xref: /openbmc/linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision e983940270f10fe8551baf0098be76ea478294a3)
1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2016 Intel Corporation.
5 
6   This program is free software; you can redistribute it and/or modify it
7   under the terms and conditions of the GNU General Public License,
8   version 2, as published by the Free Software Foundation.
9 
10   This program is distributed in the hope it will be useful, but WITHOUT
11   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13   more details.
14 
15   You should have received a copy of the GNU General Public License along with
16   this program; if not, write to the Free Software Foundation, Inc.,
17   51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 
19   The full GNU General Public License is included in this distribution in
20   the file called "COPYING".
21 
22   Contact Information:
23   Linux NICS <linux.nics@intel.com>
24   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 
27 *******************************************************************************/
28 
29 #include <linux/types.h>
30 #include <linux/module.h>
31 #include <linux/pci.h>
32 #include <linux/netdevice.h>
33 #include <linux/vmalloc.h>
34 #include <linux/string.h>
35 #include <linux/in.h>
36 #include <linux/interrupt.h>
37 #include <linux/ip.h>
38 #include <linux/tcp.h>
39 #include <linux/sctp.h>
40 #include <linux/pkt_sched.h>
41 #include <linux/ipv6.h>
42 #include <linux/slab.h>
43 #include <net/checksum.h>
44 #include <net/ip6_checksum.h>
45 #include <linux/etherdevice.h>
46 #include <linux/ethtool.h>
47 #include <linux/if.h>
48 #include <linux/if_vlan.h>
49 #include <linux/if_macvlan.h>
50 #include <linux/if_bridge.h>
51 #include <linux/prefetch.h>
52 #include <scsi/fc/fc_fcoe.h>
53 #include <net/udp_tunnel.h>
54 #include <net/pkt_cls.h>
55 #include <net/tc_act/tc_gact.h>
56 #include <net/tc_act/tc_mirred.h>
57 
58 #include "ixgbe.h"
59 #include "ixgbe_common.h"
60 #include "ixgbe_dcb_82599.h"
61 #include "ixgbe_sriov.h"
62 #include "ixgbe_model.h"
63 
64 char ixgbe_driver_name[] = "ixgbe";
65 static const char ixgbe_driver_string[] =
66 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
67 #ifdef IXGBE_FCOE
68 char ixgbe_default_device_descr[] =
69 			      "Intel(R) 10 Gigabit Network Connection";
70 #else
71 static char ixgbe_default_device_descr[] =
72 			      "Intel(R) 10 Gigabit Network Connection";
73 #endif
74 #define DRV_VERSION "4.4.0-k"
75 const char ixgbe_driver_version[] = DRV_VERSION;
76 static const char ixgbe_copyright[] =
77 				"Copyright (c) 1999-2016 Intel Corporation.";
78 
79 static const char ixgbe_overheat_msg[] = "Network adapter has been stopped because it has over heated. Restart the computer. If the problem persists, power off the system and replace the adapter";
80 
81 static const struct ixgbe_info *ixgbe_info_tbl[] = {
82 	[board_82598]		= &ixgbe_82598_info,
83 	[board_82599]		= &ixgbe_82599_info,
84 	[board_X540]		= &ixgbe_X540_info,
85 	[board_X550]		= &ixgbe_X550_info,
86 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
87 	[board_x550em_a]	= &ixgbe_x550em_a_info,
88 };
89 
90 /* ixgbe_pci_tbl - PCI Device ID Table
91  *
92  * Wildcard entries (PCI_ANY_ID) should come last
93  * Last entry must be all 0s
94  *
95  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
96  *   Class, Class Mask, private data (not used) }
97  */
98 static const struct pci_device_id ixgbe_pci_tbl[] = {
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
133 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
134 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
135 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
136 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
137 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
138 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
139 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
140 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
141 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
142 	/* required last entry */
143 	{0, }
144 };
145 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
146 
147 #ifdef CONFIG_IXGBE_DCA
148 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
149 			    void *p);
150 static struct notifier_block dca_notifier = {
151 	.notifier_call = ixgbe_notify_dca,
152 	.next          = NULL,
153 	.priority      = 0
154 };
155 #endif
156 
157 #ifdef CONFIG_PCI_IOV
158 static unsigned int max_vfs;
159 module_param(max_vfs, uint, 0);
160 MODULE_PARM_DESC(max_vfs,
161 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
162 #endif /* CONFIG_PCI_IOV */
163 
164 static unsigned int allow_unsupported_sfp;
165 module_param(allow_unsupported_sfp, uint, 0);
166 MODULE_PARM_DESC(allow_unsupported_sfp,
167 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
168 
169 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
170 static int debug = -1;
171 module_param(debug, int, 0);
172 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
173 
174 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
175 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
176 MODULE_LICENSE("GPL");
177 MODULE_VERSION(DRV_VERSION);
178 
179 static struct workqueue_struct *ixgbe_wq;
180 
181 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
182 
183 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
184 					  u32 reg, u16 *value)
185 {
186 	struct pci_dev *parent_dev;
187 	struct pci_bus *parent_bus;
188 
189 	parent_bus = adapter->pdev->bus->parent;
190 	if (!parent_bus)
191 		return -1;
192 
193 	parent_dev = parent_bus->self;
194 	if (!parent_dev)
195 		return -1;
196 
197 	if (!pci_is_pcie(parent_dev))
198 		return -1;
199 
200 	pcie_capability_read_word(parent_dev, reg, value);
201 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
202 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
203 		return -1;
204 	return 0;
205 }
206 
207 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
208 {
209 	struct ixgbe_hw *hw = &adapter->hw;
210 	u16 link_status = 0;
211 	int err;
212 
213 	hw->bus.type = ixgbe_bus_type_pci_express;
214 
215 	/* Get the negotiated link width and speed from PCI config space of the
216 	 * parent, as this device is behind a switch
217 	 */
218 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
219 
220 	/* assume caller will handle error case */
221 	if (err)
222 		return err;
223 
224 	hw->bus.width = ixgbe_convert_bus_width(link_status);
225 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
226 
227 	return 0;
228 }
229 
230 /**
231  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
232  * @hw: hw specific details
233  *
234  * This function is used by probe to determine whether a device's PCI-Express
235  * bandwidth details should be gathered from the parent bus instead of from the
236  * device. Used to ensure that various locations all have the correct device ID
237  * checks.
238  */
239 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
240 {
241 	switch (hw->device_id) {
242 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
243 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
244 		return true;
245 	default:
246 		return false;
247 	}
248 }
249 
250 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
251 				     int expected_gts)
252 {
253 	struct ixgbe_hw *hw = &adapter->hw;
254 	int max_gts = 0;
255 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
256 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
257 	struct pci_dev *pdev;
258 
259 	/* Some devices are not connected over PCIe and thus do not negotiate
260 	 * speed. These devices do not have valid bus info, and thus any report
261 	 * we generate may not be correct.
262 	 */
263 	if (hw->bus.type == ixgbe_bus_type_internal)
264 		return;
265 
266 	/* determine whether to use the parent device */
267 	if (ixgbe_pcie_from_parent(&adapter->hw))
268 		pdev = adapter->pdev->bus->parent->self;
269 	else
270 		pdev = adapter->pdev;
271 
272 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
273 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
274 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
275 		return;
276 	}
277 
278 	switch (speed) {
279 	case PCIE_SPEED_2_5GT:
280 		/* 8b/10b encoding reduces max throughput by 20% */
281 		max_gts = 2 * width;
282 		break;
283 	case PCIE_SPEED_5_0GT:
284 		/* 8b/10b encoding reduces max throughput by 20% */
285 		max_gts = 4 * width;
286 		break;
287 	case PCIE_SPEED_8_0GT:
288 		/* 128b/130b encoding reduces throughput by less than 2% */
289 		max_gts = 8 * width;
290 		break;
291 	default:
292 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
293 		return;
294 	}
295 
296 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
297 		   max_gts);
298 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
299 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
300 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
301 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
302 		    "Unknown"),
303 		   width,
304 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
305 		    speed == PCIE_SPEED_5_0GT ? "20%" :
306 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
307 		    "Unknown"));
308 
309 	if (max_gts < expected_gts) {
310 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
311 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
312 			expected_gts);
313 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
314 	}
315 }
316 
317 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
318 {
319 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
320 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
321 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
322 		queue_work(ixgbe_wq, &adapter->service_task);
323 }
324 
325 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
326 {
327 	struct ixgbe_adapter *adapter = hw->back;
328 
329 	if (!hw->hw_addr)
330 		return;
331 	hw->hw_addr = NULL;
332 	e_dev_err("Adapter removed\n");
333 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
334 		ixgbe_service_event_schedule(adapter);
335 }
336 
337 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
338 {
339 	u32 value;
340 
341 	/* The following check not only optimizes a bit by not
342 	 * performing a read on the status register when the
343 	 * register just read was a status register read that
344 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
345 	 * potential recursion.
346 	 */
347 	if (reg == IXGBE_STATUS) {
348 		ixgbe_remove_adapter(hw);
349 		return;
350 	}
351 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
352 	if (value == IXGBE_FAILED_READ_REG)
353 		ixgbe_remove_adapter(hw);
354 }
355 
356 /**
357  * ixgbe_read_reg - Read from device register
358  * @hw: hw specific details
359  * @reg: offset of register to read
360  *
361  * Returns : value read or IXGBE_FAILED_READ_REG if removed
362  *
363  * This function is used to read device registers. It checks for device
364  * removal by confirming any read that returns all ones by checking the
365  * status register value for all ones. This function avoids reading from
366  * the hardware if a removal was previously detected in which case it
367  * returns IXGBE_FAILED_READ_REG (all ones).
368  */
369 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
370 {
371 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
372 	u32 value;
373 
374 	if (ixgbe_removed(reg_addr))
375 		return IXGBE_FAILED_READ_REG;
376 	if (unlikely(hw->phy.nw_mng_if_sel &
377 		     IXGBE_NW_MNG_IF_SEL_ENABLE_10_100M)) {
378 		struct ixgbe_adapter *adapter;
379 		int i;
380 
381 		for (i = 0; i < 200; ++i) {
382 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
383 			if (likely(!value))
384 				goto writes_completed;
385 			if (value == IXGBE_FAILED_READ_REG) {
386 				ixgbe_remove_adapter(hw);
387 				return IXGBE_FAILED_READ_REG;
388 			}
389 			udelay(5);
390 		}
391 
392 		adapter = hw->back;
393 		e_warn(hw, "register writes incomplete %08x\n", value);
394 	}
395 
396 writes_completed:
397 	value = readl(reg_addr + reg);
398 	if (unlikely(value == IXGBE_FAILED_READ_REG))
399 		ixgbe_check_remove(hw, reg);
400 	return value;
401 }
402 
403 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
404 {
405 	u16 value;
406 
407 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
408 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
409 		ixgbe_remove_adapter(hw);
410 		return true;
411 	}
412 	return false;
413 }
414 
415 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
416 {
417 	struct ixgbe_adapter *adapter = hw->back;
418 	u16 value;
419 
420 	if (ixgbe_removed(hw->hw_addr))
421 		return IXGBE_FAILED_READ_CFG_WORD;
422 	pci_read_config_word(adapter->pdev, reg, &value);
423 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
424 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
425 		return IXGBE_FAILED_READ_CFG_WORD;
426 	return value;
427 }
428 
429 #ifdef CONFIG_PCI_IOV
430 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
431 {
432 	struct ixgbe_adapter *adapter = hw->back;
433 	u32 value;
434 
435 	if (ixgbe_removed(hw->hw_addr))
436 		return IXGBE_FAILED_READ_CFG_DWORD;
437 	pci_read_config_dword(adapter->pdev, reg, &value);
438 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
439 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
440 		return IXGBE_FAILED_READ_CFG_DWORD;
441 	return value;
442 }
443 #endif /* CONFIG_PCI_IOV */
444 
445 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
446 {
447 	struct ixgbe_adapter *adapter = hw->back;
448 
449 	if (ixgbe_removed(hw->hw_addr))
450 		return;
451 	pci_write_config_word(adapter->pdev, reg, value);
452 }
453 
454 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
455 {
456 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
457 
458 	/* flush memory to make sure state is correct before next watchdog */
459 	smp_mb__before_atomic();
460 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
461 }
462 
463 struct ixgbe_reg_info {
464 	u32 ofs;
465 	char *name;
466 };
467 
468 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
469 
470 	/* General Registers */
471 	{IXGBE_CTRL, "CTRL"},
472 	{IXGBE_STATUS, "STATUS"},
473 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
474 
475 	/* Interrupt Registers */
476 	{IXGBE_EICR, "EICR"},
477 
478 	/* RX Registers */
479 	{IXGBE_SRRCTL(0), "SRRCTL"},
480 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
481 	{IXGBE_RDLEN(0), "RDLEN"},
482 	{IXGBE_RDH(0), "RDH"},
483 	{IXGBE_RDT(0), "RDT"},
484 	{IXGBE_RXDCTL(0), "RXDCTL"},
485 	{IXGBE_RDBAL(0), "RDBAL"},
486 	{IXGBE_RDBAH(0), "RDBAH"},
487 
488 	/* TX Registers */
489 	{IXGBE_TDBAL(0), "TDBAL"},
490 	{IXGBE_TDBAH(0), "TDBAH"},
491 	{IXGBE_TDLEN(0), "TDLEN"},
492 	{IXGBE_TDH(0), "TDH"},
493 	{IXGBE_TDT(0), "TDT"},
494 	{IXGBE_TXDCTL(0), "TXDCTL"},
495 
496 	/* List Terminator */
497 	{ .name = NULL }
498 };
499 
500 
501 /*
502  * ixgbe_regdump - register printout routine
503  */
504 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
505 {
506 	int i = 0, j = 0;
507 	char rname[16];
508 	u32 regs[64];
509 
510 	switch (reginfo->ofs) {
511 	case IXGBE_SRRCTL(0):
512 		for (i = 0; i < 64; i++)
513 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
514 		break;
515 	case IXGBE_DCA_RXCTRL(0):
516 		for (i = 0; i < 64; i++)
517 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
518 		break;
519 	case IXGBE_RDLEN(0):
520 		for (i = 0; i < 64; i++)
521 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
522 		break;
523 	case IXGBE_RDH(0):
524 		for (i = 0; i < 64; i++)
525 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
526 		break;
527 	case IXGBE_RDT(0):
528 		for (i = 0; i < 64; i++)
529 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
530 		break;
531 	case IXGBE_RXDCTL(0):
532 		for (i = 0; i < 64; i++)
533 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
534 		break;
535 	case IXGBE_RDBAL(0):
536 		for (i = 0; i < 64; i++)
537 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
538 		break;
539 	case IXGBE_RDBAH(0):
540 		for (i = 0; i < 64; i++)
541 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
542 		break;
543 	case IXGBE_TDBAL(0):
544 		for (i = 0; i < 64; i++)
545 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
546 		break;
547 	case IXGBE_TDBAH(0):
548 		for (i = 0; i < 64; i++)
549 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
550 		break;
551 	case IXGBE_TDLEN(0):
552 		for (i = 0; i < 64; i++)
553 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
554 		break;
555 	case IXGBE_TDH(0):
556 		for (i = 0; i < 64; i++)
557 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
558 		break;
559 	case IXGBE_TDT(0):
560 		for (i = 0; i < 64; i++)
561 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
562 		break;
563 	case IXGBE_TXDCTL(0):
564 		for (i = 0; i < 64; i++)
565 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
566 		break;
567 	default:
568 		pr_info("%-15s %08x\n", reginfo->name,
569 			IXGBE_READ_REG(hw, reginfo->ofs));
570 		return;
571 	}
572 
573 	for (i = 0; i < 8; i++) {
574 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
575 		pr_err("%-15s", rname);
576 		for (j = 0; j < 8; j++)
577 			pr_cont(" %08x", regs[i*8+j]);
578 		pr_cont("\n");
579 	}
580 
581 }
582 
583 /*
584  * ixgbe_dump - Print registers, tx-rings and rx-rings
585  */
586 static void ixgbe_dump(struct ixgbe_adapter *adapter)
587 {
588 	struct net_device *netdev = adapter->netdev;
589 	struct ixgbe_hw *hw = &adapter->hw;
590 	struct ixgbe_reg_info *reginfo;
591 	int n = 0;
592 	struct ixgbe_ring *tx_ring;
593 	struct ixgbe_tx_buffer *tx_buffer;
594 	union ixgbe_adv_tx_desc *tx_desc;
595 	struct my_u0 { u64 a; u64 b; } *u0;
596 	struct ixgbe_ring *rx_ring;
597 	union ixgbe_adv_rx_desc *rx_desc;
598 	struct ixgbe_rx_buffer *rx_buffer_info;
599 	u32 staterr;
600 	int i = 0;
601 
602 	if (!netif_msg_hw(adapter))
603 		return;
604 
605 	/* Print netdevice Info */
606 	if (netdev) {
607 		dev_info(&adapter->pdev->dev, "Net device Info\n");
608 		pr_info("Device Name     state            "
609 			"trans_start      last_rx\n");
610 		pr_info("%-15s %016lX %016lX %016lX\n",
611 			netdev->name,
612 			netdev->state,
613 			dev_trans_start(netdev),
614 			netdev->last_rx);
615 	}
616 
617 	/* Print Registers */
618 	dev_info(&adapter->pdev->dev, "Register Dump\n");
619 	pr_info(" Register Name   Value\n");
620 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
621 	     reginfo->name; reginfo++) {
622 		ixgbe_regdump(hw, reginfo);
623 	}
624 
625 	/* Print TX Ring Summary */
626 	if (!netdev || !netif_running(netdev))
627 		return;
628 
629 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
630 	pr_info(" %s     %s              %s        %s\n",
631 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
632 		"leng", "ntw", "timestamp");
633 	for (n = 0; n < adapter->num_tx_queues; n++) {
634 		tx_ring = adapter->tx_ring[n];
635 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
636 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
637 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
638 			   (u64)dma_unmap_addr(tx_buffer, dma),
639 			   dma_unmap_len(tx_buffer, len),
640 			   tx_buffer->next_to_watch,
641 			   (u64)tx_buffer->time_stamp);
642 	}
643 
644 	/* Print TX Rings */
645 	if (!netif_msg_tx_done(adapter))
646 		goto rx_ring_summary;
647 
648 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
649 
650 	/* Transmit Descriptor Formats
651 	 *
652 	 * 82598 Advanced Transmit Descriptor
653 	 *   +--------------------------------------------------------------+
654 	 * 0 |         Buffer Address [63:0]                                |
655 	 *   +--------------------------------------------------------------+
656 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
657 	 *   +--------------------------------------------------------------+
658 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
659 	 *
660 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
661 	 *   +--------------------------------------------------------------+
662 	 * 0 |                          RSV [63:0]                          |
663 	 *   +--------------------------------------------------------------+
664 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
665 	 *   +--------------------------------------------------------------+
666 	 *   63                       36 35   32 31                         0
667 	 *
668 	 * 82599+ Advanced Transmit Descriptor
669 	 *   +--------------------------------------------------------------+
670 	 * 0 |         Buffer Address [63:0]                                |
671 	 *   +--------------------------------------------------------------+
672 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
673 	 *   +--------------------------------------------------------------+
674 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
675 	 *
676 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
677 	 *   +--------------------------------------------------------------+
678 	 * 0 |                          RSV [63:0]                          |
679 	 *   +--------------------------------------------------------------+
680 	 * 8 |            RSV           |  STA  |           RSV             |
681 	 *   +--------------------------------------------------------------+
682 	 *   63                       36 35   32 31                         0
683 	 */
684 
685 	for (n = 0; n < adapter->num_tx_queues; n++) {
686 		tx_ring = adapter->tx_ring[n];
687 		pr_info("------------------------------------\n");
688 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
689 		pr_info("------------------------------------\n");
690 		pr_info("%s%s    %s              %s        %s          %s\n",
691 			"T [desc]     [address 63:0  ] ",
692 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
693 			"leng", "ntw", "timestamp", "bi->skb");
694 
695 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
696 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
697 			tx_buffer = &tx_ring->tx_buffer_info[i];
698 			u0 = (struct my_u0 *)tx_desc;
699 			if (dma_unmap_len(tx_buffer, len) > 0) {
700 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
701 					i,
702 					le64_to_cpu(u0->a),
703 					le64_to_cpu(u0->b),
704 					(u64)dma_unmap_addr(tx_buffer, dma),
705 					dma_unmap_len(tx_buffer, len),
706 					tx_buffer->next_to_watch,
707 					(u64)tx_buffer->time_stamp,
708 					tx_buffer->skb);
709 				if (i == tx_ring->next_to_use &&
710 					i == tx_ring->next_to_clean)
711 					pr_cont(" NTC/U\n");
712 				else if (i == tx_ring->next_to_use)
713 					pr_cont(" NTU\n");
714 				else if (i == tx_ring->next_to_clean)
715 					pr_cont(" NTC\n");
716 				else
717 					pr_cont("\n");
718 
719 				if (netif_msg_pktdata(adapter) &&
720 				    tx_buffer->skb)
721 					print_hex_dump(KERN_INFO, "",
722 						DUMP_PREFIX_ADDRESS, 16, 1,
723 						tx_buffer->skb->data,
724 						dma_unmap_len(tx_buffer, len),
725 						true);
726 			}
727 		}
728 	}
729 
730 	/* Print RX Rings Summary */
731 rx_ring_summary:
732 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
733 	pr_info("Queue [NTU] [NTC]\n");
734 	for (n = 0; n < adapter->num_rx_queues; n++) {
735 		rx_ring = adapter->rx_ring[n];
736 		pr_info("%5d %5X %5X\n",
737 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
738 	}
739 
740 	/* Print RX Rings */
741 	if (!netif_msg_rx_status(adapter))
742 		return;
743 
744 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
745 
746 	/* Receive Descriptor Formats
747 	 *
748 	 * 82598 Advanced Receive Descriptor (Read) Format
749 	 *    63                                           1        0
750 	 *    +-----------------------------------------------------+
751 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
752 	 *    +----------------------------------------------+------+
753 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
754 	 *    +-----------------------------------------------------+
755 	 *
756 	 *
757 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
758 	 *
759 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
760 	 *   +------------------------------------------------------+
761 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
762 	 *   | Packet   | IP     |   |          |     | Type | Type |
763 	 *   | Checksum | Ident  |   |          |     |      |      |
764 	 *   +------------------------------------------------------+
765 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
766 	 *   +------------------------------------------------------+
767 	 *   63       48 47    32 31            20 19               0
768 	 *
769 	 * 82599+ Advanced Receive Descriptor (Read) Format
770 	 *    63                                           1        0
771 	 *    +-----------------------------------------------------+
772 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
773 	 *    +----------------------------------------------+------+
774 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
775 	 *    +-----------------------------------------------------+
776 	 *
777 	 *
778 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
779 	 *
780 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
781 	 *   +------------------------------------------------------+
782 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
783 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
784 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
785 	 *   +------------------------------------------------------+
786 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
787 	 *   +------------------------------------------------------+
788 	 *   63       48 47    32 31          20 19                 0
789 	 */
790 
791 	for (n = 0; n < adapter->num_rx_queues; n++) {
792 		rx_ring = adapter->rx_ring[n];
793 		pr_info("------------------------------------\n");
794 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
795 		pr_info("------------------------------------\n");
796 		pr_info("%s%s%s",
797 			"R  [desc]      [ PktBuf     A0] ",
798 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
799 			"<-- Adv Rx Read format\n");
800 		pr_info("%s%s%s",
801 			"RWB[desc]      [PcsmIpSHl PtRs] ",
802 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
803 			"<-- Adv Rx Write-Back format\n");
804 
805 		for (i = 0; i < rx_ring->count; i++) {
806 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
807 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
808 			u0 = (struct my_u0 *)rx_desc;
809 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
810 			if (staterr & IXGBE_RXD_STAT_DD) {
811 				/* Descriptor Done */
812 				pr_info("RWB[0x%03X]     %016llX "
813 					"%016llX ---------------- %p", i,
814 					le64_to_cpu(u0->a),
815 					le64_to_cpu(u0->b),
816 					rx_buffer_info->skb);
817 			} else {
818 				pr_info("R  [0x%03X]     %016llX "
819 					"%016llX %016llX %p", i,
820 					le64_to_cpu(u0->a),
821 					le64_to_cpu(u0->b),
822 					(u64)rx_buffer_info->dma,
823 					rx_buffer_info->skb);
824 
825 				if (netif_msg_pktdata(adapter) &&
826 				    rx_buffer_info->dma) {
827 					print_hex_dump(KERN_INFO, "",
828 					   DUMP_PREFIX_ADDRESS, 16, 1,
829 					   page_address(rx_buffer_info->page) +
830 						    rx_buffer_info->page_offset,
831 					   ixgbe_rx_bufsz(rx_ring), true);
832 				}
833 			}
834 
835 			if (i == rx_ring->next_to_use)
836 				pr_cont(" NTU\n");
837 			else if (i == rx_ring->next_to_clean)
838 				pr_cont(" NTC\n");
839 			else
840 				pr_cont("\n");
841 
842 		}
843 	}
844 }
845 
846 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
847 {
848 	u32 ctrl_ext;
849 
850 	/* Let firmware take over control of h/w */
851 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
852 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
853 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
854 }
855 
856 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
857 {
858 	u32 ctrl_ext;
859 
860 	/* Let firmware know the driver has taken over */
861 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
862 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
863 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
864 }
865 
866 /**
867  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
868  * @adapter: pointer to adapter struct
869  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
870  * @queue: queue to map the corresponding interrupt to
871  * @msix_vector: the vector to map to the corresponding queue
872  *
873  */
874 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
875 			   u8 queue, u8 msix_vector)
876 {
877 	u32 ivar, index;
878 	struct ixgbe_hw *hw = &adapter->hw;
879 	switch (hw->mac.type) {
880 	case ixgbe_mac_82598EB:
881 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
882 		if (direction == -1)
883 			direction = 0;
884 		index = (((direction * 64) + queue) >> 2) & 0x1F;
885 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
886 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
887 		ivar |= (msix_vector << (8 * (queue & 0x3)));
888 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
889 		break;
890 	case ixgbe_mac_82599EB:
891 	case ixgbe_mac_X540:
892 	case ixgbe_mac_X550:
893 	case ixgbe_mac_X550EM_x:
894 	case ixgbe_mac_x550em_a:
895 		if (direction == -1) {
896 			/* other causes */
897 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
898 			index = ((queue & 1) * 8);
899 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
900 			ivar &= ~(0xFF << index);
901 			ivar |= (msix_vector << index);
902 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
903 			break;
904 		} else {
905 			/* tx or rx causes */
906 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
907 			index = ((16 * (queue & 1)) + (8 * direction));
908 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
909 			ivar &= ~(0xFF << index);
910 			ivar |= (msix_vector << index);
911 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
912 			break;
913 		}
914 	default:
915 		break;
916 	}
917 }
918 
919 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
920 					  u64 qmask)
921 {
922 	u32 mask;
923 
924 	switch (adapter->hw.mac.type) {
925 	case ixgbe_mac_82598EB:
926 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
927 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
928 		break;
929 	case ixgbe_mac_82599EB:
930 	case ixgbe_mac_X540:
931 	case ixgbe_mac_X550:
932 	case ixgbe_mac_X550EM_x:
933 	case ixgbe_mac_x550em_a:
934 		mask = (qmask & 0xFFFFFFFF);
935 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
936 		mask = (qmask >> 32);
937 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
938 		break;
939 	default:
940 		break;
941 	}
942 }
943 
944 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
945 				      struct ixgbe_tx_buffer *tx_buffer)
946 {
947 	if (tx_buffer->skb) {
948 		dev_kfree_skb_any(tx_buffer->skb);
949 		if (dma_unmap_len(tx_buffer, len))
950 			dma_unmap_single(ring->dev,
951 					 dma_unmap_addr(tx_buffer, dma),
952 					 dma_unmap_len(tx_buffer, len),
953 					 DMA_TO_DEVICE);
954 	} else if (dma_unmap_len(tx_buffer, len)) {
955 		dma_unmap_page(ring->dev,
956 			       dma_unmap_addr(tx_buffer, dma),
957 			       dma_unmap_len(tx_buffer, len),
958 			       DMA_TO_DEVICE);
959 	}
960 	tx_buffer->next_to_watch = NULL;
961 	tx_buffer->skb = NULL;
962 	dma_unmap_len_set(tx_buffer, len, 0);
963 	/* tx_buffer must be completely set up in the transmit path */
964 }
965 
966 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
967 {
968 	struct ixgbe_hw *hw = &adapter->hw;
969 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
970 	int i;
971 	u32 data;
972 
973 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
974 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
975 		return;
976 
977 	switch (hw->mac.type) {
978 	case ixgbe_mac_82598EB:
979 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
980 		break;
981 	default:
982 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
983 	}
984 	hwstats->lxoffrxc += data;
985 
986 	/* refill credits (no tx hang) if we received xoff */
987 	if (!data)
988 		return;
989 
990 	for (i = 0; i < adapter->num_tx_queues; i++)
991 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
992 			  &adapter->tx_ring[i]->state);
993 }
994 
995 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
996 {
997 	struct ixgbe_hw *hw = &adapter->hw;
998 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
999 	u32 xoff[8] = {0};
1000 	u8 tc;
1001 	int i;
1002 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
1003 
1004 	if (adapter->ixgbe_ieee_pfc)
1005 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
1006 
1007 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
1008 		ixgbe_update_xoff_rx_lfc(adapter);
1009 		return;
1010 	}
1011 
1012 	/* update stats for each tc, only valid with PFC enabled */
1013 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
1014 		u32 pxoffrxc;
1015 
1016 		switch (hw->mac.type) {
1017 		case ixgbe_mac_82598EB:
1018 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1019 			break;
1020 		default:
1021 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1022 		}
1023 		hwstats->pxoffrxc[i] += pxoffrxc;
1024 		/* Get the TC for given UP */
1025 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1026 		xoff[tc] += pxoffrxc;
1027 	}
1028 
1029 	/* disarm tx queues that have received xoff frames */
1030 	for (i = 0; i < adapter->num_tx_queues; i++) {
1031 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1032 
1033 		tc = tx_ring->dcb_tc;
1034 		if (xoff[tc])
1035 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1036 	}
1037 }
1038 
1039 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1040 {
1041 	return ring->stats.packets;
1042 }
1043 
1044 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1045 {
1046 	struct ixgbe_adapter *adapter;
1047 	struct ixgbe_hw *hw;
1048 	u32 head, tail;
1049 
1050 	if (ring->l2_accel_priv)
1051 		adapter = ring->l2_accel_priv->real_adapter;
1052 	else
1053 		adapter = netdev_priv(ring->netdev);
1054 
1055 	hw = &adapter->hw;
1056 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1057 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1058 
1059 	if (head != tail)
1060 		return (head < tail) ?
1061 			tail - head : (tail + ring->count - head);
1062 
1063 	return 0;
1064 }
1065 
1066 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1067 {
1068 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1069 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1070 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1071 
1072 	clear_check_for_tx_hang(tx_ring);
1073 
1074 	/*
1075 	 * Check for a hung queue, but be thorough. This verifies
1076 	 * that a transmit has been completed since the previous
1077 	 * check AND there is at least one packet pending. The
1078 	 * ARMED bit is set to indicate a potential hang. The
1079 	 * bit is cleared if a pause frame is received to remove
1080 	 * false hang detection due to PFC or 802.3x frames. By
1081 	 * requiring this to fail twice we avoid races with
1082 	 * pfc clearing the ARMED bit and conditions where we
1083 	 * run the check_tx_hang logic with a transmit completion
1084 	 * pending but without time to complete it yet.
1085 	 */
1086 	if (tx_done_old == tx_done && tx_pending)
1087 		/* make sure it is true for two checks in a row */
1088 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1089 					&tx_ring->state);
1090 	/* update completed stats and continue */
1091 	tx_ring->tx_stats.tx_done_old = tx_done;
1092 	/* reset the countdown */
1093 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1094 
1095 	return false;
1096 }
1097 
1098 /**
1099  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1100  * @adapter: driver private struct
1101  **/
1102 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1103 {
1104 
1105 	/* Do the reset outside of interrupt context */
1106 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1107 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1108 		e_warn(drv, "initiating reset due to tx timeout\n");
1109 		ixgbe_service_event_schedule(adapter);
1110 	}
1111 }
1112 
1113 /**
1114  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1115  **/
1116 static int ixgbe_tx_maxrate(struct net_device *netdev,
1117 			    int queue_index, u32 maxrate)
1118 {
1119 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1120 	struct ixgbe_hw *hw = &adapter->hw;
1121 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1122 
1123 	if (!maxrate)
1124 		return 0;
1125 
1126 	/* Calculate the rate factor values to set */
1127 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1128 	bcnrc_val /= maxrate;
1129 
1130 	/* clear everything but the rate factor */
1131 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1132 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1133 
1134 	/* enable the rate scheduler */
1135 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1136 
1137 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1138 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1139 
1140 	return 0;
1141 }
1142 
1143 /**
1144  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1145  * @q_vector: structure containing interrupt and ring information
1146  * @tx_ring: tx ring to clean
1147  * @napi_budget: Used to determine if we are in netpoll
1148  **/
1149 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1150 			       struct ixgbe_ring *tx_ring, int napi_budget)
1151 {
1152 	struct ixgbe_adapter *adapter = q_vector->adapter;
1153 	struct ixgbe_tx_buffer *tx_buffer;
1154 	union ixgbe_adv_tx_desc *tx_desc;
1155 	unsigned int total_bytes = 0, total_packets = 0;
1156 	unsigned int budget = q_vector->tx.work_limit;
1157 	unsigned int i = tx_ring->next_to_clean;
1158 
1159 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1160 		return true;
1161 
1162 	tx_buffer = &tx_ring->tx_buffer_info[i];
1163 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1164 	i -= tx_ring->count;
1165 
1166 	do {
1167 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1168 
1169 		/* if next_to_watch is not set then there is no work pending */
1170 		if (!eop_desc)
1171 			break;
1172 
1173 		/* prevent any other reads prior to eop_desc */
1174 		read_barrier_depends();
1175 
1176 		/* if DD is not set pending work has not been completed */
1177 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1178 			break;
1179 
1180 		/* clear next_to_watch to prevent false hangs */
1181 		tx_buffer->next_to_watch = NULL;
1182 
1183 		/* update the statistics for this packet */
1184 		total_bytes += tx_buffer->bytecount;
1185 		total_packets += tx_buffer->gso_segs;
1186 
1187 		/* free the skb */
1188 		napi_consume_skb(tx_buffer->skb, napi_budget);
1189 
1190 		/* unmap skb header data */
1191 		dma_unmap_single(tx_ring->dev,
1192 				 dma_unmap_addr(tx_buffer, dma),
1193 				 dma_unmap_len(tx_buffer, len),
1194 				 DMA_TO_DEVICE);
1195 
1196 		/* clear tx_buffer data */
1197 		tx_buffer->skb = NULL;
1198 		dma_unmap_len_set(tx_buffer, len, 0);
1199 
1200 		/* unmap remaining buffers */
1201 		while (tx_desc != eop_desc) {
1202 			tx_buffer++;
1203 			tx_desc++;
1204 			i++;
1205 			if (unlikely(!i)) {
1206 				i -= tx_ring->count;
1207 				tx_buffer = tx_ring->tx_buffer_info;
1208 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1209 			}
1210 
1211 			/* unmap any remaining paged data */
1212 			if (dma_unmap_len(tx_buffer, len)) {
1213 				dma_unmap_page(tx_ring->dev,
1214 					       dma_unmap_addr(tx_buffer, dma),
1215 					       dma_unmap_len(tx_buffer, len),
1216 					       DMA_TO_DEVICE);
1217 				dma_unmap_len_set(tx_buffer, len, 0);
1218 			}
1219 		}
1220 
1221 		/* move us one more past the eop_desc for start of next pkt */
1222 		tx_buffer++;
1223 		tx_desc++;
1224 		i++;
1225 		if (unlikely(!i)) {
1226 			i -= tx_ring->count;
1227 			tx_buffer = tx_ring->tx_buffer_info;
1228 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1229 		}
1230 
1231 		/* issue prefetch for next Tx descriptor */
1232 		prefetch(tx_desc);
1233 
1234 		/* update budget accounting */
1235 		budget--;
1236 	} while (likely(budget));
1237 
1238 	i += tx_ring->count;
1239 	tx_ring->next_to_clean = i;
1240 	u64_stats_update_begin(&tx_ring->syncp);
1241 	tx_ring->stats.bytes += total_bytes;
1242 	tx_ring->stats.packets += total_packets;
1243 	u64_stats_update_end(&tx_ring->syncp);
1244 	q_vector->tx.total_bytes += total_bytes;
1245 	q_vector->tx.total_packets += total_packets;
1246 
1247 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1248 		/* schedule immediate reset if we believe we hung */
1249 		struct ixgbe_hw *hw = &adapter->hw;
1250 		e_err(drv, "Detected Tx Unit Hang\n"
1251 			"  Tx Queue             <%d>\n"
1252 			"  TDH, TDT             <%x>, <%x>\n"
1253 			"  next_to_use          <%x>\n"
1254 			"  next_to_clean        <%x>\n"
1255 			"tx_buffer_info[next_to_clean]\n"
1256 			"  time_stamp           <%lx>\n"
1257 			"  jiffies              <%lx>\n",
1258 			tx_ring->queue_index,
1259 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1260 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1261 			tx_ring->next_to_use, i,
1262 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1263 
1264 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1265 
1266 		e_info(probe,
1267 		       "tx hang %d detected on queue %d, resetting adapter\n",
1268 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1269 
1270 		/* schedule immediate reset if we believe we hung */
1271 		ixgbe_tx_timeout_reset(adapter);
1272 
1273 		/* the adapter is about to reset, no point in enabling stuff */
1274 		return true;
1275 	}
1276 
1277 	netdev_tx_completed_queue(txring_txq(tx_ring),
1278 				  total_packets, total_bytes);
1279 
1280 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1281 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1282 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1283 		/* Make sure that anybody stopping the queue after this
1284 		 * sees the new next_to_clean.
1285 		 */
1286 		smp_mb();
1287 		if (__netif_subqueue_stopped(tx_ring->netdev,
1288 					     tx_ring->queue_index)
1289 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1290 			netif_wake_subqueue(tx_ring->netdev,
1291 					    tx_ring->queue_index);
1292 			++tx_ring->tx_stats.restart_queue;
1293 		}
1294 	}
1295 
1296 	return !!budget;
1297 }
1298 
1299 #ifdef CONFIG_IXGBE_DCA
1300 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1301 				struct ixgbe_ring *tx_ring,
1302 				int cpu)
1303 {
1304 	struct ixgbe_hw *hw = &adapter->hw;
1305 	u32 txctrl = 0;
1306 	u16 reg_offset;
1307 
1308 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1309 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1310 
1311 	switch (hw->mac.type) {
1312 	case ixgbe_mac_82598EB:
1313 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1314 		break;
1315 	case ixgbe_mac_82599EB:
1316 	case ixgbe_mac_X540:
1317 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1318 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1319 		break;
1320 	default:
1321 		/* for unknown hardware do not write register */
1322 		return;
1323 	}
1324 
1325 	/*
1326 	 * We can enable relaxed ordering for reads, but not writes when
1327 	 * DCA is enabled.  This is due to a known issue in some chipsets
1328 	 * which will cause the DCA tag to be cleared.
1329 	 */
1330 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1331 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1332 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1333 
1334 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1335 }
1336 
1337 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1338 				struct ixgbe_ring *rx_ring,
1339 				int cpu)
1340 {
1341 	struct ixgbe_hw *hw = &adapter->hw;
1342 	u32 rxctrl = 0;
1343 	u8 reg_idx = rx_ring->reg_idx;
1344 
1345 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1346 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1347 
1348 	switch (hw->mac.type) {
1349 	case ixgbe_mac_82599EB:
1350 	case ixgbe_mac_X540:
1351 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1352 		break;
1353 	default:
1354 		break;
1355 	}
1356 
1357 	/*
1358 	 * We can enable relaxed ordering for reads, but not writes when
1359 	 * DCA is enabled.  This is due to a known issue in some chipsets
1360 	 * which will cause the DCA tag to be cleared.
1361 	 */
1362 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1363 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1364 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1365 
1366 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1367 }
1368 
1369 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1370 {
1371 	struct ixgbe_adapter *adapter = q_vector->adapter;
1372 	struct ixgbe_ring *ring;
1373 	int cpu = get_cpu();
1374 
1375 	if (q_vector->cpu == cpu)
1376 		goto out_no_update;
1377 
1378 	ixgbe_for_each_ring(ring, q_vector->tx)
1379 		ixgbe_update_tx_dca(adapter, ring, cpu);
1380 
1381 	ixgbe_for_each_ring(ring, q_vector->rx)
1382 		ixgbe_update_rx_dca(adapter, ring, cpu);
1383 
1384 	q_vector->cpu = cpu;
1385 out_no_update:
1386 	put_cpu();
1387 }
1388 
1389 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1390 {
1391 	int i;
1392 
1393 	/* always use CB2 mode, difference is masked in the CB driver */
1394 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1395 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1396 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1397 	else
1398 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1399 				IXGBE_DCA_CTRL_DCA_DISABLE);
1400 
1401 	for (i = 0; i < adapter->num_q_vectors; i++) {
1402 		adapter->q_vector[i]->cpu = -1;
1403 		ixgbe_update_dca(adapter->q_vector[i]);
1404 	}
1405 }
1406 
1407 static int __ixgbe_notify_dca(struct device *dev, void *data)
1408 {
1409 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1410 	unsigned long event = *(unsigned long *)data;
1411 
1412 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1413 		return 0;
1414 
1415 	switch (event) {
1416 	case DCA_PROVIDER_ADD:
1417 		/* if we're already enabled, don't do it again */
1418 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1419 			break;
1420 		if (dca_add_requester(dev) == 0) {
1421 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1422 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1423 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1424 			break;
1425 		}
1426 		/* Fall Through since DCA is disabled. */
1427 	case DCA_PROVIDER_REMOVE:
1428 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1429 			dca_remove_requester(dev);
1430 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1431 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1432 					IXGBE_DCA_CTRL_DCA_DISABLE);
1433 		}
1434 		break;
1435 	}
1436 
1437 	return 0;
1438 }
1439 
1440 #endif /* CONFIG_IXGBE_DCA */
1441 
1442 #define IXGBE_RSS_L4_TYPES_MASK \
1443 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1444 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1445 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1446 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1447 
1448 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1449 				 union ixgbe_adv_rx_desc *rx_desc,
1450 				 struct sk_buff *skb)
1451 {
1452 	u16 rss_type;
1453 
1454 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1455 		return;
1456 
1457 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1458 		   IXGBE_RXDADV_RSSTYPE_MASK;
1459 
1460 	if (!rss_type)
1461 		return;
1462 
1463 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1464 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1465 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1466 }
1467 
1468 #ifdef IXGBE_FCOE
1469 /**
1470  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1471  * @ring: structure containing ring specific data
1472  * @rx_desc: advanced rx descriptor
1473  *
1474  * Returns : true if it is FCoE pkt
1475  */
1476 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1477 				    union ixgbe_adv_rx_desc *rx_desc)
1478 {
1479 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1480 
1481 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1482 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1483 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1484 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1485 }
1486 
1487 #endif /* IXGBE_FCOE */
1488 /**
1489  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1490  * @ring: structure containing ring specific data
1491  * @rx_desc: current Rx descriptor being processed
1492  * @skb: skb currently being received and modified
1493  **/
1494 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1495 				     union ixgbe_adv_rx_desc *rx_desc,
1496 				     struct sk_buff *skb)
1497 {
1498 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1499 	bool encap_pkt = false;
1500 
1501 	skb_checksum_none_assert(skb);
1502 
1503 	/* Rx csum disabled */
1504 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1505 		return;
1506 
1507 	/* check for VXLAN and Geneve packets */
1508 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1509 		encap_pkt = true;
1510 		skb->encapsulation = 1;
1511 	}
1512 
1513 	/* if IP and error */
1514 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1515 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1516 		ring->rx_stats.csum_err++;
1517 		return;
1518 	}
1519 
1520 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1521 		return;
1522 
1523 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1524 		/*
1525 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1526 		 * checksum errors.
1527 		 */
1528 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1529 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1530 			return;
1531 
1532 		ring->rx_stats.csum_err++;
1533 		return;
1534 	}
1535 
1536 	/* It must be a TCP or UDP packet with a valid checksum */
1537 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1538 	if (encap_pkt) {
1539 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1540 			return;
1541 
1542 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1543 			skb->ip_summed = CHECKSUM_NONE;
1544 			return;
1545 		}
1546 		/* If we checked the outer header let the stack know */
1547 		skb->csum_level = 1;
1548 	}
1549 }
1550 
1551 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1552 				    struct ixgbe_rx_buffer *bi)
1553 {
1554 	struct page *page = bi->page;
1555 	dma_addr_t dma;
1556 
1557 	/* since we are recycling buffers we should seldom need to alloc */
1558 	if (likely(page))
1559 		return true;
1560 
1561 	/* alloc new page for storage */
1562 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1563 	if (unlikely(!page)) {
1564 		rx_ring->rx_stats.alloc_rx_page_failed++;
1565 		return false;
1566 	}
1567 
1568 	/* map page for use */
1569 	dma = dma_map_page(rx_ring->dev, page, 0,
1570 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1571 
1572 	/*
1573 	 * if mapping failed free memory back to system since
1574 	 * there isn't much point in holding memory we can't use
1575 	 */
1576 	if (dma_mapping_error(rx_ring->dev, dma)) {
1577 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1578 
1579 		rx_ring->rx_stats.alloc_rx_page_failed++;
1580 		return false;
1581 	}
1582 
1583 	bi->dma = dma;
1584 	bi->page = page;
1585 	bi->page_offset = 0;
1586 
1587 	return true;
1588 }
1589 
1590 /**
1591  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1592  * @rx_ring: ring to place buffers on
1593  * @cleaned_count: number of buffers to replace
1594  **/
1595 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1596 {
1597 	union ixgbe_adv_rx_desc *rx_desc;
1598 	struct ixgbe_rx_buffer *bi;
1599 	u16 i = rx_ring->next_to_use;
1600 
1601 	/* nothing to do */
1602 	if (!cleaned_count)
1603 		return;
1604 
1605 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1606 	bi = &rx_ring->rx_buffer_info[i];
1607 	i -= rx_ring->count;
1608 
1609 	do {
1610 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1611 			break;
1612 
1613 		/*
1614 		 * Refresh the desc even if buffer_addrs didn't change
1615 		 * because each write-back erases this info.
1616 		 */
1617 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1618 
1619 		rx_desc++;
1620 		bi++;
1621 		i++;
1622 		if (unlikely(!i)) {
1623 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1624 			bi = rx_ring->rx_buffer_info;
1625 			i -= rx_ring->count;
1626 		}
1627 
1628 		/* clear the status bits for the next_to_use descriptor */
1629 		rx_desc->wb.upper.status_error = 0;
1630 
1631 		cleaned_count--;
1632 	} while (cleaned_count);
1633 
1634 	i += rx_ring->count;
1635 
1636 	if (rx_ring->next_to_use != i) {
1637 		rx_ring->next_to_use = i;
1638 
1639 		/* update next to alloc since we have filled the ring */
1640 		rx_ring->next_to_alloc = i;
1641 
1642 		/* Force memory writes to complete before letting h/w
1643 		 * know there are new descriptors to fetch.  (Only
1644 		 * applicable for weak-ordered memory model archs,
1645 		 * such as IA-64).
1646 		 */
1647 		wmb();
1648 		writel(i, rx_ring->tail);
1649 	}
1650 }
1651 
1652 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1653 				   struct sk_buff *skb)
1654 {
1655 	u16 hdr_len = skb_headlen(skb);
1656 
1657 	/* set gso_size to avoid messing up TCP MSS */
1658 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1659 						 IXGBE_CB(skb)->append_cnt);
1660 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1661 }
1662 
1663 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1664 				   struct sk_buff *skb)
1665 {
1666 	/* if append_cnt is 0 then frame is not RSC */
1667 	if (!IXGBE_CB(skb)->append_cnt)
1668 		return;
1669 
1670 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1671 	rx_ring->rx_stats.rsc_flush++;
1672 
1673 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1674 
1675 	/* gso_size is computed using append_cnt so always clear it last */
1676 	IXGBE_CB(skb)->append_cnt = 0;
1677 }
1678 
1679 /**
1680  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1681  * @rx_ring: rx descriptor ring packet is being transacted on
1682  * @rx_desc: pointer to the EOP Rx descriptor
1683  * @skb: pointer to current skb being populated
1684  *
1685  * This function checks the ring, descriptor, and packet information in
1686  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1687  * other fields within the skb.
1688  **/
1689 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1690 				     union ixgbe_adv_rx_desc *rx_desc,
1691 				     struct sk_buff *skb)
1692 {
1693 	struct net_device *dev = rx_ring->netdev;
1694 	u32 flags = rx_ring->q_vector->adapter->flags;
1695 
1696 	ixgbe_update_rsc_stats(rx_ring, skb);
1697 
1698 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1699 
1700 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1701 
1702 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1703 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1704 
1705 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1706 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1707 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1708 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1709 	}
1710 
1711 	skb_record_rx_queue(skb, rx_ring->queue_index);
1712 
1713 	skb->protocol = eth_type_trans(skb, dev);
1714 }
1715 
1716 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1717 			 struct sk_buff *skb)
1718 {
1719 	skb_mark_napi_id(skb, &q_vector->napi);
1720 	if (ixgbe_qv_busy_polling(q_vector))
1721 		netif_receive_skb(skb);
1722 	else
1723 		napi_gro_receive(&q_vector->napi, skb);
1724 }
1725 
1726 /**
1727  * ixgbe_is_non_eop - process handling of non-EOP buffers
1728  * @rx_ring: Rx ring being processed
1729  * @rx_desc: Rx descriptor for current buffer
1730  * @skb: Current socket buffer containing buffer in progress
1731  *
1732  * This function updates next to clean.  If the buffer is an EOP buffer
1733  * this function exits returning false, otherwise it will place the
1734  * sk_buff in the next buffer to be chained and return true indicating
1735  * that this is in fact a non-EOP buffer.
1736  **/
1737 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1738 			     union ixgbe_adv_rx_desc *rx_desc,
1739 			     struct sk_buff *skb)
1740 {
1741 	u32 ntc = rx_ring->next_to_clean + 1;
1742 
1743 	/* fetch, update, and store next to clean */
1744 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1745 	rx_ring->next_to_clean = ntc;
1746 
1747 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1748 
1749 	/* update RSC append count if present */
1750 	if (ring_is_rsc_enabled(rx_ring)) {
1751 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1752 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1753 
1754 		if (unlikely(rsc_enabled)) {
1755 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1756 
1757 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1758 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1759 
1760 			/* update ntc based on RSC value */
1761 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1762 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1763 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1764 		}
1765 	}
1766 
1767 	/* if we are the last buffer then there is nothing else to do */
1768 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1769 		return false;
1770 
1771 	/* place skb in next buffer to be received */
1772 	rx_ring->rx_buffer_info[ntc].skb = skb;
1773 	rx_ring->rx_stats.non_eop_descs++;
1774 
1775 	return true;
1776 }
1777 
1778 /**
1779  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1780  * @rx_ring: rx descriptor ring packet is being transacted on
1781  * @skb: pointer to current skb being adjusted
1782  *
1783  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1784  * main difference between this version and the original function is that
1785  * this function can make several assumptions about the state of things
1786  * that allow for significant optimizations versus the standard function.
1787  * As a result we can do things like drop a frag and maintain an accurate
1788  * truesize for the skb.
1789  */
1790 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1791 			    struct sk_buff *skb)
1792 {
1793 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1794 	unsigned char *va;
1795 	unsigned int pull_len;
1796 
1797 	/*
1798 	 * it is valid to use page_address instead of kmap since we are
1799 	 * working with pages allocated out of the lomem pool per
1800 	 * alloc_page(GFP_ATOMIC)
1801 	 */
1802 	va = skb_frag_address(frag);
1803 
1804 	/*
1805 	 * we need the header to contain the greater of either ETH_HLEN or
1806 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1807 	 */
1808 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1809 
1810 	/* align pull length to size of long to optimize memcpy performance */
1811 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1812 
1813 	/* update all of the pointers */
1814 	skb_frag_size_sub(frag, pull_len);
1815 	frag->page_offset += pull_len;
1816 	skb->data_len -= pull_len;
1817 	skb->tail += pull_len;
1818 }
1819 
1820 /**
1821  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1822  * @rx_ring: rx descriptor ring packet is being transacted on
1823  * @skb: pointer to current skb being updated
1824  *
1825  * This function provides a basic DMA sync up for the first fragment of an
1826  * skb.  The reason for doing this is that the first fragment cannot be
1827  * unmapped until we have reached the end of packet descriptor for a buffer
1828  * chain.
1829  */
1830 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1831 				struct sk_buff *skb)
1832 {
1833 	/* if the page was released unmap it, else just sync our portion */
1834 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1835 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1836 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1837 		IXGBE_CB(skb)->page_released = false;
1838 	} else {
1839 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1840 
1841 		dma_sync_single_range_for_cpu(rx_ring->dev,
1842 					      IXGBE_CB(skb)->dma,
1843 					      frag->page_offset,
1844 					      ixgbe_rx_bufsz(rx_ring),
1845 					      DMA_FROM_DEVICE);
1846 	}
1847 	IXGBE_CB(skb)->dma = 0;
1848 }
1849 
1850 /**
1851  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1852  * @rx_ring: rx descriptor ring packet is being transacted on
1853  * @rx_desc: pointer to the EOP Rx descriptor
1854  * @skb: pointer to current skb being fixed
1855  *
1856  * Check for corrupted packet headers caused by senders on the local L2
1857  * embedded NIC switch not setting up their Tx Descriptors right.  These
1858  * should be very rare.
1859  *
1860  * Also address the case where we are pulling data in on pages only
1861  * and as such no data is present in the skb header.
1862  *
1863  * In addition if skb is not at least 60 bytes we need to pad it so that
1864  * it is large enough to qualify as a valid Ethernet frame.
1865  *
1866  * Returns true if an error was encountered and skb was freed.
1867  **/
1868 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1869 				  union ixgbe_adv_rx_desc *rx_desc,
1870 				  struct sk_buff *skb)
1871 {
1872 	struct net_device *netdev = rx_ring->netdev;
1873 
1874 	/* verify that the packet does not have any known errors */
1875 	if (unlikely(ixgbe_test_staterr(rx_desc,
1876 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1877 	    !(netdev->features & NETIF_F_RXALL))) {
1878 		dev_kfree_skb_any(skb);
1879 		return true;
1880 	}
1881 
1882 	/* place header in linear portion of buffer */
1883 	if (skb_is_nonlinear(skb))
1884 		ixgbe_pull_tail(rx_ring, skb);
1885 
1886 #ifdef IXGBE_FCOE
1887 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1888 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1889 		return false;
1890 
1891 #endif
1892 	/* if eth_skb_pad returns an error the skb was freed */
1893 	if (eth_skb_pad(skb))
1894 		return true;
1895 
1896 	return false;
1897 }
1898 
1899 /**
1900  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1901  * @rx_ring: rx descriptor ring to store buffers on
1902  * @old_buff: donor buffer to have page reused
1903  *
1904  * Synchronizes page for reuse by the adapter
1905  **/
1906 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1907 				struct ixgbe_rx_buffer *old_buff)
1908 {
1909 	struct ixgbe_rx_buffer *new_buff;
1910 	u16 nta = rx_ring->next_to_alloc;
1911 
1912 	new_buff = &rx_ring->rx_buffer_info[nta];
1913 
1914 	/* update, and store next to alloc */
1915 	nta++;
1916 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1917 
1918 	/* transfer page from old buffer to new buffer */
1919 	*new_buff = *old_buff;
1920 
1921 	/* sync the buffer for use by the device */
1922 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1923 					 new_buff->page_offset,
1924 					 ixgbe_rx_bufsz(rx_ring),
1925 					 DMA_FROM_DEVICE);
1926 }
1927 
1928 static inline bool ixgbe_page_is_reserved(struct page *page)
1929 {
1930 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
1931 }
1932 
1933 /**
1934  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1935  * @rx_ring: rx descriptor ring to transact packets on
1936  * @rx_buffer: buffer containing page to add
1937  * @rx_desc: descriptor containing length of buffer written by hardware
1938  * @skb: sk_buff to place the data into
1939  *
1940  * This function will add the data contained in rx_buffer->page to the skb.
1941  * This is done either through a direct copy if the data in the buffer is
1942  * less than the skb header size, otherwise it will just attach the page as
1943  * a frag to the skb.
1944  *
1945  * The function will then update the page offset if necessary and return
1946  * true if the buffer can be reused by the adapter.
1947  **/
1948 static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1949 			      struct ixgbe_rx_buffer *rx_buffer,
1950 			      union ixgbe_adv_rx_desc *rx_desc,
1951 			      struct sk_buff *skb)
1952 {
1953 	struct page *page = rx_buffer->page;
1954 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1955 #if (PAGE_SIZE < 8192)
1956 	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1957 #else
1958 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1959 	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1960 				   ixgbe_rx_bufsz(rx_ring);
1961 #endif
1962 
1963 	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1964 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1965 
1966 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1967 
1968 		/* page is not reserved, we can reuse buffer as-is */
1969 		if (likely(!ixgbe_page_is_reserved(page)))
1970 			return true;
1971 
1972 		/* this page cannot be reused so discard it */
1973 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1974 		return false;
1975 	}
1976 
1977 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1978 			rx_buffer->page_offset, size, truesize);
1979 
1980 	/* avoid re-using remote pages */
1981 	if (unlikely(ixgbe_page_is_reserved(page)))
1982 		return false;
1983 
1984 #if (PAGE_SIZE < 8192)
1985 	/* if we are only owner of page we can reuse it */
1986 	if (unlikely(page_count(page) != 1))
1987 		return false;
1988 
1989 	/* flip page offset to other buffer */
1990 	rx_buffer->page_offset ^= truesize;
1991 #else
1992 	/* move offset up to the next cache line */
1993 	rx_buffer->page_offset += truesize;
1994 
1995 	if (rx_buffer->page_offset > last_offset)
1996 		return false;
1997 #endif
1998 
1999 	/* Even if we own the page, we are not allowed to use atomic_set()
2000 	 * This would break get_page_unless_zero() users.
2001 	 */
2002 	page_ref_inc(page);
2003 
2004 	return true;
2005 }
2006 
2007 static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
2008 					     union ixgbe_adv_rx_desc *rx_desc)
2009 {
2010 	struct ixgbe_rx_buffer *rx_buffer;
2011 	struct sk_buff *skb;
2012 	struct page *page;
2013 
2014 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2015 	page = rx_buffer->page;
2016 	prefetchw(page);
2017 
2018 	skb = rx_buffer->skb;
2019 
2020 	if (likely(!skb)) {
2021 		void *page_addr = page_address(page) +
2022 				  rx_buffer->page_offset;
2023 
2024 		/* prefetch first cache line of first page */
2025 		prefetch(page_addr);
2026 #if L1_CACHE_BYTES < 128
2027 		prefetch(page_addr + L1_CACHE_BYTES);
2028 #endif
2029 
2030 		/* allocate a skb to store the frags */
2031 		skb = napi_alloc_skb(&rx_ring->q_vector->napi,
2032 				     IXGBE_RX_HDR_SIZE);
2033 		if (unlikely(!skb)) {
2034 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2035 			return NULL;
2036 		}
2037 
2038 		/*
2039 		 * we will be copying header into skb->data in
2040 		 * pskb_may_pull so it is in our interest to prefetch
2041 		 * it now to avoid a possible cache miss
2042 		 */
2043 		prefetchw(skb->data);
2044 
2045 		/*
2046 		 * Delay unmapping of the first packet. It carries the
2047 		 * header information, HW may still access the header
2048 		 * after the writeback.  Only unmap it when EOP is
2049 		 * reached
2050 		 */
2051 		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
2052 			goto dma_sync;
2053 
2054 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2055 	} else {
2056 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2057 			ixgbe_dma_sync_frag(rx_ring, skb);
2058 
2059 dma_sync:
2060 		/* we are reusing so sync this buffer for CPU use */
2061 		dma_sync_single_range_for_cpu(rx_ring->dev,
2062 					      rx_buffer->dma,
2063 					      rx_buffer->page_offset,
2064 					      ixgbe_rx_bufsz(rx_ring),
2065 					      DMA_FROM_DEVICE);
2066 
2067 		rx_buffer->skb = NULL;
2068 	}
2069 
2070 	/* pull page into skb */
2071 	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
2072 		/* hand second half of page back to the ring */
2073 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2074 	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
2075 		/* the page has been released from the ring */
2076 		IXGBE_CB(skb)->page_released = true;
2077 	} else {
2078 		/* we are not reusing the buffer so unmap it */
2079 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
2080 			       ixgbe_rx_pg_size(rx_ring),
2081 			       DMA_FROM_DEVICE);
2082 	}
2083 
2084 	/* clear contents of buffer_info */
2085 	rx_buffer->page = NULL;
2086 
2087 	return skb;
2088 }
2089 
2090 /**
2091  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2092  * @q_vector: structure containing interrupt and ring information
2093  * @rx_ring: rx descriptor ring to transact packets on
2094  * @budget: Total limit on number of packets to process
2095  *
2096  * This function provides a "bounce buffer" approach to Rx interrupt
2097  * processing.  The advantage to this is that on systems that have
2098  * expensive overhead for IOMMU access this provides a means of avoiding
2099  * it by maintaining the mapping of the page to the syste.
2100  *
2101  * Returns amount of work completed
2102  **/
2103 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2104 			       struct ixgbe_ring *rx_ring,
2105 			       const int budget)
2106 {
2107 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2108 #ifdef IXGBE_FCOE
2109 	struct ixgbe_adapter *adapter = q_vector->adapter;
2110 	int ddp_bytes;
2111 	unsigned int mss = 0;
2112 #endif /* IXGBE_FCOE */
2113 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2114 
2115 	while (likely(total_rx_packets < budget)) {
2116 		union ixgbe_adv_rx_desc *rx_desc;
2117 		struct sk_buff *skb;
2118 
2119 		/* return some buffers to hardware, one at a time is too slow */
2120 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2121 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2122 			cleaned_count = 0;
2123 		}
2124 
2125 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2126 
2127 		if (!rx_desc->wb.upper.status_error)
2128 			break;
2129 
2130 		/* This memory barrier is needed to keep us from reading
2131 		 * any other fields out of the rx_desc until we know the
2132 		 * descriptor has been written back
2133 		 */
2134 		dma_rmb();
2135 
2136 		/* retrieve a buffer from the ring */
2137 		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
2138 
2139 		/* exit if we failed to retrieve a buffer */
2140 		if (!skb)
2141 			break;
2142 
2143 		cleaned_count++;
2144 
2145 		/* place incomplete frames back on ring for completion */
2146 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2147 			continue;
2148 
2149 		/* verify the packet layout is correct */
2150 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2151 			continue;
2152 
2153 		/* probably a little skewed due to removing CRC */
2154 		total_rx_bytes += skb->len;
2155 
2156 		/* populate checksum, timestamp, VLAN, and protocol */
2157 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2158 
2159 #ifdef IXGBE_FCOE
2160 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2161 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2162 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2163 			/* include DDPed FCoE data */
2164 			if (ddp_bytes > 0) {
2165 				if (!mss) {
2166 					mss = rx_ring->netdev->mtu -
2167 						sizeof(struct fcoe_hdr) -
2168 						sizeof(struct fc_frame_header) -
2169 						sizeof(struct fcoe_crc_eof);
2170 					if (mss > 512)
2171 						mss &= ~511;
2172 				}
2173 				total_rx_bytes += ddp_bytes;
2174 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2175 								 mss);
2176 			}
2177 			if (!ddp_bytes) {
2178 				dev_kfree_skb_any(skb);
2179 				continue;
2180 			}
2181 		}
2182 
2183 #endif /* IXGBE_FCOE */
2184 		ixgbe_rx_skb(q_vector, skb);
2185 
2186 		/* update budget accounting */
2187 		total_rx_packets++;
2188 	}
2189 
2190 	u64_stats_update_begin(&rx_ring->syncp);
2191 	rx_ring->stats.packets += total_rx_packets;
2192 	rx_ring->stats.bytes += total_rx_bytes;
2193 	u64_stats_update_end(&rx_ring->syncp);
2194 	q_vector->rx.total_packets += total_rx_packets;
2195 	q_vector->rx.total_bytes += total_rx_bytes;
2196 
2197 	return total_rx_packets;
2198 }
2199 
2200 #ifdef CONFIG_NET_RX_BUSY_POLL
2201 /* must be called with local_bh_disable()d */
2202 static int ixgbe_low_latency_recv(struct napi_struct *napi)
2203 {
2204 	struct ixgbe_q_vector *q_vector =
2205 			container_of(napi, struct ixgbe_q_vector, napi);
2206 	struct ixgbe_adapter *adapter = q_vector->adapter;
2207 	struct ixgbe_ring  *ring;
2208 	int found = 0;
2209 
2210 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2211 		return LL_FLUSH_FAILED;
2212 
2213 	if (!ixgbe_qv_lock_poll(q_vector))
2214 		return LL_FLUSH_BUSY;
2215 
2216 	ixgbe_for_each_ring(ring, q_vector->rx) {
2217 		found = ixgbe_clean_rx_irq(q_vector, ring, 4);
2218 #ifdef BP_EXTENDED_STATS
2219 		if (found)
2220 			ring->stats.cleaned += found;
2221 		else
2222 			ring->stats.misses++;
2223 #endif
2224 		if (found)
2225 			break;
2226 	}
2227 
2228 	ixgbe_qv_unlock_poll(q_vector);
2229 
2230 	return found;
2231 }
2232 #endif	/* CONFIG_NET_RX_BUSY_POLL */
2233 
2234 /**
2235  * ixgbe_configure_msix - Configure MSI-X hardware
2236  * @adapter: board private structure
2237  *
2238  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2239  * interrupts.
2240  **/
2241 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2242 {
2243 	struct ixgbe_q_vector *q_vector;
2244 	int v_idx;
2245 	u32 mask;
2246 
2247 	/* Populate MSIX to EITR Select */
2248 	if (adapter->num_vfs > 32) {
2249 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2250 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2251 	}
2252 
2253 	/*
2254 	 * Populate the IVAR table and set the ITR values to the
2255 	 * corresponding register.
2256 	 */
2257 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2258 		struct ixgbe_ring *ring;
2259 		q_vector = adapter->q_vector[v_idx];
2260 
2261 		ixgbe_for_each_ring(ring, q_vector->rx)
2262 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2263 
2264 		ixgbe_for_each_ring(ring, q_vector->tx)
2265 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2266 
2267 		ixgbe_write_eitr(q_vector);
2268 	}
2269 
2270 	switch (adapter->hw.mac.type) {
2271 	case ixgbe_mac_82598EB:
2272 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2273 			       v_idx);
2274 		break;
2275 	case ixgbe_mac_82599EB:
2276 	case ixgbe_mac_X540:
2277 	case ixgbe_mac_X550:
2278 	case ixgbe_mac_X550EM_x:
2279 	case ixgbe_mac_x550em_a:
2280 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2281 		break;
2282 	default:
2283 		break;
2284 	}
2285 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2286 
2287 	/* set up to autoclear timer, and the vectors */
2288 	mask = IXGBE_EIMS_ENABLE_MASK;
2289 	mask &= ~(IXGBE_EIMS_OTHER |
2290 		  IXGBE_EIMS_MAILBOX |
2291 		  IXGBE_EIMS_LSC);
2292 
2293 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2294 }
2295 
2296 enum latency_range {
2297 	lowest_latency = 0,
2298 	low_latency = 1,
2299 	bulk_latency = 2,
2300 	latency_invalid = 255
2301 };
2302 
2303 /**
2304  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2305  * @q_vector: structure containing interrupt and ring information
2306  * @ring_container: structure containing ring performance data
2307  *
2308  *      Stores a new ITR value based on packets and byte
2309  *      counts during the last interrupt.  The advantage of per interrupt
2310  *      computation is faster updates and more accurate ITR for the current
2311  *      traffic pattern.  Constants in this function were computed
2312  *      based on theoretical maximum wire speed and thresholds were set based
2313  *      on testing data as well as attempting to minimize response time
2314  *      while increasing bulk throughput.
2315  *      this functionality is controlled by the InterruptThrottleRate module
2316  *      parameter (see ixgbe_param.c)
2317  **/
2318 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2319 			     struct ixgbe_ring_container *ring_container)
2320 {
2321 	int bytes = ring_container->total_bytes;
2322 	int packets = ring_container->total_packets;
2323 	u32 timepassed_us;
2324 	u64 bytes_perint;
2325 	u8 itr_setting = ring_container->itr;
2326 
2327 	if (packets == 0)
2328 		return;
2329 
2330 	/* simple throttlerate management
2331 	 *   0-10MB/s   lowest (100000 ints/s)
2332 	 *  10-20MB/s   low    (20000 ints/s)
2333 	 *  20-1249MB/s bulk   (12000 ints/s)
2334 	 */
2335 	/* what was last interrupt timeslice? */
2336 	timepassed_us = q_vector->itr >> 2;
2337 	if (timepassed_us == 0)
2338 		return;
2339 
2340 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2341 
2342 	switch (itr_setting) {
2343 	case lowest_latency:
2344 		if (bytes_perint > 10)
2345 			itr_setting = low_latency;
2346 		break;
2347 	case low_latency:
2348 		if (bytes_perint > 20)
2349 			itr_setting = bulk_latency;
2350 		else if (bytes_perint <= 10)
2351 			itr_setting = lowest_latency;
2352 		break;
2353 	case bulk_latency:
2354 		if (bytes_perint <= 20)
2355 			itr_setting = low_latency;
2356 		break;
2357 	}
2358 
2359 	/* clear work counters since we have the values we need */
2360 	ring_container->total_bytes = 0;
2361 	ring_container->total_packets = 0;
2362 
2363 	/* write updated itr to ring container */
2364 	ring_container->itr = itr_setting;
2365 }
2366 
2367 /**
2368  * ixgbe_write_eitr - write EITR register in hardware specific way
2369  * @q_vector: structure containing interrupt and ring information
2370  *
2371  * This function is made to be called by ethtool and by the driver
2372  * when it needs to update EITR registers at runtime.  Hardware
2373  * specific quirks/differences are taken care of here.
2374  */
2375 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2376 {
2377 	struct ixgbe_adapter *adapter = q_vector->adapter;
2378 	struct ixgbe_hw *hw = &adapter->hw;
2379 	int v_idx = q_vector->v_idx;
2380 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2381 
2382 	switch (adapter->hw.mac.type) {
2383 	case ixgbe_mac_82598EB:
2384 		/* must write high and low 16 bits to reset counter */
2385 		itr_reg |= (itr_reg << 16);
2386 		break;
2387 	case ixgbe_mac_82599EB:
2388 	case ixgbe_mac_X540:
2389 	case ixgbe_mac_X550:
2390 	case ixgbe_mac_X550EM_x:
2391 	case ixgbe_mac_x550em_a:
2392 		/*
2393 		 * set the WDIS bit to not clear the timer bits and cause an
2394 		 * immediate assertion of the interrupt
2395 		 */
2396 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2397 		break;
2398 	default:
2399 		break;
2400 	}
2401 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2402 }
2403 
2404 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2405 {
2406 	u32 new_itr = q_vector->itr;
2407 	u8 current_itr;
2408 
2409 	ixgbe_update_itr(q_vector, &q_vector->tx);
2410 	ixgbe_update_itr(q_vector, &q_vector->rx);
2411 
2412 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2413 
2414 	switch (current_itr) {
2415 	/* counts and packets in update_itr are dependent on these numbers */
2416 	case lowest_latency:
2417 		new_itr = IXGBE_100K_ITR;
2418 		break;
2419 	case low_latency:
2420 		new_itr = IXGBE_20K_ITR;
2421 		break;
2422 	case bulk_latency:
2423 		new_itr = IXGBE_12K_ITR;
2424 		break;
2425 	default:
2426 		break;
2427 	}
2428 
2429 	if (new_itr != q_vector->itr) {
2430 		/* do an exponential smoothing */
2431 		new_itr = (10 * new_itr * q_vector->itr) /
2432 			  ((9 * new_itr) + q_vector->itr);
2433 
2434 		/* save the algorithm value here */
2435 		q_vector->itr = new_itr;
2436 
2437 		ixgbe_write_eitr(q_vector);
2438 	}
2439 }
2440 
2441 /**
2442  * ixgbe_check_overtemp_subtask - check for over temperature
2443  * @adapter: pointer to adapter
2444  **/
2445 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2446 {
2447 	struct ixgbe_hw *hw = &adapter->hw;
2448 	u32 eicr = adapter->interrupt_event;
2449 
2450 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2451 		return;
2452 
2453 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2454 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2455 		return;
2456 
2457 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2458 
2459 	switch (hw->device_id) {
2460 	case IXGBE_DEV_ID_82599_T3_LOM:
2461 		/*
2462 		 * Since the warning interrupt is for both ports
2463 		 * we don't have to check if:
2464 		 *  - This interrupt wasn't for our port.
2465 		 *  - We may have missed the interrupt so always have to
2466 		 *    check if we  got a LSC
2467 		 */
2468 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2469 		    !(eicr & IXGBE_EICR_LSC))
2470 			return;
2471 
2472 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2473 			u32 speed;
2474 			bool link_up = false;
2475 
2476 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2477 
2478 			if (link_up)
2479 				return;
2480 		}
2481 
2482 		/* Check if this is not due to overtemp */
2483 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2484 			return;
2485 
2486 		break;
2487 	default:
2488 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2489 			return;
2490 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2491 			return;
2492 		break;
2493 	}
2494 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2495 
2496 	adapter->interrupt_event = 0;
2497 }
2498 
2499 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2500 {
2501 	struct ixgbe_hw *hw = &adapter->hw;
2502 
2503 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2504 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2505 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2506 		/* write to clear the interrupt */
2507 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2508 	}
2509 }
2510 
2511 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2512 {
2513 	struct ixgbe_hw *hw = &adapter->hw;
2514 
2515 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2516 		return;
2517 
2518 	switch (adapter->hw.mac.type) {
2519 	case ixgbe_mac_82599EB:
2520 		/*
2521 		 * Need to check link state so complete overtemp check
2522 		 * on service task
2523 		 */
2524 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2525 		     (eicr & IXGBE_EICR_LSC)) &&
2526 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2527 			adapter->interrupt_event = eicr;
2528 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2529 			ixgbe_service_event_schedule(adapter);
2530 			return;
2531 		}
2532 		return;
2533 	case ixgbe_mac_X540:
2534 		if (!(eicr & IXGBE_EICR_TS))
2535 			return;
2536 		break;
2537 	default:
2538 		return;
2539 	}
2540 
2541 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2542 }
2543 
2544 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2545 {
2546 	switch (hw->mac.type) {
2547 	case ixgbe_mac_82598EB:
2548 		if (hw->phy.type == ixgbe_phy_nl)
2549 			return true;
2550 		return false;
2551 	case ixgbe_mac_82599EB:
2552 	case ixgbe_mac_X550EM_x:
2553 	case ixgbe_mac_x550em_a:
2554 		switch (hw->mac.ops.get_media_type(hw)) {
2555 		case ixgbe_media_type_fiber:
2556 		case ixgbe_media_type_fiber_qsfp:
2557 			return true;
2558 		default:
2559 			return false;
2560 		}
2561 	default:
2562 		return false;
2563 	}
2564 }
2565 
2566 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2567 {
2568 	struct ixgbe_hw *hw = &adapter->hw;
2569 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2570 
2571 	if (!ixgbe_is_sfp(hw))
2572 		return;
2573 
2574 	/* Later MAC's use different SDP */
2575 	if (hw->mac.type >= ixgbe_mac_X540)
2576 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2577 
2578 	if (eicr & eicr_mask) {
2579 		/* Clear the interrupt */
2580 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2581 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2582 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2583 			adapter->sfp_poll_time = 0;
2584 			ixgbe_service_event_schedule(adapter);
2585 		}
2586 	}
2587 
2588 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2589 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2590 		/* Clear the interrupt */
2591 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2592 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2593 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2594 			ixgbe_service_event_schedule(adapter);
2595 		}
2596 	}
2597 }
2598 
2599 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2600 {
2601 	struct ixgbe_hw *hw = &adapter->hw;
2602 
2603 	adapter->lsc_int++;
2604 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2605 	adapter->link_check_timeout = jiffies;
2606 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2607 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2608 		IXGBE_WRITE_FLUSH(hw);
2609 		ixgbe_service_event_schedule(adapter);
2610 	}
2611 }
2612 
2613 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2614 					   u64 qmask)
2615 {
2616 	u32 mask;
2617 	struct ixgbe_hw *hw = &adapter->hw;
2618 
2619 	switch (hw->mac.type) {
2620 	case ixgbe_mac_82598EB:
2621 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2622 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2623 		break;
2624 	case ixgbe_mac_82599EB:
2625 	case ixgbe_mac_X540:
2626 	case ixgbe_mac_X550:
2627 	case ixgbe_mac_X550EM_x:
2628 	case ixgbe_mac_x550em_a:
2629 		mask = (qmask & 0xFFFFFFFF);
2630 		if (mask)
2631 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2632 		mask = (qmask >> 32);
2633 		if (mask)
2634 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2635 		break;
2636 	default:
2637 		break;
2638 	}
2639 	/* skip the flush */
2640 }
2641 
2642 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2643 					    u64 qmask)
2644 {
2645 	u32 mask;
2646 	struct ixgbe_hw *hw = &adapter->hw;
2647 
2648 	switch (hw->mac.type) {
2649 	case ixgbe_mac_82598EB:
2650 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2651 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2652 		break;
2653 	case ixgbe_mac_82599EB:
2654 	case ixgbe_mac_X540:
2655 	case ixgbe_mac_X550:
2656 	case ixgbe_mac_X550EM_x:
2657 	case ixgbe_mac_x550em_a:
2658 		mask = (qmask & 0xFFFFFFFF);
2659 		if (mask)
2660 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2661 		mask = (qmask >> 32);
2662 		if (mask)
2663 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2664 		break;
2665 	default:
2666 		break;
2667 	}
2668 	/* skip the flush */
2669 }
2670 
2671 /**
2672  * ixgbe_irq_enable - Enable default interrupt generation settings
2673  * @adapter: board private structure
2674  **/
2675 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2676 				    bool flush)
2677 {
2678 	struct ixgbe_hw *hw = &adapter->hw;
2679 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2680 
2681 	/* don't reenable LSC while waiting for link */
2682 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2683 		mask &= ~IXGBE_EIMS_LSC;
2684 
2685 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2686 		switch (adapter->hw.mac.type) {
2687 		case ixgbe_mac_82599EB:
2688 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
2689 			break;
2690 		case ixgbe_mac_X540:
2691 		case ixgbe_mac_X550:
2692 		case ixgbe_mac_X550EM_x:
2693 		case ixgbe_mac_x550em_a:
2694 			mask |= IXGBE_EIMS_TS;
2695 			break;
2696 		default:
2697 			break;
2698 		}
2699 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2700 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2701 	switch (adapter->hw.mac.type) {
2702 	case ixgbe_mac_82599EB:
2703 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2704 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
2705 		/* fall through */
2706 	case ixgbe_mac_X540:
2707 	case ixgbe_mac_X550:
2708 	case ixgbe_mac_X550EM_x:
2709 	case ixgbe_mac_x550em_a:
2710 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
2711 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
2712 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
2713 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
2714 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
2715 			mask |= IXGBE_EICR_GPI_SDP0_X540;
2716 		mask |= IXGBE_EIMS_ECC;
2717 		mask |= IXGBE_EIMS_MAILBOX;
2718 		break;
2719 	default:
2720 		break;
2721 	}
2722 
2723 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2724 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2725 		mask |= IXGBE_EIMS_FLOW_DIR;
2726 
2727 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2728 	if (queues)
2729 		ixgbe_irq_enable_queues(adapter, ~0);
2730 	if (flush)
2731 		IXGBE_WRITE_FLUSH(&adapter->hw);
2732 }
2733 
2734 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2735 {
2736 	struct ixgbe_adapter *adapter = data;
2737 	struct ixgbe_hw *hw = &adapter->hw;
2738 	u32 eicr;
2739 
2740 	/*
2741 	 * Workaround for Silicon errata.  Use clear-by-write instead
2742 	 * of clear-by-read.  Reading with EICS will return the
2743 	 * interrupt causes without clearing, which later be done
2744 	 * with the write to EICR.
2745 	 */
2746 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2747 
2748 	/* The lower 16bits of the EICR register are for the queue interrupts
2749 	 * which should be masked here in order to not accidentally clear them if
2750 	 * the bits are high when ixgbe_msix_other is called. There is a race
2751 	 * condition otherwise which results in possible performance loss
2752 	 * especially if the ixgbe_msix_other interrupt is triggering
2753 	 * consistently (as it would when PPS is turned on for the X540 device)
2754 	 */
2755 	eicr &= 0xFFFF0000;
2756 
2757 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2758 
2759 	if (eicr & IXGBE_EICR_LSC)
2760 		ixgbe_check_lsc(adapter);
2761 
2762 	if (eicr & IXGBE_EICR_MAILBOX)
2763 		ixgbe_msg_task(adapter);
2764 
2765 	switch (hw->mac.type) {
2766 	case ixgbe_mac_82599EB:
2767 	case ixgbe_mac_X540:
2768 	case ixgbe_mac_X550:
2769 	case ixgbe_mac_X550EM_x:
2770 	case ixgbe_mac_x550em_a:
2771 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
2772 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2773 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
2774 			ixgbe_service_event_schedule(adapter);
2775 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2776 					IXGBE_EICR_GPI_SDP0_X540);
2777 		}
2778 		if (eicr & IXGBE_EICR_ECC) {
2779 			e_info(link, "Received ECC Err, initiating reset\n");
2780 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
2781 			ixgbe_service_event_schedule(adapter);
2782 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2783 		}
2784 		/* Handle Flow Director Full threshold interrupt */
2785 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2786 			int reinit_count = 0;
2787 			int i;
2788 			for (i = 0; i < adapter->num_tx_queues; i++) {
2789 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2790 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2791 						       &ring->state))
2792 					reinit_count++;
2793 			}
2794 			if (reinit_count) {
2795 				/* no more flow director interrupts until after init */
2796 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2797 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2798 				ixgbe_service_event_schedule(adapter);
2799 			}
2800 		}
2801 		ixgbe_check_sfp_event(adapter, eicr);
2802 		ixgbe_check_overtemp_event(adapter, eicr);
2803 		break;
2804 	default:
2805 		break;
2806 	}
2807 
2808 	ixgbe_check_fan_failure(adapter, eicr);
2809 
2810 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2811 		ixgbe_ptp_check_pps_event(adapter);
2812 
2813 	/* re-enable the original interrupt state, no lsc, no queues */
2814 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2815 		ixgbe_irq_enable(adapter, false, false);
2816 
2817 	return IRQ_HANDLED;
2818 }
2819 
2820 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2821 {
2822 	struct ixgbe_q_vector *q_vector = data;
2823 
2824 	/* EIAM disabled interrupts (on this vector) for us */
2825 
2826 	if (q_vector->rx.ring || q_vector->tx.ring)
2827 		napi_schedule_irqoff(&q_vector->napi);
2828 
2829 	return IRQ_HANDLED;
2830 }
2831 
2832 /**
2833  * ixgbe_poll - NAPI Rx polling callback
2834  * @napi: structure for representing this polling device
2835  * @budget: how many packets driver is allowed to clean
2836  *
2837  * This function is used for legacy and MSI, NAPI mode
2838  **/
2839 int ixgbe_poll(struct napi_struct *napi, int budget)
2840 {
2841 	struct ixgbe_q_vector *q_vector =
2842 				container_of(napi, struct ixgbe_q_vector, napi);
2843 	struct ixgbe_adapter *adapter = q_vector->adapter;
2844 	struct ixgbe_ring *ring;
2845 	int per_ring_budget, work_done = 0;
2846 	bool clean_complete = true;
2847 
2848 #ifdef CONFIG_IXGBE_DCA
2849 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2850 		ixgbe_update_dca(q_vector);
2851 #endif
2852 
2853 	ixgbe_for_each_ring(ring, q_vector->tx) {
2854 		if (!ixgbe_clean_tx_irq(q_vector, ring, budget))
2855 			clean_complete = false;
2856 	}
2857 
2858 	/* Exit if we are called by netpoll or busy polling is active */
2859 	if ((budget <= 0) || !ixgbe_qv_lock_napi(q_vector))
2860 		return budget;
2861 
2862 	/* attempt to distribute budget to each queue fairly, but don't allow
2863 	 * the budget to go below 1 because we'll exit polling */
2864 	if (q_vector->rx.count > 1)
2865 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2866 	else
2867 		per_ring_budget = budget;
2868 
2869 	ixgbe_for_each_ring(ring, q_vector->rx) {
2870 		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
2871 						 per_ring_budget);
2872 
2873 		work_done += cleaned;
2874 		if (cleaned >= per_ring_budget)
2875 			clean_complete = false;
2876 	}
2877 
2878 	ixgbe_qv_unlock_napi(q_vector);
2879 	/* If all work not completed, return budget and keep polling */
2880 	if (!clean_complete)
2881 		return budget;
2882 
2883 	/* all work done, exit the polling mode */
2884 	napi_complete_done(napi, work_done);
2885 	if (adapter->rx_itr_setting & 1)
2886 		ixgbe_set_itr(q_vector);
2887 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2888 		ixgbe_irq_enable_queues(adapter, BIT_ULL(q_vector->v_idx));
2889 
2890 	return min(work_done, budget - 1);
2891 }
2892 
2893 /**
2894  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2895  * @adapter: board private structure
2896  *
2897  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2898  * interrupts from the kernel.
2899  **/
2900 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2901 {
2902 	struct net_device *netdev = adapter->netdev;
2903 	int vector, err;
2904 	int ri = 0, ti = 0;
2905 
2906 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2907 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2908 		struct msix_entry *entry = &adapter->msix_entries[vector];
2909 
2910 		if (q_vector->tx.ring && q_vector->rx.ring) {
2911 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2912 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2913 			ti++;
2914 		} else if (q_vector->rx.ring) {
2915 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2916 				 "%s-%s-%d", netdev->name, "rx", ri++);
2917 		} else if (q_vector->tx.ring) {
2918 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2919 				 "%s-%s-%d", netdev->name, "tx", ti++);
2920 		} else {
2921 			/* skip this unused q_vector */
2922 			continue;
2923 		}
2924 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2925 				  q_vector->name, q_vector);
2926 		if (err) {
2927 			e_err(probe, "request_irq failed for MSIX interrupt "
2928 			      "Error: %d\n", err);
2929 			goto free_queue_irqs;
2930 		}
2931 		/* If Flow Director is enabled, set interrupt affinity */
2932 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2933 			/* assign the mask for this irq */
2934 			irq_set_affinity_hint(entry->vector,
2935 					      &q_vector->affinity_mask);
2936 		}
2937 	}
2938 
2939 	err = request_irq(adapter->msix_entries[vector].vector,
2940 			  ixgbe_msix_other, 0, netdev->name, adapter);
2941 	if (err) {
2942 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2943 		goto free_queue_irqs;
2944 	}
2945 
2946 	return 0;
2947 
2948 free_queue_irqs:
2949 	while (vector) {
2950 		vector--;
2951 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2952 				      NULL);
2953 		free_irq(adapter->msix_entries[vector].vector,
2954 			 adapter->q_vector[vector]);
2955 	}
2956 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2957 	pci_disable_msix(adapter->pdev);
2958 	kfree(adapter->msix_entries);
2959 	adapter->msix_entries = NULL;
2960 	return err;
2961 }
2962 
2963 /**
2964  * ixgbe_intr - legacy mode Interrupt Handler
2965  * @irq: interrupt number
2966  * @data: pointer to a network interface device structure
2967  **/
2968 static irqreturn_t ixgbe_intr(int irq, void *data)
2969 {
2970 	struct ixgbe_adapter *adapter = data;
2971 	struct ixgbe_hw *hw = &adapter->hw;
2972 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2973 	u32 eicr;
2974 
2975 	/*
2976 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2977 	 * before the read of EICR.
2978 	 */
2979 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2980 
2981 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2982 	 * therefore no explicit interrupt disable is necessary */
2983 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2984 	if (!eicr) {
2985 		/*
2986 		 * shared interrupt alert!
2987 		 * make sure interrupts are enabled because the read will
2988 		 * have disabled interrupts due to EIAM
2989 		 * finish the workaround of silicon errata on 82598.  Unmask
2990 		 * the interrupt that we masked before the EICR read.
2991 		 */
2992 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2993 			ixgbe_irq_enable(adapter, true, true);
2994 		return IRQ_NONE;	/* Not our interrupt */
2995 	}
2996 
2997 	if (eicr & IXGBE_EICR_LSC)
2998 		ixgbe_check_lsc(adapter);
2999 
3000 	switch (hw->mac.type) {
3001 	case ixgbe_mac_82599EB:
3002 		ixgbe_check_sfp_event(adapter, eicr);
3003 		/* Fall through */
3004 	case ixgbe_mac_X540:
3005 	case ixgbe_mac_X550:
3006 	case ixgbe_mac_X550EM_x:
3007 	case ixgbe_mac_x550em_a:
3008 		if (eicr & IXGBE_EICR_ECC) {
3009 			e_info(link, "Received ECC Err, initiating reset\n");
3010 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3011 			ixgbe_service_event_schedule(adapter);
3012 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3013 		}
3014 		ixgbe_check_overtemp_event(adapter, eicr);
3015 		break;
3016 	default:
3017 		break;
3018 	}
3019 
3020 	ixgbe_check_fan_failure(adapter, eicr);
3021 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3022 		ixgbe_ptp_check_pps_event(adapter);
3023 
3024 	/* would disable interrupts here but EIAM disabled it */
3025 	napi_schedule_irqoff(&q_vector->napi);
3026 
3027 	/*
3028 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3029 	 * ixgbe_poll will re-enable the queue interrupts
3030 	 */
3031 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3032 		ixgbe_irq_enable(adapter, false, false);
3033 
3034 	return IRQ_HANDLED;
3035 }
3036 
3037 /**
3038  * ixgbe_request_irq - initialize interrupts
3039  * @adapter: board private structure
3040  *
3041  * Attempts to configure interrupts using the best available
3042  * capabilities of the hardware and kernel.
3043  **/
3044 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3045 {
3046 	struct net_device *netdev = adapter->netdev;
3047 	int err;
3048 
3049 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3050 		err = ixgbe_request_msix_irqs(adapter);
3051 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3052 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3053 				  netdev->name, adapter);
3054 	else
3055 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3056 				  netdev->name, adapter);
3057 
3058 	if (err)
3059 		e_err(probe, "request_irq failed, Error %d\n", err);
3060 
3061 	return err;
3062 }
3063 
3064 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3065 {
3066 	int vector;
3067 
3068 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3069 		free_irq(adapter->pdev->irq, adapter);
3070 		return;
3071 	}
3072 
3073 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3074 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3075 		struct msix_entry *entry = &adapter->msix_entries[vector];
3076 
3077 		/* free only the irqs that were actually requested */
3078 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3079 			continue;
3080 
3081 		/* clear the affinity_mask in the IRQ descriptor */
3082 		irq_set_affinity_hint(entry->vector, NULL);
3083 
3084 		free_irq(entry->vector, q_vector);
3085 	}
3086 
3087 	free_irq(adapter->msix_entries[vector].vector, adapter);
3088 }
3089 
3090 /**
3091  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3092  * @adapter: board private structure
3093  **/
3094 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3095 {
3096 	switch (adapter->hw.mac.type) {
3097 	case ixgbe_mac_82598EB:
3098 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3099 		break;
3100 	case ixgbe_mac_82599EB:
3101 	case ixgbe_mac_X540:
3102 	case ixgbe_mac_X550:
3103 	case ixgbe_mac_X550EM_x:
3104 	case ixgbe_mac_x550em_a:
3105 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3106 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3107 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3108 		break;
3109 	default:
3110 		break;
3111 	}
3112 	IXGBE_WRITE_FLUSH(&adapter->hw);
3113 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3114 		int vector;
3115 
3116 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3117 			synchronize_irq(adapter->msix_entries[vector].vector);
3118 
3119 		synchronize_irq(adapter->msix_entries[vector++].vector);
3120 	} else {
3121 		synchronize_irq(adapter->pdev->irq);
3122 	}
3123 }
3124 
3125 /**
3126  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3127  *
3128  **/
3129 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3130 {
3131 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3132 
3133 	ixgbe_write_eitr(q_vector);
3134 
3135 	ixgbe_set_ivar(adapter, 0, 0, 0);
3136 	ixgbe_set_ivar(adapter, 1, 0, 0);
3137 
3138 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3139 }
3140 
3141 /**
3142  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3143  * @adapter: board private structure
3144  * @ring: structure containing ring specific data
3145  *
3146  * Configure the Tx descriptor ring after a reset.
3147  **/
3148 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3149 			     struct ixgbe_ring *ring)
3150 {
3151 	struct ixgbe_hw *hw = &adapter->hw;
3152 	u64 tdba = ring->dma;
3153 	int wait_loop = 10;
3154 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3155 	u8 reg_idx = ring->reg_idx;
3156 
3157 	/* disable queue to avoid issues while updating state */
3158 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3159 	IXGBE_WRITE_FLUSH(hw);
3160 
3161 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3162 			(tdba & DMA_BIT_MASK(32)));
3163 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3164 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3165 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3166 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3167 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3168 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3169 
3170 	/*
3171 	 * set WTHRESH to encourage burst writeback, it should not be set
3172 	 * higher than 1 when:
3173 	 * - ITR is 0 as it could cause false TX hangs
3174 	 * - ITR is set to > 100k int/sec and BQL is enabled
3175 	 *
3176 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3177 	 * to or less than the number of on chip descriptors, which is
3178 	 * currently 40.
3179 	 */
3180 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3181 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3182 	else
3183 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3184 
3185 	/*
3186 	 * Setting PTHRESH to 32 both improves performance
3187 	 * and avoids a TX hang with DFP enabled
3188 	 */
3189 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3190 		   32;		/* PTHRESH = 32 */
3191 
3192 	/* reinitialize flowdirector state */
3193 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3194 		ring->atr_sample_rate = adapter->atr_sample_rate;
3195 		ring->atr_count = 0;
3196 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3197 	} else {
3198 		ring->atr_sample_rate = 0;
3199 	}
3200 
3201 	/* initialize XPS */
3202 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3203 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3204 
3205 		if (q_vector)
3206 			netif_set_xps_queue(ring->netdev,
3207 					    &q_vector->affinity_mask,
3208 					    ring->queue_index);
3209 	}
3210 
3211 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3212 
3213 	/* enable queue */
3214 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3215 
3216 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3217 	if (hw->mac.type == ixgbe_mac_82598EB &&
3218 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3219 		return;
3220 
3221 	/* poll to verify queue is enabled */
3222 	do {
3223 		usleep_range(1000, 2000);
3224 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3225 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3226 	if (!wait_loop)
3227 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3228 }
3229 
3230 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3231 {
3232 	struct ixgbe_hw *hw = &adapter->hw;
3233 	u32 rttdcs, mtqc;
3234 	u8 tcs = netdev_get_num_tc(adapter->netdev);
3235 
3236 	if (hw->mac.type == ixgbe_mac_82598EB)
3237 		return;
3238 
3239 	/* disable the arbiter while setting MTQC */
3240 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3241 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3242 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3243 
3244 	/* set transmit pool layout */
3245 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3246 		mtqc = IXGBE_MTQC_VT_ENA;
3247 		if (tcs > 4)
3248 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3249 		else if (tcs > 1)
3250 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3251 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3252 			 IXGBE_82599_VMDQ_4Q_MASK)
3253 			mtqc |= IXGBE_MTQC_32VF;
3254 		else
3255 			mtqc |= IXGBE_MTQC_64VF;
3256 	} else {
3257 		if (tcs > 4)
3258 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3259 		else if (tcs > 1)
3260 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3261 		else
3262 			mtqc = IXGBE_MTQC_64Q_1PB;
3263 	}
3264 
3265 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3266 
3267 	/* Enable Security TX Buffer IFG for multiple pb */
3268 	if (tcs) {
3269 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3270 		sectx |= IXGBE_SECTX_DCB;
3271 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3272 	}
3273 
3274 	/* re-enable the arbiter */
3275 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3276 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3277 }
3278 
3279 /**
3280  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3281  * @adapter: board private structure
3282  *
3283  * Configure the Tx unit of the MAC after a reset.
3284  **/
3285 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3286 {
3287 	struct ixgbe_hw *hw = &adapter->hw;
3288 	u32 dmatxctl;
3289 	u32 i;
3290 
3291 	ixgbe_setup_mtqc(adapter);
3292 
3293 	if (hw->mac.type != ixgbe_mac_82598EB) {
3294 		/* DMATXCTL.EN must be before Tx queues are enabled */
3295 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3296 		dmatxctl |= IXGBE_DMATXCTL_TE;
3297 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3298 	}
3299 
3300 	/* Setup the HW Tx Head and Tail descriptor pointers */
3301 	for (i = 0; i < adapter->num_tx_queues; i++)
3302 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3303 }
3304 
3305 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3306 				 struct ixgbe_ring *ring)
3307 {
3308 	struct ixgbe_hw *hw = &adapter->hw;
3309 	u8 reg_idx = ring->reg_idx;
3310 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3311 
3312 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3313 
3314 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3315 }
3316 
3317 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3318 				  struct ixgbe_ring *ring)
3319 {
3320 	struct ixgbe_hw *hw = &adapter->hw;
3321 	u8 reg_idx = ring->reg_idx;
3322 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3323 
3324 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3325 
3326 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3327 }
3328 
3329 #ifdef CONFIG_IXGBE_DCB
3330 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3331 #else
3332 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3333 #endif
3334 {
3335 	int i;
3336 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3337 
3338 	if (adapter->ixgbe_ieee_pfc)
3339 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3340 
3341 	/*
3342 	 * We should set the drop enable bit if:
3343 	 *  SR-IOV is enabled
3344 	 *   or
3345 	 *  Number of Rx queues > 1 and flow control is disabled
3346 	 *
3347 	 *  This allows us to avoid head of line blocking for security
3348 	 *  and performance reasons.
3349 	 */
3350 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3351 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3352 		for (i = 0; i < adapter->num_rx_queues; i++)
3353 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3354 	} else {
3355 		for (i = 0; i < adapter->num_rx_queues; i++)
3356 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3357 	}
3358 }
3359 
3360 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3361 
3362 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3363 				   struct ixgbe_ring *rx_ring)
3364 {
3365 	struct ixgbe_hw *hw = &adapter->hw;
3366 	u32 srrctl;
3367 	u8 reg_idx = rx_ring->reg_idx;
3368 
3369 	if (hw->mac.type == ixgbe_mac_82598EB) {
3370 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3371 
3372 		/*
3373 		 * if VMDq is not active we must program one srrctl register
3374 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3375 		 */
3376 		reg_idx &= mask;
3377 	}
3378 
3379 	/* configure header buffer length, needed for RSC */
3380 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3381 
3382 	/* configure the packet buffer length */
3383 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3384 
3385 	/* configure descriptor type */
3386 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3387 
3388 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3389 }
3390 
3391 /**
3392  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3393  * @adapter: device handle
3394  *
3395  *  - 82598/82599/X540:     128
3396  *  - X550(non-SRIOV mode): 512
3397  *  - X550(SRIOV mode):     64
3398  */
3399 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3400 {
3401 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3402 		return 128;
3403 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3404 		return 64;
3405 	else
3406 		return 512;
3407 }
3408 
3409 /**
3410  * ixgbe_store_reta - Write the RETA table to HW
3411  * @adapter: device handle
3412  *
3413  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3414  */
3415 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3416 {
3417 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3418 	struct ixgbe_hw *hw = &adapter->hw;
3419 	u32 reta = 0;
3420 	u32 indices_multi;
3421 	u8 *indir_tbl = adapter->rss_indir_tbl;
3422 
3423 	/* Fill out the redirection table as follows:
3424 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3425 	 *    indices.
3426 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3427 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3428 	 */
3429 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3430 		indices_multi = 0x11;
3431 	else
3432 		indices_multi = 0x1;
3433 
3434 	/* Write redirection table to HW */
3435 	for (i = 0; i < reta_entries; i++) {
3436 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3437 		if ((i & 3) == 3) {
3438 			if (i < 128)
3439 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3440 			else
3441 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3442 						reta);
3443 			reta = 0;
3444 		}
3445 	}
3446 }
3447 
3448 /**
3449  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3450  * @adapter: device handle
3451  *
3452  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3453  */
3454 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3455 {
3456 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3457 	struct ixgbe_hw *hw = &adapter->hw;
3458 	u32 vfreta = 0;
3459 	unsigned int pf_pool = adapter->num_vfs;
3460 
3461 	/* Write redirection table to HW */
3462 	for (i = 0; i < reta_entries; i++) {
3463 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3464 		if ((i & 3) == 3) {
3465 			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3466 					vfreta);
3467 			vfreta = 0;
3468 		}
3469 	}
3470 }
3471 
3472 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3473 {
3474 	struct ixgbe_hw *hw = &adapter->hw;
3475 	u32 i, j;
3476 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3477 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3478 
3479 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3480 	 * make full use of any rings they may have.  We will use the
3481 	 * PSRTYPE register to control how many rings we use within the PF.
3482 	 */
3483 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3484 		rss_i = 4;
3485 
3486 	/* Fill out hash function seeds */
3487 	for (i = 0; i < 10; i++)
3488 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3489 
3490 	/* Fill out redirection table */
3491 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3492 
3493 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3494 		if (j == rss_i)
3495 			j = 0;
3496 
3497 		adapter->rss_indir_tbl[i] = j;
3498 	}
3499 
3500 	ixgbe_store_reta(adapter);
3501 }
3502 
3503 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3504 {
3505 	struct ixgbe_hw *hw = &adapter->hw;
3506 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3507 	unsigned int pf_pool = adapter->num_vfs;
3508 	int i, j;
3509 
3510 	/* Fill out hash function seeds */
3511 	for (i = 0; i < 10; i++)
3512 		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool),
3513 				adapter->rss_key[i]);
3514 
3515 	/* Fill out the redirection table */
3516 	for (i = 0, j = 0; i < 64; i++, j++) {
3517 		if (j == rss_i)
3518 			j = 0;
3519 
3520 		adapter->rss_indir_tbl[i] = j;
3521 	}
3522 
3523 	ixgbe_store_vfreta(adapter);
3524 }
3525 
3526 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3527 {
3528 	struct ixgbe_hw *hw = &adapter->hw;
3529 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3530 	u32 rxcsum;
3531 
3532 	/* Disable indicating checksum in descriptor, enables RSS hash */
3533 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3534 	rxcsum |= IXGBE_RXCSUM_PCSD;
3535 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3536 
3537 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3538 		if (adapter->ring_feature[RING_F_RSS].mask)
3539 			mrqc = IXGBE_MRQC_RSSEN;
3540 	} else {
3541 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3542 
3543 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3544 			if (tcs > 4)
3545 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3546 			else if (tcs > 1)
3547 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3548 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3549 				 IXGBE_82599_VMDQ_4Q_MASK)
3550 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3551 			else
3552 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3553 		} else {
3554 			if (tcs > 4)
3555 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3556 			else if (tcs > 1)
3557 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3558 			else
3559 				mrqc = IXGBE_MRQC_RSSEN;
3560 		}
3561 	}
3562 
3563 	/* Perform hash on these packet types */
3564 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3565 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3566 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3567 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3568 
3569 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3570 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3571 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3572 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3573 
3574 	netdev_rss_key_fill(adapter->rss_key, sizeof(adapter->rss_key));
3575 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3576 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3577 		unsigned int pf_pool = adapter->num_vfs;
3578 
3579 		/* Enable VF RSS mode */
3580 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3581 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3582 
3583 		/* Setup RSS through the VF registers */
3584 		ixgbe_setup_vfreta(adapter);
3585 		vfmrqc = IXGBE_MRQC_RSSEN;
3586 		vfmrqc |= rss_field;
3587 		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3588 	} else {
3589 		ixgbe_setup_reta(adapter);
3590 		mrqc |= rss_field;
3591 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3592 	}
3593 }
3594 
3595 /**
3596  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3597  * @adapter:    address of board private structure
3598  * @index:      index of ring to set
3599  **/
3600 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3601 				   struct ixgbe_ring *ring)
3602 {
3603 	struct ixgbe_hw *hw = &adapter->hw;
3604 	u32 rscctrl;
3605 	u8 reg_idx = ring->reg_idx;
3606 
3607 	if (!ring_is_rsc_enabled(ring))
3608 		return;
3609 
3610 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3611 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3612 	/*
3613 	 * we must limit the number of descriptors so that the
3614 	 * total size of max desc * buf_len is not greater
3615 	 * than 65536
3616 	 */
3617 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3618 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3619 }
3620 
3621 #define IXGBE_MAX_RX_DESC_POLL 10
3622 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3623 				       struct ixgbe_ring *ring)
3624 {
3625 	struct ixgbe_hw *hw = &adapter->hw;
3626 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3627 	u32 rxdctl;
3628 	u8 reg_idx = ring->reg_idx;
3629 
3630 	if (ixgbe_removed(hw->hw_addr))
3631 		return;
3632 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3633 	if (hw->mac.type == ixgbe_mac_82598EB &&
3634 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3635 		return;
3636 
3637 	do {
3638 		usleep_range(1000, 2000);
3639 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3640 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3641 
3642 	if (!wait_loop) {
3643 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3644 		      "the polling period\n", reg_idx);
3645 	}
3646 }
3647 
3648 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3649 			    struct ixgbe_ring *ring)
3650 {
3651 	struct ixgbe_hw *hw = &adapter->hw;
3652 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3653 	u32 rxdctl;
3654 	u8 reg_idx = ring->reg_idx;
3655 
3656 	if (ixgbe_removed(hw->hw_addr))
3657 		return;
3658 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3659 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3660 
3661 	/* write value back with RXDCTL.ENABLE bit cleared */
3662 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3663 
3664 	if (hw->mac.type == ixgbe_mac_82598EB &&
3665 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3666 		return;
3667 
3668 	/* the hardware may take up to 100us to really disable the rx queue */
3669 	do {
3670 		udelay(10);
3671 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3672 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3673 
3674 	if (!wait_loop) {
3675 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3676 		      "the polling period\n", reg_idx);
3677 	}
3678 }
3679 
3680 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3681 			     struct ixgbe_ring *ring)
3682 {
3683 	struct ixgbe_hw *hw = &adapter->hw;
3684 	u64 rdba = ring->dma;
3685 	u32 rxdctl;
3686 	u8 reg_idx = ring->reg_idx;
3687 
3688 	/* disable queue to avoid issues while updating state */
3689 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3690 	ixgbe_disable_rx_queue(adapter, ring);
3691 
3692 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3693 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3694 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3695 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3696 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
3697 	IXGBE_WRITE_FLUSH(hw);
3698 
3699 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3700 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3701 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3702 
3703 	ixgbe_configure_srrctl(adapter, ring);
3704 	ixgbe_configure_rscctl(adapter, ring);
3705 
3706 	if (hw->mac.type == ixgbe_mac_82598EB) {
3707 		/*
3708 		 * enable cache line friendly hardware writes:
3709 		 * PTHRESH=32 descriptors (half the internal cache),
3710 		 * this also removes ugly rx_no_buffer_count increment
3711 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3712 		 * WTHRESH=8 burst writeback up to two cache lines
3713 		 */
3714 		rxdctl &= ~0x3FFFFF;
3715 		rxdctl |=  0x080420;
3716 	}
3717 
3718 	/* enable receive descriptor ring */
3719 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3720 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3721 
3722 	ixgbe_rx_desc_queue_enable(adapter, ring);
3723 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3724 }
3725 
3726 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3727 {
3728 	struct ixgbe_hw *hw = &adapter->hw;
3729 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3730 	u16 pool;
3731 
3732 	/* PSRTYPE must be initialized in non 82598 adapters */
3733 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3734 		      IXGBE_PSRTYPE_UDPHDR |
3735 		      IXGBE_PSRTYPE_IPV4HDR |
3736 		      IXGBE_PSRTYPE_L2HDR |
3737 		      IXGBE_PSRTYPE_IPV6HDR;
3738 
3739 	if (hw->mac.type == ixgbe_mac_82598EB)
3740 		return;
3741 
3742 	if (rss_i > 3)
3743 		psrtype |= 2u << 29;
3744 	else if (rss_i > 1)
3745 		psrtype |= 1u << 29;
3746 
3747 	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3748 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3749 }
3750 
3751 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3752 {
3753 	struct ixgbe_hw *hw = &adapter->hw;
3754 	u32 reg_offset, vf_shift;
3755 	u32 gcr_ext, vmdctl;
3756 	int i;
3757 
3758 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3759 		return;
3760 
3761 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3762 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3763 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3764 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3765 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3766 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3767 
3768 	vf_shift = VMDQ_P(0) % 32;
3769 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3770 
3771 	/* Enable only the PF's pool for Tx/Rx */
3772 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
3773 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3774 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
3775 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3776 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
3777 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3778 
3779 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3780 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3781 
3782 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
3783 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
3784 
3785 	/*
3786 	 * Set up VF register offsets for selected VT Mode,
3787 	 * i.e. 32 or 64 VFs for SR-IOV
3788 	 */
3789 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3790 	case IXGBE_82599_VMDQ_8Q_MASK:
3791 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3792 		break;
3793 	case IXGBE_82599_VMDQ_4Q_MASK:
3794 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3795 		break;
3796 	default:
3797 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3798 		break;
3799 	}
3800 
3801 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3802 
3803 	for (i = 0; i < adapter->num_vfs; i++) {
3804 		/* configure spoof checking */
3805 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
3806 					  adapter->vfinfo[i].spoofchk_enabled);
3807 
3808 		/* Enable/Disable RSS query feature  */
3809 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
3810 					  adapter->vfinfo[i].rss_query_enabled);
3811 	}
3812 }
3813 
3814 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3815 {
3816 	struct ixgbe_hw *hw = &adapter->hw;
3817 	struct net_device *netdev = adapter->netdev;
3818 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3819 	struct ixgbe_ring *rx_ring;
3820 	int i;
3821 	u32 mhadd, hlreg0;
3822 
3823 #ifdef IXGBE_FCOE
3824 	/* adjust max frame to be able to do baby jumbo for FCoE */
3825 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3826 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3827 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3828 
3829 #endif /* IXGBE_FCOE */
3830 
3831 	/* adjust max frame to be at least the size of a standard frame */
3832 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3833 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3834 
3835 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3836 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3837 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3838 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3839 
3840 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3841 	}
3842 
3843 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3844 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3845 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3846 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3847 
3848 	/*
3849 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3850 	 * the Base and Length of the Rx Descriptor Ring
3851 	 */
3852 	for (i = 0; i < adapter->num_rx_queues; i++) {
3853 		rx_ring = adapter->rx_ring[i];
3854 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3855 			set_ring_rsc_enabled(rx_ring);
3856 		else
3857 			clear_ring_rsc_enabled(rx_ring);
3858 	}
3859 }
3860 
3861 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3862 {
3863 	struct ixgbe_hw *hw = &adapter->hw;
3864 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3865 
3866 	switch (hw->mac.type) {
3867 	case ixgbe_mac_82598EB:
3868 		/*
3869 		 * For VMDq support of different descriptor types or
3870 		 * buffer sizes through the use of multiple SRRCTL
3871 		 * registers, RDRXCTL.MVMEN must be set to 1
3872 		 *
3873 		 * also, the manual doesn't mention it clearly but DCA hints
3874 		 * will only use queue 0's tags unless this bit is set.  Side
3875 		 * effects of setting this bit are only that SRRCTL must be
3876 		 * fully programmed [0..15]
3877 		 */
3878 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3879 		break;
3880 	case ixgbe_mac_X550:
3881 	case ixgbe_mac_X550EM_x:
3882 	case ixgbe_mac_x550em_a:
3883 		if (adapter->num_vfs)
3884 			rdrxctl |= IXGBE_RDRXCTL_PSP;
3885 		/* fall through for older HW */
3886 	case ixgbe_mac_82599EB:
3887 	case ixgbe_mac_X540:
3888 		/* Disable RSC for ACK packets */
3889 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3890 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3891 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3892 		/* hardware requires some bits to be set by default */
3893 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3894 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3895 		break;
3896 	default:
3897 		/* We should do nothing since we don't know this hardware */
3898 		return;
3899 	}
3900 
3901 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3902 }
3903 
3904 /**
3905  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3906  * @adapter: board private structure
3907  *
3908  * Configure the Rx unit of the MAC after a reset.
3909  **/
3910 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3911 {
3912 	struct ixgbe_hw *hw = &adapter->hw;
3913 	int i;
3914 	u32 rxctrl, rfctl;
3915 
3916 	/* disable receives while setting up the descriptors */
3917 	hw->mac.ops.disable_rx(hw);
3918 
3919 	ixgbe_setup_psrtype(adapter);
3920 	ixgbe_setup_rdrxctl(adapter);
3921 
3922 	/* RSC Setup */
3923 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
3924 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
3925 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
3926 		rfctl |= IXGBE_RFCTL_RSC_DIS;
3927 
3928 	/* disable NFS filtering */
3929 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
3930 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3931 
3932 	/* Program registers for the distribution of queues */
3933 	ixgbe_setup_mrqc(adapter);
3934 
3935 	/* set_rx_buffer_len must be called before ring initialization */
3936 	ixgbe_set_rx_buffer_len(adapter);
3937 
3938 	/*
3939 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3940 	 * the Base and Length of the Rx Descriptor Ring
3941 	 */
3942 	for (i = 0; i < adapter->num_rx_queues; i++)
3943 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3944 
3945 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3946 	/* disable drop enable for 82598 parts */
3947 	if (hw->mac.type == ixgbe_mac_82598EB)
3948 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3949 
3950 	/* enable all receives */
3951 	rxctrl |= IXGBE_RXCTRL_RXEN;
3952 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3953 }
3954 
3955 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3956 				 __be16 proto, u16 vid)
3957 {
3958 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3959 	struct ixgbe_hw *hw = &adapter->hw;
3960 
3961 	/* add VID to filter table */
3962 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
3963 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
3964 
3965 	set_bit(vid, adapter->active_vlans);
3966 
3967 	return 0;
3968 }
3969 
3970 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
3971 {
3972 	u32 vlvf;
3973 	int idx;
3974 
3975 	/* short cut the special case */
3976 	if (vlan == 0)
3977 		return 0;
3978 
3979 	/* Search for the vlan id in the VLVF entries */
3980 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
3981 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
3982 		if ((vlvf & VLAN_VID_MASK) == vlan)
3983 			break;
3984 	}
3985 
3986 	return idx;
3987 }
3988 
3989 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
3990 {
3991 	struct ixgbe_hw *hw = &adapter->hw;
3992 	u32 bits, word;
3993 	int idx;
3994 
3995 	idx = ixgbe_find_vlvf_entry(hw, vid);
3996 	if (!idx)
3997 		return;
3998 
3999 	/* See if any other pools are set for this VLAN filter
4000 	 * entry other than the PF.
4001 	 */
4002 	word = idx * 2 + (VMDQ_P(0) / 32);
4003 	bits = ~BIT(VMDQ_P(0) % 32);
4004 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4005 
4006 	/* Disable the filter so this falls into the default pool. */
4007 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4008 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4009 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4010 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4011 	}
4012 }
4013 
4014 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4015 				  __be16 proto, u16 vid)
4016 {
4017 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4018 	struct ixgbe_hw *hw = &adapter->hw;
4019 
4020 	/* remove VID from filter table */
4021 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4022 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4023 
4024 	clear_bit(vid, adapter->active_vlans);
4025 
4026 	return 0;
4027 }
4028 
4029 /**
4030  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4031  * @adapter: driver data
4032  */
4033 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4034 {
4035 	struct ixgbe_hw *hw = &adapter->hw;
4036 	u32 vlnctrl;
4037 	int i, j;
4038 
4039 	switch (hw->mac.type) {
4040 	case ixgbe_mac_82598EB:
4041 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4042 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4043 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4044 		break;
4045 	case ixgbe_mac_82599EB:
4046 	case ixgbe_mac_X540:
4047 	case ixgbe_mac_X550:
4048 	case ixgbe_mac_X550EM_x:
4049 	case ixgbe_mac_x550em_a:
4050 		for (i = 0; i < adapter->num_rx_queues; i++) {
4051 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4052 
4053 			if (ring->l2_accel_priv)
4054 				continue;
4055 			j = ring->reg_idx;
4056 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4057 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4058 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4059 		}
4060 		break;
4061 	default:
4062 		break;
4063 	}
4064 }
4065 
4066 /**
4067  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4068  * @adapter: driver data
4069  */
4070 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4071 {
4072 	struct ixgbe_hw *hw = &adapter->hw;
4073 	u32 vlnctrl;
4074 	int i, j;
4075 
4076 	switch (hw->mac.type) {
4077 	case ixgbe_mac_82598EB:
4078 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4079 		vlnctrl |= IXGBE_VLNCTRL_VME;
4080 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4081 		break;
4082 	case ixgbe_mac_82599EB:
4083 	case ixgbe_mac_X540:
4084 	case ixgbe_mac_X550:
4085 	case ixgbe_mac_X550EM_x:
4086 	case ixgbe_mac_x550em_a:
4087 		for (i = 0; i < adapter->num_rx_queues; i++) {
4088 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4089 
4090 			if (ring->l2_accel_priv)
4091 				continue;
4092 			j = ring->reg_idx;
4093 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4094 			vlnctrl |= IXGBE_RXDCTL_VME;
4095 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4096 		}
4097 		break;
4098 	default:
4099 		break;
4100 	}
4101 }
4102 
4103 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4104 {
4105 	struct ixgbe_hw *hw = &adapter->hw;
4106 	u32 vlnctrl, i;
4107 
4108 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4109 
4110 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4111 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4112 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4113 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4114 	} else {
4115 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4116 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4117 		return;
4118 	}
4119 
4120 	/* Nothing to do for 82598 */
4121 	if (hw->mac.type == ixgbe_mac_82598EB)
4122 		return;
4123 
4124 	/* We are already in VLAN promisc, nothing to do */
4125 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4126 		return;
4127 
4128 	/* Set flag so we don't redo unnecessary work */
4129 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4130 
4131 	/* Add PF to all active pools */
4132 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4133 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4134 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4135 
4136 		vlvfb |= BIT(VMDQ_P(0) % 32);
4137 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4138 	}
4139 
4140 	/* Set all bits in the VLAN filter table array */
4141 	for (i = hw->mac.vft_size; i--;)
4142 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4143 }
4144 
4145 #define VFTA_BLOCK_SIZE 8
4146 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4147 {
4148 	struct ixgbe_hw *hw = &adapter->hw;
4149 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4150 	u32 vid_start = vfta_offset * 32;
4151 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4152 	u32 i, vid, word, bits;
4153 
4154 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4155 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4156 
4157 		/* pull VLAN ID from VLVF */
4158 		vid = vlvf & VLAN_VID_MASK;
4159 
4160 		/* only concern outselves with a certain range */
4161 		if (vid < vid_start || vid >= vid_end)
4162 			continue;
4163 
4164 		if (vlvf) {
4165 			/* record VLAN ID in VFTA */
4166 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4167 
4168 			/* if PF is part of this then continue */
4169 			if (test_bit(vid, adapter->active_vlans))
4170 				continue;
4171 		}
4172 
4173 		/* remove PF from the pool */
4174 		word = i * 2 + VMDQ_P(0) / 32;
4175 		bits = ~BIT(VMDQ_P(0) % 32);
4176 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4177 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4178 	}
4179 
4180 	/* extract values from active_vlans and write back to VFTA */
4181 	for (i = VFTA_BLOCK_SIZE; i--;) {
4182 		vid = (vfta_offset + i) * 32;
4183 		word = vid / BITS_PER_LONG;
4184 		bits = vid % BITS_PER_LONG;
4185 
4186 		vfta[i] |= adapter->active_vlans[word] >> bits;
4187 
4188 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4189 	}
4190 }
4191 
4192 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4193 {
4194 	struct ixgbe_hw *hw = &adapter->hw;
4195 	u32 vlnctrl, i;
4196 
4197 	/* Set VLAN filtering to enabled */
4198 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4199 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4200 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4201 
4202 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4203 	    hw->mac.type == ixgbe_mac_82598EB)
4204 		return;
4205 
4206 	/* We are not in VLAN promisc, nothing to do */
4207 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4208 		return;
4209 
4210 	/* Set flag so we don't redo unnecessary work */
4211 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4212 
4213 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4214 		ixgbe_scrub_vfta(adapter, i);
4215 }
4216 
4217 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4218 {
4219 	u16 vid = 1;
4220 
4221 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4222 
4223 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4224 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4225 }
4226 
4227 /**
4228  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4229  * @netdev: network interface device structure
4230  *
4231  * Writes multicast address list to the MTA hash table.
4232  * Returns: -ENOMEM on failure
4233  *                0 on no addresses written
4234  *                X on writing X addresses to MTA
4235  **/
4236 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4237 {
4238 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4239 	struct ixgbe_hw *hw = &adapter->hw;
4240 
4241 	if (!netif_running(netdev))
4242 		return 0;
4243 
4244 	if (hw->mac.ops.update_mc_addr_list)
4245 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4246 	else
4247 		return -ENOMEM;
4248 
4249 #ifdef CONFIG_PCI_IOV
4250 	ixgbe_restore_vf_multicasts(adapter);
4251 #endif
4252 
4253 	return netdev_mc_count(netdev);
4254 }
4255 
4256 #ifdef CONFIG_PCI_IOV
4257 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4258 {
4259 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4260 	struct ixgbe_hw *hw = &adapter->hw;
4261 	int i;
4262 
4263 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4264 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4265 
4266 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4267 			hw->mac.ops.set_rar(hw, i,
4268 					    mac_table->addr,
4269 					    mac_table->pool,
4270 					    IXGBE_RAH_AV);
4271 		else
4272 			hw->mac.ops.clear_rar(hw, i);
4273 	}
4274 }
4275 
4276 #endif
4277 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4278 {
4279 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4280 	struct ixgbe_hw *hw = &adapter->hw;
4281 	int i;
4282 
4283 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4284 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4285 			continue;
4286 
4287 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4288 
4289 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4290 			hw->mac.ops.set_rar(hw, i,
4291 					    mac_table->addr,
4292 					    mac_table->pool,
4293 					    IXGBE_RAH_AV);
4294 		else
4295 			hw->mac.ops.clear_rar(hw, i);
4296 	}
4297 }
4298 
4299 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4300 {
4301 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4302 	struct ixgbe_hw *hw = &adapter->hw;
4303 	int i;
4304 
4305 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4306 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4307 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4308 	}
4309 
4310 	ixgbe_sync_mac_table(adapter);
4311 }
4312 
4313 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4314 {
4315 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4316 	struct ixgbe_hw *hw = &adapter->hw;
4317 	int i, count = 0;
4318 
4319 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4320 		/* do not count default RAR as available */
4321 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4322 			continue;
4323 
4324 		/* only count unused and addresses that belong to us */
4325 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4326 			if (mac_table->pool != pool)
4327 				continue;
4328 		}
4329 
4330 		count++;
4331 	}
4332 
4333 	return count;
4334 }
4335 
4336 /* this function destroys the first RAR entry */
4337 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4338 {
4339 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4340 	struct ixgbe_hw *hw = &adapter->hw;
4341 
4342 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4343 	mac_table->pool = VMDQ_P(0);
4344 
4345 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4346 
4347 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4348 			    IXGBE_RAH_AV);
4349 }
4350 
4351 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4352 			 const u8 *addr, u16 pool)
4353 {
4354 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4355 	struct ixgbe_hw *hw = &adapter->hw;
4356 	int i;
4357 
4358 	if (is_zero_ether_addr(addr))
4359 		return -EINVAL;
4360 
4361 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4362 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4363 			continue;
4364 
4365 		ether_addr_copy(mac_table->addr, addr);
4366 		mac_table->pool = pool;
4367 
4368 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4369 				    IXGBE_MAC_STATE_IN_USE;
4370 
4371 		ixgbe_sync_mac_table(adapter);
4372 
4373 		return i;
4374 	}
4375 
4376 	return -ENOMEM;
4377 }
4378 
4379 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4380 			 const u8 *addr, u16 pool)
4381 {
4382 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4383 	struct ixgbe_hw *hw = &adapter->hw;
4384 	int i;
4385 
4386 	if (is_zero_ether_addr(addr))
4387 		return -EINVAL;
4388 
4389 	/* search table for addr, if found clear IN_USE flag and sync */
4390 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4391 		/* we can only delete an entry if it is in use */
4392 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4393 			continue;
4394 		/* we only care about entries that belong to the given pool */
4395 		if (mac_table->pool != pool)
4396 			continue;
4397 		/* we only care about a specific MAC address */
4398 		if (!ether_addr_equal(addr, mac_table->addr))
4399 			continue;
4400 
4401 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4402 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4403 
4404 		ixgbe_sync_mac_table(adapter);
4405 
4406 		return 0;
4407 	}
4408 
4409 	return -ENOMEM;
4410 }
4411 /**
4412  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4413  * @netdev: network interface device structure
4414  *
4415  * Writes unicast address list to the RAR table.
4416  * Returns: -ENOMEM on failure/insufficient address space
4417  *                0 on no addresses written
4418  *                X on writing X addresses to the RAR table
4419  **/
4420 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4421 {
4422 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4423 	int count = 0;
4424 
4425 	/* return ENOMEM indicating insufficient memory for addresses */
4426 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter, vfn))
4427 		return -ENOMEM;
4428 
4429 	if (!netdev_uc_empty(netdev)) {
4430 		struct netdev_hw_addr *ha;
4431 		netdev_for_each_uc_addr(ha, netdev) {
4432 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4433 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4434 			count++;
4435 		}
4436 	}
4437 	return count;
4438 }
4439 
4440 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4441 {
4442 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4443 	int ret;
4444 
4445 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4446 
4447 	return min_t(int, ret, 0);
4448 }
4449 
4450 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4451 {
4452 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4453 
4454 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4455 
4456 	return 0;
4457 }
4458 
4459 /**
4460  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4461  * @netdev: network interface device structure
4462  *
4463  * The set_rx_method entry point is called whenever the unicast/multicast
4464  * address list or the network interface flags are updated.  This routine is
4465  * responsible for configuring the hardware for proper unicast, multicast and
4466  * promiscuous mode.
4467  **/
4468 void ixgbe_set_rx_mode(struct net_device *netdev)
4469 {
4470 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4471 	struct ixgbe_hw *hw = &adapter->hw;
4472 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4473 	netdev_features_t features = netdev->features;
4474 	int count;
4475 
4476 	/* Check for Promiscuous and All Multicast modes */
4477 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4478 
4479 	/* set all bits that we expect to always be set */
4480 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4481 	fctrl |= IXGBE_FCTRL_BAM;
4482 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4483 	fctrl |= IXGBE_FCTRL_PMCF;
4484 
4485 	/* clear the bits we are changing the status of */
4486 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4487 	if (netdev->flags & IFF_PROMISC) {
4488 		hw->addr_ctrl.user_set_promisc = true;
4489 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4490 		vmolr |= IXGBE_VMOLR_MPE;
4491 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4492 	} else {
4493 		if (netdev->flags & IFF_ALLMULTI) {
4494 			fctrl |= IXGBE_FCTRL_MPE;
4495 			vmolr |= IXGBE_VMOLR_MPE;
4496 		}
4497 		hw->addr_ctrl.user_set_promisc = false;
4498 	}
4499 
4500 	/*
4501 	 * Write addresses to available RAR registers, if there is not
4502 	 * sufficient space to store all the addresses then enable
4503 	 * unicast promiscuous mode
4504 	 */
4505 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4506 		fctrl |= IXGBE_FCTRL_UPE;
4507 		vmolr |= IXGBE_VMOLR_ROPE;
4508 	}
4509 
4510 	/* Write addresses to the MTA, if the attempt fails
4511 	 * then we should just turn on promiscuous mode so
4512 	 * that we can at least receive multicast traffic
4513 	 */
4514 	count = ixgbe_write_mc_addr_list(netdev);
4515 	if (count < 0) {
4516 		fctrl |= IXGBE_FCTRL_MPE;
4517 		vmolr |= IXGBE_VMOLR_MPE;
4518 	} else if (count) {
4519 		vmolr |= IXGBE_VMOLR_ROMPE;
4520 	}
4521 
4522 	if (hw->mac.type != ixgbe_mac_82598EB) {
4523 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4524 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4525 			   IXGBE_VMOLR_ROPE);
4526 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4527 	}
4528 
4529 	/* This is useful for sniffing bad packets. */
4530 	if (features & NETIF_F_RXALL) {
4531 		/* UPE and MPE will be handled by normal PROMISC logic
4532 		 * in e1000e_set_rx_mode */
4533 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4534 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4535 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4536 
4537 		fctrl &= ~(IXGBE_FCTRL_DPF);
4538 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4539 	}
4540 
4541 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4542 
4543 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4544 		ixgbe_vlan_strip_enable(adapter);
4545 	else
4546 		ixgbe_vlan_strip_disable(adapter);
4547 
4548 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4549 		ixgbe_vlan_promisc_disable(adapter);
4550 	else
4551 		ixgbe_vlan_promisc_enable(adapter);
4552 }
4553 
4554 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4555 {
4556 	int q_idx;
4557 
4558 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4559 		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
4560 		napi_enable(&adapter->q_vector[q_idx]->napi);
4561 	}
4562 }
4563 
4564 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4565 {
4566 	int q_idx;
4567 
4568 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4569 		napi_disable(&adapter->q_vector[q_idx]->napi);
4570 		while (!ixgbe_qv_disable(adapter->q_vector[q_idx])) {
4571 			pr_info("QV %d locked\n", q_idx);
4572 			usleep_range(1000, 20000);
4573 		}
4574 	}
4575 }
4576 
4577 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
4578 {
4579 	struct ixgbe_hw *hw = &adapter->hw;
4580 	u32 vxlanctrl;
4581 
4582 	if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
4583 				IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
4584 		return;
4585 
4586 	vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) && ~mask;
4587 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
4588 
4589 	if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
4590 		adapter->vxlan_port = 0;
4591 
4592 	if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
4593 		adapter->geneve_port = 0;
4594 }
4595 
4596 #ifdef CONFIG_IXGBE_DCB
4597 /**
4598  * ixgbe_configure_dcb - Configure DCB hardware
4599  * @adapter: ixgbe adapter struct
4600  *
4601  * This is called by the driver on open to configure the DCB hardware.
4602  * This is also called by the gennetlink interface when reconfiguring
4603  * the DCB state.
4604  */
4605 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4606 {
4607 	struct ixgbe_hw *hw = &adapter->hw;
4608 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4609 
4610 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4611 		if (hw->mac.type == ixgbe_mac_82598EB)
4612 			netif_set_gso_max_size(adapter->netdev, 65536);
4613 		return;
4614 	}
4615 
4616 	if (hw->mac.type == ixgbe_mac_82598EB)
4617 		netif_set_gso_max_size(adapter->netdev, 32768);
4618 
4619 #ifdef IXGBE_FCOE
4620 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4621 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4622 #endif
4623 
4624 	/* reconfigure the hardware */
4625 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4626 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4627 						DCB_TX_CONFIG);
4628 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4629 						DCB_RX_CONFIG);
4630 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4631 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4632 		ixgbe_dcb_hw_ets(&adapter->hw,
4633 				 adapter->ixgbe_ieee_ets,
4634 				 max_frame);
4635 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4636 					adapter->ixgbe_ieee_pfc->pfc_en,
4637 					adapter->ixgbe_ieee_ets->prio_tc);
4638 	}
4639 
4640 	/* Enable RSS Hash per TC */
4641 	if (hw->mac.type != ixgbe_mac_82598EB) {
4642 		u32 msb = 0;
4643 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4644 
4645 		while (rss_i) {
4646 			msb++;
4647 			rss_i >>= 1;
4648 		}
4649 
4650 		/* write msb to all 8 TCs in one write */
4651 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4652 	}
4653 }
4654 #endif
4655 
4656 /* Additional bittime to account for IXGBE framing */
4657 #define IXGBE_ETH_FRAMING 20
4658 
4659 /**
4660  * ixgbe_hpbthresh - calculate high water mark for flow control
4661  *
4662  * @adapter: board private structure to calculate for
4663  * @pb: packet buffer to calculate
4664  */
4665 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4666 {
4667 	struct ixgbe_hw *hw = &adapter->hw;
4668 	struct net_device *dev = adapter->netdev;
4669 	int link, tc, kb, marker;
4670 	u32 dv_id, rx_pba;
4671 
4672 	/* Calculate max LAN frame size */
4673 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4674 
4675 #ifdef IXGBE_FCOE
4676 	/* FCoE traffic class uses FCOE jumbo frames */
4677 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4678 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4679 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4680 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4681 #endif
4682 
4683 	/* Calculate delay value for device */
4684 	switch (hw->mac.type) {
4685 	case ixgbe_mac_X540:
4686 	case ixgbe_mac_X550:
4687 	case ixgbe_mac_X550EM_x:
4688 	case ixgbe_mac_x550em_a:
4689 		dv_id = IXGBE_DV_X540(link, tc);
4690 		break;
4691 	default:
4692 		dv_id = IXGBE_DV(link, tc);
4693 		break;
4694 	}
4695 
4696 	/* Loopback switch introduces additional latency */
4697 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4698 		dv_id += IXGBE_B2BT(tc);
4699 
4700 	/* Delay value is calculated in bit times convert to KB */
4701 	kb = IXGBE_BT2KB(dv_id);
4702 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4703 
4704 	marker = rx_pba - kb;
4705 
4706 	/* It is possible that the packet buffer is not large enough
4707 	 * to provide required headroom. In this case throw an error
4708 	 * to user and a do the best we can.
4709 	 */
4710 	if (marker < 0) {
4711 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4712 			    "headroom to support flow control."
4713 			    "Decrease MTU or number of traffic classes\n", pb);
4714 		marker = tc + 1;
4715 	}
4716 
4717 	return marker;
4718 }
4719 
4720 /**
4721  * ixgbe_lpbthresh - calculate low water mark for for flow control
4722  *
4723  * @adapter: board private structure to calculate for
4724  * @pb: packet buffer to calculate
4725  */
4726 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4727 {
4728 	struct ixgbe_hw *hw = &adapter->hw;
4729 	struct net_device *dev = adapter->netdev;
4730 	int tc;
4731 	u32 dv_id;
4732 
4733 	/* Calculate max LAN frame size */
4734 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4735 
4736 #ifdef IXGBE_FCOE
4737 	/* FCoE traffic class uses FCOE jumbo frames */
4738 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4739 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4740 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4741 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4742 #endif
4743 
4744 	/* Calculate delay value for device */
4745 	switch (hw->mac.type) {
4746 	case ixgbe_mac_X540:
4747 	case ixgbe_mac_X550:
4748 	case ixgbe_mac_X550EM_x:
4749 	case ixgbe_mac_x550em_a:
4750 		dv_id = IXGBE_LOW_DV_X540(tc);
4751 		break;
4752 	default:
4753 		dv_id = IXGBE_LOW_DV(tc);
4754 		break;
4755 	}
4756 
4757 	/* Delay value is calculated in bit times convert to KB */
4758 	return IXGBE_BT2KB(dv_id);
4759 }
4760 
4761 /*
4762  * ixgbe_pbthresh_setup - calculate and setup high low water marks
4763  */
4764 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4765 {
4766 	struct ixgbe_hw *hw = &adapter->hw;
4767 	int num_tc = netdev_get_num_tc(adapter->netdev);
4768 	int i;
4769 
4770 	if (!num_tc)
4771 		num_tc = 1;
4772 
4773 	for (i = 0; i < num_tc; i++) {
4774 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4775 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4776 
4777 		/* Low water marks must not be larger than high water marks */
4778 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4779 			hw->fc.low_water[i] = 0;
4780 	}
4781 
4782 	for (; i < MAX_TRAFFIC_CLASS; i++)
4783 		hw->fc.high_water[i] = 0;
4784 }
4785 
4786 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4787 {
4788 	struct ixgbe_hw *hw = &adapter->hw;
4789 	int hdrm;
4790 	u8 tc = netdev_get_num_tc(adapter->netdev);
4791 
4792 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4793 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4794 		hdrm = 32 << adapter->fdir_pballoc;
4795 	else
4796 		hdrm = 0;
4797 
4798 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4799 	ixgbe_pbthresh_setup(adapter);
4800 }
4801 
4802 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4803 {
4804 	struct ixgbe_hw *hw = &adapter->hw;
4805 	struct hlist_node *node2;
4806 	struct ixgbe_fdir_filter *filter;
4807 
4808 	spin_lock(&adapter->fdir_perfect_lock);
4809 
4810 	if (!hlist_empty(&adapter->fdir_filter_list))
4811 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4812 
4813 	hlist_for_each_entry_safe(filter, node2,
4814 				  &adapter->fdir_filter_list, fdir_node) {
4815 		ixgbe_fdir_write_perfect_filter_82599(hw,
4816 				&filter->filter,
4817 				filter->sw_idx,
4818 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4819 				IXGBE_FDIR_DROP_QUEUE :
4820 				adapter->rx_ring[filter->action]->reg_idx);
4821 	}
4822 
4823 	spin_unlock(&adapter->fdir_perfect_lock);
4824 }
4825 
4826 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4827 				      struct ixgbe_adapter *adapter)
4828 {
4829 	struct ixgbe_hw *hw = &adapter->hw;
4830 	u32 vmolr;
4831 
4832 	/* No unicast promiscuous support for VMDQ devices. */
4833 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4834 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4835 
4836 	/* clear the affected bit */
4837 	vmolr &= ~IXGBE_VMOLR_MPE;
4838 
4839 	if (dev->flags & IFF_ALLMULTI) {
4840 		vmolr |= IXGBE_VMOLR_MPE;
4841 	} else {
4842 		vmolr |= IXGBE_VMOLR_ROMPE;
4843 		hw->mac.ops.update_mc_addr_list(hw, dev);
4844 	}
4845 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4846 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4847 }
4848 
4849 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4850 {
4851 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4852 	int rss_i = adapter->num_rx_queues_per_pool;
4853 	struct ixgbe_hw *hw = &adapter->hw;
4854 	u16 pool = vadapter->pool;
4855 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4856 		      IXGBE_PSRTYPE_UDPHDR |
4857 		      IXGBE_PSRTYPE_IPV4HDR |
4858 		      IXGBE_PSRTYPE_L2HDR |
4859 		      IXGBE_PSRTYPE_IPV6HDR;
4860 
4861 	if (hw->mac.type == ixgbe_mac_82598EB)
4862 		return;
4863 
4864 	if (rss_i > 3)
4865 		psrtype |= 2u << 29;
4866 	else if (rss_i > 1)
4867 		psrtype |= 1u << 29;
4868 
4869 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4870 }
4871 
4872 /**
4873  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4874  * @rx_ring: ring to free buffers from
4875  **/
4876 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4877 {
4878 	struct device *dev = rx_ring->dev;
4879 	unsigned long size;
4880 	u16 i;
4881 
4882 	/* ring already cleared, nothing to do */
4883 	if (!rx_ring->rx_buffer_info)
4884 		return;
4885 
4886 	/* Free all the Rx ring sk_buffs */
4887 	for (i = 0; i < rx_ring->count; i++) {
4888 		struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4889 
4890 		if (rx_buffer->skb) {
4891 			struct sk_buff *skb = rx_buffer->skb;
4892 			if (IXGBE_CB(skb)->page_released)
4893 				dma_unmap_page(dev,
4894 					       IXGBE_CB(skb)->dma,
4895 					       ixgbe_rx_bufsz(rx_ring),
4896 					       DMA_FROM_DEVICE);
4897 			dev_kfree_skb(skb);
4898 			rx_buffer->skb = NULL;
4899 		}
4900 
4901 		if (!rx_buffer->page)
4902 			continue;
4903 
4904 		dma_unmap_page(dev, rx_buffer->dma,
4905 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
4906 		__free_pages(rx_buffer->page, ixgbe_rx_pg_order(rx_ring));
4907 
4908 		rx_buffer->page = NULL;
4909 	}
4910 
4911 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4912 	memset(rx_ring->rx_buffer_info, 0, size);
4913 
4914 	/* Zero out the descriptor ring */
4915 	memset(rx_ring->desc, 0, rx_ring->size);
4916 
4917 	rx_ring->next_to_alloc = 0;
4918 	rx_ring->next_to_clean = 0;
4919 	rx_ring->next_to_use = 0;
4920 }
4921 
4922 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4923 				   struct ixgbe_ring *rx_ring)
4924 {
4925 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4926 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4927 
4928 	/* shutdown specific queue receive and wait for dma to settle */
4929 	ixgbe_disable_rx_queue(adapter, rx_ring);
4930 	usleep_range(10000, 20000);
4931 	ixgbe_irq_disable_queues(adapter, BIT_ULL(index));
4932 	ixgbe_clean_rx_ring(rx_ring);
4933 	rx_ring->l2_accel_priv = NULL;
4934 }
4935 
4936 static int ixgbe_fwd_ring_down(struct net_device *vdev,
4937 			       struct ixgbe_fwd_adapter *accel)
4938 {
4939 	struct ixgbe_adapter *adapter = accel->real_adapter;
4940 	unsigned int rxbase = accel->rx_base_queue;
4941 	unsigned int txbase = accel->tx_base_queue;
4942 	int i;
4943 
4944 	netif_tx_stop_all_queues(vdev);
4945 
4946 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4947 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4948 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4949 	}
4950 
4951 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4952 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4953 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4954 	}
4955 
4956 
4957 	return 0;
4958 }
4959 
4960 static int ixgbe_fwd_ring_up(struct net_device *vdev,
4961 			     struct ixgbe_fwd_adapter *accel)
4962 {
4963 	struct ixgbe_adapter *adapter = accel->real_adapter;
4964 	unsigned int rxbase, txbase, queues;
4965 	int i, baseq, err = 0;
4966 
4967 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4968 		return 0;
4969 
4970 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4971 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4972 		   accel->pool, adapter->num_rx_pools,
4973 		   baseq, baseq + adapter->num_rx_queues_per_pool,
4974 		   adapter->fwd_bitmask);
4975 
4976 	accel->netdev = vdev;
4977 	accel->rx_base_queue = rxbase = baseq;
4978 	accel->tx_base_queue = txbase = baseq;
4979 
4980 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4981 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4982 
4983 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4984 		adapter->rx_ring[rxbase + i]->netdev = vdev;
4985 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
4986 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
4987 	}
4988 
4989 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4990 		adapter->tx_ring[txbase + i]->netdev = vdev;
4991 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
4992 	}
4993 
4994 	queues = min_t(unsigned int,
4995 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
4996 	err = netif_set_real_num_tx_queues(vdev, queues);
4997 	if (err)
4998 		goto fwd_queue_err;
4999 
5000 	err = netif_set_real_num_rx_queues(vdev, queues);
5001 	if (err)
5002 		goto fwd_queue_err;
5003 
5004 	if (is_valid_ether_addr(vdev->dev_addr))
5005 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
5006 
5007 	ixgbe_fwd_psrtype(accel);
5008 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
5009 	return err;
5010 fwd_queue_err:
5011 	ixgbe_fwd_ring_down(vdev, accel);
5012 	return err;
5013 }
5014 
5015 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5016 {
5017 	struct net_device *upper;
5018 	struct list_head *iter;
5019 	int err;
5020 
5021 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5022 		if (netif_is_macvlan(upper)) {
5023 			struct macvlan_dev *dfwd = netdev_priv(upper);
5024 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
5025 
5026 			if (dfwd->fwd_priv) {
5027 				err = ixgbe_fwd_ring_up(upper, vadapter);
5028 				if (err)
5029 					continue;
5030 			}
5031 		}
5032 	}
5033 }
5034 
5035 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5036 {
5037 	struct ixgbe_hw *hw = &adapter->hw;
5038 
5039 	ixgbe_configure_pb(adapter);
5040 #ifdef CONFIG_IXGBE_DCB
5041 	ixgbe_configure_dcb(adapter);
5042 #endif
5043 	/*
5044 	 * We must restore virtualization before VLANs or else
5045 	 * the VLVF registers will not be populated
5046 	 */
5047 	ixgbe_configure_virtualization(adapter);
5048 
5049 	ixgbe_set_rx_mode(adapter->netdev);
5050 	ixgbe_restore_vlan(adapter);
5051 
5052 	switch (hw->mac.type) {
5053 	case ixgbe_mac_82599EB:
5054 	case ixgbe_mac_X540:
5055 		hw->mac.ops.disable_rx_buff(hw);
5056 		break;
5057 	default:
5058 		break;
5059 	}
5060 
5061 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5062 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5063 						adapter->fdir_pballoc);
5064 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5065 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5066 					      adapter->fdir_pballoc);
5067 		ixgbe_fdir_filter_restore(adapter);
5068 	}
5069 
5070 	switch (hw->mac.type) {
5071 	case ixgbe_mac_82599EB:
5072 	case ixgbe_mac_X540:
5073 		hw->mac.ops.enable_rx_buff(hw);
5074 		break;
5075 	default:
5076 		break;
5077 	}
5078 
5079 #ifdef CONFIG_IXGBE_DCA
5080 	/* configure DCA */
5081 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5082 		ixgbe_setup_dca(adapter);
5083 #endif /* CONFIG_IXGBE_DCA */
5084 
5085 #ifdef IXGBE_FCOE
5086 	/* configure FCoE L2 filters, redirection table, and Rx control */
5087 	ixgbe_configure_fcoe(adapter);
5088 
5089 #endif /* IXGBE_FCOE */
5090 	ixgbe_configure_tx(adapter);
5091 	ixgbe_configure_rx(adapter);
5092 	ixgbe_configure_dfwd(adapter);
5093 }
5094 
5095 /**
5096  * ixgbe_sfp_link_config - set up SFP+ link
5097  * @adapter: pointer to private adapter struct
5098  **/
5099 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5100 {
5101 	/*
5102 	 * We are assuming the worst case scenario here, and that
5103 	 * is that an SFP was inserted/removed after the reset
5104 	 * but before SFP detection was enabled.  As such the best
5105 	 * solution is to just start searching as soon as we start
5106 	 */
5107 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5108 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5109 
5110 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5111 	adapter->sfp_poll_time = 0;
5112 }
5113 
5114 /**
5115  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5116  * @hw: pointer to private hardware struct
5117  *
5118  * Returns 0 on success, negative on failure
5119  **/
5120 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5121 {
5122 	u32 speed;
5123 	bool autoneg, link_up = false;
5124 	int ret = IXGBE_ERR_LINK_SETUP;
5125 
5126 	if (hw->mac.ops.check_link)
5127 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5128 
5129 	if (ret)
5130 		return ret;
5131 
5132 	speed = hw->phy.autoneg_advertised;
5133 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5134 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5135 							&autoneg);
5136 	if (ret)
5137 		return ret;
5138 
5139 	if (hw->mac.ops.setup_link)
5140 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5141 
5142 	return ret;
5143 }
5144 
5145 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5146 {
5147 	struct ixgbe_hw *hw = &adapter->hw;
5148 	u32 gpie = 0;
5149 
5150 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5151 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5152 		       IXGBE_GPIE_OCD;
5153 		gpie |= IXGBE_GPIE_EIAME;
5154 		/*
5155 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5156 		 * this saves a register write for every interrupt
5157 		 */
5158 		switch (hw->mac.type) {
5159 		case ixgbe_mac_82598EB:
5160 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5161 			break;
5162 		case ixgbe_mac_82599EB:
5163 		case ixgbe_mac_X540:
5164 		case ixgbe_mac_X550:
5165 		case ixgbe_mac_X550EM_x:
5166 		case ixgbe_mac_x550em_a:
5167 		default:
5168 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5169 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5170 			break;
5171 		}
5172 	} else {
5173 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5174 		 * specifically only auto mask tx and rx interrupts */
5175 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5176 	}
5177 
5178 	/* XXX: to interrupt immediately for EICS writes, enable this */
5179 	/* gpie |= IXGBE_GPIE_EIMEN; */
5180 
5181 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5182 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5183 
5184 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5185 		case IXGBE_82599_VMDQ_8Q_MASK:
5186 			gpie |= IXGBE_GPIE_VTMODE_16;
5187 			break;
5188 		case IXGBE_82599_VMDQ_4Q_MASK:
5189 			gpie |= IXGBE_GPIE_VTMODE_32;
5190 			break;
5191 		default:
5192 			gpie |= IXGBE_GPIE_VTMODE_64;
5193 			break;
5194 		}
5195 	}
5196 
5197 	/* Enable Thermal over heat sensor interrupt */
5198 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5199 		switch (adapter->hw.mac.type) {
5200 		case ixgbe_mac_82599EB:
5201 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5202 			break;
5203 		default:
5204 			break;
5205 		}
5206 	}
5207 
5208 	/* Enable fan failure interrupt */
5209 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5210 		gpie |= IXGBE_SDP1_GPIEN(hw);
5211 
5212 	switch (hw->mac.type) {
5213 	case ixgbe_mac_82599EB:
5214 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5215 		break;
5216 	case ixgbe_mac_X550EM_x:
5217 	case ixgbe_mac_x550em_a:
5218 		gpie |= IXGBE_SDP0_GPIEN_X540;
5219 		break;
5220 	default:
5221 		break;
5222 	}
5223 
5224 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5225 }
5226 
5227 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5228 {
5229 	struct ixgbe_hw *hw = &adapter->hw;
5230 	int err;
5231 	u32 ctrl_ext;
5232 
5233 	ixgbe_get_hw_control(adapter);
5234 	ixgbe_setup_gpie(adapter);
5235 
5236 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5237 		ixgbe_configure_msix(adapter);
5238 	else
5239 		ixgbe_configure_msi_and_legacy(adapter);
5240 
5241 	/* enable the optics for 82599 SFP+ fiber */
5242 	if (hw->mac.ops.enable_tx_laser)
5243 		hw->mac.ops.enable_tx_laser(hw);
5244 
5245 	if (hw->phy.ops.set_phy_power)
5246 		hw->phy.ops.set_phy_power(hw, true);
5247 
5248 	smp_mb__before_atomic();
5249 	clear_bit(__IXGBE_DOWN, &adapter->state);
5250 	ixgbe_napi_enable_all(adapter);
5251 
5252 	if (ixgbe_is_sfp(hw)) {
5253 		ixgbe_sfp_link_config(adapter);
5254 	} else {
5255 		err = ixgbe_non_sfp_link_config(hw);
5256 		if (err)
5257 			e_err(probe, "link_config FAILED %d\n", err);
5258 	}
5259 
5260 	/* clear any pending interrupts, may auto mask */
5261 	IXGBE_READ_REG(hw, IXGBE_EICR);
5262 	ixgbe_irq_enable(adapter, true, true);
5263 
5264 	/*
5265 	 * If this adapter has a fan, check to see if we had a failure
5266 	 * before we enabled the interrupt.
5267 	 */
5268 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5269 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5270 		if (esdp & IXGBE_ESDP_SDP1)
5271 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5272 	}
5273 
5274 	/* bring the link up in the watchdog, this could race with our first
5275 	 * link up interrupt but shouldn't be a problem */
5276 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5277 	adapter->link_check_timeout = jiffies;
5278 	mod_timer(&adapter->service_timer, jiffies);
5279 
5280 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5281 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5282 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5283 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5284 }
5285 
5286 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5287 {
5288 	WARN_ON(in_interrupt());
5289 	/* put off any impending NetWatchDogTimeout */
5290 	netif_trans_update(adapter->netdev);
5291 
5292 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5293 		usleep_range(1000, 2000);
5294 	ixgbe_down(adapter);
5295 	/*
5296 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5297 	 * back up to give the VFs time to respond to the reset.  The
5298 	 * two second wait is based upon the watchdog timer cycle in
5299 	 * the VF driver.
5300 	 */
5301 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5302 		msleep(2000);
5303 	ixgbe_up(adapter);
5304 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5305 }
5306 
5307 void ixgbe_up(struct ixgbe_adapter *adapter)
5308 {
5309 	/* hardware has been reset, we need to reload some things */
5310 	ixgbe_configure(adapter);
5311 
5312 	ixgbe_up_complete(adapter);
5313 }
5314 
5315 void ixgbe_reset(struct ixgbe_adapter *adapter)
5316 {
5317 	struct ixgbe_hw *hw = &adapter->hw;
5318 	struct net_device *netdev = adapter->netdev;
5319 	int err;
5320 
5321 	if (ixgbe_removed(hw->hw_addr))
5322 		return;
5323 	/* lock SFP init bit to prevent race conditions with the watchdog */
5324 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5325 		usleep_range(1000, 2000);
5326 
5327 	/* clear all SFP and link config related flags while holding SFP_INIT */
5328 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5329 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5330 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5331 
5332 	err = hw->mac.ops.init_hw(hw);
5333 	switch (err) {
5334 	case 0:
5335 	case IXGBE_ERR_SFP_NOT_PRESENT:
5336 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5337 		break;
5338 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5339 		e_dev_err("master disable timed out\n");
5340 		break;
5341 	case IXGBE_ERR_EEPROM_VERSION:
5342 		/* We are running on a pre-production device, log a warning */
5343 		e_dev_warn("This device is a pre-production adapter/LOM. "
5344 			   "Please be aware there may be issues associated with "
5345 			   "your hardware.  If you are experiencing problems "
5346 			   "please contact your Intel or hardware "
5347 			   "representative who provided you with this "
5348 			   "hardware.\n");
5349 		break;
5350 	default:
5351 		e_dev_err("Hardware Error: %d\n", err);
5352 	}
5353 
5354 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5355 
5356 	/* flush entries out of MAC table */
5357 	ixgbe_flush_sw_mac_table(adapter);
5358 	__dev_uc_unsync(netdev, NULL);
5359 
5360 	/* do not flush user set addresses */
5361 	ixgbe_mac_set_default_filter(adapter);
5362 
5363 	/* update SAN MAC vmdq pool selection */
5364 	if (hw->mac.san_mac_rar_index)
5365 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5366 
5367 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5368 		ixgbe_ptp_reset(adapter);
5369 
5370 	if (hw->phy.ops.set_phy_power) {
5371 		if (!netif_running(adapter->netdev) && !adapter->wol)
5372 			hw->phy.ops.set_phy_power(hw, false);
5373 		else
5374 			hw->phy.ops.set_phy_power(hw, true);
5375 	}
5376 }
5377 
5378 /**
5379  * ixgbe_clean_tx_ring - Free Tx Buffers
5380  * @tx_ring: ring to be cleaned
5381  **/
5382 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5383 {
5384 	struct ixgbe_tx_buffer *tx_buffer_info;
5385 	unsigned long size;
5386 	u16 i;
5387 
5388 	/* ring already cleared, nothing to do */
5389 	if (!tx_ring->tx_buffer_info)
5390 		return;
5391 
5392 	/* Free all the Tx ring sk_buffs */
5393 	for (i = 0; i < tx_ring->count; i++) {
5394 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
5395 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
5396 	}
5397 
5398 	netdev_tx_reset_queue(txring_txq(tx_ring));
5399 
5400 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5401 	memset(tx_ring->tx_buffer_info, 0, size);
5402 
5403 	/* Zero out the descriptor ring */
5404 	memset(tx_ring->desc, 0, tx_ring->size);
5405 
5406 	tx_ring->next_to_use = 0;
5407 	tx_ring->next_to_clean = 0;
5408 }
5409 
5410 /**
5411  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5412  * @adapter: board private structure
5413  **/
5414 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5415 {
5416 	int i;
5417 
5418 	for (i = 0; i < adapter->num_rx_queues; i++)
5419 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5420 }
5421 
5422 /**
5423  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5424  * @adapter: board private structure
5425  **/
5426 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5427 {
5428 	int i;
5429 
5430 	for (i = 0; i < adapter->num_tx_queues; i++)
5431 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5432 }
5433 
5434 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5435 {
5436 	struct hlist_node *node2;
5437 	struct ixgbe_fdir_filter *filter;
5438 
5439 	spin_lock(&adapter->fdir_perfect_lock);
5440 
5441 	hlist_for_each_entry_safe(filter, node2,
5442 				  &adapter->fdir_filter_list, fdir_node) {
5443 		hlist_del(&filter->fdir_node);
5444 		kfree(filter);
5445 	}
5446 	adapter->fdir_filter_count = 0;
5447 
5448 	spin_unlock(&adapter->fdir_perfect_lock);
5449 }
5450 
5451 void ixgbe_down(struct ixgbe_adapter *adapter)
5452 {
5453 	struct net_device *netdev = adapter->netdev;
5454 	struct ixgbe_hw *hw = &adapter->hw;
5455 	struct net_device *upper;
5456 	struct list_head *iter;
5457 	int i;
5458 
5459 	/* signal that we are down to the interrupt handler */
5460 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5461 		return; /* do nothing if already down */
5462 
5463 	/* disable receives */
5464 	hw->mac.ops.disable_rx(hw);
5465 
5466 	/* disable all enabled rx queues */
5467 	for (i = 0; i < adapter->num_rx_queues; i++)
5468 		/* this call also flushes the previous write */
5469 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5470 
5471 	usleep_range(10000, 20000);
5472 
5473 	netif_tx_stop_all_queues(netdev);
5474 
5475 	/* call carrier off first to avoid false dev_watchdog timeouts */
5476 	netif_carrier_off(netdev);
5477 	netif_tx_disable(netdev);
5478 
5479 	/* disable any upper devices */
5480 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5481 		if (netif_is_macvlan(upper)) {
5482 			struct macvlan_dev *vlan = netdev_priv(upper);
5483 
5484 			if (vlan->fwd_priv) {
5485 				netif_tx_stop_all_queues(upper);
5486 				netif_carrier_off(upper);
5487 				netif_tx_disable(upper);
5488 			}
5489 		}
5490 	}
5491 
5492 	ixgbe_irq_disable(adapter);
5493 
5494 	ixgbe_napi_disable_all(adapter);
5495 
5496 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
5497 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5498 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5499 
5500 	del_timer_sync(&adapter->service_timer);
5501 
5502 	if (adapter->num_vfs) {
5503 		/* Clear EITR Select mapping */
5504 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5505 
5506 		/* Mark all the VFs as inactive */
5507 		for (i = 0 ; i < adapter->num_vfs; i++)
5508 			adapter->vfinfo[i].clear_to_send = false;
5509 
5510 		/* ping all the active vfs to let them know we are going down */
5511 		ixgbe_ping_all_vfs(adapter);
5512 
5513 		/* Disable all VFTE/VFRE TX/RX */
5514 		ixgbe_disable_tx_rx(adapter);
5515 	}
5516 
5517 	/* disable transmits in the hardware now that interrupts are off */
5518 	for (i = 0; i < adapter->num_tx_queues; i++) {
5519 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5520 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5521 	}
5522 
5523 	/* Disable the Tx DMA engine on 82599 and later MAC */
5524 	switch (hw->mac.type) {
5525 	case ixgbe_mac_82599EB:
5526 	case ixgbe_mac_X540:
5527 	case ixgbe_mac_X550:
5528 	case ixgbe_mac_X550EM_x:
5529 	case ixgbe_mac_x550em_a:
5530 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5531 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5532 				 ~IXGBE_DMATXCTL_TE));
5533 		break;
5534 	default:
5535 		break;
5536 	}
5537 
5538 	if (!pci_channel_offline(adapter->pdev))
5539 		ixgbe_reset(adapter);
5540 
5541 	/* power down the optics for 82599 SFP+ fiber */
5542 	if (hw->mac.ops.disable_tx_laser)
5543 		hw->mac.ops.disable_tx_laser(hw);
5544 
5545 	ixgbe_clean_all_tx_rings(adapter);
5546 	ixgbe_clean_all_rx_rings(adapter);
5547 }
5548 
5549 /**
5550  * ixgbe_tx_timeout - Respond to a Tx Hang
5551  * @netdev: network interface device structure
5552  **/
5553 static void ixgbe_tx_timeout(struct net_device *netdev)
5554 {
5555 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5556 
5557 	/* Do the reset outside of interrupt context */
5558 	ixgbe_tx_timeout_reset(adapter);
5559 }
5560 
5561 #ifdef CONFIG_IXGBE_DCB
5562 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
5563 {
5564 	struct ixgbe_hw *hw = &adapter->hw;
5565 	struct tc_configuration *tc;
5566 	int j;
5567 
5568 	switch (hw->mac.type) {
5569 	case ixgbe_mac_82598EB:
5570 	case ixgbe_mac_82599EB:
5571 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5572 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5573 		break;
5574 	case ixgbe_mac_X540:
5575 	case ixgbe_mac_X550:
5576 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5577 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5578 		break;
5579 	case ixgbe_mac_X550EM_x:
5580 	case ixgbe_mac_x550em_a:
5581 	default:
5582 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
5583 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
5584 		break;
5585 	}
5586 
5587 	/* Configure DCB traffic classes */
5588 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5589 		tc = &adapter->dcb_cfg.tc_config[j];
5590 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5591 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5592 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5593 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5594 		tc->dcb_pfc = pfc_disabled;
5595 	}
5596 
5597 	/* Initialize default user to priority mapping, UPx->TC0 */
5598 	tc = &adapter->dcb_cfg.tc_config[0];
5599 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5600 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5601 
5602 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5603 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5604 	adapter->dcb_cfg.pfc_mode_enable = false;
5605 	adapter->dcb_set_bitmap = 0x00;
5606 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
5607 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5608 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5609 	       sizeof(adapter->temp_dcb_cfg));
5610 }
5611 #endif
5612 
5613 /**
5614  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5615  * @adapter: board private structure to initialize
5616  *
5617  * ixgbe_sw_init initializes the Adapter private data structure.
5618  * Fields are initialized based on PCI device information and
5619  * OS network device settings (MTU size).
5620  **/
5621 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
5622 {
5623 	struct ixgbe_hw *hw = &adapter->hw;
5624 	struct pci_dev *pdev = adapter->pdev;
5625 	unsigned int rss, fdir;
5626 	u32 fwsm;
5627 	int i;
5628 
5629 	/* PCI config space info */
5630 
5631 	hw->vendor_id = pdev->vendor;
5632 	hw->device_id = pdev->device;
5633 	hw->revision_id = pdev->revision;
5634 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5635 	hw->subsystem_device_id = pdev->subsystem_device;
5636 
5637 	/* Set common capability flags and settings */
5638 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5639 	adapter->ring_feature[RING_F_RSS].limit = rss;
5640 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5641 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5642 	adapter->atr_sample_rate = 20;
5643 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5644 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5645 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5646 #ifdef CONFIG_IXGBE_DCA
5647 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5648 #endif
5649 #ifdef CONFIG_IXGBE_DCB
5650 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
5651 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
5652 #endif
5653 #ifdef IXGBE_FCOE
5654 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5655 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5656 #ifdef CONFIG_IXGBE_DCB
5657 	/* Default traffic class to use for FCoE */
5658 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5659 #endif /* CONFIG_IXGBE_DCB */
5660 #endif /* IXGBE_FCOE */
5661 
5662 	/* initialize static ixgbe jump table entries */
5663 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
5664 					  GFP_KERNEL);
5665 	if (!adapter->jump_tables[0])
5666 		return -ENOMEM;
5667 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
5668 
5669 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
5670 		adapter->jump_tables[i] = NULL;
5671 
5672 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5673 				     hw->mac.num_rar_entries,
5674 				     GFP_ATOMIC);
5675 	if (!adapter->mac_table)
5676 		return -ENOMEM;
5677 
5678 	/* Set MAC specific capability flags and exceptions */
5679 	switch (hw->mac.type) {
5680 	case ixgbe_mac_82598EB:
5681 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5682 
5683 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5684 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5685 
5686 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5687 		adapter->ring_feature[RING_F_FDIR].limit = 0;
5688 		adapter->atr_sample_rate = 0;
5689 		adapter->fdir_pballoc = 0;
5690 #ifdef IXGBE_FCOE
5691 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5692 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5693 #ifdef CONFIG_IXGBE_DCB
5694 		adapter->fcoe.up = 0;
5695 #endif /* IXGBE_DCB */
5696 #endif /* IXGBE_FCOE */
5697 		break;
5698 	case ixgbe_mac_82599EB:
5699 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5700 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5701 		break;
5702 	case ixgbe_mac_X540:
5703 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
5704 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5705 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5706 		break;
5707 	case ixgbe_mac_x550em_a:
5708 		adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
5709 	/* fall through */
5710 	case ixgbe_mac_X550EM_x:
5711 #ifdef CONFIG_IXGBE_DCB
5712 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
5713 #endif
5714 #ifdef IXGBE_FCOE
5715 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5716 #ifdef CONFIG_IXGBE_DCB
5717 		adapter->fcoe.up = 0;
5718 #endif /* IXGBE_DCB */
5719 #endif /* IXGBE_FCOE */
5720 	/* Fall Through */
5721 	case ixgbe_mac_X550:
5722 #ifdef CONFIG_IXGBE_DCA
5723 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5724 #endif
5725 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
5726 		break;
5727 	default:
5728 		break;
5729 	}
5730 
5731 #ifdef IXGBE_FCOE
5732 	/* FCoE support exists, always init the FCoE lock */
5733 	spin_lock_init(&adapter->fcoe.lock);
5734 
5735 #endif
5736 	/* n-tuple support exists, always init our spinlock */
5737 	spin_lock_init(&adapter->fdir_perfect_lock);
5738 
5739 #ifdef CONFIG_IXGBE_DCB
5740 	ixgbe_init_dcb(adapter);
5741 #endif
5742 
5743 	/* default flow control settings */
5744 	hw->fc.requested_mode = ixgbe_fc_full;
5745 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5746 	ixgbe_pbthresh_setup(adapter);
5747 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5748 	hw->fc.send_xon = true;
5749 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5750 
5751 #ifdef CONFIG_PCI_IOV
5752 	if (max_vfs > 0)
5753 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5754 
5755 	/* assign number of SR-IOV VFs */
5756 	if (hw->mac.type != ixgbe_mac_82598EB) {
5757 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5758 			adapter->num_vfs = 0;
5759 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5760 		} else {
5761 			adapter->num_vfs = max_vfs;
5762 		}
5763 	}
5764 #endif /* CONFIG_PCI_IOV */
5765 
5766 	/* enable itr by default in dynamic mode */
5767 	adapter->rx_itr_setting = 1;
5768 	adapter->tx_itr_setting = 1;
5769 
5770 	/* set default ring sizes */
5771 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5772 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5773 
5774 	/* set default work limits */
5775 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5776 
5777 	/* initialize eeprom parameters */
5778 	if (ixgbe_init_eeprom_params_generic(hw)) {
5779 		e_dev_err("EEPROM initialization failed\n");
5780 		return -EIO;
5781 	}
5782 
5783 	/* PF holds first pool slot */
5784 	set_bit(0, &adapter->fwd_bitmask);
5785 	set_bit(__IXGBE_DOWN, &adapter->state);
5786 
5787 	return 0;
5788 }
5789 
5790 /**
5791  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5792  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5793  *
5794  * Return 0 on success, negative on failure
5795  **/
5796 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5797 {
5798 	struct device *dev = tx_ring->dev;
5799 	int orig_node = dev_to_node(dev);
5800 	int ring_node = -1;
5801 	int size;
5802 
5803 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5804 
5805 	if (tx_ring->q_vector)
5806 		ring_node = tx_ring->q_vector->numa_node;
5807 
5808 	tx_ring->tx_buffer_info = vzalloc_node(size, ring_node);
5809 	if (!tx_ring->tx_buffer_info)
5810 		tx_ring->tx_buffer_info = vzalloc(size);
5811 	if (!tx_ring->tx_buffer_info)
5812 		goto err;
5813 
5814 	u64_stats_init(&tx_ring->syncp);
5815 
5816 	/* round up to nearest 4K */
5817 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5818 	tx_ring->size = ALIGN(tx_ring->size, 4096);
5819 
5820 	set_dev_node(dev, ring_node);
5821 	tx_ring->desc = dma_alloc_coherent(dev,
5822 					   tx_ring->size,
5823 					   &tx_ring->dma,
5824 					   GFP_KERNEL);
5825 	set_dev_node(dev, orig_node);
5826 	if (!tx_ring->desc)
5827 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5828 						   &tx_ring->dma, GFP_KERNEL);
5829 	if (!tx_ring->desc)
5830 		goto err;
5831 
5832 	tx_ring->next_to_use = 0;
5833 	tx_ring->next_to_clean = 0;
5834 	return 0;
5835 
5836 err:
5837 	vfree(tx_ring->tx_buffer_info);
5838 	tx_ring->tx_buffer_info = NULL;
5839 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5840 	return -ENOMEM;
5841 }
5842 
5843 /**
5844  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5845  * @adapter: board private structure
5846  *
5847  * If this function returns with an error, then it's possible one or
5848  * more of the rings is populated (while the rest are not).  It is the
5849  * callers duty to clean those orphaned rings.
5850  *
5851  * Return 0 on success, negative on failure
5852  **/
5853 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5854 {
5855 	int i, err = 0;
5856 
5857 	for (i = 0; i < adapter->num_tx_queues; i++) {
5858 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5859 		if (!err)
5860 			continue;
5861 
5862 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5863 		goto err_setup_tx;
5864 	}
5865 
5866 	return 0;
5867 err_setup_tx:
5868 	/* rewind the index freeing the rings as we go */
5869 	while (i--)
5870 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5871 	return err;
5872 }
5873 
5874 /**
5875  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5876  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5877  *
5878  * Returns 0 on success, negative on failure
5879  **/
5880 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5881 {
5882 	struct device *dev = rx_ring->dev;
5883 	int orig_node = dev_to_node(dev);
5884 	int ring_node = -1;
5885 	int size;
5886 
5887 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5888 
5889 	if (rx_ring->q_vector)
5890 		ring_node = rx_ring->q_vector->numa_node;
5891 
5892 	rx_ring->rx_buffer_info = vzalloc_node(size, ring_node);
5893 	if (!rx_ring->rx_buffer_info)
5894 		rx_ring->rx_buffer_info = vzalloc(size);
5895 	if (!rx_ring->rx_buffer_info)
5896 		goto err;
5897 
5898 	u64_stats_init(&rx_ring->syncp);
5899 
5900 	/* Round up to nearest 4K */
5901 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5902 	rx_ring->size = ALIGN(rx_ring->size, 4096);
5903 
5904 	set_dev_node(dev, ring_node);
5905 	rx_ring->desc = dma_alloc_coherent(dev,
5906 					   rx_ring->size,
5907 					   &rx_ring->dma,
5908 					   GFP_KERNEL);
5909 	set_dev_node(dev, orig_node);
5910 	if (!rx_ring->desc)
5911 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5912 						   &rx_ring->dma, GFP_KERNEL);
5913 	if (!rx_ring->desc)
5914 		goto err;
5915 
5916 	rx_ring->next_to_clean = 0;
5917 	rx_ring->next_to_use = 0;
5918 
5919 	return 0;
5920 err:
5921 	vfree(rx_ring->rx_buffer_info);
5922 	rx_ring->rx_buffer_info = NULL;
5923 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5924 	return -ENOMEM;
5925 }
5926 
5927 /**
5928  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5929  * @adapter: board private structure
5930  *
5931  * If this function returns with an error, then it's possible one or
5932  * more of the rings is populated (while the rest are not).  It is the
5933  * callers duty to clean those orphaned rings.
5934  *
5935  * Return 0 on success, negative on failure
5936  **/
5937 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5938 {
5939 	int i, err = 0;
5940 
5941 	for (i = 0; i < adapter->num_rx_queues; i++) {
5942 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5943 		if (!err)
5944 			continue;
5945 
5946 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5947 		goto err_setup_rx;
5948 	}
5949 
5950 #ifdef IXGBE_FCOE
5951 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5952 	if (!err)
5953 #endif
5954 		return 0;
5955 err_setup_rx:
5956 	/* rewind the index freeing the rings as we go */
5957 	while (i--)
5958 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5959 	return err;
5960 }
5961 
5962 /**
5963  * ixgbe_free_tx_resources - Free Tx Resources per Queue
5964  * @tx_ring: Tx descriptor ring for a specific queue
5965  *
5966  * Free all transmit software resources
5967  **/
5968 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5969 {
5970 	ixgbe_clean_tx_ring(tx_ring);
5971 
5972 	vfree(tx_ring->tx_buffer_info);
5973 	tx_ring->tx_buffer_info = NULL;
5974 
5975 	/* if not set, then don't free */
5976 	if (!tx_ring->desc)
5977 		return;
5978 
5979 	dma_free_coherent(tx_ring->dev, tx_ring->size,
5980 			  tx_ring->desc, tx_ring->dma);
5981 
5982 	tx_ring->desc = NULL;
5983 }
5984 
5985 /**
5986  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5987  * @adapter: board private structure
5988  *
5989  * Free all transmit software resources
5990  **/
5991 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
5992 {
5993 	int i;
5994 
5995 	for (i = 0; i < adapter->num_tx_queues; i++)
5996 		if (adapter->tx_ring[i]->desc)
5997 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
5998 }
5999 
6000 /**
6001  * ixgbe_free_rx_resources - Free Rx Resources
6002  * @rx_ring: ring to clean the resources from
6003  *
6004  * Free all receive software resources
6005  **/
6006 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6007 {
6008 	ixgbe_clean_rx_ring(rx_ring);
6009 
6010 	vfree(rx_ring->rx_buffer_info);
6011 	rx_ring->rx_buffer_info = NULL;
6012 
6013 	/* if not set, then don't free */
6014 	if (!rx_ring->desc)
6015 		return;
6016 
6017 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6018 			  rx_ring->desc, rx_ring->dma);
6019 
6020 	rx_ring->desc = NULL;
6021 }
6022 
6023 /**
6024  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6025  * @adapter: board private structure
6026  *
6027  * Free all receive software resources
6028  **/
6029 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6030 {
6031 	int i;
6032 
6033 #ifdef IXGBE_FCOE
6034 	ixgbe_free_fcoe_ddp_resources(adapter);
6035 
6036 #endif
6037 	for (i = 0; i < adapter->num_rx_queues; i++)
6038 		if (adapter->rx_ring[i]->desc)
6039 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6040 }
6041 
6042 /**
6043  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6044  * @netdev: network interface device structure
6045  * @new_mtu: new value for maximum frame size
6046  *
6047  * Returns 0 on success, negative on failure
6048  **/
6049 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6050 {
6051 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6052 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
6053 
6054 	/* MTU < 68 is an error and causes problems on some kernels */
6055 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
6056 		return -EINVAL;
6057 
6058 	/*
6059 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6060 	 * paths when MTU greater than 1500 is configured.  So display a
6061 	 * warning that legacy VFs will be disabled.
6062 	 */
6063 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6064 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6065 	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
6066 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6067 
6068 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6069 
6070 	/* must set new MTU before calling down or up */
6071 	netdev->mtu = new_mtu;
6072 
6073 	if (netif_running(netdev))
6074 		ixgbe_reinit_locked(adapter);
6075 
6076 	return 0;
6077 }
6078 
6079 /**
6080  * ixgbe_open - Called when a network interface is made active
6081  * @netdev: network interface device structure
6082  *
6083  * Returns 0 on success, negative value on failure
6084  *
6085  * The open entry point is called when a network interface is made
6086  * active by the system (IFF_UP).  At this point all resources needed
6087  * for transmit and receive operations are allocated, the interrupt
6088  * handler is registered with the OS, the watchdog timer is started,
6089  * and the stack is notified that the interface is ready.
6090  **/
6091 int ixgbe_open(struct net_device *netdev)
6092 {
6093 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6094 	struct ixgbe_hw *hw = &adapter->hw;
6095 	int err, queues;
6096 
6097 	/* disallow open during test */
6098 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6099 		return -EBUSY;
6100 
6101 	netif_carrier_off(netdev);
6102 
6103 	/* allocate transmit descriptors */
6104 	err = ixgbe_setup_all_tx_resources(adapter);
6105 	if (err)
6106 		goto err_setup_tx;
6107 
6108 	/* allocate receive descriptors */
6109 	err = ixgbe_setup_all_rx_resources(adapter);
6110 	if (err)
6111 		goto err_setup_rx;
6112 
6113 	ixgbe_configure(adapter);
6114 
6115 	err = ixgbe_request_irq(adapter);
6116 	if (err)
6117 		goto err_req_irq;
6118 
6119 	/* Notify the stack of the actual queue counts. */
6120 	if (adapter->num_rx_pools > 1)
6121 		queues = adapter->num_rx_queues_per_pool;
6122 	else
6123 		queues = adapter->num_tx_queues;
6124 
6125 	err = netif_set_real_num_tx_queues(netdev, queues);
6126 	if (err)
6127 		goto err_set_queues;
6128 
6129 	if (adapter->num_rx_pools > 1 &&
6130 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
6131 		queues = IXGBE_MAX_L2A_QUEUES;
6132 	else
6133 		queues = adapter->num_rx_queues;
6134 	err = netif_set_real_num_rx_queues(netdev, queues);
6135 	if (err)
6136 		goto err_set_queues;
6137 
6138 	ixgbe_ptp_init(adapter);
6139 
6140 	ixgbe_up_complete(adapter);
6141 
6142 	ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6143 	udp_tunnel_get_rx_info(netdev);
6144 
6145 	return 0;
6146 
6147 err_set_queues:
6148 	ixgbe_free_irq(adapter);
6149 err_req_irq:
6150 	ixgbe_free_all_rx_resources(adapter);
6151 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6152 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6153 err_setup_rx:
6154 	ixgbe_free_all_tx_resources(adapter);
6155 err_setup_tx:
6156 	ixgbe_reset(adapter);
6157 
6158 	return err;
6159 }
6160 
6161 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6162 {
6163 	ixgbe_ptp_suspend(adapter);
6164 
6165 	if (adapter->hw.phy.ops.enter_lplu) {
6166 		adapter->hw.phy.reset_disable = true;
6167 		ixgbe_down(adapter);
6168 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6169 		adapter->hw.phy.reset_disable = false;
6170 	} else {
6171 		ixgbe_down(adapter);
6172 	}
6173 
6174 	ixgbe_free_irq(adapter);
6175 
6176 	ixgbe_free_all_tx_resources(adapter);
6177 	ixgbe_free_all_rx_resources(adapter);
6178 }
6179 
6180 /**
6181  * ixgbe_close - Disables a network interface
6182  * @netdev: network interface device structure
6183  *
6184  * Returns 0, this is not allowed to fail
6185  *
6186  * The close entry point is called when an interface is de-activated
6187  * by the OS.  The hardware is still under the drivers control, but
6188  * needs to be disabled.  A global MAC reset is issued to stop the
6189  * hardware, and all transmit and receive resources are freed.
6190  **/
6191 int ixgbe_close(struct net_device *netdev)
6192 {
6193 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6194 
6195 	ixgbe_ptp_stop(adapter);
6196 
6197 	ixgbe_close_suspend(adapter);
6198 
6199 	ixgbe_fdir_filter_exit(adapter);
6200 
6201 	ixgbe_release_hw_control(adapter);
6202 
6203 	return 0;
6204 }
6205 
6206 #ifdef CONFIG_PM
6207 static int ixgbe_resume(struct pci_dev *pdev)
6208 {
6209 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6210 	struct net_device *netdev = adapter->netdev;
6211 	u32 err;
6212 
6213 	adapter->hw.hw_addr = adapter->io_addr;
6214 	pci_set_power_state(pdev, PCI_D0);
6215 	pci_restore_state(pdev);
6216 	/*
6217 	 * pci_restore_state clears dev->state_saved so call
6218 	 * pci_save_state to restore it.
6219 	 */
6220 	pci_save_state(pdev);
6221 
6222 	err = pci_enable_device_mem(pdev);
6223 	if (err) {
6224 		e_dev_err("Cannot enable PCI device from suspend\n");
6225 		return err;
6226 	}
6227 	smp_mb__before_atomic();
6228 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6229 	pci_set_master(pdev);
6230 
6231 	pci_wake_from_d3(pdev, false);
6232 
6233 	ixgbe_reset(adapter);
6234 
6235 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6236 
6237 	rtnl_lock();
6238 	err = ixgbe_init_interrupt_scheme(adapter);
6239 	if (!err && netif_running(netdev))
6240 		err = ixgbe_open(netdev);
6241 
6242 	rtnl_unlock();
6243 
6244 	if (err)
6245 		return err;
6246 
6247 	netif_device_attach(netdev);
6248 
6249 	return 0;
6250 }
6251 #endif /* CONFIG_PM */
6252 
6253 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6254 {
6255 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6256 	struct net_device *netdev = adapter->netdev;
6257 	struct ixgbe_hw *hw = &adapter->hw;
6258 	u32 ctrl, fctrl;
6259 	u32 wufc = adapter->wol;
6260 #ifdef CONFIG_PM
6261 	int retval = 0;
6262 #endif
6263 
6264 	netif_device_detach(netdev);
6265 
6266 	rtnl_lock();
6267 	if (netif_running(netdev))
6268 		ixgbe_close_suspend(adapter);
6269 	rtnl_unlock();
6270 
6271 	ixgbe_clear_interrupt_scheme(adapter);
6272 
6273 #ifdef CONFIG_PM
6274 	retval = pci_save_state(pdev);
6275 	if (retval)
6276 		return retval;
6277 
6278 #endif
6279 	if (hw->mac.ops.stop_link_on_d3)
6280 		hw->mac.ops.stop_link_on_d3(hw);
6281 
6282 	if (wufc) {
6283 		ixgbe_set_rx_mode(netdev);
6284 
6285 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6286 		if (hw->mac.ops.enable_tx_laser)
6287 			hw->mac.ops.enable_tx_laser(hw);
6288 
6289 		/* turn on all-multi mode if wake on multicast is enabled */
6290 		if (wufc & IXGBE_WUFC_MC) {
6291 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6292 			fctrl |= IXGBE_FCTRL_MPE;
6293 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6294 		}
6295 
6296 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6297 		ctrl |= IXGBE_CTRL_GIO_DIS;
6298 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6299 
6300 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6301 	} else {
6302 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6303 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6304 	}
6305 
6306 	switch (hw->mac.type) {
6307 	case ixgbe_mac_82598EB:
6308 		pci_wake_from_d3(pdev, false);
6309 		break;
6310 	case ixgbe_mac_82599EB:
6311 	case ixgbe_mac_X540:
6312 	case ixgbe_mac_X550:
6313 	case ixgbe_mac_X550EM_x:
6314 	case ixgbe_mac_x550em_a:
6315 		pci_wake_from_d3(pdev, !!wufc);
6316 		break;
6317 	default:
6318 		break;
6319 	}
6320 
6321 	*enable_wake = !!wufc;
6322 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6323 		hw->phy.ops.set_phy_power(hw, false);
6324 
6325 	ixgbe_release_hw_control(adapter);
6326 
6327 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6328 		pci_disable_device(pdev);
6329 
6330 	return 0;
6331 }
6332 
6333 #ifdef CONFIG_PM
6334 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6335 {
6336 	int retval;
6337 	bool wake;
6338 
6339 	retval = __ixgbe_shutdown(pdev, &wake);
6340 	if (retval)
6341 		return retval;
6342 
6343 	if (wake) {
6344 		pci_prepare_to_sleep(pdev);
6345 	} else {
6346 		pci_wake_from_d3(pdev, false);
6347 		pci_set_power_state(pdev, PCI_D3hot);
6348 	}
6349 
6350 	return 0;
6351 }
6352 #endif /* CONFIG_PM */
6353 
6354 static void ixgbe_shutdown(struct pci_dev *pdev)
6355 {
6356 	bool wake;
6357 
6358 	__ixgbe_shutdown(pdev, &wake);
6359 
6360 	if (system_state == SYSTEM_POWER_OFF) {
6361 		pci_wake_from_d3(pdev, wake);
6362 		pci_set_power_state(pdev, PCI_D3hot);
6363 	}
6364 }
6365 
6366 /**
6367  * ixgbe_update_stats - Update the board statistics counters.
6368  * @adapter: board private structure
6369  **/
6370 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6371 {
6372 	struct net_device *netdev = adapter->netdev;
6373 	struct ixgbe_hw *hw = &adapter->hw;
6374 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6375 	u64 total_mpc = 0;
6376 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6377 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6378 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6379 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6380 
6381 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6382 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6383 		return;
6384 
6385 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6386 		u64 rsc_count = 0;
6387 		u64 rsc_flush = 0;
6388 		for (i = 0; i < adapter->num_rx_queues; i++) {
6389 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6390 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6391 		}
6392 		adapter->rsc_total_count = rsc_count;
6393 		adapter->rsc_total_flush = rsc_flush;
6394 	}
6395 
6396 	for (i = 0; i < adapter->num_rx_queues; i++) {
6397 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6398 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6399 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6400 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6401 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6402 		bytes += rx_ring->stats.bytes;
6403 		packets += rx_ring->stats.packets;
6404 	}
6405 	adapter->non_eop_descs = non_eop_descs;
6406 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6407 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6408 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6409 	netdev->stats.rx_bytes = bytes;
6410 	netdev->stats.rx_packets = packets;
6411 
6412 	bytes = 0;
6413 	packets = 0;
6414 	/* gather some stats to the adapter struct that are per queue */
6415 	for (i = 0; i < adapter->num_tx_queues; i++) {
6416 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6417 		restart_queue += tx_ring->tx_stats.restart_queue;
6418 		tx_busy += tx_ring->tx_stats.tx_busy;
6419 		bytes += tx_ring->stats.bytes;
6420 		packets += tx_ring->stats.packets;
6421 	}
6422 	adapter->restart_queue = restart_queue;
6423 	adapter->tx_busy = tx_busy;
6424 	netdev->stats.tx_bytes = bytes;
6425 	netdev->stats.tx_packets = packets;
6426 
6427 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6428 
6429 	/* 8 register reads */
6430 	for (i = 0; i < 8; i++) {
6431 		/* for packet buffers not used, the register should read 0 */
6432 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6433 		missed_rx += mpc;
6434 		hwstats->mpc[i] += mpc;
6435 		total_mpc += hwstats->mpc[i];
6436 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6437 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6438 		switch (hw->mac.type) {
6439 		case ixgbe_mac_82598EB:
6440 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6441 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6442 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6443 			hwstats->pxonrxc[i] +=
6444 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6445 			break;
6446 		case ixgbe_mac_82599EB:
6447 		case ixgbe_mac_X540:
6448 		case ixgbe_mac_X550:
6449 		case ixgbe_mac_X550EM_x:
6450 		case ixgbe_mac_x550em_a:
6451 			hwstats->pxonrxc[i] +=
6452 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6453 			break;
6454 		default:
6455 			break;
6456 		}
6457 	}
6458 
6459 	/*16 register reads */
6460 	for (i = 0; i < 16; i++) {
6461 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6462 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6463 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6464 		    (hw->mac.type == ixgbe_mac_X540) ||
6465 		    (hw->mac.type == ixgbe_mac_X550) ||
6466 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6467 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6468 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6469 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6470 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6471 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6472 		}
6473 	}
6474 
6475 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6476 	/* work around hardware counting issue */
6477 	hwstats->gprc -= missed_rx;
6478 
6479 	ixgbe_update_xoff_received(adapter);
6480 
6481 	/* 82598 hardware only has a 32 bit counter in the high register */
6482 	switch (hw->mac.type) {
6483 	case ixgbe_mac_82598EB:
6484 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6485 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6486 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6487 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6488 		break;
6489 	case ixgbe_mac_X540:
6490 	case ixgbe_mac_X550:
6491 	case ixgbe_mac_X550EM_x:
6492 	case ixgbe_mac_x550em_a:
6493 		/* OS2BMC stats are X540 and later */
6494 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6495 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6496 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6497 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6498 	case ixgbe_mac_82599EB:
6499 		for (i = 0; i < 16; i++)
6500 			adapter->hw_rx_no_dma_resources +=
6501 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6502 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6503 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6504 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6505 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6506 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6507 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6508 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6509 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6510 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6511 #ifdef IXGBE_FCOE
6512 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6513 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6514 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6515 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6516 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6517 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6518 		/* Add up per cpu counters for total ddp aloc fail */
6519 		if (adapter->fcoe.ddp_pool) {
6520 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6521 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6522 			unsigned int cpu;
6523 			u64 noddp = 0, noddp_ext_buff = 0;
6524 			for_each_possible_cpu(cpu) {
6525 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6526 				noddp += ddp_pool->noddp;
6527 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6528 			}
6529 			hwstats->fcoe_noddp = noddp;
6530 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6531 		}
6532 #endif /* IXGBE_FCOE */
6533 		break;
6534 	default:
6535 		break;
6536 	}
6537 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6538 	hwstats->bprc += bprc;
6539 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6540 	if (hw->mac.type == ixgbe_mac_82598EB)
6541 		hwstats->mprc -= bprc;
6542 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6543 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6544 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6545 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6546 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6547 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6548 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6549 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6550 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6551 	hwstats->lxontxc += lxon;
6552 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6553 	hwstats->lxofftxc += lxoff;
6554 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6555 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6556 	/*
6557 	 * 82598 errata - tx of flow control packets is included in tx counters
6558 	 */
6559 	xon_off_tot = lxon + lxoff;
6560 	hwstats->gptc -= xon_off_tot;
6561 	hwstats->mptc -= xon_off_tot;
6562 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6563 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6564 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6565 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6566 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6567 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6568 	hwstats->ptc64 -= xon_off_tot;
6569 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6570 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6571 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6572 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6573 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6574 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6575 
6576 	/* Fill out the OS statistics structure */
6577 	netdev->stats.multicast = hwstats->mprc;
6578 
6579 	/* Rx Errors */
6580 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6581 	netdev->stats.rx_dropped = 0;
6582 	netdev->stats.rx_length_errors = hwstats->rlec;
6583 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6584 	netdev->stats.rx_missed_errors = total_mpc;
6585 }
6586 
6587 /**
6588  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6589  * @adapter: pointer to the device adapter structure
6590  **/
6591 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6592 {
6593 	struct ixgbe_hw *hw = &adapter->hw;
6594 	int i;
6595 
6596 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6597 		return;
6598 
6599 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6600 
6601 	/* if interface is down do nothing */
6602 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6603 		return;
6604 
6605 	/* do nothing if we are not using signature filters */
6606 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6607 		return;
6608 
6609 	adapter->fdir_overflow++;
6610 
6611 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6612 		for (i = 0; i < adapter->num_tx_queues; i++)
6613 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6614 				&(adapter->tx_ring[i]->state));
6615 		/* re-enable flow director interrupts */
6616 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6617 	} else {
6618 		e_err(probe, "failed to finish FDIR re-initialization, "
6619 		      "ignored adding FDIR ATR filters\n");
6620 	}
6621 }
6622 
6623 /**
6624  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6625  * @adapter: pointer to the device adapter structure
6626  *
6627  * This function serves two purposes.  First it strobes the interrupt lines
6628  * in order to make certain interrupts are occurring.  Secondly it sets the
6629  * bits needed to check for TX hangs.  As a result we should immediately
6630  * determine if a hang has occurred.
6631  */
6632 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6633 {
6634 	struct ixgbe_hw *hw = &adapter->hw;
6635 	u64 eics = 0;
6636 	int i;
6637 
6638 	/* If we're down, removing or resetting, just bail */
6639 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6640 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6641 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6642 		return;
6643 
6644 	/* Force detection of hung controller */
6645 	if (netif_carrier_ok(adapter->netdev)) {
6646 		for (i = 0; i < adapter->num_tx_queues; i++)
6647 			set_check_for_tx_hang(adapter->tx_ring[i]);
6648 	}
6649 
6650 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6651 		/*
6652 		 * for legacy and MSI interrupts don't set any bits
6653 		 * that are enabled for EIAM, because this operation
6654 		 * would set *both* EIMS and EICS for any bit in EIAM
6655 		 */
6656 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6657 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6658 	} else {
6659 		/* get one bit for every active tx/rx interrupt vector */
6660 		for (i = 0; i < adapter->num_q_vectors; i++) {
6661 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6662 			if (qv->rx.ring || qv->tx.ring)
6663 				eics |= BIT_ULL(i);
6664 		}
6665 	}
6666 
6667 	/* Cause software interrupt to ensure rings are cleaned */
6668 	ixgbe_irq_rearm_queues(adapter, eics);
6669 }
6670 
6671 /**
6672  * ixgbe_watchdog_update_link - update the link status
6673  * @adapter: pointer to the device adapter structure
6674  * @link_speed: pointer to a u32 to store the link_speed
6675  **/
6676 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6677 {
6678 	struct ixgbe_hw *hw = &adapter->hw;
6679 	u32 link_speed = adapter->link_speed;
6680 	bool link_up = adapter->link_up;
6681 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6682 
6683 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6684 		return;
6685 
6686 	if (hw->mac.ops.check_link) {
6687 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6688 	} else {
6689 		/* always assume link is up, if no check link function */
6690 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6691 		link_up = true;
6692 	}
6693 
6694 	if (adapter->ixgbe_ieee_pfc)
6695 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6696 
6697 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6698 		hw->mac.ops.fc_enable(hw);
6699 		ixgbe_set_rx_drop_en(adapter);
6700 	}
6701 
6702 	if (link_up ||
6703 	    time_after(jiffies, (adapter->link_check_timeout +
6704 				 IXGBE_TRY_LINK_TIMEOUT))) {
6705 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6706 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6707 		IXGBE_WRITE_FLUSH(hw);
6708 	}
6709 
6710 	adapter->link_up = link_up;
6711 	adapter->link_speed = link_speed;
6712 }
6713 
6714 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6715 {
6716 #ifdef CONFIG_IXGBE_DCB
6717 	struct net_device *netdev = adapter->netdev;
6718 	struct dcb_app app = {
6719 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6720 			      .protocol = 0,
6721 			     };
6722 	u8 up = 0;
6723 
6724 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6725 		up = dcb_ieee_getapp_mask(netdev, &app);
6726 
6727 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6728 #endif
6729 }
6730 
6731 /**
6732  * ixgbe_watchdog_link_is_up - update netif_carrier status and
6733  *                             print link up message
6734  * @adapter: pointer to the device adapter structure
6735  **/
6736 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6737 {
6738 	struct net_device *netdev = adapter->netdev;
6739 	struct ixgbe_hw *hw = &adapter->hw;
6740 	struct net_device *upper;
6741 	struct list_head *iter;
6742 	u32 link_speed = adapter->link_speed;
6743 	const char *speed_str;
6744 	bool flow_rx, flow_tx;
6745 
6746 	/* only continue if link was previously down */
6747 	if (netif_carrier_ok(netdev))
6748 		return;
6749 
6750 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6751 
6752 	switch (hw->mac.type) {
6753 	case ixgbe_mac_82598EB: {
6754 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6755 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6756 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6757 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6758 	}
6759 		break;
6760 	case ixgbe_mac_X540:
6761 	case ixgbe_mac_X550:
6762 	case ixgbe_mac_X550EM_x:
6763 	case ixgbe_mac_x550em_a:
6764 	case ixgbe_mac_82599EB: {
6765 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6766 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6767 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6768 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6769 	}
6770 		break;
6771 	default:
6772 		flow_tx = false;
6773 		flow_rx = false;
6774 		break;
6775 	}
6776 
6777 	adapter->last_rx_ptp_check = jiffies;
6778 
6779 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6780 		ixgbe_ptp_start_cyclecounter(adapter);
6781 
6782 	switch (link_speed) {
6783 	case IXGBE_LINK_SPEED_10GB_FULL:
6784 		speed_str = "10 Gbps";
6785 		break;
6786 	case IXGBE_LINK_SPEED_2_5GB_FULL:
6787 		speed_str = "2.5 Gbps";
6788 		break;
6789 	case IXGBE_LINK_SPEED_1GB_FULL:
6790 		speed_str = "1 Gbps";
6791 		break;
6792 	case IXGBE_LINK_SPEED_100_FULL:
6793 		speed_str = "100 Mbps";
6794 		break;
6795 	default:
6796 		speed_str = "unknown speed";
6797 		break;
6798 	}
6799 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
6800 	       ((flow_rx && flow_tx) ? "RX/TX" :
6801 	       (flow_rx ? "RX" :
6802 	       (flow_tx ? "TX" : "None"))));
6803 
6804 	netif_carrier_on(netdev);
6805 	ixgbe_check_vf_rate_limit(adapter);
6806 
6807 	/* enable transmits */
6808 	netif_tx_wake_all_queues(adapter->netdev);
6809 
6810 	/* enable any upper devices */
6811 	rtnl_lock();
6812 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
6813 		if (netif_is_macvlan(upper)) {
6814 			struct macvlan_dev *vlan = netdev_priv(upper);
6815 
6816 			if (vlan->fwd_priv)
6817 				netif_tx_wake_all_queues(upper);
6818 		}
6819 	}
6820 	rtnl_unlock();
6821 
6822 	/* update the default user priority for VFs */
6823 	ixgbe_update_default_up(adapter);
6824 
6825 	/* ping all the active vfs to let them know link has changed */
6826 	ixgbe_ping_all_vfs(adapter);
6827 }
6828 
6829 /**
6830  * ixgbe_watchdog_link_is_down - update netif_carrier status and
6831  *                               print link down message
6832  * @adapter: pointer to the adapter structure
6833  **/
6834 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6835 {
6836 	struct net_device *netdev = adapter->netdev;
6837 	struct ixgbe_hw *hw = &adapter->hw;
6838 
6839 	adapter->link_up = false;
6840 	adapter->link_speed = 0;
6841 
6842 	/* only continue if link was up previously */
6843 	if (!netif_carrier_ok(netdev))
6844 		return;
6845 
6846 	/* poll for SFP+ cable when link is down */
6847 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6848 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6849 
6850 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6851 		ixgbe_ptp_start_cyclecounter(adapter);
6852 
6853 	e_info(drv, "NIC Link is Down\n");
6854 	netif_carrier_off(netdev);
6855 
6856 	/* ping all the active vfs to let them know link has changed */
6857 	ixgbe_ping_all_vfs(adapter);
6858 }
6859 
6860 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
6861 {
6862 	int i;
6863 
6864 	for (i = 0; i < adapter->num_tx_queues; i++) {
6865 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6866 
6867 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
6868 			return true;
6869 	}
6870 
6871 	return false;
6872 }
6873 
6874 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
6875 {
6876 	struct ixgbe_hw *hw = &adapter->hw;
6877 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
6878 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
6879 
6880 	int i, j;
6881 
6882 	if (!adapter->num_vfs)
6883 		return false;
6884 
6885 	/* resetting the PF is only needed for MAC before X550 */
6886 	if (hw->mac.type >= ixgbe_mac_X550)
6887 		return false;
6888 
6889 	for (i = 0; i < adapter->num_vfs; i++) {
6890 		for (j = 0; j < q_per_pool; j++) {
6891 			u32 h, t;
6892 
6893 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
6894 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
6895 
6896 			if (h != t)
6897 				return true;
6898 		}
6899 	}
6900 
6901 	return false;
6902 }
6903 
6904 /**
6905  * ixgbe_watchdog_flush_tx - flush queues on link down
6906  * @adapter: pointer to the device adapter structure
6907  **/
6908 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6909 {
6910 	if (!netif_carrier_ok(adapter->netdev)) {
6911 		if (ixgbe_ring_tx_pending(adapter) ||
6912 		    ixgbe_vf_tx_pending(adapter)) {
6913 			/* We've lost link, so the controller stops DMA,
6914 			 * but we've got queued Tx work that's never going
6915 			 * to get done, so reset controller to flush Tx.
6916 			 * (Do the reset outside of interrupt context).
6917 			 */
6918 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6919 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
6920 		}
6921 	}
6922 }
6923 
6924 #ifdef CONFIG_PCI_IOV
6925 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
6926 				      struct pci_dev *vfdev)
6927 {
6928 	if (!pci_wait_for_pending_transaction(vfdev))
6929 		e_dev_warn("Issuing VFLR with pending transactions\n");
6930 
6931 	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
6932 	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
6933 
6934 	msleep(100);
6935 }
6936 
6937 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6938 {
6939 	struct ixgbe_hw *hw = &adapter->hw;
6940 	struct pci_dev *pdev = adapter->pdev;
6941 	unsigned int vf;
6942 	u32 gpc;
6943 
6944 	if (!(netif_carrier_ok(adapter->netdev)))
6945 		return;
6946 
6947 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6948 	if (gpc) /* If incrementing then no need for the check below */
6949 		return;
6950 	/* Check to see if a bad DMA write target from an errant or
6951 	 * malicious VF has caused a PCIe error.  If so then we can
6952 	 * issue a VFLR to the offending VF(s) and then resume without
6953 	 * requesting a full slot reset.
6954 	 */
6955 
6956 	if (!pdev)
6957 		return;
6958 
6959 	/* check status reg for all VFs owned by this PF */
6960 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
6961 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
6962 		u16 status_reg;
6963 
6964 		if (!vfdev)
6965 			continue;
6966 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
6967 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
6968 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
6969 			ixgbe_issue_vf_flr(adapter, vfdev);
6970 	}
6971 }
6972 
6973 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6974 {
6975 	u32 ssvpc;
6976 
6977 	/* Do not perform spoof check for 82598 or if not in IOV mode */
6978 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
6979 	    adapter->num_vfs == 0)
6980 		return;
6981 
6982 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
6983 
6984 	/*
6985 	 * ssvpc register is cleared on read, if zero then no
6986 	 * spoofed packets in the last interval.
6987 	 */
6988 	if (!ssvpc)
6989 		return;
6990 
6991 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
6992 }
6993 #else
6994 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
6995 {
6996 }
6997 
6998 static void
6999 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7000 {
7001 }
7002 #endif /* CONFIG_PCI_IOV */
7003 
7004 
7005 /**
7006  * ixgbe_watchdog_subtask - check and bring link up
7007  * @adapter: pointer to the device adapter structure
7008  **/
7009 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7010 {
7011 	/* if interface is down, removing or resetting, do nothing */
7012 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7013 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7014 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7015 		return;
7016 
7017 	ixgbe_watchdog_update_link(adapter);
7018 
7019 	if (adapter->link_up)
7020 		ixgbe_watchdog_link_is_up(adapter);
7021 	else
7022 		ixgbe_watchdog_link_is_down(adapter);
7023 
7024 	ixgbe_check_for_bad_vf(adapter);
7025 	ixgbe_spoof_check(adapter);
7026 	ixgbe_update_stats(adapter);
7027 
7028 	ixgbe_watchdog_flush_tx(adapter);
7029 }
7030 
7031 /**
7032  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7033  * @adapter: the ixgbe adapter structure
7034  **/
7035 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7036 {
7037 	struct ixgbe_hw *hw = &adapter->hw;
7038 	s32 err;
7039 
7040 	/* not searching for SFP so there is nothing to do here */
7041 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7042 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7043 		return;
7044 
7045 	if (adapter->sfp_poll_time &&
7046 	    time_after(adapter->sfp_poll_time, jiffies))
7047 		return; /* If not yet time to poll for SFP */
7048 
7049 	/* someone else is in init, wait until next service event */
7050 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7051 		return;
7052 
7053 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7054 
7055 	err = hw->phy.ops.identify_sfp(hw);
7056 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7057 		goto sfp_out;
7058 
7059 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7060 		/* If no cable is present, then we need to reset
7061 		 * the next time we find a good cable. */
7062 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7063 	}
7064 
7065 	/* exit on error */
7066 	if (err)
7067 		goto sfp_out;
7068 
7069 	/* exit if reset not needed */
7070 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7071 		goto sfp_out;
7072 
7073 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7074 
7075 	/*
7076 	 * A module may be identified correctly, but the EEPROM may not have
7077 	 * support for that module.  setup_sfp() will fail in that case, so
7078 	 * we should not allow that module to load.
7079 	 */
7080 	if (hw->mac.type == ixgbe_mac_82598EB)
7081 		err = hw->phy.ops.reset(hw);
7082 	else
7083 		err = hw->mac.ops.setup_sfp(hw);
7084 
7085 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7086 		goto sfp_out;
7087 
7088 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7089 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7090 
7091 sfp_out:
7092 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7093 
7094 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7095 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7096 		e_dev_err("failed to initialize because an unsupported "
7097 			  "SFP+ module type was detected.\n");
7098 		e_dev_err("Reload the driver after installing a "
7099 			  "supported module.\n");
7100 		unregister_netdev(adapter->netdev);
7101 	}
7102 }
7103 
7104 /**
7105  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7106  * @adapter: the ixgbe adapter structure
7107  **/
7108 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7109 {
7110 	struct ixgbe_hw *hw = &adapter->hw;
7111 	u32 speed;
7112 	bool autoneg = false;
7113 
7114 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7115 		return;
7116 
7117 	/* someone else is in init, wait until next service event */
7118 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7119 		return;
7120 
7121 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7122 
7123 	speed = hw->phy.autoneg_advertised;
7124 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
7125 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
7126 
7127 		/* setup the highest link when no autoneg */
7128 		if (!autoneg) {
7129 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
7130 				speed = IXGBE_LINK_SPEED_10GB_FULL;
7131 		}
7132 	}
7133 
7134 	if (hw->mac.ops.setup_link)
7135 		hw->mac.ops.setup_link(hw, speed, true);
7136 
7137 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7138 	adapter->link_check_timeout = jiffies;
7139 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7140 }
7141 
7142 /**
7143  * ixgbe_service_timer - Timer Call-back
7144  * @data: pointer to adapter cast into an unsigned long
7145  **/
7146 static void ixgbe_service_timer(unsigned long data)
7147 {
7148 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
7149 	unsigned long next_event_offset;
7150 
7151 	/* poll faster when waiting for link */
7152 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7153 		next_event_offset = HZ / 10;
7154 	else
7155 		next_event_offset = HZ * 2;
7156 
7157 	/* Reset the timer */
7158 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7159 
7160 	ixgbe_service_event_schedule(adapter);
7161 }
7162 
7163 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7164 {
7165 	struct ixgbe_hw *hw = &adapter->hw;
7166 	u32 status;
7167 
7168 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7169 		return;
7170 
7171 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7172 
7173 	if (!hw->phy.ops.handle_lasi)
7174 		return;
7175 
7176 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7177 	if (status != IXGBE_ERR_OVERTEMP)
7178 		return;
7179 
7180 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7181 }
7182 
7183 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7184 {
7185 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7186 		return;
7187 
7188 	/* If we're already down, removing or resetting, just bail */
7189 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7190 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7191 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7192 		return;
7193 
7194 	ixgbe_dump(adapter);
7195 	netdev_err(adapter->netdev, "Reset adapter\n");
7196 	adapter->tx_timeout_count++;
7197 
7198 	rtnl_lock();
7199 	ixgbe_reinit_locked(adapter);
7200 	rtnl_unlock();
7201 }
7202 
7203 /**
7204  * ixgbe_service_task - manages and runs subtasks
7205  * @work: pointer to work_struct containing our data
7206  **/
7207 static void ixgbe_service_task(struct work_struct *work)
7208 {
7209 	struct ixgbe_adapter *adapter = container_of(work,
7210 						     struct ixgbe_adapter,
7211 						     service_task);
7212 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7213 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7214 			rtnl_lock();
7215 			ixgbe_down(adapter);
7216 			rtnl_unlock();
7217 		}
7218 		ixgbe_service_event_complete(adapter);
7219 		return;
7220 	}
7221 	if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7222 		rtnl_lock();
7223 		adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7224 		udp_tunnel_get_rx_info(adapter->netdev);
7225 		rtnl_unlock();
7226 	}
7227 	ixgbe_reset_subtask(adapter);
7228 	ixgbe_phy_interrupt_subtask(adapter);
7229 	ixgbe_sfp_detection_subtask(adapter);
7230 	ixgbe_sfp_link_config_subtask(adapter);
7231 	ixgbe_check_overtemp_subtask(adapter);
7232 	ixgbe_watchdog_subtask(adapter);
7233 	ixgbe_fdir_reinit_subtask(adapter);
7234 	ixgbe_check_hang_subtask(adapter);
7235 
7236 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7237 		ixgbe_ptp_overflow_check(adapter);
7238 		ixgbe_ptp_rx_hang(adapter);
7239 	}
7240 
7241 	ixgbe_service_event_complete(adapter);
7242 }
7243 
7244 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7245 		     struct ixgbe_tx_buffer *first,
7246 		     u8 *hdr_len)
7247 {
7248 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7249 	struct sk_buff *skb = first->skb;
7250 	union {
7251 		struct iphdr *v4;
7252 		struct ipv6hdr *v6;
7253 		unsigned char *hdr;
7254 	} ip;
7255 	union {
7256 		struct tcphdr *tcp;
7257 		unsigned char *hdr;
7258 	} l4;
7259 	u32 paylen, l4_offset;
7260 	int err;
7261 
7262 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7263 		return 0;
7264 
7265 	if (!skb_is_gso(skb))
7266 		return 0;
7267 
7268 	err = skb_cow_head(skb, 0);
7269 	if (err < 0)
7270 		return err;
7271 
7272 	ip.hdr = skb_network_header(skb);
7273 	l4.hdr = skb_checksum_start(skb);
7274 
7275 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7276 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7277 
7278 	/* initialize outer IP header fields */
7279 	if (ip.v4->version == 4) {
7280 		/* IP header will have to cancel out any data that
7281 		 * is not a part of the outer IP header
7282 		 */
7283 		ip.v4->check = csum_fold(csum_add(lco_csum(skb),
7284 						  csum_unfold(l4.tcp->check)));
7285 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7286 
7287 		ip.v4->tot_len = 0;
7288 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7289 				   IXGBE_TX_FLAGS_CSUM |
7290 				   IXGBE_TX_FLAGS_IPV4;
7291 	} else {
7292 		ip.v6->payload_len = 0;
7293 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7294 				   IXGBE_TX_FLAGS_CSUM;
7295 	}
7296 
7297 	/* determine offset of inner transport header */
7298 	l4_offset = l4.hdr - skb->data;
7299 
7300 	/* compute length of segmentation header */
7301 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7302 
7303 	/* remove payload length from inner checksum */
7304 	paylen = skb->len - l4_offset;
7305 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
7306 
7307 	/* update gso size and bytecount with header size */
7308 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7309 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7310 
7311 	/* mss_l4len_id: use 0 as index for TSO */
7312 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7313 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7314 
7315 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7316 	vlan_macip_lens = l4.hdr - ip.hdr;
7317 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7318 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7319 
7320 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
7321 			  mss_l4len_idx);
7322 
7323 	return 1;
7324 }
7325 
7326 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7327 {
7328 	unsigned int offset = 0;
7329 
7330 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7331 
7332 	return offset == skb_checksum_start_offset(skb);
7333 }
7334 
7335 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7336 			  struct ixgbe_tx_buffer *first)
7337 {
7338 	struct sk_buff *skb = first->skb;
7339 	u32 vlan_macip_lens = 0;
7340 	u32 type_tucmd = 0;
7341 
7342 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7343 csum_failed:
7344 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7345 					 IXGBE_TX_FLAGS_CC)))
7346 			return;
7347 		goto no_csum;
7348 	}
7349 
7350 	switch (skb->csum_offset) {
7351 	case offsetof(struct tcphdr, check):
7352 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7353 		/* fall through */
7354 	case offsetof(struct udphdr, check):
7355 		break;
7356 	case offsetof(struct sctphdr, checksum):
7357 		/* validate that this is actually an SCTP request */
7358 		if (((first->protocol == htons(ETH_P_IP)) &&
7359 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7360 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7361 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7362 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7363 			break;
7364 		}
7365 		/* fall through */
7366 	default:
7367 		skb_checksum_help(skb);
7368 		goto csum_failed;
7369 	}
7370 
7371 	/* update TX checksum flag */
7372 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7373 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7374 			  skb_network_offset(skb);
7375 no_csum:
7376 	/* vlan_macip_lens: MACLEN, VLAN tag */
7377 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7378 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7379 
7380 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd, 0);
7381 }
7382 
7383 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7384 	((_flag <= _result) ? \
7385 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7386 	 ((u32)(_input & _flag) / (_flag / _result)))
7387 
7388 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7389 {
7390 	/* set type for advanced descriptor with frame checksum insertion */
7391 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7392 		       IXGBE_ADVTXD_DCMD_DEXT |
7393 		       IXGBE_ADVTXD_DCMD_IFCS;
7394 
7395 	/* set HW vlan bit if vlan is present */
7396 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7397 				   IXGBE_ADVTXD_DCMD_VLE);
7398 
7399 	/* set segmentation enable bits for TSO/FSO */
7400 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7401 				   IXGBE_ADVTXD_DCMD_TSE);
7402 
7403 	/* set timestamp bit if present */
7404 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7405 				   IXGBE_ADVTXD_MAC_TSTAMP);
7406 
7407 	/* insert frame checksum */
7408 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7409 
7410 	return cmd_type;
7411 }
7412 
7413 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7414 				   u32 tx_flags, unsigned int paylen)
7415 {
7416 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7417 
7418 	/* enable L4 checksum for TSO and TX checksum offload */
7419 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7420 					IXGBE_TX_FLAGS_CSUM,
7421 					IXGBE_ADVTXD_POPTS_TXSM);
7422 
7423 	/* enble IPv4 checksum for TSO */
7424 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7425 					IXGBE_TX_FLAGS_IPV4,
7426 					IXGBE_ADVTXD_POPTS_IXSM);
7427 
7428 	/*
7429 	 * Check Context must be set if Tx switch is enabled, which it
7430 	 * always is for case where virtual functions are running
7431 	 */
7432 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7433 					IXGBE_TX_FLAGS_CC,
7434 					IXGBE_ADVTXD_CC);
7435 
7436 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7437 }
7438 
7439 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7440 {
7441 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7442 
7443 	/* Herbert's original patch had:
7444 	 *  smp_mb__after_netif_stop_queue();
7445 	 * but since that doesn't exist yet, just open code it.
7446 	 */
7447 	smp_mb();
7448 
7449 	/* We need to check again in a case another CPU has just
7450 	 * made room available.
7451 	 */
7452 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7453 		return -EBUSY;
7454 
7455 	/* A reprieve! - use start_queue because it doesn't call schedule */
7456 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7457 	++tx_ring->tx_stats.restart_queue;
7458 	return 0;
7459 }
7460 
7461 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7462 {
7463 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7464 		return 0;
7465 
7466 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7467 }
7468 
7469 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7470 		       IXGBE_TXD_CMD_RS)
7471 
7472 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7473 			 struct ixgbe_tx_buffer *first,
7474 			 const u8 hdr_len)
7475 {
7476 	struct sk_buff *skb = first->skb;
7477 	struct ixgbe_tx_buffer *tx_buffer;
7478 	union ixgbe_adv_tx_desc *tx_desc;
7479 	struct skb_frag_struct *frag;
7480 	dma_addr_t dma;
7481 	unsigned int data_len, size;
7482 	u32 tx_flags = first->tx_flags;
7483 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7484 	u16 i = tx_ring->next_to_use;
7485 
7486 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7487 
7488 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7489 
7490 	size = skb_headlen(skb);
7491 	data_len = skb->data_len;
7492 
7493 #ifdef IXGBE_FCOE
7494 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7495 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7496 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7497 			data_len = 0;
7498 		} else {
7499 			data_len -= sizeof(struct fcoe_crc_eof);
7500 		}
7501 	}
7502 
7503 #endif
7504 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7505 
7506 	tx_buffer = first;
7507 
7508 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7509 		if (dma_mapping_error(tx_ring->dev, dma))
7510 			goto dma_error;
7511 
7512 		/* record length, and DMA address */
7513 		dma_unmap_len_set(tx_buffer, len, size);
7514 		dma_unmap_addr_set(tx_buffer, dma, dma);
7515 
7516 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7517 
7518 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7519 			tx_desc->read.cmd_type_len =
7520 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7521 
7522 			i++;
7523 			tx_desc++;
7524 			if (i == tx_ring->count) {
7525 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7526 				i = 0;
7527 			}
7528 			tx_desc->read.olinfo_status = 0;
7529 
7530 			dma += IXGBE_MAX_DATA_PER_TXD;
7531 			size -= IXGBE_MAX_DATA_PER_TXD;
7532 
7533 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7534 		}
7535 
7536 		if (likely(!data_len))
7537 			break;
7538 
7539 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7540 
7541 		i++;
7542 		tx_desc++;
7543 		if (i == tx_ring->count) {
7544 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7545 			i = 0;
7546 		}
7547 		tx_desc->read.olinfo_status = 0;
7548 
7549 #ifdef IXGBE_FCOE
7550 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7551 #else
7552 		size = skb_frag_size(frag);
7553 #endif
7554 		data_len -= size;
7555 
7556 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7557 				       DMA_TO_DEVICE);
7558 
7559 		tx_buffer = &tx_ring->tx_buffer_info[i];
7560 	}
7561 
7562 	/* write last descriptor with RS and EOP bits */
7563 	cmd_type |= size | IXGBE_TXD_CMD;
7564 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7565 
7566 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7567 
7568 	/* set the timestamp */
7569 	first->time_stamp = jiffies;
7570 
7571 	/*
7572 	 * Force memory writes to complete before letting h/w know there
7573 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7574 	 * memory model archs, such as IA-64).
7575 	 *
7576 	 * We also need this memory barrier to make certain all of the
7577 	 * status bits have been updated before next_to_watch is written.
7578 	 */
7579 	wmb();
7580 
7581 	/* set next_to_watch value indicating a packet is present */
7582 	first->next_to_watch = tx_desc;
7583 
7584 	i++;
7585 	if (i == tx_ring->count)
7586 		i = 0;
7587 
7588 	tx_ring->next_to_use = i;
7589 
7590 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7591 
7592 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7593 		writel(i, tx_ring->tail);
7594 
7595 		/* we need this if more than one processor can write to our tail
7596 		 * at a time, it synchronizes IO on IA64/Altix systems
7597 		 */
7598 		mmiowb();
7599 	}
7600 
7601 	return;
7602 dma_error:
7603 	dev_err(tx_ring->dev, "TX DMA map failed\n");
7604 
7605 	/* clear dma mappings for failed tx_buffer_info map */
7606 	for (;;) {
7607 		tx_buffer = &tx_ring->tx_buffer_info[i];
7608 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
7609 		if (tx_buffer == first)
7610 			break;
7611 		if (i == 0)
7612 			i = tx_ring->count;
7613 		i--;
7614 	}
7615 
7616 	tx_ring->next_to_use = i;
7617 }
7618 
7619 static void ixgbe_atr(struct ixgbe_ring *ring,
7620 		      struct ixgbe_tx_buffer *first)
7621 {
7622 	struct ixgbe_q_vector *q_vector = ring->q_vector;
7623 	union ixgbe_atr_hash_dword input = { .dword = 0 };
7624 	union ixgbe_atr_hash_dword common = { .dword = 0 };
7625 	union {
7626 		unsigned char *network;
7627 		struct iphdr *ipv4;
7628 		struct ipv6hdr *ipv6;
7629 	} hdr;
7630 	struct tcphdr *th;
7631 	unsigned int hlen;
7632 	struct sk_buff *skb;
7633 	__be16 vlan_id;
7634 	int l4_proto;
7635 
7636 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7637 	if (!q_vector)
7638 		return;
7639 
7640 	/* do nothing if sampling is disabled */
7641 	if (!ring->atr_sample_rate)
7642 		return;
7643 
7644 	ring->atr_count++;
7645 
7646 	/* currently only IPv4/IPv6 with TCP is supported */
7647 	if ((first->protocol != htons(ETH_P_IP)) &&
7648 	    (first->protocol != htons(ETH_P_IPV6)))
7649 		return;
7650 
7651 	/* snag network header to get L4 type and address */
7652 	skb = first->skb;
7653 	hdr.network = skb_network_header(skb);
7654 	if (skb->encapsulation &&
7655 	    first->protocol == htons(ETH_P_IP) &&
7656 	    hdr.ipv4->protocol != IPPROTO_UDP) {
7657 		struct ixgbe_adapter *adapter = q_vector->adapter;
7658 
7659 		/* verify the port is recognized as VXLAN */
7660 		if (adapter->vxlan_port &&
7661 		    udp_hdr(skb)->dest == adapter->vxlan_port)
7662 			hdr.network = skb_inner_network_header(skb);
7663 
7664 		if (adapter->geneve_port &&
7665 		    udp_hdr(skb)->dest == adapter->geneve_port)
7666 			hdr.network = skb_inner_network_header(skb);
7667 	}
7668 
7669 	/* Currently only IPv4/IPv6 with TCP is supported */
7670 	switch (hdr.ipv4->version) {
7671 	case IPVERSION:
7672 		/* access ihl as u8 to avoid unaligned access on ia64 */
7673 		hlen = (hdr.network[0] & 0x0F) << 2;
7674 		l4_proto = hdr.ipv4->protocol;
7675 		break;
7676 	case 6:
7677 		hlen = hdr.network - skb->data;
7678 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
7679 		hlen -= hdr.network - skb->data;
7680 		break;
7681 	default:
7682 		return;
7683 	}
7684 
7685 	if (l4_proto != IPPROTO_TCP)
7686 		return;
7687 
7688 	th = (struct tcphdr *)(hdr.network + hlen);
7689 
7690 	/* skip this packet since the socket is closing */
7691 	if (th->fin)
7692 		return;
7693 
7694 	/* sample on all syn packets or once every atr sample count */
7695 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7696 		return;
7697 
7698 	/* reset sample count */
7699 	ring->atr_count = 0;
7700 
7701 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7702 
7703 	/*
7704 	 * src and dst are inverted, think how the receiver sees them
7705 	 *
7706 	 * The input is broken into two sections, a non-compressed section
7707 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7708 	 * is XORed together and stored in the compressed dword.
7709 	 */
7710 	input.formatted.vlan_id = vlan_id;
7711 
7712 	/*
7713 	 * since src port and flex bytes occupy the same word XOR them together
7714 	 * and write the value to source port portion of compressed dword
7715 	 */
7716 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7717 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7718 	else
7719 		common.port.src ^= th->dest ^ first->protocol;
7720 	common.port.dst ^= th->source;
7721 
7722 	switch (hdr.ipv4->version) {
7723 	case IPVERSION:
7724 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7725 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7726 		break;
7727 	case 6:
7728 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7729 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7730 			     hdr.ipv6->saddr.s6_addr32[1] ^
7731 			     hdr.ipv6->saddr.s6_addr32[2] ^
7732 			     hdr.ipv6->saddr.s6_addr32[3] ^
7733 			     hdr.ipv6->daddr.s6_addr32[0] ^
7734 			     hdr.ipv6->daddr.s6_addr32[1] ^
7735 			     hdr.ipv6->daddr.s6_addr32[2] ^
7736 			     hdr.ipv6->daddr.s6_addr32[3];
7737 		break;
7738 	default:
7739 		break;
7740 	}
7741 
7742 	if (hdr.network != skb_network_header(skb))
7743 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
7744 
7745 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7746 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7747 					      input, common, ring->queue_index);
7748 }
7749 
7750 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7751 			      void *accel_priv, select_queue_fallback_t fallback)
7752 {
7753 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7754 #ifdef IXGBE_FCOE
7755 	struct ixgbe_adapter *adapter;
7756 	struct ixgbe_ring_feature *f;
7757 	int txq;
7758 #endif
7759 
7760 	if (fwd_adapter)
7761 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7762 
7763 #ifdef IXGBE_FCOE
7764 
7765 	/*
7766 	 * only execute the code below if protocol is FCoE
7767 	 * or FIP and we have FCoE enabled on the adapter
7768 	 */
7769 	switch (vlan_get_protocol(skb)) {
7770 	case htons(ETH_P_FCOE):
7771 	case htons(ETH_P_FIP):
7772 		adapter = netdev_priv(dev);
7773 
7774 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7775 			break;
7776 	default:
7777 		return fallback(dev, skb);
7778 	}
7779 
7780 	f = &adapter->ring_feature[RING_F_FCOE];
7781 
7782 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7783 					   smp_processor_id();
7784 
7785 	while (txq >= f->indices)
7786 		txq -= f->indices;
7787 
7788 	return txq + f->offset;
7789 #else
7790 	return fallback(dev, skb);
7791 #endif
7792 }
7793 
7794 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7795 			  struct ixgbe_adapter *adapter,
7796 			  struct ixgbe_ring *tx_ring)
7797 {
7798 	struct ixgbe_tx_buffer *first;
7799 	int tso;
7800 	u32 tx_flags = 0;
7801 	unsigned short f;
7802 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7803 	__be16 protocol = skb->protocol;
7804 	u8 hdr_len = 0;
7805 
7806 	/*
7807 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7808 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7809 	 *       + 2 desc gap to keep tail from touching head,
7810 	 *       + 1 desc for context descriptor,
7811 	 * otherwise try next time
7812 	 */
7813 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7814 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7815 
7816 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7817 		tx_ring->tx_stats.tx_busy++;
7818 		return NETDEV_TX_BUSY;
7819 	}
7820 
7821 	/* record the location of the first descriptor for this packet */
7822 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7823 	first->skb = skb;
7824 	first->bytecount = skb->len;
7825 	first->gso_segs = 1;
7826 
7827 	/* if we have a HW VLAN tag being added default to the HW one */
7828 	if (skb_vlan_tag_present(skb)) {
7829 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7830 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7831 	/* else if it is a SW VLAN check the next protocol and store the tag */
7832 	} else if (protocol == htons(ETH_P_8021Q)) {
7833 		struct vlan_hdr *vhdr, _vhdr;
7834 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7835 		if (!vhdr)
7836 			goto out_drop;
7837 
7838 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7839 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7840 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7841 	}
7842 	protocol = vlan_get_protocol(skb);
7843 
7844 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
7845 	    adapter->ptp_clock &&
7846 	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7847 				   &adapter->state)) {
7848 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7849 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7850 
7851 		/* schedule check for Tx timestamp */
7852 		adapter->ptp_tx_skb = skb_get(skb);
7853 		adapter->ptp_tx_start = jiffies;
7854 		schedule_work(&adapter->ptp_tx_work);
7855 	}
7856 
7857 	skb_tx_timestamp(skb);
7858 
7859 #ifdef CONFIG_PCI_IOV
7860 	/*
7861 	 * Use the l2switch_enable flag - would be false if the DMA
7862 	 * Tx switch had been disabled.
7863 	 */
7864 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7865 		tx_flags |= IXGBE_TX_FLAGS_CC;
7866 
7867 #endif
7868 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7869 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7870 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7871 	     (skb->priority != TC_PRIO_CONTROL))) {
7872 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7873 		tx_flags |= (skb->priority & 0x7) <<
7874 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7875 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7876 			struct vlan_ethhdr *vhdr;
7877 
7878 			if (skb_cow_head(skb, 0))
7879 				goto out_drop;
7880 			vhdr = (struct vlan_ethhdr *)skb->data;
7881 			vhdr->h_vlan_TCI = htons(tx_flags >>
7882 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7883 		} else {
7884 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7885 		}
7886 	}
7887 
7888 	/* record initial flags and protocol */
7889 	first->tx_flags = tx_flags;
7890 	first->protocol = protocol;
7891 
7892 #ifdef IXGBE_FCOE
7893 	/* setup tx offload for FCoE */
7894 	if ((protocol == htons(ETH_P_FCOE)) &&
7895 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7896 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7897 		if (tso < 0)
7898 			goto out_drop;
7899 
7900 		goto xmit_fcoe;
7901 	}
7902 
7903 #endif /* IXGBE_FCOE */
7904 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7905 	if (tso < 0)
7906 		goto out_drop;
7907 	else if (!tso)
7908 		ixgbe_tx_csum(tx_ring, first);
7909 
7910 	/* add the ATR filter if ATR is on */
7911 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7912 		ixgbe_atr(tx_ring, first);
7913 
7914 #ifdef IXGBE_FCOE
7915 xmit_fcoe:
7916 #endif /* IXGBE_FCOE */
7917 	ixgbe_tx_map(tx_ring, first, hdr_len);
7918 
7919 	return NETDEV_TX_OK;
7920 
7921 out_drop:
7922 	dev_kfree_skb_any(first->skb);
7923 	first->skb = NULL;
7924 
7925 	return NETDEV_TX_OK;
7926 }
7927 
7928 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7929 				      struct net_device *netdev,
7930 				      struct ixgbe_ring *ring)
7931 {
7932 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7933 	struct ixgbe_ring *tx_ring;
7934 
7935 	/*
7936 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7937 	 * in order to meet this minimum size requirement.
7938 	 */
7939 	if (skb_put_padto(skb, 17))
7940 		return NETDEV_TX_OK;
7941 
7942 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7943 
7944 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7945 }
7946 
7947 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7948 				    struct net_device *netdev)
7949 {
7950 	return __ixgbe_xmit_frame(skb, netdev, NULL);
7951 }
7952 
7953 /**
7954  * ixgbe_set_mac - Change the Ethernet Address of the NIC
7955  * @netdev: network interface device structure
7956  * @p: pointer to an address structure
7957  *
7958  * Returns 0 on success, negative on failure
7959  **/
7960 static int ixgbe_set_mac(struct net_device *netdev, void *p)
7961 {
7962 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7963 	struct ixgbe_hw *hw = &adapter->hw;
7964 	struct sockaddr *addr = p;
7965 
7966 	if (!is_valid_ether_addr(addr->sa_data))
7967 		return -EADDRNOTAVAIL;
7968 
7969 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
7970 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
7971 
7972 	ixgbe_mac_set_default_filter(adapter);
7973 
7974 	return 0;
7975 }
7976 
7977 static int
7978 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
7979 {
7980 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7981 	struct ixgbe_hw *hw = &adapter->hw;
7982 	u16 value;
7983 	int rc;
7984 
7985 	if (prtad != hw->phy.mdio.prtad)
7986 		return -EINVAL;
7987 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
7988 	if (!rc)
7989 		rc = value;
7990 	return rc;
7991 }
7992 
7993 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
7994 			    u16 addr, u16 value)
7995 {
7996 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7997 	struct ixgbe_hw *hw = &adapter->hw;
7998 
7999 	if (prtad != hw->phy.mdio.prtad)
8000 		return -EINVAL;
8001 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8002 }
8003 
8004 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8005 {
8006 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8007 
8008 	switch (cmd) {
8009 	case SIOCSHWTSTAMP:
8010 		return ixgbe_ptp_set_ts_config(adapter, req);
8011 	case SIOCGHWTSTAMP:
8012 		return ixgbe_ptp_get_ts_config(adapter, req);
8013 	default:
8014 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8015 	}
8016 }
8017 
8018 /**
8019  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8020  * netdev->dev_addrs
8021  * @netdev: network interface device structure
8022  *
8023  * Returns non-zero on failure
8024  **/
8025 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8026 {
8027 	int err = 0;
8028 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8029 	struct ixgbe_hw *hw = &adapter->hw;
8030 
8031 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8032 		rtnl_lock();
8033 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8034 		rtnl_unlock();
8035 
8036 		/* update SAN MAC vmdq pool selection */
8037 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8038 	}
8039 	return err;
8040 }
8041 
8042 /**
8043  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8044  * netdev->dev_addrs
8045  * @netdev: network interface device structure
8046  *
8047  * Returns non-zero on failure
8048  **/
8049 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8050 {
8051 	int err = 0;
8052 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8053 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8054 
8055 	if (is_valid_ether_addr(mac->san_addr)) {
8056 		rtnl_lock();
8057 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8058 		rtnl_unlock();
8059 	}
8060 	return err;
8061 }
8062 
8063 #ifdef CONFIG_NET_POLL_CONTROLLER
8064 /*
8065  * Polling 'interrupt' - used by things like netconsole to send skbs
8066  * without having to re-enable interrupts. It's not called while
8067  * the interrupt routine is executing.
8068  */
8069 static void ixgbe_netpoll(struct net_device *netdev)
8070 {
8071 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8072 	int i;
8073 
8074 	/* if interface is down do nothing */
8075 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8076 		return;
8077 
8078 	/* loop through and schedule all active queues */
8079 	for (i = 0; i < adapter->num_q_vectors; i++)
8080 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8081 }
8082 
8083 #endif
8084 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
8085 						   struct rtnl_link_stats64 *stats)
8086 {
8087 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8088 	int i;
8089 
8090 	rcu_read_lock();
8091 	for (i = 0; i < adapter->num_rx_queues; i++) {
8092 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
8093 		u64 bytes, packets;
8094 		unsigned int start;
8095 
8096 		if (ring) {
8097 			do {
8098 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8099 				packets = ring->stats.packets;
8100 				bytes   = ring->stats.bytes;
8101 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8102 			stats->rx_packets += packets;
8103 			stats->rx_bytes   += bytes;
8104 		}
8105 	}
8106 
8107 	for (i = 0; i < adapter->num_tx_queues; i++) {
8108 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
8109 		u64 bytes, packets;
8110 		unsigned int start;
8111 
8112 		if (ring) {
8113 			do {
8114 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8115 				packets = ring->stats.packets;
8116 				bytes   = ring->stats.bytes;
8117 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8118 			stats->tx_packets += packets;
8119 			stats->tx_bytes   += bytes;
8120 		}
8121 	}
8122 	rcu_read_unlock();
8123 	/* following stats updated by ixgbe_watchdog_task() */
8124 	stats->multicast	= netdev->stats.multicast;
8125 	stats->rx_errors	= netdev->stats.rx_errors;
8126 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8127 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8128 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8129 	return stats;
8130 }
8131 
8132 #ifdef CONFIG_IXGBE_DCB
8133 /**
8134  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8135  * @adapter: pointer to ixgbe_adapter
8136  * @tc: number of traffic classes currently enabled
8137  *
8138  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8139  * 802.1Q priority maps to a packet buffer that exists.
8140  */
8141 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8142 {
8143 	struct ixgbe_hw *hw = &adapter->hw;
8144 	u32 reg, rsave;
8145 	int i;
8146 
8147 	/* 82598 have a static priority to TC mapping that can not
8148 	 * be changed so no validation is needed.
8149 	 */
8150 	if (hw->mac.type == ixgbe_mac_82598EB)
8151 		return;
8152 
8153 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8154 	rsave = reg;
8155 
8156 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8157 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8158 
8159 		/* If up2tc is out of bounds default to zero */
8160 		if (up2tc > tc)
8161 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8162 	}
8163 
8164 	if (reg != rsave)
8165 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8166 
8167 	return;
8168 }
8169 
8170 /**
8171  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8172  * @adapter: Pointer to adapter struct
8173  *
8174  * Populate the netdev user priority to tc map
8175  */
8176 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8177 {
8178 	struct net_device *dev = adapter->netdev;
8179 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8180 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8181 	u8 prio;
8182 
8183 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8184 		u8 tc = 0;
8185 
8186 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8187 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8188 		else if (ets)
8189 			tc = ets->prio_tc[prio];
8190 
8191 		netdev_set_prio_tc_map(dev, prio, tc);
8192 	}
8193 }
8194 
8195 #endif /* CONFIG_IXGBE_DCB */
8196 /**
8197  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8198  *
8199  * @netdev: net device to configure
8200  * @tc: number of traffic classes to enable
8201  */
8202 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8203 {
8204 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8205 	struct ixgbe_hw *hw = &adapter->hw;
8206 	bool pools;
8207 
8208 	/* Hardware supports up to 8 traffic classes */
8209 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8210 		return -EINVAL;
8211 
8212 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8213 		return -EINVAL;
8214 
8215 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
8216 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
8217 		return -EBUSY;
8218 
8219 	/* Hardware has to reinitialize queues and interrupts to
8220 	 * match packet buffer alignment. Unfortunately, the
8221 	 * hardware is not flexible enough to do this dynamically.
8222 	 */
8223 	if (netif_running(dev))
8224 		ixgbe_close(dev);
8225 	else
8226 		ixgbe_reset(adapter);
8227 
8228 	ixgbe_clear_interrupt_scheme(adapter);
8229 
8230 #ifdef CONFIG_IXGBE_DCB
8231 	if (tc) {
8232 		netdev_set_num_tc(dev, tc);
8233 		ixgbe_set_prio_tc_map(adapter);
8234 
8235 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8236 
8237 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8238 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8239 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8240 		}
8241 	} else {
8242 		netdev_reset_tc(dev);
8243 
8244 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8245 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8246 
8247 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8248 
8249 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8250 		adapter->dcb_cfg.pfc_mode_enable = false;
8251 	}
8252 
8253 	ixgbe_validate_rtr(adapter, tc);
8254 
8255 #endif /* CONFIG_IXGBE_DCB */
8256 	ixgbe_init_interrupt_scheme(adapter);
8257 
8258 	if (netif_running(dev))
8259 		return ixgbe_open(dev);
8260 
8261 	return 0;
8262 }
8263 
8264 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8265 			       struct tc_cls_u32_offload *cls)
8266 {
8267 	u32 hdl = cls->knode.handle;
8268 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8269 	u32 loc = cls->knode.handle & 0xfffff;
8270 	int err = 0, i, j;
8271 	struct ixgbe_jump_table *jump = NULL;
8272 
8273 	if (loc > IXGBE_MAX_HW_ENTRIES)
8274 		return -EINVAL;
8275 
8276 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8277 		return -EINVAL;
8278 
8279 	/* Clear this filter in the link data it is associated with */
8280 	if (uhtid != 0x800) {
8281 		jump = adapter->jump_tables[uhtid];
8282 		if (!jump)
8283 			return -EINVAL;
8284 		if (!test_bit(loc - 1, jump->child_loc_map))
8285 			return -EINVAL;
8286 		clear_bit(loc - 1, jump->child_loc_map);
8287 	}
8288 
8289 	/* Check if the filter being deleted is a link */
8290 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8291 		jump = adapter->jump_tables[i];
8292 		if (jump && jump->link_hdl == hdl) {
8293 			/* Delete filters in the hardware in the child hash
8294 			 * table associated with this link
8295 			 */
8296 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8297 				if (!test_bit(j, jump->child_loc_map))
8298 					continue;
8299 				spin_lock(&adapter->fdir_perfect_lock);
8300 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8301 								      NULL,
8302 								      j + 1);
8303 				spin_unlock(&adapter->fdir_perfect_lock);
8304 				clear_bit(j, jump->child_loc_map);
8305 			}
8306 			/* Remove resources for this link */
8307 			kfree(jump->input);
8308 			kfree(jump->mask);
8309 			kfree(jump);
8310 			adapter->jump_tables[i] = NULL;
8311 			return err;
8312 		}
8313 	}
8314 
8315 	spin_lock(&adapter->fdir_perfect_lock);
8316 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8317 	spin_unlock(&adapter->fdir_perfect_lock);
8318 	return err;
8319 }
8320 
8321 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8322 					    __be16 protocol,
8323 					    struct tc_cls_u32_offload *cls)
8324 {
8325 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8326 
8327 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8328 		return -EINVAL;
8329 
8330 	/* This ixgbe devices do not support hash tables at the moment
8331 	 * so abort when given hash tables.
8332 	 */
8333 	if (cls->hnode.divisor > 0)
8334 		return -EINVAL;
8335 
8336 	set_bit(uhtid - 1, &adapter->tables);
8337 	return 0;
8338 }
8339 
8340 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
8341 					    struct tc_cls_u32_offload *cls)
8342 {
8343 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8344 
8345 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8346 		return -EINVAL;
8347 
8348 	clear_bit(uhtid - 1, &adapter->tables);
8349 	return 0;
8350 }
8351 
8352 #ifdef CONFIG_NET_CLS_ACT
8353 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
8354 				  u8 *queue, u64 *action)
8355 {
8356 	unsigned int num_vfs = adapter->num_vfs, vf;
8357 	struct net_device *upper;
8358 	struct list_head *iter;
8359 
8360 	/* redirect to a SRIOV VF */
8361 	for (vf = 0; vf < num_vfs; ++vf) {
8362 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
8363 		if (upper->ifindex == ifindex) {
8364 			if (adapter->num_rx_pools > 1)
8365 				*queue = vf * 2;
8366 			else
8367 				*queue = vf * adapter->num_rx_queues_per_pool;
8368 
8369 			*action = vf + 1;
8370 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
8371 			return 0;
8372 		}
8373 	}
8374 
8375 	/* redirect to a offloaded macvlan netdev */
8376 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
8377 		if (netif_is_macvlan(upper)) {
8378 			struct macvlan_dev *dfwd = netdev_priv(upper);
8379 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
8380 
8381 			if (vadapter && vadapter->netdev->ifindex == ifindex) {
8382 				*queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
8383 				*action = *queue;
8384 				return 0;
8385 			}
8386 		}
8387 	}
8388 
8389 	return -EINVAL;
8390 }
8391 
8392 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8393 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8394 {
8395 	const struct tc_action *a;
8396 	LIST_HEAD(actions);
8397 	int err;
8398 
8399 	if (tc_no_actions(exts))
8400 		return -EINVAL;
8401 
8402 	tcf_exts_to_list(exts, &actions);
8403 	list_for_each_entry(a, &actions, list) {
8404 
8405 		/* Drop action */
8406 		if (is_tcf_gact_shot(a)) {
8407 			*action = IXGBE_FDIR_DROP_QUEUE;
8408 			*queue = IXGBE_FDIR_DROP_QUEUE;
8409 			return 0;
8410 		}
8411 
8412 		/* Redirect to a VF or a offloaded macvlan */
8413 		if (is_tcf_mirred_redirect(a)) {
8414 			int ifindex = tcf_mirred_ifindex(a);
8415 
8416 			err = handle_redirect_action(adapter, ifindex, queue,
8417 						     action);
8418 			if (err == 0)
8419 				return err;
8420 		}
8421 	}
8422 
8423 	return -EINVAL;
8424 }
8425 #else
8426 static int parse_tc_actions(struct ixgbe_adapter *adapter,
8427 			    struct tcf_exts *exts, u64 *action, u8 *queue)
8428 {
8429 	return -EINVAL;
8430 }
8431 #endif /* CONFIG_NET_CLS_ACT */
8432 
8433 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
8434 				    union ixgbe_atr_input *mask,
8435 				    struct tc_cls_u32_offload *cls,
8436 				    struct ixgbe_mat_field *field_ptr,
8437 				    struct ixgbe_nexthdr *nexthdr)
8438 {
8439 	int i, j, off;
8440 	__be32 val, m;
8441 	bool found_entry = false, found_jump_field = false;
8442 
8443 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
8444 		off = cls->knode.sel->keys[i].off;
8445 		val = cls->knode.sel->keys[i].val;
8446 		m = cls->knode.sel->keys[i].mask;
8447 
8448 		for (j = 0; field_ptr[j].val; j++) {
8449 			if (field_ptr[j].off == off) {
8450 				field_ptr[j].val(input, mask, val, m);
8451 				input->filter.formatted.flow_type |=
8452 					field_ptr[j].type;
8453 				found_entry = true;
8454 				break;
8455 			}
8456 		}
8457 		if (nexthdr) {
8458 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
8459 			    nexthdr->val == cls->knode.sel->keys[i].val &&
8460 			    nexthdr->mask == cls->knode.sel->keys[i].mask)
8461 				found_jump_field = true;
8462 			else
8463 				continue;
8464 		}
8465 	}
8466 
8467 	if (nexthdr && !found_jump_field)
8468 		return -EINVAL;
8469 
8470 	if (!found_entry)
8471 		return 0;
8472 
8473 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
8474 				    IXGBE_ATR_L4TYPE_MASK;
8475 
8476 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
8477 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
8478 
8479 	return 0;
8480 }
8481 
8482 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
8483 				  __be16 protocol,
8484 				  struct tc_cls_u32_offload *cls)
8485 {
8486 	u32 loc = cls->knode.handle & 0xfffff;
8487 	struct ixgbe_hw *hw = &adapter->hw;
8488 	struct ixgbe_mat_field *field_ptr;
8489 	struct ixgbe_fdir_filter *input = NULL;
8490 	union ixgbe_atr_input *mask = NULL;
8491 	struct ixgbe_jump_table *jump = NULL;
8492 	int i, err = -EINVAL;
8493 	u8 queue;
8494 	u32 uhtid, link_uhtid;
8495 
8496 	uhtid = TC_U32_USERHTID(cls->knode.handle);
8497 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
8498 
8499 	/* At the moment cls_u32 jumps to network layer and skips past
8500 	 * L2 headers. The canonical method to match L2 frames is to use
8501 	 * negative values. However this is error prone at best but really
8502 	 * just broken because there is no way to "know" what sort of hdr
8503 	 * is in front of the network layer. Fix cls_u32 to support L2
8504 	 * headers when needed.
8505 	 */
8506 	if (protocol != htons(ETH_P_IP))
8507 		return err;
8508 
8509 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
8510 		e_err(drv, "Location out of range\n");
8511 		return err;
8512 	}
8513 
8514 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
8515 	 * links and also the fields used to advance the parser across each
8516 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
8517 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
8518 	 * To add support for new nodes update ixgbe_model.h parse structures
8519 	 * this function _should_ be generic try not to hardcode values here.
8520 	 */
8521 	if (uhtid == 0x800) {
8522 		field_ptr = (adapter->jump_tables[0])->mat;
8523 	} else {
8524 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8525 			return err;
8526 		if (!adapter->jump_tables[uhtid])
8527 			return err;
8528 		field_ptr = (adapter->jump_tables[uhtid])->mat;
8529 	}
8530 
8531 	if (!field_ptr)
8532 		return err;
8533 
8534 	/* At this point we know the field_ptr is valid and need to either
8535 	 * build cls_u32 link or attach filter. Because adding a link to
8536 	 * a handle that does not exist is invalid and the same for adding
8537 	 * rules to handles that don't exist.
8538 	 */
8539 
8540 	if (link_uhtid) {
8541 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
8542 
8543 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
8544 			return err;
8545 
8546 		if (!test_bit(link_uhtid - 1, &adapter->tables))
8547 			return err;
8548 
8549 		/* Multiple filters as links to the same hash table are not
8550 		 * supported. To add a new filter with the same next header
8551 		 * but different match/jump conditions, create a new hash table
8552 		 * and link to it.
8553 		 */
8554 		if (adapter->jump_tables[link_uhtid] &&
8555 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
8556 			e_err(drv, "Link filter exists for link: %x\n",
8557 			      link_uhtid);
8558 			return err;
8559 		}
8560 
8561 		for (i = 0; nexthdr[i].jump; i++) {
8562 			if (nexthdr[i].o != cls->knode.sel->offoff ||
8563 			    nexthdr[i].s != cls->knode.sel->offshift ||
8564 			    nexthdr[i].m != cls->knode.sel->offmask)
8565 				return err;
8566 
8567 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
8568 			if (!jump)
8569 				return -ENOMEM;
8570 			input = kzalloc(sizeof(*input), GFP_KERNEL);
8571 			if (!input) {
8572 				err = -ENOMEM;
8573 				goto free_jump;
8574 			}
8575 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8576 			if (!mask) {
8577 				err = -ENOMEM;
8578 				goto free_input;
8579 			}
8580 			jump->input = input;
8581 			jump->mask = mask;
8582 			jump->link_hdl = cls->knode.handle;
8583 
8584 			err = ixgbe_clsu32_build_input(input, mask, cls,
8585 						       field_ptr, &nexthdr[i]);
8586 			if (!err) {
8587 				jump->mat = nexthdr[i].jump;
8588 				adapter->jump_tables[link_uhtid] = jump;
8589 				break;
8590 			}
8591 		}
8592 		return 0;
8593 	}
8594 
8595 	input = kzalloc(sizeof(*input), GFP_KERNEL);
8596 	if (!input)
8597 		return -ENOMEM;
8598 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
8599 	if (!mask) {
8600 		err = -ENOMEM;
8601 		goto free_input;
8602 	}
8603 
8604 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
8605 		if ((adapter->jump_tables[uhtid])->input)
8606 			memcpy(input, (adapter->jump_tables[uhtid])->input,
8607 			       sizeof(*input));
8608 		if ((adapter->jump_tables[uhtid])->mask)
8609 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
8610 			       sizeof(*mask));
8611 
8612 		/* Lookup in all child hash tables if this location is already
8613 		 * filled with a filter
8614 		 */
8615 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8616 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
8617 
8618 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
8619 				e_err(drv, "Filter exists in location: %x\n",
8620 				      loc);
8621 				err = -EINVAL;
8622 				goto err_out;
8623 			}
8624 		}
8625 	}
8626 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
8627 	if (err)
8628 		goto err_out;
8629 
8630 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
8631 			       &queue);
8632 	if (err < 0)
8633 		goto err_out;
8634 
8635 	input->sw_idx = loc;
8636 
8637 	spin_lock(&adapter->fdir_perfect_lock);
8638 
8639 	if (hlist_empty(&adapter->fdir_filter_list)) {
8640 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
8641 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
8642 		if (err)
8643 			goto err_out_w_lock;
8644 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
8645 		err = -EINVAL;
8646 		goto err_out_w_lock;
8647 	}
8648 
8649 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
8650 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
8651 						    input->sw_idx, queue);
8652 	if (!err)
8653 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
8654 	spin_unlock(&adapter->fdir_perfect_lock);
8655 
8656 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
8657 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
8658 
8659 	kfree(mask);
8660 	return err;
8661 err_out_w_lock:
8662 	spin_unlock(&adapter->fdir_perfect_lock);
8663 err_out:
8664 	kfree(mask);
8665 free_input:
8666 	kfree(input);
8667 free_jump:
8668 	kfree(jump);
8669 	return err;
8670 }
8671 
8672 static int __ixgbe_setup_tc(struct net_device *dev, u32 handle, __be16 proto,
8673 			    struct tc_to_netdev *tc)
8674 {
8675 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8676 
8677 	if (TC_H_MAJ(handle) == TC_H_MAJ(TC_H_INGRESS) &&
8678 	    tc->type == TC_SETUP_CLSU32) {
8679 		switch (tc->cls_u32->command) {
8680 		case TC_CLSU32_NEW_KNODE:
8681 		case TC_CLSU32_REPLACE_KNODE:
8682 			return ixgbe_configure_clsu32(adapter,
8683 						      proto, tc->cls_u32);
8684 		case TC_CLSU32_DELETE_KNODE:
8685 			return ixgbe_delete_clsu32(adapter, tc->cls_u32);
8686 		case TC_CLSU32_NEW_HNODE:
8687 		case TC_CLSU32_REPLACE_HNODE:
8688 			return ixgbe_configure_clsu32_add_hnode(adapter, proto,
8689 								tc->cls_u32);
8690 		case TC_CLSU32_DELETE_HNODE:
8691 			return ixgbe_configure_clsu32_del_hnode(adapter,
8692 								tc->cls_u32);
8693 		default:
8694 			return -EINVAL;
8695 		}
8696 	}
8697 
8698 	if (tc->type != TC_SETUP_MQPRIO)
8699 		return -EINVAL;
8700 
8701 	return ixgbe_setup_tc(dev, tc->tc);
8702 }
8703 
8704 #ifdef CONFIG_PCI_IOV
8705 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
8706 {
8707 	struct net_device *netdev = adapter->netdev;
8708 
8709 	rtnl_lock();
8710 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
8711 	rtnl_unlock();
8712 }
8713 
8714 #endif
8715 void ixgbe_do_reset(struct net_device *netdev)
8716 {
8717 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8718 
8719 	if (netif_running(netdev))
8720 		ixgbe_reinit_locked(adapter);
8721 	else
8722 		ixgbe_reset(adapter);
8723 }
8724 
8725 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
8726 					    netdev_features_t features)
8727 {
8728 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8729 
8730 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
8731 	if (!(features & NETIF_F_RXCSUM))
8732 		features &= ~NETIF_F_LRO;
8733 
8734 	/* Turn off LRO if not RSC capable */
8735 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
8736 		features &= ~NETIF_F_LRO;
8737 
8738 	return features;
8739 }
8740 
8741 static int ixgbe_set_features(struct net_device *netdev,
8742 			      netdev_features_t features)
8743 {
8744 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8745 	netdev_features_t changed = netdev->features ^ features;
8746 	bool need_reset = false;
8747 
8748 	/* Make sure RSC matches LRO, reset if change */
8749 	if (!(features & NETIF_F_LRO)) {
8750 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8751 			need_reset = true;
8752 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
8753 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
8754 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
8755 		if (adapter->rx_itr_setting == 1 ||
8756 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
8757 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
8758 			need_reset = true;
8759 		} else if ((changed ^ features) & NETIF_F_LRO) {
8760 			e_info(probe, "rx-usecs set too low, "
8761 			       "disabling RSC\n");
8762 		}
8763 	}
8764 
8765 	/*
8766 	 * Check if Flow Director n-tuple support or hw_tc support was
8767 	 * enabled or disabled.  If the state changed, we need to reset.
8768 	 */
8769 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
8770 		/* turn off ATR, enable perfect filters and reset */
8771 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
8772 			need_reset = true;
8773 
8774 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
8775 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8776 	} else {
8777 		/* turn off perfect filters, enable ATR and reset */
8778 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
8779 			need_reset = true;
8780 
8781 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8782 
8783 		/* We cannot enable ATR if SR-IOV is enabled */
8784 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
8785 		    /* We cannot enable ATR if we have 2 or more tcs */
8786 		    (netdev_get_num_tc(netdev) > 1) ||
8787 		    /* We cannot enable ATR if RSS is disabled */
8788 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
8789 		    /* A sample rate of 0 indicates ATR disabled */
8790 		    (!adapter->atr_sample_rate))
8791 			; /* do nothing not supported */
8792 		else /* otherwise supported and set the flag */
8793 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
8794 	}
8795 
8796 	if (changed & NETIF_F_RXALL)
8797 		need_reset = true;
8798 
8799 	netdev->features = features;
8800 
8801 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
8802 		if (features & NETIF_F_RXCSUM) {
8803 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
8804 		} else {
8805 			u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
8806 
8807 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
8808 		}
8809 	}
8810 
8811 	if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
8812 		if (features & NETIF_F_RXCSUM) {
8813 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
8814 		} else {
8815 			u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
8816 
8817 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
8818 		}
8819 	}
8820 
8821 	if (need_reset)
8822 		ixgbe_do_reset(netdev);
8823 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
8824 			    NETIF_F_HW_VLAN_CTAG_FILTER))
8825 		ixgbe_set_rx_mode(netdev);
8826 
8827 	return 0;
8828 }
8829 
8830 /**
8831  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
8832  * @dev: The port's netdev
8833  * @ti: Tunnel endpoint information
8834  **/
8835 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
8836 				      struct udp_tunnel_info *ti)
8837 {
8838 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8839 	struct ixgbe_hw *hw = &adapter->hw;
8840 	__be16 port = ti->port;
8841 	u32 port_shift = 0;
8842 	u32 reg;
8843 
8844 	if (ti->sa_family != AF_INET)
8845 		return;
8846 
8847 	switch (ti->type) {
8848 	case UDP_TUNNEL_TYPE_VXLAN:
8849 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8850 			return;
8851 
8852 		if (adapter->vxlan_port == port)
8853 			return;
8854 
8855 		if (adapter->vxlan_port) {
8856 			netdev_info(dev,
8857 				    "VXLAN port %d set, not adding port %d\n",
8858 				    ntohs(adapter->vxlan_port),
8859 				    ntohs(port));
8860 			return;
8861 		}
8862 
8863 		adapter->vxlan_port = port;
8864 		break;
8865 	case UDP_TUNNEL_TYPE_GENEVE:
8866 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
8867 			return;
8868 
8869 		if (adapter->geneve_port == port)
8870 			return;
8871 
8872 		if (adapter->geneve_port) {
8873 			netdev_info(dev,
8874 				    "GENEVE port %d set, not adding port %d\n",
8875 				    ntohs(adapter->geneve_port),
8876 				    ntohs(port));
8877 			return;
8878 		}
8879 
8880 		port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
8881 		adapter->geneve_port = port;
8882 		break;
8883 	default:
8884 		return;
8885 	}
8886 
8887 	reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
8888 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
8889 }
8890 
8891 /**
8892  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
8893  * @dev: The port's netdev
8894  * @ti: Tunnel endpoint information
8895  **/
8896 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
8897 				      struct udp_tunnel_info *ti)
8898 {
8899 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8900 	u32 port_mask;
8901 
8902 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
8903 	    ti->type != UDP_TUNNEL_TYPE_GENEVE)
8904 		return;
8905 
8906 	if (ti->sa_family != AF_INET)
8907 		return;
8908 
8909 	switch (ti->type) {
8910 	case UDP_TUNNEL_TYPE_VXLAN:
8911 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8912 			return;
8913 
8914 		if (adapter->vxlan_port != ti->port) {
8915 			netdev_info(dev, "VXLAN port %d not found\n",
8916 				    ntohs(ti->port));
8917 			return;
8918 		}
8919 
8920 		port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
8921 		break;
8922 	case UDP_TUNNEL_TYPE_GENEVE:
8923 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
8924 			return;
8925 
8926 		if (adapter->geneve_port != ti->port) {
8927 			netdev_info(dev, "GENEVE port %d not found\n",
8928 				    ntohs(ti->port));
8929 			return;
8930 		}
8931 
8932 		port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
8933 		break;
8934 	default:
8935 		return;
8936 	}
8937 
8938 	ixgbe_clear_udp_tunnel_port(adapter, port_mask);
8939 	adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
8940 }
8941 
8942 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8943 			     struct net_device *dev,
8944 			     const unsigned char *addr, u16 vid,
8945 			     u16 flags)
8946 {
8947 	/* guarantee we can provide a unique filter for the unicast address */
8948 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
8949 		struct ixgbe_adapter *adapter = netdev_priv(dev);
8950 		u16 pool = VMDQ_P(0);
8951 
8952 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
8953 			return -ENOMEM;
8954 	}
8955 
8956 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
8957 }
8958 
8959 /**
8960  * ixgbe_configure_bridge_mode - set various bridge modes
8961  * @adapter - the private structure
8962  * @mode - requested bridge mode
8963  *
8964  * Configure some settings require for various bridge modes.
8965  **/
8966 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
8967 				       __u16 mode)
8968 {
8969 	struct ixgbe_hw *hw = &adapter->hw;
8970 	unsigned int p, num_pools;
8971 	u32 vmdctl;
8972 
8973 	switch (mode) {
8974 	case BRIDGE_MODE_VEPA:
8975 		/* disable Tx loopback, rely on switch hairpin mode */
8976 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
8977 
8978 		/* must enable Rx switching replication to allow multicast
8979 		 * packet reception on all VFs, and to enable source address
8980 		 * pruning.
8981 		 */
8982 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
8983 		vmdctl |= IXGBE_VT_CTL_REPLEN;
8984 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
8985 
8986 		/* enable Rx source address pruning. Note, this requires
8987 		 * replication to be enabled or else it does nothing.
8988 		 */
8989 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
8990 		for (p = 0; p < num_pools; p++) {
8991 			if (hw->mac.ops.set_source_address_pruning)
8992 				hw->mac.ops.set_source_address_pruning(hw,
8993 								       true,
8994 								       p);
8995 		}
8996 		break;
8997 	case BRIDGE_MODE_VEB:
8998 		/* enable Tx loopback for internal VF/PF communication */
8999 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9000 				IXGBE_PFDTXGSWC_VT_LBEN);
9001 
9002 		/* disable Rx switching replication unless we have SR-IOV
9003 		 * virtual functions
9004 		 */
9005 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9006 		if (!adapter->num_vfs)
9007 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9008 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9009 
9010 		/* disable Rx source address pruning, since we don't expect to
9011 		 * be receiving external loopback of our transmitted frames.
9012 		 */
9013 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9014 		for (p = 0; p < num_pools; p++) {
9015 			if (hw->mac.ops.set_source_address_pruning)
9016 				hw->mac.ops.set_source_address_pruning(hw,
9017 								       false,
9018 								       p);
9019 		}
9020 		break;
9021 	default:
9022 		return -EINVAL;
9023 	}
9024 
9025 	adapter->bridge_mode = mode;
9026 
9027 	e_info(drv, "enabling bridge mode: %s\n",
9028 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9029 
9030 	return 0;
9031 }
9032 
9033 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9034 				    struct nlmsghdr *nlh, u16 flags)
9035 {
9036 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9037 	struct nlattr *attr, *br_spec;
9038 	int rem;
9039 
9040 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9041 		return -EOPNOTSUPP;
9042 
9043 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9044 	if (!br_spec)
9045 		return -EINVAL;
9046 
9047 	nla_for_each_nested(attr, br_spec, rem) {
9048 		int status;
9049 		__u16 mode;
9050 
9051 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9052 			continue;
9053 
9054 		if (nla_len(attr) < sizeof(mode))
9055 			return -EINVAL;
9056 
9057 		mode = nla_get_u16(attr);
9058 		status = ixgbe_configure_bridge_mode(adapter, mode);
9059 		if (status)
9060 			return status;
9061 
9062 		break;
9063 	}
9064 
9065 	return 0;
9066 }
9067 
9068 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9069 				    struct net_device *dev,
9070 				    u32 filter_mask, int nlflags)
9071 {
9072 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9073 
9074 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9075 		return 0;
9076 
9077 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9078 				       adapter->bridge_mode, 0, 0, nlflags,
9079 				       filter_mask, NULL);
9080 }
9081 
9082 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9083 {
9084 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
9085 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9086 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
9087 	unsigned int limit;
9088 	int pool, err;
9089 
9090 	/* Hardware has a limited number of available pools. Each VF, and the
9091 	 * PF require a pool. Check to ensure we don't attempt to use more
9092 	 * then the available number of pools.
9093 	 */
9094 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9095 		return ERR_PTR(-EINVAL);
9096 
9097 #ifdef CONFIG_RPS
9098 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
9099 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
9100 			    vdev->name);
9101 		return ERR_PTR(-EINVAL);
9102 	}
9103 #endif
9104 	/* Check for hardware restriction on number of rx/tx queues */
9105 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
9106 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
9107 		netdev_info(pdev,
9108 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
9109 			    pdev->name);
9110 		return ERR_PTR(-EINVAL);
9111 	}
9112 
9113 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9114 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
9115 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
9116 		return ERR_PTR(-EBUSY);
9117 
9118 	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
9119 	if (!fwd_adapter)
9120 		return ERR_PTR(-ENOMEM);
9121 
9122 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
9123 	adapter->num_rx_pools++;
9124 	set_bit(pool, &adapter->fwd_bitmask);
9125 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9126 
9127 	/* Enable VMDq flag so device will be set in VM mode */
9128 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
9129 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9130 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
9131 
9132 	/* Force reinit of ring allocation with VMDQ enabled */
9133 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9134 	if (err)
9135 		goto fwd_add_err;
9136 	fwd_adapter->pool = pool;
9137 	fwd_adapter->real_adapter = adapter;
9138 	err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
9139 	if (err)
9140 		goto fwd_add_err;
9141 	netif_tx_start_all_queues(vdev);
9142 	return fwd_adapter;
9143 fwd_add_err:
9144 	/* unwind counter and free adapter struct */
9145 	netdev_info(pdev,
9146 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
9147 	clear_bit(pool, &adapter->fwd_bitmask);
9148 	adapter->num_rx_pools--;
9149 	kfree(fwd_adapter);
9150 	return ERR_PTR(err);
9151 }
9152 
9153 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9154 {
9155 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
9156 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
9157 	unsigned int limit;
9158 
9159 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
9160 	adapter->num_rx_pools--;
9161 
9162 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
9163 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
9164 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
9165 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
9166 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
9167 		   fwd_adapter->pool, adapter->num_rx_pools,
9168 		   fwd_adapter->rx_base_queue,
9169 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
9170 		   adapter->fwd_bitmask);
9171 	kfree(fwd_adapter);
9172 }
9173 
9174 #define IXGBE_MAX_MAC_HDR_LEN		127
9175 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9176 
9177 static netdev_features_t
9178 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9179 		     netdev_features_t features)
9180 {
9181 	unsigned int network_hdr_len, mac_hdr_len;
9182 
9183 	/* Make certain the headers can be described by a context descriptor */
9184 	mac_hdr_len = skb_network_header(skb) - skb->data;
9185 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9186 		return features & ~(NETIF_F_HW_CSUM |
9187 				    NETIF_F_SCTP_CRC |
9188 				    NETIF_F_HW_VLAN_CTAG_TX |
9189 				    NETIF_F_TSO |
9190 				    NETIF_F_TSO6);
9191 
9192 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9193 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9194 		return features & ~(NETIF_F_HW_CSUM |
9195 				    NETIF_F_SCTP_CRC |
9196 				    NETIF_F_TSO |
9197 				    NETIF_F_TSO6);
9198 
9199 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9200 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9201 	 */
9202 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
9203 		features &= ~NETIF_F_TSO;
9204 
9205 	return features;
9206 }
9207 
9208 static const struct net_device_ops ixgbe_netdev_ops = {
9209 	.ndo_open		= ixgbe_open,
9210 	.ndo_stop		= ixgbe_close,
9211 	.ndo_start_xmit		= ixgbe_xmit_frame,
9212 	.ndo_select_queue	= ixgbe_select_queue,
9213 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
9214 	.ndo_validate_addr	= eth_validate_addr,
9215 	.ndo_set_mac_address	= ixgbe_set_mac,
9216 	.ndo_change_mtu		= ixgbe_change_mtu,
9217 	.ndo_tx_timeout		= ixgbe_tx_timeout,
9218 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
9219 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
9220 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
9221 	.ndo_do_ioctl		= ixgbe_ioctl,
9222 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
9223 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
9224 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
9225 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
9226 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
9227 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
9228 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
9229 	.ndo_get_stats64	= ixgbe_get_stats64,
9230 	.ndo_setup_tc		= __ixgbe_setup_tc,
9231 #ifdef CONFIG_NET_POLL_CONTROLLER
9232 	.ndo_poll_controller	= ixgbe_netpoll,
9233 #endif
9234 #ifdef CONFIG_NET_RX_BUSY_POLL
9235 	.ndo_busy_poll		= ixgbe_low_latency_recv,
9236 #endif
9237 #ifdef IXGBE_FCOE
9238 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
9239 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
9240 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
9241 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
9242 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
9243 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
9244 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
9245 #endif /* IXGBE_FCOE */
9246 	.ndo_set_features = ixgbe_set_features,
9247 	.ndo_fix_features = ixgbe_fix_features,
9248 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
9249 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
9250 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
9251 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
9252 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
9253 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
9254 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
9255 	.ndo_features_check	= ixgbe_features_check,
9256 };
9257 
9258 /**
9259  * ixgbe_enumerate_functions - Get the number of ports this device has
9260  * @adapter: adapter structure
9261  *
9262  * This function enumerates the phsyical functions co-located on a single slot,
9263  * in order to determine how many ports a device has. This is most useful in
9264  * determining the required GT/s of PCIe bandwidth necessary for optimal
9265  * performance.
9266  **/
9267 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
9268 {
9269 	struct pci_dev *entry, *pdev = adapter->pdev;
9270 	int physfns = 0;
9271 
9272 	/* Some cards can not use the generic count PCIe functions method,
9273 	 * because they are behind a parent switch, so we hardcode these with
9274 	 * the correct number of functions.
9275 	 */
9276 	if (ixgbe_pcie_from_parent(&adapter->hw))
9277 		physfns = 4;
9278 
9279 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
9280 		/* don't count virtual functions */
9281 		if (entry->is_virtfn)
9282 			continue;
9283 
9284 		/* When the devices on the bus don't all match our device ID,
9285 		 * we can't reliably determine the correct number of
9286 		 * functions. This can occur if a function has been direct
9287 		 * attached to a virtual machine using VT-d, for example. In
9288 		 * this case, simply return -1 to indicate this.
9289 		 */
9290 		if ((entry->vendor != pdev->vendor) ||
9291 		    (entry->device != pdev->device))
9292 			return -1;
9293 
9294 		physfns++;
9295 	}
9296 
9297 	return physfns;
9298 }
9299 
9300 /**
9301  * ixgbe_wol_supported - Check whether device supports WoL
9302  * @adapter: the adapter private structure
9303  * @device_id: the device ID
9304  * @subdev_id: the subsystem device ID
9305  *
9306  * This function is used by probe and ethtool to determine
9307  * which devices have WoL support
9308  *
9309  **/
9310 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
9311 			 u16 subdevice_id)
9312 {
9313 	struct ixgbe_hw *hw = &adapter->hw;
9314 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
9315 
9316 	/* WOL not supported on 82598 */
9317 	if (hw->mac.type == ixgbe_mac_82598EB)
9318 		return false;
9319 
9320 	/* check eeprom to see if WOL is enabled for X540 and newer */
9321 	if (hw->mac.type >= ixgbe_mac_X540) {
9322 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
9323 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
9324 		     (hw->bus.func == 0)))
9325 			return true;
9326 	}
9327 
9328 	/* WOL is determined based on device IDs for 82599 MACs */
9329 	switch (device_id) {
9330 	case IXGBE_DEV_ID_82599_SFP:
9331 		/* Only these subdevices could supports WOL */
9332 		switch (subdevice_id) {
9333 		case IXGBE_SUBDEV_ID_82599_560FLR:
9334 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
9335 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
9336 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
9337 			/* only support first port */
9338 			if (hw->bus.func != 0)
9339 				break;
9340 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
9341 		case IXGBE_SUBDEV_ID_82599_SFP:
9342 		case IXGBE_SUBDEV_ID_82599_RNDC:
9343 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
9344 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
9345 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
9346 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
9347 			return true;
9348 		}
9349 		break;
9350 	case IXGBE_DEV_ID_82599EN_SFP:
9351 		/* Only these subdevices support WOL */
9352 		switch (subdevice_id) {
9353 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
9354 			return true;
9355 		}
9356 		break;
9357 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
9358 		/* All except this subdevice support WOL */
9359 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
9360 			return true;
9361 		break;
9362 	case IXGBE_DEV_ID_82599_KX4:
9363 		return  true;
9364 	default:
9365 		break;
9366 	}
9367 
9368 	return false;
9369 }
9370 
9371 /**
9372  * ixgbe_probe - Device Initialization Routine
9373  * @pdev: PCI device information struct
9374  * @ent: entry in ixgbe_pci_tbl
9375  *
9376  * Returns 0 on success, negative on failure
9377  *
9378  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
9379  * The OS initialization, configuring of the adapter private structure,
9380  * and a hardware reset occur.
9381  **/
9382 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
9383 {
9384 	struct net_device *netdev;
9385 	struct ixgbe_adapter *adapter = NULL;
9386 	struct ixgbe_hw *hw;
9387 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
9388 	int i, err, pci_using_dac, expected_gts;
9389 	unsigned int indices = MAX_TX_QUEUES;
9390 	u8 part_str[IXGBE_PBANUM_LENGTH];
9391 	bool disable_dev = false;
9392 #ifdef IXGBE_FCOE
9393 	u16 device_caps;
9394 #endif
9395 	u32 eec;
9396 
9397 	/* Catch broken hardware that put the wrong VF device ID in
9398 	 * the PCIe SR-IOV capability.
9399 	 */
9400 	if (pdev->is_virtfn) {
9401 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
9402 		     pci_name(pdev), pdev->vendor, pdev->device);
9403 		return -EINVAL;
9404 	}
9405 
9406 	err = pci_enable_device_mem(pdev);
9407 	if (err)
9408 		return err;
9409 
9410 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
9411 		pci_using_dac = 1;
9412 	} else {
9413 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
9414 		if (err) {
9415 			dev_err(&pdev->dev,
9416 				"No usable DMA configuration, aborting\n");
9417 			goto err_dma;
9418 		}
9419 		pci_using_dac = 0;
9420 	}
9421 
9422 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
9423 	if (err) {
9424 		dev_err(&pdev->dev,
9425 			"pci_request_selected_regions failed 0x%x\n", err);
9426 		goto err_pci_reg;
9427 	}
9428 
9429 	pci_enable_pcie_error_reporting(pdev);
9430 
9431 	pci_set_master(pdev);
9432 	pci_save_state(pdev);
9433 
9434 	if (ii->mac == ixgbe_mac_82598EB) {
9435 #ifdef CONFIG_IXGBE_DCB
9436 		/* 8 TC w/ 4 queues per TC */
9437 		indices = 4 * MAX_TRAFFIC_CLASS;
9438 #else
9439 		indices = IXGBE_MAX_RSS_INDICES;
9440 #endif
9441 	}
9442 
9443 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
9444 	if (!netdev) {
9445 		err = -ENOMEM;
9446 		goto err_alloc_etherdev;
9447 	}
9448 
9449 	SET_NETDEV_DEV(netdev, &pdev->dev);
9450 
9451 	adapter = netdev_priv(netdev);
9452 
9453 	adapter->netdev = netdev;
9454 	adapter->pdev = pdev;
9455 	hw = &adapter->hw;
9456 	hw->back = adapter;
9457 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
9458 
9459 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
9460 			      pci_resource_len(pdev, 0));
9461 	adapter->io_addr = hw->hw_addr;
9462 	if (!hw->hw_addr) {
9463 		err = -EIO;
9464 		goto err_ioremap;
9465 	}
9466 
9467 	netdev->netdev_ops = &ixgbe_netdev_ops;
9468 	ixgbe_set_ethtool_ops(netdev);
9469 	netdev->watchdog_timeo = 5 * HZ;
9470 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
9471 
9472 	/* Setup hw api */
9473 	hw->mac.ops   = *ii->mac_ops;
9474 	hw->mac.type  = ii->mac;
9475 	hw->mvals     = ii->mvals;
9476 
9477 	/* EEPROM */
9478 	hw->eeprom.ops = *ii->eeprom_ops;
9479 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
9480 	if (ixgbe_removed(hw->hw_addr)) {
9481 		err = -EIO;
9482 		goto err_ioremap;
9483 	}
9484 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
9485 	if (!(eec & BIT(8)))
9486 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
9487 
9488 	/* PHY */
9489 	hw->phy.ops = *ii->phy_ops;
9490 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
9491 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
9492 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
9493 	hw->phy.mdio.mmds = 0;
9494 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
9495 	hw->phy.mdio.dev = netdev;
9496 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
9497 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
9498 
9499 	ii->get_invariants(hw);
9500 
9501 	/* setup the private structure */
9502 	err = ixgbe_sw_init(adapter);
9503 	if (err)
9504 		goto err_sw_init;
9505 
9506 	/* Make sure the SWFW semaphore is in a valid state */
9507 	if (hw->mac.ops.init_swfw_sync)
9508 		hw->mac.ops.init_swfw_sync(hw);
9509 
9510 	/* Make it possible the adapter to be woken up via WOL */
9511 	switch (adapter->hw.mac.type) {
9512 	case ixgbe_mac_82599EB:
9513 	case ixgbe_mac_X540:
9514 	case ixgbe_mac_X550:
9515 	case ixgbe_mac_X550EM_x:
9516 	case ixgbe_mac_x550em_a:
9517 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
9518 		break;
9519 	default:
9520 		break;
9521 	}
9522 
9523 	/*
9524 	 * If there is a fan on this device and it has failed log the
9525 	 * failure.
9526 	 */
9527 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
9528 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
9529 		if (esdp & IXGBE_ESDP_SDP1)
9530 			e_crit(probe, "Fan has stopped, replace the adapter\n");
9531 	}
9532 
9533 	if (allow_unsupported_sfp)
9534 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
9535 
9536 	/* reset_hw fills in the perm_addr as well */
9537 	hw->phy.reset_if_overtemp = true;
9538 	err = hw->mac.ops.reset_hw(hw);
9539 	hw->phy.reset_if_overtemp = false;
9540 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
9541 		err = 0;
9542 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
9543 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
9544 		e_dev_err("Reload the driver after installing a supported module.\n");
9545 		goto err_sw_init;
9546 	} else if (err) {
9547 		e_dev_err("HW Init failed: %d\n", err);
9548 		goto err_sw_init;
9549 	}
9550 
9551 #ifdef CONFIG_PCI_IOV
9552 	/* SR-IOV not supported on the 82598 */
9553 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9554 		goto skip_sriov;
9555 	/* Mailbox */
9556 	ixgbe_init_mbx_params_pf(hw);
9557 	hw->mbx.ops = ii->mbx_ops;
9558 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
9559 	ixgbe_enable_sriov(adapter);
9560 skip_sriov:
9561 
9562 #endif
9563 	netdev->features = NETIF_F_SG |
9564 			   NETIF_F_TSO |
9565 			   NETIF_F_TSO6 |
9566 			   NETIF_F_RXHASH |
9567 			   NETIF_F_RXCSUM |
9568 			   NETIF_F_HW_CSUM;
9569 
9570 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
9571 				    NETIF_F_GSO_GRE_CSUM | \
9572 				    NETIF_F_GSO_IPXIP4 | \
9573 				    NETIF_F_GSO_IPXIP6 | \
9574 				    NETIF_F_GSO_UDP_TUNNEL | \
9575 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
9576 
9577 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
9578 	netdev->features |= NETIF_F_GSO_PARTIAL |
9579 			    IXGBE_GSO_PARTIAL_FEATURES;
9580 
9581 	if (hw->mac.type >= ixgbe_mac_82599EB)
9582 		netdev->features |= NETIF_F_SCTP_CRC;
9583 
9584 	/* copy netdev features into list of user selectable features */
9585 	netdev->hw_features |= netdev->features |
9586 			       NETIF_F_HW_VLAN_CTAG_FILTER |
9587 			       NETIF_F_HW_VLAN_CTAG_RX |
9588 			       NETIF_F_HW_VLAN_CTAG_TX |
9589 			       NETIF_F_RXALL |
9590 			       NETIF_F_HW_L2FW_DOFFLOAD;
9591 
9592 	if (hw->mac.type >= ixgbe_mac_82599EB)
9593 		netdev->hw_features |= NETIF_F_NTUPLE |
9594 				       NETIF_F_HW_TC;
9595 
9596 	if (pci_using_dac)
9597 		netdev->features |= NETIF_F_HIGHDMA;
9598 
9599 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
9600 	netdev->hw_enc_features |= netdev->vlan_features;
9601 	netdev->mpls_features |= NETIF_F_HW_CSUM;
9602 
9603 	/* set this bit last since it cannot be part of vlan_features */
9604 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
9605 			    NETIF_F_HW_VLAN_CTAG_RX |
9606 			    NETIF_F_HW_VLAN_CTAG_TX;
9607 
9608 	netdev->priv_flags |= IFF_UNICAST_FLT;
9609 	netdev->priv_flags |= IFF_SUPP_NOFCS;
9610 
9611 #ifdef CONFIG_IXGBE_DCB
9612 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
9613 		netdev->dcbnl_ops = &dcbnl_ops;
9614 #endif
9615 
9616 #ifdef IXGBE_FCOE
9617 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
9618 		unsigned int fcoe_l;
9619 
9620 		if (hw->mac.ops.get_device_caps) {
9621 			hw->mac.ops.get_device_caps(hw, &device_caps);
9622 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
9623 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
9624 		}
9625 
9626 
9627 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
9628 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
9629 
9630 		netdev->features |= NETIF_F_FSO |
9631 				    NETIF_F_FCOE_CRC;
9632 
9633 		netdev->vlan_features |= NETIF_F_FSO |
9634 					 NETIF_F_FCOE_CRC |
9635 					 NETIF_F_FCOE_MTU;
9636 	}
9637 #endif /* IXGBE_FCOE */
9638 
9639 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
9640 		netdev->hw_features |= NETIF_F_LRO;
9641 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9642 		netdev->features |= NETIF_F_LRO;
9643 
9644 	/* make sure the EEPROM is good */
9645 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
9646 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
9647 		err = -EIO;
9648 		goto err_sw_init;
9649 	}
9650 
9651 	eth_platform_get_mac_address(&adapter->pdev->dev,
9652 				     adapter->hw.mac.perm_addr);
9653 
9654 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
9655 
9656 	if (!is_valid_ether_addr(netdev->dev_addr)) {
9657 		e_dev_err("invalid MAC address\n");
9658 		err = -EIO;
9659 		goto err_sw_init;
9660 	}
9661 
9662 	/* Set hw->mac.addr to permanent MAC address */
9663 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
9664 	ixgbe_mac_set_default_filter(adapter);
9665 
9666 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
9667 		    (unsigned long) adapter);
9668 
9669 	if (ixgbe_removed(hw->hw_addr)) {
9670 		err = -EIO;
9671 		goto err_sw_init;
9672 	}
9673 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
9674 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
9675 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
9676 
9677 	err = ixgbe_init_interrupt_scheme(adapter);
9678 	if (err)
9679 		goto err_sw_init;
9680 
9681 	/* WOL not supported for all devices */
9682 	adapter->wol = 0;
9683 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
9684 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
9685 						pdev->subsystem_device);
9686 	if (hw->wol_enabled)
9687 		adapter->wol = IXGBE_WUFC_MAG;
9688 
9689 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
9690 
9691 	/* save off EEPROM version number */
9692 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
9693 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
9694 
9695 	/* pick up the PCI bus settings for reporting later */
9696 	if (ixgbe_pcie_from_parent(hw))
9697 		ixgbe_get_parent_bus_info(adapter);
9698 	else
9699 		 hw->mac.ops.get_bus_info(hw);
9700 
9701 	/* calculate the expected PCIe bandwidth required for optimal
9702 	 * performance. Note that some older parts will never have enough
9703 	 * bandwidth due to being older generation PCIe parts. We clamp these
9704 	 * parts to ensure no warning is displayed if it can't be fixed.
9705 	 */
9706 	switch (hw->mac.type) {
9707 	case ixgbe_mac_82598EB:
9708 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
9709 		break;
9710 	default:
9711 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
9712 		break;
9713 	}
9714 
9715 	/* don't check link if we failed to enumerate functions */
9716 	if (expected_gts > 0)
9717 		ixgbe_check_minimum_link(adapter, expected_gts);
9718 
9719 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
9720 	if (err)
9721 		strlcpy(part_str, "Unknown", sizeof(part_str));
9722 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
9723 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
9724 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
9725 			   part_str);
9726 	else
9727 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
9728 			   hw->mac.type, hw->phy.type, part_str);
9729 
9730 	e_dev_info("%pM\n", netdev->dev_addr);
9731 
9732 	/* reset the hardware with the new settings */
9733 	err = hw->mac.ops.start_hw(hw);
9734 	if (err == IXGBE_ERR_EEPROM_VERSION) {
9735 		/* We are running on a pre-production device, log a warning */
9736 		e_dev_warn("This device is a pre-production adapter/LOM. "
9737 			   "Please be aware there may be issues associated "
9738 			   "with your hardware.  If you are experiencing "
9739 			   "problems please contact your Intel or hardware "
9740 			   "representative who provided you with this "
9741 			   "hardware.\n");
9742 	}
9743 	strcpy(netdev->name, "eth%d");
9744 	err = register_netdev(netdev);
9745 	if (err)
9746 		goto err_register;
9747 
9748 	pci_set_drvdata(pdev, adapter);
9749 
9750 	/* power down the optics for 82599 SFP+ fiber */
9751 	if (hw->mac.ops.disable_tx_laser)
9752 		hw->mac.ops.disable_tx_laser(hw);
9753 
9754 	/* carrier off reporting is important to ethtool even BEFORE open */
9755 	netif_carrier_off(netdev);
9756 
9757 #ifdef CONFIG_IXGBE_DCA
9758 	if (dca_add_requester(&pdev->dev) == 0) {
9759 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
9760 		ixgbe_setup_dca(adapter);
9761 	}
9762 #endif
9763 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
9764 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
9765 		for (i = 0; i < adapter->num_vfs; i++)
9766 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
9767 	}
9768 
9769 	/* firmware requires driver version to be 0xFFFFFFFF
9770 	 * since os does not support feature
9771 	 */
9772 	if (hw->mac.ops.set_fw_drv_ver)
9773 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
9774 					   0xFF);
9775 
9776 	/* add san mac addr to netdev */
9777 	ixgbe_add_sanmac_netdev(netdev);
9778 
9779 	e_dev_info("%s\n", ixgbe_default_device_descr);
9780 
9781 #ifdef CONFIG_IXGBE_HWMON
9782 	if (ixgbe_sysfs_init(adapter))
9783 		e_err(probe, "failed to allocate sysfs resources\n");
9784 #endif /* CONFIG_IXGBE_HWMON */
9785 
9786 	ixgbe_dbg_adapter_init(adapter);
9787 
9788 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
9789 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
9790 		hw->mac.ops.setup_link(hw,
9791 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
9792 			true);
9793 
9794 	return 0;
9795 
9796 err_register:
9797 	ixgbe_release_hw_control(adapter);
9798 	ixgbe_clear_interrupt_scheme(adapter);
9799 err_sw_init:
9800 	ixgbe_disable_sriov(adapter);
9801 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
9802 	iounmap(adapter->io_addr);
9803 	kfree(adapter->jump_tables[0]);
9804 	kfree(adapter->mac_table);
9805 err_ioremap:
9806 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9807 	free_netdev(netdev);
9808 err_alloc_etherdev:
9809 	pci_release_mem_regions(pdev);
9810 err_pci_reg:
9811 err_dma:
9812 	if (!adapter || disable_dev)
9813 		pci_disable_device(pdev);
9814 	return err;
9815 }
9816 
9817 /**
9818  * ixgbe_remove - Device Removal Routine
9819  * @pdev: PCI device information struct
9820  *
9821  * ixgbe_remove is called by the PCI subsystem to alert the driver
9822  * that it should release a PCI device.  The could be caused by a
9823  * Hot-Plug event, or because the driver is going to be removed from
9824  * memory.
9825  **/
9826 static void ixgbe_remove(struct pci_dev *pdev)
9827 {
9828 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9829 	struct net_device *netdev;
9830 	bool disable_dev;
9831 	int i;
9832 
9833 	/* if !adapter then we already cleaned up in probe */
9834 	if (!adapter)
9835 		return;
9836 
9837 	netdev  = adapter->netdev;
9838 	ixgbe_dbg_adapter_exit(adapter);
9839 
9840 	set_bit(__IXGBE_REMOVING, &adapter->state);
9841 	cancel_work_sync(&adapter->service_task);
9842 
9843 
9844 #ifdef CONFIG_IXGBE_DCA
9845 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
9846 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
9847 		dca_remove_requester(&pdev->dev);
9848 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
9849 				IXGBE_DCA_CTRL_DCA_DISABLE);
9850 	}
9851 
9852 #endif
9853 #ifdef CONFIG_IXGBE_HWMON
9854 	ixgbe_sysfs_exit(adapter);
9855 #endif /* CONFIG_IXGBE_HWMON */
9856 
9857 	/* remove the added san mac */
9858 	ixgbe_del_sanmac_netdev(netdev);
9859 
9860 #ifdef CONFIG_PCI_IOV
9861 	ixgbe_disable_sriov(adapter);
9862 #endif
9863 	if (netdev->reg_state == NETREG_REGISTERED)
9864 		unregister_netdev(netdev);
9865 
9866 	ixgbe_clear_interrupt_scheme(adapter);
9867 
9868 	ixgbe_release_hw_control(adapter);
9869 
9870 #ifdef CONFIG_DCB
9871 	kfree(adapter->ixgbe_ieee_pfc);
9872 	kfree(adapter->ixgbe_ieee_ets);
9873 
9874 #endif
9875 	iounmap(adapter->io_addr);
9876 	pci_release_mem_regions(pdev);
9877 
9878 	e_dev_info("complete\n");
9879 
9880 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
9881 		if (adapter->jump_tables[i]) {
9882 			kfree(adapter->jump_tables[i]->input);
9883 			kfree(adapter->jump_tables[i]->mask);
9884 		}
9885 		kfree(adapter->jump_tables[i]);
9886 	}
9887 
9888 	kfree(adapter->mac_table);
9889 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9890 	free_netdev(netdev);
9891 
9892 	pci_disable_pcie_error_reporting(pdev);
9893 
9894 	if (disable_dev)
9895 		pci_disable_device(pdev);
9896 }
9897 
9898 /**
9899  * ixgbe_io_error_detected - called when PCI error is detected
9900  * @pdev: Pointer to PCI device
9901  * @state: The current pci connection state
9902  *
9903  * This function is called after a PCI bus error affecting
9904  * this device has been detected.
9905  */
9906 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
9907 						pci_channel_state_t state)
9908 {
9909 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9910 	struct net_device *netdev = adapter->netdev;
9911 
9912 #ifdef CONFIG_PCI_IOV
9913 	struct ixgbe_hw *hw = &adapter->hw;
9914 	struct pci_dev *bdev, *vfdev;
9915 	u32 dw0, dw1, dw2, dw3;
9916 	int vf, pos;
9917 	u16 req_id, pf_func;
9918 
9919 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
9920 	    adapter->num_vfs == 0)
9921 		goto skip_bad_vf_detection;
9922 
9923 	bdev = pdev->bus->self;
9924 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
9925 		bdev = bdev->bus->self;
9926 
9927 	if (!bdev)
9928 		goto skip_bad_vf_detection;
9929 
9930 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
9931 	if (!pos)
9932 		goto skip_bad_vf_detection;
9933 
9934 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
9935 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
9936 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
9937 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
9938 	if (ixgbe_removed(hw->hw_addr))
9939 		goto skip_bad_vf_detection;
9940 
9941 	req_id = dw1 >> 16;
9942 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
9943 	if (!(req_id & 0x0080))
9944 		goto skip_bad_vf_detection;
9945 
9946 	pf_func = req_id & 0x01;
9947 	if ((pf_func & 1) == (pdev->devfn & 1)) {
9948 		unsigned int device_id;
9949 
9950 		vf = (req_id & 0x7F) >> 1;
9951 		e_dev_err("VF %d has caused a PCIe error\n", vf);
9952 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
9953 				"%8.8x\tdw3: %8.8x\n",
9954 		dw0, dw1, dw2, dw3);
9955 		switch (adapter->hw.mac.type) {
9956 		case ixgbe_mac_82599EB:
9957 			device_id = IXGBE_82599_VF_DEVICE_ID;
9958 			break;
9959 		case ixgbe_mac_X540:
9960 			device_id = IXGBE_X540_VF_DEVICE_ID;
9961 			break;
9962 		case ixgbe_mac_X550:
9963 			device_id = IXGBE_DEV_ID_X550_VF;
9964 			break;
9965 		case ixgbe_mac_X550EM_x:
9966 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
9967 			break;
9968 		case ixgbe_mac_x550em_a:
9969 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
9970 			break;
9971 		default:
9972 			device_id = 0;
9973 			break;
9974 		}
9975 
9976 		/* Find the pci device of the offending VF */
9977 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
9978 		while (vfdev) {
9979 			if (vfdev->devfn == (req_id & 0xFF))
9980 				break;
9981 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
9982 					       device_id, vfdev);
9983 		}
9984 		/*
9985 		 * There's a slim chance the VF could have been hot plugged,
9986 		 * so if it is no longer present we don't need to issue the
9987 		 * VFLR.  Just clean up the AER in that case.
9988 		 */
9989 		if (vfdev) {
9990 			ixgbe_issue_vf_flr(adapter, vfdev);
9991 			/* Free device reference count */
9992 			pci_dev_put(vfdev);
9993 		}
9994 
9995 		pci_cleanup_aer_uncorrect_error_status(pdev);
9996 	}
9997 
9998 	/*
9999 	 * Even though the error may have occurred on the other port
10000 	 * we still need to increment the vf error reference count for
10001 	 * both ports because the I/O resume function will be called
10002 	 * for both of them.
10003 	 */
10004 	adapter->vferr_refcount++;
10005 
10006 	return PCI_ERS_RESULT_RECOVERED;
10007 
10008 skip_bad_vf_detection:
10009 #endif /* CONFIG_PCI_IOV */
10010 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
10011 		return PCI_ERS_RESULT_DISCONNECT;
10012 
10013 	rtnl_lock();
10014 	netif_device_detach(netdev);
10015 
10016 	if (state == pci_channel_io_perm_failure) {
10017 		rtnl_unlock();
10018 		return PCI_ERS_RESULT_DISCONNECT;
10019 	}
10020 
10021 	if (netif_running(netdev))
10022 		ixgbe_down(adapter);
10023 
10024 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
10025 		pci_disable_device(pdev);
10026 	rtnl_unlock();
10027 
10028 	/* Request a slot reset. */
10029 	return PCI_ERS_RESULT_NEED_RESET;
10030 }
10031 
10032 /**
10033  * ixgbe_io_slot_reset - called after the pci bus has been reset.
10034  * @pdev: Pointer to PCI device
10035  *
10036  * Restart the card from scratch, as if from a cold-boot.
10037  */
10038 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
10039 {
10040 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10041 	pci_ers_result_t result;
10042 	int err;
10043 
10044 	if (pci_enable_device_mem(pdev)) {
10045 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
10046 		result = PCI_ERS_RESULT_DISCONNECT;
10047 	} else {
10048 		smp_mb__before_atomic();
10049 		clear_bit(__IXGBE_DISABLED, &adapter->state);
10050 		adapter->hw.hw_addr = adapter->io_addr;
10051 		pci_set_master(pdev);
10052 		pci_restore_state(pdev);
10053 		pci_save_state(pdev);
10054 
10055 		pci_wake_from_d3(pdev, false);
10056 
10057 		ixgbe_reset(adapter);
10058 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10059 		result = PCI_ERS_RESULT_RECOVERED;
10060 	}
10061 
10062 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10063 	if (err) {
10064 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
10065 			  "failed 0x%0x\n", err);
10066 		/* non-fatal, continue */
10067 	}
10068 
10069 	return result;
10070 }
10071 
10072 /**
10073  * ixgbe_io_resume - called when traffic can start flowing again.
10074  * @pdev: Pointer to PCI device
10075  *
10076  * This callback is called when the error recovery driver tells us that
10077  * its OK to resume normal operation.
10078  */
10079 static void ixgbe_io_resume(struct pci_dev *pdev)
10080 {
10081 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10082 	struct net_device *netdev = adapter->netdev;
10083 
10084 #ifdef CONFIG_PCI_IOV
10085 	if (adapter->vferr_refcount) {
10086 		e_info(drv, "Resuming after VF err\n");
10087 		adapter->vferr_refcount--;
10088 		return;
10089 	}
10090 
10091 #endif
10092 	if (netif_running(netdev))
10093 		ixgbe_up(adapter);
10094 
10095 	netif_device_attach(netdev);
10096 }
10097 
10098 static const struct pci_error_handlers ixgbe_err_handler = {
10099 	.error_detected = ixgbe_io_error_detected,
10100 	.slot_reset = ixgbe_io_slot_reset,
10101 	.resume = ixgbe_io_resume,
10102 };
10103 
10104 static struct pci_driver ixgbe_driver = {
10105 	.name     = ixgbe_driver_name,
10106 	.id_table = ixgbe_pci_tbl,
10107 	.probe    = ixgbe_probe,
10108 	.remove   = ixgbe_remove,
10109 #ifdef CONFIG_PM
10110 	.suspend  = ixgbe_suspend,
10111 	.resume   = ixgbe_resume,
10112 #endif
10113 	.shutdown = ixgbe_shutdown,
10114 	.sriov_configure = ixgbe_pci_sriov_configure,
10115 	.err_handler = &ixgbe_err_handler
10116 };
10117 
10118 /**
10119  * ixgbe_init_module - Driver Registration Routine
10120  *
10121  * ixgbe_init_module is the first routine called when the driver is
10122  * loaded. All it does is register with the PCI subsystem.
10123  **/
10124 static int __init ixgbe_init_module(void)
10125 {
10126 	int ret;
10127 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
10128 	pr_info("%s\n", ixgbe_copyright);
10129 
10130 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
10131 	if (!ixgbe_wq) {
10132 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
10133 		return -ENOMEM;
10134 	}
10135 
10136 	ixgbe_dbg_init();
10137 
10138 	ret = pci_register_driver(&ixgbe_driver);
10139 	if (ret) {
10140 		destroy_workqueue(ixgbe_wq);
10141 		ixgbe_dbg_exit();
10142 		return ret;
10143 	}
10144 
10145 #ifdef CONFIG_IXGBE_DCA
10146 	dca_register_notify(&dca_notifier);
10147 #endif
10148 
10149 	return 0;
10150 }
10151 
10152 module_init(ixgbe_init_module);
10153 
10154 /**
10155  * ixgbe_exit_module - Driver Exit Cleanup Routine
10156  *
10157  * ixgbe_exit_module is called just before the driver is removed
10158  * from memory.
10159  **/
10160 static void __exit ixgbe_exit_module(void)
10161 {
10162 #ifdef CONFIG_IXGBE_DCA
10163 	dca_unregister_notify(&dca_notifier);
10164 #endif
10165 	pci_unregister_driver(&ixgbe_driver);
10166 
10167 	ixgbe_dbg_exit();
10168 	if (ixgbe_wq) {
10169 		destroy_workqueue(ixgbe_wq);
10170 		ixgbe_wq = NULL;
10171 	}
10172 }
10173 
10174 #ifdef CONFIG_IXGBE_DCA
10175 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
10176 			    void *p)
10177 {
10178 	int ret_val;
10179 
10180 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
10181 					 __ixgbe_notify_dca);
10182 
10183 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
10184 }
10185 
10186 #endif /* CONFIG_IXGBE_DCA */
10187 
10188 module_exit(ixgbe_exit_module);
10189 
10190 /* ixgbe_main.c */
10191