165 #if L1_CACHE_BYTES < 128
2166 	prefetch(xdp->data + L1_CACHE_BYTES);
2167 #endif
2168 	/* Note, we get here by enabling legacy-rx via:
2169 	 *
2170 	 *    ethtool --set-priv-flags <dev> legacy-rx on
2171 	 *
2172 	 * In this mode, we currently get 0 extra XDP headroom as
2173 	 * opposed to having legacy-rx off, where we process XDP
2174 	 * packets going to stack via ixgbe_build_skb(). The latter
2175 	 * provides us currently with 192 bytes of headroom.
2176 	 *
2177 	 * For ixgbe_construct_skb() mode it means that the
2178 	 * xdp->data_meta will always point to xdp->data, since
2179 	 * the helper cannot expand the head. Should this ever
2180 	 * change in future for legacy-rx mode on, then lets also
2181 	 * add xdp->data_meta handling here.
2182 	 */
2183 
2184 	/* allocate a skb to store the frags */
2185 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2186 	if (unlikely(!skb))
2187 		return NULL;
2188 
2189 	if (size > IXGBE_RX_HDR_SIZE) {
2190 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2191 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2192 
2193 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2194 				xdp->data - page_address(rx_buffer->page),
2195 				size, truesize);
2196 #if (PAGE_SIZE < 8192)
2197 		rx_buffer->page_offset ^= truesize;
2198 #else
2199 		rx_buffer->page_offset += truesize;
2200 #endif
2201 	} else {
2202 		memcpy(__skb_put(skb, size),
2203 		       xdp->data, ALIGN(size, sizeof(long)));
2204 		rx_buffer->pagecnt_bias++;
2205 	}
2206 
2207 	return skb;
2208 }
2209 
2210 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2211 				       struct ixgbe_rx_buffer *rx_buffer,
2212 				       struct xdp_buff *xdp,
2213 				       union ixgbe_adv_rx_desc *rx_desc)
2214 {
2215 	unsigned int metasize = xdp->data - xdp->data_meta;
2216 #if (PAGE_SIZE < 8192)
2217 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2218 #else
2219 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2220 				SKB_DATA_ALIGN(xdp->data_end -
2221 					       xdp->data_hard_start);
2222 #endif
2223 	struct sk_buff *skb;
2224 
2225 	/* Prefetch first cache line of first page. If xdp->data_meta
2226 	 * is unused, this points extactly as xdp->data, otherwise we
2227 	 * likely have a consumer accessing first few bytes of meta
2228 	 * data, and then actual data.
2229 	 */
2230 	prefetch(xdp->data_meta);
2231 #if L1_CACHE_BYTES < 128
2232 	prefetch(xdp->data_meta + L1_CACHE_BYTES);
2233 #endif
2234 
2235 	/* build an skb to around the page buffer */
2236 	skb = build_skb(xdp->data_hard_start, truesize);
2237 	if (unlikely(!skb))
2238 		return NULL;
2239 
2240 	/* update pointers within the skb to store the data */
2241 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
2242 	__skb_put(skb, xdp->data_end - xdp->data);
2243 	if (metasize)
2244 		skb_metadata_set(skb, metasize);
2245 
2246 	/* record DMA address if this is the start of a chain of buffers */
2247 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2248 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2249 
2250 	/* update buffer offset */
2251 #if (PAGE_SIZE < 8192)
2252 	rx_buffer->page_offset ^= truesize;
2253 #else
2254 	rx_buffer->page_offset += truesize;
2255 #endif
2256 
2257 	return skb;
2258 }
2259 
2260 #define IXGBE_XDP_PASS 0
2261 #define IXGBE_XDP_CONSUMED 1
2262 #define IXGBE_XDP_TX 2
2263 
2264 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
2265 			       struct xdp_buff *xdp);
2266 
2267 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2268 				     struct ixgbe_ring *rx_ring,
2269 				     struct xdp_buff *xdp)
2270 {
2271 	int err, result = IXGBE_XDP_PASS;
2272 	struct bpf_prog *xdp_prog;
2273 	u32 act;
2274 
2275 	rcu_read_lock();
2276 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2277 
2278 	if (!xdp_prog)
2279 		goto xdp_out;
2280 
2281 	act = bpf_prog_run_xdp(xdp_prog, xdp);
2282 	switch (act) {
2283 	case XDP_PASS:
2284 		break;
2285 	case XDP_TX:
2286 		result = ixgbe_xmit_xdp_ring(adapter, xdp);
2287 		break;
2288 	case XDP_REDIRECT:
2289 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2290 		if (!err)
2291 			result = IXGBE_XDP_TX;
2292 		else
2293 			result = IXGBE_XDP_CONSUMED;
2294 		break;
2295 	default:
2296 		bpf_warn_invalid_xdp_action(act);
2297 		/* fallthrough */
2298 	case XDP_ABORTED:
2299 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2300 		/* fallthrough -- handle aborts by dropping packet */
2301 	case XDP_DROP:
2302 		result = IXGBE_XDP_CONSUMED;
2303 		break;
2304 	}
2305 xdp_out:
2306 	rcu_read_unlock();
2307 	return ERR_PTR(-result);
2308 }
2309 
2310 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2311 				 struct ixgbe_rx_buffer *rx_buffer,
2312 				 unsigned int size)
2313 {
2314 #if (PAGE_SIZE < 8192)
2315 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2316 
2317 	rx_buffer->page_offset ^= truesize;
2318 #else
2319 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2320 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2321 				SKB_DATA_ALIGN(size);
2322 
2323 	rx_buffer->page_offset += truesize;
2324 #endif
2325 }
2326 
2327 /**
2328  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2329  * @q_vector: structure containing interrupt and ring information
2330  * @rx_ring: rx descriptor ring to transact packets on
2331  * @budget: Total limit on number of packets to process
2332  *
2333  * This function provides a "bounce buffer" approach to Rx interrupt
2334  * processing.  The advantage to this is that on systems that have
2335  * expensive overhead for IOMMU access this provides a means of avoiding
2336  * it by maintaining the mapping of the page to the syste.
2337  *
2338  * Returns amount of work completed
2339  **/
2340 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2341 			       struct ixgbe_ring *rx_ring,
2342 			       const int budget)
2343 {
2344 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2345 	struct ixgbe_adapter *adapter = q_vector->adapter;
2346 #ifdef IXGBE_FCOE
2347 	int ddp_bytes;
2348 	unsigned int mss = 0;
2349 #endif /* IXGBE_FCOE */
2350 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2351 	bool xdp_xmit = false;
2352 	struct xdp_buff xdp;
2353 
2354 	xdp.rxq = &rx_ring->xdp_rxq;
2355 
2356 	while (likely(total_rx_packets < budget)) {
2357 		union ixgbe_adv_rx_desc *rx_desc;
2358 		struct ixgbe_rx_buffer *rx_buffer;
2359 		struct sk_buff *skb;
2360 		unsigned int size;
2361 
2362 		/* return some buffers to hardware, one at a time is too slow */
2363 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2364 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2365 			cleaned_count = 0;
2366 		}
2367 
2368 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2369 		size = le16_to_cpu(rx_desc->wb.upper.length);
2370 		if (!size)
2371 			break;
2372 
2373 		/* This memory barrier is needed to keep us from reading
2374 		 * any other fields out of the rx_desc until we know the
2375 		 * descriptor has been written back
2376 		 */
2377 		dma_rmb();
2378 
2379 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size);
2380 
2381 		/* retrieve a buffer from the ring */
2382 		if (!skb) {
2383 			xdp.data = page_address(rx_buffer->page) +
2384 				   rx_buffer->page_offset;
2385 			xdp.data_meta = xdp.data;
2386 			xdp.data_hard_start = xdp.data -
2387 					      ixgbe_rx_offset(rx_ring);
2388 			xdp.data_end = xdp.data + size;
2389 
2390 			skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2391 		}
2392 
2393 		if (IS_ERR(skb)) {
2394 			if (PTR_ERR(skb) == -IXGBE_XDP_TX) {
2395 				xdp_xmit = true;
2396 				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2397 			} else {
2398 				rx_buffer->pagecnt_bias++;
2399 			}
2400 			total_rx_packets++;
2401 			total_rx_bytes += size;
2402 		} else if (skb) {
2403 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2404 		} else if (ring_uses_build_skb(rx_ring)) {
2405 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2406 					      &xdp, rx_desc);
2407 		} else {
2408 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2409 						  &xdp, rx_desc);
2410 		}
2411 
2412 		/* exit if we failed to retrieve a buffer */
2413 		if (!skb) {
2414 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2415 			rx_buffer->pagecnt_bias++;
2416 			break;
2417 		}
2418 
2419 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb);
2420 		cleaned_count++;
2421 
2422 		/* place incomplete frames back on ring for completion */
2423 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2424 			continue;
2425 
2426 		/* verify the packet layout is correct */
2427 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2428 			continue;
2429 
2430 		/* probably a little skewed due to removing CRC */
2431 		total_rx_bytes += skb->len;
2432 
2433 		/* populate checksum, timestamp, VLAN, and protocol */
2434 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2435 
2436 #ifdef IXGBE_FCOE
2437 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2438 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2439 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2440 			/* include DDPed FCoE data */
2441 			if (ddp_bytes > 0) {
2442 				if (!mss) {
2443 					mss = rx_ring->netdev->mtu -
2444 						sizeof(struct fcoe_hdr) -
2445 						sizeof(struct fc_frame_header) -
2446 						sizeof(struct fcoe_crc_eof);
2447 					if (mss > 512)
2448 						mss &= ~511;
2449 				}
2450 				total_rx_bytes += ddp_bytes;
2451 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2452 								 mss);
2453 			}
2454 			if (!ddp_bytes) {
2455 				dev_kfree_skb_any(skb);
2456 				continue;
2457 			}
2458 		}
2459 
2460 #endif /* IXGBE_FCOE */
2461 		ixgbe_rx_skb(q_vector, skb);
2462 
2463 		/* update budget accounting */
2464 		total_rx_packets++;
2465 	}
2466 
2467 	if (xdp_xmit) {
2468 		struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
2469 
2470 		/* Force memory writes to complete before letting h/w
2471 		 * know there are new descriptors to fetch.
2472 		 */
2473 		wmb();
2474 		writel(ring->next_to_use, ring->tail);
2475 
2476 		xdp_do_flush_map();
2477 	}
2478 
2479 	u64_stats_update_begin(&rx_ring->syncp);
2480 	rx_ring->stats.packets += total_rx_packets;
2481 	rx_ring->stats.bytes += total_rx_bytes;
2482 	u64_stats_update_end(&rx_ring->syncp);
2483 	q_vector->rx.total_packets += total_rx_packets;
2484 	q_vector->rx.total_bytes += total_rx_bytes;
2485 
2486 	return total_rx_packets;
2487 }
2488 
2489 /**
2490  * ixgbe_configure_msix - Configure MSI-X hardware
2491  * @adapter: board private structure
2492  *
2493  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2494  * interrupts.
2495  **/
2496 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2497 {
2498 	struct ixgbe_q_vector *q_vector;
2499 	int v_idx;
2500 	u32 mask;
2501 
2502 	/* Populate MSIX to EITR Select */
2503 	if (adapter->num_vfs > 32) {
2504 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2505 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2506 	}
2507 
2508 	/*
2509 	 * Populate the IVAR table and set the ITR values to the
2510 	 * corresponding register.
2511 	 */
2512 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2513 		struct ixgbe_ring *ring;
2514 		q_vector = adapter->q_vector[v_idx];
2515 
2516 		ixgbe_for_each_ring(ring, q_vector->rx)
2517 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2518 
2519 		ixgbe_for_each_ring(ring, q_vector->tx)
2520 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2521 
2522 		ixgbe_write_eitr(q_vector);
2523 	}
2524 
2525 	switch (adapter->hw.mac.type) {
2526 	case ixgbe_mac_82598EB:
2527 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2528 			       v_idx);
2529 		break;
2530 	case ixgbe_mac_82599EB:
2531 	case ixgbe_mac_X540:
2532 	case ixgbe_mac_X550:
2533 	case ixgbe_mac_X550EM_x:
2534 	case ixgbe_mac_x550em_a:
2535 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2536 		break;
2537 	default:
2538 		break;
2539 	}
2540 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2541 
2542 	/* set up to autoclear timer, and the vectors */
2543 	mask = IXGBE_EIMS_ENABLE_MASK;
2544 	mask &= ~(IXGBE_EIMS_OTHER |
2545 		  IXGBE_EIMS_MAILBOX |
2546 		  IXGBE_EIMS_LSC);
2547 
2548 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2549 }
2550 
2551 /**
2552  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2553  * @q_vector: structure containing interrupt and ring information
2554  * @ring_container: structure containing ring performance data
2555  *
2556  *      Stores a new ITR value based on packets and byte
2557  *      counts during the last interrupt.  The advantage of per interrupt
2558  *      computation is faster updates and more accurate ITR for the current
2559  *      traffic pattern.  Constants in this function were computed
2560  *      based on theoretical maximum wire speed and thresholds were set based
2561  *      on testing data as well as attempting to minimize response time
2562  *      while increasing bulk throughput.
2563  **/
2564 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2565 			     struct ixgbe_ring_container *ring_container)
2566 {
2567 	unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2568 			   IXGBE_ITR_ADAPTIVE_LATENCY;
2569 	unsigned int avg_wire_size, packets, bytes;
2570 	unsigned long next_update = jiffies;
2571 
2572 	/* If we don't have any rings just leave ourselves set for maximum
2573 	 * possible latency so we take ourselves out of the equation.
2574 	 */
2575 	if (!ring_container->ring)
2576 		return;
2577 
2578 	/* If we didn't update within up to 1 - 2 jiffies we can assume
2579 	 * that either packets are coming in so slow there hasn't been
2580 	 * any work, or that there is so much work that NAPI is dealing
2581 	 * with interrupt moderation and we don't need to do anything.
2582 	 */
2583 	if (time_after(next_update, ring_container->next_update))
2584 		goto clear_counts;
2585 
2586 	packets = ring_container->total_packets;
2587 
2588 	/* We have no packets to actually measure against. This means
2589 	 * either one of the other queues on this vector is active or
2590 	 * we are a Tx queue doing TSO with too high of an interrupt rate.
2591 	 *
2592 	 * When this occurs just tick up our delay by the minimum value
2593 	 * and hope that this extra delay will prevent us from being called
2594 	 * without any work on our queue.
2595 	 */
2596 	if (!packets) {
2597 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2598 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2599 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2600 		itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2601 		goto clear_counts;
2602 	}
2603 
2604 	bytes = ring_container->total_bytes;
2605 
2606 	/* If packets are less than 4 or bytes are less than 9000 assume
2607 	 * insufficient data to use bulk rate limiting approach. We are
2608 	 * likely latency driven.
2609 	 */
2610 	if (packets < 4 && bytes < 9000) {
2611 		itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2612 		goto adjust_by_size;
2613 	}
2614 
2615 	/* Between 4 and 48 we can assume that our current interrupt delay
2616 	 * is only slightly too low. As such we should increase it by a small
2617 	 * fixed amount.
2618 	 */
2619 	if (packets < 48) {
2620 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2621 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2622 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2623 		goto clear_counts;
2624 	}
2625 
2626 	/* Between 48 and 96 is our "goldilocks" zone where we are working
2627 	 * out "just right". Just report that our current ITR is good for us.
2628 	 */
2629 	if (packets < 96) {
2630 		itr = q_vector->itr >> 2;
2631 		goto clear_counts;
2632 	}
2633 
2634 	/* If packet count is 96 or greater we are likely looking at a slight
2635 	 * overrun of the delay we want. Try halving our delay to see if that
2636 	 * will cut the number of packets in half per interrupt.
2637 	 */
2638 	if (packets < 256) {
2639 		itr = q_vector->itr >> 3;
2640 		if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2641 			itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2642 		goto clear_counts;
2643 	}
2644 
2645 	/* The paths below assume we are dealing with a bulk ITR since number
2646 	 * of packets is 256 or greater. We are just going to have to compute
2647 	 * a value and try to bring the count under control, though for smaller
2648 	 * packet sizes there isn't much we can do as NAPI polling will likely
2649 	 * be kicking in sooner rather than later.
2650 	 */
2651 	itr = IXGBE_ITR_ADAPTIVE_BULK;
2652 
2653 adjust_by_size:
2654 	/* If packet counts are 256 or greater we can assume we have a gross
2655 	 * overestimation of what the rate should be. Instead of trying to fine
2656 	 * tune it just use the formula below to try and dial in an exact value
2657 	 * give the current packet size of the frame.
2658 	 */
2659 	avg_wire_size = bytes / packets;
2660 
2661 	/* The following is a crude approximation of:
2662 	 *  wmem_default / (size + overhead) = desired_pkts_per_int
2663 	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2664 	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2665 	 *
2666 	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
2667 	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2668 	 * formula down to
2669 	 *
2670 	 *  (170 * (size + 24)) / (size + 640) = ITR
2671 	 *
2672 	 * We first do some math on the packet size and then finally bitshift
2673 	 * by 8 after rounding up. We also have to account for PCIe link speed
2674 	 * difference as ITR scales based on this.
2675 	 */
2676 	if (avg_wire_size <= 60) {
2677 		/* Start at 50k ints/sec */
2678 		avg_wire_size = 5120;
2679 	} else if (avg_wire_size <= 316) {
2680 		/* 50K ints/sec to 16K ints/sec */
2681 		avg_wire_size *= 40;
2682 		avg_wire_size += 2720;
2683 	} else if (avg_wire_size <= 1084) {
2684 		/* 16K ints/sec to 9.2K ints/sec */
2685 		avg_wire_size *= 15;
2686 		avg_wire_size += 11452;
2687 	} else if (avg_wire_size <= 1980) {
2688 		/* 9.2K ints/sec to 8K ints/sec */
2689 		avg_wire_size *= 5;
2690 		avg_wire_size += 22420;
2691 	} else {
2692 		/* plateau at a limit of 8K ints/sec */
2693 		avg_wire_size = 32256;
2694 	}
2695 
2696 	/* If we are in low latency mode half our delay which doubles the rate
2697 	 * to somewhere between 100K to 16K ints/sec
2698 	 */
2699 	if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2700 		avg_wire_size >>= 1;
2701 
2702 	/* Resultant value is 256 times larger than it needs to be. This
2703 	 * gives us room to adjust the value as needed to either increase
2704 	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2705 	 *
2706 	 * Use addition as we have already recorded the new latency flag
2707 	 * for the ITR value.
2708 	 */
2709 	switch (q_vector->adapter->link_speed) {
2710 	case IXGBE_LINK_SPEED_10GB_FULL:
2711 	case IXGBE_LINK_SPEED_100_FULL:
2712 	default:
2713 		itr += DIV_ROUND_UP(avg_wire_size,
2714 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2715 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2716 		break;
2717 	case IXGBE_LINK_SPEED_2_5GB_FULL:
2718 	case IXGBE_LINK_SPEED_1GB_FULL:
2719 	case IXGBE_LINK_SPEED_10_FULL:
2720 		itr += DIV_ROUND_UP(avg_wire_size,
2721 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2722 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2723 		break;
2724 	}
2725 
2726 clear_counts:
2727 	/* write back value */
2728 	ring_container->itr = itr;
2729 
2730 	/* next update should occur within next jiffy */
2731 	ring_container->next_update = next_update + 1;
2732 
2733 	ring_container->total_bytes = 0;
2734 	ring_container->total_packets = 0;
2735 }
2736 
2737 /**
2738  * ixgbe_write_eitr - write EITR register in hardware specific way
2739  * @q_vector: structure containing interrupt and ring information
2740  *
2741  * This function is made to be called by ethtool and by the driver
2742  * when it needs to update EITR registers at runtime.  Hardware
2743  * specific quirks/differences are taken care of here.
2744  */
2745 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2746 {
2747 	struct ixgbe_adapter *adapter = q_vector->adapter;
2748 	struct ixgbe_hw *hw = &adapter->hw;
2749 	int v_idx = q_vector->v_idx;
2750 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2751 
2752 	switch (adapter->hw.mac.type) {
2753 	case ixgbe_mac_82598EB:
2754 		/* must write high and low 16 bits to reset counter */
2755 		itr_reg |= (itr_reg << 16);
2756 		break;
2757 	case ixgbe_mac_82599EB:
2758 	case ixgbe_mac_X540:
2759 	case ixgbe_mac_X550:
2760 	case ixgbe_mac_X550EM_x:
2761 	case ixgbe_mac_x550em_a:
2762 		/*
2763 		 * set the WDIS bit to not clear the timer bits and cause an
2764 		 * immediate assertion of the interrupt
2765 		 */
2766 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2767 		break;
2768 	default:
2769 		break;
2770 	}
2771 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2772 }
2773 
2774 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2775 {
2776 	u32 new_itr;
2777 
2778 	ixgbe_update_itr(q_vector, &q_vector->tx);
2779 	ixgbe_update_itr(q_vector, &q_vector->rx);
2780 
2781 	/* use the smallest value of new ITR delay calculations */
2782 	new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2783 
2784 	/* Clear latency flag if set, shift into correct position */
2785 	new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2786 	new_itr <<= 2;
2787 
2788 	if (new_itr != q_vector->itr) {
2789 		/* save the algorithm value here */
2790 		q_vector->itr = new_itr;
2791 
2792 		ixgbe_write_eitr(q_vector);
2793 	}
2794 }
2795 
2796 /**
2797  * ixgbe_check_overtemp_subtask - check for over temperature
2798  * @adapter: pointer to adapter
2799  **/
2800 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2801 {
2802 	struct ixgbe_hw *hw = &adapter->hw;
2803 	u32 eicr = adapter->interrupt_event;
2804 	s32 rc;
2805 
2806 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2807 		return;
2808 
2809 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2810 		return;
2811 
2812 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2813 
2814 	switch (hw->device_id) {
2815 	case IXGBE_DEV_ID_82599_T3_LOM:
2816 		/*
2817 		 * Since the warning interrupt is for both ports
2818 		 * we don't have to check if:
2819 		 *  - This interrupt wasn't for our port.
2820 		 *  - We may have missed the interrupt so always have to
2821 		 *    check if we  got a LSC
2822 		 */
2823 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2824 		    !(eicr & IXGBE_EICR_LSC))
2825 			return;
2826 
2827 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2828 			u32 speed;
2829 			bool link_up = false;
2830 
2831 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2832 
2833 			if (link_up)
2834 				return;
2835 		}
2836 
2837 		/* Check if this is not due to overtemp */
2838 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2839 			return;
2840 
2841 		break;
2842 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2843 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2844 		rc = hw->phy.ops.check_overtemp(hw);
2845 		if (rc != IXGBE_ERR_OVERTEMP)
2846 			return;
2847 		break;
2848 	default:
2849 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2850 			return;
2851 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2852 			return;
2853 		break;
2854 	}
2855 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2856 
2857 	adapter->interrupt_event = 0;
2858 }
2859 
2860 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2861 {
2862 	struct ixgbe_hw *hw = &adapter->hw;
2863 
2864 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2865 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2866 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2867 		/* write to clear the interrupt */
2868 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2869 	}
2870 }
2871 
2872 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2873 {
2874 	struct ixgbe_hw *hw = &adapter->hw;
2875 
2876 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2877 		return;
2878 
2879 	switch (adapter->hw.mac.type) {
2880 	case ixgbe_mac_82599EB:
2881 		/*
2882 		 * Need to check link state so complete overtemp check
2883 		 * on service task
2884 		 */
2885 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2886 		     (eicr & IXGBE_EICR_LSC)) &&
2887 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2888 			adapter->interrupt_event = eicr;
2889 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2890 			ixgbe_service_event_schedule(adapter);
2891 			return;
2892 		}
2893 		return;
2894 	case ixgbe_mac_x550em_a:
2895 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2896 			adapter->interrupt_event = eicr;
2897 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2898 			ixgbe_service_event_schedule(adapter);
2899 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2900 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2901 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2902 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2903 		}
2904 		return;
2905 	case ixgbe_mac_X550:
2906 	case ixgbe_mac_X540:
2907 		if (!(eicr & IXGBE_EICR_TS))
2908 			return;
2909 		break;
2910 	default:
2911 		return;
2912 	}
2913 
2914 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2915 }
2916 
2917 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2918 {
2919 	switch (hw->mac.type) {
2920 	case ixgbe_mac_82598EB:
2921 		if (hw->phy.type == ixgbe_phy_nl)
2922 			return true;
2923 		return false;
2924 	case ixgbe_mac_82599EB:
2925 	case ixgbe_mac_X550EM_x:
2926 	case ixgbe_mac_x550em_a:
2927 		switch (hw->mac.ops.get_media_type(hw)) {
2928 		case ixgbe_media_type_fiber:
2929 		case ixgbe_media_type_fiber_qsfp:
2930 			return true;
2931 		default:
2932 			return false;
2933 		}
2934 	default:
2935 		return false;
2936 	}
2937 }
2938 
2939 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2940 {
2941 	struct ixgbe_hw *hw = &adapter->hw;
2942 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2943 
2944 	if (!ixgbe_is_sfp(hw))
2945 		return;
2946 
2947 	/* Later MAC's use different SDP */
2948 	if (hw->mac.type >= ixgbe_mac_X540)
2949 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2950 
2951 	if (eicr & eicr_mask) {
2952 		/* Clear the interrupt */
2953 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2954 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2955 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2956 			adapter->sfp_poll_time = 0;
2957 			ixgbe_service_event_schedule(adapter);
2958 		}
2959 	}
2960 
2961 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2962 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2963 		/* Clear the interrupt */
2964 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2965 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2966 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2967 			ixgbe_service_event_schedule(adapter);
2968 		}
2969 	}
2970 }
2971 
2972 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2973 {
2974 	struct ixgbe_hw *hw = &adapter->hw;
2975 
2976 	adapter->lsc_int++;
2977 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2978 	adapter->link_check_timeout = jiffies;
2979 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2980 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2981 		IXGBE_WRITE_FLUSH(hw);
2982 		ixgbe_service_event_schedule(adapter);
2983 	}
2984 }
2985 
2986 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2987 					   u64 qmask)
2988 {
2989 	u32 mask;
2990 	struct ixgbe_hw *hw = &adapter->hw;
2991 
2992 	switch (hw->mac.type) {
2993 	case ixgbe_mac_82598EB:
2994 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2995 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2996 		break;
2997 	case ixgbe_mac_82599EB:
2998 	case ixgbe_mac_X540:
2999 	case ixgbe_mac_X550:
3000 	case ixgbe_mac_X550EM_x:
3001 	case ixgbe_mac_x550em_a:
3002 		mask = (qmask & 0xFFFFFFFF);
3003 		if (mask)
3004 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
3005 		mask = (qmask >> 32);
3006 		if (mask)
3007 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
3008 		break;
3009 	default:
3010 		break;
3011 	}
3012 	/* skip the flush */
3013 }
3014 
3015 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
3016 					    u64 qmask)
3017 {
3018 	u32 mask;
3019 	struct ixgbe_hw *hw = &adapter->hw;
3020 
3021 	switch (hw->mac.type) {
3022 	case ixgbe_mac_82598EB:
3023 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
3024 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
3025 		break;
3026 	case ixgbe_mac_82599EB:
3027 	case ixgbe_mac_X540:
3028 	case ixgbe_mac_X550:
3029 	case ixgbe_mac_X550EM_x:
3030 	case ixgbe_mac_x550em_a:
3031 		mask = (qmask & 0xFFFFFFFF);
3032 		if (mask)
3033 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
3034 		mask = (qmask >> 32);
3035 		if (mask)
3036 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
3037 		break;
3038 	default:
3039 		break;
3040 	}
3041 	/* skip the flush */
3042 }
3043 
3044 /**
3045  * ixgbe_irq_enable - Enable default interrupt generation settings
3046  * @adapter: board private structure
3047  * @queues: enable irqs for queues
3048  * @flush: flush register write
3049  **/
3050 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
3051 				    bool flush)
3052 {
3053 	struct ixgbe_hw *hw = &adapter->hw;
3054 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
3055 
3056 	/* don't reenable LSC while waiting for link */
3057 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
3058 		mask &= ~IXGBE_EIMS_LSC;
3059 
3060 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
3061 		switch (adapter->hw.mac.type) {
3062 		case ixgbe_mac_82599EB:
3063 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
3064 			break;
3065 		case ixgbe_mac_X540:
3066 		case ixgbe_mac_X550:
3067 		case ixgbe_mac_X550EM_x:
3068 		case ixgbe_mac_x550em_a:
3069 			mask |= IXGBE_EIMS_TS;
3070 			break;
3071 		default:
3072 			break;
3073 		}
3074 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3075 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3076 	switch (adapter->hw.mac.type) {
3077 	case ixgbe_mac_82599EB:
3078 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3079 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
3080 		/* fall through */
3081 	case ixgbe_mac_X540:
3082 	case ixgbe_mac_X550:
3083 	case ixgbe_mac_X550EM_x:
3084 	case ixgbe_mac_x550em_a:
3085 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3086 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3087 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3088 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3089 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3090 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3091 		mask |= IXGBE_EIMS_ECC;
3092 		mask |= IXGBE_EIMS_MAILBOX;
3093 		break;
3094 	default:
3095 		break;
3096 	}
3097 
3098 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3099 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3100 		mask |= IXGBE_EIMS_FLOW_DIR;
3101 
3102 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3103 	if (queues)
3104 		ixgbe_irq_enable_queues(adapter, ~0);
3105 	if (flush)
3106 		IXGBE_WRITE_FLUSH(&adapter->hw);
3107 }
3108 
3109 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3110 {
3111 	struct ixgbe_adapter *adapter = data;
3112 	struct ixgbe_hw *hw = &adapter->hw;
3113 	u32 eicr;
3114 
3115 	/*
3116 	 * Workaround for Silicon errata.  Use clear-by-write instead
3117 	 * of clear-by-read.  Reading with EICS will return the
3118 	 * interrupt causes without clearing, which later be done
3119 	 * with the write to EICR.
3120 	 */
3121 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3122 
3123 	/* The lower 16bits of the EICR register are for the queue interrupts
3124 	 * which should be masked here in order to not accidentally clear them if
3125 	 * the bits are high when ixgbe_msix_other is called. There is a race
3126 	 * condition otherwise which results in possible performance loss
3127 	 * especially if the ixgbe_msix_other interrupt is triggering
3128 	 * consistently (as it would when PPS is turned on for the X540 device)
3129 	 */
3130 	eicr &= 0xFFFF0000;
3131 
3132 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3133 
3134 	if (eicr & IXGBE_EICR_LSC)
3135 		ixgbe_check_lsc(adapter);
3136 
3137 	if (eicr & IXGBE_EICR_MAILBOX)
3138 		ixgbe_msg_task(adapter);
3139 
3140 	switch (hw->mac.type) {
3141 	case ixgbe_mac_82599EB:
3142 	case ixgbe_mac_X540:
3143 	case ixgbe_mac_X550:
3144 	case ixgbe_mac_X550EM_x:
3145 	case ixgbe_mac_x550em_a:
3146 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3147 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3148 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3149 			ixgbe_service_event_schedule(adapter);
3150 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3151 					IXGBE_EICR_GPI_SDP0_X540);
3152 		}
3153 		if (eicr & IXGBE_EICR_ECC) {
3154 			e_info(link, "Received ECC Err, initiating reset\n");
3155 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3156 			ixgbe_service_event_schedule(adapter);
3157 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3158 		}
3159 		/* Handle Flow Director Full threshold interrupt */
3160 		if (eicr & IXGBE_EICR_FLOW_DIR) {
3161 			int reinit_count = 0;
3162 			int i;
3163 			for (i = 0; i < adapter->num_tx_queues; i++) {
3164 				struct ixgbe_ring *ring = adapter->tx_ring[i];
3165 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3166 						       &ring->state))
3167 					reinit_count++;
3168 			}
3169 			if (reinit_count) {
3170 				/* no more flow director interrupts until after init */
3171 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3172 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3173 				ixgbe_service_event_schedule(adapter);
3174 			}
3175 		}
3176 		ixgbe_check_sfp_event(adapter, eicr);
3177 		ixgbe_check_overtemp_event(adapter, eicr);
3178 		break;
3179 	default:
3180 		break;
3181 	}
3182 
3183 	ixgbe_check_fan_failure(adapter, eicr);
3184 
3185 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3186 		ixgbe_ptp_check_pps_event(adapter);
3187 
3188 	/* re-enable the original interrupt state, no lsc, no queues */
3189 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3190 		ixgbe_irq_enable(adapter, false, false);
3191 
3192 	return IRQ_HANDLED;
3193 }
3194 
3195 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3196 {
3197 	struct ixgbe_q_vector *q_vector = data;
3198 
3199 	/* EIAM disabled interrupts (on this vector) for us */
3200 
3201 	if (q_vector->rx.ring || q_vector->tx.ring)
3202 		napi_schedule_irqoff(&q_vector->napi);
3203 
3204 	return IRQ_HANDLED;
3205 }
3206 
3207 /**
3208  * ixgbe_poll - NAPI Rx polling callback
3209  * @napi: structure for representing this polling device
3210  * @budget: how many packets driver is allowed to clean
3211  *
3212  * This function is used for legacy and MSI, NAPI mode
3213  **/
3214 int ixgbe_poll(struct napi_struct *napi, int budget)
3215 {
3216 	struct ixgbe_q_vector *q_vector =
3217 				container_of(napi, struct ixgbe_q_vector, napi);
3218 	struct ixgbe_adapter *adapter = q_vector->adapter;
3219 	struct ixgbe_ring *ring;
3220 	int per_ring_budget, work_done = 0;
3221 	bool clean_complete = true;
3222 
3223 #ifdef CONFIG_IXGBE_DCA
3224 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3225 		ixgbe_update_dca(q_vector);
3226 #endif
3227 
3228 	ixgbe_for_each_ring(ring, q_vector->tx) {
3229 		if (!ixgbe_clean_tx_irq(q_vector, ring, budget))
3230 			clean_complete = false;
3231 	}
3232 
3233 	/* Exit if we are called by netpoll */
3234 	if (budget <= 0)
3235 		return budget;
3236 
3237 	/* attempt to distribute budget to each queue fairly, but don't allow
3238 	 * the budget to go below 1 because we'll exit polling */
3239 	if (q_vector->rx.count > 1)
3240 		per_ring_budget = max(budget/q_vector->rx.count, 1);
3241 	else
3242 		per_ring_budget = budget;
3243 
3244 	ixgbe_for_each_ring(ring, q_vector->rx) {
3245 		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
3246 						 per_ring_budget);
3247 
3248 		work_done += cleaned;
3249 		if (cleaned >= per_ring_budget)
3250 			clean_complete = false;
3251 	}
3252 
3253 	/* If all work not completed, return budget and keep polling */
3254 	if (!clean_complete)
3255 		return budget;
3256 
3257 	/* all work done, exit the polling mode */
3258 	napi_complete_done(napi, work_done);
3259 	if (adapter->rx_itr_setting & 1)
3260 		ixgbe_set_itr(q_vector);
3261 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3262 		ixgbe_irq_enable_queues(adapter, BIT_ULL(q_vector->v_idx));
3263 
3264 	return min(work_done, budget - 1);
3265 }
3266 
3267 /**
3268  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3269  * @adapter: board private structure
3270  *
3271  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3272  * interrupts from the kernel.
3273  **/
3274 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3275 {
3276 	struct net_device *netdev = adapter->netdev;
3277 	unsigned int ri = 0, ti = 0;
3278 	int vector, err;
3279 
3280 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3281 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3282 		struct msix_entry *entry = &adapter->msix_entries[vector];
3283 
3284 		if (q_vector->tx.ring && q_vector->rx.ring) {
3285 			snprintf(q_vector->name, sizeof(q_vector->name),
3286 				 "%s-TxRx-%u", netdev->name, ri++);
3287 			ti++;
3288 		} else if (q_vector->rx.ring) {
3289 			snprintf(q_vector->name, sizeof(q_vector->name),
3290 				 "%s-rx-%u", netdev->name, ri++);
3291 		} else if (q_vector->tx.ring) {
3292 			snprintf(q_vector->name, sizeof(q_vector->name),
3293 				 "%s-tx-%u", netdev->name, ti++);
3294 		} else {
3295 			/* skip this unused q_vector */
3296 			continue;
3297 		}
3298 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3299 				  q_vector->name, q_vector);
3300 		if (err) {
3301 			e_err(probe, "request_irq failed for MSIX interrupt "
3302 			      "Error: %d\n", err);
3303 			goto free_queue_irqs;
3304 		}
3305 		/* If Flow Director is enabled, set interrupt affinity */
3306 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3307 			/* assign the mask for this irq */
3308 			irq_set_affinity_hint(entry->vector,
3309 					      &q_vector->affinity_mask);
3310 		}
3311 	}
3312 
3313 	err = request_irq(adapter->msix_entries[vector].vector,
3314 			  ixgbe_msix_other, 0, netdev->name, adapter);
3315 	if (err) {
3316 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3317 		goto free_queue_irqs;
3318 	}
3319 
3320 	return 0;
3321 
3322 free_queue_irqs:
3323 	while (vector) {
3324 		vector--;
3325 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
3326 				      NULL);
3327 		free_irq(adapter->msix_entries[vector].vector,
3328 			 adapter->q_vector[vector]);
3329 	}
3330 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3331 	pci_disable_msix(adapter->pdev);
3332 	kfree(adapter->msix_entries);
3333 	adapter->msix_entries = NULL;
3334 	return err;
3335 }
3336 
3337 /**
3338  * ixgbe_intr - legacy mode Interrupt Handler
3339  * @irq: interrupt number
3340  * @data: pointer to a network interface device structure
3341  **/
3342 static irqreturn_t ixgbe_intr(int irq, void *data)
3343 {
3344 	struct ixgbe_adapter *adapter = data;
3345 	struct ixgbe_hw *hw = &adapter->hw;
3346 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3347 	u32 eicr;
3348 
3349 	/*
3350 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3351 	 * before the read of EICR.
3352 	 */
3353 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3354 
3355 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3356 	 * therefore no explicit interrupt disable is necessary */
3357 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3358 	if (!eicr) {
3359 		/*
3360 		 * shared interrupt alert!
3361 		 * make sure interrupts are enabled because the read will
3362 		 * have disabled interrupts due to EIAM
3363 		 * finish the workaround of silicon errata on 82598.  Unmask
3364 		 * the interrupt that we masked before the EICR read.
3365 		 */
3366 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3367 			ixgbe_irq_enable(adapter, true, true);
3368 		return IRQ_NONE;	/* Not our interrupt */
3369 	}
3370 
3371 	if (eicr & IXGBE_EICR_LSC)
3372 		ixgbe_check_lsc(adapter);
3373 
3374 	switch (hw->mac.type) {
3375 	case ixgbe_mac_82599EB:
3376 		ixgbe_check_sfp_event(adapter, eicr);
3377 		/* Fall through */
3378 	case ixgbe_mac_X540:
3379 	case ixgbe_mac_X550:
3380 	case ixgbe_mac_X550EM_x:
3381 	case ixgbe_mac_x550em_a:
3382 		if (eicr & IXGBE_EICR_ECC) {
3383 			e_info(link, "Received ECC Err, initiating reset\n");
3384 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3385 			ixgbe_service_event_schedule(adapter);
3386 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3387 		}
3388 		ixgbe_check_overtemp_event(adapter, eicr);
3389 		break;
3390 	default:
3391 		break;
3392 	}
3393 
3394 	ixgbe_check_fan_failure(adapter, eicr);
3395 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3396 		ixgbe_ptp_check_pps_event(adapter);
3397 
3398 	/* would disable interrupts here but EIAM disabled it */
3399 	napi_schedule_irqoff(&q_vector->napi);
3400 
3401 	/*
3402 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3403 	 * ixgbe_poll will re-enable the queue interrupts
3404 	 */
3405 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3406 		ixgbe_irq_enable(adapter, false, false);
3407 
3408 	return IRQ_HANDLED;
3409 }
3410 
3411 /**
3412  * ixgbe_request_irq - initialize interrupts
3413  * @adapter: board private structure
3414  *
3415  * Attempts to configure interrupts using the best available
3416  * capabilities of the hardware and kernel.
3417  **/
3418 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3419 {
3420 	struct net_device *netdev = adapter->netdev;
3421 	int err;
3422 
3423 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3424 		err = ixgbe_request_msix_irqs(adapter);
3425 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3426 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3427 				  netdev->name, adapter);
3428 	else
3429 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3430 				  netdev->name, adapter);
3431 
3432 	if (err)
3433 		e_err(probe, "request_irq failed, Error %d\n", err);
3434 
3435 	return err;
3436 }
3437 
3438 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3439 {
3440 	int vector;
3441 
3442 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3443 		free_irq(adapter->pdev->irq, adapter);
3444 		return;
3445 	}
3446 
3447 	if (!adapter->msix_entries)
3448 		return;
3449 
3450 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3451 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3452 		struct msix_entry *entry = &adapter->msix_entries[vector];
3453 
3454 		/* free only the irqs that were actually requested */
3455 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3456 			continue;
3457 
3458 		/* clear the affinity_mask in the IRQ descriptor */
3459 		irq_set_affinity_hint(entry->vector, NULL);
3460 
3461 		free_irq(entry->vector, q_vector);
3462 	}
3463 
3464 	free_irq(adapter->msix_entries[vector].vector, adapter);
3465 }
3466 
3467 /**
3468  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3469  * @adapter: board private structure
3470  **/
3471 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3472 {
3473 	switch (adapter->hw.mac.type) {
3474 	case ixgbe_mac_82598EB:
3475 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3476 		break;
3477 	case ixgbe_mac_82599EB:
3478 	case ixgbe_mac_X540:
3479 	case ixgbe_mac_X550:
3480 	case ixgbe_mac_X550EM_x:
3481 	case ixgbe_mac_x550em_a:
3482 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3483 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3484 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3485 		break;
3486 	default:
3487 		break;
3488 	}
3489 	IXGBE_WRITE_FLUSH(&adapter->hw);
3490 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3491 		int vector;
3492 
3493 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3494 			synchronize_irq(adapter->msix_entries[vector].vector);
3495 
3496 		synchronize_irq(adapter->msix_entries[vector++].vector);
3497 	} else {
3498 		synchronize_irq(adapter->pdev->irq);
3499 	}
3500 }
3501 
3502 /**
3503  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3504  * @adapter: board private structure
3505  *
3506  **/
3507 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3508 {
3509 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3510 
3511 	ixgbe_write_eitr(q_vector);
3512 
3513 	ixgbe_set_ivar(adapter, 0, 0, 0);
3514 	ixgbe_set_ivar(adapter, 1, 0, 0);
3515 
3516 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3517 }
3518 
3519 /**
3520  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3521  * @adapter: board private structure
3522  * @ring: structure containing ring specific data
3523  *
3524  * Configure the Tx descriptor ring after a reset.
3525  **/
3526 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3527 			     struct ixgbe_ring *ring)
3528 {
3529 	struct ixgbe_hw *hw = &adapter->hw;
3530 	u64 tdba = ring->dma;
3531 	int wait_loop = 10;
3532 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3533 	u8 reg_idx = ring->reg_idx;
3534 
3535 	/* disable queue to avoid issues while updating state */
3536 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3537 	IXGBE_WRITE_FLUSH(hw);
3538 
3539 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3540 			(tdba & DMA_BIT_MASK(32)));
3541 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3542 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3543 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3544 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3545 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3546 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3547 
3548 	/*
3549 	 * set WTHRESH to encourage burst writeback, it should not be set
3550 	 * higher than 1 when:
3551 	 * - ITR is 0 as it could cause false TX hangs
3552 	 * - ITR is set to > 100k int/sec and BQL is enabled
3553 	 *
3554 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3555 	 * to or less than the number of on chip descriptors, which is
3556 	 * currently 40.
3557 	 */
3558 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3559 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3560 	else
3561 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3562 
3563 	/*
3564 	 * Setting PTHRESH to 32 both improves performance
3565 	 * and avoids a TX hang with DFP enabled
3566 	 */
3567 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3568 		   32;		/* PTHRESH = 32 */
3569 
3570 	/* reinitialize flowdirector state */
3571 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3572 		ring->atr_sample_rate = adapter->atr_sample_rate;
3573 		ring->atr_count = 0;
3574 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3575 	} else {
3576 		ring->atr_sample_rate = 0;
3577 	}
3578 
3579 	/* initialize XPS */
3580 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3581 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3582 
3583 		if (q_vector)
3584 			netif_set_xps_queue(ring->netdev,
3585 					    &q_vector->affinity_mask,
3586 					    ring->queue_index);
3587 	}
3588 
3589 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3590 
3591 	/* reinitialize tx_buffer_info */
3592 	memset(ring->tx_buffer_info, 0,
3593 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3594 
3595 	/* enable queue */
3596 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3597 
3598 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3599 	if (hw->mac.type == ixgbe_mac_82598EB &&
3600 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3601 		return;
3602 
3603 	/* poll to verify queue is enabled */
3604 	do {
3605 		usleep_range(1000, 2000);
3606 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3607 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3608 	if (!wait_loop)
3609 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3610 }
3611 
3612 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3613 {
3614 	struct ixgbe_hw *hw = &adapter->hw;
3615 	u32 rttdcs, mtqc;
3616 	u8 tcs = adapter->hw_tcs;
3617 
3618 	if (hw->mac.type == ixgbe_mac_82598EB)
3619 		return;
3620 
3621 	/* disable the arbiter while setting MTQC */
3622 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3623 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3624 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3625 
3626 	/* set transmit pool layout */
3627 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3628 		mtqc = IXGBE_MTQC_VT_ENA;
3629 		if (tcs > 4)
3630 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3631 		else if (tcs > 1)
3632 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3633 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3634 			 IXGBE_82599_VMDQ_4Q_MASK)
3635 			mtqc |= IXGBE_MTQC_32VF;
3636 		else
3637 			mtqc |= IXGBE_MTQC_64VF;
3638 	} else {
3639 		if (tcs > 4)
3640 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3641 		else if (tcs > 1)
3642 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3643 		else
3644 			mtqc = IXGBE_MTQC_64Q_1PB;
3645 	}
3646 
3647 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3648 
3649 	/* Enable Security TX Buffer IFG for multiple pb */
3650 	if (tcs) {
3651 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3652 		sectx |= IXGBE_SECTX_DCB;
3653 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3654 	}
3655 
3656 	/* re-enable the arbiter */
3657 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3658 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3659 }
3660 
3661 /**
3662  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3663  * @adapter: board private structure
3664  *
3665  * Configure the Tx unit of the MAC after a reset.
3666  **/
3667 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3668 {
3669 	struct ixgbe_hw *hw = &adapter->hw;
3670 	u32 dmatxctl;
3671 	u32 i;
3672 
3673 	ixgbe_setup_mtqc(adapter);
3674 
3675 	if (hw->mac.type != ixgbe_mac_82598EB) {
3676 		/* DMATXCTL.EN must be before Tx queues are enabled */
3677 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3678 		dmatxctl |= IXGBE_DMATXCTL_TE;
3679 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3680 	}
3681 
3682 	/* Setup the HW Tx Head and Tail descriptor pointers */
3683 	for (i = 0; i < adapter->num_tx_queues; i++)
3684 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3685 	for (i = 0; i < adapter->num_xdp_queues; i++)
3686 		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3687 }
3688 
3689 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3690 				 struct ixgbe_ring *ring)
3691 {
3692 	struct ixgbe_hw *hw = &adapter->hw;
3693 	u8 reg_idx = ring->reg_idx;
3694 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3695 
3696 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3697 
3698 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3699 }
3700 
3701 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3702 				  struct ixgbe_ring *ring)
3703 {
3704 	struct ixgbe_hw *hw = &adapter->hw;
3705 	u8 reg_idx = ring->reg_idx;
3706 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3707 
3708 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3709 
3710 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3711 }
3712 
3713 #ifdef CONFIG_IXGBE_DCB
3714 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3715 #else
3716 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3717 #endif
3718 {
3719 	int i;
3720 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3721 
3722 	if (adapter->ixgbe_ieee_pfc)
3723 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3724 
3725 	/*
3726 	 * We should set the drop enable bit if:
3727 	 *  SR-IOV is enabled
3728 	 *   or
3729 	 *  Number of Rx queues > 1 and flow control is disabled
3730 	 *
3731 	 *  This allows us to avoid head of line blocking for security
3732 	 *  and performance reasons.
3733 	 */
3734 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3735 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3736 		for (i = 0; i < adapter->num_rx_queues; i++)
3737 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3738 	} else {
3739 		for (i = 0; i < adapter->num_rx_queues; i++)
3740 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3741 	}
3742 }
3743 
3744 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3745 
3746 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3747 				   struct ixgbe_ring *rx_ring)
3748 {
3749 	struct ixgbe_hw *hw = &adapter->hw;
3750 	u32 srrctl;
3751 	u8 reg_idx = rx_ring->reg_idx;
3752 
3753 	if (hw->mac.type == ixgbe_mac_82598EB) {
3754 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3755 
3756 		/*
3757 		 * if VMDq is not active we must program one srrctl register
3758 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3759 		 */
3760 		reg_idx &= mask;
3761 	}
3762 
3763 	/* configure header buffer length, needed for RSC */
3764 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3765 
3766 	/* configure the packet buffer length */
3767 	if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state))
3768 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3769 	else
3770 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3771 
3772 	/* configure descriptor type */
3773 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3774 
3775 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3776 }
3777 
3778 /**
3779  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3780  * @adapter: device handle
3781  *
3782  *  - 82598/82599/X540:     128
3783  *  - X550(non-SRIOV mode): 512
3784  *  - X550(SRIOV mode):     64
3785  */
3786 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3787 {
3788 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3789 		return 128;
3790 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3791 		return 64;
3792 	else
3793 		return 512;
3794 }
3795 
3796 /**
3797  * ixgbe_store_key - Write the RSS key to HW
3798  * @adapter: device handle
3799  *
3800  * Write the RSS key stored in adapter.rss_key to HW.
3801  */
3802 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3803 {
3804 	struct ixgbe_hw *hw = &adapter->hw;
3805 	int i;
3806 
3807 	for (i = 0; i < 10; i++)
3808 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3809 }
3810 
3811 /**
3812  * ixgbe_init_rss_key - Initialize adapter RSS key
3813  * @adapter: device handle
3814  *
3815  * Allocates and initializes the RSS key if it is not allocated.
3816  **/
3817 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3818 {
3819 	u32 *rss_key;
3820 
3821 	if (!adapter->rss_key) {
3822 		rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3823 		if (unlikely(!rss_key))
3824 			return -ENOMEM;
3825 
3826 		netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3827 		adapter->rss_key = rss_key;
3828 	}
3829 
3830 	return 0;
3831 }
3832 
3833 /**
3834  * ixgbe_store_reta - Write the RETA table to HW
3835  * @adapter: device handle
3836  *
3837  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3838  */
3839 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3840 {
3841 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3842 	struct ixgbe_hw *hw = &adapter->hw;
3843 	u32 reta = 0;
3844 	u32 indices_multi;
3845 	u8 *indir_tbl = adapter->rss_indir_tbl;
3846 
3847 	/* Fill out the redirection table as follows:
3848 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3849 	 *    indices.
3850 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3851 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3852 	 */
3853 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3854 		indices_multi = 0x11;
3855 	else
3856 		indices_multi = 0x1;
3857 
3858 	/* Write redirection table to HW */
3859 	for (i = 0; i < reta_entries; i++) {
3860 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3861 		if ((i & 3) == 3) {
3862 			if (i < 128)
3863 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3864 			else
3865 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3866 						reta);
3867 			reta = 0;
3868 		}
3869 	}
3870 }
3871 
3872 /**
3873  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3874  * @adapter: device handle
3875  *
3876  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3877  */
3878 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3879 {
3880 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3881 	struct ixgbe_hw *hw = &adapter->hw;
3882 	u32 vfreta = 0;
3883 
3884 	/* Write redirection table to HW */
3885 	for (i = 0; i < reta_entries; i++) {
3886 		u16 pool = adapter->num_rx_pools;
3887 
3888 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3889 		if ((i & 3) != 3)
3890 			continue;
3891 
3892 		while (pool--)
3893 			IXGBE_WRITE_REG(hw,
3894 					IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
3895 					vfreta);
3896 		vfreta = 0;
3897 	}
3898 }
3899 
3900 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3901 {
3902 	u32 i, j;
3903 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3904 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3905 
3906 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3907 	 * make full use of any rings they may have.  We will use the
3908 	 * PSRTYPE register to control how many rings we use within the PF.
3909 	 */
3910 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3911 		rss_i = 4;
3912 
3913 	/* Fill out hash function seeds */
3914 	ixgbe_store_key(adapter);
3915 
3916 	/* Fill out redirection table */
3917 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3918 
3919 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3920 		if (j == rss_i)
3921 			j = 0;
3922 
3923 		adapter->rss_indir_tbl[i] = j;
3924 	}
3925 
3926 	ixgbe_store_reta(adapter);
3927 }
3928 
3929 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3930 {
3931 	struct ixgbe_hw *hw = &adapter->hw;
3932 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3933 	int i, j;
3934 
3935 	/* Fill out hash function seeds */
3936 	for (i = 0; i < 10; i++) {
3937 		u16 pool = adapter->num_rx_pools;
3938 
3939 		while (pool--)
3940 			IXGBE_WRITE_REG(hw,
3941 					IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
3942 					*(adapter->rss_key + i));
3943 	}
3944 
3945 	/* Fill out the redirection table */
3946 	for (i = 0, j = 0; i < 64; i++, j++) {
3947 		if (j == rss_i)
3948 			j = 0;
3949 
3950 		adapter->rss_indir_tbl[i] = j;
3951 	}
3952 
3953 	ixgbe_store_vfreta(adapter);
3954 }
3955 
3956 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3957 {
3958 	struct ixgbe_hw *hw = &adapter->hw;
3959 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3960 	u32 rxcsum;
3961 
3962 	/* Disable indicating checksum in descriptor, enables RSS hash */
3963 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3964 	rxcsum |= IXGBE_RXCSUM_PCSD;
3965 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3966 
3967 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3968 		if (adapter->ring_feature[RING_F_RSS].mask)
3969 			mrqc = IXGBE_MRQC_RSSEN;
3970 	} else {
3971 		u8 tcs = adapter->hw_tcs;
3972 
3973 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3974 			if (tcs > 4)
3975 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3976 			else if (tcs > 1)
3977 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3978 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3979 				 IXGBE_82599_VMDQ_4Q_MASK)
3980 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3981 			else
3982 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3983 
3984 			/* Enable L3/L4 for Tx Switched packets */
3985 			mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3986 		} else {
3987 			if (tcs > 4)
3988 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3989 			else if (tcs > 1)
3990 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3991 			else
3992 				mrqc = IXGBE_MRQC_RSSEN;
3993 		}
3994 	}
3995 
3996 	/* Perform hash on these packet types */
3997 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3998 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3999 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
4000 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
4001 
4002 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
4003 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
4004 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
4005 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
4006 
4007 	if ((hw->mac.type >= ixgbe_mac_X550) &&
4008 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
4009 		u16 pool = adapter->num_rx_pools;
4010 
4011 		/* Enable VF RSS mode */
4012 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
4013 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4014 
4015 		/* Setup RSS through the VF registers */
4016 		ixgbe_setup_vfreta(adapter);
4017 		vfmrqc = IXGBE_MRQC_RSSEN;
4018 		vfmrqc |= rss_field;
4019 
4020 		while (pool--)
4021 			IXGBE_WRITE_REG(hw,
4022 					IXGBE_PFVFMRQC(VMDQ_P(pool)),
4023 					vfmrqc);
4024 	} else {
4025 		ixgbe_setup_reta(adapter);
4026 		mrqc |= rss_field;
4027 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4028 	}
4029 }
4030 
4031 /**
4032  * ixgbe_configure_rscctl - enable RSC for the indicated ring
4033  * @adapter: address of board private structure
4034  * @ring: structure containing ring specific data
4035  **/
4036 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
4037 				   struct ixgbe_ring *ring)
4038 {
4039 	struct ixgbe_hw *hw = &adapter->hw;
4040 	u32 rscctrl;
4041 	u8 reg_idx = ring->reg_idx;
4042 
4043 	if (!ring_is_rsc_enabled(ring))
4044 		return;
4045 
4046 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
4047 	rscctrl |= IXGBE_RSCCTL_RSCEN;
4048 	/*
4049 	 * we must limit the number of descriptors so that the
4050 	 * total size of max desc * buf_len is not greater
4051 	 * than 65536
4052 	 */
4053 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
4054 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
4055 }
4056 
4057 #define IXGBE_MAX_RX_DESC_POLL 10
4058 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4059 				       struct ixgbe_ring *ring)
4060 {
4061 	struct ixgbe_hw *hw = &adapter->hw;
4062 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4063 	u32 rxdctl;
4064 	u8 reg_idx = ring->reg_idx;
4065 
4066 	if (ixgbe_removed(hw->hw_addr))
4067 		return;
4068 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4069 	if (hw->mac.type == ixgbe_mac_82598EB &&
4070 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4071 		return;
4072 
4073 	do {
4074 		usleep_range(1000, 2000);
4075 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4076 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4077 
4078 	if (!wait_loop) {
4079 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4080 		      "the polling period\n", reg_idx);
4081 	}
4082 }
4083 
4084 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
4085 			    struct ixgbe_ring *ring)
4086 {
4087 	struct ixgbe_hw *hw = &adapter->hw;
4088 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4089 	u32 rxdctl;
4090 	u8 reg_idx = ring->reg_idx;
4091 
4092 	if (ixgbe_removed(hw->hw_addr))
4093 		return;
4094 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4095 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4096 
4097 	/* write value back with RXDCTL.ENABLE bit cleared */
4098 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4099 
4100 	if (hw->mac.type == ixgbe_mac_82598EB &&
4101 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4102 		return;
4103 
4104 	/* the hardware may take up to 100us to really disable the rx queue */
4105 	do {
4106 		udelay(10);
4107 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4108 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
4109 
4110 	if (!wait_loop) {
4111 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
4112 		      "the polling period\n", reg_idx);
4113 	}
4114 }
4115 
4116 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4117 			     struct ixgbe_ring *ring)
4118 {
4119 	struct ixgbe_hw *hw = &adapter->hw;
4120 	union ixgbe_adv_rx_desc *rx_desc;
4121 	u64 rdba = ring->dma;
4122 	u32 rxdctl;
4123 	u8 reg_idx = ring->reg_idx;
4124 
4125 	/* disable queue to avoid issues while updating state */
4126 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4127 	ixgbe_disable_rx_queue(adapter, ring);
4128 
4129 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4130 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4131 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4132 			ring->count * sizeof(union ixgbe_adv_rx_desc));
4133 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
4134 	IXGBE_WRITE_FLUSH(hw);
4135 
4136 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4137 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4138 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4139 
4140 	ixgbe_configure_srrctl(adapter, ring);
4141 	ixgbe_configure_rscctl(adapter, ring);
4142 
4143 	if (hw->mac.type == ixgbe_mac_82598EB) {
4144 		/*
4145 		 * enable cache line friendly hardware writes:
4146 		 * PTHRESH=32 descriptors (half the internal cache),
4147 		 * this also removes ugly rx_no_buffer_count increment
4148 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
4149 		 * WTHRESH=8 burst writeback up to two cache lines
4150 		 */
4151 		rxdctl &= ~0x3FFFFF;
4152 		rxdctl |=  0x080420;
4153 #if (PAGE_SIZE < 8192)
4154 	/* RXDCTL.RLPML does not work on 82599 */
4155 	} else if (hw->mac.type != ixgbe_mac_82599EB) {
4156 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4157 			    IXGBE_RXDCTL_RLPML_EN);
4158 
4159 		/* Limit the maximum frame size so we don't overrun the skb.
4160 		 * This can happen in SRIOV mode when the MTU of the VF is
4161 		 * higher than the MTU of the PF.
4162 		 */
4163 		if (ring_uses_build_skb(ring) &&
4164 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4165 			rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4166 				  IXGBE_RXDCTL_RLPML_EN;
4167 #endif
4168 	}
4169 
4170 	/* initialize rx_buffer_info */
4171 	memset(ring->rx_buffer_info, 0,
4172 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
4173 
4174 	/* initialize Rx descriptor 0 */
4175 	rx_desc = IXGBE_RX_DESC(ring, 0);
4176 	rx_desc->wb.upper.length = 0;
4177 
4178 	/* enable receive descriptor ring */
4179 	rxdctl |= IXGBE_RXDCTL_ENABLE;
4180 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4181 
4182 	ixgbe_rx_desc_queue_enable(adapter, ring);
4183 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4184 }
4185 
4186 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4187 {
4188 	struct ixgbe_hw *hw = &adapter->hw;
4189 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4190 	u16 pool = adapter->num_rx_pools;
4191 
4192 	/* PSRTYPE must be initialized in non 82598 adapters */
4193 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4194 		      IXGBE_PSRTYPE_UDPHDR |
4195 		      IXGBE_PSRTYPE_IPV4HDR |
4196 		      IXGBE_PSRTYPE_L2HDR |
4197 		      IXGBE_PSRTYPE_IPV6HDR;
4198 
4199 	if (hw->mac.type == ixgbe_mac_82598EB)
4200 		return;
4201 
4202 	if (rss_i > 3)
4203 		psrtype |= 2u << 29;
4204 	else if (rss_i > 1)
4205 		psrtype |= 1u << 29;
4206 
4207 	while (pool--)
4208 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4209 }
4210 
4211 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4212 {
4213 	struct ixgbe_hw *hw = &adapter->hw;
4214 	u32 reg_offset, vf_shift;
4215 	u32 gcr_ext, vmdctl;
4216 	int i;
4217 
4218 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4219 		return;
4220 
4221 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4222 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4223 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4224 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4225 	vmdctl |= IXGBE_VT_CTL_REPLEN;
4226 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4227 
4228 	vf_shift = VMDQ_P(0) % 32;
4229 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4230 
4231 	/* Enable only the PF's pool for Tx/Rx */
4232 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4233 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4234 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4235 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4236 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4237 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4238 
4239 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4240 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4241 
4242 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
4243 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4244 
4245 	/*
4246 	 * Set up VF register offsets for selected VT Mode,
4247 	 * i.e. 32 or 64 VFs for SR-IOV
4248 	 */
4249 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4250 	case IXGBE_82599_VMDQ_8Q_MASK:
4251 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4252 		break;
4253 	case IXGBE_82599_VMDQ_4Q_MASK:
4254 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4255 		break;
4256 	default:
4257 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4258 		break;
4259 	}
4260 
4261 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4262 
4263 	for (i = 0; i < adapter->num_vfs; i++) {
4264 		/* configure spoof checking */
4265 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4266 					  adapter->vfinfo[i].spoofchk_enabled);
4267 
4268 		/* Enable/Disable RSS query feature  */
4269 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4270 					  adapter->vfinfo[i].rss_query_enabled);
4271 	}
4272 }
4273 
4274 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4275 {
4276 	struct ixgbe_hw *hw = &adapter->hw;
4277 	struct net_device *netdev = adapter->netdev;
4278 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4279 	struct ixgbe_ring *rx_ring;
4280 	int i;
4281 	u32 mhadd, hlreg0;
4282 
4283 #ifdef IXGBE_FCOE
4284 	/* adjust max frame to be able to do baby jumbo for FCoE */
4285 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4286 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4287 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4288 
4289 #endif /* IXGBE_FCOE */
4290 
4291 	/* adjust max frame to be at least the size of a standard frame */
4292 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4293 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4294 
4295 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4296 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4297 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
4298 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4299 
4300 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4301 	}
4302 
4303 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4304 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4305 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4306 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4307 
4308 	/*
4309 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4310 	 * the Base and Length of the Rx Descriptor Ring
4311 	 */
4312 	for (i = 0; i < adapter->num_rx_queues; i++) {
4313 		rx_ring = adapter->rx_ring[i];
4314 
4315 		clear_ring_rsc_enabled(rx_ring);
4316 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4317 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4318 
4319 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4320 			set_ring_rsc_enabled(rx_ring);
4321 
4322 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4323 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4324 
4325 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4326 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4327 			continue;
4328 
4329 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4330 
4331 #if (PAGE_SIZE < 8192)
4332 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4333 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4334 
4335 		if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4336 		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4337 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4338 #endif
4339 	}
4340 }
4341 
4342 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4343 {
4344 	struct ixgbe_hw *hw = &adapter->hw;
4345 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4346 
4347 	switch (hw->mac.type) {
4348 	case ixgbe_mac_82598EB:
4349 		/*
4350 		 * For VMDq support of different descriptor types or
4351 		 * buffer sizes through the use of multiple SRRCTL
4352 		 * registers, RDRXCTL.MVMEN must be set to 1
4353 		 *
4354 		 * also, the manual doesn't mention it clearly but DCA hints
4355 		 * will only use queue 0's tags unless this bit is set.  Side
4356 		 * effects of setting this bit are only that SRRCTL must be
4357 		 * fully programmed [0..15]
4358 		 */
4359 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4360 		break;
4361 	case ixgbe_mac_X550:
4362 	case ixgbe_mac_X550EM_x:
4363 	case ixgbe_mac_x550em_a:
4364 		if (adapter->num_vfs)
4365 			rdrxctl |= IXGBE_RDRXCTL_PSP;
4366 		/* fall through */
4367 	case ixgbe_mac_82599EB:
4368 	case ixgbe_mac_X540:
4369 		/* Disable RSC for ACK packets */
4370 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4371 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4372 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4373 		/* hardware requires some bits to be set by default */
4374 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4375 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4376 		break;
4377 	default:
4378 		/* We should do nothing since we don't know this hardware */
4379 		return;
4380 	}
4381 
4382 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4383 }
4384 
4385 /**
4386  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4387  * @adapter: board private structure
4388  *
4389  * Configure the Rx unit of the MAC after a reset.
4390  **/
4391 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4392 {
4393 	struct ixgbe_hw *hw = &adapter->hw;
4394 	int i;
4395 	u32 rxctrl, rfctl;
4396 
4397 	/* disable receives while setting up the descriptors */
4398 	hw->mac.ops.disable_rx(hw);
4399 
4400 	ixgbe_setup_psrtype(adapter);
4401 	ixgbe_setup_rdrxctl(adapter);
4402 
4403 	/* RSC Setup */
4404 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4405 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4406 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4407 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4408 
4409 	/* disable NFS filtering */
4410 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4411 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4412 
4413 	/* Program registers for the distribution of queues */
4414 	ixgbe_setup_mrqc(adapter);
4415 
4416 	/* set_rx_buffer_len must be called before ring initialization */
4417 	ixgbe_set_rx_buffer_len(adapter);
4418 
4419 	/*
4420 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4421 	 * the Base and Length of the Rx Descriptor Ring
4422 	 */
4423 	for (i = 0; i < adapter->num_rx_queues; i++)
4424 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4425 
4426 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4427 	/* disable drop enable for 82598 parts */
4428 	if (hw->mac.type == ixgbe_mac_82598EB)
4429 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4430 
4431 	/* enable all receives */
4432 	rxctrl |= IXGBE_RXCTRL_RXEN;
4433 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4434 }
4435 
4436 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4437 				 __be16 proto, u16 vid)
4438 {
4439 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4440 	struct ixgbe_hw *hw = &adapter->hw;
4441 
4442 	/* add VID to filter table */
4443 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4444 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4445 
4446 	set_bit(vid, adapter->active_vlans);
4447 
4448 	return 0;
4449 }
4450 
4451 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4452 {
4453 	u32 vlvf;
4454 	int idx;
4455 
4456 	/* short cut the special case */
4457 	if (vlan == 0)
4458 		return 0;
4459 
4460 	/* Search for the vlan id in the VLVF entries */
4461 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4462 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4463 		if ((vlvf & VLAN_VID_MASK) == vlan)
4464 			break;
4465 	}
4466 
4467 	return idx;
4468 }
4469 
4470 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4471 {
4472 	struct ixgbe_hw *hw = &adapter->hw;
4473 	u32 bits, word;
4474 	int idx;
4475 
4476 	idx = ixgbe_find_vlvf_entry(hw, vid);
4477 	if (!idx)
4478 		return;
4479 
4480 	/* See if any other pools are set for this VLAN filter
4481 	 * entry other than the PF.
4482 	 */
4483 	word = idx * 2 + (VMDQ_P(0) / 32);
4484 	bits = ~BIT(VMDQ_P(0) % 32);
4485 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4486 
4487 	/* Disable the filter so this falls into the default pool. */
4488 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4489 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4490 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4491 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4492 	}
4493 }
4494 
4495 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4496 				  __be16 proto, u16 vid)
4497 {
4498 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4499 	struct ixgbe_hw *hw = &adapter->hw;
4500 
4501 	/* remove VID from filter table */
4502 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4503 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4504 
4505 	clear_bit(vid, adapter->active_vlans);
4506 
4507 	return 0;
4508 }
4509 
4510 /**
4511  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4512  * @adapter: driver data
4513  */
4514 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4515 {
4516 	struct ixgbe_hw *hw = &adapter->hw;
4517 	u32 vlnctrl;
4518 	int i, j;
4519 
4520 	switch (hw->mac.type) {
4521 	case ixgbe_mac_82598EB:
4522 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4523 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4524 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4525 		break;
4526 	case ixgbe_mac_82599EB:
4527 	case ixgbe_mac_X540:
4528 	case ixgbe_mac_X550:
4529 	case ixgbe_mac_X550EM_x:
4530 	case ixgbe_mac_x550em_a:
4531 		for (i = 0; i < adapter->num_rx_queues; i++) {
4532 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4533 
4534 			if (!netif_is_ixgbe(ring->netdev))
4535 				continue;
4536 
4537 			j = ring->reg_idx;
4538 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4539 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4540 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4541 		}
4542 		break;
4543 	default:
4544 		break;
4545 	}
4546 }
4547 
4548 /**
4549  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4550  * @adapter: driver data
4551  */
4552 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4553 {
4554 	struct ixgbe_hw *hw = &adapter->hw;
4555 	u32 vlnctrl;
4556 	int i, j;
4557 
4558 	switch (hw->mac.type) {
4559 	case ixgbe_mac_82598EB:
4560 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4561 		vlnctrl |= IXGBE_VLNCTRL_VME;
4562 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4563 		break;
4564 	case ixgbe_mac_82599EB:
4565 	case ixgbe_mac_X540:
4566 	case ixgbe_mac_X550:
4567 	case ixgbe_mac_X550EM_x:
4568 	case ixgbe_mac_x550em_a:
4569 		for (i = 0; i < adapter->num_rx_queues; i++) {
4570 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4571 
4572 			if (!netif_is_ixgbe(ring->netdev))
4573 				continue;
4574 
4575 			j = ring->reg_idx;
4576 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4577 			vlnctrl |= IXGBE_RXDCTL_VME;
4578 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4579 		}
4580 		break;
4581 	default:
4582 		break;
4583 	}
4584 }
4585 
4586 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4587 {
4588 	struct ixgbe_hw *hw = &adapter->hw;
4589 	u32 vlnctrl, i;
4590 
4591 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4592 
4593 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4594 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4595 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4596 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4597 	} else {
4598 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4599 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4600 		return;
4601 	}
4602 
4603 	/* Nothing to do for 82598 */
4604 	if (hw->mac.type == ixgbe_mac_82598EB)
4605 		return;
4606 
4607 	/* We are already in VLAN promisc, nothing to do */
4608 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4609 		return;
4610 
4611 	/* Set flag so we don't redo unnecessary work */
4612 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4613 
4614 	/* Add PF to all active pools */
4615 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4616 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4617 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4618 
4619 		vlvfb |= BIT(VMDQ_P(0) % 32);
4620 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4621 	}
4622 
4623 	/* Set all bits in the VLAN filter table array */
4624 	for (i = hw->mac.vft_size; i--;)
4625 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4626 }
4627 
4628 #define VFTA_BLOCK_SIZE 8
4629 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4630 {
4631 	struct ixgbe_hw *hw = &adapter->hw;
4632 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4633 	u32 vid_start = vfta_offset * 32;
4634 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4635 	u32 i, vid, word, bits;
4636 
4637 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4638 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4639 
4640 		/* pull VLAN ID from VLVF */
4641 		vid = vlvf & VLAN_VID_MASK;
4642 
4643 		/* only concern outselves with a certain range */
4644 		if (vid < vid_start || vid >= vid_end)
4645 			continue;
4646 
4647 		if (vlvf) {
4648 			/* record VLAN ID in VFTA */
4649 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4650 
4651 			/* if PF is part of this then continue */
4652 			if (test_bit(vid, adapter->active_vlans))
4653 				continue;
4654 		}
4655 
4656 		/* remove PF from the pool */
4657 		word = i * 2 + VMDQ_P(0) / 32;
4658 		bits = ~BIT(VMDQ_P(0) % 32);
4659 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4660 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4661 	}
4662 
4663 	/* extract values from active_vlans and write back to VFTA */
4664 	for (i = VFTA_BLOCK_SIZE; i--;) {
4665 		vid = (vfta_offset + i) * 32;
4666 		word = vid / BITS_PER_LONG;
4667 		bits = vid % BITS_PER_LONG;
4668 
4669 		vfta[i] |= adapter->active_vlans[word] >> bits;
4670 
4671 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4672 	}
4673 }
4674 
4675 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4676 {
4677 	struct ixgbe_hw *hw = &adapter->hw;
4678 	u32 vlnctrl, i;
4679 
4680 	/* Set VLAN filtering to enabled */
4681 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4682 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4683 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4684 
4685 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4686 	    hw->mac.type == ixgbe_mac_82598EB)
4687 		return;
4688 
4689 	/* We are not in VLAN promisc, nothing to do */
4690 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4691 		return;
4692 
4693 	/* Set flag so we don't redo unnecessary work */
4694 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4695 
4696 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4697 		ixgbe_scrub_vfta(adapter, i);
4698 }
4699 
4700 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4701 {
4702 	u16 vid = 1;
4703 
4704 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4705 
4706 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4707 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4708 }
4709 
4710 /**
4711  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4712  * @netdev: network interface device structure
4713  *
4714  * Writes multicast address list to the MTA hash table.
4715  * Returns: -ENOMEM on failure
4716  *                0 on no addresses written
4717  *                X on writing X addresses to MTA
4718  **/
4719 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4720 {
4721 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4722 	struct ixgbe_hw *hw = &adapter->hw;
4723 
4724 	if (!netif_running(netdev))
4725 		return 0;
4726 
4727 	if (hw->mac.ops.update_mc_addr_list)
4728 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4729 	else
4730 		return -ENOMEM;
4731 
4732 #ifdef CONFIG_PCI_IOV
4733 	ixgbe_restore_vf_multicasts(adapter);
4734 #endif
4735 
4736 	return netdev_mc_count(netdev);
4737 }
4738 
4739 #ifdef CONFIG_PCI_IOV
4740 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4741 {
4742 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4743 	struct ixgbe_hw *hw = &adapter->hw;
4744 	int i;
4745 
4746 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4747 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4748 
4749 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4750 			hw->mac.ops.set_rar(hw, i,
4751 					    mac_table->addr,
4752 					    mac_table->pool,
4753 					    IXGBE_RAH_AV);
4754 		else
4755 			hw->mac.ops.clear_rar(hw, i);
4756 	}
4757 }
4758 
4759 #endif
4760 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4761 {
4762 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4763 	struct ixgbe_hw *hw = &adapter->hw;
4764 	int i;
4765 
4766 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4767 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4768 			continue;
4769 
4770 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4771 
4772 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4773 			hw->mac.ops.set_rar(hw, i,
4774 					    mac_table->addr,
4775 					    mac_table->pool,
4776 					    IXGBE_RAH_AV);
4777 		else
4778 			hw->mac.ops.clear_rar(hw, i);
4779 	}
4780 }
4781 
4782 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4783 {
4784 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4785 	struct ixgbe_hw *hw = &adapter->hw;
4786 	int i;
4787 
4788 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4789 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4790 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4791 	}
4792 
4793 	ixgbe_sync_mac_table(adapter);
4794 }
4795 
4796 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4797 {
4798 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4799 	struct ixgbe_hw *hw = &adapter->hw;
4800 	int i, count = 0;
4801 
4802 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4803 		/* do not count default RAR as available */
4804 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4805 			continue;
4806 
4807 		/* only count unused and addresses that belong to us */
4808 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4809 			if (mac_table->pool != pool)
4810 				continue;
4811 		}
4812 
4813 		count++;
4814 	}
4815 
4816 	return count;
4817 }
4818 
4819 /* this function destroys the first RAR entry */
4820 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4821 {
4822 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4823 	struct ixgbe_hw *hw = &adapter->hw;
4824 
4825 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4826 	mac_table->pool = VMDQ_P(0);
4827 
4828 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4829 
4830 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4831 			    IXGBE_RAH_AV);
4832 }
4833 
4834 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4835 			 const u8 *addr, u16 pool)
4836 {
4837 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4838 	struct ixgbe_hw *hw = &adapter->hw;
4839 	int i;
4840 
4841 	if (is_zero_ether_addr(addr))
4842 		return -EINVAL;
4843 
4844 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4845 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4846 			continue;
4847 
4848 		ether_addr_copy(mac_table->addr, addr);
4849 		mac_table->pool = pool;
4850 
4851 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4852 				    IXGBE_MAC_STATE_IN_USE;
4853 
4854 		ixgbe_sync_mac_table(adapter);
4855 
4856 		return i;
4857 	}
4858 
4859 	return -ENOMEM;
4860 }
4861 
4862 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4863 			 const u8 *addr, u16 pool)
4864 {
4865 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4866 	struct ixgbe_hw *hw = &adapter->hw;
4867 	int i;
4868 
4869 	if (is_zero_ether_addr(addr))
4870 		return -EINVAL;
4871 
4872 	/* search table for addr, if found clear IN_USE flag and sync */
4873 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4874 		/* we can only delete an entry if it is in use */
4875 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4876 			continue;
4877 		/* we only care about entries that belong to the given pool */
4878 		if (mac_table->pool != pool)
4879 			continue;
4880 		/* we only care about a specific MAC address */
4881 		if (!ether_addr_equal(addr, mac_table->addr))
4882 			continue;
4883 
4884 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4885 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4886 
4887 		ixgbe_sync_mac_table(adapter);
4888 
4889 		return 0;
4890 	}
4891 
4892 	return -ENOMEM;
4893 }
4894 
4895 /**
4896  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4897  * @netdev: network interface device structure
4898  * @vfn: pool to associate with unicast addresses
4899  *
4900  * Writes unicast address list to the RAR table.
4901  * Returns: -ENOMEM on failure/insufficient address space
4902  *                0 on no addresses written
4903  *                X on writing X addresses to the RAR table
4904  **/
4905 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4906 {
4907 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4908 	int count = 0;
4909 
4910 	/* return ENOMEM indicating insufficient memory for addresses */
4911 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter, vfn))
4912 		return -ENOMEM;
4913 
4914 	if (!netdev_uc_empty(netdev)) {
4915 		struct netdev_hw_addr *ha;
4916 		netdev_for_each_uc_addr(ha, netdev) {
4917 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4918 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4919 			count++;
4920 		}
4921 	}
4922 	return count;
4923 }
4924 
4925 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4926 {
4927 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4928 	int ret;
4929 
4930 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4931 
4932 	return min_t(int, ret, 0);
4933 }
4934 
4935 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4936 {
4937 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4938 
4939 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4940 
4941 	return 0;
4942 }
4943 
4944 /**
4945  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4946  * @netdev: network interface device structure
4947  *
4948  * The set_rx_method entry point is called whenever the unicast/multicast
4949  * address list or the network interface flags are updated.  This routine is
4950  * responsible for configuring the hardware for proper unicast, multicast and
4951  * promiscuous mode.
4952  **/
4953 void ixgbe_set_rx_mode(struct net_device *netdev)
4954 {
4955 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4956 	struct ixgbe_hw *hw = &adapter->hw;
4957 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4958 	netdev_features_t features = netdev->features;
4959 	int count;
4960 
4961 	/* Check for Promiscuous and All Multicast modes */
4962 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4963 
4964 	/* set all bits that we expect to always be set */
4965 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4966 	fctrl |= IXGBE_FCTRL_BAM;
4967 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4968 	fctrl |= IXGBE_FCTRL_PMCF;
4969 
4970 	/* clear the bits we are changing the status of */
4971 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4972 	if (netdev->flags & IFF_PROMISC) {
4973 		hw->addr_ctrl.user_set_promisc = true;
4974 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4975 		vmolr |= IXGBE_VMOLR_MPE;
4976 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4977 	} else {
4978 		if (netdev->flags & IFF_ALLMULTI) {
4979 			fctrl |= IXGBE_FCTRL_MPE;
4980 			vmolr |= IXGBE_VMOLR_MPE;
4981 		}
4982 		hw->addr_ctrl.user_set_promisc = false;
4983 	}
4984 
4985 	/*
4986 	 * Write addresses to available RAR registers, if there is not
4987 	 * sufficient space to store all the addresses then enable
4988 	 * unicast promiscuous mode
4989 	 */
4990 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4991 		fctrl |= IXGBE_FCTRL_UPE;
4992 		vmolr |= IXGBE_VMOLR_ROPE;
4993 	}
4994 
4995 	/* Write addresses to the MTA, if the attempt fails
4996 	 * then we should just turn on promiscuous mode so
4997 	 * that we can at least receive multicast traffic
4998 	 */
4999 	count = ixgbe_write_mc_addr_list(netdev);
5000 	if (count < 0) {
5001 		fctrl |= IXGBE_FCTRL_MPE;
5002 		vmolr |= IXGBE_VMOLR_MPE;
5003 	} else if (count) {
5004 		vmolr |= IXGBE_VMOLR_ROMPE;
5005 	}
5006 
5007 	if (hw->mac.type != ixgbe_mac_82598EB) {
5008 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
5009 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
5010 			   IXGBE_VMOLR_ROPE);
5011 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
5012 	}
5013 
5014 	/* This is useful for sniffing bad packets. */
5015 	if (features & NETIF_F_RXALL) {
5016 		/* UPE and MPE will be handled by normal PROMISC logic
5017 		 * in e1000e_set_rx_mode */
5018 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
5019 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
5020 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
5021 
5022 		fctrl &= ~(IXGBE_FCTRL_DPF);
5023 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
5024 	}
5025 
5026 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5027 
5028 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
5029 		ixgbe_vlan_strip_enable(adapter);
5030 	else
5031 		ixgbe_vlan_strip_disable(adapter);
5032 
5033 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
5034 		ixgbe_vlan_promisc_disable(adapter);
5035 	else
5036 		ixgbe_vlan_promisc_enable(adapter);
5037 }
5038 
5039 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
5040 {
5041 	int q_idx;
5042 
5043 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5044 		napi_enable(&adapter->q_vector[q_idx]->napi);
5045 }
5046 
5047 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
5048 {
5049 	int q_idx;
5050 
5051 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5052 		napi_disable(&adapter->q_vector[q_idx]->napi);
5053 }
5054 
5055 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
5056 {
5057 	struct ixgbe_hw *hw = &adapter->hw;
5058 	u32 vxlanctrl;
5059 
5060 	if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
5061 				IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
5062 		return;
5063 
5064 	vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) & ~mask;
5065 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
5066 
5067 	if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
5068 		adapter->vxlan_port = 0;
5069 
5070 	if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
5071 		adapter->geneve_port = 0;
5072 }
5073 
5074 #ifdef CONFIG_IXGBE_DCB
5075 /**
5076  * ixgbe_configure_dcb - Configure DCB hardware
5077  * @adapter: ixgbe adapter struct
5078  *
5079  * This is called by the driver on open to configure the DCB hardware.
5080  * This is also called by the gennetlink interface when reconfiguring
5081  * the DCB state.
5082  */
5083 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5084 {
5085 	struct ixgbe_hw *hw = &adapter->hw;
5086 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5087 
5088 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5089 		if (hw->mac.type == ixgbe_mac_82598EB)
5090 			netif_set_gso_max_size(adapter->netdev, 65536);
5091 		return;
5092 	}
5093 
5094 	if (hw->mac.type == ixgbe_mac_82598EB)
5095 		netif_set_gso_max_size(adapter->netdev, 32768);
5096 
5097 #ifdef IXGBE_FCOE
5098 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
5099 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5100 #endif
5101 
5102 	/* reconfigure the hardware */
5103 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5104 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5105 						DCB_TX_CONFIG);
5106 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5107 						DCB_RX_CONFIG);
5108 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5109 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5110 		ixgbe_dcb_hw_ets(&adapter->hw,
5111 				 adapter->ixgbe_ieee_ets,
5112 				 max_frame);
5113 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
5114 					adapter->ixgbe_ieee_pfc->pfc_en,
5115 					adapter->ixgbe_ieee_ets->prio_tc);
5116 	}
5117 
5118 	/* Enable RSS Hash per TC */
5119 	if (hw->mac.type != ixgbe_mac_82598EB) {
5120 		u32 msb = 0;
5121 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5122 
5123 		while (rss_i) {
5124 			msb++;
5125 			rss_i >>= 1;
5126 		}
5127 
5128 		/* write msb to all 8 TCs in one write */
5129 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5130 	}
5131 }
5132 #endif
5133 
5134 /* Additional bittime to account for IXGBE framing */
5135 #define IXGBE_ETH_FRAMING 20
5136 
5137 /**
5138  * ixgbe_hpbthresh - calculate high water mark for flow control
5139  *
5140  * @adapter: board private structure to calculate for
5141  * @pb: packet buffer to calculate
5142  */
5143 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5144 {
5145 	struct ixgbe_hw *hw = &adapter->hw;
5146 	struct net_device *dev = adapter->netdev;
5147 	int link, tc, kb, marker;
5148 	u32 dv_id, rx_pba;
5149 
5150 	/* Calculate max LAN frame size */
5151 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5152 
5153 #ifdef IXGBE_FCOE
5154 	/* FCoE traffic class uses FCOE jumbo frames */
5155 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5156 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5157 	    (pb == ixgbe_fcoe_get_tc(adapter)))
5158 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5159 #endif
5160 
5161 	/* Calculate delay value for device */
5162 	switch (hw->mac.type) {
5163 	case ixgbe_mac_X540:
5164 	case ixgbe_mac_X550:
5165 	case ixgbe_mac_X550EM_x:
5166 	case ixgbe_mac_x550em_a:
5167 		dv_id = IXGBE_DV_X540(link, tc);
5168 		break;
5169 	default:
5170 		dv_id = IXGBE_DV(link, tc);
5171 		break;
5172 	}
5173 
5174 	/* Loopback switch introduces additional latency */
5175 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5176 		dv_id += IXGBE_B2BT(tc);
5177 
5178 	/* Delay value is calculated in bit times convert to KB */
5179 	kb = IXGBE_BT2KB(dv_id);
5180 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5181 
5182 	marker = rx_pba - kb;
5183 
5184 	/* It is possible that the packet buffer is not large enough
5185 	 * to provide required headroom. In this case throw an error
5186 	 * to user and a do the best we can.
5187 	 */
5188 	if (marker < 0) {
5189 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
5190 			    "headroom to support flow control."
5191 			    "Decrease MTU or number of traffic classes\n", pb);
5192 		marker = tc + 1;
5193 	}
5194 
5195 	return marker;
5196 }
5197 
5198 /**
5199  * ixgbe_lpbthresh - calculate low water mark for for flow control
5200  *
5201  * @adapter: board private structure to calculate for
5202  * @pb: packet buffer to calculate
5203  */
5204 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5205 {
5206 	struct ixgbe_hw *hw = &adapter->hw;
5207 	struct net_device *dev = adapter->netdev;
5208 	int tc;
5209 	u32 dv_id;
5210 
5211 	/* Calculate max LAN frame size */
5212 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5213 
5214 #ifdef IXGBE_FCOE
5215 	/* FCoE traffic class uses FCOE jumbo frames */
5216 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5217 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5218 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5219 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5220 #endif
5221 
5222 	/* Calculate delay value for device */
5223 	switch (hw->mac.type) {
5224 	case ixgbe_mac_X540:
5225 	case ixgbe_mac_X550:
5226 	case ixgbe_mac_X550EM_x:
5227 	case ixgbe_mac_x550em_a:
5228 		dv_id = IXGBE_LOW_DV_X540(tc);
5229 		break;
5230 	default:
5231 		dv_id = IXGBE_LOW_DV(tc);
5232 		break;
5233 	}
5234 
5235 	/* Delay value is calculated in bit times convert to KB */
5236 	return IXGBE_BT2KB(dv_id);
5237 }
5238 
5239 /*
5240  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5241  */
5242 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5243 {
5244 	struct ixgbe_hw *hw = &adapter->hw;
5245 	int num_tc = adapter->hw_tcs;
5246 	int i;
5247 
5248 	if (!num_tc)
5249 		num_tc = 1;
5250 
5251 	for (i = 0; i < num_tc; i++) {
5252 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5253 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5254 
5255 		/* Low water marks must not be larger than high water marks */
5256 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
5257 			hw->fc.low_water[i] = 0;
5258 	}
5259 
5260 	for (; i < MAX_TRAFFIC_CLASS; i++)
5261 		hw->fc.high_water[i] = 0;
5262 }
5263 
5264 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5265 {
5266 	struct ixgbe_hw *hw = &adapter->hw;
5267 	int hdrm;
5268 	u8 tc = adapter->hw_tcs;
5269 
5270 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5271 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5272 		hdrm = 32 << adapter->fdir_pballoc;
5273 	else
5274 		hdrm = 0;
5275 
5276 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5277 	ixgbe_pbthresh_setup(adapter);
5278 }
5279 
5280 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5281 {
5282 	struct ixgbe_hw *hw = &adapter->hw;
5283 	struct hlist_node *node2;
5284 	struct ixgbe_fdir_filter *filter;
5285 
5286 	spin_lock(&adapter->fdir_perfect_lock);
5287 
5288 	if (!hlist_empty(&adapter->fdir_filter_list))
5289 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5290 
5291 	hlist_for_each_entry_safe(filter, node2,
5292 				  &adapter->fdir_filter_list, fdir_node) {
5293 		ixgbe_fdir_write_perfect_filter_82599(hw,
5294 				&filter->filter,
5295 				filter->sw_idx,
5296 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
5297 				IXGBE_FDIR_DROP_QUEUE :
5298 				adapter->rx_ring[filter->action]->reg_idx);
5299 	}
5300 
5301 	spin_unlock(&adapter->fdir_perfect_lock);
5302 }
5303 
5304 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
5305 				      struct ixgbe_adapter *adapter)
5306 {
5307 	struct ixgbe_hw *hw = &adapter->hw;
5308 	u32 vmolr;
5309 
5310 	/* No unicast promiscuous support for VMDQ devices. */
5311 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
5312 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
5313 
5314 	/* clear the affected bit */
5315 	vmolr &= ~IXGBE_VMOLR_MPE;
5316 
5317 	if (dev->flags & IFF_ALLMULTI) {
5318 		vmolr |= IXGBE_VMOLR_MPE;
5319 	} else {
5320 		vmolr |= IXGBE_VMOLR_ROMPE;
5321 		hw->mac.ops.update_mc_addr_list(hw, dev);
5322 	}
5323 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
5324 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
5325 }
5326 
5327 /**
5328  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5329  * @rx_ring: ring to free buffers from
5330  **/
5331 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5332 {
5333 	u16 i = rx_ring->next_to_clean;
5334 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5335 
5336 	/* Free all the Rx ring sk_buffs */
5337 	while (i != rx_ring->next_to_alloc) {
5338 		if (rx_buffer->skb) {
5339 			struct sk_buff *skb = rx_buffer->skb;
5340 			if (IXGBE_CB(skb)->page_released)
5341 				dma_unmap_page_attrs(rx_ring->dev,
5342 						     IXGBE_CB(skb)->dma,
5343 						     ixgbe_rx_pg_size(rx_ring),
5344 						     DMA_FROM_DEVICE,
5345 						     IXGBE_RX_DMA_ATTR);
5346 			dev_kfree_skb(skb);
5347 		}
5348 
5349 		/* Invalidate cache lines that may have been written to by
5350 		 * device so that we avoid corrupting memory.
5351 		 */
5352 		dma_sync_single_range_for_cpu(rx_ring->dev,
5353 					      rx_buffer->dma,
5354 					      rx_buffer->page_offset,
5355 					      ixgbe_rx_bufsz(rx_ring),
5356 					      DMA_FROM_DEVICE);
5357 
5358 		/* free resources associated with mapping */
5359 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5360 				     ixgbe_rx_pg_size(rx_ring),
5361 				     DMA_FROM_DEVICE,
5362 				     IXGBE_RX_DMA_ATTR);
5363 		__page_frag_cache_drain(rx_buffer->page,
5364 					rx_buffer->pagecnt_bias);
5365 
5366 		i++;
5367 		rx_buffer++;
5368 		if (i == rx_ring->count) {
5369 			i = 0;
5370 			rx_buffer = rx_ring->rx_buffer_info;
5371 		}
5372 	}
5373 
5374 	rx_ring->next_to_alloc = 0;
5375 	rx_ring->next_to_clean = 0;
5376 	rx_ring->next_to_use = 0;
5377 }
5378 
5379 static int ixgbe_fwd_ring_up(struct net_device *vdev,
5380 			     struct ixgbe_fwd_adapter *accel)
5381 {
5382 	struct ixgbe_adapter *adapter = accel->real_adapter;
5383 	int i, baseq, err;
5384 
5385 	if (!test_bit(accel->pool, adapter->fwd_bitmask))
5386 		return 0;
5387 
5388 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5389 	netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5390 		   accel->pool, adapter->num_rx_pools,
5391 		   baseq, baseq + adapter->num_rx_queues_per_pool);
5392 
5393 	accel->netdev = vdev;
5394 	accel->rx_base_queue = baseq;
5395 	accel->tx_base_queue = baseq;
5396 
5397 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5398 		adapter->rx_ring[baseq + i]->netdev = vdev;
5399 
5400 	/* Guarantee all rings are updated before we update the
5401 	 * MAC address filter.
5402 	 */
5403 	wmb();
5404 
5405 	/* ixgbe_add_mac_filter will return an index if it succeeds, so we
5406 	 * need to only treat it as an error value if it is negative.
5407 	 */
5408 	err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5409 				   VMDQ_P(accel->pool));
5410 	if (err >= 0) {
5411 		ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
5412 		return 0;
5413 	}
5414 
5415 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5416 		adapter->rx_ring[baseq + i]->netdev = NULL;
5417 
5418 	return err;
5419 }
5420 
5421 static int ixgbe_upper_dev_walk(struct net_device *upper, void *data)
5422 {
5423 	if (netif_is_macvlan(upper)) {
5424 		struct macvlan_dev *dfwd = netdev_priv(upper);
5425 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
5426 
5427 		if (dfwd->fwd_priv)
5428 			ixgbe_fwd_ring_up(upper, vadapter);
5429 	}
5430 
5431 	return 0;
5432 }
5433 
5434 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5435 {
5436 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5437 				      ixgbe_upper_dev_walk, NULL);
5438 }
5439 
5440 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5441 {
5442 	struct ixgbe_hw *hw = &adapter->hw;
5443 
5444 	ixgbe_configure_pb(adapter);
5445 #ifdef CONFIG_IXGBE_DCB
5446 	ixgbe_configure_dcb(adapter);
5447 #endif
5448 	/*
5449 	 * We must restore virtualization before VLANs or else
5450 	 * the VLVF registers will not be populated
5451 	 */
5452 	ixgbe_configure_virtualization(adapter);
5453 
5454 	ixgbe_set_rx_mode(adapter->netdev);
5455 	ixgbe_restore_vlan(adapter);
5456 	ixgbe_ipsec_restore(adapter);
5457 
5458 	switch (hw->mac.type) {
5459 	case ixgbe_mac_82599EB:
5460 	case ixgbe_mac_X540:
5461 		hw->mac.ops.disable_rx_buff(hw);
5462 		break;
5463 	default:
5464 		break;
5465 	}
5466 
5467 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5468 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5469 						adapter->fdir_pballoc);
5470 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5471 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5472 					      adapter->fdir_pballoc);
5473 		ixgbe_fdir_filter_restore(adapter);
5474 	}
5475 
5476 	switch (hw->mac.type) {
5477 	case ixgbe_mac_82599EB:
5478 	case ixgbe_mac_X540:
5479 		hw->mac.ops.enable_rx_buff(hw);
5480 		break;
5481 	default:
5482 		break;
5483 	}
5484 
5485 #ifdef CONFIG_IXGBE_DCA
5486 	/* configure DCA */
5487 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5488 		ixgbe_setup_dca(adapter);
5489 #endif /* CONFIG_IXGBE_DCA */
5490 
5491 #ifdef IXGBE_FCOE
5492 	/* configure FCoE L2 filters, redirection table, and Rx control */
5493 	ixgbe_configure_fcoe(adapter);
5494 
5495 #endif /* IXGBE_FCOE */
5496 	ixgbe_configure_tx(adapter);
5497 	ixgbe_configure_rx(adapter);
5498 	ixgbe_configure_dfwd(adapter);
5499 }
5500 
5501 /**
5502  * ixgbe_sfp_link_config - set up SFP+ link
5503  * @adapter: pointer to private adapter struct
5504  **/
5505 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5506 {
5507 	/*
5508 	 * We are assuming the worst case scenario here, and that
5509 	 * is that an SFP was inserted/removed after the reset
5510 	 * but before SFP detection was enabled.  As such the best
5511 	 * solution is to just start searching as soon as we start
5512 	 */
5513 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5514 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5515 
5516 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5517 	adapter->sfp_poll_time = 0;
5518 }
5519 
5520 /**
5521  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5522  * @hw: pointer to private hardware struct
5523  *
5524  * Returns 0 on success, negative on failure
5525  **/
5526 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5527 {
5528 	u32 speed;
5529 	bool autoneg, link_up = false;
5530 	int ret = IXGBE_ERR_LINK_SETUP;
5531 
5532 	if (hw->mac.ops.check_link)
5533 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5534 
5535 	if (ret)
5536 		return ret;
5537 
5538 	speed = hw->phy.autoneg_advertised;
5539 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5540 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5541 							&autoneg);
5542 	if (ret)
5543 		return ret;
5544 
5545 	if (hw->mac.ops.setup_link)
5546 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5547 
5548 	return ret;
5549 }
5550 
5551 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5552 {
5553 	struct ixgbe_hw *hw = &adapter->hw;
5554 	u32 gpie = 0;
5555 
5556 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5557 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5558 		       IXGBE_GPIE_OCD;
5559 		gpie |= IXGBE_GPIE_EIAME;
5560 		/*
5561 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5562 		 * this saves a register write for every interrupt
5563 		 */
5564 		switch (hw->mac.type) {
5565 		case ixgbe_mac_82598EB:
5566 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5567 			break;
5568 		case ixgbe_mac_82599EB:
5569 		case ixgbe_mac_X540:
5570 		case ixgbe_mac_X550:
5571 		case ixgbe_mac_X550EM_x:
5572 		case ixgbe_mac_x550em_a:
5573 		default:
5574 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5575 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5576 			break;
5577 		}
5578 	} else {
5579 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5580 		 * specifically only auto mask tx and rx interrupts */
5581 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5582 	}
5583 
5584 	/* XXX: to interrupt immediately for EICS writes, enable this */
5585 	/* gpie |= IXGBE_GPIE_EIMEN; */
5586 
5587 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5588 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5589 
5590 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5591 		case IXGBE_82599_VMDQ_8Q_MASK:
5592 			gpie |= IXGBE_GPIE_VTMODE_16;
5593 			break;
5594 		case IXGBE_82599_VMDQ_4Q_MASK:
5595 			gpie |= IXGBE_GPIE_VTMODE_32;
5596 			break;
5597 		default:
5598 			gpie |= IXGBE_GPIE_VTMODE_64;
5599 			break;
5600 		}
5601 	}
5602 
5603 	/* Enable Thermal over heat sensor interrupt */
5604 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5605 		switch (adapter->hw.mac.type) {
5606 		case ixgbe_mac_82599EB:
5607 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5608 			break;
5609 		default:
5610 			break;
5611 		}
5612 	}
5613 
5614 	/* Enable fan failure interrupt */
5615 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5616 		gpie |= IXGBE_SDP1_GPIEN(hw);
5617 
5618 	switch (hw->mac.type) {
5619 	case ixgbe_mac_82599EB:
5620 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5621 		break;
5622 	case ixgbe_mac_X550EM_x:
5623 	case ixgbe_mac_x550em_a:
5624 		gpie |= IXGBE_SDP0_GPIEN_X540;
5625 		break;
5626 	default:
5627 		break;
5628 	}
5629 
5630 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5631 }
5632 
5633 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5634 {
5635 	struct ixgbe_hw *hw = &adapter->hw;
5636 	int err;
5637 	u32 ctrl_ext;
5638 
5639 	ixgbe_get_hw_control(adapter);
5640 	ixgbe_setup_gpie(adapter);
5641 
5642 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5643 		ixgbe_configure_msix(adapter);
5644 	else
5645 		ixgbe_configure_msi_and_legacy(adapter);
5646 
5647 	/* enable the optics for 82599 SFP+ fiber */
5648 	if (hw->mac.ops.enable_tx_laser)
5649 		hw->mac.ops.enable_tx_laser(hw);
5650 
5651 	if (hw->phy.ops.set_phy_power)
5652 		hw->phy.ops.set_phy_power(hw, true);
5653 
5654 	smp_mb__before_atomic();
5655 	clear_bit(__IXGBE_DOWN, &adapter->state);
5656 	ixgbe_napi_enable_all(adapter);
5657 
5658 	if (ixgbe_is_sfp(hw)) {
5659 		ixgbe_sfp_link_config(adapter);
5660 	} else {
5661 		err = ixgbe_non_sfp_link_config(hw);
5662 		if (err)
5663 			e_err(probe, "link_config FAILED %d\n", err);
5664 	}
5665 
5666 	/* clear any pending interrupts, may auto mask */
5667 	IXGBE_READ_REG(hw, IXGBE_EICR);
5668 	ixgbe_irq_enable(adapter, true, true);
5669 
5670 	/*
5671 	 * If this adapter has a fan, check to see if we had a failure
5672 	 * before we enabled the interrupt.
5673 	 */
5674 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5675 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5676 		if (esdp & IXGBE_ESDP_SDP1)
5677 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5678 	}
5679 
5680 	/* bring the link up in the watchdog, this could race with our first
5681 	 * link up interrupt but shouldn't be a problem */
5682 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5683 	adapter->link_check_timeout = jiffies;
5684 	mod_timer(&adapter->service_timer, jiffies);
5685 
5686 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5687 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5688 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5689 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5690 }
5691 
5692 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5693 {
5694 	WARN_ON(in_interrupt());
5695 	/* put off any impending NetWatchDogTimeout */
5696 	netif_trans_update(adapter->netdev);
5697 
5698 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5699 		usleep_range(1000, 2000);
5700 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5701 		ixgbe_watchdog_link_is_down(adapter);
5702 	ixgbe_down(adapter);
5703 	/*
5704 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5705 	 * back up to give the VFs time to respond to the reset.  The
5706 	 * two second wait is based upon the watchdog timer cycle in
5707 	 * the VF driver.
5708 	 */
5709 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5710 		msleep(2000);
5711 	ixgbe_up(adapter);
5712 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5713 }
5714 
5715 void ixgbe_up(struct ixgbe_adapter *adapter)
5716 {
5717 	/* hardware has been reset, we need to reload some things */
5718 	ixgbe_configure(adapter);
5719 
5720 	ixgbe_up_complete(adapter);
5721 }
5722 
5723 void ixgbe_reset(struct ixgbe_adapter *adapter)
5724 {
5725 	struct ixgbe_hw *hw = &adapter->hw;
5726 	struct net_device *netdev = adapter->netdev;
5727 	int err;
5728 
5729 	if (ixgbe_removed(hw->hw_addr))
5730 		return;
5731 	/* lock SFP init bit to prevent race conditions with the watchdog */
5732 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5733 		usleep_range(1000, 2000);
5734 
5735 	/* clear all SFP and link config related flags while holding SFP_INIT */
5736 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5737 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5738 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5739 
5740 	err = hw->mac.ops.init_hw(hw);
5741 	switch (err) {
5742 	case 0:
5743 	case IXGBE_ERR_SFP_NOT_PRESENT:
5744 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5745 		break;
5746 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5747 		e_dev_err("master disable timed out\n");
5748 		break;
5749 	case IXGBE_ERR_EEPROM_VERSION:
5750 		/* We are running on a pre-production device, log a warning */
5751 		e_dev_warn("This device is a pre-production adapter/LOM. "
5752 			   "Please be aware there may be issues associated with "
5753 			   "your hardware.  If you are experiencing problems "
5754 			   "please contact your Intel or hardware "
5755 			   "representative who provided you with this "
5756 			   "hardware.\n");
5757 		break;
5758 	default:
5759 		e_dev_err("Hardware Error: %d\n", err);
5760 	}
5761 
5762 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5763 
5764 	/* flush entries out of MAC table */
5765 	ixgbe_flush_sw_mac_table(adapter);
5766 	__dev_uc_unsync(netdev, NULL);
5767 
5768 	/* do not flush user set addresses */
5769 	ixgbe_mac_set_default_filter(adapter);
5770 
5771 	/* update SAN MAC vmdq pool selection */
5772 	if (hw->mac.san_mac_rar_index)
5773 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5774 
5775 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5776 		ixgbe_ptp_reset(adapter);
5777 
5778 	if (hw->phy.ops.set_phy_power) {
5779 		if (!netif_running(adapter->netdev) && !adapter->wol)
5780 			hw->phy.ops.set_phy_power(hw, false);
5781 		else
5782 			hw->phy.ops.set_phy_power(hw, true);
5783 	}
5784 }
5785 
5786 /**
5787  * ixgbe_clean_tx_ring - Free Tx Buffers
5788  * @tx_ring: ring to be cleaned
5789  **/
5790 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5791 {
5792 	u16 i = tx_ring->next_to_clean;
5793 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
5794 
5795 	while (i != tx_ring->next_to_use) {
5796 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
5797 
5798 		/* Free all the Tx ring sk_buffs */
5799 		if (ring_is_xdp(tx_ring))
5800 			page_frag_free(tx_buffer->data);
5801 		else
5802 			dev_kfree_skb_any(tx_buffer->skb);
5803 
5804 		/* unmap skb header data */
5805 		dma_unmap_single(tx_ring->dev,
5806 				 dma_unmap_addr(tx_buffer, dma),
5807 				 dma_unmap_len(tx_buffer, len),
5808 				 DMA_TO_DEVICE);
5809 
5810 		/* check for eop_desc to determine the end of the packet */
5811 		eop_desc = tx_buffer->next_to_watch;
5812 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
5813 
5814 		/* unmap remaining buffers */
5815 		while (tx_desc != eop_desc) {
5816 			tx_buffer++;
5817 			tx_desc++;
5818 			i++;
5819 			if (unlikely(i == tx_ring->count)) {
5820 				i = 0;
5821 				tx_buffer = tx_ring->tx_buffer_info;
5822 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
5823 			}
5824 
5825 			/* unmap any remaining paged data */
5826 			if (dma_unmap_len(tx_buffer, len))
5827 				dma_unmap_page(tx_ring->dev,
5828 					       dma_unmap_addr(tx_buffer, dma),
5829 					       dma_unmap_len(tx_buffer, len),
5830 					       DMA_TO_DEVICE);
5831 		}
5832 
5833 		/* move us one more past the eop_desc for start of next pkt */
5834 		tx_buffer++;
5835 		i++;
5836 		if (unlikely(i == tx_ring->count)) {
5837 			i = 0;
5838 			tx_buffer = tx_ring->tx_buffer_info;
5839 		}
5840 	}
5841 
5842 	/* reset BQL for queue */
5843 	if (!ring_is_xdp(tx_ring))
5844 		netdev_tx_reset_queue(txring_txq(tx_ring));
5845 
5846 	/* reset next_to_use and next_to_clean */
5847 	tx_ring->next_to_use = 0;
5848 	tx_ring->next_to_clean = 0;
5849 }
5850 
5851 /**
5852  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5853  * @adapter: board private structure
5854  **/
5855 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5856 {
5857 	int i;
5858 
5859 	for (i = 0; i < adapter->num_rx_queues; i++)
5860 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5861 }
5862 
5863 /**
5864  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5865  * @adapter: board private structure
5866  **/
5867 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5868 {
5869 	int i;
5870 
5871 	for (i = 0; i < adapter->num_tx_queues; i++)
5872 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5873 	for (i = 0; i < adapter->num_xdp_queues; i++)
5874 		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
5875 }
5876 
5877 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5878 {
5879 	struct hlist_node *node2;
5880 	struct ixgbe_fdir_filter *filter;
5881 
5882 	spin_lock(&adapter->fdir_perfect_lock);
5883 
5884 	hlist_for_each_entry_safe(filter, node2,
5885 				  &adapter->fdir_filter_list, fdir_node) {
5886 		hlist_del(&filter->fdir_node);
5887 		kfree(filter);
5888 	}
5889 	adapter->fdir_filter_count = 0;
5890 
5891 	spin_unlock(&adapter->fdir_perfect_lock);
5892 }
5893 
5894 void ixgbe_down(struct ixgbe_adapter *adapter)
5895 {
5896 	struct net_device *netdev = adapter->netdev;
5897 	struct ixgbe_hw *hw = &adapter->hw;
5898 	int i;
5899 
5900 	/* signal that we are down to the interrupt handler */
5901 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5902 		return; /* do nothing if already down */
5903 
5904 	/* disable receives */
5905 	hw->mac.ops.disable_rx(hw);
5906 
5907 	/* disable all enabled rx queues */
5908 	for (i = 0; i < adapter->num_rx_queues; i++)
5909 		/* this call also flushes the previous write */
5910 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5911 
5912 	usleep_range(10000, 20000);
5913 
5914 	/* synchronize_sched() needed for pending XDP buffers to drain */
5915 	if (adapter->xdp_ring[0])
5916 		synchronize_sched();
5917 	netif_tx_stop_all_queues(netdev);
5918 
5919 	/* call carrier off first to avoid false dev_watchdog timeouts */
5920 	netif_carrier_off(netdev);
5921 	netif_tx_disable(netdev);
5922 
5923 	ixgbe_irq_disable(adapter);
5924 
5925 	ixgbe_napi_disable_all(adapter);
5926 
5927 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
5928 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5929 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5930 
5931 	del_timer_sync(&adapter->service_timer);
5932 
5933 	if (adapter->num_vfs) {
5934 		/* Clear EITR Select mapping */
5935 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5936 
5937 		/* Mark all the VFs as inactive */
5938 		for (i = 0 ; i < adapter->num_vfs; i++)
5939 			adapter->vfinfo[i].clear_to_send = false;
5940 
5941 		/* ping all the active vfs to let them know we are going down */
5942 		ixgbe_ping_all_vfs(adapter);
5943 
5944 		/* Disable all VFTE/VFRE TX/RX */
5945 		ixgbe_disable_tx_rx(adapter);
5946 	}
5947 
5948 	/* disable transmits in the hardware now that interrupts are off */
5949 	for (i = 0; i < adapter->num_tx_queues; i++) {
5950 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5951 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5952 	}
5953 	for (i = 0; i < adapter->num_xdp_queues; i++) {
5954 		u8 reg_idx = adapter->xdp_ring[i]->reg_idx;
5955 
5956 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5957 	}
5958 
5959 	/* Disable the Tx DMA engine on 82599 and later MAC */
5960 	switch (hw->mac.type) {
5961 	case ixgbe_mac_82599EB:
5962 	case ixgbe_mac_X540:
5963 	case ixgbe_mac_X550:
5964 	case ixgbe_mac_X550EM_x:
5965 	case ixgbe_mac_x550em_a:
5966 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5967 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5968 				 ~IXGBE_DMATXCTL_TE));
5969 		break;
5970 	default:
5971 		break;
5972 	}
5973 
5974 	if (!pci_channel_offline(adapter->pdev))
5975 		ixgbe_reset(adapter);
5976 
5977 	/* power down the optics for 82599 SFP+ fiber */
5978 	if (hw->mac.ops.disable_tx_laser)
5979 		hw->mac.ops.disable_tx_laser(hw);
5980 
5981 	ixgbe_clean_all_tx_rings(adapter);
5982 	ixgbe_clean_all_rx_rings(adapter);
5983 }
5984 
5985 /**
5986  * ixgbe_eee_capable - helper function to determine EEE support on X550
5987  * @adapter: board private structure
5988  */
5989 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
5990 {
5991 	struct ixgbe_hw *hw = &adapter->hw;
5992 
5993 	switch (hw->device_id) {
5994 	case IXGBE_DEV_ID_X550EM_A_1G_T:
5995 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
5996 		if (!hw->phy.eee_speeds_supported)
5997 			break;
5998 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
5999 		if (!hw->phy.eee_speeds_advertised)
6000 			break;
6001 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
6002 		break;
6003 	default:
6004 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
6005 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
6006 		break;
6007 	}
6008 }
6009 
6010 /**
6011  * ixgbe_tx_timeout - Respond to a Tx Hang
6012  * @netdev: network interface device structure
6013  **/
6014 static void ixgbe_tx_timeout(struct net_device *netdev)
6015 {
6016 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6017 
6018 	/* Do the reset outside of interrupt context */
6019 	ixgbe_tx_timeout_reset(adapter);
6020 }
6021 
6022 #ifdef CONFIG_IXGBE_DCB
6023 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
6024 {
6025 	struct ixgbe_hw *hw = &adapter->hw;
6026 	struct tc_configuration *tc;
6027 	int j;
6028 
6029 	switch (hw->mac.type) {
6030 	case ixgbe_mac_82598EB:
6031 	case ixgbe_mac_82599EB:
6032 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
6033 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
6034 		break;
6035 	case ixgbe_mac_X540:
6036 	case ixgbe_mac_X550:
6037 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
6038 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
6039 		break;
6040 	case ixgbe_mac_X550EM_x:
6041 	case ixgbe_mac_x550em_a:
6042 	default:
6043 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
6044 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
6045 		break;
6046 	}
6047 
6048 	/* Configure DCB traffic classes */
6049 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
6050 		tc = &adapter->dcb_cfg.tc_config[j];
6051 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
6052 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
6053 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
6054 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
6055 		tc->dcb_pfc = pfc_disabled;
6056 	}
6057 
6058 	/* Initialize default user to priority mapping, UPx->TC0 */
6059 	tc = &adapter->dcb_cfg.tc_config[0];
6060 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
6061 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
6062 
6063 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
6064 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
6065 	adapter->dcb_cfg.pfc_mode_enable = false;
6066 	adapter->dcb_set_bitmap = 0x00;
6067 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
6068 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
6069 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
6070 	       sizeof(adapter->temp_dcb_cfg));
6071 }
6072 #endif
6073 
6074 /**
6075  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
6076  * @adapter: board private structure to initialize
6077  * @ii: pointer to ixgbe_info for device
6078  *
6079  * ixgbe_sw_init initializes the Adapter private data structure.
6080  * Fields are initialized based on PCI device information and
6081  * OS network device settings (MTU size).
6082  **/
6083 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
6084 			 const struct ixgbe_info *ii)
6085 {
6086 	struct ixgbe_hw *hw = &adapter->hw;
6087 	struct pci_dev *pdev = adapter->pdev;
6088 	unsigned int rss, fdir;
6089 	u32 fwsm;
6090 	int i;
6091 
6092 	/* PCI config space info */
6093 
6094 	hw->vendor_id = pdev->vendor;
6095 	hw->device_id = pdev->device;
6096 	hw->revision_id = pdev->revision;
6097 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6098 	hw->subsystem_device_id = pdev->subsystem_device;
6099 
6100 	/* get_invariants needs the device IDs */
6101 	ii->get_invariants(hw);
6102 
6103 	/* Set common capability flags and settings */
6104 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6105 	adapter->ring_feature[RING_F_RSS].limit = rss;
6106 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6107 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6108 	adapter->atr_sample_rate = 20;
6109 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6110 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
6111 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6112 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
6113 #ifdef CONFIG_IXGBE_DCA
6114 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6115 #endif
6116 #ifdef CONFIG_IXGBE_DCB
6117 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6118 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6119 #endif
6120 #ifdef IXGBE_FCOE
6121 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6122 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6123 #ifdef CONFIG_IXGBE_DCB
6124 	/* Default traffic class to use for FCoE */
6125 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6126 #endif /* CONFIG_IXGBE_DCB */
6127 #endif /* IXGBE_FCOE */
6128 
6129 	/* initialize static ixgbe jump table entries */
6130 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6131 					  GFP_KERNEL);
6132 	if (!adapter->jump_tables[0])
6133 		return -ENOMEM;
6134 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6135 
6136 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6137 		adapter->jump_tables[i] = NULL;
6138 
6139 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
6140 				     hw->mac.num_rar_entries,
6141 				     GFP_ATOMIC);
6142 	if (!adapter->mac_table)
6143 		return -ENOMEM;
6144 
6145 	if (ixgbe_init_rss_key(adapter))
6146 		return -ENOMEM;
6147 
6148 	/* Set MAC specific capability flags and exceptions */
6149 	switch (hw->mac.type) {
6150 	case ixgbe_mac_82598EB:
6151 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6152 
6153 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
6154 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6155 
6156 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6157 		adapter->ring_feature[RING_F_FDIR].limit = 0;
6158 		adapter->atr_sample_rate = 0;
6159 		adapter->fdir_pballoc = 0;
6160 #ifdef IXGBE_FCOE
6161 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6162 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6163 #ifdef CONFIG_IXGBE_DCB
6164 		adapter->fcoe.up = 0;
6165 #endif /* IXGBE_DCB */
6166 #endif /* IXGBE_FCOE */
6167 		break;
6168 	case ixgbe_mac_82599EB:
6169 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6170 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6171 		break;
6172 	case ixgbe_mac_X540:
6173 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6174 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
6175 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6176 		break;
6177 	case ixgbe_mac_x550em_a:
6178 		adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
6179 		switch (hw->device_id) {
6180 		case IXGBE_DEV_ID_X550EM_A_1G_T:
6181 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6182 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6183 			break;
6184 		default:
6185 			break;
6186 		}
6187 	/* fall through */
6188 	case ixgbe_mac_X550EM_x:
6189 #ifdef CONFIG_IXGBE_DCB
6190 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6191 #endif
6192 #ifdef IXGBE_FCOE
6193 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6194 #ifdef CONFIG_IXGBE_DCB
6195 		adapter->fcoe.up = 0;
6196 #endif /* IXGBE_DCB */
6197 #endif /* IXGBE_FCOE */
6198 	/* Fall Through */
6199 	case ixgbe_mac_X550:
6200 		if (hw->mac.type == ixgbe_mac_X550)
6201 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6202 #ifdef CONFIG_IXGBE_DCA
6203 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6204 #endif
6205 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
6206 		break;
6207 	default:
6208 		break;
6209 	}
6210 
6211 #ifdef IXGBE_FCOE
6212 	/* FCoE support exists, always init the FCoE lock */
6213 	spin_lock_init(&adapter->fcoe.lock);
6214 
6215 #endif
6216 	/* n-tuple support exists, always init our spinlock */
6217 	spin_lock_init(&adapter->fdir_perfect_lock);
6218 
6219 #ifdef CONFIG_IXGBE_DCB
6220 	ixgbe_init_dcb(adapter);
6221 #endif
6222 
6223 	/* default flow control settings */
6224 	hw->fc.requested_mode = ixgbe_fc_full;
6225 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
6226 	ixgbe_pbthresh_setup(adapter);
6227 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6228 	hw->fc.send_xon = true;
6229 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6230 
6231 #ifdef CONFIG_PCI_IOV
6232 	if (max_vfs > 0)
6233 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6234 
6235 	/* assign number of SR-IOV VFs */
6236 	if (hw->mac.type != ixgbe_mac_82598EB) {
6237 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6238 			max_vfs = 0;
6239 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6240 		}
6241 	}
6242 #endif /* CONFIG_PCI_IOV */
6243 
6244 	/* enable itr by default in dynamic mode */
6245 	adapter->rx_itr_setting = 1;
6246 	adapter->tx_itr_setting = 1;
6247 
6248 	/* set default ring sizes */
6249 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6250 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6251 
6252 	/* set default work limits */
6253 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6254 
6255 	/* initialize eeprom parameters */
6256 	if (ixgbe_init_eeprom_params_generic(hw)) {
6257 		e_dev_err("EEPROM initialization failed\n");
6258 		return -EIO;
6259 	}
6260 
6261 	/* PF holds first pool slot */
6262 	set_bit(0, adapter->fwd_bitmask);
6263 	set_bit(__IXGBE_DOWN, &adapter->state);
6264 
6265 	return 0;
6266 }
6267 
6268 /**
6269  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6270  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6271  *
6272  * Return 0 on success, negative on failure
6273  **/
6274 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6275 {
6276 	struct device *dev = tx_ring->dev;
6277 	int orig_node = dev_to_node(dev);
6278 	int ring_node = -1;
6279 	int size;
6280 
6281 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6282 
6283 	if (tx_ring->q_vector)
6284 		ring_node = tx_ring->q_vector->numa_node;
6285 
6286 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6287 	if (!tx_ring->tx_buffer_info)
6288 		tx_ring->tx_buffer_info = vmalloc(size);
6289 	if (!tx_ring->tx_buffer_info)
6290 		goto err;
6291 
6292 	/* round up to nearest 4K */
6293 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6294 	tx_ring->size = ALIGN(tx_ring->size, 4096);
6295 
6296 	set_dev_node(dev, ring_node);
6297 	tx_ring->desc = dma_alloc_coherent(dev,
6298 					   tx_ring->size,
6299 					   &tx_ring->dma,
6300 					   GFP_KERNEL);
6301 	set_dev_node(dev, orig_node);
6302 	if (!tx_ring->desc)
6303 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6304 						   &tx_ring->dma, GFP_KERNEL);
6305 	if (!tx_ring->desc)
6306 		goto err;
6307 
6308 	tx_ring->next_to_use = 0;
6309 	tx_ring->next_to_clean = 0;
6310 	return 0;
6311 
6312 err:
6313 	vfree(tx_ring->tx_buffer_info);
6314 	tx_ring->tx_buffer_info = NULL;
6315 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6316 	return -ENOMEM;
6317 }
6318 
6319 /**
6320  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6321  * @adapter: board private structure
6322  *
6323  * If this function returns with an error, then it's possible one or
6324  * more of the rings is populated (while the rest are not).  It is the
6325  * callers duty to clean those orphaned rings.
6326  *
6327  * Return 0 on success, negative on failure
6328  **/
6329 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6330 {
6331 	int i, j = 0, err = 0;
6332 
6333 	for (i = 0; i < adapter->num_tx_queues; i++) {
6334 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6335 		if (!err)
6336 			continue;
6337 
6338 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6339 		goto err_setup_tx;
6340 	}
6341 	for (j = 0; j < adapter->num_xdp_queues; j++) {
6342 		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6343 		if (!err)
6344 			continue;
6345 
6346 		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6347 		goto err_setup_tx;
6348 	}
6349 
6350 	return 0;
6351 err_setup_tx:
6352 	/* rewind the index freeing the rings as we go */
6353 	while (j--)
6354 		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6355 	while (i--)
6356 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
6357 	return err;
6358 }
6359 
6360 /**
6361  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6362  * @adapter: pointer to ixgbe_adapter
6363  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6364  *
6365  * Returns 0 on success, negative on failure
6366  **/
6367 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6368 			     struct ixgbe_ring *rx_ring)
6369 {
6370 	struct device *dev = rx_ring->dev;
6371 	int orig_node = dev_to_node(dev);
6372 	int ring_node = -1;
6373 	int size;
6374 
6375 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6376 
6377 	if (rx_ring->q_vector)
6378 		ring_node = rx_ring->q_vector->numa_node;
6379 
6380 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6381 	if (!rx_ring->rx_buffer_info)
6382 		rx_ring->rx_buffer_info = vmalloc(size);
6383 	if (!rx_ring->rx_buffer_info)
6384 		goto err;
6385 
6386 	/* Round up to nearest 4K */
6387 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6388 	rx_ring->size = ALIGN(rx_ring->size, 4096);
6389 
6390 	set_dev_node(dev, ring_node);
6391 	rx_ring->desc = dma_alloc_coherent(dev,
6392 					   rx_ring->size,
6393 					   &rx_ring->dma,
6394 					   GFP_KERNEL);
6395 	set_dev_node(dev, orig_node);
6396 	if (!rx_ring->desc)
6397 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6398 						   &rx_ring->dma, GFP_KERNEL);
6399 	if (!rx_ring->desc)
6400 		goto err;
6401 
6402 	rx_ring->next_to_clean = 0;
6403 	rx_ring->next_to_use = 0;
6404 
6405 	/* XDP RX-queue info */
6406 	if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
6407 			     rx_ring->queue_index) < 0)
6408 		goto err;
6409 
6410 	rx_ring->xdp_prog = adapter->xdp_prog;
6411 
6412 	return 0;
6413 err:
6414 	vfree(rx_ring->rx_buffer_info);
6415 	rx_ring->rx_buffer_info = NULL;
6416 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6417 	return -ENOMEM;
6418 }
6419 
6420 /**
6421  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6422  * @adapter: board private structure
6423  *
6424  * If this function returns with an error, then it's possible one or
6425  * more of the rings is populated (while the rest are not).  It is the
6426  * callers duty to clean those orphaned rings.
6427  *
6428  * Return 0 on success, negative on failure
6429  **/
6430 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6431 {
6432 	int i, err = 0;
6433 
6434 	for (i = 0; i < adapter->num_rx_queues; i++) {
6435 		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6436 		if (!err)
6437 			continue;
6438 
6439 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6440 		goto err_setup_rx;
6441 	}
6442 
6443 #ifdef IXGBE_FCOE
6444 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
6445 	if (!err)
6446 #endif
6447 		return 0;
6448 err_setup_rx:
6449 	/* rewind the index freeing the rings as we go */
6450 	while (i--)
6451 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
6452 	return err;
6453 }
6454 
6455 /**
6456  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6457  * @tx_ring: Tx descriptor ring for a specific queue
6458  *
6459  * Free all transmit software resources
6460  **/
6461 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6462 {
6463 	ixgbe_clean_tx_ring(tx_ring);
6464 
6465 	vfree(tx_ring->tx_buffer_info);
6466 	tx_ring->tx_buffer_info = NULL;
6467 
6468 	/* if not set, then don't free */
6469 	if (!tx_ring->desc)
6470 		return;
6471 
6472 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6473 			  tx_ring->desc, tx_ring->dma);
6474 
6475 	tx_ring->desc = NULL;
6476 }
6477 
6478 /**
6479  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6480  * @adapter: board private structure
6481  *
6482  * Free all transmit software resources
6483  **/
6484 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6485 {
6486 	int i;
6487 
6488 	for (i = 0; i < adapter->num_tx_queues; i++)
6489 		if (adapter->tx_ring[i]->desc)
6490 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6491 	for (i = 0; i < adapter->num_xdp_queues; i++)
6492 		if (adapter->xdp_ring[i]->desc)
6493 			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6494 }
6495 
6496 /**
6497  * ixgbe_free_rx_resources - Free Rx Resources
6498  * @rx_ring: ring to clean the resources from
6499  *
6500  * Free all receive software resources
6501  **/
6502 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6503 {
6504 	ixgbe_clean_rx_ring(rx_ring);
6505 
6506 	rx_ring->xdp_prog = NULL;
6507 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6508 	vfree(rx_ring->rx_buffer_info);
6509 	rx_ring->rx_buffer_info = NULL;
6510 
6511 	/* if not set, then don't free */
6512 	if (!rx_ring->desc)
6513 		return;
6514 
6515 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6516 			  rx_ring->desc, rx_ring->dma);
6517 
6518 	rx_ring->desc = NULL;
6519 }
6520 
6521 /**
6522  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6523  * @adapter: board private structure
6524  *
6525  * Free all receive software resources
6526  **/
6527 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6528 {
6529 	int i;
6530 
6531 #ifdef IXGBE_FCOE
6532 	ixgbe_free_fcoe_ddp_resources(adapter);
6533 
6534 #endif
6535 	for (i = 0; i < adapter->num_rx_queues; i++)
6536 		if (adapter->rx_ring[i]->desc)
6537 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6538 }
6539 
6540 /**
6541  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6542  * @netdev: network interface device structure
6543  * @new_mtu: new value for maximum frame size
6544  *
6545  * Returns 0 on success, negative on failure
6546  **/
6547 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6548 {
6549 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6550 
6551 	/*
6552 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6553 	 * paths when MTU greater than 1500 is configured.  So display a
6554 	 * warning that legacy VFs will be disabled.
6555 	 */
6556 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6557 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6558 	    (new_mtu > ETH_DATA_LEN))
6559 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6560 
6561 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6562 
6563 	/* must set new MTU before calling down or up */
6564 	netdev->mtu = new_mtu;
6565 
6566 	if (netif_running(netdev))
6567 		ixgbe_reinit_locked(adapter);
6568 
6569 	return 0;
6570 }
6571 
6572 /**
6573  * ixgbe_open - Called when a network interface is made active
6574  * @netdev: network interface device structure
6575  *
6576  * Returns 0 on success, negative value on failure
6577  *
6578  * The open entry point is called when a network interface is made
6579  * active by the system (IFF_UP).  At this point all resources needed
6580  * for transmit and receive operations are allocated, the interrupt
6581  * handler is registered with the OS, the watchdog timer is started,
6582  * and the stack is notified that the interface is ready.
6583  **/
6584 int ixgbe_open(struct net_device *netdev)
6585 {
6586 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6587 	struct ixgbe_hw *hw = &adapter->hw;
6588 	int err, queues;
6589 
6590 	/* disallow open during test */
6591 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6592 		return -EBUSY;
6593 
6594 	netif_carrier_off(netdev);
6595 
6596 	/* allocate transmit descriptors */
6597 	err = ixgbe_setup_all_tx_resources(adapter);
6598 	if (err)
6599 		goto err_setup_tx;
6600 
6601 	/* allocate receive descriptors */
6602 	err = ixgbe_setup_all_rx_resources(adapter);
6603 	if (err)
6604 		goto err_setup_rx;
6605 
6606 	ixgbe_configure(adapter);
6607 
6608 	err = ixgbe_request_irq(adapter);
6609 	if (err)
6610 		goto err_req_irq;
6611 
6612 	/* Notify the stack of the actual queue counts. */
6613 	queues = adapter->num_tx_queues;
6614 	err = netif_set_real_num_tx_queues(netdev, queues);
6615 	if (err)
6616 		goto err_set_queues;
6617 
6618 	queues = adapter->num_rx_queues;
6619 	err = netif_set_real_num_rx_queues(netdev, queues);
6620 	if (err)
6621 		goto err_set_queues;
6622 
6623 	ixgbe_ptp_init(adapter);
6624 
6625 	ixgbe_up_complete(adapter);
6626 
6627 	ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6628 	udp_tunnel_get_rx_info(netdev);
6629 
6630 	return 0;
6631 
6632 err_set_queues:
6633 	ixgbe_free_irq(adapter);
6634 err_req_irq:
6635 	ixgbe_free_all_rx_resources(adapter);
6636 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6637 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6638 err_setup_rx:
6639 	ixgbe_free_all_tx_resources(adapter);
6640 err_setup_tx:
6641 	ixgbe_reset(adapter);
6642 
6643 	return err;
6644 }
6645 
6646 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6647 {
6648 	ixgbe_ptp_suspend(adapter);
6649 
6650 	if (adapter->hw.phy.ops.enter_lplu) {
6651 		adapter->hw.phy.reset_disable = true;
6652 		ixgbe_down(adapter);
6653 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6654 		adapter->hw.phy.reset_disable = false;
6655 	} else {
6656 		ixgbe_down(adapter);
6657 	}
6658 
6659 	ixgbe_free_irq(adapter);
6660 
6661 	ixgbe_free_all_tx_resources(adapter);
6662 	ixgbe_free_all_rx_resources(adapter);
6663 }
6664 
6665 /**
6666  * ixgbe_close - Disables a network interface
6667  * @netdev: network interface device structure
6668  *
6669  * Returns 0, this is not allowed to fail
6670  *
6671  * The close entry point is called when an interface is de-activated
6672  * by the OS.  The hardware is still under the drivers control, but
6673  * needs to be disabled.  A global MAC reset is issued to stop the
6674  * hardware, and all transmit and receive resources are freed.
6675  **/
6676 int ixgbe_close(struct net_device *netdev)
6677 {
6678 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6679 
6680 	ixgbe_ptp_stop(adapter);
6681 
6682 	if (netif_device_present(netdev))
6683 		ixgbe_close_suspend(adapter);
6684 
6685 	ixgbe_fdir_filter_exit(adapter);
6686 
6687 	ixgbe_release_hw_control(adapter);
6688 
6689 	return 0;
6690 }
6691 
6692 #ifdef CONFIG_PM
6693 static int ixgbe_resume(struct pci_dev *pdev)
6694 {
6695 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6696 	struct net_device *netdev = adapter->netdev;
6697 	u32 err;
6698 
6699 	adapter->hw.hw_addr = adapter->io_addr;
6700 	pci_set_power_state(pdev, PCI_D0);
6701 	pci_restore_state(pdev);
6702 	/*
6703 	 * pci_restore_state clears dev->state_saved so call
6704 	 * pci_save_state to restore it.
6705 	 */
6706 	pci_save_state(pdev);
6707 
6708 	err = pci_enable_device_mem(pdev);
6709 	if (err) {
6710 		e_dev_err("Cannot enable PCI device from suspend\n");
6711 		return err;
6712 	}
6713 	smp_mb__before_atomic();
6714 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6715 	pci_set_master(pdev);
6716 
6717 	pci_wake_from_d3(pdev, false);
6718 
6719 	ixgbe_reset(adapter);
6720 
6721 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6722 
6723 	rtnl_lock();
6724 	err = ixgbe_init_interrupt_scheme(adapter);
6725 	if (!err && netif_running(netdev))
6726 		err = ixgbe_open(netdev);
6727 
6728 
6729 	if (!err)
6730 		netif_device_attach(netdev);
6731 	rtnl_unlock();
6732 
6733 	return err;
6734 }
6735 #endif /* CONFIG_PM */
6736 
6737 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6738 {
6739 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6740 	struct net_device *netdev = adapter->netdev;
6741 	struct ixgbe_hw *hw = &adapter->hw;
6742 	u32 ctrl;
6743 	u32 wufc = adapter->wol;
6744 #ifdef CONFIG_PM
6745 	int retval = 0;
6746 #endif
6747 
6748 	rtnl_lock();
6749 	netif_device_detach(netdev);
6750 
6751 	if (netif_running(netdev))
6752 		ixgbe_close_suspend(adapter);
6753 
6754 	ixgbe_clear_interrupt_scheme(adapter);
6755 	rtnl_unlock();
6756 
6757 #ifdef CONFIG_PM
6758 	retval = pci_save_state(pdev);
6759 	if (retval)
6760 		return retval;
6761 
6762 #endif
6763 	if (hw->mac.ops.stop_link_on_d3)
6764 		hw->mac.ops.stop_link_on_d3(hw);
6765 
6766 	if (wufc) {
6767 		u32 fctrl;
6768 
6769 		ixgbe_set_rx_mode(netdev);
6770 
6771 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6772 		if (hw->mac.ops.enable_tx_laser)
6773 			hw->mac.ops.enable_tx_laser(hw);
6774 
6775 		/* enable the reception of multicast packets */
6776 		fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6777 		fctrl |= IXGBE_FCTRL_MPE;
6778 		IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6779 
6780 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6781 		ctrl |= IXGBE_CTRL_GIO_DIS;
6782 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6783 
6784 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6785 	} else {
6786 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6787 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6788 	}
6789 
6790 	switch (hw->mac.type) {
6791 	case ixgbe_mac_82598EB:
6792 		pci_wake_from_d3(pdev, false);
6793 		break;
6794 	case ixgbe_mac_82599EB:
6795 	case ixgbe_mac_X540:
6796 	case ixgbe_mac_X550:
6797 	case ixgbe_mac_X550EM_x:
6798 	case ixgbe_mac_x550em_a:
6799 		pci_wake_from_d3(pdev, !!wufc);
6800 		break;
6801 	default:
6802 		break;
6803 	}
6804 
6805 	*enable_wake = !!wufc;
6806 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6807 		hw->phy.ops.set_phy_power(hw, false);
6808 
6809 	ixgbe_release_hw_control(adapter);
6810 
6811 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6812 		pci_disable_device(pdev);
6813 
6814 	return 0;
6815 }
6816 
6817 #ifdef CONFIG_PM
6818 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6819 {
6820 	int retval;
6821 	bool wake;
6822 
6823 	retval = __ixgbe_shutdown(pdev, &wake);
6824 	if (retval)
6825 		return retval;
6826 
6827 	if (wake) {
6828 		pci_prepare_to_sleep(pdev);
6829 	} else {
6830 		pci_wake_from_d3(pdev, false);
6831 		pci_set_power_state(pdev, PCI_D3hot);
6832 	}
6833 
6834 	return 0;
6835 }
6836 #endif /* CONFIG_PM */
6837 
6838 static void ixgbe_shutdown(struct pci_dev *pdev)
6839 {
6840 	bool wake;
6841 
6842 	__ixgbe_shutdown(pdev, &wake);
6843 
6844 	if (system_state == SYSTEM_POWER_OFF) {
6845 		pci_wake_from_d3(pdev, wake);
6846 		pci_set_power_state(pdev, PCI_D3hot);
6847 	}
6848 }
6849 
6850 /**
6851  * ixgbe_update_stats - Update the board statistics counters.
6852  * @adapter: board private structure
6853  **/
6854 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6855 {
6856 	struct net_device *netdev = adapter->netdev;
6857 	struct ixgbe_hw *hw = &adapter->hw;
6858 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6859 	u64 total_mpc = 0;
6860 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6861 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6862 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6863 	u64 alloc_rx_page = 0;
6864 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6865 
6866 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6867 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6868 		return;
6869 
6870 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6871 		u64 rsc_count = 0;
6872 		u64 rsc_flush = 0;
6873 		for (i = 0; i < adapter->num_rx_queues; i++) {
6874 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6875 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6876 		}
6877 		adapter->rsc_total_count = rsc_count;
6878 		adapter->rsc_total_flush = rsc_flush;
6879 	}
6880 
6881 	for (i = 0; i < adapter->num_rx_queues; i++) {
6882 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6883 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6884 		alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
6885 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6886 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6887 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6888 		bytes += rx_ring->stats.bytes;
6889 		packets += rx_ring->stats.packets;
6890 	}
6891 	adapter->non_eop_descs = non_eop_descs;
6892 	adapter->alloc_rx_page = alloc_rx_page;
6893 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6894 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6895 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6896 	netdev->stats.rx_bytes = bytes;
6897 	netdev->stats.rx_packets = packets;
6898 
6899 	bytes = 0;
6900 	packets = 0;
6901 	/* gather some stats to the adapter struct that are per queue */
6902 	for (i = 0; i < adapter->num_tx_queues; i++) {
6903 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6904 		restart_queue += tx_ring->tx_stats.restart_queue;
6905 		tx_busy += tx_ring->tx_stats.tx_busy;
6906 		bytes += tx_ring->stats.bytes;
6907 		packets += tx_ring->stats.packets;
6908 	}
6909 	for (i = 0; i < adapter->num_xdp_queues; i++) {
6910 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
6911 
6912 		restart_queue += xdp_ring->tx_stats.restart_queue;
6913 		tx_busy += xdp_ring->tx_stats.tx_busy;
6914 		bytes += xdp_ring->stats.bytes;
6915 		packets += xdp_ring->stats.packets;
6916 	}
6917 	adapter->restart_queue = restart_queue;
6918 	adapter->tx_busy = tx_busy;
6919 	netdev->stats.tx_bytes = bytes;
6920 	netdev->stats.tx_packets = packets;
6921 
6922 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6923 
6924 	/* 8 register reads */
6925 	for (i = 0; i < 8; i++) {
6926 		/* for packet buffers not used, the register should read 0 */
6927 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6928 		missed_rx += mpc;
6929 		hwstats->mpc[i] += mpc;
6930 		total_mpc += hwstats->mpc[i];
6931 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6932 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6933 		switch (hw->mac.type) {
6934 		case ixgbe_mac_82598EB:
6935 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6936 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6937 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6938 			hwstats->pxonrxc[i] +=
6939 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6940 			break;
6941 		case ixgbe_mac_82599EB:
6942 		case ixgbe_mac_X540:
6943 		case ixgbe_mac_X550:
6944 		case ixgbe_mac_X550EM_x:
6945 		case ixgbe_mac_x550em_a:
6946 			hwstats->pxonrxc[i] +=
6947 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6948 			break;
6949 		default:
6950 			break;
6951 		}
6952 	}
6953 
6954 	/*16 register reads */
6955 	for (i = 0; i < 16; i++) {
6956 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6957 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6958 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6959 		    (hw->mac.type == ixgbe_mac_X540) ||
6960 		    (hw->mac.type == ixgbe_mac_X550) ||
6961 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6962 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6963 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6964 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6965 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6966 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6967 		}
6968 	}
6969 
6970 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6971 	/* work around hardware counting issue */
6972 	hwstats->gprc -= missed_rx;
6973 
6974 	ixgbe_update_xoff_received(adapter);
6975 
6976 	/* 82598 hardware only has a 32 bit counter in the high register */
6977 	switch (hw->mac.type) {
6978 	case ixgbe_mac_82598EB:
6979 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6980 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6981 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6982 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6983 		break;
6984 	case ixgbe_mac_X540:
6985 	case ixgbe_mac_X550:
6986 	case ixgbe_mac_X550EM_x:
6987 	case ixgbe_mac_x550em_a:
6988 		/* OS2BMC stats are X540 and later */
6989 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6990 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6991 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6992 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6993 		/* fall through */
6994 	case ixgbe_mac_82599EB:
6995 		for (i = 0; i < 16; i++)
6996 			adapter->hw_rx_no_dma_resources +=
6997 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6998 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6999 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
7000 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
7001 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
7002 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
7003 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
7004 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
7005 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
7006 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
7007 #ifdef IXGBE_FCOE
7008 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
7009 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
7010 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
7011 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
7012 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
7013 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
7014 		/* Add up per cpu counters for total ddp aloc fail */
7015 		if (adapter->fcoe.ddp_pool) {
7016 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
7017 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
7018 			unsigned int cpu;
7019 			u64 noddp = 0, noddp_ext_buff = 0;
7020 			for_each_possible_cpu(cpu) {
7021 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
7022 				noddp += ddp_pool->noddp;
7023 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
7024 			}
7025 			hwstats->fcoe_noddp = noddp;
7026 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
7027 		}
7028 #endif /* IXGBE_FCOE */
7029 		break;
7030 	default:
7031 		break;
7032 	}
7033 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
7034 	hwstats->bprc += bprc;
7035 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
7036 	if (hw->mac.type == ixgbe_mac_82598EB)
7037 		hwstats->mprc -= bprc;
7038 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
7039 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
7040 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
7041 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
7042 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
7043 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
7044 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
7045 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
7046 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
7047 	hwstats->lxontxc += lxon;
7048 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
7049 	hwstats->lxofftxc += lxoff;
7050 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
7051 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
7052 	/*
7053 	 * 82598 errata - tx of flow control packets is included in tx counters
7054 	 */
7055 	xon_off_tot = lxon + lxoff;
7056 	hwstats->gptc -= xon_off_tot;
7057 	hwstats->mptc -= xon_off_tot;
7058 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
7059 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
7060 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
7061 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
7062 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
7063 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
7064 	hwstats->ptc64 -= xon_off_tot;
7065 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
7066 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
7067 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
7068 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
7069 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
7070 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
7071 
7072 	/* Fill out the OS statistics structure */
7073 	netdev->stats.multicast = hwstats->mprc;
7074 
7075 	/* Rx Errors */
7076 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
7077 	netdev->stats.rx_dropped = 0;
7078 	netdev->stats.rx_length_errors = hwstats->rlec;
7079 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
7080 	netdev->stats.rx_missed_errors = total_mpc;
7081 }
7082 
7083 /**
7084  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7085  * @adapter: pointer to the device adapter structure
7086  **/
7087 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7088 {
7089 	struct ixgbe_hw *hw = &adapter->hw;
7090 	int i;
7091 
7092 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7093 		return;
7094 
7095 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7096 
7097 	/* if interface is down do nothing */
7098 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7099 		return;
7100 
7101 	/* do nothing if we are not using signature filters */
7102 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7103 		return;
7104 
7105 	adapter->fdir_overflow++;
7106 
7107 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7108 		for (i = 0; i < adapter->num_tx_queues; i++)
7109 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7110 				&(adapter->tx_ring[i]->state));
7111 		for (i = 0; i < adapter->num_xdp_queues; i++)
7112 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7113 				&adapter->xdp_ring[i]->state);
7114 		/* re-enable flow director interrupts */
7115 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7116 	} else {
7117 		e_err(probe, "failed to finish FDIR re-initialization, "
7118 		      "ignored adding FDIR ATR filters\n");
7119 	}
7120 }
7121 
7122 /**
7123  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7124  * @adapter: pointer to the device adapter structure
7125  *
7126  * This function serves two purposes.  First it strobes the interrupt lines
7127  * in order to make certain interrupts are occurring.  Secondly it sets the
7128  * bits needed to check for TX hangs.  As a result we should immediately
7129  * determine if a hang has occurred.
7130  */
7131 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7132 {
7133 	struct ixgbe_hw *hw = &adapter->hw;
7134 	u64 eics = 0;
7135 	int i;
7136 
7137 	/* If we're down, removing or resetting, just bail */
7138 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7139 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7140 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7141 		return;
7142 
7143 	/* Force detection of hung controller */
7144 	if (netif_carrier_ok(adapter->netdev)) {
7145 		for (i = 0; i < adapter->num_tx_queues; i++)
7146 			set_check_for_tx_hang(adapter->tx_ring[i]);
7147 		for (i = 0; i < adapter->num_xdp_queues; i++)
7148 			set_check_for_tx_hang(adapter->xdp_ring[i]);
7149 	}
7150 
7151 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7152 		/*
7153 		 * for legacy and MSI interrupts don't set any bits
7154 		 * that are enabled for EIAM, because this operation
7155 		 * would set *both* EIMS and EICS for any bit in EIAM
7156 		 */
7157 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
7158 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7159 	} else {
7160 		/* get one bit for every active tx/rx interrupt vector */
7161 		for (i = 0; i < adapter->num_q_vectors; i++) {
7162 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
7163 			if (qv->rx.ring || qv->tx.ring)
7164 				eics |= BIT_ULL(i);
7165 		}
7166 	}
7167 
7168 	/* Cause software interrupt to ensure rings are cleaned */
7169 	ixgbe_irq_rearm_queues(adapter, eics);
7170 }
7171 
7172 /**
7173  * ixgbe_watchdog_update_link - update the link status
7174  * @adapter: pointer to the device adapter structure
7175  **/
7176 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7177 {
7178 	struct ixgbe_hw *hw = &adapter->hw;
7179 	u32 link_speed = adapter->link_speed;
7180 	bool link_up = adapter->link_up;
7181 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7182 
7183 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7184 		return;
7185 
7186 	if (hw->mac.ops.check_link) {
7187 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7188 	} else {
7189 		/* always assume link is up, if no check link function */
7190 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7191 		link_up = true;
7192 	}
7193 
7194 	if (adapter->ixgbe_ieee_pfc)
7195 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7196 
7197 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7198 		hw->mac.ops.fc_enable(hw);
7199 		ixgbe_set_rx_drop_en(adapter);
7200 	}
7201 
7202 	if (link_up ||
7203 	    time_after(jiffies, (adapter->link_check_timeout +
7204 				 IXGBE_TRY_LINK_TIMEOUT))) {
7205 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7206 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7207 		IXGBE_WRITE_FLUSH(hw);
7208 	}
7209 
7210 	adapter->link_up = link_up;
7211 	adapter->link_speed = link_speed;
7212 }
7213 
7214 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7215 {
7216 #ifdef CONFIG_IXGBE_DCB
7217 	struct net_device *netdev = adapter->netdev;
7218 	struct dcb_app app = {
7219 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7220 			      .protocol = 0,
7221 			     };
7222 	u8 up = 0;
7223 
7224 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7225 		up = dcb_ieee_getapp_mask(netdev, &app);
7226 
7227 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7228 #endif
7229 }
7230 
7231 /**
7232  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7233  *                             print link up message
7234  * @adapter: pointer to the device adapter structure
7235  **/
7236 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7237 {
7238 	struct net_device *netdev = adapter->netdev;
7239 	struct ixgbe_hw *hw = &adapter->hw;
7240 	u32 link_speed = adapter->link_speed;
7241 	const char *speed_str;
7242 	bool flow_rx, flow_tx;
7243 
7244 	/* only continue if link was previously down */
7245 	if (netif_carrier_ok(netdev))
7246 		return;
7247 
7248 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7249 
7250 	switch (hw->mac.type) {
7251 	case ixgbe_mac_82598EB: {
7252 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7253 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7254 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7255 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7256 	}
7257 		break;
7258 	case ixgbe_mac_X540:
7259 	case ixgbe_mac_X550:
7260 	case ixgbe_mac_X550EM_x:
7261 	case ixgbe_mac_x550em_a:
7262 	case ixgbe_mac_82599EB: {
7263 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7264 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7265 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7266 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7267 	}
7268 		break;
7269 	default:
7270 		flow_tx = false;
7271 		flow_rx = false;
7272 		break;
7273 	}
7274 
7275 	adapter->last_rx_ptp_check = jiffies;
7276 
7277 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7278 		ixgbe_ptp_start_cyclecounter(adapter);
7279 
7280 	switch (link_speed) {
7281 	case IXGBE_LINK_SPEED_10GB_FULL:
7282 		speed_str = "10 Gbps";
7283 		break;
7284 	case IXGBE_LINK_SPEED_5GB_FULL:
7285 		speed_str = "5 Gbps";
7286 		break;
7287 	case IXGBE_LINK_SPEED_2_5GB_FULL:
7288 		speed_str = "2.5 Gbps";
7289 		break;
7290 	case IXGBE_LINK_SPEED_1GB_FULL:
7291 		speed_str = "1 Gbps";
7292 		break;
7293 	case IXGBE_LINK_SPEED_100_FULL:
7294 		speed_str = "100 Mbps";
7295 		break;
7296 	case IXGBE_LINK_SPEED_10_FULL:
7297 		speed_str = "10 Mbps";
7298 		break;
7299 	default:
7300 		speed_str = "unknown speed";
7301 		break;
7302 	}
7303 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7304 	       ((flow_rx && flow_tx) ? "RX/TX" :
7305 	       (flow_rx ? "RX" :
7306 	       (flow_tx ? "TX" : "None"))));
7307 
7308 	netif_carrier_on(netdev);
7309 	ixgbe_check_vf_rate_limit(adapter);
7310 
7311 	/* enable transmits */
7312 	netif_tx_wake_all_queues(adapter->netdev);
7313 
7314 	/* update the default user priority for VFs */
7315 	ixgbe_update_default_up(adapter);
7316 
7317 	/* ping all the active vfs to let them know link has changed */
7318 	ixgbe_ping_all_vfs(adapter);
7319 }
7320 
7321 /**
7322  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7323  *                               print link down message
7324  * @adapter: pointer to the adapter structure
7325  **/
7326 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7327 {
7328 	struct net_device *netdev = adapter->netdev;
7329 	struct ixgbe_hw *hw = &adapter->hw;
7330 
7331 	adapter->link_up = false;
7332 	adapter->link_speed = 0;
7333 
7334 	/* only continue if link was up previously */
7335 	if (!netif_carrier_ok(netdev))
7336 		return;
7337 
7338 	/* poll for SFP+ cable when link is down */
7339 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7340 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7341 
7342 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7343 		ixgbe_ptp_start_cyclecounter(adapter);
7344 
7345 	e_info(drv, "NIC Link is Down\n");
7346 	netif_carrier_off(netdev);
7347 
7348 	/* ping all the active vfs to let them know link has changed */
7349 	ixgbe_ping_all_vfs(adapter);
7350 }
7351 
7352 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7353 {
7354 	int i;
7355 
7356 	for (i = 0; i < adapter->num_tx_queues; i++) {
7357 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7358 
7359 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
7360 			return true;
7361 	}
7362 
7363 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7364 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
7365 
7366 		if (ring->next_to_use != ring->next_to_clean)
7367 			return true;
7368 	}
7369 
7370 	return false;
7371 }
7372 
7373 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7374 {
7375 	struct ixgbe_hw *hw = &adapter->hw;
7376 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7377 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7378 
7379 	int i, j;
7380 
7381 	if (!adapter->num_vfs)
7382 		return false;
7383 
7384 	/* resetting the PF is only needed for MAC before X550 */
7385 	if (hw->mac.type >= ixgbe_mac_X550)
7386 		return false;
7387 
7388 	for (i = 0; i < adapter->num_vfs; i++) {
7389 		for (j = 0; j < q_per_pool; j++) {
7390 			u32 h, t;
7391 
7392 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7393 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7394 
7395 			if (h != t)
7396 				return true;
7397 		}
7398 	}
7399 
7400 	return false;
7401 }
7402 
7403 /**
7404  * ixgbe_watchdog_flush_tx - flush queues on link down
7405  * @adapter: pointer to the device adapter structure
7406  **/
7407 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7408 {
7409 	if (!netif_carrier_ok(adapter->netdev)) {
7410 		if (ixgbe_ring_tx_pending(adapter) ||
7411 		    ixgbe_vf_tx_pending(adapter)) {
7412 			/* We've lost link, so the controller stops DMA,
7413 			 * but we've got queued Tx work that's never going
7414 			 * to get done, so reset controller to flush Tx.
7415 			 * (Do the reset outside of interrupt context).
7416 			 */
7417 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7418 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7419 		}
7420 	}
7421 }
7422 
7423 #ifdef CONFIG_PCI_IOV
7424 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7425 {
7426 	struct ixgbe_hw *hw = &adapter->hw;
7427 	struct pci_dev *pdev = adapter->pdev;
7428 	unsigned int vf;
7429 	u32 gpc;
7430 
7431 	if (!(netif_carrier_ok(adapter->netdev)))
7432 		return;
7433 
7434 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7435 	if (gpc) /* If incrementing then no need for the check below */
7436 		return;
7437 	/* Check to see if a bad DMA write target from an errant or
7438 	 * malicious VF has caused a PCIe error.  If so then we can
7439 	 * issue a VFLR to the offending VF(s) and then resume without
7440 	 * requesting a full slot reset.
7441 	 */
7442 
7443 	if (!pdev)
7444 		return;
7445 
7446 	/* check status reg for all VFs owned by this PF */
7447 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
7448 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7449 		u16 status_reg;
7450 
7451 		if (!vfdev)
7452 			continue;
7453 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7454 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7455 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
7456 			pcie_flr(vfdev);
7457 	}
7458 }
7459 
7460 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7461 {
7462 	u32 ssvpc;
7463 
7464 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7465 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7466 	    adapter->num_vfs == 0)
7467 		return;
7468 
7469 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7470 
7471 	/*
7472 	 * ssvpc register is cleared on read, if zero then no
7473 	 * spoofed packets in the last interval.
7474 	 */
7475 	if (!ssvpc)
7476 		return;
7477 
7478 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7479 }
7480 #else
7481 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7482 {
7483 }
7484 
7485 static void
7486 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7487 {
7488 }
7489 #endif /* CONFIG_PCI_IOV */
7490 
7491 
7492 /**
7493  * ixgbe_watchdog_subtask - check and bring link up
7494  * @adapter: pointer to the device adapter structure
7495  **/
7496 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7497 {
7498 	/* if interface is down, removing or resetting, do nothing */
7499 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7500 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7501 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7502 		return;
7503 
7504 	ixgbe_watchdog_update_link(adapter);
7505 
7506 	if (adapter->link_up)
7507 		ixgbe_watchdog_link_is_up(adapter);
7508 	else
7509 		ixgbe_watchdog_link_is_down(adapter);
7510 
7511 	ixgbe_check_for_bad_vf(adapter);
7512 	ixgbe_spoof_check(adapter);
7513 	ixgbe_update_stats(adapter);
7514 
7515 	ixgbe_watchdog_flush_tx(adapter);
7516 }
7517 
7518 /**
7519  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7520  * @adapter: the ixgbe adapter structure
7521  **/
7522 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7523 {
7524 	struct ixgbe_hw *hw = &adapter->hw;
7525 	s32 err;
7526 
7527 	/* not searching for SFP so there is nothing to do here */
7528 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7529 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7530 		return;
7531 
7532 	if (adapter->sfp_poll_time &&
7533 	    time_after(adapter->sfp_poll_time, jiffies))
7534 		return; /* If not yet time to poll for SFP */
7535 
7536 	/* someone else is in init, wait until next service event */
7537 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7538 		return;
7539 
7540 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7541 
7542 	err = hw->phy.ops.identify_sfp(hw);
7543 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7544 		goto sfp_out;
7545 
7546 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7547 		/* If no cable is present, then we need to reset
7548 		 * the next time we find a good cable. */
7549 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7550 	}
7551 
7552 	/* exit on error */
7553 	if (err)
7554 		goto sfp_out;
7555 
7556 	/* exit if reset not needed */
7557 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7558 		goto sfp_out;
7559 
7560 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7561 
7562 	/*
7563 	 * A module may be identified correctly, but the EEPROM may not have
7564 	 * support for that module.  setup_sfp() will fail in that case, so
7565 	 * we should not allow that module to load.
7566 	 */
7567 	if (hw->mac.type == ixgbe_mac_82598EB)
7568 		err = hw->phy.ops.reset(hw);
7569 	else
7570 		err = hw->mac.ops.setup_sfp(hw);
7571 
7572 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7573 		goto sfp_out;
7574 
7575 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7576 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7577 
7578 sfp_out:
7579 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7580 
7581 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7582 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7583 		e_dev_err("failed to initialize because an unsupported "
7584 			  "SFP+ module type was detected.\n");
7585 		e_dev_err("Reload the driver after installing a "
7586 			  "supported module.\n");
7587 		unregister_netdev(adapter->netdev);
7588 	}
7589 }
7590 
7591 /**
7592  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7593  * @adapter: the ixgbe adapter structure
7594  **/
7595 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7596 {
7597 	struct ixgbe_hw *hw = &adapter->hw;
7598 	u32 cap_speed;
7599 	u32 speed;
7600 	bool autoneg = false;
7601 
7602 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7603 		return;
7604 
7605 	/* someone else is in init, wait until next service event */
7606 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7607 		return;
7608 
7609 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7610 
7611 	hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
7612 
7613 	/* advertise highest capable link speed */
7614 	if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
7615 		speed = IXGBE_LINK_SPEED_10GB_FULL;
7616 	else
7617 		speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
7618 				     IXGBE_LINK_SPEED_1GB_FULL);
7619 
7620 	if (hw->mac.ops.setup_link)
7621 		hw->mac.ops.setup_link(hw, speed, true);
7622 
7623 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7624 	adapter->link_check_timeout = jiffies;
7625 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7626 }
7627 
7628 /**
7629  * ixgbe_service_timer - Timer Call-back
7630  * @t: pointer to timer_list structure
7631  **/
7632 static void ixgbe_service_timer(struct timer_list *t)
7633 {
7634 	struct ixgbe_adapter *adapter = from_timer(adapter, t, service_timer);
7635 	unsigned long next_event_offset;
7636 
7637 	/* poll faster when waiting for link */
7638 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7639 		next_event_offset = HZ / 10;
7640 	else
7641 		next_event_offset = HZ * 2;
7642 
7643 	/* Reset the timer */
7644 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7645 
7646 	ixgbe_service_event_schedule(adapter);
7647 }
7648 
7649 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7650 {
7651 	struct ixgbe_hw *hw = &adapter->hw;
7652 	u32 status;
7653 
7654 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7655 		return;
7656 
7657 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7658 
7659 	if (!hw->phy.ops.handle_lasi)
7660 		return;
7661 
7662 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7663 	if (status != IXGBE_ERR_OVERTEMP)
7664 		return;
7665 
7666 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7667 }
7668 
7669 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7670 {
7671 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7672 		return;
7673 
7674 	/* If we're already down, removing or resetting, just bail */
7675 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7676 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7677 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7678 		return;
7679 
7680 	ixgbe_dump(adapter);
7681 	netdev_err(adapter->netdev, "Reset adapter\n");
7682 	adapter->tx_timeout_count++;
7683 
7684 	rtnl_lock();
7685 	ixgbe_reinit_locked(adapter);
7686 	rtnl_unlock();
7687 }
7688 
7689 /**
7690  * ixgbe_service_task - manages and runs subtasks
7691  * @work: pointer to work_struct containing our data
7692  **/
7693 static void ixgbe_service_task(struct work_struct *work)
7694 {
7695 	struct ixgbe_adapter *adapter = container_of(work,
7696 						     struct ixgbe_adapter,
7697 						     service_task);
7698 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7699 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7700 			rtnl_lock();
7701 			ixgbe_down(adapter);
7702 			rtnl_unlock();
7703 		}
7704 		ixgbe_service_event_complete(adapter);
7705 		return;
7706 	}
7707 	if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7708 		rtnl_lock();
7709 		adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7710 		udp_tunnel_get_rx_info(adapter->netdev);
7711 		rtnl_unlock();
7712 	}
7713 	ixgbe_reset_subtask(adapter);
7714 	ixgbe_phy_interrupt_subtask(adapter);
7715 	ixgbe_sfp_detection_subtask(adapter);
7716 	ixgbe_sfp_link_config_subtask(adapter);
7717 	ixgbe_check_overtemp_subtask(adapter);
7718 	ixgbe_watchdog_subtask(adapter);
7719 	ixgbe_fdir_reinit_subtask(adapter);
7720 	ixgbe_check_hang_subtask(adapter);
7721 
7722 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7723 		ixgbe_ptp_overflow_check(adapter);
7724 		if (adapter->flags & IXGBE_FLAG_RX_HWTSTAMP_IN_REGISTER)
7725 			ixgbe_ptp_rx_hang(adapter);
7726 		ixgbe_ptp_tx_hang(adapter);
7727 	}
7728 
7729 	ixgbe_service_event_complete(adapter);
7730 }
7731 
7732 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7733 		     struct ixgbe_tx_buffer *first,
7734 		     u8 *hdr_len,
7735 		     struct ixgbe_ipsec_tx_data *itd)
7736 {
7737 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7738 	struct sk_buff *skb = first->skb;
7739 	union {
7740 		struct iphdr *v4;
7741 		struct ipv6hdr *v6;
7742 		unsigned char *hdr;
7743 	} ip;
7744 	union {
7745 		struct tcphdr *tcp;
7746 		unsigned char *hdr;
7747 	} l4;
7748 	u32 paylen, l4_offset;
7749 	u32 fceof_saidx = 0;
7750 	int err;
7751 
7752 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7753 		return 0;
7754 
7755 	if (!skb_is_gso(skb))
7756 		return 0;
7757 
7758 	err = skb_cow_head(skb, 0);
7759 	if (err < 0)
7760 		return err;
7761 
7762 	if (eth_p_mpls(first->protocol))
7763 		ip.hdr = skb_inner_network_header(skb);
7764 	else
7765 		ip.hdr = skb_network_header(skb);
7766 	l4.hdr = skb_checksum_start(skb);
7767 
7768 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7769 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7770 
7771 	/* initialize outer IP header fields */
7772 	if (ip.v4->version == 4) {
7773 		unsigned char *csum_start = skb_checksum_start(skb);
7774 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
7775 		int len = csum_start - trans_start;
7776 
7777 		/* IP header will have to cancel out any data that
7778 		 * is not a part of the outer IP header, so set to
7779 		 * a reverse csum if needed, else init check to 0.
7780 		 */
7781 		ip.v4->check = (skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) ?
7782 					   csum_fold(csum_partial(trans_start,
7783 								  len, 0)) : 0;
7784 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7785 
7786 		ip.v4->tot_len = 0;
7787 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7788 				   IXGBE_TX_FLAGS_CSUM |
7789 				   IXGBE_TX_FLAGS_IPV4;
7790 	} else {
7791 		ip.v6->payload_len = 0;
7792 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7793 				   IXGBE_TX_FLAGS_CSUM;
7794 	}
7795 
7796 	/* determine offset of inner transport header */
7797 	l4_offset = l4.hdr - skb->data;
7798 
7799 	/* compute length of segmentation header */
7800 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7801 
7802 	/* remove payload length from inner checksum */
7803 	paylen = skb->len - l4_offset;
7804 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
7805 
7806 	/* update gso size and bytecount with header size */
7807 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7808 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7809 
7810 	/* mss_l4len_id: use 0 as index for TSO */
7811 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7812 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7813 
7814 	fceof_saidx |= itd->sa_idx;
7815 	type_tucmd |= itd->flags | itd->trailer_len;
7816 
7817 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7818 	vlan_macip_lens = l4.hdr - ip.hdr;
7819 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7820 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7821 
7822 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd,
7823 			  mss_l4len_idx);
7824 
7825 	return 1;
7826 }
7827 
7828 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7829 {
7830 	unsigned int offset = 0;
7831 
7832 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7833 
7834 	return offset == skb_checksum_start_offset(skb);
7835 }
7836 
7837 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7838 			  struct ixgbe_tx_buffer *first,
7839 			  struct ixgbe_ipsec_tx_data *itd)
7840 {
7841 	struct sk_buff *skb = first->skb;
7842 	u32 vlan_macip_lens = 0;
7843 	u32 fceof_saidx = 0;
7844 	u32 type_tucmd = 0;
7845 
7846 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7847 csum_failed:
7848 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7849 					 IXGBE_TX_FLAGS_CC)))
7850 			return;
7851 		goto no_csum;
7852 	}
7853 
7854 	switch (skb->csum_offset) {
7855 	case offsetof(struct tcphdr, check):
7856 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7857 		/* fall through */
7858 	case offsetof(struct udphdr, check):
7859 		break;
7860 	case offsetof(struct sctphdr, checksum):
7861 		/* validate that this is actually an SCTP request */
7862 		if (((first->protocol == htons(ETH_P_IP)) &&
7863 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7864 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7865 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7866 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7867 			break;
7868 		}
7869 		/* fall through */
7870 	default:
7871 		skb_checksum_help(skb);
7872 		goto csum_failed;
7873 	}
7874 
7875 	/* update TX checksum flag */
7876 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7877 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7878 			  skb_network_offset(skb);
7879 no_csum:
7880 	/* vlan_macip_lens: MACLEN, VLAN tag */
7881 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7882 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7883 
7884 	fceof_saidx |= itd->sa_idx;
7885 	type_tucmd |= itd->flags | itd->trailer_len;
7886 
7887 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
7888 }
7889 
7890 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7891 	((_flag <= _result) ? \
7892 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7893 	 ((u32)(_input & _flag) / (_flag / _result)))
7894 
7895 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7896 {
7897 	/* set type for advanced descriptor with frame checksum insertion */
7898 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7899 		       IXGBE_ADVTXD_DCMD_DEXT |
7900 		       IXGBE_ADVTXD_DCMD_IFCS;
7901 
7902 	/* set HW vlan bit if vlan is present */
7903 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7904 				   IXGBE_ADVTXD_DCMD_VLE);
7905 
7906 	/* set segmentation enable bits for TSO/FSO */
7907 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7908 				   IXGBE_ADVTXD_DCMD_TSE);
7909 
7910 	/* set timestamp bit if present */
7911 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7912 				   IXGBE_ADVTXD_MAC_TSTAMP);
7913 
7914 	/* insert frame checksum */
7915 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7916 
7917 	return cmd_type;
7918 }
7919 
7920 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7921 				   u32 tx_flags, unsigned int paylen)
7922 {
7923 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7924 
7925 	/* enable L4 checksum for TSO and TX checksum offload */
7926 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7927 					IXGBE_TX_FLAGS_CSUM,
7928 					IXGBE_ADVTXD_POPTS_TXSM);
7929 
7930 	/* enable IPv4 checksum for TSO */
7931 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7932 					IXGBE_TX_FLAGS_IPV4,
7933 					IXGBE_ADVTXD_POPTS_IXSM);
7934 
7935 	/* enable IPsec */
7936 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7937 					IXGBE_TX_FLAGS_IPSEC,
7938 					IXGBE_ADVTXD_POPTS_IPSEC);
7939 
7940 	/*
7941 	 * Check Context must be set if Tx switch is enabled, which it
7942 	 * always is for case where virtual functions are running
7943 	 */
7944 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7945 					IXGBE_TX_FLAGS_CC,
7946 					IXGBE_ADVTXD_CC);
7947 
7948 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7949 }
7950 
7951 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7952 {
7953 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7954 
7955 	/* Herbert's original patch had:
7956 	 *  smp_mb__after_netif_stop_queue();
7957 	 * but since that doesn't exist yet, just open code it.
7958 	 */
7959 	smp_mb();
7960 
7961 	/* We need to check again in a case another CPU has just
7962 	 * made room available.
7963 	 */
7964 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7965 		return -EBUSY;
7966 
7967 	/* A reprieve! - use start_queue because it doesn't call schedule */
7968 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7969 	++tx_ring->tx_stats.restart_queue;
7970 	return 0;
7971 }
7972 
7973 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7974 {
7975 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7976 		return 0;
7977 
7978 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7979 }
7980 
7981 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7982 		       IXGBE_TXD_CMD_RS)
7983 
7984 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7985 			struct ixgbe_tx_buffer *first,
7986 			const u8 hdr_len)
7987 {
7988 	struct sk_buff *skb = first->skb;
7989 	struct ixgbe_tx_buffer *tx_buffer;
7990 	union ixgbe_adv_tx_desc *tx_desc;
7991 	struct skb_frag_struct *frag;
7992 	dma_addr_t dma;
7993 	unsigned int data_len, size;
7994 	u32 tx_flags = first->tx_flags;
7995 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7996 	u16 i = tx_ring->next_to_use;
7997 
7998 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7999 
8000 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
8001 
8002 	size = skb_headlen(skb);
8003 	data_len = skb->data_len;
8004 
8005 #ifdef IXGBE_FCOE
8006 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
8007 		if (data_len < sizeof(struct fcoe_crc_eof)) {
8008 			size -= sizeof(struct fcoe_crc_eof) - data_len;
8009 			data_len = 0;
8010 		} else {
8011 			data_len -= sizeof(struct fcoe_crc_eof);
8012 		}
8013 	}
8014 
8015 #endif
8016 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
8017 
8018 	tx_buffer = first;
8019 
8020 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
8021 		if (dma_mapping_error(tx_ring->dev, dma))
8022 			goto dma_error;
8023 
8024 		/* record length, and DMA address */
8025 		dma_unmap_len_set(tx_buffer, len, size);
8026 		dma_unmap_addr_set(tx_buffer, dma, dma);
8027 
8028 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
8029 
8030 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
8031 			tx_desc->read.cmd_type_len =
8032 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
8033 
8034 			i++;
8035 			tx_desc++;
8036 			if (i == tx_ring->count) {
8037 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8038 				i = 0;
8039 			}
8040 			tx_desc->read.olinfo_status = 0;
8041 
8042 			dma += IXGBE_MAX_DATA_PER_TXD;
8043 			size -= IXGBE_MAX_DATA_PER_TXD;
8044 
8045 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
8046 		}
8047 
8048 		if (likely(!data_len))
8049 			break;
8050 
8051 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
8052 
8053 		i++;
8054 		tx_desc++;
8055 		if (i == tx_ring->count) {
8056 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8057 			i = 0;
8058 		}
8059 		tx_desc->read.olinfo_status = 0;
8060 
8061 #ifdef IXGBE_FCOE
8062 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
8063 #else
8064 		size = skb_frag_size(frag);
8065 #endif
8066 		data_len -= size;
8067 
8068 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
8069 				       DMA_TO_DEVICE);
8070 
8071 		tx_buffer = &tx_ring->tx_buffer_info[i];
8072 	}
8073 
8074 	/* write last descriptor with RS and EOP bits */
8075 	cmd_type |= size | IXGBE_TXD_CMD;
8076 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8077 
8078 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
8079 
8080 	/* set the timestamp */
8081 	first->time_stamp = jiffies;
8082 
8083 	/*
8084 	 * Force memory writes to complete before letting h/w know there
8085 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
8086 	 * memory model archs, such as IA-64).
8087 	 *
8088 	 * We also need this memory barrier to make certain all of the
8089 	 * status bits have been updated before next_to_watch is written.
8090 	 */
8091 	wmb();
8092 
8093 	/* set next_to_watch value indicating a packet is present */
8094 	first->next_to_watch = tx_desc;
8095 
8096 	i++;
8097 	if (i == tx_ring->count)
8098 		i = 0;
8099 
8100 	tx_ring->next_to_use = i;
8101 
8102 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8103 
8104 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
8105 		writel(i, tx_ring->tail);
8106 
8107 		/* we need this if more than one processor can write to our tail
8108 		 * at a time, it synchronizes IO on IA64/Altix systems
8109 		 */
8110 		mmiowb();
8111 	}
8112 
8113 	return 0;
8114 dma_error:
8115 	dev_err(tx_ring->dev, "TX DMA map failed\n");
8116 
8117 	/* clear dma mappings for failed tx_buffer_info map */
8118 	for (;;) {
8119 		tx_buffer = &tx_ring->tx_buffer_info[i];
8120 		if (dma_unmap_len(tx_buffer, len))
8121 			dma_unmap_page(tx_ring->dev,
8122 				       dma_unmap_addr(tx_buffer, dma),
8123 				       dma_unmap_len(tx_buffer, len),
8124 				       DMA_TO_DEVICE);
8125 		dma_unmap_len_set(tx_buffer, len, 0);
8126 		if (tx_buffer == first)
8127 			break;
8128 		if (i == 0)
8129 			i += tx_ring->count;
8130 		i--;
8131 	}
8132 
8133 	dev_kfree_skb_any(first->skb);
8134 	first->skb = NULL;
8135 
8136 	tx_ring->next_to_use = i;
8137 
8138 	return -1;
8139 }
8140 
8141 static void ixgbe_atr(struct ixgbe_ring *ring,
8142 		      struct ixgbe_tx_buffer *first)
8143 {
8144 	struct ixgbe_q_vector *q_vector = ring->q_vector;
8145 	union ixgbe_atr_hash_dword input = { .dword = 0 };
8146 	union ixgbe_atr_hash_dword common = { .dword = 0 };
8147 	union {
8148 		unsigned char *network;
8149 		struct iphdr *ipv4;
8150 		struct ipv6hdr *ipv6;
8151 	} hdr;
8152 	struct tcphdr *th;
8153 	unsigned int hlen;
8154 	struct sk_buff *skb;
8155 	__be16 vlan_id;
8156 	int l4_proto;
8157 
8158 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
8159 	if (!q_vector)
8160 		return;
8161 
8162 	/* do nothing if sampling is disabled */
8163 	if (!ring->atr_sample_rate)
8164 		return;
8165 
8166 	ring->atr_count++;
8167 
8168 	/* currently only IPv4/IPv6 with TCP is supported */
8169 	if ((first->protocol != htons(ETH_P_IP)) &&
8170 	    (first->protocol != htons(ETH_P_IPV6)))
8171 		return;
8172 
8173 	/* snag network header to get L4 type and address */
8174 	skb = first->skb;
8175 	hdr.network = skb_network_header(skb);
8176 	if (unlikely(hdr.network <= skb->data))
8177 		return;
8178 	if (skb->encapsulation &&
8179 	    first->protocol == htons(ETH_P_IP) &&
8180 	    hdr.ipv4->protocol == IPPROTO_UDP) {
8181 		struct ixgbe_adapter *adapter = q_vector->adapter;
8182 
8183 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
8184 			     VXLAN_HEADROOM))
8185 			return;
8186 
8187 		/* verify the port is recognized as VXLAN */
8188 		if (adapter->vxlan_port &&
8189 		    udp_hdr(skb)->dest == adapter->vxlan_port)
8190 			hdr.network = skb_inner_network_header(skb);
8191 
8192 		if (adapter->geneve_port &&
8193 		    udp_hdr(skb)->dest == adapter->geneve_port)
8194 			hdr.network = skb_inner_network_header(skb);
8195 	}
8196 
8197 	/* Make sure we have at least [minimum IPv4 header + TCP]
8198 	 * or [IPv6 header] bytes
8199 	 */
8200 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8201 		return;
8202 
8203 	/* Currently only IPv4/IPv6 with TCP is supported */
8204 	switch (hdr.ipv4->version) {
8205 	case IPVERSION:
8206 		/* access ihl as u8 to avoid unaligned access on ia64 */
8207 		hlen = (hdr.network[0] & 0x0F) << 2;
8208 		l4_proto = hdr.ipv4->protocol;
8209 		break;
8210 	case 6:
8211 		hlen = hdr.network - skb->data;
8212 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8213 		hlen -= hdr.network - skb->data;
8214 		break;
8215 	default:
8216 		return;
8217 	}
8218 
8219 	if (l4_proto != IPPROTO_TCP)
8220 		return;
8221 
8222 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
8223 		     hlen + sizeof(struct tcphdr)))
8224 		return;
8225 
8226 	th = (struct tcphdr *)(hdr.network + hlen);
8227 
8228 	/* skip this packet since the socket is closing */
8229 	if (th->fin)
8230 		return;
8231 
8232 	/* sample on all syn packets or once every atr sample count */
8233 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8234 		return;
8235 
8236 	/* reset sample count */
8237 	ring->atr_count = 0;
8238 
8239 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8240 
8241 	/*
8242 	 * src and dst are inverted, think how the receiver sees them
8243 	 *
8244 	 * The input is broken into two sections, a non-compressed section
8245 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8246 	 * is XORed together and stored in the compressed dword.
8247 	 */
8248 	input.formatted.vlan_id = vlan_id;
8249 
8250 	/*
8251 	 * since src port and flex bytes occupy the same word XOR them together
8252 	 * and write the value to source port portion of compressed dword
8253 	 */
8254 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8255 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8256 	else
8257 		common.port.src ^= th->dest ^ first->protocol;
8258 	common.port.dst ^= th->source;
8259 
8260 	switch (hdr.ipv4->version) {
8261 	case IPVERSION:
8262 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8263 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8264 		break;
8265 	case 6:
8266 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8267 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8268 			     hdr.ipv6->saddr.s6_addr32[1] ^
8269 			     hdr.ipv6->saddr.s6_addr32[2] ^
8270 			     hdr.ipv6->saddr.s6_addr32[3] ^
8271 			     hdr.ipv6->daddr.s6_addr32[0] ^
8272 			     hdr.ipv6->daddr.s6_addr32[1] ^
8273 			     hdr.ipv6->daddr.s6_addr32[2] ^
8274 			     hdr.ipv6->daddr.s6_addr32[3];
8275 		break;
8276 	default:
8277 		break;
8278 	}
8279 
8280 	if (hdr.network != skb_network_header(skb))
8281 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8282 
8283 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
8284 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8285 					      input, common, ring->queue_index);
8286 }
8287 
8288 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8289 			      void *accel_priv, select_queue_fallback_t fallback)
8290 {
8291 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
8292 	struct ixgbe_adapter *adapter;
8293 	int txq;
8294 #ifdef IXGBE_FCOE
8295 	struct ixgbe_ring_feature *f;
8296 #endif
8297 
8298 	if (fwd_adapter) {
8299 		adapter = netdev_priv(dev);
8300 		txq = reciprocal_scale(skb_get_hash(skb),
8301 				       adapter->num_rx_queues_per_pool);
8302 
8303 		return txq + fwd_adapter->tx_base_queue;
8304 	}
8305 
8306 #ifdef IXGBE_FCOE
8307 
8308 	/*
8309 	 * only execute the code below if protocol is FCoE
8310 	 * or FIP and we have FCoE enabled on the adapter
8311 	 */
8312 	switch (vlan_get_protocol(skb)) {
8313 	case htons(ETH_P_FCOE):
8314 	case htons(ETH_P_FIP):
8315 		adapter = netdev_priv(dev);
8316 
8317 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
8318 			break;
8319 		/* fall through */
8320 	default:
8321 		return fallback(dev, skb);
8322 	}
8323 
8324 	f = &adapter->ring_feature[RING_F_FCOE];
8325 
8326 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8327 					   smp_processor_id();
8328 
8329 	while (txq >= f->indices)
8330 		txq -= f->indices;
8331 
8332 	return txq + f->offset;
8333 #else
8334 	return fallback(dev, skb);
8335 #endif
8336 }
8337 
8338 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
8339 			       struct xdp_buff *xdp)
8340 {
8341 	struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
8342 	struct ixgbe_tx_buffer *tx_buffer;
8343 	union ixgbe_adv_tx_desc *tx_desc;
8344 	u32 len, cmd_type;
8345 	dma_addr_t dma;
8346 	u16 i;
8347 
8348 	len = xdp->data_end - xdp->data;
8349 
8350 	if (unlikely(!ixgbe_desc_unused(ring)))
8351 		return IXGBE_XDP_CONSUMED;
8352 
8353 	dma = dma_map_single(ring->dev, xdp->data, len, DMA_TO_DEVICE);
8354 	if (dma_mapping_error(ring->dev, dma))
8355 		return IXGBE_XDP_CONSUMED;
8356 
8357 	/* record the location of the first descriptor for this packet */
8358 	tx_buffer = &ring->tx_buffer_info[ring->next_to_use];
8359 	tx_buffer->bytecount = len;
8360 	tx_buffer->gso_segs = 1;
8361 	tx_buffer->protocol = 0;
8362 
8363 	i = ring->next_to_use;
8364 	tx_desc = IXGBE_TX_DESC(ring, i);
8365 
8366 	dma_unmap_len_set(tx_buffer, len, len);
8367 	dma_unmap_addr_set(tx_buffer, dma, dma);
8368 	tx_buffer->data = xdp->data;
8369 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
8370 
8371 	/* put descriptor type bits */
8372 	cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8373 		   IXGBE_ADVTXD_DCMD_DEXT |
8374 		   IXGBE_ADVTXD_DCMD_IFCS;
8375 	cmd_type |= len | IXGBE_TXD_CMD;
8376 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8377 	tx_desc->read.olinfo_status =
8378 		cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
8379 
8380 	/* Avoid any potential race with xdp_xmit and cleanup */
8381 	smp_wmb();
8382 
8383 	/* set next_to_watch value indicating a packet is present */
8384 	i++;
8385 	if (i == ring->count)
8386 		i = 0;
8387 
8388 	tx_buffer->next_to_watch = tx_desc;
8389 	ring->next_to_use = i;
8390 
8391 	return IXGBE_XDP_TX;
8392 }
8393 
8394 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8395 			  struct ixgbe_adapter *adapter,
8396 			  struct ixgbe_ring *tx_ring)
8397 {
8398 	struct ixgbe_tx_buffer *first;
8399 	int tso;
8400 	u32 tx_flags = 0;
8401 	unsigned short f;
8402 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
8403 	struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
8404 	__be16 protocol = skb->protocol;
8405 	u8 hdr_len = 0;
8406 
8407 	/*
8408 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8409 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8410 	 *       + 2 desc gap to keep tail from touching head,
8411 	 *       + 1 desc for context descriptor,
8412 	 * otherwise try next time
8413 	 */
8414 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8415 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
8416 
8417 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8418 		tx_ring->tx_stats.tx_busy++;
8419 		return NETDEV_TX_BUSY;
8420 	}
8421 
8422 	/* record the location of the first descriptor for this packet */
8423 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8424 	first->skb = skb;
8425 	first->bytecount = skb->len;
8426 	first->gso_segs = 1;
8427 
8428 	/* if we have a HW VLAN tag being added default to the HW one */
8429 	if (skb_vlan_tag_present(skb)) {
8430 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8431 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8432 	/* else if it is a SW VLAN check the next protocol and store the tag */
8433 	} else if (protocol == htons(ETH_P_8021Q)) {
8434 		struct vlan_hdr *vhdr, _vhdr;
8435 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8436 		if (!vhdr)
8437 			goto out_drop;
8438 
8439 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8440 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
8441 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8442 	}
8443 	protocol = vlan_get_protocol(skb);
8444 
8445 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8446 	    adapter->ptp_clock) {
8447 		if (!test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8448 					   &adapter->state)) {
8449 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8450 			tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8451 
8452 			/* schedule check for Tx timestamp */
8453 			adapter->ptp_tx_skb = skb_get(skb);
8454 			adapter->ptp_tx_start = jiffies;
8455 			schedule_work(&adapter->ptp_tx_work);
8456 		} else {
8457 			adapter->tx_hwtstamp_skipped++;
8458 		}
8459 	}
8460 
8461 	skb_tx_timestamp(skb);
8462 
8463 #ifdef CONFIG_PCI_IOV
8464 	/*
8465 	 * Use the l2switch_enable flag - would be false if the DMA
8466 	 * Tx switch had been disabled.
8467 	 */
8468 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8469 		tx_flags |= IXGBE_TX_FLAGS_CC;
8470 
8471 #endif
8472 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8473 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8474 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8475 	     (skb->priority != TC_PRIO_CONTROL))) {
8476 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8477 		tx_flags |= (skb->priority & 0x7) <<
8478 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8479 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8480 			struct vlan_ethhdr *vhdr;
8481 
8482 			if (skb_cow_head(skb, 0))
8483 				goto out_drop;
8484 			vhdr = (struct vlan_ethhdr *)skb->data;
8485 			vhdr->h_vlan_TCI = htons(tx_flags >>
8486 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
8487 		} else {
8488 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8489 		}
8490 	}
8491 
8492 	/* record initial flags and protocol */
8493 	first->tx_flags = tx_flags;
8494 	first->protocol = protocol;
8495 
8496 #ifdef IXGBE_FCOE
8497 	/* setup tx offload for FCoE */
8498 	if ((protocol == htons(ETH_P_FCOE)) &&
8499 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8500 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
8501 		if (tso < 0)
8502 			goto out_drop;
8503 
8504 		goto xmit_fcoe;
8505 	}
8506 
8507 #endif /* IXGBE_FCOE */
8508 
8509 #ifdef CONFIG_XFRM_OFFLOAD
8510 	if (skb->sp && !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
8511 		goto out_drop;
8512 #endif
8513 	tso = ixgbe_tso(tx_ring, first, &hdr_len, &ipsec_tx);
8514 	if (tso < 0)
8515 		goto out_drop;
8516 	else if (!tso)
8517 		ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
8518 
8519 	/* add the ATR filter if ATR is on */
8520 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8521 		ixgbe_atr(tx_ring, first);
8522 
8523 #ifdef IXGBE_FCOE
8524 xmit_fcoe:
8525 #endif /* IXGBE_FCOE */
8526 	if (ixgbe_tx_map(tx_ring, first, hdr_len))
8527 		goto cleanup_tx_timestamp;
8528 
8529 	return NETDEV_TX_OK;
8530 
8531 out_drop:
8532 	dev_kfree_skb_any(first->skb);
8533 	first->skb = NULL;
8534 cleanup_tx_timestamp:
8535 	if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8536 		dev_kfree_skb_any(adapter->ptp_tx_skb);
8537 		adapter->ptp_tx_skb = NULL;
8538 		cancel_work_sync(&adapter->ptp_tx_work);
8539 		clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8540 	}
8541 
8542 	return NETDEV_TX_OK;
8543 }
8544 
8545 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8546 				      struct net_device *netdev,
8547 				      struct ixgbe_ring *ring)
8548 {
8549 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8550 	struct ixgbe_ring *tx_ring;
8551 
8552 	/*
8553 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
8554 	 * in order to meet this minimum size requirement.
8555 	 */
8556 	if (skb_put_padto(skb, 17))
8557 		return NETDEV_TX_OK;
8558 
8559 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
8560 
8561 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8562 }
8563 
8564 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8565 				    struct net_device *netdev)
8566 {
8567 	return __ixgbe_xmit_frame(skb, netdev, NULL);
8568 }
8569 
8570 /**
8571  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8572  * @netdev: network interface device structure
8573  * @p: pointer to an address structure
8574  *
8575  * Returns 0 on success, negative on failure
8576  **/
8577 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8578 {
8579 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8580 	struct ixgbe_hw *hw = &adapter->hw;
8581 	struct sockaddr *addr = p;
8582 
8583 	if (!is_valid_ether_addr(addr->sa_data))
8584 		return -EADDRNOTAVAIL;
8585 
8586 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
8587 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8588 
8589 	ixgbe_mac_set_default_filter(adapter);
8590 
8591 	return 0;
8592 }
8593 
8594 static int
8595 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8596 {
8597 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8598 	struct ixgbe_hw *hw = &adapter->hw;
8599 	u16 value;
8600 	int rc;
8601 
8602 	if (prtad != hw->phy.mdio.prtad)
8603 		return -EINVAL;
8604 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8605 	if (!rc)
8606 		rc = value;
8607 	return rc;
8608 }
8609 
8610 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8611 			    u16 addr, u16 value)
8612 {
8613 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8614 	struct ixgbe_hw *hw = &adapter->hw;
8615 
8616 	if (prtad != hw->phy.mdio.prtad)
8617 		return -EINVAL;
8618 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8619 }
8620 
8621 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8622 {
8623 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8624 
8625 	switch (cmd) {
8626 	case SIOCSHWTSTAMP:
8627 		return ixgbe_ptp_set_ts_config(adapter, req);
8628 	case SIOCGHWTSTAMP:
8629 		return ixgbe_ptp_get_ts_config(adapter, req);
8630 	case SIOCGMIIPHY:
8631 		if (!adapter->hw.phy.ops.read_reg)
8632 			return -EOPNOTSUPP;
8633 		/* fall through */
8634 	default:
8635 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8636 	}
8637 }
8638 
8639 /**
8640  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8641  * netdev->dev_addrs
8642  * @dev: network interface device structure
8643  *
8644  * Returns non-zero on failure
8645  **/
8646 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8647 {
8648 	int err = 0;
8649 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8650 	struct ixgbe_hw *hw = &adapter->hw;
8651 
8652 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8653 		rtnl_lock();
8654 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8655 		rtnl_unlock();
8656 
8657 		/* update SAN MAC vmdq pool selection */
8658 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8659 	}
8660 	return err;
8661 }
8662 
8663 /**
8664  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8665  * netdev->dev_addrs
8666  * @dev: network interface device structure
8667  *
8668  * Returns non-zero on failure
8669  **/
8670 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8671 {
8672 	int err = 0;
8673 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8674 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8675 
8676 	if (is_valid_ether_addr(mac->san_addr)) {
8677 		rtnl_lock();
8678 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8679 		rtnl_unlock();
8680 	}
8681 	return err;
8682 }
8683 
8684 #ifdef CONFIG_NET_POLL_CONTROLLER
8685 /*
8686  * Polling 'interrupt' - used by things like netconsole to send skbs
8687  * without having to re-enable interrupts. It's not called while
8688  * the interrupt routine is executing.
8689  */
8690 static void ixgbe_netpoll(struct net_device *netdev)
8691 {
8692 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8693 	int i;
8694 
8695 	/* if interface is down do nothing */
8696 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8697 		return;
8698 
8699 	/* loop through and schedule all active queues */
8700 	for (i = 0; i < adapter->num_q_vectors; i++)
8701 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8702 }
8703 
8704 #endif
8705 
8706 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
8707 				   struct ixgbe_ring *ring)
8708 {
8709 	u64 bytes, packets;
8710 	unsigned int start;
8711 
8712 	if (ring) {
8713 		do {
8714 			start = u64_stats_fetch_begin_irq(&ring->syncp);
8715 			packets = ring->stats.packets;
8716 			bytes   = ring->stats.bytes;
8717 		} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8718 		stats->tx_packets += packets;
8719 		stats->tx_bytes   += bytes;
8720 	}
8721 }
8722 
8723 static void ixgbe_get_stats64(struct net_device *netdev,
8724 			      struct rtnl_link_stats64 *stats)
8725 {
8726 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8727 	int i;
8728 
8729 	rcu_read_lock();
8730 	for (i = 0; i < adapter->num_rx_queues; i++) {
8731 		struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
8732 		u64 bytes, packets;
8733 		unsigned int start;
8734 
8735 		if (ring) {
8736 			do {
8737 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8738 				packets = ring->stats.packets;
8739 				bytes   = ring->stats.bytes;
8740 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8741 			stats->rx_packets += packets;
8742 			stats->rx_bytes   += bytes;
8743 		}
8744 	}
8745 
8746 	for (i = 0; i < adapter->num_tx_queues; i++) {
8747 		struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
8748 
8749 		ixgbe_get_ring_stats64(stats, ring);
8750 	}
8751 	for (i = 0; i < adapter->num_xdp_queues; i++) {
8752 		struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
8753 
8754 		ixgbe_get_ring_stats64(stats, ring);
8755 	}
8756 	rcu_read_unlock();
8757 
8758 	/* following stats updated by ixgbe_watchdog_task() */
8759 	stats->multicast	= netdev->stats.multicast;
8760 	stats->rx_errors	= netdev->stats.rx_errors;
8761 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8762 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8763 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8764 }
8765 
8766 #ifdef CONFIG_IXGBE_DCB
8767 /**
8768  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8769  * @adapter: pointer to ixgbe_adapter
8770  * @tc: number of traffic classes currently enabled
8771  *
8772  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8773  * 802.1Q priority maps to a packet buffer that exists.
8774  */
8775 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8776 {
8777 	struct ixgbe_hw *hw = &adapter->hw;
8778 	u32 reg, rsave;
8779 	int i;
8780 
8781 	/* 82598 have a static priority to TC mapping that can not
8782 	 * be changed so no validation is needed.
8783 	 */
8784 	if (hw->mac.type == ixgbe_mac_82598EB)
8785 		return;
8786 
8787 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8788 	rsave = reg;
8789 
8790 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8791 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8792 
8793 		/* If up2tc is out of bounds default to zero */
8794 		if (up2tc > tc)
8795 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8796 	}
8797 
8798 	if (reg != rsave)
8799 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8800 
8801 	return;
8802 }
8803 
8804 /**
8805  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8806  * @adapter: Pointer to adapter struct
8807  *
8808  * Populate the netdev user priority to tc map
8809  */
8810 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8811 {
8812 	struct net_device *dev = adapter->netdev;
8813 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8814 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8815 	u8 prio;
8816 
8817 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8818 		u8 tc = 0;
8819 
8820 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8821 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8822 		else if (ets)
8823 			tc = ets->prio_tc[prio];
8824 
8825 		netdev_set_prio_tc_map(dev, prio, tc);
8826 	}
8827 }
8828 
8829 #endif /* CONFIG_IXGBE_DCB */
8830 /**
8831  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8832  *
8833  * @dev: net device to configure
8834  * @tc: number of traffic classes to enable
8835  */
8836 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8837 {
8838 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8839 	struct ixgbe_hw *hw = &adapter->hw;
8840 
8841 	/* Hardware supports up to 8 traffic classes */
8842 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8843 		return -EINVAL;
8844 
8845 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8846 		return -EINVAL;
8847 
8848 	/* Hardware has to reinitialize queues and interrupts to
8849 	 * match packet buffer alignment. Unfortunately, the
8850 	 * hardware is not flexible enough to do this dynamically.
8851 	 */
8852 	if (netif_running(dev))
8853 		ixgbe_close(dev);
8854 	else
8855 		ixgbe_reset(adapter);
8856 
8857 	ixgbe_clear_interrupt_scheme(adapter);
8858 
8859 #ifdef CONFIG_IXGBE_DCB
8860 	if (tc) {
8861 		netdev_set_num_tc(dev, tc);
8862 		ixgbe_set_prio_tc_map(adapter);
8863 
8864 		adapter->hw_tcs = tc;
8865 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8866 
8867 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8868 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8869 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8870 		}
8871 	} else {
8872 		netdev_reset_tc(dev);
8873 
8874 		/* To support macvlan offload we have to use num_tc to
8875 		 * restrict the queues that can be used by the device.
8876 		 * By doing this we can avoid reporting a false number of
8877 		 * queues.
8878 		 */
8879 		if (!tc && adapter->num_rx_pools > 1)
8880 			netdev_set_num_tc(dev, 1);
8881 
8882 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8883 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8884 
8885 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8886 		adapter->hw_tcs = tc;
8887 
8888 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8889 		adapter->dcb_cfg.pfc_mode_enable = false;
8890 	}
8891 
8892 	ixgbe_validate_rtr(adapter, tc);
8893 
8894 #endif /* CONFIG_IXGBE_DCB */
8895 	ixgbe_init_interrupt_scheme(adapter);
8896 
8897 	if (netif_running(dev))
8898 		return ixgbe_open(dev);
8899 
8900 	return 0;
8901 }
8902 
8903 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8904 			       struct tc_cls_u32_offload *cls)
8905 {
8906 	u32 hdl = cls->knode.handle;
8907 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8908 	u32 loc = cls->knode.handle & 0xfffff;
8909 	int err = 0, i, j;
8910 	struct ixgbe_jump_table *jump = NULL;
8911 
8912 	if (loc > IXGBE_MAX_HW_ENTRIES)
8913 		return -EINVAL;
8914 
8915 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8916 		return -EINVAL;
8917 
8918 	/* Clear this filter in the link data it is associated with */
8919 	if (uhtid != 0x800) {
8920 		jump = adapter->jump_tables[uhtid];
8921 		if (!jump)
8922 			return -EINVAL;
8923 		if (!test_bit(loc - 1, jump->child_loc_map))
8924 			return -EINVAL;
8925 		clear_bit(loc - 1, jump->child_loc_map);
8926 	}
8927 
8928 	/* Check if the filter being deleted is a link */
8929 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8930 		jump = adapter->jump_tables[i];
8931 		if (jump && jump->link_hdl == hdl) {
8932 			/* Delete filters in the hardware in the child hash
8933 			 * table associated with this link
8934 			 */
8935 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8936 				if (!test_bit(j, jump->child_loc_map))
8937 					continue;
8938 				spin_lock(&adapter->fdir_perfect_lock);
8939 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8940 								      NULL,
8941 								      j + 1);
8942 				spin_unlock(&adapter->fdir_perfect_lock);
8943 				clear_bit(j, jump->child_loc_map);
8944 			}
8945 			/* Remove resources for this link */
8946 			kfree(jump->input);
8947 			kfree(jump->mask);
8948 			kfree(jump);
8949 			adapter->jump_tables[i] = NULL;
8950 			return err;
8951 		}
8952 	}
8953 
8954 	spin_lock(&adapter->fdir_perfect_lock);
8955 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8956 	spin_unlock(&adapter->fdir_perfect_lock);
8957 	return err;
8958 }
8959 
8960 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8961 					    struct tc_cls_u32_offload *cls)
8962 {
8963 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8964 
8965 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8966 		return -EINVAL;
8967 
8968 	/* This ixgbe devices do not support hash tables at the moment
8969 	 * so abort when given hash tables.
8970 	 */
8971 	if (cls->hnode.divisor > 0)
8972 		return -EINVAL;
8973 
8974 	set_bit(uhtid - 1, &adapter->tables);
8975 	return 0;
8976 }
8977 
8978 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
8979 					    struct tc_cls_u32_offload *cls)
8980 {
8981 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8982 
8983 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8984 		return -EINVAL;
8985 
8986 	clear_bit(uhtid - 1, &adapter->tables);
8987 	return 0;
8988 }
8989 
8990 #ifdef CONFIG_NET_CLS_ACT
8991 struct upper_walk_data {
8992 	struct ixgbe_adapter *adapter;
8993 	u64 action;
8994 	int ifindex;
8995 	u8 queue;
8996 };
8997 
8998 static int get_macvlan_queue(struct net_device *upper, void *_data)
8999 {
9000 	if (netif_is_macvlan(upper)) {
9001 		struct macvlan_dev *dfwd = netdev_priv(upper);
9002 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
9003 		struct upper_walk_data *data = _data;
9004 		struct ixgbe_adapter *adapter = data->adapter;
9005 		int ifindex = data->ifindex;
9006 
9007 		if (vadapter && vadapter->netdev->ifindex == ifindex) {
9008 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
9009 			data->action = data->queue;
9010 			return 1;
9011 		}
9012 	}
9013 
9014 	return 0;
9015 }
9016 
9017 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
9018 				  u8 *queue, u64 *action)
9019 {
9020 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
9021 	unsigned int num_vfs = adapter->num_vfs, vf;
9022 	struct upper_walk_data data;
9023 	struct net_device *upper;
9024 
9025 	/* redirect to a SRIOV VF */
9026 	for (vf = 0; vf < num_vfs; ++vf) {
9027 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
9028 		if (upper->ifindex == ifindex) {
9029 			*queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
9030 			*action = vf + 1;
9031 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
9032 			return 0;
9033 		}
9034 	}
9035 
9036 	/* redirect to a offloaded macvlan netdev */
9037 	data.adapter = adapter;
9038 	data.ifindex = ifindex;
9039 	data.action = 0;
9040 	data.queue = 0;
9041 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
9042 					  get_macvlan_queue, &data)) {
9043 		*action = data.action;
9044 		*queue = data.queue;
9045 
9046 		return 0;
9047 	}
9048 
9049 	return -EINVAL;
9050 }
9051 
9052 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9053 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9054 {
9055 	const struct tc_action *a;
9056 	LIST_HEAD(actions);
9057 	int err;
9058 
9059 	if (!tcf_exts_has_actions(exts))
9060 		return -EINVAL;
9061 
9062 	tcf_exts_to_list(exts, &actions);
9063 	list_for_each_entry(a, &actions, list) {
9064 
9065 		/* Drop action */
9066 		if (is_tcf_gact_shot(a)) {
9067 			*action = IXGBE_FDIR_DROP_QUEUE;
9068 			*queue = IXGBE_FDIR_DROP_QUEUE;
9069 			return 0;
9070 		}
9071 
9072 		/* Redirect to a VF or a offloaded macvlan */
9073 		if (is_tcf_mirred_egress_redirect(a)) {
9074 			struct net_device *dev = tcf_mirred_dev(a);
9075 
9076 			if (!dev)
9077 				return -EINVAL;
9078 			err = handle_redirect_action(adapter, dev->ifindex, queue,
9079 						     action);
9080 			if (err == 0)
9081 				return err;
9082 		}
9083 	}
9084 
9085 	return -EINVAL;
9086 }
9087 #else
9088 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9089 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9090 {
9091 	return -EINVAL;
9092 }
9093 #endif /* CONFIG_NET_CLS_ACT */
9094 
9095 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
9096 				    union ixgbe_atr_input *mask,
9097 				    struct tc_cls_u32_offload *cls,
9098 				    struct ixgbe_mat_field *field_ptr,
9099 				    struct ixgbe_nexthdr *nexthdr)
9100 {
9101 	int i, j, off;
9102 	__be32 val, m;
9103 	bool found_entry = false, found_jump_field = false;
9104 
9105 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
9106 		off = cls->knode.sel->keys[i].off;
9107 		val = cls->knode.sel->keys[i].val;
9108 		m = cls->knode.sel->keys[i].mask;
9109 
9110 		for (j = 0; field_ptr[j].val; j++) {
9111 			if (field_ptr[j].off == off) {
9112 				field_ptr[j].val(input, mask, val, m);
9113 				input->filter.formatted.flow_type |=
9114 					field_ptr[j].type;
9115 				found_entry = true;
9116 				break;
9117 			}
9118 		}
9119 		if (nexthdr) {
9120 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
9121 			    nexthdr->val == cls->knode.sel->keys[i].val &&
9122 			    nexthdr->mask == cls->knode.sel->keys[i].mask)
9123 				found_jump_field = true;
9124 			else
9125 				continue;
9126 		}
9127 	}
9128 
9129 	if (nexthdr && !found_jump_field)
9130 		return -EINVAL;
9131 
9132 	if (!found_entry)
9133 		return 0;
9134 
9135 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9136 				    IXGBE_ATR_L4TYPE_MASK;
9137 
9138 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9139 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9140 
9141 	return 0;
9142 }
9143 
9144 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9145 				  struct tc_cls_u32_offload *cls)
9146 {
9147 	__be16 protocol = cls->common.protocol;
9148 	u32 loc = cls->knode.handle & 0xfffff;
9149 	struct ixgbe_hw *hw = &adapter->hw;
9150 	struct ixgbe_mat_field *field_ptr;
9151 	struct ixgbe_fdir_filter *input = NULL;
9152 	union ixgbe_atr_input *mask = NULL;
9153 	struct ixgbe_jump_table *jump = NULL;
9154 	int i, err = -EINVAL;
9155 	u8 queue;
9156 	u32 uhtid, link_uhtid;
9157 
9158 	uhtid = TC_U32_USERHTID(cls->knode.handle);
9159 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9160 
9161 	/* At the moment cls_u32 jumps to network layer and skips past
9162 	 * L2 headers. The canonical method to match L2 frames is to use
9163 	 * negative values. However this is error prone at best but really
9164 	 * just broken because there is no way to "know" what sort of hdr
9165 	 * is in front of the network layer. Fix cls_u32 to support L2
9166 	 * headers when needed.
9167 	 */
9168 	if (protocol != htons(ETH_P_IP))
9169 		return err;
9170 
9171 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9172 		e_err(drv, "Location out of range\n");
9173 		return err;
9174 	}
9175 
9176 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
9177 	 * links and also the fields used to advance the parser across each
9178 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9179 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9180 	 * To add support for new nodes update ixgbe_model.h parse structures
9181 	 * this function _should_ be generic try not to hardcode values here.
9182 	 */
9183 	if (uhtid == 0x800) {
9184 		field_ptr = (adapter->jump_tables[0])->mat;
9185 	} else {
9186 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9187 			return err;
9188 		if (!adapter->jump_tables[uhtid])
9189 			return err;
9190 		field_ptr = (adapter->jump_tables[uhtid])->mat;
9191 	}
9192 
9193 	if (!field_ptr)
9194 		return err;
9195 
9196 	/* At this point we know the field_ptr is valid and need to either
9197 	 * build cls_u32 link or attach filter. Because adding a link to
9198 	 * a handle that does not exist is invalid and the same for adding
9199 	 * rules to handles that don't exist.
9200 	 */
9201 
9202 	if (link_uhtid) {
9203 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9204 
9205 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9206 			return err;
9207 
9208 		if (!test_bit(link_uhtid - 1, &adapter->tables))
9209 			return err;
9210 
9211 		/* Multiple filters as links to the same hash table are not
9212 		 * supported. To add a new filter with the same next header
9213 		 * but different match/jump conditions, create a new hash table
9214 		 * and link to it.
9215 		 */
9216 		if (adapter->jump_tables[link_uhtid] &&
9217 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
9218 			e_err(drv, "Link filter exists for link: %x\n",
9219 			      link_uhtid);
9220 			return err;
9221 		}
9222 
9223 		for (i = 0; nexthdr[i].jump; i++) {
9224 			if (nexthdr[i].o != cls->knode.sel->offoff ||
9225 			    nexthdr[i].s != cls->knode.sel->offshift ||
9226 			    nexthdr[i].m != cls->knode.sel->offmask)
9227 				return err;
9228 
9229 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9230 			if (!jump)
9231 				return -ENOMEM;
9232 			input = kzalloc(sizeof(*input), GFP_KERNEL);
9233 			if (!input) {
9234 				err = -ENOMEM;
9235 				goto free_jump;
9236 			}
9237 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9238 			if (!mask) {
9239 				err = -ENOMEM;
9240 				goto free_input;
9241 			}
9242 			jump->input = input;
9243 			jump->mask = mask;
9244 			jump->link_hdl = cls->knode.handle;
9245 
9246 			err = ixgbe_clsu32_build_input(input, mask, cls,
9247 						       field_ptr, &nexthdr[i]);
9248 			if (!err) {
9249 				jump->mat = nexthdr[i].jump;
9250 				adapter->jump_tables[link_uhtid] = jump;
9251 				break;
9252 			}
9253 		}
9254 		return 0;
9255 	}
9256 
9257 	input = kzalloc(sizeof(*input), GFP_KERNEL);
9258 	if (!input)
9259 		return -ENOMEM;
9260 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9261 	if (!mask) {
9262 		err = -ENOMEM;
9263 		goto free_input;
9264 	}
9265 
9266 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9267 		if ((adapter->jump_tables[uhtid])->input)
9268 			memcpy(input, (adapter->jump_tables[uhtid])->input,
9269 			       sizeof(*input));
9270 		if ((adapter->jump_tables[uhtid])->mask)
9271 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9272 			       sizeof(*mask));
9273 
9274 		/* Lookup in all child hash tables if this location is already
9275 		 * filled with a filter
9276 		 */
9277 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9278 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
9279 
9280 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
9281 				e_err(drv, "Filter exists in location: %x\n",
9282 				      loc);
9283 				err = -EINVAL;
9284 				goto err_out;
9285 			}
9286 		}
9287 	}
9288 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9289 	if (err)
9290 		goto err_out;
9291 
9292 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9293 			       &queue);
9294 	if (err < 0)
9295 		goto err_out;
9296 
9297 	input->sw_idx = loc;
9298 
9299 	spin_lock(&adapter->fdir_perfect_lock);
9300 
9301 	if (hlist_empty(&adapter->fdir_filter_list)) {
9302 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9303 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9304 		if (err)
9305 			goto err_out_w_lock;
9306 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9307 		err = -EINVAL;
9308 		goto err_out_w_lock;
9309 	}
9310 
9311 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9312 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9313 						    input->sw_idx, queue);
9314 	if (!err)
9315 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9316 	spin_unlock(&adapter->fdir_perfect_lock);
9317 
9318 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9319 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9320 
9321 	kfree(mask);
9322 	return err;
9323 err_out_w_lock:
9324 	spin_unlock(&adapter->fdir_perfect_lock);
9325 err_out:
9326 	kfree(mask);
9327 free_input:
9328 	kfree(input);
9329 free_jump:
9330 	kfree(jump);
9331 	return err;
9332 }
9333 
9334 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
9335 				  struct tc_cls_u32_offload *cls_u32)
9336 {
9337 	switch (cls_u32->command) {
9338 	case TC_CLSU32_NEW_KNODE:
9339 	case TC_CLSU32_REPLACE_KNODE:
9340 		return ixgbe_configure_clsu32(adapter, cls_u32);
9341 	case TC_CLSU32_DELETE_KNODE:
9342 		return ixgbe_delete_clsu32(adapter, cls_u32);
9343 	case TC_CLSU32_NEW_HNODE:
9344 	case TC_CLSU32_REPLACE_HNODE:
9345 		return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9346 	case TC_CLSU32_DELETE_HNODE:
9347 		return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9348 	default:
9349 		return -EOPNOTSUPP;
9350 	}
9351 }
9352 
9353 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
9354 				   void *cb_priv)
9355 {
9356 	struct ixgbe_adapter *adapter = cb_priv;
9357 
9358 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
9359 		return -EOPNOTSUPP;
9360 
9361 	switch (type) {
9362 	case TC_SETUP_CLSU32:
9363 		return ixgbe_setup_tc_cls_u32(adapter, type_data);
9364 	default:
9365 		return -EOPNOTSUPP;
9366 	}
9367 }
9368 
9369 static int ixgbe_setup_tc_block(struct net_device *dev,
9370 				struct tc_block_offload *f)
9371 {
9372 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9373 
9374 	if (f->binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
9375 		return -EOPNOTSUPP;
9376 
9377 	switch (f->command) {
9378 	case TC_BLOCK_BIND:
9379 		return tcf_block_cb_register(f->block, ixgbe_setup_tc_block_cb,
9380 					     adapter, adapter);
9381 	case TC_BLOCK_UNBIND:
9382 		tcf_block_cb_unregister(f->block, ixgbe_setup_tc_block_cb,
9383 					adapter);
9384 		return 0;
9385 	default:
9386 		return -EOPNOTSUPP;
9387 	}
9388 }
9389 
9390 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9391 				 struct tc_mqprio_qopt *mqprio)
9392 {
9393 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9394 	return ixgbe_setup_tc(dev, mqprio->num_tc);
9395 }
9396 
9397 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9398 			    void *type_data)
9399 {
9400 	switch (type) {
9401 	case TC_SETUP_BLOCK:
9402 		return ixgbe_setup_tc_block(dev, type_data);
9403 	case TC_SETUP_QDISC_MQPRIO:
9404 		return ixgbe_setup_tc_mqprio(dev, type_data);
9405 	default:
9406 		return -EOPNOTSUPP;
9407 	}
9408 }
9409 
9410 #ifdef CONFIG_PCI_IOV
9411 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9412 {
9413 	struct net_device *netdev = adapter->netdev;
9414 
9415 	rtnl_lock();
9416 	ixgbe_setup_tc(netdev, adapter->hw_tcs);
9417 	rtnl_unlock();
9418 }
9419 
9420 #endif
9421 void ixgbe_do_reset(struct net_device *netdev)
9422 {
9423 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9424 
9425 	if (netif_running(netdev))
9426 		ixgbe_reinit_locked(adapter);
9427 	else
9428 		ixgbe_reset(adapter);
9429 }
9430 
9431 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9432 					    netdev_features_t features)
9433 {
9434 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9435 
9436 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9437 	if (!(features & NETIF_F_RXCSUM))
9438 		features &= ~NETIF_F_LRO;
9439 
9440 	/* Turn off LRO if not RSC capable */
9441 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9442 		features &= ~NETIF_F_LRO;
9443 
9444 	return features;
9445 }
9446 
9447 static int ixgbe_set_features(struct net_device *netdev,
9448 			      netdev_features_t features)
9449 {
9450 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9451 	netdev_features_t changed = netdev->features ^ features;
9452 	bool need_reset = false;
9453 
9454 	/* Make sure RSC matches LRO, reset if change */
9455 	if (!(features & NETIF_F_LRO)) {
9456 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9457 			need_reset = true;
9458 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9459 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9460 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9461 		if (adapter->rx_itr_setting == 1 ||
9462 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9463 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9464 			need_reset = true;
9465 		} else if ((changed ^ features) & NETIF_F_LRO) {
9466 			e_info(probe, "rx-usecs set too low, "
9467 			       "disabling RSC\n");
9468 		}
9469 	}
9470 
9471 	/*
9472 	 * Check if Flow Director n-tuple support or hw_tc support was
9473 	 * enabled or disabled.  If the state changed, we need to reset.
9474 	 */
9475 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9476 		/* turn off ATR, enable perfect filters and reset */
9477 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9478 			need_reset = true;
9479 
9480 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9481 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9482 	} else {
9483 		/* turn off perfect filters, enable ATR and reset */
9484 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9485 			need_reset = true;
9486 
9487 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9488 
9489 		/* We cannot enable ATR if SR-IOV is enabled */
9490 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9491 		    /* We cannot enable ATR if we have 2 or more tcs */
9492 		    (adapter->hw_tcs > 1) ||
9493 		    /* We cannot enable ATR if RSS is disabled */
9494 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9495 		    /* A sample rate of 0 indicates ATR disabled */
9496 		    (!adapter->atr_sample_rate))
9497 			; /* do nothing not supported */
9498 		else /* otherwise supported and set the flag */
9499 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9500 	}
9501 
9502 	if (changed & NETIF_F_RXALL)
9503 		need_reset = true;
9504 
9505 	netdev->features = features;
9506 
9507 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
9508 		if (features & NETIF_F_RXCSUM) {
9509 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9510 		} else {
9511 			u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9512 
9513 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9514 		}
9515 	}
9516 
9517 	if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
9518 		if (features & NETIF_F_RXCSUM) {
9519 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9520 		} else {
9521 			u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9522 
9523 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9524 		}
9525 	}
9526 
9527 	if (need_reset)
9528 		ixgbe_do_reset(netdev);
9529 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9530 			    NETIF_F_HW_VLAN_CTAG_FILTER))
9531 		ixgbe_set_rx_mode(netdev);
9532 
9533 	return 0;
9534 }
9535 
9536 /**
9537  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
9538  * @dev: The port's netdev
9539  * @ti: Tunnel endpoint information
9540  **/
9541 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
9542 				      struct udp_tunnel_info *ti)
9543 {
9544 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9545 	struct ixgbe_hw *hw = &adapter->hw;
9546 	__be16 port = ti->port;
9547 	u32 port_shift = 0;
9548 	u32 reg;
9549 
9550 	if (ti->sa_family != AF_INET)
9551 		return;
9552 
9553 	switch (ti->type) {
9554 	case UDP_TUNNEL_TYPE_VXLAN:
9555 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9556 			return;
9557 
9558 		if (adapter->vxlan_port == port)
9559 			return;
9560 
9561 		if (adapter->vxlan_port) {
9562 			netdev_info(dev,
9563 				    "VXLAN port %d set, not adding port %d\n",
9564 				    ntohs(adapter->vxlan_port),
9565 				    ntohs(port));
9566 			return;
9567 		}
9568 
9569 		adapter->vxlan_port = port;
9570 		break;
9571 	case UDP_TUNNEL_TYPE_GENEVE:
9572 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9573 			return;
9574 
9575 		if (adapter->geneve_port == port)
9576 			return;
9577 
9578 		if (adapter->geneve_port) {
9579 			netdev_info(dev,
9580 				    "GENEVE port %d set, not adding port %d\n",
9581 				    ntohs(adapter->geneve_port),
9582 				    ntohs(port));
9583 			return;
9584 		}
9585 
9586 		port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
9587 		adapter->geneve_port = port;
9588 		break;
9589 	default:
9590 		return;
9591 	}
9592 
9593 	reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
9594 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
9595 }
9596 
9597 /**
9598  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
9599  * @dev: The port's netdev
9600  * @ti: Tunnel endpoint information
9601  **/
9602 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
9603 				      struct udp_tunnel_info *ti)
9604 {
9605 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9606 	u32 port_mask;
9607 
9608 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
9609 	    ti->type != UDP_TUNNEL_TYPE_GENEVE)
9610 		return;
9611 
9612 	if (ti->sa_family != AF_INET)
9613 		return;
9614 
9615 	switch (ti->type) {
9616 	case UDP_TUNNEL_TYPE_VXLAN:
9617 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9618 			return;
9619 
9620 		if (adapter->vxlan_port != ti->port) {
9621 			netdev_info(dev, "VXLAN port %d not found\n",
9622 				    ntohs(ti->port));
9623 			return;
9624 		}
9625 
9626 		port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9627 		break;
9628 	case UDP_TUNNEL_TYPE_GENEVE:
9629 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9630 			return;
9631 
9632 		if (adapter->geneve_port != ti->port) {
9633 			netdev_info(dev, "GENEVE port %d not found\n",
9634 				    ntohs(ti->port));
9635 			return;
9636 		}
9637 
9638 		port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9639 		break;
9640 	default:
9641 		return;
9642 	}
9643 
9644 	ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9645 	adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9646 }
9647 
9648 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9649 			     struct net_device *dev,
9650 			     const unsigned char *addr, u16 vid,
9651 			     u16 flags)
9652 {
9653 	/* guarantee we can provide a unique filter for the unicast address */
9654 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9655 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9656 		u16 pool = VMDQ_P(0);
9657 
9658 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9659 			return -ENOMEM;
9660 	}
9661 
9662 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9663 }
9664 
9665 /**
9666  * ixgbe_configure_bridge_mode - set various bridge modes
9667  * @adapter: the private structure
9668  * @mode: requested bridge mode
9669  *
9670  * Configure some settings require for various bridge modes.
9671  **/
9672 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9673 				       __u16 mode)
9674 {
9675 	struct ixgbe_hw *hw = &adapter->hw;
9676 	unsigned int p, num_pools;
9677 	u32 vmdctl;
9678 
9679 	switch (mode) {
9680 	case BRIDGE_MODE_VEPA:
9681 		/* disable Tx loopback, rely on switch hairpin mode */
9682 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9683 
9684 		/* must enable Rx switching replication to allow multicast
9685 		 * packet reception on all VFs, and to enable source address
9686 		 * pruning.
9687 		 */
9688 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9689 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9690 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9691 
9692 		/* enable Rx source address pruning. Note, this requires
9693 		 * replication to be enabled or else it does nothing.
9694 		 */
9695 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9696 		for (p = 0; p < num_pools; p++) {
9697 			if (hw->mac.ops.set_source_address_pruning)
9698 				hw->mac.ops.set_source_address_pruning(hw,
9699 								       true,
9700 								       p);
9701 		}
9702 		break;
9703 	case BRIDGE_MODE_VEB:
9704 		/* enable Tx loopback for internal VF/PF communication */
9705 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9706 				IXGBE_PFDTXGSWC_VT_LBEN);
9707 
9708 		/* disable Rx switching replication unless we have SR-IOV
9709 		 * virtual functions
9710 		 */
9711 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9712 		if (!adapter->num_vfs)
9713 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9714 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9715 
9716 		/* disable Rx source address pruning, since we don't expect to
9717 		 * be receiving external loopback of our transmitted frames.
9718 		 */
9719 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9720 		for (p = 0; p < num_pools; p++) {
9721 			if (hw->mac.ops.set_source_address_pruning)
9722 				hw->mac.ops.set_source_address_pruning(hw,
9723 								       false,
9724 								       p);
9725 		}
9726 		break;
9727 	default:
9728 		return -EINVAL;
9729 	}
9730 
9731 	adapter->bridge_mode = mode;
9732 
9733 	e_info(drv, "enabling bridge mode: %s\n",
9734 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9735 
9736 	return 0;
9737 }
9738 
9739 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9740 				    struct nlmsghdr *nlh, u16 flags)
9741 {
9742 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9743 	struct nlattr *attr, *br_spec;
9744 	int rem;
9745 
9746 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9747 		return -EOPNOTSUPP;
9748 
9749 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9750 	if (!br_spec)
9751 		return -EINVAL;
9752 
9753 	nla_for_each_nested(attr, br_spec, rem) {
9754 		int status;
9755 		__u16 mode;
9756 
9757 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9758 			continue;
9759 
9760 		if (nla_len(attr) < sizeof(mode))
9761 			return -EINVAL;
9762 
9763 		mode = nla_get_u16(attr);
9764 		status = ixgbe_configure_bridge_mode(adapter, mode);
9765 		if (status)
9766 			return status;
9767 
9768 		break;
9769 	}
9770 
9771 	return 0;
9772 }
9773 
9774 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9775 				    struct net_device *dev,
9776 				    u32 filter_mask, int nlflags)
9777 {
9778 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9779 
9780 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9781 		return 0;
9782 
9783 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9784 				       adapter->bridge_mode, 0, 0, nlflags,
9785 				       filter_mask, NULL);
9786 }
9787 
9788 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9789 {
9790 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
9791 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9792 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
9793 	int tcs = adapter->hw_tcs ? : 1;
9794 	unsigned int limit;
9795 	int pool, err;
9796 
9797 	/* Hardware has a limited number of available pools. Each VF, and the
9798 	 * PF require a pool. Check to ensure we don't attempt to use more
9799 	 * then the available number of pools.
9800 	 */
9801 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9802 		return ERR_PTR(-EINVAL);
9803 
9804 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9805 	      adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
9806 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
9807 		return ERR_PTR(-EBUSY);
9808 
9809 	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
9810 	if (!fwd_adapter)
9811 		return ERR_PTR(-ENOMEM);
9812 
9813 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9814 	set_bit(pool, adapter->fwd_bitmask);
9815 	limit = find_last_bit(adapter->fwd_bitmask, adapter->num_rx_pools + 1);
9816 
9817 	/* Enable VMDq flag so device will be set in VM mode */
9818 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
9819 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9820 
9821 	fwd_adapter->pool = pool;
9822 	fwd_adapter->real_adapter = adapter;
9823 
9824 	/* Force reinit of ring allocation with VMDQ enabled */
9825 	err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
9826 
9827 	if (!err && netif_running(pdev))
9828 		err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
9829 
9830 	if (!err)
9831 		return fwd_adapter;
9832 
9833 	/* unwind counter and free adapter struct */
9834 	netdev_info(pdev,
9835 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
9836 	clear_bit(pool, adapter->fwd_bitmask);
9837 	kfree(fwd_adapter);
9838 	return ERR_PTR(err);
9839 }
9840 
9841 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9842 {
9843 	struct ixgbe_fwd_adapter *accel = priv;
9844 	struct ixgbe_adapter *adapter = accel->real_adapter;
9845 	unsigned int rxbase = accel->rx_base_queue;
9846 	unsigned int limit, i;
9847 
9848 	/* delete unicast filter associated with offloaded interface */
9849 	ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
9850 			     VMDQ_P(accel->pool));
9851 
9852 	/* disable ability to receive packets for this pool */
9853 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_VMOLR(accel->pool), 0);
9854 
9855 	/* Allow remaining Rx packets to get flushed out of the
9856 	 * Rx FIFO before we drop the netdev for the ring.
9857 	 */
9858 	usleep_range(10000, 20000);
9859 
9860 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
9861 		struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
9862 		struct ixgbe_q_vector *qv = ring->q_vector;
9863 
9864 		/* Make sure we aren't processing any packets and clear
9865 		 * netdev to shut down the ring.
9866 		 */
9867 		if (netif_running(adapter->netdev))
9868 			napi_synchronize(&qv->napi);
9869 		ring->netdev = NULL;
9870 	}
9871 
9872 	clear_bit(accel->pool, adapter->fwd_bitmask);
9873 	limit = find_last_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9874 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9875 
9876 	/* go back to full RSS if we're done with our VMQs */
9877 	if (adapter->ring_feature[RING_F_VMDQ].limit == 1) {
9878 		int rss = min_t(int, ixgbe_max_rss_indices(adapter),
9879 				num_online_cpus());
9880 
9881 		adapter->flags &= ~IXGBE_FLAG_VMDQ_ENABLED;
9882 		adapter->flags &= ~IXGBE_FLAG_SRIOV_ENABLED;
9883 		adapter->ring_feature[RING_F_RSS].limit = rss;
9884 	}
9885 
9886 	ixgbe_setup_tc(pdev, adapter->hw_tcs);
9887 	netdev_dbg(pdev, "pool %i:%i queues %i:%i\n",
9888 		   accel->pool, adapter->num_rx_pools,
9889 		   accel->rx_base_queue,
9890 		   accel->rx_base_queue +
9891 		   adapter->num_rx_queues_per_pool);
9892 	kfree(accel);
9893 }
9894 
9895 #define IXGBE_MAX_MAC_HDR_LEN		127
9896 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9897 
9898 static netdev_features_t
9899 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9900 		     netdev_features_t features)
9901 {
9902 	unsigned int network_hdr_len, mac_hdr_len;
9903 
9904 	/* Make certain the headers can be described by a context descriptor */
9905 	mac_hdr_len = skb_network_header(skb) - skb->data;
9906 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9907 		return features & ~(NETIF_F_HW_CSUM |
9908 				    NETIF_F_SCTP_CRC |
9909 				    NETIF_F_HW_VLAN_CTAG_TX |
9910 				    NETIF_F_TSO |
9911 				    NETIF_F_TSO6);
9912 
9913 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9914 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9915 		return features & ~(NETIF_F_HW_CSUM |
9916 				    NETIF_F_SCTP_CRC |
9917 				    NETIF_F_TSO |
9918 				    NETIF_F_TSO6);
9919 
9920 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9921 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9922 	 * IPsec offoad sets skb->encapsulation but still can handle
9923 	 * the TSO, so it's the exception.
9924 	 */
9925 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID)) {
9926 #ifdef CONFIG_XFRM
9927 		if (!skb->sp)
9928 #endif
9929 			features &= ~NETIF_F_TSO;
9930 	}
9931 
9932 	return features;
9933 }
9934 
9935 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
9936 {
9937 	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
9938 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9939 	struct bpf_prog *old_prog;
9940 
9941 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
9942 		return -EINVAL;
9943 
9944 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
9945 		return -EINVAL;
9946 
9947 	/* verify ixgbe ring attributes are sufficient for XDP */
9948 	for (i = 0; i < adapter->num_rx_queues; i++) {
9949 		struct ixgbe_ring *ring = adapter->rx_ring[i];
9950 
9951 		if (ring_is_rsc_enabled(ring))
9952 			return -EINVAL;
9953 
9954 		if (frame_size > ixgbe_rx_bufsz(ring))
9955 			return -EINVAL;
9956 	}
9957 
9958 	if (nr_cpu_ids > MAX_XDP_QUEUES)
9959 		return -ENOMEM;
9960 
9961 	old_prog = xchg(&adapter->xdp_prog, prog);
9962 
9963 	/* If transitioning XDP modes reconfigure rings */
9964 	if (!!prog != !!old_prog) {
9965 		int err = ixgbe_setup_tc(dev, adapter->hw_tcs);
9966 
9967 		if (err) {
9968 			rcu_assign_pointer(adapter->xdp_prog, old_prog);
9969 			return -EINVAL;
9970 		}
9971 	} else {
9972 		for (i = 0; i < adapter->num_rx_queues; i++)
9973 			xchg(&adapter->rx_ring[i]->xdp_prog, adapter->xdp_prog);
9974 	}
9975 
9976 	if (old_prog)
9977 		bpf_prog_put(old_prog);
9978 
9979 	return 0;
9980 }
9981 
9982 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
9983 {
9984 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9985 
9986 	switch (xdp->command) {
9987 	case XDP_SETUP_PROG:
9988 		return ixgbe_xdp_setup(dev, xdp->prog);
9989 	case XDP_QUERY_PROG:
9990 		xdp->prog_attached = !!(adapter->xdp_prog);
9991 		xdp->prog_id = adapter->xdp_prog ?
9992 			adapter->xdp_prog->aux->id : 0;
9993 		return 0;
9994 	default:
9995 		return -EINVAL;
9996 	}
9997 }
9998 
9999 static int ixgbe_xdp_xmit(struct net_device *dev, struct xdp_buff *xdp)
10000 {
10001 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10002 	struct ixgbe_ring *ring;
10003 	int err;
10004 
10005 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10006 		return -ENETDOWN;
10007 
10008 	/* During program transitions its possible adapter->xdp_prog is assigned
10009 	 * but ring has not been configured yet. In this case simply abort xmit.
10010 	 */
10011 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
10012 	if (unlikely(!ring))
10013 		return -ENXIO;
10014 
10015 	err = ixgbe_xmit_xdp_ring(adapter, xdp);
10016 	if (err != IXGBE_XDP_TX)
10017 		return -ENOSPC;
10018 
10019 	return 0;
10020 }
10021 
10022 static void ixgbe_xdp_flush(struct net_device *dev)
10023 {
10024 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10025 	struct ixgbe_ring *ring;
10026 
10027 	/* Its possible the device went down between xdp xmit and flush so
10028 	 * we need to ensure device is still up.
10029 	 */
10030 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10031 		return;
10032 
10033 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
10034 	if (unlikely(!ring))
10035 		return;
10036 
10037 	/* Force memory writes to complete before letting h/w know there
10038 	 * are new descriptors to fetch.
10039 	 */
10040 	wmb();
10041 	writel(ring->next_to_use, ring->tail);
10042 
10043 	return;
10044 }
10045 
10046 static const struct net_device_ops ixgbe_netdev_ops = {
10047 	.ndo_open		= ixgbe_open,
10048 	.ndo_stop		= ixgbe_close,
10049 	.ndo_start_xmit		= ixgbe_xmit_frame,
10050 	.ndo_select_queue	= ixgbe_select_queue,
10051 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
10052 	.ndo_validate_addr	= eth_validate_addr,
10053 	.ndo_set_mac_address	= ixgbe_set_mac,
10054 	.ndo_change_mtu		= ixgbe_change_mtu,
10055 	.ndo_tx_timeout		= ixgbe_tx_timeout,
10056 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
10057 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
10058 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
10059 	.ndo_do_ioctl		= ixgbe_ioctl,
10060 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
10061 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
10062 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
10063 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
10064 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
10065 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
10066 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
10067 	.ndo_get_stats64	= ixgbe_get_stats64,
10068 	.ndo_setup_tc		= __ixgbe_setup_tc,
10069 #ifdef CONFIG_NET_POLL_CONTROLLER
10070 	.ndo_poll_controller	= ixgbe_netpoll,
10071 #endif
10072 #ifdef IXGBE_FCOE
10073 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
10074 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
10075 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
10076 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
10077 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
10078 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
10079 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
10080 #endif /* IXGBE_FCOE */
10081 	.ndo_set_features = ixgbe_set_features,
10082 	.ndo_fix_features = ixgbe_fix_features,
10083 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
10084 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
10085 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
10086 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
10087 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
10088 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
10089 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
10090 	.ndo_features_check	= ixgbe_features_check,
10091 	.ndo_bpf		= ixgbe_xdp,
10092 	.ndo_xdp_xmit		= ixgbe_xdp_xmit,
10093 	.ndo_xdp_flush		= ixgbe_xdp_flush,
10094 };
10095 
10096 /**
10097  * ixgbe_enumerate_functions - Get the number of ports this device has
10098  * @adapter: adapter structure
10099  *
10100  * This function enumerates the phsyical functions co-located on a single slot,
10101  * in order to determine how many ports a device has. This is most useful in
10102  * determining the required GT/s of PCIe bandwidth necessary for optimal
10103  * performance.
10104  **/
10105 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
10106 {
10107 	struct pci_dev *entry, *pdev = adapter->pdev;
10108 	int physfns = 0;
10109 
10110 	/* Some cards can not use the generic count PCIe functions method,
10111 	 * because they are behind a parent switch, so we hardcode these with
10112 	 * the correct number of functions.
10113 	 */
10114 	if (ixgbe_pcie_from_parent(&adapter->hw))
10115 		physfns = 4;
10116 
10117 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
10118 		/* don't count virtual functions */
10119 		if (entry->is_virtfn)
10120 			continue;
10121 
10122 		/* When the devices on the bus don't all match our device ID,
10123 		 * we can't reliably determine the correct number of
10124 		 * functions. This can occur if a function has been direct
10125 		 * attached to a virtual machine using VT-d, for example. In
10126 		 * this case, simply return -1 to indicate this.
10127 		 */
10128 		if ((entry->vendor != pdev->vendor) ||
10129 		    (entry->device != pdev->device))
10130 			return -1;
10131 
10132 		physfns++;
10133 	}
10134 
10135 	return physfns;
10136 }
10137 
10138 /**
10139  * ixgbe_wol_supported - Check whether device supports WoL
10140  * @adapter: the adapter private structure
10141  * @device_id: the device ID
10142  * @subdevice_id: the subsystem device ID
10143  *
10144  * This function is used by probe and ethtool to determine
10145  * which devices have WoL support
10146  *
10147  **/
10148 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
10149 			 u16 subdevice_id)
10150 {
10151 	struct ixgbe_hw *hw = &adapter->hw;
10152 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
10153 
10154 	/* WOL not supported on 82598 */
10155 	if (hw->mac.type == ixgbe_mac_82598EB)
10156 		return false;
10157 
10158 	/* check eeprom to see if WOL is enabled for X540 and newer */
10159 	if (hw->mac.type >= ixgbe_mac_X540) {
10160 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10161 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10162 		     (hw->bus.func == 0)))
10163 			return true;
10164 	}
10165 
10166 	/* WOL is determined based on device IDs for 82599 MACs */
10167 	switch (device_id) {
10168 	case IXGBE_DEV_ID_82599_SFP:
10169 		/* Only these subdevices could supports WOL */
10170 		switch (subdevice_id) {
10171 		case IXGBE_SUBDEV_ID_82599_560FLR:
10172 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10173 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10174 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10175 			/* only support first port */
10176 			if (hw->bus.func != 0)
10177 				break;
10178 			/* fall through */
10179 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10180 		case IXGBE_SUBDEV_ID_82599_SFP:
10181 		case IXGBE_SUBDEV_ID_82599_RNDC:
10182 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10183 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10184 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10185 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10186 			return true;
10187 		}
10188 		break;
10189 	case IXGBE_DEV_ID_82599EN_SFP:
10190 		/* Only these subdevices support WOL */
10191 		switch (subdevice_id) {
10192 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10193 			return true;
10194 		}
10195 		break;
10196 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10197 		/* All except this subdevice support WOL */
10198 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10199 			return true;
10200 		break;
10201 	case IXGBE_DEV_ID_82599_KX4:
10202 		return  true;
10203 	default:
10204 		break;
10205 	}
10206 
10207 	return false;
10208 }
10209 
10210 /**
10211  * ixgbe_set_fw_version - Set FW version
10212  * @adapter: the adapter private structure
10213  *
10214  * This function is used by probe and ethtool to determine the FW version to
10215  * format to display. The FW version is taken from the EEPROM/NVM.
10216  */
10217 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
10218 {
10219 	struct ixgbe_hw *hw = &adapter->hw;
10220 	struct ixgbe_nvm_version nvm_ver;
10221 
10222 	ixgbe_get_oem_prod_version(hw, &nvm_ver);
10223 	if (nvm_ver.oem_valid) {
10224 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10225 			 "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
10226 			 nvm_ver.oem_release);
10227 		return;
10228 	}
10229 
10230 	ixgbe_get_etk_id(hw, &nvm_ver);
10231 	ixgbe_get_orom_version(hw, &nvm_ver);
10232 
10233 	if (nvm_ver.or_valid) {
10234 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10235 			 "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
10236 			 nvm_ver.or_build, nvm_ver.or_patch);
10237 		return;
10238 	}
10239 
10240 	/* Set ETrack ID format */
10241 	snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10242 		 "0x%08x", nvm_ver.etk_id);
10243 }
10244 
10245 /**
10246  * ixgbe_probe - Device Initialization Routine
10247  * @pdev: PCI device information struct
10248  * @ent: entry in ixgbe_pci_tbl
10249  *
10250  * Returns 0 on success, negative on failure
10251  *
10252  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10253  * The OS initialization, configuring of the adapter private structure,
10254  * and a hardware reset occur.
10255  **/
10256 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10257 {
10258 	struct net_device *netdev;
10259 	struct ixgbe_adapter *adapter = NULL;
10260 	struct ixgbe_hw *hw;
10261 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10262 	int i, err, pci_using_dac, expected_gts;
10263 	unsigned int indices = MAX_TX_QUEUES;
10264 	u8 part_str[IXGBE_PBANUM_LENGTH];
10265 	bool disable_dev = false;
10266 #ifdef IXGBE_FCOE
10267 	u16 device_caps;
10268 #endif
10269 	u32 eec;
10270 
10271 	/* Catch broken hardware that put the wrong VF device ID in
10272 	 * the PCIe SR-IOV capability.
10273 	 */
10274 	if (pdev->is_virtfn) {
10275 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10276 		     pci_name(pdev), pdev->vendor, pdev->device);
10277 		return -EINVAL;
10278 	}
10279 
10280 	err = pci_enable_device_mem(pdev);
10281 	if (err)
10282 		return err;
10283 
10284 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
10285 		pci_using_dac = 1;
10286 	} else {
10287 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10288 		if (err) {
10289 			dev_err(&pdev->dev,
10290 				"No usable DMA configuration, aborting\n");
10291 			goto err_dma;
10292 		}
10293 		pci_using_dac = 0;
10294 	}
10295 
10296 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10297 	if (err) {
10298 		dev_err(&pdev->dev,
10299 			"pci_request_selected_regions failed 0x%x\n", err);
10300 		goto err_pci_reg;
10301 	}
10302 
10303 	pci_enable_pcie_error_reporting(pdev);
10304 
10305 	pci_set_master(pdev);
10306 	pci_save_state(pdev);
10307 
10308 	if (ii->mac == ixgbe_mac_82598EB) {
10309 #ifdef CONFIG_IXGBE_DCB
10310 		/* 8 TC w/ 4 queues per TC */
10311 		indices = 4 * MAX_TRAFFIC_CLASS;
10312 #else
10313 		indices = IXGBE_MAX_RSS_INDICES;
10314 #endif
10315 	}
10316 
10317 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10318 	if (!netdev) {
10319 		err = -ENOMEM;
10320 		goto err_alloc_etherdev;
10321 	}
10322 
10323 	SET_NETDEV_DEV(netdev, &pdev->dev);
10324 
10325 	adapter = netdev_priv(netdev);
10326 
10327 	adapter->netdev = netdev;
10328 	adapter->pdev = pdev;
10329 	hw = &adapter->hw;
10330 	hw->back = adapter;
10331 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10332 
10333 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10334 			      pci_resource_len(pdev, 0));
10335 	adapter->io_addr = hw->hw_addr;
10336 	if (!hw->hw_addr) {
10337 		err = -EIO;
10338 		goto err_ioremap;
10339 	}
10340 
10341 	netdev->netdev_ops = &ixgbe_netdev_ops;
10342 	ixgbe_set_ethtool_ops(netdev);
10343 	netdev->watchdog_timeo = 5 * HZ;
10344 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10345 
10346 	/* Setup hw api */
10347 	hw->mac.ops   = *ii->mac_ops;
10348 	hw->mac.type  = ii->mac;
10349 	hw->mvals     = ii->mvals;
10350 	if (ii->link_ops)
10351 		hw->link.ops  = *ii->link_ops;
10352 
10353 	/* EEPROM */
10354 	hw->eeprom.ops = *ii->eeprom_ops;
10355 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10356 	if (ixgbe_removed(hw->hw_addr)) {
10357 		err = -EIO;
10358 		goto err_ioremap;
10359 	}
10360 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10361 	if (!(eec & BIT(8)))
10362 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10363 
10364 	/* PHY */
10365 	hw->phy.ops = *ii->phy_ops;
10366 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10367 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
10368 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10369 	hw->phy.mdio.mmds = 0;
10370 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10371 	hw->phy.mdio.dev = netdev;
10372 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10373 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10374 
10375 	/* setup the private structure */
10376 	err = ixgbe_sw_init(adapter, ii);
10377 	if (err)
10378 		goto err_sw_init;
10379 
10380 	/* Make sure the SWFW semaphore is in a valid state */
10381 	if (hw->mac.ops.init_swfw_sync)
10382 		hw->mac.ops.init_swfw_sync(hw);
10383 
10384 	/* Make it possible the adapter to be woken up via WOL */
10385 	switch (adapter->hw.mac.type) {
10386 	case ixgbe_mac_82599EB:
10387 	case ixgbe_mac_X540:
10388 	case ixgbe_mac_X550:
10389 	case ixgbe_mac_X550EM_x:
10390 	case ixgbe_mac_x550em_a:
10391 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10392 		break;
10393 	default:
10394 		break;
10395 	}
10396 
10397 	/*
10398 	 * If there is a fan on this device and it has failed log the
10399 	 * failure.
10400 	 */
10401 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10402 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10403 		if (esdp & IXGBE_ESDP_SDP1)
10404 			e_crit(probe, "Fan has stopped, replace the adapter\n");
10405 	}
10406 
10407 	if (allow_unsupported_sfp)
10408 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
10409 
10410 	/* reset_hw fills in the perm_addr as well */
10411 	hw->phy.reset_if_overtemp = true;
10412 	err = hw->mac.ops.reset_hw(hw);
10413 	hw->phy.reset_if_overtemp = false;
10414 	ixgbe_set_eee_capable(adapter);
10415 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
10416 		err = 0;
10417 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
10418 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10419 		e_dev_err("Reload the driver after installing a supported module.\n");
10420 		goto err_sw_init;
10421 	} else if (err) {
10422 		e_dev_err("HW Init failed: %d\n", err);
10423 		goto err_sw_init;
10424 	}
10425 
10426 #ifdef CONFIG_PCI_IOV
10427 	/* SR-IOV not supported on the 82598 */
10428 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
10429 		goto skip_sriov;
10430 	/* Mailbox */
10431 	ixgbe_init_mbx_params_pf(hw);
10432 	hw->mbx.ops = ii->mbx_ops;
10433 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
10434 	ixgbe_enable_sriov(adapter, max_vfs);
10435 skip_sriov:
10436 
10437 #endif
10438 	netdev->features = NETIF_F_SG |
10439 			   NETIF_F_TSO |
10440 			   NETIF_F_TSO6 |
10441 			   NETIF_F_RXHASH |
10442 			   NETIF_F_RXCSUM |
10443 			   NETIF_F_HW_CSUM;
10444 
10445 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
10446 				    NETIF_F_GSO_GRE_CSUM | \
10447 				    NETIF_F_GSO_IPXIP4 | \
10448 				    NETIF_F_GSO_IPXIP6 | \
10449 				    NETIF_F_GSO_UDP_TUNNEL | \
10450 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
10451 
10452 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
10453 	netdev->features |= NETIF_F_GSO_PARTIAL |
10454 			    IXGBE_GSO_PARTIAL_FEATURES;
10455 
10456 	if (hw->mac.type >= ixgbe_mac_82599EB)
10457 		netdev->features |= NETIF_F_SCTP_CRC;
10458 
10459 	/* copy netdev features into list of user selectable features */
10460 	netdev->hw_features |= netdev->features |
10461 			       NETIF_F_HW_VLAN_CTAG_FILTER |
10462 			       NETIF_F_HW_VLAN_CTAG_RX |
10463 			       NETIF_F_HW_VLAN_CTAG_TX |
10464 			       NETIF_F_RXALL |
10465 			       NETIF_F_HW_L2FW_DOFFLOAD;
10466 
10467 	if (hw->mac.type >= ixgbe_mac_82599EB)
10468 		netdev->hw_features |= NETIF_F_NTUPLE |
10469 				       NETIF_F_HW_TC;
10470 
10471 	if (pci_using_dac)
10472 		netdev->features |= NETIF_F_HIGHDMA;
10473 
10474 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
10475 	netdev->hw_enc_features |= netdev->vlan_features;
10476 	netdev->mpls_features |= NETIF_F_SG |
10477 				 NETIF_F_TSO |
10478 				 NETIF_F_TSO6 |
10479 				 NETIF_F_HW_CSUM;
10480 	netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
10481 
10482 	/* set this bit last since it cannot be part of vlan_features */
10483 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
10484 			    NETIF_F_HW_VLAN_CTAG_RX |
10485 			    NETIF_F_HW_VLAN_CTAG_TX;
10486 
10487 	netdev->priv_flags |= IFF_UNICAST_FLT;
10488 	netdev->priv_flags |= IFF_SUPP_NOFCS;
10489 
10490 	/* MTU range: 68 - 9710 */
10491 	netdev->min_mtu = ETH_MIN_MTU;
10492 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
10493 
10494 #ifdef CONFIG_IXGBE_DCB
10495 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
10496 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
10497 #endif
10498 
10499 #ifdef IXGBE_FCOE
10500 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
10501 		unsigned int fcoe_l;
10502 
10503 		if (hw->mac.ops.get_device_caps) {
10504 			hw->mac.ops.get_device_caps(hw, &device_caps);
10505 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
10506 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
10507 		}
10508 
10509 
10510 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
10511 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
10512 
10513 		netdev->features |= NETIF_F_FSO |
10514 				    NETIF_F_FCOE_CRC;
10515 
10516 		netdev->vlan_features |= NETIF_F_FSO |
10517 					 NETIF_F_FCOE_CRC |
10518 					 NETIF_F_FCOE_MTU;
10519 	}
10520 #endif /* IXGBE_FCOE */
10521 	ixgbe_init_ipsec_offload(adapter);
10522 
10523 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
10524 		netdev->hw_features |= NETIF_F_LRO;
10525 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
10526 		netdev->features |= NETIF_F_LRO;
10527 
10528 	/* make sure the EEPROM is good */
10529 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
10530 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
10531 		err = -EIO;
10532 		goto err_sw_init;
10533 	}
10534 
10535 	eth_platform_get_mac_address(&adapter->pdev->dev,
10536 				     adapter->hw.mac.perm_addr);
10537 
10538 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
10539 
10540 	if (!is_valid_ether_addr(netdev->dev_addr)) {
10541 		e_dev_err("invalid MAC address\n");
10542 		err = -EIO;
10543 		goto err_sw_init;
10544 	}
10545 
10546 	/* Set hw->mac.addr to permanent MAC address */
10547 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
10548 	ixgbe_mac_set_default_filter(adapter);
10549 
10550 	timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
10551 
10552 	if (ixgbe_removed(hw->hw_addr)) {
10553 		err = -EIO;
10554 		goto err_sw_init;
10555 	}
10556 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
10557 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
10558 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
10559 
10560 	err = ixgbe_init_interrupt_scheme(adapter);
10561 	if (err)
10562 		goto err_sw_init;
10563 
10564 	for (i = 0; i < adapter->num_rx_queues; i++)
10565 		u64_stats_init(&adapter->rx_ring[i]->syncp);
10566 	for (i = 0; i < adapter->num_tx_queues; i++)
10567 		u64_stats_init(&adapter->tx_ring[i]->syncp);
10568 	for (i = 0; i < adapter->num_xdp_queues; i++)
10569 		u64_stats_init(&adapter->xdp_ring[i]->syncp);
10570 
10571 	/* WOL not supported for all devices */
10572 	adapter->wol = 0;
10573 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
10574 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
10575 						pdev->subsystem_device);
10576 	if (hw->wol_enabled)
10577 		adapter->wol = IXGBE_WUFC_MAG;
10578 
10579 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
10580 
10581 	/* save off EEPROM version number */
10582 	ixgbe_set_fw_version(adapter);
10583 
10584 	/* pick up the PCI bus settings for reporting later */
10585 	if (ixgbe_pcie_from_parent(hw))
10586 		ixgbe_get_parent_bus_info(adapter);
10587 	else
10588 		 hw->mac.ops.get_bus_info(hw);
10589 
10590 	/* calculate the expected PCIe bandwidth required for optimal
10591 	 * performance. Note that some older parts will never have enough
10592 	 * bandwidth due to being older generation PCIe parts. We clamp these
10593 	 * parts to ensure no warning is displayed if it can't be fixed.
10594 	 */
10595 	switch (hw->mac.type) {
10596 	case ixgbe_mac_82598EB:
10597 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
10598 		break;
10599 	default:
10600 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
10601 		break;
10602 	}
10603 
10604 	/* don't check link if we failed to enumerate functions */
10605 	if (expected_gts > 0)
10606 		ixgbe_check_minimum_link(adapter, expected_gts);
10607 
10608 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
10609 	if (err)
10610 		strlcpy(part_str, "Unknown", sizeof(part_str));
10611 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
10612 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
10613 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
10614 			   part_str);
10615 	else
10616 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
10617 			   hw->mac.type, hw->phy.type, part_str);
10618 
10619 	e_dev_info("%pM\n", netdev->dev_addr);
10620 
10621 	/* reset the hardware with the new settings */
10622 	err = hw->mac.ops.start_hw(hw);
10623 	if (err == IXGBE_ERR_EEPROM_VERSION) {
10624 		/* We are running on a pre-production device, log a warning */
10625 		e_dev_warn("This device is a pre-production adapter/LOM. "
10626 			   "Please be aware there may be issues associated "
10627 			   "with your hardware.  If you are experiencing "
10628 			   "problems please contact your Intel or hardware "
10629 			   "representative who provided you with this "
10630 			   "hardware.\n");
10631 	}
10632 	strcpy(netdev->name, "eth%d");
10633 	pci_set_drvdata(pdev, adapter);
10634 	err = register_netdev(netdev);
10635 	if (err)
10636 		goto err_register;
10637 
10638 
10639 	/* power down the optics for 82599 SFP+ fiber */
10640 	if (hw->mac.ops.disable_tx_laser)
10641 		hw->mac.ops.disable_tx_laser(hw);
10642 
10643 	/* carrier off reporting is important to ethtool even BEFORE open */
10644 	netif_carrier_off(netdev);
10645 
10646 #ifdef CONFIG_IXGBE_DCA
10647 	if (dca_add_requester(&pdev->dev) == 0) {
10648 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
10649 		ixgbe_setup_dca(adapter);
10650 	}
10651 #endif
10652 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
10653 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
10654 		for (i = 0; i < adapter->num_vfs; i++)
10655 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
10656 	}
10657 
10658 	/* firmware requires driver version to be 0xFFFFFFFF
10659 	 * since os does not support feature
10660 	 */
10661 	if (hw->mac.ops.set_fw_drv_ver)
10662 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
10663 					   sizeof(ixgbe_driver_version) - 1,
10664 					   ixgbe_driver_version);
10665 
10666 	/* add san mac addr to netdev */
10667 	ixgbe_add_sanmac_netdev(netdev);
10668 
10669 	e_dev_info("%s\n", ixgbe_default_device_descr);
10670 
10671 #ifdef CONFIG_IXGBE_HWMON
10672 	if (ixgbe_sysfs_init(adapter))
10673 		e_err(probe, "failed to allocate sysfs resources\n");
10674 #endif /* CONFIG_IXGBE_HWMON */
10675 
10676 	ixgbe_dbg_adapter_init(adapter);
10677 
10678 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
10679 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
10680 		hw->mac.ops.setup_link(hw,
10681 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
10682 			true);
10683 
10684 	return 0;
10685 
10686 err_register:
10687 	ixgbe_release_hw_control(adapter);
10688 	ixgbe_clear_interrupt_scheme(adapter);
10689 err_sw_init:
10690 	ixgbe_disable_sriov(adapter);
10691 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
10692 	iounmap(adapter->io_addr);
10693 	kfree(adapter->jump_tables[0]);
10694 	kfree(adapter->mac_table);
10695 	kfree(adapter->rss_key);
10696 err_ioremap:
10697 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10698 	free_netdev(netdev);
10699 err_alloc_etherdev:
10700 	pci_release_mem_regions(pdev);
10701 err_pci_reg:
10702 err_dma:
10703 	if (!adapter || disable_dev)
10704 		pci_disable_device(pdev);
10705 	return err;
10706 }
10707 
10708 /**
10709  * ixgbe_remove - Device Removal Routine
10710  * @pdev: PCI device information struct
10711  *
10712  * ixgbe_remove is called by the PCI subsystem to alert the driver
10713  * that it should release a PCI device.  The could be caused by a
10714  * Hot-Plug event, or because the driver is going to be removed from
10715  * memory.
10716  **/
10717 static void ixgbe_remove(struct pci_dev *pdev)
10718 {
10719 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10720 	struct net_device *netdev;
10721 	bool disable_dev;
10722 	int i;
10723 
10724 	/* if !adapter then we already cleaned up in probe */
10725 	if (!adapter)
10726 		return;
10727 
10728 	netdev  = adapter->netdev;
10729 	ixgbe_dbg_adapter_exit(adapter);
10730 
10731 	set_bit(__IXGBE_REMOVING, &adapter->state);
10732 	cancel_work_sync(&adapter->service_task);
10733 
10734 
10735 #ifdef CONFIG_IXGBE_DCA
10736 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
10737 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
10738 		dca_remove_requester(&pdev->dev);
10739 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
10740 				IXGBE_DCA_CTRL_DCA_DISABLE);
10741 	}
10742 
10743 #endif
10744 #ifdef CONFIG_IXGBE_HWMON
10745 	ixgbe_sysfs_exit(adapter);
10746 #endif /* CONFIG_IXGBE_HWMON */
10747 
10748 	/* remove the added san mac */
10749 	ixgbe_del_sanmac_netdev(netdev);
10750 
10751 #ifdef CONFIG_PCI_IOV
10752 	ixgbe_disable_sriov(adapter);
10753 #endif
10754 	if (netdev->reg_state == NETREG_REGISTERED)
10755 		unregister_netdev(netdev);
10756 
10757 	ixgbe_stop_ipsec_offload(adapter);
10758 	ixgbe_clear_interrupt_scheme(adapter);
10759 
10760 	ixgbe_release_hw_control(adapter);
10761 
10762 #ifdef CONFIG_DCB
10763 	kfree(adapter->ixgbe_ieee_pfc);
10764 	kfree(adapter->ixgbe_ieee_ets);
10765 
10766 #endif
10767 	iounmap(adapter->io_addr);
10768 	pci_release_mem_regions(pdev);
10769 
10770 	e_dev_info("complete\n");
10771 
10772 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
10773 		if (adapter->jump_tables[i]) {
10774 			kfree(adapter->jump_tables[i]->input);
10775 			kfree(adapter->jump_tables[i]->mask);
10776 		}
10777 		kfree(adapter->jump_tables[i]);
10778 	}
10779 
10780 	kfree(adapter->mac_table);
10781 	kfree(adapter->rss_key);
10782 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10783 	free_netdev(netdev);
10784 
10785 	pci_disable_pcie_error_reporting(pdev);
10786 
10787 	if (disable_dev)
10788 		pci_disable_device(pdev);
10789 }
10790 
10791 /**
10792  * ixgbe_io_error_detected - called when PCI error is detected
10793  * @pdev: Pointer to PCI device
10794  * @state: The current pci connection state
10795  *
10796  * This function is called after a PCI bus error affecting
10797  * this device has been detected.
10798  */
10799 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
10800 						pci_channel_state_t state)
10801 {
10802 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10803 	struct net_device *netdev = adapter->netdev;
10804 
10805 #ifdef CONFIG_PCI_IOV
10806 	struct ixgbe_hw *hw = &adapter->hw;
10807 	struct pci_dev *bdev, *vfdev;
10808 	u32 dw0, dw1, dw2, dw3;
10809 	int vf, pos;
10810 	u16 req_id, pf_func;
10811 
10812 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
10813 	    adapter->num_vfs == 0)
10814 		goto skip_bad_vf_detection;
10815 
10816 	bdev = pdev->bus->self;
10817 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
10818 		bdev = bdev->bus->self;
10819 
10820 	if (!bdev)
10821 		goto skip_bad_vf_detection;
10822 
10823 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
10824 	if (!pos)
10825 		goto skip_bad_vf_detection;
10826 
10827 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
10828 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
10829 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
10830 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
10831 	if (ixgbe_removed(hw->hw_addr))
10832 		goto skip_bad_vf_detection;
10833 
10834 	req_id = dw1 >> 16;
10835 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
10836 	if (!(req_id & 0x0080))
10837 		goto skip_bad_vf_detection;
10838 
10839 	pf_func = req_id & 0x01;
10840 	if ((pf_func & 1) == (pdev->devfn & 1)) {
10841 		unsigned int device_id;
10842 
10843 		vf = (req_id & 0x7F) >> 1;
10844 		e_dev_err("VF %d has caused a PCIe error\n", vf);
10845 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
10846 				"%8.8x\tdw3: %8.8x\n",
10847 		dw0, dw1, dw2, dw3);
10848 		switch (adapter->hw.mac.type) {
10849 		case ixgbe_mac_82599EB:
10850 			device_id = IXGBE_82599_VF_DEVICE_ID;
10851 			break;
10852 		case ixgbe_mac_X540:
10853 			device_id = IXGBE_X540_VF_DEVICE_ID;
10854 			break;
10855 		case ixgbe_mac_X550:
10856 			device_id = IXGBE_DEV_ID_X550_VF;
10857 			break;
10858 		case ixgbe_mac_X550EM_x:
10859 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
10860 			break;
10861 		case ixgbe_mac_x550em_a:
10862 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
10863 			break;
10864 		default:
10865 			device_id = 0;
10866 			break;
10867 		}
10868 
10869 		/* Find the pci device of the offending VF */
10870 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
10871 		while (vfdev) {
10872 			if (vfdev->devfn == (req_id & 0xFF))
10873 				break;
10874 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
10875 					       device_id, vfdev);
10876 		}
10877 		/*
10878 		 * There's a slim chance the VF could have been hot plugged,
10879 		 * so if it is no longer present we don't need to issue the
10880 		 * VFLR.  Just clean up the AER in that case.
10881 		 */
10882 		if (vfdev) {
10883 			pcie_flr(vfdev);
10884 			/* Free device reference count */
10885 			pci_dev_put(vfdev);
10886 		}
10887 
10888 		pci_cleanup_aer_uncorrect_error_status(pdev);
10889 	}
10890 
10891 	/*
10892 	 * Even though the error may have occurred on the other port
10893 	 * we still need to increment the vf error reference count for
10894 	 * both ports because the I/O resume function will be called
10895 	 * for both of them.
10896 	 */
10897 	adapter->vferr_refcount++;
10898 
10899 	return PCI_ERS_RESULT_RECOVERED;
10900 
10901 skip_bad_vf_detection:
10902 #endif /* CONFIG_PCI_IOV */
10903 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
10904 		return PCI_ERS_RESULT_DISCONNECT;
10905 
10906 	if (!netif_device_present(netdev))
10907 		return PCI_ERS_RESULT_DISCONNECT;
10908 
10909 	rtnl_lock();
10910 	netif_device_detach(netdev);
10911 
10912 	if (state == pci_channel_io_perm_failure) {
10913 		rtnl_unlock();
10914 		return PCI_ERS_RESULT_DISCONNECT;
10915 	}
10916 
10917 	if (netif_running(netdev))
10918 		ixgbe_close_suspend(adapter);
10919 
10920 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
10921 		pci_disable_device(pdev);
10922 	rtnl_unlock();
10923 
10924 	/* Request a slot reset. */
10925 	return PCI_ERS_RESULT_NEED_RESET;
10926 }
10927 
10928 /**
10929  * ixgbe_io_slot_reset - called after the pci bus has been reset.
10930  * @pdev: Pointer to PCI device
10931  *
10932  * Restart the card from scratch, as if from a cold-boot.
10933  */
10934 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
10935 {
10936 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10937 	pci_ers_result_t result;
10938 	int err;
10939 
10940 	if (pci_enable_device_mem(pdev)) {
10941 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
10942 		result = PCI_ERS_RESULT_DISCONNECT;
10943 	} else {
10944 		smp_mb__before_atomic();
10945 		clear_bit(__IXGBE_DISABLED, &adapter->state);
10946 		adapter->hw.hw_addr = adapter->io_addr;
10947 		pci_set_master(pdev);
10948 		pci_restore_state(pdev);
10949 		pci_save_state(pdev);
10950 
10951 		pci_wake_from_d3(pdev, false);
10952 
10953 		ixgbe_reset(adapter);
10954 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10955 		result = PCI_ERS_RESULT_RECOVERED;
10956 	}
10957 
10958 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10959 	if (err) {
10960 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
10961 			  "failed 0x%0x\n", err);
10962 		/* non-fatal, continue */
10963 	}
10964 
10965 	return result;
10966 }
10967 
10968 /**
10969  * ixgbe_io_resume - called when traffic can start flowing again.
10970  * @pdev: Pointer to PCI device
10971  *
10972  * This callback is called when the error recovery driver tells us that
10973  * its OK to resume normal operation.
10974  */
10975 static void ixgbe_io_resume(struct pci_dev *pdev)
10976 {
10977 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10978 	struct net_device *netdev = adapter->netdev;
10979 
10980 #ifdef CONFIG_PCI_IOV
10981 	if (adapter->vferr_refcount) {
10982 		e_info(drv, "Resuming after VF err\n");
10983 		adapter->vferr_refcount--;
10984 		return;
10985 	}
10986 
10987 #endif
10988 	rtnl_lock();
10989 	if (netif_running(netdev))
10990 		ixgbe_open(netdev);
10991 
10992 	netif_device_attach(netdev);
10993 	rtnl_unlock();
10994 }
10995 
10996 static const struct pci_error_handlers ixgbe_err_handler = {
10997 	.error_detected = ixgbe_io_error_detected,
10998 	.slot_reset = ixgbe_io_slot_reset,
10999 	.resume = ixgbe_io_resume,
11000 };
11001 
11002 static struct pci_driver ixgbe_driver = {
11003 	.name     = ixgbe_driver_name,
11004 	.id_table = ixgbe_pci_tbl,
11005 	.probe    = ixgbe_probe,
11006 	.remove   = ixgbe_remove,
11007 #ifdef CONFIG_PM
11008 	.suspend  = ixgbe_suspend,
11009 	.resume   = ixgbe_resume,
11010 #endif
11011 	.shutdown = ixgbe_shutdown,
11012 	.sriov_configure = ixgbe_pci_sriov_configure,
11013 	.err_handler = &ixgbe_err_handler
11014 };
11015 
11016 /**
11017  * ixgbe_init_module - Driver Registration Routine
11018  *
11019  * ixgbe_init_module is the first routine called when the driver is
11020  * loaded. All it does is register with the PCI subsystem.
11021  **/
11022 static int __init ixgbe_init_module(void)
11023 {
11024 	int ret;
11025 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
11026 	pr_info("%s\n", ixgbe_copyright);
11027 
11028 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
11029 	if (!ixgbe_wq) {
11030 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
11031 		return -ENOMEM;
11032 	}
11033 
11034 	ixgbe_dbg_init();
11035 
11036 	ret = pci_register_driver(&ixgbe_driver);
11037 	if (ret) {
11038 		destroy_workqueue(ixgbe_wq);
11039 		ixgbe_dbg_exit();
11040 		return ret;
11041 	}
11042 
11043 #ifdef CONFIG_IXGBE_DCA
11044 	dca_register_notify(&dca_notifier);
11045 #endif
11046 
11047 	return 0;
11048 }
11049 
11050 module_init(ixgbe_init_module);
11051 
11052 /**
11053  * ixgbe_exit_module - Driver Exit Cleanup Routine
11054  *
11055  * ixgbe_exit_module is called just before the driver is removed
11056  * from memory.
11057  **/
11058 static void __exit ixgbe_exit_module(void)
11059 {
11060 #ifdef CONFIG_IXGBE_DCA
11061 	dca_unregister_notify(&dca_notifier);
11062 #endif
11063 	pci_unregister_driver(&ixgbe_driver);
11064 
11065 	ixgbe_dbg_exit();
11066 	if (ixgbe_wq) {
11067 		destroy_workqueue(ixgbe_wq);
11068 		ixgbe_wq = NULL;
11069 	}
11070 }
11071 
11072 #ifdef CONFIG_IXGBE_DCA
11073 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
11074 			    void *p)
11075 {
11076 	int ret_val;
11077 
11078 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
11079 					 __ixgbe_notify_dca);
11080 
11081 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
11082 }
11083 
11084 #endif /* CONFIG_IXGBE_DCA */
11085 
11086 module_exit(ixgbe_exit_module);
11087 
11088 /* ixgbe_main.c */
11089