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 int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
196 					  u32 reg, u16 *value)
197 {
198 	struct pci_dev *parent_dev;
199 	struct pci_bus *parent_bus;
200 
201 	parent_bus = adapter->pdev->bus->parent;
202 	if (!parent_bus)
203 		return -1;
204 
205 	parent_dev = parent_bus->self;
206 	if (!parent_dev)
207 		return -1;
208 
209 	if (!pci_is_pcie(parent_dev))
210 		return -1;
211 
212 	pcie_capability_read_word(parent_dev, reg, value);
213 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
214 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
215 		return -1;
216 	return 0;
217 }
218 
219 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
220 {
221 	struct ixgbe_hw *hw = &adapter->hw;
222 	u16 link_status = 0;
223 	int err;
224 
225 	hw->bus.type = ixgbe_bus_type_pci_express;
226 
227 	/* Get the negotiated link width and speed from PCI config space of the
228 	 * parent, as this device is behind a switch
229 	 */
230 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
231 
232 	/* assume caller will handle error case */
233 	if (err)
234 		return err;
235 
236 	hw->bus.width = ixgbe_convert_bus_width(link_status);
237 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
238 
239 	return 0;
240 }
241 
242 /**
243  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
244  * @hw: hw specific details
245  *
246  * This function is used by probe to determine whether a device's PCI-Express
247  * bandwidth details should be gathered from the parent bus instead of from the
248  * device. Used to ensure that various locations all have the correct device ID
249  * checks.
250  */
251 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
252 {
253 	switch (hw->device_id) {
254 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
255 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
256 		return true;
257 	default:
258 		return false;
259 	}
260 }
261 
262 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
263 				     int expected_gts)
264 {
265 	struct ixgbe_hw *hw = &adapter->hw;
266 	int max_gts = 0;
267 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
268 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
269 	struct pci_dev *pdev;
270 
271 	/* Some devices are not connected over PCIe and thus do not negotiate
272 	 * speed. These devices do not have valid bus info, and thus any report
273 	 * we generate may not be correct.
274 	 */
275 	if (hw->bus.type == ixgbe_bus_type_internal)
276 		return;
277 
278 	/* determine whether to use the parent device */
279 	if (ixgbe_pcie_from_parent(&adapter->hw))
280 		pdev = adapter->pdev->bus->parent->self;
281 	else
282 		pdev = adapter->pdev;
283 
284 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
285 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
286 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
287 		return;
288 	}
289 
290 	switch (speed) {
291 	case PCIE_SPEED_2_5GT:
292 		/* 8b/10b encoding reduces max throughput by 20% */
293 		max_gts = 2 * width;
294 		break;
295 	case PCIE_SPEED_5_0GT:
296 		/* 8b/10b encoding reduces max throughput by 20% */
297 		max_gts = 4 * width;
298 		break;
299 	case PCIE_SPEED_8_0GT:
300 		/* 128b/130b encoding reduces throughput by less than 2% */
301 		max_gts = 8 * width;
302 		break;
303 	default:
304 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
305 		return;
306 	}
307 
308 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
309 		   max_gts);
310 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
311 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
312 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
313 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
314 		    "Unknown"),
315 		   width,
316 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
317 		    speed == PCIE_SPEED_5_0GT ? "20%" :
318 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
319 		    "Unknown"));
320 
321 	if (max_gts < expected_gts) {
322 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
323 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
324 			expected_gts);
325 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
326 	}
327 }
328 
329 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
330 {
331 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
332 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
333 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
334 		queue_work(ixgbe_wq, &adapter->service_task);
335 }
336 
337 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
338 {
339 	struct ixgbe_adapter *adapter = hw->back;
340 
341 	if (!hw->hw_addr)
342 		return;
343 	hw->hw_addr = NULL;
344 	e_dev_err("Adapter removed\n");
345 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
346 		ixgbe_service_event_schedule(adapter);
347 }
348 
349 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
350 {
351 	u32 value;
352 
353 	/* The following check not only optimizes a bit by not
354 	 * performing a read on the status register when the
355 	 * register just read was a status register read that
356 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
357 	 * potential recursion.
358 	 */
359 	if (reg == IXGBE_STATUS) {
360 		ixgbe_remove_adapter(hw);
361 		return;
362 	}
363 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
364 	if (value == IXGBE_FAILED_READ_REG)
365 		ixgbe_remove_adapter(hw);
366 }
367 
368 /**
369  * ixgbe_read_reg - Read from device register
370  * @hw: hw specific details
371  * @reg: offset of register to read
372  *
373  * Returns : value read or IXGBE_FAILED_READ_REG if removed
374  *
375  * This function is used to read device registers. It checks for device
376  * removal by confirming any read that returns all ones by checking the
377  * status register value for all ones. This function avoids reading from
378  * the hardware if a removal was previously detected in which case it
379  * returns IXGBE_FAILED_READ_REG (all ones).
380  */
381 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
382 {
383 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
384 	u32 value;
385 
386 	if (ixgbe_removed(reg_addr))
387 		return IXGBE_FAILED_READ_REG;
388 	if (unlikely(hw->phy.nw_mng_if_sel &
389 		     IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
390 		struct ixgbe_adapter *adapter;
391 		int i;
392 
393 		for (i = 0; i < 200; ++i) {
394 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
395 			if (likely(!value))
396 				goto writes_completed;
397 			if (value == IXGBE_FAILED_READ_REG) {
398 				ixgbe_remove_adapter(hw);
399 				return IXGBE_FAILED_READ_REG;
400 			}
401 			udelay(5);
402 		}
403 
404 		adapter = hw->back;
405 		e_warn(hw, "register writes incomplete %08x\n", value);
406 	}
407 
408 writes_completed:
409 	value = readl(reg_addr + reg);
410 	if (unlikely(value == IXGBE_FAILED_READ_REG))
411 		ixgbe_check_remove(hw, reg);
412 	return value;
413 }
414 
415 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
416 {
417 	u16 value;
418 
419 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
420 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
421 		ixgbe_remove_adapter(hw);
422 		return true;
423 	}
424 	return false;
425 }
426 
427 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
428 {
429 	struct ixgbe_adapter *adapter = hw->back;
430 	u16 value;
431 
432 	if (ixgbe_removed(hw->hw_addr))
433 		return IXGBE_FAILED_READ_CFG_WORD;
434 	pci_read_config_word(adapter->pdev, reg, &value);
435 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
436 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
437 		return IXGBE_FAILED_READ_CFG_WORD;
438 	return value;
439 }
440 
441 #ifdef CONFIG_PCI_IOV
442 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
443 {
444 	struct ixgbe_adapter *adapter = hw->back;
445 	u32 value;
446 
447 	if (ixgbe_removed(hw->hw_addr))
448 		return IXGBE_FAILED_READ_CFG_DWORD;
449 	pci_read_config_dword(adapter->pdev, reg, &value);
450 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
451 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
452 		return IXGBE_FAILED_READ_CFG_DWORD;
453 	return value;
454 }
455 #endif /* CONFIG_PCI_IOV */
456 
457 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
458 {
459 	struct ixgbe_adapter *adapter = hw->back;
460 
461 	if (ixgbe_removed(hw->hw_addr))
462 		return;
463 	pci_write_config_word(adapter->pdev, reg, value);
464 }
465 
466 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
467 {
468 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
469 
470 	/* flush memory to make sure state is correct before next watchdog */
471 	smp_mb__before_atomic();
472 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
473 }
474 
475 struct ixgbe_reg_info {
476 	u32 ofs;
477 	char *name;
478 };
479 
480 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
481 
482 	/* General Registers */
483 	{IXGBE_CTRL, "CTRL"},
484 	{IXGBE_STATUS, "STATUS"},
485 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
486 
487 	/* Interrupt Registers */
488 	{IXGBE_EICR, "EICR"},
489 
490 	/* RX Registers */
491 	{IXGBE_SRRCTL(0), "SRRCTL"},
492 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
493 	{IXGBE_RDLEN(0), "RDLEN"},
494 	{IXGBE_RDH(0), "RDH"},
495 	{IXGBE_RDT(0), "RDT"},
496 	{IXGBE_RXDCTL(0), "RXDCTL"},
497 	{IXGBE_RDBAL(0), "RDBAL"},
498 	{IXGBE_RDBAH(0), "RDBAH"},
499 
500 	/* TX Registers */
501 	{IXGBE_TDBAL(0), "TDBAL"},
502 	{IXGBE_TDBAH(0), "TDBAH"},
503 	{IXGBE_TDLEN(0), "TDLEN"},
504 	{IXGBE_TDH(0), "TDH"},
505 	{IXGBE_TDT(0), "TDT"},
506 	{IXGBE_TXDCTL(0), "TXDCTL"},
507 
508 	/* List Terminator */
509 	{ .name = NULL }
510 };
511 
512 
513 /*
514  * ixgbe_regdump - register printout routine
515  */
516 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
517 {
518 	int i;
519 	char rname[16];
520 	u32 regs[64];
521 
522 	switch (reginfo->ofs) {
523 	case IXGBE_SRRCTL(0):
524 		for (i = 0; i < 64; i++)
525 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
526 		break;
527 	case IXGBE_DCA_RXCTRL(0):
528 		for (i = 0; i < 64; i++)
529 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
530 		break;
531 	case IXGBE_RDLEN(0):
532 		for (i = 0; i < 64; i++)
533 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
534 		break;
535 	case IXGBE_RDH(0):
536 		for (i = 0; i < 64; i++)
537 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
538 		break;
539 	case IXGBE_RDT(0):
540 		for (i = 0; i < 64; i++)
541 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
542 		break;
543 	case IXGBE_RXDCTL(0):
544 		for (i = 0; i < 64; i++)
545 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
546 		break;
547 	case IXGBE_RDBAL(0):
548 		for (i = 0; i < 64; i++)
549 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
550 		break;
551 	case IXGBE_RDBAH(0):
552 		for (i = 0; i < 64; i++)
553 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
554 		break;
555 	case IXGBE_TDBAL(0):
556 		for (i = 0; i < 64; i++)
557 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
558 		break;
559 	case IXGBE_TDBAH(0):
560 		for (i = 0; i < 64; i++)
561 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
562 		break;
563 	case IXGBE_TDLEN(0):
564 		for (i = 0; i < 64; i++)
565 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
566 		break;
567 	case IXGBE_TDH(0):
568 		for (i = 0; i < 64; i++)
569 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
570 		break;
571 	case IXGBE_TDT(0):
572 		for (i = 0; i < 64; i++)
573 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
574 		break;
575 	case IXGBE_TXDCTL(0):
576 		for (i = 0; i < 64; i++)
577 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
578 		break;
579 	default:
580 		pr_info("%-15s %08x\n",
581 			reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
582 		return;
583 	}
584 
585 	i = 0;
586 	while (i < 64) {
587 		int j;
588 		char buf[9 * 8 + 1];
589 		char *p = buf;
590 
591 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
592 		for (j = 0; j < 8; j++)
593 			p += sprintf(p, " %08x", regs[i++]);
594 		pr_err("%-15s%s\n", rname, buf);
595 	}
596 
597 }
598 
599 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
600 {
601 	struct ixgbe_tx_buffer *tx_buffer;
602 
603 	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
604 	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
605 		n, ring->next_to_use, ring->next_to_clean,
606 		(u64)dma_unmap_addr(tx_buffer, dma),
607 		dma_unmap_len(tx_buffer, len),
608 		tx_buffer->next_to_watch,
609 		(u64)tx_buffer->time_stamp);
610 }
611 
612 /*
613  * ixgbe_dump - Print registers, tx-rings and rx-rings
614  */
615 static void ixgbe_dump(struct ixgbe_adapter *adapter)
616 {
617 	struct net_device *netdev = adapter->netdev;
618 	struct ixgbe_hw *hw = &adapter->hw;
619 	struct ixgbe_reg_info *reginfo;
620 	int n = 0;
621 	struct ixgbe_ring *ring;
622 	struct ixgbe_tx_buffer *tx_buffer;
623 	union ixgbe_adv_tx_desc *tx_desc;
624 	struct my_u0 { u64 a; u64 b; } *u0;
625 	struct ixgbe_ring *rx_ring;
626 	union ixgbe_adv_rx_desc *rx_desc;
627 	struct ixgbe_rx_buffer *rx_buffer_info;
628 	int i = 0;
629 
630 	if (!netif_msg_hw(adapter))
631 		return;
632 
633 	/* Print netdevice Info */
634 	if (netdev) {
635 		dev_info(&adapter->pdev->dev, "Net device Info\n");
636 		pr_info("Device Name     state            "
637 			"trans_start\n");
638 		pr_info("%-15s %016lX %016lX\n",
639 			netdev->name,
640 			netdev->state,
641 			dev_trans_start(netdev));
642 	}
643 
644 	/* Print Registers */
645 	dev_info(&adapter->pdev->dev, "Register Dump\n");
646 	pr_info(" Register Name   Value\n");
647 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
648 	     reginfo->name; reginfo++) {
649 		ixgbe_regdump(hw, reginfo);
650 	}
651 
652 	/* Print TX Ring Summary */
653 	if (!netdev || !netif_running(netdev))
654 		return;
655 
656 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
657 	pr_info(" %s     %s              %s        %s\n",
658 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
659 		"leng", "ntw", "timestamp");
660 	for (n = 0; n < adapter->num_tx_queues; n++) {
661 		ring = adapter->tx_ring[n];
662 		ixgbe_print_buffer(ring, n);
663 	}
664 
665 	for (n = 0; n < adapter->num_xdp_queues; n++) {
666 		ring = adapter->xdp_ring[n];
667 		ixgbe_print_buffer(ring, n);
668 	}
669 
670 	/* Print TX Rings */
671 	if (!netif_msg_tx_done(adapter))
672 		goto rx_ring_summary;
673 
674 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
675 
676 	/* Transmit Descriptor Formats
677 	 *
678 	 * 82598 Advanced Transmit Descriptor
679 	 *   +--------------------------------------------------------------+
680 	 * 0 |         Buffer Address [63:0]                                |
681 	 *   +--------------------------------------------------------------+
682 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
683 	 *   +--------------------------------------------------------------+
684 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
685 	 *
686 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
687 	 *   +--------------------------------------------------------------+
688 	 * 0 |                          RSV [63:0]                          |
689 	 *   +--------------------------------------------------------------+
690 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
691 	 *   +--------------------------------------------------------------+
692 	 *   63                       36 35   32 31                         0
693 	 *
694 	 * 82599+ Advanced Transmit Descriptor
695 	 *   +--------------------------------------------------------------+
696 	 * 0 |         Buffer Address [63:0]                                |
697 	 *   +--------------------------------------------------------------+
698 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
699 	 *   +--------------------------------------------------------------+
700 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
701 	 *
702 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
703 	 *   +--------------------------------------------------------------+
704 	 * 0 |                          RSV [63:0]                          |
705 	 *   +--------------------------------------------------------------+
706 	 * 8 |            RSV           |  STA  |           RSV             |
707 	 *   +--------------------------------------------------------------+
708 	 *   63                       36 35   32 31                         0
709 	 */
710 
711 	for (n = 0; n < adapter->num_tx_queues; n++) {
712 		ring = adapter->tx_ring[n];
713 		pr_info("------------------------------------\n");
714 		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
715 		pr_info("------------------------------------\n");
716 		pr_info("%s%s    %s              %s        %s          %s\n",
717 			"T [desc]     [address 63:0  ] ",
718 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
719 			"leng", "ntw", "timestamp", "bi->skb");
720 
721 		for (i = 0; ring->desc && (i < ring->count); i++) {
722 			tx_desc = IXGBE_TX_DESC(ring, i);
723 			tx_buffer = &ring->tx_buffer_info[i];
724 			u0 = (struct my_u0 *)tx_desc;
725 			if (dma_unmap_len(tx_buffer, len) > 0) {
726 				const char *ring_desc;
727 
728 				if (i == ring->next_to_use &&
729 				    i == ring->next_to_clean)
730 					ring_desc = " NTC/U";
731 				else if (i == ring->next_to_use)
732 					ring_desc = " NTU";
733 				else if (i == ring->next_to_clean)
734 					ring_desc = " NTC";
735 				else
736 					ring_desc = "";
737 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
738 					i,
739 					le64_to_cpu(u0->a),
740 					le64_to_cpu(u0->b),
741 					(u64)dma_unmap_addr(tx_buffer, dma),
742 					dma_unmap_len(tx_buffer, len),
743 					tx_buffer->next_to_watch,
744 					(u64)tx_buffer->time_stamp,
745 					tx_buffer->skb,
746 					ring_desc);
747 
748 				if (netif_msg_pktdata(adapter) &&
749 				    tx_buffer->skb)
750 					print_hex_dump(KERN_INFO, "",
751 						DUMP_PREFIX_ADDRESS, 16, 1,
752 						tx_buffer->skb->data,
753 						dma_unmap_len(tx_buffer, len),
754 						true);
755 			}
756 		}
757 	}
758 
759 	/* Print RX Rings Summary */
760 rx_ring_summary:
761 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
762 	pr_info("Queue [NTU] [NTC]\n");
763 	for (n = 0; n < adapter->num_rx_queues; n++) {
764 		rx_ring = adapter->rx_ring[n];
765 		pr_info("%5d %5X %5X\n",
766 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
767 	}
768 
769 	/* Print RX Rings */
770 	if (!netif_msg_rx_status(adapter))
771 		return;
772 
773 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
774 
775 	/* Receive Descriptor Formats
776 	 *
777 	 * 82598 Advanced Receive Descriptor (Read) Format
778 	 *    63                                           1        0
779 	 *    +-----------------------------------------------------+
780 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
781 	 *    +----------------------------------------------+------+
782 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
783 	 *    +-----------------------------------------------------+
784 	 *
785 	 *
786 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
787 	 *
788 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
789 	 *   +------------------------------------------------------+
790 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
791 	 *   | Packet   | IP     |   |          |     | Type | Type |
792 	 *   | Checksum | Ident  |   |          |     |      |      |
793 	 *   +------------------------------------------------------+
794 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
795 	 *   +------------------------------------------------------+
796 	 *   63       48 47    32 31            20 19               0
797 	 *
798 	 * 82599+ Advanced Receive Descriptor (Read) Format
799 	 *    63                                           1        0
800 	 *    +-----------------------------------------------------+
801 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
802 	 *    +----------------------------------------------+------+
803 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
804 	 *    +-----------------------------------------------------+
805 	 *
806 	 *
807 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
808 	 *
809 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
810 	 *   +------------------------------------------------------+
811 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
812 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
813 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
814 	 *   +------------------------------------------------------+
815 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
816 	 *   +------------------------------------------------------+
817 	 *   63       48 47    32 31          20 19                 0
818 	 */
819 
820 	for (n = 0; n < adapter->num_rx_queues; n++) {
821 		rx_ring = adapter->rx_ring[n];
822 		pr_info("------------------------------------\n");
823 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
824 		pr_info("------------------------------------\n");
825 		pr_info("%s%s%s\n",
826 			"R  [desc]      [ PktBuf     A0] ",
827 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
828 			"<-- Adv Rx Read format");
829 		pr_info("%s%s%s\n",
830 			"RWB[desc]      [PcsmIpSHl PtRs] ",
831 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
832 			"<-- Adv Rx Write-Back format");
833 
834 		for (i = 0; i < rx_ring->count; i++) {
835 			const char *ring_desc;
836 
837 			if (i == rx_ring->next_to_use)
838 				ring_desc = " NTU";
839 			else if (i == rx_ring->next_to_clean)
840 				ring_desc = " NTC";
841 			else
842 				ring_desc = "";
843 
844 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
845 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
846 			u0 = (struct my_u0 *)rx_desc;
847 			if (rx_desc->wb.upper.length) {
848 				/* Descriptor Done */
849 				pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
850 					i,
851 					le64_to_cpu(u0->a),
852 					le64_to_cpu(u0->b),
853 					rx_buffer_info->skb,
854 					ring_desc);
855 			} else {
856 				pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
857 					i,
858 					le64_to_cpu(u0->a),
859 					le64_to_cpu(u0->b),
860 					(u64)rx_buffer_info->dma,
861 					rx_buffer_info->skb,
862 					ring_desc);
863 
864 				if (netif_msg_pktdata(adapter) &&
865 				    rx_buffer_info->dma) {
866 					print_hex_dump(KERN_INFO, "",
867 					   DUMP_PREFIX_ADDRESS, 16, 1,
868 					   page_address(rx_buffer_info->page) +
869 						    rx_buffer_info->page_offset,
870 					   ixgbe_rx_bufsz(rx_ring), true);
871 				}
872 			}
873 		}
874 	}
875 }
876 
877 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
878 {
879 	u32 ctrl_ext;
880 
881 	/* Let firmware take over control of h/w */
882 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
883 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
884 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
885 }
886 
887 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
888 {
889 	u32 ctrl_ext;
890 
891 	/* Let firmware know the driver has taken over */
892 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
893 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
894 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
895 }
896 
897 /**
898  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
899  * @adapter: pointer to adapter struct
900  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
901  * @queue: queue to map the corresponding interrupt to
902  * @msix_vector: the vector to map to the corresponding queue
903  *
904  */
905 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
906 			   u8 queue, u8 msix_vector)
907 {
908 	u32 ivar, index;
909 	struct ixgbe_hw *hw = &adapter->hw;
910 	switch (hw->mac.type) {
911 	case ixgbe_mac_82598EB:
912 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
913 		if (direction == -1)
914 			direction = 0;
915 		index = (((direction * 64) + queue) >> 2) & 0x1F;
916 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
917 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
918 		ivar |= (msix_vector << (8 * (queue & 0x3)));
919 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
920 		break;
921 	case ixgbe_mac_82599EB:
922 	case ixgbe_mac_X540:
923 	case ixgbe_mac_X550:
924 	case ixgbe_mac_X550EM_x:
925 	case ixgbe_mac_x550em_a:
926 		if (direction == -1) {
927 			/* other causes */
928 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
929 			index = ((queue & 1) * 8);
930 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
931 			ivar &= ~(0xFF << index);
932 			ivar |= (msix_vector << index);
933 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
934 			break;
935 		} else {
936 			/* tx or rx causes */
937 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
938 			index = ((16 * (queue & 1)) + (8 * direction));
939 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
940 			ivar &= ~(0xFF << index);
941 			ivar |= (msix_vector << index);
942 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
943 			break;
944 		}
945 	default:
946 		break;
947 	}
948 }
949 
950 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
951 					  u64 qmask)
952 {
953 	u32 mask;
954 
955 	switch (adapter->hw.mac.type) {
956 	case ixgbe_mac_82598EB:
957 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
958 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
959 		break;
960 	case ixgbe_mac_82599EB:
961 	case ixgbe_mac_X540:
962 	case ixgbe_mac_X550:
963 	case ixgbe_mac_X550EM_x:
964 	case ixgbe_mac_x550em_a:
965 		mask = (qmask & 0xFFFFFFFF);
966 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
967 		mask = (qmask >> 32);
968 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
969 		break;
970 	default:
971 		break;
972 	}
973 }
974 
975 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
976 {
977 	struct ixgbe_hw *hw = &adapter->hw;
978 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
979 	int i;
980 	u32 data;
981 
982 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
983 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
984 		return;
985 
986 	switch (hw->mac.type) {
987 	case ixgbe_mac_82598EB:
988 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
989 		break;
990 	default:
991 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
992 	}
993 	hwstats->lxoffrxc += data;
994 
995 	/* refill credits (no tx hang) if we received xoff */
996 	if (!data)
997 		return;
998 
999 	for (i = 0; i < adapter->num_tx_queues; i++)
1000 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
1001 			  &adapter->tx_ring[i]->state);
1002 
1003 	for (i = 0; i < adapter->num_xdp_queues; i++)
1004 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
1005 			  &adapter->xdp_ring[i]->state);
1006 }
1007 
1008 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
1009 {
1010 	struct ixgbe_hw *hw = &adapter->hw;
1011 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
1012 	u32 xoff[8] = {0};
1013 	u8 tc;
1014 	int i;
1015 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
1016 
1017 	if (adapter->ixgbe_ieee_pfc)
1018 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
1019 
1020 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
1021 		ixgbe_update_xoff_rx_lfc(adapter);
1022 		return;
1023 	}
1024 
1025 	/* update stats for each tc, only valid with PFC enabled */
1026 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
1027 		u32 pxoffrxc;
1028 
1029 		switch (hw->mac.type) {
1030 		case ixgbe_mac_82598EB:
1031 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1032 			break;
1033 		default:
1034 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1035 		}
1036 		hwstats->pxoffrxc[i] += pxoffrxc;
1037 		/* Get the TC for given UP */
1038 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1039 		xoff[tc] += pxoffrxc;
1040 	}
1041 
1042 	/* disarm tx queues that have received xoff frames */
1043 	for (i = 0; i < adapter->num_tx_queues; i++) {
1044 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1045 
1046 		tc = tx_ring->dcb_tc;
1047 		if (xoff[tc])
1048 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1049 	}
1050 
1051 	for (i = 0; i < adapter->num_xdp_queues; i++) {
1052 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1053 
1054 		tc = xdp_ring->dcb_tc;
1055 		if (xoff[tc])
1056 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1057 	}
1058 }
1059 
1060 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1061 {
1062 	return ring->stats.packets;
1063 }
1064 
1065 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1066 {
1067 	struct ixgbe_adapter *adapter;
1068 	struct ixgbe_hw *hw;
1069 	u32 head, tail;
1070 
1071 	if (ring->l2_accel_priv)
1072 		adapter = ring->l2_accel_priv->real_adapter;
1073 	else
1074 		adapter = netdev_priv(ring->netdev);
1075 
1076 	hw = &adapter->hw;
1077 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1078 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1079 
1080 	if (head != tail)
1081 		return (head < tail) ?
1082 			tail - head : (tail + ring->count - head);
1083 
1084 	return 0;
1085 }
1086 
1087 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1088 {
1089 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1090 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1091 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1092 
1093 	clear_check_for_tx_hang(tx_ring);
1094 
1095 	/*
1096 	 * Check for a hung queue, but be thorough. This verifies
1097 	 * that a transmit has been completed since the previous
1098 	 * check AND there is at least one packet pending. The
1099 	 * ARMED bit is set to indicate a potential hang. The
1100 	 * bit is cleared if a pause frame is received to remove
1101 	 * false hang detection due to PFC or 802.3x frames. By
1102 	 * requiring this to fail twice we avoid races with
1103 	 * pfc clearing the ARMED bit and conditions where we
1104 	 * run the check_tx_hang logic with a transmit completion
1105 	 * pending but without time to complete it yet.
1106 	 */
1107 	if (tx_done_old == tx_done && tx_pending)
1108 		/* make sure it is true for two checks in a row */
1109 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1110 					&tx_ring->state);
1111 	/* update completed stats and continue */
1112 	tx_ring->tx_stats.tx_done_old = tx_done;
1113 	/* reset the countdown */
1114 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1115 
1116 	return false;
1117 }
1118 
1119 /**
1120  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1121  * @adapter: driver private struct
1122  **/
1123 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1124 {
1125 
1126 	/* Do the reset outside of interrupt context */
1127 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1128 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1129 		e_warn(drv, "initiating reset due to tx timeout\n");
1130 		ixgbe_service_event_schedule(adapter);
1131 	}
1132 }
1133 
1134 /**
1135  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1136  **/
1137 static int ixgbe_tx_maxrate(struct net_device *netdev,
1138 			    int queue_index, u32 maxrate)
1139 {
1140 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1141 	struct ixgbe_hw *hw = &adapter->hw;
1142 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1143 
1144 	if (!maxrate)
1145 		return 0;
1146 
1147 	/* Calculate the rate factor values to set */
1148 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1149 	bcnrc_val /= maxrate;
1150 
1151 	/* clear everything but the rate factor */
1152 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1153 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1154 
1155 	/* enable the rate scheduler */
1156 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1157 
1158 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1159 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1160 
1161 	return 0;
1162 }
1163 
1164 /**
1165  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1166  * @q_vector: structure containing interrupt and ring information
1167  * @tx_ring: tx ring to clean
1168  * @napi_budget: Used to determine if we are in netpoll
1169  **/
1170 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1171 			       struct ixgbe_ring *tx_ring, int napi_budget)
1172 {
1173 	struct ixgbe_adapter *adapter = q_vector->adapter;
1174 	struct ixgbe_tx_buffer *tx_buffer;
1175 	union ixgbe_adv_tx_desc *tx_desc;
1176 	unsigned int total_bytes = 0, total_packets = 0;
1177 	unsigned int budget = q_vector->tx.work_limit;
1178 	unsigned int i = tx_ring->next_to_clean;
1179 
1180 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1181 		return true;
1182 
1183 	tx_buffer = &tx_ring->tx_buffer_info[i];
1184 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1185 	i -= tx_ring->count;
1186 
1187 	do {
1188 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1189 
1190 		/* if next_to_watch is not set then there is no work pending */
1191 		if (!eop_desc)
1192 			break;
1193 
1194 		/* prevent any other reads prior to eop_desc */
1195 		read_barrier_depends();
1196 
1197 		/* if DD is not set pending work has not been completed */
1198 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1199 			break;
1200 
1201 		/* clear next_to_watch to prevent false hangs */
1202 		tx_buffer->next_to_watch = NULL;
1203 
1204 		/* update the statistics for this packet */
1205 		total_bytes += tx_buffer->bytecount;
1206 		total_packets += tx_buffer->gso_segs;
1207 
1208 		/* free the skb */
1209 		if (ring_is_xdp(tx_ring))
1210 			page_frag_free(tx_buffer->data);
1211 		else
1212 			napi_consume_skb(tx_buffer->skb, napi_budget);
1213 
1214 		/* unmap skb header data */
1215 		dma_unmap_single(tx_ring->dev,
1216 				 dma_unmap_addr(tx_buffer, dma),
1217 				 dma_unmap_len(tx_buffer, len),
1218 				 DMA_TO_DEVICE);
1219 
1220 		/* clear tx_buffer data */
1221 		dma_unmap_len_set(tx_buffer, len, 0);
1222 
1223 		/* unmap remaining buffers */
1224 		while (tx_desc != eop_desc) {
1225 			tx_buffer++;
1226 			tx_desc++;
1227 			i++;
1228 			if (unlikely(!i)) {
1229 				i -= tx_ring->count;
1230 				tx_buffer = tx_ring->tx_buffer_info;
1231 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1232 			}
1233 
1234 			/* unmap any remaining paged data */
1235 			if (dma_unmap_len(tx_buffer, len)) {
1236 				dma_unmap_page(tx_ring->dev,
1237 					       dma_unmap_addr(tx_buffer, dma),
1238 					       dma_unmap_len(tx_buffer, len),
1239 					       DMA_TO_DEVICE);
1240 				dma_unmap_len_set(tx_buffer, len, 0);
1241 			}
1242 		}
1243 
1244 		/* move us one more past the eop_desc for start of next pkt */
1245 		tx_buffer++;
1246 		tx_desc++;
1247 		i++;
1248 		if (unlikely(!i)) {
1249 			i -= tx_ring->count;
1250 			tx_buffer = tx_ring->tx_buffer_info;
1251 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1252 		}
1253 
1254 		/* issue prefetch for next Tx descriptor */
1255 		prefetch(tx_desc);
1256 
1257 		/* update budget accounting */
1258 		budget--;
1259 	} while (likely(budget));
1260 
1261 	i += tx_ring->count;
1262 	tx_ring->next_to_clean = i;
1263 	u64_stats_update_begin(&tx_ring->syncp);
1264 	tx_ring->stats.bytes += total_bytes;
1265 	tx_ring->stats.packets += total_packets;
1266 	u64_stats_update_end(&tx_ring->syncp);
1267 	q_vector->tx.total_bytes += total_bytes;
1268 	q_vector->tx.total_packets += total_packets;
1269 
1270 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1271 		/* schedule immediate reset if we believe we hung */
1272 		struct ixgbe_hw *hw = &adapter->hw;
1273 		e_err(drv, "Detected Tx Unit Hang %s\n"
1274 			"  Tx Queue             <%d>\n"
1275 			"  TDH, TDT             <%x>, <%x>\n"
1276 			"  next_to_use          <%x>\n"
1277 			"  next_to_clean        <%x>\n"
1278 			"tx_buffer_info[next_to_clean]\n"
1279 			"  time_stamp           <%lx>\n"
1280 			"  jiffies              <%lx>\n",
1281 			ring_is_xdp(tx_ring) ? "(XDP)" : "",
1282 			tx_ring->queue_index,
1283 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1284 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1285 			tx_ring->next_to_use, i,
1286 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1287 
1288 		if (!ring_is_xdp(tx_ring))
1289 			netif_stop_subqueue(tx_ring->netdev,
1290 					    tx_ring->queue_index);
1291 
1292 		e_info(probe,
1293 		       "tx hang %d detected on queue %d, resetting adapter\n",
1294 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1295 
1296 		/* schedule immediate reset if we believe we hung */
1297 		ixgbe_tx_timeout_reset(adapter);
1298 
1299 		/* the adapter is about to reset, no point in enabling stuff */
1300 		return true;
1301 	}
1302 
1303 	if (ring_is_xdp(tx_ring))
1304 		return !!budget;
1305 
1306 	netdev_tx_completed_queue(txring_txq(tx_ring),
1307 				  total_packets, total_bytes);
1308 
1309 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1310 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1311 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1312 		/* Make sure that anybody stopping the queue after this
1313 		 * sees the new next_to_clean.
1314 		 */
1315 		smp_mb();
1316 		if (__netif_subqueue_stopped(tx_ring->netdev,
1317 					     tx_ring->queue_index)
1318 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1319 			netif_wake_subqueue(tx_ring->netdev,
1320 					    tx_ring->queue_index);
1321 			++tx_ring->tx_stats.restart_queue;
1322 		}
1323 	}
1324 
1325 	return !!budget;
1326 }
1327 
1328 #ifdef CONFIG_IXGBE_DCA
1329 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1330 				struct ixgbe_ring *tx_ring,
1331 				int cpu)
1332 {
1333 	struct ixgbe_hw *hw = &adapter->hw;
1334 	u32 txctrl = 0;
1335 	u16 reg_offset;
1336 
1337 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1338 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1339 
1340 	switch (hw->mac.type) {
1341 	case ixgbe_mac_82598EB:
1342 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1343 		break;
1344 	case ixgbe_mac_82599EB:
1345 	case ixgbe_mac_X540:
1346 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1347 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1348 		break;
1349 	default:
1350 		/* for unknown hardware do not write register */
1351 		return;
1352 	}
1353 
1354 	/*
1355 	 * We can enable relaxed ordering for reads, but not writes when
1356 	 * DCA is enabled.  This is due to a known issue in some chipsets
1357 	 * which will cause the DCA tag to be cleared.
1358 	 */
1359 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1360 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1361 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1362 
1363 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1364 }
1365 
1366 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1367 				struct ixgbe_ring *rx_ring,
1368 				int cpu)
1369 {
1370 	struct ixgbe_hw *hw = &adapter->hw;
1371 	u32 rxctrl = 0;
1372 	u8 reg_idx = rx_ring->reg_idx;
1373 
1374 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1375 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1376 
1377 	switch (hw->mac.type) {
1378 	case ixgbe_mac_82599EB:
1379 	case ixgbe_mac_X540:
1380 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1381 		break;
1382 	default:
1383 		break;
1384 	}
1385 
1386 	/*
1387 	 * We can enable relaxed ordering for reads, but not writes when
1388 	 * DCA is enabled.  This is due to a known issue in some chipsets
1389 	 * which will cause the DCA tag to be cleared.
1390 	 */
1391 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1392 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1393 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1394 
1395 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1396 }
1397 
1398 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1399 {
1400 	struct ixgbe_adapter *adapter = q_vector->adapter;
1401 	struct ixgbe_ring *ring;
1402 	int cpu = get_cpu();
1403 
1404 	if (q_vector->cpu == cpu)
1405 		goto out_no_update;
1406 
1407 	ixgbe_for_each_ring(ring, q_vector->tx)
1408 		ixgbe_update_tx_dca(adapter, ring, cpu);
1409 
1410 	ixgbe_for_each_ring(ring, q_vector->rx)
1411 		ixgbe_update_rx_dca(adapter, ring, cpu);
1412 
1413 	q_vector->cpu = cpu;
1414 out_no_update:
1415 	put_cpu();
1416 }
1417 
1418 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1419 {
1420 	int i;
1421 
1422 	/* always use CB2 mode, difference is masked in the CB driver */
1423 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1424 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1425 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1426 	else
1427 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1428 				IXGBE_DCA_CTRL_DCA_DISABLE);
1429 
1430 	for (i = 0; i < adapter->num_q_vectors; i++) {
1431 		adapter->q_vector[i]->cpu = -1;
1432 		ixgbe_update_dca(adapter->q_vector[i]);
1433 	}
1434 }
1435 
1436 static int __ixgbe_notify_dca(struct device *dev, void *data)
1437 {
1438 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1439 	unsigned long event = *(unsigned long *)data;
1440 
1441 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1442 		return 0;
1443 
1444 	switch (event) {
1445 	case DCA_PROVIDER_ADD:
1446 		/* if we're already enabled, don't do it again */
1447 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1448 			break;
1449 		if (dca_add_requester(dev) == 0) {
1450 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1451 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1452 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1453 			break;
1454 		}
1455 		/* fall through - DCA is disabled. */
1456 	case DCA_PROVIDER_REMOVE:
1457 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1458 			dca_remove_requester(dev);
1459 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1460 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1461 					IXGBE_DCA_CTRL_DCA_DISABLE);
1462 		}
1463 		break;
1464 	}
1465 
1466 	return 0;
1467 }
1468 
1469 #endif /* CONFIG_IXGBE_DCA */
1470 
1471 #define IXGBE_RSS_L4_TYPES_MASK \
1472 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1473 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1474 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1475 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1476 
1477 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1478 				 union ixgbe_adv_rx_desc *rx_desc,
1479 				 struct sk_buff *skb)
1480 {
1481 	u16 rss_type;
1482 
1483 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1484 		return;
1485 
1486 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1487 		   IXGBE_RXDADV_RSSTYPE_MASK;
1488 
1489 	if (!rss_type)
1490 		return;
1491 
1492 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1493 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1494 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1495 }
1496 
1497 #ifdef IXGBE_FCOE
1498 /**
1499  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1500  * @ring: structure containing ring specific data
1501  * @rx_desc: advanced rx descriptor
1502  *
1503  * Returns : true if it is FCoE pkt
1504  */
1505 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1506 				    union ixgbe_adv_rx_desc *rx_desc)
1507 {
1508 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1509 
1510 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1511 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1512 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1513 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1514 }
1515 
1516 #endif /* IXGBE_FCOE */
1517 /**
1518  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1519  * @ring: structure containing ring specific data
1520  * @rx_desc: current Rx descriptor being processed
1521  * @skb: skb currently being received and modified
1522  **/
1523 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1524 				     union ixgbe_adv_rx_desc *rx_desc,
1525 				     struct sk_buff *skb)
1526 {
1527 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1528 	bool encap_pkt = false;
1529 
1530 	skb_checksum_none_assert(skb);
1531 
1532 	/* Rx csum disabled */
1533 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1534 		return;
1535 
1536 	/* check for VXLAN and Geneve packets */
1537 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1538 		encap_pkt = true;
1539 		skb->encapsulation = 1;
1540 	}
1541 
1542 	/* if IP and error */
1543 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1544 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1545 		ring->rx_stats.csum_err++;
1546 		return;
1547 	}
1548 
1549 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1550 		return;
1551 
1552 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1553 		/*
1554 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1555 		 * checksum errors.
1556 		 */
1557 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1558 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1559 			return;
1560 
1561 		ring->rx_stats.csum_err++;
1562 		return;
1563 	}
1564 
1565 	/* It must be a TCP or UDP packet with a valid checksum */
1566 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1567 	if (encap_pkt) {
1568 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1569 			return;
1570 
1571 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1572 			skb->ip_summed = CHECKSUM_NONE;
1573 			return;
1574 		}
1575 		/* If we checked the outer header let the stack know */
1576 		skb->csum_level = 1;
1577 	}
1578 }
1579 
1580 static inline unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1581 {
1582 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1583 }
1584 
1585 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1586 				    struct ixgbe_rx_buffer *bi)
1587 {
1588 	struct page *page = bi->page;
1589 	dma_addr_t dma;
1590 
1591 	/* since we are recycling buffers we should seldom need to alloc */
1592 	if (likely(page))
1593 		return true;
1594 
1595 	/* alloc new page for storage */
1596 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1597 	if (unlikely(!page)) {
1598 		rx_ring->rx_stats.alloc_rx_page_failed++;
1599 		return false;
1600 	}
1601 
1602 	/* map page for use */
1603 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1604 				 ixgbe_rx_pg_size(rx_ring),
1605 				 DMA_FROM_DEVICE,
1606 				 IXGBE_RX_DMA_ATTR);
1607 
1608 	/*
1609 	 * if mapping failed free memory back to system since
1610 	 * there isn't much point in holding memory we can't use
1611 	 */
1612 	if (dma_mapping_error(rx_ring->dev, dma)) {
1613 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1614 
1615 		rx_ring->rx_stats.alloc_rx_page_failed++;
1616 		return false;
1617 	}
1618 
1619 	bi->dma = dma;
1620 	bi->page = page;
1621 	bi->page_offset = ixgbe_rx_offset(rx_ring);
1622 	bi->pagecnt_bias = 1;
1623 
1624 	return true;
1625 }
1626 
1627 /**
1628  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1629  * @rx_ring: ring to place buffers on
1630  * @cleaned_count: number of buffers to replace
1631  **/
1632 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1633 {
1634 	union ixgbe_adv_rx_desc *rx_desc;
1635 	struct ixgbe_rx_buffer *bi;
1636 	u16 i = rx_ring->next_to_use;
1637 	u16 bufsz;
1638 
1639 	/* nothing to do */
1640 	if (!cleaned_count)
1641 		return;
1642 
1643 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1644 	bi = &rx_ring->rx_buffer_info[i];
1645 	i -= rx_ring->count;
1646 
1647 	bufsz = ixgbe_rx_bufsz(rx_ring);
1648 
1649 	do {
1650 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1651 			break;
1652 
1653 		/* sync the buffer for use by the device */
1654 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1655 						 bi->page_offset, bufsz,
1656 						 DMA_FROM_DEVICE);
1657 
1658 		/*
1659 		 * Refresh the desc even if buffer_addrs didn't change
1660 		 * because each write-back erases this info.
1661 		 */
1662 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1663 
1664 		rx_desc++;
1665 		bi++;
1666 		i++;
1667 		if (unlikely(!i)) {
1668 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1669 			bi = rx_ring->rx_buffer_info;
1670 			i -= rx_ring->count;
1671 		}
1672 
1673 		/* clear the length for the next_to_use descriptor */
1674 		rx_desc->wb.upper.length = 0;
1675 
1676 		cleaned_count--;
1677 	} while (cleaned_count);
1678 
1679 	i += rx_ring->count;
1680 
1681 	if (rx_ring->next_to_use != i) {
1682 		rx_ring->next_to_use = i;
1683 
1684 		/* update next to alloc since we have filled the ring */
1685 		rx_ring->next_to_alloc = i;
1686 
1687 		/* Force memory writes to complete before letting h/w
1688 		 * know there are new descriptors to fetch.  (Only
1689 		 * applicable for weak-ordered memory model archs,
1690 		 * such as IA-64).
1691 		 */
1692 		wmb();
1693 		writel(i, rx_ring->tail);
1694 	}
1695 }
1696 
1697 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1698 				   struct sk_buff *skb)
1699 {
1700 	u16 hdr_len = skb_headlen(skb);
1701 
1702 	/* set gso_size to avoid messing up TCP MSS */
1703 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1704 						 IXGBE_CB(skb)->append_cnt);
1705 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1706 }
1707 
1708 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1709 				   struct sk_buff *skb)
1710 {
1711 	/* if append_cnt is 0 then frame is not RSC */
1712 	if (!IXGBE_CB(skb)->append_cnt)
1713 		return;
1714 
1715 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1716 	rx_ring->rx_stats.rsc_flush++;
1717 
1718 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1719 
1720 	/* gso_size is computed using append_cnt so always clear it last */
1721 	IXGBE_CB(skb)->append_cnt = 0;
1722 }
1723 
1724 /**
1725  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1726  * @rx_ring: rx descriptor ring packet is being transacted on
1727  * @rx_desc: pointer to the EOP Rx descriptor
1728  * @skb: pointer to current skb being populated
1729  *
1730  * This function checks the ring, descriptor, and packet information in
1731  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1732  * other fields within the skb.
1733  **/
1734 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1735 				     union ixgbe_adv_rx_desc *rx_desc,
1736 				     struct sk_buff *skb)
1737 {
1738 	struct net_device *dev = rx_ring->netdev;
1739 	u32 flags = rx_ring->q_vector->adapter->flags;
1740 
1741 	ixgbe_update_rsc_stats(rx_ring, skb);
1742 
1743 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1744 
1745 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1746 
1747 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1748 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1749 
1750 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1751 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1752 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1753 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1754 	}
1755 
1756 	skb_record_rx_queue(skb, rx_ring->queue_index);
1757 
1758 	skb->protocol = eth_type_trans(skb, dev);
1759 }
1760 
1761 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1762 			 struct sk_buff *skb)
1763 {
1764 	napi_gro_receive(&q_vector->napi, skb);
1765 }
1766 
1767 /**
1768  * ixgbe_is_non_eop - process handling of non-EOP buffers
1769  * @rx_ring: Rx ring being processed
1770  * @rx_desc: Rx descriptor for current buffer
1771  * @skb: Current socket buffer containing buffer in progress
1772  *
1773  * This function updates next to clean.  If the buffer is an EOP buffer
1774  * this function exits returning false, otherwise it will place the
1775  * sk_buff in the next buffer to be chained and return true indicating
1776  * that this is in fact a non-EOP buffer.
1777  **/
1778 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1779 			     union ixgbe_adv_rx_desc *rx_desc,
1780 			     struct sk_buff *skb)
1781 {
1782 	u32 ntc = rx_ring->next_to_clean + 1;
1783 
1784 	/* fetch, update, and store next to clean */
1785 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1786 	rx_ring->next_to_clean = ntc;
1787 
1788 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1789 
1790 	/* update RSC append count if present */
1791 	if (ring_is_rsc_enabled(rx_ring)) {
1792 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1793 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1794 
1795 		if (unlikely(rsc_enabled)) {
1796 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1797 
1798 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1799 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1800 
1801 			/* update ntc based on RSC value */
1802 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1803 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1804 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1805 		}
1806 	}
1807 
1808 	/* if we are the last buffer then there is nothing else to do */
1809 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1810 		return false;
1811 
1812 	/* place skb in next buffer to be received */
1813 	rx_ring->rx_buffer_info[ntc].skb = skb;
1814 	rx_ring->rx_stats.non_eop_descs++;
1815 
1816 	return true;
1817 }
1818 
1819 /**
1820  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1821  * @rx_ring: rx descriptor ring packet is being transacted on
1822  * @skb: pointer to current skb being adjusted
1823  *
1824  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1825  * main difference between this version and the original function is that
1826  * this function can make several assumptions about the state of things
1827  * that allow for significant optimizations versus the standard function.
1828  * As a result we can do things like drop a frag and maintain an accurate
1829  * truesize for the skb.
1830  */
1831 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1832 			    struct sk_buff *skb)
1833 {
1834 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1835 	unsigned char *va;
1836 	unsigned int pull_len;
1837 
1838 	/*
1839 	 * it is valid to use page_address instead of kmap since we are
1840 	 * working with pages allocated out of the lomem pool per
1841 	 * alloc_page(GFP_ATOMIC)
1842 	 */
1843 	va = skb_frag_address(frag);
1844 
1845 	/*
1846 	 * we need the header to contain the greater of either ETH_HLEN or
1847 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1848 	 */
1849 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1850 
1851 	/* align pull length to size of long to optimize memcpy performance */
1852 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1853 
1854 	/* update all of the pointers */
1855 	skb_frag_size_sub(frag, pull_len);
1856 	frag->page_offset += pull_len;
1857 	skb->data_len -= pull_len;
1858 	skb->tail += pull_len;
1859 }
1860 
1861 /**
1862  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1863  * @rx_ring: rx descriptor ring packet is being transacted on
1864  * @skb: pointer to current skb being updated
1865  *
1866  * This function provides a basic DMA sync up for the first fragment of an
1867  * skb.  The reason for doing this is that the first fragment cannot be
1868  * unmapped until we have reached the end of packet descriptor for a buffer
1869  * chain.
1870  */
1871 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1872 				struct sk_buff *skb)
1873 {
1874 	/* if the page was released unmap it, else just sync our portion */
1875 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1876 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1877 				     ixgbe_rx_pg_size(rx_ring),
1878 				     DMA_FROM_DEVICE,
1879 				     IXGBE_RX_DMA_ATTR);
1880 	} else {
1881 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1882 
1883 		dma_sync_single_range_for_cpu(rx_ring->dev,
1884 					      IXGBE_CB(skb)->dma,
1885 					      frag->page_offset,
1886 					      skb_frag_size(frag),
1887 					      DMA_FROM_DEVICE);
1888 	}
1889 }
1890 
1891 /**
1892  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1893  * @rx_ring: rx descriptor ring packet is being transacted on
1894  * @rx_desc: pointer to the EOP Rx descriptor
1895  * @skb: pointer to current skb being fixed
1896  *
1897  * Check if the skb is valid in the XDP case it will be an error pointer.
1898  * Return true in this case to abort processing and advance to next
1899  * descriptor.
1900  *
1901  * Check for corrupted packet headers caused by senders on the local L2
1902  * embedded NIC switch not setting up their Tx Descriptors right.  These
1903  * should be very rare.
1904  *
1905  * Also address the case where we are pulling data in on pages only
1906  * and as such no data is present in the skb header.
1907  *
1908  * In addition if skb is not at least 60 bytes we need to pad it so that
1909  * it is large enough to qualify as a valid Ethernet frame.
1910  *
1911  * Returns true if an error was encountered and skb was freed.
1912  **/
1913 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1914 				  union ixgbe_adv_rx_desc *rx_desc,
1915 				  struct sk_buff *skb)
1916 {
1917 	struct net_device *netdev = rx_ring->netdev;
1918 
1919 	/* XDP packets use error pointer so abort at this point */
1920 	if (IS_ERR(skb))
1921 		return true;
1922 
1923 	/* verify that the packet does not have any known errors */
1924 	if (unlikely(ixgbe_test_staterr(rx_desc,
1925 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1926 	    !(netdev->features & NETIF_F_RXALL))) {
1927 		dev_kfree_skb_any(skb);
1928 		return true;
1929 	}
1930 
1931 	/* place header in linear portion of buffer */
1932 	if (!skb_headlen(skb))
1933 		ixgbe_pull_tail(rx_ring, skb);
1934 
1935 #ifdef IXGBE_FCOE
1936 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1937 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1938 		return false;
1939 
1940 #endif
1941 	/* if eth_skb_pad returns an error the skb was freed */
1942 	if (eth_skb_pad(skb))
1943 		return true;
1944 
1945 	return false;
1946 }
1947 
1948 /**
1949  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1950  * @rx_ring: rx descriptor ring to store buffers on
1951  * @old_buff: donor buffer to have page reused
1952  *
1953  * Synchronizes page for reuse by the adapter
1954  **/
1955 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1956 				struct ixgbe_rx_buffer *old_buff)
1957 {
1958 	struct ixgbe_rx_buffer *new_buff;
1959 	u16 nta = rx_ring->next_to_alloc;
1960 
1961 	new_buff = &rx_ring->rx_buffer_info[nta];
1962 
1963 	/* update, and store next to alloc */
1964 	nta++;
1965 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1966 
1967 	/* Transfer page from old buffer to new buffer.
1968 	 * Move each member individually to avoid possible store
1969 	 * forwarding stalls and unnecessary copy of skb.
1970 	 */
1971 	new_buff->dma		= old_buff->dma;
1972 	new_buff->page		= old_buff->page;
1973 	new_buff->page_offset	= old_buff->page_offset;
1974 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
1975 }
1976 
1977 static inline bool ixgbe_page_is_reserved(struct page *page)
1978 {
1979 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
1980 }
1981 
1982 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer)
1983 {
1984 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
1985 	struct page *page = rx_buffer->page;
1986 
1987 	/* avoid re-using remote pages */
1988 	if (unlikely(ixgbe_page_is_reserved(page)))
1989 		return false;
1990 
1991 #if (PAGE_SIZE < 8192)
1992 	/* if we are only owner of page we can reuse it */
1993 	if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
1994 		return false;
1995 #else
1996 	/* The last offset is a bit aggressive in that we assume the
1997 	 * worst case of FCoE being enabled and using a 3K buffer.
1998 	 * However this should have minimal impact as the 1K extra is
1999 	 * still less than one buffer in size.
2000 	 */
2001 #define IXGBE_LAST_OFFSET \
2002 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
2003 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
2004 		return false;
2005 #endif
2006 
2007 	/* If we have drained the page fragment pool we need to update
2008 	 * the pagecnt_bias and page count so that we fully restock the
2009 	 * number of references the driver holds.
2010 	 */
2011 	if (unlikely(!pagecnt_bias)) {
2012 		page_ref_add(page, USHRT_MAX);
2013 		rx_buffer->pagecnt_bias = USHRT_MAX;
2014 	}
2015 
2016 	return true;
2017 }
2018 
2019 /**
2020  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
2021  * @rx_ring: rx descriptor ring to transact packets on
2022  * @rx_buffer: buffer containing page to add
2023  * @rx_desc: descriptor containing length of buffer written by hardware
2024  * @skb: sk_buff to place the data into
2025  *
2026  * This function will add the data contained in rx_buffer->page to the skb.
2027  * This is done either through a direct copy if the data in the buffer is
2028  * less than the skb header size, otherwise it will just attach the page as
2029  * a frag to the skb.
2030  *
2031  * The function will then update the page offset if necessary and return
2032  * true if the buffer can be reused by the adapter.
2033  **/
2034 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
2035 			      struct ixgbe_rx_buffer *rx_buffer,
2036 			      struct sk_buff *skb,
2037 			      unsigned int size)
2038 {
2039 #if (PAGE_SIZE < 8192)
2040 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2041 #else
2042 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2043 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2044 				SKB_DATA_ALIGN(size);
2045 #endif
2046 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2047 			rx_buffer->page_offset, size, truesize);
2048 #if (PAGE_SIZE < 8192)
2049 	rx_buffer->page_offset ^= truesize;
2050 #else
2051 	rx_buffer->page_offset += truesize;
2052 #endif
2053 }
2054 
2055 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2056 						   union ixgbe_adv_rx_desc *rx_desc,
2057 						   struct sk_buff **skb,
2058 						   const unsigned int size)
2059 {
2060 	struct ixgbe_rx_buffer *rx_buffer;
2061 
2062 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2063 	prefetchw(rx_buffer->page);
2064 	*skb = rx_buffer->skb;
2065 
2066 	/* Delay unmapping of the first packet. It carries the header
2067 	 * information, HW may still access the header after the writeback.
2068 	 * Only unmap it when EOP is reached
2069 	 */
2070 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2071 		if (!*skb)
2072 			goto skip_sync;
2073 	} else {
2074 		if (*skb)
2075 			ixgbe_dma_sync_frag(rx_ring, *skb);
2076 	}
2077 
2078 	/* we are reusing so sync this buffer for CPU use */
2079 	dma_sync_single_range_for_cpu(rx_ring->dev,
2080 				      rx_buffer->dma,
2081 				      rx_buffer->page_offset,
2082 				      size,
2083 				      DMA_FROM_DEVICE);
2084 skip_sync:
2085 	rx_buffer->pagecnt_bias--;
2086 
2087 	return rx_buffer;
2088 }
2089 
2090 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2091 				struct ixgbe_rx_buffer *rx_buffer,
2092 				struct sk_buff *skb)
2093 {
2094 	if (ixgbe_can_reuse_rx_page(rx_buffer)) {
2095 		/* hand second half of page back to the ring */
2096 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2097 	} else {
2098 		if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2099 			/* the page has been released from the ring */
2100 			IXGBE_CB(skb)->page_released = true;
2101 		} else {
2102 			/* we are not reusing the buffer so unmap it */
2103 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2104 					     ixgbe_rx_pg_size(rx_ring),
2105 					     DMA_FROM_DEVICE,
2106 					     IXGBE_RX_DMA_ATTR);
2107 		}
2108 		__page_frag_cache_drain(rx_buffer->page,
2109 					rx_buffer->pagecnt_bias);
2110 	}
2111 
2112 	/* clear contents of rx_buffer */
2113 	rx_buffer->page = NULL;
2114 	rx_buffer->skb = NULL;
2115 }
2116 
2117 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2118 					   struct ixgbe_rx_buffer *rx_buffer,
2119 					   struct xdp_buff *xdp,
2120 					   union ixgbe_adv_rx_desc *rx_desc)
2121 {
2122 	unsigned int size = xdp->data_end - xdp->data;
2123 #if (PAGE_SIZE < 8192)
2124 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2125 #else
2126 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2127 					       xdp->data_hard_start);
2128 #endif
2129 	struct sk_buff *skb;
2130 
2131 	/* prefetch first cache line of first page */
2132 	prefetch(xdp->data);
2133 #if L1_CACHE_BYTES < 128
2134 	prefetch(xdp->data + L1_CACHE_BYTES);
2135 #endif
2136 
2137 	/* allocate a skb to store the frags */
2138 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2139 	if (unlikely(!skb))
2140 		return NULL;
2141 
2142 	if (size > IXGBE_RX_HDR_SIZE) {
2143 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2144 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2145 
2146 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2147 				xdp->data - page_address(rx_buffer->page),
2148 				size, truesize);
2149 #if (PAGE_SIZE < 8192)
2150 		rx_buffer->page_offset ^= truesize;
2151 #else
2152 		rx_buffer->page_offset += truesize;
2153 #endif
2154 	} else {
2155 		memcpy(__skb_put(skb, size),
2156 		       xdp->data, ALIGN(size, sizeof(long)));
2157 		rx_buffer->pagecnt_bias++;
2158 	}
2159 
2160 	return skb;
2161 }
2162 
2163 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2164 				       struct ixgbe_rx_buffer *rx_buffer,
2165 				       struct xdp_buff *xdp,
2166 				       union ixgbe_adv_rx_desc *rx_desc)
2167 {
2168 #if (PAGE_SIZE < 8192)
2169 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2170 #else
2171 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2172 				SKB_DATA_ALIGN(xdp->data_end -
2173 					       xdp->data_hard_start);
2174 #endif
2175 	struct sk_buff *skb;
2176 
2177 	/* prefetch first cache line of first page */
2178 	prefetch(xdp->data);
2179 #if L1_CACHE_BYTES < 128
2180 	prefetch(xdp->data + L1_CACHE_BYTES);
2181 #endif
2182 
2183 	/* build an skb to around the page buffer */
2184 	skb = build_skb(xdp->data_hard_start, truesize);
2185 	if (unlikely(!skb))
2186 		return NULL;
2187 
2188 	/* update pointers within the skb to store the data */
2189 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
2190 	__skb_put(skb, xdp->data_end - xdp->data);
2191 
2192 	/* record DMA address if this is the start of a chain of buffers */
2193 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2194 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2195 
2196 	/* update buffer offset */
2197 #if (PAGE_SIZE < 8192)
2198 	rx_buffer->page_offset ^= truesize;
2199 #else
2200 	rx_buffer->page_offset += truesize;
2201 #endif
2202 
2203 	return skb;
2204 }
2205 
2206 #define IXGBE_XDP_PASS 0
2207 #define IXGBE_XDP_CONSUMED 1
2208 #define IXGBE_XDP_TX 2
2209 
2210 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
2211 			       struct xdp_buff *xdp);
2212 
2213 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2214 				     struct ixgbe_ring *rx_ring,
2215 				     struct xdp_buff *xdp)
2216 {
2217 	int err, result = IXGBE_XDP_PASS;
2218 	struct bpf_prog *xdp_prog;
2219 	u32 act;
2220 
2221 	rcu_read_lock();
2222 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2223 
2224 	if (!xdp_prog)
2225 		goto xdp_out;
2226 
2227 	act = bpf_prog_run_xdp(xdp_prog, xdp);
2228 	switch (act) {
2229 	case XDP_PASS:
2230 		break;
2231 	case XDP_TX:
2232 		result = ixgbe_xmit_xdp_ring(adapter, xdp);
2233 		break;
2234 	case XDP_REDIRECT:
2235 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2236 		if (!err)
2237 			result = IXGBE_XDP_TX;
2238 		else
2239 			result = IXGBE_XDP_CONSUMED;
2240 		break;
2241 	default:
2242 		bpf_warn_invalid_xdp_action(act);
2243 		/* fallthrough */
2244 	case XDP_ABORTED:
2245 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2246 		/* fallthrough -- handle aborts by dropping packet */
2247 	case XDP_DROP:
2248 		result = IXGBE_XDP_CONSUMED;
2249 		break;
2250 	}
2251 xdp_out:
2252 	rcu_read_unlock();
2253 	return ERR_PTR(-result);
2254 }
2255 
2256 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2257 				 struct ixgbe_rx_buffer *rx_buffer,
2258 				 unsigned int size)
2259 {
2260 #if (PAGE_SIZE < 8192)
2261 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2262 
2263 	rx_buffer->page_offset ^= truesize;
2264 #else
2265 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2266 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2267 				SKB_DATA_ALIGN(size);
2268 
2269 	rx_buffer->page_offset += truesize;
2270 #endif
2271 }
2272 
2273 /**
2274  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2275  * @q_vector: structure containing interrupt and ring information
2276  * @rx_ring: rx descriptor ring to transact packets on
2277  * @budget: Total limit on number of packets to process
2278  *
2279  * This function provides a "bounce buffer" approach to Rx interrupt
2280  * processing.  The advantage to this is that on systems that have
2281  * expensive overhead for IOMMU access this provides a means of avoiding
2282  * it by maintaining the mapping of the page to the syste.
2283  *
2284  * Returns amount of work completed
2285  **/
2286 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2287 			       struct ixgbe_ring *rx_ring,
2288 			       const int budget)
2289 {
2290 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2291 	struct ixgbe_adapter *adapter = q_vector->adapter;
2292 #ifdef IXGBE_FCOE
2293 	int ddp_bytes;
2294 	unsigned int mss = 0;
2295 #endif /* IXGBE_FCOE */
2296 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2297 	bool xdp_xmit = false;
2298 
2299 	while (likely(total_rx_packets < budget)) {
2300 		union ixgbe_adv_rx_desc *rx_desc;
2301 		struct ixgbe_rx_buffer *rx_buffer;
2302 		struct sk_buff *skb;
2303 		struct xdp_buff xdp;
2304 		unsigned int size;
2305 
2306 		/* return some buffers to hardware, one at a time is too slow */
2307 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2308 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2309 			cleaned_count = 0;
2310 		}
2311 
2312 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2313 		size = le16_to_cpu(rx_desc->wb.upper.length);
2314 		if (!size)
2315 			break;
2316 
2317 		/* This memory barrier is needed to keep us from reading
2318 		 * any other fields out of the rx_desc until we know the
2319 		 * descriptor has been written back
2320 		 */
2321 		dma_rmb();
2322 
2323 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size);
2324 
2325 		/* retrieve a buffer from the ring */
2326 		if (!skb) {
2327 			xdp.data = page_address(rx_buffer->page) +
2328 				   rx_buffer->page_offset;
2329 			xdp.data_hard_start = xdp.data -
2330 					      ixgbe_rx_offset(rx_ring);
2331 			xdp.data_end = xdp.data + size;
2332 
2333 			skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2334 		}
2335 
2336 		if (IS_ERR(skb)) {
2337 			if (PTR_ERR(skb) == -IXGBE_XDP_TX) {
2338 				xdp_xmit = true;
2339 				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2340 			} else {
2341 				rx_buffer->pagecnt_bias++;
2342 			}
2343 			total_rx_packets++;
2344 			total_rx_bytes += size;
2345 		} else if (skb) {
2346 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2347 		} else if (ring_uses_build_skb(rx_ring)) {
2348 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2349 					      &xdp, rx_desc);
2350 		} else {
2351 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2352 						  &xdp, rx_desc);
2353 		}
2354 
2355 		/* exit if we failed to retrieve a buffer */
2356 		if (!skb) {
2357 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2358 			rx_buffer->pagecnt_bias++;
2359 			break;
2360 		}
2361 
2362 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb);
2363 		cleaned_count++;
2364 
2365 		/* place incomplete frames back on ring for completion */
2366 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2367 			continue;
2368 
2369 		/* verify the packet layout is correct */
2370 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2371 			continue;
2372 
2373 		/* probably a little skewed due to removing CRC */
2374 		total_rx_bytes += skb->len;
2375 
2376 		/* populate checksum, timestamp, VLAN, and protocol */
2377 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2378 
2379 #ifdef IXGBE_FCOE
2380 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2381 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2382 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2383 			/* include DDPed FCoE data */
2384 			if (ddp_bytes > 0) {
2385 				if (!mss) {
2386 					mss = rx_ring->netdev->mtu -
2387 						sizeof(struct fcoe_hdr) -
2388 						sizeof(struct fc_frame_header) -
2389 						sizeof(struct fcoe_crc_eof);
2390 					if (mss > 512)
2391 						mss &= ~511;
2392 				}
2393 				total_rx_bytes += ddp_bytes;
2394 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2395 								 mss);
2396 			}
2397 			if (!ddp_bytes) {
2398 				dev_kfree_skb_any(skb);
2399 				continue;
2400 			}
2401 		}
2402 
2403 #endif /* IXGBE_FCOE */
2404 		ixgbe_rx_skb(q_vector, skb);
2405 
2406 		/* update budget accounting */
2407 		total_rx_packets++;
2408 	}
2409 
2410 	if (xdp_xmit) {
2411 		struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
2412 
2413 		/* Force memory writes to complete before letting h/w
2414 		 * know there are new descriptors to fetch.
2415 		 */
2416 		wmb();
2417 		writel(ring->next_to_use, ring->tail);
2418 
2419 		xdp_do_flush_map();
2420 	}
2421 
2422 	u64_stats_update_begin(&rx_ring->syncp);
2423 	rx_ring->stats.packets += total_rx_packets;
2424 	rx_ring->stats.bytes += total_rx_bytes;
2425 	u64_stats_update_end(&rx_ring->syncp);
2426 	q_vector->rx.total_packets += total_rx_packets;
2427 	q_vector->rx.total_bytes += total_rx_bytes;
2428 
2429 	return total_rx_packets;
2430 }
2431 
2432 /**
2433  * ixgbe_configure_msix - Configure MSI-X hardware
2434  * @adapter: board private structure
2435  *
2436  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2437  * interrupts.
2438  **/
2439 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2440 {
2441 	struct ixgbe_q_vector *q_vector;
2442 	int v_idx;
2443 	u32 mask;
2444 
2445 	/* Populate MSIX to EITR Select */
2446 	if (adapter->num_vfs > 32) {
2447 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2448 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2449 	}
2450 
2451 	/*
2452 	 * Populate the IVAR table and set the ITR values to the
2453 	 * corresponding register.
2454 	 */
2455 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2456 		struct ixgbe_ring *ring;
2457 		q_vector = adapter->q_vector[v_idx];
2458 
2459 		ixgbe_for_each_ring(ring, q_vector->rx)
2460 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2461 
2462 		ixgbe_for_each_ring(ring, q_vector->tx)
2463 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2464 
2465 		ixgbe_write_eitr(q_vector);
2466 	}
2467 
2468 	switch (adapter->hw.mac.type) {
2469 	case ixgbe_mac_82598EB:
2470 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2471 			       v_idx);
2472 		break;
2473 	case ixgbe_mac_82599EB:
2474 	case ixgbe_mac_X540:
2475 	case ixgbe_mac_X550:
2476 	case ixgbe_mac_X550EM_x:
2477 	case ixgbe_mac_x550em_a:
2478 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2479 		break;
2480 	default:
2481 		break;
2482 	}
2483 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2484 
2485 	/* set up to autoclear timer, and the vectors */
2486 	mask = IXGBE_EIMS_ENABLE_MASK;
2487 	mask &= ~(IXGBE_EIMS_OTHER |
2488 		  IXGBE_EIMS_MAILBOX |
2489 		  IXGBE_EIMS_LSC);
2490 
2491 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2492 }
2493 
2494 enum latency_range {
2495 	lowest_latency = 0,
2496 	low_latency = 1,
2497 	bulk_latency = 2,
2498 	latency_invalid = 255
2499 };
2500 
2501 /**
2502  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2503  * @q_vector: structure containing interrupt and ring information
2504  * @ring_container: structure containing ring performance data
2505  *
2506  *      Stores a new ITR value based on packets and byte
2507  *      counts during the last interrupt.  The advantage of per interrupt
2508  *      computation is faster updates and more accurate ITR for the current
2509  *      traffic pattern.  Constants in this function were computed
2510  *      based on theoretical maximum wire speed and thresholds were set based
2511  *      on testing data as well as attempting to minimize response time
2512  *      while increasing bulk throughput.
2513  *      this functionality is controlled by the InterruptThrottleRate module
2514  *      parameter (see ixgbe_param.c)
2515  **/
2516 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2517 			     struct ixgbe_ring_container *ring_container)
2518 {
2519 	int bytes = ring_container->total_bytes;
2520 	int packets = ring_container->total_packets;
2521 	u32 timepassed_us;
2522 	u64 bytes_perint;
2523 	u8 itr_setting = ring_container->itr;
2524 
2525 	if (packets == 0)
2526 		return;
2527 
2528 	/* simple throttlerate management
2529 	 *   0-10MB/s   lowest (100000 ints/s)
2530 	 *  10-20MB/s   low    (20000 ints/s)
2531 	 *  20-1249MB/s bulk   (12000 ints/s)
2532 	 */
2533 	/* what was last interrupt timeslice? */
2534 	timepassed_us = q_vector->itr >> 2;
2535 	if (timepassed_us == 0)
2536 		return;
2537 
2538 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2539 
2540 	switch (itr_setting) {
2541 	case lowest_latency:
2542 		if (bytes_perint > 10)
2543 			itr_setting = low_latency;
2544 		break;
2545 	case low_latency:
2546 		if (bytes_perint > 20)
2547 			itr_setting = bulk_latency;
2548 		else if (bytes_perint <= 10)
2549 			itr_setting = lowest_latency;
2550 		break;
2551 	case bulk_latency:
2552 		if (bytes_perint <= 20)
2553 			itr_setting = low_latency;
2554 		break;
2555 	}
2556 
2557 	/* clear work counters since we have the values we need */
2558 	ring_container->total_bytes = 0;
2559 	ring_container->total_packets = 0;
2560 
2561 	/* write updated itr to ring container */
2562 	ring_container->itr = itr_setting;
2563 }
2564 
2565 /**
2566  * ixgbe_write_eitr - write EITR register in hardware specific way
2567  * @q_vector: structure containing interrupt and ring information
2568  *
2569  * This function is made to be called by ethtool and by the driver
2570  * when it needs to update EITR registers at runtime.  Hardware
2571  * specific quirks/differences are taken care of here.
2572  */
2573 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2574 {
2575 	struct ixgbe_adapter *adapter = q_vector->adapter;
2576 	struct ixgbe_hw *hw = &adapter->hw;
2577 	int v_idx = q_vector->v_idx;
2578 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2579 
2580 	switch (adapter->hw.mac.type) {
2581 	case ixgbe_mac_82598EB:
2582 		/* must write high and low 16 bits to reset counter */
2583 		itr_reg |= (itr_reg << 16);
2584 		break;
2585 	case ixgbe_mac_82599EB:
2586 	case ixgbe_mac_X540:
2587 	case ixgbe_mac_X550:
2588 	case ixgbe_mac_X550EM_x:
2589 	case ixgbe_mac_x550em_a:
2590 		/*
2591 		 * set the WDIS bit to not clear the timer bits and cause an
2592 		 * immediate assertion of the interrupt
2593 		 */
2594 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2595 		break;
2596 	default:
2597 		break;
2598 	}
2599 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2600 }
2601 
2602 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2603 {
2604 	u32 new_itr = q_vector->itr;
2605 	u8 current_itr;
2606 
2607 	ixgbe_update_itr(q_vector, &q_vector->tx);
2608 	ixgbe_update_itr(q_vector, &q_vector->rx);
2609 
2610 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2611 
2612 	switch (current_itr) {
2613 	/* counts and packets in update_itr are dependent on these numbers */
2614 	case lowest_latency:
2615 		new_itr = IXGBE_100K_ITR;
2616 		break;
2617 	case low_latency:
2618 		new_itr = IXGBE_20K_ITR;
2619 		break;
2620 	case bulk_latency:
2621 		new_itr = IXGBE_12K_ITR;
2622 		break;
2623 	default:
2624 		break;
2625 	}
2626 
2627 	if (new_itr != q_vector->itr) {
2628 		/* do an exponential smoothing */
2629 		new_itr = (10 * new_itr * q_vector->itr) /
2630 			  ((9 * new_itr) + q_vector->itr);
2631 
2632 		/* save the algorithm value here */
2633 		q_vector->itr = new_itr;
2634 
2635 		ixgbe_write_eitr(q_vector);
2636 	}
2637 }
2638 
2639 /**
2640  * ixgbe_check_overtemp_subtask - check for over temperature
2641  * @adapter: pointer to adapter
2642  **/
2643 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2644 {
2645 	struct ixgbe_hw *hw = &adapter->hw;
2646 	u32 eicr = adapter->interrupt_event;
2647 	s32 rc;
2648 
2649 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2650 		return;
2651 
2652 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2653 		return;
2654 
2655 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2656 
2657 	switch (hw->device_id) {
2658 	case IXGBE_DEV_ID_82599_T3_LOM:
2659 		/*
2660 		 * Since the warning interrupt is for both ports
2661 		 * we don't have to check if:
2662 		 *  - This interrupt wasn't for our port.
2663 		 *  - We may have missed the interrupt so always have to
2664 		 *    check if we  got a LSC
2665 		 */
2666 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2667 		    !(eicr & IXGBE_EICR_LSC))
2668 			return;
2669 
2670 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2671 			u32 speed;
2672 			bool link_up = false;
2673 
2674 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2675 
2676 			if (link_up)
2677 				return;
2678 		}
2679 
2680 		/* Check if this is not due to overtemp */
2681 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2682 			return;
2683 
2684 		break;
2685 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2686 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2687 		rc = hw->phy.ops.check_overtemp(hw);
2688 		if (rc != IXGBE_ERR_OVERTEMP)
2689 			return;
2690 		break;
2691 	default:
2692 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2693 			return;
2694 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2695 			return;
2696 		break;
2697 	}
2698 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2699 
2700 	adapter->interrupt_event = 0;
2701 }
2702 
2703 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2704 {
2705 	struct ixgbe_hw *hw = &adapter->hw;
2706 
2707 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2708 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2709 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2710 		/* write to clear the interrupt */
2711 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2712 	}
2713 }
2714 
2715 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2716 {
2717 	struct ixgbe_hw *hw = &adapter->hw;
2718 
2719 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2720 		return;
2721 
2722 	switch (adapter->hw.mac.type) {
2723 	case ixgbe_mac_82599EB:
2724 		/*
2725 		 * Need to check link state so complete overtemp check
2726 		 * on service task
2727 		 */
2728 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2729 		     (eicr & IXGBE_EICR_LSC)) &&
2730 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2731 			adapter->interrupt_event = eicr;
2732 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2733 			ixgbe_service_event_schedule(adapter);
2734 			return;
2735 		}
2736 		return;
2737 	case ixgbe_mac_x550em_a:
2738 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2739 			adapter->interrupt_event = eicr;
2740 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2741 			ixgbe_service_event_schedule(adapter);
2742 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2743 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2744 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2745 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2746 		}
2747 		return;
2748 	case ixgbe_mac_X550:
2749 	case ixgbe_mac_X540:
2750 		if (!(eicr & IXGBE_EICR_TS))
2751 			return;
2752 		break;
2753 	default:
2754 		return;
2755 	}
2756 
2757 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2758 }
2759 
2760 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2761 {
2762 	switch (hw->mac.type) {
2763 	case ixgbe_mac_82598EB:
2764 		if (hw->phy.type == ixgbe_phy_nl)
2765 			return true;
2766 		return false;
2767 	case ixgbe_mac_82599EB:
2768 	case ixgbe_mac_X550EM_x:
2769 	case ixgbe_mac_x550em_a:
2770 		switch (hw->mac.ops.get_media_type(hw)) {
2771 		case ixgbe_media_type_fiber:
2772 		case ixgbe_media_type_fiber_qsfp:
2773 			return true;
2774 		default:
2775 			return false;
2776 		}
2777 	default:
2778 		return false;
2779 	}
2780 }
2781 
2782 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2783 {
2784 	struct ixgbe_hw *hw = &adapter->hw;
2785 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2786 
2787 	if (!ixgbe_is_sfp(hw))
2788 		return;
2789 
2790 	/* Later MAC's use different SDP */
2791 	if (hw->mac.type >= ixgbe_mac_X540)
2792 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2793 
2794 	if (eicr & eicr_mask) {
2795 		/* Clear the interrupt */
2796 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2797 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2798 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2799 			adapter->sfp_poll_time = 0;
2800 			ixgbe_service_event_schedule(adapter);
2801 		}
2802 	}
2803 
2804 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2805 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2806 		/* Clear the interrupt */
2807 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2808 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2809 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2810 			ixgbe_service_event_schedule(adapter);
2811 		}
2812 	}
2813 }
2814 
2815 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2816 {
2817 	struct ixgbe_hw *hw = &adapter->hw;
2818 
2819 	adapter->lsc_int++;
2820 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2821 	adapter->link_check_timeout = jiffies;
2822 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2823 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2824 		IXGBE_WRITE_FLUSH(hw);
2825 		ixgbe_service_event_schedule(adapter);
2826 	}
2827 }
2828 
2829 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2830 					   u64 qmask)
2831 {
2832 	u32 mask;
2833 	struct ixgbe_hw *hw = &adapter->hw;
2834 
2835 	switch (hw->mac.type) {
2836 	case ixgbe_mac_82598EB:
2837 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2838 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2839 		break;
2840 	case ixgbe_mac_82599EB:
2841 	case ixgbe_mac_X540:
2842 	case ixgbe_mac_X550:
2843 	case ixgbe_mac_X550EM_x:
2844 	case ixgbe_mac_x550em_a:
2845 		mask = (qmask & 0xFFFFFFFF);
2846 		if (mask)
2847 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2848 		mask = (qmask >> 32);
2849 		if (mask)
2850 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2851 		break;
2852 	default:
2853 		break;
2854 	}
2855 	/* skip the flush */
2856 }
2857 
2858 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2859 					    u64 qmask)
2860 {
2861 	u32 mask;
2862 	struct ixgbe_hw *hw = &adapter->hw;
2863 
2864 	switch (hw->mac.type) {
2865 	case ixgbe_mac_82598EB:
2866 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2867 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2868 		break;
2869 	case ixgbe_mac_82599EB:
2870 	case ixgbe_mac_X540:
2871 	case ixgbe_mac_X550:
2872 	case ixgbe_mac_X550EM_x:
2873 	case ixgbe_mac_x550em_a:
2874 		mask = (qmask & 0xFFFFFFFF);
2875 		if (mask)
2876 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2877 		mask = (qmask >> 32);
2878 		if (mask)
2879 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2880 		break;
2881 	default:
2882 		break;
2883 	}
2884 	/* skip the flush */
2885 }
2886 
2887 /**
2888  * ixgbe_irq_enable - Enable default interrupt generation settings
2889  * @adapter: board private structure
2890  **/
2891 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2892 				    bool flush)
2893 {
2894 	struct ixgbe_hw *hw = &adapter->hw;
2895 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2896 
2897 	/* don't reenable LSC while waiting for link */
2898 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2899 		mask &= ~IXGBE_EIMS_LSC;
2900 
2901 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2902 		switch (adapter->hw.mac.type) {
2903 		case ixgbe_mac_82599EB:
2904 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
2905 			break;
2906 		case ixgbe_mac_X540:
2907 		case ixgbe_mac_X550:
2908 		case ixgbe_mac_X550EM_x:
2909 		case ixgbe_mac_x550em_a:
2910 			mask |= IXGBE_EIMS_TS;
2911 			break;
2912 		default:
2913 			break;
2914 		}
2915 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2916 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2917 	switch (adapter->hw.mac.type) {
2918 	case ixgbe_mac_82599EB:
2919 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2920 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
2921 		/* fall through */
2922 	case ixgbe_mac_X540:
2923 	case ixgbe_mac_X550:
2924 	case ixgbe_mac_X550EM_x:
2925 	case ixgbe_mac_x550em_a:
2926 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
2927 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
2928 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
2929 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
2930 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
2931 			mask |= IXGBE_EICR_GPI_SDP0_X540;
2932 		mask |= IXGBE_EIMS_ECC;
2933 		mask |= IXGBE_EIMS_MAILBOX;
2934 		break;
2935 	default:
2936 		break;
2937 	}
2938 
2939 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2940 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2941 		mask |= IXGBE_EIMS_FLOW_DIR;
2942 
2943 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2944 	if (queues)
2945 		ixgbe_irq_enable_queues(adapter, ~0);
2946 	if (flush)
2947 		IXGBE_WRITE_FLUSH(&adapter->hw);
2948 }
2949 
2950 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2951 {
2952 	struct ixgbe_adapter *adapter = data;
2953 	struct ixgbe_hw *hw = &adapter->hw;
2954 	u32 eicr;
2955 
2956 	/*
2957 	 * Workaround for Silicon errata.  Use clear-by-write instead
2958 	 * of clear-by-read.  Reading with EICS will return the
2959 	 * interrupt causes without clearing, which later be done
2960 	 * with the write to EICR.
2961 	 */
2962 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2963 
2964 	/* The lower 16bits of the EICR register are for the queue interrupts
2965 	 * which should be masked here in order to not accidentally clear them if
2966 	 * the bits are high when ixgbe_msix_other is called. There is a race
2967 	 * condition otherwise which results in possible performance loss
2968 	 * especially if the ixgbe_msix_other interrupt is triggering
2969 	 * consistently (as it would when PPS is turned on for the X540 device)
2970 	 */
2971 	eicr &= 0xFFFF0000;
2972 
2973 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2974 
2975 	if (eicr & IXGBE_EICR_LSC)
2976 		ixgbe_check_lsc(adapter);
2977 
2978 	if (eicr & IXGBE_EICR_MAILBOX)
2979 		ixgbe_msg_task(adapter);
2980 
2981 	switch (hw->mac.type) {
2982 	case ixgbe_mac_82599EB:
2983 	case ixgbe_mac_X540:
2984 	case ixgbe_mac_X550:
2985 	case ixgbe_mac_X550EM_x:
2986 	case ixgbe_mac_x550em_a:
2987 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
2988 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2989 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
2990 			ixgbe_service_event_schedule(adapter);
2991 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2992 					IXGBE_EICR_GPI_SDP0_X540);
2993 		}
2994 		if (eicr & IXGBE_EICR_ECC) {
2995 			e_info(link, "Received ECC Err, initiating reset\n");
2996 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
2997 			ixgbe_service_event_schedule(adapter);
2998 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2999 		}
3000 		/* Handle Flow Director Full threshold interrupt */
3001 		if (eicr & IXGBE_EICR_FLOW_DIR) {
3002 			int reinit_count = 0;
3003 			int i;
3004 			for (i = 0; i < adapter->num_tx_queues; i++) {
3005 				struct ixgbe_ring *ring = adapter->tx_ring[i];
3006 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3007 						       &ring->state))
3008 					reinit_count++;
3009 			}
3010 			if (reinit_count) {
3011 				/* no more flow director interrupts until after init */
3012 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3013 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3014 				ixgbe_service_event_schedule(adapter);
3015 			}
3016 		}
3017 		ixgbe_check_sfp_event(adapter, eicr);
3018 		ixgbe_check_overtemp_event(adapter, eicr);
3019 		break;
3020 	default:
3021 		break;
3022 	}
3023 
3024 	ixgbe_check_fan_failure(adapter, eicr);
3025 
3026 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3027 		ixgbe_ptp_check_pps_event(adapter);
3028 
3029 	/* re-enable the original interrupt state, no lsc, no queues */
3030 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3031 		ixgbe_irq_enable(adapter, false, false);
3032 
3033 	return IRQ_HANDLED;
3034 }
3035 
3036 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3037 {
3038 	struct ixgbe_q_vector *q_vector = data;
3039 
3040 	/* EIAM disabled interrupts (on this vector) for us */
3041 
3042 	if (q_vector->rx.ring || q_vector->tx.ring)
3043 		napi_schedule_irqoff(&q_vector->napi);
3044 
3045 	return IRQ_HANDLED;
3046 }
3047 
3048 /**
3049  * ixgbe_poll - NAPI Rx polling callback
3050  * @napi: structure for representing this polling device
3051  * @budget: how many packets driver is allowed to clean
3052  *
3053  * This function is used for legacy and MSI, NAPI mode
3054  **/
3055 int ixgbe_poll(struct napi_struct *napi, int budget)
3056 {
3057 	struct ixgbe_q_vector *q_vector =
3058 				container_of(napi, struct ixgbe_q_vector, napi);
3059 	struct ixgbe_adapter *adapter = q_vector->adapter;
3060 	struct ixgbe_ring *ring;
3061 	int per_ring_budget, work_done = 0;
3062 	bool clean_complete = true;
3063 
3064 #ifdef CONFIG_IXGBE_DCA
3065 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3066 		ixgbe_update_dca(q_vector);
3067 #endif
3068 
3069 	ixgbe_for_each_ring(ring, q_vector->tx) {
3070 		if (!ixgbe_clean_tx_irq(q_vector, ring, budget))
3071 			clean_complete = false;
3072 	}
3073 
3074 	/* Exit if we are called by netpoll */
3075 	if (budget <= 0)
3076 		return budget;
3077 
3078 	/* attempt to distribute budget to each queue fairly, but don't allow
3079 	 * the budget to go below 1 because we'll exit polling */
3080 	if (q_vector->rx.count > 1)
3081 		per_ring_budget = max(budget/q_vector->rx.count, 1);
3082 	else
3083 		per_ring_budget = budget;
3084 
3085 	ixgbe_for_each_ring(ring, q_vector->rx) {
3086 		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
3087 						 per_ring_budget);
3088 
3089 		work_done += cleaned;
3090 		if (cleaned >= per_ring_budget)
3091 			clean_complete = false;
3092 	}
3093 
3094 	/* If all work not completed, return budget and keep polling */
3095 	if (!clean_complete)
3096 		return budget;
3097 
3098 	/* all work done, exit the polling mode */
3099 	napi_complete_done(napi, work_done);
3100 	if (adapter->rx_itr_setting & 1)
3101 		ixgbe_set_itr(q_vector);
3102 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3103 		ixgbe_irq_enable_queues(adapter, BIT_ULL(q_vector->v_idx));
3104 
3105 	return min(work_done, budget - 1);
3106 }
3107 
3108 /**
3109  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3110  * @adapter: board private structure
3111  *
3112  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3113  * interrupts from the kernel.
3114  **/
3115 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3116 {
3117 	struct net_device *netdev = adapter->netdev;
3118 	unsigned int ri = 0, ti = 0;
3119 	int vector, err;
3120 
3121 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3122 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3123 		struct msix_entry *entry = &adapter->msix_entries[vector];
3124 
3125 		if (q_vector->tx.ring && q_vector->rx.ring) {
3126 			snprintf(q_vector->name, sizeof(q_vector->name),
3127 				 "%s-TxRx-%u", netdev->name, ri++);
3128 			ti++;
3129 		} else if (q_vector->rx.ring) {
3130 			snprintf(q_vector->name, sizeof(q_vector->name),
3131 				 "%s-rx-%u", netdev->name, ri++);
3132 		} else if (q_vector->tx.ring) {
3133 			snprintf(q_vector->name, sizeof(q_vector->name),
3134 				 "%s-tx-%u", netdev->name, ti++);
3135 		} else {
3136 			/* skip this unused q_vector */
3137 			continue;
3138 		}
3139 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3140 				  q_vector->name, q_vector);
3141 		if (err) {
3142 			e_err(probe, "request_irq failed for MSIX interrupt "
3143 			      "Error: %d\n", err);
3144 			goto free_queue_irqs;
3145 		}
3146 		/* If Flow Director is enabled, set interrupt affinity */
3147 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3148 			/* assign the mask for this irq */
3149 			irq_set_affinity_hint(entry->vector,
3150 					      &q_vector->affinity_mask);
3151 		}
3152 	}
3153 
3154 	err = request_irq(adapter->msix_entries[vector].vector,
3155 			  ixgbe_msix_other, 0, netdev->name, adapter);
3156 	if (err) {
3157 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3158 		goto free_queue_irqs;
3159 	}
3160 
3161 	return 0;
3162 
3163 free_queue_irqs:
3164 	while (vector) {
3165 		vector--;
3166 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
3167 				      NULL);
3168 		free_irq(adapter->msix_entries[vector].vector,
3169 			 adapter->q_vector[vector]);
3170 	}
3171 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3172 	pci_disable_msix(adapter->pdev);
3173 	kfree(adapter->msix_entries);
3174 	adapter->msix_entries = NULL;
3175 	return err;
3176 }
3177 
3178 /**
3179  * ixgbe_intr - legacy mode Interrupt Handler
3180  * @irq: interrupt number
3181  * @data: pointer to a network interface device structure
3182  **/
3183 static irqreturn_t ixgbe_intr(int irq, void *data)
3184 {
3185 	struct ixgbe_adapter *adapter = data;
3186 	struct ixgbe_hw *hw = &adapter->hw;
3187 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3188 	u32 eicr;
3189 
3190 	/*
3191 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3192 	 * before the read of EICR.
3193 	 */
3194 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3195 
3196 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3197 	 * therefore no explicit interrupt disable is necessary */
3198 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3199 	if (!eicr) {
3200 		/*
3201 		 * shared interrupt alert!
3202 		 * make sure interrupts are enabled because the read will
3203 		 * have disabled interrupts due to EIAM
3204 		 * finish the workaround of silicon errata on 82598.  Unmask
3205 		 * the interrupt that we masked before the EICR read.
3206 		 */
3207 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3208 			ixgbe_irq_enable(adapter, true, true);
3209 		return IRQ_NONE;	/* Not our interrupt */
3210 	}
3211 
3212 	if (eicr & IXGBE_EICR_LSC)
3213 		ixgbe_check_lsc(adapter);
3214 
3215 	switch (hw->mac.type) {
3216 	case ixgbe_mac_82599EB:
3217 		ixgbe_check_sfp_event(adapter, eicr);
3218 		/* Fall through */
3219 	case ixgbe_mac_X540:
3220 	case ixgbe_mac_X550:
3221 	case ixgbe_mac_X550EM_x:
3222 	case ixgbe_mac_x550em_a:
3223 		if (eicr & IXGBE_EICR_ECC) {
3224 			e_info(link, "Received ECC Err, initiating reset\n");
3225 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3226 			ixgbe_service_event_schedule(adapter);
3227 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3228 		}
3229 		ixgbe_check_overtemp_event(adapter, eicr);
3230 		break;
3231 	default:
3232 		break;
3233 	}
3234 
3235 	ixgbe_check_fan_failure(adapter, eicr);
3236 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3237 		ixgbe_ptp_check_pps_event(adapter);
3238 
3239 	/* would disable interrupts here but EIAM disabled it */
3240 	napi_schedule_irqoff(&q_vector->napi);
3241 
3242 	/*
3243 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3244 	 * ixgbe_poll will re-enable the queue interrupts
3245 	 */
3246 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3247 		ixgbe_irq_enable(adapter, false, false);
3248 
3249 	return IRQ_HANDLED;
3250 }
3251 
3252 /**
3253  * ixgbe_request_irq - initialize interrupts
3254  * @adapter: board private structure
3255  *
3256  * Attempts to configure interrupts using the best available
3257  * capabilities of the hardware and kernel.
3258  **/
3259 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3260 {
3261 	struct net_device *netdev = adapter->netdev;
3262 	int err;
3263 
3264 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3265 		err = ixgbe_request_msix_irqs(adapter);
3266 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3267 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3268 				  netdev->name, adapter);
3269 	else
3270 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3271 				  netdev->name, adapter);
3272 
3273 	if (err)
3274 		e_err(probe, "request_irq failed, Error %d\n", err);
3275 
3276 	return err;
3277 }
3278 
3279 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3280 {
3281 	int vector;
3282 
3283 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3284 		free_irq(adapter->pdev->irq, adapter);
3285 		return;
3286 	}
3287 
3288 	if (!adapter->msix_entries)
3289 		return;
3290 
3291 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3292 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3293 		struct msix_entry *entry = &adapter->msix_entries[vector];
3294 
3295 		/* free only the irqs that were actually requested */
3296 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3297 			continue;
3298 
3299 		/* clear the affinity_mask in the IRQ descriptor */
3300 		irq_set_affinity_hint(entry->vector, NULL);
3301 
3302 		free_irq(entry->vector, q_vector);
3303 	}
3304 
3305 	free_irq(adapter->msix_entries[vector].vector, adapter);
3306 }
3307 
3308 /**
3309  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3310  * @adapter: board private structure
3311  **/
3312 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3313 {
3314 	switch (adapter->hw.mac.type) {
3315 	case ixgbe_mac_82598EB:
3316 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3317 		break;
3318 	case ixgbe_mac_82599EB:
3319 	case ixgbe_mac_X540:
3320 	case ixgbe_mac_X550:
3321 	case ixgbe_mac_X550EM_x:
3322 	case ixgbe_mac_x550em_a:
3323 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3324 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3325 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3326 		break;
3327 	default:
3328 		break;
3329 	}
3330 	IXGBE_WRITE_FLUSH(&adapter->hw);
3331 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3332 		int vector;
3333 
3334 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3335 			synchronize_irq(adapter->msix_entries[vector].vector);
3336 
3337 		synchronize_irq(adapter->msix_entries[vector++].vector);
3338 	} else {
3339 		synchronize_irq(adapter->pdev->irq);
3340 	}
3341 }
3342 
3343 /**
3344  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3345  *
3346  **/
3347 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3348 {
3349 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3350 
3351 	ixgbe_write_eitr(q_vector);
3352 
3353 	ixgbe_set_ivar(adapter, 0, 0, 0);
3354 	ixgbe_set_ivar(adapter, 1, 0, 0);
3355 
3356 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3357 }
3358 
3359 /**
3360  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3361  * @adapter: board private structure
3362  * @ring: structure containing ring specific data
3363  *
3364  * Configure the Tx descriptor ring after a reset.
3365  **/
3366 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3367 			     struct ixgbe_ring *ring)
3368 {
3369 	struct ixgbe_hw *hw = &adapter->hw;
3370 	u64 tdba = ring->dma;
3371 	int wait_loop = 10;
3372 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3373 	u8 reg_idx = ring->reg_idx;
3374 
3375 	/* disable queue to avoid issues while updating state */
3376 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3377 	IXGBE_WRITE_FLUSH(hw);
3378 
3379 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3380 			(tdba & DMA_BIT_MASK(32)));
3381 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3382 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3383 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3384 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3385 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3386 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3387 
3388 	/*
3389 	 * set WTHRESH to encourage burst writeback, it should not be set
3390 	 * higher than 1 when:
3391 	 * - ITR is 0 as it could cause false TX hangs
3392 	 * - ITR is set to > 100k int/sec and BQL is enabled
3393 	 *
3394 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3395 	 * to or less than the number of on chip descriptors, which is
3396 	 * currently 40.
3397 	 */
3398 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3399 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3400 	else
3401 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3402 
3403 	/*
3404 	 * Setting PTHRESH to 32 both improves performance
3405 	 * and avoids a TX hang with DFP enabled
3406 	 */
3407 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3408 		   32;		/* PTHRESH = 32 */
3409 
3410 	/* reinitialize flowdirector state */
3411 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3412 		ring->atr_sample_rate = adapter->atr_sample_rate;
3413 		ring->atr_count = 0;
3414 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3415 	} else {
3416 		ring->atr_sample_rate = 0;
3417 	}
3418 
3419 	/* initialize XPS */
3420 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3421 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3422 
3423 		if (q_vector)
3424 			netif_set_xps_queue(ring->netdev,
3425 					    &q_vector->affinity_mask,
3426 					    ring->queue_index);
3427 	}
3428 
3429 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3430 
3431 	/* reinitialize tx_buffer_info */
3432 	memset(ring->tx_buffer_info, 0,
3433 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3434 
3435 	/* enable queue */
3436 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3437 
3438 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3439 	if (hw->mac.type == ixgbe_mac_82598EB &&
3440 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3441 		return;
3442 
3443 	/* poll to verify queue is enabled */
3444 	do {
3445 		usleep_range(1000, 2000);
3446 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3447 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3448 	if (!wait_loop)
3449 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3450 }
3451 
3452 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3453 {
3454 	struct ixgbe_hw *hw = &adapter->hw;
3455 	u32 rttdcs, mtqc;
3456 	u8 tcs = netdev_get_num_tc(adapter->netdev);
3457 
3458 	if (hw->mac.type == ixgbe_mac_82598EB)
3459 		return;
3460 
3461 	/* disable the arbiter while setting MTQC */
3462 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3463 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3464 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3465 
3466 	/* set transmit pool layout */
3467 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3468 		mtqc = IXGBE_MTQC_VT_ENA;
3469 		if (tcs > 4)
3470 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3471 		else if (tcs > 1)
3472 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3473 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3474 			 IXGBE_82599_VMDQ_4Q_MASK)
3475 			mtqc |= IXGBE_MTQC_32VF;
3476 		else
3477 			mtqc |= IXGBE_MTQC_64VF;
3478 	} else {
3479 		if (tcs > 4)
3480 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3481 		else if (tcs > 1)
3482 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3483 		else
3484 			mtqc = IXGBE_MTQC_64Q_1PB;
3485 	}
3486 
3487 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3488 
3489 	/* Enable Security TX Buffer IFG for multiple pb */
3490 	if (tcs) {
3491 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3492 		sectx |= IXGBE_SECTX_DCB;
3493 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3494 	}
3495 
3496 	/* re-enable the arbiter */
3497 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3498 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3499 }
3500 
3501 /**
3502  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3503  * @adapter: board private structure
3504  *
3505  * Configure the Tx unit of the MAC after a reset.
3506  **/
3507 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3508 {
3509 	struct ixgbe_hw *hw = &adapter->hw;
3510 	u32 dmatxctl;
3511 	u32 i;
3512 
3513 	ixgbe_setup_mtqc(adapter);
3514 
3515 	if (hw->mac.type != ixgbe_mac_82598EB) {
3516 		/* DMATXCTL.EN must be before Tx queues are enabled */
3517 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3518 		dmatxctl |= IXGBE_DMATXCTL_TE;
3519 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3520 	}
3521 
3522 	/* Setup the HW Tx Head and Tail descriptor pointers */
3523 	for (i = 0; i < adapter->num_tx_queues; i++)
3524 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3525 	for (i = 0; i < adapter->num_xdp_queues; i++)
3526 		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3527 }
3528 
3529 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3530 				 struct ixgbe_ring *ring)
3531 {
3532 	struct ixgbe_hw *hw = &adapter->hw;
3533 	u8 reg_idx = ring->reg_idx;
3534 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3535 
3536 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3537 
3538 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3539 }
3540 
3541 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3542 				  struct ixgbe_ring *ring)
3543 {
3544 	struct ixgbe_hw *hw = &adapter->hw;
3545 	u8 reg_idx = ring->reg_idx;
3546 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3547 
3548 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3549 
3550 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3551 }
3552 
3553 #ifdef CONFIG_IXGBE_DCB
3554 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3555 #else
3556 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3557 #endif
3558 {
3559 	int i;
3560 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3561 
3562 	if (adapter->ixgbe_ieee_pfc)
3563 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3564 
3565 	/*
3566 	 * We should set the drop enable bit if:
3567 	 *  SR-IOV is enabled
3568 	 *   or
3569 	 *  Number of Rx queues > 1 and flow control is disabled
3570 	 *
3571 	 *  This allows us to avoid head of line blocking for security
3572 	 *  and performance reasons.
3573 	 */
3574 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3575 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3576 		for (i = 0; i < adapter->num_rx_queues; i++)
3577 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3578 	} else {
3579 		for (i = 0; i < adapter->num_rx_queues; i++)
3580 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3581 	}
3582 }
3583 
3584 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3585 
3586 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3587 				   struct ixgbe_ring *rx_ring)
3588 {
3589 	struct ixgbe_hw *hw = &adapter->hw;
3590 	u32 srrctl;
3591 	u8 reg_idx = rx_ring->reg_idx;
3592 
3593 	if (hw->mac.type == ixgbe_mac_82598EB) {
3594 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3595 
3596 		/*
3597 		 * if VMDq is not active we must program one srrctl register
3598 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3599 		 */
3600 		reg_idx &= mask;
3601 	}
3602 
3603 	/* configure header buffer length, needed for RSC */
3604 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3605 
3606 	/* configure the packet buffer length */
3607 	if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state))
3608 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3609 	else
3610 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3611 
3612 	/* configure descriptor type */
3613 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3614 
3615 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3616 }
3617 
3618 /**
3619  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3620  * @adapter: device handle
3621  *
3622  *  - 82598/82599/X540:     128
3623  *  - X550(non-SRIOV mode): 512
3624  *  - X550(SRIOV mode):     64
3625  */
3626 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3627 {
3628 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3629 		return 128;
3630 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3631 		return 64;
3632 	else
3633 		return 512;
3634 }
3635 
3636 /**
3637  * ixgbe_store_key - Write the RSS key to HW
3638  * @adapter: device handle
3639  *
3640  * Write the RSS key stored in adapter.rss_key to HW.
3641  */
3642 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3643 {
3644 	struct ixgbe_hw *hw = &adapter->hw;
3645 	int i;
3646 
3647 	for (i = 0; i < 10; i++)
3648 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3649 }
3650 
3651 /**
3652  * ixgbe_init_rss_key - Initialize adapter RSS key
3653  * @adapter: device handle
3654  *
3655  * Allocates and initializes the RSS key if it is not allocated.
3656  **/
3657 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3658 {
3659 	u32 *rss_key;
3660 
3661 	if (!adapter->rss_key) {
3662 		rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3663 		if (unlikely(!rss_key))
3664 			return -ENOMEM;
3665 
3666 		netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3667 		adapter->rss_key = rss_key;
3668 	}
3669 
3670 	return 0;
3671 }
3672 
3673 /**
3674  * ixgbe_store_reta - Write the RETA table to HW
3675  * @adapter: device handle
3676  *
3677  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3678  */
3679 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3680 {
3681 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3682 	struct ixgbe_hw *hw = &adapter->hw;
3683 	u32 reta = 0;
3684 	u32 indices_multi;
3685 	u8 *indir_tbl = adapter->rss_indir_tbl;
3686 
3687 	/* Fill out the redirection table as follows:
3688 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3689 	 *    indices.
3690 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3691 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3692 	 */
3693 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3694 		indices_multi = 0x11;
3695 	else
3696 		indices_multi = 0x1;
3697 
3698 	/* Write redirection table to HW */
3699 	for (i = 0; i < reta_entries; i++) {
3700 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3701 		if ((i & 3) == 3) {
3702 			if (i < 128)
3703 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3704 			else
3705 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3706 						reta);
3707 			reta = 0;
3708 		}
3709 	}
3710 }
3711 
3712 /**
3713  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3714  * @adapter: device handle
3715  *
3716  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3717  */
3718 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3719 {
3720 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3721 	struct ixgbe_hw *hw = &adapter->hw;
3722 	u32 vfreta = 0;
3723 	unsigned int pf_pool = adapter->num_vfs;
3724 
3725 	/* Write redirection table to HW */
3726 	for (i = 0; i < reta_entries; i++) {
3727 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3728 		if ((i & 3) == 3) {
3729 			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3730 					vfreta);
3731 			vfreta = 0;
3732 		}
3733 	}
3734 }
3735 
3736 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3737 {
3738 	u32 i, j;
3739 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3740 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3741 
3742 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3743 	 * make full use of any rings they may have.  We will use the
3744 	 * PSRTYPE register to control how many rings we use within the PF.
3745 	 */
3746 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3747 		rss_i = 4;
3748 
3749 	/* Fill out hash function seeds */
3750 	ixgbe_store_key(adapter);
3751 
3752 	/* Fill out redirection table */
3753 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3754 
3755 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3756 		if (j == rss_i)
3757 			j = 0;
3758 
3759 		adapter->rss_indir_tbl[i] = j;
3760 	}
3761 
3762 	ixgbe_store_reta(adapter);
3763 }
3764 
3765 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3766 {
3767 	struct ixgbe_hw *hw = &adapter->hw;
3768 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3769 	unsigned int pf_pool = adapter->num_vfs;
3770 	int i, j;
3771 
3772 	/* Fill out hash function seeds */
3773 	for (i = 0; i < 10; i++)
3774 		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool),
3775 				*(adapter->rss_key + i));
3776 
3777 	/* Fill out the redirection table */
3778 	for (i = 0, j = 0; i < 64; i++, j++) {
3779 		if (j == rss_i)
3780 			j = 0;
3781 
3782 		adapter->rss_indir_tbl[i] = j;
3783 	}
3784 
3785 	ixgbe_store_vfreta(adapter);
3786 }
3787 
3788 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3789 {
3790 	struct ixgbe_hw *hw = &adapter->hw;
3791 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3792 	u32 rxcsum;
3793 
3794 	/* Disable indicating checksum in descriptor, enables RSS hash */
3795 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3796 	rxcsum |= IXGBE_RXCSUM_PCSD;
3797 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3798 
3799 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3800 		if (adapter->ring_feature[RING_F_RSS].mask)
3801 			mrqc = IXGBE_MRQC_RSSEN;
3802 	} else {
3803 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3804 
3805 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3806 			if (tcs > 4)
3807 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3808 			else if (tcs > 1)
3809 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3810 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3811 				 IXGBE_82599_VMDQ_4Q_MASK)
3812 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3813 			else
3814 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3815 
3816 			/* Enable L3/L4 for Tx Switched packets */
3817 			mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3818 		} else {
3819 			if (tcs > 4)
3820 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3821 			else if (tcs > 1)
3822 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3823 			else
3824 				mrqc = IXGBE_MRQC_RSSEN;
3825 		}
3826 	}
3827 
3828 	/* Perform hash on these packet types */
3829 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3830 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3831 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3832 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3833 
3834 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3835 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3836 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3837 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3838 
3839 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3840 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3841 		unsigned int pf_pool = adapter->num_vfs;
3842 
3843 		/* Enable VF RSS mode */
3844 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3845 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3846 
3847 		/* Setup RSS through the VF registers */
3848 		ixgbe_setup_vfreta(adapter);
3849 		vfmrqc = IXGBE_MRQC_RSSEN;
3850 		vfmrqc |= rss_field;
3851 		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3852 	} else {
3853 		ixgbe_setup_reta(adapter);
3854 		mrqc |= rss_field;
3855 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3856 	}
3857 }
3858 
3859 /**
3860  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3861  * @adapter:    address of board private structure
3862  * @index:      index of ring to set
3863  **/
3864 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3865 				   struct ixgbe_ring *ring)
3866 {
3867 	struct ixgbe_hw *hw = &adapter->hw;
3868 	u32 rscctrl;
3869 	u8 reg_idx = ring->reg_idx;
3870 
3871 	if (!ring_is_rsc_enabled(ring))
3872 		return;
3873 
3874 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3875 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3876 	/*
3877 	 * we must limit the number of descriptors so that the
3878 	 * total size of max desc * buf_len is not greater
3879 	 * than 65536
3880 	 */
3881 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3882 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3883 }
3884 
3885 #define IXGBE_MAX_RX_DESC_POLL 10
3886 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3887 				       struct ixgbe_ring *ring)
3888 {
3889 	struct ixgbe_hw *hw = &adapter->hw;
3890 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3891 	u32 rxdctl;
3892 	u8 reg_idx = ring->reg_idx;
3893 
3894 	if (ixgbe_removed(hw->hw_addr))
3895 		return;
3896 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3897 	if (hw->mac.type == ixgbe_mac_82598EB &&
3898 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3899 		return;
3900 
3901 	do {
3902 		usleep_range(1000, 2000);
3903 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3904 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3905 
3906 	if (!wait_loop) {
3907 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3908 		      "the polling period\n", reg_idx);
3909 	}
3910 }
3911 
3912 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3913 			    struct ixgbe_ring *ring)
3914 {
3915 	struct ixgbe_hw *hw = &adapter->hw;
3916 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3917 	u32 rxdctl;
3918 	u8 reg_idx = ring->reg_idx;
3919 
3920 	if (ixgbe_removed(hw->hw_addr))
3921 		return;
3922 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3923 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3924 
3925 	/* write value back with RXDCTL.ENABLE bit cleared */
3926 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3927 
3928 	if (hw->mac.type == ixgbe_mac_82598EB &&
3929 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3930 		return;
3931 
3932 	/* the hardware may take up to 100us to really disable the rx queue */
3933 	do {
3934 		udelay(10);
3935 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3936 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3937 
3938 	if (!wait_loop) {
3939 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3940 		      "the polling period\n", reg_idx);
3941 	}
3942 }
3943 
3944 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3945 			     struct ixgbe_ring *ring)
3946 {
3947 	struct ixgbe_hw *hw = &adapter->hw;
3948 	union ixgbe_adv_rx_desc *rx_desc;
3949 	u64 rdba = ring->dma;
3950 	u32 rxdctl;
3951 	u8 reg_idx = ring->reg_idx;
3952 
3953 	/* disable queue to avoid issues while updating state */
3954 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3955 	ixgbe_disable_rx_queue(adapter, ring);
3956 
3957 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3958 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3959 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3960 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3961 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
3962 	IXGBE_WRITE_FLUSH(hw);
3963 
3964 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3965 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3966 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3967 
3968 	ixgbe_configure_srrctl(adapter, ring);
3969 	ixgbe_configure_rscctl(adapter, ring);
3970 
3971 	if (hw->mac.type == ixgbe_mac_82598EB) {
3972 		/*
3973 		 * enable cache line friendly hardware writes:
3974 		 * PTHRESH=32 descriptors (half the internal cache),
3975 		 * this also removes ugly rx_no_buffer_count increment
3976 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3977 		 * WTHRESH=8 burst writeback up to two cache lines
3978 		 */
3979 		rxdctl &= ~0x3FFFFF;
3980 		rxdctl |=  0x080420;
3981 #if (PAGE_SIZE < 8192)
3982 	} else {
3983 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
3984 			    IXGBE_RXDCTL_RLPML_EN);
3985 
3986 		/* Limit the maximum frame size so we don't overrun the skb */
3987 		if (ring_uses_build_skb(ring) &&
3988 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
3989 			rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
3990 				  IXGBE_RXDCTL_RLPML_EN;
3991 #endif
3992 	}
3993 
3994 	/* initialize rx_buffer_info */
3995 	memset(ring->rx_buffer_info, 0,
3996 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
3997 
3998 	/* initialize Rx descriptor 0 */
3999 	rx_desc = IXGBE_RX_DESC(ring, 0);
4000 	rx_desc->wb.upper.length = 0;
4001 
4002 	/* enable receive descriptor ring */
4003 	rxdctl |= IXGBE_RXDCTL_ENABLE;
4004 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4005 
4006 	ixgbe_rx_desc_queue_enable(adapter, ring);
4007 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4008 }
4009 
4010 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4011 {
4012 	struct ixgbe_hw *hw = &adapter->hw;
4013 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4014 	u16 pool;
4015 
4016 	/* PSRTYPE must be initialized in non 82598 adapters */
4017 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4018 		      IXGBE_PSRTYPE_UDPHDR |
4019 		      IXGBE_PSRTYPE_IPV4HDR |
4020 		      IXGBE_PSRTYPE_L2HDR |
4021 		      IXGBE_PSRTYPE_IPV6HDR;
4022 
4023 	if (hw->mac.type == ixgbe_mac_82598EB)
4024 		return;
4025 
4026 	if (rss_i > 3)
4027 		psrtype |= 2u << 29;
4028 	else if (rss_i > 1)
4029 		psrtype |= 1u << 29;
4030 
4031 	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
4032 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4033 }
4034 
4035 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4036 {
4037 	struct ixgbe_hw *hw = &adapter->hw;
4038 	u32 reg_offset, vf_shift;
4039 	u32 gcr_ext, vmdctl;
4040 	int i;
4041 
4042 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4043 		return;
4044 
4045 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4046 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4047 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4048 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4049 	vmdctl |= IXGBE_VT_CTL_REPLEN;
4050 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4051 
4052 	vf_shift = VMDQ_P(0) % 32;
4053 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4054 
4055 	/* Enable only the PF's pool for Tx/Rx */
4056 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4057 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4058 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4059 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4060 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4061 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4062 
4063 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4064 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4065 
4066 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
4067 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4068 
4069 	/*
4070 	 * Set up VF register offsets for selected VT Mode,
4071 	 * i.e. 32 or 64 VFs for SR-IOV
4072 	 */
4073 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4074 	case IXGBE_82599_VMDQ_8Q_MASK:
4075 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4076 		break;
4077 	case IXGBE_82599_VMDQ_4Q_MASK:
4078 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4079 		break;
4080 	default:
4081 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4082 		break;
4083 	}
4084 
4085 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4086 
4087 	for (i = 0; i < adapter->num_vfs; i++) {
4088 		/* configure spoof checking */
4089 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4090 					  adapter->vfinfo[i].spoofchk_enabled);
4091 
4092 		/* Enable/Disable RSS query feature  */
4093 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4094 					  adapter->vfinfo[i].rss_query_enabled);
4095 	}
4096 }
4097 
4098 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4099 {
4100 	struct ixgbe_hw *hw = &adapter->hw;
4101 	struct net_device *netdev = adapter->netdev;
4102 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4103 	struct ixgbe_ring *rx_ring;
4104 	int i;
4105 	u32 mhadd, hlreg0;
4106 
4107 #ifdef IXGBE_FCOE
4108 	/* adjust max frame to be able to do baby jumbo for FCoE */
4109 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4110 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4111 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4112 
4113 #endif /* IXGBE_FCOE */
4114 
4115 	/* adjust max frame to be at least the size of a standard frame */
4116 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4117 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4118 
4119 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4120 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4121 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
4122 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4123 
4124 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4125 	}
4126 
4127 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4128 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4129 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4130 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4131 
4132 	/*
4133 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4134 	 * the Base and Length of the Rx Descriptor Ring
4135 	 */
4136 	for (i = 0; i < adapter->num_rx_queues; i++) {
4137 		rx_ring = adapter->rx_ring[i];
4138 
4139 		clear_ring_rsc_enabled(rx_ring);
4140 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4141 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4142 
4143 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4144 			set_ring_rsc_enabled(rx_ring);
4145 
4146 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4147 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4148 
4149 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4150 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4151 			continue;
4152 
4153 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4154 
4155 #if (PAGE_SIZE < 8192)
4156 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4157 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4158 
4159 		if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4160 		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4161 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4162 #endif
4163 	}
4164 }
4165 
4166 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4167 {
4168 	struct ixgbe_hw *hw = &adapter->hw;
4169 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4170 
4171 	switch (hw->mac.type) {
4172 	case ixgbe_mac_82598EB:
4173 		/*
4174 		 * For VMDq support of different descriptor types or
4175 		 * buffer sizes through the use of multiple SRRCTL
4176 		 * registers, RDRXCTL.MVMEN must be set to 1
4177 		 *
4178 		 * also, the manual doesn't mention it clearly but DCA hints
4179 		 * will only use queue 0's tags unless this bit is set.  Side
4180 		 * effects of setting this bit are only that SRRCTL must be
4181 		 * fully programmed [0..15]
4182 		 */
4183 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4184 		break;
4185 	case ixgbe_mac_X550:
4186 	case ixgbe_mac_X550EM_x:
4187 	case ixgbe_mac_x550em_a:
4188 		if (adapter->num_vfs)
4189 			rdrxctl |= IXGBE_RDRXCTL_PSP;
4190 		/* fall through */
4191 	case ixgbe_mac_82599EB:
4192 	case ixgbe_mac_X540:
4193 		/* Disable RSC for ACK packets */
4194 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4195 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4196 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4197 		/* hardware requires some bits to be set by default */
4198 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4199 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4200 		break;
4201 	default:
4202 		/* We should do nothing since we don't know this hardware */
4203 		return;
4204 	}
4205 
4206 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4207 }
4208 
4209 /**
4210  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4211  * @adapter: board private structure
4212  *
4213  * Configure the Rx unit of the MAC after a reset.
4214  **/
4215 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4216 {
4217 	struct ixgbe_hw *hw = &adapter->hw;
4218 	int i;
4219 	u32 rxctrl, rfctl;
4220 
4221 	/* disable receives while setting up the descriptors */
4222 	hw->mac.ops.disable_rx(hw);
4223 
4224 	ixgbe_setup_psrtype(adapter);
4225 	ixgbe_setup_rdrxctl(adapter);
4226 
4227 	/* RSC Setup */
4228 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4229 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4230 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4231 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4232 
4233 	/* disable NFS filtering */
4234 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4235 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4236 
4237 	/* Program registers for the distribution of queues */
4238 	ixgbe_setup_mrqc(adapter);
4239 
4240 	/* set_rx_buffer_len must be called before ring initialization */
4241 	ixgbe_set_rx_buffer_len(adapter);
4242 
4243 	/*
4244 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4245 	 * the Base and Length of the Rx Descriptor Ring
4246 	 */
4247 	for (i = 0; i < adapter->num_rx_queues; i++)
4248 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4249 
4250 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4251 	/* disable drop enable for 82598 parts */
4252 	if (hw->mac.type == ixgbe_mac_82598EB)
4253 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4254 
4255 	/* enable all receives */
4256 	rxctrl |= IXGBE_RXCTRL_RXEN;
4257 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4258 }
4259 
4260 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4261 				 __be16 proto, u16 vid)
4262 {
4263 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4264 	struct ixgbe_hw *hw = &adapter->hw;
4265 
4266 	/* add VID to filter table */
4267 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4268 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4269 
4270 	set_bit(vid, adapter->active_vlans);
4271 
4272 	return 0;
4273 }
4274 
4275 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4276 {
4277 	u32 vlvf;
4278 	int idx;
4279 
4280 	/* short cut the special case */
4281 	if (vlan == 0)
4282 		return 0;
4283 
4284 	/* Search for the vlan id in the VLVF entries */
4285 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4286 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4287 		if ((vlvf & VLAN_VID_MASK) == vlan)
4288 			break;
4289 	}
4290 
4291 	return idx;
4292 }
4293 
4294 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4295 {
4296 	struct ixgbe_hw *hw = &adapter->hw;
4297 	u32 bits, word;
4298 	int idx;
4299 
4300 	idx = ixgbe_find_vlvf_entry(hw, vid);
4301 	if (!idx)
4302 		return;
4303 
4304 	/* See if any other pools are set for this VLAN filter
4305 	 * entry other than the PF.
4306 	 */
4307 	word = idx * 2 + (VMDQ_P(0) / 32);
4308 	bits = ~BIT(VMDQ_P(0) % 32);
4309 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4310 
4311 	/* Disable the filter so this falls into the default pool. */
4312 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4313 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4314 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4315 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4316 	}
4317 }
4318 
4319 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4320 				  __be16 proto, u16 vid)
4321 {
4322 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4323 	struct ixgbe_hw *hw = &adapter->hw;
4324 
4325 	/* remove VID from filter table */
4326 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4327 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4328 
4329 	clear_bit(vid, adapter->active_vlans);
4330 
4331 	return 0;
4332 }
4333 
4334 /**
4335  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4336  * @adapter: driver data
4337  */
4338 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4339 {
4340 	struct ixgbe_hw *hw = &adapter->hw;
4341 	u32 vlnctrl;
4342 	int i, j;
4343 
4344 	switch (hw->mac.type) {
4345 	case ixgbe_mac_82598EB:
4346 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4347 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4348 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4349 		break;
4350 	case ixgbe_mac_82599EB:
4351 	case ixgbe_mac_X540:
4352 	case ixgbe_mac_X550:
4353 	case ixgbe_mac_X550EM_x:
4354 	case ixgbe_mac_x550em_a:
4355 		for (i = 0; i < adapter->num_rx_queues; i++) {
4356 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4357 
4358 			if (ring->l2_accel_priv)
4359 				continue;
4360 			j = ring->reg_idx;
4361 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4362 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4363 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4364 		}
4365 		break;
4366 	default:
4367 		break;
4368 	}
4369 }
4370 
4371 /**
4372  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4373  * @adapter: driver data
4374  */
4375 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4376 {
4377 	struct ixgbe_hw *hw = &adapter->hw;
4378 	u32 vlnctrl;
4379 	int i, j;
4380 
4381 	switch (hw->mac.type) {
4382 	case ixgbe_mac_82598EB:
4383 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4384 		vlnctrl |= IXGBE_VLNCTRL_VME;
4385 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4386 		break;
4387 	case ixgbe_mac_82599EB:
4388 	case ixgbe_mac_X540:
4389 	case ixgbe_mac_X550:
4390 	case ixgbe_mac_X550EM_x:
4391 	case ixgbe_mac_x550em_a:
4392 		for (i = 0; i < adapter->num_rx_queues; i++) {
4393 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4394 
4395 			if (ring->l2_accel_priv)
4396 				continue;
4397 			j = ring->reg_idx;
4398 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4399 			vlnctrl |= IXGBE_RXDCTL_VME;
4400 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4401 		}
4402 		break;
4403 	default:
4404 		break;
4405 	}
4406 }
4407 
4408 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4409 {
4410 	struct ixgbe_hw *hw = &adapter->hw;
4411 	u32 vlnctrl, i;
4412 
4413 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4414 
4415 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4416 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4417 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4418 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4419 	} else {
4420 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4421 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4422 		return;
4423 	}
4424 
4425 	/* Nothing to do for 82598 */
4426 	if (hw->mac.type == ixgbe_mac_82598EB)
4427 		return;
4428 
4429 	/* We are already in VLAN promisc, nothing to do */
4430 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4431 		return;
4432 
4433 	/* Set flag so we don't redo unnecessary work */
4434 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4435 
4436 	/* Add PF to all active pools */
4437 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4438 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4439 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4440 
4441 		vlvfb |= BIT(VMDQ_P(0) % 32);
4442 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4443 	}
4444 
4445 	/* Set all bits in the VLAN filter table array */
4446 	for (i = hw->mac.vft_size; i--;)
4447 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4448 }
4449 
4450 #define VFTA_BLOCK_SIZE 8
4451 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4452 {
4453 	struct ixgbe_hw *hw = &adapter->hw;
4454 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4455 	u32 vid_start = vfta_offset * 32;
4456 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4457 	u32 i, vid, word, bits;
4458 
4459 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4460 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4461 
4462 		/* pull VLAN ID from VLVF */
4463 		vid = vlvf & VLAN_VID_MASK;
4464 
4465 		/* only concern outselves with a certain range */
4466 		if (vid < vid_start || vid >= vid_end)
4467 			continue;
4468 
4469 		if (vlvf) {
4470 			/* record VLAN ID in VFTA */
4471 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4472 
4473 			/* if PF is part of this then continue */
4474 			if (test_bit(vid, adapter->active_vlans))
4475 				continue;
4476 		}
4477 
4478 		/* remove PF from the pool */
4479 		word = i * 2 + VMDQ_P(0) / 32;
4480 		bits = ~BIT(VMDQ_P(0) % 32);
4481 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4482 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4483 	}
4484 
4485 	/* extract values from active_vlans and write back to VFTA */
4486 	for (i = VFTA_BLOCK_SIZE; i--;) {
4487 		vid = (vfta_offset + i) * 32;
4488 		word = vid / BITS_PER_LONG;
4489 		bits = vid % BITS_PER_LONG;
4490 
4491 		vfta[i] |= adapter->active_vlans[word] >> bits;
4492 
4493 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4494 	}
4495 }
4496 
4497 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4498 {
4499 	struct ixgbe_hw *hw = &adapter->hw;
4500 	u32 vlnctrl, i;
4501 
4502 	/* Set VLAN filtering to enabled */
4503 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4504 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4505 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4506 
4507 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4508 	    hw->mac.type == ixgbe_mac_82598EB)
4509 		return;
4510 
4511 	/* We are not in VLAN promisc, nothing to do */
4512 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4513 		return;
4514 
4515 	/* Set flag so we don't redo unnecessary work */
4516 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4517 
4518 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4519 		ixgbe_scrub_vfta(adapter, i);
4520 }
4521 
4522 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4523 {
4524 	u16 vid = 1;
4525 
4526 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4527 
4528 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4529 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4530 }
4531 
4532 /**
4533  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4534  * @netdev: network interface device structure
4535  *
4536  * Writes multicast address list to the MTA hash table.
4537  * Returns: -ENOMEM on failure
4538  *                0 on no addresses written
4539  *                X on writing X addresses to MTA
4540  **/
4541 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4542 {
4543 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4544 	struct ixgbe_hw *hw = &adapter->hw;
4545 
4546 	if (!netif_running(netdev))
4547 		return 0;
4548 
4549 	if (hw->mac.ops.update_mc_addr_list)
4550 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4551 	else
4552 		return -ENOMEM;
4553 
4554 #ifdef CONFIG_PCI_IOV
4555 	ixgbe_restore_vf_multicasts(adapter);
4556 #endif
4557 
4558 	return netdev_mc_count(netdev);
4559 }
4560 
4561 #ifdef CONFIG_PCI_IOV
4562 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4563 {
4564 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4565 	struct ixgbe_hw *hw = &adapter->hw;
4566 	int i;
4567 
4568 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4569 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4570 
4571 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4572 			hw->mac.ops.set_rar(hw, i,
4573 					    mac_table->addr,
4574 					    mac_table->pool,
4575 					    IXGBE_RAH_AV);
4576 		else
4577 			hw->mac.ops.clear_rar(hw, i);
4578 	}
4579 }
4580 
4581 #endif
4582 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4583 {
4584 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4585 	struct ixgbe_hw *hw = &adapter->hw;
4586 	int i;
4587 
4588 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4589 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4590 			continue;
4591 
4592 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4593 
4594 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4595 			hw->mac.ops.set_rar(hw, i,
4596 					    mac_table->addr,
4597 					    mac_table->pool,
4598 					    IXGBE_RAH_AV);
4599 		else
4600 			hw->mac.ops.clear_rar(hw, i);
4601 	}
4602 }
4603 
4604 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4605 {
4606 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4607 	struct ixgbe_hw *hw = &adapter->hw;
4608 	int i;
4609 
4610 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4611 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4612 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4613 	}
4614 
4615 	ixgbe_sync_mac_table(adapter);
4616 }
4617 
4618 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4619 {
4620 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4621 	struct ixgbe_hw *hw = &adapter->hw;
4622 	int i, count = 0;
4623 
4624 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4625 		/* do not count default RAR as available */
4626 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4627 			continue;
4628 
4629 		/* only count unused and addresses that belong to us */
4630 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4631 			if (mac_table->pool != pool)
4632 				continue;
4633 		}
4634 
4635 		count++;
4636 	}
4637 
4638 	return count;
4639 }
4640 
4641 /* this function destroys the first RAR entry */
4642 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4643 {
4644 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4645 	struct ixgbe_hw *hw = &adapter->hw;
4646 
4647 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4648 	mac_table->pool = VMDQ_P(0);
4649 
4650 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4651 
4652 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4653 			    IXGBE_RAH_AV);
4654 }
4655 
4656 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4657 			 const u8 *addr, u16 pool)
4658 {
4659 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4660 	struct ixgbe_hw *hw = &adapter->hw;
4661 	int i;
4662 
4663 	if (is_zero_ether_addr(addr))
4664 		return -EINVAL;
4665 
4666 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4667 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4668 			continue;
4669 
4670 		ether_addr_copy(mac_table->addr, addr);
4671 		mac_table->pool = pool;
4672 
4673 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4674 				    IXGBE_MAC_STATE_IN_USE;
4675 
4676 		ixgbe_sync_mac_table(adapter);
4677 
4678 		return i;
4679 	}
4680 
4681 	return -ENOMEM;
4682 }
4683 
4684 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4685 			 const u8 *addr, u16 pool)
4686 {
4687 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4688 	struct ixgbe_hw *hw = &adapter->hw;
4689 	int i;
4690 
4691 	if (is_zero_ether_addr(addr))
4692 		return -EINVAL;
4693 
4694 	/* search table for addr, if found clear IN_USE flag and sync */
4695 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4696 		/* we can only delete an entry if it is in use */
4697 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4698 			continue;
4699 		/* we only care about entries that belong to the given pool */
4700 		if (mac_table->pool != pool)
4701 			continue;
4702 		/* we only care about a specific MAC address */
4703 		if (!ether_addr_equal(addr, mac_table->addr))
4704 			continue;
4705 
4706 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4707 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4708 
4709 		ixgbe_sync_mac_table(adapter);
4710 
4711 		return 0;
4712 	}
4713 
4714 	return -ENOMEM;
4715 }
4716 /**
4717  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4718  * @netdev: network interface device structure
4719  *
4720  * Writes unicast address list to the RAR table.
4721  * Returns: -ENOMEM on failure/insufficient address space
4722  *                0 on no addresses written
4723  *                X on writing X addresses to the RAR table
4724  **/
4725 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4726 {
4727 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4728 	int count = 0;
4729 
4730 	/* return ENOMEM indicating insufficient memory for addresses */
4731 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter, vfn))
4732 		return -ENOMEM;
4733 
4734 	if (!netdev_uc_empty(netdev)) {
4735 		struct netdev_hw_addr *ha;
4736 		netdev_for_each_uc_addr(ha, netdev) {
4737 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4738 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4739 			count++;
4740 		}
4741 	}
4742 	return count;
4743 }
4744 
4745 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4746 {
4747 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4748 	int ret;
4749 
4750 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4751 
4752 	return min_t(int, ret, 0);
4753 }
4754 
4755 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4756 {
4757 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4758 
4759 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4760 
4761 	return 0;
4762 }
4763 
4764 /**
4765  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4766  * @netdev: network interface device structure
4767  *
4768  * The set_rx_method entry point is called whenever the unicast/multicast
4769  * address list or the network interface flags are updated.  This routine is
4770  * responsible for configuring the hardware for proper unicast, multicast and
4771  * promiscuous mode.
4772  **/
4773 void ixgbe_set_rx_mode(struct net_device *netdev)
4774 {
4775 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4776 	struct ixgbe_hw *hw = &adapter->hw;
4777 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4778 	netdev_features_t features = netdev->features;
4779 	int count;
4780 
4781 	/* Check for Promiscuous and All Multicast modes */
4782 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4783 
4784 	/* set all bits that we expect to always be set */
4785 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4786 	fctrl |= IXGBE_FCTRL_BAM;
4787 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4788 	fctrl |= IXGBE_FCTRL_PMCF;
4789 
4790 	/* clear the bits we are changing the status of */
4791 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4792 	if (netdev->flags & IFF_PROMISC) {
4793 		hw->addr_ctrl.user_set_promisc = true;
4794 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4795 		vmolr |= IXGBE_VMOLR_MPE;
4796 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4797 	} else {
4798 		if (netdev->flags & IFF_ALLMULTI) {
4799 			fctrl |= IXGBE_FCTRL_MPE;
4800 			vmolr |= IXGBE_VMOLR_MPE;
4801 		}
4802 		hw->addr_ctrl.user_set_promisc = false;
4803 	}
4804 
4805 	/*
4806 	 * Write addresses to available RAR registers, if there is not
4807 	 * sufficient space to store all the addresses then enable
4808 	 * unicast promiscuous mode
4809 	 */
4810 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4811 		fctrl |= IXGBE_FCTRL_UPE;
4812 		vmolr |= IXGBE_VMOLR_ROPE;
4813 	}
4814 
4815 	/* Write addresses to the MTA, if the attempt fails
4816 	 * then we should just turn on promiscuous mode so
4817 	 * that we can at least receive multicast traffic
4818 	 */
4819 	count = ixgbe_write_mc_addr_list(netdev);
4820 	if (count < 0) {
4821 		fctrl |= IXGBE_FCTRL_MPE;
4822 		vmolr |= IXGBE_VMOLR_MPE;
4823 	} else if (count) {
4824 		vmolr |= IXGBE_VMOLR_ROMPE;
4825 	}
4826 
4827 	if (hw->mac.type != ixgbe_mac_82598EB) {
4828 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4829 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4830 			   IXGBE_VMOLR_ROPE);
4831 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4832 	}
4833 
4834 	/* This is useful for sniffing bad packets. */
4835 	if (features & NETIF_F_RXALL) {
4836 		/* UPE and MPE will be handled by normal PROMISC logic
4837 		 * in e1000e_set_rx_mode */
4838 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4839 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4840 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4841 
4842 		fctrl &= ~(IXGBE_FCTRL_DPF);
4843 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4844 	}
4845 
4846 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4847 
4848 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4849 		ixgbe_vlan_strip_enable(adapter);
4850 	else
4851 		ixgbe_vlan_strip_disable(adapter);
4852 
4853 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4854 		ixgbe_vlan_promisc_disable(adapter);
4855 	else
4856 		ixgbe_vlan_promisc_enable(adapter);
4857 }
4858 
4859 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4860 {
4861 	int q_idx;
4862 
4863 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4864 		napi_enable(&adapter->q_vector[q_idx]->napi);
4865 }
4866 
4867 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4868 {
4869 	int q_idx;
4870 
4871 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4872 		napi_disable(&adapter->q_vector[q_idx]->napi);
4873 }
4874 
4875 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
4876 {
4877 	struct ixgbe_hw *hw = &adapter->hw;
4878 	u32 vxlanctrl;
4879 
4880 	if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
4881 				IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
4882 		return;
4883 
4884 	vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) && ~mask;
4885 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
4886 
4887 	if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
4888 		adapter->vxlan_port = 0;
4889 
4890 	if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
4891 		adapter->geneve_port = 0;
4892 }
4893 
4894 #ifdef CONFIG_IXGBE_DCB
4895 /**
4896  * ixgbe_configure_dcb - Configure DCB hardware
4897  * @adapter: ixgbe adapter struct
4898  *
4899  * This is called by the driver on open to configure the DCB hardware.
4900  * This is also called by the gennetlink interface when reconfiguring
4901  * the DCB state.
4902  */
4903 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4904 {
4905 	struct ixgbe_hw *hw = &adapter->hw;
4906 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4907 
4908 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4909 		if (hw->mac.type == ixgbe_mac_82598EB)
4910 			netif_set_gso_max_size(adapter->netdev, 65536);
4911 		return;
4912 	}
4913 
4914 	if (hw->mac.type == ixgbe_mac_82598EB)
4915 		netif_set_gso_max_size(adapter->netdev, 32768);
4916 
4917 #ifdef IXGBE_FCOE
4918 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4919 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4920 #endif
4921 
4922 	/* reconfigure the hardware */
4923 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4924 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4925 						DCB_TX_CONFIG);
4926 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4927 						DCB_RX_CONFIG);
4928 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4929 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4930 		ixgbe_dcb_hw_ets(&adapter->hw,
4931 				 adapter->ixgbe_ieee_ets,
4932 				 max_frame);
4933 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4934 					adapter->ixgbe_ieee_pfc->pfc_en,
4935 					adapter->ixgbe_ieee_ets->prio_tc);
4936 	}
4937 
4938 	/* Enable RSS Hash per TC */
4939 	if (hw->mac.type != ixgbe_mac_82598EB) {
4940 		u32 msb = 0;
4941 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4942 
4943 		while (rss_i) {
4944 			msb++;
4945 			rss_i >>= 1;
4946 		}
4947 
4948 		/* write msb to all 8 TCs in one write */
4949 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4950 	}
4951 }
4952 #endif
4953 
4954 /* Additional bittime to account for IXGBE framing */
4955 #define IXGBE_ETH_FRAMING 20
4956 
4957 /**
4958  * ixgbe_hpbthresh - calculate high water mark for flow control
4959  *
4960  * @adapter: board private structure to calculate for
4961  * @pb: packet buffer to calculate
4962  */
4963 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4964 {
4965 	struct ixgbe_hw *hw = &adapter->hw;
4966 	struct net_device *dev = adapter->netdev;
4967 	int link, tc, kb, marker;
4968 	u32 dv_id, rx_pba;
4969 
4970 	/* Calculate max LAN frame size */
4971 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4972 
4973 #ifdef IXGBE_FCOE
4974 	/* FCoE traffic class uses FCOE jumbo frames */
4975 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4976 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4977 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4978 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4979 #endif
4980 
4981 	/* Calculate delay value for device */
4982 	switch (hw->mac.type) {
4983 	case ixgbe_mac_X540:
4984 	case ixgbe_mac_X550:
4985 	case ixgbe_mac_X550EM_x:
4986 	case ixgbe_mac_x550em_a:
4987 		dv_id = IXGBE_DV_X540(link, tc);
4988 		break;
4989 	default:
4990 		dv_id = IXGBE_DV(link, tc);
4991 		break;
4992 	}
4993 
4994 	/* Loopback switch introduces additional latency */
4995 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4996 		dv_id += IXGBE_B2BT(tc);
4997 
4998 	/* Delay value is calculated in bit times convert to KB */
4999 	kb = IXGBE_BT2KB(dv_id);
5000 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5001 
5002 	marker = rx_pba - kb;
5003 
5004 	/* It is possible that the packet buffer is not large enough
5005 	 * to provide required headroom. In this case throw an error
5006 	 * to user and a do the best we can.
5007 	 */
5008 	if (marker < 0) {
5009 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
5010 			    "headroom to support flow control."
5011 			    "Decrease MTU or number of traffic classes\n", pb);
5012 		marker = tc + 1;
5013 	}
5014 
5015 	return marker;
5016 }
5017 
5018 /**
5019  * ixgbe_lpbthresh - calculate low water mark for for flow control
5020  *
5021  * @adapter: board private structure to calculate for
5022  * @pb: packet buffer to calculate
5023  */
5024 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5025 {
5026 	struct ixgbe_hw *hw = &adapter->hw;
5027 	struct net_device *dev = adapter->netdev;
5028 	int tc;
5029 	u32 dv_id;
5030 
5031 	/* Calculate max LAN frame size */
5032 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5033 
5034 #ifdef IXGBE_FCOE
5035 	/* FCoE traffic class uses FCOE jumbo frames */
5036 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5037 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5038 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5039 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5040 #endif
5041 
5042 	/* Calculate delay value for device */
5043 	switch (hw->mac.type) {
5044 	case ixgbe_mac_X540:
5045 	case ixgbe_mac_X550:
5046 	case ixgbe_mac_X550EM_x:
5047 	case ixgbe_mac_x550em_a:
5048 		dv_id = IXGBE_LOW_DV_X540(tc);
5049 		break;
5050 	default:
5051 		dv_id = IXGBE_LOW_DV(tc);
5052 		break;
5053 	}
5054 
5055 	/* Delay value is calculated in bit times convert to KB */
5056 	return IXGBE_BT2KB(dv_id);
5057 }
5058 
5059 /*
5060  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5061  */
5062 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5063 {
5064 	struct ixgbe_hw *hw = &adapter->hw;
5065 	int num_tc = netdev_get_num_tc(adapter->netdev);
5066 	int i;
5067 
5068 	if (!num_tc)
5069 		num_tc = 1;
5070 
5071 	for (i = 0; i < num_tc; i++) {
5072 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5073 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5074 
5075 		/* Low water marks must not be larger than high water marks */
5076 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
5077 			hw->fc.low_water[i] = 0;
5078 	}
5079 
5080 	for (; i < MAX_TRAFFIC_CLASS; i++)
5081 		hw->fc.high_water[i] = 0;
5082 }
5083 
5084 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5085 {
5086 	struct ixgbe_hw *hw = &adapter->hw;
5087 	int hdrm;
5088 	u8 tc = netdev_get_num_tc(adapter->netdev);
5089 
5090 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5091 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5092 		hdrm = 32 << adapter->fdir_pballoc;
5093 	else
5094 		hdrm = 0;
5095 
5096 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5097 	ixgbe_pbthresh_setup(adapter);
5098 }
5099 
5100 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5101 {
5102 	struct ixgbe_hw *hw = &adapter->hw;
5103 	struct hlist_node *node2;
5104 	struct ixgbe_fdir_filter *filter;
5105 
5106 	spin_lock(&adapter->fdir_perfect_lock);
5107 
5108 	if (!hlist_empty(&adapter->fdir_filter_list))
5109 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5110 
5111 	hlist_for_each_entry_safe(filter, node2,
5112 				  &adapter->fdir_filter_list, fdir_node) {
5113 		ixgbe_fdir_write_perfect_filter_82599(hw,
5114 				&filter->filter,
5115 				filter->sw_idx,
5116 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
5117 				IXGBE_FDIR_DROP_QUEUE :
5118 				adapter->rx_ring[filter->action]->reg_idx);
5119 	}
5120 
5121 	spin_unlock(&adapter->fdir_perfect_lock);
5122 }
5123 
5124 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
5125 				      struct ixgbe_adapter *adapter)
5126 {
5127 	struct ixgbe_hw *hw = &adapter->hw;
5128 	u32 vmolr;
5129 
5130 	/* No unicast promiscuous support for VMDQ devices. */
5131 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
5132 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
5133 
5134 	/* clear the affected bit */
5135 	vmolr &= ~IXGBE_VMOLR_MPE;
5136 
5137 	if (dev->flags & IFF_ALLMULTI) {
5138 		vmolr |= IXGBE_VMOLR_MPE;
5139 	} else {
5140 		vmolr |= IXGBE_VMOLR_ROMPE;
5141 		hw->mac.ops.update_mc_addr_list(hw, dev);
5142 	}
5143 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
5144 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
5145 }
5146 
5147 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
5148 {
5149 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
5150 	int rss_i = adapter->num_rx_queues_per_pool;
5151 	struct ixgbe_hw *hw = &adapter->hw;
5152 	u16 pool = vadapter->pool;
5153 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
5154 		      IXGBE_PSRTYPE_UDPHDR |
5155 		      IXGBE_PSRTYPE_IPV4HDR |
5156 		      IXGBE_PSRTYPE_L2HDR |
5157 		      IXGBE_PSRTYPE_IPV6HDR;
5158 
5159 	if (hw->mac.type == ixgbe_mac_82598EB)
5160 		return;
5161 
5162 	if (rss_i > 3)
5163 		psrtype |= 2u << 29;
5164 	else if (rss_i > 1)
5165 		psrtype |= 1u << 29;
5166 
5167 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
5168 }
5169 
5170 /**
5171  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5172  * @rx_ring: ring to free buffers from
5173  **/
5174 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5175 {
5176 	u16 i = rx_ring->next_to_clean;
5177 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5178 
5179 	/* Free all the Rx ring sk_buffs */
5180 	while (i != rx_ring->next_to_alloc) {
5181 		if (rx_buffer->skb) {
5182 			struct sk_buff *skb = rx_buffer->skb;
5183 			if (IXGBE_CB(skb)->page_released)
5184 				dma_unmap_page_attrs(rx_ring->dev,
5185 						     IXGBE_CB(skb)->dma,
5186 						     ixgbe_rx_pg_size(rx_ring),
5187 						     DMA_FROM_DEVICE,
5188 						     IXGBE_RX_DMA_ATTR);
5189 			dev_kfree_skb(skb);
5190 		}
5191 
5192 		/* Invalidate cache lines that may have been written to by
5193 		 * device so that we avoid corrupting memory.
5194 		 */
5195 		dma_sync_single_range_for_cpu(rx_ring->dev,
5196 					      rx_buffer->dma,
5197 					      rx_buffer->page_offset,
5198 					      ixgbe_rx_bufsz(rx_ring),
5199 					      DMA_FROM_DEVICE);
5200 
5201 		/* free resources associated with mapping */
5202 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5203 				     ixgbe_rx_pg_size(rx_ring),
5204 				     DMA_FROM_DEVICE,
5205 				     IXGBE_RX_DMA_ATTR);
5206 		__page_frag_cache_drain(rx_buffer->page,
5207 					rx_buffer->pagecnt_bias);
5208 
5209 		i++;
5210 		rx_buffer++;
5211 		if (i == rx_ring->count) {
5212 			i = 0;
5213 			rx_buffer = rx_ring->rx_buffer_info;
5214 		}
5215 	}
5216 
5217 	rx_ring->next_to_alloc = 0;
5218 	rx_ring->next_to_clean = 0;
5219 	rx_ring->next_to_use = 0;
5220 }
5221 
5222 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
5223 				   struct ixgbe_ring *rx_ring)
5224 {
5225 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
5226 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
5227 
5228 	/* shutdown specific queue receive and wait for dma to settle */
5229 	ixgbe_disable_rx_queue(adapter, rx_ring);
5230 	usleep_range(10000, 20000);
5231 	ixgbe_irq_disable_queues(adapter, BIT_ULL(index));
5232 	ixgbe_clean_rx_ring(rx_ring);
5233 	rx_ring->l2_accel_priv = NULL;
5234 }
5235 
5236 static int ixgbe_fwd_ring_down(struct net_device *vdev,
5237 			       struct ixgbe_fwd_adapter *accel)
5238 {
5239 	struct ixgbe_adapter *adapter = accel->real_adapter;
5240 	unsigned int rxbase = accel->rx_base_queue;
5241 	unsigned int txbase = accel->tx_base_queue;
5242 	int i;
5243 
5244 	netif_tx_stop_all_queues(vdev);
5245 
5246 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
5247 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
5248 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
5249 	}
5250 
5251 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
5252 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
5253 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
5254 	}
5255 
5256 
5257 	return 0;
5258 }
5259 
5260 static int ixgbe_fwd_ring_up(struct net_device *vdev,
5261 			     struct ixgbe_fwd_adapter *accel)
5262 {
5263 	struct ixgbe_adapter *adapter = accel->real_adapter;
5264 	unsigned int rxbase, txbase, queues;
5265 	int i, baseq, err = 0;
5266 
5267 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
5268 		return 0;
5269 
5270 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5271 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
5272 		   accel->pool, adapter->num_rx_pools,
5273 		   baseq, baseq + adapter->num_rx_queues_per_pool,
5274 		   adapter->fwd_bitmask);
5275 
5276 	accel->netdev = vdev;
5277 	accel->rx_base_queue = rxbase = baseq;
5278 	accel->tx_base_queue = txbase = baseq;
5279 
5280 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5281 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
5282 
5283 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
5284 		adapter->rx_ring[rxbase + i]->netdev = vdev;
5285 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
5286 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
5287 	}
5288 
5289 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
5290 		adapter->tx_ring[txbase + i]->netdev = vdev;
5291 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
5292 	}
5293 
5294 	queues = min_t(unsigned int,
5295 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
5296 	err = netif_set_real_num_tx_queues(vdev, queues);
5297 	if (err)
5298 		goto fwd_queue_err;
5299 
5300 	err = netif_set_real_num_rx_queues(vdev, queues);
5301 	if (err)
5302 		goto fwd_queue_err;
5303 
5304 	if (is_valid_ether_addr(vdev->dev_addr))
5305 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
5306 
5307 	ixgbe_fwd_psrtype(accel);
5308 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
5309 	return err;
5310 fwd_queue_err:
5311 	ixgbe_fwd_ring_down(vdev, accel);
5312 	return err;
5313 }
5314 
5315 static int ixgbe_upper_dev_walk(struct net_device *upper, void *data)
5316 {
5317 	if (netif_is_macvlan(upper)) {
5318 		struct macvlan_dev *dfwd = netdev_priv(upper);
5319 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
5320 
5321 		if (dfwd->fwd_priv)
5322 			ixgbe_fwd_ring_up(upper, vadapter);
5323 	}
5324 
5325 	return 0;
5326 }
5327 
5328 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5329 {
5330 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5331 				      ixgbe_upper_dev_walk, NULL);
5332 }
5333 
5334 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5335 {
5336 	struct ixgbe_hw *hw = &adapter->hw;
5337 
5338 	ixgbe_configure_pb(adapter);
5339 #ifdef CONFIG_IXGBE_DCB
5340 	ixgbe_configure_dcb(adapter);
5341 #endif
5342 	/*
5343 	 * We must restore virtualization before VLANs or else
5344 	 * the VLVF registers will not be populated
5345 	 */
5346 	ixgbe_configure_virtualization(adapter);
5347 
5348 	ixgbe_set_rx_mode(adapter->netdev);
5349 	ixgbe_restore_vlan(adapter);
5350 
5351 	switch (hw->mac.type) {
5352 	case ixgbe_mac_82599EB:
5353 	case ixgbe_mac_X540:
5354 		hw->mac.ops.disable_rx_buff(hw);
5355 		break;
5356 	default:
5357 		break;
5358 	}
5359 
5360 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5361 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5362 						adapter->fdir_pballoc);
5363 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5364 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5365 					      adapter->fdir_pballoc);
5366 		ixgbe_fdir_filter_restore(adapter);
5367 	}
5368 
5369 	switch (hw->mac.type) {
5370 	case ixgbe_mac_82599EB:
5371 	case ixgbe_mac_X540:
5372 		hw->mac.ops.enable_rx_buff(hw);
5373 		break;
5374 	default:
5375 		break;
5376 	}
5377 
5378 #ifdef CONFIG_IXGBE_DCA
5379 	/* configure DCA */
5380 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5381 		ixgbe_setup_dca(adapter);
5382 #endif /* CONFIG_IXGBE_DCA */
5383 
5384 #ifdef IXGBE_FCOE
5385 	/* configure FCoE L2 filters, redirection table, and Rx control */
5386 	ixgbe_configure_fcoe(adapter);
5387 
5388 #endif /* IXGBE_FCOE */
5389 	ixgbe_configure_tx(adapter);
5390 	ixgbe_configure_rx(adapter);
5391 	ixgbe_configure_dfwd(adapter);
5392 }
5393 
5394 /**
5395  * ixgbe_sfp_link_config - set up SFP+ link
5396  * @adapter: pointer to private adapter struct
5397  **/
5398 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5399 {
5400 	/*
5401 	 * We are assuming the worst case scenario here, and that
5402 	 * is that an SFP was inserted/removed after the reset
5403 	 * but before SFP detection was enabled.  As such the best
5404 	 * solution is to just start searching as soon as we start
5405 	 */
5406 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5407 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5408 
5409 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5410 	adapter->sfp_poll_time = 0;
5411 }
5412 
5413 /**
5414  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5415  * @hw: pointer to private hardware struct
5416  *
5417  * Returns 0 on success, negative on failure
5418  **/
5419 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5420 {
5421 	u32 speed;
5422 	bool autoneg, link_up = false;
5423 	int ret = IXGBE_ERR_LINK_SETUP;
5424 
5425 	if (hw->mac.ops.check_link)
5426 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5427 
5428 	if (ret)
5429 		return ret;
5430 
5431 	speed = hw->phy.autoneg_advertised;
5432 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5433 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5434 							&autoneg);
5435 	if (ret)
5436 		return ret;
5437 
5438 	if (hw->mac.ops.setup_link)
5439 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5440 
5441 	return ret;
5442 }
5443 
5444 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5445 {
5446 	struct ixgbe_hw *hw = &adapter->hw;
5447 	u32 gpie = 0;
5448 
5449 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5450 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5451 		       IXGBE_GPIE_OCD;
5452 		gpie |= IXGBE_GPIE_EIAME;
5453 		/*
5454 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5455 		 * this saves a register write for every interrupt
5456 		 */
5457 		switch (hw->mac.type) {
5458 		case ixgbe_mac_82598EB:
5459 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5460 			break;
5461 		case ixgbe_mac_82599EB:
5462 		case ixgbe_mac_X540:
5463 		case ixgbe_mac_X550:
5464 		case ixgbe_mac_X550EM_x:
5465 		case ixgbe_mac_x550em_a:
5466 		default:
5467 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5468 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5469 			break;
5470 		}
5471 	} else {
5472 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5473 		 * specifically only auto mask tx and rx interrupts */
5474 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5475 	}
5476 
5477 	/* XXX: to interrupt immediately for EICS writes, enable this */
5478 	/* gpie |= IXGBE_GPIE_EIMEN; */
5479 
5480 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5481 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5482 
5483 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5484 		case IXGBE_82599_VMDQ_8Q_MASK:
5485 			gpie |= IXGBE_GPIE_VTMODE_16;
5486 			break;
5487 		case IXGBE_82599_VMDQ_4Q_MASK:
5488 			gpie |= IXGBE_GPIE_VTMODE_32;
5489 			break;
5490 		default:
5491 			gpie |= IXGBE_GPIE_VTMODE_64;
5492 			break;
5493 		}
5494 	}
5495 
5496 	/* Enable Thermal over heat sensor interrupt */
5497 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5498 		switch (adapter->hw.mac.type) {
5499 		case ixgbe_mac_82599EB:
5500 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5501 			break;
5502 		default:
5503 			break;
5504 		}
5505 	}
5506 
5507 	/* Enable fan failure interrupt */
5508 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5509 		gpie |= IXGBE_SDP1_GPIEN(hw);
5510 
5511 	switch (hw->mac.type) {
5512 	case ixgbe_mac_82599EB:
5513 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5514 		break;
5515 	case ixgbe_mac_X550EM_x:
5516 	case ixgbe_mac_x550em_a:
5517 		gpie |= IXGBE_SDP0_GPIEN_X540;
5518 		break;
5519 	default:
5520 		break;
5521 	}
5522 
5523 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5524 }
5525 
5526 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5527 {
5528 	struct ixgbe_hw *hw = &adapter->hw;
5529 	int err;
5530 	u32 ctrl_ext;
5531 
5532 	ixgbe_get_hw_control(adapter);
5533 	ixgbe_setup_gpie(adapter);
5534 
5535 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5536 		ixgbe_configure_msix(adapter);
5537 	else
5538 		ixgbe_configure_msi_and_legacy(adapter);
5539 
5540 	/* enable the optics for 82599 SFP+ fiber */
5541 	if (hw->mac.ops.enable_tx_laser)
5542 		hw->mac.ops.enable_tx_laser(hw);
5543 
5544 	if (hw->phy.ops.set_phy_power)
5545 		hw->phy.ops.set_phy_power(hw, true);
5546 
5547 	smp_mb__before_atomic();
5548 	clear_bit(__IXGBE_DOWN, &adapter->state);
5549 	ixgbe_napi_enable_all(adapter);
5550 
5551 	if (ixgbe_is_sfp(hw)) {
5552 		ixgbe_sfp_link_config(adapter);
5553 	} else {
5554 		err = ixgbe_non_sfp_link_config(hw);
5555 		if (err)
5556 			e_err(probe, "link_config FAILED %d\n", err);
5557 	}
5558 
5559 	/* clear any pending interrupts, may auto mask */
5560 	IXGBE_READ_REG(hw, IXGBE_EICR);
5561 	ixgbe_irq_enable(adapter, true, true);
5562 
5563 	/*
5564 	 * If this adapter has a fan, check to see if we had a failure
5565 	 * before we enabled the interrupt.
5566 	 */
5567 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5568 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5569 		if (esdp & IXGBE_ESDP_SDP1)
5570 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5571 	}
5572 
5573 	/* bring the link up in the watchdog, this could race with our first
5574 	 * link up interrupt but shouldn't be a problem */
5575 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5576 	adapter->link_check_timeout = jiffies;
5577 	mod_timer(&adapter->service_timer, jiffies);
5578 
5579 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5580 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5581 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5582 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5583 }
5584 
5585 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5586 {
5587 	WARN_ON(in_interrupt());
5588 	/* put off any impending NetWatchDogTimeout */
5589 	netif_trans_update(adapter->netdev);
5590 
5591 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5592 		usleep_range(1000, 2000);
5593 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5594 		ixgbe_watchdog_link_is_down(adapter);
5595 	ixgbe_down(adapter);
5596 	/*
5597 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5598 	 * back up to give the VFs time to respond to the reset.  The
5599 	 * two second wait is based upon the watchdog timer cycle in
5600 	 * the VF driver.
5601 	 */
5602 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5603 		msleep(2000);
5604 	ixgbe_up(adapter);
5605 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5606 }
5607 
5608 void ixgbe_up(struct ixgbe_adapter *adapter)
5609 {
5610 	/* hardware has been reset, we need to reload some things */
5611 	ixgbe_configure(adapter);
5612 
5613 	ixgbe_up_complete(adapter);
5614 }
5615 
5616 void ixgbe_reset(struct ixgbe_adapter *adapter)
5617 {
5618 	struct ixgbe_hw *hw = &adapter->hw;
5619 	struct net_device *netdev = adapter->netdev;
5620 	int err;
5621 
5622 	if (ixgbe_removed(hw->hw_addr))
5623 		return;
5624 	/* lock SFP init bit to prevent race conditions with the watchdog */
5625 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5626 		usleep_range(1000, 2000);
5627 
5628 	/* clear all SFP and link config related flags while holding SFP_INIT */
5629 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5630 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5631 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5632 
5633 	err = hw->mac.ops.init_hw(hw);
5634 	switch (err) {
5635 	case 0:
5636 	case IXGBE_ERR_SFP_NOT_PRESENT:
5637 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5638 		break;
5639 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5640 		e_dev_err("master disable timed out\n");
5641 		break;
5642 	case IXGBE_ERR_EEPROM_VERSION:
5643 		/* We are running on a pre-production device, log a warning */
5644 		e_dev_warn("This device is a pre-production adapter/LOM. "
5645 			   "Please be aware there may be issues associated with "
5646 			   "your hardware.  If you are experiencing problems "
5647 			   "please contact your Intel or hardware "
5648 			   "representative who provided you with this "
5649 			   "hardware.\n");
5650 		break;
5651 	default:
5652 		e_dev_err("Hardware Error: %d\n", err);
5653 	}
5654 
5655 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5656 
5657 	/* flush entries out of MAC table */
5658 	ixgbe_flush_sw_mac_table(adapter);
5659 	__dev_uc_unsync(netdev, NULL);
5660 
5661 	/* do not flush user set addresses */
5662 	ixgbe_mac_set_default_filter(adapter);
5663 
5664 	/* update SAN MAC vmdq pool selection */
5665 	if (hw->mac.san_mac_rar_index)
5666 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5667 
5668 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5669 		ixgbe_ptp_reset(adapter);
5670 
5671 	if (hw->phy.ops.set_phy_power) {
5672 		if (!netif_running(adapter->netdev) && !adapter->wol)
5673 			hw->phy.ops.set_phy_power(hw, false);
5674 		else
5675 			hw->phy.ops.set_phy_power(hw, true);
5676 	}
5677 }
5678 
5679 /**
5680  * ixgbe_clean_tx_ring - Free Tx Buffers
5681  * @tx_ring: ring to be cleaned
5682  **/
5683 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5684 {
5685 	u16 i = tx_ring->next_to_clean;
5686 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
5687 
5688 	while (i != tx_ring->next_to_use) {
5689 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
5690 
5691 		/* Free all the Tx ring sk_buffs */
5692 		if (ring_is_xdp(tx_ring))
5693 			page_frag_free(tx_buffer->data);
5694 		else
5695 			dev_kfree_skb_any(tx_buffer->skb);
5696 
5697 		/* unmap skb header data */
5698 		dma_unmap_single(tx_ring->dev,
5699 				 dma_unmap_addr(tx_buffer, dma),
5700 				 dma_unmap_len(tx_buffer, len),
5701 				 DMA_TO_DEVICE);
5702 
5703 		/* check for eop_desc to determine the end of the packet */
5704 		eop_desc = tx_buffer->next_to_watch;
5705 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
5706 
5707 		/* unmap remaining buffers */
5708 		while (tx_desc != eop_desc) {
5709 			tx_buffer++;
5710 			tx_desc++;
5711 			i++;
5712 			if (unlikely(i == tx_ring->count)) {
5713 				i = 0;
5714 				tx_buffer = tx_ring->tx_buffer_info;
5715 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
5716 			}
5717 
5718 			/* unmap any remaining paged data */
5719 			if (dma_unmap_len(tx_buffer, len))
5720 				dma_unmap_page(tx_ring->dev,
5721 					       dma_unmap_addr(tx_buffer, dma),
5722 					       dma_unmap_len(tx_buffer, len),
5723 					       DMA_TO_DEVICE);
5724 		}
5725 
5726 		/* move us one more past the eop_desc for start of next pkt */
5727 		tx_buffer++;
5728 		i++;
5729 		if (unlikely(i == tx_ring->count)) {
5730 			i = 0;
5731 			tx_buffer = tx_ring->tx_buffer_info;
5732 		}
5733 	}
5734 
5735 	/* reset BQL for queue */
5736 	if (!ring_is_xdp(tx_ring))
5737 		netdev_tx_reset_queue(txring_txq(tx_ring));
5738 
5739 	/* reset next_to_use and next_to_clean */
5740 	tx_ring->next_to_use = 0;
5741 	tx_ring->next_to_clean = 0;
5742 }
5743 
5744 /**
5745  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5746  * @adapter: board private structure
5747  **/
5748 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5749 {
5750 	int i;
5751 
5752 	for (i = 0; i < adapter->num_rx_queues; i++)
5753 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5754 }
5755 
5756 /**
5757  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5758  * @adapter: board private structure
5759  **/
5760 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5761 {
5762 	int i;
5763 
5764 	for (i = 0; i < adapter->num_tx_queues; i++)
5765 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5766 	for (i = 0; i < adapter->num_xdp_queues; i++)
5767 		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
5768 }
5769 
5770 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5771 {
5772 	struct hlist_node *node2;
5773 	struct ixgbe_fdir_filter *filter;
5774 
5775 	spin_lock(&adapter->fdir_perfect_lock);
5776 
5777 	hlist_for_each_entry_safe(filter, node2,
5778 				  &adapter->fdir_filter_list, fdir_node) {
5779 		hlist_del(&filter->fdir_node);
5780 		kfree(filter);
5781 	}
5782 	adapter->fdir_filter_count = 0;
5783 
5784 	spin_unlock(&adapter->fdir_perfect_lock);
5785 }
5786 
5787 static int ixgbe_disable_macvlan(struct net_device *upper, void *data)
5788 {
5789 	if (netif_is_macvlan(upper)) {
5790 		struct macvlan_dev *vlan = netdev_priv(upper);
5791 
5792 		if (vlan->fwd_priv) {
5793 			netif_tx_stop_all_queues(upper);
5794 			netif_carrier_off(upper);
5795 			netif_tx_disable(upper);
5796 		}
5797 	}
5798 
5799 	return 0;
5800 }
5801 
5802 void ixgbe_down(struct ixgbe_adapter *adapter)
5803 {
5804 	struct net_device *netdev = adapter->netdev;
5805 	struct ixgbe_hw *hw = &adapter->hw;
5806 	int i;
5807 
5808 	/* signal that we are down to the interrupt handler */
5809 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5810 		return; /* do nothing if already down */
5811 
5812 	/* disable receives */
5813 	hw->mac.ops.disable_rx(hw);
5814 
5815 	/* disable all enabled rx queues */
5816 	for (i = 0; i < adapter->num_rx_queues; i++)
5817 		/* this call also flushes the previous write */
5818 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5819 
5820 	usleep_range(10000, 20000);
5821 
5822 	/* synchronize_sched() needed for pending XDP buffers to drain */
5823 	if (adapter->xdp_ring[0])
5824 		synchronize_sched();
5825 	netif_tx_stop_all_queues(netdev);
5826 
5827 	/* call carrier off first to avoid false dev_watchdog timeouts */
5828 	netif_carrier_off(netdev);
5829 	netif_tx_disable(netdev);
5830 
5831 	/* disable any upper devices */
5832 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5833 				      ixgbe_disable_macvlan, NULL);
5834 
5835 	ixgbe_irq_disable(adapter);
5836 
5837 	ixgbe_napi_disable_all(adapter);
5838 
5839 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
5840 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5841 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5842 
5843 	del_timer_sync(&adapter->service_timer);
5844 
5845 	if (adapter->num_vfs) {
5846 		/* Clear EITR Select mapping */
5847 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5848 
5849 		/* Mark all the VFs as inactive */
5850 		for (i = 0 ; i < adapter->num_vfs; i++)
5851 			adapter->vfinfo[i].clear_to_send = false;
5852 
5853 		/* ping all the active vfs to let them know we are going down */
5854 		ixgbe_ping_all_vfs(adapter);
5855 
5856 		/* Disable all VFTE/VFRE TX/RX */
5857 		ixgbe_disable_tx_rx(adapter);
5858 	}
5859 
5860 	/* disable transmits in the hardware now that interrupts are off */
5861 	for (i = 0; i < adapter->num_tx_queues; i++) {
5862 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5863 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5864 	}
5865 	for (i = 0; i < adapter->num_xdp_queues; i++) {
5866 		u8 reg_idx = adapter->xdp_ring[i]->reg_idx;
5867 
5868 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5869 	}
5870 
5871 	/* Disable the Tx DMA engine on 82599 and later MAC */
5872 	switch (hw->mac.type) {
5873 	case ixgbe_mac_82599EB:
5874 	case ixgbe_mac_X540:
5875 	case ixgbe_mac_X550:
5876 	case ixgbe_mac_X550EM_x:
5877 	case ixgbe_mac_x550em_a:
5878 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5879 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5880 				 ~IXGBE_DMATXCTL_TE));
5881 		break;
5882 	default:
5883 		break;
5884 	}
5885 
5886 	if (!pci_channel_offline(adapter->pdev))
5887 		ixgbe_reset(adapter);
5888 
5889 	/* power down the optics for 82599 SFP+ fiber */
5890 	if (hw->mac.ops.disable_tx_laser)
5891 		hw->mac.ops.disable_tx_laser(hw);
5892 
5893 	ixgbe_clean_all_tx_rings(adapter);
5894 	ixgbe_clean_all_rx_rings(adapter);
5895 }
5896 
5897 /**
5898  * ixgbe_eee_capable - helper function to determine EEE support on X550
5899  * @adapter: board private structure
5900  */
5901 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
5902 {
5903 	struct ixgbe_hw *hw = &adapter->hw;
5904 
5905 	switch (hw->device_id) {
5906 	case IXGBE_DEV_ID_X550EM_A_1G_T:
5907 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
5908 		if (!hw->phy.eee_speeds_supported)
5909 			break;
5910 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
5911 		if (!hw->phy.eee_speeds_advertised)
5912 			break;
5913 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
5914 		break;
5915 	default:
5916 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
5917 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
5918 		break;
5919 	}
5920 }
5921 
5922 /**
5923  * ixgbe_tx_timeout - Respond to a Tx Hang
5924  * @netdev: network interface device structure
5925  **/
5926 static void ixgbe_tx_timeout(struct net_device *netdev)
5927 {
5928 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5929 
5930 	/* Do the reset outside of interrupt context */
5931 	ixgbe_tx_timeout_reset(adapter);
5932 }
5933 
5934 #ifdef CONFIG_IXGBE_DCB
5935 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
5936 {
5937 	struct ixgbe_hw *hw = &adapter->hw;
5938 	struct tc_configuration *tc;
5939 	int j;
5940 
5941 	switch (hw->mac.type) {
5942 	case ixgbe_mac_82598EB:
5943 	case ixgbe_mac_82599EB:
5944 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5945 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5946 		break;
5947 	case ixgbe_mac_X540:
5948 	case ixgbe_mac_X550:
5949 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5950 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5951 		break;
5952 	case ixgbe_mac_X550EM_x:
5953 	case ixgbe_mac_x550em_a:
5954 	default:
5955 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
5956 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
5957 		break;
5958 	}
5959 
5960 	/* Configure DCB traffic classes */
5961 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5962 		tc = &adapter->dcb_cfg.tc_config[j];
5963 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5964 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5965 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5966 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5967 		tc->dcb_pfc = pfc_disabled;
5968 	}
5969 
5970 	/* Initialize default user to priority mapping, UPx->TC0 */
5971 	tc = &adapter->dcb_cfg.tc_config[0];
5972 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5973 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5974 
5975 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5976 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5977 	adapter->dcb_cfg.pfc_mode_enable = false;
5978 	adapter->dcb_set_bitmap = 0x00;
5979 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
5980 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5981 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5982 	       sizeof(adapter->temp_dcb_cfg));
5983 }
5984 #endif
5985 
5986 /**
5987  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5988  * @adapter: board private structure to initialize
5989  *
5990  * ixgbe_sw_init initializes the Adapter private data structure.
5991  * Fields are initialized based on PCI device information and
5992  * OS network device settings (MTU size).
5993  **/
5994 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
5995 			 const struct ixgbe_info *ii)
5996 {
5997 	struct ixgbe_hw *hw = &adapter->hw;
5998 	struct pci_dev *pdev = adapter->pdev;
5999 	unsigned int rss, fdir;
6000 	u32 fwsm;
6001 	int i;
6002 
6003 	/* PCI config space info */
6004 
6005 	hw->vendor_id = pdev->vendor;
6006 	hw->device_id = pdev->device;
6007 	hw->revision_id = pdev->revision;
6008 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6009 	hw->subsystem_device_id = pdev->subsystem_device;
6010 
6011 	/* get_invariants needs the device IDs */
6012 	ii->get_invariants(hw);
6013 
6014 	/* Set common capability flags and settings */
6015 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6016 	adapter->ring_feature[RING_F_RSS].limit = rss;
6017 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6018 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6019 	adapter->atr_sample_rate = 20;
6020 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6021 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
6022 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6023 #ifdef CONFIG_IXGBE_DCA
6024 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6025 #endif
6026 #ifdef CONFIG_IXGBE_DCB
6027 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6028 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6029 #endif
6030 #ifdef IXGBE_FCOE
6031 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6032 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6033 #ifdef CONFIG_IXGBE_DCB
6034 	/* Default traffic class to use for FCoE */
6035 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6036 #endif /* CONFIG_IXGBE_DCB */
6037 #endif /* IXGBE_FCOE */
6038 
6039 	/* initialize static ixgbe jump table entries */
6040 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6041 					  GFP_KERNEL);
6042 	if (!adapter->jump_tables[0])
6043 		return -ENOMEM;
6044 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6045 
6046 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6047 		adapter->jump_tables[i] = NULL;
6048 
6049 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
6050 				     hw->mac.num_rar_entries,
6051 				     GFP_ATOMIC);
6052 	if (!adapter->mac_table)
6053 		return -ENOMEM;
6054 
6055 	if (ixgbe_init_rss_key(adapter))
6056 		return -ENOMEM;
6057 
6058 	/* Set MAC specific capability flags and exceptions */
6059 	switch (hw->mac.type) {
6060 	case ixgbe_mac_82598EB:
6061 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6062 
6063 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
6064 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6065 
6066 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6067 		adapter->ring_feature[RING_F_FDIR].limit = 0;
6068 		adapter->atr_sample_rate = 0;
6069 		adapter->fdir_pballoc = 0;
6070 #ifdef IXGBE_FCOE
6071 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6072 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6073 #ifdef CONFIG_IXGBE_DCB
6074 		adapter->fcoe.up = 0;
6075 #endif /* IXGBE_DCB */
6076 #endif /* IXGBE_FCOE */
6077 		break;
6078 	case ixgbe_mac_82599EB:
6079 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6080 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6081 		break;
6082 	case ixgbe_mac_X540:
6083 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6084 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
6085 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6086 		break;
6087 	case ixgbe_mac_x550em_a:
6088 		adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
6089 		switch (hw->device_id) {
6090 		case IXGBE_DEV_ID_X550EM_A_1G_T:
6091 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6092 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6093 			break;
6094 		default:
6095 			break;
6096 		}
6097 	/* fall through */
6098 	case ixgbe_mac_X550EM_x:
6099 #ifdef CONFIG_IXGBE_DCB
6100 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6101 #endif
6102 #ifdef IXGBE_FCOE
6103 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6104 #ifdef CONFIG_IXGBE_DCB
6105 		adapter->fcoe.up = 0;
6106 #endif /* IXGBE_DCB */
6107 #endif /* IXGBE_FCOE */
6108 	/* Fall Through */
6109 	case ixgbe_mac_X550:
6110 		if (hw->mac.type == ixgbe_mac_X550)
6111 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6112 #ifdef CONFIG_IXGBE_DCA
6113 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6114 #endif
6115 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
6116 		break;
6117 	default:
6118 		break;
6119 	}
6120 
6121 #ifdef IXGBE_FCOE
6122 	/* FCoE support exists, always init the FCoE lock */
6123 	spin_lock_init(&adapter->fcoe.lock);
6124 
6125 #endif
6126 	/* n-tuple support exists, always init our spinlock */
6127 	spin_lock_init(&adapter->fdir_perfect_lock);
6128 
6129 #ifdef CONFIG_IXGBE_DCB
6130 	ixgbe_init_dcb(adapter);
6131 #endif
6132 
6133 	/* default flow control settings */
6134 	hw->fc.requested_mode = ixgbe_fc_full;
6135 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
6136 	ixgbe_pbthresh_setup(adapter);
6137 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6138 	hw->fc.send_xon = true;
6139 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6140 
6141 #ifdef CONFIG_PCI_IOV
6142 	if (max_vfs > 0)
6143 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6144 
6145 	/* assign number of SR-IOV VFs */
6146 	if (hw->mac.type != ixgbe_mac_82598EB) {
6147 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6148 			max_vfs = 0;
6149 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6150 		}
6151 	}
6152 #endif /* CONFIG_PCI_IOV */
6153 
6154 	/* enable itr by default in dynamic mode */
6155 	adapter->rx_itr_setting = 1;
6156 	adapter->tx_itr_setting = 1;
6157 
6158 	/* set default ring sizes */
6159 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6160 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6161 
6162 	/* set default work limits */
6163 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6164 
6165 	/* initialize eeprom parameters */
6166 	if (ixgbe_init_eeprom_params_generic(hw)) {
6167 		e_dev_err("EEPROM initialization failed\n");
6168 		return -EIO;
6169 	}
6170 
6171 	/* PF holds first pool slot */
6172 	set_bit(0, &adapter->fwd_bitmask);
6173 	set_bit(__IXGBE_DOWN, &adapter->state);
6174 
6175 	return 0;
6176 }
6177 
6178 /**
6179  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6180  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6181  *
6182  * Return 0 on success, negative on failure
6183  **/
6184 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6185 {
6186 	struct device *dev = tx_ring->dev;
6187 	int orig_node = dev_to_node(dev);
6188 	int ring_node = -1;
6189 	int size;
6190 
6191 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6192 
6193 	if (tx_ring->q_vector)
6194 		ring_node = tx_ring->q_vector->numa_node;
6195 
6196 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6197 	if (!tx_ring->tx_buffer_info)
6198 		tx_ring->tx_buffer_info = vmalloc(size);
6199 	if (!tx_ring->tx_buffer_info)
6200 		goto err;
6201 
6202 	/* round up to nearest 4K */
6203 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6204 	tx_ring->size = ALIGN(tx_ring->size, 4096);
6205 
6206 	set_dev_node(dev, ring_node);
6207 	tx_ring->desc = dma_alloc_coherent(dev,
6208 					   tx_ring->size,
6209 					   &tx_ring->dma,
6210 					   GFP_KERNEL);
6211 	set_dev_node(dev, orig_node);
6212 	if (!tx_ring->desc)
6213 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6214 						   &tx_ring->dma, GFP_KERNEL);
6215 	if (!tx_ring->desc)
6216 		goto err;
6217 
6218 	tx_ring->next_to_use = 0;
6219 	tx_ring->next_to_clean = 0;
6220 	return 0;
6221 
6222 err:
6223 	vfree(tx_ring->tx_buffer_info);
6224 	tx_ring->tx_buffer_info = NULL;
6225 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6226 	return -ENOMEM;
6227 }
6228 
6229 /**
6230  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6231  * @adapter: board private structure
6232  *
6233  * If this function returns with an error, then it's possible one or
6234  * more of the rings is populated (while the rest are not).  It is the
6235  * callers duty to clean those orphaned rings.
6236  *
6237  * Return 0 on success, negative on failure
6238  **/
6239 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6240 {
6241 	int i, j = 0, err = 0;
6242 
6243 	for (i = 0; i < adapter->num_tx_queues; i++) {
6244 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6245 		if (!err)
6246 			continue;
6247 
6248 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6249 		goto err_setup_tx;
6250 	}
6251 	for (j = 0; j < adapter->num_xdp_queues; j++) {
6252 		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6253 		if (!err)
6254 			continue;
6255 
6256 		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6257 		goto err_setup_tx;
6258 	}
6259 
6260 	return 0;
6261 err_setup_tx:
6262 	/* rewind the index freeing the rings as we go */
6263 	while (j--)
6264 		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6265 	while (i--)
6266 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
6267 	return err;
6268 }
6269 
6270 /**
6271  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6272  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6273  *
6274  * Returns 0 on success, negative on failure
6275  **/
6276 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6277 			     struct ixgbe_ring *rx_ring)
6278 {
6279 	struct device *dev = rx_ring->dev;
6280 	int orig_node = dev_to_node(dev);
6281 	int ring_node = -1;
6282 	int size;
6283 
6284 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6285 
6286 	if (rx_ring->q_vector)
6287 		ring_node = rx_ring->q_vector->numa_node;
6288 
6289 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6290 	if (!rx_ring->rx_buffer_info)
6291 		rx_ring->rx_buffer_info = vmalloc(size);
6292 	if (!rx_ring->rx_buffer_info)
6293 		goto err;
6294 
6295 	/* Round up to nearest 4K */
6296 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6297 	rx_ring->size = ALIGN(rx_ring->size, 4096);
6298 
6299 	set_dev_node(dev, ring_node);
6300 	rx_ring->desc = dma_alloc_coherent(dev,
6301 					   rx_ring->size,
6302 					   &rx_ring->dma,
6303 					   GFP_KERNEL);
6304 	set_dev_node(dev, orig_node);
6305 	if (!rx_ring->desc)
6306 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6307 						   &rx_ring->dma, GFP_KERNEL);
6308 	if (!rx_ring->desc)
6309 		goto err;
6310 
6311 	rx_ring->next_to_clean = 0;
6312 	rx_ring->next_to_use = 0;
6313 
6314 	rx_ring->xdp_prog = adapter->xdp_prog;
6315 
6316 	return 0;
6317 err:
6318 	vfree(rx_ring->rx_buffer_info);
6319 	rx_ring->rx_buffer_info = NULL;
6320 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6321 	return -ENOMEM;
6322 }
6323 
6324 /**
6325  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6326  * @adapter: board private structure
6327  *
6328  * If this function returns with an error, then it's possible one or
6329  * more of the rings is populated (while the rest are not).  It is the
6330  * callers duty to clean those orphaned rings.
6331  *
6332  * Return 0 on success, negative on failure
6333  **/
6334 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6335 {
6336 	int i, err = 0;
6337 
6338 	for (i = 0; i < adapter->num_rx_queues; i++) {
6339 		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6340 		if (!err)
6341 			continue;
6342 
6343 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6344 		goto err_setup_rx;
6345 	}
6346 
6347 #ifdef IXGBE_FCOE
6348 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
6349 	if (!err)
6350 #endif
6351 		return 0;
6352 err_setup_rx:
6353 	/* rewind the index freeing the rings as we go */
6354 	while (i--)
6355 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
6356 	return err;
6357 }
6358 
6359 /**
6360  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6361  * @tx_ring: Tx descriptor ring for a specific queue
6362  *
6363  * Free all transmit software resources
6364  **/
6365 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6366 {
6367 	ixgbe_clean_tx_ring(tx_ring);
6368 
6369 	vfree(tx_ring->tx_buffer_info);
6370 	tx_ring->tx_buffer_info = NULL;
6371 
6372 	/* if not set, then don't free */
6373 	if (!tx_ring->desc)
6374 		return;
6375 
6376 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6377 			  tx_ring->desc, tx_ring->dma);
6378 
6379 	tx_ring->desc = NULL;
6380 }
6381 
6382 /**
6383  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6384  * @adapter: board private structure
6385  *
6386  * Free all transmit software resources
6387  **/
6388 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6389 {
6390 	int i;
6391 
6392 	for (i = 0; i < adapter->num_tx_queues; i++)
6393 		if (adapter->tx_ring[i]->desc)
6394 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6395 	for (i = 0; i < adapter->num_xdp_queues; i++)
6396 		if (adapter->xdp_ring[i]->desc)
6397 			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6398 }
6399 
6400 /**
6401  * ixgbe_free_rx_resources - Free Rx Resources
6402  * @rx_ring: ring to clean the resources from
6403  *
6404  * Free all receive software resources
6405  **/
6406 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6407 {
6408 	ixgbe_clean_rx_ring(rx_ring);
6409 
6410 	rx_ring->xdp_prog = NULL;
6411 	vfree(rx_ring->rx_buffer_info);
6412 	rx_ring->rx_buffer_info = NULL;
6413 
6414 	/* if not set, then don't free */
6415 	if (!rx_ring->desc)
6416 		return;
6417 
6418 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6419 			  rx_ring->desc, rx_ring->dma);
6420 
6421 	rx_ring->desc = NULL;
6422 }
6423 
6424 /**
6425  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6426  * @adapter: board private structure
6427  *
6428  * Free all receive software resources
6429  **/
6430 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6431 {
6432 	int i;
6433 
6434 #ifdef IXGBE_FCOE
6435 	ixgbe_free_fcoe_ddp_resources(adapter);
6436 
6437 #endif
6438 	for (i = 0; i < adapter->num_rx_queues; i++)
6439 		if (adapter->rx_ring[i]->desc)
6440 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6441 }
6442 
6443 /**
6444  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6445  * @netdev: network interface device structure
6446  * @new_mtu: new value for maximum frame size
6447  *
6448  * Returns 0 on success, negative on failure
6449  **/
6450 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6451 {
6452 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6453 
6454 	/*
6455 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6456 	 * paths when MTU greater than 1500 is configured.  So display a
6457 	 * warning that legacy VFs will be disabled.
6458 	 */
6459 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6460 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6461 	    (new_mtu > ETH_DATA_LEN))
6462 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6463 
6464 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6465 
6466 	/* must set new MTU before calling down or up */
6467 	netdev->mtu = new_mtu;
6468 
6469 	if (netif_running(netdev))
6470 		ixgbe_reinit_locked(adapter);
6471 
6472 	return 0;
6473 }
6474 
6475 /**
6476  * ixgbe_open - Called when a network interface is made active
6477  * @netdev: network interface device structure
6478  *
6479  * Returns 0 on success, negative value on failure
6480  *
6481  * The open entry point is called when a network interface is made
6482  * active by the system (IFF_UP).  At this point all resources needed
6483  * for transmit and receive operations are allocated, the interrupt
6484  * handler is registered with the OS, the watchdog timer is started,
6485  * and the stack is notified that the interface is ready.
6486  **/
6487 int ixgbe_open(struct net_device *netdev)
6488 {
6489 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6490 	struct ixgbe_hw *hw = &adapter->hw;
6491 	int err, queues;
6492 
6493 	/* disallow open during test */
6494 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6495 		return -EBUSY;
6496 
6497 	netif_carrier_off(netdev);
6498 
6499 	/* allocate transmit descriptors */
6500 	err = ixgbe_setup_all_tx_resources(adapter);
6501 	if (err)
6502 		goto err_setup_tx;
6503 
6504 	/* allocate receive descriptors */
6505 	err = ixgbe_setup_all_rx_resources(adapter);
6506 	if (err)
6507 		goto err_setup_rx;
6508 
6509 	ixgbe_configure(adapter);
6510 
6511 	err = ixgbe_request_irq(adapter);
6512 	if (err)
6513 		goto err_req_irq;
6514 
6515 	/* Notify the stack of the actual queue counts. */
6516 	if (adapter->num_rx_pools > 1)
6517 		queues = adapter->num_rx_queues_per_pool;
6518 	else
6519 		queues = adapter->num_tx_queues;
6520 
6521 	err = netif_set_real_num_tx_queues(netdev, queues);
6522 	if (err)
6523 		goto err_set_queues;
6524 
6525 	if (adapter->num_rx_pools > 1 &&
6526 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
6527 		queues = IXGBE_MAX_L2A_QUEUES;
6528 	else
6529 		queues = adapter->num_rx_queues;
6530 	err = netif_set_real_num_rx_queues(netdev, queues);
6531 	if (err)
6532 		goto err_set_queues;
6533 
6534 	ixgbe_ptp_init(adapter);
6535 
6536 	ixgbe_up_complete(adapter);
6537 
6538 	ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6539 	udp_tunnel_get_rx_info(netdev);
6540 
6541 	return 0;
6542 
6543 err_set_queues:
6544 	ixgbe_free_irq(adapter);
6545 err_req_irq:
6546 	ixgbe_free_all_rx_resources(adapter);
6547 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6548 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6549 err_setup_rx:
6550 	ixgbe_free_all_tx_resources(adapter);
6551 err_setup_tx:
6552 	ixgbe_reset(adapter);
6553 
6554 	return err;
6555 }
6556 
6557 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6558 {
6559 	ixgbe_ptp_suspend(adapter);
6560 
6561 	if (adapter->hw.phy.ops.enter_lplu) {
6562 		adapter->hw.phy.reset_disable = true;
6563 		ixgbe_down(adapter);
6564 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6565 		adapter->hw.phy.reset_disable = false;
6566 	} else {
6567 		ixgbe_down(adapter);
6568 	}
6569 
6570 	ixgbe_free_irq(adapter);
6571 
6572 	ixgbe_free_all_tx_resources(adapter);
6573 	ixgbe_free_all_rx_resources(adapter);
6574 }
6575 
6576 /**
6577  * ixgbe_close - Disables a network interface
6578  * @netdev: network interface device structure
6579  *
6580  * Returns 0, this is not allowed to fail
6581  *
6582  * The close entry point is called when an interface is de-activated
6583  * by the OS.  The hardware is still under the drivers control, but
6584  * needs to be disabled.  A global MAC reset is issued to stop the
6585  * hardware, and all transmit and receive resources are freed.
6586  **/
6587 int ixgbe_close(struct net_device *netdev)
6588 {
6589 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6590 
6591 	ixgbe_ptp_stop(adapter);
6592 
6593 	if (netif_device_present(netdev))
6594 		ixgbe_close_suspend(adapter);
6595 
6596 	ixgbe_fdir_filter_exit(adapter);
6597 
6598 	ixgbe_release_hw_control(adapter);
6599 
6600 	return 0;
6601 }
6602 
6603 #ifdef CONFIG_PM
6604 static int ixgbe_resume(struct pci_dev *pdev)
6605 {
6606 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6607 	struct net_device *netdev = adapter->netdev;
6608 	u32 err;
6609 
6610 	adapter->hw.hw_addr = adapter->io_addr;
6611 	pci_set_power_state(pdev, PCI_D0);
6612 	pci_restore_state(pdev);
6613 	/*
6614 	 * pci_restore_state clears dev->state_saved so call
6615 	 * pci_save_state to restore it.
6616 	 */
6617 	pci_save_state(pdev);
6618 
6619 	err = pci_enable_device_mem(pdev);
6620 	if (err) {
6621 		e_dev_err("Cannot enable PCI device from suspend\n");
6622 		return err;
6623 	}
6624 	smp_mb__before_atomic();
6625 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6626 	pci_set_master(pdev);
6627 
6628 	pci_wake_from_d3(pdev, false);
6629 
6630 	ixgbe_reset(adapter);
6631 
6632 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6633 
6634 	rtnl_lock();
6635 	err = ixgbe_init_interrupt_scheme(adapter);
6636 	if (!err && netif_running(netdev))
6637 		err = ixgbe_open(netdev);
6638 
6639 
6640 	if (!err)
6641 		netif_device_attach(netdev);
6642 	rtnl_unlock();
6643 
6644 	return err;
6645 }
6646 #endif /* CONFIG_PM */
6647 
6648 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6649 {
6650 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6651 	struct net_device *netdev = adapter->netdev;
6652 	struct ixgbe_hw *hw = &adapter->hw;
6653 	u32 ctrl, fctrl;
6654 	u32 wufc = adapter->wol;
6655 #ifdef CONFIG_PM
6656 	int retval = 0;
6657 #endif
6658 
6659 	rtnl_lock();
6660 	netif_device_detach(netdev);
6661 
6662 	if (netif_running(netdev))
6663 		ixgbe_close_suspend(adapter);
6664 
6665 	ixgbe_clear_interrupt_scheme(adapter);
6666 	rtnl_unlock();
6667 
6668 #ifdef CONFIG_PM
6669 	retval = pci_save_state(pdev);
6670 	if (retval)
6671 		return retval;
6672 
6673 #endif
6674 	if (hw->mac.ops.stop_link_on_d3)
6675 		hw->mac.ops.stop_link_on_d3(hw);
6676 
6677 	if (wufc) {
6678 		ixgbe_set_rx_mode(netdev);
6679 
6680 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6681 		if (hw->mac.ops.enable_tx_laser)
6682 			hw->mac.ops.enable_tx_laser(hw);
6683 
6684 		/* turn on all-multi mode if wake on multicast is enabled */
6685 		if (wufc & IXGBE_WUFC_MC) {
6686 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6687 			fctrl |= IXGBE_FCTRL_MPE;
6688 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6689 		}
6690 
6691 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6692 		ctrl |= IXGBE_CTRL_GIO_DIS;
6693 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6694 
6695 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6696 	} else {
6697 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6698 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6699 	}
6700 
6701 	switch (hw->mac.type) {
6702 	case ixgbe_mac_82598EB:
6703 		pci_wake_from_d3(pdev, false);
6704 		break;
6705 	case ixgbe_mac_82599EB:
6706 	case ixgbe_mac_X540:
6707 	case ixgbe_mac_X550:
6708 	case ixgbe_mac_X550EM_x:
6709 	case ixgbe_mac_x550em_a:
6710 		pci_wake_from_d3(pdev, !!wufc);
6711 		break;
6712 	default:
6713 		break;
6714 	}
6715 
6716 	*enable_wake = !!wufc;
6717 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6718 		hw->phy.ops.set_phy_power(hw, false);
6719 
6720 	ixgbe_release_hw_control(adapter);
6721 
6722 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6723 		pci_disable_device(pdev);
6724 
6725 	return 0;
6726 }
6727 
6728 #ifdef CONFIG_PM
6729 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6730 {
6731 	int retval;
6732 	bool wake;
6733 
6734 	retval = __ixgbe_shutdown(pdev, &wake);
6735 	if (retval)
6736 		return retval;
6737 
6738 	if (wake) {
6739 		pci_prepare_to_sleep(pdev);
6740 	} else {
6741 		pci_wake_from_d3(pdev, false);
6742 		pci_set_power_state(pdev, PCI_D3hot);
6743 	}
6744 
6745 	return 0;
6746 }
6747 #endif /* CONFIG_PM */
6748 
6749 static void ixgbe_shutdown(struct pci_dev *pdev)
6750 {
6751 	bool wake;
6752 
6753 	__ixgbe_shutdown(pdev, &wake);
6754 
6755 	if (system_state == SYSTEM_POWER_OFF) {
6756 		pci_wake_from_d3(pdev, wake);
6757 		pci_set_power_state(pdev, PCI_D3hot);
6758 	}
6759 }
6760 
6761 /**
6762  * ixgbe_update_stats - Update the board statistics counters.
6763  * @adapter: board private structure
6764  **/
6765 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6766 {
6767 	struct net_device *netdev = adapter->netdev;
6768 	struct ixgbe_hw *hw = &adapter->hw;
6769 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6770 	u64 total_mpc = 0;
6771 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6772 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6773 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6774 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6775 
6776 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6777 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6778 		return;
6779 
6780 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6781 		u64 rsc_count = 0;
6782 		u64 rsc_flush = 0;
6783 		for (i = 0; i < adapter->num_rx_queues; i++) {
6784 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6785 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6786 		}
6787 		adapter->rsc_total_count = rsc_count;
6788 		adapter->rsc_total_flush = rsc_flush;
6789 	}
6790 
6791 	for (i = 0; i < adapter->num_rx_queues; i++) {
6792 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6793 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6794 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6795 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6796 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6797 		bytes += rx_ring->stats.bytes;
6798 		packets += rx_ring->stats.packets;
6799 	}
6800 	adapter->non_eop_descs = non_eop_descs;
6801 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6802 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6803 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6804 	netdev->stats.rx_bytes = bytes;
6805 	netdev->stats.rx_packets = packets;
6806 
6807 	bytes = 0;
6808 	packets = 0;
6809 	/* gather some stats to the adapter struct that are per queue */
6810 	for (i = 0; i < adapter->num_tx_queues; i++) {
6811 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6812 		restart_queue += tx_ring->tx_stats.restart_queue;
6813 		tx_busy += tx_ring->tx_stats.tx_busy;
6814 		bytes += tx_ring->stats.bytes;
6815 		packets += tx_ring->stats.packets;
6816 	}
6817 	for (i = 0; i < adapter->num_xdp_queues; i++) {
6818 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
6819 
6820 		restart_queue += xdp_ring->tx_stats.restart_queue;
6821 		tx_busy += xdp_ring->tx_stats.tx_busy;
6822 		bytes += xdp_ring->stats.bytes;
6823 		packets += xdp_ring->stats.packets;
6824 	}
6825 	adapter->restart_queue = restart_queue;
6826 	adapter->tx_busy = tx_busy;
6827 	netdev->stats.tx_bytes = bytes;
6828 	netdev->stats.tx_packets = packets;
6829 
6830 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6831 
6832 	/* 8 register reads */
6833 	for (i = 0; i < 8; i++) {
6834 		/* for packet buffers not used, the register should read 0 */
6835 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6836 		missed_rx += mpc;
6837 		hwstats->mpc[i] += mpc;
6838 		total_mpc += hwstats->mpc[i];
6839 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6840 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6841 		switch (hw->mac.type) {
6842 		case ixgbe_mac_82598EB:
6843 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6844 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6845 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6846 			hwstats->pxonrxc[i] +=
6847 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6848 			break;
6849 		case ixgbe_mac_82599EB:
6850 		case ixgbe_mac_X540:
6851 		case ixgbe_mac_X550:
6852 		case ixgbe_mac_X550EM_x:
6853 		case ixgbe_mac_x550em_a:
6854 			hwstats->pxonrxc[i] +=
6855 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6856 			break;
6857 		default:
6858 			break;
6859 		}
6860 	}
6861 
6862 	/*16 register reads */
6863 	for (i = 0; i < 16; i++) {
6864 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6865 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6866 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6867 		    (hw->mac.type == ixgbe_mac_X540) ||
6868 		    (hw->mac.type == ixgbe_mac_X550) ||
6869 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6870 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6871 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6872 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6873 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6874 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6875 		}
6876 	}
6877 
6878 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6879 	/* work around hardware counting issue */
6880 	hwstats->gprc -= missed_rx;
6881 
6882 	ixgbe_update_xoff_received(adapter);
6883 
6884 	/* 82598 hardware only has a 32 bit counter in the high register */
6885 	switch (hw->mac.type) {
6886 	case ixgbe_mac_82598EB:
6887 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6888 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6889 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6890 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6891 		break;
6892 	case ixgbe_mac_X540:
6893 	case ixgbe_mac_X550:
6894 	case ixgbe_mac_X550EM_x:
6895 	case ixgbe_mac_x550em_a:
6896 		/* OS2BMC stats are X540 and later */
6897 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6898 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6899 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6900 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6901 		/* fall through */
6902 	case ixgbe_mac_82599EB:
6903 		for (i = 0; i < 16; i++)
6904 			adapter->hw_rx_no_dma_resources +=
6905 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6906 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6907 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6908 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6909 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6910 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6911 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6912 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6913 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6914 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6915 #ifdef IXGBE_FCOE
6916 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6917 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6918 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6919 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6920 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6921 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6922 		/* Add up per cpu counters for total ddp aloc fail */
6923 		if (adapter->fcoe.ddp_pool) {
6924 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6925 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6926 			unsigned int cpu;
6927 			u64 noddp = 0, noddp_ext_buff = 0;
6928 			for_each_possible_cpu(cpu) {
6929 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6930 				noddp += ddp_pool->noddp;
6931 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6932 			}
6933 			hwstats->fcoe_noddp = noddp;
6934 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6935 		}
6936 #endif /* IXGBE_FCOE */
6937 		break;
6938 	default:
6939 		break;
6940 	}
6941 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6942 	hwstats->bprc += bprc;
6943 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6944 	if (hw->mac.type == ixgbe_mac_82598EB)
6945 		hwstats->mprc -= bprc;
6946 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6947 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6948 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6949 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6950 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6951 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6952 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6953 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6954 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6955 	hwstats->lxontxc += lxon;
6956 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6957 	hwstats->lxofftxc += lxoff;
6958 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6959 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6960 	/*
6961 	 * 82598 errata - tx of flow control packets is included in tx counters
6962 	 */
6963 	xon_off_tot = lxon + lxoff;
6964 	hwstats->gptc -= xon_off_tot;
6965 	hwstats->mptc -= xon_off_tot;
6966 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6967 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6968 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6969 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6970 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6971 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6972 	hwstats->ptc64 -= xon_off_tot;
6973 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6974 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6975 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6976 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6977 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6978 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6979 
6980 	/* Fill out the OS statistics structure */
6981 	netdev->stats.multicast = hwstats->mprc;
6982 
6983 	/* Rx Errors */
6984 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6985 	netdev->stats.rx_dropped = 0;
6986 	netdev->stats.rx_length_errors = hwstats->rlec;
6987 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6988 	netdev->stats.rx_missed_errors = total_mpc;
6989 }
6990 
6991 /**
6992  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6993  * @adapter: pointer to the device adapter structure
6994  **/
6995 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6996 {
6997 	struct ixgbe_hw *hw = &adapter->hw;
6998 	int i;
6999 
7000 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7001 		return;
7002 
7003 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7004 
7005 	/* if interface is down do nothing */
7006 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7007 		return;
7008 
7009 	/* do nothing if we are not using signature filters */
7010 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7011 		return;
7012 
7013 	adapter->fdir_overflow++;
7014 
7015 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7016 		for (i = 0; i < adapter->num_tx_queues; i++)
7017 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7018 				&(adapter->tx_ring[i]->state));
7019 		for (i = 0; i < adapter->num_xdp_queues; i++)
7020 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7021 				&adapter->xdp_ring[i]->state);
7022 		/* re-enable flow director interrupts */
7023 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7024 	} else {
7025 		e_err(probe, "failed to finish FDIR re-initialization, "
7026 		      "ignored adding FDIR ATR filters\n");
7027 	}
7028 }
7029 
7030 /**
7031  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7032  * @adapter: pointer to the device adapter structure
7033  *
7034  * This function serves two purposes.  First it strobes the interrupt lines
7035  * in order to make certain interrupts are occurring.  Secondly it sets the
7036  * bits needed to check for TX hangs.  As a result we should immediately
7037  * determine if a hang has occurred.
7038  */
7039 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7040 {
7041 	struct ixgbe_hw *hw = &adapter->hw;
7042 	u64 eics = 0;
7043 	int i;
7044 
7045 	/* If we're down, removing or resetting, just bail */
7046 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7047 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7048 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7049 		return;
7050 
7051 	/* Force detection of hung controller */
7052 	if (netif_carrier_ok(adapter->netdev)) {
7053 		for (i = 0; i < adapter->num_tx_queues; i++)
7054 			set_check_for_tx_hang(adapter->tx_ring[i]);
7055 		for (i = 0; i < adapter->num_xdp_queues; i++)
7056 			set_check_for_tx_hang(adapter->xdp_ring[i]);
7057 	}
7058 
7059 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7060 		/*
7061 		 * for legacy and MSI interrupts don't set any bits
7062 		 * that are enabled for EIAM, because this operation
7063 		 * would set *both* EIMS and EICS for any bit in EIAM
7064 		 */
7065 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
7066 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7067 	} else {
7068 		/* get one bit for every active tx/rx interrupt vector */
7069 		for (i = 0; i < adapter->num_q_vectors; i++) {
7070 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
7071 			if (qv->rx.ring || qv->tx.ring)
7072 				eics |= BIT_ULL(i);
7073 		}
7074 	}
7075 
7076 	/* Cause software interrupt to ensure rings are cleaned */
7077 	ixgbe_irq_rearm_queues(adapter, eics);
7078 }
7079 
7080 /**
7081  * ixgbe_watchdog_update_link - update the link status
7082  * @adapter: pointer to the device adapter structure
7083  * @link_speed: pointer to a u32 to store the link_speed
7084  **/
7085 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7086 {
7087 	struct ixgbe_hw *hw = &adapter->hw;
7088 	u32 link_speed = adapter->link_speed;
7089 	bool link_up = adapter->link_up;
7090 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7091 
7092 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7093 		return;
7094 
7095 	if (hw->mac.ops.check_link) {
7096 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7097 	} else {
7098 		/* always assume link is up, if no check link function */
7099 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7100 		link_up = true;
7101 	}
7102 
7103 	if (adapter->ixgbe_ieee_pfc)
7104 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7105 
7106 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7107 		hw->mac.ops.fc_enable(hw);
7108 		ixgbe_set_rx_drop_en(adapter);
7109 	}
7110 
7111 	if (link_up ||
7112 	    time_after(jiffies, (adapter->link_check_timeout +
7113 				 IXGBE_TRY_LINK_TIMEOUT))) {
7114 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7115 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7116 		IXGBE_WRITE_FLUSH(hw);
7117 	}
7118 
7119 	adapter->link_up = link_up;
7120 	adapter->link_speed = link_speed;
7121 }
7122 
7123 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7124 {
7125 #ifdef CONFIG_IXGBE_DCB
7126 	struct net_device *netdev = adapter->netdev;
7127 	struct dcb_app app = {
7128 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7129 			      .protocol = 0,
7130 			     };
7131 	u8 up = 0;
7132 
7133 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7134 		up = dcb_ieee_getapp_mask(netdev, &app);
7135 
7136 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7137 #endif
7138 }
7139 
7140 static int ixgbe_enable_macvlan(struct net_device *upper, void *data)
7141 {
7142 	if (netif_is_macvlan(upper)) {
7143 		struct macvlan_dev *vlan = netdev_priv(upper);
7144 
7145 		if (vlan->fwd_priv)
7146 			netif_tx_wake_all_queues(upper);
7147 	}
7148 
7149 	return 0;
7150 }
7151 
7152 /**
7153  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7154  *                             print link up message
7155  * @adapter: pointer to the device adapter structure
7156  **/
7157 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7158 {
7159 	struct net_device *netdev = adapter->netdev;
7160 	struct ixgbe_hw *hw = &adapter->hw;
7161 	u32 link_speed = adapter->link_speed;
7162 	const char *speed_str;
7163 	bool flow_rx, flow_tx;
7164 
7165 	/* only continue if link was previously down */
7166 	if (netif_carrier_ok(netdev))
7167 		return;
7168 
7169 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7170 
7171 	switch (hw->mac.type) {
7172 	case ixgbe_mac_82598EB: {
7173 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7174 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7175 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7176 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7177 	}
7178 		break;
7179 	case ixgbe_mac_X540:
7180 	case ixgbe_mac_X550:
7181 	case ixgbe_mac_X550EM_x:
7182 	case ixgbe_mac_x550em_a:
7183 	case ixgbe_mac_82599EB: {
7184 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7185 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7186 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7187 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7188 	}
7189 		break;
7190 	default:
7191 		flow_tx = false;
7192 		flow_rx = false;
7193 		break;
7194 	}
7195 
7196 	adapter->last_rx_ptp_check = jiffies;
7197 
7198 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7199 		ixgbe_ptp_start_cyclecounter(adapter);
7200 
7201 	switch (link_speed) {
7202 	case IXGBE_LINK_SPEED_10GB_FULL:
7203 		speed_str = "10 Gbps";
7204 		break;
7205 	case IXGBE_LINK_SPEED_2_5GB_FULL:
7206 		speed_str = "2.5 Gbps";
7207 		break;
7208 	case IXGBE_LINK_SPEED_1GB_FULL:
7209 		speed_str = "1 Gbps";
7210 		break;
7211 	case IXGBE_LINK_SPEED_100_FULL:
7212 		speed_str = "100 Mbps";
7213 		break;
7214 	case IXGBE_LINK_SPEED_10_FULL:
7215 		speed_str = "10 Mbps";
7216 		break;
7217 	default:
7218 		speed_str = "unknown speed";
7219 		break;
7220 	}
7221 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7222 	       ((flow_rx && flow_tx) ? "RX/TX" :
7223 	       (flow_rx ? "RX" :
7224 	       (flow_tx ? "TX" : "None"))));
7225 
7226 	netif_carrier_on(netdev);
7227 	ixgbe_check_vf_rate_limit(adapter);
7228 
7229 	/* enable transmits */
7230 	netif_tx_wake_all_queues(adapter->netdev);
7231 
7232 	/* enable any upper devices */
7233 	rtnl_lock();
7234 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
7235 				      ixgbe_enable_macvlan, NULL);
7236 	rtnl_unlock();
7237 
7238 	/* update the default user priority for VFs */
7239 	ixgbe_update_default_up(adapter);
7240 
7241 	/* ping all the active vfs to let them know link has changed */
7242 	ixgbe_ping_all_vfs(adapter);
7243 }
7244 
7245 /**
7246  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7247  *                               print link down message
7248  * @adapter: pointer to the adapter structure
7249  **/
7250 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7251 {
7252 	struct net_device *netdev = adapter->netdev;
7253 	struct ixgbe_hw *hw = &adapter->hw;
7254 
7255 	adapter->link_up = false;
7256 	adapter->link_speed = 0;
7257 
7258 	/* only continue if link was up previously */
7259 	if (!netif_carrier_ok(netdev))
7260 		return;
7261 
7262 	/* poll for SFP+ cable when link is down */
7263 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7264 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7265 
7266 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7267 		ixgbe_ptp_start_cyclecounter(adapter);
7268 
7269 	e_info(drv, "NIC Link is Down\n");
7270 	netif_carrier_off(netdev);
7271 
7272 	/* ping all the active vfs to let them know link has changed */
7273 	ixgbe_ping_all_vfs(adapter);
7274 }
7275 
7276 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7277 {
7278 	int i;
7279 
7280 	for (i = 0; i < adapter->num_tx_queues; i++) {
7281 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7282 
7283 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
7284 			return true;
7285 	}
7286 
7287 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7288 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
7289 
7290 		if (ring->next_to_use != ring->next_to_clean)
7291 			return true;
7292 	}
7293 
7294 	return false;
7295 }
7296 
7297 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7298 {
7299 	struct ixgbe_hw *hw = &adapter->hw;
7300 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7301 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7302 
7303 	int i, j;
7304 
7305 	if (!adapter->num_vfs)
7306 		return false;
7307 
7308 	/* resetting the PF is only needed for MAC before X550 */
7309 	if (hw->mac.type >= ixgbe_mac_X550)
7310 		return false;
7311 
7312 	for (i = 0; i < adapter->num_vfs; i++) {
7313 		for (j = 0; j < q_per_pool; j++) {
7314 			u32 h, t;
7315 
7316 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7317 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7318 
7319 			if (h != t)
7320 				return true;
7321 		}
7322 	}
7323 
7324 	return false;
7325 }
7326 
7327 /**
7328  * ixgbe_watchdog_flush_tx - flush queues on link down
7329  * @adapter: pointer to the device adapter structure
7330  **/
7331 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7332 {
7333 	if (!netif_carrier_ok(adapter->netdev)) {
7334 		if (ixgbe_ring_tx_pending(adapter) ||
7335 		    ixgbe_vf_tx_pending(adapter)) {
7336 			/* We've lost link, so the controller stops DMA,
7337 			 * but we've got queued Tx work that's never going
7338 			 * to get done, so reset controller to flush Tx.
7339 			 * (Do the reset outside of interrupt context).
7340 			 */
7341 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7342 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7343 		}
7344 	}
7345 }
7346 
7347 #ifdef CONFIG_PCI_IOV
7348 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7349 {
7350 	struct ixgbe_hw *hw = &adapter->hw;
7351 	struct pci_dev *pdev = adapter->pdev;
7352 	unsigned int vf;
7353 	u32 gpc;
7354 
7355 	if (!(netif_carrier_ok(adapter->netdev)))
7356 		return;
7357 
7358 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7359 	if (gpc) /* If incrementing then no need for the check below */
7360 		return;
7361 	/* Check to see if a bad DMA write target from an errant or
7362 	 * malicious VF has caused a PCIe error.  If so then we can
7363 	 * issue a VFLR to the offending VF(s) and then resume without
7364 	 * requesting a full slot reset.
7365 	 */
7366 
7367 	if (!pdev)
7368 		return;
7369 
7370 	/* check status reg for all VFs owned by this PF */
7371 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
7372 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7373 		u16 status_reg;
7374 
7375 		if (!vfdev)
7376 			continue;
7377 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7378 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7379 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
7380 			pcie_flr(vfdev);
7381 	}
7382 }
7383 
7384 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7385 {
7386 	u32 ssvpc;
7387 
7388 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7389 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7390 	    adapter->num_vfs == 0)
7391 		return;
7392 
7393 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7394 
7395 	/*
7396 	 * ssvpc register is cleared on read, if zero then no
7397 	 * spoofed packets in the last interval.
7398 	 */
7399 	if (!ssvpc)
7400 		return;
7401 
7402 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7403 }
7404 #else
7405 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7406 {
7407 }
7408 
7409 static void
7410 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7411 {
7412 }
7413 #endif /* CONFIG_PCI_IOV */
7414 
7415 
7416 /**
7417  * ixgbe_watchdog_subtask - check and bring link up
7418  * @adapter: pointer to the device adapter structure
7419  **/
7420 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7421 {
7422 	/* if interface is down, removing or resetting, do nothing */
7423 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7424 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7425 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7426 		return;
7427 
7428 	ixgbe_watchdog_update_link(adapter);
7429 
7430 	if (adapter->link_up)
7431 		ixgbe_watchdog_link_is_up(adapter);
7432 	else
7433 		ixgbe_watchdog_link_is_down(adapter);
7434 
7435 	ixgbe_check_for_bad_vf(adapter);
7436 	ixgbe_spoof_check(adapter);
7437 	ixgbe_update_stats(adapter);
7438 
7439 	ixgbe_watchdog_flush_tx(adapter);
7440 }
7441 
7442 /**
7443  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7444  * @adapter: the ixgbe adapter structure
7445  **/
7446 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7447 {
7448 	struct ixgbe_hw *hw = &adapter->hw;
7449 	s32 err;
7450 
7451 	/* not searching for SFP so there is nothing to do here */
7452 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7453 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7454 		return;
7455 
7456 	if (adapter->sfp_poll_time &&
7457 	    time_after(adapter->sfp_poll_time, jiffies))
7458 		return; /* If not yet time to poll for SFP */
7459 
7460 	/* someone else is in init, wait until next service event */
7461 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7462 		return;
7463 
7464 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7465 
7466 	err = hw->phy.ops.identify_sfp(hw);
7467 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7468 		goto sfp_out;
7469 
7470 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7471 		/* If no cable is present, then we need to reset
7472 		 * the next time we find a good cable. */
7473 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7474 	}
7475 
7476 	/* exit on error */
7477 	if (err)
7478 		goto sfp_out;
7479 
7480 	/* exit if reset not needed */
7481 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7482 		goto sfp_out;
7483 
7484 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7485 
7486 	/*
7487 	 * A module may be identified correctly, but the EEPROM may not have
7488 	 * support for that module.  setup_sfp() will fail in that case, so
7489 	 * we should not allow that module to load.
7490 	 */
7491 	if (hw->mac.type == ixgbe_mac_82598EB)
7492 		err = hw->phy.ops.reset(hw);
7493 	else
7494 		err = hw->mac.ops.setup_sfp(hw);
7495 
7496 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7497 		goto sfp_out;
7498 
7499 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7500 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7501 
7502 sfp_out:
7503 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7504 
7505 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7506 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7507 		e_dev_err("failed to initialize because an unsupported "
7508 			  "SFP+ module type was detected.\n");
7509 		e_dev_err("Reload the driver after installing a "
7510 			  "supported module.\n");
7511 		unregister_netdev(adapter->netdev);
7512 	}
7513 }
7514 
7515 /**
7516  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7517  * @adapter: the ixgbe adapter structure
7518  **/
7519 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7520 {
7521 	struct ixgbe_hw *hw = &adapter->hw;
7522 	u32 speed;
7523 	bool autoneg = false;
7524 
7525 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7526 		return;
7527 
7528 	/* someone else is in init, wait until next service event */
7529 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7530 		return;
7531 
7532 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7533 
7534 	speed = hw->phy.autoneg_advertised;
7535 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
7536 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
7537 
7538 		/* setup the highest link when no autoneg */
7539 		if (!autoneg) {
7540 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
7541 				speed = IXGBE_LINK_SPEED_10GB_FULL;
7542 		}
7543 	}
7544 
7545 	if (hw->mac.ops.setup_link)
7546 		hw->mac.ops.setup_link(hw, speed, true);
7547 
7548 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7549 	adapter->link_check_timeout = jiffies;
7550 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7551 }
7552 
7553 /**
7554  * ixgbe_service_timer - Timer Call-back
7555  * @data: pointer to adapter cast into an unsigned long
7556  **/
7557 static void ixgbe_service_timer(unsigned long data)
7558 {
7559 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
7560 	unsigned long next_event_offset;
7561 
7562 	/* poll faster when waiting for link */
7563 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7564 		next_event_offset = HZ / 10;
7565 	else
7566 		next_event_offset = HZ * 2;
7567 
7568 	/* Reset the timer */
7569 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7570 
7571 	ixgbe_service_event_schedule(adapter);
7572 }
7573 
7574 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7575 {
7576 	struct ixgbe_hw *hw = &adapter->hw;
7577 	u32 status;
7578 
7579 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7580 		return;
7581 
7582 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7583 
7584 	if (!hw->phy.ops.handle_lasi)
7585 		return;
7586 
7587 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7588 	if (status != IXGBE_ERR_OVERTEMP)
7589 		return;
7590 
7591 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7592 }
7593 
7594 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7595 {
7596 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7597 		return;
7598 
7599 	/* If we're already down, removing or resetting, just bail */
7600 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7601 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7602 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7603 		return;
7604 
7605 	ixgbe_dump(adapter);
7606 	netdev_err(adapter->netdev, "Reset adapter\n");
7607 	adapter->tx_timeout_count++;
7608 
7609 	rtnl_lock();
7610 	ixgbe_reinit_locked(adapter);
7611 	rtnl_unlock();
7612 }
7613 
7614 /**
7615  * ixgbe_service_task - manages and runs subtasks
7616  * @work: pointer to work_struct containing our data
7617  **/
7618 static void ixgbe_service_task(struct work_struct *work)
7619 {
7620 	struct ixgbe_adapter *adapter = container_of(work,
7621 						     struct ixgbe_adapter,
7622 						     service_task);
7623 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7624 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7625 			rtnl_lock();
7626 			ixgbe_down(adapter);
7627 			rtnl_unlock();
7628 		}
7629 		ixgbe_service_event_complete(adapter);
7630 		return;
7631 	}
7632 	if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7633 		rtnl_lock();
7634 		adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7635 		udp_tunnel_get_rx_info(adapter->netdev);
7636 		rtnl_unlock();
7637 	}
7638 	ixgbe_reset_subtask(adapter);
7639 	ixgbe_phy_interrupt_subtask(adapter);
7640 	ixgbe_sfp_detection_subtask(adapter);
7641 	ixgbe_sfp_link_config_subtask(adapter);
7642 	ixgbe_check_overtemp_subtask(adapter);
7643 	ixgbe_watchdog_subtask(adapter);
7644 	ixgbe_fdir_reinit_subtask(adapter);
7645 	ixgbe_check_hang_subtask(adapter);
7646 
7647 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7648 		ixgbe_ptp_overflow_check(adapter);
7649 		ixgbe_ptp_rx_hang(adapter);
7650 		ixgbe_ptp_tx_hang(adapter);
7651 	}
7652 
7653 	ixgbe_service_event_complete(adapter);
7654 }
7655 
7656 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7657 		     struct ixgbe_tx_buffer *first,
7658 		     u8 *hdr_len)
7659 {
7660 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7661 	struct sk_buff *skb = first->skb;
7662 	union {
7663 		struct iphdr *v4;
7664 		struct ipv6hdr *v6;
7665 		unsigned char *hdr;
7666 	} ip;
7667 	union {
7668 		struct tcphdr *tcp;
7669 		unsigned char *hdr;
7670 	} l4;
7671 	u32 paylen, l4_offset;
7672 	int err;
7673 
7674 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7675 		return 0;
7676 
7677 	if (!skb_is_gso(skb))
7678 		return 0;
7679 
7680 	err = skb_cow_head(skb, 0);
7681 	if (err < 0)
7682 		return err;
7683 
7684 	if (eth_p_mpls(first->protocol))
7685 		ip.hdr = skb_inner_network_header(skb);
7686 	else
7687 		ip.hdr = skb_network_header(skb);
7688 	l4.hdr = skb_checksum_start(skb);
7689 
7690 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7691 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7692 
7693 	/* initialize outer IP header fields */
7694 	if (ip.v4->version == 4) {
7695 		unsigned char *csum_start = skb_checksum_start(skb);
7696 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
7697 
7698 		/* IP header will have to cancel out any data that
7699 		 * is not a part of the outer IP header
7700 		 */
7701 		ip.v4->check = csum_fold(csum_partial(trans_start,
7702 						      csum_start - trans_start,
7703 						      0));
7704 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7705 
7706 		ip.v4->tot_len = 0;
7707 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7708 				   IXGBE_TX_FLAGS_CSUM |
7709 				   IXGBE_TX_FLAGS_IPV4;
7710 	} else {
7711 		ip.v6->payload_len = 0;
7712 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7713 				   IXGBE_TX_FLAGS_CSUM;
7714 	}
7715 
7716 	/* determine offset of inner transport header */
7717 	l4_offset = l4.hdr - skb->data;
7718 
7719 	/* compute length of segmentation header */
7720 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7721 
7722 	/* remove payload length from inner checksum */
7723 	paylen = skb->len - l4_offset;
7724 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
7725 
7726 	/* update gso size and bytecount with header size */
7727 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7728 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7729 
7730 	/* mss_l4len_id: use 0 as index for TSO */
7731 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7732 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7733 
7734 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7735 	vlan_macip_lens = l4.hdr - ip.hdr;
7736 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7737 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7738 
7739 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
7740 			  mss_l4len_idx);
7741 
7742 	return 1;
7743 }
7744 
7745 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7746 {
7747 	unsigned int offset = 0;
7748 
7749 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7750 
7751 	return offset == skb_checksum_start_offset(skb);
7752 }
7753 
7754 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7755 			  struct ixgbe_tx_buffer *first)
7756 {
7757 	struct sk_buff *skb = first->skb;
7758 	u32 vlan_macip_lens = 0;
7759 	u32 type_tucmd = 0;
7760 
7761 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7762 csum_failed:
7763 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7764 					 IXGBE_TX_FLAGS_CC)))
7765 			return;
7766 		goto no_csum;
7767 	}
7768 
7769 	switch (skb->csum_offset) {
7770 	case offsetof(struct tcphdr, check):
7771 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7772 		/* fall through */
7773 	case offsetof(struct udphdr, check):
7774 		break;
7775 	case offsetof(struct sctphdr, checksum):
7776 		/* validate that this is actually an SCTP request */
7777 		if (((first->protocol == htons(ETH_P_IP)) &&
7778 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7779 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7780 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7781 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7782 			break;
7783 		}
7784 		/* fall through */
7785 	default:
7786 		skb_checksum_help(skb);
7787 		goto csum_failed;
7788 	}
7789 
7790 	/* update TX checksum flag */
7791 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7792 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7793 			  skb_network_offset(skb);
7794 no_csum:
7795 	/* vlan_macip_lens: MACLEN, VLAN tag */
7796 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7797 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7798 
7799 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd, 0);
7800 }
7801 
7802 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7803 	((_flag <= _result) ? \
7804 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7805 	 ((u32)(_input & _flag) / (_flag / _result)))
7806 
7807 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7808 {
7809 	/* set type for advanced descriptor with frame checksum insertion */
7810 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7811 		       IXGBE_ADVTXD_DCMD_DEXT |
7812 		       IXGBE_ADVTXD_DCMD_IFCS;
7813 
7814 	/* set HW vlan bit if vlan is present */
7815 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7816 				   IXGBE_ADVTXD_DCMD_VLE);
7817 
7818 	/* set segmentation enable bits for TSO/FSO */
7819 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7820 				   IXGBE_ADVTXD_DCMD_TSE);
7821 
7822 	/* set timestamp bit if present */
7823 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7824 				   IXGBE_ADVTXD_MAC_TSTAMP);
7825 
7826 	/* insert frame checksum */
7827 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7828 
7829 	return cmd_type;
7830 }
7831 
7832 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7833 				   u32 tx_flags, unsigned int paylen)
7834 {
7835 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7836 
7837 	/* enable L4 checksum for TSO and TX checksum offload */
7838 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7839 					IXGBE_TX_FLAGS_CSUM,
7840 					IXGBE_ADVTXD_POPTS_TXSM);
7841 
7842 	/* enble IPv4 checksum for TSO */
7843 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7844 					IXGBE_TX_FLAGS_IPV4,
7845 					IXGBE_ADVTXD_POPTS_IXSM);
7846 
7847 	/*
7848 	 * Check Context must be set if Tx switch is enabled, which it
7849 	 * always is for case where virtual functions are running
7850 	 */
7851 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7852 					IXGBE_TX_FLAGS_CC,
7853 					IXGBE_ADVTXD_CC);
7854 
7855 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7856 }
7857 
7858 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7859 {
7860 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7861 
7862 	/* Herbert's original patch had:
7863 	 *  smp_mb__after_netif_stop_queue();
7864 	 * but since that doesn't exist yet, just open code it.
7865 	 */
7866 	smp_mb();
7867 
7868 	/* We need to check again in a case another CPU has just
7869 	 * made room available.
7870 	 */
7871 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7872 		return -EBUSY;
7873 
7874 	/* A reprieve! - use start_queue because it doesn't call schedule */
7875 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7876 	++tx_ring->tx_stats.restart_queue;
7877 	return 0;
7878 }
7879 
7880 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7881 {
7882 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7883 		return 0;
7884 
7885 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7886 }
7887 
7888 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7889 		       IXGBE_TXD_CMD_RS)
7890 
7891 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7892 			struct ixgbe_tx_buffer *first,
7893 			const u8 hdr_len)
7894 {
7895 	struct sk_buff *skb = first->skb;
7896 	struct ixgbe_tx_buffer *tx_buffer;
7897 	union ixgbe_adv_tx_desc *tx_desc;
7898 	struct skb_frag_struct *frag;
7899 	dma_addr_t dma;
7900 	unsigned int data_len, size;
7901 	u32 tx_flags = first->tx_flags;
7902 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7903 	u16 i = tx_ring->next_to_use;
7904 
7905 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7906 
7907 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7908 
7909 	size = skb_headlen(skb);
7910 	data_len = skb->data_len;
7911 
7912 #ifdef IXGBE_FCOE
7913 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7914 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7915 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7916 			data_len = 0;
7917 		} else {
7918 			data_len -= sizeof(struct fcoe_crc_eof);
7919 		}
7920 	}
7921 
7922 #endif
7923 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7924 
7925 	tx_buffer = first;
7926 
7927 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7928 		if (dma_mapping_error(tx_ring->dev, dma))
7929 			goto dma_error;
7930 
7931 		/* record length, and DMA address */
7932 		dma_unmap_len_set(tx_buffer, len, size);
7933 		dma_unmap_addr_set(tx_buffer, dma, dma);
7934 
7935 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7936 
7937 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7938 			tx_desc->read.cmd_type_len =
7939 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7940 
7941 			i++;
7942 			tx_desc++;
7943 			if (i == tx_ring->count) {
7944 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7945 				i = 0;
7946 			}
7947 			tx_desc->read.olinfo_status = 0;
7948 
7949 			dma += IXGBE_MAX_DATA_PER_TXD;
7950 			size -= IXGBE_MAX_DATA_PER_TXD;
7951 
7952 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7953 		}
7954 
7955 		if (likely(!data_len))
7956 			break;
7957 
7958 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7959 
7960 		i++;
7961 		tx_desc++;
7962 		if (i == tx_ring->count) {
7963 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7964 			i = 0;
7965 		}
7966 		tx_desc->read.olinfo_status = 0;
7967 
7968 #ifdef IXGBE_FCOE
7969 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7970 #else
7971 		size = skb_frag_size(frag);
7972 #endif
7973 		data_len -= size;
7974 
7975 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7976 				       DMA_TO_DEVICE);
7977 
7978 		tx_buffer = &tx_ring->tx_buffer_info[i];
7979 	}
7980 
7981 	/* write last descriptor with RS and EOP bits */
7982 	cmd_type |= size | IXGBE_TXD_CMD;
7983 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7984 
7985 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7986 
7987 	/* set the timestamp */
7988 	first->time_stamp = jiffies;
7989 
7990 	/*
7991 	 * Force memory writes to complete before letting h/w know there
7992 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7993 	 * memory model archs, such as IA-64).
7994 	 *
7995 	 * We also need this memory barrier to make certain all of the
7996 	 * status bits have been updated before next_to_watch is written.
7997 	 */
7998 	wmb();
7999 
8000 	/* set next_to_watch value indicating a packet is present */
8001 	first->next_to_watch = tx_desc;
8002 
8003 	i++;
8004 	if (i == tx_ring->count)
8005 		i = 0;
8006 
8007 	tx_ring->next_to_use = i;
8008 
8009 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8010 
8011 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
8012 		writel(i, tx_ring->tail);
8013 
8014 		/* we need this if more than one processor can write to our tail
8015 		 * at a time, it synchronizes IO on IA64/Altix systems
8016 		 */
8017 		mmiowb();
8018 	}
8019 
8020 	return 0;
8021 dma_error:
8022 	dev_err(tx_ring->dev, "TX DMA map failed\n");
8023 	tx_buffer = &tx_ring->tx_buffer_info[i];
8024 
8025 	/* clear dma mappings for failed tx_buffer_info map */
8026 	while (tx_buffer != first) {
8027 		if (dma_unmap_len(tx_buffer, len))
8028 			dma_unmap_page(tx_ring->dev,
8029 				       dma_unmap_addr(tx_buffer, dma),
8030 				       dma_unmap_len(tx_buffer, len),
8031 				       DMA_TO_DEVICE);
8032 		dma_unmap_len_set(tx_buffer, len, 0);
8033 
8034 		if (i--)
8035 			i += tx_ring->count;
8036 		tx_buffer = &tx_ring->tx_buffer_info[i];
8037 	}
8038 
8039 	if (dma_unmap_len(tx_buffer, len))
8040 		dma_unmap_single(tx_ring->dev,
8041 				 dma_unmap_addr(tx_buffer, dma),
8042 				 dma_unmap_len(tx_buffer, len),
8043 				 DMA_TO_DEVICE);
8044 	dma_unmap_len_set(tx_buffer, len, 0);
8045 
8046 	dev_kfree_skb_any(first->skb);
8047 	first->skb = NULL;
8048 
8049 	tx_ring->next_to_use = i;
8050 
8051 	return -1;
8052 }
8053 
8054 static void ixgbe_atr(struct ixgbe_ring *ring,
8055 		      struct ixgbe_tx_buffer *first)
8056 {
8057 	struct ixgbe_q_vector *q_vector = ring->q_vector;
8058 	union ixgbe_atr_hash_dword input = { .dword = 0 };
8059 	union ixgbe_atr_hash_dword common = { .dword = 0 };
8060 	union {
8061 		unsigned char *network;
8062 		struct iphdr *ipv4;
8063 		struct ipv6hdr *ipv6;
8064 	} hdr;
8065 	struct tcphdr *th;
8066 	unsigned int hlen;
8067 	struct sk_buff *skb;
8068 	__be16 vlan_id;
8069 	int l4_proto;
8070 
8071 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
8072 	if (!q_vector)
8073 		return;
8074 
8075 	/* do nothing if sampling is disabled */
8076 	if (!ring->atr_sample_rate)
8077 		return;
8078 
8079 	ring->atr_count++;
8080 
8081 	/* currently only IPv4/IPv6 with TCP is supported */
8082 	if ((first->protocol != htons(ETH_P_IP)) &&
8083 	    (first->protocol != htons(ETH_P_IPV6)))
8084 		return;
8085 
8086 	/* snag network header to get L4 type and address */
8087 	skb = first->skb;
8088 	hdr.network = skb_network_header(skb);
8089 	if (unlikely(hdr.network <= skb->data))
8090 		return;
8091 	if (skb->encapsulation &&
8092 	    first->protocol == htons(ETH_P_IP) &&
8093 	    hdr.ipv4->protocol == IPPROTO_UDP) {
8094 		struct ixgbe_adapter *adapter = q_vector->adapter;
8095 
8096 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
8097 			     VXLAN_HEADROOM))
8098 			return;
8099 
8100 		/* verify the port is recognized as VXLAN */
8101 		if (adapter->vxlan_port &&
8102 		    udp_hdr(skb)->dest == adapter->vxlan_port)
8103 			hdr.network = skb_inner_network_header(skb);
8104 
8105 		if (adapter->geneve_port &&
8106 		    udp_hdr(skb)->dest == adapter->geneve_port)
8107 			hdr.network = skb_inner_network_header(skb);
8108 	}
8109 
8110 	/* Make sure we have at least [minimum IPv4 header + TCP]
8111 	 * or [IPv6 header] bytes
8112 	 */
8113 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8114 		return;
8115 
8116 	/* Currently only IPv4/IPv6 with TCP is supported */
8117 	switch (hdr.ipv4->version) {
8118 	case IPVERSION:
8119 		/* access ihl as u8 to avoid unaligned access on ia64 */
8120 		hlen = (hdr.network[0] & 0x0F) << 2;
8121 		l4_proto = hdr.ipv4->protocol;
8122 		break;
8123 	case 6:
8124 		hlen = hdr.network - skb->data;
8125 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8126 		hlen -= hdr.network - skb->data;
8127 		break;
8128 	default:
8129 		return;
8130 	}
8131 
8132 	if (l4_proto != IPPROTO_TCP)
8133 		return;
8134 
8135 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
8136 		     hlen + sizeof(struct tcphdr)))
8137 		return;
8138 
8139 	th = (struct tcphdr *)(hdr.network + hlen);
8140 
8141 	/* skip this packet since the socket is closing */
8142 	if (th->fin)
8143 		return;
8144 
8145 	/* sample on all syn packets or once every atr sample count */
8146 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8147 		return;
8148 
8149 	/* reset sample count */
8150 	ring->atr_count = 0;
8151 
8152 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8153 
8154 	/*
8155 	 * src and dst are inverted, think how the receiver sees them
8156 	 *
8157 	 * The input is broken into two sections, a non-compressed section
8158 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8159 	 * is XORed together and stored in the compressed dword.
8160 	 */
8161 	input.formatted.vlan_id = vlan_id;
8162 
8163 	/*
8164 	 * since src port and flex bytes occupy the same word XOR them together
8165 	 * and write the value to source port portion of compressed dword
8166 	 */
8167 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8168 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8169 	else
8170 		common.port.src ^= th->dest ^ first->protocol;
8171 	common.port.dst ^= th->source;
8172 
8173 	switch (hdr.ipv4->version) {
8174 	case IPVERSION:
8175 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8176 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8177 		break;
8178 	case 6:
8179 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8180 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8181 			     hdr.ipv6->saddr.s6_addr32[1] ^
8182 			     hdr.ipv6->saddr.s6_addr32[2] ^
8183 			     hdr.ipv6->saddr.s6_addr32[3] ^
8184 			     hdr.ipv6->daddr.s6_addr32[0] ^
8185 			     hdr.ipv6->daddr.s6_addr32[1] ^
8186 			     hdr.ipv6->daddr.s6_addr32[2] ^
8187 			     hdr.ipv6->daddr.s6_addr32[3];
8188 		break;
8189 	default:
8190 		break;
8191 	}
8192 
8193 	if (hdr.network != skb_network_header(skb))
8194 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8195 
8196 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
8197 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8198 					      input, common, ring->queue_index);
8199 }
8200 
8201 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8202 			      void *accel_priv, select_queue_fallback_t fallback)
8203 {
8204 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
8205 #ifdef IXGBE_FCOE
8206 	struct ixgbe_adapter *adapter;
8207 	struct ixgbe_ring_feature *f;
8208 	int txq;
8209 #endif
8210 
8211 	if (fwd_adapter)
8212 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
8213 
8214 #ifdef IXGBE_FCOE
8215 
8216 	/*
8217 	 * only execute the code below if protocol is FCoE
8218 	 * or FIP and we have FCoE enabled on the adapter
8219 	 */
8220 	switch (vlan_get_protocol(skb)) {
8221 	case htons(ETH_P_FCOE):
8222 	case htons(ETH_P_FIP):
8223 		adapter = netdev_priv(dev);
8224 
8225 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
8226 			break;
8227 		/* fall through */
8228 	default:
8229 		return fallback(dev, skb);
8230 	}
8231 
8232 	f = &adapter->ring_feature[RING_F_FCOE];
8233 
8234 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8235 					   smp_processor_id();
8236 
8237 	while (txq >= f->indices)
8238 		txq -= f->indices;
8239 
8240 	return txq + f->offset;
8241 #else
8242 	return fallback(dev, skb);
8243 #endif
8244 }
8245 
8246 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
8247 			       struct xdp_buff *xdp)
8248 {
8249 	struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
8250 	struct ixgbe_tx_buffer *tx_buffer;
8251 	union ixgbe_adv_tx_desc *tx_desc;
8252 	u32 len, cmd_type;
8253 	dma_addr_t dma;
8254 	u16 i;
8255 
8256 	len = xdp->data_end - xdp->data;
8257 
8258 	if (unlikely(!ixgbe_desc_unused(ring)))
8259 		return IXGBE_XDP_CONSUMED;
8260 
8261 	dma = dma_map_single(ring->dev, xdp->data, len, DMA_TO_DEVICE);
8262 	if (dma_mapping_error(ring->dev, dma))
8263 		return IXGBE_XDP_CONSUMED;
8264 
8265 	/* record the location of the first descriptor for this packet */
8266 	tx_buffer = &ring->tx_buffer_info[ring->next_to_use];
8267 	tx_buffer->bytecount = len;
8268 	tx_buffer->gso_segs = 1;
8269 	tx_buffer->protocol = 0;
8270 
8271 	i = ring->next_to_use;
8272 	tx_desc = IXGBE_TX_DESC(ring, i);
8273 
8274 	dma_unmap_len_set(tx_buffer, len, len);
8275 	dma_unmap_addr_set(tx_buffer, dma, dma);
8276 	tx_buffer->data = xdp->data;
8277 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
8278 
8279 	/* put descriptor type bits */
8280 	cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8281 		   IXGBE_ADVTXD_DCMD_DEXT |
8282 		   IXGBE_ADVTXD_DCMD_IFCS;
8283 	cmd_type |= len | IXGBE_TXD_CMD;
8284 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8285 	tx_desc->read.olinfo_status =
8286 		cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
8287 
8288 	/* Avoid any potential race with xdp_xmit and cleanup */
8289 	smp_wmb();
8290 
8291 	/* set next_to_watch value indicating a packet is present */
8292 	i++;
8293 	if (i == ring->count)
8294 		i = 0;
8295 
8296 	tx_buffer->next_to_watch = tx_desc;
8297 	ring->next_to_use = i;
8298 
8299 	return IXGBE_XDP_TX;
8300 }
8301 
8302 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8303 			  struct ixgbe_adapter *adapter,
8304 			  struct ixgbe_ring *tx_ring)
8305 {
8306 	struct ixgbe_tx_buffer *first;
8307 	int tso;
8308 	u32 tx_flags = 0;
8309 	unsigned short f;
8310 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
8311 	__be16 protocol = skb->protocol;
8312 	u8 hdr_len = 0;
8313 
8314 	/*
8315 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8316 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8317 	 *       + 2 desc gap to keep tail from touching head,
8318 	 *       + 1 desc for context descriptor,
8319 	 * otherwise try next time
8320 	 */
8321 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8322 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
8323 
8324 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8325 		tx_ring->tx_stats.tx_busy++;
8326 		return NETDEV_TX_BUSY;
8327 	}
8328 
8329 	/* record the location of the first descriptor for this packet */
8330 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8331 	first->skb = skb;
8332 	first->bytecount = skb->len;
8333 	first->gso_segs = 1;
8334 
8335 	/* if we have a HW VLAN tag being added default to the HW one */
8336 	if (skb_vlan_tag_present(skb)) {
8337 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8338 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8339 	/* else if it is a SW VLAN check the next protocol and store the tag */
8340 	} else if (protocol == htons(ETH_P_8021Q)) {
8341 		struct vlan_hdr *vhdr, _vhdr;
8342 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8343 		if (!vhdr)
8344 			goto out_drop;
8345 
8346 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8347 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
8348 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8349 	}
8350 	protocol = vlan_get_protocol(skb);
8351 
8352 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8353 	    adapter->ptp_clock) {
8354 		if (!test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8355 					   &adapter->state)) {
8356 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8357 			tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8358 
8359 			/* schedule check for Tx timestamp */
8360 			adapter->ptp_tx_skb = skb_get(skb);
8361 			adapter->ptp_tx_start = jiffies;
8362 			schedule_work(&adapter->ptp_tx_work);
8363 		} else {
8364 			adapter->tx_hwtstamp_skipped++;
8365 		}
8366 	}
8367 
8368 	skb_tx_timestamp(skb);
8369 
8370 #ifdef CONFIG_PCI_IOV
8371 	/*
8372 	 * Use the l2switch_enable flag - would be false if the DMA
8373 	 * Tx switch had been disabled.
8374 	 */
8375 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8376 		tx_flags |= IXGBE_TX_FLAGS_CC;
8377 
8378 #endif
8379 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8380 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8381 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8382 	     (skb->priority != TC_PRIO_CONTROL))) {
8383 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8384 		tx_flags |= (skb->priority & 0x7) <<
8385 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8386 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8387 			struct vlan_ethhdr *vhdr;
8388 
8389 			if (skb_cow_head(skb, 0))
8390 				goto out_drop;
8391 			vhdr = (struct vlan_ethhdr *)skb->data;
8392 			vhdr->h_vlan_TCI = htons(tx_flags >>
8393 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
8394 		} else {
8395 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8396 		}
8397 	}
8398 
8399 	/* record initial flags and protocol */
8400 	first->tx_flags = tx_flags;
8401 	first->protocol = protocol;
8402 
8403 #ifdef IXGBE_FCOE
8404 	/* setup tx offload for FCoE */
8405 	if ((protocol == htons(ETH_P_FCOE)) &&
8406 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8407 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
8408 		if (tso < 0)
8409 			goto out_drop;
8410 
8411 		goto xmit_fcoe;
8412 	}
8413 
8414 #endif /* IXGBE_FCOE */
8415 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
8416 	if (tso < 0)
8417 		goto out_drop;
8418 	else if (!tso)
8419 		ixgbe_tx_csum(tx_ring, first);
8420 
8421 	/* add the ATR filter if ATR is on */
8422 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8423 		ixgbe_atr(tx_ring, first);
8424 
8425 #ifdef IXGBE_FCOE
8426 xmit_fcoe:
8427 #endif /* IXGBE_FCOE */
8428 	if (ixgbe_tx_map(tx_ring, first, hdr_len))
8429 		goto cleanup_tx_timestamp;
8430 
8431 	return NETDEV_TX_OK;
8432 
8433 out_drop:
8434 	dev_kfree_skb_any(first->skb);
8435 	first->skb = NULL;
8436 cleanup_tx_timestamp:
8437 	if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8438 		dev_kfree_skb_any(adapter->ptp_tx_skb);
8439 		adapter->ptp_tx_skb = NULL;
8440 		cancel_work_sync(&adapter->ptp_tx_work);
8441 		clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8442 	}
8443 
8444 	return NETDEV_TX_OK;
8445 }
8446 
8447 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8448 				      struct net_device *netdev,
8449 				      struct ixgbe_ring *ring)
8450 {
8451 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8452 	struct ixgbe_ring *tx_ring;
8453 
8454 	/*
8455 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
8456 	 * in order to meet this minimum size requirement.
8457 	 */
8458 	if (skb_put_padto(skb, 17))
8459 		return NETDEV_TX_OK;
8460 
8461 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
8462 
8463 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8464 }
8465 
8466 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8467 				    struct net_device *netdev)
8468 {
8469 	return __ixgbe_xmit_frame(skb, netdev, NULL);
8470 }
8471 
8472 /**
8473  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8474  * @netdev: network interface device structure
8475  * @p: pointer to an address structure
8476  *
8477  * Returns 0 on success, negative on failure
8478  **/
8479 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8480 {
8481 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8482 	struct ixgbe_hw *hw = &adapter->hw;
8483 	struct sockaddr *addr = p;
8484 
8485 	if (!is_valid_ether_addr(addr->sa_data))
8486 		return -EADDRNOTAVAIL;
8487 
8488 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
8489 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8490 
8491 	ixgbe_mac_set_default_filter(adapter);
8492 
8493 	return 0;
8494 }
8495 
8496 static int
8497 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8498 {
8499 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8500 	struct ixgbe_hw *hw = &adapter->hw;
8501 	u16 value;
8502 	int rc;
8503 
8504 	if (prtad != hw->phy.mdio.prtad)
8505 		return -EINVAL;
8506 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8507 	if (!rc)
8508 		rc = value;
8509 	return rc;
8510 }
8511 
8512 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8513 			    u16 addr, u16 value)
8514 {
8515 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8516 	struct ixgbe_hw *hw = &adapter->hw;
8517 
8518 	if (prtad != hw->phy.mdio.prtad)
8519 		return -EINVAL;
8520 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8521 }
8522 
8523 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8524 {
8525 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8526 
8527 	switch (cmd) {
8528 	case SIOCSHWTSTAMP:
8529 		return ixgbe_ptp_set_ts_config(adapter, req);
8530 	case SIOCGHWTSTAMP:
8531 		return ixgbe_ptp_get_ts_config(adapter, req);
8532 	default:
8533 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8534 	}
8535 }
8536 
8537 /**
8538  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8539  * netdev->dev_addrs
8540  * @netdev: network interface device structure
8541  *
8542  * Returns non-zero on failure
8543  **/
8544 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8545 {
8546 	int err = 0;
8547 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8548 	struct ixgbe_hw *hw = &adapter->hw;
8549 
8550 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8551 		rtnl_lock();
8552 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8553 		rtnl_unlock();
8554 
8555 		/* update SAN MAC vmdq pool selection */
8556 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8557 	}
8558 	return err;
8559 }
8560 
8561 /**
8562  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8563  * netdev->dev_addrs
8564  * @netdev: network interface device structure
8565  *
8566  * Returns non-zero on failure
8567  **/
8568 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8569 {
8570 	int err = 0;
8571 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8572 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8573 
8574 	if (is_valid_ether_addr(mac->san_addr)) {
8575 		rtnl_lock();
8576 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8577 		rtnl_unlock();
8578 	}
8579 	return err;
8580 }
8581 
8582 #ifdef CONFIG_NET_POLL_CONTROLLER
8583 /*
8584  * Polling 'interrupt' - used by things like netconsole to send skbs
8585  * without having to re-enable interrupts. It's not called while
8586  * the interrupt routine is executing.
8587  */
8588 static void ixgbe_netpoll(struct net_device *netdev)
8589 {
8590 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8591 	int i;
8592 
8593 	/* if interface is down do nothing */
8594 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8595 		return;
8596 
8597 	/* loop through and schedule all active queues */
8598 	for (i = 0; i < adapter->num_q_vectors; i++)
8599 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8600 }
8601 
8602 #endif
8603 
8604 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
8605 				   struct ixgbe_ring *ring)
8606 {
8607 	u64 bytes, packets;
8608 	unsigned int start;
8609 
8610 	if (ring) {
8611 		do {
8612 			start = u64_stats_fetch_begin_irq(&ring->syncp);
8613 			packets = ring->stats.packets;
8614 			bytes   = ring->stats.bytes;
8615 		} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8616 		stats->tx_packets += packets;
8617 		stats->tx_bytes   += bytes;
8618 	}
8619 }
8620 
8621 static void ixgbe_get_stats64(struct net_device *netdev,
8622 			      struct rtnl_link_stats64 *stats)
8623 {
8624 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8625 	int i;
8626 
8627 	rcu_read_lock();
8628 	for (i = 0; i < adapter->num_rx_queues; i++) {
8629 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
8630 		u64 bytes, packets;
8631 		unsigned int start;
8632 
8633 		if (ring) {
8634 			do {
8635 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8636 				packets = ring->stats.packets;
8637 				bytes   = ring->stats.bytes;
8638 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8639 			stats->rx_packets += packets;
8640 			stats->rx_bytes   += bytes;
8641 		}
8642 	}
8643 
8644 	for (i = 0; i < adapter->num_tx_queues; i++) {
8645 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
8646 
8647 		ixgbe_get_ring_stats64(stats, ring);
8648 	}
8649 	for (i = 0; i < adapter->num_xdp_queues; i++) {
8650 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->xdp_ring[i]);
8651 
8652 		ixgbe_get_ring_stats64(stats, ring);
8653 	}
8654 	rcu_read_unlock();
8655 
8656 	/* following stats updated by ixgbe_watchdog_task() */
8657 	stats->multicast	= netdev->stats.multicast;
8658 	stats->rx_errors	= netdev->stats.rx_errors;
8659 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8660 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8661 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8662 }
8663 
8664 #ifdef CONFIG_IXGBE_DCB
8665 /**
8666  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8667  * @adapter: pointer to ixgbe_adapter
8668  * @tc: number of traffic classes currently enabled
8669  *
8670  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8671  * 802.1Q priority maps to a packet buffer that exists.
8672  */
8673 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8674 {
8675 	struct ixgbe_hw *hw = &adapter->hw;
8676 	u32 reg, rsave;
8677 	int i;
8678 
8679 	/* 82598 have a static priority to TC mapping that can not
8680 	 * be changed so no validation is needed.
8681 	 */
8682 	if (hw->mac.type == ixgbe_mac_82598EB)
8683 		return;
8684 
8685 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8686 	rsave = reg;
8687 
8688 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8689 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8690 
8691 		/* If up2tc is out of bounds default to zero */
8692 		if (up2tc > tc)
8693 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8694 	}
8695 
8696 	if (reg != rsave)
8697 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8698 
8699 	return;
8700 }
8701 
8702 /**
8703  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8704  * @adapter: Pointer to adapter struct
8705  *
8706  * Populate the netdev user priority to tc map
8707  */
8708 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8709 {
8710 	struct net_device *dev = adapter->netdev;
8711 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8712 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8713 	u8 prio;
8714 
8715 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8716 		u8 tc = 0;
8717 
8718 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8719 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8720 		else if (ets)
8721 			tc = ets->prio_tc[prio];
8722 
8723 		netdev_set_prio_tc_map(dev, prio, tc);
8724 	}
8725 }
8726 
8727 #endif /* CONFIG_IXGBE_DCB */
8728 /**
8729  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8730  *
8731  * @netdev: net device to configure
8732  * @tc: number of traffic classes to enable
8733  */
8734 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8735 {
8736 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8737 	struct ixgbe_hw *hw = &adapter->hw;
8738 	bool pools;
8739 
8740 	/* Hardware supports up to 8 traffic classes */
8741 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8742 		return -EINVAL;
8743 
8744 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8745 		return -EINVAL;
8746 
8747 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
8748 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
8749 		return -EBUSY;
8750 
8751 	/* Hardware has to reinitialize queues and interrupts to
8752 	 * match packet buffer alignment. Unfortunately, the
8753 	 * hardware is not flexible enough to do this dynamically.
8754 	 */
8755 	if (netif_running(dev))
8756 		ixgbe_close(dev);
8757 	else
8758 		ixgbe_reset(adapter);
8759 
8760 	ixgbe_clear_interrupt_scheme(adapter);
8761 
8762 #ifdef CONFIG_IXGBE_DCB
8763 	if (tc) {
8764 		netdev_set_num_tc(dev, tc);
8765 		ixgbe_set_prio_tc_map(adapter);
8766 
8767 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8768 
8769 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8770 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8771 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8772 		}
8773 	} else {
8774 		netdev_reset_tc(dev);
8775 
8776 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8777 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8778 
8779 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8780 
8781 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8782 		adapter->dcb_cfg.pfc_mode_enable = false;
8783 	}
8784 
8785 	ixgbe_validate_rtr(adapter, tc);
8786 
8787 #endif /* CONFIG_IXGBE_DCB */
8788 	ixgbe_init_interrupt_scheme(adapter);
8789 
8790 	if (netif_running(dev))
8791 		return ixgbe_open(dev);
8792 
8793 	return 0;
8794 }
8795 
8796 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8797 			       struct tc_cls_u32_offload *cls)
8798 {
8799 	u32 hdl = cls->knode.handle;
8800 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8801 	u32 loc = cls->knode.handle & 0xfffff;
8802 	int err = 0, i, j;
8803 	struct ixgbe_jump_table *jump = NULL;
8804 
8805 	if (loc > IXGBE_MAX_HW_ENTRIES)
8806 		return -EINVAL;
8807 
8808 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8809 		return -EINVAL;
8810 
8811 	/* Clear this filter in the link data it is associated with */
8812 	if (uhtid != 0x800) {
8813 		jump = adapter->jump_tables[uhtid];
8814 		if (!jump)
8815 			return -EINVAL;
8816 		if (!test_bit(loc - 1, jump->child_loc_map))
8817 			return -EINVAL;
8818 		clear_bit(loc - 1, jump->child_loc_map);
8819 	}
8820 
8821 	/* Check if the filter being deleted is a link */
8822 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8823 		jump = adapter->jump_tables[i];
8824 		if (jump && jump->link_hdl == hdl) {
8825 			/* Delete filters in the hardware in the child hash
8826 			 * table associated with this link
8827 			 */
8828 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8829 				if (!test_bit(j, jump->child_loc_map))
8830 					continue;
8831 				spin_lock(&adapter->fdir_perfect_lock);
8832 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8833 								      NULL,
8834 								      j + 1);
8835 				spin_unlock(&adapter->fdir_perfect_lock);
8836 				clear_bit(j, jump->child_loc_map);
8837 			}
8838 			/* Remove resources for this link */
8839 			kfree(jump->input);
8840 			kfree(jump->mask);
8841 			kfree(jump);
8842 			adapter->jump_tables[i] = NULL;
8843 			return err;
8844 		}
8845 	}
8846 
8847 	spin_lock(&adapter->fdir_perfect_lock);
8848 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8849 	spin_unlock(&adapter->fdir_perfect_lock);
8850 	return err;
8851 }
8852 
8853 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8854 					    struct tc_cls_u32_offload *cls)
8855 {
8856 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8857 
8858 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8859 		return -EINVAL;
8860 
8861 	/* This ixgbe devices do not support hash tables at the moment
8862 	 * so abort when given hash tables.
8863 	 */
8864 	if (cls->hnode.divisor > 0)
8865 		return -EINVAL;
8866 
8867 	set_bit(uhtid - 1, &adapter->tables);
8868 	return 0;
8869 }
8870 
8871 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
8872 					    struct tc_cls_u32_offload *cls)
8873 {
8874 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8875 
8876 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8877 		return -EINVAL;
8878 
8879 	clear_bit(uhtid - 1, &adapter->tables);
8880 	return 0;
8881 }
8882 
8883 #ifdef CONFIG_NET_CLS_ACT
8884 struct upper_walk_data {
8885 	struct ixgbe_adapter *adapter;
8886 	u64 action;
8887 	int ifindex;
8888 	u8 queue;
8889 };
8890 
8891 static int get_macvlan_queue(struct net_device *upper, void *_data)
8892 {
8893 	if (netif_is_macvlan(upper)) {
8894 		struct macvlan_dev *dfwd = netdev_priv(upper);
8895 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
8896 		struct upper_walk_data *data = _data;
8897 		struct ixgbe_adapter *adapter = data->adapter;
8898 		int ifindex = data->ifindex;
8899 
8900 		if (vadapter && vadapter->netdev->ifindex == ifindex) {
8901 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
8902 			data->action = data->queue;
8903 			return 1;
8904 		}
8905 	}
8906 
8907 	return 0;
8908 }
8909 
8910 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
8911 				  u8 *queue, u64 *action)
8912 {
8913 	unsigned int num_vfs = adapter->num_vfs, vf;
8914 	struct upper_walk_data data;
8915 	struct net_device *upper;
8916 
8917 	/* redirect to a SRIOV VF */
8918 	for (vf = 0; vf < num_vfs; ++vf) {
8919 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
8920 		if (upper->ifindex == ifindex) {
8921 			if (adapter->num_rx_pools > 1)
8922 				*queue = vf * 2;
8923 			else
8924 				*queue = vf * adapter->num_rx_queues_per_pool;
8925 
8926 			*action = vf + 1;
8927 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
8928 			return 0;
8929 		}
8930 	}
8931 
8932 	/* redirect to a offloaded macvlan netdev */
8933 	data.adapter = adapter;
8934 	data.ifindex = ifindex;
8935 	data.action = 0;
8936 	data.queue = 0;
8937 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
8938 					  get_macvlan_queue, &data)) {
8939 		*action = data.action;
8940 		*queue = data.queue;
8941 
8942 		return 0;
8943 	}
8944 
8945 	return -EINVAL;
8946 }
8947 
8948 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8949 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8950 {
8951 	const struct tc_action *a;
8952 	LIST_HEAD(actions);
8953 	int err;
8954 
8955 	if (!tcf_exts_has_actions(exts))
8956 		return -EINVAL;
8957 
8958 	tcf_exts_to_list(exts, &actions);
8959 	list_for_each_entry(a, &actions, list) {
8960 
8961 		/* Drop action */
8962 		if (is_tcf_gact_shot(a)) {
8963 			*action = IXGBE_FDIR_DROP_QUEUE;
8964 			*queue = IXGBE_FDIR_DROP_QUEUE;
8965 			return 0;
8966 		}
8967 
8968 		/* Redirect to a VF or a offloaded macvlan */
8969 		if (is_tcf_mirred_egress_redirect(a)) {
8970 			int ifindex = tcf_mirred_ifindex(a);
8971 
8972 			err = handle_redirect_action(adapter, ifindex, queue,
8973 						     action);
8974 			if (err == 0)
8975 				return err;
8976 		}
8977 	}
8978 
8979 	return -EINVAL;
8980 }
8981 #else
8982 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8983 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8984 {
8985 	return -EINVAL;
8986 }
8987 #endif /* CONFIG_NET_CLS_ACT */
8988 
8989 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
8990 				    union ixgbe_atr_input *mask,
8991 				    struct tc_cls_u32_offload *cls,
8992 				    struct ixgbe_mat_field *field_ptr,
8993 				    struct ixgbe_nexthdr *nexthdr)
8994 {
8995 	int i, j, off;
8996 	__be32 val, m;
8997 	bool found_entry = false, found_jump_field = false;
8998 
8999 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
9000 		off = cls->knode.sel->keys[i].off;
9001 		val = cls->knode.sel->keys[i].val;
9002 		m = cls->knode.sel->keys[i].mask;
9003 
9004 		for (j = 0; field_ptr[j].val; j++) {
9005 			if (field_ptr[j].off == off) {
9006 				field_ptr[j].val(input, mask, val, m);
9007 				input->filter.formatted.flow_type |=
9008 					field_ptr[j].type;
9009 				found_entry = true;
9010 				break;
9011 			}
9012 		}
9013 		if (nexthdr) {
9014 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
9015 			    nexthdr->val == cls->knode.sel->keys[i].val &&
9016 			    nexthdr->mask == cls->knode.sel->keys[i].mask)
9017 				found_jump_field = true;
9018 			else
9019 				continue;
9020 		}
9021 	}
9022 
9023 	if (nexthdr && !found_jump_field)
9024 		return -EINVAL;
9025 
9026 	if (!found_entry)
9027 		return 0;
9028 
9029 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9030 				    IXGBE_ATR_L4TYPE_MASK;
9031 
9032 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9033 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9034 
9035 	return 0;
9036 }
9037 
9038 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9039 				  struct tc_cls_u32_offload *cls)
9040 {
9041 	__be16 protocol = cls->common.protocol;
9042 	u32 loc = cls->knode.handle & 0xfffff;
9043 	struct ixgbe_hw *hw = &adapter->hw;
9044 	struct ixgbe_mat_field *field_ptr;
9045 	struct ixgbe_fdir_filter *input = NULL;
9046 	union ixgbe_atr_input *mask = NULL;
9047 	struct ixgbe_jump_table *jump = NULL;
9048 	int i, err = -EINVAL;
9049 	u8 queue;
9050 	u32 uhtid, link_uhtid;
9051 
9052 	uhtid = TC_U32_USERHTID(cls->knode.handle);
9053 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9054 
9055 	/* At the moment cls_u32 jumps to network layer and skips past
9056 	 * L2 headers. The canonical method to match L2 frames is to use
9057 	 * negative values. However this is error prone at best but really
9058 	 * just broken because there is no way to "know" what sort of hdr
9059 	 * is in front of the network layer. Fix cls_u32 to support L2
9060 	 * headers when needed.
9061 	 */
9062 	if (protocol != htons(ETH_P_IP))
9063 		return err;
9064 
9065 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9066 		e_err(drv, "Location out of range\n");
9067 		return err;
9068 	}
9069 
9070 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
9071 	 * links and also the fields used to advance the parser across each
9072 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9073 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9074 	 * To add support for new nodes update ixgbe_model.h parse structures
9075 	 * this function _should_ be generic try not to hardcode values here.
9076 	 */
9077 	if (uhtid == 0x800) {
9078 		field_ptr = (adapter->jump_tables[0])->mat;
9079 	} else {
9080 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9081 			return err;
9082 		if (!adapter->jump_tables[uhtid])
9083 			return err;
9084 		field_ptr = (adapter->jump_tables[uhtid])->mat;
9085 	}
9086 
9087 	if (!field_ptr)
9088 		return err;
9089 
9090 	/* At this point we know the field_ptr is valid and need to either
9091 	 * build cls_u32 link or attach filter. Because adding a link to
9092 	 * a handle that does not exist is invalid and the same for adding
9093 	 * rules to handles that don't exist.
9094 	 */
9095 
9096 	if (link_uhtid) {
9097 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9098 
9099 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9100 			return err;
9101 
9102 		if (!test_bit(link_uhtid - 1, &adapter->tables))
9103 			return err;
9104 
9105 		/* Multiple filters as links to the same hash table are not
9106 		 * supported. To add a new filter with the same next header
9107 		 * but different match/jump conditions, create a new hash table
9108 		 * and link to it.
9109 		 */
9110 		if (adapter->jump_tables[link_uhtid] &&
9111 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
9112 			e_err(drv, "Link filter exists for link: %x\n",
9113 			      link_uhtid);
9114 			return err;
9115 		}
9116 
9117 		for (i = 0; nexthdr[i].jump; i++) {
9118 			if (nexthdr[i].o != cls->knode.sel->offoff ||
9119 			    nexthdr[i].s != cls->knode.sel->offshift ||
9120 			    nexthdr[i].m != cls->knode.sel->offmask)
9121 				return err;
9122 
9123 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9124 			if (!jump)
9125 				return -ENOMEM;
9126 			input = kzalloc(sizeof(*input), GFP_KERNEL);
9127 			if (!input) {
9128 				err = -ENOMEM;
9129 				goto free_jump;
9130 			}
9131 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9132 			if (!mask) {
9133 				err = -ENOMEM;
9134 				goto free_input;
9135 			}
9136 			jump->input = input;
9137 			jump->mask = mask;
9138 			jump->link_hdl = cls->knode.handle;
9139 
9140 			err = ixgbe_clsu32_build_input(input, mask, cls,
9141 						       field_ptr, &nexthdr[i]);
9142 			if (!err) {
9143 				jump->mat = nexthdr[i].jump;
9144 				adapter->jump_tables[link_uhtid] = jump;
9145 				break;
9146 			}
9147 		}
9148 		return 0;
9149 	}
9150 
9151 	input = kzalloc(sizeof(*input), GFP_KERNEL);
9152 	if (!input)
9153 		return -ENOMEM;
9154 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9155 	if (!mask) {
9156 		err = -ENOMEM;
9157 		goto free_input;
9158 	}
9159 
9160 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9161 		if ((adapter->jump_tables[uhtid])->input)
9162 			memcpy(input, (adapter->jump_tables[uhtid])->input,
9163 			       sizeof(*input));
9164 		if ((adapter->jump_tables[uhtid])->mask)
9165 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9166 			       sizeof(*mask));
9167 
9168 		/* Lookup in all child hash tables if this location is already
9169 		 * filled with a filter
9170 		 */
9171 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9172 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
9173 
9174 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
9175 				e_err(drv, "Filter exists in location: %x\n",
9176 				      loc);
9177 				err = -EINVAL;
9178 				goto err_out;
9179 			}
9180 		}
9181 	}
9182 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9183 	if (err)
9184 		goto err_out;
9185 
9186 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9187 			       &queue);
9188 	if (err < 0)
9189 		goto err_out;
9190 
9191 	input->sw_idx = loc;
9192 
9193 	spin_lock(&adapter->fdir_perfect_lock);
9194 
9195 	if (hlist_empty(&adapter->fdir_filter_list)) {
9196 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9197 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9198 		if (err)
9199 			goto err_out_w_lock;
9200 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9201 		err = -EINVAL;
9202 		goto err_out_w_lock;
9203 	}
9204 
9205 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9206 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9207 						    input->sw_idx, queue);
9208 	if (!err)
9209 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9210 	spin_unlock(&adapter->fdir_perfect_lock);
9211 
9212 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9213 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9214 
9215 	kfree(mask);
9216 	return err;
9217 err_out_w_lock:
9218 	spin_unlock(&adapter->fdir_perfect_lock);
9219 err_out:
9220 	kfree(mask);
9221 free_input:
9222 	kfree(input);
9223 free_jump:
9224 	kfree(jump);
9225 	return err;
9226 }
9227 
9228 static int ixgbe_setup_tc_cls_u32(struct net_device *dev,
9229 				  struct tc_cls_u32_offload *cls_u32)
9230 {
9231 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9232 
9233 	if (!is_classid_clsact_ingress(cls_u32->common.classid) ||
9234 	    cls_u32->common.chain_index)
9235 		return -EOPNOTSUPP;
9236 
9237 	switch (cls_u32->command) {
9238 	case TC_CLSU32_NEW_KNODE:
9239 	case TC_CLSU32_REPLACE_KNODE:
9240 		return ixgbe_configure_clsu32(adapter, cls_u32);
9241 	case TC_CLSU32_DELETE_KNODE:
9242 		return ixgbe_delete_clsu32(adapter, cls_u32);
9243 	case TC_CLSU32_NEW_HNODE:
9244 	case TC_CLSU32_REPLACE_HNODE:
9245 		return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9246 	case TC_CLSU32_DELETE_HNODE:
9247 		return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9248 	default:
9249 		return -EOPNOTSUPP;
9250 	}
9251 }
9252 
9253 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9254 				 struct tc_mqprio_qopt *mqprio)
9255 {
9256 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9257 	return ixgbe_setup_tc(dev, mqprio->num_tc);
9258 }
9259 
9260 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9261 			    void *type_data)
9262 {
9263 	switch (type) {
9264 	case TC_SETUP_CLSU32:
9265 		return ixgbe_setup_tc_cls_u32(dev, type_data);
9266 	case TC_SETUP_MQPRIO:
9267 		return ixgbe_setup_tc_mqprio(dev, type_data);
9268 	default:
9269 		return -EOPNOTSUPP;
9270 	}
9271 }
9272 
9273 #ifdef CONFIG_PCI_IOV
9274 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9275 {
9276 	struct net_device *netdev = adapter->netdev;
9277 
9278 	rtnl_lock();
9279 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
9280 	rtnl_unlock();
9281 }
9282 
9283 #endif
9284 void ixgbe_do_reset(struct net_device *netdev)
9285 {
9286 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9287 
9288 	if (netif_running(netdev))
9289 		ixgbe_reinit_locked(adapter);
9290 	else
9291 		ixgbe_reset(adapter);
9292 }
9293 
9294 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9295 					    netdev_features_t features)
9296 {
9297 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9298 
9299 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9300 	if (!(features & NETIF_F_RXCSUM))
9301 		features &= ~NETIF_F_LRO;
9302 
9303 	/* Turn off LRO if not RSC capable */
9304 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9305 		features &= ~NETIF_F_LRO;
9306 
9307 	return features;
9308 }
9309 
9310 static int ixgbe_set_features(struct net_device *netdev,
9311 			      netdev_features_t features)
9312 {
9313 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9314 	netdev_features_t changed = netdev->features ^ features;
9315 	bool need_reset = false;
9316 
9317 	/* Make sure RSC matches LRO, reset if change */
9318 	if (!(features & NETIF_F_LRO)) {
9319 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9320 			need_reset = true;
9321 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9322 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9323 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9324 		if (adapter->rx_itr_setting == 1 ||
9325 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9326 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9327 			need_reset = true;
9328 		} else if ((changed ^ features) & NETIF_F_LRO) {
9329 			e_info(probe, "rx-usecs set too low, "
9330 			       "disabling RSC\n");
9331 		}
9332 	}
9333 
9334 	/*
9335 	 * Check if Flow Director n-tuple support or hw_tc support was
9336 	 * enabled or disabled.  If the state changed, we need to reset.
9337 	 */
9338 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9339 		/* turn off ATR, enable perfect filters and reset */
9340 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9341 			need_reset = true;
9342 
9343 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9344 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9345 	} else {
9346 		/* turn off perfect filters, enable ATR and reset */
9347 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9348 			need_reset = true;
9349 
9350 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9351 
9352 		/* We cannot enable ATR if SR-IOV is enabled */
9353 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9354 		    /* We cannot enable ATR if we have 2 or more tcs */
9355 		    (netdev_get_num_tc(netdev) > 1) ||
9356 		    /* We cannot enable ATR if RSS is disabled */
9357 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9358 		    /* A sample rate of 0 indicates ATR disabled */
9359 		    (!adapter->atr_sample_rate))
9360 			; /* do nothing not supported */
9361 		else /* otherwise supported and set the flag */
9362 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9363 	}
9364 
9365 	if (changed & NETIF_F_RXALL)
9366 		need_reset = true;
9367 
9368 	netdev->features = features;
9369 
9370 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
9371 		if (features & NETIF_F_RXCSUM) {
9372 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9373 		} else {
9374 			u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9375 
9376 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9377 		}
9378 	}
9379 
9380 	if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
9381 		if (features & NETIF_F_RXCSUM) {
9382 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9383 		} else {
9384 			u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9385 
9386 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9387 		}
9388 	}
9389 
9390 	if (need_reset)
9391 		ixgbe_do_reset(netdev);
9392 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9393 			    NETIF_F_HW_VLAN_CTAG_FILTER))
9394 		ixgbe_set_rx_mode(netdev);
9395 
9396 	return 0;
9397 }
9398 
9399 /**
9400  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
9401  * @dev: The port's netdev
9402  * @ti: Tunnel endpoint information
9403  **/
9404 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
9405 				      struct udp_tunnel_info *ti)
9406 {
9407 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9408 	struct ixgbe_hw *hw = &adapter->hw;
9409 	__be16 port = ti->port;
9410 	u32 port_shift = 0;
9411 	u32 reg;
9412 
9413 	if (ti->sa_family != AF_INET)
9414 		return;
9415 
9416 	switch (ti->type) {
9417 	case UDP_TUNNEL_TYPE_VXLAN:
9418 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9419 			return;
9420 
9421 		if (adapter->vxlan_port == port)
9422 			return;
9423 
9424 		if (adapter->vxlan_port) {
9425 			netdev_info(dev,
9426 				    "VXLAN port %d set, not adding port %d\n",
9427 				    ntohs(adapter->vxlan_port),
9428 				    ntohs(port));
9429 			return;
9430 		}
9431 
9432 		adapter->vxlan_port = port;
9433 		break;
9434 	case UDP_TUNNEL_TYPE_GENEVE:
9435 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9436 			return;
9437 
9438 		if (adapter->geneve_port == port)
9439 			return;
9440 
9441 		if (adapter->geneve_port) {
9442 			netdev_info(dev,
9443 				    "GENEVE port %d set, not adding port %d\n",
9444 				    ntohs(adapter->geneve_port),
9445 				    ntohs(port));
9446 			return;
9447 		}
9448 
9449 		port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
9450 		adapter->geneve_port = port;
9451 		break;
9452 	default:
9453 		return;
9454 	}
9455 
9456 	reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
9457 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
9458 }
9459 
9460 /**
9461  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
9462  * @dev: The port's netdev
9463  * @ti: Tunnel endpoint information
9464  **/
9465 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
9466 				      struct udp_tunnel_info *ti)
9467 {
9468 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9469 	u32 port_mask;
9470 
9471 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
9472 	    ti->type != UDP_TUNNEL_TYPE_GENEVE)
9473 		return;
9474 
9475 	if (ti->sa_family != AF_INET)
9476 		return;
9477 
9478 	switch (ti->type) {
9479 	case UDP_TUNNEL_TYPE_VXLAN:
9480 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9481 			return;
9482 
9483 		if (adapter->vxlan_port != ti->port) {
9484 			netdev_info(dev, "VXLAN port %d not found\n",
9485 				    ntohs(ti->port));
9486 			return;
9487 		}
9488 
9489 		port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9490 		break;
9491 	case UDP_TUNNEL_TYPE_GENEVE:
9492 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9493 			return;
9494 
9495 		if (adapter->geneve_port != ti->port) {
9496 			netdev_info(dev, "GENEVE port %d not found\n",
9497 				    ntohs(ti->port));
9498 			return;
9499 		}
9500 
9501 		port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9502 		break;
9503 	default:
9504 		return;
9505 	}
9506 
9507 	ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9508 	adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9509 }
9510 
9511 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9512 			     struct net_device *dev,
9513 			     const unsigned char *addr, u16 vid,
9514 			     u16 flags)
9515 {
9516 	/* guarantee we can provide a unique filter for the unicast address */
9517 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9518 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9519 		u16 pool = VMDQ_P(0);
9520 
9521 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9522 			return -ENOMEM;
9523 	}
9524 
9525 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9526 }
9527 
9528 /**
9529  * ixgbe_configure_bridge_mode - set various bridge modes
9530  * @adapter - the private structure
9531  * @mode - requested bridge mode
9532  *
9533  * Configure some settings require for various bridge modes.
9534  **/
9535 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9536 				       __u16 mode)
9537 {
9538 	struct ixgbe_hw *hw = &adapter->hw;
9539 	unsigned int p, num_pools;
9540 	u32 vmdctl;
9541 
9542 	switch (mode) {
9543 	case BRIDGE_MODE_VEPA:
9544 		/* disable Tx loopback, rely on switch hairpin mode */
9545 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9546 
9547 		/* must enable Rx switching replication to allow multicast
9548 		 * packet reception on all VFs, and to enable source address
9549 		 * pruning.
9550 		 */
9551 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9552 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9553 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9554 
9555 		/* enable Rx source address pruning. Note, this requires
9556 		 * replication to be enabled or else it does nothing.
9557 		 */
9558 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9559 		for (p = 0; p < num_pools; p++) {
9560 			if (hw->mac.ops.set_source_address_pruning)
9561 				hw->mac.ops.set_source_address_pruning(hw,
9562 								       true,
9563 								       p);
9564 		}
9565 		break;
9566 	case BRIDGE_MODE_VEB:
9567 		/* enable Tx loopback for internal VF/PF communication */
9568 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9569 				IXGBE_PFDTXGSWC_VT_LBEN);
9570 
9571 		/* disable Rx switching replication unless we have SR-IOV
9572 		 * virtual functions
9573 		 */
9574 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9575 		if (!adapter->num_vfs)
9576 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9577 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9578 
9579 		/* disable Rx source address pruning, since we don't expect to
9580 		 * be receiving external loopback of our transmitted frames.
9581 		 */
9582 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9583 		for (p = 0; p < num_pools; p++) {
9584 			if (hw->mac.ops.set_source_address_pruning)
9585 				hw->mac.ops.set_source_address_pruning(hw,
9586 								       false,
9587 								       p);
9588 		}
9589 		break;
9590 	default:
9591 		return -EINVAL;
9592 	}
9593 
9594 	adapter->bridge_mode = mode;
9595 
9596 	e_info(drv, "enabling bridge mode: %s\n",
9597 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9598 
9599 	return 0;
9600 }
9601 
9602 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9603 				    struct nlmsghdr *nlh, u16 flags)
9604 {
9605 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9606 	struct nlattr *attr, *br_spec;
9607 	int rem;
9608 
9609 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9610 		return -EOPNOTSUPP;
9611 
9612 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9613 	if (!br_spec)
9614 		return -EINVAL;
9615 
9616 	nla_for_each_nested(attr, br_spec, rem) {
9617 		int status;
9618 		__u16 mode;
9619 
9620 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9621 			continue;
9622 
9623 		if (nla_len(attr) < sizeof(mode))
9624 			return -EINVAL;
9625 
9626 		mode = nla_get_u16(attr);
9627 		status = ixgbe_configure_bridge_mode(adapter, mode);
9628 		if (status)
9629 			return status;
9630 
9631 		break;
9632 	}
9633 
9634 	return 0;
9635 }
9636 
9637 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9638 				    struct net_device *dev,
9639 				    u32 filter_mask, int nlflags)
9640 {
9641 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9642 
9643 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9644 		return 0;
9645 
9646 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9647 				       adapter->bridge_mode, 0, 0, nlflags,
9648 				       filter_mask, NULL);
9649 }
9650 
9651 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9652 {
9653 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
9654 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9655 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
9656 	unsigned int limit;
9657 	int pool, err;
9658 
9659 	/* Hardware has a limited number of available pools. Each VF, and the
9660 	 * PF require a pool. Check to ensure we don't attempt to use more
9661 	 * then the available number of pools.
9662 	 */
9663 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9664 		return ERR_PTR(-EINVAL);
9665 
9666 #ifdef CONFIG_RPS
9667 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
9668 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
9669 			    vdev->name);
9670 		return ERR_PTR(-EINVAL);
9671 	}
9672 #endif
9673 	/* Check for hardware restriction on number of rx/tx queues */
9674 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
9675 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
9676 		netdev_info(pdev,
9677 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
9678 			    pdev->name);
9679 		return ERR_PTR(-EINVAL);
9680 	}
9681 
9682 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9683 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
9684 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
9685 		return ERR_PTR(-EBUSY);
9686 
9687 	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
9688 	if (!fwd_adapter)
9689 		return ERR_PTR(-ENOMEM);
9690 
9691 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
9692 	adapter->num_rx_pools++;
9693 	set_bit(pool, &adapter->fwd_bitmask);
9694 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9695 
9696 	/* Enable VMDq flag so device will be set in VM mode */
9697 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
9698 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9699 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
9700 
9701 	/* Force reinit of ring allocation with VMDQ enabled */
9702 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9703 	if (err)
9704 		goto fwd_add_err;
9705 	fwd_adapter->pool = pool;
9706 	fwd_adapter->real_adapter = adapter;
9707 
9708 	if (netif_running(pdev)) {
9709 		err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
9710 		if (err)
9711 			goto fwd_add_err;
9712 		netif_tx_start_all_queues(vdev);
9713 	}
9714 
9715 	return fwd_adapter;
9716 fwd_add_err:
9717 	/* unwind counter and free adapter struct */
9718 	netdev_info(pdev,
9719 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
9720 	clear_bit(pool, &adapter->fwd_bitmask);
9721 	adapter->num_rx_pools--;
9722 	kfree(fwd_adapter);
9723 	return ERR_PTR(err);
9724 }
9725 
9726 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9727 {
9728 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
9729 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
9730 	unsigned int limit;
9731 
9732 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
9733 	adapter->num_rx_pools--;
9734 
9735 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9736 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9737 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
9738 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9739 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
9740 		   fwd_adapter->pool, adapter->num_rx_pools,
9741 		   fwd_adapter->rx_base_queue,
9742 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
9743 		   adapter->fwd_bitmask);
9744 	kfree(fwd_adapter);
9745 }
9746 
9747 #define IXGBE_MAX_MAC_HDR_LEN		127
9748 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9749 
9750 static netdev_features_t
9751 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9752 		     netdev_features_t features)
9753 {
9754 	unsigned int network_hdr_len, mac_hdr_len;
9755 
9756 	/* Make certain the headers can be described by a context descriptor */
9757 	mac_hdr_len = skb_network_header(skb) - skb->data;
9758 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9759 		return features & ~(NETIF_F_HW_CSUM |
9760 				    NETIF_F_SCTP_CRC |
9761 				    NETIF_F_HW_VLAN_CTAG_TX |
9762 				    NETIF_F_TSO |
9763 				    NETIF_F_TSO6);
9764 
9765 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9766 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9767 		return features & ~(NETIF_F_HW_CSUM |
9768 				    NETIF_F_SCTP_CRC |
9769 				    NETIF_F_TSO |
9770 				    NETIF_F_TSO6);
9771 
9772 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9773 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9774 	 */
9775 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
9776 		features &= ~NETIF_F_TSO;
9777 
9778 	return features;
9779 }
9780 
9781 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
9782 {
9783 	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
9784 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9785 	struct bpf_prog *old_prog;
9786 
9787 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
9788 		return -EINVAL;
9789 
9790 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
9791 		return -EINVAL;
9792 
9793 	/* verify ixgbe ring attributes are sufficient for XDP */
9794 	for (i = 0; i < adapter->num_rx_queues; i++) {
9795 		struct ixgbe_ring *ring = adapter->rx_ring[i];
9796 
9797 		if (ring_is_rsc_enabled(ring))
9798 			return -EINVAL;
9799 
9800 		if (frame_size > ixgbe_rx_bufsz(ring))
9801 			return -EINVAL;
9802 	}
9803 
9804 	if (nr_cpu_ids > MAX_XDP_QUEUES)
9805 		return -ENOMEM;
9806 
9807 	old_prog = xchg(&adapter->xdp_prog, prog);
9808 
9809 	/* If transitioning XDP modes reconfigure rings */
9810 	if (!!prog != !!old_prog) {
9811 		int err = ixgbe_setup_tc(dev, netdev_get_num_tc(dev));
9812 
9813 		if (err) {
9814 			rcu_assign_pointer(adapter->xdp_prog, old_prog);
9815 			return -EINVAL;
9816 		}
9817 	} else {
9818 		for (i = 0; i < adapter->num_rx_queues; i++)
9819 			xchg(&adapter->rx_ring[i]->xdp_prog, adapter->xdp_prog);
9820 	}
9821 
9822 	if (old_prog)
9823 		bpf_prog_put(old_prog);
9824 
9825 	return 0;
9826 }
9827 
9828 static int ixgbe_xdp(struct net_device *dev, struct netdev_xdp *xdp)
9829 {
9830 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9831 
9832 	switch (xdp->command) {
9833 	case XDP_SETUP_PROG:
9834 		return ixgbe_xdp_setup(dev, xdp->prog);
9835 	case XDP_QUERY_PROG:
9836 		xdp->prog_attached = !!(adapter->xdp_prog);
9837 		xdp->prog_id = adapter->xdp_prog ?
9838 			adapter->xdp_prog->aux->id : 0;
9839 		return 0;
9840 	default:
9841 		return -EINVAL;
9842 	}
9843 }
9844 
9845 static int ixgbe_xdp_xmit(struct net_device *dev, struct xdp_buff *xdp)
9846 {
9847 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9848 	struct ixgbe_ring *ring;
9849 	int err;
9850 
9851 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
9852 		return -ENETDOWN;
9853 
9854 	/* During program transitions its possible adapter->xdp_prog is assigned
9855 	 * but ring has not been configured yet. In this case simply abort xmit.
9856 	 */
9857 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
9858 	if (unlikely(!ring))
9859 		return -ENXIO;
9860 
9861 	err = ixgbe_xmit_xdp_ring(adapter, xdp);
9862 	if (err != IXGBE_XDP_TX)
9863 		return -ENOSPC;
9864 
9865 	return 0;
9866 }
9867 
9868 static void ixgbe_xdp_flush(struct net_device *dev)
9869 {
9870 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9871 	struct ixgbe_ring *ring;
9872 
9873 	/* Its possible the device went down between xdp xmit and flush so
9874 	 * we need to ensure device is still up.
9875 	 */
9876 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
9877 		return;
9878 
9879 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
9880 	if (unlikely(!ring))
9881 		return;
9882 
9883 	/* Force memory writes to complete before letting h/w know there
9884 	 * are new descriptors to fetch.
9885 	 */
9886 	wmb();
9887 	writel(ring->next_to_use, ring->tail);
9888 
9889 	return;
9890 }
9891 
9892 static const struct net_device_ops ixgbe_netdev_ops = {
9893 	.ndo_open		= ixgbe_open,
9894 	.ndo_stop		= ixgbe_close,
9895 	.ndo_start_xmit		= ixgbe_xmit_frame,
9896 	.ndo_select_queue	= ixgbe_select_queue,
9897 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
9898 	.ndo_validate_addr	= eth_validate_addr,
9899 	.ndo_set_mac_address	= ixgbe_set_mac,
9900 	.ndo_change_mtu		= ixgbe_change_mtu,
9901 	.ndo_tx_timeout		= ixgbe_tx_timeout,
9902 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
9903 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
9904 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
9905 	.ndo_do_ioctl		= ixgbe_ioctl,
9906 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
9907 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
9908 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
9909 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
9910 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
9911 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
9912 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
9913 	.ndo_get_stats64	= ixgbe_get_stats64,
9914 	.ndo_setup_tc		= __ixgbe_setup_tc,
9915 #ifdef CONFIG_NET_POLL_CONTROLLER
9916 	.ndo_poll_controller	= ixgbe_netpoll,
9917 #endif
9918 #ifdef IXGBE_FCOE
9919 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
9920 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
9921 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
9922 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
9923 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
9924 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
9925 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
9926 #endif /* IXGBE_FCOE */
9927 	.ndo_set_features = ixgbe_set_features,
9928 	.ndo_fix_features = ixgbe_fix_features,
9929 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
9930 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
9931 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
9932 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
9933 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
9934 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
9935 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
9936 	.ndo_features_check	= ixgbe_features_check,
9937 	.ndo_xdp		= ixgbe_xdp,
9938 	.ndo_xdp_xmit		= ixgbe_xdp_xmit,
9939 	.ndo_xdp_flush		= ixgbe_xdp_flush,
9940 };
9941 
9942 /**
9943  * ixgbe_enumerate_functions - Get the number of ports this device has
9944  * @adapter: adapter structure
9945  *
9946  * This function enumerates the phsyical functions co-located on a single slot,
9947  * in order to determine how many ports a device has. This is most useful in
9948  * determining the required GT/s of PCIe bandwidth necessary for optimal
9949  * performance.
9950  **/
9951 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
9952 {
9953 	struct pci_dev *entry, *pdev = adapter->pdev;
9954 	int physfns = 0;
9955 
9956 	/* Some cards can not use the generic count PCIe functions method,
9957 	 * because they are behind a parent switch, so we hardcode these with
9958 	 * the correct number of functions.
9959 	 */
9960 	if (ixgbe_pcie_from_parent(&adapter->hw))
9961 		physfns = 4;
9962 
9963 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
9964 		/* don't count virtual functions */
9965 		if (entry->is_virtfn)
9966 			continue;
9967 
9968 		/* When the devices on the bus don't all match our device ID,
9969 		 * we can't reliably determine the correct number of
9970 		 * functions. This can occur if a function has been direct
9971 		 * attached to a virtual machine using VT-d, for example. In
9972 		 * this case, simply return -1 to indicate this.
9973 		 */
9974 		if ((entry->vendor != pdev->vendor) ||
9975 		    (entry->device != pdev->device))
9976 			return -1;
9977 
9978 		physfns++;
9979 	}
9980 
9981 	return physfns;
9982 }
9983 
9984 /**
9985  * ixgbe_wol_supported - Check whether device supports WoL
9986  * @adapter: the adapter private structure
9987  * @device_id: the device ID
9988  * @subdev_id: the subsystem device ID
9989  *
9990  * This function is used by probe and ethtool to determine
9991  * which devices have WoL support
9992  *
9993  **/
9994 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
9995 			 u16 subdevice_id)
9996 {
9997 	struct ixgbe_hw *hw = &adapter->hw;
9998 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
9999 
10000 	/* WOL not supported on 82598 */
10001 	if (hw->mac.type == ixgbe_mac_82598EB)
10002 		return false;
10003 
10004 	/* check eeprom to see if WOL is enabled for X540 and newer */
10005 	if (hw->mac.type >= ixgbe_mac_X540) {
10006 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10007 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10008 		     (hw->bus.func == 0)))
10009 			return true;
10010 	}
10011 
10012 	/* WOL is determined based on device IDs for 82599 MACs */
10013 	switch (device_id) {
10014 	case IXGBE_DEV_ID_82599_SFP:
10015 		/* Only these subdevices could supports WOL */
10016 		switch (subdevice_id) {
10017 		case IXGBE_SUBDEV_ID_82599_560FLR:
10018 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10019 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10020 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10021 			/* only support first port */
10022 			if (hw->bus.func != 0)
10023 				break;
10024 			/* fall through */
10025 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10026 		case IXGBE_SUBDEV_ID_82599_SFP:
10027 		case IXGBE_SUBDEV_ID_82599_RNDC:
10028 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10029 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10030 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10031 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10032 			return true;
10033 		}
10034 		break;
10035 	case IXGBE_DEV_ID_82599EN_SFP:
10036 		/* Only these subdevices support WOL */
10037 		switch (subdevice_id) {
10038 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10039 			return true;
10040 		}
10041 		break;
10042 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10043 		/* All except this subdevice support WOL */
10044 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10045 			return true;
10046 		break;
10047 	case IXGBE_DEV_ID_82599_KX4:
10048 		return  true;
10049 	default:
10050 		break;
10051 	}
10052 
10053 	return false;
10054 }
10055 
10056 /**
10057  * ixgbe_probe - Device Initialization Routine
10058  * @pdev: PCI device information struct
10059  * @ent: entry in ixgbe_pci_tbl
10060  *
10061  * Returns 0 on success, negative on failure
10062  *
10063  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10064  * The OS initialization, configuring of the adapter private structure,
10065  * and a hardware reset occur.
10066  **/
10067 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10068 {
10069 	struct net_device *netdev;
10070 	struct ixgbe_adapter *adapter = NULL;
10071 	struct ixgbe_hw *hw;
10072 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10073 	int i, err, pci_using_dac, expected_gts;
10074 	unsigned int indices = MAX_TX_QUEUES;
10075 	u8 part_str[IXGBE_PBANUM_LENGTH];
10076 	bool disable_dev = false;
10077 #ifdef IXGBE_FCOE
10078 	u16 device_caps;
10079 #endif
10080 	u32 eec;
10081 
10082 	/* Catch broken hardware that put the wrong VF device ID in
10083 	 * the PCIe SR-IOV capability.
10084 	 */
10085 	if (pdev->is_virtfn) {
10086 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10087 		     pci_name(pdev), pdev->vendor, pdev->device);
10088 		return -EINVAL;
10089 	}
10090 
10091 	err = pci_enable_device_mem(pdev);
10092 	if (err)
10093 		return err;
10094 
10095 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
10096 		pci_using_dac = 1;
10097 	} else {
10098 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10099 		if (err) {
10100 			dev_err(&pdev->dev,
10101 				"No usable DMA configuration, aborting\n");
10102 			goto err_dma;
10103 		}
10104 		pci_using_dac = 0;
10105 	}
10106 
10107 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10108 	if (err) {
10109 		dev_err(&pdev->dev,
10110 			"pci_request_selected_regions failed 0x%x\n", err);
10111 		goto err_pci_reg;
10112 	}
10113 
10114 	pci_enable_pcie_error_reporting(pdev);
10115 
10116 	pci_set_master(pdev);
10117 	pci_save_state(pdev);
10118 
10119 	if (ii->mac == ixgbe_mac_82598EB) {
10120 #ifdef CONFIG_IXGBE_DCB
10121 		/* 8 TC w/ 4 queues per TC */
10122 		indices = 4 * MAX_TRAFFIC_CLASS;
10123 #else
10124 		indices = IXGBE_MAX_RSS_INDICES;
10125 #endif
10126 	}
10127 
10128 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10129 	if (!netdev) {
10130 		err = -ENOMEM;
10131 		goto err_alloc_etherdev;
10132 	}
10133 
10134 	SET_NETDEV_DEV(netdev, &pdev->dev);
10135 
10136 	adapter = netdev_priv(netdev);
10137 
10138 	adapter->netdev = netdev;
10139 	adapter->pdev = pdev;
10140 	hw = &adapter->hw;
10141 	hw->back = adapter;
10142 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10143 
10144 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10145 			      pci_resource_len(pdev, 0));
10146 	adapter->io_addr = hw->hw_addr;
10147 	if (!hw->hw_addr) {
10148 		err = -EIO;
10149 		goto err_ioremap;
10150 	}
10151 
10152 	netdev->netdev_ops = &ixgbe_netdev_ops;
10153 	ixgbe_set_ethtool_ops(netdev);
10154 	netdev->watchdog_timeo = 5 * HZ;
10155 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10156 
10157 	/* Setup hw api */
10158 	hw->mac.ops   = *ii->mac_ops;
10159 	hw->mac.type  = ii->mac;
10160 	hw->mvals     = ii->mvals;
10161 	if (ii->link_ops)
10162 		hw->link.ops  = *ii->link_ops;
10163 
10164 	/* EEPROM */
10165 	hw->eeprom.ops = *ii->eeprom_ops;
10166 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10167 	if (ixgbe_removed(hw->hw_addr)) {
10168 		err = -EIO;
10169 		goto err_ioremap;
10170 	}
10171 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10172 	if (!(eec & BIT(8)))
10173 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10174 
10175 	/* PHY */
10176 	hw->phy.ops = *ii->phy_ops;
10177 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10178 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
10179 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10180 	hw->phy.mdio.mmds = 0;
10181 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10182 	hw->phy.mdio.dev = netdev;
10183 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10184 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10185 
10186 	/* setup the private structure */
10187 	err = ixgbe_sw_init(adapter, ii);
10188 	if (err)
10189 		goto err_sw_init;
10190 
10191 	/* Make sure the SWFW semaphore is in a valid state */
10192 	if (hw->mac.ops.init_swfw_sync)
10193 		hw->mac.ops.init_swfw_sync(hw);
10194 
10195 	/* Make it possible the adapter to be woken up via WOL */
10196 	switch (adapter->hw.mac.type) {
10197 	case ixgbe_mac_82599EB:
10198 	case ixgbe_mac_X540:
10199 	case ixgbe_mac_X550:
10200 	case ixgbe_mac_X550EM_x:
10201 	case ixgbe_mac_x550em_a:
10202 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10203 		break;
10204 	default:
10205 		break;
10206 	}
10207 
10208 	/*
10209 	 * If there is a fan on this device and it has failed log the
10210 	 * failure.
10211 	 */
10212 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10213 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10214 		if (esdp & IXGBE_ESDP_SDP1)
10215 			e_crit(probe, "Fan has stopped, replace the adapter\n");
10216 	}
10217 
10218 	if (allow_unsupported_sfp)
10219 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
10220 
10221 	/* reset_hw fills in the perm_addr as well */
10222 	hw->phy.reset_if_overtemp = true;
10223 	err = hw->mac.ops.reset_hw(hw);
10224 	hw->phy.reset_if_overtemp = false;
10225 	ixgbe_set_eee_capable(adapter);
10226 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
10227 		err = 0;
10228 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
10229 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10230 		e_dev_err("Reload the driver after installing a supported module.\n");
10231 		goto err_sw_init;
10232 	} else if (err) {
10233 		e_dev_err("HW Init failed: %d\n", err);
10234 		goto err_sw_init;
10235 	}
10236 
10237 #ifdef CONFIG_PCI_IOV
10238 	/* SR-IOV not supported on the 82598 */
10239 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
10240 		goto skip_sriov;
10241 	/* Mailbox */
10242 	ixgbe_init_mbx_params_pf(hw);
10243 	hw->mbx.ops = ii->mbx_ops;
10244 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
10245 	ixgbe_enable_sriov(adapter, max_vfs);
10246 skip_sriov:
10247 
10248 #endif
10249 	netdev->features = NETIF_F_SG |
10250 			   NETIF_F_TSO |
10251 			   NETIF_F_TSO6 |
10252 			   NETIF_F_RXHASH |
10253 			   NETIF_F_RXCSUM |
10254 			   NETIF_F_HW_CSUM;
10255 
10256 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
10257 				    NETIF_F_GSO_GRE_CSUM | \
10258 				    NETIF_F_GSO_IPXIP4 | \
10259 				    NETIF_F_GSO_IPXIP6 | \
10260 				    NETIF_F_GSO_UDP_TUNNEL | \
10261 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
10262 
10263 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
10264 	netdev->features |= NETIF_F_GSO_PARTIAL |
10265 			    IXGBE_GSO_PARTIAL_FEATURES;
10266 
10267 	if (hw->mac.type >= ixgbe_mac_82599EB)
10268 		netdev->features |= NETIF_F_SCTP_CRC;
10269 
10270 	/* copy netdev features into list of user selectable features */
10271 	netdev->hw_features |= netdev->features |
10272 			       NETIF_F_HW_VLAN_CTAG_FILTER |
10273 			       NETIF_F_HW_VLAN_CTAG_RX |
10274 			       NETIF_F_HW_VLAN_CTAG_TX |
10275 			       NETIF_F_RXALL |
10276 			       NETIF_F_HW_L2FW_DOFFLOAD;
10277 
10278 	if (hw->mac.type >= ixgbe_mac_82599EB)
10279 		netdev->hw_features |= NETIF_F_NTUPLE |
10280 				       NETIF_F_HW_TC;
10281 
10282 	if (pci_using_dac)
10283 		netdev->features |= NETIF_F_HIGHDMA;
10284 
10285 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
10286 	netdev->hw_enc_features |= netdev->vlan_features;
10287 	netdev->mpls_features |= NETIF_F_SG |
10288 				 NETIF_F_TSO |
10289 				 NETIF_F_TSO6 |
10290 				 NETIF_F_HW_CSUM;
10291 	netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
10292 
10293 	/* set this bit last since it cannot be part of vlan_features */
10294 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
10295 			    NETIF_F_HW_VLAN_CTAG_RX |
10296 			    NETIF_F_HW_VLAN_CTAG_TX;
10297 
10298 	netdev->priv_flags |= IFF_UNICAST_FLT;
10299 	netdev->priv_flags |= IFF_SUPP_NOFCS;
10300 
10301 	/* MTU range: 68 - 9710 */
10302 	netdev->min_mtu = ETH_MIN_MTU;
10303 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
10304 
10305 #ifdef CONFIG_IXGBE_DCB
10306 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
10307 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
10308 #endif
10309 
10310 #ifdef IXGBE_FCOE
10311 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
10312 		unsigned int fcoe_l;
10313 
10314 		if (hw->mac.ops.get_device_caps) {
10315 			hw->mac.ops.get_device_caps(hw, &device_caps);
10316 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
10317 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
10318 		}
10319 
10320 
10321 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
10322 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
10323 
10324 		netdev->features |= NETIF_F_FSO |
10325 				    NETIF_F_FCOE_CRC;
10326 
10327 		netdev->vlan_features |= NETIF_F_FSO |
10328 					 NETIF_F_FCOE_CRC |
10329 					 NETIF_F_FCOE_MTU;
10330 	}
10331 #endif /* IXGBE_FCOE */
10332 
10333 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
10334 		netdev->hw_features |= NETIF_F_LRO;
10335 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
10336 		netdev->features |= NETIF_F_LRO;
10337 
10338 	/* make sure the EEPROM is good */
10339 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
10340 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
10341 		err = -EIO;
10342 		goto err_sw_init;
10343 	}
10344 
10345 	eth_platform_get_mac_address(&adapter->pdev->dev,
10346 				     adapter->hw.mac.perm_addr);
10347 
10348 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
10349 
10350 	if (!is_valid_ether_addr(netdev->dev_addr)) {
10351 		e_dev_err("invalid MAC address\n");
10352 		err = -EIO;
10353 		goto err_sw_init;
10354 	}
10355 
10356 	/* Set hw->mac.addr to permanent MAC address */
10357 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
10358 	ixgbe_mac_set_default_filter(adapter);
10359 
10360 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
10361 		    (unsigned long) adapter);
10362 
10363 	if (ixgbe_removed(hw->hw_addr)) {
10364 		err = -EIO;
10365 		goto err_sw_init;
10366 	}
10367 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
10368 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
10369 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
10370 
10371 	err = ixgbe_init_interrupt_scheme(adapter);
10372 	if (err)
10373 		goto err_sw_init;
10374 
10375 	for (i = 0; i < adapter->num_rx_queues; i++)
10376 		u64_stats_init(&adapter->rx_ring[i]->syncp);
10377 	for (i = 0; i < adapter->num_tx_queues; i++)
10378 		u64_stats_init(&adapter->tx_ring[i]->syncp);
10379 	for (i = 0; i < adapter->num_xdp_queues; i++)
10380 		u64_stats_init(&adapter->xdp_ring[i]->syncp);
10381 
10382 	/* WOL not supported for all devices */
10383 	adapter->wol = 0;
10384 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
10385 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
10386 						pdev->subsystem_device);
10387 	if (hw->wol_enabled)
10388 		adapter->wol = IXGBE_WUFC_MAG;
10389 
10390 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
10391 
10392 	/* save off EEPROM version number */
10393 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
10394 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
10395 
10396 	/* pick up the PCI bus settings for reporting later */
10397 	if (ixgbe_pcie_from_parent(hw))
10398 		ixgbe_get_parent_bus_info(adapter);
10399 	else
10400 		 hw->mac.ops.get_bus_info(hw);
10401 
10402 	/* calculate the expected PCIe bandwidth required for optimal
10403 	 * performance. Note that some older parts will never have enough
10404 	 * bandwidth due to being older generation PCIe parts. We clamp these
10405 	 * parts to ensure no warning is displayed if it can't be fixed.
10406 	 */
10407 	switch (hw->mac.type) {
10408 	case ixgbe_mac_82598EB:
10409 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
10410 		break;
10411 	default:
10412 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
10413 		break;
10414 	}
10415 
10416 	/* don't check link if we failed to enumerate functions */
10417 	if (expected_gts > 0)
10418 		ixgbe_check_minimum_link(adapter, expected_gts);
10419 
10420 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
10421 	if (err)
10422 		strlcpy(part_str, "Unknown", sizeof(part_str));
10423 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
10424 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
10425 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
10426 			   part_str);
10427 	else
10428 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
10429 			   hw->mac.type, hw->phy.type, part_str);
10430 
10431 	e_dev_info("%pM\n", netdev->dev_addr);
10432 
10433 	/* reset the hardware with the new settings */
10434 	err = hw->mac.ops.start_hw(hw);
10435 	if (err == IXGBE_ERR_EEPROM_VERSION) {
10436 		/* We are running on a pre-production device, log a warning */
10437 		e_dev_warn("This device is a pre-production adapter/LOM. "
10438 			   "Please be aware there may be issues associated "
10439 			   "with your hardware.  If you are experiencing "
10440 			   "problems please contact your Intel or hardware "
10441 			   "representative who provided you with this "
10442 			   "hardware.\n");
10443 	}
10444 	strcpy(netdev->name, "eth%d");
10445 	pci_set_drvdata(pdev, adapter);
10446 	err = register_netdev(netdev);
10447 	if (err)
10448 		goto err_register;
10449 
10450 
10451 	/* power down the optics for 82599 SFP+ fiber */
10452 	if (hw->mac.ops.disable_tx_laser)
10453 		hw->mac.ops.disable_tx_laser(hw);
10454 
10455 	/* carrier off reporting is important to ethtool even BEFORE open */
10456 	netif_carrier_off(netdev);
10457 
10458 #ifdef CONFIG_IXGBE_DCA
10459 	if (dca_add_requester(&pdev->dev) == 0) {
10460 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
10461 		ixgbe_setup_dca(adapter);
10462 	}
10463 #endif
10464 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
10465 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
10466 		for (i = 0; i < adapter->num_vfs; i++)
10467 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
10468 	}
10469 
10470 	/* firmware requires driver version to be 0xFFFFFFFF
10471 	 * since os does not support feature
10472 	 */
10473 	if (hw->mac.ops.set_fw_drv_ver)
10474 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
10475 					   sizeof(ixgbe_driver_version) - 1,
10476 					   ixgbe_driver_version);
10477 
10478 	/* add san mac addr to netdev */
10479 	ixgbe_add_sanmac_netdev(netdev);
10480 
10481 	e_dev_info("%s\n", ixgbe_default_device_descr);
10482 
10483 #ifdef CONFIG_IXGBE_HWMON
10484 	if (ixgbe_sysfs_init(adapter))
10485 		e_err(probe, "failed to allocate sysfs resources\n");
10486 #endif /* CONFIG_IXGBE_HWMON */
10487 
10488 	ixgbe_dbg_adapter_init(adapter);
10489 
10490 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
10491 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
10492 		hw->mac.ops.setup_link(hw,
10493 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
10494 			true);
10495 
10496 	return 0;
10497 
10498 err_register:
10499 	ixgbe_release_hw_control(adapter);
10500 	ixgbe_clear_interrupt_scheme(adapter);
10501 err_sw_init:
10502 	ixgbe_disable_sriov(adapter);
10503 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
10504 	iounmap(adapter->io_addr);
10505 	kfree(adapter->jump_tables[0]);
10506 	kfree(adapter->mac_table);
10507 	kfree(adapter->rss_key);
10508 err_ioremap:
10509 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10510 	free_netdev(netdev);
10511 err_alloc_etherdev:
10512 	pci_release_mem_regions(pdev);
10513 err_pci_reg:
10514 err_dma:
10515 	if (!adapter || disable_dev)
10516 		pci_disable_device(pdev);
10517 	return err;
10518 }
10519 
10520 /**
10521  * ixgbe_remove - Device Removal Routine
10522  * @pdev: PCI device information struct
10523  *
10524  * ixgbe_remove is called by the PCI subsystem to alert the driver
10525  * that it should release a PCI device.  The could be caused by a
10526  * Hot-Plug event, or because the driver is going to be removed from
10527  * memory.
10528  **/
10529 static void ixgbe_remove(struct pci_dev *pdev)
10530 {
10531 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10532 	struct net_device *netdev;
10533 	bool disable_dev;
10534 	int i;
10535 
10536 	/* if !adapter then we already cleaned up in probe */
10537 	if (!adapter)
10538 		return;
10539 
10540 	netdev  = adapter->netdev;
10541 	ixgbe_dbg_adapter_exit(adapter);
10542 
10543 	set_bit(__IXGBE_REMOVING, &adapter->state);
10544 	cancel_work_sync(&adapter->service_task);
10545 
10546 
10547 #ifdef CONFIG_IXGBE_DCA
10548 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
10549 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
10550 		dca_remove_requester(&pdev->dev);
10551 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
10552 				IXGBE_DCA_CTRL_DCA_DISABLE);
10553 	}
10554 
10555 #endif
10556 #ifdef CONFIG_IXGBE_HWMON
10557 	ixgbe_sysfs_exit(adapter);
10558 #endif /* CONFIG_IXGBE_HWMON */
10559 
10560 	/* remove the added san mac */
10561 	ixgbe_del_sanmac_netdev(netdev);
10562 
10563 #ifdef CONFIG_PCI_IOV
10564 	ixgbe_disable_sriov(adapter);
10565 #endif
10566 	if (netdev->reg_state == NETREG_REGISTERED)
10567 		unregister_netdev(netdev);
10568 
10569 	ixgbe_clear_interrupt_scheme(adapter);
10570 
10571 	ixgbe_release_hw_control(adapter);
10572 
10573 #ifdef CONFIG_DCB
10574 	kfree(adapter->ixgbe_ieee_pfc);
10575 	kfree(adapter->ixgbe_ieee_ets);
10576 
10577 #endif
10578 	iounmap(adapter->io_addr);
10579 	pci_release_mem_regions(pdev);
10580 
10581 	e_dev_info("complete\n");
10582 
10583 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
10584 		if (adapter->jump_tables[i]) {
10585 			kfree(adapter->jump_tables[i]->input);
10586 			kfree(adapter->jump_tables[i]->mask);
10587 		}
10588 		kfree(adapter->jump_tables[i]);
10589 	}
10590 
10591 	kfree(adapter->mac_table);
10592 	kfree(adapter->rss_key);
10593 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10594 	free_netdev(netdev);
10595 
10596 	pci_disable_pcie_error_reporting(pdev);
10597 
10598 	if (disable_dev)
10599 		pci_disable_device(pdev);
10600 }
10601 
10602 /**
10603  * ixgbe_io_error_detected - called when PCI error is detected
10604  * @pdev: Pointer to PCI device
10605  * @state: The current pci connection state
10606  *
10607  * This function is called after a PCI bus error affecting
10608  * this device has been detected.
10609  */
10610 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
10611 						pci_channel_state_t state)
10612 {
10613 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10614 	struct net_device *netdev = adapter->netdev;
10615 
10616 #ifdef CONFIG_PCI_IOV
10617 	struct ixgbe_hw *hw = &adapter->hw;
10618 	struct pci_dev *bdev, *vfdev;
10619 	u32 dw0, dw1, dw2, dw3;
10620 	int vf, pos;
10621 	u16 req_id, pf_func;
10622 
10623 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
10624 	    adapter->num_vfs == 0)
10625 		goto skip_bad_vf_detection;
10626 
10627 	bdev = pdev->bus->self;
10628 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
10629 		bdev = bdev->bus->self;
10630 
10631 	if (!bdev)
10632 		goto skip_bad_vf_detection;
10633 
10634 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
10635 	if (!pos)
10636 		goto skip_bad_vf_detection;
10637 
10638 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
10639 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
10640 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
10641 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
10642 	if (ixgbe_removed(hw->hw_addr))
10643 		goto skip_bad_vf_detection;
10644 
10645 	req_id = dw1 >> 16;
10646 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
10647 	if (!(req_id & 0x0080))
10648 		goto skip_bad_vf_detection;
10649 
10650 	pf_func = req_id & 0x01;
10651 	if ((pf_func & 1) == (pdev->devfn & 1)) {
10652 		unsigned int device_id;
10653 
10654 		vf = (req_id & 0x7F) >> 1;
10655 		e_dev_err("VF %d has caused a PCIe error\n", vf);
10656 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
10657 				"%8.8x\tdw3: %8.8x\n",
10658 		dw0, dw1, dw2, dw3);
10659 		switch (adapter->hw.mac.type) {
10660 		case ixgbe_mac_82599EB:
10661 			device_id = IXGBE_82599_VF_DEVICE_ID;
10662 			break;
10663 		case ixgbe_mac_X540:
10664 			device_id = IXGBE_X540_VF_DEVICE_ID;
10665 			break;
10666 		case ixgbe_mac_X550:
10667 			device_id = IXGBE_DEV_ID_X550_VF;
10668 			break;
10669 		case ixgbe_mac_X550EM_x:
10670 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
10671 			break;
10672 		case ixgbe_mac_x550em_a:
10673 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
10674 			break;
10675 		default:
10676 			device_id = 0;
10677 			break;
10678 		}
10679 
10680 		/* Find the pci device of the offending VF */
10681 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
10682 		while (vfdev) {
10683 			if (vfdev->devfn == (req_id & 0xFF))
10684 				break;
10685 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
10686 					       device_id, vfdev);
10687 		}
10688 		/*
10689 		 * There's a slim chance the VF could have been hot plugged,
10690 		 * so if it is no longer present we don't need to issue the
10691 		 * VFLR.  Just clean up the AER in that case.
10692 		 */
10693 		if (vfdev) {
10694 			pcie_flr(vfdev);
10695 			/* Free device reference count */
10696 			pci_dev_put(vfdev);
10697 		}
10698 
10699 		pci_cleanup_aer_uncorrect_error_status(pdev);
10700 	}
10701 
10702 	/*
10703 	 * Even though the error may have occurred on the other port
10704 	 * we still need to increment the vf error reference count for
10705 	 * both ports because the I/O resume function will be called
10706 	 * for both of them.
10707 	 */
10708 	adapter->vferr_refcount++;
10709 
10710 	return PCI_ERS_RESULT_RECOVERED;
10711 
10712 skip_bad_vf_detection:
10713 #endif /* CONFIG_PCI_IOV */
10714 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
10715 		return PCI_ERS_RESULT_DISCONNECT;
10716 
10717 	rtnl_lock();
10718 	netif_device_detach(netdev);
10719 
10720 	if (state == pci_channel_io_perm_failure) {
10721 		rtnl_unlock();
10722 		return PCI_ERS_RESULT_DISCONNECT;
10723 	}
10724 
10725 	if (netif_running(netdev))
10726 		ixgbe_close_suspend(adapter);
10727 
10728 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
10729 		pci_disable_device(pdev);
10730 	rtnl_unlock();
10731 
10732 	/* Request a slot reset. */
10733 	return PCI_ERS_RESULT_NEED_RESET;
10734 }
10735 
10736 /**
10737  * ixgbe_io_slot_reset - called after the pci bus has been reset.
10738  * @pdev: Pointer to PCI device
10739  *
10740  * Restart the card from scratch, as if from a cold-boot.
10741  */
10742 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
10743 {
10744 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10745 	pci_ers_result_t result;
10746 	int err;
10747 
10748 	if (pci_enable_device_mem(pdev)) {
10749 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
10750 		result = PCI_ERS_RESULT_DISCONNECT;
10751 	} else {
10752 		smp_mb__before_atomic();
10753 		clear_bit(__IXGBE_DISABLED, &adapter->state);
10754 		adapter->hw.hw_addr = adapter->io_addr;
10755 		pci_set_master(pdev);
10756 		pci_restore_state(pdev);
10757 		pci_save_state(pdev);
10758 
10759 		pci_wake_from_d3(pdev, false);
10760 
10761 		ixgbe_reset(adapter);
10762 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10763 		result = PCI_ERS_RESULT_RECOVERED;
10764 	}
10765 
10766 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10767 	if (err) {
10768 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
10769 			  "failed 0x%0x\n", err);
10770 		/* non-fatal, continue */
10771 	}
10772 
10773 	return result;
10774 }
10775 
10776 /**
10777  * ixgbe_io_resume - called when traffic can start flowing again.
10778  * @pdev: Pointer to PCI device
10779  *
10780  * This callback is called when the error recovery driver tells us that
10781  * its OK to resume normal operation.
10782  */
10783 static void ixgbe_io_resume(struct pci_dev *pdev)
10784 {
10785 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10786 	struct net_device *netdev = adapter->netdev;
10787 
10788 #ifdef CONFIG_PCI_IOV
10789 	if (adapter->vferr_refcount) {
10790 		e_info(drv, "Resuming after VF err\n");
10791 		adapter->vferr_refcount--;
10792 		return;
10793 	}
10794 
10795 #endif
10796 	rtnl_lock();
10797 	if (netif_running(netdev))
10798 		ixgbe_open(netdev);
10799 
10800 	netif_device_attach(netdev);
10801 	rtnl_unlock();
10802 }
10803 
10804 static const struct pci_error_handlers ixgbe_err_handler = {
10805 	.error_detected = ixgbe_io_error_detected,
10806 	.slot_reset = ixgbe_io_slot_reset,
10807 	.resume = ixgbe_io_resume,
10808 };
10809 
10810 static struct pci_driver ixgbe_driver = {
10811 	.name     = ixgbe_driver_name,
10812 	.id_table = ixgbe_pci_tbl,
10813 	.probe    = ixgbe_probe,
10814 	.remove   = ixgbe_remove,
10815 #ifdef CONFIG_PM
10816 	.suspend  = ixgbe_suspend,
10817 	.resume   = ixgbe_resume,
10818 #endif
10819 	.shutdown = ixgbe_shutdown,
10820 	.sriov_configure = ixgbe_pci_sriov_configure,
10821 	.err_handler = &ixgbe_err_handler
10822 };
10823 
10824 /**
10825  * ixgbe_init_module - Driver Registration Routine
10826  *
10827  * ixgbe_init_module is the first routine called when the driver is
10828  * loaded. All it does is register with the PCI subsystem.
10829  **/
10830 static int __init ixgbe_init_module(void)
10831 {
10832 	int ret;
10833 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
10834 	pr_info("%s\n", ixgbe_copyright);
10835 
10836 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
10837 	if (!ixgbe_wq) {
10838 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
10839 		return -ENOMEM;
10840 	}
10841 
10842 	ixgbe_dbg_init();
10843 
10844 	ret = pci_register_driver(&ixgbe_driver);
10845 	if (ret) {
10846 		destroy_workqueue(ixgbe_wq);
10847 		ixgbe_dbg_exit();
10848 		return ret;
10849 	}
10850 
10851 #ifdef CONFIG_IXGBE_DCA
10852 	dca_register_notify(&dca_notifier);
10853 #endif
10854 
10855 	return 0;
10856 }
10857 
10858 module_init(ixgbe_init_module);
10859 
10860 /**
10861  * ixgbe_exit_module - Driver Exit Cleanup Routine
10862  *
10863  * ixgbe_exit_module is called just before the driver is removed
10864  * from memory.
10865  **/
10866 static void __exit ixgbe_exit_module(void)
10867 {
10868 #ifdef CONFIG_IXGBE_DCA
10869 	dca_unregister_notify(&dca_notifier);
10870 #endif
10871 	pci_unregister_driver(&ixgbe_driver);
10872 
10873 	ixgbe_dbg_exit();
10874 	if (ixgbe_wq) {
10875 		destroy_workqueue(ixgbe_wq);
10876 		ixgbe_wq = NULL;
10877 	}
10878 }
10879 
10880 #ifdef CONFIG_IXGBE_DCA
10881 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
10882 			    void *p)
10883 {
10884 	int ret_val;
10885 
10886 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
10887 					 __ixgbe_notify_dca);
10888 
10889 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
10890 }
10891 
10892 #endif /* CONFIG_IXGBE_DCA */
10893 
10894 module_exit(ixgbe_exit_module);
10895 
10896 /* ixgbe_main.c */
10897