xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision c1aac62f36c1e37ee81c9e09ee9ee733eef05dcb)
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 <scsi/fc/fc_fcoe.h>
53 #include <net/udp_tunnel.h>
54 #include <net/pkt_cls.h>
55 #include <net/tc_act/tc_gact.h>
56 #include <net/tc_act/tc_mirred.h>
57 #include <net/vxlan.h>
58 
59 #include "ixgbe.h"
60 #include "ixgbe_common.h"
61 #include "ixgbe_dcb_82599.h"
62 #include "ixgbe_sriov.h"
63 #include "ixgbe_model.h"
64 
65 char ixgbe_driver_name[] = "ixgbe";
66 static const char ixgbe_driver_string[] =
67 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
68 #ifdef IXGBE_FCOE
69 char ixgbe_default_device_descr[] =
70 			      "Intel(R) 10 Gigabit Network Connection";
71 #else
72 static char ixgbe_default_device_descr[] =
73 			      "Intel(R) 10 Gigabit Network Connection";
74 #endif
75 #define DRV_VERSION "5.0.0-k"
76 const char ixgbe_driver_version[] = DRV_VERSION;
77 static const char ixgbe_copyright[] =
78 				"Copyright (c) 1999-2016 Intel Corporation.";
79 
80 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";
81 
82 static const struct ixgbe_info *ixgbe_info_tbl[] = {
83 	[board_82598]		= &ixgbe_82598_info,
84 	[board_82599]		= &ixgbe_82599_info,
85 	[board_X540]		= &ixgbe_X540_info,
86 	[board_X550]		= &ixgbe_X550_info,
87 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
88 	[board_x550em_a]	= &ixgbe_x550em_a_info,
89 	[board_x550em_a_fw]	= &ixgbe_x550em_a_fw_info,
90 };
91 
92 /* ixgbe_pci_tbl - PCI Device ID Table
93  *
94  * Wildcard entries (PCI_ANY_ID) should come last
95  * Last entry must be all 0s
96  *
97  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
98  *   Class, Class Mask, private data (not used) }
99  */
100 static const struct pci_device_id ixgbe_pci_tbl[] = {
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
133 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
134 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
135 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
136 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
137 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
138 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
139 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
140 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
141 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
142 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
143 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
144 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
145 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
146 	/* required last entry */
147 	{0, }
148 };
149 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
150 
151 #ifdef CONFIG_IXGBE_DCA
152 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
153 			    void *p);
154 static struct notifier_block dca_notifier = {
155 	.notifier_call = ixgbe_notify_dca,
156 	.next          = NULL,
157 	.priority      = 0
158 };
159 #endif
160 
161 #ifdef CONFIG_PCI_IOV
162 static unsigned int max_vfs;
163 module_param(max_vfs, uint, 0);
164 MODULE_PARM_DESC(max_vfs,
165 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
166 #endif /* CONFIG_PCI_IOV */
167 
168 static unsigned int allow_unsupported_sfp;
169 module_param(allow_unsupported_sfp, uint, 0);
170 MODULE_PARM_DESC(allow_unsupported_sfp,
171 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
172 
173 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
174 static int debug = -1;
175 module_param(debug, int, 0);
176 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
177 
178 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
179 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
180 MODULE_LICENSE("GPL");
181 MODULE_VERSION(DRV_VERSION);
182 
183 static struct workqueue_struct *ixgbe_wq;
184 
185 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
186 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
187 
188 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
189 					  u32 reg, u16 *value)
190 {
191 	struct pci_dev *parent_dev;
192 	struct pci_bus *parent_bus;
193 
194 	parent_bus = adapter->pdev->bus->parent;
195 	if (!parent_bus)
196 		return -1;
197 
198 	parent_dev = parent_bus->self;
199 	if (!parent_dev)
200 		return -1;
201 
202 	if (!pci_is_pcie(parent_dev))
203 		return -1;
204 
205 	pcie_capability_read_word(parent_dev, reg, value);
206 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
207 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
208 		return -1;
209 	return 0;
210 }
211 
212 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
213 {
214 	struct ixgbe_hw *hw = &adapter->hw;
215 	u16 link_status = 0;
216 	int err;
217 
218 	hw->bus.type = ixgbe_bus_type_pci_express;
219 
220 	/* Get the negotiated link width and speed from PCI config space of the
221 	 * parent, as this device is behind a switch
222 	 */
223 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
224 
225 	/* assume caller will handle error case */
226 	if (err)
227 		return err;
228 
229 	hw->bus.width = ixgbe_convert_bus_width(link_status);
230 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
231 
232 	return 0;
233 }
234 
235 /**
236  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
237  * @hw: hw specific details
238  *
239  * This function is used by probe to determine whether a device's PCI-Express
240  * bandwidth details should be gathered from the parent bus instead of from the
241  * device. Used to ensure that various locations all have the correct device ID
242  * checks.
243  */
244 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
245 {
246 	switch (hw->device_id) {
247 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
248 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
249 		return true;
250 	default:
251 		return false;
252 	}
253 }
254 
255 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
256 				     int expected_gts)
257 {
258 	struct ixgbe_hw *hw = &adapter->hw;
259 	int max_gts = 0;
260 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
261 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
262 	struct pci_dev *pdev;
263 
264 	/* Some devices are not connected over PCIe and thus do not negotiate
265 	 * speed. These devices do not have valid bus info, and thus any report
266 	 * we generate may not be correct.
267 	 */
268 	if (hw->bus.type == ixgbe_bus_type_internal)
269 		return;
270 
271 	/* determine whether to use the parent device */
272 	if (ixgbe_pcie_from_parent(&adapter->hw))
273 		pdev = adapter->pdev->bus->parent->self;
274 	else
275 		pdev = adapter->pdev;
276 
277 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
278 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
279 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
280 		return;
281 	}
282 
283 	switch (speed) {
284 	case PCIE_SPEED_2_5GT:
285 		/* 8b/10b encoding reduces max throughput by 20% */
286 		max_gts = 2 * width;
287 		break;
288 	case PCIE_SPEED_5_0GT:
289 		/* 8b/10b encoding reduces max throughput by 20% */
290 		max_gts = 4 * width;
291 		break;
292 	case PCIE_SPEED_8_0GT:
293 		/* 128b/130b encoding reduces throughput by less than 2% */
294 		max_gts = 8 * width;
295 		break;
296 	default:
297 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
298 		return;
299 	}
300 
301 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
302 		   max_gts);
303 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
304 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
305 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
306 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
307 		    "Unknown"),
308 		   width,
309 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
310 		    speed == PCIE_SPEED_5_0GT ? "20%" :
311 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
312 		    "Unknown"));
313 
314 	if (max_gts < expected_gts) {
315 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
316 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
317 			expected_gts);
318 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
319 	}
320 }
321 
322 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
323 {
324 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
325 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
326 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
327 		queue_work(ixgbe_wq, &adapter->service_task);
328 }
329 
330 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
331 {
332 	struct ixgbe_adapter *adapter = hw->back;
333 
334 	if (!hw->hw_addr)
335 		return;
336 	hw->hw_addr = NULL;
337 	e_dev_err("Adapter removed\n");
338 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
339 		ixgbe_service_event_schedule(adapter);
340 }
341 
342 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
343 {
344 	u32 value;
345 
346 	/* The following check not only optimizes a bit by not
347 	 * performing a read on the status register when the
348 	 * register just read was a status register read that
349 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
350 	 * potential recursion.
351 	 */
352 	if (reg == IXGBE_STATUS) {
353 		ixgbe_remove_adapter(hw);
354 		return;
355 	}
356 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
357 	if (value == IXGBE_FAILED_READ_REG)
358 		ixgbe_remove_adapter(hw);
359 }
360 
361 /**
362  * ixgbe_read_reg - Read from device register
363  * @hw: hw specific details
364  * @reg: offset of register to read
365  *
366  * Returns : value read or IXGBE_FAILED_READ_REG if removed
367  *
368  * This function is used to read device registers. It checks for device
369  * removal by confirming any read that returns all ones by checking the
370  * status register value for all ones. This function avoids reading from
371  * the hardware if a removal was previously detected in which case it
372  * returns IXGBE_FAILED_READ_REG (all ones).
373  */
374 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
375 {
376 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
377 	u32 value;
378 
379 	if (ixgbe_removed(reg_addr))
380 		return IXGBE_FAILED_READ_REG;
381 	if (unlikely(hw->phy.nw_mng_if_sel &
382 		     IXGBE_NW_MNG_IF_SEL_ENABLE_10_100M)) {
383 		struct ixgbe_adapter *adapter;
384 		int i;
385 
386 		for (i = 0; i < 200; ++i) {
387 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
388 			if (likely(!value))
389 				goto writes_completed;
390 			if (value == IXGBE_FAILED_READ_REG) {
391 				ixgbe_remove_adapter(hw);
392 				return IXGBE_FAILED_READ_REG;
393 			}
394 			udelay(5);
395 		}
396 
397 		adapter = hw->back;
398 		e_warn(hw, "register writes incomplete %08x\n", value);
399 	}
400 
401 writes_completed:
402 	value = readl(reg_addr + reg);
403 	if (unlikely(value == IXGBE_FAILED_READ_REG))
404 		ixgbe_check_remove(hw, reg);
405 	return value;
406 }
407 
408 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
409 {
410 	u16 value;
411 
412 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
413 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
414 		ixgbe_remove_adapter(hw);
415 		return true;
416 	}
417 	return false;
418 }
419 
420 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
421 {
422 	struct ixgbe_adapter *adapter = hw->back;
423 	u16 value;
424 
425 	if (ixgbe_removed(hw->hw_addr))
426 		return IXGBE_FAILED_READ_CFG_WORD;
427 	pci_read_config_word(adapter->pdev, reg, &value);
428 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
429 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
430 		return IXGBE_FAILED_READ_CFG_WORD;
431 	return value;
432 }
433 
434 #ifdef CONFIG_PCI_IOV
435 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
436 {
437 	struct ixgbe_adapter *adapter = hw->back;
438 	u32 value;
439 
440 	if (ixgbe_removed(hw->hw_addr))
441 		return IXGBE_FAILED_READ_CFG_DWORD;
442 	pci_read_config_dword(adapter->pdev, reg, &value);
443 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
444 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
445 		return IXGBE_FAILED_READ_CFG_DWORD;
446 	return value;
447 }
448 #endif /* CONFIG_PCI_IOV */
449 
450 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
451 {
452 	struct ixgbe_adapter *adapter = hw->back;
453 
454 	if (ixgbe_removed(hw->hw_addr))
455 		return;
456 	pci_write_config_word(adapter->pdev, reg, value);
457 }
458 
459 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
460 {
461 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
462 
463 	/* flush memory to make sure state is correct before next watchdog */
464 	smp_mb__before_atomic();
465 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
466 }
467 
468 struct ixgbe_reg_info {
469 	u32 ofs;
470 	char *name;
471 };
472 
473 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
474 
475 	/* General Registers */
476 	{IXGBE_CTRL, "CTRL"},
477 	{IXGBE_STATUS, "STATUS"},
478 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
479 
480 	/* Interrupt Registers */
481 	{IXGBE_EICR, "EICR"},
482 
483 	/* RX Registers */
484 	{IXGBE_SRRCTL(0), "SRRCTL"},
485 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
486 	{IXGBE_RDLEN(0), "RDLEN"},
487 	{IXGBE_RDH(0), "RDH"},
488 	{IXGBE_RDT(0), "RDT"},
489 	{IXGBE_RXDCTL(0), "RXDCTL"},
490 	{IXGBE_RDBAL(0), "RDBAL"},
491 	{IXGBE_RDBAH(0), "RDBAH"},
492 
493 	/* TX Registers */
494 	{IXGBE_TDBAL(0), "TDBAL"},
495 	{IXGBE_TDBAH(0), "TDBAH"},
496 	{IXGBE_TDLEN(0), "TDLEN"},
497 	{IXGBE_TDH(0), "TDH"},
498 	{IXGBE_TDT(0), "TDT"},
499 	{IXGBE_TXDCTL(0), "TXDCTL"},
500 
501 	/* List Terminator */
502 	{ .name = NULL }
503 };
504 
505 
506 /*
507  * ixgbe_regdump - register printout routine
508  */
509 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
510 {
511 	int i = 0, j = 0;
512 	char rname[16];
513 	u32 regs[64];
514 
515 	switch (reginfo->ofs) {
516 	case IXGBE_SRRCTL(0):
517 		for (i = 0; i < 64; i++)
518 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
519 		break;
520 	case IXGBE_DCA_RXCTRL(0):
521 		for (i = 0; i < 64; i++)
522 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
523 		break;
524 	case IXGBE_RDLEN(0):
525 		for (i = 0; i < 64; i++)
526 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
527 		break;
528 	case IXGBE_RDH(0):
529 		for (i = 0; i < 64; i++)
530 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
531 		break;
532 	case IXGBE_RDT(0):
533 		for (i = 0; i < 64; i++)
534 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
535 		break;
536 	case IXGBE_RXDCTL(0):
537 		for (i = 0; i < 64; i++)
538 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
539 		break;
540 	case IXGBE_RDBAL(0):
541 		for (i = 0; i < 64; i++)
542 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
543 		break;
544 	case IXGBE_RDBAH(0):
545 		for (i = 0; i < 64; i++)
546 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
547 		break;
548 	case IXGBE_TDBAL(0):
549 		for (i = 0; i < 64; i++)
550 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
551 		break;
552 	case IXGBE_TDBAH(0):
553 		for (i = 0; i < 64; i++)
554 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
555 		break;
556 	case IXGBE_TDLEN(0):
557 		for (i = 0; i < 64; i++)
558 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
559 		break;
560 	case IXGBE_TDH(0):
561 		for (i = 0; i < 64; i++)
562 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
563 		break;
564 	case IXGBE_TDT(0):
565 		for (i = 0; i < 64; i++)
566 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
567 		break;
568 	case IXGBE_TXDCTL(0):
569 		for (i = 0; i < 64; i++)
570 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
571 		break;
572 	default:
573 		pr_info("%-15s %08x\n", reginfo->name,
574 			IXGBE_READ_REG(hw, reginfo->ofs));
575 		return;
576 	}
577 
578 	for (i = 0; i < 8; i++) {
579 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
580 		pr_err("%-15s", rname);
581 		for (j = 0; j < 8; j++)
582 			pr_cont(" %08x", regs[i*8+j]);
583 		pr_cont("\n");
584 	}
585 
586 }
587 
588 /*
589  * ixgbe_dump - Print registers, tx-rings and rx-rings
590  */
591 static void ixgbe_dump(struct ixgbe_adapter *adapter)
592 {
593 	struct net_device *netdev = adapter->netdev;
594 	struct ixgbe_hw *hw = &adapter->hw;
595 	struct ixgbe_reg_info *reginfo;
596 	int n = 0;
597 	struct ixgbe_ring *tx_ring;
598 	struct ixgbe_tx_buffer *tx_buffer;
599 	union ixgbe_adv_tx_desc *tx_desc;
600 	struct my_u0 { u64 a; u64 b; } *u0;
601 	struct ixgbe_ring *rx_ring;
602 	union ixgbe_adv_rx_desc *rx_desc;
603 	struct ixgbe_rx_buffer *rx_buffer_info;
604 	u32 staterr;
605 	int i = 0;
606 
607 	if (!netif_msg_hw(adapter))
608 		return;
609 
610 	/* Print netdevice Info */
611 	if (netdev) {
612 		dev_info(&adapter->pdev->dev, "Net device Info\n");
613 		pr_info("Device Name     state            "
614 			"trans_start\n");
615 		pr_info("%-15s %016lX %016lX\n",
616 			netdev->name,
617 			netdev->state,
618 			dev_trans_start(netdev));
619 	}
620 
621 	/* Print Registers */
622 	dev_info(&adapter->pdev->dev, "Register Dump\n");
623 	pr_info(" Register Name   Value\n");
624 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
625 	     reginfo->name; reginfo++) {
626 		ixgbe_regdump(hw, reginfo);
627 	}
628 
629 	/* Print TX Ring Summary */
630 	if (!netdev || !netif_running(netdev))
631 		return;
632 
633 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
634 	pr_info(" %s     %s              %s        %s\n",
635 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
636 		"leng", "ntw", "timestamp");
637 	for (n = 0; n < adapter->num_tx_queues; n++) {
638 		tx_ring = adapter->tx_ring[n];
639 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
640 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
641 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
642 			   (u64)dma_unmap_addr(tx_buffer, dma),
643 			   dma_unmap_len(tx_buffer, len),
644 			   tx_buffer->next_to_watch,
645 			   (u64)tx_buffer->time_stamp);
646 	}
647 
648 	/* Print TX Rings */
649 	if (!netif_msg_tx_done(adapter))
650 		goto rx_ring_summary;
651 
652 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
653 
654 	/* Transmit Descriptor Formats
655 	 *
656 	 * 82598 Advanced Transmit Descriptor
657 	 *   +--------------------------------------------------------------+
658 	 * 0 |         Buffer Address [63:0]                                |
659 	 *   +--------------------------------------------------------------+
660 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
661 	 *   +--------------------------------------------------------------+
662 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
663 	 *
664 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
665 	 *   +--------------------------------------------------------------+
666 	 * 0 |                          RSV [63:0]                          |
667 	 *   +--------------------------------------------------------------+
668 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
669 	 *   +--------------------------------------------------------------+
670 	 *   63                       36 35   32 31                         0
671 	 *
672 	 * 82599+ Advanced Transmit Descriptor
673 	 *   +--------------------------------------------------------------+
674 	 * 0 |         Buffer Address [63:0]                                |
675 	 *   +--------------------------------------------------------------+
676 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
677 	 *   +--------------------------------------------------------------+
678 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
679 	 *
680 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
681 	 *   +--------------------------------------------------------------+
682 	 * 0 |                          RSV [63:0]                          |
683 	 *   +--------------------------------------------------------------+
684 	 * 8 |            RSV           |  STA  |           RSV             |
685 	 *   +--------------------------------------------------------------+
686 	 *   63                       36 35   32 31                         0
687 	 */
688 
689 	for (n = 0; n < adapter->num_tx_queues; n++) {
690 		tx_ring = adapter->tx_ring[n];
691 		pr_info("------------------------------------\n");
692 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
693 		pr_info("------------------------------------\n");
694 		pr_info("%s%s    %s              %s        %s          %s\n",
695 			"T [desc]     [address 63:0  ] ",
696 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
697 			"leng", "ntw", "timestamp", "bi->skb");
698 
699 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
700 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
701 			tx_buffer = &tx_ring->tx_buffer_info[i];
702 			u0 = (struct my_u0 *)tx_desc;
703 			if (dma_unmap_len(tx_buffer, len) > 0) {
704 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
705 					i,
706 					le64_to_cpu(u0->a),
707 					le64_to_cpu(u0->b),
708 					(u64)dma_unmap_addr(tx_buffer, dma),
709 					dma_unmap_len(tx_buffer, len),
710 					tx_buffer->next_to_watch,
711 					(u64)tx_buffer->time_stamp,
712 					tx_buffer->skb);
713 				if (i == tx_ring->next_to_use &&
714 					i == tx_ring->next_to_clean)
715 					pr_cont(" NTC/U\n");
716 				else if (i == tx_ring->next_to_use)
717 					pr_cont(" NTU\n");
718 				else if (i == tx_ring->next_to_clean)
719 					pr_cont(" NTC\n");
720 				else
721 					pr_cont("\n");
722 
723 				if (netif_msg_pktdata(adapter) &&
724 				    tx_buffer->skb)
725 					print_hex_dump(KERN_INFO, "",
726 						DUMP_PREFIX_ADDRESS, 16, 1,
727 						tx_buffer->skb->data,
728 						dma_unmap_len(tx_buffer, len),
729 						true);
730 			}
731 		}
732 	}
733 
734 	/* Print RX Rings Summary */
735 rx_ring_summary:
736 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
737 	pr_info("Queue [NTU] [NTC]\n");
738 	for (n = 0; n < adapter->num_rx_queues; n++) {
739 		rx_ring = adapter->rx_ring[n];
740 		pr_info("%5d %5X %5X\n",
741 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
742 	}
743 
744 	/* Print RX Rings */
745 	if (!netif_msg_rx_status(adapter))
746 		return;
747 
748 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
749 
750 	/* Receive Descriptor Formats
751 	 *
752 	 * 82598 Advanced Receive Descriptor (Read) Format
753 	 *    63                                           1        0
754 	 *    +-----------------------------------------------------+
755 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
756 	 *    +----------------------------------------------+------+
757 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
758 	 *    +-----------------------------------------------------+
759 	 *
760 	 *
761 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
762 	 *
763 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
764 	 *   +------------------------------------------------------+
765 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
766 	 *   | Packet   | IP     |   |          |     | Type | Type |
767 	 *   | Checksum | Ident  |   |          |     |      |      |
768 	 *   +------------------------------------------------------+
769 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
770 	 *   +------------------------------------------------------+
771 	 *   63       48 47    32 31            20 19               0
772 	 *
773 	 * 82599+ Advanced Receive Descriptor (Read) Format
774 	 *    63                                           1        0
775 	 *    +-----------------------------------------------------+
776 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
777 	 *    +----------------------------------------------+------+
778 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
779 	 *    +-----------------------------------------------------+
780 	 *
781 	 *
782 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
783 	 *
784 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
785 	 *   +------------------------------------------------------+
786 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
787 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
788 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
789 	 *   +------------------------------------------------------+
790 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
791 	 *   +------------------------------------------------------+
792 	 *   63       48 47    32 31          20 19                 0
793 	 */
794 
795 	for (n = 0; n < adapter->num_rx_queues; n++) {
796 		rx_ring = adapter->rx_ring[n];
797 		pr_info("------------------------------------\n");
798 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
799 		pr_info("------------------------------------\n");
800 		pr_info("%s%s%s",
801 			"R  [desc]      [ PktBuf     A0] ",
802 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
803 			"<-- Adv Rx Read format\n");
804 		pr_info("%s%s%s",
805 			"RWB[desc]      [PcsmIpSHl PtRs] ",
806 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
807 			"<-- Adv Rx Write-Back format\n");
808 
809 		for (i = 0; i < rx_ring->count; i++) {
810 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
811 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
812 			u0 = (struct my_u0 *)rx_desc;
813 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
814 			if (staterr & IXGBE_RXD_STAT_DD) {
815 				/* Descriptor Done */
816 				pr_info("RWB[0x%03X]     %016llX "
817 					"%016llX ---------------- %p", i,
818 					le64_to_cpu(u0->a),
819 					le64_to_cpu(u0->b),
820 					rx_buffer_info->skb);
821 			} else {
822 				pr_info("R  [0x%03X]     %016llX "
823 					"%016llX %016llX %p", i,
824 					le64_to_cpu(u0->a),
825 					le64_to_cpu(u0->b),
826 					(u64)rx_buffer_info->dma,
827 					rx_buffer_info->skb);
828 
829 				if (netif_msg_pktdata(adapter) &&
830 				    rx_buffer_info->dma) {
831 					print_hex_dump(KERN_INFO, "",
832 					   DUMP_PREFIX_ADDRESS, 16, 1,
833 					   page_address(rx_buffer_info->page) +
834 						    rx_buffer_info->page_offset,
835 					   ixgbe_rx_bufsz(rx_ring), true);
836 				}
837 			}
838 
839 			if (i == rx_ring->next_to_use)
840 				pr_cont(" NTU\n");
841 			else if (i == rx_ring->next_to_clean)
842 				pr_cont(" NTC\n");
843 			else
844 				pr_cont("\n");
845 
846 		}
847 	}
848 }
849 
850 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
851 {
852 	u32 ctrl_ext;
853 
854 	/* Let firmware take over control of h/w */
855 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
856 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
857 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
858 }
859 
860 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
861 {
862 	u32 ctrl_ext;
863 
864 	/* Let firmware know the driver has taken over */
865 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
866 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
867 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
868 }
869 
870 /**
871  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
872  * @adapter: pointer to adapter struct
873  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
874  * @queue: queue to map the corresponding interrupt to
875  * @msix_vector: the vector to map to the corresponding queue
876  *
877  */
878 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
879 			   u8 queue, u8 msix_vector)
880 {
881 	u32 ivar, index;
882 	struct ixgbe_hw *hw = &adapter->hw;
883 	switch (hw->mac.type) {
884 	case ixgbe_mac_82598EB:
885 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
886 		if (direction == -1)
887 			direction = 0;
888 		index = (((direction * 64) + queue) >> 2) & 0x1F;
889 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
890 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
891 		ivar |= (msix_vector << (8 * (queue & 0x3)));
892 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
893 		break;
894 	case ixgbe_mac_82599EB:
895 	case ixgbe_mac_X540:
896 	case ixgbe_mac_X550:
897 	case ixgbe_mac_X550EM_x:
898 	case ixgbe_mac_x550em_a:
899 		if (direction == -1) {
900 			/* other causes */
901 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
902 			index = ((queue & 1) * 8);
903 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
904 			ivar &= ~(0xFF << index);
905 			ivar |= (msix_vector << index);
906 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
907 			break;
908 		} else {
909 			/* tx or rx causes */
910 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
911 			index = ((16 * (queue & 1)) + (8 * direction));
912 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
913 			ivar &= ~(0xFF << index);
914 			ivar |= (msix_vector << index);
915 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
916 			break;
917 		}
918 	default:
919 		break;
920 	}
921 }
922 
923 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
924 					  u64 qmask)
925 {
926 	u32 mask;
927 
928 	switch (adapter->hw.mac.type) {
929 	case ixgbe_mac_82598EB:
930 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
931 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
932 		break;
933 	case ixgbe_mac_82599EB:
934 	case ixgbe_mac_X540:
935 	case ixgbe_mac_X550:
936 	case ixgbe_mac_X550EM_x:
937 	case ixgbe_mac_x550em_a:
938 		mask = (qmask & 0xFFFFFFFF);
939 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
940 		mask = (qmask >> 32);
941 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
942 		break;
943 	default:
944 		break;
945 	}
946 }
947 
948 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
949 {
950 	struct ixgbe_hw *hw = &adapter->hw;
951 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
952 	int i;
953 	u32 data;
954 
955 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
956 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
957 		return;
958 
959 	switch (hw->mac.type) {
960 	case ixgbe_mac_82598EB:
961 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
962 		break;
963 	default:
964 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
965 	}
966 	hwstats->lxoffrxc += data;
967 
968 	/* refill credits (no tx hang) if we received xoff */
969 	if (!data)
970 		return;
971 
972 	for (i = 0; i < adapter->num_tx_queues; i++)
973 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
974 			  &adapter->tx_ring[i]->state);
975 }
976 
977 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
978 {
979 	struct ixgbe_hw *hw = &adapter->hw;
980 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
981 	u32 xoff[8] = {0};
982 	u8 tc;
983 	int i;
984 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
985 
986 	if (adapter->ixgbe_ieee_pfc)
987 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
988 
989 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
990 		ixgbe_update_xoff_rx_lfc(adapter);
991 		return;
992 	}
993 
994 	/* update stats for each tc, only valid with PFC enabled */
995 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
996 		u32 pxoffrxc;
997 
998 		switch (hw->mac.type) {
999 		case ixgbe_mac_82598EB:
1000 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1001 			break;
1002 		default:
1003 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1004 		}
1005 		hwstats->pxoffrxc[i] += pxoffrxc;
1006 		/* Get the TC for given UP */
1007 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1008 		xoff[tc] += pxoffrxc;
1009 	}
1010 
1011 	/* disarm tx queues that have received xoff frames */
1012 	for (i = 0; i < adapter->num_tx_queues; i++) {
1013 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1014 
1015 		tc = tx_ring->dcb_tc;
1016 		if (xoff[tc])
1017 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1018 	}
1019 }
1020 
1021 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1022 {
1023 	return ring->stats.packets;
1024 }
1025 
1026 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1027 {
1028 	struct ixgbe_adapter *adapter;
1029 	struct ixgbe_hw *hw;
1030 	u32 head, tail;
1031 
1032 	if (ring->l2_accel_priv)
1033 		adapter = ring->l2_accel_priv->real_adapter;
1034 	else
1035 		adapter = netdev_priv(ring->netdev);
1036 
1037 	hw = &adapter->hw;
1038 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1039 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1040 
1041 	if (head != tail)
1042 		return (head < tail) ?
1043 			tail - head : (tail + ring->count - head);
1044 
1045 	return 0;
1046 }
1047 
1048 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1049 {
1050 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1051 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1052 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1053 
1054 	clear_check_for_tx_hang(tx_ring);
1055 
1056 	/*
1057 	 * Check for a hung queue, but be thorough. This verifies
1058 	 * that a transmit has been completed since the previous
1059 	 * check AND there is at least one packet pending. The
1060 	 * ARMED bit is set to indicate a potential hang. The
1061 	 * bit is cleared if a pause frame is received to remove
1062 	 * false hang detection due to PFC or 802.3x frames. By
1063 	 * requiring this to fail twice we avoid races with
1064 	 * pfc clearing the ARMED bit and conditions where we
1065 	 * run the check_tx_hang logic with a transmit completion
1066 	 * pending but without time to complete it yet.
1067 	 */
1068 	if (tx_done_old == tx_done && tx_pending)
1069 		/* make sure it is true for two checks in a row */
1070 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1071 					&tx_ring->state);
1072 	/* update completed stats and continue */
1073 	tx_ring->tx_stats.tx_done_old = tx_done;
1074 	/* reset the countdown */
1075 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1076 
1077 	return false;
1078 }
1079 
1080 /**
1081  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1082  * @adapter: driver private struct
1083  **/
1084 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1085 {
1086 
1087 	/* Do the reset outside of interrupt context */
1088 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1089 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1090 		e_warn(drv, "initiating reset due to tx timeout\n");
1091 		ixgbe_service_event_schedule(adapter);
1092 	}
1093 }
1094 
1095 /**
1096  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1097  **/
1098 static int ixgbe_tx_maxrate(struct net_device *netdev,
1099 			    int queue_index, u32 maxrate)
1100 {
1101 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1102 	struct ixgbe_hw *hw = &adapter->hw;
1103 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1104 
1105 	if (!maxrate)
1106 		return 0;
1107 
1108 	/* Calculate the rate factor values to set */
1109 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1110 	bcnrc_val /= maxrate;
1111 
1112 	/* clear everything but the rate factor */
1113 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1114 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1115 
1116 	/* enable the rate scheduler */
1117 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1118 
1119 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1120 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1121 
1122 	return 0;
1123 }
1124 
1125 /**
1126  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1127  * @q_vector: structure containing interrupt and ring information
1128  * @tx_ring: tx ring to clean
1129  * @napi_budget: Used to determine if we are in netpoll
1130  **/
1131 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1132 			       struct ixgbe_ring *tx_ring, int napi_budget)
1133 {
1134 	struct ixgbe_adapter *adapter = q_vector->adapter;
1135 	struct ixgbe_tx_buffer *tx_buffer;
1136 	union ixgbe_adv_tx_desc *tx_desc;
1137 	unsigned int total_bytes = 0, total_packets = 0;
1138 	unsigned int budget = q_vector->tx.work_limit;
1139 	unsigned int i = tx_ring->next_to_clean;
1140 
1141 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1142 		return true;
1143 
1144 	tx_buffer = &tx_ring->tx_buffer_info[i];
1145 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1146 	i -= tx_ring->count;
1147 
1148 	do {
1149 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1150 
1151 		/* if next_to_watch is not set then there is no work pending */
1152 		if (!eop_desc)
1153 			break;
1154 
1155 		/* prevent any other reads prior to eop_desc */
1156 		read_barrier_depends();
1157 
1158 		/* if DD is not set pending work has not been completed */
1159 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1160 			break;
1161 
1162 		/* clear next_to_watch to prevent false hangs */
1163 		tx_buffer->next_to_watch = NULL;
1164 
1165 		/* update the statistics for this packet */
1166 		total_bytes += tx_buffer->bytecount;
1167 		total_packets += tx_buffer->gso_segs;
1168 
1169 		/* free the skb */
1170 		napi_consume_skb(tx_buffer->skb, napi_budget);
1171 
1172 		/* unmap skb header data */
1173 		dma_unmap_single(tx_ring->dev,
1174 				 dma_unmap_addr(tx_buffer, dma),
1175 				 dma_unmap_len(tx_buffer, len),
1176 				 DMA_TO_DEVICE);
1177 
1178 		/* clear tx_buffer data */
1179 		dma_unmap_len_set(tx_buffer, len, 0);
1180 
1181 		/* unmap remaining buffers */
1182 		while (tx_desc != eop_desc) {
1183 			tx_buffer++;
1184 			tx_desc++;
1185 			i++;
1186 			if (unlikely(!i)) {
1187 				i -= tx_ring->count;
1188 				tx_buffer = tx_ring->tx_buffer_info;
1189 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1190 			}
1191 
1192 			/* unmap any remaining paged data */
1193 			if (dma_unmap_len(tx_buffer, len)) {
1194 				dma_unmap_page(tx_ring->dev,
1195 					       dma_unmap_addr(tx_buffer, dma),
1196 					       dma_unmap_len(tx_buffer, len),
1197 					       DMA_TO_DEVICE);
1198 				dma_unmap_len_set(tx_buffer, len, 0);
1199 			}
1200 		}
1201 
1202 		/* move us one more past the eop_desc for start of next pkt */
1203 		tx_buffer++;
1204 		tx_desc++;
1205 		i++;
1206 		if (unlikely(!i)) {
1207 			i -= tx_ring->count;
1208 			tx_buffer = tx_ring->tx_buffer_info;
1209 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1210 		}
1211 
1212 		/* issue prefetch for next Tx descriptor */
1213 		prefetch(tx_desc);
1214 
1215 		/* update budget accounting */
1216 		budget--;
1217 	} while (likely(budget));
1218 
1219 	i += tx_ring->count;
1220 	tx_ring->next_to_clean = i;
1221 	u64_stats_update_begin(&tx_ring->syncp);
1222 	tx_ring->stats.bytes += total_bytes;
1223 	tx_ring->stats.packets += total_packets;
1224 	u64_stats_update_end(&tx_ring->syncp);
1225 	q_vector->tx.total_bytes += total_bytes;
1226 	q_vector->tx.total_packets += total_packets;
1227 
1228 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1229 		/* schedule immediate reset if we believe we hung */
1230 		struct ixgbe_hw *hw = &adapter->hw;
1231 		e_err(drv, "Detected Tx Unit Hang\n"
1232 			"  Tx Queue             <%d>\n"
1233 			"  TDH, TDT             <%x>, <%x>\n"
1234 			"  next_to_use          <%x>\n"
1235 			"  next_to_clean        <%x>\n"
1236 			"tx_buffer_info[next_to_clean]\n"
1237 			"  time_stamp           <%lx>\n"
1238 			"  jiffies              <%lx>\n",
1239 			tx_ring->queue_index,
1240 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1241 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1242 			tx_ring->next_to_use, i,
1243 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1244 
1245 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1246 
1247 		e_info(probe,
1248 		       "tx hang %d detected on queue %d, resetting adapter\n",
1249 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1250 
1251 		/* schedule immediate reset if we believe we hung */
1252 		ixgbe_tx_timeout_reset(adapter);
1253 
1254 		/* the adapter is about to reset, no point in enabling stuff */
1255 		return true;
1256 	}
1257 
1258 	netdev_tx_completed_queue(txring_txq(tx_ring),
1259 				  total_packets, total_bytes);
1260 
1261 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1262 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1263 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1264 		/* Make sure that anybody stopping the queue after this
1265 		 * sees the new next_to_clean.
1266 		 */
1267 		smp_mb();
1268 		if (__netif_subqueue_stopped(tx_ring->netdev,
1269 					     tx_ring->queue_index)
1270 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1271 			netif_wake_subqueue(tx_ring->netdev,
1272 					    tx_ring->queue_index);
1273 			++tx_ring->tx_stats.restart_queue;
1274 		}
1275 	}
1276 
1277 	return !!budget;
1278 }
1279 
1280 #ifdef CONFIG_IXGBE_DCA
1281 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1282 				struct ixgbe_ring *tx_ring,
1283 				int cpu)
1284 {
1285 	struct ixgbe_hw *hw = &adapter->hw;
1286 	u32 txctrl = 0;
1287 	u16 reg_offset;
1288 
1289 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1290 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1291 
1292 	switch (hw->mac.type) {
1293 	case ixgbe_mac_82598EB:
1294 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1295 		break;
1296 	case ixgbe_mac_82599EB:
1297 	case ixgbe_mac_X540:
1298 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1299 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1300 		break;
1301 	default:
1302 		/* for unknown hardware do not write register */
1303 		return;
1304 	}
1305 
1306 	/*
1307 	 * We can enable relaxed ordering for reads, but not writes when
1308 	 * DCA is enabled.  This is due to a known issue in some chipsets
1309 	 * which will cause the DCA tag to be cleared.
1310 	 */
1311 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1312 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1313 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1314 
1315 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1316 }
1317 
1318 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1319 				struct ixgbe_ring *rx_ring,
1320 				int cpu)
1321 {
1322 	struct ixgbe_hw *hw = &adapter->hw;
1323 	u32 rxctrl = 0;
1324 	u8 reg_idx = rx_ring->reg_idx;
1325 
1326 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1327 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1328 
1329 	switch (hw->mac.type) {
1330 	case ixgbe_mac_82599EB:
1331 	case ixgbe_mac_X540:
1332 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1333 		break;
1334 	default:
1335 		break;
1336 	}
1337 
1338 	/*
1339 	 * We can enable relaxed ordering for reads, but not writes when
1340 	 * DCA is enabled.  This is due to a known issue in some chipsets
1341 	 * which will cause the DCA tag to be cleared.
1342 	 */
1343 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1344 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1345 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1346 
1347 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1348 }
1349 
1350 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1351 {
1352 	struct ixgbe_adapter *adapter = q_vector->adapter;
1353 	struct ixgbe_ring *ring;
1354 	int cpu = get_cpu();
1355 
1356 	if (q_vector->cpu == cpu)
1357 		goto out_no_update;
1358 
1359 	ixgbe_for_each_ring(ring, q_vector->tx)
1360 		ixgbe_update_tx_dca(adapter, ring, cpu);
1361 
1362 	ixgbe_for_each_ring(ring, q_vector->rx)
1363 		ixgbe_update_rx_dca(adapter, ring, cpu);
1364 
1365 	q_vector->cpu = cpu;
1366 out_no_update:
1367 	put_cpu();
1368 }
1369 
1370 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1371 {
1372 	int i;
1373 
1374 	/* always use CB2 mode, difference is masked in the CB driver */
1375 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1376 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1377 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1378 	else
1379 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1380 				IXGBE_DCA_CTRL_DCA_DISABLE);
1381 
1382 	for (i = 0; i < adapter->num_q_vectors; i++) {
1383 		adapter->q_vector[i]->cpu = -1;
1384 		ixgbe_update_dca(adapter->q_vector[i]);
1385 	}
1386 }
1387 
1388 static int __ixgbe_notify_dca(struct device *dev, void *data)
1389 {
1390 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1391 	unsigned long event = *(unsigned long *)data;
1392 
1393 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1394 		return 0;
1395 
1396 	switch (event) {
1397 	case DCA_PROVIDER_ADD:
1398 		/* if we're already enabled, don't do it again */
1399 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1400 			break;
1401 		if (dca_add_requester(dev) == 0) {
1402 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1403 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1404 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1405 			break;
1406 		}
1407 		/* Fall Through since DCA is disabled. */
1408 	case DCA_PROVIDER_REMOVE:
1409 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1410 			dca_remove_requester(dev);
1411 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1412 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1413 					IXGBE_DCA_CTRL_DCA_DISABLE);
1414 		}
1415 		break;
1416 	}
1417 
1418 	return 0;
1419 }
1420 
1421 #endif /* CONFIG_IXGBE_DCA */
1422 
1423 #define IXGBE_RSS_L4_TYPES_MASK \
1424 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1425 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1426 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1427 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1428 
1429 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1430 				 union ixgbe_adv_rx_desc *rx_desc,
1431 				 struct sk_buff *skb)
1432 {
1433 	u16 rss_type;
1434 
1435 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1436 		return;
1437 
1438 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1439 		   IXGBE_RXDADV_RSSTYPE_MASK;
1440 
1441 	if (!rss_type)
1442 		return;
1443 
1444 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1445 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1446 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1447 }
1448 
1449 #ifdef IXGBE_FCOE
1450 /**
1451  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1452  * @ring: structure containing ring specific data
1453  * @rx_desc: advanced rx descriptor
1454  *
1455  * Returns : true if it is FCoE pkt
1456  */
1457 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1458 				    union ixgbe_adv_rx_desc *rx_desc)
1459 {
1460 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1461 
1462 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1463 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1464 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1465 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1466 }
1467 
1468 #endif /* IXGBE_FCOE */
1469 /**
1470  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1471  * @ring: structure containing ring specific data
1472  * @rx_desc: current Rx descriptor being processed
1473  * @skb: skb currently being received and modified
1474  **/
1475 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1476 				     union ixgbe_adv_rx_desc *rx_desc,
1477 				     struct sk_buff *skb)
1478 {
1479 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1480 	bool encap_pkt = false;
1481 
1482 	skb_checksum_none_assert(skb);
1483 
1484 	/* Rx csum disabled */
1485 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1486 		return;
1487 
1488 	/* check for VXLAN and Geneve packets */
1489 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1490 		encap_pkt = true;
1491 		skb->encapsulation = 1;
1492 	}
1493 
1494 	/* if IP and error */
1495 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1496 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1497 		ring->rx_stats.csum_err++;
1498 		return;
1499 	}
1500 
1501 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1502 		return;
1503 
1504 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1505 		/*
1506 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1507 		 * checksum errors.
1508 		 */
1509 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1510 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1511 			return;
1512 
1513 		ring->rx_stats.csum_err++;
1514 		return;
1515 	}
1516 
1517 	/* It must be a TCP or UDP packet with a valid checksum */
1518 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1519 	if (encap_pkt) {
1520 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1521 			return;
1522 
1523 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1524 			skb->ip_summed = CHECKSUM_NONE;
1525 			return;
1526 		}
1527 		/* If we checked the outer header let the stack know */
1528 		skb->csum_level = 1;
1529 	}
1530 }
1531 
1532 static inline unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1533 {
1534 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1535 }
1536 
1537 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1538 				    struct ixgbe_rx_buffer *bi)
1539 {
1540 	struct page *page = bi->page;
1541 	dma_addr_t dma;
1542 
1543 	/* since we are recycling buffers we should seldom need to alloc */
1544 	if (likely(page))
1545 		return true;
1546 
1547 	/* alloc new page for storage */
1548 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1549 	if (unlikely(!page)) {
1550 		rx_ring->rx_stats.alloc_rx_page_failed++;
1551 		return false;
1552 	}
1553 
1554 	/* map page for use */
1555 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1556 				 ixgbe_rx_pg_size(rx_ring),
1557 				 DMA_FROM_DEVICE,
1558 				 IXGBE_RX_DMA_ATTR);
1559 
1560 	/*
1561 	 * if mapping failed free memory back to system since
1562 	 * there isn't much point in holding memory we can't use
1563 	 */
1564 	if (dma_mapping_error(rx_ring->dev, dma)) {
1565 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1566 
1567 		rx_ring->rx_stats.alloc_rx_page_failed++;
1568 		return false;
1569 	}
1570 
1571 	bi->dma = dma;
1572 	bi->page = page;
1573 	bi->page_offset = ixgbe_rx_offset(rx_ring);
1574 	bi->pagecnt_bias = 1;
1575 
1576 	return true;
1577 }
1578 
1579 /**
1580  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1581  * @rx_ring: ring to place buffers on
1582  * @cleaned_count: number of buffers to replace
1583  **/
1584 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1585 {
1586 	union ixgbe_adv_rx_desc *rx_desc;
1587 	struct ixgbe_rx_buffer *bi;
1588 	u16 i = rx_ring->next_to_use;
1589 	u16 bufsz;
1590 
1591 	/* nothing to do */
1592 	if (!cleaned_count)
1593 		return;
1594 
1595 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1596 	bi = &rx_ring->rx_buffer_info[i];
1597 	i -= rx_ring->count;
1598 
1599 	bufsz = ixgbe_rx_bufsz(rx_ring);
1600 
1601 	do {
1602 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1603 			break;
1604 
1605 		/* sync the buffer for use by the device */
1606 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1607 						 bi->page_offset, bufsz,
1608 						 DMA_FROM_DEVICE);
1609 
1610 		/*
1611 		 * Refresh the desc even if buffer_addrs didn't change
1612 		 * because each write-back erases this info.
1613 		 */
1614 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1615 
1616 		rx_desc++;
1617 		bi++;
1618 		i++;
1619 		if (unlikely(!i)) {
1620 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1621 			bi = rx_ring->rx_buffer_info;
1622 			i -= rx_ring->count;
1623 		}
1624 
1625 		/* clear the length for the next_to_use descriptor */
1626 		rx_desc->wb.upper.length = 0;
1627 
1628 		cleaned_count--;
1629 	} while (cleaned_count);
1630 
1631 	i += rx_ring->count;
1632 
1633 	if (rx_ring->next_to_use != i) {
1634 		rx_ring->next_to_use = i;
1635 
1636 		/* update next to alloc since we have filled the ring */
1637 		rx_ring->next_to_alloc = i;
1638 
1639 		/* Force memory writes to complete before letting h/w
1640 		 * know there are new descriptors to fetch.  (Only
1641 		 * applicable for weak-ordered memory model archs,
1642 		 * such as IA-64).
1643 		 */
1644 		wmb();
1645 		writel(i, rx_ring->tail);
1646 	}
1647 }
1648 
1649 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1650 				   struct sk_buff *skb)
1651 {
1652 	u16 hdr_len = skb_headlen(skb);
1653 
1654 	/* set gso_size to avoid messing up TCP MSS */
1655 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1656 						 IXGBE_CB(skb)->append_cnt);
1657 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1658 }
1659 
1660 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1661 				   struct sk_buff *skb)
1662 {
1663 	/* if append_cnt is 0 then frame is not RSC */
1664 	if (!IXGBE_CB(skb)->append_cnt)
1665 		return;
1666 
1667 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1668 	rx_ring->rx_stats.rsc_flush++;
1669 
1670 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1671 
1672 	/* gso_size is computed using append_cnt so always clear it last */
1673 	IXGBE_CB(skb)->append_cnt = 0;
1674 }
1675 
1676 /**
1677  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1678  * @rx_ring: rx descriptor ring packet is being transacted on
1679  * @rx_desc: pointer to the EOP Rx descriptor
1680  * @skb: pointer to current skb being populated
1681  *
1682  * This function checks the ring, descriptor, and packet information in
1683  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1684  * other fields within the skb.
1685  **/
1686 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1687 				     union ixgbe_adv_rx_desc *rx_desc,
1688 				     struct sk_buff *skb)
1689 {
1690 	struct net_device *dev = rx_ring->netdev;
1691 	u32 flags = rx_ring->q_vector->adapter->flags;
1692 
1693 	ixgbe_update_rsc_stats(rx_ring, skb);
1694 
1695 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1696 
1697 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1698 
1699 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1700 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1701 
1702 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1703 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1704 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1705 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1706 	}
1707 
1708 	skb_record_rx_queue(skb, rx_ring->queue_index);
1709 
1710 	skb->protocol = eth_type_trans(skb, dev);
1711 }
1712 
1713 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1714 			 struct sk_buff *skb)
1715 {
1716 	napi_gro_receive(&q_vector->napi, skb);
1717 }
1718 
1719 /**
1720  * ixgbe_is_non_eop - process handling of non-EOP buffers
1721  * @rx_ring: Rx ring being processed
1722  * @rx_desc: Rx descriptor for current buffer
1723  * @skb: Current socket buffer containing buffer in progress
1724  *
1725  * This function updates next to clean.  If the buffer is an EOP buffer
1726  * this function exits returning false, otherwise it will place the
1727  * sk_buff in the next buffer to be chained and return true indicating
1728  * that this is in fact a non-EOP buffer.
1729  **/
1730 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1731 			     union ixgbe_adv_rx_desc *rx_desc,
1732 			     struct sk_buff *skb)
1733 {
1734 	u32 ntc = rx_ring->next_to_clean + 1;
1735 
1736 	/* fetch, update, and store next to clean */
1737 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1738 	rx_ring->next_to_clean = ntc;
1739 
1740 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1741 
1742 	/* update RSC append count if present */
1743 	if (ring_is_rsc_enabled(rx_ring)) {
1744 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1745 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1746 
1747 		if (unlikely(rsc_enabled)) {
1748 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1749 
1750 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1751 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1752 
1753 			/* update ntc based on RSC value */
1754 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1755 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1756 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1757 		}
1758 	}
1759 
1760 	/* if we are the last buffer then there is nothing else to do */
1761 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1762 		return false;
1763 
1764 	/* place skb in next buffer to be received */
1765 	rx_ring->rx_buffer_info[ntc].skb = skb;
1766 	rx_ring->rx_stats.non_eop_descs++;
1767 
1768 	return true;
1769 }
1770 
1771 /**
1772  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1773  * @rx_ring: rx descriptor ring packet is being transacted on
1774  * @skb: pointer to current skb being adjusted
1775  *
1776  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1777  * main difference between this version and the original function is that
1778  * this function can make several assumptions about the state of things
1779  * that allow for significant optimizations versus the standard function.
1780  * As a result we can do things like drop a frag and maintain an accurate
1781  * truesize for the skb.
1782  */
1783 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1784 			    struct sk_buff *skb)
1785 {
1786 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1787 	unsigned char *va;
1788 	unsigned int pull_len;
1789 
1790 	/*
1791 	 * it is valid to use page_address instead of kmap since we are
1792 	 * working with pages allocated out of the lomem pool per
1793 	 * alloc_page(GFP_ATOMIC)
1794 	 */
1795 	va = skb_frag_address(frag);
1796 
1797 	/*
1798 	 * we need the header to contain the greater of either ETH_HLEN or
1799 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1800 	 */
1801 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1802 
1803 	/* align pull length to size of long to optimize memcpy performance */
1804 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1805 
1806 	/* update all of the pointers */
1807 	skb_frag_size_sub(frag, pull_len);
1808 	frag->page_offset += pull_len;
1809 	skb->data_len -= pull_len;
1810 	skb->tail += pull_len;
1811 }
1812 
1813 /**
1814  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1815  * @rx_ring: rx descriptor ring packet is being transacted on
1816  * @skb: pointer to current skb being updated
1817  *
1818  * This function provides a basic DMA sync up for the first fragment of an
1819  * skb.  The reason for doing this is that the first fragment cannot be
1820  * unmapped until we have reached the end of packet descriptor for a buffer
1821  * chain.
1822  */
1823 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1824 				struct sk_buff *skb)
1825 {
1826 	/* if the page was released unmap it, else just sync our portion */
1827 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1828 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1829 				     ixgbe_rx_pg_size(rx_ring),
1830 				     DMA_FROM_DEVICE,
1831 				     IXGBE_RX_DMA_ATTR);
1832 	} else {
1833 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1834 
1835 		dma_sync_single_range_for_cpu(rx_ring->dev,
1836 					      IXGBE_CB(skb)->dma,
1837 					      frag->page_offset,
1838 					      skb_frag_size(frag),
1839 					      DMA_FROM_DEVICE);
1840 	}
1841 }
1842 
1843 /**
1844  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1845  * @rx_ring: rx descriptor ring packet is being transacted on
1846  * @rx_desc: pointer to the EOP Rx descriptor
1847  * @skb: pointer to current skb being fixed
1848  *
1849  * Check for corrupted packet headers caused by senders on the local L2
1850  * embedded NIC switch not setting up their Tx Descriptors right.  These
1851  * should be very rare.
1852  *
1853  * Also address the case where we are pulling data in on pages only
1854  * and as such no data is present in the skb header.
1855  *
1856  * In addition if skb is not at least 60 bytes we need to pad it so that
1857  * it is large enough to qualify as a valid Ethernet frame.
1858  *
1859  * Returns true if an error was encountered and skb was freed.
1860  **/
1861 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1862 				  union ixgbe_adv_rx_desc *rx_desc,
1863 				  struct sk_buff *skb)
1864 {
1865 	struct net_device *netdev = rx_ring->netdev;
1866 
1867 	/* verify that the packet does not have any known errors */
1868 	if (unlikely(ixgbe_test_staterr(rx_desc,
1869 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1870 	    !(netdev->features & NETIF_F_RXALL))) {
1871 		dev_kfree_skb_any(skb);
1872 		return true;
1873 	}
1874 
1875 	/* place header in linear portion of buffer */
1876 	if (!skb_headlen(skb))
1877 		ixgbe_pull_tail(rx_ring, skb);
1878 
1879 #ifdef IXGBE_FCOE
1880 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1881 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1882 		return false;
1883 
1884 #endif
1885 	/* if eth_skb_pad returns an error the skb was freed */
1886 	if (eth_skb_pad(skb))
1887 		return true;
1888 
1889 	return false;
1890 }
1891 
1892 /**
1893  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1894  * @rx_ring: rx descriptor ring to store buffers on
1895  * @old_buff: donor buffer to have page reused
1896  *
1897  * Synchronizes page for reuse by the adapter
1898  **/
1899 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1900 				struct ixgbe_rx_buffer *old_buff)
1901 {
1902 	struct ixgbe_rx_buffer *new_buff;
1903 	u16 nta = rx_ring->next_to_alloc;
1904 
1905 	new_buff = &rx_ring->rx_buffer_info[nta];
1906 
1907 	/* update, and store next to alloc */
1908 	nta++;
1909 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1910 
1911 	/* Transfer page from old buffer to new buffer.
1912 	 * Move each member individually to avoid possible store
1913 	 * forwarding stalls and unnecessary copy of skb.
1914 	 */
1915 	new_buff->dma		= old_buff->dma;
1916 	new_buff->page		= old_buff->page;
1917 	new_buff->page_offset	= old_buff->page_offset;
1918 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
1919 }
1920 
1921 static inline bool ixgbe_page_is_reserved(struct page *page)
1922 {
1923 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
1924 }
1925 
1926 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer)
1927 {
1928 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
1929 	struct page *page = rx_buffer->page;
1930 
1931 	/* avoid re-using remote pages */
1932 	if (unlikely(ixgbe_page_is_reserved(page)))
1933 		return false;
1934 
1935 #if (PAGE_SIZE < 8192)
1936 	/* if we are only owner of page we can reuse it */
1937 	if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
1938 		return false;
1939 #else
1940 	/* The last offset is a bit aggressive in that we assume the
1941 	 * worst case of FCoE being enabled and using a 3K buffer.
1942 	 * However this should have minimal impact as the 1K extra is
1943 	 * still less than one buffer in size.
1944 	 */
1945 #define IXGBE_LAST_OFFSET \
1946 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
1947 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
1948 		return false;
1949 #endif
1950 
1951 	/* If we have drained the page fragment pool we need to update
1952 	 * the pagecnt_bias and page count so that we fully restock the
1953 	 * number of references the driver holds.
1954 	 */
1955 	if (unlikely(!pagecnt_bias)) {
1956 		page_ref_add(page, USHRT_MAX);
1957 		rx_buffer->pagecnt_bias = USHRT_MAX;
1958 	}
1959 
1960 	return true;
1961 }
1962 
1963 /**
1964  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1965  * @rx_ring: rx descriptor ring to transact packets on
1966  * @rx_buffer: buffer containing page to add
1967  * @rx_desc: descriptor containing length of buffer written by hardware
1968  * @skb: sk_buff to place the data into
1969  *
1970  * This function will add the data contained in rx_buffer->page to the skb.
1971  * This is done either through a direct copy if the data in the buffer is
1972  * less than the skb header size, otherwise it will just attach the page as
1973  * a frag to the skb.
1974  *
1975  * The function will then update the page offset if necessary and return
1976  * true if the buffer can be reused by the adapter.
1977  **/
1978 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1979 			      struct ixgbe_rx_buffer *rx_buffer,
1980 			      struct sk_buff *skb,
1981 			      unsigned int size)
1982 {
1983 #if (PAGE_SIZE < 8192)
1984 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
1985 #else
1986 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
1987 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
1988 				SKB_DATA_ALIGN(size);
1989 #endif
1990 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
1991 			rx_buffer->page_offset, size, truesize);
1992 #if (PAGE_SIZE < 8192)
1993 	rx_buffer->page_offset ^= truesize;
1994 #else
1995 	rx_buffer->page_offset += truesize;
1996 #endif
1997 }
1998 
1999 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2000 						   union ixgbe_adv_rx_desc *rx_desc,
2001 						   struct sk_buff **skb,
2002 						   const unsigned int size)
2003 {
2004 	struct ixgbe_rx_buffer *rx_buffer;
2005 
2006 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2007 	prefetchw(rx_buffer->page);
2008 	*skb = rx_buffer->skb;
2009 
2010 	/* Delay unmapping of the first packet. It carries the header
2011 	 * information, HW may still access the header after the writeback.
2012 	 * Only unmap it when EOP is reached
2013 	 */
2014 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2015 		if (!*skb)
2016 			goto skip_sync;
2017 	} else {
2018 		if (*skb)
2019 			ixgbe_dma_sync_frag(rx_ring, *skb);
2020 	}
2021 
2022 	/* we are reusing so sync this buffer for CPU use */
2023 	dma_sync_single_range_for_cpu(rx_ring->dev,
2024 				      rx_buffer->dma,
2025 				      rx_buffer->page_offset,
2026 				      size,
2027 				      DMA_FROM_DEVICE);
2028 skip_sync:
2029 	rx_buffer->pagecnt_bias--;
2030 
2031 	return rx_buffer;
2032 }
2033 
2034 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2035 				struct ixgbe_rx_buffer *rx_buffer,
2036 				struct sk_buff *skb)
2037 {
2038 	if (ixgbe_can_reuse_rx_page(rx_buffer)) {
2039 		/* hand second half of page back to the ring */
2040 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2041 	} else {
2042 		if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
2043 			/* the page has been released from the ring */
2044 			IXGBE_CB(skb)->page_released = true;
2045 		} else {
2046 			/* we are not reusing the buffer so unmap it */
2047 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2048 					     ixgbe_rx_pg_size(rx_ring),
2049 					     DMA_FROM_DEVICE,
2050 					     IXGBE_RX_DMA_ATTR);
2051 		}
2052 		__page_frag_cache_drain(rx_buffer->page,
2053 					rx_buffer->pagecnt_bias);
2054 	}
2055 
2056 	/* clear contents of rx_buffer */
2057 	rx_buffer->page = NULL;
2058 	rx_buffer->skb = NULL;
2059 }
2060 
2061 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2062 					   struct ixgbe_rx_buffer *rx_buffer,
2063 					   union ixgbe_adv_rx_desc *rx_desc,
2064 					   unsigned int size)
2065 {
2066 	void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
2067 #if (PAGE_SIZE < 8192)
2068 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2069 #else
2070 	unsigned int truesize = SKB_DATA_ALIGN(size);
2071 #endif
2072 	struct sk_buff *skb;
2073 
2074 	/* prefetch first cache line of first page */
2075 	prefetch(va);
2076 #if L1_CACHE_BYTES < 128
2077 	prefetch(va + L1_CACHE_BYTES);
2078 #endif
2079 
2080 	/* allocate a skb to store the frags */
2081 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2082 	if (unlikely(!skb))
2083 		return NULL;
2084 
2085 	if (size > IXGBE_RX_HDR_SIZE) {
2086 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2087 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2088 
2089 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2090 				rx_buffer->page_offset,
2091 				size, truesize);
2092 #if (PAGE_SIZE < 8192)
2093 		rx_buffer->page_offset ^= truesize;
2094 #else
2095 		rx_buffer->page_offset += truesize;
2096 #endif
2097 	} else {
2098 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
2099 		rx_buffer->pagecnt_bias++;
2100 	}
2101 
2102 	return skb;
2103 }
2104 
2105 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2106 				       struct ixgbe_rx_buffer *rx_buffer,
2107 				       union ixgbe_adv_rx_desc *rx_desc,
2108 				       unsigned int size)
2109 {
2110 	void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
2111 #if (PAGE_SIZE < 8192)
2112 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2113 #else
2114 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2115 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size);
2116 #endif
2117 	struct sk_buff *skb;
2118 
2119 	/* prefetch first cache line of first page */
2120 	prefetch(va);
2121 #if L1_CACHE_BYTES < 128
2122 	prefetch(va + L1_CACHE_BYTES);
2123 #endif
2124 
2125 	/* build an skb to around the page buffer */
2126 	skb = build_skb(va - IXGBE_SKB_PAD, truesize);
2127 	if (unlikely(!skb))
2128 		return NULL;
2129 
2130 	/* update pointers within the skb to store the data */
2131 	skb_reserve(skb, IXGBE_SKB_PAD);
2132 	__skb_put(skb, size);
2133 
2134 	/* record DMA address if this is the start of a chain of buffers */
2135 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2136 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2137 
2138 	/* update buffer offset */
2139 #if (PAGE_SIZE < 8192)
2140 	rx_buffer->page_offset ^= truesize;
2141 #else
2142 	rx_buffer->page_offset += truesize;
2143 #endif
2144 
2145 	return skb;
2146 }
2147 
2148 /**
2149  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2150  * @q_vector: structure containing interrupt and ring information
2151  * @rx_ring: rx descriptor ring to transact packets on
2152  * @budget: Total limit on number of packets to process
2153  *
2154  * This function provides a "bounce buffer" approach to Rx interrupt
2155  * processing.  The advantage to this is that on systems that have
2156  * expensive overhead for IOMMU access this provides a means of avoiding
2157  * it by maintaining the mapping of the page to the syste.
2158  *
2159  * Returns amount of work completed
2160  **/
2161 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2162 			       struct ixgbe_ring *rx_ring,
2163 			       const int budget)
2164 {
2165 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2166 #ifdef IXGBE_FCOE
2167 	struct ixgbe_adapter *adapter = q_vector->adapter;
2168 	int ddp_bytes;
2169 	unsigned int mss = 0;
2170 #endif /* IXGBE_FCOE */
2171 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2172 
2173 	while (likely(total_rx_packets < budget)) {
2174 		union ixgbe_adv_rx_desc *rx_desc;
2175 		struct ixgbe_rx_buffer *rx_buffer;
2176 		struct sk_buff *skb;
2177 		unsigned int size;
2178 
2179 		/* return some buffers to hardware, one at a time is too slow */
2180 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2181 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2182 			cleaned_count = 0;
2183 		}
2184 
2185 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2186 		size = le16_to_cpu(rx_desc->wb.upper.length);
2187 		if (!size)
2188 			break;
2189 
2190 		/* This memory barrier is needed to keep us from reading
2191 		 * any other fields out of the rx_desc until we know the
2192 		 * descriptor has been written back
2193 		 */
2194 		dma_rmb();
2195 
2196 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size);
2197 
2198 		/* retrieve a buffer from the ring */
2199 		if (skb)
2200 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2201 		else if (ring_uses_build_skb(rx_ring))
2202 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2203 					      rx_desc, size);
2204 		else
2205 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2206 						  rx_desc, size);
2207 
2208 		/* exit if we failed to retrieve a buffer */
2209 		if (!skb) {
2210 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2211 			rx_buffer->pagecnt_bias++;
2212 			break;
2213 		}
2214 
2215 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb);
2216 		cleaned_count++;
2217 
2218 		/* place incomplete frames back on ring for completion */
2219 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2220 			continue;
2221 
2222 		/* verify the packet layout is correct */
2223 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2224 			continue;
2225 
2226 		/* probably a little skewed due to removing CRC */
2227 		total_rx_bytes += skb->len;
2228 
2229 		/* populate checksum, timestamp, VLAN, and protocol */
2230 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2231 
2232 #ifdef IXGBE_FCOE
2233 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2234 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2235 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2236 			/* include DDPed FCoE data */
2237 			if (ddp_bytes > 0) {
2238 				if (!mss) {
2239 					mss = rx_ring->netdev->mtu -
2240 						sizeof(struct fcoe_hdr) -
2241 						sizeof(struct fc_frame_header) -
2242 						sizeof(struct fcoe_crc_eof);
2243 					if (mss > 512)
2244 						mss &= ~511;
2245 				}
2246 				total_rx_bytes += ddp_bytes;
2247 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2248 								 mss);
2249 			}
2250 			if (!ddp_bytes) {
2251 				dev_kfree_skb_any(skb);
2252 				continue;
2253 			}
2254 		}
2255 
2256 #endif /* IXGBE_FCOE */
2257 		ixgbe_rx_skb(q_vector, skb);
2258 
2259 		/* update budget accounting */
2260 		total_rx_packets++;
2261 	}
2262 
2263 	u64_stats_update_begin(&rx_ring->syncp);
2264 	rx_ring->stats.packets += total_rx_packets;
2265 	rx_ring->stats.bytes += total_rx_bytes;
2266 	u64_stats_update_end(&rx_ring->syncp);
2267 	q_vector->rx.total_packets += total_rx_packets;
2268 	q_vector->rx.total_bytes += total_rx_bytes;
2269 
2270 	return total_rx_packets;
2271 }
2272 
2273 /**
2274  * ixgbe_configure_msix - Configure MSI-X hardware
2275  * @adapter: board private structure
2276  *
2277  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2278  * interrupts.
2279  **/
2280 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2281 {
2282 	struct ixgbe_q_vector *q_vector;
2283 	int v_idx;
2284 	u32 mask;
2285 
2286 	/* Populate MSIX to EITR Select */
2287 	if (adapter->num_vfs > 32) {
2288 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2289 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2290 	}
2291 
2292 	/*
2293 	 * Populate the IVAR table and set the ITR values to the
2294 	 * corresponding register.
2295 	 */
2296 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2297 		struct ixgbe_ring *ring;
2298 		q_vector = adapter->q_vector[v_idx];
2299 
2300 		ixgbe_for_each_ring(ring, q_vector->rx)
2301 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2302 
2303 		ixgbe_for_each_ring(ring, q_vector->tx)
2304 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2305 
2306 		ixgbe_write_eitr(q_vector);
2307 	}
2308 
2309 	switch (adapter->hw.mac.type) {
2310 	case ixgbe_mac_82598EB:
2311 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2312 			       v_idx);
2313 		break;
2314 	case ixgbe_mac_82599EB:
2315 	case ixgbe_mac_X540:
2316 	case ixgbe_mac_X550:
2317 	case ixgbe_mac_X550EM_x:
2318 	case ixgbe_mac_x550em_a:
2319 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2320 		break;
2321 	default:
2322 		break;
2323 	}
2324 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2325 
2326 	/* set up to autoclear timer, and the vectors */
2327 	mask = IXGBE_EIMS_ENABLE_MASK;
2328 	mask &= ~(IXGBE_EIMS_OTHER |
2329 		  IXGBE_EIMS_MAILBOX |
2330 		  IXGBE_EIMS_LSC);
2331 
2332 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2333 }
2334 
2335 enum latency_range {
2336 	lowest_latency = 0,
2337 	low_latency = 1,
2338 	bulk_latency = 2,
2339 	latency_invalid = 255
2340 };
2341 
2342 /**
2343  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2344  * @q_vector: structure containing interrupt and ring information
2345  * @ring_container: structure containing ring performance data
2346  *
2347  *      Stores a new ITR value based on packets and byte
2348  *      counts during the last interrupt.  The advantage of per interrupt
2349  *      computation is faster updates and more accurate ITR for the current
2350  *      traffic pattern.  Constants in this function were computed
2351  *      based on theoretical maximum wire speed and thresholds were set based
2352  *      on testing data as well as attempting to minimize response time
2353  *      while increasing bulk throughput.
2354  *      this functionality is controlled by the InterruptThrottleRate module
2355  *      parameter (see ixgbe_param.c)
2356  **/
2357 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2358 			     struct ixgbe_ring_container *ring_container)
2359 {
2360 	int bytes = ring_container->total_bytes;
2361 	int packets = ring_container->total_packets;
2362 	u32 timepassed_us;
2363 	u64 bytes_perint;
2364 	u8 itr_setting = ring_container->itr;
2365 
2366 	if (packets == 0)
2367 		return;
2368 
2369 	/* simple throttlerate management
2370 	 *   0-10MB/s   lowest (100000 ints/s)
2371 	 *  10-20MB/s   low    (20000 ints/s)
2372 	 *  20-1249MB/s bulk   (12000 ints/s)
2373 	 */
2374 	/* what was last interrupt timeslice? */
2375 	timepassed_us = q_vector->itr >> 2;
2376 	if (timepassed_us == 0)
2377 		return;
2378 
2379 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2380 
2381 	switch (itr_setting) {
2382 	case lowest_latency:
2383 		if (bytes_perint > 10)
2384 			itr_setting = low_latency;
2385 		break;
2386 	case low_latency:
2387 		if (bytes_perint > 20)
2388 			itr_setting = bulk_latency;
2389 		else if (bytes_perint <= 10)
2390 			itr_setting = lowest_latency;
2391 		break;
2392 	case bulk_latency:
2393 		if (bytes_perint <= 20)
2394 			itr_setting = low_latency;
2395 		break;
2396 	}
2397 
2398 	/* clear work counters since we have the values we need */
2399 	ring_container->total_bytes = 0;
2400 	ring_container->total_packets = 0;
2401 
2402 	/* write updated itr to ring container */
2403 	ring_container->itr = itr_setting;
2404 }
2405 
2406 /**
2407  * ixgbe_write_eitr - write EITR register in hardware specific way
2408  * @q_vector: structure containing interrupt and ring information
2409  *
2410  * This function is made to be called by ethtool and by the driver
2411  * when it needs to update EITR registers at runtime.  Hardware
2412  * specific quirks/differences are taken care of here.
2413  */
2414 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2415 {
2416 	struct ixgbe_adapter *adapter = q_vector->adapter;
2417 	struct ixgbe_hw *hw = &adapter->hw;
2418 	int v_idx = q_vector->v_idx;
2419 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2420 
2421 	switch (adapter->hw.mac.type) {
2422 	case ixgbe_mac_82598EB:
2423 		/* must write high and low 16 bits to reset counter */
2424 		itr_reg |= (itr_reg << 16);
2425 		break;
2426 	case ixgbe_mac_82599EB:
2427 	case ixgbe_mac_X540:
2428 	case ixgbe_mac_X550:
2429 	case ixgbe_mac_X550EM_x:
2430 	case ixgbe_mac_x550em_a:
2431 		/*
2432 		 * set the WDIS bit to not clear the timer bits and cause an
2433 		 * immediate assertion of the interrupt
2434 		 */
2435 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2436 		break;
2437 	default:
2438 		break;
2439 	}
2440 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2441 }
2442 
2443 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2444 {
2445 	u32 new_itr = q_vector->itr;
2446 	u8 current_itr;
2447 
2448 	ixgbe_update_itr(q_vector, &q_vector->tx);
2449 	ixgbe_update_itr(q_vector, &q_vector->rx);
2450 
2451 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2452 
2453 	switch (current_itr) {
2454 	/* counts and packets in update_itr are dependent on these numbers */
2455 	case lowest_latency:
2456 		new_itr = IXGBE_100K_ITR;
2457 		break;
2458 	case low_latency:
2459 		new_itr = IXGBE_20K_ITR;
2460 		break;
2461 	case bulk_latency:
2462 		new_itr = IXGBE_12K_ITR;
2463 		break;
2464 	default:
2465 		break;
2466 	}
2467 
2468 	if (new_itr != q_vector->itr) {
2469 		/* do an exponential smoothing */
2470 		new_itr = (10 * new_itr * q_vector->itr) /
2471 			  ((9 * new_itr) + q_vector->itr);
2472 
2473 		/* save the algorithm value here */
2474 		q_vector->itr = new_itr;
2475 
2476 		ixgbe_write_eitr(q_vector);
2477 	}
2478 }
2479 
2480 /**
2481  * ixgbe_check_overtemp_subtask - check for over temperature
2482  * @adapter: pointer to adapter
2483  **/
2484 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2485 {
2486 	struct ixgbe_hw *hw = &adapter->hw;
2487 	u32 eicr = adapter->interrupt_event;
2488 	s32 rc;
2489 
2490 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2491 		return;
2492 
2493 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2494 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2495 		return;
2496 
2497 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2498 
2499 	switch (hw->device_id) {
2500 	case IXGBE_DEV_ID_82599_T3_LOM:
2501 		/*
2502 		 * Since the warning interrupt is for both ports
2503 		 * we don't have to check if:
2504 		 *  - This interrupt wasn't for our port.
2505 		 *  - We may have missed the interrupt so always have to
2506 		 *    check if we  got a LSC
2507 		 */
2508 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2509 		    !(eicr & IXGBE_EICR_LSC))
2510 			return;
2511 
2512 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2513 			u32 speed;
2514 			bool link_up = false;
2515 
2516 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2517 
2518 			if (link_up)
2519 				return;
2520 		}
2521 
2522 		/* Check if this is not due to overtemp */
2523 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2524 			return;
2525 
2526 		break;
2527 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2528 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2529 		rc = hw->phy.ops.check_overtemp(hw);
2530 		if (rc != IXGBE_ERR_OVERTEMP)
2531 			return;
2532 		break;
2533 	default:
2534 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2535 			return;
2536 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2537 			return;
2538 		break;
2539 	}
2540 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2541 
2542 	adapter->interrupt_event = 0;
2543 }
2544 
2545 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2546 {
2547 	struct ixgbe_hw *hw = &adapter->hw;
2548 
2549 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2550 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2551 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2552 		/* write to clear the interrupt */
2553 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2554 	}
2555 }
2556 
2557 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2558 {
2559 	struct ixgbe_hw *hw = &adapter->hw;
2560 
2561 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2562 		return;
2563 
2564 	switch (adapter->hw.mac.type) {
2565 	case ixgbe_mac_82599EB:
2566 		/*
2567 		 * Need to check link state so complete overtemp check
2568 		 * on service task
2569 		 */
2570 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2571 		     (eicr & IXGBE_EICR_LSC)) &&
2572 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2573 			adapter->interrupt_event = eicr;
2574 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2575 			ixgbe_service_event_schedule(adapter);
2576 			return;
2577 		}
2578 		return;
2579 	case ixgbe_mac_x550em_a:
2580 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2581 			adapter->interrupt_event = eicr;
2582 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2583 			ixgbe_service_event_schedule(adapter);
2584 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2585 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2586 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2587 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2588 		}
2589 		return;
2590 	case ixgbe_mac_X550:
2591 	case ixgbe_mac_X540:
2592 		if (!(eicr & IXGBE_EICR_TS))
2593 			return;
2594 		break;
2595 	default:
2596 		return;
2597 	}
2598 
2599 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2600 }
2601 
2602 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2603 {
2604 	switch (hw->mac.type) {
2605 	case ixgbe_mac_82598EB:
2606 		if (hw->phy.type == ixgbe_phy_nl)
2607 			return true;
2608 		return false;
2609 	case ixgbe_mac_82599EB:
2610 	case ixgbe_mac_X550EM_x:
2611 	case ixgbe_mac_x550em_a:
2612 		switch (hw->mac.ops.get_media_type(hw)) {
2613 		case ixgbe_media_type_fiber:
2614 		case ixgbe_media_type_fiber_qsfp:
2615 			return true;
2616 		default:
2617 			return false;
2618 		}
2619 	default:
2620 		return false;
2621 	}
2622 }
2623 
2624 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2625 {
2626 	struct ixgbe_hw *hw = &adapter->hw;
2627 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2628 
2629 	if (!ixgbe_is_sfp(hw))
2630 		return;
2631 
2632 	/* Later MAC's use different SDP */
2633 	if (hw->mac.type >= ixgbe_mac_X540)
2634 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2635 
2636 	if (eicr & eicr_mask) {
2637 		/* Clear the interrupt */
2638 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2639 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2640 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2641 			adapter->sfp_poll_time = 0;
2642 			ixgbe_service_event_schedule(adapter);
2643 		}
2644 	}
2645 
2646 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2647 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2648 		/* Clear the interrupt */
2649 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2650 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2651 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2652 			ixgbe_service_event_schedule(adapter);
2653 		}
2654 	}
2655 }
2656 
2657 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2658 {
2659 	struct ixgbe_hw *hw = &adapter->hw;
2660 
2661 	adapter->lsc_int++;
2662 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2663 	adapter->link_check_timeout = jiffies;
2664 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2665 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2666 		IXGBE_WRITE_FLUSH(hw);
2667 		ixgbe_service_event_schedule(adapter);
2668 	}
2669 }
2670 
2671 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2672 					   u64 qmask)
2673 {
2674 	u32 mask;
2675 	struct ixgbe_hw *hw = &adapter->hw;
2676 
2677 	switch (hw->mac.type) {
2678 	case ixgbe_mac_82598EB:
2679 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2680 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2681 		break;
2682 	case ixgbe_mac_82599EB:
2683 	case ixgbe_mac_X540:
2684 	case ixgbe_mac_X550:
2685 	case ixgbe_mac_X550EM_x:
2686 	case ixgbe_mac_x550em_a:
2687 		mask = (qmask & 0xFFFFFFFF);
2688 		if (mask)
2689 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2690 		mask = (qmask >> 32);
2691 		if (mask)
2692 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2693 		break;
2694 	default:
2695 		break;
2696 	}
2697 	/* skip the flush */
2698 }
2699 
2700 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2701 					    u64 qmask)
2702 {
2703 	u32 mask;
2704 	struct ixgbe_hw *hw = &adapter->hw;
2705 
2706 	switch (hw->mac.type) {
2707 	case ixgbe_mac_82598EB:
2708 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2709 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2710 		break;
2711 	case ixgbe_mac_82599EB:
2712 	case ixgbe_mac_X540:
2713 	case ixgbe_mac_X550:
2714 	case ixgbe_mac_X550EM_x:
2715 	case ixgbe_mac_x550em_a:
2716 		mask = (qmask & 0xFFFFFFFF);
2717 		if (mask)
2718 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2719 		mask = (qmask >> 32);
2720 		if (mask)
2721 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2722 		break;
2723 	default:
2724 		break;
2725 	}
2726 	/* skip the flush */
2727 }
2728 
2729 /**
2730  * ixgbe_irq_enable - Enable default interrupt generation settings
2731  * @adapter: board private structure
2732  **/
2733 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2734 				    bool flush)
2735 {
2736 	struct ixgbe_hw *hw = &adapter->hw;
2737 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2738 
2739 	/* don't reenable LSC while waiting for link */
2740 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2741 		mask &= ~IXGBE_EIMS_LSC;
2742 
2743 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2744 		switch (adapter->hw.mac.type) {
2745 		case ixgbe_mac_82599EB:
2746 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
2747 			break;
2748 		case ixgbe_mac_X540:
2749 		case ixgbe_mac_X550:
2750 		case ixgbe_mac_X550EM_x:
2751 		case ixgbe_mac_x550em_a:
2752 			mask |= IXGBE_EIMS_TS;
2753 			break;
2754 		default:
2755 			break;
2756 		}
2757 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2758 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2759 	switch (adapter->hw.mac.type) {
2760 	case ixgbe_mac_82599EB:
2761 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2762 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
2763 		/* fall through */
2764 	case ixgbe_mac_X540:
2765 	case ixgbe_mac_X550:
2766 	case ixgbe_mac_X550EM_x:
2767 	case ixgbe_mac_x550em_a:
2768 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
2769 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
2770 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
2771 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
2772 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
2773 			mask |= IXGBE_EICR_GPI_SDP0_X540;
2774 		mask |= IXGBE_EIMS_ECC;
2775 		mask |= IXGBE_EIMS_MAILBOX;
2776 		break;
2777 	default:
2778 		break;
2779 	}
2780 
2781 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2782 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2783 		mask |= IXGBE_EIMS_FLOW_DIR;
2784 
2785 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2786 	if (queues)
2787 		ixgbe_irq_enable_queues(adapter, ~0);
2788 	if (flush)
2789 		IXGBE_WRITE_FLUSH(&adapter->hw);
2790 }
2791 
2792 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2793 {
2794 	struct ixgbe_adapter *adapter = data;
2795 	struct ixgbe_hw *hw = &adapter->hw;
2796 	u32 eicr;
2797 
2798 	/*
2799 	 * Workaround for Silicon errata.  Use clear-by-write instead
2800 	 * of clear-by-read.  Reading with EICS will return the
2801 	 * interrupt causes without clearing, which later be done
2802 	 * with the write to EICR.
2803 	 */
2804 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2805 
2806 	/* The lower 16bits of the EICR register are for the queue interrupts
2807 	 * which should be masked here in order to not accidentally clear them if
2808 	 * the bits are high when ixgbe_msix_other is called. There is a race
2809 	 * condition otherwise which results in possible performance loss
2810 	 * especially if the ixgbe_msix_other interrupt is triggering
2811 	 * consistently (as it would when PPS is turned on for the X540 device)
2812 	 */
2813 	eicr &= 0xFFFF0000;
2814 
2815 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2816 
2817 	if (eicr & IXGBE_EICR_LSC)
2818 		ixgbe_check_lsc(adapter);
2819 
2820 	if (eicr & IXGBE_EICR_MAILBOX)
2821 		ixgbe_msg_task(adapter);
2822 
2823 	switch (hw->mac.type) {
2824 	case ixgbe_mac_82599EB:
2825 	case ixgbe_mac_X540:
2826 	case ixgbe_mac_X550:
2827 	case ixgbe_mac_X550EM_x:
2828 	case ixgbe_mac_x550em_a:
2829 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
2830 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2831 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
2832 			ixgbe_service_event_schedule(adapter);
2833 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2834 					IXGBE_EICR_GPI_SDP0_X540);
2835 		}
2836 		if (eicr & IXGBE_EICR_ECC) {
2837 			e_info(link, "Received ECC Err, initiating reset\n");
2838 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
2839 			ixgbe_service_event_schedule(adapter);
2840 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2841 		}
2842 		/* Handle Flow Director Full threshold interrupt */
2843 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2844 			int reinit_count = 0;
2845 			int i;
2846 			for (i = 0; i < adapter->num_tx_queues; i++) {
2847 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2848 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2849 						       &ring->state))
2850 					reinit_count++;
2851 			}
2852 			if (reinit_count) {
2853 				/* no more flow director interrupts until after init */
2854 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2855 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2856 				ixgbe_service_event_schedule(adapter);
2857 			}
2858 		}
2859 		ixgbe_check_sfp_event(adapter, eicr);
2860 		ixgbe_check_overtemp_event(adapter, eicr);
2861 		break;
2862 	default:
2863 		break;
2864 	}
2865 
2866 	ixgbe_check_fan_failure(adapter, eicr);
2867 
2868 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2869 		ixgbe_ptp_check_pps_event(adapter);
2870 
2871 	/* re-enable the original interrupt state, no lsc, no queues */
2872 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2873 		ixgbe_irq_enable(adapter, false, false);
2874 
2875 	return IRQ_HANDLED;
2876 }
2877 
2878 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2879 {
2880 	struct ixgbe_q_vector *q_vector = data;
2881 
2882 	/* EIAM disabled interrupts (on this vector) for us */
2883 
2884 	if (q_vector->rx.ring || q_vector->tx.ring)
2885 		napi_schedule_irqoff(&q_vector->napi);
2886 
2887 	return IRQ_HANDLED;
2888 }
2889 
2890 /**
2891  * ixgbe_poll - NAPI Rx polling callback
2892  * @napi: structure for representing this polling device
2893  * @budget: how many packets driver is allowed to clean
2894  *
2895  * This function is used for legacy and MSI, NAPI mode
2896  **/
2897 int ixgbe_poll(struct napi_struct *napi, int budget)
2898 {
2899 	struct ixgbe_q_vector *q_vector =
2900 				container_of(napi, struct ixgbe_q_vector, napi);
2901 	struct ixgbe_adapter *adapter = q_vector->adapter;
2902 	struct ixgbe_ring *ring;
2903 	int per_ring_budget, work_done = 0;
2904 	bool clean_complete = true;
2905 
2906 #ifdef CONFIG_IXGBE_DCA
2907 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2908 		ixgbe_update_dca(q_vector);
2909 #endif
2910 
2911 	ixgbe_for_each_ring(ring, q_vector->tx) {
2912 		if (!ixgbe_clean_tx_irq(q_vector, ring, budget))
2913 			clean_complete = false;
2914 	}
2915 
2916 	/* Exit if we are called by netpoll */
2917 	if (budget <= 0)
2918 		return budget;
2919 
2920 	/* attempt to distribute budget to each queue fairly, but don't allow
2921 	 * the budget to go below 1 because we'll exit polling */
2922 	if (q_vector->rx.count > 1)
2923 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2924 	else
2925 		per_ring_budget = budget;
2926 
2927 	ixgbe_for_each_ring(ring, q_vector->rx) {
2928 		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
2929 						 per_ring_budget);
2930 
2931 		work_done += cleaned;
2932 		if (cleaned >= per_ring_budget)
2933 			clean_complete = false;
2934 	}
2935 
2936 	/* If all work not completed, return budget and keep polling */
2937 	if (!clean_complete)
2938 		return budget;
2939 
2940 	/* all work done, exit the polling mode */
2941 	napi_complete_done(napi, work_done);
2942 	if (adapter->rx_itr_setting & 1)
2943 		ixgbe_set_itr(q_vector);
2944 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2945 		ixgbe_irq_enable_queues(adapter, BIT_ULL(q_vector->v_idx));
2946 
2947 	return min(work_done, budget - 1);
2948 }
2949 
2950 /**
2951  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2952  * @adapter: board private structure
2953  *
2954  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2955  * interrupts from the kernel.
2956  **/
2957 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2958 {
2959 	struct net_device *netdev = adapter->netdev;
2960 	int vector, err;
2961 	int ri = 0, ti = 0;
2962 
2963 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2964 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2965 		struct msix_entry *entry = &adapter->msix_entries[vector];
2966 
2967 		if (q_vector->tx.ring && q_vector->rx.ring) {
2968 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2969 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2970 			ti++;
2971 		} else if (q_vector->rx.ring) {
2972 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2973 				 "%s-%s-%d", netdev->name, "rx", ri++);
2974 		} else if (q_vector->tx.ring) {
2975 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2976 				 "%s-%s-%d", netdev->name, "tx", ti++);
2977 		} else {
2978 			/* skip this unused q_vector */
2979 			continue;
2980 		}
2981 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2982 				  q_vector->name, q_vector);
2983 		if (err) {
2984 			e_err(probe, "request_irq failed for MSIX interrupt "
2985 			      "Error: %d\n", err);
2986 			goto free_queue_irqs;
2987 		}
2988 		/* If Flow Director is enabled, set interrupt affinity */
2989 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2990 			/* assign the mask for this irq */
2991 			irq_set_affinity_hint(entry->vector,
2992 					      &q_vector->affinity_mask);
2993 		}
2994 	}
2995 
2996 	err = request_irq(adapter->msix_entries[vector].vector,
2997 			  ixgbe_msix_other, 0, netdev->name, adapter);
2998 	if (err) {
2999 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3000 		goto free_queue_irqs;
3001 	}
3002 
3003 	return 0;
3004 
3005 free_queue_irqs:
3006 	while (vector) {
3007 		vector--;
3008 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
3009 				      NULL);
3010 		free_irq(adapter->msix_entries[vector].vector,
3011 			 adapter->q_vector[vector]);
3012 	}
3013 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3014 	pci_disable_msix(adapter->pdev);
3015 	kfree(adapter->msix_entries);
3016 	adapter->msix_entries = NULL;
3017 	return err;
3018 }
3019 
3020 /**
3021  * ixgbe_intr - legacy mode Interrupt Handler
3022  * @irq: interrupt number
3023  * @data: pointer to a network interface device structure
3024  **/
3025 static irqreturn_t ixgbe_intr(int irq, void *data)
3026 {
3027 	struct ixgbe_adapter *adapter = data;
3028 	struct ixgbe_hw *hw = &adapter->hw;
3029 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3030 	u32 eicr;
3031 
3032 	/*
3033 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3034 	 * before the read of EICR.
3035 	 */
3036 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3037 
3038 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3039 	 * therefore no explicit interrupt disable is necessary */
3040 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3041 	if (!eicr) {
3042 		/*
3043 		 * shared interrupt alert!
3044 		 * make sure interrupts are enabled because the read will
3045 		 * have disabled interrupts due to EIAM
3046 		 * finish the workaround of silicon errata on 82598.  Unmask
3047 		 * the interrupt that we masked before the EICR read.
3048 		 */
3049 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3050 			ixgbe_irq_enable(adapter, true, true);
3051 		return IRQ_NONE;	/* Not our interrupt */
3052 	}
3053 
3054 	if (eicr & IXGBE_EICR_LSC)
3055 		ixgbe_check_lsc(adapter);
3056 
3057 	switch (hw->mac.type) {
3058 	case ixgbe_mac_82599EB:
3059 		ixgbe_check_sfp_event(adapter, eicr);
3060 		/* Fall through */
3061 	case ixgbe_mac_X540:
3062 	case ixgbe_mac_X550:
3063 	case ixgbe_mac_X550EM_x:
3064 	case ixgbe_mac_x550em_a:
3065 		if (eicr & IXGBE_EICR_ECC) {
3066 			e_info(link, "Received ECC Err, initiating reset\n");
3067 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3068 			ixgbe_service_event_schedule(adapter);
3069 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3070 		}
3071 		ixgbe_check_overtemp_event(adapter, eicr);
3072 		break;
3073 	default:
3074 		break;
3075 	}
3076 
3077 	ixgbe_check_fan_failure(adapter, eicr);
3078 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3079 		ixgbe_ptp_check_pps_event(adapter);
3080 
3081 	/* would disable interrupts here but EIAM disabled it */
3082 	napi_schedule_irqoff(&q_vector->napi);
3083 
3084 	/*
3085 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3086 	 * ixgbe_poll will re-enable the queue interrupts
3087 	 */
3088 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3089 		ixgbe_irq_enable(adapter, false, false);
3090 
3091 	return IRQ_HANDLED;
3092 }
3093 
3094 /**
3095  * ixgbe_request_irq - initialize interrupts
3096  * @adapter: board private structure
3097  *
3098  * Attempts to configure interrupts using the best available
3099  * capabilities of the hardware and kernel.
3100  **/
3101 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3102 {
3103 	struct net_device *netdev = adapter->netdev;
3104 	int err;
3105 
3106 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3107 		err = ixgbe_request_msix_irqs(adapter);
3108 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3109 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3110 				  netdev->name, adapter);
3111 	else
3112 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3113 				  netdev->name, adapter);
3114 
3115 	if (err)
3116 		e_err(probe, "request_irq failed, Error %d\n", err);
3117 
3118 	return err;
3119 }
3120 
3121 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3122 {
3123 	int vector;
3124 
3125 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3126 		free_irq(adapter->pdev->irq, adapter);
3127 		return;
3128 	}
3129 
3130 	if (!adapter->msix_entries)
3131 		return;
3132 
3133 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3134 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3135 		struct msix_entry *entry = &adapter->msix_entries[vector];
3136 
3137 		/* free only the irqs that were actually requested */
3138 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3139 			continue;
3140 
3141 		/* clear the affinity_mask in the IRQ descriptor */
3142 		irq_set_affinity_hint(entry->vector, NULL);
3143 
3144 		free_irq(entry->vector, q_vector);
3145 	}
3146 
3147 	free_irq(adapter->msix_entries[vector].vector, adapter);
3148 }
3149 
3150 /**
3151  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3152  * @adapter: board private structure
3153  **/
3154 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3155 {
3156 	switch (adapter->hw.mac.type) {
3157 	case ixgbe_mac_82598EB:
3158 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3159 		break;
3160 	case ixgbe_mac_82599EB:
3161 	case ixgbe_mac_X540:
3162 	case ixgbe_mac_X550:
3163 	case ixgbe_mac_X550EM_x:
3164 	case ixgbe_mac_x550em_a:
3165 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3166 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3167 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3168 		break;
3169 	default:
3170 		break;
3171 	}
3172 	IXGBE_WRITE_FLUSH(&adapter->hw);
3173 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3174 		int vector;
3175 
3176 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3177 			synchronize_irq(adapter->msix_entries[vector].vector);
3178 
3179 		synchronize_irq(adapter->msix_entries[vector++].vector);
3180 	} else {
3181 		synchronize_irq(adapter->pdev->irq);
3182 	}
3183 }
3184 
3185 /**
3186  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3187  *
3188  **/
3189 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3190 {
3191 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3192 
3193 	ixgbe_write_eitr(q_vector);
3194 
3195 	ixgbe_set_ivar(adapter, 0, 0, 0);
3196 	ixgbe_set_ivar(adapter, 1, 0, 0);
3197 
3198 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3199 }
3200 
3201 /**
3202  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3203  * @adapter: board private structure
3204  * @ring: structure containing ring specific data
3205  *
3206  * Configure the Tx descriptor ring after a reset.
3207  **/
3208 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3209 			     struct ixgbe_ring *ring)
3210 {
3211 	struct ixgbe_hw *hw = &adapter->hw;
3212 	u64 tdba = ring->dma;
3213 	int wait_loop = 10;
3214 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3215 	u8 reg_idx = ring->reg_idx;
3216 
3217 	/* disable queue to avoid issues while updating state */
3218 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3219 	IXGBE_WRITE_FLUSH(hw);
3220 
3221 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3222 			(tdba & DMA_BIT_MASK(32)));
3223 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3224 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3225 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3226 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3227 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3228 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3229 
3230 	/*
3231 	 * set WTHRESH to encourage burst writeback, it should not be set
3232 	 * higher than 1 when:
3233 	 * - ITR is 0 as it could cause false TX hangs
3234 	 * - ITR is set to > 100k int/sec and BQL is enabled
3235 	 *
3236 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3237 	 * to or less than the number of on chip descriptors, which is
3238 	 * currently 40.
3239 	 */
3240 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3241 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3242 	else
3243 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3244 
3245 	/*
3246 	 * Setting PTHRESH to 32 both improves performance
3247 	 * and avoids a TX hang with DFP enabled
3248 	 */
3249 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3250 		   32;		/* PTHRESH = 32 */
3251 
3252 	/* reinitialize flowdirector state */
3253 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3254 		ring->atr_sample_rate = adapter->atr_sample_rate;
3255 		ring->atr_count = 0;
3256 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3257 	} else {
3258 		ring->atr_sample_rate = 0;
3259 	}
3260 
3261 	/* initialize XPS */
3262 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3263 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3264 
3265 		if (q_vector)
3266 			netif_set_xps_queue(ring->netdev,
3267 					    &q_vector->affinity_mask,
3268 					    ring->queue_index);
3269 	}
3270 
3271 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3272 
3273 	/* reinitialize tx_buffer_info */
3274 	memset(ring->tx_buffer_info, 0,
3275 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3276 
3277 	/* enable queue */
3278 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3279 
3280 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3281 	if (hw->mac.type == ixgbe_mac_82598EB &&
3282 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3283 		return;
3284 
3285 	/* poll to verify queue is enabled */
3286 	do {
3287 		usleep_range(1000, 2000);
3288 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3289 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3290 	if (!wait_loop)
3291 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3292 }
3293 
3294 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3295 {
3296 	struct ixgbe_hw *hw = &adapter->hw;
3297 	u32 rttdcs, mtqc;
3298 	u8 tcs = netdev_get_num_tc(adapter->netdev);
3299 
3300 	if (hw->mac.type == ixgbe_mac_82598EB)
3301 		return;
3302 
3303 	/* disable the arbiter while setting MTQC */
3304 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3305 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3306 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3307 
3308 	/* set transmit pool layout */
3309 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3310 		mtqc = IXGBE_MTQC_VT_ENA;
3311 		if (tcs > 4)
3312 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3313 		else if (tcs > 1)
3314 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3315 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3316 			 IXGBE_82599_VMDQ_4Q_MASK)
3317 			mtqc |= IXGBE_MTQC_32VF;
3318 		else
3319 			mtqc |= IXGBE_MTQC_64VF;
3320 	} else {
3321 		if (tcs > 4)
3322 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3323 		else if (tcs > 1)
3324 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3325 		else
3326 			mtqc = IXGBE_MTQC_64Q_1PB;
3327 	}
3328 
3329 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3330 
3331 	/* Enable Security TX Buffer IFG for multiple pb */
3332 	if (tcs) {
3333 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3334 		sectx |= IXGBE_SECTX_DCB;
3335 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3336 	}
3337 
3338 	/* re-enable the arbiter */
3339 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3340 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3341 }
3342 
3343 /**
3344  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3345  * @adapter: board private structure
3346  *
3347  * Configure the Tx unit of the MAC after a reset.
3348  **/
3349 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3350 {
3351 	struct ixgbe_hw *hw = &adapter->hw;
3352 	u32 dmatxctl;
3353 	u32 i;
3354 
3355 	ixgbe_setup_mtqc(adapter);
3356 
3357 	if (hw->mac.type != ixgbe_mac_82598EB) {
3358 		/* DMATXCTL.EN must be before Tx queues are enabled */
3359 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3360 		dmatxctl |= IXGBE_DMATXCTL_TE;
3361 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3362 	}
3363 
3364 	/* Setup the HW Tx Head and Tail descriptor pointers */
3365 	for (i = 0; i < adapter->num_tx_queues; i++)
3366 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3367 }
3368 
3369 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3370 				 struct ixgbe_ring *ring)
3371 {
3372 	struct ixgbe_hw *hw = &adapter->hw;
3373 	u8 reg_idx = ring->reg_idx;
3374 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3375 
3376 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3377 
3378 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3379 }
3380 
3381 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3382 				  struct ixgbe_ring *ring)
3383 {
3384 	struct ixgbe_hw *hw = &adapter->hw;
3385 	u8 reg_idx = ring->reg_idx;
3386 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3387 
3388 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3389 
3390 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3391 }
3392 
3393 #ifdef CONFIG_IXGBE_DCB
3394 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3395 #else
3396 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3397 #endif
3398 {
3399 	int i;
3400 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3401 
3402 	if (adapter->ixgbe_ieee_pfc)
3403 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3404 
3405 	/*
3406 	 * We should set the drop enable bit if:
3407 	 *  SR-IOV is enabled
3408 	 *   or
3409 	 *  Number of Rx queues > 1 and flow control is disabled
3410 	 *
3411 	 *  This allows us to avoid head of line blocking for security
3412 	 *  and performance reasons.
3413 	 */
3414 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3415 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3416 		for (i = 0; i < adapter->num_rx_queues; i++)
3417 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3418 	} else {
3419 		for (i = 0; i < adapter->num_rx_queues; i++)
3420 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3421 	}
3422 }
3423 
3424 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3425 
3426 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3427 				   struct ixgbe_ring *rx_ring)
3428 {
3429 	struct ixgbe_hw *hw = &adapter->hw;
3430 	u32 srrctl;
3431 	u8 reg_idx = rx_ring->reg_idx;
3432 
3433 	if (hw->mac.type == ixgbe_mac_82598EB) {
3434 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3435 
3436 		/*
3437 		 * if VMDq is not active we must program one srrctl register
3438 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3439 		 */
3440 		reg_idx &= mask;
3441 	}
3442 
3443 	/* configure header buffer length, needed for RSC */
3444 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3445 
3446 	/* configure the packet buffer length */
3447 	if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state))
3448 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3449 	else
3450 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3451 
3452 	/* configure descriptor type */
3453 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3454 
3455 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3456 }
3457 
3458 /**
3459  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3460  * @adapter: device handle
3461  *
3462  *  - 82598/82599/X540:     128
3463  *  - X550(non-SRIOV mode): 512
3464  *  - X550(SRIOV mode):     64
3465  */
3466 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3467 {
3468 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3469 		return 128;
3470 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3471 		return 64;
3472 	else
3473 		return 512;
3474 }
3475 
3476 /**
3477  * ixgbe_store_reta - Write the RETA table to HW
3478  * @adapter: device handle
3479  *
3480  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3481  */
3482 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3483 {
3484 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3485 	struct ixgbe_hw *hw = &adapter->hw;
3486 	u32 reta = 0;
3487 	u32 indices_multi;
3488 	u8 *indir_tbl = adapter->rss_indir_tbl;
3489 
3490 	/* Fill out the redirection table as follows:
3491 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3492 	 *    indices.
3493 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3494 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3495 	 */
3496 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3497 		indices_multi = 0x11;
3498 	else
3499 		indices_multi = 0x1;
3500 
3501 	/* Write redirection table to HW */
3502 	for (i = 0; i < reta_entries; i++) {
3503 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3504 		if ((i & 3) == 3) {
3505 			if (i < 128)
3506 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3507 			else
3508 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3509 						reta);
3510 			reta = 0;
3511 		}
3512 	}
3513 }
3514 
3515 /**
3516  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3517  * @adapter: device handle
3518  *
3519  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3520  */
3521 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3522 {
3523 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3524 	struct ixgbe_hw *hw = &adapter->hw;
3525 	u32 vfreta = 0;
3526 	unsigned int pf_pool = adapter->num_vfs;
3527 
3528 	/* Write redirection table to HW */
3529 	for (i = 0; i < reta_entries; i++) {
3530 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3531 		if ((i & 3) == 3) {
3532 			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3533 					vfreta);
3534 			vfreta = 0;
3535 		}
3536 	}
3537 }
3538 
3539 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3540 {
3541 	struct ixgbe_hw *hw = &adapter->hw;
3542 	u32 i, j;
3543 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3544 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3545 
3546 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3547 	 * make full use of any rings they may have.  We will use the
3548 	 * PSRTYPE register to control how many rings we use within the PF.
3549 	 */
3550 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3551 		rss_i = 4;
3552 
3553 	/* Fill out hash function seeds */
3554 	for (i = 0; i < 10; i++)
3555 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3556 
3557 	/* Fill out redirection table */
3558 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3559 
3560 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3561 		if (j == rss_i)
3562 			j = 0;
3563 
3564 		adapter->rss_indir_tbl[i] = j;
3565 	}
3566 
3567 	ixgbe_store_reta(adapter);
3568 }
3569 
3570 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3571 {
3572 	struct ixgbe_hw *hw = &adapter->hw;
3573 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3574 	unsigned int pf_pool = adapter->num_vfs;
3575 	int i, j;
3576 
3577 	/* Fill out hash function seeds */
3578 	for (i = 0; i < 10; i++)
3579 		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool),
3580 				adapter->rss_key[i]);
3581 
3582 	/* Fill out the redirection table */
3583 	for (i = 0, j = 0; i < 64; i++, j++) {
3584 		if (j == rss_i)
3585 			j = 0;
3586 
3587 		adapter->rss_indir_tbl[i] = j;
3588 	}
3589 
3590 	ixgbe_store_vfreta(adapter);
3591 }
3592 
3593 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3594 {
3595 	struct ixgbe_hw *hw = &adapter->hw;
3596 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3597 	u32 rxcsum;
3598 
3599 	/* Disable indicating checksum in descriptor, enables RSS hash */
3600 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3601 	rxcsum |= IXGBE_RXCSUM_PCSD;
3602 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3603 
3604 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3605 		if (adapter->ring_feature[RING_F_RSS].mask)
3606 			mrqc = IXGBE_MRQC_RSSEN;
3607 	} else {
3608 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3609 
3610 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3611 			if (tcs > 4)
3612 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3613 			else if (tcs > 1)
3614 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3615 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3616 				 IXGBE_82599_VMDQ_4Q_MASK)
3617 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3618 			else
3619 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3620 		} else {
3621 			if (tcs > 4)
3622 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3623 			else if (tcs > 1)
3624 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3625 			else
3626 				mrqc = IXGBE_MRQC_RSSEN;
3627 		}
3628 	}
3629 
3630 	/* Perform hash on these packet types */
3631 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3632 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3633 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3634 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3635 
3636 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3637 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3638 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3639 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3640 
3641 	netdev_rss_key_fill(adapter->rss_key, sizeof(adapter->rss_key));
3642 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3643 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3644 		unsigned int pf_pool = adapter->num_vfs;
3645 
3646 		/* Enable VF RSS mode */
3647 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3648 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3649 
3650 		/* Setup RSS through the VF registers */
3651 		ixgbe_setup_vfreta(adapter);
3652 		vfmrqc = IXGBE_MRQC_RSSEN;
3653 		vfmrqc |= rss_field;
3654 		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3655 	} else {
3656 		ixgbe_setup_reta(adapter);
3657 		mrqc |= rss_field;
3658 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3659 	}
3660 }
3661 
3662 /**
3663  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3664  * @adapter:    address of board private structure
3665  * @index:      index of ring to set
3666  **/
3667 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3668 				   struct ixgbe_ring *ring)
3669 {
3670 	struct ixgbe_hw *hw = &adapter->hw;
3671 	u32 rscctrl;
3672 	u8 reg_idx = ring->reg_idx;
3673 
3674 	if (!ring_is_rsc_enabled(ring))
3675 		return;
3676 
3677 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3678 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3679 	/*
3680 	 * we must limit the number of descriptors so that the
3681 	 * total size of max desc * buf_len is not greater
3682 	 * than 65536
3683 	 */
3684 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3685 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3686 }
3687 
3688 #define IXGBE_MAX_RX_DESC_POLL 10
3689 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3690 				       struct ixgbe_ring *ring)
3691 {
3692 	struct ixgbe_hw *hw = &adapter->hw;
3693 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3694 	u32 rxdctl;
3695 	u8 reg_idx = ring->reg_idx;
3696 
3697 	if (ixgbe_removed(hw->hw_addr))
3698 		return;
3699 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3700 	if (hw->mac.type == ixgbe_mac_82598EB &&
3701 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3702 		return;
3703 
3704 	do {
3705 		usleep_range(1000, 2000);
3706 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3707 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3708 
3709 	if (!wait_loop) {
3710 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3711 		      "the polling period\n", reg_idx);
3712 	}
3713 }
3714 
3715 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3716 			    struct ixgbe_ring *ring)
3717 {
3718 	struct ixgbe_hw *hw = &adapter->hw;
3719 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3720 	u32 rxdctl;
3721 	u8 reg_idx = ring->reg_idx;
3722 
3723 	if (ixgbe_removed(hw->hw_addr))
3724 		return;
3725 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3726 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3727 
3728 	/* write value back with RXDCTL.ENABLE bit cleared */
3729 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3730 
3731 	if (hw->mac.type == ixgbe_mac_82598EB &&
3732 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3733 		return;
3734 
3735 	/* the hardware may take up to 100us to really disable the rx queue */
3736 	do {
3737 		udelay(10);
3738 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3739 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3740 
3741 	if (!wait_loop) {
3742 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3743 		      "the polling period\n", reg_idx);
3744 	}
3745 }
3746 
3747 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3748 			     struct ixgbe_ring *ring)
3749 {
3750 	struct ixgbe_hw *hw = &adapter->hw;
3751 	union ixgbe_adv_rx_desc *rx_desc;
3752 	u64 rdba = ring->dma;
3753 	u32 rxdctl;
3754 	u8 reg_idx = ring->reg_idx;
3755 
3756 	/* disable queue to avoid issues while updating state */
3757 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3758 	ixgbe_disable_rx_queue(adapter, ring);
3759 
3760 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3761 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3762 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3763 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3764 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
3765 	IXGBE_WRITE_FLUSH(hw);
3766 
3767 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3768 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3769 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3770 
3771 	ixgbe_configure_srrctl(adapter, ring);
3772 	ixgbe_configure_rscctl(adapter, ring);
3773 
3774 	if (hw->mac.type == ixgbe_mac_82598EB) {
3775 		/*
3776 		 * enable cache line friendly hardware writes:
3777 		 * PTHRESH=32 descriptors (half the internal cache),
3778 		 * this also removes ugly rx_no_buffer_count increment
3779 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3780 		 * WTHRESH=8 burst writeback up to two cache lines
3781 		 */
3782 		rxdctl &= ~0x3FFFFF;
3783 		rxdctl |=  0x080420;
3784 #if (PAGE_SIZE < 8192)
3785 	} else {
3786 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
3787 			    IXGBE_RXDCTL_RLPML_EN);
3788 
3789 		/* Limit the maximum frame size so we don't overrun the skb */
3790 		if (ring_uses_build_skb(ring) &&
3791 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
3792 			rxdctl |= IXGBE_MAX_FRAME_BUILD_SKB |
3793 				  IXGBE_RXDCTL_RLPML_EN;
3794 #endif
3795 	}
3796 
3797 	/* initialize rx_buffer_info */
3798 	memset(ring->rx_buffer_info, 0,
3799 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
3800 
3801 	/* initialize Rx descriptor 0 */
3802 	rx_desc = IXGBE_RX_DESC(ring, 0);
3803 	rx_desc->wb.upper.length = 0;
3804 
3805 	/* enable receive descriptor ring */
3806 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3807 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3808 
3809 	ixgbe_rx_desc_queue_enable(adapter, ring);
3810 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3811 }
3812 
3813 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3814 {
3815 	struct ixgbe_hw *hw = &adapter->hw;
3816 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3817 	u16 pool;
3818 
3819 	/* PSRTYPE must be initialized in non 82598 adapters */
3820 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3821 		      IXGBE_PSRTYPE_UDPHDR |
3822 		      IXGBE_PSRTYPE_IPV4HDR |
3823 		      IXGBE_PSRTYPE_L2HDR |
3824 		      IXGBE_PSRTYPE_IPV6HDR;
3825 
3826 	if (hw->mac.type == ixgbe_mac_82598EB)
3827 		return;
3828 
3829 	if (rss_i > 3)
3830 		psrtype |= 2u << 29;
3831 	else if (rss_i > 1)
3832 		psrtype |= 1u << 29;
3833 
3834 	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3835 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3836 }
3837 
3838 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3839 {
3840 	struct ixgbe_hw *hw = &adapter->hw;
3841 	u32 reg_offset, vf_shift;
3842 	u32 gcr_ext, vmdctl;
3843 	int i;
3844 
3845 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3846 		return;
3847 
3848 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3849 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3850 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3851 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3852 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3853 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3854 
3855 	vf_shift = VMDQ_P(0) % 32;
3856 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3857 
3858 	/* Enable only the PF's pool for Tx/Rx */
3859 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
3860 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3861 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
3862 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3863 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
3864 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3865 
3866 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3867 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3868 
3869 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
3870 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
3871 
3872 	/*
3873 	 * Set up VF register offsets for selected VT Mode,
3874 	 * i.e. 32 or 64 VFs for SR-IOV
3875 	 */
3876 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3877 	case IXGBE_82599_VMDQ_8Q_MASK:
3878 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3879 		break;
3880 	case IXGBE_82599_VMDQ_4Q_MASK:
3881 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3882 		break;
3883 	default:
3884 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3885 		break;
3886 	}
3887 
3888 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3889 
3890 	for (i = 0; i < adapter->num_vfs; i++) {
3891 		/* configure spoof checking */
3892 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
3893 					  adapter->vfinfo[i].spoofchk_enabled);
3894 
3895 		/* Enable/Disable RSS query feature  */
3896 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
3897 					  adapter->vfinfo[i].rss_query_enabled);
3898 	}
3899 }
3900 
3901 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3902 {
3903 	struct ixgbe_hw *hw = &adapter->hw;
3904 	struct net_device *netdev = adapter->netdev;
3905 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3906 	struct ixgbe_ring *rx_ring;
3907 	int i;
3908 	u32 mhadd, hlreg0;
3909 
3910 #ifdef IXGBE_FCOE
3911 	/* adjust max frame to be able to do baby jumbo for FCoE */
3912 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3913 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3914 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3915 
3916 #endif /* IXGBE_FCOE */
3917 
3918 	/* adjust max frame to be at least the size of a standard frame */
3919 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3920 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3921 
3922 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3923 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3924 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3925 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3926 
3927 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3928 	}
3929 
3930 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3931 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3932 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3933 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3934 
3935 	/*
3936 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3937 	 * the Base and Length of the Rx Descriptor Ring
3938 	 */
3939 	for (i = 0; i < adapter->num_rx_queues; i++) {
3940 		rx_ring = adapter->rx_ring[i];
3941 
3942 		clear_ring_rsc_enabled(rx_ring);
3943 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
3944 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
3945 
3946 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3947 			set_ring_rsc_enabled(rx_ring);
3948 
3949 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
3950 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
3951 
3952 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
3953 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
3954 			continue;
3955 
3956 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
3957 
3958 #if (PAGE_SIZE < 8192)
3959 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3960 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
3961 
3962 		if (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN))
3963 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
3964 #endif
3965 	}
3966 }
3967 
3968 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3969 {
3970 	struct ixgbe_hw *hw = &adapter->hw;
3971 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3972 
3973 	switch (hw->mac.type) {
3974 	case ixgbe_mac_82598EB:
3975 		/*
3976 		 * For VMDq support of different descriptor types or
3977 		 * buffer sizes through the use of multiple SRRCTL
3978 		 * registers, RDRXCTL.MVMEN must be set to 1
3979 		 *
3980 		 * also, the manual doesn't mention it clearly but DCA hints
3981 		 * will only use queue 0's tags unless this bit is set.  Side
3982 		 * effects of setting this bit are only that SRRCTL must be
3983 		 * fully programmed [0..15]
3984 		 */
3985 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3986 		break;
3987 	case ixgbe_mac_X550:
3988 	case ixgbe_mac_X550EM_x:
3989 	case ixgbe_mac_x550em_a:
3990 		if (adapter->num_vfs)
3991 			rdrxctl |= IXGBE_RDRXCTL_PSP;
3992 		/* fall through for older HW */
3993 	case ixgbe_mac_82599EB:
3994 	case ixgbe_mac_X540:
3995 		/* Disable RSC for ACK packets */
3996 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3997 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3998 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3999 		/* hardware requires some bits to be set by default */
4000 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4001 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4002 		break;
4003 	default:
4004 		/* We should do nothing since we don't know this hardware */
4005 		return;
4006 	}
4007 
4008 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4009 }
4010 
4011 /**
4012  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4013  * @adapter: board private structure
4014  *
4015  * Configure the Rx unit of the MAC after a reset.
4016  **/
4017 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4018 {
4019 	struct ixgbe_hw *hw = &adapter->hw;
4020 	int i;
4021 	u32 rxctrl, rfctl;
4022 
4023 	/* disable receives while setting up the descriptors */
4024 	hw->mac.ops.disable_rx(hw);
4025 
4026 	ixgbe_setup_psrtype(adapter);
4027 	ixgbe_setup_rdrxctl(adapter);
4028 
4029 	/* RSC Setup */
4030 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4031 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4032 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4033 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4034 
4035 	/* disable NFS filtering */
4036 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4037 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4038 
4039 	/* Program registers for the distribution of queues */
4040 	ixgbe_setup_mrqc(adapter);
4041 
4042 	/* set_rx_buffer_len must be called before ring initialization */
4043 	ixgbe_set_rx_buffer_len(adapter);
4044 
4045 	/*
4046 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4047 	 * the Base and Length of the Rx Descriptor Ring
4048 	 */
4049 	for (i = 0; i < adapter->num_rx_queues; i++)
4050 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4051 
4052 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4053 	/* disable drop enable for 82598 parts */
4054 	if (hw->mac.type == ixgbe_mac_82598EB)
4055 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4056 
4057 	/* enable all receives */
4058 	rxctrl |= IXGBE_RXCTRL_RXEN;
4059 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4060 }
4061 
4062 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4063 				 __be16 proto, u16 vid)
4064 {
4065 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4066 	struct ixgbe_hw *hw = &adapter->hw;
4067 
4068 	/* add VID to filter table */
4069 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4070 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4071 
4072 	set_bit(vid, adapter->active_vlans);
4073 
4074 	return 0;
4075 }
4076 
4077 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4078 {
4079 	u32 vlvf;
4080 	int idx;
4081 
4082 	/* short cut the special case */
4083 	if (vlan == 0)
4084 		return 0;
4085 
4086 	/* Search for the vlan id in the VLVF entries */
4087 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4088 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4089 		if ((vlvf & VLAN_VID_MASK) == vlan)
4090 			break;
4091 	}
4092 
4093 	return idx;
4094 }
4095 
4096 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4097 {
4098 	struct ixgbe_hw *hw = &adapter->hw;
4099 	u32 bits, word;
4100 	int idx;
4101 
4102 	idx = ixgbe_find_vlvf_entry(hw, vid);
4103 	if (!idx)
4104 		return;
4105 
4106 	/* See if any other pools are set for this VLAN filter
4107 	 * entry other than the PF.
4108 	 */
4109 	word = idx * 2 + (VMDQ_P(0) / 32);
4110 	bits = ~BIT(VMDQ_P(0) % 32);
4111 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4112 
4113 	/* Disable the filter so this falls into the default pool. */
4114 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4115 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4116 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4117 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4118 	}
4119 }
4120 
4121 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4122 				  __be16 proto, u16 vid)
4123 {
4124 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4125 	struct ixgbe_hw *hw = &adapter->hw;
4126 
4127 	/* remove VID from filter table */
4128 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4129 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4130 
4131 	clear_bit(vid, adapter->active_vlans);
4132 
4133 	return 0;
4134 }
4135 
4136 /**
4137  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4138  * @adapter: driver data
4139  */
4140 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4141 {
4142 	struct ixgbe_hw *hw = &adapter->hw;
4143 	u32 vlnctrl;
4144 	int i, j;
4145 
4146 	switch (hw->mac.type) {
4147 	case ixgbe_mac_82598EB:
4148 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4149 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4150 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4151 		break;
4152 	case ixgbe_mac_82599EB:
4153 	case ixgbe_mac_X540:
4154 	case ixgbe_mac_X550:
4155 	case ixgbe_mac_X550EM_x:
4156 	case ixgbe_mac_x550em_a:
4157 		for (i = 0; i < adapter->num_rx_queues; i++) {
4158 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4159 
4160 			if (ring->l2_accel_priv)
4161 				continue;
4162 			j = ring->reg_idx;
4163 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4164 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4165 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4166 		}
4167 		break;
4168 	default:
4169 		break;
4170 	}
4171 }
4172 
4173 /**
4174  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4175  * @adapter: driver data
4176  */
4177 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4178 {
4179 	struct ixgbe_hw *hw = &adapter->hw;
4180 	u32 vlnctrl;
4181 	int i, j;
4182 
4183 	switch (hw->mac.type) {
4184 	case ixgbe_mac_82598EB:
4185 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4186 		vlnctrl |= IXGBE_VLNCTRL_VME;
4187 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4188 		break;
4189 	case ixgbe_mac_82599EB:
4190 	case ixgbe_mac_X540:
4191 	case ixgbe_mac_X550:
4192 	case ixgbe_mac_X550EM_x:
4193 	case ixgbe_mac_x550em_a:
4194 		for (i = 0; i < adapter->num_rx_queues; i++) {
4195 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4196 
4197 			if (ring->l2_accel_priv)
4198 				continue;
4199 			j = ring->reg_idx;
4200 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4201 			vlnctrl |= IXGBE_RXDCTL_VME;
4202 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4203 		}
4204 		break;
4205 	default:
4206 		break;
4207 	}
4208 }
4209 
4210 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4211 {
4212 	struct ixgbe_hw *hw = &adapter->hw;
4213 	u32 vlnctrl, i;
4214 
4215 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4216 
4217 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4218 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4219 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4220 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4221 	} else {
4222 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4223 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4224 		return;
4225 	}
4226 
4227 	/* Nothing to do for 82598 */
4228 	if (hw->mac.type == ixgbe_mac_82598EB)
4229 		return;
4230 
4231 	/* We are already in VLAN promisc, nothing to do */
4232 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4233 		return;
4234 
4235 	/* Set flag so we don't redo unnecessary work */
4236 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4237 
4238 	/* Add PF to all active pools */
4239 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4240 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4241 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4242 
4243 		vlvfb |= BIT(VMDQ_P(0) % 32);
4244 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4245 	}
4246 
4247 	/* Set all bits in the VLAN filter table array */
4248 	for (i = hw->mac.vft_size; i--;)
4249 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4250 }
4251 
4252 #define VFTA_BLOCK_SIZE 8
4253 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4254 {
4255 	struct ixgbe_hw *hw = &adapter->hw;
4256 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4257 	u32 vid_start = vfta_offset * 32;
4258 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4259 	u32 i, vid, word, bits;
4260 
4261 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4262 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4263 
4264 		/* pull VLAN ID from VLVF */
4265 		vid = vlvf & VLAN_VID_MASK;
4266 
4267 		/* only concern outselves with a certain range */
4268 		if (vid < vid_start || vid >= vid_end)
4269 			continue;
4270 
4271 		if (vlvf) {
4272 			/* record VLAN ID in VFTA */
4273 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4274 
4275 			/* if PF is part of this then continue */
4276 			if (test_bit(vid, adapter->active_vlans))
4277 				continue;
4278 		}
4279 
4280 		/* remove PF from the pool */
4281 		word = i * 2 + VMDQ_P(0) / 32;
4282 		bits = ~BIT(VMDQ_P(0) % 32);
4283 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4284 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4285 	}
4286 
4287 	/* extract values from active_vlans and write back to VFTA */
4288 	for (i = VFTA_BLOCK_SIZE; i--;) {
4289 		vid = (vfta_offset + i) * 32;
4290 		word = vid / BITS_PER_LONG;
4291 		bits = vid % BITS_PER_LONG;
4292 
4293 		vfta[i] |= adapter->active_vlans[word] >> bits;
4294 
4295 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4296 	}
4297 }
4298 
4299 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4300 {
4301 	struct ixgbe_hw *hw = &adapter->hw;
4302 	u32 vlnctrl, i;
4303 
4304 	/* Set VLAN filtering to enabled */
4305 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4306 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4307 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4308 
4309 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4310 	    hw->mac.type == ixgbe_mac_82598EB)
4311 		return;
4312 
4313 	/* We are not in VLAN promisc, nothing to do */
4314 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4315 		return;
4316 
4317 	/* Set flag so we don't redo unnecessary work */
4318 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4319 
4320 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4321 		ixgbe_scrub_vfta(adapter, i);
4322 }
4323 
4324 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4325 {
4326 	u16 vid = 1;
4327 
4328 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4329 
4330 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4331 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4332 }
4333 
4334 /**
4335  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4336  * @netdev: network interface device structure
4337  *
4338  * Writes multicast address list to the MTA hash table.
4339  * Returns: -ENOMEM on failure
4340  *                0 on no addresses written
4341  *                X on writing X addresses to MTA
4342  **/
4343 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4344 {
4345 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4346 	struct ixgbe_hw *hw = &adapter->hw;
4347 
4348 	if (!netif_running(netdev))
4349 		return 0;
4350 
4351 	if (hw->mac.ops.update_mc_addr_list)
4352 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4353 	else
4354 		return -ENOMEM;
4355 
4356 #ifdef CONFIG_PCI_IOV
4357 	ixgbe_restore_vf_multicasts(adapter);
4358 #endif
4359 
4360 	return netdev_mc_count(netdev);
4361 }
4362 
4363 #ifdef CONFIG_PCI_IOV
4364 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4365 {
4366 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4367 	struct ixgbe_hw *hw = &adapter->hw;
4368 	int i;
4369 
4370 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4371 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4372 
4373 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4374 			hw->mac.ops.set_rar(hw, i,
4375 					    mac_table->addr,
4376 					    mac_table->pool,
4377 					    IXGBE_RAH_AV);
4378 		else
4379 			hw->mac.ops.clear_rar(hw, i);
4380 	}
4381 }
4382 
4383 #endif
4384 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4385 {
4386 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4387 	struct ixgbe_hw *hw = &adapter->hw;
4388 	int i;
4389 
4390 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4391 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4392 			continue;
4393 
4394 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4395 
4396 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4397 			hw->mac.ops.set_rar(hw, i,
4398 					    mac_table->addr,
4399 					    mac_table->pool,
4400 					    IXGBE_RAH_AV);
4401 		else
4402 			hw->mac.ops.clear_rar(hw, i);
4403 	}
4404 }
4405 
4406 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4407 {
4408 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4409 	struct ixgbe_hw *hw = &adapter->hw;
4410 	int i;
4411 
4412 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4413 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4414 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4415 	}
4416 
4417 	ixgbe_sync_mac_table(adapter);
4418 }
4419 
4420 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4421 {
4422 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4423 	struct ixgbe_hw *hw = &adapter->hw;
4424 	int i, count = 0;
4425 
4426 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4427 		/* do not count default RAR as available */
4428 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4429 			continue;
4430 
4431 		/* only count unused and addresses that belong to us */
4432 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4433 			if (mac_table->pool != pool)
4434 				continue;
4435 		}
4436 
4437 		count++;
4438 	}
4439 
4440 	return count;
4441 }
4442 
4443 /* this function destroys the first RAR entry */
4444 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4445 {
4446 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4447 	struct ixgbe_hw *hw = &adapter->hw;
4448 
4449 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4450 	mac_table->pool = VMDQ_P(0);
4451 
4452 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4453 
4454 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4455 			    IXGBE_RAH_AV);
4456 }
4457 
4458 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4459 			 const u8 *addr, u16 pool)
4460 {
4461 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4462 	struct ixgbe_hw *hw = &adapter->hw;
4463 	int i;
4464 
4465 	if (is_zero_ether_addr(addr))
4466 		return -EINVAL;
4467 
4468 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4469 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4470 			continue;
4471 
4472 		ether_addr_copy(mac_table->addr, addr);
4473 		mac_table->pool = pool;
4474 
4475 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4476 				    IXGBE_MAC_STATE_IN_USE;
4477 
4478 		ixgbe_sync_mac_table(adapter);
4479 
4480 		return i;
4481 	}
4482 
4483 	return -ENOMEM;
4484 }
4485 
4486 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4487 			 const u8 *addr, u16 pool)
4488 {
4489 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4490 	struct ixgbe_hw *hw = &adapter->hw;
4491 	int i;
4492 
4493 	if (is_zero_ether_addr(addr))
4494 		return -EINVAL;
4495 
4496 	/* search table for addr, if found clear IN_USE flag and sync */
4497 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4498 		/* we can only delete an entry if it is in use */
4499 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4500 			continue;
4501 		/* we only care about entries that belong to the given pool */
4502 		if (mac_table->pool != pool)
4503 			continue;
4504 		/* we only care about a specific MAC address */
4505 		if (!ether_addr_equal(addr, mac_table->addr))
4506 			continue;
4507 
4508 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4509 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4510 
4511 		ixgbe_sync_mac_table(adapter);
4512 
4513 		return 0;
4514 	}
4515 
4516 	return -ENOMEM;
4517 }
4518 /**
4519  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4520  * @netdev: network interface device structure
4521  *
4522  * Writes unicast address list to the RAR table.
4523  * Returns: -ENOMEM on failure/insufficient address space
4524  *                0 on no addresses written
4525  *                X on writing X addresses to the RAR table
4526  **/
4527 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4528 {
4529 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4530 	int count = 0;
4531 
4532 	/* return ENOMEM indicating insufficient memory for addresses */
4533 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter, vfn))
4534 		return -ENOMEM;
4535 
4536 	if (!netdev_uc_empty(netdev)) {
4537 		struct netdev_hw_addr *ha;
4538 		netdev_for_each_uc_addr(ha, netdev) {
4539 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4540 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4541 			count++;
4542 		}
4543 	}
4544 	return count;
4545 }
4546 
4547 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4548 {
4549 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4550 	int ret;
4551 
4552 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4553 
4554 	return min_t(int, ret, 0);
4555 }
4556 
4557 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4558 {
4559 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4560 
4561 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4562 
4563 	return 0;
4564 }
4565 
4566 /**
4567  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4568  * @netdev: network interface device structure
4569  *
4570  * The set_rx_method entry point is called whenever the unicast/multicast
4571  * address list or the network interface flags are updated.  This routine is
4572  * responsible for configuring the hardware for proper unicast, multicast and
4573  * promiscuous mode.
4574  **/
4575 void ixgbe_set_rx_mode(struct net_device *netdev)
4576 {
4577 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4578 	struct ixgbe_hw *hw = &adapter->hw;
4579 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4580 	netdev_features_t features = netdev->features;
4581 	int count;
4582 
4583 	/* Check for Promiscuous and All Multicast modes */
4584 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4585 
4586 	/* set all bits that we expect to always be set */
4587 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4588 	fctrl |= IXGBE_FCTRL_BAM;
4589 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4590 	fctrl |= IXGBE_FCTRL_PMCF;
4591 
4592 	/* clear the bits we are changing the status of */
4593 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4594 	if (netdev->flags & IFF_PROMISC) {
4595 		hw->addr_ctrl.user_set_promisc = true;
4596 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4597 		vmolr |= IXGBE_VMOLR_MPE;
4598 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4599 	} else {
4600 		if (netdev->flags & IFF_ALLMULTI) {
4601 			fctrl |= IXGBE_FCTRL_MPE;
4602 			vmolr |= IXGBE_VMOLR_MPE;
4603 		}
4604 		hw->addr_ctrl.user_set_promisc = false;
4605 	}
4606 
4607 	/*
4608 	 * Write addresses to available RAR registers, if there is not
4609 	 * sufficient space to store all the addresses then enable
4610 	 * unicast promiscuous mode
4611 	 */
4612 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4613 		fctrl |= IXGBE_FCTRL_UPE;
4614 		vmolr |= IXGBE_VMOLR_ROPE;
4615 	}
4616 
4617 	/* Write addresses to the MTA, if the attempt fails
4618 	 * then we should just turn on promiscuous mode so
4619 	 * that we can at least receive multicast traffic
4620 	 */
4621 	count = ixgbe_write_mc_addr_list(netdev);
4622 	if (count < 0) {
4623 		fctrl |= IXGBE_FCTRL_MPE;
4624 		vmolr |= IXGBE_VMOLR_MPE;
4625 	} else if (count) {
4626 		vmolr |= IXGBE_VMOLR_ROMPE;
4627 	}
4628 
4629 	if (hw->mac.type != ixgbe_mac_82598EB) {
4630 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4631 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4632 			   IXGBE_VMOLR_ROPE);
4633 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4634 	}
4635 
4636 	/* This is useful for sniffing bad packets. */
4637 	if (features & NETIF_F_RXALL) {
4638 		/* UPE and MPE will be handled by normal PROMISC logic
4639 		 * in e1000e_set_rx_mode */
4640 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4641 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4642 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4643 
4644 		fctrl &= ~(IXGBE_FCTRL_DPF);
4645 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4646 	}
4647 
4648 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4649 
4650 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4651 		ixgbe_vlan_strip_enable(adapter);
4652 	else
4653 		ixgbe_vlan_strip_disable(adapter);
4654 
4655 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4656 		ixgbe_vlan_promisc_disable(adapter);
4657 	else
4658 		ixgbe_vlan_promisc_enable(adapter);
4659 }
4660 
4661 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4662 {
4663 	int q_idx;
4664 
4665 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4666 		napi_enable(&adapter->q_vector[q_idx]->napi);
4667 }
4668 
4669 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4670 {
4671 	int q_idx;
4672 
4673 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4674 		napi_disable(&adapter->q_vector[q_idx]->napi);
4675 }
4676 
4677 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
4678 {
4679 	struct ixgbe_hw *hw = &adapter->hw;
4680 	u32 vxlanctrl;
4681 
4682 	if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
4683 				IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
4684 		return;
4685 
4686 	vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) && ~mask;
4687 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
4688 
4689 	if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
4690 		adapter->vxlan_port = 0;
4691 
4692 	if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
4693 		adapter->geneve_port = 0;
4694 }
4695 
4696 #ifdef CONFIG_IXGBE_DCB
4697 /**
4698  * ixgbe_configure_dcb - Configure DCB hardware
4699  * @adapter: ixgbe adapter struct
4700  *
4701  * This is called by the driver on open to configure the DCB hardware.
4702  * This is also called by the gennetlink interface when reconfiguring
4703  * the DCB state.
4704  */
4705 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4706 {
4707 	struct ixgbe_hw *hw = &adapter->hw;
4708 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4709 
4710 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4711 		if (hw->mac.type == ixgbe_mac_82598EB)
4712 			netif_set_gso_max_size(adapter->netdev, 65536);
4713 		return;
4714 	}
4715 
4716 	if (hw->mac.type == ixgbe_mac_82598EB)
4717 		netif_set_gso_max_size(adapter->netdev, 32768);
4718 
4719 #ifdef IXGBE_FCOE
4720 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4721 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4722 #endif
4723 
4724 	/* reconfigure the hardware */
4725 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4726 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4727 						DCB_TX_CONFIG);
4728 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4729 						DCB_RX_CONFIG);
4730 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4731 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4732 		ixgbe_dcb_hw_ets(&adapter->hw,
4733 				 adapter->ixgbe_ieee_ets,
4734 				 max_frame);
4735 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4736 					adapter->ixgbe_ieee_pfc->pfc_en,
4737 					adapter->ixgbe_ieee_ets->prio_tc);
4738 	}
4739 
4740 	/* Enable RSS Hash per TC */
4741 	if (hw->mac.type != ixgbe_mac_82598EB) {
4742 		u32 msb = 0;
4743 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4744 
4745 		while (rss_i) {
4746 			msb++;
4747 			rss_i >>= 1;
4748 		}
4749 
4750 		/* write msb to all 8 TCs in one write */
4751 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4752 	}
4753 }
4754 #endif
4755 
4756 /* Additional bittime to account for IXGBE framing */
4757 #define IXGBE_ETH_FRAMING 20
4758 
4759 /**
4760  * ixgbe_hpbthresh - calculate high water mark for flow control
4761  *
4762  * @adapter: board private structure to calculate for
4763  * @pb: packet buffer to calculate
4764  */
4765 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4766 {
4767 	struct ixgbe_hw *hw = &adapter->hw;
4768 	struct net_device *dev = adapter->netdev;
4769 	int link, tc, kb, marker;
4770 	u32 dv_id, rx_pba;
4771 
4772 	/* Calculate max LAN frame size */
4773 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4774 
4775 #ifdef IXGBE_FCOE
4776 	/* FCoE traffic class uses FCOE jumbo frames */
4777 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4778 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4779 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4780 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4781 #endif
4782 
4783 	/* Calculate delay value for device */
4784 	switch (hw->mac.type) {
4785 	case ixgbe_mac_X540:
4786 	case ixgbe_mac_X550:
4787 	case ixgbe_mac_X550EM_x:
4788 	case ixgbe_mac_x550em_a:
4789 		dv_id = IXGBE_DV_X540(link, tc);
4790 		break;
4791 	default:
4792 		dv_id = IXGBE_DV(link, tc);
4793 		break;
4794 	}
4795 
4796 	/* Loopback switch introduces additional latency */
4797 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4798 		dv_id += IXGBE_B2BT(tc);
4799 
4800 	/* Delay value is calculated in bit times convert to KB */
4801 	kb = IXGBE_BT2KB(dv_id);
4802 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4803 
4804 	marker = rx_pba - kb;
4805 
4806 	/* It is possible that the packet buffer is not large enough
4807 	 * to provide required headroom. In this case throw an error
4808 	 * to user and a do the best we can.
4809 	 */
4810 	if (marker < 0) {
4811 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4812 			    "headroom to support flow control."
4813 			    "Decrease MTU or number of traffic classes\n", pb);
4814 		marker = tc + 1;
4815 	}
4816 
4817 	return marker;
4818 }
4819 
4820 /**
4821  * ixgbe_lpbthresh - calculate low water mark for for flow control
4822  *
4823  * @adapter: board private structure to calculate for
4824  * @pb: packet buffer to calculate
4825  */
4826 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4827 {
4828 	struct ixgbe_hw *hw = &adapter->hw;
4829 	struct net_device *dev = adapter->netdev;
4830 	int tc;
4831 	u32 dv_id;
4832 
4833 	/* Calculate max LAN frame size */
4834 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4835 
4836 #ifdef IXGBE_FCOE
4837 	/* FCoE traffic class uses FCOE jumbo frames */
4838 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4839 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4840 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4841 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4842 #endif
4843 
4844 	/* Calculate delay value for device */
4845 	switch (hw->mac.type) {
4846 	case ixgbe_mac_X540:
4847 	case ixgbe_mac_X550:
4848 	case ixgbe_mac_X550EM_x:
4849 	case ixgbe_mac_x550em_a:
4850 		dv_id = IXGBE_LOW_DV_X540(tc);
4851 		break;
4852 	default:
4853 		dv_id = IXGBE_LOW_DV(tc);
4854 		break;
4855 	}
4856 
4857 	/* Delay value is calculated in bit times convert to KB */
4858 	return IXGBE_BT2KB(dv_id);
4859 }
4860 
4861 /*
4862  * ixgbe_pbthresh_setup - calculate and setup high low water marks
4863  */
4864 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4865 {
4866 	struct ixgbe_hw *hw = &adapter->hw;
4867 	int num_tc = netdev_get_num_tc(adapter->netdev);
4868 	int i;
4869 
4870 	if (!num_tc)
4871 		num_tc = 1;
4872 
4873 	for (i = 0; i < num_tc; i++) {
4874 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4875 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4876 
4877 		/* Low water marks must not be larger than high water marks */
4878 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4879 			hw->fc.low_water[i] = 0;
4880 	}
4881 
4882 	for (; i < MAX_TRAFFIC_CLASS; i++)
4883 		hw->fc.high_water[i] = 0;
4884 }
4885 
4886 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4887 {
4888 	struct ixgbe_hw *hw = &adapter->hw;
4889 	int hdrm;
4890 	u8 tc = netdev_get_num_tc(adapter->netdev);
4891 
4892 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4893 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4894 		hdrm = 32 << adapter->fdir_pballoc;
4895 	else
4896 		hdrm = 0;
4897 
4898 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4899 	ixgbe_pbthresh_setup(adapter);
4900 }
4901 
4902 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4903 {
4904 	struct ixgbe_hw *hw = &adapter->hw;
4905 	struct hlist_node *node2;
4906 	struct ixgbe_fdir_filter *filter;
4907 
4908 	spin_lock(&adapter->fdir_perfect_lock);
4909 
4910 	if (!hlist_empty(&adapter->fdir_filter_list))
4911 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4912 
4913 	hlist_for_each_entry_safe(filter, node2,
4914 				  &adapter->fdir_filter_list, fdir_node) {
4915 		ixgbe_fdir_write_perfect_filter_82599(hw,
4916 				&filter->filter,
4917 				filter->sw_idx,
4918 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4919 				IXGBE_FDIR_DROP_QUEUE :
4920 				adapter->rx_ring[filter->action]->reg_idx);
4921 	}
4922 
4923 	spin_unlock(&adapter->fdir_perfect_lock);
4924 }
4925 
4926 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4927 				      struct ixgbe_adapter *adapter)
4928 {
4929 	struct ixgbe_hw *hw = &adapter->hw;
4930 	u32 vmolr;
4931 
4932 	/* No unicast promiscuous support for VMDQ devices. */
4933 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4934 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4935 
4936 	/* clear the affected bit */
4937 	vmolr &= ~IXGBE_VMOLR_MPE;
4938 
4939 	if (dev->flags & IFF_ALLMULTI) {
4940 		vmolr |= IXGBE_VMOLR_MPE;
4941 	} else {
4942 		vmolr |= IXGBE_VMOLR_ROMPE;
4943 		hw->mac.ops.update_mc_addr_list(hw, dev);
4944 	}
4945 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4946 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4947 }
4948 
4949 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4950 {
4951 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4952 	int rss_i = adapter->num_rx_queues_per_pool;
4953 	struct ixgbe_hw *hw = &adapter->hw;
4954 	u16 pool = vadapter->pool;
4955 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4956 		      IXGBE_PSRTYPE_UDPHDR |
4957 		      IXGBE_PSRTYPE_IPV4HDR |
4958 		      IXGBE_PSRTYPE_L2HDR |
4959 		      IXGBE_PSRTYPE_IPV6HDR;
4960 
4961 	if (hw->mac.type == ixgbe_mac_82598EB)
4962 		return;
4963 
4964 	if (rss_i > 3)
4965 		psrtype |= 2u << 29;
4966 	else if (rss_i > 1)
4967 		psrtype |= 1u << 29;
4968 
4969 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4970 }
4971 
4972 /**
4973  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4974  * @rx_ring: ring to free buffers from
4975  **/
4976 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4977 {
4978 	u16 i = rx_ring->next_to_clean;
4979 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4980 
4981 	/* Free all the Rx ring sk_buffs */
4982 	while (i != rx_ring->next_to_alloc) {
4983 		if (rx_buffer->skb) {
4984 			struct sk_buff *skb = rx_buffer->skb;
4985 			if (IXGBE_CB(skb)->page_released)
4986 				dma_unmap_page_attrs(rx_ring->dev,
4987 						     IXGBE_CB(skb)->dma,
4988 						     ixgbe_rx_pg_size(rx_ring),
4989 						     DMA_FROM_DEVICE,
4990 						     IXGBE_RX_DMA_ATTR);
4991 			dev_kfree_skb(skb);
4992 		}
4993 
4994 		/* Invalidate cache lines that may have been written to by
4995 		 * device so that we avoid corrupting memory.
4996 		 */
4997 		dma_sync_single_range_for_cpu(rx_ring->dev,
4998 					      rx_buffer->dma,
4999 					      rx_buffer->page_offset,
5000 					      ixgbe_rx_bufsz(rx_ring),
5001 					      DMA_FROM_DEVICE);
5002 
5003 		/* free resources associated with mapping */
5004 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5005 				     ixgbe_rx_pg_size(rx_ring),
5006 				     DMA_FROM_DEVICE,
5007 				     IXGBE_RX_DMA_ATTR);
5008 		__page_frag_cache_drain(rx_buffer->page,
5009 					rx_buffer->pagecnt_bias);
5010 
5011 		i++;
5012 		rx_buffer++;
5013 		if (i == rx_ring->count) {
5014 			i = 0;
5015 			rx_buffer = rx_ring->rx_buffer_info;
5016 		}
5017 	}
5018 
5019 	rx_ring->next_to_alloc = 0;
5020 	rx_ring->next_to_clean = 0;
5021 	rx_ring->next_to_use = 0;
5022 }
5023 
5024 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
5025 				   struct ixgbe_ring *rx_ring)
5026 {
5027 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
5028 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
5029 
5030 	/* shutdown specific queue receive and wait for dma to settle */
5031 	ixgbe_disable_rx_queue(adapter, rx_ring);
5032 	usleep_range(10000, 20000);
5033 	ixgbe_irq_disable_queues(adapter, BIT_ULL(index));
5034 	ixgbe_clean_rx_ring(rx_ring);
5035 	rx_ring->l2_accel_priv = NULL;
5036 }
5037 
5038 static int ixgbe_fwd_ring_down(struct net_device *vdev,
5039 			       struct ixgbe_fwd_adapter *accel)
5040 {
5041 	struct ixgbe_adapter *adapter = accel->real_adapter;
5042 	unsigned int rxbase = accel->rx_base_queue;
5043 	unsigned int txbase = accel->tx_base_queue;
5044 	int i;
5045 
5046 	netif_tx_stop_all_queues(vdev);
5047 
5048 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
5049 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
5050 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
5051 	}
5052 
5053 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
5054 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
5055 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
5056 	}
5057 
5058 
5059 	return 0;
5060 }
5061 
5062 static int ixgbe_fwd_ring_up(struct net_device *vdev,
5063 			     struct ixgbe_fwd_adapter *accel)
5064 {
5065 	struct ixgbe_adapter *adapter = accel->real_adapter;
5066 	unsigned int rxbase, txbase, queues;
5067 	int i, baseq, err = 0;
5068 
5069 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
5070 		return 0;
5071 
5072 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5073 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
5074 		   accel->pool, adapter->num_rx_pools,
5075 		   baseq, baseq + adapter->num_rx_queues_per_pool,
5076 		   adapter->fwd_bitmask);
5077 
5078 	accel->netdev = vdev;
5079 	accel->rx_base_queue = rxbase = baseq;
5080 	accel->tx_base_queue = txbase = baseq;
5081 
5082 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5083 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
5084 
5085 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
5086 		adapter->rx_ring[rxbase + i]->netdev = vdev;
5087 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
5088 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
5089 	}
5090 
5091 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
5092 		adapter->tx_ring[txbase + i]->netdev = vdev;
5093 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
5094 	}
5095 
5096 	queues = min_t(unsigned int,
5097 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
5098 	err = netif_set_real_num_tx_queues(vdev, queues);
5099 	if (err)
5100 		goto fwd_queue_err;
5101 
5102 	err = netif_set_real_num_rx_queues(vdev, queues);
5103 	if (err)
5104 		goto fwd_queue_err;
5105 
5106 	if (is_valid_ether_addr(vdev->dev_addr))
5107 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
5108 
5109 	ixgbe_fwd_psrtype(accel);
5110 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
5111 	return err;
5112 fwd_queue_err:
5113 	ixgbe_fwd_ring_down(vdev, accel);
5114 	return err;
5115 }
5116 
5117 static int ixgbe_upper_dev_walk(struct net_device *upper, void *data)
5118 {
5119 	if (netif_is_macvlan(upper)) {
5120 		struct macvlan_dev *dfwd = netdev_priv(upper);
5121 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
5122 
5123 		if (dfwd->fwd_priv)
5124 			ixgbe_fwd_ring_up(upper, vadapter);
5125 	}
5126 
5127 	return 0;
5128 }
5129 
5130 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5131 {
5132 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5133 				      ixgbe_upper_dev_walk, NULL);
5134 }
5135 
5136 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5137 {
5138 	struct ixgbe_hw *hw = &adapter->hw;
5139 
5140 	ixgbe_configure_pb(adapter);
5141 #ifdef CONFIG_IXGBE_DCB
5142 	ixgbe_configure_dcb(adapter);
5143 #endif
5144 	/*
5145 	 * We must restore virtualization before VLANs or else
5146 	 * the VLVF registers will not be populated
5147 	 */
5148 	ixgbe_configure_virtualization(adapter);
5149 
5150 	ixgbe_set_rx_mode(adapter->netdev);
5151 	ixgbe_restore_vlan(adapter);
5152 
5153 	switch (hw->mac.type) {
5154 	case ixgbe_mac_82599EB:
5155 	case ixgbe_mac_X540:
5156 		hw->mac.ops.disable_rx_buff(hw);
5157 		break;
5158 	default:
5159 		break;
5160 	}
5161 
5162 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5163 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5164 						adapter->fdir_pballoc);
5165 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5166 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5167 					      adapter->fdir_pballoc);
5168 		ixgbe_fdir_filter_restore(adapter);
5169 	}
5170 
5171 	switch (hw->mac.type) {
5172 	case ixgbe_mac_82599EB:
5173 	case ixgbe_mac_X540:
5174 		hw->mac.ops.enable_rx_buff(hw);
5175 		break;
5176 	default:
5177 		break;
5178 	}
5179 
5180 #ifdef CONFIG_IXGBE_DCA
5181 	/* configure DCA */
5182 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5183 		ixgbe_setup_dca(adapter);
5184 #endif /* CONFIG_IXGBE_DCA */
5185 
5186 #ifdef IXGBE_FCOE
5187 	/* configure FCoE L2 filters, redirection table, and Rx control */
5188 	ixgbe_configure_fcoe(adapter);
5189 
5190 #endif /* IXGBE_FCOE */
5191 	ixgbe_configure_tx(adapter);
5192 	ixgbe_configure_rx(adapter);
5193 	ixgbe_configure_dfwd(adapter);
5194 }
5195 
5196 /**
5197  * ixgbe_sfp_link_config - set up SFP+ link
5198  * @adapter: pointer to private adapter struct
5199  **/
5200 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5201 {
5202 	/*
5203 	 * We are assuming the worst case scenario here, and that
5204 	 * is that an SFP was inserted/removed after the reset
5205 	 * but before SFP detection was enabled.  As such the best
5206 	 * solution is to just start searching as soon as we start
5207 	 */
5208 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5209 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5210 
5211 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5212 	adapter->sfp_poll_time = 0;
5213 }
5214 
5215 /**
5216  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5217  * @hw: pointer to private hardware struct
5218  *
5219  * Returns 0 on success, negative on failure
5220  **/
5221 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5222 {
5223 	u32 speed;
5224 	bool autoneg, link_up = false;
5225 	int ret = IXGBE_ERR_LINK_SETUP;
5226 
5227 	if (hw->mac.ops.check_link)
5228 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5229 
5230 	if (ret)
5231 		return ret;
5232 
5233 	speed = hw->phy.autoneg_advertised;
5234 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5235 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5236 							&autoneg);
5237 	if (ret)
5238 		return ret;
5239 
5240 	if (hw->mac.ops.setup_link)
5241 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5242 
5243 	return ret;
5244 }
5245 
5246 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5247 {
5248 	struct ixgbe_hw *hw = &adapter->hw;
5249 	u32 gpie = 0;
5250 
5251 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5252 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5253 		       IXGBE_GPIE_OCD;
5254 		gpie |= IXGBE_GPIE_EIAME;
5255 		/*
5256 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5257 		 * this saves a register write for every interrupt
5258 		 */
5259 		switch (hw->mac.type) {
5260 		case ixgbe_mac_82598EB:
5261 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5262 			break;
5263 		case ixgbe_mac_82599EB:
5264 		case ixgbe_mac_X540:
5265 		case ixgbe_mac_X550:
5266 		case ixgbe_mac_X550EM_x:
5267 		case ixgbe_mac_x550em_a:
5268 		default:
5269 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5270 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5271 			break;
5272 		}
5273 	} else {
5274 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5275 		 * specifically only auto mask tx and rx interrupts */
5276 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5277 	}
5278 
5279 	/* XXX: to interrupt immediately for EICS writes, enable this */
5280 	/* gpie |= IXGBE_GPIE_EIMEN; */
5281 
5282 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5283 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5284 
5285 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5286 		case IXGBE_82599_VMDQ_8Q_MASK:
5287 			gpie |= IXGBE_GPIE_VTMODE_16;
5288 			break;
5289 		case IXGBE_82599_VMDQ_4Q_MASK:
5290 			gpie |= IXGBE_GPIE_VTMODE_32;
5291 			break;
5292 		default:
5293 			gpie |= IXGBE_GPIE_VTMODE_64;
5294 			break;
5295 		}
5296 	}
5297 
5298 	/* Enable Thermal over heat sensor interrupt */
5299 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5300 		switch (adapter->hw.mac.type) {
5301 		case ixgbe_mac_82599EB:
5302 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5303 			break;
5304 		default:
5305 			break;
5306 		}
5307 	}
5308 
5309 	/* Enable fan failure interrupt */
5310 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5311 		gpie |= IXGBE_SDP1_GPIEN(hw);
5312 
5313 	switch (hw->mac.type) {
5314 	case ixgbe_mac_82599EB:
5315 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5316 		break;
5317 	case ixgbe_mac_X550EM_x:
5318 	case ixgbe_mac_x550em_a:
5319 		gpie |= IXGBE_SDP0_GPIEN_X540;
5320 		break;
5321 	default:
5322 		break;
5323 	}
5324 
5325 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5326 }
5327 
5328 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5329 {
5330 	struct ixgbe_hw *hw = &adapter->hw;
5331 	int err;
5332 	u32 ctrl_ext;
5333 
5334 	ixgbe_get_hw_control(adapter);
5335 	ixgbe_setup_gpie(adapter);
5336 
5337 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5338 		ixgbe_configure_msix(adapter);
5339 	else
5340 		ixgbe_configure_msi_and_legacy(adapter);
5341 
5342 	/* enable the optics for 82599 SFP+ fiber */
5343 	if (hw->mac.ops.enable_tx_laser)
5344 		hw->mac.ops.enable_tx_laser(hw);
5345 
5346 	if (hw->phy.ops.set_phy_power)
5347 		hw->phy.ops.set_phy_power(hw, true);
5348 
5349 	smp_mb__before_atomic();
5350 	clear_bit(__IXGBE_DOWN, &adapter->state);
5351 	ixgbe_napi_enable_all(adapter);
5352 
5353 	if (ixgbe_is_sfp(hw)) {
5354 		ixgbe_sfp_link_config(adapter);
5355 	} else {
5356 		err = ixgbe_non_sfp_link_config(hw);
5357 		if (err)
5358 			e_err(probe, "link_config FAILED %d\n", err);
5359 	}
5360 
5361 	/* clear any pending interrupts, may auto mask */
5362 	IXGBE_READ_REG(hw, IXGBE_EICR);
5363 	ixgbe_irq_enable(adapter, true, true);
5364 
5365 	/*
5366 	 * If this adapter has a fan, check to see if we had a failure
5367 	 * before we enabled the interrupt.
5368 	 */
5369 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5370 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5371 		if (esdp & IXGBE_ESDP_SDP1)
5372 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5373 	}
5374 
5375 	/* bring the link up in the watchdog, this could race with our first
5376 	 * link up interrupt but shouldn't be a problem */
5377 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5378 	adapter->link_check_timeout = jiffies;
5379 	mod_timer(&adapter->service_timer, jiffies);
5380 
5381 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5382 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5383 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5384 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5385 }
5386 
5387 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5388 {
5389 	WARN_ON(in_interrupt());
5390 	/* put off any impending NetWatchDogTimeout */
5391 	netif_trans_update(adapter->netdev);
5392 
5393 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5394 		usleep_range(1000, 2000);
5395 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5396 		ixgbe_watchdog_link_is_down(adapter);
5397 	ixgbe_down(adapter);
5398 	/*
5399 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5400 	 * back up to give the VFs time to respond to the reset.  The
5401 	 * two second wait is based upon the watchdog timer cycle in
5402 	 * the VF driver.
5403 	 */
5404 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5405 		msleep(2000);
5406 	ixgbe_up(adapter);
5407 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5408 }
5409 
5410 void ixgbe_up(struct ixgbe_adapter *adapter)
5411 {
5412 	/* hardware has been reset, we need to reload some things */
5413 	ixgbe_configure(adapter);
5414 
5415 	ixgbe_up_complete(adapter);
5416 }
5417 
5418 void ixgbe_reset(struct ixgbe_adapter *adapter)
5419 {
5420 	struct ixgbe_hw *hw = &adapter->hw;
5421 	struct net_device *netdev = adapter->netdev;
5422 	int err;
5423 
5424 	if (ixgbe_removed(hw->hw_addr))
5425 		return;
5426 	/* lock SFP init bit to prevent race conditions with the watchdog */
5427 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5428 		usleep_range(1000, 2000);
5429 
5430 	/* clear all SFP and link config related flags while holding SFP_INIT */
5431 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5432 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5433 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5434 
5435 	err = hw->mac.ops.init_hw(hw);
5436 	switch (err) {
5437 	case 0:
5438 	case IXGBE_ERR_SFP_NOT_PRESENT:
5439 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5440 		break;
5441 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5442 		e_dev_err("master disable timed out\n");
5443 		break;
5444 	case IXGBE_ERR_EEPROM_VERSION:
5445 		/* We are running on a pre-production device, log a warning */
5446 		e_dev_warn("This device is a pre-production adapter/LOM. "
5447 			   "Please be aware there may be issues associated with "
5448 			   "your hardware.  If you are experiencing problems "
5449 			   "please contact your Intel or hardware "
5450 			   "representative who provided you with this "
5451 			   "hardware.\n");
5452 		break;
5453 	default:
5454 		e_dev_err("Hardware Error: %d\n", err);
5455 	}
5456 
5457 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5458 
5459 	/* flush entries out of MAC table */
5460 	ixgbe_flush_sw_mac_table(adapter);
5461 	__dev_uc_unsync(netdev, NULL);
5462 
5463 	/* do not flush user set addresses */
5464 	ixgbe_mac_set_default_filter(adapter);
5465 
5466 	/* update SAN MAC vmdq pool selection */
5467 	if (hw->mac.san_mac_rar_index)
5468 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5469 
5470 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5471 		ixgbe_ptp_reset(adapter);
5472 
5473 	if (hw->phy.ops.set_phy_power) {
5474 		if (!netif_running(adapter->netdev) && !adapter->wol)
5475 			hw->phy.ops.set_phy_power(hw, false);
5476 		else
5477 			hw->phy.ops.set_phy_power(hw, true);
5478 	}
5479 }
5480 
5481 /**
5482  * ixgbe_clean_tx_ring - Free Tx Buffers
5483  * @tx_ring: ring to be cleaned
5484  **/
5485 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5486 {
5487 	u16 i = tx_ring->next_to_clean;
5488 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
5489 
5490 	while (i != tx_ring->next_to_use) {
5491 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
5492 
5493 		/* Free all the Tx ring sk_buffs */
5494 		dev_kfree_skb_any(tx_buffer->skb);
5495 
5496 		/* unmap skb header data */
5497 		dma_unmap_single(tx_ring->dev,
5498 				 dma_unmap_addr(tx_buffer, dma),
5499 				 dma_unmap_len(tx_buffer, len),
5500 				 DMA_TO_DEVICE);
5501 
5502 		/* check for eop_desc to determine the end of the packet */
5503 		eop_desc = tx_buffer->next_to_watch;
5504 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
5505 
5506 		/* unmap remaining buffers */
5507 		while (tx_desc != eop_desc) {
5508 			tx_buffer++;
5509 			tx_desc++;
5510 			i++;
5511 			if (unlikely(i == tx_ring->count)) {
5512 				i = 0;
5513 				tx_buffer = tx_ring->tx_buffer_info;
5514 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
5515 			}
5516 
5517 			/* unmap any remaining paged data */
5518 			if (dma_unmap_len(tx_buffer, len))
5519 				dma_unmap_page(tx_ring->dev,
5520 					       dma_unmap_addr(tx_buffer, dma),
5521 					       dma_unmap_len(tx_buffer, len),
5522 					       DMA_TO_DEVICE);
5523 		}
5524 
5525 		/* move us one more past the eop_desc for start of next pkt */
5526 		tx_buffer++;
5527 		i++;
5528 		if (unlikely(i == tx_ring->count)) {
5529 			i = 0;
5530 			tx_buffer = tx_ring->tx_buffer_info;
5531 		}
5532 	}
5533 
5534 	/* reset BQL for queue */
5535 	netdev_tx_reset_queue(txring_txq(tx_ring));
5536 
5537 	/* reset next_to_use and next_to_clean */
5538 	tx_ring->next_to_use = 0;
5539 	tx_ring->next_to_clean = 0;
5540 }
5541 
5542 /**
5543  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5544  * @adapter: board private structure
5545  **/
5546 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5547 {
5548 	int i;
5549 
5550 	for (i = 0; i < adapter->num_rx_queues; i++)
5551 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5552 }
5553 
5554 /**
5555  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5556  * @adapter: board private structure
5557  **/
5558 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5559 {
5560 	int i;
5561 
5562 	for (i = 0; i < adapter->num_tx_queues; i++)
5563 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5564 }
5565 
5566 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5567 {
5568 	struct hlist_node *node2;
5569 	struct ixgbe_fdir_filter *filter;
5570 
5571 	spin_lock(&adapter->fdir_perfect_lock);
5572 
5573 	hlist_for_each_entry_safe(filter, node2,
5574 				  &adapter->fdir_filter_list, fdir_node) {
5575 		hlist_del(&filter->fdir_node);
5576 		kfree(filter);
5577 	}
5578 	adapter->fdir_filter_count = 0;
5579 
5580 	spin_unlock(&adapter->fdir_perfect_lock);
5581 }
5582 
5583 static int ixgbe_disable_macvlan(struct net_device *upper, void *data)
5584 {
5585 	if (netif_is_macvlan(upper)) {
5586 		struct macvlan_dev *vlan = netdev_priv(upper);
5587 
5588 		if (vlan->fwd_priv) {
5589 			netif_tx_stop_all_queues(upper);
5590 			netif_carrier_off(upper);
5591 			netif_tx_disable(upper);
5592 		}
5593 	}
5594 
5595 	return 0;
5596 }
5597 
5598 void ixgbe_down(struct ixgbe_adapter *adapter)
5599 {
5600 	struct net_device *netdev = adapter->netdev;
5601 	struct ixgbe_hw *hw = &adapter->hw;
5602 	int i;
5603 
5604 	/* signal that we are down to the interrupt handler */
5605 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5606 		return; /* do nothing if already down */
5607 
5608 	/* disable receives */
5609 	hw->mac.ops.disable_rx(hw);
5610 
5611 	/* disable all enabled rx queues */
5612 	for (i = 0; i < adapter->num_rx_queues; i++)
5613 		/* this call also flushes the previous write */
5614 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5615 
5616 	usleep_range(10000, 20000);
5617 
5618 	netif_tx_stop_all_queues(netdev);
5619 
5620 	/* call carrier off first to avoid false dev_watchdog timeouts */
5621 	netif_carrier_off(netdev);
5622 	netif_tx_disable(netdev);
5623 
5624 	/* disable any upper devices */
5625 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5626 				      ixgbe_disable_macvlan, NULL);
5627 
5628 	ixgbe_irq_disable(adapter);
5629 
5630 	ixgbe_napi_disable_all(adapter);
5631 
5632 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
5633 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5634 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5635 
5636 	del_timer_sync(&adapter->service_timer);
5637 
5638 	if (adapter->num_vfs) {
5639 		/* Clear EITR Select mapping */
5640 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5641 
5642 		/* Mark all the VFs as inactive */
5643 		for (i = 0 ; i < adapter->num_vfs; i++)
5644 			adapter->vfinfo[i].clear_to_send = false;
5645 
5646 		/* ping all the active vfs to let them know we are going down */
5647 		ixgbe_ping_all_vfs(adapter);
5648 
5649 		/* Disable all VFTE/VFRE TX/RX */
5650 		ixgbe_disable_tx_rx(adapter);
5651 	}
5652 
5653 	/* disable transmits in the hardware now that interrupts are off */
5654 	for (i = 0; i < adapter->num_tx_queues; i++) {
5655 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5656 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5657 	}
5658 
5659 	/* Disable the Tx DMA engine on 82599 and later MAC */
5660 	switch (hw->mac.type) {
5661 	case ixgbe_mac_82599EB:
5662 	case ixgbe_mac_X540:
5663 	case ixgbe_mac_X550:
5664 	case ixgbe_mac_X550EM_x:
5665 	case ixgbe_mac_x550em_a:
5666 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5667 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5668 				 ~IXGBE_DMATXCTL_TE));
5669 		break;
5670 	default:
5671 		break;
5672 	}
5673 
5674 	if (!pci_channel_offline(adapter->pdev))
5675 		ixgbe_reset(adapter);
5676 
5677 	/* power down the optics for 82599 SFP+ fiber */
5678 	if (hw->mac.ops.disable_tx_laser)
5679 		hw->mac.ops.disable_tx_laser(hw);
5680 
5681 	ixgbe_clean_all_tx_rings(adapter);
5682 	ixgbe_clean_all_rx_rings(adapter);
5683 }
5684 
5685 /**
5686  * ixgbe_eee_capable - helper function to determine EEE support on X550
5687  * @adapter: board private structure
5688  */
5689 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
5690 {
5691 	struct ixgbe_hw *hw = &adapter->hw;
5692 
5693 	switch (hw->device_id) {
5694 	case IXGBE_DEV_ID_X550EM_A_1G_T:
5695 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
5696 		if (!hw->phy.eee_speeds_supported)
5697 			break;
5698 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
5699 		if (!hw->phy.eee_speeds_advertised)
5700 			break;
5701 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
5702 		break;
5703 	default:
5704 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
5705 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
5706 		break;
5707 	}
5708 }
5709 
5710 /**
5711  * ixgbe_tx_timeout - Respond to a Tx Hang
5712  * @netdev: network interface device structure
5713  **/
5714 static void ixgbe_tx_timeout(struct net_device *netdev)
5715 {
5716 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5717 
5718 	/* Do the reset outside of interrupt context */
5719 	ixgbe_tx_timeout_reset(adapter);
5720 }
5721 
5722 #ifdef CONFIG_IXGBE_DCB
5723 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
5724 {
5725 	struct ixgbe_hw *hw = &adapter->hw;
5726 	struct tc_configuration *tc;
5727 	int j;
5728 
5729 	switch (hw->mac.type) {
5730 	case ixgbe_mac_82598EB:
5731 	case ixgbe_mac_82599EB:
5732 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5733 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5734 		break;
5735 	case ixgbe_mac_X540:
5736 	case ixgbe_mac_X550:
5737 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5738 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5739 		break;
5740 	case ixgbe_mac_X550EM_x:
5741 	case ixgbe_mac_x550em_a:
5742 	default:
5743 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
5744 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
5745 		break;
5746 	}
5747 
5748 	/* Configure DCB traffic classes */
5749 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5750 		tc = &adapter->dcb_cfg.tc_config[j];
5751 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5752 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5753 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5754 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5755 		tc->dcb_pfc = pfc_disabled;
5756 	}
5757 
5758 	/* Initialize default user to priority mapping, UPx->TC0 */
5759 	tc = &adapter->dcb_cfg.tc_config[0];
5760 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5761 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5762 
5763 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5764 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5765 	adapter->dcb_cfg.pfc_mode_enable = false;
5766 	adapter->dcb_set_bitmap = 0x00;
5767 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
5768 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5769 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5770 	       sizeof(adapter->temp_dcb_cfg));
5771 }
5772 #endif
5773 
5774 /**
5775  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5776  * @adapter: board private structure to initialize
5777  *
5778  * ixgbe_sw_init initializes the Adapter private data structure.
5779  * Fields are initialized based on PCI device information and
5780  * OS network device settings (MTU size).
5781  **/
5782 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
5783 			 const struct ixgbe_info *ii)
5784 {
5785 	struct ixgbe_hw *hw = &adapter->hw;
5786 	struct pci_dev *pdev = adapter->pdev;
5787 	unsigned int rss, fdir;
5788 	u32 fwsm;
5789 	int i;
5790 
5791 	/* PCI config space info */
5792 
5793 	hw->vendor_id = pdev->vendor;
5794 	hw->device_id = pdev->device;
5795 	hw->revision_id = pdev->revision;
5796 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5797 	hw->subsystem_device_id = pdev->subsystem_device;
5798 
5799 	/* get_invariants needs the device IDs */
5800 	ii->get_invariants(hw);
5801 
5802 	/* Set common capability flags and settings */
5803 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5804 	adapter->ring_feature[RING_F_RSS].limit = rss;
5805 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5806 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5807 	adapter->atr_sample_rate = 20;
5808 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5809 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5810 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5811 #ifdef CONFIG_IXGBE_DCA
5812 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5813 #endif
5814 #ifdef CONFIG_IXGBE_DCB
5815 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
5816 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
5817 #endif
5818 #ifdef IXGBE_FCOE
5819 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5820 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5821 #ifdef CONFIG_IXGBE_DCB
5822 	/* Default traffic class to use for FCoE */
5823 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5824 #endif /* CONFIG_IXGBE_DCB */
5825 #endif /* IXGBE_FCOE */
5826 
5827 	/* initialize static ixgbe jump table entries */
5828 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
5829 					  GFP_KERNEL);
5830 	if (!adapter->jump_tables[0])
5831 		return -ENOMEM;
5832 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
5833 
5834 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
5835 		adapter->jump_tables[i] = NULL;
5836 
5837 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5838 				     hw->mac.num_rar_entries,
5839 				     GFP_ATOMIC);
5840 	if (!adapter->mac_table)
5841 		return -ENOMEM;
5842 
5843 	/* Set MAC specific capability flags and exceptions */
5844 	switch (hw->mac.type) {
5845 	case ixgbe_mac_82598EB:
5846 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5847 
5848 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5849 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5850 
5851 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5852 		adapter->ring_feature[RING_F_FDIR].limit = 0;
5853 		adapter->atr_sample_rate = 0;
5854 		adapter->fdir_pballoc = 0;
5855 #ifdef IXGBE_FCOE
5856 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5857 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5858 #ifdef CONFIG_IXGBE_DCB
5859 		adapter->fcoe.up = 0;
5860 #endif /* IXGBE_DCB */
5861 #endif /* IXGBE_FCOE */
5862 		break;
5863 	case ixgbe_mac_82599EB:
5864 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5865 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5866 		break;
5867 	case ixgbe_mac_X540:
5868 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
5869 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5870 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5871 		break;
5872 	case ixgbe_mac_x550em_a:
5873 		adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
5874 		switch (hw->device_id) {
5875 		case IXGBE_DEV_ID_X550EM_A_1G_T:
5876 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
5877 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5878 			break;
5879 		default:
5880 			break;
5881 		}
5882 	/* fall through */
5883 	case ixgbe_mac_X550EM_x:
5884 #ifdef CONFIG_IXGBE_DCB
5885 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
5886 #endif
5887 #ifdef IXGBE_FCOE
5888 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5889 #ifdef CONFIG_IXGBE_DCB
5890 		adapter->fcoe.up = 0;
5891 #endif /* IXGBE_DCB */
5892 #endif /* IXGBE_FCOE */
5893 	/* Fall Through */
5894 	case ixgbe_mac_X550:
5895 		if (hw->mac.type == ixgbe_mac_X550)
5896 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5897 #ifdef CONFIG_IXGBE_DCA
5898 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5899 #endif
5900 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
5901 		break;
5902 	default:
5903 		break;
5904 	}
5905 
5906 #ifdef IXGBE_FCOE
5907 	/* FCoE support exists, always init the FCoE lock */
5908 	spin_lock_init(&adapter->fcoe.lock);
5909 
5910 #endif
5911 	/* n-tuple support exists, always init our spinlock */
5912 	spin_lock_init(&adapter->fdir_perfect_lock);
5913 
5914 #ifdef CONFIG_IXGBE_DCB
5915 	ixgbe_init_dcb(adapter);
5916 #endif
5917 
5918 	/* default flow control settings */
5919 	hw->fc.requested_mode = ixgbe_fc_full;
5920 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5921 	ixgbe_pbthresh_setup(adapter);
5922 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5923 	hw->fc.send_xon = true;
5924 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5925 
5926 #ifdef CONFIG_PCI_IOV
5927 	if (max_vfs > 0)
5928 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5929 
5930 	/* assign number of SR-IOV VFs */
5931 	if (hw->mac.type != ixgbe_mac_82598EB) {
5932 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5933 			adapter->num_vfs = 0;
5934 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5935 		} else {
5936 			adapter->num_vfs = max_vfs;
5937 		}
5938 	}
5939 #endif /* CONFIG_PCI_IOV */
5940 
5941 	/* enable itr by default in dynamic mode */
5942 	adapter->rx_itr_setting = 1;
5943 	adapter->tx_itr_setting = 1;
5944 
5945 	/* set default ring sizes */
5946 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5947 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5948 
5949 	/* set default work limits */
5950 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5951 
5952 	/* initialize eeprom parameters */
5953 	if (ixgbe_init_eeprom_params_generic(hw)) {
5954 		e_dev_err("EEPROM initialization failed\n");
5955 		return -EIO;
5956 	}
5957 
5958 	/* PF holds first pool slot */
5959 	set_bit(0, &adapter->fwd_bitmask);
5960 	set_bit(__IXGBE_DOWN, &adapter->state);
5961 
5962 	return 0;
5963 }
5964 
5965 /**
5966  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5967  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5968  *
5969  * Return 0 on success, negative on failure
5970  **/
5971 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5972 {
5973 	struct device *dev = tx_ring->dev;
5974 	int orig_node = dev_to_node(dev);
5975 	int ring_node = -1;
5976 	int size;
5977 
5978 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5979 
5980 	if (tx_ring->q_vector)
5981 		ring_node = tx_ring->q_vector->numa_node;
5982 
5983 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
5984 	if (!tx_ring->tx_buffer_info)
5985 		tx_ring->tx_buffer_info = vmalloc(size);
5986 	if (!tx_ring->tx_buffer_info)
5987 		goto err;
5988 
5989 	u64_stats_init(&tx_ring->syncp);
5990 
5991 	/* round up to nearest 4K */
5992 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5993 	tx_ring->size = ALIGN(tx_ring->size, 4096);
5994 
5995 	set_dev_node(dev, ring_node);
5996 	tx_ring->desc = dma_alloc_coherent(dev,
5997 					   tx_ring->size,
5998 					   &tx_ring->dma,
5999 					   GFP_KERNEL);
6000 	set_dev_node(dev, orig_node);
6001 	if (!tx_ring->desc)
6002 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6003 						   &tx_ring->dma, GFP_KERNEL);
6004 	if (!tx_ring->desc)
6005 		goto err;
6006 
6007 	tx_ring->next_to_use = 0;
6008 	tx_ring->next_to_clean = 0;
6009 	return 0;
6010 
6011 err:
6012 	vfree(tx_ring->tx_buffer_info);
6013 	tx_ring->tx_buffer_info = NULL;
6014 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6015 	return -ENOMEM;
6016 }
6017 
6018 /**
6019  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6020  * @adapter: board private structure
6021  *
6022  * If this function returns with an error, then it's possible one or
6023  * more of the rings is populated (while the rest are not).  It is the
6024  * callers duty to clean those orphaned rings.
6025  *
6026  * Return 0 on success, negative on failure
6027  **/
6028 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6029 {
6030 	int i, err = 0;
6031 
6032 	for (i = 0; i < adapter->num_tx_queues; i++) {
6033 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6034 		if (!err)
6035 			continue;
6036 
6037 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6038 		goto err_setup_tx;
6039 	}
6040 
6041 	return 0;
6042 err_setup_tx:
6043 	/* rewind the index freeing the rings as we go */
6044 	while (i--)
6045 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
6046 	return err;
6047 }
6048 
6049 /**
6050  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6051  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6052  *
6053  * Returns 0 on success, negative on failure
6054  **/
6055 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
6056 {
6057 	struct device *dev = rx_ring->dev;
6058 	int orig_node = dev_to_node(dev);
6059 	int ring_node = -1;
6060 	int size;
6061 
6062 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6063 
6064 	if (rx_ring->q_vector)
6065 		ring_node = rx_ring->q_vector->numa_node;
6066 
6067 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6068 	if (!rx_ring->rx_buffer_info)
6069 		rx_ring->rx_buffer_info = vmalloc(size);
6070 	if (!rx_ring->rx_buffer_info)
6071 		goto err;
6072 
6073 	u64_stats_init(&rx_ring->syncp);
6074 
6075 	/* Round up to nearest 4K */
6076 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6077 	rx_ring->size = ALIGN(rx_ring->size, 4096);
6078 
6079 	set_dev_node(dev, ring_node);
6080 	rx_ring->desc = dma_alloc_coherent(dev,
6081 					   rx_ring->size,
6082 					   &rx_ring->dma,
6083 					   GFP_KERNEL);
6084 	set_dev_node(dev, orig_node);
6085 	if (!rx_ring->desc)
6086 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6087 						   &rx_ring->dma, GFP_KERNEL);
6088 	if (!rx_ring->desc)
6089 		goto err;
6090 
6091 	rx_ring->next_to_clean = 0;
6092 	rx_ring->next_to_use = 0;
6093 
6094 	return 0;
6095 err:
6096 	vfree(rx_ring->rx_buffer_info);
6097 	rx_ring->rx_buffer_info = NULL;
6098 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6099 	return -ENOMEM;
6100 }
6101 
6102 /**
6103  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6104  * @adapter: board private structure
6105  *
6106  * If this function returns with an error, then it's possible one or
6107  * more of the rings is populated (while the rest are not).  It is the
6108  * callers duty to clean those orphaned rings.
6109  *
6110  * Return 0 on success, negative on failure
6111  **/
6112 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6113 {
6114 	int i, err = 0;
6115 
6116 	for (i = 0; i < adapter->num_rx_queues; i++) {
6117 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
6118 		if (!err)
6119 			continue;
6120 
6121 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6122 		goto err_setup_rx;
6123 	}
6124 
6125 #ifdef IXGBE_FCOE
6126 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
6127 	if (!err)
6128 #endif
6129 		return 0;
6130 err_setup_rx:
6131 	/* rewind the index freeing the rings as we go */
6132 	while (i--)
6133 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
6134 	return err;
6135 }
6136 
6137 /**
6138  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6139  * @tx_ring: Tx descriptor ring for a specific queue
6140  *
6141  * Free all transmit software resources
6142  **/
6143 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6144 {
6145 	ixgbe_clean_tx_ring(tx_ring);
6146 
6147 	vfree(tx_ring->tx_buffer_info);
6148 	tx_ring->tx_buffer_info = NULL;
6149 
6150 	/* if not set, then don't free */
6151 	if (!tx_ring->desc)
6152 		return;
6153 
6154 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6155 			  tx_ring->desc, tx_ring->dma);
6156 
6157 	tx_ring->desc = NULL;
6158 }
6159 
6160 /**
6161  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6162  * @adapter: board private structure
6163  *
6164  * Free all transmit software resources
6165  **/
6166 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6167 {
6168 	int i;
6169 
6170 	for (i = 0; i < adapter->num_tx_queues; i++)
6171 		if (adapter->tx_ring[i]->desc)
6172 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6173 }
6174 
6175 /**
6176  * ixgbe_free_rx_resources - Free Rx Resources
6177  * @rx_ring: ring to clean the resources from
6178  *
6179  * Free all receive software resources
6180  **/
6181 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6182 {
6183 	ixgbe_clean_rx_ring(rx_ring);
6184 
6185 	vfree(rx_ring->rx_buffer_info);
6186 	rx_ring->rx_buffer_info = NULL;
6187 
6188 	/* if not set, then don't free */
6189 	if (!rx_ring->desc)
6190 		return;
6191 
6192 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6193 			  rx_ring->desc, rx_ring->dma);
6194 
6195 	rx_ring->desc = NULL;
6196 }
6197 
6198 /**
6199  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6200  * @adapter: board private structure
6201  *
6202  * Free all receive software resources
6203  **/
6204 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6205 {
6206 	int i;
6207 
6208 #ifdef IXGBE_FCOE
6209 	ixgbe_free_fcoe_ddp_resources(adapter);
6210 
6211 #endif
6212 	for (i = 0; i < adapter->num_rx_queues; i++)
6213 		if (adapter->rx_ring[i]->desc)
6214 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6215 }
6216 
6217 /**
6218  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6219  * @netdev: network interface device structure
6220  * @new_mtu: new value for maximum frame size
6221  *
6222  * Returns 0 on success, negative on failure
6223  **/
6224 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6225 {
6226 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6227 
6228 	/*
6229 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6230 	 * paths when MTU greater than 1500 is configured.  So display a
6231 	 * warning that legacy VFs will be disabled.
6232 	 */
6233 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6234 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6235 	    (new_mtu > ETH_DATA_LEN))
6236 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6237 
6238 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6239 
6240 	/* must set new MTU before calling down or up */
6241 	netdev->mtu = new_mtu;
6242 
6243 	if (netif_running(netdev))
6244 		ixgbe_reinit_locked(adapter);
6245 
6246 	return 0;
6247 }
6248 
6249 /**
6250  * ixgbe_open - Called when a network interface is made active
6251  * @netdev: network interface device structure
6252  *
6253  * Returns 0 on success, negative value on failure
6254  *
6255  * The open entry point is called when a network interface is made
6256  * active by the system (IFF_UP).  At this point all resources needed
6257  * for transmit and receive operations are allocated, the interrupt
6258  * handler is registered with the OS, the watchdog timer is started,
6259  * and the stack is notified that the interface is ready.
6260  **/
6261 int ixgbe_open(struct net_device *netdev)
6262 {
6263 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6264 	struct ixgbe_hw *hw = &adapter->hw;
6265 	int err, queues;
6266 
6267 	/* disallow open during test */
6268 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6269 		return -EBUSY;
6270 
6271 	netif_carrier_off(netdev);
6272 
6273 	/* allocate transmit descriptors */
6274 	err = ixgbe_setup_all_tx_resources(adapter);
6275 	if (err)
6276 		goto err_setup_tx;
6277 
6278 	/* allocate receive descriptors */
6279 	err = ixgbe_setup_all_rx_resources(adapter);
6280 	if (err)
6281 		goto err_setup_rx;
6282 
6283 	ixgbe_configure(adapter);
6284 
6285 	err = ixgbe_request_irq(adapter);
6286 	if (err)
6287 		goto err_req_irq;
6288 
6289 	/* Notify the stack of the actual queue counts. */
6290 	if (adapter->num_rx_pools > 1)
6291 		queues = adapter->num_rx_queues_per_pool;
6292 	else
6293 		queues = adapter->num_tx_queues;
6294 
6295 	err = netif_set_real_num_tx_queues(netdev, queues);
6296 	if (err)
6297 		goto err_set_queues;
6298 
6299 	if (adapter->num_rx_pools > 1 &&
6300 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
6301 		queues = IXGBE_MAX_L2A_QUEUES;
6302 	else
6303 		queues = adapter->num_rx_queues;
6304 	err = netif_set_real_num_rx_queues(netdev, queues);
6305 	if (err)
6306 		goto err_set_queues;
6307 
6308 	ixgbe_ptp_init(adapter);
6309 
6310 	ixgbe_up_complete(adapter);
6311 
6312 	ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6313 	udp_tunnel_get_rx_info(netdev);
6314 
6315 	return 0;
6316 
6317 err_set_queues:
6318 	ixgbe_free_irq(adapter);
6319 err_req_irq:
6320 	ixgbe_free_all_rx_resources(adapter);
6321 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6322 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6323 err_setup_rx:
6324 	ixgbe_free_all_tx_resources(adapter);
6325 err_setup_tx:
6326 	ixgbe_reset(adapter);
6327 
6328 	return err;
6329 }
6330 
6331 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6332 {
6333 	ixgbe_ptp_suspend(adapter);
6334 
6335 	if (adapter->hw.phy.ops.enter_lplu) {
6336 		adapter->hw.phy.reset_disable = true;
6337 		ixgbe_down(adapter);
6338 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6339 		adapter->hw.phy.reset_disable = false;
6340 	} else {
6341 		ixgbe_down(adapter);
6342 	}
6343 
6344 	ixgbe_free_irq(adapter);
6345 
6346 	ixgbe_free_all_tx_resources(adapter);
6347 	ixgbe_free_all_rx_resources(adapter);
6348 }
6349 
6350 /**
6351  * ixgbe_close - Disables a network interface
6352  * @netdev: network interface device structure
6353  *
6354  * Returns 0, this is not allowed to fail
6355  *
6356  * The close entry point is called when an interface is de-activated
6357  * by the OS.  The hardware is still under the drivers control, but
6358  * needs to be disabled.  A global MAC reset is issued to stop the
6359  * hardware, and all transmit and receive resources are freed.
6360  **/
6361 int ixgbe_close(struct net_device *netdev)
6362 {
6363 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6364 
6365 	ixgbe_ptp_stop(adapter);
6366 
6367 	if (netif_device_present(netdev))
6368 		ixgbe_close_suspend(adapter);
6369 
6370 	ixgbe_fdir_filter_exit(adapter);
6371 
6372 	ixgbe_release_hw_control(adapter);
6373 
6374 	return 0;
6375 }
6376 
6377 #ifdef CONFIG_PM
6378 static int ixgbe_resume(struct pci_dev *pdev)
6379 {
6380 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6381 	struct net_device *netdev = adapter->netdev;
6382 	u32 err;
6383 
6384 	adapter->hw.hw_addr = adapter->io_addr;
6385 	pci_set_power_state(pdev, PCI_D0);
6386 	pci_restore_state(pdev);
6387 	/*
6388 	 * pci_restore_state clears dev->state_saved so call
6389 	 * pci_save_state to restore it.
6390 	 */
6391 	pci_save_state(pdev);
6392 
6393 	err = pci_enable_device_mem(pdev);
6394 	if (err) {
6395 		e_dev_err("Cannot enable PCI device from suspend\n");
6396 		return err;
6397 	}
6398 	smp_mb__before_atomic();
6399 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6400 	pci_set_master(pdev);
6401 
6402 	pci_wake_from_d3(pdev, false);
6403 
6404 	ixgbe_reset(adapter);
6405 
6406 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6407 
6408 	rtnl_lock();
6409 	err = ixgbe_init_interrupt_scheme(adapter);
6410 	if (!err && netif_running(netdev))
6411 		err = ixgbe_open(netdev);
6412 
6413 
6414 	if (!err)
6415 		netif_device_attach(netdev);
6416 	rtnl_unlock();
6417 
6418 	return err;
6419 }
6420 #endif /* CONFIG_PM */
6421 
6422 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6423 {
6424 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6425 	struct net_device *netdev = adapter->netdev;
6426 	struct ixgbe_hw *hw = &adapter->hw;
6427 	u32 ctrl, fctrl;
6428 	u32 wufc = adapter->wol;
6429 #ifdef CONFIG_PM
6430 	int retval = 0;
6431 #endif
6432 
6433 	rtnl_lock();
6434 	netif_device_detach(netdev);
6435 
6436 	if (netif_running(netdev))
6437 		ixgbe_close_suspend(adapter);
6438 
6439 	ixgbe_clear_interrupt_scheme(adapter);
6440 	rtnl_unlock();
6441 
6442 #ifdef CONFIG_PM
6443 	retval = pci_save_state(pdev);
6444 	if (retval)
6445 		return retval;
6446 
6447 #endif
6448 	if (hw->mac.ops.stop_link_on_d3)
6449 		hw->mac.ops.stop_link_on_d3(hw);
6450 
6451 	if (wufc) {
6452 		ixgbe_set_rx_mode(netdev);
6453 
6454 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6455 		if (hw->mac.ops.enable_tx_laser)
6456 			hw->mac.ops.enable_tx_laser(hw);
6457 
6458 		/* turn on all-multi mode if wake on multicast is enabled */
6459 		if (wufc & IXGBE_WUFC_MC) {
6460 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6461 			fctrl |= IXGBE_FCTRL_MPE;
6462 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6463 		}
6464 
6465 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6466 		ctrl |= IXGBE_CTRL_GIO_DIS;
6467 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6468 
6469 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6470 	} else {
6471 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6472 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6473 	}
6474 
6475 	switch (hw->mac.type) {
6476 	case ixgbe_mac_82598EB:
6477 		pci_wake_from_d3(pdev, false);
6478 		break;
6479 	case ixgbe_mac_82599EB:
6480 	case ixgbe_mac_X540:
6481 	case ixgbe_mac_X550:
6482 	case ixgbe_mac_X550EM_x:
6483 	case ixgbe_mac_x550em_a:
6484 		pci_wake_from_d3(pdev, !!wufc);
6485 		break;
6486 	default:
6487 		break;
6488 	}
6489 
6490 	*enable_wake = !!wufc;
6491 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6492 		hw->phy.ops.set_phy_power(hw, false);
6493 
6494 	ixgbe_release_hw_control(adapter);
6495 
6496 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6497 		pci_disable_device(pdev);
6498 
6499 	return 0;
6500 }
6501 
6502 #ifdef CONFIG_PM
6503 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6504 {
6505 	int retval;
6506 	bool wake;
6507 
6508 	retval = __ixgbe_shutdown(pdev, &wake);
6509 	if (retval)
6510 		return retval;
6511 
6512 	if (wake) {
6513 		pci_prepare_to_sleep(pdev);
6514 	} else {
6515 		pci_wake_from_d3(pdev, false);
6516 		pci_set_power_state(pdev, PCI_D3hot);
6517 	}
6518 
6519 	return 0;
6520 }
6521 #endif /* CONFIG_PM */
6522 
6523 static void ixgbe_shutdown(struct pci_dev *pdev)
6524 {
6525 	bool wake;
6526 
6527 	__ixgbe_shutdown(pdev, &wake);
6528 
6529 	if (system_state == SYSTEM_POWER_OFF) {
6530 		pci_wake_from_d3(pdev, wake);
6531 		pci_set_power_state(pdev, PCI_D3hot);
6532 	}
6533 }
6534 
6535 /**
6536  * ixgbe_update_stats - Update the board statistics counters.
6537  * @adapter: board private structure
6538  **/
6539 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6540 {
6541 	struct net_device *netdev = adapter->netdev;
6542 	struct ixgbe_hw *hw = &adapter->hw;
6543 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6544 	u64 total_mpc = 0;
6545 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6546 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6547 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6548 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6549 
6550 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6551 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6552 		return;
6553 
6554 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6555 		u64 rsc_count = 0;
6556 		u64 rsc_flush = 0;
6557 		for (i = 0; i < adapter->num_rx_queues; i++) {
6558 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6559 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6560 		}
6561 		adapter->rsc_total_count = rsc_count;
6562 		adapter->rsc_total_flush = rsc_flush;
6563 	}
6564 
6565 	for (i = 0; i < adapter->num_rx_queues; i++) {
6566 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6567 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6568 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6569 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6570 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6571 		bytes += rx_ring->stats.bytes;
6572 		packets += rx_ring->stats.packets;
6573 	}
6574 	adapter->non_eop_descs = non_eop_descs;
6575 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6576 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6577 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6578 	netdev->stats.rx_bytes = bytes;
6579 	netdev->stats.rx_packets = packets;
6580 
6581 	bytes = 0;
6582 	packets = 0;
6583 	/* gather some stats to the adapter struct that are per queue */
6584 	for (i = 0; i < adapter->num_tx_queues; i++) {
6585 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6586 		restart_queue += tx_ring->tx_stats.restart_queue;
6587 		tx_busy += tx_ring->tx_stats.tx_busy;
6588 		bytes += tx_ring->stats.bytes;
6589 		packets += tx_ring->stats.packets;
6590 	}
6591 	adapter->restart_queue = restart_queue;
6592 	adapter->tx_busy = tx_busy;
6593 	netdev->stats.tx_bytes = bytes;
6594 	netdev->stats.tx_packets = packets;
6595 
6596 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6597 
6598 	/* 8 register reads */
6599 	for (i = 0; i < 8; i++) {
6600 		/* for packet buffers not used, the register should read 0 */
6601 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6602 		missed_rx += mpc;
6603 		hwstats->mpc[i] += mpc;
6604 		total_mpc += hwstats->mpc[i];
6605 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6606 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6607 		switch (hw->mac.type) {
6608 		case ixgbe_mac_82598EB:
6609 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6610 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6611 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6612 			hwstats->pxonrxc[i] +=
6613 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6614 			break;
6615 		case ixgbe_mac_82599EB:
6616 		case ixgbe_mac_X540:
6617 		case ixgbe_mac_X550:
6618 		case ixgbe_mac_X550EM_x:
6619 		case ixgbe_mac_x550em_a:
6620 			hwstats->pxonrxc[i] +=
6621 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6622 			break;
6623 		default:
6624 			break;
6625 		}
6626 	}
6627 
6628 	/*16 register reads */
6629 	for (i = 0; i < 16; i++) {
6630 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6631 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6632 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6633 		    (hw->mac.type == ixgbe_mac_X540) ||
6634 		    (hw->mac.type == ixgbe_mac_X550) ||
6635 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6636 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6637 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6638 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6639 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6640 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6641 		}
6642 	}
6643 
6644 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6645 	/* work around hardware counting issue */
6646 	hwstats->gprc -= missed_rx;
6647 
6648 	ixgbe_update_xoff_received(adapter);
6649 
6650 	/* 82598 hardware only has a 32 bit counter in the high register */
6651 	switch (hw->mac.type) {
6652 	case ixgbe_mac_82598EB:
6653 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6654 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6655 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6656 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6657 		break;
6658 	case ixgbe_mac_X540:
6659 	case ixgbe_mac_X550:
6660 	case ixgbe_mac_X550EM_x:
6661 	case ixgbe_mac_x550em_a:
6662 		/* OS2BMC stats are X540 and later */
6663 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6664 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6665 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6666 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6667 	case ixgbe_mac_82599EB:
6668 		for (i = 0; i < 16; i++)
6669 			adapter->hw_rx_no_dma_resources +=
6670 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6671 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6672 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6673 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6674 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6675 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6676 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6677 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6678 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6679 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6680 #ifdef IXGBE_FCOE
6681 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6682 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6683 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6684 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6685 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6686 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6687 		/* Add up per cpu counters for total ddp aloc fail */
6688 		if (adapter->fcoe.ddp_pool) {
6689 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6690 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6691 			unsigned int cpu;
6692 			u64 noddp = 0, noddp_ext_buff = 0;
6693 			for_each_possible_cpu(cpu) {
6694 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6695 				noddp += ddp_pool->noddp;
6696 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6697 			}
6698 			hwstats->fcoe_noddp = noddp;
6699 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6700 		}
6701 #endif /* IXGBE_FCOE */
6702 		break;
6703 	default:
6704 		break;
6705 	}
6706 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6707 	hwstats->bprc += bprc;
6708 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6709 	if (hw->mac.type == ixgbe_mac_82598EB)
6710 		hwstats->mprc -= bprc;
6711 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6712 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6713 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6714 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6715 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6716 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6717 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6718 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6719 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6720 	hwstats->lxontxc += lxon;
6721 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6722 	hwstats->lxofftxc += lxoff;
6723 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6724 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6725 	/*
6726 	 * 82598 errata - tx of flow control packets is included in tx counters
6727 	 */
6728 	xon_off_tot = lxon + lxoff;
6729 	hwstats->gptc -= xon_off_tot;
6730 	hwstats->mptc -= xon_off_tot;
6731 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6732 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6733 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6734 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6735 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6736 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6737 	hwstats->ptc64 -= xon_off_tot;
6738 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6739 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6740 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6741 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6742 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6743 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6744 
6745 	/* Fill out the OS statistics structure */
6746 	netdev->stats.multicast = hwstats->mprc;
6747 
6748 	/* Rx Errors */
6749 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6750 	netdev->stats.rx_dropped = 0;
6751 	netdev->stats.rx_length_errors = hwstats->rlec;
6752 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6753 	netdev->stats.rx_missed_errors = total_mpc;
6754 }
6755 
6756 /**
6757  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6758  * @adapter: pointer to the device adapter structure
6759  **/
6760 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6761 {
6762 	struct ixgbe_hw *hw = &adapter->hw;
6763 	int i;
6764 
6765 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6766 		return;
6767 
6768 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6769 
6770 	/* if interface is down do nothing */
6771 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6772 		return;
6773 
6774 	/* do nothing if we are not using signature filters */
6775 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6776 		return;
6777 
6778 	adapter->fdir_overflow++;
6779 
6780 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6781 		for (i = 0; i < adapter->num_tx_queues; i++)
6782 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6783 				&(adapter->tx_ring[i]->state));
6784 		/* re-enable flow director interrupts */
6785 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6786 	} else {
6787 		e_err(probe, "failed to finish FDIR re-initialization, "
6788 		      "ignored adding FDIR ATR filters\n");
6789 	}
6790 }
6791 
6792 /**
6793  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6794  * @adapter: pointer to the device adapter structure
6795  *
6796  * This function serves two purposes.  First it strobes the interrupt lines
6797  * in order to make certain interrupts are occurring.  Secondly it sets the
6798  * bits needed to check for TX hangs.  As a result we should immediately
6799  * determine if a hang has occurred.
6800  */
6801 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6802 {
6803 	struct ixgbe_hw *hw = &adapter->hw;
6804 	u64 eics = 0;
6805 	int i;
6806 
6807 	/* If we're down, removing or resetting, just bail */
6808 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6809 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6810 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6811 		return;
6812 
6813 	/* Force detection of hung controller */
6814 	if (netif_carrier_ok(adapter->netdev)) {
6815 		for (i = 0; i < adapter->num_tx_queues; i++)
6816 			set_check_for_tx_hang(adapter->tx_ring[i]);
6817 	}
6818 
6819 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6820 		/*
6821 		 * for legacy and MSI interrupts don't set any bits
6822 		 * that are enabled for EIAM, because this operation
6823 		 * would set *both* EIMS and EICS for any bit in EIAM
6824 		 */
6825 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6826 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6827 	} else {
6828 		/* get one bit for every active tx/rx interrupt vector */
6829 		for (i = 0; i < adapter->num_q_vectors; i++) {
6830 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6831 			if (qv->rx.ring || qv->tx.ring)
6832 				eics |= BIT_ULL(i);
6833 		}
6834 	}
6835 
6836 	/* Cause software interrupt to ensure rings are cleaned */
6837 	ixgbe_irq_rearm_queues(adapter, eics);
6838 }
6839 
6840 /**
6841  * ixgbe_watchdog_update_link - update the link status
6842  * @adapter: pointer to the device adapter structure
6843  * @link_speed: pointer to a u32 to store the link_speed
6844  **/
6845 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6846 {
6847 	struct ixgbe_hw *hw = &adapter->hw;
6848 	u32 link_speed = adapter->link_speed;
6849 	bool link_up = adapter->link_up;
6850 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6851 
6852 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6853 		return;
6854 
6855 	if (hw->mac.ops.check_link) {
6856 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6857 	} else {
6858 		/* always assume link is up, if no check link function */
6859 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6860 		link_up = true;
6861 	}
6862 
6863 	if (adapter->ixgbe_ieee_pfc)
6864 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6865 
6866 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6867 		hw->mac.ops.fc_enable(hw);
6868 		ixgbe_set_rx_drop_en(adapter);
6869 	}
6870 
6871 	if (link_up ||
6872 	    time_after(jiffies, (adapter->link_check_timeout +
6873 				 IXGBE_TRY_LINK_TIMEOUT))) {
6874 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6875 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6876 		IXGBE_WRITE_FLUSH(hw);
6877 	}
6878 
6879 	adapter->link_up = link_up;
6880 	adapter->link_speed = link_speed;
6881 }
6882 
6883 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6884 {
6885 #ifdef CONFIG_IXGBE_DCB
6886 	struct net_device *netdev = adapter->netdev;
6887 	struct dcb_app app = {
6888 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6889 			      .protocol = 0,
6890 			     };
6891 	u8 up = 0;
6892 
6893 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6894 		up = dcb_ieee_getapp_mask(netdev, &app);
6895 
6896 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6897 #endif
6898 }
6899 
6900 static int ixgbe_enable_macvlan(struct net_device *upper, void *data)
6901 {
6902 	if (netif_is_macvlan(upper)) {
6903 		struct macvlan_dev *vlan = netdev_priv(upper);
6904 
6905 		if (vlan->fwd_priv)
6906 			netif_tx_wake_all_queues(upper);
6907 	}
6908 
6909 	return 0;
6910 }
6911 
6912 /**
6913  * ixgbe_watchdog_link_is_up - update netif_carrier status and
6914  *                             print link up message
6915  * @adapter: pointer to the device adapter structure
6916  **/
6917 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6918 {
6919 	struct net_device *netdev = adapter->netdev;
6920 	struct ixgbe_hw *hw = &adapter->hw;
6921 	u32 link_speed = adapter->link_speed;
6922 	const char *speed_str;
6923 	bool flow_rx, flow_tx;
6924 
6925 	/* only continue if link was previously down */
6926 	if (netif_carrier_ok(netdev))
6927 		return;
6928 
6929 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6930 
6931 	switch (hw->mac.type) {
6932 	case ixgbe_mac_82598EB: {
6933 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6934 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6935 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6936 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6937 	}
6938 		break;
6939 	case ixgbe_mac_X540:
6940 	case ixgbe_mac_X550:
6941 	case ixgbe_mac_X550EM_x:
6942 	case ixgbe_mac_x550em_a:
6943 	case ixgbe_mac_82599EB: {
6944 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6945 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6946 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6947 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6948 	}
6949 		break;
6950 	default:
6951 		flow_tx = false;
6952 		flow_rx = false;
6953 		break;
6954 	}
6955 
6956 	adapter->last_rx_ptp_check = jiffies;
6957 
6958 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6959 		ixgbe_ptp_start_cyclecounter(adapter);
6960 
6961 	switch (link_speed) {
6962 	case IXGBE_LINK_SPEED_10GB_FULL:
6963 		speed_str = "10 Gbps";
6964 		break;
6965 	case IXGBE_LINK_SPEED_2_5GB_FULL:
6966 		speed_str = "2.5 Gbps";
6967 		break;
6968 	case IXGBE_LINK_SPEED_1GB_FULL:
6969 		speed_str = "1 Gbps";
6970 		break;
6971 	case IXGBE_LINK_SPEED_100_FULL:
6972 		speed_str = "100 Mbps";
6973 		break;
6974 	case IXGBE_LINK_SPEED_10_FULL:
6975 		speed_str = "10 Mbps";
6976 		break;
6977 	default:
6978 		speed_str = "unknown speed";
6979 		break;
6980 	}
6981 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
6982 	       ((flow_rx && flow_tx) ? "RX/TX" :
6983 	       (flow_rx ? "RX" :
6984 	       (flow_tx ? "TX" : "None"))));
6985 
6986 	netif_carrier_on(netdev);
6987 	ixgbe_check_vf_rate_limit(adapter);
6988 
6989 	/* enable transmits */
6990 	netif_tx_wake_all_queues(adapter->netdev);
6991 
6992 	/* enable any upper devices */
6993 	rtnl_lock();
6994 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
6995 				      ixgbe_enable_macvlan, NULL);
6996 	rtnl_unlock();
6997 
6998 	/* update the default user priority for VFs */
6999 	ixgbe_update_default_up(adapter);
7000 
7001 	/* ping all the active vfs to let them know link has changed */
7002 	ixgbe_ping_all_vfs(adapter);
7003 }
7004 
7005 /**
7006  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7007  *                               print link down message
7008  * @adapter: pointer to the adapter structure
7009  **/
7010 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7011 {
7012 	struct net_device *netdev = adapter->netdev;
7013 	struct ixgbe_hw *hw = &adapter->hw;
7014 
7015 	adapter->link_up = false;
7016 	adapter->link_speed = 0;
7017 
7018 	/* only continue if link was up previously */
7019 	if (!netif_carrier_ok(netdev))
7020 		return;
7021 
7022 	/* poll for SFP+ cable when link is down */
7023 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7024 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7025 
7026 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7027 		ixgbe_ptp_start_cyclecounter(adapter);
7028 
7029 	e_info(drv, "NIC Link is Down\n");
7030 	netif_carrier_off(netdev);
7031 
7032 	/* ping all the active vfs to let them know link has changed */
7033 	ixgbe_ping_all_vfs(adapter);
7034 }
7035 
7036 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7037 {
7038 	int i;
7039 
7040 	for (i = 0; i < adapter->num_tx_queues; i++) {
7041 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7042 
7043 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
7044 			return true;
7045 	}
7046 
7047 	return false;
7048 }
7049 
7050 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7051 {
7052 	struct ixgbe_hw *hw = &adapter->hw;
7053 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7054 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7055 
7056 	int i, j;
7057 
7058 	if (!adapter->num_vfs)
7059 		return false;
7060 
7061 	/* resetting the PF is only needed for MAC before X550 */
7062 	if (hw->mac.type >= ixgbe_mac_X550)
7063 		return false;
7064 
7065 	for (i = 0; i < adapter->num_vfs; i++) {
7066 		for (j = 0; j < q_per_pool; j++) {
7067 			u32 h, t;
7068 
7069 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7070 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7071 
7072 			if (h != t)
7073 				return true;
7074 		}
7075 	}
7076 
7077 	return false;
7078 }
7079 
7080 /**
7081  * ixgbe_watchdog_flush_tx - flush queues on link down
7082  * @adapter: pointer to the device adapter structure
7083  **/
7084 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7085 {
7086 	if (!netif_carrier_ok(adapter->netdev)) {
7087 		if (ixgbe_ring_tx_pending(adapter) ||
7088 		    ixgbe_vf_tx_pending(adapter)) {
7089 			/* We've lost link, so the controller stops DMA,
7090 			 * but we've got queued Tx work that's never going
7091 			 * to get done, so reset controller to flush Tx.
7092 			 * (Do the reset outside of interrupt context).
7093 			 */
7094 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7095 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7096 		}
7097 	}
7098 }
7099 
7100 #ifdef CONFIG_PCI_IOV
7101 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
7102 				      struct pci_dev *vfdev)
7103 {
7104 	if (!pci_wait_for_pending_transaction(vfdev))
7105 		e_dev_warn("Issuing VFLR with pending transactions\n");
7106 
7107 	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
7108 	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
7109 
7110 	msleep(100);
7111 }
7112 
7113 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7114 {
7115 	struct ixgbe_hw *hw = &adapter->hw;
7116 	struct pci_dev *pdev = adapter->pdev;
7117 	unsigned int vf;
7118 	u32 gpc;
7119 
7120 	if (!(netif_carrier_ok(adapter->netdev)))
7121 		return;
7122 
7123 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7124 	if (gpc) /* If incrementing then no need for the check below */
7125 		return;
7126 	/* Check to see if a bad DMA write target from an errant or
7127 	 * malicious VF has caused a PCIe error.  If so then we can
7128 	 * issue a VFLR to the offending VF(s) and then resume without
7129 	 * requesting a full slot reset.
7130 	 */
7131 
7132 	if (!pdev)
7133 		return;
7134 
7135 	/* check status reg for all VFs owned by this PF */
7136 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
7137 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7138 		u16 status_reg;
7139 
7140 		if (!vfdev)
7141 			continue;
7142 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7143 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7144 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
7145 			ixgbe_issue_vf_flr(adapter, vfdev);
7146 	}
7147 }
7148 
7149 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7150 {
7151 	u32 ssvpc;
7152 
7153 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7154 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7155 	    adapter->num_vfs == 0)
7156 		return;
7157 
7158 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7159 
7160 	/*
7161 	 * ssvpc register is cleared on read, if zero then no
7162 	 * spoofed packets in the last interval.
7163 	 */
7164 	if (!ssvpc)
7165 		return;
7166 
7167 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7168 }
7169 #else
7170 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7171 {
7172 }
7173 
7174 static void
7175 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7176 {
7177 }
7178 #endif /* CONFIG_PCI_IOV */
7179 
7180 
7181 /**
7182  * ixgbe_watchdog_subtask - check and bring link up
7183  * @adapter: pointer to the device adapter structure
7184  **/
7185 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7186 {
7187 	/* if interface is down, removing or resetting, do nothing */
7188 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7189 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7190 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7191 		return;
7192 
7193 	ixgbe_watchdog_update_link(adapter);
7194 
7195 	if (adapter->link_up)
7196 		ixgbe_watchdog_link_is_up(adapter);
7197 	else
7198 		ixgbe_watchdog_link_is_down(adapter);
7199 
7200 	ixgbe_check_for_bad_vf(adapter);
7201 	ixgbe_spoof_check(adapter);
7202 	ixgbe_update_stats(adapter);
7203 
7204 	ixgbe_watchdog_flush_tx(adapter);
7205 }
7206 
7207 /**
7208  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7209  * @adapter: the ixgbe adapter structure
7210  **/
7211 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7212 {
7213 	struct ixgbe_hw *hw = &adapter->hw;
7214 	s32 err;
7215 
7216 	/* not searching for SFP so there is nothing to do here */
7217 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7218 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7219 		return;
7220 
7221 	if (adapter->sfp_poll_time &&
7222 	    time_after(adapter->sfp_poll_time, jiffies))
7223 		return; /* If not yet time to poll for SFP */
7224 
7225 	/* someone else is in init, wait until next service event */
7226 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7227 		return;
7228 
7229 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7230 
7231 	err = hw->phy.ops.identify_sfp(hw);
7232 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7233 		goto sfp_out;
7234 
7235 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7236 		/* If no cable is present, then we need to reset
7237 		 * the next time we find a good cable. */
7238 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7239 	}
7240 
7241 	/* exit on error */
7242 	if (err)
7243 		goto sfp_out;
7244 
7245 	/* exit if reset not needed */
7246 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7247 		goto sfp_out;
7248 
7249 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7250 
7251 	/*
7252 	 * A module may be identified correctly, but the EEPROM may not have
7253 	 * support for that module.  setup_sfp() will fail in that case, so
7254 	 * we should not allow that module to load.
7255 	 */
7256 	if (hw->mac.type == ixgbe_mac_82598EB)
7257 		err = hw->phy.ops.reset(hw);
7258 	else
7259 		err = hw->mac.ops.setup_sfp(hw);
7260 
7261 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7262 		goto sfp_out;
7263 
7264 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7265 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7266 
7267 sfp_out:
7268 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7269 
7270 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7271 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7272 		e_dev_err("failed to initialize because an unsupported "
7273 			  "SFP+ module type was detected.\n");
7274 		e_dev_err("Reload the driver after installing a "
7275 			  "supported module.\n");
7276 		unregister_netdev(adapter->netdev);
7277 	}
7278 }
7279 
7280 /**
7281  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7282  * @adapter: the ixgbe adapter structure
7283  **/
7284 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7285 {
7286 	struct ixgbe_hw *hw = &adapter->hw;
7287 	u32 speed;
7288 	bool autoneg = false;
7289 
7290 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7291 		return;
7292 
7293 	/* someone else is in init, wait until next service event */
7294 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7295 		return;
7296 
7297 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7298 
7299 	speed = hw->phy.autoneg_advertised;
7300 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
7301 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
7302 
7303 		/* setup the highest link when no autoneg */
7304 		if (!autoneg) {
7305 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
7306 				speed = IXGBE_LINK_SPEED_10GB_FULL;
7307 		}
7308 	}
7309 
7310 	if (hw->mac.ops.setup_link)
7311 		hw->mac.ops.setup_link(hw, speed, true);
7312 
7313 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7314 	adapter->link_check_timeout = jiffies;
7315 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7316 }
7317 
7318 /**
7319  * ixgbe_service_timer - Timer Call-back
7320  * @data: pointer to adapter cast into an unsigned long
7321  **/
7322 static void ixgbe_service_timer(unsigned long data)
7323 {
7324 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
7325 	unsigned long next_event_offset;
7326 
7327 	/* poll faster when waiting for link */
7328 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7329 		next_event_offset = HZ / 10;
7330 	else
7331 		next_event_offset = HZ * 2;
7332 
7333 	/* Reset the timer */
7334 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7335 
7336 	ixgbe_service_event_schedule(adapter);
7337 }
7338 
7339 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7340 {
7341 	struct ixgbe_hw *hw = &adapter->hw;
7342 	u32 status;
7343 
7344 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7345 		return;
7346 
7347 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7348 
7349 	if (!hw->phy.ops.handle_lasi)
7350 		return;
7351 
7352 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7353 	if (status != IXGBE_ERR_OVERTEMP)
7354 		return;
7355 
7356 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7357 }
7358 
7359 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7360 {
7361 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7362 		return;
7363 
7364 	/* If we're already down, removing or resetting, just bail */
7365 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7366 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7367 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7368 		return;
7369 
7370 	ixgbe_dump(adapter);
7371 	netdev_err(adapter->netdev, "Reset adapter\n");
7372 	adapter->tx_timeout_count++;
7373 
7374 	rtnl_lock();
7375 	ixgbe_reinit_locked(adapter);
7376 	rtnl_unlock();
7377 }
7378 
7379 /**
7380  * ixgbe_service_task - manages and runs subtasks
7381  * @work: pointer to work_struct containing our data
7382  **/
7383 static void ixgbe_service_task(struct work_struct *work)
7384 {
7385 	struct ixgbe_adapter *adapter = container_of(work,
7386 						     struct ixgbe_adapter,
7387 						     service_task);
7388 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7389 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7390 			rtnl_lock();
7391 			ixgbe_down(adapter);
7392 			rtnl_unlock();
7393 		}
7394 		ixgbe_service_event_complete(adapter);
7395 		return;
7396 	}
7397 	if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7398 		rtnl_lock();
7399 		adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7400 		udp_tunnel_get_rx_info(adapter->netdev);
7401 		rtnl_unlock();
7402 	}
7403 	ixgbe_reset_subtask(adapter);
7404 	ixgbe_phy_interrupt_subtask(adapter);
7405 	ixgbe_sfp_detection_subtask(adapter);
7406 	ixgbe_sfp_link_config_subtask(adapter);
7407 	ixgbe_check_overtemp_subtask(adapter);
7408 	ixgbe_watchdog_subtask(adapter);
7409 	ixgbe_fdir_reinit_subtask(adapter);
7410 	ixgbe_check_hang_subtask(adapter);
7411 
7412 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7413 		ixgbe_ptp_overflow_check(adapter);
7414 		ixgbe_ptp_rx_hang(adapter);
7415 	}
7416 
7417 	ixgbe_service_event_complete(adapter);
7418 }
7419 
7420 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7421 		     struct ixgbe_tx_buffer *first,
7422 		     u8 *hdr_len)
7423 {
7424 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7425 	struct sk_buff *skb = first->skb;
7426 	union {
7427 		struct iphdr *v4;
7428 		struct ipv6hdr *v6;
7429 		unsigned char *hdr;
7430 	} ip;
7431 	union {
7432 		struct tcphdr *tcp;
7433 		unsigned char *hdr;
7434 	} l4;
7435 	u32 paylen, l4_offset;
7436 	int err;
7437 
7438 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7439 		return 0;
7440 
7441 	if (!skb_is_gso(skb))
7442 		return 0;
7443 
7444 	err = skb_cow_head(skb, 0);
7445 	if (err < 0)
7446 		return err;
7447 
7448 	ip.hdr = skb_network_header(skb);
7449 	l4.hdr = skb_checksum_start(skb);
7450 
7451 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7452 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7453 
7454 	/* initialize outer IP header fields */
7455 	if (ip.v4->version == 4) {
7456 		unsigned char *csum_start = skb_checksum_start(skb);
7457 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
7458 
7459 		/* IP header will have to cancel out any data that
7460 		 * is not a part of the outer IP header
7461 		 */
7462 		ip.v4->check = csum_fold(csum_partial(trans_start,
7463 						      csum_start - trans_start,
7464 						      0));
7465 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7466 
7467 		ip.v4->tot_len = 0;
7468 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7469 				   IXGBE_TX_FLAGS_CSUM |
7470 				   IXGBE_TX_FLAGS_IPV4;
7471 	} else {
7472 		ip.v6->payload_len = 0;
7473 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7474 				   IXGBE_TX_FLAGS_CSUM;
7475 	}
7476 
7477 	/* determine offset of inner transport header */
7478 	l4_offset = l4.hdr - skb->data;
7479 
7480 	/* compute length of segmentation header */
7481 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7482 
7483 	/* remove payload length from inner checksum */
7484 	paylen = skb->len - l4_offset;
7485 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
7486 
7487 	/* update gso size and bytecount with header size */
7488 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7489 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7490 
7491 	/* mss_l4len_id: use 0 as index for TSO */
7492 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7493 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7494 
7495 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7496 	vlan_macip_lens = l4.hdr - ip.hdr;
7497 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7498 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7499 
7500 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
7501 			  mss_l4len_idx);
7502 
7503 	return 1;
7504 }
7505 
7506 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7507 {
7508 	unsigned int offset = 0;
7509 
7510 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7511 
7512 	return offset == skb_checksum_start_offset(skb);
7513 }
7514 
7515 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7516 			  struct ixgbe_tx_buffer *first)
7517 {
7518 	struct sk_buff *skb = first->skb;
7519 	u32 vlan_macip_lens = 0;
7520 	u32 type_tucmd = 0;
7521 
7522 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7523 csum_failed:
7524 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7525 					 IXGBE_TX_FLAGS_CC)))
7526 			return;
7527 		goto no_csum;
7528 	}
7529 
7530 	switch (skb->csum_offset) {
7531 	case offsetof(struct tcphdr, check):
7532 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7533 		/* fall through */
7534 	case offsetof(struct udphdr, check):
7535 		break;
7536 	case offsetof(struct sctphdr, checksum):
7537 		/* validate that this is actually an SCTP request */
7538 		if (((first->protocol == htons(ETH_P_IP)) &&
7539 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7540 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7541 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7542 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7543 			break;
7544 		}
7545 		/* fall through */
7546 	default:
7547 		skb_checksum_help(skb);
7548 		goto csum_failed;
7549 	}
7550 
7551 	/* update TX checksum flag */
7552 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7553 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7554 			  skb_network_offset(skb);
7555 no_csum:
7556 	/* vlan_macip_lens: MACLEN, VLAN tag */
7557 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7558 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7559 
7560 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd, 0);
7561 }
7562 
7563 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7564 	((_flag <= _result) ? \
7565 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7566 	 ((u32)(_input & _flag) / (_flag / _result)))
7567 
7568 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7569 {
7570 	/* set type for advanced descriptor with frame checksum insertion */
7571 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7572 		       IXGBE_ADVTXD_DCMD_DEXT |
7573 		       IXGBE_ADVTXD_DCMD_IFCS;
7574 
7575 	/* set HW vlan bit if vlan is present */
7576 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7577 				   IXGBE_ADVTXD_DCMD_VLE);
7578 
7579 	/* set segmentation enable bits for TSO/FSO */
7580 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7581 				   IXGBE_ADVTXD_DCMD_TSE);
7582 
7583 	/* set timestamp bit if present */
7584 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7585 				   IXGBE_ADVTXD_MAC_TSTAMP);
7586 
7587 	/* insert frame checksum */
7588 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7589 
7590 	return cmd_type;
7591 }
7592 
7593 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7594 				   u32 tx_flags, unsigned int paylen)
7595 {
7596 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7597 
7598 	/* enable L4 checksum for TSO and TX checksum offload */
7599 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7600 					IXGBE_TX_FLAGS_CSUM,
7601 					IXGBE_ADVTXD_POPTS_TXSM);
7602 
7603 	/* enble IPv4 checksum for TSO */
7604 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7605 					IXGBE_TX_FLAGS_IPV4,
7606 					IXGBE_ADVTXD_POPTS_IXSM);
7607 
7608 	/*
7609 	 * Check Context must be set if Tx switch is enabled, which it
7610 	 * always is for case where virtual functions are running
7611 	 */
7612 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7613 					IXGBE_TX_FLAGS_CC,
7614 					IXGBE_ADVTXD_CC);
7615 
7616 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7617 }
7618 
7619 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7620 {
7621 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7622 
7623 	/* Herbert's original patch had:
7624 	 *  smp_mb__after_netif_stop_queue();
7625 	 * but since that doesn't exist yet, just open code it.
7626 	 */
7627 	smp_mb();
7628 
7629 	/* We need to check again in a case another CPU has just
7630 	 * made room available.
7631 	 */
7632 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7633 		return -EBUSY;
7634 
7635 	/* A reprieve! - use start_queue because it doesn't call schedule */
7636 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7637 	++tx_ring->tx_stats.restart_queue;
7638 	return 0;
7639 }
7640 
7641 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7642 {
7643 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7644 		return 0;
7645 
7646 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7647 }
7648 
7649 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7650 		       IXGBE_TXD_CMD_RS)
7651 
7652 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7653 			 struct ixgbe_tx_buffer *first,
7654 			 const u8 hdr_len)
7655 {
7656 	struct sk_buff *skb = first->skb;
7657 	struct ixgbe_tx_buffer *tx_buffer;
7658 	union ixgbe_adv_tx_desc *tx_desc;
7659 	struct skb_frag_struct *frag;
7660 	dma_addr_t dma;
7661 	unsigned int data_len, size;
7662 	u32 tx_flags = first->tx_flags;
7663 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7664 	u16 i = tx_ring->next_to_use;
7665 
7666 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7667 
7668 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7669 
7670 	size = skb_headlen(skb);
7671 	data_len = skb->data_len;
7672 
7673 #ifdef IXGBE_FCOE
7674 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7675 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7676 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7677 			data_len = 0;
7678 		} else {
7679 			data_len -= sizeof(struct fcoe_crc_eof);
7680 		}
7681 	}
7682 
7683 #endif
7684 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7685 
7686 	tx_buffer = first;
7687 
7688 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7689 		if (dma_mapping_error(tx_ring->dev, dma))
7690 			goto dma_error;
7691 
7692 		/* record length, and DMA address */
7693 		dma_unmap_len_set(tx_buffer, len, size);
7694 		dma_unmap_addr_set(tx_buffer, dma, dma);
7695 
7696 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7697 
7698 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7699 			tx_desc->read.cmd_type_len =
7700 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7701 
7702 			i++;
7703 			tx_desc++;
7704 			if (i == tx_ring->count) {
7705 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7706 				i = 0;
7707 			}
7708 			tx_desc->read.olinfo_status = 0;
7709 
7710 			dma += IXGBE_MAX_DATA_PER_TXD;
7711 			size -= IXGBE_MAX_DATA_PER_TXD;
7712 
7713 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7714 		}
7715 
7716 		if (likely(!data_len))
7717 			break;
7718 
7719 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7720 
7721 		i++;
7722 		tx_desc++;
7723 		if (i == tx_ring->count) {
7724 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7725 			i = 0;
7726 		}
7727 		tx_desc->read.olinfo_status = 0;
7728 
7729 #ifdef IXGBE_FCOE
7730 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7731 #else
7732 		size = skb_frag_size(frag);
7733 #endif
7734 		data_len -= size;
7735 
7736 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7737 				       DMA_TO_DEVICE);
7738 
7739 		tx_buffer = &tx_ring->tx_buffer_info[i];
7740 	}
7741 
7742 	/* write last descriptor with RS and EOP bits */
7743 	cmd_type |= size | IXGBE_TXD_CMD;
7744 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7745 
7746 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7747 
7748 	/* set the timestamp */
7749 	first->time_stamp = jiffies;
7750 
7751 	/*
7752 	 * Force memory writes to complete before letting h/w know there
7753 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7754 	 * memory model archs, such as IA-64).
7755 	 *
7756 	 * We also need this memory barrier to make certain all of the
7757 	 * status bits have been updated before next_to_watch is written.
7758 	 */
7759 	wmb();
7760 
7761 	/* set next_to_watch value indicating a packet is present */
7762 	first->next_to_watch = tx_desc;
7763 
7764 	i++;
7765 	if (i == tx_ring->count)
7766 		i = 0;
7767 
7768 	tx_ring->next_to_use = i;
7769 
7770 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7771 
7772 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7773 		writel(i, tx_ring->tail);
7774 
7775 		/* we need this if more than one processor can write to our tail
7776 		 * at a time, it synchronizes IO on IA64/Altix systems
7777 		 */
7778 		mmiowb();
7779 	}
7780 
7781 	return;
7782 dma_error:
7783 	dev_err(tx_ring->dev, "TX DMA map failed\n");
7784 	tx_buffer = &tx_ring->tx_buffer_info[i];
7785 
7786 	/* clear dma mappings for failed tx_buffer_info map */
7787 	while (tx_buffer != first) {
7788 		if (dma_unmap_len(tx_buffer, len))
7789 			dma_unmap_page(tx_ring->dev,
7790 				       dma_unmap_addr(tx_buffer, dma),
7791 				       dma_unmap_len(tx_buffer, len),
7792 				       DMA_TO_DEVICE);
7793 		dma_unmap_len_set(tx_buffer, len, 0);
7794 
7795 		if (i--)
7796 			i += tx_ring->count;
7797 		tx_buffer = &tx_ring->tx_buffer_info[i];
7798 	}
7799 
7800 	if (dma_unmap_len(tx_buffer, len))
7801 		dma_unmap_single(tx_ring->dev,
7802 				 dma_unmap_addr(tx_buffer, dma),
7803 				 dma_unmap_len(tx_buffer, len),
7804 				 DMA_TO_DEVICE);
7805 	dma_unmap_len_set(tx_buffer, len, 0);
7806 
7807 	dev_kfree_skb_any(first->skb);
7808 	first->skb = NULL;
7809 
7810 	tx_ring->next_to_use = i;
7811 }
7812 
7813 static void ixgbe_atr(struct ixgbe_ring *ring,
7814 		      struct ixgbe_tx_buffer *first)
7815 {
7816 	struct ixgbe_q_vector *q_vector = ring->q_vector;
7817 	union ixgbe_atr_hash_dword input = { .dword = 0 };
7818 	union ixgbe_atr_hash_dword common = { .dword = 0 };
7819 	union {
7820 		unsigned char *network;
7821 		struct iphdr *ipv4;
7822 		struct ipv6hdr *ipv6;
7823 	} hdr;
7824 	struct tcphdr *th;
7825 	unsigned int hlen;
7826 	struct sk_buff *skb;
7827 	__be16 vlan_id;
7828 	int l4_proto;
7829 
7830 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7831 	if (!q_vector)
7832 		return;
7833 
7834 	/* do nothing if sampling is disabled */
7835 	if (!ring->atr_sample_rate)
7836 		return;
7837 
7838 	ring->atr_count++;
7839 
7840 	/* currently only IPv4/IPv6 with TCP is supported */
7841 	if ((first->protocol != htons(ETH_P_IP)) &&
7842 	    (first->protocol != htons(ETH_P_IPV6)))
7843 		return;
7844 
7845 	/* snag network header to get L4 type and address */
7846 	skb = first->skb;
7847 	hdr.network = skb_network_header(skb);
7848 	if (unlikely(hdr.network <= skb->data))
7849 		return;
7850 	if (skb->encapsulation &&
7851 	    first->protocol == htons(ETH_P_IP) &&
7852 	    hdr.ipv4->protocol == IPPROTO_UDP) {
7853 		struct ixgbe_adapter *adapter = q_vector->adapter;
7854 
7855 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
7856 			     VXLAN_HEADROOM))
7857 			return;
7858 
7859 		/* verify the port is recognized as VXLAN */
7860 		if (adapter->vxlan_port &&
7861 		    udp_hdr(skb)->dest == adapter->vxlan_port)
7862 			hdr.network = skb_inner_network_header(skb);
7863 
7864 		if (adapter->geneve_port &&
7865 		    udp_hdr(skb)->dest == adapter->geneve_port)
7866 			hdr.network = skb_inner_network_header(skb);
7867 	}
7868 
7869 	/* Make sure we have at least [minimum IPv4 header + TCP]
7870 	 * or [IPv6 header] bytes
7871 	 */
7872 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
7873 		return;
7874 
7875 	/* Currently only IPv4/IPv6 with TCP is supported */
7876 	switch (hdr.ipv4->version) {
7877 	case IPVERSION:
7878 		/* access ihl as u8 to avoid unaligned access on ia64 */
7879 		hlen = (hdr.network[0] & 0x0F) << 2;
7880 		l4_proto = hdr.ipv4->protocol;
7881 		break;
7882 	case 6:
7883 		hlen = hdr.network - skb->data;
7884 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
7885 		hlen -= hdr.network - skb->data;
7886 		break;
7887 	default:
7888 		return;
7889 	}
7890 
7891 	if (l4_proto != IPPROTO_TCP)
7892 		return;
7893 
7894 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
7895 		     hlen + sizeof(struct tcphdr)))
7896 		return;
7897 
7898 	th = (struct tcphdr *)(hdr.network + hlen);
7899 
7900 	/* skip this packet since the socket is closing */
7901 	if (th->fin)
7902 		return;
7903 
7904 	/* sample on all syn packets or once every atr sample count */
7905 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7906 		return;
7907 
7908 	/* reset sample count */
7909 	ring->atr_count = 0;
7910 
7911 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7912 
7913 	/*
7914 	 * src and dst are inverted, think how the receiver sees them
7915 	 *
7916 	 * The input is broken into two sections, a non-compressed section
7917 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7918 	 * is XORed together and stored in the compressed dword.
7919 	 */
7920 	input.formatted.vlan_id = vlan_id;
7921 
7922 	/*
7923 	 * since src port and flex bytes occupy the same word XOR them together
7924 	 * and write the value to source port portion of compressed dword
7925 	 */
7926 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7927 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7928 	else
7929 		common.port.src ^= th->dest ^ first->protocol;
7930 	common.port.dst ^= th->source;
7931 
7932 	switch (hdr.ipv4->version) {
7933 	case IPVERSION:
7934 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7935 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7936 		break;
7937 	case 6:
7938 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7939 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7940 			     hdr.ipv6->saddr.s6_addr32[1] ^
7941 			     hdr.ipv6->saddr.s6_addr32[2] ^
7942 			     hdr.ipv6->saddr.s6_addr32[3] ^
7943 			     hdr.ipv6->daddr.s6_addr32[0] ^
7944 			     hdr.ipv6->daddr.s6_addr32[1] ^
7945 			     hdr.ipv6->daddr.s6_addr32[2] ^
7946 			     hdr.ipv6->daddr.s6_addr32[3];
7947 		break;
7948 	default:
7949 		break;
7950 	}
7951 
7952 	if (hdr.network != skb_network_header(skb))
7953 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
7954 
7955 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7956 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7957 					      input, common, ring->queue_index);
7958 }
7959 
7960 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7961 			      void *accel_priv, select_queue_fallback_t fallback)
7962 {
7963 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7964 #ifdef IXGBE_FCOE
7965 	struct ixgbe_adapter *adapter;
7966 	struct ixgbe_ring_feature *f;
7967 	int txq;
7968 #endif
7969 
7970 	if (fwd_adapter)
7971 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7972 
7973 #ifdef IXGBE_FCOE
7974 
7975 	/*
7976 	 * only execute the code below if protocol is FCoE
7977 	 * or FIP and we have FCoE enabled on the adapter
7978 	 */
7979 	switch (vlan_get_protocol(skb)) {
7980 	case htons(ETH_P_FCOE):
7981 	case htons(ETH_P_FIP):
7982 		adapter = netdev_priv(dev);
7983 
7984 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7985 			break;
7986 	default:
7987 		return fallback(dev, skb);
7988 	}
7989 
7990 	f = &adapter->ring_feature[RING_F_FCOE];
7991 
7992 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7993 					   smp_processor_id();
7994 
7995 	while (txq >= f->indices)
7996 		txq -= f->indices;
7997 
7998 	return txq + f->offset;
7999 #else
8000 	return fallback(dev, skb);
8001 #endif
8002 }
8003 
8004 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8005 			  struct ixgbe_adapter *adapter,
8006 			  struct ixgbe_ring *tx_ring)
8007 {
8008 	struct ixgbe_tx_buffer *first;
8009 	int tso;
8010 	u32 tx_flags = 0;
8011 	unsigned short f;
8012 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
8013 	__be16 protocol = skb->protocol;
8014 	u8 hdr_len = 0;
8015 
8016 	/*
8017 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8018 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8019 	 *       + 2 desc gap to keep tail from touching head,
8020 	 *       + 1 desc for context descriptor,
8021 	 * otherwise try next time
8022 	 */
8023 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8024 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
8025 
8026 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8027 		tx_ring->tx_stats.tx_busy++;
8028 		return NETDEV_TX_BUSY;
8029 	}
8030 
8031 	/* record the location of the first descriptor for this packet */
8032 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8033 	first->skb = skb;
8034 	first->bytecount = skb->len;
8035 	first->gso_segs = 1;
8036 
8037 	/* if we have a HW VLAN tag being added default to the HW one */
8038 	if (skb_vlan_tag_present(skb)) {
8039 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8040 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8041 	/* else if it is a SW VLAN check the next protocol and store the tag */
8042 	} else if (protocol == htons(ETH_P_8021Q)) {
8043 		struct vlan_hdr *vhdr, _vhdr;
8044 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8045 		if (!vhdr)
8046 			goto out_drop;
8047 
8048 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8049 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
8050 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8051 	}
8052 	protocol = vlan_get_protocol(skb);
8053 
8054 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8055 	    adapter->ptp_clock &&
8056 	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8057 				   &adapter->state)) {
8058 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8059 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8060 
8061 		/* schedule check for Tx timestamp */
8062 		adapter->ptp_tx_skb = skb_get(skb);
8063 		adapter->ptp_tx_start = jiffies;
8064 		schedule_work(&adapter->ptp_tx_work);
8065 	}
8066 
8067 	skb_tx_timestamp(skb);
8068 
8069 #ifdef CONFIG_PCI_IOV
8070 	/*
8071 	 * Use the l2switch_enable flag - would be false if the DMA
8072 	 * Tx switch had been disabled.
8073 	 */
8074 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8075 		tx_flags |= IXGBE_TX_FLAGS_CC;
8076 
8077 #endif
8078 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8079 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8080 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8081 	     (skb->priority != TC_PRIO_CONTROL))) {
8082 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8083 		tx_flags |= (skb->priority & 0x7) <<
8084 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8085 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8086 			struct vlan_ethhdr *vhdr;
8087 
8088 			if (skb_cow_head(skb, 0))
8089 				goto out_drop;
8090 			vhdr = (struct vlan_ethhdr *)skb->data;
8091 			vhdr->h_vlan_TCI = htons(tx_flags >>
8092 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
8093 		} else {
8094 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8095 		}
8096 	}
8097 
8098 	/* record initial flags and protocol */
8099 	first->tx_flags = tx_flags;
8100 	first->protocol = protocol;
8101 
8102 #ifdef IXGBE_FCOE
8103 	/* setup tx offload for FCoE */
8104 	if ((protocol == htons(ETH_P_FCOE)) &&
8105 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8106 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
8107 		if (tso < 0)
8108 			goto out_drop;
8109 
8110 		goto xmit_fcoe;
8111 	}
8112 
8113 #endif /* IXGBE_FCOE */
8114 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
8115 	if (tso < 0)
8116 		goto out_drop;
8117 	else if (!tso)
8118 		ixgbe_tx_csum(tx_ring, first);
8119 
8120 	/* add the ATR filter if ATR is on */
8121 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8122 		ixgbe_atr(tx_ring, first);
8123 
8124 #ifdef IXGBE_FCOE
8125 xmit_fcoe:
8126 #endif /* IXGBE_FCOE */
8127 	ixgbe_tx_map(tx_ring, first, hdr_len);
8128 
8129 	return NETDEV_TX_OK;
8130 
8131 out_drop:
8132 	dev_kfree_skb_any(first->skb);
8133 	first->skb = NULL;
8134 
8135 	return NETDEV_TX_OK;
8136 }
8137 
8138 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8139 				      struct net_device *netdev,
8140 				      struct ixgbe_ring *ring)
8141 {
8142 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8143 	struct ixgbe_ring *tx_ring;
8144 
8145 	/*
8146 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
8147 	 * in order to meet this minimum size requirement.
8148 	 */
8149 	if (skb_put_padto(skb, 17))
8150 		return NETDEV_TX_OK;
8151 
8152 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
8153 
8154 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8155 }
8156 
8157 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8158 				    struct net_device *netdev)
8159 {
8160 	return __ixgbe_xmit_frame(skb, netdev, NULL);
8161 }
8162 
8163 /**
8164  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8165  * @netdev: network interface device structure
8166  * @p: pointer to an address structure
8167  *
8168  * Returns 0 on success, negative on failure
8169  **/
8170 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8171 {
8172 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8173 	struct ixgbe_hw *hw = &adapter->hw;
8174 	struct sockaddr *addr = p;
8175 
8176 	if (!is_valid_ether_addr(addr->sa_data))
8177 		return -EADDRNOTAVAIL;
8178 
8179 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
8180 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8181 
8182 	ixgbe_mac_set_default_filter(adapter);
8183 
8184 	return 0;
8185 }
8186 
8187 static int
8188 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8189 {
8190 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8191 	struct ixgbe_hw *hw = &adapter->hw;
8192 	u16 value;
8193 	int rc;
8194 
8195 	if (prtad != hw->phy.mdio.prtad)
8196 		return -EINVAL;
8197 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8198 	if (!rc)
8199 		rc = value;
8200 	return rc;
8201 }
8202 
8203 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8204 			    u16 addr, u16 value)
8205 {
8206 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8207 	struct ixgbe_hw *hw = &adapter->hw;
8208 
8209 	if (prtad != hw->phy.mdio.prtad)
8210 		return -EINVAL;
8211 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8212 }
8213 
8214 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8215 {
8216 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8217 
8218 	switch (cmd) {
8219 	case SIOCSHWTSTAMP:
8220 		return ixgbe_ptp_set_ts_config(adapter, req);
8221 	case SIOCGHWTSTAMP:
8222 		return ixgbe_ptp_get_ts_config(adapter, req);
8223 	default:
8224 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8225 	}
8226 }
8227 
8228 /**
8229  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8230  * netdev->dev_addrs
8231  * @netdev: network interface device structure
8232  *
8233  * Returns non-zero on failure
8234  **/
8235 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8236 {
8237 	int err = 0;
8238 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8239 	struct ixgbe_hw *hw = &adapter->hw;
8240 
8241 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8242 		rtnl_lock();
8243 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8244 		rtnl_unlock();
8245 
8246 		/* update SAN MAC vmdq pool selection */
8247 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8248 	}
8249 	return err;
8250 }
8251 
8252 /**
8253  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8254  * netdev->dev_addrs
8255  * @netdev: network interface device structure
8256  *
8257  * Returns non-zero on failure
8258  **/
8259 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8260 {
8261 	int err = 0;
8262 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8263 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8264 
8265 	if (is_valid_ether_addr(mac->san_addr)) {
8266 		rtnl_lock();
8267 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8268 		rtnl_unlock();
8269 	}
8270 	return err;
8271 }
8272 
8273 #ifdef CONFIG_NET_POLL_CONTROLLER
8274 /*
8275  * Polling 'interrupt' - used by things like netconsole to send skbs
8276  * without having to re-enable interrupts. It's not called while
8277  * the interrupt routine is executing.
8278  */
8279 static void ixgbe_netpoll(struct net_device *netdev)
8280 {
8281 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8282 	int i;
8283 
8284 	/* if interface is down do nothing */
8285 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8286 		return;
8287 
8288 	/* loop through and schedule all active queues */
8289 	for (i = 0; i < adapter->num_q_vectors; i++)
8290 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8291 }
8292 
8293 #endif
8294 
8295 static void ixgbe_get_stats64(struct net_device *netdev,
8296 			      struct rtnl_link_stats64 *stats)
8297 {
8298 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8299 	int i;
8300 
8301 	rcu_read_lock();
8302 	for (i = 0; i < adapter->num_rx_queues; i++) {
8303 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
8304 		u64 bytes, packets;
8305 		unsigned int start;
8306 
8307 		if (ring) {
8308 			do {
8309 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8310 				packets = ring->stats.packets;
8311 				bytes   = ring->stats.bytes;
8312 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8313 			stats->rx_packets += packets;
8314 			stats->rx_bytes   += bytes;
8315 		}
8316 	}
8317 
8318 	for (i = 0; i < adapter->num_tx_queues; i++) {
8319 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
8320 		u64 bytes, packets;
8321 		unsigned int start;
8322 
8323 		if (ring) {
8324 			do {
8325 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8326 				packets = ring->stats.packets;
8327 				bytes   = ring->stats.bytes;
8328 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8329 			stats->tx_packets += packets;
8330 			stats->tx_bytes   += bytes;
8331 		}
8332 	}
8333 	rcu_read_unlock();
8334 
8335 	/* following stats updated by ixgbe_watchdog_task() */
8336 	stats->multicast	= netdev->stats.multicast;
8337 	stats->rx_errors	= netdev->stats.rx_errors;
8338 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8339 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8340 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8341 }
8342 
8343 #ifdef CONFIG_IXGBE_DCB
8344 /**
8345  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8346  * @adapter: pointer to ixgbe_adapter
8347  * @tc: number of traffic classes currently enabled
8348  *
8349  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8350  * 802.1Q priority maps to a packet buffer that exists.
8351  */
8352 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8353 {
8354 	struct ixgbe_hw *hw = &adapter->hw;
8355 	u32 reg, rsave;
8356 	int i;
8357 
8358 	/* 82598 have a static priority to TC mapping that can not
8359 	 * be changed so no validation is needed.
8360 	 */
8361 	if (hw->mac.type == ixgbe_mac_82598EB)
8362 		return;
8363 
8364 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8365 	rsave = reg;
8366 
8367 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8368 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8369 
8370 		/* If up2tc is out of bounds default to zero */
8371 		if (up2tc > tc)
8372 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8373 	}
8374 
8375 	if (reg != rsave)
8376 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8377 
8378 	return;
8379 }
8380 
8381 /**
8382  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8383  * @adapter: Pointer to adapter struct
8384  *
8385  * Populate the netdev user priority to tc map
8386  */
8387 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8388 {
8389 	struct net_device *dev = adapter->netdev;
8390 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8391 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8392 	u8 prio;
8393 
8394 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8395 		u8 tc = 0;
8396 
8397 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8398 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8399 		else if (ets)
8400 			tc = ets->prio_tc[prio];
8401 
8402 		netdev_set_prio_tc_map(dev, prio, tc);
8403 	}
8404 }
8405 
8406 #endif /* CONFIG_IXGBE_DCB */
8407 /**
8408  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8409  *
8410  * @netdev: net device to configure
8411  * @tc: number of traffic classes to enable
8412  */
8413 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8414 {
8415 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8416 	struct ixgbe_hw *hw = &adapter->hw;
8417 	bool pools;
8418 
8419 	/* Hardware supports up to 8 traffic classes */
8420 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8421 		return -EINVAL;
8422 
8423 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8424 		return -EINVAL;
8425 
8426 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
8427 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
8428 		return -EBUSY;
8429 
8430 	/* Hardware has to reinitialize queues and interrupts to
8431 	 * match packet buffer alignment. Unfortunately, the
8432 	 * hardware is not flexible enough to do this dynamically.
8433 	 */
8434 	if (netif_running(dev))
8435 		ixgbe_close(dev);
8436 	else
8437 		ixgbe_reset(adapter);
8438 
8439 	ixgbe_clear_interrupt_scheme(adapter);
8440 
8441 #ifdef CONFIG_IXGBE_DCB
8442 	if (tc) {
8443 		netdev_set_num_tc(dev, tc);
8444 		ixgbe_set_prio_tc_map(adapter);
8445 
8446 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8447 
8448 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8449 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8450 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8451 		}
8452 	} else {
8453 		netdev_reset_tc(dev);
8454 
8455 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8456 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8457 
8458 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8459 
8460 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8461 		adapter->dcb_cfg.pfc_mode_enable = false;
8462 	}
8463 
8464 	ixgbe_validate_rtr(adapter, tc);
8465 
8466 #endif /* CONFIG_IXGBE_DCB */
8467 	ixgbe_init_interrupt_scheme(adapter);
8468 
8469 	if (netif_running(dev))
8470 		return ixgbe_open(dev);
8471 
8472 	return 0;
8473 }
8474 
8475 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8476 			       struct tc_cls_u32_offload *cls)
8477 {
8478 	u32 hdl = cls->knode.handle;
8479 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8480 	u32 loc = cls->knode.handle & 0xfffff;
8481 	int err = 0, i, j;
8482 	struct ixgbe_jump_table *jump = NULL;
8483 
8484 	if (loc > IXGBE_MAX_HW_ENTRIES)
8485 		return -EINVAL;
8486 
8487 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8488 		return -EINVAL;
8489 
8490 	/* Clear this filter in the link data it is associated with */
8491 	if (uhtid != 0x800) {
8492 		jump = adapter->jump_tables[uhtid];
8493 		if (!jump)
8494 			return -EINVAL;
8495 		if (!test_bit(loc - 1, jump->child_loc_map))
8496 			return -EINVAL;
8497 		clear_bit(loc - 1, jump->child_loc_map);
8498 	}
8499 
8500 	/* Check if the filter being deleted is a link */
8501 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8502 		jump = adapter->jump_tables[i];
8503 		if (jump && jump->link_hdl == hdl) {
8504 			/* Delete filters in the hardware in the child hash
8505 			 * table associated with this link
8506 			 */
8507 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8508 				if (!test_bit(j, jump->child_loc_map))
8509 					continue;
8510 				spin_lock(&adapter->fdir_perfect_lock);
8511 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8512 								      NULL,
8513 								      j + 1);
8514 				spin_unlock(&adapter->fdir_perfect_lock);
8515 				clear_bit(j, jump->child_loc_map);
8516 			}
8517 			/* Remove resources for this link */
8518 			kfree(jump->input);
8519 			kfree(jump->mask);
8520 			kfree(jump);
8521 			adapter->jump_tables[i] = NULL;
8522 			return err;
8523 		}
8524 	}
8525 
8526 	spin_lock(&adapter->fdir_perfect_lock);
8527 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8528 	spin_unlock(&adapter->fdir_perfect_lock);
8529 	return err;
8530 }
8531 
8532 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8533 					    __be16 protocol,
8534 					    struct tc_cls_u32_offload *cls)
8535 {
8536 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8537 
8538 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8539 		return -EINVAL;
8540 
8541 	/* This ixgbe devices do not support hash tables at the moment
8542 	 * so abort when given hash tables.
8543 	 */
8544 	if (cls->hnode.divisor > 0)
8545 		return -EINVAL;
8546 
8547 	set_bit(uhtid - 1, &adapter->tables);
8548 	return 0;
8549 }
8550 
8551 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
8552 					    struct tc_cls_u32_offload *cls)
8553 {
8554 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8555 
8556 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8557 		return -EINVAL;
8558 
8559 	clear_bit(uhtid - 1, &adapter->tables);
8560 	return 0;
8561 }
8562 
8563 #ifdef CONFIG_NET_CLS_ACT
8564 struct upper_walk_data {
8565 	struct ixgbe_adapter *adapter;
8566 	u64 action;
8567 	int ifindex;
8568 	u8 queue;
8569 };
8570 
8571 static int get_macvlan_queue(struct net_device *upper, void *_data)
8572 {
8573 	if (netif_is_macvlan(upper)) {
8574 		struct macvlan_dev *dfwd = netdev_priv(upper);
8575 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
8576 		struct upper_walk_data *data = _data;
8577 		struct ixgbe_adapter *adapter = data->adapter;
8578 		int ifindex = data->ifindex;
8579 
8580 		if (vadapter && vadapter->netdev->ifindex == ifindex) {
8581 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
8582 			data->action = data->queue;
8583 			return 1;
8584 		}
8585 	}
8586 
8587 	return 0;
8588 }
8589 
8590 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
8591 				  u8 *queue, u64 *action)
8592 {
8593 	unsigned int num_vfs = adapter->num_vfs, vf;
8594 	struct upper_walk_data data;
8595 	struct net_device *upper;
8596 
8597 	/* redirect to a SRIOV VF */
8598 	for (vf = 0; vf < num_vfs; ++vf) {
8599 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
8600 		if (upper->ifindex == ifindex) {
8601 			if (adapter->num_rx_pools > 1)
8602 				*queue = vf * 2;
8603 			else
8604 				*queue = vf * adapter->num_rx_queues_per_pool;
8605 
8606 			*action = vf + 1;
8607 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
8608 			return 0;
8609 		}
8610 	}
8611 
8612 	/* redirect to a offloaded macvlan netdev */
8613 	data.adapter = adapter;
8614 	data.ifindex = ifindex;
8615 	data.action = 0;
8616 	data.queue = 0;
8617 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
8618 					  get_macvlan_queue, &data)) {
8619 		*action = data.action;
8620 		*queue = data.queue;
8621 
8622 		return 0;
8623 	}
8624 
8625 	return -EINVAL;
8626 }
8627 
8628 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8629 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8630 {
8631 	const struct tc_action *a;
8632 	LIST_HEAD(actions);
8633 	int err;
8634 
8635 	if (tc_no_actions(exts))
8636 		return -EINVAL;
8637 
8638 	tcf_exts_to_list(exts, &actions);
8639 	list_for_each_entry(a, &actions, list) {
8640 
8641 		/* Drop action */
8642 		if (is_tcf_gact_shot(a)) {
8643 			*action = IXGBE_FDIR_DROP_QUEUE;
8644 			*queue = IXGBE_FDIR_DROP_QUEUE;
8645 			return 0;
8646 		}
8647 
8648 		/* Redirect to a VF or a offloaded macvlan */
8649 		if (is_tcf_mirred_egress_redirect(a)) {
8650 			int ifindex = tcf_mirred_ifindex(a);
8651 
8652 			err = handle_redirect_action(adapter, ifindex, queue,
8653 						     action);
8654 			if (err == 0)
8655 				return err;
8656 		}
8657 	}
8658 
8659 	return -EINVAL;
8660 }
8661 #else
8662 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8663 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8664 {
8665 	return -EINVAL;
8666 }
8667 #endif /* CONFIG_NET_CLS_ACT */
8668 
8669 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
8670 				    union ixgbe_atr_input *mask,
8671 				    struct tc_cls_u32_offload *cls,
8672 				    struct ixgbe_mat_field *field_ptr,
8673 				    struct ixgbe_nexthdr *nexthdr)
8674 {
8675 	int i, j, off;
8676 	__be32 val, m;
8677 	bool found_entry = false, found_jump_field = false;
8678 
8679 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
8680 		off = cls->knode.sel->keys[i].off;
8681 		val = cls->knode.sel->keys[i].val;
8682 		m = cls->knode.sel->keys[i].mask;
8683 
8684 		for (j = 0; field_ptr[j].val; j++) {
8685 			if (field_ptr[j].off == off) {
8686 				field_ptr[j].val(input, mask, val, m);
8687 				input->filter.formatted.flow_type |=
8688 					field_ptr[j].type;
8689 				found_entry = true;
8690 				break;
8691 			}
8692 		}
8693 		if (nexthdr) {
8694 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
8695 			    nexthdr->val == cls->knode.sel->keys[i].val &&
8696 			    nexthdr->mask == cls->knode.sel->keys[i].mask)
8697 				found_jump_field = true;
8698 			else
8699 				continue;
8700 		}
8701 	}
8702 
8703 	if (nexthdr && !found_jump_field)
8704 		return -EINVAL;
8705 
8706 	if (!found_entry)
8707 		return 0;
8708 
8709 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
8710 				    IXGBE_ATR_L4TYPE_MASK;
8711 
8712 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
8713 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
8714 
8715 	return 0;
8716 }
8717 
8718 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
8719 				  __be16 protocol,
8720 				  struct tc_cls_u32_offload *cls)
8721 {
8722 	u32 loc = cls->knode.handle & 0xfffff;
8723 	struct ixgbe_hw *hw = &adapter->hw;
8724 	struct ixgbe_mat_field *field_ptr;
8725 	struct ixgbe_fdir_filter *input = NULL;
8726 	union ixgbe_atr_input *mask = NULL;
8727 	struct ixgbe_jump_table *jump = NULL;
8728 	int i, err = -EINVAL;
8729 	u8 queue;
8730 	u32 uhtid, link_uhtid;
8731 
8732 	uhtid = TC_U32_USERHTID(cls->knode.handle);
8733 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
8734 
8735 	/* At the moment cls_u32 jumps to network layer and skips past
8736 	 * L2 headers. The canonical method to match L2 frames is to use
8737 	 * negative values. However this is error prone at best but really
8738 	 * just broken because there is no way to "know" what sort of hdr
8739 	 * is in front of the network layer. Fix cls_u32 to support L2
8740 	 * headers when needed.
8741 	 */
8742 	if (protocol != htons(ETH_P_IP))
8743 		return err;
8744 
8745 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
8746 		e_err(drv, "Location out of range\n");
8747 		return err;
8748 	}
8749 
8750 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
8751 	 * links and also the fields used to advance the parser across each
8752 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
8753 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
8754 	 * To add support for new nodes update ixgbe_model.h parse structures
8755 	 * this function _should_ be generic try not to hardcode values here.
8756 	 */
8757 	if (uhtid == 0x800) {
8758 		field_ptr = (adapter->jump_tables[0])->mat;
8759 	} else {
8760 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8761 			return err;
8762 		if (!adapter->jump_tables[uhtid])
8763 			return err;
8764 		field_ptr = (adapter->jump_tables[uhtid])->mat;
8765 	}
8766 
8767 	if (!field_ptr)
8768 		return err;
8769 
8770 	/* At this point we know the field_ptr is valid and need to either
8771 	 * build cls_u32 link or attach filter. Because adding a link to
8772 	 * a handle that does not exist is invalid and the same for adding
8773 	 * rules to handles that don't exist.
8774 	 */
8775 
8776 	if (link_uhtid) {
8777 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
8778 
8779 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
8780 			return err;
8781 
8782 		if (!test_bit(link_uhtid - 1, &adapter->tables))
8783 			return err;
8784 
8785 		/* Multiple filters as links to the same hash table are not
8786 		 * supported. To add a new filter with the same next header
8787 		 * but different match/jump conditions, create a new hash table
8788 		 * and link to it.
8789 		 */
8790 		if (adapter->jump_tables[link_uhtid] &&
8791 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
8792 			e_err(drv, "Link filter exists for link: %x\n",
8793 			      link_uhtid);
8794 			return err;
8795 		}
8796 
8797 		for (i = 0; nexthdr[i].jump; i++) {
8798 			if (nexthdr[i].o != cls->knode.sel->offoff ||
8799 			    nexthdr[i].s != cls->knode.sel->offshift ||
8800 			    nexthdr[i].m != cls->knode.sel->offmask)
8801 				return err;
8802 
8803 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
8804 			if (!jump)
8805 				return -ENOMEM;
8806 			input = kzalloc(sizeof(*input), GFP_KERNEL);
8807 			if (!input) {
8808 				err = -ENOMEM;
8809 				goto free_jump;
8810 			}
8811 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8812 			if (!mask) {
8813 				err = -ENOMEM;
8814 				goto free_input;
8815 			}
8816 			jump->input = input;
8817 			jump->mask = mask;
8818 			jump->link_hdl = cls->knode.handle;
8819 
8820 			err = ixgbe_clsu32_build_input(input, mask, cls,
8821 						       field_ptr, &nexthdr[i]);
8822 			if (!err) {
8823 				jump->mat = nexthdr[i].jump;
8824 				adapter->jump_tables[link_uhtid] = jump;
8825 				break;
8826 			}
8827 		}
8828 		return 0;
8829 	}
8830 
8831 	input = kzalloc(sizeof(*input), GFP_KERNEL);
8832 	if (!input)
8833 		return -ENOMEM;
8834 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8835 	if (!mask) {
8836 		err = -ENOMEM;
8837 		goto free_input;
8838 	}
8839 
8840 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
8841 		if ((adapter->jump_tables[uhtid])->input)
8842 			memcpy(input, (adapter->jump_tables[uhtid])->input,
8843 			       sizeof(*input));
8844 		if ((adapter->jump_tables[uhtid])->mask)
8845 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
8846 			       sizeof(*mask));
8847 
8848 		/* Lookup in all child hash tables if this location is already
8849 		 * filled with a filter
8850 		 */
8851 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8852 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
8853 
8854 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
8855 				e_err(drv, "Filter exists in location: %x\n",
8856 				      loc);
8857 				err = -EINVAL;
8858 				goto err_out;
8859 			}
8860 		}
8861 	}
8862 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
8863 	if (err)
8864 		goto err_out;
8865 
8866 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
8867 			       &queue);
8868 	if (err < 0)
8869 		goto err_out;
8870 
8871 	input->sw_idx = loc;
8872 
8873 	spin_lock(&adapter->fdir_perfect_lock);
8874 
8875 	if (hlist_empty(&adapter->fdir_filter_list)) {
8876 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
8877 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
8878 		if (err)
8879 			goto err_out_w_lock;
8880 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
8881 		err = -EINVAL;
8882 		goto err_out_w_lock;
8883 	}
8884 
8885 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
8886 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
8887 						    input->sw_idx, queue);
8888 	if (!err)
8889 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
8890 	spin_unlock(&adapter->fdir_perfect_lock);
8891 
8892 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
8893 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
8894 
8895 	kfree(mask);
8896 	return err;
8897 err_out_w_lock:
8898 	spin_unlock(&adapter->fdir_perfect_lock);
8899 err_out:
8900 	kfree(mask);
8901 free_input:
8902 	kfree(input);
8903 free_jump:
8904 	kfree(jump);
8905 	return err;
8906 }
8907 
8908 static int __ixgbe_setup_tc(struct net_device *dev, u32 handle, __be16 proto,
8909 			    struct tc_to_netdev *tc)
8910 {
8911 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8912 
8913 	if (TC_H_MAJ(handle) == TC_H_MAJ(TC_H_INGRESS) &&
8914 	    tc->type == TC_SETUP_CLSU32) {
8915 		switch (tc->cls_u32->command) {
8916 		case TC_CLSU32_NEW_KNODE:
8917 		case TC_CLSU32_REPLACE_KNODE:
8918 			return ixgbe_configure_clsu32(adapter,
8919 						      proto, tc->cls_u32);
8920 		case TC_CLSU32_DELETE_KNODE:
8921 			return ixgbe_delete_clsu32(adapter, tc->cls_u32);
8922 		case TC_CLSU32_NEW_HNODE:
8923 		case TC_CLSU32_REPLACE_HNODE:
8924 			return ixgbe_configure_clsu32_add_hnode(adapter, proto,
8925 								tc->cls_u32);
8926 		case TC_CLSU32_DELETE_HNODE:
8927 			return ixgbe_configure_clsu32_del_hnode(adapter,
8928 								tc->cls_u32);
8929 		default:
8930 			return -EINVAL;
8931 		}
8932 	}
8933 
8934 	if (tc->type != TC_SETUP_MQPRIO)
8935 		return -EINVAL;
8936 
8937 	return ixgbe_setup_tc(dev, tc->tc);
8938 }
8939 
8940 #ifdef CONFIG_PCI_IOV
8941 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
8942 {
8943 	struct net_device *netdev = adapter->netdev;
8944 
8945 	rtnl_lock();
8946 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
8947 	rtnl_unlock();
8948 }
8949 
8950 #endif
8951 void ixgbe_do_reset(struct net_device *netdev)
8952 {
8953 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8954 
8955 	if (netif_running(netdev))
8956 		ixgbe_reinit_locked(adapter);
8957 	else
8958 		ixgbe_reset(adapter);
8959 }
8960 
8961 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
8962 					    netdev_features_t features)
8963 {
8964 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8965 
8966 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
8967 	if (!(features & NETIF_F_RXCSUM))
8968 		features &= ~NETIF_F_LRO;
8969 
8970 	/* Turn off LRO if not RSC capable */
8971 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
8972 		features &= ~NETIF_F_LRO;
8973 
8974 	return features;
8975 }
8976 
8977 static int ixgbe_set_features(struct net_device *netdev,
8978 			      netdev_features_t features)
8979 {
8980 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8981 	netdev_features_t changed = netdev->features ^ features;
8982 	bool need_reset = false;
8983 
8984 	/* Make sure RSC matches LRO, reset if change */
8985 	if (!(features & NETIF_F_LRO)) {
8986 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8987 			need_reset = true;
8988 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
8989 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
8990 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
8991 		if (adapter->rx_itr_setting == 1 ||
8992 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
8993 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
8994 			need_reset = true;
8995 		} else if ((changed ^ features) & NETIF_F_LRO) {
8996 			e_info(probe, "rx-usecs set too low, "
8997 			       "disabling RSC\n");
8998 		}
8999 	}
9000 
9001 	/*
9002 	 * Check if Flow Director n-tuple support or hw_tc support was
9003 	 * enabled or disabled.  If the state changed, we need to reset.
9004 	 */
9005 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9006 		/* turn off ATR, enable perfect filters and reset */
9007 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9008 			need_reset = true;
9009 
9010 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9011 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9012 	} else {
9013 		/* turn off perfect filters, enable ATR and reset */
9014 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9015 			need_reset = true;
9016 
9017 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9018 
9019 		/* We cannot enable ATR if SR-IOV is enabled */
9020 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9021 		    /* We cannot enable ATR if we have 2 or more tcs */
9022 		    (netdev_get_num_tc(netdev) > 1) ||
9023 		    /* We cannot enable ATR if RSS is disabled */
9024 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9025 		    /* A sample rate of 0 indicates ATR disabled */
9026 		    (!adapter->atr_sample_rate))
9027 			; /* do nothing not supported */
9028 		else /* otherwise supported and set the flag */
9029 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9030 	}
9031 
9032 	if (changed & NETIF_F_RXALL)
9033 		need_reset = true;
9034 
9035 	netdev->features = features;
9036 
9037 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
9038 		if (features & NETIF_F_RXCSUM) {
9039 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9040 		} else {
9041 			u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9042 
9043 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9044 		}
9045 	}
9046 
9047 	if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
9048 		if (features & NETIF_F_RXCSUM) {
9049 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9050 		} else {
9051 			u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9052 
9053 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9054 		}
9055 	}
9056 
9057 	if (need_reset)
9058 		ixgbe_do_reset(netdev);
9059 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9060 			    NETIF_F_HW_VLAN_CTAG_FILTER))
9061 		ixgbe_set_rx_mode(netdev);
9062 
9063 	return 0;
9064 }
9065 
9066 /**
9067  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
9068  * @dev: The port's netdev
9069  * @ti: Tunnel endpoint information
9070  **/
9071 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
9072 				      struct udp_tunnel_info *ti)
9073 {
9074 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9075 	struct ixgbe_hw *hw = &adapter->hw;
9076 	__be16 port = ti->port;
9077 	u32 port_shift = 0;
9078 	u32 reg;
9079 
9080 	if (ti->sa_family != AF_INET)
9081 		return;
9082 
9083 	switch (ti->type) {
9084 	case UDP_TUNNEL_TYPE_VXLAN:
9085 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9086 			return;
9087 
9088 		if (adapter->vxlan_port == port)
9089 			return;
9090 
9091 		if (adapter->vxlan_port) {
9092 			netdev_info(dev,
9093 				    "VXLAN port %d set, not adding port %d\n",
9094 				    ntohs(adapter->vxlan_port),
9095 				    ntohs(port));
9096 			return;
9097 		}
9098 
9099 		adapter->vxlan_port = port;
9100 		break;
9101 	case UDP_TUNNEL_TYPE_GENEVE:
9102 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9103 			return;
9104 
9105 		if (adapter->geneve_port == port)
9106 			return;
9107 
9108 		if (adapter->geneve_port) {
9109 			netdev_info(dev,
9110 				    "GENEVE port %d set, not adding port %d\n",
9111 				    ntohs(adapter->geneve_port),
9112 				    ntohs(port));
9113 			return;
9114 		}
9115 
9116 		port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
9117 		adapter->geneve_port = port;
9118 		break;
9119 	default:
9120 		return;
9121 	}
9122 
9123 	reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
9124 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
9125 }
9126 
9127 /**
9128  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
9129  * @dev: The port's netdev
9130  * @ti: Tunnel endpoint information
9131  **/
9132 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
9133 				      struct udp_tunnel_info *ti)
9134 {
9135 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9136 	u32 port_mask;
9137 
9138 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
9139 	    ti->type != UDP_TUNNEL_TYPE_GENEVE)
9140 		return;
9141 
9142 	if (ti->sa_family != AF_INET)
9143 		return;
9144 
9145 	switch (ti->type) {
9146 	case UDP_TUNNEL_TYPE_VXLAN:
9147 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9148 			return;
9149 
9150 		if (adapter->vxlan_port != ti->port) {
9151 			netdev_info(dev, "VXLAN port %d not found\n",
9152 				    ntohs(ti->port));
9153 			return;
9154 		}
9155 
9156 		port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9157 		break;
9158 	case UDP_TUNNEL_TYPE_GENEVE:
9159 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9160 			return;
9161 
9162 		if (adapter->geneve_port != ti->port) {
9163 			netdev_info(dev, "GENEVE port %d not found\n",
9164 				    ntohs(ti->port));
9165 			return;
9166 		}
9167 
9168 		port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9169 		break;
9170 	default:
9171 		return;
9172 	}
9173 
9174 	ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9175 	adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9176 }
9177 
9178 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9179 			     struct net_device *dev,
9180 			     const unsigned char *addr, u16 vid,
9181 			     u16 flags)
9182 {
9183 	/* guarantee we can provide a unique filter for the unicast address */
9184 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9185 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9186 		u16 pool = VMDQ_P(0);
9187 
9188 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9189 			return -ENOMEM;
9190 	}
9191 
9192 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9193 }
9194 
9195 /**
9196  * ixgbe_configure_bridge_mode - set various bridge modes
9197  * @adapter - the private structure
9198  * @mode - requested bridge mode
9199  *
9200  * Configure some settings require for various bridge modes.
9201  **/
9202 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9203 				       __u16 mode)
9204 {
9205 	struct ixgbe_hw *hw = &adapter->hw;
9206 	unsigned int p, num_pools;
9207 	u32 vmdctl;
9208 
9209 	switch (mode) {
9210 	case BRIDGE_MODE_VEPA:
9211 		/* disable Tx loopback, rely on switch hairpin mode */
9212 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9213 
9214 		/* must enable Rx switching replication to allow multicast
9215 		 * packet reception on all VFs, and to enable source address
9216 		 * pruning.
9217 		 */
9218 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9219 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9220 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9221 
9222 		/* enable Rx source address pruning. Note, this requires
9223 		 * replication to be enabled or else it does nothing.
9224 		 */
9225 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9226 		for (p = 0; p < num_pools; p++) {
9227 			if (hw->mac.ops.set_source_address_pruning)
9228 				hw->mac.ops.set_source_address_pruning(hw,
9229 								       true,
9230 								       p);
9231 		}
9232 		break;
9233 	case BRIDGE_MODE_VEB:
9234 		/* enable Tx loopback for internal VF/PF communication */
9235 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9236 				IXGBE_PFDTXGSWC_VT_LBEN);
9237 
9238 		/* disable Rx switching replication unless we have SR-IOV
9239 		 * virtual functions
9240 		 */
9241 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9242 		if (!adapter->num_vfs)
9243 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9244 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9245 
9246 		/* disable Rx source address pruning, since we don't expect to
9247 		 * be receiving external loopback of our transmitted frames.
9248 		 */
9249 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9250 		for (p = 0; p < num_pools; p++) {
9251 			if (hw->mac.ops.set_source_address_pruning)
9252 				hw->mac.ops.set_source_address_pruning(hw,
9253 								       false,
9254 								       p);
9255 		}
9256 		break;
9257 	default:
9258 		return -EINVAL;
9259 	}
9260 
9261 	adapter->bridge_mode = mode;
9262 
9263 	e_info(drv, "enabling bridge mode: %s\n",
9264 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9265 
9266 	return 0;
9267 }
9268 
9269 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9270 				    struct nlmsghdr *nlh, u16 flags)
9271 {
9272 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9273 	struct nlattr *attr, *br_spec;
9274 	int rem;
9275 
9276 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9277 		return -EOPNOTSUPP;
9278 
9279 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9280 	if (!br_spec)
9281 		return -EINVAL;
9282 
9283 	nla_for_each_nested(attr, br_spec, rem) {
9284 		int status;
9285 		__u16 mode;
9286 
9287 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9288 			continue;
9289 
9290 		if (nla_len(attr) < sizeof(mode))
9291 			return -EINVAL;
9292 
9293 		mode = nla_get_u16(attr);
9294 		status = ixgbe_configure_bridge_mode(adapter, mode);
9295 		if (status)
9296 			return status;
9297 
9298 		break;
9299 	}
9300 
9301 	return 0;
9302 }
9303 
9304 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9305 				    struct net_device *dev,
9306 				    u32 filter_mask, int nlflags)
9307 {
9308 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9309 
9310 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9311 		return 0;
9312 
9313 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9314 				       adapter->bridge_mode, 0, 0, nlflags,
9315 				       filter_mask, NULL);
9316 }
9317 
9318 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9319 {
9320 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
9321 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9322 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
9323 	unsigned int limit;
9324 	int pool, err;
9325 
9326 	/* Hardware has a limited number of available pools. Each VF, and the
9327 	 * PF require a pool. Check to ensure we don't attempt to use more
9328 	 * then the available number of pools.
9329 	 */
9330 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9331 		return ERR_PTR(-EINVAL);
9332 
9333 #ifdef CONFIG_RPS
9334 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
9335 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
9336 			    vdev->name);
9337 		return ERR_PTR(-EINVAL);
9338 	}
9339 #endif
9340 	/* Check for hardware restriction on number of rx/tx queues */
9341 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
9342 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
9343 		netdev_info(pdev,
9344 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
9345 			    pdev->name);
9346 		return ERR_PTR(-EINVAL);
9347 	}
9348 
9349 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9350 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
9351 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
9352 		return ERR_PTR(-EBUSY);
9353 
9354 	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
9355 	if (!fwd_adapter)
9356 		return ERR_PTR(-ENOMEM);
9357 
9358 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
9359 	adapter->num_rx_pools++;
9360 	set_bit(pool, &adapter->fwd_bitmask);
9361 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9362 
9363 	/* Enable VMDq flag so device will be set in VM mode */
9364 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
9365 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9366 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
9367 
9368 	/* Force reinit of ring allocation with VMDQ enabled */
9369 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9370 	if (err)
9371 		goto fwd_add_err;
9372 	fwd_adapter->pool = pool;
9373 	fwd_adapter->real_adapter = adapter;
9374 
9375 	if (netif_running(pdev)) {
9376 		err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
9377 		if (err)
9378 			goto fwd_add_err;
9379 		netif_tx_start_all_queues(vdev);
9380 	}
9381 
9382 	return fwd_adapter;
9383 fwd_add_err:
9384 	/* unwind counter and free adapter struct */
9385 	netdev_info(pdev,
9386 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
9387 	clear_bit(pool, &adapter->fwd_bitmask);
9388 	adapter->num_rx_pools--;
9389 	kfree(fwd_adapter);
9390 	return ERR_PTR(err);
9391 }
9392 
9393 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9394 {
9395 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
9396 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
9397 	unsigned int limit;
9398 
9399 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
9400 	adapter->num_rx_pools--;
9401 
9402 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9403 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9404 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
9405 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9406 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
9407 		   fwd_adapter->pool, adapter->num_rx_pools,
9408 		   fwd_adapter->rx_base_queue,
9409 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
9410 		   adapter->fwd_bitmask);
9411 	kfree(fwd_adapter);
9412 }
9413 
9414 #define IXGBE_MAX_MAC_HDR_LEN		127
9415 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9416 
9417 static netdev_features_t
9418 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9419 		     netdev_features_t features)
9420 {
9421 	unsigned int network_hdr_len, mac_hdr_len;
9422 
9423 	/* Make certain the headers can be described by a context descriptor */
9424 	mac_hdr_len = skb_network_header(skb) - skb->data;
9425 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9426 		return features & ~(NETIF_F_HW_CSUM |
9427 				    NETIF_F_SCTP_CRC |
9428 				    NETIF_F_HW_VLAN_CTAG_TX |
9429 				    NETIF_F_TSO |
9430 				    NETIF_F_TSO6);
9431 
9432 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9433 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9434 		return features & ~(NETIF_F_HW_CSUM |
9435 				    NETIF_F_SCTP_CRC |
9436 				    NETIF_F_TSO |
9437 				    NETIF_F_TSO6);
9438 
9439 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9440 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9441 	 */
9442 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
9443 		features &= ~NETIF_F_TSO;
9444 
9445 	return features;
9446 }
9447 
9448 static const struct net_device_ops ixgbe_netdev_ops = {
9449 	.ndo_open		= ixgbe_open,
9450 	.ndo_stop		= ixgbe_close,
9451 	.ndo_start_xmit		= ixgbe_xmit_frame,
9452 	.ndo_select_queue	= ixgbe_select_queue,
9453 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
9454 	.ndo_validate_addr	= eth_validate_addr,
9455 	.ndo_set_mac_address	= ixgbe_set_mac,
9456 	.ndo_change_mtu		= ixgbe_change_mtu,
9457 	.ndo_tx_timeout		= ixgbe_tx_timeout,
9458 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
9459 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
9460 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
9461 	.ndo_do_ioctl		= ixgbe_ioctl,
9462 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
9463 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
9464 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
9465 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
9466 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
9467 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
9468 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
9469 	.ndo_get_stats64	= ixgbe_get_stats64,
9470 	.ndo_setup_tc		= __ixgbe_setup_tc,
9471 #ifdef CONFIG_NET_POLL_CONTROLLER
9472 	.ndo_poll_controller	= ixgbe_netpoll,
9473 #endif
9474 #ifdef IXGBE_FCOE
9475 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
9476 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
9477 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
9478 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
9479 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
9480 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
9481 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
9482 #endif /* IXGBE_FCOE */
9483 	.ndo_set_features = ixgbe_set_features,
9484 	.ndo_fix_features = ixgbe_fix_features,
9485 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
9486 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
9487 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
9488 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
9489 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
9490 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
9491 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
9492 	.ndo_features_check	= ixgbe_features_check,
9493 };
9494 
9495 /**
9496  * ixgbe_enumerate_functions - Get the number of ports this device has
9497  * @adapter: adapter structure
9498  *
9499  * This function enumerates the phsyical functions co-located on a single slot,
9500  * in order to determine how many ports a device has. This is most useful in
9501  * determining the required GT/s of PCIe bandwidth necessary for optimal
9502  * performance.
9503  **/
9504 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
9505 {
9506 	struct pci_dev *entry, *pdev = adapter->pdev;
9507 	int physfns = 0;
9508 
9509 	/* Some cards can not use the generic count PCIe functions method,
9510 	 * because they are behind a parent switch, so we hardcode these with
9511 	 * the correct number of functions.
9512 	 */
9513 	if (ixgbe_pcie_from_parent(&adapter->hw))
9514 		physfns = 4;
9515 
9516 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
9517 		/* don't count virtual functions */
9518 		if (entry->is_virtfn)
9519 			continue;
9520 
9521 		/* When the devices on the bus don't all match our device ID,
9522 		 * we can't reliably determine the correct number of
9523 		 * functions. This can occur if a function has been direct
9524 		 * attached to a virtual machine using VT-d, for example. In
9525 		 * this case, simply return -1 to indicate this.
9526 		 */
9527 		if ((entry->vendor != pdev->vendor) ||
9528 		    (entry->device != pdev->device))
9529 			return -1;
9530 
9531 		physfns++;
9532 	}
9533 
9534 	return physfns;
9535 }
9536 
9537 /**
9538  * ixgbe_wol_supported - Check whether device supports WoL
9539  * @adapter: the adapter private structure
9540  * @device_id: the device ID
9541  * @subdev_id: the subsystem device ID
9542  *
9543  * This function is used by probe and ethtool to determine
9544  * which devices have WoL support
9545  *
9546  **/
9547 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
9548 			 u16 subdevice_id)
9549 {
9550 	struct ixgbe_hw *hw = &adapter->hw;
9551 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
9552 
9553 	/* WOL not supported on 82598 */
9554 	if (hw->mac.type == ixgbe_mac_82598EB)
9555 		return false;
9556 
9557 	/* check eeprom to see if WOL is enabled for X540 and newer */
9558 	if (hw->mac.type >= ixgbe_mac_X540) {
9559 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
9560 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
9561 		     (hw->bus.func == 0)))
9562 			return true;
9563 	}
9564 
9565 	/* WOL is determined based on device IDs for 82599 MACs */
9566 	switch (device_id) {
9567 	case IXGBE_DEV_ID_82599_SFP:
9568 		/* Only these subdevices could supports WOL */
9569 		switch (subdevice_id) {
9570 		case IXGBE_SUBDEV_ID_82599_560FLR:
9571 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
9572 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
9573 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
9574 			/* only support first port */
9575 			if (hw->bus.func != 0)
9576 				break;
9577 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
9578 		case IXGBE_SUBDEV_ID_82599_SFP:
9579 		case IXGBE_SUBDEV_ID_82599_RNDC:
9580 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
9581 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
9582 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
9583 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
9584 			return true;
9585 		}
9586 		break;
9587 	case IXGBE_DEV_ID_82599EN_SFP:
9588 		/* Only these subdevices support WOL */
9589 		switch (subdevice_id) {
9590 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
9591 			return true;
9592 		}
9593 		break;
9594 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
9595 		/* All except this subdevice support WOL */
9596 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
9597 			return true;
9598 		break;
9599 	case IXGBE_DEV_ID_82599_KX4:
9600 		return  true;
9601 	default:
9602 		break;
9603 	}
9604 
9605 	return false;
9606 }
9607 
9608 /**
9609  * ixgbe_probe - Device Initialization Routine
9610  * @pdev: PCI device information struct
9611  * @ent: entry in ixgbe_pci_tbl
9612  *
9613  * Returns 0 on success, negative on failure
9614  *
9615  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
9616  * The OS initialization, configuring of the adapter private structure,
9617  * and a hardware reset occur.
9618  **/
9619 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
9620 {
9621 	struct net_device *netdev;
9622 	struct ixgbe_adapter *adapter = NULL;
9623 	struct ixgbe_hw *hw;
9624 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
9625 	int i, err, pci_using_dac, expected_gts;
9626 	unsigned int indices = MAX_TX_QUEUES;
9627 	u8 part_str[IXGBE_PBANUM_LENGTH];
9628 	bool disable_dev = false;
9629 #ifdef IXGBE_FCOE
9630 	u16 device_caps;
9631 #endif
9632 	u32 eec;
9633 
9634 	/* Catch broken hardware that put the wrong VF device ID in
9635 	 * the PCIe SR-IOV capability.
9636 	 */
9637 	if (pdev->is_virtfn) {
9638 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
9639 		     pci_name(pdev), pdev->vendor, pdev->device);
9640 		return -EINVAL;
9641 	}
9642 
9643 	err = pci_enable_device_mem(pdev);
9644 	if (err)
9645 		return err;
9646 
9647 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
9648 		pci_using_dac = 1;
9649 	} else {
9650 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
9651 		if (err) {
9652 			dev_err(&pdev->dev,
9653 				"No usable DMA configuration, aborting\n");
9654 			goto err_dma;
9655 		}
9656 		pci_using_dac = 0;
9657 	}
9658 
9659 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
9660 	if (err) {
9661 		dev_err(&pdev->dev,
9662 			"pci_request_selected_regions failed 0x%x\n", err);
9663 		goto err_pci_reg;
9664 	}
9665 
9666 	pci_enable_pcie_error_reporting(pdev);
9667 
9668 	pci_set_master(pdev);
9669 	pci_save_state(pdev);
9670 
9671 	if (ii->mac == ixgbe_mac_82598EB) {
9672 #ifdef CONFIG_IXGBE_DCB
9673 		/* 8 TC w/ 4 queues per TC */
9674 		indices = 4 * MAX_TRAFFIC_CLASS;
9675 #else
9676 		indices = IXGBE_MAX_RSS_INDICES;
9677 #endif
9678 	}
9679 
9680 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
9681 	if (!netdev) {
9682 		err = -ENOMEM;
9683 		goto err_alloc_etherdev;
9684 	}
9685 
9686 	SET_NETDEV_DEV(netdev, &pdev->dev);
9687 
9688 	adapter = netdev_priv(netdev);
9689 
9690 	adapter->netdev = netdev;
9691 	adapter->pdev = pdev;
9692 	hw = &adapter->hw;
9693 	hw->back = adapter;
9694 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
9695 
9696 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
9697 			      pci_resource_len(pdev, 0));
9698 	adapter->io_addr = hw->hw_addr;
9699 	if (!hw->hw_addr) {
9700 		err = -EIO;
9701 		goto err_ioremap;
9702 	}
9703 
9704 	netdev->netdev_ops = &ixgbe_netdev_ops;
9705 	ixgbe_set_ethtool_ops(netdev);
9706 	netdev->watchdog_timeo = 5 * HZ;
9707 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
9708 
9709 	/* Setup hw api */
9710 	hw->mac.ops   = *ii->mac_ops;
9711 	hw->mac.type  = ii->mac;
9712 	hw->mvals     = ii->mvals;
9713 	if (ii->link_ops)
9714 		hw->link.ops  = *ii->link_ops;
9715 
9716 	/* EEPROM */
9717 	hw->eeprom.ops = *ii->eeprom_ops;
9718 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
9719 	if (ixgbe_removed(hw->hw_addr)) {
9720 		err = -EIO;
9721 		goto err_ioremap;
9722 	}
9723 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
9724 	if (!(eec & BIT(8)))
9725 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
9726 
9727 	/* PHY */
9728 	hw->phy.ops = *ii->phy_ops;
9729 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
9730 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
9731 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
9732 	hw->phy.mdio.mmds = 0;
9733 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
9734 	hw->phy.mdio.dev = netdev;
9735 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
9736 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
9737 
9738 	/* setup the private structure */
9739 	err = ixgbe_sw_init(adapter, ii);
9740 	if (err)
9741 		goto err_sw_init;
9742 
9743 	/* Make sure the SWFW semaphore is in a valid state */
9744 	if (hw->mac.ops.init_swfw_sync)
9745 		hw->mac.ops.init_swfw_sync(hw);
9746 
9747 	/* Make it possible the adapter to be woken up via WOL */
9748 	switch (adapter->hw.mac.type) {
9749 	case ixgbe_mac_82599EB:
9750 	case ixgbe_mac_X540:
9751 	case ixgbe_mac_X550:
9752 	case ixgbe_mac_X550EM_x:
9753 	case ixgbe_mac_x550em_a:
9754 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
9755 		break;
9756 	default:
9757 		break;
9758 	}
9759 
9760 	/*
9761 	 * If there is a fan on this device and it has failed log the
9762 	 * failure.
9763 	 */
9764 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
9765 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
9766 		if (esdp & IXGBE_ESDP_SDP1)
9767 			e_crit(probe, "Fan has stopped, replace the adapter\n");
9768 	}
9769 
9770 	if (allow_unsupported_sfp)
9771 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
9772 
9773 	/* reset_hw fills in the perm_addr as well */
9774 	hw->phy.reset_if_overtemp = true;
9775 	err = hw->mac.ops.reset_hw(hw);
9776 	hw->phy.reset_if_overtemp = false;
9777 	ixgbe_set_eee_capable(adapter);
9778 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
9779 		err = 0;
9780 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
9781 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
9782 		e_dev_err("Reload the driver after installing a supported module.\n");
9783 		goto err_sw_init;
9784 	} else if (err) {
9785 		e_dev_err("HW Init failed: %d\n", err);
9786 		goto err_sw_init;
9787 	}
9788 
9789 #ifdef CONFIG_PCI_IOV
9790 	/* SR-IOV not supported on the 82598 */
9791 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9792 		goto skip_sriov;
9793 	/* Mailbox */
9794 	ixgbe_init_mbx_params_pf(hw);
9795 	hw->mbx.ops = ii->mbx_ops;
9796 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
9797 	ixgbe_enable_sriov(adapter);
9798 skip_sriov:
9799 
9800 #endif
9801 	netdev->features = NETIF_F_SG |
9802 			   NETIF_F_TSO |
9803 			   NETIF_F_TSO6 |
9804 			   NETIF_F_RXHASH |
9805 			   NETIF_F_RXCSUM |
9806 			   NETIF_F_HW_CSUM;
9807 
9808 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
9809 				    NETIF_F_GSO_GRE_CSUM | \
9810 				    NETIF_F_GSO_IPXIP4 | \
9811 				    NETIF_F_GSO_IPXIP6 | \
9812 				    NETIF_F_GSO_UDP_TUNNEL | \
9813 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
9814 
9815 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
9816 	netdev->features |= NETIF_F_GSO_PARTIAL |
9817 			    IXGBE_GSO_PARTIAL_FEATURES;
9818 
9819 	if (hw->mac.type >= ixgbe_mac_82599EB)
9820 		netdev->features |= NETIF_F_SCTP_CRC;
9821 
9822 	/* copy netdev features into list of user selectable features */
9823 	netdev->hw_features |= netdev->features |
9824 			       NETIF_F_HW_VLAN_CTAG_FILTER |
9825 			       NETIF_F_HW_VLAN_CTAG_RX |
9826 			       NETIF_F_HW_VLAN_CTAG_TX |
9827 			       NETIF_F_RXALL |
9828 			       NETIF_F_HW_L2FW_DOFFLOAD;
9829 
9830 	if (hw->mac.type >= ixgbe_mac_82599EB)
9831 		netdev->hw_features |= NETIF_F_NTUPLE |
9832 				       NETIF_F_HW_TC;
9833 
9834 	if (pci_using_dac)
9835 		netdev->features |= NETIF_F_HIGHDMA;
9836 
9837 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
9838 	netdev->hw_enc_features |= netdev->vlan_features;
9839 	netdev->mpls_features |= NETIF_F_HW_CSUM;
9840 
9841 	/* set this bit last since it cannot be part of vlan_features */
9842 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
9843 			    NETIF_F_HW_VLAN_CTAG_RX |
9844 			    NETIF_F_HW_VLAN_CTAG_TX;
9845 
9846 	netdev->priv_flags |= IFF_UNICAST_FLT;
9847 	netdev->priv_flags |= IFF_SUPP_NOFCS;
9848 
9849 	/* MTU range: 68 - 9710 */
9850 	netdev->min_mtu = ETH_MIN_MTU;
9851 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
9852 
9853 #ifdef CONFIG_IXGBE_DCB
9854 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
9855 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
9856 #endif
9857 
9858 #ifdef IXGBE_FCOE
9859 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
9860 		unsigned int fcoe_l;
9861 
9862 		if (hw->mac.ops.get_device_caps) {
9863 			hw->mac.ops.get_device_caps(hw, &device_caps);
9864 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
9865 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
9866 		}
9867 
9868 
9869 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
9870 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
9871 
9872 		netdev->features |= NETIF_F_FSO |
9873 				    NETIF_F_FCOE_CRC;
9874 
9875 		netdev->vlan_features |= NETIF_F_FSO |
9876 					 NETIF_F_FCOE_CRC |
9877 					 NETIF_F_FCOE_MTU;
9878 	}
9879 #endif /* IXGBE_FCOE */
9880 
9881 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
9882 		netdev->hw_features |= NETIF_F_LRO;
9883 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9884 		netdev->features |= NETIF_F_LRO;
9885 
9886 	/* make sure the EEPROM is good */
9887 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
9888 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
9889 		err = -EIO;
9890 		goto err_sw_init;
9891 	}
9892 
9893 	eth_platform_get_mac_address(&adapter->pdev->dev,
9894 				     adapter->hw.mac.perm_addr);
9895 
9896 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
9897 
9898 	if (!is_valid_ether_addr(netdev->dev_addr)) {
9899 		e_dev_err("invalid MAC address\n");
9900 		err = -EIO;
9901 		goto err_sw_init;
9902 	}
9903 
9904 	/* Set hw->mac.addr to permanent MAC address */
9905 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
9906 	ixgbe_mac_set_default_filter(adapter);
9907 
9908 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
9909 		    (unsigned long) adapter);
9910 
9911 	if (ixgbe_removed(hw->hw_addr)) {
9912 		err = -EIO;
9913 		goto err_sw_init;
9914 	}
9915 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
9916 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
9917 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
9918 
9919 	err = ixgbe_init_interrupt_scheme(adapter);
9920 	if (err)
9921 		goto err_sw_init;
9922 
9923 	/* WOL not supported for all devices */
9924 	adapter->wol = 0;
9925 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
9926 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
9927 						pdev->subsystem_device);
9928 	if (hw->wol_enabled)
9929 		adapter->wol = IXGBE_WUFC_MAG;
9930 
9931 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
9932 
9933 	/* save off EEPROM version number */
9934 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
9935 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
9936 
9937 	/* pick up the PCI bus settings for reporting later */
9938 	if (ixgbe_pcie_from_parent(hw))
9939 		ixgbe_get_parent_bus_info(adapter);
9940 	else
9941 		 hw->mac.ops.get_bus_info(hw);
9942 
9943 	/* calculate the expected PCIe bandwidth required for optimal
9944 	 * performance. Note that some older parts will never have enough
9945 	 * bandwidth due to being older generation PCIe parts. We clamp these
9946 	 * parts to ensure no warning is displayed if it can't be fixed.
9947 	 */
9948 	switch (hw->mac.type) {
9949 	case ixgbe_mac_82598EB:
9950 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
9951 		break;
9952 	default:
9953 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
9954 		break;
9955 	}
9956 
9957 	/* don't check link if we failed to enumerate functions */
9958 	if (expected_gts > 0)
9959 		ixgbe_check_minimum_link(adapter, expected_gts);
9960 
9961 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
9962 	if (err)
9963 		strlcpy(part_str, "Unknown", sizeof(part_str));
9964 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
9965 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
9966 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
9967 			   part_str);
9968 	else
9969 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
9970 			   hw->mac.type, hw->phy.type, part_str);
9971 
9972 	e_dev_info("%pM\n", netdev->dev_addr);
9973 
9974 	/* reset the hardware with the new settings */
9975 	err = hw->mac.ops.start_hw(hw);
9976 	if (err == IXGBE_ERR_EEPROM_VERSION) {
9977 		/* We are running on a pre-production device, log a warning */
9978 		e_dev_warn("This device is a pre-production adapter/LOM. "
9979 			   "Please be aware there may be issues associated "
9980 			   "with your hardware.  If you are experiencing "
9981 			   "problems please contact your Intel or hardware "
9982 			   "representative who provided you with this "
9983 			   "hardware.\n");
9984 	}
9985 	strcpy(netdev->name, "eth%d");
9986 	err = register_netdev(netdev);
9987 	if (err)
9988 		goto err_register;
9989 
9990 	pci_set_drvdata(pdev, adapter);
9991 
9992 	/* power down the optics for 82599 SFP+ fiber */
9993 	if (hw->mac.ops.disable_tx_laser)
9994 		hw->mac.ops.disable_tx_laser(hw);
9995 
9996 	/* carrier off reporting is important to ethtool even BEFORE open */
9997 	netif_carrier_off(netdev);
9998 
9999 #ifdef CONFIG_IXGBE_DCA
10000 	if (dca_add_requester(&pdev->dev) == 0) {
10001 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
10002 		ixgbe_setup_dca(adapter);
10003 	}
10004 #endif
10005 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
10006 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
10007 		for (i = 0; i < adapter->num_vfs; i++)
10008 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
10009 	}
10010 
10011 	/* firmware requires driver version to be 0xFFFFFFFF
10012 	 * since os does not support feature
10013 	 */
10014 	if (hw->mac.ops.set_fw_drv_ver)
10015 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
10016 					   sizeof(ixgbe_driver_version) - 1,
10017 					   ixgbe_driver_version);
10018 
10019 	/* add san mac addr to netdev */
10020 	ixgbe_add_sanmac_netdev(netdev);
10021 
10022 	e_dev_info("%s\n", ixgbe_default_device_descr);
10023 
10024 #ifdef CONFIG_IXGBE_HWMON
10025 	if (ixgbe_sysfs_init(adapter))
10026 		e_err(probe, "failed to allocate sysfs resources\n");
10027 #endif /* CONFIG_IXGBE_HWMON */
10028 
10029 	ixgbe_dbg_adapter_init(adapter);
10030 
10031 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
10032 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
10033 		hw->mac.ops.setup_link(hw,
10034 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
10035 			true);
10036 
10037 	return 0;
10038 
10039 err_register:
10040 	ixgbe_release_hw_control(adapter);
10041 	ixgbe_clear_interrupt_scheme(adapter);
10042 err_sw_init:
10043 	ixgbe_disable_sriov(adapter);
10044 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
10045 	iounmap(adapter->io_addr);
10046 	kfree(adapter->jump_tables[0]);
10047 	kfree(adapter->mac_table);
10048 err_ioremap:
10049 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10050 	free_netdev(netdev);
10051 err_alloc_etherdev:
10052 	pci_release_mem_regions(pdev);
10053 err_pci_reg:
10054 err_dma:
10055 	if (!adapter || disable_dev)
10056 		pci_disable_device(pdev);
10057 	return err;
10058 }
10059 
10060 /**
10061  * ixgbe_remove - Device Removal Routine
10062  * @pdev: PCI device information struct
10063  *
10064  * ixgbe_remove is called by the PCI subsystem to alert the driver
10065  * that it should release a PCI device.  The could be caused by a
10066  * Hot-Plug event, or because the driver is going to be removed from
10067  * memory.
10068  **/
10069 static void ixgbe_remove(struct pci_dev *pdev)
10070 {
10071 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10072 	struct net_device *netdev;
10073 	bool disable_dev;
10074 	int i;
10075 
10076 	/* if !adapter then we already cleaned up in probe */
10077 	if (!adapter)
10078 		return;
10079 
10080 	netdev  = adapter->netdev;
10081 	ixgbe_dbg_adapter_exit(adapter);
10082 
10083 	set_bit(__IXGBE_REMOVING, &adapter->state);
10084 	cancel_work_sync(&adapter->service_task);
10085 
10086 
10087 #ifdef CONFIG_IXGBE_DCA
10088 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
10089 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
10090 		dca_remove_requester(&pdev->dev);
10091 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
10092 				IXGBE_DCA_CTRL_DCA_DISABLE);
10093 	}
10094 
10095 #endif
10096 #ifdef CONFIG_IXGBE_HWMON
10097 	ixgbe_sysfs_exit(adapter);
10098 #endif /* CONFIG_IXGBE_HWMON */
10099 
10100 	/* remove the added san mac */
10101 	ixgbe_del_sanmac_netdev(netdev);
10102 
10103 #ifdef CONFIG_PCI_IOV
10104 	ixgbe_disable_sriov(adapter);
10105 #endif
10106 	if (netdev->reg_state == NETREG_REGISTERED)
10107 		unregister_netdev(netdev);
10108 
10109 	ixgbe_clear_interrupt_scheme(adapter);
10110 
10111 	ixgbe_release_hw_control(adapter);
10112 
10113 #ifdef CONFIG_DCB
10114 	kfree(adapter->ixgbe_ieee_pfc);
10115 	kfree(adapter->ixgbe_ieee_ets);
10116 
10117 #endif
10118 	iounmap(adapter->io_addr);
10119 	pci_release_mem_regions(pdev);
10120 
10121 	e_dev_info("complete\n");
10122 
10123 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
10124 		if (adapter->jump_tables[i]) {
10125 			kfree(adapter->jump_tables[i]->input);
10126 			kfree(adapter->jump_tables[i]->mask);
10127 		}
10128 		kfree(adapter->jump_tables[i]);
10129 	}
10130 
10131 	kfree(adapter->mac_table);
10132 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10133 	free_netdev(netdev);
10134 
10135 	pci_disable_pcie_error_reporting(pdev);
10136 
10137 	if (disable_dev)
10138 		pci_disable_device(pdev);
10139 }
10140 
10141 /**
10142  * ixgbe_io_error_detected - called when PCI error is detected
10143  * @pdev: Pointer to PCI device
10144  * @state: The current pci connection state
10145  *
10146  * This function is called after a PCI bus error affecting
10147  * this device has been detected.
10148  */
10149 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
10150 						pci_channel_state_t state)
10151 {
10152 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10153 	struct net_device *netdev = adapter->netdev;
10154 
10155 #ifdef CONFIG_PCI_IOV
10156 	struct ixgbe_hw *hw = &adapter->hw;
10157 	struct pci_dev *bdev, *vfdev;
10158 	u32 dw0, dw1, dw2, dw3;
10159 	int vf, pos;
10160 	u16 req_id, pf_func;
10161 
10162 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
10163 	    adapter->num_vfs == 0)
10164 		goto skip_bad_vf_detection;
10165 
10166 	bdev = pdev->bus->self;
10167 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
10168 		bdev = bdev->bus->self;
10169 
10170 	if (!bdev)
10171 		goto skip_bad_vf_detection;
10172 
10173 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
10174 	if (!pos)
10175 		goto skip_bad_vf_detection;
10176 
10177 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
10178 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
10179 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
10180 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
10181 	if (ixgbe_removed(hw->hw_addr))
10182 		goto skip_bad_vf_detection;
10183 
10184 	req_id = dw1 >> 16;
10185 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
10186 	if (!(req_id & 0x0080))
10187 		goto skip_bad_vf_detection;
10188 
10189 	pf_func = req_id & 0x01;
10190 	if ((pf_func & 1) == (pdev->devfn & 1)) {
10191 		unsigned int device_id;
10192 
10193 		vf = (req_id & 0x7F) >> 1;
10194 		e_dev_err("VF %d has caused a PCIe error\n", vf);
10195 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
10196 				"%8.8x\tdw3: %8.8x\n",
10197 		dw0, dw1, dw2, dw3);
10198 		switch (adapter->hw.mac.type) {
10199 		case ixgbe_mac_82599EB:
10200 			device_id = IXGBE_82599_VF_DEVICE_ID;
10201 			break;
10202 		case ixgbe_mac_X540:
10203 			device_id = IXGBE_X540_VF_DEVICE_ID;
10204 			break;
10205 		case ixgbe_mac_X550:
10206 			device_id = IXGBE_DEV_ID_X550_VF;
10207 			break;
10208 		case ixgbe_mac_X550EM_x:
10209 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
10210 			break;
10211 		case ixgbe_mac_x550em_a:
10212 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
10213 			break;
10214 		default:
10215 			device_id = 0;
10216 			break;
10217 		}
10218 
10219 		/* Find the pci device of the offending VF */
10220 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
10221 		while (vfdev) {
10222 			if (vfdev->devfn == (req_id & 0xFF))
10223 				break;
10224 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
10225 					       device_id, vfdev);
10226 		}
10227 		/*
10228 		 * There's a slim chance the VF could have been hot plugged,
10229 		 * so if it is no longer present we don't need to issue the
10230 		 * VFLR.  Just clean up the AER in that case.
10231 		 */
10232 		if (vfdev) {
10233 			ixgbe_issue_vf_flr(adapter, vfdev);
10234 			/* Free device reference count */
10235 			pci_dev_put(vfdev);
10236 		}
10237 
10238 		pci_cleanup_aer_uncorrect_error_status(pdev);
10239 	}
10240 
10241 	/*
10242 	 * Even though the error may have occurred on the other port
10243 	 * we still need to increment the vf error reference count for
10244 	 * both ports because the I/O resume function will be called
10245 	 * for both of them.
10246 	 */
10247 	adapter->vferr_refcount++;
10248 
10249 	return PCI_ERS_RESULT_RECOVERED;
10250 
10251 skip_bad_vf_detection:
10252 #endif /* CONFIG_PCI_IOV */
10253 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
10254 		return PCI_ERS_RESULT_DISCONNECT;
10255 
10256 	rtnl_lock();
10257 	netif_device_detach(netdev);
10258 
10259 	if (state == pci_channel_io_perm_failure) {
10260 		rtnl_unlock();
10261 		return PCI_ERS_RESULT_DISCONNECT;
10262 	}
10263 
10264 	if (netif_running(netdev))
10265 		ixgbe_close_suspend(adapter);
10266 
10267 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
10268 		pci_disable_device(pdev);
10269 	rtnl_unlock();
10270 
10271 	/* Request a slot reset. */
10272 	return PCI_ERS_RESULT_NEED_RESET;
10273 }
10274 
10275 /**
10276  * ixgbe_io_slot_reset - called after the pci bus has been reset.
10277  * @pdev: Pointer to PCI device
10278  *
10279  * Restart the card from scratch, as if from a cold-boot.
10280  */
10281 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
10282 {
10283 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10284 	pci_ers_result_t result;
10285 	int err;
10286 
10287 	if (pci_enable_device_mem(pdev)) {
10288 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
10289 		result = PCI_ERS_RESULT_DISCONNECT;
10290 	} else {
10291 		smp_mb__before_atomic();
10292 		clear_bit(__IXGBE_DISABLED, &adapter->state);
10293 		adapter->hw.hw_addr = adapter->io_addr;
10294 		pci_set_master(pdev);
10295 		pci_restore_state(pdev);
10296 		pci_save_state(pdev);
10297 
10298 		pci_wake_from_d3(pdev, false);
10299 
10300 		ixgbe_reset(adapter);
10301 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10302 		result = PCI_ERS_RESULT_RECOVERED;
10303 	}
10304 
10305 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10306 	if (err) {
10307 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
10308 			  "failed 0x%0x\n", err);
10309 		/* non-fatal, continue */
10310 	}
10311 
10312 	return result;
10313 }
10314 
10315 /**
10316  * ixgbe_io_resume - called when traffic can start flowing again.
10317  * @pdev: Pointer to PCI device
10318  *
10319  * This callback is called when the error recovery driver tells us that
10320  * its OK to resume normal operation.
10321  */
10322 static void ixgbe_io_resume(struct pci_dev *pdev)
10323 {
10324 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10325 	struct net_device *netdev = adapter->netdev;
10326 
10327 #ifdef CONFIG_PCI_IOV
10328 	if (adapter->vferr_refcount) {
10329 		e_info(drv, "Resuming after VF err\n");
10330 		adapter->vferr_refcount--;
10331 		return;
10332 	}
10333 
10334 #endif
10335 	rtnl_lock();
10336 	if (netif_running(netdev))
10337 		ixgbe_open(netdev);
10338 
10339 	netif_device_attach(netdev);
10340 	rtnl_unlock();
10341 }
10342 
10343 static const struct pci_error_handlers ixgbe_err_handler = {
10344 	.error_detected = ixgbe_io_error_detected,
10345 	.slot_reset = ixgbe_io_slot_reset,
10346 	.resume = ixgbe_io_resume,
10347 };
10348 
10349 static struct pci_driver ixgbe_driver = {
10350 	.name     = ixgbe_driver_name,
10351 	.id_table = ixgbe_pci_tbl,
10352 	.probe    = ixgbe_probe,
10353 	.remove   = ixgbe_remove,
10354 #ifdef CONFIG_PM
10355 	.suspend  = ixgbe_suspend,
10356 	.resume   = ixgbe_resume,
10357 #endif
10358 	.shutdown = ixgbe_shutdown,
10359 	.sriov_configure = ixgbe_pci_sriov_configure,
10360 	.err_handler = &ixgbe_err_handler
10361 };
10362 
10363 /**
10364  * ixgbe_init_module - Driver Registration Routine
10365  *
10366  * ixgbe_init_module is the first routine called when the driver is
10367  * loaded. All it does is register with the PCI subsystem.
10368  **/
10369 static int __init ixgbe_init_module(void)
10370 {
10371 	int ret;
10372 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
10373 	pr_info("%s\n", ixgbe_copyright);
10374 
10375 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
10376 	if (!ixgbe_wq) {
10377 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
10378 		return -ENOMEM;
10379 	}
10380 
10381 	ixgbe_dbg_init();
10382 
10383 	ret = pci_register_driver(&ixgbe_driver);
10384 	if (ret) {
10385 		destroy_workqueue(ixgbe_wq);
10386 		ixgbe_dbg_exit();
10387 		return ret;
10388 	}
10389 
10390 #ifdef CONFIG_IXGBE_DCA
10391 	dca_register_notify(&dca_notifier);
10392 #endif
10393 
10394 	return 0;
10395 }
10396 
10397 module_init(ixgbe_init_module);
10398 
10399 /**
10400  * ixgbe_exit_module - Driver Exit Cleanup Routine
10401  *
10402  * ixgbe_exit_module is called just before the driver is removed
10403  * from memory.
10404  **/
10405 static void __exit ixgbe_exit_module(void)
10406 {
10407 #ifdef CONFIG_IXGBE_DCA
10408 	dca_unregister_notify(&dca_notifier);
10409 #endif
10410 	pci_unregister_driver(&ixgbe_driver);
10411 
10412 	ixgbe_dbg_exit();
10413 	if (ixgbe_wq) {
10414 		destroy_workqueue(ixgbe_wq);
10415 		ixgbe_wq = NULL;
10416 	}
10417 }
10418 
10419 #ifdef CONFIG_IXGBE_DCA
10420 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
10421 			    void *p)
10422 {
10423 	int ret_val;
10424 
10425 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
10426 					 __ixgbe_notify_dca);
10427 
10428 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
10429 }
10430 
10431 #endif /* CONFIG_IXGBE_DCA */
10432 
10433 module_exit(ixgbe_exit_module);
10434 
10435 /* ixgbe_main.c */
10436