1/*******************************************************************************
2
3  Intel 10 Gigabit PCI Express Linux driver
4  Copyright(c) 1999 - 2015 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/vxlan.h>
54
55#ifdef CONFIG_OF
56#include <linux/of_net.h>
57#endif
58
59#ifdef CONFIG_SPARC
60#include <asm/idprom.h>
61#include <asm/prom.h>
62#endif
63
64#include "ixgbe.h"
65#include "ixgbe_common.h"
66#include "ixgbe_dcb_82599.h"
67#include "ixgbe_sriov.h"
68#ifdef CONFIG_IXGBE_VXLAN
69#include <net/vxlan.h>
70#endif
71
72char ixgbe_driver_name[] = "ixgbe";
73static const char ixgbe_driver_string[] =
74			      "Intel(R) 10 Gigabit PCI Express Network Driver";
75#ifdef IXGBE_FCOE
76char ixgbe_default_device_descr[] =
77			      "Intel(R) 10 Gigabit Network Connection";
78#else
79static char ixgbe_default_device_descr[] =
80			      "Intel(R) 10 Gigabit Network Connection";
81#endif
82#define DRV_VERSION "4.2.1-k"
83const char ixgbe_driver_version[] = DRV_VERSION;
84static const char ixgbe_copyright[] =
85				"Copyright (c) 1999-2015 Intel Corporation.";
86
87static 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";
88
89static const struct ixgbe_info *ixgbe_info_tbl[] = {
90	[board_82598]		= &ixgbe_82598_info,
91	[board_82599]		= &ixgbe_82599_info,
92	[board_X540]		= &ixgbe_X540_info,
93	[board_X550]		= &ixgbe_X550_info,
94	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
95};
96
97/* ixgbe_pci_tbl - PCI Device ID Table
98 *
99 * Wildcard entries (PCI_ANY_ID) should come last
100 * Last entry must be all 0s
101 *
102 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
103 *   Class, Class Mask, private data (not used) }
104 */
105static const struct pci_device_id ixgbe_pci_tbl[] = {
106	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
107	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
108	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
109	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
110	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
111	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
112	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
113	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
114	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
115	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
116	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
117	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
118	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
119	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
120	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
121	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
122	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
123	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
124	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
125	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
126	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
127	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
128	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
129	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
130	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
131	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
132	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
133	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
134	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
135	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
136	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
137	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
138	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
139	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
140	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
141	/* required last entry */
142	{0, }
143};
144MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
145
146#ifdef CONFIG_IXGBE_DCA
147static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
148			    void *p);
149static struct notifier_block dca_notifier = {
150	.notifier_call = ixgbe_notify_dca,
151	.next          = NULL,
152	.priority      = 0
153};
154#endif
155
156#ifdef CONFIG_PCI_IOV
157static unsigned int max_vfs;
158module_param(max_vfs, uint, 0);
159MODULE_PARM_DESC(max_vfs,
160		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
161#endif /* CONFIG_PCI_IOV */
162
163static unsigned int allow_unsupported_sfp;
164module_param(allow_unsupported_sfp, uint, 0);
165MODULE_PARM_DESC(allow_unsupported_sfp,
166		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
167
168#define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
169static int debug = -1;
170module_param(debug, int, 0);
171MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
172
173MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
174MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
175MODULE_LICENSE("GPL");
176MODULE_VERSION(DRV_VERSION);
177
178static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
179
180static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
181					  u32 reg, u16 *value)
182{
183	struct pci_dev *parent_dev;
184	struct pci_bus *parent_bus;
185
186	parent_bus = adapter->pdev->bus->parent;
187	if (!parent_bus)
188		return -1;
189
190	parent_dev = parent_bus->self;
191	if (!parent_dev)
192		return -1;
193
194	if (!pci_is_pcie(parent_dev))
195		return -1;
196
197	pcie_capability_read_word(parent_dev, reg, value);
198	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
199	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
200		return -1;
201	return 0;
202}
203
204static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
205{
206	struct ixgbe_hw *hw = &adapter->hw;
207	u16 link_status = 0;
208	int err;
209
210	hw->bus.type = ixgbe_bus_type_pci_express;
211
212	/* Get the negotiated link width and speed from PCI config space of the
213	 * parent, as this device is behind a switch
214	 */
215	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
216
217	/* assume caller will handle error case */
218	if (err)
219		return err;
220
221	hw->bus.width = ixgbe_convert_bus_width(link_status);
222	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
223
224	return 0;
225}
226
227/**
228 * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
229 * @hw: hw specific details
230 *
231 * This function is used by probe to determine whether a device's PCI-Express
232 * bandwidth details should be gathered from the parent bus instead of from the
233 * device. Used to ensure that various locations all have the correct device ID
234 * checks.
235 */
236static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
237{
238	switch (hw->device_id) {
239	case IXGBE_DEV_ID_82599_SFP_SF_QP:
240	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
241		return true;
242	default:
243		return false;
244	}
245}
246
247static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
248				     int expected_gts)
249{
250	struct ixgbe_hw *hw = &adapter->hw;
251	int max_gts = 0;
252	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
253	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
254	struct pci_dev *pdev;
255
256	/* Some devices are not connected over PCIe and thus do not negotiate
257	 * speed. These devices do not have valid bus info, and thus any report
258	 * we generate may not be correct.
259	 */
260	if (hw->bus.type == ixgbe_bus_type_internal)
261		return;
262
263	/* determine whether to use the parent device */
264	if (ixgbe_pcie_from_parent(&adapter->hw))
265		pdev = adapter->pdev->bus->parent->self;
266	else
267		pdev = adapter->pdev;
268
269	if (pcie_get_minimum_link(pdev, &speed, &width) ||
270	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
271		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
272		return;
273	}
274
275	switch (speed) {
276	case PCIE_SPEED_2_5GT:
277		/* 8b/10b encoding reduces max throughput by 20% */
278		max_gts = 2 * width;
279		break;
280	case PCIE_SPEED_5_0GT:
281		/* 8b/10b encoding reduces max throughput by 20% */
282		max_gts = 4 * width;
283		break;
284	case PCIE_SPEED_8_0GT:
285		/* 128b/130b encoding reduces throughput by less than 2% */
286		max_gts = 8 * width;
287		break;
288	default:
289		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
290		return;
291	}
292
293	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
294		   max_gts);
295	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
296		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
297		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
298		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
299		    "Unknown"),
300		   width,
301		   (speed == PCIE_SPEED_2_5GT ? "20%" :
302		    speed == PCIE_SPEED_5_0GT ? "20%" :
303		    speed == PCIE_SPEED_8_0GT ? "<2%" :
304		    "Unknown"));
305
306	if (max_gts < expected_gts) {
307		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
308		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
309			expected_gts);
310		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
311	}
312}
313
314static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
315{
316	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
317	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
318	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
319		schedule_work(&adapter->service_task);
320}
321
322static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
323{
324	struct ixgbe_adapter *adapter = hw->back;
325
326	if (!hw->hw_addr)
327		return;
328	hw->hw_addr = NULL;
329	e_dev_err("Adapter removed\n");
330	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
331		ixgbe_service_event_schedule(adapter);
332}
333
334static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
335{
336	u32 value;
337
338	/* The following check not only optimizes a bit by not
339	 * performing a read on the status register when the
340	 * register just read was a status register read that
341	 * returned IXGBE_FAILED_READ_REG. It also blocks any
342	 * potential recursion.
343	 */
344	if (reg == IXGBE_STATUS) {
345		ixgbe_remove_adapter(hw);
346		return;
347	}
348	value = ixgbe_read_reg(hw, IXGBE_STATUS);
349	if (value == IXGBE_FAILED_READ_REG)
350		ixgbe_remove_adapter(hw);
351}
352
353/**
354 * ixgbe_read_reg - Read from device register
355 * @hw: hw specific details
356 * @reg: offset of register to read
357 *
358 * Returns : value read or IXGBE_FAILED_READ_REG if removed
359 *
360 * This function is used to read device registers. It checks for device
361 * removal by confirming any read that returns all ones by checking the
362 * status register value for all ones. This function avoids reading from
363 * the hardware if a removal was previously detected in which case it
364 * returns IXGBE_FAILED_READ_REG (all ones).
365 */
366u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
367{
368	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
369	u32 value;
370
371	if (ixgbe_removed(reg_addr))
372		return IXGBE_FAILED_READ_REG;
373	value = readl(reg_addr + reg);
374	if (unlikely(value == IXGBE_FAILED_READ_REG))
375		ixgbe_check_remove(hw, reg);
376	return value;
377}
378
379static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
380{
381	u16 value;
382
383	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
384	if (value == IXGBE_FAILED_READ_CFG_WORD) {
385		ixgbe_remove_adapter(hw);
386		return true;
387	}
388	return false;
389}
390
391u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
392{
393	struct ixgbe_adapter *adapter = hw->back;
394	u16 value;
395
396	if (ixgbe_removed(hw->hw_addr))
397		return IXGBE_FAILED_READ_CFG_WORD;
398	pci_read_config_word(adapter->pdev, reg, &value);
399	if (value == IXGBE_FAILED_READ_CFG_WORD &&
400	    ixgbe_check_cfg_remove(hw, adapter->pdev))
401		return IXGBE_FAILED_READ_CFG_WORD;
402	return value;
403}
404
405#ifdef CONFIG_PCI_IOV
406static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
407{
408	struct ixgbe_adapter *adapter = hw->back;
409	u32 value;
410
411	if (ixgbe_removed(hw->hw_addr))
412		return IXGBE_FAILED_READ_CFG_DWORD;
413	pci_read_config_dword(adapter->pdev, reg, &value);
414	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
415	    ixgbe_check_cfg_remove(hw, adapter->pdev))
416		return IXGBE_FAILED_READ_CFG_DWORD;
417	return value;
418}
419#endif /* CONFIG_PCI_IOV */
420
421void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
422{
423	struct ixgbe_adapter *adapter = hw->back;
424
425	if (ixgbe_removed(hw->hw_addr))
426		return;
427	pci_write_config_word(adapter->pdev, reg, value);
428}
429
430static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
431{
432	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
433
434	/* flush memory to make sure state is correct before next watchdog */
435	smp_mb__before_atomic();
436	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
437}
438
439struct ixgbe_reg_info {
440	u32 ofs;
441	char *name;
442};
443
444static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
445
446	/* General Registers */
447	{IXGBE_CTRL, "CTRL"},
448	{IXGBE_STATUS, "STATUS"},
449	{IXGBE_CTRL_EXT, "CTRL_EXT"},
450
451	/* Interrupt Registers */
452	{IXGBE_EICR, "EICR"},
453
454	/* RX Registers */
455	{IXGBE_SRRCTL(0), "SRRCTL"},
456	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
457	{IXGBE_RDLEN(0), "RDLEN"},
458	{IXGBE_RDH(0), "RDH"},
459	{IXGBE_RDT(0), "RDT"},
460	{IXGBE_RXDCTL(0), "RXDCTL"},
461	{IXGBE_RDBAL(0), "RDBAL"},
462	{IXGBE_RDBAH(0), "RDBAH"},
463
464	/* TX Registers */
465	{IXGBE_TDBAL(0), "TDBAL"},
466	{IXGBE_TDBAH(0), "TDBAH"},
467	{IXGBE_TDLEN(0), "TDLEN"},
468	{IXGBE_TDH(0), "TDH"},
469	{IXGBE_TDT(0), "TDT"},
470	{IXGBE_TXDCTL(0), "TXDCTL"},
471
472	/* List Terminator */
473	{ .name = NULL }
474};
475
476
477/*
478 * ixgbe_regdump - register printout routine
479 */
480static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
481{
482	int i = 0, j = 0;
483	char rname[16];
484	u32 regs[64];
485
486	switch (reginfo->ofs) {
487	case IXGBE_SRRCTL(0):
488		for (i = 0; i < 64; i++)
489			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
490		break;
491	case IXGBE_DCA_RXCTRL(0):
492		for (i = 0; i < 64; i++)
493			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
494		break;
495	case IXGBE_RDLEN(0):
496		for (i = 0; i < 64; i++)
497			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
498		break;
499	case IXGBE_RDH(0):
500		for (i = 0; i < 64; i++)
501			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
502		break;
503	case IXGBE_RDT(0):
504		for (i = 0; i < 64; i++)
505			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
506		break;
507	case IXGBE_RXDCTL(0):
508		for (i = 0; i < 64; i++)
509			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
510		break;
511	case IXGBE_RDBAL(0):
512		for (i = 0; i < 64; i++)
513			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
514		break;
515	case IXGBE_RDBAH(0):
516		for (i = 0; i < 64; i++)
517			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
518		break;
519	case IXGBE_TDBAL(0):
520		for (i = 0; i < 64; i++)
521			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
522		break;
523	case IXGBE_TDBAH(0):
524		for (i = 0; i < 64; i++)
525			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
526		break;
527	case IXGBE_TDLEN(0):
528		for (i = 0; i < 64; i++)
529			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
530		break;
531	case IXGBE_TDH(0):
532		for (i = 0; i < 64; i++)
533			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
534		break;
535	case IXGBE_TDT(0):
536		for (i = 0; i < 64; i++)
537			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
538		break;
539	case IXGBE_TXDCTL(0):
540		for (i = 0; i < 64; i++)
541			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
542		break;
543	default:
544		pr_info("%-15s %08x\n", reginfo->name,
545			IXGBE_READ_REG(hw, reginfo->ofs));
546		return;
547	}
548
549	for (i = 0; i < 8; i++) {
550		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
551		pr_err("%-15s", rname);
552		for (j = 0; j < 8; j++)
553			pr_cont(" %08x", regs[i*8+j]);
554		pr_cont("\n");
555	}
556
557}
558
559/*
560 * ixgbe_dump - Print registers, tx-rings and rx-rings
561 */
562static void ixgbe_dump(struct ixgbe_adapter *adapter)
563{
564	struct net_device *netdev = adapter->netdev;
565	struct ixgbe_hw *hw = &adapter->hw;
566	struct ixgbe_reg_info *reginfo;
567	int n = 0;
568	struct ixgbe_ring *tx_ring;
569	struct ixgbe_tx_buffer *tx_buffer;
570	union ixgbe_adv_tx_desc *tx_desc;
571	struct my_u0 { u64 a; u64 b; } *u0;
572	struct ixgbe_ring *rx_ring;
573	union ixgbe_adv_rx_desc *rx_desc;
574	struct ixgbe_rx_buffer *rx_buffer_info;
575	u32 staterr;
576	int i = 0;
577
578	if (!netif_msg_hw(adapter))
579		return;
580
581	/* Print netdevice Info */
582	if (netdev) {
583		dev_info(&adapter->pdev->dev, "Net device Info\n");
584		pr_info("Device Name     state            "
585			"trans_start      last_rx\n");
586		pr_info("%-15s %016lX %016lX %016lX\n",
587			netdev->name,
588			netdev->state,
589			netdev->trans_start,
590			netdev->last_rx);
591	}
592
593	/* Print Registers */
594	dev_info(&adapter->pdev->dev, "Register Dump\n");
595	pr_info(" Register Name   Value\n");
596	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
597	     reginfo->name; reginfo++) {
598		ixgbe_regdump(hw, reginfo);
599	}
600
601	/* Print TX Ring Summary */
602	if (!netdev || !netif_running(netdev))
603		return;
604
605	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
606	pr_info(" %s     %s              %s        %s\n",
607		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
608		"leng", "ntw", "timestamp");
609	for (n = 0; n < adapter->num_tx_queues; n++) {
610		tx_ring = adapter->tx_ring[n];
611		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
612		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
613			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
614			   (u64)dma_unmap_addr(tx_buffer, dma),
615			   dma_unmap_len(tx_buffer, len),
616			   tx_buffer->next_to_watch,
617			   (u64)tx_buffer->time_stamp);
618	}
619
620	/* Print TX Rings */
621	if (!netif_msg_tx_done(adapter))
622		goto rx_ring_summary;
623
624	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
625
626	/* Transmit Descriptor Formats
627	 *
628	 * 82598 Advanced Transmit Descriptor
629	 *   +--------------------------------------------------------------+
630	 * 0 |         Buffer Address [63:0]                                |
631	 *   +--------------------------------------------------------------+
632	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
633	 *   +--------------------------------------------------------------+
634	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
635	 *
636	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
637	 *   +--------------------------------------------------------------+
638	 * 0 |                          RSV [63:0]                          |
639	 *   +--------------------------------------------------------------+
640	 * 8 |            RSV           |  STA  |          NXTSEQ           |
641	 *   +--------------------------------------------------------------+
642	 *   63                       36 35   32 31                         0
643	 *
644	 * 82599+ Advanced Transmit Descriptor
645	 *   +--------------------------------------------------------------+
646	 * 0 |         Buffer Address [63:0]                                |
647	 *   +--------------------------------------------------------------+
648	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
649	 *   +--------------------------------------------------------------+
650	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
651	 *
652	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
653	 *   +--------------------------------------------------------------+
654	 * 0 |                          RSV [63:0]                          |
655	 *   +--------------------------------------------------------------+
656	 * 8 |            RSV           |  STA  |           RSV             |
657	 *   +--------------------------------------------------------------+
658	 *   63                       36 35   32 31                         0
659	 */
660
661	for (n = 0; n < adapter->num_tx_queues; n++) {
662		tx_ring = adapter->tx_ring[n];
663		pr_info("------------------------------------\n");
664		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
665		pr_info("------------------------------------\n");
666		pr_info("%s%s    %s              %s        %s          %s\n",
667			"T [desc]     [address 63:0  ] ",
668			"[PlPOIdStDDt Ln] [bi->dma       ] ",
669			"leng", "ntw", "timestamp", "bi->skb");
670
671		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
672			tx_desc = IXGBE_TX_DESC(tx_ring, i);
673			tx_buffer = &tx_ring->tx_buffer_info[i];
674			u0 = (struct my_u0 *)tx_desc;
675			if (dma_unmap_len(tx_buffer, len) > 0) {
676				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
677					i,
678					le64_to_cpu(u0->a),
679					le64_to_cpu(u0->b),
680					(u64)dma_unmap_addr(tx_buffer, dma),
681					dma_unmap_len(tx_buffer, len),
682					tx_buffer->next_to_watch,
683					(u64)tx_buffer->time_stamp,
684					tx_buffer->skb);
685				if (i == tx_ring->next_to_use &&
686					i == tx_ring->next_to_clean)
687					pr_cont(" NTC/U\n");
688				else if (i == tx_ring->next_to_use)
689					pr_cont(" NTU\n");
690				else if (i == tx_ring->next_to_clean)
691					pr_cont(" NTC\n");
692				else
693					pr_cont("\n");
694
695				if (netif_msg_pktdata(adapter) &&
696				    tx_buffer->skb)
697					print_hex_dump(KERN_INFO, "",
698						DUMP_PREFIX_ADDRESS, 16, 1,
699						tx_buffer->skb->data,
700						dma_unmap_len(tx_buffer, len),
701						true);
702			}
703		}
704	}
705
706	/* Print RX Rings Summary */
707rx_ring_summary:
708	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
709	pr_info("Queue [NTU] [NTC]\n");
710	for (n = 0; n < adapter->num_rx_queues; n++) {
711		rx_ring = adapter->rx_ring[n];
712		pr_info("%5d %5X %5X\n",
713			n, rx_ring->next_to_use, rx_ring->next_to_clean);
714	}
715
716	/* Print RX Rings */
717	if (!netif_msg_rx_status(adapter))
718		return;
719
720	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
721
722	/* Receive Descriptor Formats
723	 *
724	 * 82598 Advanced Receive Descriptor (Read) Format
725	 *    63                                           1        0
726	 *    +-----------------------------------------------------+
727	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
728	 *    +----------------------------------------------+------+
729	 *  8 |       Header Buffer Address [63:1]           |  DD  |
730	 *    +-----------------------------------------------------+
731	 *
732	 *
733	 * 82598 Advanced Receive Descriptor (Write-Back) Format
734	 *
735	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
736	 *   +------------------------------------------------------+
737	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
738	 *   | Packet   | IP     |   |          |     | Type | Type |
739	 *   | Checksum | Ident  |   |          |     |      |      |
740	 *   +------------------------------------------------------+
741	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
742	 *   +------------------------------------------------------+
743	 *   63       48 47    32 31            20 19               0
744	 *
745	 * 82599+ Advanced Receive Descriptor (Read) Format
746	 *    63                                           1        0
747	 *    +-----------------------------------------------------+
748	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
749	 *    +----------------------------------------------+------+
750	 *  8 |       Header Buffer Address [63:1]           |  DD  |
751	 *    +-----------------------------------------------------+
752	 *
753	 *
754	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
755	 *
756	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
757	 *   +------------------------------------------------------+
758	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
759	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
760	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
761	 *   +------------------------------------------------------+
762	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
763	 *   +------------------------------------------------------+
764	 *   63       48 47    32 31          20 19                 0
765	 */
766
767	for (n = 0; n < adapter->num_rx_queues; n++) {
768		rx_ring = adapter->rx_ring[n];
769		pr_info("------------------------------------\n");
770		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
771		pr_info("------------------------------------\n");
772		pr_info("%s%s%s",
773			"R  [desc]      [ PktBuf     A0] ",
774			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
775			"<-- Adv Rx Read format\n");
776		pr_info("%s%s%s",
777			"RWB[desc]      [PcsmIpSHl PtRs] ",
778			"[vl er S cks ln] ---------------- [bi->skb       ] ",
779			"<-- Adv Rx Write-Back format\n");
780
781		for (i = 0; i < rx_ring->count; i++) {
782			rx_buffer_info = &rx_ring->rx_buffer_info[i];
783			rx_desc = IXGBE_RX_DESC(rx_ring, i);
784			u0 = (struct my_u0 *)rx_desc;
785			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
786			if (staterr & IXGBE_RXD_STAT_DD) {
787				/* Descriptor Done */
788				pr_info("RWB[0x%03X]     %016llX "
789					"%016llX ---------------- %p", i,
790					le64_to_cpu(u0->a),
791					le64_to_cpu(u0->b),
792					rx_buffer_info->skb);
793			} else {
794				pr_info("R  [0x%03X]     %016llX "
795					"%016llX %016llX %p", i,
796					le64_to_cpu(u0->a),
797					le64_to_cpu(u0->b),
798					(u64)rx_buffer_info->dma,
799					rx_buffer_info->skb);
800
801				if (netif_msg_pktdata(adapter) &&
802				    rx_buffer_info->dma) {
803					print_hex_dump(KERN_INFO, "",
804					   DUMP_PREFIX_ADDRESS, 16, 1,
805					   page_address(rx_buffer_info->page) +
806						    rx_buffer_info->page_offset,
807					   ixgbe_rx_bufsz(rx_ring), true);
808				}
809			}
810
811			if (i == rx_ring->next_to_use)
812				pr_cont(" NTU\n");
813			else if (i == rx_ring->next_to_clean)
814				pr_cont(" NTC\n");
815			else
816				pr_cont("\n");
817
818		}
819	}
820}
821
822static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
823{
824	u32 ctrl_ext;
825
826	/* Let firmware take over control of h/w */
827	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
828	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
829			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
830}
831
832static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
833{
834	u32 ctrl_ext;
835
836	/* Let firmware know the driver has taken over */
837	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
838	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
839			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
840}
841
842/**
843 * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
844 * @adapter: pointer to adapter struct
845 * @direction: 0 for Rx, 1 for Tx, -1 for other causes
846 * @queue: queue to map the corresponding interrupt to
847 * @msix_vector: the vector to map to the corresponding queue
848 *
849 */
850static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
851			   u8 queue, u8 msix_vector)
852{
853	u32 ivar, index;
854	struct ixgbe_hw *hw = &adapter->hw;
855	switch (hw->mac.type) {
856	case ixgbe_mac_82598EB:
857		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
858		if (direction == -1)
859			direction = 0;
860		index = (((direction * 64) + queue) >> 2) & 0x1F;
861		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
862		ivar &= ~(0xFF << (8 * (queue & 0x3)));
863		ivar |= (msix_vector << (8 * (queue & 0x3)));
864		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
865		break;
866	case ixgbe_mac_82599EB:
867	case ixgbe_mac_X540:
868	case ixgbe_mac_X550:
869	case ixgbe_mac_X550EM_x:
870		if (direction == -1) {
871			/* other causes */
872			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
873			index = ((queue & 1) * 8);
874			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
875			ivar &= ~(0xFF << index);
876			ivar |= (msix_vector << index);
877			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
878			break;
879		} else {
880			/* tx or rx causes */
881			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
882			index = ((16 * (queue & 1)) + (8 * direction));
883			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
884			ivar &= ~(0xFF << index);
885			ivar |= (msix_vector << index);
886			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
887			break;
888		}
889	default:
890		break;
891	}
892}
893
894static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
895					  u64 qmask)
896{
897	u32 mask;
898
899	switch (adapter->hw.mac.type) {
900	case ixgbe_mac_82598EB:
901		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
902		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
903		break;
904	case ixgbe_mac_82599EB:
905	case ixgbe_mac_X540:
906	case ixgbe_mac_X550:
907	case ixgbe_mac_X550EM_x:
908		mask = (qmask & 0xFFFFFFFF);
909		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
910		mask = (qmask >> 32);
911		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
912		break;
913	default:
914		break;
915	}
916}
917
918void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
919				      struct ixgbe_tx_buffer *tx_buffer)
920{
921	if (tx_buffer->skb) {
922		dev_kfree_skb_any(tx_buffer->skb);
923		if (dma_unmap_len(tx_buffer, len))
924			dma_unmap_single(ring->dev,
925					 dma_unmap_addr(tx_buffer, dma),
926					 dma_unmap_len(tx_buffer, len),
927					 DMA_TO_DEVICE);
928	} else if (dma_unmap_len(tx_buffer, len)) {
929		dma_unmap_page(ring->dev,
930			       dma_unmap_addr(tx_buffer, dma),
931			       dma_unmap_len(tx_buffer, len),
932			       DMA_TO_DEVICE);
933	}
934	tx_buffer->next_to_watch = NULL;
935	tx_buffer->skb = NULL;
936	dma_unmap_len_set(tx_buffer, len, 0);
937	/* tx_buffer must be completely set up in the transmit path */
938}
939
940static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
941{
942	struct ixgbe_hw *hw = &adapter->hw;
943	struct ixgbe_hw_stats *hwstats = &adapter->stats;
944	int i;
945	u32 data;
946
947	if ((hw->fc.current_mode != ixgbe_fc_full) &&
948	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
949		return;
950
951	switch (hw->mac.type) {
952	case ixgbe_mac_82598EB:
953		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
954		break;
955	default:
956		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
957	}
958	hwstats->lxoffrxc += data;
959
960	/* refill credits (no tx hang) if we received xoff */
961	if (!data)
962		return;
963
964	for (i = 0; i < adapter->num_tx_queues; i++)
965		clear_bit(__IXGBE_HANG_CHECK_ARMED,
966			  &adapter->tx_ring[i]->state);
967}
968
969static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
970{
971	struct ixgbe_hw *hw = &adapter->hw;
972	struct ixgbe_hw_stats *hwstats = &adapter->stats;
973	u32 xoff[8] = {0};
974	u8 tc;
975	int i;
976	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
977
978	if (adapter->ixgbe_ieee_pfc)
979		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
980
981	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
982		ixgbe_update_xoff_rx_lfc(adapter);
983		return;
984	}
985
986	/* update stats for each tc, only valid with PFC enabled */
987	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
988		u32 pxoffrxc;
989
990		switch (hw->mac.type) {
991		case ixgbe_mac_82598EB:
992			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
993			break;
994		default:
995			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
996		}
997		hwstats->pxoffrxc[i] += pxoffrxc;
998		/* Get the TC for given UP */
999		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1000		xoff[tc] += pxoffrxc;
1001	}
1002
1003	/* disarm tx queues that have received xoff frames */
1004	for (i = 0; i < adapter->num_tx_queues; i++) {
1005		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1006
1007		tc = tx_ring->dcb_tc;
1008		if (xoff[tc])
1009			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1010	}
1011}
1012
1013static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1014{
1015	return ring->stats.packets;
1016}
1017
1018static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1019{
1020	struct ixgbe_adapter *adapter;
1021	struct ixgbe_hw *hw;
1022	u32 head, tail;
1023
1024	if (ring->l2_accel_priv)
1025		adapter = ring->l2_accel_priv->real_adapter;
1026	else
1027		adapter = netdev_priv(ring->netdev);
1028
1029	hw = &adapter->hw;
1030	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1031	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1032
1033	if (head != tail)
1034		return (head < tail) ?
1035			tail - head : (tail + ring->count - head);
1036
1037	return 0;
1038}
1039
1040static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1041{
1042	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1043	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1044	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1045
1046	clear_check_for_tx_hang(tx_ring);
1047
1048	/*
1049	 * Check for a hung queue, but be thorough. This verifies
1050	 * that a transmit has been completed since the previous
1051	 * check AND there is at least one packet pending. The
1052	 * ARMED bit is set to indicate a potential hang. The
1053	 * bit is cleared if a pause frame is received to remove
1054	 * false hang detection due to PFC or 802.3x frames. By
1055	 * requiring this to fail twice we avoid races with
1056	 * pfc clearing the ARMED bit and conditions where we
1057	 * run the check_tx_hang logic with a transmit completion
1058	 * pending but without time to complete it yet.
1059	 */
1060	if (tx_done_old == tx_done && tx_pending)
1061		/* make sure it is true for two checks in a row */
1062		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1063					&tx_ring->state);
1064	/* update completed stats and continue */
1065	tx_ring->tx_stats.tx_done_old = tx_done;
1066	/* reset the countdown */
1067	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1068
1069	return false;
1070}
1071
1072/**
1073 * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1074 * @adapter: driver private struct
1075 **/
1076static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1077{
1078
1079	/* Do the reset outside of interrupt context */
1080	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1081		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
1082		e_warn(drv, "initiating reset due to tx timeout\n");
1083		ixgbe_service_event_schedule(adapter);
1084	}
1085}
1086
1087/**
1088 * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1089 * @q_vector: structure containing interrupt and ring information
1090 * @tx_ring: tx ring to clean
1091 **/
1092static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1093			       struct ixgbe_ring *tx_ring)
1094{
1095	struct ixgbe_adapter *adapter = q_vector->adapter;
1096	struct ixgbe_tx_buffer *tx_buffer;
1097	union ixgbe_adv_tx_desc *tx_desc;
1098	unsigned int total_bytes = 0, total_packets = 0;
1099	unsigned int budget = q_vector->tx.work_limit;
1100	unsigned int i = tx_ring->next_to_clean;
1101
1102	if (test_bit(__IXGBE_DOWN, &adapter->state))
1103		return true;
1104
1105	tx_buffer = &tx_ring->tx_buffer_info[i];
1106	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1107	i -= tx_ring->count;
1108
1109	do {
1110		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1111
1112		/* if next_to_watch is not set then there is no work pending */
1113		if (!eop_desc)
1114			break;
1115
1116		/* prevent any other reads prior to eop_desc */
1117		read_barrier_depends();
1118
1119		/* if DD is not set pending work has not been completed */
1120		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1121			break;
1122
1123		/* clear next_to_watch to prevent false hangs */
1124		tx_buffer->next_to_watch = NULL;
1125
1126		/* update the statistics for this packet */
1127		total_bytes += tx_buffer->bytecount;
1128		total_packets += tx_buffer->gso_segs;
1129
1130		/* free the skb */
1131		dev_consume_skb_any(tx_buffer->skb);
1132
1133		/* unmap skb header data */
1134		dma_unmap_single(tx_ring->dev,
1135				 dma_unmap_addr(tx_buffer, dma),
1136				 dma_unmap_len(tx_buffer, len),
1137				 DMA_TO_DEVICE);
1138
1139		/* clear tx_buffer data */
1140		tx_buffer->skb = NULL;
1141		dma_unmap_len_set(tx_buffer, len, 0);
1142
1143		/* unmap remaining buffers */
1144		while (tx_desc != eop_desc) {
1145			tx_buffer++;
1146			tx_desc++;
1147			i++;
1148			if (unlikely(!i)) {
1149				i -= tx_ring->count;
1150				tx_buffer = tx_ring->tx_buffer_info;
1151				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1152			}
1153
1154			/* unmap any remaining paged data */
1155			if (dma_unmap_len(tx_buffer, len)) {
1156				dma_unmap_page(tx_ring->dev,
1157					       dma_unmap_addr(tx_buffer, dma),
1158					       dma_unmap_len(tx_buffer, len),
1159					       DMA_TO_DEVICE);
1160				dma_unmap_len_set(tx_buffer, len, 0);
1161			}
1162		}
1163
1164		/* move us one more past the eop_desc for start of next pkt */
1165		tx_buffer++;
1166		tx_desc++;
1167		i++;
1168		if (unlikely(!i)) {
1169			i -= tx_ring->count;
1170			tx_buffer = tx_ring->tx_buffer_info;
1171			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1172		}
1173
1174		/* issue prefetch for next Tx descriptor */
1175		prefetch(tx_desc);
1176
1177		/* update budget accounting */
1178		budget--;
1179	} while (likely(budget));
1180
1181	i += tx_ring->count;
1182	tx_ring->next_to_clean = i;
1183	u64_stats_update_begin(&tx_ring->syncp);
1184	tx_ring->stats.bytes += total_bytes;
1185	tx_ring->stats.packets += total_packets;
1186	u64_stats_update_end(&tx_ring->syncp);
1187	q_vector->tx.total_bytes += total_bytes;
1188	q_vector->tx.total_packets += total_packets;
1189
1190	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1191		/* schedule immediate reset if we believe we hung */
1192		struct ixgbe_hw *hw = &adapter->hw;
1193		e_err(drv, "Detected Tx Unit Hang\n"
1194			"  Tx Queue             <%d>\n"
1195			"  TDH, TDT             <%x>, <%x>\n"
1196			"  next_to_use          <%x>\n"
1197			"  next_to_clean        <%x>\n"
1198			"tx_buffer_info[next_to_clean]\n"
1199			"  time_stamp           <%lx>\n"
1200			"  jiffies              <%lx>\n",
1201			tx_ring->queue_index,
1202			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1203			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1204			tx_ring->next_to_use, i,
1205			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1206
1207		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1208
1209		e_info(probe,
1210		       "tx hang %d detected on queue %d, resetting adapter\n",
1211			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1212
1213		/* schedule immediate reset if we believe we hung */
1214		ixgbe_tx_timeout_reset(adapter);
1215
1216		/* the adapter is about to reset, no point in enabling stuff */
1217		return true;
1218	}
1219
1220	netdev_tx_completed_queue(txring_txq(tx_ring),
1221				  total_packets, total_bytes);
1222
1223#define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1224	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1225		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1226		/* Make sure that anybody stopping the queue after this
1227		 * sees the new next_to_clean.
1228		 */
1229		smp_mb();
1230		if (__netif_subqueue_stopped(tx_ring->netdev,
1231					     tx_ring->queue_index)
1232		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1233			netif_wake_subqueue(tx_ring->netdev,
1234					    tx_ring->queue_index);
1235			++tx_ring->tx_stats.restart_queue;
1236		}
1237	}
1238
1239	return !!budget;
1240}
1241
1242#ifdef CONFIG_IXGBE_DCA
1243static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1244				struct ixgbe_ring *tx_ring,
1245				int cpu)
1246{
1247	struct ixgbe_hw *hw = &adapter->hw;
1248	u32 txctrl = 0;
1249	u16 reg_offset;
1250
1251	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1252		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1253
1254	switch (hw->mac.type) {
1255	case ixgbe_mac_82598EB:
1256		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1257		break;
1258	case ixgbe_mac_82599EB:
1259	case ixgbe_mac_X540:
1260		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1261		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1262		break;
1263	default:
1264		/* for unknown hardware do not write register */
1265		return;
1266	}
1267
1268	/*
1269	 * We can enable relaxed ordering for reads, but not writes when
1270	 * DCA is enabled.  This is due to a known issue in some chipsets
1271	 * which will cause the DCA tag to be cleared.
1272	 */
1273	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1274		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1275		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1276
1277	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1278}
1279
1280static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1281				struct ixgbe_ring *rx_ring,
1282				int cpu)
1283{
1284	struct ixgbe_hw *hw = &adapter->hw;
1285	u32 rxctrl = 0;
1286	u8 reg_idx = rx_ring->reg_idx;
1287
1288	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1289		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1290
1291	switch (hw->mac.type) {
1292	case ixgbe_mac_82599EB:
1293	case ixgbe_mac_X540:
1294		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1295		break;
1296	default:
1297		break;
1298	}
1299
1300	/*
1301	 * We can enable relaxed ordering for reads, but not writes when
1302	 * DCA is enabled.  This is due to a known issue in some chipsets
1303	 * which will cause the DCA tag to be cleared.
1304	 */
1305	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1306		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1307		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1308
1309	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1310}
1311
1312static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1313{
1314	struct ixgbe_adapter *adapter = q_vector->adapter;
1315	struct ixgbe_ring *ring;
1316	int cpu = get_cpu();
1317
1318	if (q_vector->cpu == cpu)
1319		goto out_no_update;
1320
1321	ixgbe_for_each_ring(ring, q_vector->tx)
1322		ixgbe_update_tx_dca(adapter, ring, cpu);
1323
1324	ixgbe_for_each_ring(ring, q_vector->rx)
1325		ixgbe_update_rx_dca(adapter, ring, cpu);
1326
1327	q_vector->cpu = cpu;
1328out_no_update:
1329	put_cpu();
1330}
1331
1332static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1333{
1334	int i;
1335
1336	/* always use CB2 mode, difference is masked in the CB driver */
1337	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1338		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1339				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1340	else
1341		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1342				IXGBE_DCA_CTRL_DCA_DISABLE);
1343
1344	for (i = 0; i < adapter->num_q_vectors; i++) {
1345		adapter->q_vector[i]->cpu = -1;
1346		ixgbe_update_dca(adapter->q_vector[i]);
1347	}
1348}
1349
1350static int __ixgbe_notify_dca(struct device *dev, void *data)
1351{
1352	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1353	unsigned long event = *(unsigned long *)data;
1354
1355	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1356		return 0;
1357
1358	switch (event) {
1359	case DCA_PROVIDER_ADD:
1360		/* if we're already enabled, don't do it again */
1361		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1362			break;
1363		if (dca_add_requester(dev) == 0) {
1364			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1365			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1366					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1367			break;
1368		}
1369		/* Fall Through since DCA is disabled. */
1370	case DCA_PROVIDER_REMOVE:
1371		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1372			dca_remove_requester(dev);
1373			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1374			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1375					IXGBE_DCA_CTRL_DCA_DISABLE);
1376		}
1377		break;
1378	}
1379
1380	return 0;
1381}
1382
1383#endif /* CONFIG_IXGBE_DCA */
1384
1385#define IXGBE_RSS_L4_TYPES_MASK \
1386	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1387	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1388	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1389	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1390
1391static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1392				 union ixgbe_adv_rx_desc *rx_desc,
1393				 struct sk_buff *skb)
1394{
1395	u16 rss_type;
1396
1397	if (!(ring->netdev->features & NETIF_F_RXHASH))
1398		return;
1399
1400	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1401		   IXGBE_RXDADV_RSSTYPE_MASK;
1402
1403	if (!rss_type)
1404		return;
1405
1406	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1407		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1408		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1409}
1410
1411#ifdef IXGBE_FCOE
1412/**
1413 * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1414 * @ring: structure containing ring specific data
1415 * @rx_desc: advanced rx descriptor
1416 *
1417 * Returns : true if it is FCoE pkt
1418 */
1419static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1420				    union ixgbe_adv_rx_desc *rx_desc)
1421{
1422	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1423
1424	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1425	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1426		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1427			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1428}
1429
1430#endif /* IXGBE_FCOE */
1431/**
1432 * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1433 * @ring: structure containing ring specific data
1434 * @rx_desc: current Rx descriptor being processed
1435 * @skb: skb currently being received and modified
1436 **/
1437static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1438				     union ixgbe_adv_rx_desc *rx_desc,
1439				     struct sk_buff *skb)
1440{
1441	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1442	__le16 hdr_info = rx_desc->wb.lower.lo_dword.hs_rss.hdr_info;
1443	bool encap_pkt = false;
1444
1445	skb_checksum_none_assert(skb);
1446
1447	/* Rx csum disabled */
1448	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1449		return;
1450
1451	if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) &&
1452	    (hdr_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_TUNNEL >> 16))) {
1453		encap_pkt = true;
1454		skb->encapsulation = 1;
1455	}
1456
1457	/* if IP and error */
1458	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1459	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1460		ring->rx_stats.csum_err++;
1461		return;
1462	}
1463
1464	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1465		return;
1466
1467	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1468		/*
1469		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1470		 * checksum errors.
1471		 */
1472		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1473		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1474			return;
1475
1476		ring->rx_stats.csum_err++;
1477		return;
1478	}
1479
1480	/* It must be a TCP or UDP packet with a valid checksum */
1481	skb->ip_summed = CHECKSUM_UNNECESSARY;
1482	if (encap_pkt) {
1483		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1484			return;
1485
1486		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1487			ring->rx_stats.csum_err++;
1488			return;
1489		}
1490		/* If we checked the outer header let the stack know */
1491		skb->csum_level = 1;
1492	}
1493}
1494
1495static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1496				    struct ixgbe_rx_buffer *bi)
1497{
1498	struct page *page = bi->page;
1499	dma_addr_t dma;
1500
1501	/* since we are recycling buffers we should seldom need to alloc */
1502	if (likely(page))
1503		return true;
1504
1505	/* alloc new page for storage */
1506	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1507	if (unlikely(!page)) {
1508		rx_ring->rx_stats.alloc_rx_page_failed++;
1509		return false;
1510	}
1511
1512	/* map page for use */
1513	dma = dma_map_page(rx_ring->dev, page, 0,
1514			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1515
1516	/*
1517	 * if mapping failed free memory back to system since
1518	 * there isn't much point in holding memory we can't use
1519	 */
1520	if (dma_mapping_error(rx_ring->dev, dma)) {
1521		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1522
1523		rx_ring->rx_stats.alloc_rx_page_failed++;
1524		return false;
1525	}
1526
1527	bi->dma = dma;
1528	bi->page = page;
1529	bi->page_offset = 0;
1530
1531	return true;
1532}
1533
1534/**
1535 * ixgbe_alloc_rx_buffers - Replace used receive buffers
1536 * @rx_ring: ring to place buffers on
1537 * @cleaned_count: number of buffers to replace
1538 **/
1539void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1540{
1541	union ixgbe_adv_rx_desc *rx_desc;
1542	struct ixgbe_rx_buffer *bi;
1543	u16 i = rx_ring->next_to_use;
1544
1545	/* nothing to do */
1546	if (!cleaned_count)
1547		return;
1548
1549	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1550	bi = &rx_ring->rx_buffer_info[i];
1551	i -= rx_ring->count;
1552
1553	do {
1554		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1555			break;
1556
1557		/*
1558		 * Refresh the desc even if buffer_addrs didn't change
1559		 * because each write-back erases this info.
1560		 */
1561		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1562
1563		rx_desc++;
1564		bi++;
1565		i++;
1566		if (unlikely(!i)) {
1567			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1568			bi = rx_ring->rx_buffer_info;
1569			i -= rx_ring->count;
1570		}
1571
1572		/* clear the status bits for the next_to_use descriptor */
1573		rx_desc->wb.upper.status_error = 0;
1574
1575		cleaned_count--;
1576	} while (cleaned_count);
1577
1578	i += rx_ring->count;
1579
1580	if (rx_ring->next_to_use != i) {
1581		rx_ring->next_to_use = i;
1582
1583		/* update next to alloc since we have filled the ring */
1584		rx_ring->next_to_alloc = i;
1585
1586		/* Force memory writes to complete before letting h/w
1587		 * know there are new descriptors to fetch.  (Only
1588		 * applicable for weak-ordered memory model archs,
1589		 * such as IA-64).
1590		 */
1591		wmb();
1592		writel(i, rx_ring->tail);
1593	}
1594}
1595
1596static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1597				   struct sk_buff *skb)
1598{
1599	u16 hdr_len = skb_headlen(skb);
1600
1601	/* set gso_size to avoid messing up TCP MSS */
1602	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1603						 IXGBE_CB(skb)->append_cnt);
1604	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1605}
1606
1607static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1608				   struct sk_buff *skb)
1609{
1610	/* if append_cnt is 0 then frame is not RSC */
1611	if (!IXGBE_CB(skb)->append_cnt)
1612		return;
1613
1614	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1615	rx_ring->rx_stats.rsc_flush++;
1616
1617	ixgbe_set_rsc_gso_size(rx_ring, skb);
1618
1619	/* gso_size is computed using append_cnt so always clear it last */
1620	IXGBE_CB(skb)->append_cnt = 0;
1621}
1622
1623/**
1624 * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1625 * @rx_ring: rx descriptor ring packet is being transacted on
1626 * @rx_desc: pointer to the EOP Rx descriptor
1627 * @skb: pointer to current skb being populated
1628 *
1629 * This function checks the ring, descriptor, and packet information in
1630 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1631 * other fields within the skb.
1632 **/
1633static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1634				     union ixgbe_adv_rx_desc *rx_desc,
1635				     struct sk_buff *skb)
1636{
1637	struct net_device *dev = rx_ring->netdev;
1638
1639	ixgbe_update_rsc_stats(rx_ring, skb);
1640
1641	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1642
1643	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1644
1645	if (unlikely(ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_TS)))
1646		ixgbe_ptp_rx_hwtstamp(rx_ring->q_vector->adapter, skb);
1647
1648	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1649	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1650		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1651		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1652	}
1653
1654	skb_record_rx_queue(skb, rx_ring->queue_index);
1655
1656	skb->protocol = eth_type_trans(skb, dev);
1657}
1658
1659static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1660			 struct sk_buff *skb)
1661{
1662	if (ixgbe_qv_busy_polling(q_vector))
1663		netif_receive_skb(skb);
1664	else
1665		napi_gro_receive(&q_vector->napi, skb);
1666}
1667
1668/**
1669 * ixgbe_is_non_eop - process handling of non-EOP buffers
1670 * @rx_ring: Rx ring being processed
1671 * @rx_desc: Rx descriptor for current buffer
1672 * @skb: Current socket buffer containing buffer in progress
1673 *
1674 * This function updates next to clean.  If the buffer is an EOP buffer
1675 * this function exits returning false, otherwise it will place the
1676 * sk_buff in the next buffer to be chained and return true indicating
1677 * that this is in fact a non-EOP buffer.
1678 **/
1679static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1680			     union ixgbe_adv_rx_desc *rx_desc,
1681			     struct sk_buff *skb)
1682{
1683	u32 ntc = rx_ring->next_to_clean + 1;
1684
1685	/* fetch, update, and store next to clean */
1686	ntc = (ntc < rx_ring->count) ? ntc : 0;
1687	rx_ring->next_to_clean = ntc;
1688
1689	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1690
1691	/* update RSC append count if present */
1692	if (ring_is_rsc_enabled(rx_ring)) {
1693		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1694				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1695
1696		if (unlikely(rsc_enabled)) {
1697			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1698
1699			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1700			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1701
1702			/* update ntc based on RSC value */
1703			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1704			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1705			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1706		}
1707	}
1708
1709	/* if we are the last buffer then there is nothing else to do */
1710	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1711		return false;
1712
1713	/* place skb in next buffer to be received */
1714	rx_ring->rx_buffer_info[ntc].skb = skb;
1715	rx_ring->rx_stats.non_eop_descs++;
1716
1717	return true;
1718}
1719
1720/**
1721 * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1722 * @rx_ring: rx descriptor ring packet is being transacted on
1723 * @skb: pointer to current skb being adjusted
1724 *
1725 * This function is an ixgbe specific version of __pskb_pull_tail.  The
1726 * main difference between this version and the original function is that
1727 * this function can make several assumptions about the state of things
1728 * that allow for significant optimizations versus the standard function.
1729 * As a result we can do things like drop a frag and maintain an accurate
1730 * truesize for the skb.
1731 */
1732static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1733			    struct sk_buff *skb)
1734{
1735	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1736	unsigned char *va;
1737	unsigned int pull_len;
1738
1739	/*
1740	 * it is valid to use page_address instead of kmap since we are
1741	 * working with pages allocated out of the lomem pool per
1742	 * alloc_page(GFP_ATOMIC)
1743	 */
1744	va = skb_frag_address(frag);
1745
1746	/*
1747	 * we need the header to contain the greater of either ETH_HLEN or
1748	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1749	 */
1750	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1751
1752	/* align pull length to size of long to optimize memcpy performance */
1753	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1754
1755	/* update all of the pointers */
1756	skb_frag_size_sub(frag, pull_len);
1757	frag->page_offset += pull_len;
1758	skb->data_len -= pull_len;
1759	skb->tail += pull_len;
1760}
1761
1762/**
1763 * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1764 * @rx_ring: rx descriptor ring packet is being transacted on
1765 * @skb: pointer to current skb being updated
1766 *
1767 * This function provides a basic DMA sync up for the first fragment of an
1768 * skb.  The reason for doing this is that the first fragment cannot be
1769 * unmapped until we have reached the end of packet descriptor for a buffer
1770 * chain.
1771 */
1772static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1773				struct sk_buff *skb)
1774{
1775	/* if the page was released unmap it, else just sync our portion */
1776	if (unlikely(IXGBE_CB(skb)->page_released)) {
1777		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1778			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1779		IXGBE_CB(skb)->page_released = false;
1780	} else {
1781		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1782
1783		dma_sync_single_range_for_cpu(rx_ring->dev,
1784					      IXGBE_CB(skb)->dma,
1785					      frag->page_offset,
1786					      ixgbe_rx_bufsz(rx_ring),
1787					      DMA_FROM_DEVICE);
1788	}
1789	IXGBE_CB(skb)->dma = 0;
1790}
1791
1792/**
1793 * ixgbe_cleanup_headers - Correct corrupted or empty headers
1794 * @rx_ring: rx descriptor ring packet is being transacted on
1795 * @rx_desc: pointer to the EOP Rx descriptor
1796 * @skb: pointer to current skb being fixed
1797 *
1798 * Check for corrupted packet headers caused by senders on the local L2
1799 * embedded NIC switch not setting up their Tx Descriptors right.  These
1800 * should be very rare.
1801 *
1802 * Also address the case where we are pulling data in on pages only
1803 * and as such no data is present in the skb header.
1804 *
1805 * In addition if skb is not at least 60 bytes we need to pad it so that
1806 * it is large enough to qualify as a valid Ethernet frame.
1807 *
1808 * Returns true if an error was encountered and skb was freed.
1809 **/
1810static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1811				  union ixgbe_adv_rx_desc *rx_desc,
1812				  struct sk_buff *skb)
1813{
1814	struct net_device *netdev = rx_ring->netdev;
1815
1816	/* verify that the packet does not have any known errors */
1817	if (unlikely(ixgbe_test_staterr(rx_desc,
1818					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1819	    !(netdev->features & NETIF_F_RXALL))) {
1820		dev_kfree_skb_any(skb);
1821		return true;
1822	}
1823
1824	/* place header in linear portion of buffer */
1825	if (skb_is_nonlinear(skb))
1826		ixgbe_pull_tail(rx_ring, skb);
1827
1828#ifdef IXGBE_FCOE
1829	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1830	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1831		return false;
1832
1833#endif
1834	/* if eth_skb_pad returns an error the skb was freed */
1835	if (eth_skb_pad(skb))
1836		return true;
1837
1838	return false;
1839}
1840
1841/**
1842 * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1843 * @rx_ring: rx descriptor ring to store buffers on
1844 * @old_buff: donor buffer to have page reused
1845 *
1846 * Synchronizes page for reuse by the adapter
1847 **/
1848static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1849				struct ixgbe_rx_buffer *old_buff)
1850{
1851	struct ixgbe_rx_buffer *new_buff;
1852	u16 nta = rx_ring->next_to_alloc;
1853
1854	new_buff = &rx_ring->rx_buffer_info[nta];
1855
1856	/* update, and store next to alloc */
1857	nta++;
1858	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1859
1860	/* transfer page from old buffer to new buffer */
1861	*new_buff = *old_buff;
1862
1863	/* sync the buffer for use by the device */
1864	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1865					 new_buff->page_offset,
1866					 ixgbe_rx_bufsz(rx_ring),
1867					 DMA_FROM_DEVICE);
1868}
1869
1870static inline bool ixgbe_page_is_reserved(struct page *page)
1871{
1872	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
1873}
1874
1875/**
1876 * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1877 * @rx_ring: rx descriptor ring to transact packets on
1878 * @rx_buffer: buffer containing page to add
1879 * @rx_desc: descriptor containing length of buffer written by hardware
1880 * @skb: sk_buff to place the data into
1881 *
1882 * This function will add the data contained in rx_buffer->page to the skb.
1883 * This is done either through a direct copy if the data in the buffer is
1884 * less than the skb header size, otherwise it will just attach the page as
1885 * a frag to the skb.
1886 *
1887 * The function will then update the page offset if necessary and return
1888 * true if the buffer can be reused by the adapter.
1889 **/
1890static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1891			      struct ixgbe_rx_buffer *rx_buffer,
1892			      union ixgbe_adv_rx_desc *rx_desc,
1893			      struct sk_buff *skb)
1894{
1895	struct page *page = rx_buffer->page;
1896	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1897#if (PAGE_SIZE < 8192)
1898	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1899#else
1900	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1901	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1902				   ixgbe_rx_bufsz(rx_ring);
1903#endif
1904
1905	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1906		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1907
1908		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1909
1910		/* page is not reserved, we can reuse buffer as-is */
1911		if (likely(!ixgbe_page_is_reserved(page)))
1912			return true;
1913
1914		/* this page cannot be reused so discard it */
1915		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1916		return false;
1917	}
1918
1919	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1920			rx_buffer->page_offset, size, truesize);
1921
1922	/* avoid re-using remote pages */
1923	if (unlikely(ixgbe_page_is_reserved(page)))
1924		return false;
1925
1926#if (PAGE_SIZE < 8192)
1927	/* if we are only owner of page we can reuse it */
1928	if (unlikely(page_count(page) != 1))
1929		return false;
1930
1931	/* flip page offset to other buffer */
1932	rx_buffer->page_offset ^= truesize;
1933#else
1934	/* move offset up to the next cache line */
1935	rx_buffer->page_offset += truesize;
1936
1937	if (rx_buffer->page_offset > last_offset)
1938		return false;
1939#endif
1940
1941	/* Even if we own the page, we are not allowed to use atomic_set()
1942	 * This would break get_page_unless_zero() users.
1943	 */
1944	atomic_inc(&page->_count);
1945
1946	return true;
1947}
1948
1949static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
1950					     union ixgbe_adv_rx_desc *rx_desc)
1951{
1952	struct ixgbe_rx_buffer *rx_buffer;
1953	struct sk_buff *skb;
1954	struct page *page;
1955
1956	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
1957	page = rx_buffer->page;
1958	prefetchw(page);
1959
1960	skb = rx_buffer->skb;
1961
1962	if (likely(!skb)) {
1963		void *page_addr = page_address(page) +
1964				  rx_buffer->page_offset;
1965
1966		/* prefetch first cache line of first page */
1967		prefetch(page_addr);
1968#if L1_CACHE_BYTES < 128
1969		prefetch(page_addr + L1_CACHE_BYTES);
1970#endif
1971
1972		/* allocate a skb to store the frags */
1973		skb = napi_alloc_skb(&rx_ring->q_vector->napi,
1974				     IXGBE_RX_HDR_SIZE);
1975		if (unlikely(!skb)) {
1976			rx_ring->rx_stats.alloc_rx_buff_failed++;
1977			return NULL;
1978		}
1979
1980		/*
1981		 * we will be copying header into skb->data in
1982		 * pskb_may_pull so it is in our interest to prefetch
1983		 * it now to avoid a possible cache miss
1984		 */
1985		prefetchw(skb->data);
1986
1987		/*
1988		 * Delay unmapping of the first packet. It carries the
1989		 * header information, HW may still access the header
1990		 * after the writeback.  Only unmap it when EOP is
1991		 * reached
1992		 */
1993		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1994			goto dma_sync;
1995
1996		IXGBE_CB(skb)->dma = rx_buffer->dma;
1997	} else {
1998		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
1999			ixgbe_dma_sync_frag(rx_ring, skb);
2000
2001dma_sync:
2002		/* we are reusing so sync this buffer for CPU use */
2003		dma_sync_single_range_for_cpu(rx_ring->dev,
2004					      rx_buffer->dma,
2005					      rx_buffer->page_offset,
2006					      ixgbe_rx_bufsz(rx_ring),
2007					      DMA_FROM_DEVICE);
2008
2009		rx_buffer->skb = NULL;
2010	}
2011
2012	/* pull page into skb */
2013	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
2014		/* hand second half of page back to the ring */
2015		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2016	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
2017		/* the page has been released from the ring */
2018		IXGBE_CB(skb)->page_released = true;
2019	} else {
2020		/* we are not reusing the buffer so unmap it */
2021		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
2022			       ixgbe_rx_pg_size(rx_ring),
2023			       DMA_FROM_DEVICE);
2024	}
2025
2026	/* clear contents of buffer_info */
2027	rx_buffer->page = NULL;
2028
2029	return skb;
2030}
2031
2032/**
2033 * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2034 * @q_vector: structure containing interrupt and ring information
2035 * @rx_ring: rx descriptor ring to transact packets on
2036 * @budget: Total limit on number of packets to process
2037 *
2038 * This function provides a "bounce buffer" approach to Rx interrupt
2039 * processing.  The advantage to this is that on systems that have
2040 * expensive overhead for IOMMU access this provides a means of avoiding
2041 * it by maintaining the mapping of the page to the syste.
2042 *
2043 * Returns amount of work completed
2044 **/
2045static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2046			       struct ixgbe_ring *rx_ring,
2047			       const int budget)
2048{
2049	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2050#ifdef IXGBE_FCOE
2051	struct ixgbe_adapter *adapter = q_vector->adapter;
2052	int ddp_bytes;
2053	unsigned int mss = 0;
2054#endif /* IXGBE_FCOE */
2055	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2056
2057	while (likely(total_rx_packets < budget)) {
2058		union ixgbe_adv_rx_desc *rx_desc;
2059		struct sk_buff *skb;
2060
2061		/* return some buffers to hardware, one at a time is too slow */
2062		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2063			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2064			cleaned_count = 0;
2065		}
2066
2067		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2068
2069		if (!rx_desc->wb.upper.status_error)
2070			break;
2071
2072		/* This memory barrier is needed to keep us from reading
2073		 * any other fields out of the rx_desc until we know the
2074		 * descriptor has been written back
2075		 */
2076		dma_rmb();
2077
2078		/* retrieve a buffer from the ring */
2079		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
2080
2081		/* exit if we failed to retrieve a buffer */
2082		if (!skb)
2083			break;
2084
2085		cleaned_count++;
2086
2087		/* place incomplete frames back on ring for completion */
2088		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2089			continue;
2090
2091		/* verify the packet layout is correct */
2092		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2093			continue;
2094
2095		/* probably a little skewed due to removing CRC */
2096		total_rx_bytes += skb->len;
2097
2098		/* populate checksum, timestamp, VLAN, and protocol */
2099		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2100
2101#ifdef IXGBE_FCOE
2102		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2103		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2104			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2105			/* include DDPed FCoE data */
2106			if (ddp_bytes > 0) {
2107				if (!mss) {
2108					mss = rx_ring->netdev->mtu -
2109						sizeof(struct fcoe_hdr) -
2110						sizeof(struct fc_frame_header) -
2111						sizeof(struct fcoe_crc_eof);
2112					if (mss > 512)
2113						mss &= ~511;
2114				}
2115				total_rx_bytes += ddp_bytes;
2116				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2117								 mss);
2118			}
2119			if (!ddp_bytes) {
2120				dev_kfree_skb_any(skb);
2121				continue;
2122			}
2123		}
2124
2125#endif /* IXGBE_FCOE */
2126		skb_mark_napi_id(skb, &q_vector->napi);
2127		ixgbe_rx_skb(q_vector, skb);
2128
2129		/* update budget accounting */
2130		total_rx_packets++;
2131	}
2132
2133	u64_stats_update_begin(&rx_ring->syncp);
2134	rx_ring->stats.packets += total_rx_packets;
2135	rx_ring->stats.bytes += total_rx_bytes;
2136	u64_stats_update_end(&rx_ring->syncp);
2137	q_vector->rx.total_packets += total_rx_packets;
2138	q_vector->rx.total_bytes += total_rx_bytes;
2139
2140	return total_rx_packets;
2141}
2142
2143#ifdef CONFIG_NET_RX_BUSY_POLL
2144/* must be called with local_bh_disable()d */
2145static int ixgbe_low_latency_recv(struct napi_struct *napi)
2146{
2147	struct ixgbe_q_vector *q_vector =
2148			container_of(napi, struct ixgbe_q_vector, napi);
2149	struct ixgbe_adapter *adapter = q_vector->adapter;
2150	struct ixgbe_ring  *ring;
2151	int found = 0;
2152
2153	if (test_bit(__IXGBE_DOWN, &adapter->state))
2154		return LL_FLUSH_FAILED;
2155
2156	if (!ixgbe_qv_lock_poll(q_vector))
2157		return LL_FLUSH_BUSY;
2158
2159	ixgbe_for_each_ring(ring, q_vector->rx) {
2160		found = ixgbe_clean_rx_irq(q_vector, ring, 4);
2161#ifdef BP_EXTENDED_STATS
2162		if (found)
2163			ring->stats.cleaned += found;
2164		else
2165			ring->stats.misses++;
2166#endif
2167		if (found)
2168			break;
2169	}
2170
2171	ixgbe_qv_unlock_poll(q_vector);
2172
2173	return found;
2174}
2175#endif	/* CONFIG_NET_RX_BUSY_POLL */
2176
2177/**
2178 * ixgbe_configure_msix - Configure MSI-X hardware
2179 * @adapter: board private structure
2180 *
2181 * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2182 * interrupts.
2183 **/
2184static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2185{
2186	struct ixgbe_q_vector *q_vector;
2187	int v_idx;
2188	u32 mask;
2189
2190	/* Populate MSIX to EITR Select */
2191	if (adapter->num_vfs > 32) {
2192		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
2193		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2194	}
2195
2196	/*
2197	 * Populate the IVAR table and set the ITR values to the
2198	 * corresponding register.
2199	 */
2200	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2201		struct ixgbe_ring *ring;
2202		q_vector = adapter->q_vector[v_idx];
2203
2204		ixgbe_for_each_ring(ring, q_vector->rx)
2205			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2206
2207		ixgbe_for_each_ring(ring, q_vector->tx)
2208			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2209
2210		ixgbe_write_eitr(q_vector);
2211	}
2212
2213	switch (adapter->hw.mac.type) {
2214	case ixgbe_mac_82598EB:
2215		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2216			       v_idx);
2217		break;
2218	case ixgbe_mac_82599EB:
2219	case ixgbe_mac_X540:
2220	case ixgbe_mac_X550:
2221	case ixgbe_mac_X550EM_x:
2222		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2223		break;
2224	default:
2225		break;
2226	}
2227	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2228
2229	/* set up to autoclear timer, and the vectors */
2230	mask = IXGBE_EIMS_ENABLE_MASK;
2231	mask &= ~(IXGBE_EIMS_OTHER |
2232		  IXGBE_EIMS_MAILBOX |
2233		  IXGBE_EIMS_LSC);
2234
2235	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2236}
2237
2238enum latency_range {
2239	lowest_latency = 0,
2240	low_latency = 1,
2241	bulk_latency = 2,
2242	latency_invalid = 255
2243};
2244
2245/**
2246 * ixgbe_update_itr - update the dynamic ITR value based on statistics
2247 * @q_vector: structure containing interrupt and ring information
2248 * @ring_container: structure containing ring performance data
2249 *
2250 *      Stores a new ITR value based on packets and byte
2251 *      counts during the last interrupt.  The advantage of per interrupt
2252 *      computation is faster updates and more accurate ITR for the current
2253 *      traffic pattern.  Constants in this function were computed
2254 *      based on theoretical maximum wire speed and thresholds were set based
2255 *      on testing data as well as attempting to minimize response time
2256 *      while increasing bulk throughput.
2257 *      this functionality is controlled by the InterruptThrottleRate module
2258 *      parameter (see ixgbe_param.c)
2259 **/
2260static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2261			     struct ixgbe_ring_container *ring_container)
2262{
2263	int bytes = ring_container->total_bytes;
2264	int packets = ring_container->total_packets;
2265	u32 timepassed_us;
2266	u64 bytes_perint;
2267	u8 itr_setting = ring_container->itr;
2268
2269	if (packets == 0)
2270		return;
2271
2272	/* simple throttlerate management
2273	 *   0-10MB/s   lowest (100000 ints/s)
2274	 *  10-20MB/s   low    (20000 ints/s)
2275	 *  20-1249MB/s bulk   (12000 ints/s)
2276	 */
2277	/* what was last interrupt timeslice? */
2278	timepassed_us = q_vector->itr >> 2;
2279	if (timepassed_us == 0)
2280		return;
2281
2282	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2283
2284	switch (itr_setting) {
2285	case lowest_latency:
2286		if (bytes_perint > 10)
2287			itr_setting = low_latency;
2288		break;
2289	case low_latency:
2290		if (bytes_perint > 20)
2291			itr_setting = bulk_latency;
2292		else if (bytes_perint <= 10)
2293			itr_setting = lowest_latency;
2294		break;
2295	case bulk_latency:
2296		if (bytes_perint <= 20)
2297			itr_setting = low_latency;
2298		break;
2299	}
2300
2301	/* clear work counters since we have the values we need */
2302	ring_container->total_bytes = 0;
2303	ring_container->total_packets = 0;
2304
2305	/* write updated itr to ring container */
2306	ring_container->itr = itr_setting;
2307}
2308
2309/**
2310 * ixgbe_write_eitr - write EITR register in hardware specific way
2311 * @q_vector: structure containing interrupt and ring information
2312 *
2313 * This function is made to be called by ethtool and by the driver
2314 * when it needs to update EITR registers at runtime.  Hardware
2315 * specific quirks/differences are taken care of here.
2316 */
2317void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2318{
2319	struct ixgbe_adapter *adapter = q_vector->adapter;
2320	struct ixgbe_hw *hw = &adapter->hw;
2321	int v_idx = q_vector->v_idx;
2322	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2323
2324	switch (adapter->hw.mac.type) {
2325	case ixgbe_mac_82598EB:
2326		/* must write high and low 16 bits to reset counter */
2327		itr_reg |= (itr_reg << 16);
2328		break;
2329	case ixgbe_mac_82599EB:
2330	case ixgbe_mac_X540:
2331	case ixgbe_mac_X550:
2332	case ixgbe_mac_X550EM_x:
2333		/*
2334		 * set the WDIS bit to not clear the timer bits and cause an
2335		 * immediate assertion of the interrupt
2336		 */
2337		itr_reg |= IXGBE_EITR_CNT_WDIS;
2338		break;
2339	default:
2340		break;
2341	}
2342	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2343}
2344
2345static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2346{
2347	u32 new_itr = q_vector->itr;
2348	u8 current_itr;
2349
2350	ixgbe_update_itr(q_vector, &q_vector->tx);
2351	ixgbe_update_itr(q_vector, &q_vector->rx);
2352
2353	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2354
2355	switch (current_itr) {
2356	/* counts and packets in update_itr are dependent on these numbers */
2357	case lowest_latency:
2358		new_itr = IXGBE_100K_ITR;
2359		break;
2360	case low_latency:
2361		new_itr = IXGBE_20K_ITR;
2362		break;
2363	case bulk_latency:
2364		new_itr = IXGBE_12K_ITR;
2365		break;
2366	default:
2367		break;
2368	}
2369
2370	if (new_itr != q_vector->itr) {
2371		/* do an exponential smoothing */
2372		new_itr = (10 * new_itr * q_vector->itr) /
2373			  ((9 * new_itr) + q_vector->itr);
2374
2375		/* save the algorithm value here */
2376		q_vector->itr = new_itr;
2377
2378		ixgbe_write_eitr(q_vector);
2379	}
2380}
2381
2382/**
2383 * ixgbe_check_overtemp_subtask - check for over temperature
2384 * @adapter: pointer to adapter
2385 **/
2386static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2387{
2388	struct ixgbe_hw *hw = &adapter->hw;
2389	u32 eicr = adapter->interrupt_event;
2390
2391	if (test_bit(__IXGBE_DOWN, &adapter->state))
2392		return;
2393
2394	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2395	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2396		return;
2397
2398	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2399
2400	switch (hw->device_id) {
2401	case IXGBE_DEV_ID_82599_T3_LOM:
2402		/*
2403		 * Since the warning interrupt is for both ports
2404		 * we don't have to check if:
2405		 *  - This interrupt wasn't for our port.
2406		 *  - We may have missed the interrupt so always have to
2407		 *    check if we  got a LSC
2408		 */
2409		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2410		    !(eicr & IXGBE_EICR_LSC))
2411			return;
2412
2413		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2414			u32 speed;
2415			bool link_up = false;
2416
2417			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2418
2419			if (link_up)
2420				return;
2421		}
2422
2423		/* Check if this is not due to overtemp */
2424		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2425			return;
2426
2427		break;
2428	default:
2429		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2430			return;
2431		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2432			return;
2433		break;
2434	}
2435	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2436
2437	adapter->interrupt_event = 0;
2438}
2439
2440static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2441{
2442	struct ixgbe_hw *hw = &adapter->hw;
2443
2444	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2445	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2446		e_crit(probe, "Fan has stopped, replace the adapter\n");
2447		/* write to clear the interrupt */
2448		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2449	}
2450}
2451
2452static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2453{
2454	struct ixgbe_hw *hw = &adapter->hw;
2455
2456	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2457		return;
2458
2459	switch (adapter->hw.mac.type) {
2460	case ixgbe_mac_82599EB:
2461		/*
2462		 * Need to check link state so complete overtemp check
2463		 * on service task
2464		 */
2465		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2466		     (eicr & IXGBE_EICR_LSC)) &&
2467		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2468			adapter->interrupt_event = eicr;
2469			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2470			ixgbe_service_event_schedule(adapter);
2471			return;
2472		}
2473		return;
2474	case ixgbe_mac_X540:
2475		if (!(eicr & IXGBE_EICR_TS))
2476			return;
2477		break;
2478	default:
2479		return;
2480	}
2481
2482	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2483}
2484
2485static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2486{
2487	switch (hw->mac.type) {
2488	case ixgbe_mac_82598EB:
2489		if (hw->phy.type == ixgbe_phy_nl)
2490			return true;
2491		return false;
2492	case ixgbe_mac_82599EB:
2493	case ixgbe_mac_X550EM_x:
2494		switch (hw->mac.ops.get_media_type(hw)) {
2495		case ixgbe_media_type_fiber:
2496		case ixgbe_media_type_fiber_qsfp:
2497			return true;
2498		default:
2499			return false;
2500		}
2501	default:
2502		return false;
2503	}
2504}
2505
2506static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2507{
2508	struct ixgbe_hw *hw = &adapter->hw;
2509	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2510
2511	if (!ixgbe_is_sfp(hw))
2512		return;
2513
2514	/* Later MAC's use different SDP */
2515	if (hw->mac.type >= ixgbe_mac_X540)
2516		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2517
2518	if (eicr & eicr_mask) {
2519		/* Clear the interrupt */
2520		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2521		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2522			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2523			adapter->sfp_poll_time = 0;
2524			ixgbe_service_event_schedule(adapter);
2525		}
2526	}
2527
2528	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2529	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2530		/* Clear the interrupt */
2531		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2532		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2533			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2534			ixgbe_service_event_schedule(adapter);
2535		}
2536	}
2537}
2538
2539static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2540{
2541	struct ixgbe_hw *hw = &adapter->hw;
2542
2543	adapter->lsc_int++;
2544	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2545	adapter->link_check_timeout = jiffies;
2546	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2547		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2548		IXGBE_WRITE_FLUSH(hw);
2549		ixgbe_service_event_schedule(adapter);
2550	}
2551}
2552
2553static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2554					   u64 qmask)
2555{
2556	u32 mask;
2557	struct ixgbe_hw *hw = &adapter->hw;
2558
2559	switch (hw->mac.type) {
2560	case ixgbe_mac_82598EB:
2561		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2562		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2563		break;
2564	case ixgbe_mac_82599EB:
2565	case ixgbe_mac_X540:
2566	case ixgbe_mac_X550:
2567	case ixgbe_mac_X550EM_x:
2568		mask = (qmask & 0xFFFFFFFF);
2569		if (mask)
2570			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2571		mask = (qmask >> 32);
2572		if (mask)
2573			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2574		break;
2575	default:
2576		break;
2577	}
2578	/* skip the flush */
2579}
2580
2581static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2582					    u64 qmask)
2583{
2584	u32 mask;
2585	struct ixgbe_hw *hw = &adapter->hw;
2586
2587	switch (hw->mac.type) {
2588	case ixgbe_mac_82598EB:
2589		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2590		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2591		break;
2592	case ixgbe_mac_82599EB:
2593	case ixgbe_mac_X540:
2594	case ixgbe_mac_X550:
2595	case ixgbe_mac_X550EM_x:
2596		mask = (qmask & 0xFFFFFFFF);
2597		if (mask)
2598			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2599		mask = (qmask >> 32);
2600		if (mask)
2601			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2602		break;
2603	default:
2604		break;
2605	}
2606	/* skip the flush */
2607}
2608
2609/**
2610 * ixgbe_irq_enable - Enable default interrupt generation settings
2611 * @adapter: board private structure
2612 **/
2613static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2614				    bool flush)
2615{
2616	struct ixgbe_hw *hw = &adapter->hw;
2617	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2618
2619	/* don't reenable LSC while waiting for link */
2620	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2621		mask &= ~IXGBE_EIMS_LSC;
2622
2623	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2624		switch (adapter->hw.mac.type) {
2625		case ixgbe_mac_82599EB:
2626			mask |= IXGBE_EIMS_GPI_SDP0(hw);
2627			break;
2628		case ixgbe_mac_X540:
2629		case ixgbe_mac_X550:
2630		case ixgbe_mac_X550EM_x:
2631			mask |= IXGBE_EIMS_TS;
2632			break;
2633		default:
2634			break;
2635		}
2636	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2637		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2638	switch (adapter->hw.mac.type) {
2639	case ixgbe_mac_82599EB:
2640		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2641		mask |= IXGBE_EIMS_GPI_SDP2(hw);
2642		/* fall through */
2643	case ixgbe_mac_X540:
2644	case ixgbe_mac_X550:
2645	case ixgbe_mac_X550EM_x:
2646		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP)
2647			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
2648		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
2649			mask |= IXGBE_EICR_GPI_SDP0_X540;
2650		mask |= IXGBE_EIMS_ECC;
2651		mask |= IXGBE_EIMS_MAILBOX;
2652		break;
2653	default:
2654		break;
2655	}
2656
2657	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2658	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2659		mask |= IXGBE_EIMS_FLOW_DIR;
2660
2661	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2662	if (queues)
2663		ixgbe_irq_enable_queues(adapter, ~0);
2664	if (flush)
2665		IXGBE_WRITE_FLUSH(&adapter->hw);
2666}
2667
2668static irqreturn_t ixgbe_msix_other(int irq, void *data)
2669{
2670	struct ixgbe_adapter *adapter = data;
2671	struct ixgbe_hw *hw = &adapter->hw;
2672	u32 eicr;
2673
2674	/*
2675	 * Workaround for Silicon errata.  Use clear-by-write instead
2676	 * of clear-by-read.  Reading with EICS will return the
2677	 * interrupt causes without clearing, which later be done
2678	 * with the write to EICR.
2679	 */
2680	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2681
2682	/* The lower 16bits of the EICR register are for the queue interrupts
2683	 * which should be masked here in order to not accidentally clear them if
2684	 * the bits are high when ixgbe_msix_other is called. There is a race
2685	 * condition otherwise which results in possible performance loss
2686	 * especially if the ixgbe_msix_other interrupt is triggering
2687	 * consistently (as it would when PPS is turned on for the X540 device)
2688	 */
2689	eicr &= 0xFFFF0000;
2690
2691	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2692
2693	if (eicr & IXGBE_EICR_LSC)
2694		ixgbe_check_lsc(adapter);
2695
2696	if (eicr & IXGBE_EICR_MAILBOX)
2697		ixgbe_msg_task(adapter);
2698
2699	switch (hw->mac.type) {
2700	case ixgbe_mac_82599EB:
2701	case ixgbe_mac_X540:
2702	case ixgbe_mac_X550:
2703	case ixgbe_mac_X550EM_x:
2704		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
2705		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2706			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
2707			ixgbe_service_event_schedule(adapter);
2708			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2709					IXGBE_EICR_GPI_SDP0_X540);
2710		}
2711		if (eicr & IXGBE_EICR_ECC) {
2712			e_info(link, "Received ECC Err, initiating reset\n");
2713			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2714			ixgbe_service_event_schedule(adapter);
2715			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2716		}
2717		/* Handle Flow Director Full threshold interrupt */
2718		if (eicr & IXGBE_EICR_FLOW_DIR) {
2719			int reinit_count = 0;
2720			int i;
2721			for (i = 0; i < adapter->num_tx_queues; i++) {
2722				struct ixgbe_ring *ring = adapter->tx_ring[i];
2723				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2724						       &ring->state))
2725					reinit_count++;
2726			}
2727			if (reinit_count) {
2728				/* no more flow director interrupts until after init */
2729				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2730				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2731				ixgbe_service_event_schedule(adapter);
2732			}
2733		}
2734		ixgbe_check_sfp_event(adapter, eicr);
2735		ixgbe_check_overtemp_event(adapter, eicr);
2736		break;
2737	default:
2738		break;
2739	}
2740
2741	ixgbe_check_fan_failure(adapter, eicr);
2742
2743	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2744		ixgbe_ptp_check_pps_event(adapter, eicr);
2745
2746	/* re-enable the original interrupt state, no lsc, no queues */
2747	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2748		ixgbe_irq_enable(adapter, false, false);
2749
2750	return IRQ_HANDLED;
2751}
2752
2753static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2754{
2755	struct ixgbe_q_vector *q_vector = data;
2756
2757	/* EIAM disabled interrupts (on this vector) for us */
2758
2759	if (q_vector->rx.ring || q_vector->tx.ring)
2760		napi_schedule(&q_vector->napi);
2761
2762	return IRQ_HANDLED;
2763}
2764
2765/**
2766 * ixgbe_poll - NAPI Rx polling callback
2767 * @napi: structure for representing this polling device
2768 * @budget: how many packets driver is allowed to clean
2769 *
2770 * This function is used for legacy and MSI, NAPI mode
2771 **/
2772int ixgbe_poll(struct napi_struct *napi, int budget)
2773{
2774	struct ixgbe_q_vector *q_vector =
2775				container_of(napi, struct ixgbe_q_vector, napi);
2776	struct ixgbe_adapter *adapter = q_vector->adapter;
2777	struct ixgbe_ring *ring;
2778	int per_ring_budget, work_done = 0;
2779	bool clean_complete = true;
2780
2781#ifdef CONFIG_IXGBE_DCA
2782	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2783		ixgbe_update_dca(q_vector);
2784#endif
2785
2786	ixgbe_for_each_ring(ring, q_vector->tx)
2787		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2788
2789	if (!ixgbe_qv_lock_napi(q_vector))
2790		return budget;
2791
2792	/* attempt to distribute budget to each queue fairly, but don't allow
2793	 * the budget to go below 1 because we'll exit polling */
2794	if (q_vector->rx.count > 1)
2795		per_ring_budget = max(budget/q_vector->rx.count, 1);
2796	else
2797		per_ring_budget = budget;
2798
2799	ixgbe_for_each_ring(ring, q_vector->rx) {
2800		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
2801						 per_ring_budget);
2802
2803		work_done += cleaned;
2804		clean_complete &= (cleaned < per_ring_budget);
2805	}
2806
2807	ixgbe_qv_unlock_napi(q_vector);
2808	/* If all work not completed, return budget and keep polling */
2809	if (!clean_complete)
2810		return budget;
2811
2812	/* all work done, exit the polling mode */
2813	napi_complete_done(napi, work_done);
2814	if (adapter->rx_itr_setting & 1)
2815		ixgbe_set_itr(q_vector);
2816	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2817		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2818
2819	return 0;
2820}
2821
2822/**
2823 * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2824 * @adapter: board private structure
2825 *
2826 * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2827 * interrupts from the kernel.
2828 **/
2829static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2830{
2831	struct net_device *netdev = adapter->netdev;
2832	int vector, err;
2833	int ri = 0, ti = 0;
2834
2835	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2836		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2837		struct msix_entry *entry = &adapter->msix_entries[vector];
2838
2839		if (q_vector->tx.ring && q_vector->rx.ring) {
2840			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2841				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2842			ti++;
2843		} else if (q_vector->rx.ring) {
2844			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2845				 "%s-%s-%d", netdev->name, "rx", ri++);
2846		} else if (q_vector->tx.ring) {
2847			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2848				 "%s-%s-%d", netdev->name, "tx", ti++);
2849		} else {
2850			/* skip this unused q_vector */
2851			continue;
2852		}
2853		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2854				  q_vector->name, q_vector);
2855		if (err) {
2856			e_err(probe, "request_irq failed for MSIX interrupt "
2857			      "Error: %d\n", err);
2858			goto free_queue_irqs;
2859		}
2860		/* If Flow Director is enabled, set interrupt affinity */
2861		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2862			/* assign the mask for this irq */
2863			irq_set_affinity_hint(entry->vector,
2864					      &q_vector->affinity_mask);
2865		}
2866	}
2867
2868	err = request_irq(adapter->msix_entries[vector].vector,
2869			  ixgbe_msix_other, 0, netdev->name, adapter);
2870	if (err) {
2871		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2872		goto free_queue_irqs;
2873	}
2874
2875	return 0;
2876
2877free_queue_irqs:
2878	while (vector) {
2879		vector--;
2880		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2881				      NULL);
2882		free_irq(adapter->msix_entries[vector].vector,
2883			 adapter->q_vector[vector]);
2884	}
2885	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2886	pci_disable_msix(adapter->pdev);
2887	kfree(adapter->msix_entries);
2888	adapter->msix_entries = NULL;
2889	return err;
2890}
2891
2892/**
2893 * ixgbe_intr - legacy mode Interrupt Handler
2894 * @irq: interrupt number
2895 * @data: pointer to a network interface device structure
2896 **/
2897static irqreturn_t ixgbe_intr(int irq, void *data)
2898{
2899	struct ixgbe_adapter *adapter = data;
2900	struct ixgbe_hw *hw = &adapter->hw;
2901	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2902	u32 eicr;
2903
2904	/*
2905	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2906	 * before the read of EICR.
2907	 */
2908	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2909
2910	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2911	 * therefore no explicit interrupt disable is necessary */
2912	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2913	if (!eicr) {
2914		/*
2915		 * shared interrupt alert!
2916		 * make sure interrupts are enabled because the read will
2917		 * have disabled interrupts due to EIAM
2918		 * finish the workaround of silicon errata on 82598.  Unmask
2919		 * the interrupt that we masked before the EICR read.
2920		 */
2921		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2922			ixgbe_irq_enable(adapter, true, true);
2923		return IRQ_NONE;	/* Not our interrupt */
2924	}
2925
2926	if (eicr & IXGBE_EICR_LSC)
2927		ixgbe_check_lsc(adapter);
2928
2929	switch (hw->mac.type) {
2930	case ixgbe_mac_82599EB:
2931		ixgbe_check_sfp_event(adapter, eicr);
2932		/* Fall through */
2933	case ixgbe_mac_X540:
2934	case ixgbe_mac_X550:
2935	case ixgbe_mac_X550EM_x:
2936		if (eicr & IXGBE_EICR_ECC) {
2937			e_info(link, "Received ECC Err, initiating reset\n");
2938			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2939			ixgbe_service_event_schedule(adapter);
2940			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2941		}
2942		ixgbe_check_overtemp_event(adapter, eicr);
2943		break;
2944	default:
2945		break;
2946	}
2947
2948	ixgbe_check_fan_failure(adapter, eicr);
2949	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2950		ixgbe_ptp_check_pps_event(adapter, eicr);
2951
2952	/* would disable interrupts here but EIAM disabled it */
2953	napi_schedule(&q_vector->napi);
2954
2955	/*
2956	 * re-enable link(maybe) and non-queue interrupts, no flush.
2957	 * ixgbe_poll will re-enable the queue interrupts
2958	 */
2959	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2960		ixgbe_irq_enable(adapter, false, false);
2961
2962	return IRQ_HANDLED;
2963}
2964
2965/**
2966 * ixgbe_request_irq - initialize interrupts
2967 * @adapter: board private structure
2968 *
2969 * Attempts to configure interrupts using the best available
2970 * capabilities of the hardware and kernel.
2971 **/
2972static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2973{
2974	struct net_device *netdev = adapter->netdev;
2975	int err;
2976
2977	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2978		err = ixgbe_request_msix_irqs(adapter);
2979	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2980		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2981				  netdev->name, adapter);
2982	else
2983		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2984				  netdev->name, adapter);
2985
2986	if (err)
2987		e_err(probe, "request_irq failed, Error %d\n", err);
2988
2989	return err;
2990}
2991
2992static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2993{
2994	int vector;
2995
2996	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2997		free_irq(adapter->pdev->irq, adapter);
2998		return;
2999	}
3000
3001	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3002		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3003		struct msix_entry *entry = &adapter->msix_entries[vector];
3004
3005		/* free only the irqs that were actually requested */
3006		if (!q_vector->rx.ring && !q_vector->tx.ring)
3007			continue;
3008
3009		/* clear the affinity_mask in the IRQ descriptor */
3010		irq_set_affinity_hint(entry->vector, NULL);
3011
3012		free_irq(entry->vector, q_vector);
3013	}
3014
3015	free_irq(adapter->msix_entries[vector++].vector, adapter);
3016}
3017
3018/**
3019 * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3020 * @adapter: board private structure
3021 **/
3022static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3023{
3024	switch (adapter->hw.mac.type) {
3025	case ixgbe_mac_82598EB:
3026		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3027		break;
3028	case ixgbe_mac_82599EB:
3029	case ixgbe_mac_X540:
3030	case ixgbe_mac_X550:
3031	case ixgbe_mac_X550EM_x:
3032		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3033		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3034		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3035		break;
3036	default:
3037		break;
3038	}
3039	IXGBE_WRITE_FLUSH(&adapter->hw);
3040	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3041		int vector;
3042
3043		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3044			synchronize_irq(adapter->msix_entries[vector].vector);
3045
3046		synchronize_irq(adapter->msix_entries[vector++].vector);
3047	} else {
3048		synchronize_irq(adapter->pdev->irq);
3049	}
3050}
3051
3052/**
3053 * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3054 *
3055 **/
3056static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3057{
3058	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3059
3060	ixgbe_write_eitr(q_vector);
3061
3062	ixgbe_set_ivar(adapter, 0, 0, 0);
3063	ixgbe_set_ivar(adapter, 1, 0, 0);
3064
3065	e_info(hw, "Legacy interrupt IVAR setup done\n");
3066}
3067
3068/**
3069 * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3070 * @adapter: board private structure
3071 * @ring: structure containing ring specific data
3072 *
3073 * Configure the Tx descriptor ring after a reset.
3074 **/
3075void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3076			     struct ixgbe_ring *ring)
3077{
3078	struct ixgbe_hw *hw = &adapter->hw;
3079	u64 tdba = ring->dma;
3080	int wait_loop = 10;
3081	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3082	u8 reg_idx = ring->reg_idx;
3083
3084	/* disable queue to avoid issues while updating state */
3085	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3086	IXGBE_WRITE_FLUSH(hw);
3087
3088	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3089			(tdba & DMA_BIT_MASK(32)));
3090	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3091	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3092			ring->count * sizeof(union ixgbe_adv_tx_desc));
3093	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3094	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3095	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3096
3097	/*
3098	 * set WTHRESH to encourage burst writeback, it should not be set
3099	 * higher than 1 when:
3100	 * - ITR is 0 as it could cause false TX hangs
3101	 * - ITR is set to > 100k int/sec and BQL is enabled
3102	 *
3103	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3104	 * to or less than the number of on chip descriptors, which is
3105	 * currently 40.
3106	 */
3107	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3108		txdctl |= (1 << 16);	/* WTHRESH = 1 */
3109	else
3110		txdctl |= (8 << 16);	/* WTHRESH = 8 */
3111
3112	/*
3113	 * Setting PTHRESH to 32 both improves performance
3114	 * and avoids a TX hang with DFP enabled
3115	 */
3116	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
3117		   32;		/* PTHRESH = 32 */
3118
3119	/* reinitialize flowdirector state */
3120	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3121		ring->atr_sample_rate = adapter->atr_sample_rate;
3122		ring->atr_count = 0;
3123		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3124	} else {
3125		ring->atr_sample_rate = 0;
3126	}
3127
3128	/* initialize XPS */
3129	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3130		struct ixgbe_q_vector *q_vector = ring->q_vector;
3131
3132		if (q_vector)
3133			netif_set_xps_queue(ring->netdev,
3134					    &q_vector->affinity_mask,
3135					    ring->queue_index);
3136	}
3137
3138	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3139
3140	/* enable queue */
3141	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3142
3143	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3144	if (hw->mac.type == ixgbe_mac_82598EB &&
3145	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3146		return;
3147
3148	/* poll to verify queue is enabled */
3149	do {
3150		usleep_range(1000, 2000);
3151		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3152	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3153	if (!wait_loop)
3154		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
3155}
3156
3157static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3158{
3159	struct ixgbe_hw *hw = &adapter->hw;
3160	u32 rttdcs, mtqc;
3161	u8 tcs = netdev_get_num_tc(adapter->netdev);
3162
3163	if (hw->mac.type == ixgbe_mac_82598EB)
3164		return;
3165
3166	/* disable the arbiter while setting MTQC */
3167	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3168	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3169	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3170
3171	/* set transmit pool layout */
3172	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3173		mtqc = IXGBE_MTQC_VT_ENA;
3174		if (tcs > 4)
3175			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3176		else if (tcs > 1)
3177			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3178		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3179			mtqc |= IXGBE_MTQC_32VF;
3180		else
3181			mtqc |= IXGBE_MTQC_64VF;
3182	} else {
3183		if (tcs > 4)
3184			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3185		else if (tcs > 1)
3186			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3187		else
3188			mtqc = IXGBE_MTQC_64Q_1PB;
3189	}
3190
3191	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3192
3193	/* Enable Security TX Buffer IFG for multiple pb */
3194	if (tcs) {
3195		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3196		sectx |= IXGBE_SECTX_DCB;
3197		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3198	}
3199
3200	/* re-enable the arbiter */
3201	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3202	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3203}
3204
3205/**
3206 * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3207 * @adapter: board private structure
3208 *
3209 * Configure the Tx unit of the MAC after a reset.
3210 **/
3211static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3212{
3213	struct ixgbe_hw *hw = &adapter->hw;
3214	u32 dmatxctl;
3215	u32 i;
3216
3217	ixgbe_setup_mtqc(adapter);
3218
3219	if (hw->mac.type != ixgbe_mac_82598EB) {
3220		/* DMATXCTL.EN must be before Tx queues are enabled */
3221		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3222		dmatxctl |= IXGBE_DMATXCTL_TE;
3223		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3224	}
3225
3226	/* Setup the HW Tx Head and Tail descriptor pointers */
3227	for (i = 0; i < adapter->num_tx_queues; i++)
3228		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3229}
3230
3231static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3232				 struct ixgbe_ring *ring)
3233{
3234	struct ixgbe_hw *hw = &adapter->hw;
3235	u8 reg_idx = ring->reg_idx;
3236	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3237
3238	srrctl |= IXGBE_SRRCTL_DROP_EN;
3239
3240	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3241}
3242
3243static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3244				  struct ixgbe_ring *ring)
3245{
3246	struct ixgbe_hw *hw = &adapter->hw;
3247	u8 reg_idx = ring->reg_idx;
3248	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3249
3250	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3251
3252	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3253}
3254
3255#ifdef CONFIG_IXGBE_DCB
3256void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3257#else
3258static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3259#endif
3260{
3261	int i;
3262	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3263
3264	if (adapter->ixgbe_ieee_pfc)
3265		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3266
3267	/*
3268	 * We should set the drop enable bit if:
3269	 *  SR-IOV is enabled
3270	 *   or
3271	 *  Number of Rx queues > 1 and flow control is disabled
3272	 *
3273	 *  This allows us to avoid head of line blocking for security
3274	 *  and performance reasons.
3275	 */
3276	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3277	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3278		for (i = 0; i < adapter->num_rx_queues; i++)
3279			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3280	} else {
3281		for (i = 0; i < adapter->num_rx_queues; i++)
3282			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3283	}
3284}
3285
3286#define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3287
3288static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3289				   struct ixgbe_ring *rx_ring)
3290{
3291	struct ixgbe_hw *hw = &adapter->hw;
3292	u32 srrctl;
3293	u8 reg_idx = rx_ring->reg_idx;
3294
3295	if (hw->mac.type == ixgbe_mac_82598EB) {
3296		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3297
3298		/*
3299		 * if VMDq is not active we must program one srrctl register
3300		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3301		 */
3302		reg_idx &= mask;
3303	}
3304
3305	/* configure header buffer length, needed for RSC */
3306	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3307
3308	/* configure the packet buffer length */
3309	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3310
3311	/* configure descriptor type */
3312	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3313
3314	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3315}
3316
3317/**
3318 * Return a number of entries in the RSS indirection table
3319 *
3320 * @adapter: device handle
3321 *
3322 *  - 82598/82599/X540:     128
3323 *  - X550(non-SRIOV mode): 512
3324 *  - X550(SRIOV mode):     64
3325 */
3326u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3327{
3328	if (adapter->hw.mac.type < ixgbe_mac_X550)
3329		return 128;
3330	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3331		return 64;
3332	else
3333		return 512;
3334}
3335
3336/**
3337 * Write the RETA table to HW
3338 *
3339 * @adapter: device handle
3340 *
3341 * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3342 */
3343void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3344{
3345	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3346	struct ixgbe_hw *hw = &adapter->hw;
3347	u32 reta = 0;
3348	u32 indices_multi;
3349	u8 *indir_tbl = adapter->rss_indir_tbl;
3350
3351	/* Fill out the redirection table as follows:
3352	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3353	 *    indices.
3354	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3355	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3356	 */
3357	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3358		indices_multi = 0x11;
3359	else
3360		indices_multi = 0x1;
3361
3362	/* Write redirection table to HW */
3363	for (i = 0; i < reta_entries; i++) {
3364		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3365		if ((i & 3) == 3) {
3366			if (i < 128)
3367				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3368			else
3369				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3370						reta);
3371			reta = 0;
3372		}
3373	}
3374}
3375
3376/**
3377 * Write the RETA table to HW (for x550 devices in SRIOV mode)
3378 *
3379 * @adapter: device handle
3380 *
3381 * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3382 */
3383static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3384{
3385	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3386	struct ixgbe_hw *hw = &adapter->hw;
3387	u32 vfreta = 0;
3388	unsigned int pf_pool = adapter->num_vfs;
3389
3390	/* Write redirection table to HW */
3391	for (i = 0; i < reta_entries; i++) {
3392		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3393		if ((i & 3) == 3) {
3394			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3395					vfreta);
3396			vfreta = 0;
3397		}
3398	}
3399}
3400
3401static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3402{
3403	struct ixgbe_hw *hw = &adapter->hw;
3404	u32 i, j;
3405	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3406	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3407
3408	/* Program table for at least 2 queues w/ SR-IOV so that VFs can
3409	 * make full use of any rings they may have.  We will use the
3410	 * PSRTYPE register to control how many rings we use within the PF.
3411	 */
3412	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3413		rss_i = 2;
3414
3415	/* Fill out hash function seeds */
3416	for (i = 0; i < 10; i++)
3417		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3418
3419	/* Fill out redirection table */
3420	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3421
3422	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3423		if (j == rss_i)
3424			j = 0;
3425
3426		adapter->rss_indir_tbl[i] = j;
3427	}
3428
3429	ixgbe_store_reta(adapter);
3430}
3431
3432static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3433{
3434	struct ixgbe_hw *hw = &adapter->hw;
3435	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3436	unsigned int pf_pool = adapter->num_vfs;
3437	int i, j;
3438
3439	/* Fill out hash function seeds */
3440	for (i = 0; i < 10; i++)
3441		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool),
3442				adapter->rss_key[i]);
3443
3444	/* Fill out the redirection table */
3445	for (i = 0, j = 0; i < 64; i++, j++) {
3446		if (j == rss_i)
3447			j = 0;
3448
3449		adapter->rss_indir_tbl[i] = j;
3450	}
3451
3452	ixgbe_store_vfreta(adapter);
3453}
3454
3455static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3456{
3457	struct ixgbe_hw *hw = &adapter->hw;
3458	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3459	u32 rxcsum;
3460
3461	/* Disable indicating checksum in descriptor, enables RSS hash */
3462	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3463	rxcsum |= IXGBE_RXCSUM_PCSD;
3464	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3465
3466	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3467		if (adapter->ring_feature[RING_F_RSS].mask)
3468			mrqc = IXGBE_MRQC_RSSEN;
3469	} else {
3470		u8 tcs = netdev_get_num_tc(adapter->netdev);
3471
3472		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3473			if (tcs > 4)
3474				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3475			else if (tcs > 1)
3476				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3477			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3478				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3479			else
3480				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3481		} else {
3482			if (tcs > 4)
3483				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3484			else if (tcs > 1)
3485				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3486			else
3487				mrqc = IXGBE_MRQC_RSSEN;
3488		}
3489	}
3490
3491	/* Perform hash on these packet types */
3492	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3493		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3494		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3495		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3496
3497	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3498		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3499	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3500		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3501
3502	netdev_rss_key_fill(adapter->rss_key, sizeof(adapter->rss_key));
3503	if ((hw->mac.type >= ixgbe_mac_X550) &&
3504	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3505		unsigned int pf_pool = adapter->num_vfs;
3506
3507		/* Enable VF RSS mode */
3508		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3509		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3510
3511		/* Setup RSS through the VF registers */
3512		ixgbe_setup_vfreta(adapter);
3513		vfmrqc = IXGBE_MRQC_RSSEN;
3514		vfmrqc |= rss_field;
3515		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3516	} else {
3517		ixgbe_setup_reta(adapter);
3518		mrqc |= rss_field;
3519		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3520	}
3521}
3522
3523/**
3524 * ixgbe_configure_rscctl - enable RSC for the indicated ring
3525 * @adapter:    address of board private structure
3526 * @index:      index of ring to set
3527 **/
3528static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3529				   struct ixgbe_ring *ring)
3530{
3531	struct ixgbe_hw *hw = &adapter->hw;
3532	u32 rscctrl;
3533	u8 reg_idx = ring->reg_idx;
3534
3535	if (!ring_is_rsc_enabled(ring))
3536		return;
3537
3538	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3539	rscctrl |= IXGBE_RSCCTL_RSCEN;
3540	/*
3541	 * we must limit the number of descriptors so that the
3542	 * total size of max desc * buf_len is not greater
3543	 * than 65536
3544	 */
3545	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3546	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3547}
3548
3549#define IXGBE_MAX_RX_DESC_POLL 10
3550static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3551				       struct ixgbe_ring *ring)
3552{
3553	struct ixgbe_hw *hw = &adapter->hw;
3554	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3555	u32 rxdctl;
3556	u8 reg_idx = ring->reg_idx;
3557
3558	if (ixgbe_removed(hw->hw_addr))
3559		return;
3560	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3561	if (hw->mac.type == ixgbe_mac_82598EB &&
3562	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3563		return;
3564
3565	do {
3566		usleep_range(1000, 2000);
3567		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3568	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3569
3570	if (!wait_loop) {
3571		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3572		      "the polling period\n", reg_idx);
3573	}
3574}
3575
3576void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3577			    struct ixgbe_ring *ring)
3578{
3579	struct ixgbe_hw *hw = &adapter->hw;
3580	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3581	u32 rxdctl;
3582	u8 reg_idx = ring->reg_idx;
3583
3584	if (ixgbe_removed(hw->hw_addr))
3585		return;
3586	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3587	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3588
3589	/* write value back with RXDCTL.ENABLE bit cleared */
3590	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3591
3592	if (hw->mac.type == ixgbe_mac_82598EB &&
3593	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3594		return;
3595
3596	/* the hardware may take up to 100us to really disable the rx queue */
3597	do {
3598		udelay(10);
3599		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3600	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3601
3602	if (!wait_loop) {
3603		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3604		      "the polling period\n", reg_idx);
3605	}
3606}
3607
3608void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3609			     struct ixgbe_ring *ring)
3610{
3611	struct ixgbe_hw *hw = &adapter->hw;
3612	u64 rdba = ring->dma;
3613	u32 rxdctl;
3614	u8 reg_idx = ring->reg_idx;
3615
3616	/* disable queue to avoid issues while updating state */
3617	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3618	ixgbe_disable_rx_queue(adapter, ring);
3619
3620	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3621	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3622	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3623			ring->count * sizeof(union ixgbe_adv_rx_desc));
3624	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3625	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3626	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3627
3628	ixgbe_configure_srrctl(adapter, ring);
3629	ixgbe_configure_rscctl(adapter, ring);
3630
3631	if (hw->mac.type == ixgbe_mac_82598EB) {
3632		/*
3633		 * enable cache line friendly hardware writes:
3634		 * PTHRESH=32 descriptors (half the internal cache),
3635		 * this also removes ugly rx_no_buffer_count increment
3636		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3637		 * WTHRESH=8 burst writeback up to two cache lines
3638		 */
3639		rxdctl &= ~0x3FFFFF;
3640		rxdctl |=  0x080420;
3641	}
3642
3643	/* enable receive descriptor ring */
3644	rxdctl |= IXGBE_RXDCTL_ENABLE;
3645	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3646
3647	ixgbe_rx_desc_queue_enable(adapter, ring);
3648	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3649}
3650
3651static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3652{
3653	struct ixgbe_hw *hw = &adapter->hw;
3654	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3655	u16 pool;
3656
3657	/* PSRTYPE must be initialized in non 82598 adapters */
3658	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3659		      IXGBE_PSRTYPE_UDPHDR |
3660		      IXGBE_PSRTYPE_IPV4HDR |
3661		      IXGBE_PSRTYPE_L2HDR |
3662		      IXGBE_PSRTYPE_IPV6HDR;
3663
3664	if (hw->mac.type == ixgbe_mac_82598EB)
3665		return;
3666
3667	if (rss_i > 3)
3668		psrtype |= 2 << 29;
3669	else if (rss_i > 1)
3670		psrtype |= 1 << 29;
3671
3672	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3673		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3674}
3675
3676static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3677{
3678	struct ixgbe_hw *hw = &adapter->hw;
3679	u32 reg_offset, vf_shift;
3680	u32 gcr_ext, vmdctl;
3681	int i;
3682
3683	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3684		return;
3685
3686	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3687	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3688	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3689	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3690	vmdctl |= IXGBE_VT_CTL_REPLEN;
3691	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3692
3693	vf_shift = VMDQ_P(0) % 32;
3694	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3695
3696	/* Enable only the PF's pool for Tx/Rx */
3697	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (~0) << vf_shift);
3698	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3699	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (~0) << vf_shift);
3700	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3701	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
3702		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3703
3704	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3705	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3706
3707	/*
3708	 * Set up VF register offsets for selected VT Mode,
3709	 * i.e. 32 or 64 VFs for SR-IOV
3710	 */
3711	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3712	case IXGBE_82599_VMDQ_8Q_MASK:
3713		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3714		break;
3715	case IXGBE_82599_VMDQ_4Q_MASK:
3716		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3717		break;
3718	default:
3719		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3720		break;
3721	}
3722
3723	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3724
3725
3726	/* Enable MAC Anti-Spoofing */
3727	hw->mac.ops.set_mac_anti_spoofing(hw, (adapter->num_vfs != 0),
3728					  adapter->num_vfs);
3729
3730	/* Ensure LLDP and FC is set for Ethertype Antispoofing if we will be
3731	 * calling set_ethertype_anti_spoofing for each VF in loop below
3732	 */
3733	if (hw->mac.ops.set_ethertype_anti_spoofing) {
3734		IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_LLDP),
3735				(IXGBE_ETQF_FILTER_EN    |
3736				 IXGBE_ETQF_TX_ANTISPOOF |
3737				 IXGBE_ETH_P_LLDP));
3738
3739		IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_FC),
3740				(IXGBE_ETQF_FILTER_EN |
3741				 IXGBE_ETQF_TX_ANTISPOOF |
3742				 ETH_P_PAUSE));
3743	}
3744
3745	/* For VFs that have spoof checking turned off */
3746	for (i = 0; i < adapter->num_vfs; i++) {
3747		if (!adapter->vfinfo[i].spoofchk_enabled)
3748			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3749
3750		/* enable ethertype anti spoofing if hw supports it */
3751		if (hw->mac.ops.set_ethertype_anti_spoofing)
3752			hw->mac.ops.set_ethertype_anti_spoofing(hw, true, i);
3753
3754		/* Enable/Disable RSS query feature  */
3755		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
3756					  adapter->vfinfo[i].rss_query_enabled);
3757	}
3758}
3759
3760static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3761{
3762	struct ixgbe_hw *hw = &adapter->hw;
3763	struct net_device *netdev = adapter->netdev;
3764	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3765	struct ixgbe_ring *rx_ring;
3766	int i;
3767	u32 mhadd, hlreg0;
3768
3769#ifdef IXGBE_FCOE
3770	/* adjust max frame to be able to do baby jumbo for FCoE */
3771	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3772	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3773		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3774
3775#endif /* IXGBE_FCOE */
3776
3777	/* adjust max frame to be at least the size of a standard frame */
3778	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3779		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3780
3781	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3782	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3783		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3784		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3785
3786		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3787	}
3788
3789	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3790	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3791	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3792	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3793
3794	/*
3795	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3796	 * the Base and Length of the Rx Descriptor Ring
3797	 */
3798	for (i = 0; i < adapter->num_rx_queues; i++) {
3799		rx_ring = adapter->rx_ring[i];
3800		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3801			set_ring_rsc_enabled(rx_ring);
3802		else
3803			clear_ring_rsc_enabled(rx_ring);
3804	}
3805}
3806
3807static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3808{
3809	struct ixgbe_hw *hw = &adapter->hw;
3810	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3811
3812	switch (hw->mac.type) {
3813	case ixgbe_mac_82598EB:
3814		/*
3815		 * For VMDq support of different descriptor types or
3816		 * buffer sizes through the use of multiple SRRCTL
3817		 * registers, RDRXCTL.MVMEN must be set to 1
3818		 *
3819		 * also, the manual doesn't mention it clearly but DCA hints
3820		 * will only use queue 0's tags unless this bit is set.  Side
3821		 * effects of setting this bit are only that SRRCTL must be
3822		 * fully programmed [0..15]
3823		 */
3824		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3825		break;
3826	case ixgbe_mac_X550:
3827	case ixgbe_mac_X550EM_x:
3828		if (adapter->num_vfs)
3829			rdrxctl |= IXGBE_RDRXCTL_PSP;
3830		/* fall through for older HW */
3831	case ixgbe_mac_82599EB:
3832	case ixgbe_mac_X540:
3833		/* Disable RSC for ACK packets */
3834		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3835		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3836		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3837		/* hardware requires some bits to be set by default */
3838		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3839		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3840		break;
3841	default:
3842		/* We should do nothing since we don't know this hardware */
3843		return;
3844	}
3845
3846	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3847}
3848
3849/**
3850 * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3851 * @adapter: board private structure
3852 *
3853 * Configure the Rx unit of the MAC after a reset.
3854 **/
3855static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3856{
3857	struct ixgbe_hw *hw = &adapter->hw;
3858	int i;
3859	u32 rxctrl, rfctl;
3860
3861	/* disable receives while setting up the descriptors */
3862	hw->mac.ops.disable_rx(hw);
3863
3864	ixgbe_setup_psrtype(adapter);
3865	ixgbe_setup_rdrxctl(adapter);
3866
3867	/* RSC Setup */
3868	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
3869	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
3870	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
3871		rfctl |= IXGBE_RFCTL_RSC_DIS;
3872	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3873
3874	/* Program registers for the distribution of queues */
3875	ixgbe_setup_mrqc(adapter);
3876
3877	/* set_rx_buffer_len must be called before ring initialization */
3878	ixgbe_set_rx_buffer_len(adapter);
3879
3880	/*
3881	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3882	 * the Base and Length of the Rx Descriptor Ring
3883	 */
3884	for (i = 0; i < adapter->num_rx_queues; i++)
3885		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3886
3887	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3888	/* disable drop enable for 82598 parts */
3889	if (hw->mac.type == ixgbe_mac_82598EB)
3890		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3891
3892	/* enable all receives */
3893	rxctrl |= IXGBE_RXCTRL_RXEN;
3894	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3895}
3896
3897static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3898				 __be16 proto, u16 vid)
3899{
3900	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3901	struct ixgbe_hw *hw = &adapter->hw;
3902
3903	/* add VID to filter table */
3904	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true);
3905	set_bit(vid, adapter->active_vlans);
3906
3907	return 0;
3908}
3909
3910static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
3911				  __be16 proto, u16 vid)
3912{
3913	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3914	struct ixgbe_hw *hw = &adapter->hw;
3915
3916	/* remove VID from filter table */
3917	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), false);
3918	clear_bit(vid, adapter->active_vlans);
3919
3920	return 0;
3921}
3922
3923/**
3924 * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3925 * @adapter: driver data
3926 */
3927static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3928{
3929	struct ixgbe_hw *hw = &adapter->hw;
3930	u32 vlnctrl;
3931	int i, j;
3932
3933	switch (hw->mac.type) {
3934	case ixgbe_mac_82598EB:
3935		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3936		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3937		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3938		break;
3939	case ixgbe_mac_82599EB:
3940	case ixgbe_mac_X540:
3941	case ixgbe_mac_X550:
3942	case ixgbe_mac_X550EM_x:
3943		for (i = 0; i < adapter->num_rx_queues; i++) {
3944			struct ixgbe_ring *ring = adapter->rx_ring[i];
3945
3946			if (ring->l2_accel_priv)
3947				continue;
3948			j = ring->reg_idx;
3949			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3950			vlnctrl &= ~IXGBE_RXDCTL_VME;
3951			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3952		}
3953		break;
3954	default:
3955		break;
3956	}
3957}
3958
3959/**
3960 * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3961 * @adapter: driver data
3962 */
3963static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3964{
3965	struct ixgbe_hw *hw = &adapter->hw;
3966	u32 vlnctrl;
3967	int i, j;
3968
3969	switch (hw->mac.type) {
3970	case ixgbe_mac_82598EB:
3971		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3972		vlnctrl |= IXGBE_VLNCTRL_VME;
3973		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3974		break;
3975	case ixgbe_mac_82599EB:
3976	case ixgbe_mac_X540:
3977	case ixgbe_mac_X550:
3978	case ixgbe_mac_X550EM_x:
3979		for (i = 0; i < adapter->num_rx_queues; i++) {
3980			struct ixgbe_ring *ring = adapter->rx_ring[i];
3981
3982			if (ring->l2_accel_priv)
3983				continue;
3984			j = ring->reg_idx;
3985			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3986			vlnctrl |= IXGBE_RXDCTL_VME;
3987			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3988		}
3989		break;
3990	default:
3991		break;
3992	}
3993}
3994
3995static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3996{
3997	u16 vid;
3998
3999	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4000
4001	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
4002		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4003}
4004
4005/**
4006 * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4007 * @netdev: network interface device structure
4008 *
4009 * Writes multicast address list to the MTA hash table.
4010 * Returns: -ENOMEM on failure
4011 *                0 on no addresses written
4012 *                X on writing X addresses to MTA
4013 **/
4014static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4015{
4016	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4017	struct ixgbe_hw *hw = &adapter->hw;
4018
4019	if (!netif_running(netdev))
4020		return 0;
4021
4022	if (hw->mac.ops.update_mc_addr_list)
4023		hw->mac.ops.update_mc_addr_list(hw, netdev);
4024	else
4025		return -ENOMEM;
4026
4027#ifdef CONFIG_PCI_IOV
4028	ixgbe_restore_vf_multicasts(adapter);
4029#endif
4030
4031	return netdev_mc_count(netdev);
4032}
4033
4034#ifdef CONFIG_PCI_IOV
4035void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4036{
4037	struct ixgbe_hw *hw = &adapter->hw;
4038	int i;
4039	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4040		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
4041			hw->mac.ops.set_rar(hw, i, adapter->mac_table[i].addr,
4042					    adapter->mac_table[i].queue,
4043					    IXGBE_RAH_AV);
4044		else
4045			hw->mac.ops.clear_rar(hw, i);
4046
4047		adapter->mac_table[i].state &= ~(IXGBE_MAC_STATE_MODIFIED);
4048	}
4049}
4050#endif
4051
4052static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4053{
4054	struct ixgbe_hw *hw = &adapter->hw;
4055	int i;
4056	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4057		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_MODIFIED) {
4058			if (adapter->mac_table[i].state &
4059			    IXGBE_MAC_STATE_IN_USE)
4060				hw->mac.ops.set_rar(hw, i,
4061						adapter->mac_table[i].addr,
4062						adapter->mac_table[i].queue,
4063						IXGBE_RAH_AV);
4064			else
4065				hw->mac.ops.clear_rar(hw, i);
4066
4067			adapter->mac_table[i].state &=
4068						~(IXGBE_MAC_STATE_MODIFIED);
4069		}
4070	}
4071}
4072
4073static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4074{
4075	int i;
4076	struct ixgbe_hw *hw = &adapter->hw;
4077
4078	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4079		adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
4080		adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
4081		eth_zero_addr(adapter->mac_table[i].addr);
4082		adapter->mac_table[i].queue = 0;
4083	}
4084	ixgbe_sync_mac_table(adapter);
4085}
4086
4087static int ixgbe_available_rars(struct ixgbe_adapter *adapter)
4088{
4089	struct ixgbe_hw *hw = &adapter->hw;
4090	int i, count = 0;
4091
4092	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4093		if (adapter->mac_table[i].state == 0)
4094			count++;
4095	}
4096	return count;
4097}
4098
4099/* this function destroys the first RAR entry */
4100static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter,
4101					 u8 *addr)
4102{
4103	struct ixgbe_hw *hw = &adapter->hw;
4104
4105	memcpy(&adapter->mac_table[0].addr, addr, ETH_ALEN);
4106	adapter->mac_table[0].queue = VMDQ_P(0);
4107	adapter->mac_table[0].state = (IXGBE_MAC_STATE_DEFAULT |
4108				       IXGBE_MAC_STATE_IN_USE);
4109	hw->mac.ops.set_rar(hw, 0, adapter->mac_table[0].addr,
4110			    adapter->mac_table[0].queue,
4111			    IXGBE_RAH_AV);
4112}
4113
4114int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
4115{
4116	struct ixgbe_hw *hw = &adapter->hw;
4117	int i;
4118
4119	if (is_zero_ether_addr(addr))
4120		return -EINVAL;
4121
4122	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4123		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
4124			continue;
4125		adapter->mac_table[i].state |= (IXGBE_MAC_STATE_MODIFIED |
4126						IXGBE_MAC_STATE_IN_USE);
4127		ether_addr_copy(adapter->mac_table[i].addr, addr);
4128		adapter->mac_table[i].queue = queue;
4129		ixgbe_sync_mac_table(adapter);
4130		return i;
4131	}
4132	return -ENOMEM;
4133}
4134
4135int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
4136{
4137	/* search table for addr, if found, set to 0 and sync */
4138	int i;
4139	struct ixgbe_hw *hw = &adapter->hw;
4140
4141	if (is_zero_ether_addr(addr))
4142		return -EINVAL;
4143
4144	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4145		if (ether_addr_equal(addr, adapter->mac_table[i].addr) &&
4146		    adapter->mac_table[i].queue == queue) {
4147			adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
4148			adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
4149			eth_zero_addr(adapter->mac_table[i].addr);
4150			adapter->mac_table[i].queue = 0;
4151			ixgbe_sync_mac_table(adapter);
4152			return 0;
4153		}
4154	}
4155	return -ENOMEM;
4156}
4157/**
4158 * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4159 * @netdev: network interface device structure
4160 *
4161 * Writes unicast address list to the RAR table.
4162 * Returns: -ENOMEM on failure/insufficient address space
4163 *                0 on no addresses written
4164 *                X on writing X addresses to the RAR table
4165 **/
4166static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4167{
4168	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4169	int count = 0;
4170
4171	/* return ENOMEM indicating insufficient memory for addresses */
4172	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter))
4173		return -ENOMEM;
4174
4175	if (!netdev_uc_empty(netdev)) {
4176		struct netdev_hw_addr *ha;
4177		netdev_for_each_uc_addr(ha, netdev) {
4178			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4179			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4180			count++;
4181		}
4182	}
4183	return count;
4184}
4185
4186/**
4187 * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4188 * @netdev: network interface device structure
4189 *
4190 * The set_rx_method entry point is called whenever the unicast/multicast
4191 * address list or the network interface flags are updated.  This routine is
4192 * responsible for configuring the hardware for proper unicast, multicast and
4193 * promiscuous mode.
4194 **/
4195void ixgbe_set_rx_mode(struct net_device *netdev)
4196{
4197	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4198	struct ixgbe_hw *hw = &adapter->hw;
4199	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4200	u32 vlnctrl;
4201	int count;
4202
4203	/* Check for Promiscuous and All Multicast modes */
4204	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4205	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4206
4207	/* set all bits that we expect to always be set */
4208	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4209	fctrl |= IXGBE_FCTRL_BAM;
4210	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4211	fctrl |= IXGBE_FCTRL_PMCF;
4212
4213	/* clear the bits we are changing the status of */
4214	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4215	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4216	if (netdev->flags & IFF_PROMISC) {
4217		hw->addr_ctrl.user_set_promisc = true;
4218		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4219		vmolr |= IXGBE_VMOLR_MPE;
4220		/* Only disable hardware filter vlans in promiscuous mode
4221		 * if SR-IOV and VMDQ are disabled - otherwise ensure
4222		 * that hardware VLAN filters remain enabled.
4223		 */
4224		if (adapter->flags & (IXGBE_FLAG_VMDQ_ENABLED |
4225				      IXGBE_FLAG_SRIOV_ENABLED))
4226			vlnctrl |= (IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4227	} else {
4228		if (netdev->flags & IFF_ALLMULTI) {
4229			fctrl |= IXGBE_FCTRL_MPE;
4230			vmolr |= IXGBE_VMOLR_MPE;
4231		}
4232		vlnctrl |= IXGBE_VLNCTRL_VFE;
4233		hw->addr_ctrl.user_set_promisc = false;
4234	}
4235
4236	/*
4237	 * Write addresses to available RAR registers, if there is not
4238	 * sufficient space to store all the addresses then enable
4239	 * unicast promiscuous mode
4240	 */
4241	count = ixgbe_write_uc_addr_list(netdev, VMDQ_P(0));
4242	if (count < 0) {
4243		fctrl |= IXGBE_FCTRL_UPE;
4244		vmolr |= IXGBE_VMOLR_ROPE;
4245	}
4246
4247	/* Write addresses to the MTA, if the attempt fails
4248	 * then we should just turn on promiscuous mode so
4249	 * that we can at least receive multicast traffic
4250	 */
4251	count = ixgbe_write_mc_addr_list(netdev);
4252	if (count < 0) {
4253		fctrl |= IXGBE_FCTRL_MPE;
4254		vmolr |= IXGBE_VMOLR_MPE;
4255	} else if (count) {
4256		vmolr |= IXGBE_VMOLR_ROMPE;
4257	}
4258
4259	if (hw->mac.type != ixgbe_mac_82598EB) {
4260		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4261			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4262			   IXGBE_VMOLR_ROPE);
4263		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4264	}
4265
4266	/* This is useful for sniffing bad packets. */
4267	if (adapter->netdev->features & NETIF_F_RXALL) {
4268		/* UPE and MPE will be handled by normal PROMISC logic
4269		 * in e1000e_set_rx_mode */
4270		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4271			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4272			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4273
4274		fctrl &= ~(IXGBE_FCTRL_DPF);
4275		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4276	}
4277
4278	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4279	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4280
4281	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
4282		ixgbe_vlan_strip_enable(adapter);
4283	else
4284		ixgbe_vlan_strip_disable(adapter);
4285}
4286
4287static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4288{
4289	int q_idx;
4290
4291	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4292		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
4293		napi_enable(&adapter->q_vector[q_idx]->napi);
4294	}
4295}
4296
4297static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4298{
4299	int q_idx;
4300
4301	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4302		napi_disable(&adapter->q_vector[q_idx]->napi);
4303		while (!ixgbe_qv_disable(adapter->q_vector[q_idx])) {
4304			pr_info("QV %d locked\n", q_idx);
4305			usleep_range(1000, 20000);
4306		}
4307	}
4308}
4309
4310static void ixgbe_clear_vxlan_port(struct ixgbe_adapter *adapter)
4311{
4312	switch (adapter->hw.mac.type) {
4313	case ixgbe_mac_X550:
4314	case ixgbe_mac_X550EM_x:
4315		IXGBE_WRITE_REG(&adapter->hw, IXGBE_VXLANCTRL, 0);
4316#ifdef CONFIG_IXGBE_VXLAN
4317		adapter->vxlan_port = 0;
4318#endif
4319		break;
4320	default:
4321		break;
4322	}
4323}
4324
4325#ifdef CONFIG_IXGBE_DCB
4326/**
4327 * ixgbe_configure_dcb - Configure DCB hardware
4328 * @adapter: ixgbe adapter struct
4329 *
4330 * This is called by the driver on open to configure the DCB hardware.
4331 * This is also called by the gennetlink interface when reconfiguring
4332 * the DCB state.
4333 */
4334static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4335{
4336	struct ixgbe_hw *hw = &adapter->hw;
4337	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4338
4339	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4340		if (hw->mac.type == ixgbe_mac_82598EB)
4341			netif_set_gso_max_size(adapter->netdev, 65536);
4342		return;
4343	}
4344
4345	if (hw->mac.type == ixgbe_mac_82598EB)
4346		netif_set_gso_max_size(adapter->netdev, 32768);
4347
4348#ifdef IXGBE_FCOE
4349	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4350		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4351#endif
4352
4353	/* reconfigure the hardware */
4354	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4355		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4356						DCB_TX_CONFIG);
4357		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4358						DCB_RX_CONFIG);
4359		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4360	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4361		ixgbe_dcb_hw_ets(&adapter->hw,
4362				 adapter->ixgbe_ieee_ets,
4363				 max_frame);
4364		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4365					adapter->ixgbe_ieee_pfc->pfc_en,
4366					adapter->ixgbe_ieee_ets->prio_tc);
4367	}
4368
4369	/* Enable RSS Hash per TC */
4370	if (hw->mac.type != ixgbe_mac_82598EB) {
4371		u32 msb = 0;
4372		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4373
4374		while (rss_i) {
4375			msb++;
4376			rss_i >>= 1;
4377		}
4378
4379		/* write msb to all 8 TCs in one write */
4380		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4381	}
4382}
4383#endif
4384
4385/* Additional bittime to account for IXGBE framing */
4386#define IXGBE_ETH_FRAMING 20
4387
4388/**
4389 * ixgbe_hpbthresh - calculate high water mark for flow control
4390 *
4391 * @adapter: board private structure to calculate for
4392 * @pb: packet buffer to calculate
4393 */
4394static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4395{
4396	struct ixgbe_hw *hw = &adapter->hw;
4397	struct net_device *dev = adapter->netdev;
4398	int link, tc, kb, marker;
4399	u32 dv_id, rx_pba;
4400
4401	/* Calculate max LAN frame size */
4402	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4403
4404#ifdef IXGBE_FCOE
4405	/* FCoE traffic class uses FCOE jumbo frames */
4406	if ((dev->features & NETIF_F_FCOE_MTU) &&
4407	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4408	    (pb == ixgbe_fcoe_get_tc(adapter)))
4409		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4410#endif
4411
4412	/* Calculate delay value for device */
4413	switch (hw->mac.type) {
4414	case ixgbe_mac_X540:
4415	case ixgbe_mac_X550:
4416	case ixgbe_mac_X550EM_x:
4417		dv_id = IXGBE_DV_X540(link, tc);
4418		break;
4419	default:
4420		dv_id = IXGBE_DV(link, tc);
4421		break;
4422	}
4423
4424	/* Loopback switch introduces additional latency */
4425	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4426		dv_id += IXGBE_B2BT(tc);
4427
4428	/* Delay value is calculated in bit times convert to KB */
4429	kb = IXGBE_BT2KB(dv_id);
4430	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4431
4432	marker = rx_pba - kb;
4433
4434	/* It is possible that the packet buffer is not large enough
4435	 * to provide required headroom. In this case throw an error
4436	 * to user and a do the best we can.
4437	 */
4438	if (marker < 0) {
4439		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4440			    "headroom to support flow control."
4441			    "Decrease MTU or number of traffic classes\n", pb);
4442		marker = tc + 1;
4443	}
4444
4445	return marker;
4446}
4447
4448/**
4449 * ixgbe_lpbthresh - calculate low water mark for for flow control
4450 *
4451 * @adapter: board private structure to calculate for
4452 * @pb: packet buffer to calculate
4453 */
4454static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4455{
4456	struct ixgbe_hw *hw = &adapter->hw;
4457	struct net_device *dev = adapter->netdev;
4458	int tc;
4459	u32 dv_id;
4460
4461	/* Calculate max LAN frame size */
4462	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4463
4464#ifdef IXGBE_FCOE
4465	/* FCoE traffic class uses FCOE jumbo frames */
4466	if ((dev->features & NETIF_F_FCOE_MTU) &&
4467	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4468	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4469		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4470#endif
4471
4472	/* Calculate delay value for device */
4473	switch (hw->mac.type) {
4474	case ixgbe_mac_X540:
4475	case ixgbe_mac_X550:
4476	case ixgbe_mac_X550EM_x:
4477		dv_id = IXGBE_LOW_DV_X540(tc);
4478		break;
4479	default:
4480		dv_id = IXGBE_LOW_DV(tc);
4481		break;
4482	}
4483
4484	/* Delay value is calculated in bit times convert to KB */
4485	return IXGBE_BT2KB(dv_id);
4486}
4487
4488/*
4489 * ixgbe_pbthresh_setup - calculate and setup high low water marks
4490 */
4491static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4492{
4493	struct ixgbe_hw *hw = &adapter->hw;
4494	int num_tc = netdev_get_num_tc(adapter->netdev);
4495	int i;
4496
4497	if (!num_tc)
4498		num_tc = 1;
4499
4500	for (i = 0; i < num_tc; i++) {
4501		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4502		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4503
4504		/* Low water marks must not be larger than high water marks */
4505		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4506			hw->fc.low_water[i] = 0;
4507	}
4508
4509	for (; i < MAX_TRAFFIC_CLASS; i++)
4510		hw->fc.high_water[i] = 0;
4511}
4512
4513static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4514{
4515	struct ixgbe_hw *hw = &adapter->hw;
4516	int hdrm;
4517	u8 tc = netdev_get_num_tc(adapter->netdev);
4518
4519	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4520	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4521		hdrm = 32 << adapter->fdir_pballoc;
4522	else
4523		hdrm = 0;
4524
4525	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4526	ixgbe_pbthresh_setup(adapter);
4527}
4528
4529static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4530{
4531	struct ixgbe_hw *hw = &adapter->hw;
4532	struct hlist_node *node2;
4533	struct ixgbe_fdir_filter *filter;
4534
4535	spin_lock(&adapter->fdir_perfect_lock);
4536
4537	if (!hlist_empty(&adapter->fdir_filter_list))
4538		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4539
4540	hlist_for_each_entry_safe(filter, node2,
4541				  &adapter->fdir_filter_list, fdir_node) {
4542		ixgbe_fdir_write_perfect_filter_82599(hw,
4543				&filter->filter,
4544				filter->sw_idx,
4545				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4546				IXGBE_FDIR_DROP_QUEUE :
4547				adapter->rx_ring[filter->action]->reg_idx);
4548	}
4549
4550	spin_unlock(&adapter->fdir_perfect_lock);
4551}
4552
4553static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4554				      struct ixgbe_adapter *adapter)
4555{
4556	struct ixgbe_hw *hw = &adapter->hw;
4557	u32 vmolr;
4558
4559	/* No unicast promiscuous support for VMDQ devices. */
4560	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4561	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4562
4563	/* clear the affected bit */
4564	vmolr &= ~IXGBE_VMOLR_MPE;
4565
4566	if (dev->flags & IFF_ALLMULTI) {
4567		vmolr |= IXGBE_VMOLR_MPE;
4568	} else {
4569		vmolr |= IXGBE_VMOLR_ROMPE;
4570		hw->mac.ops.update_mc_addr_list(hw, dev);
4571	}
4572	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4573	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4574}
4575
4576static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4577{
4578	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4579	int rss_i = adapter->num_rx_queues_per_pool;
4580	struct ixgbe_hw *hw = &adapter->hw;
4581	u16 pool = vadapter->pool;
4582	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4583		      IXGBE_PSRTYPE_UDPHDR |
4584		      IXGBE_PSRTYPE_IPV4HDR |
4585		      IXGBE_PSRTYPE_L2HDR |
4586		      IXGBE_PSRTYPE_IPV6HDR;
4587
4588	if (hw->mac.type == ixgbe_mac_82598EB)
4589		return;
4590
4591	if (rss_i > 3)
4592		psrtype |= 2 << 29;
4593	else if (rss_i > 1)
4594		psrtype |= 1 << 29;
4595
4596	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4597}
4598
4599/**
4600 * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4601 * @rx_ring: ring to free buffers from
4602 **/
4603static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4604{
4605	struct device *dev = rx_ring->dev;
4606	unsigned long size;
4607	u16 i;
4608
4609	/* ring already cleared, nothing to do */
4610	if (!rx_ring->rx_buffer_info)
4611		return;
4612
4613	/* Free all the Rx ring sk_buffs */
4614	for (i = 0; i < rx_ring->count; i++) {
4615		struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4616
4617		if (rx_buffer->skb) {
4618			struct sk_buff *skb = rx_buffer->skb;
4619			if (IXGBE_CB(skb)->page_released)
4620				dma_unmap_page(dev,
4621					       IXGBE_CB(skb)->dma,
4622					       ixgbe_rx_bufsz(rx_ring),
4623					       DMA_FROM_DEVICE);
4624			dev_kfree_skb(skb);
4625			rx_buffer->skb = NULL;
4626		}
4627
4628		if (!rx_buffer->page)
4629			continue;
4630
4631		dma_unmap_page(dev, rx_buffer->dma,
4632			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
4633		__free_pages(rx_buffer->page, ixgbe_rx_pg_order(rx_ring));
4634
4635		rx_buffer->page = NULL;
4636	}
4637
4638	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4639	memset(rx_ring->rx_buffer_info, 0, size);
4640
4641	/* Zero out the descriptor ring */
4642	memset(rx_ring->desc, 0, rx_ring->size);
4643
4644	rx_ring->next_to_alloc = 0;
4645	rx_ring->next_to_clean = 0;
4646	rx_ring->next_to_use = 0;
4647}
4648
4649static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4650				   struct ixgbe_ring *rx_ring)
4651{
4652	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4653	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4654
4655	/* shutdown specific queue receive and wait for dma to settle */
4656	ixgbe_disable_rx_queue(adapter, rx_ring);
4657	usleep_range(10000, 20000);
4658	ixgbe_irq_disable_queues(adapter, ((u64)1 << index));
4659	ixgbe_clean_rx_ring(rx_ring);
4660	rx_ring->l2_accel_priv = NULL;
4661}
4662
4663static int ixgbe_fwd_ring_down(struct net_device *vdev,
4664			       struct ixgbe_fwd_adapter *accel)
4665{
4666	struct ixgbe_adapter *adapter = accel->real_adapter;
4667	unsigned int rxbase = accel->rx_base_queue;
4668	unsigned int txbase = accel->tx_base_queue;
4669	int i;
4670
4671	netif_tx_stop_all_queues(vdev);
4672
4673	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4674		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4675		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4676	}
4677
4678	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4679		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4680		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4681	}
4682
4683
4684	return 0;
4685}
4686
4687static int ixgbe_fwd_ring_up(struct net_device *vdev,
4688			     struct ixgbe_fwd_adapter *accel)
4689{
4690	struct ixgbe_adapter *adapter = accel->real_adapter;
4691	unsigned int rxbase, txbase, queues;
4692	int i, baseq, err = 0;
4693
4694	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4695		return 0;
4696
4697	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4698	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4699		   accel->pool, adapter->num_rx_pools,
4700		   baseq, baseq + adapter->num_rx_queues_per_pool,
4701		   adapter->fwd_bitmask);
4702
4703	accel->netdev = vdev;
4704	accel->rx_base_queue = rxbase = baseq;
4705	accel->tx_base_queue = txbase = baseq;
4706
4707	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4708		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4709
4710	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4711		adapter->rx_ring[rxbase + i]->netdev = vdev;
4712		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
4713		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
4714	}
4715
4716	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4717		adapter->tx_ring[txbase + i]->netdev = vdev;
4718		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
4719	}
4720
4721	queues = min_t(unsigned int,
4722		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
4723	err = netif_set_real_num_tx_queues(vdev, queues);
4724	if (err)
4725		goto fwd_queue_err;
4726
4727	err = netif_set_real_num_rx_queues(vdev, queues);
4728	if (err)
4729		goto fwd_queue_err;
4730
4731	if (is_valid_ether_addr(vdev->dev_addr))
4732		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
4733
4734	ixgbe_fwd_psrtype(accel);
4735	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
4736	return err;
4737fwd_queue_err:
4738	ixgbe_fwd_ring_down(vdev, accel);
4739	return err;
4740}
4741
4742static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
4743{
4744	struct net_device *upper;
4745	struct list_head *iter;
4746	int err;
4747
4748	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
4749		if (netif_is_macvlan(upper)) {
4750			struct macvlan_dev *dfwd = netdev_priv(upper);
4751			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
4752
4753			if (dfwd->fwd_priv) {
4754				err = ixgbe_fwd_ring_up(upper, vadapter);
4755				if (err)
4756					continue;
4757			}
4758		}
4759	}
4760}
4761
4762static void ixgbe_configure(struct ixgbe_adapter *adapter)
4763{
4764	struct ixgbe_hw *hw = &adapter->hw;
4765
4766	ixgbe_configure_pb(adapter);
4767#ifdef CONFIG_IXGBE_DCB
4768	ixgbe_configure_dcb(adapter);
4769#endif
4770	/*
4771	 * We must restore virtualization before VLANs or else
4772	 * the VLVF registers will not be populated
4773	 */
4774	ixgbe_configure_virtualization(adapter);
4775
4776	ixgbe_set_rx_mode(adapter->netdev);
4777	ixgbe_restore_vlan(adapter);
4778
4779	switch (hw->mac.type) {
4780	case ixgbe_mac_82599EB:
4781	case ixgbe_mac_X540:
4782		hw->mac.ops.disable_rx_buff(hw);
4783		break;
4784	default:
4785		break;
4786	}
4787
4788	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
4789		ixgbe_init_fdir_signature_82599(&adapter->hw,
4790						adapter->fdir_pballoc);
4791	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
4792		ixgbe_init_fdir_perfect_82599(&adapter->hw,
4793					      adapter->fdir_pballoc);
4794		ixgbe_fdir_filter_restore(adapter);
4795	}
4796
4797	switch (hw->mac.type) {
4798	case ixgbe_mac_82599EB:
4799	case ixgbe_mac_X540:
4800		hw->mac.ops.enable_rx_buff(hw);
4801		break;
4802	default:
4803		break;
4804	}
4805
4806#ifdef CONFIG_IXGBE_DCA
4807	/* configure DCA */
4808	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
4809		ixgbe_setup_dca(adapter);
4810#endif /* CONFIG_IXGBE_DCA */
4811
4812#ifdef IXGBE_FCOE
4813	/* configure FCoE L2 filters, redirection table, and Rx control */
4814	ixgbe_configure_fcoe(adapter);
4815
4816#endif /* IXGBE_FCOE */
4817	ixgbe_configure_tx(adapter);
4818	ixgbe_configure_rx(adapter);
4819	ixgbe_configure_dfwd(adapter);
4820}
4821
4822/**
4823 * ixgbe_sfp_link_config - set up SFP+ link
4824 * @adapter: pointer to private adapter struct
4825 **/
4826static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
4827{
4828	/*
4829	 * We are assuming the worst case scenario here, and that
4830	 * is that an SFP was inserted/removed after the reset
4831	 * but before SFP detection was enabled.  As such the best
4832	 * solution is to just start searching as soon as we start
4833	 */
4834	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
4835		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
4836
4837	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
4838	adapter->sfp_poll_time = 0;
4839}
4840
4841/**
4842 * ixgbe_non_sfp_link_config - set up non-SFP+ link
4843 * @hw: pointer to private hardware struct
4844 *
4845 * Returns 0 on success, negative on failure
4846 **/
4847static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
4848{
4849	u32 speed;
4850	bool autoneg, link_up = false;
4851	int ret = IXGBE_ERR_LINK_SETUP;
4852
4853	if (hw->mac.ops.check_link)
4854		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
4855
4856	if (ret)
4857		return ret;
4858
4859	speed = hw->phy.autoneg_advertised;
4860	if ((!speed) && (hw->mac.ops.get_link_capabilities))
4861		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
4862							&autoneg);
4863	if (ret)
4864		return ret;
4865
4866	if (hw->mac.ops.setup_link)
4867		ret = hw->mac.ops.setup_link(hw, speed, link_up);
4868
4869	return ret;
4870}
4871
4872static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
4873{
4874	struct ixgbe_hw *hw = &adapter->hw;
4875	u32 gpie = 0;
4876
4877	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
4878		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4879		       IXGBE_GPIE_OCD;
4880		gpie |= IXGBE_GPIE_EIAME;
4881		/*
4882		 * use EIAM to auto-mask when MSI-X interrupt is asserted
4883		 * this saves a register write for every interrupt
4884		 */
4885		switch (hw->mac.type) {
4886		case ixgbe_mac_82598EB:
4887			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4888			break;
4889		case ixgbe_mac_82599EB:
4890		case ixgbe_mac_X540:
4891		case ixgbe_mac_X550:
4892		case ixgbe_mac_X550EM_x:
4893		default:
4894			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4895			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4896			break;
4897		}
4898	} else {
4899		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
4900		 * specifically only auto mask tx and rx interrupts */
4901		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4902	}
4903
4904	/* XXX: to interrupt immediately for EICS writes, enable this */
4905	/* gpie |= IXGBE_GPIE_EIMEN; */
4906
4907	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4908		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
4909
4910		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4911		case IXGBE_82599_VMDQ_8Q_MASK:
4912			gpie |= IXGBE_GPIE_VTMODE_16;
4913			break;
4914		case IXGBE_82599_VMDQ_4Q_MASK:
4915			gpie |= IXGBE_GPIE_VTMODE_32;
4916			break;
4917		default:
4918			gpie |= IXGBE_GPIE_VTMODE_64;
4919			break;
4920		}
4921	}
4922
4923	/* Enable Thermal over heat sensor interrupt */
4924	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
4925		switch (adapter->hw.mac.type) {
4926		case ixgbe_mac_82599EB:
4927			gpie |= IXGBE_SDP0_GPIEN_8259X;
4928			break;
4929		default:
4930			break;
4931		}
4932	}
4933
4934	/* Enable fan failure interrupt */
4935	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
4936		gpie |= IXGBE_SDP1_GPIEN(hw);
4937
4938	switch (hw->mac.type) {
4939	case ixgbe_mac_82599EB:
4940		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
4941		break;
4942	case ixgbe_mac_X550EM_x:
4943		gpie |= IXGBE_SDP0_GPIEN_X540;
4944		break;
4945	default:
4946		break;
4947	}
4948
4949	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4950}
4951
4952static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4953{
4954	struct ixgbe_hw *hw = &adapter->hw;
4955	int err;
4956	u32 ctrl_ext;
4957
4958	ixgbe_get_hw_control(adapter);
4959	ixgbe_setup_gpie(adapter);
4960
4961	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4962		ixgbe_configure_msix(adapter);
4963	else
4964		ixgbe_configure_msi_and_legacy(adapter);
4965
4966	/* enable the optics for 82599 SFP+ fiber */
4967	if (hw->mac.ops.enable_tx_laser)
4968		hw->mac.ops.enable_tx_laser(hw);
4969
4970	if (hw->phy.ops.set_phy_power)
4971		hw->phy.ops.set_phy_power(hw, true);
4972
4973	smp_mb__before_atomic();
4974	clear_bit(__IXGBE_DOWN, &adapter->state);
4975	ixgbe_napi_enable_all(adapter);
4976
4977	if (ixgbe_is_sfp(hw)) {
4978		ixgbe_sfp_link_config(adapter);
4979	} else {
4980		err = ixgbe_non_sfp_link_config(hw);
4981		if (err)
4982			e_err(probe, "link_config FAILED %d\n", err);
4983	}
4984
4985	/* clear any pending interrupts, may auto mask */
4986	IXGBE_READ_REG(hw, IXGBE_EICR);
4987	ixgbe_irq_enable(adapter, true, true);
4988
4989	/*
4990	 * If this adapter has a fan, check to see if we had a failure
4991	 * before we enabled the interrupt.
4992	 */
4993	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4994		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4995		if (esdp & IXGBE_ESDP_SDP1)
4996			e_crit(drv, "Fan has stopped, replace the adapter\n");
4997	}
4998
4999	/* bring the link up in the watchdog, this could race with our first
5000	 * link up interrupt but shouldn't be a problem */
5001	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5002	adapter->link_check_timeout = jiffies;
5003	mod_timer(&adapter->service_timer, jiffies);
5004
5005	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5006	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5007	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5008	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5009}
5010
5011void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5012{
5013	WARN_ON(in_interrupt());
5014	/* put off any impending NetWatchDogTimeout */
5015	adapter->netdev->trans_start = jiffies;
5016
5017	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5018		usleep_range(1000, 2000);
5019	ixgbe_down(adapter);
5020	/*
5021	 * If SR-IOV enabled then wait a bit before bringing the adapter
5022	 * back up to give the VFs time to respond to the reset.  The
5023	 * two second wait is based upon the watchdog timer cycle in
5024	 * the VF driver.
5025	 */
5026	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5027		msleep(2000);
5028	ixgbe_up(adapter);
5029	clear_bit(__IXGBE_RESETTING, &adapter->state);
5030}
5031
5032void ixgbe_up(struct ixgbe_adapter *adapter)
5033{
5034	/* hardware has been reset, we need to reload some things */
5035	ixgbe_configure(adapter);
5036
5037	ixgbe_up_complete(adapter);
5038}
5039
5040void ixgbe_reset(struct ixgbe_adapter *adapter)
5041{
5042	struct ixgbe_hw *hw = &adapter->hw;
5043	struct net_device *netdev = adapter->netdev;
5044	int err;
5045	u8 old_addr[ETH_ALEN];
5046
5047	if (ixgbe_removed(hw->hw_addr))
5048		return;
5049	/* lock SFP init bit to prevent race conditions with the watchdog */
5050	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5051		usleep_range(1000, 2000);
5052
5053	/* clear all SFP and link config related flags while holding SFP_INIT */
5054	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5055			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5056	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5057
5058	err = hw->mac.ops.init_hw(hw);
5059	switch (err) {
5060	case 0:
5061	case IXGBE_ERR_SFP_NOT_PRESENT:
5062	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5063		break;
5064	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5065		e_dev_err("master disable timed out\n");
5066		break;
5067	case IXGBE_ERR_EEPROM_VERSION:
5068		/* We are running on a pre-production device, log a warning */
5069		e_dev_warn("This device is a pre-production adapter/LOM. "
5070			   "Please be aware there may be issues associated with "
5071			   "your hardware.  If you are experiencing problems "
5072			   "please contact your Intel or hardware "
5073			   "representative who provided you with this "
5074			   "hardware.\n");
5075		break;
5076	default:
5077		e_dev_err("Hardware Error: %d\n", err);
5078	}
5079
5080	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5081	/* do not flush user set addresses */
5082	memcpy(old_addr, &adapter->mac_table[0].addr, netdev->addr_len);
5083	ixgbe_flush_sw_mac_table(adapter);
5084	ixgbe_mac_set_default_filter(adapter, old_addr);
5085
5086	/* update SAN MAC vmdq pool selection */
5087	if (hw->mac.san_mac_rar_index)
5088		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5089
5090	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5091		ixgbe_ptp_reset(adapter);
5092
5093	if (hw->phy.ops.set_phy_power) {
5094		if (!netif_running(adapter->netdev) && !adapter->wol)
5095			hw->phy.ops.set_phy_power(hw, false);
5096		else
5097			hw->phy.ops.set_phy_power(hw, true);
5098	}
5099}
5100
5101/**
5102 * ixgbe_clean_tx_ring - Free Tx Buffers
5103 * @tx_ring: ring to be cleaned
5104 **/
5105static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5106{
5107	struct ixgbe_tx_buffer *tx_buffer_info;
5108	unsigned long size;
5109	u16 i;
5110
5111	/* ring already cleared, nothing to do */
5112	if (!tx_ring->tx_buffer_info)
5113		return;
5114
5115	/* Free all the Tx ring sk_buffs */
5116	for (i = 0; i < tx_ring->count; i++) {
5117		tx_buffer_info = &tx_ring->tx_buffer_info[i];
5118		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
5119	}
5120
5121	netdev_tx_reset_queue(txring_txq(tx_ring));
5122
5123	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5124	memset(tx_ring->tx_buffer_info, 0, size);
5125
5126	/* Zero out the descriptor ring */
5127	memset(tx_ring->desc, 0, tx_ring->size);
5128
5129	tx_ring->next_to_use = 0;
5130	tx_ring->next_to_clean = 0;
5131}
5132
5133/**
5134 * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5135 * @adapter: board private structure
5136 **/
5137static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5138{
5139	int i;
5140
5141	for (i = 0; i < adapter->num_rx_queues; i++)
5142		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5143}
5144
5145/**
5146 * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5147 * @adapter: board private structure
5148 **/
5149static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5150{
5151	int i;
5152
5153	for (i = 0; i < adapter->num_tx_queues; i++)
5154		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5155}
5156
5157static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5158{
5159	struct hlist_node *node2;
5160	struct ixgbe_fdir_filter *filter;
5161
5162	spin_lock(&adapter->fdir_perfect_lock);
5163
5164	hlist_for_each_entry_safe(filter, node2,
5165				  &adapter->fdir_filter_list, fdir_node) {
5166		hlist_del(&filter->fdir_node);
5167		kfree(filter);
5168	}
5169	adapter->fdir_filter_count = 0;
5170
5171	spin_unlock(&adapter->fdir_perfect_lock);
5172}
5173
5174void ixgbe_down(struct ixgbe_adapter *adapter)
5175{
5176	struct net_device *netdev = adapter->netdev;
5177	struct ixgbe_hw *hw = &adapter->hw;
5178	struct net_device *upper;
5179	struct list_head *iter;
5180	int i;
5181
5182	/* signal that we are down to the interrupt handler */
5183	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5184		return; /* do nothing if already down */
5185
5186	/* disable receives */
5187	hw->mac.ops.disable_rx(hw);
5188
5189	/* disable all enabled rx queues */
5190	for (i = 0; i < adapter->num_rx_queues; i++)
5191		/* this call also flushes the previous write */
5192		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5193
5194	usleep_range(10000, 20000);
5195
5196	netif_tx_stop_all_queues(netdev);
5197
5198	/* call carrier off first to avoid false dev_watchdog timeouts */
5199	netif_carrier_off(netdev);
5200	netif_tx_disable(netdev);
5201
5202	/* disable any upper devices */
5203	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5204		if (netif_is_macvlan(upper)) {
5205			struct macvlan_dev *vlan = netdev_priv(upper);
5206
5207			if (vlan->fwd_priv) {
5208				netif_tx_stop_all_queues(upper);
5209				netif_carrier_off(upper);
5210				netif_tx_disable(upper);
5211			}
5212		}
5213	}
5214
5215	ixgbe_irq_disable(adapter);
5216
5217	ixgbe_napi_disable_all(adapter);
5218
5219	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
5220			     IXGBE_FLAG2_RESET_REQUESTED);
5221	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5222
5223	del_timer_sync(&adapter->service_timer);
5224
5225	if (adapter->num_vfs) {
5226		/* Clear EITR Select mapping */
5227		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5228
5229		/* Mark all the VFs as inactive */
5230		for (i = 0 ; i < adapter->num_vfs; i++)
5231			adapter->vfinfo[i].clear_to_send = false;
5232
5233		/* ping all the active vfs to let them know we are going down */
5234		ixgbe_ping_all_vfs(adapter);
5235
5236		/* Disable all VFTE/VFRE TX/RX */
5237		ixgbe_disable_tx_rx(adapter);
5238	}
5239
5240	/* disable transmits in the hardware now that interrupts are off */
5241	for (i = 0; i < adapter->num_tx_queues; i++) {
5242		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5243		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5244	}
5245
5246	/* Disable the Tx DMA engine on 82599 and later MAC */
5247	switch (hw->mac.type) {
5248	case ixgbe_mac_82599EB:
5249	case ixgbe_mac_X540:
5250	case ixgbe_mac_X550:
5251	case ixgbe_mac_X550EM_x:
5252		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5253				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5254				 ~IXGBE_DMATXCTL_TE));
5255		break;
5256	default:
5257		break;
5258	}
5259
5260	if (!pci_channel_offline(adapter->pdev))
5261		ixgbe_reset(adapter);
5262
5263	/* power down the optics for 82599 SFP+ fiber */
5264	if (hw->mac.ops.disable_tx_laser)
5265		hw->mac.ops.disable_tx_laser(hw);
5266
5267	ixgbe_clean_all_tx_rings(adapter);
5268	ixgbe_clean_all_rx_rings(adapter);
5269}
5270
5271/**
5272 * ixgbe_tx_timeout - Respond to a Tx Hang
5273 * @netdev: network interface device structure
5274 **/
5275static void ixgbe_tx_timeout(struct net_device *netdev)
5276{
5277	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5278
5279	/* Do the reset outside of interrupt context */
5280	ixgbe_tx_timeout_reset(adapter);
5281}
5282
5283/**
5284 * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5285 * @adapter: board private structure to initialize
5286 *
5287 * ixgbe_sw_init initializes the Adapter private data structure.
5288 * Fields are initialized based on PCI device information and
5289 * OS network device settings (MTU size).
5290 **/
5291static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
5292{
5293	struct ixgbe_hw *hw = &adapter->hw;
5294	struct pci_dev *pdev = adapter->pdev;
5295	unsigned int rss, fdir;
5296	u32 fwsm;
5297#ifdef CONFIG_IXGBE_DCB
5298	int j;
5299	struct tc_configuration *tc;
5300#endif
5301
5302	/* PCI config space info */
5303
5304	hw->vendor_id = pdev->vendor;
5305	hw->device_id = pdev->device;
5306	hw->revision_id = pdev->revision;
5307	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5308	hw->subsystem_device_id = pdev->subsystem_device;
5309
5310	/* Set common capability flags and settings */
5311	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5312	adapter->ring_feature[RING_F_RSS].limit = rss;
5313	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5314	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5315	adapter->atr_sample_rate = 20;
5316	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5317	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5318	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5319#ifdef CONFIG_IXGBE_DCA
5320	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5321#endif
5322#ifdef IXGBE_FCOE
5323	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5324	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5325#ifdef CONFIG_IXGBE_DCB
5326	/* Default traffic class to use for FCoE */
5327	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5328#endif /* CONFIG_IXGBE_DCB */
5329#endif /* IXGBE_FCOE */
5330
5331	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5332				     hw->mac.num_rar_entries,
5333				     GFP_ATOMIC);
5334
5335	/* Set MAC specific capability flags and exceptions */
5336	switch (hw->mac.type) {
5337	case ixgbe_mac_82598EB:
5338		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5339
5340		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5341			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5342
5343		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5344		adapter->ring_feature[RING_F_FDIR].limit = 0;
5345		adapter->atr_sample_rate = 0;
5346		adapter->fdir_pballoc = 0;
5347#ifdef IXGBE_FCOE
5348		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5349		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5350#ifdef CONFIG_IXGBE_DCB
5351		adapter->fcoe.up = 0;
5352#endif /* IXGBE_DCB */
5353#endif /* IXGBE_FCOE */
5354		break;
5355	case ixgbe_mac_82599EB:
5356		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5357			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5358		break;
5359	case ixgbe_mac_X540:
5360		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
5361		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5362			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5363		break;
5364	case ixgbe_mac_X550EM_x:
5365	case ixgbe_mac_X550:
5366#ifdef CONFIG_IXGBE_DCA
5367		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5368#endif
5369#ifdef CONFIG_IXGBE_VXLAN
5370		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
5371#endif
5372		break;
5373	default:
5374		break;
5375	}
5376
5377#ifdef IXGBE_FCOE
5378	/* FCoE support exists, always init the FCoE lock */
5379	spin_lock_init(&adapter->fcoe.lock);
5380
5381#endif
5382	/* n-tuple support exists, always init our spinlock */
5383	spin_lock_init(&adapter->fdir_perfect_lock);
5384
5385#ifdef CONFIG_IXGBE_DCB
5386	switch (hw->mac.type) {
5387	case ixgbe_mac_X540:
5388	case ixgbe_mac_X550:
5389	case ixgbe_mac_X550EM_x:
5390		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5391		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5392		break;
5393	default:
5394		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5395		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5396		break;
5397	}
5398
5399	/* Configure DCB traffic classes */
5400	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5401		tc = &adapter->dcb_cfg.tc_config[j];
5402		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5403		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5404		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5405		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5406		tc->dcb_pfc = pfc_disabled;
5407	}
5408
5409	/* Initialize default user to priority mapping, UPx->TC0 */
5410	tc = &adapter->dcb_cfg.tc_config[0];
5411	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5412	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5413
5414	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5415	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5416	adapter->dcb_cfg.pfc_mode_enable = false;
5417	adapter->dcb_set_bitmap = 0x00;
5418	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5419	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5420	       sizeof(adapter->temp_dcb_cfg));
5421
5422#endif
5423
5424	/* default flow control settings */
5425	hw->fc.requested_mode = ixgbe_fc_full;
5426	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5427	ixgbe_pbthresh_setup(adapter);
5428	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5429	hw->fc.send_xon = true;
5430	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5431
5432#ifdef CONFIG_PCI_IOV
5433	if (max_vfs > 0)
5434		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5435
5436	/* assign number of SR-IOV VFs */
5437	if (hw->mac.type != ixgbe_mac_82598EB) {
5438		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5439			adapter->num_vfs = 0;
5440			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5441		} else {
5442			adapter->num_vfs = max_vfs;
5443		}
5444	}
5445#endif /* CONFIG_PCI_IOV */
5446
5447	/* enable itr by default in dynamic mode */
5448	adapter->rx_itr_setting = 1;
5449	adapter->tx_itr_setting = 1;
5450
5451	/* set default ring sizes */
5452	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5453	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5454
5455	/* set default work limits */
5456	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5457
5458	/* initialize eeprom parameters */
5459	if (ixgbe_init_eeprom_params_generic(hw)) {
5460		e_dev_err("EEPROM initialization failed\n");
5461		return -EIO;
5462	}
5463
5464	/* PF holds first pool slot */
5465	set_bit(0, &adapter->fwd_bitmask);
5466	set_bit(__IXGBE_DOWN, &adapter->state);
5467
5468	return 0;
5469}
5470
5471/**
5472 * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5473 * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5474 *
5475 * Return 0 on success, negative on failure
5476 **/
5477int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5478{
5479	struct device *dev = tx_ring->dev;
5480	int orig_node = dev_to_node(dev);
5481	int ring_node = -1;
5482	int size;
5483
5484	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5485
5486	if (tx_ring->q_vector)
5487		ring_node = tx_ring->q_vector->numa_node;
5488
5489	tx_ring->tx_buffer_info = vzalloc_node(size, ring_node);
5490	if (!tx_ring->tx_buffer_info)
5491		tx_ring->tx_buffer_info = vzalloc(size);
5492	if (!tx_ring->tx_buffer_info)
5493		goto err;
5494
5495	u64_stats_init(&tx_ring->syncp);
5496
5497	/* round up to nearest 4K */
5498	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5499	tx_ring->size = ALIGN(tx_ring->size, 4096);
5500
5501	set_dev_node(dev, ring_node);
5502	tx_ring->desc = dma_alloc_coherent(dev,
5503					   tx_ring->size,
5504					   &tx_ring->dma,
5505					   GFP_KERNEL);
5506	set_dev_node(dev, orig_node);
5507	if (!tx_ring->desc)
5508		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5509						   &tx_ring->dma, GFP_KERNEL);
5510	if (!tx_ring->desc)
5511		goto err;
5512
5513	tx_ring->next_to_use = 0;
5514	tx_ring->next_to_clean = 0;
5515	return 0;
5516
5517err:
5518	vfree(tx_ring->tx_buffer_info);
5519	tx_ring->tx_buffer_info = NULL;
5520	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5521	return -ENOMEM;
5522}
5523
5524/**
5525 * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5526 * @adapter: board private structure
5527 *
5528 * If this function returns with an error, then it's possible one or
5529 * more of the rings is populated (while the rest are not).  It is the
5530 * callers duty to clean those orphaned rings.
5531 *
5532 * Return 0 on success, negative on failure
5533 **/
5534static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5535{
5536	int i, err = 0;
5537
5538	for (i = 0; i < adapter->num_tx_queues; i++) {
5539		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5540		if (!err)
5541			continue;
5542
5543		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5544		goto err_setup_tx;
5545	}
5546
5547	return 0;
5548err_setup_tx:
5549	/* rewind the index freeing the rings as we go */
5550	while (i--)
5551		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5552	return err;
5553}
5554
5555/**
5556 * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5557 * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5558 *
5559 * Returns 0 on success, negative on failure
5560 **/
5561int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5562{
5563	struct device *dev = rx_ring->dev;
5564	int orig_node = dev_to_node(dev);
5565	int ring_node = -1;
5566	int size;
5567
5568	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5569
5570	if (rx_ring->q_vector)
5571		ring_node = rx_ring->q_vector->numa_node;
5572
5573	rx_ring->rx_buffer_info = vzalloc_node(size, ring_node);
5574	if (!rx_ring->rx_buffer_info)
5575		rx_ring->rx_buffer_info = vzalloc(size);
5576	if (!rx_ring->rx_buffer_info)
5577		goto err;
5578
5579	u64_stats_init(&rx_ring->syncp);
5580
5581	/* Round up to nearest 4K */
5582	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5583	rx_ring->size = ALIGN(rx_ring->size, 4096);
5584
5585	set_dev_node(dev, ring_node);
5586	rx_ring->desc = dma_alloc_coherent(dev,
5587					   rx_ring->size,
5588					   &rx_ring->dma,
5589					   GFP_KERNEL);
5590	set_dev_node(dev, orig_node);
5591	if (!rx_ring->desc)
5592		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5593						   &rx_ring->dma, GFP_KERNEL);
5594	if (!rx_ring->desc)
5595		goto err;
5596
5597	rx_ring->next_to_clean = 0;
5598	rx_ring->next_to_use = 0;
5599
5600	return 0;
5601err:
5602	vfree(rx_ring->rx_buffer_info);
5603	rx_ring->rx_buffer_info = NULL;
5604	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5605	return -ENOMEM;
5606}
5607
5608/**
5609 * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5610 * @adapter: board private structure
5611 *
5612 * If this function returns with an error, then it's possible one or
5613 * more of the rings is populated (while the rest are not).  It is the
5614 * callers duty to clean those orphaned rings.
5615 *
5616 * Return 0 on success, negative on failure
5617 **/
5618static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5619{
5620	int i, err = 0;
5621
5622	for (i = 0; i < adapter->num_rx_queues; i++) {
5623		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5624		if (!err)
5625			continue;
5626
5627		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5628		goto err_setup_rx;
5629	}
5630
5631#ifdef IXGBE_FCOE
5632	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5633	if (!err)
5634#endif
5635		return 0;
5636err_setup_rx:
5637	/* rewind the index freeing the rings as we go */
5638	while (i--)
5639		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5640	return err;
5641}
5642
5643/**
5644 * ixgbe_free_tx_resources - Free Tx Resources per Queue
5645 * @tx_ring: Tx descriptor ring for a specific queue
5646 *
5647 * Free all transmit software resources
5648 **/
5649void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5650{
5651	ixgbe_clean_tx_ring(tx_ring);
5652
5653	vfree(tx_ring->tx_buffer_info);
5654	tx_ring->tx_buffer_info = NULL;
5655
5656	/* if not set, then don't free */
5657	if (!tx_ring->desc)
5658		return;
5659
5660	dma_free_coherent(tx_ring->dev, tx_ring->size,
5661			  tx_ring->desc, tx_ring->dma);
5662
5663	tx_ring->desc = NULL;
5664}
5665
5666/**
5667 * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5668 * @adapter: board private structure
5669 *
5670 * Free all transmit software resources
5671 **/
5672static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
5673{
5674	int i;
5675
5676	for (i = 0; i < adapter->num_tx_queues; i++)
5677		if (adapter->tx_ring[i]->desc)
5678			ixgbe_free_tx_resources(adapter->tx_ring[i]);
5679}
5680
5681/**
5682 * ixgbe_free_rx_resources - Free Rx Resources
5683 * @rx_ring: ring to clean the resources from
5684 *
5685 * Free all receive software resources
5686 **/
5687void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
5688{
5689	ixgbe_clean_rx_ring(rx_ring);
5690
5691	vfree(rx_ring->rx_buffer_info);
5692	rx_ring->rx_buffer_info = NULL;
5693
5694	/* if not set, then don't free */
5695	if (!rx_ring->desc)
5696		return;
5697
5698	dma_free_coherent(rx_ring->dev, rx_ring->size,
5699			  rx_ring->desc, rx_ring->dma);
5700
5701	rx_ring->desc = NULL;
5702}
5703
5704/**
5705 * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
5706 * @adapter: board private structure
5707 *
5708 * Free all receive software resources
5709 **/
5710static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
5711{
5712	int i;
5713
5714#ifdef IXGBE_FCOE
5715	ixgbe_free_fcoe_ddp_resources(adapter);
5716
5717#endif
5718	for (i = 0; i < adapter->num_rx_queues; i++)
5719		if (adapter->rx_ring[i]->desc)
5720			ixgbe_free_rx_resources(adapter->rx_ring[i]);
5721}
5722
5723/**
5724 * ixgbe_change_mtu - Change the Maximum Transfer Unit
5725 * @netdev: network interface device structure
5726 * @new_mtu: new value for maximum frame size
5727 *
5728 * Returns 0 on success, negative on failure
5729 **/
5730static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
5731{
5732	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5733	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5734
5735	/* MTU < 68 is an error and causes problems on some kernels */
5736	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
5737		return -EINVAL;
5738
5739	/*
5740	 * For 82599EB we cannot allow legacy VFs to enable their receive
5741	 * paths when MTU greater than 1500 is configured.  So display a
5742	 * warning that legacy VFs will be disabled.
5743	 */
5744	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
5745	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
5746	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
5747		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
5748
5749	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5750
5751	/* must set new MTU before calling down or up */
5752	netdev->mtu = new_mtu;
5753
5754	if (netif_running(netdev))
5755		ixgbe_reinit_locked(adapter);
5756
5757	return 0;
5758}
5759
5760/**
5761 * ixgbe_open - Called when a network interface is made active
5762 * @netdev: network interface device structure
5763 *
5764 * Returns 0 on success, negative value on failure
5765 *
5766 * The open entry point is called when a network interface is made
5767 * active by the system (IFF_UP).  At this point all resources needed
5768 * for transmit and receive operations are allocated, the interrupt
5769 * handler is registered with the OS, the watchdog timer is started,
5770 * and the stack is notified that the interface is ready.
5771 **/
5772static int ixgbe_open(struct net_device *netdev)
5773{
5774	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5775	struct ixgbe_hw *hw = &adapter->hw;
5776	int err, queues;
5777
5778	/* disallow open during test */
5779	if (test_bit(__IXGBE_TESTING, &adapter->state))
5780		return -EBUSY;
5781
5782	netif_carrier_off(netdev);
5783
5784	/* allocate transmit descriptors */
5785	err = ixgbe_setup_all_tx_resources(adapter);
5786	if (err)
5787		goto err_setup_tx;
5788
5789	/* allocate receive descriptors */
5790	err = ixgbe_setup_all_rx_resources(adapter);
5791	if (err)
5792		goto err_setup_rx;
5793
5794	ixgbe_configure(adapter);
5795
5796	err = ixgbe_request_irq(adapter);
5797	if (err)
5798		goto err_req_irq;
5799
5800	/* Notify the stack of the actual queue counts. */
5801	if (adapter->num_rx_pools > 1)
5802		queues = adapter->num_rx_queues_per_pool;
5803	else
5804		queues = adapter->num_tx_queues;
5805
5806	err = netif_set_real_num_tx_queues(netdev, queues);
5807	if (err)
5808		goto err_set_queues;
5809
5810	if (adapter->num_rx_pools > 1 &&
5811	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
5812		queues = IXGBE_MAX_L2A_QUEUES;
5813	else
5814		queues = adapter->num_rx_queues;
5815	err = netif_set_real_num_rx_queues(netdev, queues);
5816	if (err)
5817		goto err_set_queues;
5818
5819	ixgbe_ptp_init(adapter);
5820
5821	ixgbe_up_complete(adapter);
5822
5823	ixgbe_clear_vxlan_port(adapter);
5824#ifdef CONFIG_IXGBE_VXLAN
5825	vxlan_get_rx_port(netdev);
5826#endif
5827
5828	return 0;
5829
5830err_set_queues:
5831	ixgbe_free_irq(adapter);
5832err_req_irq:
5833	ixgbe_free_all_rx_resources(adapter);
5834	if (hw->phy.ops.set_phy_power && !adapter->wol)
5835		hw->phy.ops.set_phy_power(&adapter->hw, false);
5836err_setup_rx:
5837	ixgbe_free_all_tx_resources(adapter);
5838err_setup_tx:
5839	ixgbe_reset(adapter);
5840
5841	return err;
5842}
5843
5844static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
5845{
5846	ixgbe_ptp_suspend(adapter);
5847
5848	if (adapter->hw.phy.ops.enter_lplu) {
5849		adapter->hw.phy.reset_disable = true;
5850		ixgbe_down(adapter);
5851		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
5852		adapter->hw.phy.reset_disable = false;
5853	} else {
5854		ixgbe_down(adapter);
5855	}
5856
5857	ixgbe_free_irq(adapter);
5858
5859	ixgbe_free_all_tx_resources(adapter);
5860	ixgbe_free_all_rx_resources(adapter);
5861}
5862
5863/**
5864 * ixgbe_close - Disables a network interface
5865 * @netdev: network interface device structure
5866 *
5867 * Returns 0, this is not allowed to fail
5868 *
5869 * The close entry point is called when an interface is de-activated
5870 * by the OS.  The hardware is still under the drivers control, but
5871 * needs to be disabled.  A global MAC reset is issued to stop the
5872 * hardware, and all transmit and receive resources are freed.
5873 **/
5874static int ixgbe_close(struct net_device *netdev)
5875{
5876	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5877
5878	ixgbe_ptp_stop(adapter);
5879
5880	ixgbe_close_suspend(adapter);
5881
5882	ixgbe_fdir_filter_exit(adapter);
5883
5884	ixgbe_release_hw_control(adapter);
5885
5886	return 0;
5887}
5888
5889#ifdef CONFIG_PM
5890static int ixgbe_resume(struct pci_dev *pdev)
5891{
5892	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5893	struct net_device *netdev = adapter->netdev;
5894	u32 err;
5895
5896	adapter->hw.hw_addr = adapter->io_addr;
5897	pci_set_power_state(pdev, PCI_D0);
5898	pci_restore_state(pdev);
5899	/*
5900	 * pci_restore_state clears dev->state_saved so call
5901	 * pci_save_state to restore it.
5902	 */
5903	pci_save_state(pdev);
5904
5905	err = pci_enable_device_mem(pdev);
5906	if (err) {
5907		e_dev_err("Cannot enable PCI device from suspend\n");
5908		return err;
5909	}
5910	smp_mb__before_atomic();
5911	clear_bit(__IXGBE_DISABLED, &adapter->state);
5912	pci_set_master(pdev);
5913
5914	pci_wake_from_d3(pdev, false);
5915
5916	ixgbe_reset(adapter);
5917
5918	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
5919
5920	rtnl_lock();
5921	err = ixgbe_init_interrupt_scheme(adapter);
5922	if (!err && netif_running(netdev))
5923		err = ixgbe_open(netdev);
5924
5925	rtnl_unlock();
5926
5927	if (err)
5928		return err;
5929
5930	netif_device_attach(netdev);
5931
5932	return 0;
5933}
5934#endif /* CONFIG_PM */
5935
5936static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
5937{
5938	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5939	struct net_device *netdev = adapter->netdev;
5940	struct ixgbe_hw *hw = &adapter->hw;
5941	u32 ctrl, fctrl;
5942	u32 wufc = adapter->wol;
5943#ifdef CONFIG_PM
5944	int retval = 0;
5945#endif
5946
5947	netif_device_detach(netdev);
5948
5949	rtnl_lock();
5950	if (netif_running(netdev))
5951		ixgbe_close_suspend(adapter);
5952	rtnl_unlock();
5953
5954	ixgbe_clear_interrupt_scheme(adapter);
5955
5956#ifdef CONFIG_PM
5957	retval = pci_save_state(pdev);
5958	if (retval)
5959		return retval;
5960
5961#endif
5962	if (hw->mac.ops.stop_link_on_d3)
5963		hw->mac.ops.stop_link_on_d3(hw);
5964
5965	if (wufc) {
5966		ixgbe_set_rx_mode(netdev);
5967
5968		/* enable the optics for 82599 SFP+ fiber as we can WoL */
5969		if (hw->mac.ops.enable_tx_laser)
5970			hw->mac.ops.enable_tx_laser(hw);
5971
5972		/* turn on all-multi mode if wake on multicast is enabled */
5973		if (wufc & IXGBE_WUFC_MC) {
5974			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5975			fctrl |= IXGBE_FCTRL_MPE;
5976			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5977		}
5978
5979		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
5980		ctrl |= IXGBE_CTRL_GIO_DIS;
5981		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
5982
5983		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
5984	} else {
5985		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
5986		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
5987	}
5988
5989	switch (hw->mac.type) {
5990	case ixgbe_mac_82598EB:
5991		pci_wake_from_d3(pdev, false);
5992		break;
5993	case ixgbe_mac_82599EB:
5994	case ixgbe_mac_X540:
5995	case ixgbe_mac_X550:
5996	case ixgbe_mac_X550EM_x:
5997		pci_wake_from_d3(pdev, !!wufc);
5998		break;
5999	default:
6000		break;
6001	}
6002
6003	*enable_wake = !!wufc;
6004	if (hw->phy.ops.set_phy_power && !*enable_wake)
6005		hw->phy.ops.set_phy_power(hw, false);
6006
6007	ixgbe_release_hw_control(adapter);
6008
6009	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6010		pci_disable_device(pdev);
6011
6012	return 0;
6013}
6014
6015#ifdef CONFIG_PM
6016static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6017{
6018	int retval;
6019	bool wake;
6020
6021	retval = __ixgbe_shutdown(pdev, &wake);
6022	if (retval)
6023		return retval;
6024
6025	if (wake) {
6026		pci_prepare_to_sleep(pdev);
6027	} else {
6028		pci_wake_from_d3(pdev, false);
6029		pci_set_power_state(pdev, PCI_D3hot);
6030	}
6031
6032	return 0;
6033}
6034#endif /* CONFIG_PM */
6035
6036static void ixgbe_shutdown(struct pci_dev *pdev)
6037{
6038	bool wake;
6039
6040	__ixgbe_shutdown(pdev, &wake);
6041
6042	if (system_state == SYSTEM_POWER_OFF) {
6043		pci_wake_from_d3(pdev, wake);
6044		pci_set_power_state(pdev, PCI_D3hot);
6045	}
6046}
6047
6048/**
6049 * ixgbe_update_stats - Update the board statistics counters.
6050 * @adapter: board private structure
6051 **/
6052void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6053{
6054	struct net_device *netdev = adapter->netdev;
6055	struct ixgbe_hw *hw = &adapter->hw;
6056	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6057	u64 total_mpc = 0;
6058	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6059	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6060	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6061	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6062
6063	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6064	    test_bit(__IXGBE_RESETTING, &adapter->state))
6065		return;
6066
6067	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6068		u64 rsc_count = 0;
6069		u64 rsc_flush = 0;
6070		for (i = 0; i < adapter->num_rx_queues; i++) {
6071			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6072			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6073		}
6074		adapter->rsc_total_count = rsc_count;
6075		adapter->rsc_total_flush = rsc_flush;
6076	}
6077
6078	for (i = 0; i < adapter->num_rx_queues; i++) {
6079		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6080		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6081		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6082		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6083		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6084		bytes += rx_ring->stats.bytes;
6085		packets += rx_ring->stats.packets;
6086	}
6087	adapter->non_eop_descs = non_eop_descs;
6088	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6089	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6090	adapter->hw_csum_rx_error = hw_csum_rx_error;
6091	netdev->stats.rx_bytes = bytes;
6092	netdev->stats.rx_packets = packets;
6093
6094	bytes = 0;
6095	packets = 0;
6096	/* gather some stats to the adapter struct that are per queue */
6097	for (i = 0; i < adapter->num_tx_queues; i++) {
6098		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6099		restart_queue += tx_ring->tx_stats.restart_queue;
6100		tx_busy += tx_ring->tx_stats.tx_busy;
6101		bytes += tx_ring->stats.bytes;
6102		packets += tx_ring->stats.packets;
6103	}
6104	adapter->restart_queue = restart_queue;
6105	adapter->tx_busy = tx_busy;
6106	netdev->stats.tx_bytes = bytes;
6107	netdev->stats.tx_packets = packets;
6108
6109	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6110
6111	/* 8 register reads */
6112	for (i = 0; i < 8; i++) {
6113		/* for packet buffers not used, the register should read 0 */
6114		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6115		missed_rx += mpc;
6116		hwstats->mpc[i] += mpc;
6117		total_mpc += hwstats->mpc[i];
6118		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6119		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6120		switch (hw->mac.type) {
6121		case ixgbe_mac_82598EB:
6122			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6123			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6124			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6125			hwstats->pxonrxc[i] +=
6126				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6127			break;
6128		case ixgbe_mac_82599EB:
6129		case ixgbe_mac_X540:
6130		case ixgbe_mac_X550:
6131		case ixgbe_mac_X550EM_x:
6132			hwstats->pxonrxc[i] +=
6133				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6134			break;
6135		default:
6136			break;
6137		}
6138	}
6139
6140	/*16 register reads */
6141	for (i = 0; i < 16; i++) {
6142		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6143		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6144		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6145		    (hw->mac.type == ixgbe_mac_X540) ||
6146		    (hw->mac.type == ixgbe_mac_X550) ||
6147		    (hw->mac.type == ixgbe_mac_X550EM_x)) {
6148			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6149			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6150			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6151			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6152		}
6153	}
6154
6155	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6156	/* work around hardware counting issue */
6157	hwstats->gprc -= missed_rx;
6158
6159	ixgbe_update_xoff_received(adapter);
6160
6161	/* 82598 hardware only has a 32 bit counter in the high register */
6162	switch (hw->mac.type) {
6163	case ixgbe_mac_82598EB:
6164		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6165		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6166		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6167		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6168		break;
6169	case ixgbe_mac_X540:
6170	case ixgbe_mac_X550:
6171	case ixgbe_mac_X550EM_x:
6172		/* OS2BMC stats are X540 and later */
6173		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6174		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6175		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6176		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6177	case ixgbe_mac_82599EB:
6178		for (i = 0; i < 16; i++)
6179			adapter->hw_rx_no_dma_resources +=
6180					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6181		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6182		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6183		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6184		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6185		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6186		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6187		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6188		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6189		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6190#ifdef IXGBE_FCOE
6191		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6192		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6193		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6194		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6195		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6196		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6197		/* Add up per cpu counters for total ddp aloc fail */
6198		if (adapter->fcoe.ddp_pool) {
6199			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6200			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6201			unsigned int cpu;
6202			u64 noddp = 0, noddp_ext_buff = 0;
6203			for_each_possible_cpu(cpu) {
6204				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6205				noddp += ddp_pool->noddp;
6206				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6207			}
6208			hwstats->fcoe_noddp = noddp;
6209			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6210		}
6211#endif /* IXGBE_FCOE */
6212		break;
6213	default:
6214		break;
6215	}
6216	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6217	hwstats->bprc += bprc;
6218	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6219	if (hw->mac.type == ixgbe_mac_82598EB)
6220		hwstats->mprc -= bprc;
6221	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6222	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6223	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6224	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6225	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6226	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6227	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6228	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6229	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6230	hwstats->lxontxc += lxon;
6231	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6232	hwstats->lxofftxc += lxoff;
6233	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6234	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6235	/*
6236	 * 82598 errata - tx of flow control packets is included in tx counters
6237	 */
6238	xon_off_tot = lxon + lxoff;
6239	hwstats->gptc -= xon_off_tot;
6240	hwstats->mptc -= xon_off_tot;
6241	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6242	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6243	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6244	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6245	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6246	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6247	hwstats->ptc64 -= xon_off_tot;
6248	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6249	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6250	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6251	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6252	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6253	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6254
6255	/* Fill out the OS statistics structure */
6256	netdev->stats.multicast = hwstats->mprc;
6257
6258	/* Rx Errors */
6259	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6260	netdev->stats.rx_dropped = 0;
6261	netdev->stats.rx_length_errors = hwstats->rlec;
6262	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6263	netdev->stats.rx_missed_errors = total_mpc;
6264}
6265
6266/**
6267 * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6268 * @adapter: pointer to the device adapter structure
6269 **/
6270static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6271{
6272	struct ixgbe_hw *hw = &adapter->hw;
6273	int i;
6274
6275	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6276		return;
6277
6278	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6279
6280	/* if interface is down do nothing */
6281	if (test_bit(__IXGBE_DOWN, &adapter->state))
6282		return;
6283
6284	/* do nothing if we are not using signature filters */
6285	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6286		return;
6287
6288	adapter->fdir_overflow++;
6289
6290	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6291		for (i = 0; i < adapter->num_tx_queues; i++)
6292			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6293				&(adapter->tx_ring[i]->state));
6294		/* re-enable flow director interrupts */
6295		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6296	} else {
6297		e_err(probe, "failed to finish FDIR re-initialization, "
6298		      "ignored adding FDIR ATR filters\n");
6299	}
6300}
6301
6302/**
6303 * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6304 * @adapter: pointer to the device adapter structure
6305 *
6306 * This function serves two purposes.  First it strobes the interrupt lines
6307 * in order to make certain interrupts are occurring.  Secondly it sets the
6308 * bits needed to check for TX hangs.  As a result we should immediately
6309 * determine if a hang has occurred.
6310 */
6311static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6312{
6313	struct ixgbe_hw *hw = &adapter->hw;
6314	u64 eics = 0;
6315	int i;
6316
6317	/* If we're down, removing or resetting, just bail */
6318	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6319	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6320	    test_bit(__IXGBE_RESETTING, &adapter->state))
6321		return;
6322
6323	/* Force detection of hung controller */
6324	if (netif_carrier_ok(adapter->netdev)) {
6325		for (i = 0; i < adapter->num_tx_queues; i++)
6326			set_check_for_tx_hang(adapter->tx_ring[i]);
6327	}
6328
6329	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6330		/*
6331		 * for legacy and MSI interrupts don't set any bits
6332		 * that are enabled for EIAM, because this operation
6333		 * would set *both* EIMS and EICS for any bit in EIAM
6334		 */
6335		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6336			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6337	} else {
6338		/* get one bit for every active tx/rx interrupt vector */
6339		for (i = 0; i < adapter->num_q_vectors; i++) {
6340			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6341			if (qv->rx.ring || qv->tx.ring)
6342				eics |= ((u64)1 << i);
6343		}
6344	}
6345
6346	/* Cause software interrupt to ensure rings are cleaned */
6347	ixgbe_irq_rearm_queues(adapter, eics);
6348}
6349
6350/**
6351 * ixgbe_watchdog_update_link - update the link status
6352 * @adapter: pointer to the device adapter structure
6353 * @link_speed: pointer to a u32 to store the link_speed
6354 **/
6355static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6356{
6357	struct ixgbe_hw *hw = &adapter->hw;
6358	u32 link_speed = adapter->link_speed;
6359	bool link_up = adapter->link_up;
6360	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6361
6362	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6363		return;
6364
6365	if (hw->mac.ops.check_link) {
6366		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6367	} else {
6368		/* always assume link is up, if no check link function */
6369		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6370		link_up = true;
6371	}
6372
6373	if (adapter->ixgbe_ieee_pfc)
6374		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6375
6376	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6377		hw->mac.ops.fc_enable(hw);
6378		ixgbe_set_rx_drop_en(adapter);
6379	}
6380
6381	if (link_up ||
6382	    time_after(jiffies, (adapter->link_check_timeout +
6383				 IXGBE_TRY_LINK_TIMEOUT))) {
6384		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6385		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6386		IXGBE_WRITE_FLUSH(hw);
6387	}
6388
6389	adapter->link_up = link_up;
6390	adapter->link_speed = link_speed;
6391}
6392
6393static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6394{
6395#ifdef CONFIG_IXGBE_DCB
6396	struct net_device *netdev = adapter->netdev;
6397	struct dcb_app app = {
6398			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6399			      .protocol = 0,
6400			     };
6401	u8 up = 0;
6402
6403	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6404		up = dcb_ieee_getapp_mask(netdev, &app);
6405
6406	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6407#endif
6408}
6409
6410/**
6411 * ixgbe_watchdog_link_is_up - update netif_carrier status and
6412 *                             print link up message
6413 * @adapter: pointer to the device adapter structure
6414 **/
6415static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6416{
6417	struct net_device *netdev = adapter->netdev;
6418	struct ixgbe_hw *hw = &adapter->hw;
6419	struct net_device *upper;
6420	struct list_head *iter;
6421	u32 link_speed = adapter->link_speed;
6422	const char *speed_str;
6423	bool flow_rx, flow_tx;
6424
6425	/* only continue if link was previously down */
6426	if (netif_carrier_ok(netdev))
6427		return;
6428
6429	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6430
6431	switch (hw->mac.type) {
6432	case ixgbe_mac_82598EB: {
6433		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6434		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6435		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6436		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6437	}
6438		break;
6439	case ixgbe_mac_X540:
6440	case ixgbe_mac_X550:
6441	case ixgbe_mac_X550EM_x:
6442	case ixgbe_mac_82599EB: {
6443		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6444		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6445		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6446		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6447	}
6448		break;
6449	default:
6450		flow_tx = false;
6451		flow_rx = false;
6452		break;
6453	}
6454
6455	adapter->last_rx_ptp_check = jiffies;
6456
6457	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6458		ixgbe_ptp_start_cyclecounter(adapter);
6459
6460	switch (link_speed) {
6461	case IXGBE_LINK_SPEED_10GB_FULL:
6462		speed_str = "10 Gbps";
6463		break;
6464	case IXGBE_LINK_SPEED_2_5GB_FULL:
6465		speed_str = "2.5 Gbps";
6466		break;
6467	case IXGBE_LINK_SPEED_1GB_FULL:
6468		speed_str = "1 Gbps";
6469		break;
6470	case IXGBE_LINK_SPEED_100_FULL:
6471		speed_str = "100 Mbps";
6472		break;
6473	default:
6474		speed_str = "unknown speed";
6475		break;
6476	}
6477	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
6478	       ((flow_rx && flow_tx) ? "RX/TX" :
6479	       (flow_rx ? "RX" :
6480	       (flow_tx ? "TX" : "None"))));
6481
6482	netif_carrier_on(netdev);
6483	ixgbe_check_vf_rate_limit(adapter);
6484
6485	/* enable transmits */
6486	netif_tx_wake_all_queues(adapter->netdev);
6487
6488	/* enable any upper devices */
6489	rtnl_lock();
6490	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
6491		if (netif_is_macvlan(upper)) {
6492			struct macvlan_dev *vlan = netdev_priv(upper);
6493
6494			if (vlan->fwd_priv)
6495				netif_tx_wake_all_queues(upper);
6496		}
6497	}
6498	rtnl_unlock();
6499
6500	/* update the default user priority for VFs */
6501	ixgbe_update_default_up(adapter);
6502
6503	/* ping all the active vfs to let them know link has changed */
6504	ixgbe_ping_all_vfs(adapter);
6505}
6506
6507/**
6508 * ixgbe_watchdog_link_is_down - update netif_carrier status and
6509 *                               print link down message
6510 * @adapter: pointer to the adapter structure
6511 **/
6512static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6513{
6514	struct net_device *netdev = adapter->netdev;
6515	struct ixgbe_hw *hw = &adapter->hw;
6516
6517	adapter->link_up = false;
6518	adapter->link_speed = 0;
6519
6520	/* only continue if link was up previously */
6521	if (!netif_carrier_ok(netdev))
6522		return;
6523
6524	/* poll for SFP+ cable when link is down */
6525	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6526		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6527
6528	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6529		ixgbe_ptp_start_cyclecounter(adapter);
6530
6531	e_info(drv, "NIC Link is Down\n");
6532	netif_carrier_off(netdev);
6533
6534	/* ping all the active vfs to let them know link has changed */
6535	ixgbe_ping_all_vfs(adapter);
6536}
6537
6538static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
6539{
6540	int i;
6541
6542	for (i = 0; i < adapter->num_tx_queues; i++) {
6543		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6544
6545		if (tx_ring->next_to_use != tx_ring->next_to_clean)
6546			return true;
6547	}
6548
6549	return false;
6550}
6551
6552static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
6553{
6554	struct ixgbe_hw *hw = &adapter->hw;
6555	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
6556	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
6557
6558	int i, j;
6559
6560	if (!adapter->num_vfs)
6561		return false;
6562
6563	/* resetting the PF is only needed for MAC before X550 */
6564	if (hw->mac.type >= ixgbe_mac_X550)
6565		return false;
6566
6567	for (i = 0; i < adapter->num_vfs; i++) {
6568		for (j = 0; j < q_per_pool; j++) {
6569			u32 h, t;
6570
6571			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
6572			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
6573
6574			if (h != t)
6575				return true;
6576		}
6577	}
6578
6579	return false;
6580}
6581
6582/**
6583 * ixgbe_watchdog_flush_tx - flush queues on link down
6584 * @adapter: pointer to the device adapter structure
6585 **/
6586static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6587{
6588	if (!netif_carrier_ok(adapter->netdev)) {
6589		if (ixgbe_ring_tx_pending(adapter) ||
6590		    ixgbe_vf_tx_pending(adapter)) {
6591			/* We've lost link, so the controller stops DMA,
6592			 * but we've got queued Tx work that's never going
6593			 * to get done, so reset controller to flush Tx.
6594			 * (Do the reset outside of interrupt context).
6595			 */
6596			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6597			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
6598		}
6599	}
6600}
6601
6602#ifdef CONFIG_PCI_IOV
6603static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
6604				      struct pci_dev *vfdev)
6605{
6606	if (!pci_wait_for_pending_transaction(vfdev))
6607		e_dev_warn("Issuing VFLR with pending transactions\n");
6608
6609	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
6610	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
6611
6612	msleep(100);
6613}
6614
6615static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6616{
6617	struct ixgbe_hw *hw = &adapter->hw;
6618	struct pci_dev *pdev = adapter->pdev;
6619	struct pci_dev *vfdev;
6620	u32 gpc;
6621	int pos;
6622	unsigned short vf_id;
6623
6624	if (!(netif_carrier_ok(adapter->netdev)))
6625		return;
6626
6627	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6628	if (gpc) /* If incrementing then no need for the check below */
6629		return;
6630	/* Check to see if a bad DMA write target from an errant or
6631	 * malicious VF has caused a PCIe error.  If so then we can
6632	 * issue a VFLR to the offending VF(s) and then resume without
6633	 * requesting a full slot reset.
6634	 */
6635
6636	if (!pdev)
6637		return;
6638
6639	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
6640	if (!pos)
6641		return;
6642
6643	/* get the device ID for the VF */
6644	pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_id);
6645
6646	/* check status reg for all VFs owned by this PF */
6647	vfdev = pci_get_device(pdev->vendor, vf_id, NULL);
6648	while (vfdev) {
6649		if (vfdev->is_virtfn && (vfdev->physfn == pdev)) {
6650			u16 status_reg;
6651
6652			pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
6653			if (status_reg & PCI_STATUS_REC_MASTER_ABORT)
6654				/* issue VFLR */
6655				ixgbe_issue_vf_flr(adapter, vfdev);
6656		}
6657
6658		vfdev = pci_get_device(pdev->vendor, vf_id, vfdev);
6659	}
6660}
6661
6662static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6663{
6664	u32 ssvpc;
6665
6666	/* Do not perform spoof check for 82598 or if not in IOV mode */
6667	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
6668	    adapter->num_vfs == 0)
6669		return;
6670
6671	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
6672
6673	/*
6674	 * ssvpc register is cleared on read, if zero then no
6675	 * spoofed packets in the last interval.
6676	 */
6677	if (!ssvpc)
6678		return;
6679
6680	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
6681}
6682#else
6683static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
6684{
6685}
6686
6687static void
6688ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
6689{
6690}
6691#endif /* CONFIG_PCI_IOV */
6692
6693
6694/**
6695 * ixgbe_watchdog_subtask - check and bring link up
6696 * @adapter: pointer to the device adapter structure
6697 **/
6698static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
6699{
6700	/* if interface is down, removing or resetting, do nothing */
6701	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6702	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6703	    test_bit(__IXGBE_RESETTING, &adapter->state))
6704		return;
6705
6706	ixgbe_watchdog_update_link(adapter);
6707
6708	if (adapter->link_up)
6709		ixgbe_watchdog_link_is_up(adapter);
6710	else
6711		ixgbe_watchdog_link_is_down(adapter);
6712
6713	ixgbe_check_for_bad_vf(adapter);
6714	ixgbe_spoof_check(adapter);
6715	ixgbe_update_stats(adapter);
6716
6717	ixgbe_watchdog_flush_tx(adapter);
6718}
6719
6720/**
6721 * ixgbe_sfp_detection_subtask - poll for SFP+ cable
6722 * @adapter: the ixgbe adapter structure
6723 **/
6724static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
6725{
6726	struct ixgbe_hw *hw = &adapter->hw;
6727	s32 err;
6728
6729	/* not searching for SFP so there is nothing to do here */
6730	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
6731	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6732		return;
6733
6734	if (adapter->sfp_poll_time &&
6735	    time_after(adapter->sfp_poll_time, jiffies))
6736		return; /* If not yet time to poll for SFP */
6737
6738	/* someone else is in init, wait until next service event */
6739	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6740		return;
6741
6742	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
6743
6744	err = hw->phy.ops.identify_sfp(hw);
6745	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6746		goto sfp_out;
6747
6748	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
6749		/* If no cable is present, then we need to reset
6750		 * the next time we find a good cable. */
6751		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
6752	}
6753
6754	/* exit on error */
6755	if (err)
6756		goto sfp_out;
6757
6758	/* exit if reset not needed */
6759	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6760		goto sfp_out;
6761
6762	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
6763
6764	/*
6765	 * A module may be identified correctly, but the EEPROM may not have
6766	 * support for that module.  setup_sfp() will fail in that case, so
6767	 * we should not allow that module to load.
6768	 */
6769	if (hw->mac.type == ixgbe_mac_82598EB)
6770		err = hw->phy.ops.reset(hw);
6771	else
6772		err = hw->mac.ops.setup_sfp(hw);
6773
6774	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6775		goto sfp_out;
6776
6777	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
6778	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
6779
6780sfp_out:
6781	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6782
6783	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
6784	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
6785		e_dev_err("failed to initialize because an unsupported "
6786			  "SFP+ module type was detected.\n");
6787		e_dev_err("Reload the driver after installing a "
6788			  "supported module.\n");
6789		unregister_netdev(adapter->netdev);
6790	}
6791}
6792
6793/**
6794 * ixgbe_sfp_link_config_subtask - set up link SFP after module install
6795 * @adapter: the ixgbe adapter structure
6796 **/
6797static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
6798{
6799	struct ixgbe_hw *hw = &adapter->hw;
6800	u32 speed;
6801	bool autoneg = false;
6802
6803	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
6804		return;
6805
6806	/* someone else is in init, wait until next service event */
6807	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6808		return;
6809
6810	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
6811
6812	speed = hw->phy.autoneg_advertised;
6813	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
6814		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
6815
6816		/* setup the highest link when no autoneg */
6817		if (!autoneg) {
6818			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
6819				speed = IXGBE_LINK_SPEED_10GB_FULL;
6820		}
6821	}
6822
6823	if (hw->mac.ops.setup_link)
6824		hw->mac.ops.setup_link(hw, speed, true);
6825
6826	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
6827	adapter->link_check_timeout = jiffies;
6828	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6829}
6830
6831/**
6832 * ixgbe_service_timer - Timer Call-back
6833 * @data: pointer to adapter cast into an unsigned long
6834 **/
6835static void ixgbe_service_timer(unsigned long data)
6836{
6837	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
6838	unsigned long next_event_offset;
6839
6840	/* poll faster when waiting for link */
6841	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
6842		next_event_offset = HZ / 10;
6843	else
6844		next_event_offset = HZ * 2;
6845
6846	/* Reset the timer */
6847	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
6848
6849	ixgbe_service_event_schedule(adapter);
6850}
6851
6852static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
6853{
6854	struct ixgbe_hw *hw = &adapter->hw;
6855	u32 status;
6856
6857	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
6858		return;
6859
6860	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
6861
6862	if (!hw->phy.ops.handle_lasi)
6863		return;
6864
6865	status = hw->phy.ops.handle_lasi(&adapter->hw);
6866	if (status != IXGBE_ERR_OVERTEMP)
6867		return;
6868
6869	e_crit(drv, "%s\n", ixgbe_overheat_msg);
6870}
6871
6872static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
6873{
6874	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
6875		return;
6876
6877	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
6878
6879	/* If we're already down, removing or resetting, just bail */
6880	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6881	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6882	    test_bit(__IXGBE_RESETTING, &adapter->state))
6883		return;
6884
6885	ixgbe_dump(adapter);
6886	netdev_err(adapter->netdev, "Reset adapter\n");
6887	adapter->tx_timeout_count++;
6888
6889	rtnl_lock();
6890	ixgbe_reinit_locked(adapter);
6891	rtnl_unlock();
6892}
6893
6894/**
6895 * ixgbe_service_task - manages and runs subtasks
6896 * @work: pointer to work_struct containing our data
6897 **/
6898static void ixgbe_service_task(struct work_struct *work)
6899{
6900	struct ixgbe_adapter *adapter = container_of(work,
6901						     struct ixgbe_adapter,
6902						     service_task);
6903	if (ixgbe_removed(adapter->hw.hw_addr)) {
6904		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
6905			rtnl_lock();
6906			ixgbe_down(adapter);
6907			rtnl_unlock();
6908		}
6909		ixgbe_service_event_complete(adapter);
6910		return;
6911	}
6912#ifdef CONFIG_IXGBE_VXLAN
6913	if (adapter->flags2 & IXGBE_FLAG2_VXLAN_REREG_NEEDED) {
6914		adapter->flags2 &= ~IXGBE_FLAG2_VXLAN_REREG_NEEDED;
6915		vxlan_get_rx_port(adapter->netdev);
6916	}
6917#endif /* CONFIG_IXGBE_VXLAN */
6918	ixgbe_reset_subtask(adapter);
6919	ixgbe_phy_interrupt_subtask(adapter);
6920	ixgbe_sfp_detection_subtask(adapter);
6921	ixgbe_sfp_link_config_subtask(adapter);
6922	ixgbe_check_overtemp_subtask(adapter);
6923	ixgbe_watchdog_subtask(adapter);
6924	ixgbe_fdir_reinit_subtask(adapter);
6925	ixgbe_check_hang_subtask(adapter);
6926
6927	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
6928		ixgbe_ptp_overflow_check(adapter);
6929		ixgbe_ptp_rx_hang(adapter);
6930	}
6931
6932	ixgbe_service_event_complete(adapter);
6933}
6934
6935static int ixgbe_tso(struct ixgbe_ring *tx_ring,
6936		     struct ixgbe_tx_buffer *first,
6937		     u8 *hdr_len)
6938{
6939	struct sk_buff *skb = first->skb;
6940	u32 vlan_macip_lens, type_tucmd;
6941	u32 mss_l4len_idx, l4len;
6942	int err;
6943
6944	if (skb->ip_summed != CHECKSUM_PARTIAL)
6945		return 0;
6946
6947	if (!skb_is_gso(skb))
6948		return 0;
6949
6950	err = skb_cow_head(skb, 0);
6951	if (err < 0)
6952		return err;
6953
6954	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6955	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
6956
6957	if (first->protocol == htons(ETH_P_IP)) {
6958		struct iphdr *iph = ip_hdr(skb);
6959		iph->tot_len = 0;
6960		iph->check = 0;
6961		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
6962							 iph->daddr, 0,
6963							 IPPROTO_TCP,
6964							 0);
6965		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6966		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6967				   IXGBE_TX_FLAGS_CSUM |
6968				   IXGBE_TX_FLAGS_IPV4;
6969	} else if (skb_is_gso_v6(skb)) {
6970		ipv6_hdr(skb)->payload_len = 0;
6971		tcp_hdr(skb)->check =
6972		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
6973				     &ipv6_hdr(skb)->daddr,
6974				     0, IPPROTO_TCP, 0);
6975		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6976				   IXGBE_TX_FLAGS_CSUM;
6977	}
6978
6979	/* compute header lengths */
6980	l4len = tcp_hdrlen(skb);
6981	*hdr_len = skb_transport_offset(skb) + l4len;
6982
6983	/* update gso size and bytecount with header size */
6984	first->gso_segs = skb_shinfo(skb)->gso_segs;
6985	first->bytecount += (first->gso_segs - 1) * *hdr_len;
6986
6987	/* mss_l4len_id: use 0 as index for TSO */
6988	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
6989	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
6990
6991	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6992	vlan_macip_lens = skb_network_header_len(skb);
6993	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6994	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6995
6996	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
6997			  mss_l4len_idx);
6998
6999	return 1;
7000}
7001
7002static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7003			  struct ixgbe_tx_buffer *first)
7004{
7005	struct sk_buff *skb = first->skb;
7006	u32 vlan_macip_lens = 0;
7007	u32 mss_l4len_idx = 0;
7008	u32 type_tucmd = 0;
7009
7010	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7011		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN) &&
7012		    !(first->tx_flags & IXGBE_TX_FLAGS_CC))
7013			return;
7014		vlan_macip_lens = skb_network_offset(skb) <<
7015				  IXGBE_ADVTXD_MACLEN_SHIFT;
7016	} else {
7017		u8 l4_hdr = 0;
7018		union {
7019			struct iphdr *ipv4;
7020			struct ipv6hdr *ipv6;
7021			u8 *raw;
7022		} network_hdr;
7023		union {
7024			struct tcphdr *tcphdr;
7025			u8 *raw;
7026		} transport_hdr;
7027
7028		if (skb->encapsulation) {
7029			network_hdr.raw = skb_inner_network_header(skb);
7030			transport_hdr.raw = skb_inner_transport_header(skb);
7031			vlan_macip_lens = skb_inner_network_offset(skb) <<
7032					  IXGBE_ADVTXD_MACLEN_SHIFT;
7033		} else {
7034			network_hdr.raw = skb_network_header(skb);
7035			transport_hdr.raw = skb_transport_header(skb);
7036			vlan_macip_lens = skb_network_offset(skb) <<
7037					  IXGBE_ADVTXD_MACLEN_SHIFT;
7038		}
7039
7040		/* use first 4 bits to determine IP version */
7041		switch (network_hdr.ipv4->version) {
7042		case IPVERSION:
7043			vlan_macip_lens |= transport_hdr.raw - network_hdr.raw;
7044			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7045			l4_hdr = network_hdr.ipv4->protocol;
7046			break;
7047		case 6:
7048			vlan_macip_lens |= transport_hdr.raw - network_hdr.raw;
7049			l4_hdr = network_hdr.ipv6->nexthdr;
7050			break;
7051		default:
7052			if (unlikely(net_ratelimit())) {
7053				dev_warn(tx_ring->dev,
7054					 "partial checksum but version=%d\n",
7055					 network_hdr.ipv4->version);
7056			}
7057		}
7058
7059		switch (l4_hdr) {
7060		case IPPROTO_TCP:
7061			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
7062			mss_l4len_idx = (transport_hdr.tcphdr->doff * 4) <<
7063					IXGBE_ADVTXD_L4LEN_SHIFT;
7064			break;
7065		case IPPROTO_SCTP:
7066			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7067			mss_l4len_idx = sizeof(struct sctphdr) <<
7068					IXGBE_ADVTXD_L4LEN_SHIFT;
7069			break;
7070		case IPPROTO_UDP:
7071			mss_l4len_idx = sizeof(struct udphdr) <<
7072					IXGBE_ADVTXD_L4LEN_SHIFT;
7073			break;
7074		default:
7075			if (unlikely(net_ratelimit())) {
7076				dev_warn(tx_ring->dev,
7077				 "partial checksum but l4 proto=%x!\n",
7078				 l4_hdr);
7079			}
7080			break;
7081		}
7082
7083		/* update TX checksum flag */
7084		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7085	}
7086
7087	/* vlan_macip_lens: MACLEN, VLAN tag */
7088	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7089
7090	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
7091			  type_tucmd, mss_l4len_idx);
7092}
7093
7094#define IXGBE_SET_FLAG(_input, _flag, _result) \
7095	((_flag <= _result) ? \
7096	 ((u32)(_input & _flag) * (_result / _flag)) : \
7097	 ((u32)(_input & _flag) / (_flag / _result)))
7098
7099static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7100{
7101	/* set type for advanced descriptor with frame checksum insertion */
7102	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7103		       IXGBE_ADVTXD_DCMD_DEXT |
7104		       IXGBE_ADVTXD_DCMD_IFCS;
7105
7106	/* set HW vlan bit if vlan is present */
7107	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7108				   IXGBE_ADVTXD_DCMD_VLE);
7109
7110	/* set segmentation enable bits for TSO/FSO */
7111	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7112				   IXGBE_ADVTXD_DCMD_TSE);
7113
7114	/* set timestamp bit if present */
7115	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7116				   IXGBE_ADVTXD_MAC_TSTAMP);
7117
7118	/* insert frame checksum */
7119	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7120
7121	return cmd_type;
7122}
7123
7124static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7125				   u32 tx_flags, unsigned int paylen)
7126{
7127	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7128
7129	/* enable L4 checksum for TSO and TX checksum offload */
7130	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7131					IXGBE_TX_FLAGS_CSUM,
7132					IXGBE_ADVTXD_POPTS_TXSM);
7133
7134	/* enble IPv4 checksum for TSO */
7135	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7136					IXGBE_TX_FLAGS_IPV4,
7137					IXGBE_ADVTXD_POPTS_IXSM);
7138
7139	/*
7140	 * Check Context must be set if Tx switch is enabled, which it
7141	 * always is for case where virtual functions are running
7142	 */
7143	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7144					IXGBE_TX_FLAGS_CC,
7145					IXGBE_ADVTXD_CC);
7146
7147	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7148}
7149
7150static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7151{
7152	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7153
7154	/* Herbert's original patch had:
7155	 *  smp_mb__after_netif_stop_queue();
7156	 * but since that doesn't exist yet, just open code it.
7157	 */
7158	smp_mb();
7159
7160	/* We need to check again in a case another CPU has just
7161	 * made room available.
7162	 */
7163	if (likely(ixgbe_desc_unused(tx_ring) < size))
7164		return -EBUSY;
7165
7166	/* A reprieve! - use start_queue because it doesn't call schedule */
7167	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7168	++tx_ring->tx_stats.restart_queue;
7169	return 0;
7170}
7171
7172static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7173{
7174	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7175		return 0;
7176
7177	return __ixgbe_maybe_stop_tx(tx_ring, size);
7178}
7179
7180#define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7181		       IXGBE_TXD_CMD_RS)
7182
7183static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7184			 struct ixgbe_tx_buffer *first,
7185			 const u8 hdr_len)
7186{
7187	struct sk_buff *skb = first->skb;
7188	struct ixgbe_tx_buffer *tx_buffer;
7189	union ixgbe_adv_tx_desc *tx_desc;
7190	struct skb_frag_struct *frag;
7191	dma_addr_t dma;
7192	unsigned int data_len, size;
7193	u32 tx_flags = first->tx_flags;
7194	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7195	u16 i = tx_ring->next_to_use;
7196
7197	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7198
7199	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7200
7201	size = skb_headlen(skb);
7202	data_len = skb->data_len;
7203
7204#ifdef IXGBE_FCOE
7205	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7206		if (data_len < sizeof(struct fcoe_crc_eof)) {
7207			size -= sizeof(struct fcoe_crc_eof) - data_len;
7208			data_len = 0;
7209		} else {
7210			data_len -= sizeof(struct fcoe_crc_eof);
7211		}
7212	}
7213
7214#endif
7215	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7216
7217	tx_buffer = first;
7218
7219	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7220		if (dma_mapping_error(tx_ring->dev, dma))
7221			goto dma_error;
7222
7223		/* record length, and DMA address */
7224		dma_unmap_len_set(tx_buffer, len, size);
7225		dma_unmap_addr_set(tx_buffer, dma, dma);
7226
7227		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7228
7229		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7230			tx_desc->read.cmd_type_len =
7231				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7232
7233			i++;
7234			tx_desc++;
7235			if (i == tx_ring->count) {
7236				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7237				i = 0;
7238			}
7239			tx_desc->read.olinfo_status = 0;
7240
7241			dma += IXGBE_MAX_DATA_PER_TXD;
7242			size -= IXGBE_MAX_DATA_PER_TXD;
7243
7244			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7245		}
7246
7247		if (likely(!data_len))
7248			break;
7249
7250		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7251
7252		i++;
7253		tx_desc++;
7254		if (i == tx_ring->count) {
7255			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7256			i = 0;
7257		}
7258		tx_desc->read.olinfo_status = 0;
7259
7260#ifdef IXGBE_FCOE
7261		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7262#else
7263		size = skb_frag_size(frag);
7264#endif
7265		data_len -= size;
7266
7267		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7268				       DMA_TO_DEVICE);
7269
7270		tx_buffer = &tx_ring->tx_buffer_info[i];
7271	}
7272
7273	/* write last descriptor with RS and EOP bits */
7274	cmd_type |= size | IXGBE_TXD_CMD;
7275	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7276
7277	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7278
7279	/* set the timestamp */
7280	first->time_stamp = jiffies;
7281
7282	/*
7283	 * Force memory writes to complete before letting h/w know there
7284	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7285	 * memory model archs, such as IA-64).
7286	 *
7287	 * We also need this memory barrier to make certain all of the
7288	 * status bits have been updated before next_to_watch is written.
7289	 */
7290	wmb();
7291
7292	/* set next_to_watch value indicating a packet is present */
7293	first->next_to_watch = tx_desc;
7294
7295	i++;
7296	if (i == tx_ring->count)
7297		i = 0;
7298
7299	tx_ring->next_to_use = i;
7300
7301	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7302
7303	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7304		writel(i, tx_ring->tail);
7305
7306		/* we need this if more than one processor can write to our tail
7307		 * at a time, it synchronizes IO on IA64/Altix systems
7308		 */
7309		mmiowb();
7310	}
7311
7312	return;
7313dma_error:
7314	dev_err(tx_ring->dev, "TX DMA map failed\n");
7315
7316	/* clear dma mappings for failed tx_buffer_info map */
7317	for (;;) {
7318		tx_buffer = &tx_ring->tx_buffer_info[i];
7319		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
7320		if (tx_buffer == first)
7321			break;
7322		if (i == 0)
7323			i = tx_ring->count;
7324		i--;
7325	}
7326
7327	tx_ring->next_to_use = i;
7328}
7329
7330static void ixgbe_atr(struct ixgbe_ring *ring,
7331		      struct ixgbe_tx_buffer *first)
7332{
7333	struct ixgbe_q_vector *q_vector = ring->q_vector;
7334	union ixgbe_atr_hash_dword input = { .dword = 0 };
7335	union ixgbe_atr_hash_dword common = { .dword = 0 };
7336	union {
7337		unsigned char *network;
7338		struct iphdr *ipv4;
7339		struct ipv6hdr *ipv6;
7340	} hdr;
7341	struct tcphdr *th;
7342	struct sk_buff *skb;
7343#ifdef CONFIG_IXGBE_VXLAN
7344	u8 encap = false;
7345#endif /* CONFIG_IXGBE_VXLAN */
7346	__be16 vlan_id;
7347
7348	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7349	if (!q_vector)
7350		return;
7351
7352	/* do nothing if sampling is disabled */
7353	if (!ring->atr_sample_rate)
7354		return;
7355
7356	ring->atr_count++;
7357
7358	/* snag network header to get L4 type and address */
7359	skb = first->skb;
7360	hdr.network = skb_network_header(skb);
7361	if (skb->encapsulation) {
7362#ifdef CONFIG_IXGBE_VXLAN
7363		struct ixgbe_adapter *adapter = q_vector->adapter;
7364
7365		if (!adapter->vxlan_port)
7366			return;
7367		if (first->protocol != htons(ETH_P_IP) ||
7368		    hdr.ipv4->version != IPVERSION ||
7369		    hdr.ipv4->protocol != IPPROTO_UDP) {
7370			return;
7371		}
7372		if (ntohs(udp_hdr(skb)->dest) != adapter->vxlan_port)
7373			return;
7374		encap = true;
7375		hdr.network = skb_inner_network_header(skb);
7376		th = inner_tcp_hdr(skb);
7377#else
7378		return;
7379#endif /* CONFIG_IXGBE_VXLAN */
7380	} else {
7381		/* Currently only IPv4/IPv6 with TCP is supported */
7382		if ((first->protocol != htons(ETH_P_IPV6) ||
7383		     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
7384		    (first->protocol != htons(ETH_P_IP) ||
7385		     hdr.ipv4->protocol != IPPROTO_TCP))
7386			return;
7387		th = tcp_hdr(skb);
7388	}
7389
7390	/* skip this packet since it is invalid or the socket is closing */
7391	if (!th || th->fin)
7392		return;
7393
7394	/* sample on all syn packets or once every atr sample count */
7395	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7396		return;
7397
7398	/* reset sample count */
7399	ring->atr_count = 0;
7400
7401	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7402
7403	/*
7404	 * src and dst are inverted, think how the receiver sees them
7405	 *
7406	 * The input is broken into two sections, a non-compressed section
7407	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7408	 * is XORed together and stored in the compressed dword.
7409	 */
7410	input.formatted.vlan_id = vlan_id;
7411
7412	/*
7413	 * since src port and flex bytes occupy the same word XOR them together
7414	 * and write the value to source port portion of compressed dword
7415	 */
7416	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7417		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7418	else
7419		common.port.src ^= th->dest ^ first->protocol;
7420	common.port.dst ^= th->source;
7421
7422	if (first->protocol == htons(ETH_P_IP)) {
7423		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7424		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7425	} else {
7426		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7427		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7428			     hdr.ipv6->saddr.s6_addr32[1] ^
7429			     hdr.ipv6->saddr.s6_addr32[2] ^
7430			     hdr.ipv6->saddr.s6_addr32[3] ^
7431			     hdr.ipv6->daddr.s6_addr32[0] ^
7432			     hdr.ipv6->daddr.s6_addr32[1] ^
7433			     hdr.ipv6->daddr.s6_addr32[2] ^
7434			     hdr.ipv6->daddr.s6_addr32[3];
7435	}
7436
7437#ifdef CONFIG_IXGBE_VXLAN
7438	if (encap)
7439		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
7440#endif /* CONFIG_IXGBE_VXLAN */
7441
7442	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7443	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7444					      input, common, ring->queue_index);
7445}
7446
7447static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7448			      void *accel_priv, select_queue_fallback_t fallback)
7449{
7450	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7451#ifdef IXGBE_FCOE
7452	struct ixgbe_adapter *adapter;
7453	struct ixgbe_ring_feature *f;
7454	int txq;
7455#endif
7456
7457	if (fwd_adapter)
7458		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7459
7460#ifdef IXGBE_FCOE
7461
7462	/*
7463	 * only execute the code below if protocol is FCoE
7464	 * or FIP and we have FCoE enabled on the adapter
7465	 */
7466	switch (vlan_get_protocol(skb)) {
7467	case htons(ETH_P_FCOE):
7468	case htons(ETH_P_FIP):
7469		adapter = netdev_priv(dev);
7470
7471		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7472			break;
7473	default:
7474		return fallback(dev, skb);
7475	}
7476
7477	f = &adapter->ring_feature[RING_F_FCOE];
7478
7479	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7480					   smp_processor_id();
7481
7482	while (txq >= f->indices)
7483		txq -= f->indices;
7484
7485	return txq + f->offset;
7486#else
7487	return fallback(dev, skb);
7488#endif
7489}
7490
7491netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7492			  struct ixgbe_adapter *adapter,
7493			  struct ixgbe_ring *tx_ring)
7494{
7495	struct ixgbe_tx_buffer *first;
7496	int tso;
7497	u32 tx_flags = 0;
7498	unsigned short f;
7499	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7500	__be16 protocol = skb->protocol;
7501	u8 hdr_len = 0;
7502
7503	/*
7504	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7505	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7506	 *       + 2 desc gap to keep tail from touching head,
7507	 *       + 1 desc for context descriptor,
7508	 * otherwise try next time
7509	 */
7510	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7511		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7512
7513	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7514		tx_ring->tx_stats.tx_busy++;
7515		return NETDEV_TX_BUSY;
7516	}
7517
7518	/* record the location of the first descriptor for this packet */
7519	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7520	first->skb = skb;
7521	first->bytecount = skb->len;
7522	first->gso_segs = 1;
7523
7524	/* if we have a HW VLAN tag being added default to the HW one */
7525	if (skb_vlan_tag_present(skb)) {
7526		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7527		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7528	/* else if it is a SW VLAN check the next protocol and store the tag */
7529	} else if (protocol == htons(ETH_P_8021Q)) {
7530		struct vlan_hdr *vhdr, _vhdr;
7531		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7532		if (!vhdr)
7533			goto out_drop;
7534
7535		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7536				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7537		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7538	}
7539	protocol = vlan_get_protocol(skb);
7540
7541	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
7542	    adapter->ptp_clock &&
7543	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7544				   &adapter->state)) {
7545		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7546		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7547
7548		/* schedule check for Tx timestamp */
7549		adapter->ptp_tx_skb = skb_get(skb);
7550		adapter->ptp_tx_start = jiffies;
7551		schedule_work(&adapter->ptp_tx_work);
7552	}
7553
7554	skb_tx_timestamp(skb);
7555
7556#ifdef CONFIG_PCI_IOV
7557	/*
7558	 * Use the l2switch_enable flag - would be false if the DMA
7559	 * Tx switch had been disabled.
7560	 */
7561	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7562		tx_flags |= IXGBE_TX_FLAGS_CC;
7563
7564#endif
7565	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7566	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7567	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7568	     (skb->priority != TC_PRIO_CONTROL))) {
7569		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7570		tx_flags |= (skb->priority & 0x7) <<
7571					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7572		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7573			struct vlan_ethhdr *vhdr;
7574
7575			if (skb_cow_head(skb, 0))
7576				goto out_drop;
7577			vhdr = (struct vlan_ethhdr *)skb->data;
7578			vhdr->h_vlan_TCI = htons(tx_flags >>
7579						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7580		} else {
7581			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7582		}
7583	}
7584
7585	/* record initial flags and protocol */
7586	first->tx_flags = tx_flags;
7587	first->protocol = protocol;
7588
7589#ifdef IXGBE_FCOE
7590	/* setup tx offload for FCoE */
7591	if ((protocol == htons(ETH_P_FCOE)) &&
7592	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7593		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7594		if (tso < 0)
7595			goto out_drop;
7596
7597		goto xmit_fcoe;
7598	}
7599
7600#endif /* IXGBE_FCOE */
7601	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7602	if (tso < 0)
7603		goto out_drop;
7604	else if (!tso)
7605		ixgbe_tx_csum(tx_ring, first);
7606
7607	/* add the ATR filter if ATR is on */
7608	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7609		ixgbe_atr(tx_ring, first);
7610
7611#ifdef IXGBE_FCOE
7612xmit_fcoe:
7613#endif /* IXGBE_FCOE */
7614	ixgbe_tx_map(tx_ring, first, hdr_len);
7615
7616	return NETDEV_TX_OK;
7617
7618out_drop:
7619	dev_kfree_skb_any(first->skb);
7620	first->skb = NULL;
7621
7622	return NETDEV_TX_OK;
7623}
7624
7625static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7626				      struct net_device *netdev,
7627				      struct ixgbe_ring *ring)
7628{
7629	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7630	struct ixgbe_ring *tx_ring;
7631
7632	/*
7633	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7634	 * in order to meet this minimum size requirement.
7635	 */
7636	if (skb_put_padto(skb, 17))
7637		return NETDEV_TX_OK;
7638
7639	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7640
7641	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7642}
7643
7644static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7645				    struct net_device *netdev)
7646{
7647	return __ixgbe_xmit_frame(skb, netdev, NULL);
7648}
7649
7650/**
7651 * ixgbe_set_mac - Change the Ethernet Address of the NIC
7652 * @netdev: network interface device structure
7653 * @p: pointer to an address structure
7654 *
7655 * Returns 0 on success, negative on failure
7656 **/
7657static int ixgbe_set_mac(struct net_device *netdev, void *p)
7658{
7659	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7660	struct ixgbe_hw *hw = &adapter->hw;
7661	struct sockaddr *addr = p;
7662	int ret;
7663
7664	if (!is_valid_ether_addr(addr->sa_data))
7665		return -EADDRNOTAVAIL;
7666
7667	ixgbe_del_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7668	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
7669	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
7670
7671	ret = ixgbe_add_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7672	return ret > 0 ? 0 : ret;
7673}
7674
7675static int
7676ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
7677{
7678	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7679	struct ixgbe_hw *hw = &adapter->hw;
7680	u16 value;
7681	int rc;
7682
7683	if (prtad != hw->phy.mdio.prtad)
7684		return -EINVAL;
7685	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
7686	if (!rc)
7687		rc = value;
7688	return rc;
7689}
7690
7691static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
7692			    u16 addr, u16 value)
7693{
7694	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7695	struct ixgbe_hw *hw = &adapter->hw;
7696
7697	if (prtad != hw->phy.mdio.prtad)
7698		return -EINVAL;
7699	return hw->phy.ops.write_reg(hw, addr, devad, value);
7700}
7701
7702static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
7703{
7704	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7705
7706	switch (cmd) {
7707	case SIOCSHWTSTAMP:
7708		return ixgbe_ptp_set_ts_config(adapter, req);
7709	case SIOCGHWTSTAMP:
7710		return ixgbe_ptp_get_ts_config(adapter, req);
7711	default:
7712		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
7713	}
7714}
7715
7716/**
7717 * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
7718 * netdev->dev_addrs
7719 * @netdev: network interface device structure
7720 *
7721 * Returns non-zero on failure
7722 **/
7723static int ixgbe_add_sanmac_netdev(struct net_device *dev)
7724{
7725	int err = 0;
7726	struct ixgbe_adapter *adapter = netdev_priv(dev);
7727	struct ixgbe_hw *hw = &adapter->hw;
7728
7729	if (is_valid_ether_addr(hw->mac.san_addr)) {
7730		rtnl_lock();
7731		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
7732		rtnl_unlock();
7733
7734		/* update SAN MAC vmdq pool selection */
7735		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
7736	}
7737	return err;
7738}
7739
7740/**
7741 * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
7742 * netdev->dev_addrs
7743 * @netdev: network interface device structure
7744 *
7745 * Returns non-zero on failure
7746 **/
7747static int ixgbe_del_sanmac_netdev(struct net_device *dev)
7748{
7749	int err = 0;
7750	struct ixgbe_adapter *adapter = netdev_priv(dev);
7751	struct ixgbe_mac_info *mac = &adapter->hw.mac;
7752
7753	if (is_valid_ether_addr(mac->san_addr)) {
7754		rtnl_lock();
7755		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
7756		rtnl_unlock();
7757	}
7758	return err;
7759}
7760
7761#ifdef CONFIG_NET_POLL_CONTROLLER
7762/*
7763 * Polling 'interrupt' - used by things like netconsole to send skbs
7764 * without having to re-enable interrupts. It's not called while
7765 * the interrupt routine is executing.
7766 */
7767static void ixgbe_netpoll(struct net_device *netdev)
7768{
7769	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7770	int i;
7771
7772	/* if interface is down do nothing */
7773	if (test_bit(__IXGBE_DOWN, &adapter->state))
7774		return;
7775
7776	/* loop through and schedule all active queues */
7777	for (i = 0; i < adapter->num_q_vectors; i++)
7778		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
7779}
7780
7781#endif
7782static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
7783						   struct rtnl_link_stats64 *stats)
7784{
7785	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7786	int i;
7787
7788	rcu_read_lock();
7789	for (i = 0; i < adapter->num_rx_queues; i++) {
7790		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
7791		u64 bytes, packets;
7792		unsigned int start;
7793
7794		if (ring) {
7795			do {
7796				start = u64_stats_fetch_begin_irq(&ring->syncp);
7797				packets = ring->stats.packets;
7798				bytes   = ring->stats.bytes;
7799			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7800			stats->rx_packets += packets;
7801			stats->rx_bytes   += bytes;
7802		}
7803	}
7804
7805	for (i = 0; i < adapter->num_tx_queues; i++) {
7806		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
7807		u64 bytes, packets;
7808		unsigned int start;
7809
7810		if (ring) {
7811			do {
7812				start = u64_stats_fetch_begin_irq(&ring->syncp);
7813				packets = ring->stats.packets;
7814				bytes   = ring->stats.bytes;
7815			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7816			stats->tx_packets += packets;
7817			stats->tx_bytes   += bytes;
7818		}
7819	}
7820	rcu_read_unlock();
7821	/* following stats updated by ixgbe_watchdog_task() */
7822	stats->multicast	= netdev->stats.multicast;
7823	stats->rx_errors	= netdev->stats.rx_errors;
7824	stats->rx_length_errors	= netdev->stats.rx_length_errors;
7825	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
7826	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
7827	return stats;
7828}
7829
7830#ifdef CONFIG_IXGBE_DCB
7831/**
7832 * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
7833 * @adapter: pointer to ixgbe_adapter
7834 * @tc: number of traffic classes currently enabled
7835 *
7836 * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
7837 * 802.1Q priority maps to a packet buffer that exists.
7838 */
7839static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
7840{
7841	struct ixgbe_hw *hw = &adapter->hw;
7842	u32 reg, rsave;
7843	int i;
7844
7845	/* 82598 have a static priority to TC mapping that can not
7846	 * be changed so no validation is needed.
7847	 */
7848	if (hw->mac.type == ixgbe_mac_82598EB)
7849		return;
7850
7851	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
7852	rsave = reg;
7853
7854	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
7855		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
7856
7857		/* If up2tc is out of bounds default to zero */
7858		if (up2tc > tc)
7859			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
7860	}
7861
7862	if (reg != rsave)
7863		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
7864
7865	return;
7866}
7867
7868/**
7869 * ixgbe_set_prio_tc_map - Configure netdev prio tc map
7870 * @adapter: Pointer to adapter struct
7871 *
7872 * Populate the netdev user priority to tc map
7873 */
7874static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
7875{
7876	struct net_device *dev = adapter->netdev;
7877	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
7878	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
7879	u8 prio;
7880
7881	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
7882		u8 tc = 0;
7883
7884		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
7885			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
7886		else if (ets)
7887			tc = ets->prio_tc[prio];
7888
7889		netdev_set_prio_tc_map(dev, prio, tc);
7890	}
7891}
7892
7893#endif /* CONFIG_IXGBE_DCB */
7894/**
7895 * ixgbe_setup_tc - configure net_device for multiple traffic classes
7896 *
7897 * @netdev: net device to configure
7898 * @tc: number of traffic classes to enable
7899 */
7900int ixgbe_setup_tc(struct net_device *dev, u8 tc)
7901{
7902	struct ixgbe_adapter *adapter = netdev_priv(dev);
7903	struct ixgbe_hw *hw = &adapter->hw;
7904	bool pools;
7905
7906	/* Hardware supports up to 8 traffic classes */
7907	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
7908		return -EINVAL;
7909
7910	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
7911		return -EINVAL;
7912
7913	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
7914	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
7915		return -EBUSY;
7916
7917	/* Hardware has to reinitialize queues and interrupts to
7918	 * match packet buffer alignment. Unfortunately, the
7919	 * hardware is not flexible enough to do this dynamically.
7920	 */
7921	if (netif_running(dev))
7922		ixgbe_close(dev);
7923	else
7924		ixgbe_reset(adapter);
7925
7926	ixgbe_clear_interrupt_scheme(adapter);
7927
7928#ifdef CONFIG_IXGBE_DCB
7929	if (tc) {
7930		netdev_set_num_tc(dev, tc);
7931		ixgbe_set_prio_tc_map(adapter);
7932
7933		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
7934
7935		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
7936			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
7937			adapter->hw.fc.requested_mode = ixgbe_fc_none;
7938		}
7939	} else {
7940		netdev_reset_tc(dev);
7941
7942		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
7943			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
7944
7945		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
7946
7947		adapter->temp_dcb_cfg.pfc_mode_enable = false;
7948		adapter->dcb_cfg.pfc_mode_enable = false;
7949	}
7950
7951	ixgbe_validate_rtr(adapter, tc);
7952
7953#endif /* CONFIG_IXGBE_DCB */
7954	ixgbe_init_interrupt_scheme(adapter);
7955
7956	if (netif_running(dev))
7957		return ixgbe_open(dev);
7958
7959	return 0;
7960}
7961
7962#ifdef CONFIG_PCI_IOV
7963void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
7964{
7965	struct net_device *netdev = adapter->netdev;
7966
7967	rtnl_lock();
7968	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
7969	rtnl_unlock();
7970}
7971
7972#endif
7973void ixgbe_do_reset(struct net_device *netdev)
7974{
7975	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7976
7977	if (netif_running(netdev))
7978		ixgbe_reinit_locked(adapter);
7979	else
7980		ixgbe_reset(adapter);
7981}
7982
7983static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
7984					    netdev_features_t features)
7985{
7986	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7987
7988	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
7989	if (!(features & NETIF_F_RXCSUM))
7990		features &= ~NETIF_F_LRO;
7991
7992	/* Turn off LRO if not RSC capable */
7993	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
7994		features &= ~NETIF_F_LRO;
7995
7996	return features;
7997}
7998
7999static int ixgbe_set_features(struct net_device *netdev,
8000			      netdev_features_t features)
8001{
8002	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8003	netdev_features_t changed = netdev->features ^ features;
8004	bool need_reset = false;
8005
8006	/* Make sure RSC matches LRO, reset if change */
8007	if (!(features & NETIF_F_LRO)) {
8008		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8009			need_reset = true;
8010		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
8011	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
8012		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
8013		if (adapter->rx_itr_setting == 1 ||
8014		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
8015			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
8016			need_reset = true;
8017		} else if ((changed ^ features) & NETIF_F_LRO) {
8018			e_info(probe, "rx-usecs set too low, "
8019			       "disabling RSC\n");
8020		}
8021	}
8022
8023	/*
8024	 * Check if Flow Director n-tuple support was enabled or disabled.  If
8025	 * the state changed, we need to reset.
8026	 */
8027	switch (features & NETIF_F_NTUPLE) {
8028	case NETIF_F_NTUPLE:
8029		/* turn off ATR, enable perfect filters and reset */
8030		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
8031			need_reset = true;
8032
8033		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
8034		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8035		break;
8036	default:
8037		/* turn off perfect filters, enable ATR and reset */
8038		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
8039			need_reset = true;
8040
8041		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8042
8043		/* We cannot enable ATR if SR-IOV is enabled */
8044		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8045			break;
8046
8047		/* We cannot enable ATR if we have 2 or more traffic classes */
8048		if (netdev_get_num_tc(netdev) > 1)
8049			break;
8050
8051		/* We cannot enable ATR if RSS is disabled */
8052		if (adapter->ring_feature[RING_F_RSS].limit <= 1)
8053			break;
8054
8055		/* A sample rate of 0 indicates ATR disabled */
8056		if (!adapter->atr_sample_rate)
8057			break;
8058
8059		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
8060		break;
8061	}
8062
8063	if (features & NETIF_F_HW_VLAN_CTAG_RX)
8064		ixgbe_vlan_strip_enable(adapter);
8065	else
8066		ixgbe_vlan_strip_disable(adapter);
8067
8068	if (changed & NETIF_F_RXALL)
8069		need_reset = true;
8070
8071	netdev->features = features;
8072
8073#ifdef CONFIG_IXGBE_VXLAN
8074	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
8075		if (features & NETIF_F_RXCSUM)
8076			adapter->flags2 |= IXGBE_FLAG2_VXLAN_REREG_NEEDED;
8077		else
8078			ixgbe_clear_vxlan_port(adapter);
8079	}
8080#endif /* CONFIG_IXGBE_VXLAN */
8081
8082	if (need_reset)
8083		ixgbe_do_reset(netdev);
8084
8085	return 0;
8086}
8087
8088#ifdef CONFIG_IXGBE_VXLAN
8089/**
8090 * ixgbe_add_vxlan_port - Get notifications about VXLAN ports that come up
8091 * @dev: The port's netdev
8092 * @sa_family: Socket Family that VXLAN is notifiying us about
8093 * @port: New UDP port number that VXLAN started listening to
8094 **/
8095static void ixgbe_add_vxlan_port(struct net_device *dev, sa_family_t sa_family,
8096				 __be16 port)
8097{
8098	struct ixgbe_adapter *adapter = netdev_priv(dev);
8099	struct ixgbe_hw *hw = &adapter->hw;
8100	u16 new_port = ntohs(port);
8101
8102	if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8103		return;
8104
8105	if (sa_family == AF_INET6)
8106		return;
8107
8108	if (adapter->vxlan_port == new_port)
8109		return;
8110
8111	if (adapter->vxlan_port) {
8112		netdev_info(dev,
8113			    "Hit Max num of VXLAN ports, not adding port %d\n",
8114			    new_port);
8115		return;
8116	}
8117
8118	adapter->vxlan_port = new_port;
8119	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, new_port);
8120}
8121
8122/**
8123 * ixgbe_del_vxlan_port - Get notifications about VXLAN ports that go away
8124 * @dev: The port's netdev
8125 * @sa_family: Socket Family that VXLAN is notifying us about
8126 * @port: UDP port number that VXLAN stopped listening to
8127 **/
8128static void ixgbe_del_vxlan_port(struct net_device *dev, sa_family_t sa_family,
8129				 __be16 port)
8130{
8131	struct ixgbe_adapter *adapter = netdev_priv(dev);
8132	u16 new_port = ntohs(port);
8133
8134	if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8135		return;
8136
8137	if (sa_family == AF_INET6)
8138		return;
8139
8140	if (adapter->vxlan_port != new_port) {
8141		netdev_info(dev, "Port %d was not found, not deleting\n",
8142			    new_port);
8143		return;
8144	}
8145
8146	ixgbe_clear_vxlan_port(adapter);
8147	adapter->flags2 |= IXGBE_FLAG2_VXLAN_REREG_NEEDED;
8148}
8149#endif /* CONFIG_IXGBE_VXLAN */
8150
8151static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8152			     struct net_device *dev,
8153			     const unsigned char *addr, u16 vid,
8154			     u16 flags)
8155{
8156	/* guarantee we can provide a unique filter for the unicast address */
8157	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
8158		if (IXGBE_MAX_PF_MACVLANS <= netdev_uc_count(dev))
8159			return -ENOMEM;
8160	}
8161
8162	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
8163}
8164
8165/**
8166 * ixgbe_configure_bridge_mode - set various bridge modes
8167 * @adapter - the private structure
8168 * @mode - requested bridge mode
8169 *
8170 * Configure some settings require for various bridge modes.
8171 **/
8172static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
8173				       __u16 mode)
8174{
8175	struct ixgbe_hw *hw = &adapter->hw;
8176	unsigned int p, num_pools;
8177	u32 vmdctl;
8178
8179	switch (mode) {
8180	case BRIDGE_MODE_VEPA:
8181		/* disable Tx loopback, rely on switch hairpin mode */
8182		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
8183
8184		/* must enable Rx switching replication to allow multicast
8185		 * packet reception on all VFs, and to enable source address
8186		 * pruning.
8187		 */
8188		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
8189		vmdctl |= IXGBE_VT_CTL_REPLEN;
8190		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
8191
8192		/* enable Rx source address pruning. Note, this requires
8193		 * replication to be enabled or else it does nothing.
8194		 */
8195		num_pools = adapter->num_vfs + adapter->num_rx_pools;
8196		for (p = 0; p < num_pools; p++) {
8197			if (hw->mac.ops.set_source_address_pruning)
8198				hw->mac.ops.set_source_address_pruning(hw,
8199								       true,
8200								       p);
8201		}
8202		break;
8203	case BRIDGE_MODE_VEB:
8204		/* enable Tx loopback for internal VF/PF communication */
8205		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
8206				IXGBE_PFDTXGSWC_VT_LBEN);
8207
8208		/* disable Rx switching replication unless we have SR-IOV
8209		 * virtual functions
8210		 */
8211		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
8212		if (!adapter->num_vfs)
8213			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
8214		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
8215
8216		/* disable Rx source address pruning, since we don't expect to
8217		 * be receiving external loopback of our transmitted frames.
8218		 */
8219		num_pools = adapter->num_vfs + adapter->num_rx_pools;
8220		for (p = 0; p < num_pools; p++) {
8221			if (hw->mac.ops.set_source_address_pruning)
8222				hw->mac.ops.set_source_address_pruning(hw,
8223								       false,
8224								       p);
8225		}
8226		break;
8227	default:
8228		return -EINVAL;
8229	}
8230
8231	adapter->bridge_mode = mode;
8232
8233	e_info(drv, "enabling bridge mode: %s\n",
8234	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
8235
8236	return 0;
8237}
8238
8239static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
8240				    struct nlmsghdr *nlh, u16 flags)
8241{
8242	struct ixgbe_adapter *adapter = netdev_priv(dev);
8243	struct nlattr *attr, *br_spec;
8244	int rem;
8245
8246	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
8247		return -EOPNOTSUPP;
8248
8249	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8250	if (!br_spec)
8251		return -EINVAL;
8252
8253	nla_for_each_nested(attr, br_spec, rem) {
8254		int status;
8255		__u16 mode;
8256
8257		if (nla_type(attr) != IFLA_BRIDGE_MODE)
8258			continue;
8259
8260		if (nla_len(attr) < sizeof(mode))
8261			return -EINVAL;
8262
8263		mode = nla_get_u16(attr);
8264		status = ixgbe_configure_bridge_mode(adapter, mode);
8265		if (status)
8266			return status;
8267
8268		break;
8269	}
8270
8271	return 0;
8272}
8273
8274static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8275				    struct net_device *dev,
8276				    u32 filter_mask, int nlflags)
8277{
8278	struct ixgbe_adapter *adapter = netdev_priv(dev);
8279
8280	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
8281		return 0;
8282
8283	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
8284				       adapter->bridge_mode, 0, 0, nlflags,
8285				       filter_mask, NULL);
8286}
8287
8288static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
8289{
8290	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
8291	struct ixgbe_adapter *adapter = netdev_priv(pdev);
8292	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
8293	unsigned int limit;
8294	int pool, err;
8295
8296	/* Hardware has a limited number of available pools. Each VF, and the
8297	 * PF require a pool. Check to ensure we don't attempt to use more
8298	 * then the available number of pools.
8299	 */
8300	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
8301		return ERR_PTR(-EINVAL);
8302
8303#ifdef CONFIG_RPS
8304	if (vdev->num_rx_queues != vdev->num_tx_queues) {
8305		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
8306			    vdev->name);
8307		return ERR_PTR(-EINVAL);
8308	}
8309#endif
8310	/* Check for hardware restriction on number of rx/tx queues */
8311	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
8312	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
8313		netdev_info(pdev,
8314			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
8315			    pdev->name);
8316		return ERR_PTR(-EINVAL);
8317	}
8318
8319	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8320	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
8321	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
8322		return ERR_PTR(-EBUSY);
8323
8324	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
8325	if (!fwd_adapter)
8326		return ERR_PTR(-ENOMEM);
8327
8328	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
8329	adapter->num_rx_pools++;
8330	set_bit(pool, &adapter->fwd_bitmask);
8331	limit = find_last_bit(&adapter->fwd_bitmask, 32);
8332
8333	/* Enable VMDq flag so device will be set in VM mode */
8334	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
8335	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
8336	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
8337
8338	/* Force reinit of ring allocation with VMDQ enabled */
8339	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
8340	if (err)
8341		goto fwd_add_err;
8342	fwd_adapter->pool = pool;
8343	fwd_adapter->real_adapter = adapter;
8344	err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
8345	if (err)
8346		goto fwd_add_err;
8347	netif_tx_start_all_queues(vdev);
8348	return fwd_adapter;
8349fwd_add_err:
8350	/* unwind counter and free adapter struct */
8351	netdev_info(pdev,
8352		    "%s: dfwd hardware acceleration failed\n", vdev->name);
8353	clear_bit(pool, &adapter->fwd_bitmask);
8354	adapter->num_rx_pools--;
8355	kfree(fwd_adapter);
8356	return ERR_PTR(err);
8357}
8358
8359static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
8360{
8361	struct ixgbe_fwd_adapter *fwd_adapter = priv;
8362	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
8363	unsigned int limit;
8364
8365	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
8366	adapter->num_rx_pools--;
8367
8368	limit = find_last_bit(&adapter->fwd_bitmask, 32);
8369	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
8370	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
8371	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
8372	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
8373		   fwd_adapter->pool, adapter->num_rx_pools,
8374		   fwd_adapter->rx_base_queue,
8375		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
8376		   adapter->fwd_bitmask);
8377	kfree(fwd_adapter);
8378}
8379
8380#define IXGBE_MAX_TUNNEL_HDR_LEN 80
8381static netdev_features_t
8382ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
8383		     netdev_features_t features)
8384{
8385	if (!skb->encapsulation)
8386		return features;
8387
8388	if (unlikely(skb_inner_mac_header(skb) - skb_transport_header(skb) >
8389		     IXGBE_MAX_TUNNEL_HDR_LEN))
8390		return features & ~NETIF_F_ALL_CSUM;
8391
8392	return features;
8393}
8394
8395static const struct net_device_ops ixgbe_netdev_ops = {
8396	.ndo_open		= ixgbe_open,
8397	.ndo_stop		= ixgbe_close,
8398	.ndo_start_xmit		= ixgbe_xmit_frame,
8399	.ndo_select_queue	= ixgbe_select_queue,
8400	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
8401	.ndo_validate_addr	= eth_validate_addr,
8402	.ndo_set_mac_address	= ixgbe_set_mac,
8403	.ndo_change_mtu		= ixgbe_change_mtu,
8404	.ndo_tx_timeout		= ixgbe_tx_timeout,
8405	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
8406	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
8407	.ndo_do_ioctl		= ixgbe_ioctl,
8408	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
8409	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
8410	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
8411	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
8412	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
8413	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
8414	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
8415	.ndo_get_stats64	= ixgbe_get_stats64,
8416#ifdef CONFIG_IXGBE_DCB
8417	.ndo_setup_tc		= ixgbe_setup_tc,
8418#endif
8419#ifdef CONFIG_NET_POLL_CONTROLLER
8420	.ndo_poll_controller	= ixgbe_netpoll,
8421#endif
8422#ifdef CONFIG_NET_RX_BUSY_POLL
8423	.ndo_busy_poll		= ixgbe_low_latency_recv,
8424#endif
8425#ifdef IXGBE_FCOE
8426	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
8427	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
8428	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
8429	.ndo_fcoe_enable = ixgbe_fcoe_enable,
8430	.ndo_fcoe_disable = ixgbe_fcoe_disable,
8431	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
8432	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
8433#endif /* IXGBE_FCOE */
8434	.ndo_set_features = ixgbe_set_features,
8435	.ndo_fix_features = ixgbe_fix_features,
8436	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
8437	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
8438	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
8439	.ndo_dfwd_add_station	= ixgbe_fwd_add,
8440	.ndo_dfwd_del_station	= ixgbe_fwd_del,
8441#ifdef CONFIG_IXGBE_VXLAN
8442	.ndo_add_vxlan_port	= ixgbe_add_vxlan_port,
8443	.ndo_del_vxlan_port	= ixgbe_del_vxlan_port,
8444#endif /* CONFIG_IXGBE_VXLAN */
8445	.ndo_features_check	= ixgbe_features_check,
8446};
8447
8448/**
8449 * ixgbe_enumerate_functions - Get the number of ports this device has
8450 * @adapter: adapter structure
8451 *
8452 * This function enumerates the phsyical functions co-located on a single slot,
8453 * in order to determine how many ports a device has. This is most useful in
8454 * determining the required GT/s of PCIe bandwidth necessary for optimal
8455 * performance.
8456 **/
8457static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
8458{
8459	struct pci_dev *entry, *pdev = adapter->pdev;
8460	int physfns = 0;
8461
8462	/* Some cards can not use the generic count PCIe functions method,
8463	 * because they are behind a parent switch, so we hardcode these with
8464	 * the correct number of functions.
8465	 */
8466	if (ixgbe_pcie_from_parent(&adapter->hw))
8467		physfns = 4;
8468
8469	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
8470		/* don't count virtual functions */
8471		if (entry->is_virtfn)
8472			continue;
8473
8474		/* When the devices on the bus don't all match our device ID,
8475		 * we can't reliably determine the correct number of
8476		 * functions. This can occur if a function has been direct
8477		 * attached to a virtual machine using VT-d, for example. In
8478		 * this case, simply return -1 to indicate this.
8479		 */
8480		if ((entry->vendor != pdev->vendor) ||
8481		    (entry->device != pdev->device))
8482			return -1;
8483
8484		physfns++;
8485	}
8486
8487	return physfns;
8488}
8489
8490/**
8491 * ixgbe_wol_supported - Check whether device supports WoL
8492 * @hw: hw specific details
8493 * @device_id: the device ID
8494 * @subdev_id: the subsystem device ID
8495 *
8496 * This function is used by probe and ethtool to determine
8497 * which devices have WoL support
8498 *
8499 **/
8500int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
8501			u16 subdevice_id)
8502{
8503	struct ixgbe_hw *hw = &adapter->hw;
8504	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
8505	int is_wol_supported = 0;
8506
8507	switch (device_id) {
8508	case IXGBE_DEV_ID_82599_SFP:
8509		/* Only these subdevices could supports WOL */
8510		switch (subdevice_id) {
8511		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
8512		case IXGBE_SUBDEV_ID_82599_560FLR:
8513			/* only support first port */
8514			if (hw->bus.func != 0)
8515				break;
8516		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
8517		case IXGBE_SUBDEV_ID_82599_SFP:
8518		case IXGBE_SUBDEV_ID_82599_RNDC:
8519		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
8520		case IXGBE_SUBDEV_ID_82599_LOM_SFP:
8521			is_wol_supported = 1;
8522			break;
8523		}
8524		break;
8525	case IXGBE_DEV_ID_82599EN_SFP:
8526		/* Only this subdevice supports WOL */
8527		switch (subdevice_id) {
8528		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
8529			is_wol_supported = 1;
8530			break;
8531		}
8532		break;
8533	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
8534		/* All except this subdevice support WOL */
8535		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
8536			is_wol_supported = 1;
8537		break;
8538	case IXGBE_DEV_ID_82599_KX4:
8539		is_wol_supported = 1;
8540		break;
8541	case IXGBE_DEV_ID_X540T:
8542	case IXGBE_DEV_ID_X540T1:
8543	case IXGBE_DEV_ID_X550T:
8544	case IXGBE_DEV_ID_X550EM_X_KX4:
8545	case IXGBE_DEV_ID_X550EM_X_KR:
8546	case IXGBE_DEV_ID_X550EM_X_10G_T:
8547		/* check eeprom to see if enabled wol */
8548		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
8549		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
8550		     (hw->bus.func == 0))) {
8551			is_wol_supported = 1;
8552		}
8553		break;
8554	}
8555
8556	return is_wol_supported;
8557}
8558
8559/**
8560 * ixgbe_get_platform_mac_addr - Look up MAC address in Open Firmware / IDPROM
8561 * @adapter: Pointer to adapter struct
8562 */
8563static void ixgbe_get_platform_mac_addr(struct ixgbe_adapter *adapter)
8564{
8565#ifdef CONFIG_OF
8566	struct device_node *dp = pci_device_to_OF_node(adapter->pdev);
8567	struct ixgbe_hw *hw = &adapter->hw;
8568	const unsigned char *addr;
8569
8570	addr = of_get_mac_address(dp);
8571	if (addr) {
8572		ether_addr_copy(hw->mac.perm_addr, addr);
8573		return;
8574	}
8575#endif /* CONFIG_OF */
8576
8577#ifdef CONFIG_SPARC
8578	ether_addr_copy(hw->mac.perm_addr, idprom->id_ethaddr);
8579#endif /* CONFIG_SPARC */
8580}
8581
8582/**
8583 * ixgbe_probe - Device Initialization Routine
8584 * @pdev: PCI device information struct
8585 * @ent: entry in ixgbe_pci_tbl
8586 *
8587 * Returns 0 on success, negative on failure
8588 *
8589 * ixgbe_probe initializes an adapter identified by a pci_dev structure.
8590 * The OS initialization, configuring of the adapter private structure,
8591 * and a hardware reset occur.
8592 **/
8593static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
8594{
8595	struct net_device *netdev;
8596	struct ixgbe_adapter *adapter = NULL;
8597	struct ixgbe_hw *hw;
8598	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
8599	int i, err, pci_using_dac, expected_gts;
8600	unsigned int indices = MAX_TX_QUEUES;
8601	u8 part_str[IXGBE_PBANUM_LENGTH];
8602	bool disable_dev = false;
8603#ifdef IXGBE_FCOE
8604	u16 device_caps;
8605#endif
8606	u32 eec;
8607
8608	/* Catch broken hardware that put the wrong VF device ID in
8609	 * the PCIe SR-IOV capability.
8610	 */
8611	if (pdev->is_virtfn) {
8612		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
8613		     pci_name(pdev), pdev->vendor, pdev->device);
8614		return -EINVAL;
8615	}
8616
8617	err = pci_enable_device_mem(pdev);
8618	if (err)
8619		return err;
8620
8621	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
8622		pci_using_dac = 1;
8623	} else {
8624		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
8625		if (err) {
8626			dev_err(&pdev->dev,
8627				"No usable DMA configuration, aborting\n");
8628			goto err_dma;
8629		}
8630		pci_using_dac = 0;
8631	}
8632
8633	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
8634					   IORESOURCE_MEM), ixgbe_driver_name);
8635	if (err) {
8636		dev_err(&pdev->dev,
8637			"pci_request_selected_regions failed 0x%x\n", err);
8638		goto err_pci_reg;
8639	}
8640
8641	pci_enable_pcie_error_reporting(pdev);
8642
8643	pci_set_master(pdev);
8644	pci_save_state(pdev);
8645
8646	if (ii->mac == ixgbe_mac_82598EB) {
8647#ifdef CONFIG_IXGBE_DCB
8648		/* 8 TC w/ 4 queues per TC */
8649		indices = 4 * MAX_TRAFFIC_CLASS;
8650#else
8651		indices = IXGBE_MAX_RSS_INDICES;
8652#endif
8653	}
8654
8655	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
8656	if (!netdev) {
8657		err = -ENOMEM;
8658		goto err_alloc_etherdev;
8659	}
8660
8661	SET_NETDEV_DEV(netdev, &pdev->dev);
8662
8663	adapter = netdev_priv(netdev);
8664
8665	adapter->netdev = netdev;
8666	adapter->pdev = pdev;
8667	hw = &adapter->hw;
8668	hw->back = adapter;
8669	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
8670
8671	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
8672			      pci_resource_len(pdev, 0));
8673	adapter->io_addr = hw->hw_addr;
8674	if (!hw->hw_addr) {
8675		err = -EIO;
8676		goto err_ioremap;
8677	}
8678
8679	netdev->netdev_ops = &ixgbe_netdev_ops;
8680	ixgbe_set_ethtool_ops(netdev);
8681	netdev->watchdog_timeo = 5 * HZ;
8682	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
8683
8684	/* Setup hw api */
8685	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
8686	hw->mac.type  = ii->mac;
8687	hw->mvals     = ii->mvals;
8688
8689	/* EEPROM */
8690	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
8691	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
8692	if (ixgbe_removed(hw->hw_addr)) {
8693		err = -EIO;
8694		goto err_ioremap;
8695	}
8696	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
8697	if (!(eec & (1 << 8)))
8698		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
8699
8700	/* PHY */
8701	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
8702	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
8703	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
8704	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
8705	hw->phy.mdio.mmds = 0;
8706	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
8707	hw->phy.mdio.dev = netdev;
8708	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
8709	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
8710
8711	ii->get_invariants(hw);
8712
8713	/* setup the private structure */
8714	err = ixgbe_sw_init(adapter);
8715	if (err)
8716		goto err_sw_init;
8717
8718	/* Make it possible the adapter to be woken up via WOL */
8719	switch (adapter->hw.mac.type) {
8720	case ixgbe_mac_82599EB:
8721	case ixgbe_mac_X540:
8722	case ixgbe_mac_X550:
8723	case ixgbe_mac_X550EM_x:
8724		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
8725		break;
8726	default:
8727		break;
8728	}
8729
8730	/*
8731	 * If there is a fan on this device and it has failed log the
8732	 * failure.
8733	 */
8734	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
8735		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
8736		if (esdp & IXGBE_ESDP_SDP1)
8737			e_crit(probe, "Fan has stopped, replace the adapter\n");
8738	}
8739
8740	if (allow_unsupported_sfp)
8741		hw->allow_unsupported_sfp = allow_unsupported_sfp;
8742
8743	/* reset_hw fills in the perm_addr as well */
8744	hw->phy.reset_if_overtemp = true;
8745	err = hw->mac.ops.reset_hw(hw);
8746	hw->phy.reset_if_overtemp = false;
8747	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
8748		err = 0;
8749	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
8750		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
8751		e_dev_err("Reload the driver after installing a supported module.\n");
8752		goto err_sw_init;
8753	} else if (err) {
8754		e_dev_err("HW Init failed: %d\n", err);
8755		goto err_sw_init;
8756	}
8757
8758#ifdef CONFIG_PCI_IOV
8759	/* SR-IOV not supported on the 82598 */
8760	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8761		goto skip_sriov;
8762	/* Mailbox */
8763	ixgbe_init_mbx_params_pf(hw);
8764	memcpy(&hw->mbx.ops, ii->mbx_ops, sizeof(hw->mbx.ops));
8765	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
8766	ixgbe_enable_sriov(adapter);
8767skip_sriov:
8768
8769#endif
8770	netdev->features = NETIF_F_SG |
8771			   NETIF_F_IP_CSUM |
8772			   NETIF_F_IPV6_CSUM |
8773			   NETIF_F_HW_VLAN_CTAG_TX |
8774			   NETIF_F_HW_VLAN_CTAG_RX |
8775			   NETIF_F_TSO |
8776			   NETIF_F_TSO6 |
8777			   NETIF_F_RXHASH |
8778			   NETIF_F_RXCSUM;
8779
8780	netdev->hw_features = netdev->features | NETIF_F_HW_L2FW_DOFFLOAD;
8781
8782	switch (adapter->hw.mac.type) {
8783	case ixgbe_mac_82599EB:
8784	case ixgbe_mac_X540:
8785	case ixgbe_mac_X550:
8786	case ixgbe_mac_X550EM_x:
8787		netdev->features |= NETIF_F_SCTP_CSUM;
8788		netdev->hw_features |= NETIF_F_SCTP_CSUM |
8789				       NETIF_F_NTUPLE;
8790		break;
8791	default:
8792		break;
8793	}
8794
8795	netdev->hw_features |= NETIF_F_RXALL;
8796	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
8797
8798	netdev->vlan_features |= NETIF_F_TSO;
8799	netdev->vlan_features |= NETIF_F_TSO6;
8800	netdev->vlan_features |= NETIF_F_IP_CSUM;
8801	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
8802	netdev->vlan_features |= NETIF_F_SG;
8803
8804	netdev->hw_enc_features |= NETIF_F_SG | NETIF_F_IP_CSUM |
8805				   NETIF_F_IPV6_CSUM;
8806
8807	netdev->priv_flags |= IFF_UNICAST_FLT;
8808	netdev->priv_flags |= IFF_SUPP_NOFCS;
8809
8810#ifdef CONFIG_IXGBE_VXLAN
8811	switch (adapter->hw.mac.type) {
8812	case ixgbe_mac_X550:
8813	case ixgbe_mac_X550EM_x:
8814		netdev->hw_enc_features |= NETIF_F_RXCSUM |
8815					   NETIF_F_IP_CSUM |
8816					   NETIF_F_IPV6_CSUM;
8817		break;
8818	default:
8819		break;
8820	}
8821#endif /* CONFIG_IXGBE_VXLAN */
8822
8823#ifdef CONFIG_IXGBE_DCB
8824	netdev->dcbnl_ops = &dcbnl_ops;
8825#endif
8826
8827#ifdef IXGBE_FCOE
8828	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
8829		unsigned int fcoe_l;
8830
8831		if (hw->mac.ops.get_device_caps) {
8832			hw->mac.ops.get_device_caps(hw, &device_caps);
8833			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
8834				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
8835		}
8836
8837
8838		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
8839		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
8840
8841		netdev->features |= NETIF_F_FSO |
8842				    NETIF_F_FCOE_CRC;
8843
8844		netdev->vlan_features |= NETIF_F_FSO |
8845					 NETIF_F_FCOE_CRC |
8846					 NETIF_F_FCOE_MTU;
8847	}
8848#endif /* IXGBE_FCOE */
8849	if (pci_using_dac) {
8850		netdev->features |= NETIF_F_HIGHDMA;
8851		netdev->vlan_features |= NETIF_F_HIGHDMA;
8852	}
8853
8854	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
8855		netdev->hw_features |= NETIF_F_LRO;
8856	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8857		netdev->features |= NETIF_F_LRO;
8858
8859	/* make sure the EEPROM is good */
8860	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
8861		e_dev_err("The EEPROM Checksum Is Not Valid\n");
8862		err = -EIO;
8863		goto err_sw_init;
8864	}
8865
8866	ixgbe_get_platform_mac_addr(adapter);
8867
8868	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
8869
8870	if (!is_valid_ether_addr(netdev->dev_addr)) {
8871		e_dev_err("invalid MAC address\n");
8872		err = -EIO;
8873		goto err_sw_init;
8874	}
8875
8876	ixgbe_mac_set_default_filter(adapter, hw->mac.perm_addr);
8877
8878	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
8879		    (unsigned long) adapter);
8880
8881	if (ixgbe_removed(hw->hw_addr)) {
8882		err = -EIO;
8883		goto err_sw_init;
8884	}
8885	INIT_WORK(&adapter->service_task, ixgbe_service_task);
8886	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
8887	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
8888
8889	err = ixgbe_init_interrupt_scheme(adapter);
8890	if (err)
8891		goto err_sw_init;
8892
8893	/* WOL not supported for all devices */
8894	adapter->wol = 0;
8895	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
8896	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
8897						pdev->subsystem_device);
8898	if (hw->wol_enabled)
8899		adapter->wol = IXGBE_WUFC_MAG;
8900
8901	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
8902
8903	/* save off EEPROM version number */
8904	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
8905	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
8906
8907	/* pick up the PCI bus settings for reporting later */
8908	if (ixgbe_pcie_from_parent(hw))
8909		ixgbe_get_parent_bus_info(adapter);
8910	else
8911		 hw->mac.ops.get_bus_info(hw);
8912
8913	/* calculate the expected PCIe bandwidth required for optimal
8914	 * performance. Note that some older parts will never have enough
8915	 * bandwidth due to being older generation PCIe parts. We clamp these
8916	 * parts to ensure no warning is displayed if it can't be fixed.
8917	 */
8918	switch (hw->mac.type) {
8919	case ixgbe_mac_82598EB:
8920		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
8921		break;
8922	default:
8923		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
8924		break;
8925	}
8926
8927	/* don't check link if we failed to enumerate functions */
8928	if (expected_gts > 0)
8929		ixgbe_check_minimum_link(adapter, expected_gts);
8930
8931	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
8932	if (err)
8933		strlcpy(part_str, "Unknown", sizeof(part_str));
8934	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
8935		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
8936			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
8937			   part_str);
8938	else
8939		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
8940			   hw->mac.type, hw->phy.type, part_str);
8941
8942	e_dev_info("%pM\n", netdev->dev_addr);
8943
8944	/* reset the hardware with the new settings */
8945	err = hw->mac.ops.start_hw(hw);
8946	if (err == IXGBE_ERR_EEPROM_VERSION) {
8947		/* We are running on a pre-production device, log a warning */
8948		e_dev_warn("This device is a pre-production adapter/LOM. "
8949			   "Please be aware there may be issues associated "
8950			   "with your hardware.  If you are experiencing "
8951			   "problems please contact your Intel or hardware "
8952			   "representative who provided you with this "
8953			   "hardware.\n");
8954	}
8955	strcpy(netdev->name, "eth%d");
8956	err = register_netdev(netdev);
8957	if (err)
8958		goto err_register;
8959
8960	pci_set_drvdata(pdev, adapter);
8961
8962	/* power down the optics for 82599 SFP+ fiber */
8963	if (hw->mac.ops.disable_tx_laser)
8964		hw->mac.ops.disable_tx_laser(hw);
8965
8966	/* carrier off reporting is important to ethtool even BEFORE open */
8967	netif_carrier_off(netdev);
8968
8969#ifdef CONFIG_IXGBE_DCA
8970	if (dca_add_requester(&pdev->dev) == 0) {
8971		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
8972		ixgbe_setup_dca(adapter);
8973	}
8974#endif
8975	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
8976		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
8977		for (i = 0; i < adapter->num_vfs; i++)
8978			ixgbe_vf_configuration(pdev, (i | 0x10000000));
8979	}
8980
8981	/* firmware requires driver version to be 0xFFFFFFFF
8982	 * since os does not support feature
8983	 */
8984	if (hw->mac.ops.set_fw_drv_ver)
8985		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
8986					   0xFF);
8987
8988	/* add san mac addr to netdev */
8989	ixgbe_add_sanmac_netdev(netdev);
8990
8991	e_dev_info("%s\n", ixgbe_default_device_descr);
8992
8993#ifdef CONFIG_IXGBE_HWMON
8994	if (ixgbe_sysfs_init(adapter))
8995		e_err(probe, "failed to allocate sysfs resources\n");
8996#endif /* CONFIG_IXGBE_HWMON */
8997
8998	ixgbe_dbg_adapter_init(adapter);
8999
9000	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
9001	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
9002		hw->mac.ops.setup_link(hw,
9003			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
9004			true);
9005
9006	return 0;
9007
9008err_register:
9009	ixgbe_release_hw_control(adapter);
9010	ixgbe_clear_interrupt_scheme(adapter);
9011err_sw_init:
9012	ixgbe_disable_sriov(adapter);
9013	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
9014	iounmap(adapter->io_addr);
9015	kfree(adapter->mac_table);
9016err_ioremap:
9017	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9018	free_netdev(netdev);
9019err_alloc_etherdev:
9020	pci_release_selected_regions(pdev,
9021				     pci_select_bars(pdev, IORESOURCE_MEM));
9022err_pci_reg:
9023err_dma:
9024	if (!adapter || disable_dev)
9025		pci_disable_device(pdev);
9026	return err;
9027}
9028
9029/**
9030 * ixgbe_remove - Device Removal Routine
9031 * @pdev: PCI device information struct
9032 *
9033 * ixgbe_remove is called by the PCI subsystem to alert the driver
9034 * that it should release a PCI device.  The could be caused by a
9035 * Hot-Plug event, or because the driver is going to be removed from
9036 * memory.
9037 **/
9038static void ixgbe_remove(struct pci_dev *pdev)
9039{
9040	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9041	struct net_device *netdev;
9042	bool disable_dev;
9043
9044	/* if !adapter then we already cleaned up in probe */
9045	if (!adapter)
9046		return;
9047
9048	netdev  = adapter->netdev;
9049	ixgbe_dbg_adapter_exit(adapter);
9050
9051	set_bit(__IXGBE_REMOVING, &adapter->state);
9052	cancel_work_sync(&adapter->service_task);
9053
9054
9055#ifdef CONFIG_IXGBE_DCA
9056	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
9057		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
9058		dca_remove_requester(&pdev->dev);
9059		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
9060				IXGBE_DCA_CTRL_DCA_DISABLE);
9061	}
9062
9063#endif
9064#ifdef CONFIG_IXGBE_HWMON
9065	ixgbe_sysfs_exit(adapter);
9066#endif /* CONFIG_IXGBE_HWMON */
9067
9068	/* remove the added san mac */
9069	ixgbe_del_sanmac_netdev(netdev);
9070
9071#ifdef CONFIG_PCI_IOV
9072	ixgbe_disable_sriov(adapter);
9073#endif
9074	if (netdev->reg_state == NETREG_REGISTERED)
9075		unregister_netdev(netdev);
9076
9077	ixgbe_clear_interrupt_scheme(adapter);
9078
9079	ixgbe_release_hw_control(adapter);
9080
9081#ifdef CONFIG_DCB
9082	kfree(adapter->ixgbe_ieee_pfc);
9083	kfree(adapter->ixgbe_ieee_ets);
9084
9085#endif
9086	iounmap(adapter->io_addr);
9087	pci_release_selected_regions(pdev, pci_select_bars(pdev,
9088				     IORESOURCE_MEM));
9089
9090	e_dev_info("complete\n");
9091
9092	kfree(adapter->mac_table);
9093	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9094	free_netdev(netdev);
9095
9096	pci_disable_pcie_error_reporting(pdev);
9097
9098	if (disable_dev)
9099		pci_disable_device(pdev);
9100}
9101
9102/**
9103 * ixgbe_io_error_detected - called when PCI error is detected
9104 * @pdev: Pointer to PCI device
9105 * @state: The current pci connection state
9106 *
9107 * This function is called after a PCI bus error affecting
9108 * this device has been detected.
9109 */
9110static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
9111						pci_channel_state_t state)
9112{
9113	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9114	struct net_device *netdev = adapter->netdev;
9115
9116#ifdef CONFIG_PCI_IOV
9117	struct ixgbe_hw *hw = &adapter->hw;
9118	struct pci_dev *bdev, *vfdev;
9119	u32 dw0, dw1, dw2, dw3;
9120	int vf, pos;
9121	u16 req_id, pf_func;
9122
9123	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
9124	    adapter->num_vfs == 0)
9125		goto skip_bad_vf_detection;
9126
9127	bdev = pdev->bus->self;
9128	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
9129		bdev = bdev->bus->self;
9130
9131	if (!bdev)
9132		goto skip_bad_vf_detection;
9133
9134	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
9135	if (!pos)
9136		goto skip_bad_vf_detection;
9137
9138	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
9139	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
9140	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
9141	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
9142	if (ixgbe_removed(hw->hw_addr))
9143		goto skip_bad_vf_detection;
9144
9145	req_id = dw1 >> 16;
9146	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
9147	if (!(req_id & 0x0080))
9148		goto skip_bad_vf_detection;
9149
9150	pf_func = req_id & 0x01;
9151	if ((pf_func & 1) == (pdev->devfn & 1)) {
9152		unsigned int device_id;
9153
9154		vf = (req_id & 0x7F) >> 1;
9155		e_dev_err("VF %d has caused a PCIe error\n", vf);
9156		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
9157				"%8.8x\tdw3: %8.8x\n",
9158		dw0, dw1, dw2, dw3);
9159		switch (adapter->hw.mac.type) {
9160		case ixgbe_mac_82599EB:
9161			device_id = IXGBE_82599_VF_DEVICE_ID;
9162			break;
9163		case ixgbe_mac_X540:
9164			device_id = IXGBE_X540_VF_DEVICE_ID;
9165			break;
9166		case ixgbe_mac_X550:
9167			device_id = IXGBE_DEV_ID_X550_VF;
9168			break;
9169		case ixgbe_mac_X550EM_x:
9170			device_id = IXGBE_DEV_ID_X550EM_X_VF;
9171			break;
9172		default:
9173			device_id = 0;
9174			break;
9175		}
9176
9177		/* Find the pci device of the offending VF */
9178		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
9179		while (vfdev) {
9180			if (vfdev->devfn == (req_id & 0xFF))
9181				break;
9182			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
9183					       device_id, vfdev);
9184		}
9185		/*
9186		 * There's a slim chance the VF could have been hot plugged,
9187		 * so if it is no longer present we don't need to issue the
9188		 * VFLR.  Just clean up the AER in that case.
9189		 */
9190		if (vfdev) {
9191			ixgbe_issue_vf_flr(adapter, vfdev);
9192			/* Free device reference count */
9193			pci_dev_put(vfdev);
9194		}
9195
9196		pci_cleanup_aer_uncorrect_error_status(pdev);
9197	}
9198
9199	/*
9200	 * Even though the error may have occurred on the other port
9201	 * we still need to increment the vf error reference count for
9202	 * both ports because the I/O resume function will be called
9203	 * for both of them.
9204	 */
9205	adapter->vferr_refcount++;
9206
9207	return PCI_ERS_RESULT_RECOVERED;
9208
9209skip_bad_vf_detection:
9210#endif /* CONFIG_PCI_IOV */
9211	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
9212		return PCI_ERS_RESULT_DISCONNECT;
9213
9214	rtnl_lock();
9215	netif_device_detach(netdev);
9216
9217	if (state == pci_channel_io_perm_failure) {
9218		rtnl_unlock();
9219		return PCI_ERS_RESULT_DISCONNECT;
9220	}
9221
9222	if (netif_running(netdev))
9223		ixgbe_down(adapter);
9224
9225	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
9226		pci_disable_device(pdev);
9227	rtnl_unlock();
9228
9229	/* Request a slot reset. */
9230	return PCI_ERS_RESULT_NEED_RESET;
9231}
9232
9233/**
9234 * ixgbe_io_slot_reset - called after the pci bus has been reset.
9235 * @pdev: Pointer to PCI device
9236 *
9237 * Restart the card from scratch, as if from a cold-boot.
9238 */
9239static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
9240{
9241	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9242	pci_ers_result_t result;
9243	int err;
9244
9245	if (pci_enable_device_mem(pdev)) {
9246		e_err(probe, "Cannot re-enable PCI device after reset.\n");
9247		result = PCI_ERS_RESULT_DISCONNECT;
9248	} else {
9249		smp_mb__before_atomic();
9250		clear_bit(__IXGBE_DISABLED, &adapter->state);
9251		adapter->hw.hw_addr = adapter->io_addr;
9252		pci_set_master(pdev);
9253		pci_restore_state(pdev);
9254		pci_save_state(pdev);
9255
9256		pci_wake_from_d3(pdev, false);
9257
9258		ixgbe_reset(adapter);
9259		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
9260		result = PCI_ERS_RESULT_RECOVERED;
9261	}
9262
9263	err = pci_cleanup_aer_uncorrect_error_status(pdev);
9264	if (err) {
9265		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
9266			  "failed 0x%0x\n", err);
9267		/* non-fatal, continue */
9268	}
9269
9270	return result;
9271}
9272
9273/**
9274 * ixgbe_io_resume - called when traffic can start flowing again.
9275 * @pdev: Pointer to PCI device
9276 *
9277 * This callback is called when the error recovery driver tells us that
9278 * its OK to resume normal operation.
9279 */
9280static void ixgbe_io_resume(struct pci_dev *pdev)
9281{
9282	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9283	struct net_device *netdev = adapter->netdev;
9284
9285#ifdef CONFIG_PCI_IOV
9286	if (adapter->vferr_refcount) {
9287		e_info(drv, "Resuming after VF err\n");
9288		adapter->vferr_refcount--;
9289		return;
9290	}
9291
9292#endif
9293	if (netif_running(netdev))
9294		ixgbe_up(adapter);
9295
9296	netif_device_attach(netdev);
9297}
9298
9299static const struct pci_error_handlers ixgbe_err_handler = {
9300	.error_detected = ixgbe_io_error_detected,
9301	.slot_reset = ixgbe_io_slot_reset,
9302	.resume = ixgbe_io_resume,
9303};
9304
9305static struct pci_driver ixgbe_driver = {
9306	.name     = ixgbe_driver_name,
9307	.id_table = ixgbe_pci_tbl,
9308	.probe    = ixgbe_probe,
9309	.remove   = ixgbe_remove,
9310#ifdef CONFIG_PM
9311	.suspend  = ixgbe_suspend,
9312	.resume   = ixgbe_resume,
9313#endif
9314	.shutdown = ixgbe_shutdown,
9315	.sriov_configure = ixgbe_pci_sriov_configure,
9316	.err_handler = &ixgbe_err_handler
9317};
9318
9319/**
9320 * ixgbe_init_module - Driver Registration Routine
9321 *
9322 * ixgbe_init_module is the first routine called when the driver is
9323 * loaded. All it does is register with the PCI subsystem.
9324 **/
9325static int __init ixgbe_init_module(void)
9326{
9327	int ret;
9328	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
9329	pr_info("%s\n", ixgbe_copyright);
9330
9331	ixgbe_dbg_init();
9332
9333	ret = pci_register_driver(&ixgbe_driver);
9334	if (ret) {
9335		ixgbe_dbg_exit();
9336		return ret;
9337	}
9338
9339#ifdef CONFIG_IXGBE_DCA
9340	dca_register_notify(&dca_notifier);
9341#endif
9342
9343	return 0;
9344}
9345
9346module_init(ixgbe_init_module);
9347
9348/**
9349 * ixgbe_exit_module - Driver Exit Cleanup Routine
9350 *
9351 * ixgbe_exit_module is called just before the driver is removed
9352 * from memory.
9353 **/
9354static void __exit ixgbe_exit_module(void)
9355{
9356#ifdef CONFIG_IXGBE_DCA
9357	dca_unregister_notify(&dca_notifier);
9358#endif
9359	pci_unregister_driver(&ixgbe_driver);
9360
9361	ixgbe_dbg_exit();
9362}
9363
9364#ifdef CONFIG_IXGBE_DCA
9365static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
9366			    void *p)
9367{
9368	int ret_val;
9369
9370	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
9371					 __ixgbe_notify_dca);
9372
9373	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
9374}
9375
9376#endif /* CONFIG_IXGBE_DCA */
9377
9378module_exit(ixgbe_exit_module);
9379
9380/* ixgbe_main.c */
9381