xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 04d8a0a5f3b6887543850d991a5e37c4ec90e250)
1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2016 Intel Corporation.
5 
6   This program is free software; you can redistribute it and/or modify it
7   under the terms and conditions of the GNU General Public License,
8   version 2, as published by the Free Software Foundation.
9 
10   This program is distributed in the hope it will be useful, but WITHOUT
11   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13   more details.
14 
15   You should have received a copy of the GNU General Public License along with
16   this program; if not, write to the Free Software Foundation, Inc.,
17   51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 
19   The full GNU General Public License is included in this distribution in
20   the file called "COPYING".
21 
22   Contact Information:
23   Linux NICS <linux.nics@intel.com>
24   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 
27 *******************************************************************************/
28 
29 #include <linux/types.h>
30 #include <linux/module.h>
31 #include <linux/pci.h>
32 #include <linux/netdevice.h>
33 #include <linux/vmalloc.h>
34 #include <linux/string.h>
35 #include <linux/in.h>
36 #include <linux/interrupt.h>
37 #include <linux/ip.h>
38 #include <linux/tcp.h>
39 #include <linux/sctp.h>
40 #include <linux/pkt_sched.h>
41 #include <linux/ipv6.h>
42 #include <linux/slab.h>
43 #include <net/checksum.h>
44 #include <net/ip6_checksum.h>
45 #include <linux/etherdevice.h>
46 #include <linux/ethtool.h>
47 #include <linux/if.h>
48 #include <linux/if_vlan.h>
49 #include <linux/if_macvlan.h>
50 #include <linux/if_bridge.h>
51 #include <linux/prefetch.h>
52 #include <scsi/fc/fc_fcoe.h>
53 #include <net/udp_tunnel.h>
54 #include <net/pkt_cls.h>
55 #include <net/tc_act/tc_gact.h>
56 #include <net/tc_act/tc_mirred.h>
57 #include <net/vxlan.h>
58 
59 #include "ixgbe.h"
60 #include "ixgbe_common.h"
61 #include "ixgbe_dcb_82599.h"
62 #include "ixgbe_sriov.h"
63 #include "ixgbe_model.h"
64 
65 char ixgbe_driver_name[] = "ixgbe";
66 static const char ixgbe_driver_string[] =
67 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
68 #ifdef IXGBE_FCOE
69 char ixgbe_default_device_descr[] =
70 			      "Intel(R) 10 Gigabit Network Connection";
71 #else
72 static char ixgbe_default_device_descr[] =
73 			      "Intel(R) 10 Gigabit Network Connection";
74 #endif
75 #define DRV_VERSION "4.4.0-k"
76 const char ixgbe_driver_version[] = DRV_VERSION;
77 static const char ixgbe_copyright[] =
78 				"Copyright (c) 1999-2016 Intel Corporation.";
79 
80 static const char ixgbe_overheat_msg[] = "Network adapter has been stopped because it has over heated. Restart the computer. If the problem persists, power off the system and replace the adapter";
81 
82 static const struct ixgbe_info *ixgbe_info_tbl[] = {
83 	[board_82598]		= &ixgbe_82598_info,
84 	[board_82599]		= &ixgbe_82599_info,
85 	[board_X540]		= &ixgbe_X540_info,
86 	[board_X550]		= &ixgbe_X550_info,
87 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
88 	[board_x550em_a]	= &ixgbe_x550em_a_info,
89 	[board_x550em_a_fw]	= &ixgbe_x550em_a_fw_info,
90 };
91 
92 /* ixgbe_pci_tbl - PCI Device ID Table
93  *
94  * Wildcard entries (PCI_ANY_ID) should come last
95  * Last entry must be all 0s
96  *
97  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
98  *   Class, Class Mask, private data (not used) }
99  */
100 static const struct pci_device_id ixgbe_pci_tbl[] = {
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
133 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
134 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
135 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
136 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
137 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
138 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
139 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
140 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
141 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
142 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
143 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
144 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
145 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
146 	/* required last entry */
147 	{0, }
148 };
149 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
150 
151 #ifdef CONFIG_IXGBE_DCA
152 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
153 			    void *p);
154 static struct notifier_block dca_notifier = {
155 	.notifier_call = ixgbe_notify_dca,
156 	.next          = NULL,
157 	.priority      = 0
158 };
159 #endif
160 
161 #ifdef CONFIG_PCI_IOV
162 static unsigned int max_vfs;
163 module_param(max_vfs, uint, 0);
164 MODULE_PARM_DESC(max_vfs,
165 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
166 #endif /* CONFIG_PCI_IOV */
167 
168 static unsigned int allow_unsupported_sfp;
169 module_param(allow_unsupported_sfp, uint, 0);
170 MODULE_PARM_DESC(allow_unsupported_sfp,
171 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
172 
173 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
174 static int debug = -1;
175 module_param(debug, int, 0);
176 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
177 
178 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
179 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
180 MODULE_LICENSE("GPL");
181 MODULE_VERSION(DRV_VERSION);
182 
183 static struct workqueue_struct *ixgbe_wq;
184 
185 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
186 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
187 
188 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
189 					  u32 reg, u16 *value)
190 {
191 	struct pci_dev *parent_dev;
192 	struct pci_bus *parent_bus;
193 
194 	parent_bus = adapter->pdev->bus->parent;
195 	if (!parent_bus)
196 		return -1;
197 
198 	parent_dev = parent_bus->self;
199 	if (!parent_dev)
200 		return -1;
201 
202 	if (!pci_is_pcie(parent_dev))
203 		return -1;
204 
205 	pcie_capability_read_word(parent_dev, reg, value);
206 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
207 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
208 		return -1;
209 	return 0;
210 }
211 
212 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
213 {
214 	struct ixgbe_hw *hw = &adapter->hw;
215 	u16 link_status = 0;
216 	int err;
217 
218 	hw->bus.type = ixgbe_bus_type_pci_express;
219 
220 	/* Get the negotiated link width and speed from PCI config space of the
221 	 * parent, as this device is behind a switch
222 	 */
223 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
224 
225 	/* assume caller will handle error case */
226 	if (err)
227 		return err;
228 
229 	hw->bus.width = ixgbe_convert_bus_width(link_status);
230 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
231 
232 	return 0;
233 }
234 
235 /**
236  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
237  * @hw: hw specific details
238  *
239  * This function is used by probe to determine whether a device's PCI-Express
240  * bandwidth details should be gathered from the parent bus instead of from the
241  * device. Used to ensure that various locations all have the correct device ID
242  * checks.
243  */
244 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
245 {
246 	switch (hw->device_id) {
247 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
248 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
249 		return true;
250 	default:
251 		return false;
252 	}
253 }
254 
255 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
256 				     int expected_gts)
257 {
258 	struct ixgbe_hw *hw = &adapter->hw;
259 	int max_gts = 0;
260 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
261 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
262 	struct pci_dev *pdev;
263 
264 	/* Some devices are not connected over PCIe and thus do not negotiate
265 	 * speed. These devices do not have valid bus info, and thus any report
266 	 * we generate may not be correct.
267 	 */
268 	if (hw->bus.type == ixgbe_bus_type_internal)
269 		return;
270 
271 	/* determine whether to use the parent device */
272 	if (ixgbe_pcie_from_parent(&adapter->hw))
273 		pdev = adapter->pdev->bus->parent->self;
274 	else
275 		pdev = adapter->pdev;
276 
277 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
278 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
279 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
280 		return;
281 	}
282 
283 	switch (speed) {
284 	case PCIE_SPEED_2_5GT:
285 		/* 8b/10b encoding reduces max throughput by 20% */
286 		max_gts = 2 * width;
287 		break;
288 	case PCIE_SPEED_5_0GT:
289 		/* 8b/10b encoding reduces max throughput by 20% */
290 		max_gts = 4 * width;
291 		break;
292 	case PCIE_SPEED_8_0GT:
293 		/* 128b/130b encoding reduces throughput by less than 2% */
294 		max_gts = 8 * width;
295 		break;
296 	default:
297 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
298 		return;
299 	}
300 
301 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
302 		   max_gts);
303 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
304 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
305 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
306 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
307 		    "Unknown"),
308 		   width,
309 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
310 		    speed == PCIE_SPEED_5_0GT ? "20%" :
311 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
312 		    "Unknown"));
313 
314 	if (max_gts < expected_gts) {
315 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
316 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
317 			expected_gts);
318 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
319 	}
320 }
321 
322 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
323 {
324 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
325 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
326 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
327 		queue_work(ixgbe_wq, &adapter->service_task);
328 }
329 
330 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
331 {
332 	struct ixgbe_adapter *adapter = hw->back;
333 
334 	if (!hw->hw_addr)
335 		return;
336 	hw->hw_addr = NULL;
337 	e_dev_err("Adapter removed\n");
338 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
339 		ixgbe_service_event_schedule(adapter);
340 }
341 
342 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
343 {
344 	u32 value;
345 
346 	/* The following check not only optimizes a bit by not
347 	 * performing a read on the status register when the
348 	 * register just read was a status register read that
349 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
350 	 * potential recursion.
351 	 */
352 	if (reg == IXGBE_STATUS) {
353 		ixgbe_remove_adapter(hw);
354 		return;
355 	}
356 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
357 	if (value == IXGBE_FAILED_READ_REG)
358 		ixgbe_remove_adapter(hw);
359 }
360 
361 /**
362  * ixgbe_read_reg - Read from device register
363  * @hw: hw specific details
364  * @reg: offset of register to read
365  *
366  * Returns : value read or IXGBE_FAILED_READ_REG if removed
367  *
368  * This function is used to read device registers. It checks for device
369  * removal by confirming any read that returns all ones by checking the
370  * status register value for all ones. This function avoids reading from
371  * the hardware if a removal was previously detected in which case it
372  * returns IXGBE_FAILED_READ_REG (all ones).
373  */
374 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
375 {
376 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
377 	u32 value;
378 
379 	if (ixgbe_removed(reg_addr))
380 		return IXGBE_FAILED_READ_REG;
381 	if (unlikely(hw->phy.nw_mng_if_sel &
382 		     IXGBE_NW_MNG_IF_SEL_ENABLE_10_100M)) {
383 		struct ixgbe_adapter *adapter;
384 		int i;
385 
386 		for (i = 0; i < 200; ++i) {
387 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
388 			if (likely(!value))
389 				goto writes_completed;
390 			if (value == IXGBE_FAILED_READ_REG) {
391 				ixgbe_remove_adapter(hw);
392 				return IXGBE_FAILED_READ_REG;
393 			}
394 			udelay(5);
395 		}
396 
397 		adapter = hw->back;
398 		e_warn(hw, "register writes incomplete %08x\n", value);
399 	}
400 
401 writes_completed:
402 	value = readl(reg_addr + reg);
403 	if (unlikely(value == IXGBE_FAILED_READ_REG))
404 		ixgbe_check_remove(hw, reg);
405 	return value;
406 }
407 
408 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
409 {
410 	u16 value;
411 
412 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
413 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
414 		ixgbe_remove_adapter(hw);
415 		return true;
416 	}
417 	return false;
418 }
419 
420 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
421 {
422 	struct ixgbe_adapter *adapter = hw->back;
423 	u16 value;
424 
425 	if (ixgbe_removed(hw->hw_addr))
426 		return IXGBE_FAILED_READ_CFG_WORD;
427 	pci_read_config_word(adapter->pdev, reg, &value);
428 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
429 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
430 		return IXGBE_FAILED_READ_CFG_WORD;
431 	return value;
432 }
433 
434 #ifdef CONFIG_PCI_IOV
435 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
436 {
437 	struct ixgbe_adapter *adapter = hw->back;
438 	u32 value;
439 
440 	if (ixgbe_removed(hw->hw_addr))
441 		return IXGBE_FAILED_READ_CFG_DWORD;
442 	pci_read_config_dword(adapter->pdev, reg, &value);
443 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
444 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
445 		return IXGBE_FAILED_READ_CFG_DWORD;
446 	return value;
447 }
448 #endif /* CONFIG_PCI_IOV */
449 
450 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
451 {
452 	struct ixgbe_adapter *adapter = hw->back;
453 
454 	if (ixgbe_removed(hw->hw_addr))
455 		return;
456 	pci_write_config_word(adapter->pdev, reg, value);
457 }
458 
459 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
460 {
461 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
462 
463 	/* flush memory to make sure state is correct before next watchdog */
464 	smp_mb__before_atomic();
465 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
466 }
467 
468 struct ixgbe_reg_info {
469 	u32 ofs;
470 	char *name;
471 };
472 
473 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
474 
475 	/* General Registers */
476 	{IXGBE_CTRL, "CTRL"},
477 	{IXGBE_STATUS, "STATUS"},
478 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
479 
480 	/* Interrupt Registers */
481 	{IXGBE_EICR, "EICR"},
482 
483 	/* RX Registers */
484 	{IXGBE_SRRCTL(0), "SRRCTL"},
485 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
486 	{IXGBE_RDLEN(0), "RDLEN"},
487 	{IXGBE_RDH(0), "RDH"},
488 	{IXGBE_RDT(0), "RDT"},
489 	{IXGBE_RXDCTL(0), "RXDCTL"},
490 	{IXGBE_RDBAL(0), "RDBAL"},
491 	{IXGBE_RDBAH(0), "RDBAH"},
492 
493 	/* TX Registers */
494 	{IXGBE_TDBAL(0), "TDBAL"},
495 	{IXGBE_TDBAH(0), "TDBAH"},
496 	{IXGBE_TDLEN(0), "TDLEN"},
497 	{IXGBE_TDH(0), "TDH"},
498 	{IXGBE_TDT(0), "TDT"},
499 	{IXGBE_TXDCTL(0), "TXDCTL"},
500 
501 	/* List Terminator */
502 	{ .name = NULL }
503 };
504 
505 
506 /*
507  * ixgbe_regdump - register printout routine
508  */
509 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
510 {
511 	int i = 0, j = 0;
512 	char rname[16];
513 	u32 regs[64];
514 
515 	switch (reginfo->ofs) {
516 	case IXGBE_SRRCTL(0):
517 		for (i = 0; i < 64; i++)
518 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
519 		break;
520 	case IXGBE_DCA_RXCTRL(0):
521 		for (i = 0; i < 64; i++)
522 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
523 		break;
524 	case IXGBE_RDLEN(0):
525 		for (i = 0; i < 64; i++)
526 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
527 		break;
528 	case IXGBE_RDH(0):
529 		for (i = 0; i < 64; i++)
530 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
531 		break;
532 	case IXGBE_RDT(0):
533 		for (i = 0; i < 64; i++)
534 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
535 		break;
536 	case IXGBE_RXDCTL(0):
537 		for (i = 0; i < 64; i++)
538 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
539 		break;
540 	case IXGBE_RDBAL(0):
541 		for (i = 0; i < 64; i++)
542 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
543 		break;
544 	case IXGBE_RDBAH(0):
545 		for (i = 0; i < 64; i++)
546 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
547 		break;
548 	case IXGBE_TDBAL(0):
549 		for (i = 0; i < 64; i++)
550 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
551 		break;
552 	case IXGBE_TDBAH(0):
553 		for (i = 0; i < 64; i++)
554 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
555 		break;
556 	case IXGBE_TDLEN(0):
557 		for (i = 0; i < 64; i++)
558 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
559 		break;
560 	case IXGBE_TDH(0):
561 		for (i = 0; i < 64; i++)
562 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
563 		break;
564 	case IXGBE_TDT(0):
565 		for (i = 0; i < 64; i++)
566 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
567 		break;
568 	case IXGBE_TXDCTL(0):
569 		for (i = 0; i < 64; i++)
570 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
571 		break;
572 	default:
573 		pr_info("%-15s %08x\n", reginfo->name,
574 			IXGBE_READ_REG(hw, reginfo->ofs));
575 		return;
576 	}
577 
578 	for (i = 0; i < 8; i++) {
579 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
580 		pr_err("%-15s", rname);
581 		for (j = 0; j < 8; j++)
582 			pr_cont(" %08x", regs[i*8+j]);
583 		pr_cont("\n");
584 	}
585 
586 }
587 
588 /*
589  * ixgbe_dump - Print registers, tx-rings and rx-rings
590  */
591 static void ixgbe_dump(struct ixgbe_adapter *adapter)
592 {
593 	struct net_device *netdev = adapter->netdev;
594 	struct ixgbe_hw *hw = &adapter->hw;
595 	struct ixgbe_reg_info *reginfo;
596 	int n = 0;
597 	struct ixgbe_ring *tx_ring;
598 	struct ixgbe_tx_buffer *tx_buffer;
599 	union ixgbe_adv_tx_desc *tx_desc;
600 	struct my_u0 { u64 a; u64 b; } *u0;
601 	struct ixgbe_ring *rx_ring;
602 	union ixgbe_adv_rx_desc *rx_desc;
603 	struct ixgbe_rx_buffer *rx_buffer_info;
604 	u32 staterr;
605 	int i = 0;
606 
607 	if (!netif_msg_hw(adapter))
608 		return;
609 
610 	/* Print netdevice Info */
611 	if (netdev) {
612 		dev_info(&adapter->pdev->dev, "Net device Info\n");
613 		pr_info("Device Name     state            "
614 			"trans_start\n");
615 		pr_info("%-15s %016lX %016lX\n",
616 			netdev->name,
617 			netdev->state,
618 			dev_trans_start(netdev));
619 	}
620 
621 	/* Print Registers */
622 	dev_info(&adapter->pdev->dev, "Register Dump\n");
623 	pr_info(" Register Name   Value\n");
624 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
625 	     reginfo->name; reginfo++) {
626 		ixgbe_regdump(hw, reginfo);
627 	}
628 
629 	/* Print TX Ring Summary */
630 	if (!netdev || !netif_running(netdev))
631 		return;
632 
633 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
634 	pr_info(" %s     %s              %s        %s\n",
635 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
636 		"leng", "ntw", "timestamp");
637 	for (n = 0; n < adapter->num_tx_queues; n++) {
638 		tx_ring = adapter->tx_ring[n];
639 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
640 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
641 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
642 			   (u64)dma_unmap_addr(tx_buffer, dma),
643 			   dma_unmap_len(tx_buffer, len),
644 			   tx_buffer->next_to_watch,
645 			   (u64)tx_buffer->time_stamp);
646 	}
647 
648 	/* Print TX Rings */
649 	if (!netif_msg_tx_done(adapter))
650 		goto rx_ring_summary;
651 
652 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
653 
654 	/* Transmit Descriptor Formats
655 	 *
656 	 * 82598 Advanced Transmit Descriptor
657 	 *   +--------------------------------------------------------------+
658 	 * 0 |         Buffer Address [63:0]                                |
659 	 *   +--------------------------------------------------------------+
660 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
661 	 *   +--------------------------------------------------------------+
662 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
663 	 *
664 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
665 	 *   +--------------------------------------------------------------+
666 	 * 0 |                          RSV [63:0]                          |
667 	 *   +--------------------------------------------------------------+
668 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
669 	 *   +--------------------------------------------------------------+
670 	 *   63                       36 35   32 31                         0
671 	 *
672 	 * 82599+ Advanced Transmit Descriptor
673 	 *   +--------------------------------------------------------------+
674 	 * 0 |         Buffer Address [63:0]                                |
675 	 *   +--------------------------------------------------------------+
676 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
677 	 *   +--------------------------------------------------------------+
678 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
679 	 *
680 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
681 	 *   +--------------------------------------------------------------+
682 	 * 0 |                          RSV [63:0]                          |
683 	 *   +--------------------------------------------------------------+
684 	 * 8 |            RSV           |  STA  |           RSV             |
685 	 *   +--------------------------------------------------------------+
686 	 *   63                       36 35   32 31                         0
687 	 */
688 
689 	for (n = 0; n < adapter->num_tx_queues; n++) {
690 		tx_ring = adapter->tx_ring[n];
691 		pr_info("------------------------------------\n");
692 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
693 		pr_info("------------------------------------\n");
694 		pr_info("%s%s    %s              %s        %s          %s\n",
695 			"T [desc]     [address 63:0  ] ",
696 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
697 			"leng", "ntw", "timestamp", "bi->skb");
698 
699 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
700 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
701 			tx_buffer = &tx_ring->tx_buffer_info[i];
702 			u0 = (struct my_u0 *)tx_desc;
703 			if (dma_unmap_len(tx_buffer, len) > 0) {
704 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
705 					i,
706 					le64_to_cpu(u0->a),
707 					le64_to_cpu(u0->b),
708 					(u64)dma_unmap_addr(tx_buffer, dma),
709 					dma_unmap_len(tx_buffer, len),
710 					tx_buffer->next_to_watch,
711 					(u64)tx_buffer->time_stamp,
712 					tx_buffer->skb);
713 				if (i == tx_ring->next_to_use &&
714 					i == tx_ring->next_to_clean)
715 					pr_cont(" NTC/U\n");
716 				else if (i == tx_ring->next_to_use)
717 					pr_cont(" NTU\n");
718 				else if (i == tx_ring->next_to_clean)
719 					pr_cont(" NTC\n");
720 				else
721 					pr_cont("\n");
722 
723 				if (netif_msg_pktdata(adapter) &&
724 				    tx_buffer->skb)
725 					print_hex_dump(KERN_INFO, "",
726 						DUMP_PREFIX_ADDRESS, 16, 1,
727 						tx_buffer->skb->data,
728 						dma_unmap_len(tx_buffer, len),
729 						true);
730 			}
731 		}
732 	}
733 
734 	/* Print RX Rings Summary */
735 rx_ring_summary:
736 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
737 	pr_info("Queue [NTU] [NTC]\n");
738 	for (n = 0; n < adapter->num_rx_queues; n++) {
739 		rx_ring = adapter->rx_ring[n];
740 		pr_info("%5d %5X %5X\n",
741 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
742 	}
743 
744 	/* Print RX Rings */
745 	if (!netif_msg_rx_status(adapter))
746 		return;
747 
748 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
749 
750 	/* Receive Descriptor Formats
751 	 *
752 	 * 82598 Advanced Receive Descriptor (Read) Format
753 	 *    63                                           1        0
754 	 *    +-----------------------------------------------------+
755 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
756 	 *    +----------------------------------------------+------+
757 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
758 	 *    +-----------------------------------------------------+
759 	 *
760 	 *
761 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
762 	 *
763 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
764 	 *   +------------------------------------------------------+
765 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
766 	 *   | Packet   | IP     |   |          |     | Type | Type |
767 	 *   | Checksum | Ident  |   |          |     |      |      |
768 	 *   +------------------------------------------------------+
769 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
770 	 *   +------------------------------------------------------+
771 	 *   63       48 47    32 31            20 19               0
772 	 *
773 	 * 82599+ Advanced Receive Descriptor (Read) Format
774 	 *    63                                           1        0
775 	 *    +-----------------------------------------------------+
776 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
777 	 *    +----------------------------------------------+------+
778 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
779 	 *    +-----------------------------------------------------+
780 	 *
781 	 *
782 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
783 	 *
784 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
785 	 *   +------------------------------------------------------+
786 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
787 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
788 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
789 	 *   +------------------------------------------------------+
790 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
791 	 *   +------------------------------------------------------+
792 	 *   63       48 47    32 31          20 19                 0
793 	 */
794 
795 	for (n = 0; n < adapter->num_rx_queues; n++) {
796 		rx_ring = adapter->rx_ring[n];
797 		pr_info("------------------------------------\n");
798 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
799 		pr_info("------------------------------------\n");
800 		pr_info("%s%s%s",
801 			"R  [desc]      [ PktBuf     A0] ",
802 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
803 			"<-- Adv Rx Read format\n");
804 		pr_info("%s%s%s",
805 			"RWB[desc]      [PcsmIpSHl PtRs] ",
806 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
807 			"<-- Adv Rx Write-Back format\n");
808 
809 		for (i = 0; i < rx_ring->count; i++) {
810 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
811 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
812 			u0 = (struct my_u0 *)rx_desc;
813 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
814 			if (staterr & IXGBE_RXD_STAT_DD) {
815 				/* Descriptor Done */
816 				pr_info("RWB[0x%03X]     %016llX "
817 					"%016llX ---------------- %p", i,
818 					le64_to_cpu(u0->a),
819 					le64_to_cpu(u0->b),
820 					rx_buffer_info->skb);
821 			} else {
822 				pr_info("R  [0x%03X]     %016llX "
823 					"%016llX %016llX %p", i,
824 					le64_to_cpu(u0->a),
825 					le64_to_cpu(u0->b),
826 					(u64)rx_buffer_info->dma,
827 					rx_buffer_info->skb);
828 
829 				if (netif_msg_pktdata(adapter) &&
830 				    rx_buffer_info->dma) {
831 					print_hex_dump(KERN_INFO, "",
832 					   DUMP_PREFIX_ADDRESS, 16, 1,
833 					   page_address(rx_buffer_info->page) +
834 						    rx_buffer_info->page_offset,
835 					   ixgbe_rx_bufsz(rx_ring), true);
836 				}
837 			}
838 
839 			if (i == rx_ring->next_to_use)
840 				pr_cont(" NTU\n");
841 			else if (i == rx_ring->next_to_clean)
842 				pr_cont(" NTC\n");
843 			else
844 				pr_cont("\n");
845 
846 		}
847 	}
848 }
849 
850 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
851 {
852 	u32 ctrl_ext;
853 
854 	/* Let firmware take over control of h/w */
855 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
856 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
857 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
858 }
859 
860 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
861 {
862 	u32 ctrl_ext;
863 
864 	/* Let firmware know the driver has taken over */
865 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
866 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
867 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
868 }
869 
870 /**
871  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
872  * @adapter: pointer to adapter struct
873  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
874  * @queue: queue to map the corresponding interrupt to
875  * @msix_vector: the vector to map to the corresponding queue
876  *
877  */
878 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
879 			   u8 queue, u8 msix_vector)
880 {
881 	u32 ivar, index;
882 	struct ixgbe_hw *hw = &adapter->hw;
883 	switch (hw->mac.type) {
884 	case ixgbe_mac_82598EB:
885 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
886 		if (direction == -1)
887 			direction = 0;
888 		index = (((direction * 64) + queue) >> 2) & 0x1F;
889 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
890 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
891 		ivar |= (msix_vector << (8 * (queue & 0x3)));
892 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
893 		break;
894 	case ixgbe_mac_82599EB:
895 	case ixgbe_mac_X540:
896 	case ixgbe_mac_X550:
897 	case ixgbe_mac_X550EM_x:
898 	case ixgbe_mac_x550em_a:
899 		if (direction == -1) {
900 			/* other causes */
901 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
902 			index = ((queue & 1) * 8);
903 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
904 			ivar &= ~(0xFF << index);
905 			ivar |= (msix_vector << index);
906 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
907 			break;
908 		} else {
909 			/* tx or rx causes */
910 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
911 			index = ((16 * (queue & 1)) + (8 * direction));
912 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
913 			ivar &= ~(0xFF << index);
914 			ivar |= (msix_vector << index);
915 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
916 			break;
917 		}
918 	default:
919 		break;
920 	}
921 }
922 
923 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
924 					  u64 qmask)
925 {
926 	u32 mask;
927 
928 	switch (adapter->hw.mac.type) {
929 	case ixgbe_mac_82598EB:
930 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
931 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
932 		break;
933 	case ixgbe_mac_82599EB:
934 	case ixgbe_mac_X540:
935 	case ixgbe_mac_X550:
936 	case ixgbe_mac_X550EM_x:
937 	case ixgbe_mac_x550em_a:
938 		mask = (qmask & 0xFFFFFFFF);
939 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
940 		mask = (qmask >> 32);
941 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
942 		break;
943 	default:
944 		break;
945 	}
946 }
947 
948 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
949 				      struct ixgbe_tx_buffer *tx_buffer)
950 {
951 	if (tx_buffer->skb) {
952 		dev_kfree_skb_any(tx_buffer->skb);
953 		if (dma_unmap_len(tx_buffer, len))
954 			dma_unmap_single(ring->dev,
955 					 dma_unmap_addr(tx_buffer, dma),
956 					 dma_unmap_len(tx_buffer, len),
957 					 DMA_TO_DEVICE);
958 	} else if (dma_unmap_len(tx_buffer, len)) {
959 		dma_unmap_page(ring->dev,
960 			       dma_unmap_addr(tx_buffer, dma),
961 			       dma_unmap_len(tx_buffer, len),
962 			       DMA_TO_DEVICE);
963 	}
964 	tx_buffer->next_to_watch = NULL;
965 	tx_buffer->skb = NULL;
966 	dma_unmap_len_set(tx_buffer, len, 0);
967 	/* tx_buffer must be completely set up in the transmit path */
968 }
969 
970 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
971 {
972 	struct ixgbe_hw *hw = &adapter->hw;
973 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
974 	int i;
975 	u32 data;
976 
977 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
978 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
979 		return;
980 
981 	switch (hw->mac.type) {
982 	case ixgbe_mac_82598EB:
983 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
984 		break;
985 	default:
986 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
987 	}
988 	hwstats->lxoffrxc += data;
989 
990 	/* refill credits (no tx hang) if we received xoff */
991 	if (!data)
992 		return;
993 
994 	for (i = 0; i < adapter->num_tx_queues; i++)
995 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
996 			  &adapter->tx_ring[i]->state);
997 }
998 
999 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
1000 {
1001 	struct ixgbe_hw *hw = &adapter->hw;
1002 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
1003 	u32 xoff[8] = {0};
1004 	u8 tc;
1005 	int i;
1006 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
1007 
1008 	if (adapter->ixgbe_ieee_pfc)
1009 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
1010 
1011 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
1012 		ixgbe_update_xoff_rx_lfc(adapter);
1013 		return;
1014 	}
1015 
1016 	/* update stats for each tc, only valid with PFC enabled */
1017 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
1018 		u32 pxoffrxc;
1019 
1020 		switch (hw->mac.type) {
1021 		case ixgbe_mac_82598EB:
1022 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1023 			break;
1024 		default:
1025 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1026 		}
1027 		hwstats->pxoffrxc[i] += pxoffrxc;
1028 		/* Get the TC for given UP */
1029 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1030 		xoff[tc] += pxoffrxc;
1031 	}
1032 
1033 	/* disarm tx queues that have received xoff frames */
1034 	for (i = 0; i < adapter->num_tx_queues; i++) {
1035 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1036 
1037 		tc = tx_ring->dcb_tc;
1038 		if (xoff[tc])
1039 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1040 	}
1041 }
1042 
1043 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1044 {
1045 	return ring->stats.packets;
1046 }
1047 
1048 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1049 {
1050 	struct ixgbe_adapter *adapter;
1051 	struct ixgbe_hw *hw;
1052 	u32 head, tail;
1053 
1054 	if (ring->l2_accel_priv)
1055 		adapter = ring->l2_accel_priv->real_adapter;
1056 	else
1057 		adapter = netdev_priv(ring->netdev);
1058 
1059 	hw = &adapter->hw;
1060 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1061 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1062 
1063 	if (head != tail)
1064 		return (head < tail) ?
1065 			tail - head : (tail + ring->count - head);
1066 
1067 	return 0;
1068 }
1069 
1070 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1071 {
1072 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1073 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1074 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1075 
1076 	clear_check_for_tx_hang(tx_ring);
1077 
1078 	/*
1079 	 * Check for a hung queue, but be thorough. This verifies
1080 	 * that a transmit has been completed since the previous
1081 	 * check AND there is at least one packet pending. The
1082 	 * ARMED bit is set to indicate a potential hang. The
1083 	 * bit is cleared if a pause frame is received to remove
1084 	 * false hang detection due to PFC or 802.3x frames. By
1085 	 * requiring this to fail twice we avoid races with
1086 	 * pfc clearing the ARMED bit and conditions where we
1087 	 * run the check_tx_hang logic with a transmit completion
1088 	 * pending but without time to complete it yet.
1089 	 */
1090 	if (tx_done_old == tx_done && tx_pending)
1091 		/* make sure it is true for two checks in a row */
1092 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1093 					&tx_ring->state);
1094 	/* update completed stats and continue */
1095 	tx_ring->tx_stats.tx_done_old = tx_done;
1096 	/* reset the countdown */
1097 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1098 
1099 	return false;
1100 }
1101 
1102 /**
1103  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1104  * @adapter: driver private struct
1105  **/
1106 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1107 {
1108 
1109 	/* Do the reset outside of interrupt context */
1110 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1111 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1112 		e_warn(drv, "initiating reset due to tx timeout\n");
1113 		ixgbe_service_event_schedule(adapter);
1114 	}
1115 }
1116 
1117 /**
1118  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1119  **/
1120 static int ixgbe_tx_maxrate(struct net_device *netdev,
1121 			    int queue_index, u32 maxrate)
1122 {
1123 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1124 	struct ixgbe_hw *hw = &adapter->hw;
1125 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1126 
1127 	if (!maxrate)
1128 		return 0;
1129 
1130 	/* Calculate the rate factor values to set */
1131 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1132 	bcnrc_val /= maxrate;
1133 
1134 	/* clear everything but the rate factor */
1135 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1136 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1137 
1138 	/* enable the rate scheduler */
1139 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1140 
1141 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1142 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1143 
1144 	return 0;
1145 }
1146 
1147 /**
1148  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1149  * @q_vector: structure containing interrupt and ring information
1150  * @tx_ring: tx ring to clean
1151  * @napi_budget: Used to determine if we are in netpoll
1152  **/
1153 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1154 			       struct ixgbe_ring *tx_ring, int napi_budget)
1155 {
1156 	struct ixgbe_adapter *adapter = q_vector->adapter;
1157 	struct ixgbe_tx_buffer *tx_buffer;
1158 	union ixgbe_adv_tx_desc *tx_desc;
1159 	unsigned int total_bytes = 0, total_packets = 0;
1160 	unsigned int budget = q_vector->tx.work_limit;
1161 	unsigned int i = tx_ring->next_to_clean;
1162 
1163 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1164 		return true;
1165 
1166 	tx_buffer = &tx_ring->tx_buffer_info[i];
1167 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1168 	i -= tx_ring->count;
1169 
1170 	do {
1171 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1172 
1173 		/* if next_to_watch is not set then there is no work pending */
1174 		if (!eop_desc)
1175 			break;
1176 
1177 		/* prevent any other reads prior to eop_desc */
1178 		read_barrier_depends();
1179 
1180 		/* if DD is not set pending work has not been completed */
1181 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1182 			break;
1183 
1184 		/* clear next_to_watch to prevent false hangs */
1185 		tx_buffer->next_to_watch = NULL;
1186 
1187 		/* update the statistics for this packet */
1188 		total_bytes += tx_buffer->bytecount;
1189 		total_packets += tx_buffer->gso_segs;
1190 
1191 		/* free the skb */
1192 		napi_consume_skb(tx_buffer->skb, napi_budget);
1193 
1194 		/* unmap skb header data */
1195 		dma_unmap_single(tx_ring->dev,
1196 				 dma_unmap_addr(tx_buffer, dma),
1197 				 dma_unmap_len(tx_buffer, len),
1198 				 DMA_TO_DEVICE);
1199 
1200 		/* clear tx_buffer data */
1201 		tx_buffer->skb = NULL;
1202 		dma_unmap_len_set(tx_buffer, len, 0);
1203 
1204 		/* unmap remaining buffers */
1205 		while (tx_desc != eop_desc) {
1206 			tx_buffer++;
1207 			tx_desc++;
1208 			i++;
1209 			if (unlikely(!i)) {
1210 				i -= tx_ring->count;
1211 				tx_buffer = tx_ring->tx_buffer_info;
1212 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1213 			}
1214 
1215 			/* unmap any remaining paged data */
1216 			if (dma_unmap_len(tx_buffer, len)) {
1217 				dma_unmap_page(tx_ring->dev,
1218 					       dma_unmap_addr(tx_buffer, dma),
1219 					       dma_unmap_len(tx_buffer, len),
1220 					       DMA_TO_DEVICE);
1221 				dma_unmap_len_set(tx_buffer, len, 0);
1222 			}
1223 		}
1224 
1225 		/* move us one more past the eop_desc for start of next pkt */
1226 		tx_buffer++;
1227 		tx_desc++;
1228 		i++;
1229 		if (unlikely(!i)) {
1230 			i -= tx_ring->count;
1231 			tx_buffer = tx_ring->tx_buffer_info;
1232 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1233 		}
1234 
1235 		/* issue prefetch for next Tx descriptor */
1236 		prefetch(tx_desc);
1237 
1238 		/* update budget accounting */
1239 		budget--;
1240 	} while (likely(budget));
1241 
1242 	i += tx_ring->count;
1243 	tx_ring->next_to_clean = i;
1244 	u64_stats_update_begin(&tx_ring->syncp);
1245 	tx_ring->stats.bytes += total_bytes;
1246 	tx_ring->stats.packets += total_packets;
1247 	u64_stats_update_end(&tx_ring->syncp);
1248 	q_vector->tx.total_bytes += total_bytes;
1249 	q_vector->tx.total_packets += total_packets;
1250 
1251 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1252 		/* schedule immediate reset if we believe we hung */
1253 		struct ixgbe_hw *hw = &adapter->hw;
1254 		e_err(drv, "Detected Tx Unit Hang\n"
1255 			"  Tx Queue             <%d>\n"
1256 			"  TDH, TDT             <%x>, <%x>\n"
1257 			"  next_to_use          <%x>\n"
1258 			"  next_to_clean        <%x>\n"
1259 			"tx_buffer_info[next_to_clean]\n"
1260 			"  time_stamp           <%lx>\n"
1261 			"  jiffies              <%lx>\n",
1262 			tx_ring->queue_index,
1263 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1264 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1265 			tx_ring->next_to_use, i,
1266 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1267 
1268 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1269 
1270 		e_info(probe,
1271 		       "tx hang %d detected on queue %d, resetting adapter\n",
1272 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1273 
1274 		/* schedule immediate reset if we believe we hung */
1275 		ixgbe_tx_timeout_reset(adapter);
1276 
1277 		/* the adapter is about to reset, no point in enabling stuff */
1278 		return true;
1279 	}
1280 
1281 	netdev_tx_completed_queue(txring_txq(tx_ring),
1282 				  total_packets, total_bytes);
1283 
1284 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1285 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1286 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1287 		/* Make sure that anybody stopping the queue after this
1288 		 * sees the new next_to_clean.
1289 		 */
1290 		smp_mb();
1291 		if (__netif_subqueue_stopped(tx_ring->netdev,
1292 					     tx_ring->queue_index)
1293 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1294 			netif_wake_subqueue(tx_ring->netdev,
1295 					    tx_ring->queue_index);
1296 			++tx_ring->tx_stats.restart_queue;
1297 		}
1298 	}
1299 
1300 	return !!budget;
1301 }
1302 
1303 #ifdef CONFIG_IXGBE_DCA
1304 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1305 				struct ixgbe_ring *tx_ring,
1306 				int cpu)
1307 {
1308 	struct ixgbe_hw *hw = &adapter->hw;
1309 	u32 txctrl = 0;
1310 	u16 reg_offset;
1311 
1312 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1313 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1314 
1315 	switch (hw->mac.type) {
1316 	case ixgbe_mac_82598EB:
1317 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1318 		break;
1319 	case ixgbe_mac_82599EB:
1320 	case ixgbe_mac_X540:
1321 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1322 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1323 		break;
1324 	default:
1325 		/* for unknown hardware do not write register */
1326 		return;
1327 	}
1328 
1329 	/*
1330 	 * We can enable relaxed ordering for reads, but not writes when
1331 	 * DCA is enabled.  This is due to a known issue in some chipsets
1332 	 * which will cause the DCA tag to be cleared.
1333 	 */
1334 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1335 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1336 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1337 
1338 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1339 }
1340 
1341 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1342 				struct ixgbe_ring *rx_ring,
1343 				int cpu)
1344 {
1345 	struct ixgbe_hw *hw = &adapter->hw;
1346 	u32 rxctrl = 0;
1347 	u8 reg_idx = rx_ring->reg_idx;
1348 
1349 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1350 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1351 
1352 	switch (hw->mac.type) {
1353 	case ixgbe_mac_82599EB:
1354 	case ixgbe_mac_X540:
1355 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1356 		break;
1357 	default:
1358 		break;
1359 	}
1360 
1361 	/*
1362 	 * We can enable relaxed ordering for reads, but not writes when
1363 	 * DCA is enabled.  This is due to a known issue in some chipsets
1364 	 * which will cause the DCA tag to be cleared.
1365 	 */
1366 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1367 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1368 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1369 
1370 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1371 }
1372 
1373 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1374 {
1375 	struct ixgbe_adapter *adapter = q_vector->adapter;
1376 	struct ixgbe_ring *ring;
1377 	int cpu = get_cpu();
1378 
1379 	if (q_vector->cpu == cpu)
1380 		goto out_no_update;
1381 
1382 	ixgbe_for_each_ring(ring, q_vector->tx)
1383 		ixgbe_update_tx_dca(adapter, ring, cpu);
1384 
1385 	ixgbe_for_each_ring(ring, q_vector->rx)
1386 		ixgbe_update_rx_dca(adapter, ring, cpu);
1387 
1388 	q_vector->cpu = cpu;
1389 out_no_update:
1390 	put_cpu();
1391 }
1392 
1393 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1394 {
1395 	int i;
1396 
1397 	/* always use CB2 mode, difference is masked in the CB driver */
1398 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1399 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1400 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1401 	else
1402 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1403 				IXGBE_DCA_CTRL_DCA_DISABLE);
1404 
1405 	for (i = 0; i < adapter->num_q_vectors; i++) {
1406 		adapter->q_vector[i]->cpu = -1;
1407 		ixgbe_update_dca(adapter->q_vector[i]);
1408 	}
1409 }
1410 
1411 static int __ixgbe_notify_dca(struct device *dev, void *data)
1412 {
1413 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1414 	unsigned long event = *(unsigned long *)data;
1415 
1416 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1417 		return 0;
1418 
1419 	switch (event) {
1420 	case DCA_PROVIDER_ADD:
1421 		/* if we're already enabled, don't do it again */
1422 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1423 			break;
1424 		if (dca_add_requester(dev) == 0) {
1425 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1426 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1427 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1428 			break;
1429 		}
1430 		/* Fall Through since DCA is disabled. */
1431 	case DCA_PROVIDER_REMOVE:
1432 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1433 			dca_remove_requester(dev);
1434 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1435 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1436 					IXGBE_DCA_CTRL_DCA_DISABLE);
1437 		}
1438 		break;
1439 	}
1440 
1441 	return 0;
1442 }
1443 
1444 #endif /* CONFIG_IXGBE_DCA */
1445 
1446 #define IXGBE_RSS_L4_TYPES_MASK \
1447 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1448 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1449 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1450 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1451 
1452 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1453 				 union ixgbe_adv_rx_desc *rx_desc,
1454 				 struct sk_buff *skb)
1455 {
1456 	u16 rss_type;
1457 
1458 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1459 		return;
1460 
1461 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1462 		   IXGBE_RXDADV_RSSTYPE_MASK;
1463 
1464 	if (!rss_type)
1465 		return;
1466 
1467 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1468 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1469 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1470 }
1471 
1472 #ifdef IXGBE_FCOE
1473 /**
1474  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1475  * @ring: structure containing ring specific data
1476  * @rx_desc: advanced rx descriptor
1477  *
1478  * Returns : true if it is FCoE pkt
1479  */
1480 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1481 				    union ixgbe_adv_rx_desc *rx_desc)
1482 {
1483 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1484 
1485 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1486 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1487 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1488 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1489 }
1490 
1491 #endif /* IXGBE_FCOE */
1492 /**
1493  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1494  * @ring: structure containing ring specific data
1495  * @rx_desc: current Rx descriptor being processed
1496  * @skb: skb currently being received and modified
1497  **/
1498 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1499 				     union ixgbe_adv_rx_desc *rx_desc,
1500 				     struct sk_buff *skb)
1501 {
1502 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1503 	bool encap_pkt = false;
1504 
1505 	skb_checksum_none_assert(skb);
1506 
1507 	/* Rx csum disabled */
1508 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1509 		return;
1510 
1511 	/* check for VXLAN and Geneve packets */
1512 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1513 		encap_pkt = true;
1514 		skb->encapsulation = 1;
1515 	}
1516 
1517 	/* if IP and error */
1518 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1519 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1520 		ring->rx_stats.csum_err++;
1521 		return;
1522 	}
1523 
1524 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1525 		return;
1526 
1527 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1528 		/*
1529 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1530 		 * checksum errors.
1531 		 */
1532 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1533 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1534 			return;
1535 
1536 		ring->rx_stats.csum_err++;
1537 		return;
1538 	}
1539 
1540 	/* It must be a TCP or UDP packet with a valid checksum */
1541 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1542 	if (encap_pkt) {
1543 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1544 			return;
1545 
1546 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1547 			skb->ip_summed = CHECKSUM_NONE;
1548 			return;
1549 		}
1550 		/* If we checked the outer header let the stack know */
1551 		skb->csum_level = 1;
1552 	}
1553 }
1554 
1555 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1556 				    struct ixgbe_rx_buffer *bi)
1557 {
1558 	struct page *page = bi->page;
1559 	dma_addr_t dma;
1560 
1561 	/* since we are recycling buffers we should seldom need to alloc */
1562 	if (likely(page))
1563 		return true;
1564 
1565 	/* alloc new page for storage */
1566 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1567 	if (unlikely(!page)) {
1568 		rx_ring->rx_stats.alloc_rx_page_failed++;
1569 		return false;
1570 	}
1571 
1572 	/* map page for use */
1573 	dma = dma_map_page(rx_ring->dev, page, 0,
1574 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1575 
1576 	/*
1577 	 * if mapping failed free memory back to system since
1578 	 * there isn't much point in holding memory we can't use
1579 	 */
1580 	if (dma_mapping_error(rx_ring->dev, dma)) {
1581 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1582 
1583 		rx_ring->rx_stats.alloc_rx_page_failed++;
1584 		return false;
1585 	}
1586 
1587 	bi->dma = dma;
1588 	bi->page = page;
1589 	bi->page_offset = 0;
1590 
1591 	return true;
1592 }
1593 
1594 /**
1595  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1596  * @rx_ring: ring to place buffers on
1597  * @cleaned_count: number of buffers to replace
1598  **/
1599 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1600 {
1601 	union ixgbe_adv_rx_desc *rx_desc;
1602 	struct ixgbe_rx_buffer *bi;
1603 	u16 i = rx_ring->next_to_use;
1604 
1605 	/* nothing to do */
1606 	if (!cleaned_count)
1607 		return;
1608 
1609 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1610 	bi = &rx_ring->rx_buffer_info[i];
1611 	i -= rx_ring->count;
1612 
1613 	do {
1614 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1615 			break;
1616 
1617 		/*
1618 		 * Refresh the desc even if buffer_addrs didn't change
1619 		 * because each write-back erases this info.
1620 		 */
1621 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1622 
1623 		rx_desc++;
1624 		bi++;
1625 		i++;
1626 		if (unlikely(!i)) {
1627 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1628 			bi = rx_ring->rx_buffer_info;
1629 			i -= rx_ring->count;
1630 		}
1631 
1632 		/* clear the status bits for the next_to_use descriptor */
1633 		rx_desc->wb.upper.status_error = 0;
1634 
1635 		cleaned_count--;
1636 	} while (cleaned_count);
1637 
1638 	i += rx_ring->count;
1639 
1640 	if (rx_ring->next_to_use != i) {
1641 		rx_ring->next_to_use = i;
1642 
1643 		/* update next to alloc since we have filled the ring */
1644 		rx_ring->next_to_alloc = i;
1645 
1646 		/* Force memory writes to complete before letting h/w
1647 		 * know there are new descriptors to fetch.  (Only
1648 		 * applicable for weak-ordered memory model archs,
1649 		 * such as IA-64).
1650 		 */
1651 		wmb();
1652 		writel(i, rx_ring->tail);
1653 	}
1654 }
1655 
1656 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1657 				   struct sk_buff *skb)
1658 {
1659 	u16 hdr_len = skb_headlen(skb);
1660 
1661 	/* set gso_size to avoid messing up TCP MSS */
1662 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1663 						 IXGBE_CB(skb)->append_cnt);
1664 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1665 }
1666 
1667 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1668 				   struct sk_buff *skb)
1669 {
1670 	/* if append_cnt is 0 then frame is not RSC */
1671 	if (!IXGBE_CB(skb)->append_cnt)
1672 		return;
1673 
1674 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1675 	rx_ring->rx_stats.rsc_flush++;
1676 
1677 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1678 
1679 	/* gso_size is computed using append_cnt so always clear it last */
1680 	IXGBE_CB(skb)->append_cnt = 0;
1681 }
1682 
1683 /**
1684  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1685  * @rx_ring: rx descriptor ring packet is being transacted on
1686  * @rx_desc: pointer to the EOP Rx descriptor
1687  * @skb: pointer to current skb being populated
1688  *
1689  * This function checks the ring, descriptor, and packet information in
1690  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1691  * other fields within the skb.
1692  **/
1693 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1694 				     union ixgbe_adv_rx_desc *rx_desc,
1695 				     struct sk_buff *skb)
1696 {
1697 	struct net_device *dev = rx_ring->netdev;
1698 	u32 flags = rx_ring->q_vector->adapter->flags;
1699 
1700 	ixgbe_update_rsc_stats(rx_ring, skb);
1701 
1702 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1703 
1704 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1705 
1706 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1707 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1708 
1709 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1710 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1711 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1712 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1713 	}
1714 
1715 	skb_record_rx_queue(skb, rx_ring->queue_index);
1716 
1717 	skb->protocol = eth_type_trans(skb, dev);
1718 }
1719 
1720 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1721 			 struct sk_buff *skb)
1722 {
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 /**
2201  * ixgbe_configure_msix - Configure MSI-X hardware
2202  * @adapter: board private structure
2203  *
2204  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2205  * interrupts.
2206  **/
2207 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2208 {
2209 	struct ixgbe_q_vector *q_vector;
2210 	int v_idx;
2211 	u32 mask;
2212 
2213 	/* Populate MSIX to EITR Select */
2214 	if (adapter->num_vfs > 32) {
2215 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2216 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2217 	}
2218 
2219 	/*
2220 	 * Populate the IVAR table and set the ITR values to the
2221 	 * corresponding register.
2222 	 */
2223 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2224 		struct ixgbe_ring *ring;
2225 		q_vector = adapter->q_vector[v_idx];
2226 
2227 		ixgbe_for_each_ring(ring, q_vector->rx)
2228 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2229 
2230 		ixgbe_for_each_ring(ring, q_vector->tx)
2231 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2232 
2233 		ixgbe_write_eitr(q_vector);
2234 	}
2235 
2236 	switch (adapter->hw.mac.type) {
2237 	case ixgbe_mac_82598EB:
2238 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2239 			       v_idx);
2240 		break;
2241 	case ixgbe_mac_82599EB:
2242 	case ixgbe_mac_X540:
2243 	case ixgbe_mac_X550:
2244 	case ixgbe_mac_X550EM_x:
2245 	case ixgbe_mac_x550em_a:
2246 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2247 		break;
2248 	default:
2249 		break;
2250 	}
2251 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2252 
2253 	/* set up to autoclear timer, and the vectors */
2254 	mask = IXGBE_EIMS_ENABLE_MASK;
2255 	mask &= ~(IXGBE_EIMS_OTHER |
2256 		  IXGBE_EIMS_MAILBOX |
2257 		  IXGBE_EIMS_LSC);
2258 
2259 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2260 }
2261 
2262 enum latency_range {
2263 	lowest_latency = 0,
2264 	low_latency = 1,
2265 	bulk_latency = 2,
2266 	latency_invalid = 255
2267 };
2268 
2269 /**
2270  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2271  * @q_vector: structure containing interrupt and ring information
2272  * @ring_container: structure containing ring performance data
2273  *
2274  *      Stores a new ITR value based on packets and byte
2275  *      counts during the last interrupt.  The advantage of per interrupt
2276  *      computation is faster updates and more accurate ITR for the current
2277  *      traffic pattern.  Constants in this function were computed
2278  *      based on theoretical maximum wire speed and thresholds were set based
2279  *      on testing data as well as attempting to minimize response time
2280  *      while increasing bulk throughput.
2281  *      this functionality is controlled by the InterruptThrottleRate module
2282  *      parameter (see ixgbe_param.c)
2283  **/
2284 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2285 			     struct ixgbe_ring_container *ring_container)
2286 {
2287 	int bytes = ring_container->total_bytes;
2288 	int packets = ring_container->total_packets;
2289 	u32 timepassed_us;
2290 	u64 bytes_perint;
2291 	u8 itr_setting = ring_container->itr;
2292 
2293 	if (packets == 0)
2294 		return;
2295 
2296 	/* simple throttlerate management
2297 	 *   0-10MB/s   lowest (100000 ints/s)
2298 	 *  10-20MB/s   low    (20000 ints/s)
2299 	 *  20-1249MB/s bulk   (12000 ints/s)
2300 	 */
2301 	/* what was last interrupt timeslice? */
2302 	timepassed_us = q_vector->itr >> 2;
2303 	if (timepassed_us == 0)
2304 		return;
2305 
2306 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2307 
2308 	switch (itr_setting) {
2309 	case lowest_latency:
2310 		if (bytes_perint > 10)
2311 			itr_setting = low_latency;
2312 		break;
2313 	case low_latency:
2314 		if (bytes_perint > 20)
2315 			itr_setting = bulk_latency;
2316 		else if (bytes_perint <= 10)
2317 			itr_setting = lowest_latency;
2318 		break;
2319 	case bulk_latency:
2320 		if (bytes_perint <= 20)
2321 			itr_setting = low_latency;
2322 		break;
2323 	}
2324 
2325 	/* clear work counters since we have the values we need */
2326 	ring_container->total_bytes = 0;
2327 	ring_container->total_packets = 0;
2328 
2329 	/* write updated itr to ring container */
2330 	ring_container->itr = itr_setting;
2331 }
2332 
2333 /**
2334  * ixgbe_write_eitr - write EITR register in hardware specific way
2335  * @q_vector: structure containing interrupt and ring information
2336  *
2337  * This function is made to be called by ethtool and by the driver
2338  * when it needs to update EITR registers at runtime.  Hardware
2339  * specific quirks/differences are taken care of here.
2340  */
2341 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2342 {
2343 	struct ixgbe_adapter *adapter = q_vector->adapter;
2344 	struct ixgbe_hw *hw = &adapter->hw;
2345 	int v_idx = q_vector->v_idx;
2346 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2347 
2348 	switch (adapter->hw.mac.type) {
2349 	case ixgbe_mac_82598EB:
2350 		/* must write high and low 16 bits to reset counter */
2351 		itr_reg |= (itr_reg << 16);
2352 		break;
2353 	case ixgbe_mac_82599EB:
2354 	case ixgbe_mac_X540:
2355 	case ixgbe_mac_X550:
2356 	case ixgbe_mac_X550EM_x:
2357 	case ixgbe_mac_x550em_a:
2358 		/*
2359 		 * set the WDIS bit to not clear the timer bits and cause an
2360 		 * immediate assertion of the interrupt
2361 		 */
2362 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2363 		break;
2364 	default:
2365 		break;
2366 	}
2367 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2368 }
2369 
2370 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2371 {
2372 	u32 new_itr = q_vector->itr;
2373 	u8 current_itr;
2374 
2375 	ixgbe_update_itr(q_vector, &q_vector->tx);
2376 	ixgbe_update_itr(q_vector, &q_vector->rx);
2377 
2378 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2379 
2380 	switch (current_itr) {
2381 	/* counts and packets in update_itr are dependent on these numbers */
2382 	case lowest_latency:
2383 		new_itr = IXGBE_100K_ITR;
2384 		break;
2385 	case low_latency:
2386 		new_itr = IXGBE_20K_ITR;
2387 		break;
2388 	case bulk_latency:
2389 		new_itr = IXGBE_12K_ITR;
2390 		break;
2391 	default:
2392 		break;
2393 	}
2394 
2395 	if (new_itr != q_vector->itr) {
2396 		/* do an exponential smoothing */
2397 		new_itr = (10 * new_itr * q_vector->itr) /
2398 			  ((9 * new_itr) + q_vector->itr);
2399 
2400 		/* save the algorithm value here */
2401 		q_vector->itr = new_itr;
2402 
2403 		ixgbe_write_eitr(q_vector);
2404 	}
2405 }
2406 
2407 /**
2408  * ixgbe_check_overtemp_subtask - check for over temperature
2409  * @adapter: pointer to adapter
2410  **/
2411 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2412 {
2413 	struct ixgbe_hw *hw = &adapter->hw;
2414 	u32 eicr = adapter->interrupt_event;
2415 	s32 rc;
2416 
2417 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2418 		return;
2419 
2420 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2421 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2422 		return;
2423 
2424 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2425 
2426 	switch (hw->device_id) {
2427 	case IXGBE_DEV_ID_82599_T3_LOM:
2428 		/*
2429 		 * Since the warning interrupt is for both ports
2430 		 * we don't have to check if:
2431 		 *  - This interrupt wasn't for our port.
2432 		 *  - We may have missed the interrupt so always have to
2433 		 *    check if we  got a LSC
2434 		 */
2435 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2436 		    !(eicr & IXGBE_EICR_LSC))
2437 			return;
2438 
2439 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2440 			u32 speed;
2441 			bool link_up = false;
2442 
2443 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2444 
2445 			if (link_up)
2446 				return;
2447 		}
2448 
2449 		/* Check if this is not due to overtemp */
2450 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2451 			return;
2452 
2453 		break;
2454 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2455 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2456 		rc = hw->phy.ops.check_overtemp(hw);
2457 		if (rc != IXGBE_ERR_OVERTEMP)
2458 			return;
2459 		break;
2460 	default:
2461 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2462 			return;
2463 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2464 			return;
2465 		break;
2466 	}
2467 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2468 
2469 	adapter->interrupt_event = 0;
2470 }
2471 
2472 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2473 {
2474 	struct ixgbe_hw *hw = &adapter->hw;
2475 
2476 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2477 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2478 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2479 		/* write to clear the interrupt */
2480 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2481 	}
2482 }
2483 
2484 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2485 {
2486 	struct ixgbe_hw *hw = &adapter->hw;
2487 
2488 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2489 		return;
2490 
2491 	switch (adapter->hw.mac.type) {
2492 	case ixgbe_mac_82599EB:
2493 		/*
2494 		 * Need to check link state so complete overtemp check
2495 		 * on service task
2496 		 */
2497 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2498 		     (eicr & IXGBE_EICR_LSC)) &&
2499 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2500 			adapter->interrupt_event = eicr;
2501 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2502 			ixgbe_service_event_schedule(adapter);
2503 			return;
2504 		}
2505 		return;
2506 	case ixgbe_mac_x550em_a:
2507 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2508 			adapter->interrupt_event = eicr;
2509 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2510 			ixgbe_service_event_schedule(adapter);
2511 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2512 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2513 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2514 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2515 		}
2516 		return;
2517 	case ixgbe_mac_X550:
2518 	case ixgbe_mac_X540:
2519 		if (!(eicr & IXGBE_EICR_TS))
2520 			return;
2521 		break;
2522 	default:
2523 		return;
2524 	}
2525 
2526 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2527 }
2528 
2529 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2530 {
2531 	switch (hw->mac.type) {
2532 	case ixgbe_mac_82598EB:
2533 		if (hw->phy.type == ixgbe_phy_nl)
2534 			return true;
2535 		return false;
2536 	case ixgbe_mac_82599EB:
2537 	case ixgbe_mac_X550EM_x:
2538 	case ixgbe_mac_x550em_a:
2539 		switch (hw->mac.ops.get_media_type(hw)) {
2540 		case ixgbe_media_type_fiber:
2541 		case ixgbe_media_type_fiber_qsfp:
2542 			return true;
2543 		default:
2544 			return false;
2545 		}
2546 	default:
2547 		return false;
2548 	}
2549 }
2550 
2551 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2552 {
2553 	struct ixgbe_hw *hw = &adapter->hw;
2554 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2555 
2556 	if (!ixgbe_is_sfp(hw))
2557 		return;
2558 
2559 	/* Later MAC's use different SDP */
2560 	if (hw->mac.type >= ixgbe_mac_X540)
2561 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2562 
2563 	if (eicr & eicr_mask) {
2564 		/* Clear the interrupt */
2565 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2566 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2567 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2568 			adapter->sfp_poll_time = 0;
2569 			ixgbe_service_event_schedule(adapter);
2570 		}
2571 	}
2572 
2573 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2574 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2575 		/* Clear the interrupt */
2576 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2577 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2578 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2579 			ixgbe_service_event_schedule(adapter);
2580 		}
2581 	}
2582 }
2583 
2584 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2585 {
2586 	struct ixgbe_hw *hw = &adapter->hw;
2587 
2588 	adapter->lsc_int++;
2589 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2590 	adapter->link_check_timeout = jiffies;
2591 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2592 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2593 		IXGBE_WRITE_FLUSH(hw);
2594 		ixgbe_service_event_schedule(adapter);
2595 	}
2596 }
2597 
2598 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2599 					   u64 qmask)
2600 {
2601 	u32 mask;
2602 	struct ixgbe_hw *hw = &adapter->hw;
2603 
2604 	switch (hw->mac.type) {
2605 	case ixgbe_mac_82598EB:
2606 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2607 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2608 		break;
2609 	case ixgbe_mac_82599EB:
2610 	case ixgbe_mac_X540:
2611 	case ixgbe_mac_X550:
2612 	case ixgbe_mac_X550EM_x:
2613 	case ixgbe_mac_x550em_a:
2614 		mask = (qmask & 0xFFFFFFFF);
2615 		if (mask)
2616 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2617 		mask = (qmask >> 32);
2618 		if (mask)
2619 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2620 		break;
2621 	default:
2622 		break;
2623 	}
2624 	/* skip the flush */
2625 }
2626 
2627 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2628 					    u64 qmask)
2629 {
2630 	u32 mask;
2631 	struct ixgbe_hw *hw = &adapter->hw;
2632 
2633 	switch (hw->mac.type) {
2634 	case ixgbe_mac_82598EB:
2635 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2636 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2637 		break;
2638 	case ixgbe_mac_82599EB:
2639 	case ixgbe_mac_X540:
2640 	case ixgbe_mac_X550:
2641 	case ixgbe_mac_X550EM_x:
2642 	case ixgbe_mac_x550em_a:
2643 		mask = (qmask & 0xFFFFFFFF);
2644 		if (mask)
2645 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2646 		mask = (qmask >> 32);
2647 		if (mask)
2648 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2649 		break;
2650 	default:
2651 		break;
2652 	}
2653 	/* skip the flush */
2654 }
2655 
2656 /**
2657  * ixgbe_irq_enable - Enable default interrupt generation settings
2658  * @adapter: board private structure
2659  **/
2660 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2661 				    bool flush)
2662 {
2663 	struct ixgbe_hw *hw = &adapter->hw;
2664 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2665 
2666 	/* don't reenable LSC while waiting for link */
2667 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2668 		mask &= ~IXGBE_EIMS_LSC;
2669 
2670 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2671 		switch (adapter->hw.mac.type) {
2672 		case ixgbe_mac_82599EB:
2673 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
2674 			break;
2675 		case ixgbe_mac_X540:
2676 		case ixgbe_mac_X550:
2677 		case ixgbe_mac_X550EM_x:
2678 		case ixgbe_mac_x550em_a:
2679 			mask |= IXGBE_EIMS_TS;
2680 			break;
2681 		default:
2682 			break;
2683 		}
2684 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2685 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2686 	switch (adapter->hw.mac.type) {
2687 	case ixgbe_mac_82599EB:
2688 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2689 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
2690 		/* fall through */
2691 	case ixgbe_mac_X540:
2692 	case ixgbe_mac_X550:
2693 	case ixgbe_mac_X550EM_x:
2694 	case ixgbe_mac_x550em_a:
2695 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
2696 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
2697 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
2698 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
2699 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
2700 			mask |= IXGBE_EICR_GPI_SDP0_X540;
2701 		mask |= IXGBE_EIMS_ECC;
2702 		mask |= IXGBE_EIMS_MAILBOX;
2703 		break;
2704 	default:
2705 		break;
2706 	}
2707 
2708 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2709 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2710 		mask |= IXGBE_EIMS_FLOW_DIR;
2711 
2712 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2713 	if (queues)
2714 		ixgbe_irq_enable_queues(adapter, ~0);
2715 	if (flush)
2716 		IXGBE_WRITE_FLUSH(&adapter->hw);
2717 }
2718 
2719 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2720 {
2721 	struct ixgbe_adapter *adapter = data;
2722 	struct ixgbe_hw *hw = &adapter->hw;
2723 	u32 eicr;
2724 
2725 	/*
2726 	 * Workaround for Silicon errata.  Use clear-by-write instead
2727 	 * of clear-by-read.  Reading with EICS will return the
2728 	 * interrupt causes without clearing, which later be done
2729 	 * with the write to EICR.
2730 	 */
2731 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2732 
2733 	/* The lower 16bits of the EICR register are for the queue interrupts
2734 	 * which should be masked here in order to not accidentally clear them if
2735 	 * the bits are high when ixgbe_msix_other is called. There is a race
2736 	 * condition otherwise which results in possible performance loss
2737 	 * especially if the ixgbe_msix_other interrupt is triggering
2738 	 * consistently (as it would when PPS is turned on for the X540 device)
2739 	 */
2740 	eicr &= 0xFFFF0000;
2741 
2742 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2743 
2744 	if (eicr & IXGBE_EICR_LSC)
2745 		ixgbe_check_lsc(adapter);
2746 
2747 	if (eicr & IXGBE_EICR_MAILBOX)
2748 		ixgbe_msg_task(adapter);
2749 
2750 	switch (hw->mac.type) {
2751 	case ixgbe_mac_82599EB:
2752 	case ixgbe_mac_X540:
2753 	case ixgbe_mac_X550:
2754 	case ixgbe_mac_X550EM_x:
2755 	case ixgbe_mac_x550em_a:
2756 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
2757 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2758 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
2759 			ixgbe_service_event_schedule(adapter);
2760 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2761 					IXGBE_EICR_GPI_SDP0_X540);
2762 		}
2763 		if (eicr & IXGBE_EICR_ECC) {
2764 			e_info(link, "Received ECC Err, initiating reset\n");
2765 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
2766 			ixgbe_service_event_schedule(adapter);
2767 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2768 		}
2769 		/* Handle Flow Director Full threshold interrupt */
2770 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2771 			int reinit_count = 0;
2772 			int i;
2773 			for (i = 0; i < adapter->num_tx_queues; i++) {
2774 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2775 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2776 						       &ring->state))
2777 					reinit_count++;
2778 			}
2779 			if (reinit_count) {
2780 				/* no more flow director interrupts until after init */
2781 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2782 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2783 				ixgbe_service_event_schedule(adapter);
2784 			}
2785 		}
2786 		ixgbe_check_sfp_event(adapter, eicr);
2787 		ixgbe_check_overtemp_event(adapter, eicr);
2788 		break;
2789 	default:
2790 		break;
2791 	}
2792 
2793 	ixgbe_check_fan_failure(adapter, eicr);
2794 
2795 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2796 		ixgbe_ptp_check_pps_event(adapter);
2797 
2798 	/* re-enable the original interrupt state, no lsc, no queues */
2799 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2800 		ixgbe_irq_enable(adapter, false, false);
2801 
2802 	return IRQ_HANDLED;
2803 }
2804 
2805 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2806 {
2807 	struct ixgbe_q_vector *q_vector = data;
2808 
2809 	/* EIAM disabled interrupts (on this vector) for us */
2810 
2811 	if (q_vector->rx.ring || q_vector->tx.ring)
2812 		napi_schedule_irqoff(&q_vector->napi);
2813 
2814 	return IRQ_HANDLED;
2815 }
2816 
2817 /**
2818  * ixgbe_poll - NAPI Rx polling callback
2819  * @napi: structure for representing this polling device
2820  * @budget: how many packets driver is allowed to clean
2821  *
2822  * This function is used for legacy and MSI, NAPI mode
2823  **/
2824 int ixgbe_poll(struct napi_struct *napi, int budget)
2825 {
2826 	struct ixgbe_q_vector *q_vector =
2827 				container_of(napi, struct ixgbe_q_vector, napi);
2828 	struct ixgbe_adapter *adapter = q_vector->adapter;
2829 	struct ixgbe_ring *ring;
2830 	int per_ring_budget, work_done = 0;
2831 	bool clean_complete = true;
2832 
2833 #ifdef CONFIG_IXGBE_DCA
2834 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2835 		ixgbe_update_dca(q_vector);
2836 #endif
2837 
2838 	ixgbe_for_each_ring(ring, q_vector->tx) {
2839 		if (!ixgbe_clean_tx_irq(q_vector, ring, budget))
2840 			clean_complete = false;
2841 	}
2842 
2843 	/* Exit if we are called by netpoll */
2844 	if (budget <= 0)
2845 		return budget;
2846 
2847 	/* attempt to distribute budget to each queue fairly, but don't allow
2848 	 * the budget to go below 1 because we'll exit polling */
2849 	if (q_vector->rx.count > 1)
2850 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2851 	else
2852 		per_ring_budget = budget;
2853 
2854 	ixgbe_for_each_ring(ring, q_vector->rx) {
2855 		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
2856 						 per_ring_budget);
2857 
2858 		work_done += cleaned;
2859 		if (cleaned >= per_ring_budget)
2860 			clean_complete = false;
2861 	}
2862 
2863 	/* If all work not completed, return budget and keep polling */
2864 	if (!clean_complete)
2865 		return budget;
2866 
2867 	/* all work done, exit the polling mode */
2868 	napi_complete_done(napi, work_done);
2869 	if (adapter->rx_itr_setting & 1)
2870 		ixgbe_set_itr(q_vector);
2871 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2872 		ixgbe_irq_enable_queues(adapter, BIT_ULL(q_vector->v_idx));
2873 
2874 	return min(work_done, budget - 1);
2875 }
2876 
2877 /**
2878  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2879  * @adapter: board private structure
2880  *
2881  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2882  * interrupts from the kernel.
2883  **/
2884 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2885 {
2886 	struct net_device *netdev = adapter->netdev;
2887 	int vector, err;
2888 	int ri = 0, ti = 0;
2889 
2890 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2891 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2892 		struct msix_entry *entry = &adapter->msix_entries[vector];
2893 
2894 		if (q_vector->tx.ring && q_vector->rx.ring) {
2895 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2896 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2897 			ti++;
2898 		} else if (q_vector->rx.ring) {
2899 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2900 				 "%s-%s-%d", netdev->name, "rx", ri++);
2901 		} else if (q_vector->tx.ring) {
2902 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2903 				 "%s-%s-%d", netdev->name, "tx", ti++);
2904 		} else {
2905 			/* skip this unused q_vector */
2906 			continue;
2907 		}
2908 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2909 				  q_vector->name, q_vector);
2910 		if (err) {
2911 			e_err(probe, "request_irq failed for MSIX interrupt "
2912 			      "Error: %d\n", err);
2913 			goto free_queue_irqs;
2914 		}
2915 		/* If Flow Director is enabled, set interrupt affinity */
2916 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2917 			/* assign the mask for this irq */
2918 			irq_set_affinity_hint(entry->vector,
2919 					      &q_vector->affinity_mask);
2920 		}
2921 	}
2922 
2923 	err = request_irq(adapter->msix_entries[vector].vector,
2924 			  ixgbe_msix_other, 0, netdev->name, adapter);
2925 	if (err) {
2926 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2927 		goto free_queue_irqs;
2928 	}
2929 
2930 	return 0;
2931 
2932 free_queue_irqs:
2933 	while (vector) {
2934 		vector--;
2935 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2936 				      NULL);
2937 		free_irq(adapter->msix_entries[vector].vector,
2938 			 adapter->q_vector[vector]);
2939 	}
2940 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2941 	pci_disable_msix(adapter->pdev);
2942 	kfree(adapter->msix_entries);
2943 	adapter->msix_entries = NULL;
2944 	return err;
2945 }
2946 
2947 /**
2948  * ixgbe_intr - legacy mode Interrupt Handler
2949  * @irq: interrupt number
2950  * @data: pointer to a network interface device structure
2951  **/
2952 static irqreturn_t ixgbe_intr(int irq, void *data)
2953 {
2954 	struct ixgbe_adapter *adapter = data;
2955 	struct ixgbe_hw *hw = &adapter->hw;
2956 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2957 	u32 eicr;
2958 
2959 	/*
2960 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2961 	 * before the read of EICR.
2962 	 */
2963 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2964 
2965 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2966 	 * therefore no explicit interrupt disable is necessary */
2967 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2968 	if (!eicr) {
2969 		/*
2970 		 * shared interrupt alert!
2971 		 * make sure interrupts are enabled because the read will
2972 		 * have disabled interrupts due to EIAM
2973 		 * finish the workaround of silicon errata on 82598.  Unmask
2974 		 * the interrupt that we masked before the EICR read.
2975 		 */
2976 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2977 			ixgbe_irq_enable(adapter, true, true);
2978 		return IRQ_NONE;	/* Not our interrupt */
2979 	}
2980 
2981 	if (eicr & IXGBE_EICR_LSC)
2982 		ixgbe_check_lsc(adapter);
2983 
2984 	switch (hw->mac.type) {
2985 	case ixgbe_mac_82599EB:
2986 		ixgbe_check_sfp_event(adapter, eicr);
2987 		/* Fall through */
2988 	case ixgbe_mac_X540:
2989 	case ixgbe_mac_X550:
2990 	case ixgbe_mac_X550EM_x:
2991 	case ixgbe_mac_x550em_a:
2992 		if (eicr & IXGBE_EICR_ECC) {
2993 			e_info(link, "Received ECC Err, initiating reset\n");
2994 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
2995 			ixgbe_service_event_schedule(adapter);
2996 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2997 		}
2998 		ixgbe_check_overtemp_event(adapter, eicr);
2999 		break;
3000 	default:
3001 		break;
3002 	}
3003 
3004 	ixgbe_check_fan_failure(adapter, eicr);
3005 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3006 		ixgbe_ptp_check_pps_event(adapter);
3007 
3008 	/* would disable interrupts here but EIAM disabled it */
3009 	napi_schedule_irqoff(&q_vector->napi);
3010 
3011 	/*
3012 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3013 	 * ixgbe_poll will re-enable the queue interrupts
3014 	 */
3015 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3016 		ixgbe_irq_enable(adapter, false, false);
3017 
3018 	return IRQ_HANDLED;
3019 }
3020 
3021 /**
3022  * ixgbe_request_irq - initialize interrupts
3023  * @adapter: board private structure
3024  *
3025  * Attempts to configure interrupts using the best available
3026  * capabilities of the hardware and kernel.
3027  **/
3028 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3029 {
3030 	struct net_device *netdev = adapter->netdev;
3031 	int err;
3032 
3033 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3034 		err = ixgbe_request_msix_irqs(adapter);
3035 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3036 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3037 				  netdev->name, adapter);
3038 	else
3039 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3040 				  netdev->name, adapter);
3041 
3042 	if (err)
3043 		e_err(probe, "request_irq failed, Error %d\n", err);
3044 
3045 	return err;
3046 }
3047 
3048 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3049 {
3050 	int vector;
3051 
3052 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3053 		free_irq(adapter->pdev->irq, adapter);
3054 		return;
3055 	}
3056 
3057 	if (!adapter->msix_entries)
3058 		return;
3059 
3060 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3061 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3062 		struct msix_entry *entry = &adapter->msix_entries[vector];
3063 
3064 		/* free only the irqs that were actually requested */
3065 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3066 			continue;
3067 
3068 		/* clear the affinity_mask in the IRQ descriptor */
3069 		irq_set_affinity_hint(entry->vector, NULL);
3070 
3071 		free_irq(entry->vector, q_vector);
3072 	}
3073 
3074 	free_irq(adapter->msix_entries[vector].vector, adapter);
3075 }
3076 
3077 /**
3078  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3079  * @adapter: board private structure
3080  **/
3081 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3082 {
3083 	switch (adapter->hw.mac.type) {
3084 	case ixgbe_mac_82598EB:
3085 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3086 		break;
3087 	case ixgbe_mac_82599EB:
3088 	case ixgbe_mac_X540:
3089 	case ixgbe_mac_X550:
3090 	case ixgbe_mac_X550EM_x:
3091 	case ixgbe_mac_x550em_a:
3092 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3093 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3094 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3095 		break;
3096 	default:
3097 		break;
3098 	}
3099 	IXGBE_WRITE_FLUSH(&adapter->hw);
3100 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3101 		int vector;
3102 
3103 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3104 			synchronize_irq(adapter->msix_entries[vector].vector);
3105 
3106 		synchronize_irq(adapter->msix_entries[vector++].vector);
3107 	} else {
3108 		synchronize_irq(adapter->pdev->irq);
3109 	}
3110 }
3111 
3112 /**
3113  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3114  *
3115  **/
3116 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3117 {
3118 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3119 
3120 	ixgbe_write_eitr(q_vector);
3121 
3122 	ixgbe_set_ivar(adapter, 0, 0, 0);
3123 	ixgbe_set_ivar(adapter, 1, 0, 0);
3124 
3125 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3126 }
3127 
3128 /**
3129  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3130  * @adapter: board private structure
3131  * @ring: structure containing ring specific data
3132  *
3133  * Configure the Tx descriptor ring after a reset.
3134  **/
3135 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3136 			     struct ixgbe_ring *ring)
3137 {
3138 	struct ixgbe_hw *hw = &adapter->hw;
3139 	u64 tdba = ring->dma;
3140 	int wait_loop = 10;
3141 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3142 	u8 reg_idx = ring->reg_idx;
3143 
3144 	/* disable queue to avoid issues while updating state */
3145 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3146 	IXGBE_WRITE_FLUSH(hw);
3147 
3148 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3149 			(tdba & DMA_BIT_MASK(32)));
3150 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3151 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3152 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3153 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3154 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3155 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3156 
3157 	/*
3158 	 * set WTHRESH to encourage burst writeback, it should not be set
3159 	 * higher than 1 when:
3160 	 * - ITR is 0 as it could cause false TX hangs
3161 	 * - ITR is set to > 100k int/sec and BQL is enabled
3162 	 *
3163 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3164 	 * to or less than the number of on chip descriptors, which is
3165 	 * currently 40.
3166 	 */
3167 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3168 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3169 	else
3170 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3171 
3172 	/*
3173 	 * Setting PTHRESH to 32 both improves performance
3174 	 * and avoids a TX hang with DFP enabled
3175 	 */
3176 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3177 		   32;		/* PTHRESH = 32 */
3178 
3179 	/* reinitialize flowdirector state */
3180 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3181 		ring->atr_sample_rate = adapter->atr_sample_rate;
3182 		ring->atr_count = 0;
3183 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3184 	} else {
3185 		ring->atr_sample_rate = 0;
3186 	}
3187 
3188 	/* initialize XPS */
3189 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3190 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3191 
3192 		if (q_vector)
3193 			netif_set_xps_queue(ring->netdev,
3194 					    &q_vector->affinity_mask,
3195 					    ring->queue_index);
3196 	}
3197 
3198 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3199 
3200 	/* enable queue */
3201 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3202 
3203 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3204 	if (hw->mac.type == ixgbe_mac_82598EB &&
3205 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3206 		return;
3207 
3208 	/* poll to verify queue is enabled */
3209 	do {
3210 		usleep_range(1000, 2000);
3211 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3212 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3213 	if (!wait_loop)
3214 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3215 }
3216 
3217 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3218 {
3219 	struct ixgbe_hw *hw = &adapter->hw;
3220 	u32 rttdcs, mtqc;
3221 	u8 tcs = netdev_get_num_tc(adapter->netdev);
3222 
3223 	if (hw->mac.type == ixgbe_mac_82598EB)
3224 		return;
3225 
3226 	/* disable the arbiter while setting MTQC */
3227 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3228 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3229 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3230 
3231 	/* set transmit pool layout */
3232 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3233 		mtqc = IXGBE_MTQC_VT_ENA;
3234 		if (tcs > 4)
3235 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3236 		else if (tcs > 1)
3237 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3238 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3239 			 IXGBE_82599_VMDQ_4Q_MASK)
3240 			mtqc |= IXGBE_MTQC_32VF;
3241 		else
3242 			mtqc |= IXGBE_MTQC_64VF;
3243 	} else {
3244 		if (tcs > 4)
3245 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3246 		else if (tcs > 1)
3247 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3248 		else
3249 			mtqc = IXGBE_MTQC_64Q_1PB;
3250 	}
3251 
3252 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3253 
3254 	/* Enable Security TX Buffer IFG for multiple pb */
3255 	if (tcs) {
3256 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3257 		sectx |= IXGBE_SECTX_DCB;
3258 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3259 	}
3260 
3261 	/* re-enable the arbiter */
3262 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3263 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3264 }
3265 
3266 /**
3267  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3268  * @adapter: board private structure
3269  *
3270  * Configure the Tx unit of the MAC after a reset.
3271  **/
3272 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3273 {
3274 	struct ixgbe_hw *hw = &adapter->hw;
3275 	u32 dmatxctl;
3276 	u32 i;
3277 
3278 	ixgbe_setup_mtqc(adapter);
3279 
3280 	if (hw->mac.type != ixgbe_mac_82598EB) {
3281 		/* DMATXCTL.EN must be before Tx queues are enabled */
3282 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3283 		dmatxctl |= IXGBE_DMATXCTL_TE;
3284 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3285 	}
3286 
3287 	/* Setup the HW Tx Head and Tail descriptor pointers */
3288 	for (i = 0; i < adapter->num_tx_queues; i++)
3289 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3290 }
3291 
3292 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3293 				 struct ixgbe_ring *ring)
3294 {
3295 	struct ixgbe_hw *hw = &adapter->hw;
3296 	u8 reg_idx = ring->reg_idx;
3297 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3298 
3299 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3300 
3301 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3302 }
3303 
3304 static void ixgbe_disable_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 #ifdef CONFIG_IXGBE_DCB
3317 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3318 #else
3319 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3320 #endif
3321 {
3322 	int i;
3323 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3324 
3325 	if (adapter->ixgbe_ieee_pfc)
3326 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3327 
3328 	/*
3329 	 * We should set the drop enable bit if:
3330 	 *  SR-IOV is enabled
3331 	 *   or
3332 	 *  Number of Rx queues > 1 and flow control is disabled
3333 	 *
3334 	 *  This allows us to avoid head of line blocking for security
3335 	 *  and performance reasons.
3336 	 */
3337 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3338 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3339 		for (i = 0; i < adapter->num_rx_queues; i++)
3340 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3341 	} else {
3342 		for (i = 0; i < adapter->num_rx_queues; i++)
3343 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3344 	}
3345 }
3346 
3347 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3348 
3349 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3350 				   struct ixgbe_ring *rx_ring)
3351 {
3352 	struct ixgbe_hw *hw = &adapter->hw;
3353 	u32 srrctl;
3354 	u8 reg_idx = rx_ring->reg_idx;
3355 
3356 	if (hw->mac.type == ixgbe_mac_82598EB) {
3357 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3358 
3359 		/*
3360 		 * if VMDq is not active we must program one srrctl register
3361 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3362 		 */
3363 		reg_idx &= mask;
3364 	}
3365 
3366 	/* configure header buffer length, needed for RSC */
3367 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3368 
3369 	/* configure the packet buffer length */
3370 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3371 
3372 	/* configure descriptor type */
3373 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3374 
3375 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3376 }
3377 
3378 /**
3379  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3380  * @adapter: device handle
3381  *
3382  *  - 82598/82599/X540:     128
3383  *  - X550(non-SRIOV mode): 512
3384  *  - X550(SRIOV mode):     64
3385  */
3386 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3387 {
3388 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3389 		return 128;
3390 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3391 		return 64;
3392 	else
3393 		return 512;
3394 }
3395 
3396 /**
3397  * ixgbe_store_reta - Write the RETA table to HW
3398  * @adapter: device handle
3399  *
3400  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3401  */
3402 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3403 {
3404 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3405 	struct ixgbe_hw *hw = &adapter->hw;
3406 	u32 reta = 0;
3407 	u32 indices_multi;
3408 	u8 *indir_tbl = adapter->rss_indir_tbl;
3409 
3410 	/* Fill out the redirection table as follows:
3411 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3412 	 *    indices.
3413 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3414 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3415 	 */
3416 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3417 		indices_multi = 0x11;
3418 	else
3419 		indices_multi = 0x1;
3420 
3421 	/* Write redirection table to HW */
3422 	for (i = 0; i < reta_entries; i++) {
3423 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3424 		if ((i & 3) == 3) {
3425 			if (i < 128)
3426 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3427 			else
3428 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3429 						reta);
3430 			reta = 0;
3431 		}
3432 	}
3433 }
3434 
3435 /**
3436  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3437  * @adapter: device handle
3438  *
3439  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3440  */
3441 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3442 {
3443 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3444 	struct ixgbe_hw *hw = &adapter->hw;
3445 	u32 vfreta = 0;
3446 	unsigned int pf_pool = adapter->num_vfs;
3447 
3448 	/* Write redirection table to HW */
3449 	for (i = 0; i < reta_entries; i++) {
3450 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3451 		if ((i & 3) == 3) {
3452 			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3453 					vfreta);
3454 			vfreta = 0;
3455 		}
3456 	}
3457 }
3458 
3459 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3460 {
3461 	struct ixgbe_hw *hw = &adapter->hw;
3462 	u32 i, j;
3463 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3464 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3465 
3466 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3467 	 * make full use of any rings they may have.  We will use the
3468 	 * PSRTYPE register to control how many rings we use within the PF.
3469 	 */
3470 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3471 		rss_i = 4;
3472 
3473 	/* Fill out hash function seeds */
3474 	for (i = 0; i < 10; i++)
3475 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3476 
3477 	/* Fill out redirection table */
3478 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3479 
3480 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3481 		if (j == rss_i)
3482 			j = 0;
3483 
3484 		adapter->rss_indir_tbl[i] = j;
3485 	}
3486 
3487 	ixgbe_store_reta(adapter);
3488 }
3489 
3490 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3491 {
3492 	struct ixgbe_hw *hw = &adapter->hw;
3493 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3494 	unsigned int pf_pool = adapter->num_vfs;
3495 	int i, j;
3496 
3497 	/* Fill out hash function seeds */
3498 	for (i = 0; i < 10; i++)
3499 		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool),
3500 				adapter->rss_key[i]);
3501 
3502 	/* Fill out the redirection table */
3503 	for (i = 0, j = 0; i < 64; i++, j++) {
3504 		if (j == rss_i)
3505 			j = 0;
3506 
3507 		adapter->rss_indir_tbl[i] = j;
3508 	}
3509 
3510 	ixgbe_store_vfreta(adapter);
3511 }
3512 
3513 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3514 {
3515 	struct ixgbe_hw *hw = &adapter->hw;
3516 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3517 	u32 rxcsum;
3518 
3519 	/* Disable indicating checksum in descriptor, enables RSS hash */
3520 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3521 	rxcsum |= IXGBE_RXCSUM_PCSD;
3522 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3523 
3524 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3525 		if (adapter->ring_feature[RING_F_RSS].mask)
3526 			mrqc = IXGBE_MRQC_RSSEN;
3527 	} else {
3528 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3529 
3530 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3531 			if (tcs > 4)
3532 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3533 			else if (tcs > 1)
3534 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3535 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3536 				 IXGBE_82599_VMDQ_4Q_MASK)
3537 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3538 			else
3539 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3540 		} else {
3541 			if (tcs > 4)
3542 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3543 			else if (tcs > 1)
3544 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3545 			else
3546 				mrqc = IXGBE_MRQC_RSSEN;
3547 		}
3548 	}
3549 
3550 	/* Perform hash on these packet types */
3551 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3552 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3553 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3554 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3555 
3556 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3557 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3558 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3559 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3560 
3561 	netdev_rss_key_fill(adapter->rss_key, sizeof(adapter->rss_key));
3562 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3563 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3564 		unsigned int pf_pool = adapter->num_vfs;
3565 
3566 		/* Enable VF RSS mode */
3567 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3568 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3569 
3570 		/* Setup RSS through the VF registers */
3571 		ixgbe_setup_vfreta(adapter);
3572 		vfmrqc = IXGBE_MRQC_RSSEN;
3573 		vfmrqc |= rss_field;
3574 		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3575 	} else {
3576 		ixgbe_setup_reta(adapter);
3577 		mrqc |= rss_field;
3578 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3579 	}
3580 }
3581 
3582 /**
3583  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3584  * @adapter:    address of board private structure
3585  * @index:      index of ring to set
3586  **/
3587 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3588 				   struct ixgbe_ring *ring)
3589 {
3590 	struct ixgbe_hw *hw = &adapter->hw;
3591 	u32 rscctrl;
3592 	u8 reg_idx = ring->reg_idx;
3593 
3594 	if (!ring_is_rsc_enabled(ring))
3595 		return;
3596 
3597 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3598 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3599 	/*
3600 	 * we must limit the number of descriptors so that the
3601 	 * total size of max desc * buf_len is not greater
3602 	 * than 65536
3603 	 */
3604 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3605 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3606 }
3607 
3608 #define IXGBE_MAX_RX_DESC_POLL 10
3609 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3610 				       struct ixgbe_ring *ring)
3611 {
3612 	struct ixgbe_hw *hw = &adapter->hw;
3613 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3614 	u32 rxdctl;
3615 	u8 reg_idx = ring->reg_idx;
3616 
3617 	if (ixgbe_removed(hw->hw_addr))
3618 		return;
3619 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3620 	if (hw->mac.type == ixgbe_mac_82598EB &&
3621 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3622 		return;
3623 
3624 	do {
3625 		usleep_range(1000, 2000);
3626 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3627 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3628 
3629 	if (!wait_loop) {
3630 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3631 		      "the polling period\n", reg_idx);
3632 	}
3633 }
3634 
3635 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3636 			    struct ixgbe_ring *ring)
3637 {
3638 	struct ixgbe_hw *hw = &adapter->hw;
3639 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3640 	u32 rxdctl;
3641 	u8 reg_idx = ring->reg_idx;
3642 
3643 	if (ixgbe_removed(hw->hw_addr))
3644 		return;
3645 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3646 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3647 
3648 	/* write value back with RXDCTL.ENABLE bit cleared */
3649 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3650 
3651 	if (hw->mac.type == ixgbe_mac_82598EB &&
3652 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3653 		return;
3654 
3655 	/* the hardware may take up to 100us to really disable the rx queue */
3656 	do {
3657 		udelay(10);
3658 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3659 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3660 
3661 	if (!wait_loop) {
3662 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3663 		      "the polling period\n", reg_idx);
3664 	}
3665 }
3666 
3667 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3668 			     struct ixgbe_ring *ring)
3669 {
3670 	struct ixgbe_hw *hw = &adapter->hw;
3671 	u64 rdba = ring->dma;
3672 	u32 rxdctl;
3673 	u8 reg_idx = ring->reg_idx;
3674 
3675 	/* disable queue to avoid issues while updating state */
3676 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3677 	ixgbe_disable_rx_queue(adapter, ring);
3678 
3679 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3680 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3681 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3682 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3683 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
3684 	IXGBE_WRITE_FLUSH(hw);
3685 
3686 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3687 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3688 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3689 
3690 	ixgbe_configure_srrctl(adapter, ring);
3691 	ixgbe_configure_rscctl(adapter, ring);
3692 
3693 	if (hw->mac.type == ixgbe_mac_82598EB) {
3694 		/*
3695 		 * enable cache line friendly hardware writes:
3696 		 * PTHRESH=32 descriptors (half the internal cache),
3697 		 * this also removes ugly rx_no_buffer_count increment
3698 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3699 		 * WTHRESH=8 burst writeback up to two cache lines
3700 		 */
3701 		rxdctl &= ~0x3FFFFF;
3702 		rxdctl |=  0x080420;
3703 	}
3704 
3705 	/* enable receive descriptor ring */
3706 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3707 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3708 
3709 	ixgbe_rx_desc_queue_enable(adapter, ring);
3710 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3711 }
3712 
3713 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3714 {
3715 	struct ixgbe_hw *hw = &adapter->hw;
3716 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3717 	u16 pool;
3718 
3719 	/* PSRTYPE must be initialized in non 82598 adapters */
3720 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3721 		      IXGBE_PSRTYPE_UDPHDR |
3722 		      IXGBE_PSRTYPE_IPV4HDR |
3723 		      IXGBE_PSRTYPE_L2HDR |
3724 		      IXGBE_PSRTYPE_IPV6HDR;
3725 
3726 	if (hw->mac.type == ixgbe_mac_82598EB)
3727 		return;
3728 
3729 	if (rss_i > 3)
3730 		psrtype |= 2u << 29;
3731 	else if (rss_i > 1)
3732 		psrtype |= 1u << 29;
3733 
3734 	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3735 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3736 }
3737 
3738 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3739 {
3740 	struct ixgbe_hw *hw = &adapter->hw;
3741 	u32 reg_offset, vf_shift;
3742 	u32 gcr_ext, vmdctl;
3743 	int i;
3744 
3745 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3746 		return;
3747 
3748 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3749 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3750 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3751 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3752 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3753 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3754 
3755 	vf_shift = VMDQ_P(0) % 32;
3756 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3757 
3758 	/* Enable only the PF's pool for Tx/Rx */
3759 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
3760 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3761 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
3762 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3763 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
3764 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3765 
3766 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3767 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3768 
3769 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
3770 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
3771 
3772 	/*
3773 	 * Set up VF register offsets for selected VT Mode,
3774 	 * i.e. 32 or 64 VFs for SR-IOV
3775 	 */
3776 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3777 	case IXGBE_82599_VMDQ_8Q_MASK:
3778 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3779 		break;
3780 	case IXGBE_82599_VMDQ_4Q_MASK:
3781 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3782 		break;
3783 	default:
3784 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3785 		break;
3786 	}
3787 
3788 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3789 
3790 	for (i = 0; i < adapter->num_vfs; i++) {
3791 		/* configure spoof checking */
3792 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
3793 					  adapter->vfinfo[i].spoofchk_enabled);
3794 
3795 		/* Enable/Disable RSS query feature  */
3796 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
3797 					  adapter->vfinfo[i].rss_query_enabled);
3798 	}
3799 }
3800 
3801 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3802 {
3803 	struct ixgbe_hw *hw = &adapter->hw;
3804 	struct net_device *netdev = adapter->netdev;
3805 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3806 	struct ixgbe_ring *rx_ring;
3807 	int i;
3808 	u32 mhadd, hlreg0;
3809 
3810 #ifdef IXGBE_FCOE
3811 	/* adjust max frame to be able to do baby jumbo for FCoE */
3812 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3813 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3814 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3815 
3816 #endif /* IXGBE_FCOE */
3817 
3818 	/* adjust max frame to be at least the size of a standard frame */
3819 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3820 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3821 
3822 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3823 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3824 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3825 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3826 
3827 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3828 	}
3829 
3830 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3831 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3832 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3833 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3834 
3835 	/*
3836 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3837 	 * the Base and Length of the Rx Descriptor Ring
3838 	 */
3839 	for (i = 0; i < adapter->num_rx_queues; i++) {
3840 		rx_ring = adapter->rx_ring[i];
3841 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3842 			set_ring_rsc_enabled(rx_ring);
3843 		else
3844 			clear_ring_rsc_enabled(rx_ring);
3845 	}
3846 }
3847 
3848 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3849 {
3850 	struct ixgbe_hw *hw = &adapter->hw;
3851 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3852 
3853 	switch (hw->mac.type) {
3854 	case ixgbe_mac_82598EB:
3855 		/*
3856 		 * For VMDq support of different descriptor types or
3857 		 * buffer sizes through the use of multiple SRRCTL
3858 		 * registers, RDRXCTL.MVMEN must be set to 1
3859 		 *
3860 		 * also, the manual doesn't mention it clearly but DCA hints
3861 		 * will only use queue 0's tags unless this bit is set.  Side
3862 		 * effects of setting this bit are only that SRRCTL must be
3863 		 * fully programmed [0..15]
3864 		 */
3865 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3866 		break;
3867 	case ixgbe_mac_X550:
3868 	case ixgbe_mac_X550EM_x:
3869 	case ixgbe_mac_x550em_a:
3870 		if (adapter->num_vfs)
3871 			rdrxctl |= IXGBE_RDRXCTL_PSP;
3872 		/* fall through for older HW */
3873 	case ixgbe_mac_82599EB:
3874 	case ixgbe_mac_X540:
3875 		/* Disable RSC for ACK packets */
3876 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3877 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3878 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3879 		/* hardware requires some bits to be set by default */
3880 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3881 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3882 		break;
3883 	default:
3884 		/* We should do nothing since we don't know this hardware */
3885 		return;
3886 	}
3887 
3888 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3889 }
3890 
3891 /**
3892  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3893  * @adapter: board private structure
3894  *
3895  * Configure the Rx unit of the MAC after a reset.
3896  **/
3897 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3898 {
3899 	struct ixgbe_hw *hw = &adapter->hw;
3900 	int i;
3901 	u32 rxctrl, rfctl;
3902 
3903 	/* disable receives while setting up the descriptors */
3904 	hw->mac.ops.disable_rx(hw);
3905 
3906 	ixgbe_setup_psrtype(adapter);
3907 	ixgbe_setup_rdrxctl(adapter);
3908 
3909 	/* RSC Setup */
3910 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
3911 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
3912 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
3913 		rfctl |= IXGBE_RFCTL_RSC_DIS;
3914 
3915 	/* disable NFS filtering */
3916 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
3917 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3918 
3919 	/* Program registers for the distribution of queues */
3920 	ixgbe_setup_mrqc(adapter);
3921 
3922 	/* set_rx_buffer_len must be called before ring initialization */
3923 	ixgbe_set_rx_buffer_len(adapter);
3924 
3925 	/*
3926 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3927 	 * the Base and Length of the Rx Descriptor Ring
3928 	 */
3929 	for (i = 0; i < adapter->num_rx_queues; i++)
3930 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3931 
3932 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3933 	/* disable drop enable for 82598 parts */
3934 	if (hw->mac.type == ixgbe_mac_82598EB)
3935 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3936 
3937 	/* enable all receives */
3938 	rxctrl |= IXGBE_RXCTRL_RXEN;
3939 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3940 }
3941 
3942 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3943 				 __be16 proto, u16 vid)
3944 {
3945 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3946 	struct ixgbe_hw *hw = &adapter->hw;
3947 
3948 	/* add VID to filter table */
3949 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
3950 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
3951 
3952 	set_bit(vid, adapter->active_vlans);
3953 
3954 	return 0;
3955 }
3956 
3957 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
3958 {
3959 	u32 vlvf;
3960 	int idx;
3961 
3962 	/* short cut the special case */
3963 	if (vlan == 0)
3964 		return 0;
3965 
3966 	/* Search for the vlan id in the VLVF entries */
3967 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
3968 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
3969 		if ((vlvf & VLAN_VID_MASK) == vlan)
3970 			break;
3971 	}
3972 
3973 	return idx;
3974 }
3975 
3976 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
3977 {
3978 	struct ixgbe_hw *hw = &adapter->hw;
3979 	u32 bits, word;
3980 	int idx;
3981 
3982 	idx = ixgbe_find_vlvf_entry(hw, vid);
3983 	if (!idx)
3984 		return;
3985 
3986 	/* See if any other pools are set for this VLAN filter
3987 	 * entry other than the PF.
3988 	 */
3989 	word = idx * 2 + (VMDQ_P(0) / 32);
3990 	bits = ~BIT(VMDQ_P(0) % 32);
3991 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
3992 
3993 	/* Disable the filter so this falls into the default pool. */
3994 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
3995 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
3996 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
3997 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
3998 	}
3999 }
4000 
4001 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4002 				  __be16 proto, u16 vid)
4003 {
4004 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4005 	struct ixgbe_hw *hw = &adapter->hw;
4006 
4007 	/* remove VID from filter table */
4008 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4009 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4010 
4011 	clear_bit(vid, adapter->active_vlans);
4012 
4013 	return 0;
4014 }
4015 
4016 /**
4017  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4018  * @adapter: driver data
4019  */
4020 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4021 {
4022 	struct ixgbe_hw *hw = &adapter->hw;
4023 	u32 vlnctrl;
4024 	int i, j;
4025 
4026 	switch (hw->mac.type) {
4027 	case ixgbe_mac_82598EB:
4028 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4029 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4030 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4031 		break;
4032 	case ixgbe_mac_82599EB:
4033 	case ixgbe_mac_X540:
4034 	case ixgbe_mac_X550:
4035 	case ixgbe_mac_X550EM_x:
4036 	case ixgbe_mac_x550em_a:
4037 		for (i = 0; i < adapter->num_rx_queues; i++) {
4038 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4039 
4040 			if (ring->l2_accel_priv)
4041 				continue;
4042 			j = ring->reg_idx;
4043 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4044 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4045 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4046 		}
4047 		break;
4048 	default:
4049 		break;
4050 	}
4051 }
4052 
4053 /**
4054  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4055  * @adapter: driver data
4056  */
4057 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4058 {
4059 	struct ixgbe_hw *hw = &adapter->hw;
4060 	u32 vlnctrl;
4061 	int i, j;
4062 
4063 	switch (hw->mac.type) {
4064 	case ixgbe_mac_82598EB:
4065 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4066 		vlnctrl |= IXGBE_VLNCTRL_VME;
4067 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4068 		break;
4069 	case ixgbe_mac_82599EB:
4070 	case ixgbe_mac_X540:
4071 	case ixgbe_mac_X550:
4072 	case ixgbe_mac_X550EM_x:
4073 	case ixgbe_mac_x550em_a:
4074 		for (i = 0; i < adapter->num_rx_queues; i++) {
4075 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4076 
4077 			if (ring->l2_accel_priv)
4078 				continue;
4079 			j = ring->reg_idx;
4080 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4081 			vlnctrl |= IXGBE_RXDCTL_VME;
4082 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4083 		}
4084 		break;
4085 	default:
4086 		break;
4087 	}
4088 }
4089 
4090 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4091 {
4092 	struct ixgbe_hw *hw = &adapter->hw;
4093 	u32 vlnctrl, i;
4094 
4095 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4096 
4097 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4098 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4099 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4100 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4101 	} else {
4102 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4103 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4104 		return;
4105 	}
4106 
4107 	/* Nothing to do for 82598 */
4108 	if (hw->mac.type == ixgbe_mac_82598EB)
4109 		return;
4110 
4111 	/* We are already in VLAN promisc, nothing to do */
4112 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4113 		return;
4114 
4115 	/* Set flag so we don't redo unnecessary work */
4116 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4117 
4118 	/* Add PF to all active pools */
4119 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4120 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4121 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4122 
4123 		vlvfb |= BIT(VMDQ_P(0) % 32);
4124 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4125 	}
4126 
4127 	/* Set all bits in the VLAN filter table array */
4128 	for (i = hw->mac.vft_size; i--;)
4129 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4130 }
4131 
4132 #define VFTA_BLOCK_SIZE 8
4133 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4134 {
4135 	struct ixgbe_hw *hw = &adapter->hw;
4136 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4137 	u32 vid_start = vfta_offset * 32;
4138 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4139 	u32 i, vid, word, bits;
4140 
4141 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4142 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4143 
4144 		/* pull VLAN ID from VLVF */
4145 		vid = vlvf & VLAN_VID_MASK;
4146 
4147 		/* only concern outselves with a certain range */
4148 		if (vid < vid_start || vid >= vid_end)
4149 			continue;
4150 
4151 		if (vlvf) {
4152 			/* record VLAN ID in VFTA */
4153 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4154 
4155 			/* if PF is part of this then continue */
4156 			if (test_bit(vid, adapter->active_vlans))
4157 				continue;
4158 		}
4159 
4160 		/* remove PF from the pool */
4161 		word = i * 2 + VMDQ_P(0) / 32;
4162 		bits = ~BIT(VMDQ_P(0) % 32);
4163 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4164 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4165 	}
4166 
4167 	/* extract values from active_vlans and write back to VFTA */
4168 	for (i = VFTA_BLOCK_SIZE; i--;) {
4169 		vid = (vfta_offset + i) * 32;
4170 		word = vid / BITS_PER_LONG;
4171 		bits = vid % BITS_PER_LONG;
4172 
4173 		vfta[i] |= adapter->active_vlans[word] >> bits;
4174 
4175 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4176 	}
4177 }
4178 
4179 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4180 {
4181 	struct ixgbe_hw *hw = &adapter->hw;
4182 	u32 vlnctrl, i;
4183 
4184 	/* Set VLAN filtering to enabled */
4185 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4186 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4187 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4188 
4189 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4190 	    hw->mac.type == ixgbe_mac_82598EB)
4191 		return;
4192 
4193 	/* We are not in VLAN promisc, nothing to do */
4194 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4195 		return;
4196 
4197 	/* Set flag so we don't redo unnecessary work */
4198 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4199 
4200 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4201 		ixgbe_scrub_vfta(adapter, i);
4202 }
4203 
4204 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4205 {
4206 	u16 vid = 1;
4207 
4208 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4209 
4210 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4211 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4212 }
4213 
4214 /**
4215  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4216  * @netdev: network interface device structure
4217  *
4218  * Writes multicast address list to the MTA hash table.
4219  * Returns: -ENOMEM on failure
4220  *                0 on no addresses written
4221  *                X on writing X addresses to MTA
4222  **/
4223 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4224 {
4225 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4226 	struct ixgbe_hw *hw = &adapter->hw;
4227 
4228 	if (!netif_running(netdev))
4229 		return 0;
4230 
4231 	if (hw->mac.ops.update_mc_addr_list)
4232 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4233 	else
4234 		return -ENOMEM;
4235 
4236 #ifdef CONFIG_PCI_IOV
4237 	ixgbe_restore_vf_multicasts(adapter);
4238 #endif
4239 
4240 	return netdev_mc_count(netdev);
4241 }
4242 
4243 #ifdef CONFIG_PCI_IOV
4244 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4245 {
4246 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4247 	struct ixgbe_hw *hw = &adapter->hw;
4248 	int i;
4249 
4250 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4251 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4252 
4253 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4254 			hw->mac.ops.set_rar(hw, i,
4255 					    mac_table->addr,
4256 					    mac_table->pool,
4257 					    IXGBE_RAH_AV);
4258 		else
4259 			hw->mac.ops.clear_rar(hw, i);
4260 	}
4261 }
4262 
4263 #endif
4264 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4265 {
4266 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4267 	struct ixgbe_hw *hw = &adapter->hw;
4268 	int i;
4269 
4270 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4271 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4272 			continue;
4273 
4274 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4275 
4276 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4277 			hw->mac.ops.set_rar(hw, i,
4278 					    mac_table->addr,
4279 					    mac_table->pool,
4280 					    IXGBE_RAH_AV);
4281 		else
4282 			hw->mac.ops.clear_rar(hw, i);
4283 	}
4284 }
4285 
4286 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4287 {
4288 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4289 	struct ixgbe_hw *hw = &adapter->hw;
4290 	int i;
4291 
4292 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4293 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4294 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4295 	}
4296 
4297 	ixgbe_sync_mac_table(adapter);
4298 }
4299 
4300 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4301 {
4302 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4303 	struct ixgbe_hw *hw = &adapter->hw;
4304 	int i, count = 0;
4305 
4306 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4307 		/* do not count default RAR as available */
4308 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4309 			continue;
4310 
4311 		/* only count unused and addresses that belong to us */
4312 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4313 			if (mac_table->pool != pool)
4314 				continue;
4315 		}
4316 
4317 		count++;
4318 	}
4319 
4320 	return count;
4321 }
4322 
4323 /* this function destroys the first RAR entry */
4324 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4325 {
4326 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4327 	struct ixgbe_hw *hw = &adapter->hw;
4328 
4329 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4330 	mac_table->pool = VMDQ_P(0);
4331 
4332 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4333 
4334 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4335 			    IXGBE_RAH_AV);
4336 }
4337 
4338 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4339 			 const u8 *addr, u16 pool)
4340 {
4341 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4342 	struct ixgbe_hw *hw = &adapter->hw;
4343 	int i;
4344 
4345 	if (is_zero_ether_addr(addr))
4346 		return -EINVAL;
4347 
4348 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4349 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4350 			continue;
4351 
4352 		ether_addr_copy(mac_table->addr, addr);
4353 		mac_table->pool = pool;
4354 
4355 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4356 				    IXGBE_MAC_STATE_IN_USE;
4357 
4358 		ixgbe_sync_mac_table(adapter);
4359 
4360 		return i;
4361 	}
4362 
4363 	return -ENOMEM;
4364 }
4365 
4366 int ixgbe_del_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 	/* search table for addr, if found clear IN_USE flag and sync */
4377 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4378 		/* we can only delete an entry if it is in use */
4379 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4380 			continue;
4381 		/* we only care about entries that belong to the given pool */
4382 		if (mac_table->pool != pool)
4383 			continue;
4384 		/* we only care about a specific MAC address */
4385 		if (!ether_addr_equal(addr, mac_table->addr))
4386 			continue;
4387 
4388 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4389 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4390 
4391 		ixgbe_sync_mac_table(adapter);
4392 
4393 		return 0;
4394 	}
4395 
4396 	return -ENOMEM;
4397 }
4398 /**
4399  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4400  * @netdev: network interface device structure
4401  *
4402  * Writes unicast address list to the RAR table.
4403  * Returns: -ENOMEM on failure/insufficient address space
4404  *                0 on no addresses written
4405  *                X on writing X addresses to the RAR table
4406  **/
4407 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4408 {
4409 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4410 	int count = 0;
4411 
4412 	/* return ENOMEM indicating insufficient memory for addresses */
4413 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter, vfn))
4414 		return -ENOMEM;
4415 
4416 	if (!netdev_uc_empty(netdev)) {
4417 		struct netdev_hw_addr *ha;
4418 		netdev_for_each_uc_addr(ha, netdev) {
4419 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4420 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4421 			count++;
4422 		}
4423 	}
4424 	return count;
4425 }
4426 
4427 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4428 {
4429 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4430 	int ret;
4431 
4432 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4433 
4434 	return min_t(int, ret, 0);
4435 }
4436 
4437 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4438 {
4439 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4440 
4441 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4442 
4443 	return 0;
4444 }
4445 
4446 /**
4447  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4448  * @netdev: network interface device structure
4449  *
4450  * The set_rx_method entry point is called whenever the unicast/multicast
4451  * address list or the network interface flags are updated.  This routine is
4452  * responsible for configuring the hardware for proper unicast, multicast and
4453  * promiscuous mode.
4454  **/
4455 void ixgbe_set_rx_mode(struct net_device *netdev)
4456 {
4457 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4458 	struct ixgbe_hw *hw = &adapter->hw;
4459 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4460 	netdev_features_t features = netdev->features;
4461 	int count;
4462 
4463 	/* Check for Promiscuous and All Multicast modes */
4464 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4465 
4466 	/* set all bits that we expect to always be set */
4467 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4468 	fctrl |= IXGBE_FCTRL_BAM;
4469 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4470 	fctrl |= IXGBE_FCTRL_PMCF;
4471 
4472 	/* clear the bits we are changing the status of */
4473 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4474 	if (netdev->flags & IFF_PROMISC) {
4475 		hw->addr_ctrl.user_set_promisc = true;
4476 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4477 		vmolr |= IXGBE_VMOLR_MPE;
4478 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4479 	} else {
4480 		if (netdev->flags & IFF_ALLMULTI) {
4481 			fctrl |= IXGBE_FCTRL_MPE;
4482 			vmolr |= IXGBE_VMOLR_MPE;
4483 		}
4484 		hw->addr_ctrl.user_set_promisc = false;
4485 	}
4486 
4487 	/*
4488 	 * Write addresses to available RAR registers, if there is not
4489 	 * sufficient space to store all the addresses then enable
4490 	 * unicast promiscuous mode
4491 	 */
4492 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4493 		fctrl |= IXGBE_FCTRL_UPE;
4494 		vmolr |= IXGBE_VMOLR_ROPE;
4495 	}
4496 
4497 	/* Write addresses to the MTA, if the attempt fails
4498 	 * then we should just turn on promiscuous mode so
4499 	 * that we can at least receive multicast traffic
4500 	 */
4501 	count = ixgbe_write_mc_addr_list(netdev);
4502 	if (count < 0) {
4503 		fctrl |= IXGBE_FCTRL_MPE;
4504 		vmolr |= IXGBE_VMOLR_MPE;
4505 	} else if (count) {
4506 		vmolr |= IXGBE_VMOLR_ROMPE;
4507 	}
4508 
4509 	if (hw->mac.type != ixgbe_mac_82598EB) {
4510 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4511 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4512 			   IXGBE_VMOLR_ROPE);
4513 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4514 	}
4515 
4516 	/* This is useful for sniffing bad packets. */
4517 	if (features & NETIF_F_RXALL) {
4518 		/* UPE and MPE will be handled by normal PROMISC logic
4519 		 * in e1000e_set_rx_mode */
4520 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4521 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4522 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4523 
4524 		fctrl &= ~(IXGBE_FCTRL_DPF);
4525 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4526 	}
4527 
4528 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4529 
4530 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4531 		ixgbe_vlan_strip_enable(adapter);
4532 	else
4533 		ixgbe_vlan_strip_disable(adapter);
4534 
4535 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4536 		ixgbe_vlan_promisc_disable(adapter);
4537 	else
4538 		ixgbe_vlan_promisc_enable(adapter);
4539 }
4540 
4541 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4542 {
4543 	int q_idx;
4544 
4545 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4546 		napi_enable(&adapter->q_vector[q_idx]->napi);
4547 }
4548 
4549 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4550 {
4551 	int q_idx;
4552 
4553 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4554 		napi_disable(&adapter->q_vector[q_idx]->napi);
4555 }
4556 
4557 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
4558 {
4559 	struct ixgbe_hw *hw = &adapter->hw;
4560 	u32 vxlanctrl;
4561 
4562 	if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
4563 				IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
4564 		return;
4565 
4566 	vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) && ~mask;
4567 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
4568 
4569 	if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
4570 		adapter->vxlan_port = 0;
4571 
4572 	if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
4573 		adapter->geneve_port = 0;
4574 }
4575 
4576 #ifdef CONFIG_IXGBE_DCB
4577 /**
4578  * ixgbe_configure_dcb - Configure DCB hardware
4579  * @adapter: ixgbe adapter struct
4580  *
4581  * This is called by the driver on open to configure the DCB hardware.
4582  * This is also called by the gennetlink interface when reconfiguring
4583  * the DCB state.
4584  */
4585 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4586 {
4587 	struct ixgbe_hw *hw = &adapter->hw;
4588 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4589 
4590 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4591 		if (hw->mac.type == ixgbe_mac_82598EB)
4592 			netif_set_gso_max_size(adapter->netdev, 65536);
4593 		return;
4594 	}
4595 
4596 	if (hw->mac.type == ixgbe_mac_82598EB)
4597 		netif_set_gso_max_size(adapter->netdev, 32768);
4598 
4599 #ifdef IXGBE_FCOE
4600 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4601 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4602 #endif
4603 
4604 	/* reconfigure the hardware */
4605 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4606 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4607 						DCB_TX_CONFIG);
4608 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4609 						DCB_RX_CONFIG);
4610 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4611 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4612 		ixgbe_dcb_hw_ets(&adapter->hw,
4613 				 adapter->ixgbe_ieee_ets,
4614 				 max_frame);
4615 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4616 					adapter->ixgbe_ieee_pfc->pfc_en,
4617 					adapter->ixgbe_ieee_ets->prio_tc);
4618 	}
4619 
4620 	/* Enable RSS Hash per TC */
4621 	if (hw->mac.type != ixgbe_mac_82598EB) {
4622 		u32 msb = 0;
4623 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4624 
4625 		while (rss_i) {
4626 			msb++;
4627 			rss_i >>= 1;
4628 		}
4629 
4630 		/* write msb to all 8 TCs in one write */
4631 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4632 	}
4633 }
4634 #endif
4635 
4636 /* Additional bittime to account for IXGBE framing */
4637 #define IXGBE_ETH_FRAMING 20
4638 
4639 /**
4640  * ixgbe_hpbthresh - calculate high water mark for flow control
4641  *
4642  * @adapter: board private structure to calculate for
4643  * @pb: packet buffer to calculate
4644  */
4645 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4646 {
4647 	struct ixgbe_hw *hw = &adapter->hw;
4648 	struct net_device *dev = adapter->netdev;
4649 	int link, tc, kb, marker;
4650 	u32 dv_id, rx_pba;
4651 
4652 	/* Calculate max LAN frame size */
4653 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4654 
4655 #ifdef IXGBE_FCOE
4656 	/* FCoE traffic class uses FCOE jumbo frames */
4657 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4658 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4659 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4660 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4661 #endif
4662 
4663 	/* Calculate delay value for device */
4664 	switch (hw->mac.type) {
4665 	case ixgbe_mac_X540:
4666 	case ixgbe_mac_X550:
4667 	case ixgbe_mac_X550EM_x:
4668 	case ixgbe_mac_x550em_a:
4669 		dv_id = IXGBE_DV_X540(link, tc);
4670 		break;
4671 	default:
4672 		dv_id = IXGBE_DV(link, tc);
4673 		break;
4674 	}
4675 
4676 	/* Loopback switch introduces additional latency */
4677 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4678 		dv_id += IXGBE_B2BT(tc);
4679 
4680 	/* Delay value is calculated in bit times convert to KB */
4681 	kb = IXGBE_BT2KB(dv_id);
4682 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4683 
4684 	marker = rx_pba - kb;
4685 
4686 	/* It is possible that the packet buffer is not large enough
4687 	 * to provide required headroom. In this case throw an error
4688 	 * to user and a do the best we can.
4689 	 */
4690 	if (marker < 0) {
4691 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4692 			    "headroom to support flow control."
4693 			    "Decrease MTU or number of traffic classes\n", pb);
4694 		marker = tc + 1;
4695 	}
4696 
4697 	return marker;
4698 }
4699 
4700 /**
4701  * ixgbe_lpbthresh - calculate low water mark for for flow control
4702  *
4703  * @adapter: board private structure to calculate for
4704  * @pb: packet buffer to calculate
4705  */
4706 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4707 {
4708 	struct ixgbe_hw *hw = &adapter->hw;
4709 	struct net_device *dev = adapter->netdev;
4710 	int tc;
4711 	u32 dv_id;
4712 
4713 	/* Calculate max LAN frame size */
4714 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4715 
4716 #ifdef IXGBE_FCOE
4717 	/* FCoE traffic class uses FCOE jumbo frames */
4718 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4719 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4720 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4721 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4722 #endif
4723 
4724 	/* Calculate delay value for device */
4725 	switch (hw->mac.type) {
4726 	case ixgbe_mac_X540:
4727 	case ixgbe_mac_X550:
4728 	case ixgbe_mac_X550EM_x:
4729 	case ixgbe_mac_x550em_a:
4730 		dv_id = IXGBE_LOW_DV_X540(tc);
4731 		break;
4732 	default:
4733 		dv_id = IXGBE_LOW_DV(tc);
4734 		break;
4735 	}
4736 
4737 	/* Delay value is calculated in bit times convert to KB */
4738 	return IXGBE_BT2KB(dv_id);
4739 }
4740 
4741 /*
4742  * ixgbe_pbthresh_setup - calculate and setup high low water marks
4743  */
4744 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4745 {
4746 	struct ixgbe_hw *hw = &adapter->hw;
4747 	int num_tc = netdev_get_num_tc(adapter->netdev);
4748 	int i;
4749 
4750 	if (!num_tc)
4751 		num_tc = 1;
4752 
4753 	for (i = 0; i < num_tc; i++) {
4754 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4755 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4756 
4757 		/* Low water marks must not be larger than high water marks */
4758 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4759 			hw->fc.low_water[i] = 0;
4760 	}
4761 
4762 	for (; i < MAX_TRAFFIC_CLASS; i++)
4763 		hw->fc.high_water[i] = 0;
4764 }
4765 
4766 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4767 {
4768 	struct ixgbe_hw *hw = &adapter->hw;
4769 	int hdrm;
4770 	u8 tc = netdev_get_num_tc(adapter->netdev);
4771 
4772 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4773 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4774 		hdrm = 32 << adapter->fdir_pballoc;
4775 	else
4776 		hdrm = 0;
4777 
4778 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4779 	ixgbe_pbthresh_setup(adapter);
4780 }
4781 
4782 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4783 {
4784 	struct ixgbe_hw *hw = &adapter->hw;
4785 	struct hlist_node *node2;
4786 	struct ixgbe_fdir_filter *filter;
4787 
4788 	spin_lock(&adapter->fdir_perfect_lock);
4789 
4790 	if (!hlist_empty(&adapter->fdir_filter_list))
4791 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4792 
4793 	hlist_for_each_entry_safe(filter, node2,
4794 				  &adapter->fdir_filter_list, fdir_node) {
4795 		ixgbe_fdir_write_perfect_filter_82599(hw,
4796 				&filter->filter,
4797 				filter->sw_idx,
4798 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4799 				IXGBE_FDIR_DROP_QUEUE :
4800 				adapter->rx_ring[filter->action]->reg_idx);
4801 	}
4802 
4803 	spin_unlock(&adapter->fdir_perfect_lock);
4804 }
4805 
4806 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4807 				      struct ixgbe_adapter *adapter)
4808 {
4809 	struct ixgbe_hw *hw = &adapter->hw;
4810 	u32 vmolr;
4811 
4812 	/* No unicast promiscuous support for VMDQ devices. */
4813 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4814 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4815 
4816 	/* clear the affected bit */
4817 	vmolr &= ~IXGBE_VMOLR_MPE;
4818 
4819 	if (dev->flags & IFF_ALLMULTI) {
4820 		vmolr |= IXGBE_VMOLR_MPE;
4821 	} else {
4822 		vmolr |= IXGBE_VMOLR_ROMPE;
4823 		hw->mac.ops.update_mc_addr_list(hw, dev);
4824 	}
4825 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4826 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4827 }
4828 
4829 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4830 {
4831 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4832 	int rss_i = adapter->num_rx_queues_per_pool;
4833 	struct ixgbe_hw *hw = &adapter->hw;
4834 	u16 pool = vadapter->pool;
4835 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4836 		      IXGBE_PSRTYPE_UDPHDR |
4837 		      IXGBE_PSRTYPE_IPV4HDR |
4838 		      IXGBE_PSRTYPE_L2HDR |
4839 		      IXGBE_PSRTYPE_IPV6HDR;
4840 
4841 	if (hw->mac.type == ixgbe_mac_82598EB)
4842 		return;
4843 
4844 	if (rss_i > 3)
4845 		psrtype |= 2u << 29;
4846 	else if (rss_i > 1)
4847 		psrtype |= 1u << 29;
4848 
4849 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4850 }
4851 
4852 /**
4853  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4854  * @rx_ring: ring to free buffers from
4855  **/
4856 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4857 {
4858 	struct device *dev = rx_ring->dev;
4859 	unsigned long size;
4860 	u16 i;
4861 
4862 	/* ring already cleared, nothing to do */
4863 	if (!rx_ring->rx_buffer_info)
4864 		return;
4865 
4866 	/* Free all the Rx ring sk_buffs */
4867 	for (i = 0; i < rx_ring->count; i++) {
4868 		struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4869 
4870 		if (rx_buffer->skb) {
4871 			struct sk_buff *skb = rx_buffer->skb;
4872 			if (IXGBE_CB(skb)->page_released)
4873 				dma_unmap_page(dev,
4874 					       IXGBE_CB(skb)->dma,
4875 					       ixgbe_rx_bufsz(rx_ring),
4876 					       DMA_FROM_DEVICE);
4877 			dev_kfree_skb(skb);
4878 			rx_buffer->skb = NULL;
4879 		}
4880 
4881 		if (!rx_buffer->page)
4882 			continue;
4883 
4884 		dma_unmap_page(dev, rx_buffer->dma,
4885 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
4886 		__free_pages(rx_buffer->page, ixgbe_rx_pg_order(rx_ring));
4887 
4888 		rx_buffer->page = NULL;
4889 	}
4890 
4891 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4892 	memset(rx_ring->rx_buffer_info, 0, size);
4893 
4894 	/* Zero out the descriptor ring */
4895 	memset(rx_ring->desc, 0, rx_ring->size);
4896 
4897 	rx_ring->next_to_alloc = 0;
4898 	rx_ring->next_to_clean = 0;
4899 	rx_ring->next_to_use = 0;
4900 }
4901 
4902 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4903 				   struct ixgbe_ring *rx_ring)
4904 {
4905 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4906 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4907 
4908 	/* shutdown specific queue receive and wait for dma to settle */
4909 	ixgbe_disable_rx_queue(adapter, rx_ring);
4910 	usleep_range(10000, 20000);
4911 	ixgbe_irq_disable_queues(adapter, BIT_ULL(index));
4912 	ixgbe_clean_rx_ring(rx_ring);
4913 	rx_ring->l2_accel_priv = NULL;
4914 }
4915 
4916 static int ixgbe_fwd_ring_down(struct net_device *vdev,
4917 			       struct ixgbe_fwd_adapter *accel)
4918 {
4919 	struct ixgbe_adapter *adapter = accel->real_adapter;
4920 	unsigned int rxbase = accel->rx_base_queue;
4921 	unsigned int txbase = accel->tx_base_queue;
4922 	int i;
4923 
4924 	netif_tx_stop_all_queues(vdev);
4925 
4926 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4927 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4928 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4929 	}
4930 
4931 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4932 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4933 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4934 	}
4935 
4936 
4937 	return 0;
4938 }
4939 
4940 static int ixgbe_fwd_ring_up(struct net_device *vdev,
4941 			     struct ixgbe_fwd_adapter *accel)
4942 {
4943 	struct ixgbe_adapter *adapter = accel->real_adapter;
4944 	unsigned int rxbase, txbase, queues;
4945 	int i, baseq, err = 0;
4946 
4947 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4948 		return 0;
4949 
4950 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4951 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4952 		   accel->pool, adapter->num_rx_pools,
4953 		   baseq, baseq + adapter->num_rx_queues_per_pool,
4954 		   adapter->fwd_bitmask);
4955 
4956 	accel->netdev = vdev;
4957 	accel->rx_base_queue = rxbase = baseq;
4958 	accel->tx_base_queue = txbase = baseq;
4959 
4960 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4961 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4962 
4963 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4964 		adapter->rx_ring[rxbase + i]->netdev = vdev;
4965 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
4966 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
4967 	}
4968 
4969 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4970 		adapter->tx_ring[txbase + i]->netdev = vdev;
4971 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
4972 	}
4973 
4974 	queues = min_t(unsigned int,
4975 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
4976 	err = netif_set_real_num_tx_queues(vdev, queues);
4977 	if (err)
4978 		goto fwd_queue_err;
4979 
4980 	err = netif_set_real_num_rx_queues(vdev, queues);
4981 	if (err)
4982 		goto fwd_queue_err;
4983 
4984 	if (is_valid_ether_addr(vdev->dev_addr))
4985 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
4986 
4987 	ixgbe_fwd_psrtype(accel);
4988 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
4989 	return err;
4990 fwd_queue_err:
4991 	ixgbe_fwd_ring_down(vdev, accel);
4992 	return err;
4993 }
4994 
4995 static int ixgbe_upper_dev_walk(struct net_device *upper, void *data)
4996 {
4997 	if (netif_is_macvlan(upper)) {
4998 		struct macvlan_dev *dfwd = netdev_priv(upper);
4999 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
5000 
5001 		if (dfwd->fwd_priv)
5002 			ixgbe_fwd_ring_up(upper, vadapter);
5003 	}
5004 
5005 	return 0;
5006 }
5007 
5008 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5009 {
5010 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5011 				      ixgbe_upper_dev_walk, NULL);
5012 }
5013 
5014 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5015 {
5016 	struct ixgbe_hw *hw = &adapter->hw;
5017 
5018 	ixgbe_configure_pb(adapter);
5019 #ifdef CONFIG_IXGBE_DCB
5020 	ixgbe_configure_dcb(adapter);
5021 #endif
5022 	/*
5023 	 * We must restore virtualization before VLANs or else
5024 	 * the VLVF registers will not be populated
5025 	 */
5026 	ixgbe_configure_virtualization(adapter);
5027 
5028 	ixgbe_set_rx_mode(adapter->netdev);
5029 	ixgbe_restore_vlan(adapter);
5030 
5031 	switch (hw->mac.type) {
5032 	case ixgbe_mac_82599EB:
5033 	case ixgbe_mac_X540:
5034 		hw->mac.ops.disable_rx_buff(hw);
5035 		break;
5036 	default:
5037 		break;
5038 	}
5039 
5040 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5041 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5042 						adapter->fdir_pballoc);
5043 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5044 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5045 					      adapter->fdir_pballoc);
5046 		ixgbe_fdir_filter_restore(adapter);
5047 	}
5048 
5049 	switch (hw->mac.type) {
5050 	case ixgbe_mac_82599EB:
5051 	case ixgbe_mac_X540:
5052 		hw->mac.ops.enable_rx_buff(hw);
5053 		break;
5054 	default:
5055 		break;
5056 	}
5057 
5058 #ifdef CONFIG_IXGBE_DCA
5059 	/* configure DCA */
5060 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5061 		ixgbe_setup_dca(adapter);
5062 #endif /* CONFIG_IXGBE_DCA */
5063 
5064 #ifdef IXGBE_FCOE
5065 	/* configure FCoE L2 filters, redirection table, and Rx control */
5066 	ixgbe_configure_fcoe(adapter);
5067 
5068 #endif /* IXGBE_FCOE */
5069 	ixgbe_configure_tx(adapter);
5070 	ixgbe_configure_rx(adapter);
5071 	ixgbe_configure_dfwd(adapter);
5072 }
5073 
5074 /**
5075  * ixgbe_sfp_link_config - set up SFP+ link
5076  * @adapter: pointer to private adapter struct
5077  **/
5078 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5079 {
5080 	/*
5081 	 * We are assuming the worst case scenario here, and that
5082 	 * is that an SFP was inserted/removed after the reset
5083 	 * but before SFP detection was enabled.  As such the best
5084 	 * solution is to just start searching as soon as we start
5085 	 */
5086 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5087 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5088 
5089 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5090 	adapter->sfp_poll_time = 0;
5091 }
5092 
5093 /**
5094  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5095  * @hw: pointer to private hardware struct
5096  *
5097  * Returns 0 on success, negative on failure
5098  **/
5099 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5100 {
5101 	u32 speed;
5102 	bool autoneg, link_up = false;
5103 	int ret = IXGBE_ERR_LINK_SETUP;
5104 
5105 	if (hw->mac.ops.check_link)
5106 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5107 
5108 	if (ret)
5109 		return ret;
5110 
5111 	speed = hw->phy.autoneg_advertised;
5112 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5113 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5114 							&autoneg);
5115 	if (ret)
5116 		return ret;
5117 
5118 	if (hw->mac.ops.setup_link)
5119 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5120 
5121 	return ret;
5122 }
5123 
5124 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5125 {
5126 	struct ixgbe_hw *hw = &adapter->hw;
5127 	u32 gpie = 0;
5128 
5129 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5130 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5131 		       IXGBE_GPIE_OCD;
5132 		gpie |= IXGBE_GPIE_EIAME;
5133 		/*
5134 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5135 		 * this saves a register write for every interrupt
5136 		 */
5137 		switch (hw->mac.type) {
5138 		case ixgbe_mac_82598EB:
5139 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5140 			break;
5141 		case ixgbe_mac_82599EB:
5142 		case ixgbe_mac_X540:
5143 		case ixgbe_mac_X550:
5144 		case ixgbe_mac_X550EM_x:
5145 		case ixgbe_mac_x550em_a:
5146 		default:
5147 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5148 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5149 			break;
5150 		}
5151 	} else {
5152 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5153 		 * specifically only auto mask tx and rx interrupts */
5154 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5155 	}
5156 
5157 	/* XXX: to interrupt immediately for EICS writes, enable this */
5158 	/* gpie |= IXGBE_GPIE_EIMEN; */
5159 
5160 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5161 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5162 
5163 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5164 		case IXGBE_82599_VMDQ_8Q_MASK:
5165 			gpie |= IXGBE_GPIE_VTMODE_16;
5166 			break;
5167 		case IXGBE_82599_VMDQ_4Q_MASK:
5168 			gpie |= IXGBE_GPIE_VTMODE_32;
5169 			break;
5170 		default:
5171 			gpie |= IXGBE_GPIE_VTMODE_64;
5172 			break;
5173 		}
5174 	}
5175 
5176 	/* Enable Thermal over heat sensor interrupt */
5177 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5178 		switch (adapter->hw.mac.type) {
5179 		case ixgbe_mac_82599EB:
5180 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5181 			break;
5182 		default:
5183 			break;
5184 		}
5185 	}
5186 
5187 	/* Enable fan failure interrupt */
5188 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5189 		gpie |= IXGBE_SDP1_GPIEN(hw);
5190 
5191 	switch (hw->mac.type) {
5192 	case ixgbe_mac_82599EB:
5193 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5194 		break;
5195 	case ixgbe_mac_X550EM_x:
5196 	case ixgbe_mac_x550em_a:
5197 		gpie |= IXGBE_SDP0_GPIEN_X540;
5198 		break;
5199 	default:
5200 		break;
5201 	}
5202 
5203 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5204 }
5205 
5206 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5207 {
5208 	struct ixgbe_hw *hw = &adapter->hw;
5209 	int err;
5210 	u32 ctrl_ext;
5211 
5212 	ixgbe_get_hw_control(adapter);
5213 	ixgbe_setup_gpie(adapter);
5214 
5215 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5216 		ixgbe_configure_msix(adapter);
5217 	else
5218 		ixgbe_configure_msi_and_legacy(adapter);
5219 
5220 	/* enable the optics for 82599 SFP+ fiber */
5221 	if (hw->mac.ops.enable_tx_laser)
5222 		hw->mac.ops.enable_tx_laser(hw);
5223 
5224 	if (hw->phy.ops.set_phy_power)
5225 		hw->phy.ops.set_phy_power(hw, true);
5226 
5227 	smp_mb__before_atomic();
5228 	clear_bit(__IXGBE_DOWN, &adapter->state);
5229 	ixgbe_napi_enable_all(adapter);
5230 
5231 	if (ixgbe_is_sfp(hw)) {
5232 		ixgbe_sfp_link_config(adapter);
5233 	} else {
5234 		err = ixgbe_non_sfp_link_config(hw);
5235 		if (err)
5236 			e_err(probe, "link_config FAILED %d\n", err);
5237 	}
5238 
5239 	/* clear any pending interrupts, may auto mask */
5240 	IXGBE_READ_REG(hw, IXGBE_EICR);
5241 	ixgbe_irq_enable(adapter, true, true);
5242 
5243 	/*
5244 	 * If this adapter has a fan, check to see if we had a failure
5245 	 * before we enabled the interrupt.
5246 	 */
5247 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5248 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5249 		if (esdp & IXGBE_ESDP_SDP1)
5250 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5251 	}
5252 
5253 	/* bring the link up in the watchdog, this could race with our first
5254 	 * link up interrupt but shouldn't be a problem */
5255 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5256 	adapter->link_check_timeout = jiffies;
5257 	mod_timer(&adapter->service_timer, jiffies);
5258 
5259 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5260 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5261 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5262 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5263 }
5264 
5265 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5266 {
5267 	WARN_ON(in_interrupt());
5268 	/* put off any impending NetWatchDogTimeout */
5269 	netif_trans_update(adapter->netdev);
5270 
5271 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5272 		usleep_range(1000, 2000);
5273 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5274 		ixgbe_watchdog_link_is_down(adapter);
5275 	ixgbe_down(adapter);
5276 	/*
5277 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5278 	 * back up to give the VFs time to respond to the reset.  The
5279 	 * two second wait is based upon the watchdog timer cycle in
5280 	 * the VF driver.
5281 	 */
5282 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5283 		msleep(2000);
5284 	ixgbe_up(adapter);
5285 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5286 }
5287 
5288 void ixgbe_up(struct ixgbe_adapter *adapter)
5289 {
5290 	/* hardware has been reset, we need to reload some things */
5291 	ixgbe_configure(adapter);
5292 
5293 	ixgbe_up_complete(adapter);
5294 }
5295 
5296 void ixgbe_reset(struct ixgbe_adapter *adapter)
5297 {
5298 	struct ixgbe_hw *hw = &adapter->hw;
5299 	struct net_device *netdev = adapter->netdev;
5300 	int err;
5301 
5302 	if (ixgbe_removed(hw->hw_addr))
5303 		return;
5304 	/* lock SFP init bit to prevent race conditions with the watchdog */
5305 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5306 		usleep_range(1000, 2000);
5307 
5308 	/* clear all SFP and link config related flags while holding SFP_INIT */
5309 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5310 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5311 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5312 
5313 	err = hw->mac.ops.init_hw(hw);
5314 	switch (err) {
5315 	case 0:
5316 	case IXGBE_ERR_SFP_NOT_PRESENT:
5317 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5318 		break;
5319 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5320 		e_dev_err("master disable timed out\n");
5321 		break;
5322 	case IXGBE_ERR_EEPROM_VERSION:
5323 		/* We are running on a pre-production device, log a warning */
5324 		e_dev_warn("This device is a pre-production adapter/LOM. "
5325 			   "Please be aware there may be issues associated with "
5326 			   "your hardware.  If you are experiencing problems "
5327 			   "please contact your Intel or hardware "
5328 			   "representative who provided you with this "
5329 			   "hardware.\n");
5330 		break;
5331 	default:
5332 		e_dev_err("Hardware Error: %d\n", err);
5333 	}
5334 
5335 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5336 
5337 	/* flush entries out of MAC table */
5338 	ixgbe_flush_sw_mac_table(adapter);
5339 	__dev_uc_unsync(netdev, NULL);
5340 
5341 	/* do not flush user set addresses */
5342 	ixgbe_mac_set_default_filter(adapter);
5343 
5344 	/* update SAN MAC vmdq pool selection */
5345 	if (hw->mac.san_mac_rar_index)
5346 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5347 
5348 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5349 		ixgbe_ptp_reset(adapter);
5350 
5351 	if (hw->phy.ops.set_phy_power) {
5352 		if (!netif_running(adapter->netdev) && !adapter->wol)
5353 			hw->phy.ops.set_phy_power(hw, false);
5354 		else
5355 			hw->phy.ops.set_phy_power(hw, true);
5356 	}
5357 }
5358 
5359 /**
5360  * ixgbe_clean_tx_ring - Free Tx Buffers
5361  * @tx_ring: ring to be cleaned
5362  **/
5363 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5364 {
5365 	struct ixgbe_tx_buffer *tx_buffer_info;
5366 	unsigned long size;
5367 	u16 i;
5368 
5369 	/* ring already cleared, nothing to do */
5370 	if (!tx_ring->tx_buffer_info)
5371 		return;
5372 
5373 	/* Free all the Tx ring sk_buffs */
5374 	for (i = 0; i < tx_ring->count; i++) {
5375 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
5376 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
5377 	}
5378 
5379 	netdev_tx_reset_queue(txring_txq(tx_ring));
5380 
5381 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5382 	memset(tx_ring->tx_buffer_info, 0, size);
5383 
5384 	/* Zero out the descriptor ring */
5385 	memset(tx_ring->desc, 0, tx_ring->size);
5386 
5387 	tx_ring->next_to_use = 0;
5388 	tx_ring->next_to_clean = 0;
5389 }
5390 
5391 /**
5392  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5393  * @adapter: board private structure
5394  **/
5395 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5396 {
5397 	int i;
5398 
5399 	for (i = 0; i < adapter->num_rx_queues; i++)
5400 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5401 }
5402 
5403 /**
5404  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5405  * @adapter: board private structure
5406  **/
5407 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5408 {
5409 	int i;
5410 
5411 	for (i = 0; i < adapter->num_tx_queues; i++)
5412 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5413 }
5414 
5415 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5416 {
5417 	struct hlist_node *node2;
5418 	struct ixgbe_fdir_filter *filter;
5419 
5420 	spin_lock(&adapter->fdir_perfect_lock);
5421 
5422 	hlist_for_each_entry_safe(filter, node2,
5423 				  &adapter->fdir_filter_list, fdir_node) {
5424 		hlist_del(&filter->fdir_node);
5425 		kfree(filter);
5426 	}
5427 	adapter->fdir_filter_count = 0;
5428 
5429 	spin_unlock(&adapter->fdir_perfect_lock);
5430 }
5431 
5432 static int ixgbe_disable_macvlan(struct net_device *upper, void *data)
5433 {
5434 	if (netif_is_macvlan(upper)) {
5435 		struct macvlan_dev *vlan = netdev_priv(upper);
5436 
5437 		if (vlan->fwd_priv) {
5438 			netif_tx_stop_all_queues(upper);
5439 			netif_carrier_off(upper);
5440 			netif_tx_disable(upper);
5441 		}
5442 	}
5443 
5444 	return 0;
5445 }
5446 
5447 void ixgbe_down(struct ixgbe_adapter *adapter)
5448 {
5449 	struct net_device *netdev = adapter->netdev;
5450 	struct ixgbe_hw *hw = &adapter->hw;
5451 	int i;
5452 
5453 	/* signal that we are down to the interrupt handler */
5454 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5455 		return; /* do nothing if already down */
5456 
5457 	/* disable receives */
5458 	hw->mac.ops.disable_rx(hw);
5459 
5460 	/* disable all enabled rx queues */
5461 	for (i = 0; i < adapter->num_rx_queues; i++)
5462 		/* this call also flushes the previous write */
5463 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5464 
5465 	usleep_range(10000, 20000);
5466 
5467 	netif_tx_stop_all_queues(netdev);
5468 
5469 	/* call carrier off first to avoid false dev_watchdog timeouts */
5470 	netif_carrier_off(netdev);
5471 	netif_tx_disable(netdev);
5472 
5473 	/* disable any upper devices */
5474 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5475 				      ixgbe_disable_macvlan, NULL);
5476 
5477 	ixgbe_irq_disable(adapter);
5478 
5479 	ixgbe_napi_disable_all(adapter);
5480 
5481 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
5482 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5483 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5484 
5485 	del_timer_sync(&adapter->service_timer);
5486 
5487 	if (adapter->num_vfs) {
5488 		/* Clear EITR Select mapping */
5489 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5490 
5491 		/* Mark all the VFs as inactive */
5492 		for (i = 0 ; i < adapter->num_vfs; i++)
5493 			adapter->vfinfo[i].clear_to_send = false;
5494 
5495 		/* ping all the active vfs to let them know we are going down */
5496 		ixgbe_ping_all_vfs(adapter);
5497 
5498 		/* Disable all VFTE/VFRE TX/RX */
5499 		ixgbe_disable_tx_rx(adapter);
5500 	}
5501 
5502 	/* disable transmits in the hardware now that interrupts are off */
5503 	for (i = 0; i < adapter->num_tx_queues; i++) {
5504 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5505 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5506 	}
5507 
5508 	/* Disable the Tx DMA engine on 82599 and later MAC */
5509 	switch (hw->mac.type) {
5510 	case ixgbe_mac_82599EB:
5511 	case ixgbe_mac_X540:
5512 	case ixgbe_mac_X550:
5513 	case ixgbe_mac_X550EM_x:
5514 	case ixgbe_mac_x550em_a:
5515 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5516 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5517 				 ~IXGBE_DMATXCTL_TE));
5518 		break;
5519 	default:
5520 		break;
5521 	}
5522 
5523 	if (!pci_channel_offline(adapter->pdev))
5524 		ixgbe_reset(adapter);
5525 
5526 	/* power down the optics for 82599 SFP+ fiber */
5527 	if (hw->mac.ops.disable_tx_laser)
5528 		hw->mac.ops.disable_tx_laser(hw);
5529 
5530 	ixgbe_clean_all_tx_rings(adapter);
5531 	ixgbe_clean_all_rx_rings(adapter);
5532 }
5533 
5534 /**
5535  * ixgbe_eee_capable - helper function to determine EEE support on X550
5536  * @adapter: board private structure
5537  */
5538 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
5539 {
5540 	struct ixgbe_hw *hw = &adapter->hw;
5541 
5542 	switch (hw->device_id) {
5543 	case IXGBE_DEV_ID_X550EM_A_1G_T:
5544 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
5545 		if (!hw->phy.eee_speeds_supported)
5546 			break;
5547 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
5548 		if (!hw->phy.eee_speeds_advertised)
5549 			break;
5550 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
5551 		break;
5552 	default:
5553 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
5554 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
5555 		break;
5556 	}
5557 }
5558 
5559 /**
5560  * ixgbe_tx_timeout - Respond to a Tx Hang
5561  * @netdev: network interface device structure
5562  **/
5563 static void ixgbe_tx_timeout(struct net_device *netdev)
5564 {
5565 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5566 
5567 	/* Do the reset outside of interrupt context */
5568 	ixgbe_tx_timeout_reset(adapter);
5569 }
5570 
5571 #ifdef CONFIG_IXGBE_DCB
5572 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
5573 {
5574 	struct ixgbe_hw *hw = &adapter->hw;
5575 	struct tc_configuration *tc;
5576 	int j;
5577 
5578 	switch (hw->mac.type) {
5579 	case ixgbe_mac_82598EB:
5580 	case ixgbe_mac_82599EB:
5581 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5582 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5583 		break;
5584 	case ixgbe_mac_X540:
5585 	case ixgbe_mac_X550:
5586 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5587 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5588 		break;
5589 	case ixgbe_mac_X550EM_x:
5590 	case ixgbe_mac_x550em_a:
5591 	default:
5592 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
5593 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
5594 		break;
5595 	}
5596 
5597 	/* Configure DCB traffic classes */
5598 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5599 		tc = &adapter->dcb_cfg.tc_config[j];
5600 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5601 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5602 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5603 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5604 		tc->dcb_pfc = pfc_disabled;
5605 	}
5606 
5607 	/* Initialize default user to priority mapping, UPx->TC0 */
5608 	tc = &adapter->dcb_cfg.tc_config[0];
5609 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5610 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5611 
5612 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5613 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5614 	adapter->dcb_cfg.pfc_mode_enable = false;
5615 	adapter->dcb_set_bitmap = 0x00;
5616 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
5617 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5618 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5619 	       sizeof(adapter->temp_dcb_cfg));
5620 }
5621 #endif
5622 
5623 /**
5624  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5625  * @adapter: board private structure to initialize
5626  *
5627  * ixgbe_sw_init initializes the Adapter private data structure.
5628  * Fields are initialized based on PCI device information and
5629  * OS network device settings (MTU size).
5630  **/
5631 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
5632 			 const struct ixgbe_info *ii)
5633 {
5634 	struct ixgbe_hw *hw = &adapter->hw;
5635 	struct pci_dev *pdev = adapter->pdev;
5636 	unsigned int rss, fdir;
5637 	u32 fwsm;
5638 	int i;
5639 
5640 	/* PCI config space info */
5641 
5642 	hw->vendor_id = pdev->vendor;
5643 	hw->device_id = pdev->device;
5644 	hw->revision_id = pdev->revision;
5645 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5646 	hw->subsystem_device_id = pdev->subsystem_device;
5647 
5648 	/* get_invariants needs the device IDs */
5649 	ii->get_invariants(hw);
5650 
5651 	/* Set common capability flags and settings */
5652 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5653 	adapter->ring_feature[RING_F_RSS].limit = rss;
5654 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5655 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5656 	adapter->atr_sample_rate = 20;
5657 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5658 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5659 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5660 #ifdef CONFIG_IXGBE_DCA
5661 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5662 #endif
5663 #ifdef CONFIG_IXGBE_DCB
5664 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
5665 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
5666 #endif
5667 #ifdef IXGBE_FCOE
5668 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5669 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5670 #ifdef CONFIG_IXGBE_DCB
5671 	/* Default traffic class to use for FCoE */
5672 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5673 #endif /* CONFIG_IXGBE_DCB */
5674 #endif /* IXGBE_FCOE */
5675 
5676 	/* initialize static ixgbe jump table entries */
5677 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
5678 					  GFP_KERNEL);
5679 	if (!adapter->jump_tables[0])
5680 		return -ENOMEM;
5681 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
5682 
5683 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
5684 		adapter->jump_tables[i] = NULL;
5685 
5686 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5687 				     hw->mac.num_rar_entries,
5688 				     GFP_ATOMIC);
5689 	if (!adapter->mac_table)
5690 		return -ENOMEM;
5691 
5692 	/* Set MAC specific capability flags and exceptions */
5693 	switch (hw->mac.type) {
5694 	case ixgbe_mac_82598EB:
5695 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5696 
5697 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5698 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5699 
5700 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5701 		adapter->ring_feature[RING_F_FDIR].limit = 0;
5702 		adapter->atr_sample_rate = 0;
5703 		adapter->fdir_pballoc = 0;
5704 #ifdef IXGBE_FCOE
5705 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5706 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5707 #ifdef CONFIG_IXGBE_DCB
5708 		adapter->fcoe.up = 0;
5709 #endif /* IXGBE_DCB */
5710 #endif /* IXGBE_FCOE */
5711 		break;
5712 	case ixgbe_mac_82599EB:
5713 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5714 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5715 		break;
5716 	case ixgbe_mac_X540:
5717 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
5718 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5719 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5720 		break;
5721 	case ixgbe_mac_x550em_a:
5722 		adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
5723 		switch (hw->device_id) {
5724 		case IXGBE_DEV_ID_X550EM_A_1G_T:
5725 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
5726 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5727 			break;
5728 		default:
5729 			break;
5730 		}
5731 	/* fall through */
5732 	case ixgbe_mac_X550EM_x:
5733 #ifdef CONFIG_IXGBE_DCB
5734 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
5735 #endif
5736 #ifdef IXGBE_FCOE
5737 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5738 #ifdef CONFIG_IXGBE_DCB
5739 		adapter->fcoe.up = 0;
5740 #endif /* IXGBE_DCB */
5741 #endif /* IXGBE_FCOE */
5742 	/* Fall Through */
5743 	case ixgbe_mac_X550:
5744 		if (hw->mac.type == ixgbe_mac_X550)
5745 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5746 #ifdef CONFIG_IXGBE_DCA
5747 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5748 #endif
5749 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
5750 		break;
5751 	default:
5752 		break;
5753 	}
5754 
5755 #ifdef IXGBE_FCOE
5756 	/* FCoE support exists, always init the FCoE lock */
5757 	spin_lock_init(&adapter->fcoe.lock);
5758 
5759 #endif
5760 	/* n-tuple support exists, always init our spinlock */
5761 	spin_lock_init(&adapter->fdir_perfect_lock);
5762 
5763 #ifdef CONFIG_IXGBE_DCB
5764 	ixgbe_init_dcb(adapter);
5765 #endif
5766 
5767 	/* default flow control settings */
5768 	hw->fc.requested_mode = ixgbe_fc_full;
5769 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5770 	ixgbe_pbthresh_setup(adapter);
5771 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5772 	hw->fc.send_xon = true;
5773 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5774 
5775 #ifdef CONFIG_PCI_IOV
5776 	if (max_vfs > 0)
5777 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5778 
5779 	/* assign number of SR-IOV VFs */
5780 	if (hw->mac.type != ixgbe_mac_82598EB) {
5781 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5782 			adapter->num_vfs = 0;
5783 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5784 		} else {
5785 			adapter->num_vfs = max_vfs;
5786 		}
5787 	}
5788 #endif /* CONFIG_PCI_IOV */
5789 
5790 	/* enable itr by default in dynamic mode */
5791 	adapter->rx_itr_setting = 1;
5792 	adapter->tx_itr_setting = 1;
5793 
5794 	/* set default ring sizes */
5795 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5796 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5797 
5798 	/* set default work limits */
5799 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5800 
5801 	/* initialize eeprom parameters */
5802 	if (ixgbe_init_eeprom_params_generic(hw)) {
5803 		e_dev_err("EEPROM initialization failed\n");
5804 		return -EIO;
5805 	}
5806 
5807 	/* PF holds first pool slot */
5808 	set_bit(0, &adapter->fwd_bitmask);
5809 	set_bit(__IXGBE_DOWN, &adapter->state);
5810 
5811 	return 0;
5812 }
5813 
5814 /**
5815  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5816  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5817  *
5818  * Return 0 on success, negative on failure
5819  **/
5820 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5821 {
5822 	struct device *dev = tx_ring->dev;
5823 	int orig_node = dev_to_node(dev);
5824 	int ring_node = -1;
5825 	int size;
5826 
5827 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5828 
5829 	if (tx_ring->q_vector)
5830 		ring_node = tx_ring->q_vector->numa_node;
5831 
5832 	tx_ring->tx_buffer_info = vzalloc_node(size, ring_node);
5833 	if (!tx_ring->tx_buffer_info)
5834 		tx_ring->tx_buffer_info = vzalloc(size);
5835 	if (!tx_ring->tx_buffer_info)
5836 		goto err;
5837 
5838 	u64_stats_init(&tx_ring->syncp);
5839 
5840 	/* round up to nearest 4K */
5841 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5842 	tx_ring->size = ALIGN(tx_ring->size, 4096);
5843 
5844 	set_dev_node(dev, ring_node);
5845 	tx_ring->desc = dma_alloc_coherent(dev,
5846 					   tx_ring->size,
5847 					   &tx_ring->dma,
5848 					   GFP_KERNEL);
5849 	set_dev_node(dev, orig_node);
5850 	if (!tx_ring->desc)
5851 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5852 						   &tx_ring->dma, GFP_KERNEL);
5853 	if (!tx_ring->desc)
5854 		goto err;
5855 
5856 	tx_ring->next_to_use = 0;
5857 	tx_ring->next_to_clean = 0;
5858 	return 0;
5859 
5860 err:
5861 	vfree(tx_ring->tx_buffer_info);
5862 	tx_ring->tx_buffer_info = NULL;
5863 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5864 	return -ENOMEM;
5865 }
5866 
5867 /**
5868  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5869  * @adapter: board private structure
5870  *
5871  * If this function returns with an error, then it's possible one or
5872  * more of the rings is populated (while the rest are not).  It is the
5873  * callers duty to clean those orphaned rings.
5874  *
5875  * Return 0 on success, negative on failure
5876  **/
5877 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5878 {
5879 	int i, err = 0;
5880 
5881 	for (i = 0; i < adapter->num_tx_queues; i++) {
5882 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5883 		if (!err)
5884 			continue;
5885 
5886 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5887 		goto err_setup_tx;
5888 	}
5889 
5890 	return 0;
5891 err_setup_tx:
5892 	/* rewind the index freeing the rings as we go */
5893 	while (i--)
5894 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5895 	return err;
5896 }
5897 
5898 /**
5899  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5900  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5901  *
5902  * Returns 0 on success, negative on failure
5903  **/
5904 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5905 {
5906 	struct device *dev = rx_ring->dev;
5907 	int orig_node = dev_to_node(dev);
5908 	int ring_node = -1;
5909 	int size;
5910 
5911 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5912 
5913 	if (rx_ring->q_vector)
5914 		ring_node = rx_ring->q_vector->numa_node;
5915 
5916 	rx_ring->rx_buffer_info = vzalloc_node(size, ring_node);
5917 	if (!rx_ring->rx_buffer_info)
5918 		rx_ring->rx_buffer_info = vzalloc(size);
5919 	if (!rx_ring->rx_buffer_info)
5920 		goto err;
5921 
5922 	u64_stats_init(&rx_ring->syncp);
5923 
5924 	/* Round up to nearest 4K */
5925 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5926 	rx_ring->size = ALIGN(rx_ring->size, 4096);
5927 
5928 	set_dev_node(dev, ring_node);
5929 	rx_ring->desc = dma_alloc_coherent(dev,
5930 					   rx_ring->size,
5931 					   &rx_ring->dma,
5932 					   GFP_KERNEL);
5933 	set_dev_node(dev, orig_node);
5934 	if (!rx_ring->desc)
5935 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5936 						   &rx_ring->dma, GFP_KERNEL);
5937 	if (!rx_ring->desc)
5938 		goto err;
5939 
5940 	rx_ring->next_to_clean = 0;
5941 	rx_ring->next_to_use = 0;
5942 
5943 	return 0;
5944 err:
5945 	vfree(rx_ring->rx_buffer_info);
5946 	rx_ring->rx_buffer_info = NULL;
5947 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5948 	return -ENOMEM;
5949 }
5950 
5951 /**
5952  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5953  * @adapter: board private structure
5954  *
5955  * If this function returns with an error, then it's possible one or
5956  * more of the rings is populated (while the rest are not).  It is the
5957  * callers duty to clean those orphaned rings.
5958  *
5959  * Return 0 on success, negative on failure
5960  **/
5961 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5962 {
5963 	int i, err = 0;
5964 
5965 	for (i = 0; i < adapter->num_rx_queues; i++) {
5966 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5967 		if (!err)
5968 			continue;
5969 
5970 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5971 		goto err_setup_rx;
5972 	}
5973 
5974 #ifdef IXGBE_FCOE
5975 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5976 	if (!err)
5977 #endif
5978 		return 0;
5979 err_setup_rx:
5980 	/* rewind the index freeing the rings as we go */
5981 	while (i--)
5982 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5983 	return err;
5984 }
5985 
5986 /**
5987  * ixgbe_free_tx_resources - Free Tx Resources per Queue
5988  * @tx_ring: Tx descriptor ring for a specific queue
5989  *
5990  * Free all transmit software resources
5991  **/
5992 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5993 {
5994 	ixgbe_clean_tx_ring(tx_ring);
5995 
5996 	vfree(tx_ring->tx_buffer_info);
5997 	tx_ring->tx_buffer_info = NULL;
5998 
5999 	/* if not set, then don't free */
6000 	if (!tx_ring->desc)
6001 		return;
6002 
6003 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6004 			  tx_ring->desc, tx_ring->dma);
6005 
6006 	tx_ring->desc = NULL;
6007 }
6008 
6009 /**
6010  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6011  * @adapter: board private structure
6012  *
6013  * Free all transmit software resources
6014  **/
6015 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6016 {
6017 	int i;
6018 
6019 	for (i = 0; i < adapter->num_tx_queues; i++)
6020 		if (adapter->tx_ring[i]->desc)
6021 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6022 }
6023 
6024 /**
6025  * ixgbe_free_rx_resources - Free Rx Resources
6026  * @rx_ring: ring to clean the resources from
6027  *
6028  * Free all receive software resources
6029  **/
6030 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6031 {
6032 	ixgbe_clean_rx_ring(rx_ring);
6033 
6034 	vfree(rx_ring->rx_buffer_info);
6035 	rx_ring->rx_buffer_info = NULL;
6036 
6037 	/* if not set, then don't free */
6038 	if (!rx_ring->desc)
6039 		return;
6040 
6041 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6042 			  rx_ring->desc, rx_ring->dma);
6043 
6044 	rx_ring->desc = NULL;
6045 }
6046 
6047 /**
6048  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6049  * @adapter: board private structure
6050  *
6051  * Free all receive software resources
6052  **/
6053 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6054 {
6055 	int i;
6056 
6057 #ifdef IXGBE_FCOE
6058 	ixgbe_free_fcoe_ddp_resources(adapter);
6059 
6060 #endif
6061 	for (i = 0; i < adapter->num_rx_queues; i++)
6062 		if (adapter->rx_ring[i]->desc)
6063 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6064 }
6065 
6066 /**
6067  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6068  * @netdev: network interface device structure
6069  * @new_mtu: new value for maximum frame size
6070  *
6071  * Returns 0 on success, negative on failure
6072  **/
6073 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6074 {
6075 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6076 
6077 	/*
6078 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6079 	 * paths when MTU greater than 1500 is configured.  So display a
6080 	 * warning that legacy VFs will be disabled.
6081 	 */
6082 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6083 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6084 	    (new_mtu > ETH_DATA_LEN))
6085 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6086 
6087 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6088 
6089 	/* must set new MTU before calling down or up */
6090 	netdev->mtu = new_mtu;
6091 
6092 	if (netif_running(netdev))
6093 		ixgbe_reinit_locked(adapter);
6094 
6095 	return 0;
6096 }
6097 
6098 /**
6099  * ixgbe_open - Called when a network interface is made active
6100  * @netdev: network interface device structure
6101  *
6102  * Returns 0 on success, negative value on failure
6103  *
6104  * The open entry point is called when a network interface is made
6105  * active by the system (IFF_UP).  At this point all resources needed
6106  * for transmit and receive operations are allocated, the interrupt
6107  * handler is registered with the OS, the watchdog timer is started,
6108  * and the stack is notified that the interface is ready.
6109  **/
6110 int ixgbe_open(struct net_device *netdev)
6111 {
6112 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6113 	struct ixgbe_hw *hw = &adapter->hw;
6114 	int err, queues;
6115 
6116 	/* disallow open during test */
6117 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6118 		return -EBUSY;
6119 
6120 	netif_carrier_off(netdev);
6121 
6122 	/* allocate transmit descriptors */
6123 	err = ixgbe_setup_all_tx_resources(adapter);
6124 	if (err)
6125 		goto err_setup_tx;
6126 
6127 	/* allocate receive descriptors */
6128 	err = ixgbe_setup_all_rx_resources(adapter);
6129 	if (err)
6130 		goto err_setup_rx;
6131 
6132 	ixgbe_configure(adapter);
6133 
6134 	err = ixgbe_request_irq(adapter);
6135 	if (err)
6136 		goto err_req_irq;
6137 
6138 	/* Notify the stack of the actual queue counts. */
6139 	if (adapter->num_rx_pools > 1)
6140 		queues = adapter->num_rx_queues_per_pool;
6141 	else
6142 		queues = adapter->num_tx_queues;
6143 
6144 	err = netif_set_real_num_tx_queues(netdev, queues);
6145 	if (err)
6146 		goto err_set_queues;
6147 
6148 	if (adapter->num_rx_pools > 1 &&
6149 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
6150 		queues = IXGBE_MAX_L2A_QUEUES;
6151 	else
6152 		queues = adapter->num_rx_queues;
6153 	err = netif_set_real_num_rx_queues(netdev, queues);
6154 	if (err)
6155 		goto err_set_queues;
6156 
6157 	ixgbe_ptp_init(adapter);
6158 
6159 	ixgbe_up_complete(adapter);
6160 
6161 	ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6162 	udp_tunnel_get_rx_info(netdev);
6163 
6164 	return 0;
6165 
6166 err_set_queues:
6167 	ixgbe_free_irq(adapter);
6168 err_req_irq:
6169 	ixgbe_free_all_rx_resources(adapter);
6170 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6171 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6172 err_setup_rx:
6173 	ixgbe_free_all_tx_resources(adapter);
6174 err_setup_tx:
6175 	ixgbe_reset(adapter);
6176 
6177 	return err;
6178 }
6179 
6180 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6181 {
6182 	ixgbe_ptp_suspend(adapter);
6183 
6184 	if (adapter->hw.phy.ops.enter_lplu) {
6185 		adapter->hw.phy.reset_disable = true;
6186 		ixgbe_down(adapter);
6187 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6188 		adapter->hw.phy.reset_disable = false;
6189 	} else {
6190 		ixgbe_down(adapter);
6191 	}
6192 
6193 	ixgbe_free_irq(adapter);
6194 
6195 	ixgbe_free_all_tx_resources(adapter);
6196 	ixgbe_free_all_rx_resources(adapter);
6197 }
6198 
6199 /**
6200  * ixgbe_close - Disables a network interface
6201  * @netdev: network interface device structure
6202  *
6203  * Returns 0, this is not allowed to fail
6204  *
6205  * The close entry point is called when an interface is de-activated
6206  * by the OS.  The hardware is still under the drivers control, but
6207  * needs to be disabled.  A global MAC reset is issued to stop the
6208  * hardware, and all transmit and receive resources are freed.
6209  **/
6210 int ixgbe_close(struct net_device *netdev)
6211 {
6212 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6213 
6214 	ixgbe_ptp_stop(adapter);
6215 
6216 	if (netif_device_present(netdev))
6217 		ixgbe_close_suspend(adapter);
6218 
6219 	ixgbe_fdir_filter_exit(adapter);
6220 
6221 	ixgbe_release_hw_control(adapter);
6222 
6223 	return 0;
6224 }
6225 
6226 #ifdef CONFIG_PM
6227 static int ixgbe_resume(struct pci_dev *pdev)
6228 {
6229 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6230 	struct net_device *netdev = adapter->netdev;
6231 	u32 err;
6232 
6233 	adapter->hw.hw_addr = adapter->io_addr;
6234 	pci_set_power_state(pdev, PCI_D0);
6235 	pci_restore_state(pdev);
6236 	/*
6237 	 * pci_restore_state clears dev->state_saved so call
6238 	 * pci_save_state to restore it.
6239 	 */
6240 	pci_save_state(pdev);
6241 
6242 	err = pci_enable_device_mem(pdev);
6243 	if (err) {
6244 		e_dev_err("Cannot enable PCI device from suspend\n");
6245 		return err;
6246 	}
6247 	smp_mb__before_atomic();
6248 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6249 	pci_set_master(pdev);
6250 
6251 	pci_wake_from_d3(pdev, false);
6252 
6253 	ixgbe_reset(adapter);
6254 
6255 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6256 
6257 	rtnl_lock();
6258 	err = ixgbe_init_interrupt_scheme(adapter);
6259 	if (!err && netif_running(netdev))
6260 		err = ixgbe_open(netdev);
6261 
6262 
6263 	if (!err)
6264 		netif_device_attach(netdev);
6265 	rtnl_unlock();
6266 
6267 	return err;
6268 }
6269 #endif /* CONFIG_PM */
6270 
6271 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6272 {
6273 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6274 	struct net_device *netdev = adapter->netdev;
6275 	struct ixgbe_hw *hw = &adapter->hw;
6276 	u32 ctrl, fctrl;
6277 	u32 wufc = adapter->wol;
6278 #ifdef CONFIG_PM
6279 	int retval = 0;
6280 #endif
6281 
6282 	rtnl_lock();
6283 	netif_device_detach(netdev);
6284 
6285 	if (netif_running(netdev))
6286 		ixgbe_close_suspend(adapter);
6287 
6288 	ixgbe_clear_interrupt_scheme(adapter);
6289 	rtnl_unlock();
6290 
6291 #ifdef CONFIG_PM
6292 	retval = pci_save_state(pdev);
6293 	if (retval)
6294 		return retval;
6295 
6296 #endif
6297 	if (hw->mac.ops.stop_link_on_d3)
6298 		hw->mac.ops.stop_link_on_d3(hw);
6299 
6300 	if (wufc) {
6301 		ixgbe_set_rx_mode(netdev);
6302 
6303 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6304 		if (hw->mac.ops.enable_tx_laser)
6305 			hw->mac.ops.enable_tx_laser(hw);
6306 
6307 		/* turn on all-multi mode if wake on multicast is enabled */
6308 		if (wufc & IXGBE_WUFC_MC) {
6309 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6310 			fctrl |= IXGBE_FCTRL_MPE;
6311 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6312 		}
6313 
6314 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6315 		ctrl |= IXGBE_CTRL_GIO_DIS;
6316 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6317 
6318 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6319 	} else {
6320 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6321 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6322 	}
6323 
6324 	switch (hw->mac.type) {
6325 	case ixgbe_mac_82598EB:
6326 		pci_wake_from_d3(pdev, false);
6327 		break;
6328 	case ixgbe_mac_82599EB:
6329 	case ixgbe_mac_X540:
6330 	case ixgbe_mac_X550:
6331 	case ixgbe_mac_X550EM_x:
6332 	case ixgbe_mac_x550em_a:
6333 		pci_wake_from_d3(pdev, !!wufc);
6334 		break;
6335 	default:
6336 		break;
6337 	}
6338 
6339 	*enable_wake = !!wufc;
6340 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6341 		hw->phy.ops.set_phy_power(hw, false);
6342 
6343 	ixgbe_release_hw_control(adapter);
6344 
6345 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6346 		pci_disable_device(pdev);
6347 
6348 	return 0;
6349 }
6350 
6351 #ifdef CONFIG_PM
6352 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6353 {
6354 	int retval;
6355 	bool wake;
6356 
6357 	retval = __ixgbe_shutdown(pdev, &wake);
6358 	if (retval)
6359 		return retval;
6360 
6361 	if (wake) {
6362 		pci_prepare_to_sleep(pdev);
6363 	} else {
6364 		pci_wake_from_d3(pdev, false);
6365 		pci_set_power_state(pdev, PCI_D3hot);
6366 	}
6367 
6368 	return 0;
6369 }
6370 #endif /* CONFIG_PM */
6371 
6372 static void ixgbe_shutdown(struct pci_dev *pdev)
6373 {
6374 	bool wake;
6375 
6376 	__ixgbe_shutdown(pdev, &wake);
6377 
6378 	if (system_state == SYSTEM_POWER_OFF) {
6379 		pci_wake_from_d3(pdev, wake);
6380 		pci_set_power_state(pdev, PCI_D3hot);
6381 	}
6382 }
6383 
6384 /**
6385  * ixgbe_update_stats - Update the board statistics counters.
6386  * @adapter: board private structure
6387  **/
6388 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6389 {
6390 	struct net_device *netdev = adapter->netdev;
6391 	struct ixgbe_hw *hw = &adapter->hw;
6392 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6393 	u64 total_mpc = 0;
6394 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6395 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6396 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6397 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6398 
6399 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6400 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6401 		return;
6402 
6403 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6404 		u64 rsc_count = 0;
6405 		u64 rsc_flush = 0;
6406 		for (i = 0; i < adapter->num_rx_queues; i++) {
6407 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6408 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6409 		}
6410 		adapter->rsc_total_count = rsc_count;
6411 		adapter->rsc_total_flush = rsc_flush;
6412 	}
6413 
6414 	for (i = 0; i < adapter->num_rx_queues; i++) {
6415 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6416 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6417 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6418 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6419 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6420 		bytes += rx_ring->stats.bytes;
6421 		packets += rx_ring->stats.packets;
6422 	}
6423 	adapter->non_eop_descs = non_eop_descs;
6424 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6425 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6426 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6427 	netdev->stats.rx_bytes = bytes;
6428 	netdev->stats.rx_packets = packets;
6429 
6430 	bytes = 0;
6431 	packets = 0;
6432 	/* gather some stats to the adapter struct that are per queue */
6433 	for (i = 0; i < adapter->num_tx_queues; i++) {
6434 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6435 		restart_queue += tx_ring->tx_stats.restart_queue;
6436 		tx_busy += tx_ring->tx_stats.tx_busy;
6437 		bytes += tx_ring->stats.bytes;
6438 		packets += tx_ring->stats.packets;
6439 	}
6440 	adapter->restart_queue = restart_queue;
6441 	adapter->tx_busy = tx_busy;
6442 	netdev->stats.tx_bytes = bytes;
6443 	netdev->stats.tx_packets = packets;
6444 
6445 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6446 
6447 	/* 8 register reads */
6448 	for (i = 0; i < 8; i++) {
6449 		/* for packet buffers not used, the register should read 0 */
6450 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6451 		missed_rx += mpc;
6452 		hwstats->mpc[i] += mpc;
6453 		total_mpc += hwstats->mpc[i];
6454 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6455 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6456 		switch (hw->mac.type) {
6457 		case ixgbe_mac_82598EB:
6458 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6459 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6460 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6461 			hwstats->pxonrxc[i] +=
6462 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6463 			break;
6464 		case ixgbe_mac_82599EB:
6465 		case ixgbe_mac_X540:
6466 		case ixgbe_mac_X550:
6467 		case ixgbe_mac_X550EM_x:
6468 		case ixgbe_mac_x550em_a:
6469 			hwstats->pxonrxc[i] +=
6470 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6471 			break;
6472 		default:
6473 			break;
6474 		}
6475 	}
6476 
6477 	/*16 register reads */
6478 	for (i = 0; i < 16; i++) {
6479 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6480 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6481 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6482 		    (hw->mac.type == ixgbe_mac_X540) ||
6483 		    (hw->mac.type == ixgbe_mac_X550) ||
6484 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6485 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6486 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6487 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6488 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6489 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6490 		}
6491 	}
6492 
6493 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6494 	/* work around hardware counting issue */
6495 	hwstats->gprc -= missed_rx;
6496 
6497 	ixgbe_update_xoff_received(adapter);
6498 
6499 	/* 82598 hardware only has a 32 bit counter in the high register */
6500 	switch (hw->mac.type) {
6501 	case ixgbe_mac_82598EB:
6502 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6503 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6504 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6505 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6506 		break;
6507 	case ixgbe_mac_X540:
6508 	case ixgbe_mac_X550:
6509 	case ixgbe_mac_X550EM_x:
6510 	case ixgbe_mac_x550em_a:
6511 		/* OS2BMC stats are X540 and later */
6512 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6513 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6514 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6515 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6516 	case ixgbe_mac_82599EB:
6517 		for (i = 0; i < 16; i++)
6518 			adapter->hw_rx_no_dma_resources +=
6519 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6520 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6521 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6522 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6523 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6524 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6525 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6526 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6527 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6528 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6529 #ifdef IXGBE_FCOE
6530 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6531 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6532 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6533 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6534 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6535 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6536 		/* Add up per cpu counters for total ddp aloc fail */
6537 		if (adapter->fcoe.ddp_pool) {
6538 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6539 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6540 			unsigned int cpu;
6541 			u64 noddp = 0, noddp_ext_buff = 0;
6542 			for_each_possible_cpu(cpu) {
6543 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6544 				noddp += ddp_pool->noddp;
6545 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6546 			}
6547 			hwstats->fcoe_noddp = noddp;
6548 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6549 		}
6550 #endif /* IXGBE_FCOE */
6551 		break;
6552 	default:
6553 		break;
6554 	}
6555 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6556 	hwstats->bprc += bprc;
6557 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6558 	if (hw->mac.type == ixgbe_mac_82598EB)
6559 		hwstats->mprc -= bprc;
6560 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6561 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6562 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6563 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6564 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6565 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6566 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6567 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6568 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6569 	hwstats->lxontxc += lxon;
6570 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6571 	hwstats->lxofftxc += lxoff;
6572 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6573 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6574 	/*
6575 	 * 82598 errata - tx of flow control packets is included in tx counters
6576 	 */
6577 	xon_off_tot = lxon + lxoff;
6578 	hwstats->gptc -= xon_off_tot;
6579 	hwstats->mptc -= xon_off_tot;
6580 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6581 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6582 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6583 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6584 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6585 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6586 	hwstats->ptc64 -= xon_off_tot;
6587 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6588 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6589 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6590 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6591 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6592 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6593 
6594 	/* Fill out the OS statistics structure */
6595 	netdev->stats.multicast = hwstats->mprc;
6596 
6597 	/* Rx Errors */
6598 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6599 	netdev->stats.rx_dropped = 0;
6600 	netdev->stats.rx_length_errors = hwstats->rlec;
6601 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6602 	netdev->stats.rx_missed_errors = total_mpc;
6603 }
6604 
6605 /**
6606  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6607  * @adapter: pointer to the device adapter structure
6608  **/
6609 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6610 {
6611 	struct ixgbe_hw *hw = &adapter->hw;
6612 	int i;
6613 
6614 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6615 		return;
6616 
6617 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6618 
6619 	/* if interface is down do nothing */
6620 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6621 		return;
6622 
6623 	/* do nothing if we are not using signature filters */
6624 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6625 		return;
6626 
6627 	adapter->fdir_overflow++;
6628 
6629 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6630 		for (i = 0; i < adapter->num_tx_queues; i++)
6631 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6632 				&(adapter->tx_ring[i]->state));
6633 		/* re-enable flow director interrupts */
6634 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6635 	} else {
6636 		e_err(probe, "failed to finish FDIR re-initialization, "
6637 		      "ignored adding FDIR ATR filters\n");
6638 	}
6639 }
6640 
6641 /**
6642  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6643  * @adapter: pointer to the device adapter structure
6644  *
6645  * This function serves two purposes.  First it strobes the interrupt lines
6646  * in order to make certain interrupts are occurring.  Secondly it sets the
6647  * bits needed to check for TX hangs.  As a result we should immediately
6648  * determine if a hang has occurred.
6649  */
6650 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6651 {
6652 	struct ixgbe_hw *hw = &adapter->hw;
6653 	u64 eics = 0;
6654 	int i;
6655 
6656 	/* If we're down, removing or resetting, just bail */
6657 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6658 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6659 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6660 		return;
6661 
6662 	/* Force detection of hung controller */
6663 	if (netif_carrier_ok(adapter->netdev)) {
6664 		for (i = 0; i < adapter->num_tx_queues; i++)
6665 			set_check_for_tx_hang(adapter->tx_ring[i]);
6666 	}
6667 
6668 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6669 		/*
6670 		 * for legacy and MSI interrupts don't set any bits
6671 		 * that are enabled for EIAM, because this operation
6672 		 * would set *both* EIMS and EICS for any bit in EIAM
6673 		 */
6674 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6675 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6676 	} else {
6677 		/* get one bit for every active tx/rx interrupt vector */
6678 		for (i = 0; i < adapter->num_q_vectors; i++) {
6679 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6680 			if (qv->rx.ring || qv->tx.ring)
6681 				eics |= BIT_ULL(i);
6682 		}
6683 	}
6684 
6685 	/* Cause software interrupt to ensure rings are cleaned */
6686 	ixgbe_irq_rearm_queues(adapter, eics);
6687 }
6688 
6689 /**
6690  * ixgbe_watchdog_update_link - update the link status
6691  * @adapter: pointer to the device adapter structure
6692  * @link_speed: pointer to a u32 to store the link_speed
6693  **/
6694 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6695 {
6696 	struct ixgbe_hw *hw = &adapter->hw;
6697 	u32 link_speed = adapter->link_speed;
6698 	bool link_up = adapter->link_up;
6699 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6700 
6701 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6702 		return;
6703 
6704 	if (hw->mac.ops.check_link) {
6705 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6706 	} else {
6707 		/* always assume link is up, if no check link function */
6708 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6709 		link_up = true;
6710 	}
6711 
6712 	if (adapter->ixgbe_ieee_pfc)
6713 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6714 
6715 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6716 		hw->mac.ops.fc_enable(hw);
6717 		ixgbe_set_rx_drop_en(adapter);
6718 	}
6719 
6720 	if (link_up ||
6721 	    time_after(jiffies, (adapter->link_check_timeout +
6722 				 IXGBE_TRY_LINK_TIMEOUT))) {
6723 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6724 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6725 		IXGBE_WRITE_FLUSH(hw);
6726 	}
6727 
6728 	adapter->link_up = link_up;
6729 	adapter->link_speed = link_speed;
6730 }
6731 
6732 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6733 {
6734 #ifdef CONFIG_IXGBE_DCB
6735 	struct net_device *netdev = adapter->netdev;
6736 	struct dcb_app app = {
6737 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6738 			      .protocol = 0,
6739 			     };
6740 	u8 up = 0;
6741 
6742 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6743 		up = dcb_ieee_getapp_mask(netdev, &app);
6744 
6745 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6746 #endif
6747 }
6748 
6749 static int ixgbe_enable_macvlan(struct net_device *upper, void *data)
6750 {
6751 	if (netif_is_macvlan(upper)) {
6752 		struct macvlan_dev *vlan = netdev_priv(upper);
6753 
6754 		if (vlan->fwd_priv)
6755 			netif_tx_wake_all_queues(upper);
6756 	}
6757 
6758 	return 0;
6759 }
6760 
6761 /**
6762  * ixgbe_watchdog_link_is_up - update netif_carrier status and
6763  *                             print link up message
6764  * @adapter: pointer to the device adapter structure
6765  **/
6766 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6767 {
6768 	struct net_device *netdev = adapter->netdev;
6769 	struct ixgbe_hw *hw = &adapter->hw;
6770 	u32 link_speed = adapter->link_speed;
6771 	const char *speed_str;
6772 	bool flow_rx, flow_tx;
6773 
6774 	/* only continue if link was previously down */
6775 	if (netif_carrier_ok(netdev))
6776 		return;
6777 
6778 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6779 
6780 	switch (hw->mac.type) {
6781 	case ixgbe_mac_82598EB: {
6782 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6783 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6784 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6785 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6786 	}
6787 		break;
6788 	case ixgbe_mac_X540:
6789 	case ixgbe_mac_X550:
6790 	case ixgbe_mac_X550EM_x:
6791 	case ixgbe_mac_x550em_a:
6792 	case ixgbe_mac_82599EB: {
6793 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6794 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6795 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6796 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6797 	}
6798 		break;
6799 	default:
6800 		flow_tx = false;
6801 		flow_rx = false;
6802 		break;
6803 	}
6804 
6805 	adapter->last_rx_ptp_check = jiffies;
6806 
6807 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6808 		ixgbe_ptp_start_cyclecounter(adapter);
6809 
6810 	switch (link_speed) {
6811 	case IXGBE_LINK_SPEED_10GB_FULL:
6812 		speed_str = "10 Gbps";
6813 		break;
6814 	case IXGBE_LINK_SPEED_2_5GB_FULL:
6815 		speed_str = "2.5 Gbps";
6816 		break;
6817 	case IXGBE_LINK_SPEED_1GB_FULL:
6818 		speed_str = "1 Gbps";
6819 		break;
6820 	case IXGBE_LINK_SPEED_100_FULL:
6821 		speed_str = "100 Mbps";
6822 		break;
6823 	case IXGBE_LINK_SPEED_10_FULL:
6824 		speed_str = "10 Mbps";
6825 		break;
6826 	default:
6827 		speed_str = "unknown speed";
6828 		break;
6829 	}
6830 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
6831 	       ((flow_rx && flow_tx) ? "RX/TX" :
6832 	       (flow_rx ? "RX" :
6833 	       (flow_tx ? "TX" : "None"))));
6834 
6835 	netif_carrier_on(netdev);
6836 	ixgbe_check_vf_rate_limit(adapter);
6837 
6838 	/* enable transmits */
6839 	netif_tx_wake_all_queues(adapter->netdev);
6840 
6841 	/* enable any upper devices */
6842 	rtnl_lock();
6843 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
6844 				      ixgbe_enable_macvlan, NULL);
6845 	rtnl_unlock();
6846 
6847 	/* update the default user priority for VFs */
6848 	ixgbe_update_default_up(adapter);
6849 
6850 	/* ping all the active vfs to let them know link has changed */
6851 	ixgbe_ping_all_vfs(adapter);
6852 }
6853 
6854 /**
6855  * ixgbe_watchdog_link_is_down - update netif_carrier status and
6856  *                               print link down message
6857  * @adapter: pointer to the adapter structure
6858  **/
6859 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6860 {
6861 	struct net_device *netdev = adapter->netdev;
6862 	struct ixgbe_hw *hw = &adapter->hw;
6863 
6864 	adapter->link_up = false;
6865 	adapter->link_speed = 0;
6866 
6867 	/* only continue if link was up previously */
6868 	if (!netif_carrier_ok(netdev))
6869 		return;
6870 
6871 	/* poll for SFP+ cable when link is down */
6872 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6873 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6874 
6875 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6876 		ixgbe_ptp_start_cyclecounter(adapter);
6877 
6878 	e_info(drv, "NIC Link is Down\n");
6879 	netif_carrier_off(netdev);
6880 
6881 	/* ping all the active vfs to let them know link has changed */
6882 	ixgbe_ping_all_vfs(adapter);
6883 }
6884 
6885 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
6886 {
6887 	int i;
6888 
6889 	for (i = 0; i < adapter->num_tx_queues; i++) {
6890 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6891 
6892 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
6893 			return true;
6894 	}
6895 
6896 	return false;
6897 }
6898 
6899 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
6900 {
6901 	struct ixgbe_hw *hw = &adapter->hw;
6902 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
6903 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
6904 
6905 	int i, j;
6906 
6907 	if (!adapter->num_vfs)
6908 		return false;
6909 
6910 	/* resetting the PF is only needed for MAC before X550 */
6911 	if (hw->mac.type >= ixgbe_mac_X550)
6912 		return false;
6913 
6914 	for (i = 0; i < adapter->num_vfs; i++) {
6915 		for (j = 0; j < q_per_pool; j++) {
6916 			u32 h, t;
6917 
6918 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
6919 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
6920 
6921 			if (h != t)
6922 				return true;
6923 		}
6924 	}
6925 
6926 	return false;
6927 }
6928 
6929 /**
6930  * ixgbe_watchdog_flush_tx - flush queues on link down
6931  * @adapter: pointer to the device adapter structure
6932  **/
6933 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6934 {
6935 	if (!netif_carrier_ok(adapter->netdev)) {
6936 		if (ixgbe_ring_tx_pending(adapter) ||
6937 		    ixgbe_vf_tx_pending(adapter)) {
6938 			/* We've lost link, so the controller stops DMA,
6939 			 * but we've got queued Tx work that's never going
6940 			 * to get done, so reset controller to flush Tx.
6941 			 * (Do the reset outside of interrupt context).
6942 			 */
6943 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6944 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
6945 		}
6946 	}
6947 }
6948 
6949 #ifdef CONFIG_PCI_IOV
6950 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
6951 				      struct pci_dev *vfdev)
6952 {
6953 	if (!pci_wait_for_pending_transaction(vfdev))
6954 		e_dev_warn("Issuing VFLR with pending transactions\n");
6955 
6956 	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
6957 	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
6958 
6959 	msleep(100);
6960 }
6961 
6962 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6963 {
6964 	struct ixgbe_hw *hw = &adapter->hw;
6965 	struct pci_dev *pdev = adapter->pdev;
6966 	unsigned int vf;
6967 	u32 gpc;
6968 
6969 	if (!(netif_carrier_ok(adapter->netdev)))
6970 		return;
6971 
6972 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6973 	if (gpc) /* If incrementing then no need for the check below */
6974 		return;
6975 	/* Check to see if a bad DMA write target from an errant or
6976 	 * malicious VF has caused a PCIe error.  If so then we can
6977 	 * issue a VFLR to the offending VF(s) and then resume without
6978 	 * requesting a full slot reset.
6979 	 */
6980 
6981 	if (!pdev)
6982 		return;
6983 
6984 	/* check status reg for all VFs owned by this PF */
6985 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
6986 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
6987 		u16 status_reg;
6988 
6989 		if (!vfdev)
6990 			continue;
6991 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
6992 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
6993 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
6994 			ixgbe_issue_vf_flr(adapter, vfdev);
6995 	}
6996 }
6997 
6998 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6999 {
7000 	u32 ssvpc;
7001 
7002 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7003 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7004 	    adapter->num_vfs == 0)
7005 		return;
7006 
7007 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7008 
7009 	/*
7010 	 * ssvpc register is cleared on read, if zero then no
7011 	 * spoofed packets in the last interval.
7012 	 */
7013 	if (!ssvpc)
7014 		return;
7015 
7016 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7017 }
7018 #else
7019 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7020 {
7021 }
7022 
7023 static void
7024 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7025 {
7026 }
7027 #endif /* CONFIG_PCI_IOV */
7028 
7029 
7030 /**
7031  * ixgbe_watchdog_subtask - check and bring link up
7032  * @adapter: pointer to the device adapter structure
7033  **/
7034 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7035 {
7036 	/* if interface is down, removing or resetting, do nothing */
7037 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7038 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7039 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7040 		return;
7041 
7042 	ixgbe_watchdog_update_link(adapter);
7043 
7044 	if (adapter->link_up)
7045 		ixgbe_watchdog_link_is_up(adapter);
7046 	else
7047 		ixgbe_watchdog_link_is_down(adapter);
7048 
7049 	ixgbe_check_for_bad_vf(adapter);
7050 	ixgbe_spoof_check(adapter);
7051 	ixgbe_update_stats(adapter);
7052 
7053 	ixgbe_watchdog_flush_tx(adapter);
7054 }
7055 
7056 /**
7057  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7058  * @adapter: the ixgbe adapter structure
7059  **/
7060 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7061 {
7062 	struct ixgbe_hw *hw = &adapter->hw;
7063 	s32 err;
7064 
7065 	/* not searching for SFP so there is nothing to do here */
7066 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7067 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7068 		return;
7069 
7070 	if (adapter->sfp_poll_time &&
7071 	    time_after(adapter->sfp_poll_time, jiffies))
7072 		return; /* If not yet time to poll for SFP */
7073 
7074 	/* someone else is in init, wait until next service event */
7075 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7076 		return;
7077 
7078 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7079 
7080 	err = hw->phy.ops.identify_sfp(hw);
7081 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7082 		goto sfp_out;
7083 
7084 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7085 		/* If no cable is present, then we need to reset
7086 		 * the next time we find a good cable. */
7087 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7088 	}
7089 
7090 	/* exit on error */
7091 	if (err)
7092 		goto sfp_out;
7093 
7094 	/* exit if reset not needed */
7095 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7096 		goto sfp_out;
7097 
7098 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7099 
7100 	/*
7101 	 * A module may be identified correctly, but the EEPROM may not have
7102 	 * support for that module.  setup_sfp() will fail in that case, so
7103 	 * we should not allow that module to load.
7104 	 */
7105 	if (hw->mac.type == ixgbe_mac_82598EB)
7106 		err = hw->phy.ops.reset(hw);
7107 	else
7108 		err = hw->mac.ops.setup_sfp(hw);
7109 
7110 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7111 		goto sfp_out;
7112 
7113 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7114 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7115 
7116 sfp_out:
7117 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7118 
7119 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7120 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7121 		e_dev_err("failed to initialize because an unsupported "
7122 			  "SFP+ module type was detected.\n");
7123 		e_dev_err("Reload the driver after installing a "
7124 			  "supported module.\n");
7125 		unregister_netdev(adapter->netdev);
7126 	}
7127 }
7128 
7129 /**
7130  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7131  * @adapter: the ixgbe adapter structure
7132  **/
7133 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7134 {
7135 	struct ixgbe_hw *hw = &adapter->hw;
7136 	u32 speed;
7137 	bool autoneg = false;
7138 
7139 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7140 		return;
7141 
7142 	/* someone else is in init, wait until next service event */
7143 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7144 		return;
7145 
7146 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7147 
7148 	speed = hw->phy.autoneg_advertised;
7149 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
7150 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
7151 
7152 		/* setup the highest link when no autoneg */
7153 		if (!autoneg) {
7154 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
7155 				speed = IXGBE_LINK_SPEED_10GB_FULL;
7156 		}
7157 	}
7158 
7159 	if (hw->mac.ops.setup_link)
7160 		hw->mac.ops.setup_link(hw, speed, true);
7161 
7162 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7163 	adapter->link_check_timeout = jiffies;
7164 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7165 }
7166 
7167 /**
7168  * ixgbe_service_timer - Timer Call-back
7169  * @data: pointer to adapter cast into an unsigned long
7170  **/
7171 static void ixgbe_service_timer(unsigned long data)
7172 {
7173 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
7174 	unsigned long next_event_offset;
7175 
7176 	/* poll faster when waiting for link */
7177 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7178 		next_event_offset = HZ / 10;
7179 	else
7180 		next_event_offset = HZ * 2;
7181 
7182 	/* Reset the timer */
7183 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7184 
7185 	ixgbe_service_event_schedule(adapter);
7186 }
7187 
7188 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7189 {
7190 	struct ixgbe_hw *hw = &adapter->hw;
7191 	u32 status;
7192 
7193 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7194 		return;
7195 
7196 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7197 
7198 	if (!hw->phy.ops.handle_lasi)
7199 		return;
7200 
7201 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7202 	if (status != IXGBE_ERR_OVERTEMP)
7203 		return;
7204 
7205 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7206 }
7207 
7208 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7209 {
7210 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7211 		return;
7212 
7213 	/* If we're already down, removing or resetting, just bail */
7214 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7215 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7216 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7217 		return;
7218 
7219 	ixgbe_dump(adapter);
7220 	netdev_err(adapter->netdev, "Reset adapter\n");
7221 	adapter->tx_timeout_count++;
7222 
7223 	rtnl_lock();
7224 	ixgbe_reinit_locked(adapter);
7225 	rtnl_unlock();
7226 }
7227 
7228 /**
7229  * ixgbe_service_task - manages and runs subtasks
7230  * @work: pointer to work_struct containing our data
7231  **/
7232 static void ixgbe_service_task(struct work_struct *work)
7233 {
7234 	struct ixgbe_adapter *adapter = container_of(work,
7235 						     struct ixgbe_adapter,
7236 						     service_task);
7237 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7238 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7239 			rtnl_lock();
7240 			ixgbe_down(adapter);
7241 			rtnl_unlock();
7242 		}
7243 		ixgbe_service_event_complete(adapter);
7244 		return;
7245 	}
7246 	if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7247 		rtnl_lock();
7248 		adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7249 		udp_tunnel_get_rx_info(adapter->netdev);
7250 		rtnl_unlock();
7251 	}
7252 	ixgbe_reset_subtask(adapter);
7253 	ixgbe_phy_interrupt_subtask(adapter);
7254 	ixgbe_sfp_detection_subtask(adapter);
7255 	ixgbe_sfp_link_config_subtask(adapter);
7256 	ixgbe_check_overtemp_subtask(adapter);
7257 	ixgbe_watchdog_subtask(adapter);
7258 	ixgbe_fdir_reinit_subtask(adapter);
7259 	ixgbe_check_hang_subtask(adapter);
7260 
7261 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7262 		ixgbe_ptp_overflow_check(adapter);
7263 		ixgbe_ptp_rx_hang(adapter);
7264 	}
7265 
7266 	ixgbe_service_event_complete(adapter);
7267 }
7268 
7269 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7270 		     struct ixgbe_tx_buffer *first,
7271 		     u8 *hdr_len)
7272 {
7273 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7274 	struct sk_buff *skb = first->skb;
7275 	union {
7276 		struct iphdr *v4;
7277 		struct ipv6hdr *v6;
7278 		unsigned char *hdr;
7279 	} ip;
7280 	union {
7281 		struct tcphdr *tcp;
7282 		unsigned char *hdr;
7283 	} l4;
7284 	u32 paylen, l4_offset;
7285 	int err;
7286 
7287 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7288 		return 0;
7289 
7290 	if (!skb_is_gso(skb))
7291 		return 0;
7292 
7293 	err = skb_cow_head(skb, 0);
7294 	if (err < 0)
7295 		return err;
7296 
7297 	ip.hdr = skb_network_header(skb);
7298 	l4.hdr = skb_checksum_start(skb);
7299 
7300 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7301 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7302 
7303 	/* initialize outer IP header fields */
7304 	if (ip.v4->version == 4) {
7305 		unsigned char *csum_start = skb_checksum_start(skb);
7306 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
7307 
7308 		/* IP header will have to cancel out any data that
7309 		 * is not a part of the outer IP header
7310 		 */
7311 		ip.v4->check = csum_fold(csum_partial(trans_start,
7312 						      csum_start - trans_start,
7313 						      0));
7314 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7315 
7316 		ip.v4->tot_len = 0;
7317 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7318 				   IXGBE_TX_FLAGS_CSUM |
7319 				   IXGBE_TX_FLAGS_IPV4;
7320 	} else {
7321 		ip.v6->payload_len = 0;
7322 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7323 				   IXGBE_TX_FLAGS_CSUM;
7324 	}
7325 
7326 	/* determine offset of inner transport header */
7327 	l4_offset = l4.hdr - skb->data;
7328 
7329 	/* compute length of segmentation header */
7330 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7331 
7332 	/* remove payload length from inner checksum */
7333 	paylen = skb->len - l4_offset;
7334 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
7335 
7336 	/* update gso size and bytecount with header size */
7337 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7338 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7339 
7340 	/* mss_l4len_id: use 0 as index for TSO */
7341 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7342 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7343 
7344 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7345 	vlan_macip_lens = l4.hdr - ip.hdr;
7346 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7347 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7348 
7349 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
7350 			  mss_l4len_idx);
7351 
7352 	return 1;
7353 }
7354 
7355 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7356 {
7357 	unsigned int offset = 0;
7358 
7359 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7360 
7361 	return offset == skb_checksum_start_offset(skb);
7362 }
7363 
7364 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7365 			  struct ixgbe_tx_buffer *first)
7366 {
7367 	struct sk_buff *skb = first->skb;
7368 	u32 vlan_macip_lens = 0;
7369 	u32 type_tucmd = 0;
7370 
7371 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7372 csum_failed:
7373 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7374 					 IXGBE_TX_FLAGS_CC)))
7375 			return;
7376 		goto no_csum;
7377 	}
7378 
7379 	switch (skb->csum_offset) {
7380 	case offsetof(struct tcphdr, check):
7381 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7382 		/* fall through */
7383 	case offsetof(struct udphdr, check):
7384 		break;
7385 	case offsetof(struct sctphdr, checksum):
7386 		/* validate that this is actually an SCTP request */
7387 		if (((first->protocol == htons(ETH_P_IP)) &&
7388 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7389 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7390 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7391 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7392 			break;
7393 		}
7394 		/* fall through */
7395 	default:
7396 		skb_checksum_help(skb);
7397 		goto csum_failed;
7398 	}
7399 
7400 	/* update TX checksum flag */
7401 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7402 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7403 			  skb_network_offset(skb);
7404 no_csum:
7405 	/* vlan_macip_lens: MACLEN, VLAN tag */
7406 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7407 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7408 
7409 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd, 0);
7410 }
7411 
7412 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7413 	((_flag <= _result) ? \
7414 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7415 	 ((u32)(_input & _flag) / (_flag / _result)))
7416 
7417 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7418 {
7419 	/* set type for advanced descriptor with frame checksum insertion */
7420 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7421 		       IXGBE_ADVTXD_DCMD_DEXT |
7422 		       IXGBE_ADVTXD_DCMD_IFCS;
7423 
7424 	/* set HW vlan bit if vlan is present */
7425 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7426 				   IXGBE_ADVTXD_DCMD_VLE);
7427 
7428 	/* set segmentation enable bits for TSO/FSO */
7429 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7430 				   IXGBE_ADVTXD_DCMD_TSE);
7431 
7432 	/* set timestamp bit if present */
7433 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7434 				   IXGBE_ADVTXD_MAC_TSTAMP);
7435 
7436 	/* insert frame checksum */
7437 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7438 
7439 	return cmd_type;
7440 }
7441 
7442 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7443 				   u32 tx_flags, unsigned int paylen)
7444 {
7445 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7446 
7447 	/* enable L4 checksum for TSO and TX checksum offload */
7448 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7449 					IXGBE_TX_FLAGS_CSUM,
7450 					IXGBE_ADVTXD_POPTS_TXSM);
7451 
7452 	/* enble IPv4 checksum for TSO */
7453 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7454 					IXGBE_TX_FLAGS_IPV4,
7455 					IXGBE_ADVTXD_POPTS_IXSM);
7456 
7457 	/*
7458 	 * Check Context must be set if Tx switch is enabled, which it
7459 	 * always is for case where virtual functions are running
7460 	 */
7461 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7462 					IXGBE_TX_FLAGS_CC,
7463 					IXGBE_ADVTXD_CC);
7464 
7465 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7466 }
7467 
7468 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7469 {
7470 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7471 
7472 	/* Herbert's original patch had:
7473 	 *  smp_mb__after_netif_stop_queue();
7474 	 * but since that doesn't exist yet, just open code it.
7475 	 */
7476 	smp_mb();
7477 
7478 	/* We need to check again in a case another CPU has just
7479 	 * made room available.
7480 	 */
7481 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7482 		return -EBUSY;
7483 
7484 	/* A reprieve! - use start_queue because it doesn't call schedule */
7485 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7486 	++tx_ring->tx_stats.restart_queue;
7487 	return 0;
7488 }
7489 
7490 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7491 {
7492 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7493 		return 0;
7494 
7495 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7496 }
7497 
7498 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7499 		       IXGBE_TXD_CMD_RS)
7500 
7501 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7502 			 struct ixgbe_tx_buffer *first,
7503 			 const u8 hdr_len)
7504 {
7505 	struct sk_buff *skb = first->skb;
7506 	struct ixgbe_tx_buffer *tx_buffer;
7507 	union ixgbe_adv_tx_desc *tx_desc;
7508 	struct skb_frag_struct *frag;
7509 	dma_addr_t dma;
7510 	unsigned int data_len, size;
7511 	u32 tx_flags = first->tx_flags;
7512 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7513 	u16 i = tx_ring->next_to_use;
7514 
7515 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7516 
7517 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7518 
7519 	size = skb_headlen(skb);
7520 	data_len = skb->data_len;
7521 
7522 #ifdef IXGBE_FCOE
7523 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7524 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7525 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7526 			data_len = 0;
7527 		} else {
7528 			data_len -= sizeof(struct fcoe_crc_eof);
7529 		}
7530 	}
7531 
7532 #endif
7533 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7534 
7535 	tx_buffer = first;
7536 
7537 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7538 		if (dma_mapping_error(tx_ring->dev, dma))
7539 			goto dma_error;
7540 
7541 		/* record length, and DMA address */
7542 		dma_unmap_len_set(tx_buffer, len, size);
7543 		dma_unmap_addr_set(tx_buffer, dma, dma);
7544 
7545 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7546 
7547 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7548 			tx_desc->read.cmd_type_len =
7549 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7550 
7551 			i++;
7552 			tx_desc++;
7553 			if (i == tx_ring->count) {
7554 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7555 				i = 0;
7556 			}
7557 			tx_desc->read.olinfo_status = 0;
7558 
7559 			dma += IXGBE_MAX_DATA_PER_TXD;
7560 			size -= IXGBE_MAX_DATA_PER_TXD;
7561 
7562 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7563 		}
7564 
7565 		if (likely(!data_len))
7566 			break;
7567 
7568 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7569 
7570 		i++;
7571 		tx_desc++;
7572 		if (i == tx_ring->count) {
7573 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7574 			i = 0;
7575 		}
7576 		tx_desc->read.olinfo_status = 0;
7577 
7578 #ifdef IXGBE_FCOE
7579 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7580 #else
7581 		size = skb_frag_size(frag);
7582 #endif
7583 		data_len -= size;
7584 
7585 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7586 				       DMA_TO_DEVICE);
7587 
7588 		tx_buffer = &tx_ring->tx_buffer_info[i];
7589 	}
7590 
7591 	/* write last descriptor with RS and EOP bits */
7592 	cmd_type |= size | IXGBE_TXD_CMD;
7593 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7594 
7595 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7596 
7597 	/* set the timestamp */
7598 	first->time_stamp = jiffies;
7599 
7600 	/*
7601 	 * Force memory writes to complete before letting h/w know there
7602 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7603 	 * memory model archs, such as IA-64).
7604 	 *
7605 	 * We also need this memory barrier to make certain all of the
7606 	 * status bits have been updated before next_to_watch is written.
7607 	 */
7608 	wmb();
7609 
7610 	/* set next_to_watch value indicating a packet is present */
7611 	first->next_to_watch = tx_desc;
7612 
7613 	i++;
7614 	if (i == tx_ring->count)
7615 		i = 0;
7616 
7617 	tx_ring->next_to_use = i;
7618 
7619 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7620 
7621 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7622 		writel(i, tx_ring->tail);
7623 
7624 		/* we need this if more than one processor can write to our tail
7625 		 * at a time, it synchronizes IO on IA64/Altix systems
7626 		 */
7627 		mmiowb();
7628 	}
7629 
7630 	return;
7631 dma_error:
7632 	dev_err(tx_ring->dev, "TX DMA map failed\n");
7633 
7634 	/* clear dma mappings for failed tx_buffer_info map */
7635 	for (;;) {
7636 		tx_buffer = &tx_ring->tx_buffer_info[i];
7637 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
7638 		if (tx_buffer == first)
7639 			break;
7640 		if (i == 0)
7641 			i = tx_ring->count;
7642 		i--;
7643 	}
7644 
7645 	tx_ring->next_to_use = i;
7646 }
7647 
7648 static void ixgbe_atr(struct ixgbe_ring *ring,
7649 		      struct ixgbe_tx_buffer *first)
7650 {
7651 	struct ixgbe_q_vector *q_vector = ring->q_vector;
7652 	union ixgbe_atr_hash_dword input = { .dword = 0 };
7653 	union ixgbe_atr_hash_dword common = { .dword = 0 };
7654 	union {
7655 		unsigned char *network;
7656 		struct iphdr *ipv4;
7657 		struct ipv6hdr *ipv6;
7658 	} hdr;
7659 	struct tcphdr *th;
7660 	unsigned int hlen;
7661 	struct sk_buff *skb;
7662 	__be16 vlan_id;
7663 	int l4_proto;
7664 
7665 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7666 	if (!q_vector)
7667 		return;
7668 
7669 	/* do nothing if sampling is disabled */
7670 	if (!ring->atr_sample_rate)
7671 		return;
7672 
7673 	ring->atr_count++;
7674 
7675 	/* currently only IPv4/IPv6 with TCP is supported */
7676 	if ((first->protocol != htons(ETH_P_IP)) &&
7677 	    (first->protocol != htons(ETH_P_IPV6)))
7678 		return;
7679 
7680 	/* snag network header to get L4 type and address */
7681 	skb = first->skb;
7682 	hdr.network = skb_network_header(skb);
7683 	if (unlikely(hdr.network <= skb->data))
7684 		return;
7685 	if (skb->encapsulation &&
7686 	    first->protocol == htons(ETH_P_IP) &&
7687 	    hdr.ipv4->protocol == IPPROTO_UDP) {
7688 		struct ixgbe_adapter *adapter = q_vector->adapter;
7689 
7690 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
7691 			     VXLAN_HEADROOM))
7692 			return;
7693 
7694 		/* verify the port is recognized as VXLAN */
7695 		if (adapter->vxlan_port &&
7696 		    udp_hdr(skb)->dest == adapter->vxlan_port)
7697 			hdr.network = skb_inner_network_header(skb);
7698 
7699 		if (adapter->geneve_port &&
7700 		    udp_hdr(skb)->dest == adapter->geneve_port)
7701 			hdr.network = skb_inner_network_header(skb);
7702 	}
7703 
7704 	/* Make sure we have at least [minimum IPv4 header + TCP]
7705 	 * or [IPv6 header] bytes
7706 	 */
7707 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
7708 		return;
7709 
7710 	/* Currently only IPv4/IPv6 with TCP is supported */
7711 	switch (hdr.ipv4->version) {
7712 	case IPVERSION:
7713 		/* access ihl as u8 to avoid unaligned access on ia64 */
7714 		hlen = (hdr.network[0] & 0x0F) << 2;
7715 		l4_proto = hdr.ipv4->protocol;
7716 		break;
7717 	case 6:
7718 		hlen = hdr.network - skb->data;
7719 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
7720 		hlen -= hdr.network - skb->data;
7721 		break;
7722 	default:
7723 		return;
7724 	}
7725 
7726 	if (l4_proto != IPPROTO_TCP)
7727 		return;
7728 
7729 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
7730 		     hlen + sizeof(struct tcphdr)))
7731 		return;
7732 
7733 	th = (struct tcphdr *)(hdr.network + hlen);
7734 
7735 	/* skip this packet since the socket is closing */
7736 	if (th->fin)
7737 		return;
7738 
7739 	/* sample on all syn packets or once every atr sample count */
7740 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7741 		return;
7742 
7743 	/* reset sample count */
7744 	ring->atr_count = 0;
7745 
7746 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7747 
7748 	/*
7749 	 * src and dst are inverted, think how the receiver sees them
7750 	 *
7751 	 * The input is broken into two sections, a non-compressed section
7752 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7753 	 * is XORed together and stored in the compressed dword.
7754 	 */
7755 	input.formatted.vlan_id = vlan_id;
7756 
7757 	/*
7758 	 * since src port and flex bytes occupy the same word XOR them together
7759 	 * and write the value to source port portion of compressed dword
7760 	 */
7761 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7762 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7763 	else
7764 		common.port.src ^= th->dest ^ first->protocol;
7765 	common.port.dst ^= th->source;
7766 
7767 	switch (hdr.ipv4->version) {
7768 	case IPVERSION:
7769 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7770 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7771 		break;
7772 	case 6:
7773 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7774 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7775 			     hdr.ipv6->saddr.s6_addr32[1] ^
7776 			     hdr.ipv6->saddr.s6_addr32[2] ^
7777 			     hdr.ipv6->saddr.s6_addr32[3] ^
7778 			     hdr.ipv6->daddr.s6_addr32[0] ^
7779 			     hdr.ipv6->daddr.s6_addr32[1] ^
7780 			     hdr.ipv6->daddr.s6_addr32[2] ^
7781 			     hdr.ipv6->daddr.s6_addr32[3];
7782 		break;
7783 	default:
7784 		break;
7785 	}
7786 
7787 	if (hdr.network != skb_network_header(skb))
7788 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
7789 
7790 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7791 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7792 					      input, common, ring->queue_index);
7793 }
7794 
7795 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7796 			      void *accel_priv, select_queue_fallback_t fallback)
7797 {
7798 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7799 #ifdef IXGBE_FCOE
7800 	struct ixgbe_adapter *adapter;
7801 	struct ixgbe_ring_feature *f;
7802 	int txq;
7803 #endif
7804 
7805 	if (fwd_adapter)
7806 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7807 
7808 #ifdef IXGBE_FCOE
7809 
7810 	/*
7811 	 * only execute the code below if protocol is FCoE
7812 	 * or FIP and we have FCoE enabled on the adapter
7813 	 */
7814 	switch (vlan_get_protocol(skb)) {
7815 	case htons(ETH_P_FCOE):
7816 	case htons(ETH_P_FIP):
7817 		adapter = netdev_priv(dev);
7818 
7819 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7820 			break;
7821 	default:
7822 		return fallback(dev, skb);
7823 	}
7824 
7825 	f = &adapter->ring_feature[RING_F_FCOE];
7826 
7827 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7828 					   smp_processor_id();
7829 
7830 	while (txq >= f->indices)
7831 		txq -= f->indices;
7832 
7833 	return txq + f->offset;
7834 #else
7835 	return fallback(dev, skb);
7836 #endif
7837 }
7838 
7839 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7840 			  struct ixgbe_adapter *adapter,
7841 			  struct ixgbe_ring *tx_ring)
7842 {
7843 	struct ixgbe_tx_buffer *first;
7844 	int tso;
7845 	u32 tx_flags = 0;
7846 	unsigned short f;
7847 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7848 	__be16 protocol = skb->protocol;
7849 	u8 hdr_len = 0;
7850 
7851 	/*
7852 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7853 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7854 	 *       + 2 desc gap to keep tail from touching head,
7855 	 *       + 1 desc for context descriptor,
7856 	 * otherwise try next time
7857 	 */
7858 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7859 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7860 
7861 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7862 		tx_ring->tx_stats.tx_busy++;
7863 		return NETDEV_TX_BUSY;
7864 	}
7865 
7866 	/* record the location of the first descriptor for this packet */
7867 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7868 	first->skb = skb;
7869 	first->bytecount = skb->len;
7870 	first->gso_segs = 1;
7871 
7872 	/* if we have a HW VLAN tag being added default to the HW one */
7873 	if (skb_vlan_tag_present(skb)) {
7874 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7875 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7876 	/* else if it is a SW VLAN check the next protocol and store the tag */
7877 	} else if (protocol == htons(ETH_P_8021Q)) {
7878 		struct vlan_hdr *vhdr, _vhdr;
7879 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7880 		if (!vhdr)
7881 			goto out_drop;
7882 
7883 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7884 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7885 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7886 	}
7887 	protocol = vlan_get_protocol(skb);
7888 
7889 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
7890 	    adapter->ptp_clock &&
7891 	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7892 				   &adapter->state)) {
7893 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7894 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7895 
7896 		/* schedule check for Tx timestamp */
7897 		adapter->ptp_tx_skb = skb_get(skb);
7898 		adapter->ptp_tx_start = jiffies;
7899 		schedule_work(&adapter->ptp_tx_work);
7900 	}
7901 
7902 	skb_tx_timestamp(skb);
7903 
7904 #ifdef CONFIG_PCI_IOV
7905 	/*
7906 	 * Use the l2switch_enable flag - would be false if the DMA
7907 	 * Tx switch had been disabled.
7908 	 */
7909 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7910 		tx_flags |= IXGBE_TX_FLAGS_CC;
7911 
7912 #endif
7913 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7914 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7915 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7916 	     (skb->priority != TC_PRIO_CONTROL))) {
7917 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7918 		tx_flags |= (skb->priority & 0x7) <<
7919 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7920 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7921 			struct vlan_ethhdr *vhdr;
7922 
7923 			if (skb_cow_head(skb, 0))
7924 				goto out_drop;
7925 			vhdr = (struct vlan_ethhdr *)skb->data;
7926 			vhdr->h_vlan_TCI = htons(tx_flags >>
7927 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7928 		} else {
7929 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7930 		}
7931 	}
7932 
7933 	/* record initial flags and protocol */
7934 	first->tx_flags = tx_flags;
7935 	first->protocol = protocol;
7936 
7937 #ifdef IXGBE_FCOE
7938 	/* setup tx offload for FCoE */
7939 	if ((protocol == htons(ETH_P_FCOE)) &&
7940 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7941 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7942 		if (tso < 0)
7943 			goto out_drop;
7944 
7945 		goto xmit_fcoe;
7946 	}
7947 
7948 #endif /* IXGBE_FCOE */
7949 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7950 	if (tso < 0)
7951 		goto out_drop;
7952 	else if (!tso)
7953 		ixgbe_tx_csum(tx_ring, first);
7954 
7955 	/* add the ATR filter if ATR is on */
7956 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7957 		ixgbe_atr(tx_ring, first);
7958 
7959 #ifdef IXGBE_FCOE
7960 xmit_fcoe:
7961 #endif /* IXGBE_FCOE */
7962 	ixgbe_tx_map(tx_ring, first, hdr_len);
7963 
7964 	return NETDEV_TX_OK;
7965 
7966 out_drop:
7967 	dev_kfree_skb_any(first->skb);
7968 	first->skb = NULL;
7969 
7970 	return NETDEV_TX_OK;
7971 }
7972 
7973 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7974 				      struct net_device *netdev,
7975 				      struct ixgbe_ring *ring)
7976 {
7977 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7978 	struct ixgbe_ring *tx_ring;
7979 
7980 	/*
7981 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7982 	 * in order to meet this minimum size requirement.
7983 	 */
7984 	if (skb_put_padto(skb, 17))
7985 		return NETDEV_TX_OK;
7986 
7987 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7988 
7989 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7990 }
7991 
7992 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7993 				    struct net_device *netdev)
7994 {
7995 	return __ixgbe_xmit_frame(skb, netdev, NULL);
7996 }
7997 
7998 /**
7999  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8000  * @netdev: network interface device structure
8001  * @p: pointer to an address structure
8002  *
8003  * Returns 0 on success, negative on failure
8004  **/
8005 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8006 {
8007 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8008 	struct ixgbe_hw *hw = &adapter->hw;
8009 	struct sockaddr *addr = p;
8010 
8011 	if (!is_valid_ether_addr(addr->sa_data))
8012 		return -EADDRNOTAVAIL;
8013 
8014 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
8015 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8016 
8017 	ixgbe_mac_set_default_filter(adapter);
8018 
8019 	return 0;
8020 }
8021 
8022 static int
8023 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8024 {
8025 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8026 	struct ixgbe_hw *hw = &adapter->hw;
8027 	u16 value;
8028 	int rc;
8029 
8030 	if (prtad != hw->phy.mdio.prtad)
8031 		return -EINVAL;
8032 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8033 	if (!rc)
8034 		rc = value;
8035 	return rc;
8036 }
8037 
8038 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8039 			    u16 addr, u16 value)
8040 {
8041 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8042 	struct ixgbe_hw *hw = &adapter->hw;
8043 
8044 	if (prtad != hw->phy.mdio.prtad)
8045 		return -EINVAL;
8046 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8047 }
8048 
8049 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8050 {
8051 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8052 
8053 	switch (cmd) {
8054 	case SIOCSHWTSTAMP:
8055 		return ixgbe_ptp_set_ts_config(adapter, req);
8056 	case SIOCGHWTSTAMP:
8057 		return ixgbe_ptp_get_ts_config(adapter, req);
8058 	default:
8059 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8060 	}
8061 }
8062 
8063 /**
8064  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8065  * netdev->dev_addrs
8066  * @netdev: network interface device structure
8067  *
8068  * Returns non-zero on failure
8069  **/
8070 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8071 {
8072 	int err = 0;
8073 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8074 	struct ixgbe_hw *hw = &adapter->hw;
8075 
8076 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8077 		rtnl_lock();
8078 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8079 		rtnl_unlock();
8080 
8081 		/* update SAN MAC vmdq pool selection */
8082 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8083 	}
8084 	return err;
8085 }
8086 
8087 /**
8088  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8089  * netdev->dev_addrs
8090  * @netdev: network interface device structure
8091  *
8092  * Returns non-zero on failure
8093  **/
8094 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8095 {
8096 	int err = 0;
8097 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8098 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8099 
8100 	if (is_valid_ether_addr(mac->san_addr)) {
8101 		rtnl_lock();
8102 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8103 		rtnl_unlock();
8104 	}
8105 	return err;
8106 }
8107 
8108 #ifdef CONFIG_NET_POLL_CONTROLLER
8109 /*
8110  * Polling 'interrupt' - used by things like netconsole to send skbs
8111  * without having to re-enable interrupts. It's not called while
8112  * the interrupt routine is executing.
8113  */
8114 static void ixgbe_netpoll(struct net_device *netdev)
8115 {
8116 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8117 	int i;
8118 
8119 	/* if interface is down do nothing */
8120 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8121 		return;
8122 
8123 	/* loop through and schedule all active queues */
8124 	for (i = 0; i < adapter->num_q_vectors; i++)
8125 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8126 }
8127 
8128 #endif
8129 
8130 static void ixgbe_get_stats64(struct net_device *netdev,
8131 			      struct rtnl_link_stats64 *stats)
8132 {
8133 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8134 	int i;
8135 
8136 	rcu_read_lock();
8137 	for (i = 0; i < adapter->num_rx_queues; i++) {
8138 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
8139 		u64 bytes, packets;
8140 		unsigned int start;
8141 
8142 		if (ring) {
8143 			do {
8144 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8145 				packets = ring->stats.packets;
8146 				bytes   = ring->stats.bytes;
8147 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8148 			stats->rx_packets += packets;
8149 			stats->rx_bytes   += bytes;
8150 		}
8151 	}
8152 
8153 	for (i = 0; i < adapter->num_tx_queues; i++) {
8154 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
8155 		u64 bytes, packets;
8156 		unsigned int start;
8157 
8158 		if (ring) {
8159 			do {
8160 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8161 				packets = ring->stats.packets;
8162 				bytes   = ring->stats.bytes;
8163 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8164 			stats->tx_packets += packets;
8165 			stats->tx_bytes   += bytes;
8166 		}
8167 	}
8168 	rcu_read_unlock();
8169 
8170 	/* following stats updated by ixgbe_watchdog_task() */
8171 	stats->multicast	= netdev->stats.multicast;
8172 	stats->rx_errors	= netdev->stats.rx_errors;
8173 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8174 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8175 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8176 }
8177 
8178 #ifdef CONFIG_IXGBE_DCB
8179 /**
8180  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8181  * @adapter: pointer to ixgbe_adapter
8182  * @tc: number of traffic classes currently enabled
8183  *
8184  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8185  * 802.1Q priority maps to a packet buffer that exists.
8186  */
8187 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8188 {
8189 	struct ixgbe_hw *hw = &adapter->hw;
8190 	u32 reg, rsave;
8191 	int i;
8192 
8193 	/* 82598 have a static priority to TC mapping that can not
8194 	 * be changed so no validation is needed.
8195 	 */
8196 	if (hw->mac.type == ixgbe_mac_82598EB)
8197 		return;
8198 
8199 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8200 	rsave = reg;
8201 
8202 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8203 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8204 
8205 		/* If up2tc is out of bounds default to zero */
8206 		if (up2tc > tc)
8207 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8208 	}
8209 
8210 	if (reg != rsave)
8211 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8212 
8213 	return;
8214 }
8215 
8216 /**
8217  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8218  * @adapter: Pointer to adapter struct
8219  *
8220  * Populate the netdev user priority to tc map
8221  */
8222 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8223 {
8224 	struct net_device *dev = adapter->netdev;
8225 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8226 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8227 	u8 prio;
8228 
8229 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8230 		u8 tc = 0;
8231 
8232 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8233 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8234 		else if (ets)
8235 			tc = ets->prio_tc[prio];
8236 
8237 		netdev_set_prio_tc_map(dev, prio, tc);
8238 	}
8239 }
8240 
8241 #endif /* CONFIG_IXGBE_DCB */
8242 /**
8243  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8244  *
8245  * @netdev: net device to configure
8246  * @tc: number of traffic classes to enable
8247  */
8248 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8249 {
8250 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8251 	struct ixgbe_hw *hw = &adapter->hw;
8252 	bool pools;
8253 
8254 	/* Hardware supports up to 8 traffic classes */
8255 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8256 		return -EINVAL;
8257 
8258 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8259 		return -EINVAL;
8260 
8261 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
8262 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
8263 		return -EBUSY;
8264 
8265 	/* Hardware has to reinitialize queues and interrupts to
8266 	 * match packet buffer alignment. Unfortunately, the
8267 	 * hardware is not flexible enough to do this dynamically.
8268 	 */
8269 	if (netif_running(dev))
8270 		ixgbe_close(dev);
8271 	else
8272 		ixgbe_reset(adapter);
8273 
8274 	ixgbe_clear_interrupt_scheme(adapter);
8275 
8276 #ifdef CONFIG_IXGBE_DCB
8277 	if (tc) {
8278 		netdev_set_num_tc(dev, tc);
8279 		ixgbe_set_prio_tc_map(adapter);
8280 
8281 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8282 
8283 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8284 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8285 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8286 		}
8287 	} else {
8288 		netdev_reset_tc(dev);
8289 
8290 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8291 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8292 
8293 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8294 
8295 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8296 		adapter->dcb_cfg.pfc_mode_enable = false;
8297 	}
8298 
8299 	ixgbe_validate_rtr(adapter, tc);
8300 
8301 #endif /* CONFIG_IXGBE_DCB */
8302 	ixgbe_init_interrupt_scheme(adapter);
8303 
8304 	if (netif_running(dev))
8305 		return ixgbe_open(dev);
8306 
8307 	return 0;
8308 }
8309 
8310 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8311 			       struct tc_cls_u32_offload *cls)
8312 {
8313 	u32 hdl = cls->knode.handle;
8314 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8315 	u32 loc = cls->knode.handle & 0xfffff;
8316 	int err = 0, i, j;
8317 	struct ixgbe_jump_table *jump = NULL;
8318 
8319 	if (loc > IXGBE_MAX_HW_ENTRIES)
8320 		return -EINVAL;
8321 
8322 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8323 		return -EINVAL;
8324 
8325 	/* Clear this filter in the link data it is associated with */
8326 	if (uhtid != 0x800) {
8327 		jump = adapter->jump_tables[uhtid];
8328 		if (!jump)
8329 			return -EINVAL;
8330 		if (!test_bit(loc - 1, jump->child_loc_map))
8331 			return -EINVAL;
8332 		clear_bit(loc - 1, jump->child_loc_map);
8333 	}
8334 
8335 	/* Check if the filter being deleted is a link */
8336 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8337 		jump = adapter->jump_tables[i];
8338 		if (jump && jump->link_hdl == hdl) {
8339 			/* Delete filters in the hardware in the child hash
8340 			 * table associated with this link
8341 			 */
8342 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8343 				if (!test_bit(j, jump->child_loc_map))
8344 					continue;
8345 				spin_lock(&adapter->fdir_perfect_lock);
8346 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8347 								      NULL,
8348 								      j + 1);
8349 				spin_unlock(&adapter->fdir_perfect_lock);
8350 				clear_bit(j, jump->child_loc_map);
8351 			}
8352 			/* Remove resources for this link */
8353 			kfree(jump->input);
8354 			kfree(jump->mask);
8355 			kfree(jump);
8356 			adapter->jump_tables[i] = NULL;
8357 			return err;
8358 		}
8359 	}
8360 
8361 	spin_lock(&adapter->fdir_perfect_lock);
8362 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8363 	spin_unlock(&adapter->fdir_perfect_lock);
8364 	return err;
8365 }
8366 
8367 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8368 					    __be16 protocol,
8369 					    struct tc_cls_u32_offload *cls)
8370 {
8371 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8372 
8373 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8374 		return -EINVAL;
8375 
8376 	/* This ixgbe devices do not support hash tables at the moment
8377 	 * so abort when given hash tables.
8378 	 */
8379 	if (cls->hnode.divisor > 0)
8380 		return -EINVAL;
8381 
8382 	set_bit(uhtid - 1, &adapter->tables);
8383 	return 0;
8384 }
8385 
8386 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
8387 					    struct tc_cls_u32_offload *cls)
8388 {
8389 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8390 
8391 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8392 		return -EINVAL;
8393 
8394 	clear_bit(uhtid - 1, &adapter->tables);
8395 	return 0;
8396 }
8397 
8398 #ifdef CONFIG_NET_CLS_ACT
8399 struct upper_walk_data {
8400 	struct ixgbe_adapter *adapter;
8401 	u64 action;
8402 	int ifindex;
8403 	u8 queue;
8404 };
8405 
8406 static int get_macvlan_queue(struct net_device *upper, void *_data)
8407 {
8408 	if (netif_is_macvlan(upper)) {
8409 		struct macvlan_dev *dfwd = netdev_priv(upper);
8410 		struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
8411 		struct upper_walk_data *data = _data;
8412 		struct ixgbe_adapter *adapter = data->adapter;
8413 		int ifindex = data->ifindex;
8414 
8415 		if (vadapter && vadapter->netdev->ifindex == ifindex) {
8416 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
8417 			data->action = data->queue;
8418 			return 1;
8419 		}
8420 	}
8421 
8422 	return 0;
8423 }
8424 
8425 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
8426 				  u8 *queue, u64 *action)
8427 {
8428 	unsigned int num_vfs = adapter->num_vfs, vf;
8429 	struct upper_walk_data data;
8430 	struct net_device *upper;
8431 
8432 	/* redirect to a SRIOV VF */
8433 	for (vf = 0; vf < num_vfs; ++vf) {
8434 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
8435 		if (upper->ifindex == ifindex) {
8436 			if (adapter->num_rx_pools > 1)
8437 				*queue = vf * 2;
8438 			else
8439 				*queue = vf * adapter->num_rx_queues_per_pool;
8440 
8441 			*action = vf + 1;
8442 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
8443 			return 0;
8444 		}
8445 	}
8446 
8447 	/* redirect to a offloaded macvlan netdev */
8448 	data.adapter = adapter;
8449 	data.ifindex = ifindex;
8450 	data.action = 0;
8451 	data.queue = 0;
8452 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
8453 					  get_macvlan_queue, &data)) {
8454 		*action = data.action;
8455 		*queue = data.queue;
8456 
8457 		return 0;
8458 	}
8459 
8460 	return -EINVAL;
8461 }
8462 
8463 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8464 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8465 {
8466 	const struct tc_action *a;
8467 	LIST_HEAD(actions);
8468 	int err;
8469 
8470 	if (tc_no_actions(exts))
8471 		return -EINVAL;
8472 
8473 	tcf_exts_to_list(exts, &actions);
8474 	list_for_each_entry(a, &actions, list) {
8475 
8476 		/* Drop action */
8477 		if (is_tcf_gact_shot(a)) {
8478 			*action = IXGBE_FDIR_DROP_QUEUE;
8479 			*queue = IXGBE_FDIR_DROP_QUEUE;
8480 			return 0;
8481 		}
8482 
8483 		/* Redirect to a VF or a offloaded macvlan */
8484 		if (is_tcf_mirred_egress_redirect(a)) {
8485 			int ifindex = tcf_mirred_ifindex(a);
8486 
8487 			err = handle_redirect_action(adapter, ifindex, queue,
8488 						     action);
8489 			if (err == 0)
8490 				return err;
8491 		}
8492 	}
8493 
8494 	return -EINVAL;
8495 }
8496 #else
8497 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8498 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8499 {
8500 	return -EINVAL;
8501 }
8502 #endif /* CONFIG_NET_CLS_ACT */
8503 
8504 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
8505 				    union ixgbe_atr_input *mask,
8506 				    struct tc_cls_u32_offload *cls,
8507 				    struct ixgbe_mat_field *field_ptr,
8508 				    struct ixgbe_nexthdr *nexthdr)
8509 {
8510 	int i, j, off;
8511 	__be32 val, m;
8512 	bool found_entry = false, found_jump_field = false;
8513 
8514 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
8515 		off = cls->knode.sel->keys[i].off;
8516 		val = cls->knode.sel->keys[i].val;
8517 		m = cls->knode.sel->keys[i].mask;
8518 
8519 		for (j = 0; field_ptr[j].val; j++) {
8520 			if (field_ptr[j].off == off) {
8521 				field_ptr[j].val(input, mask, val, m);
8522 				input->filter.formatted.flow_type |=
8523 					field_ptr[j].type;
8524 				found_entry = true;
8525 				break;
8526 			}
8527 		}
8528 		if (nexthdr) {
8529 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
8530 			    nexthdr->val == cls->knode.sel->keys[i].val &&
8531 			    nexthdr->mask == cls->knode.sel->keys[i].mask)
8532 				found_jump_field = true;
8533 			else
8534 				continue;
8535 		}
8536 	}
8537 
8538 	if (nexthdr && !found_jump_field)
8539 		return -EINVAL;
8540 
8541 	if (!found_entry)
8542 		return 0;
8543 
8544 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
8545 				    IXGBE_ATR_L4TYPE_MASK;
8546 
8547 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
8548 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
8549 
8550 	return 0;
8551 }
8552 
8553 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
8554 				  __be16 protocol,
8555 				  struct tc_cls_u32_offload *cls)
8556 {
8557 	u32 loc = cls->knode.handle & 0xfffff;
8558 	struct ixgbe_hw *hw = &adapter->hw;
8559 	struct ixgbe_mat_field *field_ptr;
8560 	struct ixgbe_fdir_filter *input = NULL;
8561 	union ixgbe_atr_input *mask = NULL;
8562 	struct ixgbe_jump_table *jump = NULL;
8563 	int i, err = -EINVAL;
8564 	u8 queue;
8565 	u32 uhtid, link_uhtid;
8566 
8567 	uhtid = TC_U32_USERHTID(cls->knode.handle);
8568 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
8569 
8570 	/* At the moment cls_u32 jumps to network layer and skips past
8571 	 * L2 headers. The canonical method to match L2 frames is to use
8572 	 * negative values. However this is error prone at best but really
8573 	 * just broken because there is no way to "know" what sort of hdr
8574 	 * is in front of the network layer. Fix cls_u32 to support L2
8575 	 * headers when needed.
8576 	 */
8577 	if (protocol != htons(ETH_P_IP))
8578 		return err;
8579 
8580 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
8581 		e_err(drv, "Location out of range\n");
8582 		return err;
8583 	}
8584 
8585 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
8586 	 * links and also the fields used to advance the parser across each
8587 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
8588 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
8589 	 * To add support for new nodes update ixgbe_model.h parse structures
8590 	 * this function _should_ be generic try not to hardcode values here.
8591 	 */
8592 	if (uhtid == 0x800) {
8593 		field_ptr = (adapter->jump_tables[0])->mat;
8594 	} else {
8595 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8596 			return err;
8597 		if (!adapter->jump_tables[uhtid])
8598 			return err;
8599 		field_ptr = (adapter->jump_tables[uhtid])->mat;
8600 	}
8601 
8602 	if (!field_ptr)
8603 		return err;
8604 
8605 	/* At this point we know the field_ptr is valid and need to either
8606 	 * build cls_u32 link or attach filter. Because adding a link to
8607 	 * a handle that does not exist is invalid and the same for adding
8608 	 * rules to handles that don't exist.
8609 	 */
8610 
8611 	if (link_uhtid) {
8612 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
8613 
8614 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
8615 			return err;
8616 
8617 		if (!test_bit(link_uhtid - 1, &adapter->tables))
8618 			return err;
8619 
8620 		/* Multiple filters as links to the same hash table are not
8621 		 * supported. To add a new filter with the same next header
8622 		 * but different match/jump conditions, create a new hash table
8623 		 * and link to it.
8624 		 */
8625 		if (adapter->jump_tables[link_uhtid] &&
8626 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
8627 			e_err(drv, "Link filter exists for link: %x\n",
8628 			      link_uhtid);
8629 			return err;
8630 		}
8631 
8632 		for (i = 0; nexthdr[i].jump; i++) {
8633 			if (nexthdr[i].o != cls->knode.sel->offoff ||
8634 			    nexthdr[i].s != cls->knode.sel->offshift ||
8635 			    nexthdr[i].m != cls->knode.sel->offmask)
8636 				return err;
8637 
8638 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
8639 			if (!jump)
8640 				return -ENOMEM;
8641 			input = kzalloc(sizeof(*input), GFP_KERNEL);
8642 			if (!input) {
8643 				err = -ENOMEM;
8644 				goto free_jump;
8645 			}
8646 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8647 			if (!mask) {
8648 				err = -ENOMEM;
8649 				goto free_input;
8650 			}
8651 			jump->input = input;
8652 			jump->mask = mask;
8653 			jump->link_hdl = cls->knode.handle;
8654 
8655 			err = ixgbe_clsu32_build_input(input, mask, cls,
8656 						       field_ptr, &nexthdr[i]);
8657 			if (!err) {
8658 				jump->mat = nexthdr[i].jump;
8659 				adapter->jump_tables[link_uhtid] = jump;
8660 				break;
8661 			}
8662 		}
8663 		return 0;
8664 	}
8665 
8666 	input = kzalloc(sizeof(*input), GFP_KERNEL);
8667 	if (!input)
8668 		return -ENOMEM;
8669 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8670 	if (!mask) {
8671 		err = -ENOMEM;
8672 		goto free_input;
8673 	}
8674 
8675 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
8676 		if ((adapter->jump_tables[uhtid])->input)
8677 			memcpy(input, (adapter->jump_tables[uhtid])->input,
8678 			       sizeof(*input));
8679 		if ((adapter->jump_tables[uhtid])->mask)
8680 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
8681 			       sizeof(*mask));
8682 
8683 		/* Lookup in all child hash tables if this location is already
8684 		 * filled with a filter
8685 		 */
8686 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8687 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
8688 
8689 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
8690 				e_err(drv, "Filter exists in location: %x\n",
8691 				      loc);
8692 				err = -EINVAL;
8693 				goto err_out;
8694 			}
8695 		}
8696 	}
8697 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
8698 	if (err)
8699 		goto err_out;
8700 
8701 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
8702 			       &queue);
8703 	if (err < 0)
8704 		goto err_out;
8705 
8706 	input->sw_idx = loc;
8707 
8708 	spin_lock(&adapter->fdir_perfect_lock);
8709 
8710 	if (hlist_empty(&adapter->fdir_filter_list)) {
8711 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
8712 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
8713 		if (err)
8714 			goto err_out_w_lock;
8715 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
8716 		err = -EINVAL;
8717 		goto err_out_w_lock;
8718 	}
8719 
8720 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
8721 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
8722 						    input->sw_idx, queue);
8723 	if (!err)
8724 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
8725 	spin_unlock(&adapter->fdir_perfect_lock);
8726 
8727 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
8728 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
8729 
8730 	kfree(mask);
8731 	return err;
8732 err_out_w_lock:
8733 	spin_unlock(&adapter->fdir_perfect_lock);
8734 err_out:
8735 	kfree(mask);
8736 free_input:
8737 	kfree(input);
8738 free_jump:
8739 	kfree(jump);
8740 	return err;
8741 }
8742 
8743 static int __ixgbe_setup_tc(struct net_device *dev, u32 handle, __be16 proto,
8744 			    struct tc_to_netdev *tc)
8745 {
8746 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8747 
8748 	if (TC_H_MAJ(handle) == TC_H_MAJ(TC_H_INGRESS) &&
8749 	    tc->type == TC_SETUP_CLSU32) {
8750 		switch (tc->cls_u32->command) {
8751 		case TC_CLSU32_NEW_KNODE:
8752 		case TC_CLSU32_REPLACE_KNODE:
8753 			return ixgbe_configure_clsu32(adapter,
8754 						      proto, tc->cls_u32);
8755 		case TC_CLSU32_DELETE_KNODE:
8756 			return ixgbe_delete_clsu32(adapter, tc->cls_u32);
8757 		case TC_CLSU32_NEW_HNODE:
8758 		case TC_CLSU32_REPLACE_HNODE:
8759 			return ixgbe_configure_clsu32_add_hnode(adapter, proto,
8760 								tc->cls_u32);
8761 		case TC_CLSU32_DELETE_HNODE:
8762 			return ixgbe_configure_clsu32_del_hnode(adapter,
8763 								tc->cls_u32);
8764 		default:
8765 			return -EINVAL;
8766 		}
8767 	}
8768 
8769 	if (tc->type != TC_SETUP_MQPRIO)
8770 		return -EINVAL;
8771 
8772 	return ixgbe_setup_tc(dev, tc->tc);
8773 }
8774 
8775 #ifdef CONFIG_PCI_IOV
8776 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
8777 {
8778 	struct net_device *netdev = adapter->netdev;
8779 
8780 	rtnl_lock();
8781 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
8782 	rtnl_unlock();
8783 }
8784 
8785 #endif
8786 void ixgbe_do_reset(struct net_device *netdev)
8787 {
8788 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8789 
8790 	if (netif_running(netdev))
8791 		ixgbe_reinit_locked(adapter);
8792 	else
8793 		ixgbe_reset(adapter);
8794 }
8795 
8796 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
8797 					    netdev_features_t features)
8798 {
8799 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8800 
8801 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
8802 	if (!(features & NETIF_F_RXCSUM))
8803 		features &= ~NETIF_F_LRO;
8804 
8805 	/* Turn off LRO if not RSC capable */
8806 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
8807 		features &= ~NETIF_F_LRO;
8808 
8809 	return features;
8810 }
8811 
8812 static int ixgbe_set_features(struct net_device *netdev,
8813 			      netdev_features_t features)
8814 {
8815 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8816 	netdev_features_t changed = netdev->features ^ features;
8817 	bool need_reset = false;
8818 
8819 	/* Make sure RSC matches LRO, reset if change */
8820 	if (!(features & NETIF_F_LRO)) {
8821 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8822 			need_reset = true;
8823 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
8824 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
8825 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
8826 		if (adapter->rx_itr_setting == 1 ||
8827 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
8828 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
8829 			need_reset = true;
8830 		} else if ((changed ^ features) & NETIF_F_LRO) {
8831 			e_info(probe, "rx-usecs set too low, "
8832 			       "disabling RSC\n");
8833 		}
8834 	}
8835 
8836 	/*
8837 	 * Check if Flow Director n-tuple support or hw_tc support was
8838 	 * enabled or disabled.  If the state changed, we need to reset.
8839 	 */
8840 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
8841 		/* turn off ATR, enable perfect filters and reset */
8842 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
8843 			need_reset = true;
8844 
8845 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
8846 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8847 	} else {
8848 		/* turn off perfect filters, enable ATR and reset */
8849 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
8850 			need_reset = true;
8851 
8852 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8853 
8854 		/* We cannot enable ATR if SR-IOV is enabled */
8855 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
8856 		    /* We cannot enable ATR if we have 2 or more tcs */
8857 		    (netdev_get_num_tc(netdev) > 1) ||
8858 		    /* We cannot enable ATR if RSS is disabled */
8859 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
8860 		    /* A sample rate of 0 indicates ATR disabled */
8861 		    (!adapter->atr_sample_rate))
8862 			; /* do nothing not supported */
8863 		else /* otherwise supported and set the flag */
8864 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
8865 	}
8866 
8867 	if (changed & NETIF_F_RXALL)
8868 		need_reset = true;
8869 
8870 	netdev->features = features;
8871 
8872 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
8873 		if (features & NETIF_F_RXCSUM) {
8874 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
8875 		} else {
8876 			u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
8877 
8878 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
8879 		}
8880 	}
8881 
8882 	if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
8883 		if (features & NETIF_F_RXCSUM) {
8884 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
8885 		} else {
8886 			u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
8887 
8888 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
8889 		}
8890 	}
8891 
8892 	if (need_reset)
8893 		ixgbe_do_reset(netdev);
8894 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
8895 			    NETIF_F_HW_VLAN_CTAG_FILTER))
8896 		ixgbe_set_rx_mode(netdev);
8897 
8898 	return 0;
8899 }
8900 
8901 /**
8902  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
8903  * @dev: The port's netdev
8904  * @ti: Tunnel endpoint information
8905  **/
8906 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
8907 				      struct udp_tunnel_info *ti)
8908 {
8909 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8910 	struct ixgbe_hw *hw = &adapter->hw;
8911 	__be16 port = ti->port;
8912 	u32 port_shift = 0;
8913 	u32 reg;
8914 
8915 	if (ti->sa_family != AF_INET)
8916 		return;
8917 
8918 	switch (ti->type) {
8919 	case UDP_TUNNEL_TYPE_VXLAN:
8920 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8921 			return;
8922 
8923 		if (adapter->vxlan_port == port)
8924 			return;
8925 
8926 		if (adapter->vxlan_port) {
8927 			netdev_info(dev,
8928 				    "VXLAN port %d set, not adding port %d\n",
8929 				    ntohs(adapter->vxlan_port),
8930 				    ntohs(port));
8931 			return;
8932 		}
8933 
8934 		adapter->vxlan_port = port;
8935 		break;
8936 	case UDP_TUNNEL_TYPE_GENEVE:
8937 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
8938 			return;
8939 
8940 		if (adapter->geneve_port == port)
8941 			return;
8942 
8943 		if (adapter->geneve_port) {
8944 			netdev_info(dev,
8945 				    "GENEVE port %d set, not adding port %d\n",
8946 				    ntohs(adapter->geneve_port),
8947 				    ntohs(port));
8948 			return;
8949 		}
8950 
8951 		port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
8952 		adapter->geneve_port = port;
8953 		break;
8954 	default:
8955 		return;
8956 	}
8957 
8958 	reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
8959 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
8960 }
8961 
8962 /**
8963  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
8964  * @dev: The port's netdev
8965  * @ti: Tunnel endpoint information
8966  **/
8967 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
8968 				      struct udp_tunnel_info *ti)
8969 {
8970 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8971 	u32 port_mask;
8972 
8973 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
8974 	    ti->type != UDP_TUNNEL_TYPE_GENEVE)
8975 		return;
8976 
8977 	if (ti->sa_family != AF_INET)
8978 		return;
8979 
8980 	switch (ti->type) {
8981 	case UDP_TUNNEL_TYPE_VXLAN:
8982 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8983 			return;
8984 
8985 		if (adapter->vxlan_port != ti->port) {
8986 			netdev_info(dev, "VXLAN port %d not found\n",
8987 				    ntohs(ti->port));
8988 			return;
8989 		}
8990 
8991 		port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
8992 		break;
8993 	case UDP_TUNNEL_TYPE_GENEVE:
8994 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
8995 			return;
8996 
8997 		if (adapter->geneve_port != ti->port) {
8998 			netdev_info(dev, "GENEVE port %d not found\n",
8999 				    ntohs(ti->port));
9000 			return;
9001 		}
9002 
9003 		port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9004 		break;
9005 	default:
9006 		return;
9007 	}
9008 
9009 	ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9010 	adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9011 }
9012 
9013 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9014 			     struct net_device *dev,
9015 			     const unsigned char *addr, u16 vid,
9016 			     u16 flags)
9017 {
9018 	/* guarantee we can provide a unique filter for the unicast address */
9019 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9020 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9021 		u16 pool = VMDQ_P(0);
9022 
9023 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9024 			return -ENOMEM;
9025 	}
9026 
9027 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9028 }
9029 
9030 /**
9031  * ixgbe_configure_bridge_mode - set various bridge modes
9032  * @adapter - the private structure
9033  * @mode - requested bridge mode
9034  *
9035  * Configure some settings require for various bridge modes.
9036  **/
9037 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9038 				       __u16 mode)
9039 {
9040 	struct ixgbe_hw *hw = &adapter->hw;
9041 	unsigned int p, num_pools;
9042 	u32 vmdctl;
9043 
9044 	switch (mode) {
9045 	case BRIDGE_MODE_VEPA:
9046 		/* disable Tx loopback, rely on switch hairpin mode */
9047 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9048 
9049 		/* must enable Rx switching replication to allow multicast
9050 		 * packet reception on all VFs, and to enable source address
9051 		 * pruning.
9052 		 */
9053 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9054 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9055 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9056 
9057 		/* enable Rx source address pruning. Note, this requires
9058 		 * replication to be enabled or else it does nothing.
9059 		 */
9060 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9061 		for (p = 0; p < num_pools; p++) {
9062 			if (hw->mac.ops.set_source_address_pruning)
9063 				hw->mac.ops.set_source_address_pruning(hw,
9064 								       true,
9065 								       p);
9066 		}
9067 		break;
9068 	case BRIDGE_MODE_VEB:
9069 		/* enable Tx loopback for internal VF/PF communication */
9070 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9071 				IXGBE_PFDTXGSWC_VT_LBEN);
9072 
9073 		/* disable Rx switching replication unless we have SR-IOV
9074 		 * virtual functions
9075 		 */
9076 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9077 		if (!adapter->num_vfs)
9078 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9079 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9080 
9081 		/* disable Rx source address pruning, since we don't expect to
9082 		 * be receiving external loopback of our transmitted frames.
9083 		 */
9084 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9085 		for (p = 0; p < num_pools; p++) {
9086 			if (hw->mac.ops.set_source_address_pruning)
9087 				hw->mac.ops.set_source_address_pruning(hw,
9088 								       false,
9089 								       p);
9090 		}
9091 		break;
9092 	default:
9093 		return -EINVAL;
9094 	}
9095 
9096 	adapter->bridge_mode = mode;
9097 
9098 	e_info(drv, "enabling bridge mode: %s\n",
9099 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9100 
9101 	return 0;
9102 }
9103 
9104 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9105 				    struct nlmsghdr *nlh, u16 flags)
9106 {
9107 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9108 	struct nlattr *attr, *br_spec;
9109 	int rem;
9110 
9111 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9112 		return -EOPNOTSUPP;
9113 
9114 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9115 	if (!br_spec)
9116 		return -EINVAL;
9117 
9118 	nla_for_each_nested(attr, br_spec, rem) {
9119 		int status;
9120 		__u16 mode;
9121 
9122 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9123 			continue;
9124 
9125 		if (nla_len(attr) < sizeof(mode))
9126 			return -EINVAL;
9127 
9128 		mode = nla_get_u16(attr);
9129 		status = ixgbe_configure_bridge_mode(adapter, mode);
9130 		if (status)
9131 			return status;
9132 
9133 		break;
9134 	}
9135 
9136 	return 0;
9137 }
9138 
9139 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9140 				    struct net_device *dev,
9141 				    u32 filter_mask, int nlflags)
9142 {
9143 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9144 
9145 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9146 		return 0;
9147 
9148 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9149 				       adapter->bridge_mode, 0, 0, nlflags,
9150 				       filter_mask, NULL);
9151 }
9152 
9153 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9154 {
9155 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
9156 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9157 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
9158 	unsigned int limit;
9159 	int pool, err;
9160 
9161 	/* Hardware has a limited number of available pools. Each VF, and the
9162 	 * PF require a pool. Check to ensure we don't attempt to use more
9163 	 * then the available number of pools.
9164 	 */
9165 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9166 		return ERR_PTR(-EINVAL);
9167 
9168 #ifdef CONFIG_RPS
9169 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
9170 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
9171 			    vdev->name);
9172 		return ERR_PTR(-EINVAL);
9173 	}
9174 #endif
9175 	/* Check for hardware restriction on number of rx/tx queues */
9176 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
9177 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
9178 		netdev_info(pdev,
9179 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
9180 			    pdev->name);
9181 		return ERR_PTR(-EINVAL);
9182 	}
9183 
9184 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9185 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
9186 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
9187 		return ERR_PTR(-EBUSY);
9188 
9189 	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
9190 	if (!fwd_adapter)
9191 		return ERR_PTR(-ENOMEM);
9192 
9193 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
9194 	adapter->num_rx_pools++;
9195 	set_bit(pool, &adapter->fwd_bitmask);
9196 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9197 
9198 	/* Enable VMDq flag so device will be set in VM mode */
9199 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
9200 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9201 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
9202 
9203 	/* Force reinit of ring allocation with VMDQ enabled */
9204 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9205 	if (err)
9206 		goto fwd_add_err;
9207 	fwd_adapter->pool = pool;
9208 	fwd_adapter->real_adapter = adapter;
9209 
9210 	if (netif_running(pdev)) {
9211 		err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
9212 		if (err)
9213 			goto fwd_add_err;
9214 		netif_tx_start_all_queues(vdev);
9215 	}
9216 
9217 	return fwd_adapter;
9218 fwd_add_err:
9219 	/* unwind counter and free adapter struct */
9220 	netdev_info(pdev,
9221 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
9222 	clear_bit(pool, &adapter->fwd_bitmask);
9223 	adapter->num_rx_pools--;
9224 	kfree(fwd_adapter);
9225 	return ERR_PTR(err);
9226 }
9227 
9228 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9229 {
9230 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
9231 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
9232 	unsigned int limit;
9233 
9234 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
9235 	adapter->num_rx_pools--;
9236 
9237 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9238 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9239 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
9240 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9241 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
9242 		   fwd_adapter->pool, adapter->num_rx_pools,
9243 		   fwd_adapter->rx_base_queue,
9244 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
9245 		   adapter->fwd_bitmask);
9246 	kfree(fwd_adapter);
9247 }
9248 
9249 #define IXGBE_MAX_MAC_HDR_LEN		127
9250 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9251 
9252 static netdev_features_t
9253 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9254 		     netdev_features_t features)
9255 {
9256 	unsigned int network_hdr_len, mac_hdr_len;
9257 
9258 	/* Make certain the headers can be described by a context descriptor */
9259 	mac_hdr_len = skb_network_header(skb) - skb->data;
9260 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9261 		return features & ~(NETIF_F_HW_CSUM |
9262 				    NETIF_F_SCTP_CRC |
9263 				    NETIF_F_HW_VLAN_CTAG_TX |
9264 				    NETIF_F_TSO |
9265 				    NETIF_F_TSO6);
9266 
9267 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9268 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9269 		return features & ~(NETIF_F_HW_CSUM |
9270 				    NETIF_F_SCTP_CRC |
9271 				    NETIF_F_TSO |
9272 				    NETIF_F_TSO6);
9273 
9274 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9275 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9276 	 */
9277 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
9278 		features &= ~NETIF_F_TSO;
9279 
9280 	return features;
9281 }
9282 
9283 static const struct net_device_ops ixgbe_netdev_ops = {
9284 	.ndo_open		= ixgbe_open,
9285 	.ndo_stop		= ixgbe_close,
9286 	.ndo_start_xmit		= ixgbe_xmit_frame,
9287 	.ndo_select_queue	= ixgbe_select_queue,
9288 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
9289 	.ndo_validate_addr	= eth_validate_addr,
9290 	.ndo_set_mac_address	= ixgbe_set_mac,
9291 	.ndo_change_mtu		= ixgbe_change_mtu,
9292 	.ndo_tx_timeout		= ixgbe_tx_timeout,
9293 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
9294 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
9295 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
9296 	.ndo_do_ioctl		= ixgbe_ioctl,
9297 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
9298 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
9299 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
9300 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
9301 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
9302 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
9303 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
9304 	.ndo_get_stats64	= ixgbe_get_stats64,
9305 	.ndo_setup_tc		= __ixgbe_setup_tc,
9306 #ifdef CONFIG_NET_POLL_CONTROLLER
9307 	.ndo_poll_controller	= ixgbe_netpoll,
9308 #endif
9309 #ifdef IXGBE_FCOE
9310 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
9311 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
9312 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
9313 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
9314 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
9315 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
9316 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
9317 #endif /* IXGBE_FCOE */
9318 	.ndo_set_features = ixgbe_set_features,
9319 	.ndo_fix_features = ixgbe_fix_features,
9320 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
9321 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
9322 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
9323 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
9324 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
9325 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
9326 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
9327 	.ndo_features_check	= ixgbe_features_check,
9328 };
9329 
9330 /**
9331  * ixgbe_enumerate_functions - Get the number of ports this device has
9332  * @adapter: adapter structure
9333  *
9334  * This function enumerates the phsyical functions co-located on a single slot,
9335  * in order to determine how many ports a device has. This is most useful in
9336  * determining the required GT/s of PCIe bandwidth necessary for optimal
9337  * performance.
9338  **/
9339 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
9340 {
9341 	struct pci_dev *entry, *pdev = adapter->pdev;
9342 	int physfns = 0;
9343 
9344 	/* Some cards can not use the generic count PCIe functions method,
9345 	 * because they are behind a parent switch, so we hardcode these with
9346 	 * the correct number of functions.
9347 	 */
9348 	if (ixgbe_pcie_from_parent(&adapter->hw))
9349 		physfns = 4;
9350 
9351 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
9352 		/* don't count virtual functions */
9353 		if (entry->is_virtfn)
9354 			continue;
9355 
9356 		/* When the devices on the bus don't all match our device ID,
9357 		 * we can't reliably determine the correct number of
9358 		 * functions. This can occur if a function has been direct
9359 		 * attached to a virtual machine using VT-d, for example. In
9360 		 * this case, simply return -1 to indicate this.
9361 		 */
9362 		if ((entry->vendor != pdev->vendor) ||
9363 		    (entry->device != pdev->device))
9364 			return -1;
9365 
9366 		physfns++;
9367 	}
9368 
9369 	return physfns;
9370 }
9371 
9372 /**
9373  * ixgbe_wol_supported - Check whether device supports WoL
9374  * @adapter: the adapter private structure
9375  * @device_id: the device ID
9376  * @subdev_id: the subsystem device ID
9377  *
9378  * This function is used by probe and ethtool to determine
9379  * which devices have WoL support
9380  *
9381  **/
9382 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
9383 			 u16 subdevice_id)
9384 {
9385 	struct ixgbe_hw *hw = &adapter->hw;
9386 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
9387 
9388 	/* WOL not supported on 82598 */
9389 	if (hw->mac.type == ixgbe_mac_82598EB)
9390 		return false;
9391 
9392 	/* check eeprom to see if WOL is enabled for X540 and newer */
9393 	if (hw->mac.type >= ixgbe_mac_X540) {
9394 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
9395 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
9396 		     (hw->bus.func == 0)))
9397 			return true;
9398 	}
9399 
9400 	/* WOL is determined based on device IDs for 82599 MACs */
9401 	switch (device_id) {
9402 	case IXGBE_DEV_ID_82599_SFP:
9403 		/* Only these subdevices could supports WOL */
9404 		switch (subdevice_id) {
9405 		case IXGBE_SUBDEV_ID_82599_560FLR:
9406 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
9407 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
9408 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
9409 			/* only support first port */
9410 			if (hw->bus.func != 0)
9411 				break;
9412 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
9413 		case IXGBE_SUBDEV_ID_82599_SFP:
9414 		case IXGBE_SUBDEV_ID_82599_RNDC:
9415 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
9416 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
9417 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
9418 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
9419 			return true;
9420 		}
9421 		break;
9422 	case IXGBE_DEV_ID_82599EN_SFP:
9423 		/* Only these subdevices support WOL */
9424 		switch (subdevice_id) {
9425 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
9426 			return true;
9427 		}
9428 		break;
9429 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
9430 		/* All except this subdevice support WOL */
9431 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
9432 			return true;
9433 		break;
9434 	case IXGBE_DEV_ID_82599_KX4:
9435 		return  true;
9436 	default:
9437 		break;
9438 	}
9439 
9440 	return false;
9441 }
9442 
9443 /**
9444  * ixgbe_probe - Device Initialization Routine
9445  * @pdev: PCI device information struct
9446  * @ent: entry in ixgbe_pci_tbl
9447  *
9448  * Returns 0 on success, negative on failure
9449  *
9450  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
9451  * The OS initialization, configuring of the adapter private structure,
9452  * and a hardware reset occur.
9453  **/
9454 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
9455 {
9456 	struct net_device *netdev;
9457 	struct ixgbe_adapter *adapter = NULL;
9458 	struct ixgbe_hw *hw;
9459 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
9460 	int i, err, pci_using_dac, expected_gts;
9461 	unsigned int indices = MAX_TX_QUEUES;
9462 	u8 part_str[IXGBE_PBANUM_LENGTH];
9463 	bool disable_dev = false;
9464 #ifdef IXGBE_FCOE
9465 	u16 device_caps;
9466 #endif
9467 	u32 eec;
9468 
9469 	/* Catch broken hardware that put the wrong VF device ID in
9470 	 * the PCIe SR-IOV capability.
9471 	 */
9472 	if (pdev->is_virtfn) {
9473 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
9474 		     pci_name(pdev), pdev->vendor, pdev->device);
9475 		return -EINVAL;
9476 	}
9477 
9478 	err = pci_enable_device_mem(pdev);
9479 	if (err)
9480 		return err;
9481 
9482 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
9483 		pci_using_dac = 1;
9484 	} else {
9485 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
9486 		if (err) {
9487 			dev_err(&pdev->dev,
9488 				"No usable DMA configuration, aborting\n");
9489 			goto err_dma;
9490 		}
9491 		pci_using_dac = 0;
9492 	}
9493 
9494 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
9495 	if (err) {
9496 		dev_err(&pdev->dev,
9497 			"pci_request_selected_regions failed 0x%x\n", err);
9498 		goto err_pci_reg;
9499 	}
9500 
9501 	pci_enable_pcie_error_reporting(pdev);
9502 
9503 	pci_set_master(pdev);
9504 	pci_save_state(pdev);
9505 
9506 	if (ii->mac == ixgbe_mac_82598EB) {
9507 #ifdef CONFIG_IXGBE_DCB
9508 		/* 8 TC w/ 4 queues per TC */
9509 		indices = 4 * MAX_TRAFFIC_CLASS;
9510 #else
9511 		indices = IXGBE_MAX_RSS_INDICES;
9512 #endif
9513 	}
9514 
9515 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
9516 	if (!netdev) {
9517 		err = -ENOMEM;
9518 		goto err_alloc_etherdev;
9519 	}
9520 
9521 	SET_NETDEV_DEV(netdev, &pdev->dev);
9522 
9523 	adapter = netdev_priv(netdev);
9524 
9525 	adapter->netdev = netdev;
9526 	adapter->pdev = pdev;
9527 	hw = &adapter->hw;
9528 	hw->back = adapter;
9529 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
9530 
9531 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
9532 			      pci_resource_len(pdev, 0));
9533 	adapter->io_addr = hw->hw_addr;
9534 	if (!hw->hw_addr) {
9535 		err = -EIO;
9536 		goto err_ioremap;
9537 	}
9538 
9539 	netdev->netdev_ops = &ixgbe_netdev_ops;
9540 	ixgbe_set_ethtool_ops(netdev);
9541 	netdev->watchdog_timeo = 5 * HZ;
9542 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
9543 
9544 	/* Setup hw api */
9545 	hw->mac.ops   = *ii->mac_ops;
9546 	hw->mac.type  = ii->mac;
9547 	hw->mvals     = ii->mvals;
9548 	if (ii->link_ops)
9549 		hw->link.ops  = *ii->link_ops;
9550 
9551 	/* EEPROM */
9552 	hw->eeprom.ops = *ii->eeprom_ops;
9553 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
9554 	if (ixgbe_removed(hw->hw_addr)) {
9555 		err = -EIO;
9556 		goto err_ioremap;
9557 	}
9558 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
9559 	if (!(eec & BIT(8)))
9560 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
9561 
9562 	/* PHY */
9563 	hw->phy.ops = *ii->phy_ops;
9564 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
9565 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
9566 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
9567 	hw->phy.mdio.mmds = 0;
9568 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
9569 	hw->phy.mdio.dev = netdev;
9570 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
9571 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
9572 
9573 	/* setup the private structure */
9574 	err = ixgbe_sw_init(adapter, ii);
9575 	if (err)
9576 		goto err_sw_init;
9577 
9578 	/* Make sure the SWFW semaphore is in a valid state */
9579 	if (hw->mac.ops.init_swfw_sync)
9580 		hw->mac.ops.init_swfw_sync(hw);
9581 
9582 	/* Make it possible the adapter to be woken up via WOL */
9583 	switch (adapter->hw.mac.type) {
9584 	case ixgbe_mac_82599EB:
9585 	case ixgbe_mac_X540:
9586 	case ixgbe_mac_X550:
9587 	case ixgbe_mac_X550EM_x:
9588 	case ixgbe_mac_x550em_a:
9589 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
9590 		break;
9591 	default:
9592 		break;
9593 	}
9594 
9595 	/*
9596 	 * If there is a fan on this device and it has failed log the
9597 	 * failure.
9598 	 */
9599 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
9600 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
9601 		if (esdp & IXGBE_ESDP_SDP1)
9602 			e_crit(probe, "Fan has stopped, replace the adapter\n");
9603 	}
9604 
9605 	if (allow_unsupported_sfp)
9606 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
9607 
9608 	/* reset_hw fills in the perm_addr as well */
9609 	hw->phy.reset_if_overtemp = true;
9610 	err = hw->mac.ops.reset_hw(hw);
9611 	hw->phy.reset_if_overtemp = false;
9612 	ixgbe_set_eee_capable(adapter);
9613 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
9614 		err = 0;
9615 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
9616 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
9617 		e_dev_err("Reload the driver after installing a supported module.\n");
9618 		goto err_sw_init;
9619 	} else if (err) {
9620 		e_dev_err("HW Init failed: %d\n", err);
9621 		goto err_sw_init;
9622 	}
9623 
9624 #ifdef CONFIG_PCI_IOV
9625 	/* SR-IOV not supported on the 82598 */
9626 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9627 		goto skip_sriov;
9628 	/* Mailbox */
9629 	ixgbe_init_mbx_params_pf(hw);
9630 	hw->mbx.ops = ii->mbx_ops;
9631 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
9632 	ixgbe_enable_sriov(adapter);
9633 skip_sriov:
9634 
9635 #endif
9636 	netdev->features = NETIF_F_SG |
9637 			   NETIF_F_TSO |
9638 			   NETIF_F_TSO6 |
9639 			   NETIF_F_RXHASH |
9640 			   NETIF_F_RXCSUM |
9641 			   NETIF_F_HW_CSUM;
9642 
9643 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
9644 				    NETIF_F_GSO_GRE_CSUM | \
9645 				    NETIF_F_GSO_IPXIP4 | \
9646 				    NETIF_F_GSO_IPXIP6 | \
9647 				    NETIF_F_GSO_UDP_TUNNEL | \
9648 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
9649 
9650 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
9651 	netdev->features |= NETIF_F_GSO_PARTIAL |
9652 			    IXGBE_GSO_PARTIAL_FEATURES;
9653 
9654 	if (hw->mac.type >= ixgbe_mac_82599EB)
9655 		netdev->features |= NETIF_F_SCTP_CRC;
9656 
9657 	/* copy netdev features into list of user selectable features */
9658 	netdev->hw_features |= netdev->features |
9659 			       NETIF_F_HW_VLAN_CTAG_FILTER |
9660 			       NETIF_F_HW_VLAN_CTAG_RX |
9661 			       NETIF_F_HW_VLAN_CTAG_TX |
9662 			       NETIF_F_RXALL |
9663 			       NETIF_F_HW_L2FW_DOFFLOAD;
9664 
9665 	if (hw->mac.type >= ixgbe_mac_82599EB)
9666 		netdev->hw_features |= NETIF_F_NTUPLE |
9667 				       NETIF_F_HW_TC;
9668 
9669 	if (pci_using_dac)
9670 		netdev->features |= NETIF_F_HIGHDMA;
9671 
9672 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
9673 	netdev->hw_enc_features |= netdev->vlan_features;
9674 	netdev->mpls_features |= NETIF_F_HW_CSUM;
9675 
9676 	/* set this bit last since it cannot be part of vlan_features */
9677 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
9678 			    NETIF_F_HW_VLAN_CTAG_RX |
9679 			    NETIF_F_HW_VLAN_CTAG_TX;
9680 
9681 	netdev->priv_flags |= IFF_UNICAST_FLT;
9682 	netdev->priv_flags |= IFF_SUPP_NOFCS;
9683 
9684 	/* MTU range: 68 - 9710 */
9685 	netdev->min_mtu = ETH_MIN_MTU;
9686 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
9687 
9688 #ifdef CONFIG_IXGBE_DCB
9689 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
9690 		netdev->dcbnl_ops = &dcbnl_ops;
9691 #endif
9692 
9693 #ifdef IXGBE_FCOE
9694 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
9695 		unsigned int fcoe_l;
9696 
9697 		if (hw->mac.ops.get_device_caps) {
9698 			hw->mac.ops.get_device_caps(hw, &device_caps);
9699 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
9700 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
9701 		}
9702 
9703 
9704 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
9705 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
9706 
9707 		netdev->features |= NETIF_F_FSO |
9708 				    NETIF_F_FCOE_CRC;
9709 
9710 		netdev->vlan_features |= NETIF_F_FSO |
9711 					 NETIF_F_FCOE_CRC |
9712 					 NETIF_F_FCOE_MTU;
9713 	}
9714 #endif /* IXGBE_FCOE */
9715 
9716 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
9717 		netdev->hw_features |= NETIF_F_LRO;
9718 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9719 		netdev->features |= NETIF_F_LRO;
9720 
9721 	/* make sure the EEPROM is good */
9722 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
9723 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
9724 		err = -EIO;
9725 		goto err_sw_init;
9726 	}
9727 
9728 	eth_platform_get_mac_address(&adapter->pdev->dev,
9729 				     adapter->hw.mac.perm_addr);
9730 
9731 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
9732 
9733 	if (!is_valid_ether_addr(netdev->dev_addr)) {
9734 		e_dev_err("invalid MAC address\n");
9735 		err = -EIO;
9736 		goto err_sw_init;
9737 	}
9738 
9739 	/* Set hw->mac.addr to permanent MAC address */
9740 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
9741 	ixgbe_mac_set_default_filter(adapter);
9742 
9743 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
9744 		    (unsigned long) adapter);
9745 
9746 	if (ixgbe_removed(hw->hw_addr)) {
9747 		err = -EIO;
9748 		goto err_sw_init;
9749 	}
9750 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
9751 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
9752 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
9753 
9754 	err = ixgbe_init_interrupt_scheme(adapter);
9755 	if (err)
9756 		goto err_sw_init;
9757 
9758 	/* WOL not supported for all devices */
9759 	adapter->wol = 0;
9760 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
9761 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
9762 						pdev->subsystem_device);
9763 	if (hw->wol_enabled)
9764 		adapter->wol = IXGBE_WUFC_MAG;
9765 
9766 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
9767 
9768 	/* save off EEPROM version number */
9769 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
9770 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
9771 
9772 	/* pick up the PCI bus settings for reporting later */
9773 	if (ixgbe_pcie_from_parent(hw))
9774 		ixgbe_get_parent_bus_info(adapter);
9775 	else
9776 		 hw->mac.ops.get_bus_info(hw);
9777 
9778 	/* calculate the expected PCIe bandwidth required for optimal
9779 	 * performance. Note that some older parts will never have enough
9780 	 * bandwidth due to being older generation PCIe parts. We clamp these
9781 	 * parts to ensure no warning is displayed if it can't be fixed.
9782 	 */
9783 	switch (hw->mac.type) {
9784 	case ixgbe_mac_82598EB:
9785 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
9786 		break;
9787 	default:
9788 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
9789 		break;
9790 	}
9791 
9792 	/* don't check link if we failed to enumerate functions */
9793 	if (expected_gts > 0)
9794 		ixgbe_check_minimum_link(adapter, expected_gts);
9795 
9796 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
9797 	if (err)
9798 		strlcpy(part_str, "Unknown", sizeof(part_str));
9799 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
9800 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
9801 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
9802 			   part_str);
9803 	else
9804 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
9805 			   hw->mac.type, hw->phy.type, part_str);
9806 
9807 	e_dev_info("%pM\n", netdev->dev_addr);
9808 
9809 	/* reset the hardware with the new settings */
9810 	err = hw->mac.ops.start_hw(hw);
9811 	if (err == IXGBE_ERR_EEPROM_VERSION) {
9812 		/* We are running on a pre-production device, log a warning */
9813 		e_dev_warn("This device is a pre-production adapter/LOM. "
9814 			   "Please be aware there may be issues associated "
9815 			   "with your hardware.  If you are experiencing "
9816 			   "problems please contact your Intel or hardware "
9817 			   "representative who provided you with this "
9818 			   "hardware.\n");
9819 	}
9820 	strcpy(netdev->name, "eth%d");
9821 	err = register_netdev(netdev);
9822 	if (err)
9823 		goto err_register;
9824 
9825 	pci_set_drvdata(pdev, adapter);
9826 
9827 	/* power down the optics for 82599 SFP+ fiber */
9828 	if (hw->mac.ops.disable_tx_laser)
9829 		hw->mac.ops.disable_tx_laser(hw);
9830 
9831 	/* carrier off reporting is important to ethtool even BEFORE open */
9832 	netif_carrier_off(netdev);
9833 
9834 #ifdef CONFIG_IXGBE_DCA
9835 	if (dca_add_requester(&pdev->dev) == 0) {
9836 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
9837 		ixgbe_setup_dca(adapter);
9838 	}
9839 #endif
9840 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
9841 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
9842 		for (i = 0; i < adapter->num_vfs; i++)
9843 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
9844 	}
9845 
9846 	/* firmware requires driver version to be 0xFFFFFFFF
9847 	 * since os does not support feature
9848 	 */
9849 	if (hw->mac.ops.set_fw_drv_ver)
9850 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
9851 					   sizeof(ixgbe_driver_version) - 1,
9852 					   ixgbe_driver_version);
9853 
9854 	/* add san mac addr to netdev */
9855 	ixgbe_add_sanmac_netdev(netdev);
9856 
9857 	e_dev_info("%s\n", ixgbe_default_device_descr);
9858 
9859 #ifdef CONFIG_IXGBE_HWMON
9860 	if (ixgbe_sysfs_init(adapter))
9861 		e_err(probe, "failed to allocate sysfs resources\n");
9862 #endif /* CONFIG_IXGBE_HWMON */
9863 
9864 	ixgbe_dbg_adapter_init(adapter);
9865 
9866 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
9867 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
9868 		hw->mac.ops.setup_link(hw,
9869 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
9870 			true);
9871 
9872 	return 0;
9873 
9874 err_register:
9875 	ixgbe_release_hw_control(adapter);
9876 	ixgbe_clear_interrupt_scheme(adapter);
9877 err_sw_init:
9878 	ixgbe_disable_sriov(adapter);
9879 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
9880 	iounmap(adapter->io_addr);
9881 	kfree(adapter->jump_tables[0]);
9882 	kfree(adapter->mac_table);
9883 err_ioremap:
9884 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9885 	free_netdev(netdev);
9886 err_alloc_etherdev:
9887 	pci_release_mem_regions(pdev);
9888 err_pci_reg:
9889 err_dma:
9890 	if (!adapter || disable_dev)
9891 		pci_disable_device(pdev);
9892 	return err;
9893 }
9894 
9895 /**
9896  * ixgbe_remove - Device Removal Routine
9897  * @pdev: PCI device information struct
9898  *
9899  * ixgbe_remove is called by the PCI subsystem to alert the driver
9900  * that it should release a PCI device.  The could be caused by a
9901  * Hot-Plug event, or because the driver is going to be removed from
9902  * memory.
9903  **/
9904 static void ixgbe_remove(struct pci_dev *pdev)
9905 {
9906 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9907 	struct net_device *netdev;
9908 	bool disable_dev;
9909 	int i;
9910 
9911 	/* if !adapter then we already cleaned up in probe */
9912 	if (!adapter)
9913 		return;
9914 
9915 	netdev  = adapter->netdev;
9916 	ixgbe_dbg_adapter_exit(adapter);
9917 
9918 	set_bit(__IXGBE_REMOVING, &adapter->state);
9919 	cancel_work_sync(&adapter->service_task);
9920 
9921 
9922 #ifdef CONFIG_IXGBE_DCA
9923 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
9924 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
9925 		dca_remove_requester(&pdev->dev);
9926 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
9927 				IXGBE_DCA_CTRL_DCA_DISABLE);
9928 	}
9929 
9930 #endif
9931 #ifdef CONFIG_IXGBE_HWMON
9932 	ixgbe_sysfs_exit(adapter);
9933 #endif /* CONFIG_IXGBE_HWMON */
9934 
9935 	/* remove the added san mac */
9936 	ixgbe_del_sanmac_netdev(netdev);
9937 
9938 #ifdef CONFIG_PCI_IOV
9939 	ixgbe_disable_sriov(adapter);
9940 #endif
9941 	if (netdev->reg_state == NETREG_REGISTERED)
9942 		unregister_netdev(netdev);
9943 
9944 	ixgbe_clear_interrupt_scheme(adapter);
9945 
9946 	ixgbe_release_hw_control(adapter);
9947 
9948 #ifdef CONFIG_DCB
9949 	kfree(adapter->ixgbe_ieee_pfc);
9950 	kfree(adapter->ixgbe_ieee_ets);
9951 
9952 #endif
9953 	iounmap(adapter->io_addr);
9954 	pci_release_mem_regions(pdev);
9955 
9956 	e_dev_info("complete\n");
9957 
9958 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
9959 		if (adapter->jump_tables[i]) {
9960 			kfree(adapter->jump_tables[i]->input);
9961 			kfree(adapter->jump_tables[i]->mask);
9962 		}
9963 		kfree(adapter->jump_tables[i]);
9964 	}
9965 
9966 	kfree(adapter->mac_table);
9967 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9968 	free_netdev(netdev);
9969 
9970 	pci_disable_pcie_error_reporting(pdev);
9971 
9972 	if (disable_dev)
9973 		pci_disable_device(pdev);
9974 }
9975 
9976 /**
9977  * ixgbe_io_error_detected - called when PCI error is detected
9978  * @pdev: Pointer to PCI device
9979  * @state: The current pci connection state
9980  *
9981  * This function is called after a PCI bus error affecting
9982  * this device has been detected.
9983  */
9984 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
9985 						pci_channel_state_t state)
9986 {
9987 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9988 	struct net_device *netdev = adapter->netdev;
9989 
9990 #ifdef CONFIG_PCI_IOV
9991 	struct ixgbe_hw *hw = &adapter->hw;
9992 	struct pci_dev *bdev, *vfdev;
9993 	u32 dw0, dw1, dw2, dw3;
9994 	int vf, pos;
9995 	u16 req_id, pf_func;
9996 
9997 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
9998 	    adapter->num_vfs == 0)
9999 		goto skip_bad_vf_detection;
10000 
10001 	bdev = pdev->bus->self;
10002 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
10003 		bdev = bdev->bus->self;
10004 
10005 	if (!bdev)
10006 		goto skip_bad_vf_detection;
10007 
10008 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
10009 	if (!pos)
10010 		goto skip_bad_vf_detection;
10011 
10012 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
10013 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
10014 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
10015 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
10016 	if (ixgbe_removed(hw->hw_addr))
10017 		goto skip_bad_vf_detection;
10018 
10019 	req_id = dw1 >> 16;
10020 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
10021 	if (!(req_id & 0x0080))
10022 		goto skip_bad_vf_detection;
10023 
10024 	pf_func = req_id & 0x01;
10025 	if ((pf_func & 1) == (pdev->devfn & 1)) {
10026 		unsigned int device_id;
10027 
10028 		vf = (req_id & 0x7F) >> 1;
10029 		e_dev_err("VF %d has caused a PCIe error\n", vf);
10030 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
10031 				"%8.8x\tdw3: %8.8x\n",
10032 		dw0, dw1, dw2, dw3);
10033 		switch (adapter->hw.mac.type) {
10034 		case ixgbe_mac_82599EB:
10035 			device_id = IXGBE_82599_VF_DEVICE_ID;
10036 			break;
10037 		case ixgbe_mac_X540:
10038 			device_id = IXGBE_X540_VF_DEVICE_ID;
10039 			break;
10040 		case ixgbe_mac_X550:
10041 			device_id = IXGBE_DEV_ID_X550_VF;
10042 			break;
10043 		case ixgbe_mac_X550EM_x:
10044 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
10045 			break;
10046 		case ixgbe_mac_x550em_a:
10047 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
10048 			break;
10049 		default:
10050 			device_id = 0;
10051 			break;
10052 		}
10053 
10054 		/* Find the pci device of the offending VF */
10055 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
10056 		while (vfdev) {
10057 			if (vfdev->devfn == (req_id & 0xFF))
10058 				break;
10059 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
10060 					       device_id, vfdev);
10061 		}
10062 		/*
10063 		 * There's a slim chance the VF could have been hot plugged,
10064 		 * so if it is no longer present we don't need to issue the
10065 		 * VFLR.  Just clean up the AER in that case.
10066 		 */
10067 		if (vfdev) {
10068 			ixgbe_issue_vf_flr(adapter, vfdev);
10069 			/* Free device reference count */
10070 			pci_dev_put(vfdev);
10071 		}
10072 
10073 		pci_cleanup_aer_uncorrect_error_status(pdev);
10074 	}
10075 
10076 	/*
10077 	 * Even though the error may have occurred on the other port
10078 	 * we still need to increment the vf error reference count for
10079 	 * both ports because the I/O resume function will be called
10080 	 * for both of them.
10081 	 */
10082 	adapter->vferr_refcount++;
10083 
10084 	return PCI_ERS_RESULT_RECOVERED;
10085 
10086 skip_bad_vf_detection:
10087 #endif /* CONFIG_PCI_IOV */
10088 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
10089 		return PCI_ERS_RESULT_DISCONNECT;
10090 
10091 	rtnl_lock();
10092 	netif_device_detach(netdev);
10093 
10094 	if (state == pci_channel_io_perm_failure) {
10095 		rtnl_unlock();
10096 		return PCI_ERS_RESULT_DISCONNECT;
10097 	}
10098 
10099 	if (netif_running(netdev))
10100 		ixgbe_close_suspend(adapter);
10101 
10102 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
10103 		pci_disable_device(pdev);
10104 	rtnl_unlock();
10105 
10106 	/* Request a slot reset. */
10107 	return PCI_ERS_RESULT_NEED_RESET;
10108 }
10109 
10110 /**
10111  * ixgbe_io_slot_reset - called after the pci bus has been reset.
10112  * @pdev: Pointer to PCI device
10113  *
10114  * Restart the card from scratch, as if from a cold-boot.
10115  */
10116 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
10117 {
10118 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10119 	pci_ers_result_t result;
10120 	int err;
10121 
10122 	if (pci_enable_device_mem(pdev)) {
10123 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
10124 		result = PCI_ERS_RESULT_DISCONNECT;
10125 	} else {
10126 		smp_mb__before_atomic();
10127 		clear_bit(__IXGBE_DISABLED, &adapter->state);
10128 		adapter->hw.hw_addr = adapter->io_addr;
10129 		pci_set_master(pdev);
10130 		pci_restore_state(pdev);
10131 		pci_save_state(pdev);
10132 
10133 		pci_wake_from_d3(pdev, false);
10134 
10135 		ixgbe_reset(adapter);
10136 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10137 		result = PCI_ERS_RESULT_RECOVERED;
10138 	}
10139 
10140 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10141 	if (err) {
10142 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
10143 			  "failed 0x%0x\n", err);
10144 		/* non-fatal, continue */
10145 	}
10146 
10147 	return result;
10148 }
10149 
10150 /**
10151  * ixgbe_io_resume - called when traffic can start flowing again.
10152  * @pdev: Pointer to PCI device
10153  *
10154  * This callback is called when the error recovery driver tells us that
10155  * its OK to resume normal operation.
10156  */
10157 static void ixgbe_io_resume(struct pci_dev *pdev)
10158 {
10159 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10160 	struct net_device *netdev = adapter->netdev;
10161 
10162 #ifdef CONFIG_PCI_IOV
10163 	if (adapter->vferr_refcount) {
10164 		e_info(drv, "Resuming after VF err\n");
10165 		adapter->vferr_refcount--;
10166 		return;
10167 	}
10168 
10169 #endif
10170 	rtnl_lock();
10171 	if (netif_running(netdev))
10172 		ixgbe_open(netdev);
10173 
10174 	netif_device_attach(netdev);
10175 	rtnl_unlock();
10176 }
10177 
10178 static const struct pci_error_handlers ixgbe_err_handler = {
10179 	.error_detected = ixgbe_io_error_detected,
10180 	.slot_reset = ixgbe_io_slot_reset,
10181 	.resume = ixgbe_io_resume,
10182 };
10183 
10184 static struct pci_driver ixgbe_driver = {
10185 	.name     = ixgbe_driver_name,
10186 	.id_table = ixgbe_pci_tbl,
10187 	.probe    = ixgbe_probe,
10188 	.remove   = ixgbe_remove,
10189 #ifdef CONFIG_PM
10190 	.suspend  = ixgbe_suspend,
10191 	.resume   = ixgbe_resume,
10192 #endif
10193 	.shutdown = ixgbe_shutdown,
10194 	.sriov_configure = ixgbe_pci_sriov_configure,
10195 	.err_handler = &ixgbe_err_handler
10196 };
10197 
10198 /**
10199  * ixgbe_init_module - Driver Registration Routine
10200  *
10201  * ixgbe_init_module is the first routine called when the driver is
10202  * loaded. All it does is register with the PCI subsystem.
10203  **/
10204 static int __init ixgbe_init_module(void)
10205 {
10206 	int ret;
10207 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
10208 	pr_info("%s\n", ixgbe_copyright);
10209 
10210 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
10211 	if (!ixgbe_wq) {
10212 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
10213 		return -ENOMEM;
10214 	}
10215 
10216 	ixgbe_dbg_init();
10217 
10218 	ret = pci_register_driver(&ixgbe_driver);
10219 	if (ret) {
10220 		destroy_workqueue(ixgbe_wq);
10221 		ixgbe_dbg_exit();
10222 		return ret;
10223 	}
10224 
10225 #ifdef CONFIG_IXGBE_DCA
10226 	dca_register_notify(&dca_notifier);
10227 #endif
10228 
10229 	return 0;
10230 }
10231 
10232 module_init(ixgbe_init_module);
10233 
10234 /**
10235  * ixgbe_exit_module - Driver Exit Cleanup Routine
10236  *
10237  * ixgbe_exit_module is called just before the driver is removed
10238  * from memory.
10239  **/
10240 static void __exit ixgbe_exit_module(void)
10241 {
10242 #ifdef CONFIG_IXGBE_DCA
10243 	dca_unregister_notify(&dca_notifier);
10244 #endif
10245 	pci_unregister_driver(&ixgbe_driver);
10246 
10247 	ixgbe_dbg_exit();
10248 	if (ixgbe_wq) {
10249 		destroy_workqueue(ixgbe_wq);
10250 		ixgbe_wq = NULL;
10251 	}
10252 }
10253 
10254 #ifdef CONFIG_IXGBE_DCA
10255 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
10256 			    void *p)
10257 {
10258 	int ret_val;
10259 
10260 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
10261 					 __ixgbe_notify_dca);
10262 
10263 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
10264 }
10265 
10266 #endif /* CONFIG_IXGBE_DCA */
10267 
10268 module_exit(ixgbe_exit_module);
10269 
10270 /* ixgbe_main.c */
10271