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 
58 #include "ixgbe.h"
59 #include "ixgbe_common.h"
60 #include "ixgbe_dcb_82599.h"
61 #include "ixgbe_sriov.h"
62 #include "ixgbe_model.h"
63 
64 char ixgbe_driver_name[] = "ixgbe";
65 static const char ixgbe_driver_string[] =
66 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
67 #ifdef IXGBE_FCOE
68 char ixgbe_default_device_descr[] =
69 			      "Intel(R) 10 Gigabit Network Connection";
70 #else
71 static char ixgbe_default_device_descr[] =
72 			      "Intel(R) 10 Gigabit Network Connection";
73 #endif
74 #define DRV_VERSION "4.4.0-k"
75 const char ixgbe_driver_version[] = DRV_VERSION;
76 static const char ixgbe_copyright[] =
77 				"Copyright (c) 1999-2016 Intel Corporation.";
78 
79 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";
80 
81 static const struct ixgbe_info *ixgbe_info_tbl[] = {
82 	[board_82598]		= &ixgbe_82598_info,
83 	[board_82599]		= &ixgbe_82599_info,
84 	[board_X540]		= &ixgbe_X540_info,
85 	[board_X550]		= &ixgbe_X550_info,
86 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
87 	[board_x550em_a]	= &ixgbe_x550em_a_info,
88 };
89 
90 /* ixgbe_pci_tbl - PCI Device ID Table
91  *
92  * Wildcard entries (PCI_ANY_ID) should come last
93  * Last entry must be all 0s
94  *
95  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
96  *   Class, Class Mask, private data (not used) }
97  */
98 static const struct pci_device_id ixgbe_pci_tbl[] = {
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
133 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
134 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
135 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
136 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
137 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
138 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
139 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
140 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
141 	/* required last entry */
142 	{0, }
143 };
144 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
145 
146 #ifdef CONFIG_IXGBE_DCA
147 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
148 			    void *p);
149 static struct notifier_block dca_notifier = {
150 	.notifier_call = ixgbe_notify_dca,
151 	.next          = NULL,
152 	.priority      = 0
153 };
154 #endif
155 
156 #ifdef CONFIG_PCI_IOV
157 static unsigned int max_vfs;
158 module_param(max_vfs, uint, 0);
159 MODULE_PARM_DESC(max_vfs,
160 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
161 #endif /* CONFIG_PCI_IOV */
162 
163 static unsigned int allow_unsupported_sfp;
164 module_param(allow_unsupported_sfp, uint, 0);
165 MODULE_PARM_DESC(allow_unsupported_sfp,
166 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
167 
168 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
169 static int debug = -1;
170 module_param(debug, int, 0);
171 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
172 
173 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
174 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
175 MODULE_LICENSE("GPL");
176 MODULE_VERSION(DRV_VERSION);
177 
178 static struct workqueue_struct *ixgbe_wq;
179 
180 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
181 
182 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
183 					  u32 reg, u16 *value)
184 {
185 	struct pci_dev *parent_dev;
186 	struct pci_bus *parent_bus;
187 
188 	parent_bus = adapter->pdev->bus->parent;
189 	if (!parent_bus)
190 		return -1;
191 
192 	parent_dev = parent_bus->self;
193 	if (!parent_dev)
194 		return -1;
195 
196 	if (!pci_is_pcie(parent_dev))
197 		return -1;
198 
199 	pcie_capability_read_word(parent_dev, reg, value);
200 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
201 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
202 		return -1;
203 	return 0;
204 }
205 
206 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
207 {
208 	struct ixgbe_hw *hw = &adapter->hw;
209 	u16 link_status = 0;
210 	int err;
211 
212 	hw->bus.type = ixgbe_bus_type_pci_express;
213 
214 	/* Get the negotiated link width and speed from PCI config space of the
215 	 * parent, as this device is behind a switch
216 	 */
217 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
218 
219 	/* assume caller will handle error case */
220 	if (err)
221 		return err;
222 
223 	hw->bus.width = ixgbe_convert_bus_width(link_status);
224 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
225 
226 	return 0;
227 }
228 
229 /**
230  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
231  * @hw: hw specific details
232  *
233  * This function is used by probe to determine whether a device's PCI-Express
234  * bandwidth details should be gathered from the parent bus instead of from the
235  * device. Used to ensure that various locations all have the correct device ID
236  * checks.
237  */
238 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
239 {
240 	switch (hw->device_id) {
241 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
242 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
243 		return true;
244 	default:
245 		return false;
246 	}
247 }
248 
249 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
250 				     int expected_gts)
251 {
252 	struct ixgbe_hw *hw = &adapter->hw;
253 	int max_gts = 0;
254 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
255 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
256 	struct pci_dev *pdev;
257 
258 	/* Some devices are not connected over PCIe and thus do not negotiate
259 	 * speed. These devices do not have valid bus info, and thus any report
260 	 * we generate may not be correct.
261 	 */
262 	if (hw->bus.type == ixgbe_bus_type_internal)
263 		return;
264 
265 	/* determine whether to use the parent device */
266 	if (ixgbe_pcie_from_parent(&adapter->hw))
267 		pdev = adapter->pdev->bus->parent->self;
268 	else
269 		pdev = adapter->pdev;
270 
271 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
272 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
273 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
274 		return;
275 	}
276 
277 	switch (speed) {
278 	case PCIE_SPEED_2_5GT:
279 		/* 8b/10b encoding reduces max throughput by 20% */
280 		max_gts = 2 * width;
281 		break;
282 	case PCIE_SPEED_5_0GT:
283 		/* 8b/10b encoding reduces max throughput by 20% */
284 		max_gts = 4 * width;
285 		break;
286 	case PCIE_SPEED_8_0GT:
287 		/* 128b/130b encoding reduces throughput by less than 2% */
288 		max_gts = 8 * width;
289 		break;
290 	default:
291 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
292 		return;
293 	}
294 
295 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
296 		   max_gts);
297 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
298 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
299 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
300 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
301 		    "Unknown"),
302 		   width,
303 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
304 		    speed == PCIE_SPEED_5_0GT ? "20%" :
305 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
306 		    "Unknown"));
307 
308 	if (max_gts < expected_gts) {
309 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
310 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
311 			expected_gts);
312 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
313 	}
314 }
315 
316 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
317 {
318 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
319 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
320 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
321 		queue_work(ixgbe_wq, &adapter->service_task);
322 }
323 
324 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
325 {
326 	struct ixgbe_adapter *adapter = hw->back;
327 
328 	if (!hw->hw_addr)
329 		return;
330 	hw->hw_addr = NULL;
331 	e_dev_err("Adapter removed\n");
332 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
333 		ixgbe_service_event_schedule(adapter);
334 }
335 
336 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
337 {
338 	u32 value;
339 
340 	/* The following check not only optimizes a bit by not
341 	 * performing a read on the status register when the
342 	 * register just read was a status register read that
343 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
344 	 * potential recursion.
345 	 */
346 	if (reg == IXGBE_STATUS) {
347 		ixgbe_remove_adapter(hw);
348 		return;
349 	}
350 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
351 	if (value == IXGBE_FAILED_READ_REG)
352 		ixgbe_remove_adapter(hw);
353 }
354 
355 /**
356  * ixgbe_read_reg - Read from device register
357  * @hw: hw specific details
358  * @reg: offset of register to read
359  *
360  * Returns : value read or IXGBE_FAILED_READ_REG if removed
361  *
362  * This function is used to read device registers. It checks for device
363  * removal by confirming any read that returns all ones by checking the
364  * status register value for all ones. This function avoids reading from
365  * the hardware if a removal was previously detected in which case it
366  * returns IXGBE_FAILED_READ_REG (all ones).
367  */
368 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
369 {
370 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
371 	u32 value;
372 
373 	if (ixgbe_removed(reg_addr))
374 		return IXGBE_FAILED_READ_REG;
375 	if (unlikely(hw->phy.nw_mng_if_sel &
376 		     IXGBE_NW_MNG_IF_SEL_ENABLE_10_100M)) {
377 		struct ixgbe_adapter *adapter;
378 		int i;
379 
380 		for (i = 0; i < 200; ++i) {
381 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
382 			if (likely(!value))
383 				goto writes_completed;
384 			if (value == IXGBE_FAILED_READ_REG) {
385 				ixgbe_remove_adapter(hw);
386 				return IXGBE_FAILED_READ_REG;
387 			}
388 			udelay(5);
389 		}
390 
391 		adapter = hw->back;
392 		e_warn(hw, "register writes incomplete %08x\n", value);
393 	}
394 
395 writes_completed:
396 	value = readl(reg_addr + reg);
397 	if (unlikely(value == IXGBE_FAILED_READ_REG))
398 		ixgbe_check_remove(hw, reg);
399 	return value;
400 }
401 
402 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
403 {
404 	u16 value;
405 
406 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
407 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
408 		ixgbe_remove_adapter(hw);
409 		return true;
410 	}
411 	return false;
412 }
413 
414 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
415 {
416 	struct ixgbe_adapter *adapter = hw->back;
417 	u16 value;
418 
419 	if (ixgbe_removed(hw->hw_addr))
420 		return IXGBE_FAILED_READ_CFG_WORD;
421 	pci_read_config_word(adapter->pdev, reg, &value);
422 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
423 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
424 		return IXGBE_FAILED_READ_CFG_WORD;
425 	return value;
426 }
427 
428 #ifdef CONFIG_PCI_IOV
429 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
430 {
431 	struct ixgbe_adapter *adapter = hw->back;
432 	u32 value;
433 
434 	if (ixgbe_removed(hw->hw_addr))
435 		return IXGBE_FAILED_READ_CFG_DWORD;
436 	pci_read_config_dword(adapter->pdev, reg, &value);
437 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
438 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
439 		return IXGBE_FAILED_READ_CFG_DWORD;
440 	return value;
441 }
442 #endif /* CONFIG_PCI_IOV */
443 
444 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
445 {
446 	struct ixgbe_adapter *adapter = hw->back;
447 
448 	if (ixgbe_removed(hw->hw_addr))
449 		return;
450 	pci_write_config_word(adapter->pdev, reg, value);
451 }
452 
453 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
454 {
455 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
456 
457 	/* flush memory to make sure state is correct before next watchdog */
458 	smp_mb__before_atomic();
459 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
460 }
461 
462 struct ixgbe_reg_info {
463 	u32 ofs;
464 	char *name;
465 };
466 
467 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
468 
469 	/* General Registers */
470 	{IXGBE_CTRL, "CTRL"},
471 	{IXGBE_STATUS, "STATUS"},
472 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
473 
474 	/* Interrupt Registers */
475 	{IXGBE_EICR, "EICR"},
476 
477 	/* RX Registers */
478 	{IXGBE_SRRCTL(0), "SRRCTL"},
479 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
480 	{IXGBE_RDLEN(0), "RDLEN"},
481 	{IXGBE_RDH(0), "RDH"},
482 	{IXGBE_RDT(0), "RDT"},
483 	{IXGBE_RXDCTL(0), "RXDCTL"},
484 	{IXGBE_RDBAL(0), "RDBAL"},
485 	{IXGBE_RDBAH(0), "RDBAH"},
486 
487 	/* TX Registers */
488 	{IXGBE_TDBAL(0), "TDBAL"},
489 	{IXGBE_TDBAH(0), "TDBAH"},
490 	{IXGBE_TDLEN(0), "TDLEN"},
491 	{IXGBE_TDH(0), "TDH"},
492 	{IXGBE_TDT(0), "TDT"},
493 	{IXGBE_TXDCTL(0), "TXDCTL"},
494 
495 	/* List Terminator */
496 	{ .name = NULL }
497 };
498 
499 
500 /*
501  * ixgbe_regdump - register printout routine
502  */
503 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
504 {
505 	int i = 0, j = 0;
506 	char rname[16];
507 	u32 regs[64];
508 
509 	switch (reginfo->ofs) {
510 	case IXGBE_SRRCTL(0):
511 		for (i = 0; i < 64; i++)
512 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
513 		break;
514 	case IXGBE_DCA_RXCTRL(0):
515 		for (i = 0; i < 64; i++)
516 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
517 		break;
518 	case IXGBE_RDLEN(0):
519 		for (i = 0; i < 64; i++)
520 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
521 		break;
522 	case IXGBE_RDH(0):
523 		for (i = 0; i < 64; i++)
524 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
525 		break;
526 	case IXGBE_RDT(0):
527 		for (i = 0; i < 64; i++)
528 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
529 		break;
530 	case IXGBE_RXDCTL(0):
531 		for (i = 0; i < 64; i++)
532 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
533 		break;
534 	case IXGBE_RDBAL(0):
535 		for (i = 0; i < 64; i++)
536 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
537 		break;
538 	case IXGBE_RDBAH(0):
539 		for (i = 0; i < 64; i++)
540 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
541 		break;
542 	case IXGBE_TDBAL(0):
543 		for (i = 0; i < 64; i++)
544 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
545 		break;
546 	case IXGBE_TDBAH(0):
547 		for (i = 0; i < 64; i++)
548 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
549 		break;
550 	case IXGBE_TDLEN(0):
551 		for (i = 0; i < 64; i++)
552 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
553 		break;
554 	case IXGBE_TDH(0):
555 		for (i = 0; i < 64; i++)
556 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
557 		break;
558 	case IXGBE_TDT(0):
559 		for (i = 0; i < 64; i++)
560 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
561 		break;
562 	case IXGBE_TXDCTL(0):
563 		for (i = 0; i < 64; i++)
564 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
565 		break;
566 	default:
567 		pr_info("%-15s %08x\n", reginfo->name,
568 			IXGBE_READ_REG(hw, reginfo->ofs));
569 		return;
570 	}
571 
572 	for (i = 0; i < 8; i++) {
573 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
574 		pr_err("%-15s", rname);
575 		for (j = 0; j < 8; j++)
576 			pr_cont(" %08x", regs[i*8+j]);
577 		pr_cont("\n");
578 	}
579 
580 }
581 
582 /*
583  * ixgbe_dump - Print registers, tx-rings and rx-rings
584  */
585 static void ixgbe_dump(struct ixgbe_adapter *adapter)
586 {
587 	struct net_device *netdev = adapter->netdev;
588 	struct ixgbe_hw *hw = &adapter->hw;
589 	struct ixgbe_reg_info *reginfo;
590 	int n = 0;
591 	struct ixgbe_ring *tx_ring;
592 	struct ixgbe_tx_buffer *tx_buffer;
593 	union ixgbe_adv_tx_desc *tx_desc;
594 	struct my_u0 { u64 a; u64 b; } *u0;
595 	struct ixgbe_ring *rx_ring;
596 	union ixgbe_adv_rx_desc *rx_desc;
597 	struct ixgbe_rx_buffer *rx_buffer_info;
598 	u32 staterr;
599 	int i = 0;
600 
601 	if (!netif_msg_hw(adapter))
602 		return;
603 
604 	/* Print netdevice Info */
605 	if (netdev) {
606 		dev_info(&adapter->pdev->dev, "Net device Info\n");
607 		pr_info("Device Name     state            "
608 			"trans_start      last_rx\n");
609 		pr_info("%-15s %016lX %016lX %016lX\n",
610 			netdev->name,
611 			netdev->state,
612 			dev_trans_start(netdev),
613 			netdev->last_rx);
614 	}
615 
616 	/* Print Registers */
617 	dev_info(&adapter->pdev->dev, "Register Dump\n");
618 	pr_info(" Register Name   Value\n");
619 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
620 	     reginfo->name; reginfo++) {
621 		ixgbe_regdump(hw, reginfo);
622 	}
623 
624 	/* Print TX Ring Summary */
625 	if (!netdev || !netif_running(netdev))
626 		return;
627 
628 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
629 	pr_info(" %s     %s              %s        %s\n",
630 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
631 		"leng", "ntw", "timestamp");
632 	for (n = 0; n < adapter->num_tx_queues; n++) {
633 		tx_ring = adapter->tx_ring[n];
634 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
635 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
636 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
637 			   (u64)dma_unmap_addr(tx_buffer, dma),
638 			   dma_unmap_len(tx_buffer, len),
639 			   tx_buffer->next_to_watch,
640 			   (u64)tx_buffer->time_stamp);
641 	}
642 
643 	/* Print TX Rings */
644 	if (!netif_msg_tx_done(adapter))
645 		goto rx_ring_summary;
646 
647 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
648 
649 	/* Transmit Descriptor Formats
650 	 *
651 	 * 82598 Advanced Transmit Descriptor
652 	 *   +--------------------------------------------------------------+
653 	 * 0 |         Buffer Address [63:0]                                |
654 	 *   +--------------------------------------------------------------+
655 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
656 	 *   +--------------------------------------------------------------+
657 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
658 	 *
659 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
660 	 *   +--------------------------------------------------------------+
661 	 * 0 |                          RSV [63:0]                          |
662 	 *   +--------------------------------------------------------------+
663 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
664 	 *   +--------------------------------------------------------------+
665 	 *   63                       36 35   32 31                         0
666 	 *
667 	 * 82599+ Advanced Transmit Descriptor
668 	 *   +--------------------------------------------------------------+
669 	 * 0 |         Buffer Address [63:0]                                |
670 	 *   +--------------------------------------------------------------+
671 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
672 	 *   +--------------------------------------------------------------+
673 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
674 	 *
675 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
676 	 *   +--------------------------------------------------------------+
677 	 * 0 |                          RSV [63:0]                          |
678 	 *   +--------------------------------------------------------------+
679 	 * 8 |            RSV           |  STA  |           RSV             |
680 	 *   +--------------------------------------------------------------+
681 	 *   63                       36 35   32 31                         0
682 	 */
683 
684 	for (n = 0; n < adapter->num_tx_queues; n++) {
685 		tx_ring = adapter->tx_ring[n];
686 		pr_info("------------------------------------\n");
687 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
688 		pr_info("------------------------------------\n");
689 		pr_info("%s%s    %s              %s        %s          %s\n",
690 			"T [desc]     [address 63:0  ] ",
691 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
692 			"leng", "ntw", "timestamp", "bi->skb");
693 
694 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
695 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
696 			tx_buffer = &tx_ring->tx_buffer_info[i];
697 			u0 = (struct my_u0 *)tx_desc;
698 			if (dma_unmap_len(tx_buffer, len) > 0) {
699 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
700 					i,
701 					le64_to_cpu(u0->a),
702 					le64_to_cpu(u0->b),
703 					(u64)dma_unmap_addr(tx_buffer, dma),
704 					dma_unmap_len(tx_buffer, len),
705 					tx_buffer->next_to_watch,
706 					(u64)tx_buffer->time_stamp,
707 					tx_buffer->skb);
708 				if (i == tx_ring->next_to_use &&
709 					i == tx_ring->next_to_clean)
710 					pr_cont(" NTC/U\n");
711 				else if (i == tx_ring->next_to_use)
712 					pr_cont(" NTU\n");
713 				else if (i == tx_ring->next_to_clean)
714 					pr_cont(" NTC\n");
715 				else
716 					pr_cont("\n");
717 
718 				if (netif_msg_pktdata(adapter) &&
719 				    tx_buffer->skb)
720 					print_hex_dump(KERN_INFO, "",
721 						DUMP_PREFIX_ADDRESS, 16, 1,
722 						tx_buffer->skb->data,
723 						dma_unmap_len(tx_buffer, len),
724 						true);
725 			}
726 		}
727 	}
728 
729 	/* Print RX Rings Summary */
730 rx_ring_summary:
731 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
732 	pr_info("Queue [NTU] [NTC]\n");
733 	for (n = 0; n < adapter->num_rx_queues; n++) {
734 		rx_ring = adapter->rx_ring[n];
735 		pr_info("%5d %5X %5X\n",
736 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
737 	}
738 
739 	/* Print RX Rings */
740 	if (!netif_msg_rx_status(adapter))
741 		return;
742 
743 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
744 
745 	/* Receive Descriptor Formats
746 	 *
747 	 * 82598 Advanced Receive Descriptor (Read) Format
748 	 *    63                                           1        0
749 	 *    +-----------------------------------------------------+
750 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
751 	 *    +----------------------------------------------+------+
752 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
753 	 *    +-----------------------------------------------------+
754 	 *
755 	 *
756 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
757 	 *
758 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
759 	 *   +------------------------------------------------------+
760 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
761 	 *   | Packet   | IP     |   |          |     | Type | Type |
762 	 *   | Checksum | Ident  |   |          |     |      |      |
763 	 *   +------------------------------------------------------+
764 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
765 	 *   +------------------------------------------------------+
766 	 *   63       48 47    32 31            20 19               0
767 	 *
768 	 * 82599+ Advanced Receive Descriptor (Read) Format
769 	 *    63                                           1        0
770 	 *    +-----------------------------------------------------+
771 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
772 	 *    +----------------------------------------------+------+
773 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
774 	 *    +-----------------------------------------------------+
775 	 *
776 	 *
777 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
778 	 *
779 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
780 	 *   +------------------------------------------------------+
781 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
782 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
783 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
784 	 *   +------------------------------------------------------+
785 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
786 	 *   +------------------------------------------------------+
787 	 *   63       48 47    32 31          20 19                 0
788 	 */
789 
790 	for (n = 0; n < adapter->num_rx_queues; n++) {
791 		rx_ring = adapter->rx_ring[n];
792 		pr_info("------------------------------------\n");
793 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
794 		pr_info("------------------------------------\n");
795 		pr_info("%s%s%s",
796 			"R  [desc]      [ PktBuf     A0] ",
797 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
798 			"<-- Adv Rx Read format\n");
799 		pr_info("%s%s%s",
800 			"RWB[desc]      [PcsmIpSHl PtRs] ",
801 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
802 			"<-- Adv Rx Write-Back format\n");
803 
804 		for (i = 0; i < rx_ring->count; i++) {
805 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
806 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
807 			u0 = (struct my_u0 *)rx_desc;
808 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
809 			if (staterr & IXGBE_RXD_STAT_DD) {
810 				/* Descriptor Done */
811 				pr_info("RWB[0x%03X]     %016llX "
812 					"%016llX ---------------- %p", i,
813 					le64_to_cpu(u0->a),
814 					le64_to_cpu(u0->b),
815 					rx_buffer_info->skb);
816 			} else {
817 				pr_info("R  [0x%03X]     %016llX "
818 					"%016llX %016llX %p", i,
819 					le64_to_cpu(u0->a),
820 					le64_to_cpu(u0->b),
821 					(u64)rx_buffer_info->dma,
822 					rx_buffer_info->skb);
823 
824 				if (netif_msg_pktdata(adapter) &&
825 				    rx_buffer_info->dma) {
826 					print_hex_dump(KERN_INFO, "",
827 					   DUMP_PREFIX_ADDRESS, 16, 1,
828 					   page_address(rx_buffer_info->page) +
829 						    rx_buffer_info->page_offset,
830 					   ixgbe_rx_bufsz(rx_ring), true);
831 				}
832 			}
833 
834 			if (i == rx_ring->next_to_use)
835 				pr_cont(" NTU\n");
836 			else if (i == rx_ring->next_to_clean)
837 				pr_cont(" NTC\n");
838 			else
839 				pr_cont("\n");
840 
841 		}
842 	}
843 }
844 
845 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
846 {
847 	u32 ctrl_ext;
848 
849 	/* Let firmware take over control of h/w */
850 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
851 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
852 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
853 }
854 
855 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
856 {
857 	u32 ctrl_ext;
858 
859 	/* Let firmware know the driver has taken over */
860 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
861 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
862 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
863 }
864 
865 /**
866  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
867  * @adapter: pointer to adapter struct
868  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
869  * @queue: queue to map the corresponding interrupt to
870  * @msix_vector: the vector to map to the corresponding queue
871  *
872  */
873 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
874 			   u8 queue, u8 msix_vector)
875 {
876 	u32 ivar, index;
877 	struct ixgbe_hw *hw = &adapter->hw;
878 	switch (hw->mac.type) {
879 	case ixgbe_mac_82598EB:
880 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
881 		if (direction == -1)
882 			direction = 0;
883 		index = (((direction * 64) + queue) >> 2) & 0x1F;
884 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
885 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
886 		ivar |= (msix_vector << (8 * (queue & 0x3)));
887 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
888 		break;
889 	case ixgbe_mac_82599EB:
890 	case ixgbe_mac_X540:
891 	case ixgbe_mac_X550:
892 	case ixgbe_mac_X550EM_x:
893 	case ixgbe_mac_x550em_a:
894 		if (direction == -1) {
895 			/* other causes */
896 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
897 			index = ((queue & 1) * 8);
898 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
899 			ivar &= ~(0xFF << index);
900 			ivar |= (msix_vector << index);
901 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
902 			break;
903 		} else {
904 			/* tx or rx causes */
905 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
906 			index = ((16 * (queue & 1)) + (8 * direction));
907 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
908 			ivar &= ~(0xFF << index);
909 			ivar |= (msix_vector << index);
910 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
911 			break;
912 		}
913 	default:
914 		break;
915 	}
916 }
917 
918 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
919 					  u64 qmask)
920 {
921 	u32 mask;
922 
923 	switch (adapter->hw.mac.type) {
924 	case ixgbe_mac_82598EB:
925 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
926 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
927 		break;
928 	case ixgbe_mac_82599EB:
929 	case ixgbe_mac_X540:
930 	case ixgbe_mac_X550:
931 	case ixgbe_mac_X550EM_x:
932 	case ixgbe_mac_x550em_a:
933 		mask = (qmask & 0xFFFFFFFF);
934 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
935 		mask = (qmask >> 32);
936 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
937 		break;
938 	default:
939 		break;
940 	}
941 }
942 
943 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
944 				      struct ixgbe_tx_buffer *tx_buffer)
945 {
946 	if (tx_buffer->skb) {
947 		dev_kfree_skb_any(tx_buffer->skb);
948 		if (dma_unmap_len(tx_buffer, len))
949 			dma_unmap_single(ring->dev,
950 					 dma_unmap_addr(tx_buffer, dma),
951 					 dma_unmap_len(tx_buffer, len),
952 					 DMA_TO_DEVICE);
953 	} else if (dma_unmap_len(tx_buffer, len)) {
954 		dma_unmap_page(ring->dev,
955 			       dma_unmap_addr(tx_buffer, dma),
956 			       dma_unmap_len(tx_buffer, len),
957 			       DMA_TO_DEVICE);
958 	}
959 	tx_buffer->next_to_watch = NULL;
960 	tx_buffer->skb = NULL;
961 	dma_unmap_len_set(tx_buffer, len, 0);
962 	/* tx_buffer must be completely set up in the transmit path */
963 }
964 
965 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
966 {
967 	struct ixgbe_hw *hw = &adapter->hw;
968 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
969 	int i;
970 	u32 data;
971 
972 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
973 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
974 		return;
975 
976 	switch (hw->mac.type) {
977 	case ixgbe_mac_82598EB:
978 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
979 		break;
980 	default:
981 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
982 	}
983 	hwstats->lxoffrxc += data;
984 
985 	/* refill credits (no tx hang) if we received xoff */
986 	if (!data)
987 		return;
988 
989 	for (i = 0; i < adapter->num_tx_queues; i++)
990 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
991 			  &adapter->tx_ring[i]->state);
992 }
993 
994 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
995 {
996 	struct ixgbe_hw *hw = &adapter->hw;
997 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
998 	u32 xoff[8] = {0};
999 	u8 tc;
1000 	int i;
1001 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
1002 
1003 	if (adapter->ixgbe_ieee_pfc)
1004 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
1005 
1006 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
1007 		ixgbe_update_xoff_rx_lfc(adapter);
1008 		return;
1009 	}
1010 
1011 	/* update stats for each tc, only valid with PFC enabled */
1012 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
1013 		u32 pxoffrxc;
1014 
1015 		switch (hw->mac.type) {
1016 		case ixgbe_mac_82598EB:
1017 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1018 			break;
1019 		default:
1020 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1021 		}
1022 		hwstats->pxoffrxc[i] += pxoffrxc;
1023 		/* Get the TC for given UP */
1024 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1025 		xoff[tc] += pxoffrxc;
1026 	}
1027 
1028 	/* disarm tx queues that have received xoff frames */
1029 	for (i = 0; i < adapter->num_tx_queues; i++) {
1030 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1031 
1032 		tc = tx_ring->dcb_tc;
1033 		if (xoff[tc])
1034 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1035 	}
1036 }
1037 
1038 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1039 {
1040 	return ring->stats.packets;
1041 }
1042 
1043 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1044 {
1045 	struct ixgbe_adapter *adapter;
1046 	struct ixgbe_hw *hw;
1047 	u32 head, tail;
1048 
1049 	if (ring->l2_accel_priv)
1050 		adapter = ring->l2_accel_priv->real_adapter;
1051 	else
1052 		adapter = netdev_priv(ring->netdev);
1053 
1054 	hw = &adapter->hw;
1055 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1056 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1057 
1058 	if (head != tail)
1059 		return (head < tail) ?
1060 			tail - head : (tail + ring->count - head);
1061 
1062 	return 0;
1063 }
1064 
1065 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1066 {
1067 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1068 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1069 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1070 
1071 	clear_check_for_tx_hang(tx_ring);
1072 
1073 	/*
1074 	 * Check for a hung queue, but be thorough. This verifies
1075 	 * that a transmit has been completed since the previous
1076 	 * check AND there is at least one packet pending. The
1077 	 * ARMED bit is set to indicate a potential hang. The
1078 	 * bit is cleared if a pause frame is received to remove
1079 	 * false hang detection due to PFC or 802.3x frames. By
1080 	 * requiring this to fail twice we avoid races with
1081 	 * pfc clearing the ARMED bit and conditions where we
1082 	 * run the check_tx_hang logic with a transmit completion
1083 	 * pending but without time to complete it yet.
1084 	 */
1085 	if (tx_done_old == tx_done && tx_pending)
1086 		/* make sure it is true for two checks in a row */
1087 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1088 					&tx_ring->state);
1089 	/* update completed stats and continue */
1090 	tx_ring->tx_stats.tx_done_old = tx_done;
1091 	/* reset the countdown */
1092 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1093 
1094 	return false;
1095 }
1096 
1097 /**
1098  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1099  * @adapter: driver private struct
1100  **/
1101 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1102 {
1103 
1104 	/* Do the reset outside of interrupt context */
1105 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1106 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
1107 		e_warn(drv, "initiating reset due to tx timeout\n");
1108 		ixgbe_service_event_schedule(adapter);
1109 	}
1110 }
1111 
1112 /**
1113  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1114  **/
1115 static int ixgbe_tx_maxrate(struct net_device *netdev,
1116 			    int queue_index, u32 maxrate)
1117 {
1118 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1119 	struct ixgbe_hw *hw = &adapter->hw;
1120 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1121 
1122 	if (!maxrate)
1123 		return 0;
1124 
1125 	/* Calculate the rate factor values to set */
1126 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1127 	bcnrc_val /= maxrate;
1128 
1129 	/* clear everything but the rate factor */
1130 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1131 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1132 
1133 	/* enable the rate scheduler */
1134 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1135 
1136 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1137 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1138 
1139 	return 0;
1140 }
1141 
1142 /**
1143  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1144  * @q_vector: structure containing interrupt and ring information
1145  * @tx_ring: tx ring to clean
1146  * @napi_budget: Used to determine if we are in netpoll
1147  **/
1148 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1149 			       struct ixgbe_ring *tx_ring, int napi_budget)
1150 {
1151 	struct ixgbe_adapter *adapter = q_vector->adapter;
1152 	struct ixgbe_tx_buffer *tx_buffer;
1153 	union ixgbe_adv_tx_desc *tx_desc;
1154 	unsigned int total_bytes = 0, total_packets = 0;
1155 	unsigned int budget = q_vector->tx.work_limit;
1156 	unsigned int i = tx_ring->next_to_clean;
1157 
1158 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1159 		return true;
1160 
1161 	tx_buffer = &tx_ring->tx_buffer_info[i];
1162 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1163 	i -= tx_ring->count;
1164 
1165 	do {
1166 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1167 
1168 		/* if next_to_watch is not set then there is no work pending */
1169 		if (!eop_desc)
1170 			break;
1171 
1172 		/* prevent any other reads prior to eop_desc */
1173 		read_barrier_depends();
1174 
1175 		/* if DD is not set pending work has not been completed */
1176 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1177 			break;
1178 
1179 		/* clear next_to_watch to prevent false hangs */
1180 		tx_buffer->next_to_watch = NULL;
1181 
1182 		/* update the statistics for this packet */
1183 		total_bytes += tx_buffer->bytecount;
1184 		total_packets += tx_buffer->gso_segs;
1185 
1186 		/* free the skb */
1187 		napi_consume_skb(tx_buffer->skb, napi_budget);
1188 
1189 		/* unmap skb header data */
1190 		dma_unmap_single(tx_ring->dev,
1191 				 dma_unmap_addr(tx_buffer, dma),
1192 				 dma_unmap_len(tx_buffer, len),
1193 				 DMA_TO_DEVICE);
1194 
1195 		/* clear tx_buffer data */
1196 		tx_buffer->skb = NULL;
1197 		dma_unmap_len_set(tx_buffer, len, 0);
1198 
1199 		/* unmap remaining buffers */
1200 		while (tx_desc != eop_desc) {
1201 			tx_buffer++;
1202 			tx_desc++;
1203 			i++;
1204 			if (unlikely(!i)) {
1205 				i -= tx_ring->count;
1206 				tx_buffer = tx_ring->tx_buffer_info;
1207 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1208 			}
1209 
1210 			/* unmap any remaining paged data */
1211 			if (dma_unmap_len(tx_buffer, len)) {
1212 				dma_unmap_page(tx_ring->dev,
1213 					       dma_unmap_addr(tx_buffer, dma),
1214 					       dma_unmap_len(tx_buffer, len),
1215 					       DMA_TO_DEVICE);
1216 				dma_unmap_len_set(tx_buffer, len, 0);
1217 			}
1218 		}
1219 
1220 		/* move us one more past the eop_desc for start of next pkt */
1221 		tx_buffer++;
1222 		tx_desc++;
1223 		i++;
1224 		if (unlikely(!i)) {
1225 			i -= tx_ring->count;
1226 			tx_buffer = tx_ring->tx_buffer_info;
1227 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1228 		}
1229 
1230 		/* issue prefetch for next Tx descriptor */
1231 		prefetch(tx_desc);
1232 
1233 		/* update budget accounting */
1234 		budget--;
1235 	} while (likely(budget));
1236 
1237 	i += tx_ring->count;
1238 	tx_ring->next_to_clean = i;
1239 	u64_stats_update_begin(&tx_ring->syncp);
1240 	tx_ring->stats.bytes += total_bytes;
1241 	tx_ring->stats.packets += total_packets;
1242 	u64_stats_update_end(&tx_ring->syncp);
1243 	q_vector->tx.total_bytes += total_bytes;
1244 	q_vector->tx.total_packets += total_packets;
1245 
1246 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1247 		/* schedule immediate reset if we believe we hung */
1248 		struct ixgbe_hw *hw = &adapter->hw;
1249 		e_err(drv, "Detected Tx Unit Hang\n"
1250 			"  Tx Queue             <%d>\n"
1251 			"  TDH, TDT             <%x>, <%x>\n"
1252 			"  next_to_use          <%x>\n"
1253 			"  next_to_clean        <%x>\n"
1254 			"tx_buffer_info[next_to_clean]\n"
1255 			"  time_stamp           <%lx>\n"
1256 			"  jiffies              <%lx>\n",
1257 			tx_ring->queue_index,
1258 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1259 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1260 			tx_ring->next_to_use, i,
1261 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1262 
1263 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1264 
1265 		e_info(probe,
1266 		       "tx hang %d detected on queue %d, resetting adapter\n",
1267 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1268 
1269 		/* schedule immediate reset if we believe we hung */
1270 		ixgbe_tx_timeout_reset(adapter);
1271 
1272 		/* the adapter is about to reset, no point in enabling stuff */
1273 		return true;
1274 	}
1275 
1276 	netdev_tx_completed_queue(txring_txq(tx_ring),
1277 				  total_packets, total_bytes);
1278 
1279 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1280 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1281 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1282 		/* Make sure that anybody stopping the queue after this
1283 		 * sees the new next_to_clean.
1284 		 */
1285 		smp_mb();
1286 		if (__netif_subqueue_stopped(tx_ring->netdev,
1287 					     tx_ring->queue_index)
1288 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1289 			netif_wake_subqueue(tx_ring->netdev,
1290 					    tx_ring->queue_index);
1291 			++tx_ring->tx_stats.restart_queue;
1292 		}
1293 	}
1294 
1295 	return !!budget;
1296 }
1297 
1298 #ifdef CONFIG_IXGBE_DCA
1299 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1300 				struct ixgbe_ring *tx_ring,
1301 				int cpu)
1302 {
1303 	struct ixgbe_hw *hw = &adapter->hw;
1304 	u32 txctrl = 0;
1305 	u16 reg_offset;
1306 
1307 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1308 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1309 
1310 	switch (hw->mac.type) {
1311 	case ixgbe_mac_82598EB:
1312 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1313 		break;
1314 	case ixgbe_mac_82599EB:
1315 	case ixgbe_mac_X540:
1316 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1317 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1318 		break;
1319 	default:
1320 		/* for unknown hardware do not write register */
1321 		return;
1322 	}
1323 
1324 	/*
1325 	 * We can enable relaxed ordering for reads, but not writes when
1326 	 * DCA is enabled.  This is due to a known issue in some chipsets
1327 	 * which will cause the DCA tag to be cleared.
1328 	 */
1329 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1330 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1331 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1332 
1333 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1334 }
1335 
1336 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1337 				struct ixgbe_ring *rx_ring,
1338 				int cpu)
1339 {
1340 	struct ixgbe_hw *hw = &adapter->hw;
1341 	u32 rxctrl = 0;
1342 	u8 reg_idx = rx_ring->reg_idx;
1343 
1344 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1345 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1346 
1347 	switch (hw->mac.type) {
1348 	case ixgbe_mac_82599EB:
1349 	case ixgbe_mac_X540:
1350 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1351 		break;
1352 	default:
1353 		break;
1354 	}
1355 
1356 	/*
1357 	 * We can enable relaxed ordering for reads, but not writes when
1358 	 * DCA is enabled.  This is due to a known issue in some chipsets
1359 	 * which will cause the DCA tag to be cleared.
1360 	 */
1361 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1362 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1363 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1364 
1365 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1366 }
1367 
1368 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1369 {
1370 	struct ixgbe_adapter *adapter = q_vector->adapter;
1371 	struct ixgbe_ring *ring;
1372 	int cpu = get_cpu();
1373 
1374 	if (q_vector->cpu == cpu)
1375 		goto out_no_update;
1376 
1377 	ixgbe_for_each_ring(ring, q_vector->tx)
1378 		ixgbe_update_tx_dca(adapter, ring, cpu);
1379 
1380 	ixgbe_for_each_ring(ring, q_vector->rx)
1381 		ixgbe_update_rx_dca(adapter, ring, cpu);
1382 
1383 	q_vector->cpu = cpu;
1384 out_no_update:
1385 	put_cpu();
1386 }
1387 
1388 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1389 {
1390 	int i;
1391 
1392 	/* always use CB2 mode, difference is masked in the CB driver */
1393 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1394 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1395 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1396 	else
1397 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1398 				IXGBE_DCA_CTRL_DCA_DISABLE);
1399 
1400 	for (i = 0; i < adapter->num_q_vectors; i++) {
1401 		adapter->q_vector[i]->cpu = -1;
1402 		ixgbe_update_dca(adapter->q_vector[i]);
1403 	}
1404 }
1405 
1406 static int __ixgbe_notify_dca(struct device *dev, void *data)
1407 {
1408 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1409 	unsigned long event = *(unsigned long *)data;
1410 
1411 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1412 		return 0;
1413 
1414 	switch (event) {
1415 	case DCA_PROVIDER_ADD:
1416 		/* if we're already enabled, don't do it again */
1417 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1418 			break;
1419 		if (dca_add_requester(dev) == 0) {
1420 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1421 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1422 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1423 			break;
1424 		}
1425 		/* Fall Through since DCA is disabled. */
1426 	case DCA_PROVIDER_REMOVE:
1427 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1428 			dca_remove_requester(dev);
1429 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1430 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1431 					IXGBE_DCA_CTRL_DCA_DISABLE);
1432 		}
1433 		break;
1434 	}
1435 
1436 	return 0;
1437 }
1438 
1439 #endif /* CONFIG_IXGBE_DCA */
1440 
1441 #define IXGBE_RSS_L4_TYPES_MASK \
1442 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1443 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1444 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1445 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1446 
1447 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1448 				 union ixgbe_adv_rx_desc *rx_desc,
1449 				 struct sk_buff *skb)
1450 {
1451 	u16 rss_type;
1452 
1453 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1454 		return;
1455 
1456 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1457 		   IXGBE_RXDADV_RSSTYPE_MASK;
1458 
1459 	if (!rss_type)
1460 		return;
1461 
1462 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1463 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1464 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1465 }
1466 
1467 #ifdef IXGBE_FCOE
1468 /**
1469  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1470  * @ring: structure containing ring specific data
1471  * @rx_desc: advanced rx descriptor
1472  *
1473  * Returns : true if it is FCoE pkt
1474  */
1475 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1476 				    union ixgbe_adv_rx_desc *rx_desc)
1477 {
1478 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1479 
1480 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1481 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1482 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1483 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1484 }
1485 
1486 #endif /* IXGBE_FCOE */
1487 /**
1488  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1489  * @ring: structure containing ring specific data
1490  * @rx_desc: current Rx descriptor being processed
1491  * @skb: skb currently being received and modified
1492  **/
1493 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1494 				     union ixgbe_adv_rx_desc *rx_desc,
1495 				     struct sk_buff *skb)
1496 {
1497 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1498 	__le16 hdr_info = rx_desc->wb.lower.lo_dword.hs_rss.hdr_info;
1499 	bool encap_pkt = false;
1500 
1501 	skb_checksum_none_assert(skb);
1502 
1503 	/* Rx csum disabled */
1504 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1505 		return;
1506 
1507 	if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) &&
1508 	    (hdr_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_TUNNEL >> 16))) {
1509 		encap_pkt = true;
1510 		skb->encapsulation = 1;
1511 	}
1512 
1513 	/* if IP and error */
1514 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1515 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1516 		ring->rx_stats.csum_err++;
1517 		return;
1518 	}
1519 
1520 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1521 		return;
1522 
1523 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1524 		/*
1525 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1526 		 * checksum errors.
1527 		 */
1528 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1529 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1530 			return;
1531 
1532 		ring->rx_stats.csum_err++;
1533 		return;
1534 	}
1535 
1536 	/* It must be a TCP or UDP packet with a valid checksum */
1537 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1538 	if (encap_pkt) {
1539 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1540 			return;
1541 
1542 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1543 			skb->ip_summed = CHECKSUM_NONE;
1544 			return;
1545 		}
1546 		/* If we checked the outer header let the stack know */
1547 		skb->csum_level = 1;
1548 	}
1549 }
1550 
1551 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1552 				    struct ixgbe_rx_buffer *bi)
1553 {
1554 	struct page *page = bi->page;
1555 	dma_addr_t dma;
1556 
1557 	/* since we are recycling buffers we should seldom need to alloc */
1558 	if (likely(page))
1559 		return true;
1560 
1561 	/* alloc new page for storage */
1562 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1563 	if (unlikely(!page)) {
1564 		rx_ring->rx_stats.alloc_rx_page_failed++;
1565 		return false;
1566 	}
1567 
1568 	/* map page for use */
1569 	dma = dma_map_page(rx_ring->dev, page, 0,
1570 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1571 
1572 	/*
1573 	 * if mapping failed free memory back to system since
1574 	 * there isn't much point in holding memory we can't use
1575 	 */
1576 	if (dma_mapping_error(rx_ring->dev, dma)) {
1577 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1578 
1579 		rx_ring->rx_stats.alloc_rx_page_failed++;
1580 		return false;
1581 	}
1582 
1583 	bi->dma = dma;
1584 	bi->page = page;
1585 	bi->page_offset = 0;
1586 
1587 	return true;
1588 }
1589 
1590 /**
1591  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1592  * @rx_ring: ring to place buffers on
1593  * @cleaned_count: number of buffers to replace
1594  **/
1595 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1596 {
1597 	union ixgbe_adv_rx_desc *rx_desc;
1598 	struct ixgbe_rx_buffer *bi;
1599 	u16 i = rx_ring->next_to_use;
1600 
1601 	/* nothing to do */
1602 	if (!cleaned_count)
1603 		return;
1604 
1605 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1606 	bi = &rx_ring->rx_buffer_info[i];
1607 	i -= rx_ring->count;
1608 
1609 	do {
1610 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1611 			break;
1612 
1613 		/*
1614 		 * Refresh the desc even if buffer_addrs didn't change
1615 		 * because each write-back erases this info.
1616 		 */
1617 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1618 
1619 		rx_desc++;
1620 		bi++;
1621 		i++;
1622 		if (unlikely(!i)) {
1623 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1624 			bi = rx_ring->rx_buffer_info;
1625 			i -= rx_ring->count;
1626 		}
1627 
1628 		/* clear the status bits for the next_to_use descriptor */
1629 		rx_desc->wb.upper.status_error = 0;
1630 
1631 		cleaned_count--;
1632 	} while (cleaned_count);
1633 
1634 	i += rx_ring->count;
1635 
1636 	if (rx_ring->next_to_use != i) {
1637 		rx_ring->next_to_use = i;
1638 
1639 		/* update next to alloc since we have filled the ring */
1640 		rx_ring->next_to_alloc = i;
1641 
1642 		/* Force memory writes to complete before letting h/w
1643 		 * know there are new descriptors to fetch.  (Only
1644 		 * applicable for weak-ordered memory model archs,
1645 		 * such as IA-64).
1646 		 */
1647 		wmb();
1648 		writel(i, rx_ring->tail);
1649 	}
1650 }
1651 
1652 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1653 				   struct sk_buff *skb)
1654 {
1655 	u16 hdr_len = skb_headlen(skb);
1656 
1657 	/* set gso_size to avoid messing up TCP MSS */
1658 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1659 						 IXGBE_CB(skb)->append_cnt);
1660 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1661 }
1662 
1663 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1664 				   struct sk_buff *skb)
1665 {
1666 	/* if append_cnt is 0 then frame is not RSC */
1667 	if (!IXGBE_CB(skb)->append_cnt)
1668 		return;
1669 
1670 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1671 	rx_ring->rx_stats.rsc_flush++;
1672 
1673 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1674 
1675 	/* gso_size is computed using append_cnt so always clear it last */
1676 	IXGBE_CB(skb)->append_cnt = 0;
1677 }
1678 
1679 /**
1680  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1681  * @rx_ring: rx descriptor ring packet is being transacted on
1682  * @rx_desc: pointer to the EOP Rx descriptor
1683  * @skb: pointer to current skb being populated
1684  *
1685  * This function checks the ring, descriptor, and packet information in
1686  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1687  * other fields within the skb.
1688  **/
1689 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1690 				     union ixgbe_adv_rx_desc *rx_desc,
1691 				     struct sk_buff *skb)
1692 {
1693 	struct net_device *dev = rx_ring->netdev;
1694 	u32 flags = rx_ring->q_vector->adapter->flags;
1695 
1696 	ixgbe_update_rsc_stats(rx_ring, skb);
1697 
1698 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1699 
1700 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1701 
1702 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1703 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1704 
1705 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1706 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1707 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1708 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1709 	}
1710 
1711 	skb_record_rx_queue(skb, rx_ring->queue_index);
1712 
1713 	skb->protocol = eth_type_trans(skb, dev);
1714 }
1715 
1716 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1717 			 struct sk_buff *skb)
1718 {
1719 	skb_mark_napi_id(skb, &q_vector->napi);
1720 	if (ixgbe_qv_busy_polling(q_vector))
1721 		netif_receive_skb(skb);
1722 	else
1723 		napi_gro_receive(&q_vector->napi, skb);
1724 }
1725 
1726 /**
1727  * ixgbe_is_non_eop - process handling of non-EOP buffers
1728  * @rx_ring: Rx ring being processed
1729  * @rx_desc: Rx descriptor for current buffer
1730  * @skb: Current socket buffer containing buffer in progress
1731  *
1732  * This function updates next to clean.  If the buffer is an EOP buffer
1733  * this function exits returning false, otherwise it will place the
1734  * sk_buff in the next buffer to be chained and return true indicating
1735  * that this is in fact a non-EOP buffer.
1736  **/
1737 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1738 			     union ixgbe_adv_rx_desc *rx_desc,
1739 			     struct sk_buff *skb)
1740 {
1741 	u32 ntc = rx_ring->next_to_clean + 1;
1742 
1743 	/* fetch, update, and store next to clean */
1744 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1745 	rx_ring->next_to_clean = ntc;
1746 
1747 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1748 
1749 	/* update RSC append count if present */
1750 	if (ring_is_rsc_enabled(rx_ring)) {
1751 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1752 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1753 
1754 		if (unlikely(rsc_enabled)) {
1755 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1756 
1757 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1758 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1759 
1760 			/* update ntc based on RSC value */
1761 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1762 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1763 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1764 		}
1765 	}
1766 
1767 	/* if we are the last buffer then there is nothing else to do */
1768 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1769 		return false;
1770 
1771 	/* place skb in next buffer to be received */
1772 	rx_ring->rx_buffer_info[ntc].skb = skb;
1773 	rx_ring->rx_stats.non_eop_descs++;
1774 
1775 	return true;
1776 }
1777 
1778 /**
1779  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1780  * @rx_ring: rx descriptor ring packet is being transacted on
1781  * @skb: pointer to current skb being adjusted
1782  *
1783  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1784  * main difference between this version and the original function is that
1785  * this function can make several assumptions about the state of things
1786  * that allow for significant optimizations versus the standard function.
1787  * As a result we can do things like drop a frag and maintain an accurate
1788  * truesize for the skb.
1789  */
1790 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1791 			    struct sk_buff *skb)
1792 {
1793 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1794 	unsigned char *va;
1795 	unsigned int pull_len;
1796 
1797 	/*
1798 	 * it is valid to use page_address instead of kmap since we are
1799 	 * working with pages allocated out of the lomem pool per
1800 	 * alloc_page(GFP_ATOMIC)
1801 	 */
1802 	va = skb_frag_address(frag);
1803 
1804 	/*
1805 	 * we need the header to contain the greater of either ETH_HLEN or
1806 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1807 	 */
1808 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1809 
1810 	/* align pull length to size of long to optimize memcpy performance */
1811 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1812 
1813 	/* update all of the pointers */
1814 	skb_frag_size_sub(frag, pull_len);
1815 	frag->page_offset += pull_len;
1816 	skb->data_len -= pull_len;
1817 	skb->tail += pull_len;
1818 }
1819 
1820 /**
1821  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1822  * @rx_ring: rx descriptor ring packet is being transacted on
1823  * @skb: pointer to current skb being updated
1824  *
1825  * This function provides a basic DMA sync up for the first fragment of an
1826  * skb.  The reason for doing this is that the first fragment cannot be
1827  * unmapped until we have reached the end of packet descriptor for a buffer
1828  * chain.
1829  */
1830 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1831 				struct sk_buff *skb)
1832 {
1833 	/* if the page was released unmap it, else just sync our portion */
1834 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1835 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1836 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1837 		IXGBE_CB(skb)->page_released = false;
1838 	} else {
1839 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1840 
1841 		dma_sync_single_range_for_cpu(rx_ring->dev,
1842 					      IXGBE_CB(skb)->dma,
1843 					      frag->page_offset,
1844 					      ixgbe_rx_bufsz(rx_ring),
1845 					      DMA_FROM_DEVICE);
1846 	}
1847 	IXGBE_CB(skb)->dma = 0;
1848 }
1849 
1850 /**
1851  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1852  * @rx_ring: rx descriptor ring packet is being transacted on
1853  * @rx_desc: pointer to the EOP Rx descriptor
1854  * @skb: pointer to current skb being fixed
1855  *
1856  * Check for corrupted packet headers caused by senders on the local L2
1857  * embedded NIC switch not setting up their Tx Descriptors right.  These
1858  * should be very rare.
1859  *
1860  * Also address the case where we are pulling data in on pages only
1861  * and as such no data is present in the skb header.
1862  *
1863  * In addition if skb is not at least 60 bytes we need to pad it so that
1864  * it is large enough to qualify as a valid Ethernet frame.
1865  *
1866  * Returns true if an error was encountered and skb was freed.
1867  **/
1868 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1869 				  union ixgbe_adv_rx_desc *rx_desc,
1870 				  struct sk_buff *skb)
1871 {
1872 	struct net_device *netdev = rx_ring->netdev;
1873 
1874 	/* verify that the packet does not have any known errors */
1875 	if (unlikely(ixgbe_test_staterr(rx_desc,
1876 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1877 	    !(netdev->features & NETIF_F_RXALL))) {
1878 		dev_kfree_skb_any(skb);
1879 		return true;
1880 	}
1881 
1882 	/* place header in linear portion of buffer */
1883 	if (skb_is_nonlinear(skb))
1884 		ixgbe_pull_tail(rx_ring, skb);
1885 
1886 #ifdef IXGBE_FCOE
1887 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1888 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1889 		return false;
1890 
1891 #endif
1892 	/* if eth_skb_pad returns an error the skb was freed */
1893 	if (eth_skb_pad(skb))
1894 		return true;
1895 
1896 	return false;
1897 }
1898 
1899 /**
1900  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1901  * @rx_ring: rx descriptor ring to store buffers on
1902  * @old_buff: donor buffer to have page reused
1903  *
1904  * Synchronizes page for reuse by the adapter
1905  **/
1906 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1907 				struct ixgbe_rx_buffer *old_buff)
1908 {
1909 	struct ixgbe_rx_buffer *new_buff;
1910 	u16 nta = rx_ring->next_to_alloc;
1911 
1912 	new_buff = &rx_ring->rx_buffer_info[nta];
1913 
1914 	/* update, and store next to alloc */
1915 	nta++;
1916 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1917 
1918 	/* transfer page from old buffer to new buffer */
1919 	*new_buff = *old_buff;
1920 
1921 	/* sync the buffer for use by the device */
1922 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1923 					 new_buff->page_offset,
1924 					 ixgbe_rx_bufsz(rx_ring),
1925 					 DMA_FROM_DEVICE);
1926 }
1927 
1928 static inline bool ixgbe_page_is_reserved(struct page *page)
1929 {
1930 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
1931 }
1932 
1933 /**
1934  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1935  * @rx_ring: rx descriptor ring to transact packets on
1936  * @rx_buffer: buffer containing page to add
1937  * @rx_desc: descriptor containing length of buffer written by hardware
1938  * @skb: sk_buff to place the data into
1939  *
1940  * This function will add the data contained in rx_buffer->page to the skb.
1941  * This is done either through a direct copy if the data in the buffer is
1942  * less than the skb header size, otherwise it will just attach the page as
1943  * a frag to the skb.
1944  *
1945  * The function will then update the page offset if necessary and return
1946  * true if the buffer can be reused by the adapter.
1947  **/
1948 static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1949 			      struct ixgbe_rx_buffer *rx_buffer,
1950 			      union ixgbe_adv_rx_desc *rx_desc,
1951 			      struct sk_buff *skb)
1952 {
1953 	struct page *page = rx_buffer->page;
1954 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1955 #if (PAGE_SIZE < 8192)
1956 	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1957 #else
1958 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1959 	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1960 				   ixgbe_rx_bufsz(rx_ring);
1961 #endif
1962 
1963 	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1964 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1965 
1966 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1967 
1968 		/* page is not reserved, we can reuse buffer as-is */
1969 		if (likely(!ixgbe_page_is_reserved(page)))
1970 			return true;
1971 
1972 		/* this page cannot be reused so discard it */
1973 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1974 		return false;
1975 	}
1976 
1977 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1978 			rx_buffer->page_offset, size, truesize);
1979 
1980 	/* avoid re-using remote pages */
1981 	if (unlikely(ixgbe_page_is_reserved(page)))
1982 		return false;
1983 
1984 #if (PAGE_SIZE < 8192)
1985 	/* if we are only owner of page we can reuse it */
1986 	if (unlikely(page_count(page) != 1))
1987 		return false;
1988 
1989 	/* flip page offset to other buffer */
1990 	rx_buffer->page_offset ^= truesize;
1991 #else
1992 	/* move offset up to the next cache line */
1993 	rx_buffer->page_offset += truesize;
1994 
1995 	if (rx_buffer->page_offset > last_offset)
1996 		return false;
1997 #endif
1998 
1999 	/* Even if we own the page, we are not allowed to use atomic_set()
2000 	 * This would break get_page_unless_zero() users.
2001 	 */
2002 	page_ref_inc(page);
2003 
2004 	return true;
2005 }
2006 
2007 static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
2008 					     union ixgbe_adv_rx_desc *rx_desc)
2009 {
2010 	struct ixgbe_rx_buffer *rx_buffer;
2011 	struct sk_buff *skb;
2012 	struct page *page;
2013 
2014 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2015 	page = rx_buffer->page;
2016 	prefetchw(page);
2017 
2018 	skb = rx_buffer->skb;
2019 
2020 	if (likely(!skb)) {
2021 		void *page_addr = page_address(page) +
2022 				  rx_buffer->page_offset;
2023 
2024 		/* prefetch first cache line of first page */
2025 		prefetch(page_addr);
2026 #if L1_CACHE_BYTES < 128
2027 		prefetch(page_addr + L1_CACHE_BYTES);
2028 #endif
2029 
2030 		/* allocate a skb to store the frags */
2031 		skb = napi_alloc_skb(&rx_ring->q_vector->napi,
2032 				     IXGBE_RX_HDR_SIZE);
2033 		if (unlikely(!skb)) {
2034 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2035 			return NULL;
2036 		}
2037 
2038 		/*
2039 		 * we will be copying header into skb->data in
2040 		 * pskb_may_pull so it is in our interest to prefetch
2041 		 * it now to avoid a possible cache miss
2042 		 */
2043 		prefetchw(skb->data);
2044 
2045 		/*
2046 		 * Delay unmapping of the first packet. It carries the
2047 		 * header information, HW may still access the header
2048 		 * after the writeback.  Only unmap it when EOP is
2049 		 * reached
2050 		 */
2051 		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
2052 			goto dma_sync;
2053 
2054 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2055 	} else {
2056 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2057 			ixgbe_dma_sync_frag(rx_ring, skb);
2058 
2059 dma_sync:
2060 		/* we are reusing so sync this buffer for CPU use */
2061 		dma_sync_single_range_for_cpu(rx_ring->dev,
2062 					      rx_buffer->dma,
2063 					      rx_buffer->page_offset,
2064 					      ixgbe_rx_bufsz(rx_ring),
2065 					      DMA_FROM_DEVICE);
2066 
2067 		rx_buffer->skb = NULL;
2068 	}
2069 
2070 	/* pull page into skb */
2071 	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
2072 		/* hand second half of page back to the ring */
2073 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2074 	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
2075 		/* the page has been released from the ring */
2076 		IXGBE_CB(skb)->page_released = true;
2077 	} else {
2078 		/* we are not reusing the buffer so unmap it */
2079 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
2080 			       ixgbe_rx_pg_size(rx_ring),
2081 			       DMA_FROM_DEVICE);
2082 	}
2083 
2084 	/* clear contents of buffer_info */
2085 	rx_buffer->page = NULL;
2086 
2087 	return skb;
2088 }
2089 
2090 /**
2091  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2092  * @q_vector: structure containing interrupt and ring information
2093  * @rx_ring: rx descriptor ring to transact packets on
2094  * @budget: Total limit on number of packets to process
2095  *
2096  * This function provides a "bounce buffer" approach to Rx interrupt
2097  * processing.  The advantage to this is that on systems that have
2098  * expensive overhead for IOMMU access this provides a means of avoiding
2099  * it by maintaining the mapping of the page to the syste.
2100  *
2101  * Returns amount of work completed
2102  **/
2103 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2104 			       struct ixgbe_ring *rx_ring,
2105 			       const int budget)
2106 {
2107 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2108 #ifdef IXGBE_FCOE
2109 	struct ixgbe_adapter *adapter = q_vector->adapter;
2110 	int ddp_bytes;
2111 	unsigned int mss = 0;
2112 #endif /* IXGBE_FCOE */
2113 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2114 
2115 	while (likely(total_rx_packets < budget)) {
2116 		union ixgbe_adv_rx_desc *rx_desc;
2117 		struct sk_buff *skb;
2118 
2119 		/* return some buffers to hardware, one at a time is too slow */
2120 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2121 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2122 			cleaned_count = 0;
2123 		}
2124 
2125 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2126 
2127 		if (!rx_desc->wb.upper.status_error)
2128 			break;
2129 
2130 		/* This memory barrier is needed to keep us from reading
2131 		 * any other fields out of the rx_desc until we know the
2132 		 * descriptor has been written back
2133 		 */
2134 		dma_rmb();
2135 
2136 		/* retrieve a buffer from the ring */
2137 		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
2138 
2139 		/* exit if we failed to retrieve a buffer */
2140 		if (!skb)
2141 			break;
2142 
2143 		cleaned_count++;
2144 
2145 		/* place incomplete frames back on ring for completion */
2146 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2147 			continue;
2148 
2149 		/* verify the packet layout is correct */
2150 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2151 			continue;
2152 
2153 		/* probably a little skewed due to removing CRC */
2154 		total_rx_bytes += skb->len;
2155 
2156 		/* populate checksum, timestamp, VLAN, and protocol */
2157 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2158 
2159 #ifdef IXGBE_FCOE
2160 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2161 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2162 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2163 			/* include DDPed FCoE data */
2164 			if (ddp_bytes > 0) {
2165 				if (!mss) {
2166 					mss = rx_ring->netdev->mtu -
2167 						sizeof(struct fcoe_hdr) -
2168 						sizeof(struct fc_frame_header) -
2169 						sizeof(struct fcoe_crc_eof);
2170 					if (mss > 512)
2171 						mss &= ~511;
2172 				}
2173 				total_rx_bytes += ddp_bytes;
2174 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2175 								 mss);
2176 			}
2177 			if (!ddp_bytes) {
2178 				dev_kfree_skb_any(skb);
2179 				continue;
2180 			}
2181 		}
2182 
2183 #endif /* IXGBE_FCOE */
2184 		ixgbe_rx_skb(q_vector, skb);
2185 
2186 		/* update budget accounting */
2187 		total_rx_packets++;
2188 	}
2189 
2190 	u64_stats_update_begin(&rx_ring->syncp);
2191 	rx_ring->stats.packets += total_rx_packets;
2192 	rx_ring->stats.bytes += total_rx_bytes;
2193 	u64_stats_update_end(&rx_ring->syncp);
2194 	q_vector->rx.total_packets += total_rx_packets;
2195 	q_vector->rx.total_bytes += total_rx_bytes;
2196 
2197 	return total_rx_packets;
2198 }
2199 
2200 #ifdef CONFIG_NET_RX_BUSY_POLL
2201 /* must be called with local_bh_disable()d */
2202 static int ixgbe_low_latency_recv(struct napi_struct *napi)
2203 {
2204 	struct ixgbe_q_vector *q_vector =
2205 			container_of(napi, struct ixgbe_q_vector, napi);
2206 	struct ixgbe_adapter *adapter = q_vector->adapter;
2207 	struct ixgbe_ring  *ring;
2208 	int found = 0;
2209 
2210 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2211 		return LL_FLUSH_FAILED;
2212 
2213 	if (!ixgbe_qv_lock_poll(q_vector))
2214 		return LL_FLUSH_BUSY;
2215 
2216 	ixgbe_for_each_ring(ring, q_vector->rx) {
2217 		found = ixgbe_clean_rx_irq(q_vector, ring, 4);
2218 #ifdef BP_EXTENDED_STATS
2219 		if (found)
2220 			ring->stats.cleaned += found;
2221 		else
2222 			ring->stats.misses++;
2223 #endif
2224 		if (found)
2225 			break;
2226 	}
2227 
2228 	ixgbe_qv_unlock_poll(q_vector);
2229 
2230 	return found;
2231 }
2232 #endif	/* CONFIG_NET_RX_BUSY_POLL */
2233 
2234 /**
2235  * ixgbe_configure_msix - Configure MSI-X hardware
2236  * @adapter: board private structure
2237  *
2238  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2239  * interrupts.
2240  **/
2241 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2242 {
2243 	struct ixgbe_q_vector *q_vector;
2244 	int v_idx;
2245 	u32 mask;
2246 
2247 	/* Populate MSIX to EITR Select */
2248 	if (adapter->num_vfs > 32) {
2249 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2250 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2251 	}
2252 
2253 	/*
2254 	 * Populate the IVAR table and set the ITR values to the
2255 	 * corresponding register.
2256 	 */
2257 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2258 		struct ixgbe_ring *ring;
2259 		q_vector = adapter->q_vector[v_idx];
2260 
2261 		ixgbe_for_each_ring(ring, q_vector->rx)
2262 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2263 
2264 		ixgbe_for_each_ring(ring, q_vector->tx)
2265 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2266 
2267 		ixgbe_write_eitr(q_vector);
2268 	}
2269 
2270 	switch (adapter->hw.mac.type) {
2271 	case ixgbe_mac_82598EB:
2272 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2273 			       v_idx);
2274 		break;
2275 	case ixgbe_mac_82599EB:
2276 	case ixgbe_mac_X540:
2277 	case ixgbe_mac_X550:
2278 	case ixgbe_mac_X550EM_x:
2279 	case ixgbe_mac_x550em_a:
2280 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2281 		break;
2282 	default:
2283 		break;
2284 	}
2285 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2286 
2287 	/* set up to autoclear timer, and the vectors */
2288 	mask = IXGBE_EIMS_ENABLE_MASK;
2289 	mask &= ~(IXGBE_EIMS_OTHER |
2290 		  IXGBE_EIMS_MAILBOX |
2291 		  IXGBE_EIMS_LSC);
2292 
2293 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2294 }
2295 
2296 enum latency_range {
2297 	lowest_latency = 0,
2298 	low_latency = 1,
2299 	bulk_latency = 2,
2300 	latency_invalid = 255
2301 };
2302 
2303 /**
2304  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2305  * @q_vector: structure containing interrupt and ring information
2306  * @ring_container: structure containing ring performance data
2307  *
2308  *      Stores a new ITR value based on packets and byte
2309  *      counts during the last interrupt.  The advantage of per interrupt
2310  *      computation is faster updates and more accurate ITR for the current
2311  *      traffic pattern.  Constants in this function were computed
2312  *      based on theoretical maximum wire speed and thresholds were set based
2313  *      on testing data as well as attempting to minimize response time
2314  *      while increasing bulk throughput.
2315  *      this functionality is controlled by the InterruptThrottleRate module
2316  *      parameter (see ixgbe_param.c)
2317  **/
2318 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2319 			     struct ixgbe_ring_container *ring_container)
2320 {
2321 	int bytes = ring_container->total_bytes;
2322 	int packets = ring_container->total_packets;
2323 	u32 timepassed_us;
2324 	u64 bytes_perint;
2325 	u8 itr_setting = ring_container->itr;
2326 
2327 	if (packets == 0)
2328 		return;
2329 
2330 	/* simple throttlerate management
2331 	 *   0-10MB/s   lowest (100000 ints/s)
2332 	 *  10-20MB/s   low    (20000 ints/s)
2333 	 *  20-1249MB/s bulk   (12000 ints/s)
2334 	 */
2335 	/* what was last interrupt timeslice? */
2336 	timepassed_us = q_vector->itr >> 2;
2337 	if (timepassed_us == 0)
2338 		return;
2339 
2340 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2341 
2342 	switch (itr_setting) {
2343 	case lowest_latency:
2344 		if (bytes_perint > 10)
2345 			itr_setting = low_latency;
2346 		break;
2347 	case low_latency:
2348 		if (bytes_perint > 20)
2349 			itr_setting = bulk_latency;
2350 		else if (bytes_perint <= 10)
2351 			itr_setting = lowest_latency;
2352 		break;
2353 	case bulk_latency:
2354 		if (bytes_perint <= 20)
2355 			itr_setting = low_latency;
2356 		break;
2357 	}
2358 
2359 	/* clear work counters since we have the values we need */
2360 	ring_container->total_bytes = 0;
2361 	ring_container->total_packets = 0;
2362 
2363 	/* write updated itr to ring container */
2364 	ring_container->itr = itr_setting;
2365 }
2366 
2367 /**
2368  * ixgbe_write_eitr - write EITR register in hardware specific way
2369  * @q_vector: structure containing interrupt and ring information
2370  *
2371  * This function is made to be called by ethtool and by the driver
2372  * when it needs to update EITR registers at runtime.  Hardware
2373  * specific quirks/differences are taken care of here.
2374  */
2375 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2376 {
2377 	struct ixgbe_adapter *adapter = q_vector->adapter;
2378 	struct ixgbe_hw *hw = &adapter->hw;
2379 	int v_idx = q_vector->v_idx;
2380 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2381 
2382 	switch (adapter->hw.mac.type) {
2383 	case ixgbe_mac_82598EB:
2384 		/* must write high and low 16 bits to reset counter */
2385 		itr_reg |= (itr_reg << 16);
2386 		break;
2387 	case ixgbe_mac_82599EB:
2388 	case ixgbe_mac_X540:
2389 	case ixgbe_mac_X550:
2390 	case ixgbe_mac_X550EM_x:
2391 	case ixgbe_mac_x550em_a:
2392 		/*
2393 		 * set the WDIS bit to not clear the timer bits and cause an
2394 		 * immediate assertion of the interrupt
2395 		 */
2396 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2397 		break;
2398 	default:
2399 		break;
2400 	}
2401 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2402 }
2403 
2404 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2405 {
2406 	u32 new_itr = q_vector->itr;
2407 	u8 current_itr;
2408 
2409 	ixgbe_update_itr(q_vector, &q_vector->tx);
2410 	ixgbe_update_itr(q_vector, &q_vector->rx);
2411 
2412 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2413 
2414 	switch (current_itr) {
2415 	/* counts and packets in update_itr are dependent on these numbers */
2416 	case lowest_latency:
2417 		new_itr = IXGBE_100K_ITR;
2418 		break;
2419 	case low_latency:
2420 		new_itr = IXGBE_20K_ITR;
2421 		break;
2422 	case bulk_latency:
2423 		new_itr = IXGBE_12K_ITR;
2424 		break;
2425 	default:
2426 		break;
2427 	}
2428 
2429 	if (new_itr != q_vector->itr) {
2430 		/* do an exponential smoothing */
2431 		new_itr = (10 * new_itr * q_vector->itr) /
2432 			  ((9 * new_itr) + q_vector->itr);
2433 
2434 		/* save the algorithm value here */
2435 		q_vector->itr = new_itr;
2436 
2437 		ixgbe_write_eitr(q_vector);
2438 	}
2439 }
2440 
2441 /**
2442  * ixgbe_check_overtemp_subtask - check for over temperature
2443  * @adapter: pointer to adapter
2444  **/
2445 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2446 {
2447 	struct ixgbe_hw *hw = &adapter->hw;
2448 	u32 eicr = adapter->interrupt_event;
2449 
2450 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2451 		return;
2452 
2453 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2454 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2455 		return;
2456 
2457 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2458 
2459 	switch (hw->device_id) {
2460 	case IXGBE_DEV_ID_82599_T3_LOM:
2461 		/*
2462 		 * Since the warning interrupt is for both ports
2463 		 * we don't have to check if:
2464 		 *  - This interrupt wasn't for our port.
2465 		 *  - We may have missed the interrupt so always have to
2466 		 *    check if we  got a LSC
2467 		 */
2468 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2469 		    !(eicr & IXGBE_EICR_LSC))
2470 			return;
2471 
2472 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2473 			u32 speed;
2474 			bool link_up = false;
2475 
2476 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2477 
2478 			if (link_up)
2479 				return;
2480 		}
2481 
2482 		/* Check if this is not due to overtemp */
2483 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2484 			return;
2485 
2486 		break;
2487 	default:
2488 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2489 			return;
2490 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2491 			return;
2492 		break;
2493 	}
2494 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2495 
2496 	adapter->interrupt_event = 0;
2497 }
2498 
2499 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2500 {
2501 	struct ixgbe_hw *hw = &adapter->hw;
2502 
2503 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2504 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2505 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2506 		/* write to clear the interrupt */
2507 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2508 	}
2509 }
2510 
2511 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2512 {
2513 	struct ixgbe_hw *hw = &adapter->hw;
2514 
2515 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2516 		return;
2517 
2518 	switch (adapter->hw.mac.type) {
2519 	case ixgbe_mac_82599EB:
2520 		/*
2521 		 * Need to check link state so complete overtemp check
2522 		 * on service task
2523 		 */
2524 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2525 		     (eicr & IXGBE_EICR_LSC)) &&
2526 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2527 			adapter->interrupt_event = eicr;
2528 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2529 			ixgbe_service_event_schedule(adapter);
2530 			return;
2531 		}
2532 		return;
2533 	case ixgbe_mac_X540:
2534 		if (!(eicr & IXGBE_EICR_TS))
2535 			return;
2536 		break;
2537 	default:
2538 		return;
2539 	}
2540 
2541 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2542 }
2543 
2544 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2545 {
2546 	switch (hw->mac.type) {
2547 	case ixgbe_mac_82598EB:
2548 		if (hw->phy.type == ixgbe_phy_nl)
2549 			return true;
2550 		return false;
2551 	case ixgbe_mac_82599EB:
2552 	case ixgbe_mac_X550EM_x:
2553 	case ixgbe_mac_x550em_a:
2554 		switch (hw->mac.ops.get_media_type(hw)) {
2555 		case ixgbe_media_type_fiber:
2556 		case ixgbe_media_type_fiber_qsfp:
2557 			return true;
2558 		default:
2559 			return false;
2560 		}
2561 	default:
2562 		return false;
2563 	}
2564 }
2565 
2566 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2567 {
2568 	struct ixgbe_hw *hw = &adapter->hw;
2569 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2570 
2571 	if (!ixgbe_is_sfp(hw))
2572 		return;
2573 
2574 	/* Later MAC's use different SDP */
2575 	if (hw->mac.type >= ixgbe_mac_X540)
2576 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2577 
2578 	if (eicr & eicr_mask) {
2579 		/* Clear the interrupt */
2580 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2581 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2582 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2583 			adapter->sfp_poll_time = 0;
2584 			ixgbe_service_event_schedule(adapter);
2585 		}
2586 	}
2587 
2588 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2589 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2590 		/* Clear the interrupt */
2591 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2592 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2593 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2594 			ixgbe_service_event_schedule(adapter);
2595 		}
2596 	}
2597 }
2598 
2599 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2600 {
2601 	struct ixgbe_hw *hw = &adapter->hw;
2602 
2603 	adapter->lsc_int++;
2604 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2605 	adapter->link_check_timeout = jiffies;
2606 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2607 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2608 		IXGBE_WRITE_FLUSH(hw);
2609 		ixgbe_service_event_schedule(adapter);
2610 	}
2611 }
2612 
2613 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2614 					   u64 qmask)
2615 {
2616 	u32 mask;
2617 	struct ixgbe_hw *hw = &adapter->hw;
2618 
2619 	switch (hw->mac.type) {
2620 	case ixgbe_mac_82598EB:
2621 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2622 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2623 		break;
2624 	case ixgbe_mac_82599EB:
2625 	case ixgbe_mac_X540:
2626 	case ixgbe_mac_X550:
2627 	case ixgbe_mac_X550EM_x:
2628 	case ixgbe_mac_x550em_a:
2629 		mask = (qmask & 0xFFFFFFFF);
2630 		if (mask)
2631 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2632 		mask = (qmask >> 32);
2633 		if (mask)
2634 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2635 		break;
2636 	default:
2637 		break;
2638 	}
2639 	/* skip the flush */
2640 }
2641 
2642 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2643 					    u64 qmask)
2644 {
2645 	u32 mask;
2646 	struct ixgbe_hw *hw = &adapter->hw;
2647 
2648 	switch (hw->mac.type) {
2649 	case ixgbe_mac_82598EB:
2650 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2651 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2652 		break;
2653 	case ixgbe_mac_82599EB:
2654 	case ixgbe_mac_X540:
2655 	case ixgbe_mac_X550:
2656 	case ixgbe_mac_X550EM_x:
2657 	case ixgbe_mac_x550em_a:
2658 		mask = (qmask & 0xFFFFFFFF);
2659 		if (mask)
2660 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2661 		mask = (qmask >> 32);
2662 		if (mask)
2663 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2664 		break;
2665 	default:
2666 		break;
2667 	}
2668 	/* skip the flush */
2669 }
2670 
2671 /**
2672  * ixgbe_irq_enable - Enable default interrupt generation settings
2673  * @adapter: board private structure
2674  **/
2675 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2676 				    bool flush)
2677 {
2678 	struct ixgbe_hw *hw = &adapter->hw;
2679 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2680 
2681 	/* don't reenable LSC while waiting for link */
2682 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2683 		mask &= ~IXGBE_EIMS_LSC;
2684 
2685 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2686 		switch (adapter->hw.mac.type) {
2687 		case ixgbe_mac_82599EB:
2688 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
2689 			break;
2690 		case ixgbe_mac_X540:
2691 		case ixgbe_mac_X550:
2692 		case ixgbe_mac_X550EM_x:
2693 		case ixgbe_mac_x550em_a:
2694 			mask |= IXGBE_EIMS_TS;
2695 			break;
2696 		default:
2697 			break;
2698 		}
2699 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2700 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2701 	switch (adapter->hw.mac.type) {
2702 	case ixgbe_mac_82599EB:
2703 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2704 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
2705 		/* fall through */
2706 	case ixgbe_mac_X540:
2707 	case ixgbe_mac_X550:
2708 	case ixgbe_mac_X550EM_x:
2709 	case ixgbe_mac_x550em_a:
2710 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
2711 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
2712 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
2713 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
2714 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
2715 			mask |= IXGBE_EICR_GPI_SDP0_X540;
2716 		mask |= IXGBE_EIMS_ECC;
2717 		mask |= IXGBE_EIMS_MAILBOX;
2718 		break;
2719 	default:
2720 		break;
2721 	}
2722 
2723 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2724 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2725 		mask |= IXGBE_EIMS_FLOW_DIR;
2726 
2727 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2728 	if (queues)
2729 		ixgbe_irq_enable_queues(adapter, ~0);
2730 	if (flush)
2731 		IXGBE_WRITE_FLUSH(&adapter->hw);
2732 }
2733 
2734 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2735 {
2736 	struct ixgbe_adapter *adapter = data;
2737 	struct ixgbe_hw *hw = &adapter->hw;
2738 	u32 eicr;
2739 
2740 	/*
2741 	 * Workaround for Silicon errata.  Use clear-by-write instead
2742 	 * of clear-by-read.  Reading with EICS will return the
2743 	 * interrupt causes without clearing, which later be done
2744 	 * with the write to EICR.
2745 	 */
2746 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2747 
2748 	/* The lower 16bits of the EICR register are for the queue interrupts
2749 	 * which should be masked here in order to not accidentally clear them if
2750 	 * the bits are high when ixgbe_msix_other is called. There is a race
2751 	 * condition otherwise which results in possible performance loss
2752 	 * especially if the ixgbe_msix_other interrupt is triggering
2753 	 * consistently (as it would when PPS is turned on for the X540 device)
2754 	 */
2755 	eicr &= 0xFFFF0000;
2756 
2757 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2758 
2759 	if (eicr & IXGBE_EICR_LSC)
2760 		ixgbe_check_lsc(adapter);
2761 
2762 	if (eicr & IXGBE_EICR_MAILBOX)
2763 		ixgbe_msg_task(adapter);
2764 
2765 	switch (hw->mac.type) {
2766 	case ixgbe_mac_82599EB:
2767 	case ixgbe_mac_X540:
2768 	case ixgbe_mac_X550:
2769 	case ixgbe_mac_X550EM_x:
2770 	case ixgbe_mac_x550em_a:
2771 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
2772 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2773 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
2774 			ixgbe_service_event_schedule(adapter);
2775 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2776 					IXGBE_EICR_GPI_SDP0_X540);
2777 		}
2778 		if (eicr & IXGBE_EICR_ECC) {
2779 			e_info(link, "Received ECC Err, initiating reset\n");
2780 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2781 			ixgbe_service_event_schedule(adapter);
2782 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2783 		}
2784 		/* Handle Flow Director Full threshold interrupt */
2785 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2786 			int reinit_count = 0;
2787 			int i;
2788 			for (i = 0; i < adapter->num_tx_queues; i++) {
2789 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2790 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2791 						       &ring->state))
2792 					reinit_count++;
2793 			}
2794 			if (reinit_count) {
2795 				/* no more flow director interrupts until after init */
2796 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2797 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2798 				ixgbe_service_event_schedule(adapter);
2799 			}
2800 		}
2801 		ixgbe_check_sfp_event(adapter, eicr);
2802 		ixgbe_check_overtemp_event(adapter, eicr);
2803 		break;
2804 	default:
2805 		break;
2806 	}
2807 
2808 	ixgbe_check_fan_failure(adapter, eicr);
2809 
2810 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2811 		ixgbe_ptp_check_pps_event(adapter);
2812 
2813 	/* re-enable the original interrupt state, no lsc, no queues */
2814 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2815 		ixgbe_irq_enable(adapter, false, false);
2816 
2817 	return IRQ_HANDLED;
2818 }
2819 
2820 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2821 {
2822 	struct ixgbe_q_vector *q_vector = data;
2823 
2824 	/* EIAM disabled interrupts (on this vector) for us */
2825 
2826 	if (q_vector->rx.ring || q_vector->tx.ring)
2827 		napi_schedule_irqoff(&q_vector->napi);
2828 
2829 	return IRQ_HANDLED;
2830 }
2831 
2832 /**
2833  * ixgbe_poll - NAPI Rx polling callback
2834  * @napi: structure for representing this polling device
2835  * @budget: how many packets driver is allowed to clean
2836  *
2837  * This function is used for legacy and MSI, NAPI mode
2838  **/
2839 int ixgbe_poll(struct napi_struct *napi, int budget)
2840 {
2841 	struct ixgbe_q_vector *q_vector =
2842 				container_of(napi, struct ixgbe_q_vector, napi);
2843 	struct ixgbe_adapter *adapter = q_vector->adapter;
2844 	struct ixgbe_ring *ring;
2845 	int per_ring_budget, work_done = 0;
2846 	bool clean_complete = true;
2847 
2848 #ifdef CONFIG_IXGBE_DCA
2849 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2850 		ixgbe_update_dca(q_vector);
2851 #endif
2852 
2853 	ixgbe_for_each_ring(ring, q_vector->tx) {
2854 		if (!ixgbe_clean_tx_irq(q_vector, ring, budget))
2855 			clean_complete = false;
2856 	}
2857 
2858 	/* Exit if we are called by netpoll or busy polling is active */
2859 	if ((budget <= 0) || !ixgbe_qv_lock_napi(q_vector))
2860 		return budget;
2861 
2862 	/* attempt to distribute budget to each queue fairly, but don't allow
2863 	 * the budget to go below 1 because we'll exit polling */
2864 	if (q_vector->rx.count > 1)
2865 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2866 	else
2867 		per_ring_budget = budget;
2868 
2869 	ixgbe_for_each_ring(ring, q_vector->rx) {
2870 		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
2871 						 per_ring_budget);
2872 
2873 		work_done += cleaned;
2874 		if (cleaned >= per_ring_budget)
2875 			clean_complete = false;
2876 	}
2877 
2878 	ixgbe_qv_unlock_napi(q_vector);
2879 	/* If all work not completed, return budget and keep polling */
2880 	if (!clean_complete)
2881 		return budget;
2882 
2883 	/* all work done, exit the polling mode */
2884 	napi_complete_done(napi, work_done);
2885 	if (adapter->rx_itr_setting & 1)
2886 		ixgbe_set_itr(q_vector);
2887 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2888 		ixgbe_irq_enable_queues(adapter, BIT_ULL(q_vector->v_idx));
2889 
2890 	return min(work_done, budget - 1);
2891 }
2892 
2893 /**
2894  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2895  * @adapter: board private structure
2896  *
2897  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2898  * interrupts from the kernel.
2899  **/
2900 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2901 {
2902 	struct net_device *netdev = adapter->netdev;
2903 	int vector, err;
2904 	int ri = 0, ti = 0;
2905 
2906 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2907 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2908 		struct msix_entry *entry = &adapter->msix_entries[vector];
2909 
2910 		if (q_vector->tx.ring && q_vector->rx.ring) {
2911 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2912 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2913 			ti++;
2914 		} else if (q_vector->rx.ring) {
2915 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2916 				 "%s-%s-%d", netdev->name, "rx", ri++);
2917 		} else if (q_vector->tx.ring) {
2918 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2919 				 "%s-%s-%d", netdev->name, "tx", ti++);
2920 		} else {
2921 			/* skip this unused q_vector */
2922 			continue;
2923 		}
2924 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2925 				  q_vector->name, q_vector);
2926 		if (err) {
2927 			e_err(probe, "request_irq failed for MSIX interrupt "
2928 			      "Error: %d\n", err);
2929 			goto free_queue_irqs;
2930 		}
2931 		/* If Flow Director is enabled, set interrupt affinity */
2932 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2933 			/* assign the mask for this irq */
2934 			irq_set_affinity_hint(entry->vector,
2935 					      &q_vector->affinity_mask);
2936 		}
2937 	}
2938 
2939 	err = request_irq(adapter->msix_entries[vector].vector,
2940 			  ixgbe_msix_other, 0, netdev->name, adapter);
2941 	if (err) {
2942 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2943 		goto free_queue_irqs;
2944 	}
2945 
2946 	return 0;
2947 
2948 free_queue_irqs:
2949 	while (vector) {
2950 		vector--;
2951 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2952 				      NULL);
2953 		free_irq(adapter->msix_entries[vector].vector,
2954 			 adapter->q_vector[vector]);
2955 	}
2956 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2957 	pci_disable_msix(adapter->pdev);
2958 	kfree(adapter->msix_entries);
2959 	adapter->msix_entries = NULL;
2960 	return err;
2961 }
2962 
2963 /**
2964  * ixgbe_intr - legacy mode Interrupt Handler
2965  * @irq: interrupt number
2966  * @data: pointer to a network interface device structure
2967  **/
2968 static irqreturn_t ixgbe_intr(int irq, void *data)
2969 {
2970 	struct ixgbe_adapter *adapter = data;
2971 	struct ixgbe_hw *hw = &adapter->hw;
2972 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2973 	u32 eicr;
2974 
2975 	/*
2976 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2977 	 * before the read of EICR.
2978 	 */
2979 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2980 
2981 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2982 	 * therefore no explicit interrupt disable is necessary */
2983 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2984 	if (!eicr) {
2985 		/*
2986 		 * shared interrupt alert!
2987 		 * make sure interrupts are enabled because the read will
2988 		 * have disabled interrupts due to EIAM
2989 		 * finish the workaround of silicon errata on 82598.  Unmask
2990 		 * the interrupt that we masked before the EICR read.
2991 		 */
2992 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2993 			ixgbe_irq_enable(adapter, true, true);
2994 		return IRQ_NONE;	/* Not our interrupt */
2995 	}
2996 
2997 	if (eicr & IXGBE_EICR_LSC)
2998 		ixgbe_check_lsc(adapter);
2999 
3000 	switch (hw->mac.type) {
3001 	case ixgbe_mac_82599EB:
3002 		ixgbe_check_sfp_event(adapter, eicr);
3003 		/* Fall through */
3004 	case ixgbe_mac_X540:
3005 	case ixgbe_mac_X550:
3006 	case ixgbe_mac_X550EM_x:
3007 	case ixgbe_mac_x550em_a:
3008 		if (eicr & IXGBE_EICR_ECC) {
3009 			e_info(link, "Received ECC Err, initiating reset\n");
3010 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
3011 			ixgbe_service_event_schedule(adapter);
3012 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3013 		}
3014 		ixgbe_check_overtemp_event(adapter, eicr);
3015 		break;
3016 	default:
3017 		break;
3018 	}
3019 
3020 	ixgbe_check_fan_failure(adapter, eicr);
3021 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3022 		ixgbe_ptp_check_pps_event(adapter);
3023 
3024 	/* would disable interrupts here but EIAM disabled it */
3025 	napi_schedule_irqoff(&q_vector->napi);
3026 
3027 	/*
3028 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3029 	 * ixgbe_poll will re-enable the queue interrupts
3030 	 */
3031 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3032 		ixgbe_irq_enable(adapter, false, false);
3033 
3034 	return IRQ_HANDLED;
3035 }
3036 
3037 /**
3038  * ixgbe_request_irq - initialize interrupts
3039  * @adapter: board private structure
3040  *
3041  * Attempts to configure interrupts using the best available
3042  * capabilities of the hardware and kernel.
3043  **/
3044 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3045 {
3046 	struct net_device *netdev = adapter->netdev;
3047 	int err;
3048 
3049 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3050 		err = ixgbe_request_msix_irqs(adapter);
3051 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3052 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3053 				  netdev->name, adapter);
3054 	else
3055 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3056 				  netdev->name, adapter);
3057 
3058 	if (err)
3059 		e_err(probe, "request_irq failed, Error %d\n", err);
3060 
3061 	return err;
3062 }
3063 
3064 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3065 {
3066 	int vector;
3067 
3068 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3069 		free_irq(adapter->pdev->irq, adapter);
3070 		return;
3071 	}
3072 
3073 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3074 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3075 		struct msix_entry *entry = &adapter->msix_entries[vector];
3076 
3077 		/* free only the irqs that were actually requested */
3078 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3079 			continue;
3080 
3081 		/* clear the affinity_mask in the IRQ descriptor */
3082 		irq_set_affinity_hint(entry->vector, NULL);
3083 
3084 		free_irq(entry->vector, q_vector);
3085 	}
3086 
3087 	free_irq(adapter->msix_entries[vector].vector, adapter);
3088 }
3089 
3090 /**
3091  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3092  * @adapter: board private structure
3093  **/
3094 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3095 {
3096 	switch (adapter->hw.mac.type) {
3097 	case ixgbe_mac_82598EB:
3098 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3099 		break;
3100 	case ixgbe_mac_82599EB:
3101 	case ixgbe_mac_X540:
3102 	case ixgbe_mac_X550:
3103 	case ixgbe_mac_X550EM_x:
3104 	case ixgbe_mac_x550em_a:
3105 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3106 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3107 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3108 		break;
3109 	default:
3110 		break;
3111 	}
3112 	IXGBE_WRITE_FLUSH(&adapter->hw);
3113 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3114 		int vector;
3115 
3116 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3117 			synchronize_irq(adapter->msix_entries[vector].vector);
3118 
3119 		synchronize_irq(adapter->msix_entries[vector++].vector);
3120 	} else {
3121 		synchronize_irq(adapter->pdev->irq);
3122 	}
3123 }
3124 
3125 /**
3126  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3127  *
3128  **/
3129 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3130 {
3131 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3132 
3133 	ixgbe_write_eitr(q_vector);
3134 
3135 	ixgbe_set_ivar(adapter, 0, 0, 0);
3136 	ixgbe_set_ivar(adapter, 1, 0, 0);
3137 
3138 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3139 }
3140 
3141 /**
3142  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3143  * @adapter: board private structure
3144  * @ring: structure containing ring specific data
3145  *
3146  * Configure the Tx descriptor ring after a reset.
3147  **/
3148 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3149 			     struct ixgbe_ring *ring)
3150 {
3151 	struct ixgbe_hw *hw = &adapter->hw;
3152 	u64 tdba = ring->dma;
3153 	int wait_loop = 10;
3154 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3155 	u8 reg_idx = ring->reg_idx;
3156 
3157 	/* disable queue to avoid issues while updating state */
3158 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3159 	IXGBE_WRITE_FLUSH(hw);
3160 
3161 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3162 			(tdba & DMA_BIT_MASK(32)));
3163 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3164 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3165 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3166 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3167 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3168 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3169 
3170 	/*
3171 	 * set WTHRESH to encourage burst writeback, it should not be set
3172 	 * higher than 1 when:
3173 	 * - ITR is 0 as it could cause false TX hangs
3174 	 * - ITR is set to > 100k int/sec and BQL is enabled
3175 	 *
3176 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3177 	 * to or less than the number of on chip descriptors, which is
3178 	 * currently 40.
3179 	 */
3180 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3181 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3182 	else
3183 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3184 
3185 	/*
3186 	 * Setting PTHRESH to 32 both improves performance
3187 	 * and avoids a TX hang with DFP enabled
3188 	 */
3189 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3190 		   32;		/* PTHRESH = 32 */
3191 
3192 	/* reinitialize flowdirector state */
3193 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3194 		ring->atr_sample_rate = adapter->atr_sample_rate;
3195 		ring->atr_count = 0;
3196 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3197 	} else {
3198 		ring->atr_sample_rate = 0;
3199 	}
3200 
3201 	/* initialize XPS */
3202 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3203 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3204 
3205 		if (q_vector)
3206 			netif_set_xps_queue(ring->netdev,
3207 					    &q_vector->affinity_mask,
3208 					    ring->queue_index);
3209 	}
3210 
3211 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3212 
3213 	/* enable queue */
3214 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3215 
3216 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3217 	if (hw->mac.type == ixgbe_mac_82598EB &&
3218 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3219 		return;
3220 
3221 	/* poll to verify queue is enabled */
3222 	do {
3223 		usleep_range(1000, 2000);
3224 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3225 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3226 	if (!wait_loop)
3227 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
3228 }
3229 
3230 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3231 {
3232 	struct ixgbe_hw *hw = &adapter->hw;
3233 	u32 rttdcs, mtqc;
3234 	u8 tcs = netdev_get_num_tc(adapter->netdev);
3235 
3236 	if (hw->mac.type == ixgbe_mac_82598EB)
3237 		return;
3238 
3239 	/* disable the arbiter while setting MTQC */
3240 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3241 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3242 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3243 
3244 	/* set transmit pool layout */
3245 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3246 		mtqc = IXGBE_MTQC_VT_ENA;
3247 		if (tcs > 4)
3248 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3249 		else if (tcs > 1)
3250 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3251 		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3252 			mtqc |= IXGBE_MTQC_32VF;
3253 		else
3254 			mtqc |= IXGBE_MTQC_64VF;
3255 	} else {
3256 		if (tcs > 4)
3257 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3258 		else if (tcs > 1)
3259 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3260 		else
3261 			mtqc = IXGBE_MTQC_64Q_1PB;
3262 	}
3263 
3264 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3265 
3266 	/* Enable Security TX Buffer IFG for multiple pb */
3267 	if (tcs) {
3268 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3269 		sectx |= IXGBE_SECTX_DCB;
3270 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3271 	}
3272 
3273 	/* re-enable the arbiter */
3274 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3275 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3276 }
3277 
3278 /**
3279  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3280  * @adapter: board private structure
3281  *
3282  * Configure the Tx unit of the MAC after a reset.
3283  **/
3284 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3285 {
3286 	struct ixgbe_hw *hw = &adapter->hw;
3287 	u32 dmatxctl;
3288 	u32 i;
3289 
3290 	ixgbe_setup_mtqc(adapter);
3291 
3292 	if (hw->mac.type != ixgbe_mac_82598EB) {
3293 		/* DMATXCTL.EN must be before Tx queues are enabled */
3294 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3295 		dmatxctl |= IXGBE_DMATXCTL_TE;
3296 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3297 	}
3298 
3299 	/* Setup the HW Tx Head and Tail descriptor pointers */
3300 	for (i = 0; i < adapter->num_tx_queues; i++)
3301 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3302 }
3303 
3304 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3305 				 struct ixgbe_ring *ring)
3306 {
3307 	struct ixgbe_hw *hw = &adapter->hw;
3308 	u8 reg_idx = ring->reg_idx;
3309 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3310 
3311 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3312 
3313 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3314 }
3315 
3316 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3317 				  struct ixgbe_ring *ring)
3318 {
3319 	struct ixgbe_hw *hw = &adapter->hw;
3320 	u8 reg_idx = ring->reg_idx;
3321 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3322 
3323 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3324 
3325 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3326 }
3327 
3328 #ifdef CONFIG_IXGBE_DCB
3329 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3330 #else
3331 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3332 #endif
3333 {
3334 	int i;
3335 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3336 
3337 	if (adapter->ixgbe_ieee_pfc)
3338 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3339 
3340 	/*
3341 	 * We should set the drop enable bit if:
3342 	 *  SR-IOV is enabled
3343 	 *   or
3344 	 *  Number of Rx queues > 1 and flow control is disabled
3345 	 *
3346 	 *  This allows us to avoid head of line blocking for security
3347 	 *  and performance reasons.
3348 	 */
3349 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3350 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3351 		for (i = 0; i < adapter->num_rx_queues; i++)
3352 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3353 	} else {
3354 		for (i = 0; i < adapter->num_rx_queues; i++)
3355 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3356 	}
3357 }
3358 
3359 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3360 
3361 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3362 				   struct ixgbe_ring *rx_ring)
3363 {
3364 	struct ixgbe_hw *hw = &adapter->hw;
3365 	u32 srrctl;
3366 	u8 reg_idx = rx_ring->reg_idx;
3367 
3368 	if (hw->mac.type == ixgbe_mac_82598EB) {
3369 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3370 
3371 		/*
3372 		 * if VMDq is not active we must program one srrctl register
3373 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3374 		 */
3375 		reg_idx &= mask;
3376 	}
3377 
3378 	/* configure header buffer length, needed for RSC */
3379 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3380 
3381 	/* configure the packet buffer length */
3382 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3383 
3384 	/* configure descriptor type */
3385 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3386 
3387 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3388 }
3389 
3390 /**
3391  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3392  * @adapter: device handle
3393  *
3394  *  - 82598/82599/X540:     128
3395  *  - X550(non-SRIOV mode): 512
3396  *  - X550(SRIOV mode):     64
3397  */
3398 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3399 {
3400 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3401 		return 128;
3402 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3403 		return 64;
3404 	else
3405 		return 512;
3406 }
3407 
3408 /**
3409  * ixgbe_store_reta - Write the RETA table to HW
3410  * @adapter: device handle
3411  *
3412  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3413  */
3414 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3415 {
3416 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3417 	struct ixgbe_hw *hw = &adapter->hw;
3418 	u32 reta = 0;
3419 	u32 indices_multi;
3420 	u8 *indir_tbl = adapter->rss_indir_tbl;
3421 
3422 	/* Fill out the redirection table as follows:
3423 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3424 	 *    indices.
3425 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3426 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3427 	 */
3428 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3429 		indices_multi = 0x11;
3430 	else
3431 		indices_multi = 0x1;
3432 
3433 	/* Write redirection table to HW */
3434 	for (i = 0; i < reta_entries; i++) {
3435 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3436 		if ((i & 3) == 3) {
3437 			if (i < 128)
3438 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3439 			else
3440 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3441 						reta);
3442 			reta = 0;
3443 		}
3444 	}
3445 }
3446 
3447 /**
3448  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3449  * @adapter: device handle
3450  *
3451  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3452  */
3453 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3454 {
3455 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3456 	struct ixgbe_hw *hw = &adapter->hw;
3457 	u32 vfreta = 0;
3458 	unsigned int pf_pool = adapter->num_vfs;
3459 
3460 	/* Write redirection table to HW */
3461 	for (i = 0; i < reta_entries; i++) {
3462 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3463 		if ((i & 3) == 3) {
3464 			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3465 					vfreta);
3466 			vfreta = 0;
3467 		}
3468 	}
3469 }
3470 
3471 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3472 {
3473 	struct ixgbe_hw *hw = &adapter->hw;
3474 	u32 i, j;
3475 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3476 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3477 
3478 	/* Program table for at least 2 queues w/ SR-IOV so that VFs can
3479 	 * make full use of any rings they may have.  We will use the
3480 	 * PSRTYPE register to control how many rings we use within the PF.
3481 	 */
3482 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3483 		rss_i = 2;
3484 
3485 	/* Fill out hash function seeds */
3486 	for (i = 0; i < 10; i++)
3487 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3488 
3489 	/* Fill out redirection table */
3490 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3491 
3492 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3493 		if (j == rss_i)
3494 			j = 0;
3495 
3496 		adapter->rss_indir_tbl[i] = j;
3497 	}
3498 
3499 	ixgbe_store_reta(adapter);
3500 }
3501 
3502 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3503 {
3504 	struct ixgbe_hw *hw = &adapter->hw;
3505 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3506 	unsigned int pf_pool = adapter->num_vfs;
3507 	int i, j;
3508 
3509 	/* Fill out hash function seeds */
3510 	for (i = 0; i < 10; i++)
3511 		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool),
3512 				adapter->rss_key[i]);
3513 
3514 	/* Fill out the redirection table */
3515 	for (i = 0, j = 0; i < 64; i++, j++) {
3516 		if (j == rss_i)
3517 			j = 0;
3518 
3519 		adapter->rss_indir_tbl[i] = j;
3520 	}
3521 
3522 	ixgbe_store_vfreta(adapter);
3523 }
3524 
3525 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3526 {
3527 	struct ixgbe_hw *hw = &adapter->hw;
3528 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3529 	u32 rxcsum;
3530 
3531 	/* Disable indicating checksum in descriptor, enables RSS hash */
3532 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3533 	rxcsum |= IXGBE_RXCSUM_PCSD;
3534 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3535 
3536 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3537 		if (adapter->ring_feature[RING_F_RSS].mask)
3538 			mrqc = IXGBE_MRQC_RSSEN;
3539 	} else {
3540 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3541 
3542 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3543 			if (tcs > 4)
3544 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3545 			else if (tcs > 1)
3546 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3547 			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3548 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3549 			else
3550 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3551 		} else {
3552 			if (tcs > 4)
3553 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3554 			else if (tcs > 1)
3555 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3556 			else
3557 				mrqc = IXGBE_MRQC_RSSEN;
3558 		}
3559 	}
3560 
3561 	/* Perform hash on these packet types */
3562 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3563 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3564 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3565 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3566 
3567 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3568 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3569 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3570 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3571 
3572 	netdev_rss_key_fill(adapter->rss_key, sizeof(adapter->rss_key));
3573 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3574 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3575 		unsigned int pf_pool = adapter->num_vfs;
3576 
3577 		/* Enable VF RSS mode */
3578 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3579 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3580 
3581 		/* Setup RSS through the VF registers */
3582 		ixgbe_setup_vfreta(adapter);
3583 		vfmrqc = IXGBE_MRQC_RSSEN;
3584 		vfmrqc |= rss_field;
3585 		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3586 	} else {
3587 		ixgbe_setup_reta(adapter);
3588 		mrqc |= rss_field;
3589 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3590 	}
3591 }
3592 
3593 /**
3594  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3595  * @adapter:    address of board private structure
3596  * @index:      index of ring to set
3597  **/
3598 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3599 				   struct ixgbe_ring *ring)
3600 {
3601 	struct ixgbe_hw *hw = &adapter->hw;
3602 	u32 rscctrl;
3603 	u8 reg_idx = ring->reg_idx;
3604 
3605 	if (!ring_is_rsc_enabled(ring))
3606 		return;
3607 
3608 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3609 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3610 	/*
3611 	 * we must limit the number of descriptors so that the
3612 	 * total size of max desc * buf_len is not greater
3613 	 * than 65536
3614 	 */
3615 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3616 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3617 }
3618 
3619 #define IXGBE_MAX_RX_DESC_POLL 10
3620 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3621 				       struct ixgbe_ring *ring)
3622 {
3623 	struct ixgbe_hw *hw = &adapter->hw;
3624 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3625 	u32 rxdctl;
3626 	u8 reg_idx = ring->reg_idx;
3627 
3628 	if (ixgbe_removed(hw->hw_addr))
3629 		return;
3630 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3631 	if (hw->mac.type == ixgbe_mac_82598EB &&
3632 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3633 		return;
3634 
3635 	do {
3636 		usleep_range(1000, 2000);
3637 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3638 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3639 
3640 	if (!wait_loop) {
3641 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3642 		      "the polling period\n", reg_idx);
3643 	}
3644 }
3645 
3646 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3647 			    struct ixgbe_ring *ring)
3648 {
3649 	struct ixgbe_hw *hw = &adapter->hw;
3650 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3651 	u32 rxdctl;
3652 	u8 reg_idx = ring->reg_idx;
3653 
3654 	if (ixgbe_removed(hw->hw_addr))
3655 		return;
3656 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3657 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3658 
3659 	/* write value back with RXDCTL.ENABLE bit cleared */
3660 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3661 
3662 	if (hw->mac.type == ixgbe_mac_82598EB &&
3663 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3664 		return;
3665 
3666 	/* the hardware may take up to 100us to really disable the rx queue */
3667 	do {
3668 		udelay(10);
3669 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3670 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3671 
3672 	if (!wait_loop) {
3673 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3674 		      "the polling period\n", reg_idx);
3675 	}
3676 }
3677 
3678 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3679 			     struct ixgbe_ring *ring)
3680 {
3681 	struct ixgbe_hw *hw = &adapter->hw;
3682 	u64 rdba = ring->dma;
3683 	u32 rxdctl;
3684 	u8 reg_idx = ring->reg_idx;
3685 
3686 	/* disable queue to avoid issues while updating state */
3687 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3688 	ixgbe_disable_rx_queue(adapter, ring);
3689 
3690 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3691 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3692 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3693 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3694 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
3695 	IXGBE_WRITE_FLUSH(hw);
3696 
3697 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3698 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3699 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3700 
3701 	ixgbe_configure_srrctl(adapter, ring);
3702 	ixgbe_configure_rscctl(adapter, ring);
3703 
3704 	if (hw->mac.type == ixgbe_mac_82598EB) {
3705 		/*
3706 		 * enable cache line friendly hardware writes:
3707 		 * PTHRESH=32 descriptors (half the internal cache),
3708 		 * this also removes ugly rx_no_buffer_count increment
3709 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3710 		 * WTHRESH=8 burst writeback up to two cache lines
3711 		 */
3712 		rxdctl &= ~0x3FFFFF;
3713 		rxdctl |=  0x080420;
3714 	}
3715 
3716 	/* enable receive descriptor ring */
3717 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3718 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3719 
3720 	ixgbe_rx_desc_queue_enable(adapter, ring);
3721 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3722 }
3723 
3724 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3725 {
3726 	struct ixgbe_hw *hw = &adapter->hw;
3727 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3728 	u16 pool;
3729 
3730 	/* PSRTYPE must be initialized in non 82598 adapters */
3731 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3732 		      IXGBE_PSRTYPE_UDPHDR |
3733 		      IXGBE_PSRTYPE_IPV4HDR |
3734 		      IXGBE_PSRTYPE_L2HDR |
3735 		      IXGBE_PSRTYPE_IPV6HDR;
3736 
3737 	if (hw->mac.type == ixgbe_mac_82598EB)
3738 		return;
3739 
3740 	if (rss_i > 3)
3741 		psrtype |= 2u << 29;
3742 	else if (rss_i > 1)
3743 		psrtype |= 1u << 29;
3744 
3745 	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3746 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3747 }
3748 
3749 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3750 {
3751 	struct ixgbe_hw *hw = &adapter->hw;
3752 	u32 reg_offset, vf_shift;
3753 	u32 gcr_ext, vmdctl;
3754 	int i;
3755 
3756 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3757 		return;
3758 
3759 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3760 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3761 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3762 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3763 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3764 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3765 
3766 	vf_shift = VMDQ_P(0) % 32;
3767 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3768 
3769 	/* Enable only the PF's pool for Tx/Rx */
3770 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
3771 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3772 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
3773 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3774 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
3775 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3776 
3777 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3778 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3779 
3780 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
3781 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
3782 
3783 	/*
3784 	 * Set up VF register offsets for selected VT Mode,
3785 	 * i.e. 32 or 64 VFs for SR-IOV
3786 	 */
3787 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3788 	case IXGBE_82599_VMDQ_8Q_MASK:
3789 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3790 		break;
3791 	case IXGBE_82599_VMDQ_4Q_MASK:
3792 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3793 		break;
3794 	default:
3795 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3796 		break;
3797 	}
3798 
3799 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3800 
3801 	for (i = 0; i < adapter->num_vfs; i++) {
3802 		/* configure spoof checking */
3803 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
3804 					  adapter->vfinfo[i].spoofchk_enabled);
3805 
3806 		/* Enable/Disable RSS query feature  */
3807 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
3808 					  adapter->vfinfo[i].rss_query_enabled);
3809 	}
3810 }
3811 
3812 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3813 {
3814 	struct ixgbe_hw *hw = &adapter->hw;
3815 	struct net_device *netdev = adapter->netdev;
3816 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3817 	struct ixgbe_ring *rx_ring;
3818 	int i;
3819 	u32 mhadd, hlreg0;
3820 
3821 #ifdef IXGBE_FCOE
3822 	/* adjust max frame to be able to do baby jumbo for FCoE */
3823 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3824 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3825 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3826 
3827 #endif /* IXGBE_FCOE */
3828 
3829 	/* adjust max frame to be at least the size of a standard frame */
3830 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3831 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3832 
3833 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3834 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3835 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3836 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3837 
3838 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3839 	}
3840 
3841 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3842 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3843 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3844 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3845 
3846 	/*
3847 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3848 	 * the Base and Length of the Rx Descriptor Ring
3849 	 */
3850 	for (i = 0; i < adapter->num_rx_queues; i++) {
3851 		rx_ring = adapter->rx_ring[i];
3852 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3853 			set_ring_rsc_enabled(rx_ring);
3854 		else
3855 			clear_ring_rsc_enabled(rx_ring);
3856 	}
3857 }
3858 
3859 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3860 {
3861 	struct ixgbe_hw *hw = &adapter->hw;
3862 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3863 
3864 	switch (hw->mac.type) {
3865 	case ixgbe_mac_82598EB:
3866 		/*
3867 		 * For VMDq support of different descriptor types or
3868 		 * buffer sizes through the use of multiple SRRCTL
3869 		 * registers, RDRXCTL.MVMEN must be set to 1
3870 		 *
3871 		 * also, the manual doesn't mention it clearly but DCA hints
3872 		 * will only use queue 0's tags unless this bit is set.  Side
3873 		 * effects of setting this bit are only that SRRCTL must be
3874 		 * fully programmed [0..15]
3875 		 */
3876 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3877 		break;
3878 	case ixgbe_mac_X550:
3879 	case ixgbe_mac_X550EM_x:
3880 	case ixgbe_mac_x550em_a:
3881 		if (adapter->num_vfs)
3882 			rdrxctl |= IXGBE_RDRXCTL_PSP;
3883 		/* fall through for older HW */
3884 	case ixgbe_mac_82599EB:
3885 	case ixgbe_mac_X540:
3886 		/* Disable RSC for ACK packets */
3887 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3888 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3889 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3890 		/* hardware requires some bits to be set by default */
3891 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3892 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3893 		break;
3894 	default:
3895 		/* We should do nothing since we don't know this hardware */
3896 		return;
3897 	}
3898 
3899 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3900 }
3901 
3902 /**
3903  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3904  * @adapter: board private structure
3905  *
3906  * Configure the Rx unit of the MAC after a reset.
3907  **/
3908 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3909 {
3910 	struct ixgbe_hw *hw = &adapter->hw;
3911 	int i;
3912 	u32 rxctrl, rfctl;
3913 
3914 	/* disable receives while setting up the descriptors */
3915 	hw->mac.ops.disable_rx(hw);
3916 
3917 	ixgbe_setup_psrtype(adapter);
3918 	ixgbe_setup_rdrxctl(adapter);
3919 
3920 	/* RSC Setup */
3921 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
3922 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
3923 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
3924 		rfctl |= IXGBE_RFCTL_RSC_DIS;
3925 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3926 
3927 	/* Program registers for the distribution of queues */
3928 	ixgbe_setup_mrqc(adapter);
3929 
3930 	/* set_rx_buffer_len must be called before ring initialization */
3931 	ixgbe_set_rx_buffer_len(adapter);
3932 
3933 	/*
3934 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3935 	 * the Base and Length of the Rx Descriptor Ring
3936 	 */
3937 	for (i = 0; i < adapter->num_rx_queues; i++)
3938 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3939 
3940 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3941 	/* disable drop enable for 82598 parts */
3942 	if (hw->mac.type == ixgbe_mac_82598EB)
3943 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3944 
3945 	/* enable all receives */
3946 	rxctrl |= IXGBE_RXCTRL_RXEN;
3947 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3948 }
3949 
3950 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3951 				 __be16 proto, u16 vid)
3952 {
3953 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3954 	struct ixgbe_hw *hw = &adapter->hw;
3955 
3956 	/* add VID to filter table */
3957 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
3958 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
3959 
3960 	set_bit(vid, adapter->active_vlans);
3961 
3962 	return 0;
3963 }
3964 
3965 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
3966 {
3967 	u32 vlvf;
3968 	int idx;
3969 
3970 	/* short cut the special case */
3971 	if (vlan == 0)
3972 		return 0;
3973 
3974 	/* Search for the vlan id in the VLVF entries */
3975 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
3976 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
3977 		if ((vlvf & VLAN_VID_MASK) == vlan)
3978 			break;
3979 	}
3980 
3981 	return idx;
3982 }
3983 
3984 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
3985 {
3986 	struct ixgbe_hw *hw = &adapter->hw;
3987 	u32 bits, word;
3988 	int idx;
3989 
3990 	idx = ixgbe_find_vlvf_entry(hw, vid);
3991 	if (!idx)
3992 		return;
3993 
3994 	/* See if any other pools are set for this VLAN filter
3995 	 * entry other than the PF.
3996 	 */
3997 	word = idx * 2 + (VMDQ_P(0) / 32);
3998 	bits = ~BIT(VMDQ_P(0) % 32);
3999 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4000 
4001 	/* Disable the filter so this falls into the default pool. */
4002 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4003 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4004 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4005 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4006 	}
4007 }
4008 
4009 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4010 				  __be16 proto, u16 vid)
4011 {
4012 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4013 	struct ixgbe_hw *hw = &adapter->hw;
4014 
4015 	/* remove VID from filter table */
4016 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4017 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4018 
4019 	clear_bit(vid, adapter->active_vlans);
4020 
4021 	return 0;
4022 }
4023 
4024 /**
4025  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4026  * @adapter: driver data
4027  */
4028 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4029 {
4030 	struct ixgbe_hw *hw = &adapter->hw;
4031 	u32 vlnctrl;
4032 	int i, j;
4033 
4034 	switch (hw->mac.type) {
4035 	case ixgbe_mac_82598EB:
4036 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4037 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4038 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4039 		break;
4040 	case ixgbe_mac_82599EB:
4041 	case ixgbe_mac_X540:
4042 	case ixgbe_mac_X550:
4043 	case ixgbe_mac_X550EM_x:
4044 	case ixgbe_mac_x550em_a:
4045 		for (i = 0; i < adapter->num_rx_queues; i++) {
4046 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4047 
4048 			if (ring->l2_accel_priv)
4049 				continue;
4050 			j = ring->reg_idx;
4051 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4052 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4053 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4054 		}
4055 		break;
4056 	default:
4057 		break;
4058 	}
4059 }
4060 
4061 /**
4062  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4063  * @adapter: driver data
4064  */
4065 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4066 {
4067 	struct ixgbe_hw *hw = &adapter->hw;
4068 	u32 vlnctrl;
4069 	int i, j;
4070 
4071 	switch (hw->mac.type) {
4072 	case ixgbe_mac_82598EB:
4073 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4074 		vlnctrl |= IXGBE_VLNCTRL_VME;
4075 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4076 		break;
4077 	case ixgbe_mac_82599EB:
4078 	case ixgbe_mac_X540:
4079 	case ixgbe_mac_X550:
4080 	case ixgbe_mac_X550EM_x:
4081 	case ixgbe_mac_x550em_a:
4082 		for (i = 0; i < adapter->num_rx_queues; i++) {
4083 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4084 
4085 			if (ring->l2_accel_priv)
4086 				continue;
4087 			j = ring->reg_idx;
4088 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4089 			vlnctrl |= IXGBE_RXDCTL_VME;
4090 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4091 		}
4092 		break;
4093 	default:
4094 		break;
4095 	}
4096 }
4097 
4098 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4099 {
4100 	struct ixgbe_hw *hw = &adapter->hw;
4101 	u32 vlnctrl, i;
4102 
4103 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4104 
4105 	switch (hw->mac.type) {
4106 	case ixgbe_mac_82599EB:
4107 	case ixgbe_mac_X540:
4108 	case ixgbe_mac_X550:
4109 	case ixgbe_mac_X550EM_x:
4110 	case ixgbe_mac_x550em_a:
4111 	default:
4112 		if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED)
4113 			break;
4114 		/* fall through */
4115 	case ixgbe_mac_82598EB:
4116 		/* legacy case, we can just disable VLAN filtering */
4117 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4118 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4119 		return;
4120 	}
4121 
4122 	/* We are already in VLAN promisc, nothing to do */
4123 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4124 		return;
4125 
4126 	/* Set flag so we don't redo unnecessary work */
4127 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4128 
4129 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4130 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4131 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4132 
4133 	/* Add PF to all active pools */
4134 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4135 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4136 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4137 
4138 		vlvfb |= BIT(VMDQ_P(0) % 32);
4139 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4140 	}
4141 
4142 	/* Set all bits in the VLAN filter table array */
4143 	for (i = hw->mac.vft_size; i--;)
4144 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4145 }
4146 
4147 #define VFTA_BLOCK_SIZE 8
4148 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4149 {
4150 	struct ixgbe_hw *hw = &adapter->hw;
4151 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4152 	u32 vid_start = vfta_offset * 32;
4153 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4154 	u32 i, vid, word, bits;
4155 
4156 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4157 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4158 
4159 		/* pull VLAN ID from VLVF */
4160 		vid = vlvf & VLAN_VID_MASK;
4161 
4162 		/* only concern outselves with a certain range */
4163 		if (vid < vid_start || vid >= vid_end)
4164 			continue;
4165 
4166 		if (vlvf) {
4167 			/* record VLAN ID in VFTA */
4168 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4169 
4170 			/* if PF is part of this then continue */
4171 			if (test_bit(vid, adapter->active_vlans))
4172 				continue;
4173 		}
4174 
4175 		/* remove PF from the pool */
4176 		word = i * 2 + VMDQ_P(0) / 32;
4177 		bits = ~BIT(VMDQ_P(0) % 32);
4178 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4179 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4180 	}
4181 
4182 	/* extract values from active_vlans and write back to VFTA */
4183 	for (i = VFTA_BLOCK_SIZE; i--;) {
4184 		vid = (vfta_offset + i) * 32;
4185 		word = vid / BITS_PER_LONG;
4186 		bits = vid % BITS_PER_LONG;
4187 
4188 		vfta[i] |= adapter->active_vlans[word] >> bits;
4189 
4190 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4191 	}
4192 }
4193 
4194 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4195 {
4196 	struct ixgbe_hw *hw = &adapter->hw;
4197 	u32 vlnctrl, i;
4198 
4199 	/* Set VLAN filtering to enabled */
4200 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4201 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4202 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4203 
4204 	switch (hw->mac.type) {
4205 	case ixgbe_mac_82599EB:
4206 	case ixgbe_mac_X540:
4207 	case ixgbe_mac_X550:
4208 	case ixgbe_mac_X550EM_x:
4209 	case ixgbe_mac_x550em_a:
4210 	default:
4211 		if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED)
4212 			break;
4213 		/* fall through */
4214 	case ixgbe_mac_82598EB:
4215 		return;
4216 	}
4217 
4218 	/* We are not in VLAN promisc, nothing to do */
4219 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4220 		return;
4221 
4222 	/* Set flag so we don't redo unnecessary work */
4223 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4224 
4225 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4226 		ixgbe_scrub_vfta(adapter, i);
4227 }
4228 
4229 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4230 {
4231 	u16 vid = 1;
4232 
4233 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4234 
4235 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4236 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4237 }
4238 
4239 /**
4240  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4241  * @netdev: network interface device structure
4242  *
4243  * Writes multicast address list to the MTA hash table.
4244  * Returns: -ENOMEM on failure
4245  *                0 on no addresses written
4246  *                X on writing X addresses to MTA
4247  **/
4248 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4249 {
4250 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4251 	struct ixgbe_hw *hw = &adapter->hw;
4252 
4253 	if (!netif_running(netdev))
4254 		return 0;
4255 
4256 	if (hw->mac.ops.update_mc_addr_list)
4257 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4258 	else
4259 		return -ENOMEM;
4260 
4261 #ifdef CONFIG_PCI_IOV
4262 	ixgbe_restore_vf_multicasts(adapter);
4263 #endif
4264 
4265 	return netdev_mc_count(netdev);
4266 }
4267 
4268 #ifdef CONFIG_PCI_IOV
4269 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4270 {
4271 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4272 	struct ixgbe_hw *hw = &adapter->hw;
4273 	int i;
4274 
4275 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4276 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4277 
4278 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4279 			hw->mac.ops.set_rar(hw, i,
4280 					    mac_table->addr,
4281 					    mac_table->pool,
4282 					    IXGBE_RAH_AV);
4283 		else
4284 			hw->mac.ops.clear_rar(hw, i);
4285 	}
4286 }
4287 
4288 #endif
4289 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4290 {
4291 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4292 	struct ixgbe_hw *hw = &adapter->hw;
4293 	int i;
4294 
4295 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4296 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4297 			continue;
4298 
4299 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4300 
4301 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4302 			hw->mac.ops.set_rar(hw, i,
4303 					    mac_table->addr,
4304 					    mac_table->pool,
4305 					    IXGBE_RAH_AV);
4306 		else
4307 			hw->mac.ops.clear_rar(hw, i);
4308 	}
4309 }
4310 
4311 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4312 {
4313 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4314 	struct ixgbe_hw *hw = &adapter->hw;
4315 	int i;
4316 
4317 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4318 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4319 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4320 	}
4321 
4322 	ixgbe_sync_mac_table(adapter);
4323 }
4324 
4325 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4326 {
4327 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4328 	struct ixgbe_hw *hw = &adapter->hw;
4329 	int i, count = 0;
4330 
4331 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4332 		/* do not count default RAR as available */
4333 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4334 			continue;
4335 
4336 		/* only count unused and addresses that belong to us */
4337 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4338 			if (mac_table->pool != pool)
4339 				continue;
4340 		}
4341 
4342 		count++;
4343 	}
4344 
4345 	return count;
4346 }
4347 
4348 /* this function destroys the first RAR entry */
4349 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4350 {
4351 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4352 	struct ixgbe_hw *hw = &adapter->hw;
4353 
4354 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4355 	mac_table->pool = VMDQ_P(0);
4356 
4357 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4358 
4359 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4360 			    IXGBE_RAH_AV);
4361 }
4362 
4363 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4364 			 const u8 *addr, u16 pool)
4365 {
4366 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4367 	struct ixgbe_hw *hw = &adapter->hw;
4368 	int i;
4369 
4370 	if (is_zero_ether_addr(addr))
4371 		return -EINVAL;
4372 
4373 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4374 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4375 			continue;
4376 
4377 		ether_addr_copy(mac_table->addr, addr);
4378 		mac_table->pool = pool;
4379 
4380 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4381 				    IXGBE_MAC_STATE_IN_USE;
4382 
4383 		ixgbe_sync_mac_table(adapter);
4384 
4385 		return i;
4386 	}
4387 
4388 	return -ENOMEM;
4389 }
4390 
4391 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4392 			 const u8 *addr, u16 pool)
4393 {
4394 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4395 	struct ixgbe_hw *hw = &adapter->hw;
4396 	int i;
4397 
4398 	if (is_zero_ether_addr(addr))
4399 		return -EINVAL;
4400 
4401 	/* search table for addr, if found clear IN_USE flag and sync */
4402 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4403 		/* we can only delete an entry if it is in use */
4404 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4405 			continue;
4406 		/* we only care about entries that belong to the given pool */
4407 		if (mac_table->pool != pool)
4408 			continue;
4409 		/* we only care about a specific MAC address */
4410 		if (!ether_addr_equal(addr, mac_table->addr))
4411 			continue;
4412 
4413 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4414 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4415 
4416 		ixgbe_sync_mac_table(adapter);
4417 
4418 		return 0;
4419 	}
4420 
4421 	return -ENOMEM;
4422 }
4423 /**
4424  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4425  * @netdev: network interface device structure
4426  *
4427  * Writes unicast address list to the RAR table.
4428  * Returns: -ENOMEM on failure/insufficient address space
4429  *                0 on no addresses written
4430  *                X on writing X addresses to the RAR table
4431  **/
4432 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4433 {
4434 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4435 	int count = 0;
4436 
4437 	/* return ENOMEM indicating insufficient memory for addresses */
4438 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter, vfn))
4439 		return -ENOMEM;
4440 
4441 	if (!netdev_uc_empty(netdev)) {
4442 		struct netdev_hw_addr *ha;
4443 		netdev_for_each_uc_addr(ha, netdev) {
4444 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4445 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4446 			count++;
4447 		}
4448 	}
4449 	return count;
4450 }
4451 
4452 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4453 {
4454 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4455 	int ret;
4456 
4457 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4458 
4459 	return min_t(int, ret, 0);
4460 }
4461 
4462 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4463 {
4464 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4465 
4466 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4467 
4468 	return 0;
4469 }
4470 
4471 /**
4472  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4473  * @netdev: network interface device structure
4474  *
4475  * The set_rx_method entry point is called whenever the unicast/multicast
4476  * address list or the network interface flags are updated.  This routine is
4477  * responsible for configuring the hardware for proper unicast, multicast and
4478  * promiscuous mode.
4479  **/
4480 void ixgbe_set_rx_mode(struct net_device *netdev)
4481 {
4482 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4483 	struct ixgbe_hw *hw = &adapter->hw;
4484 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4485 	netdev_features_t features = netdev->features;
4486 	int count;
4487 
4488 	/* Check for Promiscuous and All Multicast modes */
4489 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4490 
4491 	/* set all bits that we expect to always be set */
4492 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4493 	fctrl |= IXGBE_FCTRL_BAM;
4494 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4495 	fctrl |= IXGBE_FCTRL_PMCF;
4496 
4497 	/* clear the bits we are changing the status of */
4498 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4499 	if (netdev->flags & IFF_PROMISC) {
4500 		hw->addr_ctrl.user_set_promisc = true;
4501 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4502 		vmolr |= IXGBE_VMOLR_MPE;
4503 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4504 	} else {
4505 		if (netdev->flags & IFF_ALLMULTI) {
4506 			fctrl |= IXGBE_FCTRL_MPE;
4507 			vmolr |= IXGBE_VMOLR_MPE;
4508 		}
4509 		hw->addr_ctrl.user_set_promisc = false;
4510 	}
4511 
4512 	/*
4513 	 * Write addresses to available RAR registers, if there is not
4514 	 * sufficient space to store all the addresses then enable
4515 	 * unicast promiscuous mode
4516 	 */
4517 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4518 		fctrl |= IXGBE_FCTRL_UPE;
4519 		vmolr |= IXGBE_VMOLR_ROPE;
4520 	}
4521 
4522 	/* Write addresses to the MTA, if the attempt fails
4523 	 * then we should just turn on promiscuous mode so
4524 	 * that we can at least receive multicast traffic
4525 	 */
4526 	count = ixgbe_write_mc_addr_list(netdev);
4527 	if (count < 0) {
4528 		fctrl |= IXGBE_FCTRL_MPE;
4529 		vmolr |= IXGBE_VMOLR_MPE;
4530 	} else if (count) {
4531 		vmolr |= IXGBE_VMOLR_ROMPE;
4532 	}
4533 
4534 	if (hw->mac.type != ixgbe_mac_82598EB) {
4535 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4536 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4537 			   IXGBE_VMOLR_ROPE);
4538 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4539 	}
4540 
4541 	/* This is useful for sniffing bad packets. */
4542 	if (features & NETIF_F_RXALL) {
4543 		/* UPE and MPE will be handled by normal PROMISC logic
4544 		 * in e1000e_set_rx_mode */
4545 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4546 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4547 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4548 
4549 		fctrl &= ~(IXGBE_FCTRL_DPF);
4550 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4551 	}
4552 
4553 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4554 
4555 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4556 		ixgbe_vlan_strip_enable(adapter);
4557 	else
4558 		ixgbe_vlan_strip_disable(adapter);
4559 
4560 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4561 		ixgbe_vlan_promisc_disable(adapter);
4562 	else
4563 		ixgbe_vlan_promisc_enable(adapter);
4564 }
4565 
4566 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4567 {
4568 	int q_idx;
4569 
4570 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4571 		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
4572 		napi_enable(&adapter->q_vector[q_idx]->napi);
4573 	}
4574 }
4575 
4576 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4577 {
4578 	int q_idx;
4579 
4580 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4581 		napi_disable(&adapter->q_vector[q_idx]->napi);
4582 		while (!ixgbe_qv_disable(adapter->q_vector[q_idx])) {
4583 			pr_info("QV %d locked\n", q_idx);
4584 			usleep_range(1000, 20000);
4585 		}
4586 	}
4587 }
4588 
4589 static void ixgbe_clear_vxlan_port(struct ixgbe_adapter *adapter)
4590 {
4591 	switch (adapter->hw.mac.type) {
4592 	case ixgbe_mac_X550:
4593 	case ixgbe_mac_X550EM_x:
4594 	case ixgbe_mac_x550em_a:
4595 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_VXLANCTRL, 0);
4596 		adapter->vxlan_port = 0;
4597 		break;
4598 	default:
4599 		break;
4600 	}
4601 }
4602 
4603 #ifdef CONFIG_IXGBE_DCB
4604 /**
4605  * ixgbe_configure_dcb - Configure DCB hardware
4606  * @adapter: ixgbe adapter struct
4607  *
4608  * This is called by the driver on open to configure the DCB hardware.
4609  * This is also called by the gennetlink interface when reconfiguring
4610  * the DCB state.
4611  */
4612 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4613 {
4614 	struct ixgbe_hw *hw = &adapter->hw;
4615 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4616 
4617 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4618 		if (hw->mac.type == ixgbe_mac_82598EB)
4619 			netif_set_gso_max_size(adapter->netdev, 65536);
4620 		return;
4621 	}
4622 
4623 	if (hw->mac.type == ixgbe_mac_82598EB)
4624 		netif_set_gso_max_size(adapter->netdev, 32768);
4625 
4626 #ifdef IXGBE_FCOE
4627 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4628 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4629 #endif
4630 
4631 	/* reconfigure the hardware */
4632 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4633 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4634 						DCB_TX_CONFIG);
4635 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4636 						DCB_RX_CONFIG);
4637 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4638 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4639 		ixgbe_dcb_hw_ets(&adapter->hw,
4640 				 adapter->ixgbe_ieee_ets,
4641 				 max_frame);
4642 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4643 					adapter->ixgbe_ieee_pfc->pfc_en,
4644 					adapter->ixgbe_ieee_ets->prio_tc);
4645 	}
4646 
4647 	/* Enable RSS Hash per TC */
4648 	if (hw->mac.type != ixgbe_mac_82598EB) {
4649 		u32 msb = 0;
4650 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4651 
4652 		while (rss_i) {
4653 			msb++;
4654 			rss_i >>= 1;
4655 		}
4656 
4657 		/* write msb to all 8 TCs in one write */
4658 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4659 	}
4660 }
4661 #endif
4662 
4663 /* Additional bittime to account for IXGBE framing */
4664 #define IXGBE_ETH_FRAMING 20
4665 
4666 /**
4667  * ixgbe_hpbthresh - calculate high water mark for flow control
4668  *
4669  * @adapter: board private structure to calculate for
4670  * @pb: packet buffer to calculate
4671  */
4672 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4673 {
4674 	struct ixgbe_hw *hw = &adapter->hw;
4675 	struct net_device *dev = adapter->netdev;
4676 	int link, tc, kb, marker;
4677 	u32 dv_id, rx_pba;
4678 
4679 	/* Calculate max LAN frame size */
4680 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4681 
4682 #ifdef IXGBE_FCOE
4683 	/* FCoE traffic class uses FCOE jumbo frames */
4684 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4685 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4686 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4687 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4688 #endif
4689 
4690 	/* Calculate delay value for device */
4691 	switch (hw->mac.type) {
4692 	case ixgbe_mac_X540:
4693 	case ixgbe_mac_X550:
4694 	case ixgbe_mac_X550EM_x:
4695 	case ixgbe_mac_x550em_a:
4696 		dv_id = IXGBE_DV_X540(link, tc);
4697 		break;
4698 	default:
4699 		dv_id = IXGBE_DV(link, tc);
4700 		break;
4701 	}
4702 
4703 	/* Loopback switch introduces additional latency */
4704 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4705 		dv_id += IXGBE_B2BT(tc);
4706 
4707 	/* Delay value is calculated in bit times convert to KB */
4708 	kb = IXGBE_BT2KB(dv_id);
4709 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4710 
4711 	marker = rx_pba - kb;
4712 
4713 	/* It is possible that the packet buffer is not large enough
4714 	 * to provide required headroom. In this case throw an error
4715 	 * to user and a do the best we can.
4716 	 */
4717 	if (marker < 0) {
4718 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4719 			    "headroom to support flow control."
4720 			    "Decrease MTU or number of traffic classes\n", pb);
4721 		marker = tc + 1;
4722 	}
4723 
4724 	return marker;
4725 }
4726 
4727 /**
4728  * ixgbe_lpbthresh - calculate low water mark for for flow control
4729  *
4730  * @adapter: board private structure to calculate for
4731  * @pb: packet buffer to calculate
4732  */
4733 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4734 {
4735 	struct ixgbe_hw *hw = &adapter->hw;
4736 	struct net_device *dev = adapter->netdev;
4737 	int tc;
4738 	u32 dv_id;
4739 
4740 	/* Calculate max LAN frame size */
4741 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4742 
4743 #ifdef IXGBE_FCOE
4744 	/* FCoE traffic class uses FCOE jumbo frames */
4745 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4746 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4747 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4748 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4749 #endif
4750 
4751 	/* Calculate delay value for device */
4752 	switch (hw->mac.type) {
4753 	case ixgbe_mac_X540:
4754 	case ixgbe_mac_X550:
4755 	case ixgbe_mac_X550EM_x:
4756 	case ixgbe_mac_x550em_a:
4757 		dv_id = IXGBE_LOW_DV_X540(tc);
4758 		break;
4759 	default:
4760 		dv_id = IXGBE_LOW_DV(tc);
4761 		break;
4762 	}
4763 
4764 	/* Delay value is calculated in bit times convert to KB */
4765 	return IXGBE_BT2KB(dv_id);
4766 }
4767 
4768 /*
4769  * ixgbe_pbthresh_setup - calculate and setup high low water marks
4770  */
4771 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4772 {
4773 	struct ixgbe_hw *hw = &adapter->hw;
4774 	int num_tc = netdev_get_num_tc(adapter->netdev);
4775 	int i;
4776 
4777 	if (!num_tc)
4778 		num_tc = 1;
4779 
4780 	for (i = 0; i < num_tc; i++) {
4781 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4782 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4783 
4784 		/* Low water marks must not be larger than high water marks */
4785 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4786 			hw->fc.low_water[i] = 0;
4787 	}
4788 
4789 	for (; i < MAX_TRAFFIC_CLASS; i++)
4790 		hw->fc.high_water[i] = 0;
4791 }
4792 
4793 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4794 {
4795 	struct ixgbe_hw *hw = &adapter->hw;
4796 	int hdrm;
4797 	u8 tc = netdev_get_num_tc(adapter->netdev);
4798 
4799 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4800 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4801 		hdrm = 32 << adapter->fdir_pballoc;
4802 	else
4803 		hdrm = 0;
4804 
4805 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4806 	ixgbe_pbthresh_setup(adapter);
4807 }
4808 
4809 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4810 {
4811 	struct ixgbe_hw *hw = &adapter->hw;
4812 	struct hlist_node *node2;
4813 	struct ixgbe_fdir_filter *filter;
4814 
4815 	spin_lock(&adapter->fdir_perfect_lock);
4816 
4817 	if (!hlist_empty(&adapter->fdir_filter_list))
4818 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4819 
4820 	hlist_for_each_entry_safe(filter, node2,
4821 				  &adapter->fdir_filter_list, fdir_node) {
4822 		ixgbe_fdir_write_perfect_filter_82599(hw,
4823 				&filter->filter,
4824 				filter->sw_idx,
4825 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4826 				IXGBE_FDIR_DROP_QUEUE :
4827 				adapter->rx_ring[filter->action]->reg_idx);
4828 	}
4829 
4830 	spin_unlock(&adapter->fdir_perfect_lock);
4831 }
4832 
4833 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4834 				      struct ixgbe_adapter *adapter)
4835 {
4836 	struct ixgbe_hw *hw = &adapter->hw;
4837 	u32 vmolr;
4838 
4839 	/* No unicast promiscuous support for VMDQ devices. */
4840 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4841 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4842 
4843 	/* clear the affected bit */
4844 	vmolr &= ~IXGBE_VMOLR_MPE;
4845 
4846 	if (dev->flags & IFF_ALLMULTI) {
4847 		vmolr |= IXGBE_VMOLR_MPE;
4848 	} else {
4849 		vmolr |= IXGBE_VMOLR_ROMPE;
4850 		hw->mac.ops.update_mc_addr_list(hw, dev);
4851 	}
4852 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4853 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4854 }
4855 
4856 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4857 {
4858 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4859 	int rss_i = adapter->num_rx_queues_per_pool;
4860 	struct ixgbe_hw *hw = &adapter->hw;
4861 	u16 pool = vadapter->pool;
4862 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4863 		      IXGBE_PSRTYPE_UDPHDR |
4864 		      IXGBE_PSRTYPE_IPV4HDR |
4865 		      IXGBE_PSRTYPE_L2HDR |
4866 		      IXGBE_PSRTYPE_IPV6HDR;
4867 
4868 	if (hw->mac.type == ixgbe_mac_82598EB)
4869 		return;
4870 
4871 	if (rss_i > 3)
4872 		psrtype |= 2u << 29;
4873 	else if (rss_i > 1)
4874 		psrtype |= 1u << 29;
4875 
4876 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4877 }
4878 
4879 /**
4880  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4881  * @rx_ring: ring to free buffers from
4882  **/
4883 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4884 {
4885 	struct device *dev = rx_ring->dev;
4886 	unsigned long size;
4887 	u16 i;
4888 
4889 	/* ring already cleared, nothing to do */
4890 	if (!rx_ring->rx_buffer_info)
4891 		return;
4892 
4893 	/* Free all the Rx ring sk_buffs */
4894 	for (i = 0; i < rx_ring->count; i++) {
4895 		struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4896 
4897 		if (rx_buffer->skb) {
4898 			struct sk_buff *skb = rx_buffer->skb;
4899 			if (IXGBE_CB(skb)->page_released)
4900 				dma_unmap_page(dev,
4901 					       IXGBE_CB(skb)->dma,
4902 					       ixgbe_rx_bufsz(rx_ring),
4903 					       DMA_FROM_DEVICE);
4904 			dev_kfree_skb(skb);
4905 			rx_buffer->skb = NULL;
4906 		}
4907 
4908 		if (!rx_buffer->page)
4909 			continue;
4910 
4911 		dma_unmap_page(dev, rx_buffer->dma,
4912 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
4913 		__free_pages(rx_buffer->page, ixgbe_rx_pg_order(rx_ring));
4914 
4915 		rx_buffer->page = NULL;
4916 	}
4917 
4918 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4919 	memset(rx_ring->rx_buffer_info, 0, size);
4920 
4921 	/* Zero out the descriptor ring */
4922 	memset(rx_ring->desc, 0, rx_ring->size);
4923 
4924 	rx_ring->next_to_alloc = 0;
4925 	rx_ring->next_to_clean = 0;
4926 	rx_ring->next_to_use = 0;
4927 }
4928 
4929 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4930 				   struct ixgbe_ring *rx_ring)
4931 {
4932 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4933 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4934 
4935 	/* shutdown specific queue receive and wait for dma to settle */
4936 	ixgbe_disable_rx_queue(adapter, rx_ring);
4937 	usleep_range(10000, 20000);
4938 	ixgbe_irq_disable_queues(adapter, BIT_ULL(index));
4939 	ixgbe_clean_rx_ring(rx_ring);
4940 	rx_ring->l2_accel_priv = NULL;
4941 }
4942 
4943 static int ixgbe_fwd_ring_down(struct net_device *vdev,
4944 			       struct ixgbe_fwd_adapter *accel)
4945 {
4946 	struct ixgbe_adapter *adapter = accel->real_adapter;
4947 	unsigned int rxbase = accel->rx_base_queue;
4948 	unsigned int txbase = accel->tx_base_queue;
4949 	int i;
4950 
4951 	netif_tx_stop_all_queues(vdev);
4952 
4953 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4954 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4955 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4956 	}
4957 
4958 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4959 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4960 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4961 	}
4962 
4963 
4964 	return 0;
4965 }
4966 
4967 static int ixgbe_fwd_ring_up(struct net_device *vdev,
4968 			     struct ixgbe_fwd_adapter *accel)
4969 {
4970 	struct ixgbe_adapter *adapter = accel->real_adapter;
4971 	unsigned int rxbase, txbase, queues;
4972 	int i, baseq, err = 0;
4973 
4974 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4975 		return 0;
4976 
4977 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4978 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4979 		   accel->pool, adapter->num_rx_pools,
4980 		   baseq, baseq + adapter->num_rx_queues_per_pool,
4981 		   adapter->fwd_bitmask);
4982 
4983 	accel->netdev = vdev;
4984 	accel->rx_base_queue = rxbase = baseq;
4985 	accel->tx_base_queue = txbase = baseq;
4986 
4987 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4988 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4989 
4990 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4991 		adapter->rx_ring[rxbase + i]->netdev = vdev;
4992 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
4993 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
4994 	}
4995 
4996 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4997 		adapter->tx_ring[txbase + i]->netdev = vdev;
4998 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
4999 	}
5000 
5001 	queues = min_t(unsigned int,
5002 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
5003 	err = netif_set_real_num_tx_queues(vdev, queues);
5004 	if (err)
5005 		goto fwd_queue_err;
5006 
5007 	err = netif_set_real_num_rx_queues(vdev, queues);
5008 	if (err)
5009 		goto fwd_queue_err;
5010 
5011 	if (is_valid_ether_addr(vdev->dev_addr))
5012 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
5013 
5014 	ixgbe_fwd_psrtype(accel);
5015 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
5016 	return err;
5017 fwd_queue_err:
5018 	ixgbe_fwd_ring_down(vdev, accel);
5019 	return err;
5020 }
5021 
5022 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5023 {
5024 	struct net_device *upper;
5025 	struct list_head *iter;
5026 	int err;
5027 
5028 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5029 		if (netif_is_macvlan(upper)) {
5030 			struct macvlan_dev *dfwd = netdev_priv(upper);
5031 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
5032 
5033 			if (dfwd->fwd_priv) {
5034 				err = ixgbe_fwd_ring_up(upper, vadapter);
5035 				if (err)
5036 					continue;
5037 			}
5038 		}
5039 	}
5040 }
5041 
5042 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5043 {
5044 	struct ixgbe_hw *hw = &adapter->hw;
5045 
5046 	ixgbe_configure_pb(adapter);
5047 #ifdef CONFIG_IXGBE_DCB
5048 	ixgbe_configure_dcb(adapter);
5049 #endif
5050 	/*
5051 	 * We must restore virtualization before VLANs or else
5052 	 * the VLVF registers will not be populated
5053 	 */
5054 	ixgbe_configure_virtualization(adapter);
5055 
5056 	ixgbe_set_rx_mode(adapter->netdev);
5057 	ixgbe_restore_vlan(adapter);
5058 
5059 	switch (hw->mac.type) {
5060 	case ixgbe_mac_82599EB:
5061 	case ixgbe_mac_X540:
5062 		hw->mac.ops.disable_rx_buff(hw);
5063 		break;
5064 	default:
5065 		break;
5066 	}
5067 
5068 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5069 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5070 						adapter->fdir_pballoc);
5071 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5072 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5073 					      adapter->fdir_pballoc);
5074 		ixgbe_fdir_filter_restore(adapter);
5075 	}
5076 
5077 	switch (hw->mac.type) {
5078 	case ixgbe_mac_82599EB:
5079 	case ixgbe_mac_X540:
5080 		hw->mac.ops.enable_rx_buff(hw);
5081 		break;
5082 	default:
5083 		break;
5084 	}
5085 
5086 #ifdef CONFIG_IXGBE_DCA
5087 	/* configure DCA */
5088 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5089 		ixgbe_setup_dca(adapter);
5090 #endif /* CONFIG_IXGBE_DCA */
5091 
5092 #ifdef IXGBE_FCOE
5093 	/* configure FCoE L2 filters, redirection table, and Rx control */
5094 	ixgbe_configure_fcoe(adapter);
5095 
5096 #endif /* IXGBE_FCOE */
5097 	ixgbe_configure_tx(adapter);
5098 	ixgbe_configure_rx(adapter);
5099 	ixgbe_configure_dfwd(adapter);
5100 }
5101 
5102 /**
5103  * ixgbe_sfp_link_config - set up SFP+ link
5104  * @adapter: pointer to private adapter struct
5105  **/
5106 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5107 {
5108 	/*
5109 	 * We are assuming the worst case scenario here, and that
5110 	 * is that an SFP was inserted/removed after the reset
5111 	 * but before SFP detection was enabled.  As such the best
5112 	 * solution is to just start searching as soon as we start
5113 	 */
5114 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5115 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5116 
5117 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5118 	adapter->sfp_poll_time = 0;
5119 }
5120 
5121 /**
5122  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5123  * @hw: pointer to private hardware struct
5124  *
5125  * Returns 0 on success, negative on failure
5126  **/
5127 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5128 {
5129 	u32 speed;
5130 	bool autoneg, link_up = false;
5131 	int ret = IXGBE_ERR_LINK_SETUP;
5132 
5133 	if (hw->mac.ops.check_link)
5134 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5135 
5136 	if (ret)
5137 		return ret;
5138 
5139 	speed = hw->phy.autoneg_advertised;
5140 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5141 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5142 							&autoneg);
5143 	if (ret)
5144 		return ret;
5145 
5146 	if (hw->mac.ops.setup_link)
5147 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5148 
5149 	return ret;
5150 }
5151 
5152 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5153 {
5154 	struct ixgbe_hw *hw = &adapter->hw;
5155 	u32 gpie = 0;
5156 
5157 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5158 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5159 		       IXGBE_GPIE_OCD;
5160 		gpie |= IXGBE_GPIE_EIAME;
5161 		/*
5162 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5163 		 * this saves a register write for every interrupt
5164 		 */
5165 		switch (hw->mac.type) {
5166 		case ixgbe_mac_82598EB:
5167 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5168 			break;
5169 		case ixgbe_mac_82599EB:
5170 		case ixgbe_mac_X540:
5171 		case ixgbe_mac_X550:
5172 		case ixgbe_mac_X550EM_x:
5173 		case ixgbe_mac_x550em_a:
5174 		default:
5175 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5176 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5177 			break;
5178 		}
5179 	} else {
5180 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5181 		 * specifically only auto mask tx and rx interrupts */
5182 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5183 	}
5184 
5185 	/* XXX: to interrupt immediately for EICS writes, enable this */
5186 	/* gpie |= IXGBE_GPIE_EIMEN; */
5187 
5188 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5189 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5190 
5191 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5192 		case IXGBE_82599_VMDQ_8Q_MASK:
5193 			gpie |= IXGBE_GPIE_VTMODE_16;
5194 			break;
5195 		case IXGBE_82599_VMDQ_4Q_MASK:
5196 			gpie |= IXGBE_GPIE_VTMODE_32;
5197 			break;
5198 		default:
5199 			gpie |= IXGBE_GPIE_VTMODE_64;
5200 			break;
5201 		}
5202 	}
5203 
5204 	/* Enable Thermal over heat sensor interrupt */
5205 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5206 		switch (adapter->hw.mac.type) {
5207 		case ixgbe_mac_82599EB:
5208 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5209 			break;
5210 		default:
5211 			break;
5212 		}
5213 	}
5214 
5215 	/* Enable fan failure interrupt */
5216 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5217 		gpie |= IXGBE_SDP1_GPIEN(hw);
5218 
5219 	switch (hw->mac.type) {
5220 	case ixgbe_mac_82599EB:
5221 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5222 		break;
5223 	case ixgbe_mac_X550EM_x:
5224 	case ixgbe_mac_x550em_a:
5225 		gpie |= IXGBE_SDP0_GPIEN_X540;
5226 		break;
5227 	default:
5228 		break;
5229 	}
5230 
5231 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5232 }
5233 
5234 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5235 {
5236 	struct ixgbe_hw *hw = &adapter->hw;
5237 	int err;
5238 	u32 ctrl_ext;
5239 
5240 	ixgbe_get_hw_control(adapter);
5241 	ixgbe_setup_gpie(adapter);
5242 
5243 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5244 		ixgbe_configure_msix(adapter);
5245 	else
5246 		ixgbe_configure_msi_and_legacy(adapter);
5247 
5248 	/* enable the optics for 82599 SFP+ fiber */
5249 	if (hw->mac.ops.enable_tx_laser)
5250 		hw->mac.ops.enable_tx_laser(hw);
5251 
5252 	if (hw->phy.ops.set_phy_power)
5253 		hw->phy.ops.set_phy_power(hw, true);
5254 
5255 	smp_mb__before_atomic();
5256 	clear_bit(__IXGBE_DOWN, &adapter->state);
5257 	ixgbe_napi_enable_all(adapter);
5258 
5259 	if (ixgbe_is_sfp(hw)) {
5260 		ixgbe_sfp_link_config(adapter);
5261 	} else {
5262 		err = ixgbe_non_sfp_link_config(hw);
5263 		if (err)
5264 			e_err(probe, "link_config FAILED %d\n", err);
5265 	}
5266 
5267 	/* clear any pending interrupts, may auto mask */
5268 	IXGBE_READ_REG(hw, IXGBE_EICR);
5269 	ixgbe_irq_enable(adapter, true, true);
5270 
5271 	/*
5272 	 * If this adapter has a fan, check to see if we had a failure
5273 	 * before we enabled the interrupt.
5274 	 */
5275 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5276 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5277 		if (esdp & IXGBE_ESDP_SDP1)
5278 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5279 	}
5280 
5281 	/* bring the link up in the watchdog, this could race with our first
5282 	 * link up interrupt but shouldn't be a problem */
5283 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5284 	adapter->link_check_timeout = jiffies;
5285 	mod_timer(&adapter->service_timer, jiffies);
5286 
5287 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5288 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5289 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5290 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5291 }
5292 
5293 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5294 {
5295 	WARN_ON(in_interrupt());
5296 	/* put off any impending NetWatchDogTimeout */
5297 	netif_trans_update(adapter->netdev);
5298 
5299 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5300 		usleep_range(1000, 2000);
5301 	ixgbe_down(adapter);
5302 	/*
5303 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5304 	 * back up to give the VFs time to respond to the reset.  The
5305 	 * two second wait is based upon the watchdog timer cycle in
5306 	 * the VF driver.
5307 	 */
5308 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5309 		msleep(2000);
5310 	ixgbe_up(adapter);
5311 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5312 }
5313 
5314 void ixgbe_up(struct ixgbe_adapter *adapter)
5315 {
5316 	/* hardware has been reset, we need to reload some things */
5317 	ixgbe_configure(adapter);
5318 
5319 	ixgbe_up_complete(adapter);
5320 }
5321 
5322 void ixgbe_reset(struct ixgbe_adapter *adapter)
5323 {
5324 	struct ixgbe_hw *hw = &adapter->hw;
5325 	struct net_device *netdev = adapter->netdev;
5326 	int err;
5327 
5328 	if (ixgbe_removed(hw->hw_addr))
5329 		return;
5330 	/* lock SFP init bit to prevent race conditions with the watchdog */
5331 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5332 		usleep_range(1000, 2000);
5333 
5334 	/* clear all SFP and link config related flags while holding SFP_INIT */
5335 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5336 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5337 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5338 
5339 	err = hw->mac.ops.init_hw(hw);
5340 	switch (err) {
5341 	case 0:
5342 	case IXGBE_ERR_SFP_NOT_PRESENT:
5343 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5344 		break;
5345 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5346 		e_dev_err("master disable timed out\n");
5347 		break;
5348 	case IXGBE_ERR_EEPROM_VERSION:
5349 		/* We are running on a pre-production device, log a warning */
5350 		e_dev_warn("This device is a pre-production adapter/LOM. "
5351 			   "Please be aware there may be issues associated with "
5352 			   "your hardware.  If you are experiencing problems "
5353 			   "please contact your Intel or hardware "
5354 			   "representative who provided you with this "
5355 			   "hardware.\n");
5356 		break;
5357 	default:
5358 		e_dev_err("Hardware Error: %d\n", err);
5359 	}
5360 
5361 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5362 
5363 	/* flush entries out of MAC table */
5364 	ixgbe_flush_sw_mac_table(adapter);
5365 	__dev_uc_unsync(netdev, NULL);
5366 
5367 	/* do not flush user set addresses */
5368 	ixgbe_mac_set_default_filter(adapter);
5369 
5370 	/* update SAN MAC vmdq pool selection */
5371 	if (hw->mac.san_mac_rar_index)
5372 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5373 
5374 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5375 		ixgbe_ptp_reset(adapter);
5376 
5377 	if (hw->phy.ops.set_phy_power) {
5378 		if (!netif_running(adapter->netdev) && !adapter->wol)
5379 			hw->phy.ops.set_phy_power(hw, false);
5380 		else
5381 			hw->phy.ops.set_phy_power(hw, true);
5382 	}
5383 }
5384 
5385 /**
5386  * ixgbe_clean_tx_ring - Free Tx Buffers
5387  * @tx_ring: ring to be cleaned
5388  **/
5389 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5390 {
5391 	struct ixgbe_tx_buffer *tx_buffer_info;
5392 	unsigned long size;
5393 	u16 i;
5394 
5395 	/* ring already cleared, nothing to do */
5396 	if (!tx_ring->tx_buffer_info)
5397 		return;
5398 
5399 	/* Free all the Tx ring sk_buffs */
5400 	for (i = 0; i < tx_ring->count; i++) {
5401 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
5402 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
5403 	}
5404 
5405 	netdev_tx_reset_queue(txring_txq(tx_ring));
5406 
5407 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5408 	memset(tx_ring->tx_buffer_info, 0, size);
5409 
5410 	/* Zero out the descriptor ring */
5411 	memset(tx_ring->desc, 0, tx_ring->size);
5412 
5413 	tx_ring->next_to_use = 0;
5414 	tx_ring->next_to_clean = 0;
5415 }
5416 
5417 /**
5418  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5419  * @adapter: board private structure
5420  **/
5421 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5422 {
5423 	int i;
5424 
5425 	for (i = 0; i < adapter->num_rx_queues; i++)
5426 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5427 }
5428 
5429 /**
5430  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5431  * @adapter: board private structure
5432  **/
5433 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5434 {
5435 	int i;
5436 
5437 	for (i = 0; i < adapter->num_tx_queues; i++)
5438 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5439 }
5440 
5441 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5442 {
5443 	struct hlist_node *node2;
5444 	struct ixgbe_fdir_filter *filter;
5445 
5446 	spin_lock(&adapter->fdir_perfect_lock);
5447 
5448 	hlist_for_each_entry_safe(filter, node2,
5449 				  &adapter->fdir_filter_list, fdir_node) {
5450 		hlist_del(&filter->fdir_node);
5451 		kfree(filter);
5452 	}
5453 	adapter->fdir_filter_count = 0;
5454 
5455 	spin_unlock(&adapter->fdir_perfect_lock);
5456 }
5457 
5458 void ixgbe_down(struct ixgbe_adapter *adapter)
5459 {
5460 	struct net_device *netdev = adapter->netdev;
5461 	struct ixgbe_hw *hw = &adapter->hw;
5462 	struct net_device *upper;
5463 	struct list_head *iter;
5464 	int i;
5465 
5466 	/* signal that we are down to the interrupt handler */
5467 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5468 		return; /* do nothing if already down */
5469 
5470 	/* disable receives */
5471 	hw->mac.ops.disable_rx(hw);
5472 
5473 	/* disable all enabled rx queues */
5474 	for (i = 0; i < adapter->num_rx_queues; i++)
5475 		/* this call also flushes the previous write */
5476 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5477 
5478 	usleep_range(10000, 20000);
5479 
5480 	netif_tx_stop_all_queues(netdev);
5481 
5482 	/* call carrier off first to avoid false dev_watchdog timeouts */
5483 	netif_carrier_off(netdev);
5484 	netif_tx_disable(netdev);
5485 
5486 	/* disable any upper devices */
5487 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5488 		if (netif_is_macvlan(upper)) {
5489 			struct macvlan_dev *vlan = netdev_priv(upper);
5490 
5491 			if (vlan->fwd_priv) {
5492 				netif_tx_stop_all_queues(upper);
5493 				netif_carrier_off(upper);
5494 				netif_tx_disable(upper);
5495 			}
5496 		}
5497 	}
5498 
5499 	ixgbe_irq_disable(adapter);
5500 
5501 	ixgbe_napi_disable_all(adapter);
5502 
5503 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
5504 			     IXGBE_FLAG2_RESET_REQUESTED);
5505 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5506 
5507 	del_timer_sync(&adapter->service_timer);
5508 
5509 	if (adapter->num_vfs) {
5510 		/* Clear EITR Select mapping */
5511 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5512 
5513 		/* Mark all the VFs as inactive */
5514 		for (i = 0 ; i < adapter->num_vfs; i++)
5515 			adapter->vfinfo[i].clear_to_send = false;
5516 
5517 		/* ping all the active vfs to let them know we are going down */
5518 		ixgbe_ping_all_vfs(adapter);
5519 
5520 		/* Disable all VFTE/VFRE TX/RX */
5521 		ixgbe_disable_tx_rx(adapter);
5522 	}
5523 
5524 	/* disable transmits in the hardware now that interrupts are off */
5525 	for (i = 0; i < adapter->num_tx_queues; i++) {
5526 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5527 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5528 	}
5529 
5530 	/* Disable the Tx DMA engine on 82599 and later MAC */
5531 	switch (hw->mac.type) {
5532 	case ixgbe_mac_82599EB:
5533 	case ixgbe_mac_X540:
5534 	case ixgbe_mac_X550:
5535 	case ixgbe_mac_X550EM_x:
5536 	case ixgbe_mac_x550em_a:
5537 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5538 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5539 				 ~IXGBE_DMATXCTL_TE));
5540 		break;
5541 	default:
5542 		break;
5543 	}
5544 
5545 	if (!pci_channel_offline(adapter->pdev))
5546 		ixgbe_reset(adapter);
5547 
5548 	/* power down the optics for 82599 SFP+ fiber */
5549 	if (hw->mac.ops.disable_tx_laser)
5550 		hw->mac.ops.disable_tx_laser(hw);
5551 
5552 	ixgbe_clean_all_tx_rings(adapter);
5553 	ixgbe_clean_all_rx_rings(adapter);
5554 }
5555 
5556 /**
5557  * ixgbe_tx_timeout - Respond to a Tx Hang
5558  * @netdev: network interface device structure
5559  **/
5560 static void ixgbe_tx_timeout(struct net_device *netdev)
5561 {
5562 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5563 
5564 	/* Do the reset outside of interrupt context */
5565 	ixgbe_tx_timeout_reset(adapter);
5566 }
5567 
5568 #ifdef CONFIG_IXGBE_DCB
5569 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
5570 {
5571 	struct ixgbe_hw *hw = &adapter->hw;
5572 	struct tc_configuration *tc;
5573 	int j;
5574 
5575 	switch (hw->mac.type) {
5576 	case ixgbe_mac_82598EB:
5577 	case ixgbe_mac_82599EB:
5578 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5579 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5580 		break;
5581 	case ixgbe_mac_X540:
5582 	case ixgbe_mac_X550:
5583 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5584 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5585 		break;
5586 	case ixgbe_mac_X550EM_x:
5587 	case ixgbe_mac_x550em_a:
5588 	default:
5589 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
5590 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
5591 		break;
5592 	}
5593 
5594 	/* Configure DCB traffic classes */
5595 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5596 		tc = &adapter->dcb_cfg.tc_config[j];
5597 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5598 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5599 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5600 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5601 		tc->dcb_pfc = pfc_disabled;
5602 	}
5603 
5604 	/* Initialize default user to priority mapping, UPx->TC0 */
5605 	tc = &adapter->dcb_cfg.tc_config[0];
5606 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5607 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5608 
5609 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5610 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5611 	adapter->dcb_cfg.pfc_mode_enable = false;
5612 	adapter->dcb_set_bitmap = 0x00;
5613 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
5614 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5615 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5616 	       sizeof(adapter->temp_dcb_cfg));
5617 }
5618 #endif
5619 
5620 /**
5621  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5622  * @adapter: board private structure to initialize
5623  *
5624  * ixgbe_sw_init initializes the Adapter private data structure.
5625  * Fields are initialized based on PCI device information and
5626  * OS network device settings (MTU size).
5627  **/
5628 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
5629 {
5630 	struct ixgbe_hw *hw = &adapter->hw;
5631 	struct pci_dev *pdev = adapter->pdev;
5632 	unsigned int rss, fdir;
5633 	u32 fwsm;
5634 	int i;
5635 
5636 	/* PCI config space info */
5637 
5638 	hw->vendor_id = pdev->vendor;
5639 	hw->device_id = pdev->device;
5640 	hw->revision_id = pdev->revision;
5641 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5642 	hw->subsystem_device_id = pdev->subsystem_device;
5643 
5644 	/* Set common capability flags and settings */
5645 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5646 	adapter->ring_feature[RING_F_RSS].limit = rss;
5647 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5648 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5649 	adapter->atr_sample_rate = 20;
5650 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5651 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5652 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5653 #ifdef CONFIG_IXGBE_DCA
5654 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5655 #endif
5656 #ifdef CONFIG_IXGBE_DCB
5657 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
5658 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
5659 #endif
5660 #ifdef IXGBE_FCOE
5661 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5662 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5663 #ifdef CONFIG_IXGBE_DCB
5664 	/* Default traffic class to use for FCoE */
5665 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5666 #endif /* CONFIG_IXGBE_DCB */
5667 #endif /* IXGBE_FCOE */
5668 
5669 	/* initialize static ixgbe jump table entries */
5670 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
5671 					  GFP_KERNEL);
5672 	if (!adapter->jump_tables[0])
5673 		return -ENOMEM;
5674 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
5675 
5676 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
5677 		adapter->jump_tables[i] = NULL;
5678 
5679 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5680 				     hw->mac.num_rar_entries,
5681 				     GFP_ATOMIC);
5682 	if (!adapter->mac_table)
5683 		return -ENOMEM;
5684 
5685 	/* Set MAC specific capability flags and exceptions */
5686 	switch (hw->mac.type) {
5687 	case ixgbe_mac_82598EB:
5688 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5689 
5690 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5691 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5692 
5693 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5694 		adapter->ring_feature[RING_F_FDIR].limit = 0;
5695 		adapter->atr_sample_rate = 0;
5696 		adapter->fdir_pballoc = 0;
5697 #ifdef IXGBE_FCOE
5698 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5699 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5700 #ifdef CONFIG_IXGBE_DCB
5701 		adapter->fcoe.up = 0;
5702 #endif /* IXGBE_DCB */
5703 #endif /* IXGBE_FCOE */
5704 		break;
5705 	case ixgbe_mac_82599EB:
5706 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5707 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5708 		break;
5709 	case ixgbe_mac_X540:
5710 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
5711 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5712 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5713 		break;
5714 	case ixgbe_mac_X550EM_x:
5715 	case ixgbe_mac_x550em_a:
5716 #ifdef CONFIG_IXGBE_DCB
5717 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
5718 #endif
5719 #ifdef IXGBE_FCOE
5720 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5721 #ifdef CONFIG_IXGBE_DCB
5722 		adapter->fcoe.up = 0;
5723 #endif /* IXGBE_DCB */
5724 #endif /* IXGBE_FCOE */
5725 	/* Fall Through */
5726 	case ixgbe_mac_X550:
5727 #ifdef CONFIG_IXGBE_DCA
5728 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5729 #endif
5730 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
5731 		break;
5732 	default:
5733 		break;
5734 	}
5735 
5736 #ifdef IXGBE_FCOE
5737 	/* FCoE support exists, always init the FCoE lock */
5738 	spin_lock_init(&adapter->fcoe.lock);
5739 
5740 #endif
5741 	/* n-tuple support exists, always init our spinlock */
5742 	spin_lock_init(&adapter->fdir_perfect_lock);
5743 
5744 #ifdef CONFIG_IXGBE_DCB
5745 	ixgbe_init_dcb(adapter);
5746 #endif
5747 
5748 	/* default flow control settings */
5749 	hw->fc.requested_mode = ixgbe_fc_full;
5750 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5751 	ixgbe_pbthresh_setup(adapter);
5752 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5753 	hw->fc.send_xon = true;
5754 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5755 
5756 #ifdef CONFIG_PCI_IOV
5757 	if (max_vfs > 0)
5758 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5759 
5760 	/* assign number of SR-IOV VFs */
5761 	if (hw->mac.type != ixgbe_mac_82598EB) {
5762 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5763 			adapter->num_vfs = 0;
5764 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5765 		} else {
5766 			adapter->num_vfs = max_vfs;
5767 		}
5768 	}
5769 #endif /* CONFIG_PCI_IOV */
5770 
5771 	/* enable itr by default in dynamic mode */
5772 	adapter->rx_itr_setting = 1;
5773 	adapter->tx_itr_setting = 1;
5774 
5775 	/* set default ring sizes */
5776 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5777 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5778 
5779 	/* set default work limits */
5780 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5781 
5782 	/* initialize eeprom parameters */
5783 	if (ixgbe_init_eeprom_params_generic(hw)) {
5784 		e_dev_err("EEPROM initialization failed\n");
5785 		return -EIO;
5786 	}
5787 
5788 	/* PF holds first pool slot */
5789 	set_bit(0, &adapter->fwd_bitmask);
5790 	set_bit(__IXGBE_DOWN, &adapter->state);
5791 
5792 	return 0;
5793 }
5794 
5795 /**
5796  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5797  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5798  *
5799  * Return 0 on success, negative on failure
5800  **/
5801 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5802 {
5803 	struct device *dev = tx_ring->dev;
5804 	int orig_node = dev_to_node(dev);
5805 	int ring_node = -1;
5806 	int size;
5807 
5808 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5809 
5810 	if (tx_ring->q_vector)
5811 		ring_node = tx_ring->q_vector->numa_node;
5812 
5813 	tx_ring->tx_buffer_info = vzalloc_node(size, ring_node);
5814 	if (!tx_ring->tx_buffer_info)
5815 		tx_ring->tx_buffer_info = vzalloc(size);
5816 	if (!tx_ring->tx_buffer_info)
5817 		goto err;
5818 
5819 	u64_stats_init(&tx_ring->syncp);
5820 
5821 	/* round up to nearest 4K */
5822 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5823 	tx_ring->size = ALIGN(tx_ring->size, 4096);
5824 
5825 	set_dev_node(dev, ring_node);
5826 	tx_ring->desc = dma_alloc_coherent(dev,
5827 					   tx_ring->size,
5828 					   &tx_ring->dma,
5829 					   GFP_KERNEL);
5830 	set_dev_node(dev, orig_node);
5831 	if (!tx_ring->desc)
5832 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5833 						   &tx_ring->dma, GFP_KERNEL);
5834 	if (!tx_ring->desc)
5835 		goto err;
5836 
5837 	tx_ring->next_to_use = 0;
5838 	tx_ring->next_to_clean = 0;
5839 	return 0;
5840 
5841 err:
5842 	vfree(tx_ring->tx_buffer_info);
5843 	tx_ring->tx_buffer_info = NULL;
5844 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5845 	return -ENOMEM;
5846 }
5847 
5848 /**
5849  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5850  * @adapter: board private structure
5851  *
5852  * If this function returns with an error, then it's possible one or
5853  * more of the rings is populated (while the rest are not).  It is the
5854  * callers duty to clean those orphaned rings.
5855  *
5856  * Return 0 on success, negative on failure
5857  **/
5858 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5859 {
5860 	int i, err = 0;
5861 
5862 	for (i = 0; i < adapter->num_tx_queues; i++) {
5863 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5864 		if (!err)
5865 			continue;
5866 
5867 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5868 		goto err_setup_tx;
5869 	}
5870 
5871 	return 0;
5872 err_setup_tx:
5873 	/* rewind the index freeing the rings as we go */
5874 	while (i--)
5875 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5876 	return err;
5877 }
5878 
5879 /**
5880  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5881  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5882  *
5883  * Returns 0 on success, negative on failure
5884  **/
5885 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5886 {
5887 	struct device *dev = rx_ring->dev;
5888 	int orig_node = dev_to_node(dev);
5889 	int ring_node = -1;
5890 	int size;
5891 
5892 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5893 
5894 	if (rx_ring->q_vector)
5895 		ring_node = rx_ring->q_vector->numa_node;
5896 
5897 	rx_ring->rx_buffer_info = vzalloc_node(size, ring_node);
5898 	if (!rx_ring->rx_buffer_info)
5899 		rx_ring->rx_buffer_info = vzalloc(size);
5900 	if (!rx_ring->rx_buffer_info)
5901 		goto err;
5902 
5903 	u64_stats_init(&rx_ring->syncp);
5904 
5905 	/* Round up to nearest 4K */
5906 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5907 	rx_ring->size = ALIGN(rx_ring->size, 4096);
5908 
5909 	set_dev_node(dev, ring_node);
5910 	rx_ring->desc = dma_alloc_coherent(dev,
5911 					   rx_ring->size,
5912 					   &rx_ring->dma,
5913 					   GFP_KERNEL);
5914 	set_dev_node(dev, orig_node);
5915 	if (!rx_ring->desc)
5916 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5917 						   &rx_ring->dma, GFP_KERNEL);
5918 	if (!rx_ring->desc)
5919 		goto err;
5920 
5921 	rx_ring->next_to_clean = 0;
5922 	rx_ring->next_to_use = 0;
5923 
5924 	return 0;
5925 err:
5926 	vfree(rx_ring->rx_buffer_info);
5927 	rx_ring->rx_buffer_info = NULL;
5928 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5929 	return -ENOMEM;
5930 }
5931 
5932 /**
5933  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5934  * @adapter: board private structure
5935  *
5936  * If this function returns with an error, then it's possible one or
5937  * more of the rings is populated (while the rest are not).  It is the
5938  * callers duty to clean those orphaned rings.
5939  *
5940  * Return 0 on success, negative on failure
5941  **/
5942 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5943 {
5944 	int i, err = 0;
5945 
5946 	for (i = 0; i < adapter->num_rx_queues; i++) {
5947 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5948 		if (!err)
5949 			continue;
5950 
5951 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5952 		goto err_setup_rx;
5953 	}
5954 
5955 #ifdef IXGBE_FCOE
5956 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5957 	if (!err)
5958 #endif
5959 		return 0;
5960 err_setup_rx:
5961 	/* rewind the index freeing the rings as we go */
5962 	while (i--)
5963 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5964 	return err;
5965 }
5966 
5967 /**
5968  * ixgbe_free_tx_resources - Free Tx Resources per Queue
5969  * @tx_ring: Tx descriptor ring for a specific queue
5970  *
5971  * Free all transmit software resources
5972  **/
5973 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5974 {
5975 	ixgbe_clean_tx_ring(tx_ring);
5976 
5977 	vfree(tx_ring->tx_buffer_info);
5978 	tx_ring->tx_buffer_info = NULL;
5979 
5980 	/* if not set, then don't free */
5981 	if (!tx_ring->desc)
5982 		return;
5983 
5984 	dma_free_coherent(tx_ring->dev, tx_ring->size,
5985 			  tx_ring->desc, tx_ring->dma);
5986 
5987 	tx_ring->desc = NULL;
5988 }
5989 
5990 /**
5991  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5992  * @adapter: board private structure
5993  *
5994  * Free all transmit software resources
5995  **/
5996 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
5997 {
5998 	int i;
5999 
6000 	for (i = 0; i < adapter->num_tx_queues; i++)
6001 		if (adapter->tx_ring[i]->desc)
6002 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6003 }
6004 
6005 /**
6006  * ixgbe_free_rx_resources - Free Rx Resources
6007  * @rx_ring: ring to clean the resources from
6008  *
6009  * Free all receive software resources
6010  **/
6011 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6012 {
6013 	ixgbe_clean_rx_ring(rx_ring);
6014 
6015 	vfree(rx_ring->rx_buffer_info);
6016 	rx_ring->rx_buffer_info = NULL;
6017 
6018 	/* if not set, then don't free */
6019 	if (!rx_ring->desc)
6020 		return;
6021 
6022 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6023 			  rx_ring->desc, rx_ring->dma);
6024 
6025 	rx_ring->desc = NULL;
6026 }
6027 
6028 /**
6029  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6030  * @adapter: board private structure
6031  *
6032  * Free all receive software resources
6033  **/
6034 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6035 {
6036 	int i;
6037 
6038 #ifdef IXGBE_FCOE
6039 	ixgbe_free_fcoe_ddp_resources(adapter);
6040 
6041 #endif
6042 	for (i = 0; i < adapter->num_rx_queues; i++)
6043 		if (adapter->rx_ring[i]->desc)
6044 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6045 }
6046 
6047 /**
6048  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6049  * @netdev: network interface device structure
6050  * @new_mtu: new value for maximum frame size
6051  *
6052  * Returns 0 on success, negative on failure
6053  **/
6054 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6055 {
6056 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6057 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
6058 
6059 	/* MTU < 68 is an error and causes problems on some kernels */
6060 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
6061 		return -EINVAL;
6062 
6063 	/*
6064 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6065 	 * paths when MTU greater than 1500 is configured.  So display a
6066 	 * warning that legacy VFs will be disabled.
6067 	 */
6068 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6069 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6070 	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
6071 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6072 
6073 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6074 
6075 	/* must set new MTU before calling down or up */
6076 	netdev->mtu = new_mtu;
6077 
6078 	if (netif_running(netdev))
6079 		ixgbe_reinit_locked(adapter);
6080 
6081 	return 0;
6082 }
6083 
6084 /**
6085  * ixgbe_open - Called when a network interface is made active
6086  * @netdev: network interface device structure
6087  *
6088  * Returns 0 on success, negative value on failure
6089  *
6090  * The open entry point is called when a network interface is made
6091  * active by the system (IFF_UP).  At this point all resources needed
6092  * for transmit and receive operations are allocated, the interrupt
6093  * handler is registered with the OS, the watchdog timer is started,
6094  * and the stack is notified that the interface is ready.
6095  **/
6096 int ixgbe_open(struct net_device *netdev)
6097 {
6098 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6099 	struct ixgbe_hw *hw = &adapter->hw;
6100 	int err, queues;
6101 
6102 	/* disallow open during test */
6103 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6104 		return -EBUSY;
6105 
6106 	netif_carrier_off(netdev);
6107 
6108 	/* allocate transmit descriptors */
6109 	err = ixgbe_setup_all_tx_resources(adapter);
6110 	if (err)
6111 		goto err_setup_tx;
6112 
6113 	/* allocate receive descriptors */
6114 	err = ixgbe_setup_all_rx_resources(adapter);
6115 	if (err)
6116 		goto err_setup_rx;
6117 
6118 	ixgbe_configure(adapter);
6119 
6120 	err = ixgbe_request_irq(adapter);
6121 	if (err)
6122 		goto err_req_irq;
6123 
6124 	/* Notify the stack of the actual queue counts. */
6125 	if (adapter->num_rx_pools > 1)
6126 		queues = adapter->num_rx_queues_per_pool;
6127 	else
6128 		queues = adapter->num_tx_queues;
6129 
6130 	err = netif_set_real_num_tx_queues(netdev, queues);
6131 	if (err)
6132 		goto err_set_queues;
6133 
6134 	if (adapter->num_rx_pools > 1 &&
6135 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
6136 		queues = IXGBE_MAX_L2A_QUEUES;
6137 	else
6138 		queues = adapter->num_rx_queues;
6139 	err = netif_set_real_num_rx_queues(netdev, queues);
6140 	if (err)
6141 		goto err_set_queues;
6142 
6143 	ixgbe_ptp_init(adapter);
6144 
6145 	ixgbe_up_complete(adapter);
6146 
6147 	ixgbe_clear_vxlan_port(adapter);
6148 	udp_tunnel_get_rx_info(netdev);
6149 
6150 	return 0;
6151 
6152 err_set_queues:
6153 	ixgbe_free_irq(adapter);
6154 err_req_irq:
6155 	ixgbe_free_all_rx_resources(adapter);
6156 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6157 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6158 err_setup_rx:
6159 	ixgbe_free_all_tx_resources(adapter);
6160 err_setup_tx:
6161 	ixgbe_reset(adapter);
6162 
6163 	return err;
6164 }
6165 
6166 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6167 {
6168 	ixgbe_ptp_suspend(adapter);
6169 
6170 	if (adapter->hw.phy.ops.enter_lplu) {
6171 		adapter->hw.phy.reset_disable = true;
6172 		ixgbe_down(adapter);
6173 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6174 		adapter->hw.phy.reset_disable = false;
6175 	} else {
6176 		ixgbe_down(adapter);
6177 	}
6178 
6179 	ixgbe_free_irq(adapter);
6180 
6181 	ixgbe_free_all_tx_resources(adapter);
6182 	ixgbe_free_all_rx_resources(adapter);
6183 }
6184 
6185 /**
6186  * ixgbe_close - Disables a network interface
6187  * @netdev: network interface device structure
6188  *
6189  * Returns 0, this is not allowed to fail
6190  *
6191  * The close entry point is called when an interface is de-activated
6192  * by the OS.  The hardware is still under the drivers control, but
6193  * needs to be disabled.  A global MAC reset is issued to stop the
6194  * hardware, and all transmit and receive resources are freed.
6195  **/
6196 int ixgbe_close(struct net_device *netdev)
6197 {
6198 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6199 
6200 	ixgbe_ptp_stop(adapter);
6201 
6202 	ixgbe_close_suspend(adapter);
6203 
6204 	ixgbe_fdir_filter_exit(adapter);
6205 
6206 	ixgbe_release_hw_control(adapter);
6207 
6208 	return 0;
6209 }
6210 
6211 #ifdef CONFIG_PM
6212 static int ixgbe_resume(struct pci_dev *pdev)
6213 {
6214 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6215 	struct net_device *netdev = adapter->netdev;
6216 	u32 err;
6217 
6218 	adapter->hw.hw_addr = adapter->io_addr;
6219 	pci_set_power_state(pdev, PCI_D0);
6220 	pci_restore_state(pdev);
6221 	/*
6222 	 * pci_restore_state clears dev->state_saved so call
6223 	 * pci_save_state to restore it.
6224 	 */
6225 	pci_save_state(pdev);
6226 
6227 	err = pci_enable_device_mem(pdev);
6228 	if (err) {
6229 		e_dev_err("Cannot enable PCI device from suspend\n");
6230 		return err;
6231 	}
6232 	smp_mb__before_atomic();
6233 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6234 	pci_set_master(pdev);
6235 
6236 	pci_wake_from_d3(pdev, false);
6237 
6238 	ixgbe_reset(adapter);
6239 
6240 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6241 
6242 	rtnl_lock();
6243 	err = ixgbe_init_interrupt_scheme(adapter);
6244 	if (!err && netif_running(netdev))
6245 		err = ixgbe_open(netdev);
6246 
6247 	rtnl_unlock();
6248 
6249 	if (err)
6250 		return err;
6251 
6252 	netif_device_attach(netdev);
6253 
6254 	return 0;
6255 }
6256 #endif /* CONFIG_PM */
6257 
6258 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6259 {
6260 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6261 	struct net_device *netdev = adapter->netdev;
6262 	struct ixgbe_hw *hw = &adapter->hw;
6263 	u32 ctrl, fctrl;
6264 	u32 wufc = adapter->wol;
6265 #ifdef CONFIG_PM
6266 	int retval = 0;
6267 #endif
6268 
6269 	netif_device_detach(netdev);
6270 
6271 	rtnl_lock();
6272 	if (netif_running(netdev))
6273 		ixgbe_close_suspend(adapter);
6274 	rtnl_unlock();
6275 
6276 	ixgbe_clear_interrupt_scheme(adapter);
6277 
6278 #ifdef CONFIG_PM
6279 	retval = pci_save_state(pdev);
6280 	if (retval)
6281 		return retval;
6282 
6283 #endif
6284 	if (hw->mac.ops.stop_link_on_d3)
6285 		hw->mac.ops.stop_link_on_d3(hw);
6286 
6287 	if (wufc) {
6288 		ixgbe_set_rx_mode(netdev);
6289 
6290 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6291 		if (hw->mac.ops.enable_tx_laser)
6292 			hw->mac.ops.enable_tx_laser(hw);
6293 
6294 		/* turn on all-multi mode if wake on multicast is enabled */
6295 		if (wufc & IXGBE_WUFC_MC) {
6296 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6297 			fctrl |= IXGBE_FCTRL_MPE;
6298 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6299 		}
6300 
6301 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6302 		ctrl |= IXGBE_CTRL_GIO_DIS;
6303 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6304 
6305 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6306 	} else {
6307 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6308 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6309 	}
6310 
6311 	switch (hw->mac.type) {
6312 	case ixgbe_mac_82598EB:
6313 		pci_wake_from_d3(pdev, false);
6314 		break;
6315 	case ixgbe_mac_82599EB:
6316 	case ixgbe_mac_X540:
6317 	case ixgbe_mac_X550:
6318 	case ixgbe_mac_X550EM_x:
6319 	case ixgbe_mac_x550em_a:
6320 		pci_wake_from_d3(pdev, !!wufc);
6321 		break;
6322 	default:
6323 		break;
6324 	}
6325 
6326 	*enable_wake = !!wufc;
6327 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6328 		hw->phy.ops.set_phy_power(hw, false);
6329 
6330 	ixgbe_release_hw_control(adapter);
6331 
6332 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6333 		pci_disable_device(pdev);
6334 
6335 	return 0;
6336 }
6337 
6338 #ifdef CONFIG_PM
6339 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6340 {
6341 	int retval;
6342 	bool wake;
6343 
6344 	retval = __ixgbe_shutdown(pdev, &wake);
6345 	if (retval)
6346 		return retval;
6347 
6348 	if (wake) {
6349 		pci_prepare_to_sleep(pdev);
6350 	} else {
6351 		pci_wake_from_d3(pdev, false);
6352 		pci_set_power_state(pdev, PCI_D3hot);
6353 	}
6354 
6355 	return 0;
6356 }
6357 #endif /* CONFIG_PM */
6358 
6359 static void ixgbe_shutdown(struct pci_dev *pdev)
6360 {
6361 	bool wake;
6362 
6363 	__ixgbe_shutdown(pdev, &wake);
6364 
6365 	if (system_state == SYSTEM_POWER_OFF) {
6366 		pci_wake_from_d3(pdev, wake);
6367 		pci_set_power_state(pdev, PCI_D3hot);
6368 	}
6369 }
6370 
6371 /**
6372  * ixgbe_update_stats - Update the board statistics counters.
6373  * @adapter: board private structure
6374  **/
6375 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6376 {
6377 	struct net_device *netdev = adapter->netdev;
6378 	struct ixgbe_hw *hw = &adapter->hw;
6379 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6380 	u64 total_mpc = 0;
6381 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6382 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6383 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6384 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6385 
6386 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6387 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6388 		return;
6389 
6390 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6391 		u64 rsc_count = 0;
6392 		u64 rsc_flush = 0;
6393 		for (i = 0; i < adapter->num_rx_queues; i++) {
6394 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6395 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6396 		}
6397 		adapter->rsc_total_count = rsc_count;
6398 		adapter->rsc_total_flush = rsc_flush;
6399 	}
6400 
6401 	for (i = 0; i < adapter->num_rx_queues; i++) {
6402 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6403 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6404 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6405 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6406 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6407 		bytes += rx_ring->stats.bytes;
6408 		packets += rx_ring->stats.packets;
6409 	}
6410 	adapter->non_eop_descs = non_eop_descs;
6411 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6412 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6413 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6414 	netdev->stats.rx_bytes = bytes;
6415 	netdev->stats.rx_packets = packets;
6416 
6417 	bytes = 0;
6418 	packets = 0;
6419 	/* gather some stats to the adapter struct that are per queue */
6420 	for (i = 0; i < adapter->num_tx_queues; i++) {
6421 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6422 		restart_queue += tx_ring->tx_stats.restart_queue;
6423 		tx_busy += tx_ring->tx_stats.tx_busy;
6424 		bytes += tx_ring->stats.bytes;
6425 		packets += tx_ring->stats.packets;
6426 	}
6427 	adapter->restart_queue = restart_queue;
6428 	adapter->tx_busy = tx_busy;
6429 	netdev->stats.tx_bytes = bytes;
6430 	netdev->stats.tx_packets = packets;
6431 
6432 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6433 
6434 	/* 8 register reads */
6435 	for (i = 0; i < 8; i++) {
6436 		/* for packet buffers not used, the register should read 0 */
6437 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6438 		missed_rx += mpc;
6439 		hwstats->mpc[i] += mpc;
6440 		total_mpc += hwstats->mpc[i];
6441 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6442 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6443 		switch (hw->mac.type) {
6444 		case ixgbe_mac_82598EB:
6445 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6446 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6447 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6448 			hwstats->pxonrxc[i] +=
6449 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6450 			break;
6451 		case ixgbe_mac_82599EB:
6452 		case ixgbe_mac_X540:
6453 		case ixgbe_mac_X550:
6454 		case ixgbe_mac_X550EM_x:
6455 		case ixgbe_mac_x550em_a:
6456 			hwstats->pxonrxc[i] +=
6457 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6458 			break;
6459 		default:
6460 			break;
6461 		}
6462 	}
6463 
6464 	/*16 register reads */
6465 	for (i = 0; i < 16; i++) {
6466 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6467 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6468 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6469 		    (hw->mac.type == ixgbe_mac_X540) ||
6470 		    (hw->mac.type == ixgbe_mac_X550) ||
6471 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6472 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6473 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6474 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6475 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6476 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6477 		}
6478 	}
6479 
6480 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6481 	/* work around hardware counting issue */
6482 	hwstats->gprc -= missed_rx;
6483 
6484 	ixgbe_update_xoff_received(adapter);
6485 
6486 	/* 82598 hardware only has a 32 bit counter in the high register */
6487 	switch (hw->mac.type) {
6488 	case ixgbe_mac_82598EB:
6489 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6490 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6491 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6492 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6493 		break;
6494 	case ixgbe_mac_X540:
6495 	case ixgbe_mac_X550:
6496 	case ixgbe_mac_X550EM_x:
6497 	case ixgbe_mac_x550em_a:
6498 		/* OS2BMC stats are X540 and later */
6499 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6500 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6501 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6502 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6503 	case ixgbe_mac_82599EB:
6504 		for (i = 0; i < 16; i++)
6505 			adapter->hw_rx_no_dma_resources +=
6506 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6507 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6508 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6509 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6510 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6511 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6512 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6513 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6514 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6515 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6516 #ifdef IXGBE_FCOE
6517 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6518 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6519 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6520 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6521 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6522 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6523 		/* Add up per cpu counters for total ddp aloc fail */
6524 		if (adapter->fcoe.ddp_pool) {
6525 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6526 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6527 			unsigned int cpu;
6528 			u64 noddp = 0, noddp_ext_buff = 0;
6529 			for_each_possible_cpu(cpu) {
6530 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6531 				noddp += ddp_pool->noddp;
6532 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6533 			}
6534 			hwstats->fcoe_noddp = noddp;
6535 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6536 		}
6537 #endif /* IXGBE_FCOE */
6538 		break;
6539 	default:
6540 		break;
6541 	}
6542 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6543 	hwstats->bprc += bprc;
6544 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6545 	if (hw->mac.type == ixgbe_mac_82598EB)
6546 		hwstats->mprc -= bprc;
6547 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6548 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6549 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6550 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6551 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6552 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6553 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6554 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6555 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6556 	hwstats->lxontxc += lxon;
6557 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6558 	hwstats->lxofftxc += lxoff;
6559 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6560 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6561 	/*
6562 	 * 82598 errata - tx of flow control packets is included in tx counters
6563 	 */
6564 	xon_off_tot = lxon + lxoff;
6565 	hwstats->gptc -= xon_off_tot;
6566 	hwstats->mptc -= xon_off_tot;
6567 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6568 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6569 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6570 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6571 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6572 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6573 	hwstats->ptc64 -= xon_off_tot;
6574 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6575 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6576 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6577 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6578 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6579 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6580 
6581 	/* Fill out the OS statistics structure */
6582 	netdev->stats.multicast = hwstats->mprc;
6583 
6584 	/* Rx Errors */
6585 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6586 	netdev->stats.rx_dropped = 0;
6587 	netdev->stats.rx_length_errors = hwstats->rlec;
6588 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6589 	netdev->stats.rx_missed_errors = total_mpc;
6590 }
6591 
6592 /**
6593  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6594  * @adapter: pointer to the device adapter structure
6595  **/
6596 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6597 {
6598 	struct ixgbe_hw *hw = &adapter->hw;
6599 	int i;
6600 
6601 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6602 		return;
6603 
6604 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6605 
6606 	/* if interface is down do nothing */
6607 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6608 		return;
6609 
6610 	/* do nothing if we are not using signature filters */
6611 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6612 		return;
6613 
6614 	adapter->fdir_overflow++;
6615 
6616 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6617 		for (i = 0; i < adapter->num_tx_queues; i++)
6618 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6619 				&(adapter->tx_ring[i]->state));
6620 		/* re-enable flow director interrupts */
6621 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6622 	} else {
6623 		e_err(probe, "failed to finish FDIR re-initialization, "
6624 		      "ignored adding FDIR ATR filters\n");
6625 	}
6626 }
6627 
6628 /**
6629  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6630  * @adapter: pointer to the device adapter structure
6631  *
6632  * This function serves two purposes.  First it strobes the interrupt lines
6633  * in order to make certain interrupts are occurring.  Secondly it sets the
6634  * bits needed to check for TX hangs.  As a result we should immediately
6635  * determine if a hang has occurred.
6636  */
6637 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6638 {
6639 	struct ixgbe_hw *hw = &adapter->hw;
6640 	u64 eics = 0;
6641 	int i;
6642 
6643 	/* If we're down, removing or resetting, just bail */
6644 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6645 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6646 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6647 		return;
6648 
6649 	/* Force detection of hung controller */
6650 	if (netif_carrier_ok(adapter->netdev)) {
6651 		for (i = 0; i < adapter->num_tx_queues; i++)
6652 			set_check_for_tx_hang(adapter->tx_ring[i]);
6653 	}
6654 
6655 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6656 		/*
6657 		 * for legacy and MSI interrupts don't set any bits
6658 		 * that are enabled for EIAM, because this operation
6659 		 * would set *both* EIMS and EICS for any bit in EIAM
6660 		 */
6661 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6662 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6663 	} else {
6664 		/* get one bit for every active tx/rx interrupt vector */
6665 		for (i = 0; i < adapter->num_q_vectors; i++) {
6666 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6667 			if (qv->rx.ring || qv->tx.ring)
6668 				eics |= BIT_ULL(i);
6669 		}
6670 	}
6671 
6672 	/* Cause software interrupt to ensure rings are cleaned */
6673 	ixgbe_irq_rearm_queues(adapter, eics);
6674 }
6675 
6676 /**
6677  * ixgbe_watchdog_update_link - update the link status
6678  * @adapter: pointer to the device adapter structure
6679  * @link_speed: pointer to a u32 to store the link_speed
6680  **/
6681 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6682 {
6683 	struct ixgbe_hw *hw = &adapter->hw;
6684 	u32 link_speed = adapter->link_speed;
6685 	bool link_up = adapter->link_up;
6686 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6687 
6688 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6689 		return;
6690 
6691 	if (hw->mac.ops.check_link) {
6692 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6693 	} else {
6694 		/* always assume link is up, if no check link function */
6695 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6696 		link_up = true;
6697 	}
6698 
6699 	if (adapter->ixgbe_ieee_pfc)
6700 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6701 
6702 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6703 		hw->mac.ops.fc_enable(hw);
6704 		ixgbe_set_rx_drop_en(adapter);
6705 	}
6706 
6707 	if (link_up ||
6708 	    time_after(jiffies, (adapter->link_check_timeout +
6709 				 IXGBE_TRY_LINK_TIMEOUT))) {
6710 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6711 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6712 		IXGBE_WRITE_FLUSH(hw);
6713 	}
6714 
6715 	adapter->link_up = link_up;
6716 	adapter->link_speed = link_speed;
6717 }
6718 
6719 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6720 {
6721 #ifdef CONFIG_IXGBE_DCB
6722 	struct net_device *netdev = adapter->netdev;
6723 	struct dcb_app app = {
6724 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6725 			      .protocol = 0,
6726 			     };
6727 	u8 up = 0;
6728 
6729 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6730 		up = dcb_ieee_getapp_mask(netdev, &app);
6731 
6732 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6733 #endif
6734 }
6735 
6736 /**
6737  * ixgbe_watchdog_link_is_up - update netif_carrier status and
6738  *                             print link up message
6739  * @adapter: pointer to the device adapter structure
6740  **/
6741 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6742 {
6743 	struct net_device *netdev = adapter->netdev;
6744 	struct ixgbe_hw *hw = &adapter->hw;
6745 	struct net_device *upper;
6746 	struct list_head *iter;
6747 	u32 link_speed = adapter->link_speed;
6748 	const char *speed_str;
6749 	bool flow_rx, flow_tx;
6750 
6751 	/* only continue if link was previously down */
6752 	if (netif_carrier_ok(netdev))
6753 		return;
6754 
6755 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6756 
6757 	switch (hw->mac.type) {
6758 	case ixgbe_mac_82598EB: {
6759 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6760 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6761 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6762 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6763 	}
6764 		break;
6765 	case ixgbe_mac_X540:
6766 	case ixgbe_mac_X550:
6767 	case ixgbe_mac_X550EM_x:
6768 	case ixgbe_mac_x550em_a:
6769 	case ixgbe_mac_82599EB: {
6770 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6771 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6772 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6773 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6774 	}
6775 		break;
6776 	default:
6777 		flow_tx = false;
6778 		flow_rx = false;
6779 		break;
6780 	}
6781 
6782 	adapter->last_rx_ptp_check = jiffies;
6783 
6784 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6785 		ixgbe_ptp_start_cyclecounter(adapter);
6786 
6787 	switch (link_speed) {
6788 	case IXGBE_LINK_SPEED_10GB_FULL:
6789 		speed_str = "10 Gbps";
6790 		break;
6791 	case IXGBE_LINK_SPEED_2_5GB_FULL:
6792 		speed_str = "2.5 Gbps";
6793 		break;
6794 	case IXGBE_LINK_SPEED_1GB_FULL:
6795 		speed_str = "1 Gbps";
6796 		break;
6797 	case IXGBE_LINK_SPEED_100_FULL:
6798 		speed_str = "100 Mbps";
6799 		break;
6800 	default:
6801 		speed_str = "unknown speed";
6802 		break;
6803 	}
6804 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
6805 	       ((flow_rx && flow_tx) ? "RX/TX" :
6806 	       (flow_rx ? "RX" :
6807 	       (flow_tx ? "TX" : "None"))));
6808 
6809 	netif_carrier_on(netdev);
6810 	ixgbe_check_vf_rate_limit(adapter);
6811 
6812 	/* enable transmits */
6813 	netif_tx_wake_all_queues(adapter->netdev);
6814 
6815 	/* enable any upper devices */
6816 	rtnl_lock();
6817 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
6818 		if (netif_is_macvlan(upper)) {
6819 			struct macvlan_dev *vlan = netdev_priv(upper);
6820 
6821 			if (vlan->fwd_priv)
6822 				netif_tx_wake_all_queues(upper);
6823 		}
6824 	}
6825 	rtnl_unlock();
6826 
6827 	/* update the default user priority for VFs */
6828 	ixgbe_update_default_up(adapter);
6829 
6830 	/* ping all the active vfs to let them know link has changed */
6831 	ixgbe_ping_all_vfs(adapter);
6832 }
6833 
6834 /**
6835  * ixgbe_watchdog_link_is_down - update netif_carrier status and
6836  *                               print link down message
6837  * @adapter: pointer to the adapter structure
6838  **/
6839 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6840 {
6841 	struct net_device *netdev = adapter->netdev;
6842 	struct ixgbe_hw *hw = &adapter->hw;
6843 
6844 	adapter->link_up = false;
6845 	adapter->link_speed = 0;
6846 
6847 	/* only continue if link was up previously */
6848 	if (!netif_carrier_ok(netdev))
6849 		return;
6850 
6851 	/* poll for SFP+ cable when link is down */
6852 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6853 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6854 
6855 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6856 		ixgbe_ptp_start_cyclecounter(adapter);
6857 
6858 	e_info(drv, "NIC Link is Down\n");
6859 	netif_carrier_off(netdev);
6860 
6861 	/* ping all the active vfs to let them know link has changed */
6862 	ixgbe_ping_all_vfs(adapter);
6863 }
6864 
6865 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
6866 {
6867 	int i;
6868 
6869 	for (i = 0; i < adapter->num_tx_queues; i++) {
6870 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6871 
6872 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
6873 			return true;
6874 	}
6875 
6876 	return false;
6877 }
6878 
6879 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
6880 {
6881 	struct ixgbe_hw *hw = &adapter->hw;
6882 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
6883 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
6884 
6885 	int i, j;
6886 
6887 	if (!adapter->num_vfs)
6888 		return false;
6889 
6890 	/* resetting the PF is only needed for MAC before X550 */
6891 	if (hw->mac.type >= ixgbe_mac_X550)
6892 		return false;
6893 
6894 	for (i = 0; i < adapter->num_vfs; i++) {
6895 		for (j = 0; j < q_per_pool; j++) {
6896 			u32 h, t;
6897 
6898 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
6899 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
6900 
6901 			if (h != t)
6902 				return true;
6903 		}
6904 	}
6905 
6906 	return false;
6907 }
6908 
6909 /**
6910  * ixgbe_watchdog_flush_tx - flush queues on link down
6911  * @adapter: pointer to the device adapter structure
6912  **/
6913 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6914 {
6915 	if (!netif_carrier_ok(adapter->netdev)) {
6916 		if (ixgbe_ring_tx_pending(adapter) ||
6917 		    ixgbe_vf_tx_pending(adapter)) {
6918 			/* We've lost link, so the controller stops DMA,
6919 			 * but we've got queued Tx work that's never going
6920 			 * to get done, so reset controller to flush Tx.
6921 			 * (Do the reset outside of interrupt context).
6922 			 */
6923 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6924 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
6925 		}
6926 	}
6927 }
6928 
6929 #ifdef CONFIG_PCI_IOV
6930 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
6931 				      struct pci_dev *vfdev)
6932 {
6933 	if (!pci_wait_for_pending_transaction(vfdev))
6934 		e_dev_warn("Issuing VFLR with pending transactions\n");
6935 
6936 	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
6937 	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
6938 
6939 	msleep(100);
6940 }
6941 
6942 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6943 {
6944 	struct ixgbe_hw *hw = &adapter->hw;
6945 	struct pci_dev *pdev = adapter->pdev;
6946 	unsigned int vf;
6947 	u32 gpc;
6948 
6949 	if (!(netif_carrier_ok(adapter->netdev)))
6950 		return;
6951 
6952 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6953 	if (gpc) /* If incrementing then no need for the check below */
6954 		return;
6955 	/* Check to see if a bad DMA write target from an errant or
6956 	 * malicious VF has caused a PCIe error.  If so then we can
6957 	 * issue a VFLR to the offending VF(s) and then resume without
6958 	 * requesting a full slot reset.
6959 	 */
6960 
6961 	if (!pdev)
6962 		return;
6963 
6964 	/* check status reg for all VFs owned by this PF */
6965 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
6966 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
6967 		u16 status_reg;
6968 
6969 		if (!vfdev)
6970 			continue;
6971 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
6972 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
6973 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
6974 			ixgbe_issue_vf_flr(adapter, vfdev);
6975 	}
6976 }
6977 
6978 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6979 {
6980 	u32 ssvpc;
6981 
6982 	/* Do not perform spoof check for 82598 or if not in IOV mode */
6983 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
6984 	    adapter->num_vfs == 0)
6985 		return;
6986 
6987 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
6988 
6989 	/*
6990 	 * ssvpc register is cleared on read, if zero then no
6991 	 * spoofed packets in the last interval.
6992 	 */
6993 	if (!ssvpc)
6994 		return;
6995 
6996 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
6997 }
6998 #else
6999 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7000 {
7001 }
7002 
7003 static void
7004 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7005 {
7006 }
7007 #endif /* CONFIG_PCI_IOV */
7008 
7009 
7010 /**
7011  * ixgbe_watchdog_subtask - check and bring link up
7012  * @adapter: pointer to the device adapter structure
7013  **/
7014 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7015 {
7016 	/* if interface is down, removing or resetting, do nothing */
7017 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7018 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7019 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7020 		return;
7021 
7022 	ixgbe_watchdog_update_link(adapter);
7023 
7024 	if (adapter->link_up)
7025 		ixgbe_watchdog_link_is_up(adapter);
7026 	else
7027 		ixgbe_watchdog_link_is_down(adapter);
7028 
7029 	ixgbe_check_for_bad_vf(adapter);
7030 	ixgbe_spoof_check(adapter);
7031 	ixgbe_update_stats(adapter);
7032 
7033 	ixgbe_watchdog_flush_tx(adapter);
7034 }
7035 
7036 /**
7037  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7038  * @adapter: the ixgbe adapter structure
7039  **/
7040 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7041 {
7042 	struct ixgbe_hw *hw = &adapter->hw;
7043 	s32 err;
7044 
7045 	/* not searching for SFP so there is nothing to do here */
7046 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7047 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7048 		return;
7049 
7050 	if (adapter->sfp_poll_time &&
7051 	    time_after(adapter->sfp_poll_time, jiffies))
7052 		return; /* If not yet time to poll for SFP */
7053 
7054 	/* someone else is in init, wait until next service event */
7055 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7056 		return;
7057 
7058 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7059 
7060 	err = hw->phy.ops.identify_sfp(hw);
7061 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7062 		goto sfp_out;
7063 
7064 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7065 		/* If no cable is present, then we need to reset
7066 		 * the next time we find a good cable. */
7067 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7068 	}
7069 
7070 	/* exit on error */
7071 	if (err)
7072 		goto sfp_out;
7073 
7074 	/* exit if reset not needed */
7075 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7076 		goto sfp_out;
7077 
7078 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7079 
7080 	/*
7081 	 * A module may be identified correctly, but the EEPROM may not have
7082 	 * support for that module.  setup_sfp() will fail in that case, so
7083 	 * we should not allow that module to load.
7084 	 */
7085 	if (hw->mac.type == ixgbe_mac_82598EB)
7086 		err = hw->phy.ops.reset(hw);
7087 	else
7088 		err = hw->mac.ops.setup_sfp(hw);
7089 
7090 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7091 		goto sfp_out;
7092 
7093 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7094 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7095 
7096 sfp_out:
7097 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7098 
7099 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7100 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7101 		e_dev_err("failed to initialize because an unsupported "
7102 			  "SFP+ module type was detected.\n");
7103 		e_dev_err("Reload the driver after installing a "
7104 			  "supported module.\n");
7105 		unregister_netdev(adapter->netdev);
7106 	}
7107 }
7108 
7109 /**
7110  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7111  * @adapter: the ixgbe adapter structure
7112  **/
7113 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7114 {
7115 	struct ixgbe_hw *hw = &adapter->hw;
7116 	u32 speed;
7117 	bool autoneg = false;
7118 
7119 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7120 		return;
7121 
7122 	/* someone else is in init, wait until next service event */
7123 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7124 		return;
7125 
7126 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7127 
7128 	speed = hw->phy.autoneg_advertised;
7129 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
7130 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
7131 
7132 		/* setup the highest link when no autoneg */
7133 		if (!autoneg) {
7134 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
7135 				speed = IXGBE_LINK_SPEED_10GB_FULL;
7136 		}
7137 	}
7138 
7139 	if (hw->mac.ops.setup_link)
7140 		hw->mac.ops.setup_link(hw, speed, true);
7141 
7142 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7143 	adapter->link_check_timeout = jiffies;
7144 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7145 }
7146 
7147 /**
7148  * ixgbe_service_timer - Timer Call-back
7149  * @data: pointer to adapter cast into an unsigned long
7150  **/
7151 static void ixgbe_service_timer(unsigned long data)
7152 {
7153 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
7154 	unsigned long next_event_offset;
7155 
7156 	/* poll faster when waiting for link */
7157 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7158 		next_event_offset = HZ / 10;
7159 	else
7160 		next_event_offset = HZ * 2;
7161 
7162 	/* Reset the timer */
7163 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7164 
7165 	ixgbe_service_event_schedule(adapter);
7166 }
7167 
7168 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7169 {
7170 	struct ixgbe_hw *hw = &adapter->hw;
7171 	u32 status;
7172 
7173 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7174 		return;
7175 
7176 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7177 
7178 	if (!hw->phy.ops.handle_lasi)
7179 		return;
7180 
7181 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7182 	if (status != IXGBE_ERR_OVERTEMP)
7183 		return;
7184 
7185 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7186 }
7187 
7188 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7189 {
7190 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
7191 		return;
7192 
7193 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
7194 
7195 	/* If we're already down, removing or resetting, just bail */
7196 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7197 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7198 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7199 		return;
7200 
7201 	ixgbe_dump(adapter);
7202 	netdev_err(adapter->netdev, "Reset adapter\n");
7203 	adapter->tx_timeout_count++;
7204 
7205 	rtnl_lock();
7206 	ixgbe_reinit_locked(adapter);
7207 	rtnl_unlock();
7208 }
7209 
7210 /**
7211  * ixgbe_service_task - manages and runs subtasks
7212  * @work: pointer to work_struct containing our data
7213  **/
7214 static void ixgbe_service_task(struct work_struct *work)
7215 {
7216 	struct ixgbe_adapter *adapter = container_of(work,
7217 						     struct ixgbe_adapter,
7218 						     service_task);
7219 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7220 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7221 			rtnl_lock();
7222 			ixgbe_down(adapter);
7223 			rtnl_unlock();
7224 		}
7225 		ixgbe_service_event_complete(adapter);
7226 		return;
7227 	}
7228 	if (adapter->flags2 & IXGBE_FLAG2_VXLAN_REREG_NEEDED) {
7229 		rtnl_lock();
7230 		adapter->flags2 &= ~IXGBE_FLAG2_VXLAN_REREG_NEEDED;
7231 		udp_tunnel_get_rx_info(adapter->netdev);
7232 		rtnl_unlock();
7233 	}
7234 	ixgbe_reset_subtask(adapter);
7235 	ixgbe_phy_interrupt_subtask(adapter);
7236 	ixgbe_sfp_detection_subtask(adapter);
7237 	ixgbe_sfp_link_config_subtask(adapter);
7238 	ixgbe_check_overtemp_subtask(adapter);
7239 	ixgbe_watchdog_subtask(adapter);
7240 	ixgbe_fdir_reinit_subtask(adapter);
7241 	ixgbe_check_hang_subtask(adapter);
7242 
7243 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7244 		ixgbe_ptp_overflow_check(adapter);
7245 		ixgbe_ptp_rx_hang(adapter);
7246 	}
7247 
7248 	ixgbe_service_event_complete(adapter);
7249 }
7250 
7251 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7252 		     struct ixgbe_tx_buffer *first,
7253 		     u8 *hdr_len)
7254 {
7255 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7256 	struct sk_buff *skb = first->skb;
7257 	union {
7258 		struct iphdr *v4;
7259 		struct ipv6hdr *v6;
7260 		unsigned char *hdr;
7261 	} ip;
7262 	union {
7263 		struct tcphdr *tcp;
7264 		unsigned char *hdr;
7265 	} l4;
7266 	u32 paylen, l4_offset;
7267 	int err;
7268 
7269 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7270 		return 0;
7271 
7272 	if (!skb_is_gso(skb))
7273 		return 0;
7274 
7275 	err = skb_cow_head(skb, 0);
7276 	if (err < 0)
7277 		return err;
7278 
7279 	ip.hdr = skb_network_header(skb);
7280 	l4.hdr = skb_checksum_start(skb);
7281 
7282 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7283 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7284 
7285 	/* initialize outer IP header fields */
7286 	if (ip.v4->version == 4) {
7287 		/* IP header will have to cancel out any data that
7288 		 * is not a part of the outer IP header
7289 		 */
7290 		ip.v4->check = csum_fold(csum_add(lco_csum(skb),
7291 						  csum_unfold(l4.tcp->check)));
7292 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7293 
7294 		ip.v4->tot_len = 0;
7295 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7296 				   IXGBE_TX_FLAGS_CSUM |
7297 				   IXGBE_TX_FLAGS_IPV4;
7298 	} else {
7299 		ip.v6->payload_len = 0;
7300 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7301 				   IXGBE_TX_FLAGS_CSUM;
7302 	}
7303 
7304 	/* determine offset of inner transport header */
7305 	l4_offset = l4.hdr - skb->data;
7306 
7307 	/* compute length of segmentation header */
7308 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7309 
7310 	/* remove payload length from inner checksum */
7311 	paylen = skb->len - l4_offset;
7312 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
7313 
7314 	/* update gso size and bytecount with header size */
7315 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7316 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7317 
7318 	/* mss_l4len_id: use 0 as index for TSO */
7319 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7320 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7321 
7322 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7323 	vlan_macip_lens = l4.hdr - ip.hdr;
7324 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7325 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7326 
7327 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
7328 			  mss_l4len_idx);
7329 
7330 	return 1;
7331 }
7332 
7333 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7334 {
7335 	unsigned int offset = 0;
7336 
7337 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7338 
7339 	return offset == skb_checksum_start_offset(skb);
7340 }
7341 
7342 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7343 			  struct ixgbe_tx_buffer *first)
7344 {
7345 	struct sk_buff *skb = first->skb;
7346 	u32 vlan_macip_lens = 0;
7347 	u32 type_tucmd = 0;
7348 
7349 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7350 csum_failed:
7351 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7352 					 IXGBE_TX_FLAGS_CC)))
7353 			return;
7354 		goto no_csum;
7355 	}
7356 
7357 	switch (skb->csum_offset) {
7358 	case offsetof(struct tcphdr, check):
7359 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7360 		/* fall through */
7361 	case offsetof(struct udphdr, check):
7362 		break;
7363 	case offsetof(struct sctphdr, checksum):
7364 		/* validate that this is actually an SCTP request */
7365 		if (((first->protocol == htons(ETH_P_IP)) &&
7366 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7367 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7368 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7369 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7370 			break;
7371 		}
7372 		/* fall through */
7373 	default:
7374 		skb_checksum_help(skb);
7375 		goto csum_failed;
7376 	}
7377 
7378 	/* update TX checksum flag */
7379 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7380 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7381 			  skb_network_offset(skb);
7382 no_csum:
7383 	/* vlan_macip_lens: MACLEN, VLAN tag */
7384 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7385 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7386 
7387 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd, 0);
7388 }
7389 
7390 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7391 	((_flag <= _result) ? \
7392 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7393 	 ((u32)(_input & _flag) / (_flag / _result)))
7394 
7395 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7396 {
7397 	/* set type for advanced descriptor with frame checksum insertion */
7398 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7399 		       IXGBE_ADVTXD_DCMD_DEXT |
7400 		       IXGBE_ADVTXD_DCMD_IFCS;
7401 
7402 	/* set HW vlan bit if vlan is present */
7403 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7404 				   IXGBE_ADVTXD_DCMD_VLE);
7405 
7406 	/* set segmentation enable bits for TSO/FSO */
7407 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7408 				   IXGBE_ADVTXD_DCMD_TSE);
7409 
7410 	/* set timestamp bit if present */
7411 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7412 				   IXGBE_ADVTXD_MAC_TSTAMP);
7413 
7414 	/* insert frame checksum */
7415 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7416 
7417 	return cmd_type;
7418 }
7419 
7420 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7421 				   u32 tx_flags, unsigned int paylen)
7422 {
7423 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7424 
7425 	/* enable L4 checksum for TSO and TX checksum offload */
7426 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7427 					IXGBE_TX_FLAGS_CSUM,
7428 					IXGBE_ADVTXD_POPTS_TXSM);
7429 
7430 	/* enble IPv4 checksum for TSO */
7431 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7432 					IXGBE_TX_FLAGS_IPV4,
7433 					IXGBE_ADVTXD_POPTS_IXSM);
7434 
7435 	/*
7436 	 * Check Context must be set if Tx switch is enabled, which it
7437 	 * always is for case where virtual functions are running
7438 	 */
7439 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7440 					IXGBE_TX_FLAGS_CC,
7441 					IXGBE_ADVTXD_CC);
7442 
7443 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7444 }
7445 
7446 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7447 {
7448 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7449 
7450 	/* Herbert's original patch had:
7451 	 *  smp_mb__after_netif_stop_queue();
7452 	 * but since that doesn't exist yet, just open code it.
7453 	 */
7454 	smp_mb();
7455 
7456 	/* We need to check again in a case another CPU has just
7457 	 * made room available.
7458 	 */
7459 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7460 		return -EBUSY;
7461 
7462 	/* A reprieve! - use start_queue because it doesn't call schedule */
7463 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7464 	++tx_ring->tx_stats.restart_queue;
7465 	return 0;
7466 }
7467 
7468 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7469 {
7470 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7471 		return 0;
7472 
7473 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7474 }
7475 
7476 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7477 		       IXGBE_TXD_CMD_RS)
7478 
7479 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7480 			 struct ixgbe_tx_buffer *first,
7481 			 const u8 hdr_len)
7482 {
7483 	struct sk_buff *skb = first->skb;
7484 	struct ixgbe_tx_buffer *tx_buffer;
7485 	union ixgbe_adv_tx_desc *tx_desc;
7486 	struct skb_frag_struct *frag;
7487 	dma_addr_t dma;
7488 	unsigned int data_len, size;
7489 	u32 tx_flags = first->tx_flags;
7490 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7491 	u16 i = tx_ring->next_to_use;
7492 
7493 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7494 
7495 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7496 
7497 	size = skb_headlen(skb);
7498 	data_len = skb->data_len;
7499 
7500 #ifdef IXGBE_FCOE
7501 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7502 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7503 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7504 			data_len = 0;
7505 		} else {
7506 			data_len -= sizeof(struct fcoe_crc_eof);
7507 		}
7508 	}
7509 
7510 #endif
7511 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7512 
7513 	tx_buffer = first;
7514 
7515 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7516 		if (dma_mapping_error(tx_ring->dev, dma))
7517 			goto dma_error;
7518 
7519 		/* record length, and DMA address */
7520 		dma_unmap_len_set(tx_buffer, len, size);
7521 		dma_unmap_addr_set(tx_buffer, dma, dma);
7522 
7523 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7524 
7525 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7526 			tx_desc->read.cmd_type_len =
7527 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7528 
7529 			i++;
7530 			tx_desc++;
7531 			if (i == tx_ring->count) {
7532 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7533 				i = 0;
7534 			}
7535 			tx_desc->read.olinfo_status = 0;
7536 
7537 			dma += IXGBE_MAX_DATA_PER_TXD;
7538 			size -= IXGBE_MAX_DATA_PER_TXD;
7539 
7540 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7541 		}
7542 
7543 		if (likely(!data_len))
7544 			break;
7545 
7546 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7547 
7548 		i++;
7549 		tx_desc++;
7550 		if (i == tx_ring->count) {
7551 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7552 			i = 0;
7553 		}
7554 		tx_desc->read.olinfo_status = 0;
7555 
7556 #ifdef IXGBE_FCOE
7557 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7558 #else
7559 		size = skb_frag_size(frag);
7560 #endif
7561 		data_len -= size;
7562 
7563 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7564 				       DMA_TO_DEVICE);
7565 
7566 		tx_buffer = &tx_ring->tx_buffer_info[i];
7567 	}
7568 
7569 	/* write last descriptor with RS and EOP bits */
7570 	cmd_type |= size | IXGBE_TXD_CMD;
7571 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7572 
7573 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7574 
7575 	/* set the timestamp */
7576 	first->time_stamp = jiffies;
7577 
7578 	/*
7579 	 * Force memory writes to complete before letting h/w know there
7580 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7581 	 * memory model archs, such as IA-64).
7582 	 *
7583 	 * We also need this memory barrier to make certain all of the
7584 	 * status bits have been updated before next_to_watch is written.
7585 	 */
7586 	wmb();
7587 
7588 	/* set next_to_watch value indicating a packet is present */
7589 	first->next_to_watch = tx_desc;
7590 
7591 	i++;
7592 	if (i == tx_ring->count)
7593 		i = 0;
7594 
7595 	tx_ring->next_to_use = i;
7596 
7597 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7598 
7599 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7600 		writel(i, tx_ring->tail);
7601 
7602 		/* we need this if more than one processor can write to our tail
7603 		 * at a time, it synchronizes IO on IA64/Altix systems
7604 		 */
7605 		mmiowb();
7606 	}
7607 
7608 	return;
7609 dma_error:
7610 	dev_err(tx_ring->dev, "TX DMA map failed\n");
7611 
7612 	/* clear dma mappings for failed tx_buffer_info map */
7613 	for (;;) {
7614 		tx_buffer = &tx_ring->tx_buffer_info[i];
7615 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
7616 		if (tx_buffer == first)
7617 			break;
7618 		if (i == 0)
7619 			i = tx_ring->count;
7620 		i--;
7621 	}
7622 
7623 	tx_ring->next_to_use = i;
7624 }
7625 
7626 static void ixgbe_atr(struct ixgbe_ring *ring,
7627 		      struct ixgbe_tx_buffer *first)
7628 {
7629 	struct ixgbe_q_vector *q_vector = ring->q_vector;
7630 	union ixgbe_atr_hash_dword input = { .dword = 0 };
7631 	union ixgbe_atr_hash_dword common = { .dword = 0 };
7632 	union {
7633 		unsigned char *network;
7634 		struct iphdr *ipv4;
7635 		struct ipv6hdr *ipv6;
7636 	} hdr;
7637 	struct tcphdr *th;
7638 	unsigned int hlen;
7639 	struct sk_buff *skb;
7640 	__be16 vlan_id;
7641 	int l4_proto;
7642 
7643 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7644 	if (!q_vector)
7645 		return;
7646 
7647 	/* do nothing if sampling is disabled */
7648 	if (!ring->atr_sample_rate)
7649 		return;
7650 
7651 	ring->atr_count++;
7652 
7653 	/* currently only IPv4/IPv6 with TCP is supported */
7654 	if ((first->protocol != htons(ETH_P_IP)) &&
7655 	    (first->protocol != htons(ETH_P_IPV6)))
7656 		return;
7657 
7658 	/* snag network header to get L4 type and address */
7659 	skb = first->skb;
7660 	hdr.network = skb_network_header(skb);
7661 	if (skb->encapsulation &&
7662 	    first->protocol == htons(ETH_P_IP) &&
7663 	    hdr.ipv4->protocol != IPPROTO_UDP) {
7664 		struct ixgbe_adapter *adapter = q_vector->adapter;
7665 
7666 		/* verify the port is recognized as VXLAN */
7667 		if (adapter->vxlan_port &&
7668 		    udp_hdr(skb)->dest == adapter->vxlan_port)
7669 			hdr.network = skb_inner_network_header(skb);
7670 	}
7671 
7672 	/* Currently only IPv4/IPv6 with TCP is supported */
7673 	switch (hdr.ipv4->version) {
7674 	case IPVERSION:
7675 		/* access ihl as u8 to avoid unaligned access on ia64 */
7676 		hlen = (hdr.network[0] & 0x0F) << 2;
7677 		l4_proto = hdr.ipv4->protocol;
7678 		break;
7679 	case 6:
7680 		hlen = hdr.network - skb->data;
7681 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
7682 		hlen -= hdr.network - skb->data;
7683 		break;
7684 	default:
7685 		return;
7686 	}
7687 
7688 	if (l4_proto != IPPROTO_TCP)
7689 		return;
7690 
7691 	th = (struct tcphdr *)(hdr.network + hlen);
7692 
7693 	/* skip this packet since the socket is closing */
7694 	if (th->fin)
7695 		return;
7696 
7697 	/* sample on all syn packets or once every atr sample count */
7698 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7699 		return;
7700 
7701 	/* reset sample count */
7702 	ring->atr_count = 0;
7703 
7704 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7705 
7706 	/*
7707 	 * src and dst are inverted, think how the receiver sees them
7708 	 *
7709 	 * The input is broken into two sections, a non-compressed section
7710 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7711 	 * is XORed together and stored in the compressed dword.
7712 	 */
7713 	input.formatted.vlan_id = vlan_id;
7714 
7715 	/*
7716 	 * since src port and flex bytes occupy the same word XOR them together
7717 	 * and write the value to source port portion of compressed dword
7718 	 */
7719 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7720 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7721 	else
7722 		common.port.src ^= th->dest ^ first->protocol;
7723 	common.port.dst ^= th->source;
7724 
7725 	switch (hdr.ipv4->version) {
7726 	case IPVERSION:
7727 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7728 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7729 		break;
7730 	case 6:
7731 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7732 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7733 			     hdr.ipv6->saddr.s6_addr32[1] ^
7734 			     hdr.ipv6->saddr.s6_addr32[2] ^
7735 			     hdr.ipv6->saddr.s6_addr32[3] ^
7736 			     hdr.ipv6->daddr.s6_addr32[0] ^
7737 			     hdr.ipv6->daddr.s6_addr32[1] ^
7738 			     hdr.ipv6->daddr.s6_addr32[2] ^
7739 			     hdr.ipv6->daddr.s6_addr32[3];
7740 		break;
7741 	default:
7742 		break;
7743 	}
7744 
7745 	if (hdr.network != skb_network_header(skb))
7746 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
7747 
7748 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7749 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7750 					      input, common, ring->queue_index);
7751 }
7752 
7753 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7754 			      void *accel_priv, select_queue_fallback_t fallback)
7755 {
7756 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7757 #ifdef IXGBE_FCOE
7758 	struct ixgbe_adapter *adapter;
7759 	struct ixgbe_ring_feature *f;
7760 	int txq;
7761 #endif
7762 
7763 	if (fwd_adapter)
7764 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7765 
7766 #ifdef IXGBE_FCOE
7767 
7768 	/*
7769 	 * only execute the code below if protocol is FCoE
7770 	 * or FIP and we have FCoE enabled on the adapter
7771 	 */
7772 	switch (vlan_get_protocol(skb)) {
7773 	case htons(ETH_P_FCOE):
7774 	case htons(ETH_P_FIP):
7775 		adapter = netdev_priv(dev);
7776 
7777 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7778 			break;
7779 	default:
7780 		return fallback(dev, skb);
7781 	}
7782 
7783 	f = &adapter->ring_feature[RING_F_FCOE];
7784 
7785 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7786 					   smp_processor_id();
7787 
7788 	while (txq >= f->indices)
7789 		txq -= f->indices;
7790 
7791 	return txq + f->offset;
7792 #else
7793 	return fallback(dev, skb);
7794 #endif
7795 }
7796 
7797 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7798 			  struct ixgbe_adapter *adapter,
7799 			  struct ixgbe_ring *tx_ring)
7800 {
7801 	struct ixgbe_tx_buffer *first;
7802 	int tso;
7803 	u32 tx_flags = 0;
7804 	unsigned short f;
7805 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7806 	__be16 protocol = skb->protocol;
7807 	u8 hdr_len = 0;
7808 
7809 	/*
7810 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7811 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7812 	 *       + 2 desc gap to keep tail from touching head,
7813 	 *       + 1 desc for context descriptor,
7814 	 * otherwise try next time
7815 	 */
7816 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7817 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7818 
7819 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7820 		tx_ring->tx_stats.tx_busy++;
7821 		return NETDEV_TX_BUSY;
7822 	}
7823 
7824 	/* record the location of the first descriptor for this packet */
7825 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7826 	first->skb = skb;
7827 	first->bytecount = skb->len;
7828 	first->gso_segs = 1;
7829 
7830 	/* if we have a HW VLAN tag being added default to the HW one */
7831 	if (skb_vlan_tag_present(skb)) {
7832 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7833 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7834 	/* else if it is a SW VLAN check the next protocol and store the tag */
7835 	} else if (protocol == htons(ETH_P_8021Q)) {
7836 		struct vlan_hdr *vhdr, _vhdr;
7837 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7838 		if (!vhdr)
7839 			goto out_drop;
7840 
7841 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7842 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7843 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7844 	}
7845 	protocol = vlan_get_protocol(skb);
7846 
7847 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
7848 	    adapter->ptp_clock &&
7849 	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7850 				   &adapter->state)) {
7851 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7852 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7853 
7854 		/* schedule check for Tx timestamp */
7855 		adapter->ptp_tx_skb = skb_get(skb);
7856 		adapter->ptp_tx_start = jiffies;
7857 		schedule_work(&adapter->ptp_tx_work);
7858 	}
7859 
7860 	skb_tx_timestamp(skb);
7861 
7862 #ifdef CONFIG_PCI_IOV
7863 	/*
7864 	 * Use the l2switch_enable flag - would be false if the DMA
7865 	 * Tx switch had been disabled.
7866 	 */
7867 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7868 		tx_flags |= IXGBE_TX_FLAGS_CC;
7869 
7870 #endif
7871 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7872 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7873 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7874 	     (skb->priority != TC_PRIO_CONTROL))) {
7875 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7876 		tx_flags |= (skb->priority & 0x7) <<
7877 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7878 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7879 			struct vlan_ethhdr *vhdr;
7880 
7881 			if (skb_cow_head(skb, 0))
7882 				goto out_drop;
7883 			vhdr = (struct vlan_ethhdr *)skb->data;
7884 			vhdr->h_vlan_TCI = htons(tx_flags >>
7885 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7886 		} else {
7887 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7888 		}
7889 	}
7890 
7891 	/* record initial flags and protocol */
7892 	first->tx_flags = tx_flags;
7893 	first->protocol = protocol;
7894 
7895 #ifdef IXGBE_FCOE
7896 	/* setup tx offload for FCoE */
7897 	if ((protocol == htons(ETH_P_FCOE)) &&
7898 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7899 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7900 		if (tso < 0)
7901 			goto out_drop;
7902 
7903 		goto xmit_fcoe;
7904 	}
7905 
7906 #endif /* IXGBE_FCOE */
7907 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7908 	if (tso < 0)
7909 		goto out_drop;
7910 	else if (!tso)
7911 		ixgbe_tx_csum(tx_ring, first);
7912 
7913 	/* add the ATR filter if ATR is on */
7914 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7915 		ixgbe_atr(tx_ring, first);
7916 
7917 #ifdef IXGBE_FCOE
7918 xmit_fcoe:
7919 #endif /* IXGBE_FCOE */
7920 	ixgbe_tx_map(tx_ring, first, hdr_len);
7921 
7922 	return NETDEV_TX_OK;
7923 
7924 out_drop:
7925 	dev_kfree_skb_any(first->skb);
7926 	first->skb = NULL;
7927 
7928 	return NETDEV_TX_OK;
7929 }
7930 
7931 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7932 				      struct net_device *netdev,
7933 				      struct ixgbe_ring *ring)
7934 {
7935 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7936 	struct ixgbe_ring *tx_ring;
7937 
7938 	/*
7939 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7940 	 * in order to meet this minimum size requirement.
7941 	 */
7942 	if (skb_put_padto(skb, 17))
7943 		return NETDEV_TX_OK;
7944 
7945 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7946 
7947 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7948 }
7949 
7950 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7951 				    struct net_device *netdev)
7952 {
7953 	return __ixgbe_xmit_frame(skb, netdev, NULL);
7954 }
7955 
7956 /**
7957  * ixgbe_set_mac - Change the Ethernet Address of the NIC
7958  * @netdev: network interface device structure
7959  * @p: pointer to an address structure
7960  *
7961  * Returns 0 on success, negative on failure
7962  **/
7963 static int ixgbe_set_mac(struct net_device *netdev, void *p)
7964 {
7965 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7966 	struct ixgbe_hw *hw = &adapter->hw;
7967 	struct sockaddr *addr = p;
7968 
7969 	if (!is_valid_ether_addr(addr->sa_data))
7970 		return -EADDRNOTAVAIL;
7971 
7972 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
7973 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
7974 
7975 	ixgbe_mac_set_default_filter(adapter);
7976 
7977 	return 0;
7978 }
7979 
7980 static int
7981 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
7982 {
7983 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7984 	struct ixgbe_hw *hw = &adapter->hw;
7985 	u16 value;
7986 	int rc;
7987 
7988 	if (prtad != hw->phy.mdio.prtad)
7989 		return -EINVAL;
7990 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
7991 	if (!rc)
7992 		rc = value;
7993 	return rc;
7994 }
7995 
7996 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
7997 			    u16 addr, u16 value)
7998 {
7999 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8000 	struct ixgbe_hw *hw = &adapter->hw;
8001 
8002 	if (prtad != hw->phy.mdio.prtad)
8003 		return -EINVAL;
8004 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8005 }
8006 
8007 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8008 {
8009 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8010 
8011 	switch (cmd) {
8012 	case SIOCSHWTSTAMP:
8013 		return ixgbe_ptp_set_ts_config(adapter, req);
8014 	case SIOCGHWTSTAMP:
8015 		return ixgbe_ptp_get_ts_config(adapter, req);
8016 	default:
8017 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8018 	}
8019 }
8020 
8021 /**
8022  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8023  * netdev->dev_addrs
8024  * @netdev: network interface device structure
8025  *
8026  * Returns non-zero on failure
8027  **/
8028 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8029 {
8030 	int err = 0;
8031 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8032 	struct ixgbe_hw *hw = &adapter->hw;
8033 
8034 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8035 		rtnl_lock();
8036 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8037 		rtnl_unlock();
8038 
8039 		/* update SAN MAC vmdq pool selection */
8040 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8041 	}
8042 	return err;
8043 }
8044 
8045 /**
8046  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8047  * netdev->dev_addrs
8048  * @netdev: network interface device structure
8049  *
8050  * Returns non-zero on failure
8051  **/
8052 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8053 {
8054 	int err = 0;
8055 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8056 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8057 
8058 	if (is_valid_ether_addr(mac->san_addr)) {
8059 		rtnl_lock();
8060 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8061 		rtnl_unlock();
8062 	}
8063 	return err;
8064 }
8065 
8066 #ifdef CONFIG_NET_POLL_CONTROLLER
8067 /*
8068  * Polling 'interrupt' - used by things like netconsole to send skbs
8069  * without having to re-enable interrupts. It's not called while
8070  * the interrupt routine is executing.
8071  */
8072 static void ixgbe_netpoll(struct net_device *netdev)
8073 {
8074 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8075 	int i;
8076 
8077 	/* if interface is down do nothing */
8078 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8079 		return;
8080 
8081 	/* loop through and schedule all active queues */
8082 	for (i = 0; i < adapter->num_q_vectors; i++)
8083 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8084 }
8085 
8086 #endif
8087 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
8088 						   struct rtnl_link_stats64 *stats)
8089 {
8090 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8091 	int i;
8092 
8093 	rcu_read_lock();
8094 	for (i = 0; i < adapter->num_rx_queues; i++) {
8095 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
8096 		u64 bytes, packets;
8097 		unsigned int start;
8098 
8099 		if (ring) {
8100 			do {
8101 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8102 				packets = ring->stats.packets;
8103 				bytes   = ring->stats.bytes;
8104 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8105 			stats->rx_packets += packets;
8106 			stats->rx_bytes   += bytes;
8107 		}
8108 	}
8109 
8110 	for (i = 0; i < adapter->num_tx_queues; i++) {
8111 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
8112 		u64 bytes, packets;
8113 		unsigned int start;
8114 
8115 		if (ring) {
8116 			do {
8117 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8118 				packets = ring->stats.packets;
8119 				bytes   = ring->stats.bytes;
8120 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8121 			stats->tx_packets += packets;
8122 			stats->tx_bytes   += bytes;
8123 		}
8124 	}
8125 	rcu_read_unlock();
8126 	/* following stats updated by ixgbe_watchdog_task() */
8127 	stats->multicast	= netdev->stats.multicast;
8128 	stats->rx_errors	= netdev->stats.rx_errors;
8129 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8130 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8131 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8132 	return stats;
8133 }
8134 
8135 #ifdef CONFIG_IXGBE_DCB
8136 /**
8137  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8138  * @adapter: pointer to ixgbe_adapter
8139  * @tc: number of traffic classes currently enabled
8140  *
8141  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8142  * 802.1Q priority maps to a packet buffer that exists.
8143  */
8144 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8145 {
8146 	struct ixgbe_hw *hw = &adapter->hw;
8147 	u32 reg, rsave;
8148 	int i;
8149 
8150 	/* 82598 have a static priority to TC mapping that can not
8151 	 * be changed so no validation is needed.
8152 	 */
8153 	if (hw->mac.type == ixgbe_mac_82598EB)
8154 		return;
8155 
8156 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8157 	rsave = reg;
8158 
8159 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8160 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8161 
8162 		/* If up2tc is out of bounds default to zero */
8163 		if (up2tc > tc)
8164 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8165 	}
8166 
8167 	if (reg != rsave)
8168 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8169 
8170 	return;
8171 }
8172 
8173 /**
8174  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8175  * @adapter: Pointer to adapter struct
8176  *
8177  * Populate the netdev user priority to tc map
8178  */
8179 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8180 {
8181 	struct net_device *dev = adapter->netdev;
8182 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8183 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8184 	u8 prio;
8185 
8186 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8187 		u8 tc = 0;
8188 
8189 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8190 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8191 		else if (ets)
8192 			tc = ets->prio_tc[prio];
8193 
8194 		netdev_set_prio_tc_map(dev, prio, tc);
8195 	}
8196 }
8197 
8198 #endif /* CONFIG_IXGBE_DCB */
8199 /**
8200  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8201  *
8202  * @netdev: net device to configure
8203  * @tc: number of traffic classes to enable
8204  */
8205 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8206 {
8207 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8208 	struct ixgbe_hw *hw = &adapter->hw;
8209 	bool pools;
8210 
8211 	/* Hardware supports up to 8 traffic classes */
8212 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8213 		return -EINVAL;
8214 
8215 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8216 		return -EINVAL;
8217 
8218 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
8219 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
8220 		return -EBUSY;
8221 
8222 	/* Hardware has to reinitialize queues and interrupts to
8223 	 * match packet buffer alignment. Unfortunately, the
8224 	 * hardware is not flexible enough to do this dynamically.
8225 	 */
8226 	if (netif_running(dev))
8227 		ixgbe_close(dev);
8228 	else
8229 		ixgbe_reset(adapter);
8230 
8231 	ixgbe_clear_interrupt_scheme(adapter);
8232 
8233 #ifdef CONFIG_IXGBE_DCB
8234 	if (tc) {
8235 		netdev_set_num_tc(dev, tc);
8236 		ixgbe_set_prio_tc_map(adapter);
8237 
8238 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8239 
8240 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8241 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8242 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8243 		}
8244 	} else {
8245 		netdev_reset_tc(dev);
8246 
8247 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8248 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8249 
8250 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8251 
8252 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8253 		adapter->dcb_cfg.pfc_mode_enable = false;
8254 	}
8255 
8256 	ixgbe_validate_rtr(adapter, tc);
8257 
8258 #endif /* CONFIG_IXGBE_DCB */
8259 	ixgbe_init_interrupt_scheme(adapter);
8260 
8261 	if (netif_running(dev))
8262 		return ixgbe_open(dev);
8263 
8264 	return 0;
8265 }
8266 
8267 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8268 			       struct tc_cls_u32_offload *cls)
8269 {
8270 	u32 hdl = cls->knode.handle;
8271 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8272 	u32 loc = cls->knode.handle & 0xfffff;
8273 	int err = 0, i, j;
8274 	struct ixgbe_jump_table *jump = NULL;
8275 
8276 	if (loc > IXGBE_MAX_HW_ENTRIES)
8277 		return -EINVAL;
8278 
8279 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8280 		return -EINVAL;
8281 
8282 	/* Clear this filter in the link data it is associated with */
8283 	if (uhtid != 0x800) {
8284 		jump = adapter->jump_tables[uhtid];
8285 		if (!jump)
8286 			return -EINVAL;
8287 		if (!test_bit(loc - 1, jump->child_loc_map))
8288 			return -EINVAL;
8289 		clear_bit(loc - 1, jump->child_loc_map);
8290 	}
8291 
8292 	/* Check if the filter being deleted is a link */
8293 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8294 		jump = adapter->jump_tables[i];
8295 		if (jump && jump->link_hdl == hdl) {
8296 			/* Delete filters in the hardware in the child hash
8297 			 * table associated with this link
8298 			 */
8299 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8300 				if (!test_bit(j, jump->child_loc_map))
8301 					continue;
8302 				spin_lock(&adapter->fdir_perfect_lock);
8303 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8304 								      NULL,
8305 								      j + 1);
8306 				spin_unlock(&adapter->fdir_perfect_lock);
8307 				clear_bit(j, jump->child_loc_map);
8308 			}
8309 			/* Remove resources for this link */
8310 			kfree(jump->input);
8311 			kfree(jump->mask);
8312 			kfree(jump);
8313 			adapter->jump_tables[i] = NULL;
8314 			return err;
8315 		}
8316 	}
8317 
8318 	spin_lock(&adapter->fdir_perfect_lock);
8319 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8320 	spin_unlock(&adapter->fdir_perfect_lock);
8321 	return err;
8322 }
8323 
8324 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8325 					    __be16 protocol,
8326 					    struct tc_cls_u32_offload *cls)
8327 {
8328 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8329 
8330 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8331 		return -EINVAL;
8332 
8333 	/* This ixgbe devices do not support hash tables at the moment
8334 	 * so abort when given hash tables.
8335 	 */
8336 	if (cls->hnode.divisor > 0)
8337 		return -EINVAL;
8338 
8339 	set_bit(uhtid - 1, &adapter->tables);
8340 	return 0;
8341 }
8342 
8343 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
8344 					    struct tc_cls_u32_offload *cls)
8345 {
8346 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8347 
8348 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8349 		return -EINVAL;
8350 
8351 	clear_bit(uhtid - 1, &adapter->tables);
8352 	return 0;
8353 }
8354 
8355 #ifdef CONFIG_NET_CLS_ACT
8356 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
8357 				  u8 *queue, u64 *action)
8358 {
8359 	unsigned int num_vfs = adapter->num_vfs, vf;
8360 	struct net_device *upper;
8361 	struct list_head *iter;
8362 
8363 	/* redirect to a SRIOV VF */
8364 	for (vf = 0; vf < num_vfs; ++vf) {
8365 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
8366 		if (upper->ifindex == ifindex) {
8367 			if (adapter->num_rx_pools > 1)
8368 				*queue = vf * 2;
8369 			else
8370 				*queue = vf * adapter->num_rx_queues_per_pool;
8371 
8372 			*action = vf + 1;
8373 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
8374 			return 0;
8375 		}
8376 	}
8377 
8378 	/* redirect to a offloaded macvlan netdev */
8379 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
8380 		if (netif_is_macvlan(upper)) {
8381 			struct macvlan_dev *dfwd = netdev_priv(upper);
8382 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
8383 
8384 			if (vadapter && vadapter->netdev->ifindex == ifindex) {
8385 				*queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
8386 				*action = *queue;
8387 				return 0;
8388 			}
8389 		}
8390 	}
8391 
8392 	return -EINVAL;
8393 }
8394 
8395 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8396 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8397 {
8398 	const struct tc_action *a;
8399 	LIST_HEAD(actions);
8400 	int err;
8401 
8402 	if (tc_no_actions(exts))
8403 		return -EINVAL;
8404 
8405 	tcf_exts_to_list(exts, &actions);
8406 	list_for_each_entry(a, &actions, list) {
8407 
8408 		/* Drop action */
8409 		if (is_tcf_gact_shot(a)) {
8410 			*action = IXGBE_FDIR_DROP_QUEUE;
8411 			*queue = IXGBE_FDIR_DROP_QUEUE;
8412 			return 0;
8413 		}
8414 
8415 		/* Redirect to a VF or a offloaded macvlan */
8416 		if (is_tcf_mirred_redirect(a)) {
8417 			int ifindex = tcf_mirred_ifindex(a);
8418 
8419 			err = handle_redirect_action(adapter, ifindex, queue,
8420 						     action);
8421 			if (err == 0)
8422 				return err;
8423 		}
8424 	}
8425 
8426 	return -EINVAL;
8427 }
8428 #else
8429 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8430 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8431 {
8432 	return -EINVAL;
8433 }
8434 #endif /* CONFIG_NET_CLS_ACT */
8435 
8436 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
8437 				    union ixgbe_atr_input *mask,
8438 				    struct tc_cls_u32_offload *cls,
8439 				    struct ixgbe_mat_field *field_ptr,
8440 				    struct ixgbe_nexthdr *nexthdr)
8441 {
8442 	int i, j, off;
8443 	__be32 val, m;
8444 	bool found_entry = false, found_jump_field = false;
8445 
8446 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
8447 		off = cls->knode.sel->keys[i].off;
8448 		val = cls->knode.sel->keys[i].val;
8449 		m = cls->knode.sel->keys[i].mask;
8450 
8451 		for (j = 0; field_ptr[j].val; j++) {
8452 			if (field_ptr[j].off == off) {
8453 				field_ptr[j].val(input, mask, val, m);
8454 				input->filter.formatted.flow_type |=
8455 					field_ptr[j].type;
8456 				found_entry = true;
8457 				break;
8458 			}
8459 		}
8460 		if (nexthdr) {
8461 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
8462 			    nexthdr->val == cls->knode.sel->keys[i].val &&
8463 			    nexthdr->mask == cls->knode.sel->keys[i].mask)
8464 				found_jump_field = true;
8465 			else
8466 				continue;
8467 		}
8468 	}
8469 
8470 	if (nexthdr && !found_jump_field)
8471 		return -EINVAL;
8472 
8473 	if (!found_entry)
8474 		return 0;
8475 
8476 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
8477 				    IXGBE_ATR_L4TYPE_MASK;
8478 
8479 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
8480 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
8481 
8482 	return 0;
8483 }
8484 
8485 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
8486 				  __be16 protocol,
8487 				  struct tc_cls_u32_offload *cls)
8488 {
8489 	u32 loc = cls->knode.handle & 0xfffff;
8490 	struct ixgbe_hw *hw = &adapter->hw;
8491 	struct ixgbe_mat_field *field_ptr;
8492 	struct ixgbe_fdir_filter *input = NULL;
8493 	union ixgbe_atr_input *mask = NULL;
8494 	struct ixgbe_jump_table *jump = NULL;
8495 	int i, err = -EINVAL;
8496 	u8 queue;
8497 	u32 uhtid, link_uhtid;
8498 
8499 	uhtid = TC_U32_USERHTID(cls->knode.handle);
8500 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
8501 
8502 	/* At the moment cls_u32 jumps to network layer and skips past
8503 	 * L2 headers. The canonical method to match L2 frames is to use
8504 	 * negative values. However this is error prone at best but really
8505 	 * just broken because there is no way to "know" what sort of hdr
8506 	 * is in front of the network layer. Fix cls_u32 to support L2
8507 	 * headers when needed.
8508 	 */
8509 	if (protocol != htons(ETH_P_IP))
8510 		return err;
8511 
8512 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
8513 		e_err(drv, "Location out of range\n");
8514 		return err;
8515 	}
8516 
8517 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
8518 	 * links and also the fields used to advance the parser across each
8519 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
8520 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
8521 	 * To add support for new nodes update ixgbe_model.h parse structures
8522 	 * this function _should_ be generic try not to hardcode values here.
8523 	 */
8524 	if (uhtid == 0x800) {
8525 		field_ptr = (adapter->jump_tables[0])->mat;
8526 	} else {
8527 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8528 			return err;
8529 		if (!adapter->jump_tables[uhtid])
8530 			return err;
8531 		field_ptr = (adapter->jump_tables[uhtid])->mat;
8532 	}
8533 
8534 	if (!field_ptr)
8535 		return err;
8536 
8537 	/* At this point we know the field_ptr is valid and need to either
8538 	 * build cls_u32 link or attach filter. Because adding a link to
8539 	 * a handle that does not exist is invalid and the same for adding
8540 	 * rules to handles that don't exist.
8541 	 */
8542 
8543 	if (link_uhtid) {
8544 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
8545 
8546 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
8547 			return err;
8548 
8549 		if (!test_bit(link_uhtid - 1, &adapter->tables))
8550 			return err;
8551 
8552 		/* Multiple filters as links to the same hash table are not
8553 		 * supported. To add a new filter with the same next header
8554 		 * but different match/jump conditions, create a new hash table
8555 		 * and link to it.
8556 		 */
8557 		if (adapter->jump_tables[link_uhtid] &&
8558 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
8559 			e_err(drv, "Link filter exists for link: %x\n",
8560 			      link_uhtid);
8561 			return err;
8562 		}
8563 
8564 		for (i = 0; nexthdr[i].jump; i++) {
8565 			if (nexthdr[i].o != cls->knode.sel->offoff ||
8566 			    nexthdr[i].s != cls->knode.sel->offshift ||
8567 			    nexthdr[i].m != cls->knode.sel->offmask)
8568 				return err;
8569 
8570 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
8571 			if (!jump)
8572 				return -ENOMEM;
8573 			input = kzalloc(sizeof(*input), GFP_KERNEL);
8574 			if (!input) {
8575 				err = -ENOMEM;
8576 				goto free_jump;
8577 			}
8578 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8579 			if (!mask) {
8580 				err = -ENOMEM;
8581 				goto free_input;
8582 			}
8583 			jump->input = input;
8584 			jump->mask = mask;
8585 			jump->link_hdl = cls->knode.handle;
8586 
8587 			err = ixgbe_clsu32_build_input(input, mask, cls,
8588 						       field_ptr, &nexthdr[i]);
8589 			if (!err) {
8590 				jump->mat = nexthdr[i].jump;
8591 				adapter->jump_tables[link_uhtid] = jump;
8592 				break;
8593 			}
8594 		}
8595 		return 0;
8596 	}
8597 
8598 	input = kzalloc(sizeof(*input), GFP_KERNEL);
8599 	if (!input)
8600 		return -ENOMEM;
8601 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8602 	if (!mask) {
8603 		err = -ENOMEM;
8604 		goto free_input;
8605 	}
8606 
8607 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
8608 		if ((adapter->jump_tables[uhtid])->input)
8609 			memcpy(input, (adapter->jump_tables[uhtid])->input,
8610 			       sizeof(*input));
8611 		if ((adapter->jump_tables[uhtid])->mask)
8612 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
8613 			       sizeof(*mask));
8614 
8615 		/* Lookup in all child hash tables if this location is already
8616 		 * filled with a filter
8617 		 */
8618 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8619 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
8620 
8621 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
8622 				e_err(drv, "Filter exists in location: %x\n",
8623 				      loc);
8624 				err = -EINVAL;
8625 				goto err_out;
8626 			}
8627 		}
8628 	}
8629 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
8630 	if (err)
8631 		goto err_out;
8632 
8633 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
8634 			       &queue);
8635 	if (err < 0)
8636 		goto err_out;
8637 
8638 	input->sw_idx = loc;
8639 
8640 	spin_lock(&adapter->fdir_perfect_lock);
8641 
8642 	if (hlist_empty(&adapter->fdir_filter_list)) {
8643 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
8644 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
8645 		if (err)
8646 			goto err_out_w_lock;
8647 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
8648 		err = -EINVAL;
8649 		goto err_out_w_lock;
8650 	}
8651 
8652 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
8653 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
8654 						    input->sw_idx, queue);
8655 	if (!err)
8656 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
8657 	spin_unlock(&adapter->fdir_perfect_lock);
8658 
8659 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
8660 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
8661 
8662 	kfree(mask);
8663 	return err;
8664 err_out_w_lock:
8665 	spin_unlock(&adapter->fdir_perfect_lock);
8666 err_out:
8667 	kfree(mask);
8668 free_input:
8669 	kfree(input);
8670 free_jump:
8671 	kfree(jump);
8672 	return err;
8673 }
8674 
8675 static int __ixgbe_setup_tc(struct net_device *dev, u32 handle, __be16 proto,
8676 			    struct tc_to_netdev *tc)
8677 {
8678 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8679 
8680 	if (TC_H_MAJ(handle) == TC_H_MAJ(TC_H_INGRESS) &&
8681 	    tc->type == TC_SETUP_CLSU32) {
8682 		switch (tc->cls_u32->command) {
8683 		case TC_CLSU32_NEW_KNODE:
8684 		case TC_CLSU32_REPLACE_KNODE:
8685 			return ixgbe_configure_clsu32(adapter,
8686 						      proto, tc->cls_u32);
8687 		case TC_CLSU32_DELETE_KNODE:
8688 			return ixgbe_delete_clsu32(adapter, tc->cls_u32);
8689 		case TC_CLSU32_NEW_HNODE:
8690 		case TC_CLSU32_REPLACE_HNODE:
8691 			return ixgbe_configure_clsu32_add_hnode(adapter, proto,
8692 								tc->cls_u32);
8693 		case TC_CLSU32_DELETE_HNODE:
8694 			return ixgbe_configure_clsu32_del_hnode(adapter,
8695 								tc->cls_u32);
8696 		default:
8697 			return -EINVAL;
8698 		}
8699 	}
8700 
8701 	if (tc->type != TC_SETUP_MQPRIO)
8702 		return -EINVAL;
8703 
8704 	return ixgbe_setup_tc(dev, tc->tc);
8705 }
8706 
8707 #ifdef CONFIG_PCI_IOV
8708 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
8709 {
8710 	struct net_device *netdev = adapter->netdev;
8711 
8712 	rtnl_lock();
8713 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
8714 	rtnl_unlock();
8715 }
8716 
8717 #endif
8718 void ixgbe_do_reset(struct net_device *netdev)
8719 {
8720 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8721 
8722 	if (netif_running(netdev))
8723 		ixgbe_reinit_locked(adapter);
8724 	else
8725 		ixgbe_reset(adapter);
8726 }
8727 
8728 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
8729 					    netdev_features_t features)
8730 {
8731 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8732 
8733 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
8734 	if (!(features & NETIF_F_RXCSUM))
8735 		features &= ~NETIF_F_LRO;
8736 
8737 	/* Turn off LRO if not RSC capable */
8738 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
8739 		features &= ~NETIF_F_LRO;
8740 
8741 	return features;
8742 }
8743 
8744 static int ixgbe_set_features(struct net_device *netdev,
8745 			      netdev_features_t features)
8746 {
8747 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8748 	netdev_features_t changed = netdev->features ^ features;
8749 	bool need_reset = false;
8750 
8751 	/* Make sure RSC matches LRO, reset if change */
8752 	if (!(features & NETIF_F_LRO)) {
8753 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8754 			need_reset = true;
8755 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
8756 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
8757 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
8758 		if (adapter->rx_itr_setting == 1 ||
8759 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
8760 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
8761 			need_reset = true;
8762 		} else if ((changed ^ features) & NETIF_F_LRO) {
8763 			e_info(probe, "rx-usecs set too low, "
8764 			       "disabling RSC\n");
8765 		}
8766 	}
8767 
8768 	/*
8769 	 * Check if Flow Director n-tuple support or hw_tc support was
8770 	 * enabled or disabled.  If the state changed, we need to reset.
8771 	 */
8772 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
8773 		/* turn off ATR, enable perfect filters and reset */
8774 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
8775 			need_reset = true;
8776 
8777 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
8778 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8779 	} else {
8780 		/* turn off perfect filters, enable ATR and reset */
8781 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
8782 			need_reset = true;
8783 
8784 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8785 
8786 		/* We cannot enable ATR if SR-IOV is enabled */
8787 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
8788 		    /* We cannot enable ATR if we have 2 or more tcs */
8789 		    (netdev_get_num_tc(netdev) > 1) ||
8790 		    /* We cannot enable ATR if RSS is disabled */
8791 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
8792 		    /* A sample rate of 0 indicates ATR disabled */
8793 		    (!adapter->atr_sample_rate))
8794 			; /* do nothing not supported */
8795 		else /* otherwise supported and set the flag */
8796 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
8797 	}
8798 
8799 	if (changed & NETIF_F_RXALL)
8800 		need_reset = true;
8801 
8802 	netdev->features = features;
8803 
8804 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
8805 		if (features & NETIF_F_RXCSUM)
8806 			adapter->flags2 |= IXGBE_FLAG2_VXLAN_REREG_NEEDED;
8807 		else
8808 			ixgbe_clear_vxlan_port(adapter);
8809 	}
8810 
8811 	if (need_reset)
8812 		ixgbe_do_reset(netdev);
8813 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
8814 			    NETIF_F_HW_VLAN_CTAG_FILTER))
8815 		ixgbe_set_rx_mode(netdev);
8816 
8817 	return 0;
8818 }
8819 
8820 /**
8821  * ixgbe_add_vxlan_port - Get notifications about VXLAN ports that come up
8822  * @dev: The port's netdev
8823  * @ti: Tunnel endpoint information
8824  **/
8825 static void ixgbe_add_vxlan_port(struct net_device *dev,
8826 				 struct udp_tunnel_info *ti)
8827 {
8828 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8829 	struct ixgbe_hw *hw = &adapter->hw;
8830 	__be16 port = ti->port;
8831 
8832 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN)
8833 		return;
8834 
8835 	if (ti->sa_family != AF_INET)
8836 		return;
8837 
8838 	if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8839 		return;
8840 
8841 	if (adapter->vxlan_port == port)
8842 		return;
8843 
8844 	if (adapter->vxlan_port) {
8845 		netdev_info(dev,
8846 			    "Hit Max num of VXLAN ports, not adding port %d\n",
8847 			    ntohs(port));
8848 		return;
8849 	}
8850 
8851 	adapter->vxlan_port = port;
8852 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, ntohs(port));
8853 }
8854 
8855 /**
8856  * ixgbe_del_vxlan_port - Get notifications about VXLAN ports that go away
8857  * @dev: The port's netdev
8858  * @ti: Tunnel endpoint information
8859  **/
8860 static void ixgbe_del_vxlan_port(struct net_device *dev,
8861 				 struct udp_tunnel_info *ti)
8862 {
8863 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8864 
8865 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN)
8866 		return;
8867 
8868 	if (ti->sa_family != AF_INET)
8869 		return;
8870 
8871 	if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8872 		return;
8873 
8874 	if (adapter->vxlan_port != ti->port) {
8875 		netdev_info(dev, "Port %d was not found, not deleting\n",
8876 			    ntohs(ti->port));
8877 		return;
8878 	}
8879 
8880 	ixgbe_clear_vxlan_port(adapter);
8881 	adapter->flags2 |= IXGBE_FLAG2_VXLAN_REREG_NEEDED;
8882 }
8883 
8884 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8885 			     struct net_device *dev,
8886 			     const unsigned char *addr, u16 vid,
8887 			     u16 flags)
8888 {
8889 	/* guarantee we can provide a unique filter for the unicast address */
8890 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
8891 		struct ixgbe_adapter *adapter = netdev_priv(dev);
8892 		u16 pool = VMDQ_P(0);
8893 
8894 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
8895 			return -ENOMEM;
8896 	}
8897 
8898 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
8899 }
8900 
8901 /**
8902  * ixgbe_configure_bridge_mode - set various bridge modes
8903  * @adapter - the private structure
8904  * @mode - requested bridge mode
8905  *
8906  * Configure some settings require for various bridge modes.
8907  **/
8908 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
8909 				       __u16 mode)
8910 {
8911 	struct ixgbe_hw *hw = &adapter->hw;
8912 	unsigned int p, num_pools;
8913 	u32 vmdctl;
8914 
8915 	switch (mode) {
8916 	case BRIDGE_MODE_VEPA:
8917 		/* disable Tx loopback, rely on switch hairpin mode */
8918 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
8919 
8920 		/* must enable Rx switching replication to allow multicast
8921 		 * packet reception on all VFs, and to enable source address
8922 		 * pruning.
8923 		 */
8924 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
8925 		vmdctl |= IXGBE_VT_CTL_REPLEN;
8926 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
8927 
8928 		/* enable Rx source address pruning. Note, this requires
8929 		 * replication to be enabled or else it does nothing.
8930 		 */
8931 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
8932 		for (p = 0; p < num_pools; p++) {
8933 			if (hw->mac.ops.set_source_address_pruning)
8934 				hw->mac.ops.set_source_address_pruning(hw,
8935 								       true,
8936 								       p);
8937 		}
8938 		break;
8939 	case BRIDGE_MODE_VEB:
8940 		/* enable Tx loopback for internal VF/PF communication */
8941 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
8942 				IXGBE_PFDTXGSWC_VT_LBEN);
8943 
8944 		/* disable Rx switching replication unless we have SR-IOV
8945 		 * virtual functions
8946 		 */
8947 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
8948 		if (!adapter->num_vfs)
8949 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
8950 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
8951 
8952 		/* disable Rx source address pruning, since we don't expect to
8953 		 * be receiving external loopback of our transmitted frames.
8954 		 */
8955 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
8956 		for (p = 0; p < num_pools; p++) {
8957 			if (hw->mac.ops.set_source_address_pruning)
8958 				hw->mac.ops.set_source_address_pruning(hw,
8959 								       false,
8960 								       p);
8961 		}
8962 		break;
8963 	default:
8964 		return -EINVAL;
8965 	}
8966 
8967 	adapter->bridge_mode = mode;
8968 
8969 	e_info(drv, "enabling bridge mode: %s\n",
8970 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
8971 
8972 	return 0;
8973 }
8974 
8975 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
8976 				    struct nlmsghdr *nlh, u16 flags)
8977 {
8978 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8979 	struct nlattr *attr, *br_spec;
8980 	int rem;
8981 
8982 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
8983 		return -EOPNOTSUPP;
8984 
8985 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8986 	if (!br_spec)
8987 		return -EINVAL;
8988 
8989 	nla_for_each_nested(attr, br_spec, rem) {
8990 		int status;
8991 		__u16 mode;
8992 
8993 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
8994 			continue;
8995 
8996 		if (nla_len(attr) < sizeof(mode))
8997 			return -EINVAL;
8998 
8999 		mode = nla_get_u16(attr);
9000 		status = ixgbe_configure_bridge_mode(adapter, mode);
9001 		if (status)
9002 			return status;
9003 
9004 		break;
9005 	}
9006 
9007 	return 0;
9008 }
9009 
9010 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9011 				    struct net_device *dev,
9012 				    u32 filter_mask, int nlflags)
9013 {
9014 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9015 
9016 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9017 		return 0;
9018 
9019 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9020 				       adapter->bridge_mode, 0, 0, nlflags,
9021 				       filter_mask, NULL);
9022 }
9023 
9024 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9025 {
9026 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
9027 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9028 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
9029 	unsigned int limit;
9030 	int pool, err;
9031 
9032 	/* Hardware has a limited number of available pools. Each VF, and the
9033 	 * PF require a pool. Check to ensure we don't attempt to use more
9034 	 * then the available number of pools.
9035 	 */
9036 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9037 		return ERR_PTR(-EINVAL);
9038 
9039 #ifdef CONFIG_RPS
9040 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
9041 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
9042 			    vdev->name);
9043 		return ERR_PTR(-EINVAL);
9044 	}
9045 #endif
9046 	/* Check for hardware restriction on number of rx/tx queues */
9047 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
9048 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
9049 		netdev_info(pdev,
9050 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
9051 			    pdev->name);
9052 		return ERR_PTR(-EINVAL);
9053 	}
9054 
9055 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9056 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
9057 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
9058 		return ERR_PTR(-EBUSY);
9059 
9060 	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
9061 	if (!fwd_adapter)
9062 		return ERR_PTR(-ENOMEM);
9063 
9064 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
9065 	adapter->num_rx_pools++;
9066 	set_bit(pool, &adapter->fwd_bitmask);
9067 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9068 
9069 	/* Enable VMDq flag so device will be set in VM mode */
9070 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
9071 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9072 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
9073 
9074 	/* Force reinit of ring allocation with VMDQ enabled */
9075 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9076 	if (err)
9077 		goto fwd_add_err;
9078 	fwd_adapter->pool = pool;
9079 	fwd_adapter->real_adapter = adapter;
9080 	err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
9081 	if (err)
9082 		goto fwd_add_err;
9083 	netif_tx_start_all_queues(vdev);
9084 	return fwd_adapter;
9085 fwd_add_err:
9086 	/* unwind counter and free adapter struct */
9087 	netdev_info(pdev,
9088 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
9089 	clear_bit(pool, &adapter->fwd_bitmask);
9090 	adapter->num_rx_pools--;
9091 	kfree(fwd_adapter);
9092 	return ERR_PTR(err);
9093 }
9094 
9095 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9096 {
9097 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
9098 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
9099 	unsigned int limit;
9100 
9101 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
9102 	adapter->num_rx_pools--;
9103 
9104 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9105 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9106 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
9107 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9108 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
9109 		   fwd_adapter->pool, adapter->num_rx_pools,
9110 		   fwd_adapter->rx_base_queue,
9111 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
9112 		   adapter->fwd_bitmask);
9113 	kfree(fwd_adapter);
9114 }
9115 
9116 #define IXGBE_MAX_MAC_HDR_LEN		127
9117 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9118 
9119 static netdev_features_t
9120 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9121 		     netdev_features_t features)
9122 {
9123 	unsigned int network_hdr_len, mac_hdr_len;
9124 
9125 	/* Make certain the headers can be described by a context descriptor */
9126 	mac_hdr_len = skb_network_header(skb) - skb->data;
9127 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9128 		return features & ~(NETIF_F_HW_CSUM |
9129 				    NETIF_F_SCTP_CRC |
9130 				    NETIF_F_HW_VLAN_CTAG_TX |
9131 				    NETIF_F_TSO |
9132 				    NETIF_F_TSO6);
9133 
9134 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9135 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9136 		return features & ~(NETIF_F_HW_CSUM |
9137 				    NETIF_F_SCTP_CRC |
9138 				    NETIF_F_TSO |
9139 				    NETIF_F_TSO6);
9140 
9141 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9142 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9143 	 */
9144 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
9145 		features &= ~NETIF_F_TSO;
9146 
9147 	return features;
9148 }
9149 
9150 static const struct net_device_ops ixgbe_netdev_ops = {
9151 	.ndo_open		= ixgbe_open,
9152 	.ndo_stop		= ixgbe_close,
9153 	.ndo_start_xmit		= ixgbe_xmit_frame,
9154 	.ndo_select_queue	= ixgbe_select_queue,
9155 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
9156 	.ndo_validate_addr	= eth_validate_addr,
9157 	.ndo_set_mac_address	= ixgbe_set_mac,
9158 	.ndo_change_mtu		= ixgbe_change_mtu,
9159 	.ndo_tx_timeout		= ixgbe_tx_timeout,
9160 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
9161 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
9162 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
9163 	.ndo_do_ioctl		= ixgbe_ioctl,
9164 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
9165 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
9166 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
9167 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
9168 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
9169 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
9170 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
9171 	.ndo_get_stats64	= ixgbe_get_stats64,
9172 	.ndo_setup_tc		= __ixgbe_setup_tc,
9173 #ifdef CONFIG_NET_POLL_CONTROLLER
9174 	.ndo_poll_controller	= ixgbe_netpoll,
9175 #endif
9176 #ifdef CONFIG_NET_RX_BUSY_POLL
9177 	.ndo_busy_poll		= ixgbe_low_latency_recv,
9178 #endif
9179 #ifdef IXGBE_FCOE
9180 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
9181 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
9182 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
9183 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
9184 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
9185 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
9186 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
9187 #endif /* IXGBE_FCOE */
9188 	.ndo_set_features = ixgbe_set_features,
9189 	.ndo_fix_features = ixgbe_fix_features,
9190 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
9191 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
9192 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
9193 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
9194 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
9195 	.ndo_udp_tunnel_add	= ixgbe_add_vxlan_port,
9196 	.ndo_udp_tunnel_del	= ixgbe_del_vxlan_port,
9197 	.ndo_features_check	= ixgbe_features_check,
9198 };
9199 
9200 /**
9201  * ixgbe_enumerate_functions - Get the number of ports this device has
9202  * @adapter: adapter structure
9203  *
9204  * This function enumerates the phsyical functions co-located on a single slot,
9205  * in order to determine how many ports a device has. This is most useful in
9206  * determining the required GT/s of PCIe bandwidth necessary for optimal
9207  * performance.
9208  **/
9209 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
9210 {
9211 	struct pci_dev *entry, *pdev = adapter->pdev;
9212 	int physfns = 0;
9213 
9214 	/* Some cards can not use the generic count PCIe functions method,
9215 	 * because they are behind a parent switch, so we hardcode these with
9216 	 * the correct number of functions.
9217 	 */
9218 	if (ixgbe_pcie_from_parent(&adapter->hw))
9219 		physfns = 4;
9220 
9221 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
9222 		/* don't count virtual functions */
9223 		if (entry->is_virtfn)
9224 			continue;
9225 
9226 		/* When the devices on the bus don't all match our device ID,
9227 		 * we can't reliably determine the correct number of
9228 		 * functions. This can occur if a function has been direct
9229 		 * attached to a virtual machine using VT-d, for example. In
9230 		 * this case, simply return -1 to indicate this.
9231 		 */
9232 		if ((entry->vendor != pdev->vendor) ||
9233 		    (entry->device != pdev->device))
9234 			return -1;
9235 
9236 		physfns++;
9237 	}
9238 
9239 	return physfns;
9240 }
9241 
9242 /**
9243  * ixgbe_wol_supported - Check whether device supports WoL
9244  * @adapter: the adapter private structure
9245  * @device_id: the device ID
9246  * @subdev_id: the subsystem device ID
9247  *
9248  * This function is used by probe and ethtool to determine
9249  * which devices have WoL support
9250  *
9251  **/
9252 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
9253 			 u16 subdevice_id)
9254 {
9255 	struct ixgbe_hw *hw = &adapter->hw;
9256 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
9257 
9258 	/* WOL not supported on 82598 */
9259 	if (hw->mac.type == ixgbe_mac_82598EB)
9260 		return false;
9261 
9262 	/* check eeprom to see if WOL is enabled for X540 and newer */
9263 	if (hw->mac.type >= ixgbe_mac_X540) {
9264 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
9265 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
9266 		     (hw->bus.func == 0)))
9267 			return true;
9268 	}
9269 
9270 	/* WOL is determined based on device IDs for 82599 MACs */
9271 	switch (device_id) {
9272 	case IXGBE_DEV_ID_82599_SFP:
9273 		/* Only these subdevices could supports WOL */
9274 		switch (subdevice_id) {
9275 		case IXGBE_SUBDEV_ID_82599_560FLR:
9276 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
9277 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
9278 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
9279 			/* only support first port */
9280 			if (hw->bus.func != 0)
9281 				break;
9282 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
9283 		case IXGBE_SUBDEV_ID_82599_SFP:
9284 		case IXGBE_SUBDEV_ID_82599_RNDC:
9285 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
9286 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
9287 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
9288 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
9289 			return true;
9290 		}
9291 		break;
9292 	case IXGBE_DEV_ID_82599EN_SFP:
9293 		/* Only these subdevices support WOL */
9294 		switch (subdevice_id) {
9295 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
9296 			return true;
9297 		}
9298 		break;
9299 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
9300 		/* All except this subdevice support WOL */
9301 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
9302 			return true;
9303 		break;
9304 	case IXGBE_DEV_ID_82599_KX4:
9305 		return  true;
9306 	default:
9307 		break;
9308 	}
9309 
9310 	return false;
9311 }
9312 
9313 /**
9314  * ixgbe_probe - Device Initialization Routine
9315  * @pdev: PCI device information struct
9316  * @ent: entry in ixgbe_pci_tbl
9317  *
9318  * Returns 0 on success, negative on failure
9319  *
9320  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
9321  * The OS initialization, configuring of the adapter private structure,
9322  * and a hardware reset occur.
9323  **/
9324 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
9325 {
9326 	struct net_device *netdev;
9327 	struct ixgbe_adapter *adapter = NULL;
9328 	struct ixgbe_hw *hw;
9329 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
9330 	int i, err, pci_using_dac, expected_gts;
9331 	unsigned int indices = MAX_TX_QUEUES;
9332 	u8 part_str[IXGBE_PBANUM_LENGTH];
9333 	bool disable_dev = false;
9334 #ifdef IXGBE_FCOE
9335 	u16 device_caps;
9336 #endif
9337 	u32 eec;
9338 
9339 	/* Catch broken hardware that put the wrong VF device ID in
9340 	 * the PCIe SR-IOV capability.
9341 	 */
9342 	if (pdev->is_virtfn) {
9343 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
9344 		     pci_name(pdev), pdev->vendor, pdev->device);
9345 		return -EINVAL;
9346 	}
9347 
9348 	err = pci_enable_device_mem(pdev);
9349 	if (err)
9350 		return err;
9351 
9352 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
9353 		pci_using_dac = 1;
9354 	} else {
9355 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
9356 		if (err) {
9357 			dev_err(&pdev->dev,
9358 				"No usable DMA configuration, aborting\n");
9359 			goto err_dma;
9360 		}
9361 		pci_using_dac = 0;
9362 	}
9363 
9364 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
9365 	if (err) {
9366 		dev_err(&pdev->dev,
9367 			"pci_request_selected_regions failed 0x%x\n", err);
9368 		goto err_pci_reg;
9369 	}
9370 
9371 	pci_enable_pcie_error_reporting(pdev);
9372 
9373 	pci_set_master(pdev);
9374 	pci_save_state(pdev);
9375 
9376 	if (ii->mac == ixgbe_mac_82598EB) {
9377 #ifdef CONFIG_IXGBE_DCB
9378 		/* 8 TC w/ 4 queues per TC */
9379 		indices = 4 * MAX_TRAFFIC_CLASS;
9380 #else
9381 		indices = IXGBE_MAX_RSS_INDICES;
9382 #endif
9383 	}
9384 
9385 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
9386 	if (!netdev) {
9387 		err = -ENOMEM;
9388 		goto err_alloc_etherdev;
9389 	}
9390 
9391 	SET_NETDEV_DEV(netdev, &pdev->dev);
9392 
9393 	adapter = netdev_priv(netdev);
9394 
9395 	adapter->netdev = netdev;
9396 	adapter->pdev = pdev;
9397 	hw = &adapter->hw;
9398 	hw->back = adapter;
9399 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
9400 
9401 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
9402 			      pci_resource_len(pdev, 0));
9403 	adapter->io_addr = hw->hw_addr;
9404 	if (!hw->hw_addr) {
9405 		err = -EIO;
9406 		goto err_ioremap;
9407 	}
9408 
9409 	netdev->netdev_ops = &ixgbe_netdev_ops;
9410 	ixgbe_set_ethtool_ops(netdev);
9411 	netdev->watchdog_timeo = 5 * HZ;
9412 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
9413 
9414 	/* Setup hw api */
9415 	hw->mac.ops   = *ii->mac_ops;
9416 	hw->mac.type  = ii->mac;
9417 	hw->mvals     = ii->mvals;
9418 
9419 	/* EEPROM */
9420 	hw->eeprom.ops = *ii->eeprom_ops;
9421 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
9422 	if (ixgbe_removed(hw->hw_addr)) {
9423 		err = -EIO;
9424 		goto err_ioremap;
9425 	}
9426 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
9427 	if (!(eec & BIT(8)))
9428 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
9429 
9430 	/* PHY */
9431 	hw->phy.ops = *ii->phy_ops;
9432 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
9433 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
9434 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
9435 	hw->phy.mdio.mmds = 0;
9436 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
9437 	hw->phy.mdio.dev = netdev;
9438 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
9439 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
9440 
9441 	ii->get_invariants(hw);
9442 
9443 	/* setup the private structure */
9444 	err = ixgbe_sw_init(adapter);
9445 	if (err)
9446 		goto err_sw_init;
9447 
9448 	/* Make sure the SWFW semaphore is in a valid state */
9449 	if (hw->mac.ops.init_swfw_sync)
9450 		hw->mac.ops.init_swfw_sync(hw);
9451 
9452 	/* Make it possible the adapter to be woken up via WOL */
9453 	switch (adapter->hw.mac.type) {
9454 	case ixgbe_mac_82599EB:
9455 	case ixgbe_mac_X540:
9456 	case ixgbe_mac_X550:
9457 	case ixgbe_mac_X550EM_x:
9458 	case ixgbe_mac_x550em_a:
9459 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
9460 		break;
9461 	default:
9462 		break;
9463 	}
9464 
9465 	/*
9466 	 * If there is a fan on this device and it has failed log the
9467 	 * failure.
9468 	 */
9469 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
9470 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
9471 		if (esdp & IXGBE_ESDP_SDP1)
9472 			e_crit(probe, "Fan has stopped, replace the adapter\n");
9473 	}
9474 
9475 	if (allow_unsupported_sfp)
9476 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
9477 
9478 	/* reset_hw fills in the perm_addr as well */
9479 	hw->phy.reset_if_overtemp = true;
9480 	err = hw->mac.ops.reset_hw(hw);
9481 	hw->phy.reset_if_overtemp = false;
9482 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
9483 		err = 0;
9484 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
9485 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
9486 		e_dev_err("Reload the driver after installing a supported module.\n");
9487 		goto err_sw_init;
9488 	} else if (err) {
9489 		e_dev_err("HW Init failed: %d\n", err);
9490 		goto err_sw_init;
9491 	}
9492 
9493 #ifdef CONFIG_PCI_IOV
9494 	/* SR-IOV not supported on the 82598 */
9495 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9496 		goto skip_sriov;
9497 	/* Mailbox */
9498 	ixgbe_init_mbx_params_pf(hw);
9499 	hw->mbx.ops = ii->mbx_ops;
9500 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
9501 	ixgbe_enable_sriov(adapter);
9502 skip_sriov:
9503 
9504 #endif
9505 	netdev->features = NETIF_F_SG |
9506 			   NETIF_F_TSO |
9507 			   NETIF_F_TSO6 |
9508 			   NETIF_F_RXHASH |
9509 			   NETIF_F_RXCSUM |
9510 			   NETIF_F_HW_CSUM;
9511 
9512 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
9513 				    NETIF_F_GSO_GRE_CSUM | \
9514 				    NETIF_F_GSO_IPXIP4 | \
9515 				    NETIF_F_GSO_IPXIP6 | \
9516 				    NETIF_F_GSO_UDP_TUNNEL | \
9517 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
9518 
9519 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
9520 	netdev->features |= NETIF_F_GSO_PARTIAL |
9521 			    IXGBE_GSO_PARTIAL_FEATURES;
9522 
9523 	if (hw->mac.type >= ixgbe_mac_82599EB)
9524 		netdev->features |= NETIF_F_SCTP_CRC;
9525 
9526 	/* copy netdev features into list of user selectable features */
9527 	netdev->hw_features |= netdev->features |
9528 			       NETIF_F_HW_VLAN_CTAG_FILTER |
9529 			       NETIF_F_HW_VLAN_CTAG_RX |
9530 			       NETIF_F_HW_VLAN_CTAG_TX |
9531 			       NETIF_F_RXALL |
9532 			       NETIF_F_HW_L2FW_DOFFLOAD;
9533 
9534 	if (hw->mac.type >= ixgbe_mac_82599EB)
9535 		netdev->hw_features |= NETIF_F_NTUPLE |
9536 				       NETIF_F_HW_TC;
9537 
9538 	if (pci_using_dac)
9539 		netdev->features |= NETIF_F_HIGHDMA;
9540 
9541 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
9542 	netdev->hw_enc_features |= netdev->vlan_features;
9543 	netdev->mpls_features |= NETIF_F_HW_CSUM;
9544 
9545 	/* set this bit last since it cannot be part of vlan_features */
9546 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
9547 			    NETIF_F_HW_VLAN_CTAG_RX |
9548 			    NETIF_F_HW_VLAN_CTAG_TX;
9549 
9550 	netdev->priv_flags |= IFF_UNICAST_FLT;
9551 	netdev->priv_flags |= IFF_SUPP_NOFCS;
9552 
9553 #ifdef CONFIG_IXGBE_DCB
9554 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
9555 		netdev->dcbnl_ops = &dcbnl_ops;
9556 #endif
9557 
9558 #ifdef IXGBE_FCOE
9559 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
9560 		unsigned int fcoe_l;
9561 
9562 		if (hw->mac.ops.get_device_caps) {
9563 			hw->mac.ops.get_device_caps(hw, &device_caps);
9564 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
9565 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
9566 		}
9567 
9568 
9569 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
9570 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
9571 
9572 		netdev->features |= NETIF_F_FSO |
9573 				    NETIF_F_FCOE_CRC;
9574 
9575 		netdev->vlan_features |= NETIF_F_FSO |
9576 					 NETIF_F_FCOE_CRC |
9577 					 NETIF_F_FCOE_MTU;
9578 	}
9579 #endif /* IXGBE_FCOE */
9580 
9581 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
9582 		netdev->hw_features |= NETIF_F_LRO;
9583 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9584 		netdev->features |= NETIF_F_LRO;
9585 
9586 	/* make sure the EEPROM is good */
9587 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
9588 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
9589 		err = -EIO;
9590 		goto err_sw_init;
9591 	}
9592 
9593 	eth_platform_get_mac_address(&adapter->pdev->dev,
9594 				     adapter->hw.mac.perm_addr);
9595 
9596 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
9597 
9598 	if (!is_valid_ether_addr(netdev->dev_addr)) {
9599 		e_dev_err("invalid MAC address\n");
9600 		err = -EIO;
9601 		goto err_sw_init;
9602 	}
9603 
9604 	/* Set hw->mac.addr to permanent MAC address */
9605 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
9606 	ixgbe_mac_set_default_filter(adapter);
9607 
9608 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
9609 		    (unsigned long) adapter);
9610 
9611 	if (ixgbe_removed(hw->hw_addr)) {
9612 		err = -EIO;
9613 		goto err_sw_init;
9614 	}
9615 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
9616 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
9617 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
9618 
9619 	err = ixgbe_init_interrupt_scheme(adapter);
9620 	if (err)
9621 		goto err_sw_init;
9622 
9623 	/* WOL not supported for all devices */
9624 	adapter->wol = 0;
9625 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
9626 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
9627 						pdev->subsystem_device);
9628 	if (hw->wol_enabled)
9629 		adapter->wol = IXGBE_WUFC_MAG;
9630 
9631 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
9632 
9633 	/* save off EEPROM version number */
9634 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
9635 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
9636 
9637 	/* pick up the PCI bus settings for reporting later */
9638 	if (ixgbe_pcie_from_parent(hw))
9639 		ixgbe_get_parent_bus_info(adapter);
9640 	else
9641 		 hw->mac.ops.get_bus_info(hw);
9642 
9643 	/* calculate the expected PCIe bandwidth required for optimal
9644 	 * performance. Note that some older parts will never have enough
9645 	 * bandwidth due to being older generation PCIe parts. We clamp these
9646 	 * parts to ensure no warning is displayed if it can't be fixed.
9647 	 */
9648 	switch (hw->mac.type) {
9649 	case ixgbe_mac_82598EB:
9650 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
9651 		break;
9652 	default:
9653 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
9654 		break;
9655 	}
9656 
9657 	/* don't check link if we failed to enumerate functions */
9658 	if (expected_gts > 0)
9659 		ixgbe_check_minimum_link(adapter, expected_gts);
9660 
9661 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
9662 	if (err)
9663 		strlcpy(part_str, "Unknown", sizeof(part_str));
9664 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
9665 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
9666 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
9667 			   part_str);
9668 	else
9669 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
9670 			   hw->mac.type, hw->phy.type, part_str);
9671 
9672 	e_dev_info("%pM\n", netdev->dev_addr);
9673 
9674 	/* reset the hardware with the new settings */
9675 	err = hw->mac.ops.start_hw(hw);
9676 	if (err == IXGBE_ERR_EEPROM_VERSION) {
9677 		/* We are running on a pre-production device, log a warning */
9678 		e_dev_warn("This device is a pre-production adapter/LOM. "
9679 			   "Please be aware there may be issues associated "
9680 			   "with your hardware.  If you are experiencing "
9681 			   "problems please contact your Intel or hardware "
9682 			   "representative who provided you with this "
9683 			   "hardware.\n");
9684 	}
9685 	strcpy(netdev->name, "eth%d");
9686 	err = register_netdev(netdev);
9687 	if (err)
9688 		goto err_register;
9689 
9690 	pci_set_drvdata(pdev, adapter);
9691 
9692 	/* power down the optics for 82599 SFP+ fiber */
9693 	if (hw->mac.ops.disable_tx_laser)
9694 		hw->mac.ops.disable_tx_laser(hw);
9695 
9696 	/* carrier off reporting is important to ethtool even BEFORE open */
9697 	netif_carrier_off(netdev);
9698 
9699 #ifdef CONFIG_IXGBE_DCA
9700 	if (dca_add_requester(&pdev->dev) == 0) {
9701 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
9702 		ixgbe_setup_dca(adapter);
9703 	}
9704 #endif
9705 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
9706 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
9707 		for (i = 0; i < adapter->num_vfs; i++)
9708 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
9709 	}
9710 
9711 	/* firmware requires driver version to be 0xFFFFFFFF
9712 	 * since os does not support feature
9713 	 */
9714 	if (hw->mac.ops.set_fw_drv_ver)
9715 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
9716 					   0xFF);
9717 
9718 	/* add san mac addr to netdev */
9719 	ixgbe_add_sanmac_netdev(netdev);
9720 
9721 	e_dev_info("%s\n", ixgbe_default_device_descr);
9722 
9723 #ifdef CONFIG_IXGBE_HWMON
9724 	if (ixgbe_sysfs_init(adapter))
9725 		e_err(probe, "failed to allocate sysfs resources\n");
9726 #endif /* CONFIG_IXGBE_HWMON */
9727 
9728 	ixgbe_dbg_adapter_init(adapter);
9729 
9730 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
9731 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
9732 		hw->mac.ops.setup_link(hw,
9733 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
9734 			true);
9735 
9736 	return 0;
9737 
9738 err_register:
9739 	ixgbe_release_hw_control(adapter);
9740 	ixgbe_clear_interrupt_scheme(adapter);
9741 err_sw_init:
9742 	ixgbe_disable_sriov(adapter);
9743 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
9744 	iounmap(adapter->io_addr);
9745 	kfree(adapter->jump_tables[0]);
9746 	kfree(adapter->mac_table);
9747 err_ioremap:
9748 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9749 	free_netdev(netdev);
9750 err_alloc_etherdev:
9751 	pci_release_mem_regions(pdev);
9752 err_pci_reg:
9753 err_dma:
9754 	if (!adapter || disable_dev)
9755 		pci_disable_device(pdev);
9756 	return err;
9757 }
9758 
9759 /**
9760  * ixgbe_remove - Device Removal Routine
9761  * @pdev: PCI device information struct
9762  *
9763  * ixgbe_remove is called by the PCI subsystem to alert the driver
9764  * that it should release a PCI device.  The could be caused by a
9765  * Hot-Plug event, or because the driver is going to be removed from
9766  * memory.
9767  **/
9768 static void ixgbe_remove(struct pci_dev *pdev)
9769 {
9770 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9771 	struct net_device *netdev;
9772 	bool disable_dev;
9773 	int i;
9774 
9775 	/* if !adapter then we already cleaned up in probe */
9776 	if (!adapter)
9777 		return;
9778 
9779 	netdev  = adapter->netdev;
9780 	ixgbe_dbg_adapter_exit(adapter);
9781 
9782 	set_bit(__IXGBE_REMOVING, &adapter->state);
9783 	cancel_work_sync(&adapter->service_task);
9784 
9785 
9786 #ifdef CONFIG_IXGBE_DCA
9787 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
9788 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
9789 		dca_remove_requester(&pdev->dev);
9790 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
9791 				IXGBE_DCA_CTRL_DCA_DISABLE);
9792 	}
9793 
9794 #endif
9795 #ifdef CONFIG_IXGBE_HWMON
9796 	ixgbe_sysfs_exit(adapter);
9797 #endif /* CONFIG_IXGBE_HWMON */
9798 
9799 	/* remove the added san mac */
9800 	ixgbe_del_sanmac_netdev(netdev);
9801 
9802 #ifdef CONFIG_PCI_IOV
9803 	ixgbe_disable_sriov(adapter);
9804 #endif
9805 	if (netdev->reg_state == NETREG_REGISTERED)
9806 		unregister_netdev(netdev);
9807 
9808 	ixgbe_clear_interrupt_scheme(adapter);
9809 
9810 	ixgbe_release_hw_control(adapter);
9811 
9812 #ifdef CONFIG_DCB
9813 	kfree(adapter->ixgbe_ieee_pfc);
9814 	kfree(adapter->ixgbe_ieee_ets);
9815 
9816 #endif
9817 	iounmap(adapter->io_addr);
9818 	pci_release_mem_regions(pdev);
9819 
9820 	e_dev_info("complete\n");
9821 
9822 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
9823 		if (adapter->jump_tables[i]) {
9824 			kfree(adapter->jump_tables[i]->input);
9825 			kfree(adapter->jump_tables[i]->mask);
9826 		}
9827 		kfree(adapter->jump_tables[i]);
9828 	}
9829 
9830 	kfree(adapter->mac_table);
9831 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9832 	free_netdev(netdev);
9833 
9834 	pci_disable_pcie_error_reporting(pdev);
9835 
9836 	if (disable_dev)
9837 		pci_disable_device(pdev);
9838 }
9839 
9840 /**
9841  * ixgbe_io_error_detected - called when PCI error is detected
9842  * @pdev: Pointer to PCI device
9843  * @state: The current pci connection state
9844  *
9845  * This function is called after a PCI bus error affecting
9846  * this device has been detected.
9847  */
9848 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
9849 						pci_channel_state_t state)
9850 {
9851 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9852 	struct net_device *netdev = adapter->netdev;
9853 
9854 #ifdef CONFIG_PCI_IOV
9855 	struct ixgbe_hw *hw = &adapter->hw;
9856 	struct pci_dev *bdev, *vfdev;
9857 	u32 dw0, dw1, dw2, dw3;
9858 	int vf, pos;
9859 	u16 req_id, pf_func;
9860 
9861 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
9862 	    adapter->num_vfs == 0)
9863 		goto skip_bad_vf_detection;
9864 
9865 	bdev = pdev->bus->self;
9866 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
9867 		bdev = bdev->bus->self;
9868 
9869 	if (!bdev)
9870 		goto skip_bad_vf_detection;
9871 
9872 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
9873 	if (!pos)
9874 		goto skip_bad_vf_detection;
9875 
9876 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
9877 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
9878 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
9879 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
9880 	if (ixgbe_removed(hw->hw_addr))
9881 		goto skip_bad_vf_detection;
9882 
9883 	req_id = dw1 >> 16;
9884 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
9885 	if (!(req_id & 0x0080))
9886 		goto skip_bad_vf_detection;
9887 
9888 	pf_func = req_id & 0x01;
9889 	if ((pf_func & 1) == (pdev->devfn & 1)) {
9890 		unsigned int device_id;
9891 
9892 		vf = (req_id & 0x7F) >> 1;
9893 		e_dev_err("VF %d has caused a PCIe error\n", vf);
9894 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
9895 				"%8.8x\tdw3: %8.8x\n",
9896 		dw0, dw1, dw2, dw3);
9897 		switch (adapter->hw.mac.type) {
9898 		case ixgbe_mac_82599EB:
9899 			device_id = IXGBE_82599_VF_DEVICE_ID;
9900 			break;
9901 		case ixgbe_mac_X540:
9902 			device_id = IXGBE_X540_VF_DEVICE_ID;
9903 			break;
9904 		case ixgbe_mac_X550:
9905 			device_id = IXGBE_DEV_ID_X550_VF;
9906 			break;
9907 		case ixgbe_mac_X550EM_x:
9908 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
9909 			break;
9910 		case ixgbe_mac_x550em_a:
9911 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
9912 			break;
9913 		default:
9914 			device_id = 0;
9915 			break;
9916 		}
9917 
9918 		/* Find the pci device of the offending VF */
9919 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
9920 		while (vfdev) {
9921 			if (vfdev->devfn == (req_id & 0xFF))
9922 				break;
9923 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
9924 					       device_id, vfdev);
9925 		}
9926 		/*
9927 		 * There's a slim chance the VF could have been hot plugged,
9928 		 * so if it is no longer present we don't need to issue the
9929 		 * VFLR.  Just clean up the AER in that case.
9930 		 */
9931 		if (vfdev) {
9932 			ixgbe_issue_vf_flr(adapter, vfdev);
9933 			/* Free device reference count */
9934 			pci_dev_put(vfdev);
9935 		}
9936 
9937 		pci_cleanup_aer_uncorrect_error_status(pdev);
9938 	}
9939 
9940 	/*
9941 	 * Even though the error may have occurred on the other port
9942 	 * we still need to increment the vf error reference count for
9943 	 * both ports because the I/O resume function will be called
9944 	 * for both of them.
9945 	 */
9946 	adapter->vferr_refcount++;
9947 
9948 	return PCI_ERS_RESULT_RECOVERED;
9949 
9950 skip_bad_vf_detection:
9951 #endif /* CONFIG_PCI_IOV */
9952 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
9953 		return PCI_ERS_RESULT_DISCONNECT;
9954 
9955 	rtnl_lock();
9956 	netif_device_detach(netdev);
9957 
9958 	if (state == pci_channel_io_perm_failure) {
9959 		rtnl_unlock();
9960 		return PCI_ERS_RESULT_DISCONNECT;
9961 	}
9962 
9963 	if (netif_running(netdev))
9964 		ixgbe_down(adapter);
9965 
9966 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
9967 		pci_disable_device(pdev);
9968 	rtnl_unlock();
9969 
9970 	/* Request a slot reset. */
9971 	return PCI_ERS_RESULT_NEED_RESET;
9972 }
9973 
9974 /**
9975  * ixgbe_io_slot_reset - called after the pci bus has been reset.
9976  * @pdev: Pointer to PCI device
9977  *
9978  * Restart the card from scratch, as if from a cold-boot.
9979  */
9980 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
9981 {
9982 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9983 	pci_ers_result_t result;
9984 	int err;
9985 
9986 	if (pci_enable_device_mem(pdev)) {
9987 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
9988 		result = PCI_ERS_RESULT_DISCONNECT;
9989 	} else {
9990 		smp_mb__before_atomic();
9991 		clear_bit(__IXGBE_DISABLED, &adapter->state);
9992 		adapter->hw.hw_addr = adapter->io_addr;
9993 		pci_set_master(pdev);
9994 		pci_restore_state(pdev);
9995 		pci_save_state(pdev);
9996 
9997 		pci_wake_from_d3(pdev, false);
9998 
9999 		ixgbe_reset(adapter);
10000 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10001 		result = PCI_ERS_RESULT_RECOVERED;
10002 	}
10003 
10004 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10005 	if (err) {
10006 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
10007 			  "failed 0x%0x\n", err);
10008 		/* non-fatal, continue */
10009 	}
10010 
10011 	return result;
10012 }
10013 
10014 /**
10015  * ixgbe_io_resume - called when traffic can start flowing again.
10016  * @pdev: Pointer to PCI device
10017  *
10018  * This callback is called when the error recovery driver tells us that
10019  * its OK to resume normal operation.
10020  */
10021 static void ixgbe_io_resume(struct pci_dev *pdev)
10022 {
10023 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10024 	struct net_device *netdev = adapter->netdev;
10025 
10026 #ifdef CONFIG_PCI_IOV
10027 	if (adapter->vferr_refcount) {
10028 		e_info(drv, "Resuming after VF err\n");
10029 		adapter->vferr_refcount--;
10030 		return;
10031 	}
10032 
10033 #endif
10034 	if (netif_running(netdev))
10035 		ixgbe_up(adapter);
10036 
10037 	netif_device_attach(netdev);
10038 }
10039 
10040 static const struct pci_error_handlers ixgbe_err_handler = {
10041 	.error_detected = ixgbe_io_error_detected,
10042 	.slot_reset = ixgbe_io_slot_reset,
10043 	.resume = ixgbe_io_resume,
10044 };
10045 
10046 static struct pci_driver ixgbe_driver = {
10047 	.name     = ixgbe_driver_name,
10048 	.id_table = ixgbe_pci_tbl,
10049 	.probe    = ixgbe_probe,
10050 	.remove   = ixgbe_remove,
10051 #ifdef CONFIG_PM
10052 	.suspend  = ixgbe_suspend,
10053 	.resume   = ixgbe_resume,
10054 #endif
10055 	.shutdown = ixgbe_shutdown,
10056 	.sriov_configure = ixgbe_pci_sriov_configure,
10057 	.err_handler = &ixgbe_err_handler
10058 };
10059 
10060 /**
10061  * ixgbe_init_module - Driver Registration Routine
10062  *
10063  * ixgbe_init_module is the first routine called when the driver is
10064  * loaded. All it does is register with the PCI subsystem.
10065  **/
10066 static int __init ixgbe_init_module(void)
10067 {
10068 	int ret;
10069 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
10070 	pr_info("%s\n", ixgbe_copyright);
10071 
10072 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
10073 	if (!ixgbe_wq) {
10074 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
10075 		return -ENOMEM;
10076 	}
10077 
10078 	ixgbe_dbg_init();
10079 
10080 	ret = pci_register_driver(&ixgbe_driver);
10081 	if (ret) {
10082 		destroy_workqueue(ixgbe_wq);
10083 		ixgbe_dbg_exit();
10084 		return ret;
10085 	}
10086 
10087 #ifdef CONFIG_IXGBE_DCA
10088 	dca_register_notify(&dca_notifier);
10089 #endif
10090 
10091 	return 0;
10092 }
10093 
10094 module_init(ixgbe_init_module);
10095 
10096 /**
10097  * ixgbe_exit_module - Driver Exit Cleanup Routine
10098  *
10099  * ixgbe_exit_module is called just before the driver is removed
10100  * from memory.
10101  **/
10102 static void __exit ixgbe_exit_module(void)
10103 {
10104 #ifdef CONFIG_IXGBE_DCA
10105 	dca_unregister_notify(&dca_notifier);
10106 #endif
10107 	pci_unregister_driver(&ixgbe_driver);
10108 
10109 	ixgbe_dbg_exit();
10110 	if (ixgbe_wq) {
10111 		destroy_workqueue(ixgbe_wq);
10112 		ixgbe_wq = NULL;
10113 	}
10114 }
10115 
10116 #ifdef CONFIG_IXGBE_DCA
10117 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
10118 			    void *p)
10119 {
10120 	int ret_val;
10121 
10122 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
10123 					 __ixgbe_notify_dca);
10124 
10125 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
10126 }
10127 
10128 #endif /* CONFIG_IXGBE_DCA */
10129 
10130 module_exit(ixgbe_exit_module);
10131 
10132 /* ixgbe_main.c */
10133