xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 9ee0034b8f49aaaa7e7c2da8db1038915db99c19)
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_10G_T), board_x550em_a},
141 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
142 	/* required last entry */
143 	{0, }
144 };
145 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
146 
147 #ifdef CONFIG_IXGBE_DCA
148 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
149 			    void *p);
150 static struct notifier_block dca_notifier = {
151 	.notifier_call = ixgbe_notify_dca,
152 	.next          = NULL,
153 	.priority      = 0
154 };
155 #endif
156 
157 #ifdef CONFIG_PCI_IOV
158 static unsigned int max_vfs;
159 module_param(max_vfs, uint, 0);
160 MODULE_PARM_DESC(max_vfs,
161 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
162 #endif /* CONFIG_PCI_IOV */
163 
164 static unsigned int allow_unsupported_sfp;
165 module_param(allow_unsupported_sfp, uint, 0);
166 MODULE_PARM_DESC(allow_unsupported_sfp,
167 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
168 
169 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
170 static int debug = -1;
171 module_param(debug, int, 0);
172 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
173 
174 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
175 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
176 MODULE_LICENSE("GPL");
177 MODULE_VERSION(DRV_VERSION);
178 
179 static struct workqueue_struct *ixgbe_wq;
180 
181 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
182 
183 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
184 					  u32 reg, u16 *value)
185 {
186 	struct pci_dev *parent_dev;
187 	struct pci_bus *parent_bus;
188 
189 	parent_bus = adapter->pdev->bus->parent;
190 	if (!parent_bus)
191 		return -1;
192 
193 	parent_dev = parent_bus->self;
194 	if (!parent_dev)
195 		return -1;
196 
197 	if (!pci_is_pcie(parent_dev))
198 		return -1;
199 
200 	pcie_capability_read_word(parent_dev, reg, value);
201 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
202 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
203 		return -1;
204 	return 0;
205 }
206 
207 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
208 {
209 	struct ixgbe_hw *hw = &adapter->hw;
210 	u16 link_status = 0;
211 	int err;
212 
213 	hw->bus.type = ixgbe_bus_type_pci_express;
214 
215 	/* Get the negotiated link width and speed from PCI config space of the
216 	 * parent, as this device is behind a switch
217 	 */
218 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
219 
220 	/* assume caller will handle error case */
221 	if (err)
222 		return err;
223 
224 	hw->bus.width = ixgbe_convert_bus_width(link_status);
225 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
226 
227 	return 0;
228 }
229 
230 /**
231  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
232  * @hw: hw specific details
233  *
234  * This function is used by probe to determine whether a device's PCI-Express
235  * bandwidth details should be gathered from the parent bus instead of from the
236  * device. Used to ensure that various locations all have the correct device ID
237  * checks.
238  */
239 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
240 {
241 	switch (hw->device_id) {
242 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
243 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
244 		return true;
245 	default:
246 		return false;
247 	}
248 }
249 
250 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
251 				     int expected_gts)
252 {
253 	struct ixgbe_hw *hw = &adapter->hw;
254 	int max_gts = 0;
255 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
256 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
257 	struct pci_dev *pdev;
258 
259 	/* Some devices are not connected over PCIe and thus do not negotiate
260 	 * speed. These devices do not have valid bus info, and thus any report
261 	 * we generate may not be correct.
262 	 */
263 	if (hw->bus.type == ixgbe_bus_type_internal)
264 		return;
265 
266 	/* determine whether to use the parent device */
267 	if (ixgbe_pcie_from_parent(&adapter->hw))
268 		pdev = adapter->pdev->bus->parent->self;
269 	else
270 		pdev = adapter->pdev;
271 
272 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
273 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
274 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
275 		return;
276 	}
277 
278 	switch (speed) {
279 	case PCIE_SPEED_2_5GT:
280 		/* 8b/10b encoding reduces max throughput by 20% */
281 		max_gts = 2 * width;
282 		break;
283 	case PCIE_SPEED_5_0GT:
284 		/* 8b/10b encoding reduces max throughput by 20% */
285 		max_gts = 4 * width;
286 		break;
287 	case PCIE_SPEED_8_0GT:
288 		/* 128b/130b encoding reduces throughput by less than 2% */
289 		max_gts = 8 * width;
290 		break;
291 	default:
292 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
293 		return;
294 	}
295 
296 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
297 		   max_gts);
298 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
299 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
300 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
301 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
302 		    "Unknown"),
303 		   width,
304 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
305 		    speed == PCIE_SPEED_5_0GT ? "20%" :
306 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
307 		    "Unknown"));
308 
309 	if (max_gts < expected_gts) {
310 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
311 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
312 			expected_gts);
313 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
314 	}
315 }
316 
317 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
318 {
319 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
320 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
321 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
322 		queue_work(ixgbe_wq, &adapter->service_task);
323 }
324 
325 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
326 {
327 	struct ixgbe_adapter *adapter = hw->back;
328 
329 	if (!hw->hw_addr)
330 		return;
331 	hw->hw_addr = NULL;
332 	e_dev_err("Adapter removed\n");
333 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
334 		ixgbe_service_event_schedule(adapter);
335 }
336 
337 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
338 {
339 	u32 value;
340 
341 	/* The following check not only optimizes a bit by not
342 	 * performing a read on the status register when the
343 	 * register just read was a status register read that
344 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
345 	 * potential recursion.
346 	 */
347 	if (reg == IXGBE_STATUS) {
348 		ixgbe_remove_adapter(hw);
349 		return;
350 	}
351 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
352 	if (value == IXGBE_FAILED_READ_REG)
353 		ixgbe_remove_adapter(hw);
354 }
355 
356 /**
357  * ixgbe_read_reg - Read from device register
358  * @hw: hw specific details
359  * @reg: offset of register to read
360  *
361  * Returns : value read or IXGBE_FAILED_READ_REG if removed
362  *
363  * This function is used to read device registers. It checks for device
364  * removal by confirming any read that returns all ones by checking the
365  * status register value for all ones. This function avoids reading from
366  * the hardware if a removal was previously detected in which case it
367  * returns IXGBE_FAILED_READ_REG (all ones).
368  */
369 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
370 {
371 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
372 	u32 value;
373 
374 	if (ixgbe_removed(reg_addr))
375 		return IXGBE_FAILED_READ_REG;
376 	if (unlikely(hw->phy.nw_mng_if_sel &
377 		     IXGBE_NW_MNG_IF_SEL_ENABLE_10_100M)) {
378 		struct ixgbe_adapter *adapter;
379 		int i;
380 
381 		for (i = 0; i < 200; ++i) {
382 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
383 			if (likely(!value))
384 				goto writes_completed;
385 			if (value == IXGBE_FAILED_READ_REG) {
386 				ixgbe_remove_adapter(hw);
387 				return IXGBE_FAILED_READ_REG;
388 			}
389 			udelay(5);
390 		}
391 
392 		adapter = hw->back;
393 		e_warn(hw, "register writes incomplete %08x\n", value);
394 	}
395 
396 writes_completed:
397 	value = readl(reg_addr + reg);
398 	if (unlikely(value == IXGBE_FAILED_READ_REG))
399 		ixgbe_check_remove(hw, reg);
400 	return value;
401 }
402 
403 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
404 {
405 	u16 value;
406 
407 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
408 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
409 		ixgbe_remove_adapter(hw);
410 		return true;
411 	}
412 	return false;
413 }
414 
415 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
416 {
417 	struct ixgbe_adapter *adapter = hw->back;
418 	u16 value;
419 
420 	if (ixgbe_removed(hw->hw_addr))
421 		return IXGBE_FAILED_READ_CFG_WORD;
422 	pci_read_config_word(adapter->pdev, reg, &value);
423 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
424 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
425 		return IXGBE_FAILED_READ_CFG_WORD;
426 	return value;
427 }
428 
429 #ifdef CONFIG_PCI_IOV
430 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
431 {
432 	struct ixgbe_adapter *adapter = hw->back;
433 	u32 value;
434 
435 	if (ixgbe_removed(hw->hw_addr))
436 		return IXGBE_FAILED_READ_CFG_DWORD;
437 	pci_read_config_dword(adapter->pdev, reg, &value);
438 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
439 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
440 		return IXGBE_FAILED_READ_CFG_DWORD;
441 	return value;
442 }
443 #endif /* CONFIG_PCI_IOV */
444 
445 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
446 {
447 	struct ixgbe_adapter *adapter = hw->back;
448 
449 	if (ixgbe_removed(hw->hw_addr))
450 		return;
451 	pci_write_config_word(adapter->pdev, reg, value);
452 }
453 
454 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
455 {
456 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
457 
458 	/* flush memory to make sure state is correct before next watchdog */
459 	smp_mb__before_atomic();
460 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
461 }
462 
463 struct ixgbe_reg_info {
464 	u32 ofs;
465 	char *name;
466 };
467 
468 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
469 
470 	/* General Registers */
471 	{IXGBE_CTRL, "CTRL"},
472 	{IXGBE_STATUS, "STATUS"},
473 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
474 
475 	/* Interrupt Registers */
476 	{IXGBE_EICR, "EICR"},
477 
478 	/* RX Registers */
479 	{IXGBE_SRRCTL(0), "SRRCTL"},
480 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
481 	{IXGBE_RDLEN(0), "RDLEN"},
482 	{IXGBE_RDH(0), "RDH"},
483 	{IXGBE_RDT(0), "RDT"},
484 	{IXGBE_RXDCTL(0), "RXDCTL"},
485 	{IXGBE_RDBAL(0), "RDBAL"},
486 	{IXGBE_RDBAH(0), "RDBAH"},
487 
488 	/* TX Registers */
489 	{IXGBE_TDBAL(0), "TDBAL"},
490 	{IXGBE_TDBAH(0), "TDBAH"},
491 	{IXGBE_TDLEN(0), "TDLEN"},
492 	{IXGBE_TDH(0), "TDH"},
493 	{IXGBE_TDT(0), "TDT"},
494 	{IXGBE_TXDCTL(0), "TXDCTL"},
495 
496 	/* List Terminator */
497 	{ .name = NULL }
498 };
499 
500 
501 /*
502  * ixgbe_regdump - register printout routine
503  */
504 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
505 {
506 	int i = 0, j = 0;
507 	char rname[16];
508 	u32 regs[64];
509 
510 	switch (reginfo->ofs) {
511 	case IXGBE_SRRCTL(0):
512 		for (i = 0; i < 64; i++)
513 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
514 		break;
515 	case IXGBE_DCA_RXCTRL(0):
516 		for (i = 0; i < 64; i++)
517 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
518 		break;
519 	case IXGBE_RDLEN(0):
520 		for (i = 0; i < 64; i++)
521 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
522 		break;
523 	case IXGBE_RDH(0):
524 		for (i = 0; i < 64; i++)
525 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
526 		break;
527 	case IXGBE_RDT(0):
528 		for (i = 0; i < 64; i++)
529 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
530 		break;
531 	case IXGBE_RXDCTL(0):
532 		for (i = 0; i < 64; i++)
533 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
534 		break;
535 	case IXGBE_RDBAL(0):
536 		for (i = 0; i < 64; i++)
537 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
538 		break;
539 	case IXGBE_RDBAH(0):
540 		for (i = 0; i < 64; i++)
541 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
542 		break;
543 	case IXGBE_TDBAL(0):
544 		for (i = 0; i < 64; i++)
545 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
546 		break;
547 	case IXGBE_TDBAH(0):
548 		for (i = 0; i < 64; i++)
549 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
550 		break;
551 	case IXGBE_TDLEN(0):
552 		for (i = 0; i < 64; i++)
553 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
554 		break;
555 	case IXGBE_TDH(0):
556 		for (i = 0; i < 64; i++)
557 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
558 		break;
559 	case IXGBE_TDT(0):
560 		for (i = 0; i < 64; i++)
561 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
562 		break;
563 	case IXGBE_TXDCTL(0):
564 		for (i = 0; i < 64; i++)
565 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
566 		break;
567 	default:
568 		pr_info("%-15s %08x\n", reginfo->name,
569 			IXGBE_READ_REG(hw, reginfo->ofs));
570 		return;
571 	}
572 
573 	for (i = 0; i < 8; i++) {
574 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
575 		pr_err("%-15s", rname);
576 		for (j = 0; j < 8; j++)
577 			pr_cont(" %08x", regs[i*8+j]);
578 		pr_cont("\n");
579 	}
580 
581 }
582 
583 /*
584  * ixgbe_dump - Print registers, tx-rings and rx-rings
585  */
586 static void ixgbe_dump(struct ixgbe_adapter *adapter)
587 {
588 	struct net_device *netdev = adapter->netdev;
589 	struct ixgbe_hw *hw = &adapter->hw;
590 	struct ixgbe_reg_info *reginfo;
591 	int n = 0;
592 	struct ixgbe_ring *tx_ring;
593 	struct ixgbe_tx_buffer *tx_buffer;
594 	union ixgbe_adv_tx_desc *tx_desc;
595 	struct my_u0 { u64 a; u64 b; } *u0;
596 	struct ixgbe_ring *rx_ring;
597 	union ixgbe_adv_rx_desc *rx_desc;
598 	struct ixgbe_rx_buffer *rx_buffer_info;
599 	u32 staterr;
600 	int i = 0;
601 
602 	if (!netif_msg_hw(adapter))
603 		return;
604 
605 	/* Print netdevice Info */
606 	if (netdev) {
607 		dev_info(&adapter->pdev->dev, "Net device Info\n");
608 		pr_info("Device Name     state            "
609 			"trans_start      last_rx\n");
610 		pr_info("%-15s %016lX %016lX %016lX\n",
611 			netdev->name,
612 			netdev->state,
613 			dev_trans_start(netdev),
614 			netdev->last_rx);
615 	}
616 
617 	/* Print Registers */
618 	dev_info(&adapter->pdev->dev, "Register Dump\n");
619 	pr_info(" Register Name   Value\n");
620 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
621 	     reginfo->name; reginfo++) {
622 		ixgbe_regdump(hw, reginfo);
623 	}
624 
625 	/* Print TX Ring Summary */
626 	if (!netdev || !netif_running(netdev))
627 		return;
628 
629 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
630 	pr_info(" %s     %s              %s        %s\n",
631 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
632 		"leng", "ntw", "timestamp");
633 	for (n = 0; n < adapter->num_tx_queues; n++) {
634 		tx_ring = adapter->tx_ring[n];
635 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
636 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
637 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
638 			   (u64)dma_unmap_addr(tx_buffer, dma),
639 			   dma_unmap_len(tx_buffer, len),
640 			   tx_buffer->next_to_watch,
641 			   (u64)tx_buffer->time_stamp);
642 	}
643 
644 	/* Print TX Rings */
645 	if (!netif_msg_tx_done(adapter))
646 		goto rx_ring_summary;
647 
648 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
649 
650 	/* Transmit Descriptor Formats
651 	 *
652 	 * 82598 Advanced Transmit Descriptor
653 	 *   +--------------------------------------------------------------+
654 	 * 0 |         Buffer Address [63:0]                                |
655 	 *   +--------------------------------------------------------------+
656 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
657 	 *   +--------------------------------------------------------------+
658 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
659 	 *
660 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
661 	 *   +--------------------------------------------------------------+
662 	 * 0 |                          RSV [63:0]                          |
663 	 *   +--------------------------------------------------------------+
664 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
665 	 *   +--------------------------------------------------------------+
666 	 *   63                       36 35   32 31                         0
667 	 *
668 	 * 82599+ Advanced Transmit Descriptor
669 	 *   +--------------------------------------------------------------+
670 	 * 0 |         Buffer Address [63:0]                                |
671 	 *   +--------------------------------------------------------------+
672 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
673 	 *   +--------------------------------------------------------------+
674 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
675 	 *
676 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
677 	 *   +--------------------------------------------------------------+
678 	 * 0 |                          RSV [63:0]                          |
679 	 *   +--------------------------------------------------------------+
680 	 * 8 |            RSV           |  STA  |           RSV             |
681 	 *   +--------------------------------------------------------------+
682 	 *   63                       36 35   32 31                         0
683 	 */
684 
685 	for (n = 0; n < adapter->num_tx_queues; n++) {
686 		tx_ring = adapter->tx_ring[n];
687 		pr_info("------------------------------------\n");
688 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
689 		pr_info("------------------------------------\n");
690 		pr_info("%s%s    %s              %s        %s          %s\n",
691 			"T [desc]     [address 63:0  ] ",
692 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
693 			"leng", "ntw", "timestamp", "bi->skb");
694 
695 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
696 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
697 			tx_buffer = &tx_ring->tx_buffer_info[i];
698 			u0 = (struct my_u0 *)tx_desc;
699 			if (dma_unmap_len(tx_buffer, len) > 0) {
700 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
701 					i,
702 					le64_to_cpu(u0->a),
703 					le64_to_cpu(u0->b),
704 					(u64)dma_unmap_addr(tx_buffer, dma),
705 					dma_unmap_len(tx_buffer, len),
706 					tx_buffer->next_to_watch,
707 					(u64)tx_buffer->time_stamp,
708 					tx_buffer->skb);
709 				if (i == tx_ring->next_to_use &&
710 					i == tx_ring->next_to_clean)
711 					pr_cont(" NTC/U\n");
712 				else if (i == tx_ring->next_to_use)
713 					pr_cont(" NTU\n");
714 				else if (i == tx_ring->next_to_clean)
715 					pr_cont(" NTC\n");
716 				else
717 					pr_cont("\n");
718 
719 				if (netif_msg_pktdata(adapter) &&
720 				    tx_buffer->skb)
721 					print_hex_dump(KERN_INFO, "",
722 						DUMP_PREFIX_ADDRESS, 16, 1,
723 						tx_buffer->skb->data,
724 						dma_unmap_len(tx_buffer, len),
725 						true);
726 			}
727 		}
728 	}
729 
730 	/* Print RX Rings Summary */
731 rx_ring_summary:
732 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
733 	pr_info("Queue [NTU] [NTC]\n");
734 	for (n = 0; n < adapter->num_rx_queues; n++) {
735 		rx_ring = adapter->rx_ring[n];
736 		pr_info("%5d %5X %5X\n",
737 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
738 	}
739 
740 	/* Print RX Rings */
741 	if (!netif_msg_rx_status(adapter))
742 		return;
743 
744 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
745 
746 	/* Receive Descriptor Formats
747 	 *
748 	 * 82598 Advanced Receive Descriptor (Read) Format
749 	 *    63                                           1        0
750 	 *    +-----------------------------------------------------+
751 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
752 	 *    +----------------------------------------------+------+
753 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
754 	 *    +-----------------------------------------------------+
755 	 *
756 	 *
757 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
758 	 *
759 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
760 	 *   +------------------------------------------------------+
761 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
762 	 *   | Packet   | IP     |   |          |     | Type | Type |
763 	 *   | Checksum | Ident  |   |          |     |      |      |
764 	 *   +------------------------------------------------------+
765 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
766 	 *   +------------------------------------------------------+
767 	 *   63       48 47    32 31            20 19               0
768 	 *
769 	 * 82599+ Advanced Receive Descriptor (Read) Format
770 	 *    63                                           1        0
771 	 *    +-----------------------------------------------------+
772 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
773 	 *    +----------------------------------------------+------+
774 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
775 	 *    +-----------------------------------------------------+
776 	 *
777 	 *
778 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
779 	 *
780 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
781 	 *   +------------------------------------------------------+
782 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
783 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
784 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
785 	 *   +------------------------------------------------------+
786 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
787 	 *   +------------------------------------------------------+
788 	 *   63       48 47    32 31          20 19                 0
789 	 */
790 
791 	for (n = 0; n < adapter->num_rx_queues; n++) {
792 		rx_ring = adapter->rx_ring[n];
793 		pr_info("------------------------------------\n");
794 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
795 		pr_info("------------------------------------\n");
796 		pr_info("%s%s%s",
797 			"R  [desc]      [ PktBuf     A0] ",
798 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
799 			"<-- Adv Rx Read format\n");
800 		pr_info("%s%s%s",
801 			"RWB[desc]      [PcsmIpSHl PtRs] ",
802 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
803 			"<-- Adv Rx Write-Back format\n");
804 
805 		for (i = 0; i < rx_ring->count; i++) {
806 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
807 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
808 			u0 = (struct my_u0 *)rx_desc;
809 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
810 			if (staterr & IXGBE_RXD_STAT_DD) {
811 				/* Descriptor Done */
812 				pr_info("RWB[0x%03X]     %016llX "
813 					"%016llX ---------------- %p", i,
814 					le64_to_cpu(u0->a),
815 					le64_to_cpu(u0->b),
816 					rx_buffer_info->skb);
817 			} else {
818 				pr_info("R  [0x%03X]     %016llX "
819 					"%016llX %016llX %p", i,
820 					le64_to_cpu(u0->a),
821 					le64_to_cpu(u0->b),
822 					(u64)rx_buffer_info->dma,
823 					rx_buffer_info->skb);
824 
825 				if (netif_msg_pktdata(adapter) &&
826 				    rx_buffer_info->dma) {
827 					print_hex_dump(KERN_INFO, "",
828 					   DUMP_PREFIX_ADDRESS, 16, 1,
829 					   page_address(rx_buffer_info->page) +
830 						    rx_buffer_info->page_offset,
831 					   ixgbe_rx_bufsz(rx_ring), true);
832 				}
833 			}
834 
835 			if (i == rx_ring->next_to_use)
836 				pr_cont(" NTU\n");
837 			else if (i == rx_ring->next_to_clean)
838 				pr_cont(" NTC\n");
839 			else
840 				pr_cont("\n");
841 
842 		}
843 	}
844 }
845 
846 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
847 {
848 	u32 ctrl_ext;
849 
850 	/* Let firmware take over control of h/w */
851 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
852 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
853 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
854 }
855 
856 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
857 {
858 	u32 ctrl_ext;
859 
860 	/* Let firmware know the driver has taken over */
861 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
862 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
863 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
864 }
865 
866 /**
867  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
868  * @adapter: pointer to adapter struct
869  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
870  * @queue: queue to map the corresponding interrupt to
871  * @msix_vector: the vector to map to the corresponding queue
872  *
873  */
874 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
875 			   u8 queue, u8 msix_vector)
876 {
877 	u32 ivar, index;
878 	struct ixgbe_hw *hw = &adapter->hw;
879 	switch (hw->mac.type) {
880 	case ixgbe_mac_82598EB:
881 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
882 		if (direction == -1)
883 			direction = 0;
884 		index = (((direction * 64) + queue) >> 2) & 0x1F;
885 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
886 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
887 		ivar |= (msix_vector << (8 * (queue & 0x3)));
888 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
889 		break;
890 	case ixgbe_mac_82599EB:
891 	case ixgbe_mac_X540:
892 	case ixgbe_mac_X550:
893 	case ixgbe_mac_X550EM_x:
894 	case ixgbe_mac_x550em_a:
895 		if (direction == -1) {
896 			/* other causes */
897 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
898 			index = ((queue & 1) * 8);
899 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
900 			ivar &= ~(0xFF << index);
901 			ivar |= (msix_vector << index);
902 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
903 			break;
904 		} else {
905 			/* tx or rx causes */
906 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
907 			index = ((16 * (queue & 1)) + (8 * direction));
908 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
909 			ivar &= ~(0xFF << index);
910 			ivar |= (msix_vector << index);
911 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
912 			break;
913 		}
914 	default:
915 		break;
916 	}
917 }
918 
919 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
920 					  u64 qmask)
921 {
922 	u32 mask;
923 
924 	switch (adapter->hw.mac.type) {
925 	case ixgbe_mac_82598EB:
926 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
927 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
928 		break;
929 	case ixgbe_mac_82599EB:
930 	case ixgbe_mac_X540:
931 	case ixgbe_mac_X550:
932 	case ixgbe_mac_X550EM_x:
933 	case ixgbe_mac_x550em_a:
934 		mask = (qmask & 0xFFFFFFFF);
935 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
936 		mask = (qmask >> 32);
937 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
938 		break;
939 	default:
940 		break;
941 	}
942 }
943 
944 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
945 				      struct ixgbe_tx_buffer *tx_buffer)
946 {
947 	if (tx_buffer->skb) {
948 		dev_kfree_skb_any(tx_buffer->skb);
949 		if (dma_unmap_len(tx_buffer, len))
950 			dma_unmap_single(ring->dev,
951 					 dma_unmap_addr(tx_buffer, dma),
952 					 dma_unmap_len(tx_buffer, len),
953 					 DMA_TO_DEVICE);
954 	} else if (dma_unmap_len(tx_buffer, len)) {
955 		dma_unmap_page(ring->dev,
956 			       dma_unmap_addr(tx_buffer, dma),
957 			       dma_unmap_len(tx_buffer, len),
958 			       DMA_TO_DEVICE);
959 	}
960 	tx_buffer->next_to_watch = NULL;
961 	tx_buffer->skb = NULL;
962 	dma_unmap_len_set(tx_buffer, len, 0);
963 	/* tx_buffer must be completely set up in the transmit path */
964 }
965 
966 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
967 {
968 	struct ixgbe_hw *hw = &adapter->hw;
969 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
970 	int i;
971 	u32 data;
972 
973 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
974 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
975 		return;
976 
977 	switch (hw->mac.type) {
978 	case ixgbe_mac_82598EB:
979 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
980 		break;
981 	default:
982 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
983 	}
984 	hwstats->lxoffrxc += data;
985 
986 	/* refill credits (no tx hang) if we received xoff */
987 	if (!data)
988 		return;
989 
990 	for (i = 0; i < adapter->num_tx_queues; i++)
991 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
992 			  &adapter->tx_ring[i]->state);
993 }
994 
995 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
996 {
997 	struct ixgbe_hw *hw = &adapter->hw;
998 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
999 	u32 xoff[8] = {0};
1000 	u8 tc;
1001 	int i;
1002 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
1003 
1004 	if (adapter->ixgbe_ieee_pfc)
1005 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
1006 
1007 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
1008 		ixgbe_update_xoff_rx_lfc(adapter);
1009 		return;
1010 	}
1011 
1012 	/* update stats for each tc, only valid with PFC enabled */
1013 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
1014 		u32 pxoffrxc;
1015 
1016 		switch (hw->mac.type) {
1017 		case ixgbe_mac_82598EB:
1018 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1019 			break;
1020 		default:
1021 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1022 		}
1023 		hwstats->pxoffrxc[i] += pxoffrxc;
1024 		/* Get the TC for given UP */
1025 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1026 		xoff[tc] += pxoffrxc;
1027 	}
1028 
1029 	/* disarm tx queues that have received xoff frames */
1030 	for (i = 0; i < adapter->num_tx_queues; i++) {
1031 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1032 
1033 		tc = tx_ring->dcb_tc;
1034 		if (xoff[tc])
1035 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1036 	}
1037 }
1038 
1039 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1040 {
1041 	return ring->stats.packets;
1042 }
1043 
1044 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1045 {
1046 	struct ixgbe_adapter *adapter;
1047 	struct ixgbe_hw *hw;
1048 	u32 head, tail;
1049 
1050 	if (ring->l2_accel_priv)
1051 		adapter = ring->l2_accel_priv->real_adapter;
1052 	else
1053 		adapter = netdev_priv(ring->netdev);
1054 
1055 	hw = &adapter->hw;
1056 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1057 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1058 
1059 	if (head != tail)
1060 		return (head < tail) ?
1061 			tail - head : (tail + ring->count - head);
1062 
1063 	return 0;
1064 }
1065 
1066 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1067 {
1068 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1069 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1070 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1071 
1072 	clear_check_for_tx_hang(tx_ring);
1073 
1074 	/*
1075 	 * Check for a hung queue, but be thorough. This verifies
1076 	 * that a transmit has been completed since the previous
1077 	 * check AND there is at least one packet pending. The
1078 	 * ARMED bit is set to indicate a potential hang. The
1079 	 * bit is cleared if a pause frame is received to remove
1080 	 * false hang detection due to PFC or 802.3x frames. By
1081 	 * requiring this to fail twice we avoid races with
1082 	 * pfc clearing the ARMED bit and conditions where we
1083 	 * run the check_tx_hang logic with a transmit completion
1084 	 * pending but without time to complete it yet.
1085 	 */
1086 	if (tx_done_old == tx_done && tx_pending)
1087 		/* make sure it is true for two checks in a row */
1088 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1089 					&tx_ring->state);
1090 	/* update completed stats and continue */
1091 	tx_ring->tx_stats.tx_done_old = tx_done;
1092 	/* reset the countdown */
1093 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1094 
1095 	return false;
1096 }
1097 
1098 /**
1099  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1100  * @adapter: driver private struct
1101  **/
1102 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1103 {
1104 
1105 	/* Do the reset outside of interrupt context */
1106 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1107 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1108 		e_warn(drv, "initiating reset due to tx timeout\n");
1109 		ixgbe_service_event_schedule(adapter);
1110 	}
1111 }
1112 
1113 /**
1114  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1115  **/
1116 static int ixgbe_tx_maxrate(struct net_device *netdev,
1117 			    int queue_index, u32 maxrate)
1118 {
1119 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1120 	struct ixgbe_hw *hw = &adapter->hw;
1121 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1122 
1123 	if (!maxrate)
1124 		return 0;
1125 
1126 	/* Calculate the rate factor values to set */
1127 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1128 	bcnrc_val /= maxrate;
1129 
1130 	/* clear everything but the rate factor */
1131 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1132 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1133 
1134 	/* enable the rate scheduler */
1135 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1136 
1137 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1138 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1139 
1140 	return 0;
1141 }
1142 
1143 /**
1144  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1145  * @q_vector: structure containing interrupt and ring information
1146  * @tx_ring: tx ring to clean
1147  * @napi_budget: Used to determine if we are in netpoll
1148  **/
1149 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1150 			       struct ixgbe_ring *tx_ring, int napi_budget)
1151 {
1152 	struct ixgbe_adapter *adapter = q_vector->adapter;
1153 	struct ixgbe_tx_buffer *tx_buffer;
1154 	union ixgbe_adv_tx_desc *tx_desc;
1155 	unsigned int total_bytes = 0, total_packets = 0;
1156 	unsigned int budget = q_vector->tx.work_limit;
1157 	unsigned int i = tx_ring->next_to_clean;
1158 
1159 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1160 		return true;
1161 
1162 	tx_buffer = &tx_ring->tx_buffer_info[i];
1163 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1164 	i -= tx_ring->count;
1165 
1166 	do {
1167 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1168 
1169 		/* if next_to_watch is not set then there is no work pending */
1170 		if (!eop_desc)
1171 			break;
1172 
1173 		/* prevent any other reads prior to eop_desc */
1174 		read_barrier_depends();
1175 
1176 		/* if DD is not set pending work has not been completed */
1177 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1178 			break;
1179 
1180 		/* clear next_to_watch to prevent false hangs */
1181 		tx_buffer->next_to_watch = NULL;
1182 
1183 		/* update the statistics for this packet */
1184 		total_bytes += tx_buffer->bytecount;
1185 		total_packets += tx_buffer->gso_segs;
1186 
1187 		/* free the skb */
1188 		napi_consume_skb(tx_buffer->skb, napi_budget);
1189 
1190 		/* unmap skb header data */
1191 		dma_unmap_single(tx_ring->dev,
1192 				 dma_unmap_addr(tx_buffer, dma),
1193 				 dma_unmap_len(tx_buffer, len),
1194 				 DMA_TO_DEVICE);
1195 
1196 		/* clear tx_buffer data */
1197 		tx_buffer->skb = NULL;
1198 		dma_unmap_len_set(tx_buffer, len, 0);
1199 
1200 		/* unmap remaining buffers */
1201 		while (tx_desc != eop_desc) {
1202 			tx_buffer++;
1203 			tx_desc++;
1204 			i++;
1205 			if (unlikely(!i)) {
1206 				i -= tx_ring->count;
1207 				tx_buffer = tx_ring->tx_buffer_info;
1208 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1209 			}
1210 
1211 			/* unmap any remaining paged data */
1212 			if (dma_unmap_len(tx_buffer, len)) {
1213 				dma_unmap_page(tx_ring->dev,
1214 					       dma_unmap_addr(tx_buffer, dma),
1215 					       dma_unmap_len(tx_buffer, len),
1216 					       DMA_TO_DEVICE);
1217 				dma_unmap_len_set(tx_buffer, len, 0);
1218 			}
1219 		}
1220 
1221 		/* move us one more past the eop_desc for start of next pkt */
1222 		tx_buffer++;
1223 		tx_desc++;
1224 		i++;
1225 		if (unlikely(!i)) {
1226 			i -= tx_ring->count;
1227 			tx_buffer = tx_ring->tx_buffer_info;
1228 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1229 		}
1230 
1231 		/* issue prefetch for next Tx descriptor */
1232 		prefetch(tx_desc);
1233 
1234 		/* update budget accounting */
1235 		budget--;
1236 	} while (likely(budget));
1237 
1238 	i += tx_ring->count;
1239 	tx_ring->next_to_clean = i;
1240 	u64_stats_update_begin(&tx_ring->syncp);
1241 	tx_ring->stats.bytes += total_bytes;
1242 	tx_ring->stats.packets += total_packets;
1243 	u64_stats_update_end(&tx_ring->syncp);
1244 	q_vector->tx.total_bytes += total_bytes;
1245 	q_vector->tx.total_packets += total_packets;
1246 
1247 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1248 		/* schedule immediate reset if we believe we hung */
1249 		struct ixgbe_hw *hw = &adapter->hw;
1250 		e_err(drv, "Detected Tx Unit Hang\n"
1251 			"  Tx Queue             <%d>\n"
1252 			"  TDH, TDT             <%x>, <%x>\n"
1253 			"  next_to_use          <%x>\n"
1254 			"  next_to_clean        <%x>\n"
1255 			"tx_buffer_info[next_to_clean]\n"
1256 			"  time_stamp           <%lx>\n"
1257 			"  jiffies              <%lx>\n",
1258 			tx_ring->queue_index,
1259 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1260 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1261 			tx_ring->next_to_use, i,
1262 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1263 
1264 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1265 
1266 		e_info(probe,
1267 		       "tx hang %d detected on queue %d, resetting adapter\n",
1268 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1269 
1270 		/* schedule immediate reset if we believe we hung */
1271 		ixgbe_tx_timeout_reset(adapter);
1272 
1273 		/* the adapter is about to reset, no point in enabling stuff */
1274 		return true;
1275 	}
1276 
1277 	netdev_tx_completed_queue(txring_txq(tx_ring),
1278 				  total_packets, total_bytes);
1279 
1280 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1281 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1282 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1283 		/* Make sure that anybody stopping the queue after this
1284 		 * sees the new next_to_clean.
1285 		 */
1286 		smp_mb();
1287 		if (__netif_subqueue_stopped(tx_ring->netdev,
1288 					     tx_ring->queue_index)
1289 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1290 			netif_wake_subqueue(tx_ring->netdev,
1291 					    tx_ring->queue_index);
1292 			++tx_ring->tx_stats.restart_queue;
1293 		}
1294 	}
1295 
1296 	return !!budget;
1297 }
1298 
1299 #ifdef CONFIG_IXGBE_DCA
1300 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1301 				struct ixgbe_ring *tx_ring,
1302 				int cpu)
1303 {
1304 	struct ixgbe_hw *hw = &adapter->hw;
1305 	u32 txctrl = 0;
1306 	u16 reg_offset;
1307 
1308 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1309 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1310 
1311 	switch (hw->mac.type) {
1312 	case ixgbe_mac_82598EB:
1313 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1314 		break;
1315 	case ixgbe_mac_82599EB:
1316 	case ixgbe_mac_X540:
1317 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1318 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1319 		break;
1320 	default:
1321 		/* for unknown hardware do not write register */
1322 		return;
1323 	}
1324 
1325 	/*
1326 	 * We can enable relaxed ordering for reads, but not writes when
1327 	 * DCA is enabled.  This is due to a known issue in some chipsets
1328 	 * which will cause the DCA tag to be cleared.
1329 	 */
1330 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1331 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1332 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1333 
1334 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1335 }
1336 
1337 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1338 				struct ixgbe_ring *rx_ring,
1339 				int cpu)
1340 {
1341 	struct ixgbe_hw *hw = &adapter->hw;
1342 	u32 rxctrl = 0;
1343 	u8 reg_idx = rx_ring->reg_idx;
1344 
1345 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1346 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1347 
1348 	switch (hw->mac.type) {
1349 	case ixgbe_mac_82599EB:
1350 	case ixgbe_mac_X540:
1351 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1352 		break;
1353 	default:
1354 		break;
1355 	}
1356 
1357 	/*
1358 	 * We can enable relaxed ordering for reads, but not writes when
1359 	 * DCA is enabled.  This is due to a known issue in some chipsets
1360 	 * which will cause the DCA tag to be cleared.
1361 	 */
1362 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1363 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1364 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1365 
1366 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1367 }
1368 
1369 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1370 {
1371 	struct ixgbe_adapter *adapter = q_vector->adapter;
1372 	struct ixgbe_ring *ring;
1373 	int cpu = get_cpu();
1374 
1375 	if (q_vector->cpu == cpu)
1376 		goto out_no_update;
1377 
1378 	ixgbe_for_each_ring(ring, q_vector->tx)
1379 		ixgbe_update_tx_dca(adapter, ring, cpu);
1380 
1381 	ixgbe_for_each_ring(ring, q_vector->rx)
1382 		ixgbe_update_rx_dca(adapter, ring, cpu);
1383 
1384 	q_vector->cpu = cpu;
1385 out_no_update:
1386 	put_cpu();
1387 }
1388 
1389 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1390 {
1391 	int i;
1392 
1393 	/* always use CB2 mode, difference is masked in the CB driver */
1394 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1395 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1396 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1397 	else
1398 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1399 				IXGBE_DCA_CTRL_DCA_DISABLE);
1400 
1401 	for (i = 0; i < adapter->num_q_vectors; i++) {
1402 		adapter->q_vector[i]->cpu = -1;
1403 		ixgbe_update_dca(adapter->q_vector[i]);
1404 	}
1405 }
1406 
1407 static int __ixgbe_notify_dca(struct device *dev, void *data)
1408 {
1409 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1410 	unsigned long event = *(unsigned long *)data;
1411 
1412 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1413 		return 0;
1414 
1415 	switch (event) {
1416 	case DCA_PROVIDER_ADD:
1417 		/* if we're already enabled, don't do it again */
1418 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1419 			break;
1420 		if (dca_add_requester(dev) == 0) {
1421 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1422 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1423 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1424 			break;
1425 		}
1426 		/* Fall Through since DCA is disabled. */
1427 	case DCA_PROVIDER_REMOVE:
1428 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1429 			dca_remove_requester(dev);
1430 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1431 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1432 					IXGBE_DCA_CTRL_DCA_DISABLE);
1433 		}
1434 		break;
1435 	}
1436 
1437 	return 0;
1438 }
1439 
1440 #endif /* CONFIG_IXGBE_DCA */
1441 
1442 #define IXGBE_RSS_L4_TYPES_MASK \
1443 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1444 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1445 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1446 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1447 
1448 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1449 				 union ixgbe_adv_rx_desc *rx_desc,
1450 				 struct sk_buff *skb)
1451 {
1452 	u16 rss_type;
1453 
1454 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1455 		return;
1456 
1457 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1458 		   IXGBE_RXDADV_RSSTYPE_MASK;
1459 
1460 	if (!rss_type)
1461 		return;
1462 
1463 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1464 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1465 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1466 }
1467 
1468 #ifdef IXGBE_FCOE
1469 /**
1470  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1471  * @ring: structure containing ring specific data
1472  * @rx_desc: advanced rx descriptor
1473  *
1474  * Returns : true if it is FCoE pkt
1475  */
1476 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1477 				    union ixgbe_adv_rx_desc *rx_desc)
1478 {
1479 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1480 
1481 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1482 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1483 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1484 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1485 }
1486 
1487 #endif /* IXGBE_FCOE */
1488 /**
1489  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1490  * @ring: structure containing ring specific data
1491  * @rx_desc: current Rx descriptor being processed
1492  * @skb: skb currently being received and modified
1493  **/
1494 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1495 				     union ixgbe_adv_rx_desc *rx_desc,
1496 				     struct sk_buff *skb)
1497 {
1498 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_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 	/* check for VXLAN and Geneve packets */
1508 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
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 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
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 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
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 		hw_dbg(hw, "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 
3926 	/* disable NFS filtering */
3927 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
3928 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3929 
3930 	/* Program registers for the distribution of queues */
3931 	ixgbe_setup_mrqc(adapter);
3932 
3933 	/* set_rx_buffer_len must be called before ring initialization */
3934 	ixgbe_set_rx_buffer_len(adapter);
3935 
3936 	/*
3937 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3938 	 * the Base and Length of the Rx Descriptor Ring
3939 	 */
3940 	for (i = 0; i < adapter->num_rx_queues; i++)
3941 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3942 
3943 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3944 	/* disable drop enable for 82598 parts */
3945 	if (hw->mac.type == ixgbe_mac_82598EB)
3946 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3947 
3948 	/* enable all receives */
3949 	rxctrl |= IXGBE_RXCTRL_RXEN;
3950 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3951 }
3952 
3953 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3954 				 __be16 proto, u16 vid)
3955 {
3956 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3957 	struct ixgbe_hw *hw = &adapter->hw;
3958 
3959 	/* add VID to filter table */
3960 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
3961 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
3962 
3963 	set_bit(vid, adapter->active_vlans);
3964 
3965 	return 0;
3966 }
3967 
3968 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
3969 {
3970 	u32 vlvf;
3971 	int idx;
3972 
3973 	/* short cut the special case */
3974 	if (vlan == 0)
3975 		return 0;
3976 
3977 	/* Search for the vlan id in the VLVF entries */
3978 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
3979 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
3980 		if ((vlvf & VLAN_VID_MASK) == vlan)
3981 			break;
3982 	}
3983 
3984 	return idx;
3985 }
3986 
3987 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
3988 {
3989 	struct ixgbe_hw *hw = &adapter->hw;
3990 	u32 bits, word;
3991 	int idx;
3992 
3993 	idx = ixgbe_find_vlvf_entry(hw, vid);
3994 	if (!idx)
3995 		return;
3996 
3997 	/* See if any other pools are set for this VLAN filter
3998 	 * entry other than the PF.
3999 	 */
4000 	word = idx * 2 + (VMDQ_P(0) / 32);
4001 	bits = ~BIT(VMDQ_P(0) % 32);
4002 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4003 
4004 	/* Disable the filter so this falls into the default pool. */
4005 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4006 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4007 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4008 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4009 	}
4010 }
4011 
4012 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4013 				  __be16 proto, u16 vid)
4014 {
4015 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4016 	struct ixgbe_hw *hw = &adapter->hw;
4017 
4018 	/* remove VID from filter table */
4019 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4020 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4021 
4022 	clear_bit(vid, adapter->active_vlans);
4023 
4024 	return 0;
4025 }
4026 
4027 /**
4028  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4029  * @adapter: driver data
4030  */
4031 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4032 {
4033 	struct ixgbe_hw *hw = &adapter->hw;
4034 	u32 vlnctrl;
4035 	int i, j;
4036 
4037 	switch (hw->mac.type) {
4038 	case ixgbe_mac_82598EB:
4039 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4040 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4041 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4042 		break;
4043 	case ixgbe_mac_82599EB:
4044 	case ixgbe_mac_X540:
4045 	case ixgbe_mac_X550:
4046 	case ixgbe_mac_X550EM_x:
4047 	case ixgbe_mac_x550em_a:
4048 		for (i = 0; i < adapter->num_rx_queues; i++) {
4049 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4050 
4051 			if (ring->l2_accel_priv)
4052 				continue;
4053 			j = ring->reg_idx;
4054 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4055 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4056 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4057 		}
4058 		break;
4059 	default:
4060 		break;
4061 	}
4062 }
4063 
4064 /**
4065  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4066  * @adapter: driver data
4067  */
4068 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4069 {
4070 	struct ixgbe_hw *hw = &adapter->hw;
4071 	u32 vlnctrl;
4072 	int i, j;
4073 
4074 	switch (hw->mac.type) {
4075 	case ixgbe_mac_82598EB:
4076 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4077 		vlnctrl |= IXGBE_VLNCTRL_VME;
4078 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4079 		break;
4080 	case ixgbe_mac_82599EB:
4081 	case ixgbe_mac_X540:
4082 	case ixgbe_mac_X550:
4083 	case ixgbe_mac_X550EM_x:
4084 	case ixgbe_mac_x550em_a:
4085 		for (i = 0; i < adapter->num_rx_queues; i++) {
4086 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4087 
4088 			if (ring->l2_accel_priv)
4089 				continue;
4090 			j = ring->reg_idx;
4091 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4092 			vlnctrl |= IXGBE_RXDCTL_VME;
4093 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4094 		}
4095 		break;
4096 	default:
4097 		break;
4098 	}
4099 }
4100 
4101 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4102 {
4103 	struct ixgbe_hw *hw = &adapter->hw;
4104 	u32 vlnctrl, i;
4105 
4106 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4107 
4108 	switch (hw->mac.type) {
4109 	case ixgbe_mac_82599EB:
4110 	case ixgbe_mac_X540:
4111 	case ixgbe_mac_X550:
4112 	case ixgbe_mac_X550EM_x:
4113 	case ixgbe_mac_x550em_a:
4114 	default:
4115 		if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED)
4116 			break;
4117 		/* fall through */
4118 	case ixgbe_mac_82598EB:
4119 		/* legacy case, we can just disable VLAN filtering */
4120 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4121 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4122 		return;
4123 	}
4124 
4125 	/* We are already in VLAN promisc, nothing to do */
4126 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4127 		return;
4128 
4129 	/* Set flag so we don't redo unnecessary work */
4130 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4131 
4132 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4133 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4134 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4135 
4136 	/* Add PF to all active pools */
4137 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4138 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4139 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4140 
4141 		vlvfb |= BIT(VMDQ_P(0) % 32);
4142 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4143 	}
4144 
4145 	/* Set all bits in the VLAN filter table array */
4146 	for (i = hw->mac.vft_size; i--;)
4147 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4148 }
4149 
4150 #define VFTA_BLOCK_SIZE 8
4151 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4152 {
4153 	struct ixgbe_hw *hw = &adapter->hw;
4154 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4155 	u32 vid_start = vfta_offset * 32;
4156 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4157 	u32 i, vid, word, bits;
4158 
4159 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4160 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4161 
4162 		/* pull VLAN ID from VLVF */
4163 		vid = vlvf & VLAN_VID_MASK;
4164 
4165 		/* only concern outselves with a certain range */
4166 		if (vid < vid_start || vid >= vid_end)
4167 			continue;
4168 
4169 		if (vlvf) {
4170 			/* record VLAN ID in VFTA */
4171 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4172 
4173 			/* if PF is part of this then continue */
4174 			if (test_bit(vid, adapter->active_vlans))
4175 				continue;
4176 		}
4177 
4178 		/* remove PF from the pool */
4179 		word = i * 2 + VMDQ_P(0) / 32;
4180 		bits = ~BIT(VMDQ_P(0) % 32);
4181 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4182 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4183 	}
4184 
4185 	/* extract values from active_vlans and write back to VFTA */
4186 	for (i = VFTA_BLOCK_SIZE; i--;) {
4187 		vid = (vfta_offset + i) * 32;
4188 		word = vid / BITS_PER_LONG;
4189 		bits = vid % BITS_PER_LONG;
4190 
4191 		vfta[i] |= adapter->active_vlans[word] >> bits;
4192 
4193 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4194 	}
4195 }
4196 
4197 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4198 {
4199 	struct ixgbe_hw *hw = &adapter->hw;
4200 	u32 vlnctrl, i;
4201 
4202 	/* Set VLAN filtering to enabled */
4203 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4204 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4205 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4206 
4207 	switch (hw->mac.type) {
4208 	case ixgbe_mac_82599EB:
4209 	case ixgbe_mac_X540:
4210 	case ixgbe_mac_X550:
4211 	case ixgbe_mac_X550EM_x:
4212 	case ixgbe_mac_x550em_a:
4213 	default:
4214 		if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED)
4215 			break;
4216 		/* fall through */
4217 	case ixgbe_mac_82598EB:
4218 		return;
4219 	}
4220 
4221 	/* We are not in VLAN promisc, nothing to do */
4222 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4223 		return;
4224 
4225 	/* Set flag so we don't redo unnecessary work */
4226 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4227 
4228 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4229 		ixgbe_scrub_vfta(adapter, i);
4230 }
4231 
4232 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4233 {
4234 	u16 vid = 1;
4235 
4236 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4237 
4238 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4239 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4240 }
4241 
4242 /**
4243  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4244  * @netdev: network interface device structure
4245  *
4246  * Writes multicast address list to the MTA hash table.
4247  * Returns: -ENOMEM on failure
4248  *                0 on no addresses written
4249  *                X on writing X addresses to MTA
4250  **/
4251 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4252 {
4253 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4254 	struct ixgbe_hw *hw = &adapter->hw;
4255 
4256 	if (!netif_running(netdev))
4257 		return 0;
4258 
4259 	if (hw->mac.ops.update_mc_addr_list)
4260 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4261 	else
4262 		return -ENOMEM;
4263 
4264 #ifdef CONFIG_PCI_IOV
4265 	ixgbe_restore_vf_multicasts(adapter);
4266 #endif
4267 
4268 	return netdev_mc_count(netdev);
4269 }
4270 
4271 #ifdef CONFIG_PCI_IOV
4272 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4273 {
4274 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4275 	struct ixgbe_hw *hw = &adapter->hw;
4276 	int i;
4277 
4278 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4279 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4280 
4281 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4282 			hw->mac.ops.set_rar(hw, i,
4283 					    mac_table->addr,
4284 					    mac_table->pool,
4285 					    IXGBE_RAH_AV);
4286 		else
4287 			hw->mac.ops.clear_rar(hw, i);
4288 	}
4289 }
4290 
4291 #endif
4292 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4293 {
4294 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4295 	struct ixgbe_hw *hw = &adapter->hw;
4296 	int i;
4297 
4298 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4299 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4300 			continue;
4301 
4302 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4303 
4304 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4305 			hw->mac.ops.set_rar(hw, i,
4306 					    mac_table->addr,
4307 					    mac_table->pool,
4308 					    IXGBE_RAH_AV);
4309 		else
4310 			hw->mac.ops.clear_rar(hw, i);
4311 	}
4312 }
4313 
4314 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4315 {
4316 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4317 	struct ixgbe_hw *hw = &adapter->hw;
4318 	int i;
4319 
4320 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4321 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4322 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4323 	}
4324 
4325 	ixgbe_sync_mac_table(adapter);
4326 }
4327 
4328 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4329 {
4330 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4331 	struct ixgbe_hw *hw = &adapter->hw;
4332 	int i, count = 0;
4333 
4334 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4335 		/* do not count default RAR as available */
4336 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4337 			continue;
4338 
4339 		/* only count unused and addresses that belong to us */
4340 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4341 			if (mac_table->pool != pool)
4342 				continue;
4343 		}
4344 
4345 		count++;
4346 	}
4347 
4348 	return count;
4349 }
4350 
4351 /* this function destroys the first RAR entry */
4352 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4353 {
4354 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4355 	struct ixgbe_hw *hw = &adapter->hw;
4356 
4357 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4358 	mac_table->pool = VMDQ_P(0);
4359 
4360 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4361 
4362 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4363 			    IXGBE_RAH_AV);
4364 }
4365 
4366 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4367 			 const u8 *addr, u16 pool)
4368 {
4369 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4370 	struct ixgbe_hw *hw = &adapter->hw;
4371 	int i;
4372 
4373 	if (is_zero_ether_addr(addr))
4374 		return -EINVAL;
4375 
4376 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4377 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4378 			continue;
4379 
4380 		ether_addr_copy(mac_table->addr, addr);
4381 		mac_table->pool = pool;
4382 
4383 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4384 				    IXGBE_MAC_STATE_IN_USE;
4385 
4386 		ixgbe_sync_mac_table(adapter);
4387 
4388 		return i;
4389 	}
4390 
4391 	return -ENOMEM;
4392 }
4393 
4394 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4395 			 const u8 *addr, u16 pool)
4396 {
4397 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4398 	struct ixgbe_hw *hw = &adapter->hw;
4399 	int i;
4400 
4401 	if (is_zero_ether_addr(addr))
4402 		return -EINVAL;
4403 
4404 	/* search table for addr, if found clear IN_USE flag and sync */
4405 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4406 		/* we can only delete an entry if it is in use */
4407 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4408 			continue;
4409 		/* we only care about entries that belong to the given pool */
4410 		if (mac_table->pool != pool)
4411 			continue;
4412 		/* we only care about a specific MAC address */
4413 		if (!ether_addr_equal(addr, mac_table->addr))
4414 			continue;
4415 
4416 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4417 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4418 
4419 		ixgbe_sync_mac_table(adapter);
4420 
4421 		return 0;
4422 	}
4423 
4424 	return -ENOMEM;
4425 }
4426 /**
4427  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4428  * @netdev: network interface device structure
4429  *
4430  * Writes unicast address list to the RAR table.
4431  * Returns: -ENOMEM on failure/insufficient address space
4432  *                0 on no addresses written
4433  *                X on writing X addresses to the RAR table
4434  **/
4435 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4436 {
4437 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4438 	int count = 0;
4439 
4440 	/* return ENOMEM indicating insufficient memory for addresses */
4441 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter, vfn))
4442 		return -ENOMEM;
4443 
4444 	if (!netdev_uc_empty(netdev)) {
4445 		struct netdev_hw_addr *ha;
4446 		netdev_for_each_uc_addr(ha, netdev) {
4447 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4448 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4449 			count++;
4450 		}
4451 	}
4452 	return count;
4453 }
4454 
4455 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4456 {
4457 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4458 	int ret;
4459 
4460 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4461 
4462 	return min_t(int, ret, 0);
4463 }
4464 
4465 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4466 {
4467 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4468 
4469 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4470 
4471 	return 0;
4472 }
4473 
4474 /**
4475  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4476  * @netdev: network interface device structure
4477  *
4478  * The set_rx_method entry point is called whenever the unicast/multicast
4479  * address list or the network interface flags are updated.  This routine is
4480  * responsible for configuring the hardware for proper unicast, multicast and
4481  * promiscuous mode.
4482  **/
4483 void ixgbe_set_rx_mode(struct net_device *netdev)
4484 {
4485 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4486 	struct ixgbe_hw *hw = &adapter->hw;
4487 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4488 	netdev_features_t features = netdev->features;
4489 	int count;
4490 
4491 	/* Check for Promiscuous and All Multicast modes */
4492 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4493 
4494 	/* set all bits that we expect to always be set */
4495 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4496 	fctrl |= IXGBE_FCTRL_BAM;
4497 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4498 	fctrl |= IXGBE_FCTRL_PMCF;
4499 
4500 	/* clear the bits we are changing the status of */
4501 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4502 	if (netdev->flags & IFF_PROMISC) {
4503 		hw->addr_ctrl.user_set_promisc = true;
4504 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4505 		vmolr |= IXGBE_VMOLR_MPE;
4506 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4507 	} else {
4508 		if (netdev->flags & IFF_ALLMULTI) {
4509 			fctrl |= IXGBE_FCTRL_MPE;
4510 			vmolr |= IXGBE_VMOLR_MPE;
4511 		}
4512 		hw->addr_ctrl.user_set_promisc = false;
4513 	}
4514 
4515 	/*
4516 	 * Write addresses to available RAR registers, if there is not
4517 	 * sufficient space to store all the addresses then enable
4518 	 * unicast promiscuous mode
4519 	 */
4520 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4521 		fctrl |= IXGBE_FCTRL_UPE;
4522 		vmolr |= IXGBE_VMOLR_ROPE;
4523 	}
4524 
4525 	/* Write addresses to the MTA, if the attempt fails
4526 	 * then we should just turn on promiscuous mode so
4527 	 * that we can at least receive multicast traffic
4528 	 */
4529 	count = ixgbe_write_mc_addr_list(netdev);
4530 	if (count < 0) {
4531 		fctrl |= IXGBE_FCTRL_MPE;
4532 		vmolr |= IXGBE_VMOLR_MPE;
4533 	} else if (count) {
4534 		vmolr |= IXGBE_VMOLR_ROMPE;
4535 	}
4536 
4537 	if (hw->mac.type != ixgbe_mac_82598EB) {
4538 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4539 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4540 			   IXGBE_VMOLR_ROPE);
4541 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4542 	}
4543 
4544 	/* This is useful for sniffing bad packets. */
4545 	if (features & NETIF_F_RXALL) {
4546 		/* UPE and MPE will be handled by normal PROMISC logic
4547 		 * in e1000e_set_rx_mode */
4548 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4549 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4550 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4551 
4552 		fctrl &= ~(IXGBE_FCTRL_DPF);
4553 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4554 	}
4555 
4556 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4557 
4558 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4559 		ixgbe_vlan_strip_enable(adapter);
4560 	else
4561 		ixgbe_vlan_strip_disable(adapter);
4562 
4563 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4564 		ixgbe_vlan_promisc_disable(adapter);
4565 	else
4566 		ixgbe_vlan_promisc_enable(adapter);
4567 }
4568 
4569 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4570 {
4571 	int q_idx;
4572 
4573 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4574 		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
4575 		napi_enable(&adapter->q_vector[q_idx]->napi);
4576 	}
4577 }
4578 
4579 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4580 {
4581 	int q_idx;
4582 
4583 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4584 		napi_disable(&adapter->q_vector[q_idx]->napi);
4585 		while (!ixgbe_qv_disable(adapter->q_vector[q_idx])) {
4586 			pr_info("QV %d locked\n", q_idx);
4587 			usleep_range(1000, 20000);
4588 		}
4589 	}
4590 }
4591 
4592 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
4593 {
4594 	struct ixgbe_hw *hw = &adapter->hw;
4595 	u32 vxlanctrl;
4596 
4597 	if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
4598 				IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
4599 		return;
4600 
4601 	vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) && ~mask;
4602 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
4603 
4604 	if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
4605 		adapter->vxlan_port = 0;
4606 
4607 	if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
4608 		adapter->geneve_port = 0;
4609 }
4610 
4611 #ifdef CONFIG_IXGBE_DCB
4612 /**
4613  * ixgbe_configure_dcb - Configure DCB hardware
4614  * @adapter: ixgbe adapter struct
4615  *
4616  * This is called by the driver on open to configure the DCB hardware.
4617  * This is also called by the gennetlink interface when reconfiguring
4618  * the DCB state.
4619  */
4620 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4621 {
4622 	struct ixgbe_hw *hw = &adapter->hw;
4623 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4624 
4625 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4626 		if (hw->mac.type == ixgbe_mac_82598EB)
4627 			netif_set_gso_max_size(adapter->netdev, 65536);
4628 		return;
4629 	}
4630 
4631 	if (hw->mac.type == ixgbe_mac_82598EB)
4632 		netif_set_gso_max_size(adapter->netdev, 32768);
4633 
4634 #ifdef IXGBE_FCOE
4635 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4636 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4637 #endif
4638 
4639 	/* reconfigure the hardware */
4640 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4641 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4642 						DCB_TX_CONFIG);
4643 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4644 						DCB_RX_CONFIG);
4645 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4646 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4647 		ixgbe_dcb_hw_ets(&adapter->hw,
4648 				 adapter->ixgbe_ieee_ets,
4649 				 max_frame);
4650 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4651 					adapter->ixgbe_ieee_pfc->pfc_en,
4652 					adapter->ixgbe_ieee_ets->prio_tc);
4653 	}
4654 
4655 	/* Enable RSS Hash per TC */
4656 	if (hw->mac.type != ixgbe_mac_82598EB) {
4657 		u32 msb = 0;
4658 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4659 
4660 		while (rss_i) {
4661 			msb++;
4662 			rss_i >>= 1;
4663 		}
4664 
4665 		/* write msb to all 8 TCs in one write */
4666 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4667 	}
4668 }
4669 #endif
4670 
4671 /* Additional bittime to account for IXGBE framing */
4672 #define IXGBE_ETH_FRAMING 20
4673 
4674 /**
4675  * ixgbe_hpbthresh - calculate high water mark for flow control
4676  *
4677  * @adapter: board private structure to calculate for
4678  * @pb: packet buffer to calculate
4679  */
4680 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4681 {
4682 	struct ixgbe_hw *hw = &adapter->hw;
4683 	struct net_device *dev = adapter->netdev;
4684 	int link, tc, kb, marker;
4685 	u32 dv_id, rx_pba;
4686 
4687 	/* Calculate max LAN frame size */
4688 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4689 
4690 #ifdef IXGBE_FCOE
4691 	/* FCoE traffic class uses FCOE jumbo frames */
4692 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4693 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4694 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4695 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4696 #endif
4697 
4698 	/* Calculate delay value for device */
4699 	switch (hw->mac.type) {
4700 	case ixgbe_mac_X540:
4701 	case ixgbe_mac_X550:
4702 	case ixgbe_mac_X550EM_x:
4703 	case ixgbe_mac_x550em_a:
4704 		dv_id = IXGBE_DV_X540(link, tc);
4705 		break;
4706 	default:
4707 		dv_id = IXGBE_DV(link, tc);
4708 		break;
4709 	}
4710 
4711 	/* Loopback switch introduces additional latency */
4712 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4713 		dv_id += IXGBE_B2BT(tc);
4714 
4715 	/* Delay value is calculated in bit times convert to KB */
4716 	kb = IXGBE_BT2KB(dv_id);
4717 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4718 
4719 	marker = rx_pba - kb;
4720 
4721 	/* It is possible that the packet buffer is not large enough
4722 	 * to provide required headroom. In this case throw an error
4723 	 * to user and a do the best we can.
4724 	 */
4725 	if (marker < 0) {
4726 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4727 			    "headroom to support flow control."
4728 			    "Decrease MTU or number of traffic classes\n", pb);
4729 		marker = tc + 1;
4730 	}
4731 
4732 	return marker;
4733 }
4734 
4735 /**
4736  * ixgbe_lpbthresh - calculate low water mark for for flow control
4737  *
4738  * @adapter: board private structure to calculate for
4739  * @pb: packet buffer to calculate
4740  */
4741 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4742 {
4743 	struct ixgbe_hw *hw = &adapter->hw;
4744 	struct net_device *dev = adapter->netdev;
4745 	int tc;
4746 	u32 dv_id;
4747 
4748 	/* Calculate max LAN frame size */
4749 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4750 
4751 #ifdef IXGBE_FCOE
4752 	/* FCoE traffic class uses FCOE jumbo frames */
4753 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4754 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4755 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4756 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4757 #endif
4758 
4759 	/* Calculate delay value for device */
4760 	switch (hw->mac.type) {
4761 	case ixgbe_mac_X540:
4762 	case ixgbe_mac_X550:
4763 	case ixgbe_mac_X550EM_x:
4764 	case ixgbe_mac_x550em_a:
4765 		dv_id = IXGBE_LOW_DV_X540(tc);
4766 		break;
4767 	default:
4768 		dv_id = IXGBE_LOW_DV(tc);
4769 		break;
4770 	}
4771 
4772 	/* Delay value is calculated in bit times convert to KB */
4773 	return IXGBE_BT2KB(dv_id);
4774 }
4775 
4776 /*
4777  * ixgbe_pbthresh_setup - calculate and setup high low water marks
4778  */
4779 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4780 {
4781 	struct ixgbe_hw *hw = &adapter->hw;
4782 	int num_tc = netdev_get_num_tc(adapter->netdev);
4783 	int i;
4784 
4785 	if (!num_tc)
4786 		num_tc = 1;
4787 
4788 	for (i = 0; i < num_tc; i++) {
4789 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4790 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4791 
4792 		/* Low water marks must not be larger than high water marks */
4793 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4794 			hw->fc.low_water[i] = 0;
4795 	}
4796 
4797 	for (; i < MAX_TRAFFIC_CLASS; i++)
4798 		hw->fc.high_water[i] = 0;
4799 }
4800 
4801 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4802 {
4803 	struct ixgbe_hw *hw = &adapter->hw;
4804 	int hdrm;
4805 	u8 tc = netdev_get_num_tc(adapter->netdev);
4806 
4807 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4808 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4809 		hdrm = 32 << adapter->fdir_pballoc;
4810 	else
4811 		hdrm = 0;
4812 
4813 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4814 	ixgbe_pbthresh_setup(adapter);
4815 }
4816 
4817 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4818 {
4819 	struct ixgbe_hw *hw = &adapter->hw;
4820 	struct hlist_node *node2;
4821 	struct ixgbe_fdir_filter *filter;
4822 
4823 	spin_lock(&adapter->fdir_perfect_lock);
4824 
4825 	if (!hlist_empty(&adapter->fdir_filter_list))
4826 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4827 
4828 	hlist_for_each_entry_safe(filter, node2,
4829 				  &adapter->fdir_filter_list, fdir_node) {
4830 		ixgbe_fdir_write_perfect_filter_82599(hw,
4831 				&filter->filter,
4832 				filter->sw_idx,
4833 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4834 				IXGBE_FDIR_DROP_QUEUE :
4835 				adapter->rx_ring[filter->action]->reg_idx);
4836 	}
4837 
4838 	spin_unlock(&adapter->fdir_perfect_lock);
4839 }
4840 
4841 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4842 				      struct ixgbe_adapter *adapter)
4843 {
4844 	struct ixgbe_hw *hw = &adapter->hw;
4845 	u32 vmolr;
4846 
4847 	/* No unicast promiscuous support for VMDQ devices. */
4848 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4849 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4850 
4851 	/* clear the affected bit */
4852 	vmolr &= ~IXGBE_VMOLR_MPE;
4853 
4854 	if (dev->flags & IFF_ALLMULTI) {
4855 		vmolr |= IXGBE_VMOLR_MPE;
4856 	} else {
4857 		vmolr |= IXGBE_VMOLR_ROMPE;
4858 		hw->mac.ops.update_mc_addr_list(hw, dev);
4859 	}
4860 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4861 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4862 }
4863 
4864 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4865 {
4866 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4867 	int rss_i = adapter->num_rx_queues_per_pool;
4868 	struct ixgbe_hw *hw = &adapter->hw;
4869 	u16 pool = vadapter->pool;
4870 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4871 		      IXGBE_PSRTYPE_UDPHDR |
4872 		      IXGBE_PSRTYPE_IPV4HDR |
4873 		      IXGBE_PSRTYPE_L2HDR |
4874 		      IXGBE_PSRTYPE_IPV6HDR;
4875 
4876 	if (hw->mac.type == ixgbe_mac_82598EB)
4877 		return;
4878 
4879 	if (rss_i > 3)
4880 		psrtype |= 2u << 29;
4881 	else if (rss_i > 1)
4882 		psrtype |= 1u << 29;
4883 
4884 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4885 }
4886 
4887 /**
4888  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4889  * @rx_ring: ring to free buffers from
4890  **/
4891 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4892 {
4893 	struct device *dev = rx_ring->dev;
4894 	unsigned long size;
4895 	u16 i;
4896 
4897 	/* ring already cleared, nothing to do */
4898 	if (!rx_ring->rx_buffer_info)
4899 		return;
4900 
4901 	/* Free all the Rx ring sk_buffs */
4902 	for (i = 0; i < rx_ring->count; i++) {
4903 		struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4904 
4905 		if (rx_buffer->skb) {
4906 			struct sk_buff *skb = rx_buffer->skb;
4907 			if (IXGBE_CB(skb)->page_released)
4908 				dma_unmap_page(dev,
4909 					       IXGBE_CB(skb)->dma,
4910 					       ixgbe_rx_bufsz(rx_ring),
4911 					       DMA_FROM_DEVICE);
4912 			dev_kfree_skb(skb);
4913 			rx_buffer->skb = NULL;
4914 		}
4915 
4916 		if (!rx_buffer->page)
4917 			continue;
4918 
4919 		dma_unmap_page(dev, rx_buffer->dma,
4920 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
4921 		__free_pages(rx_buffer->page, ixgbe_rx_pg_order(rx_ring));
4922 
4923 		rx_buffer->page = NULL;
4924 	}
4925 
4926 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4927 	memset(rx_ring->rx_buffer_info, 0, size);
4928 
4929 	/* Zero out the descriptor ring */
4930 	memset(rx_ring->desc, 0, rx_ring->size);
4931 
4932 	rx_ring->next_to_alloc = 0;
4933 	rx_ring->next_to_clean = 0;
4934 	rx_ring->next_to_use = 0;
4935 }
4936 
4937 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4938 				   struct ixgbe_ring *rx_ring)
4939 {
4940 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4941 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4942 
4943 	/* shutdown specific queue receive and wait for dma to settle */
4944 	ixgbe_disable_rx_queue(adapter, rx_ring);
4945 	usleep_range(10000, 20000);
4946 	ixgbe_irq_disable_queues(adapter, BIT_ULL(index));
4947 	ixgbe_clean_rx_ring(rx_ring);
4948 	rx_ring->l2_accel_priv = NULL;
4949 }
4950 
4951 static int ixgbe_fwd_ring_down(struct net_device *vdev,
4952 			       struct ixgbe_fwd_adapter *accel)
4953 {
4954 	struct ixgbe_adapter *adapter = accel->real_adapter;
4955 	unsigned int rxbase = accel->rx_base_queue;
4956 	unsigned int txbase = accel->tx_base_queue;
4957 	int i;
4958 
4959 	netif_tx_stop_all_queues(vdev);
4960 
4961 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4962 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4963 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4964 	}
4965 
4966 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4967 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4968 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4969 	}
4970 
4971 
4972 	return 0;
4973 }
4974 
4975 static int ixgbe_fwd_ring_up(struct net_device *vdev,
4976 			     struct ixgbe_fwd_adapter *accel)
4977 {
4978 	struct ixgbe_adapter *adapter = accel->real_adapter;
4979 	unsigned int rxbase, txbase, queues;
4980 	int i, baseq, err = 0;
4981 
4982 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4983 		return 0;
4984 
4985 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4986 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4987 		   accel->pool, adapter->num_rx_pools,
4988 		   baseq, baseq + adapter->num_rx_queues_per_pool,
4989 		   adapter->fwd_bitmask);
4990 
4991 	accel->netdev = vdev;
4992 	accel->rx_base_queue = rxbase = baseq;
4993 	accel->tx_base_queue = txbase = baseq;
4994 
4995 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4996 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4997 
4998 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4999 		adapter->rx_ring[rxbase + i]->netdev = vdev;
5000 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
5001 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
5002 	}
5003 
5004 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
5005 		adapter->tx_ring[txbase + i]->netdev = vdev;
5006 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
5007 	}
5008 
5009 	queues = min_t(unsigned int,
5010 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
5011 	err = netif_set_real_num_tx_queues(vdev, queues);
5012 	if (err)
5013 		goto fwd_queue_err;
5014 
5015 	err = netif_set_real_num_rx_queues(vdev, queues);
5016 	if (err)
5017 		goto fwd_queue_err;
5018 
5019 	if (is_valid_ether_addr(vdev->dev_addr))
5020 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
5021 
5022 	ixgbe_fwd_psrtype(accel);
5023 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
5024 	return err;
5025 fwd_queue_err:
5026 	ixgbe_fwd_ring_down(vdev, accel);
5027 	return err;
5028 }
5029 
5030 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5031 {
5032 	struct net_device *upper;
5033 	struct list_head *iter;
5034 	int err;
5035 
5036 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5037 		if (netif_is_macvlan(upper)) {
5038 			struct macvlan_dev *dfwd = netdev_priv(upper);
5039 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
5040 
5041 			if (dfwd->fwd_priv) {
5042 				err = ixgbe_fwd_ring_up(upper, vadapter);
5043 				if (err)
5044 					continue;
5045 			}
5046 		}
5047 	}
5048 }
5049 
5050 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5051 {
5052 	struct ixgbe_hw *hw = &adapter->hw;
5053 
5054 	ixgbe_configure_pb(adapter);
5055 #ifdef CONFIG_IXGBE_DCB
5056 	ixgbe_configure_dcb(adapter);
5057 #endif
5058 	/*
5059 	 * We must restore virtualization before VLANs or else
5060 	 * the VLVF registers will not be populated
5061 	 */
5062 	ixgbe_configure_virtualization(adapter);
5063 
5064 	ixgbe_set_rx_mode(adapter->netdev);
5065 	ixgbe_restore_vlan(adapter);
5066 
5067 	switch (hw->mac.type) {
5068 	case ixgbe_mac_82599EB:
5069 	case ixgbe_mac_X540:
5070 		hw->mac.ops.disable_rx_buff(hw);
5071 		break;
5072 	default:
5073 		break;
5074 	}
5075 
5076 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5077 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5078 						adapter->fdir_pballoc);
5079 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5080 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5081 					      adapter->fdir_pballoc);
5082 		ixgbe_fdir_filter_restore(adapter);
5083 	}
5084 
5085 	switch (hw->mac.type) {
5086 	case ixgbe_mac_82599EB:
5087 	case ixgbe_mac_X540:
5088 		hw->mac.ops.enable_rx_buff(hw);
5089 		break;
5090 	default:
5091 		break;
5092 	}
5093 
5094 #ifdef CONFIG_IXGBE_DCA
5095 	/* configure DCA */
5096 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5097 		ixgbe_setup_dca(adapter);
5098 #endif /* CONFIG_IXGBE_DCA */
5099 
5100 #ifdef IXGBE_FCOE
5101 	/* configure FCoE L2 filters, redirection table, and Rx control */
5102 	ixgbe_configure_fcoe(adapter);
5103 
5104 #endif /* IXGBE_FCOE */
5105 	ixgbe_configure_tx(adapter);
5106 	ixgbe_configure_rx(adapter);
5107 	ixgbe_configure_dfwd(adapter);
5108 }
5109 
5110 /**
5111  * ixgbe_sfp_link_config - set up SFP+ link
5112  * @adapter: pointer to private adapter struct
5113  **/
5114 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5115 {
5116 	/*
5117 	 * We are assuming the worst case scenario here, and that
5118 	 * is that an SFP was inserted/removed after the reset
5119 	 * but before SFP detection was enabled.  As such the best
5120 	 * solution is to just start searching as soon as we start
5121 	 */
5122 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5123 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5124 
5125 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5126 	adapter->sfp_poll_time = 0;
5127 }
5128 
5129 /**
5130  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5131  * @hw: pointer to private hardware struct
5132  *
5133  * Returns 0 on success, negative on failure
5134  **/
5135 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5136 {
5137 	u32 speed;
5138 	bool autoneg, link_up = false;
5139 	int ret = IXGBE_ERR_LINK_SETUP;
5140 
5141 	if (hw->mac.ops.check_link)
5142 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5143 
5144 	if (ret)
5145 		return ret;
5146 
5147 	speed = hw->phy.autoneg_advertised;
5148 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5149 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5150 							&autoneg);
5151 	if (ret)
5152 		return ret;
5153 
5154 	if (hw->mac.ops.setup_link)
5155 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5156 
5157 	return ret;
5158 }
5159 
5160 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5161 {
5162 	struct ixgbe_hw *hw = &adapter->hw;
5163 	u32 gpie = 0;
5164 
5165 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5166 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5167 		       IXGBE_GPIE_OCD;
5168 		gpie |= IXGBE_GPIE_EIAME;
5169 		/*
5170 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5171 		 * this saves a register write for every interrupt
5172 		 */
5173 		switch (hw->mac.type) {
5174 		case ixgbe_mac_82598EB:
5175 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5176 			break;
5177 		case ixgbe_mac_82599EB:
5178 		case ixgbe_mac_X540:
5179 		case ixgbe_mac_X550:
5180 		case ixgbe_mac_X550EM_x:
5181 		case ixgbe_mac_x550em_a:
5182 		default:
5183 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5184 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5185 			break;
5186 		}
5187 	} else {
5188 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5189 		 * specifically only auto mask tx and rx interrupts */
5190 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5191 	}
5192 
5193 	/* XXX: to interrupt immediately for EICS writes, enable this */
5194 	/* gpie |= IXGBE_GPIE_EIMEN; */
5195 
5196 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5197 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5198 
5199 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5200 		case IXGBE_82599_VMDQ_8Q_MASK:
5201 			gpie |= IXGBE_GPIE_VTMODE_16;
5202 			break;
5203 		case IXGBE_82599_VMDQ_4Q_MASK:
5204 			gpie |= IXGBE_GPIE_VTMODE_32;
5205 			break;
5206 		default:
5207 			gpie |= IXGBE_GPIE_VTMODE_64;
5208 			break;
5209 		}
5210 	}
5211 
5212 	/* Enable Thermal over heat sensor interrupt */
5213 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5214 		switch (adapter->hw.mac.type) {
5215 		case ixgbe_mac_82599EB:
5216 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5217 			break;
5218 		default:
5219 			break;
5220 		}
5221 	}
5222 
5223 	/* Enable fan failure interrupt */
5224 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5225 		gpie |= IXGBE_SDP1_GPIEN(hw);
5226 
5227 	switch (hw->mac.type) {
5228 	case ixgbe_mac_82599EB:
5229 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5230 		break;
5231 	case ixgbe_mac_X550EM_x:
5232 	case ixgbe_mac_x550em_a:
5233 		gpie |= IXGBE_SDP0_GPIEN_X540;
5234 		break;
5235 	default:
5236 		break;
5237 	}
5238 
5239 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5240 }
5241 
5242 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5243 {
5244 	struct ixgbe_hw *hw = &adapter->hw;
5245 	int err;
5246 	u32 ctrl_ext;
5247 
5248 	ixgbe_get_hw_control(adapter);
5249 	ixgbe_setup_gpie(adapter);
5250 
5251 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5252 		ixgbe_configure_msix(adapter);
5253 	else
5254 		ixgbe_configure_msi_and_legacy(adapter);
5255 
5256 	/* enable the optics for 82599 SFP+ fiber */
5257 	if (hw->mac.ops.enable_tx_laser)
5258 		hw->mac.ops.enable_tx_laser(hw);
5259 
5260 	if (hw->phy.ops.set_phy_power)
5261 		hw->phy.ops.set_phy_power(hw, true);
5262 
5263 	smp_mb__before_atomic();
5264 	clear_bit(__IXGBE_DOWN, &adapter->state);
5265 	ixgbe_napi_enable_all(adapter);
5266 
5267 	if (ixgbe_is_sfp(hw)) {
5268 		ixgbe_sfp_link_config(adapter);
5269 	} else {
5270 		err = ixgbe_non_sfp_link_config(hw);
5271 		if (err)
5272 			e_err(probe, "link_config FAILED %d\n", err);
5273 	}
5274 
5275 	/* clear any pending interrupts, may auto mask */
5276 	IXGBE_READ_REG(hw, IXGBE_EICR);
5277 	ixgbe_irq_enable(adapter, true, true);
5278 
5279 	/*
5280 	 * If this adapter has a fan, check to see if we had a failure
5281 	 * before we enabled the interrupt.
5282 	 */
5283 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5284 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5285 		if (esdp & IXGBE_ESDP_SDP1)
5286 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5287 	}
5288 
5289 	/* bring the link up in the watchdog, this could race with our first
5290 	 * link up interrupt but shouldn't be a problem */
5291 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5292 	adapter->link_check_timeout = jiffies;
5293 	mod_timer(&adapter->service_timer, jiffies);
5294 
5295 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5296 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5297 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5298 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5299 }
5300 
5301 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5302 {
5303 	WARN_ON(in_interrupt());
5304 	/* put off any impending NetWatchDogTimeout */
5305 	netif_trans_update(adapter->netdev);
5306 
5307 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5308 		usleep_range(1000, 2000);
5309 	ixgbe_down(adapter);
5310 	/*
5311 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5312 	 * back up to give the VFs time to respond to the reset.  The
5313 	 * two second wait is based upon the watchdog timer cycle in
5314 	 * the VF driver.
5315 	 */
5316 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5317 		msleep(2000);
5318 	ixgbe_up(adapter);
5319 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5320 }
5321 
5322 void ixgbe_up(struct ixgbe_adapter *adapter)
5323 {
5324 	/* hardware has been reset, we need to reload some things */
5325 	ixgbe_configure(adapter);
5326 
5327 	ixgbe_up_complete(adapter);
5328 }
5329 
5330 void ixgbe_reset(struct ixgbe_adapter *adapter)
5331 {
5332 	struct ixgbe_hw *hw = &adapter->hw;
5333 	struct net_device *netdev = adapter->netdev;
5334 	int err;
5335 
5336 	if (ixgbe_removed(hw->hw_addr))
5337 		return;
5338 	/* lock SFP init bit to prevent race conditions with the watchdog */
5339 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5340 		usleep_range(1000, 2000);
5341 
5342 	/* clear all SFP and link config related flags while holding SFP_INIT */
5343 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5344 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5345 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5346 
5347 	err = hw->mac.ops.init_hw(hw);
5348 	switch (err) {
5349 	case 0:
5350 	case IXGBE_ERR_SFP_NOT_PRESENT:
5351 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5352 		break;
5353 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5354 		e_dev_err("master disable timed out\n");
5355 		break;
5356 	case IXGBE_ERR_EEPROM_VERSION:
5357 		/* We are running on a pre-production device, log a warning */
5358 		e_dev_warn("This device is a pre-production adapter/LOM. "
5359 			   "Please be aware there may be issues associated with "
5360 			   "your hardware.  If you are experiencing problems "
5361 			   "please contact your Intel or hardware "
5362 			   "representative who provided you with this "
5363 			   "hardware.\n");
5364 		break;
5365 	default:
5366 		e_dev_err("Hardware Error: %d\n", err);
5367 	}
5368 
5369 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5370 
5371 	/* flush entries out of MAC table */
5372 	ixgbe_flush_sw_mac_table(adapter);
5373 	__dev_uc_unsync(netdev, NULL);
5374 
5375 	/* do not flush user set addresses */
5376 	ixgbe_mac_set_default_filter(adapter);
5377 
5378 	/* update SAN MAC vmdq pool selection */
5379 	if (hw->mac.san_mac_rar_index)
5380 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5381 
5382 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5383 		ixgbe_ptp_reset(adapter);
5384 
5385 	if (hw->phy.ops.set_phy_power) {
5386 		if (!netif_running(adapter->netdev) && !adapter->wol)
5387 			hw->phy.ops.set_phy_power(hw, false);
5388 		else
5389 			hw->phy.ops.set_phy_power(hw, true);
5390 	}
5391 }
5392 
5393 /**
5394  * ixgbe_clean_tx_ring - Free Tx Buffers
5395  * @tx_ring: ring to be cleaned
5396  **/
5397 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5398 {
5399 	struct ixgbe_tx_buffer *tx_buffer_info;
5400 	unsigned long size;
5401 	u16 i;
5402 
5403 	/* ring already cleared, nothing to do */
5404 	if (!tx_ring->tx_buffer_info)
5405 		return;
5406 
5407 	/* Free all the Tx ring sk_buffs */
5408 	for (i = 0; i < tx_ring->count; i++) {
5409 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
5410 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
5411 	}
5412 
5413 	netdev_tx_reset_queue(txring_txq(tx_ring));
5414 
5415 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5416 	memset(tx_ring->tx_buffer_info, 0, size);
5417 
5418 	/* Zero out the descriptor ring */
5419 	memset(tx_ring->desc, 0, tx_ring->size);
5420 
5421 	tx_ring->next_to_use = 0;
5422 	tx_ring->next_to_clean = 0;
5423 }
5424 
5425 /**
5426  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5427  * @adapter: board private structure
5428  **/
5429 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5430 {
5431 	int i;
5432 
5433 	for (i = 0; i < adapter->num_rx_queues; i++)
5434 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5435 }
5436 
5437 /**
5438  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5439  * @adapter: board private structure
5440  **/
5441 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5442 {
5443 	int i;
5444 
5445 	for (i = 0; i < adapter->num_tx_queues; i++)
5446 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5447 }
5448 
5449 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5450 {
5451 	struct hlist_node *node2;
5452 	struct ixgbe_fdir_filter *filter;
5453 
5454 	spin_lock(&adapter->fdir_perfect_lock);
5455 
5456 	hlist_for_each_entry_safe(filter, node2,
5457 				  &adapter->fdir_filter_list, fdir_node) {
5458 		hlist_del(&filter->fdir_node);
5459 		kfree(filter);
5460 	}
5461 	adapter->fdir_filter_count = 0;
5462 
5463 	spin_unlock(&adapter->fdir_perfect_lock);
5464 }
5465 
5466 void ixgbe_down(struct ixgbe_adapter *adapter)
5467 {
5468 	struct net_device *netdev = adapter->netdev;
5469 	struct ixgbe_hw *hw = &adapter->hw;
5470 	struct net_device *upper;
5471 	struct list_head *iter;
5472 	int i;
5473 
5474 	/* signal that we are down to the interrupt handler */
5475 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5476 		return; /* do nothing if already down */
5477 
5478 	/* disable receives */
5479 	hw->mac.ops.disable_rx(hw);
5480 
5481 	/* disable all enabled rx queues */
5482 	for (i = 0; i < adapter->num_rx_queues; i++)
5483 		/* this call also flushes the previous write */
5484 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5485 
5486 	usleep_range(10000, 20000);
5487 
5488 	netif_tx_stop_all_queues(netdev);
5489 
5490 	/* call carrier off first to avoid false dev_watchdog timeouts */
5491 	netif_carrier_off(netdev);
5492 	netif_tx_disable(netdev);
5493 
5494 	/* disable any upper devices */
5495 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5496 		if (netif_is_macvlan(upper)) {
5497 			struct macvlan_dev *vlan = netdev_priv(upper);
5498 
5499 			if (vlan->fwd_priv) {
5500 				netif_tx_stop_all_queues(upper);
5501 				netif_carrier_off(upper);
5502 				netif_tx_disable(upper);
5503 			}
5504 		}
5505 	}
5506 
5507 	ixgbe_irq_disable(adapter);
5508 
5509 	ixgbe_napi_disable_all(adapter);
5510 
5511 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
5512 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5513 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5514 
5515 	del_timer_sync(&adapter->service_timer);
5516 
5517 	if (adapter->num_vfs) {
5518 		/* Clear EITR Select mapping */
5519 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5520 
5521 		/* Mark all the VFs as inactive */
5522 		for (i = 0 ; i < adapter->num_vfs; i++)
5523 			adapter->vfinfo[i].clear_to_send = false;
5524 
5525 		/* ping all the active vfs to let them know we are going down */
5526 		ixgbe_ping_all_vfs(adapter);
5527 
5528 		/* Disable all VFTE/VFRE TX/RX */
5529 		ixgbe_disable_tx_rx(adapter);
5530 	}
5531 
5532 	/* disable transmits in the hardware now that interrupts are off */
5533 	for (i = 0; i < adapter->num_tx_queues; i++) {
5534 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5535 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5536 	}
5537 
5538 	/* Disable the Tx DMA engine on 82599 and later MAC */
5539 	switch (hw->mac.type) {
5540 	case ixgbe_mac_82599EB:
5541 	case ixgbe_mac_X540:
5542 	case ixgbe_mac_X550:
5543 	case ixgbe_mac_X550EM_x:
5544 	case ixgbe_mac_x550em_a:
5545 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5546 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5547 				 ~IXGBE_DMATXCTL_TE));
5548 		break;
5549 	default:
5550 		break;
5551 	}
5552 
5553 	if (!pci_channel_offline(adapter->pdev))
5554 		ixgbe_reset(adapter);
5555 
5556 	/* power down the optics for 82599 SFP+ fiber */
5557 	if (hw->mac.ops.disable_tx_laser)
5558 		hw->mac.ops.disable_tx_laser(hw);
5559 
5560 	ixgbe_clean_all_tx_rings(adapter);
5561 	ixgbe_clean_all_rx_rings(adapter);
5562 }
5563 
5564 /**
5565  * ixgbe_tx_timeout - Respond to a Tx Hang
5566  * @netdev: network interface device structure
5567  **/
5568 static void ixgbe_tx_timeout(struct net_device *netdev)
5569 {
5570 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5571 
5572 	/* Do the reset outside of interrupt context */
5573 	ixgbe_tx_timeout_reset(adapter);
5574 }
5575 
5576 #ifdef CONFIG_IXGBE_DCB
5577 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
5578 {
5579 	struct ixgbe_hw *hw = &adapter->hw;
5580 	struct tc_configuration *tc;
5581 	int j;
5582 
5583 	switch (hw->mac.type) {
5584 	case ixgbe_mac_82598EB:
5585 	case ixgbe_mac_82599EB:
5586 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5587 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5588 		break;
5589 	case ixgbe_mac_X540:
5590 	case ixgbe_mac_X550:
5591 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5592 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5593 		break;
5594 	case ixgbe_mac_X550EM_x:
5595 	case ixgbe_mac_x550em_a:
5596 	default:
5597 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
5598 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
5599 		break;
5600 	}
5601 
5602 	/* Configure DCB traffic classes */
5603 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5604 		tc = &adapter->dcb_cfg.tc_config[j];
5605 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5606 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5607 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5608 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5609 		tc->dcb_pfc = pfc_disabled;
5610 	}
5611 
5612 	/* Initialize default user to priority mapping, UPx->TC0 */
5613 	tc = &adapter->dcb_cfg.tc_config[0];
5614 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5615 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5616 
5617 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5618 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5619 	adapter->dcb_cfg.pfc_mode_enable = false;
5620 	adapter->dcb_set_bitmap = 0x00;
5621 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
5622 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5623 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5624 	       sizeof(adapter->temp_dcb_cfg));
5625 }
5626 #endif
5627 
5628 /**
5629  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5630  * @adapter: board private structure to initialize
5631  *
5632  * ixgbe_sw_init initializes the Adapter private data structure.
5633  * Fields are initialized based on PCI device information and
5634  * OS network device settings (MTU size).
5635  **/
5636 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
5637 {
5638 	struct ixgbe_hw *hw = &adapter->hw;
5639 	struct pci_dev *pdev = adapter->pdev;
5640 	unsigned int rss, fdir;
5641 	u32 fwsm;
5642 	int i;
5643 
5644 	/* PCI config space info */
5645 
5646 	hw->vendor_id = pdev->vendor;
5647 	hw->device_id = pdev->device;
5648 	hw->revision_id = pdev->revision;
5649 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5650 	hw->subsystem_device_id = pdev->subsystem_device;
5651 
5652 	/* Set common capability flags and settings */
5653 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5654 	adapter->ring_feature[RING_F_RSS].limit = rss;
5655 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5656 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5657 	adapter->atr_sample_rate = 20;
5658 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5659 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5660 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5661 #ifdef CONFIG_IXGBE_DCA
5662 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5663 #endif
5664 #ifdef CONFIG_IXGBE_DCB
5665 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
5666 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
5667 #endif
5668 #ifdef IXGBE_FCOE
5669 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5670 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5671 #ifdef CONFIG_IXGBE_DCB
5672 	/* Default traffic class to use for FCoE */
5673 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5674 #endif /* CONFIG_IXGBE_DCB */
5675 #endif /* IXGBE_FCOE */
5676 
5677 	/* initialize static ixgbe jump table entries */
5678 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
5679 					  GFP_KERNEL);
5680 	if (!adapter->jump_tables[0])
5681 		return -ENOMEM;
5682 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
5683 
5684 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
5685 		adapter->jump_tables[i] = NULL;
5686 
5687 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5688 				     hw->mac.num_rar_entries,
5689 				     GFP_ATOMIC);
5690 	if (!adapter->mac_table)
5691 		return -ENOMEM;
5692 
5693 	/* Set MAC specific capability flags and exceptions */
5694 	switch (hw->mac.type) {
5695 	case ixgbe_mac_82598EB:
5696 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5697 
5698 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5699 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5700 
5701 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5702 		adapter->ring_feature[RING_F_FDIR].limit = 0;
5703 		adapter->atr_sample_rate = 0;
5704 		adapter->fdir_pballoc = 0;
5705 #ifdef IXGBE_FCOE
5706 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5707 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5708 #ifdef CONFIG_IXGBE_DCB
5709 		adapter->fcoe.up = 0;
5710 #endif /* IXGBE_DCB */
5711 #endif /* IXGBE_FCOE */
5712 		break;
5713 	case ixgbe_mac_82599EB:
5714 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5715 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5716 		break;
5717 	case ixgbe_mac_X540:
5718 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
5719 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5720 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5721 		break;
5722 	case ixgbe_mac_x550em_a:
5723 		adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
5724 	/* fall through */
5725 	case ixgbe_mac_X550EM_x:
5726 #ifdef CONFIG_IXGBE_DCB
5727 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
5728 #endif
5729 #ifdef IXGBE_FCOE
5730 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5731 #ifdef CONFIG_IXGBE_DCB
5732 		adapter->fcoe.up = 0;
5733 #endif /* IXGBE_DCB */
5734 #endif /* IXGBE_FCOE */
5735 	/* Fall Through */
5736 	case ixgbe_mac_X550:
5737 #ifdef CONFIG_IXGBE_DCA
5738 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5739 #endif
5740 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
5741 		break;
5742 	default:
5743 		break;
5744 	}
5745 
5746 #ifdef IXGBE_FCOE
5747 	/* FCoE support exists, always init the FCoE lock */
5748 	spin_lock_init(&adapter->fcoe.lock);
5749 
5750 #endif
5751 	/* n-tuple support exists, always init our spinlock */
5752 	spin_lock_init(&adapter->fdir_perfect_lock);
5753 
5754 #ifdef CONFIG_IXGBE_DCB
5755 	ixgbe_init_dcb(adapter);
5756 #endif
5757 
5758 	/* default flow control settings */
5759 	hw->fc.requested_mode = ixgbe_fc_full;
5760 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5761 	ixgbe_pbthresh_setup(adapter);
5762 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5763 	hw->fc.send_xon = true;
5764 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5765 
5766 #ifdef CONFIG_PCI_IOV
5767 	if (max_vfs > 0)
5768 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5769 
5770 	/* assign number of SR-IOV VFs */
5771 	if (hw->mac.type != ixgbe_mac_82598EB) {
5772 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5773 			adapter->num_vfs = 0;
5774 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5775 		} else {
5776 			adapter->num_vfs = max_vfs;
5777 		}
5778 	}
5779 #endif /* CONFIG_PCI_IOV */
5780 
5781 	/* enable itr by default in dynamic mode */
5782 	adapter->rx_itr_setting = 1;
5783 	adapter->tx_itr_setting = 1;
5784 
5785 	/* set default ring sizes */
5786 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5787 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5788 
5789 	/* set default work limits */
5790 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5791 
5792 	/* initialize eeprom parameters */
5793 	if (ixgbe_init_eeprom_params_generic(hw)) {
5794 		e_dev_err("EEPROM initialization failed\n");
5795 		return -EIO;
5796 	}
5797 
5798 	/* PF holds first pool slot */
5799 	set_bit(0, &adapter->fwd_bitmask);
5800 	set_bit(__IXGBE_DOWN, &adapter->state);
5801 
5802 	return 0;
5803 }
5804 
5805 /**
5806  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5807  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5808  *
5809  * Return 0 on success, negative on failure
5810  **/
5811 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5812 {
5813 	struct device *dev = tx_ring->dev;
5814 	int orig_node = dev_to_node(dev);
5815 	int ring_node = -1;
5816 	int size;
5817 
5818 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5819 
5820 	if (tx_ring->q_vector)
5821 		ring_node = tx_ring->q_vector->numa_node;
5822 
5823 	tx_ring->tx_buffer_info = vzalloc_node(size, ring_node);
5824 	if (!tx_ring->tx_buffer_info)
5825 		tx_ring->tx_buffer_info = vzalloc(size);
5826 	if (!tx_ring->tx_buffer_info)
5827 		goto err;
5828 
5829 	u64_stats_init(&tx_ring->syncp);
5830 
5831 	/* round up to nearest 4K */
5832 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5833 	tx_ring->size = ALIGN(tx_ring->size, 4096);
5834 
5835 	set_dev_node(dev, ring_node);
5836 	tx_ring->desc = dma_alloc_coherent(dev,
5837 					   tx_ring->size,
5838 					   &tx_ring->dma,
5839 					   GFP_KERNEL);
5840 	set_dev_node(dev, orig_node);
5841 	if (!tx_ring->desc)
5842 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5843 						   &tx_ring->dma, GFP_KERNEL);
5844 	if (!tx_ring->desc)
5845 		goto err;
5846 
5847 	tx_ring->next_to_use = 0;
5848 	tx_ring->next_to_clean = 0;
5849 	return 0;
5850 
5851 err:
5852 	vfree(tx_ring->tx_buffer_info);
5853 	tx_ring->tx_buffer_info = NULL;
5854 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5855 	return -ENOMEM;
5856 }
5857 
5858 /**
5859  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5860  * @adapter: board private structure
5861  *
5862  * If this function returns with an error, then it's possible one or
5863  * more of the rings is populated (while the rest are not).  It is the
5864  * callers duty to clean those orphaned rings.
5865  *
5866  * Return 0 on success, negative on failure
5867  **/
5868 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5869 {
5870 	int i, err = 0;
5871 
5872 	for (i = 0; i < adapter->num_tx_queues; i++) {
5873 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5874 		if (!err)
5875 			continue;
5876 
5877 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5878 		goto err_setup_tx;
5879 	}
5880 
5881 	return 0;
5882 err_setup_tx:
5883 	/* rewind the index freeing the rings as we go */
5884 	while (i--)
5885 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5886 	return err;
5887 }
5888 
5889 /**
5890  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5891  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5892  *
5893  * Returns 0 on success, negative on failure
5894  **/
5895 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5896 {
5897 	struct device *dev = rx_ring->dev;
5898 	int orig_node = dev_to_node(dev);
5899 	int ring_node = -1;
5900 	int size;
5901 
5902 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5903 
5904 	if (rx_ring->q_vector)
5905 		ring_node = rx_ring->q_vector->numa_node;
5906 
5907 	rx_ring->rx_buffer_info = vzalloc_node(size, ring_node);
5908 	if (!rx_ring->rx_buffer_info)
5909 		rx_ring->rx_buffer_info = vzalloc(size);
5910 	if (!rx_ring->rx_buffer_info)
5911 		goto err;
5912 
5913 	u64_stats_init(&rx_ring->syncp);
5914 
5915 	/* Round up to nearest 4K */
5916 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5917 	rx_ring->size = ALIGN(rx_ring->size, 4096);
5918 
5919 	set_dev_node(dev, ring_node);
5920 	rx_ring->desc = dma_alloc_coherent(dev,
5921 					   rx_ring->size,
5922 					   &rx_ring->dma,
5923 					   GFP_KERNEL);
5924 	set_dev_node(dev, orig_node);
5925 	if (!rx_ring->desc)
5926 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5927 						   &rx_ring->dma, GFP_KERNEL);
5928 	if (!rx_ring->desc)
5929 		goto err;
5930 
5931 	rx_ring->next_to_clean = 0;
5932 	rx_ring->next_to_use = 0;
5933 
5934 	return 0;
5935 err:
5936 	vfree(rx_ring->rx_buffer_info);
5937 	rx_ring->rx_buffer_info = NULL;
5938 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5939 	return -ENOMEM;
5940 }
5941 
5942 /**
5943  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5944  * @adapter: board private structure
5945  *
5946  * If this function returns with an error, then it's possible one or
5947  * more of the rings is populated (while the rest are not).  It is the
5948  * callers duty to clean those orphaned rings.
5949  *
5950  * Return 0 on success, negative on failure
5951  **/
5952 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5953 {
5954 	int i, err = 0;
5955 
5956 	for (i = 0; i < adapter->num_rx_queues; i++) {
5957 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5958 		if (!err)
5959 			continue;
5960 
5961 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5962 		goto err_setup_rx;
5963 	}
5964 
5965 #ifdef IXGBE_FCOE
5966 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5967 	if (!err)
5968 #endif
5969 		return 0;
5970 err_setup_rx:
5971 	/* rewind the index freeing the rings as we go */
5972 	while (i--)
5973 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5974 	return err;
5975 }
5976 
5977 /**
5978  * ixgbe_free_tx_resources - Free Tx Resources per Queue
5979  * @tx_ring: Tx descriptor ring for a specific queue
5980  *
5981  * Free all transmit software resources
5982  **/
5983 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5984 {
5985 	ixgbe_clean_tx_ring(tx_ring);
5986 
5987 	vfree(tx_ring->tx_buffer_info);
5988 	tx_ring->tx_buffer_info = NULL;
5989 
5990 	/* if not set, then don't free */
5991 	if (!tx_ring->desc)
5992 		return;
5993 
5994 	dma_free_coherent(tx_ring->dev, tx_ring->size,
5995 			  tx_ring->desc, tx_ring->dma);
5996 
5997 	tx_ring->desc = NULL;
5998 }
5999 
6000 /**
6001  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6002  * @adapter: board private structure
6003  *
6004  * Free all transmit software resources
6005  **/
6006 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6007 {
6008 	int i;
6009 
6010 	for (i = 0; i < adapter->num_tx_queues; i++)
6011 		if (adapter->tx_ring[i]->desc)
6012 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6013 }
6014 
6015 /**
6016  * ixgbe_free_rx_resources - Free Rx Resources
6017  * @rx_ring: ring to clean the resources from
6018  *
6019  * Free all receive software resources
6020  **/
6021 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6022 {
6023 	ixgbe_clean_rx_ring(rx_ring);
6024 
6025 	vfree(rx_ring->rx_buffer_info);
6026 	rx_ring->rx_buffer_info = NULL;
6027 
6028 	/* if not set, then don't free */
6029 	if (!rx_ring->desc)
6030 		return;
6031 
6032 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6033 			  rx_ring->desc, rx_ring->dma);
6034 
6035 	rx_ring->desc = NULL;
6036 }
6037 
6038 /**
6039  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6040  * @adapter: board private structure
6041  *
6042  * Free all receive software resources
6043  **/
6044 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6045 {
6046 	int i;
6047 
6048 #ifdef IXGBE_FCOE
6049 	ixgbe_free_fcoe_ddp_resources(adapter);
6050 
6051 #endif
6052 	for (i = 0; i < adapter->num_rx_queues; i++)
6053 		if (adapter->rx_ring[i]->desc)
6054 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6055 }
6056 
6057 /**
6058  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6059  * @netdev: network interface device structure
6060  * @new_mtu: new value for maximum frame size
6061  *
6062  * Returns 0 on success, negative on failure
6063  **/
6064 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6065 {
6066 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6067 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
6068 
6069 	/* MTU < 68 is an error and causes problems on some kernels */
6070 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
6071 		return -EINVAL;
6072 
6073 	/*
6074 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6075 	 * paths when MTU greater than 1500 is configured.  So display a
6076 	 * warning that legacy VFs will be disabled.
6077 	 */
6078 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6079 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6080 	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
6081 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6082 
6083 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6084 
6085 	/* must set new MTU before calling down or up */
6086 	netdev->mtu = new_mtu;
6087 
6088 	if (netif_running(netdev))
6089 		ixgbe_reinit_locked(adapter);
6090 
6091 	return 0;
6092 }
6093 
6094 /**
6095  * ixgbe_open - Called when a network interface is made active
6096  * @netdev: network interface device structure
6097  *
6098  * Returns 0 on success, negative value on failure
6099  *
6100  * The open entry point is called when a network interface is made
6101  * active by the system (IFF_UP).  At this point all resources needed
6102  * for transmit and receive operations are allocated, the interrupt
6103  * handler is registered with the OS, the watchdog timer is started,
6104  * and the stack is notified that the interface is ready.
6105  **/
6106 int ixgbe_open(struct net_device *netdev)
6107 {
6108 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6109 	struct ixgbe_hw *hw = &adapter->hw;
6110 	int err, queues;
6111 
6112 	/* disallow open during test */
6113 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6114 		return -EBUSY;
6115 
6116 	netif_carrier_off(netdev);
6117 
6118 	/* allocate transmit descriptors */
6119 	err = ixgbe_setup_all_tx_resources(adapter);
6120 	if (err)
6121 		goto err_setup_tx;
6122 
6123 	/* allocate receive descriptors */
6124 	err = ixgbe_setup_all_rx_resources(adapter);
6125 	if (err)
6126 		goto err_setup_rx;
6127 
6128 	ixgbe_configure(adapter);
6129 
6130 	err = ixgbe_request_irq(adapter);
6131 	if (err)
6132 		goto err_req_irq;
6133 
6134 	/* Notify the stack of the actual queue counts. */
6135 	if (adapter->num_rx_pools > 1)
6136 		queues = adapter->num_rx_queues_per_pool;
6137 	else
6138 		queues = adapter->num_tx_queues;
6139 
6140 	err = netif_set_real_num_tx_queues(netdev, queues);
6141 	if (err)
6142 		goto err_set_queues;
6143 
6144 	if (adapter->num_rx_pools > 1 &&
6145 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
6146 		queues = IXGBE_MAX_L2A_QUEUES;
6147 	else
6148 		queues = adapter->num_rx_queues;
6149 	err = netif_set_real_num_rx_queues(netdev, queues);
6150 	if (err)
6151 		goto err_set_queues;
6152 
6153 	ixgbe_ptp_init(adapter);
6154 
6155 	ixgbe_up_complete(adapter);
6156 
6157 	ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6158 	udp_tunnel_get_rx_info(netdev);
6159 
6160 	return 0;
6161 
6162 err_set_queues:
6163 	ixgbe_free_irq(adapter);
6164 err_req_irq:
6165 	ixgbe_free_all_rx_resources(adapter);
6166 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6167 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6168 err_setup_rx:
6169 	ixgbe_free_all_tx_resources(adapter);
6170 err_setup_tx:
6171 	ixgbe_reset(adapter);
6172 
6173 	return err;
6174 }
6175 
6176 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6177 {
6178 	ixgbe_ptp_suspend(adapter);
6179 
6180 	if (adapter->hw.phy.ops.enter_lplu) {
6181 		adapter->hw.phy.reset_disable = true;
6182 		ixgbe_down(adapter);
6183 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6184 		adapter->hw.phy.reset_disable = false;
6185 	} else {
6186 		ixgbe_down(adapter);
6187 	}
6188 
6189 	ixgbe_free_irq(adapter);
6190 
6191 	ixgbe_free_all_tx_resources(adapter);
6192 	ixgbe_free_all_rx_resources(adapter);
6193 }
6194 
6195 /**
6196  * ixgbe_close - Disables a network interface
6197  * @netdev: network interface device structure
6198  *
6199  * Returns 0, this is not allowed to fail
6200  *
6201  * The close entry point is called when an interface is de-activated
6202  * by the OS.  The hardware is still under the drivers control, but
6203  * needs to be disabled.  A global MAC reset is issued to stop the
6204  * hardware, and all transmit and receive resources are freed.
6205  **/
6206 int ixgbe_close(struct net_device *netdev)
6207 {
6208 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6209 
6210 	ixgbe_ptp_stop(adapter);
6211 
6212 	ixgbe_close_suspend(adapter);
6213 
6214 	ixgbe_fdir_filter_exit(adapter);
6215 
6216 	ixgbe_release_hw_control(adapter);
6217 
6218 	return 0;
6219 }
6220 
6221 #ifdef CONFIG_PM
6222 static int ixgbe_resume(struct pci_dev *pdev)
6223 {
6224 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6225 	struct net_device *netdev = adapter->netdev;
6226 	u32 err;
6227 
6228 	adapter->hw.hw_addr = adapter->io_addr;
6229 	pci_set_power_state(pdev, PCI_D0);
6230 	pci_restore_state(pdev);
6231 	/*
6232 	 * pci_restore_state clears dev->state_saved so call
6233 	 * pci_save_state to restore it.
6234 	 */
6235 	pci_save_state(pdev);
6236 
6237 	err = pci_enable_device_mem(pdev);
6238 	if (err) {
6239 		e_dev_err("Cannot enable PCI device from suspend\n");
6240 		return err;
6241 	}
6242 	smp_mb__before_atomic();
6243 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6244 	pci_set_master(pdev);
6245 
6246 	pci_wake_from_d3(pdev, false);
6247 
6248 	ixgbe_reset(adapter);
6249 
6250 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6251 
6252 	rtnl_lock();
6253 	err = ixgbe_init_interrupt_scheme(adapter);
6254 	if (!err && netif_running(netdev))
6255 		err = ixgbe_open(netdev);
6256 
6257 	rtnl_unlock();
6258 
6259 	if (err)
6260 		return err;
6261 
6262 	netif_device_attach(netdev);
6263 
6264 	return 0;
6265 }
6266 #endif /* CONFIG_PM */
6267 
6268 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6269 {
6270 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6271 	struct net_device *netdev = adapter->netdev;
6272 	struct ixgbe_hw *hw = &adapter->hw;
6273 	u32 ctrl, fctrl;
6274 	u32 wufc = adapter->wol;
6275 #ifdef CONFIG_PM
6276 	int retval = 0;
6277 #endif
6278 
6279 	netif_device_detach(netdev);
6280 
6281 	rtnl_lock();
6282 	if (netif_running(netdev))
6283 		ixgbe_close_suspend(adapter);
6284 	rtnl_unlock();
6285 
6286 	ixgbe_clear_interrupt_scheme(adapter);
6287 
6288 #ifdef CONFIG_PM
6289 	retval = pci_save_state(pdev);
6290 	if (retval)
6291 		return retval;
6292 
6293 #endif
6294 	if (hw->mac.ops.stop_link_on_d3)
6295 		hw->mac.ops.stop_link_on_d3(hw);
6296 
6297 	if (wufc) {
6298 		ixgbe_set_rx_mode(netdev);
6299 
6300 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6301 		if (hw->mac.ops.enable_tx_laser)
6302 			hw->mac.ops.enable_tx_laser(hw);
6303 
6304 		/* turn on all-multi mode if wake on multicast is enabled */
6305 		if (wufc & IXGBE_WUFC_MC) {
6306 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6307 			fctrl |= IXGBE_FCTRL_MPE;
6308 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6309 		}
6310 
6311 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6312 		ctrl |= IXGBE_CTRL_GIO_DIS;
6313 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6314 
6315 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6316 	} else {
6317 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6318 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6319 	}
6320 
6321 	switch (hw->mac.type) {
6322 	case ixgbe_mac_82598EB:
6323 		pci_wake_from_d3(pdev, false);
6324 		break;
6325 	case ixgbe_mac_82599EB:
6326 	case ixgbe_mac_X540:
6327 	case ixgbe_mac_X550:
6328 	case ixgbe_mac_X550EM_x:
6329 	case ixgbe_mac_x550em_a:
6330 		pci_wake_from_d3(pdev, !!wufc);
6331 		break;
6332 	default:
6333 		break;
6334 	}
6335 
6336 	*enable_wake = !!wufc;
6337 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6338 		hw->phy.ops.set_phy_power(hw, false);
6339 
6340 	ixgbe_release_hw_control(adapter);
6341 
6342 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6343 		pci_disable_device(pdev);
6344 
6345 	return 0;
6346 }
6347 
6348 #ifdef CONFIG_PM
6349 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6350 {
6351 	int retval;
6352 	bool wake;
6353 
6354 	retval = __ixgbe_shutdown(pdev, &wake);
6355 	if (retval)
6356 		return retval;
6357 
6358 	if (wake) {
6359 		pci_prepare_to_sleep(pdev);
6360 	} else {
6361 		pci_wake_from_d3(pdev, false);
6362 		pci_set_power_state(pdev, PCI_D3hot);
6363 	}
6364 
6365 	return 0;
6366 }
6367 #endif /* CONFIG_PM */
6368 
6369 static void ixgbe_shutdown(struct pci_dev *pdev)
6370 {
6371 	bool wake;
6372 
6373 	__ixgbe_shutdown(pdev, &wake);
6374 
6375 	if (system_state == SYSTEM_POWER_OFF) {
6376 		pci_wake_from_d3(pdev, wake);
6377 		pci_set_power_state(pdev, PCI_D3hot);
6378 	}
6379 }
6380 
6381 /**
6382  * ixgbe_update_stats - Update the board statistics counters.
6383  * @adapter: board private structure
6384  **/
6385 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6386 {
6387 	struct net_device *netdev = adapter->netdev;
6388 	struct ixgbe_hw *hw = &adapter->hw;
6389 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6390 	u64 total_mpc = 0;
6391 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6392 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6393 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6394 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6395 
6396 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6397 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6398 		return;
6399 
6400 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6401 		u64 rsc_count = 0;
6402 		u64 rsc_flush = 0;
6403 		for (i = 0; i < adapter->num_rx_queues; i++) {
6404 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6405 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6406 		}
6407 		adapter->rsc_total_count = rsc_count;
6408 		adapter->rsc_total_flush = rsc_flush;
6409 	}
6410 
6411 	for (i = 0; i < adapter->num_rx_queues; i++) {
6412 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6413 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6414 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6415 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6416 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6417 		bytes += rx_ring->stats.bytes;
6418 		packets += rx_ring->stats.packets;
6419 	}
6420 	adapter->non_eop_descs = non_eop_descs;
6421 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6422 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6423 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6424 	netdev->stats.rx_bytes = bytes;
6425 	netdev->stats.rx_packets = packets;
6426 
6427 	bytes = 0;
6428 	packets = 0;
6429 	/* gather some stats to the adapter struct that are per queue */
6430 	for (i = 0; i < adapter->num_tx_queues; i++) {
6431 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6432 		restart_queue += tx_ring->tx_stats.restart_queue;
6433 		tx_busy += tx_ring->tx_stats.tx_busy;
6434 		bytes += tx_ring->stats.bytes;
6435 		packets += tx_ring->stats.packets;
6436 	}
6437 	adapter->restart_queue = restart_queue;
6438 	adapter->tx_busy = tx_busy;
6439 	netdev->stats.tx_bytes = bytes;
6440 	netdev->stats.tx_packets = packets;
6441 
6442 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6443 
6444 	/* 8 register reads */
6445 	for (i = 0; i < 8; i++) {
6446 		/* for packet buffers not used, the register should read 0 */
6447 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6448 		missed_rx += mpc;
6449 		hwstats->mpc[i] += mpc;
6450 		total_mpc += hwstats->mpc[i];
6451 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6452 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6453 		switch (hw->mac.type) {
6454 		case ixgbe_mac_82598EB:
6455 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6456 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6457 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6458 			hwstats->pxonrxc[i] +=
6459 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6460 			break;
6461 		case ixgbe_mac_82599EB:
6462 		case ixgbe_mac_X540:
6463 		case ixgbe_mac_X550:
6464 		case ixgbe_mac_X550EM_x:
6465 		case ixgbe_mac_x550em_a:
6466 			hwstats->pxonrxc[i] +=
6467 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6468 			break;
6469 		default:
6470 			break;
6471 		}
6472 	}
6473 
6474 	/*16 register reads */
6475 	for (i = 0; i < 16; i++) {
6476 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6477 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6478 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6479 		    (hw->mac.type == ixgbe_mac_X540) ||
6480 		    (hw->mac.type == ixgbe_mac_X550) ||
6481 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6482 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6483 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6484 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6485 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6486 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6487 		}
6488 	}
6489 
6490 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6491 	/* work around hardware counting issue */
6492 	hwstats->gprc -= missed_rx;
6493 
6494 	ixgbe_update_xoff_received(adapter);
6495 
6496 	/* 82598 hardware only has a 32 bit counter in the high register */
6497 	switch (hw->mac.type) {
6498 	case ixgbe_mac_82598EB:
6499 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6500 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6501 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6502 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6503 		break;
6504 	case ixgbe_mac_X540:
6505 	case ixgbe_mac_X550:
6506 	case ixgbe_mac_X550EM_x:
6507 	case ixgbe_mac_x550em_a:
6508 		/* OS2BMC stats are X540 and later */
6509 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6510 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6511 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6512 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6513 	case ixgbe_mac_82599EB:
6514 		for (i = 0; i < 16; i++)
6515 			adapter->hw_rx_no_dma_resources +=
6516 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6517 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6518 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6519 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6520 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6521 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6522 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6523 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6524 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6525 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6526 #ifdef IXGBE_FCOE
6527 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6528 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6529 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6530 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6531 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6532 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6533 		/* Add up per cpu counters for total ddp aloc fail */
6534 		if (adapter->fcoe.ddp_pool) {
6535 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6536 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6537 			unsigned int cpu;
6538 			u64 noddp = 0, noddp_ext_buff = 0;
6539 			for_each_possible_cpu(cpu) {
6540 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6541 				noddp += ddp_pool->noddp;
6542 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6543 			}
6544 			hwstats->fcoe_noddp = noddp;
6545 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6546 		}
6547 #endif /* IXGBE_FCOE */
6548 		break;
6549 	default:
6550 		break;
6551 	}
6552 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6553 	hwstats->bprc += bprc;
6554 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6555 	if (hw->mac.type == ixgbe_mac_82598EB)
6556 		hwstats->mprc -= bprc;
6557 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6558 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6559 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6560 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6561 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6562 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6563 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6564 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6565 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6566 	hwstats->lxontxc += lxon;
6567 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6568 	hwstats->lxofftxc += lxoff;
6569 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6570 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6571 	/*
6572 	 * 82598 errata - tx of flow control packets is included in tx counters
6573 	 */
6574 	xon_off_tot = lxon + lxoff;
6575 	hwstats->gptc -= xon_off_tot;
6576 	hwstats->mptc -= xon_off_tot;
6577 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6578 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6579 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6580 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6581 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6582 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6583 	hwstats->ptc64 -= xon_off_tot;
6584 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6585 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6586 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6587 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6588 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6589 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6590 
6591 	/* Fill out the OS statistics structure */
6592 	netdev->stats.multicast = hwstats->mprc;
6593 
6594 	/* Rx Errors */
6595 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6596 	netdev->stats.rx_dropped = 0;
6597 	netdev->stats.rx_length_errors = hwstats->rlec;
6598 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6599 	netdev->stats.rx_missed_errors = total_mpc;
6600 }
6601 
6602 /**
6603  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6604  * @adapter: pointer to the device adapter structure
6605  **/
6606 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6607 {
6608 	struct ixgbe_hw *hw = &adapter->hw;
6609 	int i;
6610 
6611 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6612 		return;
6613 
6614 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6615 
6616 	/* if interface is down do nothing */
6617 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6618 		return;
6619 
6620 	/* do nothing if we are not using signature filters */
6621 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6622 		return;
6623 
6624 	adapter->fdir_overflow++;
6625 
6626 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6627 		for (i = 0; i < adapter->num_tx_queues; i++)
6628 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6629 				&(adapter->tx_ring[i]->state));
6630 		/* re-enable flow director interrupts */
6631 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6632 	} else {
6633 		e_err(probe, "failed to finish FDIR re-initialization, "
6634 		      "ignored adding FDIR ATR filters\n");
6635 	}
6636 }
6637 
6638 /**
6639  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6640  * @adapter: pointer to the device adapter structure
6641  *
6642  * This function serves two purposes.  First it strobes the interrupt lines
6643  * in order to make certain interrupts are occurring.  Secondly it sets the
6644  * bits needed to check for TX hangs.  As a result we should immediately
6645  * determine if a hang has occurred.
6646  */
6647 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6648 {
6649 	struct ixgbe_hw *hw = &adapter->hw;
6650 	u64 eics = 0;
6651 	int i;
6652 
6653 	/* If we're down, removing or resetting, just bail */
6654 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6655 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6656 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6657 		return;
6658 
6659 	/* Force detection of hung controller */
6660 	if (netif_carrier_ok(adapter->netdev)) {
6661 		for (i = 0; i < adapter->num_tx_queues; i++)
6662 			set_check_for_tx_hang(adapter->tx_ring[i]);
6663 	}
6664 
6665 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6666 		/*
6667 		 * for legacy and MSI interrupts don't set any bits
6668 		 * that are enabled for EIAM, because this operation
6669 		 * would set *both* EIMS and EICS for any bit in EIAM
6670 		 */
6671 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6672 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6673 	} else {
6674 		/* get one bit for every active tx/rx interrupt vector */
6675 		for (i = 0; i < adapter->num_q_vectors; i++) {
6676 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6677 			if (qv->rx.ring || qv->tx.ring)
6678 				eics |= BIT_ULL(i);
6679 		}
6680 	}
6681 
6682 	/* Cause software interrupt to ensure rings are cleaned */
6683 	ixgbe_irq_rearm_queues(adapter, eics);
6684 }
6685 
6686 /**
6687  * ixgbe_watchdog_update_link - update the link status
6688  * @adapter: pointer to the device adapter structure
6689  * @link_speed: pointer to a u32 to store the link_speed
6690  **/
6691 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6692 {
6693 	struct ixgbe_hw *hw = &adapter->hw;
6694 	u32 link_speed = adapter->link_speed;
6695 	bool link_up = adapter->link_up;
6696 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6697 
6698 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6699 		return;
6700 
6701 	if (hw->mac.ops.check_link) {
6702 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6703 	} else {
6704 		/* always assume link is up, if no check link function */
6705 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6706 		link_up = true;
6707 	}
6708 
6709 	if (adapter->ixgbe_ieee_pfc)
6710 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6711 
6712 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6713 		hw->mac.ops.fc_enable(hw);
6714 		ixgbe_set_rx_drop_en(adapter);
6715 	}
6716 
6717 	if (link_up ||
6718 	    time_after(jiffies, (adapter->link_check_timeout +
6719 				 IXGBE_TRY_LINK_TIMEOUT))) {
6720 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6721 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6722 		IXGBE_WRITE_FLUSH(hw);
6723 	}
6724 
6725 	adapter->link_up = link_up;
6726 	adapter->link_speed = link_speed;
6727 }
6728 
6729 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6730 {
6731 #ifdef CONFIG_IXGBE_DCB
6732 	struct net_device *netdev = adapter->netdev;
6733 	struct dcb_app app = {
6734 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6735 			      .protocol = 0,
6736 			     };
6737 	u8 up = 0;
6738 
6739 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6740 		up = dcb_ieee_getapp_mask(netdev, &app);
6741 
6742 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6743 #endif
6744 }
6745 
6746 /**
6747  * ixgbe_watchdog_link_is_up - update netif_carrier status and
6748  *                             print link up message
6749  * @adapter: pointer to the device adapter structure
6750  **/
6751 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6752 {
6753 	struct net_device *netdev = adapter->netdev;
6754 	struct ixgbe_hw *hw = &adapter->hw;
6755 	struct net_device *upper;
6756 	struct list_head *iter;
6757 	u32 link_speed = adapter->link_speed;
6758 	const char *speed_str;
6759 	bool flow_rx, flow_tx;
6760 
6761 	/* only continue if link was previously down */
6762 	if (netif_carrier_ok(netdev))
6763 		return;
6764 
6765 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6766 
6767 	switch (hw->mac.type) {
6768 	case ixgbe_mac_82598EB: {
6769 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6770 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6771 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6772 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6773 	}
6774 		break;
6775 	case ixgbe_mac_X540:
6776 	case ixgbe_mac_X550:
6777 	case ixgbe_mac_X550EM_x:
6778 	case ixgbe_mac_x550em_a:
6779 	case ixgbe_mac_82599EB: {
6780 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6781 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6782 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6783 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6784 	}
6785 		break;
6786 	default:
6787 		flow_tx = false;
6788 		flow_rx = false;
6789 		break;
6790 	}
6791 
6792 	adapter->last_rx_ptp_check = jiffies;
6793 
6794 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6795 		ixgbe_ptp_start_cyclecounter(adapter);
6796 
6797 	switch (link_speed) {
6798 	case IXGBE_LINK_SPEED_10GB_FULL:
6799 		speed_str = "10 Gbps";
6800 		break;
6801 	case IXGBE_LINK_SPEED_2_5GB_FULL:
6802 		speed_str = "2.5 Gbps";
6803 		break;
6804 	case IXGBE_LINK_SPEED_1GB_FULL:
6805 		speed_str = "1 Gbps";
6806 		break;
6807 	case IXGBE_LINK_SPEED_100_FULL:
6808 		speed_str = "100 Mbps";
6809 		break;
6810 	default:
6811 		speed_str = "unknown speed";
6812 		break;
6813 	}
6814 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
6815 	       ((flow_rx && flow_tx) ? "RX/TX" :
6816 	       (flow_rx ? "RX" :
6817 	       (flow_tx ? "TX" : "None"))));
6818 
6819 	netif_carrier_on(netdev);
6820 	ixgbe_check_vf_rate_limit(adapter);
6821 
6822 	/* enable transmits */
6823 	netif_tx_wake_all_queues(adapter->netdev);
6824 
6825 	/* enable any upper devices */
6826 	rtnl_lock();
6827 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
6828 		if (netif_is_macvlan(upper)) {
6829 			struct macvlan_dev *vlan = netdev_priv(upper);
6830 
6831 			if (vlan->fwd_priv)
6832 				netif_tx_wake_all_queues(upper);
6833 		}
6834 	}
6835 	rtnl_unlock();
6836 
6837 	/* update the default user priority for VFs */
6838 	ixgbe_update_default_up(adapter);
6839 
6840 	/* ping all the active vfs to let them know link has changed */
6841 	ixgbe_ping_all_vfs(adapter);
6842 }
6843 
6844 /**
6845  * ixgbe_watchdog_link_is_down - update netif_carrier status and
6846  *                               print link down message
6847  * @adapter: pointer to the adapter structure
6848  **/
6849 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6850 {
6851 	struct net_device *netdev = adapter->netdev;
6852 	struct ixgbe_hw *hw = &adapter->hw;
6853 
6854 	adapter->link_up = false;
6855 	adapter->link_speed = 0;
6856 
6857 	/* only continue if link was up previously */
6858 	if (!netif_carrier_ok(netdev))
6859 		return;
6860 
6861 	/* poll for SFP+ cable when link is down */
6862 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6863 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6864 
6865 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6866 		ixgbe_ptp_start_cyclecounter(adapter);
6867 
6868 	e_info(drv, "NIC Link is Down\n");
6869 	netif_carrier_off(netdev);
6870 
6871 	/* ping all the active vfs to let them know link has changed */
6872 	ixgbe_ping_all_vfs(adapter);
6873 }
6874 
6875 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
6876 {
6877 	int i;
6878 
6879 	for (i = 0; i < adapter->num_tx_queues; i++) {
6880 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6881 
6882 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
6883 			return true;
6884 	}
6885 
6886 	return false;
6887 }
6888 
6889 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
6890 {
6891 	struct ixgbe_hw *hw = &adapter->hw;
6892 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
6893 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
6894 
6895 	int i, j;
6896 
6897 	if (!adapter->num_vfs)
6898 		return false;
6899 
6900 	/* resetting the PF is only needed for MAC before X550 */
6901 	if (hw->mac.type >= ixgbe_mac_X550)
6902 		return false;
6903 
6904 	for (i = 0; i < adapter->num_vfs; i++) {
6905 		for (j = 0; j < q_per_pool; j++) {
6906 			u32 h, t;
6907 
6908 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
6909 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
6910 
6911 			if (h != t)
6912 				return true;
6913 		}
6914 	}
6915 
6916 	return false;
6917 }
6918 
6919 /**
6920  * ixgbe_watchdog_flush_tx - flush queues on link down
6921  * @adapter: pointer to the device adapter structure
6922  **/
6923 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6924 {
6925 	if (!netif_carrier_ok(adapter->netdev)) {
6926 		if (ixgbe_ring_tx_pending(adapter) ||
6927 		    ixgbe_vf_tx_pending(adapter)) {
6928 			/* We've lost link, so the controller stops DMA,
6929 			 * but we've got queued Tx work that's never going
6930 			 * to get done, so reset controller to flush Tx.
6931 			 * (Do the reset outside of interrupt context).
6932 			 */
6933 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6934 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
6935 		}
6936 	}
6937 }
6938 
6939 #ifdef CONFIG_PCI_IOV
6940 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
6941 				      struct pci_dev *vfdev)
6942 {
6943 	if (!pci_wait_for_pending_transaction(vfdev))
6944 		e_dev_warn("Issuing VFLR with pending transactions\n");
6945 
6946 	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
6947 	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
6948 
6949 	msleep(100);
6950 }
6951 
6952 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6953 {
6954 	struct ixgbe_hw *hw = &adapter->hw;
6955 	struct pci_dev *pdev = adapter->pdev;
6956 	unsigned int vf;
6957 	u32 gpc;
6958 
6959 	if (!(netif_carrier_ok(adapter->netdev)))
6960 		return;
6961 
6962 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6963 	if (gpc) /* If incrementing then no need for the check below */
6964 		return;
6965 	/* Check to see if a bad DMA write target from an errant or
6966 	 * malicious VF has caused a PCIe error.  If so then we can
6967 	 * issue a VFLR to the offending VF(s) and then resume without
6968 	 * requesting a full slot reset.
6969 	 */
6970 
6971 	if (!pdev)
6972 		return;
6973 
6974 	/* check status reg for all VFs owned by this PF */
6975 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
6976 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
6977 		u16 status_reg;
6978 
6979 		if (!vfdev)
6980 			continue;
6981 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
6982 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
6983 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
6984 			ixgbe_issue_vf_flr(adapter, vfdev);
6985 	}
6986 }
6987 
6988 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6989 {
6990 	u32 ssvpc;
6991 
6992 	/* Do not perform spoof check for 82598 or if not in IOV mode */
6993 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
6994 	    adapter->num_vfs == 0)
6995 		return;
6996 
6997 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
6998 
6999 	/*
7000 	 * ssvpc register is cleared on read, if zero then no
7001 	 * spoofed packets in the last interval.
7002 	 */
7003 	if (!ssvpc)
7004 		return;
7005 
7006 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7007 }
7008 #else
7009 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7010 {
7011 }
7012 
7013 static void
7014 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7015 {
7016 }
7017 #endif /* CONFIG_PCI_IOV */
7018 
7019 
7020 /**
7021  * ixgbe_watchdog_subtask - check and bring link up
7022  * @adapter: pointer to the device adapter structure
7023  **/
7024 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7025 {
7026 	/* if interface is down, removing or resetting, do nothing */
7027 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7028 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7029 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7030 		return;
7031 
7032 	ixgbe_watchdog_update_link(adapter);
7033 
7034 	if (adapter->link_up)
7035 		ixgbe_watchdog_link_is_up(adapter);
7036 	else
7037 		ixgbe_watchdog_link_is_down(adapter);
7038 
7039 	ixgbe_check_for_bad_vf(adapter);
7040 	ixgbe_spoof_check(adapter);
7041 	ixgbe_update_stats(adapter);
7042 
7043 	ixgbe_watchdog_flush_tx(adapter);
7044 }
7045 
7046 /**
7047  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7048  * @adapter: the ixgbe adapter structure
7049  **/
7050 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7051 {
7052 	struct ixgbe_hw *hw = &adapter->hw;
7053 	s32 err;
7054 
7055 	/* not searching for SFP so there is nothing to do here */
7056 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7057 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7058 		return;
7059 
7060 	if (adapter->sfp_poll_time &&
7061 	    time_after(adapter->sfp_poll_time, jiffies))
7062 		return; /* If not yet time to poll for SFP */
7063 
7064 	/* someone else is in init, wait until next service event */
7065 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7066 		return;
7067 
7068 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7069 
7070 	err = hw->phy.ops.identify_sfp(hw);
7071 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7072 		goto sfp_out;
7073 
7074 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7075 		/* If no cable is present, then we need to reset
7076 		 * the next time we find a good cable. */
7077 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7078 	}
7079 
7080 	/* exit on error */
7081 	if (err)
7082 		goto sfp_out;
7083 
7084 	/* exit if reset not needed */
7085 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7086 		goto sfp_out;
7087 
7088 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7089 
7090 	/*
7091 	 * A module may be identified correctly, but the EEPROM may not have
7092 	 * support for that module.  setup_sfp() will fail in that case, so
7093 	 * we should not allow that module to load.
7094 	 */
7095 	if (hw->mac.type == ixgbe_mac_82598EB)
7096 		err = hw->phy.ops.reset(hw);
7097 	else
7098 		err = hw->mac.ops.setup_sfp(hw);
7099 
7100 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7101 		goto sfp_out;
7102 
7103 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7104 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7105 
7106 sfp_out:
7107 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7108 
7109 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7110 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7111 		e_dev_err("failed to initialize because an unsupported "
7112 			  "SFP+ module type was detected.\n");
7113 		e_dev_err("Reload the driver after installing a "
7114 			  "supported module.\n");
7115 		unregister_netdev(adapter->netdev);
7116 	}
7117 }
7118 
7119 /**
7120  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7121  * @adapter: the ixgbe adapter structure
7122  **/
7123 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7124 {
7125 	struct ixgbe_hw *hw = &adapter->hw;
7126 	u32 speed;
7127 	bool autoneg = false;
7128 
7129 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7130 		return;
7131 
7132 	/* someone else is in init, wait until next service event */
7133 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7134 		return;
7135 
7136 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7137 
7138 	speed = hw->phy.autoneg_advertised;
7139 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
7140 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
7141 
7142 		/* setup the highest link when no autoneg */
7143 		if (!autoneg) {
7144 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
7145 				speed = IXGBE_LINK_SPEED_10GB_FULL;
7146 		}
7147 	}
7148 
7149 	if (hw->mac.ops.setup_link)
7150 		hw->mac.ops.setup_link(hw, speed, true);
7151 
7152 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7153 	adapter->link_check_timeout = jiffies;
7154 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7155 }
7156 
7157 /**
7158  * ixgbe_service_timer - Timer Call-back
7159  * @data: pointer to adapter cast into an unsigned long
7160  **/
7161 static void ixgbe_service_timer(unsigned long data)
7162 {
7163 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
7164 	unsigned long next_event_offset;
7165 
7166 	/* poll faster when waiting for link */
7167 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7168 		next_event_offset = HZ / 10;
7169 	else
7170 		next_event_offset = HZ * 2;
7171 
7172 	/* Reset the timer */
7173 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7174 
7175 	ixgbe_service_event_schedule(adapter);
7176 }
7177 
7178 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7179 {
7180 	struct ixgbe_hw *hw = &adapter->hw;
7181 	u32 status;
7182 
7183 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7184 		return;
7185 
7186 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7187 
7188 	if (!hw->phy.ops.handle_lasi)
7189 		return;
7190 
7191 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7192 	if (status != IXGBE_ERR_OVERTEMP)
7193 		return;
7194 
7195 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7196 }
7197 
7198 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7199 {
7200 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7201 		return;
7202 
7203 	/* If we're already down, removing or resetting, just bail */
7204 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7205 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7206 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7207 		return;
7208 
7209 	ixgbe_dump(adapter);
7210 	netdev_err(adapter->netdev, "Reset adapter\n");
7211 	adapter->tx_timeout_count++;
7212 
7213 	rtnl_lock();
7214 	ixgbe_reinit_locked(adapter);
7215 	rtnl_unlock();
7216 }
7217 
7218 /**
7219  * ixgbe_service_task - manages and runs subtasks
7220  * @work: pointer to work_struct containing our data
7221  **/
7222 static void ixgbe_service_task(struct work_struct *work)
7223 {
7224 	struct ixgbe_adapter *adapter = container_of(work,
7225 						     struct ixgbe_adapter,
7226 						     service_task);
7227 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7228 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7229 			rtnl_lock();
7230 			ixgbe_down(adapter);
7231 			rtnl_unlock();
7232 		}
7233 		ixgbe_service_event_complete(adapter);
7234 		return;
7235 	}
7236 	if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7237 		rtnl_lock();
7238 		adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7239 		udp_tunnel_get_rx_info(adapter->netdev);
7240 		rtnl_unlock();
7241 	}
7242 	ixgbe_reset_subtask(adapter);
7243 	ixgbe_phy_interrupt_subtask(adapter);
7244 	ixgbe_sfp_detection_subtask(adapter);
7245 	ixgbe_sfp_link_config_subtask(adapter);
7246 	ixgbe_check_overtemp_subtask(adapter);
7247 	ixgbe_watchdog_subtask(adapter);
7248 	ixgbe_fdir_reinit_subtask(adapter);
7249 	ixgbe_check_hang_subtask(adapter);
7250 
7251 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7252 		ixgbe_ptp_overflow_check(adapter);
7253 		ixgbe_ptp_rx_hang(adapter);
7254 	}
7255 
7256 	ixgbe_service_event_complete(adapter);
7257 }
7258 
7259 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7260 		     struct ixgbe_tx_buffer *first,
7261 		     u8 *hdr_len)
7262 {
7263 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7264 	struct sk_buff *skb = first->skb;
7265 	union {
7266 		struct iphdr *v4;
7267 		struct ipv6hdr *v6;
7268 		unsigned char *hdr;
7269 	} ip;
7270 	union {
7271 		struct tcphdr *tcp;
7272 		unsigned char *hdr;
7273 	} l4;
7274 	u32 paylen, l4_offset;
7275 	int err;
7276 
7277 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7278 		return 0;
7279 
7280 	if (!skb_is_gso(skb))
7281 		return 0;
7282 
7283 	err = skb_cow_head(skb, 0);
7284 	if (err < 0)
7285 		return err;
7286 
7287 	ip.hdr = skb_network_header(skb);
7288 	l4.hdr = skb_checksum_start(skb);
7289 
7290 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7291 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7292 
7293 	/* initialize outer IP header fields */
7294 	if (ip.v4->version == 4) {
7295 		/* IP header will have to cancel out any data that
7296 		 * is not a part of the outer IP header
7297 		 */
7298 		ip.v4->check = csum_fold(csum_add(lco_csum(skb),
7299 						  csum_unfold(l4.tcp->check)));
7300 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7301 
7302 		ip.v4->tot_len = 0;
7303 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7304 				   IXGBE_TX_FLAGS_CSUM |
7305 				   IXGBE_TX_FLAGS_IPV4;
7306 	} else {
7307 		ip.v6->payload_len = 0;
7308 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7309 				   IXGBE_TX_FLAGS_CSUM;
7310 	}
7311 
7312 	/* determine offset of inner transport header */
7313 	l4_offset = l4.hdr - skb->data;
7314 
7315 	/* compute length of segmentation header */
7316 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7317 
7318 	/* remove payload length from inner checksum */
7319 	paylen = skb->len - l4_offset;
7320 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
7321 
7322 	/* update gso size and bytecount with header size */
7323 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7324 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7325 
7326 	/* mss_l4len_id: use 0 as index for TSO */
7327 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7328 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7329 
7330 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7331 	vlan_macip_lens = l4.hdr - ip.hdr;
7332 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7333 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7334 
7335 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
7336 			  mss_l4len_idx);
7337 
7338 	return 1;
7339 }
7340 
7341 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7342 {
7343 	unsigned int offset = 0;
7344 
7345 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7346 
7347 	return offset == skb_checksum_start_offset(skb);
7348 }
7349 
7350 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7351 			  struct ixgbe_tx_buffer *first)
7352 {
7353 	struct sk_buff *skb = first->skb;
7354 	u32 vlan_macip_lens = 0;
7355 	u32 type_tucmd = 0;
7356 
7357 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7358 csum_failed:
7359 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7360 					 IXGBE_TX_FLAGS_CC)))
7361 			return;
7362 		goto no_csum;
7363 	}
7364 
7365 	switch (skb->csum_offset) {
7366 	case offsetof(struct tcphdr, check):
7367 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7368 		/* fall through */
7369 	case offsetof(struct udphdr, check):
7370 		break;
7371 	case offsetof(struct sctphdr, checksum):
7372 		/* validate that this is actually an SCTP request */
7373 		if (((first->protocol == htons(ETH_P_IP)) &&
7374 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7375 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7376 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7377 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7378 			break;
7379 		}
7380 		/* fall through */
7381 	default:
7382 		skb_checksum_help(skb);
7383 		goto csum_failed;
7384 	}
7385 
7386 	/* update TX checksum flag */
7387 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7388 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7389 			  skb_network_offset(skb);
7390 no_csum:
7391 	/* vlan_macip_lens: MACLEN, VLAN tag */
7392 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7393 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7394 
7395 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd, 0);
7396 }
7397 
7398 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7399 	((_flag <= _result) ? \
7400 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7401 	 ((u32)(_input & _flag) / (_flag / _result)))
7402 
7403 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7404 {
7405 	/* set type for advanced descriptor with frame checksum insertion */
7406 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7407 		       IXGBE_ADVTXD_DCMD_DEXT |
7408 		       IXGBE_ADVTXD_DCMD_IFCS;
7409 
7410 	/* set HW vlan bit if vlan is present */
7411 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7412 				   IXGBE_ADVTXD_DCMD_VLE);
7413 
7414 	/* set segmentation enable bits for TSO/FSO */
7415 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7416 				   IXGBE_ADVTXD_DCMD_TSE);
7417 
7418 	/* set timestamp bit if present */
7419 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7420 				   IXGBE_ADVTXD_MAC_TSTAMP);
7421 
7422 	/* insert frame checksum */
7423 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7424 
7425 	return cmd_type;
7426 }
7427 
7428 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7429 				   u32 tx_flags, unsigned int paylen)
7430 {
7431 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7432 
7433 	/* enable L4 checksum for TSO and TX checksum offload */
7434 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7435 					IXGBE_TX_FLAGS_CSUM,
7436 					IXGBE_ADVTXD_POPTS_TXSM);
7437 
7438 	/* enble IPv4 checksum for TSO */
7439 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7440 					IXGBE_TX_FLAGS_IPV4,
7441 					IXGBE_ADVTXD_POPTS_IXSM);
7442 
7443 	/*
7444 	 * Check Context must be set if Tx switch is enabled, which it
7445 	 * always is for case where virtual functions are running
7446 	 */
7447 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7448 					IXGBE_TX_FLAGS_CC,
7449 					IXGBE_ADVTXD_CC);
7450 
7451 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7452 }
7453 
7454 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7455 {
7456 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7457 
7458 	/* Herbert's original patch had:
7459 	 *  smp_mb__after_netif_stop_queue();
7460 	 * but since that doesn't exist yet, just open code it.
7461 	 */
7462 	smp_mb();
7463 
7464 	/* We need to check again in a case another CPU has just
7465 	 * made room available.
7466 	 */
7467 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7468 		return -EBUSY;
7469 
7470 	/* A reprieve! - use start_queue because it doesn't call schedule */
7471 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7472 	++tx_ring->tx_stats.restart_queue;
7473 	return 0;
7474 }
7475 
7476 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7477 {
7478 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7479 		return 0;
7480 
7481 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7482 }
7483 
7484 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7485 		       IXGBE_TXD_CMD_RS)
7486 
7487 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7488 			 struct ixgbe_tx_buffer *first,
7489 			 const u8 hdr_len)
7490 {
7491 	struct sk_buff *skb = first->skb;
7492 	struct ixgbe_tx_buffer *tx_buffer;
7493 	union ixgbe_adv_tx_desc *tx_desc;
7494 	struct skb_frag_struct *frag;
7495 	dma_addr_t dma;
7496 	unsigned int data_len, size;
7497 	u32 tx_flags = first->tx_flags;
7498 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7499 	u16 i = tx_ring->next_to_use;
7500 
7501 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7502 
7503 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7504 
7505 	size = skb_headlen(skb);
7506 	data_len = skb->data_len;
7507 
7508 #ifdef IXGBE_FCOE
7509 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7510 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7511 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7512 			data_len = 0;
7513 		} else {
7514 			data_len -= sizeof(struct fcoe_crc_eof);
7515 		}
7516 	}
7517 
7518 #endif
7519 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7520 
7521 	tx_buffer = first;
7522 
7523 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7524 		if (dma_mapping_error(tx_ring->dev, dma))
7525 			goto dma_error;
7526 
7527 		/* record length, and DMA address */
7528 		dma_unmap_len_set(tx_buffer, len, size);
7529 		dma_unmap_addr_set(tx_buffer, dma, dma);
7530 
7531 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7532 
7533 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7534 			tx_desc->read.cmd_type_len =
7535 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7536 
7537 			i++;
7538 			tx_desc++;
7539 			if (i == tx_ring->count) {
7540 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7541 				i = 0;
7542 			}
7543 			tx_desc->read.olinfo_status = 0;
7544 
7545 			dma += IXGBE_MAX_DATA_PER_TXD;
7546 			size -= IXGBE_MAX_DATA_PER_TXD;
7547 
7548 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7549 		}
7550 
7551 		if (likely(!data_len))
7552 			break;
7553 
7554 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7555 
7556 		i++;
7557 		tx_desc++;
7558 		if (i == tx_ring->count) {
7559 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7560 			i = 0;
7561 		}
7562 		tx_desc->read.olinfo_status = 0;
7563 
7564 #ifdef IXGBE_FCOE
7565 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7566 #else
7567 		size = skb_frag_size(frag);
7568 #endif
7569 		data_len -= size;
7570 
7571 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7572 				       DMA_TO_DEVICE);
7573 
7574 		tx_buffer = &tx_ring->tx_buffer_info[i];
7575 	}
7576 
7577 	/* write last descriptor with RS and EOP bits */
7578 	cmd_type |= size | IXGBE_TXD_CMD;
7579 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7580 
7581 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7582 
7583 	/* set the timestamp */
7584 	first->time_stamp = jiffies;
7585 
7586 	/*
7587 	 * Force memory writes to complete before letting h/w know there
7588 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7589 	 * memory model archs, such as IA-64).
7590 	 *
7591 	 * We also need this memory barrier to make certain all of the
7592 	 * status bits have been updated before next_to_watch is written.
7593 	 */
7594 	wmb();
7595 
7596 	/* set next_to_watch value indicating a packet is present */
7597 	first->next_to_watch = tx_desc;
7598 
7599 	i++;
7600 	if (i == tx_ring->count)
7601 		i = 0;
7602 
7603 	tx_ring->next_to_use = i;
7604 
7605 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7606 
7607 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7608 		writel(i, tx_ring->tail);
7609 
7610 		/* we need this if more than one processor can write to our tail
7611 		 * at a time, it synchronizes IO on IA64/Altix systems
7612 		 */
7613 		mmiowb();
7614 	}
7615 
7616 	return;
7617 dma_error:
7618 	dev_err(tx_ring->dev, "TX DMA map failed\n");
7619 
7620 	/* clear dma mappings for failed tx_buffer_info map */
7621 	for (;;) {
7622 		tx_buffer = &tx_ring->tx_buffer_info[i];
7623 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
7624 		if (tx_buffer == first)
7625 			break;
7626 		if (i == 0)
7627 			i = tx_ring->count;
7628 		i--;
7629 	}
7630 
7631 	tx_ring->next_to_use = i;
7632 }
7633 
7634 static void ixgbe_atr(struct ixgbe_ring *ring,
7635 		      struct ixgbe_tx_buffer *first)
7636 {
7637 	struct ixgbe_q_vector *q_vector = ring->q_vector;
7638 	union ixgbe_atr_hash_dword input = { .dword = 0 };
7639 	union ixgbe_atr_hash_dword common = { .dword = 0 };
7640 	union {
7641 		unsigned char *network;
7642 		struct iphdr *ipv4;
7643 		struct ipv6hdr *ipv6;
7644 	} hdr;
7645 	struct tcphdr *th;
7646 	unsigned int hlen;
7647 	struct sk_buff *skb;
7648 	__be16 vlan_id;
7649 	int l4_proto;
7650 
7651 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7652 	if (!q_vector)
7653 		return;
7654 
7655 	/* do nothing if sampling is disabled */
7656 	if (!ring->atr_sample_rate)
7657 		return;
7658 
7659 	ring->atr_count++;
7660 
7661 	/* currently only IPv4/IPv6 with TCP is supported */
7662 	if ((first->protocol != htons(ETH_P_IP)) &&
7663 	    (first->protocol != htons(ETH_P_IPV6)))
7664 		return;
7665 
7666 	/* snag network header to get L4 type and address */
7667 	skb = first->skb;
7668 	hdr.network = skb_network_header(skb);
7669 	if (skb->encapsulation &&
7670 	    first->protocol == htons(ETH_P_IP) &&
7671 	    hdr.ipv4->protocol != IPPROTO_UDP) {
7672 		struct ixgbe_adapter *adapter = q_vector->adapter;
7673 
7674 		/* verify the port is recognized as VXLAN */
7675 		if (adapter->vxlan_port &&
7676 		    udp_hdr(skb)->dest == adapter->vxlan_port)
7677 			hdr.network = skb_inner_network_header(skb);
7678 
7679 		if (adapter->geneve_port &&
7680 		    udp_hdr(skb)->dest == adapter->geneve_port)
7681 			hdr.network = skb_inner_network_header(skb);
7682 	}
7683 
7684 	/* Currently only IPv4/IPv6 with TCP is supported */
7685 	switch (hdr.ipv4->version) {
7686 	case IPVERSION:
7687 		/* access ihl as u8 to avoid unaligned access on ia64 */
7688 		hlen = (hdr.network[0] & 0x0F) << 2;
7689 		l4_proto = hdr.ipv4->protocol;
7690 		break;
7691 	case 6:
7692 		hlen = hdr.network - skb->data;
7693 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
7694 		hlen -= hdr.network - skb->data;
7695 		break;
7696 	default:
7697 		return;
7698 	}
7699 
7700 	if (l4_proto != IPPROTO_TCP)
7701 		return;
7702 
7703 	th = (struct tcphdr *)(hdr.network + hlen);
7704 
7705 	/* skip this packet since the socket is closing */
7706 	if (th->fin)
7707 		return;
7708 
7709 	/* sample on all syn packets or once every atr sample count */
7710 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7711 		return;
7712 
7713 	/* reset sample count */
7714 	ring->atr_count = 0;
7715 
7716 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7717 
7718 	/*
7719 	 * src and dst are inverted, think how the receiver sees them
7720 	 *
7721 	 * The input is broken into two sections, a non-compressed section
7722 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7723 	 * is XORed together and stored in the compressed dword.
7724 	 */
7725 	input.formatted.vlan_id = vlan_id;
7726 
7727 	/*
7728 	 * since src port and flex bytes occupy the same word XOR them together
7729 	 * and write the value to source port portion of compressed dword
7730 	 */
7731 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7732 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7733 	else
7734 		common.port.src ^= th->dest ^ first->protocol;
7735 	common.port.dst ^= th->source;
7736 
7737 	switch (hdr.ipv4->version) {
7738 	case IPVERSION:
7739 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7740 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7741 		break;
7742 	case 6:
7743 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7744 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7745 			     hdr.ipv6->saddr.s6_addr32[1] ^
7746 			     hdr.ipv6->saddr.s6_addr32[2] ^
7747 			     hdr.ipv6->saddr.s6_addr32[3] ^
7748 			     hdr.ipv6->daddr.s6_addr32[0] ^
7749 			     hdr.ipv6->daddr.s6_addr32[1] ^
7750 			     hdr.ipv6->daddr.s6_addr32[2] ^
7751 			     hdr.ipv6->daddr.s6_addr32[3];
7752 		break;
7753 	default:
7754 		break;
7755 	}
7756 
7757 	if (hdr.network != skb_network_header(skb))
7758 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
7759 
7760 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7761 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7762 					      input, common, ring->queue_index);
7763 }
7764 
7765 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7766 			      void *accel_priv, select_queue_fallback_t fallback)
7767 {
7768 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7769 #ifdef IXGBE_FCOE
7770 	struct ixgbe_adapter *adapter;
7771 	struct ixgbe_ring_feature *f;
7772 	int txq;
7773 #endif
7774 
7775 	if (fwd_adapter)
7776 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7777 
7778 #ifdef IXGBE_FCOE
7779 
7780 	/*
7781 	 * only execute the code below if protocol is FCoE
7782 	 * or FIP and we have FCoE enabled on the adapter
7783 	 */
7784 	switch (vlan_get_protocol(skb)) {
7785 	case htons(ETH_P_FCOE):
7786 	case htons(ETH_P_FIP):
7787 		adapter = netdev_priv(dev);
7788 
7789 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7790 			break;
7791 	default:
7792 		return fallback(dev, skb);
7793 	}
7794 
7795 	f = &adapter->ring_feature[RING_F_FCOE];
7796 
7797 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7798 					   smp_processor_id();
7799 
7800 	while (txq >= f->indices)
7801 		txq -= f->indices;
7802 
7803 	return txq + f->offset;
7804 #else
7805 	return fallback(dev, skb);
7806 #endif
7807 }
7808 
7809 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7810 			  struct ixgbe_adapter *adapter,
7811 			  struct ixgbe_ring *tx_ring)
7812 {
7813 	struct ixgbe_tx_buffer *first;
7814 	int tso;
7815 	u32 tx_flags = 0;
7816 	unsigned short f;
7817 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7818 	__be16 protocol = skb->protocol;
7819 	u8 hdr_len = 0;
7820 
7821 	/*
7822 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7823 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7824 	 *       + 2 desc gap to keep tail from touching head,
7825 	 *       + 1 desc for context descriptor,
7826 	 * otherwise try next time
7827 	 */
7828 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7829 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7830 
7831 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7832 		tx_ring->tx_stats.tx_busy++;
7833 		return NETDEV_TX_BUSY;
7834 	}
7835 
7836 	/* record the location of the first descriptor for this packet */
7837 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7838 	first->skb = skb;
7839 	first->bytecount = skb->len;
7840 	first->gso_segs = 1;
7841 
7842 	/* if we have a HW VLAN tag being added default to the HW one */
7843 	if (skb_vlan_tag_present(skb)) {
7844 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7845 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7846 	/* else if it is a SW VLAN check the next protocol and store the tag */
7847 	} else if (protocol == htons(ETH_P_8021Q)) {
7848 		struct vlan_hdr *vhdr, _vhdr;
7849 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7850 		if (!vhdr)
7851 			goto out_drop;
7852 
7853 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7854 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7855 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7856 	}
7857 	protocol = vlan_get_protocol(skb);
7858 
7859 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
7860 	    adapter->ptp_clock &&
7861 	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7862 				   &adapter->state)) {
7863 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7864 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7865 
7866 		/* schedule check for Tx timestamp */
7867 		adapter->ptp_tx_skb = skb_get(skb);
7868 		adapter->ptp_tx_start = jiffies;
7869 		schedule_work(&adapter->ptp_tx_work);
7870 	}
7871 
7872 	skb_tx_timestamp(skb);
7873 
7874 #ifdef CONFIG_PCI_IOV
7875 	/*
7876 	 * Use the l2switch_enable flag - would be false if the DMA
7877 	 * Tx switch had been disabled.
7878 	 */
7879 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7880 		tx_flags |= IXGBE_TX_FLAGS_CC;
7881 
7882 #endif
7883 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7884 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7885 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7886 	     (skb->priority != TC_PRIO_CONTROL))) {
7887 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7888 		tx_flags |= (skb->priority & 0x7) <<
7889 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7890 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7891 			struct vlan_ethhdr *vhdr;
7892 
7893 			if (skb_cow_head(skb, 0))
7894 				goto out_drop;
7895 			vhdr = (struct vlan_ethhdr *)skb->data;
7896 			vhdr->h_vlan_TCI = htons(tx_flags >>
7897 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7898 		} else {
7899 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7900 		}
7901 	}
7902 
7903 	/* record initial flags and protocol */
7904 	first->tx_flags = tx_flags;
7905 	first->protocol = protocol;
7906 
7907 #ifdef IXGBE_FCOE
7908 	/* setup tx offload for FCoE */
7909 	if ((protocol == htons(ETH_P_FCOE)) &&
7910 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7911 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7912 		if (tso < 0)
7913 			goto out_drop;
7914 
7915 		goto xmit_fcoe;
7916 	}
7917 
7918 #endif /* IXGBE_FCOE */
7919 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7920 	if (tso < 0)
7921 		goto out_drop;
7922 	else if (!tso)
7923 		ixgbe_tx_csum(tx_ring, first);
7924 
7925 	/* add the ATR filter if ATR is on */
7926 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7927 		ixgbe_atr(tx_ring, first);
7928 
7929 #ifdef IXGBE_FCOE
7930 xmit_fcoe:
7931 #endif /* IXGBE_FCOE */
7932 	ixgbe_tx_map(tx_ring, first, hdr_len);
7933 
7934 	return NETDEV_TX_OK;
7935 
7936 out_drop:
7937 	dev_kfree_skb_any(first->skb);
7938 	first->skb = NULL;
7939 
7940 	return NETDEV_TX_OK;
7941 }
7942 
7943 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7944 				      struct net_device *netdev,
7945 				      struct ixgbe_ring *ring)
7946 {
7947 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7948 	struct ixgbe_ring *tx_ring;
7949 
7950 	/*
7951 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7952 	 * in order to meet this minimum size requirement.
7953 	 */
7954 	if (skb_put_padto(skb, 17))
7955 		return NETDEV_TX_OK;
7956 
7957 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7958 
7959 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7960 }
7961 
7962 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7963 				    struct net_device *netdev)
7964 {
7965 	return __ixgbe_xmit_frame(skb, netdev, NULL);
7966 }
7967 
7968 /**
7969  * ixgbe_set_mac - Change the Ethernet Address of the NIC
7970  * @netdev: network interface device structure
7971  * @p: pointer to an address structure
7972  *
7973  * Returns 0 on success, negative on failure
7974  **/
7975 static int ixgbe_set_mac(struct net_device *netdev, void *p)
7976 {
7977 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7978 	struct ixgbe_hw *hw = &adapter->hw;
7979 	struct sockaddr *addr = p;
7980 
7981 	if (!is_valid_ether_addr(addr->sa_data))
7982 		return -EADDRNOTAVAIL;
7983 
7984 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
7985 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
7986 
7987 	ixgbe_mac_set_default_filter(adapter);
7988 
7989 	return 0;
7990 }
7991 
7992 static int
7993 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
7994 {
7995 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7996 	struct ixgbe_hw *hw = &adapter->hw;
7997 	u16 value;
7998 	int rc;
7999 
8000 	if (prtad != hw->phy.mdio.prtad)
8001 		return -EINVAL;
8002 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8003 	if (!rc)
8004 		rc = value;
8005 	return rc;
8006 }
8007 
8008 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8009 			    u16 addr, u16 value)
8010 {
8011 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8012 	struct ixgbe_hw *hw = &adapter->hw;
8013 
8014 	if (prtad != hw->phy.mdio.prtad)
8015 		return -EINVAL;
8016 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8017 }
8018 
8019 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8020 {
8021 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8022 
8023 	switch (cmd) {
8024 	case SIOCSHWTSTAMP:
8025 		return ixgbe_ptp_set_ts_config(adapter, req);
8026 	case SIOCGHWTSTAMP:
8027 		return ixgbe_ptp_get_ts_config(adapter, req);
8028 	default:
8029 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8030 	}
8031 }
8032 
8033 /**
8034  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8035  * netdev->dev_addrs
8036  * @netdev: network interface device structure
8037  *
8038  * Returns non-zero on failure
8039  **/
8040 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8041 {
8042 	int err = 0;
8043 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8044 	struct ixgbe_hw *hw = &adapter->hw;
8045 
8046 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8047 		rtnl_lock();
8048 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8049 		rtnl_unlock();
8050 
8051 		/* update SAN MAC vmdq pool selection */
8052 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8053 	}
8054 	return err;
8055 }
8056 
8057 /**
8058  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8059  * netdev->dev_addrs
8060  * @netdev: network interface device structure
8061  *
8062  * Returns non-zero on failure
8063  **/
8064 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8065 {
8066 	int err = 0;
8067 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8068 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8069 
8070 	if (is_valid_ether_addr(mac->san_addr)) {
8071 		rtnl_lock();
8072 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8073 		rtnl_unlock();
8074 	}
8075 	return err;
8076 }
8077 
8078 #ifdef CONFIG_NET_POLL_CONTROLLER
8079 /*
8080  * Polling 'interrupt' - used by things like netconsole to send skbs
8081  * without having to re-enable interrupts. It's not called while
8082  * the interrupt routine is executing.
8083  */
8084 static void ixgbe_netpoll(struct net_device *netdev)
8085 {
8086 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8087 	int i;
8088 
8089 	/* if interface is down do nothing */
8090 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8091 		return;
8092 
8093 	/* loop through and schedule all active queues */
8094 	for (i = 0; i < adapter->num_q_vectors; i++)
8095 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8096 }
8097 
8098 #endif
8099 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
8100 						   struct rtnl_link_stats64 *stats)
8101 {
8102 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8103 	int i;
8104 
8105 	rcu_read_lock();
8106 	for (i = 0; i < adapter->num_rx_queues; i++) {
8107 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
8108 		u64 bytes, packets;
8109 		unsigned int start;
8110 
8111 		if (ring) {
8112 			do {
8113 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8114 				packets = ring->stats.packets;
8115 				bytes   = ring->stats.bytes;
8116 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8117 			stats->rx_packets += packets;
8118 			stats->rx_bytes   += bytes;
8119 		}
8120 	}
8121 
8122 	for (i = 0; i < adapter->num_tx_queues; i++) {
8123 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
8124 		u64 bytes, packets;
8125 		unsigned int start;
8126 
8127 		if (ring) {
8128 			do {
8129 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8130 				packets = ring->stats.packets;
8131 				bytes   = ring->stats.bytes;
8132 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8133 			stats->tx_packets += packets;
8134 			stats->tx_bytes   += bytes;
8135 		}
8136 	}
8137 	rcu_read_unlock();
8138 	/* following stats updated by ixgbe_watchdog_task() */
8139 	stats->multicast	= netdev->stats.multicast;
8140 	stats->rx_errors	= netdev->stats.rx_errors;
8141 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8142 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8143 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8144 	return stats;
8145 }
8146 
8147 #ifdef CONFIG_IXGBE_DCB
8148 /**
8149  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8150  * @adapter: pointer to ixgbe_adapter
8151  * @tc: number of traffic classes currently enabled
8152  *
8153  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8154  * 802.1Q priority maps to a packet buffer that exists.
8155  */
8156 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8157 {
8158 	struct ixgbe_hw *hw = &adapter->hw;
8159 	u32 reg, rsave;
8160 	int i;
8161 
8162 	/* 82598 have a static priority to TC mapping that can not
8163 	 * be changed so no validation is needed.
8164 	 */
8165 	if (hw->mac.type == ixgbe_mac_82598EB)
8166 		return;
8167 
8168 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8169 	rsave = reg;
8170 
8171 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8172 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8173 
8174 		/* If up2tc is out of bounds default to zero */
8175 		if (up2tc > tc)
8176 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8177 	}
8178 
8179 	if (reg != rsave)
8180 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8181 
8182 	return;
8183 }
8184 
8185 /**
8186  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8187  * @adapter: Pointer to adapter struct
8188  *
8189  * Populate the netdev user priority to tc map
8190  */
8191 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8192 {
8193 	struct net_device *dev = adapter->netdev;
8194 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8195 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8196 	u8 prio;
8197 
8198 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8199 		u8 tc = 0;
8200 
8201 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8202 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8203 		else if (ets)
8204 			tc = ets->prio_tc[prio];
8205 
8206 		netdev_set_prio_tc_map(dev, prio, tc);
8207 	}
8208 }
8209 
8210 #endif /* CONFIG_IXGBE_DCB */
8211 /**
8212  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8213  *
8214  * @netdev: net device to configure
8215  * @tc: number of traffic classes to enable
8216  */
8217 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8218 {
8219 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8220 	struct ixgbe_hw *hw = &adapter->hw;
8221 	bool pools;
8222 
8223 	/* Hardware supports up to 8 traffic classes */
8224 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8225 		return -EINVAL;
8226 
8227 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8228 		return -EINVAL;
8229 
8230 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
8231 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
8232 		return -EBUSY;
8233 
8234 	/* Hardware has to reinitialize queues and interrupts to
8235 	 * match packet buffer alignment. Unfortunately, the
8236 	 * hardware is not flexible enough to do this dynamically.
8237 	 */
8238 	if (netif_running(dev))
8239 		ixgbe_close(dev);
8240 	else
8241 		ixgbe_reset(adapter);
8242 
8243 	ixgbe_clear_interrupt_scheme(adapter);
8244 
8245 #ifdef CONFIG_IXGBE_DCB
8246 	if (tc) {
8247 		netdev_set_num_tc(dev, tc);
8248 		ixgbe_set_prio_tc_map(adapter);
8249 
8250 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8251 
8252 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8253 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8254 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8255 		}
8256 	} else {
8257 		netdev_reset_tc(dev);
8258 
8259 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8260 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8261 
8262 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8263 
8264 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8265 		adapter->dcb_cfg.pfc_mode_enable = false;
8266 	}
8267 
8268 	ixgbe_validate_rtr(adapter, tc);
8269 
8270 #endif /* CONFIG_IXGBE_DCB */
8271 	ixgbe_init_interrupt_scheme(adapter);
8272 
8273 	if (netif_running(dev))
8274 		return ixgbe_open(dev);
8275 
8276 	return 0;
8277 }
8278 
8279 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8280 			       struct tc_cls_u32_offload *cls)
8281 {
8282 	u32 hdl = cls->knode.handle;
8283 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8284 	u32 loc = cls->knode.handle & 0xfffff;
8285 	int err = 0, i, j;
8286 	struct ixgbe_jump_table *jump = NULL;
8287 
8288 	if (loc > IXGBE_MAX_HW_ENTRIES)
8289 		return -EINVAL;
8290 
8291 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8292 		return -EINVAL;
8293 
8294 	/* Clear this filter in the link data it is associated with */
8295 	if (uhtid != 0x800) {
8296 		jump = adapter->jump_tables[uhtid];
8297 		if (!jump)
8298 			return -EINVAL;
8299 		if (!test_bit(loc - 1, jump->child_loc_map))
8300 			return -EINVAL;
8301 		clear_bit(loc - 1, jump->child_loc_map);
8302 	}
8303 
8304 	/* Check if the filter being deleted is a link */
8305 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8306 		jump = adapter->jump_tables[i];
8307 		if (jump && jump->link_hdl == hdl) {
8308 			/* Delete filters in the hardware in the child hash
8309 			 * table associated with this link
8310 			 */
8311 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8312 				if (!test_bit(j, jump->child_loc_map))
8313 					continue;
8314 				spin_lock(&adapter->fdir_perfect_lock);
8315 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8316 								      NULL,
8317 								      j + 1);
8318 				spin_unlock(&adapter->fdir_perfect_lock);
8319 				clear_bit(j, jump->child_loc_map);
8320 			}
8321 			/* Remove resources for this link */
8322 			kfree(jump->input);
8323 			kfree(jump->mask);
8324 			kfree(jump);
8325 			adapter->jump_tables[i] = NULL;
8326 			return err;
8327 		}
8328 	}
8329 
8330 	spin_lock(&adapter->fdir_perfect_lock);
8331 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8332 	spin_unlock(&adapter->fdir_perfect_lock);
8333 	return err;
8334 }
8335 
8336 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8337 					    __be16 protocol,
8338 					    struct tc_cls_u32_offload *cls)
8339 {
8340 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8341 
8342 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8343 		return -EINVAL;
8344 
8345 	/* This ixgbe devices do not support hash tables at the moment
8346 	 * so abort when given hash tables.
8347 	 */
8348 	if (cls->hnode.divisor > 0)
8349 		return -EINVAL;
8350 
8351 	set_bit(uhtid - 1, &adapter->tables);
8352 	return 0;
8353 }
8354 
8355 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
8356 					    struct tc_cls_u32_offload *cls)
8357 {
8358 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8359 
8360 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8361 		return -EINVAL;
8362 
8363 	clear_bit(uhtid - 1, &adapter->tables);
8364 	return 0;
8365 }
8366 
8367 #ifdef CONFIG_NET_CLS_ACT
8368 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
8369 				  u8 *queue, u64 *action)
8370 {
8371 	unsigned int num_vfs = adapter->num_vfs, vf;
8372 	struct net_device *upper;
8373 	struct list_head *iter;
8374 
8375 	/* redirect to a SRIOV VF */
8376 	for (vf = 0; vf < num_vfs; ++vf) {
8377 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
8378 		if (upper->ifindex == ifindex) {
8379 			if (adapter->num_rx_pools > 1)
8380 				*queue = vf * 2;
8381 			else
8382 				*queue = vf * adapter->num_rx_queues_per_pool;
8383 
8384 			*action = vf + 1;
8385 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
8386 			return 0;
8387 		}
8388 	}
8389 
8390 	/* redirect to a offloaded macvlan netdev */
8391 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
8392 		if (netif_is_macvlan(upper)) {
8393 			struct macvlan_dev *dfwd = netdev_priv(upper);
8394 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
8395 
8396 			if (vadapter && vadapter->netdev->ifindex == ifindex) {
8397 				*queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
8398 				*action = *queue;
8399 				return 0;
8400 			}
8401 		}
8402 	}
8403 
8404 	return -EINVAL;
8405 }
8406 
8407 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8408 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8409 {
8410 	const struct tc_action *a;
8411 	LIST_HEAD(actions);
8412 	int err;
8413 
8414 	if (tc_no_actions(exts))
8415 		return -EINVAL;
8416 
8417 	tcf_exts_to_list(exts, &actions);
8418 	list_for_each_entry(a, &actions, list) {
8419 
8420 		/* Drop action */
8421 		if (is_tcf_gact_shot(a)) {
8422 			*action = IXGBE_FDIR_DROP_QUEUE;
8423 			*queue = IXGBE_FDIR_DROP_QUEUE;
8424 			return 0;
8425 		}
8426 
8427 		/* Redirect to a VF or a offloaded macvlan */
8428 		if (is_tcf_mirred_redirect(a)) {
8429 			int ifindex = tcf_mirred_ifindex(a);
8430 
8431 			err = handle_redirect_action(adapter, ifindex, queue,
8432 						     action);
8433 			if (err == 0)
8434 				return err;
8435 		}
8436 	}
8437 
8438 	return -EINVAL;
8439 }
8440 #else
8441 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8442 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8443 {
8444 	return -EINVAL;
8445 }
8446 #endif /* CONFIG_NET_CLS_ACT */
8447 
8448 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
8449 				    union ixgbe_atr_input *mask,
8450 				    struct tc_cls_u32_offload *cls,
8451 				    struct ixgbe_mat_field *field_ptr,
8452 				    struct ixgbe_nexthdr *nexthdr)
8453 {
8454 	int i, j, off;
8455 	__be32 val, m;
8456 	bool found_entry = false, found_jump_field = false;
8457 
8458 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
8459 		off = cls->knode.sel->keys[i].off;
8460 		val = cls->knode.sel->keys[i].val;
8461 		m = cls->knode.sel->keys[i].mask;
8462 
8463 		for (j = 0; field_ptr[j].val; j++) {
8464 			if (field_ptr[j].off == off) {
8465 				field_ptr[j].val(input, mask, val, m);
8466 				input->filter.formatted.flow_type |=
8467 					field_ptr[j].type;
8468 				found_entry = true;
8469 				break;
8470 			}
8471 		}
8472 		if (nexthdr) {
8473 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
8474 			    nexthdr->val == cls->knode.sel->keys[i].val &&
8475 			    nexthdr->mask == cls->knode.sel->keys[i].mask)
8476 				found_jump_field = true;
8477 			else
8478 				continue;
8479 		}
8480 	}
8481 
8482 	if (nexthdr && !found_jump_field)
8483 		return -EINVAL;
8484 
8485 	if (!found_entry)
8486 		return 0;
8487 
8488 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
8489 				    IXGBE_ATR_L4TYPE_MASK;
8490 
8491 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
8492 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
8493 
8494 	return 0;
8495 }
8496 
8497 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
8498 				  __be16 protocol,
8499 				  struct tc_cls_u32_offload *cls)
8500 {
8501 	u32 loc = cls->knode.handle & 0xfffff;
8502 	struct ixgbe_hw *hw = &adapter->hw;
8503 	struct ixgbe_mat_field *field_ptr;
8504 	struct ixgbe_fdir_filter *input = NULL;
8505 	union ixgbe_atr_input *mask = NULL;
8506 	struct ixgbe_jump_table *jump = NULL;
8507 	int i, err = -EINVAL;
8508 	u8 queue;
8509 	u32 uhtid, link_uhtid;
8510 
8511 	uhtid = TC_U32_USERHTID(cls->knode.handle);
8512 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
8513 
8514 	/* At the moment cls_u32 jumps to network layer and skips past
8515 	 * L2 headers. The canonical method to match L2 frames is to use
8516 	 * negative values. However this is error prone at best but really
8517 	 * just broken because there is no way to "know" what sort of hdr
8518 	 * is in front of the network layer. Fix cls_u32 to support L2
8519 	 * headers when needed.
8520 	 */
8521 	if (protocol != htons(ETH_P_IP))
8522 		return err;
8523 
8524 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
8525 		e_err(drv, "Location out of range\n");
8526 		return err;
8527 	}
8528 
8529 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
8530 	 * links and also the fields used to advance the parser across each
8531 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
8532 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
8533 	 * To add support for new nodes update ixgbe_model.h parse structures
8534 	 * this function _should_ be generic try not to hardcode values here.
8535 	 */
8536 	if (uhtid == 0x800) {
8537 		field_ptr = (adapter->jump_tables[0])->mat;
8538 	} else {
8539 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8540 			return err;
8541 		if (!adapter->jump_tables[uhtid])
8542 			return err;
8543 		field_ptr = (adapter->jump_tables[uhtid])->mat;
8544 	}
8545 
8546 	if (!field_ptr)
8547 		return err;
8548 
8549 	/* At this point we know the field_ptr is valid and need to either
8550 	 * build cls_u32 link or attach filter. Because adding a link to
8551 	 * a handle that does not exist is invalid and the same for adding
8552 	 * rules to handles that don't exist.
8553 	 */
8554 
8555 	if (link_uhtid) {
8556 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
8557 
8558 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
8559 			return err;
8560 
8561 		if (!test_bit(link_uhtid - 1, &adapter->tables))
8562 			return err;
8563 
8564 		/* Multiple filters as links to the same hash table are not
8565 		 * supported. To add a new filter with the same next header
8566 		 * but different match/jump conditions, create a new hash table
8567 		 * and link to it.
8568 		 */
8569 		if (adapter->jump_tables[link_uhtid] &&
8570 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
8571 			e_err(drv, "Link filter exists for link: %x\n",
8572 			      link_uhtid);
8573 			return err;
8574 		}
8575 
8576 		for (i = 0; nexthdr[i].jump; i++) {
8577 			if (nexthdr[i].o != cls->knode.sel->offoff ||
8578 			    nexthdr[i].s != cls->knode.sel->offshift ||
8579 			    nexthdr[i].m != cls->knode.sel->offmask)
8580 				return err;
8581 
8582 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
8583 			if (!jump)
8584 				return -ENOMEM;
8585 			input = kzalloc(sizeof(*input), GFP_KERNEL);
8586 			if (!input) {
8587 				err = -ENOMEM;
8588 				goto free_jump;
8589 			}
8590 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8591 			if (!mask) {
8592 				err = -ENOMEM;
8593 				goto free_input;
8594 			}
8595 			jump->input = input;
8596 			jump->mask = mask;
8597 			jump->link_hdl = cls->knode.handle;
8598 
8599 			err = ixgbe_clsu32_build_input(input, mask, cls,
8600 						       field_ptr, &nexthdr[i]);
8601 			if (!err) {
8602 				jump->mat = nexthdr[i].jump;
8603 				adapter->jump_tables[link_uhtid] = jump;
8604 				break;
8605 			}
8606 		}
8607 		return 0;
8608 	}
8609 
8610 	input = kzalloc(sizeof(*input), GFP_KERNEL);
8611 	if (!input)
8612 		return -ENOMEM;
8613 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8614 	if (!mask) {
8615 		err = -ENOMEM;
8616 		goto free_input;
8617 	}
8618 
8619 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
8620 		if ((adapter->jump_tables[uhtid])->input)
8621 			memcpy(input, (adapter->jump_tables[uhtid])->input,
8622 			       sizeof(*input));
8623 		if ((adapter->jump_tables[uhtid])->mask)
8624 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
8625 			       sizeof(*mask));
8626 
8627 		/* Lookup in all child hash tables if this location is already
8628 		 * filled with a filter
8629 		 */
8630 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8631 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
8632 
8633 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
8634 				e_err(drv, "Filter exists in location: %x\n",
8635 				      loc);
8636 				err = -EINVAL;
8637 				goto err_out;
8638 			}
8639 		}
8640 	}
8641 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
8642 	if (err)
8643 		goto err_out;
8644 
8645 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
8646 			       &queue);
8647 	if (err < 0)
8648 		goto err_out;
8649 
8650 	input->sw_idx = loc;
8651 
8652 	spin_lock(&adapter->fdir_perfect_lock);
8653 
8654 	if (hlist_empty(&adapter->fdir_filter_list)) {
8655 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
8656 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
8657 		if (err)
8658 			goto err_out_w_lock;
8659 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
8660 		err = -EINVAL;
8661 		goto err_out_w_lock;
8662 	}
8663 
8664 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
8665 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
8666 						    input->sw_idx, queue);
8667 	if (!err)
8668 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
8669 	spin_unlock(&adapter->fdir_perfect_lock);
8670 
8671 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
8672 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
8673 
8674 	kfree(mask);
8675 	return err;
8676 err_out_w_lock:
8677 	spin_unlock(&adapter->fdir_perfect_lock);
8678 err_out:
8679 	kfree(mask);
8680 free_input:
8681 	kfree(input);
8682 free_jump:
8683 	kfree(jump);
8684 	return err;
8685 }
8686 
8687 static int __ixgbe_setup_tc(struct net_device *dev, u32 handle, __be16 proto,
8688 			    struct tc_to_netdev *tc)
8689 {
8690 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8691 
8692 	if (TC_H_MAJ(handle) == TC_H_MAJ(TC_H_INGRESS) &&
8693 	    tc->type == TC_SETUP_CLSU32) {
8694 		switch (tc->cls_u32->command) {
8695 		case TC_CLSU32_NEW_KNODE:
8696 		case TC_CLSU32_REPLACE_KNODE:
8697 			return ixgbe_configure_clsu32(adapter,
8698 						      proto, tc->cls_u32);
8699 		case TC_CLSU32_DELETE_KNODE:
8700 			return ixgbe_delete_clsu32(adapter, tc->cls_u32);
8701 		case TC_CLSU32_NEW_HNODE:
8702 		case TC_CLSU32_REPLACE_HNODE:
8703 			return ixgbe_configure_clsu32_add_hnode(adapter, proto,
8704 								tc->cls_u32);
8705 		case TC_CLSU32_DELETE_HNODE:
8706 			return ixgbe_configure_clsu32_del_hnode(adapter,
8707 								tc->cls_u32);
8708 		default:
8709 			return -EINVAL;
8710 		}
8711 	}
8712 
8713 	if (tc->type != TC_SETUP_MQPRIO)
8714 		return -EINVAL;
8715 
8716 	return ixgbe_setup_tc(dev, tc->tc);
8717 }
8718 
8719 #ifdef CONFIG_PCI_IOV
8720 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
8721 {
8722 	struct net_device *netdev = adapter->netdev;
8723 
8724 	rtnl_lock();
8725 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
8726 	rtnl_unlock();
8727 }
8728 
8729 #endif
8730 void ixgbe_do_reset(struct net_device *netdev)
8731 {
8732 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8733 
8734 	if (netif_running(netdev))
8735 		ixgbe_reinit_locked(adapter);
8736 	else
8737 		ixgbe_reset(adapter);
8738 }
8739 
8740 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
8741 					    netdev_features_t features)
8742 {
8743 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8744 
8745 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
8746 	if (!(features & NETIF_F_RXCSUM))
8747 		features &= ~NETIF_F_LRO;
8748 
8749 	/* Turn off LRO if not RSC capable */
8750 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
8751 		features &= ~NETIF_F_LRO;
8752 
8753 	return features;
8754 }
8755 
8756 static int ixgbe_set_features(struct net_device *netdev,
8757 			      netdev_features_t features)
8758 {
8759 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8760 	netdev_features_t changed = netdev->features ^ features;
8761 	bool need_reset = false;
8762 
8763 	/* Make sure RSC matches LRO, reset if change */
8764 	if (!(features & NETIF_F_LRO)) {
8765 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8766 			need_reset = true;
8767 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
8768 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
8769 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
8770 		if (adapter->rx_itr_setting == 1 ||
8771 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
8772 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
8773 			need_reset = true;
8774 		} else if ((changed ^ features) & NETIF_F_LRO) {
8775 			e_info(probe, "rx-usecs set too low, "
8776 			       "disabling RSC\n");
8777 		}
8778 	}
8779 
8780 	/*
8781 	 * Check if Flow Director n-tuple support or hw_tc support was
8782 	 * enabled or disabled.  If the state changed, we need to reset.
8783 	 */
8784 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
8785 		/* turn off ATR, enable perfect filters and reset */
8786 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
8787 			need_reset = true;
8788 
8789 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
8790 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8791 	} else {
8792 		/* turn off perfect filters, enable ATR and reset */
8793 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
8794 			need_reset = true;
8795 
8796 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8797 
8798 		/* We cannot enable ATR if SR-IOV is enabled */
8799 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
8800 		    /* We cannot enable ATR if we have 2 or more tcs */
8801 		    (netdev_get_num_tc(netdev) > 1) ||
8802 		    /* We cannot enable ATR if RSS is disabled */
8803 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
8804 		    /* A sample rate of 0 indicates ATR disabled */
8805 		    (!adapter->atr_sample_rate))
8806 			; /* do nothing not supported */
8807 		else /* otherwise supported and set the flag */
8808 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
8809 	}
8810 
8811 	if (changed & NETIF_F_RXALL)
8812 		need_reset = true;
8813 
8814 	netdev->features = features;
8815 
8816 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
8817 		if (features & NETIF_F_RXCSUM) {
8818 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
8819 		} else {
8820 			u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
8821 
8822 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
8823 		}
8824 	}
8825 
8826 	if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
8827 		if (features & NETIF_F_RXCSUM) {
8828 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
8829 		} else {
8830 			u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
8831 
8832 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
8833 		}
8834 	}
8835 
8836 	if (need_reset)
8837 		ixgbe_do_reset(netdev);
8838 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
8839 			    NETIF_F_HW_VLAN_CTAG_FILTER))
8840 		ixgbe_set_rx_mode(netdev);
8841 
8842 	return 0;
8843 }
8844 
8845 /**
8846  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
8847  * @dev: The port's netdev
8848  * @ti: Tunnel endpoint information
8849  **/
8850 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
8851 				      struct udp_tunnel_info *ti)
8852 {
8853 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8854 	struct ixgbe_hw *hw = &adapter->hw;
8855 	__be16 port = ti->port;
8856 	u32 port_shift = 0;
8857 	u32 reg;
8858 
8859 	if (ti->sa_family != AF_INET)
8860 		return;
8861 
8862 	switch (ti->type) {
8863 	case UDP_TUNNEL_TYPE_VXLAN:
8864 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8865 			return;
8866 
8867 		if (adapter->vxlan_port == port)
8868 			return;
8869 
8870 		if (adapter->vxlan_port) {
8871 			netdev_info(dev,
8872 				    "VXLAN port %d set, not adding port %d\n",
8873 				    ntohs(adapter->vxlan_port),
8874 				    ntohs(port));
8875 			return;
8876 		}
8877 
8878 		adapter->vxlan_port = port;
8879 		break;
8880 	case UDP_TUNNEL_TYPE_GENEVE:
8881 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
8882 			return;
8883 
8884 		if (adapter->geneve_port == port)
8885 			return;
8886 
8887 		if (adapter->geneve_port) {
8888 			netdev_info(dev,
8889 				    "GENEVE port %d set, not adding port %d\n",
8890 				    ntohs(adapter->geneve_port),
8891 				    ntohs(port));
8892 			return;
8893 		}
8894 
8895 		port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
8896 		adapter->geneve_port = port;
8897 		break;
8898 	default:
8899 		return;
8900 	}
8901 
8902 	reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
8903 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
8904 }
8905 
8906 /**
8907  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
8908  * @dev: The port's netdev
8909  * @ti: Tunnel endpoint information
8910  **/
8911 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
8912 				      struct udp_tunnel_info *ti)
8913 {
8914 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8915 	u32 port_mask;
8916 
8917 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
8918 	    ti->type != UDP_TUNNEL_TYPE_GENEVE)
8919 		return;
8920 
8921 	if (ti->sa_family != AF_INET)
8922 		return;
8923 
8924 	switch (ti->type) {
8925 	case UDP_TUNNEL_TYPE_VXLAN:
8926 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8927 			return;
8928 
8929 		if (adapter->vxlan_port != ti->port) {
8930 			netdev_info(dev, "VXLAN port %d not found\n",
8931 				    ntohs(ti->port));
8932 			return;
8933 		}
8934 
8935 		port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
8936 		break;
8937 	case UDP_TUNNEL_TYPE_GENEVE:
8938 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
8939 			return;
8940 
8941 		if (adapter->geneve_port != ti->port) {
8942 			netdev_info(dev, "GENEVE port %d not found\n",
8943 				    ntohs(ti->port));
8944 			return;
8945 		}
8946 
8947 		port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
8948 		break;
8949 	default:
8950 		return;
8951 	}
8952 
8953 	ixgbe_clear_udp_tunnel_port(adapter, port_mask);
8954 	adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
8955 }
8956 
8957 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8958 			     struct net_device *dev,
8959 			     const unsigned char *addr, u16 vid,
8960 			     u16 flags)
8961 {
8962 	/* guarantee we can provide a unique filter for the unicast address */
8963 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
8964 		struct ixgbe_adapter *adapter = netdev_priv(dev);
8965 		u16 pool = VMDQ_P(0);
8966 
8967 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
8968 			return -ENOMEM;
8969 	}
8970 
8971 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
8972 }
8973 
8974 /**
8975  * ixgbe_configure_bridge_mode - set various bridge modes
8976  * @adapter - the private structure
8977  * @mode - requested bridge mode
8978  *
8979  * Configure some settings require for various bridge modes.
8980  **/
8981 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
8982 				       __u16 mode)
8983 {
8984 	struct ixgbe_hw *hw = &adapter->hw;
8985 	unsigned int p, num_pools;
8986 	u32 vmdctl;
8987 
8988 	switch (mode) {
8989 	case BRIDGE_MODE_VEPA:
8990 		/* disable Tx loopback, rely on switch hairpin mode */
8991 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
8992 
8993 		/* must enable Rx switching replication to allow multicast
8994 		 * packet reception on all VFs, and to enable source address
8995 		 * pruning.
8996 		 */
8997 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
8998 		vmdctl |= IXGBE_VT_CTL_REPLEN;
8999 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9000 
9001 		/* enable Rx source address pruning. Note, this requires
9002 		 * replication to be enabled or else it does nothing.
9003 		 */
9004 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9005 		for (p = 0; p < num_pools; p++) {
9006 			if (hw->mac.ops.set_source_address_pruning)
9007 				hw->mac.ops.set_source_address_pruning(hw,
9008 								       true,
9009 								       p);
9010 		}
9011 		break;
9012 	case BRIDGE_MODE_VEB:
9013 		/* enable Tx loopback for internal VF/PF communication */
9014 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9015 				IXGBE_PFDTXGSWC_VT_LBEN);
9016 
9017 		/* disable Rx switching replication unless we have SR-IOV
9018 		 * virtual functions
9019 		 */
9020 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9021 		if (!adapter->num_vfs)
9022 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9023 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9024 
9025 		/* disable Rx source address pruning, since we don't expect to
9026 		 * be receiving external loopback of our transmitted frames.
9027 		 */
9028 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9029 		for (p = 0; p < num_pools; p++) {
9030 			if (hw->mac.ops.set_source_address_pruning)
9031 				hw->mac.ops.set_source_address_pruning(hw,
9032 								       false,
9033 								       p);
9034 		}
9035 		break;
9036 	default:
9037 		return -EINVAL;
9038 	}
9039 
9040 	adapter->bridge_mode = mode;
9041 
9042 	e_info(drv, "enabling bridge mode: %s\n",
9043 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9044 
9045 	return 0;
9046 }
9047 
9048 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9049 				    struct nlmsghdr *nlh, u16 flags)
9050 {
9051 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9052 	struct nlattr *attr, *br_spec;
9053 	int rem;
9054 
9055 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9056 		return -EOPNOTSUPP;
9057 
9058 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9059 	if (!br_spec)
9060 		return -EINVAL;
9061 
9062 	nla_for_each_nested(attr, br_spec, rem) {
9063 		int status;
9064 		__u16 mode;
9065 
9066 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9067 			continue;
9068 
9069 		if (nla_len(attr) < sizeof(mode))
9070 			return -EINVAL;
9071 
9072 		mode = nla_get_u16(attr);
9073 		status = ixgbe_configure_bridge_mode(adapter, mode);
9074 		if (status)
9075 			return status;
9076 
9077 		break;
9078 	}
9079 
9080 	return 0;
9081 }
9082 
9083 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9084 				    struct net_device *dev,
9085 				    u32 filter_mask, int nlflags)
9086 {
9087 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9088 
9089 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9090 		return 0;
9091 
9092 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9093 				       adapter->bridge_mode, 0, 0, nlflags,
9094 				       filter_mask, NULL);
9095 }
9096 
9097 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9098 {
9099 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
9100 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9101 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
9102 	unsigned int limit;
9103 	int pool, err;
9104 
9105 	/* Hardware has a limited number of available pools. Each VF, and the
9106 	 * PF require a pool. Check to ensure we don't attempt to use more
9107 	 * then the available number of pools.
9108 	 */
9109 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9110 		return ERR_PTR(-EINVAL);
9111 
9112 #ifdef CONFIG_RPS
9113 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
9114 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
9115 			    vdev->name);
9116 		return ERR_PTR(-EINVAL);
9117 	}
9118 #endif
9119 	/* Check for hardware restriction on number of rx/tx queues */
9120 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
9121 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
9122 		netdev_info(pdev,
9123 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
9124 			    pdev->name);
9125 		return ERR_PTR(-EINVAL);
9126 	}
9127 
9128 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9129 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
9130 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
9131 		return ERR_PTR(-EBUSY);
9132 
9133 	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
9134 	if (!fwd_adapter)
9135 		return ERR_PTR(-ENOMEM);
9136 
9137 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
9138 	adapter->num_rx_pools++;
9139 	set_bit(pool, &adapter->fwd_bitmask);
9140 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9141 
9142 	/* Enable VMDq flag so device will be set in VM mode */
9143 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
9144 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9145 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
9146 
9147 	/* Force reinit of ring allocation with VMDQ enabled */
9148 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9149 	if (err)
9150 		goto fwd_add_err;
9151 	fwd_adapter->pool = pool;
9152 	fwd_adapter->real_adapter = adapter;
9153 	err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
9154 	if (err)
9155 		goto fwd_add_err;
9156 	netif_tx_start_all_queues(vdev);
9157 	return fwd_adapter;
9158 fwd_add_err:
9159 	/* unwind counter and free adapter struct */
9160 	netdev_info(pdev,
9161 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
9162 	clear_bit(pool, &adapter->fwd_bitmask);
9163 	adapter->num_rx_pools--;
9164 	kfree(fwd_adapter);
9165 	return ERR_PTR(err);
9166 }
9167 
9168 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9169 {
9170 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
9171 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
9172 	unsigned int limit;
9173 
9174 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
9175 	adapter->num_rx_pools--;
9176 
9177 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9178 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9179 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
9180 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9181 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
9182 		   fwd_adapter->pool, adapter->num_rx_pools,
9183 		   fwd_adapter->rx_base_queue,
9184 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
9185 		   adapter->fwd_bitmask);
9186 	kfree(fwd_adapter);
9187 }
9188 
9189 #define IXGBE_MAX_MAC_HDR_LEN		127
9190 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9191 
9192 static netdev_features_t
9193 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9194 		     netdev_features_t features)
9195 {
9196 	unsigned int network_hdr_len, mac_hdr_len;
9197 
9198 	/* Make certain the headers can be described by a context descriptor */
9199 	mac_hdr_len = skb_network_header(skb) - skb->data;
9200 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9201 		return features & ~(NETIF_F_HW_CSUM |
9202 				    NETIF_F_SCTP_CRC |
9203 				    NETIF_F_HW_VLAN_CTAG_TX |
9204 				    NETIF_F_TSO |
9205 				    NETIF_F_TSO6);
9206 
9207 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9208 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9209 		return features & ~(NETIF_F_HW_CSUM |
9210 				    NETIF_F_SCTP_CRC |
9211 				    NETIF_F_TSO |
9212 				    NETIF_F_TSO6);
9213 
9214 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9215 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9216 	 */
9217 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
9218 		features &= ~NETIF_F_TSO;
9219 
9220 	return features;
9221 }
9222 
9223 static const struct net_device_ops ixgbe_netdev_ops = {
9224 	.ndo_open		= ixgbe_open,
9225 	.ndo_stop		= ixgbe_close,
9226 	.ndo_start_xmit		= ixgbe_xmit_frame,
9227 	.ndo_select_queue	= ixgbe_select_queue,
9228 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
9229 	.ndo_validate_addr	= eth_validate_addr,
9230 	.ndo_set_mac_address	= ixgbe_set_mac,
9231 	.ndo_change_mtu		= ixgbe_change_mtu,
9232 	.ndo_tx_timeout		= ixgbe_tx_timeout,
9233 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
9234 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
9235 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
9236 	.ndo_do_ioctl		= ixgbe_ioctl,
9237 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
9238 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
9239 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
9240 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
9241 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
9242 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
9243 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
9244 	.ndo_get_stats64	= ixgbe_get_stats64,
9245 	.ndo_setup_tc		= __ixgbe_setup_tc,
9246 #ifdef CONFIG_NET_POLL_CONTROLLER
9247 	.ndo_poll_controller	= ixgbe_netpoll,
9248 #endif
9249 #ifdef CONFIG_NET_RX_BUSY_POLL
9250 	.ndo_busy_poll		= ixgbe_low_latency_recv,
9251 #endif
9252 #ifdef IXGBE_FCOE
9253 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
9254 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
9255 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
9256 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
9257 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
9258 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
9259 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
9260 #endif /* IXGBE_FCOE */
9261 	.ndo_set_features = ixgbe_set_features,
9262 	.ndo_fix_features = ixgbe_fix_features,
9263 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
9264 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
9265 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
9266 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
9267 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
9268 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
9269 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
9270 	.ndo_features_check	= ixgbe_features_check,
9271 };
9272 
9273 /**
9274  * ixgbe_enumerate_functions - Get the number of ports this device has
9275  * @adapter: adapter structure
9276  *
9277  * This function enumerates the phsyical functions co-located on a single slot,
9278  * in order to determine how many ports a device has. This is most useful in
9279  * determining the required GT/s of PCIe bandwidth necessary for optimal
9280  * performance.
9281  **/
9282 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
9283 {
9284 	struct pci_dev *entry, *pdev = adapter->pdev;
9285 	int physfns = 0;
9286 
9287 	/* Some cards can not use the generic count PCIe functions method,
9288 	 * because they are behind a parent switch, so we hardcode these with
9289 	 * the correct number of functions.
9290 	 */
9291 	if (ixgbe_pcie_from_parent(&adapter->hw))
9292 		physfns = 4;
9293 
9294 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
9295 		/* don't count virtual functions */
9296 		if (entry->is_virtfn)
9297 			continue;
9298 
9299 		/* When the devices on the bus don't all match our device ID,
9300 		 * we can't reliably determine the correct number of
9301 		 * functions. This can occur if a function has been direct
9302 		 * attached to a virtual machine using VT-d, for example. In
9303 		 * this case, simply return -1 to indicate this.
9304 		 */
9305 		if ((entry->vendor != pdev->vendor) ||
9306 		    (entry->device != pdev->device))
9307 			return -1;
9308 
9309 		physfns++;
9310 	}
9311 
9312 	return physfns;
9313 }
9314 
9315 /**
9316  * ixgbe_wol_supported - Check whether device supports WoL
9317  * @adapter: the adapter private structure
9318  * @device_id: the device ID
9319  * @subdev_id: the subsystem device ID
9320  *
9321  * This function is used by probe and ethtool to determine
9322  * which devices have WoL support
9323  *
9324  **/
9325 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
9326 			 u16 subdevice_id)
9327 {
9328 	struct ixgbe_hw *hw = &adapter->hw;
9329 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
9330 
9331 	/* WOL not supported on 82598 */
9332 	if (hw->mac.type == ixgbe_mac_82598EB)
9333 		return false;
9334 
9335 	/* check eeprom to see if WOL is enabled for X540 and newer */
9336 	if (hw->mac.type >= ixgbe_mac_X540) {
9337 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
9338 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
9339 		     (hw->bus.func == 0)))
9340 			return true;
9341 	}
9342 
9343 	/* WOL is determined based on device IDs for 82599 MACs */
9344 	switch (device_id) {
9345 	case IXGBE_DEV_ID_82599_SFP:
9346 		/* Only these subdevices could supports WOL */
9347 		switch (subdevice_id) {
9348 		case IXGBE_SUBDEV_ID_82599_560FLR:
9349 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
9350 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
9351 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
9352 			/* only support first port */
9353 			if (hw->bus.func != 0)
9354 				break;
9355 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
9356 		case IXGBE_SUBDEV_ID_82599_SFP:
9357 		case IXGBE_SUBDEV_ID_82599_RNDC:
9358 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
9359 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
9360 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
9361 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
9362 			return true;
9363 		}
9364 		break;
9365 	case IXGBE_DEV_ID_82599EN_SFP:
9366 		/* Only these subdevices support WOL */
9367 		switch (subdevice_id) {
9368 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
9369 			return true;
9370 		}
9371 		break;
9372 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
9373 		/* All except this subdevice support WOL */
9374 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
9375 			return true;
9376 		break;
9377 	case IXGBE_DEV_ID_82599_KX4:
9378 		return  true;
9379 	default:
9380 		break;
9381 	}
9382 
9383 	return false;
9384 }
9385 
9386 /**
9387  * ixgbe_probe - Device Initialization Routine
9388  * @pdev: PCI device information struct
9389  * @ent: entry in ixgbe_pci_tbl
9390  *
9391  * Returns 0 on success, negative on failure
9392  *
9393  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
9394  * The OS initialization, configuring of the adapter private structure,
9395  * and a hardware reset occur.
9396  **/
9397 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
9398 {
9399 	struct net_device *netdev;
9400 	struct ixgbe_adapter *adapter = NULL;
9401 	struct ixgbe_hw *hw;
9402 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
9403 	int i, err, pci_using_dac, expected_gts;
9404 	unsigned int indices = MAX_TX_QUEUES;
9405 	u8 part_str[IXGBE_PBANUM_LENGTH];
9406 	bool disable_dev = false;
9407 #ifdef IXGBE_FCOE
9408 	u16 device_caps;
9409 #endif
9410 	u32 eec;
9411 
9412 	/* Catch broken hardware that put the wrong VF device ID in
9413 	 * the PCIe SR-IOV capability.
9414 	 */
9415 	if (pdev->is_virtfn) {
9416 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
9417 		     pci_name(pdev), pdev->vendor, pdev->device);
9418 		return -EINVAL;
9419 	}
9420 
9421 	err = pci_enable_device_mem(pdev);
9422 	if (err)
9423 		return err;
9424 
9425 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
9426 		pci_using_dac = 1;
9427 	} else {
9428 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
9429 		if (err) {
9430 			dev_err(&pdev->dev,
9431 				"No usable DMA configuration, aborting\n");
9432 			goto err_dma;
9433 		}
9434 		pci_using_dac = 0;
9435 	}
9436 
9437 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
9438 	if (err) {
9439 		dev_err(&pdev->dev,
9440 			"pci_request_selected_regions failed 0x%x\n", err);
9441 		goto err_pci_reg;
9442 	}
9443 
9444 	pci_enable_pcie_error_reporting(pdev);
9445 
9446 	pci_set_master(pdev);
9447 	pci_save_state(pdev);
9448 
9449 	if (ii->mac == ixgbe_mac_82598EB) {
9450 #ifdef CONFIG_IXGBE_DCB
9451 		/* 8 TC w/ 4 queues per TC */
9452 		indices = 4 * MAX_TRAFFIC_CLASS;
9453 #else
9454 		indices = IXGBE_MAX_RSS_INDICES;
9455 #endif
9456 	}
9457 
9458 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
9459 	if (!netdev) {
9460 		err = -ENOMEM;
9461 		goto err_alloc_etherdev;
9462 	}
9463 
9464 	SET_NETDEV_DEV(netdev, &pdev->dev);
9465 
9466 	adapter = netdev_priv(netdev);
9467 
9468 	adapter->netdev = netdev;
9469 	adapter->pdev = pdev;
9470 	hw = &adapter->hw;
9471 	hw->back = adapter;
9472 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
9473 
9474 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
9475 			      pci_resource_len(pdev, 0));
9476 	adapter->io_addr = hw->hw_addr;
9477 	if (!hw->hw_addr) {
9478 		err = -EIO;
9479 		goto err_ioremap;
9480 	}
9481 
9482 	netdev->netdev_ops = &ixgbe_netdev_ops;
9483 	ixgbe_set_ethtool_ops(netdev);
9484 	netdev->watchdog_timeo = 5 * HZ;
9485 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
9486 
9487 	/* Setup hw api */
9488 	hw->mac.ops   = *ii->mac_ops;
9489 	hw->mac.type  = ii->mac;
9490 	hw->mvals     = ii->mvals;
9491 
9492 	/* EEPROM */
9493 	hw->eeprom.ops = *ii->eeprom_ops;
9494 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
9495 	if (ixgbe_removed(hw->hw_addr)) {
9496 		err = -EIO;
9497 		goto err_ioremap;
9498 	}
9499 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
9500 	if (!(eec & BIT(8)))
9501 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
9502 
9503 	/* PHY */
9504 	hw->phy.ops = *ii->phy_ops;
9505 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
9506 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
9507 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
9508 	hw->phy.mdio.mmds = 0;
9509 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
9510 	hw->phy.mdio.dev = netdev;
9511 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
9512 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
9513 
9514 	ii->get_invariants(hw);
9515 
9516 	/* setup the private structure */
9517 	err = ixgbe_sw_init(adapter);
9518 	if (err)
9519 		goto err_sw_init;
9520 
9521 	/* Make sure the SWFW semaphore is in a valid state */
9522 	if (hw->mac.ops.init_swfw_sync)
9523 		hw->mac.ops.init_swfw_sync(hw);
9524 
9525 	/* Make it possible the adapter to be woken up via WOL */
9526 	switch (adapter->hw.mac.type) {
9527 	case ixgbe_mac_82599EB:
9528 	case ixgbe_mac_X540:
9529 	case ixgbe_mac_X550:
9530 	case ixgbe_mac_X550EM_x:
9531 	case ixgbe_mac_x550em_a:
9532 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
9533 		break;
9534 	default:
9535 		break;
9536 	}
9537 
9538 	/*
9539 	 * If there is a fan on this device and it has failed log the
9540 	 * failure.
9541 	 */
9542 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
9543 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
9544 		if (esdp & IXGBE_ESDP_SDP1)
9545 			e_crit(probe, "Fan has stopped, replace the adapter\n");
9546 	}
9547 
9548 	if (allow_unsupported_sfp)
9549 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
9550 
9551 	/* reset_hw fills in the perm_addr as well */
9552 	hw->phy.reset_if_overtemp = true;
9553 	err = hw->mac.ops.reset_hw(hw);
9554 	hw->phy.reset_if_overtemp = false;
9555 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
9556 		err = 0;
9557 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
9558 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
9559 		e_dev_err("Reload the driver after installing a supported module.\n");
9560 		goto err_sw_init;
9561 	} else if (err) {
9562 		e_dev_err("HW Init failed: %d\n", err);
9563 		goto err_sw_init;
9564 	}
9565 
9566 #ifdef CONFIG_PCI_IOV
9567 	/* SR-IOV not supported on the 82598 */
9568 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9569 		goto skip_sriov;
9570 	/* Mailbox */
9571 	ixgbe_init_mbx_params_pf(hw);
9572 	hw->mbx.ops = ii->mbx_ops;
9573 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
9574 	ixgbe_enable_sriov(adapter);
9575 skip_sriov:
9576 
9577 #endif
9578 	netdev->features = NETIF_F_SG |
9579 			   NETIF_F_TSO |
9580 			   NETIF_F_TSO6 |
9581 			   NETIF_F_RXHASH |
9582 			   NETIF_F_RXCSUM |
9583 			   NETIF_F_HW_CSUM;
9584 
9585 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
9586 				    NETIF_F_GSO_GRE_CSUM | \
9587 				    NETIF_F_GSO_IPXIP4 | \
9588 				    NETIF_F_GSO_IPXIP6 | \
9589 				    NETIF_F_GSO_UDP_TUNNEL | \
9590 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
9591 
9592 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
9593 	netdev->features |= NETIF_F_GSO_PARTIAL |
9594 			    IXGBE_GSO_PARTIAL_FEATURES;
9595 
9596 	if (hw->mac.type >= ixgbe_mac_82599EB)
9597 		netdev->features |= NETIF_F_SCTP_CRC;
9598 
9599 	/* copy netdev features into list of user selectable features */
9600 	netdev->hw_features |= netdev->features |
9601 			       NETIF_F_HW_VLAN_CTAG_FILTER |
9602 			       NETIF_F_HW_VLAN_CTAG_RX |
9603 			       NETIF_F_HW_VLAN_CTAG_TX |
9604 			       NETIF_F_RXALL |
9605 			       NETIF_F_HW_L2FW_DOFFLOAD;
9606 
9607 	if (hw->mac.type >= ixgbe_mac_82599EB)
9608 		netdev->hw_features |= NETIF_F_NTUPLE |
9609 				       NETIF_F_HW_TC;
9610 
9611 	if (pci_using_dac)
9612 		netdev->features |= NETIF_F_HIGHDMA;
9613 
9614 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
9615 	netdev->hw_enc_features |= netdev->vlan_features;
9616 	netdev->mpls_features |= NETIF_F_HW_CSUM;
9617 
9618 	/* set this bit last since it cannot be part of vlan_features */
9619 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
9620 			    NETIF_F_HW_VLAN_CTAG_RX |
9621 			    NETIF_F_HW_VLAN_CTAG_TX;
9622 
9623 	netdev->priv_flags |= IFF_UNICAST_FLT;
9624 	netdev->priv_flags |= IFF_SUPP_NOFCS;
9625 
9626 #ifdef CONFIG_IXGBE_DCB
9627 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
9628 		netdev->dcbnl_ops = &dcbnl_ops;
9629 #endif
9630 
9631 #ifdef IXGBE_FCOE
9632 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
9633 		unsigned int fcoe_l;
9634 
9635 		if (hw->mac.ops.get_device_caps) {
9636 			hw->mac.ops.get_device_caps(hw, &device_caps);
9637 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
9638 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
9639 		}
9640 
9641 
9642 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
9643 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
9644 
9645 		netdev->features |= NETIF_F_FSO |
9646 				    NETIF_F_FCOE_CRC;
9647 
9648 		netdev->vlan_features |= NETIF_F_FSO |
9649 					 NETIF_F_FCOE_CRC |
9650 					 NETIF_F_FCOE_MTU;
9651 	}
9652 #endif /* IXGBE_FCOE */
9653 
9654 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
9655 		netdev->hw_features |= NETIF_F_LRO;
9656 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9657 		netdev->features |= NETIF_F_LRO;
9658 
9659 	/* make sure the EEPROM is good */
9660 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
9661 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
9662 		err = -EIO;
9663 		goto err_sw_init;
9664 	}
9665 
9666 	eth_platform_get_mac_address(&adapter->pdev->dev,
9667 				     adapter->hw.mac.perm_addr);
9668 
9669 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
9670 
9671 	if (!is_valid_ether_addr(netdev->dev_addr)) {
9672 		e_dev_err("invalid MAC address\n");
9673 		err = -EIO;
9674 		goto err_sw_init;
9675 	}
9676 
9677 	/* Set hw->mac.addr to permanent MAC address */
9678 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
9679 	ixgbe_mac_set_default_filter(adapter);
9680 
9681 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
9682 		    (unsigned long) adapter);
9683 
9684 	if (ixgbe_removed(hw->hw_addr)) {
9685 		err = -EIO;
9686 		goto err_sw_init;
9687 	}
9688 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
9689 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
9690 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
9691 
9692 	err = ixgbe_init_interrupt_scheme(adapter);
9693 	if (err)
9694 		goto err_sw_init;
9695 
9696 	/* WOL not supported for all devices */
9697 	adapter->wol = 0;
9698 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
9699 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
9700 						pdev->subsystem_device);
9701 	if (hw->wol_enabled)
9702 		adapter->wol = IXGBE_WUFC_MAG;
9703 
9704 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
9705 
9706 	/* save off EEPROM version number */
9707 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
9708 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
9709 
9710 	/* pick up the PCI bus settings for reporting later */
9711 	if (ixgbe_pcie_from_parent(hw))
9712 		ixgbe_get_parent_bus_info(adapter);
9713 	else
9714 		 hw->mac.ops.get_bus_info(hw);
9715 
9716 	/* calculate the expected PCIe bandwidth required for optimal
9717 	 * performance. Note that some older parts will never have enough
9718 	 * bandwidth due to being older generation PCIe parts. We clamp these
9719 	 * parts to ensure no warning is displayed if it can't be fixed.
9720 	 */
9721 	switch (hw->mac.type) {
9722 	case ixgbe_mac_82598EB:
9723 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
9724 		break;
9725 	default:
9726 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
9727 		break;
9728 	}
9729 
9730 	/* don't check link if we failed to enumerate functions */
9731 	if (expected_gts > 0)
9732 		ixgbe_check_minimum_link(adapter, expected_gts);
9733 
9734 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
9735 	if (err)
9736 		strlcpy(part_str, "Unknown", sizeof(part_str));
9737 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
9738 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
9739 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
9740 			   part_str);
9741 	else
9742 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
9743 			   hw->mac.type, hw->phy.type, part_str);
9744 
9745 	e_dev_info("%pM\n", netdev->dev_addr);
9746 
9747 	/* reset the hardware with the new settings */
9748 	err = hw->mac.ops.start_hw(hw);
9749 	if (err == IXGBE_ERR_EEPROM_VERSION) {
9750 		/* We are running on a pre-production device, log a warning */
9751 		e_dev_warn("This device is a pre-production adapter/LOM. "
9752 			   "Please be aware there may be issues associated "
9753 			   "with your hardware.  If you are experiencing "
9754 			   "problems please contact your Intel or hardware "
9755 			   "representative who provided you with this "
9756 			   "hardware.\n");
9757 	}
9758 	strcpy(netdev->name, "eth%d");
9759 	err = register_netdev(netdev);
9760 	if (err)
9761 		goto err_register;
9762 
9763 	pci_set_drvdata(pdev, adapter);
9764 
9765 	/* power down the optics for 82599 SFP+ fiber */
9766 	if (hw->mac.ops.disable_tx_laser)
9767 		hw->mac.ops.disable_tx_laser(hw);
9768 
9769 	/* carrier off reporting is important to ethtool even BEFORE open */
9770 	netif_carrier_off(netdev);
9771 
9772 #ifdef CONFIG_IXGBE_DCA
9773 	if (dca_add_requester(&pdev->dev) == 0) {
9774 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
9775 		ixgbe_setup_dca(adapter);
9776 	}
9777 #endif
9778 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
9779 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
9780 		for (i = 0; i < adapter->num_vfs; i++)
9781 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
9782 	}
9783 
9784 	/* firmware requires driver version to be 0xFFFFFFFF
9785 	 * since os does not support feature
9786 	 */
9787 	if (hw->mac.ops.set_fw_drv_ver)
9788 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
9789 					   0xFF);
9790 
9791 	/* add san mac addr to netdev */
9792 	ixgbe_add_sanmac_netdev(netdev);
9793 
9794 	e_dev_info("%s\n", ixgbe_default_device_descr);
9795 
9796 #ifdef CONFIG_IXGBE_HWMON
9797 	if (ixgbe_sysfs_init(adapter))
9798 		e_err(probe, "failed to allocate sysfs resources\n");
9799 #endif /* CONFIG_IXGBE_HWMON */
9800 
9801 	ixgbe_dbg_adapter_init(adapter);
9802 
9803 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
9804 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
9805 		hw->mac.ops.setup_link(hw,
9806 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
9807 			true);
9808 
9809 	return 0;
9810 
9811 err_register:
9812 	ixgbe_release_hw_control(adapter);
9813 	ixgbe_clear_interrupt_scheme(adapter);
9814 err_sw_init:
9815 	ixgbe_disable_sriov(adapter);
9816 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
9817 	iounmap(adapter->io_addr);
9818 	kfree(adapter->jump_tables[0]);
9819 	kfree(adapter->mac_table);
9820 err_ioremap:
9821 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9822 	free_netdev(netdev);
9823 err_alloc_etherdev:
9824 	pci_release_mem_regions(pdev);
9825 err_pci_reg:
9826 err_dma:
9827 	if (!adapter || disable_dev)
9828 		pci_disable_device(pdev);
9829 	return err;
9830 }
9831 
9832 /**
9833  * ixgbe_remove - Device Removal Routine
9834  * @pdev: PCI device information struct
9835  *
9836  * ixgbe_remove is called by the PCI subsystem to alert the driver
9837  * that it should release a PCI device.  The could be caused by a
9838  * Hot-Plug event, or because the driver is going to be removed from
9839  * memory.
9840  **/
9841 static void ixgbe_remove(struct pci_dev *pdev)
9842 {
9843 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9844 	struct net_device *netdev;
9845 	bool disable_dev;
9846 	int i;
9847 
9848 	/* if !adapter then we already cleaned up in probe */
9849 	if (!adapter)
9850 		return;
9851 
9852 	netdev  = adapter->netdev;
9853 	ixgbe_dbg_adapter_exit(adapter);
9854 
9855 	set_bit(__IXGBE_REMOVING, &adapter->state);
9856 	cancel_work_sync(&adapter->service_task);
9857 
9858 
9859 #ifdef CONFIG_IXGBE_DCA
9860 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
9861 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
9862 		dca_remove_requester(&pdev->dev);
9863 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
9864 				IXGBE_DCA_CTRL_DCA_DISABLE);
9865 	}
9866 
9867 #endif
9868 #ifdef CONFIG_IXGBE_HWMON
9869 	ixgbe_sysfs_exit(adapter);
9870 #endif /* CONFIG_IXGBE_HWMON */
9871 
9872 	/* remove the added san mac */
9873 	ixgbe_del_sanmac_netdev(netdev);
9874 
9875 #ifdef CONFIG_PCI_IOV
9876 	ixgbe_disable_sriov(adapter);
9877 #endif
9878 	if (netdev->reg_state == NETREG_REGISTERED)
9879 		unregister_netdev(netdev);
9880 
9881 	ixgbe_clear_interrupt_scheme(adapter);
9882 
9883 	ixgbe_release_hw_control(adapter);
9884 
9885 #ifdef CONFIG_DCB
9886 	kfree(adapter->ixgbe_ieee_pfc);
9887 	kfree(adapter->ixgbe_ieee_ets);
9888 
9889 #endif
9890 	iounmap(adapter->io_addr);
9891 	pci_release_mem_regions(pdev);
9892 
9893 	e_dev_info("complete\n");
9894 
9895 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
9896 		if (adapter->jump_tables[i]) {
9897 			kfree(adapter->jump_tables[i]->input);
9898 			kfree(adapter->jump_tables[i]->mask);
9899 		}
9900 		kfree(adapter->jump_tables[i]);
9901 	}
9902 
9903 	kfree(adapter->mac_table);
9904 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9905 	free_netdev(netdev);
9906 
9907 	pci_disable_pcie_error_reporting(pdev);
9908 
9909 	if (disable_dev)
9910 		pci_disable_device(pdev);
9911 }
9912 
9913 /**
9914  * ixgbe_io_error_detected - called when PCI error is detected
9915  * @pdev: Pointer to PCI device
9916  * @state: The current pci connection state
9917  *
9918  * This function is called after a PCI bus error affecting
9919  * this device has been detected.
9920  */
9921 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
9922 						pci_channel_state_t state)
9923 {
9924 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9925 	struct net_device *netdev = adapter->netdev;
9926 
9927 #ifdef CONFIG_PCI_IOV
9928 	struct ixgbe_hw *hw = &adapter->hw;
9929 	struct pci_dev *bdev, *vfdev;
9930 	u32 dw0, dw1, dw2, dw3;
9931 	int vf, pos;
9932 	u16 req_id, pf_func;
9933 
9934 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
9935 	    adapter->num_vfs == 0)
9936 		goto skip_bad_vf_detection;
9937 
9938 	bdev = pdev->bus->self;
9939 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
9940 		bdev = bdev->bus->self;
9941 
9942 	if (!bdev)
9943 		goto skip_bad_vf_detection;
9944 
9945 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
9946 	if (!pos)
9947 		goto skip_bad_vf_detection;
9948 
9949 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
9950 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
9951 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
9952 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
9953 	if (ixgbe_removed(hw->hw_addr))
9954 		goto skip_bad_vf_detection;
9955 
9956 	req_id = dw1 >> 16;
9957 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
9958 	if (!(req_id & 0x0080))
9959 		goto skip_bad_vf_detection;
9960 
9961 	pf_func = req_id & 0x01;
9962 	if ((pf_func & 1) == (pdev->devfn & 1)) {
9963 		unsigned int device_id;
9964 
9965 		vf = (req_id & 0x7F) >> 1;
9966 		e_dev_err("VF %d has caused a PCIe error\n", vf);
9967 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
9968 				"%8.8x\tdw3: %8.8x\n",
9969 		dw0, dw1, dw2, dw3);
9970 		switch (adapter->hw.mac.type) {
9971 		case ixgbe_mac_82599EB:
9972 			device_id = IXGBE_82599_VF_DEVICE_ID;
9973 			break;
9974 		case ixgbe_mac_X540:
9975 			device_id = IXGBE_X540_VF_DEVICE_ID;
9976 			break;
9977 		case ixgbe_mac_X550:
9978 			device_id = IXGBE_DEV_ID_X550_VF;
9979 			break;
9980 		case ixgbe_mac_X550EM_x:
9981 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
9982 			break;
9983 		case ixgbe_mac_x550em_a:
9984 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
9985 			break;
9986 		default:
9987 			device_id = 0;
9988 			break;
9989 		}
9990 
9991 		/* Find the pci device of the offending VF */
9992 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
9993 		while (vfdev) {
9994 			if (vfdev->devfn == (req_id & 0xFF))
9995 				break;
9996 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
9997 					       device_id, vfdev);
9998 		}
9999 		/*
10000 		 * There's a slim chance the VF could have been hot plugged,
10001 		 * so if it is no longer present we don't need to issue the
10002 		 * VFLR.  Just clean up the AER in that case.
10003 		 */
10004 		if (vfdev) {
10005 			ixgbe_issue_vf_flr(adapter, vfdev);
10006 			/* Free device reference count */
10007 			pci_dev_put(vfdev);
10008 		}
10009 
10010 		pci_cleanup_aer_uncorrect_error_status(pdev);
10011 	}
10012 
10013 	/*
10014 	 * Even though the error may have occurred on the other port
10015 	 * we still need to increment the vf error reference count for
10016 	 * both ports because the I/O resume function will be called
10017 	 * for both of them.
10018 	 */
10019 	adapter->vferr_refcount++;
10020 
10021 	return PCI_ERS_RESULT_RECOVERED;
10022 
10023 skip_bad_vf_detection:
10024 #endif /* CONFIG_PCI_IOV */
10025 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
10026 		return PCI_ERS_RESULT_DISCONNECT;
10027 
10028 	rtnl_lock();
10029 	netif_device_detach(netdev);
10030 
10031 	if (state == pci_channel_io_perm_failure) {
10032 		rtnl_unlock();
10033 		return PCI_ERS_RESULT_DISCONNECT;
10034 	}
10035 
10036 	if (netif_running(netdev))
10037 		ixgbe_down(adapter);
10038 
10039 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
10040 		pci_disable_device(pdev);
10041 	rtnl_unlock();
10042 
10043 	/* Request a slot reset. */
10044 	return PCI_ERS_RESULT_NEED_RESET;
10045 }
10046 
10047 /**
10048  * ixgbe_io_slot_reset - called after the pci bus has been reset.
10049  * @pdev: Pointer to PCI device
10050  *
10051  * Restart the card from scratch, as if from a cold-boot.
10052  */
10053 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
10054 {
10055 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10056 	pci_ers_result_t result;
10057 	int err;
10058 
10059 	if (pci_enable_device_mem(pdev)) {
10060 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
10061 		result = PCI_ERS_RESULT_DISCONNECT;
10062 	} else {
10063 		smp_mb__before_atomic();
10064 		clear_bit(__IXGBE_DISABLED, &adapter->state);
10065 		adapter->hw.hw_addr = adapter->io_addr;
10066 		pci_set_master(pdev);
10067 		pci_restore_state(pdev);
10068 		pci_save_state(pdev);
10069 
10070 		pci_wake_from_d3(pdev, false);
10071 
10072 		ixgbe_reset(adapter);
10073 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10074 		result = PCI_ERS_RESULT_RECOVERED;
10075 	}
10076 
10077 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10078 	if (err) {
10079 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
10080 			  "failed 0x%0x\n", err);
10081 		/* non-fatal, continue */
10082 	}
10083 
10084 	return result;
10085 }
10086 
10087 /**
10088  * ixgbe_io_resume - called when traffic can start flowing again.
10089  * @pdev: Pointer to PCI device
10090  *
10091  * This callback is called when the error recovery driver tells us that
10092  * its OK to resume normal operation.
10093  */
10094 static void ixgbe_io_resume(struct pci_dev *pdev)
10095 {
10096 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10097 	struct net_device *netdev = adapter->netdev;
10098 
10099 #ifdef CONFIG_PCI_IOV
10100 	if (adapter->vferr_refcount) {
10101 		e_info(drv, "Resuming after VF err\n");
10102 		adapter->vferr_refcount--;
10103 		return;
10104 	}
10105 
10106 #endif
10107 	if (netif_running(netdev))
10108 		ixgbe_up(adapter);
10109 
10110 	netif_device_attach(netdev);
10111 }
10112 
10113 static const struct pci_error_handlers ixgbe_err_handler = {
10114 	.error_detected = ixgbe_io_error_detected,
10115 	.slot_reset = ixgbe_io_slot_reset,
10116 	.resume = ixgbe_io_resume,
10117 };
10118 
10119 static struct pci_driver ixgbe_driver = {
10120 	.name     = ixgbe_driver_name,
10121 	.id_table = ixgbe_pci_tbl,
10122 	.probe    = ixgbe_probe,
10123 	.remove   = ixgbe_remove,
10124 #ifdef CONFIG_PM
10125 	.suspend  = ixgbe_suspend,
10126 	.resume   = ixgbe_resume,
10127 #endif
10128 	.shutdown = ixgbe_shutdown,
10129 	.sriov_configure = ixgbe_pci_sriov_configure,
10130 	.err_handler = &ixgbe_err_handler
10131 };
10132 
10133 /**
10134  * ixgbe_init_module - Driver Registration Routine
10135  *
10136  * ixgbe_init_module is the first routine called when the driver is
10137  * loaded. All it does is register with the PCI subsystem.
10138  **/
10139 static int __init ixgbe_init_module(void)
10140 {
10141 	int ret;
10142 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
10143 	pr_info("%s\n", ixgbe_copyright);
10144 
10145 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
10146 	if (!ixgbe_wq) {
10147 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
10148 		return -ENOMEM;
10149 	}
10150 
10151 	ixgbe_dbg_init();
10152 
10153 	ret = pci_register_driver(&ixgbe_driver);
10154 	if (ret) {
10155 		destroy_workqueue(ixgbe_wq);
10156 		ixgbe_dbg_exit();
10157 		return ret;
10158 	}
10159 
10160 #ifdef CONFIG_IXGBE_DCA
10161 	dca_register_notify(&dca_notifier);
10162 #endif
10163 
10164 	return 0;
10165 }
10166 
10167 module_init(ixgbe_init_module);
10168 
10169 /**
10170  * ixgbe_exit_module - Driver Exit Cleanup Routine
10171  *
10172  * ixgbe_exit_module is called just before the driver is removed
10173  * from memory.
10174  **/
10175 static void __exit ixgbe_exit_module(void)
10176 {
10177 #ifdef CONFIG_IXGBE_DCA
10178 	dca_unregister_notify(&dca_notifier);
10179 #endif
10180 	pci_unregister_driver(&ixgbe_driver);
10181 
10182 	ixgbe_dbg_exit();
10183 	if (ixgbe_wq) {
10184 		destroy_workqueue(ixgbe_wq);
10185 		ixgbe_wq = NULL;
10186 	}
10187 }
10188 
10189 #ifdef CONFIG_IXGBE_DCA
10190 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
10191 			    void *p)
10192 {
10193 	int ret_val;
10194 
10195 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
10196 					 __ixgbe_notify_dca);
10197 
10198 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
10199 }
10200 
10201 #endif /* CONFIG_IXGBE_DCA */
10202 
10203 module_exit(ixgbe_exit_module);
10204 
10205 /* ixgbe_main.c */
10206