xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision f412eed9dfdeeb6becd7de2ffe8b5d0a8b3f81ca)
1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2016 Intel Corporation.
5 
6   This program is free software; you can redistribute it and/or modify it
7   under the terms and conditions of the GNU General Public License,
8   version 2, as published by the Free Software Foundation.
9 
10   This program is distributed in the hope it will be useful, but WITHOUT
11   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13   more details.
14 
15   You should have received a copy of the GNU General Public License along with
16   this program; if not, write to the Free Software Foundation, Inc.,
17   51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 
19   The full GNU General Public License is included in this distribution in
20   the file called "COPYING".
21 
22   Contact Information:
23   Linux NICS <linux.nics@intel.com>
24   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 
27 *******************************************************************************/
28 
29 #include <linux/types.h>
30 #include <linux/module.h>
31 #include <linux/pci.h>
32 #include <linux/netdevice.h>
33 #include <linux/vmalloc.h>
34 #include <linux/string.h>
35 #include <linux/in.h>
36 #include <linux/interrupt.h>
37 #include <linux/ip.h>
38 #include <linux/tcp.h>
39 #include <linux/sctp.h>
40 #include <linux/pkt_sched.h>
41 #include <linux/ipv6.h>
42 #include <linux/slab.h>
43 #include <net/checksum.h>
44 #include <net/ip6_checksum.h>
45 #include <linux/etherdevice.h>
46 #include <linux/ethtool.h>
47 #include <linux/if.h>
48 #include <linux/if_vlan.h>
49 #include <linux/if_macvlan.h>
50 #include <linux/if_bridge.h>
51 #include <linux/prefetch.h>
52 #include <linux/bpf.h>
53 #include <linux/bpf_trace.h>
54 #include <linux/atomic.h>
55 #include <scsi/fc/fc_fcoe.h>
56 #include <net/udp_tunnel.h>
57 #include <net/pkt_cls.h>
58 #include <net/tc_act/tc_gact.h>
59 #include <net/tc_act/tc_mirred.h>
60 #include <net/vxlan.h>
61 #include <net/mpls.h>
62 
63 #include "ixgbe.h"
64 #include "ixgbe_common.h"
65 #include "ixgbe_dcb_82599.h"
66 #include "ixgbe_sriov.h"
67 #include "ixgbe_model.h"
68 
69 char ixgbe_driver_name[] = "ixgbe";
70 static const char ixgbe_driver_string[] =
71 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
72 #ifdef IXGBE_FCOE
73 char ixgbe_default_device_descr[] =
74 			      "Intel(R) 10 Gigabit Network Connection";
75 #else
76 static char ixgbe_default_device_descr[] =
77 			      "Intel(R) 10 Gigabit Network Connection";
78 #endif
79 #define DRV_VERSION "5.1.0-k"
80 const char ixgbe_driver_version[] = DRV_VERSION;
81 static const char ixgbe_copyright[] =
82 				"Copyright (c) 1999-2016 Intel Corporation.";
83 
84 static const char ixgbe_overheat_msg[] = "Network adapter has been stopped because it has over heated. Restart the computer. If the problem persists, power off the system and replace the adapter";
85 
86 static const struct ixgbe_info *ixgbe_info_tbl[] = {
87 	[board_82598]		= &ixgbe_82598_info,
88 	[board_82599]		= &ixgbe_82599_info,
89 	[board_X540]		= &ixgbe_X540_info,
90 	[board_X550]		= &ixgbe_X550_info,
91 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
92 	[board_x550em_x_fw]	= &ixgbe_x550em_x_fw_info,
93 	[board_x550em_a]	= &ixgbe_x550em_a_info,
94 	[board_x550em_a_fw]	= &ixgbe_x550em_a_fw_info,
95 };
96 
97 /* ixgbe_pci_tbl - PCI Device ID Table
98  *
99  * Wildcard entries (PCI_ANY_ID) should come last
100  * Last entry must be all 0s
101  *
102  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
103  *   Class, Class Mask, private data (not used) }
104  */
105 static const struct pci_device_id ixgbe_pci_tbl[] = {
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
133 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
134 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
135 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
136 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
137 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
138 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
139 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_XFI), board_X550EM_x},
140 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
141 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
142 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
143 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_1G_T), board_x550em_x_fw},
144 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
145 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
146 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
147 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
148 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
149 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
150 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
151 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
152 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
153 	/* required last entry */
154 	{0, }
155 };
156 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
157 
158 #ifdef CONFIG_IXGBE_DCA
159 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
160 			    void *p);
161 static struct notifier_block dca_notifier = {
162 	.notifier_call = ixgbe_notify_dca,
163 	.next          = NULL,
164 	.priority      = 0
165 };
166 #endif
167 
168 #ifdef CONFIG_PCI_IOV
169 static unsigned int max_vfs;
170 module_param(max_vfs, uint, 0);
171 MODULE_PARM_DESC(max_vfs,
172 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
173 #endif /* CONFIG_PCI_IOV */
174 
175 static unsigned int allow_unsupported_sfp;
176 module_param(allow_unsupported_sfp, uint, 0);
177 MODULE_PARM_DESC(allow_unsupported_sfp,
178 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
179 
180 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
181 static int debug = -1;
182 module_param(debug, int, 0);
183 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
184 
185 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
186 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
187 MODULE_LICENSE("GPL");
188 MODULE_VERSION(DRV_VERSION);
189 
190 static struct workqueue_struct *ixgbe_wq;
191 
192 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
193 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
194 
195 static const struct net_device_ops ixgbe_netdev_ops;
196 
197 static bool netif_is_ixgbe(struct net_device *dev)
198 {
199 	return dev && (dev->netdev_ops == &ixgbe_netdev_ops);
200 }
201 
202 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
203 					  u32 reg, u16 *value)
204 {
205 	struct pci_dev *parent_dev;
206 	struct pci_bus *parent_bus;
207 
208 	parent_bus = adapter->pdev->bus->parent;
209 	if (!parent_bus)
210 		return -1;
211 
212 	parent_dev = parent_bus->self;
213 	if (!parent_dev)
214 		return -1;
215 
216 	if (!pci_is_pcie(parent_dev))
217 		return -1;
218 
219 	pcie_capability_read_word(parent_dev, reg, value);
220 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
221 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
222 		return -1;
223 	return 0;
224 }
225 
226 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
227 {
228 	struct ixgbe_hw *hw = &adapter->hw;
229 	u16 link_status = 0;
230 	int err;
231 
232 	hw->bus.type = ixgbe_bus_type_pci_express;
233 
234 	/* Get the negotiated link width and speed from PCI config space of the
235 	 * parent, as this device is behind a switch
236 	 */
237 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
238 
239 	/* assume caller will handle error case */
240 	if (err)
241 		return err;
242 
243 	hw->bus.width = ixgbe_convert_bus_width(link_status);
244 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
245 
246 	return 0;
247 }
248 
249 /**
250  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
251  * @hw: hw specific details
252  *
253  * This function is used by probe to determine whether a device's PCI-Express
254  * bandwidth details should be gathered from the parent bus instead of from the
255  * device. Used to ensure that various locations all have the correct device ID
256  * checks.
257  */
258 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
259 {
260 	switch (hw->device_id) {
261 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
262 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
263 		return true;
264 	default:
265 		return false;
266 	}
267 }
268 
269 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
270 				     int expected_gts)
271 {
272 	struct ixgbe_hw *hw = &adapter->hw;
273 	int max_gts = 0;
274 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
275 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
276 	struct pci_dev *pdev;
277 
278 	/* Some devices are not connected over PCIe and thus do not negotiate
279 	 * speed. These devices do not have valid bus info, and thus any report
280 	 * we generate may not be correct.
281 	 */
282 	if (hw->bus.type == ixgbe_bus_type_internal)
283 		return;
284 
285 	/* determine whether to use the parent device */
286 	if (ixgbe_pcie_from_parent(&adapter->hw))
287 		pdev = adapter->pdev->bus->parent->self;
288 	else
289 		pdev = adapter->pdev;
290 
291 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
292 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
293 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
294 		return;
295 	}
296 
297 	switch (speed) {
298 	case PCIE_SPEED_2_5GT:
299 		/* 8b/10b encoding reduces max throughput by 20% */
300 		max_gts = 2 * width;
301 		break;
302 	case PCIE_SPEED_5_0GT:
303 		/* 8b/10b encoding reduces max throughput by 20% */
304 		max_gts = 4 * width;
305 		break;
306 	case PCIE_SPEED_8_0GT:
307 		/* 128b/130b encoding reduces throughput by less than 2% */
308 		max_gts = 8 * width;
309 		break;
310 	default:
311 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
312 		return;
313 	}
314 
315 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
316 		   max_gts);
317 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
318 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
319 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
320 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
321 		    "Unknown"),
322 		   width,
323 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
324 		    speed == PCIE_SPEED_5_0GT ? "20%" :
325 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
326 		    "Unknown"));
327 
328 	if (max_gts < expected_gts) {
329 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
330 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
331 			expected_gts);
332 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
333 	}
334 }
335 
336 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
337 {
338 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
339 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
340 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
341 		queue_work(ixgbe_wq, &adapter->service_task);
342 }
343 
344 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
345 {
346 	struct ixgbe_adapter *adapter = hw->back;
347 
348 	if (!hw->hw_addr)
349 		return;
350 	hw->hw_addr = NULL;
351 	e_dev_err("Adapter removed\n");
352 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
353 		ixgbe_service_event_schedule(adapter);
354 }
355 
356 static u32 ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
357 {
358 	u8 __iomem *reg_addr;
359 	u32 value;
360 	int i;
361 
362 	reg_addr = READ_ONCE(hw->hw_addr);
363 	if (ixgbe_removed(reg_addr))
364 		return IXGBE_FAILED_READ_REG;
365 
366 	/* Register read of 0xFFFFFFF can indicate the adapter has been removed,
367 	 * so perform several status register reads to determine if the adapter
368 	 * has been removed.
369 	 */
370 	for (i = 0; i < IXGBE_FAILED_READ_RETRIES; i++) {
371 		value = readl(reg_addr + IXGBE_STATUS);
372 		if (value != IXGBE_FAILED_READ_REG)
373 			break;
374 		mdelay(3);
375 	}
376 
377 	if (value == IXGBE_FAILED_READ_REG)
378 		ixgbe_remove_adapter(hw);
379 	else
380 		value = readl(reg_addr + reg);
381 	return value;
382 }
383 
384 /**
385  * ixgbe_read_reg - Read from device register
386  * @hw: hw specific details
387  * @reg: offset of register to read
388  *
389  * Returns : value read or IXGBE_FAILED_READ_REG if removed
390  *
391  * This function is used to read device registers. It checks for device
392  * removal by confirming any read that returns all ones by checking the
393  * status register value for all ones. This function avoids reading from
394  * the hardware if a removal was previously detected in which case it
395  * returns IXGBE_FAILED_READ_REG (all ones).
396  */
397 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
398 {
399 	u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
400 	u32 value;
401 
402 	if (ixgbe_removed(reg_addr))
403 		return IXGBE_FAILED_READ_REG;
404 	if (unlikely(hw->phy.nw_mng_if_sel &
405 		     IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
406 		struct ixgbe_adapter *adapter;
407 		int i;
408 
409 		for (i = 0; i < 200; ++i) {
410 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
411 			if (likely(!value))
412 				goto writes_completed;
413 			if (value == IXGBE_FAILED_READ_REG) {
414 				ixgbe_remove_adapter(hw);
415 				return IXGBE_FAILED_READ_REG;
416 			}
417 			udelay(5);
418 		}
419 
420 		adapter = hw->back;
421 		e_warn(hw, "register writes incomplete %08x\n", value);
422 	}
423 
424 writes_completed:
425 	value = readl(reg_addr + reg);
426 	if (unlikely(value == IXGBE_FAILED_READ_REG))
427 		value = ixgbe_check_remove(hw, reg);
428 	return value;
429 }
430 
431 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
432 {
433 	u16 value;
434 
435 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
436 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
437 		ixgbe_remove_adapter(hw);
438 		return true;
439 	}
440 	return false;
441 }
442 
443 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
444 {
445 	struct ixgbe_adapter *adapter = hw->back;
446 	u16 value;
447 
448 	if (ixgbe_removed(hw->hw_addr))
449 		return IXGBE_FAILED_READ_CFG_WORD;
450 	pci_read_config_word(adapter->pdev, reg, &value);
451 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
452 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
453 		return IXGBE_FAILED_READ_CFG_WORD;
454 	return value;
455 }
456 
457 #ifdef CONFIG_PCI_IOV
458 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
459 {
460 	struct ixgbe_adapter *adapter = hw->back;
461 	u32 value;
462 
463 	if (ixgbe_removed(hw->hw_addr))
464 		return IXGBE_FAILED_READ_CFG_DWORD;
465 	pci_read_config_dword(adapter->pdev, reg, &value);
466 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
467 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
468 		return IXGBE_FAILED_READ_CFG_DWORD;
469 	return value;
470 }
471 #endif /* CONFIG_PCI_IOV */
472 
473 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
474 {
475 	struct ixgbe_adapter *adapter = hw->back;
476 
477 	if (ixgbe_removed(hw->hw_addr))
478 		return;
479 	pci_write_config_word(adapter->pdev, reg, value);
480 }
481 
482 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
483 {
484 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
485 
486 	/* flush memory to make sure state is correct before next watchdog */
487 	smp_mb__before_atomic();
488 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
489 }
490 
491 struct ixgbe_reg_info {
492 	u32 ofs;
493 	char *name;
494 };
495 
496 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
497 
498 	/* General Registers */
499 	{IXGBE_CTRL, "CTRL"},
500 	{IXGBE_STATUS, "STATUS"},
501 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
502 
503 	/* Interrupt Registers */
504 	{IXGBE_EICR, "EICR"},
505 
506 	/* RX Registers */
507 	{IXGBE_SRRCTL(0), "SRRCTL"},
508 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
509 	{IXGBE_RDLEN(0), "RDLEN"},
510 	{IXGBE_RDH(0), "RDH"},
511 	{IXGBE_RDT(0), "RDT"},
512 	{IXGBE_RXDCTL(0), "RXDCTL"},
513 	{IXGBE_RDBAL(0), "RDBAL"},
514 	{IXGBE_RDBAH(0), "RDBAH"},
515 
516 	/* TX Registers */
517 	{IXGBE_TDBAL(0), "TDBAL"},
518 	{IXGBE_TDBAH(0), "TDBAH"},
519 	{IXGBE_TDLEN(0), "TDLEN"},
520 	{IXGBE_TDH(0), "TDH"},
521 	{IXGBE_TDT(0), "TDT"},
522 	{IXGBE_TXDCTL(0), "TXDCTL"},
523 
524 	/* List Terminator */
525 	{ .name = NULL }
526 };
527 
528 
529 /*
530  * ixgbe_regdump - register printout routine
531  */
532 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
533 {
534 	int i;
535 	char rname[16];
536 	u32 regs[64];
537 
538 	switch (reginfo->ofs) {
539 	case IXGBE_SRRCTL(0):
540 		for (i = 0; i < 64; i++)
541 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
542 		break;
543 	case IXGBE_DCA_RXCTRL(0):
544 		for (i = 0; i < 64; i++)
545 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
546 		break;
547 	case IXGBE_RDLEN(0):
548 		for (i = 0; i < 64; i++)
549 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
550 		break;
551 	case IXGBE_RDH(0):
552 		for (i = 0; i < 64; i++)
553 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
554 		break;
555 	case IXGBE_RDT(0):
556 		for (i = 0; i < 64; i++)
557 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
558 		break;
559 	case IXGBE_RXDCTL(0):
560 		for (i = 0; i < 64; i++)
561 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
562 		break;
563 	case IXGBE_RDBAL(0):
564 		for (i = 0; i < 64; i++)
565 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
566 		break;
567 	case IXGBE_RDBAH(0):
568 		for (i = 0; i < 64; i++)
569 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
570 		break;
571 	case IXGBE_TDBAL(0):
572 		for (i = 0; i < 64; i++)
573 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
574 		break;
575 	case IXGBE_TDBAH(0):
576 		for (i = 0; i < 64; i++)
577 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
578 		break;
579 	case IXGBE_TDLEN(0):
580 		for (i = 0; i < 64; i++)
581 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
582 		break;
583 	case IXGBE_TDH(0):
584 		for (i = 0; i < 64; i++)
585 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
586 		break;
587 	case IXGBE_TDT(0):
588 		for (i = 0; i < 64; i++)
589 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
590 		break;
591 	case IXGBE_TXDCTL(0):
592 		for (i = 0; i < 64; i++)
593 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
594 		break;
595 	default:
596 		pr_info("%-15s %08x\n",
597 			reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
598 		return;
599 	}
600 
601 	i = 0;
602 	while (i < 64) {
603 		int j;
604 		char buf[9 * 8 + 1];
605 		char *p = buf;
606 
607 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
608 		for (j = 0; j < 8; j++)
609 			p += sprintf(p, " %08x", regs[i++]);
610 		pr_err("%-15s%s\n", rname, buf);
611 	}
612 
613 }
614 
615 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
616 {
617 	struct ixgbe_tx_buffer *tx_buffer;
618 
619 	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
620 	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
621 		n, ring->next_to_use, ring->next_to_clean,
622 		(u64)dma_unmap_addr(tx_buffer, dma),
623 		dma_unmap_len(tx_buffer, len),
624 		tx_buffer->next_to_watch,
625 		(u64)tx_buffer->time_stamp);
626 }
627 
628 /*
629  * ixgbe_dump - Print registers, tx-rings and rx-rings
630  */
631 static void ixgbe_dump(struct ixgbe_adapter *adapter)
632 {
633 	struct net_device *netdev = adapter->netdev;
634 	struct ixgbe_hw *hw = &adapter->hw;
635 	struct ixgbe_reg_info *reginfo;
636 	int n = 0;
637 	struct ixgbe_ring *ring;
638 	struct ixgbe_tx_buffer *tx_buffer;
639 	union ixgbe_adv_tx_desc *tx_desc;
640 	struct my_u0 { u64 a; u64 b; } *u0;
641 	struct ixgbe_ring *rx_ring;
642 	union ixgbe_adv_rx_desc *rx_desc;
643 	struct ixgbe_rx_buffer *rx_buffer_info;
644 	int i = 0;
645 
646 	if (!netif_msg_hw(adapter))
647 		return;
648 
649 	/* Print netdevice Info */
650 	if (netdev) {
651 		dev_info(&adapter->pdev->dev, "Net device Info\n");
652 		pr_info("Device Name     state            "
653 			"trans_start\n");
654 		pr_info("%-15s %016lX %016lX\n",
655 			netdev->name,
656 			netdev->state,
657 			dev_trans_start(netdev));
658 	}
659 
660 	/* Print Registers */
661 	dev_info(&adapter->pdev->dev, "Register Dump\n");
662 	pr_info(" Register Name   Value\n");
663 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
664 	     reginfo->name; reginfo++) {
665 		ixgbe_regdump(hw, reginfo);
666 	}
667 
668 	/* Print TX Ring Summary */
669 	if (!netdev || !netif_running(netdev))
670 		return;
671 
672 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
673 	pr_info(" %s     %s              %s        %s\n",
674 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
675 		"leng", "ntw", "timestamp");
676 	for (n = 0; n < adapter->num_tx_queues; n++) {
677 		ring = adapter->tx_ring[n];
678 		ixgbe_print_buffer(ring, n);
679 	}
680 
681 	for (n = 0; n < adapter->num_xdp_queues; n++) {
682 		ring = adapter->xdp_ring[n];
683 		ixgbe_print_buffer(ring, n);
684 	}
685 
686 	/* Print TX Rings */
687 	if (!netif_msg_tx_done(adapter))
688 		goto rx_ring_summary;
689 
690 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
691 
692 	/* Transmit Descriptor Formats
693 	 *
694 	 * 82598 Advanced Transmit Descriptor
695 	 *   +--------------------------------------------------------------+
696 	 * 0 |         Buffer Address [63:0]                                |
697 	 *   +--------------------------------------------------------------+
698 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
699 	 *   +--------------------------------------------------------------+
700 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
701 	 *
702 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
703 	 *   +--------------------------------------------------------------+
704 	 * 0 |                          RSV [63:0]                          |
705 	 *   +--------------------------------------------------------------+
706 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
707 	 *   +--------------------------------------------------------------+
708 	 *   63                       36 35   32 31                         0
709 	 *
710 	 * 82599+ Advanced Transmit Descriptor
711 	 *   +--------------------------------------------------------------+
712 	 * 0 |         Buffer Address [63:0]                                |
713 	 *   +--------------------------------------------------------------+
714 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
715 	 *   +--------------------------------------------------------------+
716 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
717 	 *
718 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
719 	 *   +--------------------------------------------------------------+
720 	 * 0 |                          RSV [63:0]                          |
721 	 *   +--------------------------------------------------------------+
722 	 * 8 |            RSV           |  STA  |           RSV             |
723 	 *   +--------------------------------------------------------------+
724 	 *   63                       36 35   32 31                         0
725 	 */
726 
727 	for (n = 0; n < adapter->num_tx_queues; n++) {
728 		ring = adapter->tx_ring[n];
729 		pr_info("------------------------------------\n");
730 		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
731 		pr_info("------------------------------------\n");
732 		pr_info("%s%s    %s              %s        %s          %s\n",
733 			"T [desc]     [address 63:0  ] ",
734 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
735 			"leng", "ntw", "timestamp", "bi->skb");
736 
737 		for (i = 0; ring->desc && (i < ring->count); i++) {
738 			tx_desc = IXGBE_TX_DESC(ring, i);
739 			tx_buffer = &ring->tx_buffer_info[i];
740 			u0 = (struct my_u0 *)tx_desc;
741 			if (dma_unmap_len(tx_buffer, len) > 0) {
742 				const char *ring_desc;
743 
744 				if (i == ring->next_to_use &&
745 				    i == ring->next_to_clean)
746 					ring_desc = " NTC/U";
747 				else if (i == ring->next_to_use)
748 					ring_desc = " NTU";
749 				else if (i == ring->next_to_clean)
750 					ring_desc = " NTC";
751 				else
752 					ring_desc = "";
753 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
754 					i,
755 					le64_to_cpu(u0->a),
756 					le64_to_cpu(u0->b),
757 					(u64)dma_unmap_addr(tx_buffer, dma),
758 					dma_unmap_len(tx_buffer, len),
759 					tx_buffer->next_to_watch,
760 					(u64)tx_buffer->time_stamp,
761 					tx_buffer->skb,
762 					ring_desc);
763 
764 				if (netif_msg_pktdata(adapter) &&
765 				    tx_buffer->skb)
766 					print_hex_dump(KERN_INFO, "",
767 						DUMP_PREFIX_ADDRESS, 16, 1,
768 						tx_buffer->skb->data,
769 						dma_unmap_len(tx_buffer, len),
770 						true);
771 			}
772 		}
773 	}
774 
775 	/* Print RX Rings Summary */
776 rx_ring_summary:
777 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
778 	pr_info("Queue [NTU] [NTC]\n");
779 	for (n = 0; n < adapter->num_rx_queues; n++) {
780 		rx_ring = adapter->rx_ring[n];
781 		pr_info("%5d %5X %5X\n",
782 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
783 	}
784 
785 	/* Print RX Rings */
786 	if (!netif_msg_rx_status(adapter))
787 		return;
788 
789 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
790 
791 	/* Receive Descriptor Formats
792 	 *
793 	 * 82598 Advanced Receive Descriptor (Read) Format
794 	 *    63                                           1        0
795 	 *    +-----------------------------------------------------+
796 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
797 	 *    +----------------------------------------------+------+
798 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
799 	 *    +-----------------------------------------------------+
800 	 *
801 	 *
802 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
803 	 *
804 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
805 	 *   +------------------------------------------------------+
806 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
807 	 *   | Packet   | IP     |   |          |     | Type | Type |
808 	 *   | Checksum | Ident  |   |          |     |      |      |
809 	 *   +------------------------------------------------------+
810 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
811 	 *   +------------------------------------------------------+
812 	 *   63       48 47    32 31            20 19               0
813 	 *
814 	 * 82599+ Advanced Receive Descriptor (Read) Format
815 	 *    63                                           1        0
816 	 *    +-----------------------------------------------------+
817 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
818 	 *    +----------------------------------------------+------+
819 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
820 	 *    +-----------------------------------------------------+
821 	 *
822 	 *
823 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
824 	 *
825 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
826 	 *   +------------------------------------------------------+
827 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
828 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
829 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
830 	 *   +------------------------------------------------------+
831 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
832 	 *   +------------------------------------------------------+
833 	 *   63       48 47    32 31          20 19                 0
834 	 */
835 
836 	for (n = 0; n < adapter->num_rx_queues; n++) {
837 		rx_ring = adapter->rx_ring[n];
838 		pr_info("------------------------------------\n");
839 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
840 		pr_info("------------------------------------\n");
841 		pr_info("%s%s%s\n",
842 			"R  [desc]      [ PktBuf     A0] ",
843 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
844 			"<-- Adv Rx Read format");
845 		pr_info("%s%s%s\n",
846 			"RWB[desc]      [PcsmIpSHl PtRs] ",
847 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
848 			"<-- Adv Rx Write-Back format");
849 
850 		for (i = 0; i < rx_ring->count; i++) {
851 			const char *ring_desc;
852 
853 			if (i == rx_ring->next_to_use)
854 				ring_desc = " NTU";
855 			else if (i == rx_ring->next_to_clean)
856 				ring_desc = " NTC";
857 			else
858 				ring_desc = "";
859 
860 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
861 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
862 			u0 = (struct my_u0 *)rx_desc;
863 			if (rx_desc->wb.upper.length) {
864 				/* Descriptor Done */
865 				pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
866 					i,
867 					le64_to_cpu(u0->a),
868 					le64_to_cpu(u0->b),
869 					rx_buffer_info->skb,
870 					ring_desc);
871 			} else {
872 				pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
873 					i,
874 					le64_to_cpu(u0->a),
875 					le64_to_cpu(u0->b),
876 					(u64)rx_buffer_info->dma,
877 					rx_buffer_info->skb,
878 					ring_desc);
879 
880 				if (netif_msg_pktdata(adapter) &&
881 				    rx_buffer_info->dma) {
882 					print_hex_dump(KERN_INFO, "",
883 					   DUMP_PREFIX_ADDRESS, 16, 1,
884 					   page_address(rx_buffer_info->page) +
885 						    rx_buffer_info->page_offset,
886 					   ixgbe_rx_bufsz(rx_ring), true);
887 				}
888 			}
889 		}
890 	}
891 }
892 
893 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
894 {
895 	u32 ctrl_ext;
896 
897 	/* Let firmware take over control of h/w */
898 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
899 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
900 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
901 }
902 
903 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
904 {
905 	u32 ctrl_ext;
906 
907 	/* Let firmware know the driver has taken over */
908 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
909 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
910 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
911 }
912 
913 /**
914  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
915  * @adapter: pointer to adapter struct
916  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
917  * @queue: queue to map the corresponding interrupt to
918  * @msix_vector: the vector to map to the corresponding queue
919  *
920  */
921 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
922 			   u8 queue, u8 msix_vector)
923 {
924 	u32 ivar, index;
925 	struct ixgbe_hw *hw = &adapter->hw;
926 	switch (hw->mac.type) {
927 	case ixgbe_mac_82598EB:
928 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
929 		if (direction == -1)
930 			direction = 0;
931 		index = (((direction * 64) + queue) >> 2) & 0x1F;
932 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
933 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
934 		ivar |= (msix_vector << (8 * (queue & 0x3)));
935 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
936 		break;
937 	case ixgbe_mac_82599EB:
938 	case ixgbe_mac_X540:
939 	case ixgbe_mac_X550:
940 	case ixgbe_mac_X550EM_x:
941 	case ixgbe_mac_x550em_a:
942 		if (direction == -1) {
943 			/* other causes */
944 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
945 			index = ((queue & 1) * 8);
946 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
947 			ivar &= ~(0xFF << index);
948 			ivar |= (msix_vector << index);
949 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
950 			break;
951 		} else {
952 			/* tx or rx causes */
953 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
954 			index = ((16 * (queue & 1)) + (8 * direction));
955 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
956 			ivar &= ~(0xFF << index);
957 			ivar |= (msix_vector << index);
958 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
959 			break;
960 		}
961 	default:
962 		break;
963 	}
964 }
965 
966 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
967 					  u64 qmask)
968 {
969 	u32 mask;
970 
971 	switch (adapter->hw.mac.type) {
972 	case ixgbe_mac_82598EB:
973 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
974 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
975 		break;
976 	case ixgbe_mac_82599EB:
977 	case ixgbe_mac_X540:
978 	case ixgbe_mac_X550:
979 	case ixgbe_mac_X550EM_x:
980 	case ixgbe_mac_x550em_a:
981 		mask = (qmask & 0xFFFFFFFF);
982 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
983 		mask = (qmask >> 32);
984 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
985 		break;
986 	default:
987 		break;
988 	}
989 }
990 
991 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
992 {
993 	struct ixgbe_hw *hw = &adapter->hw;
994 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
995 	int i;
996 	u32 data;
997 
998 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
999 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
1000 		return;
1001 
1002 	switch (hw->mac.type) {
1003 	case ixgbe_mac_82598EB:
1004 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
1005 		break;
1006 	default:
1007 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
1008 	}
1009 	hwstats->lxoffrxc += data;
1010 
1011 	/* refill credits (no tx hang) if we received xoff */
1012 	if (!data)
1013 		return;
1014 
1015 	for (i = 0; i < adapter->num_tx_queues; i++)
1016 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
1017 			  &adapter->tx_ring[i]->state);
1018 
1019 	for (i = 0; i < adapter->num_xdp_queues; i++)
1020 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
1021 			  &adapter->xdp_ring[i]->state);
1022 }
1023 
1024 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
1025 {
1026 	struct ixgbe_hw *hw = &adapter->hw;
1027 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
1028 	u32 xoff[8] = {0};
1029 	u8 tc;
1030 	int i;
1031 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
1032 
1033 	if (adapter->ixgbe_ieee_pfc)
1034 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
1035 
1036 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
1037 		ixgbe_update_xoff_rx_lfc(adapter);
1038 		return;
1039 	}
1040 
1041 	/* update stats for each tc, only valid with PFC enabled */
1042 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
1043 		u32 pxoffrxc;
1044 
1045 		switch (hw->mac.type) {
1046 		case ixgbe_mac_82598EB:
1047 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1048 			break;
1049 		default:
1050 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1051 		}
1052 		hwstats->pxoffrxc[i] += pxoffrxc;
1053 		/* Get the TC for given UP */
1054 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1055 		xoff[tc] += pxoffrxc;
1056 	}
1057 
1058 	/* disarm tx queues that have received xoff frames */
1059 	for (i = 0; i < adapter->num_tx_queues; i++) {
1060 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1061 
1062 		tc = tx_ring->dcb_tc;
1063 		if (xoff[tc])
1064 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1065 	}
1066 
1067 	for (i = 0; i < adapter->num_xdp_queues; i++) {
1068 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1069 
1070 		tc = xdp_ring->dcb_tc;
1071 		if (xoff[tc])
1072 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1073 	}
1074 }
1075 
1076 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1077 {
1078 	return ring->stats.packets;
1079 }
1080 
1081 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1082 {
1083 	unsigned int head, tail;
1084 
1085 	head = ring->next_to_clean;
1086 	tail = ring->next_to_use;
1087 
1088 	return ((head <= tail) ? tail : tail + ring->count) - head;
1089 }
1090 
1091 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1092 {
1093 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1094 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1095 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1096 
1097 	clear_check_for_tx_hang(tx_ring);
1098 
1099 	/*
1100 	 * Check for a hung queue, but be thorough. This verifies
1101 	 * that a transmit has been completed since the previous
1102 	 * check AND there is at least one packet pending. The
1103 	 * ARMED bit is set to indicate a potential hang. The
1104 	 * bit is cleared if a pause frame is received to remove
1105 	 * false hang detection due to PFC or 802.3x frames. By
1106 	 * requiring this to fail twice we avoid races with
1107 	 * pfc clearing the ARMED bit and conditions where we
1108 	 * run the check_tx_hang logic with a transmit completion
1109 	 * pending but without time to complete it yet.
1110 	 */
1111 	if (tx_done_old == tx_done && tx_pending)
1112 		/* make sure it is true for two checks in a row */
1113 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1114 					&tx_ring->state);
1115 	/* update completed stats and continue */
1116 	tx_ring->tx_stats.tx_done_old = tx_done;
1117 	/* reset the countdown */
1118 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1119 
1120 	return false;
1121 }
1122 
1123 /**
1124  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1125  * @adapter: driver private struct
1126  **/
1127 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1128 {
1129 
1130 	/* Do the reset outside of interrupt context */
1131 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1132 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1133 		e_warn(drv, "initiating reset due to tx timeout\n");
1134 		ixgbe_service_event_schedule(adapter);
1135 	}
1136 }
1137 
1138 /**
1139  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1140  * @netdev: network interface device structure
1141  * @queue_index: Tx queue to set
1142  * @maxrate: desired maximum transmit bitrate
1143  **/
1144 static int ixgbe_tx_maxrate(struct net_device *netdev,
1145 			    int queue_index, u32 maxrate)
1146 {
1147 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1148 	struct ixgbe_hw *hw = &adapter->hw;
1149 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1150 
1151 	if (!maxrate)
1152 		return 0;
1153 
1154 	/* Calculate the rate factor values to set */
1155 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1156 	bcnrc_val /= maxrate;
1157 
1158 	/* clear everything but the rate factor */
1159 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1160 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1161 
1162 	/* enable the rate scheduler */
1163 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1164 
1165 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1166 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1167 
1168 	return 0;
1169 }
1170 
1171 /**
1172  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1173  * @q_vector: structure containing interrupt and ring information
1174  * @tx_ring: tx ring to clean
1175  * @napi_budget: Used to determine if we are in netpoll
1176  **/
1177 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1178 			       struct ixgbe_ring *tx_ring, int napi_budget)
1179 {
1180 	struct ixgbe_adapter *adapter = q_vector->adapter;
1181 	struct ixgbe_tx_buffer *tx_buffer;
1182 	union ixgbe_adv_tx_desc *tx_desc;
1183 	unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1184 	unsigned int budget = q_vector->tx.work_limit;
1185 	unsigned int i = tx_ring->next_to_clean;
1186 
1187 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1188 		return true;
1189 
1190 	tx_buffer = &tx_ring->tx_buffer_info[i];
1191 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1192 	i -= tx_ring->count;
1193 
1194 	do {
1195 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1196 
1197 		/* if next_to_watch is not set then there is no work pending */
1198 		if (!eop_desc)
1199 			break;
1200 
1201 		/* prevent any other reads prior to eop_desc */
1202 		smp_rmb();
1203 
1204 		/* if DD is not set pending work has not been completed */
1205 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1206 			break;
1207 
1208 		/* clear next_to_watch to prevent false hangs */
1209 		tx_buffer->next_to_watch = NULL;
1210 
1211 		/* update the statistics for this packet */
1212 		total_bytes += tx_buffer->bytecount;
1213 		total_packets += tx_buffer->gso_segs;
1214 		if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1215 			total_ipsec++;
1216 
1217 		/* free the skb */
1218 		if (ring_is_xdp(tx_ring))
1219 			xdp_return_frame(tx_buffer->xdpf);
1220 		else
1221 			napi_consume_skb(tx_buffer->skb, napi_budget);
1222 
1223 		/* unmap skb header data */
1224 		dma_unmap_single(tx_ring->dev,
1225 				 dma_unmap_addr(tx_buffer, dma),
1226 				 dma_unmap_len(tx_buffer, len),
1227 				 DMA_TO_DEVICE);
1228 
1229 		/* clear tx_buffer data */
1230 		dma_unmap_len_set(tx_buffer, len, 0);
1231 
1232 		/* unmap remaining buffers */
1233 		while (tx_desc != eop_desc) {
1234 			tx_buffer++;
1235 			tx_desc++;
1236 			i++;
1237 			if (unlikely(!i)) {
1238 				i -= tx_ring->count;
1239 				tx_buffer = tx_ring->tx_buffer_info;
1240 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1241 			}
1242 
1243 			/* unmap any remaining paged data */
1244 			if (dma_unmap_len(tx_buffer, len)) {
1245 				dma_unmap_page(tx_ring->dev,
1246 					       dma_unmap_addr(tx_buffer, dma),
1247 					       dma_unmap_len(tx_buffer, len),
1248 					       DMA_TO_DEVICE);
1249 				dma_unmap_len_set(tx_buffer, len, 0);
1250 			}
1251 		}
1252 
1253 		/* move us one more past the eop_desc for start of next pkt */
1254 		tx_buffer++;
1255 		tx_desc++;
1256 		i++;
1257 		if (unlikely(!i)) {
1258 			i -= tx_ring->count;
1259 			tx_buffer = tx_ring->tx_buffer_info;
1260 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1261 		}
1262 
1263 		/* issue prefetch for next Tx descriptor */
1264 		prefetch(tx_desc);
1265 
1266 		/* update budget accounting */
1267 		budget--;
1268 	} while (likely(budget));
1269 
1270 	i += tx_ring->count;
1271 	tx_ring->next_to_clean = i;
1272 	u64_stats_update_begin(&tx_ring->syncp);
1273 	tx_ring->stats.bytes += total_bytes;
1274 	tx_ring->stats.packets += total_packets;
1275 	u64_stats_update_end(&tx_ring->syncp);
1276 	q_vector->tx.total_bytes += total_bytes;
1277 	q_vector->tx.total_packets += total_packets;
1278 	adapter->tx_ipsec += total_ipsec;
1279 
1280 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1281 		/* schedule immediate reset if we believe we hung */
1282 		struct ixgbe_hw *hw = &adapter->hw;
1283 		e_err(drv, "Detected Tx Unit Hang %s\n"
1284 			"  Tx Queue             <%d>\n"
1285 			"  TDH, TDT             <%x>, <%x>\n"
1286 			"  next_to_use          <%x>\n"
1287 			"  next_to_clean        <%x>\n"
1288 			"tx_buffer_info[next_to_clean]\n"
1289 			"  time_stamp           <%lx>\n"
1290 			"  jiffies              <%lx>\n",
1291 			ring_is_xdp(tx_ring) ? "(XDP)" : "",
1292 			tx_ring->queue_index,
1293 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1294 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1295 			tx_ring->next_to_use, i,
1296 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1297 
1298 		if (!ring_is_xdp(tx_ring))
1299 			netif_stop_subqueue(tx_ring->netdev,
1300 					    tx_ring->queue_index);
1301 
1302 		e_info(probe,
1303 		       "tx hang %d detected on queue %d, resetting adapter\n",
1304 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1305 
1306 		/* schedule immediate reset if we believe we hung */
1307 		ixgbe_tx_timeout_reset(adapter);
1308 
1309 		/* the adapter is about to reset, no point in enabling stuff */
1310 		return true;
1311 	}
1312 
1313 	if (ring_is_xdp(tx_ring))
1314 		return !!budget;
1315 
1316 	netdev_tx_completed_queue(txring_txq(tx_ring),
1317 				  total_packets, total_bytes);
1318 
1319 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1320 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1321 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1322 		/* Make sure that anybody stopping the queue after this
1323 		 * sees the new next_to_clean.
1324 		 */
1325 		smp_mb();
1326 		if (__netif_subqueue_stopped(tx_ring->netdev,
1327 					     tx_ring->queue_index)
1328 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1329 			netif_wake_subqueue(tx_ring->netdev,
1330 					    tx_ring->queue_index);
1331 			++tx_ring->tx_stats.restart_queue;
1332 		}
1333 	}
1334 
1335 	return !!budget;
1336 }
1337 
1338 #ifdef CONFIG_IXGBE_DCA
1339 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1340 				struct ixgbe_ring *tx_ring,
1341 				int cpu)
1342 {
1343 	struct ixgbe_hw *hw = &adapter->hw;
1344 	u32 txctrl = 0;
1345 	u16 reg_offset;
1346 
1347 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1348 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1349 
1350 	switch (hw->mac.type) {
1351 	case ixgbe_mac_82598EB:
1352 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1353 		break;
1354 	case ixgbe_mac_82599EB:
1355 	case ixgbe_mac_X540:
1356 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1357 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1358 		break;
1359 	default:
1360 		/* for unknown hardware do not write register */
1361 		return;
1362 	}
1363 
1364 	/*
1365 	 * We can enable relaxed ordering for reads, but not writes when
1366 	 * DCA is enabled.  This is due to a known issue in some chipsets
1367 	 * which will cause the DCA tag to be cleared.
1368 	 */
1369 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1370 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1371 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1372 
1373 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1374 }
1375 
1376 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1377 				struct ixgbe_ring *rx_ring,
1378 				int cpu)
1379 {
1380 	struct ixgbe_hw *hw = &adapter->hw;
1381 	u32 rxctrl = 0;
1382 	u8 reg_idx = rx_ring->reg_idx;
1383 
1384 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1385 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1386 
1387 	switch (hw->mac.type) {
1388 	case ixgbe_mac_82599EB:
1389 	case ixgbe_mac_X540:
1390 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1391 		break;
1392 	default:
1393 		break;
1394 	}
1395 
1396 	/*
1397 	 * We can enable relaxed ordering for reads, but not writes when
1398 	 * DCA is enabled.  This is due to a known issue in some chipsets
1399 	 * which will cause the DCA tag to be cleared.
1400 	 */
1401 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1402 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1403 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1404 
1405 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1406 }
1407 
1408 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1409 {
1410 	struct ixgbe_adapter *adapter = q_vector->adapter;
1411 	struct ixgbe_ring *ring;
1412 	int cpu = get_cpu();
1413 
1414 	if (q_vector->cpu == cpu)
1415 		goto out_no_update;
1416 
1417 	ixgbe_for_each_ring(ring, q_vector->tx)
1418 		ixgbe_update_tx_dca(adapter, ring, cpu);
1419 
1420 	ixgbe_for_each_ring(ring, q_vector->rx)
1421 		ixgbe_update_rx_dca(adapter, ring, cpu);
1422 
1423 	q_vector->cpu = cpu;
1424 out_no_update:
1425 	put_cpu();
1426 }
1427 
1428 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1429 {
1430 	int i;
1431 
1432 	/* always use CB2 mode, difference is masked in the CB driver */
1433 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1434 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1435 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1436 	else
1437 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1438 				IXGBE_DCA_CTRL_DCA_DISABLE);
1439 
1440 	for (i = 0; i < adapter->num_q_vectors; i++) {
1441 		adapter->q_vector[i]->cpu = -1;
1442 		ixgbe_update_dca(adapter->q_vector[i]);
1443 	}
1444 }
1445 
1446 static int __ixgbe_notify_dca(struct device *dev, void *data)
1447 {
1448 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1449 	unsigned long event = *(unsigned long *)data;
1450 
1451 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1452 		return 0;
1453 
1454 	switch (event) {
1455 	case DCA_PROVIDER_ADD:
1456 		/* if we're already enabled, don't do it again */
1457 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1458 			break;
1459 		if (dca_add_requester(dev) == 0) {
1460 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1461 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1462 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1463 			break;
1464 		}
1465 		/* fall through - DCA is disabled. */
1466 	case DCA_PROVIDER_REMOVE:
1467 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1468 			dca_remove_requester(dev);
1469 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1470 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1471 					IXGBE_DCA_CTRL_DCA_DISABLE);
1472 		}
1473 		break;
1474 	}
1475 
1476 	return 0;
1477 }
1478 
1479 #endif /* CONFIG_IXGBE_DCA */
1480 
1481 #define IXGBE_RSS_L4_TYPES_MASK \
1482 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1483 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1484 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1485 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1486 
1487 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1488 				 union ixgbe_adv_rx_desc *rx_desc,
1489 				 struct sk_buff *skb)
1490 {
1491 	u16 rss_type;
1492 
1493 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1494 		return;
1495 
1496 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1497 		   IXGBE_RXDADV_RSSTYPE_MASK;
1498 
1499 	if (!rss_type)
1500 		return;
1501 
1502 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1503 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1504 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1505 }
1506 
1507 #ifdef IXGBE_FCOE
1508 /**
1509  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1510  * @ring: structure containing ring specific data
1511  * @rx_desc: advanced rx descriptor
1512  *
1513  * Returns : true if it is FCoE pkt
1514  */
1515 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1516 				    union ixgbe_adv_rx_desc *rx_desc)
1517 {
1518 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1519 
1520 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1521 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1522 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1523 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1524 }
1525 
1526 #endif /* IXGBE_FCOE */
1527 /**
1528  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1529  * @ring: structure containing ring specific data
1530  * @rx_desc: current Rx descriptor being processed
1531  * @skb: skb currently being received and modified
1532  **/
1533 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1534 				     union ixgbe_adv_rx_desc *rx_desc,
1535 				     struct sk_buff *skb)
1536 {
1537 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1538 	bool encap_pkt = false;
1539 
1540 	skb_checksum_none_assert(skb);
1541 
1542 	/* Rx csum disabled */
1543 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1544 		return;
1545 
1546 	/* check for VXLAN and Geneve packets */
1547 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1548 		encap_pkt = true;
1549 		skb->encapsulation = 1;
1550 	}
1551 
1552 	/* if IP and error */
1553 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1554 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1555 		ring->rx_stats.csum_err++;
1556 		return;
1557 	}
1558 
1559 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1560 		return;
1561 
1562 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1563 		/*
1564 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1565 		 * checksum errors.
1566 		 */
1567 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1568 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1569 			return;
1570 
1571 		ring->rx_stats.csum_err++;
1572 		return;
1573 	}
1574 
1575 	/* It must be a TCP or UDP packet with a valid checksum */
1576 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1577 	if (encap_pkt) {
1578 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1579 			return;
1580 
1581 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1582 			skb->ip_summed = CHECKSUM_NONE;
1583 			return;
1584 		}
1585 		/* If we checked the outer header let the stack know */
1586 		skb->csum_level = 1;
1587 	}
1588 }
1589 
1590 static inline unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1591 {
1592 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1593 }
1594 
1595 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1596 				    struct ixgbe_rx_buffer *bi)
1597 {
1598 	struct page *page = bi->page;
1599 	dma_addr_t dma;
1600 
1601 	/* since we are recycling buffers we should seldom need to alloc */
1602 	if (likely(page))
1603 		return true;
1604 
1605 	/* alloc new page for storage */
1606 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1607 	if (unlikely(!page)) {
1608 		rx_ring->rx_stats.alloc_rx_page_failed++;
1609 		return false;
1610 	}
1611 
1612 	/* map page for use */
1613 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1614 				 ixgbe_rx_pg_size(rx_ring),
1615 				 DMA_FROM_DEVICE,
1616 				 IXGBE_RX_DMA_ATTR);
1617 
1618 	/*
1619 	 * if mapping failed free memory back to system since
1620 	 * there isn't much point in holding memory we can't use
1621 	 */
1622 	if (dma_mapping_error(rx_ring->dev, dma)) {
1623 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1624 
1625 		rx_ring->rx_stats.alloc_rx_page_failed++;
1626 		return false;
1627 	}
1628 
1629 	bi->dma = dma;
1630 	bi->page = page;
1631 	bi->page_offset = ixgbe_rx_offset(rx_ring);
1632 	page_ref_add(page, USHRT_MAX - 1);
1633 	bi->pagecnt_bias = USHRT_MAX;
1634 	rx_ring->rx_stats.alloc_rx_page++;
1635 
1636 	return true;
1637 }
1638 
1639 /**
1640  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1641  * @rx_ring: ring to place buffers on
1642  * @cleaned_count: number of buffers to replace
1643  **/
1644 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1645 {
1646 	union ixgbe_adv_rx_desc *rx_desc;
1647 	struct ixgbe_rx_buffer *bi;
1648 	u16 i = rx_ring->next_to_use;
1649 	u16 bufsz;
1650 
1651 	/* nothing to do */
1652 	if (!cleaned_count)
1653 		return;
1654 
1655 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1656 	bi = &rx_ring->rx_buffer_info[i];
1657 	i -= rx_ring->count;
1658 
1659 	bufsz = ixgbe_rx_bufsz(rx_ring);
1660 
1661 	do {
1662 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1663 			break;
1664 
1665 		/* sync the buffer for use by the device */
1666 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1667 						 bi->page_offset, bufsz,
1668 						 DMA_FROM_DEVICE);
1669 
1670 		/*
1671 		 * Refresh the desc even if buffer_addrs didn't change
1672 		 * because each write-back erases this info.
1673 		 */
1674 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1675 
1676 		rx_desc++;
1677 		bi++;
1678 		i++;
1679 		if (unlikely(!i)) {
1680 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1681 			bi = rx_ring->rx_buffer_info;
1682 			i -= rx_ring->count;
1683 		}
1684 
1685 		/* clear the length for the next_to_use descriptor */
1686 		rx_desc->wb.upper.length = 0;
1687 
1688 		cleaned_count--;
1689 	} while (cleaned_count);
1690 
1691 	i += rx_ring->count;
1692 
1693 	if (rx_ring->next_to_use != i) {
1694 		rx_ring->next_to_use = i;
1695 
1696 		/* update next to alloc since we have filled the ring */
1697 		rx_ring->next_to_alloc = i;
1698 
1699 		/* Force memory writes to complete before letting h/w
1700 		 * know there are new descriptors to fetch.  (Only
1701 		 * applicable for weak-ordered memory model archs,
1702 		 * such as IA-64).
1703 		 */
1704 		wmb();
1705 		writel(i, rx_ring->tail);
1706 	}
1707 }
1708 
1709 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1710 				   struct sk_buff *skb)
1711 {
1712 	u16 hdr_len = skb_headlen(skb);
1713 
1714 	/* set gso_size to avoid messing up TCP MSS */
1715 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1716 						 IXGBE_CB(skb)->append_cnt);
1717 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1718 }
1719 
1720 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1721 				   struct sk_buff *skb)
1722 {
1723 	/* if append_cnt is 0 then frame is not RSC */
1724 	if (!IXGBE_CB(skb)->append_cnt)
1725 		return;
1726 
1727 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1728 	rx_ring->rx_stats.rsc_flush++;
1729 
1730 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1731 
1732 	/* gso_size is computed using append_cnt so always clear it last */
1733 	IXGBE_CB(skb)->append_cnt = 0;
1734 }
1735 
1736 /**
1737  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1738  * @rx_ring: rx descriptor ring packet is being transacted on
1739  * @rx_desc: pointer to the EOP Rx descriptor
1740  * @skb: pointer to current skb being populated
1741  *
1742  * This function checks the ring, descriptor, and packet information in
1743  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1744  * other fields within the skb.
1745  **/
1746 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1747 				     union ixgbe_adv_rx_desc *rx_desc,
1748 				     struct sk_buff *skb)
1749 {
1750 	struct net_device *dev = rx_ring->netdev;
1751 	u32 flags = rx_ring->q_vector->adapter->flags;
1752 
1753 	ixgbe_update_rsc_stats(rx_ring, skb);
1754 
1755 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1756 
1757 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1758 
1759 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1760 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1761 
1762 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1763 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1764 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1765 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1766 	}
1767 
1768 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1769 		ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1770 
1771 	/* record Rx queue, or update MACVLAN statistics */
1772 	if (netif_is_ixgbe(dev))
1773 		skb_record_rx_queue(skb, rx_ring->queue_index);
1774 	else
1775 		macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1776 				 false);
1777 
1778 	skb->protocol = eth_type_trans(skb, dev);
1779 }
1780 
1781 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1782 			 struct sk_buff *skb)
1783 {
1784 	napi_gro_receive(&q_vector->napi, skb);
1785 }
1786 
1787 /**
1788  * ixgbe_is_non_eop - process handling of non-EOP buffers
1789  * @rx_ring: Rx ring being processed
1790  * @rx_desc: Rx descriptor for current buffer
1791  * @skb: Current socket buffer containing buffer in progress
1792  *
1793  * This function updates next to clean.  If the buffer is an EOP buffer
1794  * this function exits returning false, otherwise it will place the
1795  * sk_buff in the next buffer to be chained and return true indicating
1796  * that this is in fact a non-EOP buffer.
1797  **/
1798 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1799 			     union ixgbe_adv_rx_desc *rx_desc,
1800 			     struct sk_buff *skb)
1801 {
1802 	u32 ntc = rx_ring->next_to_clean + 1;
1803 
1804 	/* fetch, update, and store next to clean */
1805 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1806 	rx_ring->next_to_clean = ntc;
1807 
1808 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1809 
1810 	/* update RSC append count if present */
1811 	if (ring_is_rsc_enabled(rx_ring)) {
1812 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1813 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1814 
1815 		if (unlikely(rsc_enabled)) {
1816 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1817 
1818 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1819 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1820 
1821 			/* update ntc based on RSC value */
1822 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1823 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1824 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1825 		}
1826 	}
1827 
1828 	/* if we are the last buffer then there is nothing else to do */
1829 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1830 		return false;
1831 
1832 	/* place skb in next buffer to be received */
1833 	rx_ring->rx_buffer_info[ntc].skb = skb;
1834 	rx_ring->rx_stats.non_eop_descs++;
1835 
1836 	return true;
1837 }
1838 
1839 /**
1840  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1841  * @rx_ring: rx descriptor ring packet is being transacted on
1842  * @skb: pointer to current skb being adjusted
1843  *
1844  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1845  * main difference between this version and the original function is that
1846  * this function can make several assumptions about the state of things
1847  * that allow for significant optimizations versus the standard function.
1848  * As a result we can do things like drop a frag and maintain an accurate
1849  * truesize for the skb.
1850  */
1851 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1852 			    struct sk_buff *skb)
1853 {
1854 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1855 	unsigned char *va;
1856 	unsigned int pull_len;
1857 
1858 	/*
1859 	 * it is valid to use page_address instead of kmap since we are
1860 	 * working with pages allocated out of the lomem pool per
1861 	 * alloc_page(GFP_ATOMIC)
1862 	 */
1863 	va = skb_frag_address(frag);
1864 
1865 	/*
1866 	 * we need the header to contain the greater of either ETH_HLEN or
1867 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1868 	 */
1869 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1870 
1871 	/* align pull length to size of long to optimize memcpy performance */
1872 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1873 
1874 	/* update all of the pointers */
1875 	skb_frag_size_sub(frag, pull_len);
1876 	frag->page_offset += pull_len;
1877 	skb->data_len -= pull_len;
1878 	skb->tail += pull_len;
1879 }
1880 
1881 /**
1882  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1883  * @rx_ring: rx descriptor ring packet is being transacted on
1884  * @skb: pointer to current skb being updated
1885  *
1886  * This function provides a basic DMA sync up for the first fragment of an
1887  * skb.  The reason for doing this is that the first fragment cannot be
1888  * unmapped until we have reached the end of packet descriptor for a buffer
1889  * chain.
1890  */
1891 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1892 				struct sk_buff *skb)
1893 {
1894 	/* if the page was released unmap it, else just sync our portion */
1895 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1896 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1897 				     ixgbe_rx_pg_size(rx_ring),
1898 				     DMA_FROM_DEVICE,
1899 				     IXGBE_RX_DMA_ATTR);
1900 	} else if (ring_uses_build_skb(rx_ring)) {
1901 		unsigned long offset = (unsigned long)(skb->data) & ~PAGE_MASK;
1902 
1903 		dma_sync_single_range_for_cpu(rx_ring->dev,
1904 					      IXGBE_CB(skb)->dma,
1905 					      offset,
1906 					      skb_headlen(skb),
1907 					      DMA_FROM_DEVICE);
1908 	} else {
1909 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1910 
1911 		dma_sync_single_range_for_cpu(rx_ring->dev,
1912 					      IXGBE_CB(skb)->dma,
1913 					      frag->page_offset,
1914 					      skb_frag_size(frag),
1915 					      DMA_FROM_DEVICE);
1916 	}
1917 }
1918 
1919 /**
1920  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1921  * @rx_ring: rx descriptor ring packet is being transacted on
1922  * @rx_desc: pointer to the EOP Rx descriptor
1923  * @skb: pointer to current skb being fixed
1924  *
1925  * Check if the skb is valid in the XDP case it will be an error pointer.
1926  * Return true in this case to abort processing and advance to next
1927  * descriptor.
1928  *
1929  * Check for corrupted packet headers caused by senders on the local L2
1930  * embedded NIC switch not setting up their Tx Descriptors right.  These
1931  * should be very rare.
1932  *
1933  * Also address the case where we are pulling data in on pages only
1934  * and as such no data is present in the skb header.
1935  *
1936  * In addition if skb is not at least 60 bytes we need to pad it so that
1937  * it is large enough to qualify as a valid Ethernet frame.
1938  *
1939  * Returns true if an error was encountered and skb was freed.
1940  **/
1941 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1942 				  union ixgbe_adv_rx_desc *rx_desc,
1943 				  struct sk_buff *skb)
1944 {
1945 	struct net_device *netdev = rx_ring->netdev;
1946 
1947 	/* XDP packets use error pointer so abort at this point */
1948 	if (IS_ERR(skb))
1949 		return true;
1950 
1951 	/* Verify netdev is present, and that packet does not have any
1952 	 * errors that would be unacceptable to the netdev.
1953 	 */
1954 	if (!netdev ||
1955 	    (unlikely(ixgbe_test_staterr(rx_desc,
1956 					 IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1957 	     !(netdev->features & NETIF_F_RXALL)))) {
1958 		dev_kfree_skb_any(skb);
1959 		return true;
1960 	}
1961 
1962 	/* place header in linear portion of buffer */
1963 	if (!skb_headlen(skb))
1964 		ixgbe_pull_tail(rx_ring, skb);
1965 
1966 #ifdef IXGBE_FCOE
1967 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1968 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1969 		return false;
1970 
1971 #endif
1972 	/* if eth_skb_pad returns an error the skb was freed */
1973 	if (eth_skb_pad(skb))
1974 		return true;
1975 
1976 	return false;
1977 }
1978 
1979 /**
1980  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1981  * @rx_ring: rx descriptor ring to store buffers on
1982  * @old_buff: donor buffer to have page reused
1983  *
1984  * Synchronizes page for reuse by the adapter
1985  **/
1986 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1987 				struct ixgbe_rx_buffer *old_buff)
1988 {
1989 	struct ixgbe_rx_buffer *new_buff;
1990 	u16 nta = rx_ring->next_to_alloc;
1991 
1992 	new_buff = &rx_ring->rx_buffer_info[nta];
1993 
1994 	/* update, and store next to alloc */
1995 	nta++;
1996 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1997 
1998 	/* Transfer page from old buffer to new buffer.
1999 	 * Move each member individually to avoid possible store
2000 	 * forwarding stalls and unnecessary copy of skb.
2001 	 */
2002 	new_buff->dma		= old_buff->dma;
2003 	new_buff->page		= old_buff->page;
2004 	new_buff->page_offset	= old_buff->page_offset;
2005 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
2006 }
2007 
2008 static inline bool ixgbe_page_is_reserved(struct page *page)
2009 {
2010 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
2011 }
2012 
2013 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer)
2014 {
2015 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
2016 	struct page *page = rx_buffer->page;
2017 
2018 	/* avoid re-using remote pages */
2019 	if (unlikely(ixgbe_page_is_reserved(page)))
2020 		return false;
2021 
2022 #if (PAGE_SIZE < 8192)
2023 	/* if we are only owner of page we can reuse it */
2024 	if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
2025 		return false;
2026 #else
2027 	/* The last offset is a bit aggressive in that we assume the
2028 	 * worst case of FCoE being enabled and using a 3K buffer.
2029 	 * However this should have minimal impact as the 1K extra is
2030 	 * still less than one buffer in size.
2031 	 */
2032 #define IXGBE_LAST_OFFSET \
2033 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
2034 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
2035 		return false;
2036 #endif
2037 
2038 	/* If we have drained the page fragment pool we need to update
2039 	 * the pagecnt_bias and page count so that we fully restock the
2040 	 * number of references the driver holds.
2041 	 */
2042 	if (unlikely(pagecnt_bias == 1)) {
2043 		page_ref_add(page, USHRT_MAX - 1);
2044 		rx_buffer->pagecnt_bias = USHRT_MAX;
2045 	}
2046 
2047 	return true;
2048 }
2049 
2050 /**
2051  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
2052  * @rx_ring: rx descriptor ring to transact packets on
2053  * @rx_buffer: buffer containing page to add
2054  * @skb: sk_buff to place the data into
2055  * @size: size of data in rx_buffer
2056  *
2057  * This function will add the data contained in rx_buffer->page to the skb.
2058  * This is done either through a direct copy if the data in the buffer is
2059  * less than the skb header size, otherwise it will just attach the page as
2060  * a frag to the skb.
2061  *
2062  * The function will then update the page offset if necessary and return
2063  * true if the buffer can be reused by the adapter.
2064  **/
2065 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
2066 			      struct ixgbe_rx_buffer *rx_buffer,
2067 			      struct sk_buff *skb,
2068 			      unsigned int size)
2069 {
2070 #if (PAGE_SIZE < 8192)
2071 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2072 #else
2073 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2074 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2075 				SKB_DATA_ALIGN(size);
2076 #endif
2077 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2078 			rx_buffer->page_offset, size, truesize);
2079 #if (PAGE_SIZE < 8192)
2080 	rx_buffer->page_offset ^= truesize;
2081 #else
2082 	rx_buffer->page_offset += truesize;
2083 #endif
2084 }
2085 
2086 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2087 						   union ixgbe_adv_rx_desc *rx_desc,
2088 						   struct sk_buff **skb,
2089 						   const unsigned int size)
2090 {
2091 	struct ixgbe_rx_buffer *rx_buffer;
2092 
2093 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2094 	prefetchw(rx_buffer->page);
2095 	*skb = rx_buffer->skb;
2096 
2097 	/* Delay unmapping of the first packet. It carries the header
2098 	 * information, HW may still access the header after the writeback.
2099 	 * Only unmap it when EOP is reached
2100 	 */
2101 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2102 		if (!*skb)
2103 			goto skip_sync;
2104 	} else {
2105 		if (*skb)
2106 			ixgbe_dma_sync_frag(rx_ring, *skb);
2107 	}
2108 
2109 	/* we are reusing so sync this buffer for CPU use */
2110 	dma_sync_single_range_for_cpu(rx_ring->dev,
2111 				      rx_buffer->dma,
2112 				      rx_buffer->page_offset,
2113 				      size,
2114 				      DMA_FROM_DEVICE);
2115 skip_sync:
2116 	rx_buffer->pagecnt_bias--;
2117 
2118 	return rx_buffer;
2119 }
2120 
2121 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2122 				struct ixgbe_rx_buffer *rx_buffer,
2123 				struct sk_buff *skb)
2124 {
2125 	if (ixgbe_can_reuse_rx_page(rx_buffer)) {
2126 		/* hand second half of page back to the ring */
2127 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2128 	} else {
2129 		if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2130 			/* the page has been released from the ring */
2131 			IXGBE_CB(skb)->page_released = true;
2132 		} else {
2133 			/* we are not reusing the buffer so unmap it */
2134 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2135 					     ixgbe_rx_pg_size(rx_ring),
2136 					     DMA_FROM_DEVICE,
2137 					     IXGBE_RX_DMA_ATTR);
2138 		}
2139 		__page_frag_cache_drain(rx_buffer->page,
2140 					rx_buffer->pagecnt_bias);
2141 	}
2142 
2143 	/* clear contents of rx_buffer */
2144 	rx_buffer->page = NULL;
2145 	rx_buffer->skb = NULL;
2146 }
2147 
2148 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2149 					   struct ixgbe_rx_buffer *rx_buffer,
2150 					   struct xdp_buff *xdp,
2151 					   union ixgbe_adv_rx_desc *rx_desc)
2152 {
2153 	unsigned int size = xdp->data_end - xdp->data;
2154 #if (PAGE_SIZE < 8192)
2155 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2156 #else
2157 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2158 					       xdp->data_hard_start);
2159 #endif
2160 	struct sk_buff *skb;
2161 
2162 	/* prefetch first cache line of first page */
2163 	prefetch(xdp->data);
2164 #if L1_CACHE_BYTES < 128
2165 	prefetch(xdp->data + L1_CACHE_BYTES);
2166 #endif
2167 	/* Note, we get here by enabling legacy-rx via:
2168 	 *
2169 	 *    ethtool --set-priv-flags <dev> legacy-rx on
2170 	 *
2171 	 * In this mode, we currently get 0 extra XDP headroom as
2172 	 * opposed to having legacy-rx off, where we process XDP
2173 	 * packets going to stack via ixgbe_build_skb(). The latter
2174 	 * provides us currently with 192 bytes of headroom.
2175 	 *
2176 	 * For ixgbe_construct_skb() mode it means that the
2177 	 * xdp->data_meta will always point to xdp->data, since
2178 	 * the helper cannot expand the head. Should this ever
2179 	 * change in future for legacy-rx mode on, then lets also
2180 	 * add xdp->data_meta handling here.
2181 	 */
2182 
2183 	/* allocate a skb to store the frags */
2184 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2185 	if (unlikely(!skb))
2186 		return NULL;
2187 
2188 	if (size > IXGBE_RX_HDR_SIZE) {
2189 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2190 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2191 
2192 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2193 				xdp->data - page_address(rx_buffer->page),
2194 				size, truesize);
2195 #if (PAGE_SIZE < 8192)
2196 		rx_buffer->page_offset ^= truesize;
2197 #else
2198 		rx_buffer->page_offset += truesize;
2199 #endif
2200 	} else {
2201 		memcpy(__skb_put(skb, size),
2202 		       xdp->data, ALIGN(size, sizeof(long)));
2203 		rx_buffer->pagecnt_bias++;
2204 	}
2205 
2206 	return skb;
2207 }
2208 
2209 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2210 				       struct ixgbe_rx_buffer *rx_buffer,
2211 				       struct xdp_buff *xdp,
2212 				       union ixgbe_adv_rx_desc *rx_desc)
2213 {
2214 	unsigned int metasize = xdp->data - xdp->data_meta;
2215 #if (PAGE_SIZE < 8192)
2216 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2217 #else
2218 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2219 				SKB_DATA_ALIGN(xdp->data_end -
2220 					       xdp->data_hard_start);
2221 #endif
2222 	struct sk_buff *skb;
2223 
2224 	/* Prefetch first cache line of first page. If xdp->data_meta
2225 	 * is unused, this points extactly as xdp->data, otherwise we
2226 	 * likely have a consumer accessing first few bytes of meta
2227 	 * data, and then actual data.
2228 	 */
2229 	prefetch(xdp->data_meta);
2230 #if L1_CACHE_BYTES < 128
2231 	prefetch(xdp->data_meta + L1_CACHE_BYTES);
2232 #endif
2233 
2234 	/* build an skb to around the page buffer */
2235 	skb = build_skb(xdp->data_hard_start, truesize);
2236 	if (unlikely(!skb))
2237 		return NULL;
2238 
2239 	/* update pointers within the skb to store the data */
2240 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
2241 	__skb_put(skb, xdp->data_end - xdp->data);
2242 	if (metasize)
2243 		skb_metadata_set(skb, metasize);
2244 
2245 	/* record DMA address if this is the start of a chain of buffers */
2246 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2247 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2248 
2249 	/* update buffer offset */
2250 #if (PAGE_SIZE < 8192)
2251 	rx_buffer->page_offset ^= truesize;
2252 #else
2253 	rx_buffer->page_offset += truesize;
2254 #endif
2255 
2256 	return skb;
2257 }
2258 
2259 #define IXGBE_XDP_PASS 0
2260 #define IXGBE_XDP_CONSUMED 1
2261 #define IXGBE_XDP_TX 2
2262 
2263 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
2264 			       struct xdp_frame *xdpf);
2265 
2266 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2267 				     struct ixgbe_ring *rx_ring,
2268 				     struct xdp_buff *xdp)
2269 {
2270 	int err, result = IXGBE_XDP_PASS;
2271 	struct bpf_prog *xdp_prog;
2272 	struct xdp_frame *xdpf;
2273 	u32 act;
2274 
2275 	rcu_read_lock();
2276 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2277 
2278 	if (!xdp_prog)
2279 		goto xdp_out;
2280 
2281 	prefetchw(xdp->data_hard_start); /* xdp_frame write */
2282 
2283 	act = bpf_prog_run_xdp(xdp_prog, xdp);
2284 	switch (act) {
2285 	case XDP_PASS:
2286 		break;
2287 	case XDP_TX:
2288 		xdpf = convert_to_xdp_frame(xdp);
2289 		if (unlikely(!xdpf)) {
2290 			result = IXGBE_XDP_CONSUMED;
2291 			break;
2292 		}
2293 		result = ixgbe_xmit_xdp_ring(adapter, xdpf);
2294 		break;
2295 	case XDP_REDIRECT:
2296 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2297 		if (!err)
2298 			result = IXGBE_XDP_TX;
2299 		else
2300 			result = IXGBE_XDP_CONSUMED;
2301 		break;
2302 	default:
2303 		bpf_warn_invalid_xdp_action(act);
2304 		/* fallthrough */
2305 	case XDP_ABORTED:
2306 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2307 		/* fallthrough -- handle aborts by dropping packet */
2308 	case XDP_DROP:
2309 		result = IXGBE_XDP_CONSUMED;
2310 		break;
2311 	}
2312 xdp_out:
2313 	rcu_read_unlock();
2314 	return ERR_PTR(-result);
2315 }
2316 
2317 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2318 				 struct ixgbe_rx_buffer *rx_buffer,
2319 				 unsigned int size)
2320 {
2321 #if (PAGE_SIZE < 8192)
2322 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2323 
2324 	rx_buffer->page_offset ^= truesize;
2325 #else
2326 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2327 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2328 				SKB_DATA_ALIGN(size);
2329 
2330 	rx_buffer->page_offset += truesize;
2331 #endif
2332 }
2333 
2334 /**
2335  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2336  * @q_vector: structure containing interrupt and ring information
2337  * @rx_ring: rx descriptor ring to transact packets on
2338  * @budget: Total limit on number of packets to process
2339  *
2340  * This function provides a "bounce buffer" approach to Rx interrupt
2341  * processing.  The advantage to this is that on systems that have
2342  * expensive overhead for IOMMU access this provides a means of avoiding
2343  * it by maintaining the mapping of the page to the syste.
2344  *
2345  * Returns amount of work completed
2346  **/
2347 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2348 			       struct ixgbe_ring *rx_ring,
2349 			       const int budget)
2350 {
2351 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2352 	struct ixgbe_adapter *adapter = q_vector->adapter;
2353 #ifdef IXGBE_FCOE
2354 	int ddp_bytes;
2355 	unsigned int mss = 0;
2356 #endif /* IXGBE_FCOE */
2357 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2358 	bool xdp_xmit = false;
2359 	struct xdp_buff xdp;
2360 
2361 	xdp.rxq = &rx_ring->xdp_rxq;
2362 
2363 	while (likely(total_rx_packets < budget)) {
2364 		union ixgbe_adv_rx_desc *rx_desc;
2365 		struct ixgbe_rx_buffer *rx_buffer;
2366 		struct sk_buff *skb;
2367 		unsigned int size;
2368 
2369 		/* return some buffers to hardware, one at a time is too slow */
2370 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2371 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2372 			cleaned_count = 0;
2373 		}
2374 
2375 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2376 		size = le16_to_cpu(rx_desc->wb.upper.length);
2377 		if (!size)
2378 			break;
2379 
2380 		/* This memory barrier is needed to keep us from reading
2381 		 * any other fields out of the rx_desc until we know the
2382 		 * descriptor has been written back
2383 		 */
2384 		dma_rmb();
2385 
2386 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size);
2387 
2388 		/* retrieve a buffer from the ring */
2389 		if (!skb) {
2390 			xdp.data = page_address(rx_buffer->page) +
2391 				   rx_buffer->page_offset;
2392 			xdp.data_meta = xdp.data;
2393 			xdp.data_hard_start = xdp.data -
2394 					      ixgbe_rx_offset(rx_ring);
2395 			xdp.data_end = xdp.data + size;
2396 
2397 			skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2398 		}
2399 
2400 		if (IS_ERR(skb)) {
2401 			if (PTR_ERR(skb) == -IXGBE_XDP_TX) {
2402 				xdp_xmit = true;
2403 				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2404 			} else {
2405 				rx_buffer->pagecnt_bias++;
2406 			}
2407 			total_rx_packets++;
2408 			total_rx_bytes += size;
2409 		} else if (skb) {
2410 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2411 		} else if (ring_uses_build_skb(rx_ring)) {
2412 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2413 					      &xdp, rx_desc);
2414 		} else {
2415 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2416 						  &xdp, rx_desc);
2417 		}
2418 
2419 		/* exit if we failed to retrieve a buffer */
2420 		if (!skb) {
2421 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2422 			rx_buffer->pagecnt_bias++;
2423 			break;
2424 		}
2425 
2426 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb);
2427 		cleaned_count++;
2428 
2429 		/* place incomplete frames back on ring for completion */
2430 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2431 			continue;
2432 
2433 		/* verify the packet layout is correct */
2434 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2435 			continue;
2436 
2437 		/* probably a little skewed due to removing CRC */
2438 		total_rx_bytes += skb->len;
2439 
2440 		/* populate checksum, timestamp, VLAN, and protocol */
2441 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2442 
2443 #ifdef IXGBE_FCOE
2444 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2445 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2446 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2447 			/* include DDPed FCoE data */
2448 			if (ddp_bytes > 0) {
2449 				if (!mss) {
2450 					mss = rx_ring->netdev->mtu -
2451 						sizeof(struct fcoe_hdr) -
2452 						sizeof(struct fc_frame_header) -
2453 						sizeof(struct fcoe_crc_eof);
2454 					if (mss > 512)
2455 						mss &= ~511;
2456 				}
2457 				total_rx_bytes += ddp_bytes;
2458 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2459 								 mss);
2460 			}
2461 			if (!ddp_bytes) {
2462 				dev_kfree_skb_any(skb);
2463 				continue;
2464 			}
2465 		}
2466 
2467 #endif /* IXGBE_FCOE */
2468 		ixgbe_rx_skb(q_vector, skb);
2469 
2470 		/* update budget accounting */
2471 		total_rx_packets++;
2472 	}
2473 
2474 	if (xdp_xmit) {
2475 		struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
2476 
2477 		/* Force memory writes to complete before letting h/w
2478 		 * know there are new descriptors to fetch.
2479 		 */
2480 		wmb();
2481 		writel(ring->next_to_use, ring->tail);
2482 
2483 		xdp_do_flush_map();
2484 	}
2485 
2486 	u64_stats_update_begin(&rx_ring->syncp);
2487 	rx_ring->stats.packets += total_rx_packets;
2488 	rx_ring->stats.bytes += total_rx_bytes;
2489 	u64_stats_update_end(&rx_ring->syncp);
2490 	q_vector->rx.total_packets += total_rx_packets;
2491 	q_vector->rx.total_bytes += total_rx_bytes;
2492 
2493 	return total_rx_packets;
2494 }
2495 
2496 /**
2497  * ixgbe_configure_msix - Configure MSI-X hardware
2498  * @adapter: board private structure
2499  *
2500  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2501  * interrupts.
2502  **/
2503 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2504 {
2505 	struct ixgbe_q_vector *q_vector;
2506 	int v_idx;
2507 	u32 mask;
2508 
2509 	/* Populate MSIX to EITR Select */
2510 	if (adapter->num_vfs > 32) {
2511 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2512 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2513 	}
2514 
2515 	/*
2516 	 * Populate the IVAR table and set the ITR values to the
2517 	 * corresponding register.
2518 	 */
2519 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2520 		struct ixgbe_ring *ring;
2521 		q_vector = adapter->q_vector[v_idx];
2522 
2523 		ixgbe_for_each_ring(ring, q_vector->rx)
2524 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2525 
2526 		ixgbe_for_each_ring(ring, q_vector->tx)
2527 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2528 
2529 		ixgbe_write_eitr(q_vector);
2530 	}
2531 
2532 	switch (adapter->hw.mac.type) {
2533 	case ixgbe_mac_82598EB:
2534 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2535 			       v_idx);
2536 		break;
2537 	case ixgbe_mac_82599EB:
2538 	case ixgbe_mac_X540:
2539 	case ixgbe_mac_X550:
2540 	case ixgbe_mac_X550EM_x:
2541 	case ixgbe_mac_x550em_a:
2542 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2543 		break;
2544 	default:
2545 		break;
2546 	}
2547 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2548 
2549 	/* set up to autoclear timer, and the vectors */
2550 	mask = IXGBE_EIMS_ENABLE_MASK;
2551 	mask &= ~(IXGBE_EIMS_OTHER |
2552 		  IXGBE_EIMS_MAILBOX |
2553 		  IXGBE_EIMS_LSC);
2554 
2555 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2556 }
2557 
2558 /**
2559  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2560  * @q_vector: structure containing interrupt and ring information
2561  * @ring_container: structure containing ring performance data
2562  *
2563  *      Stores a new ITR value based on packets and byte
2564  *      counts during the last interrupt.  The advantage of per interrupt
2565  *      computation is faster updates and more accurate ITR for the current
2566  *      traffic pattern.  Constants in this function were computed
2567  *      based on theoretical maximum wire speed and thresholds were set based
2568  *      on testing data as well as attempting to minimize response time
2569  *      while increasing bulk throughput.
2570  **/
2571 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2572 			     struct ixgbe_ring_container *ring_container)
2573 {
2574 	unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2575 			   IXGBE_ITR_ADAPTIVE_LATENCY;
2576 	unsigned int avg_wire_size, packets, bytes;
2577 	unsigned long next_update = jiffies;
2578 
2579 	/* If we don't have any rings just leave ourselves set for maximum
2580 	 * possible latency so we take ourselves out of the equation.
2581 	 */
2582 	if (!ring_container->ring)
2583 		return;
2584 
2585 	/* If we didn't update within up to 1 - 2 jiffies we can assume
2586 	 * that either packets are coming in so slow there hasn't been
2587 	 * any work, or that there is so much work that NAPI is dealing
2588 	 * with interrupt moderation and we don't need to do anything.
2589 	 */
2590 	if (time_after(next_update, ring_container->next_update))
2591 		goto clear_counts;
2592 
2593 	packets = ring_container->total_packets;
2594 
2595 	/* We have no packets to actually measure against. This means
2596 	 * either one of the other queues on this vector is active or
2597 	 * we are a Tx queue doing TSO with too high of an interrupt rate.
2598 	 *
2599 	 * When this occurs just tick up our delay by the minimum value
2600 	 * and hope that this extra delay will prevent us from being called
2601 	 * without any work on our queue.
2602 	 */
2603 	if (!packets) {
2604 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2605 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2606 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2607 		itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2608 		goto clear_counts;
2609 	}
2610 
2611 	bytes = ring_container->total_bytes;
2612 
2613 	/* If packets are less than 4 or bytes are less than 9000 assume
2614 	 * insufficient data to use bulk rate limiting approach. We are
2615 	 * likely latency driven.
2616 	 */
2617 	if (packets < 4 && bytes < 9000) {
2618 		itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2619 		goto adjust_by_size;
2620 	}
2621 
2622 	/* Between 4 and 48 we can assume that our current interrupt delay
2623 	 * is only slightly too low. As such we should increase it by a small
2624 	 * fixed amount.
2625 	 */
2626 	if (packets < 48) {
2627 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2628 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2629 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2630 		goto clear_counts;
2631 	}
2632 
2633 	/* Between 48 and 96 is our "goldilocks" zone where we are working
2634 	 * out "just right". Just report that our current ITR is good for us.
2635 	 */
2636 	if (packets < 96) {
2637 		itr = q_vector->itr >> 2;
2638 		goto clear_counts;
2639 	}
2640 
2641 	/* If packet count is 96 or greater we are likely looking at a slight
2642 	 * overrun of the delay we want. Try halving our delay to see if that
2643 	 * will cut the number of packets in half per interrupt.
2644 	 */
2645 	if (packets < 256) {
2646 		itr = q_vector->itr >> 3;
2647 		if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2648 			itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2649 		goto clear_counts;
2650 	}
2651 
2652 	/* The paths below assume we are dealing with a bulk ITR since number
2653 	 * of packets is 256 or greater. We are just going to have to compute
2654 	 * a value and try to bring the count under control, though for smaller
2655 	 * packet sizes there isn't much we can do as NAPI polling will likely
2656 	 * be kicking in sooner rather than later.
2657 	 */
2658 	itr = IXGBE_ITR_ADAPTIVE_BULK;
2659 
2660 adjust_by_size:
2661 	/* If packet counts are 256 or greater we can assume we have a gross
2662 	 * overestimation of what the rate should be. Instead of trying to fine
2663 	 * tune it just use the formula below to try and dial in an exact value
2664 	 * give the current packet size of the frame.
2665 	 */
2666 	avg_wire_size = bytes / packets;
2667 
2668 	/* The following is a crude approximation of:
2669 	 *  wmem_default / (size + overhead) = desired_pkts_per_int
2670 	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2671 	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2672 	 *
2673 	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
2674 	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2675 	 * formula down to
2676 	 *
2677 	 *  (170 * (size + 24)) / (size + 640) = ITR
2678 	 *
2679 	 * We first do some math on the packet size and then finally bitshift
2680 	 * by 8 after rounding up. We also have to account for PCIe link speed
2681 	 * difference as ITR scales based on this.
2682 	 */
2683 	if (avg_wire_size <= 60) {
2684 		/* Start at 50k ints/sec */
2685 		avg_wire_size = 5120;
2686 	} else if (avg_wire_size <= 316) {
2687 		/* 50K ints/sec to 16K ints/sec */
2688 		avg_wire_size *= 40;
2689 		avg_wire_size += 2720;
2690 	} else if (avg_wire_size <= 1084) {
2691 		/* 16K ints/sec to 9.2K ints/sec */
2692 		avg_wire_size *= 15;
2693 		avg_wire_size += 11452;
2694 	} else if (avg_wire_size <= 1980) {
2695 		/* 9.2K ints/sec to 8K ints/sec */
2696 		avg_wire_size *= 5;
2697 		avg_wire_size += 22420;
2698 	} else {
2699 		/* plateau at a limit of 8K ints/sec */
2700 		avg_wire_size = 32256;
2701 	}
2702 
2703 	/* If we are in low latency mode half our delay which doubles the rate
2704 	 * to somewhere between 100K to 16K ints/sec
2705 	 */
2706 	if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2707 		avg_wire_size >>= 1;
2708 
2709 	/* Resultant value is 256 times larger than it needs to be. This
2710 	 * gives us room to adjust the value as needed to either increase
2711 	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2712 	 *
2713 	 * Use addition as we have already recorded the new latency flag
2714 	 * for the ITR value.
2715 	 */
2716 	switch (q_vector->adapter->link_speed) {
2717 	case IXGBE_LINK_SPEED_10GB_FULL:
2718 	case IXGBE_LINK_SPEED_100_FULL:
2719 	default:
2720 		itr += DIV_ROUND_UP(avg_wire_size,
2721 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2722 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2723 		break;
2724 	case IXGBE_LINK_SPEED_2_5GB_FULL:
2725 	case IXGBE_LINK_SPEED_1GB_FULL:
2726 	case IXGBE_LINK_SPEED_10_FULL:
2727 		itr += DIV_ROUND_UP(avg_wire_size,
2728 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2729 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2730 		break;
2731 	}
2732 
2733 clear_counts:
2734 	/* write back value */
2735 	ring_container->itr = itr;
2736 
2737 	/* next update should occur within next jiffy */
2738 	ring_container->next_update = next_update + 1;
2739 
2740 	ring_container->total_bytes = 0;
2741 	ring_container->total_packets = 0;
2742 }
2743 
2744 /**
2745  * ixgbe_write_eitr - write EITR register in hardware specific way
2746  * @q_vector: structure containing interrupt and ring information
2747  *
2748  * This function is made to be called by ethtool and by the driver
2749  * when it needs to update EITR registers at runtime.  Hardware
2750  * specific quirks/differences are taken care of here.
2751  */
2752 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2753 {
2754 	struct ixgbe_adapter *adapter = q_vector->adapter;
2755 	struct ixgbe_hw *hw = &adapter->hw;
2756 	int v_idx = q_vector->v_idx;
2757 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2758 
2759 	switch (adapter->hw.mac.type) {
2760 	case ixgbe_mac_82598EB:
2761 		/* must write high and low 16 bits to reset counter */
2762 		itr_reg |= (itr_reg << 16);
2763 		break;
2764 	case ixgbe_mac_82599EB:
2765 	case ixgbe_mac_X540:
2766 	case ixgbe_mac_X550:
2767 	case ixgbe_mac_X550EM_x:
2768 	case ixgbe_mac_x550em_a:
2769 		/*
2770 		 * set the WDIS bit to not clear the timer bits and cause an
2771 		 * immediate assertion of the interrupt
2772 		 */
2773 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2774 		break;
2775 	default:
2776 		break;
2777 	}
2778 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2779 }
2780 
2781 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2782 {
2783 	u32 new_itr;
2784 
2785 	ixgbe_update_itr(q_vector, &q_vector->tx);
2786 	ixgbe_update_itr(q_vector, &q_vector->rx);
2787 
2788 	/* use the smallest value of new ITR delay calculations */
2789 	new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2790 
2791 	/* Clear latency flag if set, shift into correct position */
2792 	new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2793 	new_itr <<= 2;
2794 
2795 	if (new_itr != q_vector->itr) {
2796 		/* save the algorithm value here */
2797 		q_vector->itr = new_itr;
2798 
2799 		ixgbe_write_eitr(q_vector);
2800 	}
2801 }
2802 
2803 /**
2804  * ixgbe_check_overtemp_subtask - check for over temperature
2805  * @adapter: pointer to adapter
2806  **/
2807 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2808 {
2809 	struct ixgbe_hw *hw = &adapter->hw;
2810 	u32 eicr = adapter->interrupt_event;
2811 	s32 rc;
2812 
2813 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2814 		return;
2815 
2816 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2817 		return;
2818 
2819 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2820 
2821 	switch (hw->device_id) {
2822 	case IXGBE_DEV_ID_82599_T3_LOM:
2823 		/*
2824 		 * Since the warning interrupt is for both ports
2825 		 * we don't have to check if:
2826 		 *  - This interrupt wasn't for our port.
2827 		 *  - We may have missed the interrupt so always have to
2828 		 *    check if we  got a LSC
2829 		 */
2830 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2831 		    !(eicr & IXGBE_EICR_LSC))
2832 			return;
2833 
2834 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2835 			u32 speed;
2836 			bool link_up = false;
2837 
2838 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2839 
2840 			if (link_up)
2841 				return;
2842 		}
2843 
2844 		/* Check if this is not due to overtemp */
2845 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2846 			return;
2847 
2848 		break;
2849 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2850 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2851 		rc = hw->phy.ops.check_overtemp(hw);
2852 		if (rc != IXGBE_ERR_OVERTEMP)
2853 			return;
2854 		break;
2855 	default:
2856 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2857 			return;
2858 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2859 			return;
2860 		break;
2861 	}
2862 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2863 
2864 	adapter->interrupt_event = 0;
2865 }
2866 
2867 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2868 {
2869 	struct ixgbe_hw *hw = &adapter->hw;
2870 
2871 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2872 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2873 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2874 		/* write to clear the interrupt */
2875 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2876 	}
2877 }
2878 
2879 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2880 {
2881 	struct ixgbe_hw *hw = &adapter->hw;
2882 
2883 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2884 		return;
2885 
2886 	switch (adapter->hw.mac.type) {
2887 	case ixgbe_mac_82599EB:
2888 		/*
2889 		 * Need to check link state so complete overtemp check
2890 		 * on service task
2891 		 */
2892 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2893 		     (eicr & IXGBE_EICR_LSC)) &&
2894 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2895 			adapter->interrupt_event = eicr;
2896 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2897 			ixgbe_service_event_schedule(adapter);
2898 			return;
2899 		}
2900 		return;
2901 	case ixgbe_mac_x550em_a:
2902 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2903 			adapter->interrupt_event = eicr;
2904 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2905 			ixgbe_service_event_schedule(adapter);
2906 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2907 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2908 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2909 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2910 		}
2911 		return;
2912 	case ixgbe_mac_X550:
2913 	case ixgbe_mac_X540:
2914 		if (!(eicr & IXGBE_EICR_TS))
2915 			return;
2916 		break;
2917 	default:
2918 		return;
2919 	}
2920 
2921 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2922 }
2923 
2924 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2925 {
2926 	switch (hw->mac.type) {
2927 	case ixgbe_mac_82598EB:
2928 		if (hw->phy.type == ixgbe_phy_nl)
2929 			return true;
2930 		return false;
2931 	case ixgbe_mac_82599EB:
2932 	case ixgbe_mac_X550EM_x:
2933 	case ixgbe_mac_x550em_a:
2934 		switch (hw->mac.ops.get_media_type(hw)) {
2935 		case ixgbe_media_type_fiber:
2936 		case ixgbe_media_type_fiber_qsfp:
2937 			return true;
2938 		default:
2939 			return false;
2940 		}
2941 	default:
2942 		return false;
2943 	}
2944 }
2945 
2946 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2947 {
2948 	struct ixgbe_hw *hw = &adapter->hw;
2949 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2950 
2951 	if (!ixgbe_is_sfp(hw))
2952 		return;
2953 
2954 	/* Later MAC's use different SDP */
2955 	if (hw->mac.type >= ixgbe_mac_X540)
2956 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2957 
2958 	if (eicr & eicr_mask) {
2959 		/* Clear the interrupt */
2960 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2961 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2962 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2963 			adapter->sfp_poll_time = 0;
2964 			ixgbe_service_event_schedule(adapter);
2965 		}
2966 	}
2967 
2968 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2969 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2970 		/* Clear the interrupt */
2971 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2972 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2973 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2974 			ixgbe_service_event_schedule(adapter);
2975 		}
2976 	}
2977 }
2978 
2979 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2980 {
2981 	struct ixgbe_hw *hw = &adapter->hw;
2982 
2983 	adapter->lsc_int++;
2984 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2985 	adapter->link_check_timeout = jiffies;
2986 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2987 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2988 		IXGBE_WRITE_FLUSH(hw);
2989 		ixgbe_service_event_schedule(adapter);
2990 	}
2991 }
2992 
2993 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2994 					   u64 qmask)
2995 {
2996 	u32 mask;
2997 	struct ixgbe_hw *hw = &adapter->hw;
2998 
2999 	switch (hw->mac.type) {
3000 	case ixgbe_mac_82598EB:
3001 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
3002 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
3003 		break;
3004 	case ixgbe_mac_82599EB:
3005 	case ixgbe_mac_X540:
3006 	case ixgbe_mac_X550:
3007 	case ixgbe_mac_X550EM_x:
3008 	case ixgbe_mac_x550em_a:
3009 		mask = (qmask & 0xFFFFFFFF);
3010 		if (mask)
3011 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
3012 		mask = (qmask >> 32);
3013 		if (mask)
3014 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
3015 		break;
3016 	default:
3017 		break;
3018 	}
3019 	/* skip the flush */
3020 }
3021 
3022 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
3023 					    u64 qmask)
3024 {
3025 	u32 mask;
3026 	struct ixgbe_hw *hw = &adapter->hw;
3027 
3028 	switch (hw->mac.type) {
3029 	case ixgbe_mac_82598EB:
3030 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
3031 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
3032 		break;
3033 	case ixgbe_mac_82599EB:
3034 	case ixgbe_mac_X540:
3035 	case ixgbe_mac_X550:
3036 	case ixgbe_mac_X550EM_x:
3037 	case ixgbe_mac_x550em_a:
3038 		mask = (qmask & 0xFFFFFFFF);
3039 		if (mask)
3040 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
3041 		mask = (qmask >> 32);
3042 		if (mask)
3043 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
3044 		break;
3045 	default:
3046 		break;
3047 	}
3048 	/* skip the flush */
3049 }
3050 
3051 /**
3052  * ixgbe_irq_enable - Enable default interrupt generation settings
3053  * @adapter: board private structure
3054  * @queues: enable irqs for queues
3055  * @flush: flush register write
3056  **/
3057 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
3058 				    bool flush)
3059 {
3060 	struct ixgbe_hw *hw = &adapter->hw;
3061 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
3062 
3063 	/* don't reenable LSC while waiting for link */
3064 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
3065 		mask &= ~IXGBE_EIMS_LSC;
3066 
3067 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
3068 		switch (adapter->hw.mac.type) {
3069 		case ixgbe_mac_82599EB:
3070 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
3071 			break;
3072 		case ixgbe_mac_X540:
3073 		case ixgbe_mac_X550:
3074 		case ixgbe_mac_X550EM_x:
3075 		case ixgbe_mac_x550em_a:
3076 			mask |= IXGBE_EIMS_TS;
3077 			break;
3078 		default:
3079 			break;
3080 		}
3081 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3082 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3083 	switch (adapter->hw.mac.type) {
3084 	case ixgbe_mac_82599EB:
3085 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3086 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
3087 		/* fall through */
3088 	case ixgbe_mac_X540:
3089 	case ixgbe_mac_X550:
3090 	case ixgbe_mac_X550EM_x:
3091 	case ixgbe_mac_x550em_a:
3092 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3093 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3094 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3095 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3096 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3097 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3098 		mask |= IXGBE_EIMS_ECC;
3099 		mask |= IXGBE_EIMS_MAILBOX;
3100 		break;
3101 	default:
3102 		break;
3103 	}
3104 
3105 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3106 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3107 		mask |= IXGBE_EIMS_FLOW_DIR;
3108 
3109 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3110 	if (queues)
3111 		ixgbe_irq_enable_queues(adapter, ~0);
3112 	if (flush)
3113 		IXGBE_WRITE_FLUSH(&adapter->hw);
3114 }
3115 
3116 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3117 {
3118 	struct ixgbe_adapter *adapter = data;
3119 	struct ixgbe_hw *hw = &adapter->hw;
3120 	u32 eicr;
3121 
3122 	/*
3123 	 * Workaround for Silicon errata.  Use clear-by-write instead
3124 	 * of clear-by-read.  Reading with EICS will return the
3125 	 * interrupt causes without clearing, which later be done
3126 	 * with the write to EICR.
3127 	 */
3128 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3129 
3130 	/* The lower 16bits of the EICR register are for the queue interrupts
3131 	 * which should be masked here in order to not accidentally clear them if
3132 	 * the bits are high when ixgbe_msix_other is called. There is a race
3133 	 * condition otherwise which results in possible performance loss
3134 	 * especially if the ixgbe_msix_other interrupt is triggering
3135 	 * consistently (as it would when PPS is turned on for the X540 device)
3136 	 */
3137 	eicr &= 0xFFFF0000;
3138 
3139 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3140 
3141 	if (eicr & IXGBE_EICR_LSC)
3142 		ixgbe_check_lsc(adapter);
3143 
3144 	if (eicr & IXGBE_EICR_MAILBOX)
3145 		ixgbe_msg_task(adapter);
3146 
3147 	switch (hw->mac.type) {
3148 	case ixgbe_mac_82599EB:
3149 	case ixgbe_mac_X540:
3150 	case ixgbe_mac_X550:
3151 	case ixgbe_mac_X550EM_x:
3152 	case ixgbe_mac_x550em_a:
3153 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3154 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3155 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3156 			ixgbe_service_event_schedule(adapter);
3157 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3158 					IXGBE_EICR_GPI_SDP0_X540);
3159 		}
3160 		if (eicr & IXGBE_EICR_ECC) {
3161 			e_info(link, "Received ECC Err, initiating reset\n");
3162 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3163 			ixgbe_service_event_schedule(adapter);
3164 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3165 		}
3166 		/* Handle Flow Director Full threshold interrupt */
3167 		if (eicr & IXGBE_EICR_FLOW_DIR) {
3168 			int reinit_count = 0;
3169 			int i;
3170 			for (i = 0; i < adapter->num_tx_queues; i++) {
3171 				struct ixgbe_ring *ring = adapter->tx_ring[i];
3172 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3173 						       &ring->state))
3174 					reinit_count++;
3175 			}
3176 			if (reinit_count) {
3177 				/* no more flow director interrupts until after init */
3178 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3179 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3180 				ixgbe_service_event_schedule(adapter);
3181 			}
3182 		}
3183 		ixgbe_check_sfp_event(adapter, eicr);
3184 		ixgbe_check_overtemp_event(adapter, eicr);
3185 		break;
3186 	default:
3187 		break;
3188 	}
3189 
3190 	ixgbe_check_fan_failure(adapter, eicr);
3191 
3192 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3193 		ixgbe_ptp_check_pps_event(adapter);
3194 
3195 	/* re-enable the original interrupt state, no lsc, no queues */
3196 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3197 		ixgbe_irq_enable(adapter, false, false);
3198 
3199 	return IRQ_HANDLED;
3200 }
3201 
3202 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3203 {
3204 	struct ixgbe_q_vector *q_vector = data;
3205 
3206 	/* EIAM disabled interrupts (on this vector) for us */
3207 
3208 	if (q_vector->rx.ring || q_vector->tx.ring)
3209 		napi_schedule_irqoff(&q_vector->napi);
3210 
3211 	return IRQ_HANDLED;
3212 }
3213 
3214 /**
3215  * ixgbe_poll - NAPI Rx polling callback
3216  * @napi: structure for representing this polling device
3217  * @budget: how many packets driver is allowed to clean
3218  *
3219  * This function is used for legacy and MSI, NAPI mode
3220  **/
3221 int ixgbe_poll(struct napi_struct *napi, int budget)
3222 {
3223 	struct ixgbe_q_vector *q_vector =
3224 				container_of(napi, struct ixgbe_q_vector, napi);
3225 	struct ixgbe_adapter *adapter = q_vector->adapter;
3226 	struct ixgbe_ring *ring;
3227 	int per_ring_budget, work_done = 0;
3228 	bool clean_complete = true;
3229 
3230 #ifdef CONFIG_IXGBE_DCA
3231 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3232 		ixgbe_update_dca(q_vector);
3233 #endif
3234 
3235 	ixgbe_for_each_ring(ring, q_vector->tx) {
3236 		if (!ixgbe_clean_tx_irq(q_vector, ring, budget))
3237 			clean_complete = false;
3238 	}
3239 
3240 	/* Exit if we are called by netpoll */
3241 	if (budget <= 0)
3242 		return budget;
3243 
3244 	/* attempt to distribute budget to each queue fairly, but don't allow
3245 	 * the budget to go below 1 because we'll exit polling */
3246 	if (q_vector->rx.count > 1)
3247 		per_ring_budget = max(budget/q_vector->rx.count, 1);
3248 	else
3249 		per_ring_budget = budget;
3250 
3251 	ixgbe_for_each_ring(ring, q_vector->rx) {
3252 		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
3253 						 per_ring_budget);
3254 
3255 		work_done += cleaned;
3256 		if (cleaned >= per_ring_budget)
3257 			clean_complete = false;
3258 	}
3259 
3260 	/* If all work not completed, return budget and keep polling */
3261 	if (!clean_complete)
3262 		return budget;
3263 
3264 	/* all work done, exit the polling mode */
3265 	napi_complete_done(napi, work_done);
3266 	if (adapter->rx_itr_setting & 1)
3267 		ixgbe_set_itr(q_vector);
3268 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3269 		ixgbe_irq_enable_queues(adapter, BIT_ULL(q_vector->v_idx));
3270 
3271 	return min(work_done, budget - 1);
3272 }
3273 
3274 /**
3275  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3276  * @adapter: board private structure
3277  *
3278  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3279  * interrupts from the kernel.
3280  **/
3281 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3282 {
3283 	struct net_device *netdev = adapter->netdev;
3284 	unsigned int ri = 0, ti = 0;
3285 	int vector, err;
3286 
3287 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3288 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3289 		struct msix_entry *entry = &adapter->msix_entries[vector];
3290 
3291 		if (q_vector->tx.ring && q_vector->rx.ring) {
3292 			snprintf(q_vector->name, sizeof(q_vector->name),
3293 				 "%s-TxRx-%u", netdev->name, ri++);
3294 			ti++;
3295 		} else if (q_vector->rx.ring) {
3296 			snprintf(q_vector->name, sizeof(q_vector->name),
3297 				 "%s-rx-%u", netdev->name, ri++);
3298 		} else if (q_vector->tx.ring) {
3299 			snprintf(q_vector->name, sizeof(q_vector->name),
3300 				 "%s-tx-%u", netdev->name, ti++);
3301 		} else {
3302 			/* skip this unused q_vector */
3303 			continue;
3304 		}
3305 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3306 				  q_vector->name, q_vector);
3307 		if (err) {
3308 			e_err(probe, "request_irq failed for MSIX interrupt "
3309 			      "Error: %d\n", err);
3310 			goto free_queue_irqs;
3311 		}
3312 		/* If Flow Director is enabled, set interrupt affinity */
3313 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3314 			/* assign the mask for this irq */
3315 			irq_set_affinity_hint(entry->vector,
3316 					      &q_vector->affinity_mask);
3317 		}
3318 	}
3319 
3320 	err = request_irq(adapter->msix_entries[vector].vector,
3321 			  ixgbe_msix_other, 0, netdev->name, adapter);
3322 	if (err) {
3323 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3324 		goto free_queue_irqs;
3325 	}
3326 
3327 	return 0;
3328 
3329 free_queue_irqs:
3330 	while (vector) {
3331 		vector--;
3332 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
3333 				      NULL);
3334 		free_irq(adapter->msix_entries[vector].vector,
3335 			 adapter->q_vector[vector]);
3336 	}
3337 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3338 	pci_disable_msix(adapter->pdev);
3339 	kfree(adapter->msix_entries);
3340 	adapter->msix_entries = NULL;
3341 	return err;
3342 }
3343 
3344 /**
3345  * ixgbe_intr - legacy mode Interrupt Handler
3346  * @irq: interrupt number
3347  * @data: pointer to a network interface device structure
3348  **/
3349 static irqreturn_t ixgbe_intr(int irq, void *data)
3350 {
3351 	struct ixgbe_adapter *adapter = data;
3352 	struct ixgbe_hw *hw = &adapter->hw;
3353 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3354 	u32 eicr;
3355 
3356 	/*
3357 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3358 	 * before the read of EICR.
3359 	 */
3360 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3361 
3362 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3363 	 * therefore no explicit interrupt disable is necessary */
3364 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3365 	if (!eicr) {
3366 		/*
3367 		 * shared interrupt alert!
3368 		 * make sure interrupts are enabled because the read will
3369 		 * have disabled interrupts due to EIAM
3370 		 * finish the workaround of silicon errata on 82598.  Unmask
3371 		 * the interrupt that we masked before the EICR read.
3372 		 */
3373 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3374 			ixgbe_irq_enable(adapter, true, true);
3375 		return IRQ_NONE;	/* Not our interrupt */
3376 	}
3377 
3378 	if (eicr & IXGBE_EICR_LSC)
3379 		ixgbe_check_lsc(adapter);
3380 
3381 	switch (hw->mac.type) {
3382 	case ixgbe_mac_82599EB:
3383 		ixgbe_check_sfp_event(adapter, eicr);
3384 		/* Fall through */
3385 	case ixgbe_mac_X540:
3386 	case ixgbe_mac_X550:
3387 	case ixgbe_mac_X550EM_x:
3388 	case ixgbe_mac_x550em_a:
3389 		if (eicr & IXGBE_EICR_ECC) {
3390 			e_info(link, "Received ECC Err, initiating reset\n");
3391 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3392 			ixgbe_service_event_schedule(adapter);
3393 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3394 		}
3395 		ixgbe_check_overtemp_event(adapter, eicr);
3396 		break;
3397 	default:
3398 		break;
3399 	}
3400 
3401 	ixgbe_check_fan_failure(adapter, eicr);
3402 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3403 		ixgbe_ptp_check_pps_event(adapter);
3404 
3405 	/* would disable interrupts here but EIAM disabled it */
3406 	napi_schedule_irqoff(&q_vector->napi);
3407 
3408 	/*
3409 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3410 	 * ixgbe_poll will re-enable the queue interrupts
3411 	 */
3412 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3413 		ixgbe_irq_enable(adapter, false, false);
3414 
3415 	return IRQ_HANDLED;
3416 }
3417 
3418 /**
3419  * ixgbe_request_irq - initialize interrupts
3420  * @adapter: board private structure
3421  *
3422  * Attempts to configure interrupts using the best available
3423  * capabilities of the hardware and kernel.
3424  **/
3425 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3426 {
3427 	struct net_device *netdev = adapter->netdev;
3428 	int err;
3429 
3430 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3431 		err = ixgbe_request_msix_irqs(adapter);
3432 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3433 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3434 				  netdev->name, adapter);
3435 	else
3436 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3437 				  netdev->name, adapter);
3438 
3439 	if (err)
3440 		e_err(probe, "request_irq failed, Error %d\n", err);
3441 
3442 	return err;
3443 }
3444 
3445 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3446 {
3447 	int vector;
3448 
3449 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3450 		free_irq(adapter->pdev->irq, adapter);
3451 		return;
3452 	}
3453 
3454 	if (!adapter->msix_entries)
3455 		return;
3456 
3457 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3458 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3459 		struct msix_entry *entry = &adapter->msix_entries[vector];
3460 
3461 		/* free only the irqs that were actually requested */
3462 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3463 			continue;
3464 
3465 		/* clear the affinity_mask in the IRQ descriptor */
3466 		irq_set_affinity_hint(entry->vector, NULL);
3467 
3468 		free_irq(entry->vector, q_vector);
3469 	}
3470 
3471 	free_irq(adapter->msix_entries[vector].vector, adapter);
3472 }
3473 
3474 /**
3475  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3476  * @adapter: board private structure
3477  **/
3478 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3479 {
3480 	switch (adapter->hw.mac.type) {
3481 	case ixgbe_mac_82598EB:
3482 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3483 		break;
3484 	case ixgbe_mac_82599EB:
3485 	case ixgbe_mac_X540:
3486 	case ixgbe_mac_X550:
3487 	case ixgbe_mac_X550EM_x:
3488 	case ixgbe_mac_x550em_a:
3489 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3490 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3491 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3492 		break;
3493 	default:
3494 		break;
3495 	}
3496 	IXGBE_WRITE_FLUSH(&adapter->hw);
3497 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3498 		int vector;
3499 
3500 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3501 			synchronize_irq(adapter->msix_entries[vector].vector);
3502 
3503 		synchronize_irq(adapter->msix_entries[vector++].vector);
3504 	} else {
3505 		synchronize_irq(adapter->pdev->irq);
3506 	}
3507 }
3508 
3509 /**
3510  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3511  * @adapter: board private structure
3512  *
3513  **/
3514 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3515 {
3516 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3517 
3518 	ixgbe_write_eitr(q_vector);
3519 
3520 	ixgbe_set_ivar(adapter, 0, 0, 0);
3521 	ixgbe_set_ivar(adapter, 1, 0, 0);
3522 
3523 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3524 }
3525 
3526 /**
3527  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3528  * @adapter: board private structure
3529  * @ring: structure containing ring specific data
3530  *
3531  * Configure the Tx descriptor ring after a reset.
3532  **/
3533 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3534 			     struct ixgbe_ring *ring)
3535 {
3536 	struct ixgbe_hw *hw = &adapter->hw;
3537 	u64 tdba = ring->dma;
3538 	int wait_loop = 10;
3539 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3540 	u8 reg_idx = ring->reg_idx;
3541 
3542 	/* disable queue to avoid issues while updating state */
3543 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3544 	IXGBE_WRITE_FLUSH(hw);
3545 
3546 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3547 			(tdba & DMA_BIT_MASK(32)));
3548 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3549 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3550 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3551 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3552 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3553 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3554 
3555 	/*
3556 	 * set WTHRESH to encourage burst writeback, it should not be set
3557 	 * higher than 1 when:
3558 	 * - ITR is 0 as it could cause false TX hangs
3559 	 * - ITR is set to > 100k int/sec and BQL is enabled
3560 	 *
3561 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3562 	 * to or less than the number of on chip descriptors, which is
3563 	 * currently 40.
3564 	 */
3565 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3566 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3567 	else
3568 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3569 
3570 	/*
3571 	 * Setting PTHRESH to 32 both improves performance
3572 	 * and avoids a TX hang with DFP enabled
3573 	 */
3574 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3575 		   32;		/* PTHRESH = 32 */
3576 
3577 	/* reinitialize flowdirector state */
3578 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3579 		ring->atr_sample_rate = adapter->atr_sample_rate;
3580 		ring->atr_count = 0;
3581 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3582 	} else {
3583 		ring->atr_sample_rate = 0;
3584 	}
3585 
3586 	/* initialize XPS */
3587 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3588 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3589 
3590 		if (q_vector)
3591 			netif_set_xps_queue(ring->netdev,
3592 					    &q_vector->affinity_mask,
3593 					    ring->queue_index);
3594 	}
3595 
3596 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3597 
3598 	/* reinitialize tx_buffer_info */
3599 	memset(ring->tx_buffer_info, 0,
3600 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3601 
3602 	/* enable queue */
3603 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3604 
3605 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3606 	if (hw->mac.type == ixgbe_mac_82598EB &&
3607 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3608 		return;
3609 
3610 	/* poll to verify queue is enabled */
3611 	do {
3612 		usleep_range(1000, 2000);
3613 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3614 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3615 	if (!wait_loop)
3616 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3617 }
3618 
3619 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3620 {
3621 	struct ixgbe_hw *hw = &adapter->hw;
3622 	u32 rttdcs, mtqc;
3623 	u8 tcs = adapter->hw_tcs;
3624 
3625 	if (hw->mac.type == ixgbe_mac_82598EB)
3626 		return;
3627 
3628 	/* disable the arbiter while setting MTQC */
3629 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3630 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3631 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3632 
3633 	/* set transmit pool layout */
3634 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3635 		mtqc = IXGBE_MTQC_VT_ENA;
3636 		if (tcs > 4)
3637 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3638 		else if (tcs > 1)
3639 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3640 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3641 			 IXGBE_82599_VMDQ_4Q_MASK)
3642 			mtqc |= IXGBE_MTQC_32VF;
3643 		else
3644 			mtqc |= IXGBE_MTQC_64VF;
3645 	} else {
3646 		if (tcs > 4)
3647 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3648 		else if (tcs > 1)
3649 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3650 		else
3651 			mtqc = IXGBE_MTQC_64Q_1PB;
3652 	}
3653 
3654 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3655 
3656 	/* Enable Security TX Buffer IFG for multiple pb */
3657 	if (tcs) {
3658 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3659 		sectx |= IXGBE_SECTX_DCB;
3660 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3661 	}
3662 
3663 	/* re-enable the arbiter */
3664 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3665 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3666 }
3667 
3668 /**
3669  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3670  * @adapter: board private structure
3671  *
3672  * Configure the Tx unit of the MAC after a reset.
3673  **/
3674 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3675 {
3676 	struct ixgbe_hw *hw = &adapter->hw;
3677 	u32 dmatxctl;
3678 	u32 i;
3679 
3680 	ixgbe_setup_mtqc(adapter);
3681 
3682 	if (hw->mac.type != ixgbe_mac_82598EB) {
3683 		/* DMATXCTL.EN must be before Tx queues are enabled */
3684 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3685 		dmatxctl |= IXGBE_DMATXCTL_TE;
3686 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3687 	}
3688 
3689 	/* Setup the HW Tx Head and Tail descriptor pointers */
3690 	for (i = 0; i < adapter->num_tx_queues; i++)
3691 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3692 	for (i = 0; i < adapter->num_xdp_queues; i++)
3693 		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3694 }
3695 
3696 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3697 				 struct ixgbe_ring *ring)
3698 {
3699 	struct ixgbe_hw *hw = &adapter->hw;
3700 	u8 reg_idx = ring->reg_idx;
3701 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3702 
3703 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3704 
3705 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3706 }
3707 
3708 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3709 				  struct ixgbe_ring *ring)
3710 {
3711 	struct ixgbe_hw *hw = &adapter->hw;
3712 	u8 reg_idx = ring->reg_idx;
3713 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3714 
3715 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3716 
3717 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3718 }
3719 
3720 #ifdef CONFIG_IXGBE_DCB
3721 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3722 #else
3723 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3724 #endif
3725 {
3726 	int i;
3727 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3728 
3729 	if (adapter->ixgbe_ieee_pfc)
3730 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3731 
3732 	/*
3733 	 * We should set the drop enable bit if:
3734 	 *  SR-IOV is enabled
3735 	 *   or
3736 	 *  Number of Rx queues > 1 and flow control is disabled
3737 	 *
3738 	 *  This allows us to avoid head of line blocking for security
3739 	 *  and performance reasons.
3740 	 */
3741 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3742 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3743 		for (i = 0; i < adapter->num_rx_queues; i++)
3744 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3745 	} else {
3746 		for (i = 0; i < adapter->num_rx_queues; i++)
3747 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3748 	}
3749 }
3750 
3751 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3752 
3753 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3754 				   struct ixgbe_ring *rx_ring)
3755 {
3756 	struct ixgbe_hw *hw = &adapter->hw;
3757 	u32 srrctl;
3758 	u8 reg_idx = rx_ring->reg_idx;
3759 
3760 	if (hw->mac.type == ixgbe_mac_82598EB) {
3761 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3762 
3763 		/*
3764 		 * if VMDq is not active we must program one srrctl register
3765 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3766 		 */
3767 		reg_idx &= mask;
3768 	}
3769 
3770 	/* configure header buffer length, needed for RSC */
3771 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3772 
3773 	/* configure the packet buffer length */
3774 	if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state))
3775 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3776 	else
3777 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3778 
3779 	/* configure descriptor type */
3780 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3781 
3782 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3783 }
3784 
3785 /**
3786  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3787  * @adapter: device handle
3788  *
3789  *  - 82598/82599/X540:     128
3790  *  - X550(non-SRIOV mode): 512
3791  *  - X550(SRIOV mode):     64
3792  */
3793 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3794 {
3795 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3796 		return 128;
3797 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3798 		return 64;
3799 	else
3800 		return 512;
3801 }
3802 
3803 /**
3804  * ixgbe_store_key - Write the RSS key to HW
3805  * @adapter: device handle
3806  *
3807  * Write the RSS key stored in adapter.rss_key to HW.
3808  */
3809 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3810 {
3811 	struct ixgbe_hw *hw = &adapter->hw;
3812 	int i;
3813 
3814 	for (i = 0; i < 10; i++)
3815 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3816 }
3817 
3818 /**
3819  * ixgbe_init_rss_key - Initialize adapter RSS key
3820  * @adapter: device handle
3821  *
3822  * Allocates and initializes the RSS key if it is not allocated.
3823  **/
3824 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3825 {
3826 	u32 *rss_key;
3827 
3828 	if (!adapter->rss_key) {
3829 		rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3830 		if (unlikely(!rss_key))
3831 			return -ENOMEM;
3832 
3833 		netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3834 		adapter->rss_key = rss_key;
3835 	}
3836 
3837 	return 0;
3838 }
3839 
3840 /**
3841  * ixgbe_store_reta - Write the RETA table to HW
3842  * @adapter: device handle
3843  *
3844  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3845  */
3846 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3847 {
3848 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3849 	struct ixgbe_hw *hw = &adapter->hw;
3850 	u32 reta = 0;
3851 	u32 indices_multi;
3852 	u8 *indir_tbl = adapter->rss_indir_tbl;
3853 
3854 	/* Fill out the redirection table as follows:
3855 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3856 	 *    indices.
3857 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3858 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3859 	 */
3860 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3861 		indices_multi = 0x11;
3862 	else
3863 		indices_multi = 0x1;
3864 
3865 	/* Write redirection table to HW */
3866 	for (i = 0; i < reta_entries; i++) {
3867 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3868 		if ((i & 3) == 3) {
3869 			if (i < 128)
3870 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3871 			else
3872 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3873 						reta);
3874 			reta = 0;
3875 		}
3876 	}
3877 }
3878 
3879 /**
3880  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3881  * @adapter: device handle
3882  *
3883  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3884  */
3885 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3886 {
3887 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3888 	struct ixgbe_hw *hw = &adapter->hw;
3889 	u32 vfreta = 0;
3890 
3891 	/* Write redirection table to HW */
3892 	for (i = 0; i < reta_entries; i++) {
3893 		u16 pool = adapter->num_rx_pools;
3894 
3895 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3896 		if ((i & 3) != 3)
3897 			continue;
3898 
3899 		while (pool--)
3900 			IXGBE_WRITE_REG(hw,
3901 					IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
3902 					vfreta);
3903 		vfreta = 0;
3904 	}
3905 }
3906 
3907 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3908 {
3909 	u32 i, j;
3910 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3911 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3912 
3913 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3914 	 * make full use of any rings they may have.  We will use the
3915 	 * PSRTYPE register to control how many rings we use within the PF.
3916 	 */
3917 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3918 		rss_i = 4;
3919 
3920 	/* Fill out hash function seeds */
3921 	ixgbe_store_key(adapter);
3922 
3923 	/* Fill out redirection table */
3924 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3925 
3926 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3927 		if (j == rss_i)
3928 			j = 0;
3929 
3930 		adapter->rss_indir_tbl[i] = j;
3931 	}
3932 
3933 	ixgbe_store_reta(adapter);
3934 }
3935 
3936 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3937 {
3938 	struct ixgbe_hw *hw = &adapter->hw;
3939 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3940 	int i, j;
3941 
3942 	/* Fill out hash function seeds */
3943 	for (i = 0; i < 10; i++) {
3944 		u16 pool = adapter->num_rx_pools;
3945 
3946 		while (pool--)
3947 			IXGBE_WRITE_REG(hw,
3948 					IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
3949 					*(adapter->rss_key + i));
3950 	}
3951 
3952 	/* Fill out the redirection table */
3953 	for (i = 0, j = 0; i < 64; i++, j++) {
3954 		if (j == rss_i)
3955 			j = 0;
3956 
3957 		adapter->rss_indir_tbl[i] = j;
3958 	}
3959 
3960 	ixgbe_store_vfreta(adapter);
3961 }
3962 
3963 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3964 {
3965 	struct ixgbe_hw *hw = &adapter->hw;
3966 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3967 	u32 rxcsum;
3968 
3969 	/* Disable indicating checksum in descriptor, enables RSS hash */
3970 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3971 	rxcsum |= IXGBE_RXCSUM_PCSD;
3972 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3973 
3974 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3975 		if (adapter->ring_feature[RING_F_RSS].mask)
3976 			mrqc = IXGBE_MRQC_RSSEN;
3977 	} else {
3978 		u8 tcs = adapter->hw_tcs;
3979 
3980 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3981 			if (tcs > 4)
3982 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3983 			else if (tcs > 1)
3984 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3985 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3986 				 IXGBE_82599_VMDQ_4Q_MASK)
3987 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3988 			else
3989 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3990 
3991 			/* Enable L3/L4 for Tx Switched packets */
3992 			mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3993 		} else {
3994 			if (tcs > 4)
3995 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3996 			else if (tcs > 1)
3997 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3998 			else
3999 				mrqc = IXGBE_MRQC_RSSEN;
4000 		}
4001 	}
4002 
4003 	/* Perform hash on these packet types */
4004 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
4005 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
4006 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
4007 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
4008 
4009 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
4010 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
4011 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
4012 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
4013 
4014 	if ((hw->mac.type >= ixgbe_mac_X550) &&
4015 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
4016 		u16 pool = adapter->num_rx_pools;
4017 
4018 		/* Enable VF RSS mode */
4019 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
4020 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4021 
4022 		/* Setup RSS through the VF registers */
4023 		ixgbe_setup_vfreta(adapter);
4024 		vfmrqc = IXGBE_MRQC_RSSEN;
4025 		vfmrqc |= rss_field;
4026 
4027 		while (pool--)
4028 			IXGBE_WRITE_REG(hw,
4029 					IXGBE_PFVFMRQC(VMDQ_P(pool)),
4030 					vfmrqc);
4031 	} else {
4032 		ixgbe_setup_reta(adapter);
4033 		mrqc |= rss_field;
4034 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4035 	}
4036 }
4037 
4038 /**
4039  * ixgbe_configure_rscctl - enable RSC for the indicated ring
4040  * @adapter: address of board private structure
4041  * @ring: structure containing ring specific data
4042  **/
4043 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
4044 				   struct ixgbe_ring *ring)
4045 {
4046 	struct ixgbe_hw *hw = &adapter->hw;
4047 	u32 rscctrl;
4048 	u8 reg_idx = ring->reg_idx;
4049 
4050 	if (!ring_is_rsc_enabled(ring))
4051 		return;
4052 
4053 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
4054 	rscctrl |= IXGBE_RSCCTL_RSCEN;
4055 	/*
4056 	 * we must limit the number of descriptors so that the
4057 	 * total size of max desc * buf_len is not greater
4058 	 * than 65536
4059 	 */
4060 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
4061 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
4062 }
4063 
4064 #define IXGBE_MAX_RX_DESC_POLL 10
4065 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4066 				       struct ixgbe_ring *ring)
4067 {
4068 	struct ixgbe_hw *hw = &adapter->hw;
4069 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4070 	u32 rxdctl;
4071 	u8 reg_idx = ring->reg_idx;
4072 
4073 	if (ixgbe_removed(hw->hw_addr))
4074 		return;
4075 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4076 	if (hw->mac.type == ixgbe_mac_82598EB &&
4077 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4078 		return;
4079 
4080 	do {
4081 		usleep_range(1000, 2000);
4082 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4083 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4084 
4085 	if (!wait_loop) {
4086 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4087 		      "the polling period\n", reg_idx);
4088 	}
4089 }
4090 
4091 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
4092 			    struct ixgbe_ring *ring)
4093 {
4094 	struct ixgbe_hw *hw = &adapter->hw;
4095 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4096 	u32 rxdctl;
4097 	u8 reg_idx = ring->reg_idx;
4098 
4099 	if (ixgbe_removed(hw->hw_addr))
4100 		return;
4101 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4102 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4103 
4104 	/* write value back with RXDCTL.ENABLE bit cleared */
4105 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4106 
4107 	if (hw->mac.type == ixgbe_mac_82598EB &&
4108 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4109 		return;
4110 
4111 	/* the hardware may take up to 100us to really disable the rx queue */
4112 	do {
4113 		udelay(10);
4114 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4115 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
4116 
4117 	if (!wait_loop) {
4118 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
4119 		      "the polling period\n", reg_idx);
4120 	}
4121 }
4122 
4123 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4124 			     struct ixgbe_ring *ring)
4125 {
4126 	struct ixgbe_hw *hw = &adapter->hw;
4127 	union ixgbe_adv_rx_desc *rx_desc;
4128 	u64 rdba = ring->dma;
4129 	u32 rxdctl;
4130 	u8 reg_idx = ring->reg_idx;
4131 
4132 	/* disable queue to avoid issues while updating state */
4133 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4134 	ixgbe_disable_rx_queue(adapter, ring);
4135 
4136 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4137 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4138 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4139 			ring->count * sizeof(union ixgbe_adv_rx_desc));
4140 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
4141 	IXGBE_WRITE_FLUSH(hw);
4142 
4143 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4144 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4145 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4146 
4147 	ixgbe_configure_srrctl(adapter, ring);
4148 	ixgbe_configure_rscctl(adapter, ring);
4149 
4150 	if (hw->mac.type == ixgbe_mac_82598EB) {
4151 		/*
4152 		 * enable cache line friendly hardware writes:
4153 		 * PTHRESH=32 descriptors (half the internal cache),
4154 		 * this also removes ugly rx_no_buffer_count increment
4155 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
4156 		 * WTHRESH=8 burst writeback up to two cache lines
4157 		 */
4158 		rxdctl &= ~0x3FFFFF;
4159 		rxdctl |=  0x080420;
4160 #if (PAGE_SIZE < 8192)
4161 	/* RXDCTL.RLPML does not work on 82599 */
4162 	} else if (hw->mac.type != ixgbe_mac_82599EB) {
4163 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4164 			    IXGBE_RXDCTL_RLPML_EN);
4165 
4166 		/* Limit the maximum frame size so we don't overrun the skb.
4167 		 * This can happen in SRIOV mode when the MTU of the VF is
4168 		 * higher than the MTU of the PF.
4169 		 */
4170 		if (ring_uses_build_skb(ring) &&
4171 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4172 			rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4173 				  IXGBE_RXDCTL_RLPML_EN;
4174 #endif
4175 	}
4176 
4177 	/* initialize rx_buffer_info */
4178 	memset(ring->rx_buffer_info, 0,
4179 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
4180 
4181 	/* initialize Rx descriptor 0 */
4182 	rx_desc = IXGBE_RX_DESC(ring, 0);
4183 	rx_desc->wb.upper.length = 0;
4184 
4185 	/* enable receive descriptor ring */
4186 	rxdctl |= IXGBE_RXDCTL_ENABLE;
4187 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4188 
4189 	ixgbe_rx_desc_queue_enable(adapter, ring);
4190 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4191 }
4192 
4193 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4194 {
4195 	struct ixgbe_hw *hw = &adapter->hw;
4196 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4197 	u16 pool = adapter->num_rx_pools;
4198 
4199 	/* PSRTYPE must be initialized in non 82598 adapters */
4200 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4201 		      IXGBE_PSRTYPE_UDPHDR |
4202 		      IXGBE_PSRTYPE_IPV4HDR |
4203 		      IXGBE_PSRTYPE_L2HDR |
4204 		      IXGBE_PSRTYPE_IPV6HDR;
4205 
4206 	if (hw->mac.type == ixgbe_mac_82598EB)
4207 		return;
4208 
4209 	if (rss_i > 3)
4210 		psrtype |= 2u << 29;
4211 	else if (rss_i > 1)
4212 		psrtype |= 1u << 29;
4213 
4214 	while (pool--)
4215 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4216 }
4217 
4218 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4219 {
4220 	struct ixgbe_hw *hw = &adapter->hw;
4221 	u16 pool = adapter->num_rx_pools;
4222 	u32 reg_offset, vf_shift, vmolr;
4223 	u32 gcr_ext, vmdctl;
4224 	int i;
4225 
4226 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4227 		return;
4228 
4229 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4230 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4231 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4232 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4233 	vmdctl |= IXGBE_VT_CTL_REPLEN;
4234 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4235 
4236 	/* accept untagged packets until a vlan tag is
4237 	 * specifically set for the VMDQ queue/pool
4238 	 */
4239 	vmolr = IXGBE_VMOLR_AUPE;
4240 	while (pool--)
4241 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(pool)), vmolr);
4242 
4243 	vf_shift = VMDQ_P(0) % 32;
4244 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4245 
4246 	/* Enable only the PF's pool for Tx/Rx */
4247 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4248 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4249 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4250 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4251 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4252 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4253 
4254 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4255 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4256 
4257 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
4258 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4259 
4260 	/*
4261 	 * Set up VF register offsets for selected VT Mode,
4262 	 * i.e. 32 or 64 VFs for SR-IOV
4263 	 */
4264 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4265 	case IXGBE_82599_VMDQ_8Q_MASK:
4266 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4267 		break;
4268 	case IXGBE_82599_VMDQ_4Q_MASK:
4269 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4270 		break;
4271 	default:
4272 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4273 		break;
4274 	}
4275 
4276 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4277 
4278 	for (i = 0; i < adapter->num_vfs; i++) {
4279 		/* configure spoof checking */
4280 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4281 					  adapter->vfinfo[i].spoofchk_enabled);
4282 
4283 		/* Enable/Disable RSS query feature  */
4284 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4285 					  adapter->vfinfo[i].rss_query_enabled);
4286 	}
4287 }
4288 
4289 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4290 {
4291 	struct ixgbe_hw *hw = &adapter->hw;
4292 	struct net_device *netdev = adapter->netdev;
4293 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4294 	struct ixgbe_ring *rx_ring;
4295 	int i;
4296 	u32 mhadd, hlreg0;
4297 
4298 #ifdef IXGBE_FCOE
4299 	/* adjust max frame to be able to do baby jumbo for FCoE */
4300 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4301 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4302 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4303 
4304 #endif /* IXGBE_FCOE */
4305 
4306 	/* adjust max frame to be at least the size of a standard frame */
4307 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4308 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4309 
4310 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4311 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4312 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
4313 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4314 
4315 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4316 	}
4317 
4318 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4319 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4320 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4321 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4322 
4323 	/*
4324 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4325 	 * the Base and Length of the Rx Descriptor Ring
4326 	 */
4327 	for (i = 0; i < adapter->num_rx_queues; i++) {
4328 		rx_ring = adapter->rx_ring[i];
4329 
4330 		clear_ring_rsc_enabled(rx_ring);
4331 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4332 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4333 
4334 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4335 			set_ring_rsc_enabled(rx_ring);
4336 
4337 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4338 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4339 
4340 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4341 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4342 			continue;
4343 
4344 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4345 
4346 #if (PAGE_SIZE < 8192)
4347 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4348 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4349 
4350 		if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4351 		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4352 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4353 #endif
4354 	}
4355 }
4356 
4357 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4358 {
4359 	struct ixgbe_hw *hw = &adapter->hw;
4360 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4361 
4362 	switch (hw->mac.type) {
4363 	case ixgbe_mac_82598EB:
4364 		/*
4365 		 * For VMDq support of different descriptor types or
4366 		 * buffer sizes through the use of multiple SRRCTL
4367 		 * registers, RDRXCTL.MVMEN must be set to 1
4368 		 *
4369 		 * also, the manual doesn't mention it clearly but DCA hints
4370 		 * will only use queue 0's tags unless this bit is set.  Side
4371 		 * effects of setting this bit are only that SRRCTL must be
4372 		 * fully programmed [0..15]
4373 		 */
4374 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4375 		break;
4376 	case ixgbe_mac_X550:
4377 	case ixgbe_mac_X550EM_x:
4378 	case ixgbe_mac_x550em_a:
4379 		if (adapter->num_vfs)
4380 			rdrxctl |= IXGBE_RDRXCTL_PSP;
4381 		/* fall through */
4382 	case ixgbe_mac_82599EB:
4383 	case ixgbe_mac_X540:
4384 		/* Disable RSC for ACK packets */
4385 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4386 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4387 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4388 		/* hardware requires some bits to be set by default */
4389 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4390 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4391 		break;
4392 	default:
4393 		/* We should do nothing since we don't know this hardware */
4394 		return;
4395 	}
4396 
4397 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4398 }
4399 
4400 /**
4401  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4402  * @adapter: board private structure
4403  *
4404  * Configure the Rx unit of the MAC after a reset.
4405  **/
4406 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4407 {
4408 	struct ixgbe_hw *hw = &adapter->hw;
4409 	int i;
4410 	u32 rxctrl, rfctl;
4411 
4412 	/* disable receives while setting up the descriptors */
4413 	hw->mac.ops.disable_rx(hw);
4414 
4415 	ixgbe_setup_psrtype(adapter);
4416 	ixgbe_setup_rdrxctl(adapter);
4417 
4418 	/* RSC Setup */
4419 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4420 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4421 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4422 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4423 
4424 	/* disable NFS filtering */
4425 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4426 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4427 
4428 	/* Program registers for the distribution of queues */
4429 	ixgbe_setup_mrqc(adapter);
4430 
4431 	/* set_rx_buffer_len must be called before ring initialization */
4432 	ixgbe_set_rx_buffer_len(adapter);
4433 
4434 	/*
4435 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4436 	 * the Base and Length of the Rx Descriptor Ring
4437 	 */
4438 	for (i = 0; i < adapter->num_rx_queues; i++)
4439 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4440 
4441 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4442 	/* disable drop enable for 82598 parts */
4443 	if (hw->mac.type == ixgbe_mac_82598EB)
4444 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4445 
4446 	/* enable all receives */
4447 	rxctrl |= IXGBE_RXCTRL_RXEN;
4448 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4449 }
4450 
4451 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4452 				 __be16 proto, u16 vid)
4453 {
4454 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4455 	struct ixgbe_hw *hw = &adapter->hw;
4456 
4457 	/* add VID to filter table */
4458 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4459 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4460 
4461 	set_bit(vid, adapter->active_vlans);
4462 
4463 	return 0;
4464 }
4465 
4466 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4467 {
4468 	u32 vlvf;
4469 	int idx;
4470 
4471 	/* short cut the special case */
4472 	if (vlan == 0)
4473 		return 0;
4474 
4475 	/* Search for the vlan id in the VLVF entries */
4476 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4477 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4478 		if ((vlvf & VLAN_VID_MASK) == vlan)
4479 			break;
4480 	}
4481 
4482 	return idx;
4483 }
4484 
4485 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4486 {
4487 	struct ixgbe_hw *hw = &adapter->hw;
4488 	u32 bits, word;
4489 	int idx;
4490 
4491 	idx = ixgbe_find_vlvf_entry(hw, vid);
4492 	if (!idx)
4493 		return;
4494 
4495 	/* See if any other pools are set for this VLAN filter
4496 	 * entry other than the PF.
4497 	 */
4498 	word = idx * 2 + (VMDQ_P(0) / 32);
4499 	bits = ~BIT(VMDQ_P(0) % 32);
4500 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4501 
4502 	/* Disable the filter so this falls into the default pool. */
4503 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4504 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4505 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4506 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4507 	}
4508 }
4509 
4510 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4511 				  __be16 proto, u16 vid)
4512 {
4513 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4514 	struct ixgbe_hw *hw = &adapter->hw;
4515 
4516 	/* remove VID from filter table */
4517 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4518 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4519 
4520 	clear_bit(vid, adapter->active_vlans);
4521 
4522 	return 0;
4523 }
4524 
4525 /**
4526  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4527  * @adapter: driver data
4528  */
4529 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4530 {
4531 	struct ixgbe_hw *hw = &adapter->hw;
4532 	u32 vlnctrl;
4533 	int i, j;
4534 
4535 	switch (hw->mac.type) {
4536 	case ixgbe_mac_82598EB:
4537 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4538 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4539 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4540 		break;
4541 	case ixgbe_mac_82599EB:
4542 	case ixgbe_mac_X540:
4543 	case ixgbe_mac_X550:
4544 	case ixgbe_mac_X550EM_x:
4545 	case ixgbe_mac_x550em_a:
4546 		for (i = 0; i < adapter->num_rx_queues; i++) {
4547 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4548 
4549 			if (!netif_is_ixgbe(ring->netdev))
4550 				continue;
4551 
4552 			j = ring->reg_idx;
4553 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4554 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4555 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4556 		}
4557 		break;
4558 	default:
4559 		break;
4560 	}
4561 }
4562 
4563 /**
4564  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4565  * @adapter: driver data
4566  */
4567 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4568 {
4569 	struct ixgbe_hw *hw = &adapter->hw;
4570 	u32 vlnctrl;
4571 	int i, j;
4572 
4573 	switch (hw->mac.type) {
4574 	case ixgbe_mac_82598EB:
4575 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4576 		vlnctrl |= IXGBE_VLNCTRL_VME;
4577 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4578 		break;
4579 	case ixgbe_mac_82599EB:
4580 	case ixgbe_mac_X540:
4581 	case ixgbe_mac_X550:
4582 	case ixgbe_mac_X550EM_x:
4583 	case ixgbe_mac_x550em_a:
4584 		for (i = 0; i < adapter->num_rx_queues; i++) {
4585 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4586 
4587 			if (!netif_is_ixgbe(ring->netdev))
4588 				continue;
4589 
4590 			j = ring->reg_idx;
4591 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4592 			vlnctrl |= IXGBE_RXDCTL_VME;
4593 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4594 		}
4595 		break;
4596 	default:
4597 		break;
4598 	}
4599 }
4600 
4601 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4602 {
4603 	struct ixgbe_hw *hw = &adapter->hw;
4604 	u32 vlnctrl, i;
4605 
4606 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4607 
4608 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4609 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4610 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4611 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4612 	} else {
4613 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4614 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4615 		return;
4616 	}
4617 
4618 	/* Nothing to do for 82598 */
4619 	if (hw->mac.type == ixgbe_mac_82598EB)
4620 		return;
4621 
4622 	/* We are already in VLAN promisc, nothing to do */
4623 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4624 		return;
4625 
4626 	/* Set flag so we don't redo unnecessary work */
4627 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4628 
4629 	/* Add PF to all active pools */
4630 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4631 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4632 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4633 
4634 		vlvfb |= BIT(VMDQ_P(0) % 32);
4635 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4636 	}
4637 
4638 	/* Set all bits in the VLAN filter table array */
4639 	for (i = hw->mac.vft_size; i--;)
4640 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4641 }
4642 
4643 #define VFTA_BLOCK_SIZE 8
4644 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4645 {
4646 	struct ixgbe_hw *hw = &adapter->hw;
4647 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4648 	u32 vid_start = vfta_offset * 32;
4649 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4650 	u32 i, vid, word, bits;
4651 
4652 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4653 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4654 
4655 		/* pull VLAN ID from VLVF */
4656 		vid = vlvf & VLAN_VID_MASK;
4657 
4658 		/* only concern outselves with a certain range */
4659 		if (vid < vid_start || vid >= vid_end)
4660 			continue;
4661 
4662 		if (vlvf) {
4663 			/* record VLAN ID in VFTA */
4664 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4665 
4666 			/* if PF is part of this then continue */
4667 			if (test_bit(vid, adapter->active_vlans))
4668 				continue;
4669 		}
4670 
4671 		/* remove PF from the pool */
4672 		word = i * 2 + VMDQ_P(0) / 32;
4673 		bits = ~BIT(VMDQ_P(0) % 32);
4674 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4675 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4676 	}
4677 
4678 	/* extract values from active_vlans and write back to VFTA */
4679 	for (i = VFTA_BLOCK_SIZE; i--;) {
4680 		vid = (vfta_offset + i) * 32;
4681 		word = vid / BITS_PER_LONG;
4682 		bits = vid % BITS_PER_LONG;
4683 
4684 		vfta[i] |= adapter->active_vlans[word] >> bits;
4685 
4686 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4687 	}
4688 }
4689 
4690 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4691 {
4692 	struct ixgbe_hw *hw = &adapter->hw;
4693 	u32 vlnctrl, i;
4694 
4695 	/* Set VLAN filtering to enabled */
4696 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4697 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4698 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4699 
4700 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4701 	    hw->mac.type == ixgbe_mac_82598EB)
4702 		return;
4703 
4704 	/* We are not in VLAN promisc, nothing to do */
4705 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4706 		return;
4707 
4708 	/* Set flag so we don't redo unnecessary work */
4709 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4710 
4711 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4712 		ixgbe_scrub_vfta(adapter, i);
4713 }
4714 
4715 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4716 {
4717 	u16 vid = 1;
4718 
4719 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4720 
4721 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4722 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4723 }
4724 
4725 /**
4726  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4727  * @netdev: network interface device structure
4728  *
4729  * Writes multicast address list to the MTA hash table.
4730  * Returns: -ENOMEM on failure
4731  *                0 on no addresses written
4732  *                X on writing X addresses to MTA
4733  **/
4734 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4735 {
4736 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4737 	struct ixgbe_hw *hw = &adapter->hw;
4738 
4739 	if (!netif_running(netdev))
4740 		return 0;
4741 
4742 	if (hw->mac.ops.update_mc_addr_list)
4743 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4744 	else
4745 		return -ENOMEM;
4746 
4747 #ifdef CONFIG_PCI_IOV
4748 	ixgbe_restore_vf_multicasts(adapter);
4749 #endif
4750 
4751 	return netdev_mc_count(netdev);
4752 }
4753 
4754 #ifdef CONFIG_PCI_IOV
4755 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4756 {
4757 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4758 	struct ixgbe_hw *hw = &adapter->hw;
4759 	int i;
4760 
4761 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4762 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4763 
4764 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4765 			hw->mac.ops.set_rar(hw, i,
4766 					    mac_table->addr,
4767 					    mac_table->pool,
4768 					    IXGBE_RAH_AV);
4769 		else
4770 			hw->mac.ops.clear_rar(hw, i);
4771 	}
4772 }
4773 
4774 #endif
4775 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4776 {
4777 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4778 	struct ixgbe_hw *hw = &adapter->hw;
4779 	int i;
4780 
4781 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4782 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4783 			continue;
4784 
4785 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4786 
4787 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4788 			hw->mac.ops.set_rar(hw, i,
4789 					    mac_table->addr,
4790 					    mac_table->pool,
4791 					    IXGBE_RAH_AV);
4792 		else
4793 			hw->mac.ops.clear_rar(hw, i);
4794 	}
4795 }
4796 
4797 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4798 {
4799 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4800 	struct ixgbe_hw *hw = &adapter->hw;
4801 	int i;
4802 
4803 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4804 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4805 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4806 	}
4807 
4808 	ixgbe_sync_mac_table(adapter);
4809 }
4810 
4811 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4812 {
4813 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4814 	struct ixgbe_hw *hw = &adapter->hw;
4815 	int i, count = 0;
4816 
4817 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4818 		/* do not count default RAR as available */
4819 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4820 			continue;
4821 
4822 		/* only count unused and addresses that belong to us */
4823 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4824 			if (mac_table->pool != pool)
4825 				continue;
4826 		}
4827 
4828 		count++;
4829 	}
4830 
4831 	return count;
4832 }
4833 
4834 /* this function destroys the first RAR entry */
4835 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4836 {
4837 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4838 	struct ixgbe_hw *hw = &adapter->hw;
4839 
4840 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4841 	mac_table->pool = VMDQ_P(0);
4842 
4843 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4844 
4845 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4846 			    IXGBE_RAH_AV);
4847 }
4848 
4849 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4850 			 const u8 *addr, u16 pool)
4851 {
4852 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4853 	struct ixgbe_hw *hw = &adapter->hw;
4854 	int i;
4855 
4856 	if (is_zero_ether_addr(addr))
4857 		return -EINVAL;
4858 
4859 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4860 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4861 			continue;
4862 
4863 		ether_addr_copy(mac_table->addr, addr);
4864 		mac_table->pool = pool;
4865 
4866 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4867 				    IXGBE_MAC_STATE_IN_USE;
4868 
4869 		ixgbe_sync_mac_table(adapter);
4870 
4871 		return i;
4872 	}
4873 
4874 	return -ENOMEM;
4875 }
4876 
4877 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4878 			 const u8 *addr, u16 pool)
4879 {
4880 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4881 	struct ixgbe_hw *hw = &adapter->hw;
4882 	int i;
4883 
4884 	if (is_zero_ether_addr(addr))
4885 		return -EINVAL;
4886 
4887 	/* search table for addr, if found clear IN_USE flag and sync */
4888 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4889 		/* we can only delete an entry if it is in use */
4890 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4891 			continue;
4892 		/* we only care about entries that belong to the given pool */
4893 		if (mac_table->pool != pool)
4894 			continue;
4895 		/* we only care about a specific MAC address */
4896 		if (!ether_addr_equal(addr, mac_table->addr))
4897 			continue;
4898 
4899 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4900 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4901 
4902 		ixgbe_sync_mac_table(adapter);
4903 
4904 		return 0;
4905 	}
4906 
4907 	return -ENOMEM;
4908 }
4909 
4910 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4911 {
4912 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4913 	int ret;
4914 
4915 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4916 
4917 	return min_t(int, ret, 0);
4918 }
4919 
4920 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4921 {
4922 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4923 
4924 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4925 
4926 	return 0;
4927 }
4928 
4929 /**
4930  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4931  * @netdev: network interface device structure
4932  *
4933  * The set_rx_method entry point is called whenever the unicast/multicast
4934  * address list or the network interface flags are updated.  This routine is
4935  * responsible for configuring the hardware for proper unicast, multicast and
4936  * promiscuous mode.
4937  **/
4938 void ixgbe_set_rx_mode(struct net_device *netdev)
4939 {
4940 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4941 	struct ixgbe_hw *hw = &adapter->hw;
4942 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4943 	netdev_features_t features = netdev->features;
4944 	int count;
4945 
4946 	/* Check for Promiscuous and All Multicast modes */
4947 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4948 
4949 	/* set all bits that we expect to always be set */
4950 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4951 	fctrl |= IXGBE_FCTRL_BAM;
4952 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4953 	fctrl |= IXGBE_FCTRL_PMCF;
4954 
4955 	/* clear the bits we are changing the status of */
4956 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4957 	if (netdev->flags & IFF_PROMISC) {
4958 		hw->addr_ctrl.user_set_promisc = true;
4959 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4960 		vmolr |= IXGBE_VMOLR_MPE;
4961 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4962 	} else {
4963 		if (netdev->flags & IFF_ALLMULTI) {
4964 			fctrl |= IXGBE_FCTRL_MPE;
4965 			vmolr |= IXGBE_VMOLR_MPE;
4966 		}
4967 		hw->addr_ctrl.user_set_promisc = false;
4968 	}
4969 
4970 	/*
4971 	 * Write addresses to available RAR registers, if there is not
4972 	 * sufficient space to store all the addresses then enable
4973 	 * unicast promiscuous mode
4974 	 */
4975 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4976 		fctrl |= IXGBE_FCTRL_UPE;
4977 		vmolr |= IXGBE_VMOLR_ROPE;
4978 	}
4979 
4980 	/* Write addresses to the MTA, if the attempt fails
4981 	 * then we should just turn on promiscuous mode so
4982 	 * that we can at least receive multicast traffic
4983 	 */
4984 	count = ixgbe_write_mc_addr_list(netdev);
4985 	if (count < 0) {
4986 		fctrl |= IXGBE_FCTRL_MPE;
4987 		vmolr |= IXGBE_VMOLR_MPE;
4988 	} else if (count) {
4989 		vmolr |= IXGBE_VMOLR_ROMPE;
4990 	}
4991 
4992 	if (hw->mac.type != ixgbe_mac_82598EB) {
4993 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4994 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4995 			   IXGBE_VMOLR_ROPE);
4996 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4997 	}
4998 
4999 	/* This is useful for sniffing bad packets. */
5000 	if (features & NETIF_F_RXALL) {
5001 		/* UPE and MPE will be handled by normal PROMISC logic
5002 		 * in e1000e_set_rx_mode */
5003 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
5004 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
5005 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
5006 
5007 		fctrl &= ~(IXGBE_FCTRL_DPF);
5008 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
5009 	}
5010 
5011 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5012 
5013 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
5014 		ixgbe_vlan_strip_enable(adapter);
5015 	else
5016 		ixgbe_vlan_strip_disable(adapter);
5017 
5018 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
5019 		ixgbe_vlan_promisc_disable(adapter);
5020 	else
5021 		ixgbe_vlan_promisc_enable(adapter);
5022 }
5023 
5024 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
5025 {
5026 	int q_idx;
5027 
5028 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5029 		napi_enable(&adapter->q_vector[q_idx]->napi);
5030 }
5031 
5032 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
5033 {
5034 	int q_idx;
5035 
5036 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5037 		napi_disable(&adapter->q_vector[q_idx]->napi);
5038 }
5039 
5040 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
5041 {
5042 	struct ixgbe_hw *hw = &adapter->hw;
5043 	u32 vxlanctrl;
5044 
5045 	if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
5046 				IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
5047 		return;
5048 
5049 	vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) & ~mask;
5050 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
5051 
5052 	if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
5053 		adapter->vxlan_port = 0;
5054 
5055 	if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
5056 		adapter->geneve_port = 0;
5057 }
5058 
5059 #ifdef CONFIG_IXGBE_DCB
5060 /**
5061  * ixgbe_configure_dcb - Configure DCB hardware
5062  * @adapter: ixgbe adapter struct
5063  *
5064  * This is called by the driver on open to configure the DCB hardware.
5065  * This is also called by the gennetlink interface when reconfiguring
5066  * the DCB state.
5067  */
5068 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5069 {
5070 	struct ixgbe_hw *hw = &adapter->hw;
5071 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5072 
5073 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5074 		if (hw->mac.type == ixgbe_mac_82598EB)
5075 			netif_set_gso_max_size(adapter->netdev, 65536);
5076 		return;
5077 	}
5078 
5079 	if (hw->mac.type == ixgbe_mac_82598EB)
5080 		netif_set_gso_max_size(adapter->netdev, 32768);
5081 
5082 #ifdef IXGBE_FCOE
5083 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
5084 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5085 #endif
5086 
5087 	/* reconfigure the hardware */
5088 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5089 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5090 						DCB_TX_CONFIG);
5091 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5092 						DCB_RX_CONFIG);
5093 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5094 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5095 		ixgbe_dcb_hw_ets(&adapter->hw,
5096 				 adapter->ixgbe_ieee_ets,
5097 				 max_frame);
5098 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
5099 					adapter->ixgbe_ieee_pfc->pfc_en,
5100 					adapter->ixgbe_ieee_ets->prio_tc);
5101 	}
5102 
5103 	/* Enable RSS Hash per TC */
5104 	if (hw->mac.type != ixgbe_mac_82598EB) {
5105 		u32 msb = 0;
5106 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5107 
5108 		while (rss_i) {
5109 			msb++;
5110 			rss_i >>= 1;
5111 		}
5112 
5113 		/* write msb to all 8 TCs in one write */
5114 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5115 	}
5116 }
5117 #endif
5118 
5119 /* Additional bittime to account for IXGBE framing */
5120 #define IXGBE_ETH_FRAMING 20
5121 
5122 /**
5123  * ixgbe_hpbthresh - calculate high water mark for flow control
5124  *
5125  * @adapter: board private structure to calculate for
5126  * @pb: packet buffer to calculate
5127  */
5128 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5129 {
5130 	struct ixgbe_hw *hw = &adapter->hw;
5131 	struct net_device *dev = adapter->netdev;
5132 	int link, tc, kb, marker;
5133 	u32 dv_id, rx_pba;
5134 
5135 	/* Calculate max LAN frame size */
5136 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5137 
5138 #ifdef IXGBE_FCOE
5139 	/* FCoE traffic class uses FCOE jumbo frames */
5140 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5141 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5142 	    (pb == ixgbe_fcoe_get_tc(adapter)))
5143 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5144 #endif
5145 
5146 	/* Calculate delay value for device */
5147 	switch (hw->mac.type) {
5148 	case ixgbe_mac_X540:
5149 	case ixgbe_mac_X550:
5150 	case ixgbe_mac_X550EM_x:
5151 	case ixgbe_mac_x550em_a:
5152 		dv_id = IXGBE_DV_X540(link, tc);
5153 		break;
5154 	default:
5155 		dv_id = IXGBE_DV(link, tc);
5156 		break;
5157 	}
5158 
5159 	/* Loopback switch introduces additional latency */
5160 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5161 		dv_id += IXGBE_B2BT(tc);
5162 
5163 	/* Delay value is calculated in bit times convert to KB */
5164 	kb = IXGBE_BT2KB(dv_id);
5165 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5166 
5167 	marker = rx_pba - kb;
5168 
5169 	/* It is possible that the packet buffer is not large enough
5170 	 * to provide required headroom. In this case throw an error
5171 	 * to user and a do the best we can.
5172 	 */
5173 	if (marker < 0) {
5174 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
5175 			    "headroom to support flow control."
5176 			    "Decrease MTU or number of traffic classes\n", pb);
5177 		marker = tc + 1;
5178 	}
5179 
5180 	return marker;
5181 }
5182 
5183 /**
5184  * ixgbe_lpbthresh - calculate low water mark for for flow control
5185  *
5186  * @adapter: board private structure to calculate for
5187  * @pb: packet buffer to calculate
5188  */
5189 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5190 {
5191 	struct ixgbe_hw *hw = &adapter->hw;
5192 	struct net_device *dev = adapter->netdev;
5193 	int tc;
5194 	u32 dv_id;
5195 
5196 	/* Calculate max LAN frame size */
5197 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5198 
5199 #ifdef IXGBE_FCOE
5200 	/* FCoE traffic class uses FCOE jumbo frames */
5201 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5202 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5203 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5204 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5205 #endif
5206 
5207 	/* Calculate delay value for device */
5208 	switch (hw->mac.type) {
5209 	case ixgbe_mac_X540:
5210 	case ixgbe_mac_X550:
5211 	case ixgbe_mac_X550EM_x:
5212 	case ixgbe_mac_x550em_a:
5213 		dv_id = IXGBE_LOW_DV_X540(tc);
5214 		break;
5215 	default:
5216 		dv_id = IXGBE_LOW_DV(tc);
5217 		break;
5218 	}
5219 
5220 	/* Delay value is calculated in bit times convert to KB */
5221 	return IXGBE_BT2KB(dv_id);
5222 }
5223 
5224 /*
5225  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5226  */
5227 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5228 {
5229 	struct ixgbe_hw *hw = &adapter->hw;
5230 	int num_tc = adapter->hw_tcs;
5231 	int i;
5232 
5233 	if (!num_tc)
5234 		num_tc = 1;
5235 
5236 	for (i = 0; i < num_tc; i++) {
5237 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5238 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5239 
5240 		/* Low water marks must not be larger than high water marks */
5241 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
5242 			hw->fc.low_water[i] = 0;
5243 	}
5244 
5245 	for (; i < MAX_TRAFFIC_CLASS; i++)
5246 		hw->fc.high_water[i] = 0;
5247 }
5248 
5249 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5250 {
5251 	struct ixgbe_hw *hw = &adapter->hw;
5252 	int hdrm;
5253 	u8 tc = adapter->hw_tcs;
5254 
5255 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5256 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5257 		hdrm = 32 << adapter->fdir_pballoc;
5258 	else
5259 		hdrm = 0;
5260 
5261 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5262 	ixgbe_pbthresh_setup(adapter);
5263 }
5264 
5265 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5266 {
5267 	struct ixgbe_hw *hw = &adapter->hw;
5268 	struct hlist_node *node2;
5269 	struct ixgbe_fdir_filter *filter;
5270 
5271 	spin_lock(&adapter->fdir_perfect_lock);
5272 
5273 	if (!hlist_empty(&adapter->fdir_filter_list))
5274 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5275 
5276 	hlist_for_each_entry_safe(filter, node2,
5277 				  &adapter->fdir_filter_list, fdir_node) {
5278 		ixgbe_fdir_write_perfect_filter_82599(hw,
5279 				&filter->filter,
5280 				filter->sw_idx,
5281 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
5282 				IXGBE_FDIR_DROP_QUEUE :
5283 				adapter->rx_ring[filter->action]->reg_idx);
5284 	}
5285 
5286 	spin_unlock(&adapter->fdir_perfect_lock);
5287 }
5288 
5289 /**
5290  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5291  * @rx_ring: ring to free buffers from
5292  **/
5293 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5294 {
5295 	u16 i = rx_ring->next_to_clean;
5296 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5297 
5298 	/* Free all the Rx ring sk_buffs */
5299 	while (i != rx_ring->next_to_alloc) {
5300 		if (rx_buffer->skb) {
5301 			struct sk_buff *skb = rx_buffer->skb;
5302 			if (IXGBE_CB(skb)->page_released)
5303 				dma_unmap_page_attrs(rx_ring->dev,
5304 						     IXGBE_CB(skb)->dma,
5305 						     ixgbe_rx_pg_size(rx_ring),
5306 						     DMA_FROM_DEVICE,
5307 						     IXGBE_RX_DMA_ATTR);
5308 			dev_kfree_skb(skb);
5309 		}
5310 
5311 		/* Invalidate cache lines that may have been written to by
5312 		 * device so that we avoid corrupting memory.
5313 		 */
5314 		dma_sync_single_range_for_cpu(rx_ring->dev,
5315 					      rx_buffer->dma,
5316 					      rx_buffer->page_offset,
5317 					      ixgbe_rx_bufsz(rx_ring),
5318 					      DMA_FROM_DEVICE);
5319 
5320 		/* free resources associated with mapping */
5321 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5322 				     ixgbe_rx_pg_size(rx_ring),
5323 				     DMA_FROM_DEVICE,
5324 				     IXGBE_RX_DMA_ATTR);
5325 		__page_frag_cache_drain(rx_buffer->page,
5326 					rx_buffer->pagecnt_bias);
5327 
5328 		i++;
5329 		rx_buffer++;
5330 		if (i == rx_ring->count) {
5331 			i = 0;
5332 			rx_buffer = rx_ring->rx_buffer_info;
5333 		}
5334 	}
5335 
5336 	rx_ring->next_to_alloc = 0;
5337 	rx_ring->next_to_clean = 0;
5338 	rx_ring->next_to_use = 0;
5339 }
5340 
5341 static int ixgbe_fwd_ring_up(struct ixgbe_adapter *adapter,
5342 			     struct ixgbe_fwd_adapter *accel)
5343 {
5344 	struct net_device *vdev = accel->netdev;
5345 	int i, baseq, err;
5346 
5347 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5348 	netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5349 		   accel->pool, adapter->num_rx_pools,
5350 		   baseq, baseq + adapter->num_rx_queues_per_pool);
5351 
5352 	accel->rx_base_queue = baseq;
5353 	accel->tx_base_queue = baseq;
5354 
5355 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5356 		adapter->rx_ring[baseq + i]->netdev = vdev;
5357 
5358 	/* Guarantee all rings are updated before we update the
5359 	 * MAC address filter.
5360 	 */
5361 	wmb();
5362 
5363 	/* ixgbe_add_mac_filter will return an index if it succeeds, so we
5364 	 * need to only treat it as an error value if it is negative.
5365 	 */
5366 	err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5367 				   VMDQ_P(accel->pool));
5368 	if (err >= 0)
5369 		return 0;
5370 
5371 	/* if we cannot add the MAC rule then disable the offload */
5372 	macvlan_release_l2fw_offload(vdev);
5373 
5374 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5375 		adapter->rx_ring[baseq + i]->netdev = NULL;
5376 
5377 	netdev_err(vdev, "L2FW offload disabled due to L2 filter error\n");
5378 
5379 	clear_bit(accel->pool, adapter->fwd_bitmask);
5380 	kfree(accel);
5381 
5382 	return err;
5383 }
5384 
5385 static int ixgbe_macvlan_up(struct net_device *vdev, void *data)
5386 {
5387 	struct ixgbe_adapter *adapter = data;
5388 	struct ixgbe_fwd_adapter *accel;
5389 
5390 	if (!netif_is_macvlan(vdev))
5391 		return 0;
5392 
5393 	accel = macvlan_accel_priv(vdev);
5394 	if (!accel)
5395 		return 0;
5396 
5397 	ixgbe_fwd_ring_up(adapter, accel);
5398 
5399 	return 0;
5400 }
5401 
5402 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5403 {
5404 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5405 				      ixgbe_macvlan_up, adapter);
5406 }
5407 
5408 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5409 {
5410 	struct ixgbe_hw *hw = &adapter->hw;
5411 
5412 	ixgbe_configure_pb(adapter);
5413 #ifdef CONFIG_IXGBE_DCB
5414 	ixgbe_configure_dcb(adapter);
5415 #endif
5416 	/*
5417 	 * We must restore virtualization before VLANs or else
5418 	 * the VLVF registers will not be populated
5419 	 */
5420 	ixgbe_configure_virtualization(adapter);
5421 
5422 	ixgbe_set_rx_mode(adapter->netdev);
5423 	ixgbe_restore_vlan(adapter);
5424 	ixgbe_ipsec_restore(adapter);
5425 
5426 	switch (hw->mac.type) {
5427 	case ixgbe_mac_82599EB:
5428 	case ixgbe_mac_X540:
5429 		hw->mac.ops.disable_rx_buff(hw);
5430 		break;
5431 	default:
5432 		break;
5433 	}
5434 
5435 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5436 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5437 						adapter->fdir_pballoc);
5438 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5439 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5440 					      adapter->fdir_pballoc);
5441 		ixgbe_fdir_filter_restore(adapter);
5442 	}
5443 
5444 	switch (hw->mac.type) {
5445 	case ixgbe_mac_82599EB:
5446 	case ixgbe_mac_X540:
5447 		hw->mac.ops.enable_rx_buff(hw);
5448 		break;
5449 	default:
5450 		break;
5451 	}
5452 
5453 #ifdef CONFIG_IXGBE_DCA
5454 	/* configure DCA */
5455 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5456 		ixgbe_setup_dca(adapter);
5457 #endif /* CONFIG_IXGBE_DCA */
5458 
5459 #ifdef IXGBE_FCOE
5460 	/* configure FCoE L2 filters, redirection table, and Rx control */
5461 	ixgbe_configure_fcoe(adapter);
5462 
5463 #endif /* IXGBE_FCOE */
5464 	ixgbe_configure_tx(adapter);
5465 	ixgbe_configure_rx(adapter);
5466 	ixgbe_configure_dfwd(adapter);
5467 }
5468 
5469 /**
5470  * ixgbe_sfp_link_config - set up SFP+ link
5471  * @adapter: pointer to private adapter struct
5472  **/
5473 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5474 {
5475 	/*
5476 	 * We are assuming the worst case scenario here, and that
5477 	 * is that an SFP was inserted/removed after the reset
5478 	 * but before SFP detection was enabled.  As such the best
5479 	 * solution is to just start searching as soon as we start
5480 	 */
5481 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5482 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5483 
5484 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5485 	adapter->sfp_poll_time = 0;
5486 }
5487 
5488 /**
5489  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5490  * @hw: pointer to private hardware struct
5491  *
5492  * Returns 0 on success, negative on failure
5493  **/
5494 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5495 {
5496 	u32 speed;
5497 	bool autoneg, link_up = false;
5498 	int ret = IXGBE_ERR_LINK_SETUP;
5499 
5500 	if (hw->mac.ops.check_link)
5501 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5502 
5503 	if (ret)
5504 		return ret;
5505 
5506 	speed = hw->phy.autoneg_advertised;
5507 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5508 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5509 							&autoneg);
5510 	if (ret)
5511 		return ret;
5512 
5513 	if (hw->mac.ops.setup_link)
5514 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5515 
5516 	return ret;
5517 }
5518 
5519 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5520 {
5521 	struct ixgbe_hw *hw = &adapter->hw;
5522 	u32 gpie = 0;
5523 
5524 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5525 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5526 		       IXGBE_GPIE_OCD;
5527 		gpie |= IXGBE_GPIE_EIAME;
5528 		/*
5529 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5530 		 * this saves a register write for every interrupt
5531 		 */
5532 		switch (hw->mac.type) {
5533 		case ixgbe_mac_82598EB:
5534 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5535 			break;
5536 		case ixgbe_mac_82599EB:
5537 		case ixgbe_mac_X540:
5538 		case ixgbe_mac_X550:
5539 		case ixgbe_mac_X550EM_x:
5540 		case ixgbe_mac_x550em_a:
5541 		default:
5542 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5543 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5544 			break;
5545 		}
5546 	} else {
5547 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5548 		 * specifically only auto mask tx and rx interrupts */
5549 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5550 	}
5551 
5552 	/* XXX: to interrupt immediately for EICS writes, enable this */
5553 	/* gpie |= IXGBE_GPIE_EIMEN; */
5554 
5555 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5556 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5557 
5558 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5559 		case IXGBE_82599_VMDQ_8Q_MASK:
5560 			gpie |= IXGBE_GPIE_VTMODE_16;
5561 			break;
5562 		case IXGBE_82599_VMDQ_4Q_MASK:
5563 			gpie |= IXGBE_GPIE_VTMODE_32;
5564 			break;
5565 		default:
5566 			gpie |= IXGBE_GPIE_VTMODE_64;
5567 			break;
5568 		}
5569 	}
5570 
5571 	/* Enable Thermal over heat sensor interrupt */
5572 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5573 		switch (adapter->hw.mac.type) {
5574 		case ixgbe_mac_82599EB:
5575 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5576 			break;
5577 		default:
5578 			break;
5579 		}
5580 	}
5581 
5582 	/* Enable fan failure interrupt */
5583 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5584 		gpie |= IXGBE_SDP1_GPIEN(hw);
5585 
5586 	switch (hw->mac.type) {
5587 	case ixgbe_mac_82599EB:
5588 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5589 		break;
5590 	case ixgbe_mac_X550EM_x:
5591 	case ixgbe_mac_x550em_a:
5592 		gpie |= IXGBE_SDP0_GPIEN_X540;
5593 		break;
5594 	default:
5595 		break;
5596 	}
5597 
5598 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5599 }
5600 
5601 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5602 {
5603 	struct ixgbe_hw *hw = &adapter->hw;
5604 	int err;
5605 	u32 ctrl_ext;
5606 
5607 	ixgbe_get_hw_control(adapter);
5608 	ixgbe_setup_gpie(adapter);
5609 
5610 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5611 		ixgbe_configure_msix(adapter);
5612 	else
5613 		ixgbe_configure_msi_and_legacy(adapter);
5614 
5615 	/* enable the optics for 82599 SFP+ fiber */
5616 	if (hw->mac.ops.enable_tx_laser)
5617 		hw->mac.ops.enable_tx_laser(hw);
5618 
5619 	if (hw->phy.ops.set_phy_power)
5620 		hw->phy.ops.set_phy_power(hw, true);
5621 
5622 	smp_mb__before_atomic();
5623 	clear_bit(__IXGBE_DOWN, &adapter->state);
5624 	ixgbe_napi_enable_all(adapter);
5625 
5626 	if (ixgbe_is_sfp(hw)) {
5627 		ixgbe_sfp_link_config(adapter);
5628 	} else {
5629 		err = ixgbe_non_sfp_link_config(hw);
5630 		if (err)
5631 			e_err(probe, "link_config FAILED %d\n", err);
5632 	}
5633 
5634 	/* clear any pending interrupts, may auto mask */
5635 	IXGBE_READ_REG(hw, IXGBE_EICR);
5636 	ixgbe_irq_enable(adapter, true, true);
5637 
5638 	/*
5639 	 * If this adapter has a fan, check to see if we had a failure
5640 	 * before we enabled the interrupt.
5641 	 */
5642 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5643 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5644 		if (esdp & IXGBE_ESDP_SDP1)
5645 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5646 	}
5647 
5648 	/* bring the link up in the watchdog, this could race with our first
5649 	 * link up interrupt but shouldn't be a problem */
5650 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5651 	adapter->link_check_timeout = jiffies;
5652 	mod_timer(&adapter->service_timer, jiffies);
5653 
5654 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5655 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5656 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5657 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5658 }
5659 
5660 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5661 {
5662 	WARN_ON(in_interrupt());
5663 	/* put off any impending NetWatchDogTimeout */
5664 	netif_trans_update(adapter->netdev);
5665 
5666 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5667 		usleep_range(1000, 2000);
5668 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5669 		ixgbe_watchdog_link_is_down(adapter);
5670 	ixgbe_down(adapter);
5671 	/*
5672 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5673 	 * back up to give the VFs time to respond to the reset.  The
5674 	 * two second wait is based upon the watchdog timer cycle in
5675 	 * the VF driver.
5676 	 */
5677 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5678 		msleep(2000);
5679 	ixgbe_up(adapter);
5680 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5681 }
5682 
5683 void ixgbe_up(struct ixgbe_adapter *adapter)
5684 {
5685 	/* hardware has been reset, we need to reload some things */
5686 	ixgbe_configure(adapter);
5687 
5688 	ixgbe_up_complete(adapter);
5689 }
5690 
5691 void ixgbe_reset(struct ixgbe_adapter *adapter)
5692 {
5693 	struct ixgbe_hw *hw = &adapter->hw;
5694 	struct net_device *netdev = adapter->netdev;
5695 	int err;
5696 
5697 	if (ixgbe_removed(hw->hw_addr))
5698 		return;
5699 	/* lock SFP init bit to prevent race conditions with the watchdog */
5700 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5701 		usleep_range(1000, 2000);
5702 
5703 	/* clear all SFP and link config related flags while holding SFP_INIT */
5704 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5705 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5706 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5707 
5708 	err = hw->mac.ops.init_hw(hw);
5709 	switch (err) {
5710 	case 0:
5711 	case IXGBE_ERR_SFP_NOT_PRESENT:
5712 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5713 		break;
5714 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5715 		e_dev_err("master disable timed out\n");
5716 		break;
5717 	case IXGBE_ERR_EEPROM_VERSION:
5718 		/* We are running on a pre-production device, log a warning */
5719 		e_dev_warn("This device is a pre-production adapter/LOM. "
5720 			   "Please be aware there may be issues associated with "
5721 			   "your hardware.  If you are experiencing problems "
5722 			   "please contact your Intel or hardware "
5723 			   "representative who provided you with this "
5724 			   "hardware.\n");
5725 		break;
5726 	default:
5727 		e_dev_err("Hardware Error: %d\n", err);
5728 	}
5729 
5730 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5731 
5732 	/* flush entries out of MAC table */
5733 	ixgbe_flush_sw_mac_table(adapter);
5734 	__dev_uc_unsync(netdev, NULL);
5735 
5736 	/* do not flush user set addresses */
5737 	ixgbe_mac_set_default_filter(adapter);
5738 
5739 	/* update SAN MAC vmdq pool selection */
5740 	if (hw->mac.san_mac_rar_index)
5741 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5742 
5743 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5744 		ixgbe_ptp_reset(adapter);
5745 
5746 	if (hw->phy.ops.set_phy_power) {
5747 		if (!netif_running(adapter->netdev) && !adapter->wol)
5748 			hw->phy.ops.set_phy_power(hw, false);
5749 		else
5750 			hw->phy.ops.set_phy_power(hw, true);
5751 	}
5752 }
5753 
5754 /**
5755  * ixgbe_clean_tx_ring - Free Tx Buffers
5756  * @tx_ring: ring to be cleaned
5757  **/
5758 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5759 {
5760 	u16 i = tx_ring->next_to_clean;
5761 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
5762 
5763 	while (i != tx_ring->next_to_use) {
5764 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
5765 
5766 		/* Free all the Tx ring sk_buffs */
5767 		if (ring_is_xdp(tx_ring))
5768 			xdp_return_frame(tx_buffer->xdpf);
5769 		else
5770 			dev_kfree_skb_any(tx_buffer->skb);
5771 
5772 		/* unmap skb header data */
5773 		dma_unmap_single(tx_ring->dev,
5774 				 dma_unmap_addr(tx_buffer, dma),
5775 				 dma_unmap_len(tx_buffer, len),
5776 				 DMA_TO_DEVICE);
5777 
5778 		/* check for eop_desc to determine the end of the packet */
5779 		eop_desc = tx_buffer->next_to_watch;
5780 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
5781 
5782 		/* unmap remaining buffers */
5783 		while (tx_desc != eop_desc) {
5784 			tx_buffer++;
5785 			tx_desc++;
5786 			i++;
5787 			if (unlikely(i == tx_ring->count)) {
5788 				i = 0;
5789 				tx_buffer = tx_ring->tx_buffer_info;
5790 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
5791 			}
5792 
5793 			/* unmap any remaining paged data */
5794 			if (dma_unmap_len(tx_buffer, len))
5795 				dma_unmap_page(tx_ring->dev,
5796 					       dma_unmap_addr(tx_buffer, dma),
5797 					       dma_unmap_len(tx_buffer, len),
5798 					       DMA_TO_DEVICE);
5799 		}
5800 
5801 		/* move us one more past the eop_desc for start of next pkt */
5802 		tx_buffer++;
5803 		i++;
5804 		if (unlikely(i == tx_ring->count)) {
5805 			i = 0;
5806 			tx_buffer = tx_ring->tx_buffer_info;
5807 		}
5808 	}
5809 
5810 	/* reset BQL for queue */
5811 	if (!ring_is_xdp(tx_ring))
5812 		netdev_tx_reset_queue(txring_txq(tx_ring));
5813 
5814 	/* reset next_to_use and next_to_clean */
5815 	tx_ring->next_to_use = 0;
5816 	tx_ring->next_to_clean = 0;
5817 }
5818 
5819 /**
5820  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5821  * @adapter: board private structure
5822  **/
5823 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5824 {
5825 	int i;
5826 
5827 	for (i = 0; i < adapter->num_rx_queues; i++)
5828 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5829 }
5830 
5831 /**
5832  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5833  * @adapter: board private structure
5834  **/
5835 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5836 {
5837 	int i;
5838 
5839 	for (i = 0; i < adapter->num_tx_queues; i++)
5840 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5841 	for (i = 0; i < adapter->num_xdp_queues; i++)
5842 		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
5843 }
5844 
5845 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5846 {
5847 	struct hlist_node *node2;
5848 	struct ixgbe_fdir_filter *filter;
5849 
5850 	spin_lock(&adapter->fdir_perfect_lock);
5851 
5852 	hlist_for_each_entry_safe(filter, node2,
5853 				  &adapter->fdir_filter_list, fdir_node) {
5854 		hlist_del(&filter->fdir_node);
5855 		kfree(filter);
5856 	}
5857 	adapter->fdir_filter_count = 0;
5858 
5859 	spin_unlock(&adapter->fdir_perfect_lock);
5860 }
5861 
5862 void ixgbe_down(struct ixgbe_adapter *adapter)
5863 {
5864 	struct net_device *netdev = adapter->netdev;
5865 	struct ixgbe_hw *hw = &adapter->hw;
5866 	int i;
5867 
5868 	/* signal that we are down to the interrupt handler */
5869 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5870 		return; /* do nothing if already down */
5871 
5872 	/* disable receives */
5873 	hw->mac.ops.disable_rx(hw);
5874 
5875 	/* disable all enabled rx queues */
5876 	for (i = 0; i < adapter->num_rx_queues; i++)
5877 		/* this call also flushes the previous write */
5878 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5879 
5880 	usleep_range(10000, 20000);
5881 
5882 	/* synchronize_sched() needed for pending XDP buffers to drain */
5883 	if (adapter->xdp_ring[0])
5884 		synchronize_sched();
5885 	netif_tx_stop_all_queues(netdev);
5886 
5887 	/* call carrier off first to avoid false dev_watchdog timeouts */
5888 	netif_carrier_off(netdev);
5889 	netif_tx_disable(netdev);
5890 
5891 	ixgbe_irq_disable(adapter);
5892 
5893 	ixgbe_napi_disable_all(adapter);
5894 
5895 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
5896 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5897 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5898 
5899 	del_timer_sync(&adapter->service_timer);
5900 
5901 	if (adapter->num_vfs) {
5902 		/* Clear EITR Select mapping */
5903 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5904 
5905 		/* Mark all the VFs as inactive */
5906 		for (i = 0 ; i < adapter->num_vfs; i++)
5907 			adapter->vfinfo[i].clear_to_send = false;
5908 
5909 		/* ping all the active vfs to let them know we are going down */
5910 		ixgbe_ping_all_vfs(adapter);
5911 
5912 		/* Disable all VFTE/VFRE TX/RX */
5913 		ixgbe_disable_tx_rx(adapter);
5914 	}
5915 
5916 	/* disable transmits in the hardware now that interrupts are off */
5917 	for (i = 0; i < adapter->num_tx_queues; i++) {
5918 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5919 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5920 	}
5921 	for (i = 0; i < adapter->num_xdp_queues; i++) {
5922 		u8 reg_idx = adapter->xdp_ring[i]->reg_idx;
5923 
5924 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5925 	}
5926 
5927 	/* Disable the Tx DMA engine on 82599 and later MAC */
5928 	switch (hw->mac.type) {
5929 	case ixgbe_mac_82599EB:
5930 	case ixgbe_mac_X540:
5931 	case ixgbe_mac_X550:
5932 	case ixgbe_mac_X550EM_x:
5933 	case ixgbe_mac_x550em_a:
5934 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5935 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5936 				 ~IXGBE_DMATXCTL_TE));
5937 		break;
5938 	default:
5939 		break;
5940 	}
5941 
5942 	if (!pci_channel_offline(adapter->pdev))
5943 		ixgbe_reset(adapter);
5944 
5945 	/* power down the optics for 82599 SFP+ fiber */
5946 	if (hw->mac.ops.disable_tx_laser)
5947 		hw->mac.ops.disable_tx_laser(hw);
5948 
5949 	ixgbe_clean_all_tx_rings(adapter);
5950 	ixgbe_clean_all_rx_rings(adapter);
5951 }
5952 
5953 /**
5954  * ixgbe_eee_capable - helper function to determine EEE support on X550
5955  * @adapter: board private structure
5956  */
5957 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
5958 {
5959 	struct ixgbe_hw *hw = &adapter->hw;
5960 
5961 	switch (hw->device_id) {
5962 	case IXGBE_DEV_ID_X550EM_A_1G_T:
5963 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
5964 		if (!hw->phy.eee_speeds_supported)
5965 			break;
5966 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
5967 		if (!hw->phy.eee_speeds_advertised)
5968 			break;
5969 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
5970 		break;
5971 	default:
5972 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
5973 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
5974 		break;
5975 	}
5976 }
5977 
5978 /**
5979  * ixgbe_tx_timeout - Respond to a Tx Hang
5980  * @netdev: network interface device structure
5981  **/
5982 static void ixgbe_tx_timeout(struct net_device *netdev)
5983 {
5984 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5985 
5986 	/* Do the reset outside of interrupt context */
5987 	ixgbe_tx_timeout_reset(adapter);
5988 }
5989 
5990 #ifdef CONFIG_IXGBE_DCB
5991 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
5992 {
5993 	struct ixgbe_hw *hw = &adapter->hw;
5994 	struct tc_configuration *tc;
5995 	int j;
5996 
5997 	switch (hw->mac.type) {
5998 	case ixgbe_mac_82598EB:
5999 	case ixgbe_mac_82599EB:
6000 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
6001 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
6002 		break;
6003 	case ixgbe_mac_X540:
6004 	case ixgbe_mac_X550:
6005 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
6006 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
6007 		break;
6008 	case ixgbe_mac_X550EM_x:
6009 	case ixgbe_mac_x550em_a:
6010 	default:
6011 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
6012 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
6013 		break;
6014 	}
6015 
6016 	/* Configure DCB traffic classes */
6017 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
6018 		tc = &adapter->dcb_cfg.tc_config[j];
6019 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
6020 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
6021 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
6022 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
6023 		tc->dcb_pfc = pfc_disabled;
6024 	}
6025 
6026 	/* Initialize default user to priority mapping, UPx->TC0 */
6027 	tc = &adapter->dcb_cfg.tc_config[0];
6028 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
6029 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
6030 
6031 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
6032 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
6033 	adapter->dcb_cfg.pfc_mode_enable = false;
6034 	adapter->dcb_set_bitmap = 0x00;
6035 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
6036 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
6037 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
6038 	       sizeof(adapter->temp_dcb_cfg));
6039 }
6040 #endif
6041 
6042 /**
6043  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
6044  * @adapter: board private structure to initialize
6045  * @ii: pointer to ixgbe_info for device
6046  *
6047  * ixgbe_sw_init initializes the Adapter private data structure.
6048  * Fields are initialized based on PCI device information and
6049  * OS network device settings (MTU size).
6050  **/
6051 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
6052 			 const struct ixgbe_info *ii)
6053 {
6054 	struct ixgbe_hw *hw = &adapter->hw;
6055 	struct pci_dev *pdev = adapter->pdev;
6056 	unsigned int rss, fdir;
6057 	u32 fwsm;
6058 	int i;
6059 
6060 	/* PCI config space info */
6061 
6062 	hw->vendor_id = pdev->vendor;
6063 	hw->device_id = pdev->device;
6064 	hw->revision_id = pdev->revision;
6065 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6066 	hw->subsystem_device_id = pdev->subsystem_device;
6067 
6068 	/* get_invariants needs the device IDs */
6069 	ii->get_invariants(hw);
6070 
6071 	/* Set common capability flags and settings */
6072 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6073 	adapter->ring_feature[RING_F_RSS].limit = rss;
6074 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6075 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6076 	adapter->atr_sample_rate = 20;
6077 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6078 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
6079 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6080 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
6081 #ifdef CONFIG_IXGBE_DCA
6082 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6083 #endif
6084 #ifdef CONFIG_IXGBE_DCB
6085 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6086 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6087 #endif
6088 #ifdef IXGBE_FCOE
6089 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6090 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6091 #ifdef CONFIG_IXGBE_DCB
6092 	/* Default traffic class to use for FCoE */
6093 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6094 #endif /* CONFIG_IXGBE_DCB */
6095 #endif /* IXGBE_FCOE */
6096 
6097 	/* initialize static ixgbe jump table entries */
6098 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6099 					  GFP_KERNEL);
6100 	if (!adapter->jump_tables[0])
6101 		return -ENOMEM;
6102 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6103 
6104 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6105 		adapter->jump_tables[i] = NULL;
6106 
6107 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
6108 				     hw->mac.num_rar_entries,
6109 				     GFP_ATOMIC);
6110 	if (!adapter->mac_table)
6111 		return -ENOMEM;
6112 
6113 	if (ixgbe_init_rss_key(adapter))
6114 		return -ENOMEM;
6115 
6116 	/* Set MAC specific capability flags and exceptions */
6117 	switch (hw->mac.type) {
6118 	case ixgbe_mac_82598EB:
6119 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6120 
6121 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
6122 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6123 
6124 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6125 		adapter->ring_feature[RING_F_FDIR].limit = 0;
6126 		adapter->atr_sample_rate = 0;
6127 		adapter->fdir_pballoc = 0;
6128 #ifdef IXGBE_FCOE
6129 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6130 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6131 #ifdef CONFIG_IXGBE_DCB
6132 		adapter->fcoe.up = 0;
6133 #endif /* IXGBE_DCB */
6134 #endif /* IXGBE_FCOE */
6135 		break;
6136 	case ixgbe_mac_82599EB:
6137 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6138 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6139 		break;
6140 	case ixgbe_mac_X540:
6141 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6142 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
6143 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6144 		break;
6145 	case ixgbe_mac_x550em_a:
6146 		adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
6147 		switch (hw->device_id) {
6148 		case IXGBE_DEV_ID_X550EM_A_1G_T:
6149 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6150 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6151 			break;
6152 		default:
6153 			break;
6154 		}
6155 	/* fall through */
6156 	case ixgbe_mac_X550EM_x:
6157 #ifdef CONFIG_IXGBE_DCB
6158 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6159 #endif
6160 #ifdef IXGBE_FCOE
6161 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6162 #ifdef CONFIG_IXGBE_DCB
6163 		adapter->fcoe.up = 0;
6164 #endif /* IXGBE_DCB */
6165 #endif /* IXGBE_FCOE */
6166 	/* Fall Through */
6167 	case ixgbe_mac_X550:
6168 		if (hw->mac.type == ixgbe_mac_X550)
6169 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6170 #ifdef CONFIG_IXGBE_DCA
6171 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6172 #endif
6173 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
6174 		break;
6175 	default:
6176 		break;
6177 	}
6178 
6179 #ifdef IXGBE_FCOE
6180 	/* FCoE support exists, always init the FCoE lock */
6181 	spin_lock_init(&adapter->fcoe.lock);
6182 
6183 #endif
6184 	/* n-tuple support exists, always init our spinlock */
6185 	spin_lock_init(&adapter->fdir_perfect_lock);
6186 
6187 #ifdef CONFIG_IXGBE_DCB
6188 	ixgbe_init_dcb(adapter);
6189 #endif
6190 
6191 	/* default flow control settings */
6192 	hw->fc.requested_mode = ixgbe_fc_full;
6193 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
6194 	ixgbe_pbthresh_setup(adapter);
6195 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6196 	hw->fc.send_xon = true;
6197 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6198 
6199 #ifdef CONFIG_PCI_IOV
6200 	if (max_vfs > 0)
6201 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6202 
6203 	/* assign number of SR-IOV VFs */
6204 	if (hw->mac.type != ixgbe_mac_82598EB) {
6205 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6206 			max_vfs = 0;
6207 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6208 		}
6209 	}
6210 #endif /* CONFIG_PCI_IOV */
6211 
6212 	/* enable itr by default in dynamic mode */
6213 	adapter->rx_itr_setting = 1;
6214 	adapter->tx_itr_setting = 1;
6215 
6216 	/* set default ring sizes */
6217 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6218 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6219 
6220 	/* set default work limits */
6221 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6222 
6223 	/* initialize eeprom parameters */
6224 	if (ixgbe_init_eeprom_params_generic(hw)) {
6225 		e_dev_err("EEPROM initialization failed\n");
6226 		return -EIO;
6227 	}
6228 
6229 	/* PF holds first pool slot */
6230 	set_bit(0, adapter->fwd_bitmask);
6231 	set_bit(__IXGBE_DOWN, &adapter->state);
6232 
6233 	return 0;
6234 }
6235 
6236 /**
6237  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6238  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6239  *
6240  * Return 0 on success, negative on failure
6241  **/
6242 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6243 {
6244 	struct device *dev = tx_ring->dev;
6245 	int orig_node = dev_to_node(dev);
6246 	int ring_node = -1;
6247 	int size;
6248 
6249 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6250 
6251 	if (tx_ring->q_vector)
6252 		ring_node = tx_ring->q_vector->numa_node;
6253 
6254 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6255 	if (!tx_ring->tx_buffer_info)
6256 		tx_ring->tx_buffer_info = vmalloc(size);
6257 	if (!tx_ring->tx_buffer_info)
6258 		goto err;
6259 
6260 	/* round up to nearest 4K */
6261 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6262 	tx_ring->size = ALIGN(tx_ring->size, 4096);
6263 
6264 	set_dev_node(dev, ring_node);
6265 	tx_ring->desc = dma_alloc_coherent(dev,
6266 					   tx_ring->size,
6267 					   &tx_ring->dma,
6268 					   GFP_KERNEL);
6269 	set_dev_node(dev, orig_node);
6270 	if (!tx_ring->desc)
6271 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6272 						   &tx_ring->dma, GFP_KERNEL);
6273 	if (!tx_ring->desc)
6274 		goto err;
6275 
6276 	tx_ring->next_to_use = 0;
6277 	tx_ring->next_to_clean = 0;
6278 	return 0;
6279 
6280 err:
6281 	vfree(tx_ring->tx_buffer_info);
6282 	tx_ring->tx_buffer_info = NULL;
6283 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6284 	return -ENOMEM;
6285 }
6286 
6287 /**
6288  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6289  * @adapter: board private structure
6290  *
6291  * If this function returns with an error, then it's possible one or
6292  * more of the rings is populated (while the rest are not).  It is the
6293  * callers duty to clean those orphaned rings.
6294  *
6295  * Return 0 on success, negative on failure
6296  **/
6297 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6298 {
6299 	int i, j = 0, err = 0;
6300 
6301 	for (i = 0; i < adapter->num_tx_queues; i++) {
6302 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6303 		if (!err)
6304 			continue;
6305 
6306 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6307 		goto err_setup_tx;
6308 	}
6309 	for (j = 0; j < adapter->num_xdp_queues; j++) {
6310 		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6311 		if (!err)
6312 			continue;
6313 
6314 		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6315 		goto err_setup_tx;
6316 	}
6317 
6318 	return 0;
6319 err_setup_tx:
6320 	/* rewind the index freeing the rings as we go */
6321 	while (j--)
6322 		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6323 	while (i--)
6324 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
6325 	return err;
6326 }
6327 
6328 /**
6329  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6330  * @adapter: pointer to ixgbe_adapter
6331  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6332  *
6333  * Returns 0 on success, negative on failure
6334  **/
6335 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6336 			     struct ixgbe_ring *rx_ring)
6337 {
6338 	struct device *dev = rx_ring->dev;
6339 	int orig_node = dev_to_node(dev);
6340 	int ring_node = -1;
6341 	int size, err;
6342 
6343 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6344 
6345 	if (rx_ring->q_vector)
6346 		ring_node = rx_ring->q_vector->numa_node;
6347 
6348 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6349 	if (!rx_ring->rx_buffer_info)
6350 		rx_ring->rx_buffer_info = vmalloc(size);
6351 	if (!rx_ring->rx_buffer_info)
6352 		goto err;
6353 
6354 	/* Round up to nearest 4K */
6355 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6356 	rx_ring->size = ALIGN(rx_ring->size, 4096);
6357 
6358 	set_dev_node(dev, ring_node);
6359 	rx_ring->desc = dma_alloc_coherent(dev,
6360 					   rx_ring->size,
6361 					   &rx_ring->dma,
6362 					   GFP_KERNEL);
6363 	set_dev_node(dev, orig_node);
6364 	if (!rx_ring->desc)
6365 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6366 						   &rx_ring->dma, GFP_KERNEL);
6367 	if (!rx_ring->desc)
6368 		goto err;
6369 
6370 	rx_ring->next_to_clean = 0;
6371 	rx_ring->next_to_use = 0;
6372 
6373 	/* XDP RX-queue info */
6374 	if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
6375 			     rx_ring->queue_index) < 0)
6376 		goto err;
6377 
6378 	err = xdp_rxq_info_reg_mem_model(&rx_ring->xdp_rxq,
6379 					 MEM_TYPE_PAGE_SHARED, NULL);
6380 	if (err) {
6381 		xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6382 		goto err;
6383 	}
6384 
6385 	rx_ring->xdp_prog = adapter->xdp_prog;
6386 
6387 	return 0;
6388 err:
6389 	vfree(rx_ring->rx_buffer_info);
6390 	rx_ring->rx_buffer_info = NULL;
6391 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6392 	return -ENOMEM;
6393 }
6394 
6395 /**
6396  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6397  * @adapter: board private structure
6398  *
6399  * If this function returns with an error, then it's possible one or
6400  * more of the rings is populated (while the rest are not).  It is the
6401  * callers duty to clean those orphaned rings.
6402  *
6403  * Return 0 on success, negative on failure
6404  **/
6405 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6406 {
6407 	int i, err = 0;
6408 
6409 	for (i = 0; i < adapter->num_rx_queues; i++) {
6410 		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6411 		if (!err)
6412 			continue;
6413 
6414 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6415 		goto err_setup_rx;
6416 	}
6417 
6418 #ifdef IXGBE_FCOE
6419 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
6420 	if (!err)
6421 #endif
6422 		return 0;
6423 err_setup_rx:
6424 	/* rewind the index freeing the rings as we go */
6425 	while (i--)
6426 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
6427 	return err;
6428 }
6429 
6430 /**
6431  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6432  * @tx_ring: Tx descriptor ring for a specific queue
6433  *
6434  * Free all transmit software resources
6435  **/
6436 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6437 {
6438 	ixgbe_clean_tx_ring(tx_ring);
6439 
6440 	vfree(tx_ring->tx_buffer_info);
6441 	tx_ring->tx_buffer_info = NULL;
6442 
6443 	/* if not set, then don't free */
6444 	if (!tx_ring->desc)
6445 		return;
6446 
6447 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6448 			  tx_ring->desc, tx_ring->dma);
6449 
6450 	tx_ring->desc = NULL;
6451 }
6452 
6453 /**
6454  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6455  * @adapter: board private structure
6456  *
6457  * Free all transmit software resources
6458  **/
6459 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6460 {
6461 	int i;
6462 
6463 	for (i = 0; i < adapter->num_tx_queues; i++)
6464 		if (adapter->tx_ring[i]->desc)
6465 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6466 	for (i = 0; i < adapter->num_xdp_queues; i++)
6467 		if (adapter->xdp_ring[i]->desc)
6468 			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6469 }
6470 
6471 /**
6472  * ixgbe_free_rx_resources - Free Rx Resources
6473  * @rx_ring: ring to clean the resources from
6474  *
6475  * Free all receive software resources
6476  **/
6477 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6478 {
6479 	ixgbe_clean_rx_ring(rx_ring);
6480 
6481 	rx_ring->xdp_prog = NULL;
6482 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6483 	vfree(rx_ring->rx_buffer_info);
6484 	rx_ring->rx_buffer_info = NULL;
6485 
6486 	/* if not set, then don't free */
6487 	if (!rx_ring->desc)
6488 		return;
6489 
6490 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6491 			  rx_ring->desc, rx_ring->dma);
6492 
6493 	rx_ring->desc = NULL;
6494 }
6495 
6496 /**
6497  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6498  * @adapter: board private structure
6499  *
6500  * Free all receive software resources
6501  **/
6502 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6503 {
6504 	int i;
6505 
6506 #ifdef IXGBE_FCOE
6507 	ixgbe_free_fcoe_ddp_resources(adapter);
6508 
6509 #endif
6510 	for (i = 0; i < adapter->num_rx_queues; i++)
6511 		if (adapter->rx_ring[i]->desc)
6512 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6513 }
6514 
6515 /**
6516  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6517  * @netdev: network interface device structure
6518  * @new_mtu: new value for maximum frame size
6519  *
6520  * Returns 0 on success, negative on failure
6521  **/
6522 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6523 {
6524 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6525 
6526 	/*
6527 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6528 	 * paths when MTU greater than 1500 is configured.  So display a
6529 	 * warning that legacy VFs will be disabled.
6530 	 */
6531 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6532 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6533 	    (new_mtu > ETH_DATA_LEN))
6534 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6535 
6536 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6537 
6538 	/* must set new MTU before calling down or up */
6539 	netdev->mtu = new_mtu;
6540 
6541 	if (netif_running(netdev))
6542 		ixgbe_reinit_locked(adapter);
6543 
6544 	return 0;
6545 }
6546 
6547 /**
6548  * ixgbe_open - Called when a network interface is made active
6549  * @netdev: network interface device structure
6550  *
6551  * Returns 0 on success, negative value on failure
6552  *
6553  * The open entry point is called when a network interface is made
6554  * active by the system (IFF_UP).  At this point all resources needed
6555  * for transmit and receive operations are allocated, the interrupt
6556  * handler is registered with the OS, the watchdog timer is started,
6557  * and the stack is notified that the interface is ready.
6558  **/
6559 int ixgbe_open(struct net_device *netdev)
6560 {
6561 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6562 	struct ixgbe_hw *hw = &adapter->hw;
6563 	int err, queues;
6564 
6565 	/* disallow open during test */
6566 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6567 		return -EBUSY;
6568 
6569 	netif_carrier_off(netdev);
6570 
6571 	/* allocate transmit descriptors */
6572 	err = ixgbe_setup_all_tx_resources(adapter);
6573 	if (err)
6574 		goto err_setup_tx;
6575 
6576 	/* allocate receive descriptors */
6577 	err = ixgbe_setup_all_rx_resources(adapter);
6578 	if (err)
6579 		goto err_setup_rx;
6580 
6581 	ixgbe_configure(adapter);
6582 
6583 	err = ixgbe_request_irq(adapter);
6584 	if (err)
6585 		goto err_req_irq;
6586 
6587 	/* Notify the stack of the actual queue counts. */
6588 	queues = adapter->num_tx_queues;
6589 	err = netif_set_real_num_tx_queues(netdev, queues);
6590 	if (err)
6591 		goto err_set_queues;
6592 
6593 	queues = adapter->num_rx_queues;
6594 	err = netif_set_real_num_rx_queues(netdev, queues);
6595 	if (err)
6596 		goto err_set_queues;
6597 
6598 	ixgbe_ptp_init(adapter);
6599 
6600 	ixgbe_up_complete(adapter);
6601 
6602 	ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6603 	udp_tunnel_get_rx_info(netdev);
6604 
6605 	return 0;
6606 
6607 err_set_queues:
6608 	ixgbe_free_irq(adapter);
6609 err_req_irq:
6610 	ixgbe_free_all_rx_resources(adapter);
6611 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6612 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6613 err_setup_rx:
6614 	ixgbe_free_all_tx_resources(adapter);
6615 err_setup_tx:
6616 	ixgbe_reset(adapter);
6617 
6618 	return err;
6619 }
6620 
6621 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6622 {
6623 	ixgbe_ptp_suspend(adapter);
6624 
6625 	if (adapter->hw.phy.ops.enter_lplu) {
6626 		adapter->hw.phy.reset_disable = true;
6627 		ixgbe_down(adapter);
6628 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6629 		adapter->hw.phy.reset_disable = false;
6630 	} else {
6631 		ixgbe_down(adapter);
6632 	}
6633 
6634 	ixgbe_free_irq(adapter);
6635 
6636 	ixgbe_free_all_tx_resources(adapter);
6637 	ixgbe_free_all_rx_resources(adapter);
6638 }
6639 
6640 /**
6641  * ixgbe_close - Disables a network interface
6642  * @netdev: network interface device structure
6643  *
6644  * Returns 0, this is not allowed to fail
6645  *
6646  * The close entry point is called when an interface is de-activated
6647  * by the OS.  The hardware is still under the drivers control, but
6648  * needs to be disabled.  A global MAC reset is issued to stop the
6649  * hardware, and all transmit and receive resources are freed.
6650  **/
6651 int ixgbe_close(struct net_device *netdev)
6652 {
6653 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6654 
6655 	ixgbe_ptp_stop(adapter);
6656 
6657 	if (netif_device_present(netdev))
6658 		ixgbe_close_suspend(adapter);
6659 
6660 	ixgbe_fdir_filter_exit(adapter);
6661 
6662 	ixgbe_release_hw_control(adapter);
6663 
6664 	return 0;
6665 }
6666 
6667 #ifdef CONFIG_PM
6668 static int ixgbe_resume(struct pci_dev *pdev)
6669 {
6670 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6671 	struct net_device *netdev = adapter->netdev;
6672 	u32 err;
6673 
6674 	adapter->hw.hw_addr = adapter->io_addr;
6675 	pci_set_power_state(pdev, PCI_D0);
6676 	pci_restore_state(pdev);
6677 	/*
6678 	 * pci_restore_state clears dev->state_saved so call
6679 	 * pci_save_state to restore it.
6680 	 */
6681 	pci_save_state(pdev);
6682 
6683 	err = pci_enable_device_mem(pdev);
6684 	if (err) {
6685 		e_dev_err("Cannot enable PCI device from suspend\n");
6686 		return err;
6687 	}
6688 	smp_mb__before_atomic();
6689 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6690 	pci_set_master(pdev);
6691 
6692 	pci_wake_from_d3(pdev, false);
6693 
6694 	ixgbe_reset(adapter);
6695 
6696 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6697 
6698 	rtnl_lock();
6699 	err = ixgbe_init_interrupt_scheme(adapter);
6700 	if (!err && netif_running(netdev))
6701 		err = ixgbe_open(netdev);
6702 
6703 
6704 	if (!err)
6705 		netif_device_attach(netdev);
6706 	rtnl_unlock();
6707 
6708 	return err;
6709 }
6710 #endif /* CONFIG_PM */
6711 
6712 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6713 {
6714 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6715 	struct net_device *netdev = adapter->netdev;
6716 	struct ixgbe_hw *hw = &adapter->hw;
6717 	u32 ctrl;
6718 	u32 wufc = adapter->wol;
6719 #ifdef CONFIG_PM
6720 	int retval = 0;
6721 #endif
6722 
6723 	rtnl_lock();
6724 	netif_device_detach(netdev);
6725 
6726 	if (netif_running(netdev))
6727 		ixgbe_close_suspend(adapter);
6728 
6729 	ixgbe_clear_interrupt_scheme(adapter);
6730 	rtnl_unlock();
6731 
6732 #ifdef CONFIG_PM
6733 	retval = pci_save_state(pdev);
6734 	if (retval)
6735 		return retval;
6736 
6737 #endif
6738 	if (hw->mac.ops.stop_link_on_d3)
6739 		hw->mac.ops.stop_link_on_d3(hw);
6740 
6741 	if (wufc) {
6742 		u32 fctrl;
6743 
6744 		ixgbe_set_rx_mode(netdev);
6745 
6746 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6747 		if (hw->mac.ops.enable_tx_laser)
6748 			hw->mac.ops.enable_tx_laser(hw);
6749 
6750 		/* enable the reception of multicast packets */
6751 		fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6752 		fctrl |= IXGBE_FCTRL_MPE;
6753 		IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6754 
6755 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6756 		ctrl |= IXGBE_CTRL_GIO_DIS;
6757 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6758 
6759 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6760 	} else {
6761 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6762 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6763 	}
6764 
6765 	switch (hw->mac.type) {
6766 	case ixgbe_mac_82598EB:
6767 		pci_wake_from_d3(pdev, false);
6768 		break;
6769 	case ixgbe_mac_82599EB:
6770 	case ixgbe_mac_X540:
6771 	case ixgbe_mac_X550:
6772 	case ixgbe_mac_X550EM_x:
6773 	case ixgbe_mac_x550em_a:
6774 		pci_wake_from_d3(pdev, !!wufc);
6775 		break;
6776 	default:
6777 		break;
6778 	}
6779 
6780 	*enable_wake = !!wufc;
6781 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6782 		hw->phy.ops.set_phy_power(hw, false);
6783 
6784 	ixgbe_release_hw_control(adapter);
6785 
6786 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6787 		pci_disable_device(pdev);
6788 
6789 	return 0;
6790 }
6791 
6792 #ifdef CONFIG_PM
6793 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6794 {
6795 	int retval;
6796 	bool wake;
6797 
6798 	retval = __ixgbe_shutdown(pdev, &wake);
6799 	if (retval)
6800 		return retval;
6801 
6802 	if (wake) {
6803 		pci_prepare_to_sleep(pdev);
6804 	} else {
6805 		pci_wake_from_d3(pdev, false);
6806 		pci_set_power_state(pdev, PCI_D3hot);
6807 	}
6808 
6809 	return 0;
6810 }
6811 #endif /* CONFIG_PM */
6812 
6813 static void ixgbe_shutdown(struct pci_dev *pdev)
6814 {
6815 	bool wake;
6816 
6817 	__ixgbe_shutdown(pdev, &wake);
6818 
6819 	if (system_state == SYSTEM_POWER_OFF) {
6820 		pci_wake_from_d3(pdev, wake);
6821 		pci_set_power_state(pdev, PCI_D3hot);
6822 	}
6823 }
6824 
6825 /**
6826  * ixgbe_update_stats - Update the board statistics counters.
6827  * @adapter: board private structure
6828  **/
6829 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6830 {
6831 	struct net_device *netdev = adapter->netdev;
6832 	struct ixgbe_hw *hw = &adapter->hw;
6833 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6834 	u64 total_mpc = 0;
6835 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6836 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6837 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6838 	u64 alloc_rx_page = 0;
6839 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6840 
6841 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6842 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6843 		return;
6844 
6845 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6846 		u64 rsc_count = 0;
6847 		u64 rsc_flush = 0;
6848 		for (i = 0; i < adapter->num_rx_queues; i++) {
6849 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6850 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6851 		}
6852 		adapter->rsc_total_count = rsc_count;
6853 		adapter->rsc_total_flush = rsc_flush;
6854 	}
6855 
6856 	for (i = 0; i < adapter->num_rx_queues; i++) {
6857 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6858 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6859 		alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
6860 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6861 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6862 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6863 		bytes += rx_ring->stats.bytes;
6864 		packets += rx_ring->stats.packets;
6865 	}
6866 	adapter->non_eop_descs = non_eop_descs;
6867 	adapter->alloc_rx_page = alloc_rx_page;
6868 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6869 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6870 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6871 	netdev->stats.rx_bytes = bytes;
6872 	netdev->stats.rx_packets = packets;
6873 
6874 	bytes = 0;
6875 	packets = 0;
6876 	/* gather some stats to the adapter struct that are per queue */
6877 	for (i = 0; i < adapter->num_tx_queues; i++) {
6878 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6879 		restart_queue += tx_ring->tx_stats.restart_queue;
6880 		tx_busy += tx_ring->tx_stats.tx_busy;
6881 		bytes += tx_ring->stats.bytes;
6882 		packets += tx_ring->stats.packets;
6883 	}
6884 	for (i = 0; i < adapter->num_xdp_queues; i++) {
6885 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
6886 
6887 		restart_queue += xdp_ring->tx_stats.restart_queue;
6888 		tx_busy += xdp_ring->tx_stats.tx_busy;
6889 		bytes += xdp_ring->stats.bytes;
6890 		packets += xdp_ring->stats.packets;
6891 	}
6892 	adapter->restart_queue = restart_queue;
6893 	adapter->tx_busy = tx_busy;
6894 	netdev->stats.tx_bytes = bytes;
6895 	netdev->stats.tx_packets = packets;
6896 
6897 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6898 
6899 	/* 8 register reads */
6900 	for (i = 0; i < 8; i++) {
6901 		/* for packet buffers not used, the register should read 0 */
6902 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6903 		missed_rx += mpc;
6904 		hwstats->mpc[i] += mpc;
6905 		total_mpc += hwstats->mpc[i];
6906 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6907 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6908 		switch (hw->mac.type) {
6909 		case ixgbe_mac_82598EB:
6910 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6911 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6912 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6913 			hwstats->pxonrxc[i] +=
6914 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6915 			break;
6916 		case ixgbe_mac_82599EB:
6917 		case ixgbe_mac_X540:
6918 		case ixgbe_mac_X550:
6919 		case ixgbe_mac_X550EM_x:
6920 		case ixgbe_mac_x550em_a:
6921 			hwstats->pxonrxc[i] +=
6922 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6923 			break;
6924 		default:
6925 			break;
6926 		}
6927 	}
6928 
6929 	/*16 register reads */
6930 	for (i = 0; i < 16; i++) {
6931 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6932 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6933 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6934 		    (hw->mac.type == ixgbe_mac_X540) ||
6935 		    (hw->mac.type == ixgbe_mac_X550) ||
6936 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6937 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6938 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6939 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6940 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6941 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6942 		}
6943 	}
6944 
6945 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6946 	/* work around hardware counting issue */
6947 	hwstats->gprc -= missed_rx;
6948 
6949 	ixgbe_update_xoff_received(adapter);
6950 
6951 	/* 82598 hardware only has a 32 bit counter in the high register */
6952 	switch (hw->mac.type) {
6953 	case ixgbe_mac_82598EB:
6954 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6955 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6956 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6957 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6958 		break;
6959 	case ixgbe_mac_X540:
6960 	case ixgbe_mac_X550:
6961 	case ixgbe_mac_X550EM_x:
6962 	case ixgbe_mac_x550em_a:
6963 		/* OS2BMC stats are X540 and later */
6964 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6965 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6966 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6967 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6968 		/* fall through */
6969 	case ixgbe_mac_82599EB:
6970 		for (i = 0; i < 16; i++)
6971 			adapter->hw_rx_no_dma_resources +=
6972 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6973 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6974 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6975 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6976 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6977 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6978 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6979 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6980 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6981 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6982 #ifdef IXGBE_FCOE
6983 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6984 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6985 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6986 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6987 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6988 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6989 		/* Add up per cpu counters for total ddp aloc fail */
6990 		if (adapter->fcoe.ddp_pool) {
6991 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6992 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6993 			unsigned int cpu;
6994 			u64 noddp = 0, noddp_ext_buff = 0;
6995 			for_each_possible_cpu(cpu) {
6996 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6997 				noddp += ddp_pool->noddp;
6998 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6999 			}
7000 			hwstats->fcoe_noddp = noddp;
7001 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
7002 		}
7003 #endif /* IXGBE_FCOE */
7004 		break;
7005 	default:
7006 		break;
7007 	}
7008 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
7009 	hwstats->bprc += bprc;
7010 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
7011 	if (hw->mac.type == ixgbe_mac_82598EB)
7012 		hwstats->mprc -= bprc;
7013 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
7014 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
7015 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
7016 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
7017 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
7018 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
7019 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
7020 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
7021 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
7022 	hwstats->lxontxc += lxon;
7023 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
7024 	hwstats->lxofftxc += lxoff;
7025 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
7026 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
7027 	/*
7028 	 * 82598 errata - tx of flow control packets is included in tx counters
7029 	 */
7030 	xon_off_tot = lxon + lxoff;
7031 	hwstats->gptc -= xon_off_tot;
7032 	hwstats->mptc -= xon_off_tot;
7033 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
7034 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
7035 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
7036 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
7037 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
7038 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
7039 	hwstats->ptc64 -= xon_off_tot;
7040 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
7041 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
7042 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
7043 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
7044 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
7045 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
7046 
7047 	/* Fill out the OS statistics structure */
7048 	netdev->stats.multicast = hwstats->mprc;
7049 
7050 	/* Rx Errors */
7051 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
7052 	netdev->stats.rx_dropped = 0;
7053 	netdev->stats.rx_length_errors = hwstats->rlec;
7054 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
7055 	netdev->stats.rx_missed_errors = total_mpc;
7056 }
7057 
7058 /**
7059  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7060  * @adapter: pointer to the device adapter structure
7061  **/
7062 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7063 {
7064 	struct ixgbe_hw *hw = &adapter->hw;
7065 	int i;
7066 
7067 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7068 		return;
7069 
7070 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7071 
7072 	/* if interface is down do nothing */
7073 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7074 		return;
7075 
7076 	/* do nothing if we are not using signature filters */
7077 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7078 		return;
7079 
7080 	adapter->fdir_overflow++;
7081 
7082 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7083 		for (i = 0; i < adapter->num_tx_queues; i++)
7084 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7085 				&(adapter->tx_ring[i]->state));
7086 		for (i = 0; i < adapter->num_xdp_queues; i++)
7087 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7088 				&adapter->xdp_ring[i]->state);
7089 		/* re-enable flow director interrupts */
7090 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7091 	} else {
7092 		e_err(probe, "failed to finish FDIR re-initialization, "
7093 		      "ignored adding FDIR ATR filters\n");
7094 	}
7095 }
7096 
7097 /**
7098  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7099  * @adapter: pointer to the device adapter structure
7100  *
7101  * This function serves two purposes.  First it strobes the interrupt lines
7102  * in order to make certain interrupts are occurring.  Secondly it sets the
7103  * bits needed to check for TX hangs.  As a result we should immediately
7104  * determine if a hang has occurred.
7105  */
7106 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7107 {
7108 	struct ixgbe_hw *hw = &adapter->hw;
7109 	u64 eics = 0;
7110 	int i;
7111 
7112 	/* If we're down, removing or resetting, just bail */
7113 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7114 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7115 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7116 		return;
7117 
7118 	/* Force detection of hung controller */
7119 	if (netif_carrier_ok(adapter->netdev)) {
7120 		for (i = 0; i < adapter->num_tx_queues; i++)
7121 			set_check_for_tx_hang(adapter->tx_ring[i]);
7122 		for (i = 0; i < adapter->num_xdp_queues; i++)
7123 			set_check_for_tx_hang(adapter->xdp_ring[i]);
7124 	}
7125 
7126 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7127 		/*
7128 		 * for legacy and MSI interrupts don't set any bits
7129 		 * that are enabled for EIAM, because this operation
7130 		 * would set *both* EIMS and EICS for any bit in EIAM
7131 		 */
7132 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
7133 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7134 	} else {
7135 		/* get one bit for every active tx/rx interrupt vector */
7136 		for (i = 0; i < adapter->num_q_vectors; i++) {
7137 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
7138 			if (qv->rx.ring || qv->tx.ring)
7139 				eics |= BIT_ULL(i);
7140 		}
7141 	}
7142 
7143 	/* Cause software interrupt to ensure rings are cleaned */
7144 	ixgbe_irq_rearm_queues(adapter, eics);
7145 }
7146 
7147 /**
7148  * ixgbe_watchdog_update_link - update the link status
7149  * @adapter: pointer to the device adapter structure
7150  **/
7151 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7152 {
7153 	struct ixgbe_hw *hw = &adapter->hw;
7154 	u32 link_speed = adapter->link_speed;
7155 	bool link_up = adapter->link_up;
7156 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7157 
7158 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7159 		return;
7160 
7161 	if (hw->mac.ops.check_link) {
7162 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7163 	} else {
7164 		/* always assume link is up, if no check link function */
7165 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7166 		link_up = true;
7167 	}
7168 
7169 	if (adapter->ixgbe_ieee_pfc)
7170 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7171 
7172 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7173 		hw->mac.ops.fc_enable(hw);
7174 		ixgbe_set_rx_drop_en(adapter);
7175 	}
7176 
7177 	if (link_up ||
7178 	    time_after(jiffies, (adapter->link_check_timeout +
7179 				 IXGBE_TRY_LINK_TIMEOUT))) {
7180 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7181 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7182 		IXGBE_WRITE_FLUSH(hw);
7183 	}
7184 
7185 	adapter->link_up = link_up;
7186 	adapter->link_speed = link_speed;
7187 }
7188 
7189 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7190 {
7191 #ifdef CONFIG_IXGBE_DCB
7192 	struct net_device *netdev = adapter->netdev;
7193 	struct dcb_app app = {
7194 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7195 			      .protocol = 0,
7196 			     };
7197 	u8 up = 0;
7198 
7199 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7200 		up = dcb_ieee_getapp_mask(netdev, &app);
7201 
7202 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7203 #endif
7204 }
7205 
7206 /**
7207  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7208  *                             print link up message
7209  * @adapter: pointer to the device adapter structure
7210  **/
7211 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7212 {
7213 	struct net_device *netdev = adapter->netdev;
7214 	struct ixgbe_hw *hw = &adapter->hw;
7215 	u32 link_speed = adapter->link_speed;
7216 	const char *speed_str;
7217 	bool flow_rx, flow_tx;
7218 
7219 	/* only continue if link was previously down */
7220 	if (netif_carrier_ok(netdev))
7221 		return;
7222 
7223 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7224 
7225 	switch (hw->mac.type) {
7226 	case ixgbe_mac_82598EB: {
7227 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7228 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7229 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7230 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7231 	}
7232 		break;
7233 	case ixgbe_mac_X540:
7234 	case ixgbe_mac_X550:
7235 	case ixgbe_mac_X550EM_x:
7236 	case ixgbe_mac_x550em_a:
7237 	case ixgbe_mac_82599EB: {
7238 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7239 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7240 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7241 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7242 	}
7243 		break;
7244 	default:
7245 		flow_tx = false;
7246 		flow_rx = false;
7247 		break;
7248 	}
7249 
7250 	adapter->last_rx_ptp_check = jiffies;
7251 
7252 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7253 		ixgbe_ptp_start_cyclecounter(adapter);
7254 
7255 	switch (link_speed) {
7256 	case IXGBE_LINK_SPEED_10GB_FULL:
7257 		speed_str = "10 Gbps";
7258 		break;
7259 	case IXGBE_LINK_SPEED_5GB_FULL:
7260 		speed_str = "5 Gbps";
7261 		break;
7262 	case IXGBE_LINK_SPEED_2_5GB_FULL:
7263 		speed_str = "2.5 Gbps";
7264 		break;
7265 	case IXGBE_LINK_SPEED_1GB_FULL:
7266 		speed_str = "1 Gbps";
7267 		break;
7268 	case IXGBE_LINK_SPEED_100_FULL:
7269 		speed_str = "100 Mbps";
7270 		break;
7271 	case IXGBE_LINK_SPEED_10_FULL:
7272 		speed_str = "10 Mbps";
7273 		break;
7274 	default:
7275 		speed_str = "unknown speed";
7276 		break;
7277 	}
7278 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7279 	       ((flow_rx && flow_tx) ? "RX/TX" :
7280 	       (flow_rx ? "RX" :
7281 	       (flow_tx ? "TX" : "None"))));
7282 
7283 	netif_carrier_on(netdev);
7284 	ixgbe_check_vf_rate_limit(adapter);
7285 
7286 	/* enable transmits */
7287 	netif_tx_wake_all_queues(adapter->netdev);
7288 
7289 	/* update the default user priority for VFs */
7290 	ixgbe_update_default_up(adapter);
7291 
7292 	/* ping all the active vfs to let them know link has changed */
7293 	ixgbe_ping_all_vfs(adapter);
7294 }
7295 
7296 /**
7297  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7298  *                               print link down message
7299  * @adapter: pointer to the adapter structure
7300  **/
7301 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7302 {
7303 	struct net_device *netdev = adapter->netdev;
7304 	struct ixgbe_hw *hw = &adapter->hw;
7305 
7306 	adapter->link_up = false;
7307 	adapter->link_speed = 0;
7308 
7309 	/* only continue if link was up previously */
7310 	if (!netif_carrier_ok(netdev))
7311 		return;
7312 
7313 	/* poll for SFP+ cable when link is down */
7314 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7315 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7316 
7317 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7318 		ixgbe_ptp_start_cyclecounter(adapter);
7319 
7320 	e_info(drv, "NIC Link is Down\n");
7321 	netif_carrier_off(netdev);
7322 
7323 	/* ping all the active vfs to let them know link has changed */
7324 	ixgbe_ping_all_vfs(adapter);
7325 }
7326 
7327 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7328 {
7329 	int i;
7330 
7331 	for (i = 0; i < adapter->num_tx_queues; i++) {
7332 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7333 
7334 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
7335 			return true;
7336 	}
7337 
7338 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7339 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
7340 
7341 		if (ring->next_to_use != ring->next_to_clean)
7342 			return true;
7343 	}
7344 
7345 	return false;
7346 }
7347 
7348 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7349 {
7350 	struct ixgbe_hw *hw = &adapter->hw;
7351 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7352 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7353 
7354 	int i, j;
7355 
7356 	if (!adapter->num_vfs)
7357 		return false;
7358 
7359 	/* resetting the PF is only needed for MAC before X550 */
7360 	if (hw->mac.type >= ixgbe_mac_X550)
7361 		return false;
7362 
7363 	for (i = 0; i < adapter->num_vfs; i++) {
7364 		for (j = 0; j < q_per_pool; j++) {
7365 			u32 h, t;
7366 
7367 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7368 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7369 
7370 			if (h != t)
7371 				return true;
7372 		}
7373 	}
7374 
7375 	return false;
7376 }
7377 
7378 /**
7379  * ixgbe_watchdog_flush_tx - flush queues on link down
7380  * @adapter: pointer to the device adapter structure
7381  **/
7382 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7383 {
7384 	if (!netif_carrier_ok(adapter->netdev)) {
7385 		if (ixgbe_ring_tx_pending(adapter) ||
7386 		    ixgbe_vf_tx_pending(adapter)) {
7387 			/* We've lost link, so the controller stops DMA,
7388 			 * but we've got queued Tx work that's never going
7389 			 * to get done, so reset controller to flush Tx.
7390 			 * (Do the reset outside of interrupt context).
7391 			 */
7392 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7393 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7394 		}
7395 	}
7396 }
7397 
7398 #ifdef CONFIG_PCI_IOV
7399 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7400 {
7401 	struct ixgbe_hw *hw = &adapter->hw;
7402 	struct pci_dev *pdev = adapter->pdev;
7403 	unsigned int vf;
7404 	u32 gpc;
7405 
7406 	if (!(netif_carrier_ok(adapter->netdev)))
7407 		return;
7408 
7409 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7410 	if (gpc) /* If incrementing then no need for the check below */
7411 		return;
7412 	/* Check to see if a bad DMA write target from an errant or
7413 	 * malicious VF has caused a PCIe error.  If so then we can
7414 	 * issue a VFLR to the offending VF(s) and then resume without
7415 	 * requesting a full slot reset.
7416 	 */
7417 
7418 	if (!pdev)
7419 		return;
7420 
7421 	/* check status reg for all VFs owned by this PF */
7422 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
7423 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7424 		u16 status_reg;
7425 
7426 		if (!vfdev)
7427 			continue;
7428 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7429 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7430 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
7431 			pcie_flr(vfdev);
7432 	}
7433 }
7434 
7435 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7436 {
7437 	u32 ssvpc;
7438 
7439 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7440 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7441 	    adapter->num_vfs == 0)
7442 		return;
7443 
7444 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7445 
7446 	/*
7447 	 * ssvpc register is cleared on read, if zero then no
7448 	 * spoofed packets in the last interval.
7449 	 */
7450 	if (!ssvpc)
7451 		return;
7452 
7453 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7454 }
7455 #else
7456 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7457 {
7458 }
7459 
7460 static void
7461 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7462 {
7463 }
7464 #endif /* CONFIG_PCI_IOV */
7465 
7466 
7467 /**
7468  * ixgbe_watchdog_subtask - check and bring link up
7469  * @adapter: pointer to the device adapter structure
7470  **/
7471 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7472 {
7473 	/* if interface is down, removing or resetting, do nothing */
7474 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7475 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7476 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7477 		return;
7478 
7479 	ixgbe_watchdog_update_link(adapter);
7480 
7481 	if (adapter->link_up)
7482 		ixgbe_watchdog_link_is_up(adapter);
7483 	else
7484 		ixgbe_watchdog_link_is_down(adapter);
7485 
7486 	ixgbe_check_for_bad_vf(adapter);
7487 	ixgbe_spoof_check(adapter);
7488 	ixgbe_update_stats(adapter);
7489 
7490 	ixgbe_watchdog_flush_tx(adapter);
7491 }
7492 
7493 /**
7494  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7495  * @adapter: the ixgbe adapter structure
7496  **/
7497 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7498 {
7499 	struct ixgbe_hw *hw = &adapter->hw;
7500 	s32 err;
7501 
7502 	/* not searching for SFP so there is nothing to do here */
7503 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7504 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7505 		return;
7506 
7507 	if (adapter->sfp_poll_time &&
7508 	    time_after(adapter->sfp_poll_time, jiffies))
7509 		return; /* If not yet time to poll for SFP */
7510 
7511 	/* someone else is in init, wait until next service event */
7512 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7513 		return;
7514 
7515 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7516 
7517 	err = hw->phy.ops.identify_sfp(hw);
7518 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7519 		goto sfp_out;
7520 
7521 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7522 		/* If no cable is present, then we need to reset
7523 		 * the next time we find a good cable. */
7524 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7525 	}
7526 
7527 	/* exit on error */
7528 	if (err)
7529 		goto sfp_out;
7530 
7531 	/* exit if reset not needed */
7532 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7533 		goto sfp_out;
7534 
7535 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7536 
7537 	/*
7538 	 * A module may be identified correctly, but the EEPROM may not have
7539 	 * support for that module.  setup_sfp() will fail in that case, so
7540 	 * we should not allow that module to load.
7541 	 */
7542 	if (hw->mac.type == ixgbe_mac_82598EB)
7543 		err = hw->phy.ops.reset(hw);
7544 	else
7545 		err = hw->mac.ops.setup_sfp(hw);
7546 
7547 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7548 		goto sfp_out;
7549 
7550 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7551 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7552 
7553 sfp_out:
7554 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7555 
7556 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7557 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7558 		e_dev_err("failed to initialize because an unsupported "
7559 			  "SFP+ module type was detected.\n");
7560 		e_dev_err("Reload the driver after installing a "
7561 			  "supported module.\n");
7562 		unregister_netdev(adapter->netdev);
7563 	}
7564 }
7565 
7566 /**
7567  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7568  * @adapter: the ixgbe adapter structure
7569  **/
7570 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7571 {
7572 	struct ixgbe_hw *hw = &adapter->hw;
7573 	u32 cap_speed;
7574 	u32 speed;
7575 	bool autoneg = false;
7576 
7577 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7578 		return;
7579 
7580 	/* someone else is in init, wait until next service event */
7581 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7582 		return;
7583 
7584 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7585 
7586 	hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
7587 
7588 	/* advertise highest capable link speed */
7589 	if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
7590 		speed = IXGBE_LINK_SPEED_10GB_FULL;
7591 	else
7592 		speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
7593 				     IXGBE_LINK_SPEED_1GB_FULL);
7594 
7595 	if (hw->mac.ops.setup_link)
7596 		hw->mac.ops.setup_link(hw, speed, true);
7597 
7598 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7599 	adapter->link_check_timeout = jiffies;
7600 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7601 }
7602 
7603 /**
7604  * ixgbe_service_timer - Timer Call-back
7605  * @t: pointer to timer_list structure
7606  **/
7607 static void ixgbe_service_timer(struct timer_list *t)
7608 {
7609 	struct ixgbe_adapter *adapter = from_timer(adapter, t, service_timer);
7610 	unsigned long next_event_offset;
7611 
7612 	/* poll faster when waiting for link */
7613 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7614 		next_event_offset = HZ / 10;
7615 	else
7616 		next_event_offset = HZ * 2;
7617 
7618 	/* Reset the timer */
7619 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7620 
7621 	ixgbe_service_event_schedule(adapter);
7622 }
7623 
7624 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7625 {
7626 	struct ixgbe_hw *hw = &adapter->hw;
7627 	u32 status;
7628 
7629 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7630 		return;
7631 
7632 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7633 
7634 	if (!hw->phy.ops.handle_lasi)
7635 		return;
7636 
7637 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7638 	if (status != IXGBE_ERR_OVERTEMP)
7639 		return;
7640 
7641 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7642 }
7643 
7644 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7645 {
7646 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7647 		return;
7648 
7649 	/* If we're already down, removing or resetting, just bail */
7650 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7651 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7652 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7653 		return;
7654 
7655 	ixgbe_dump(adapter);
7656 	netdev_err(adapter->netdev, "Reset adapter\n");
7657 	adapter->tx_timeout_count++;
7658 
7659 	rtnl_lock();
7660 	ixgbe_reinit_locked(adapter);
7661 	rtnl_unlock();
7662 }
7663 
7664 /**
7665  * ixgbe_service_task - manages and runs subtasks
7666  * @work: pointer to work_struct containing our data
7667  **/
7668 static void ixgbe_service_task(struct work_struct *work)
7669 {
7670 	struct ixgbe_adapter *adapter = container_of(work,
7671 						     struct ixgbe_adapter,
7672 						     service_task);
7673 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7674 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7675 			rtnl_lock();
7676 			ixgbe_down(adapter);
7677 			rtnl_unlock();
7678 		}
7679 		ixgbe_service_event_complete(adapter);
7680 		return;
7681 	}
7682 	if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7683 		rtnl_lock();
7684 		adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7685 		udp_tunnel_get_rx_info(adapter->netdev);
7686 		rtnl_unlock();
7687 	}
7688 	ixgbe_reset_subtask(adapter);
7689 	ixgbe_phy_interrupt_subtask(adapter);
7690 	ixgbe_sfp_detection_subtask(adapter);
7691 	ixgbe_sfp_link_config_subtask(adapter);
7692 	ixgbe_check_overtemp_subtask(adapter);
7693 	ixgbe_watchdog_subtask(adapter);
7694 	ixgbe_fdir_reinit_subtask(adapter);
7695 	ixgbe_check_hang_subtask(adapter);
7696 
7697 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7698 		ixgbe_ptp_overflow_check(adapter);
7699 		if (adapter->flags & IXGBE_FLAG_RX_HWTSTAMP_IN_REGISTER)
7700 			ixgbe_ptp_rx_hang(adapter);
7701 		ixgbe_ptp_tx_hang(adapter);
7702 	}
7703 
7704 	ixgbe_service_event_complete(adapter);
7705 }
7706 
7707 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7708 		     struct ixgbe_tx_buffer *first,
7709 		     u8 *hdr_len,
7710 		     struct ixgbe_ipsec_tx_data *itd)
7711 {
7712 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7713 	struct sk_buff *skb = first->skb;
7714 	union {
7715 		struct iphdr *v4;
7716 		struct ipv6hdr *v6;
7717 		unsigned char *hdr;
7718 	} ip;
7719 	union {
7720 		struct tcphdr *tcp;
7721 		unsigned char *hdr;
7722 	} l4;
7723 	u32 paylen, l4_offset;
7724 	u32 fceof_saidx = 0;
7725 	int err;
7726 
7727 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7728 		return 0;
7729 
7730 	if (!skb_is_gso(skb))
7731 		return 0;
7732 
7733 	err = skb_cow_head(skb, 0);
7734 	if (err < 0)
7735 		return err;
7736 
7737 	if (eth_p_mpls(first->protocol))
7738 		ip.hdr = skb_inner_network_header(skb);
7739 	else
7740 		ip.hdr = skb_network_header(skb);
7741 	l4.hdr = skb_checksum_start(skb);
7742 
7743 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7744 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7745 
7746 	/* initialize outer IP header fields */
7747 	if (ip.v4->version == 4) {
7748 		unsigned char *csum_start = skb_checksum_start(skb);
7749 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
7750 		int len = csum_start - trans_start;
7751 
7752 		/* IP header will have to cancel out any data that
7753 		 * is not a part of the outer IP header, so set to
7754 		 * a reverse csum if needed, else init check to 0.
7755 		 */
7756 		ip.v4->check = (skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) ?
7757 					   csum_fold(csum_partial(trans_start,
7758 								  len, 0)) : 0;
7759 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7760 
7761 		ip.v4->tot_len = 0;
7762 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7763 				   IXGBE_TX_FLAGS_CSUM |
7764 				   IXGBE_TX_FLAGS_IPV4;
7765 	} else {
7766 		ip.v6->payload_len = 0;
7767 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7768 				   IXGBE_TX_FLAGS_CSUM;
7769 	}
7770 
7771 	/* determine offset of inner transport header */
7772 	l4_offset = l4.hdr - skb->data;
7773 
7774 	/* compute length of segmentation header */
7775 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7776 
7777 	/* remove payload length from inner checksum */
7778 	paylen = skb->len - l4_offset;
7779 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
7780 
7781 	/* update gso size and bytecount with header size */
7782 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7783 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7784 
7785 	/* mss_l4len_id: use 0 as index for TSO */
7786 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7787 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7788 
7789 	fceof_saidx |= itd->sa_idx;
7790 	type_tucmd |= itd->flags | itd->trailer_len;
7791 
7792 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7793 	vlan_macip_lens = l4.hdr - ip.hdr;
7794 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7795 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7796 
7797 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd,
7798 			  mss_l4len_idx);
7799 
7800 	return 1;
7801 }
7802 
7803 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7804 {
7805 	unsigned int offset = 0;
7806 
7807 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7808 
7809 	return offset == skb_checksum_start_offset(skb);
7810 }
7811 
7812 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7813 			  struct ixgbe_tx_buffer *first,
7814 			  struct ixgbe_ipsec_tx_data *itd)
7815 {
7816 	struct sk_buff *skb = first->skb;
7817 	u32 vlan_macip_lens = 0;
7818 	u32 fceof_saidx = 0;
7819 	u32 type_tucmd = 0;
7820 
7821 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7822 csum_failed:
7823 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7824 					 IXGBE_TX_FLAGS_CC)))
7825 			return;
7826 		goto no_csum;
7827 	}
7828 
7829 	switch (skb->csum_offset) {
7830 	case offsetof(struct tcphdr, check):
7831 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7832 		/* fall through */
7833 	case offsetof(struct udphdr, check):
7834 		break;
7835 	case offsetof(struct sctphdr, checksum):
7836 		/* validate that this is actually an SCTP request */
7837 		if (((first->protocol == htons(ETH_P_IP)) &&
7838 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7839 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7840 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7841 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7842 			break;
7843 		}
7844 		/* fall through */
7845 	default:
7846 		skb_checksum_help(skb);
7847 		goto csum_failed;
7848 	}
7849 
7850 	/* update TX checksum flag */
7851 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7852 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7853 			  skb_network_offset(skb);
7854 no_csum:
7855 	/* vlan_macip_lens: MACLEN, VLAN tag */
7856 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7857 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7858 
7859 	fceof_saidx |= itd->sa_idx;
7860 	type_tucmd |= itd->flags | itd->trailer_len;
7861 
7862 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
7863 }
7864 
7865 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7866 	((_flag <= _result) ? \
7867 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7868 	 ((u32)(_input & _flag) / (_flag / _result)))
7869 
7870 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7871 {
7872 	/* set type for advanced descriptor with frame checksum insertion */
7873 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7874 		       IXGBE_ADVTXD_DCMD_DEXT |
7875 		       IXGBE_ADVTXD_DCMD_IFCS;
7876 
7877 	/* set HW vlan bit if vlan is present */
7878 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7879 				   IXGBE_ADVTXD_DCMD_VLE);
7880 
7881 	/* set segmentation enable bits for TSO/FSO */
7882 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7883 				   IXGBE_ADVTXD_DCMD_TSE);
7884 
7885 	/* set timestamp bit if present */
7886 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7887 				   IXGBE_ADVTXD_MAC_TSTAMP);
7888 
7889 	/* insert frame checksum */
7890 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7891 
7892 	return cmd_type;
7893 }
7894 
7895 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7896 				   u32 tx_flags, unsigned int paylen)
7897 {
7898 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7899 
7900 	/* enable L4 checksum for TSO and TX checksum offload */
7901 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7902 					IXGBE_TX_FLAGS_CSUM,
7903 					IXGBE_ADVTXD_POPTS_TXSM);
7904 
7905 	/* enable IPv4 checksum for TSO */
7906 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7907 					IXGBE_TX_FLAGS_IPV4,
7908 					IXGBE_ADVTXD_POPTS_IXSM);
7909 
7910 	/* enable IPsec */
7911 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7912 					IXGBE_TX_FLAGS_IPSEC,
7913 					IXGBE_ADVTXD_POPTS_IPSEC);
7914 
7915 	/*
7916 	 * Check Context must be set if Tx switch is enabled, which it
7917 	 * always is for case where virtual functions are running
7918 	 */
7919 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7920 					IXGBE_TX_FLAGS_CC,
7921 					IXGBE_ADVTXD_CC);
7922 
7923 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7924 }
7925 
7926 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7927 {
7928 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7929 
7930 	/* Herbert's original patch had:
7931 	 *  smp_mb__after_netif_stop_queue();
7932 	 * but since that doesn't exist yet, just open code it.
7933 	 */
7934 	smp_mb();
7935 
7936 	/* We need to check again in a case another CPU has just
7937 	 * made room available.
7938 	 */
7939 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7940 		return -EBUSY;
7941 
7942 	/* A reprieve! - use start_queue because it doesn't call schedule */
7943 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7944 	++tx_ring->tx_stats.restart_queue;
7945 	return 0;
7946 }
7947 
7948 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7949 {
7950 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7951 		return 0;
7952 
7953 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7954 }
7955 
7956 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7957 		       IXGBE_TXD_CMD_RS)
7958 
7959 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7960 			struct ixgbe_tx_buffer *first,
7961 			const u8 hdr_len)
7962 {
7963 	struct sk_buff *skb = first->skb;
7964 	struct ixgbe_tx_buffer *tx_buffer;
7965 	union ixgbe_adv_tx_desc *tx_desc;
7966 	struct skb_frag_struct *frag;
7967 	dma_addr_t dma;
7968 	unsigned int data_len, size;
7969 	u32 tx_flags = first->tx_flags;
7970 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7971 	u16 i = tx_ring->next_to_use;
7972 
7973 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7974 
7975 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7976 
7977 	size = skb_headlen(skb);
7978 	data_len = skb->data_len;
7979 
7980 #ifdef IXGBE_FCOE
7981 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7982 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7983 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7984 			data_len = 0;
7985 		} else {
7986 			data_len -= sizeof(struct fcoe_crc_eof);
7987 		}
7988 	}
7989 
7990 #endif
7991 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7992 
7993 	tx_buffer = first;
7994 
7995 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7996 		if (dma_mapping_error(tx_ring->dev, dma))
7997 			goto dma_error;
7998 
7999 		/* record length, and DMA address */
8000 		dma_unmap_len_set(tx_buffer, len, size);
8001 		dma_unmap_addr_set(tx_buffer, dma, dma);
8002 
8003 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
8004 
8005 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
8006 			tx_desc->read.cmd_type_len =
8007 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
8008 
8009 			i++;
8010 			tx_desc++;
8011 			if (i == tx_ring->count) {
8012 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8013 				i = 0;
8014 			}
8015 			tx_desc->read.olinfo_status = 0;
8016 
8017 			dma += IXGBE_MAX_DATA_PER_TXD;
8018 			size -= IXGBE_MAX_DATA_PER_TXD;
8019 
8020 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
8021 		}
8022 
8023 		if (likely(!data_len))
8024 			break;
8025 
8026 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
8027 
8028 		i++;
8029 		tx_desc++;
8030 		if (i == tx_ring->count) {
8031 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8032 			i = 0;
8033 		}
8034 		tx_desc->read.olinfo_status = 0;
8035 
8036 #ifdef IXGBE_FCOE
8037 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
8038 #else
8039 		size = skb_frag_size(frag);
8040 #endif
8041 		data_len -= size;
8042 
8043 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
8044 				       DMA_TO_DEVICE);
8045 
8046 		tx_buffer = &tx_ring->tx_buffer_info[i];
8047 	}
8048 
8049 	/* write last descriptor with RS and EOP bits */
8050 	cmd_type |= size | IXGBE_TXD_CMD;
8051 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8052 
8053 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
8054 
8055 	/* set the timestamp */
8056 	first->time_stamp = jiffies;
8057 
8058 	/*
8059 	 * Force memory writes to complete before letting h/w know there
8060 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
8061 	 * memory model archs, such as IA-64).
8062 	 *
8063 	 * We also need this memory barrier to make certain all of the
8064 	 * status bits have been updated before next_to_watch is written.
8065 	 */
8066 	wmb();
8067 
8068 	/* set next_to_watch value indicating a packet is present */
8069 	first->next_to_watch = tx_desc;
8070 
8071 	i++;
8072 	if (i == tx_ring->count)
8073 		i = 0;
8074 
8075 	tx_ring->next_to_use = i;
8076 
8077 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8078 
8079 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
8080 		writel(i, tx_ring->tail);
8081 
8082 		/* we need this if more than one processor can write to our tail
8083 		 * at a time, it synchronizes IO on IA64/Altix systems
8084 		 */
8085 		mmiowb();
8086 	}
8087 
8088 	return 0;
8089 dma_error:
8090 	dev_err(tx_ring->dev, "TX DMA map failed\n");
8091 
8092 	/* clear dma mappings for failed tx_buffer_info map */
8093 	for (;;) {
8094 		tx_buffer = &tx_ring->tx_buffer_info[i];
8095 		if (dma_unmap_len(tx_buffer, len))
8096 			dma_unmap_page(tx_ring->dev,
8097 				       dma_unmap_addr(tx_buffer, dma),
8098 				       dma_unmap_len(tx_buffer, len),
8099 				       DMA_TO_DEVICE);
8100 		dma_unmap_len_set(tx_buffer, len, 0);
8101 		if (tx_buffer == first)
8102 			break;
8103 		if (i == 0)
8104 			i += tx_ring->count;
8105 		i--;
8106 	}
8107 
8108 	dev_kfree_skb_any(first->skb);
8109 	first->skb = NULL;
8110 
8111 	tx_ring->next_to_use = i;
8112 
8113 	return -1;
8114 }
8115 
8116 static void ixgbe_atr(struct ixgbe_ring *ring,
8117 		      struct ixgbe_tx_buffer *first)
8118 {
8119 	struct ixgbe_q_vector *q_vector = ring->q_vector;
8120 	union ixgbe_atr_hash_dword input = { .dword = 0 };
8121 	union ixgbe_atr_hash_dword common = { .dword = 0 };
8122 	union {
8123 		unsigned char *network;
8124 		struct iphdr *ipv4;
8125 		struct ipv6hdr *ipv6;
8126 	} hdr;
8127 	struct tcphdr *th;
8128 	unsigned int hlen;
8129 	struct sk_buff *skb;
8130 	__be16 vlan_id;
8131 	int l4_proto;
8132 
8133 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
8134 	if (!q_vector)
8135 		return;
8136 
8137 	/* do nothing if sampling is disabled */
8138 	if (!ring->atr_sample_rate)
8139 		return;
8140 
8141 	ring->atr_count++;
8142 
8143 	/* currently only IPv4/IPv6 with TCP is supported */
8144 	if ((first->protocol != htons(ETH_P_IP)) &&
8145 	    (first->protocol != htons(ETH_P_IPV6)))
8146 		return;
8147 
8148 	/* snag network header to get L4 type and address */
8149 	skb = first->skb;
8150 	hdr.network = skb_network_header(skb);
8151 	if (unlikely(hdr.network <= skb->data))
8152 		return;
8153 	if (skb->encapsulation &&
8154 	    first->protocol == htons(ETH_P_IP) &&
8155 	    hdr.ipv4->protocol == IPPROTO_UDP) {
8156 		struct ixgbe_adapter *adapter = q_vector->adapter;
8157 
8158 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
8159 			     VXLAN_HEADROOM))
8160 			return;
8161 
8162 		/* verify the port is recognized as VXLAN */
8163 		if (adapter->vxlan_port &&
8164 		    udp_hdr(skb)->dest == adapter->vxlan_port)
8165 			hdr.network = skb_inner_network_header(skb);
8166 
8167 		if (adapter->geneve_port &&
8168 		    udp_hdr(skb)->dest == adapter->geneve_port)
8169 			hdr.network = skb_inner_network_header(skb);
8170 	}
8171 
8172 	/* Make sure we have at least [minimum IPv4 header + TCP]
8173 	 * or [IPv6 header] bytes
8174 	 */
8175 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8176 		return;
8177 
8178 	/* Currently only IPv4/IPv6 with TCP is supported */
8179 	switch (hdr.ipv4->version) {
8180 	case IPVERSION:
8181 		/* access ihl as u8 to avoid unaligned access on ia64 */
8182 		hlen = (hdr.network[0] & 0x0F) << 2;
8183 		l4_proto = hdr.ipv4->protocol;
8184 		break;
8185 	case 6:
8186 		hlen = hdr.network - skb->data;
8187 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8188 		hlen -= hdr.network - skb->data;
8189 		break;
8190 	default:
8191 		return;
8192 	}
8193 
8194 	if (l4_proto != IPPROTO_TCP)
8195 		return;
8196 
8197 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
8198 		     hlen + sizeof(struct tcphdr)))
8199 		return;
8200 
8201 	th = (struct tcphdr *)(hdr.network + hlen);
8202 
8203 	/* skip this packet since the socket is closing */
8204 	if (th->fin)
8205 		return;
8206 
8207 	/* sample on all syn packets or once every atr sample count */
8208 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8209 		return;
8210 
8211 	/* reset sample count */
8212 	ring->atr_count = 0;
8213 
8214 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8215 
8216 	/*
8217 	 * src and dst are inverted, think how the receiver sees them
8218 	 *
8219 	 * The input is broken into two sections, a non-compressed section
8220 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8221 	 * is XORed together and stored in the compressed dword.
8222 	 */
8223 	input.formatted.vlan_id = vlan_id;
8224 
8225 	/*
8226 	 * since src port and flex bytes occupy the same word XOR them together
8227 	 * and write the value to source port portion of compressed dword
8228 	 */
8229 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8230 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8231 	else
8232 		common.port.src ^= th->dest ^ first->protocol;
8233 	common.port.dst ^= th->source;
8234 
8235 	switch (hdr.ipv4->version) {
8236 	case IPVERSION:
8237 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8238 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8239 		break;
8240 	case 6:
8241 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8242 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8243 			     hdr.ipv6->saddr.s6_addr32[1] ^
8244 			     hdr.ipv6->saddr.s6_addr32[2] ^
8245 			     hdr.ipv6->saddr.s6_addr32[3] ^
8246 			     hdr.ipv6->daddr.s6_addr32[0] ^
8247 			     hdr.ipv6->daddr.s6_addr32[1] ^
8248 			     hdr.ipv6->daddr.s6_addr32[2] ^
8249 			     hdr.ipv6->daddr.s6_addr32[3];
8250 		break;
8251 	default:
8252 		break;
8253 	}
8254 
8255 	if (hdr.network != skb_network_header(skb))
8256 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8257 
8258 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
8259 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8260 					      input, common, ring->queue_index);
8261 }
8262 
8263 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8264 			      void *accel_priv, select_queue_fallback_t fallback)
8265 {
8266 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
8267 	struct ixgbe_adapter *adapter;
8268 	int txq;
8269 #ifdef IXGBE_FCOE
8270 	struct ixgbe_ring_feature *f;
8271 #endif
8272 
8273 	if (fwd_adapter) {
8274 		adapter = netdev_priv(dev);
8275 		txq = reciprocal_scale(skb_get_hash(skb),
8276 				       adapter->num_rx_queues_per_pool);
8277 
8278 		return txq + fwd_adapter->tx_base_queue;
8279 	}
8280 
8281 #ifdef IXGBE_FCOE
8282 
8283 	/*
8284 	 * only execute the code below if protocol is FCoE
8285 	 * or FIP and we have FCoE enabled on the adapter
8286 	 */
8287 	switch (vlan_get_protocol(skb)) {
8288 	case htons(ETH_P_FCOE):
8289 	case htons(ETH_P_FIP):
8290 		adapter = netdev_priv(dev);
8291 
8292 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
8293 			break;
8294 		/* fall through */
8295 	default:
8296 		return fallback(dev, skb);
8297 	}
8298 
8299 	f = &adapter->ring_feature[RING_F_FCOE];
8300 
8301 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8302 					   smp_processor_id();
8303 
8304 	while (txq >= f->indices)
8305 		txq -= f->indices;
8306 
8307 	return txq + f->offset;
8308 #else
8309 	return fallback(dev, skb);
8310 #endif
8311 }
8312 
8313 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
8314 			       struct xdp_frame *xdpf)
8315 {
8316 	struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
8317 	struct ixgbe_tx_buffer *tx_buffer;
8318 	union ixgbe_adv_tx_desc *tx_desc;
8319 	u32 len, cmd_type;
8320 	dma_addr_t dma;
8321 	u16 i;
8322 
8323 	len = xdpf->len;
8324 
8325 	if (unlikely(!ixgbe_desc_unused(ring)))
8326 		return IXGBE_XDP_CONSUMED;
8327 
8328 	dma = dma_map_single(ring->dev, xdpf->data, len, DMA_TO_DEVICE);
8329 	if (dma_mapping_error(ring->dev, dma))
8330 		return IXGBE_XDP_CONSUMED;
8331 
8332 	/* record the location of the first descriptor for this packet */
8333 	tx_buffer = &ring->tx_buffer_info[ring->next_to_use];
8334 	tx_buffer->bytecount = len;
8335 	tx_buffer->gso_segs = 1;
8336 	tx_buffer->protocol = 0;
8337 
8338 	i = ring->next_to_use;
8339 	tx_desc = IXGBE_TX_DESC(ring, i);
8340 
8341 	dma_unmap_len_set(tx_buffer, len, len);
8342 	dma_unmap_addr_set(tx_buffer, dma, dma);
8343 	tx_buffer->xdpf = xdpf;
8344 
8345 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
8346 
8347 	/* put descriptor type bits */
8348 	cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8349 		   IXGBE_ADVTXD_DCMD_DEXT |
8350 		   IXGBE_ADVTXD_DCMD_IFCS;
8351 	cmd_type |= len | IXGBE_TXD_CMD;
8352 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8353 	tx_desc->read.olinfo_status =
8354 		cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
8355 
8356 	/* Avoid any potential race with xdp_xmit and cleanup */
8357 	smp_wmb();
8358 
8359 	/* set next_to_watch value indicating a packet is present */
8360 	i++;
8361 	if (i == ring->count)
8362 		i = 0;
8363 
8364 	tx_buffer->next_to_watch = tx_desc;
8365 	ring->next_to_use = i;
8366 
8367 	return IXGBE_XDP_TX;
8368 }
8369 
8370 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8371 			  struct ixgbe_adapter *adapter,
8372 			  struct ixgbe_ring *tx_ring)
8373 {
8374 	struct ixgbe_tx_buffer *first;
8375 	int tso;
8376 	u32 tx_flags = 0;
8377 	unsigned short f;
8378 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
8379 	struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
8380 	__be16 protocol = skb->protocol;
8381 	u8 hdr_len = 0;
8382 
8383 	/*
8384 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8385 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8386 	 *       + 2 desc gap to keep tail from touching head,
8387 	 *       + 1 desc for context descriptor,
8388 	 * otherwise try next time
8389 	 */
8390 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8391 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
8392 
8393 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8394 		tx_ring->tx_stats.tx_busy++;
8395 		return NETDEV_TX_BUSY;
8396 	}
8397 
8398 	/* record the location of the first descriptor for this packet */
8399 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8400 	first->skb = skb;
8401 	first->bytecount = skb->len;
8402 	first->gso_segs = 1;
8403 
8404 	/* if we have a HW VLAN tag being added default to the HW one */
8405 	if (skb_vlan_tag_present(skb)) {
8406 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8407 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8408 	/* else if it is a SW VLAN check the next protocol and store the tag */
8409 	} else if (protocol == htons(ETH_P_8021Q)) {
8410 		struct vlan_hdr *vhdr, _vhdr;
8411 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8412 		if (!vhdr)
8413 			goto out_drop;
8414 
8415 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8416 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
8417 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8418 	}
8419 	protocol = vlan_get_protocol(skb);
8420 
8421 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8422 	    adapter->ptp_clock) {
8423 		if (!test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8424 					   &adapter->state)) {
8425 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8426 			tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8427 
8428 			/* schedule check for Tx timestamp */
8429 			adapter->ptp_tx_skb = skb_get(skb);
8430 			adapter->ptp_tx_start = jiffies;
8431 			schedule_work(&adapter->ptp_tx_work);
8432 		} else {
8433 			adapter->tx_hwtstamp_skipped++;
8434 		}
8435 	}
8436 
8437 	skb_tx_timestamp(skb);
8438 
8439 #ifdef CONFIG_PCI_IOV
8440 	/*
8441 	 * Use the l2switch_enable flag - would be false if the DMA
8442 	 * Tx switch had been disabled.
8443 	 */
8444 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8445 		tx_flags |= IXGBE_TX_FLAGS_CC;
8446 
8447 #endif
8448 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8449 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8450 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8451 	     (skb->priority != TC_PRIO_CONTROL))) {
8452 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8453 		tx_flags |= (skb->priority & 0x7) <<
8454 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8455 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8456 			struct vlan_ethhdr *vhdr;
8457 
8458 			if (skb_cow_head(skb, 0))
8459 				goto out_drop;
8460 			vhdr = (struct vlan_ethhdr *)skb->data;
8461 			vhdr->h_vlan_TCI = htons(tx_flags >>
8462 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
8463 		} else {
8464 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8465 		}
8466 	}
8467 
8468 	/* record initial flags and protocol */
8469 	first->tx_flags = tx_flags;
8470 	first->protocol = protocol;
8471 
8472 #ifdef IXGBE_FCOE
8473 	/* setup tx offload for FCoE */
8474 	if ((protocol == htons(ETH_P_FCOE)) &&
8475 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8476 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
8477 		if (tso < 0)
8478 			goto out_drop;
8479 
8480 		goto xmit_fcoe;
8481 	}
8482 
8483 #endif /* IXGBE_FCOE */
8484 
8485 #ifdef CONFIG_XFRM_OFFLOAD
8486 	if (skb->sp && !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
8487 		goto out_drop;
8488 #endif
8489 	tso = ixgbe_tso(tx_ring, first, &hdr_len, &ipsec_tx);
8490 	if (tso < 0)
8491 		goto out_drop;
8492 	else if (!tso)
8493 		ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
8494 
8495 	/* add the ATR filter if ATR is on */
8496 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8497 		ixgbe_atr(tx_ring, first);
8498 
8499 #ifdef IXGBE_FCOE
8500 xmit_fcoe:
8501 #endif /* IXGBE_FCOE */
8502 	if (ixgbe_tx_map(tx_ring, first, hdr_len))
8503 		goto cleanup_tx_timestamp;
8504 
8505 	return NETDEV_TX_OK;
8506 
8507 out_drop:
8508 	dev_kfree_skb_any(first->skb);
8509 	first->skb = NULL;
8510 cleanup_tx_timestamp:
8511 	if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8512 		dev_kfree_skb_any(adapter->ptp_tx_skb);
8513 		adapter->ptp_tx_skb = NULL;
8514 		cancel_work_sync(&adapter->ptp_tx_work);
8515 		clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8516 	}
8517 
8518 	return NETDEV_TX_OK;
8519 }
8520 
8521 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8522 				      struct net_device *netdev,
8523 				      struct ixgbe_ring *ring)
8524 {
8525 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8526 	struct ixgbe_ring *tx_ring;
8527 
8528 	/*
8529 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
8530 	 * in order to meet this minimum size requirement.
8531 	 */
8532 	if (skb_put_padto(skb, 17))
8533 		return NETDEV_TX_OK;
8534 
8535 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
8536 
8537 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8538 }
8539 
8540 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8541 				    struct net_device *netdev)
8542 {
8543 	return __ixgbe_xmit_frame(skb, netdev, NULL);
8544 }
8545 
8546 /**
8547  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8548  * @netdev: network interface device structure
8549  * @p: pointer to an address structure
8550  *
8551  * Returns 0 on success, negative on failure
8552  **/
8553 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8554 {
8555 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8556 	struct ixgbe_hw *hw = &adapter->hw;
8557 	struct sockaddr *addr = p;
8558 
8559 	if (!is_valid_ether_addr(addr->sa_data))
8560 		return -EADDRNOTAVAIL;
8561 
8562 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
8563 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8564 
8565 	ixgbe_mac_set_default_filter(adapter);
8566 
8567 	return 0;
8568 }
8569 
8570 static int
8571 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8572 {
8573 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8574 	struct ixgbe_hw *hw = &adapter->hw;
8575 	u16 value;
8576 	int rc;
8577 
8578 	if (prtad != hw->phy.mdio.prtad)
8579 		return -EINVAL;
8580 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8581 	if (!rc)
8582 		rc = value;
8583 	return rc;
8584 }
8585 
8586 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8587 			    u16 addr, u16 value)
8588 {
8589 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8590 	struct ixgbe_hw *hw = &adapter->hw;
8591 
8592 	if (prtad != hw->phy.mdio.prtad)
8593 		return -EINVAL;
8594 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8595 }
8596 
8597 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8598 {
8599 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8600 
8601 	switch (cmd) {
8602 	case SIOCSHWTSTAMP:
8603 		return ixgbe_ptp_set_ts_config(adapter, req);
8604 	case SIOCGHWTSTAMP:
8605 		return ixgbe_ptp_get_ts_config(adapter, req);
8606 	case SIOCGMIIPHY:
8607 		if (!adapter->hw.phy.ops.read_reg)
8608 			return -EOPNOTSUPP;
8609 		/* fall through */
8610 	default:
8611 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8612 	}
8613 }
8614 
8615 /**
8616  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8617  * netdev->dev_addrs
8618  * @dev: network interface device structure
8619  *
8620  * Returns non-zero on failure
8621  **/
8622 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8623 {
8624 	int err = 0;
8625 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8626 	struct ixgbe_hw *hw = &adapter->hw;
8627 
8628 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8629 		rtnl_lock();
8630 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8631 		rtnl_unlock();
8632 
8633 		/* update SAN MAC vmdq pool selection */
8634 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8635 	}
8636 	return err;
8637 }
8638 
8639 /**
8640  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8641  * netdev->dev_addrs
8642  * @dev: network interface device structure
8643  *
8644  * Returns non-zero on failure
8645  **/
8646 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8647 {
8648 	int err = 0;
8649 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8650 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8651 
8652 	if (is_valid_ether_addr(mac->san_addr)) {
8653 		rtnl_lock();
8654 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8655 		rtnl_unlock();
8656 	}
8657 	return err;
8658 }
8659 
8660 #ifdef CONFIG_NET_POLL_CONTROLLER
8661 /*
8662  * Polling 'interrupt' - used by things like netconsole to send skbs
8663  * without having to re-enable interrupts. It's not called while
8664  * the interrupt routine is executing.
8665  */
8666 static void ixgbe_netpoll(struct net_device *netdev)
8667 {
8668 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8669 	int i;
8670 
8671 	/* if interface is down do nothing */
8672 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8673 		return;
8674 
8675 	/* loop through and schedule all active queues */
8676 	for (i = 0; i < adapter->num_q_vectors; i++)
8677 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8678 }
8679 
8680 #endif
8681 
8682 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
8683 				   struct ixgbe_ring *ring)
8684 {
8685 	u64 bytes, packets;
8686 	unsigned int start;
8687 
8688 	if (ring) {
8689 		do {
8690 			start = u64_stats_fetch_begin_irq(&ring->syncp);
8691 			packets = ring->stats.packets;
8692 			bytes   = ring->stats.bytes;
8693 		} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8694 		stats->tx_packets += packets;
8695 		stats->tx_bytes   += bytes;
8696 	}
8697 }
8698 
8699 static void ixgbe_get_stats64(struct net_device *netdev,
8700 			      struct rtnl_link_stats64 *stats)
8701 {
8702 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8703 	int i;
8704 
8705 	rcu_read_lock();
8706 	for (i = 0; i < adapter->num_rx_queues; i++) {
8707 		struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
8708 		u64 bytes, packets;
8709 		unsigned int start;
8710 
8711 		if (ring) {
8712 			do {
8713 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8714 				packets = ring->stats.packets;
8715 				bytes   = ring->stats.bytes;
8716 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8717 			stats->rx_packets += packets;
8718 			stats->rx_bytes   += bytes;
8719 		}
8720 	}
8721 
8722 	for (i = 0; i < adapter->num_tx_queues; i++) {
8723 		struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
8724 
8725 		ixgbe_get_ring_stats64(stats, ring);
8726 	}
8727 	for (i = 0; i < adapter->num_xdp_queues; i++) {
8728 		struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
8729 
8730 		ixgbe_get_ring_stats64(stats, ring);
8731 	}
8732 	rcu_read_unlock();
8733 
8734 	/* following stats updated by ixgbe_watchdog_task() */
8735 	stats->multicast	= netdev->stats.multicast;
8736 	stats->rx_errors	= netdev->stats.rx_errors;
8737 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8738 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8739 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8740 }
8741 
8742 #ifdef CONFIG_IXGBE_DCB
8743 /**
8744  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8745  * @adapter: pointer to ixgbe_adapter
8746  * @tc: number of traffic classes currently enabled
8747  *
8748  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8749  * 802.1Q priority maps to a packet buffer that exists.
8750  */
8751 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8752 {
8753 	struct ixgbe_hw *hw = &adapter->hw;
8754 	u32 reg, rsave;
8755 	int i;
8756 
8757 	/* 82598 have a static priority to TC mapping that can not
8758 	 * be changed so no validation is needed.
8759 	 */
8760 	if (hw->mac.type == ixgbe_mac_82598EB)
8761 		return;
8762 
8763 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8764 	rsave = reg;
8765 
8766 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8767 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8768 
8769 		/* If up2tc is out of bounds default to zero */
8770 		if (up2tc > tc)
8771 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8772 	}
8773 
8774 	if (reg != rsave)
8775 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8776 
8777 	return;
8778 }
8779 
8780 /**
8781  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8782  * @adapter: Pointer to adapter struct
8783  *
8784  * Populate the netdev user priority to tc map
8785  */
8786 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8787 {
8788 	struct net_device *dev = adapter->netdev;
8789 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8790 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8791 	u8 prio;
8792 
8793 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8794 		u8 tc = 0;
8795 
8796 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8797 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8798 		else if (ets)
8799 			tc = ets->prio_tc[prio];
8800 
8801 		netdev_set_prio_tc_map(dev, prio, tc);
8802 	}
8803 }
8804 
8805 #endif /* CONFIG_IXGBE_DCB */
8806 static int ixgbe_reassign_macvlan_pool(struct net_device *vdev, void *data)
8807 {
8808 	struct ixgbe_adapter *adapter = data;
8809 	struct ixgbe_fwd_adapter *accel;
8810 	int pool;
8811 
8812 	/* we only care about macvlans... */
8813 	if (!netif_is_macvlan(vdev))
8814 		return 0;
8815 
8816 	/* that have hardware offload enabled... */
8817 	accel = macvlan_accel_priv(vdev);
8818 	if (!accel)
8819 		return 0;
8820 
8821 	/* If we can relocate to a different bit do so */
8822 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
8823 	if (pool < adapter->num_rx_pools) {
8824 		set_bit(pool, adapter->fwd_bitmask);
8825 		accel->pool = pool;
8826 		return 0;
8827 	}
8828 
8829 	/* if we cannot find a free pool then disable the offload */
8830 	netdev_err(vdev, "L2FW offload disabled due to lack of queue resources\n");
8831 	macvlan_release_l2fw_offload(vdev);
8832 	kfree(accel);
8833 
8834 	return 0;
8835 }
8836 
8837 static void ixgbe_defrag_macvlan_pools(struct net_device *dev)
8838 {
8839 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8840 
8841 	/* flush any stale bits out of the fwd bitmask */
8842 	bitmap_clear(adapter->fwd_bitmask, 1, 63);
8843 
8844 	/* walk through upper devices reassigning pools */
8845 	netdev_walk_all_upper_dev_rcu(dev, ixgbe_reassign_macvlan_pool,
8846 				      adapter);
8847 }
8848 
8849 /**
8850  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8851  *
8852  * @dev: net device to configure
8853  * @tc: number of traffic classes to enable
8854  */
8855 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8856 {
8857 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8858 	struct ixgbe_hw *hw = &adapter->hw;
8859 
8860 	/* Hardware supports up to 8 traffic classes */
8861 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8862 		return -EINVAL;
8863 
8864 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8865 		return -EINVAL;
8866 
8867 	/* Hardware has to reinitialize queues and interrupts to
8868 	 * match packet buffer alignment. Unfortunately, the
8869 	 * hardware is not flexible enough to do this dynamically.
8870 	 */
8871 	if (netif_running(dev))
8872 		ixgbe_close(dev);
8873 	else
8874 		ixgbe_reset(adapter);
8875 
8876 	ixgbe_clear_interrupt_scheme(adapter);
8877 
8878 #ifdef CONFIG_IXGBE_DCB
8879 	if (tc) {
8880 		netdev_set_num_tc(dev, tc);
8881 		ixgbe_set_prio_tc_map(adapter);
8882 
8883 		adapter->hw_tcs = tc;
8884 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8885 
8886 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8887 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8888 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8889 		}
8890 	} else {
8891 		netdev_reset_tc(dev);
8892 
8893 		/* To support macvlan offload we have to use num_tc to
8894 		 * restrict the queues that can be used by the device.
8895 		 * By doing this we can avoid reporting a false number of
8896 		 * queues.
8897 		 */
8898 		if (!tc && adapter->num_rx_pools > 1)
8899 			netdev_set_num_tc(dev, 1);
8900 
8901 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8902 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8903 
8904 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8905 		adapter->hw_tcs = tc;
8906 
8907 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8908 		adapter->dcb_cfg.pfc_mode_enable = false;
8909 	}
8910 
8911 	ixgbe_validate_rtr(adapter, tc);
8912 
8913 #endif /* CONFIG_IXGBE_DCB */
8914 	ixgbe_init_interrupt_scheme(adapter);
8915 
8916 	ixgbe_defrag_macvlan_pools(dev);
8917 
8918 	if (netif_running(dev))
8919 		return ixgbe_open(dev);
8920 
8921 	return 0;
8922 }
8923 
8924 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8925 			       struct tc_cls_u32_offload *cls)
8926 {
8927 	u32 hdl = cls->knode.handle;
8928 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8929 	u32 loc = cls->knode.handle & 0xfffff;
8930 	int err = 0, i, j;
8931 	struct ixgbe_jump_table *jump = NULL;
8932 
8933 	if (loc > IXGBE_MAX_HW_ENTRIES)
8934 		return -EINVAL;
8935 
8936 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8937 		return -EINVAL;
8938 
8939 	/* Clear this filter in the link data it is associated with */
8940 	if (uhtid != 0x800) {
8941 		jump = adapter->jump_tables[uhtid];
8942 		if (!jump)
8943 			return -EINVAL;
8944 		if (!test_bit(loc - 1, jump->child_loc_map))
8945 			return -EINVAL;
8946 		clear_bit(loc - 1, jump->child_loc_map);
8947 	}
8948 
8949 	/* Check if the filter being deleted is a link */
8950 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8951 		jump = adapter->jump_tables[i];
8952 		if (jump && jump->link_hdl == hdl) {
8953 			/* Delete filters in the hardware in the child hash
8954 			 * table associated with this link
8955 			 */
8956 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8957 				if (!test_bit(j, jump->child_loc_map))
8958 					continue;
8959 				spin_lock(&adapter->fdir_perfect_lock);
8960 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8961 								      NULL,
8962 								      j + 1);
8963 				spin_unlock(&adapter->fdir_perfect_lock);
8964 				clear_bit(j, jump->child_loc_map);
8965 			}
8966 			/* Remove resources for this link */
8967 			kfree(jump->input);
8968 			kfree(jump->mask);
8969 			kfree(jump);
8970 			adapter->jump_tables[i] = NULL;
8971 			return err;
8972 		}
8973 	}
8974 
8975 	spin_lock(&adapter->fdir_perfect_lock);
8976 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8977 	spin_unlock(&adapter->fdir_perfect_lock);
8978 	return err;
8979 }
8980 
8981 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8982 					    struct tc_cls_u32_offload *cls)
8983 {
8984 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8985 
8986 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8987 		return -EINVAL;
8988 
8989 	/* This ixgbe devices do not support hash tables at the moment
8990 	 * so abort when given hash tables.
8991 	 */
8992 	if (cls->hnode.divisor > 0)
8993 		return -EINVAL;
8994 
8995 	set_bit(uhtid - 1, &adapter->tables);
8996 	return 0;
8997 }
8998 
8999 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
9000 					    struct tc_cls_u32_offload *cls)
9001 {
9002 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9003 
9004 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9005 		return -EINVAL;
9006 
9007 	clear_bit(uhtid - 1, &adapter->tables);
9008 	return 0;
9009 }
9010 
9011 #ifdef CONFIG_NET_CLS_ACT
9012 struct upper_walk_data {
9013 	struct ixgbe_adapter *adapter;
9014 	u64 action;
9015 	int ifindex;
9016 	u8 queue;
9017 };
9018 
9019 static int get_macvlan_queue(struct net_device *upper, void *_data)
9020 {
9021 	if (netif_is_macvlan(upper)) {
9022 		struct ixgbe_fwd_adapter *vadapter = macvlan_accel_priv(upper);
9023 		struct upper_walk_data *data = _data;
9024 		struct ixgbe_adapter *adapter = data->adapter;
9025 		int ifindex = data->ifindex;
9026 
9027 		if (vadapter && upper->ifindex == ifindex) {
9028 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
9029 			data->action = data->queue;
9030 			return 1;
9031 		}
9032 	}
9033 
9034 	return 0;
9035 }
9036 
9037 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
9038 				  u8 *queue, u64 *action)
9039 {
9040 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
9041 	unsigned int num_vfs = adapter->num_vfs, vf;
9042 	struct upper_walk_data data;
9043 	struct net_device *upper;
9044 
9045 	/* redirect to a SRIOV VF */
9046 	for (vf = 0; vf < num_vfs; ++vf) {
9047 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
9048 		if (upper->ifindex == ifindex) {
9049 			*queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
9050 			*action = vf + 1;
9051 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
9052 			return 0;
9053 		}
9054 	}
9055 
9056 	/* redirect to a offloaded macvlan netdev */
9057 	data.adapter = adapter;
9058 	data.ifindex = ifindex;
9059 	data.action = 0;
9060 	data.queue = 0;
9061 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
9062 					  get_macvlan_queue, &data)) {
9063 		*action = data.action;
9064 		*queue = data.queue;
9065 
9066 		return 0;
9067 	}
9068 
9069 	return -EINVAL;
9070 }
9071 
9072 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9073 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9074 {
9075 	const struct tc_action *a;
9076 	LIST_HEAD(actions);
9077 	int err;
9078 
9079 	if (!tcf_exts_has_actions(exts))
9080 		return -EINVAL;
9081 
9082 	tcf_exts_to_list(exts, &actions);
9083 	list_for_each_entry(a, &actions, list) {
9084 
9085 		/* Drop action */
9086 		if (is_tcf_gact_shot(a)) {
9087 			*action = IXGBE_FDIR_DROP_QUEUE;
9088 			*queue = IXGBE_FDIR_DROP_QUEUE;
9089 			return 0;
9090 		}
9091 
9092 		/* Redirect to a VF or a offloaded macvlan */
9093 		if (is_tcf_mirred_egress_redirect(a)) {
9094 			struct net_device *dev = tcf_mirred_dev(a);
9095 
9096 			if (!dev)
9097 				return -EINVAL;
9098 			err = handle_redirect_action(adapter, dev->ifindex, queue,
9099 						     action);
9100 			if (err == 0)
9101 				return err;
9102 		}
9103 	}
9104 
9105 	return -EINVAL;
9106 }
9107 #else
9108 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9109 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9110 {
9111 	return -EINVAL;
9112 }
9113 #endif /* CONFIG_NET_CLS_ACT */
9114 
9115 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
9116 				    union ixgbe_atr_input *mask,
9117 				    struct tc_cls_u32_offload *cls,
9118 				    struct ixgbe_mat_field *field_ptr,
9119 				    struct ixgbe_nexthdr *nexthdr)
9120 {
9121 	int i, j, off;
9122 	__be32 val, m;
9123 	bool found_entry = false, found_jump_field = false;
9124 
9125 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
9126 		off = cls->knode.sel->keys[i].off;
9127 		val = cls->knode.sel->keys[i].val;
9128 		m = cls->knode.sel->keys[i].mask;
9129 
9130 		for (j = 0; field_ptr[j].val; j++) {
9131 			if (field_ptr[j].off == off) {
9132 				field_ptr[j].val(input, mask, val, m);
9133 				input->filter.formatted.flow_type |=
9134 					field_ptr[j].type;
9135 				found_entry = true;
9136 				break;
9137 			}
9138 		}
9139 		if (nexthdr) {
9140 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
9141 			    nexthdr->val == cls->knode.sel->keys[i].val &&
9142 			    nexthdr->mask == cls->knode.sel->keys[i].mask)
9143 				found_jump_field = true;
9144 			else
9145 				continue;
9146 		}
9147 	}
9148 
9149 	if (nexthdr && !found_jump_field)
9150 		return -EINVAL;
9151 
9152 	if (!found_entry)
9153 		return 0;
9154 
9155 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9156 				    IXGBE_ATR_L4TYPE_MASK;
9157 
9158 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9159 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9160 
9161 	return 0;
9162 }
9163 
9164 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9165 				  struct tc_cls_u32_offload *cls)
9166 {
9167 	__be16 protocol = cls->common.protocol;
9168 	u32 loc = cls->knode.handle & 0xfffff;
9169 	struct ixgbe_hw *hw = &adapter->hw;
9170 	struct ixgbe_mat_field *field_ptr;
9171 	struct ixgbe_fdir_filter *input = NULL;
9172 	union ixgbe_atr_input *mask = NULL;
9173 	struct ixgbe_jump_table *jump = NULL;
9174 	int i, err = -EINVAL;
9175 	u8 queue;
9176 	u32 uhtid, link_uhtid;
9177 
9178 	uhtid = TC_U32_USERHTID(cls->knode.handle);
9179 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9180 
9181 	/* At the moment cls_u32 jumps to network layer and skips past
9182 	 * L2 headers. The canonical method to match L2 frames is to use
9183 	 * negative values. However this is error prone at best but really
9184 	 * just broken because there is no way to "know" what sort of hdr
9185 	 * is in front of the network layer. Fix cls_u32 to support L2
9186 	 * headers when needed.
9187 	 */
9188 	if (protocol != htons(ETH_P_IP))
9189 		return err;
9190 
9191 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9192 		e_err(drv, "Location out of range\n");
9193 		return err;
9194 	}
9195 
9196 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
9197 	 * links and also the fields used to advance the parser across each
9198 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9199 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9200 	 * To add support for new nodes update ixgbe_model.h parse structures
9201 	 * this function _should_ be generic try not to hardcode values here.
9202 	 */
9203 	if (uhtid == 0x800) {
9204 		field_ptr = (adapter->jump_tables[0])->mat;
9205 	} else {
9206 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9207 			return err;
9208 		if (!adapter->jump_tables[uhtid])
9209 			return err;
9210 		field_ptr = (adapter->jump_tables[uhtid])->mat;
9211 	}
9212 
9213 	if (!field_ptr)
9214 		return err;
9215 
9216 	/* At this point we know the field_ptr is valid and need to either
9217 	 * build cls_u32 link or attach filter. Because adding a link to
9218 	 * a handle that does not exist is invalid and the same for adding
9219 	 * rules to handles that don't exist.
9220 	 */
9221 
9222 	if (link_uhtid) {
9223 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9224 
9225 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9226 			return err;
9227 
9228 		if (!test_bit(link_uhtid - 1, &adapter->tables))
9229 			return err;
9230 
9231 		/* Multiple filters as links to the same hash table are not
9232 		 * supported. To add a new filter with the same next header
9233 		 * but different match/jump conditions, create a new hash table
9234 		 * and link to it.
9235 		 */
9236 		if (adapter->jump_tables[link_uhtid] &&
9237 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
9238 			e_err(drv, "Link filter exists for link: %x\n",
9239 			      link_uhtid);
9240 			return err;
9241 		}
9242 
9243 		for (i = 0; nexthdr[i].jump; i++) {
9244 			if (nexthdr[i].o != cls->knode.sel->offoff ||
9245 			    nexthdr[i].s != cls->knode.sel->offshift ||
9246 			    nexthdr[i].m != cls->knode.sel->offmask)
9247 				return err;
9248 
9249 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9250 			if (!jump)
9251 				return -ENOMEM;
9252 			input = kzalloc(sizeof(*input), GFP_KERNEL);
9253 			if (!input) {
9254 				err = -ENOMEM;
9255 				goto free_jump;
9256 			}
9257 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9258 			if (!mask) {
9259 				err = -ENOMEM;
9260 				goto free_input;
9261 			}
9262 			jump->input = input;
9263 			jump->mask = mask;
9264 			jump->link_hdl = cls->knode.handle;
9265 
9266 			err = ixgbe_clsu32_build_input(input, mask, cls,
9267 						       field_ptr, &nexthdr[i]);
9268 			if (!err) {
9269 				jump->mat = nexthdr[i].jump;
9270 				adapter->jump_tables[link_uhtid] = jump;
9271 				break;
9272 			}
9273 		}
9274 		return 0;
9275 	}
9276 
9277 	input = kzalloc(sizeof(*input), GFP_KERNEL);
9278 	if (!input)
9279 		return -ENOMEM;
9280 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9281 	if (!mask) {
9282 		err = -ENOMEM;
9283 		goto free_input;
9284 	}
9285 
9286 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9287 		if ((adapter->jump_tables[uhtid])->input)
9288 			memcpy(input, (adapter->jump_tables[uhtid])->input,
9289 			       sizeof(*input));
9290 		if ((adapter->jump_tables[uhtid])->mask)
9291 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9292 			       sizeof(*mask));
9293 
9294 		/* Lookup in all child hash tables if this location is already
9295 		 * filled with a filter
9296 		 */
9297 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9298 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
9299 
9300 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
9301 				e_err(drv, "Filter exists in location: %x\n",
9302 				      loc);
9303 				err = -EINVAL;
9304 				goto err_out;
9305 			}
9306 		}
9307 	}
9308 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9309 	if (err)
9310 		goto err_out;
9311 
9312 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9313 			       &queue);
9314 	if (err < 0)
9315 		goto err_out;
9316 
9317 	input->sw_idx = loc;
9318 
9319 	spin_lock(&adapter->fdir_perfect_lock);
9320 
9321 	if (hlist_empty(&adapter->fdir_filter_list)) {
9322 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9323 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9324 		if (err)
9325 			goto err_out_w_lock;
9326 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9327 		err = -EINVAL;
9328 		goto err_out_w_lock;
9329 	}
9330 
9331 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9332 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9333 						    input->sw_idx, queue);
9334 	if (!err)
9335 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9336 	spin_unlock(&adapter->fdir_perfect_lock);
9337 
9338 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9339 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9340 
9341 	kfree(mask);
9342 	return err;
9343 err_out_w_lock:
9344 	spin_unlock(&adapter->fdir_perfect_lock);
9345 err_out:
9346 	kfree(mask);
9347 free_input:
9348 	kfree(input);
9349 free_jump:
9350 	kfree(jump);
9351 	return err;
9352 }
9353 
9354 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
9355 				  struct tc_cls_u32_offload *cls_u32)
9356 {
9357 	switch (cls_u32->command) {
9358 	case TC_CLSU32_NEW_KNODE:
9359 	case TC_CLSU32_REPLACE_KNODE:
9360 		return ixgbe_configure_clsu32(adapter, cls_u32);
9361 	case TC_CLSU32_DELETE_KNODE:
9362 		return ixgbe_delete_clsu32(adapter, cls_u32);
9363 	case TC_CLSU32_NEW_HNODE:
9364 	case TC_CLSU32_REPLACE_HNODE:
9365 		return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9366 	case TC_CLSU32_DELETE_HNODE:
9367 		return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9368 	default:
9369 		return -EOPNOTSUPP;
9370 	}
9371 }
9372 
9373 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
9374 				   void *cb_priv)
9375 {
9376 	struct ixgbe_adapter *adapter = cb_priv;
9377 
9378 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
9379 		return -EOPNOTSUPP;
9380 
9381 	switch (type) {
9382 	case TC_SETUP_CLSU32:
9383 		return ixgbe_setup_tc_cls_u32(adapter, type_data);
9384 	default:
9385 		return -EOPNOTSUPP;
9386 	}
9387 }
9388 
9389 static int ixgbe_setup_tc_block(struct net_device *dev,
9390 				struct tc_block_offload *f)
9391 {
9392 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9393 
9394 	if (f->binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
9395 		return -EOPNOTSUPP;
9396 
9397 	switch (f->command) {
9398 	case TC_BLOCK_BIND:
9399 		return tcf_block_cb_register(f->block, ixgbe_setup_tc_block_cb,
9400 					     adapter, adapter);
9401 	case TC_BLOCK_UNBIND:
9402 		tcf_block_cb_unregister(f->block, ixgbe_setup_tc_block_cb,
9403 					adapter);
9404 		return 0;
9405 	default:
9406 		return -EOPNOTSUPP;
9407 	}
9408 }
9409 
9410 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9411 				 struct tc_mqprio_qopt *mqprio)
9412 {
9413 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9414 	return ixgbe_setup_tc(dev, mqprio->num_tc);
9415 }
9416 
9417 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9418 			    void *type_data)
9419 {
9420 	switch (type) {
9421 	case TC_SETUP_BLOCK:
9422 		return ixgbe_setup_tc_block(dev, type_data);
9423 	case TC_SETUP_QDISC_MQPRIO:
9424 		return ixgbe_setup_tc_mqprio(dev, type_data);
9425 	default:
9426 		return -EOPNOTSUPP;
9427 	}
9428 }
9429 
9430 #ifdef CONFIG_PCI_IOV
9431 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9432 {
9433 	struct net_device *netdev = adapter->netdev;
9434 
9435 	rtnl_lock();
9436 	ixgbe_setup_tc(netdev, adapter->hw_tcs);
9437 	rtnl_unlock();
9438 }
9439 
9440 #endif
9441 void ixgbe_do_reset(struct net_device *netdev)
9442 {
9443 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9444 
9445 	if (netif_running(netdev))
9446 		ixgbe_reinit_locked(adapter);
9447 	else
9448 		ixgbe_reset(adapter);
9449 }
9450 
9451 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9452 					    netdev_features_t features)
9453 {
9454 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9455 
9456 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9457 	if (!(features & NETIF_F_RXCSUM))
9458 		features &= ~NETIF_F_LRO;
9459 
9460 	/* Turn off LRO if not RSC capable */
9461 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9462 		features &= ~NETIF_F_LRO;
9463 
9464 	return features;
9465 }
9466 
9467 static void ixgbe_reset_l2fw_offload(struct ixgbe_adapter *adapter)
9468 {
9469 	int rss = min_t(int, ixgbe_max_rss_indices(adapter),
9470 			num_online_cpus());
9471 
9472 	/* go back to full RSS if we're not running SR-IOV */
9473 	if (!adapter->ring_feature[RING_F_VMDQ].offset)
9474 		adapter->flags &= ~(IXGBE_FLAG_VMDQ_ENABLED |
9475 				    IXGBE_FLAG_SRIOV_ENABLED);
9476 
9477 	adapter->ring_feature[RING_F_RSS].limit = rss;
9478 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
9479 
9480 	ixgbe_setup_tc(adapter->netdev, adapter->hw_tcs);
9481 }
9482 
9483 static int ixgbe_set_features(struct net_device *netdev,
9484 			      netdev_features_t features)
9485 {
9486 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9487 	netdev_features_t changed = netdev->features ^ features;
9488 	bool need_reset = false;
9489 
9490 	/* Make sure RSC matches LRO, reset if change */
9491 	if (!(features & NETIF_F_LRO)) {
9492 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9493 			need_reset = true;
9494 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9495 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9496 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9497 		if (adapter->rx_itr_setting == 1 ||
9498 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9499 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9500 			need_reset = true;
9501 		} else if ((changed ^ features) & NETIF_F_LRO) {
9502 			e_info(probe, "rx-usecs set too low, "
9503 			       "disabling RSC\n");
9504 		}
9505 	}
9506 
9507 	/*
9508 	 * Check if Flow Director n-tuple support or hw_tc support was
9509 	 * enabled or disabled.  If the state changed, we need to reset.
9510 	 */
9511 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9512 		/* turn off ATR, enable perfect filters and reset */
9513 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9514 			need_reset = true;
9515 
9516 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9517 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9518 	} else {
9519 		/* turn off perfect filters, enable ATR and reset */
9520 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9521 			need_reset = true;
9522 
9523 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9524 
9525 		/* We cannot enable ATR if SR-IOV is enabled */
9526 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9527 		    /* We cannot enable ATR if we have 2 or more tcs */
9528 		    (adapter->hw_tcs > 1) ||
9529 		    /* We cannot enable ATR if RSS is disabled */
9530 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9531 		    /* A sample rate of 0 indicates ATR disabled */
9532 		    (!adapter->atr_sample_rate))
9533 			; /* do nothing not supported */
9534 		else /* otherwise supported and set the flag */
9535 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9536 	}
9537 
9538 	if (changed & NETIF_F_RXALL)
9539 		need_reset = true;
9540 
9541 	netdev->features = features;
9542 
9543 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
9544 		if (features & NETIF_F_RXCSUM) {
9545 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9546 		} else {
9547 			u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9548 
9549 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9550 		}
9551 	}
9552 
9553 	if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
9554 		if (features & NETIF_F_RXCSUM) {
9555 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9556 		} else {
9557 			u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9558 
9559 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9560 		}
9561 	}
9562 
9563 	if ((changed & NETIF_F_HW_L2FW_DOFFLOAD) && adapter->num_rx_pools > 1)
9564 		ixgbe_reset_l2fw_offload(adapter);
9565 	else if (need_reset)
9566 		ixgbe_do_reset(netdev);
9567 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9568 			    NETIF_F_HW_VLAN_CTAG_FILTER))
9569 		ixgbe_set_rx_mode(netdev);
9570 
9571 	return 0;
9572 }
9573 
9574 /**
9575  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
9576  * @dev: The port's netdev
9577  * @ti: Tunnel endpoint information
9578  **/
9579 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
9580 				      struct udp_tunnel_info *ti)
9581 {
9582 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9583 	struct ixgbe_hw *hw = &adapter->hw;
9584 	__be16 port = ti->port;
9585 	u32 port_shift = 0;
9586 	u32 reg;
9587 
9588 	if (ti->sa_family != AF_INET)
9589 		return;
9590 
9591 	switch (ti->type) {
9592 	case UDP_TUNNEL_TYPE_VXLAN:
9593 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9594 			return;
9595 
9596 		if (adapter->vxlan_port == port)
9597 			return;
9598 
9599 		if (adapter->vxlan_port) {
9600 			netdev_info(dev,
9601 				    "VXLAN port %d set, not adding port %d\n",
9602 				    ntohs(adapter->vxlan_port),
9603 				    ntohs(port));
9604 			return;
9605 		}
9606 
9607 		adapter->vxlan_port = port;
9608 		break;
9609 	case UDP_TUNNEL_TYPE_GENEVE:
9610 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9611 			return;
9612 
9613 		if (adapter->geneve_port == port)
9614 			return;
9615 
9616 		if (adapter->geneve_port) {
9617 			netdev_info(dev,
9618 				    "GENEVE port %d set, not adding port %d\n",
9619 				    ntohs(adapter->geneve_port),
9620 				    ntohs(port));
9621 			return;
9622 		}
9623 
9624 		port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
9625 		adapter->geneve_port = port;
9626 		break;
9627 	default:
9628 		return;
9629 	}
9630 
9631 	reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
9632 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
9633 }
9634 
9635 /**
9636  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
9637  * @dev: The port's netdev
9638  * @ti: Tunnel endpoint information
9639  **/
9640 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
9641 				      struct udp_tunnel_info *ti)
9642 {
9643 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9644 	u32 port_mask;
9645 
9646 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
9647 	    ti->type != UDP_TUNNEL_TYPE_GENEVE)
9648 		return;
9649 
9650 	if (ti->sa_family != AF_INET)
9651 		return;
9652 
9653 	switch (ti->type) {
9654 	case UDP_TUNNEL_TYPE_VXLAN:
9655 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9656 			return;
9657 
9658 		if (adapter->vxlan_port != ti->port) {
9659 			netdev_info(dev, "VXLAN port %d not found\n",
9660 				    ntohs(ti->port));
9661 			return;
9662 		}
9663 
9664 		port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9665 		break;
9666 	case UDP_TUNNEL_TYPE_GENEVE:
9667 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9668 			return;
9669 
9670 		if (adapter->geneve_port != ti->port) {
9671 			netdev_info(dev, "GENEVE port %d not found\n",
9672 				    ntohs(ti->port));
9673 			return;
9674 		}
9675 
9676 		port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9677 		break;
9678 	default:
9679 		return;
9680 	}
9681 
9682 	ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9683 	adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9684 }
9685 
9686 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9687 			     struct net_device *dev,
9688 			     const unsigned char *addr, u16 vid,
9689 			     u16 flags)
9690 {
9691 	/* guarantee we can provide a unique filter for the unicast address */
9692 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9693 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9694 		u16 pool = VMDQ_P(0);
9695 
9696 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9697 			return -ENOMEM;
9698 	}
9699 
9700 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9701 }
9702 
9703 /**
9704  * ixgbe_configure_bridge_mode - set various bridge modes
9705  * @adapter: the private structure
9706  * @mode: requested bridge mode
9707  *
9708  * Configure some settings require for various bridge modes.
9709  **/
9710 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9711 				       __u16 mode)
9712 {
9713 	struct ixgbe_hw *hw = &adapter->hw;
9714 	unsigned int p, num_pools;
9715 	u32 vmdctl;
9716 
9717 	switch (mode) {
9718 	case BRIDGE_MODE_VEPA:
9719 		/* disable Tx loopback, rely on switch hairpin mode */
9720 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9721 
9722 		/* must enable Rx switching replication to allow multicast
9723 		 * packet reception on all VFs, and to enable source address
9724 		 * pruning.
9725 		 */
9726 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9727 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9728 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9729 
9730 		/* enable Rx source address pruning. Note, this requires
9731 		 * replication to be enabled or else it does nothing.
9732 		 */
9733 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9734 		for (p = 0; p < num_pools; p++) {
9735 			if (hw->mac.ops.set_source_address_pruning)
9736 				hw->mac.ops.set_source_address_pruning(hw,
9737 								       true,
9738 								       p);
9739 		}
9740 		break;
9741 	case BRIDGE_MODE_VEB:
9742 		/* enable Tx loopback for internal VF/PF communication */
9743 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9744 				IXGBE_PFDTXGSWC_VT_LBEN);
9745 
9746 		/* disable Rx switching replication unless we have SR-IOV
9747 		 * virtual functions
9748 		 */
9749 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9750 		if (!adapter->num_vfs)
9751 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9752 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9753 
9754 		/* disable Rx source address pruning, since we don't expect to
9755 		 * be receiving external loopback of our transmitted frames.
9756 		 */
9757 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9758 		for (p = 0; p < num_pools; p++) {
9759 			if (hw->mac.ops.set_source_address_pruning)
9760 				hw->mac.ops.set_source_address_pruning(hw,
9761 								       false,
9762 								       p);
9763 		}
9764 		break;
9765 	default:
9766 		return -EINVAL;
9767 	}
9768 
9769 	adapter->bridge_mode = mode;
9770 
9771 	e_info(drv, "enabling bridge mode: %s\n",
9772 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9773 
9774 	return 0;
9775 }
9776 
9777 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9778 				    struct nlmsghdr *nlh, u16 flags)
9779 {
9780 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9781 	struct nlattr *attr, *br_spec;
9782 	int rem;
9783 
9784 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9785 		return -EOPNOTSUPP;
9786 
9787 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9788 	if (!br_spec)
9789 		return -EINVAL;
9790 
9791 	nla_for_each_nested(attr, br_spec, rem) {
9792 		int status;
9793 		__u16 mode;
9794 
9795 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9796 			continue;
9797 
9798 		if (nla_len(attr) < sizeof(mode))
9799 			return -EINVAL;
9800 
9801 		mode = nla_get_u16(attr);
9802 		status = ixgbe_configure_bridge_mode(adapter, mode);
9803 		if (status)
9804 			return status;
9805 
9806 		break;
9807 	}
9808 
9809 	return 0;
9810 }
9811 
9812 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9813 				    struct net_device *dev,
9814 				    u32 filter_mask, int nlflags)
9815 {
9816 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9817 
9818 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9819 		return 0;
9820 
9821 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9822 				       adapter->bridge_mode, 0, 0, nlflags,
9823 				       filter_mask, NULL);
9824 }
9825 
9826 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9827 {
9828 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9829 	struct ixgbe_fwd_adapter *accel;
9830 	int tcs = adapter->hw_tcs ? : 1;
9831 	int pool, err;
9832 
9833 	/* The hardware supported by ixgbe only filters on the destination MAC
9834 	 * address. In order to avoid issues we only support offloading modes
9835 	 * where the hardware can actually provide the functionality.
9836 	 */
9837 	if (!macvlan_supports_dest_filter(vdev))
9838 		return ERR_PTR(-EMEDIUMTYPE);
9839 
9840 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9841 	if (pool == adapter->num_rx_pools) {
9842 		u16 used_pools = adapter->num_vfs + adapter->num_rx_pools;
9843 		u16 reserved_pools;
9844 
9845 		if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9846 		     adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
9847 		    adapter->num_rx_pools > IXGBE_MAX_MACVLANS)
9848 			return ERR_PTR(-EBUSY);
9849 
9850 		/* Hardware has a limited number of available pools. Each VF,
9851 		 * and the PF require a pool. Check to ensure we don't
9852 		 * attempt to use more then the available number of pools.
9853 		 */
9854 		if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9855 			return ERR_PTR(-EBUSY);
9856 
9857 		/* Enable VMDq flag so device will be set in VM mode */
9858 		adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED |
9859 				  IXGBE_FLAG_SRIOV_ENABLED;
9860 
9861 		/* Try to reserve as many queues per pool as possible,
9862 		 * we start with the configurations that support 4 queues
9863 		 * per pools, followed by 2, and then by just 1 per pool.
9864 		 */
9865 		if (used_pools < 32 && adapter->num_rx_pools < 16)
9866 			reserved_pools = min_t(u16,
9867 					       32 - used_pools,
9868 					       16 - adapter->num_rx_pools);
9869 		else if (adapter->num_rx_pools < 32)
9870 			reserved_pools = min_t(u16,
9871 					       64 - used_pools,
9872 					       32 - adapter->num_rx_pools);
9873 		else
9874 			reserved_pools = 64 - used_pools;
9875 
9876 
9877 		if (!reserved_pools)
9878 			return ERR_PTR(-EBUSY);
9879 
9880 		adapter->ring_feature[RING_F_VMDQ].limit += reserved_pools;
9881 
9882 		/* Force reinit of ring allocation with VMDQ enabled */
9883 		err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
9884 		if (err)
9885 			return ERR_PTR(err);
9886 
9887 		if (pool >= adapter->num_rx_pools)
9888 			return ERR_PTR(-ENOMEM);
9889 	}
9890 
9891 	accel = kzalloc(sizeof(*accel), GFP_KERNEL);
9892 	if (!accel)
9893 		return ERR_PTR(-ENOMEM);
9894 
9895 	set_bit(pool, adapter->fwd_bitmask);
9896 	accel->pool = pool;
9897 	accel->netdev = vdev;
9898 
9899 	if (!netif_running(pdev))
9900 		return accel;
9901 
9902 	err = ixgbe_fwd_ring_up(adapter, accel);
9903 	if (err)
9904 		return ERR_PTR(err);
9905 
9906 	return accel;
9907 }
9908 
9909 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9910 {
9911 	struct ixgbe_fwd_adapter *accel = priv;
9912 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9913 	unsigned int rxbase = accel->rx_base_queue;
9914 	unsigned int i;
9915 
9916 	/* delete unicast filter associated with offloaded interface */
9917 	ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
9918 			     VMDQ_P(accel->pool));
9919 
9920 	/* Allow remaining Rx packets to get flushed out of the
9921 	 * Rx FIFO before we drop the netdev for the ring.
9922 	 */
9923 	usleep_range(10000, 20000);
9924 
9925 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
9926 		struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
9927 		struct ixgbe_q_vector *qv = ring->q_vector;
9928 
9929 		/* Make sure we aren't processing any packets and clear
9930 		 * netdev to shut down the ring.
9931 		 */
9932 		if (netif_running(adapter->netdev))
9933 			napi_synchronize(&qv->napi);
9934 		ring->netdev = NULL;
9935 	}
9936 
9937 	clear_bit(accel->pool, adapter->fwd_bitmask);
9938 	kfree(accel);
9939 }
9940 
9941 #define IXGBE_MAX_MAC_HDR_LEN		127
9942 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9943 
9944 static netdev_features_t
9945 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9946 		     netdev_features_t features)
9947 {
9948 	unsigned int network_hdr_len, mac_hdr_len;
9949 
9950 	/* Make certain the headers can be described by a context descriptor */
9951 	mac_hdr_len = skb_network_header(skb) - skb->data;
9952 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9953 		return features & ~(NETIF_F_HW_CSUM |
9954 				    NETIF_F_SCTP_CRC |
9955 				    NETIF_F_HW_VLAN_CTAG_TX |
9956 				    NETIF_F_TSO |
9957 				    NETIF_F_TSO6);
9958 
9959 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9960 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9961 		return features & ~(NETIF_F_HW_CSUM |
9962 				    NETIF_F_SCTP_CRC |
9963 				    NETIF_F_TSO |
9964 				    NETIF_F_TSO6);
9965 
9966 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9967 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9968 	 * IPsec offoad sets skb->encapsulation but still can handle
9969 	 * the TSO, so it's the exception.
9970 	 */
9971 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID)) {
9972 #ifdef CONFIG_XFRM
9973 		if (!skb->sp)
9974 #endif
9975 			features &= ~NETIF_F_TSO;
9976 	}
9977 
9978 	return features;
9979 }
9980 
9981 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
9982 {
9983 	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
9984 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9985 	struct bpf_prog *old_prog;
9986 
9987 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
9988 		return -EINVAL;
9989 
9990 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
9991 		return -EINVAL;
9992 
9993 	/* verify ixgbe ring attributes are sufficient for XDP */
9994 	for (i = 0; i < adapter->num_rx_queues; i++) {
9995 		struct ixgbe_ring *ring = adapter->rx_ring[i];
9996 
9997 		if (ring_is_rsc_enabled(ring))
9998 			return -EINVAL;
9999 
10000 		if (frame_size > ixgbe_rx_bufsz(ring))
10001 			return -EINVAL;
10002 	}
10003 
10004 	if (nr_cpu_ids > MAX_XDP_QUEUES)
10005 		return -ENOMEM;
10006 
10007 	old_prog = xchg(&adapter->xdp_prog, prog);
10008 
10009 	/* If transitioning XDP modes reconfigure rings */
10010 	if (!!prog != !!old_prog) {
10011 		int err = ixgbe_setup_tc(dev, adapter->hw_tcs);
10012 
10013 		if (err) {
10014 			rcu_assign_pointer(adapter->xdp_prog, old_prog);
10015 			return -EINVAL;
10016 		}
10017 	} else {
10018 		for (i = 0; i < adapter->num_rx_queues; i++)
10019 			xchg(&adapter->rx_ring[i]->xdp_prog, adapter->xdp_prog);
10020 	}
10021 
10022 	if (old_prog)
10023 		bpf_prog_put(old_prog);
10024 
10025 	return 0;
10026 }
10027 
10028 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
10029 {
10030 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10031 
10032 	switch (xdp->command) {
10033 	case XDP_SETUP_PROG:
10034 		return ixgbe_xdp_setup(dev, xdp->prog);
10035 	case XDP_QUERY_PROG:
10036 		xdp->prog_attached = !!(adapter->xdp_prog);
10037 		xdp->prog_id = adapter->xdp_prog ?
10038 			adapter->xdp_prog->aux->id : 0;
10039 		return 0;
10040 	default:
10041 		return -EINVAL;
10042 	}
10043 }
10044 
10045 static int ixgbe_xdp_xmit(struct net_device *dev, struct xdp_frame *xdpf)
10046 {
10047 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10048 	struct ixgbe_ring *ring;
10049 	int err;
10050 
10051 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10052 		return -ENETDOWN;
10053 
10054 	/* During program transitions its possible adapter->xdp_prog is assigned
10055 	 * but ring has not been configured yet. In this case simply abort xmit.
10056 	 */
10057 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
10058 	if (unlikely(!ring))
10059 		return -ENXIO;
10060 
10061 	err = ixgbe_xmit_xdp_ring(adapter, xdpf);
10062 	if (err != IXGBE_XDP_TX)
10063 		return -ENOSPC;
10064 
10065 	return 0;
10066 }
10067 
10068 static void ixgbe_xdp_flush(struct net_device *dev)
10069 {
10070 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10071 	struct ixgbe_ring *ring;
10072 
10073 	/* Its possible the device went down between xdp xmit and flush so
10074 	 * we need to ensure device is still up.
10075 	 */
10076 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10077 		return;
10078 
10079 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
10080 	if (unlikely(!ring))
10081 		return;
10082 
10083 	/* Force memory writes to complete before letting h/w know there
10084 	 * are new descriptors to fetch.
10085 	 */
10086 	wmb();
10087 	writel(ring->next_to_use, ring->tail);
10088 
10089 	return;
10090 }
10091 
10092 static const struct net_device_ops ixgbe_netdev_ops = {
10093 	.ndo_open		= ixgbe_open,
10094 	.ndo_stop		= ixgbe_close,
10095 	.ndo_start_xmit		= ixgbe_xmit_frame,
10096 	.ndo_select_queue	= ixgbe_select_queue,
10097 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
10098 	.ndo_validate_addr	= eth_validate_addr,
10099 	.ndo_set_mac_address	= ixgbe_set_mac,
10100 	.ndo_change_mtu		= ixgbe_change_mtu,
10101 	.ndo_tx_timeout		= ixgbe_tx_timeout,
10102 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
10103 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
10104 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
10105 	.ndo_do_ioctl		= ixgbe_ioctl,
10106 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
10107 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
10108 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
10109 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
10110 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
10111 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
10112 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
10113 	.ndo_get_stats64	= ixgbe_get_stats64,
10114 	.ndo_setup_tc		= __ixgbe_setup_tc,
10115 #ifdef CONFIG_NET_POLL_CONTROLLER
10116 	.ndo_poll_controller	= ixgbe_netpoll,
10117 #endif
10118 #ifdef IXGBE_FCOE
10119 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
10120 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
10121 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
10122 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
10123 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
10124 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
10125 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
10126 #endif /* IXGBE_FCOE */
10127 	.ndo_set_features = ixgbe_set_features,
10128 	.ndo_fix_features = ixgbe_fix_features,
10129 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
10130 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
10131 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
10132 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
10133 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
10134 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
10135 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
10136 	.ndo_features_check	= ixgbe_features_check,
10137 	.ndo_bpf		= ixgbe_xdp,
10138 	.ndo_xdp_xmit		= ixgbe_xdp_xmit,
10139 	.ndo_xdp_flush		= ixgbe_xdp_flush,
10140 };
10141 
10142 /**
10143  * ixgbe_enumerate_functions - Get the number of ports this device has
10144  * @adapter: adapter structure
10145  *
10146  * This function enumerates the phsyical functions co-located on a single slot,
10147  * in order to determine how many ports a device has. This is most useful in
10148  * determining the required GT/s of PCIe bandwidth necessary for optimal
10149  * performance.
10150  **/
10151 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
10152 {
10153 	struct pci_dev *entry, *pdev = adapter->pdev;
10154 	int physfns = 0;
10155 
10156 	/* Some cards can not use the generic count PCIe functions method,
10157 	 * because they are behind a parent switch, so we hardcode these with
10158 	 * the correct number of functions.
10159 	 */
10160 	if (ixgbe_pcie_from_parent(&adapter->hw))
10161 		physfns = 4;
10162 
10163 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
10164 		/* don't count virtual functions */
10165 		if (entry->is_virtfn)
10166 			continue;
10167 
10168 		/* When the devices on the bus don't all match our device ID,
10169 		 * we can't reliably determine the correct number of
10170 		 * functions. This can occur if a function has been direct
10171 		 * attached to a virtual machine using VT-d, for example. In
10172 		 * this case, simply return -1 to indicate this.
10173 		 */
10174 		if ((entry->vendor != pdev->vendor) ||
10175 		    (entry->device != pdev->device))
10176 			return -1;
10177 
10178 		physfns++;
10179 	}
10180 
10181 	return physfns;
10182 }
10183 
10184 /**
10185  * ixgbe_wol_supported - Check whether device supports WoL
10186  * @adapter: the adapter private structure
10187  * @device_id: the device ID
10188  * @subdevice_id: the subsystem device ID
10189  *
10190  * This function is used by probe and ethtool to determine
10191  * which devices have WoL support
10192  *
10193  **/
10194 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
10195 			 u16 subdevice_id)
10196 {
10197 	struct ixgbe_hw *hw = &adapter->hw;
10198 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
10199 
10200 	/* WOL not supported on 82598 */
10201 	if (hw->mac.type == ixgbe_mac_82598EB)
10202 		return false;
10203 
10204 	/* check eeprom to see if WOL is enabled for X540 and newer */
10205 	if (hw->mac.type >= ixgbe_mac_X540) {
10206 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10207 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10208 		     (hw->bus.func == 0)))
10209 			return true;
10210 	}
10211 
10212 	/* WOL is determined based on device IDs for 82599 MACs */
10213 	switch (device_id) {
10214 	case IXGBE_DEV_ID_82599_SFP:
10215 		/* Only these subdevices could supports WOL */
10216 		switch (subdevice_id) {
10217 		case IXGBE_SUBDEV_ID_82599_560FLR:
10218 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10219 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10220 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10221 			/* only support first port */
10222 			if (hw->bus.func != 0)
10223 				break;
10224 			/* fall through */
10225 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10226 		case IXGBE_SUBDEV_ID_82599_SFP:
10227 		case IXGBE_SUBDEV_ID_82599_RNDC:
10228 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10229 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10230 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10231 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10232 			return true;
10233 		}
10234 		break;
10235 	case IXGBE_DEV_ID_82599EN_SFP:
10236 		/* Only these subdevices support WOL */
10237 		switch (subdevice_id) {
10238 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10239 			return true;
10240 		}
10241 		break;
10242 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10243 		/* All except this subdevice support WOL */
10244 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10245 			return true;
10246 		break;
10247 	case IXGBE_DEV_ID_82599_KX4:
10248 		return  true;
10249 	default:
10250 		break;
10251 	}
10252 
10253 	return false;
10254 }
10255 
10256 /**
10257  * ixgbe_set_fw_version - Set FW version
10258  * @adapter: the adapter private structure
10259  *
10260  * This function is used by probe and ethtool to determine the FW version to
10261  * format to display. The FW version is taken from the EEPROM/NVM.
10262  */
10263 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
10264 {
10265 	struct ixgbe_hw *hw = &adapter->hw;
10266 	struct ixgbe_nvm_version nvm_ver;
10267 
10268 	ixgbe_get_oem_prod_version(hw, &nvm_ver);
10269 	if (nvm_ver.oem_valid) {
10270 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10271 			 "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
10272 			 nvm_ver.oem_release);
10273 		return;
10274 	}
10275 
10276 	ixgbe_get_etk_id(hw, &nvm_ver);
10277 	ixgbe_get_orom_version(hw, &nvm_ver);
10278 
10279 	if (nvm_ver.or_valid) {
10280 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10281 			 "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
10282 			 nvm_ver.or_build, nvm_ver.or_patch);
10283 		return;
10284 	}
10285 
10286 	/* Set ETrack ID format */
10287 	snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10288 		 "0x%08x", nvm_ver.etk_id);
10289 }
10290 
10291 /**
10292  * ixgbe_probe - Device Initialization Routine
10293  * @pdev: PCI device information struct
10294  * @ent: entry in ixgbe_pci_tbl
10295  *
10296  * Returns 0 on success, negative on failure
10297  *
10298  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10299  * The OS initialization, configuring of the adapter private structure,
10300  * and a hardware reset occur.
10301  **/
10302 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10303 {
10304 	struct net_device *netdev;
10305 	struct ixgbe_adapter *adapter = NULL;
10306 	struct ixgbe_hw *hw;
10307 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10308 	int i, err, pci_using_dac, expected_gts;
10309 	unsigned int indices = MAX_TX_QUEUES;
10310 	u8 part_str[IXGBE_PBANUM_LENGTH];
10311 	bool disable_dev = false;
10312 #ifdef IXGBE_FCOE
10313 	u16 device_caps;
10314 #endif
10315 	u32 eec;
10316 
10317 	/* Catch broken hardware that put the wrong VF device ID in
10318 	 * the PCIe SR-IOV capability.
10319 	 */
10320 	if (pdev->is_virtfn) {
10321 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10322 		     pci_name(pdev), pdev->vendor, pdev->device);
10323 		return -EINVAL;
10324 	}
10325 
10326 	err = pci_enable_device_mem(pdev);
10327 	if (err)
10328 		return err;
10329 
10330 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
10331 		pci_using_dac = 1;
10332 	} else {
10333 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10334 		if (err) {
10335 			dev_err(&pdev->dev,
10336 				"No usable DMA configuration, aborting\n");
10337 			goto err_dma;
10338 		}
10339 		pci_using_dac = 0;
10340 	}
10341 
10342 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10343 	if (err) {
10344 		dev_err(&pdev->dev,
10345 			"pci_request_selected_regions failed 0x%x\n", err);
10346 		goto err_pci_reg;
10347 	}
10348 
10349 	pci_enable_pcie_error_reporting(pdev);
10350 
10351 	pci_set_master(pdev);
10352 	pci_save_state(pdev);
10353 
10354 	if (ii->mac == ixgbe_mac_82598EB) {
10355 #ifdef CONFIG_IXGBE_DCB
10356 		/* 8 TC w/ 4 queues per TC */
10357 		indices = 4 * MAX_TRAFFIC_CLASS;
10358 #else
10359 		indices = IXGBE_MAX_RSS_INDICES;
10360 #endif
10361 	}
10362 
10363 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10364 	if (!netdev) {
10365 		err = -ENOMEM;
10366 		goto err_alloc_etherdev;
10367 	}
10368 
10369 	SET_NETDEV_DEV(netdev, &pdev->dev);
10370 
10371 	adapter = netdev_priv(netdev);
10372 
10373 	adapter->netdev = netdev;
10374 	adapter->pdev = pdev;
10375 	hw = &adapter->hw;
10376 	hw->back = adapter;
10377 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10378 
10379 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10380 			      pci_resource_len(pdev, 0));
10381 	adapter->io_addr = hw->hw_addr;
10382 	if (!hw->hw_addr) {
10383 		err = -EIO;
10384 		goto err_ioremap;
10385 	}
10386 
10387 	netdev->netdev_ops = &ixgbe_netdev_ops;
10388 	ixgbe_set_ethtool_ops(netdev);
10389 	netdev->watchdog_timeo = 5 * HZ;
10390 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10391 
10392 	/* Setup hw api */
10393 	hw->mac.ops   = *ii->mac_ops;
10394 	hw->mac.type  = ii->mac;
10395 	hw->mvals     = ii->mvals;
10396 	if (ii->link_ops)
10397 		hw->link.ops  = *ii->link_ops;
10398 
10399 	/* EEPROM */
10400 	hw->eeprom.ops = *ii->eeprom_ops;
10401 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10402 	if (ixgbe_removed(hw->hw_addr)) {
10403 		err = -EIO;
10404 		goto err_ioremap;
10405 	}
10406 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10407 	if (!(eec & BIT(8)))
10408 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10409 
10410 	/* PHY */
10411 	hw->phy.ops = *ii->phy_ops;
10412 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10413 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
10414 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10415 	hw->phy.mdio.mmds = 0;
10416 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10417 	hw->phy.mdio.dev = netdev;
10418 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10419 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10420 
10421 	/* setup the private structure */
10422 	err = ixgbe_sw_init(adapter, ii);
10423 	if (err)
10424 		goto err_sw_init;
10425 
10426 	/* Make sure the SWFW semaphore is in a valid state */
10427 	if (hw->mac.ops.init_swfw_sync)
10428 		hw->mac.ops.init_swfw_sync(hw);
10429 
10430 	/* Make it possible the adapter to be woken up via WOL */
10431 	switch (adapter->hw.mac.type) {
10432 	case ixgbe_mac_82599EB:
10433 	case ixgbe_mac_X540:
10434 	case ixgbe_mac_X550:
10435 	case ixgbe_mac_X550EM_x:
10436 	case ixgbe_mac_x550em_a:
10437 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10438 		break;
10439 	default:
10440 		break;
10441 	}
10442 
10443 	/*
10444 	 * If there is a fan on this device and it has failed log the
10445 	 * failure.
10446 	 */
10447 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10448 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10449 		if (esdp & IXGBE_ESDP_SDP1)
10450 			e_crit(probe, "Fan has stopped, replace the adapter\n");
10451 	}
10452 
10453 	if (allow_unsupported_sfp)
10454 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
10455 
10456 	/* reset_hw fills in the perm_addr as well */
10457 	hw->phy.reset_if_overtemp = true;
10458 	err = hw->mac.ops.reset_hw(hw);
10459 	hw->phy.reset_if_overtemp = false;
10460 	ixgbe_set_eee_capable(adapter);
10461 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
10462 		err = 0;
10463 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
10464 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10465 		e_dev_err("Reload the driver after installing a supported module.\n");
10466 		goto err_sw_init;
10467 	} else if (err) {
10468 		e_dev_err("HW Init failed: %d\n", err);
10469 		goto err_sw_init;
10470 	}
10471 
10472 #ifdef CONFIG_PCI_IOV
10473 	/* SR-IOV not supported on the 82598 */
10474 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
10475 		goto skip_sriov;
10476 	/* Mailbox */
10477 	ixgbe_init_mbx_params_pf(hw);
10478 	hw->mbx.ops = ii->mbx_ops;
10479 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
10480 	ixgbe_enable_sriov(adapter, max_vfs);
10481 skip_sriov:
10482 
10483 #endif
10484 	netdev->features = NETIF_F_SG |
10485 			   NETIF_F_TSO |
10486 			   NETIF_F_TSO6 |
10487 			   NETIF_F_RXHASH |
10488 			   NETIF_F_RXCSUM |
10489 			   NETIF_F_HW_CSUM;
10490 
10491 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
10492 				    NETIF_F_GSO_GRE_CSUM | \
10493 				    NETIF_F_GSO_IPXIP4 | \
10494 				    NETIF_F_GSO_IPXIP6 | \
10495 				    NETIF_F_GSO_UDP_TUNNEL | \
10496 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
10497 
10498 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
10499 	netdev->features |= NETIF_F_GSO_PARTIAL |
10500 			    IXGBE_GSO_PARTIAL_FEATURES;
10501 
10502 	if (hw->mac.type >= ixgbe_mac_82599EB)
10503 		netdev->features |= NETIF_F_SCTP_CRC;
10504 
10505 	/* copy netdev features into list of user selectable features */
10506 	netdev->hw_features |= netdev->features |
10507 			       NETIF_F_HW_VLAN_CTAG_FILTER |
10508 			       NETIF_F_HW_VLAN_CTAG_RX |
10509 			       NETIF_F_HW_VLAN_CTAG_TX |
10510 			       NETIF_F_RXALL |
10511 			       NETIF_F_HW_L2FW_DOFFLOAD;
10512 
10513 	if (hw->mac.type >= ixgbe_mac_82599EB)
10514 		netdev->hw_features |= NETIF_F_NTUPLE |
10515 				       NETIF_F_HW_TC;
10516 
10517 	if (pci_using_dac)
10518 		netdev->features |= NETIF_F_HIGHDMA;
10519 
10520 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
10521 	netdev->hw_enc_features |= netdev->vlan_features;
10522 	netdev->mpls_features |= NETIF_F_SG |
10523 				 NETIF_F_TSO |
10524 				 NETIF_F_TSO6 |
10525 				 NETIF_F_HW_CSUM;
10526 	netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
10527 
10528 	/* set this bit last since it cannot be part of vlan_features */
10529 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
10530 			    NETIF_F_HW_VLAN_CTAG_RX |
10531 			    NETIF_F_HW_VLAN_CTAG_TX;
10532 
10533 	netdev->priv_flags |= IFF_UNICAST_FLT;
10534 	netdev->priv_flags |= IFF_SUPP_NOFCS;
10535 
10536 	/* MTU range: 68 - 9710 */
10537 	netdev->min_mtu = ETH_MIN_MTU;
10538 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
10539 
10540 #ifdef CONFIG_IXGBE_DCB
10541 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
10542 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
10543 #endif
10544 
10545 #ifdef IXGBE_FCOE
10546 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
10547 		unsigned int fcoe_l;
10548 
10549 		if (hw->mac.ops.get_device_caps) {
10550 			hw->mac.ops.get_device_caps(hw, &device_caps);
10551 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
10552 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
10553 		}
10554 
10555 
10556 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
10557 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
10558 
10559 		netdev->features |= NETIF_F_FSO |
10560 				    NETIF_F_FCOE_CRC;
10561 
10562 		netdev->vlan_features |= NETIF_F_FSO |
10563 					 NETIF_F_FCOE_CRC |
10564 					 NETIF_F_FCOE_MTU;
10565 	}
10566 #endif /* IXGBE_FCOE */
10567 	ixgbe_init_ipsec_offload(adapter);
10568 
10569 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
10570 		netdev->hw_features |= NETIF_F_LRO;
10571 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
10572 		netdev->features |= NETIF_F_LRO;
10573 
10574 	/* make sure the EEPROM is good */
10575 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
10576 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
10577 		err = -EIO;
10578 		goto err_sw_init;
10579 	}
10580 
10581 	eth_platform_get_mac_address(&adapter->pdev->dev,
10582 				     adapter->hw.mac.perm_addr);
10583 
10584 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
10585 
10586 	if (!is_valid_ether_addr(netdev->dev_addr)) {
10587 		e_dev_err("invalid MAC address\n");
10588 		err = -EIO;
10589 		goto err_sw_init;
10590 	}
10591 
10592 	/* Set hw->mac.addr to permanent MAC address */
10593 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
10594 	ixgbe_mac_set_default_filter(adapter);
10595 
10596 	timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
10597 
10598 	if (ixgbe_removed(hw->hw_addr)) {
10599 		err = -EIO;
10600 		goto err_sw_init;
10601 	}
10602 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
10603 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
10604 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
10605 
10606 	err = ixgbe_init_interrupt_scheme(adapter);
10607 	if (err)
10608 		goto err_sw_init;
10609 
10610 	for (i = 0; i < adapter->num_rx_queues; i++)
10611 		u64_stats_init(&adapter->rx_ring[i]->syncp);
10612 	for (i = 0; i < adapter->num_tx_queues; i++)
10613 		u64_stats_init(&adapter->tx_ring[i]->syncp);
10614 	for (i = 0; i < adapter->num_xdp_queues; i++)
10615 		u64_stats_init(&adapter->xdp_ring[i]->syncp);
10616 
10617 	/* WOL not supported for all devices */
10618 	adapter->wol = 0;
10619 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
10620 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
10621 						pdev->subsystem_device);
10622 	if (hw->wol_enabled)
10623 		adapter->wol = IXGBE_WUFC_MAG;
10624 
10625 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
10626 
10627 	/* save off EEPROM version number */
10628 	ixgbe_set_fw_version(adapter);
10629 
10630 	/* pick up the PCI bus settings for reporting later */
10631 	if (ixgbe_pcie_from_parent(hw))
10632 		ixgbe_get_parent_bus_info(adapter);
10633 	else
10634 		 hw->mac.ops.get_bus_info(hw);
10635 
10636 	/* calculate the expected PCIe bandwidth required for optimal
10637 	 * performance. Note that some older parts will never have enough
10638 	 * bandwidth due to being older generation PCIe parts. We clamp these
10639 	 * parts to ensure no warning is displayed if it can't be fixed.
10640 	 */
10641 	switch (hw->mac.type) {
10642 	case ixgbe_mac_82598EB:
10643 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
10644 		break;
10645 	default:
10646 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
10647 		break;
10648 	}
10649 
10650 	/* don't check link if we failed to enumerate functions */
10651 	if (expected_gts > 0)
10652 		ixgbe_check_minimum_link(adapter, expected_gts);
10653 
10654 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
10655 	if (err)
10656 		strlcpy(part_str, "Unknown", sizeof(part_str));
10657 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
10658 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
10659 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
10660 			   part_str);
10661 	else
10662 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
10663 			   hw->mac.type, hw->phy.type, part_str);
10664 
10665 	e_dev_info("%pM\n", netdev->dev_addr);
10666 
10667 	/* reset the hardware with the new settings */
10668 	err = hw->mac.ops.start_hw(hw);
10669 	if (err == IXGBE_ERR_EEPROM_VERSION) {
10670 		/* We are running on a pre-production device, log a warning */
10671 		e_dev_warn("This device is a pre-production adapter/LOM. "
10672 			   "Please be aware there may be issues associated "
10673 			   "with your hardware.  If you are experiencing "
10674 			   "problems please contact your Intel or hardware "
10675 			   "representative who provided you with this "
10676 			   "hardware.\n");
10677 	}
10678 	strcpy(netdev->name, "eth%d");
10679 	pci_set_drvdata(pdev, adapter);
10680 	err = register_netdev(netdev);
10681 	if (err)
10682 		goto err_register;
10683 
10684 
10685 	/* power down the optics for 82599 SFP+ fiber */
10686 	if (hw->mac.ops.disable_tx_laser)
10687 		hw->mac.ops.disable_tx_laser(hw);
10688 
10689 	/* carrier off reporting is important to ethtool even BEFORE open */
10690 	netif_carrier_off(netdev);
10691 
10692 #ifdef CONFIG_IXGBE_DCA
10693 	if (dca_add_requester(&pdev->dev) == 0) {
10694 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
10695 		ixgbe_setup_dca(adapter);
10696 	}
10697 #endif
10698 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
10699 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
10700 		for (i = 0; i < adapter->num_vfs; i++)
10701 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
10702 	}
10703 
10704 	/* firmware requires driver version to be 0xFFFFFFFF
10705 	 * since os does not support feature
10706 	 */
10707 	if (hw->mac.ops.set_fw_drv_ver)
10708 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
10709 					   sizeof(ixgbe_driver_version) - 1,
10710 					   ixgbe_driver_version);
10711 
10712 	/* add san mac addr to netdev */
10713 	ixgbe_add_sanmac_netdev(netdev);
10714 
10715 	e_dev_info("%s\n", ixgbe_default_device_descr);
10716 
10717 #ifdef CONFIG_IXGBE_HWMON
10718 	if (ixgbe_sysfs_init(adapter))
10719 		e_err(probe, "failed to allocate sysfs resources\n");
10720 #endif /* CONFIG_IXGBE_HWMON */
10721 
10722 	ixgbe_dbg_adapter_init(adapter);
10723 
10724 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
10725 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
10726 		hw->mac.ops.setup_link(hw,
10727 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
10728 			true);
10729 
10730 	return 0;
10731 
10732 err_register:
10733 	ixgbe_release_hw_control(adapter);
10734 	ixgbe_clear_interrupt_scheme(adapter);
10735 err_sw_init:
10736 	ixgbe_disable_sriov(adapter);
10737 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
10738 	iounmap(adapter->io_addr);
10739 	kfree(adapter->jump_tables[0]);
10740 	kfree(adapter->mac_table);
10741 	kfree(adapter->rss_key);
10742 err_ioremap:
10743 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10744 	free_netdev(netdev);
10745 err_alloc_etherdev:
10746 	pci_release_mem_regions(pdev);
10747 err_pci_reg:
10748 err_dma:
10749 	if (!adapter || disable_dev)
10750 		pci_disable_device(pdev);
10751 	return err;
10752 }
10753 
10754 /**
10755  * ixgbe_remove - Device Removal Routine
10756  * @pdev: PCI device information struct
10757  *
10758  * ixgbe_remove is called by the PCI subsystem to alert the driver
10759  * that it should release a PCI device.  The could be caused by a
10760  * Hot-Plug event, or because the driver is going to be removed from
10761  * memory.
10762  **/
10763 static void ixgbe_remove(struct pci_dev *pdev)
10764 {
10765 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10766 	struct net_device *netdev;
10767 	bool disable_dev;
10768 	int i;
10769 
10770 	/* if !adapter then we already cleaned up in probe */
10771 	if (!adapter)
10772 		return;
10773 
10774 	netdev  = adapter->netdev;
10775 	ixgbe_dbg_adapter_exit(adapter);
10776 
10777 	set_bit(__IXGBE_REMOVING, &adapter->state);
10778 	cancel_work_sync(&adapter->service_task);
10779 
10780 
10781 #ifdef CONFIG_IXGBE_DCA
10782 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
10783 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
10784 		dca_remove_requester(&pdev->dev);
10785 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
10786 				IXGBE_DCA_CTRL_DCA_DISABLE);
10787 	}
10788 
10789 #endif
10790 #ifdef CONFIG_IXGBE_HWMON
10791 	ixgbe_sysfs_exit(adapter);
10792 #endif /* CONFIG_IXGBE_HWMON */
10793 
10794 	/* remove the added san mac */
10795 	ixgbe_del_sanmac_netdev(netdev);
10796 
10797 #ifdef CONFIG_PCI_IOV
10798 	ixgbe_disable_sriov(adapter);
10799 #endif
10800 	if (netdev->reg_state == NETREG_REGISTERED)
10801 		unregister_netdev(netdev);
10802 
10803 	ixgbe_stop_ipsec_offload(adapter);
10804 	ixgbe_clear_interrupt_scheme(adapter);
10805 
10806 	ixgbe_release_hw_control(adapter);
10807 
10808 #ifdef CONFIG_DCB
10809 	kfree(adapter->ixgbe_ieee_pfc);
10810 	kfree(adapter->ixgbe_ieee_ets);
10811 
10812 #endif
10813 	iounmap(adapter->io_addr);
10814 	pci_release_mem_regions(pdev);
10815 
10816 	e_dev_info("complete\n");
10817 
10818 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
10819 		if (adapter->jump_tables[i]) {
10820 			kfree(adapter->jump_tables[i]->input);
10821 			kfree(adapter->jump_tables[i]->mask);
10822 		}
10823 		kfree(adapter->jump_tables[i]);
10824 	}
10825 
10826 	kfree(adapter->mac_table);
10827 	kfree(adapter->rss_key);
10828 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10829 	free_netdev(netdev);
10830 
10831 	pci_disable_pcie_error_reporting(pdev);
10832 
10833 	if (disable_dev)
10834 		pci_disable_device(pdev);
10835 }
10836 
10837 /**
10838  * ixgbe_io_error_detected - called when PCI error is detected
10839  * @pdev: Pointer to PCI device
10840  * @state: The current pci connection state
10841  *
10842  * This function is called after a PCI bus error affecting
10843  * this device has been detected.
10844  */
10845 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
10846 						pci_channel_state_t state)
10847 {
10848 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10849 	struct net_device *netdev = adapter->netdev;
10850 
10851 #ifdef CONFIG_PCI_IOV
10852 	struct ixgbe_hw *hw = &adapter->hw;
10853 	struct pci_dev *bdev, *vfdev;
10854 	u32 dw0, dw1, dw2, dw3;
10855 	int vf, pos;
10856 	u16 req_id, pf_func;
10857 
10858 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
10859 	    adapter->num_vfs == 0)
10860 		goto skip_bad_vf_detection;
10861 
10862 	bdev = pdev->bus->self;
10863 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
10864 		bdev = bdev->bus->self;
10865 
10866 	if (!bdev)
10867 		goto skip_bad_vf_detection;
10868 
10869 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
10870 	if (!pos)
10871 		goto skip_bad_vf_detection;
10872 
10873 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
10874 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
10875 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
10876 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
10877 	if (ixgbe_removed(hw->hw_addr))
10878 		goto skip_bad_vf_detection;
10879 
10880 	req_id = dw1 >> 16;
10881 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
10882 	if (!(req_id & 0x0080))
10883 		goto skip_bad_vf_detection;
10884 
10885 	pf_func = req_id & 0x01;
10886 	if ((pf_func & 1) == (pdev->devfn & 1)) {
10887 		unsigned int device_id;
10888 
10889 		vf = (req_id & 0x7F) >> 1;
10890 		e_dev_err("VF %d has caused a PCIe error\n", vf);
10891 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
10892 				"%8.8x\tdw3: %8.8x\n",
10893 		dw0, dw1, dw2, dw3);
10894 		switch (adapter->hw.mac.type) {
10895 		case ixgbe_mac_82599EB:
10896 			device_id = IXGBE_82599_VF_DEVICE_ID;
10897 			break;
10898 		case ixgbe_mac_X540:
10899 			device_id = IXGBE_X540_VF_DEVICE_ID;
10900 			break;
10901 		case ixgbe_mac_X550:
10902 			device_id = IXGBE_DEV_ID_X550_VF;
10903 			break;
10904 		case ixgbe_mac_X550EM_x:
10905 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
10906 			break;
10907 		case ixgbe_mac_x550em_a:
10908 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
10909 			break;
10910 		default:
10911 			device_id = 0;
10912 			break;
10913 		}
10914 
10915 		/* Find the pci device of the offending VF */
10916 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
10917 		while (vfdev) {
10918 			if (vfdev->devfn == (req_id & 0xFF))
10919 				break;
10920 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
10921 					       device_id, vfdev);
10922 		}
10923 		/*
10924 		 * There's a slim chance the VF could have been hot plugged,
10925 		 * so if it is no longer present we don't need to issue the
10926 		 * VFLR.  Just clean up the AER in that case.
10927 		 */
10928 		if (vfdev) {
10929 			pcie_flr(vfdev);
10930 			/* Free device reference count */
10931 			pci_dev_put(vfdev);
10932 		}
10933 
10934 		pci_cleanup_aer_uncorrect_error_status(pdev);
10935 	}
10936 
10937 	/*
10938 	 * Even though the error may have occurred on the other port
10939 	 * we still need to increment the vf error reference count for
10940 	 * both ports because the I/O resume function will be called
10941 	 * for both of them.
10942 	 */
10943 	adapter->vferr_refcount++;
10944 
10945 	return PCI_ERS_RESULT_RECOVERED;
10946 
10947 skip_bad_vf_detection:
10948 #endif /* CONFIG_PCI_IOV */
10949 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
10950 		return PCI_ERS_RESULT_DISCONNECT;
10951 
10952 	if (!netif_device_present(netdev))
10953 		return PCI_ERS_RESULT_DISCONNECT;
10954 
10955 	rtnl_lock();
10956 	netif_device_detach(netdev);
10957 
10958 	if (state == pci_channel_io_perm_failure) {
10959 		rtnl_unlock();
10960 		return PCI_ERS_RESULT_DISCONNECT;
10961 	}
10962 
10963 	if (netif_running(netdev))
10964 		ixgbe_close_suspend(adapter);
10965 
10966 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
10967 		pci_disable_device(pdev);
10968 	rtnl_unlock();
10969 
10970 	/* Request a slot reset. */
10971 	return PCI_ERS_RESULT_NEED_RESET;
10972 }
10973 
10974 /**
10975  * ixgbe_io_slot_reset - called after the pci bus has been reset.
10976  * @pdev: Pointer to PCI device
10977  *
10978  * Restart the card from scratch, as if from a cold-boot.
10979  */
10980 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
10981 {
10982 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10983 	pci_ers_result_t result;
10984 	int err;
10985 
10986 	if (pci_enable_device_mem(pdev)) {
10987 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
10988 		result = PCI_ERS_RESULT_DISCONNECT;
10989 	} else {
10990 		smp_mb__before_atomic();
10991 		clear_bit(__IXGBE_DISABLED, &adapter->state);
10992 		adapter->hw.hw_addr = adapter->io_addr;
10993 		pci_set_master(pdev);
10994 		pci_restore_state(pdev);
10995 		pci_save_state(pdev);
10996 
10997 		pci_wake_from_d3(pdev, false);
10998 
10999 		ixgbe_reset(adapter);
11000 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
11001 		result = PCI_ERS_RESULT_RECOVERED;
11002 	}
11003 
11004 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
11005 	if (err) {
11006 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
11007 			  "failed 0x%0x\n", err);
11008 		/* non-fatal, continue */
11009 	}
11010 
11011 	return result;
11012 }
11013 
11014 /**
11015  * ixgbe_io_resume - called when traffic can start flowing again.
11016  * @pdev: Pointer to PCI device
11017  *
11018  * This callback is called when the error recovery driver tells us that
11019  * its OK to resume normal operation.
11020  */
11021 static void ixgbe_io_resume(struct pci_dev *pdev)
11022 {
11023 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11024 	struct net_device *netdev = adapter->netdev;
11025 
11026 #ifdef CONFIG_PCI_IOV
11027 	if (adapter->vferr_refcount) {
11028 		e_info(drv, "Resuming after VF err\n");
11029 		adapter->vferr_refcount--;
11030 		return;
11031 	}
11032 
11033 #endif
11034 	rtnl_lock();
11035 	if (netif_running(netdev))
11036 		ixgbe_open(netdev);
11037 
11038 	netif_device_attach(netdev);
11039 	rtnl_unlock();
11040 }
11041 
11042 static const struct pci_error_handlers ixgbe_err_handler = {
11043 	.error_detected = ixgbe_io_error_detected,
11044 	.slot_reset = ixgbe_io_slot_reset,
11045 	.resume = ixgbe_io_resume,
11046 };
11047 
11048 static struct pci_driver ixgbe_driver = {
11049 	.name     = ixgbe_driver_name,
11050 	.id_table = ixgbe_pci_tbl,
11051 	.probe    = ixgbe_probe,
11052 	.remove   = ixgbe_remove,
11053 #ifdef CONFIG_PM
11054 	.suspend  = ixgbe_suspend,
11055 	.resume   = ixgbe_resume,
11056 #endif
11057 	.shutdown = ixgbe_shutdown,
11058 	.sriov_configure = ixgbe_pci_sriov_configure,
11059 	.err_handler = &ixgbe_err_handler
11060 };
11061 
11062 /**
11063  * ixgbe_init_module - Driver Registration Routine
11064  *
11065  * ixgbe_init_module is the first routine called when the driver is
11066  * loaded. All it does is register with the PCI subsystem.
11067  **/
11068 static int __init ixgbe_init_module(void)
11069 {
11070 	int ret;
11071 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
11072 	pr_info("%s\n", ixgbe_copyright);
11073 
11074 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
11075 	if (!ixgbe_wq) {
11076 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
11077 		return -ENOMEM;
11078 	}
11079 
11080 	ixgbe_dbg_init();
11081 
11082 	ret = pci_register_driver(&ixgbe_driver);
11083 	if (ret) {
11084 		destroy_workqueue(ixgbe_wq);
11085 		ixgbe_dbg_exit();
11086 		return ret;
11087 	}
11088 
11089 #ifdef CONFIG_IXGBE_DCA
11090 	dca_register_notify(&dca_notifier);
11091 #endif
11092 
11093 	return 0;
11094 }
11095 
11096 module_init(ixgbe_init_module);
11097 
11098 /**
11099  * ixgbe_exit_module - Driver Exit Cleanup Routine
11100  *
11101  * ixgbe_exit_module is called just before the driver is removed
11102  * from memory.
11103  **/
11104 static void __exit ixgbe_exit_module(void)
11105 {
11106 #ifdef CONFIG_IXGBE_DCA
11107 	dca_unregister_notify(&dca_notifier);
11108 #endif
11109 	pci_unregister_driver(&ixgbe_driver);
11110 
11111 	ixgbe_dbg_exit();
11112 	if (ixgbe_wq) {
11113 		destroy_workqueue(ixgbe_wq);
11114 		ixgbe_wq = NULL;
11115 	}
11116 }
11117 
11118 #ifdef CONFIG_IXGBE_DCA
11119 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
11120 			    void *p)
11121 {
11122 	int ret_val;
11123 
11124 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
11125 					 __ixgbe_notify_dca);
11126 
11127 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
11128 }
11129 
11130 #endif /* CONFIG_IXGBE_DCA */
11131 
11132 module_exit(ixgbe_exit_module);
11133 
11134 /* ixgbe_main.c */
11135