1/*******************************************************************************
2
3  Intel 10 Gigabit PCI Express Linux driver
4  Copyright(c) 1999 - 2014 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
69char ixgbe_driver_name[] = "ixgbe";
70static const char ixgbe_driver_string[] =
71			      "Intel(R) 10 Gigabit PCI Express Network Driver";
72#ifdef IXGBE_FCOE
73char ixgbe_default_device_descr[] =
74			      "Intel(R) 10 Gigabit Network Connection";
75#else
76static char ixgbe_default_device_descr[] =
77			      "Intel(R) 10 Gigabit Network Connection";
78#endif
79#define DRV_VERSION "4.0.1-k"
80const char ixgbe_driver_version[] = DRV_VERSION;
81static const char ixgbe_copyright[] =
82				"Copyright (c) 1999-2014 Intel Corporation.";
83
84static const struct ixgbe_info *ixgbe_info_tbl[] = {
85	[board_82598]		= &ixgbe_82598_info,
86	[board_82599]		= &ixgbe_82599_info,
87	[board_X540]		= &ixgbe_X540_info,
88	[board_X550]		= &ixgbe_X550_info,
89	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
90};
91
92/* ixgbe_pci_tbl - PCI Device ID Table
93 *
94 * Wildcard entries (PCI_ANY_ID) should come last
95 * Last entry must be all 0s
96 *
97 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
98 *   Class, Class Mask, private data (not used) }
99 */
100static const struct pci_device_id ixgbe_pci_tbl[] = {
101	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
102	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
103	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
104	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
105	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
106	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
107	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
108	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
109	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
110	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
111	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
112	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
113	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
114	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
115	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
116	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
117	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
118	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
119	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
120	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
121	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
122	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
123	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
124	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
125	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
126	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
127	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
128	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
129	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
130	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
131	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
132	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
133	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
134	/* required last entry */
135	{0, }
136};
137MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
138
139#ifdef CONFIG_IXGBE_DCA
140static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
141			    void *p);
142static struct notifier_block dca_notifier = {
143	.notifier_call = ixgbe_notify_dca,
144	.next          = NULL,
145	.priority      = 0
146};
147#endif
148
149#ifdef CONFIG_PCI_IOV
150static unsigned int max_vfs;
151module_param(max_vfs, uint, 0);
152MODULE_PARM_DESC(max_vfs,
153		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
154#endif /* CONFIG_PCI_IOV */
155
156static unsigned int allow_unsupported_sfp;
157module_param(allow_unsupported_sfp, uint, 0);
158MODULE_PARM_DESC(allow_unsupported_sfp,
159		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
160
161#define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
162static int debug = -1;
163module_param(debug, int, 0);
164MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
165
166MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
167MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
168MODULE_LICENSE("GPL");
169MODULE_VERSION(DRV_VERSION);
170
171static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
172
173static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
174					  u32 reg, u16 *value)
175{
176	struct pci_dev *parent_dev;
177	struct pci_bus *parent_bus;
178
179	parent_bus = adapter->pdev->bus->parent;
180	if (!parent_bus)
181		return -1;
182
183	parent_dev = parent_bus->self;
184	if (!parent_dev)
185		return -1;
186
187	if (!pci_is_pcie(parent_dev))
188		return -1;
189
190	pcie_capability_read_word(parent_dev, reg, value);
191	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
192	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
193		return -1;
194	return 0;
195}
196
197static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
198{
199	struct ixgbe_hw *hw = &adapter->hw;
200	u16 link_status = 0;
201	int err;
202
203	hw->bus.type = ixgbe_bus_type_pci_express;
204
205	/* Get the negotiated link width and speed from PCI config space of the
206	 * parent, as this device is behind a switch
207	 */
208	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
209
210	/* assume caller will handle error case */
211	if (err)
212		return err;
213
214	hw->bus.width = ixgbe_convert_bus_width(link_status);
215	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
216
217	return 0;
218}
219
220/**
221 * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
222 * @hw: hw specific details
223 *
224 * This function is used by probe to determine whether a device's PCI-Express
225 * bandwidth details should be gathered from the parent bus instead of from the
226 * device. Used to ensure that various locations all have the correct device ID
227 * checks.
228 */
229static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
230{
231	switch (hw->device_id) {
232	case IXGBE_DEV_ID_82599_SFP_SF_QP:
233	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
234		return true;
235	default:
236		return false;
237	}
238}
239
240static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
241				     int expected_gts)
242{
243	int max_gts = 0;
244	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
245	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
246	struct pci_dev *pdev;
247
248	/* determine whether to use the the parent device
249	 */
250	if (ixgbe_pcie_from_parent(&adapter->hw))
251		pdev = adapter->pdev->bus->parent->self;
252	else
253		pdev = adapter->pdev;
254
255	if (pcie_get_minimum_link(pdev, &speed, &width) ||
256	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
257		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
258		return;
259	}
260
261	switch (speed) {
262	case PCIE_SPEED_2_5GT:
263		/* 8b/10b encoding reduces max throughput by 20% */
264		max_gts = 2 * width;
265		break;
266	case PCIE_SPEED_5_0GT:
267		/* 8b/10b encoding reduces max throughput by 20% */
268		max_gts = 4 * width;
269		break;
270	case PCIE_SPEED_8_0GT:
271		/* 128b/130b encoding reduces throughput by less than 2% */
272		max_gts = 8 * width;
273		break;
274	default:
275		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
276		return;
277	}
278
279	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
280		   max_gts);
281	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
282		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
283		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
284		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
285		    "Unknown"),
286		   width,
287		   (speed == PCIE_SPEED_2_5GT ? "20%" :
288		    speed == PCIE_SPEED_5_0GT ? "20%" :
289		    speed == PCIE_SPEED_8_0GT ? "<2%" :
290		    "Unknown"));
291
292	if (max_gts < expected_gts) {
293		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
294		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
295			expected_gts);
296		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
297	}
298}
299
300static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
301{
302	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
303	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
304	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
305		schedule_work(&adapter->service_task);
306}
307
308static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
309{
310	struct ixgbe_adapter *adapter = hw->back;
311
312	if (!hw->hw_addr)
313		return;
314	hw->hw_addr = NULL;
315	e_dev_err("Adapter removed\n");
316	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
317		ixgbe_service_event_schedule(adapter);
318}
319
320static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
321{
322	u32 value;
323
324	/* The following check not only optimizes a bit by not
325	 * performing a read on the status register when the
326	 * register just read was a status register read that
327	 * returned IXGBE_FAILED_READ_REG. It also blocks any
328	 * potential recursion.
329	 */
330	if (reg == IXGBE_STATUS) {
331		ixgbe_remove_adapter(hw);
332		return;
333	}
334	value = ixgbe_read_reg(hw, IXGBE_STATUS);
335	if (value == IXGBE_FAILED_READ_REG)
336		ixgbe_remove_adapter(hw);
337}
338
339/**
340 * ixgbe_read_reg - Read from device register
341 * @hw: hw specific details
342 * @reg: offset of register to read
343 *
344 * Returns : value read or IXGBE_FAILED_READ_REG if removed
345 *
346 * This function is used to read device registers. It checks for device
347 * removal by confirming any read that returns all ones by checking the
348 * status register value for all ones. This function avoids reading from
349 * the hardware if a removal was previously detected in which case it
350 * returns IXGBE_FAILED_READ_REG (all ones).
351 */
352u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
353{
354	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
355	u32 value;
356
357	if (ixgbe_removed(reg_addr))
358		return IXGBE_FAILED_READ_REG;
359	value = readl(reg_addr + reg);
360	if (unlikely(value == IXGBE_FAILED_READ_REG))
361		ixgbe_check_remove(hw, reg);
362	return value;
363}
364
365static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
366{
367	u16 value;
368
369	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
370	if (value == IXGBE_FAILED_READ_CFG_WORD) {
371		ixgbe_remove_adapter(hw);
372		return true;
373	}
374	return false;
375}
376
377u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
378{
379	struct ixgbe_adapter *adapter = hw->back;
380	u16 value;
381
382	if (ixgbe_removed(hw->hw_addr))
383		return IXGBE_FAILED_READ_CFG_WORD;
384	pci_read_config_word(adapter->pdev, reg, &value);
385	if (value == IXGBE_FAILED_READ_CFG_WORD &&
386	    ixgbe_check_cfg_remove(hw, adapter->pdev))
387		return IXGBE_FAILED_READ_CFG_WORD;
388	return value;
389}
390
391#ifdef CONFIG_PCI_IOV
392static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
393{
394	struct ixgbe_adapter *adapter = hw->back;
395	u32 value;
396
397	if (ixgbe_removed(hw->hw_addr))
398		return IXGBE_FAILED_READ_CFG_DWORD;
399	pci_read_config_dword(adapter->pdev, reg, &value);
400	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
401	    ixgbe_check_cfg_remove(hw, adapter->pdev))
402		return IXGBE_FAILED_READ_CFG_DWORD;
403	return value;
404}
405#endif /* CONFIG_PCI_IOV */
406
407void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
408{
409	struct ixgbe_adapter *adapter = hw->back;
410
411	if (ixgbe_removed(hw->hw_addr))
412		return;
413	pci_write_config_word(adapter->pdev, reg, value);
414}
415
416static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
417{
418	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
419
420	/* flush memory to make sure state is correct before next watchdog */
421	smp_mb__before_atomic();
422	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
423}
424
425struct ixgbe_reg_info {
426	u32 ofs;
427	char *name;
428};
429
430static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
431
432	/* General Registers */
433	{IXGBE_CTRL, "CTRL"},
434	{IXGBE_STATUS, "STATUS"},
435	{IXGBE_CTRL_EXT, "CTRL_EXT"},
436
437	/* Interrupt Registers */
438	{IXGBE_EICR, "EICR"},
439
440	/* RX Registers */
441	{IXGBE_SRRCTL(0), "SRRCTL"},
442	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
443	{IXGBE_RDLEN(0), "RDLEN"},
444	{IXGBE_RDH(0), "RDH"},
445	{IXGBE_RDT(0), "RDT"},
446	{IXGBE_RXDCTL(0), "RXDCTL"},
447	{IXGBE_RDBAL(0), "RDBAL"},
448	{IXGBE_RDBAH(0), "RDBAH"},
449
450	/* TX Registers */
451	{IXGBE_TDBAL(0), "TDBAL"},
452	{IXGBE_TDBAH(0), "TDBAH"},
453	{IXGBE_TDLEN(0), "TDLEN"},
454	{IXGBE_TDH(0), "TDH"},
455	{IXGBE_TDT(0), "TDT"},
456	{IXGBE_TXDCTL(0), "TXDCTL"},
457
458	/* List Terminator */
459	{ .name = NULL }
460};
461
462
463/*
464 * ixgbe_regdump - register printout routine
465 */
466static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
467{
468	int i = 0, j = 0;
469	char rname[16];
470	u32 regs[64];
471
472	switch (reginfo->ofs) {
473	case IXGBE_SRRCTL(0):
474		for (i = 0; i < 64; i++)
475			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
476		break;
477	case IXGBE_DCA_RXCTRL(0):
478		for (i = 0; i < 64; i++)
479			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
480		break;
481	case IXGBE_RDLEN(0):
482		for (i = 0; i < 64; i++)
483			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
484		break;
485	case IXGBE_RDH(0):
486		for (i = 0; i < 64; i++)
487			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
488		break;
489	case IXGBE_RDT(0):
490		for (i = 0; i < 64; i++)
491			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
492		break;
493	case IXGBE_RXDCTL(0):
494		for (i = 0; i < 64; i++)
495			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
496		break;
497	case IXGBE_RDBAL(0):
498		for (i = 0; i < 64; i++)
499			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
500		break;
501	case IXGBE_RDBAH(0):
502		for (i = 0; i < 64; i++)
503			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
504		break;
505	case IXGBE_TDBAL(0):
506		for (i = 0; i < 64; i++)
507			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
508		break;
509	case IXGBE_TDBAH(0):
510		for (i = 0; i < 64; i++)
511			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
512		break;
513	case IXGBE_TDLEN(0):
514		for (i = 0; i < 64; i++)
515			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
516		break;
517	case IXGBE_TDH(0):
518		for (i = 0; i < 64; i++)
519			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
520		break;
521	case IXGBE_TDT(0):
522		for (i = 0; i < 64; i++)
523			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
524		break;
525	case IXGBE_TXDCTL(0):
526		for (i = 0; i < 64; i++)
527			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
528		break;
529	default:
530		pr_info("%-15s %08x\n", reginfo->name,
531			IXGBE_READ_REG(hw, reginfo->ofs));
532		return;
533	}
534
535	for (i = 0; i < 8; i++) {
536		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
537		pr_err("%-15s", rname);
538		for (j = 0; j < 8; j++)
539			pr_cont(" %08x", regs[i*8+j]);
540		pr_cont("\n");
541	}
542
543}
544
545/*
546 * ixgbe_dump - Print registers, tx-rings and rx-rings
547 */
548static void ixgbe_dump(struct ixgbe_adapter *adapter)
549{
550	struct net_device *netdev = adapter->netdev;
551	struct ixgbe_hw *hw = &adapter->hw;
552	struct ixgbe_reg_info *reginfo;
553	int n = 0;
554	struct ixgbe_ring *tx_ring;
555	struct ixgbe_tx_buffer *tx_buffer;
556	union ixgbe_adv_tx_desc *tx_desc;
557	struct my_u0 { u64 a; u64 b; } *u0;
558	struct ixgbe_ring *rx_ring;
559	union ixgbe_adv_rx_desc *rx_desc;
560	struct ixgbe_rx_buffer *rx_buffer_info;
561	u32 staterr;
562	int i = 0;
563
564	if (!netif_msg_hw(adapter))
565		return;
566
567	/* Print netdevice Info */
568	if (netdev) {
569		dev_info(&adapter->pdev->dev, "Net device Info\n");
570		pr_info("Device Name     state            "
571			"trans_start      last_rx\n");
572		pr_info("%-15s %016lX %016lX %016lX\n",
573			netdev->name,
574			netdev->state,
575			netdev->trans_start,
576			netdev->last_rx);
577	}
578
579	/* Print Registers */
580	dev_info(&adapter->pdev->dev, "Register Dump\n");
581	pr_info(" Register Name   Value\n");
582	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
583	     reginfo->name; reginfo++) {
584		ixgbe_regdump(hw, reginfo);
585	}
586
587	/* Print TX Ring Summary */
588	if (!netdev || !netif_running(netdev))
589		return;
590
591	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
592	pr_info(" %s     %s              %s        %s\n",
593		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
594		"leng", "ntw", "timestamp");
595	for (n = 0; n < adapter->num_tx_queues; n++) {
596		tx_ring = adapter->tx_ring[n];
597		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
598		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
599			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
600			   (u64)dma_unmap_addr(tx_buffer, dma),
601			   dma_unmap_len(tx_buffer, len),
602			   tx_buffer->next_to_watch,
603			   (u64)tx_buffer->time_stamp);
604	}
605
606	/* Print TX Rings */
607	if (!netif_msg_tx_done(adapter))
608		goto rx_ring_summary;
609
610	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
611
612	/* Transmit Descriptor Formats
613	 *
614	 * 82598 Advanced Transmit Descriptor
615	 *   +--------------------------------------------------------------+
616	 * 0 |         Buffer Address [63:0]                                |
617	 *   +--------------------------------------------------------------+
618	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
619	 *   +--------------------------------------------------------------+
620	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
621	 *
622	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
623	 *   +--------------------------------------------------------------+
624	 * 0 |                          RSV [63:0]                          |
625	 *   +--------------------------------------------------------------+
626	 * 8 |            RSV           |  STA  |          NXTSEQ           |
627	 *   +--------------------------------------------------------------+
628	 *   63                       36 35   32 31                         0
629	 *
630	 * 82599+ Advanced Transmit Descriptor
631	 *   +--------------------------------------------------------------+
632	 * 0 |         Buffer Address [63:0]                                |
633	 *   +--------------------------------------------------------------+
634	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
635	 *   +--------------------------------------------------------------+
636	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
637	 *
638	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
639	 *   +--------------------------------------------------------------+
640	 * 0 |                          RSV [63:0]                          |
641	 *   +--------------------------------------------------------------+
642	 * 8 |            RSV           |  STA  |           RSV             |
643	 *   +--------------------------------------------------------------+
644	 *   63                       36 35   32 31                         0
645	 */
646
647	for (n = 0; n < adapter->num_tx_queues; n++) {
648		tx_ring = adapter->tx_ring[n];
649		pr_info("------------------------------------\n");
650		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
651		pr_info("------------------------------------\n");
652		pr_info("%s%s    %s              %s        %s          %s\n",
653			"T [desc]     [address 63:0  ] ",
654			"[PlPOIdStDDt Ln] [bi->dma       ] ",
655			"leng", "ntw", "timestamp", "bi->skb");
656
657		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
658			tx_desc = IXGBE_TX_DESC(tx_ring, i);
659			tx_buffer = &tx_ring->tx_buffer_info[i];
660			u0 = (struct my_u0 *)tx_desc;
661			if (dma_unmap_len(tx_buffer, len) > 0) {
662				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
663					i,
664					le64_to_cpu(u0->a),
665					le64_to_cpu(u0->b),
666					(u64)dma_unmap_addr(tx_buffer, dma),
667					dma_unmap_len(tx_buffer, len),
668					tx_buffer->next_to_watch,
669					(u64)tx_buffer->time_stamp,
670					tx_buffer->skb);
671				if (i == tx_ring->next_to_use &&
672					i == tx_ring->next_to_clean)
673					pr_cont(" NTC/U\n");
674				else if (i == tx_ring->next_to_use)
675					pr_cont(" NTU\n");
676				else if (i == tx_ring->next_to_clean)
677					pr_cont(" NTC\n");
678				else
679					pr_cont("\n");
680
681				if (netif_msg_pktdata(adapter) &&
682				    tx_buffer->skb)
683					print_hex_dump(KERN_INFO, "",
684						DUMP_PREFIX_ADDRESS, 16, 1,
685						tx_buffer->skb->data,
686						dma_unmap_len(tx_buffer, len),
687						true);
688			}
689		}
690	}
691
692	/* Print RX Rings Summary */
693rx_ring_summary:
694	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
695	pr_info("Queue [NTU] [NTC]\n");
696	for (n = 0; n < adapter->num_rx_queues; n++) {
697		rx_ring = adapter->rx_ring[n];
698		pr_info("%5d %5X %5X\n",
699			n, rx_ring->next_to_use, rx_ring->next_to_clean);
700	}
701
702	/* Print RX Rings */
703	if (!netif_msg_rx_status(adapter))
704		return;
705
706	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
707
708	/* Receive Descriptor Formats
709	 *
710	 * 82598 Advanced Receive Descriptor (Read) Format
711	 *    63                                           1        0
712	 *    +-----------------------------------------------------+
713	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
714	 *    +----------------------------------------------+------+
715	 *  8 |       Header Buffer Address [63:1]           |  DD  |
716	 *    +-----------------------------------------------------+
717	 *
718	 *
719	 * 82598 Advanced Receive Descriptor (Write-Back) Format
720	 *
721	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
722	 *   +------------------------------------------------------+
723	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
724	 *   | Packet   | IP     |   |          |     | Type | Type |
725	 *   | Checksum | Ident  |   |          |     |      |      |
726	 *   +------------------------------------------------------+
727	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
728	 *   +------------------------------------------------------+
729	 *   63       48 47    32 31            20 19               0
730	 *
731	 * 82599+ Advanced Receive Descriptor (Read) Format
732	 *    63                                           1        0
733	 *    +-----------------------------------------------------+
734	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
735	 *    +----------------------------------------------+------+
736	 *  8 |       Header Buffer Address [63:1]           |  DD  |
737	 *    +-----------------------------------------------------+
738	 *
739	 *
740	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
741	 *
742	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
743	 *   +------------------------------------------------------+
744	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
745	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
746	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
747	 *   +------------------------------------------------------+
748	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
749	 *   +------------------------------------------------------+
750	 *   63       48 47    32 31          20 19                 0
751	 */
752
753	for (n = 0; n < adapter->num_rx_queues; n++) {
754		rx_ring = adapter->rx_ring[n];
755		pr_info("------------------------------------\n");
756		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
757		pr_info("------------------------------------\n");
758		pr_info("%s%s%s",
759			"R  [desc]      [ PktBuf     A0] ",
760			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
761			"<-- Adv Rx Read format\n");
762		pr_info("%s%s%s",
763			"RWB[desc]      [PcsmIpSHl PtRs] ",
764			"[vl er S cks ln] ---------------- [bi->skb       ] ",
765			"<-- Adv Rx Write-Back format\n");
766
767		for (i = 0; i < rx_ring->count; i++) {
768			rx_buffer_info = &rx_ring->rx_buffer_info[i];
769			rx_desc = IXGBE_RX_DESC(rx_ring, i);
770			u0 = (struct my_u0 *)rx_desc;
771			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
772			if (staterr & IXGBE_RXD_STAT_DD) {
773				/* Descriptor Done */
774				pr_info("RWB[0x%03X]     %016llX "
775					"%016llX ---------------- %p", i,
776					le64_to_cpu(u0->a),
777					le64_to_cpu(u0->b),
778					rx_buffer_info->skb);
779			} else {
780				pr_info("R  [0x%03X]     %016llX "
781					"%016llX %016llX %p", i,
782					le64_to_cpu(u0->a),
783					le64_to_cpu(u0->b),
784					(u64)rx_buffer_info->dma,
785					rx_buffer_info->skb);
786
787				if (netif_msg_pktdata(adapter) &&
788				    rx_buffer_info->dma) {
789					print_hex_dump(KERN_INFO, "",
790					   DUMP_PREFIX_ADDRESS, 16, 1,
791					   page_address(rx_buffer_info->page) +
792						    rx_buffer_info->page_offset,
793					   ixgbe_rx_bufsz(rx_ring), true);
794				}
795			}
796
797			if (i == rx_ring->next_to_use)
798				pr_cont(" NTU\n");
799			else if (i == rx_ring->next_to_clean)
800				pr_cont(" NTC\n");
801			else
802				pr_cont("\n");
803
804		}
805	}
806}
807
808static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
809{
810	u32 ctrl_ext;
811
812	/* Let firmware take over control of h/w */
813	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
814	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
815			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
816}
817
818static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
819{
820	u32 ctrl_ext;
821
822	/* Let firmware know the driver has taken over */
823	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
824	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
825			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
826}
827
828/**
829 * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
830 * @adapter: pointer to adapter struct
831 * @direction: 0 for Rx, 1 for Tx, -1 for other causes
832 * @queue: queue to map the corresponding interrupt to
833 * @msix_vector: the vector to map to the corresponding queue
834 *
835 */
836static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
837			   u8 queue, u8 msix_vector)
838{
839	u32 ivar, index;
840	struct ixgbe_hw *hw = &adapter->hw;
841	switch (hw->mac.type) {
842	case ixgbe_mac_82598EB:
843		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
844		if (direction == -1)
845			direction = 0;
846		index = (((direction * 64) + queue) >> 2) & 0x1F;
847		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
848		ivar &= ~(0xFF << (8 * (queue & 0x3)));
849		ivar |= (msix_vector << (8 * (queue & 0x3)));
850		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
851		break;
852	case ixgbe_mac_82599EB:
853	case ixgbe_mac_X540:
854	case ixgbe_mac_X550:
855	case ixgbe_mac_X550EM_x:
856		if (direction == -1) {
857			/* other causes */
858			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
859			index = ((queue & 1) * 8);
860			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
861			ivar &= ~(0xFF << index);
862			ivar |= (msix_vector << index);
863			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
864			break;
865		} else {
866			/* tx or rx causes */
867			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
868			index = ((16 * (queue & 1)) + (8 * direction));
869			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
870			ivar &= ~(0xFF << index);
871			ivar |= (msix_vector << index);
872			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
873			break;
874		}
875	default:
876		break;
877	}
878}
879
880static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
881					  u64 qmask)
882{
883	u32 mask;
884
885	switch (adapter->hw.mac.type) {
886	case ixgbe_mac_82598EB:
887		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
888		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
889		break;
890	case ixgbe_mac_82599EB:
891	case ixgbe_mac_X540:
892	case ixgbe_mac_X550:
893	case ixgbe_mac_X550EM_x:
894		mask = (qmask & 0xFFFFFFFF);
895		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
896		mask = (qmask >> 32);
897		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
898		break;
899	default:
900		break;
901	}
902}
903
904void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
905				      struct ixgbe_tx_buffer *tx_buffer)
906{
907	if (tx_buffer->skb) {
908		dev_kfree_skb_any(tx_buffer->skb);
909		if (dma_unmap_len(tx_buffer, len))
910			dma_unmap_single(ring->dev,
911					 dma_unmap_addr(tx_buffer, dma),
912					 dma_unmap_len(tx_buffer, len),
913					 DMA_TO_DEVICE);
914	} else if (dma_unmap_len(tx_buffer, len)) {
915		dma_unmap_page(ring->dev,
916			       dma_unmap_addr(tx_buffer, dma),
917			       dma_unmap_len(tx_buffer, len),
918			       DMA_TO_DEVICE);
919	}
920	tx_buffer->next_to_watch = NULL;
921	tx_buffer->skb = NULL;
922	dma_unmap_len_set(tx_buffer, len, 0);
923	/* tx_buffer must be completely set up in the transmit path */
924}
925
926static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
927{
928	struct ixgbe_hw *hw = &adapter->hw;
929	struct ixgbe_hw_stats *hwstats = &adapter->stats;
930	int i;
931	u32 data;
932
933	if ((hw->fc.current_mode != ixgbe_fc_full) &&
934	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
935		return;
936
937	switch (hw->mac.type) {
938	case ixgbe_mac_82598EB:
939		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
940		break;
941	default:
942		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
943	}
944	hwstats->lxoffrxc += data;
945
946	/* refill credits (no tx hang) if we received xoff */
947	if (!data)
948		return;
949
950	for (i = 0; i < adapter->num_tx_queues; i++)
951		clear_bit(__IXGBE_HANG_CHECK_ARMED,
952			  &adapter->tx_ring[i]->state);
953}
954
955static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
956{
957	struct ixgbe_hw *hw = &adapter->hw;
958	struct ixgbe_hw_stats *hwstats = &adapter->stats;
959	u32 xoff[8] = {0};
960	u8 tc;
961	int i;
962	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
963
964	if (adapter->ixgbe_ieee_pfc)
965		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
966
967	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
968		ixgbe_update_xoff_rx_lfc(adapter);
969		return;
970	}
971
972	/* update stats for each tc, only valid with PFC enabled */
973	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
974		u32 pxoffrxc;
975
976		switch (hw->mac.type) {
977		case ixgbe_mac_82598EB:
978			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
979			break;
980		default:
981			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
982		}
983		hwstats->pxoffrxc[i] += pxoffrxc;
984		/* Get the TC for given UP */
985		tc = netdev_get_prio_tc_map(adapter->netdev, i);
986		xoff[tc] += pxoffrxc;
987	}
988
989	/* disarm tx queues that have received xoff frames */
990	for (i = 0; i < adapter->num_tx_queues; i++) {
991		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
992
993		tc = tx_ring->dcb_tc;
994		if (xoff[tc])
995			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
996	}
997}
998
999static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1000{
1001	return ring->stats.packets;
1002}
1003
1004static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1005{
1006	struct ixgbe_adapter *adapter;
1007	struct ixgbe_hw *hw;
1008	u32 head, tail;
1009
1010	if (ring->l2_accel_priv)
1011		adapter = ring->l2_accel_priv->real_adapter;
1012	else
1013		adapter = netdev_priv(ring->netdev);
1014
1015	hw = &adapter->hw;
1016	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1017	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1018
1019	if (head != tail)
1020		return (head < tail) ?
1021			tail - head : (tail + ring->count - head);
1022
1023	return 0;
1024}
1025
1026static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1027{
1028	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1029	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1030	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1031
1032	clear_check_for_tx_hang(tx_ring);
1033
1034	/*
1035	 * Check for a hung queue, but be thorough. This verifies
1036	 * that a transmit has been completed since the previous
1037	 * check AND there is at least one packet pending. The
1038	 * ARMED bit is set to indicate a potential hang. The
1039	 * bit is cleared if a pause frame is received to remove
1040	 * false hang detection due to PFC or 802.3x frames. By
1041	 * requiring this to fail twice we avoid races with
1042	 * pfc clearing the ARMED bit and conditions where we
1043	 * run the check_tx_hang logic with a transmit completion
1044	 * pending but without time to complete it yet.
1045	 */
1046	if (tx_done_old == tx_done && tx_pending)
1047		/* make sure it is true for two checks in a row */
1048		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1049					&tx_ring->state);
1050	/* update completed stats and continue */
1051	tx_ring->tx_stats.tx_done_old = tx_done;
1052	/* reset the countdown */
1053	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1054
1055	return false;
1056}
1057
1058/**
1059 * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1060 * @adapter: driver private struct
1061 **/
1062static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1063{
1064
1065	/* Do the reset outside of interrupt context */
1066	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1067		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
1068		e_warn(drv, "initiating reset due to tx timeout\n");
1069		ixgbe_service_event_schedule(adapter);
1070	}
1071}
1072
1073/**
1074 * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1075 * @q_vector: structure containing interrupt and ring information
1076 * @tx_ring: tx ring to clean
1077 **/
1078static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1079			       struct ixgbe_ring *tx_ring)
1080{
1081	struct ixgbe_adapter *adapter = q_vector->adapter;
1082	struct ixgbe_tx_buffer *tx_buffer;
1083	union ixgbe_adv_tx_desc *tx_desc;
1084	unsigned int total_bytes = 0, total_packets = 0;
1085	unsigned int budget = q_vector->tx.work_limit;
1086	unsigned int i = tx_ring->next_to_clean;
1087
1088	if (test_bit(__IXGBE_DOWN, &adapter->state))
1089		return true;
1090
1091	tx_buffer = &tx_ring->tx_buffer_info[i];
1092	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1093	i -= tx_ring->count;
1094
1095	do {
1096		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1097
1098		/* if next_to_watch is not set then there is no work pending */
1099		if (!eop_desc)
1100			break;
1101
1102		/* prevent any other reads prior to eop_desc */
1103		read_barrier_depends();
1104
1105		/* if DD is not set pending work has not been completed */
1106		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1107			break;
1108
1109		/* clear next_to_watch to prevent false hangs */
1110		tx_buffer->next_to_watch = NULL;
1111
1112		/* update the statistics for this packet */
1113		total_bytes += tx_buffer->bytecount;
1114		total_packets += tx_buffer->gso_segs;
1115
1116		/* free the skb */
1117		dev_consume_skb_any(tx_buffer->skb);
1118
1119		/* unmap skb header data */
1120		dma_unmap_single(tx_ring->dev,
1121				 dma_unmap_addr(tx_buffer, dma),
1122				 dma_unmap_len(tx_buffer, len),
1123				 DMA_TO_DEVICE);
1124
1125		/* clear tx_buffer data */
1126		tx_buffer->skb = NULL;
1127		dma_unmap_len_set(tx_buffer, len, 0);
1128
1129		/* unmap remaining buffers */
1130		while (tx_desc != eop_desc) {
1131			tx_buffer++;
1132			tx_desc++;
1133			i++;
1134			if (unlikely(!i)) {
1135				i -= tx_ring->count;
1136				tx_buffer = tx_ring->tx_buffer_info;
1137				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1138			}
1139
1140			/* unmap any remaining paged data */
1141			if (dma_unmap_len(tx_buffer, len)) {
1142				dma_unmap_page(tx_ring->dev,
1143					       dma_unmap_addr(tx_buffer, dma),
1144					       dma_unmap_len(tx_buffer, len),
1145					       DMA_TO_DEVICE);
1146				dma_unmap_len_set(tx_buffer, len, 0);
1147			}
1148		}
1149
1150		/* move us one more past the eop_desc for start of next pkt */
1151		tx_buffer++;
1152		tx_desc++;
1153		i++;
1154		if (unlikely(!i)) {
1155			i -= tx_ring->count;
1156			tx_buffer = tx_ring->tx_buffer_info;
1157			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1158		}
1159
1160		/* issue prefetch for next Tx descriptor */
1161		prefetch(tx_desc);
1162
1163		/* update budget accounting */
1164		budget--;
1165	} while (likely(budget));
1166
1167	i += tx_ring->count;
1168	tx_ring->next_to_clean = i;
1169	u64_stats_update_begin(&tx_ring->syncp);
1170	tx_ring->stats.bytes += total_bytes;
1171	tx_ring->stats.packets += total_packets;
1172	u64_stats_update_end(&tx_ring->syncp);
1173	q_vector->tx.total_bytes += total_bytes;
1174	q_vector->tx.total_packets += total_packets;
1175
1176	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1177		/* schedule immediate reset if we believe we hung */
1178		struct ixgbe_hw *hw = &adapter->hw;
1179		e_err(drv, "Detected Tx Unit Hang\n"
1180			"  Tx Queue             <%d>\n"
1181			"  TDH, TDT             <%x>, <%x>\n"
1182			"  next_to_use          <%x>\n"
1183			"  next_to_clean        <%x>\n"
1184			"tx_buffer_info[next_to_clean]\n"
1185			"  time_stamp           <%lx>\n"
1186			"  jiffies              <%lx>\n",
1187			tx_ring->queue_index,
1188			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1189			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1190			tx_ring->next_to_use, i,
1191			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1192
1193		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1194
1195		e_info(probe,
1196		       "tx hang %d detected on queue %d, resetting adapter\n",
1197			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1198
1199		/* schedule immediate reset if we believe we hung */
1200		ixgbe_tx_timeout_reset(adapter);
1201
1202		/* the adapter is about to reset, no point in enabling stuff */
1203		return true;
1204	}
1205
1206	netdev_tx_completed_queue(txring_txq(tx_ring),
1207				  total_packets, total_bytes);
1208
1209#define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1210	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1211		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1212		/* Make sure that anybody stopping the queue after this
1213		 * sees the new next_to_clean.
1214		 */
1215		smp_mb();
1216		if (__netif_subqueue_stopped(tx_ring->netdev,
1217					     tx_ring->queue_index)
1218		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1219			netif_wake_subqueue(tx_ring->netdev,
1220					    tx_ring->queue_index);
1221			++tx_ring->tx_stats.restart_queue;
1222		}
1223	}
1224
1225	return !!budget;
1226}
1227
1228#ifdef CONFIG_IXGBE_DCA
1229static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1230				struct ixgbe_ring *tx_ring,
1231				int cpu)
1232{
1233	struct ixgbe_hw *hw = &adapter->hw;
1234	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
1235	u16 reg_offset;
1236
1237	switch (hw->mac.type) {
1238	case ixgbe_mac_82598EB:
1239		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1240		break;
1241	case ixgbe_mac_82599EB:
1242	case ixgbe_mac_X540:
1243		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1244		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1245		break;
1246	default:
1247		/* for unknown hardware do not write register */
1248		return;
1249	}
1250
1251	/*
1252	 * We can enable relaxed ordering for reads, but not writes when
1253	 * DCA is enabled.  This is due to a known issue in some chipsets
1254	 * which will cause the DCA tag to be cleared.
1255	 */
1256	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1257		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1258		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1259
1260	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1261}
1262
1263static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1264				struct ixgbe_ring *rx_ring,
1265				int cpu)
1266{
1267	struct ixgbe_hw *hw = &adapter->hw;
1268	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1269	u8 reg_idx = rx_ring->reg_idx;
1270
1271
1272	switch (hw->mac.type) {
1273	case ixgbe_mac_82599EB:
1274	case ixgbe_mac_X540:
1275		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1276		break;
1277	default:
1278		break;
1279	}
1280
1281	/*
1282	 * We can enable relaxed ordering for reads, but not writes when
1283	 * DCA is enabled.  This is due to a known issue in some chipsets
1284	 * which will cause the DCA tag to be cleared.
1285	 */
1286	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1287		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1288
1289	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1290}
1291
1292static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1293{
1294	struct ixgbe_adapter *adapter = q_vector->adapter;
1295	struct ixgbe_ring *ring;
1296	int cpu = get_cpu();
1297
1298	if (q_vector->cpu == cpu)
1299		goto out_no_update;
1300
1301	ixgbe_for_each_ring(ring, q_vector->tx)
1302		ixgbe_update_tx_dca(adapter, ring, cpu);
1303
1304	ixgbe_for_each_ring(ring, q_vector->rx)
1305		ixgbe_update_rx_dca(adapter, ring, cpu);
1306
1307	q_vector->cpu = cpu;
1308out_no_update:
1309	put_cpu();
1310}
1311
1312static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1313{
1314	int i;
1315
1316	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
1317		return;
1318
1319	/* always use CB2 mode, difference is masked in the CB driver */
1320	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
1321
1322	for (i = 0; i < adapter->num_q_vectors; i++) {
1323		adapter->q_vector[i]->cpu = -1;
1324		ixgbe_update_dca(adapter->q_vector[i]);
1325	}
1326}
1327
1328static int __ixgbe_notify_dca(struct device *dev, void *data)
1329{
1330	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1331	unsigned long event = *(unsigned long *)data;
1332
1333	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1334		return 0;
1335
1336	switch (event) {
1337	case DCA_PROVIDER_ADD:
1338		/* if we're already enabled, don't do it again */
1339		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1340			break;
1341		if (dca_add_requester(dev) == 0) {
1342			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1343			ixgbe_setup_dca(adapter);
1344			break;
1345		}
1346		/* Fall Through since DCA is disabled. */
1347	case DCA_PROVIDER_REMOVE:
1348		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1349			dca_remove_requester(dev);
1350			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1351			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1352		}
1353		break;
1354	}
1355
1356	return 0;
1357}
1358
1359#endif /* CONFIG_IXGBE_DCA */
1360static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1361				 union ixgbe_adv_rx_desc *rx_desc,
1362				 struct sk_buff *skb)
1363{
1364	if (ring->netdev->features & NETIF_F_RXHASH)
1365		skb_set_hash(skb,
1366			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1367			     PKT_HASH_TYPE_L3);
1368}
1369
1370#ifdef IXGBE_FCOE
1371/**
1372 * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1373 * @ring: structure containing ring specific data
1374 * @rx_desc: advanced rx descriptor
1375 *
1376 * Returns : true if it is FCoE pkt
1377 */
1378static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1379				    union ixgbe_adv_rx_desc *rx_desc)
1380{
1381	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1382
1383	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1384	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1385		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1386			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1387}
1388
1389#endif /* IXGBE_FCOE */
1390/**
1391 * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1392 * @ring: structure containing ring specific data
1393 * @rx_desc: current Rx descriptor being processed
1394 * @skb: skb currently being received and modified
1395 **/
1396static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1397				     union ixgbe_adv_rx_desc *rx_desc,
1398				     struct sk_buff *skb)
1399{
1400	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1401	__le16 hdr_info = rx_desc->wb.lower.lo_dword.hs_rss.hdr_info;
1402	bool encap_pkt = false;
1403
1404	skb_checksum_none_assert(skb);
1405
1406	/* Rx csum disabled */
1407	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1408		return;
1409
1410	if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) &&
1411	    (hdr_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_TUNNEL >> 16))) {
1412		encap_pkt = true;
1413		skb->encapsulation = 1;
1414		skb->ip_summed = CHECKSUM_NONE;
1415	}
1416
1417	/* if IP and error */
1418	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1419	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1420		ring->rx_stats.csum_err++;
1421		return;
1422	}
1423
1424	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1425		return;
1426
1427	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1428		/*
1429		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1430		 * checksum errors.
1431		 */
1432		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1433		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1434			return;
1435
1436		ring->rx_stats.csum_err++;
1437		return;
1438	}
1439
1440	/* It must be a TCP or UDP packet with a valid checksum */
1441	skb->ip_summed = CHECKSUM_UNNECESSARY;
1442	if (encap_pkt) {
1443		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1444			return;
1445
1446		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1447			ring->rx_stats.csum_err++;
1448			return;
1449		}
1450		/* If we checked the outer header let the stack know */
1451		skb->csum_level = 1;
1452	}
1453}
1454
1455static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1456				    struct ixgbe_rx_buffer *bi)
1457{
1458	struct page *page = bi->page;
1459	dma_addr_t dma;
1460
1461	/* since we are recycling buffers we should seldom need to alloc */
1462	if (likely(page))
1463		return true;
1464
1465	/* alloc new page for storage */
1466	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1467	if (unlikely(!page)) {
1468		rx_ring->rx_stats.alloc_rx_page_failed++;
1469		return false;
1470	}
1471
1472	/* map page for use */
1473	dma = dma_map_page(rx_ring->dev, page, 0,
1474			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1475
1476	/*
1477	 * if mapping failed free memory back to system since
1478	 * there isn't much point in holding memory we can't use
1479	 */
1480	if (dma_mapping_error(rx_ring->dev, dma)) {
1481		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1482
1483		rx_ring->rx_stats.alloc_rx_page_failed++;
1484		return false;
1485	}
1486
1487	bi->dma = dma;
1488	bi->page = page;
1489	bi->page_offset = 0;
1490
1491	return true;
1492}
1493
1494/**
1495 * ixgbe_alloc_rx_buffers - Replace used receive buffers
1496 * @rx_ring: ring to place buffers on
1497 * @cleaned_count: number of buffers to replace
1498 **/
1499void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1500{
1501	union ixgbe_adv_rx_desc *rx_desc;
1502	struct ixgbe_rx_buffer *bi;
1503	u16 i = rx_ring->next_to_use;
1504
1505	/* nothing to do */
1506	if (!cleaned_count)
1507		return;
1508
1509	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1510	bi = &rx_ring->rx_buffer_info[i];
1511	i -= rx_ring->count;
1512
1513	do {
1514		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1515			break;
1516
1517		/*
1518		 * Refresh the desc even if buffer_addrs didn't change
1519		 * because each write-back erases this info.
1520		 */
1521		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1522
1523		rx_desc++;
1524		bi++;
1525		i++;
1526		if (unlikely(!i)) {
1527			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1528			bi = rx_ring->rx_buffer_info;
1529			i -= rx_ring->count;
1530		}
1531
1532		/* clear the status bits for the next_to_use descriptor */
1533		rx_desc->wb.upper.status_error = 0;
1534
1535		cleaned_count--;
1536	} while (cleaned_count);
1537
1538	i += rx_ring->count;
1539
1540	if (rx_ring->next_to_use != i) {
1541		rx_ring->next_to_use = i;
1542
1543		/* update next to alloc since we have filled the ring */
1544		rx_ring->next_to_alloc = i;
1545
1546		/* Force memory writes to complete before letting h/w
1547		 * know there are new descriptors to fetch.  (Only
1548		 * applicable for weak-ordered memory model archs,
1549		 * such as IA-64).
1550		 */
1551		wmb();
1552		writel(i, rx_ring->tail);
1553	}
1554}
1555
1556static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1557				   struct sk_buff *skb)
1558{
1559	u16 hdr_len = skb_headlen(skb);
1560
1561	/* set gso_size to avoid messing up TCP MSS */
1562	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1563						 IXGBE_CB(skb)->append_cnt);
1564	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1565}
1566
1567static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1568				   struct sk_buff *skb)
1569{
1570	/* if append_cnt is 0 then frame is not RSC */
1571	if (!IXGBE_CB(skb)->append_cnt)
1572		return;
1573
1574	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1575	rx_ring->rx_stats.rsc_flush++;
1576
1577	ixgbe_set_rsc_gso_size(rx_ring, skb);
1578
1579	/* gso_size is computed using append_cnt so always clear it last */
1580	IXGBE_CB(skb)->append_cnt = 0;
1581}
1582
1583/**
1584 * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1585 * @rx_ring: rx descriptor ring packet is being transacted on
1586 * @rx_desc: pointer to the EOP Rx descriptor
1587 * @skb: pointer to current skb being populated
1588 *
1589 * This function checks the ring, descriptor, and packet information in
1590 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1591 * other fields within the skb.
1592 **/
1593static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1594				     union ixgbe_adv_rx_desc *rx_desc,
1595				     struct sk_buff *skb)
1596{
1597	struct net_device *dev = rx_ring->netdev;
1598
1599	ixgbe_update_rsc_stats(rx_ring, skb);
1600
1601	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1602
1603	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1604
1605	if (unlikely(ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_TS)))
1606		ixgbe_ptp_rx_hwtstamp(rx_ring->q_vector->adapter, skb);
1607
1608	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1609	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1610		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1611		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1612	}
1613
1614	skb_record_rx_queue(skb, rx_ring->queue_index);
1615
1616	skb->protocol = eth_type_trans(skb, dev);
1617}
1618
1619static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1620			 struct sk_buff *skb)
1621{
1622	if (ixgbe_qv_busy_polling(q_vector))
1623		netif_receive_skb(skb);
1624	else
1625		napi_gro_receive(&q_vector->napi, skb);
1626}
1627
1628/**
1629 * ixgbe_is_non_eop - process handling of non-EOP buffers
1630 * @rx_ring: Rx ring being processed
1631 * @rx_desc: Rx descriptor for current buffer
1632 * @skb: Current socket buffer containing buffer in progress
1633 *
1634 * This function updates next to clean.  If the buffer is an EOP buffer
1635 * this function exits returning false, otherwise it will place the
1636 * sk_buff in the next buffer to be chained and return true indicating
1637 * that this is in fact a non-EOP buffer.
1638 **/
1639static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1640			     union ixgbe_adv_rx_desc *rx_desc,
1641			     struct sk_buff *skb)
1642{
1643	u32 ntc = rx_ring->next_to_clean + 1;
1644
1645	/* fetch, update, and store next to clean */
1646	ntc = (ntc < rx_ring->count) ? ntc : 0;
1647	rx_ring->next_to_clean = ntc;
1648
1649	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1650
1651	/* update RSC append count if present */
1652	if (ring_is_rsc_enabled(rx_ring)) {
1653		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1654				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1655
1656		if (unlikely(rsc_enabled)) {
1657			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1658
1659			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1660			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1661
1662			/* update ntc based on RSC value */
1663			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1664			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1665			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1666		}
1667	}
1668
1669	/* if we are the last buffer then there is nothing else to do */
1670	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1671		return false;
1672
1673	/* place skb in next buffer to be received */
1674	rx_ring->rx_buffer_info[ntc].skb = skb;
1675	rx_ring->rx_stats.non_eop_descs++;
1676
1677	return true;
1678}
1679
1680/**
1681 * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1682 * @rx_ring: rx descriptor ring packet is being transacted on
1683 * @skb: pointer to current skb being adjusted
1684 *
1685 * This function is an ixgbe specific version of __pskb_pull_tail.  The
1686 * main difference between this version and the original function is that
1687 * this function can make several assumptions about the state of things
1688 * that allow for significant optimizations versus the standard function.
1689 * As a result we can do things like drop a frag and maintain an accurate
1690 * truesize for the skb.
1691 */
1692static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1693			    struct sk_buff *skb)
1694{
1695	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1696	unsigned char *va;
1697	unsigned int pull_len;
1698
1699	/*
1700	 * it is valid to use page_address instead of kmap since we are
1701	 * working with pages allocated out of the lomem pool per
1702	 * alloc_page(GFP_ATOMIC)
1703	 */
1704	va = skb_frag_address(frag);
1705
1706	/*
1707	 * we need the header to contain the greater of either ETH_HLEN or
1708	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1709	 */
1710	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1711
1712	/* align pull length to size of long to optimize memcpy performance */
1713	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1714
1715	/* update all of the pointers */
1716	skb_frag_size_sub(frag, pull_len);
1717	frag->page_offset += pull_len;
1718	skb->data_len -= pull_len;
1719	skb->tail += pull_len;
1720}
1721
1722/**
1723 * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1724 * @rx_ring: rx descriptor ring packet is being transacted on
1725 * @skb: pointer to current skb being updated
1726 *
1727 * This function provides a basic DMA sync up for the first fragment of an
1728 * skb.  The reason for doing this is that the first fragment cannot be
1729 * unmapped until we have reached the end of packet descriptor for a buffer
1730 * chain.
1731 */
1732static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1733				struct sk_buff *skb)
1734{
1735	/* if the page was released unmap it, else just sync our portion */
1736	if (unlikely(IXGBE_CB(skb)->page_released)) {
1737		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1738			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1739		IXGBE_CB(skb)->page_released = false;
1740	} else {
1741		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1742
1743		dma_sync_single_range_for_cpu(rx_ring->dev,
1744					      IXGBE_CB(skb)->dma,
1745					      frag->page_offset,
1746					      ixgbe_rx_bufsz(rx_ring),
1747					      DMA_FROM_DEVICE);
1748	}
1749	IXGBE_CB(skb)->dma = 0;
1750}
1751
1752/**
1753 * ixgbe_cleanup_headers - Correct corrupted or empty headers
1754 * @rx_ring: rx descriptor ring packet is being transacted on
1755 * @rx_desc: pointer to the EOP Rx descriptor
1756 * @skb: pointer to current skb being fixed
1757 *
1758 * Check for corrupted packet headers caused by senders on the local L2
1759 * embedded NIC switch not setting up their Tx Descriptors right.  These
1760 * should be very rare.
1761 *
1762 * Also address the case where we are pulling data in on pages only
1763 * and as such no data is present in the skb header.
1764 *
1765 * In addition if skb is not at least 60 bytes we need to pad it so that
1766 * it is large enough to qualify as a valid Ethernet frame.
1767 *
1768 * Returns true if an error was encountered and skb was freed.
1769 **/
1770static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1771				  union ixgbe_adv_rx_desc *rx_desc,
1772				  struct sk_buff *skb)
1773{
1774	struct net_device *netdev = rx_ring->netdev;
1775
1776	/* verify that the packet does not have any known errors */
1777	if (unlikely(ixgbe_test_staterr(rx_desc,
1778					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1779	    !(netdev->features & NETIF_F_RXALL))) {
1780		dev_kfree_skb_any(skb);
1781		return true;
1782	}
1783
1784	/* place header in linear portion of buffer */
1785	if (skb_is_nonlinear(skb))
1786		ixgbe_pull_tail(rx_ring, skb);
1787
1788#ifdef IXGBE_FCOE
1789	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1790	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1791		return false;
1792
1793#endif
1794	/* if eth_skb_pad returns an error the skb was freed */
1795	if (eth_skb_pad(skb))
1796		return true;
1797
1798	return false;
1799}
1800
1801/**
1802 * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1803 * @rx_ring: rx descriptor ring to store buffers on
1804 * @old_buff: donor buffer to have page reused
1805 *
1806 * Synchronizes page for reuse by the adapter
1807 **/
1808static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1809				struct ixgbe_rx_buffer *old_buff)
1810{
1811	struct ixgbe_rx_buffer *new_buff;
1812	u16 nta = rx_ring->next_to_alloc;
1813
1814	new_buff = &rx_ring->rx_buffer_info[nta];
1815
1816	/* update, and store next to alloc */
1817	nta++;
1818	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1819
1820	/* transfer page from old buffer to new buffer */
1821	*new_buff = *old_buff;
1822
1823	/* sync the buffer for use by the device */
1824	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1825					 new_buff->page_offset,
1826					 ixgbe_rx_bufsz(rx_ring),
1827					 DMA_FROM_DEVICE);
1828}
1829
1830static inline bool ixgbe_page_is_reserved(struct page *page)
1831{
1832	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
1833}
1834
1835/**
1836 * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1837 * @rx_ring: rx descriptor ring to transact packets on
1838 * @rx_buffer: buffer containing page to add
1839 * @rx_desc: descriptor containing length of buffer written by hardware
1840 * @skb: sk_buff to place the data into
1841 *
1842 * This function will add the data contained in rx_buffer->page to the skb.
1843 * This is done either through a direct copy if the data in the buffer is
1844 * less than the skb header size, otherwise it will just attach the page as
1845 * a frag to the skb.
1846 *
1847 * The function will then update the page offset if necessary and return
1848 * true if the buffer can be reused by the adapter.
1849 **/
1850static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1851			      struct ixgbe_rx_buffer *rx_buffer,
1852			      union ixgbe_adv_rx_desc *rx_desc,
1853			      struct sk_buff *skb)
1854{
1855	struct page *page = rx_buffer->page;
1856	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1857#if (PAGE_SIZE < 8192)
1858	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1859#else
1860	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1861	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1862				   ixgbe_rx_bufsz(rx_ring);
1863#endif
1864
1865	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1866		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1867
1868		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1869
1870		/* page is not reserved, we can reuse buffer as-is */
1871		if (likely(!ixgbe_page_is_reserved(page)))
1872			return true;
1873
1874		/* this page cannot be reused so discard it */
1875		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1876		return false;
1877	}
1878
1879	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1880			rx_buffer->page_offset, size, truesize);
1881
1882	/* avoid re-using remote pages */
1883	if (unlikely(ixgbe_page_is_reserved(page)))
1884		return false;
1885
1886#if (PAGE_SIZE < 8192)
1887	/* if we are only owner of page we can reuse it */
1888	if (unlikely(page_count(page) != 1))
1889		return false;
1890
1891	/* flip page offset to other buffer */
1892	rx_buffer->page_offset ^= truesize;
1893#else
1894	/* move offset up to the next cache line */
1895	rx_buffer->page_offset += truesize;
1896
1897	if (rx_buffer->page_offset > last_offset)
1898		return false;
1899#endif
1900
1901	/* Even if we own the page, we are not allowed to use atomic_set()
1902	 * This would break get_page_unless_zero() users.
1903	 */
1904	atomic_inc(&page->_count);
1905
1906	return true;
1907}
1908
1909static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
1910					     union ixgbe_adv_rx_desc *rx_desc)
1911{
1912	struct ixgbe_rx_buffer *rx_buffer;
1913	struct sk_buff *skb;
1914	struct page *page;
1915
1916	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
1917	page = rx_buffer->page;
1918	prefetchw(page);
1919
1920	skb = rx_buffer->skb;
1921
1922	if (likely(!skb)) {
1923		void *page_addr = page_address(page) +
1924				  rx_buffer->page_offset;
1925
1926		/* prefetch first cache line of first page */
1927		prefetch(page_addr);
1928#if L1_CACHE_BYTES < 128
1929		prefetch(page_addr + L1_CACHE_BYTES);
1930#endif
1931
1932		/* allocate a skb to store the frags */
1933		skb = napi_alloc_skb(&rx_ring->q_vector->napi,
1934				     IXGBE_RX_HDR_SIZE);
1935		if (unlikely(!skb)) {
1936			rx_ring->rx_stats.alloc_rx_buff_failed++;
1937			return NULL;
1938		}
1939
1940		/*
1941		 * we will be copying header into skb->data in
1942		 * pskb_may_pull so it is in our interest to prefetch
1943		 * it now to avoid a possible cache miss
1944		 */
1945		prefetchw(skb->data);
1946
1947		/*
1948		 * Delay unmapping of the first packet. It carries the
1949		 * header information, HW may still access the header
1950		 * after the writeback.  Only unmap it when EOP is
1951		 * reached
1952		 */
1953		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1954			goto dma_sync;
1955
1956		IXGBE_CB(skb)->dma = rx_buffer->dma;
1957	} else {
1958		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
1959			ixgbe_dma_sync_frag(rx_ring, skb);
1960
1961dma_sync:
1962		/* we are reusing so sync this buffer for CPU use */
1963		dma_sync_single_range_for_cpu(rx_ring->dev,
1964					      rx_buffer->dma,
1965					      rx_buffer->page_offset,
1966					      ixgbe_rx_bufsz(rx_ring),
1967					      DMA_FROM_DEVICE);
1968
1969		rx_buffer->skb = NULL;
1970	}
1971
1972	/* pull page into skb */
1973	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
1974		/* hand second half of page back to the ring */
1975		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
1976	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
1977		/* the page has been released from the ring */
1978		IXGBE_CB(skb)->page_released = true;
1979	} else {
1980		/* we are not reusing the buffer so unmap it */
1981		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
1982			       ixgbe_rx_pg_size(rx_ring),
1983			       DMA_FROM_DEVICE);
1984	}
1985
1986	/* clear contents of buffer_info */
1987	rx_buffer->page = NULL;
1988
1989	return skb;
1990}
1991
1992/**
1993 * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1994 * @q_vector: structure containing interrupt and ring information
1995 * @rx_ring: rx descriptor ring to transact packets on
1996 * @budget: Total limit on number of packets to process
1997 *
1998 * This function provides a "bounce buffer" approach to Rx interrupt
1999 * processing.  The advantage to this is that on systems that have
2000 * expensive overhead for IOMMU access this provides a means of avoiding
2001 * it by maintaining the mapping of the page to the syste.
2002 *
2003 * Returns amount of work completed
2004 **/
2005static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2006			       struct ixgbe_ring *rx_ring,
2007			       const int budget)
2008{
2009	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2010#ifdef IXGBE_FCOE
2011	struct ixgbe_adapter *adapter = q_vector->adapter;
2012	int ddp_bytes;
2013	unsigned int mss = 0;
2014#endif /* IXGBE_FCOE */
2015	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2016
2017	while (likely(total_rx_packets < budget)) {
2018		union ixgbe_adv_rx_desc *rx_desc;
2019		struct sk_buff *skb;
2020
2021		/* return some buffers to hardware, one at a time is too slow */
2022		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2023			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2024			cleaned_count = 0;
2025		}
2026
2027		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2028
2029		if (!rx_desc->wb.upper.status_error)
2030			break;
2031
2032		/* This memory barrier is needed to keep us from reading
2033		 * any other fields out of the rx_desc until we know the
2034		 * descriptor has been written back
2035		 */
2036		dma_rmb();
2037
2038		/* retrieve a buffer from the ring */
2039		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
2040
2041		/* exit if we failed to retrieve a buffer */
2042		if (!skb)
2043			break;
2044
2045		cleaned_count++;
2046
2047		/* place incomplete frames back on ring for completion */
2048		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2049			continue;
2050
2051		/* verify the packet layout is correct */
2052		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2053			continue;
2054
2055		/* probably a little skewed due to removing CRC */
2056		total_rx_bytes += skb->len;
2057
2058		/* populate checksum, timestamp, VLAN, and protocol */
2059		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2060
2061#ifdef IXGBE_FCOE
2062		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2063		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2064			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2065			/* include DDPed FCoE data */
2066			if (ddp_bytes > 0) {
2067				if (!mss) {
2068					mss = rx_ring->netdev->mtu -
2069						sizeof(struct fcoe_hdr) -
2070						sizeof(struct fc_frame_header) -
2071						sizeof(struct fcoe_crc_eof);
2072					if (mss > 512)
2073						mss &= ~511;
2074				}
2075				total_rx_bytes += ddp_bytes;
2076				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2077								 mss);
2078			}
2079			if (!ddp_bytes) {
2080				dev_kfree_skb_any(skb);
2081				continue;
2082			}
2083		}
2084
2085#endif /* IXGBE_FCOE */
2086		skb_mark_napi_id(skb, &q_vector->napi);
2087		ixgbe_rx_skb(q_vector, skb);
2088
2089		/* update budget accounting */
2090		total_rx_packets++;
2091	}
2092
2093	u64_stats_update_begin(&rx_ring->syncp);
2094	rx_ring->stats.packets += total_rx_packets;
2095	rx_ring->stats.bytes += total_rx_bytes;
2096	u64_stats_update_end(&rx_ring->syncp);
2097	q_vector->rx.total_packets += total_rx_packets;
2098	q_vector->rx.total_bytes += total_rx_bytes;
2099
2100	return total_rx_packets;
2101}
2102
2103#ifdef CONFIG_NET_RX_BUSY_POLL
2104/* must be called with local_bh_disable()d */
2105static int ixgbe_low_latency_recv(struct napi_struct *napi)
2106{
2107	struct ixgbe_q_vector *q_vector =
2108			container_of(napi, struct ixgbe_q_vector, napi);
2109	struct ixgbe_adapter *adapter = q_vector->adapter;
2110	struct ixgbe_ring  *ring;
2111	int found = 0;
2112
2113	if (test_bit(__IXGBE_DOWN, &adapter->state))
2114		return LL_FLUSH_FAILED;
2115
2116	if (!ixgbe_qv_lock_poll(q_vector))
2117		return LL_FLUSH_BUSY;
2118
2119	ixgbe_for_each_ring(ring, q_vector->rx) {
2120		found = ixgbe_clean_rx_irq(q_vector, ring, 4);
2121#ifdef BP_EXTENDED_STATS
2122		if (found)
2123			ring->stats.cleaned += found;
2124		else
2125			ring->stats.misses++;
2126#endif
2127		if (found)
2128			break;
2129	}
2130
2131	ixgbe_qv_unlock_poll(q_vector);
2132
2133	return found;
2134}
2135#endif	/* CONFIG_NET_RX_BUSY_POLL */
2136
2137/**
2138 * ixgbe_configure_msix - Configure MSI-X hardware
2139 * @adapter: board private structure
2140 *
2141 * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2142 * interrupts.
2143 **/
2144static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2145{
2146	struct ixgbe_q_vector *q_vector;
2147	int v_idx;
2148	u32 mask;
2149
2150	/* Populate MSIX to EITR Select */
2151	if (adapter->num_vfs > 32) {
2152		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
2153		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2154	}
2155
2156	/*
2157	 * Populate the IVAR table and set the ITR values to the
2158	 * corresponding register.
2159	 */
2160	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2161		struct ixgbe_ring *ring;
2162		q_vector = adapter->q_vector[v_idx];
2163
2164		ixgbe_for_each_ring(ring, q_vector->rx)
2165			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2166
2167		ixgbe_for_each_ring(ring, q_vector->tx)
2168			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2169
2170		ixgbe_write_eitr(q_vector);
2171	}
2172
2173	switch (adapter->hw.mac.type) {
2174	case ixgbe_mac_82598EB:
2175		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2176			       v_idx);
2177		break;
2178	case ixgbe_mac_82599EB:
2179	case ixgbe_mac_X540:
2180	case ixgbe_mac_X550:
2181	case ixgbe_mac_X550EM_x:
2182		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2183		break;
2184	default:
2185		break;
2186	}
2187	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2188
2189	/* set up to autoclear timer, and the vectors */
2190	mask = IXGBE_EIMS_ENABLE_MASK;
2191	mask &= ~(IXGBE_EIMS_OTHER |
2192		  IXGBE_EIMS_MAILBOX |
2193		  IXGBE_EIMS_LSC);
2194
2195	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2196}
2197
2198enum latency_range {
2199	lowest_latency = 0,
2200	low_latency = 1,
2201	bulk_latency = 2,
2202	latency_invalid = 255
2203};
2204
2205/**
2206 * ixgbe_update_itr - update the dynamic ITR value based on statistics
2207 * @q_vector: structure containing interrupt and ring information
2208 * @ring_container: structure containing ring performance data
2209 *
2210 *      Stores a new ITR value based on packets and byte
2211 *      counts during the last interrupt.  The advantage of per interrupt
2212 *      computation is faster updates and more accurate ITR for the current
2213 *      traffic pattern.  Constants in this function were computed
2214 *      based on theoretical maximum wire speed and thresholds were set based
2215 *      on testing data as well as attempting to minimize response time
2216 *      while increasing bulk throughput.
2217 *      this functionality is controlled by the InterruptThrottleRate module
2218 *      parameter (see ixgbe_param.c)
2219 **/
2220static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2221			     struct ixgbe_ring_container *ring_container)
2222{
2223	int bytes = ring_container->total_bytes;
2224	int packets = ring_container->total_packets;
2225	u32 timepassed_us;
2226	u64 bytes_perint;
2227	u8 itr_setting = ring_container->itr;
2228
2229	if (packets == 0)
2230		return;
2231
2232	/* simple throttlerate management
2233	 *   0-10MB/s   lowest (100000 ints/s)
2234	 *  10-20MB/s   low    (20000 ints/s)
2235	 *  20-1249MB/s bulk   (8000 ints/s)
2236	 */
2237	/* what was last interrupt timeslice? */
2238	timepassed_us = q_vector->itr >> 2;
2239	if (timepassed_us == 0)
2240		return;
2241
2242	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2243
2244	switch (itr_setting) {
2245	case lowest_latency:
2246		if (bytes_perint > 10)
2247			itr_setting = low_latency;
2248		break;
2249	case low_latency:
2250		if (bytes_perint > 20)
2251			itr_setting = bulk_latency;
2252		else if (bytes_perint <= 10)
2253			itr_setting = lowest_latency;
2254		break;
2255	case bulk_latency:
2256		if (bytes_perint <= 20)
2257			itr_setting = low_latency;
2258		break;
2259	}
2260
2261	/* clear work counters since we have the values we need */
2262	ring_container->total_bytes = 0;
2263	ring_container->total_packets = 0;
2264
2265	/* write updated itr to ring container */
2266	ring_container->itr = itr_setting;
2267}
2268
2269/**
2270 * ixgbe_write_eitr - write EITR register in hardware specific way
2271 * @q_vector: structure containing interrupt and ring information
2272 *
2273 * This function is made to be called by ethtool and by the driver
2274 * when it needs to update EITR registers at runtime.  Hardware
2275 * specific quirks/differences are taken care of here.
2276 */
2277void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2278{
2279	struct ixgbe_adapter *adapter = q_vector->adapter;
2280	struct ixgbe_hw *hw = &adapter->hw;
2281	int v_idx = q_vector->v_idx;
2282	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2283
2284	switch (adapter->hw.mac.type) {
2285	case ixgbe_mac_82598EB:
2286		/* must write high and low 16 bits to reset counter */
2287		itr_reg |= (itr_reg << 16);
2288		break;
2289	case ixgbe_mac_82599EB:
2290	case ixgbe_mac_X540:
2291	case ixgbe_mac_X550:
2292	case ixgbe_mac_X550EM_x:
2293		/*
2294		 * set the WDIS bit to not clear the timer bits and cause an
2295		 * immediate assertion of the interrupt
2296		 */
2297		itr_reg |= IXGBE_EITR_CNT_WDIS;
2298		break;
2299	default:
2300		break;
2301	}
2302	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2303}
2304
2305static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2306{
2307	u32 new_itr = q_vector->itr;
2308	u8 current_itr;
2309
2310	ixgbe_update_itr(q_vector, &q_vector->tx);
2311	ixgbe_update_itr(q_vector, &q_vector->rx);
2312
2313	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2314
2315	switch (current_itr) {
2316	/* counts and packets in update_itr are dependent on these numbers */
2317	case lowest_latency:
2318		new_itr = IXGBE_100K_ITR;
2319		break;
2320	case low_latency:
2321		new_itr = IXGBE_20K_ITR;
2322		break;
2323	case bulk_latency:
2324		new_itr = IXGBE_8K_ITR;
2325		break;
2326	default:
2327		break;
2328	}
2329
2330	if (new_itr != q_vector->itr) {
2331		/* do an exponential smoothing */
2332		new_itr = (10 * new_itr * q_vector->itr) /
2333			  ((9 * new_itr) + q_vector->itr);
2334
2335		/* save the algorithm value here */
2336		q_vector->itr = new_itr;
2337
2338		ixgbe_write_eitr(q_vector);
2339	}
2340}
2341
2342/**
2343 * ixgbe_check_overtemp_subtask - check for over temperature
2344 * @adapter: pointer to adapter
2345 **/
2346static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2347{
2348	struct ixgbe_hw *hw = &adapter->hw;
2349	u32 eicr = adapter->interrupt_event;
2350
2351	if (test_bit(__IXGBE_DOWN, &adapter->state))
2352		return;
2353
2354	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2355	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2356		return;
2357
2358	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2359
2360	switch (hw->device_id) {
2361	case IXGBE_DEV_ID_82599_T3_LOM:
2362		/*
2363		 * Since the warning interrupt is for both ports
2364		 * we don't have to check if:
2365		 *  - This interrupt wasn't for our port.
2366		 *  - We may have missed the interrupt so always have to
2367		 *    check if we  got a LSC
2368		 */
2369		if (!(eicr & IXGBE_EICR_GPI_SDP0) &&
2370		    !(eicr & IXGBE_EICR_LSC))
2371			return;
2372
2373		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2374			u32 speed;
2375			bool link_up = false;
2376
2377			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2378
2379			if (link_up)
2380				return;
2381		}
2382
2383		/* Check if this is not due to overtemp */
2384		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2385			return;
2386
2387		break;
2388	default:
2389		if (!(eicr & IXGBE_EICR_GPI_SDP0))
2390			return;
2391		break;
2392	}
2393	e_crit(drv,
2394	       "Network adapter has been stopped because it has over heated. "
2395	       "Restart the computer. If the problem persists, "
2396	       "power off the system and replace the adapter\n");
2397
2398	adapter->interrupt_event = 0;
2399}
2400
2401static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2402{
2403	struct ixgbe_hw *hw = &adapter->hw;
2404
2405	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2406	    (eicr & IXGBE_EICR_GPI_SDP1)) {
2407		e_crit(probe, "Fan has stopped, replace the adapter\n");
2408		/* write to clear the interrupt */
2409		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2410	}
2411}
2412
2413static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2414{
2415	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2416		return;
2417
2418	switch (adapter->hw.mac.type) {
2419	case ixgbe_mac_82599EB:
2420		/*
2421		 * Need to check link state so complete overtemp check
2422		 * on service task
2423		 */
2424		if (((eicr & IXGBE_EICR_GPI_SDP0) || (eicr & IXGBE_EICR_LSC)) &&
2425		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2426			adapter->interrupt_event = eicr;
2427			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2428			ixgbe_service_event_schedule(adapter);
2429			return;
2430		}
2431		return;
2432	case ixgbe_mac_X540:
2433		if (!(eicr & IXGBE_EICR_TS))
2434			return;
2435		break;
2436	default:
2437		return;
2438	}
2439
2440	e_crit(drv,
2441	       "Network adapter has been stopped because it has over heated. "
2442	       "Restart the computer. If the problem persists, "
2443	       "power off the system and replace the adapter\n");
2444}
2445
2446static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2447{
2448	struct ixgbe_hw *hw = &adapter->hw;
2449
2450	if (eicr & IXGBE_EICR_GPI_SDP2) {
2451		/* Clear the interrupt */
2452		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP2);
2453		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2454			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2455			ixgbe_service_event_schedule(adapter);
2456		}
2457	}
2458
2459	if (eicr & IXGBE_EICR_GPI_SDP1) {
2460		/* Clear the interrupt */
2461		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2462		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2463			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2464			ixgbe_service_event_schedule(adapter);
2465		}
2466	}
2467}
2468
2469static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2470{
2471	struct ixgbe_hw *hw = &adapter->hw;
2472
2473	adapter->lsc_int++;
2474	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2475	adapter->link_check_timeout = jiffies;
2476	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2477		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2478		IXGBE_WRITE_FLUSH(hw);
2479		ixgbe_service_event_schedule(adapter);
2480	}
2481}
2482
2483static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2484					   u64 qmask)
2485{
2486	u32 mask;
2487	struct ixgbe_hw *hw = &adapter->hw;
2488
2489	switch (hw->mac.type) {
2490	case ixgbe_mac_82598EB:
2491		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2492		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2493		break;
2494	case ixgbe_mac_82599EB:
2495	case ixgbe_mac_X540:
2496	case ixgbe_mac_X550:
2497	case ixgbe_mac_X550EM_x:
2498		mask = (qmask & 0xFFFFFFFF);
2499		if (mask)
2500			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2501		mask = (qmask >> 32);
2502		if (mask)
2503			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2504		break;
2505	default:
2506		break;
2507	}
2508	/* skip the flush */
2509}
2510
2511static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2512					    u64 qmask)
2513{
2514	u32 mask;
2515	struct ixgbe_hw *hw = &adapter->hw;
2516
2517	switch (hw->mac.type) {
2518	case ixgbe_mac_82598EB:
2519		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2520		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2521		break;
2522	case ixgbe_mac_82599EB:
2523	case ixgbe_mac_X540:
2524	case ixgbe_mac_X550:
2525	case ixgbe_mac_X550EM_x:
2526		mask = (qmask & 0xFFFFFFFF);
2527		if (mask)
2528			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2529		mask = (qmask >> 32);
2530		if (mask)
2531			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2532		break;
2533	default:
2534		break;
2535	}
2536	/* skip the flush */
2537}
2538
2539/**
2540 * ixgbe_irq_enable - Enable default interrupt generation settings
2541 * @adapter: board private structure
2542 **/
2543static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2544				    bool flush)
2545{
2546	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2547
2548	/* don't reenable LSC while waiting for link */
2549	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2550		mask &= ~IXGBE_EIMS_LSC;
2551
2552	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2553		switch (adapter->hw.mac.type) {
2554		case ixgbe_mac_82599EB:
2555			mask |= IXGBE_EIMS_GPI_SDP0;
2556			break;
2557		case ixgbe_mac_X540:
2558		case ixgbe_mac_X550:
2559		case ixgbe_mac_X550EM_x:
2560			mask |= IXGBE_EIMS_TS;
2561			break;
2562		default:
2563			break;
2564		}
2565	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2566		mask |= IXGBE_EIMS_GPI_SDP1;
2567	switch (adapter->hw.mac.type) {
2568	case ixgbe_mac_82599EB:
2569		mask |= IXGBE_EIMS_GPI_SDP1;
2570		mask |= IXGBE_EIMS_GPI_SDP2;
2571		/* fall through */
2572	case ixgbe_mac_X540:
2573	case ixgbe_mac_X550:
2574	case ixgbe_mac_X550EM_x:
2575		mask |= IXGBE_EIMS_ECC;
2576		mask |= IXGBE_EIMS_MAILBOX;
2577		break;
2578	default:
2579		break;
2580	}
2581
2582	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2583	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2584		mask |= IXGBE_EIMS_FLOW_DIR;
2585
2586	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2587	if (queues)
2588		ixgbe_irq_enable_queues(adapter, ~0);
2589	if (flush)
2590		IXGBE_WRITE_FLUSH(&adapter->hw);
2591}
2592
2593static irqreturn_t ixgbe_msix_other(int irq, void *data)
2594{
2595	struct ixgbe_adapter *adapter = data;
2596	struct ixgbe_hw *hw = &adapter->hw;
2597	u32 eicr;
2598
2599	/*
2600	 * Workaround for Silicon errata.  Use clear-by-write instead
2601	 * of clear-by-read.  Reading with EICS will return the
2602	 * interrupt causes without clearing, which later be done
2603	 * with the write to EICR.
2604	 */
2605	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2606
2607	/* The lower 16bits of the EICR register are for the queue interrupts
2608	 * which should be masked here in order to not accidentally clear them if
2609	 * the bits are high when ixgbe_msix_other is called. There is a race
2610	 * condition otherwise which results in possible performance loss
2611	 * especially if the ixgbe_msix_other interrupt is triggering
2612	 * consistently (as it would when PPS is turned on for the X540 device)
2613	 */
2614	eicr &= 0xFFFF0000;
2615
2616	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2617
2618	if (eicr & IXGBE_EICR_LSC)
2619		ixgbe_check_lsc(adapter);
2620
2621	if (eicr & IXGBE_EICR_MAILBOX)
2622		ixgbe_msg_task(adapter);
2623
2624	switch (hw->mac.type) {
2625	case ixgbe_mac_82599EB:
2626	case ixgbe_mac_X540:
2627	case ixgbe_mac_X550:
2628	case ixgbe_mac_X550EM_x:
2629		if (eicr & IXGBE_EICR_ECC) {
2630			e_info(link, "Received ECC Err, initiating reset\n");
2631			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2632			ixgbe_service_event_schedule(adapter);
2633			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2634		}
2635		/* Handle Flow Director Full threshold interrupt */
2636		if (eicr & IXGBE_EICR_FLOW_DIR) {
2637			int reinit_count = 0;
2638			int i;
2639			for (i = 0; i < adapter->num_tx_queues; i++) {
2640				struct ixgbe_ring *ring = adapter->tx_ring[i];
2641				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2642						       &ring->state))
2643					reinit_count++;
2644			}
2645			if (reinit_count) {
2646				/* no more flow director interrupts until after init */
2647				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2648				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2649				ixgbe_service_event_schedule(adapter);
2650			}
2651		}
2652		ixgbe_check_sfp_event(adapter, eicr);
2653		ixgbe_check_overtemp_event(adapter, eicr);
2654		break;
2655	default:
2656		break;
2657	}
2658
2659	ixgbe_check_fan_failure(adapter, eicr);
2660
2661	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2662		ixgbe_ptp_check_pps_event(adapter, eicr);
2663
2664	/* re-enable the original interrupt state, no lsc, no queues */
2665	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2666		ixgbe_irq_enable(adapter, false, false);
2667
2668	return IRQ_HANDLED;
2669}
2670
2671static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2672{
2673	struct ixgbe_q_vector *q_vector = data;
2674
2675	/* EIAM disabled interrupts (on this vector) for us */
2676
2677	if (q_vector->rx.ring || q_vector->tx.ring)
2678		napi_schedule(&q_vector->napi);
2679
2680	return IRQ_HANDLED;
2681}
2682
2683/**
2684 * ixgbe_poll - NAPI Rx polling callback
2685 * @napi: structure for representing this polling device
2686 * @budget: how many packets driver is allowed to clean
2687 *
2688 * This function is used for legacy and MSI, NAPI mode
2689 **/
2690int ixgbe_poll(struct napi_struct *napi, int budget)
2691{
2692	struct ixgbe_q_vector *q_vector =
2693				container_of(napi, struct ixgbe_q_vector, napi);
2694	struct ixgbe_adapter *adapter = q_vector->adapter;
2695	struct ixgbe_ring *ring;
2696	int per_ring_budget;
2697	bool clean_complete = true;
2698
2699#ifdef CONFIG_IXGBE_DCA
2700	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2701		ixgbe_update_dca(q_vector);
2702#endif
2703
2704	ixgbe_for_each_ring(ring, q_vector->tx)
2705		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2706
2707	if (!ixgbe_qv_lock_napi(q_vector))
2708		return budget;
2709
2710	/* attempt to distribute budget to each queue fairly, but don't allow
2711	 * the budget to go below 1 because we'll exit polling */
2712	if (q_vector->rx.count > 1)
2713		per_ring_budget = max(budget/q_vector->rx.count, 1);
2714	else
2715		per_ring_budget = budget;
2716
2717	ixgbe_for_each_ring(ring, q_vector->rx)
2718		clean_complete &= (ixgbe_clean_rx_irq(q_vector, ring,
2719				   per_ring_budget) < per_ring_budget);
2720
2721	ixgbe_qv_unlock_napi(q_vector);
2722	/* If all work not completed, return budget and keep polling */
2723	if (!clean_complete)
2724		return budget;
2725
2726	/* all work done, exit the polling mode */
2727	napi_complete(napi);
2728	if (adapter->rx_itr_setting & 1)
2729		ixgbe_set_itr(q_vector);
2730	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2731		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2732
2733	return 0;
2734}
2735
2736/**
2737 * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2738 * @adapter: board private structure
2739 *
2740 * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2741 * interrupts from the kernel.
2742 **/
2743static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2744{
2745	struct net_device *netdev = adapter->netdev;
2746	int vector, err;
2747	int ri = 0, ti = 0;
2748
2749	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2750		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2751		struct msix_entry *entry = &adapter->msix_entries[vector];
2752
2753		if (q_vector->tx.ring && q_vector->rx.ring) {
2754			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2755				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2756			ti++;
2757		} else if (q_vector->rx.ring) {
2758			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2759				 "%s-%s-%d", netdev->name, "rx", ri++);
2760		} else if (q_vector->tx.ring) {
2761			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2762				 "%s-%s-%d", netdev->name, "tx", ti++);
2763		} else {
2764			/* skip this unused q_vector */
2765			continue;
2766		}
2767		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2768				  q_vector->name, q_vector);
2769		if (err) {
2770			e_err(probe, "request_irq failed for MSIX interrupt "
2771			      "Error: %d\n", err);
2772			goto free_queue_irqs;
2773		}
2774		/* If Flow Director is enabled, set interrupt affinity */
2775		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2776			/* assign the mask for this irq */
2777			irq_set_affinity_hint(entry->vector,
2778					      &q_vector->affinity_mask);
2779		}
2780	}
2781
2782	err = request_irq(adapter->msix_entries[vector].vector,
2783			  ixgbe_msix_other, 0, netdev->name, adapter);
2784	if (err) {
2785		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2786		goto free_queue_irqs;
2787	}
2788
2789	return 0;
2790
2791free_queue_irqs:
2792	while (vector) {
2793		vector--;
2794		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2795				      NULL);
2796		free_irq(adapter->msix_entries[vector].vector,
2797			 adapter->q_vector[vector]);
2798	}
2799	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2800	pci_disable_msix(adapter->pdev);
2801	kfree(adapter->msix_entries);
2802	adapter->msix_entries = NULL;
2803	return err;
2804}
2805
2806/**
2807 * ixgbe_intr - legacy mode Interrupt Handler
2808 * @irq: interrupt number
2809 * @data: pointer to a network interface device structure
2810 **/
2811static irqreturn_t ixgbe_intr(int irq, void *data)
2812{
2813	struct ixgbe_adapter *adapter = data;
2814	struct ixgbe_hw *hw = &adapter->hw;
2815	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2816	u32 eicr;
2817
2818	/*
2819	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2820	 * before the read of EICR.
2821	 */
2822	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2823
2824	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2825	 * therefore no explicit interrupt disable is necessary */
2826	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2827	if (!eicr) {
2828		/*
2829		 * shared interrupt alert!
2830		 * make sure interrupts are enabled because the read will
2831		 * have disabled interrupts due to EIAM
2832		 * finish the workaround of silicon errata on 82598.  Unmask
2833		 * the interrupt that we masked before the EICR read.
2834		 */
2835		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2836			ixgbe_irq_enable(adapter, true, true);
2837		return IRQ_NONE;	/* Not our interrupt */
2838	}
2839
2840	if (eicr & IXGBE_EICR_LSC)
2841		ixgbe_check_lsc(adapter);
2842
2843	switch (hw->mac.type) {
2844	case ixgbe_mac_82599EB:
2845		ixgbe_check_sfp_event(adapter, eicr);
2846		/* Fall through */
2847	case ixgbe_mac_X540:
2848	case ixgbe_mac_X550:
2849	case ixgbe_mac_X550EM_x:
2850		if (eicr & IXGBE_EICR_ECC) {
2851			e_info(link, "Received ECC Err, initiating reset\n");
2852			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2853			ixgbe_service_event_schedule(adapter);
2854			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2855		}
2856		ixgbe_check_overtemp_event(adapter, eicr);
2857		break;
2858	default:
2859		break;
2860	}
2861
2862	ixgbe_check_fan_failure(adapter, eicr);
2863	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2864		ixgbe_ptp_check_pps_event(adapter, eicr);
2865
2866	/* would disable interrupts here but EIAM disabled it */
2867	napi_schedule(&q_vector->napi);
2868
2869	/*
2870	 * re-enable link(maybe) and non-queue interrupts, no flush.
2871	 * ixgbe_poll will re-enable the queue interrupts
2872	 */
2873	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2874		ixgbe_irq_enable(adapter, false, false);
2875
2876	return IRQ_HANDLED;
2877}
2878
2879/**
2880 * ixgbe_request_irq - initialize interrupts
2881 * @adapter: board private structure
2882 *
2883 * Attempts to configure interrupts using the best available
2884 * capabilities of the hardware and kernel.
2885 **/
2886static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2887{
2888	struct net_device *netdev = adapter->netdev;
2889	int err;
2890
2891	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2892		err = ixgbe_request_msix_irqs(adapter);
2893	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2894		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2895				  netdev->name, adapter);
2896	else
2897		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2898				  netdev->name, adapter);
2899
2900	if (err)
2901		e_err(probe, "request_irq failed, Error %d\n", err);
2902
2903	return err;
2904}
2905
2906static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2907{
2908	int vector;
2909
2910	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2911		free_irq(adapter->pdev->irq, adapter);
2912		return;
2913	}
2914
2915	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2916		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2917		struct msix_entry *entry = &adapter->msix_entries[vector];
2918
2919		/* free only the irqs that were actually requested */
2920		if (!q_vector->rx.ring && !q_vector->tx.ring)
2921			continue;
2922
2923		/* clear the affinity_mask in the IRQ descriptor */
2924		irq_set_affinity_hint(entry->vector, NULL);
2925
2926		free_irq(entry->vector, q_vector);
2927	}
2928
2929	free_irq(adapter->msix_entries[vector++].vector, adapter);
2930}
2931
2932/**
2933 * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2934 * @adapter: board private structure
2935 **/
2936static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
2937{
2938	switch (adapter->hw.mac.type) {
2939	case ixgbe_mac_82598EB:
2940		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
2941		break;
2942	case ixgbe_mac_82599EB:
2943	case ixgbe_mac_X540:
2944	case ixgbe_mac_X550:
2945	case ixgbe_mac_X550EM_x:
2946		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
2947		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
2948		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
2949		break;
2950	default:
2951		break;
2952	}
2953	IXGBE_WRITE_FLUSH(&adapter->hw);
2954	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2955		int vector;
2956
2957		for (vector = 0; vector < adapter->num_q_vectors; vector++)
2958			synchronize_irq(adapter->msix_entries[vector].vector);
2959
2960		synchronize_irq(adapter->msix_entries[vector++].vector);
2961	} else {
2962		synchronize_irq(adapter->pdev->irq);
2963	}
2964}
2965
2966/**
2967 * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2968 *
2969 **/
2970static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
2971{
2972	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2973
2974	ixgbe_write_eitr(q_vector);
2975
2976	ixgbe_set_ivar(adapter, 0, 0, 0);
2977	ixgbe_set_ivar(adapter, 1, 0, 0);
2978
2979	e_info(hw, "Legacy interrupt IVAR setup done\n");
2980}
2981
2982/**
2983 * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2984 * @adapter: board private structure
2985 * @ring: structure containing ring specific data
2986 *
2987 * Configure the Tx descriptor ring after a reset.
2988 **/
2989void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
2990			     struct ixgbe_ring *ring)
2991{
2992	struct ixgbe_hw *hw = &adapter->hw;
2993	u64 tdba = ring->dma;
2994	int wait_loop = 10;
2995	u32 txdctl = IXGBE_TXDCTL_ENABLE;
2996	u8 reg_idx = ring->reg_idx;
2997
2998	/* disable queue to avoid issues while updating state */
2999	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3000	IXGBE_WRITE_FLUSH(hw);
3001
3002	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3003			(tdba & DMA_BIT_MASK(32)));
3004	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3005	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3006			ring->count * sizeof(union ixgbe_adv_tx_desc));
3007	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3008	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3009	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3010
3011	/*
3012	 * set WTHRESH to encourage burst writeback, it should not be set
3013	 * higher than 1 when:
3014	 * - ITR is 0 as it could cause false TX hangs
3015	 * - ITR is set to > 100k int/sec and BQL is enabled
3016	 *
3017	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3018	 * to or less than the number of on chip descriptors, which is
3019	 * currently 40.
3020	 */
3021	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3022		txdctl |= (1 << 16);	/* WTHRESH = 1 */
3023	else
3024		txdctl |= (8 << 16);	/* WTHRESH = 8 */
3025
3026	/*
3027	 * Setting PTHRESH to 32 both improves performance
3028	 * and avoids a TX hang with DFP enabled
3029	 */
3030	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
3031		   32;		/* PTHRESH = 32 */
3032
3033	/* reinitialize flowdirector state */
3034	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3035		ring->atr_sample_rate = adapter->atr_sample_rate;
3036		ring->atr_count = 0;
3037		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3038	} else {
3039		ring->atr_sample_rate = 0;
3040	}
3041
3042	/* initialize XPS */
3043	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3044		struct ixgbe_q_vector *q_vector = ring->q_vector;
3045
3046		if (q_vector)
3047			netif_set_xps_queue(ring->netdev,
3048					    &q_vector->affinity_mask,
3049					    ring->queue_index);
3050	}
3051
3052	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3053
3054	/* enable queue */
3055	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3056
3057	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3058	if (hw->mac.type == ixgbe_mac_82598EB &&
3059	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3060		return;
3061
3062	/* poll to verify queue is enabled */
3063	do {
3064		usleep_range(1000, 2000);
3065		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3066	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3067	if (!wait_loop)
3068		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
3069}
3070
3071static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3072{
3073	struct ixgbe_hw *hw = &adapter->hw;
3074	u32 rttdcs, mtqc;
3075	u8 tcs = netdev_get_num_tc(adapter->netdev);
3076
3077	if (hw->mac.type == ixgbe_mac_82598EB)
3078		return;
3079
3080	/* disable the arbiter while setting MTQC */
3081	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3082	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3083	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3084
3085	/* set transmit pool layout */
3086	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3087		mtqc = IXGBE_MTQC_VT_ENA;
3088		if (tcs > 4)
3089			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3090		else if (tcs > 1)
3091			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3092		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3093			mtqc |= IXGBE_MTQC_32VF;
3094		else
3095			mtqc |= IXGBE_MTQC_64VF;
3096	} else {
3097		if (tcs > 4)
3098			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3099		else if (tcs > 1)
3100			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3101		else
3102			mtqc = IXGBE_MTQC_64Q_1PB;
3103	}
3104
3105	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3106
3107	/* Enable Security TX Buffer IFG for multiple pb */
3108	if (tcs) {
3109		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3110		sectx |= IXGBE_SECTX_DCB;
3111		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3112	}
3113
3114	/* re-enable the arbiter */
3115	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3116	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3117}
3118
3119/**
3120 * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3121 * @adapter: board private structure
3122 *
3123 * Configure the Tx unit of the MAC after a reset.
3124 **/
3125static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3126{
3127	struct ixgbe_hw *hw = &adapter->hw;
3128	u32 dmatxctl;
3129	u32 i;
3130
3131	ixgbe_setup_mtqc(adapter);
3132
3133	if (hw->mac.type != ixgbe_mac_82598EB) {
3134		/* DMATXCTL.EN must be before Tx queues are enabled */
3135		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3136		dmatxctl |= IXGBE_DMATXCTL_TE;
3137		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3138	}
3139
3140	/* Setup the HW Tx Head and Tail descriptor pointers */
3141	for (i = 0; i < adapter->num_tx_queues; i++)
3142		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3143}
3144
3145static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3146				 struct ixgbe_ring *ring)
3147{
3148	struct ixgbe_hw *hw = &adapter->hw;
3149	u8 reg_idx = ring->reg_idx;
3150	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3151
3152	srrctl |= IXGBE_SRRCTL_DROP_EN;
3153
3154	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3155}
3156
3157static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3158				  struct ixgbe_ring *ring)
3159{
3160	struct ixgbe_hw *hw = &adapter->hw;
3161	u8 reg_idx = ring->reg_idx;
3162	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3163
3164	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3165
3166	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3167}
3168
3169#ifdef CONFIG_IXGBE_DCB
3170void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3171#else
3172static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3173#endif
3174{
3175	int i;
3176	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3177
3178	if (adapter->ixgbe_ieee_pfc)
3179		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3180
3181	/*
3182	 * We should set the drop enable bit if:
3183	 *  SR-IOV is enabled
3184	 *   or
3185	 *  Number of Rx queues > 1 and flow control is disabled
3186	 *
3187	 *  This allows us to avoid head of line blocking for security
3188	 *  and performance reasons.
3189	 */
3190	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3191	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3192		for (i = 0; i < adapter->num_rx_queues; i++)
3193			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3194	} else {
3195		for (i = 0; i < adapter->num_rx_queues; i++)
3196			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3197	}
3198}
3199
3200#define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3201
3202static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3203				   struct ixgbe_ring *rx_ring)
3204{
3205	struct ixgbe_hw *hw = &adapter->hw;
3206	u32 srrctl;
3207	u8 reg_idx = rx_ring->reg_idx;
3208
3209	if (hw->mac.type == ixgbe_mac_82598EB) {
3210		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3211
3212		/*
3213		 * if VMDq is not active we must program one srrctl register
3214		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3215		 */
3216		reg_idx &= mask;
3217	}
3218
3219	/* configure header buffer length, needed for RSC */
3220	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3221
3222	/* configure the packet buffer length */
3223	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3224
3225	/* configure descriptor type */
3226	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3227
3228	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3229}
3230
3231/**
3232 * Return a number of entries in the RSS indirection table
3233 *
3234 * @adapter: device handle
3235 *
3236 *  - 82598/82599/X540:     128
3237 *  - X550(non-SRIOV mode): 512
3238 *  - X550(SRIOV mode):     64
3239 */
3240u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3241{
3242	if (adapter->hw.mac.type < ixgbe_mac_X550)
3243		return 128;
3244	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3245		return 64;
3246	else
3247		return 512;
3248}
3249
3250/**
3251 * Write the RETA table to HW
3252 *
3253 * @adapter: device handle
3254 *
3255 * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3256 */
3257static void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3258{
3259	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3260	struct ixgbe_hw *hw = &adapter->hw;
3261	u32 reta = 0;
3262	u32 indices_multi;
3263	u8 *indir_tbl = adapter->rss_indir_tbl;
3264
3265	/* Fill out the redirection table as follows:
3266	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3267	 *    indices.
3268	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3269	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3270	 */
3271	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3272		indices_multi = 0x11;
3273	else
3274		indices_multi = 0x1;
3275
3276	/* Write redirection table to HW */
3277	for (i = 0; i < reta_entries; i++) {
3278		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3279		if ((i & 3) == 3) {
3280			if (i < 128)
3281				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3282			else
3283				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3284						reta);
3285			reta = 0;
3286		}
3287	}
3288}
3289
3290/**
3291 * Write the RETA table to HW (for x550 devices in SRIOV mode)
3292 *
3293 * @adapter: device handle
3294 *
3295 * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3296 */
3297static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3298{
3299	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3300	struct ixgbe_hw *hw = &adapter->hw;
3301	u32 vfreta = 0;
3302	unsigned int pf_pool = adapter->num_vfs;
3303
3304	/* Write redirection table to HW */
3305	for (i = 0; i < reta_entries; i++) {
3306		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3307		if ((i & 3) == 3) {
3308			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3309					vfreta);
3310			vfreta = 0;
3311		}
3312	}
3313}
3314
3315static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3316{
3317	struct ixgbe_hw *hw = &adapter->hw;
3318	u32 i, j;
3319	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3320	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3321
3322	/* Program table for at least 2 queues w/ SR-IOV so that VFs can
3323	 * make full use of any rings they may have.  We will use the
3324	 * PSRTYPE register to control how many rings we use within the PF.
3325	 */
3326	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3327		rss_i = 2;
3328
3329	/* Fill out hash function seeds */
3330	for (i = 0; i < 10; i++)
3331		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3332
3333	/* Fill out redirection table */
3334	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3335
3336	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3337		if (j == rss_i)
3338			j = 0;
3339
3340		adapter->rss_indir_tbl[i] = j;
3341	}
3342
3343	ixgbe_store_reta(adapter);
3344}
3345
3346static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3347{
3348	struct ixgbe_hw *hw = &adapter->hw;
3349	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3350	unsigned int pf_pool = adapter->num_vfs;
3351	int i, j;
3352
3353	/* Fill out hash function seeds */
3354	for (i = 0; i < 10; i++)
3355		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool),
3356				adapter->rss_key[i]);
3357
3358	/* Fill out the redirection table */
3359	for (i = 0, j = 0; i < 64; i++, j++) {
3360		if (j == rss_i)
3361			j = 0;
3362
3363		adapter->rss_indir_tbl[i] = j;
3364	}
3365
3366	ixgbe_store_vfreta(adapter);
3367}
3368
3369static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3370{
3371	struct ixgbe_hw *hw = &adapter->hw;
3372	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3373	u32 rxcsum;
3374
3375	/* Disable indicating checksum in descriptor, enables RSS hash */
3376	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3377	rxcsum |= IXGBE_RXCSUM_PCSD;
3378	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3379
3380	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3381		if (adapter->ring_feature[RING_F_RSS].mask)
3382			mrqc = IXGBE_MRQC_RSSEN;
3383	} else {
3384		u8 tcs = netdev_get_num_tc(adapter->netdev);
3385
3386		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3387			if (tcs > 4)
3388				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3389			else if (tcs > 1)
3390				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3391			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3392				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3393			else
3394				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3395		} else {
3396			if (tcs > 4)
3397				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3398			else if (tcs > 1)
3399				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3400			else
3401				mrqc = IXGBE_MRQC_RSSEN;
3402		}
3403	}
3404
3405	/* Perform hash on these packet types */
3406	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3407		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3408		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3409		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3410
3411	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3412		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3413	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3414		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3415
3416	netdev_rss_key_fill(adapter->rss_key, sizeof(adapter->rss_key));
3417	if ((hw->mac.type >= ixgbe_mac_X550) &&
3418	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3419		unsigned int pf_pool = adapter->num_vfs;
3420
3421		/* Enable VF RSS mode */
3422		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3423		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3424
3425		/* Setup RSS through the VF registers */
3426		ixgbe_setup_vfreta(adapter);
3427		vfmrqc = IXGBE_MRQC_RSSEN;
3428		vfmrqc |= rss_field;
3429		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3430	} else {
3431		ixgbe_setup_reta(adapter);
3432		mrqc |= rss_field;
3433		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3434	}
3435}
3436
3437/**
3438 * ixgbe_configure_rscctl - enable RSC for the indicated ring
3439 * @adapter:    address of board private structure
3440 * @index:      index of ring to set
3441 **/
3442static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3443				   struct ixgbe_ring *ring)
3444{
3445	struct ixgbe_hw *hw = &adapter->hw;
3446	u32 rscctrl;
3447	u8 reg_idx = ring->reg_idx;
3448
3449	if (!ring_is_rsc_enabled(ring))
3450		return;
3451
3452	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3453	rscctrl |= IXGBE_RSCCTL_RSCEN;
3454	/*
3455	 * we must limit the number of descriptors so that the
3456	 * total size of max desc * buf_len is not greater
3457	 * than 65536
3458	 */
3459	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3460	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3461}
3462
3463#define IXGBE_MAX_RX_DESC_POLL 10
3464static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3465				       struct ixgbe_ring *ring)
3466{
3467	struct ixgbe_hw *hw = &adapter->hw;
3468	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3469	u32 rxdctl;
3470	u8 reg_idx = ring->reg_idx;
3471
3472	if (ixgbe_removed(hw->hw_addr))
3473		return;
3474	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3475	if (hw->mac.type == ixgbe_mac_82598EB &&
3476	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3477		return;
3478
3479	do {
3480		usleep_range(1000, 2000);
3481		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3482	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3483
3484	if (!wait_loop) {
3485		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3486		      "the polling period\n", reg_idx);
3487	}
3488}
3489
3490void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3491			    struct ixgbe_ring *ring)
3492{
3493	struct ixgbe_hw *hw = &adapter->hw;
3494	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3495	u32 rxdctl;
3496	u8 reg_idx = ring->reg_idx;
3497
3498	if (ixgbe_removed(hw->hw_addr))
3499		return;
3500	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3501	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3502
3503	/* write value back with RXDCTL.ENABLE bit cleared */
3504	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3505
3506	if (hw->mac.type == ixgbe_mac_82598EB &&
3507	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3508		return;
3509
3510	/* the hardware may take up to 100us to really disable the rx queue */
3511	do {
3512		udelay(10);
3513		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3514	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3515
3516	if (!wait_loop) {
3517		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3518		      "the polling period\n", reg_idx);
3519	}
3520}
3521
3522void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3523			     struct ixgbe_ring *ring)
3524{
3525	struct ixgbe_hw *hw = &adapter->hw;
3526	u64 rdba = ring->dma;
3527	u32 rxdctl;
3528	u8 reg_idx = ring->reg_idx;
3529
3530	/* disable queue to avoid issues while updating state */
3531	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3532	ixgbe_disable_rx_queue(adapter, ring);
3533
3534	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3535	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3536	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3537			ring->count * sizeof(union ixgbe_adv_rx_desc));
3538	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3539	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3540	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3541
3542	ixgbe_configure_srrctl(adapter, ring);
3543	ixgbe_configure_rscctl(adapter, ring);
3544
3545	if (hw->mac.type == ixgbe_mac_82598EB) {
3546		/*
3547		 * enable cache line friendly hardware writes:
3548		 * PTHRESH=32 descriptors (half the internal cache),
3549		 * this also removes ugly rx_no_buffer_count increment
3550		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3551		 * WTHRESH=8 burst writeback up to two cache lines
3552		 */
3553		rxdctl &= ~0x3FFFFF;
3554		rxdctl |=  0x080420;
3555	}
3556
3557	/* enable receive descriptor ring */
3558	rxdctl |= IXGBE_RXDCTL_ENABLE;
3559	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3560
3561	ixgbe_rx_desc_queue_enable(adapter, ring);
3562	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3563}
3564
3565static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3566{
3567	struct ixgbe_hw *hw = &adapter->hw;
3568	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3569	u16 pool;
3570
3571	/* PSRTYPE must be initialized in non 82598 adapters */
3572	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3573		      IXGBE_PSRTYPE_UDPHDR |
3574		      IXGBE_PSRTYPE_IPV4HDR |
3575		      IXGBE_PSRTYPE_L2HDR |
3576		      IXGBE_PSRTYPE_IPV6HDR;
3577
3578	if (hw->mac.type == ixgbe_mac_82598EB)
3579		return;
3580
3581	if (rss_i > 3)
3582		psrtype |= 2 << 29;
3583	else if (rss_i > 1)
3584		psrtype |= 1 << 29;
3585
3586	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3587		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3588}
3589
3590static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3591{
3592	struct ixgbe_hw *hw = &adapter->hw;
3593	u32 reg_offset, vf_shift;
3594	u32 gcr_ext, vmdctl;
3595	int i;
3596
3597	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3598		return;
3599
3600	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3601	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3602	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3603	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3604	vmdctl |= IXGBE_VT_CTL_REPLEN;
3605	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3606
3607	vf_shift = VMDQ_P(0) % 32;
3608	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3609
3610	/* Enable only the PF's pool for Tx/Rx */
3611	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (~0) << vf_shift);
3612	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3613	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (~0) << vf_shift);
3614	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3615	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
3616		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3617
3618	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3619	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3620
3621	/*
3622	 * Set up VF register offsets for selected VT Mode,
3623	 * i.e. 32 or 64 VFs for SR-IOV
3624	 */
3625	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3626	case IXGBE_82599_VMDQ_8Q_MASK:
3627		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3628		break;
3629	case IXGBE_82599_VMDQ_4Q_MASK:
3630		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3631		break;
3632	default:
3633		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3634		break;
3635	}
3636
3637	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3638
3639
3640	/* Enable MAC Anti-Spoofing */
3641	hw->mac.ops.set_mac_anti_spoofing(hw, (adapter->num_vfs != 0),
3642					  adapter->num_vfs);
3643
3644	/* Ensure LLDP is set for Ethertype Antispoofing if we will be
3645	 * calling set_ethertype_anti_spoofing for each VF in loop below
3646	 */
3647	if (hw->mac.ops.set_ethertype_anti_spoofing)
3648		IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_LLDP),
3649				(IXGBE_ETQF_FILTER_EN    | /* enable filter */
3650				 IXGBE_ETQF_TX_ANTISPOOF | /* tx antispoof */
3651				 IXGBE_ETH_P_LLDP));	   /* LLDP eth type */
3652
3653	/* For VFs that have spoof checking turned off */
3654	for (i = 0; i < adapter->num_vfs; i++) {
3655		if (!adapter->vfinfo[i].spoofchk_enabled)
3656			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3657
3658		/* enable ethertype anti spoofing if hw supports it */
3659		if (hw->mac.ops.set_ethertype_anti_spoofing)
3660			hw->mac.ops.set_ethertype_anti_spoofing(hw, true, i);
3661
3662		/* Enable/Disable RSS query feature  */
3663		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
3664					  adapter->vfinfo[i].rss_query_enabled);
3665	}
3666}
3667
3668static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3669{
3670	struct ixgbe_hw *hw = &adapter->hw;
3671	struct net_device *netdev = adapter->netdev;
3672	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3673	struct ixgbe_ring *rx_ring;
3674	int i;
3675	u32 mhadd, hlreg0;
3676
3677#ifdef IXGBE_FCOE
3678	/* adjust max frame to be able to do baby jumbo for FCoE */
3679	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3680	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3681		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3682
3683#endif /* IXGBE_FCOE */
3684
3685	/* adjust max frame to be at least the size of a standard frame */
3686	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3687		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3688
3689	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3690	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3691		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3692		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3693
3694		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3695	}
3696
3697	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3698	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3699	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3700	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3701
3702	/*
3703	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3704	 * the Base and Length of the Rx Descriptor Ring
3705	 */
3706	for (i = 0; i < adapter->num_rx_queues; i++) {
3707		rx_ring = adapter->rx_ring[i];
3708		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3709			set_ring_rsc_enabled(rx_ring);
3710		else
3711			clear_ring_rsc_enabled(rx_ring);
3712	}
3713}
3714
3715static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3716{
3717	struct ixgbe_hw *hw = &adapter->hw;
3718	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3719
3720	switch (hw->mac.type) {
3721	case ixgbe_mac_X550:
3722	case ixgbe_mac_X550EM_x:
3723	case ixgbe_mac_82598EB:
3724		/*
3725		 * For VMDq support of different descriptor types or
3726		 * buffer sizes through the use of multiple SRRCTL
3727		 * registers, RDRXCTL.MVMEN must be set to 1
3728		 *
3729		 * also, the manual doesn't mention it clearly but DCA hints
3730		 * will only use queue 0's tags unless this bit is set.  Side
3731		 * effects of setting this bit are only that SRRCTL must be
3732		 * fully programmed [0..15]
3733		 */
3734		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3735		break;
3736	case ixgbe_mac_82599EB:
3737	case ixgbe_mac_X540:
3738		/* Disable RSC for ACK packets */
3739		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3740		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3741		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3742		/* hardware requires some bits to be set by default */
3743		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3744		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3745		break;
3746	default:
3747		/* We should do nothing since we don't know this hardware */
3748		return;
3749	}
3750
3751	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3752}
3753
3754/**
3755 * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3756 * @adapter: board private structure
3757 *
3758 * Configure the Rx unit of the MAC after a reset.
3759 **/
3760static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3761{
3762	struct ixgbe_hw *hw = &adapter->hw;
3763	int i;
3764	u32 rxctrl, rfctl;
3765
3766	/* disable receives while setting up the descriptors */
3767	hw->mac.ops.disable_rx(hw);
3768
3769	ixgbe_setup_psrtype(adapter);
3770	ixgbe_setup_rdrxctl(adapter);
3771
3772	/* RSC Setup */
3773	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
3774	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
3775	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
3776		rfctl |= IXGBE_RFCTL_RSC_DIS;
3777	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3778
3779	/* Program registers for the distribution of queues */
3780	ixgbe_setup_mrqc(adapter);
3781
3782	/* set_rx_buffer_len must be called before ring initialization */
3783	ixgbe_set_rx_buffer_len(adapter);
3784
3785	/*
3786	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3787	 * the Base and Length of the Rx Descriptor Ring
3788	 */
3789	for (i = 0; i < adapter->num_rx_queues; i++)
3790		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3791
3792	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3793	/* disable drop enable for 82598 parts */
3794	if (hw->mac.type == ixgbe_mac_82598EB)
3795		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3796
3797	/* enable all receives */
3798	rxctrl |= IXGBE_RXCTRL_RXEN;
3799	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3800}
3801
3802static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3803				 __be16 proto, u16 vid)
3804{
3805	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3806	struct ixgbe_hw *hw = &adapter->hw;
3807
3808	/* add VID to filter table */
3809	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true);
3810	set_bit(vid, adapter->active_vlans);
3811
3812	return 0;
3813}
3814
3815static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
3816				  __be16 proto, u16 vid)
3817{
3818	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3819	struct ixgbe_hw *hw = &adapter->hw;
3820
3821	/* remove VID from filter table */
3822	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), false);
3823	clear_bit(vid, adapter->active_vlans);
3824
3825	return 0;
3826}
3827
3828/**
3829 * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3830 * @adapter: driver data
3831 */
3832static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3833{
3834	struct ixgbe_hw *hw = &adapter->hw;
3835	u32 vlnctrl;
3836	int i, j;
3837
3838	switch (hw->mac.type) {
3839	case ixgbe_mac_82598EB:
3840		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3841		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3842		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3843		break;
3844	case ixgbe_mac_82599EB:
3845	case ixgbe_mac_X540:
3846	case ixgbe_mac_X550:
3847	case ixgbe_mac_X550EM_x:
3848		for (i = 0; i < adapter->num_rx_queues; i++) {
3849			struct ixgbe_ring *ring = adapter->rx_ring[i];
3850
3851			if (ring->l2_accel_priv)
3852				continue;
3853			j = ring->reg_idx;
3854			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3855			vlnctrl &= ~IXGBE_RXDCTL_VME;
3856			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3857		}
3858		break;
3859	default:
3860		break;
3861	}
3862}
3863
3864/**
3865 * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3866 * @adapter: driver data
3867 */
3868static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3869{
3870	struct ixgbe_hw *hw = &adapter->hw;
3871	u32 vlnctrl;
3872	int i, j;
3873
3874	switch (hw->mac.type) {
3875	case ixgbe_mac_82598EB:
3876		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3877		vlnctrl |= IXGBE_VLNCTRL_VME;
3878		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3879		break;
3880	case ixgbe_mac_82599EB:
3881	case ixgbe_mac_X540:
3882	case ixgbe_mac_X550:
3883	case ixgbe_mac_X550EM_x:
3884		for (i = 0; i < adapter->num_rx_queues; i++) {
3885			struct ixgbe_ring *ring = adapter->rx_ring[i];
3886
3887			if (ring->l2_accel_priv)
3888				continue;
3889			j = ring->reg_idx;
3890			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3891			vlnctrl |= IXGBE_RXDCTL_VME;
3892			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3893		}
3894		break;
3895	default:
3896		break;
3897	}
3898}
3899
3900static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3901{
3902	u16 vid;
3903
3904	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
3905
3906	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3907		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
3908}
3909
3910/**
3911 * ixgbe_write_mc_addr_list - write multicast addresses to MTA
3912 * @netdev: network interface device structure
3913 *
3914 * Writes multicast address list to the MTA hash table.
3915 * Returns: -ENOMEM on failure
3916 *                0 on no addresses written
3917 *                X on writing X addresses to MTA
3918 **/
3919static int ixgbe_write_mc_addr_list(struct net_device *netdev)
3920{
3921	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3922	struct ixgbe_hw *hw = &adapter->hw;
3923
3924	if (!netif_running(netdev))
3925		return 0;
3926
3927	if (hw->mac.ops.update_mc_addr_list)
3928		hw->mac.ops.update_mc_addr_list(hw, netdev);
3929	else
3930		return -ENOMEM;
3931
3932#ifdef CONFIG_PCI_IOV
3933	ixgbe_restore_vf_multicasts(adapter);
3934#endif
3935
3936	return netdev_mc_count(netdev);
3937}
3938
3939#ifdef CONFIG_PCI_IOV
3940void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
3941{
3942	struct ixgbe_hw *hw = &adapter->hw;
3943	int i;
3944	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3945		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
3946			hw->mac.ops.set_rar(hw, i, adapter->mac_table[i].addr,
3947					    adapter->mac_table[i].queue,
3948					    IXGBE_RAH_AV);
3949		else
3950			hw->mac.ops.clear_rar(hw, i);
3951
3952		adapter->mac_table[i].state &= ~(IXGBE_MAC_STATE_MODIFIED);
3953	}
3954}
3955#endif
3956
3957static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
3958{
3959	struct ixgbe_hw *hw = &adapter->hw;
3960	int i;
3961	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3962		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_MODIFIED) {
3963			if (adapter->mac_table[i].state &
3964			    IXGBE_MAC_STATE_IN_USE)
3965				hw->mac.ops.set_rar(hw, i,
3966						adapter->mac_table[i].addr,
3967						adapter->mac_table[i].queue,
3968						IXGBE_RAH_AV);
3969			else
3970				hw->mac.ops.clear_rar(hw, i);
3971
3972			adapter->mac_table[i].state &=
3973						~(IXGBE_MAC_STATE_MODIFIED);
3974		}
3975	}
3976}
3977
3978static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
3979{
3980	int i;
3981	struct ixgbe_hw *hw = &adapter->hw;
3982
3983	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3984		adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
3985		adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
3986		eth_zero_addr(adapter->mac_table[i].addr);
3987		adapter->mac_table[i].queue = 0;
3988	}
3989	ixgbe_sync_mac_table(adapter);
3990}
3991
3992static int ixgbe_available_rars(struct ixgbe_adapter *adapter)
3993{
3994	struct ixgbe_hw *hw = &adapter->hw;
3995	int i, count = 0;
3996
3997	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3998		if (adapter->mac_table[i].state == 0)
3999			count++;
4000	}
4001	return count;
4002}
4003
4004/* this function destroys the first RAR entry */
4005static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter,
4006					 u8 *addr)
4007{
4008	struct ixgbe_hw *hw = &adapter->hw;
4009
4010	memcpy(&adapter->mac_table[0].addr, addr, ETH_ALEN);
4011	adapter->mac_table[0].queue = VMDQ_P(0);
4012	adapter->mac_table[0].state = (IXGBE_MAC_STATE_DEFAULT |
4013				       IXGBE_MAC_STATE_IN_USE);
4014	hw->mac.ops.set_rar(hw, 0, adapter->mac_table[0].addr,
4015			    adapter->mac_table[0].queue,
4016			    IXGBE_RAH_AV);
4017}
4018
4019int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
4020{
4021	struct ixgbe_hw *hw = &adapter->hw;
4022	int i;
4023
4024	if (is_zero_ether_addr(addr))
4025		return -EINVAL;
4026
4027	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4028		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
4029			continue;
4030		adapter->mac_table[i].state |= (IXGBE_MAC_STATE_MODIFIED |
4031						IXGBE_MAC_STATE_IN_USE);
4032		ether_addr_copy(adapter->mac_table[i].addr, addr);
4033		adapter->mac_table[i].queue = queue;
4034		ixgbe_sync_mac_table(adapter);
4035		return i;
4036	}
4037	return -ENOMEM;
4038}
4039
4040int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
4041{
4042	/* search table for addr, if found, set to 0 and sync */
4043	int i;
4044	struct ixgbe_hw *hw = &adapter->hw;
4045
4046	if (is_zero_ether_addr(addr))
4047		return -EINVAL;
4048
4049	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4050		if (ether_addr_equal(addr, adapter->mac_table[i].addr) &&
4051		    adapter->mac_table[i].queue == queue) {
4052			adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
4053			adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
4054			eth_zero_addr(adapter->mac_table[i].addr);
4055			adapter->mac_table[i].queue = 0;
4056			ixgbe_sync_mac_table(adapter);
4057			return 0;
4058		}
4059	}
4060	return -ENOMEM;
4061}
4062/**
4063 * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4064 * @netdev: network interface device structure
4065 *
4066 * Writes unicast address list to the RAR table.
4067 * Returns: -ENOMEM on failure/insufficient address space
4068 *                0 on no addresses written
4069 *                X on writing X addresses to the RAR table
4070 **/
4071static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4072{
4073	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4074	int count = 0;
4075
4076	/* return ENOMEM indicating insufficient memory for addresses */
4077	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter))
4078		return -ENOMEM;
4079
4080	if (!netdev_uc_empty(netdev)) {
4081		struct netdev_hw_addr *ha;
4082		netdev_for_each_uc_addr(ha, netdev) {
4083			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4084			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4085			count++;
4086		}
4087	}
4088	return count;
4089}
4090
4091/**
4092 * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4093 * @netdev: network interface device structure
4094 *
4095 * The set_rx_method entry point is called whenever the unicast/multicast
4096 * address list or the network interface flags are updated.  This routine is
4097 * responsible for configuring the hardware for proper unicast, multicast and
4098 * promiscuous mode.
4099 **/
4100void ixgbe_set_rx_mode(struct net_device *netdev)
4101{
4102	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4103	struct ixgbe_hw *hw = &adapter->hw;
4104	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4105	u32 vlnctrl;
4106	int count;
4107
4108	/* Check for Promiscuous and All Multicast modes */
4109	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4110	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4111
4112	/* set all bits that we expect to always be set */
4113	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4114	fctrl |= IXGBE_FCTRL_BAM;
4115	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4116	fctrl |= IXGBE_FCTRL_PMCF;
4117
4118	/* clear the bits we are changing the status of */
4119	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4120	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4121	if (netdev->flags & IFF_PROMISC) {
4122		hw->addr_ctrl.user_set_promisc = true;
4123		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4124		vmolr |= IXGBE_VMOLR_MPE;
4125		/* Only disable hardware filter vlans in promiscuous mode
4126		 * if SR-IOV and VMDQ are disabled - otherwise ensure
4127		 * that hardware VLAN filters remain enabled.
4128		 */
4129		if (adapter->flags & (IXGBE_FLAG_VMDQ_ENABLED |
4130				      IXGBE_FLAG_SRIOV_ENABLED))
4131			vlnctrl |= (IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4132	} else {
4133		if (netdev->flags & IFF_ALLMULTI) {
4134			fctrl |= IXGBE_FCTRL_MPE;
4135			vmolr |= IXGBE_VMOLR_MPE;
4136		}
4137		vlnctrl |= IXGBE_VLNCTRL_VFE;
4138		hw->addr_ctrl.user_set_promisc = false;
4139	}
4140
4141	/*
4142	 * Write addresses to available RAR registers, if there is not
4143	 * sufficient space to store all the addresses then enable
4144	 * unicast promiscuous mode
4145	 */
4146	count = ixgbe_write_uc_addr_list(netdev, VMDQ_P(0));
4147	if (count < 0) {
4148		fctrl |= IXGBE_FCTRL_UPE;
4149		vmolr |= IXGBE_VMOLR_ROPE;
4150	}
4151
4152	/* Write addresses to the MTA, if the attempt fails
4153	 * then we should just turn on promiscuous mode so
4154	 * that we can at least receive multicast traffic
4155	 */
4156	count = ixgbe_write_mc_addr_list(netdev);
4157	if (count < 0) {
4158		fctrl |= IXGBE_FCTRL_MPE;
4159		vmolr |= IXGBE_VMOLR_MPE;
4160	} else if (count) {
4161		vmolr |= IXGBE_VMOLR_ROMPE;
4162	}
4163
4164	if (hw->mac.type != ixgbe_mac_82598EB) {
4165		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4166			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4167			   IXGBE_VMOLR_ROPE);
4168		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4169	}
4170
4171	/* This is useful for sniffing bad packets. */
4172	if (adapter->netdev->features & NETIF_F_RXALL) {
4173		/* UPE and MPE will be handled by normal PROMISC logic
4174		 * in e1000e_set_rx_mode */
4175		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4176			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4177			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4178
4179		fctrl &= ~(IXGBE_FCTRL_DPF);
4180		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4181	}
4182
4183	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4184	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4185
4186	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
4187		ixgbe_vlan_strip_enable(adapter);
4188	else
4189		ixgbe_vlan_strip_disable(adapter);
4190}
4191
4192static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4193{
4194	int q_idx;
4195
4196	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4197		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
4198		napi_enable(&adapter->q_vector[q_idx]->napi);
4199	}
4200}
4201
4202static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4203{
4204	int q_idx;
4205
4206	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4207		napi_disable(&adapter->q_vector[q_idx]->napi);
4208		while (!ixgbe_qv_disable(adapter->q_vector[q_idx])) {
4209			pr_info("QV %d locked\n", q_idx);
4210			usleep_range(1000, 20000);
4211		}
4212	}
4213}
4214
4215#ifdef CONFIG_IXGBE_DCB
4216/**
4217 * ixgbe_configure_dcb - Configure DCB hardware
4218 * @adapter: ixgbe adapter struct
4219 *
4220 * This is called by the driver on open to configure the DCB hardware.
4221 * This is also called by the gennetlink interface when reconfiguring
4222 * the DCB state.
4223 */
4224static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4225{
4226	struct ixgbe_hw *hw = &adapter->hw;
4227	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4228
4229	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4230		if (hw->mac.type == ixgbe_mac_82598EB)
4231			netif_set_gso_max_size(adapter->netdev, 65536);
4232		return;
4233	}
4234
4235	if (hw->mac.type == ixgbe_mac_82598EB)
4236		netif_set_gso_max_size(adapter->netdev, 32768);
4237
4238#ifdef IXGBE_FCOE
4239	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4240		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4241#endif
4242
4243	/* reconfigure the hardware */
4244	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4245		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4246						DCB_TX_CONFIG);
4247		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4248						DCB_RX_CONFIG);
4249		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4250	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4251		ixgbe_dcb_hw_ets(&adapter->hw,
4252				 adapter->ixgbe_ieee_ets,
4253				 max_frame);
4254		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4255					adapter->ixgbe_ieee_pfc->pfc_en,
4256					adapter->ixgbe_ieee_ets->prio_tc);
4257	}
4258
4259	/* Enable RSS Hash per TC */
4260	if (hw->mac.type != ixgbe_mac_82598EB) {
4261		u32 msb = 0;
4262		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4263
4264		while (rss_i) {
4265			msb++;
4266			rss_i >>= 1;
4267		}
4268
4269		/* write msb to all 8 TCs in one write */
4270		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4271	}
4272}
4273#endif
4274
4275/* Additional bittime to account for IXGBE framing */
4276#define IXGBE_ETH_FRAMING 20
4277
4278/**
4279 * ixgbe_hpbthresh - calculate high water mark for flow control
4280 *
4281 * @adapter: board private structure to calculate for
4282 * @pb: packet buffer to calculate
4283 */
4284static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4285{
4286	struct ixgbe_hw *hw = &adapter->hw;
4287	struct net_device *dev = adapter->netdev;
4288	int link, tc, kb, marker;
4289	u32 dv_id, rx_pba;
4290
4291	/* Calculate max LAN frame size */
4292	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4293
4294#ifdef IXGBE_FCOE
4295	/* FCoE traffic class uses FCOE jumbo frames */
4296	if ((dev->features & NETIF_F_FCOE_MTU) &&
4297	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4298	    (pb == ixgbe_fcoe_get_tc(adapter)))
4299		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4300#endif
4301
4302	/* Calculate delay value for device */
4303	switch (hw->mac.type) {
4304	case ixgbe_mac_X540:
4305	case ixgbe_mac_X550:
4306	case ixgbe_mac_X550EM_x:
4307		dv_id = IXGBE_DV_X540(link, tc);
4308		break;
4309	default:
4310		dv_id = IXGBE_DV(link, tc);
4311		break;
4312	}
4313
4314	/* Loopback switch introduces additional latency */
4315	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4316		dv_id += IXGBE_B2BT(tc);
4317
4318	/* Delay value is calculated in bit times convert to KB */
4319	kb = IXGBE_BT2KB(dv_id);
4320	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4321
4322	marker = rx_pba - kb;
4323
4324	/* It is possible that the packet buffer is not large enough
4325	 * to provide required headroom. In this case throw an error
4326	 * to user and a do the best we can.
4327	 */
4328	if (marker < 0) {
4329		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4330			    "headroom to support flow control."
4331			    "Decrease MTU or number of traffic classes\n", pb);
4332		marker = tc + 1;
4333	}
4334
4335	return marker;
4336}
4337
4338/**
4339 * ixgbe_lpbthresh - calculate low water mark for for flow control
4340 *
4341 * @adapter: board private structure to calculate for
4342 * @pb: packet buffer to calculate
4343 */
4344static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4345{
4346	struct ixgbe_hw *hw = &adapter->hw;
4347	struct net_device *dev = adapter->netdev;
4348	int tc;
4349	u32 dv_id;
4350
4351	/* Calculate max LAN frame size */
4352	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4353
4354#ifdef IXGBE_FCOE
4355	/* FCoE traffic class uses FCOE jumbo frames */
4356	if ((dev->features & NETIF_F_FCOE_MTU) &&
4357	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4358	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4359		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4360#endif
4361
4362	/* Calculate delay value for device */
4363	switch (hw->mac.type) {
4364	case ixgbe_mac_X540:
4365	case ixgbe_mac_X550:
4366	case ixgbe_mac_X550EM_x:
4367		dv_id = IXGBE_LOW_DV_X540(tc);
4368		break;
4369	default:
4370		dv_id = IXGBE_LOW_DV(tc);
4371		break;
4372	}
4373
4374	/* Delay value is calculated in bit times convert to KB */
4375	return IXGBE_BT2KB(dv_id);
4376}
4377
4378/*
4379 * ixgbe_pbthresh_setup - calculate and setup high low water marks
4380 */
4381static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4382{
4383	struct ixgbe_hw *hw = &adapter->hw;
4384	int num_tc = netdev_get_num_tc(adapter->netdev);
4385	int i;
4386
4387	if (!num_tc)
4388		num_tc = 1;
4389
4390	for (i = 0; i < num_tc; i++) {
4391		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4392		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4393
4394		/* Low water marks must not be larger than high water marks */
4395		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4396			hw->fc.low_water[i] = 0;
4397	}
4398
4399	for (; i < MAX_TRAFFIC_CLASS; i++)
4400		hw->fc.high_water[i] = 0;
4401}
4402
4403static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4404{
4405	struct ixgbe_hw *hw = &adapter->hw;
4406	int hdrm;
4407	u8 tc = netdev_get_num_tc(adapter->netdev);
4408
4409	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4410	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4411		hdrm = 32 << adapter->fdir_pballoc;
4412	else
4413		hdrm = 0;
4414
4415	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4416	ixgbe_pbthresh_setup(adapter);
4417}
4418
4419static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4420{
4421	struct ixgbe_hw *hw = &adapter->hw;
4422	struct hlist_node *node2;
4423	struct ixgbe_fdir_filter *filter;
4424
4425	spin_lock(&adapter->fdir_perfect_lock);
4426
4427	if (!hlist_empty(&adapter->fdir_filter_list))
4428		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4429
4430	hlist_for_each_entry_safe(filter, node2,
4431				  &adapter->fdir_filter_list, fdir_node) {
4432		ixgbe_fdir_write_perfect_filter_82599(hw,
4433				&filter->filter,
4434				filter->sw_idx,
4435				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4436				IXGBE_FDIR_DROP_QUEUE :
4437				adapter->rx_ring[filter->action]->reg_idx);
4438	}
4439
4440	spin_unlock(&adapter->fdir_perfect_lock);
4441}
4442
4443static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4444				      struct ixgbe_adapter *adapter)
4445{
4446	struct ixgbe_hw *hw = &adapter->hw;
4447	u32 vmolr;
4448
4449	/* No unicast promiscuous support for VMDQ devices. */
4450	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4451	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4452
4453	/* clear the affected bit */
4454	vmolr &= ~IXGBE_VMOLR_MPE;
4455
4456	if (dev->flags & IFF_ALLMULTI) {
4457		vmolr |= IXGBE_VMOLR_MPE;
4458	} else {
4459		vmolr |= IXGBE_VMOLR_ROMPE;
4460		hw->mac.ops.update_mc_addr_list(hw, dev);
4461	}
4462	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4463	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4464}
4465
4466static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4467{
4468	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4469	int rss_i = adapter->num_rx_queues_per_pool;
4470	struct ixgbe_hw *hw = &adapter->hw;
4471	u16 pool = vadapter->pool;
4472	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4473		      IXGBE_PSRTYPE_UDPHDR |
4474		      IXGBE_PSRTYPE_IPV4HDR |
4475		      IXGBE_PSRTYPE_L2HDR |
4476		      IXGBE_PSRTYPE_IPV6HDR;
4477
4478	if (hw->mac.type == ixgbe_mac_82598EB)
4479		return;
4480
4481	if (rss_i > 3)
4482		psrtype |= 2 << 29;
4483	else if (rss_i > 1)
4484		psrtype |= 1 << 29;
4485
4486	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4487}
4488
4489/**
4490 * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4491 * @rx_ring: ring to free buffers from
4492 **/
4493static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4494{
4495	struct device *dev = rx_ring->dev;
4496	unsigned long size;
4497	u16 i;
4498
4499	/* ring already cleared, nothing to do */
4500	if (!rx_ring->rx_buffer_info)
4501		return;
4502
4503	/* Free all the Rx ring sk_buffs */
4504	for (i = 0; i < rx_ring->count; i++) {
4505		struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4506
4507		if (rx_buffer->skb) {
4508			struct sk_buff *skb = rx_buffer->skb;
4509			if (IXGBE_CB(skb)->page_released)
4510				dma_unmap_page(dev,
4511					       IXGBE_CB(skb)->dma,
4512					       ixgbe_rx_bufsz(rx_ring),
4513					       DMA_FROM_DEVICE);
4514			dev_kfree_skb(skb);
4515			rx_buffer->skb = NULL;
4516		}
4517
4518		if (!rx_buffer->page)
4519			continue;
4520
4521		dma_unmap_page(dev, rx_buffer->dma,
4522			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
4523		__free_pages(rx_buffer->page, ixgbe_rx_pg_order(rx_ring));
4524
4525		rx_buffer->page = NULL;
4526	}
4527
4528	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4529	memset(rx_ring->rx_buffer_info, 0, size);
4530
4531	/* Zero out the descriptor ring */
4532	memset(rx_ring->desc, 0, rx_ring->size);
4533
4534	rx_ring->next_to_alloc = 0;
4535	rx_ring->next_to_clean = 0;
4536	rx_ring->next_to_use = 0;
4537}
4538
4539static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4540				   struct ixgbe_ring *rx_ring)
4541{
4542	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4543	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4544
4545	/* shutdown specific queue receive and wait for dma to settle */
4546	ixgbe_disable_rx_queue(adapter, rx_ring);
4547	usleep_range(10000, 20000);
4548	ixgbe_irq_disable_queues(adapter, ((u64)1 << index));
4549	ixgbe_clean_rx_ring(rx_ring);
4550	rx_ring->l2_accel_priv = NULL;
4551}
4552
4553static int ixgbe_fwd_ring_down(struct net_device *vdev,
4554			       struct ixgbe_fwd_adapter *accel)
4555{
4556	struct ixgbe_adapter *adapter = accel->real_adapter;
4557	unsigned int rxbase = accel->rx_base_queue;
4558	unsigned int txbase = accel->tx_base_queue;
4559	int i;
4560
4561	netif_tx_stop_all_queues(vdev);
4562
4563	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4564		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4565		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4566	}
4567
4568	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4569		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4570		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4571	}
4572
4573
4574	return 0;
4575}
4576
4577static int ixgbe_fwd_ring_up(struct net_device *vdev,
4578			     struct ixgbe_fwd_adapter *accel)
4579{
4580	struct ixgbe_adapter *adapter = accel->real_adapter;
4581	unsigned int rxbase, txbase, queues;
4582	int i, baseq, err = 0;
4583
4584	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4585		return 0;
4586
4587	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4588	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4589		   accel->pool, adapter->num_rx_pools,
4590		   baseq, baseq + adapter->num_rx_queues_per_pool,
4591		   adapter->fwd_bitmask);
4592
4593	accel->netdev = vdev;
4594	accel->rx_base_queue = rxbase = baseq;
4595	accel->tx_base_queue = txbase = baseq;
4596
4597	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4598		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4599
4600	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4601		adapter->rx_ring[rxbase + i]->netdev = vdev;
4602		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
4603		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
4604	}
4605
4606	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4607		adapter->tx_ring[txbase + i]->netdev = vdev;
4608		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
4609	}
4610
4611	queues = min_t(unsigned int,
4612		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
4613	err = netif_set_real_num_tx_queues(vdev, queues);
4614	if (err)
4615		goto fwd_queue_err;
4616
4617	err = netif_set_real_num_rx_queues(vdev, queues);
4618	if (err)
4619		goto fwd_queue_err;
4620
4621	if (is_valid_ether_addr(vdev->dev_addr))
4622		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
4623
4624	ixgbe_fwd_psrtype(accel);
4625	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
4626	return err;
4627fwd_queue_err:
4628	ixgbe_fwd_ring_down(vdev, accel);
4629	return err;
4630}
4631
4632static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
4633{
4634	struct net_device *upper;
4635	struct list_head *iter;
4636	int err;
4637
4638	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
4639		if (netif_is_macvlan(upper)) {
4640			struct macvlan_dev *dfwd = netdev_priv(upper);
4641			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
4642
4643			if (dfwd->fwd_priv) {
4644				err = ixgbe_fwd_ring_up(upper, vadapter);
4645				if (err)
4646					continue;
4647			}
4648		}
4649	}
4650}
4651
4652static void ixgbe_configure(struct ixgbe_adapter *adapter)
4653{
4654	struct ixgbe_hw *hw = &adapter->hw;
4655
4656	ixgbe_configure_pb(adapter);
4657#ifdef CONFIG_IXGBE_DCB
4658	ixgbe_configure_dcb(adapter);
4659#endif
4660	/*
4661	 * We must restore virtualization before VLANs or else
4662	 * the VLVF registers will not be populated
4663	 */
4664	ixgbe_configure_virtualization(adapter);
4665
4666	ixgbe_set_rx_mode(adapter->netdev);
4667	ixgbe_restore_vlan(adapter);
4668
4669	switch (hw->mac.type) {
4670	case ixgbe_mac_82599EB:
4671	case ixgbe_mac_X540:
4672		hw->mac.ops.disable_rx_buff(hw);
4673		break;
4674	default:
4675		break;
4676	}
4677
4678	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
4679		ixgbe_init_fdir_signature_82599(&adapter->hw,
4680						adapter->fdir_pballoc);
4681	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
4682		ixgbe_init_fdir_perfect_82599(&adapter->hw,
4683					      adapter->fdir_pballoc);
4684		ixgbe_fdir_filter_restore(adapter);
4685	}
4686
4687	switch (hw->mac.type) {
4688	case ixgbe_mac_82599EB:
4689	case ixgbe_mac_X540:
4690		hw->mac.ops.enable_rx_buff(hw);
4691		break;
4692	default:
4693		break;
4694	}
4695
4696#ifdef IXGBE_FCOE
4697	/* configure FCoE L2 filters, redirection table, and Rx control */
4698	ixgbe_configure_fcoe(adapter);
4699
4700#endif /* IXGBE_FCOE */
4701	ixgbe_configure_tx(adapter);
4702	ixgbe_configure_rx(adapter);
4703	ixgbe_configure_dfwd(adapter);
4704}
4705
4706static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
4707{
4708	switch (hw->phy.type) {
4709	case ixgbe_phy_sfp_avago:
4710	case ixgbe_phy_sfp_ftl:
4711	case ixgbe_phy_sfp_intel:
4712	case ixgbe_phy_sfp_unknown:
4713	case ixgbe_phy_sfp_passive_tyco:
4714	case ixgbe_phy_sfp_passive_unknown:
4715	case ixgbe_phy_sfp_active_unknown:
4716	case ixgbe_phy_sfp_ftl_active:
4717	case ixgbe_phy_qsfp_passive_unknown:
4718	case ixgbe_phy_qsfp_active_unknown:
4719	case ixgbe_phy_qsfp_intel:
4720	case ixgbe_phy_qsfp_unknown:
4721	/* ixgbe_phy_none is set when no SFP module is present */
4722	case ixgbe_phy_none:
4723		return true;
4724	case ixgbe_phy_nl:
4725		if (hw->mac.type == ixgbe_mac_82598EB)
4726			return true;
4727	default:
4728		return false;
4729	}
4730}
4731
4732/**
4733 * ixgbe_sfp_link_config - set up SFP+ link
4734 * @adapter: pointer to private adapter struct
4735 **/
4736static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
4737{
4738	/*
4739	 * We are assuming the worst case scenario here, and that
4740	 * is that an SFP was inserted/removed after the reset
4741	 * but before SFP detection was enabled.  As such the best
4742	 * solution is to just start searching as soon as we start
4743	 */
4744	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
4745		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
4746
4747	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
4748}
4749
4750/**
4751 * ixgbe_non_sfp_link_config - set up non-SFP+ link
4752 * @hw: pointer to private hardware struct
4753 *
4754 * Returns 0 on success, negative on failure
4755 **/
4756static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
4757{
4758	u32 speed;
4759	bool autoneg, link_up = false;
4760	u32 ret = IXGBE_ERR_LINK_SETUP;
4761
4762	if (hw->mac.ops.check_link)
4763		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
4764
4765	if (ret)
4766		return ret;
4767
4768	speed = hw->phy.autoneg_advertised;
4769	if ((!speed) && (hw->mac.ops.get_link_capabilities))
4770		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
4771							&autoneg);
4772	if (ret)
4773		return ret;
4774
4775	if (hw->mac.ops.setup_link)
4776		ret = hw->mac.ops.setup_link(hw, speed, link_up);
4777
4778	return ret;
4779}
4780
4781static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
4782{
4783	struct ixgbe_hw *hw = &adapter->hw;
4784	u32 gpie = 0;
4785
4786	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
4787		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4788		       IXGBE_GPIE_OCD;
4789		gpie |= IXGBE_GPIE_EIAME;
4790		/*
4791		 * use EIAM to auto-mask when MSI-X interrupt is asserted
4792		 * this saves a register write for every interrupt
4793		 */
4794		switch (hw->mac.type) {
4795		case ixgbe_mac_82598EB:
4796			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4797			break;
4798		case ixgbe_mac_82599EB:
4799		case ixgbe_mac_X540:
4800		case ixgbe_mac_X550:
4801		case ixgbe_mac_X550EM_x:
4802		default:
4803			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4804			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4805			break;
4806		}
4807	} else {
4808		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
4809		 * specifically only auto mask tx and rx interrupts */
4810		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4811	}
4812
4813	/* XXX: to interrupt immediately for EICS writes, enable this */
4814	/* gpie |= IXGBE_GPIE_EIMEN; */
4815
4816	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4817		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
4818
4819		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4820		case IXGBE_82599_VMDQ_8Q_MASK:
4821			gpie |= IXGBE_GPIE_VTMODE_16;
4822			break;
4823		case IXGBE_82599_VMDQ_4Q_MASK:
4824			gpie |= IXGBE_GPIE_VTMODE_32;
4825			break;
4826		default:
4827			gpie |= IXGBE_GPIE_VTMODE_64;
4828			break;
4829		}
4830	}
4831
4832	/* Enable Thermal over heat sensor interrupt */
4833	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
4834		switch (adapter->hw.mac.type) {
4835		case ixgbe_mac_82599EB:
4836			gpie |= IXGBE_SDP0_GPIEN;
4837			break;
4838		case ixgbe_mac_X540:
4839			gpie |= IXGBE_EIMS_TS;
4840			break;
4841		default:
4842			break;
4843		}
4844	}
4845
4846	/* Enable fan failure interrupt */
4847	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
4848		gpie |= IXGBE_SDP1_GPIEN;
4849
4850	if (hw->mac.type == ixgbe_mac_82599EB) {
4851		gpie |= IXGBE_SDP1_GPIEN;
4852		gpie |= IXGBE_SDP2_GPIEN;
4853	}
4854
4855	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4856}
4857
4858static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4859{
4860	struct ixgbe_hw *hw = &adapter->hw;
4861	int err;
4862	u32 ctrl_ext;
4863
4864	ixgbe_get_hw_control(adapter);
4865	ixgbe_setup_gpie(adapter);
4866
4867	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4868		ixgbe_configure_msix(adapter);
4869	else
4870		ixgbe_configure_msi_and_legacy(adapter);
4871
4872	/* enable the optics for 82599 SFP+ fiber */
4873	if (hw->mac.ops.enable_tx_laser)
4874		hw->mac.ops.enable_tx_laser(hw);
4875
4876	smp_mb__before_atomic();
4877	clear_bit(__IXGBE_DOWN, &adapter->state);
4878	ixgbe_napi_enable_all(adapter);
4879
4880	if (ixgbe_is_sfp(hw)) {
4881		ixgbe_sfp_link_config(adapter);
4882	} else {
4883		err = ixgbe_non_sfp_link_config(hw);
4884		if (err)
4885			e_err(probe, "link_config FAILED %d\n", err);
4886	}
4887
4888	/* clear any pending interrupts, may auto mask */
4889	IXGBE_READ_REG(hw, IXGBE_EICR);
4890	ixgbe_irq_enable(adapter, true, true);
4891
4892	/*
4893	 * If this adapter has a fan, check to see if we had a failure
4894	 * before we enabled the interrupt.
4895	 */
4896	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4897		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4898		if (esdp & IXGBE_ESDP_SDP1)
4899			e_crit(drv, "Fan has stopped, replace the adapter\n");
4900	}
4901
4902	/* bring the link up in the watchdog, this could race with our first
4903	 * link up interrupt but shouldn't be a problem */
4904	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4905	adapter->link_check_timeout = jiffies;
4906	mod_timer(&adapter->service_timer, jiffies);
4907
4908	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
4909	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
4910	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4911	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4912}
4913
4914void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4915{
4916	WARN_ON(in_interrupt());
4917	/* put off any impending NetWatchDogTimeout */
4918	adapter->netdev->trans_start = jiffies;
4919
4920	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4921		usleep_range(1000, 2000);
4922	ixgbe_down(adapter);
4923	/*
4924	 * If SR-IOV enabled then wait a bit before bringing the adapter
4925	 * back up to give the VFs time to respond to the reset.  The
4926	 * two second wait is based upon the watchdog timer cycle in
4927	 * the VF driver.
4928	 */
4929	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4930		msleep(2000);
4931	ixgbe_up(adapter);
4932	clear_bit(__IXGBE_RESETTING, &adapter->state);
4933}
4934
4935void ixgbe_up(struct ixgbe_adapter *adapter)
4936{
4937	/* hardware has been reset, we need to reload some things */
4938	ixgbe_configure(adapter);
4939
4940	ixgbe_up_complete(adapter);
4941}
4942
4943void ixgbe_reset(struct ixgbe_adapter *adapter)
4944{
4945	struct ixgbe_hw *hw = &adapter->hw;
4946	struct net_device *netdev = adapter->netdev;
4947	int err;
4948	u8 old_addr[ETH_ALEN];
4949
4950	if (ixgbe_removed(hw->hw_addr))
4951		return;
4952	/* lock SFP init bit to prevent race conditions with the watchdog */
4953	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
4954		usleep_range(1000, 2000);
4955
4956	/* clear all SFP and link config related flags while holding SFP_INIT */
4957	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
4958			     IXGBE_FLAG2_SFP_NEEDS_RESET);
4959	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4960
4961	err = hw->mac.ops.init_hw(hw);
4962	switch (err) {
4963	case 0:
4964	case IXGBE_ERR_SFP_NOT_PRESENT:
4965	case IXGBE_ERR_SFP_NOT_SUPPORTED:
4966		break;
4967	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
4968		e_dev_err("master disable timed out\n");
4969		break;
4970	case IXGBE_ERR_EEPROM_VERSION:
4971		/* We are running on a pre-production device, log a warning */
4972		e_dev_warn("This device is a pre-production adapter/LOM. "
4973			   "Please be aware there may be issues associated with "
4974			   "your hardware.  If you are experiencing problems "
4975			   "please contact your Intel or hardware "
4976			   "representative who provided you with this "
4977			   "hardware.\n");
4978		break;
4979	default:
4980		e_dev_err("Hardware Error: %d\n", err);
4981	}
4982
4983	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
4984	/* do not flush user set addresses */
4985	memcpy(old_addr, &adapter->mac_table[0].addr, netdev->addr_len);
4986	ixgbe_flush_sw_mac_table(adapter);
4987	ixgbe_mac_set_default_filter(adapter, old_addr);
4988
4989	/* update SAN MAC vmdq pool selection */
4990	if (hw->mac.san_mac_rar_index)
4991		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
4992
4993	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
4994		ixgbe_ptp_reset(adapter);
4995}
4996
4997/**
4998 * ixgbe_clean_tx_ring - Free Tx Buffers
4999 * @tx_ring: ring to be cleaned
5000 **/
5001static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5002{
5003	struct ixgbe_tx_buffer *tx_buffer_info;
5004	unsigned long size;
5005	u16 i;
5006
5007	/* ring already cleared, nothing to do */
5008	if (!tx_ring->tx_buffer_info)
5009		return;
5010
5011	/* Free all the Tx ring sk_buffs */
5012	for (i = 0; i < tx_ring->count; i++) {
5013		tx_buffer_info = &tx_ring->tx_buffer_info[i];
5014		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
5015	}
5016
5017	netdev_tx_reset_queue(txring_txq(tx_ring));
5018
5019	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5020	memset(tx_ring->tx_buffer_info, 0, size);
5021
5022	/* Zero out the descriptor ring */
5023	memset(tx_ring->desc, 0, tx_ring->size);
5024
5025	tx_ring->next_to_use = 0;
5026	tx_ring->next_to_clean = 0;
5027}
5028
5029/**
5030 * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5031 * @adapter: board private structure
5032 **/
5033static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5034{
5035	int i;
5036
5037	for (i = 0; i < adapter->num_rx_queues; i++)
5038		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5039}
5040
5041/**
5042 * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5043 * @adapter: board private structure
5044 **/
5045static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5046{
5047	int i;
5048
5049	for (i = 0; i < adapter->num_tx_queues; i++)
5050		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5051}
5052
5053static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5054{
5055	struct hlist_node *node2;
5056	struct ixgbe_fdir_filter *filter;
5057
5058	spin_lock(&adapter->fdir_perfect_lock);
5059
5060	hlist_for_each_entry_safe(filter, node2,
5061				  &adapter->fdir_filter_list, fdir_node) {
5062		hlist_del(&filter->fdir_node);
5063		kfree(filter);
5064	}
5065	adapter->fdir_filter_count = 0;
5066
5067	spin_unlock(&adapter->fdir_perfect_lock);
5068}
5069
5070void ixgbe_down(struct ixgbe_adapter *adapter)
5071{
5072	struct net_device *netdev = adapter->netdev;
5073	struct ixgbe_hw *hw = &adapter->hw;
5074	struct net_device *upper;
5075	struct list_head *iter;
5076	int i;
5077
5078	/* signal that we are down to the interrupt handler */
5079	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5080		return; /* do nothing if already down */
5081
5082	/* disable receives */
5083	hw->mac.ops.disable_rx(hw);
5084
5085	/* disable all enabled rx queues */
5086	for (i = 0; i < adapter->num_rx_queues; i++)
5087		/* this call also flushes the previous write */
5088		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5089
5090	usleep_range(10000, 20000);
5091
5092	netif_tx_stop_all_queues(netdev);
5093
5094	/* call carrier off first to avoid false dev_watchdog timeouts */
5095	netif_carrier_off(netdev);
5096	netif_tx_disable(netdev);
5097
5098	/* disable any upper devices */
5099	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5100		if (netif_is_macvlan(upper)) {
5101			struct macvlan_dev *vlan = netdev_priv(upper);
5102
5103			if (vlan->fwd_priv) {
5104				netif_tx_stop_all_queues(upper);
5105				netif_carrier_off(upper);
5106				netif_tx_disable(upper);
5107			}
5108		}
5109	}
5110
5111	ixgbe_irq_disable(adapter);
5112
5113	ixgbe_napi_disable_all(adapter);
5114
5115	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
5116			     IXGBE_FLAG2_RESET_REQUESTED);
5117	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5118
5119	del_timer_sync(&adapter->service_timer);
5120
5121	if (adapter->num_vfs) {
5122		/* Clear EITR Select mapping */
5123		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5124
5125		/* Mark all the VFs as inactive */
5126		for (i = 0 ; i < adapter->num_vfs; i++)
5127			adapter->vfinfo[i].clear_to_send = false;
5128
5129		/* ping all the active vfs to let them know we are going down */
5130		ixgbe_ping_all_vfs(adapter);
5131
5132		/* Disable all VFTE/VFRE TX/RX */
5133		ixgbe_disable_tx_rx(adapter);
5134	}
5135
5136	/* disable transmits in the hardware now that interrupts are off */
5137	for (i = 0; i < adapter->num_tx_queues; i++) {
5138		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5139		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5140	}
5141
5142	/* Disable the Tx DMA engine on 82599 and later MAC */
5143	switch (hw->mac.type) {
5144	case ixgbe_mac_82599EB:
5145	case ixgbe_mac_X540:
5146	case ixgbe_mac_X550:
5147	case ixgbe_mac_X550EM_x:
5148		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5149				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5150				 ~IXGBE_DMATXCTL_TE));
5151		break;
5152	default:
5153		break;
5154	}
5155
5156	if (!pci_channel_offline(adapter->pdev))
5157		ixgbe_reset(adapter);
5158
5159	/* power down the optics for 82599 SFP+ fiber */
5160	if (hw->mac.ops.disable_tx_laser)
5161		hw->mac.ops.disable_tx_laser(hw);
5162
5163	ixgbe_clean_all_tx_rings(adapter);
5164	ixgbe_clean_all_rx_rings(adapter);
5165
5166#ifdef CONFIG_IXGBE_DCA
5167	/* since we reset the hardware DCA settings were cleared */
5168	ixgbe_setup_dca(adapter);
5169#endif
5170}
5171
5172/**
5173 * ixgbe_tx_timeout - Respond to a Tx Hang
5174 * @netdev: network interface device structure
5175 **/
5176static void ixgbe_tx_timeout(struct net_device *netdev)
5177{
5178	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5179
5180	/* Do the reset outside of interrupt context */
5181	ixgbe_tx_timeout_reset(adapter);
5182}
5183
5184/**
5185 * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5186 * @adapter: board private structure to initialize
5187 *
5188 * ixgbe_sw_init initializes the Adapter private data structure.
5189 * Fields are initialized based on PCI device information and
5190 * OS network device settings (MTU size).
5191 **/
5192static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
5193{
5194	struct ixgbe_hw *hw = &adapter->hw;
5195	struct pci_dev *pdev = adapter->pdev;
5196	unsigned int rss, fdir;
5197	u32 fwsm;
5198#ifdef CONFIG_IXGBE_DCB
5199	int j;
5200	struct tc_configuration *tc;
5201#endif
5202
5203	/* PCI config space info */
5204
5205	hw->vendor_id = pdev->vendor;
5206	hw->device_id = pdev->device;
5207	hw->revision_id = pdev->revision;
5208	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5209	hw->subsystem_device_id = pdev->subsystem_device;
5210
5211	/* Set common capability flags and settings */
5212	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5213	adapter->ring_feature[RING_F_RSS].limit = rss;
5214	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5215	adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
5216	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5217	adapter->atr_sample_rate = 20;
5218	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5219	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5220	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5221#ifdef CONFIG_IXGBE_DCA
5222	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5223#endif
5224#ifdef IXGBE_FCOE
5225	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5226	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5227#ifdef CONFIG_IXGBE_DCB
5228	/* Default traffic class to use for FCoE */
5229	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5230#endif /* CONFIG_IXGBE_DCB */
5231#endif /* IXGBE_FCOE */
5232
5233	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5234				     hw->mac.num_rar_entries,
5235				     GFP_ATOMIC);
5236
5237	/* Set MAC specific capability flags and exceptions */
5238	switch (hw->mac.type) {
5239	case ixgbe_mac_82598EB:
5240		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5241		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
5242
5243		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5244			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5245
5246		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5247		adapter->ring_feature[RING_F_FDIR].limit = 0;
5248		adapter->atr_sample_rate = 0;
5249		adapter->fdir_pballoc = 0;
5250#ifdef IXGBE_FCOE
5251		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5252		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5253#ifdef CONFIG_IXGBE_DCB
5254		adapter->fcoe.up = 0;
5255#endif /* IXGBE_DCB */
5256#endif /* IXGBE_FCOE */
5257		break;
5258	case ixgbe_mac_82599EB:
5259		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5260			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5261		break;
5262	case ixgbe_mac_X540:
5263		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
5264		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5265			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5266		break;
5267	case ixgbe_mac_X550EM_x:
5268	case ixgbe_mac_X550:
5269#ifdef CONFIG_IXGBE_DCA
5270		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5271#endif
5272		break;
5273	default:
5274		break;
5275	}
5276
5277#ifdef IXGBE_FCOE
5278	/* FCoE support exists, always init the FCoE lock */
5279	spin_lock_init(&adapter->fcoe.lock);
5280
5281#endif
5282	/* n-tuple support exists, always init our spinlock */
5283	spin_lock_init(&adapter->fdir_perfect_lock);
5284
5285#ifdef CONFIG_IXGBE_DCB
5286	switch (hw->mac.type) {
5287	case ixgbe_mac_X540:
5288	case ixgbe_mac_X550:
5289	case ixgbe_mac_X550EM_x:
5290		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5291		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5292		break;
5293	default:
5294		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5295		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5296		break;
5297	}
5298
5299	/* Configure DCB traffic classes */
5300	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5301		tc = &adapter->dcb_cfg.tc_config[j];
5302		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5303		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5304		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5305		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5306		tc->dcb_pfc = pfc_disabled;
5307	}
5308
5309	/* Initialize default user to priority mapping, UPx->TC0 */
5310	tc = &adapter->dcb_cfg.tc_config[0];
5311	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5312	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5313
5314	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5315	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5316	adapter->dcb_cfg.pfc_mode_enable = false;
5317	adapter->dcb_set_bitmap = 0x00;
5318	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5319	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5320	       sizeof(adapter->temp_dcb_cfg));
5321
5322#endif
5323
5324	/* default flow control settings */
5325	hw->fc.requested_mode = ixgbe_fc_full;
5326	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5327	ixgbe_pbthresh_setup(adapter);
5328	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5329	hw->fc.send_xon = true;
5330	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5331
5332#ifdef CONFIG_PCI_IOV
5333	if (max_vfs > 0)
5334		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5335
5336	/* assign number of SR-IOV VFs */
5337	if (hw->mac.type != ixgbe_mac_82598EB) {
5338		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5339			adapter->num_vfs = 0;
5340			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5341		} else {
5342			adapter->num_vfs = max_vfs;
5343		}
5344	}
5345#endif /* CONFIG_PCI_IOV */
5346
5347	/* enable itr by default in dynamic mode */
5348	adapter->rx_itr_setting = 1;
5349	adapter->tx_itr_setting = 1;
5350
5351	/* set default ring sizes */
5352	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5353	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5354
5355	/* set default work limits */
5356	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5357
5358	/* initialize eeprom parameters */
5359	if (ixgbe_init_eeprom_params_generic(hw)) {
5360		e_dev_err("EEPROM initialization failed\n");
5361		return -EIO;
5362	}
5363
5364	/* PF holds first pool slot */
5365	set_bit(0, &adapter->fwd_bitmask);
5366	set_bit(__IXGBE_DOWN, &adapter->state);
5367
5368	return 0;
5369}
5370
5371/**
5372 * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5373 * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5374 *
5375 * Return 0 on success, negative on failure
5376 **/
5377int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5378{
5379	struct device *dev = tx_ring->dev;
5380	int orig_node = dev_to_node(dev);
5381	int ring_node = -1;
5382	int size;
5383
5384	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5385
5386	if (tx_ring->q_vector)
5387		ring_node = tx_ring->q_vector->numa_node;
5388
5389	tx_ring->tx_buffer_info = vzalloc_node(size, ring_node);
5390	if (!tx_ring->tx_buffer_info)
5391		tx_ring->tx_buffer_info = vzalloc(size);
5392	if (!tx_ring->tx_buffer_info)
5393		goto err;
5394
5395	u64_stats_init(&tx_ring->syncp);
5396
5397	/* round up to nearest 4K */
5398	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5399	tx_ring->size = ALIGN(tx_ring->size, 4096);
5400
5401	set_dev_node(dev, ring_node);
5402	tx_ring->desc = dma_alloc_coherent(dev,
5403					   tx_ring->size,
5404					   &tx_ring->dma,
5405					   GFP_KERNEL);
5406	set_dev_node(dev, orig_node);
5407	if (!tx_ring->desc)
5408		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5409						   &tx_ring->dma, GFP_KERNEL);
5410	if (!tx_ring->desc)
5411		goto err;
5412
5413	tx_ring->next_to_use = 0;
5414	tx_ring->next_to_clean = 0;
5415	return 0;
5416
5417err:
5418	vfree(tx_ring->tx_buffer_info);
5419	tx_ring->tx_buffer_info = NULL;
5420	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5421	return -ENOMEM;
5422}
5423
5424/**
5425 * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5426 * @adapter: board private structure
5427 *
5428 * If this function returns with an error, then it's possible one or
5429 * more of the rings is populated (while the rest are not).  It is the
5430 * callers duty to clean those orphaned rings.
5431 *
5432 * Return 0 on success, negative on failure
5433 **/
5434static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5435{
5436	int i, err = 0;
5437
5438	for (i = 0; i < adapter->num_tx_queues; i++) {
5439		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5440		if (!err)
5441			continue;
5442
5443		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5444		goto err_setup_tx;
5445	}
5446
5447	return 0;
5448err_setup_tx:
5449	/* rewind the index freeing the rings as we go */
5450	while (i--)
5451		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5452	return err;
5453}
5454
5455/**
5456 * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5457 * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5458 *
5459 * Returns 0 on success, negative on failure
5460 **/
5461int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5462{
5463	struct device *dev = rx_ring->dev;
5464	int orig_node = dev_to_node(dev);
5465	int ring_node = -1;
5466	int size;
5467
5468	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5469
5470	if (rx_ring->q_vector)
5471		ring_node = rx_ring->q_vector->numa_node;
5472
5473	rx_ring->rx_buffer_info = vzalloc_node(size, ring_node);
5474	if (!rx_ring->rx_buffer_info)
5475		rx_ring->rx_buffer_info = vzalloc(size);
5476	if (!rx_ring->rx_buffer_info)
5477		goto err;
5478
5479	u64_stats_init(&rx_ring->syncp);
5480
5481	/* Round up to nearest 4K */
5482	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5483	rx_ring->size = ALIGN(rx_ring->size, 4096);
5484
5485	set_dev_node(dev, ring_node);
5486	rx_ring->desc = dma_alloc_coherent(dev,
5487					   rx_ring->size,
5488					   &rx_ring->dma,
5489					   GFP_KERNEL);
5490	set_dev_node(dev, orig_node);
5491	if (!rx_ring->desc)
5492		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5493						   &rx_ring->dma, GFP_KERNEL);
5494	if (!rx_ring->desc)
5495		goto err;
5496
5497	rx_ring->next_to_clean = 0;
5498	rx_ring->next_to_use = 0;
5499
5500	return 0;
5501err:
5502	vfree(rx_ring->rx_buffer_info);
5503	rx_ring->rx_buffer_info = NULL;
5504	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5505	return -ENOMEM;
5506}
5507
5508/**
5509 * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5510 * @adapter: board private structure
5511 *
5512 * If this function returns with an error, then it's possible one or
5513 * more of the rings is populated (while the rest are not).  It is the
5514 * callers duty to clean those orphaned rings.
5515 *
5516 * Return 0 on success, negative on failure
5517 **/
5518static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5519{
5520	int i, err = 0;
5521
5522	for (i = 0; i < adapter->num_rx_queues; i++) {
5523		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5524		if (!err)
5525			continue;
5526
5527		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5528		goto err_setup_rx;
5529	}
5530
5531#ifdef IXGBE_FCOE
5532	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5533	if (!err)
5534#endif
5535		return 0;
5536err_setup_rx:
5537	/* rewind the index freeing the rings as we go */
5538	while (i--)
5539		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5540	return err;
5541}
5542
5543/**
5544 * ixgbe_free_tx_resources - Free Tx Resources per Queue
5545 * @tx_ring: Tx descriptor ring for a specific queue
5546 *
5547 * Free all transmit software resources
5548 **/
5549void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5550{
5551	ixgbe_clean_tx_ring(tx_ring);
5552
5553	vfree(tx_ring->tx_buffer_info);
5554	tx_ring->tx_buffer_info = NULL;
5555
5556	/* if not set, then don't free */
5557	if (!tx_ring->desc)
5558		return;
5559
5560	dma_free_coherent(tx_ring->dev, tx_ring->size,
5561			  tx_ring->desc, tx_ring->dma);
5562
5563	tx_ring->desc = NULL;
5564}
5565
5566/**
5567 * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5568 * @adapter: board private structure
5569 *
5570 * Free all transmit software resources
5571 **/
5572static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
5573{
5574	int i;
5575
5576	for (i = 0; i < adapter->num_tx_queues; i++)
5577		if (adapter->tx_ring[i]->desc)
5578			ixgbe_free_tx_resources(adapter->tx_ring[i]);
5579}
5580
5581/**
5582 * ixgbe_free_rx_resources - Free Rx Resources
5583 * @rx_ring: ring to clean the resources from
5584 *
5585 * Free all receive software resources
5586 **/
5587void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
5588{
5589	ixgbe_clean_rx_ring(rx_ring);
5590
5591	vfree(rx_ring->rx_buffer_info);
5592	rx_ring->rx_buffer_info = NULL;
5593
5594	/* if not set, then don't free */
5595	if (!rx_ring->desc)
5596		return;
5597
5598	dma_free_coherent(rx_ring->dev, rx_ring->size,
5599			  rx_ring->desc, rx_ring->dma);
5600
5601	rx_ring->desc = NULL;
5602}
5603
5604/**
5605 * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
5606 * @adapter: board private structure
5607 *
5608 * Free all receive software resources
5609 **/
5610static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
5611{
5612	int i;
5613
5614#ifdef IXGBE_FCOE
5615	ixgbe_free_fcoe_ddp_resources(adapter);
5616
5617#endif
5618	for (i = 0; i < adapter->num_rx_queues; i++)
5619		if (adapter->rx_ring[i]->desc)
5620			ixgbe_free_rx_resources(adapter->rx_ring[i]);
5621}
5622
5623/**
5624 * ixgbe_change_mtu - Change the Maximum Transfer Unit
5625 * @netdev: network interface device structure
5626 * @new_mtu: new value for maximum frame size
5627 *
5628 * Returns 0 on success, negative on failure
5629 **/
5630static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
5631{
5632	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5633	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5634
5635	/* MTU < 68 is an error and causes problems on some kernels */
5636	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
5637		return -EINVAL;
5638
5639	/*
5640	 * For 82599EB we cannot allow legacy VFs to enable their receive
5641	 * paths when MTU greater than 1500 is configured.  So display a
5642	 * warning that legacy VFs will be disabled.
5643	 */
5644	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
5645	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
5646	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
5647		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
5648
5649	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5650
5651	/* must set new MTU before calling down or up */
5652	netdev->mtu = new_mtu;
5653
5654	if (netif_running(netdev))
5655		ixgbe_reinit_locked(adapter);
5656
5657	return 0;
5658}
5659
5660/**
5661 * ixgbe_open - Called when a network interface is made active
5662 * @netdev: network interface device structure
5663 *
5664 * Returns 0 on success, negative value on failure
5665 *
5666 * The open entry point is called when a network interface is made
5667 * active by the system (IFF_UP).  At this point all resources needed
5668 * for transmit and receive operations are allocated, the interrupt
5669 * handler is registered with the OS, the watchdog timer is started,
5670 * and the stack is notified that the interface is ready.
5671 **/
5672static int ixgbe_open(struct net_device *netdev)
5673{
5674	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5675	int err, queues;
5676
5677	/* disallow open during test */
5678	if (test_bit(__IXGBE_TESTING, &adapter->state))
5679		return -EBUSY;
5680
5681	netif_carrier_off(netdev);
5682
5683	/* allocate transmit descriptors */
5684	err = ixgbe_setup_all_tx_resources(adapter);
5685	if (err)
5686		goto err_setup_tx;
5687
5688	/* allocate receive descriptors */
5689	err = ixgbe_setup_all_rx_resources(adapter);
5690	if (err)
5691		goto err_setup_rx;
5692
5693	ixgbe_configure(adapter);
5694
5695	err = ixgbe_request_irq(adapter);
5696	if (err)
5697		goto err_req_irq;
5698
5699	/* Notify the stack of the actual queue counts. */
5700	if (adapter->num_rx_pools > 1)
5701		queues = adapter->num_rx_queues_per_pool;
5702	else
5703		queues = adapter->num_tx_queues;
5704
5705	err = netif_set_real_num_tx_queues(netdev, queues);
5706	if (err)
5707		goto err_set_queues;
5708
5709	if (adapter->num_rx_pools > 1 &&
5710	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
5711		queues = IXGBE_MAX_L2A_QUEUES;
5712	else
5713		queues = adapter->num_rx_queues;
5714	err = netif_set_real_num_rx_queues(netdev, queues);
5715	if (err)
5716		goto err_set_queues;
5717
5718	ixgbe_ptp_init(adapter);
5719
5720	ixgbe_up_complete(adapter);
5721
5722#if IS_ENABLED(CONFIG_IXGBE_VXLAN)
5723	vxlan_get_rx_port(netdev);
5724
5725#endif
5726	return 0;
5727
5728err_set_queues:
5729	ixgbe_free_irq(adapter);
5730err_req_irq:
5731	ixgbe_free_all_rx_resources(adapter);
5732err_setup_rx:
5733	ixgbe_free_all_tx_resources(adapter);
5734err_setup_tx:
5735	ixgbe_reset(adapter);
5736
5737	return err;
5738}
5739
5740static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
5741{
5742	ixgbe_ptp_suspend(adapter);
5743
5744	ixgbe_down(adapter);
5745	ixgbe_free_irq(adapter);
5746
5747	ixgbe_free_all_tx_resources(adapter);
5748	ixgbe_free_all_rx_resources(adapter);
5749}
5750
5751/**
5752 * ixgbe_close - Disables a network interface
5753 * @netdev: network interface device structure
5754 *
5755 * Returns 0, this is not allowed to fail
5756 *
5757 * The close entry point is called when an interface is de-activated
5758 * by the OS.  The hardware is still under the drivers control, but
5759 * needs to be disabled.  A global MAC reset is issued to stop the
5760 * hardware, and all transmit and receive resources are freed.
5761 **/
5762static int ixgbe_close(struct net_device *netdev)
5763{
5764	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5765
5766	ixgbe_ptp_stop(adapter);
5767
5768	ixgbe_close_suspend(adapter);
5769
5770	ixgbe_fdir_filter_exit(adapter);
5771
5772	ixgbe_release_hw_control(adapter);
5773
5774	return 0;
5775}
5776
5777#ifdef CONFIG_PM
5778static int ixgbe_resume(struct pci_dev *pdev)
5779{
5780	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5781	struct net_device *netdev = adapter->netdev;
5782	u32 err;
5783
5784	adapter->hw.hw_addr = adapter->io_addr;
5785	pci_set_power_state(pdev, PCI_D0);
5786	pci_restore_state(pdev);
5787	/*
5788	 * pci_restore_state clears dev->state_saved so call
5789	 * pci_save_state to restore it.
5790	 */
5791	pci_save_state(pdev);
5792
5793	err = pci_enable_device_mem(pdev);
5794	if (err) {
5795		e_dev_err("Cannot enable PCI device from suspend\n");
5796		return err;
5797	}
5798	smp_mb__before_atomic();
5799	clear_bit(__IXGBE_DISABLED, &adapter->state);
5800	pci_set_master(pdev);
5801
5802	pci_wake_from_d3(pdev, false);
5803
5804	ixgbe_reset(adapter);
5805
5806	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
5807
5808	rtnl_lock();
5809	err = ixgbe_init_interrupt_scheme(adapter);
5810	if (!err && netif_running(netdev))
5811		err = ixgbe_open(netdev);
5812
5813	rtnl_unlock();
5814
5815	if (err)
5816		return err;
5817
5818	netif_device_attach(netdev);
5819
5820	return 0;
5821}
5822#endif /* CONFIG_PM */
5823
5824static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
5825{
5826	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5827	struct net_device *netdev = adapter->netdev;
5828	struct ixgbe_hw *hw = &adapter->hw;
5829	u32 ctrl, fctrl;
5830	u32 wufc = adapter->wol;
5831#ifdef CONFIG_PM
5832	int retval = 0;
5833#endif
5834
5835	netif_device_detach(netdev);
5836
5837	rtnl_lock();
5838	if (netif_running(netdev))
5839		ixgbe_close_suspend(adapter);
5840	rtnl_unlock();
5841
5842	ixgbe_clear_interrupt_scheme(adapter);
5843
5844#ifdef CONFIG_PM
5845	retval = pci_save_state(pdev);
5846	if (retval)
5847		return retval;
5848
5849#endif
5850	if (hw->mac.ops.stop_link_on_d3)
5851		hw->mac.ops.stop_link_on_d3(hw);
5852
5853	if (wufc) {
5854		ixgbe_set_rx_mode(netdev);
5855
5856		/* enable the optics for 82599 SFP+ fiber as we can WoL */
5857		if (hw->mac.ops.enable_tx_laser)
5858			hw->mac.ops.enable_tx_laser(hw);
5859
5860		/* turn on all-multi mode if wake on multicast is enabled */
5861		if (wufc & IXGBE_WUFC_MC) {
5862			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5863			fctrl |= IXGBE_FCTRL_MPE;
5864			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5865		}
5866
5867		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
5868		ctrl |= IXGBE_CTRL_GIO_DIS;
5869		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
5870
5871		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
5872	} else {
5873		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
5874		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
5875	}
5876
5877	switch (hw->mac.type) {
5878	case ixgbe_mac_82598EB:
5879		pci_wake_from_d3(pdev, false);
5880		break;
5881	case ixgbe_mac_82599EB:
5882	case ixgbe_mac_X540:
5883	case ixgbe_mac_X550:
5884	case ixgbe_mac_X550EM_x:
5885		pci_wake_from_d3(pdev, !!wufc);
5886		break;
5887	default:
5888		break;
5889	}
5890
5891	*enable_wake = !!wufc;
5892
5893	ixgbe_release_hw_control(adapter);
5894
5895	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
5896		pci_disable_device(pdev);
5897
5898	return 0;
5899}
5900
5901#ifdef CONFIG_PM
5902static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
5903{
5904	int retval;
5905	bool wake;
5906
5907	retval = __ixgbe_shutdown(pdev, &wake);
5908	if (retval)
5909		return retval;
5910
5911	if (wake) {
5912		pci_prepare_to_sleep(pdev);
5913	} else {
5914		pci_wake_from_d3(pdev, false);
5915		pci_set_power_state(pdev, PCI_D3hot);
5916	}
5917
5918	return 0;
5919}
5920#endif /* CONFIG_PM */
5921
5922static void ixgbe_shutdown(struct pci_dev *pdev)
5923{
5924	bool wake;
5925
5926	__ixgbe_shutdown(pdev, &wake);
5927
5928	if (system_state == SYSTEM_POWER_OFF) {
5929		pci_wake_from_d3(pdev, wake);
5930		pci_set_power_state(pdev, PCI_D3hot);
5931	}
5932}
5933
5934/**
5935 * ixgbe_update_stats - Update the board statistics counters.
5936 * @adapter: board private structure
5937 **/
5938void ixgbe_update_stats(struct ixgbe_adapter *adapter)
5939{
5940	struct net_device *netdev = adapter->netdev;
5941	struct ixgbe_hw *hw = &adapter->hw;
5942	struct ixgbe_hw_stats *hwstats = &adapter->stats;
5943	u64 total_mpc = 0;
5944	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
5945	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
5946	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
5947	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
5948
5949	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5950	    test_bit(__IXGBE_RESETTING, &adapter->state))
5951		return;
5952
5953	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
5954		u64 rsc_count = 0;
5955		u64 rsc_flush = 0;
5956		for (i = 0; i < adapter->num_rx_queues; i++) {
5957			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
5958			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
5959		}
5960		adapter->rsc_total_count = rsc_count;
5961		adapter->rsc_total_flush = rsc_flush;
5962	}
5963
5964	for (i = 0; i < adapter->num_rx_queues; i++) {
5965		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
5966		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
5967		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
5968		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
5969		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
5970		bytes += rx_ring->stats.bytes;
5971		packets += rx_ring->stats.packets;
5972	}
5973	adapter->non_eop_descs = non_eop_descs;
5974	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
5975	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
5976	adapter->hw_csum_rx_error = hw_csum_rx_error;
5977	netdev->stats.rx_bytes = bytes;
5978	netdev->stats.rx_packets = packets;
5979
5980	bytes = 0;
5981	packets = 0;
5982	/* gather some stats to the adapter struct that are per queue */
5983	for (i = 0; i < adapter->num_tx_queues; i++) {
5984		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5985		restart_queue += tx_ring->tx_stats.restart_queue;
5986		tx_busy += tx_ring->tx_stats.tx_busy;
5987		bytes += tx_ring->stats.bytes;
5988		packets += tx_ring->stats.packets;
5989	}
5990	adapter->restart_queue = restart_queue;
5991	adapter->tx_busy = tx_busy;
5992	netdev->stats.tx_bytes = bytes;
5993	netdev->stats.tx_packets = packets;
5994
5995	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
5996
5997	/* 8 register reads */
5998	for (i = 0; i < 8; i++) {
5999		/* for packet buffers not used, the register should read 0 */
6000		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6001		missed_rx += mpc;
6002		hwstats->mpc[i] += mpc;
6003		total_mpc += hwstats->mpc[i];
6004		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6005		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6006		switch (hw->mac.type) {
6007		case ixgbe_mac_82598EB:
6008			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6009			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6010			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6011			hwstats->pxonrxc[i] +=
6012				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6013			break;
6014		case ixgbe_mac_82599EB:
6015		case ixgbe_mac_X540:
6016		case ixgbe_mac_X550:
6017		case ixgbe_mac_X550EM_x:
6018			hwstats->pxonrxc[i] +=
6019				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6020			break;
6021		default:
6022			break;
6023		}
6024	}
6025
6026	/*16 register reads */
6027	for (i = 0; i < 16; i++) {
6028		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6029		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6030		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6031		    (hw->mac.type == ixgbe_mac_X540) ||
6032		    (hw->mac.type == ixgbe_mac_X550) ||
6033		    (hw->mac.type == ixgbe_mac_X550EM_x)) {
6034			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6035			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6036			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6037			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6038		}
6039	}
6040
6041	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6042	/* work around hardware counting issue */
6043	hwstats->gprc -= missed_rx;
6044
6045	ixgbe_update_xoff_received(adapter);
6046
6047	/* 82598 hardware only has a 32 bit counter in the high register */
6048	switch (hw->mac.type) {
6049	case ixgbe_mac_82598EB:
6050		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6051		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6052		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6053		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6054		break;
6055	case ixgbe_mac_X540:
6056	case ixgbe_mac_X550:
6057	case ixgbe_mac_X550EM_x:
6058		/* OS2BMC stats are X540 and later */
6059		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6060		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6061		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6062		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6063	case ixgbe_mac_82599EB:
6064		for (i = 0; i < 16; i++)
6065			adapter->hw_rx_no_dma_resources +=
6066					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6067		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6068		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6069		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6070		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6071		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6072		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6073		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6074		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6075		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6076#ifdef IXGBE_FCOE
6077		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6078		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6079		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6080		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6081		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6082		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6083		/* Add up per cpu counters for total ddp aloc fail */
6084		if (adapter->fcoe.ddp_pool) {
6085			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6086			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6087			unsigned int cpu;
6088			u64 noddp = 0, noddp_ext_buff = 0;
6089			for_each_possible_cpu(cpu) {
6090				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6091				noddp += ddp_pool->noddp;
6092				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6093			}
6094			hwstats->fcoe_noddp = noddp;
6095			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6096		}
6097#endif /* IXGBE_FCOE */
6098		break;
6099	default:
6100		break;
6101	}
6102	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6103	hwstats->bprc += bprc;
6104	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6105	if (hw->mac.type == ixgbe_mac_82598EB)
6106		hwstats->mprc -= bprc;
6107	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6108	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6109	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6110	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6111	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6112	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6113	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6114	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6115	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6116	hwstats->lxontxc += lxon;
6117	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6118	hwstats->lxofftxc += lxoff;
6119	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6120	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6121	/*
6122	 * 82598 errata - tx of flow control packets is included in tx counters
6123	 */
6124	xon_off_tot = lxon + lxoff;
6125	hwstats->gptc -= xon_off_tot;
6126	hwstats->mptc -= xon_off_tot;
6127	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6128	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6129	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6130	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6131	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6132	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6133	hwstats->ptc64 -= xon_off_tot;
6134	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6135	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6136	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6137	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6138	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6139	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6140
6141	/* Fill out the OS statistics structure */
6142	netdev->stats.multicast = hwstats->mprc;
6143
6144	/* Rx Errors */
6145	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6146	netdev->stats.rx_dropped = 0;
6147	netdev->stats.rx_length_errors = hwstats->rlec;
6148	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6149	netdev->stats.rx_missed_errors = total_mpc;
6150}
6151
6152/**
6153 * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6154 * @adapter: pointer to the device adapter structure
6155 **/
6156static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6157{
6158	struct ixgbe_hw *hw = &adapter->hw;
6159	int i;
6160
6161	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6162		return;
6163
6164	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6165
6166	/* if interface is down do nothing */
6167	if (test_bit(__IXGBE_DOWN, &adapter->state))
6168		return;
6169
6170	/* do nothing if we are not using signature filters */
6171	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6172		return;
6173
6174	adapter->fdir_overflow++;
6175
6176	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6177		for (i = 0; i < adapter->num_tx_queues; i++)
6178			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6179				&(adapter->tx_ring[i]->state));
6180		/* re-enable flow director interrupts */
6181		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6182	} else {
6183		e_err(probe, "failed to finish FDIR re-initialization, "
6184		      "ignored adding FDIR ATR filters\n");
6185	}
6186}
6187
6188/**
6189 * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6190 * @adapter: pointer to the device adapter structure
6191 *
6192 * This function serves two purposes.  First it strobes the interrupt lines
6193 * in order to make certain interrupts are occurring.  Secondly it sets the
6194 * bits needed to check for TX hangs.  As a result we should immediately
6195 * determine if a hang has occurred.
6196 */
6197static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6198{
6199	struct ixgbe_hw *hw = &adapter->hw;
6200	u64 eics = 0;
6201	int i;
6202
6203	/* If we're down, removing or resetting, just bail */
6204	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6205	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6206	    test_bit(__IXGBE_RESETTING, &adapter->state))
6207		return;
6208
6209	/* Force detection of hung controller */
6210	if (netif_carrier_ok(adapter->netdev)) {
6211		for (i = 0; i < adapter->num_tx_queues; i++)
6212			set_check_for_tx_hang(adapter->tx_ring[i]);
6213	}
6214
6215	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6216		/*
6217		 * for legacy and MSI interrupts don't set any bits
6218		 * that are enabled for EIAM, because this operation
6219		 * would set *both* EIMS and EICS for any bit in EIAM
6220		 */
6221		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6222			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6223	} else {
6224		/* get one bit for every active tx/rx interrupt vector */
6225		for (i = 0; i < adapter->num_q_vectors; i++) {
6226			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6227			if (qv->rx.ring || qv->tx.ring)
6228				eics |= ((u64)1 << i);
6229		}
6230	}
6231
6232	/* Cause software interrupt to ensure rings are cleaned */
6233	ixgbe_irq_rearm_queues(adapter, eics);
6234}
6235
6236/**
6237 * ixgbe_watchdog_update_link - update the link status
6238 * @adapter: pointer to the device adapter structure
6239 * @link_speed: pointer to a u32 to store the link_speed
6240 **/
6241static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6242{
6243	struct ixgbe_hw *hw = &adapter->hw;
6244	u32 link_speed = adapter->link_speed;
6245	bool link_up = adapter->link_up;
6246	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6247
6248	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6249		return;
6250
6251	if (hw->mac.ops.check_link) {
6252		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6253	} else {
6254		/* always assume link is up, if no check link function */
6255		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6256		link_up = true;
6257	}
6258
6259	if (adapter->ixgbe_ieee_pfc)
6260		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6261
6262	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6263		hw->mac.ops.fc_enable(hw);
6264		ixgbe_set_rx_drop_en(adapter);
6265	}
6266
6267	if (link_up ||
6268	    time_after(jiffies, (adapter->link_check_timeout +
6269				 IXGBE_TRY_LINK_TIMEOUT))) {
6270		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6271		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6272		IXGBE_WRITE_FLUSH(hw);
6273	}
6274
6275	adapter->link_up = link_up;
6276	adapter->link_speed = link_speed;
6277}
6278
6279static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6280{
6281#ifdef CONFIG_IXGBE_DCB
6282	struct net_device *netdev = adapter->netdev;
6283	struct dcb_app app = {
6284			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6285			      .protocol = 0,
6286			     };
6287	u8 up = 0;
6288
6289	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6290		up = dcb_ieee_getapp_mask(netdev, &app);
6291
6292	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6293#endif
6294}
6295
6296/**
6297 * ixgbe_watchdog_link_is_up - update netif_carrier status and
6298 *                             print link up message
6299 * @adapter: pointer to the device adapter structure
6300 **/
6301static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6302{
6303	struct net_device *netdev = adapter->netdev;
6304	struct ixgbe_hw *hw = &adapter->hw;
6305	struct net_device *upper;
6306	struct list_head *iter;
6307	u32 link_speed = adapter->link_speed;
6308	bool flow_rx, flow_tx;
6309
6310	/* only continue if link was previously down */
6311	if (netif_carrier_ok(netdev))
6312		return;
6313
6314	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6315
6316	switch (hw->mac.type) {
6317	case ixgbe_mac_82598EB: {
6318		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6319		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6320		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6321		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6322	}
6323		break;
6324	case ixgbe_mac_X540:
6325	case ixgbe_mac_X550:
6326	case ixgbe_mac_X550EM_x:
6327	case ixgbe_mac_82599EB: {
6328		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6329		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6330		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6331		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6332	}
6333		break;
6334	default:
6335		flow_tx = false;
6336		flow_rx = false;
6337		break;
6338	}
6339
6340	adapter->last_rx_ptp_check = jiffies;
6341
6342	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6343		ixgbe_ptp_start_cyclecounter(adapter);
6344
6345	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
6346	       (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
6347	       "10 Gbps" :
6348	       (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
6349	       "1 Gbps" :
6350	       (link_speed == IXGBE_LINK_SPEED_100_FULL ?
6351	       "100 Mbps" :
6352	       "unknown speed"))),
6353	       ((flow_rx && flow_tx) ? "RX/TX" :
6354	       (flow_rx ? "RX" :
6355	       (flow_tx ? "TX" : "None"))));
6356
6357	netif_carrier_on(netdev);
6358	ixgbe_check_vf_rate_limit(adapter);
6359
6360	/* enable transmits */
6361	netif_tx_wake_all_queues(adapter->netdev);
6362
6363	/* enable any upper devices */
6364	rtnl_lock();
6365	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
6366		if (netif_is_macvlan(upper)) {
6367			struct macvlan_dev *vlan = netdev_priv(upper);
6368
6369			if (vlan->fwd_priv)
6370				netif_tx_wake_all_queues(upper);
6371		}
6372	}
6373	rtnl_unlock();
6374
6375	/* update the default user priority for VFs */
6376	ixgbe_update_default_up(adapter);
6377
6378	/* ping all the active vfs to let them know link has changed */
6379	ixgbe_ping_all_vfs(adapter);
6380}
6381
6382/**
6383 * ixgbe_watchdog_link_is_down - update netif_carrier status and
6384 *                               print link down message
6385 * @adapter: pointer to the adapter structure
6386 **/
6387static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6388{
6389	struct net_device *netdev = adapter->netdev;
6390	struct ixgbe_hw *hw = &adapter->hw;
6391
6392	adapter->link_up = false;
6393	adapter->link_speed = 0;
6394
6395	/* only continue if link was up previously */
6396	if (!netif_carrier_ok(netdev))
6397		return;
6398
6399	/* poll for SFP+ cable when link is down */
6400	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6401		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6402
6403	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6404		ixgbe_ptp_start_cyclecounter(adapter);
6405
6406	e_info(drv, "NIC Link is Down\n");
6407	netif_carrier_off(netdev);
6408
6409	/* ping all the active vfs to let them know link has changed */
6410	ixgbe_ping_all_vfs(adapter);
6411}
6412
6413static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
6414{
6415	int i;
6416
6417	for (i = 0; i < adapter->num_tx_queues; i++) {
6418		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6419
6420		if (tx_ring->next_to_use != tx_ring->next_to_clean)
6421			return true;
6422	}
6423
6424	return false;
6425}
6426
6427static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
6428{
6429	struct ixgbe_hw *hw = &adapter->hw;
6430	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
6431	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
6432
6433	int i, j;
6434
6435	if (!adapter->num_vfs)
6436		return false;
6437
6438	/* resetting the PF is only needed for MAC before X550 */
6439	if (hw->mac.type >= ixgbe_mac_X550)
6440		return false;
6441
6442	for (i = 0; i < adapter->num_vfs; i++) {
6443		for (j = 0; j < q_per_pool; j++) {
6444			u32 h, t;
6445
6446			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
6447			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
6448
6449			if (h != t)
6450				return true;
6451		}
6452	}
6453
6454	return false;
6455}
6456
6457/**
6458 * ixgbe_watchdog_flush_tx - flush queues on link down
6459 * @adapter: pointer to the device adapter structure
6460 **/
6461static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6462{
6463	if (!netif_carrier_ok(adapter->netdev)) {
6464		if (ixgbe_ring_tx_pending(adapter) ||
6465		    ixgbe_vf_tx_pending(adapter)) {
6466			/* We've lost link, so the controller stops DMA,
6467			 * but we've got queued Tx work that's never going
6468			 * to get done, so reset controller to flush Tx.
6469			 * (Do the reset outside of interrupt context).
6470			 */
6471			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6472			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
6473		}
6474	}
6475}
6476
6477#ifdef CONFIG_PCI_IOV
6478static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
6479				      struct pci_dev *vfdev)
6480{
6481	if (!pci_wait_for_pending_transaction(vfdev))
6482		e_dev_warn("Issuing VFLR with pending transactions\n");
6483
6484	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
6485	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
6486
6487	msleep(100);
6488}
6489
6490static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6491{
6492	struct ixgbe_hw *hw = &adapter->hw;
6493	struct pci_dev *pdev = adapter->pdev;
6494	struct pci_dev *vfdev;
6495	u32 gpc;
6496	int pos;
6497	unsigned short vf_id;
6498
6499	if (!(netif_carrier_ok(adapter->netdev)))
6500		return;
6501
6502	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6503	if (gpc) /* If incrementing then no need for the check below */
6504		return;
6505	/* Check to see if a bad DMA write target from an errant or
6506	 * malicious VF has caused a PCIe error.  If so then we can
6507	 * issue a VFLR to the offending VF(s) and then resume without
6508	 * requesting a full slot reset.
6509	 */
6510
6511	if (!pdev)
6512		return;
6513
6514	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
6515	if (!pos)
6516		return;
6517
6518	/* get the device ID for the VF */
6519	pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_id);
6520
6521	/* check status reg for all VFs owned by this PF */
6522	vfdev = pci_get_device(pdev->vendor, vf_id, NULL);
6523	while (vfdev) {
6524		if (vfdev->is_virtfn && (vfdev->physfn == pdev)) {
6525			u16 status_reg;
6526
6527			pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
6528			if (status_reg & PCI_STATUS_REC_MASTER_ABORT)
6529				/* issue VFLR */
6530				ixgbe_issue_vf_flr(adapter, vfdev);
6531		}
6532
6533		vfdev = pci_get_device(pdev->vendor, vf_id, vfdev);
6534	}
6535}
6536
6537static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6538{
6539	u32 ssvpc;
6540
6541	/* Do not perform spoof check for 82598 or if not in IOV mode */
6542	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
6543	    adapter->num_vfs == 0)
6544		return;
6545
6546	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
6547
6548	/*
6549	 * ssvpc register is cleared on read, if zero then no
6550	 * spoofed packets in the last interval.
6551	 */
6552	if (!ssvpc)
6553		return;
6554
6555	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
6556}
6557#else
6558static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
6559{
6560}
6561
6562static void
6563ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
6564{
6565}
6566#endif /* CONFIG_PCI_IOV */
6567
6568
6569/**
6570 * ixgbe_watchdog_subtask - check and bring link up
6571 * @adapter: pointer to the device adapter structure
6572 **/
6573static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
6574{
6575	/* if interface is down, removing or resetting, do nothing */
6576	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6577	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6578	    test_bit(__IXGBE_RESETTING, &adapter->state))
6579		return;
6580
6581	ixgbe_watchdog_update_link(adapter);
6582
6583	if (adapter->link_up)
6584		ixgbe_watchdog_link_is_up(adapter);
6585	else
6586		ixgbe_watchdog_link_is_down(adapter);
6587
6588	ixgbe_check_for_bad_vf(adapter);
6589	ixgbe_spoof_check(adapter);
6590	ixgbe_update_stats(adapter);
6591
6592	ixgbe_watchdog_flush_tx(adapter);
6593}
6594
6595/**
6596 * ixgbe_sfp_detection_subtask - poll for SFP+ cable
6597 * @adapter: the ixgbe adapter structure
6598 **/
6599static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
6600{
6601	struct ixgbe_hw *hw = &adapter->hw;
6602	s32 err;
6603
6604	/* not searching for SFP so there is nothing to do here */
6605	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
6606	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6607		return;
6608
6609	/* someone else is in init, wait until next service event */
6610	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6611		return;
6612
6613	err = hw->phy.ops.identify_sfp(hw);
6614	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6615		goto sfp_out;
6616
6617	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
6618		/* If no cable is present, then we need to reset
6619		 * the next time we find a good cable. */
6620		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
6621	}
6622
6623	/* exit on error */
6624	if (err)
6625		goto sfp_out;
6626
6627	/* exit if reset not needed */
6628	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6629		goto sfp_out;
6630
6631	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
6632
6633	/*
6634	 * A module may be identified correctly, but the EEPROM may not have
6635	 * support for that module.  setup_sfp() will fail in that case, so
6636	 * we should not allow that module to load.
6637	 */
6638	if (hw->mac.type == ixgbe_mac_82598EB)
6639		err = hw->phy.ops.reset(hw);
6640	else
6641		err = hw->mac.ops.setup_sfp(hw);
6642
6643	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6644		goto sfp_out;
6645
6646	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
6647	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
6648
6649sfp_out:
6650	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6651
6652	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
6653	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
6654		e_dev_err("failed to initialize because an unsupported "
6655			  "SFP+ module type was detected.\n");
6656		e_dev_err("Reload the driver after installing a "
6657			  "supported module.\n");
6658		unregister_netdev(adapter->netdev);
6659	}
6660}
6661
6662/**
6663 * ixgbe_sfp_link_config_subtask - set up link SFP after module install
6664 * @adapter: the ixgbe adapter structure
6665 **/
6666static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
6667{
6668	struct ixgbe_hw *hw = &adapter->hw;
6669	u32 speed;
6670	bool autoneg = false;
6671
6672	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
6673		return;
6674
6675	/* someone else is in init, wait until next service event */
6676	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6677		return;
6678
6679	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
6680
6681	speed = hw->phy.autoneg_advertised;
6682	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
6683		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
6684
6685		/* setup the highest link when no autoneg */
6686		if (!autoneg) {
6687			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
6688				speed = IXGBE_LINK_SPEED_10GB_FULL;
6689		}
6690	}
6691
6692	if (hw->mac.ops.setup_link)
6693		hw->mac.ops.setup_link(hw, speed, true);
6694
6695	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
6696	adapter->link_check_timeout = jiffies;
6697	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6698}
6699
6700/**
6701 * ixgbe_service_timer - Timer Call-back
6702 * @data: pointer to adapter cast into an unsigned long
6703 **/
6704static void ixgbe_service_timer(unsigned long data)
6705{
6706	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
6707	unsigned long next_event_offset;
6708
6709	/* poll faster when waiting for link */
6710	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
6711		next_event_offset = HZ / 10;
6712	else
6713		next_event_offset = HZ * 2;
6714
6715	/* Reset the timer */
6716	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
6717
6718	ixgbe_service_event_schedule(adapter);
6719}
6720
6721static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
6722{
6723	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
6724		return;
6725
6726	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
6727
6728	/* If we're already down, removing or resetting, just bail */
6729	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6730	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6731	    test_bit(__IXGBE_RESETTING, &adapter->state))
6732		return;
6733
6734	ixgbe_dump(adapter);
6735	netdev_err(adapter->netdev, "Reset adapter\n");
6736	adapter->tx_timeout_count++;
6737
6738	rtnl_lock();
6739	ixgbe_reinit_locked(adapter);
6740	rtnl_unlock();
6741}
6742
6743/**
6744 * ixgbe_service_task - manages and runs subtasks
6745 * @work: pointer to work_struct containing our data
6746 **/
6747static void ixgbe_service_task(struct work_struct *work)
6748{
6749	struct ixgbe_adapter *adapter = container_of(work,
6750						     struct ixgbe_adapter,
6751						     service_task);
6752	if (ixgbe_removed(adapter->hw.hw_addr)) {
6753		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
6754			rtnl_lock();
6755			ixgbe_down(adapter);
6756			rtnl_unlock();
6757		}
6758		ixgbe_service_event_complete(adapter);
6759		return;
6760	}
6761	ixgbe_reset_subtask(adapter);
6762	ixgbe_sfp_detection_subtask(adapter);
6763	ixgbe_sfp_link_config_subtask(adapter);
6764	ixgbe_check_overtemp_subtask(adapter);
6765	ixgbe_watchdog_subtask(adapter);
6766	ixgbe_fdir_reinit_subtask(adapter);
6767	ixgbe_check_hang_subtask(adapter);
6768
6769	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
6770		ixgbe_ptp_overflow_check(adapter);
6771		ixgbe_ptp_rx_hang(adapter);
6772	}
6773
6774	ixgbe_service_event_complete(adapter);
6775}
6776
6777static int ixgbe_tso(struct ixgbe_ring *tx_ring,
6778		     struct ixgbe_tx_buffer *first,
6779		     u8 *hdr_len)
6780{
6781	struct sk_buff *skb = first->skb;
6782	u32 vlan_macip_lens, type_tucmd;
6783	u32 mss_l4len_idx, l4len;
6784	int err;
6785
6786	if (skb->ip_summed != CHECKSUM_PARTIAL)
6787		return 0;
6788
6789	if (!skb_is_gso(skb))
6790		return 0;
6791
6792	err = skb_cow_head(skb, 0);
6793	if (err < 0)
6794		return err;
6795
6796	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6797	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
6798
6799	if (first->protocol == htons(ETH_P_IP)) {
6800		struct iphdr *iph = ip_hdr(skb);
6801		iph->tot_len = 0;
6802		iph->check = 0;
6803		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
6804							 iph->daddr, 0,
6805							 IPPROTO_TCP,
6806							 0);
6807		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6808		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6809				   IXGBE_TX_FLAGS_CSUM |
6810				   IXGBE_TX_FLAGS_IPV4;
6811	} else if (skb_is_gso_v6(skb)) {
6812		ipv6_hdr(skb)->payload_len = 0;
6813		tcp_hdr(skb)->check =
6814		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
6815				     &ipv6_hdr(skb)->daddr,
6816				     0, IPPROTO_TCP, 0);
6817		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6818				   IXGBE_TX_FLAGS_CSUM;
6819	}
6820
6821	/* compute header lengths */
6822	l4len = tcp_hdrlen(skb);
6823	*hdr_len = skb_transport_offset(skb) + l4len;
6824
6825	/* update gso size and bytecount with header size */
6826	first->gso_segs = skb_shinfo(skb)->gso_segs;
6827	first->bytecount += (first->gso_segs - 1) * *hdr_len;
6828
6829	/* mss_l4len_id: use 0 as index for TSO */
6830	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
6831	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
6832
6833	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6834	vlan_macip_lens = skb_network_header_len(skb);
6835	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6836	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6837
6838	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
6839			  mss_l4len_idx);
6840
6841	return 1;
6842}
6843
6844static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
6845			  struct ixgbe_tx_buffer *first)
6846{
6847	struct sk_buff *skb = first->skb;
6848	u32 vlan_macip_lens = 0;
6849	u32 mss_l4len_idx = 0;
6850	u32 type_tucmd = 0;
6851
6852	if (skb->ip_summed != CHECKSUM_PARTIAL) {
6853		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN) &&
6854		    !(first->tx_flags & IXGBE_TX_FLAGS_CC))
6855			return;
6856	} else {
6857		u8 l4_hdr = 0;
6858		switch (first->protocol) {
6859		case htons(ETH_P_IP):
6860			vlan_macip_lens |= skb_network_header_len(skb);
6861			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6862			l4_hdr = ip_hdr(skb)->protocol;
6863			break;
6864		case htons(ETH_P_IPV6):
6865			vlan_macip_lens |= skb_network_header_len(skb);
6866			l4_hdr = ipv6_hdr(skb)->nexthdr;
6867			break;
6868		default:
6869			if (unlikely(net_ratelimit())) {
6870				dev_warn(tx_ring->dev,
6871				 "partial checksum but proto=%x!\n",
6872				 first->protocol);
6873			}
6874			break;
6875		}
6876
6877		switch (l4_hdr) {
6878		case IPPROTO_TCP:
6879			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
6880			mss_l4len_idx = tcp_hdrlen(skb) <<
6881					IXGBE_ADVTXD_L4LEN_SHIFT;
6882			break;
6883		case IPPROTO_SCTP:
6884			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
6885			mss_l4len_idx = sizeof(struct sctphdr) <<
6886					IXGBE_ADVTXD_L4LEN_SHIFT;
6887			break;
6888		case IPPROTO_UDP:
6889			mss_l4len_idx = sizeof(struct udphdr) <<
6890					IXGBE_ADVTXD_L4LEN_SHIFT;
6891			break;
6892		default:
6893			if (unlikely(net_ratelimit())) {
6894				dev_warn(tx_ring->dev,
6895				 "partial checksum but l4 proto=%x!\n",
6896				 l4_hdr);
6897			}
6898			break;
6899		}
6900
6901		/* update TX checksum flag */
6902		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
6903	}
6904
6905	/* vlan_macip_lens: MACLEN, VLAN tag */
6906	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6907	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6908
6909	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
6910			  type_tucmd, mss_l4len_idx);
6911}
6912
6913#define IXGBE_SET_FLAG(_input, _flag, _result) \
6914	((_flag <= _result) ? \
6915	 ((u32)(_input & _flag) * (_result / _flag)) : \
6916	 ((u32)(_input & _flag) / (_flag / _result)))
6917
6918static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
6919{
6920	/* set type for advanced descriptor with frame checksum insertion */
6921	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
6922		       IXGBE_ADVTXD_DCMD_DEXT |
6923		       IXGBE_ADVTXD_DCMD_IFCS;
6924
6925	/* set HW vlan bit if vlan is present */
6926	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
6927				   IXGBE_ADVTXD_DCMD_VLE);
6928
6929	/* set segmentation enable bits for TSO/FSO */
6930	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
6931				   IXGBE_ADVTXD_DCMD_TSE);
6932
6933	/* set timestamp bit if present */
6934	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
6935				   IXGBE_ADVTXD_MAC_TSTAMP);
6936
6937	/* insert frame checksum */
6938	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
6939
6940	return cmd_type;
6941}
6942
6943static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
6944				   u32 tx_flags, unsigned int paylen)
6945{
6946	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
6947
6948	/* enable L4 checksum for TSO and TX checksum offload */
6949	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6950					IXGBE_TX_FLAGS_CSUM,
6951					IXGBE_ADVTXD_POPTS_TXSM);
6952
6953	/* enble IPv4 checksum for TSO */
6954	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6955					IXGBE_TX_FLAGS_IPV4,
6956					IXGBE_ADVTXD_POPTS_IXSM);
6957
6958	/*
6959	 * Check Context must be set if Tx switch is enabled, which it
6960	 * always is for case where virtual functions are running
6961	 */
6962	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6963					IXGBE_TX_FLAGS_CC,
6964					IXGBE_ADVTXD_CC);
6965
6966	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6967}
6968
6969static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6970{
6971	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
6972
6973	/* Herbert's original patch had:
6974	 *  smp_mb__after_netif_stop_queue();
6975	 * but since that doesn't exist yet, just open code it.
6976	 */
6977	smp_mb();
6978
6979	/* We need to check again in a case another CPU has just
6980	 * made room available.
6981	 */
6982	if (likely(ixgbe_desc_unused(tx_ring) < size))
6983		return -EBUSY;
6984
6985	/* A reprieve! - use start_queue because it doesn't call schedule */
6986	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
6987	++tx_ring->tx_stats.restart_queue;
6988	return 0;
6989}
6990
6991static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6992{
6993	if (likely(ixgbe_desc_unused(tx_ring) >= size))
6994		return 0;
6995
6996	return __ixgbe_maybe_stop_tx(tx_ring, size);
6997}
6998
6999#define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7000		       IXGBE_TXD_CMD_RS)
7001
7002static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7003			 struct ixgbe_tx_buffer *first,
7004			 const u8 hdr_len)
7005{
7006	struct sk_buff *skb = first->skb;
7007	struct ixgbe_tx_buffer *tx_buffer;
7008	union ixgbe_adv_tx_desc *tx_desc;
7009	struct skb_frag_struct *frag;
7010	dma_addr_t dma;
7011	unsigned int data_len, size;
7012	u32 tx_flags = first->tx_flags;
7013	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7014	u16 i = tx_ring->next_to_use;
7015
7016	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7017
7018	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7019
7020	size = skb_headlen(skb);
7021	data_len = skb->data_len;
7022
7023#ifdef IXGBE_FCOE
7024	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7025		if (data_len < sizeof(struct fcoe_crc_eof)) {
7026			size -= sizeof(struct fcoe_crc_eof) - data_len;
7027			data_len = 0;
7028		} else {
7029			data_len -= sizeof(struct fcoe_crc_eof);
7030		}
7031	}
7032
7033#endif
7034	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7035
7036	tx_buffer = first;
7037
7038	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7039		if (dma_mapping_error(tx_ring->dev, dma))
7040			goto dma_error;
7041
7042		/* record length, and DMA address */
7043		dma_unmap_len_set(tx_buffer, len, size);
7044		dma_unmap_addr_set(tx_buffer, dma, dma);
7045
7046		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7047
7048		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7049			tx_desc->read.cmd_type_len =
7050				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7051
7052			i++;
7053			tx_desc++;
7054			if (i == tx_ring->count) {
7055				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7056				i = 0;
7057			}
7058			tx_desc->read.olinfo_status = 0;
7059
7060			dma += IXGBE_MAX_DATA_PER_TXD;
7061			size -= IXGBE_MAX_DATA_PER_TXD;
7062
7063			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7064		}
7065
7066		if (likely(!data_len))
7067			break;
7068
7069		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7070
7071		i++;
7072		tx_desc++;
7073		if (i == tx_ring->count) {
7074			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7075			i = 0;
7076		}
7077		tx_desc->read.olinfo_status = 0;
7078
7079#ifdef IXGBE_FCOE
7080		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7081#else
7082		size = skb_frag_size(frag);
7083#endif
7084		data_len -= size;
7085
7086		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7087				       DMA_TO_DEVICE);
7088
7089		tx_buffer = &tx_ring->tx_buffer_info[i];
7090	}
7091
7092	/* write last descriptor with RS and EOP bits */
7093	cmd_type |= size | IXGBE_TXD_CMD;
7094	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7095
7096	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7097
7098	/* set the timestamp */
7099	first->time_stamp = jiffies;
7100
7101	/*
7102	 * Force memory writes to complete before letting h/w know there
7103	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7104	 * memory model archs, such as IA-64).
7105	 *
7106	 * We also need this memory barrier to make certain all of the
7107	 * status bits have been updated before next_to_watch is written.
7108	 */
7109	wmb();
7110
7111	/* set next_to_watch value indicating a packet is present */
7112	first->next_to_watch = tx_desc;
7113
7114	i++;
7115	if (i == tx_ring->count)
7116		i = 0;
7117
7118	tx_ring->next_to_use = i;
7119
7120	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7121
7122	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7123		writel(i, tx_ring->tail);
7124
7125		/* we need this if more than one processor can write to our tail
7126		 * at a time, it synchronizes IO on IA64/Altix systems
7127		 */
7128		mmiowb();
7129	}
7130
7131	return;
7132dma_error:
7133	dev_err(tx_ring->dev, "TX DMA map failed\n");
7134
7135	/* clear dma mappings for failed tx_buffer_info map */
7136	for (;;) {
7137		tx_buffer = &tx_ring->tx_buffer_info[i];
7138		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
7139		if (tx_buffer == first)
7140			break;
7141		if (i == 0)
7142			i = tx_ring->count;
7143		i--;
7144	}
7145
7146	tx_ring->next_to_use = i;
7147}
7148
7149static void ixgbe_atr(struct ixgbe_ring *ring,
7150		      struct ixgbe_tx_buffer *first)
7151{
7152	struct ixgbe_q_vector *q_vector = ring->q_vector;
7153	union ixgbe_atr_hash_dword input = { .dword = 0 };
7154	union ixgbe_atr_hash_dword common = { .dword = 0 };
7155	union {
7156		unsigned char *network;
7157		struct iphdr *ipv4;
7158		struct ipv6hdr *ipv6;
7159	} hdr;
7160	struct tcphdr *th;
7161	__be16 vlan_id;
7162
7163	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7164	if (!q_vector)
7165		return;
7166
7167	/* do nothing if sampling is disabled */
7168	if (!ring->atr_sample_rate)
7169		return;
7170
7171	ring->atr_count++;
7172
7173	/* snag network header to get L4 type and address */
7174	hdr.network = skb_network_header(first->skb);
7175
7176	/* Currently only IPv4/IPv6 with TCP is supported */
7177	if ((first->protocol != htons(ETH_P_IPV6) ||
7178	     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
7179	    (first->protocol != htons(ETH_P_IP) ||
7180	     hdr.ipv4->protocol != IPPROTO_TCP))
7181		return;
7182
7183	th = tcp_hdr(first->skb);
7184
7185	/* skip this packet since it is invalid or the socket is closing */
7186	if (!th || th->fin)
7187		return;
7188
7189	/* sample on all syn packets or once every atr sample count */
7190	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7191		return;
7192
7193	/* reset sample count */
7194	ring->atr_count = 0;
7195
7196	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7197
7198	/*
7199	 * src and dst are inverted, think how the receiver sees them
7200	 *
7201	 * The input is broken into two sections, a non-compressed section
7202	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7203	 * is XORed together and stored in the compressed dword.
7204	 */
7205	input.formatted.vlan_id = vlan_id;
7206
7207	/*
7208	 * since src port and flex bytes occupy the same word XOR them together
7209	 * and write the value to source port portion of compressed dword
7210	 */
7211	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7212		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7213	else
7214		common.port.src ^= th->dest ^ first->protocol;
7215	common.port.dst ^= th->source;
7216
7217	if (first->protocol == htons(ETH_P_IP)) {
7218		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7219		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7220	} else {
7221		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7222		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7223			     hdr.ipv6->saddr.s6_addr32[1] ^
7224			     hdr.ipv6->saddr.s6_addr32[2] ^
7225			     hdr.ipv6->saddr.s6_addr32[3] ^
7226			     hdr.ipv6->daddr.s6_addr32[0] ^
7227			     hdr.ipv6->daddr.s6_addr32[1] ^
7228			     hdr.ipv6->daddr.s6_addr32[2] ^
7229			     hdr.ipv6->daddr.s6_addr32[3];
7230	}
7231
7232	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7233	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7234					      input, common, ring->queue_index);
7235}
7236
7237static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7238			      void *accel_priv, select_queue_fallback_t fallback)
7239{
7240	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7241#ifdef IXGBE_FCOE
7242	struct ixgbe_adapter *adapter;
7243	struct ixgbe_ring_feature *f;
7244	int txq;
7245#endif
7246
7247	if (fwd_adapter)
7248		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7249
7250#ifdef IXGBE_FCOE
7251
7252	/*
7253	 * only execute the code below if protocol is FCoE
7254	 * or FIP and we have FCoE enabled on the adapter
7255	 */
7256	switch (vlan_get_protocol(skb)) {
7257	case htons(ETH_P_FCOE):
7258	case htons(ETH_P_FIP):
7259		adapter = netdev_priv(dev);
7260
7261		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7262			break;
7263	default:
7264		return fallback(dev, skb);
7265	}
7266
7267	f = &adapter->ring_feature[RING_F_FCOE];
7268
7269	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7270					   smp_processor_id();
7271
7272	while (txq >= f->indices)
7273		txq -= f->indices;
7274
7275	return txq + f->offset;
7276#else
7277	return fallback(dev, skb);
7278#endif
7279}
7280
7281netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7282			  struct ixgbe_adapter *adapter,
7283			  struct ixgbe_ring *tx_ring)
7284{
7285	struct ixgbe_tx_buffer *first;
7286	int tso;
7287	u32 tx_flags = 0;
7288	unsigned short f;
7289	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7290	__be16 protocol = skb->protocol;
7291	u8 hdr_len = 0;
7292
7293	/*
7294	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7295	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7296	 *       + 2 desc gap to keep tail from touching head,
7297	 *       + 1 desc for context descriptor,
7298	 * otherwise try next time
7299	 */
7300	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7301		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7302
7303	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7304		tx_ring->tx_stats.tx_busy++;
7305		return NETDEV_TX_BUSY;
7306	}
7307
7308	/* record the location of the first descriptor for this packet */
7309	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7310	first->skb = skb;
7311	first->bytecount = skb->len;
7312	first->gso_segs = 1;
7313
7314	/* if we have a HW VLAN tag being added default to the HW one */
7315	if (skb_vlan_tag_present(skb)) {
7316		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7317		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7318	/* else if it is a SW VLAN check the next protocol and store the tag */
7319	} else if (protocol == htons(ETH_P_8021Q)) {
7320		struct vlan_hdr *vhdr, _vhdr;
7321		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7322		if (!vhdr)
7323			goto out_drop;
7324
7325		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7326				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7327		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7328	}
7329	protocol = vlan_get_protocol(skb);
7330
7331	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
7332	    adapter->ptp_clock &&
7333	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7334				   &adapter->state)) {
7335		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7336		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7337
7338		/* schedule check for Tx timestamp */
7339		adapter->ptp_tx_skb = skb_get(skb);
7340		adapter->ptp_tx_start = jiffies;
7341		schedule_work(&adapter->ptp_tx_work);
7342	}
7343
7344	skb_tx_timestamp(skb);
7345
7346#ifdef CONFIG_PCI_IOV
7347	/*
7348	 * Use the l2switch_enable flag - would be false if the DMA
7349	 * Tx switch had been disabled.
7350	 */
7351	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7352		tx_flags |= IXGBE_TX_FLAGS_CC;
7353
7354#endif
7355	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7356	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7357	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7358	     (skb->priority != TC_PRIO_CONTROL))) {
7359		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7360		tx_flags |= (skb->priority & 0x7) <<
7361					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7362		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7363			struct vlan_ethhdr *vhdr;
7364
7365			if (skb_cow_head(skb, 0))
7366				goto out_drop;
7367			vhdr = (struct vlan_ethhdr *)skb->data;
7368			vhdr->h_vlan_TCI = htons(tx_flags >>
7369						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7370		} else {
7371			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7372		}
7373	}
7374
7375	/* record initial flags and protocol */
7376	first->tx_flags = tx_flags;
7377	first->protocol = protocol;
7378
7379#ifdef IXGBE_FCOE
7380	/* setup tx offload for FCoE */
7381	if ((protocol == htons(ETH_P_FCOE)) &&
7382	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7383		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7384		if (tso < 0)
7385			goto out_drop;
7386
7387		goto xmit_fcoe;
7388	}
7389
7390#endif /* IXGBE_FCOE */
7391	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7392	if (tso < 0)
7393		goto out_drop;
7394	else if (!tso)
7395		ixgbe_tx_csum(tx_ring, first);
7396
7397	/* add the ATR filter if ATR is on */
7398	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7399		ixgbe_atr(tx_ring, first);
7400
7401#ifdef IXGBE_FCOE
7402xmit_fcoe:
7403#endif /* IXGBE_FCOE */
7404	ixgbe_tx_map(tx_ring, first, hdr_len);
7405
7406	return NETDEV_TX_OK;
7407
7408out_drop:
7409	dev_kfree_skb_any(first->skb);
7410	first->skb = NULL;
7411
7412	return NETDEV_TX_OK;
7413}
7414
7415static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7416				      struct net_device *netdev,
7417				      struct ixgbe_ring *ring)
7418{
7419	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7420	struct ixgbe_ring *tx_ring;
7421
7422	/*
7423	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7424	 * in order to meet this minimum size requirement.
7425	 */
7426	if (skb_put_padto(skb, 17))
7427		return NETDEV_TX_OK;
7428
7429	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7430
7431	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7432}
7433
7434static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7435				    struct net_device *netdev)
7436{
7437	return __ixgbe_xmit_frame(skb, netdev, NULL);
7438}
7439
7440/**
7441 * ixgbe_set_mac - Change the Ethernet Address of the NIC
7442 * @netdev: network interface device structure
7443 * @p: pointer to an address structure
7444 *
7445 * Returns 0 on success, negative on failure
7446 **/
7447static int ixgbe_set_mac(struct net_device *netdev, void *p)
7448{
7449	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7450	struct ixgbe_hw *hw = &adapter->hw;
7451	struct sockaddr *addr = p;
7452	int ret;
7453
7454	if (!is_valid_ether_addr(addr->sa_data))
7455		return -EADDRNOTAVAIL;
7456
7457	ixgbe_del_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7458	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
7459	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
7460
7461	ret = ixgbe_add_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7462	return ret > 0 ? 0 : ret;
7463}
7464
7465static int
7466ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
7467{
7468	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7469	struct ixgbe_hw *hw = &adapter->hw;
7470	u16 value;
7471	int rc;
7472
7473	if (prtad != hw->phy.mdio.prtad)
7474		return -EINVAL;
7475	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
7476	if (!rc)
7477		rc = value;
7478	return rc;
7479}
7480
7481static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
7482			    u16 addr, u16 value)
7483{
7484	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7485	struct ixgbe_hw *hw = &adapter->hw;
7486
7487	if (prtad != hw->phy.mdio.prtad)
7488		return -EINVAL;
7489	return hw->phy.ops.write_reg(hw, addr, devad, value);
7490}
7491
7492static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
7493{
7494	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7495
7496	switch (cmd) {
7497	case SIOCSHWTSTAMP:
7498		return ixgbe_ptp_set_ts_config(adapter, req);
7499	case SIOCGHWTSTAMP:
7500		return ixgbe_ptp_get_ts_config(adapter, req);
7501	default:
7502		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
7503	}
7504}
7505
7506/**
7507 * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
7508 * netdev->dev_addrs
7509 * @netdev: network interface device structure
7510 *
7511 * Returns non-zero on failure
7512 **/
7513static int ixgbe_add_sanmac_netdev(struct net_device *dev)
7514{
7515	int err = 0;
7516	struct ixgbe_adapter *adapter = netdev_priv(dev);
7517	struct ixgbe_hw *hw = &adapter->hw;
7518
7519	if (is_valid_ether_addr(hw->mac.san_addr)) {
7520		rtnl_lock();
7521		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
7522		rtnl_unlock();
7523
7524		/* update SAN MAC vmdq pool selection */
7525		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
7526	}
7527	return err;
7528}
7529
7530/**
7531 * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
7532 * netdev->dev_addrs
7533 * @netdev: network interface device structure
7534 *
7535 * Returns non-zero on failure
7536 **/
7537static int ixgbe_del_sanmac_netdev(struct net_device *dev)
7538{
7539	int err = 0;
7540	struct ixgbe_adapter *adapter = netdev_priv(dev);
7541	struct ixgbe_mac_info *mac = &adapter->hw.mac;
7542
7543	if (is_valid_ether_addr(mac->san_addr)) {
7544		rtnl_lock();
7545		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
7546		rtnl_unlock();
7547	}
7548	return err;
7549}
7550
7551#ifdef CONFIG_NET_POLL_CONTROLLER
7552/*
7553 * Polling 'interrupt' - used by things like netconsole to send skbs
7554 * without having to re-enable interrupts. It's not called while
7555 * the interrupt routine is executing.
7556 */
7557static void ixgbe_netpoll(struct net_device *netdev)
7558{
7559	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7560	int i;
7561
7562	/* if interface is down do nothing */
7563	if (test_bit(__IXGBE_DOWN, &adapter->state))
7564		return;
7565
7566	/* loop through and schedule all active queues */
7567	for (i = 0; i < adapter->num_q_vectors; i++)
7568		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
7569}
7570
7571#endif
7572static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
7573						   struct rtnl_link_stats64 *stats)
7574{
7575	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7576	int i;
7577
7578	rcu_read_lock();
7579	for (i = 0; i < adapter->num_rx_queues; i++) {
7580		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
7581		u64 bytes, packets;
7582		unsigned int start;
7583
7584		if (ring) {
7585			do {
7586				start = u64_stats_fetch_begin_irq(&ring->syncp);
7587				packets = ring->stats.packets;
7588				bytes   = ring->stats.bytes;
7589			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7590			stats->rx_packets += packets;
7591			stats->rx_bytes   += bytes;
7592		}
7593	}
7594
7595	for (i = 0; i < adapter->num_tx_queues; i++) {
7596		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
7597		u64 bytes, packets;
7598		unsigned int start;
7599
7600		if (ring) {
7601			do {
7602				start = u64_stats_fetch_begin_irq(&ring->syncp);
7603				packets = ring->stats.packets;
7604				bytes   = ring->stats.bytes;
7605			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7606			stats->tx_packets += packets;
7607			stats->tx_bytes   += bytes;
7608		}
7609	}
7610	rcu_read_unlock();
7611	/* following stats updated by ixgbe_watchdog_task() */
7612	stats->multicast	= netdev->stats.multicast;
7613	stats->rx_errors	= netdev->stats.rx_errors;
7614	stats->rx_length_errors	= netdev->stats.rx_length_errors;
7615	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
7616	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
7617	return stats;
7618}
7619
7620#ifdef CONFIG_IXGBE_DCB
7621/**
7622 * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
7623 * @adapter: pointer to ixgbe_adapter
7624 * @tc: number of traffic classes currently enabled
7625 *
7626 * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
7627 * 802.1Q priority maps to a packet buffer that exists.
7628 */
7629static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
7630{
7631	struct ixgbe_hw *hw = &adapter->hw;
7632	u32 reg, rsave;
7633	int i;
7634
7635	/* 82598 have a static priority to TC mapping that can not
7636	 * be changed so no validation is needed.
7637	 */
7638	if (hw->mac.type == ixgbe_mac_82598EB)
7639		return;
7640
7641	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
7642	rsave = reg;
7643
7644	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
7645		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
7646
7647		/* If up2tc is out of bounds default to zero */
7648		if (up2tc > tc)
7649			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
7650	}
7651
7652	if (reg != rsave)
7653		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
7654
7655	return;
7656}
7657
7658/**
7659 * ixgbe_set_prio_tc_map - Configure netdev prio tc map
7660 * @adapter: Pointer to adapter struct
7661 *
7662 * Populate the netdev user priority to tc map
7663 */
7664static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
7665{
7666	struct net_device *dev = adapter->netdev;
7667	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
7668	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
7669	u8 prio;
7670
7671	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
7672		u8 tc = 0;
7673
7674		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
7675			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
7676		else if (ets)
7677			tc = ets->prio_tc[prio];
7678
7679		netdev_set_prio_tc_map(dev, prio, tc);
7680	}
7681}
7682
7683#endif /* CONFIG_IXGBE_DCB */
7684/**
7685 * ixgbe_setup_tc - configure net_device for multiple traffic classes
7686 *
7687 * @netdev: net device to configure
7688 * @tc: number of traffic classes to enable
7689 */
7690int ixgbe_setup_tc(struct net_device *dev, u8 tc)
7691{
7692	struct ixgbe_adapter *adapter = netdev_priv(dev);
7693	struct ixgbe_hw *hw = &adapter->hw;
7694	bool pools;
7695
7696	/* Hardware supports up to 8 traffic classes */
7697	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs ||
7698	    (hw->mac.type == ixgbe_mac_82598EB &&
7699	     tc < MAX_TRAFFIC_CLASS))
7700		return -EINVAL;
7701
7702	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
7703	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
7704		return -EBUSY;
7705
7706	/* Hardware has to reinitialize queues and interrupts to
7707	 * match packet buffer alignment. Unfortunately, the
7708	 * hardware is not flexible enough to do this dynamically.
7709	 */
7710	if (netif_running(dev))
7711		ixgbe_close(dev);
7712	ixgbe_clear_interrupt_scheme(adapter);
7713
7714#ifdef CONFIG_IXGBE_DCB
7715	if (tc) {
7716		netdev_set_num_tc(dev, tc);
7717		ixgbe_set_prio_tc_map(adapter);
7718
7719		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
7720
7721		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
7722			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
7723			adapter->hw.fc.requested_mode = ixgbe_fc_none;
7724		}
7725	} else {
7726		netdev_reset_tc(dev);
7727
7728		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
7729			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
7730
7731		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
7732
7733		adapter->temp_dcb_cfg.pfc_mode_enable = false;
7734		adapter->dcb_cfg.pfc_mode_enable = false;
7735	}
7736
7737	ixgbe_validate_rtr(adapter, tc);
7738
7739#endif /* CONFIG_IXGBE_DCB */
7740	ixgbe_init_interrupt_scheme(adapter);
7741
7742	if (netif_running(dev))
7743		return ixgbe_open(dev);
7744
7745	return 0;
7746}
7747
7748#ifdef CONFIG_PCI_IOV
7749void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
7750{
7751	struct net_device *netdev = adapter->netdev;
7752
7753	rtnl_lock();
7754	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
7755	rtnl_unlock();
7756}
7757
7758#endif
7759void ixgbe_do_reset(struct net_device *netdev)
7760{
7761	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7762
7763	if (netif_running(netdev))
7764		ixgbe_reinit_locked(adapter);
7765	else
7766		ixgbe_reset(adapter);
7767}
7768
7769static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
7770					    netdev_features_t features)
7771{
7772	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7773
7774	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
7775	if (!(features & NETIF_F_RXCSUM))
7776		features &= ~NETIF_F_LRO;
7777
7778	/* Turn off LRO if not RSC capable */
7779	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
7780		features &= ~NETIF_F_LRO;
7781
7782	return features;
7783}
7784
7785static int ixgbe_set_features(struct net_device *netdev,
7786			      netdev_features_t features)
7787{
7788	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7789	netdev_features_t changed = netdev->features ^ features;
7790	bool need_reset = false;
7791
7792	/* Make sure RSC matches LRO, reset if change */
7793	if (!(features & NETIF_F_LRO)) {
7794		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7795			need_reset = true;
7796		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
7797	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
7798		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
7799		if (adapter->rx_itr_setting == 1 ||
7800		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
7801			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
7802			need_reset = true;
7803		} else if ((changed ^ features) & NETIF_F_LRO) {
7804			e_info(probe, "rx-usecs set too low, "
7805			       "disabling RSC\n");
7806		}
7807	}
7808
7809	/*
7810	 * Check if Flow Director n-tuple support was enabled or disabled.  If
7811	 * the state changed, we need to reset.
7812	 */
7813	switch (features & NETIF_F_NTUPLE) {
7814	case NETIF_F_NTUPLE:
7815		/* turn off ATR, enable perfect filters and reset */
7816		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
7817			need_reset = true;
7818
7819		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
7820		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7821		break;
7822	default:
7823		/* turn off perfect filters, enable ATR and reset */
7824		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
7825			need_reset = true;
7826
7827		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7828
7829		/* We cannot enable ATR if SR-IOV is enabled */
7830		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7831			break;
7832
7833		/* We cannot enable ATR if we have 2 or more traffic classes */
7834		if (netdev_get_num_tc(netdev) > 1)
7835			break;
7836
7837		/* We cannot enable ATR if RSS is disabled */
7838		if (adapter->ring_feature[RING_F_RSS].limit <= 1)
7839			break;
7840
7841		/* A sample rate of 0 indicates ATR disabled */
7842		if (!adapter->atr_sample_rate)
7843			break;
7844
7845		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
7846		break;
7847	}
7848
7849	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7850		ixgbe_vlan_strip_enable(adapter);
7851	else
7852		ixgbe_vlan_strip_disable(adapter);
7853
7854	if (changed & NETIF_F_RXALL)
7855		need_reset = true;
7856
7857	netdev->features = features;
7858	if (need_reset)
7859		ixgbe_do_reset(netdev);
7860
7861	return 0;
7862}
7863
7864/**
7865 * ixgbe_add_vxlan_port - Get notifications about VXLAN ports that come up
7866 * @dev: The port's netdev
7867 * @sa_family: Socket Family that VXLAN is notifiying us about
7868 * @port: New UDP port number that VXLAN started listening to
7869 **/
7870static void ixgbe_add_vxlan_port(struct net_device *dev, sa_family_t sa_family,
7871				 __be16 port)
7872{
7873	struct ixgbe_adapter *adapter = netdev_priv(dev);
7874	struct ixgbe_hw *hw = &adapter->hw;
7875	u16 new_port = ntohs(port);
7876
7877	if (sa_family == AF_INET6)
7878		return;
7879
7880	if (adapter->vxlan_port == new_port) {
7881		netdev_info(dev, "Port %d already offloaded\n", new_port);
7882		return;
7883	}
7884
7885	if (adapter->vxlan_port) {
7886		netdev_info(dev,
7887			    "Hit Max num of UDP ports, not adding port %d\n",
7888			    new_port);
7889		return;
7890	}
7891
7892	adapter->vxlan_port = new_port;
7893	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, new_port);
7894}
7895
7896/**
7897 * ixgbe_del_vxlan_port - Get notifications about VXLAN ports that go away
7898 * @dev: The port's netdev
7899 * @sa_family: Socket Family that VXLAN is notifying us about
7900 * @port: UDP port number that VXLAN stopped listening to
7901 **/
7902static void ixgbe_del_vxlan_port(struct net_device *dev, sa_family_t sa_family,
7903				 __be16 port)
7904{
7905	struct ixgbe_adapter *adapter = netdev_priv(dev);
7906	struct ixgbe_hw *hw = &adapter->hw;
7907	u16 new_port = ntohs(port);
7908
7909	if (sa_family == AF_INET6)
7910		return;
7911
7912	if (adapter->vxlan_port != new_port) {
7913		netdev_info(dev, "Port %d was not found, not deleting\n",
7914			    new_port);
7915		return;
7916	}
7917
7918	adapter->vxlan_port = 0;
7919	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, 0);
7920}
7921
7922static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
7923			     struct net_device *dev,
7924			     const unsigned char *addr, u16 vid,
7925			     u16 flags)
7926{
7927	/* guarantee we can provide a unique filter for the unicast address */
7928	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
7929		if (IXGBE_MAX_PF_MACVLANS <= netdev_uc_count(dev))
7930			return -ENOMEM;
7931	}
7932
7933	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
7934}
7935
7936/**
7937 * ixgbe_configure_bridge_mode - set various bridge modes
7938 * @adapter - the private structure
7939 * @mode - requested bridge mode
7940 *
7941 * Configure some settings require for various bridge modes.
7942 **/
7943static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
7944				       __u16 mode)
7945{
7946	struct ixgbe_hw *hw = &adapter->hw;
7947	unsigned int p, num_pools;
7948	u32 vmdctl;
7949
7950	switch (mode) {
7951	case BRIDGE_MODE_VEPA:
7952		/* disable Tx loopback, rely on switch hairpin mode */
7953		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
7954
7955		/* must enable Rx switching replication to allow multicast
7956		 * packet reception on all VFs, and to enable source address
7957		 * pruning.
7958		 */
7959		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
7960		vmdctl |= IXGBE_VT_CTL_REPLEN;
7961		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
7962
7963		/* enable Rx source address pruning. Note, this requires
7964		 * replication to be enabled or else it does nothing.
7965		 */
7966		num_pools = adapter->num_vfs + adapter->num_rx_pools;
7967		for (p = 0; p < num_pools; p++) {
7968			if (hw->mac.ops.set_source_address_pruning)
7969				hw->mac.ops.set_source_address_pruning(hw,
7970								       true,
7971								       p);
7972		}
7973		break;
7974	case BRIDGE_MODE_VEB:
7975		/* enable Tx loopback for internal VF/PF communication */
7976		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
7977				IXGBE_PFDTXGSWC_VT_LBEN);
7978
7979		/* disable Rx switching replication unless we have SR-IOV
7980		 * virtual functions
7981		 */
7982		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
7983		if (!adapter->num_vfs)
7984			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
7985		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
7986
7987		/* disable Rx source address pruning, since we don't expect to
7988		 * be receiving external loopback of our transmitted frames.
7989		 */
7990		num_pools = adapter->num_vfs + adapter->num_rx_pools;
7991		for (p = 0; p < num_pools; p++) {
7992			if (hw->mac.ops.set_source_address_pruning)
7993				hw->mac.ops.set_source_address_pruning(hw,
7994								       false,
7995								       p);
7996		}
7997		break;
7998	default:
7999		return -EINVAL;
8000	}
8001
8002	adapter->bridge_mode = mode;
8003
8004	e_info(drv, "enabling bridge mode: %s\n",
8005	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
8006
8007	return 0;
8008}
8009
8010static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
8011				    struct nlmsghdr *nlh, u16 flags)
8012{
8013	struct ixgbe_adapter *adapter = netdev_priv(dev);
8014	struct nlattr *attr, *br_spec;
8015	int rem;
8016
8017	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
8018		return -EOPNOTSUPP;
8019
8020	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8021	if (!br_spec)
8022		return -EINVAL;
8023
8024	nla_for_each_nested(attr, br_spec, rem) {
8025		u32 status;
8026		__u16 mode;
8027
8028		if (nla_type(attr) != IFLA_BRIDGE_MODE)
8029			continue;
8030
8031		if (nla_len(attr) < sizeof(mode))
8032			return -EINVAL;
8033
8034		mode = nla_get_u16(attr);
8035		status = ixgbe_configure_bridge_mode(adapter, mode);
8036		if (status)
8037			return status;
8038
8039		break;
8040	}
8041
8042	return 0;
8043}
8044
8045static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8046				    struct net_device *dev,
8047				    u32 filter_mask, int nlflags)
8048{
8049	struct ixgbe_adapter *adapter = netdev_priv(dev);
8050
8051	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
8052		return 0;
8053
8054	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
8055				       adapter->bridge_mode, 0, 0, nlflags);
8056}
8057
8058static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
8059{
8060	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
8061	struct ixgbe_adapter *adapter = netdev_priv(pdev);
8062	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
8063	unsigned int limit;
8064	int pool, err;
8065
8066	/* Hardware has a limited number of available pools. Each VF, and the
8067	 * PF require a pool. Check to ensure we don't attempt to use more
8068	 * then the available number of pools.
8069	 */
8070	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
8071		return ERR_PTR(-EINVAL);
8072
8073#ifdef CONFIG_RPS
8074	if (vdev->num_rx_queues != vdev->num_tx_queues) {
8075		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
8076			    vdev->name);
8077		return ERR_PTR(-EINVAL);
8078	}
8079#endif
8080	/* Check for hardware restriction on number of rx/tx queues */
8081	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
8082	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
8083		netdev_info(pdev,
8084			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
8085			    pdev->name);
8086		return ERR_PTR(-EINVAL);
8087	}
8088
8089	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8090	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
8091	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
8092		return ERR_PTR(-EBUSY);
8093
8094	fwd_adapter = kcalloc(1, sizeof(struct ixgbe_fwd_adapter), GFP_KERNEL);
8095	if (!fwd_adapter)
8096		return ERR_PTR(-ENOMEM);
8097
8098	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
8099	adapter->num_rx_pools++;
8100	set_bit(pool, &adapter->fwd_bitmask);
8101	limit = find_last_bit(&adapter->fwd_bitmask, 32);
8102
8103	/* Enable VMDq flag so device will be set in VM mode */
8104	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
8105	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
8106	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
8107
8108	/* Force reinit of ring allocation with VMDQ enabled */
8109	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
8110	if (err)
8111		goto fwd_add_err;
8112	fwd_adapter->pool = pool;
8113	fwd_adapter->real_adapter = adapter;
8114	err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
8115	if (err)
8116		goto fwd_add_err;
8117	netif_tx_start_all_queues(vdev);
8118	return fwd_adapter;
8119fwd_add_err:
8120	/* unwind counter and free adapter struct */
8121	netdev_info(pdev,
8122		    "%s: dfwd hardware acceleration failed\n", vdev->name);
8123	clear_bit(pool, &adapter->fwd_bitmask);
8124	adapter->num_rx_pools--;
8125	kfree(fwd_adapter);
8126	return ERR_PTR(err);
8127}
8128
8129static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
8130{
8131	struct ixgbe_fwd_adapter *fwd_adapter = priv;
8132	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
8133	unsigned int limit;
8134
8135	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
8136	adapter->num_rx_pools--;
8137
8138	limit = find_last_bit(&adapter->fwd_bitmask, 32);
8139	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
8140	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
8141	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
8142	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
8143		   fwd_adapter->pool, adapter->num_rx_pools,
8144		   fwd_adapter->rx_base_queue,
8145		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
8146		   adapter->fwd_bitmask);
8147	kfree(fwd_adapter);
8148}
8149
8150static const struct net_device_ops ixgbe_netdev_ops = {
8151	.ndo_open		= ixgbe_open,
8152	.ndo_stop		= ixgbe_close,
8153	.ndo_start_xmit		= ixgbe_xmit_frame,
8154	.ndo_select_queue	= ixgbe_select_queue,
8155	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
8156	.ndo_validate_addr	= eth_validate_addr,
8157	.ndo_set_mac_address	= ixgbe_set_mac,
8158	.ndo_change_mtu		= ixgbe_change_mtu,
8159	.ndo_tx_timeout		= ixgbe_tx_timeout,
8160	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
8161	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
8162	.ndo_do_ioctl		= ixgbe_ioctl,
8163	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
8164	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
8165	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
8166	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
8167	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
8168	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
8169	.ndo_get_stats64	= ixgbe_get_stats64,
8170#ifdef CONFIG_IXGBE_DCB
8171	.ndo_setup_tc		= ixgbe_setup_tc,
8172#endif
8173#ifdef CONFIG_NET_POLL_CONTROLLER
8174	.ndo_poll_controller	= ixgbe_netpoll,
8175#endif
8176#ifdef CONFIG_NET_RX_BUSY_POLL
8177	.ndo_busy_poll		= ixgbe_low_latency_recv,
8178#endif
8179#ifdef IXGBE_FCOE
8180	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
8181	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
8182	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
8183	.ndo_fcoe_enable = ixgbe_fcoe_enable,
8184	.ndo_fcoe_disable = ixgbe_fcoe_disable,
8185	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
8186	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
8187#endif /* IXGBE_FCOE */
8188	.ndo_set_features = ixgbe_set_features,
8189	.ndo_fix_features = ixgbe_fix_features,
8190	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
8191	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
8192	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
8193	.ndo_dfwd_add_station	= ixgbe_fwd_add,
8194	.ndo_dfwd_del_station	= ixgbe_fwd_del,
8195	.ndo_add_vxlan_port	= ixgbe_add_vxlan_port,
8196	.ndo_del_vxlan_port	= ixgbe_del_vxlan_port,
8197};
8198
8199/**
8200 * ixgbe_enumerate_functions - Get the number of ports this device has
8201 * @adapter: adapter structure
8202 *
8203 * This function enumerates the phsyical functions co-located on a single slot,
8204 * in order to determine how many ports a device has. This is most useful in
8205 * determining the required GT/s of PCIe bandwidth necessary for optimal
8206 * performance.
8207 **/
8208static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
8209{
8210	struct pci_dev *entry, *pdev = adapter->pdev;
8211	int physfns = 0;
8212
8213	/* Some cards can not use the generic count PCIe functions method,
8214	 * because they are behind a parent switch, so we hardcode these with
8215	 * the correct number of functions.
8216	 */
8217	if (ixgbe_pcie_from_parent(&adapter->hw))
8218		physfns = 4;
8219
8220	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
8221		/* don't count virtual functions */
8222		if (entry->is_virtfn)
8223			continue;
8224
8225		/* When the devices on the bus don't all match our device ID,
8226		 * we can't reliably determine the correct number of
8227		 * functions. This can occur if a function has been direct
8228		 * attached to a virtual machine using VT-d, for example. In
8229		 * this case, simply return -1 to indicate this.
8230		 */
8231		if ((entry->vendor != pdev->vendor) ||
8232		    (entry->device != pdev->device))
8233			return -1;
8234
8235		physfns++;
8236	}
8237
8238	return physfns;
8239}
8240
8241/**
8242 * ixgbe_wol_supported - Check whether device supports WoL
8243 * @hw: hw specific details
8244 * @device_id: the device ID
8245 * @subdev_id: the subsystem device ID
8246 *
8247 * This function is used by probe and ethtool to determine
8248 * which devices have WoL support
8249 *
8250 **/
8251int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
8252			u16 subdevice_id)
8253{
8254	struct ixgbe_hw *hw = &adapter->hw;
8255	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
8256	int is_wol_supported = 0;
8257
8258	switch (device_id) {
8259	case IXGBE_DEV_ID_82599_SFP:
8260		/* Only these subdevices could supports WOL */
8261		switch (subdevice_id) {
8262		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
8263		case IXGBE_SUBDEV_ID_82599_560FLR:
8264			/* only support first port */
8265			if (hw->bus.func != 0)
8266				break;
8267		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
8268		case IXGBE_SUBDEV_ID_82599_SFP:
8269		case IXGBE_SUBDEV_ID_82599_RNDC:
8270		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
8271		case IXGBE_SUBDEV_ID_82599_LOM_SFP:
8272			is_wol_supported = 1;
8273			break;
8274		}
8275		break;
8276	case IXGBE_DEV_ID_82599EN_SFP:
8277		/* Only this subdevice supports WOL */
8278		switch (subdevice_id) {
8279		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
8280			is_wol_supported = 1;
8281			break;
8282		}
8283		break;
8284	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
8285		/* All except this subdevice support WOL */
8286		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
8287			is_wol_supported = 1;
8288		break;
8289	case IXGBE_DEV_ID_82599_KX4:
8290		is_wol_supported = 1;
8291		break;
8292	case IXGBE_DEV_ID_X540T:
8293	case IXGBE_DEV_ID_X540T1:
8294		/* check eeprom to see if enabled wol */
8295		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
8296		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
8297		     (hw->bus.func == 0))) {
8298			is_wol_supported = 1;
8299		}
8300		break;
8301	}
8302
8303	return is_wol_supported;
8304}
8305
8306/**
8307 * ixgbe_get_platform_mac_addr - Look up MAC address in Open Firmware / IDPROM
8308 * @adapter: Pointer to adapter struct
8309 */
8310static void ixgbe_get_platform_mac_addr(struct ixgbe_adapter *adapter)
8311{
8312#ifdef CONFIG_OF
8313	struct device_node *dp = pci_device_to_OF_node(adapter->pdev);
8314	struct ixgbe_hw *hw = &adapter->hw;
8315	const unsigned char *addr;
8316
8317	addr = of_get_mac_address(dp);
8318	if (addr) {
8319		ether_addr_copy(hw->mac.perm_addr, addr);
8320		return;
8321	}
8322#endif /* CONFIG_OF */
8323
8324#ifdef CONFIG_SPARC
8325	ether_addr_copy(hw->mac.perm_addr, idprom->id_ethaddr);
8326#endif /* CONFIG_SPARC */
8327}
8328
8329/**
8330 * ixgbe_probe - Device Initialization Routine
8331 * @pdev: PCI device information struct
8332 * @ent: entry in ixgbe_pci_tbl
8333 *
8334 * Returns 0 on success, negative on failure
8335 *
8336 * ixgbe_probe initializes an adapter identified by a pci_dev structure.
8337 * The OS initialization, configuring of the adapter private structure,
8338 * and a hardware reset occur.
8339 **/
8340static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
8341{
8342	struct net_device *netdev;
8343	struct ixgbe_adapter *adapter = NULL;
8344	struct ixgbe_hw *hw;
8345	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
8346	int i, err, pci_using_dac, expected_gts;
8347	unsigned int indices = MAX_TX_QUEUES;
8348	u8 part_str[IXGBE_PBANUM_LENGTH];
8349	bool disable_dev = false;
8350#ifdef IXGBE_FCOE
8351	u16 device_caps;
8352#endif
8353	u32 eec;
8354
8355	/* Catch broken hardware that put the wrong VF device ID in
8356	 * the PCIe SR-IOV capability.
8357	 */
8358	if (pdev->is_virtfn) {
8359		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
8360		     pci_name(pdev), pdev->vendor, pdev->device);
8361		return -EINVAL;
8362	}
8363
8364	err = pci_enable_device_mem(pdev);
8365	if (err)
8366		return err;
8367
8368	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
8369		pci_using_dac = 1;
8370	} else {
8371		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
8372		if (err) {
8373			dev_err(&pdev->dev,
8374				"No usable DMA configuration, aborting\n");
8375			goto err_dma;
8376		}
8377		pci_using_dac = 0;
8378	}
8379
8380	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
8381					   IORESOURCE_MEM), ixgbe_driver_name);
8382	if (err) {
8383		dev_err(&pdev->dev,
8384			"pci_request_selected_regions failed 0x%x\n", err);
8385		goto err_pci_reg;
8386	}
8387
8388	pci_enable_pcie_error_reporting(pdev);
8389
8390	pci_set_master(pdev);
8391	pci_save_state(pdev);
8392
8393	if (ii->mac == ixgbe_mac_82598EB) {
8394#ifdef CONFIG_IXGBE_DCB
8395		/* 8 TC w/ 4 queues per TC */
8396		indices = 4 * MAX_TRAFFIC_CLASS;
8397#else
8398		indices = IXGBE_MAX_RSS_INDICES;
8399#endif
8400	}
8401
8402	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
8403	if (!netdev) {
8404		err = -ENOMEM;
8405		goto err_alloc_etherdev;
8406	}
8407
8408	SET_NETDEV_DEV(netdev, &pdev->dev);
8409
8410	adapter = netdev_priv(netdev);
8411
8412	adapter->netdev = netdev;
8413	adapter->pdev = pdev;
8414	hw = &adapter->hw;
8415	hw->back = adapter;
8416	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
8417
8418	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
8419			      pci_resource_len(pdev, 0));
8420	adapter->io_addr = hw->hw_addr;
8421	if (!hw->hw_addr) {
8422		err = -EIO;
8423		goto err_ioremap;
8424	}
8425
8426	netdev->netdev_ops = &ixgbe_netdev_ops;
8427	ixgbe_set_ethtool_ops(netdev);
8428	netdev->watchdog_timeo = 5 * HZ;
8429	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
8430
8431	/* Setup hw api */
8432	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
8433	hw->mac.type  = ii->mac;
8434
8435	/* EEPROM */
8436	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
8437	eec = IXGBE_READ_REG(hw, IXGBE_EEC);
8438	if (ixgbe_removed(hw->hw_addr)) {
8439		err = -EIO;
8440		goto err_ioremap;
8441	}
8442	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
8443	if (!(eec & (1 << 8)))
8444		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
8445
8446	/* PHY */
8447	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
8448	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
8449	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
8450	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
8451	hw->phy.mdio.mmds = 0;
8452	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
8453	hw->phy.mdio.dev = netdev;
8454	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
8455	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
8456
8457	ii->get_invariants(hw);
8458
8459	/* setup the private structure */
8460	err = ixgbe_sw_init(adapter);
8461	if (err)
8462		goto err_sw_init;
8463
8464	/* Make it possible the adapter to be woken up via WOL */
8465	switch (adapter->hw.mac.type) {
8466	case ixgbe_mac_82599EB:
8467	case ixgbe_mac_X540:
8468	case ixgbe_mac_X550:
8469	case ixgbe_mac_X550EM_x:
8470		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
8471		break;
8472	default:
8473		break;
8474	}
8475
8476	/*
8477	 * If there is a fan on this device and it has failed log the
8478	 * failure.
8479	 */
8480	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
8481		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
8482		if (esdp & IXGBE_ESDP_SDP1)
8483			e_crit(probe, "Fan has stopped, replace the adapter\n");
8484	}
8485
8486	if (allow_unsupported_sfp)
8487		hw->allow_unsupported_sfp = allow_unsupported_sfp;
8488
8489	/* reset_hw fills in the perm_addr as well */
8490	hw->phy.reset_if_overtemp = true;
8491	err = hw->mac.ops.reset_hw(hw);
8492	hw->phy.reset_if_overtemp = false;
8493	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
8494	    hw->mac.type == ixgbe_mac_82598EB) {
8495		err = 0;
8496	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
8497		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
8498		e_dev_err("Reload the driver after installing a supported module.\n");
8499		goto err_sw_init;
8500	} else if (err) {
8501		e_dev_err("HW Init failed: %d\n", err);
8502		goto err_sw_init;
8503	}
8504
8505#ifdef CONFIG_PCI_IOV
8506	/* SR-IOV not supported on the 82598 */
8507	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8508		goto skip_sriov;
8509	/* Mailbox */
8510	ixgbe_init_mbx_params_pf(hw);
8511	memcpy(&hw->mbx.ops, ii->mbx_ops, sizeof(hw->mbx.ops));
8512	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
8513	ixgbe_enable_sriov(adapter);
8514skip_sriov:
8515
8516#endif
8517	netdev->features = NETIF_F_SG |
8518			   NETIF_F_IP_CSUM |
8519			   NETIF_F_IPV6_CSUM |
8520			   NETIF_F_HW_VLAN_CTAG_TX |
8521			   NETIF_F_HW_VLAN_CTAG_RX |
8522			   NETIF_F_TSO |
8523			   NETIF_F_TSO6 |
8524			   NETIF_F_RXHASH |
8525			   NETIF_F_RXCSUM;
8526
8527	netdev->hw_features = netdev->features | NETIF_F_HW_L2FW_DOFFLOAD;
8528
8529	switch (adapter->hw.mac.type) {
8530	case ixgbe_mac_82599EB:
8531	case ixgbe_mac_X540:
8532	case ixgbe_mac_X550:
8533	case ixgbe_mac_X550EM_x:
8534		netdev->features |= NETIF_F_SCTP_CSUM;
8535		netdev->hw_features |= NETIF_F_SCTP_CSUM |
8536				       NETIF_F_NTUPLE;
8537		break;
8538	default:
8539		break;
8540	}
8541
8542	netdev->hw_features |= NETIF_F_RXALL;
8543	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
8544
8545	netdev->vlan_features |= NETIF_F_TSO;
8546	netdev->vlan_features |= NETIF_F_TSO6;
8547	netdev->vlan_features |= NETIF_F_IP_CSUM;
8548	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
8549	netdev->vlan_features |= NETIF_F_SG;
8550
8551	netdev->priv_flags |= IFF_UNICAST_FLT;
8552	netdev->priv_flags |= IFF_SUPP_NOFCS;
8553
8554	switch (adapter->hw.mac.type) {
8555	case ixgbe_mac_X550:
8556	case ixgbe_mac_X550EM_x:
8557		netdev->hw_enc_features |= NETIF_F_RXCSUM;
8558		break;
8559	default:
8560		break;
8561	}
8562
8563#ifdef CONFIG_IXGBE_DCB
8564	netdev->dcbnl_ops = &dcbnl_ops;
8565#endif
8566
8567#ifdef IXGBE_FCOE
8568	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
8569		unsigned int fcoe_l;
8570
8571		if (hw->mac.ops.get_device_caps) {
8572			hw->mac.ops.get_device_caps(hw, &device_caps);
8573			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
8574				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
8575		}
8576
8577
8578		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
8579		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
8580
8581		netdev->features |= NETIF_F_FSO |
8582				    NETIF_F_FCOE_CRC;
8583
8584		netdev->vlan_features |= NETIF_F_FSO |
8585					 NETIF_F_FCOE_CRC |
8586					 NETIF_F_FCOE_MTU;
8587	}
8588#endif /* IXGBE_FCOE */
8589	if (pci_using_dac) {
8590		netdev->features |= NETIF_F_HIGHDMA;
8591		netdev->vlan_features |= NETIF_F_HIGHDMA;
8592	}
8593
8594	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
8595		netdev->hw_features |= NETIF_F_LRO;
8596	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8597		netdev->features |= NETIF_F_LRO;
8598
8599	/* make sure the EEPROM is good */
8600	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
8601		e_dev_err("The EEPROM Checksum Is Not Valid\n");
8602		err = -EIO;
8603		goto err_sw_init;
8604	}
8605
8606	ixgbe_get_platform_mac_addr(adapter);
8607
8608	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
8609
8610	if (!is_valid_ether_addr(netdev->dev_addr)) {
8611		e_dev_err("invalid MAC address\n");
8612		err = -EIO;
8613		goto err_sw_init;
8614	}
8615
8616	ixgbe_mac_set_default_filter(adapter, hw->mac.perm_addr);
8617
8618	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
8619		    (unsigned long) adapter);
8620
8621	if (ixgbe_removed(hw->hw_addr)) {
8622		err = -EIO;
8623		goto err_sw_init;
8624	}
8625	INIT_WORK(&adapter->service_task, ixgbe_service_task);
8626	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
8627	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
8628
8629	err = ixgbe_init_interrupt_scheme(adapter);
8630	if (err)
8631		goto err_sw_init;
8632
8633	/* WOL not supported for all devices */
8634	adapter->wol = 0;
8635	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
8636	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
8637						pdev->subsystem_device);
8638	if (hw->wol_enabled)
8639		adapter->wol = IXGBE_WUFC_MAG;
8640
8641	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
8642
8643	/* save off EEPROM version number */
8644	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
8645	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
8646
8647	/* pick up the PCI bus settings for reporting later */
8648	hw->mac.ops.get_bus_info(hw);
8649	if (ixgbe_pcie_from_parent(hw))
8650		ixgbe_get_parent_bus_info(adapter);
8651
8652	/* calculate the expected PCIe bandwidth required for optimal
8653	 * performance. Note that some older parts will never have enough
8654	 * bandwidth due to being older generation PCIe parts. We clamp these
8655	 * parts to ensure no warning is displayed if it can't be fixed.
8656	 */
8657	switch (hw->mac.type) {
8658	case ixgbe_mac_82598EB:
8659		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
8660		break;
8661	default:
8662		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
8663		break;
8664	}
8665
8666	/* don't check link if we failed to enumerate functions */
8667	if (expected_gts > 0)
8668		ixgbe_check_minimum_link(adapter, expected_gts);
8669
8670	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
8671	if (err)
8672		strlcpy(part_str, "Unknown", sizeof(part_str));
8673	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
8674		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
8675			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
8676			   part_str);
8677	else
8678		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
8679			   hw->mac.type, hw->phy.type, part_str);
8680
8681	e_dev_info("%pM\n", netdev->dev_addr);
8682
8683	/* reset the hardware with the new settings */
8684	err = hw->mac.ops.start_hw(hw);
8685	if (err == IXGBE_ERR_EEPROM_VERSION) {
8686		/* We are running on a pre-production device, log a warning */
8687		e_dev_warn("This device is a pre-production adapter/LOM. "
8688			   "Please be aware there may be issues associated "
8689			   "with your hardware.  If you are experiencing "
8690			   "problems please contact your Intel or hardware "
8691			   "representative who provided you with this "
8692			   "hardware.\n");
8693	}
8694	strcpy(netdev->name, "eth%d");
8695	err = register_netdev(netdev);
8696	if (err)
8697		goto err_register;
8698
8699	pci_set_drvdata(pdev, adapter);
8700
8701	/* power down the optics for 82599 SFP+ fiber */
8702	if (hw->mac.ops.disable_tx_laser)
8703		hw->mac.ops.disable_tx_laser(hw);
8704
8705	/* carrier off reporting is important to ethtool even BEFORE open */
8706	netif_carrier_off(netdev);
8707
8708#ifdef CONFIG_IXGBE_DCA
8709	if (dca_add_requester(&pdev->dev) == 0) {
8710		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
8711		ixgbe_setup_dca(adapter);
8712	}
8713#endif
8714	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
8715		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
8716		for (i = 0; i < adapter->num_vfs; i++)
8717			ixgbe_vf_configuration(pdev, (i | 0x10000000));
8718	}
8719
8720	/* firmware requires driver version to be 0xFFFFFFFF
8721	 * since os does not support feature
8722	 */
8723	if (hw->mac.ops.set_fw_drv_ver)
8724		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
8725					   0xFF);
8726
8727	/* add san mac addr to netdev */
8728	ixgbe_add_sanmac_netdev(netdev);
8729
8730	e_dev_info("%s\n", ixgbe_default_device_descr);
8731
8732#ifdef CONFIG_IXGBE_HWMON
8733	if (ixgbe_sysfs_init(adapter))
8734		e_err(probe, "failed to allocate sysfs resources\n");
8735#endif /* CONFIG_IXGBE_HWMON */
8736
8737	ixgbe_dbg_adapter_init(adapter);
8738
8739	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
8740	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
8741		hw->mac.ops.setup_link(hw,
8742			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
8743			true);
8744
8745	return 0;
8746
8747err_register:
8748	ixgbe_release_hw_control(adapter);
8749	ixgbe_clear_interrupt_scheme(adapter);
8750err_sw_init:
8751	ixgbe_disable_sriov(adapter);
8752	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
8753	iounmap(adapter->io_addr);
8754	kfree(adapter->mac_table);
8755err_ioremap:
8756	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
8757	free_netdev(netdev);
8758err_alloc_etherdev:
8759	pci_release_selected_regions(pdev,
8760				     pci_select_bars(pdev, IORESOURCE_MEM));
8761err_pci_reg:
8762err_dma:
8763	if (!adapter || disable_dev)
8764		pci_disable_device(pdev);
8765	return err;
8766}
8767
8768/**
8769 * ixgbe_remove - Device Removal Routine
8770 * @pdev: PCI device information struct
8771 *
8772 * ixgbe_remove is called by the PCI subsystem to alert the driver
8773 * that it should release a PCI device.  The could be caused by a
8774 * Hot-Plug event, or because the driver is going to be removed from
8775 * memory.
8776 **/
8777static void ixgbe_remove(struct pci_dev *pdev)
8778{
8779	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8780	struct net_device *netdev;
8781	bool disable_dev;
8782
8783	/* if !adapter then we already cleaned up in probe */
8784	if (!adapter)
8785		return;
8786
8787	netdev  = adapter->netdev;
8788	ixgbe_dbg_adapter_exit(adapter);
8789
8790	set_bit(__IXGBE_REMOVING, &adapter->state);
8791	cancel_work_sync(&adapter->service_task);
8792
8793
8794#ifdef CONFIG_IXGBE_DCA
8795	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
8796		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
8797		dca_remove_requester(&pdev->dev);
8798		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
8799	}
8800
8801#endif
8802#ifdef CONFIG_IXGBE_HWMON
8803	ixgbe_sysfs_exit(adapter);
8804#endif /* CONFIG_IXGBE_HWMON */
8805
8806	/* remove the added san mac */
8807	ixgbe_del_sanmac_netdev(netdev);
8808
8809	if (netdev->reg_state == NETREG_REGISTERED)
8810		unregister_netdev(netdev);
8811
8812#ifdef CONFIG_PCI_IOV
8813	/*
8814	 * Only disable SR-IOV on unload if the user specified the now
8815	 * deprecated max_vfs module parameter.
8816	 */
8817	if (max_vfs)
8818		ixgbe_disable_sriov(adapter);
8819#endif
8820	ixgbe_clear_interrupt_scheme(adapter);
8821
8822	ixgbe_release_hw_control(adapter);
8823
8824#ifdef CONFIG_DCB
8825	kfree(adapter->ixgbe_ieee_pfc);
8826	kfree(adapter->ixgbe_ieee_ets);
8827
8828#endif
8829	iounmap(adapter->io_addr);
8830	pci_release_selected_regions(pdev, pci_select_bars(pdev,
8831				     IORESOURCE_MEM));
8832
8833	e_dev_info("complete\n");
8834
8835	kfree(adapter->mac_table);
8836	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
8837	free_netdev(netdev);
8838
8839	pci_disable_pcie_error_reporting(pdev);
8840
8841	if (disable_dev)
8842		pci_disable_device(pdev);
8843}
8844
8845/**
8846 * ixgbe_io_error_detected - called when PCI error is detected
8847 * @pdev: Pointer to PCI device
8848 * @state: The current pci connection state
8849 *
8850 * This function is called after a PCI bus error affecting
8851 * this device has been detected.
8852 */
8853static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
8854						pci_channel_state_t state)
8855{
8856	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8857	struct net_device *netdev = adapter->netdev;
8858
8859#ifdef CONFIG_PCI_IOV
8860	struct ixgbe_hw *hw = &adapter->hw;
8861	struct pci_dev *bdev, *vfdev;
8862	u32 dw0, dw1, dw2, dw3;
8863	int vf, pos;
8864	u16 req_id, pf_func;
8865
8866	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
8867	    adapter->num_vfs == 0)
8868		goto skip_bad_vf_detection;
8869
8870	bdev = pdev->bus->self;
8871	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
8872		bdev = bdev->bus->self;
8873
8874	if (!bdev)
8875		goto skip_bad_vf_detection;
8876
8877	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
8878	if (!pos)
8879		goto skip_bad_vf_detection;
8880
8881	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
8882	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
8883	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
8884	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
8885	if (ixgbe_removed(hw->hw_addr))
8886		goto skip_bad_vf_detection;
8887
8888	req_id = dw1 >> 16;
8889	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
8890	if (!(req_id & 0x0080))
8891		goto skip_bad_vf_detection;
8892
8893	pf_func = req_id & 0x01;
8894	if ((pf_func & 1) == (pdev->devfn & 1)) {
8895		unsigned int device_id;
8896
8897		vf = (req_id & 0x7F) >> 1;
8898		e_dev_err("VF %d has caused a PCIe error\n", vf);
8899		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
8900				"%8.8x\tdw3: %8.8x\n",
8901		dw0, dw1, dw2, dw3);
8902		switch (adapter->hw.mac.type) {
8903		case ixgbe_mac_82599EB:
8904			device_id = IXGBE_82599_VF_DEVICE_ID;
8905			break;
8906		case ixgbe_mac_X540:
8907			device_id = IXGBE_X540_VF_DEVICE_ID;
8908			break;
8909		case ixgbe_mac_X550:
8910			device_id = IXGBE_DEV_ID_X550_VF;
8911			break;
8912		case ixgbe_mac_X550EM_x:
8913			device_id = IXGBE_DEV_ID_X550EM_X_VF;
8914			break;
8915		default:
8916			device_id = 0;
8917			break;
8918		}
8919
8920		/* Find the pci device of the offending VF */
8921		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
8922		while (vfdev) {
8923			if (vfdev->devfn == (req_id & 0xFF))
8924				break;
8925			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
8926					       device_id, vfdev);
8927		}
8928		/*
8929		 * There's a slim chance the VF could have been hot plugged,
8930		 * so if it is no longer present we don't need to issue the
8931		 * VFLR.  Just clean up the AER in that case.
8932		 */
8933		if (vfdev) {
8934			ixgbe_issue_vf_flr(adapter, vfdev);
8935			/* Free device reference count */
8936			pci_dev_put(vfdev);
8937		}
8938
8939		pci_cleanup_aer_uncorrect_error_status(pdev);
8940	}
8941
8942	/*
8943	 * Even though the error may have occurred on the other port
8944	 * we still need to increment the vf error reference count for
8945	 * both ports because the I/O resume function will be called
8946	 * for both of them.
8947	 */
8948	adapter->vferr_refcount++;
8949
8950	return PCI_ERS_RESULT_RECOVERED;
8951
8952skip_bad_vf_detection:
8953#endif /* CONFIG_PCI_IOV */
8954	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
8955		return PCI_ERS_RESULT_DISCONNECT;
8956
8957	rtnl_lock();
8958	netif_device_detach(netdev);
8959
8960	if (state == pci_channel_io_perm_failure) {
8961		rtnl_unlock();
8962		return PCI_ERS_RESULT_DISCONNECT;
8963	}
8964
8965	if (netif_running(netdev))
8966		ixgbe_down(adapter);
8967
8968	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
8969		pci_disable_device(pdev);
8970	rtnl_unlock();
8971
8972	/* Request a slot reset. */
8973	return PCI_ERS_RESULT_NEED_RESET;
8974}
8975
8976/**
8977 * ixgbe_io_slot_reset - called after the pci bus has been reset.
8978 * @pdev: Pointer to PCI device
8979 *
8980 * Restart the card from scratch, as if from a cold-boot.
8981 */
8982static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
8983{
8984	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8985	pci_ers_result_t result;
8986	int err;
8987
8988	if (pci_enable_device_mem(pdev)) {
8989		e_err(probe, "Cannot re-enable PCI device after reset.\n");
8990		result = PCI_ERS_RESULT_DISCONNECT;
8991	} else {
8992		smp_mb__before_atomic();
8993		clear_bit(__IXGBE_DISABLED, &adapter->state);
8994		adapter->hw.hw_addr = adapter->io_addr;
8995		pci_set_master(pdev);
8996		pci_restore_state(pdev);
8997		pci_save_state(pdev);
8998
8999		pci_wake_from_d3(pdev, false);
9000
9001		ixgbe_reset(adapter);
9002		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
9003		result = PCI_ERS_RESULT_RECOVERED;
9004	}
9005
9006	err = pci_cleanup_aer_uncorrect_error_status(pdev);
9007	if (err) {
9008		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
9009			  "failed 0x%0x\n", err);
9010		/* non-fatal, continue */
9011	}
9012
9013	return result;
9014}
9015
9016/**
9017 * ixgbe_io_resume - called when traffic can start flowing again.
9018 * @pdev: Pointer to PCI device
9019 *
9020 * This callback is called when the error recovery driver tells us that
9021 * its OK to resume normal operation.
9022 */
9023static void ixgbe_io_resume(struct pci_dev *pdev)
9024{
9025	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9026	struct net_device *netdev = adapter->netdev;
9027
9028#ifdef CONFIG_PCI_IOV
9029	if (adapter->vferr_refcount) {
9030		e_info(drv, "Resuming after VF err\n");
9031		adapter->vferr_refcount--;
9032		return;
9033	}
9034
9035#endif
9036	if (netif_running(netdev))
9037		ixgbe_up(adapter);
9038
9039	netif_device_attach(netdev);
9040}
9041
9042static const struct pci_error_handlers ixgbe_err_handler = {
9043	.error_detected = ixgbe_io_error_detected,
9044	.slot_reset = ixgbe_io_slot_reset,
9045	.resume = ixgbe_io_resume,
9046};
9047
9048static struct pci_driver ixgbe_driver = {
9049	.name     = ixgbe_driver_name,
9050	.id_table = ixgbe_pci_tbl,
9051	.probe    = ixgbe_probe,
9052	.remove   = ixgbe_remove,
9053#ifdef CONFIG_PM
9054	.suspend  = ixgbe_suspend,
9055	.resume   = ixgbe_resume,
9056#endif
9057	.shutdown = ixgbe_shutdown,
9058	.sriov_configure = ixgbe_pci_sriov_configure,
9059	.err_handler = &ixgbe_err_handler
9060};
9061
9062/**
9063 * ixgbe_init_module - Driver Registration Routine
9064 *
9065 * ixgbe_init_module is the first routine called when the driver is
9066 * loaded. All it does is register with the PCI subsystem.
9067 **/
9068static int __init ixgbe_init_module(void)
9069{
9070	int ret;
9071	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
9072	pr_info("%s\n", ixgbe_copyright);
9073
9074	ixgbe_dbg_init();
9075
9076	ret = pci_register_driver(&ixgbe_driver);
9077	if (ret) {
9078		ixgbe_dbg_exit();
9079		return ret;
9080	}
9081
9082#ifdef CONFIG_IXGBE_DCA
9083	dca_register_notify(&dca_notifier);
9084#endif
9085
9086	return 0;
9087}
9088
9089module_init(ixgbe_init_module);
9090
9091/**
9092 * ixgbe_exit_module - Driver Exit Cleanup Routine
9093 *
9094 * ixgbe_exit_module is called just before the driver is removed
9095 * from memory.
9096 **/
9097static void __exit ixgbe_exit_module(void)
9098{
9099#ifdef CONFIG_IXGBE_DCA
9100	dca_unregister_notify(&dca_notifier);
9101#endif
9102	pci_unregister_driver(&ixgbe_driver);
9103
9104	ixgbe_dbg_exit();
9105}
9106
9107#ifdef CONFIG_IXGBE_DCA
9108static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
9109			    void *p)
9110{
9111	int ret_val;
9112
9113	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
9114					 __ixgbe_notify_dca);
9115
9116	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
9117}
9118
9119#endif /* CONFIG_IXGBE_DCA */
9120
9121module_exit(ixgbe_exit_module);
9122
9123/* ixgbe_main.c */
9124