xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision b76960c0f6b25d447a1493c4388defb9e8e76c63)
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 void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
357 {
358 	u32 value;
359 
360 	/* The following check not only optimizes a bit by not
361 	 * performing a read on the status register when the
362 	 * register just read was a status register read that
363 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
364 	 * potential recursion.
365 	 */
366 	if (reg == IXGBE_STATUS) {
367 		ixgbe_remove_adapter(hw);
368 		return;
369 	}
370 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
371 	if (value == IXGBE_FAILED_READ_REG)
372 		ixgbe_remove_adapter(hw);
373 }
374 
375 /**
376  * ixgbe_read_reg - Read from device register
377  * @hw: hw specific details
378  * @reg: offset of register to read
379  *
380  * Returns : value read or IXGBE_FAILED_READ_REG if removed
381  *
382  * This function is used to read device registers. It checks for device
383  * removal by confirming any read that returns all ones by checking the
384  * status register value for all ones. This function avoids reading from
385  * the hardware if a removal was previously detected in which case it
386  * returns IXGBE_FAILED_READ_REG (all ones).
387  */
388 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
389 {
390 	u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
391 	u32 value;
392 
393 	if (ixgbe_removed(reg_addr))
394 		return IXGBE_FAILED_READ_REG;
395 	if (unlikely(hw->phy.nw_mng_if_sel &
396 		     IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
397 		struct ixgbe_adapter *adapter;
398 		int i;
399 
400 		for (i = 0; i < 200; ++i) {
401 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
402 			if (likely(!value))
403 				goto writes_completed;
404 			if (value == IXGBE_FAILED_READ_REG) {
405 				ixgbe_remove_adapter(hw);
406 				return IXGBE_FAILED_READ_REG;
407 			}
408 			udelay(5);
409 		}
410 
411 		adapter = hw->back;
412 		e_warn(hw, "register writes incomplete %08x\n", value);
413 	}
414 
415 writes_completed:
416 	value = readl(reg_addr + reg);
417 	if (unlikely(value == IXGBE_FAILED_READ_REG))
418 		ixgbe_check_remove(hw, reg);
419 	return value;
420 }
421 
422 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
423 {
424 	u16 value;
425 
426 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
427 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
428 		ixgbe_remove_adapter(hw);
429 		return true;
430 	}
431 	return false;
432 }
433 
434 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
435 {
436 	struct ixgbe_adapter *adapter = hw->back;
437 	u16 value;
438 
439 	if (ixgbe_removed(hw->hw_addr))
440 		return IXGBE_FAILED_READ_CFG_WORD;
441 	pci_read_config_word(adapter->pdev, reg, &value);
442 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
443 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
444 		return IXGBE_FAILED_READ_CFG_WORD;
445 	return value;
446 }
447 
448 #ifdef CONFIG_PCI_IOV
449 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
450 {
451 	struct ixgbe_adapter *adapter = hw->back;
452 	u32 value;
453 
454 	if (ixgbe_removed(hw->hw_addr))
455 		return IXGBE_FAILED_READ_CFG_DWORD;
456 	pci_read_config_dword(adapter->pdev, reg, &value);
457 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
458 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
459 		return IXGBE_FAILED_READ_CFG_DWORD;
460 	return value;
461 }
462 #endif /* CONFIG_PCI_IOV */
463 
464 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
465 {
466 	struct ixgbe_adapter *adapter = hw->back;
467 
468 	if (ixgbe_removed(hw->hw_addr))
469 		return;
470 	pci_write_config_word(adapter->pdev, reg, value);
471 }
472 
473 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
474 {
475 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
476 
477 	/* flush memory to make sure state is correct before next watchdog */
478 	smp_mb__before_atomic();
479 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
480 }
481 
482 struct ixgbe_reg_info {
483 	u32 ofs;
484 	char *name;
485 };
486 
487 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
488 
489 	/* General Registers */
490 	{IXGBE_CTRL, "CTRL"},
491 	{IXGBE_STATUS, "STATUS"},
492 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
493 
494 	/* Interrupt Registers */
495 	{IXGBE_EICR, "EICR"},
496 
497 	/* RX Registers */
498 	{IXGBE_SRRCTL(0), "SRRCTL"},
499 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
500 	{IXGBE_RDLEN(0), "RDLEN"},
501 	{IXGBE_RDH(0), "RDH"},
502 	{IXGBE_RDT(0), "RDT"},
503 	{IXGBE_RXDCTL(0), "RXDCTL"},
504 	{IXGBE_RDBAL(0), "RDBAL"},
505 	{IXGBE_RDBAH(0), "RDBAH"},
506 
507 	/* TX Registers */
508 	{IXGBE_TDBAL(0), "TDBAL"},
509 	{IXGBE_TDBAH(0), "TDBAH"},
510 	{IXGBE_TDLEN(0), "TDLEN"},
511 	{IXGBE_TDH(0), "TDH"},
512 	{IXGBE_TDT(0), "TDT"},
513 	{IXGBE_TXDCTL(0), "TXDCTL"},
514 
515 	/* List Terminator */
516 	{ .name = NULL }
517 };
518 
519 
520 /*
521  * ixgbe_regdump - register printout routine
522  */
523 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
524 {
525 	int i;
526 	char rname[16];
527 	u32 regs[64];
528 
529 	switch (reginfo->ofs) {
530 	case IXGBE_SRRCTL(0):
531 		for (i = 0; i < 64; i++)
532 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
533 		break;
534 	case IXGBE_DCA_RXCTRL(0):
535 		for (i = 0; i < 64; i++)
536 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
537 		break;
538 	case IXGBE_RDLEN(0):
539 		for (i = 0; i < 64; i++)
540 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
541 		break;
542 	case IXGBE_RDH(0):
543 		for (i = 0; i < 64; i++)
544 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
545 		break;
546 	case IXGBE_RDT(0):
547 		for (i = 0; i < 64; i++)
548 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
549 		break;
550 	case IXGBE_RXDCTL(0):
551 		for (i = 0; i < 64; i++)
552 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
553 		break;
554 	case IXGBE_RDBAL(0):
555 		for (i = 0; i < 64; i++)
556 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
557 		break;
558 	case IXGBE_RDBAH(0):
559 		for (i = 0; i < 64; i++)
560 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
561 		break;
562 	case IXGBE_TDBAL(0):
563 		for (i = 0; i < 64; i++)
564 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
565 		break;
566 	case IXGBE_TDBAH(0):
567 		for (i = 0; i < 64; i++)
568 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
569 		break;
570 	case IXGBE_TDLEN(0):
571 		for (i = 0; i < 64; i++)
572 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
573 		break;
574 	case IXGBE_TDH(0):
575 		for (i = 0; i < 64; i++)
576 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
577 		break;
578 	case IXGBE_TDT(0):
579 		for (i = 0; i < 64; i++)
580 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
581 		break;
582 	case IXGBE_TXDCTL(0):
583 		for (i = 0; i < 64; i++)
584 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
585 		break;
586 	default:
587 		pr_info("%-15s %08x\n",
588 			reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
589 		return;
590 	}
591 
592 	i = 0;
593 	while (i < 64) {
594 		int j;
595 		char buf[9 * 8 + 1];
596 		char *p = buf;
597 
598 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
599 		for (j = 0; j < 8; j++)
600 			p += sprintf(p, " %08x", regs[i++]);
601 		pr_err("%-15s%s\n", rname, buf);
602 	}
603 
604 }
605 
606 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
607 {
608 	struct ixgbe_tx_buffer *tx_buffer;
609 
610 	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
611 	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
612 		n, ring->next_to_use, ring->next_to_clean,
613 		(u64)dma_unmap_addr(tx_buffer, dma),
614 		dma_unmap_len(tx_buffer, len),
615 		tx_buffer->next_to_watch,
616 		(u64)tx_buffer->time_stamp);
617 }
618 
619 /*
620  * ixgbe_dump - Print registers, tx-rings and rx-rings
621  */
622 static void ixgbe_dump(struct ixgbe_adapter *adapter)
623 {
624 	struct net_device *netdev = adapter->netdev;
625 	struct ixgbe_hw *hw = &adapter->hw;
626 	struct ixgbe_reg_info *reginfo;
627 	int n = 0;
628 	struct ixgbe_ring *ring;
629 	struct ixgbe_tx_buffer *tx_buffer;
630 	union ixgbe_adv_tx_desc *tx_desc;
631 	struct my_u0 { u64 a; u64 b; } *u0;
632 	struct ixgbe_ring *rx_ring;
633 	union ixgbe_adv_rx_desc *rx_desc;
634 	struct ixgbe_rx_buffer *rx_buffer_info;
635 	int i = 0;
636 
637 	if (!netif_msg_hw(adapter))
638 		return;
639 
640 	/* Print netdevice Info */
641 	if (netdev) {
642 		dev_info(&adapter->pdev->dev, "Net device Info\n");
643 		pr_info("Device Name     state            "
644 			"trans_start\n");
645 		pr_info("%-15s %016lX %016lX\n",
646 			netdev->name,
647 			netdev->state,
648 			dev_trans_start(netdev));
649 	}
650 
651 	/* Print Registers */
652 	dev_info(&adapter->pdev->dev, "Register Dump\n");
653 	pr_info(" Register Name   Value\n");
654 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
655 	     reginfo->name; reginfo++) {
656 		ixgbe_regdump(hw, reginfo);
657 	}
658 
659 	/* Print TX Ring Summary */
660 	if (!netdev || !netif_running(netdev))
661 		return;
662 
663 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
664 	pr_info(" %s     %s              %s        %s\n",
665 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
666 		"leng", "ntw", "timestamp");
667 	for (n = 0; n < adapter->num_tx_queues; n++) {
668 		ring = adapter->tx_ring[n];
669 		ixgbe_print_buffer(ring, n);
670 	}
671 
672 	for (n = 0; n < adapter->num_xdp_queues; n++) {
673 		ring = adapter->xdp_ring[n];
674 		ixgbe_print_buffer(ring, n);
675 	}
676 
677 	/* Print TX Rings */
678 	if (!netif_msg_tx_done(adapter))
679 		goto rx_ring_summary;
680 
681 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
682 
683 	/* Transmit Descriptor Formats
684 	 *
685 	 * 82598 Advanced Transmit Descriptor
686 	 *   +--------------------------------------------------------------+
687 	 * 0 |         Buffer Address [63:0]                                |
688 	 *   +--------------------------------------------------------------+
689 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
690 	 *   +--------------------------------------------------------------+
691 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
692 	 *
693 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
694 	 *   +--------------------------------------------------------------+
695 	 * 0 |                          RSV [63:0]                          |
696 	 *   +--------------------------------------------------------------+
697 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
698 	 *   +--------------------------------------------------------------+
699 	 *   63                       36 35   32 31                         0
700 	 *
701 	 * 82599+ Advanced Transmit Descriptor
702 	 *   +--------------------------------------------------------------+
703 	 * 0 |         Buffer Address [63:0]                                |
704 	 *   +--------------------------------------------------------------+
705 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
706 	 *   +--------------------------------------------------------------+
707 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
708 	 *
709 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
710 	 *   +--------------------------------------------------------------+
711 	 * 0 |                          RSV [63:0]                          |
712 	 *   +--------------------------------------------------------------+
713 	 * 8 |            RSV           |  STA  |           RSV             |
714 	 *   +--------------------------------------------------------------+
715 	 *   63                       36 35   32 31                         0
716 	 */
717 
718 	for (n = 0; n < adapter->num_tx_queues; n++) {
719 		ring = adapter->tx_ring[n];
720 		pr_info("------------------------------------\n");
721 		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
722 		pr_info("------------------------------------\n");
723 		pr_info("%s%s    %s              %s        %s          %s\n",
724 			"T [desc]     [address 63:0  ] ",
725 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
726 			"leng", "ntw", "timestamp", "bi->skb");
727 
728 		for (i = 0; ring->desc && (i < ring->count); i++) {
729 			tx_desc = IXGBE_TX_DESC(ring, i);
730 			tx_buffer = &ring->tx_buffer_info[i];
731 			u0 = (struct my_u0 *)tx_desc;
732 			if (dma_unmap_len(tx_buffer, len) > 0) {
733 				const char *ring_desc;
734 
735 				if (i == ring->next_to_use &&
736 				    i == ring->next_to_clean)
737 					ring_desc = " NTC/U";
738 				else if (i == ring->next_to_use)
739 					ring_desc = " NTU";
740 				else if (i == ring->next_to_clean)
741 					ring_desc = " NTC";
742 				else
743 					ring_desc = "";
744 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
745 					i,
746 					le64_to_cpu(u0->a),
747 					le64_to_cpu(u0->b),
748 					(u64)dma_unmap_addr(tx_buffer, dma),
749 					dma_unmap_len(tx_buffer, len),
750 					tx_buffer->next_to_watch,
751 					(u64)tx_buffer->time_stamp,
752 					tx_buffer->skb,
753 					ring_desc);
754 
755 				if (netif_msg_pktdata(adapter) &&
756 				    tx_buffer->skb)
757 					print_hex_dump(KERN_INFO, "",
758 						DUMP_PREFIX_ADDRESS, 16, 1,
759 						tx_buffer->skb->data,
760 						dma_unmap_len(tx_buffer, len),
761 						true);
762 			}
763 		}
764 	}
765 
766 	/* Print RX Rings Summary */
767 rx_ring_summary:
768 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
769 	pr_info("Queue [NTU] [NTC]\n");
770 	for (n = 0; n < adapter->num_rx_queues; n++) {
771 		rx_ring = adapter->rx_ring[n];
772 		pr_info("%5d %5X %5X\n",
773 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
774 	}
775 
776 	/* Print RX Rings */
777 	if (!netif_msg_rx_status(adapter))
778 		return;
779 
780 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
781 
782 	/* Receive Descriptor Formats
783 	 *
784 	 * 82598 Advanced Receive Descriptor (Read) Format
785 	 *    63                                           1        0
786 	 *    +-----------------------------------------------------+
787 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
788 	 *    +----------------------------------------------+------+
789 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
790 	 *    +-----------------------------------------------------+
791 	 *
792 	 *
793 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
794 	 *
795 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
796 	 *   +------------------------------------------------------+
797 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
798 	 *   | Packet   | IP     |   |          |     | Type | Type |
799 	 *   | Checksum | Ident  |   |          |     |      |      |
800 	 *   +------------------------------------------------------+
801 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
802 	 *   +------------------------------------------------------+
803 	 *   63       48 47    32 31            20 19               0
804 	 *
805 	 * 82599+ Advanced Receive Descriptor (Read) Format
806 	 *    63                                           1        0
807 	 *    +-----------------------------------------------------+
808 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
809 	 *    +----------------------------------------------+------+
810 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
811 	 *    +-----------------------------------------------------+
812 	 *
813 	 *
814 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
815 	 *
816 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
817 	 *   +------------------------------------------------------+
818 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
819 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
820 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
821 	 *   +------------------------------------------------------+
822 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
823 	 *   +------------------------------------------------------+
824 	 *   63       48 47    32 31          20 19                 0
825 	 */
826 
827 	for (n = 0; n < adapter->num_rx_queues; n++) {
828 		rx_ring = adapter->rx_ring[n];
829 		pr_info("------------------------------------\n");
830 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
831 		pr_info("------------------------------------\n");
832 		pr_info("%s%s%s\n",
833 			"R  [desc]      [ PktBuf     A0] ",
834 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
835 			"<-- Adv Rx Read format");
836 		pr_info("%s%s%s\n",
837 			"RWB[desc]      [PcsmIpSHl PtRs] ",
838 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
839 			"<-- Adv Rx Write-Back format");
840 
841 		for (i = 0; i < rx_ring->count; i++) {
842 			const char *ring_desc;
843 
844 			if (i == rx_ring->next_to_use)
845 				ring_desc = " NTU";
846 			else if (i == rx_ring->next_to_clean)
847 				ring_desc = " NTC";
848 			else
849 				ring_desc = "";
850 
851 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
852 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
853 			u0 = (struct my_u0 *)rx_desc;
854 			if (rx_desc->wb.upper.length) {
855 				/* Descriptor Done */
856 				pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
857 					i,
858 					le64_to_cpu(u0->a),
859 					le64_to_cpu(u0->b),
860 					rx_buffer_info->skb,
861 					ring_desc);
862 			} else {
863 				pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
864 					i,
865 					le64_to_cpu(u0->a),
866 					le64_to_cpu(u0->b),
867 					(u64)rx_buffer_info->dma,
868 					rx_buffer_info->skb,
869 					ring_desc);
870 
871 				if (netif_msg_pktdata(adapter) &&
872 				    rx_buffer_info->dma) {
873 					print_hex_dump(KERN_INFO, "",
874 					   DUMP_PREFIX_ADDRESS, 16, 1,
875 					   page_address(rx_buffer_info->page) +
876 						    rx_buffer_info->page_offset,
877 					   ixgbe_rx_bufsz(rx_ring), true);
878 				}
879 			}
880 		}
881 	}
882 }
883 
884 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
885 {
886 	u32 ctrl_ext;
887 
888 	/* Let firmware take over control of h/w */
889 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
890 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
891 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
892 }
893 
894 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
895 {
896 	u32 ctrl_ext;
897 
898 	/* Let firmware know the driver has taken over */
899 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
900 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
901 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
902 }
903 
904 /**
905  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
906  * @adapter: pointer to adapter struct
907  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
908  * @queue: queue to map the corresponding interrupt to
909  * @msix_vector: the vector to map to the corresponding queue
910  *
911  */
912 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
913 			   u8 queue, u8 msix_vector)
914 {
915 	u32 ivar, index;
916 	struct ixgbe_hw *hw = &adapter->hw;
917 	switch (hw->mac.type) {
918 	case ixgbe_mac_82598EB:
919 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
920 		if (direction == -1)
921 			direction = 0;
922 		index = (((direction * 64) + queue) >> 2) & 0x1F;
923 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
924 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
925 		ivar |= (msix_vector << (8 * (queue & 0x3)));
926 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
927 		break;
928 	case ixgbe_mac_82599EB:
929 	case ixgbe_mac_X540:
930 	case ixgbe_mac_X550:
931 	case ixgbe_mac_X550EM_x:
932 	case ixgbe_mac_x550em_a:
933 		if (direction == -1) {
934 			/* other causes */
935 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
936 			index = ((queue & 1) * 8);
937 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
938 			ivar &= ~(0xFF << index);
939 			ivar |= (msix_vector << index);
940 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
941 			break;
942 		} else {
943 			/* tx or rx causes */
944 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
945 			index = ((16 * (queue & 1)) + (8 * direction));
946 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
947 			ivar &= ~(0xFF << index);
948 			ivar |= (msix_vector << index);
949 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
950 			break;
951 		}
952 	default:
953 		break;
954 	}
955 }
956 
957 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
958 					  u64 qmask)
959 {
960 	u32 mask;
961 
962 	switch (adapter->hw.mac.type) {
963 	case ixgbe_mac_82598EB:
964 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
965 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
966 		break;
967 	case ixgbe_mac_82599EB:
968 	case ixgbe_mac_X540:
969 	case ixgbe_mac_X550:
970 	case ixgbe_mac_X550EM_x:
971 	case ixgbe_mac_x550em_a:
972 		mask = (qmask & 0xFFFFFFFF);
973 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
974 		mask = (qmask >> 32);
975 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
976 		break;
977 	default:
978 		break;
979 	}
980 }
981 
982 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
983 {
984 	struct ixgbe_hw *hw = &adapter->hw;
985 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
986 	int i;
987 	u32 data;
988 
989 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
990 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
991 		return;
992 
993 	switch (hw->mac.type) {
994 	case ixgbe_mac_82598EB:
995 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
996 		break;
997 	default:
998 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
999 	}
1000 	hwstats->lxoffrxc += data;
1001 
1002 	/* refill credits (no tx hang) if we received xoff */
1003 	if (!data)
1004 		return;
1005 
1006 	for (i = 0; i < adapter->num_tx_queues; i++)
1007 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
1008 			  &adapter->tx_ring[i]->state);
1009 
1010 	for (i = 0; i < adapter->num_xdp_queues; i++)
1011 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
1012 			  &adapter->xdp_ring[i]->state);
1013 }
1014 
1015 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
1016 {
1017 	struct ixgbe_hw *hw = &adapter->hw;
1018 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
1019 	u32 xoff[8] = {0};
1020 	u8 tc;
1021 	int i;
1022 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
1023 
1024 	if (adapter->ixgbe_ieee_pfc)
1025 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
1026 
1027 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
1028 		ixgbe_update_xoff_rx_lfc(adapter);
1029 		return;
1030 	}
1031 
1032 	/* update stats for each tc, only valid with PFC enabled */
1033 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
1034 		u32 pxoffrxc;
1035 
1036 		switch (hw->mac.type) {
1037 		case ixgbe_mac_82598EB:
1038 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1039 			break;
1040 		default:
1041 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1042 		}
1043 		hwstats->pxoffrxc[i] += pxoffrxc;
1044 		/* Get the TC for given UP */
1045 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1046 		xoff[tc] += pxoffrxc;
1047 	}
1048 
1049 	/* disarm tx queues that have received xoff frames */
1050 	for (i = 0; i < adapter->num_tx_queues; i++) {
1051 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1052 
1053 		tc = tx_ring->dcb_tc;
1054 		if (xoff[tc])
1055 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1056 	}
1057 
1058 	for (i = 0; i < adapter->num_xdp_queues; i++) {
1059 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1060 
1061 		tc = xdp_ring->dcb_tc;
1062 		if (xoff[tc])
1063 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1064 	}
1065 }
1066 
1067 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1068 {
1069 	return ring->stats.packets;
1070 }
1071 
1072 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1073 {
1074 	unsigned int head, tail;
1075 
1076 	head = ring->next_to_clean;
1077 	tail = ring->next_to_use;
1078 
1079 	return ((head <= tail) ? tail : tail + ring->count) - head;
1080 }
1081 
1082 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1083 {
1084 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1085 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1086 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1087 
1088 	clear_check_for_tx_hang(tx_ring);
1089 
1090 	/*
1091 	 * Check for a hung queue, but be thorough. This verifies
1092 	 * that a transmit has been completed since the previous
1093 	 * check AND there is at least one packet pending. The
1094 	 * ARMED bit is set to indicate a potential hang. The
1095 	 * bit is cleared if a pause frame is received to remove
1096 	 * false hang detection due to PFC or 802.3x frames. By
1097 	 * requiring this to fail twice we avoid races with
1098 	 * pfc clearing the ARMED bit and conditions where we
1099 	 * run the check_tx_hang logic with a transmit completion
1100 	 * pending but without time to complete it yet.
1101 	 */
1102 	if (tx_done_old == tx_done && tx_pending)
1103 		/* make sure it is true for two checks in a row */
1104 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1105 					&tx_ring->state);
1106 	/* update completed stats and continue */
1107 	tx_ring->tx_stats.tx_done_old = tx_done;
1108 	/* reset the countdown */
1109 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1110 
1111 	return false;
1112 }
1113 
1114 /**
1115  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1116  * @adapter: driver private struct
1117  **/
1118 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1119 {
1120 
1121 	/* Do the reset outside of interrupt context */
1122 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1123 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1124 		e_warn(drv, "initiating reset due to tx timeout\n");
1125 		ixgbe_service_event_schedule(adapter);
1126 	}
1127 }
1128 
1129 /**
1130  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1131  * @netdev: network interface device structure
1132  * @queue_index: Tx queue to set
1133  * @maxrate: desired maximum transmit bitrate
1134  **/
1135 static int ixgbe_tx_maxrate(struct net_device *netdev,
1136 			    int queue_index, u32 maxrate)
1137 {
1138 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1139 	struct ixgbe_hw *hw = &adapter->hw;
1140 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1141 
1142 	if (!maxrate)
1143 		return 0;
1144 
1145 	/* Calculate the rate factor values to set */
1146 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1147 	bcnrc_val /= maxrate;
1148 
1149 	/* clear everything but the rate factor */
1150 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1151 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1152 
1153 	/* enable the rate scheduler */
1154 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1155 
1156 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1157 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1158 
1159 	return 0;
1160 }
1161 
1162 /**
1163  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1164  * @q_vector: structure containing interrupt and ring information
1165  * @tx_ring: tx ring to clean
1166  * @napi_budget: Used to determine if we are in netpoll
1167  **/
1168 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1169 			       struct ixgbe_ring *tx_ring, int napi_budget)
1170 {
1171 	struct ixgbe_adapter *adapter = q_vector->adapter;
1172 	struct ixgbe_tx_buffer *tx_buffer;
1173 	union ixgbe_adv_tx_desc *tx_desc;
1174 	unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1175 	unsigned int budget = q_vector->tx.work_limit;
1176 	unsigned int i = tx_ring->next_to_clean;
1177 
1178 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1179 		return true;
1180 
1181 	tx_buffer = &tx_ring->tx_buffer_info[i];
1182 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1183 	i -= tx_ring->count;
1184 
1185 	do {
1186 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1187 
1188 		/* if next_to_watch is not set then there is no work pending */
1189 		if (!eop_desc)
1190 			break;
1191 
1192 		/* prevent any other reads prior to eop_desc */
1193 		smp_rmb();
1194 
1195 		/* if DD is not set pending work has not been completed */
1196 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1197 			break;
1198 
1199 		/* clear next_to_watch to prevent false hangs */
1200 		tx_buffer->next_to_watch = NULL;
1201 
1202 		/* update the statistics for this packet */
1203 		total_bytes += tx_buffer->bytecount;
1204 		total_packets += tx_buffer->gso_segs;
1205 		if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1206 			total_ipsec++;
1207 
1208 		/* free the skb */
1209 		if (ring_is_xdp(tx_ring))
1210 			page_frag_free(tx_buffer->data);
1211 		else
1212 			napi_consume_skb(tx_buffer->skb, napi_budget);
1213 
1214 		/* unmap skb header data */
1215 		dma_unmap_single(tx_ring->dev,
1216 				 dma_unmap_addr(tx_buffer, dma),
1217 				 dma_unmap_len(tx_buffer, len),
1218 				 DMA_TO_DEVICE);
1219 
1220 		/* clear tx_buffer data */
1221 		dma_unmap_len_set(tx_buffer, len, 0);
1222 
1223 		/* unmap remaining buffers */
1224 		while (tx_desc != eop_desc) {
1225 			tx_buffer++;
1226 			tx_desc++;
1227 			i++;
1228 			if (unlikely(!i)) {
1229 				i -= tx_ring->count;
1230 				tx_buffer = tx_ring->tx_buffer_info;
1231 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1232 			}
1233 
1234 			/* unmap any remaining paged data */
1235 			if (dma_unmap_len(tx_buffer, len)) {
1236 				dma_unmap_page(tx_ring->dev,
1237 					       dma_unmap_addr(tx_buffer, dma),
1238 					       dma_unmap_len(tx_buffer, len),
1239 					       DMA_TO_DEVICE);
1240 				dma_unmap_len_set(tx_buffer, len, 0);
1241 			}
1242 		}
1243 
1244 		/* move us one more past the eop_desc for start of next pkt */
1245 		tx_buffer++;
1246 		tx_desc++;
1247 		i++;
1248 		if (unlikely(!i)) {
1249 			i -= tx_ring->count;
1250 			tx_buffer = tx_ring->tx_buffer_info;
1251 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1252 		}
1253 
1254 		/* issue prefetch for next Tx descriptor */
1255 		prefetch(tx_desc);
1256 
1257 		/* update budget accounting */
1258 		budget--;
1259 	} while (likely(budget));
1260 
1261 	i += tx_ring->count;
1262 	tx_ring->next_to_clean = i;
1263 	u64_stats_update_begin(&tx_ring->syncp);
1264 	tx_ring->stats.bytes += total_bytes;
1265 	tx_ring->stats.packets += total_packets;
1266 	u64_stats_update_end(&tx_ring->syncp);
1267 	q_vector->tx.total_bytes += total_bytes;
1268 	q_vector->tx.total_packets += total_packets;
1269 	adapter->tx_ipsec += total_ipsec;
1270 
1271 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1272 		/* schedule immediate reset if we believe we hung */
1273 		struct ixgbe_hw *hw = &adapter->hw;
1274 		e_err(drv, "Detected Tx Unit Hang %s\n"
1275 			"  Tx Queue             <%d>\n"
1276 			"  TDH, TDT             <%x>, <%x>\n"
1277 			"  next_to_use          <%x>\n"
1278 			"  next_to_clean        <%x>\n"
1279 			"tx_buffer_info[next_to_clean]\n"
1280 			"  time_stamp           <%lx>\n"
1281 			"  jiffies              <%lx>\n",
1282 			ring_is_xdp(tx_ring) ? "(XDP)" : "",
1283 			tx_ring->queue_index,
1284 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1285 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1286 			tx_ring->next_to_use, i,
1287 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1288 
1289 		if (!ring_is_xdp(tx_ring))
1290 			netif_stop_subqueue(tx_ring->netdev,
1291 					    tx_ring->queue_index);
1292 
1293 		e_info(probe,
1294 		       "tx hang %d detected on queue %d, resetting adapter\n",
1295 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1296 
1297 		/* schedule immediate reset if we believe we hung */
1298 		ixgbe_tx_timeout_reset(adapter);
1299 
1300 		/* the adapter is about to reset, no point in enabling stuff */
1301 		return true;
1302 	}
1303 
1304 	if (ring_is_xdp(tx_ring))
1305 		return !!budget;
1306 
1307 	netdev_tx_completed_queue(txring_txq(tx_ring),
1308 				  total_packets, total_bytes);
1309 
1310 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1311 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1312 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1313 		/* Make sure that anybody stopping the queue after this
1314 		 * sees the new next_to_clean.
1315 		 */
1316 		smp_mb();
1317 		if (__netif_subqueue_stopped(tx_ring->netdev,
1318 					     tx_ring->queue_index)
1319 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1320 			netif_wake_subqueue(tx_ring->netdev,
1321 					    tx_ring->queue_index);
1322 			++tx_ring->tx_stats.restart_queue;
1323 		}
1324 	}
1325 
1326 	return !!budget;
1327 }
1328 
1329 #ifdef CONFIG_IXGBE_DCA
1330 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1331 				struct ixgbe_ring *tx_ring,
1332 				int cpu)
1333 {
1334 	struct ixgbe_hw *hw = &adapter->hw;
1335 	u32 txctrl = 0;
1336 	u16 reg_offset;
1337 
1338 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1339 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1340 
1341 	switch (hw->mac.type) {
1342 	case ixgbe_mac_82598EB:
1343 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1344 		break;
1345 	case ixgbe_mac_82599EB:
1346 	case ixgbe_mac_X540:
1347 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1348 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1349 		break;
1350 	default:
1351 		/* for unknown hardware do not write register */
1352 		return;
1353 	}
1354 
1355 	/*
1356 	 * We can enable relaxed ordering for reads, but not writes when
1357 	 * DCA is enabled.  This is due to a known issue in some chipsets
1358 	 * which will cause the DCA tag to be cleared.
1359 	 */
1360 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1361 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1362 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1363 
1364 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1365 }
1366 
1367 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1368 				struct ixgbe_ring *rx_ring,
1369 				int cpu)
1370 {
1371 	struct ixgbe_hw *hw = &adapter->hw;
1372 	u32 rxctrl = 0;
1373 	u8 reg_idx = rx_ring->reg_idx;
1374 
1375 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1376 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1377 
1378 	switch (hw->mac.type) {
1379 	case ixgbe_mac_82599EB:
1380 	case ixgbe_mac_X540:
1381 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1382 		break;
1383 	default:
1384 		break;
1385 	}
1386 
1387 	/*
1388 	 * We can enable relaxed ordering for reads, but not writes when
1389 	 * DCA is enabled.  This is due to a known issue in some chipsets
1390 	 * which will cause the DCA tag to be cleared.
1391 	 */
1392 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1393 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1394 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1395 
1396 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1397 }
1398 
1399 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1400 {
1401 	struct ixgbe_adapter *adapter = q_vector->adapter;
1402 	struct ixgbe_ring *ring;
1403 	int cpu = get_cpu();
1404 
1405 	if (q_vector->cpu == cpu)
1406 		goto out_no_update;
1407 
1408 	ixgbe_for_each_ring(ring, q_vector->tx)
1409 		ixgbe_update_tx_dca(adapter, ring, cpu);
1410 
1411 	ixgbe_for_each_ring(ring, q_vector->rx)
1412 		ixgbe_update_rx_dca(adapter, ring, cpu);
1413 
1414 	q_vector->cpu = cpu;
1415 out_no_update:
1416 	put_cpu();
1417 }
1418 
1419 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1420 {
1421 	int i;
1422 
1423 	/* always use CB2 mode, difference is masked in the CB driver */
1424 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1425 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1426 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1427 	else
1428 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1429 				IXGBE_DCA_CTRL_DCA_DISABLE);
1430 
1431 	for (i = 0; i < adapter->num_q_vectors; i++) {
1432 		adapter->q_vector[i]->cpu = -1;
1433 		ixgbe_update_dca(adapter->q_vector[i]);
1434 	}
1435 }
1436 
1437 static int __ixgbe_notify_dca(struct device *dev, void *data)
1438 {
1439 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1440 	unsigned long event = *(unsigned long *)data;
1441 
1442 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1443 		return 0;
1444 
1445 	switch (event) {
1446 	case DCA_PROVIDER_ADD:
1447 		/* if we're already enabled, don't do it again */
1448 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1449 			break;
1450 		if (dca_add_requester(dev) == 0) {
1451 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1452 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1453 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1454 			break;
1455 		}
1456 		/* fall through - DCA is disabled. */
1457 	case DCA_PROVIDER_REMOVE:
1458 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1459 			dca_remove_requester(dev);
1460 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1461 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1462 					IXGBE_DCA_CTRL_DCA_DISABLE);
1463 		}
1464 		break;
1465 	}
1466 
1467 	return 0;
1468 }
1469 
1470 #endif /* CONFIG_IXGBE_DCA */
1471 
1472 #define IXGBE_RSS_L4_TYPES_MASK \
1473 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1474 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1475 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1476 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1477 
1478 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1479 				 union ixgbe_adv_rx_desc *rx_desc,
1480 				 struct sk_buff *skb)
1481 {
1482 	u16 rss_type;
1483 
1484 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1485 		return;
1486 
1487 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1488 		   IXGBE_RXDADV_RSSTYPE_MASK;
1489 
1490 	if (!rss_type)
1491 		return;
1492 
1493 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1494 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1495 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1496 }
1497 
1498 #ifdef IXGBE_FCOE
1499 /**
1500  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1501  * @ring: structure containing ring specific data
1502  * @rx_desc: advanced rx descriptor
1503  *
1504  * Returns : true if it is FCoE pkt
1505  */
1506 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1507 				    union ixgbe_adv_rx_desc *rx_desc)
1508 {
1509 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1510 
1511 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1512 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1513 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1514 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1515 }
1516 
1517 #endif /* IXGBE_FCOE */
1518 /**
1519  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1520  * @ring: structure containing ring specific data
1521  * @rx_desc: current Rx descriptor being processed
1522  * @skb: skb currently being received and modified
1523  **/
1524 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1525 				     union ixgbe_adv_rx_desc *rx_desc,
1526 				     struct sk_buff *skb)
1527 {
1528 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1529 	bool encap_pkt = false;
1530 
1531 	skb_checksum_none_assert(skb);
1532 
1533 	/* Rx csum disabled */
1534 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1535 		return;
1536 
1537 	/* check for VXLAN and Geneve packets */
1538 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1539 		encap_pkt = true;
1540 		skb->encapsulation = 1;
1541 	}
1542 
1543 	/* if IP and error */
1544 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1545 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1546 		ring->rx_stats.csum_err++;
1547 		return;
1548 	}
1549 
1550 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1551 		return;
1552 
1553 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1554 		/*
1555 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1556 		 * checksum errors.
1557 		 */
1558 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1559 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1560 			return;
1561 
1562 		ring->rx_stats.csum_err++;
1563 		return;
1564 	}
1565 
1566 	/* It must be a TCP or UDP packet with a valid checksum */
1567 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1568 	if (encap_pkt) {
1569 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1570 			return;
1571 
1572 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1573 			skb->ip_summed = CHECKSUM_NONE;
1574 			return;
1575 		}
1576 		/* If we checked the outer header let the stack know */
1577 		skb->csum_level = 1;
1578 	}
1579 }
1580 
1581 static inline unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1582 {
1583 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1584 }
1585 
1586 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1587 				    struct ixgbe_rx_buffer *bi)
1588 {
1589 	struct page *page = bi->page;
1590 	dma_addr_t dma;
1591 
1592 	/* since we are recycling buffers we should seldom need to alloc */
1593 	if (likely(page))
1594 		return true;
1595 
1596 	/* alloc new page for storage */
1597 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1598 	if (unlikely(!page)) {
1599 		rx_ring->rx_stats.alloc_rx_page_failed++;
1600 		return false;
1601 	}
1602 
1603 	/* map page for use */
1604 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1605 				 ixgbe_rx_pg_size(rx_ring),
1606 				 DMA_FROM_DEVICE,
1607 				 IXGBE_RX_DMA_ATTR);
1608 
1609 	/*
1610 	 * if mapping failed free memory back to system since
1611 	 * there isn't much point in holding memory we can't use
1612 	 */
1613 	if (dma_mapping_error(rx_ring->dev, dma)) {
1614 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1615 
1616 		rx_ring->rx_stats.alloc_rx_page_failed++;
1617 		return false;
1618 	}
1619 
1620 	bi->dma = dma;
1621 	bi->page = page;
1622 	bi->page_offset = ixgbe_rx_offset(rx_ring);
1623 	bi->pagecnt_bias = 1;
1624 	rx_ring->rx_stats.alloc_rx_page++;
1625 
1626 	return true;
1627 }
1628 
1629 /**
1630  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1631  * @rx_ring: ring to place buffers on
1632  * @cleaned_count: number of buffers to replace
1633  **/
1634 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1635 {
1636 	union ixgbe_adv_rx_desc *rx_desc;
1637 	struct ixgbe_rx_buffer *bi;
1638 	u16 i = rx_ring->next_to_use;
1639 	u16 bufsz;
1640 
1641 	/* nothing to do */
1642 	if (!cleaned_count)
1643 		return;
1644 
1645 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1646 	bi = &rx_ring->rx_buffer_info[i];
1647 	i -= rx_ring->count;
1648 
1649 	bufsz = ixgbe_rx_bufsz(rx_ring);
1650 
1651 	do {
1652 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1653 			break;
1654 
1655 		/* sync the buffer for use by the device */
1656 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1657 						 bi->page_offset, bufsz,
1658 						 DMA_FROM_DEVICE);
1659 
1660 		/*
1661 		 * Refresh the desc even if buffer_addrs didn't change
1662 		 * because each write-back erases this info.
1663 		 */
1664 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1665 
1666 		rx_desc++;
1667 		bi++;
1668 		i++;
1669 		if (unlikely(!i)) {
1670 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1671 			bi = rx_ring->rx_buffer_info;
1672 			i -= rx_ring->count;
1673 		}
1674 
1675 		/* clear the length for the next_to_use descriptor */
1676 		rx_desc->wb.upper.length = 0;
1677 
1678 		cleaned_count--;
1679 	} while (cleaned_count);
1680 
1681 	i += rx_ring->count;
1682 
1683 	if (rx_ring->next_to_use != i) {
1684 		rx_ring->next_to_use = i;
1685 
1686 		/* update next to alloc since we have filled the ring */
1687 		rx_ring->next_to_alloc = i;
1688 
1689 		/* Force memory writes to complete before letting h/w
1690 		 * know there are new descriptors to fetch.  (Only
1691 		 * applicable for weak-ordered memory model archs,
1692 		 * such as IA-64).
1693 		 */
1694 		wmb();
1695 		writel(i, rx_ring->tail);
1696 	}
1697 }
1698 
1699 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1700 				   struct sk_buff *skb)
1701 {
1702 	u16 hdr_len = skb_headlen(skb);
1703 
1704 	/* set gso_size to avoid messing up TCP MSS */
1705 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1706 						 IXGBE_CB(skb)->append_cnt);
1707 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1708 }
1709 
1710 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1711 				   struct sk_buff *skb)
1712 {
1713 	/* if append_cnt is 0 then frame is not RSC */
1714 	if (!IXGBE_CB(skb)->append_cnt)
1715 		return;
1716 
1717 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1718 	rx_ring->rx_stats.rsc_flush++;
1719 
1720 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1721 
1722 	/* gso_size is computed using append_cnt so always clear it last */
1723 	IXGBE_CB(skb)->append_cnt = 0;
1724 }
1725 
1726 /**
1727  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1728  * @rx_ring: rx descriptor ring packet is being transacted on
1729  * @rx_desc: pointer to the EOP Rx descriptor
1730  * @skb: pointer to current skb being populated
1731  *
1732  * This function checks the ring, descriptor, and packet information in
1733  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1734  * other fields within the skb.
1735  **/
1736 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1737 				     union ixgbe_adv_rx_desc *rx_desc,
1738 				     struct sk_buff *skb)
1739 {
1740 	struct net_device *dev = rx_ring->netdev;
1741 	u32 flags = rx_ring->q_vector->adapter->flags;
1742 
1743 	ixgbe_update_rsc_stats(rx_ring, skb);
1744 
1745 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1746 
1747 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1748 
1749 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1750 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1751 
1752 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1753 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1754 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1755 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1756 	}
1757 
1758 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1759 		ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1760 
1761 	skb->protocol = eth_type_trans(skb, dev);
1762 
1763 	/* record Rx queue, or update MACVLAN statistics */
1764 	if (netif_is_ixgbe(dev))
1765 		skb_record_rx_queue(skb, rx_ring->queue_index);
1766 	else
1767 		macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1768 				 (skb->pkt_type == PACKET_BROADCAST) ||
1769 				 (skb->pkt_type == PACKET_MULTICAST));
1770 }
1771 
1772 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1773 			 struct sk_buff *skb)
1774 {
1775 	napi_gro_receive(&q_vector->napi, skb);
1776 }
1777 
1778 /**
1779  * ixgbe_is_non_eop - process handling of non-EOP buffers
1780  * @rx_ring: Rx ring being processed
1781  * @rx_desc: Rx descriptor for current buffer
1782  * @skb: Current socket buffer containing buffer in progress
1783  *
1784  * This function updates next to clean.  If the buffer is an EOP buffer
1785  * this function exits returning false, otherwise it will place the
1786  * sk_buff in the next buffer to be chained and return true indicating
1787  * that this is in fact a non-EOP buffer.
1788  **/
1789 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1790 			     union ixgbe_adv_rx_desc *rx_desc,
1791 			     struct sk_buff *skb)
1792 {
1793 	u32 ntc = rx_ring->next_to_clean + 1;
1794 
1795 	/* fetch, update, and store next to clean */
1796 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1797 	rx_ring->next_to_clean = ntc;
1798 
1799 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1800 
1801 	/* update RSC append count if present */
1802 	if (ring_is_rsc_enabled(rx_ring)) {
1803 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1804 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1805 
1806 		if (unlikely(rsc_enabled)) {
1807 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1808 
1809 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1810 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1811 
1812 			/* update ntc based on RSC value */
1813 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1814 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1815 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1816 		}
1817 	}
1818 
1819 	/* if we are the last buffer then there is nothing else to do */
1820 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1821 		return false;
1822 
1823 	/* place skb in next buffer to be received */
1824 	rx_ring->rx_buffer_info[ntc].skb = skb;
1825 	rx_ring->rx_stats.non_eop_descs++;
1826 
1827 	return true;
1828 }
1829 
1830 /**
1831  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1832  * @rx_ring: rx descriptor ring packet is being transacted on
1833  * @skb: pointer to current skb being adjusted
1834  *
1835  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1836  * main difference between this version and the original function is that
1837  * this function can make several assumptions about the state of things
1838  * that allow for significant optimizations versus the standard function.
1839  * As a result we can do things like drop a frag and maintain an accurate
1840  * truesize for the skb.
1841  */
1842 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1843 			    struct sk_buff *skb)
1844 {
1845 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1846 	unsigned char *va;
1847 	unsigned int pull_len;
1848 
1849 	/*
1850 	 * it is valid to use page_address instead of kmap since we are
1851 	 * working with pages allocated out of the lomem pool per
1852 	 * alloc_page(GFP_ATOMIC)
1853 	 */
1854 	va = skb_frag_address(frag);
1855 
1856 	/*
1857 	 * we need the header to contain the greater of either ETH_HLEN or
1858 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1859 	 */
1860 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1861 
1862 	/* align pull length to size of long to optimize memcpy performance */
1863 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1864 
1865 	/* update all of the pointers */
1866 	skb_frag_size_sub(frag, pull_len);
1867 	frag->page_offset += pull_len;
1868 	skb->data_len -= pull_len;
1869 	skb->tail += pull_len;
1870 }
1871 
1872 /**
1873  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1874  * @rx_ring: rx descriptor ring packet is being transacted on
1875  * @skb: pointer to current skb being updated
1876  *
1877  * This function provides a basic DMA sync up for the first fragment of an
1878  * skb.  The reason for doing this is that the first fragment cannot be
1879  * unmapped until we have reached the end of packet descriptor for a buffer
1880  * chain.
1881  */
1882 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1883 				struct sk_buff *skb)
1884 {
1885 	/* if the page was released unmap it, else just sync our portion */
1886 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1887 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1888 				     ixgbe_rx_pg_size(rx_ring),
1889 				     DMA_FROM_DEVICE,
1890 				     IXGBE_RX_DMA_ATTR);
1891 	} else if (ring_uses_build_skb(rx_ring)) {
1892 		unsigned long offset = (unsigned long)(skb->data) & ~PAGE_MASK;
1893 
1894 		dma_sync_single_range_for_cpu(rx_ring->dev,
1895 					      IXGBE_CB(skb)->dma,
1896 					      offset,
1897 					      skb_headlen(skb),
1898 					      DMA_FROM_DEVICE);
1899 	} else {
1900 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1901 
1902 		dma_sync_single_range_for_cpu(rx_ring->dev,
1903 					      IXGBE_CB(skb)->dma,
1904 					      frag->page_offset,
1905 					      skb_frag_size(frag),
1906 					      DMA_FROM_DEVICE);
1907 	}
1908 }
1909 
1910 /**
1911  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1912  * @rx_ring: rx descriptor ring packet is being transacted on
1913  * @rx_desc: pointer to the EOP Rx descriptor
1914  * @skb: pointer to current skb being fixed
1915  *
1916  * Check if the skb is valid in the XDP case it will be an error pointer.
1917  * Return true in this case to abort processing and advance to next
1918  * descriptor.
1919  *
1920  * Check for corrupted packet headers caused by senders on the local L2
1921  * embedded NIC switch not setting up their Tx Descriptors right.  These
1922  * should be very rare.
1923  *
1924  * Also address the case where we are pulling data in on pages only
1925  * and as such no data is present in the skb header.
1926  *
1927  * In addition if skb is not at least 60 bytes we need to pad it so that
1928  * it is large enough to qualify as a valid Ethernet frame.
1929  *
1930  * Returns true if an error was encountered and skb was freed.
1931  **/
1932 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1933 				  union ixgbe_adv_rx_desc *rx_desc,
1934 				  struct sk_buff *skb)
1935 {
1936 	struct net_device *netdev = rx_ring->netdev;
1937 
1938 	/* XDP packets use error pointer so abort at this point */
1939 	if (IS_ERR(skb))
1940 		return true;
1941 
1942 	/* Verify netdev is present, and that packet does not have any
1943 	 * errors that would be unacceptable to the netdev.
1944 	 */
1945 	if (!netdev ||
1946 	    (unlikely(ixgbe_test_staterr(rx_desc,
1947 					 IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1948 	     !(netdev->features & NETIF_F_RXALL)))) {
1949 		dev_kfree_skb_any(skb);
1950 		return true;
1951 	}
1952 
1953 	/* place header in linear portion of buffer */
1954 	if (!skb_headlen(skb))
1955 		ixgbe_pull_tail(rx_ring, skb);
1956 
1957 #ifdef IXGBE_FCOE
1958 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1959 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1960 		return false;
1961 
1962 #endif
1963 	/* if eth_skb_pad returns an error the skb was freed */
1964 	if (eth_skb_pad(skb))
1965 		return true;
1966 
1967 	return false;
1968 }
1969 
1970 /**
1971  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1972  * @rx_ring: rx descriptor ring to store buffers on
1973  * @old_buff: donor buffer to have page reused
1974  *
1975  * Synchronizes page for reuse by the adapter
1976  **/
1977 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1978 				struct ixgbe_rx_buffer *old_buff)
1979 {
1980 	struct ixgbe_rx_buffer *new_buff;
1981 	u16 nta = rx_ring->next_to_alloc;
1982 
1983 	new_buff = &rx_ring->rx_buffer_info[nta];
1984 
1985 	/* update, and store next to alloc */
1986 	nta++;
1987 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1988 
1989 	/* Transfer page from old buffer to new buffer.
1990 	 * Move each member individually to avoid possible store
1991 	 * forwarding stalls and unnecessary copy of skb.
1992 	 */
1993 	new_buff->dma		= old_buff->dma;
1994 	new_buff->page		= old_buff->page;
1995 	new_buff->page_offset	= old_buff->page_offset;
1996 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
1997 }
1998 
1999 static inline bool ixgbe_page_is_reserved(struct page *page)
2000 {
2001 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
2002 }
2003 
2004 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer)
2005 {
2006 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
2007 	struct page *page = rx_buffer->page;
2008 
2009 	/* avoid re-using remote pages */
2010 	if (unlikely(ixgbe_page_is_reserved(page)))
2011 		return false;
2012 
2013 #if (PAGE_SIZE < 8192)
2014 	/* if we are only owner of page we can reuse it */
2015 	if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
2016 		return false;
2017 #else
2018 	/* The last offset is a bit aggressive in that we assume the
2019 	 * worst case of FCoE being enabled and using a 3K buffer.
2020 	 * However this should have minimal impact as the 1K extra is
2021 	 * still less than one buffer in size.
2022 	 */
2023 #define IXGBE_LAST_OFFSET \
2024 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
2025 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
2026 		return false;
2027 #endif
2028 
2029 	/* If we have drained the page fragment pool we need to update
2030 	 * the pagecnt_bias and page count so that we fully restock the
2031 	 * number of references the driver holds.
2032 	 */
2033 	if (unlikely(!pagecnt_bias)) {
2034 		page_ref_add(page, USHRT_MAX);
2035 		rx_buffer->pagecnt_bias = USHRT_MAX;
2036 	}
2037 
2038 	return true;
2039 }
2040 
2041 /**
2042  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
2043  * @rx_ring: rx descriptor ring to transact packets on
2044  * @rx_buffer: buffer containing page to add
2045  * @skb: sk_buff to place the data into
2046  * @size: size of data in rx_buffer
2047  *
2048  * This function will add the data contained in rx_buffer->page to the skb.
2049  * This is done either through a direct copy if the data in the buffer is
2050  * less than the skb header size, otherwise it will just attach the page as
2051  * a frag to the skb.
2052  *
2053  * The function will then update the page offset if necessary and return
2054  * true if the buffer can be reused by the adapter.
2055  **/
2056 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
2057 			      struct ixgbe_rx_buffer *rx_buffer,
2058 			      struct sk_buff *skb,
2059 			      unsigned int size)
2060 {
2061 #if (PAGE_SIZE < 8192)
2062 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2063 #else
2064 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2065 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2066 				SKB_DATA_ALIGN(size);
2067 #endif
2068 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2069 			rx_buffer->page_offset, size, truesize);
2070 #if (PAGE_SIZE < 8192)
2071 	rx_buffer->page_offset ^= truesize;
2072 #else
2073 	rx_buffer->page_offset += truesize;
2074 #endif
2075 }
2076 
2077 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2078 						   union ixgbe_adv_rx_desc *rx_desc,
2079 						   struct sk_buff **skb,
2080 						   const unsigned int size)
2081 {
2082 	struct ixgbe_rx_buffer *rx_buffer;
2083 
2084 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2085 	prefetchw(rx_buffer->page);
2086 	*skb = rx_buffer->skb;
2087 
2088 	/* Delay unmapping of the first packet. It carries the header
2089 	 * information, HW may still access the header after the writeback.
2090 	 * Only unmap it when EOP is reached
2091 	 */
2092 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2093 		if (!*skb)
2094 			goto skip_sync;
2095 	} else {
2096 		if (*skb)
2097 			ixgbe_dma_sync_frag(rx_ring, *skb);
2098 	}
2099 
2100 	/* we are reusing so sync this buffer for CPU use */
2101 	dma_sync_single_range_for_cpu(rx_ring->dev,
2102 				      rx_buffer->dma,
2103 				      rx_buffer->page_offset,
2104 				      size,
2105 				      DMA_FROM_DEVICE);
2106 skip_sync:
2107 	rx_buffer->pagecnt_bias--;
2108 
2109 	return rx_buffer;
2110 }
2111 
2112 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2113 				struct ixgbe_rx_buffer *rx_buffer,
2114 				struct sk_buff *skb)
2115 {
2116 	if (ixgbe_can_reuse_rx_page(rx_buffer)) {
2117 		/* hand second half of page back to the ring */
2118 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2119 	} else {
2120 		if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2121 			/* the page has been released from the ring */
2122 			IXGBE_CB(skb)->page_released = true;
2123 		} else {
2124 			/* we are not reusing the buffer so unmap it */
2125 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2126 					     ixgbe_rx_pg_size(rx_ring),
2127 					     DMA_FROM_DEVICE,
2128 					     IXGBE_RX_DMA_ATTR);
2129 		}
2130 		__page_frag_cache_drain(rx_buffer->page,
2131 					rx_buffer->pagecnt_bias);
2132 	}
2133 
2134 	/* clear contents of rx_buffer */
2135 	rx_buffer->page = NULL;
2136 	rx_buffer->skb = NULL;
2137 }
2138 
2139 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2140 					   struct ixgbe_rx_buffer *rx_buffer,
2141 					   struct xdp_buff *xdp,
2142 					   union ixgbe_adv_rx_desc *rx_desc)
2143 {
2144 	unsigned int size = xdp->data_end - xdp->data;
2145 #if (PAGE_SIZE < 8192)
2146 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2147 #else
2148 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2149 					       xdp->data_hard_start);
2150 #endif
2151 	struct sk_buff *skb;
2152 
2153 	/* prefetch first cache line of first page */
2154 	prefetch(xdp->data);
2155 #if L1_CACHE_BYTES < 128
2156 	prefetch(xdp->data + L1_CACHE_BYTES);
2157 #endif
2158 	/* Note, we get here by enabling legacy-rx via:
2159 	 *
2160 	 *    ethtool --set-priv-flags <dev> legacy-rx on
2161 	 *
2162 	 * In this mode, we currently get 0 extra XDP headroom as
2163 	 * opposed to having legacy-rx off, where we process XDP
2164 	 * packets going to stack via ixgbe_build_skb(). The latter
2165 	 * provides us currently with 192 bytes of headroom.
2166 	 *
2167 	 * For ixgbe_construct_skb() mode it means that the
2168 	 * xdp->data_meta will always point to xdp->data, since
2169 	 * the helper cannot expand the head. Should this ever
2170 	 * change in future for legacy-rx mode on, then lets also
2171 	 * add xdp->data_meta handling here.
2172 	 */
2173 
2174 	/* allocate a skb to store the frags */
2175 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2176 	if (unlikely(!skb))
2177 		return NULL;
2178 
2179 	if (size > IXGBE_RX_HDR_SIZE) {
2180 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2181 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2182 
2183 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2184 				xdp->data - page_address(rx_buffer->page),
2185 				size, truesize);
2186 #if (PAGE_SIZE < 8192)
2187 		rx_buffer->page_offset ^= truesize;
2188 #else
2189 		rx_buffer->page_offset += truesize;
2190 #endif
2191 	} else {
2192 		memcpy(__skb_put(skb, size),
2193 		       xdp->data, ALIGN(size, sizeof(long)));
2194 		rx_buffer->pagecnt_bias++;
2195 	}
2196 
2197 	return skb;
2198 }
2199 
2200 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2201 				       struct ixgbe_rx_buffer *rx_buffer,
2202 				       struct xdp_buff *xdp,
2203 				       union ixgbe_adv_rx_desc *rx_desc)
2204 {
2205 	unsigned int metasize = xdp->data - xdp->data_meta;
2206 #if (PAGE_SIZE < 8192)
2207 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2208 #else
2209 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2210 				SKB_DATA_ALIGN(xdp->data_end -
2211 					       xdp->data_hard_start);
2212 #endif
2213 	struct sk_buff *skb;
2214 
2215 	/* Prefetch first cache line of first page. If xdp->data_meta
2216 	 * is unused, this points extactly as xdp->data, otherwise we
2217 	 * likely have a consumer accessing first few bytes of meta
2218 	 * data, and then actual data.
2219 	 */
2220 	prefetch(xdp->data_meta);
2221 #if L1_CACHE_BYTES < 128
2222 	prefetch(xdp->data_meta + L1_CACHE_BYTES);
2223 #endif
2224 
2225 	/* build an skb to around the page buffer */
2226 	skb = build_skb(xdp->data_hard_start, truesize);
2227 	if (unlikely(!skb))
2228 		return NULL;
2229 
2230 	/* update pointers within the skb to store the data */
2231 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
2232 	__skb_put(skb, xdp->data_end - xdp->data);
2233 	if (metasize)
2234 		skb_metadata_set(skb, metasize);
2235 
2236 	/* record DMA address if this is the start of a chain of buffers */
2237 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2238 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2239 
2240 	/* update buffer offset */
2241 #if (PAGE_SIZE < 8192)
2242 	rx_buffer->page_offset ^= truesize;
2243 #else
2244 	rx_buffer->page_offset += truesize;
2245 #endif
2246 
2247 	return skb;
2248 }
2249 
2250 #define IXGBE_XDP_PASS 0
2251 #define IXGBE_XDP_CONSUMED 1
2252 #define IXGBE_XDP_TX 2
2253 
2254 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
2255 			       struct xdp_buff *xdp);
2256 
2257 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2258 				     struct ixgbe_ring *rx_ring,
2259 				     struct xdp_buff *xdp)
2260 {
2261 	int err, result = IXGBE_XDP_PASS;
2262 	struct bpf_prog *xdp_prog;
2263 	u32 act;
2264 
2265 	rcu_read_lock();
2266 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2267 
2268 	if (!xdp_prog)
2269 		goto xdp_out;
2270 
2271 	act = bpf_prog_run_xdp(xdp_prog, xdp);
2272 	switch (act) {
2273 	case XDP_PASS:
2274 		break;
2275 	case XDP_TX:
2276 		result = ixgbe_xmit_xdp_ring(adapter, xdp);
2277 		break;
2278 	case XDP_REDIRECT:
2279 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2280 		if (!err)
2281 			result = IXGBE_XDP_TX;
2282 		else
2283 			result = IXGBE_XDP_CONSUMED;
2284 		break;
2285 	default:
2286 		bpf_warn_invalid_xdp_action(act);
2287 		/* fallthrough */
2288 	case XDP_ABORTED:
2289 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2290 		/* fallthrough -- handle aborts by dropping packet */
2291 	case XDP_DROP:
2292 		result = IXGBE_XDP_CONSUMED;
2293 		break;
2294 	}
2295 xdp_out:
2296 	rcu_read_unlock();
2297 	return ERR_PTR(-result);
2298 }
2299 
2300 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2301 				 struct ixgbe_rx_buffer *rx_buffer,
2302 				 unsigned int size)
2303 {
2304 #if (PAGE_SIZE < 8192)
2305 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2306 
2307 	rx_buffer->page_offset ^= truesize;
2308 #else
2309 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2310 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2311 				SKB_DATA_ALIGN(size);
2312 
2313 	rx_buffer->page_offset += truesize;
2314 #endif
2315 }
2316 
2317 /**
2318  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2319  * @q_vector: structure containing interrupt and ring information
2320  * @rx_ring: rx descriptor ring to transact packets on
2321  * @budget: Total limit on number of packets to process
2322  *
2323  * This function provides a "bounce buffer" approach to Rx interrupt
2324  * processing.  The advantage to this is that on systems that have
2325  * expensive overhead for IOMMU access this provides a means of avoiding
2326  * it by maintaining the mapping of the page to the syste.
2327  *
2328  * Returns amount of work completed
2329  **/
2330 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2331 			       struct ixgbe_ring *rx_ring,
2332 			       const int budget)
2333 {
2334 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2335 	struct ixgbe_adapter *adapter = q_vector->adapter;
2336 #ifdef IXGBE_FCOE
2337 	int ddp_bytes;
2338 	unsigned int mss = 0;
2339 #endif /* IXGBE_FCOE */
2340 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2341 	bool xdp_xmit = false;
2342 	struct xdp_buff xdp;
2343 
2344 	xdp.rxq = &rx_ring->xdp_rxq;
2345 
2346 	while (likely(total_rx_packets < budget)) {
2347 		union ixgbe_adv_rx_desc *rx_desc;
2348 		struct ixgbe_rx_buffer *rx_buffer;
2349 		struct sk_buff *skb;
2350 		unsigned int size;
2351 
2352 		/* return some buffers to hardware, one at a time is too slow */
2353 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2354 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2355 			cleaned_count = 0;
2356 		}
2357 
2358 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2359 		size = le16_to_cpu(rx_desc->wb.upper.length);
2360 		if (!size)
2361 			break;
2362 
2363 		/* This memory barrier is needed to keep us from reading
2364 		 * any other fields out of the rx_desc until we know the
2365 		 * descriptor has been written back
2366 		 */
2367 		dma_rmb();
2368 
2369 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size);
2370 
2371 		/* retrieve a buffer from the ring */
2372 		if (!skb) {
2373 			xdp.data = page_address(rx_buffer->page) +
2374 				   rx_buffer->page_offset;
2375 			xdp.data_meta = xdp.data;
2376 			xdp.data_hard_start = xdp.data -
2377 					      ixgbe_rx_offset(rx_ring);
2378 			xdp.data_end = xdp.data + size;
2379 
2380 			skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2381 		}
2382 
2383 		if (IS_ERR(skb)) {
2384 			if (PTR_ERR(skb) == -IXGBE_XDP_TX) {
2385 				xdp_xmit = true;
2386 				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2387 			} else {
2388 				rx_buffer->pagecnt_bias++;
2389 			}
2390 			total_rx_packets++;
2391 			total_rx_bytes += size;
2392 		} else if (skb) {
2393 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2394 		} else if (ring_uses_build_skb(rx_ring)) {
2395 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2396 					      &xdp, rx_desc);
2397 		} else {
2398 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2399 						  &xdp, rx_desc);
2400 		}
2401 
2402 		/* exit if we failed to retrieve a buffer */
2403 		if (!skb) {
2404 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2405 			rx_buffer->pagecnt_bias++;
2406 			break;
2407 		}
2408 
2409 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb);
2410 		cleaned_count++;
2411 
2412 		/* place incomplete frames back on ring for completion */
2413 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2414 			continue;
2415 
2416 		/* verify the packet layout is correct */
2417 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2418 			continue;
2419 
2420 		/* probably a little skewed due to removing CRC */
2421 		total_rx_bytes += skb->len;
2422 
2423 		/* populate checksum, timestamp, VLAN, and protocol */
2424 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2425 
2426 #ifdef IXGBE_FCOE
2427 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2428 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2429 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2430 			/* include DDPed FCoE data */
2431 			if (ddp_bytes > 0) {
2432 				if (!mss) {
2433 					mss = rx_ring->netdev->mtu -
2434 						sizeof(struct fcoe_hdr) -
2435 						sizeof(struct fc_frame_header) -
2436 						sizeof(struct fcoe_crc_eof);
2437 					if (mss > 512)
2438 						mss &= ~511;
2439 				}
2440 				total_rx_bytes += ddp_bytes;
2441 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2442 								 mss);
2443 			}
2444 			if (!ddp_bytes) {
2445 				dev_kfree_skb_any(skb);
2446 				continue;
2447 			}
2448 		}
2449 
2450 #endif /* IXGBE_FCOE */
2451 		ixgbe_rx_skb(q_vector, skb);
2452 
2453 		/* update budget accounting */
2454 		total_rx_packets++;
2455 	}
2456 
2457 	if (xdp_xmit) {
2458 		struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
2459 
2460 		/* Force memory writes to complete before letting h/w
2461 		 * know there are new descriptors to fetch.
2462 		 */
2463 		wmb();
2464 		writel(ring->next_to_use, ring->tail);
2465 
2466 		xdp_do_flush_map();
2467 	}
2468 
2469 	u64_stats_update_begin(&rx_ring->syncp);
2470 	rx_ring->stats.packets += total_rx_packets;
2471 	rx_ring->stats.bytes += total_rx_bytes;
2472 	u64_stats_update_end(&rx_ring->syncp);
2473 	q_vector->rx.total_packets += total_rx_packets;
2474 	q_vector->rx.total_bytes += total_rx_bytes;
2475 
2476 	return total_rx_packets;
2477 }
2478 
2479 /**
2480  * ixgbe_configure_msix - Configure MSI-X hardware
2481  * @adapter: board private structure
2482  *
2483  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2484  * interrupts.
2485  **/
2486 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2487 {
2488 	struct ixgbe_q_vector *q_vector;
2489 	int v_idx;
2490 	u32 mask;
2491 
2492 	/* Populate MSIX to EITR Select */
2493 	if (adapter->num_vfs > 32) {
2494 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2495 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2496 	}
2497 
2498 	/*
2499 	 * Populate the IVAR table and set the ITR values to the
2500 	 * corresponding register.
2501 	 */
2502 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2503 		struct ixgbe_ring *ring;
2504 		q_vector = adapter->q_vector[v_idx];
2505 
2506 		ixgbe_for_each_ring(ring, q_vector->rx)
2507 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2508 
2509 		ixgbe_for_each_ring(ring, q_vector->tx)
2510 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2511 
2512 		ixgbe_write_eitr(q_vector);
2513 	}
2514 
2515 	switch (adapter->hw.mac.type) {
2516 	case ixgbe_mac_82598EB:
2517 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2518 			       v_idx);
2519 		break;
2520 	case ixgbe_mac_82599EB:
2521 	case ixgbe_mac_X540:
2522 	case ixgbe_mac_X550:
2523 	case ixgbe_mac_X550EM_x:
2524 	case ixgbe_mac_x550em_a:
2525 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2526 		break;
2527 	default:
2528 		break;
2529 	}
2530 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2531 
2532 	/* set up to autoclear timer, and the vectors */
2533 	mask = IXGBE_EIMS_ENABLE_MASK;
2534 	mask &= ~(IXGBE_EIMS_OTHER |
2535 		  IXGBE_EIMS_MAILBOX |
2536 		  IXGBE_EIMS_LSC);
2537 
2538 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2539 }
2540 
2541 /**
2542  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2543  * @q_vector: structure containing interrupt and ring information
2544  * @ring_container: structure containing ring performance data
2545  *
2546  *      Stores a new ITR value based on packets and byte
2547  *      counts during the last interrupt.  The advantage of per interrupt
2548  *      computation is faster updates and more accurate ITR for the current
2549  *      traffic pattern.  Constants in this function were computed
2550  *      based on theoretical maximum wire speed and thresholds were set based
2551  *      on testing data as well as attempting to minimize response time
2552  *      while increasing bulk throughput.
2553  **/
2554 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2555 			     struct ixgbe_ring_container *ring_container)
2556 {
2557 	unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2558 			   IXGBE_ITR_ADAPTIVE_LATENCY;
2559 	unsigned int avg_wire_size, packets, bytes;
2560 	unsigned long next_update = jiffies;
2561 
2562 	/* If we don't have any rings just leave ourselves set for maximum
2563 	 * possible latency so we take ourselves out of the equation.
2564 	 */
2565 	if (!ring_container->ring)
2566 		return;
2567 
2568 	/* If we didn't update within up to 1 - 2 jiffies we can assume
2569 	 * that either packets are coming in so slow there hasn't been
2570 	 * any work, or that there is so much work that NAPI is dealing
2571 	 * with interrupt moderation and we don't need to do anything.
2572 	 */
2573 	if (time_after(next_update, ring_container->next_update))
2574 		goto clear_counts;
2575 
2576 	packets = ring_container->total_packets;
2577 
2578 	/* We have no packets to actually measure against. This means
2579 	 * either one of the other queues on this vector is active or
2580 	 * we are a Tx queue doing TSO with too high of an interrupt rate.
2581 	 *
2582 	 * When this occurs just tick up our delay by the minimum value
2583 	 * and hope that this extra delay will prevent us from being called
2584 	 * without any work on our queue.
2585 	 */
2586 	if (!packets) {
2587 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2588 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2589 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2590 		itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2591 		goto clear_counts;
2592 	}
2593 
2594 	bytes = ring_container->total_bytes;
2595 
2596 	/* If packets are less than 4 or bytes are less than 9000 assume
2597 	 * insufficient data to use bulk rate limiting approach. We are
2598 	 * likely latency driven.
2599 	 */
2600 	if (packets < 4 && bytes < 9000) {
2601 		itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2602 		goto adjust_by_size;
2603 	}
2604 
2605 	/* Between 4 and 48 we can assume that our current interrupt delay
2606 	 * is only slightly too low. As such we should increase it by a small
2607 	 * fixed amount.
2608 	 */
2609 	if (packets < 48) {
2610 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2611 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2612 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2613 		goto clear_counts;
2614 	}
2615 
2616 	/* Between 48 and 96 is our "goldilocks" zone where we are working
2617 	 * out "just right". Just report that our current ITR is good for us.
2618 	 */
2619 	if (packets < 96) {
2620 		itr = q_vector->itr >> 2;
2621 		goto clear_counts;
2622 	}
2623 
2624 	/* If packet count is 96 or greater we are likely looking at a slight
2625 	 * overrun of the delay we want. Try halving our delay to see if that
2626 	 * will cut the number of packets in half per interrupt.
2627 	 */
2628 	if (packets < 256) {
2629 		itr = q_vector->itr >> 3;
2630 		if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2631 			itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2632 		goto clear_counts;
2633 	}
2634 
2635 	/* The paths below assume we are dealing with a bulk ITR since number
2636 	 * of packets is 256 or greater. We are just going to have to compute
2637 	 * a value and try to bring the count under control, though for smaller
2638 	 * packet sizes there isn't much we can do as NAPI polling will likely
2639 	 * be kicking in sooner rather than later.
2640 	 */
2641 	itr = IXGBE_ITR_ADAPTIVE_BULK;
2642 
2643 adjust_by_size:
2644 	/* If packet counts are 256 or greater we can assume we have a gross
2645 	 * overestimation of what the rate should be. Instead of trying to fine
2646 	 * tune it just use the formula below to try and dial in an exact value
2647 	 * give the current packet size of the frame.
2648 	 */
2649 	avg_wire_size = bytes / packets;
2650 
2651 	/* The following is a crude approximation of:
2652 	 *  wmem_default / (size + overhead) = desired_pkts_per_int
2653 	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2654 	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2655 	 *
2656 	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
2657 	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2658 	 * formula down to
2659 	 *
2660 	 *  (170 * (size + 24)) / (size + 640) = ITR
2661 	 *
2662 	 * We first do some math on the packet size and then finally bitshift
2663 	 * by 8 after rounding up. We also have to account for PCIe link speed
2664 	 * difference as ITR scales based on this.
2665 	 */
2666 	if (avg_wire_size <= 60) {
2667 		/* Start at 50k ints/sec */
2668 		avg_wire_size = 5120;
2669 	} else if (avg_wire_size <= 316) {
2670 		/* 50K ints/sec to 16K ints/sec */
2671 		avg_wire_size *= 40;
2672 		avg_wire_size += 2720;
2673 	} else if (avg_wire_size <= 1084) {
2674 		/* 16K ints/sec to 9.2K ints/sec */
2675 		avg_wire_size *= 15;
2676 		avg_wire_size += 11452;
2677 	} else if (avg_wire_size <= 1980) {
2678 		/* 9.2K ints/sec to 8K ints/sec */
2679 		avg_wire_size *= 5;
2680 		avg_wire_size += 22420;
2681 	} else {
2682 		/* plateau at a limit of 8K ints/sec */
2683 		avg_wire_size = 32256;
2684 	}
2685 
2686 	/* If we are in low latency mode half our delay which doubles the rate
2687 	 * to somewhere between 100K to 16K ints/sec
2688 	 */
2689 	if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2690 		avg_wire_size >>= 1;
2691 
2692 	/* Resultant value is 256 times larger than it needs to be. This
2693 	 * gives us room to adjust the value as needed to either increase
2694 	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2695 	 *
2696 	 * Use addition as we have already recorded the new latency flag
2697 	 * for the ITR value.
2698 	 */
2699 	switch (q_vector->adapter->link_speed) {
2700 	case IXGBE_LINK_SPEED_10GB_FULL:
2701 	case IXGBE_LINK_SPEED_100_FULL:
2702 	default:
2703 		itr += DIV_ROUND_UP(avg_wire_size,
2704 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2705 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2706 		break;
2707 	case IXGBE_LINK_SPEED_2_5GB_FULL:
2708 	case IXGBE_LINK_SPEED_1GB_FULL:
2709 	case IXGBE_LINK_SPEED_10_FULL:
2710 		itr += DIV_ROUND_UP(avg_wire_size,
2711 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2712 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2713 		break;
2714 	}
2715 
2716 clear_counts:
2717 	/* write back value */
2718 	ring_container->itr = itr;
2719 
2720 	/* next update should occur within next jiffy */
2721 	ring_container->next_update = next_update + 1;
2722 
2723 	ring_container->total_bytes = 0;
2724 	ring_container->total_packets = 0;
2725 }
2726 
2727 /**
2728  * ixgbe_write_eitr - write EITR register in hardware specific way
2729  * @q_vector: structure containing interrupt and ring information
2730  *
2731  * This function is made to be called by ethtool and by the driver
2732  * when it needs to update EITR registers at runtime.  Hardware
2733  * specific quirks/differences are taken care of here.
2734  */
2735 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2736 {
2737 	struct ixgbe_adapter *adapter = q_vector->adapter;
2738 	struct ixgbe_hw *hw = &adapter->hw;
2739 	int v_idx = q_vector->v_idx;
2740 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2741 
2742 	switch (adapter->hw.mac.type) {
2743 	case ixgbe_mac_82598EB:
2744 		/* must write high and low 16 bits to reset counter */
2745 		itr_reg |= (itr_reg << 16);
2746 		break;
2747 	case ixgbe_mac_82599EB:
2748 	case ixgbe_mac_X540:
2749 	case ixgbe_mac_X550:
2750 	case ixgbe_mac_X550EM_x:
2751 	case ixgbe_mac_x550em_a:
2752 		/*
2753 		 * set the WDIS bit to not clear the timer bits and cause an
2754 		 * immediate assertion of the interrupt
2755 		 */
2756 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2757 		break;
2758 	default:
2759 		break;
2760 	}
2761 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2762 }
2763 
2764 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2765 {
2766 	u32 new_itr;
2767 
2768 	ixgbe_update_itr(q_vector, &q_vector->tx);
2769 	ixgbe_update_itr(q_vector, &q_vector->rx);
2770 
2771 	/* use the smallest value of new ITR delay calculations */
2772 	new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2773 
2774 	/* Clear latency flag if set, shift into correct position */
2775 	new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2776 	new_itr <<= 2;
2777 
2778 	if (new_itr != q_vector->itr) {
2779 		/* save the algorithm value here */
2780 		q_vector->itr = new_itr;
2781 
2782 		ixgbe_write_eitr(q_vector);
2783 	}
2784 }
2785 
2786 /**
2787  * ixgbe_check_overtemp_subtask - check for over temperature
2788  * @adapter: pointer to adapter
2789  **/
2790 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2791 {
2792 	struct ixgbe_hw *hw = &adapter->hw;
2793 	u32 eicr = adapter->interrupt_event;
2794 	s32 rc;
2795 
2796 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2797 		return;
2798 
2799 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2800 		return;
2801 
2802 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2803 
2804 	switch (hw->device_id) {
2805 	case IXGBE_DEV_ID_82599_T3_LOM:
2806 		/*
2807 		 * Since the warning interrupt is for both ports
2808 		 * we don't have to check if:
2809 		 *  - This interrupt wasn't for our port.
2810 		 *  - We may have missed the interrupt so always have to
2811 		 *    check if we  got a LSC
2812 		 */
2813 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2814 		    !(eicr & IXGBE_EICR_LSC))
2815 			return;
2816 
2817 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2818 			u32 speed;
2819 			bool link_up = false;
2820 
2821 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2822 
2823 			if (link_up)
2824 				return;
2825 		}
2826 
2827 		/* Check if this is not due to overtemp */
2828 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2829 			return;
2830 
2831 		break;
2832 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2833 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2834 		rc = hw->phy.ops.check_overtemp(hw);
2835 		if (rc != IXGBE_ERR_OVERTEMP)
2836 			return;
2837 		break;
2838 	default:
2839 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2840 			return;
2841 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2842 			return;
2843 		break;
2844 	}
2845 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2846 
2847 	adapter->interrupt_event = 0;
2848 }
2849 
2850 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2851 {
2852 	struct ixgbe_hw *hw = &adapter->hw;
2853 
2854 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2855 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2856 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2857 		/* write to clear the interrupt */
2858 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2859 	}
2860 }
2861 
2862 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2863 {
2864 	struct ixgbe_hw *hw = &adapter->hw;
2865 
2866 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2867 		return;
2868 
2869 	switch (adapter->hw.mac.type) {
2870 	case ixgbe_mac_82599EB:
2871 		/*
2872 		 * Need to check link state so complete overtemp check
2873 		 * on service task
2874 		 */
2875 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2876 		     (eicr & IXGBE_EICR_LSC)) &&
2877 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2878 			adapter->interrupt_event = eicr;
2879 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2880 			ixgbe_service_event_schedule(adapter);
2881 			return;
2882 		}
2883 		return;
2884 	case ixgbe_mac_x550em_a:
2885 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2886 			adapter->interrupt_event = eicr;
2887 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2888 			ixgbe_service_event_schedule(adapter);
2889 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2890 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2891 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2892 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2893 		}
2894 		return;
2895 	case ixgbe_mac_X550:
2896 	case ixgbe_mac_X540:
2897 		if (!(eicr & IXGBE_EICR_TS))
2898 			return;
2899 		break;
2900 	default:
2901 		return;
2902 	}
2903 
2904 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2905 }
2906 
2907 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2908 {
2909 	switch (hw->mac.type) {
2910 	case ixgbe_mac_82598EB:
2911 		if (hw->phy.type == ixgbe_phy_nl)
2912 			return true;
2913 		return false;
2914 	case ixgbe_mac_82599EB:
2915 	case ixgbe_mac_X550EM_x:
2916 	case ixgbe_mac_x550em_a:
2917 		switch (hw->mac.ops.get_media_type(hw)) {
2918 		case ixgbe_media_type_fiber:
2919 		case ixgbe_media_type_fiber_qsfp:
2920 			return true;
2921 		default:
2922 			return false;
2923 		}
2924 	default:
2925 		return false;
2926 	}
2927 }
2928 
2929 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2930 {
2931 	struct ixgbe_hw *hw = &adapter->hw;
2932 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2933 
2934 	if (!ixgbe_is_sfp(hw))
2935 		return;
2936 
2937 	/* Later MAC's use different SDP */
2938 	if (hw->mac.type >= ixgbe_mac_X540)
2939 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2940 
2941 	if (eicr & eicr_mask) {
2942 		/* Clear the interrupt */
2943 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2944 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2945 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2946 			adapter->sfp_poll_time = 0;
2947 			ixgbe_service_event_schedule(adapter);
2948 		}
2949 	}
2950 
2951 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2952 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2953 		/* Clear the interrupt */
2954 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2955 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2956 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2957 			ixgbe_service_event_schedule(adapter);
2958 		}
2959 	}
2960 }
2961 
2962 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2963 {
2964 	struct ixgbe_hw *hw = &adapter->hw;
2965 
2966 	adapter->lsc_int++;
2967 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2968 	adapter->link_check_timeout = jiffies;
2969 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2970 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2971 		IXGBE_WRITE_FLUSH(hw);
2972 		ixgbe_service_event_schedule(adapter);
2973 	}
2974 }
2975 
2976 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2977 					   u64 qmask)
2978 {
2979 	u32 mask;
2980 	struct ixgbe_hw *hw = &adapter->hw;
2981 
2982 	switch (hw->mac.type) {
2983 	case ixgbe_mac_82598EB:
2984 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2985 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2986 		break;
2987 	case ixgbe_mac_82599EB:
2988 	case ixgbe_mac_X540:
2989 	case ixgbe_mac_X550:
2990 	case ixgbe_mac_X550EM_x:
2991 	case ixgbe_mac_x550em_a:
2992 		mask = (qmask & 0xFFFFFFFF);
2993 		if (mask)
2994 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2995 		mask = (qmask >> 32);
2996 		if (mask)
2997 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2998 		break;
2999 	default:
3000 		break;
3001 	}
3002 	/* skip the flush */
3003 }
3004 
3005 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
3006 					    u64 qmask)
3007 {
3008 	u32 mask;
3009 	struct ixgbe_hw *hw = &adapter->hw;
3010 
3011 	switch (hw->mac.type) {
3012 	case ixgbe_mac_82598EB:
3013 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
3014 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
3015 		break;
3016 	case ixgbe_mac_82599EB:
3017 	case ixgbe_mac_X540:
3018 	case ixgbe_mac_X550:
3019 	case ixgbe_mac_X550EM_x:
3020 	case ixgbe_mac_x550em_a:
3021 		mask = (qmask & 0xFFFFFFFF);
3022 		if (mask)
3023 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
3024 		mask = (qmask >> 32);
3025 		if (mask)
3026 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
3027 		break;
3028 	default:
3029 		break;
3030 	}
3031 	/* skip the flush */
3032 }
3033 
3034 /**
3035  * ixgbe_irq_enable - Enable default interrupt generation settings
3036  * @adapter: board private structure
3037  * @queues: enable irqs for queues
3038  * @flush: flush register write
3039  **/
3040 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
3041 				    bool flush)
3042 {
3043 	struct ixgbe_hw *hw = &adapter->hw;
3044 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
3045 
3046 	/* don't reenable LSC while waiting for link */
3047 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
3048 		mask &= ~IXGBE_EIMS_LSC;
3049 
3050 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
3051 		switch (adapter->hw.mac.type) {
3052 		case ixgbe_mac_82599EB:
3053 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
3054 			break;
3055 		case ixgbe_mac_X540:
3056 		case ixgbe_mac_X550:
3057 		case ixgbe_mac_X550EM_x:
3058 		case ixgbe_mac_x550em_a:
3059 			mask |= IXGBE_EIMS_TS;
3060 			break;
3061 		default:
3062 			break;
3063 		}
3064 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3065 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3066 	switch (adapter->hw.mac.type) {
3067 	case ixgbe_mac_82599EB:
3068 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3069 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
3070 		/* fall through */
3071 	case ixgbe_mac_X540:
3072 	case ixgbe_mac_X550:
3073 	case ixgbe_mac_X550EM_x:
3074 	case ixgbe_mac_x550em_a:
3075 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3076 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3077 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3078 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3079 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3080 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3081 		mask |= IXGBE_EIMS_ECC;
3082 		mask |= IXGBE_EIMS_MAILBOX;
3083 		break;
3084 	default:
3085 		break;
3086 	}
3087 
3088 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3089 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3090 		mask |= IXGBE_EIMS_FLOW_DIR;
3091 
3092 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3093 	if (queues)
3094 		ixgbe_irq_enable_queues(adapter, ~0);
3095 	if (flush)
3096 		IXGBE_WRITE_FLUSH(&adapter->hw);
3097 }
3098 
3099 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3100 {
3101 	struct ixgbe_adapter *adapter = data;
3102 	struct ixgbe_hw *hw = &adapter->hw;
3103 	u32 eicr;
3104 
3105 	/*
3106 	 * Workaround for Silicon errata.  Use clear-by-write instead
3107 	 * of clear-by-read.  Reading with EICS will return the
3108 	 * interrupt causes without clearing, which later be done
3109 	 * with the write to EICR.
3110 	 */
3111 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3112 
3113 	/* The lower 16bits of the EICR register are for the queue interrupts
3114 	 * which should be masked here in order to not accidentally clear them if
3115 	 * the bits are high when ixgbe_msix_other is called. There is a race
3116 	 * condition otherwise which results in possible performance loss
3117 	 * especially if the ixgbe_msix_other interrupt is triggering
3118 	 * consistently (as it would when PPS is turned on for the X540 device)
3119 	 */
3120 	eicr &= 0xFFFF0000;
3121 
3122 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3123 
3124 	if (eicr & IXGBE_EICR_LSC)
3125 		ixgbe_check_lsc(adapter);
3126 
3127 	if (eicr & IXGBE_EICR_MAILBOX)
3128 		ixgbe_msg_task(adapter);
3129 
3130 	switch (hw->mac.type) {
3131 	case ixgbe_mac_82599EB:
3132 	case ixgbe_mac_X540:
3133 	case ixgbe_mac_X550:
3134 	case ixgbe_mac_X550EM_x:
3135 	case ixgbe_mac_x550em_a:
3136 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3137 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3138 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3139 			ixgbe_service_event_schedule(adapter);
3140 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3141 					IXGBE_EICR_GPI_SDP0_X540);
3142 		}
3143 		if (eicr & IXGBE_EICR_ECC) {
3144 			e_info(link, "Received ECC Err, initiating reset\n");
3145 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3146 			ixgbe_service_event_schedule(adapter);
3147 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3148 		}
3149 		/* Handle Flow Director Full threshold interrupt */
3150 		if (eicr & IXGBE_EICR_FLOW_DIR) {
3151 			int reinit_count = 0;
3152 			int i;
3153 			for (i = 0; i < adapter->num_tx_queues; i++) {
3154 				struct ixgbe_ring *ring = adapter->tx_ring[i];
3155 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3156 						       &ring->state))
3157 					reinit_count++;
3158 			}
3159 			if (reinit_count) {
3160 				/* no more flow director interrupts until after init */
3161 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3162 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3163 				ixgbe_service_event_schedule(adapter);
3164 			}
3165 		}
3166 		ixgbe_check_sfp_event(adapter, eicr);
3167 		ixgbe_check_overtemp_event(adapter, eicr);
3168 		break;
3169 	default:
3170 		break;
3171 	}
3172 
3173 	ixgbe_check_fan_failure(adapter, eicr);
3174 
3175 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3176 		ixgbe_ptp_check_pps_event(adapter);
3177 
3178 	/* re-enable the original interrupt state, no lsc, no queues */
3179 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3180 		ixgbe_irq_enable(adapter, false, false);
3181 
3182 	return IRQ_HANDLED;
3183 }
3184 
3185 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3186 {
3187 	struct ixgbe_q_vector *q_vector = data;
3188 
3189 	/* EIAM disabled interrupts (on this vector) for us */
3190 
3191 	if (q_vector->rx.ring || q_vector->tx.ring)
3192 		napi_schedule_irqoff(&q_vector->napi);
3193 
3194 	return IRQ_HANDLED;
3195 }
3196 
3197 /**
3198  * ixgbe_poll - NAPI Rx polling callback
3199  * @napi: structure for representing this polling device
3200  * @budget: how many packets driver is allowed to clean
3201  *
3202  * This function is used for legacy and MSI, NAPI mode
3203  **/
3204 int ixgbe_poll(struct napi_struct *napi, int budget)
3205 {
3206 	struct ixgbe_q_vector *q_vector =
3207 				container_of(napi, struct ixgbe_q_vector, napi);
3208 	struct ixgbe_adapter *adapter = q_vector->adapter;
3209 	struct ixgbe_ring *ring;
3210 	int per_ring_budget, work_done = 0;
3211 	bool clean_complete = true;
3212 
3213 #ifdef CONFIG_IXGBE_DCA
3214 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3215 		ixgbe_update_dca(q_vector);
3216 #endif
3217 
3218 	ixgbe_for_each_ring(ring, q_vector->tx) {
3219 		if (!ixgbe_clean_tx_irq(q_vector, ring, budget))
3220 			clean_complete = false;
3221 	}
3222 
3223 	/* Exit if we are called by netpoll */
3224 	if (budget <= 0)
3225 		return budget;
3226 
3227 	/* attempt to distribute budget to each queue fairly, but don't allow
3228 	 * the budget to go below 1 because we'll exit polling */
3229 	if (q_vector->rx.count > 1)
3230 		per_ring_budget = max(budget/q_vector->rx.count, 1);
3231 	else
3232 		per_ring_budget = budget;
3233 
3234 	ixgbe_for_each_ring(ring, q_vector->rx) {
3235 		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
3236 						 per_ring_budget);
3237 
3238 		work_done += cleaned;
3239 		if (cleaned >= per_ring_budget)
3240 			clean_complete = false;
3241 	}
3242 
3243 	/* If all work not completed, return budget and keep polling */
3244 	if (!clean_complete)
3245 		return budget;
3246 
3247 	/* all work done, exit the polling mode */
3248 	napi_complete_done(napi, work_done);
3249 	if (adapter->rx_itr_setting & 1)
3250 		ixgbe_set_itr(q_vector);
3251 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3252 		ixgbe_irq_enable_queues(adapter, BIT_ULL(q_vector->v_idx));
3253 
3254 	return min(work_done, budget - 1);
3255 }
3256 
3257 /**
3258  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3259  * @adapter: board private structure
3260  *
3261  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3262  * interrupts from the kernel.
3263  **/
3264 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3265 {
3266 	struct net_device *netdev = adapter->netdev;
3267 	unsigned int ri = 0, ti = 0;
3268 	int vector, err;
3269 
3270 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3271 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3272 		struct msix_entry *entry = &adapter->msix_entries[vector];
3273 
3274 		if (q_vector->tx.ring && q_vector->rx.ring) {
3275 			snprintf(q_vector->name, sizeof(q_vector->name),
3276 				 "%s-TxRx-%u", netdev->name, ri++);
3277 			ti++;
3278 		} else if (q_vector->rx.ring) {
3279 			snprintf(q_vector->name, sizeof(q_vector->name),
3280 				 "%s-rx-%u", netdev->name, ri++);
3281 		} else if (q_vector->tx.ring) {
3282 			snprintf(q_vector->name, sizeof(q_vector->name),
3283 				 "%s-tx-%u", netdev->name, ti++);
3284 		} else {
3285 			/* skip this unused q_vector */
3286 			continue;
3287 		}
3288 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3289 				  q_vector->name, q_vector);
3290 		if (err) {
3291 			e_err(probe, "request_irq failed for MSIX interrupt "
3292 			      "Error: %d\n", err);
3293 			goto free_queue_irqs;
3294 		}
3295 		/* If Flow Director is enabled, set interrupt affinity */
3296 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3297 			/* assign the mask for this irq */
3298 			irq_set_affinity_hint(entry->vector,
3299 					      &q_vector->affinity_mask);
3300 		}
3301 	}
3302 
3303 	err = request_irq(adapter->msix_entries[vector].vector,
3304 			  ixgbe_msix_other, 0, netdev->name, adapter);
3305 	if (err) {
3306 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3307 		goto free_queue_irqs;
3308 	}
3309 
3310 	return 0;
3311 
3312 free_queue_irqs:
3313 	while (vector) {
3314 		vector--;
3315 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
3316 				      NULL);
3317 		free_irq(adapter->msix_entries[vector].vector,
3318 			 adapter->q_vector[vector]);
3319 	}
3320 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3321 	pci_disable_msix(adapter->pdev);
3322 	kfree(adapter->msix_entries);
3323 	adapter->msix_entries = NULL;
3324 	return err;
3325 }
3326 
3327 /**
3328  * ixgbe_intr - legacy mode Interrupt Handler
3329  * @irq: interrupt number
3330  * @data: pointer to a network interface device structure
3331  **/
3332 static irqreturn_t ixgbe_intr(int irq, void *data)
3333 {
3334 	struct ixgbe_adapter *adapter = data;
3335 	struct ixgbe_hw *hw = &adapter->hw;
3336 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3337 	u32 eicr;
3338 
3339 	/*
3340 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3341 	 * before the read of EICR.
3342 	 */
3343 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3344 
3345 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3346 	 * therefore no explicit interrupt disable is necessary */
3347 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3348 	if (!eicr) {
3349 		/*
3350 		 * shared interrupt alert!
3351 		 * make sure interrupts are enabled because the read will
3352 		 * have disabled interrupts due to EIAM
3353 		 * finish the workaround of silicon errata on 82598.  Unmask
3354 		 * the interrupt that we masked before the EICR read.
3355 		 */
3356 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3357 			ixgbe_irq_enable(adapter, true, true);
3358 		return IRQ_NONE;	/* Not our interrupt */
3359 	}
3360 
3361 	if (eicr & IXGBE_EICR_LSC)
3362 		ixgbe_check_lsc(adapter);
3363 
3364 	switch (hw->mac.type) {
3365 	case ixgbe_mac_82599EB:
3366 		ixgbe_check_sfp_event(adapter, eicr);
3367 		/* Fall through */
3368 	case ixgbe_mac_X540:
3369 	case ixgbe_mac_X550:
3370 	case ixgbe_mac_X550EM_x:
3371 	case ixgbe_mac_x550em_a:
3372 		if (eicr & IXGBE_EICR_ECC) {
3373 			e_info(link, "Received ECC Err, initiating reset\n");
3374 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3375 			ixgbe_service_event_schedule(adapter);
3376 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3377 		}
3378 		ixgbe_check_overtemp_event(adapter, eicr);
3379 		break;
3380 	default:
3381 		break;
3382 	}
3383 
3384 	ixgbe_check_fan_failure(adapter, eicr);
3385 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3386 		ixgbe_ptp_check_pps_event(adapter);
3387 
3388 	/* would disable interrupts here but EIAM disabled it */
3389 	napi_schedule_irqoff(&q_vector->napi);
3390 
3391 	/*
3392 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3393 	 * ixgbe_poll will re-enable the queue interrupts
3394 	 */
3395 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3396 		ixgbe_irq_enable(adapter, false, false);
3397 
3398 	return IRQ_HANDLED;
3399 }
3400 
3401 /**
3402  * ixgbe_request_irq - initialize interrupts
3403  * @adapter: board private structure
3404  *
3405  * Attempts to configure interrupts using the best available
3406  * capabilities of the hardware and kernel.
3407  **/
3408 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3409 {
3410 	struct net_device *netdev = adapter->netdev;
3411 	int err;
3412 
3413 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3414 		err = ixgbe_request_msix_irqs(adapter);
3415 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3416 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3417 				  netdev->name, adapter);
3418 	else
3419 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3420 				  netdev->name, adapter);
3421 
3422 	if (err)
3423 		e_err(probe, "request_irq failed, Error %d\n", err);
3424 
3425 	return err;
3426 }
3427 
3428 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3429 {
3430 	int vector;
3431 
3432 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3433 		free_irq(adapter->pdev->irq, adapter);
3434 		return;
3435 	}
3436 
3437 	if (!adapter->msix_entries)
3438 		return;
3439 
3440 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3441 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3442 		struct msix_entry *entry = &adapter->msix_entries[vector];
3443 
3444 		/* free only the irqs that were actually requested */
3445 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3446 			continue;
3447 
3448 		/* clear the affinity_mask in the IRQ descriptor */
3449 		irq_set_affinity_hint(entry->vector, NULL);
3450 
3451 		free_irq(entry->vector, q_vector);
3452 	}
3453 
3454 	free_irq(adapter->msix_entries[vector].vector, adapter);
3455 }
3456 
3457 /**
3458  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3459  * @adapter: board private structure
3460  **/
3461 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3462 {
3463 	switch (adapter->hw.mac.type) {
3464 	case ixgbe_mac_82598EB:
3465 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3466 		break;
3467 	case ixgbe_mac_82599EB:
3468 	case ixgbe_mac_X540:
3469 	case ixgbe_mac_X550:
3470 	case ixgbe_mac_X550EM_x:
3471 	case ixgbe_mac_x550em_a:
3472 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3473 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3474 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3475 		break;
3476 	default:
3477 		break;
3478 	}
3479 	IXGBE_WRITE_FLUSH(&adapter->hw);
3480 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3481 		int vector;
3482 
3483 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3484 			synchronize_irq(adapter->msix_entries[vector].vector);
3485 
3486 		synchronize_irq(adapter->msix_entries[vector++].vector);
3487 	} else {
3488 		synchronize_irq(adapter->pdev->irq);
3489 	}
3490 }
3491 
3492 /**
3493  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3494  * @adapter: board private structure
3495  *
3496  **/
3497 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3498 {
3499 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3500 
3501 	ixgbe_write_eitr(q_vector);
3502 
3503 	ixgbe_set_ivar(adapter, 0, 0, 0);
3504 	ixgbe_set_ivar(adapter, 1, 0, 0);
3505 
3506 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3507 }
3508 
3509 /**
3510  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3511  * @adapter: board private structure
3512  * @ring: structure containing ring specific data
3513  *
3514  * Configure the Tx descriptor ring after a reset.
3515  **/
3516 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3517 			     struct ixgbe_ring *ring)
3518 {
3519 	struct ixgbe_hw *hw = &adapter->hw;
3520 	u64 tdba = ring->dma;
3521 	int wait_loop = 10;
3522 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3523 	u8 reg_idx = ring->reg_idx;
3524 
3525 	/* disable queue to avoid issues while updating state */
3526 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3527 	IXGBE_WRITE_FLUSH(hw);
3528 
3529 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3530 			(tdba & DMA_BIT_MASK(32)));
3531 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3532 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3533 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3534 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3535 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3536 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3537 
3538 	/*
3539 	 * set WTHRESH to encourage burst writeback, it should not be set
3540 	 * higher than 1 when:
3541 	 * - ITR is 0 as it could cause false TX hangs
3542 	 * - ITR is set to > 100k int/sec and BQL is enabled
3543 	 *
3544 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3545 	 * to or less than the number of on chip descriptors, which is
3546 	 * currently 40.
3547 	 */
3548 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3549 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3550 	else
3551 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3552 
3553 	/*
3554 	 * Setting PTHRESH to 32 both improves performance
3555 	 * and avoids a TX hang with DFP enabled
3556 	 */
3557 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3558 		   32;		/* PTHRESH = 32 */
3559 
3560 	/* reinitialize flowdirector state */
3561 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3562 		ring->atr_sample_rate = adapter->atr_sample_rate;
3563 		ring->atr_count = 0;
3564 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3565 	} else {
3566 		ring->atr_sample_rate = 0;
3567 	}
3568 
3569 	/* initialize XPS */
3570 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3571 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3572 
3573 		if (q_vector)
3574 			netif_set_xps_queue(ring->netdev,
3575 					    &q_vector->affinity_mask,
3576 					    ring->queue_index);
3577 	}
3578 
3579 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3580 
3581 	/* reinitialize tx_buffer_info */
3582 	memset(ring->tx_buffer_info, 0,
3583 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3584 
3585 	/* enable queue */
3586 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3587 
3588 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3589 	if (hw->mac.type == ixgbe_mac_82598EB &&
3590 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3591 		return;
3592 
3593 	/* poll to verify queue is enabled */
3594 	do {
3595 		usleep_range(1000, 2000);
3596 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3597 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3598 	if (!wait_loop)
3599 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3600 }
3601 
3602 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3603 {
3604 	struct ixgbe_hw *hw = &adapter->hw;
3605 	u32 rttdcs, mtqc;
3606 	u8 tcs = adapter->hw_tcs;
3607 
3608 	if (hw->mac.type == ixgbe_mac_82598EB)
3609 		return;
3610 
3611 	/* disable the arbiter while setting MTQC */
3612 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3613 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3614 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3615 
3616 	/* set transmit pool layout */
3617 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3618 		mtqc = IXGBE_MTQC_VT_ENA;
3619 		if (tcs > 4)
3620 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3621 		else if (tcs > 1)
3622 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3623 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3624 			 IXGBE_82599_VMDQ_4Q_MASK)
3625 			mtqc |= IXGBE_MTQC_32VF;
3626 		else
3627 			mtqc |= IXGBE_MTQC_64VF;
3628 	} else {
3629 		if (tcs > 4)
3630 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3631 		else if (tcs > 1)
3632 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3633 		else
3634 			mtqc = IXGBE_MTQC_64Q_1PB;
3635 	}
3636 
3637 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3638 
3639 	/* Enable Security TX Buffer IFG for multiple pb */
3640 	if (tcs) {
3641 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3642 		sectx |= IXGBE_SECTX_DCB;
3643 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3644 	}
3645 
3646 	/* re-enable the arbiter */
3647 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3648 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3649 }
3650 
3651 /**
3652  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3653  * @adapter: board private structure
3654  *
3655  * Configure the Tx unit of the MAC after a reset.
3656  **/
3657 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3658 {
3659 	struct ixgbe_hw *hw = &adapter->hw;
3660 	u32 dmatxctl;
3661 	u32 i;
3662 
3663 	ixgbe_setup_mtqc(adapter);
3664 
3665 	if (hw->mac.type != ixgbe_mac_82598EB) {
3666 		/* DMATXCTL.EN must be before Tx queues are enabled */
3667 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3668 		dmatxctl |= IXGBE_DMATXCTL_TE;
3669 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3670 	}
3671 
3672 	/* Setup the HW Tx Head and Tail descriptor pointers */
3673 	for (i = 0; i < adapter->num_tx_queues; i++)
3674 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3675 	for (i = 0; i < adapter->num_xdp_queues; i++)
3676 		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3677 }
3678 
3679 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3680 				 struct ixgbe_ring *ring)
3681 {
3682 	struct ixgbe_hw *hw = &adapter->hw;
3683 	u8 reg_idx = ring->reg_idx;
3684 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3685 
3686 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3687 
3688 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3689 }
3690 
3691 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3692 				  struct ixgbe_ring *ring)
3693 {
3694 	struct ixgbe_hw *hw = &adapter->hw;
3695 	u8 reg_idx = ring->reg_idx;
3696 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3697 
3698 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3699 
3700 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3701 }
3702 
3703 #ifdef CONFIG_IXGBE_DCB
3704 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3705 #else
3706 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3707 #endif
3708 {
3709 	int i;
3710 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3711 
3712 	if (adapter->ixgbe_ieee_pfc)
3713 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3714 
3715 	/*
3716 	 * We should set the drop enable bit if:
3717 	 *  SR-IOV is enabled
3718 	 *   or
3719 	 *  Number of Rx queues > 1 and flow control is disabled
3720 	 *
3721 	 *  This allows us to avoid head of line blocking for security
3722 	 *  and performance reasons.
3723 	 */
3724 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3725 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3726 		for (i = 0; i < adapter->num_rx_queues; i++)
3727 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3728 	} else {
3729 		for (i = 0; i < adapter->num_rx_queues; i++)
3730 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3731 	}
3732 }
3733 
3734 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3735 
3736 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3737 				   struct ixgbe_ring *rx_ring)
3738 {
3739 	struct ixgbe_hw *hw = &adapter->hw;
3740 	u32 srrctl;
3741 	u8 reg_idx = rx_ring->reg_idx;
3742 
3743 	if (hw->mac.type == ixgbe_mac_82598EB) {
3744 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3745 
3746 		/*
3747 		 * if VMDq is not active we must program one srrctl register
3748 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3749 		 */
3750 		reg_idx &= mask;
3751 	}
3752 
3753 	/* configure header buffer length, needed for RSC */
3754 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3755 
3756 	/* configure the packet buffer length */
3757 	if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state))
3758 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3759 	else
3760 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3761 
3762 	/* configure descriptor type */
3763 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3764 
3765 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3766 }
3767 
3768 /**
3769  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3770  * @adapter: device handle
3771  *
3772  *  - 82598/82599/X540:     128
3773  *  - X550(non-SRIOV mode): 512
3774  *  - X550(SRIOV mode):     64
3775  */
3776 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3777 {
3778 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3779 		return 128;
3780 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3781 		return 64;
3782 	else
3783 		return 512;
3784 }
3785 
3786 /**
3787  * ixgbe_store_key - Write the RSS key to HW
3788  * @adapter: device handle
3789  *
3790  * Write the RSS key stored in adapter.rss_key to HW.
3791  */
3792 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3793 {
3794 	struct ixgbe_hw *hw = &adapter->hw;
3795 	int i;
3796 
3797 	for (i = 0; i < 10; i++)
3798 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3799 }
3800 
3801 /**
3802  * ixgbe_init_rss_key - Initialize adapter RSS key
3803  * @adapter: device handle
3804  *
3805  * Allocates and initializes the RSS key if it is not allocated.
3806  **/
3807 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3808 {
3809 	u32 *rss_key;
3810 
3811 	if (!adapter->rss_key) {
3812 		rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3813 		if (unlikely(!rss_key))
3814 			return -ENOMEM;
3815 
3816 		netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3817 		adapter->rss_key = rss_key;
3818 	}
3819 
3820 	return 0;
3821 }
3822 
3823 /**
3824  * ixgbe_store_reta - Write the RETA table to HW
3825  * @adapter: device handle
3826  *
3827  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3828  */
3829 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3830 {
3831 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3832 	struct ixgbe_hw *hw = &adapter->hw;
3833 	u32 reta = 0;
3834 	u32 indices_multi;
3835 	u8 *indir_tbl = adapter->rss_indir_tbl;
3836 
3837 	/* Fill out the redirection table as follows:
3838 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3839 	 *    indices.
3840 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3841 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3842 	 */
3843 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3844 		indices_multi = 0x11;
3845 	else
3846 		indices_multi = 0x1;
3847 
3848 	/* Write redirection table to HW */
3849 	for (i = 0; i < reta_entries; i++) {
3850 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3851 		if ((i & 3) == 3) {
3852 			if (i < 128)
3853 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3854 			else
3855 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3856 						reta);
3857 			reta = 0;
3858 		}
3859 	}
3860 }
3861 
3862 /**
3863  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3864  * @adapter: device handle
3865  *
3866  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3867  */
3868 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3869 {
3870 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3871 	struct ixgbe_hw *hw = &adapter->hw;
3872 	u32 vfreta = 0;
3873 
3874 	/* Write redirection table to HW */
3875 	for (i = 0; i < reta_entries; i++) {
3876 		u16 pool = adapter->num_rx_pools;
3877 
3878 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3879 		if ((i & 3) != 3)
3880 			continue;
3881 
3882 		while (pool--)
3883 			IXGBE_WRITE_REG(hw,
3884 					IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
3885 					vfreta);
3886 		vfreta = 0;
3887 	}
3888 }
3889 
3890 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3891 {
3892 	u32 i, j;
3893 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3894 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3895 
3896 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3897 	 * make full use of any rings they may have.  We will use the
3898 	 * PSRTYPE register to control how many rings we use within the PF.
3899 	 */
3900 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3901 		rss_i = 4;
3902 
3903 	/* Fill out hash function seeds */
3904 	ixgbe_store_key(adapter);
3905 
3906 	/* Fill out redirection table */
3907 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3908 
3909 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3910 		if (j == rss_i)
3911 			j = 0;
3912 
3913 		adapter->rss_indir_tbl[i] = j;
3914 	}
3915 
3916 	ixgbe_store_reta(adapter);
3917 }
3918 
3919 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3920 {
3921 	struct ixgbe_hw *hw = &adapter->hw;
3922 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3923 	int i, j;
3924 
3925 	/* Fill out hash function seeds */
3926 	for (i = 0; i < 10; i++) {
3927 		u16 pool = adapter->num_rx_pools;
3928 
3929 		while (pool--)
3930 			IXGBE_WRITE_REG(hw,
3931 					IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
3932 					*(adapter->rss_key + i));
3933 	}
3934 
3935 	/* Fill out the redirection table */
3936 	for (i = 0, j = 0; i < 64; i++, j++) {
3937 		if (j == rss_i)
3938 			j = 0;
3939 
3940 		adapter->rss_indir_tbl[i] = j;
3941 	}
3942 
3943 	ixgbe_store_vfreta(adapter);
3944 }
3945 
3946 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3947 {
3948 	struct ixgbe_hw *hw = &adapter->hw;
3949 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3950 	u32 rxcsum;
3951 
3952 	/* Disable indicating checksum in descriptor, enables RSS hash */
3953 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3954 	rxcsum |= IXGBE_RXCSUM_PCSD;
3955 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3956 
3957 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3958 		if (adapter->ring_feature[RING_F_RSS].mask)
3959 			mrqc = IXGBE_MRQC_RSSEN;
3960 	} else {
3961 		u8 tcs = adapter->hw_tcs;
3962 
3963 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3964 			if (tcs > 4)
3965 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3966 			else if (tcs > 1)
3967 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3968 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3969 				 IXGBE_82599_VMDQ_4Q_MASK)
3970 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3971 			else
3972 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3973 
3974 			/* Enable L3/L4 for Tx Switched packets */
3975 			mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3976 		} else {
3977 			if (tcs > 4)
3978 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3979 			else if (tcs > 1)
3980 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3981 			else
3982 				mrqc = IXGBE_MRQC_RSSEN;
3983 		}
3984 	}
3985 
3986 	/* Perform hash on these packet types */
3987 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3988 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3989 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3990 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3991 
3992 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3993 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3994 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3995 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3996 
3997 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3998 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3999 		u16 pool = adapter->num_rx_pools;
4000 
4001 		/* Enable VF RSS mode */
4002 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
4003 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4004 
4005 		/* Setup RSS through the VF registers */
4006 		ixgbe_setup_vfreta(adapter);
4007 		vfmrqc = IXGBE_MRQC_RSSEN;
4008 		vfmrqc |= rss_field;
4009 
4010 		while (pool--)
4011 			IXGBE_WRITE_REG(hw,
4012 					IXGBE_PFVFMRQC(VMDQ_P(pool)),
4013 					vfmrqc);
4014 	} else {
4015 		ixgbe_setup_reta(adapter);
4016 		mrqc |= rss_field;
4017 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4018 	}
4019 }
4020 
4021 /**
4022  * ixgbe_configure_rscctl - enable RSC for the indicated ring
4023  * @adapter: address of board private structure
4024  * @ring: structure containing ring specific data
4025  **/
4026 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
4027 				   struct ixgbe_ring *ring)
4028 {
4029 	struct ixgbe_hw *hw = &adapter->hw;
4030 	u32 rscctrl;
4031 	u8 reg_idx = ring->reg_idx;
4032 
4033 	if (!ring_is_rsc_enabled(ring))
4034 		return;
4035 
4036 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
4037 	rscctrl |= IXGBE_RSCCTL_RSCEN;
4038 	/*
4039 	 * we must limit the number of descriptors so that the
4040 	 * total size of max desc * buf_len is not greater
4041 	 * than 65536
4042 	 */
4043 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
4044 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
4045 }
4046 
4047 #define IXGBE_MAX_RX_DESC_POLL 10
4048 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4049 				       struct ixgbe_ring *ring)
4050 {
4051 	struct ixgbe_hw *hw = &adapter->hw;
4052 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4053 	u32 rxdctl;
4054 	u8 reg_idx = ring->reg_idx;
4055 
4056 	if (ixgbe_removed(hw->hw_addr))
4057 		return;
4058 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4059 	if (hw->mac.type == ixgbe_mac_82598EB &&
4060 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4061 		return;
4062 
4063 	do {
4064 		usleep_range(1000, 2000);
4065 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4066 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4067 
4068 	if (!wait_loop) {
4069 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4070 		      "the polling period\n", reg_idx);
4071 	}
4072 }
4073 
4074 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
4075 			    struct ixgbe_ring *ring)
4076 {
4077 	struct ixgbe_hw *hw = &adapter->hw;
4078 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4079 	u32 rxdctl;
4080 	u8 reg_idx = ring->reg_idx;
4081 
4082 	if (ixgbe_removed(hw->hw_addr))
4083 		return;
4084 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4085 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4086 
4087 	/* write value back with RXDCTL.ENABLE bit cleared */
4088 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4089 
4090 	if (hw->mac.type == ixgbe_mac_82598EB &&
4091 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4092 		return;
4093 
4094 	/* the hardware may take up to 100us to really disable the rx queue */
4095 	do {
4096 		udelay(10);
4097 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4098 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
4099 
4100 	if (!wait_loop) {
4101 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
4102 		      "the polling period\n", reg_idx);
4103 	}
4104 }
4105 
4106 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4107 			     struct ixgbe_ring *ring)
4108 {
4109 	struct ixgbe_hw *hw = &adapter->hw;
4110 	union ixgbe_adv_rx_desc *rx_desc;
4111 	u64 rdba = ring->dma;
4112 	u32 rxdctl;
4113 	u8 reg_idx = ring->reg_idx;
4114 
4115 	/* disable queue to avoid issues while updating state */
4116 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4117 	ixgbe_disable_rx_queue(adapter, ring);
4118 
4119 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4120 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4121 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4122 			ring->count * sizeof(union ixgbe_adv_rx_desc));
4123 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
4124 	IXGBE_WRITE_FLUSH(hw);
4125 
4126 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4127 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4128 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4129 
4130 	ixgbe_configure_srrctl(adapter, ring);
4131 	ixgbe_configure_rscctl(adapter, ring);
4132 
4133 	if (hw->mac.type == ixgbe_mac_82598EB) {
4134 		/*
4135 		 * enable cache line friendly hardware writes:
4136 		 * PTHRESH=32 descriptors (half the internal cache),
4137 		 * this also removes ugly rx_no_buffer_count increment
4138 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
4139 		 * WTHRESH=8 burst writeback up to two cache lines
4140 		 */
4141 		rxdctl &= ~0x3FFFFF;
4142 		rxdctl |=  0x080420;
4143 #if (PAGE_SIZE < 8192)
4144 	/* RXDCTL.RLPML does not work on 82599 */
4145 	} else if (hw->mac.type != ixgbe_mac_82599EB) {
4146 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4147 			    IXGBE_RXDCTL_RLPML_EN);
4148 
4149 		/* Limit the maximum frame size so we don't overrun the skb.
4150 		 * This can happen in SRIOV mode when the MTU of the VF is
4151 		 * higher than the MTU of the PF.
4152 		 */
4153 		if (ring_uses_build_skb(ring) &&
4154 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4155 			rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4156 				  IXGBE_RXDCTL_RLPML_EN;
4157 #endif
4158 	}
4159 
4160 	/* initialize rx_buffer_info */
4161 	memset(ring->rx_buffer_info, 0,
4162 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
4163 
4164 	/* initialize Rx descriptor 0 */
4165 	rx_desc = IXGBE_RX_DESC(ring, 0);
4166 	rx_desc->wb.upper.length = 0;
4167 
4168 	/* enable receive descriptor ring */
4169 	rxdctl |= IXGBE_RXDCTL_ENABLE;
4170 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4171 
4172 	ixgbe_rx_desc_queue_enable(adapter, ring);
4173 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4174 }
4175 
4176 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4177 {
4178 	struct ixgbe_hw *hw = &adapter->hw;
4179 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4180 	u16 pool = adapter->num_rx_pools;
4181 
4182 	/* PSRTYPE must be initialized in non 82598 adapters */
4183 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4184 		      IXGBE_PSRTYPE_UDPHDR |
4185 		      IXGBE_PSRTYPE_IPV4HDR |
4186 		      IXGBE_PSRTYPE_L2HDR |
4187 		      IXGBE_PSRTYPE_IPV6HDR;
4188 
4189 	if (hw->mac.type == ixgbe_mac_82598EB)
4190 		return;
4191 
4192 	if (rss_i > 3)
4193 		psrtype |= 2u << 29;
4194 	else if (rss_i > 1)
4195 		psrtype |= 1u << 29;
4196 
4197 	while (pool--)
4198 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4199 }
4200 
4201 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4202 {
4203 	struct ixgbe_hw *hw = &adapter->hw;
4204 	u32 reg_offset, vf_shift;
4205 	u32 gcr_ext, vmdctl;
4206 	int i;
4207 
4208 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4209 		return;
4210 
4211 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4212 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4213 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4214 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4215 	vmdctl |= IXGBE_VT_CTL_REPLEN;
4216 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4217 
4218 	vf_shift = VMDQ_P(0) % 32;
4219 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4220 
4221 	/* Enable only the PF's pool for Tx/Rx */
4222 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4223 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4224 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4225 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4226 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4227 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4228 
4229 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4230 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4231 
4232 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
4233 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4234 
4235 	/*
4236 	 * Set up VF register offsets for selected VT Mode,
4237 	 * i.e. 32 or 64 VFs for SR-IOV
4238 	 */
4239 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4240 	case IXGBE_82599_VMDQ_8Q_MASK:
4241 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4242 		break;
4243 	case IXGBE_82599_VMDQ_4Q_MASK:
4244 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4245 		break;
4246 	default:
4247 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4248 		break;
4249 	}
4250 
4251 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4252 
4253 	for (i = 0; i < adapter->num_vfs; i++) {
4254 		/* configure spoof checking */
4255 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4256 					  adapter->vfinfo[i].spoofchk_enabled);
4257 
4258 		/* Enable/Disable RSS query feature  */
4259 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4260 					  adapter->vfinfo[i].rss_query_enabled);
4261 	}
4262 }
4263 
4264 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4265 {
4266 	struct ixgbe_hw *hw = &adapter->hw;
4267 	struct net_device *netdev = adapter->netdev;
4268 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4269 	struct ixgbe_ring *rx_ring;
4270 	int i;
4271 	u32 mhadd, hlreg0;
4272 
4273 #ifdef IXGBE_FCOE
4274 	/* adjust max frame to be able to do baby jumbo for FCoE */
4275 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4276 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4277 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4278 
4279 #endif /* IXGBE_FCOE */
4280 
4281 	/* adjust max frame to be at least the size of a standard frame */
4282 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4283 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4284 
4285 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4286 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4287 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
4288 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4289 
4290 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4291 	}
4292 
4293 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4294 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4295 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4296 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4297 
4298 	/*
4299 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4300 	 * the Base and Length of the Rx Descriptor Ring
4301 	 */
4302 	for (i = 0; i < adapter->num_rx_queues; i++) {
4303 		rx_ring = adapter->rx_ring[i];
4304 
4305 		clear_ring_rsc_enabled(rx_ring);
4306 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4307 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4308 
4309 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4310 			set_ring_rsc_enabled(rx_ring);
4311 
4312 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4313 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4314 
4315 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4316 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4317 			continue;
4318 
4319 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4320 
4321 #if (PAGE_SIZE < 8192)
4322 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4323 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4324 
4325 		if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4326 		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4327 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4328 #endif
4329 	}
4330 }
4331 
4332 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4333 {
4334 	struct ixgbe_hw *hw = &adapter->hw;
4335 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4336 
4337 	switch (hw->mac.type) {
4338 	case ixgbe_mac_82598EB:
4339 		/*
4340 		 * For VMDq support of different descriptor types or
4341 		 * buffer sizes through the use of multiple SRRCTL
4342 		 * registers, RDRXCTL.MVMEN must be set to 1
4343 		 *
4344 		 * also, the manual doesn't mention it clearly but DCA hints
4345 		 * will only use queue 0's tags unless this bit is set.  Side
4346 		 * effects of setting this bit are only that SRRCTL must be
4347 		 * fully programmed [0..15]
4348 		 */
4349 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4350 		break;
4351 	case ixgbe_mac_X550:
4352 	case ixgbe_mac_X550EM_x:
4353 	case ixgbe_mac_x550em_a:
4354 		if (adapter->num_vfs)
4355 			rdrxctl |= IXGBE_RDRXCTL_PSP;
4356 		/* fall through */
4357 	case ixgbe_mac_82599EB:
4358 	case ixgbe_mac_X540:
4359 		/* Disable RSC for ACK packets */
4360 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4361 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4362 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4363 		/* hardware requires some bits to be set by default */
4364 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4365 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4366 		break;
4367 	default:
4368 		/* We should do nothing since we don't know this hardware */
4369 		return;
4370 	}
4371 
4372 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4373 }
4374 
4375 /**
4376  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4377  * @adapter: board private structure
4378  *
4379  * Configure the Rx unit of the MAC after a reset.
4380  **/
4381 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4382 {
4383 	struct ixgbe_hw *hw = &adapter->hw;
4384 	int i;
4385 	u32 rxctrl, rfctl;
4386 
4387 	/* disable receives while setting up the descriptors */
4388 	hw->mac.ops.disable_rx(hw);
4389 
4390 	ixgbe_setup_psrtype(adapter);
4391 	ixgbe_setup_rdrxctl(adapter);
4392 
4393 	/* RSC Setup */
4394 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4395 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4396 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4397 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4398 
4399 	/* disable NFS filtering */
4400 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4401 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4402 
4403 	/* Program registers for the distribution of queues */
4404 	ixgbe_setup_mrqc(adapter);
4405 
4406 	/* set_rx_buffer_len must be called before ring initialization */
4407 	ixgbe_set_rx_buffer_len(adapter);
4408 
4409 	/*
4410 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4411 	 * the Base and Length of the Rx Descriptor Ring
4412 	 */
4413 	for (i = 0; i < adapter->num_rx_queues; i++)
4414 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4415 
4416 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4417 	/* disable drop enable for 82598 parts */
4418 	if (hw->mac.type == ixgbe_mac_82598EB)
4419 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4420 
4421 	/* enable all receives */
4422 	rxctrl |= IXGBE_RXCTRL_RXEN;
4423 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4424 }
4425 
4426 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4427 				 __be16 proto, u16 vid)
4428 {
4429 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4430 	struct ixgbe_hw *hw = &adapter->hw;
4431 
4432 	/* add VID to filter table */
4433 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4434 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4435 
4436 	set_bit(vid, adapter->active_vlans);
4437 
4438 	return 0;
4439 }
4440 
4441 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4442 {
4443 	u32 vlvf;
4444 	int idx;
4445 
4446 	/* short cut the special case */
4447 	if (vlan == 0)
4448 		return 0;
4449 
4450 	/* Search for the vlan id in the VLVF entries */
4451 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4452 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4453 		if ((vlvf & VLAN_VID_MASK) == vlan)
4454 			break;
4455 	}
4456 
4457 	return idx;
4458 }
4459 
4460 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4461 {
4462 	struct ixgbe_hw *hw = &adapter->hw;
4463 	u32 bits, word;
4464 	int idx;
4465 
4466 	idx = ixgbe_find_vlvf_entry(hw, vid);
4467 	if (!idx)
4468 		return;
4469 
4470 	/* See if any other pools are set for this VLAN filter
4471 	 * entry other than the PF.
4472 	 */
4473 	word = idx * 2 + (VMDQ_P(0) / 32);
4474 	bits = ~BIT(VMDQ_P(0) % 32);
4475 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4476 
4477 	/* Disable the filter so this falls into the default pool. */
4478 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4479 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4480 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4481 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4482 	}
4483 }
4484 
4485 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4486 				  __be16 proto, u16 vid)
4487 {
4488 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4489 	struct ixgbe_hw *hw = &adapter->hw;
4490 
4491 	/* remove VID from filter table */
4492 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4493 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4494 
4495 	clear_bit(vid, adapter->active_vlans);
4496 
4497 	return 0;
4498 }
4499 
4500 /**
4501  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4502  * @adapter: driver data
4503  */
4504 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4505 {
4506 	struct ixgbe_hw *hw = &adapter->hw;
4507 	u32 vlnctrl;
4508 	int i, j;
4509 
4510 	switch (hw->mac.type) {
4511 	case ixgbe_mac_82598EB:
4512 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4513 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4514 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4515 		break;
4516 	case ixgbe_mac_82599EB:
4517 	case ixgbe_mac_X540:
4518 	case ixgbe_mac_X550:
4519 	case ixgbe_mac_X550EM_x:
4520 	case ixgbe_mac_x550em_a:
4521 		for (i = 0; i < adapter->num_rx_queues; i++) {
4522 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4523 
4524 			if (!netif_is_ixgbe(ring->netdev))
4525 				continue;
4526 
4527 			j = ring->reg_idx;
4528 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4529 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4530 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4531 		}
4532 		break;
4533 	default:
4534 		break;
4535 	}
4536 }
4537 
4538 /**
4539  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4540  * @adapter: driver data
4541  */
4542 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4543 {
4544 	struct ixgbe_hw *hw = &adapter->hw;
4545 	u32 vlnctrl;
4546 	int i, j;
4547 
4548 	switch (hw->mac.type) {
4549 	case ixgbe_mac_82598EB:
4550 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4551 		vlnctrl |= IXGBE_VLNCTRL_VME;
4552 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4553 		break;
4554 	case ixgbe_mac_82599EB:
4555 	case ixgbe_mac_X540:
4556 	case ixgbe_mac_X550:
4557 	case ixgbe_mac_X550EM_x:
4558 	case ixgbe_mac_x550em_a:
4559 		for (i = 0; i < adapter->num_rx_queues; i++) {
4560 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4561 
4562 			if (!netif_is_ixgbe(ring->netdev))
4563 				continue;
4564 
4565 			j = ring->reg_idx;
4566 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4567 			vlnctrl |= IXGBE_RXDCTL_VME;
4568 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4569 		}
4570 		break;
4571 	default:
4572 		break;
4573 	}
4574 }
4575 
4576 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4577 {
4578 	struct ixgbe_hw *hw = &adapter->hw;
4579 	u32 vlnctrl, i;
4580 
4581 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4582 
4583 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4584 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4585 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4586 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4587 	} else {
4588 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4589 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4590 		return;
4591 	}
4592 
4593 	/* Nothing to do for 82598 */
4594 	if (hw->mac.type == ixgbe_mac_82598EB)
4595 		return;
4596 
4597 	/* We are already in VLAN promisc, nothing to do */
4598 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4599 		return;
4600 
4601 	/* Set flag so we don't redo unnecessary work */
4602 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4603 
4604 	/* Add PF to all active pools */
4605 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4606 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4607 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4608 
4609 		vlvfb |= BIT(VMDQ_P(0) % 32);
4610 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4611 	}
4612 
4613 	/* Set all bits in the VLAN filter table array */
4614 	for (i = hw->mac.vft_size; i--;)
4615 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4616 }
4617 
4618 #define VFTA_BLOCK_SIZE 8
4619 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4620 {
4621 	struct ixgbe_hw *hw = &adapter->hw;
4622 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4623 	u32 vid_start = vfta_offset * 32;
4624 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4625 	u32 i, vid, word, bits;
4626 
4627 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4628 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4629 
4630 		/* pull VLAN ID from VLVF */
4631 		vid = vlvf & VLAN_VID_MASK;
4632 
4633 		/* only concern outselves with a certain range */
4634 		if (vid < vid_start || vid >= vid_end)
4635 			continue;
4636 
4637 		if (vlvf) {
4638 			/* record VLAN ID in VFTA */
4639 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4640 
4641 			/* if PF is part of this then continue */
4642 			if (test_bit(vid, adapter->active_vlans))
4643 				continue;
4644 		}
4645 
4646 		/* remove PF from the pool */
4647 		word = i * 2 + VMDQ_P(0) / 32;
4648 		bits = ~BIT(VMDQ_P(0) % 32);
4649 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4650 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4651 	}
4652 
4653 	/* extract values from active_vlans and write back to VFTA */
4654 	for (i = VFTA_BLOCK_SIZE; i--;) {
4655 		vid = (vfta_offset + i) * 32;
4656 		word = vid / BITS_PER_LONG;
4657 		bits = vid % BITS_PER_LONG;
4658 
4659 		vfta[i] |= adapter->active_vlans[word] >> bits;
4660 
4661 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4662 	}
4663 }
4664 
4665 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4666 {
4667 	struct ixgbe_hw *hw = &adapter->hw;
4668 	u32 vlnctrl, i;
4669 
4670 	/* Set VLAN filtering to enabled */
4671 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4672 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4673 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4674 
4675 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4676 	    hw->mac.type == ixgbe_mac_82598EB)
4677 		return;
4678 
4679 	/* We are not in VLAN promisc, nothing to do */
4680 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4681 		return;
4682 
4683 	/* Set flag so we don't redo unnecessary work */
4684 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4685 
4686 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4687 		ixgbe_scrub_vfta(adapter, i);
4688 }
4689 
4690 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4691 {
4692 	u16 vid = 1;
4693 
4694 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4695 
4696 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4697 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4698 }
4699 
4700 /**
4701  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4702  * @netdev: network interface device structure
4703  *
4704  * Writes multicast address list to the MTA hash table.
4705  * Returns: -ENOMEM on failure
4706  *                0 on no addresses written
4707  *                X on writing X addresses to MTA
4708  **/
4709 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4710 {
4711 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4712 	struct ixgbe_hw *hw = &adapter->hw;
4713 
4714 	if (!netif_running(netdev))
4715 		return 0;
4716 
4717 	if (hw->mac.ops.update_mc_addr_list)
4718 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4719 	else
4720 		return -ENOMEM;
4721 
4722 #ifdef CONFIG_PCI_IOV
4723 	ixgbe_restore_vf_multicasts(adapter);
4724 #endif
4725 
4726 	return netdev_mc_count(netdev);
4727 }
4728 
4729 #ifdef CONFIG_PCI_IOV
4730 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4731 {
4732 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4733 	struct ixgbe_hw *hw = &adapter->hw;
4734 	int i;
4735 
4736 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4737 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4738 
4739 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4740 			hw->mac.ops.set_rar(hw, i,
4741 					    mac_table->addr,
4742 					    mac_table->pool,
4743 					    IXGBE_RAH_AV);
4744 		else
4745 			hw->mac.ops.clear_rar(hw, i);
4746 	}
4747 }
4748 
4749 #endif
4750 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4751 {
4752 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4753 	struct ixgbe_hw *hw = &adapter->hw;
4754 	int i;
4755 
4756 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4757 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4758 			continue;
4759 
4760 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4761 
4762 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4763 			hw->mac.ops.set_rar(hw, i,
4764 					    mac_table->addr,
4765 					    mac_table->pool,
4766 					    IXGBE_RAH_AV);
4767 		else
4768 			hw->mac.ops.clear_rar(hw, i);
4769 	}
4770 }
4771 
4772 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4773 {
4774 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4775 	struct ixgbe_hw *hw = &adapter->hw;
4776 	int i;
4777 
4778 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4779 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4780 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4781 	}
4782 
4783 	ixgbe_sync_mac_table(adapter);
4784 }
4785 
4786 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4787 {
4788 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4789 	struct ixgbe_hw *hw = &adapter->hw;
4790 	int i, count = 0;
4791 
4792 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4793 		/* do not count default RAR as available */
4794 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4795 			continue;
4796 
4797 		/* only count unused and addresses that belong to us */
4798 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4799 			if (mac_table->pool != pool)
4800 				continue;
4801 		}
4802 
4803 		count++;
4804 	}
4805 
4806 	return count;
4807 }
4808 
4809 /* this function destroys the first RAR entry */
4810 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4811 {
4812 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4813 	struct ixgbe_hw *hw = &adapter->hw;
4814 
4815 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4816 	mac_table->pool = VMDQ_P(0);
4817 
4818 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4819 
4820 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4821 			    IXGBE_RAH_AV);
4822 }
4823 
4824 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4825 			 const u8 *addr, u16 pool)
4826 {
4827 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4828 	struct ixgbe_hw *hw = &adapter->hw;
4829 	int i;
4830 
4831 	if (is_zero_ether_addr(addr))
4832 		return -EINVAL;
4833 
4834 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4835 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4836 			continue;
4837 
4838 		ether_addr_copy(mac_table->addr, addr);
4839 		mac_table->pool = pool;
4840 
4841 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4842 				    IXGBE_MAC_STATE_IN_USE;
4843 
4844 		ixgbe_sync_mac_table(adapter);
4845 
4846 		return i;
4847 	}
4848 
4849 	return -ENOMEM;
4850 }
4851 
4852 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4853 			 const u8 *addr, u16 pool)
4854 {
4855 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4856 	struct ixgbe_hw *hw = &adapter->hw;
4857 	int i;
4858 
4859 	if (is_zero_ether_addr(addr))
4860 		return -EINVAL;
4861 
4862 	/* search table for addr, if found clear IN_USE flag and sync */
4863 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4864 		/* we can only delete an entry if it is in use */
4865 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4866 			continue;
4867 		/* we only care about entries that belong to the given pool */
4868 		if (mac_table->pool != pool)
4869 			continue;
4870 		/* we only care about a specific MAC address */
4871 		if (!ether_addr_equal(addr, mac_table->addr))
4872 			continue;
4873 
4874 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4875 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4876 
4877 		ixgbe_sync_mac_table(adapter);
4878 
4879 		return 0;
4880 	}
4881 
4882 	return -ENOMEM;
4883 }
4884 
4885 /**
4886  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4887  * @netdev: network interface device structure
4888  * @vfn: pool to associate with unicast addresses
4889  *
4890  * Writes unicast address list to the RAR table.
4891  * Returns: -ENOMEM on failure/insufficient address space
4892  *                0 on no addresses written
4893  *                X on writing X addresses to the RAR table
4894  **/
4895 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4896 {
4897 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4898 	int count = 0;
4899 
4900 	/* return ENOMEM indicating insufficient memory for addresses */
4901 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter, vfn))
4902 		return -ENOMEM;
4903 
4904 	if (!netdev_uc_empty(netdev)) {
4905 		struct netdev_hw_addr *ha;
4906 		netdev_for_each_uc_addr(ha, netdev) {
4907 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4908 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4909 			count++;
4910 		}
4911 	}
4912 	return count;
4913 }
4914 
4915 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4916 {
4917 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4918 	int ret;
4919 
4920 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4921 
4922 	return min_t(int, ret, 0);
4923 }
4924 
4925 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4926 {
4927 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4928 
4929 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4930 
4931 	return 0;
4932 }
4933 
4934 /**
4935  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4936  * @netdev: network interface device structure
4937  *
4938  * The set_rx_method entry point is called whenever the unicast/multicast
4939  * address list or the network interface flags are updated.  This routine is
4940  * responsible for configuring the hardware for proper unicast, multicast and
4941  * promiscuous mode.
4942  **/
4943 void ixgbe_set_rx_mode(struct net_device *netdev)
4944 {
4945 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4946 	struct ixgbe_hw *hw = &adapter->hw;
4947 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4948 	netdev_features_t features = netdev->features;
4949 	int count;
4950 
4951 	/* Check for Promiscuous and All Multicast modes */
4952 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4953 
4954 	/* set all bits that we expect to always be set */
4955 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4956 	fctrl |= IXGBE_FCTRL_BAM;
4957 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4958 	fctrl |= IXGBE_FCTRL_PMCF;
4959 
4960 	/* clear the bits we are changing the status of */
4961 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4962 	if (netdev->flags & IFF_PROMISC) {
4963 		hw->addr_ctrl.user_set_promisc = true;
4964 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4965 		vmolr |= IXGBE_VMOLR_MPE;
4966 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4967 	} else {
4968 		if (netdev->flags & IFF_ALLMULTI) {
4969 			fctrl |= IXGBE_FCTRL_MPE;
4970 			vmolr |= IXGBE_VMOLR_MPE;
4971 		}
4972 		hw->addr_ctrl.user_set_promisc = false;
4973 	}
4974 
4975 	/*
4976 	 * Write addresses to available RAR registers, if there is not
4977 	 * sufficient space to store all the addresses then enable
4978 	 * unicast promiscuous mode
4979 	 */
4980 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4981 		fctrl |= IXGBE_FCTRL_UPE;
4982 		vmolr |= IXGBE_VMOLR_ROPE;
4983 	}
4984 
4985 	/* Write addresses to the MTA, if the attempt fails
4986 	 * then we should just turn on promiscuous mode so
4987 	 * that we can at least receive multicast traffic
4988 	 */
4989 	count = ixgbe_write_mc_addr_list(netdev);
4990 	if (count < 0) {
4991 		fctrl |= IXGBE_FCTRL_MPE;
4992 		vmolr |= IXGBE_VMOLR_MPE;
4993 	} else if (count) {
4994 		vmolr |= IXGBE_VMOLR_ROMPE;
4995 	}
4996 
4997 	if (hw->mac.type != ixgbe_mac_82598EB) {
4998 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4999 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
5000 			   IXGBE_VMOLR_ROPE);
5001 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
5002 	}
5003 
5004 	/* This is useful for sniffing bad packets. */
5005 	if (features & NETIF_F_RXALL) {
5006 		/* UPE and MPE will be handled by normal PROMISC logic
5007 		 * in e1000e_set_rx_mode */
5008 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
5009 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
5010 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
5011 
5012 		fctrl &= ~(IXGBE_FCTRL_DPF);
5013 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
5014 	}
5015 
5016 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5017 
5018 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
5019 		ixgbe_vlan_strip_enable(adapter);
5020 	else
5021 		ixgbe_vlan_strip_disable(adapter);
5022 
5023 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
5024 		ixgbe_vlan_promisc_disable(adapter);
5025 	else
5026 		ixgbe_vlan_promisc_enable(adapter);
5027 }
5028 
5029 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
5030 {
5031 	int q_idx;
5032 
5033 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5034 		napi_enable(&adapter->q_vector[q_idx]->napi);
5035 }
5036 
5037 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
5038 {
5039 	int q_idx;
5040 
5041 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5042 		napi_disable(&adapter->q_vector[q_idx]->napi);
5043 }
5044 
5045 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
5046 {
5047 	struct ixgbe_hw *hw = &adapter->hw;
5048 	u32 vxlanctrl;
5049 
5050 	if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
5051 				IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
5052 		return;
5053 
5054 	vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) & ~mask;
5055 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
5056 
5057 	if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
5058 		adapter->vxlan_port = 0;
5059 
5060 	if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
5061 		adapter->geneve_port = 0;
5062 }
5063 
5064 #ifdef CONFIG_IXGBE_DCB
5065 /**
5066  * ixgbe_configure_dcb - Configure DCB hardware
5067  * @adapter: ixgbe adapter struct
5068  *
5069  * This is called by the driver on open to configure the DCB hardware.
5070  * This is also called by the gennetlink interface when reconfiguring
5071  * the DCB state.
5072  */
5073 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5074 {
5075 	struct ixgbe_hw *hw = &adapter->hw;
5076 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5077 
5078 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5079 		if (hw->mac.type == ixgbe_mac_82598EB)
5080 			netif_set_gso_max_size(adapter->netdev, 65536);
5081 		return;
5082 	}
5083 
5084 	if (hw->mac.type == ixgbe_mac_82598EB)
5085 		netif_set_gso_max_size(adapter->netdev, 32768);
5086 
5087 #ifdef IXGBE_FCOE
5088 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
5089 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5090 #endif
5091 
5092 	/* reconfigure the hardware */
5093 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5094 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5095 						DCB_TX_CONFIG);
5096 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5097 						DCB_RX_CONFIG);
5098 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5099 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5100 		ixgbe_dcb_hw_ets(&adapter->hw,
5101 				 adapter->ixgbe_ieee_ets,
5102 				 max_frame);
5103 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
5104 					adapter->ixgbe_ieee_pfc->pfc_en,
5105 					adapter->ixgbe_ieee_ets->prio_tc);
5106 	}
5107 
5108 	/* Enable RSS Hash per TC */
5109 	if (hw->mac.type != ixgbe_mac_82598EB) {
5110 		u32 msb = 0;
5111 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5112 
5113 		while (rss_i) {
5114 			msb++;
5115 			rss_i >>= 1;
5116 		}
5117 
5118 		/* write msb to all 8 TCs in one write */
5119 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5120 	}
5121 }
5122 #endif
5123 
5124 /* Additional bittime to account for IXGBE framing */
5125 #define IXGBE_ETH_FRAMING 20
5126 
5127 /**
5128  * ixgbe_hpbthresh - calculate high water mark for flow control
5129  *
5130  * @adapter: board private structure to calculate for
5131  * @pb: packet buffer to calculate
5132  */
5133 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5134 {
5135 	struct ixgbe_hw *hw = &adapter->hw;
5136 	struct net_device *dev = adapter->netdev;
5137 	int link, tc, kb, marker;
5138 	u32 dv_id, rx_pba;
5139 
5140 	/* Calculate max LAN frame size */
5141 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5142 
5143 #ifdef IXGBE_FCOE
5144 	/* FCoE traffic class uses FCOE jumbo frames */
5145 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5146 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5147 	    (pb == ixgbe_fcoe_get_tc(adapter)))
5148 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5149 #endif
5150 
5151 	/* Calculate delay value for device */
5152 	switch (hw->mac.type) {
5153 	case ixgbe_mac_X540:
5154 	case ixgbe_mac_X550:
5155 	case ixgbe_mac_X550EM_x:
5156 	case ixgbe_mac_x550em_a:
5157 		dv_id = IXGBE_DV_X540(link, tc);
5158 		break;
5159 	default:
5160 		dv_id = IXGBE_DV(link, tc);
5161 		break;
5162 	}
5163 
5164 	/* Loopback switch introduces additional latency */
5165 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5166 		dv_id += IXGBE_B2BT(tc);
5167 
5168 	/* Delay value is calculated in bit times convert to KB */
5169 	kb = IXGBE_BT2KB(dv_id);
5170 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5171 
5172 	marker = rx_pba - kb;
5173 
5174 	/* It is possible that the packet buffer is not large enough
5175 	 * to provide required headroom. In this case throw an error
5176 	 * to user and a do the best we can.
5177 	 */
5178 	if (marker < 0) {
5179 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
5180 			    "headroom to support flow control."
5181 			    "Decrease MTU or number of traffic classes\n", pb);
5182 		marker = tc + 1;
5183 	}
5184 
5185 	return marker;
5186 }
5187 
5188 /**
5189  * ixgbe_lpbthresh - calculate low water mark for for flow control
5190  *
5191  * @adapter: board private structure to calculate for
5192  * @pb: packet buffer to calculate
5193  */
5194 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5195 {
5196 	struct ixgbe_hw *hw = &adapter->hw;
5197 	struct net_device *dev = adapter->netdev;
5198 	int tc;
5199 	u32 dv_id;
5200 
5201 	/* Calculate max LAN frame size */
5202 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5203 
5204 #ifdef IXGBE_FCOE
5205 	/* FCoE traffic class uses FCOE jumbo frames */
5206 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5207 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5208 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5209 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5210 #endif
5211 
5212 	/* Calculate delay value for device */
5213 	switch (hw->mac.type) {
5214 	case ixgbe_mac_X540:
5215 	case ixgbe_mac_X550:
5216 	case ixgbe_mac_X550EM_x:
5217 	case ixgbe_mac_x550em_a:
5218 		dv_id = IXGBE_LOW_DV_X540(tc);
5219 		break;
5220 	default:
5221 		dv_id = IXGBE_LOW_DV(tc);
5222 		break;
5223 	}
5224 
5225 	/* Delay value is calculated in bit times convert to KB */
5226 	return IXGBE_BT2KB(dv_id);
5227 }
5228 
5229 /*
5230  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5231  */
5232 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5233 {
5234 	struct ixgbe_hw *hw = &adapter->hw;
5235 	int num_tc = adapter->hw_tcs;
5236 	int i;
5237 
5238 	if (!num_tc)
5239 		num_tc = 1;
5240 
5241 	for (i = 0; i < num_tc; i++) {
5242 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5243 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5244 
5245 		/* Low water marks must not be larger than high water marks */
5246 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
5247 			hw->fc.low_water[i] = 0;
5248 	}
5249 
5250 	for (; i < MAX_TRAFFIC_CLASS; i++)
5251 		hw->fc.high_water[i] = 0;
5252 }
5253 
5254 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5255 {
5256 	struct ixgbe_hw *hw = &adapter->hw;
5257 	int hdrm;
5258 	u8 tc = adapter->hw_tcs;
5259 
5260 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5261 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5262 		hdrm = 32 << adapter->fdir_pballoc;
5263 	else
5264 		hdrm = 0;
5265 
5266 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5267 	ixgbe_pbthresh_setup(adapter);
5268 }
5269 
5270 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5271 {
5272 	struct ixgbe_hw *hw = &adapter->hw;
5273 	struct hlist_node *node2;
5274 	struct ixgbe_fdir_filter *filter;
5275 
5276 	spin_lock(&adapter->fdir_perfect_lock);
5277 
5278 	if (!hlist_empty(&adapter->fdir_filter_list))
5279 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5280 
5281 	hlist_for_each_entry_safe(filter, node2,
5282 				  &adapter->fdir_filter_list, fdir_node) {
5283 		ixgbe_fdir_write_perfect_filter_82599(hw,
5284 				&filter->filter,
5285 				filter->sw_idx,
5286 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
5287 				IXGBE_FDIR_DROP_QUEUE :
5288 				adapter->rx_ring[filter->action]->reg_idx);
5289 	}
5290 
5291 	spin_unlock(&adapter->fdir_perfect_lock);
5292 }
5293 
5294 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
5295 				      struct ixgbe_adapter *adapter)
5296 {
5297 	struct ixgbe_hw *hw = &adapter->hw;
5298 	u32 vmolr;
5299 
5300 	/* No unicast promiscuous support for VMDQ devices. */
5301 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
5302 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
5303 
5304 	/* clear the affected bit */
5305 	vmolr &= ~IXGBE_VMOLR_MPE;
5306 
5307 	if (dev->flags & IFF_ALLMULTI) {
5308 		vmolr |= IXGBE_VMOLR_MPE;
5309 	} else {
5310 		vmolr |= IXGBE_VMOLR_ROMPE;
5311 		hw->mac.ops.update_mc_addr_list(hw, dev);
5312 	}
5313 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
5314 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
5315 }
5316 
5317 /**
5318  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5319  * @rx_ring: ring to free buffers from
5320  **/
5321 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5322 {
5323 	u16 i = rx_ring->next_to_clean;
5324 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5325 
5326 	/* Free all the Rx ring sk_buffs */
5327 	while (i != rx_ring->next_to_alloc) {
5328 		if (rx_buffer->skb) {
5329 			struct sk_buff *skb = rx_buffer->skb;
5330 			if (IXGBE_CB(skb)->page_released)
5331 				dma_unmap_page_attrs(rx_ring->dev,
5332 						     IXGBE_CB(skb)->dma,
5333 						     ixgbe_rx_pg_size(rx_ring),
5334 						     DMA_FROM_DEVICE,
5335 						     IXGBE_RX_DMA_ATTR);
5336 			dev_kfree_skb(skb);
5337 		}
5338 
5339 		/* Invalidate cache lines that may have been written to by
5340 		 * device so that we avoid corrupting memory.
5341 		 */
5342 		dma_sync_single_range_for_cpu(rx_ring->dev,
5343 					      rx_buffer->dma,
5344 					      rx_buffer->page_offset,
5345 					      ixgbe_rx_bufsz(rx_ring),
5346 					      DMA_FROM_DEVICE);
5347 
5348 		/* free resources associated with mapping */
5349 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5350 				     ixgbe_rx_pg_size(rx_ring),
5351 				     DMA_FROM_DEVICE,
5352 				     IXGBE_RX_DMA_ATTR);
5353 		__page_frag_cache_drain(rx_buffer->page,
5354 					rx_buffer->pagecnt_bias);
5355 
5356 		i++;
5357 		rx_buffer++;
5358 		if (i == rx_ring->count) {
5359 			i = 0;
5360 			rx_buffer = rx_ring->rx_buffer_info;
5361 		}
5362 	}
5363 
5364 	rx_ring->next_to_alloc = 0;
5365 	rx_ring->next_to_clean = 0;
5366 	rx_ring->next_to_use = 0;
5367 }
5368 
5369 static int ixgbe_fwd_ring_up(struct net_device *vdev,
5370 			     struct ixgbe_fwd_adapter *accel)
5371 {
5372 	struct ixgbe_adapter *adapter = accel->real_adapter;
5373 	int i, baseq, err;
5374 
5375 	if (!test_bit(accel->pool, adapter->fwd_bitmask))
5376 		return 0;
5377 
5378 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5379 	netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5380 		   accel->pool, adapter->num_rx_pools,
5381 		   baseq, baseq + adapter->num_rx_queues_per_pool);
5382 
5383 	accel->netdev = vdev;
5384 	accel->rx_base_queue = baseq;
5385 	accel->tx_base_queue = baseq;
5386 
5387 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5388 		adapter->rx_ring[baseq + i]->netdev = vdev;
5389 
5390 	/* Guarantee all rings are updated before we update the
5391 	 * MAC address filter.
5392 	 */
5393 	wmb();
5394 
5395 	/* ixgbe_add_mac_filter will return an index if it succeeds, so we
5396 	 * need to only treat it as an error value if it is negative.
5397 	 */
5398 	err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5399 				   VMDQ_P(accel->pool));
5400 	if (err >= 0) {
5401 		ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
5402 		return 0;
5403 	}
5404 
5405 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5406 		adapter->rx_ring[baseq + i]->netdev = NULL;
5407 
5408 	return err;
5409 }
5410 
5411 static int ixgbe_upper_dev_walk(struct net_device *upper, void *data)
5412 {
5413 	if (netif_is_macvlan(upper)) {
5414 		struct macvlan_dev *dfwd = netdev_priv(upper);
5415 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
5416 
5417 		if (dfwd->fwd_priv)
5418 			ixgbe_fwd_ring_up(upper, vadapter);
5419 	}
5420 
5421 	return 0;
5422 }
5423 
5424 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5425 {
5426 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5427 				      ixgbe_upper_dev_walk, NULL);
5428 }
5429 
5430 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5431 {
5432 	struct ixgbe_hw *hw = &adapter->hw;
5433 
5434 	ixgbe_configure_pb(adapter);
5435 #ifdef CONFIG_IXGBE_DCB
5436 	ixgbe_configure_dcb(adapter);
5437 #endif
5438 	/*
5439 	 * We must restore virtualization before VLANs or else
5440 	 * the VLVF registers will not be populated
5441 	 */
5442 	ixgbe_configure_virtualization(adapter);
5443 
5444 	ixgbe_set_rx_mode(adapter->netdev);
5445 	ixgbe_restore_vlan(adapter);
5446 	ixgbe_ipsec_restore(adapter);
5447 
5448 	switch (hw->mac.type) {
5449 	case ixgbe_mac_82599EB:
5450 	case ixgbe_mac_X540:
5451 		hw->mac.ops.disable_rx_buff(hw);
5452 		break;
5453 	default:
5454 		break;
5455 	}
5456 
5457 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5458 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5459 						adapter->fdir_pballoc);
5460 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5461 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5462 					      adapter->fdir_pballoc);
5463 		ixgbe_fdir_filter_restore(adapter);
5464 	}
5465 
5466 	switch (hw->mac.type) {
5467 	case ixgbe_mac_82599EB:
5468 	case ixgbe_mac_X540:
5469 		hw->mac.ops.enable_rx_buff(hw);
5470 		break;
5471 	default:
5472 		break;
5473 	}
5474 
5475 #ifdef CONFIG_IXGBE_DCA
5476 	/* configure DCA */
5477 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5478 		ixgbe_setup_dca(adapter);
5479 #endif /* CONFIG_IXGBE_DCA */
5480 
5481 #ifdef IXGBE_FCOE
5482 	/* configure FCoE L2 filters, redirection table, and Rx control */
5483 	ixgbe_configure_fcoe(adapter);
5484 
5485 #endif /* IXGBE_FCOE */
5486 	ixgbe_configure_tx(adapter);
5487 	ixgbe_configure_rx(adapter);
5488 	ixgbe_configure_dfwd(adapter);
5489 }
5490 
5491 /**
5492  * ixgbe_sfp_link_config - set up SFP+ link
5493  * @adapter: pointer to private adapter struct
5494  **/
5495 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5496 {
5497 	/*
5498 	 * We are assuming the worst case scenario here, and that
5499 	 * is that an SFP was inserted/removed after the reset
5500 	 * but before SFP detection was enabled.  As such the best
5501 	 * solution is to just start searching as soon as we start
5502 	 */
5503 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5504 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5505 
5506 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5507 	adapter->sfp_poll_time = 0;
5508 }
5509 
5510 /**
5511  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5512  * @hw: pointer to private hardware struct
5513  *
5514  * Returns 0 on success, negative on failure
5515  **/
5516 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5517 {
5518 	u32 speed;
5519 	bool autoneg, link_up = false;
5520 	int ret = IXGBE_ERR_LINK_SETUP;
5521 
5522 	if (hw->mac.ops.check_link)
5523 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5524 
5525 	if (ret)
5526 		return ret;
5527 
5528 	speed = hw->phy.autoneg_advertised;
5529 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5530 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5531 							&autoneg);
5532 	if (ret)
5533 		return ret;
5534 
5535 	if (hw->mac.ops.setup_link)
5536 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5537 
5538 	return ret;
5539 }
5540 
5541 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5542 {
5543 	struct ixgbe_hw *hw = &adapter->hw;
5544 	u32 gpie = 0;
5545 
5546 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5547 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5548 		       IXGBE_GPIE_OCD;
5549 		gpie |= IXGBE_GPIE_EIAME;
5550 		/*
5551 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5552 		 * this saves a register write for every interrupt
5553 		 */
5554 		switch (hw->mac.type) {
5555 		case ixgbe_mac_82598EB:
5556 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5557 			break;
5558 		case ixgbe_mac_82599EB:
5559 		case ixgbe_mac_X540:
5560 		case ixgbe_mac_X550:
5561 		case ixgbe_mac_X550EM_x:
5562 		case ixgbe_mac_x550em_a:
5563 		default:
5564 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5565 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5566 			break;
5567 		}
5568 	} else {
5569 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5570 		 * specifically only auto mask tx and rx interrupts */
5571 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5572 	}
5573 
5574 	/* XXX: to interrupt immediately for EICS writes, enable this */
5575 	/* gpie |= IXGBE_GPIE_EIMEN; */
5576 
5577 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5578 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5579 
5580 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5581 		case IXGBE_82599_VMDQ_8Q_MASK:
5582 			gpie |= IXGBE_GPIE_VTMODE_16;
5583 			break;
5584 		case IXGBE_82599_VMDQ_4Q_MASK:
5585 			gpie |= IXGBE_GPIE_VTMODE_32;
5586 			break;
5587 		default:
5588 			gpie |= IXGBE_GPIE_VTMODE_64;
5589 			break;
5590 		}
5591 	}
5592 
5593 	/* Enable Thermal over heat sensor interrupt */
5594 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5595 		switch (adapter->hw.mac.type) {
5596 		case ixgbe_mac_82599EB:
5597 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5598 			break;
5599 		default:
5600 			break;
5601 		}
5602 	}
5603 
5604 	/* Enable fan failure interrupt */
5605 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5606 		gpie |= IXGBE_SDP1_GPIEN(hw);
5607 
5608 	switch (hw->mac.type) {
5609 	case ixgbe_mac_82599EB:
5610 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5611 		break;
5612 	case ixgbe_mac_X550EM_x:
5613 	case ixgbe_mac_x550em_a:
5614 		gpie |= IXGBE_SDP0_GPIEN_X540;
5615 		break;
5616 	default:
5617 		break;
5618 	}
5619 
5620 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5621 }
5622 
5623 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5624 {
5625 	struct ixgbe_hw *hw = &adapter->hw;
5626 	int err;
5627 	u32 ctrl_ext;
5628 
5629 	ixgbe_get_hw_control(adapter);
5630 	ixgbe_setup_gpie(adapter);
5631 
5632 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5633 		ixgbe_configure_msix(adapter);
5634 	else
5635 		ixgbe_configure_msi_and_legacy(adapter);
5636 
5637 	/* enable the optics for 82599 SFP+ fiber */
5638 	if (hw->mac.ops.enable_tx_laser)
5639 		hw->mac.ops.enable_tx_laser(hw);
5640 
5641 	if (hw->phy.ops.set_phy_power)
5642 		hw->phy.ops.set_phy_power(hw, true);
5643 
5644 	smp_mb__before_atomic();
5645 	clear_bit(__IXGBE_DOWN, &adapter->state);
5646 	ixgbe_napi_enable_all(adapter);
5647 
5648 	if (ixgbe_is_sfp(hw)) {
5649 		ixgbe_sfp_link_config(adapter);
5650 	} else {
5651 		err = ixgbe_non_sfp_link_config(hw);
5652 		if (err)
5653 			e_err(probe, "link_config FAILED %d\n", err);
5654 	}
5655 
5656 	/* clear any pending interrupts, may auto mask */
5657 	IXGBE_READ_REG(hw, IXGBE_EICR);
5658 	ixgbe_irq_enable(adapter, true, true);
5659 
5660 	/*
5661 	 * If this adapter has a fan, check to see if we had a failure
5662 	 * before we enabled the interrupt.
5663 	 */
5664 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5665 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5666 		if (esdp & IXGBE_ESDP_SDP1)
5667 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5668 	}
5669 
5670 	/* bring the link up in the watchdog, this could race with our first
5671 	 * link up interrupt but shouldn't be a problem */
5672 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5673 	adapter->link_check_timeout = jiffies;
5674 	mod_timer(&adapter->service_timer, jiffies);
5675 
5676 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5677 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5678 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5679 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5680 }
5681 
5682 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5683 {
5684 	WARN_ON(in_interrupt());
5685 	/* put off any impending NetWatchDogTimeout */
5686 	netif_trans_update(adapter->netdev);
5687 
5688 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5689 		usleep_range(1000, 2000);
5690 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5691 		ixgbe_watchdog_link_is_down(adapter);
5692 	ixgbe_down(adapter);
5693 	/*
5694 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5695 	 * back up to give the VFs time to respond to the reset.  The
5696 	 * two second wait is based upon the watchdog timer cycle in
5697 	 * the VF driver.
5698 	 */
5699 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5700 		msleep(2000);
5701 	ixgbe_up(adapter);
5702 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5703 }
5704 
5705 void ixgbe_up(struct ixgbe_adapter *adapter)
5706 {
5707 	/* hardware has been reset, we need to reload some things */
5708 	ixgbe_configure(adapter);
5709 
5710 	ixgbe_up_complete(adapter);
5711 }
5712 
5713 void ixgbe_reset(struct ixgbe_adapter *adapter)
5714 {
5715 	struct ixgbe_hw *hw = &adapter->hw;
5716 	struct net_device *netdev = adapter->netdev;
5717 	int err;
5718 
5719 	if (ixgbe_removed(hw->hw_addr))
5720 		return;
5721 	/* lock SFP init bit to prevent race conditions with the watchdog */
5722 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5723 		usleep_range(1000, 2000);
5724 
5725 	/* clear all SFP and link config related flags while holding SFP_INIT */
5726 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5727 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5728 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5729 
5730 	err = hw->mac.ops.init_hw(hw);
5731 	switch (err) {
5732 	case 0:
5733 	case IXGBE_ERR_SFP_NOT_PRESENT:
5734 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5735 		break;
5736 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5737 		e_dev_err("master disable timed out\n");
5738 		break;
5739 	case IXGBE_ERR_EEPROM_VERSION:
5740 		/* We are running on a pre-production device, log a warning */
5741 		e_dev_warn("This device is a pre-production adapter/LOM. "
5742 			   "Please be aware there may be issues associated with "
5743 			   "your hardware.  If you are experiencing problems "
5744 			   "please contact your Intel or hardware "
5745 			   "representative who provided you with this "
5746 			   "hardware.\n");
5747 		break;
5748 	default:
5749 		e_dev_err("Hardware Error: %d\n", err);
5750 	}
5751 
5752 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5753 
5754 	/* flush entries out of MAC table */
5755 	ixgbe_flush_sw_mac_table(adapter);
5756 	__dev_uc_unsync(netdev, NULL);
5757 
5758 	/* do not flush user set addresses */
5759 	ixgbe_mac_set_default_filter(adapter);
5760 
5761 	/* update SAN MAC vmdq pool selection */
5762 	if (hw->mac.san_mac_rar_index)
5763 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5764 
5765 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5766 		ixgbe_ptp_reset(adapter);
5767 
5768 	if (hw->phy.ops.set_phy_power) {
5769 		if (!netif_running(adapter->netdev) && !adapter->wol)
5770 			hw->phy.ops.set_phy_power(hw, false);
5771 		else
5772 			hw->phy.ops.set_phy_power(hw, true);
5773 	}
5774 }
5775 
5776 /**
5777  * ixgbe_clean_tx_ring - Free Tx Buffers
5778  * @tx_ring: ring to be cleaned
5779  **/
5780 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5781 {
5782 	u16 i = tx_ring->next_to_clean;
5783 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
5784 
5785 	while (i != tx_ring->next_to_use) {
5786 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
5787 
5788 		/* Free all the Tx ring sk_buffs */
5789 		if (ring_is_xdp(tx_ring))
5790 			page_frag_free(tx_buffer->data);
5791 		else
5792 			dev_kfree_skb_any(tx_buffer->skb);
5793 
5794 		/* unmap skb header data */
5795 		dma_unmap_single(tx_ring->dev,
5796 				 dma_unmap_addr(tx_buffer, dma),
5797 				 dma_unmap_len(tx_buffer, len),
5798 				 DMA_TO_DEVICE);
5799 
5800 		/* check for eop_desc to determine the end of the packet */
5801 		eop_desc = tx_buffer->next_to_watch;
5802 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
5803 
5804 		/* unmap remaining buffers */
5805 		while (tx_desc != eop_desc) {
5806 			tx_buffer++;
5807 			tx_desc++;
5808 			i++;
5809 			if (unlikely(i == tx_ring->count)) {
5810 				i = 0;
5811 				tx_buffer = tx_ring->tx_buffer_info;
5812 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
5813 			}
5814 
5815 			/* unmap any remaining paged data */
5816 			if (dma_unmap_len(tx_buffer, len))
5817 				dma_unmap_page(tx_ring->dev,
5818 					       dma_unmap_addr(tx_buffer, dma),
5819 					       dma_unmap_len(tx_buffer, len),
5820 					       DMA_TO_DEVICE);
5821 		}
5822 
5823 		/* move us one more past the eop_desc for start of next pkt */
5824 		tx_buffer++;
5825 		i++;
5826 		if (unlikely(i == tx_ring->count)) {
5827 			i = 0;
5828 			tx_buffer = tx_ring->tx_buffer_info;
5829 		}
5830 	}
5831 
5832 	/* reset BQL for queue */
5833 	if (!ring_is_xdp(tx_ring))
5834 		netdev_tx_reset_queue(txring_txq(tx_ring));
5835 
5836 	/* reset next_to_use and next_to_clean */
5837 	tx_ring->next_to_use = 0;
5838 	tx_ring->next_to_clean = 0;
5839 }
5840 
5841 /**
5842  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5843  * @adapter: board private structure
5844  **/
5845 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5846 {
5847 	int i;
5848 
5849 	for (i = 0; i < adapter->num_rx_queues; i++)
5850 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5851 }
5852 
5853 /**
5854  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5855  * @adapter: board private structure
5856  **/
5857 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5858 {
5859 	int i;
5860 
5861 	for (i = 0; i < adapter->num_tx_queues; i++)
5862 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5863 	for (i = 0; i < adapter->num_xdp_queues; i++)
5864 		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
5865 }
5866 
5867 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5868 {
5869 	struct hlist_node *node2;
5870 	struct ixgbe_fdir_filter *filter;
5871 
5872 	spin_lock(&adapter->fdir_perfect_lock);
5873 
5874 	hlist_for_each_entry_safe(filter, node2,
5875 				  &adapter->fdir_filter_list, fdir_node) {
5876 		hlist_del(&filter->fdir_node);
5877 		kfree(filter);
5878 	}
5879 	adapter->fdir_filter_count = 0;
5880 
5881 	spin_unlock(&adapter->fdir_perfect_lock);
5882 }
5883 
5884 void ixgbe_down(struct ixgbe_adapter *adapter)
5885 {
5886 	struct net_device *netdev = adapter->netdev;
5887 	struct ixgbe_hw *hw = &adapter->hw;
5888 	int i;
5889 
5890 	/* signal that we are down to the interrupt handler */
5891 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5892 		return; /* do nothing if already down */
5893 
5894 	/* disable receives */
5895 	hw->mac.ops.disable_rx(hw);
5896 
5897 	/* disable all enabled rx queues */
5898 	for (i = 0; i < adapter->num_rx_queues; i++)
5899 		/* this call also flushes the previous write */
5900 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5901 
5902 	usleep_range(10000, 20000);
5903 
5904 	/* synchronize_sched() needed for pending XDP buffers to drain */
5905 	if (adapter->xdp_ring[0])
5906 		synchronize_sched();
5907 	netif_tx_stop_all_queues(netdev);
5908 
5909 	/* call carrier off first to avoid false dev_watchdog timeouts */
5910 	netif_carrier_off(netdev);
5911 	netif_tx_disable(netdev);
5912 
5913 	ixgbe_irq_disable(adapter);
5914 
5915 	ixgbe_napi_disable_all(adapter);
5916 
5917 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
5918 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5919 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5920 
5921 	del_timer_sync(&adapter->service_timer);
5922 
5923 	if (adapter->num_vfs) {
5924 		/* Clear EITR Select mapping */
5925 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5926 
5927 		/* Mark all the VFs as inactive */
5928 		for (i = 0 ; i < adapter->num_vfs; i++)
5929 			adapter->vfinfo[i].clear_to_send = false;
5930 
5931 		/* ping all the active vfs to let them know we are going down */
5932 		ixgbe_ping_all_vfs(adapter);
5933 
5934 		/* Disable all VFTE/VFRE TX/RX */
5935 		ixgbe_disable_tx_rx(adapter);
5936 	}
5937 
5938 	/* disable transmits in the hardware now that interrupts are off */
5939 	for (i = 0; i < adapter->num_tx_queues; i++) {
5940 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5941 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5942 	}
5943 	for (i = 0; i < adapter->num_xdp_queues; i++) {
5944 		u8 reg_idx = adapter->xdp_ring[i]->reg_idx;
5945 
5946 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5947 	}
5948 
5949 	/* Disable the Tx DMA engine on 82599 and later MAC */
5950 	switch (hw->mac.type) {
5951 	case ixgbe_mac_82599EB:
5952 	case ixgbe_mac_X540:
5953 	case ixgbe_mac_X550:
5954 	case ixgbe_mac_X550EM_x:
5955 	case ixgbe_mac_x550em_a:
5956 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5957 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5958 				 ~IXGBE_DMATXCTL_TE));
5959 		break;
5960 	default:
5961 		break;
5962 	}
5963 
5964 	if (!pci_channel_offline(adapter->pdev))
5965 		ixgbe_reset(adapter);
5966 
5967 	/* power down the optics for 82599 SFP+ fiber */
5968 	if (hw->mac.ops.disable_tx_laser)
5969 		hw->mac.ops.disable_tx_laser(hw);
5970 
5971 	ixgbe_clean_all_tx_rings(adapter);
5972 	ixgbe_clean_all_rx_rings(adapter);
5973 }
5974 
5975 /**
5976  * ixgbe_eee_capable - helper function to determine EEE support on X550
5977  * @adapter: board private structure
5978  */
5979 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
5980 {
5981 	struct ixgbe_hw *hw = &adapter->hw;
5982 
5983 	switch (hw->device_id) {
5984 	case IXGBE_DEV_ID_X550EM_A_1G_T:
5985 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
5986 		if (!hw->phy.eee_speeds_supported)
5987 			break;
5988 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
5989 		if (!hw->phy.eee_speeds_advertised)
5990 			break;
5991 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
5992 		break;
5993 	default:
5994 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
5995 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
5996 		break;
5997 	}
5998 }
5999 
6000 /**
6001  * ixgbe_tx_timeout - Respond to a Tx Hang
6002  * @netdev: network interface device structure
6003  **/
6004 static void ixgbe_tx_timeout(struct net_device *netdev)
6005 {
6006 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6007 
6008 	/* Do the reset outside of interrupt context */
6009 	ixgbe_tx_timeout_reset(adapter);
6010 }
6011 
6012 #ifdef CONFIG_IXGBE_DCB
6013 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
6014 {
6015 	struct ixgbe_hw *hw = &adapter->hw;
6016 	struct tc_configuration *tc;
6017 	int j;
6018 
6019 	switch (hw->mac.type) {
6020 	case ixgbe_mac_82598EB:
6021 	case ixgbe_mac_82599EB:
6022 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
6023 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
6024 		break;
6025 	case ixgbe_mac_X540:
6026 	case ixgbe_mac_X550:
6027 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
6028 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
6029 		break;
6030 	case ixgbe_mac_X550EM_x:
6031 	case ixgbe_mac_x550em_a:
6032 	default:
6033 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
6034 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
6035 		break;
6036 	}
6037 
6038 	/* Configure DCB traffic classes */
6039 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
6040 		tc = &adapter->dcb_cfg.tc_config[j];
6041 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
6042 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
6043 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
6044 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
6045 		tc->dcb_pfc = pfc_disabled;
6046 	}
6047 
6048 	/* Initialize default user to priority mapping, UPx->TC0 */
6049 	tc = &adapter->dcb_cfg.tc_config[0];
6050 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
6051 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
6052 
6053 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
6054 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
6055 	adapter->dcb_cfg.pfc_mode_enable = false;
6056 	adapter->dcb_set_bitmap = 0x00;
6057 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
6058 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
6059 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
6060 	       sizeof(adapter->temp_dcb_cfg));
6061 }
6062 #endif
6063 
6064 /**
6065  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
6066  * @adapter: board private structure to initialize
6067  * @ii: pointer to ixgbe_info for device
6068  *
6069  * ixgbe_sw_init initializes the Adapter private data structure.
6070  * Fields are initialized based on PCI device information and
6071  * OS network device settings (MTU size).
6072  **/
6073 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
6074 			 const struct ixgbe_info *ii)
6075 {
6076 	struct ixgbe_hw *hw = &adapter->hw;
6077 	struct pci_dev *pdev = adapter->pdev;
6078 	unsigned int rss, fdir;
6079 	u32 fwsm;
6080 	int i;
6081 
6082 	/* PCI config space info */
6083 
6084 	hw->vendor_id = pdev->vendor;
6085 	hw->device_id = pdev->device;
6086 	hw->revision_id = pdev->revision;
6087 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6088 	hw->subsystem_device_id = pdev->subsystem_device;
6089 
6090 	/* get_invariants needs the device IDs */
6091 	ii->get_invariants(hw);
6092 
6093 	/* Set common capability flags and settings */
6094 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6095 	adapter->ring_feature[RING_F_RSS].limit = rss;
6096 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6097 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6098 	adapter->atr_sample_rate = 20;
6099 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6100 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
6101 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6102 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
6103 #ifdef CONFIG_IXGBE_DCA
6104 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6105 #endif
6106 #ifdef CONFIG_IXGBE_DCB
6107 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6108 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6109 #endif
6110 #ifdef IXGBE_FCOE
6111 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6112 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6113 #ifdef CONFIG_IXGBE_DCB
6114 	/* Default traffic class to use for FCoE */
6115 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6116 #endif /* CONFIG_IXGBE_DCB */
6117 #endif /* IXGBE_FCOE */
6118 
6119 	/* initialize static ixgbe jump table entries */
6120 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6121 					  GFP_KERNEL);
6122 	if (!adapter->jump_tables[0])
6123 		return -ENOMEM;
6124 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6125 
6126 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6127 		adapter->jump_tables[i] = NULL;
6128 
6129 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
6130 				     hw->mac.num_rar_entries,
6131 				     GFP_ATOMIC);
6132 	if (!adapter->mac_table)
6133 		return -ENOMEM;
6134 
6135 	if (ixgbe_init_rss_key(adapter))
6136 		return -ENOMEM;
6137 
6138 	/* Set MAC specific capability flags and exceptions */
6139 	switch (hw->mac.type) {
6140 	case ixgbe_mac_82598EB:
6141 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6142 
6143 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
6144 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6145 
6146 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6147 		adapter->ring_feature[RING_F_FDIR].limit = 0;
6148 		adapter->atr_sample_rate = 0;
6149 		adapter->fdir_pballoc = 0;
6150 #ifdef IXGBE_FCOE
6151 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6152 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6153 #ifdef CONFIG_IXGBE_DCB
6154 		adapter->fcoe.up = 0;
6155 #endif /* IXGBE_DCB */
6156 #endif /* IXGBE_FCOE */
6157 		break;
6158 	case ixgbe_mac_82599EB:
6159 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6160 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6161 		break;
6162 	case ixgbe_mac_X540:
6163 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6164 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
6165 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6166 		break;
6167 	case ixgbe_mac_x550em_a:
6168 		adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
6169 		switch (hw->device_id) {
6170 		case IXGBE_DEV_ID_X550EM_A_1G_T:
6171 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6172 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6173 			break;
6174 		default:
6175 			break;
6176 		}
6177 	/* fall through */
6178 	case ixgbe_mac_X550EM_x:
6179 #ifdef CONFIG_IXGBE_DCB
6180 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6181 #endif
6182 #ifdef IXGBE_FCOE
6183 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6184 #ifdef CONFIG_IXGBE_DCB
6185 		adapter->fcoe.up = 0;
6186 #endif /* IXGBE_DCB */
6187 #endif /* IXGBE_FCOE */
6188 	/* Fall Through */
6189 	case ixgbe_mac_X550:
6190 		if (hw->mac.type == ixgbe_mac_X550)
6191 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6192 #ifdef CONFIG_IXGBE_DCA
6193 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6194 #endif
6195 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
6196 		break;
6197 	default:
6198 		break;
6199 	}
6200 
6201 #ifdef IXGBE_FCOE
6202 	/* FCoE support exists, always init the FCoE lock */
6203 	spin_lock_init(&adapter->fcoe.lock);
6204 
6205 #endif
6206 	/* n-tuple support exists, always init our spinlock */
6207 	spin_lock_init(&adapter->fdir_perfect_lock);
6208 
6209 #ifdef CONFIG_IXGBE_DCB
6210 	ixgbe_init_dcb(adapter);
6211 #endif
6212 
6213 	/* default flow control settings */
6214 	hw->fc.requested_mode = ixgbe_fc_full;
6215 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
6216 	ixgbe_pbthresh_setup(adapter);
6217 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6218 	hw->fc.send_xon = true;
6219 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6220 
6221 #ifdef CONFIG_PCI_IOV
6222 	if (max_vfs > 0)
6223 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6224 
6225 	/* assign number of SR-IOV VFs */
6226 	if (hw->mac.type != ixgbe_mac_82598EB) {
6227 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6228 			max_vfs = 0;
6229 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6230 		}
6231 	}
6232 #endif /* CONFIG_PCI_IOV */
6233 
6234 	/* enable itr by default in dynamic mode */
6235 	adapter->rx_itr_setting = 1;
6236 	adapter->tx_itr_setting = 1;
6237 
6238 	/* set default ring sizes */
6239 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6240 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6241 
6242 	/* set default work limits */
6243 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6244 
6245 	/* initialize eeprom parameters */
6246 	if (ixgbe_init_eeprom_params_generic(hw)) {
6247 		e_dev_err("EEPROM initialization failed\n");
6248 		return -EIO;
6249 	}
6250 
6251 	/* PF holds first pool slot */
6252 	set_bit(0, adapter->fwd_bitmask);
6253 	set_bit(__IXGBE_DOWN, &adapter->state);
6254 
6255 	return 0;
6256 }
6257 
6258 /**
6259  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6260  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6261  *
6262  * Return 0 on success, negative on failure
6263  **/
6264 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6265 {
6266 	struct device *dev = tx_ring->dev;
6267 	int orig_node = dev_to_node(dev);
6268 	int ring_node = -1;
6269 	int size;
6270 
6271 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6272 
6273 	if (tx_ring->q_vector)
6274 		ring_node = tx_ring->q_vector->numa_node;
6275 
6276 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6277 	if (!tx_ring->tx_buffer_info)
6278 		tx_ring->tx_buffer_info = vmalloc(size);
6279 	if (!tx_ring->tx_buffer_info)
6280 		goto err;
6281 
6282 	/* round up to nearest 4K */
6283 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6284 	tx_ring->size = ALIGN(tx_ring->size, 4096);
6285 
6286 	set_dev_node(dev, ring_node);
6287 	tx_ring->desc = dma_alloc_coherent(dev,
6288 					   tx_ring->size,
6289 					   &tx_ring->dma,
6290 					   GFP_KERNEL);
6291 	set_dev_node(dev, orig_node);
6292 	if (!tx_ring->desc)
6293 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6294 						   &tx_ring->dma, GFP_KERNEL);
6295 	if (!tx_ring->desc)
6296 		goto err;
6297 
6298 	tx_ring->next_to_use = 0;
6299 	tx_ring->next_to_clean = 0;
6300 	return 0;
6301 
6302 err:
6303 	vfree(tx_ring->tx_buffer_info);
6304 	tx_ring->tx_buffer_info = NULL;
6305 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6306 	return -ENOMEM;
6307 }
6308 
6309 /**
6310  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6311  * @adapter: board private structure
6312  *
6313  * If this function returns with an error, then it's possible one or
6314  * more of the rings is populated (while the rest are not).  It is the
6315  * callers duty to clean those orphaned rings.
6316  *
6317  * Return 0 on success, negative on failure
6318  **/
6319 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6320 {
6321 	int i, j = 0, err = 0;
6322 
6323 	for (i = 0; i < adapter->num_tx_queues; i++) {
6324 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6325 		if (!err)
6326 			continue;
6327 
6328 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6329 		goto err_setup_tx;
6330 	}
6331 	for (j = 0; j < adapter->num_xdp_queues; j++) {
6332 		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6333 		if (!err)
6334 			continue;
6335 
6336 		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6337 		goto err_setup_tx;
6338 	}
6339 
6340 	return 0;
6341 err_setup_tx:
6342 	/* rewind the index freeing the rings as we go */
6343 	while (j--)
6344 		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6345 	while (i--)
6346 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
6347 	return err;
6348 }
6349 
6350 /**
6351  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6352  * @adapter: pointer to ixgbe_adapter
6353  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6354  *
6355  * Returns 0 on success, negative on failure
6356  **/
6357 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6358 			     struct ixgbe_ring *rx_ring)
6359 {
6360 	struct device *dev = rx_ring->dev;
6361 	int orig_node = dev_to_node(dev);
6362 	int ring_node = -1;
6363 	int size;
6364 
6365 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6366 
6367 	if (rx_ring->q_vector)
6368 		ring_node = rx_ring->q_vector->numa_node;
6369 
6370 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6371 	if (!rx_ring->rx_buffer_info)
6372 		rx_ring->rx_buffer_info = vmalloc(size);
6373 	if (!rx_ring->rx_buffer_info)
6374 		goto err;
6375 
6376 	/* Round up to nearest 4K */
6377 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6378 	rx_ring->size = ALIGN(rx_ring->size, 4096);
6379 
6380 	set_dev_node(dev, ring_node);
6381 	rx_ring->desc = dma_alloc_coherent(dev,
6382 					   rx_ring->size,
6383 					   &rx_ring->dma,
6384 					   GFP_KERNEL);
6385 	set_dev_node(dev, orig_node);
6386 	if (!rx_ring->desc)
6387 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6388 						   &rx_ring->dma, GFP_KERNEL);
6389 	if (!rx_ring->desc)
6390 		goto err;
6391 
6392 	rx_ring->next_to_clean = 0;
6393 	rx_ring->next_to_use = 0;
6394 
6395 	/* XDP RX-queue info */
6396 	if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
6397 			     rx_ring->queue_index) < 0)
6398 		goto err;
6399 
6400 	rx_ring->xdp_prog = adapter->xdp_prog;
6401 
6402 	return 0;
6403 err:
6404 	vfree(rx_ring->rx_buffer_info);
6405 	rx_ring->rx_buffer_info = NULL;
6406 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6407 	return -ENOMEM;
6408 }
6409 
6410 /**
6411  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6412  * @adapter: board private structure
6413  *
6414  * If this function returns with an error, then it's possible one or
6415  * more of the rings is populated (while the rest are not).  It is the
6416  * callers duty to clean those orphaned rings.
6417  *
6418  * Return 0 on success, negative on failure
6419  **/
6420 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6421 {
6422 	int i, err = 0;
6423 
6424 	for (i = 0; i < adapter->num_rx_queues; i++) {
6425 		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6426 		if (!err)
6427 			continue;
6428 
6429 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6430 		goto err_setup_rx;
6431 	}
6432 
6433 #ifdef IXGBE_FCOE
6434 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
6435 	if (!err)
6436 #endif
6437 		return 0;
6438 err_setup_rx:
6439 	/* rewind the index freeing the rings as we go */
6440 	while (i--)
6441 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
6442 	return err;
6443 }
6444 
6445 /**
6446  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6447  * @tx_ring: Tx descriptor ring for a specific queue
6448  *
6449  * Free all transmit software resources
6450  **/
6451 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6452 {
6453 	ixgbe_clean_tx_ring(tx_ring);
6454 
6455 	vfree(tx_ring->tx_buffer_info);
6456 	tx_ring->tx_buffer_info = NULL;
6457 
6458 	/* if not set, then don't free */
6459 	if (!tx_ring->desc)
6460 		return;
6461 
6462 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6463 			  tx_ring->desc, tx_ring->dma);
6464 
6465 	tx_ring->desc = NULL;
6466 }
6467 
6468 /**
6469  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6470  * @adapter: board private structure
6471  *
6472  * Free all transmit software resources
6473  **/
6474 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6475 {
6476 	int i;
6477 
6478 	for (i = 0; i < adapter->num_tx_queues; i++)
6479 		if (adapter->tx_ring[i]->desc)
6480 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6481 	for (i = 0; i < adapter->num_xdp_queues; i++)
6482 		if (adapter->xdp_ring[i]->desc)
6483 			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6484 }
6485 
6486 /**
6487  * ixgbe_free_rx_resources - Free Rx Resources
6488  * @rx_ring: ring to clean the resources from
6489  *
6490  * Free all receive software resources
6491  **/
6492 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6493 {
6494 	ixgbe_clean_rx_ring(rx_ring);
6495 
6496 	rx_ring->xdp_prog = NULL;
6497 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6498 	vfree(rx_ring->rx_buffer_info);
6499 	rx_ring->rx_buffer_info = NULL;
6500 
6501 	/* if not set, then don't free */
6502 	if (!rx_ring->desc)
6503 		return;
6504 
6505 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6506 			  rx_ring->desc, rx_ring->dma);
6507 
6508 	rx_ring->desc = NULL;
6509 }
6510 
6511 /**
6512  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6513  * @adapter: board private structure
6514  *
6515  * Free all receive software resources
6516  **/
6517 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6518 {
6519 	int i;
6520 
6521 #ifdef IXGBE_FCOE
6522 	ixgbe_free_fcoe_ddp_resources(adapter);
6523 
6524 #endif
6525 	for (i = 0; i < adapter->num_rx_queues; i++)
6526 		if (adapter->rx_ring[i]->desc)
6527 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6528 }
6529 
6530 /**
6531  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6532  * @netdev: network interface device structure
6533  * @new_mtu: new value for maximum frame size
6534  *
6535  * Returns 0 on success, negative on failure
6536  **/
6537 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6538 {
6539 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6540 
6541 	/*
6542 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6543 	 * paths when MTU greater than 1500 is configured.  So display a
6544 	 * warning that legacy VFs will be disabled.
6545 	 */
6546 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6547 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6548 	    (new_mtu > ETH_DATA_LEN))
6549 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6550 
6551 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6552 
6553 	/* must set new MTU before calling down or up */
6554 	netdev->mtu = new_mtu;
6555 
6556 	if (netif_running(netdev))
6557 		ixgbe_reinit_locked(adapter);
6558 
6559 	return 0;
6560 }
6561 
6562 /**
6563  * ixgbe_open - Called when a network interface is made active
6564  * @netdev: network interface device structure
6565  *
6566  * Returns 0 on success, negative value on failure
6567  *
6568  * The open entry point is called when a network interface is made
6569  * active by the system (IFF_UP).  At this point all resources needed
6570  * for transmit and receive operations are allocated, the interrupt
6571  * handler is registered with the OS, the watchdog timer is started,
6572  * and the stack is notified that the interface is ready.
6573  **/
6574 int ixgbe_open(struct net_device *netdev)
6575 {
6576 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6577 	struct ixgbe_hw *hw = &adapter->hw;
6578 	int err, queues;
6579 
6580 	/* disallow open during test */
6581 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6582 		return -EBUSY;
6583 
6584 	netif_carrier_off(netdev);
6585 
6586 	/* allocate transmit descriptors */
6587 	err = ixgbe_setup_all_tx_resources(adapter);
6588 	if (err)
6589 		goto err_setup_tx;
6590 
6591 	/* allocate receive descriptors */
6592 	err = ixgbe_setup_all_rx_resources(adapter);
6593 	if (err)
6594 		goto err_setup_rx;
6595 
6596 	ixgbe_configure(adapter);
6597 
6598 	err = ixgbe_request_irq(adapter);
6599 	if (err)
6600 		goto err_req_irq;
6601 
6602 	/* Notify the stack of the actual queue counts. */
6603 	queues = adapter->num_tx_queues;
6604 	err = netif_set_real_num_tx_queues(netdev, queues);
6605 	if (err)
6606 		goto err_set_queues;
6607 
6608 	queues = adapter->num_rx_queues;
6609 	err = netif_set_real_num_rx_queues(netdev, queues);
6610 	if (err)
6611 		goto err_set_queues;
6612 
6613 	ixgbe_ptp_init(adapter);
6614 
6615 	ixgbe_up_complete(adapter);
6616 
6617 	ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6618 	udp_tunnel_get_rx_info(netdev);
6619 
6620 	return 0;
6621 
6622 err_set_queues:
6623 	ixgbe_free_irq(adapter);
6624 err_req_irq:
6625 	ixgbe_free_all_rx_resources(adapter);
6626 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6627 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6628 err_setup_rx:
6629 	ixgbe_free_all_tx_resources(adapter);
6630 err_setup_tx:
6631 	ixgbe_reset(adapter);
6632 
6633 	return err;
6634 }
6635 
6636 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6637 {
6638 	ixgbe_ptp_suspend(adapter);
6639 
6640 	if (adapter->hw.phy.ops.enter_lplu) {
6641 		adapter->hw.phy.reset_disable = true;
6642 		ixgbe_down(adapter);
6643 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6644 		adapter->hw.phy.reset_disable = false;
6645 	} else {
6646 		ixgbe_down(adapter);
6647 	}
6648 
6649 	ixgbe_free_irq(adapter);
6650 
6651 	ixgbe_free_all_tx_resources(adapter);
6652 	ixgbe_free_all_rx_resources(adapter);
6653 }
6654 
6655 /**
6656  * ixgbe_close - Disables a network interface
6657  * @netdev: network interface device structure
6658  *
6659  * Returns 0, this is not allowed to fail
6660  *
6661  * The close entry point is called when an interface is de-activated
6662  * by the OS.  The hardware is still under the drivers control, but
6663  * needs to be disabled.  A global MAC reset is issued to stop the
6664  * hardware, and all transmit and receive resources are freed.
6665  **/
6666 int ixgbe_close(struct net_device *netdev)
6667 {
6668 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6669 
6670 	ixgbe_ptp_stop(adapter);
6671 
6672 	if (netif_device_present(netdev))
6673 		ixgbe_close_suspend(adapter);
6674 
6675 	ixgbe_fdir_filter_exit(adapter);
6676 
6677 	ixgbe_release_hw_control(adapter);
6678 
6679 	return 0;
6680 }
6681 
6682 #ifdef CONFIG_PM
6683 static int ixgbe_resume(struct pci_dev *pdev)
6684 {
6685 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6686 	struct net_device *netdev = adapter->netdev;
6687 	u32 err;
6688 
6689 	adapter->hw.hw_addr = adapter->io_addr;
6690 	pci_set_power_state(pdev, PCI_D0);
6691 	pci_restore_state(pdev);
6692 	/*
6693 	 * pci_restore_state clears dev->state_saved so call
6694 	 * pci_save_state to restore it.
6695 	 */
6696 	pci_save_state(pdev);
6697 
6698 	err = pci_enable_device_mem(pdev);
6699 	if (err) {
6700 		e_dev_err("Cannot enable PCI device from suspend\n");
6701 		return err;
6702 	}
6703 	smp_mb__before_atomic();
6704 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6705 	pci_set_master(pdev);
6706 
6707 	pci_wake_from_d3(pdev, false);
6708 
6709 	ixgbe_reset(adapter);
6710 
6711 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6712 
6713 	rtnl_lock();
6714 	err = ixgbe_init_interrupt_scheme(adapter);
6715 	if (!err && netif_running(netdev))
6716 		err = ixgbe_open(netdev);
6717 
6718 
6719 	if (!err)
6720 		netif_device_attach(netdev);
6721 	rtnl_unlock();
6722 
6723 	return err;
6724 }
6725 #endif /* CONFIG_PM */
6726 
6727 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6728 {
6729 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6730 	struct net_device *netdev = adapter->netdev;
6731 	struct ixgbe_hw *hw = &adapter->hw;
6732 	u32 ctrl;
6733 	u32 wufc = adapter->wol;
6734 #ifdef CONFIG_PM
6735 	int retval = 0;
6736 #endif
6737 
6738 	rtnl_lock();
6739 	netif_device_detach(netdev);
6740 
6741 	if (netif_running(netdev))
6742 		ixgbe_close_suspend(adapter);
6743 
6744 	ixgbe_clear_interrupt_scheme(adapter);
6745 	rtnl_unlock();
6746 
6747 #ifdef CONFIG_PM
6748 	retval = pci_save_state(pdev);
6749 	if (retval)
6750 		return retval;
6751 
6752 #endif
6753 	if (hw->mac.ops.stop_link_on_d3)
6754 		hw->mac.ops.stop_link_on_d3(hw);
6755 
6756 	if (wufc) {
6757 		u32 fctrl;
6758 
6759 		ixgbe_set_rx_mode(netdev);
6760 
6761 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6762 		if (hw->mac.ops.enable_tx_laser)
6763 			hw->mac.ops.enable_tx_laser(hw);
6764 
6765 		/* enable the reception of multicast packets */
6766 		fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6767 		fctrl |= IXGBE_FCTRL_MPE;
6768 		IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6769 
6770 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6771 		ctrl |= IXGBE_CTRL_GIO_DIS;
6772 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6773 
6774 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6775 	} else {
6776 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6777 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6778 	}
6779 
6780 	switch (hw->mac.type) {
6781 	case ixgbe_mac_82598EB:
6782 		pci_wake_from_d3(pdev, false);
6783 		break;
6784 	case ixgbe_mac_82599EB:
6785 	case ixgbe_mac_X540:
6786 	case ixgbe_mac_X550:
6787 	case ixgbe_mac_X550EM_x:
6788 	case ixgbe_mac_x550em_a:
6789 		pci_wake_from_d3(pdev, !!wufc);
6790 		break;
6791 	default:
6792 		break;
6793 	}
6794 
6795 	*enable_wake = !!wufc;
6796 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6797 		hw->phy.ops.set_phy_power(hw, false);
6798 
6799 	ixgbe_release_hw_control(adapter);
6800 
6801 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6802 		pci_disable_device(pdev);
6803 
6804 	return 0;
6805 }
6806 
6807 #ifdef CONFIG_PM
6808 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6809 {
6810 	int retval;
6811 	bool wake;
6812 
6813 	retval = __ixgbe_shutdown(pdev, &wake);
6814 	if (retval)
6815 		return retval;
6816 
6817 	if (wake) {
6818 		pci_prepare_to_sleep(pdev);
6819 	} else {
6820 		pci_wake_from_d3(pdev, false);
6821 		pci_set_power_state(pdev, PCI_D3hot);
6822 	}
6823 
6824 	return 0;
6825 }
6826 #endif /* CONFIG_PM */
6827 
6828 static void ixgbe_shutdown(struct pci_dev *pdev)
6829 {
6830 	bool wake;
6831 
6832 	__ixgbe_shutdown(pdev, &wake);
6833 
6834 	if (system_state == SYSTEM_POWER_OFF) {
6835 		pci_wake_from_d3(pdev, wake);
6836 		pci_set_power_state(pdev, PCI_D3hot);
6837 	}
6838 }
6839 
6840 /**
6841  * ixgbe_update_stats - Update the board statistics counters.
6842  * @adapter: board private structure
6843  **/
6844 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6845 {
6846 	struct net_device *netdev = adapter->netdev;
6847 	struct ixgbe_hw *hw = &adapter->hw;
6848 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6849 	u64 total_mpc = 0;
6850 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6851 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6852 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6853 	u64 alloc_rx_page = 0;
6854 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6855 
6856 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6857 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6858 		return;
6859 
6860 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6861 		u64 rsc_count = 0;
6862 		u64 rsc_flush = 0;
6863 		for (i = 0; i < adapter->num_rx_queues; i++) {
6864 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6865 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6866 		}
6867 		adapter->rsc_total_count = rsc_count;
6868 		adapter->rsc_total_flush = rsc_flush;
6869 	}
6870 
6871 	for (i = 0; i < adapter->num_rx_queues; i++) {
6872 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6873 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6874 		alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
6875 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6876 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6877 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6878 		bytes += rx_ring->stats.bytes;
6879 		packets += rx_ring->stats.packets;
6880 	}
6881 	adapter->non_eop_descs = non_eop_descs;
6882 	adapter->alloc_rx_page = alloc_rx_page;
6883 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6884 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6885 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6886 	netdev->stats.rx_bytes = bytes;
6887 	netdev->stats.rx_packets = packets;
6888 
6889 	bytes = 0;
6890 	packets = 0;
6891 	/* gather some stats to the adapter struct that are per queue */
6892 	for (i = 0; i < adapter->num_tx_queues; i++) {
6893 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6894 		restart_queue += tx_ring->tx_stats.restart_queue;
6895 		tx_busy += tx_ring->tx_stats.tx_busy;
6896 		bytes += tx_ring->stats.bytes;
6897 		packets += tx_ring->stats.packets;
6898 	}
6899 	for (i = 0; i < adapter->num_xdp_queues; i++) {
6900 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
6901 
6902 		restart_queue += xdp_ring->tx_stats.restart_queue;
6903 		tx_busy += xdp_ring->tx_stats.tx_busy;
6904 		bytes += xdp_ring->stats.bytes;
6905 		packets += xdp_ring->stats.packets;
6906 	}
6907 	adapter->restart_queue = restart_queue;
6908 	adapter->tx_busy = tx_busy;
6909 	netdev->stats.tx_bytes = bytes;
6910 	netdev->stats.tx_packets = packets;
6911 
6912 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6913 
6914 	/* 8 register reads */
6915 	for (i = 0; i < 8; i++) {
6916 		/* for packet buffers not used, the register should read 0 */
6917 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6918 		missed_rx += mpc;
6919 		hwstats->mpc[i] += mpc;
6920 		total_mpc += hwstats->mpc[i];
6921 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6922 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6923 		switch (hw->mac.type) {
6924 		case ixgbe_mac_82598EB:
6925 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6926 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6927 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6928 			hwstats->pxonrxc[i] +=
6929 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6930 			break;
6931 		case ixgbe_mac_82599EB:
6932 		case ixgbe_mac_X540:
6933 		case ixgbe_mac_X550:
6934 		case ixgbe_mac_X550EM_x:
6935 		case ixgbe_mac_x550em_a:
6936 			hwstats->pxonrxc[i] +=
6937 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6938 			break;
6939 		default:
6940 			break;
6941 		}
6942 	}
6943 
6944 	/*16 register reads */
6945 	for (i = 0; i < 16; i++) {
6946 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6947 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6948 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6949 		    (hw->mac.type == ixgbe_mac_X540) ||
6950 		    (hw->mac.type == ixgbe_mac_X550) ||
6951 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6952 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6953 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6954 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6955 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6956 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6957 		}
6958 	}
6959 
6960 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6961 	/* work around hardware counting issue */
6962 	hwstats->gprc -= missed_rx;
6963 
6964 	ixgbe_update_xoff_received(adapter);
6965 
6966 	/* 82598 hardware only has a 32 bit counter in the high register */
6967 	switch (hw->mac.type) {
6968 	case ixgbe_mac_82598EB:
6969 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6970 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6971 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6972 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6973 		break;
6974 	case ixgbe_mac_X540:
6975 	case ixgbe_mac_X550:
6976 	case ixgbe_mac_X550EM_x:
6977 	case ixgbe_mac_x550em_a:
6978 		/* OS2BMC stats are X540 and later */
6979 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6980 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6981 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6982 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6983 		/* fall through */
6984 	case ixgbe_mac_82599EB:
6985 		for (i = 0; i < 16; i++)
6986 			adapter->hw_rx_no_dma_resources +=
6987 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6988 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6989 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6990 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6991 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6992 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6993 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6994 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6995 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6996 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6997 #ifdef IXGBE_FCOE
6998 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6999 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
7000 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
7001 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
7002 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
7003 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
7004 		/* Add up per cpu counters for total ddp aloc fail */
7005 		if (adapter->fcoe.ddp_pool) {
7006 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
7007 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
7008 			unsigned int cpu;
7009 			u64 noddp = 0, noddp_ext_buff = 0;
7010 			for_each_possible_cpu(cpu) {
7011 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
7012 				noddp += ddp_pool->noddp;
7013 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
7014 			}
7015 			hwstats->fcoe_noddp = noddp;
7016 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
7017 		}
7018 #endif /* IXGBE_FCOE */
7019 		break;
7020 	default:
7021 		break;
7022 	}
7023 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
7024 	hwstats->bprc += bprc;
7025 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
7026 	if (hw->mac.type == ixgbe_mac_82598EB)
7027 		hwstats->mprc -= bprc;
7028 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
7029 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
7030 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
7031 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
7032 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
7033 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
7034 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
7035 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
7036 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
7037 	hwstats->lxontxc += lxon;
7038 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
7039 	hwstats->lxofftxc += lxoff;
7040 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
7041 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
7042 	/*
7043 	 * 82598 errata - tx of flow control packets is included in tx counters
7044 	 */
7045 	xon_off_tot = lxon + lxoff;
7046 	hwstats->gptc -= xon_off_tot;
7047 	hwstats->mptc -= xon_off_tot;
7048 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
7049 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
7050 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
7051 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
7052 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
7053 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
7054 	hwstats->ptc64 -= xon_off_tot;
7055 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
7056 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
7057 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
7058 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
7059 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
7060 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
7061 
7062 	/* Fill out the OS statistics structure */
7063 	netdev->stats.multicast = hwstats->mprc;
7064 
7065 	/* Rx Errors */
7066 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
7067 	netdev->stats.rx_dropped = 0;
7068 	netdev->stats.rx_length_errors = hwstats->rlec;
7069 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
7070 	netdev->stats.rx_missed_errors = total_mpc;
7071 }
7072 
7073 /**
7074  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7075  * @adapter: pointer to the device adapter structure
7076  **/
7077 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7078 {
7079 	struct ixgbe_hw *hw = &adapter->hw;
7080 	int i;
7081 
7082 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7083 		return;
7084 
7085 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7086 
7087 	/* if interface is down do nothing */
7088 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7089 		return;
7090 
7091 	/* do nothing if we are not using signature filters */
7092 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7093 		return;
7094 
7095 	adapter->fdir_overflow++;
7096 
7097 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7098 		for (i = 0; i < adapter->num_tx_queues; i++)
7099 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7100 				&(adapter->tx_ring[i]->state));
7101 		for (i = 0; i < adapter->num_xdp_queues; i++)
7102 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7103 				&adapter->xdp_ring[i]->state);
7104 		/* re-enable flow director interrupts */
7105 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7106 	} else {
7107 		e_err(probe, "failed to finish FDIR re-initialization, "
7108 		      "ignored adding FDIR ATR filters\n");
7109 	}
7110 }
7111 
7112 /**
7113  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7114  * @adapter: pointer to the device adapter structure
7115  *
7116  * This function serves two purposes.  First it strobes the interrupt lines
7117  * in order to make certain interrupts are occurring.  Secondly it sets the
7118  * bits needed to check for TX hangs.  As a result we should immediately
7119  * determine if a hang has occurred.
7120  */
7121 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7122 {
7123 	struct ixgbe_hw *hw = &adapter->hw;
7124 	u64 eics = 0;
7125 	int i;
7126 
7127 	/* If we're down, removing or resetting, just bail */
7128 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7129 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7130 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7131 		return;
7132 
7133 	/* Force detection of hung controller */
7134 	if (netif_carrier_ok(adapter->netdev)) {
7135 		for (i = 0; i < adapter->num_tx_queues; i++)
7136 			set_check_for_tx_hang(adapter->tx_ring[i]);
7137 		for (i = 0; i < adapter->num_xdp_queues; i++)
7138 			set_check_for_tx_hang(adapter->xdp_ring[i]);
7139 	}
7140 
7141 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7142 		/*
7143 		 * for legacy and MSI interrupts don't set any bits
7144 		 * that are enabled for EIAM, because this operation
7145 		 * would set *both* EIMS and EICS for any bit in EIAM
7146 		 */
7147 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
7148 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7149 	} else {
7150 		/* get one bit for every active tx/rx interrupt vector */
7151 		for (i = 0; i < adapter->num_q_vectors; i++) {
7152 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
7153 			if (qv->rx.ring || qv->tx.ring)
7154 				eics |= BIT_ULL(i);
7155 		}
7156 	}
7157 
7158 	/* Cause software interrupt to ensure rings are cleaned */
7159 	ixgbe_irq_rearm_queues(adapter, eics);
7160 }
7161 
7162 /**
7163  * ixgbe_watchdog_update_link - update the link status
7164  * @adapter: pointer to the device adapter structure
7165  **/
7166 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7167 {
7168 	struct ixgbe_hw *hw = &adapter->hw;
7169 	u32 link_speed = adapter->link_speed;
7170 	bool link_up = adapter->link_up;
7171 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7172 
7173 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7174 		return;
7175 
7176 	if (hw->mac.ops.check_link) {
7177 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7178 	} else {
7179 		/* always assume link is up, if no check link function */
7180 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7181 		link_up = true;
7182 	}
7183 
7184 	if (adapter->ixgbe_ieee_pfc)
7185 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7186 
7187 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7188 		hw->mac.ops.fc_enable(hw);
7189 		ixgbe_set_rx_drop_en(adapter);
7190 	}
7191 
7192 	if (link_up ||
7193 	    time_after(jiffies, (adapter->link_check_timeout +
7194 				 IXGBE_TRY_LINK_TIMEOUT))) {
7195 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7196 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7197 		IXGBE_WRITE_FLUSH(hw);
7198 	}
7199 
7200 	adapter->link_up = link_up;
7201 	adapter->link_speed = link_speed;
7202 }
7203 
7204 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7205 {
7206 #ifdef CONFIG_IXGBE_DCB
7207 	struct net_device *netdev = adapter->netdev;
7208 	struct dcb_app app = {
7209 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7210 			      .protocol = 0,
7211 			     };
7212 	u8 up = 0;
7213 
7214 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7215 		up = dcb_ieee_getapp_mask(netdev, &app);
7216 
7217 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7218 #endif
7219 }
7220 
7221 /**
7222  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7223  *                             print link up message
7224  * @adapter: pointer to the device adapter structure
7225  **/
7226 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7227 {
7228 	struct net_device *netdev = adapter->netdev;
7229 	struct ixgbe_hw *hw = &adapter->hw;
7230 	u32 link_speed = adapter->link_speed;
7231 	const char *speed_str;
7232 	bool flow_rx, flow_tx;
7233 
7234 	/* only continue if link was previously down */
7235 	if (netif_carrier_ok(netdev))
7236 		return;
7237 
7238 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7239 
7240 	switch (hw->mac.type) {
7241 	case ixgbe_mac_82598EB: {
7242 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7243 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7244 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7245 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7246 	}
7247 		break;
7248 	case ixgbe_mac_X540:
7249 	case ixgbe_mac_X550:
7250 	case ixgbe_mac_X550EM_x:
7251 	case ixgbe_mac_x550em_a:
7252 	case ixgbe_mac_82599EB: {
7253 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7254 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7255 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7256 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7257 	}
7258 		break;
7259 	default:
7260 		flow_tx = false;
7261 		flow_rx = false;
7262 		break;
7263 	}
7264 
7265 	adapter->last_rx_ptp_check = jiffies;
7266 
7267 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7268 		ixgbe_ptp_start_cyclecounter(adapter);
7269 
7270 	switch (link_speed) {
7271 	case IXGBE_LINK_SPEED_10GB_FULL:
7272 		speed_str = "10 Gbps";
7273 		break;
7274 	case IXGBE_LINK_SPEED_5GB_FULL:
7275 		speed_str = "5 Gbps";
7276 		break;
7277 	case IXGBE_LINK_SPEED_2_5GB_FULL:
7278 		speed_str = "2.5 Gbps";
7279 		break;
7280 	case IXGBE_LINK_SPEED_1GB_FULL:
7281 		speed_str = "1 Gbps";
7282 		break;
7283 	case IXGBE_LINK_SPEED_100_FULL:
7284 		speed_str = "100 Mbps";
7285 		break;
7286 	case IXGBE_LINK_SPEED_10_FULL:
7287 		speed_str = "10 Mbps";
7288 		break;
7289 	default:
7290 		speed_str = "unknown speed";
7291 		break;
7292 	}
7293 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7294 	       ((flow_rx && flow_tx) ? "RX/TX" :
7295 	       (flow_rx ? "RX" :
7296 	       (flow_tx ? "TX" : "None"))));
7297 
7298 	netif_carrier_on(netdev);
7299 	ixgbe_check_vf_rate_limit(adapter);
7300 
7301 	/* enable transmits */
7302 	netif_tx_wake_all_queues(adapter->netdev);
7303 
7304 	/* update the default user priority for VFs */
7305 	ixgbe_update_default_up(adapter);
7306 
7307 	/* ping all the active vfs to let them know link has changed */
7308 	ixgbe_ping_all_vfs(adapter);
7309 }
7310 
7311 /**
7312  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7313  *                               print link down message
7314  * @adapter: pointer to the adapter structure
7315  **/
7316 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7317 {
7318 	struct net_device *netdev = adapter->netdev;
7319 	struct ixgbe_hw *hw = &adapter->hw;
7320 
7321 	adapter->link_up = false;
7322 	adapter->link_speed = 0;
7323 
7324 	/* only continue if link was up previously */
7325 	if (!netif_carrier_ok(netdev))
7326 		return;
7327 
7328 	/* poll for SFP+ cable when link is down */
7329 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7330 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7331 
7332 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7333 		ixgbe_ptp_start_cyclecounter(adapter);
7334 
7335 	e_info(drv, "NIC Link is Down\n");
7336 	netif_carrier_off(netdev);
7337 
7338 	/* ping all the active vfs to let them know link has changed */
7339 	ixgbe_ping_all_vfs(adapter);
7340 }
7341 
7342 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7343 {
7344 	int i;
7345 
7346 	for (i = 0; i < adapter->num_tx_queues; i++) {
7347 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7348 
7349 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
7350 			return true;
7351 	}
7352 
7353 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7354 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
7355 
7356 		if (ring->next_to_use != ring->next_to_clean)
7357 			return true;
7358 	}
7359 
7360 	return false;
7361 }
7362 
7363 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7364 {
7365 	struct ixgbe_hw *hw = &adapter->hw;
7366 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7367 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7368 
7369 	int i, j;
7370 
7371 	if (!adapter->num_vfs)
7372 		return false;
7373 
7374 	/* resetting the PF is only needed for MAC before X550 */
7375 	if (hw->mac.type >= ixgbe_mac_X550)
7376 		return false;
7377 
7378 	for (i = 0; i < adapter->num_vfs; i++) {
7379 		for (j = 0; j < q_per_pool; j++) {
7380 			u32 h, t;
7381 
7382 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7383 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7384 
7385 			if (h != t)
7386 				return true;
7387 		}
7388 	}
7389 
7390 	return false;
7391 }
7392 
7393 /**
7394  * ixgbe_watchdog_flush_tx - flush queues on link down
7395  * @adapter: pointer to the device adapter structure
7396  **/
7397 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7398 {
7399 	if (!netif_carrier_ok(adapter->netdev)) {
7400 		if (ixgbe_ring_tx_pending(adapter) ||
7401 		    ixgbe_vf_tx_pending(adapter)) {
7402 			/* We've lost link, so the controller stops DMA,
7403 			 * but we've got queued Tx work that's never going
7404 			 * to get done, so reset controller to flush Tx.
7405 			 * (Do the reset outside of interrupt context).
7406 			 */
7407 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7408 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7409 		}
7410 	}
7411 }
7412 
7413 #ifdef CONFIG_PCI_IOV
7414 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7415 {
7416 	struct ixgbe_hw *hw = &adapter->hw;
7417 	struct pci_dev *pdev = adapter->pdev;
7418 	unsigned int vf;
7419 	u32 gpc;
7420 
7421 	if (!(netif_carrier_ok(adapter->netdev)))
7422 		return;
7423 
7424 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7425 	if (gpc) /* If incrementing then no need for the check below */
7426 		return;
7427 	/* Check to see if a bad DMA write target from an errant or
7428 	 * malicious VF has caused a PCIe error.  If so then we can
7429 	 * issue a VFLR to the offending VF(s) and then resume without
7430 	 * requesting a full slot reset.
7431 	 */
7432 
7433 	if (!pdev)
7434 		return;
7435 
7436 	/* check status reg for all VFs owned by this PF */
7437 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
7438 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7439 		u16 status_reg;
7440 
7441 		if (!vfdev)
7442 			continue;
7443 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7444 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7445 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
7446 			pcie_flr(vfdev);
7447 	}
7448 }
7449 
7450 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7451 {
7452 	u32 ssvpc;
7453 
7454 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7455 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7456 	    adapter->num_vfs == 0)
7457 		return;
7458 
7459 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7460 
7461 	/*
7462 	 * ssvpc register is cleared on read, if zero then no
7463 	 * spoofed packets in the last interval.
7464 	 */
7465 	if (!ssvpc)
7466 		return;
7467 
7468 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7469 }
7470 #else
7471 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7472 {
7473 }
7474 
7475 static void
7476 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7477 {
7478 }
7479 #endif /* CONFIG_PCI_IOV */
7480 
7481 
7482 /**
7483  * ixgbe_watchdog_subtask - check and bring link up
7484  * @adapter: pointer to the device adapter structure
7485  **/
7486 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7487 {
7488 	/* if interface is down, removing or resetting, do nothing */
7489 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7490 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7491 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7492 		return;
7493 
7494 	ixgbe_watchdog_update_link(adapter);
7495 
7496 	if (adapter->link_up)
7497 		ixgbe_watchdog_link_is_up(adapter);
7498 	else
7499 		ixgbe_watchdog_link_is_down(adapter);
7500 
7501 	ixgbe_check_for_bad_vf(adapter);
7502 	ixgbe_spoof_check(adapter);
7503 	ixgbe_update_stats(adapter);
7504 
7505 	ixgbe_watchdog_flush_tx(adapter);
7506 }
7507 
7508 /**
7509  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7510  * @adapter: the ixgbe adapter structure
7511  **/
7512 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7513 {
7514 	struct ixgbe_hw *hw = &adapter->hw;
7515 	s32 err;
7516 
7517 	/* not searching for SFP so there is nothing to do here */
7518 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7519 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7520 		return;
7521 
7522 	if (adapter->sfp_poll_time &&
7523 	    time_after(adapter->sfp_poll_time, jiffies))
7524 		return; /* If not yet time to poll for SFP */
7525 
7526 	/* someone else is in init, wait until next service event */
7527 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7528 		return;
7529 
7530 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7531 
7532 	err = hw->phy.ops.identify_sfp(hw);
7533 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7534 		goto sfp_out;
7535 
7536 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7537 		/* If no cable is present, then we need to reset
7538 		 * the next time we find a good cable. */
7539 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7540 	}
7541 
7542 	/* exit on error */
7543 	if (err)
7544 		goto sfp_out;
7545 
7546 	/* exit if reset not needed */
7547 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7548 		goto sfp_out;
7549 
7550 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7551 
7552 	/*
7553 	 * A module may be identified correctly, but the EEPROM may not have
7554 	 * support for that module.  setup_sfp() will fail in that case, so
7555 	 * we should not allow that module to load.
7556 	 */
7557 	if (hw->mac.type == ixgbe_mac_82598EB)
7558 		err = hw->phy.ops.reset(hw);
7559 	else
7560 		err = hw->mac.ops.setup_sfp(hw);
7561 
7562 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7563 		goto sfp_out;
7564 
7565 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7566 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7567 
7568 sfp_out:
7569 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7570 
7571 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7572 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7573 		e_dev_err("failed to initialize because an unsupported "
7574 			  "SFP+ module type was detected.\n");
7575 		e_dev_err("Reload the driver after installing a "
7576 			  "supported module.\n");
7577 		unregister_netdev(adapter->netdev);
7578 	}
7579 }
7580 
7581 /**
7582  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7583  * @adapter: the ixgbe adapter structure
7584  **/
7585 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7586 {
7587 	struct ixgbe_hw *hw = &adapter->hw;
7588 	u32 cap_speed;
7589 	u32 speed;
7590 	bool autoneg = false;
7591 
7592 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7593 		return;
7594 
7595 	/* someone else is in init, wait until next service event */
7596 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7597 		return;
7598 
7599 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7600 
7601 	hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
7602 
7603 	/* advertise highest capable link speed */
7604 	if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
7605 		speed = IXGBE_LINK_SPEED_10GB_FULL;
7606 	else
7607 		speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
7608 				     IXGBE_LINK_SPEED_1GB_FULL);
7609 
7610 	if (hw->mac.ops.setup_link)
7611 		hw->mac.ops.setup_link(hw, speed, true);
7612 
7613 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7614 	adapter->link_check_timeout = jiffies;
7615 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7616 }
7617 
7618 /**
7619  * ixgbe_service_timer - Timer Call-back
7620  * @t: pointer to timer_list structure
7621  **/
7622 static void ixgbe_service_timer(struct timer_list *t)
7623 {
7624 	struct ixgbe_adapter *adapter = from_timer(adapter, t, service_timer);
7625 	unsigned long next_event_offset;
7626 
7627 	/* poll faster when waiting for link */
7628 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7629 		next_event_offset = HZ / 10;
7630 	else
7631 		next_event_offset = HZ * 2;
7632 
7633 	/* Reset the timer */
7634 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7635 
7636 	ixgbe_service_event_schedule(adapter);
7637 }
7638 
7639 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7640 {
7641 	struct ixgbe_hw *hw = &adapter->hw;
7642 	u32 status;
7643 
7644 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7645 		return;
7646 
7647 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7648 
7649 	if (!hw->phy.ops.handle_lasi)
7650 		return;
7651 
7652 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7653 	if (status != IXGBE_ERR_OVERTEMP)
7654 		return;
7655 
7656 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7657 }
7658 
7659 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7660 {
7661 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7662 		return;
7663 
7664 	/* If we're already down, removing or resetting, just bail */
7665 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7666 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7667 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7668 		return;
7669 
7670 	ixgbe_dump(adapter);
7671 	netdev_err(adapter->netdev, "Reset adapter\n");
7672 	adapter->tx_timeout_count++;
7673 
7674 	rtnl_lock();
7675 	ixgbe_reinit_locked(adapter);
7676 	rtnl_unlock();
7677 }
7678 
7679 /**
7680  * ixgbe_service_task - manages and runs subtasks
7681  * @work: pointer to work_struct containing our data
7682  **/
7683 static void ixgbe_service_task(struct work_struct *work)
7684 {
7685 	struct ixgbe_adapter *adapter = container_of(work,
7686 						     struct ixgbe_adapter,
7687 						     service_task);
7688 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7689 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7690 			rtnl_lock();
7691 			ixgbe_down(adapter);
7692 			rtnl_unlock();
7693 		}
7694 		ixgbe_service_event_complete(adapter);
7695 		return;
7696 	}
7697 	if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7698 		rtnl_lock();
7699 		adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7700 		udp_tunnel_get_rx_info(adapter->netdev);
7701 		rtnl_unlock();
7702 	}
7703 	ixgbe_reset_subtask(adapter);
7704 	ixgbe_phy_interrupt_subtask(adapter);
7705 	ixgbe_sfp_detection_subtask(adapter);
7706 	ixgbe_sfp_link_config_subtask(adapter);
7707 	ixgbe_check_overtemp_subtask(adapter);
7708 	ixgbe_watchdog_subtask(adapter);
7709 	ixgbe_fdir_reinit_subtask(adapter);
7710 	ixgbe_check_hang_subtask(adapter);
7711 
7712 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7713 		ixgbe_ptp_overflow_check(adapter);
7714 		ixgbe_ptp_rx_hang(adapter);
7715 		ixgbe_ptp_tx_hang(adapter);
7716 	}
7717 
7718 	ixgbe_service_event_complete(adapter);
7719 }
7720 
7721 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7722 		     struct ixgbe_tx_buffer *first,
7723 		     u8 *hdr_len)
7724 {
7725 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7726 	struct sk_buff *skb = first->skb;
7727 	union {
7728 		struct iphdr *v4;
7729 		struct ipv6hdr *v6;
7730 		unsigned char *hdr;
7731 	} ip;
7732 	union {
7733 		struct tcphdr *tcp;
7734 		unsigned char *hdr;
7735 	} l4;
7736 	u32 paylen, l4_offset;
7737 	int err;
7738 
7739 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7740 		return 0;
7741 
7742 	if (!skb_is_gso(skb))
7743 		return 0;
7744 
7745 	err = skb_cow_head(skb, 0);
7746 	if (err < 0)
7747 		return err;
7748 
7749 	if (eth_p_mpls(first->protocol))
7750 		ip.hdr = skb_inner_network_header(skb);
7751 	else
7752 		ip.hdr = skb_network_header(skb);
7753 	l4.hdr = skb_checksum_start(skb);
7754 
7755 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7756 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7757 
7758 	/* initialize outer IP header fields */
7759 	if (ip.v4->version == 4) {
7760 		unsigned char *csum_start = skb_checksum_start(skb);
7761 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
7762 
7763 		/* IP header will have to cancel out any data that
7764 		 * is not a part of the outer IP header
7765 		 */
7766 		ip.v4->check = csum_fold(csum_partial(trans_start,
7767 						      csum_start - trans_start,
7768 						      0));
7769 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7770 
7771 		ip.v4->tot_len = 0;
7772 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7773 				   IXGBE_TX_FLAGS_CSUM |
7774 				   IXGBE_TX_FLAGS_IPV4;
7775 	} else {
7776 		ip.v6->payload_len = 0;
7777 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7778 				   IXGBE_TX_FLAGS_CSUM;
7779 	}
7780 
7781 	/* determine offset of inner transport header */
7782 	l4_offset = l4.hdr - skb->data;
7783 
7784 	/* compute length of segmentation header */
7785 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7786 
7787 	/* remove payload length from inner checksum */
7788 	paylen = skb->len - l4_offset;
7789 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
7790 
7791 	/* update gso size and bytecount with header size */
7792 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7793 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7794 
7795 	/* mss_l4len_id: use 0 as index for TSO */
7796 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7797 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7798 
7799 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7800 	vlan_macip_lens = l4.hdr - ip.hdr;
7801 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7802 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7803 
7804 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
7805 			  mss_l4len_idx);
7806 
7807 	return 1;
7808 }
7809 
7810 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7811 {
7812 	unsigned int offset = 0;
7813 
7814 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7815 
7816 	return offset == skb_checksum_start_offset(skb);
7817 }
7818 
7819 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7820 			  struct ixgbe_tx_buffer *first,
7821 			  struct ixgbe_ipsec_tx_data *itd)
7822 {
7823 	struct sk_buff *skb = first->skb;
7824 	u32 vlan_macip_lens = 0;
7825 	u32 fceof_saidx = 0;
7826 	u32 type_tucmd = 0;
7827 
7828 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7829 csum_failed:
7830 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7831 					 IXGBE_TX_FLAGS_CC)))
7832 			return;
7833 		goto no_csum;
7834 	}
7835 
7836 	switch (skb->csum_offset) {
7837 	case offsetof(struct tcphdr, check):
7838 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7839 		/* fall through */
7840 	case offsetof(struct udphdr, check):
7841 		break;
7842 	case offsetof(struct sctphdr, checksum):
7843 		/* validate that this is actually an SCTP request */
7844 		if (((first->protocol == htons(ETH_P_IP)) &&
7845 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7846 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7847 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7848 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7849 			break;
7850 		}
7851 		/* fall through */
7852 	default:
7853 		skb_checksum_help(skb);
7854 		goto csum_failed;
7855 	}
7856 
7857 	/* update TX checksum flag */
7858 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7859 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7860 			  skb_network_offset(skb);
7861 no_csum:
7862 	/* vlan_macip_lens: MACLEN, VLAN tag */
7863 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7864 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7865 
7866 	if (first->tx_flags & IXGBE_TX_FLAGS_IPSEC) {
7867 		fceof_saidx |= itd->sa_idx;
7868 		type_tucmd |= itd->flags | itd->trailer_len;
7869 	}
7870 
7871 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
7872 }
7873 
7874 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7875 	((_flag <= _result) ? \
7876 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7877 	 ((u32)(_input & _flag) / (_flag / _result)))
7878 
7879 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7880 {
7881 	/* set type for advanced descriptor with frame checksum insertion */
7882 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7883 		       IXGBE_ADVTXD_DCMD_DEXT |
7884 		       IXGBE_ADVTXD_DCMD_IFCS;
7885 
7886 	/* set HW vlan bit if vlan is present */
7887 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7888 				   IXGBE_ADVTXD_DCMD_VLE);
7889 
7890 	/* set segmentation enable bits for TSO/FSO */
7891 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7892 				   IXGBE_ADVTXD_DCMD_TSE);
7893 
7894 	/* set timestamp bit if present */
7895 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7896 				   IXGBE_ADVTXD_MAC_TSTAMP);
7897 
7898 	/* insert frame checksum */
7899 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7900 
7901 	return cmd_type;
7902 }
7903 
7904 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7905 				   u32 tx_flags, unsigned int paylen)
7906 {
7907 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7908 
7909 	/* enable L4 checksum for TSO and TX checksum offload */
7910 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7911 					IXGBE_TX_FLAGS_CSUM,
7912 					IXGBE_ADVTXD_POPTS_TXSM);
7913 
7914 	/* enable IPv4 checksum for TSO */
7915 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7916 					IXGBE_TX_FLAGS_IPV4,
7917 					IXGBE_ADVTXD_POPTS_IXSM);
7918 
7919 	/* enable IPsec */
7920 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7921 					IXGBE_TX_FLAGS_IPSEC,
7922 					IXGBE_ADVTXD_POPTS_IPSEC);
7923 
7924 	/*
7925 	 * Check Context must be set if Tx switch is enabled, which it
7926 	 * always is for case where virtual functions are running
7927 	 */
7928 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7929 					IXGBE_TX_FLAGS_CC,
7930 					IXGBE_ADVTXD_CC);
7931 
7932 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7933 }
7934 
7935 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7936 {
7937 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7938 
7939 	/* Herbert's original patch had:
7940 	 *  smp_mb__after_netif_stop_queue();
7941 	 * but since that doesn't exist yet, just open code it.
7942 	 */
7943 	smp_mb();
7944 
7945 	/* We need to check again in a case another CPU has just
7946 	 * made room available.
7947 	 */
7948 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7949 		return -EBUSY;
7950 
7951 	/* A reprieve! - use start_queue because it doesn't call schedule */
7952 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7953 	++tx_ring->tx_stats.restart_queue;
7954 	return 0;
7955 }
7956 
7957 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7958 {
7959 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7960 		return 0;
7961 
7962 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7963 }
7964 
7965 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7966 		       IXGBE_TXD_CMD_RS)
7967 
7968 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7969 			struct ixgbe_tx_buffer *first,
7970 			const u8 hdr_len)
7971 {
7972 	struct sk_buff *skb = first->skb;
7973 	struct ixgbe_tx_buffer *tx_buffer;
7974 	union ixgbe_adv_tx_desc *tx_desc;
7975 	struct skb_frag_struct *frag;
7976 	dma_addr_t dma;
7977 	unsigned int data_len, size;
7978 	u32 tx_flags = first->tx_flags;
7979 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7980 	u16 i = tx_ring->next_to_use;
7981 
7982 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7983 
7984 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7985 
7986 	size = skb_headlen(skb);
7987 	data_len = skb->data_len;
7988 
7989 #ifdef IXGBE_FCOE
7990 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7991 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7992 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7993 			data_len = 0;
7994 		} else {
7995 			data_len -= sizeof(struct fcoe_crc_eof);
7996 		}
7997 	}
7998 
7999 #endif
8000 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
8001 
8002 	tx_buffer = first;
8003 
8004 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
8005 		if (dma_mapping_error(tx_ring->dev, dma))
8006 			goto dma_error;
8007 
8008 		/* record length, and DMA address */
8009 		dma_unmap_len_set(tx_buffer, len, size);
8010 		dma_unmap_addr_set(tx_buffer, dma, dma);
8011 
8012 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
8013 
8014 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
8015 			tx_desc->read.cmd_type_len =
8016 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
8017 
8018 			i++;
8019 			tx_desc++;
8020 			if (i == tx_ring->count) {
8021 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8022 				i = 0;
8023 			}
8024 			tx_desc->read.olinfo_status = 0;
8025 
8026 			dma += IXGBE_MAX_DATA_PER_TXD;
8027 			size -= IXGBE_MAX_DATA_PER_TXD;
8028 
8029 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
8030 		}
8031 
8032 		if (likely(!data_len))
8033 			break;
8034 
8035 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
8036 
8037 		i++;
8038 		tx_desc++;
8039 		if (i == tx_ring->count) {
8040 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8041 			i = 0;
8042 		}
8043 		tx_desc->read.olinfo_status = 0;
8044 
8045 #ifdef IXGBE_FCOE
8046 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
8047 #else
8048 		size = skb_frag_size(frag);
8049 #endif
8050 		data_len -= size;
8051 
8052 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
8053 				       DMA_TO_DEVICE);
8054 
8055 		tx_buffer = &tx_ring->tx_buffer_info[i];
8056 	}
8057 
8058 	/* write last descriptor with RS and EOP bits */
8059 	cmd_type |= size | IXGBE_TXD_CMD;
8060 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8061 
8062 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
8063 
8064 	/* set the timestamp */
8065 	first->time_stamp = jiffies;
8066 
8067 	/*
8068 	 * Force memory writes to complete before letting h/w know there
8069 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
8070 	 * memory model archs, such as IA-64).
8071 	 *
8072 	 * We also need this memory barrier to make certain all of the
8073 	 * status bits have been updated before next_to_watch is written.
8074 	 */
8075 	wmb();
8076 
8077 	/* set next_to_watch value indicating a packet is present */
8078 	first->next_to_watch = tx_desc;
8079 
8080 	i++;
8081 	if (i == tx_ring->count)
8082 		i = 0;
8083 
8084 	tx_ring->next_to_use = i;
8085 
8086 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8087 
8088 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
8089 		writel(i, tx_ring->tail);
8090 
8091 		/* we need this if more than one processor can write to our tail
8092 		 * at a time, it synchronizes IO on IA64/Altix systems
8093 		 */
8094 		mmiowb();
8095 	}
8096 
8097 	return 0;
8098 dma_error:
8099 	dev_err(tx_ring->dev, "TX DMA map failed\n");
8100 
8101 	/* clear dma mappings for failed tx_buffer_info map */
8102 	for (;;) {
8103 		tx_buffer = &tx_ring->tx_buffer_info[i];
8104 		if (dma_unmap_len(tx_buffer, len))
8105 			dma_unmap_page(tx_ring->dev,
8106 				       dma_unmap_addr(tx_buffer, dma),
8107 				       dma_unmap_len(tx_buffer, len),
8108 				       DMA_TO_DEVICE);
8109 		dma_unmap_len_set(tx_buffer, len, 0);
8110 		if (tx_buffer == first)
8111 			break;
8112 		if (i == 0)
8113 			i += tx_ring->count;
8114 		i--;
8115 	}
8116 
8117 	dev_kfree_skb_any(first->skb);
8118 	first->skb = NULL;
8119 
8120 	tx_ring->next_to_use = i;
8121 
8122 	return -1;
8123 }
8124 
8125 static void ixgbe_atr(struct ixgbe_ring *ring,
8126 		      struct ixgbe_tx_buffer *first)
8127 {
8128 	struct ixgbe_q_vector *q_vector = ring->q_vector;
8129 	union ixgbe_atr_hash_dword input = { .dword = 0 };
8130 	union ixgbe_atr_hash_dword common = { .dword = 0 };
8131 	union {
8132 		unsigned char *network;
8133 		struct iphdr *ipv4;
8134 		struct ipv6hdr *ipv6;
8135 	} hdr;
8136 	struct tcphdr *th;
8137 	unsigned int hlen;
8138 	struct sk_buff *skb;
8139 	__be16 vlan_id;
8140 	int l4_proto;
8141 
8142 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
8143 	if (!q_vector)
8144 		return;
8145 
8146 	/* do nothing if sampling is disabled */
8147 	if (!ring->atr_sample_rate)
8148 		return;
8149 
8150 	ring->atr_count++;
8151 
8152 	/* currently only IPv4/IPv6 with TCP is supported */
8153 	if ((first->protocol != htons(ETH_P_IP)) &&
8154 	    (first->protocol != htons(ETH_P_IPV6)))
8155 		return;
8156 
8157 	/* snag network header to get L4 type and address */
8158 	skb = first->skb;
8159 	hdr.network = skb_network_header(skb);
8160 	if (unlikely(hdr.network <= skb->data))
8161 		return;
8162 	if (skb->encapsulation &&
8163 	    first->protocol == htons(ETH_P_IP) &&
8164 	    hdr.ipv4->protocol == IPPROTO_UDP) {
8165 		struct ixgbe_adapter *adapter = q_vector->adapter;
8166 
8167 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
8168 			     VXLAN_HEADROOM))
8169 			return;
8170 
8171 		/* verify the port is recognized as VXLAN */
8172 		if (adapter->vxlan_port &&
8173 		    udp_hdr(skb)->dest == adapter->vxlan_port)
8174 			hdr.network = skb_inner_network_header(skb);
8175 
8176 		if (adapter->geneve_port &&
8177 		    udp_hdr(skb)->dest == adapter->geneve_port)
8178 			hdr.network = skb_inner_network_header(skb);
8179 	}
8180 
8181 	/* Make sure we have at least [minimum IPv4 header + TCP]
8182 	 * or [IPv6 header] bytes
8183 	 */
8184 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8185 		return;
8186 
8187 	/* Currently only IPv4/IPv6 with TCP is supported */
8188 	switch (hdr.ipv4->version) {
8189 	case IPVERSION:
8190 		/* access ihl as u8 to avoid unaligned access on ia64 */
8191 		hlen = (hdr.network[0] & 0x0F) << 2;
8192 		l4_proto = hdr.ipv4->protocol;
8193 		break;
8194 	case 6:
8195 		hlen = hdr.network - skb->data;
8196 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8197 		hlen -= hdr.network - skb->data;
8198 		break;
8199 	default:
8200 		return;
8201 	}
8202 
8203 	if (l4_proto != IPPROTO_TCP)
8204 		return;
8205 
8206 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
8207 		     hlen + sizeof(struct tcphdr)))
8208 		return;
8209 
8210 	th = (struct tcphdr *)(hdr.network + hlen);
8211 
8212 	/* skip this packet since the socket is closing */
8213 	if (th->fin)
8214 		return;
8215 
8216 	/* sample on all syn packets or once every atr sample count */
8217 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8218 		return;
8219 
8220 	/* reset sample count */
8221 	ring->atr_count = 0;
8222 
8223 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8224 
8225 	/*
8226 	 * src and dst are inverted, think how the receiver sees them
8227 	 *
8228 	 * The input is broken into two sections, a non-compressed section
8229 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8230 	 * is XORed together and stored in the compressed dword.
8231 	 */
8232 	input.formatted.vlan_id = vlan_id;
8233 
8234 	/*
8235 	 * since src port and flex bytes occupy the same word XOR them together
8236 	 * and write the value to source port portion of compressed dword
8237 	 */
8238 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8239 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8240 	else
8241 		common.port.src ^= th->dest ^ first->protocol;
8242 	common.port.dst ^= th->source;
8243 
8244 	switch (hdr.ipv4->version) {
8245 	case IPVERSION:
8246 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8247 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8248 		break;
8249 	case 6:
8250 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8251 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8252 			     hdr.ipv6->saddr.s6_addr32[1] ^
8253 			     hdr.ipv6->saddr.s6_addr32[2] ^
8254 			     hdr.ipv6->saddr.s6_addr32[3] ^
8255 			     hdr.ipv6->daddr.s6_addr32[0] ^
8256 			     hdr.ipv6->daddr.s6_addr32[1] ^
8257 			     hdr.ipv6->daddr.s6_addr32[2] ^
8258 			     hdr.ipv6->daddr.s6_addr32[3];
8259 		break;
8260 	default:
8261 		break;
8262 	}
8263 
8264 	if (hdr.network != skb_network_header(skb))
8265 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8266 
8267 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
8268 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8269 					      input, common, ring->queue_index);
8270 }
8271 
8272 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8273 			      void *accel_priv, select_queue_fallback_t fallback)
8274 {
8275 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
8276 	struct ixgbe_adapter *adapter;
8277 	int txq;
8278 #ifdef IXGBE_FCOE
8279 	struct ixgbe_ring_feature *f;
8280 #endif
8281 
8282 	if (fwd_adapter) {
8283 		adapter = netdev_priv(dev);
8284 		txq = reciprocal_scale(skb_get_hash(skb),
8285 				       adapter->num_rx_queues_per_pool);
8286 
8287 		return txq + fwd_adapter->tx_base_queue;
8288 	}
8289 
8290 #ifdef IXGBE_FCOE
8291 
8292 	/*
8293 	 * only execute the code below if protocol is FCoE
8294 	 * or FIP and we have FCoE enabled on the adapter
8295 	 */
8296 	switch (vlan_get_protocol(skb)) {
8297 	case htons(ETH_P_FCOE):
8298 	case htons(ETH_P_FIP):
8299 		adapter = netdev_priv(dev);
8300 
8301 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
8302 			break;
8303 		/* fall through */
8304 	default:
8305 		return fallback(dev, skb);
8306 	}
8307 
8308 	f = &adapter->ring_feature[RING_F_FCOE];
8309 
8310 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8311 					   smp_processor_id();
8312 
8313 	while (txq >= f->indices)
8314 		txq -= f->indices;
8315 
8316 	return txq + f->offset;
8317 #else
8318 	return fallback(dev, skb);
8319 #endif
8320 }
8321 
8322 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
8323 			       struct xdp_buff *xdp)
8324 {
8325 	struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
8326 	struct ixgbe_tx_buffer *tx_buffer;
8327 	union ixgbe_adv_tx_desc *tx_desc;
8328 	u32 len, cmd_type;
8329 	dma_addr_t dma;
8330 	u16 i;
8331 
8332 	len = xdp->data_end - xdp->data;
8333 
8334 	if (unlikely(!ixgbe_desc_unused(ring)))
8335 		return IXGBE_XDP_CONSUMED;
8336 
8337 	dma = dma_map_single(ring->dev, xdp->data, len, DMA_TO_DEVICE);
8338 	if (dma_mapping_error(ring->dev, dma))
8339 		return IXGBE_XDP_CONSUMED;
8340 
8341 	/* record the location of the first descriptor for this packet */
8342 	tx_buffer = &ring->tx_buffer_info[ring->next_to_use];
8343 	tx_buffer->bytecount = len;
8344 	tx_buffer->gso_segs = 1;
8345 	tx_buffer->protocol = 0;
8346 
8347 	i = ring->next_to_use;
8348 	tx_desc = IXGBE_TX_DESC(ring, i);
8349 
8350 	dma_unmap_len_set(tx_buffer, len, len);
8351 	dma_unmap_addr_set(tx_buffer, dma, dma);
8352 	tx_buffer->data = xdp->data;
8353 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
8354 
8355 	/* put descriptor type bits */
8356 	cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8357 		   IXGBE_ADVTXD_DCMD_DEXT |
8358 		   IXGBE_ADVTXD_DCMD_IFCS;
8359 	cmd_type |= len | IXGBE_TXD_CMD;
8360 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8361 	tx_desc->read.olinfo_status =
8362 		cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
8363 
8364 	/* Avoid any potential race with xdp_xmit and cleanup */
8365 	smp_wmb();
8366 
8367 	/* set next_to_watch value indicating a packet is present */
8368 	i++;
8369 	if (i == ring->count)
8370 		i = 0;
8371 
8372 	tx_buffer->next_to_watch = tx_desc;
8373 	ring->next_to_use = i;
8374 
8375 	return IXGBE_XDP_TX;
8376 }
8377 
8378 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8379 			  struct ixgbe_adapter *adapter,
8380 			  struct ixgbe_ring *tx_ring)
8381 {
8382 	struct ixgbe_tx_buffer *first;
8383 	int tso;
8384 	u32 tx_flags = 0;
8385 	unsigned short f;
8386 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
8387 	struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
8388 	__be16 protocol = skb->protocol;
8389 	u8 hdr_len = 0;
8390 
8391 	/*
8392 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8393 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8394 	 *       + 2 desc gap to keep tail from touching head,
8395 	 *       + 1 desc for context descriptor,
8396 	 * otherwise try next time
8397 	 */
8398 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8399 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
8400 
8401 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8402 		tx_ring->tx_stats.tx_busy++;
8403 		return NETDEV_TX_BUSY;
8404 	}
8405 
8406 	/* record the location of the first descriptor for this packet */
8407 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8408 	first->skb = skb;
8409 	first->bytecount = skb->len;
8410 	first->gso_segs = 1;
8411 
8412 	/* if we have a HW VLAN tag being added default to the HW one */
8413 	if (skb_vlan_tag_present(skb)) {
8414 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8415 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8416 	/* else if it is a SW VLAN check the next protocol and store the tag */
8417 	} else if (protocol == htons(ETH_P_8021Q)) {
8418 		struct vlan_hdr *vhdr, _vhdr;
8419 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8420 		if (!vhdr)
8421 			goto out_drop;
8422 
8423 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8424 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
8425 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8426 	}
8427 	protocol = vlan_get_protocol(skb);
8428 
8429 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8430 	    adapter->ptp_clock) {
8431 		if (!test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8432 					   &adapter->state)) {
8433 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8434 			tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8435 
8436 			/* schedule check for Tx timestamp */
8437 			adapter->ptp_tx_skb = skb_get(skb);
8438 			adapter->ptp_tx_start = jiffies;
8439 			schedule_work(&adapter->ptp_tx_work);
8440 		} else {
8441 			adapter->tx_hwtstamp_skipped++;
8442 		}
8443 	}
8444 
8445 	skb_tx_timestamp(skb);
8446 
8447 #ifdef CONFIG_PCI_IOV
8448 	/*
8449 	 * Use the l2switch_enable flag - would be false if the DMA
8450 	 * Tx switch had been disabled.
8451 	 */
8452 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8453 		tx_flags |= IXGBE_TX_FLAGS_CC;
8454 
8455 #endif
8456 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8457 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8458 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8459 	     (skb->priority != TC_PRIO_CONTROL))) {
8460 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8461 		tx_flags |= (skb->priority & 0x7) <<
8462 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8463 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8464 			struct vlan_ethhdr *vhdr;
8465 
8466 			if (skb_cow_head(skb, 0))
8467 				goto out_drop;
8468 			vhdr = (struct vlan_ethhdr *)skb->data;
8469 			vhdr->h_vlan_TCI = htons(tx_flags >>
8470 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
8471 		} else {
8472 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8473 		}
8474 	}
8475 
8476 	/* record initial flags and protocol */
8477 	first->tx_flags = tx_flags;
8478 	first->protocol = protocol;
8479 
8480 #ifdef IXGBE_FCOE
8481 	/* setup tx offload for FCoE */
8482 	if ((protocol == htons(ETH_P_FCOE)) &&
8483 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8484 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
8485 		if (tso < 0)
8486 			goto out_drop;
8487 
8488 		goto xmit_fcoe;
8489 	}
8490 
8491 #endif /* IXGBE_FCOE */
8492 
8493 #ifdef CONFIG_XFRM_OFFLOAD
8494 	if (skb->sp && !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
8495 		goto out_drop;
8496 #endif
8497 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
8498 	if (tso < 0)
8499 		goto out_drop;
8500 	else if (!tso)
8501 		ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
8502 
8503 	/* add the ATR filter if ATR is on */
8504 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8505 		ixgbe_atr(tx_ring, first);
8506 
8507 #ifdef IXGBE_FCOE
8508 xmit_fcoe:
8509 #endif /* IXGBE_FCOE */
8510 	if (ixgbe_tx_map(tx_ring, first, hdr_len))
8511 		goto cleanup_tx_timestamp;
8512 
8513 	return NETDEV_TX_OK;
8514 
8515 out_drop:
8516 	dev_kfree_skb_any(first->skb);
8517 	first->skb = NULL;
8518 cleanup_tx_timestamp:
8519 	if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8520 		dev_kfree_skb_any(adapter->ptp_tx_skb);
8521 		adapter->ptp_tx_skb = NULL;
8522 		cancel_work_sync(&adapter->ptp_tx_work);
8523 		clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8524 	}
8525 
8526 	return NETDEV_TX_OK;
8527 }
8528 
8529 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8530 				      struct net_device *netdev,
8531 				      struct ixgbe_ring *ring)
8532 {
8533 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8534 	struct ixgbe_ring *tx_ring;
8535 
8536 	/*
8537 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
8538 	 * in order to meet this minimum size requirement.
8539 	 */
8540 	if (skb_put_padto(skb, 17))
8541 		return NETDEV_TX_OK;
8542 
8543 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
8544 
8545 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8546 }
8547 
8548 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8549 				    struct net_device *netdev)
8550 {
8551 	return __ixgbe_xmit_frame(skb, netdev, NULL);
8552 }
8553 
8554 /**
8555  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8556  * @netdev: network interface device structure
8557  * @p: pointer to an address structure
8558  *
8559  * Returns 0 on success, negative on failure
8560  **/
8561 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8562 {
8563 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8564 	struct ixgbe_hw *hw = &adapter->hw;
8565 	struct sockaddr *addr = p;
8566 
8567 	if (!is_valid_ether_addr(addr->sa_data))
8568 		return -EADDRNOTAVAIL;
8569 
8570 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
8571 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8572 
8573 	ixgbe_mac_set_default_filter(adapter);
8574 
8575 	return 0;
8576 }
8577 
8578 static int
8579 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8580 {
8581 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8582 	struct ixgbe_hw *hw = &adapter->hw;
8583 	u16 value;
8584 	int rc;
8585 
8586 	if (prtad != hw->phy.mdio.prtad)
8587 		return -EINVAL;
8588 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8589 	if (!rc)
8590 		rc = value;
8591 	return rc;
8592 }
8593 
8594 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8595 			    u16 addr, u16 value)
8596 {
8597 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8598 	struct ixgbe_hw *hw = &adapter->hw;
8599 
8600 	if (prtad != hw->phy.mdio.prtad)
8601 		return -EINVAL;
8602 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8603 }
8604 
8605 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8606 {
8607 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8608 
8609 	switch (cmd) {
8610 	case SIOCSHWTSTAMP:
8611 		return ixgbe_ptp_set_ts_config(adapter, req);
8612 	case SIOCGHWTSTAMP:
8613 		return ixgbe_ptp_get_ts_config(adapter, req);
8614 	case SIOCGMIIPHY:
8615 		if (!adapter->hw.phy.ops.read_reg)
8616 			return -EOPNOTSUPP;
8617 		/* fall through */
8618 	default:
8619 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8620 	}
8621 }
8622 
8623 /**
8624  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8625  * netdev->dev_addrs
8626  * @dev: network interface device structure
8627  *
8628  * Returns non-zero on failure
8629  **/
8630 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8631 {
8632 	int err = 0;
8633 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8634 	struct ixgbe_hw *hw = &adapter->hw;
8635 
8636 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8637 		rtnl_lock();
8638 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8639 		rtnl_unlock();
8640 
8641 		/* update SAN MAC vmdq pool selection */
8642 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8643 	}
8644 	return err;
8645 }
8646 
8647 /**
8648  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8649  * netdev->dev_addrs
8650  * @dev: network interface device structure
8651  *
8652  * Returns non-zero on failure
8653  **/
8654 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8655 {
8656 	int err = 0;
8657 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8658 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8659 
8660 	if (is_valid_ether_addr(mac->san_addr)) {
8661 		rtnl_lock();
8662 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8663 		rtnl_unlock();
8664 	}
8665 	return err;
8666 }
8667 
8668 #ifdef CONFIG_NET_POLL_CONTROLLER
8669 /*
8670  * Polling 'interrupt' - used by things like netconsole to send skbs
8671  * without having to re-enable interrupts. It's not called while
8672  * the interrupt routine is executing.
8673  */
8674 static void ixgbe_netpoll(struct net_device *netdev)
8675 {
8676 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8677 	int i;
8678 
8679 	/* if interface is down do nothing */
8680 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8681 		return;
8682 
8683 	/* loop through and schedule all active queues */
8684 	for (i = 0; i < adapter->num_q_vectors; i++)
8685 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8686 }
8687 
8688 #endif
8689 
8690 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
8691 				   struct ixgbe_ring *ring)
8692 {
8693 	u64 bytes, packets;
8694 	unsigned int start;
8695 
8696 	if (ring) {
8697 		do {
8698 			start = u64_stats_fetch_begin_irq(&ring->syncp);
8699 			packets = ring->stats.packets;
8700 			bytes   = ring->stats.bytes;
8701 		} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8702 		stats->tx_packets += packets;
8703 		stats->tx_bytes   += bytes;
8704 	}
8705 }
8706 
8707 static void ixgbe_get_stats64(struct net_device *netdev,
8708 			      struct rtnl_link_stats64 *stats)
8709 {
8710 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8711 	int i;
8712 
8713 	rcu_read_lock();
8714 	for (i = 0; i < adapter->num_rx_queues; i++) {
8715 		struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
8716 		u64 bytes, packets;
8717 		unsigned int start;
8718 
8719 		if (ring) {
8720 			do {
8721 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8722 				packets = ring->stats.packets;
8723 				bytes   = ring->stats.bytes;
8724 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8725 			stats->rx_packets += packets;
8726 			stats->rx_bytes   += bytes;
8727 		}
8728 	}
8729 
8730 	for (i = 0; i < adapter->num_tx_queues; i++) {
8731 		struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
8732 
8733 		ixgbe_get_ring_stats64(stats, ring);
8734 	}
8735 	for (i = 0; i < adapter->num_xdp_queues; i++) {
8736 		struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
8737 
8738 		ixgbe_get_ring_stats64(stats, ring);
8739 	}
8740 	rcu_read_unlock();
8741 
8742 	/* following stats updated by ixgbe_watchdog_task() */
8743 	stats->multicast	= netdev->stats.multicast;
8744 	stats->rx_errors	= netdev->stats.rx_errors;
8745 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8746 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8747 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8748 }
8749 
8750 #ifdef CONFIG_IXGBE_DCB
8751 /**
8752  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8753  * @adapter: pointer to ixgbe_adapter
8754  * @tc: number of traffic classes currently enabled
8755  *
8756  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8757  * 802.1Q priority maps to a packet buffer that exists.
8758  */
8759 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8760 {
8761 	struct ixgbe_hw *hw = &adapter->hw;
8762 	u32 reg, rsave;
8763 	int i;
8764 
8765 	/* 82598 have a static priority to TC mapping that can not
8766 	 * be changed so no validation is needed.
8767 	 */
8768 	if (hw->mac.type == ixgbe_mac_82598EB)
8769 		return;
8770 
8771 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8772 	rsave = reg;
8773 
8774 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8775 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8776 
8777 		/* If up2tc is out of bounds default to zero */
8778 		if (up2tc > tc)
8779 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8780 	}
8781 
8782 	if (reg != rsave)
8783 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8784 
8785 	return;
8786 }
8787 
8788 /**
8789  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8790  * @adapter: Pointer to adapter struct
8791  *
8792  * Populate the netdev user priority to tc map
8793  */
8794 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8795 {
8796 	struct net_device *dev = adapter->netdev;
8797 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8798 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8799 	u8 prio;
8800 
8801 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8802 		u8 tc = 0;
8803 
8804 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8805 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8806 		else if (ets)
8807 			tc = ets->prio_tc[prio];
8808 
8809 		netdev_set_prio_tc_map(dev, prio, tc);
8810 	}
8811 }
8812 
8813 #endif /* CONFIG_IXGBE_DCB */
8814 /**
8815  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8816  *
8817  * @dev: net device to configure
8818  * @tc: number of traffic classes to enable
8819  */
8820 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8821 {
8822 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8823 	struct ixgbe_hw *hw = &adapter->hw;
8824 
8825 	/* Hardware supports up to 8 traffic classes */
8826 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8827 		return -EINVAL;
8828 
8829 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8830 		return -EINVAL;
8831 
8832 	/* Hardware has to reinitialize queues and interrupts to
8833 	 * match packet buffer alignment. Unfortunately, the
8834 	 * hardware is not flexible enough to do this dynamically.
8835 	 */
8836 	if (netif_running(dev))
8837 		ixgbe_close(dev);
8838 	else
8839 		ixgbe_reset(adapter);
8840 
8841 	ixgbe_clear_interrupt_scheme(adapter);
8842 
8843 #ifdef CONFIG_IXGBE_DCB
8844 	if (tc) {
8845 		netdev_set_num_tc(dev, tc);
8846 		ixgbe_set_prio_tc_map(adapter);
8847 
8848 		adapter->hw_tcs = tc;
8849 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8850 
8851 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8852 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8853 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8854 		}
8855 	} else {
8856 		netdev_reset_tc(dev);
8857 
8858 		/* To support macvlan offload we have to use num_tc to
8859 		 * restrict the queues that can be used by the device.
8860 		 * By doing this we can avoid reporting a false number of
8861 		 * queues.
8862 		 */
8863 		if (!tc && adapter->num_rx_pools > 1)
8864 			netdev_set_num_tc(dev, 1);
8865 
8866 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8867 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8868 
8869 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8870 		adapter->hw_tcs = tc;
8871 
8872 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8873 		adapter->dcb_cfg.pfc_mode_enable = false;
8874 	}
8875 
8876 	ixgbe_validate_rtr(adapter, tc);
8877 
8878 #endif /* CONFIG_IXGBE_DCB */
8879 	ixgbe_init_interrupt_scheme(adapter);
8880 
8881 	if (netif_running(dev))
8882 		return ixgbe_open(dev);
8883 
8884 	return 0;
8885 }
8886 
8887 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8888 			       struct tc_cls_u32_offload *cls)
8889 {
8890 	u32 hdl = cls->knode.handle;
8891 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8892 	u32 loc = cls->knode.handle & 0xfffff;
8893 	int err = 0, i, j;
8894 	struct ixgbe_jump_table *jump = NULL;
8895 
8896 	if (loc > IXGBE_MAX_HW_ENTRIES)
8897 		return -EINVAL;
8898 
8899 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8900 		return -EINVAL;
8901 
8902 	/* Clear this filter in the link data it is associated with */
8903 	if (uhtid != 0x800) {
8904 		jump = adapter->jump_tables[uhtid];
8905 		if (!jump)
8906 			return -EINVAL;
8907 		if (!test_bit(loc - 1, jump->child_loc_map))
8908 			return -EINVAL;
8909 		clear_bit(loc - 1, jump->child_loc_map);
8910 	}
8911 
8912 	/* Check if the filter being deleted is a link */
8913 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8914 		jump = adapter->jump_tables[i];
8915 		if (jump && jump->link_hdl == hdl) {
8916 			/* Delete filters in the hardware in the child hash
8917 			 * table associated with this link
8918 			 */
8919 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8920 				if (!test_bit(j, jump->child_loc_map))
8921 					continue;
8922 				spin_lock(&adapter->fdir_perfect_lock);
8923 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8924 								      NULL,
8925 								      j + 1);
8926 				spin_unlock(&adapter->fdir_perfect_lock);
8927 				clear_bit(j, jump->child_loc_map);
8928 			}
8929 			/* Remove resources for this link */
8930 			kfree(jump->input);
8931 			kfree(jump->mask);
8932 			kfree(jump);
8933 			adapter->jump_tables[i] = NULL;
8934 			return err;
8935 		}
8936 	}
8937 
8938 	spin_lock(&adapter->fdir_perfect_lock);
8939 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8940 	spin_unlock(&adapter->fdir_perfect_lock);
8941 	return err;
8942 }
8943 
8944 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8945 					    struct tc_cls_u32_offload *cls)
8946 {
8947 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8948 
8949 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8950 		return -EINVAL;
8951 
8952 	/* This ixgbe devices do not support hash tables at the moment
8953 	 * so abort when given hash tables.
8954 	 */
8955 	if (cls->hnode.divisor > 0)
8956 		return -EINVAL;
8957 
8958 	set_bit(uhtid - 1, &adapter->tables);
8959 	return 0;
8960 }
8961 
8962 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
8963 					    struct tc_cls_u32_offload *cls)
8964 {
8965 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8966 
8967 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8968 		return -EINVAL;
8969 
8970 	clear_bit(uhtid - 1, &adapter->tables);
8971 	return 0;
8972 }
8973 
8974 #ifdef CONFIG_NET_CLS_ACT
8975 struct upper_walk_data {
8976 	struct ixgbe_adapter *adapter;
8977 	u64 action;
8978 	int ifindex;
8979 	u8 queue;
8980 };
8981 
8982 static int get_macvlan_queue(struct net_device *upper, void *_data)
8983 {
8984 	if (netif_is_macvlan(upper)) {
8985 		struct macvlan_dev *dfwd = netdev_priv(upper);
8986 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
8987 		struct upper_walk_data *data = _data;
8988 		struct ixgbe_adapter *adapter = data->adapter;
8989 		int ifindex = data->ifindex;
8990 
8991 		if (vadapter && vadapter->netdev->ifindex == ifindex) {
8992 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
8993 			data->action = data->queue;
8994 			return 1;
8995 		}
8996 	}
8997 
8998 	return 0;
8999 }
9000 
9001 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
9002 				  u8 *queue, u64 *action)
9003 {
9004 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
9005 	unsigned int num_vfs = adapter->num_vfs, vf;
9006 	struct upper_walk_data data;
9007 	struct net_device *upper;
9008 
9009 	/* redirect to a SRIOV VF */
9010 	for (vf = 0; vf < num_vfs; ++vf) {
9011 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
9012 		if (upper->ifindex == ifindex) {
9013 			*queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
9014 			*action = vf + 1;
9015 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
9016 			return 0;
9017 		}
9018 	}
9019 
9020 	/* redirect to a offloaded macvlan netdev */
9021 	data.adapter = adapter;
9022 	data.ifindex = ifindex;
9023 	data.action = 0;
9024 	data.queue = 0;
9025 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
9026 					  get_macvlan_queue, &data)) {
9027 		*action = data.action;
9028 		*queue = data.queue;
9029 
9030 		return 0;
9031 	}
9032 
9033 	return -EINVAL;
9034 }
9035 
9036 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9037 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9038 {
9039 	const struct tc_action *a;
9040 	LIST_HEAD(actions);
9041 	int err;
9042 
9043 	if (!tcf_exts_has_actions(exts))
9044 		return -EINVAL;
9045 
9046 	tcf_exts_to_list(exts, &actions);
9047 	list_for_each_entry(a, &actions, list) {
9048 
9049 		/* Drop action */
9050 		if (is_tcf_gact_shot(a)) {
9051 			*action = IXGBE_FDIR_DROP_QUEUE;
9052 			*queue = IXGBE_FDIR_DROP_QUEUE;
9053 			return 0;
9054 		}
9055 
9056 		/* Redirect to a VF or a offloaded macvlan */
9057 		if (is_tcf_mirred_egress_redirect(a)) {
9058 			struct net_device *dev = tcf_mirred_dev(a);
9059 
9060 			if (!dev)
9061 				return -EINVAL;
9062 			err = handle_redirect_action(adapter, dev->ifindex, queue,
9063 						     action);
9064 			if (err == 0)
9065 				return err;
9066 		}
9067 	}
9068 
9069 	return -EINVAL;
9070 }
9071 #else
9072 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9073 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9074 {
9075 	return -EINVAL;
9076 }
9077 #endif /* CONFIG_NET_CLS_ACT */
9078 
9079 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
9080 				    union ixgbe_atr_input *mask,
9081 				    struct tc_cls_u32_offload *cls,
9082 				    struct ixgbe_mat_field *field_ptr,
9083 				    struct ixgbe_nexthdr *nexthdr)
9084 {
9085 	int i, j, off;
9086 	__be32 val, m;
9087 	bool found_entry = false, found_jump_field = false;
9088 
9089 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
9090 		off = cls->knode.sel->keys[i].off;
9091 		val = cls->knode.sel->keys[i].val;
9092 		m = cls->knode.sel->keys[i].mask;
9093 
9094 		for (j = 0; field_ptr[j].val; j++) {
9095 			if (field_ptr[j].off == off) {
9096 				field_ptr[j].val(input, mask, val, m);
9097 				input->filter.formatted.flow_type |=
9098 					field_ptr[j].type;
9099 				found_entry = true;
9100 				break;
9101 			}
9102 		}
9103 		if (nexthdr) {
9104 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
9105 			    nexthdr->val == cls->knode.sel->keys[i].val &&
9106 			    nexthdr->mask == cls->knode.sel->keys[i].mask)
9107 				found_jump_field = true;
9108 			else
9109 				continue;
9110 		}
9111 	}
9112 
9113 	if (nexthdr && !found_jump_field)
9114 		return -EINVAL;
9115 
9116 	if (!found_entry)
9117 		return 0;
9118 
9119 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9120 				    IXGBE_ATR_L4TYPE_MASK;
9121 
9122 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9123 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9124 
9125 	return 0;
9126 }
9127 
9128 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9129 				  struct tc_cls_u32_offload *cls)
9130 {
9131 	__be16 protocol = cls->common.protocol;
9132 	u32 loc = cls->knode.handle & 0xfffff;
9133 	struct ixgbe_hw *hw = &adapter->hw;
9134 	struct ixgbe_mat_field *field_ptr;
9135 	struct ixgbe_fdir_filter *input = NULL;
9136 	union ixgbe_atr_input *mask = NULL;
9137 	struct ixgbe_jump_table *jump = NULL;
9138 	int i, err = -EINVAL;
9139 	u8 queue;
9140 	u32 uhtid, link_uhtid;
9141 
9142 	uhtid = TC_U32_USERHTID(cls->knode.handle);
9143 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9144 
9145 	/* At the moment cls_u32 jumps to network layer and skips past
9146 	 * L2 headers. The canonical method to match L2 frames is to use
9147 	 * negative values. However this is error prone at best but really
9148 	 * just broken because there is no way to "know" what sort of hdr
9149 	 * is in front of the network layer. Fix cls_u32 to support L2
9150 	 * headers when needed.
9151 	 */
9152 	if (protocol != htons(ETH_P_IP))
9153 		return err;
9154 
9155 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9156 		e_err(drv, "Location out of range\n");
9157 		return err;
9158 	}
9159 
9160 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
9161 	 * links and also the fields used to advance the parser across each
9162 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9163 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9164 	 * To add support for new nodes update ixgbe_model.h parse structures
9165 	 * this function _should_ be generic try not to hardcode values here.
9166 	 */
9167 	if (uhtid == 0x800) {
9168 		field_ptr = (adapter->jump_tables[0])->mat;
9169 	} else {
9170 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9171 			return err;
9172 		if (!adapter->jump_tables[uhtid])
9173 			return err;
9174 		field_ptr = (adapter->jump_tables[uhtid])->mat;
9175 	}
9176 
9177 	if (!field_ptr)
9178 		return err;
9179 
9180 	/* At this point we know the field_ptr is valid and need to either
9181 	 * build cls_u32 link or attach filter. Because adding a link to
9182 	 * a handle that does not exist is invalid and the same for adding
9183 	 * rules to handles that don't exist.
9184 	 */
9185 
9186 	if (link_uhtid) {
9187 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9188 
9189 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9190 			return err;
9191 
9192 		if (!test_bit(link_uhtid - 1, &adapter->tables))
9193 			return err;
9194 
9195 		/* Multiple filters as links to the same hash table are not
9196 		 * supported. To add a new filter with the same next header
9197 		 * but different match/jump conditions, create a new hash table
9198 		 * and link to it.
9199 		 */
9200 		if (adapter->jump_tables[link_uhtid] &&
9201 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
9202 			e_err(drv, "Link filter exists for link: %x\n",
9203 			      link_uhtid);
9204 			return err;
9205 		}
9206 
9207 		for (i = 0; nexthdr[i].jump; i++) {
9208 			if (nexthdr[i].o != cls->knode.sel->offoff ||
9209 			    nexthdr[i].s != cls->knode.sel->offshift ||
9210 			    nexthdr[i].m != cls->knode.sel->offmask)
9211 				return err;
9212 
9213 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9214 			if (!jump)
9215 				return -ENOMEM;
9216 			input = kzalloc(sizeof(*input), GFP_KERNEL);
9217 			if (!input) {
9218 				err = -ENOMEM;
9219 				goto free_jump;
9220 			}
9221 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9222 			if (!mask) {
9223 				err = -ENOMEM;
9224 				goto free_input;
9225 			}
9226 			jump->input = input;
9227 			jump->mask = mask;
9228 			jump->link_hdl = cls->knode.handle;
9229 
9230 			err = ixgbe_clsu32_build_input(input, mask, cls,
9231 						       field_ptr, &nexthdr[i]);
9232 			if (!err) {
9233 				jump->mat = nexthdr[i].jump;
9234 				adapter->jump_tables[link_uhtid] = jump;
9235 				break;
9236 			}
9237 		}
9238 		return 0;
9239 	}
9240 
9241 	input = kzalloc(sizeof(*input), GFP_KERNEL);
9242 	if (!input)
9243 		return -ENOMEM;
9244 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9245 	if (!mask) {
9246 		err = -ENOMEM;
9247 		goto free_input;
9248 	}
9249 
9250 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9251 		if ((adapter->jump_tables[uhtid])->input)
9252 			memcpy(input, (adapter->jump_tables[uhtid])->input,
9253 			       sizeof(*input));
9254 		if ((adapter->jump_tables[uhtid])->mask)
9255 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9256 			       sizeof(*mask));
9257 
9258 		/* Lookup in all child hash tables if this location is already
9259 		 * filled with a filter
9260 		 */
9261 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9262 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
9263 
9264 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
9265 				e_err(drv, "Filter exists in location: %x\n",
9266 				      loc);
9267 				err = -EINVAL;
9268 				goto err_out;
9269 			}
9270 		}
9271 	}
9272 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9273 	if (err)
9274 		goto err_out;
9275 
9276 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9277 			       &queue);
9278 	if (err < 0)
9279 		goto err_out;
9280 
9281 	input->sw_idx = loc;
9282 
9283 	spin_lock(&adapter->fdir_perfect_lock);
9284 
9285 	if (hlist_empty(&adapter->fdir_filter_list)) {
9286 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9287 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9288 		if (err)
9289 			goto err_out_w_lock;
9290 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9291 		err = -EINVAL;
9292 		goto err_out_w_lock;
9293 	}
9294 
9295 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9296 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9297 						    input->sw_idx, queue);
9298 	if (!err)
9299 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9300 	spin_unlock(&adapter->fdir_perfect_lock);
9301 
9302 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9303 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9304 
9305 	kfree(mask);
9306 	return err;
9307 err_out_w_lock:
9308 	spin_unlock(&adapter->fdir_perfect_lock);
9309 err_out:
9310 	kfree(mask);
9311 free_input:
9312 	kfree(input);
9313 free_jump:
9314 	kfree(jump);
9315 	return err;
9316 }
9317 
9318 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
9319 				  struct tc_cls_u32_offload *cls_u32)
9320 {
9321 	switch (cls_u32->command) {
9322 	case TC_CLSU32_NEW_KNODE:
9323 	case TC_CLSU32_REPLACE_KNODE:
9324 		return ixgbe_configure_clsu32(adapter, cls_u32);
9325 	case TC_CLSU32_DELETE_KNODE:
9326 		return ixgbe_delete_clsu32(adapter, cls_u32);
9327 	case TC_CLSU32_NEW_HNODE:
9328 	case TC_CLSU32_REPLACE_HNODE:
9329 		return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9330 	case TC_CLSU32_DELETE_HNODE:
9331 		return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9332 	default:
9333 		return -EOPNOTSUPP;
9334 	}
9335 }
9336 
9337 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
9338 				   void *cb_priv)
9339 {
9340 	struct ixgbe_adapter *adapter = cb_priv;
9341 
9342 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
9343 		return -EOPNOTSUPP;
9344 
9345 	switch (type) {
9346 	case TC_SETUP_CLSU32:
9347 		return ixgbe_setup_tc_cls_u32(adapter, type_data);
9348 	default:
9349 		return -EOPNOTSUPP;
9350 	}
9351 }
9352 
9353 static int ixgbe_setup_tc_block(struct net_device *dev,
9354 				struct tc_block_offload *f)
9355 {
9356 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9357 
9358 	if (f->binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
9359 		return -EOPNOTSUPP;
9360 
9361 	switch (f->command) {
9362 	case TC_BLOCK_BIND:
9363 		return tcf_block_cb_register(f->block, ixgbe_setup_tc_block_cb,
9364 					     adapter, adapter);
9365 	case TC_BLOCK_UNBIND:
9366 		tcf_block_cb_unregister(f->block, ixgbe_setup_tc_block_cb,
9367 					adapter);
9368 		return 0;
9369 	default:
9370 		return -EOPNOTSUPP;
9371 	}
9372 }
9373 
9374 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9375 				 struct tc_mqprio_qopt *mqprio)
9376 {
9377 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9378 	return ixgbe_setup_tc(dev, mqprio->num_tc);
9379 }
9380 
9381 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9382 			    void *type_data)
9383 {
9384 	switch (type) {
9385 	case TC_SETUP_BLOCK:
9386 		return ixgbe_setup_tc_block(dev, type_data);
9387 	case TC_SETUP_QDISC_MQPRIO:
9388 		return ixgbe_setup_tc_mqprio(dev, type_data);
9389 	default:
9390 		return -EOPNOTSUPP;
9391 	}
9392 }
9393 
9394 #ifdef CONFIG_PCI_IOV
9395 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9396 {
9397 	struct net_device *netdev = adapter->netdev;
9398 
9399 	rtnl_lock();
9400 	ixgbe_setup_tc(netdev, adapter->hw_tcs);
9401 	rtnl_unlock();
9402 }
9403 
9404 #endif
9405 void ixgbe_do_reset(struct net_device *netdev)
9406 {
9407 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9408 
9409 	if (netif_running(netdev))
9410 		ixgbe_reinit_locked(adapter);
9411 	else
9412 		ixgbe_reset(adapter);
9413 }
9414 
9415 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9416 					    netdev_features_t features)
9417 {
9418 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9419 
9420 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9421 	if (!(features & NETIF_F_RXCSUM))
9422 		features &= ~NETIF_F_LRO;
9423 
9424 	/* Turn off LRO if not RSC capable */
9425 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9426 		features &= ~NETIF_F_LRO;
9427 
9428 	return features;
9429 }
9430 
9431 static int ixgbe_set_features(struct net_device *netdev,
9432 			      netdev_features_t features)
9433 {
9434 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9435 	netdev_features_t changed = netdev->features ^ features;
9436 	bool need_reset = false;
9437 
9438 	/* Make sure RSC matches LRO, reset if change */
9439 	if (!(features & NETIF_F_LRO)) {
9440 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9441 			need_reset = true;
9442 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9443 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9444 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9445 		if (adapter->rx_itr_setting == 1 ||
9446 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9447 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9448 			need_reset = true;
9449 		} else if ((changed ^ features) & NETIF_F_LRO) {
9450 			e_info(probe, "rx-usecs set too low, "
9451 			       "disabling RSC\n");
9452 		}
9453 	}
9454 
9455 	/*
9456 	 * Check if Flow Director n-tuple support or hw_tc support was
9457 	 * enabled or disabled.  If the state changed, we need to reset.
9458 	 */
9459 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9460 		/* turn off ATR, enable perfect filters and reset */
9461 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9462 			need_reset = true;
9463 
9464 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9465 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9466 	} else {
9467 		/* turn off perfect filters, enable ATR and reset */
9468 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9469 			need_reset = true;
9470 
9471 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9472 
9473 		/* We cannot enable ATR if SR-IOV is enabled */
9474 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9475 		    /* We cannot enable ATR if we have 2 or more tcs */
9476 		    (adapter->hw_tcs > 1) ||
9477 		    /* We cannot enable ATR if RSS is disabled */
9478 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9479 		    /* A sample rate of 0 indicates ATR disabled */
9480 		    (!adapter->atr_sample_rate))
9481 			; /* do nothing not supported */
9482 		else /* otherwise supported and set the flag */
9483 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9484 	}
9485 
9486 	if (changed & NETIF_F_RXALL)
9487 		need_reset = true;
9488 
9489 	netdev->features = features;
9490 
9491 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
9492 		if (features & NETIF_F_RXCSUM) {
9493 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9494 		} else {
9495 			u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9496 
9497 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9498 		}
9499 	}
9500 
9501 	if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
9502 		if (features & NETIF_F_RXCSUM) {
9503 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9504 		} else {
9505 			u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9506 
9507 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9508 		}
9509 	}
9510 
9511 	if (need_reset)
9512 		ixgbe_do_reset(netdev);
9513 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9514 			    NETIF_F_HW_VLAN_CTAG_FILTER))
9515 		ixgbe_set_rx_mode(netdev);
9516 
9517 	return 0;
9518 }
9519 
9520 /**
9521  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
9522  * @dev: The port's netdev
9523  * @ti: Tunnel endpoint information
9524  **/
9525 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
9526 				      struct udp_tunnel_info *ti)
9527 {
9528 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9529 	struct ixgbe_hw *hw = &adapter->hw;
9530 	__be16 port = ti->port;
9531 	u32 port_shift = 0;
9532 	u32 reg;
9533 
9534 	if (ti->sa_family != AF_INET)
9535 		return;
9536 
9537 	switch (ti->type) {
9538 	case UDP_TUNNEL_TYPE_VXLAN:
9539 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9540 			return;
9541 
9542 		if (adapter->vxlan_port == port)
9543 			return;
9544 
9545 		if (adapter->vxlan_port) {
9546 			netdev_info(dev,
9547 				    "VXLAN port %d set, not adding port %d\n",
9548 				    ntohs(adapter->vxlan_port),
9549 				    ntohs(port));
9550 			return;
9551 		}
9552 
9553 		adapter->vxlan_port = port;
9554 		break;
9555 	case UDP_TUNNEL_TYPE_GENEVE:
9556 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9557 			return;
9558 
9559 		if (adapter->geneve_port == port)
9560 			return;
9561 
9562 		if (adapter->geneve_port) {
9563 			netdev_info(dev,
9564 				    "GENEVE port %d set, not adding port %d\n",
9565 				    ntohs(adapter->geneve_port),
9566 				    ntohs(port));
9567 			return;
9568 		}
9569 
9570 		port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
9571 		adapter->geneve_port = port;
9572 		break;
9573 	default:
9574 		return;
9575 	}
9576 
9577 	reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
9578 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
9579 }
9580 
9581 /**
9582  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
9583  * @dev: The port's netdev
9584  * @ti: Tunnel endpoint information
9585  **/
9586 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
9587 				      struct udp_tunnel_info *ti)
9588 {
9589 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9590 	u32 port_mask;
9591 
9592 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
9593 	    ti->type != UDP_TUNNEL_TYPE_GENEVE)
9594 		return;
9595 
9596 	if (ti->sa_family != AF_INET)
9597 		return;
9598 
9599 	switch (ti->type) {
9600 	case UDP_TUNNEL_TYPE_VXLAN:
9601 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9602 			return;
9603 
9604 		if (adapter->vxlan_port != ti->port) {
9605 			netdev_info(dev, "VXLAN port %d not found\n",
9606 				    ntohs(ti->port));
9607 			return;
9608 		}
9609 
9610 		port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9611 		break;
9612 	case UDP_TUNNEL_TYPE_GENEVE:
9613 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9614 			return;
9615 
9616 		if (adapter->geneve_port != ti->port) {
9617 			netdev_info(dev, "GENEVE port %d not found\n",
9618 				    ntohs(ti->port));
9619 			return;
9620 		}
9621 
9622 		port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9623 		break;
9624 	default:
9625 		return;
9626 	}
9627 
9628 	ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9629 	adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9630 }
9631 
9632 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9633 			     struct net_device *dev,
9634 			     const unsigned char *addr, u16 vid,
9635 			     u16 flags)
9636 {
9637 	/* guarantee we can provide a unique filter for the unicast address */
9638 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9639 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9640 		u16 pool = VMDQ_P(0);
9641 
9642 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9643 			return -ENOMEM;
9644 	}
9645 
9646 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9647 }
9648 
9649 /**
9650  * ixgbe_configure_bridge_mode - set various bridge modes
9651  * @adapter: the private structure
9652  * @mode: requested bridge mode
9653  *
9654  * Configure some settings require for various bridge modes.
9655  **/
9656 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9657 				       __u16 mode)
9658 {
9659 	struct ixgbe_hw *hw = &adapter->hw;
9660 	unsigned int p, num_pools;
9661 	u32 vmdctl;
9662 
9663 	switch (mode) {
9664 	case BRIDGE_MODE_VEPA:
9665 		/* disable Tx loopback, rely on switch hairpin mode */
9666 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9667 
9668 		/* must enable Rx switching replication to allow multicast
9669 		 * packet reception on all VFs, and to enable source address
9670 		 * pruning.
9671 		 */
9672 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9673 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9674 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9675 
9676 		/* enable Rx source address pruning. Note, this requires
9677 		 * replication to be enabled or else it does nothing.
9678 		 */
9679 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9680 		for (p = 0; p < num_pools; p++) {
9681 			if (hw->mac.ops.set_source_address_pruning)
9682 				hw->mac.ops.set_source_address_pruning(hw,
9683 								       true,
9684 								       p);
9685 		}
9686 		break;
9687 	case BRIDGE_MODE_VEB:
9688 		/* enable Tx loopback for internal VF/PF communication */
9689 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9690 				IXGBE_PFDTXGSWC_VT_LBEN);
9691 
9692 		/* disable Rx switching replication unless we have SR-IOV
9693 		 * virtual functions
9694 		 */
9695 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9696 		if (!adapter->num_vfs)
9697 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9698 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9699 
9700 		/* disable Rx source address pruning, since we don't expect to
9701 		 * be receiving external loopback of our transmitted frames.
9702 		 */
9703 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9704 		for (p = 0; p < num_pools; p++) {
9705 			if (hw->mac.ops.set_source_address_pruning)
9706 				hw->mac.ops.set_source_address_pruning(hw,
9707 								       false,
9708 								       p);
9709 		}
9710 		break;
9711 	default:
9712 		return -EINVAL;
9713 	}
9714 
9715 	adapter->bridge_mode = mode;
9716 
9717 	e_info(drv, "enabling bridge mode: %s\n",
9718 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9719 
9720 	return 0;
9721 }
9722 
9723 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9724 				    struct nlmsghdr *nlh, u16 flags)
9725 {
9726 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9727 	struct nlattr *attr, *br_spec;
9728 	int rem;
9729 
9730 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9731 		return -EOPNOTSUPP;
9732 
9733 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9734 	if (!br_spec)
9735 		return -EINVAL;
9736 
9737 	nla_for_each_nested(attr, br_spec, rem) {
9738 		int status;
9739 		__u16 mode;
9740 
9741 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9742 			continue;
9743 
9744 		if (nla_len(attr) < sizeof(mode))
9745 			return -EINVAL;
9746 
9747 		mode = nla_get_u16(attr);
9748 		status = ixgbe_configure_bridge_mode(adapter, mode);
9749 		if (status)
9750 			return status;
9751 
9752 		break;
9753 	}
9754 
9755 	return 0;
9756 }
9757 
9758 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9759 				    struct net_device *dev,
9760 				    u32 filter_mask, int nlflags)
9761 {
9762 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9763 
9764 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9765 		return 0;
9766 
9767 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9768 				       adapter->bridge_mode, 0, 0, nlflags,
9769 				       filter_mask, NULL);
9770 }
9771 
9772 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9773 {
9774 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
9775 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9776 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
9777 	int tcs = adapter->hw_tcs ? : 1;
9778 	unsigned int limit;
9779 	int pool, err;
9780 
9781 	/* Hardware has a limited number of available pools. Each VF, and the
9782 	 * PF require a pool. Check to ensure we don't attempt to use more
9783 	 * then the available number of pools.
9784 	 */
9785 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9786 		return ERR_PTR(-EINVAL);
9787 
9788 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9789 	      adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
9790 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
9791 		return ERR_PTR(-EBUSY);
9792 
9793 	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
9794 	if (!fwd_adapter)
9795 		return ERR_PTR(-ENOMEM);
9796 
9797 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9798 	set_bit(pool, adapter->fwd_bitmask);
9799 	limit = find_last_bit(adapter->fwd_bitmask, adapter->num_rx_pools + 1);
9800 
9801 	/* Enable VMDq flag so device will be set in VM mode */
9802 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
9803 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9804 
9805 	fwd_adapter->pool = pool;
9806 	fwd_adapter->real_adapter = adapter;
9807 
9808 	/* Force reinit of ring allocation with VMDQ enabled */
9809 	err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
9810 
9811 	if (!err && netif_running(pdev))
9812 		err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
9813 
9814 	if (!err)
9815 		return fwd_adapter;
9816 
9817 	/* unwind counter and free adapter struct */
9818 	netdev_info(pdev,
9819 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
9820 	clear_bit(pool, adapter->fwd_bitmask);
9821 	kfree(fwd_adapter);
9822 	return ERR_PTR(err);
9823 }
9824 
9825 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9826 {
9827 	struct ixgbe_fwd_adapter *accel = priv;
9828 	struct ixgbe_adapter *adapter = accel->real_adapter;
9829 	unsigned int rxbase = accel->rx_base_queue;
9830 	unsigned int limit, i;
9831 
9832 	/* delete unicast filter associated with offloaded interface */
9833 	ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
9834 			     VMDQ_P(accel->pool));
9835 
9836 	/* disable ability to receive packets for this pool */
9837 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_VMOLR(accel->pool), 0);
9838 
9839 	/* Allow remaining Rx packets to get flushed out of the
9840 	 * Rx FIFO before we drop the netdev for the ring.
9841 	 */
9842 	usleep_range(10000, 20000);
9843 
9844 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
9845 		struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
9846 		struct ixgbe_q_vector *qv = ring->q_vector;
9847 
9848 		/* Make sure we aren't processing any packets and clear
9849 		 * netdev to shut down the ring.
9850 		 */
9851 		if (netif_running(adapter->netdev))
9852 			napi_synchronize(&qv->napi);
9853 		ring->netdev = NULL;
9854 	}
9855 
9856 	clear_bit(accel->pool, adapter->fwd_bitmask);
9857 	limit = find_last_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9858 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9859 
9860 	/* go back to full RSS if we're done with our VMQs */
9861 	if (adapter->ring_feature[RING_F_VMDQ].limit == 1) {
9862 		int rss = min_t(int, ixgbe_max_rss_indices(adapter),
9863 				num_online_cpus());
9864 
9865 		adapter->flags &= ~IXGBE_FLAG_VMDQ_ENABLED;
9866 		adapter->flags &= ~IXGBE_FLAG_SRIOV_ENABLED;
9867 		adapter->ring_feature[RING_F_RSS].limit = rss;
9868 	}
9869 
9870 	ixgbe_setup_tc(pdev, adapter->hw_tcs);
9871 	netdev_dbg(pdev, "pool %i:%i queues %i:%i\n",
9872 		   accel->pool, adapter->num_rx_pools,
9873 		   accel->rx_base_queue,
9874 		   accel->rx_base_queue +
9875 		   adapter->num_rx_queues_per_pool);
9876 	kfree(accel);
9877 }
9878 
9879 #define IXGBE_MAX_MAC_HDR_LEN		127
9880 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9881 
9882 static netdev_features_t
9883 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9884 		     netdev_features_t features)
9885 {
9886 	unsigned int network_hdr_len, mac_hdr_len;
9887 
9888 	/* Make certain the headers can be described by a context descriptor */
9889 	mac_hdr_len = skb_network_header(skb) - skb->data;
9890 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9891 		return features & ~(NETIF_F_HW_CSUM |
9892 				    NETIF_F_SCTP_CRC |
9893 				    NETIF_F_HW_VLAN_CTAG_TX |
9894 				    NETIF_F_TSO |
9895 				    NETIF_F_TSO6);
9896 
9897 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9898 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9899 		return features & ~(NETIF_F_HW_CSUM |
9900 				    NETIF_F_SCTP_CRC |
9901 				    NETIF_F_TSO |
9902 				    NETIF_F_TSO6);
9903 
9904 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9905 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9906 	 */
9907 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
9908 		features &= ~NETIF_F_TSO;
9909 
9910 #ifdef CONFIG_XFRM_OFFLOAD
9911 	/* IPsec offload doesn't get along well with others *yet* */
9912 	if (skb->sp)
9913 		features &= ~(NETIF_F_TSO | NETIF_F_HW_CSUM);
9914 #endif
9915 
9916 	return features;
9917 }
9918 
9919 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
9920 {
9921 	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
9922 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9923 	struct bpf_prog *old_prog;
9924 
9925 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
9926 		return -EINVAL;
9927 
9928 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
9929 		return -EINVAL;
9930 
9931 	/* verify ixgbe ring attributes are sufficient for XDP */
9932 	for (i = 0; i < adapter->num_rx_queues; i++) {
9933 		struct ixgbe_ring *ring = adapter->rx_ring[i];
9934 
9935 		if (ring_is_rsc_enabled(ring))
9936 			return -EINVAL;
9937 
9938 		if (frame_size > ixgbe_rx_bufsz(ring))
9939 			return -EINVAL;
9940 	}
9941 
9942 	if (nr_cpu_ids > MAX_XDP_QUEUES)
9943 		return -ENOMEM;
9944 
9945 	old_prog = xchg(&adapter->xdp_prog, prog);
9946 
9947 	/* If transitioning XDP modes reconfigure rings */
9948 	if (!!prog != !!old_prog) {
9949 		int err = ixgbe_setup_tc(dev, adapter->hw_tcs);
9950 
9951 		if (err) {
9952 			rcu_assign_pointer(adapter->xdp_prog, old_prog);
9953 			return -EINVAL;
9954 		}
9955 	} else {
9956 		for (i = 0; i < adapter->num_rx_queues; i++)
9957 			xchg(&adapter->rx_ring[i]->xdp_prog, adapter->xdp_prog);
9958 	}
9959 
9960 	if (old_prog)
9961 		bpf_prog_put(old_prog);
9962 
9963 	return 0;
9964 }
9965 
9966 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
9967 {
9968 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9969 
9970 	switch (xdp->command) {
9971 	case XDP_SETUP_PROG:
9972 		return ixgbe_xdp_setup(dev, xdp->prog);
9973 	case XDP_QUERY_PROG:
9974 		xdp->prog_attached = !!(adapter->xdp_prog);
9975 		xdp->prog_id = adapter->xdp_prog ?
9976 			adapter->xdp_prog->aux->id : 0;
9977 		return 0;
9978 	default:
9979 		return -EINVAL;
9980 	}
9981 }
9982 
9983 static int ixgbe_xdp_xmit(struct net_device *dev, struct xdp_buff *xdp)
9984 {
9985 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9986 	struct ixgbe_ring *ring;
9987 	int err;
9988 
9989 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
9990 		return -ENETDOWN;
9991 
9992 	/* During program transitions its possible adapter->xdp_prog is assigned
9993 	 * but ring has not been configured yet. In this case simply abort xmit.
9994 	 */
9995 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
9996 	if (unlikely(!ring))
9997 		return -ENXIO;
9998 
9999 	err = ixgbe_xmit_xdp_ring(adapter, xdp);
10000 	if (err != IXGBE_XDP_TX)
10001 		return -ENOSPC;
10002 
10003 	return 0;
10004 }
10005 
10006 static void ixgbe_xdp_flush(struct net_device *dev)
10007 {
10008 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10009 	struct ixgbe_ring *ring;
10010 
10011 	/* Its possible the device went down between xdp xmit and flush so
10012 	 * we need to ensure device is still up.
10013 	 */
10014 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10015 		return;
10016 
10017 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
10018 	if (unlikely(!ring))
10019 		return;
10020 
10021 	/* Force memory writes to complete before letting h/w know there
10022 	 * are new descriptors to fetch.
10023 	 */
10024 	wmb();
10025 	writel(ring->next_to_use, ring->tail);
10026 
10027 	return;
10028 }
10029 
10030 static const struct net_device_ops ixgbe_netdev_ops = {
10031 	.ndo_open		= ixgbe_open,
10032 	.ndo_stop		= ixgbe_close,
10033 	.ndo_start_xmit		= ixgbe_xmit_frame,
10034 	.ndo_select_queue	= ixgbe_select_queue,
10035 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
10036 	.ndo_validate_addr	= eth_validate_addr,
10037 	.ndo_set_mac_address	= ixgbe_set_mac,
10038 	.ndo_change_mtu		= ixgbe_change_mtu,
10039 	.ndo_tx_timeout		= ixgbe_tx_timeout,
10040 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
10041 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
10042 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
10043 	.ndo_do_ioctl		= ixgbe_ioctl,
10044 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
10045 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
10046 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
10047 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
10048 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
10049 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
10050 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
10051 	.ndo_get_stats64	= ixgbe_get_stats64,
10052 	.ndo_setup_tc		= __ixgbe_setup_tc,
10053 #ifdef CONFIG_NET_POLL_CONTROLLER
10054 	.ndo_poll_controller	= ixgbe_netpoll,
10055 #endif
10056 #ifdef IXGBE_FCOE
10057 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
10058 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
10059 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
10060 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
10061 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
10062 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
10063 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
10064 #endif /* IXGBE_FCOE */
10065 	.ndo_set_features = ixgbe_set_features,
10066 	.ndo_fix_features = ixgbe_fix_features,
10067 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
10068 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
10069 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
10070 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
10071 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
10072 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
10073 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
10074 	.ndo_features_check	= ixgbe_features_check,
10075 	.ndo_bpf		= ixgbe_xdp,
10076 	.ndo_xdp_xmit		= ixgbe_xdp_xmit,
10077 	.ndo_xdp_flush		= ixgbe_xdp_flush,
10078 };
10079 
10080 /**
10081  * ixgbe_enumerate_functions - Get the number of ports this device has
10082  * @adapter: adapter structure
10083  *
10084  * This function enumerates the phsyical functions co-located on a single slot,
10085  * in order to determine how many ports a device has. This is most useful in
10086  * determining the required GT/s of PCIe bandwidth necessary for optimal
10087  * performance.
10088  **/
10089 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
10090 {
10091 	struct pci_dev *entry, *pdev = adapter->pdev;
10092 	int physfns = 0;
10093 
10094 	/* Some cards can not use the generic count PCIe functions method,
10095 	 * because they are behind a parent switch, so we hardcode these with
10096 	 * the correct number of functions.
10097 	 */
10098 	if (ixgbe_pcie_from_parent(&adapter->hw))
10099 		physfns = 4;
10100 
10101 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
10102 		/* don't count virtual functions */
10103 		if (entry->is_virtfn)
10104 			continue;
10105 
10106 		/* When the devices on the bus don't all match our device ID,
10107 		 * we can't reliably determine the correct number of
10108 		 * functions. This can occur if a function has been direct
10109 		 * attached to a virtual machine using VT-d, for example. In
10110 		 * this case, simply return -1 to indicate this.
10111 		 */
10112 		if ((entry->vendor != pdev->vendor) ||
10113 		    (entry->device != pdev->device))
10114 			return -1;
10115 
10116 		physfns++;
10117 	}
10118 
10119 	return physfns;
10120 }
10121 
10122 /**
10123  * ixgbe_wol_supported - Check whether device supports WoL
10124  * @adapter: the adapter private structure
10125  * @device_id: the device ID
10126  * @subdevice_id: the subsystem device ID
10127  *
10128  * This function is used by probe and ethtool to determine
10129  * which devices have WoL support
10130  *
10131  **/
10132 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
10133 			 u16 subdevice_id)
10134 {
10135 	struct ixgbe_hw *hw = &adapter->hw;
10136 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
10137 
10138 	/* WOL not supported on 82598 */
10139 	if (hw->mac.type == ixgbe_mac_82598EB)
10140 		return false;
10141 
10142 	/* check eeprom to see if WOL is enabled for X540 and newer */
10143 	if (hw->mac.type >= ixgbe_mac_X540) {
10144 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10145 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10146 		     (hw->bus.func == 0)))
10147 			return true;
10148 	}
10149 
10150 	/* WOL is determined based on device IDs for 82599 MACs */
10151 	switch (device_id) {
10152 	case IXGBE_DEV_ID_82599_SFP:
10153 		/* Only these subdevices could supports WOL */
10154 		switch (subdevice_id) {
10155 		case IXGBE_SUBDEV_ID_82599_560FLR:
10156 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10157 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10158 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10159 			/* only support first port */
10160 			if (hw->bus.func != 0)
10161 				break;
10162 			/* fall through */
10163 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10164 		case IXGBE_SUBDEV_ID_82599_SFP:
10165 		case IXGBE_SUBDEV_ID_82599_RNDC:
10166 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10167 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10168 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10169 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10170 			return true;
10171 		}
10172 		break;
10173 	case IXGBE_DEV_ID_82599EN_SFP:
10174 		/* Only these subdevices support WOL */
10175 		switch (subdevice_id) {
10176 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10177 			return true;
10178 		}
10179 		break;
10180 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10181 		/* All except this subdevice support WOL */
10182 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10183 			return true;
10184 		break;
10185 	case IXGBE_DEV_ID_82599_KX4:
10186 		return  true;
10187 	default:
10188 		break;
10189 	}
10190 
10191 	return false;
10192 }
10193 
10194 /**
10195  * ixgbe_set_fw_version - Set FW version
10196  * @adapter: the adapter private structure
10197  *
10198  * This function is used by probe and ethtool to determine the FW version to
10199  * format to display. The FW version is taken from the EEPROM/NVM.
10200  */
10201 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
10202 {
10203 	struct ixgbe_hw *hw = &adapter->hw;
10204 	struct ixgbe_nvm_version nvm_ver;
10205 
10206 	ixgbe_get_oem_prod_version(hw, &nvm_ver);
10207 	if (nvm_ver.oem_valid) {
10208 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10209 			 "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
10210 			 nvm_ver.oem_release);
10211 		return;
10212 	}
10213 
10214 	ixgbe_get_etk_id(hw, &nvm_ver);
10215 	ixgbe_get_orom_version(hw, &nvm_ver);
10216 
10217 	if (nvm_ver.or_valid) {
10218 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10219 			 "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
10220 			 nvm_ver.or_build, nvm_ver.or_patch);
10221 		return;
10222 	}
10223 
10224 	/* Set ETrack ID format */
10225 	snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10226 		 "0x%08x", nvm_ver.etk_id);
10227 }
10228 
10229 /**
10230  * ixgbe_probe - Device Initialization Routine
10231  * @pdev: PCI device information struct
10232  * @ent: entry in ixgbe_pci_tbl
10233  *
10234  * Returns 0 on success, negative on failure
10235  *
10236  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10237  * The OS initialization, configuring of the adapter private structure,
10238  * and a hardware reset occur.
10239  **/
10240 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10241 {
10242 	struct net_device *netdev;
10243 	struct ixgbe_adapter *adapter = NULL;
10244 	struct ixgbe_hw *hw;
10245 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10246 	int i, err, pci_using_dac, expected_gts;
10247 	unsigned int indices = MAX_TX_QUEUES;
10248 	u8 part_str[IXGBE_PBANUM_LENGTH];
10249 	bool disable_dev = false;
10250 #ifdef IXGBE_FCOE
10251 	u16 device_caps;
10252 #endif
10253 	u32 eec;
10254 
10255 	/* Catch broken hardware that put the wrong VF device ID in
10256 	 * the PCIe SR-IOV capability.
10257 	 */
10258 	if (pdev->is_virtfn) {
10259 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10260 		     pci_name(pdev), pdev->vendor, pdev->device);
10261 		return -EINVAL;
10262 	}
10263 
10264 	err = pci_enable_device_mem(pdev);
10265 	if (err)
10266 		return err;
10267 
10268 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
10269 		pci_using_dac = 1;
10270 	} else {
10271 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10272 		if (err) {
10273 			dev_err(&pdev->dev,
10274 				"No usable DMA configuration, aborting\n");
10275 			goto err_dma;
10276 		}
10277 		pci_using_dac = 0;
10278 	}
10279 
10280 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10281 	if (err) {
10282 		dev_err(&pdev->dev,
10283 			"pci_request_selected_regions failed 0x%x\n", err);
10284 		goto err_pci_reg;
10285 	}
10286 
10287 	pci_enable_pcie_error_reporting(pdev);
10288 
10289 	pci_set_master(pdev);
10290 	pci_save_state(pdev);
10291 
10292 	if (ii->mac == ixgbe_mac_82598EB) {
10293 #ifdef CONFIG_IXGBE_DCB
10294 		/* 8 TC w/ 4 queues per TC */
10295 		indices = 4 * MAX_TRAFFIC_CLASS;
10296 #else
10297 		indices = IXGBE_MAX_RSS_INDICES;
10298 #endif
10299 	}
10300 
10301 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10302 	if (!netdev) {
10303 		err = -ENOMEM;
10304 		goto err_alloc_etherdev;
10305 	}
10306 
10307 	SET_NETDEV_DEV(netdev, &pdev->dev);
10308 
10309 	adapter = netdev_priv(netdev);
10310 
10311 	adapter->netdev = netdev;
10312 	adapter->pdev = pdev;
10313 	hw = &adapter->hw;
10314 	hw->back = adapter;
10315 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10316 
10317 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10318 			      pci_resource_len(pdev, 0));
10319 	adapter->io_addr = hw->hw_addr;
10320 	if (!hw->hw_addr) {
10321 		err = -EIO;
10322 		goto err_ioremap;
10323 	}
10324 
10325 	netdev->netdev_ops = &ixgbe_netdev_ops;
10326 	ixgbe_set_ethtool_ops(netdev);
10327 	netdev->watchdog_timeo = 5 * HZ;
10328 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10329 
10330 	/* Setup hw api */
10331 	hw->mac.ops   = *ii->mac_ops;
10332 	hw->mac.type  = ii->mac;
10333 	hw->mvals     = ii->mvals;
10334 	if (ii->link_ops)
10335 		hw->link.ops  = *ii->link_ops;
10336 
10337 	/* EEPROM */
10338 	hw->eeprom.ops = *ii->eeprom_ops;
10339 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10340 	if (ixgbe_removed(hw->hw_addr)) {
10341 		err = -EIO;
10342 		goto err_ioremap;
10343 	}
10344 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10345 	if (!(eec & BIT(8)))
10346 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10347 
10348 	/* PHY */
10349 	hw->phy.ops = *ii->phy_ops;
10350 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10351 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
10352 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10353 	hw->phy.mdio.mmds = 0;
10354 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10355 	hw->phy.mdio.dev = netdev;
10356 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10357 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10358 
10359 	/* setup the private structure */
10360 	err = ixgbe_sw_init(adapter, ii);
10361 	if (err)
10362 		goto err_sw_init;
10363 
10364 	/* Make sure the SWFW semaphore is in a valid state */
10365 	if (hw->mac.ops.init_swfw_sync)
10366 		hw->mac.ops.init_swfw_sync(hw);
10367 
10368 	/* Make it possible the adapter to be woken up via WOL */
10369 	switch (adapter->hw.mac.type) {
10370 	case ixgbe_mac_82599EB:
10371 	case ixgbe_mac_X540:
10372 	case ixgbe_mac_X550:
10373 	case ixgbe_mac_X550EM_x:
10374 	case ixgbe_mac_x550em_a:
10375 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10376 		break;
10377 	default:
10378 		break;
10379 	}
10380 
10381 	/*
10382 	 * If there is a fan on this device and it has failed log the
10383 	 * failure.
10384 	 */
10385 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10386 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10387 		if (esdp & IXGBE_ESDP_SDP1)
10388 			e_crit(probe, "Fan has stopped, replace the adapter\n");
10389 	}
10390 
10391 	if (allow_unsupported_sfp)
10392 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
10393 
10394 	/* reset_hw fills in the perm_addr as well */
10395 	hw->phy.reset_if_overtemp = true;
10396 	err = hw->mac.ops.reset_hw(hw);
10397 	hw->phy.reset_if_overtemp = false;
10398 	ixgbe_set_eee_capable(adapter);
10399 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
10400 		err = 0;
10401 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
10402 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10403 		e_dev_err("Reload the driver after installing a supported module.\n");
10404 		goto err_sw_init;
10405 	} else if (err) {
10406 		e_dev_err("HW Init failed: %d\n", err);
10407 		goto err_sw_init;
10408 	}
10409 
10410 #ifdef CONFIG_PCI_IOV
10411 	/* SR-IOV not supported on the 82598 */
10412 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
10413 		goto skip_sriov;
10414 	/* Mailbox */
10415 	ixgbe_init_mbx_params_pf(hw);
10416 	hw->mbx.ops = ii->mbx_ops;
10417 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
10418 	ixgbe_enable_sriov(adapter, max_vfs);
10419 skip_sriov:
10420 
10421 #endif
10422 	netdev->features = NETIF_F_SG |
10423 			   NETIF_F_TSO |
10424 			   NETIF_F_TSO6 |
10425 			   NETIF_F_RXHASH |
10426 			   NETIF_F_RXCSUM |
10427 			   NETIF_F_HW_CSUM;
10428 
10429 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
10430 				    NETIF_F_GSO_GRE_CSUM | \
10431 				    NETIF_F_GSO_IPXIP4 | \
10432 				    NETIF_F_GSO_IPXIP6 | \
10433 				    NETIF_F_GSO_UDP_TUNNEL | \
10434 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
10435 
10436 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
10437 	netdev->features |= NETIF_F_GSO_PARTIAL |
10438 			    IXGBE_GSO_PARTIAL_FEATURES;
10439 
10440 	if (hw->mac.type >= ixgbe_mac_82599EB)
10441 		netdev->features |= NETIF_F_SCTP_CRC;
10442 
10443 	/* copy netdev features into list of user selectable features */
10444 	netdev->hw_features |= netdev->features |
10445 			       NETIF_F_HW_VLAN_CTAG_FILTER |
10446 			       NETIF_F_HW_VLAN_CTAG_RX |
10447 			       NETIF_F_HW_VLAN_CTAG_TX |
10448 			       NETIF_F_RXALL |
10449 			       NETIF_F_HW_L2FW_DOFFLOAD;
10450 
10451 	if (hw->mac.type >= ixgbe_mac_82599EB)
10452 		netdev->hw_features |= NETIF_F_NTUPLE |
10453 				       NETIF_F_HW_TC;
10454 
10455 	if (pci_using_dac)
10456 		netdev->features |= NETIF_F_HIGHDMA;
10457 
10458 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
10459 	netdev->hw_enc_features |= netdev->vlan_features;
10460 	netdev->mpls_features |= NETIF_F_SG |
10461 				 NETIF_F_TSO |
10462 				 NETIF_F_TSO6 |
10463 				 NETIF_F_HW_CSUM;
10464 	netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
10465 
10466 	/* set this bit last since it cannot be part of vlan_features */
10467 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
10468 			    NETIF_F_HW_VLAN_CTAG_RX |
10469 			    NETIF_F_HW_VLAN_CTAG_TX;
10470 
10471 	netdev->priv_flags |= IFF_UNICAST_FLT;
10472 	netdev->priv_flags |= IFF_SUPP_NOFCS;
10473 
10474 	/* MTU range: 68 - 9710 */
10475 	netdev->min_mtu = ETH_MIN_MTU;
10476 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
10477 
10478 #ifdef CONFIG_IXGBE_DCB
10479 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
10480 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
10481 #endif
10482 
10483 #ifdef IXGBE_FCOE
10484 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
10485 		unsigned int fcoe_l;
10486 
10487 		if (hw->mac.ops.get_device_caps) {
10488 			hw->mac.ops.get_device_caps(hw, &device_caps);
10489 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
10490 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
10491 		}
10492 
10493 
10494 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
10495 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
10496 
10497 		netdev->features |= NETIF_F_FSO |
10498 				    NETIF_F_FCOE_CRC;
10499 
10500 		netdev->vlan_features |= NETIF_F_FSO |
10501 					 NETIF_F_FCOE_CRC |
10502 					 NETIF_F_FCOE_MTU;
10503 	}
10504 #endif /* IXGBE_FCOE */
10505 	ixgbe_init_ipsec_offload(adapter);
10506 
10507 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
10508 		netdev->hw_features |= NETIF_F_LRO;
10509 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
10510 		netdev->features |= NETIF_F_LRO;
10511 
10512 	/* make sure the EEPROM is good */
10513 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
10514 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
10515 		err = -EIO;
10516 		goto err_sw_init;
10517 	}
10518 
10519 	eth_platform_get_mac_address(&adapter->pdev->dev,
10520 				     adapter->hw.mac.perm_addr);
10521 
10522 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
10523 
10524 	if (!is_valid_ether_addr(netdev->dev_addr)) {
10525 		e_dev_err("invalid MAC address\n");
10526 		err = -EIO;
10527 		goto err_sw_init;
10528 	}
10529 
10530 	/* Set hw->mac.addr to permanent MAC address */
10531 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
10532 	ixgbe_mac_set_default_filter(adapter);
10533 
10534 	timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
10535 
10536 	if (ixgbe_removed(hw->hw_addr)) {
10537 		err = -EIO;
10538 		goto err_sw_init;
10539 	}
10540 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
10541 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
10542 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
10543 
10544 	err = ixgbe_init_interrupt_scheme(adapter);
10545 	if (err)
10546 		goto err_sw_init;
10547 
10548 	for (i = 0; i < adapter->num_rx_queues; i++)
10549 		u64_stats_init(&adapter->rx_ring[i]->syncp);
10550 	for (i = 0; i < adapter->num_tx_queues; i++)
10551 		u64_stats_init(&adapter->tx_ring[i]->syncp);
10552 	for (i = 0; i < adapter->num_xdp_queues; i++)
10553 		u64_stats_init(&adapter->xdp_ring[i]->syncp);
10554 
10555 	/* WOL not supported for all devices */
10556 	adapter->wol = 0;
10557 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
10558 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
10559 						pdev->subsystem_device);
10560 	if (hw->wol_enabled)
10561 		adapter->wol = IXGBE_WUFC_MAG;
10562 
10563 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
10564 
10565 	/* save off EEPROM version number */
10566 	ixgbe_set_fw_version(adapter);
10567 
10568 	/* pick up the PCI bus settings for reporting later */
10569 	if (ixgbe_pcie_from_parent(hw))
10570 		ixgbe_get_parent_bus_info(adapter);
10571 	else
10572 		 hw->mac.ops.get_bus_info(hw);
10573 
10574 	/* calculate the expected PCIe bandwidth required for optimal
10575 	 * performance. Note that some older parts will never have enough
10576 	 * bandwidth due to being older generation PCIe parts. We clamp these
10577 	 * parts to ensure no warning is displayed if it can't be fixed.
10578 	 */
10579 	switch (hw->mac.type) {
10580 	case ixgbe_mac_82598EB:
10581 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
10582 		break;
10583 	default:
10584 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
10585 		break;
10586 	}
10587 
10588 	/* don't check link if we failed to enumerate functions */
10589 	if (expected_gts > 0)
10590 		ixgbe_check_minimum_link(adapter, expected_gts);
10591 
10592 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
10593 	if (err)
10594 		strlcpy(part_str, "Unknown", sizeof(part_str));
10595 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
10596 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
10597 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
10598 			   part_str);
10599 	else
10600 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
10601 			   hw->mac.type, hw->phy.type, part_str);
10602 
10603 	e_dev_info("%pM\n", netdev->dev_addr);
10604 
10605 	/* reset the hardware with the new settings */
10606 	err = hw->mac.ops.start_hw(hw);
10607 	if (err == IXGBE_ERR_EEPROM_VERSION) {
10608 		/* We are running on a pre-production device, log a warning */
10609 		e_dev_warn("This device is a pre-production adapter/LOM. "
10610 			   "Please be aware there may be issues associated "
10611 			   "with your hardware.  If you are experiencing "
10612 			   "problems please contact your Intel or hardware "
10613 			   "representative who provided you with this "
10614 			   "hardware.\n");
10615 	}
10616 	strcpy(netdev->name, "eth%d");
10617 	pci_set_drvdata(pdev, adapter);
10618 	err = register_netdev(netdev);
10619 	if (err)
10620 		goto err_register;
10621 
10622 
10623 	/* power down the optics for 82599 SFP+ fiber */
10624 	if (hw->mac.ops.disable_tx_laser)
10625 		hw->mac.ops.disable_tx_laser(hw);
10626 
10627 	/* carrier off reporting is important to ethtool even BEFORE open */
10628 	netif_carrier_off(netdev);
10629 
10630 #ifdef CONFIG_IXGBE_DCA
10631 	if (dca_add_requester(&pdev->dev) == 0) {
10632 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
10633 		ixgbe_setup_dca(adapter);
10634 	}
10635 #endif
10636 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
10637 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
10638 		for (i = 0; i < adapter->num_vfs; i++)
10639 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
10640 	}
10641 
10642 	/* firmware requires driver version to be 0xFFFFFFFF
10643 	 * since os does not support feature
10644 	 */
10645 	if (hw->mac.ops.set_fw_drv_ver)
10646 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
10647 					   sizeof(ixgbe_driver_version) - 1,
10648 					   ixgbe_driver_version);
10649 
10650 	/* add san mac addr to netdev */
10651 	ixgbe_add_sanmac_netdev(netdev);
10652 
10653 	e_dev_info("%s\n", ixgbe_default_device_descr);
10654 
10655 #ifdef CONFIG_IXGBE_HWMON
10656 	if (ixgbe_sysfs_init(adapter))
10657 		e_err(probe, "failed to allocate sysfs resources\n");
10658 #endif /* CONFIG_IXGBE_HWMON */
10659 
10660 	ixgbe_dbg_adapter_init(adapter);
10661 
10662 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
10663 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
10664 		hw->mac.ops.setup_link(hw,
10665 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
10666 			true);
10667 
10668 	return 0;
10669 
10670 err_register:
10671 	ixgbe_release_hw_control(adapter);
10672 	ixgbe_clear_interrupt_scheme(adapter);
10673 err_sw_init:
10674 	ixgbe_disable_sriov(adapter);
10675 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
10676 	iounmap(adapter->io_addr);
10677 	kfree(adapter->jump_tables[0]);
10678 	kfree(adapter->mac_table);
10679 	kfree(adapter->rss_key);
10680 err_ioremap:
10681 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10682 	free_netdev(netdev);
10683 err_alloc_etherdev:
10684 	pci_release_mem_regions(pdev);
10685 err_pci_reg:
10686 err_dma:
10687 	if (!adapter || disable_dev)
10688 		pci_disable_device(pdev);
10689 	return err;
10690 }
10691 
10692 /**
10693  * ixgbe_remove - Device Removal Routine
10694  * @pdev: PCI device information struct
10695  *
10696  * ixgbe_remove is called by the PCI subsystem to alert the driver
10697  * that it should release a PCI device.  The could be caused by a
10698  * Hot-Plug event, or because the driver is going to be removed from
10699  * memory.
10700  **/
10701 static void ixgbe_remove(struct pci_dev *pdev)
10702 {
10703 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10704 	struct net_device *netdev;
10705 	bool disable_dev;
10706 	int i;
10707 
10708 	/* if !adapter then we already cleaned up in probe */
10709 	if (!adapter)
10710 		return;
10711 
10712 	netdev  = adapter->netdev;
10713 	ixgbe_dbg_adapter_exit(adapter);
10714 
10715 	set_bit(__IXGBE_REMOVING, &adapter->state);
10716 	cancel_work_sync(&adapter->service_task);
10717 
10718 
10719 #ifdef CONFIG_IXGBE_DCA
10720 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
10721 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
10722 		dca_remove_requester(&pdev->dev);
10723 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
10724 				IXGBE_DCA_CTRL_DCA_DISABLE);
10725 	}
10726 
10727 #endif
10728 #ifdef CONFIG_IXGBE_HWMON
10729 	ixgbe_sysfs_exit(adapter);
10730 #endif /* CONFIG_IXGBE_HWMON */
10731 
10732 	/* remove the added san mac */
10733 	ixgbe_del_sanmac_netdev(netdev);
10734 
10735 #ifdef CONFIG_PCI_IOV
10736 	ixgbe_disable_sriov(adapter);
10737 #endif
10738 	if (netdev->reg_state == NETREG_REGISTERED)
10739 		unregister_netdev(netdev);
10740 
10741 	ixgbe_stop_ipsec_offload(adapter);
10742 	ixgbe_clear_interrupt_scheme(adapter);
10743 
10744 	ixgbe_release_hw_control(adapter);
10745 
10746 #ifdef CONFIG_DCB
10747 	kfree(adapter->ixgbe_ieee_pfc);
10748 	kfree(adapter->ixgbe_ieee_ets);
10749 
10750 #endif
10751 	iounmap(adapter->io_addr);
10752 	pci_release_mem_regions(pdev);
10753 
10754 	e_dev_info("complete\n");
10755 
10756 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
10757 		if (adapter->jump_tables[i]) {
10758 			kfree(adapter->jump_tables[i]->input);
10759 			kfree(adapter->jump_tables[i]->mask);
10760 		}
10761 		kfree(adapter->jump_tables[i]);
10762 	}
10763 
10764 	kfree(adapter->mac_table);
10765 	kfree(adapter->rss_key);
10766 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10767 	free_netdev(netdev);
10768 
10769 	pci_disable_pcie_error_reporting(pdev);
10770 
10771 	if (disable_dev)
10772 		pci_disable_device(pdev);
10773 }
10774 
10775 /**
10776  * ixgbe_io_error_detected - called when PCI error is detected
10777  * @pdev: Pointer to PCI device
10778  * @state: The current pci connection state
10779  *
10780  * This function is called after a PCI bus error affecting
10781  * this device has been detected.
10782  */
10783 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
10784 						pci_channel_state_t state)
10785 {
10786 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10787 	struct net_device *netdev = adapter->netdev;
10788 
10789 #ifdef CONFIG_PCI_IOV
10790 	struct ixgbe_hw *hw = &adapter->hw;
10791 	struct pci_dev *bdev, *vfdev;
10792 	u32 dw0, dw1, dw2, dw3;
10793 	int vf, pos;
10794 	u16 req_id, pf_func;
10795 
10796 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
10797 	    adapter->num_vfs == 0)
10798 		goto skip_bad_vf_detection;
10799 
10800 	bdev = pdev->bus->self;
10801 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
10802 		bdev = bdev->bus->self;
10803 
10804 	if (!bdev)
10805 		goto skip_bad_vf_detection;
10806 
10807 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
10808 	if (!pos)
10809 		goto skip_bad_vf_detection;
10810 
10811 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
10812 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
10813 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
10814 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
10815 	if (ixgbe_removed(hw->hw_addr))
10816 		goto skip_bad_vf_detection;
10817 
10818 	req_id = dw1 >> 16;
10819 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
10820 	if (!(req_id & 0x0080))
10821 		goto skip_bad_vf_detection;
10822 
10823 	pf_func = req_id & 0x01;
10824 	if ((pf_func & 1) == (pdev->devfn & 1)) {
10825 		unsigned int device_id;
10826 
10827 		vf = (req_id & 0x7F) >> 1;
10828 		e_dev_err("VF %d has caused a PCIe error\n", vf);
10829 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
10830 				"%8.8x\tdw3: %8.8x\n",
10831 		dw0, dw1, dw2, dw3);
10832 		switch (adapter->hw.mac.type) {
10833 		case ixgbe_mac_82599EB:
10834 			device_id = IXGBE_82599_VF_DEVICE_ID;
10835 			break;
10836 		case ixgbe_mac_X540:
10837 			device_id = IXGBE_X540_VF_DEVICE_ID;
10838 			break;
10839 		case ixgbe_mac_X550:
10840 			device_id = IXGBE_DEV_ID_X550_VF;
10841 			break;
10842 		case ixgbe_mac_X550EM_x:
10843 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
10844 			break;
10845 		case ixgbe_mac_x550em_a:
10846 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
10847 			break;
10848 		default:
10849 			device_id = 0;
10850 			break;
10851 		}
10852 
10853 		/* Find the pci device of the offending VF */
10854 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
10855 		while (vfdev) {
10856 			if (vfdev->devfn == (req_id & 0xFF))
10857 				break;
10858 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
10859 					       device_id, vfdev);
10860 		}
10861 		/*
10862 		 * There's a slim chance the VF could have been hot plugged,
10863 		 * so if it is no longer present we don't need to issue the
10864 		 * VFLR.  Just clean up the AER in that case.
10865 		 */
10866 		if (vfdev) {
10867 			pcie_flr(vfdev);
10868 			/* Free device reference count */
10869 			pci_dev_put(vfdev);
10870 		}
10871 
10872 		pci_cleanup_aer_uncorrect_error_status(pdev);
10873 	}
10874 
10875 	/*
10876 	 * Even though the error may have occurred on the other port
10877 	 * we still need to increment the vf error reference count for
10878 	 * both ports because the I/O resume function will be called
10879 	 * for both of them.
10880 	 */
10881 	adapter->vferr_refcount++;
10882 
10883 	return PCI_ERS_RESULT_RECOVERED;
10884 
10885 skip_bad_vf_detection:
10886 #endif /* CONFIG_PCI_IOV */
10887 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
10888 		return PCI_ERS_RESULT_DISCONNECT;
10889 
10890 	if (!netif_device_present(netdev))
10891 		return PCI_ERS_RESULT_DISCONNECT;
10892 
10893 	rtnl_lock();
10894 	netif_device_detach(netdev);
10895 
10896 	if (state == pci_channel_io_perm_failure) {
10897 		rtnl_unlock();
10898 		return PCI_ERS_RESULT_DISCONNECT;
10899 	}
10900 
10901 	if (netif_running(netdev))
10902 		ixgbe_close_suspend(adapter);
10903 
10904 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
10905 		pci_disable_device(pdev);
10906 	rtnl_unlock();
10907 
10908 	/* Request a slot reset. */
10909 	return PCI_ERS_RESULT_NEED_RESET;
10910 }
10911 
10912 /**
10913  * ixgbe_io_slot_reset - called after the pci bus has been reset.
10914  * @pdev: Pointer to PCI device
10915  *
10916  * Restart the card from scratch, as if from a cold-boot.
10917  */
10918 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
10919 {
10920 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10921 	pci_ers_result_t result;
10922 	int err;
10923 
10924 	if (pci_enable_device_mem(pdev)) {
10925 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
10926 		result = PCI_ERS_RESULT_DISCONNECT;
10927 	} else {
10928 		smp_mb__before_atomic();
10929 		clear_bit(__IXGBE_DISABLED, &adapter->state);
10930 		adapter->hw.hw_addr = adapter->io_addr;
10931 		pci_set_master(pdev);
10932 		pci_restore_state(pdev);
10933 		pci_save_state(pdev);
10934 
10935 		pci_wake_from_d3(pdev, false);
10936 
10937 		ixgbe_reset(adapter);
10938 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10939 		result = PCI_ERS_RESULT_RECOVERED;
10940 	}
10941 
10942 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10943 	if (err) {
10944 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
10945 			  "failed 0x%0x\n", err);
10946 		/* non-fatal, continue */
10947 	}
10948 
10949 	return result;
10950 }
10951 
10952 /**
10953  * ixgbe_io_resume - called when traffic can start flowing again.
10954  * @pdev: Pointer to PCI device
10955  *
10956  * This callback is called when the error recovery driver tells us that
10957  * its OK to resume normal operation.
10958  */
10959 static void ixgbe_io_resume(struct pci_dev *pdev)
10960 {
10961 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10962 	struct net_device *netdev = adapter->netdev;
10963 
10964 #ifdef CONFIG_PCI_IOV
10965 	if (adapter->vferr_refcount) {
10966 		e_info(drv, "Resuming after VF err\n");
10967 		adapter->vferr_refcount--;
10968 		return;
10969 	}
10970 
10971 #endif
10972 	rtnl_lock();
10973 	if (netif_running(netdev))
10974 		ixgbe_open(netdev);
10975 
10976 	netif_device_attach(netdev);
10977 	rtnl_unlock();
10978 }
10979 
10980 static const struct pci_error_handlers ixgbe_err_handler = {
10981 	.error_detected = ixgbe_io_error_detected,
10982 	.slot_reset = ixgbe_io_slot_reset,
10983 	.resume = ixgbe_io_resume,
10984 };
10985 
10986 static struct pci_driver ixgbe_driver = {
10987 	.name     = ixgbe_driver_name,
10988 	.id_table = ixgbe_pci_tbl,
10989 	.probe    = ixgbe_probe,
10990 	.remove   = ixgbe_remove,
10991 #ifdef CONFIG_PM
10992 	.suspend  = ixgbe_suspend,
10993 	.resume   = ixgbe_resume,
10994 #endif
10995 	.shutdown = ixgbe_shutdown,
10996 	.sriov_configure = ixgbe_pci_sriov_configure,
10997 	.err_handler = &ixgbe_err_handler
10998 };
10999 
11000 /**
11001  * ixgbe_init_module - Driver Registration Routine
11002  *
11003  * ixgbe_init_module is the first routine called when the driver is
11004  * loaded. All it does is register with the PCI subsystem.
11005  **/
11006 static int __init ixgbe_init_module(void)
11007 {
11008 	int ret;
11009 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
11010 	pr_info("%s\n", ixgbe_copyright);
11011 
11012 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
11013 	if (!ixgbe_wq) {
11014 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
11015 		return -ENOMEM;
11016 	}
11017 
11018 	ixgbe_dbg_init();
11019 
11020 	ret = pci_register_driver(&ixgbe_driver);
11021 	if (ret) {
11022 		destroy_workqueue(ixgbe_wq);
11023 		ixgbe_dbg_exit();
11024 		return ret;
11025 	}
11026 
11027 #ifdef CONFIG_IXGBE_DCA
11028 	dca_register_notify(&dca_notifier);
11029 #endif
11030 
11031 	return 0;
11032 }
11033 
11034 module_init(ixgbe_init_module);
11035 
11036 /**
11037  * ixgbe_exit_module - Driver Exit Cleanup Routine
11038  *
11039  * ixgbe_exit_module is called just before the driver is removed
11040  * from memory.
11041  **/
11042 static void __exit ixgbe_exit_module(void)
11043 {
11044 #ifdef CONFIG_IXGBE_DCA
11045 	dca_unregister_notify(&dca_notifier);
11046 #endif
11047 	pci_unregister_driver(&ixgbe_driver);
11048 
11049 	ixgbe_dbg_exit();
11050 	if (ixgbe_wq) {
11051 		destroy_workqueue(ixgbe_wq);
11052 		ixgbe_wq = NULL;
11053 	}
11054 }
11055 
11056 #ifdef CONFIG_IXGBE_DCA
11057 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
11058 			    void *p)
11059 {
11060 	int ret_val;
11061 
11062 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
11063 					 __ixgbe_notify_dca);
11064 
11065 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
11066 }
11067 
11068 #endif /* CONFIG_IXGBE_DCA */
11069 
11070 module_exit(ixgbe_exit_module);
11071 
11072 /* ixgbe_main.c */
11073