xref: /openbmc/linux/drivers/net/ethernet/intel/igb/igb_main.c (revision f2d51e579359b708d7063eef543bec6a57d2b5e9)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright(c) 2007 - 2018 Intel Corporation. */
3 
4 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
5 
6 #include <linux/module.h>
7 #include <linux/types.h>
8 #include <linux/init.h>
9 #include <linux/bitops.h>
10 #include <linux/vmalloc.h>
11 #include <linux/pagemap.h>
12 #include <linux/netdevice.h>
13 #include <linux/ipv6.h>
14 #include <linux/slab.h>
15 #include <net/checksum.h>
16 #include <net/ip6_checksum.h>
17 #include <net/pkt_sched.h>
18 #include <net/pkt_cls.h>
19 #include <linux/net_tstamp.h>
20 #include <linux/mii.h>
21 #include <linux/ethtool.h>
22 #include <linux/if.h>
23 #include <linux/if_vlan.h>
24 #include <linux/pci.h>
25 #include <linux/delay.h>
26 #include <linux/interrupt.h>
27 #include <linux/ip.h>
28 #include <linux/tcp.h>
29 #include <linux/sctp.h>
30 #include <linux/if_ether.h>
31 #include <linux/aer.h>
32 #include <linux/prefetch.h>
33 #include <linux/bpf.h>
34 #include <linux/bpf_trace.h>
35 #include <linux/pm_runtime.h>
36 #include <linux/etherdevice.h>
37 #ifdef CONFIG_IGB_DCA
38 #include <linux/dca.h>
39 #endif
40 #include <linux/i2c.h>
41 #include "igb.h"
42 
43 enum queue_mode {
44 	QUEUE_MODE_STRICT_PRIORITY,
45 	QUEUE_MODE_STREAM_RESERVATION,
46 };
47 
48 enum tx_queue_prio {
49 	TX_QUEUE_PRIO_HIGH,
50 	TX_QUEUE_PRIO_LOW,
51 };
52 
53 char igb_driver_name[] = "igb";
54 static const char igb_driver_string[] =
55 				"Intel(R) Gigabit Ethernet Network Driver";
56 static const char igb_copyright[] =
57 				"Copyright (c) 2007-2014 Intel Corporation.";
58 
59 static const struct e1000_info *igb_info_tbl[] = {
60 	[board_82575] = &e1000_82575_info,
61 };
62 
63 static const struct pci_device_id igb_pci_tbl[] = {
64 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
65 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_SGMII) },
66 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
67 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I211_COPPER), board_82575 },
68 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER), board_82575 },
69 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_FIBER), board_82575 },
70 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES), board_82575 },
71 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SGMII), board_82575 },
72 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER_FLASHLESS), board_82575 },
73 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES_FLASHLESS), board_82575 },
74 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_COPPER), board_82575 },
75 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_FIBER), board_82575 },
76 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SERDES), board_82575 },
77 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SGMII), board_82575 },
78 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER), board_82575 },
79 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_FIBER), board_82575 },
80 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_QUAD_FIBER), board_82575 },
81 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SERDES), board_82575 },
82 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SGMII), board_82575 },
83 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER_DUAL), board_82575 },
84 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SGMII), board_82575 },
85 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SERDES), board_82575 },
86 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_BACKPLANE), board_82575 },
87 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SFP), board_82575 },
88 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
89 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
90 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
91 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
92 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
93 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
94 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER_ET2), board_82575 },
95 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
96 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
97 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
98 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
99 	/* required last entry */
100 	{0, }
101 };
102 
103 MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
104 
105 static int igb_setup_all_tx_resources(struct igb_adapter *);
106 static int igb_setup_all_rx_resources(struct igb_adapter *);
107 static void igb_free_all_tx_resources(struct igb_adapter *);
108 static void igb_free_all_rx_resources(struct igb_adapter *);
109 static void igb_setup_mrqc(struct igb_adapter *);
110 static int igb_probe(struct pci_dev *, const struct pci_device_id *);
111 static void igb_remove(struct pci_dev *pdev);
112 static int igb_sw_init(struct igb_adapter *);
113 int igb_open(struct net_device *);
114 int igb_close(struct net_device *);
115 static void igb_configure(struct igb_adapter *);
116 static void igb_configure_tx(struct igb_adapter *);
117 static void igb_configure_rx(struct igb_adapter *);
118 static void igb_clean_all_tx_rings(struct igb_adapter *);
119 static void igb_clean_all_rx_rings(struct igb_adapter *);
120 static void igb_clean_tx_ring(struct igb_ring *);
121 static void igb_clean_rx_ring(struct igb_ring *);
122 static void igb_set_rx_mode(struct net_device *);
123 static void igb_update_phy_info(struct timer_list *);
124 static void igb_watchdog(struct timer_list *);
125 static void igb_watchdog_task(struct work_struct *);
126 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb, struct net_device *);
127 static void igb_get_stats64(struct net_device *dev,
128 			    struct rtnl_link_stats64 *stats);
129 static int igb_change_mtu(struct net_device *, int);
130 static int igb_set_mac(struct net_device *, void *);
131 static void igb_set_uta(struct igb_adapter *adapter, bool set);
132 static irqreturn_t igb_intr(int irq, void *);
133 static irqreturn_t igb_intr_msi(int irq, void *);
134 static irqreturn_t igb_msix_other(int irq, void *);
135 static irqreturn_t igb_msix_ring(int irq, void *);
136 #ifdef CONFIG_IGB_DCA
137 static void igb_update_dca(struct igb_q_vector *);
138 static void igb_setup_dca(struct igb_adapter *);
139 #endif /* CONFIG_IGB_DCA */
140 static int igb_poll(struct napi_struct *, int);
141 static bool igb_clean_tx_irq(struct igb_q_vector *, int);
142 static int igb_clean_rx_irq(struct igb_q_vector *, int);
143 static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
144 static void igb_tx_timeout(struct net_device *, unsigned int txqueue);
145 static void igb_reset_task(struct work_struct *);
146 static void igb_vlan_mode(struct net_device *netdev,
147 			  netdev_features_t features);
148 static int igb_vlan_rx_add_vid(struct net_device *, __be16, u16);
149 static int igb_vlan_rx_kill_vid(struct net_device *, __be16, u16);
150 static void igb_restore_vlan(struct igb_adapter *);
151 static void igb_rar_set_index(struct igb_adapter *, u32);
152 static void igb_ping_all_vfs(struct igb_adapter *);
153 static void igb_msg_task(struct igb_adapter *);
154 static void igb_vmm_control(struct igb_adapter *);
155 static int igb_set_vf_mac(struct igb_adapter *, int, unsigned char *);
156 static void igb_flush_mac_table(struct igb_adapter *);
157 static int igb_available_rars(struct igb_adapter *, u8);
158 static void igb_set_default_mac_filter(struct igb_adapter *);
159 static int igb_uc_sync(struct net_device *, const unsigned char *);
160 static int igb_uc_unsync(struct net_device *, const unsigned char *);
161 static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
162 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac);
163 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
164 			       int vf, u16 vlan, u8 qos, __be16 vlan_proto);
165 static int igb_ndo_set_vf_bw(struct net_device *, int, int, int);
166 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
167 				   bool setting);
168 static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf,
169 				bool setting);
170 static int igb_ndo_get_vf_config(struct net_device *netdev, int vf,
171 				 struct ifla_vf_info *ivi);
172 static void igb_check_vf_rate_limit(struct igb_adapter *);
173 static void igb_nfc_filter_exit(struct igb_adapter *adapter);
174 static void igb_nfc_filter_restore(struct igb_adapter *adapter);
175 
176 #ifdef CONFIG_PCI_IOV
177 static int igb_vf_configure(struct igb_adapter *adapter, int vf);
178 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs);
179 static int igb_disable_sriov(struct pci_dev *dev);
180 static int igb_pci_disable_sriov(struct pci_dev *dev);
181 #endif
182 
183 static int igb_suspend(struct device *);
184 static int igb_resume(struct device *);
185 static int igb_runtime_suspend(struct device *dev);
186 static int igb_runtime_resume(struct device *dev);
187 static int igb_runtime_idle(struct device *dev);
188 static const struct dev_pm_ops igb_pm_ops = {
189 	SET_SYSTEM_SLEEP_PM_OPS(igb_suspend, igb_resume)
190 	SET_RUNTIME_PM_OPS(igb_runtime_suspend, igb_runtime_resume,
191 			igb_runtime_idle)
192 };
193 static void igb_shutdown(struct pci_dev *);
194 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs);
195 #ifdef CONFIG_IGB_DCA
196 static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
197 static struct notifier_block dca_notifier = {
198 	.notifier_call	= igb_notify_dca,
199 	.next		= NULL,
200 	.priority	= 0
201 };
202 #endif
203 #ifdef CONFIG_PCI_IOV
204 static unsigned int max_vfs;
205 module_param(max_vfs, uint, 0);
206 MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate per physical function");
207 #endif /* CONFIG_PCI_IOV */
208 
209 static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
210 		     pci_channel_state_t);
211 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
212 static void igb_io_resume(struct pci_dev *);
213 
214 static const struct pci_error_handlers igb_err_handler = {
215 	.error_detected = igb_io_error_detected,
216 	.slot_reset = igb_io_slot_reset,
217 	.resume = igb_io_resume,
218 };
219 
220 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba);
221 
222 static struct pci_driver igb_driver = {
223 	.name     = igb_driver_name,
224 	.id_table = igb_pci_tbl,
225 	.probe    = igb_probe,
226 	.remove   = igb_remove,
227 #ifdef CONFIG_PM
228 	.driver.pm = &igb_pm_ops,
229 #endif
230 	.shutdown = igb_shutdown,
231 	.sriov_configure = igb_pci_sriov_configure,
232 	.err_handler = &igb_err_handler
233 };
234 
235 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
236 MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
237 MODULE_LICENSE("GPL v2");
238 
239 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
240 static int debug = -1;
241 module_param(debug, int, 0);
242 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
243 
244 struct igb_reg_info {
245 	u32 ofs;
246 	char *name;
247 };
248 
249 static const struct igb_reg_info igb_reg_info_tbl[] = {
250 
251 	/* General Registers */
252 	{E1000_CTRL, "CTRL"},
253 	{E1000_STATUS, "STATUS"},
254 	{E1000_CTRL_EXT, "CTRL_EXT"},
255 
256 	/* Interrupt Registers */
257 	{E1000_ICR, "ICR"},
258 
259 	/* RX Registers */
260 	{E1000_RCTL, "RCTL"},
261 	{E1000_RDLEN(0), "RDLEN"},
262 	{E1000_RDH(0), "RDH"},
263 	{E1000_RDT(0), "RDT"},
264 	{E1000_RXDCTL(0), "RXDCTL"},
265 	{E1000_RDBAL(0), "RDBAL"},
266 	{E1000_RDBAH(0), "RDBAH"},
267 
268 	/* TX Registers */
269 	{E1000_TCTL, "TCTL"},
270 	{E1000_TDBAL(0), "TDBAL"},
271 	{E1000_TDBAH(0), "TDBAH"},
272 	{E1000_TDLEN(0), "TDLEN"},
273 	{E1000_TDH(0), "TDH"},
274 	{E1000_TDT(0), "TDT"},
275 	{E1000_TXDCTL(0), "TXDCTL"},
276 	{E1000_TDFH, "TDFH"},
277 	{E1000_TDFT, "TDFT"},
278 	{E1000_TDFHS, "TDFHS"},
279 	{E1000_TDFPC, "TDFPC"},
280 
281 	/* List Terminator */
282 	{}
283 };
284 
285 /* igb_regdump - register printout routine */
286 static void igb_regdump(struct e1000_hw *hw, struct igb_reg_info *reginfo)
287 {
288 	int n = 0;
289 	char rname[16];
290 	u32 regs[8];
291 
292 	switch (reginfo->ofs) {
293 	case E1000_RDLEN(0):
294 		for (n = 0; n < 4; n++)
295 			regs[n] = rd32(E1000_RDLEN(n));
296 		break;
297 	case E1000_RDH(0):
298 		for (n = 0; n < 4; n++)
299 			regs[n] = rd32(E1000_RDH(n));
300 		break;
301 	case E1000_RDT(0):
302 		for (n = 0; n < 4; n++)
303 			regs[n] = rd32(E1000_RDT(n));
304 		break;
305 	case E1000_RXDCTL(0):
306 		for (n = 0; n < 4; n++)
307 			regs[n] = rd32(E1000_RXDCTL(n));
308 		break;
309 	case E1000_RDBAL(0):
310 		for (n = 0; n < 4; n++)
311 			regs[n] = rd32(E1000_RDBAL(n));
312 		break;
313 	case E1000_RDBAH(0):
314 		for (n = 0; n < 4; n++)
315 			regs[n] = rd32(E1000_RDBAH(n));
316 		break;
317 	case E1000_TDBAL(0):
318 		for (n = 0; n < 4; n++)
319 			regs[n] = rd32(E1000_TDBAL(n));
320 		break;
321 	case E1000_TDBAH(0):
322 		for (n = 0; n < 4; n++)
323 			regs[n] = rd32(E1000_TDBAH(n));
324 		break;
325 	case E1000_TDLEN(0):
326 		for (n = 0; n < 4; n++)
327 			regs[n] = rd32(E1000_TDLEN(n));
328 		break;
329 	case E1000_TDH(0):
330 		for (n = 0; n < 4; n++)
331 			regs[n] = rd32(E1000_TDH(n));
332 		break;
333 	case E1000_TDT(0):
334 		for (n = 0; n < 4; n++)
335 			regs[n] = rd32(E1000_TDT(n));
336 		break;
337 	case E1000_TXDCTL(0):
338 		for (n = 0; n < 4; n++)
339 			regs[n] = rd32(E1000_TXDCTL(n));
340 		break;
341 	default:
342 		pr_info("%-15s %08x\n", reginfo->name, rd32(reginfo->ofs));
343 		return;
344 	}
345 
346 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-3]");
347 	pr_info("%-15s %08x %08x %08x %08x\n", rname, regs[0], regs[1],
348 		regs[2], regs[3]);
349 }
350 
351 /* igb_dump - Print registers, Tx-rings and Rx-rings */
352 static void igb_dump(struct igb_adapter *adapter)
353 {
354 	struct net_device *netdev = adapter->netdev;
355 	struct e1000_hw *hw = &adapter->hw;
356 	struct igb_reg_info *reginfo;
357 	struct igb_ring *tx_ring;
358 	union e1000_adv_tx_desc *tx_desc;
359 	struct my_u0 { __le64 a; __le64 b; } *u0;
360 	struct igb_ring *rx_ring;
361 	union e1000_adv_rx_desc *rx_desc;
362 	u32 staterr;
363 	u16 i, n;
364 
365 	if (!netif_msg_hw(adapter))
366 		return;
367 
368 	/* Print netdevice Info */
369 	if (netdev) {
370 		dev_info(&adapter->pdev->dev, "Net device Info\n");
371 		pr_info("Device Name     state            trans_start\n");
372 		pr_info("%-15s %016lX %016lX\n", netdev->name,
373 			netdev->state, dev_trans_start(netdev));
374 	}
375 
376 	/* Print Registers */
377 	dev_info(&adapter->pdev->dev, "Register Dump\n");
378 	pr_info(" Register Name   Value\n");
379 	for (reginfo = (struct igb_reg_info *)igb_reg_info_tbl;
380 	     reginfo->name; reginfo++) {
381 		igb_regdump(hw, reginfo);
382 	}
383 
384 	/* Print TX Ring Summary */
385 	if (!netdev || !netif_running(netdev))
386 		goto exit;
387 
388 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
389 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
390 	for (n = 0; n < adapter->num_tx_queues; n++) {
391 		struct igb_tx_buffer *buffer_info;
392 		tx_ring = adapter->tx_ring[n];
393 		buffer_info = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
394 		pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
395 			n, tx_ring->next_to_use, tx_ring->next_to_clean,
396 			(u64)dma_unmap_addr(buffer_info, dma),
397 			dma_unmap_len(buffer_info, len),
398 			buffer_info->next_to_watch,
399 			(u64)buffer_info->time_stamp);
400 	}
401 
402 	/* Print TX Rings */
403 	if (!netif_msg_tx_done(adapter))
404 		goto rx_ring_summary;
405 
406 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
407 
408 	/* Transmit Descriptor Formats
409 	 *
410 	 * Advanced Transmit Descriptor
411 	 *   +--------------------------------------------------------------+
412 	 * 0 |         Buffer Address [63:0]                                |
413 	 *   +--------------------------------------------------------------+
414 	 * 8 | PAYLEN  | PORTS  |CC|IDX | STA | DCMD  |DTYP|MAC|RSV| DTALEN |
415 	 *   +--------------------------------------------------------------+
416 	 *   63      46 45    40 39 38 36 35 32 31   24             15       0
417 	 */
418 
419 	for (n = 0; n < adapter->num_tx_queues; n++) {
420 		tx_ring = adapter->tx_ring[n];
421 		pr_info("------------------------------------\n");
422 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
423 		pr_info("------------------------------------\n");
424 		pr_info("T [desc]     [address 63:0  ] [PlPOCIStDDM Ln] [bi->dma       ] leng  ntw timestamp        bi->skb\n");
425 
426 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
427 			const char *next_desc;
428 			struct igb_tx_buffer *buffer_info;
429 			tx_desc = IGB_TX_DESC(tx_ring, i);
430 			buffer_info = &tx_ring->tx_buffer_info[i];
431 			u0 = (struct my_u0 *)tx_desc;
432 			if (i == tx_ring->next_to_use &&
433 			    i == tx_ring->next_to_clean)
434 				next_desc = " NTC/U";
435 			else if (i == tx_ring->next_to_use)
436 				next_desc = " NTU";
437 			else if (i == tx_ring->next_to_clean)
438 				next_desc = " NTC";
439 			else
440 				next_desc = "";
441 
442 			pr_info("T [0x%03X]    %016llX %016llX %016llX %04X  %p %016llX %p%s\n",
443 				i, le64_to_cpu(u0->a),
444 				le64_to_cpu(u0->b),
445 				(u64)dma_unmap_addr(buffer_info, dma),
446 				dma_unmap_len(buffer_info, len),
447 				buffer_info->next_to_watch,
448 				(u64)buffer_info->time_stamp,
449 				buffer_info->skb, next_desc);
450 
451 			if (netif_msg_pktdata(adapter) && buffer_info->skb)
452 				print_hex_dump(KERN_INFO, "",
453 					DUMP_PREFIX_ADDRESS,
454 					16, 1, buffer_info->skb->data,
455 					dma_unmap_len(buffer_info, len),
456 					true);
457 		}
458 	}
459 
460 	/* Print RX Rings Summary */
461 rx_ring_summary:
462 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
463 	pr_info("Queue [NTU] [NTC]\n");
464 	for (n = 0; n < adapter->num_rx_queues; n++) {
465 		rx_ring = adapter->rx_ring[n];
466 		pr_info(" %5d %5X %5X\n",
467 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
468 	}
469 
470 	/* Print RX Rings */
471 	if (!netif_msg_rx_status(adapter))
472 		goto exit;
473 
474 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
475 
476 	/* Advanced Receive Descriptor (Read) Format
477 	 *    63                                           1        0
478 	 *    +-----------------------------------------------------+
479 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
480 	 *    +----------------------------------------------+------+
481 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
482 	 *    +-----------------------------------------------------+
483 	 *
484 	 *
485 	 * Advanced Receive Descriptor (Write-Back) Format
486 	 *
487 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
488 	 *   +------------------------------------------------------+
489 	 * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
490 	 *   | Checksum   Ident  |   |           |    | Type | Type |
491 	 *   +------------------------------------------------------+
492 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
493 	 *   +------------------------------------------------------+
494 	 *   63       48 47    32 31            20 19               0
495 	 */
496 
497 	for (n = 0; n < adapter->num_rx_queues; n++) {
498 		rx_ring = adapter->rx_ring[n];
499 		pr_info("------------------------------------\n");
500 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
501 		pr_info("------------------------------------\n");
502 		pr_info("R  [desc]      [ PktBuf     A0] [  HeadBuf   DD] [bi->dma       ] [bi->skb] <-- Adv Rx Read format\n");
503 		pr_info("RWB[desc]      [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
504 
505 		for (i = 0; i < rx_ring->count; i++) {
506 			const char *next_desc;
507 			struct igb_rx_buffer *buffer_info;
508 			buffer_info = &rx_ring->rx_buffer_info[i];
509 			rx_desc = IGB_RX_DESC(rx_ring, i);
510 			u0 = (struct my_u0 *)rx_desc;
511 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
512 
513 			if (i == rx_ring->next_to_use)
514 				next_desc = " NTU";
515 			else if (i == rx_ring->next_to_clean)
516 				next_desc = " NTC";
517 			else
518 				next_desc = "";
519 
520 			if (staterr & E1000_RXD_STAT_DD) {
521 				/* Descriptor Done */
522 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %s\n",
523 					"RWB", i,
524 					le64_to_cpu(u0->a),
525 					le64_to_cpu(u0->b),
526 					next_desc);
527 			} else {
528 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %s\n",
529 					"R  ", i,
530 					le64_to_cpu(u0->a),
531 					le64_to_cpu(u0->b),
532 					(u64)buffer_info->dma,
533 					next_desc);
534 
535 				if (netif_msg_pktdata(adapter) &&
536 				    buffer_info->dma && buffer_info->page) {
537 					print_hex_dump(KERN_INFO, "",
538 					  DUMP_PREFIX_ADDRESS,
539 					  16, 1,
540 					  page_address(buffer_info->page) +
541 						      buffer_info->page_offset,
542 					  igb_rx_bufsz(rx_ring), true);
543 				}
544 			}
545 		}
546 	}
547 
548 exit:
549 	return;
550 }
551 
552 /**
553  *  igb_get_i2c_data - Reads the I2C SDA data bit
554  *  @data: opaque pointer to adapter struct
555  *
556  *  Returns the I2C data bit value
557  **/
558 static int igb_get_i2c_data(void *data)
559 {
560 	struct igb_adapter *adapter = (struct igb_adapter *)data;
561 	struct e1000_hw *hw = &adapter->hw;
562 	s32 i2cctl = rd32(E1000_I2CPARAMS);
563 
564 	return !!(i2cctl & E1000_I2C_DATA_IN);
565 }
566 
567 /**
568  *  igb_set_i2c_data - Sets the I2C data bit
569  *  @data: pointer to hardware structure
570  *  @state: I2C data value (0 or 1) to set
571  *
572  *  Sets the I2C data bit
573  **/
574 static void igb_set_i2c_data(void *data, int state)
575 {
576 	struct igb_adapter *adapter = (struct igb_adapter *)data;
577 	struct e1000_hw *hw = &adapter->hw;
578 	s32 i2cctl = rd32(E1000_I2CPARAMS);
579 
580 	if (state) {
581 		i2cctl |= E1000_I2C_DATA_OUT | E1000_I2C_DATA_OE_N;
582 	} else {
583 		i2cctl &= ~E1000_I2C_DATA_OE_N;
584 		i2cctl &= ~E1000_I2C_DATA_OUT;
585 	}
586 
587 	wr32(E1000_I2CPARAMS, i2cctl);
588 	wrfl();
589 }
590 
591 /**
592  *  igb_set_i2c_clk - Sets the I2C SCL clock
593  *  @data: pointer to hardware structure
594  *  @state: state to set clock
595  *
596  *  Sets the I2C clock line to state
597  **/
598 static void igb_set_i2c_clk(void *data, int state)
599 {
600 	struct igb_adapter *adapter = (struct igb_adapter *)data;
601 	struct e1000_hw *hw = &adapter->hw;
602 	s32 i2cctl = rd32(E1000_I2CPARAMS);
603 
604 	if (state) {
605 		i2cctl |= E1000_I2C_CLK_OUT | E1000_I2C_CLK_OE_N;
606 	} else {
607 		i2cctl &= ~E1000_I2C_CLK_OUT;
608 		i2cctl &= ~E1000_I2C_CLK_OE_N;
609 	}
610 	wr32(E1000_I2CPARAMS, i2cctl);
611 	wrfl();
612 }
613 
614 /**
615  *  igb_get_i2c_clk - Gets the I2C SCL clock state
616  *  @data: pointer to hardware structure
617  *
618  *  Gets the I2C clock state
619  **/
620 static int igb_get_i2c_clk(void *data)
621 {
622 	struct igb_adapter *adapter = (struct igb_adapter *)data;
623 	struct e1000_hw *hw = &adapter->hw;
624 	s32 i2cctl = rd32(E1000_I2CPARAMS);
625 
626 	return !!(i2cctl & E1000_I2C_CLK_IN);
627 }
628 
629 static const struct i2c_algo_bit_data igb_i2c_algo = {
630 	.setsda		= igb_set_i2c_data,
631 	.setscl		= igb_set_i2c_clk,
632 	.getsda		= igb_get_i2c_data,
633 	.getscl		= igb_get_i2c_clk,
634 	.udelay		= 5,
635 	.timeout	= 20,
636 };
637 
638 /**
639  *  igb_get_hw_dev - return device
640  *  @hw: pointer to hardware structure
641  *
642  *  used by hardware layer to print debugging information
643  **/
644 struct net_device *igb_get_hw_dev(struct e1000_hw *hw)
645 {
646 	struct igb_adapter *adapter = hw->back;
647 	return adapter->netdev;
648 }
649 
650 /**
651  *  igb_init_module - Driver Registration Routine
652  *
653  *  igb_init_module is the first routine called when the driver is
654  *  loaded. All it does is register with the PCI subsystem.
655  **/
656 static int __init igb_init_module(void)
657 {
658 	int ret;
659 
660 	pr_info("%s\n", igb_driver_string);
661 	pr_info("%s\n", igb_copyright);
662 
663 #ifdef CONFIG_IGB_DCA
664 	dca_register_notify(&dca_notifier);
665 #endif
666 	ret = pci_register_driver(&igb_driver);
667 	return ret;
668 }
669 
670 module_init(igb_init_module);
671 
672 /**
673  *  igb_exit_module - Driver Exit Cleanup Routine
674  *
675  *  igb_exit_module is called just before the driver is removed
676  *  from memory.
677  **/
678 static void __exit igb_exit_module(void)
679 {
680 #ifdef CONFIG_IGB_DCA
681 	dca_unregister_notify(&dca_notifier);
682 #endif
683 	pci_unregister_driver(&igb_driver);
684 }
685 
686 module_exit(igb_exit_module);
687 
688 #define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
689 /**
690  *  igb_cache_ring_register - Descriptor ring to register mapping
691  *  @adapter: board private structure to initialize
692  *
693  *  Once we know the feature-set enabled for the device, we'll cache
694  *  the register offset the descriptor ring is assigned to.
695  **/
696 static void igb_cache_ring_register(struct igb_adapter *adapter)
697 {
698 	int i = 0, j = 0;
699 	u32 rbase_offset = adapter->vfs_allocated_count;
700 
701 	switch (adapter->hw.mac.type) {
702 	case e1000_82576:
703 		/* The queues are allocated for virtualization such that VF 0
704 		 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
705 		 * In order to avoid collision we start at the first free queue
706 		 * and continue consuming queues in the same sequence
707 		 */
708 		if (adapter->vfs_allocated_count) {
709 			for (; i < adapter->rss_queues; i++)
710 				adapter->rx_ring[i]->reg_idx = rbase_offset +
711 							       Q_IDX_82576(i);
712 		}
713 		fallthrough;
714 	case e1000_82575:
715 	case e1000_82580:
716 	case e1000_i350:
717 	case e1000_i354:
718 	case e1000_i210:
719 	case e1000_i211:
720 	default:
721 		for (; i < adapter->num_rx_queues; i++)
722 			adapter->rx_ring[i]->reg_idx = rbase_offset + i;
723 		for (; j < adapter->num_tx_queues; j++)
724 			adapter->tx_ring[j]->reg_idx = rbase_offset + j;
725 		break;
726 	}
727 }
728 
729 u32 igb_rd32(struct e1000_hw *hw, u32 reg)
730 {
731 	struct igb_adapter *igb = container_of(hw, struct igb_adapter, hw);
732 	u8 __iomem *hw_addr = READ_ONCE(hw->hw_addr);
733 	u32 value = 0;
734 
735 	if (E1000_REMOVED(hw_addr))
736 		return ~value;
737 
738 	value = readl(&hw_addr[reg]);
739 
740 	/* reads should not return all F's */
741 	if (!(~value) && (!reg || !(~readl(hw_addr)))) {
742 		struct net_device *netdev = igb->netdev;
743 		hw->hw_addr = NULL;
744 		netdev_err(netdev, "PCIe link lost\n");
745 		WARN(pci_device_is_present(igb->pdev),
746 		     "igb: Failed to read reg 0x%x!\n", reg);
747 	}
748 
749 	return value;
750 }
751 
752 /**
753  *  igb_write_ivar - configure ivar for given MSI-X vector
754  *  @hw: pointer to the HW structure
755  *  @msix_vector: vector number we are allocating to a given ring
756  *  @index: row index of IVAR register to write within IVAR table
757  *  @offset: column offset of in IVAR, should be multiple of 8
758  *
759  *  This function is intended to handle the writing of the IVAR register
760  *  for adapters 82576 and newer.  The IVAR table consists of 2 columns,
761  *  each containing an cause allocation for an Rx and Tx ring, and a
762  *  variable number of rows depending on the number of queues supported.
763  **/
764 static void igb_write_ivar(struct e1000_hw *hw, int msix_vector,
765 			   int index, int offset)
766 {
767 	u32 ivar = array_rd32(E1000_IVAR0, index);
768 
769 	/* clear any bits that are currently set */
770 	ivar &= ~((u32)0xFF << offset);
771 
772 	/* write vector and valid bit */
773 	ivar |= (msix_vector | E1000_IVAR_VALID) << offset;
774 
775 	array_wr32(E1000_IVAR0, index, ivar);
776 }
777 
778 #define IGB_N0_QUEUE -1
779 static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
780 {
781 	struct igb_adapter *adapter = q_vector->adapter;
782 	struct e1000_hw *hw = &adapter->hw;
783 	int rx_queue = IGB_N0_QUEUE;
784 	int tx_queue = IGB_N0_QUEUE;
785 	u32 msixbm = 0;
786 
787 	if (q_vector->rx.ring)
788 		rx_queue = q_vector->rx.ring->reg_idx;
789 	if (q_vector->tx.ring)
790 		tx_queue = q_vector->tx.ring->reg_idx;
791 
792 	switch (hw->mac.type) {
793 	case e1000_82575:
794 		/* The 82575 assigns vectors using a bitmask, which matches the
795 		 * bitmask for the EICR/EIMS/EIMC registers.  To assign one
796 		 * or more queues to a vector, we write the appropriate bits
797 		 * into the MSIXBM register for that vector.
798 		 */
799 		if (rx_queue > IGB_N0_QUEUE)
800 			msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
801 		if (tx_queue > IGB_N0_QUEUE)
802 			msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
803 		if (!(adapter->flags & IGB_FLAG_HAS_MSIX) && msix_vector == 0)
804 			msixbm |= E1000_EIMS_OTHER;
805 		array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
806 		q_vector->eims_value = msixbm;
807 		break;
808 	case e1000_82576:
809 		/* 82576 uses a table that essentially consists of 2 columns
810 		 * with 8 rows.  The ordering is column-major so we use the
811 		 * lower 3 bits as the row index, and the 4th bit as the
812 		 * column offset.
813 		 */
814 		if (rx_queue > IGB_N0_QUEUE)
815 			igb_write_ivar(hw, msix_vector,
816 				       rx_queue & 0x7,
817 				       (rx_queue & 0x8) << 1);
818 		if (tx_queue > IGB_N0_QUEUE)
819 			igb_write_ivar(hw, msix_vector,
820 				       tx_queue & 0x7,
821 				       ((tx_queue & 0x8) << 1) + 8);
822 		q_vector->eims_value = BIT(msix_vector);
823 		break;
824 	case e1000_82580:
825 	case e1000_i350:
826 	case e1000_i354:
827 	case e1000_i210:
828 	case e1000_i211:
829 		/* On 82580 and newer adapters the scheme is similar to 82576
830 		 * however instead of ordering column-major we have things
831 		 * ordered row-major.  So we traverse the table by using
832 		 * bit 0 as the column offset, and the remaining bits as the
833 		 * row index.
834 		 */
835 		if (rx_queue > IGB_N0_QUEUE)
836 			igb_write_ivar(hw, msix_vector,
837 				       rx_queue >> 1,
838 				       (rx_queue & 0x1) << 4);
839 		if (tx_queue > IGB_N0_QUEUE)
840 			igb_write_ivar(hw, msix_vector,
841 				       tx_queue >> 1,
842 				       ((tx_queue & 0x1) << 4) + 8);
843 		q_vector->eims_value = BIT(msix_vector);
844 		break;
845 	default:
846 		BUG();
847 		break;
848 	}
849 
850 	/* add q_vector eims value to global eims_enable_mask */
851 	adapter->eims_enable_mask |= q_vector->eims_value;
852 
853 	/* configure q_vector to set itr on first interrupt */
854 	q_vector->set_itr = 1;
855 }
856 
857 /**
858  *  igb_configure_msix - Configure MSI-X hardware
859  *  @adapter: board private structure to initialize
860  *
861  *  igb_configure_msix sets up the hardware to properly
862  *  generate MSI-X interrupts.
863  **/
864 static void igb_configure_msix(struct igb_adapter *adapter)
865 {
866 	u32 tmp;
867 	int i, vector = 0;
868 	struct e1000_hw *hw = &adapter->hw;
869 
870 	adapter->eims_enable_mask = 0;
871 
872 	/* set vector for other causes, i.e. link changes */
873 	switch (hw->mac.type) {
874 	case e1000_82575:
875 		tmp = rd32(E1000_CTRL_EXT);
876 		/* enable MSI-X PBA support*/
877 		tmp |= E1000_CTRL_EXT_PBA_CLR;
878 
879 		/* Auto-Mask interrupts upon ICR read. */
880 		tmp |= E1000_CTRL_EXT_EIAME;
881 		tmp |= E1000_CTRL_EXT_IRCA;
882 
883 		wr32(E1000_CTRL_EXT, tmp);
884 
885 		/* enable msix_other interrupt */
886 		array_wr32(E1000_MSIXBM(0), vector++, E1000_EIMS_OTHER);
887 		adapter->eims_other = E1000_EIMS_OTHER;
888 
889 		break;
890 
891 	case e1000_82576:
892 	case e1000_82580:
893 	case e1000_i350:
894 	case e1000_i354:
895 	case e1000_i210:
896 	case e1000_i211:
897 		/* Turn on MSI-X capability first, or our settings
898 		 * won't stick.  And it will take days to debug.
899 		 */
900 		wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
901 		     E1000_GPIE_PBA | E1000_GPIE_EIAME |
902 		     E1000_GPIE_NSICR);
903 
904 		/* enable msix_other interrupt */
905 		adapter->eims_other = BIT(vector);
906 		tmp = (vector++ | E1000_IVAR_VALID) << 8;
907 
908 		wr32(E1000_IVAR_MISC, tmp);
909 		break;
910 	default:
911 		/* do nothing, since nothing else supports MSI-X */
912 		break;
913 	} /* switch (hw->mac.type) */
914 
915 	adapter->eims_enable_mask |= adapter->eims_other;
916 
917 	for (i = 0; i < adapter->num_q_vectors; i++)
918 		igb_assign_vector(adapter->q_vector[i], vector++);
919 
920 	wrfl();
921 }
922 
923 /**
924  *  igb_request_msix - Initialize MSI-X interrupts
925  *  @adapter: board private structure to initialize
926  *
927  *  igb_request_msix allocates MSI-X vectors and requests interrupts from the
928  *  kernel.
929  **/
930 static int igb_request_msix(struct igb_adapter *adapter)
931 {
932 	unsigned int num_q_vectors = adapter->num_q_vectors;
933 	struct net_device *netdev = adapter->netdev;
934 	int i, err = 0, vector = 0, free_vector = 0;
935 
936 	err = request_irq(adapter->msix_entries[vector].vector,
937 			  igb_msix_other, 0, netdev->name, adapter);
938 	if (err)
939 		goto err_out;
940 
941 	if (num_q_vectors > MAX_Q_VECTORS) {
942 		num_q_vectors = MAX_Q_VECTORS;
943 		dev_warn(&adapter->pdev->dev,
944 			 "The number of queue vectors (%d) is higher than max allowed (%d)\n",
945 			 adapter->num_q_vectors, MAX_Q_VECTORS);
946 	}
947 	for (i = 0; i < num_q_vectors; i++) {
948 		struct igb_q_vector *q_vector = adapter->q_vector[i];
949 
950 		vector++;
951 
952 		q_vector->itr_register = adapter->io_addr + E1000_EITR(vector);
953 
954 		if (q_vector->rx.ring && q_vector->tx.ring)
955 			sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
956 				q_vector->rx.ring->queue_index);
957 		else if (q_vector->tx.ring)
958 			sprintf(q_vector->name, "%s-tx-%u", netdev->name,
959 				q_vector->tx.ring->queue_index);
960 		else if (q_vector->rx.ring)
961 			sprintf(q_vector->name, "%s-rx-%u", netdev->name,
962 				q_vector->rx.ring->queue_index);
963 		else
964 			sprintf(q_vector->name, "%s-unused", netdev->name);
965 
966 		err = request_irq(adapter->msix_entries[vector].vector,
967 				  igb_msix_ring, 0, q_vector->name,
968 				  q_vector);
969 		if (err)
970 			goto err_free;
971 	}
972 
973 	igb_configure_msix(adapter);
974 	return 0;
975 
976 err_free:
977 	/* free already assigned IRQs */
978 	free_irq(adapter->msix_entries[free_vector++].vector, adapter);
979 
980 	vector--;
981 	for (i = 0; i < vector; i++) {
982 		free_irq(adapter->msix_entries[free_vector++].vector,
983 			 adapter->q_vector[i]);
984 	}
985 err_out:
986 	return err;
987 }
988 
989 /**
990  *  igb_free_q_vector - Free memory allocated for specific interrupt vector
991  *  @adapter: board private structure to initialize
992  *  @v_idx: Index of vector to be freed
993  *
994  *  This function frees the memory allocated to the q_vector.
995  **/
996 static void igb_free_q_vector(struct igb_adapter *adapter, int v_idx)
997 {
998 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
999 
1000 	adapter->q_vector[v_idx] = NULL;
1001 
1002 	/* igb_get_stats64() might access the rings on this vector,
1003 	 * we must wait a grace period before freeing it.
1004 	 */
1005 	if (q_vector)
1006 		kfree_rcu(q_vector, rcu);
1007 }
1008 
1009 /**
1010  *  igb_reset_q_vector - Reset config for interrupt vector
1011  *  @adapter: board private structure to initialize
1012  *  @v_idx: Index of vector to be reset
1013  *
1014  *  If NAPI is enabled it will delete any references to the
1015  *  NAPI struct. This is preparation for igb_free_q_vector.
1016  **/
1017 static void igb_reset_q_vector(struct igb_adapter *adapter, int v_idx)
1018 {
1019 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1020 
1021 	/* Coming from igb_set_interrupt_capability, the vectors are not yet
1022 	 * allocated. So, q_vector is NULL so we should stop here.
1023 	 */
1024 	if (!q_vector)
1025 		return;
1026 
1027 	if (q_vector->tx.ring)
1028 		adapter->tx_ring[q_vector->tx.ring->queue_index] = NULL;
1029 
1030 	if (q_vector->rx.ring)
1031 		adapter->rx_ring[q_vector->rx.ring->queue_index] = NULL;
1032 
1033 	netif_napi_del(&q_vector->napi);
1034 
1035 }
1036 
1037 static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
1038 {
1039 	int v_idx = adapter->num_q_vectors;
1040 
1041 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
1042 		pci_disable_msix(adapter->pdev);
1043 	else if (adapter->flags & IGB_FLAG_HAS_MSI)
1044 		pci_disable_msi(adapter->pdev);
1045 
1046 	while (v_idx--)
1047 		igb_reset_q_vector(adapter, v_idx);
1048 }
1049 
1050 /**
1051  *  igb_free_q_vectors - Free memory allocated for interrupt vectors
1052  *  @adapter: board private structure to initialize
1053  *
1054  *  This function frees the memory allocated to the q_vectors.  In addition if
1055  *  NAPI is enabled it will delete any references to the NAPI struct prior
1056  *  to freeing the q_vector.
1057  **/
1058 static void igb_free_q_vectors(struct igb_adapter *adapter)
1059 {
1060 	int v_idx = adapter->num_q_vectors;
1061 
1062 	adapter->num_tx_queues = 0;
1063 	adapter->num_rx_queues = 0;
1064 	adapter->num_q_vectors = 0;
1065 
1066 	while (v_idx--) {
1067 		igb_reset_q_vector(adapter, v_idx);
1068 		igb_free_q_vector(adapter, v_idx);
1069 	}
1070 }
1071 
1072 /**
1073  *  igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1074  *  @adapter: board private structure to initialize
1075  *
1076  *  This function resets the device so that it has 0 Rx queues, Tx queues, and
1077  *  MSI-X interrupts allocated.
1078  */
1079 static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
1080 {
1081 	igb_free_q_vectors(adapter);
1082 	igb_reset_interrupt_capability(adapter);
1083 }
1084 
1085 /**
1086  *  igb_set_interrupt_capability - set MSI or MSI-X if supported
1087  *  @adapter: board private structure to initialize
1088  *  @msix: boolean value of MSIX capability
1089  *
1090  *  Attempt to configure interrupts using the best available
1091  *  capabilities of the hardware and kernel.
1092  **/
1093 static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
1094 {
1095 	int err;
1096 	int numvecs, i;
1097 
1098 	if (!msix)
1099 		goto msi_only;
1100 	adapter->flags |= IGB_FLAG_HAS_MSIX;
1101 
1102 	/* Number of supported queues. */
1103 	adapter->num_rx_queues = adapter->rss_queues;
1104 	if (adapter->vfs_allocated_count)
1105 		adapter->num_tx_queues = 1;
1106 	else
1107 		adapter->num_tx_queues = adapter->rss_queues;
1108 
1109 	/* start with one vector for every Rx queue */
1110 	numvecs = adapter->num_rx_queues;
1111 
1112 	/* if Tx handler is separate add 1 for every Tx queue */
1113 	if (!(adapter->flags & IGB_FLAG_QUEUE_PAIRS))
1114 		numvecs += adapter->num_tx_queues;
1115 
1116 	/* store the number of vectors reserved for queues */
1117 	adapter->num_q_vectors = numvecs;
1118 
1119 	/* add 1 vector for link status interrupts */
1120 	numvecs++;
1121 	for (i = 0; i < numvecs; i++)
1122 		adapter->msix_entries[i].entry = i;
1123 
1124 	err = pci_enable_msix_range(adapter->pdev,
1125 				    adapter->msix_entries,
1126 				    numvecs,
1127 				    numvecs);
1128 	if (err > 0)
1129 		return;
1130 
1131 	igb_reset_interrupt_capability(adapter);
1132 
1133 	/* If we can't do MSI-X, try MSI */
1134 msi_only:
1135 	adapter->flags &= ~IGB_FLAG_HAS_MSIX;
1136 #ifdef CONFIG_PCI_IOV
1137 	/* disable SR-IOV for non MSI-X configurations */
1138 	if (adapter->vf_data) {
1139 		struct e1000_hw *hw = &adapter->hw;
1140 		/* disable iov and allow time for transactions to clear */
1141 		pci_disable_sriov(adapter->pdev);
1142 		msleep(500);
1143 
1144 		kfree(adapter->vf_mac_list);
1145 		adapter->vf_mac_list = NULL;
1146 		kfree(adapter->vf_data);
1147 		adapter->vf_data = NULL;
1148 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1149 		wrfl();
1150 		msleep(100);
1151 		dev_info(&adapter->pdev->dev, "IOV Disabled\n");
1152 	}
1153 #endif
1154 	adapter->vfs_allocated_count = 0;
1155 	adapter->rss_queues = 1;
1156 	adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
1157 	adapter->num_rx_queues = 1;
1158 	adapter->num_tx_queues = 1;
1159 	adapter->num_q_vectors = 1;
1160 	if (!pci_enable_msi(adapter->pdev))
1161 		adapter->flags |= IGB_FLAG_HAS_MSI;
1162 }
1163 
1164 static void igb_add_ring(struct igb_ring *ring,
1165 			 struct igb_ring_container *head)
1166 {
1167 	head->ring = ring;
1168 	head->count++;
1169 }
1170 
1171 /**
1172  *  igb_alloc_q_vector - Allocate memory for a single interrupt vector
1173  *  @adapter: board private structure to initialize
1174  *  @v_count: q_vectors allocated on adapter, used for ring interleaving
1175  *  @v_idx: index of vector in adapter struct
1176  *  @txr_count: total number of Tx rings to allocate
1177  *  @txr_idx: index of first Tx ring to allocate
1178  *  @rxr_count: total number of Rx rings to allocate
1179  *  @rxr_idx: index of first Rx ring to allocate
1180  *
1181  *  We allocate one q_vector.  If allocation fails we return -ENOMEM.
1182  **/
1183 static int igb_alloc_q_vector(struct igb_adapter *adapter,
1184 			      int v_count, int v_idx,
1185 			      int txr_count, int txr_idx,
1186 			      int rxr_count, int rxr_idx)
1187 {
1188 	struct igb_q_vector *q_vector;
1189 	struct igb_ring *ring;
1190 	int ring_count;
1191 	size_t size;
1192 
1193 	/* igb only supports 1 Tx and/or 1 Rx queue per vector */
1194 	if (txr_count > 1 || rxr_count > 1)
1195 		return -ENOMEM;
1196 
1197 	ring_count = txr_count + rxr_count;
1198 	size = kmalloc_size_roundup(struct_size(q_vector, ring, ring_count));
1199 
1200 	/* allocate q_vector and rings */
1201 	q_vector = adapter->q_vector[v_idx];
1202 	if (!q_vector) {
1203 		q_vector = kzalloc(size, GFP_KERNEL);
1204 	} else if (size > ksize(q_vector)) {
1205 		struct igb_q_vector *new_q_vector;
1206 
1207 		new_q_vector = kzalloc(size, GFP_KERNEL);
1208 		if (new_q_vector)
1209 			kfree_rcu(q_vector, rcu);
1210 		q_vector = new_q_vector;
1211 	} else {
1212 		memset(q_vector, 0, size);
1213 	}
1214 	if (!q_vector)
1215 		return -ENOMEM;
1216 
1217 	/* initialize NAPI */
1218 	netif_napi_add(adapter->netdev, &q_vector->napi, igb_poll);
1219 
1220 	/* tie q_vector and adapter together */
1221 	adapter->q_vector[v_idx] = q_vector;
1222 	q_vector->adapter = adapter;
1223 
1224 	/* initialize work limits */
1225 	q_vector->tx.work_limit = adapter->tx_work_limit;
1226 
1227 	/* initialize ITR configuration */
1228 	q_vector->itr_register = adapter->io_addr + E1000_EITR(0);
1229 	q_vector->itr_val = IGB_START_ITR;
1230 
1231 	/* initialize pointer to rings */
1232 	ring = q_vector->ring;
1233 
1234 	/* intialize ITR */
1235 	if (rxr_count) {
1236 		/* rx or rx/tx vector */
1237 		if (!adapter->rx_itr_setting || adapter->rx_itr_setting > 3)
1238 			q_vector->itr_val = adapter->rx_itr_setting;
1239 	} else {
1240 		/* tx only vector */
1241 		if (!adapter->tx_itr_setting || adapter->tx_itr_setting > 3)
1242 			q_vector->itr_val = adapter->tx_itr_setting;
1243 	}
1244 
1245 	if (txr_count) {
1246 		/* assign generic ring traits */
1247 		ring->dev = &adapter->pdev->dev;
1248 		ring->netdev = adapter->netdev;
1249 
1250 		/* configure backlink on ring */
1251 		ring->q_vector = q_vector;
1252 
1253 		/* update q_vector Tx values */
1254 		igb_add_ring(ring, &q_vector->tx);
1255 
1256 		/* For 82575, context index must be unique per ring. */
1257 		if (adapter->hw.mac.type == e1000_82575)
1258 			set_bit(IGB_RING_FLAG_TX_CTX_IDX, &ring->flags);
1259 
1260 		/* apply Tx specific ring traits */
1261 		ring->count = adapter->tx_ring_count;
1262 		ring->queue_index = txr_idx;
1263 
1264 		ring->cbs_enable = false;
1265 		ring->idleslope = 0;
1266 		ring->sendslope = 0;
1267 		ring->hicredit = 0;
1268 		ring->locredit = 0;
1269 
1270 		u64_stats_init(&ring->tx_syncp);
1271 		u64_stats_init(&ring->tx_syncp2);
1272 
1273 		/* assign ring to adapter */
1274 		adapter->tx_ring[txr_idx] = ring;
1275 
1276 		/* push pointer to next ring */
1277 		ring++;
1278 	}
1279 
1280 	if (rxr_count) {
1281 		/* assign generic ring traits */
1282 		ring->dev = &adapter->pdev->dev;
1283 		ring->netdev = adapter->netdev;
1284 
1285 		/* configure backlink on ring */
1286 		ring->q_vector = q_vector;
1287 
1288 		/* update q_vector Rx values */
1289 		igb_add_ring(ring, &q_vector->rx);
1290 
1291 		/* set flag indicating ring supports SCTP checksum offload */
1292 		if (adapter->hw.mac.type >= e1000_82576)
1293 			set_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags);
1294 
1295 		/* On i350, i354, i210, and i211, loopback VLAN packets
1296 		 * have the tag byte-swapped.
1297 		 */
1298 		if (adapter->hw.mac.type >= e1000_i350)
1299 			set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &ring->flags);
1300 
1301 		/* apply Rx specific ring traits */
1302 		ring->count = adapter->rx_ring_count;
1303 		ring->queue_index = rxr_idx;
1304 
1305 		u64_stats_init(&ring->rx_syncp);
1306 
1307 		/* assign ring to adapter */
1308 		adapter->rx_ring[rxr_idx] = ring;
1309 	}
1310 
1311 	return 0;
1312 }
1313 
1314 
1315 /**
1316  *  igb_alloc_q_vectors - Allocate memory for interrupt vectors
1317  *  @adapter: board private structure to initialize
1318  *
1319  *  We allocate one q_vector per queue interrupt.  If allocation fails we
1320  *  return -ENOMEM.
1321  **/
1322 static int igb_alloc_q_vectors(struct igb_adapter *adapter)
1323 {
1324 	int q_vectors = adapter->num_q_vectors;
1325 	int rxr_remaining = adapter->num_rx_queues;
1326 	int txr_remaining = adapter->num_tx_queues;
1327 	int rxr_idx = 0, txr_idx = 0, v_idx = 0;
1328 	int err;
1329 
1330 	if (q_vectors >= (rxr_remaining + txr_remaining)) {
1331 		for (; rxr_remaining; v_idx++) {
1332 			err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1333 						 0, 0, 1, rxr_idx);
1334 
1335 			if (err)
1336 				goto err_out;
1337 
1338 			/* update counts and index */
1339 			rxr_remaining--;
1340 			rxr_idx++;
1341 		}
1342 	}
1343 
1344 	for (; v_idx < q_vectors; v_idx++) {
1345 		int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
1346 		int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);
1347 
1348 		err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1349 					 tqpv, txr_idx, rqpv, rxr_idx);
1350 
1351 		if (err)
1352 			goto err_out;
1353 
1354 		/* update counts and index */
1355 		rxr_remaining -= rqpv;
1356 		txr_remaining -= tqpv;
1357 		rxr_idx++;
1358 		txr_idx++;
1359 	}
1360 
1361 	return 0;
1362 
1363 err_out:
1364 	adapter->num_tx_queues = 0;
1365 	adapter->num_rx_queues = 0;
1366 	adapter->num_q_vectors = 0;
1367 
1368 	while (v_idx--)
1369 		igb_free_q_vector(adapter, v_idx);
1370 
1371 	return -ENOMEM;
1372 }
1373 
1374 /**
1375  *  igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1376  *  @adapter: board private structure to initialize
1377  *  @msix: boolean value of MSIX capability
1378  *
1379  *  This function initializes the interrupts and allocates all of the queues.
1380  **/
1381 static int igb_init_interrupt_scheme(struct igb_adapter *adapter, bool msix)
1382 {
1383 	struct pci_dev *pdev = adapter->pdev;
1384 	int err;
1385 
1386 	igb_set_interrupt_capability(adapter, msix);
1387 
1388 	err = igb_alloc_q_vectors(adapter);
1389 	if (err) {
1390 		dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
1391 		goto err_alloc_q_vectors;
1392 	}
1393 
1394 	igb_cache_ring_register(adapter);
1395 
1396 	return 0;
1397 
1398 err_alloc_q_vectors:
1399 	igb_reset_interrupt_capability(adapter);
1400 	return err;
1401 }
1402 
1403 /**
1404  *  igb_request_irq - initialize interrupts
1405  *  @adapter: board private structure to initialize
1406  *
1407  *  Attempts to configure interrupts using the best available
1408  *  capabilities of the hardware and kernel.
1409  **/
1410 static int igb_request_irq(struct igb_adapter *adapter)
1411 {
1412 	struct net_device *netdev = adapter->netdev;
1413 	struct pci_dev *pdev = adapter->pdev;
1414 	int err = 0;
1415 
1416 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1417 		err = igb_request_msix(adapter);
1418 		if (!err)
1419 			goto request_done;
1420 		/* fall back to MSI */
1421 		igb_free_all_tx_resources(adapter);
1422 		igb_free_all_rx_resources(adapter);
1423 
1424 		igb_clear_interrupt_scheme(adapter);
1425 		err = igb_init_interrupt_scheme(adapter, false);
1426 		if (err)
1427 			goto request_done;
1428 
1429 		igb_setup_all_tx_resources(adapter);
1430 		igb_setup_all_rx_resources(adapter);
1431 		igb_configure(adapter);
1432 	}
1433 
1434 	igb_assign_vector(adapter->q_vector[0], 0);
1435 
1436 	if (adapter->flags & IGB_FLAG_HAS_MSI) {
1437 		err = request_irq(pdev->irq, igb_intr_msi, 0,
1438 				  netdev->name, adapter);
1439 		if (!err)
1440 			goto request_done;
1441 
1442 		/* fall back to legacy interrupts */
1443 		igb_reset_interrupt_capability(adapter);
1444 		adapter->flags &= ~IGB_FLAG_HAS_MSI;
1445 	}
1446 
1447 	err = request_irq(pdev->irq, igb_intr, IRQF_SHARED,
1448 			  netdev->name, adapter);
1449 
1450 	if (err)
1451 		dev_err(&pdev->dev, "Error %d getting interrupt\n",
1452 			err);
1453 
1454 request_done:
1455 	return err;
1456 }
1457 
1458 static void igb_free_irq(struct igb_adapter *adapter)
1459 {
1460 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1461 		int vector = 0, i;
1462 
1463 		free_irq(adapter->msix_entries[vector++].vector, adapter);
1464 
1465 		for (i = 0; i < adapter->num_q_vectors; i++)
1466 			free_irq(adapter->msix_entries[vector++].vector,
1467 				 adapter->q_vector[i]);
1468 	} else {
1469 		free_irq(adapter->pdev->irq, adapter);
1470 	}
1471 }
1472 
1473 /**
1474  *  igb_irq_disable - Mask off interrupt generation on the NIC
1475  *  @adapter: board private structure
1476  **/
1477 static void igb_irq_disable(struct igb_adapter *adapter)
1478 {
1479 	struct e1000_hw *hw = &adapter->hw;
1480 
1481 	/* we need to be careful when disabling interrupts.  The VFs are also
1482 	 * mapped into these registers and so clearing the bits can cause
1483 	 * issues on the VF drivers so we only need to clear what we set
1484 	 */
1485 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1486 		u32 regval = rd32(E1000_EIAM);
1487 
1488 		wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
1489 		wr32(E1000_EIMC, adapter->eims_enable_mask);
1490 		regval = rd32(E1000_EIAC);
1491 		wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
1492 	}
1493 
1494 	wr32(E1000_IAM, 0);
1495 	wr32(E1000_IMC, ~0);
1496 	wrfl();
1497 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1498 		int i;
1499 
1500 		for (i = 0; i < adapter->num_q_vectors; i++)
1501 			synchronize_irq(adapter->msix_entries[i].vector);
1502 	} else {
1503 		synchronize_irq(adapter->pdev->irq);
1504 	}
1505 }
1506 
1507 /**
1508  *  igb_irq_enable - Enable default interrupt generation settings
1509  *  @adapter: board private structure
1510  **/
1511 static void igb_irq_enable(struct igb_adapter *adapter)
1512 {
1513 	struct e1000_hw *hw = &adapter->hw;
1514 
1515 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1516 		u32 ims = E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_DRSTA;
1517 		u32 regval = rd32(E1000_EIAC);
1518 
1519 		wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
1520 		regval = rd32(E1000_EIAM);
1521 		wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
1522 		wr32(E1000_EIMS, adapter->eims_enable_mask);
1523 		if (adapter->vfs_allocated_count) {
1524 			wr32(E1000_MBVFIMR, 0xFF);
1525 			ims |= E1000_IMS_VMMB;
1526 		}
1527 		wr32(E1000_IMS, ims);
1528 	} else {
1529 		wr32(E1000_IMS, IMS_ENABLE_MASK |
1530 				E1000_IMS_DRSTA);
1531 		wr32(E1000_IAM, IMS_ENABLE_MASK |
1532 				E1000_IMS_DRSTA);
1533 	}
1534 }
1535 
1536 static void igb_update_mng_vlan(struct igb_adapter *adapter)
1537 {
1538 	struct e1000_hw *hw = &adapter->hw;
1539 	u16 pf_id = adapter->vfs_allocated_count;
1540 	u16 vid = adapter->hw.mng_cookie.vlan_id;
1541 	u16 old_vid = adapter->mng_vlan_id;
1542 
1543 	if (hw->mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1544 		/* add VID to filter table */
1545 		igb_vfta_set(hw, vid, pf_id, true, true);
1546 		adapter->mng_vlan_id = vid;
1547 	} else {
1548 		adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1549 	}
1550 
1551 	if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1552 	    (vid != old_vid) &&
1553 	    !test_bit(old_vid, adapter->active_vlans)) {
1554 		/* remove VID from filter table */
1555 		igb_vfta_set(hw, vid, pf_id, false, true);
1556 	}
1557 }
1558 
1559 /**
1560  *  igb_release_hw_control - release control of the h/w to f/w
1561  *  @adapter: address of board private structure
1562  *
1563  *  igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1564  *  For ASF and Pass Through versions of f/w this means that the
1565  *  driver is no longer loaded.
1566  **/
1567 static void igb_release_hw_control(struct igb_adapter *adapter)
1568 {
1569 	struct e1000_hw *hw = &adapter->hw;
1570 	u32 ctrl_ext;
1571 
1572 	/* Let firmware take over control of h/w */
1573 	ctrl_ext = rd32(E1000_CTRL_EXT);
1574 	wr32(E1000_CTRL_EXT,
1575 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1576 }
1577 
1578 /**
1579  *  igb_get_hw_control - get control of the h/w from f/w
1580  *  @adapter: address of board private structure
1581  *
1582  *  igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1583  *  For ASF and Pass Through versions of f/w this means that
1584  *  the driver is loaded.
1585  **/
1586 static void igb_get_hw_control(struct igb_adapter *adapter)
1587 {
1588 	struct e1000_hw *hw = &adapter->hw;
1589 	u32 ctrl_ext;
1590 
1591 	/* Let firmware know the driver has taken over */
1592 	ctrl_ext = rd32(E1000_CTRL_EXT);
1593 	wr32(E1000_CTRL_EXT,
1594 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1595 }
1596 
1597 static void enable_fqtss(struct igb_adapter *adapter, bool enable)
1598 {
1599 	struct net_device *netdev = adapter->netdev;
1600 	struct e1000_hw *hw = &adapter->hw;
1601 
1602 	WARN_ON(hw->mac.type != e1000_i210);
1603 
1604 	if (enable)
1605 		adapter->flags |= IGB_FLAG_FQTSS;
1606 	else
1607 		adapter->flags &= ~IGB_FLAG_FQTSS;
1608 
1609 	if (netif_running(netdev))
1610 		schedule_work(&adapter->reset_task);
1611 }
1612 
1613 static bool is_fqtss_enabled(struct igb_adapter *adapter)
1614 {
1615 	return (adapter->flags & IGB_FLAG_FQTSS) ? true : false;
1616 }
1617 
1618 static void set_tx_desc_fetch_prio(struct e1000_hw *hw, int queue,
1619 				   enum tx_queue_prio prio)
1620 {
1621 	u32 val;
1622 
1623 	WARN_ON(hw->mac.type != e1000_i210);
1624 	WARN_ON(queue < 0 || queue > 4);
1625 
1626 	val = rd32(E1000_I210_TXDCTL(queue));
1627 
1628 	if (prio == TX_QUEUE_PRIO_HIGH)
1629 		val |= E1000_TXDCTL_PRIORITY;
1630 	else
1631 		val &= ~E1000_TXDCTL_PRIORITY;
1632 
1633 	wr32(E1000_I210_TXDCTL(queue), val);
1634 }
1635 
1636 static void set_queue_mode(struct e1000_hw *hw, int queue, enum queue_mode mode)
1637 {
1638 	u32 val;
1639 
1640 	WARN_ON(hw->mac.type != e1000_i210);
1641 	WARN_ON(queue < 0 || queue > 1);
1642 
1643 	val = rd32(E1000_I210_TQAVCC(queue));
1644 
1645 	if (mode == QUEUE_MODE_STREAM_RESERVATION)
1646 		val |= E1000_TQAVCC_QUEUEMODE;
1647 	else
1648 		val &= ~E1000_TQAVCC_QUEUEMODE;
1649 
1650 	wr32(E1000_I210_TQAVCC(queue), val);
1651 }
1652 
1653 static bool is_any_cbs_enabled(struct igb_adapter *adapter)
1654 {
1655 	int i;
1656 
1657 	for (i = 0; i < adapter->num_tx_queues; i++) {
1658 		if (adapter->tx_ring[i]->cbs_enable)
1659 			return true;
1660 	}
1661 
1662 	return false;
1663 }
1664 
1665 static bool is_any_txtime_enabled(struct igb_adapter *adapter)
1666 {
1667 	int i;
1668 
1669 	for (i = 0; i < adapter->num_tx_queues; i++) {
1670 		if (adapter->tx_ring[i]->launchtime_enable)
1671 			return true;
1672 	}
1673 
1674 	return false;
1675 }
1676 
1677 /**
1678  *  igb_config_tx_modes - Configure "Qav Tx mode" features on igb
1679  *  @adapter: pointer to adapter struct
1680  *  @queue: queue number
1681  *
1682  *  Configure CBS and Launchtime for a given hardware queue.
1683  *  Parameters are retrieved from the correct Tx ring, so
1684  *  igb_save_cbs_params() and igb_save_txtime_params() should be used
1685  *  for setting those correctly prior to this function being called.
1686  **/
1687 static void igb_config_tx_modes(struct igb_adapter *adapter, int queue)
1688 {
1689 	struct net_device *netdev = adapter->netdev;
1690 	struct e1000_hw *hw = &adapter->hw;
1691 	struct igb_ring *ring;
1692 	u32 tqavcc, tqavctrl;
1693 	u16 value;
1694 
1695 	WARN_ON(hw->mac.type != e1000_i210);
1696 	WARN_ON(queue < 0 || queue > 1);
1697 	ring = adapter->tx_ring[queue];
1698 
1699 	/* If any of the Qav features is enabled, configure queues as SR and
1700 	 * with HIGH PRIO. If none is, then configure them with LOW PRIO and
1701 	 * as SP.
1702 	 */
1703 	if (ring->cbs_enable || ring->launchtime_enable) {
1704 		set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_HIGH);
1705 		set_queue_mode(hw, queue, QUEUE_MODE_STREAM_RESERVATION);
1706 	} else {
1707 		set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_LOW);
1708 		set_queue_mode(hw, queue, QUEUE_MODE_STRICT_PRIORITY);
1709 	}
1710 
1711 	/* If CBS is enabled, set DataTranARB and config its parameters. */
1712 	if (ring->cbs_enable || queue == 0) {
1713 		/* i210 does not allow the queue 0 to be in the Strict
1714 		 * Priority mode while the Qav mode is enabled, so,
1715 		 * instead of disabling strict priority mode, we give
1716 		 * queue 0 the maximum of credits possible.
1717 		 *
1718 		 * See section 8.12.19 of the i210 datasheet, "Note:
1719 		 * Queue0 QueueMode must be set to 1b when
1720 		 * TransmitMode is set to Qav."
1721 		 */
1722 		if (queue == 0 && !ring->cbs_enable) {
1723 			/* max "linkspeed" idleslope in kbps */
1724 			ring->idleslope = 1000000;
1725 			ring->hicredit = ETH_FRAME_LEN;
1726 		}
1727 
1728 		/* Always set data transfer arbitration to credit-based
1729 		 * shaper algorithm on TQAVCTRL if CBS is enabled for any of
1730 		 * the queues.
1731 		 */
1732 		tqavctrl = rd32(E1000_I210_TQAVCTRL);
1733 		tqavctrl |= E1000_TQAVCTRL_DATATRANARB;
1734 		wr32(E1000_I210_TQAVCTRL, tqavctrl);
1735 
1736 		/* According to i210 datasheet section 7.2.7.7, we should set
1737 		 * the 'idleSlope' field from TQAVCC register following the
1738 		 * equation:
1739 		 *
1740 		 * For 100 Mbps link speed:
1741 		 *
1742 		 *     value = BW * 0x7735 * 0.2                          (E1)
1743 		 *
1744 		 * For 1000Mbps link speed:
1745 		 *
1746 		 *     value = BW * 0x7735 * 2                            (E2)
1747 		 *
1748 		 * E1 and E2 can be merged into one equation as shown below.
1749 		 * Note that 'link-speed' is in Mbps.
1750 		 *
1751 		 *     value = BW * 0x7735 * 2 * link-speed
1752 		 *                           --------------               (E3)
1753 		 *                                1000
1754 		 *
1755 		 * 'BW' is the percentage bandwidth out of full link speed
1756 		 * which can be found with the following equation. Note that
1757 		 * idleSlope here is the parameter from this function which
1758 		 * is in kbps.
1759 		 *
1760 		 *     BW =     idleSlope
1761 		 *          -----------------                             (E4)
1762 		 *          link-speed * 1000
1763 		 *
1764 		 * That said, we can come up with a generic equation to
1765 		 * calculate the value we should set it TQAVCC register by
1766 		 * replacing 'BW' in E3 by E4. The resulting equation is:
1767 		 *
1768 		 * value =     idleSlope     * 0x7735 * 2 * link-speed
1769 		 *         -----------------            --------------    (E5)
1770 		 *         link-speed * 1000                 1000
1771 		 *
1772 		 * 'link-speed' is present in both sides of the fraction so
1773 		 * it is canceled out. The final equation is the following:
1774 		 *
1775 		 *     value = idleSlope * 61034
1776 		 *             -----------------                          (E6)
1777 		 *                  1000000
1778 		 *
1779 		 * NOTE: For i210, given the above, we can see that idleslope
1780 		 *       is represented in 16.38431 kbps units by the value at
1781 		 *       the TQAVCC register (1Gbps / 61034), which reduces
1782 		 *       the granularity for idleslope increments.
1783 		 *       For instance, if you want to configure a 2576kbps
1784 		 *       idleslope, the value to be written on the register
1785 		 *       would have to be 157.23. If rounded down, you end
1786 		 *       up with less bandwidth available than originally
1787 		 *       required (~2572 kbps). If rounded up, you end up
1788 		 *       with a higher bandwidth (~2589 kbps). Below the
1789 		 *       approach we take is to always round up the
1790 		 *       calculated value, so the resulting bandwidth might
1791 		 *       be slightly higher for some configurations.
1792 		 */
1793 		value = DIV_ROUND_UP_ULL(ring->idleslope * 61034ULL, 1000000);
1794 
1795 		tqavcc = rd32(E1000_I210_TQAVCC(queue));
1796 		tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1797 		tqavcc |= value;
1798 		wr32(E1000_I210_TQAVCC(queue), tqavcc);
1799 
1800 		wr32(E1000_I210_TQAVHC(queue),
1801 		     0x80000000 + ring->hicredit * 0x7735);
1802 	} else {
1803 
1804 		/* Set idleSlope to zero. */
1805 		tqavcc = rd32(E1000_I210_TQAVCC(queue));
1806 		tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1807 		wr32(E1000_I210_TQAVCC(queue), tqavcc);
1808 
1809 		/* Set hiCredit to zero. */
1810 		wr32(E1000_I210_TQAVHC(queue), 0);
1811 
1812 		/* If CBS is not enabled for any queues anymore, then return to
1813 		 * the default state of Data Transmission Arbitration on
1814 		 * TQAVCTRL.
1815 		 */
1816 		if (!is_any_cbs_enabled(adapter)) {
1817 			tqavctrl = rd32(E1000_I210_TQAVCTRL);
1818 			tqavctrl &= ~E1000_TQAVCTRL_DATATRANARB;
1819 			wr32(E1000_I210_TQAVCTRL, tqavctrl);
1820 		}
1821 	}
1822 
1823 	/* If LaunchTime is enabled, set DataTranTIM. */
1824 	if (ring->launchtime_enable) {
1825 		/* Always set DataTranTIM on TQAVCTRL if LaunchTime is enabled
1826 		 * for any of the SR queues, and configure fetchtime delta.
1827 		 * XXX NOTE:
1828 		 *     - LaunchTime will be enabled for all SR queues.
1829 		 *     - A fixed offset can be added relative to the launch
1830 		 *       time of all packets if configured at reg LAUNCH_OS0.
1831 		 *       We are keeping it as 0 for now (default value).
1832 		 */
1833 		tqavctrl = rd32(E1000_I210_TQAVCTRL);
1834 		tqavctrl |= E1000_TQAVCTRL_DATATRANTIM |
1835 		       E1000_TQAVCTRL_FETCHTIME_DELTA;
1836 		wr32(E1000_I210_TQAVCTRL, tqavctrl);
1837 	} else {
1838 		/* If Launchtime is not enabled for any SR queues anymore,
1839 		 * then clear DataTranTIM on TQAVCTRL and clear fetchtime delta,
1840 		 * effectively disabling Launchtime.
1841 		 */
1842 		if (!is_any_txtime_enabled(adapter)) {
1843 			tqavctrl = rd32(E1000_I210_TQAVCTRL);
1844 			tqavctrl &= ~E1000_TQAVCTRL_DATATRANTIM;
1845 			tqavctrl &= ~E1000_TQAVCTRL_FETCHTIME_DELTA;
1846 			wr32(E1000_I210_TQAVCTRL, tqavctrl);
1847 		}
1848 	}
1849 
1850 	/* XXX: In i210 controller the sendSlope and loCredit parameters from
1851 	 * CBS are not configurable by software so we don't do any 'controller
1852 	 * configuration' in respect to these parameters.
1853 	 */
1854 
1855 	netdev_dbg(netdev, "Qav Tx mode: cbs %s, launchtime %s, queue %d idleslope %d sendslope %d hiCredit %d locredit %d\n",
1856 		   ring->cbs_enable ? "enabled" : "disabled",
1857 		   ring->launchtime_enable ? "enabled" : "disabled",
1858 		   queue,
1859 		   ring->idleslope, ring->sendslope,
1860 		   ring->hicredit, ring->locredit);
1861 }
1862 
1863 static int igb_save_txtime_params(struct igb_adapter *adapter, int queue,
1864 				  bool enable)
1865 {
1866 	struct igb_ring *ring;
1867 
1868 	if (queue < 0 || queue > adapter->num_tx_queues)
1869 		return -EINVAL;
1870 
1871 	ring = adapter->tx_ring[queue];
1872 	ring->launchtime_enable = enable;
1873 
1874 	return 0;
1875 }
1876 
1877 static int igb_save_cbs_params(struct igb_adapter *adapter, int queue,
1878 			       bool enable, int idleslope, int sendslope,
1879 			       int hicredit, int locredit)
1880 {
1881 	struct igb_ring *ring;
1882 
1883 	if (queue < 0 || queue > adapter->num_tx_queues)
1884 		return -EINVAL;
1885 
1886 	ring = adapter->tx_ring[queue];
1887 
1888 	ring->cbs_enable = enable;
1889 	ring->idleslope = idleslope;
1890 	ring->sendslope = sendslope;
1891 	ring->hicredit = hicredit;
1892 	ring->locredit = locredit;
1893 
1894 	return 0;
1895 }
1896 
1897 /**
1898  *  igb_setup_tx_mode - Switch to/from Qav Tx mode when applicable
1899  *  @adapter: pointer to adapter struct
1900  *
1901  *  Configure TQAVCTRL register switching the controller's Tx mode
1902  *  if FQTSS mode is enabled or disabled. Additionally, will issue
1903  *  a call to igb_config_tx_modes() per queue so any previously saved
1904  *  Tx parameters are applied.
1905  **/
1906 static void igb_setup_tx_mode(struct igb_adapter *adapter)
1907 {
1908 	struct net_device *netdev = adapter->netdev;
1909 	struct e1000_hw *hw = &adapter->hw;
1910 	u32 val;
1911 
1912 	/* Only i210 controller supports changing the transmission mode. */
1913 	if (hw->mac.type != e1000_i210)
1914 		return;
1915 
1916 	if (is_fqtss_enabled(adapter)) {
1917 		int i, max_queue;
1918 
1919 		/* Configure TQAVCTRL register: set transmit mode to 'Qav',
1920 		 * set data fetch arbitration to 'round robin', set SP_WAIT_SR
1921 		 * so SP queues wait for SR ones.
1922 		 */
1923 		val = rd32(E1000_I210_TQAVCTRL);
1924 		val |= E1000_TQAVCTRL_XMIT_MODE | E1000_TQAVCTRL_SP_WAIT_SR;
1925 		val &= ~E1000_TQAVCTRL_DATAFETCHARB;
1926 		wr32(E1000_I210_TQAVCTRL, val);
1927 
1928 		/* Configure Tx and Rx packet buffers sizes as described in
1929 		 * i210 datasheet section 7.2.7.7.
1930 		 */
1931 		val = rd32(E1000_TXPBS);
1932 		val &= ~I210_TXPBSIZE_MASK;
1933 		val |= I210_TXPBSIZE_PB0_6KB | I210_TXPBSIZE_PB1_6KB |
1934 			I210_TXPBSIZE_PB2_6KB | I210_TXPBSIZE_PB3_6KB;
1935 		wr32(E1000_TXPBS, val);
1936 
1937 		val = rd32(E1000_RXPBS);
1938 		val &= ~I210_RXPBSIZE_MASK;
1939 		val |= I210_RXPBSIZE_PB_30KB;
1940 		wr32(E1000_RXPBS, val);
1941 
1942 		/* Section 8.12.9 states that MAX_TPKT_SIZE from DTXMXPKTSZ
1943 		 * register should not exceed the buffer size programmed in
1944 		 * TXPBS. The smallest buffer size programmed in TXPBS is 4kB
1945 		 * so according to the datasheet we should set MAX_TPKT_SIZE to
1946 		 * 4kB / 64.
1947 		 *
1948 		 * However, when we do so, no frame from queue 2 and 3 are
1949 		 * transmitted.  It seems the MAX_TPKT_SIZE should not be great
1950 		 * or _equal_ to the buffer size programmed in TXPBS. For this
1951 		 * reason, we set MAX_ TPKT_SIZE to (4kB - 1) / 64.
1952 		 */
1953 		val = (4096 - 1) / 64;
1954 		wr32(E1000_I210_DTXMXPKTSZ, val);
1955 
1956 		/* Since FQTSS mode is enabled, apply any CBS configuration
1957 		 * previously set. If no previous CBS configuration has been
1958 		 * done, then the initial configuration is applied, which means
1959 		 * CBS is disabled.
1960 		 */
1961 		max_queue = (adapter->num_tx_queues < I210_SR_QUEUES_NUM) ?
1962 			    adapter->num_tx_queues : I210_SR_QUEUES_NUM;
1963 
1964 		for (i = 0; i < max_queue; i++) {
1965 			igb_config_tx_modes(adapter, i);
1966 		}
1967 	} else {
1968 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
1969 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
1970 		wr32(E1000_I210_DTXMXPKTSZ, I210_DTXMXPKTSZ_DEFAULT);
1971 
1972 		val = rd32(E1000_I210_TQAVCTRL);
1973 		/* According to Section 8.12.21, the other flags we've set when
1974 		 * enabling FQTSS are not relevant when disabling FQTSS so we
1975 		 * don't set they here.
1976 		 */
1977 		val &= ~E1000_TQAVCTRL_XMIT_MODE;
1978 		wr32(E1000_I210_TQAVCTRL, val);
1979 	}
1980 
1981 	netdev_dbg(netdev, "FQTSS %s\n", (is_fqtss_enabled(adapter)) ?
1982 		   "enabled" : "disabled");
1983 }
1984 
1985 /**
1986  *  igb_configure - configure the hardware for RX and TX
1987  *  @adapter: private board structure
1988  **/
1989 static void igb_configure(struct igb_adapter *adapter)
1990 {
1991 	struct net_device *netdev = adapter->netdev;
1992 	int i;
1993 
1994 	igb_get_hw_control(adapter);
1995 	igb_set_rx_mode(netdev);
1996 	igb_setup_tx_mode(adapter);
1997 
1998 	igb_restore_vlan(adapter);
1999 
2000 	igb_setup_tctl(adapter);
2001 	igb_setup_mrqc(adapter);
2002 	igb_setup_rctl(adapter);
2003 
2004 	igb_nfc_filter_restore(adapter);
2005 	igb_configure_tx(adapter);
2006 	igb_configure_rx(adapter);
2007 
2008 	igb_rx_fifo_flush_82575(&adapter->hw);
2009 
2010 	/* call igb_desc_unused which always leaves
2011 	 * at least 1 descriptor unused to make sure
2012 	 * next_to_use != next_to_clean
2013 	 */
2014 	for (i = 0; i < adapter->num_rx_queues; i++) {
2015 		struct igb_ring *ring = adapter->rx_ring[i];
2016 		igb_alloc_rx_buffers(ring, igb_desc_unused(ring));
2017 	}
2018 }
2019 
2020 /**
2021  *  igb_power_up_link - Power up the phy/serdes link
2022  *  @adapter: address of board private structure
2023  **/
2024 void igb_power_up_link(struct igb_adapter *adapter)
2025 {
2026 	igb_reset_phy(&adapter->hw);
2027 
2028 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
2029 		igb_power_up_phy_copper(&adapter->hw);
2030 	else
2031 		igb_power_up_serdes_link_82575(&adapter->hw);
2032 
2033 	igb_setup_link(&adapter->hw);
2034 }
2035 
2036 /**
2037  *  igb_power_down_link - Power down the phy/serdes link
2038  *  @adapter: address of board private structure
2039  */
2040 static void igb_power_down_link(struct igb_adapter *adapter)
2041 {
2042 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
2043 		igb_power_down_phy_copper_82575(&adapter->hw);
2044 	else
2045 		igb_shutdown_serdes_link_82575(&adapter->hw);
2046 }
2047 
2048 /**
2049  * igb_check_swap_media -  Detect and switch function for Media Auto Sense
2050  * @adapter: address of the board private structure
2051  **/
2052 static void igb_check_swap_media(struct igb_adapter *adapter)
2053 {
2054 	struct e1000_hw *hw = &adapter->hw;
2055 	u32 ctrl_ext, connsw;
2056 	bool swap_now = false;
2057 
2058 	ctrl_ext = rd32(E1000_CTRL_EXT);
2059 	connsw = rd32(E1000_CONNSW);
2060 
2061 	/* need to live swap if current media is copper and we have fiber/serdes
2062 	 * to go to.
2063 	 */
2064 
2065 	if ((hw->phy.media_type == e1000_media_type_copper) &&
2066 	    (!(connsw & E1000_CONNSW_AUTOSENSE_EN))) {
2067 		swap_now = true;
2068 	} else if ((hw->phy.media_type != e1000_media_type_copper) &&
2069 		   !(connsw & E1000_CONNSW_SERDESD)) {
2070 		/* copper signal takes time to appear */
2071 		if (adapter->copper_tries < 4) {
2072 			adapter->copper_tries++;
2073 			connsw |= E1000_CONNSW_AUTOSENSE_CONF;
2074 			wr32(E1000_CONNSW, connsw);
2075 			return;
2076 		} else {
2077 			adapter->copper_tries = 0;
2078 			if ((connsw & E1000_CONNSW_PHYSD) &&
2079 			    (!(connsw & E1000_CONNSW_PHY_PDN))) {
2080 				swap_now = true;
2081 				connsw &= ~E1000_CONNSW_AUTOSENSE_CONF;
2082 				wr32(E1000_CONNSW, connsw);
2083 			}
2084 		}
2085 	}
2086 
2087 	if (!swap_now)
2088 		return;
2089 
2090 	switch (hw->phy.media_type) {
2091 	case e1000_media_type_copper:
2092 		netdev_info(adapter->netdev,
2093 			"MAS: changing media to fiber/serdes\n");
2094 		ctrl_ext |=
2095 			E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2096 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
2097 		adapter->copper_tries = 0;
2098 		break;
2099 	case e1000_media_type_internal_serdes:
2100 	case e1000_media_type_fiber:
2101 		netdev_info(adapter->netdev,
2102 			"MAS: changing media to copper\n");
2103 		ctrl_ext &=
2104 			~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2105 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
2106 		break;
2107 	default:
2108 		/* shouldn't get here during regular operation */
2109 		netdev_err(adapter->netdev,
2110 			"AMS: Invalid media type found, returning\n");
2111 		break;
2112 	}
2113 	wr32(E1000_CTRL_EXT, ctrl_ext);
2114 }
2115 
2116 /**
2117  *  igb_up - Open the interface and prepare it to handle traffic
2118  *  @adapter: board private structure
2119  **/
2120 int igb_up(struct igb_adapter *adapter)
2121 {
2122 	struct e1000_hw *hw = &adapter->hw;
2123 	int i;
2124 
2125 	/* hardware has been reset, we need to reload some things */
2126 	igb_configure(adapter);
2127 
2128 	clear_bit(__IGB_DOWN, &adapter->state);
2129 
2130 	for (i = 0; i < adapter->num_q_vectors; i++)
2131 		napi_enable(&(adapter->q_vector[i]->napi));
2132 
2133 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
2134 		igb_configure_msix(adapter);
2135 	else
2136 		igb_assign_vector(adapter->q_vector[0], 0);
2137 
2138 	/* Clear any pending interrupts. */
2139 	rd32(E1000_TSICR);
2140 	rd32(E1000_ICR);
2141 	igb_irq_enable(adapter);
2142 
2143 	/* notify VFs that reset has been completed */
2144 	if (adapter->vfs_allocated_count) {
2145 		u32 reg_data = rd32(E1000_CTRL_EXT);
2146 
2147 		reg_data |= E1000_CTRL_EXT_PFRSTD;
2148 		wr32(E1000_CTRL_EXT, reg_data);
2149 	}
2150 
2151 	netif_tx_start_all_queues(adapter->netdev);
2152 
2153 	/* start the watchdog. */
2154 	hw->mac.get_link_status = 1;
2155 	schedule_work(&adapter->watchdog_task);
2156 
2157 	if ((adapter->flags & IGB_FLAG_EEE) &&
2158 	    (!hw->dev_spec._82575.eee_disable))
2159 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
2160 
2161 	return 0;
2162 }
2163 
2164 void igb_down(struct igb_adapter *adapter)
2165 {
2166 	struct net_device *netdev = adapter->netdev;
2167 	struct e1000_hw *hw = &adapter->hw;
2168 	u32 tctl, rctl;
2169 	int i;
2170 
2171 	/* signal that we're down so the interrupt handler does not
2172 	 * reschedule our watchdog timer
2173 	 */
2174 	set_bit(__IGB_DOWN, &adapter->state);
2175 
2176 	/* disable receives in the hardware */
2177 	rctl = rd32(E1000_RCTL);
2178 	wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
2179 	/* flush and sleep below */
2180 
2181 	igb_nfc_filter_exit(adapter);
2182 
2183 	netif_carrier_off(netdev);
2184 	netif_tx_stop_all_queues(netdev);
2185 
2186 	/* disable transmits in the hardware */
2187 	tctl = rd32(E1000_TCTL);
2188 	tctl &= ~E1000_TCTL_EN;
2189 	wr32(E1000_TCTL, tctl);
2190 	/* flush both disables and wait for them to finish */
2191 	wrfl();
2192 	usleep_range(10000, 11000);
2193 
2194 	igb_irq_disable(adapter);
2195 
2196 	adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
2197 
2198 	for (i = 0; i < adapter->num_q_vectors; i++) {
2199 		if (adapter->q_vector[i]) {
2200 			napi_synchronize(&adapter->q_vector[i]->napi);
2201 			napi_disable(&adapter->q_vector[i]->napi);
2202 		}
2203 	}
2204 
2205 	del_timer_sync(&adapter->watchdog_timer);
2206 	del_timer_sync(&adapter->phy_info_timer);
2207 
2208 	/* record the stats before reset*/
2209 	spin_lock(&adapter->stats64_lock);
2210 	igb_update_stats(adapter);
2211 	spin_unlock(&adapter->stats64_lock);
2212 
2213 	adapter->link_speed = 0;
2214 	adapter->link_duplex = 0;
2215 
2216 	if (!pci_channel_offline(adapter->pdev))
2217 		igb_reset(adapter);
2218 
2219 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
2220 	adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
2221 
2222 	igb_clean_all_tx_rings(adapter);
2223 	igb_clean_all_rx_rings(adapter);
2224 #ifdef CONFIG_IGB_DCA
2225 
2226 	/* since we reset the hardware DCA settings were cleared */
2227 	igb_setup_dca(adapter);
2228 #endif
2229 }
2230 
2231 void igb_reinit_locked(struct igb_adapter *adapter)
2232 {
2233 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
2234 		usleep_range(1000, 2000);
2235 	igb_down(adapter);
2236 	igb_up(adapter);
2237 	clear_bit(__IGB_RESETTING, &adapter->state);
2238 }
2239 
2240 /** igb_enable_mas - Media Autosense re-enable after swap
2241  *
2242  * @adapter: adapter struct
2243  **/
2244 static void igb_enable_mas(struct igb_adapter *adapter)
2245 {
2246 	struct e1000_hw *hw = &adapter->hw;
2247 	u32 connsw = rd32(E1000_CONNSW);
2248 
2249 	/* configure for SerDes media detect */
2250 	if ((hw->phy.media_type == e1000_media_type_copper) &&
2251 	    (!(connsw & E1000_CONNSW_SERDESD))) {
2252 		connsw |= E1000_CONNSW_ENRGSRC;
2253 		connsw |= E1000_CONNSW_AUTOSENSE_EN;
2254 		wr32(E1000_CONNSW, connsw);
2255 		wrfl();
2256 	}
2257 }
2258 
2259 void igb_reset(struct igb_adapter *adapter)
2260 {
2261 	struct pci_dev *pdev = adapter->pdev;
2262 	struct e1000_hw *hw = &adapter->hw;
2263 	struct e1000_mac_info *mac = &hw->mac;
2264 	struct e1000_fc_info *fc = &hw->fc;
2265 	u32 pba, hwm;
2266 
2267 	/* Repartition Pba for greater than 9k mtu
2268 	 * To take effect CTRL.RST is required.
2269 	 */
2270 	switch (mac->type) {
2271 	case e1000_i350:
2272 	case e1000_i354:
2273 	case e1000_82580:
2274 		pba = rd32(E1000_RXPBS);
2275 		pba = igb_rxpbs_adjust_82580(pba);
2276 		break;
2277 	case e1000_82576:
2278 		pba = rd32(E1000_RXPBS);
2279 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2280 		break;
2281 	case e1000_82575:
2282 	case e1000_i210:
2283 	case e1000_i211:
2284 	default:
2285 		pba = E1000_PBA_34K;
2286 		break;
2287 	}
2288 
2289 	if (mac->type == e1000_82575) {
2290 		u32 min_rx_space, min_tx_space, needed_tx_space;
2291 
2292 		/* write Rx PBA so that hardware can report correct Tx PBA */
2293 		wr32(E1000_PBA, pba);
2294 
2295 		/* To maintain wire speed transmits, the Tx FIFO should be
2296 		 * large enough to accommodate two full transmit packets,
2297 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
2298 		 * the Rx FIFO should be large enough to accommodate at least
2299 		 * one full receive packet and is similarly rounded up and
2300 		 * expressed in KB.
2301 		 */
2302 		min_rx_space = DIV_ROUND_UP(MAX_JUMBO_FRAME_SIZE, 1024);
2303 
2304 		/* The Tx FIFO also stores 16 bytes of information about the Tx
2305 		 * but don't include Ethernet FCS because hardware appends it.
2306 		 * We only need to round down to the nearest 512 byte block
2307 		 * count since the value we care about is 2 frames, not 1.
2308 		 */
2309 		min_tx_space = adapter->max_frame_size;
2310 		min_tx_space += sizeof(union e1000_adv_tx_desc) - ETH_FCS_LEN;
2311 		min_tx_space = DIV_ROUND_UP(min_tx_space, 512);
2312 
2313 		/* upper 16 bits has Tx packet buffer allocation size in KB */
2314 		needed_tx_space = min_tx_space - (rd32(E1000_PBA) >> 16);
2315 
2316 		/* If current Tx allocation is less than the min Tx FIFO size,
2317 		 * and the min Tx FIFO size is less than the current Rx FIFO
2318 		 * allocation, take space away from current Rx allocation.
2319 		 */
2320 		if (needed_tx_space < pba) {
2321 			pba -= needed_tx_space;
2322 
2323 			/* if short on Rx space, Rx wins and must trump Tx
2324 			 * adjustment
2325 			 */
2326 			if (pba < min_rx_space)
2327 				pba = min_rx_space;
2328 		}
2329 
2330 		/* adjust PBA for jumbo frames */
2331 		wr32(E1000_PBA, pba);
2332 	}
2333 
2334 	/* flow control settings
2335 	 * The high water mark must be low enough to fit one full frame
2336 	 * after transmitting the pause frame.  As such we must have enough
2337 	 * space to allow for us to complete our current transmit and then
2338 	 * receive the frame that is in progress from the link partner.
2339 	 * Set it to:
2340 	 * - the full Rx FIFO size minus one full Tx plus one full Rx frame
2341 	 */
2342 	hwm = (pba << 10) - (adapter->max_frame_size + MAX_JUMBO_FRAME_SIZE);
2343 
2344 	fc->high_water = hwm & 0xFFFFFFF0;	/* 16-byte granularity */
2345 	fc->low_water = fc->high_water - 16;
2346 	fc->pause_time = 0xFFFF;
2347 	fc->send_xon = 1;
2348 	fc->current_mode = fc->requested_mode;
2349 
2350 	/* disable receive for all VFs and wait one second */
2351 	if (adapter->vfs_allocated_count) {
2352 		int i;
2353 
2354 		for (i = 0 ; i < adapter->vfs_allocated_count; i++)
2355 			adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
2356 
2357 		/* ping all the active vfs to let them know we are going down */
2358 		igb_ping_all_vfs(adapter);
2359 
2360 		/* disable transmits and receives */
2361 		wr32(E1000_VFRE, 0);
2362 		wr32(E1000_VFTE, 0);
2363 	}
2364 
2365 	/* Allow time for pending master requests to run */
2366 	hw->mac.ops.reset_hw(hw);
2367 	wr32(E1000_WUC, 0);
2368 
2369 	if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
2370 		/* need to resetup here after media swap */
2371 		adapter->ei.get_invariants(hw);
2372 		adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
2373 	}
2374 	if ((mac->type == e1000_82575 || mac->type == e1000_i350) &&
2375 	    (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
2376 		igb_enable_mas(adapter);
2377 	}
2378 	if (hw->mac.ops.init_hw(hw))
2379 		dev_err(&pdev->dev, "Hardware Error\n");
2380 
2381 	/* RAR registers were cleared during init_hw, clear mac table */
2382 	igb_flush_mac_table(adapter);
2383 	__dev_uc_unsync(adapter->netdev, NULL);
2384 
2385 	/* Recover default RAR entry */
2386 	igb_set_default_mac_filter(adapter);
2387 
2388 	/* Flow control settings reset on hardware reset, so guarantee flow
2389 	 * control is off when forcing speed.
2390 	 */
2391 	if (!hw->mac.autoneg)
2392 		igb_force_mac_fc(hw);
2393 
2394 	igb_init_dmac(adapter, pba);
2395 #ifdef CONFIG_IGB_HWMON
2396 	/* Re-initialize the thermal sensor on i350 devices. */
2397 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
2398 		if (mac->type == e1000_i350 && hw->bus.func == 0) {
2399 			/* If present, re-initialize the external thermal sensor
2400 			 * interface.
2401 			 */
2402 			if (adapter->ets)
2403 				mac->ops.init_thermal_sensor_thresh(hw);
2404 		}
2405 	}
2406 #endif
2407 	/* Re-establish EEE setting */
2408 	if (hw->phy.media_type == e1000_media_type_copper) {
2409 		switch (mac->type) {
2410 		case e1000_i350:
2411 		case e1000_i210:
2412 		case e1000_i211:
2413 			igb_set_eee_i350(hw, true, true);
2414 			break;
2415 		case e1000_i354:
2416 			igb_set_eee_i354(hw, true, true);
2417 			break;
2418 		default:
2419 			break;
2420 		}
2421 	}
2422 	if (!netif_running(adapter->netdev))
2423 		igb_power_down_link(adapter);
2424 
2425 	igb_update_mng_vlan(adapter);
2426 
2427 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2428 	wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2429 
2430 	/* Re-enable PTP, where applicable. */
2431 	if (adapter->ptp_flags & IGB_PTP_ENABLED)
2432 		igb_ptp_reset(adapter);
2433 
2434 	igb_get_phy_info(hw);
2435 }
2436 
2437 static netdev_features_t igb_fix_features(struct net_device *netdev,
2438 	netdev_features_t features)
2439 {
2440 	/* Since there is no support for separate Rx/Tx vlan accel
2441 	 * enable/disable make sure Tx flag is always in same state as Rx.
2442 	 */
2443 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2444 		features |= NETIF_F_HW_VLAN_CTAG_TX;
2445 	else
2446 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2447 
2448 	return features;
2449 }
2450 
2451 static int igb_set_features(struct net_device *netdev,
2452 	netdev_features_t features)
2453 {
2454 	netdev_features_t changed = netdev->features ^ features;
2455 	struct igb_adapter *adapter = netdev_priv(netdev);
2456 
2457 	if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2458 		igb_vlan_mode(netdev, features);
2459 
2460 	if (!(changed & (NETIF_F_RXALL | NETIF_F_NTUPLE)))
2461 		return 0;
2462 
2463 	if (!(features & NETIF_F_NTUPLE)) {
2464 		struct hlist_node *node2;
2465 		struct igb_nfc_filter *rule;
2466 
2467 		spin_lock(&adapter->nfc_lock);
2468 		hlist_for_each_entry_safe(rule, node2,
2469 					  &adapter->nfc_filter_list, nfc_node) {
2470 			igb_erase_filter(adapter, rule);
2471 			hlist_del(&rule->nfc_node);
2472 			kfree(rule);
2473 		}
2474 		spin_unlock(&adapter->nfc_lock);
2475 		adapter->nfc_filter_count = 0;
2476 	}
2477 
2478 	netdev->features = features;
2479 
2480 	if (netif_running(netdev))
2481 		igb_reinit_locked(adapter);
2482 	else
2483 		igb_reset(adapter);
2484 
2485 	return 1;
2486 }
2487 
2488 static int igb_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
2489 			   struct net_device *dev,
2490 			   const unsigned char *addr, u16 vid,
2491 			   u16 flags,
2492 			   struct netlink_ext_ack *extack)
2493 {
2494 	/* guarantee we can provide a unique filter for the unicast address */
2495 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
2496 		struct igb_adapter *adapter = netdev_priv(dev);
2497 		int vfn = adapter->vfs_allocated_count;
2498 
2499 		if (netdev_uc_count(dev) >= igb_available_rars(adapter, vfn))
2500 			return -ENOMEM;
2501 	}
2502 
2503 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
2504 }
2505 
2506 #define IGB_MAX_MAC_HDR_LEN	127
2507 #define IGB_MAX_NETWORK_HDR_LEN	511
2508 
2509 static netdev_features_t
2510 igb_features_check(struct sk_buff *skb, struct net_device *dev,
2511 		   netdev_features_t features)
2512 {
2513 	unsigned int network_hdr_len, mac_hdr_len;
2514 
2515 	/* Make certain the headers can be described by a context descriptor */
2516 	mac_hdr_len = skb_network_header(skb) - skb->data;
2517 	if (unlikely(mac_hdr_len > IGB_MAX_MAC_HDR_LEN))
2518 		return features & ~(NETIF_F_HW_CSUM |
2519 				    NETIF_F_SCTP_CRC |
2520 				    NETIF_F_GSO_UDP_L4 |
2521 				    NETIF_F_HW_VLAN_CTAG_TX |
2522 				    NETIF_F_TSO |
2523 				    NETIF_F_TSO6);
2524 
2525 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
2526 	if (unlikely(network_hdr_len >  IGB_MAX_NETWORK_HDR_LEN))
2527 		return features & ~(NETIF_F_HW_CSUM |
2528 				    NETIF_F_SCTP_CRC |
2529 				    NETIF_F_GSO_UDP_L4 |
2530 				    NETIF_F_TSO |
2531 				    NETIF_F_TSO6);
2532 
2533 	/* We can only support IPV4 TSO in tunnels if we can mangle the
2534 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
2535 	 */
2536 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
2537 		features &= ~NETIF_F_TSO;
2538 
2539 	return features;
2540 }
2541 
2542 static void igb_offload_apply(struct igb_adapter *adapter, s32 queue)
2543 {
2544 	if (!is_fqtss_enabled(adapter)) {
2545 		enable_fqtss(adapter, true);
2546 		return;
2547 	}
2548 
2549 	igb_config_tx_modes(adapter, queue);
2550 
2551 	if (!is_any_cbs_enabled(adapter) && !is_any_txtime_enabled(adapter))
2552 		enable_fqtss(adapter, false);
2553 }
2554 
2555 static int igb_offload_cbs(struct igb_adapter *adapter,
2556 			   struct tc_cbs_qopt_offload *qopt)
2557 {
2558 	struct e1000_hw *hw = &adapter->hw;
2559 	int err;
2560 
2561 	/* CBS offloading is only supported by i210 controller. */
2562 	if (hw->mac.type != e1000_i210)
2563 		return -EOPNOTSUPP;
2564 
2565 	/* CBS offloading is only supported by queue 0 and queue 1. */
2566 	if (qopt->queue < 0 || qopt->queue > 1)
2567 		return -EINVAL;
2568 
2569 	err = igb_save_cbs_params(adapter, qopt->queue, qopt->enable,
2570 				  qopt->idleslope, qopt->sendslope,
2571 				  qopt->hicredit, qopt->locredit);
2572 	if (err)
2573 		return err;
2574 
2575 	igb_offload_apply(adapter, qopt->queue);
2576 
2577 	return 0;
2578 }
2579 
2580 #define ETHER_TYPE_FULL_MASK ((__force __be16)~0)
2581 #define VLAN_PRIO_FULL_MASK (0x07)
2582 
2583 static int igb_parse_cls_flower(struct igb_adapter *adapter,
2584 				struct flow_cls_offload *f,
2585 				int traffic_class,
2586 				struct igb_nfc_filter *input)
2587 {
2588 	struct flow_rule *rule = flow_cls_offload_flow_rule(f);
2589 	struct flow_dissector *dissector = rule->match.dissector;
2590 	struct netlink_ext_ack *extack = f->common.extack;
2591 
2592 	if (dissector->used_keys &
2593 	    ~(BIT(FLOW_DISSECTOR_KEY_BASIC) |
2594 	      BIT(FLOW_DISSECTOR_KEY_CONTROL) |
2595 	      BIT(FLOW_DISSECTOR_KEY_ETH_ADDRS) |
2596 	      BIT(FLOW_DISSECTOR_KEY_VLAN))) {
2597 		NL_SET_ERR_MSG_MOD(extack,
2598 				   "Unsupported key used, only BASIC, CONTROL, ETH_ADDRS and VLAN are supported");
2599 		return -EOPNOTSUPP;
2600 	}
2601 
2602 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
2603 		struct flow_match_eth_addrs match;
2604 
2605 		flow_rule_match_eth_addrs(rule, &match);
2606 		if (!is_zero_ether_addr(match.mask->dst)) {
2607 			if (!is_broadcast_ether_addr(match.mask->dst)) {
2608 				NL_SET_ERR_MSG_MOD(extack, "Only full masks are supported for destination MAC address");
2609 				return -EINVAL;
2610 			}
2611 
2612 			input->filter.match_flags |=
2613 				IGB_FILTER_FLAG_DST_MAC_ADDR;
2614 			ether_addr_copy(input->filter.dst_addr, match.key->dst);
2615 		}
2616 
2617 		if (!is_zero_ether_addr(match.mask->src)) {
2618 			if (!is_broadcast_ether_addr(match.mask->src)) {
2619 				NL_SET_ERR_MSG_MOD(extack, "Only full masks are supported for source MAC address");
2620 				return -EINVAL;
2621 			}
2622 
2623 			input->filter.match_flags |=
2624 				IGB_FILTER_FLAG_SRC_MAC_ADDR;
2625 			ether_addr_copy(input->filter.src_addr, match.key->src);
2626 		}
2627 	}
2628 
2629 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
2630 		struct flow_match_basic match;
2631 
2632 		flow_rule_match_basic(rule, &match);
2633 		if (match.mask->n_proto) {
2634 			if (match.mask->n_proto != ETHER_TYPE_FULL_MASK) {
2635 				NL_SET_ERR_MSG_MOD(extack, "Only full mask is supported for EtherType filter");
2636 				return -EINVAL;
2637 			}
2638 
2639 			input->filter.match_flags |= IGB_FILTER_FLAG_ETHER_TYPE;
2640 			input->filter.etype = match.key->n_proto;
2641 		}
2642 	}
2643 
2644 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
2645 		struct flow_match_vlan match;
2646 
2647 		flow_rule_match_vlan(rule, &match);
2648 		if (match.mask->vlan_priority) {
2649 			if (match.mask->vlan_priority != VLAN_PRIO_FULL_MASK) {
2650 				NL_SET_ERR_MSG_MOD(extack, "Only full mask is supported for VLAN priority");
2651 				return -EINVAL;
2652 			}
2653 
2654 			input->filter.match_flags |= IGB_FILTER_FLAG_VLAN_TCI;
2655 			input->filter.vlan_tci =
2656 				(__force __be16)match.key->vlan_priority;
2657 		}
2658 	}
2659 
2660 	input->action = traffic_class;
2661 	input->cookie = f->cookie;
2662 
2663 	return 0;
2664 }
2665 
2666 static int igb_configure_clsflower(struct igb_adapter *adapter,
2667 				   struct flow_cls_offload *cls_flower)
2668 {
2669 	struct netlink_ext_ack *extack = cls_flower->common.extack;
2670 	struct igb_nfc_filter *filter, *f;
2671 	int err, tc;
2672 
2673 	tc = tc_classid_to_hwtc(adapter->netdev, cls_flower->classid);
2674 	if (tc < 0) {
2675 		NL_SET_ERR_MSG_MOD(extack, "Invalid traffic class");
2676 		return -EINVAL;
2677 	}
2678 
2679 	filter = kzalloc(sizeof(*filter), GFP_KERNEL);
2680 	if (!filter)
2681 		return -ENOMEM;
2682 
2683 	err = igb_parse_cls_flower(adapter, cls_flower, tc, filter);
2684 	if (err < 0)
2685 		goto err_parse;
2686 
2687 	spin_lock(&adapter->nfc_lock);
2688 
2689 	hlist_for_each_entry(f, &adapter->nfc_filter_list, nfc_node) {
2690 		if (!memcmp(&f->filter, &filter->filter, sizeof(f->filter))) {
2691 			err = -EEXIST;
2692 			NL_SET_ERR_MSG_MOD(extack,
2693 					   "This filter is already set in ethtool");
2694 			goto err_locked;
2695 		}
2696 	}
2697 
2698 	hlist_for_each_entry(f, &adapter->cls_flower_list, nfc_node) {
2699 		if (!memcmp(&f->filter, &filter->filter, sizeof(f->filter))) {
2700 			err = -EEXIST;
2701 			NL_SET_ERR_MSG_MOD(extack,
2702 					   "This filter is already set in cls_flower");
2703 			goto err_locked;
2704 		}
2705 	}
2706 
2707 	err = igb_add_filter(adapter, filter);
2708 	if (err < 0) {
2709 		NL_SET_ERR_MSG_MOD(extack, "Could not add filter to the adapter");
2710 		goto err_locked;
2711 	}
2712 
2713 	hlist_add_head(&filter->nfc_node, &adapter->cls_flower_list);
2714 
2715 	spin_unlock(&adapter->nfc_lock);
2716 
2717 	return 0;
2718 
2719 err_locked:
2720 	spin_unlock(&adapter->nfc_lock);
2721 
2722 err_parse:
2723 	kfree(filter);
2724 
2725 	return err;
2726 }
2727 
2728 static int igb_delete_clsflower(struct igb_adapter *adapter,
2729 				struct flow_cls_offload *cls_flower)
2730 {
2731 	struct igb_nfc_filter *filter;
2732 	int err;
2733 
2734 	spin_lock(&adapter->nfc_lock);
2735 
2736 	hlist_for_each_entry(filter, &adapter->cls_flower_list, nfc_node)
2737 		if (filter->cookie == cls_flower->cookie)
2738 			break;
2739 
2740 	if (!filter) {
2741 		err = -ENOENT;
2742 		goto out;
2743 	}
2744 
2745 	err = igb_erase_filter(adapter, filter);
2746 	if (err < 0)
2747 		goto out;
2748 
2749 	hlist_del(&filter->nfc_node);
2750 	kfree(filter);
2751 
2752 out:
2753 	spin_unlock(&adapter->nfc_lock);
2754 
2755 	return err;
2756 }
2757 
2758 static int igb_setup_tc_cls_flower(struct igb_adapter *adapter,
2759 				   struct flow_cls_offload *cls_flower)
2760 {
2761 	switch (cls_flower->command) {
2762 	case FLOW_CLS_REPLACE:
2763 		return igb_configure_clsflower(adapter, cls_flower);
2764 	case FLOW_CLS_DESTROY:
2765 		return igb_delete_clsflower(adapter, cls_flower);
2766 	case FLOW_CLS_STATS:
2767 		return -EOPNOTSUPP;
2768 	default:
2769 		return -EOPNOTSUPP;
2770 	}
2771 }
2772 
2773 static int igb_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
2774 				 void *cb_priv)
2775 {
2776 	struct igb_adapter *adapter = cb_priv;
2777 
2778 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
2779 		return -EOPNOTSUPP;
2780 
2781 	switch (type) {
2782 	case TC_SETUP_CLSFLOWER:
2783 		return igb_setup_tc_cls_flower(adapter, type_data);
2784 
2785 	default:
2786 		return -EOPNOTSUPP;
2787 	}
2788 }
2789 
2790 static int igb_offload_txtime(struct igb_adapter *adapter,
2791 			      struct tc_etf_qopt_offload *qopt)
2792 {
2793 	struct e1000_hw *hw = &adapter->hw;
2794 	int err;
2795 
2796 	/* Launchtime offloading is only supported by i210 controller. */
2797 	if (hw->mac.type != e1000_i210)
2798 		return -EOPNOTSUPP;
2799 
2800 	/* Launchtime offloading is only supported by queues 0 and 1. */
2801 	if (qopt->queue < 0 || qopt->queue > 1)
2802 		return -EINVAL;
2803 
2804 	err = igb_save_txtime_params(adapter, qopt->queue, qopt->enable);
2805 	if (err)
2806 		return err;
2807 
2808 	igb_offload_apply(adapter, qopt->queue);
2809 
2810 	return 0;
2811 }
2812 
2813 static LIST_HEAD(igb_block_cb_list);
2814 
2815 static int igb_setup_tc(struct net_device *dev, enum tc_setup_type type,
2816 			void *type_data)
2817 {
2818 	struct igb_adapter *adapter = netdev_priv(dev);
2819 
2820 	switch (type) {
2821 	case TC_SETUP_QDISC_CBS:
2822 		return igb_offload_cbs(adapter, type_data);
2823 	case TC_SETUP_BLOCK:
2824 		return flow_block_cb_setup_simple(type_data,
2825 						  &igb_block_cb_list,
2826 						  igb_setup_tc_block_cb,
2827 						  adapter, adapter, true);
2828 
2829 	case TC_SETUP_QDISC_ETF:
2830 		return igb_offload_txtime(adapter, type_data);
2831 
2832 	default:
2833 		return -EOPNOTSUPP;
2834 	}
2835 }
2836 
2837 static int igb_xdp_setup(struct net_device *dev, struct netdev_bpf *bpf)
2838 {
2839 	int i, frame_size = dev->mtu + IGB_ETH_PKT_HDR_PAD;
2840 	struct igb_adapter *adapter = netdev_priv(dev);
2841 	struct bpf_prog *prog = bpf->prog, *old_prog;
2842 	bool running = netif_running(dev);
2843 	bool need_reset;
2844 
2845 	/* verify igb ring attributes are sufficient for XDP */
2846 	for (i = 0; i < adapter->num_rx_queues; i++) {
2847 		struct igb_ring *ring = adapter->rx_ring[i];
2848 
2849 		if (frame_size > igb_rx_bufsz(ring)) {
2850 			NL_SET_ERR_MSG_MOD(bpf->extack,
2851 					   "The RX buffer size is too small for the frame size");
2852 			netdev_warn(dev, "XDP RX buffer size %d is too small for the frame size %d\n",
2853 				    igb_rx_bufsz(ring), frame_size);
2854 			return -EINVAL;
2855 		}
2856 	}
2857 
2858 	old_prog = xchg(&adapter->xdp_prog, prog);
2859 	need_reset = (!!prog != !!old_prog);
2860 
2861 	/* device is up and bpf is added/removed, must setup the RX queues */
2862 	if (need_reset && running) {
2863 		igb_close(dev);
2864 	} else {
2865 		for (i = 0; i < adapter->num_rx_queues; i++)
2866 			(void)xchg(&adapter->rx_ring[i]->xdp_prog,
2867 			    adapter->xdp_prog);
2868 	}
2869 
2870 	if (old_prog)
2871 		bpf_prog_put(old_prog);
2872 
2873 	/* bpf is just replaced, RXQ and MTU are already setup */
2874 	if (!need_reset)
2875 		return 0;
2876 
2877 	if (running)
2878 		igb_open(dev);
2879 
2880 	return 0;
2881 }
2882 
2883 static int igb_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2884 {
2885 	switch (xdp->command) {
2886 	case XDP_SETUP_PROG:
2887 		return igb_xdp_setup(dev, xdp);
2888 	default:
2889 		return -EINVAL;
2890 	}
2891 }
2892 
2893 static void igb_xdp_ring_update_tail(struct igb_ring *ring)
2894 {
2895 	/* Force memory writes to complete before letting h/w know there
2896 	 * are new descriptors to fetch.
2897 	 */
2898 	wmb();
2899 	writel(ring->next_to_use, ring->tail);
2900 }
2901 
2902 static struct igb_ring *igb_xdp_tx_queue_mapping(struct igb_adapter *adapter)
2903 {
2904 	unsigned int r_idx = smp_processor_id();
2905 
2906 	if (r_idx >= adapter->num_tx_queues)
2907 		r_idx = r_idx % adapter->num_tx_queues;
2908 
2909 	return adapter->tx_ring[r_idx];
2910 }
2911 
2912 static int igb_xdp_xmit_back(struct igb_adapter *adapter, struct xdp_buff *xdp)
2913 {
2914 	struct xdp_frame *xdpf = xdp_convert_buff_to_frame(xdp);
2915 	int cpu = smp_processor_id();
2916 	struct igb_ring *tx_ring;
2917 	struct netdev_queue *nq;
2918 	u32 ret;
2919 
2920 	if (unlikely(!xdpf))
2921 		return IGB_XDP_CONSUMED;
2922 
2923 	/* During program transitions its possible adapter->xdp_prog is assigned
2924 	 * but ring has not been configured yet. In this case simply abort xmit.
2925 	 */
2926 	tx_ring = adapter->xdp_prog ? igb_xdp_tx_queue_mapping(adapter) : NULL;
2927 	if (unlikely(!tx_ring))
2928 		return IGB_XDP_CONSUMED;
2929 
2930 	nq = txring_txq(tx_ring);
2931 	__netif_tx_lock(nq, cpu);
2932 	/* Avoid transmit queue timeout since we share it with the slow path */
2933 	txq_trans_cond_update(nq);
2934 	ret = igb_xmit_xdp_ring(adapter, tx_ring, xdpf);
2935 	__netif_tx_unlock(nq);
2936 
2937 	return ret;
2938 }
2939 
2940 static int igb_xdp_xmit(struct net_device *dev, int n,
2941 			struct xdp_frame **frames, u32 flags)
2942 {
2943 	struct igb_adapter *adapter = netdev_priv(dev);
2944 	int cpu = smp_processor_id();
2945 	struct igb_ring *tx_ring;
2946 	struct netdev_queue *nq;
2947 	int nxmit = 0;
2948 	int i;
2949 
2950 	if (unlikely(test_bit(__IGB_DOWN, &adapter->state)))
2951 		return -ENETDOWN;
2952 
2953 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
2954 		return -EINVAL;
2955 
2956 	/* During program transitions its possible adapter->xdp_prog is assigned
2957 	 * but ring has not been configured yet. In this case simply abort xmit.
2958 	 */
2959 	tx_ring = adapter->xdp_prog ? igb_xdp_tx_queue_mapping(adapter) : NULL;
2960 	if (unlikely(!tx_ring))
2961 		return -ENXIO;
2962 
2963 	nq = txring_txq(tx_ring);
2964 	__netif_tx_lock(nq, cpu);
2965 
2966 	/* Avoid transmit queue timeout since we share it with the slow path */
2967 	txq_trans_cond_update(nq);
2968 
2969 	for (i = 0; i < n; i++) {
2970 		struct xdp_frame *xdpf = frames[i];
2971 		int err;
2972 
2973 		err = igb_xmit_xdp_ring(adapter, tx_ring, xdpf);
2974 		if (err != IGB_XDP_TX)
2975 			break;
2976 		nxmit++;
2977 	}
2978 
2979 	__netif_tx_unlock(nq);
2980 
2981 	if (unlikely(flags & XDP_XMIT_FLUSH))
2982 		igb_xdp_ring_update_tail(tx_ring);
2983 
2984 	return nxmit;
2985 }
2986 
2987 static const struct net_device_ops igb_netdev_ops = {
2988 	.ndo_open		= igb_open,
2989 	.ndo_stop		= igb_close,
2990 	.ndo_start_xmit		= igb_xmit_frame,
2991 	.ndo_get_stats64	= igb_get_stats64,
2992 	.ndo_set_rx_mode	= igb_set_rx_mode,
2993 	.ndo_set_mac_address	= igb_set_mac,
2994 	.ndo_change_mtu		= igb_change_mtu,
2995 	.ndo_eth_ioctl		= igb_ioctl,
2996 	.ndo_tx_timeout		= igb_tx_timeout,
2997 	.ndo_validate_addr	= eth_validate_addr,
2998 	.ndo_vlan_rx_add_vid	= igb_vlan_rx_add_vid,
2999 	.ndo_vlan_rx_kill_vid	= igb_vlan_rx_kill_vid,
3000 	.ndo_set_vf_mac		= igb_ndo_set_vf_mac,
3001 	.ndo_set_vf_vlan	= igb_ndo_set_vf_vlan,
3002 	.ndo_set_vf_rate	= igb_ndo_set_vf_bw,
3003 	.ndo_set_vf_spoofchk	= igb_ndo_set_vf_spoofchk,
3004 	.ndo_set_vf_trust	= igb_ndo_set_vf_trust,
3005 	.ndo_get_vf_config	= igb_ndo_get_vf_config,
3006 	.ndo_fix_features	= igb_fix_features,
3007 	.ndo_set_features	= igb_set_features,
3008 	.ndo_fdb_add		= igb_ndo_fdb_add,
3009 	.ndo_features_check	= igb_features_check,
3010 	.ndo_setup_tc		= igb_setup_tc,
3011 	.ndo_bpf		= igb_xdp,
3012 	.ndo_xdp_xmit		= igb_xdp_xmit,
3013 };
3014 
3015 /**
3016  * igb_set_fw_version - Configure version string for ethtool
3017  * @adapter: adapter struct
3018  **/
3019 void igb_set_fw_version(struct igb_adapter *adapter)
3020 {
3021 	struct e1000_hw *hw = &adapter->hw;
3022 	struct e1000_fw_version fw;
3023 
3024 	igb_get_fw_version(hw, &fw);
3025 
3026 	switch (hw->mac.type) {
3027 	case e1000_i210:
3028 	case e1000_i211:
3029 		if (!(igb_get_flash_presence_i210(hw))) {
3030 			snprintf(adapter->fw_version,
3031 				 sizeof(adapter->fw_version),
3032 				 "%2d.%2d-%d",
3033 				 fw.invm_major, fw.invm_minor,
3034 				 fw.invm_img_type);
3035 			break;
3036 		}
3037 		fallthrough;
3038 	default:
3039 		/* if option is rom valid, display its version too */
3040 		if (fw.or_valid) {
3041 			snprintf(adapter->fw_version,
3042 				 sizeof(adapter->fw_version),
3043 				 "%d.%d, 0x%08x, %d.%d.%d",
3044 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
3045 				 fw.or_major, fw.or_build, fw.or_patch);
3046 		/* no option rom */
3047 		} else if (fw.etrack_id != 0X0000) {
3048 			snprintf(adapter->fw_version,
3049 			    sizeof(adapter->fw_version),
3050 			    "%d.%d, 0x%08x",
3051 			    fw.eep_major, fw.eep_minor, fw.etrack_id);
3052 		} else {
3053 		snprintf(adapter->fw_version,
3054 		    sizeof(adapter->fw_version),
3055 		    "%d.%d.%d",
3056 		    fw.eep_major, fw.eep_minor, fw.eep_build);
3057 		}
3058 		break;
3059 	}
3060 }
3061 
3062 /**
3063  * igb_init_mas - init Media Autosense feature if enabled in the NVM
3064  *
3065  * @adapter: adapter struct
3066  **/
3067 static void igb_init_mas(struct igb_adapter *adapter)
3068 {
3069 	struct e1000_hw *hw = &adapter->hw;
3070 	u16 eeprom_data;
3071 
3072 	hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
3073 	switch (hw->bus.func) {
3074 	case E1000_FUNC_0:
3075 		if (eeprom_data & IGB_MAS_ENABLE_0) {
3076 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3077 			netdev_info(adapter->netdev,
3078 				"MAS: Enabling Media Autosense for port %d\n",
3079 				hw->bus.func);
3080 		}
3081 		break;
3082 	case E1000_FUNC_1:
3083 		if (eeprom_data & IGB_MAS_ENABLE_1) {
3084 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3085 			netdev_info(adapter->netdev,
3086 				"MAS: Enabling Media Autosense for port %d\n",
3087 				hw->bus.func);
3088 		}
3089 		break;
3090 	case E1000_FUNC_2:
3091 		if (eeprom_data & IGB_MAS_ENABLE_2) {
3092 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3093 			netdev_info(adapter->netdev,
3094 				"MAS: Enabling Media Autosense for port %d\n",
3095 				hw->bus.func);
3096 		}
3097 		break;
3098 	case E1000_FUNC_3:
3099 		if (eeprom_data & IGB_MAS_ENABLE_3) {
3100 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3101 			netdev_info(adapter->netdev,
3102 				"MAS: Enabling Media Autosense for port %d\n",
3103 				hw->bus.func);
3104 		}
3105 		break;
3106 	default:
3107 		/* Shouldn't get here */
3108 		netdev_err(adapter->netdev,
3109 			"MAS: Invalid port configuration, returning\n");
3110 		break;
3111 	}
3112 }
3113 
3114 /**
3115  *  igb_init_i2c - Init I2C interface
3116  *  @adapter: pointer to adapter structure
3117  **/
3118 static s32 igb_init_i2c(struct igb_adapter *adapter)
3119 {
3120 	struct e1000_hw *hw = &adapter->hw;
3121 	s32 status = 0;
3122 	s32 i2cctl;
3123 
3124 	/* I2C interface supported on i350 devices */
3125 	if (adapter->hw.mac.type != e1000_i350)
3126 		return 0;
3127 
3128 	i2cctl = rd32(E1000_I2CPARAMS);
3129 	i2cctl |= E1000_I2CBB_EN
3130 		| E1000_I2C_CLK_OUT | E1000_I2C_CLK_OE_N
3131 		| E1000_I2C_DATA_OUT | E1000_I2C_DATA_OE_N;
3132 	wr32(E1000_I2CPARAMS, i2cctl);
3133 	wrfl();
3134 
3135 	/* Initialize the i2c bus which is controlled by the registers.
3136 	 * This bus will use the i2c_algo_bit structure that implements
3137 	 * the protocol through toggling of the 4 bits in the register.
3138 	 */
3139 	adapter->i2c_adap.owner = THIS_MODULE;
3140 	adapter->i2c_algo = igb_i2c_algo;
3141 	adapter->i2c_algo.data = adapter;
3142 	adapter->i2c_adap.algo_data = &adapter->i2c_algo;
3143 	adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
3144 	strscpy(adapter->i2c_adap.name, "igb BB",
3145 		sizeof(adapter->i2c_adap.name));
3146 	status = i2c_bit_add_bus(&adapter->i2c_adap);
3147 	return status;
3148 }
3149 
3150 /**
3151  *  igb_probe - Device Initialization Routine
3152  *  @pdev: PCI device information struct
3153  *  @ent: entry in igb_pci_tbl
3154  *
3155  *  Returns 0 on success, negative on failure
3156  *
3157  *  igb_probe initializes an adapter identified by a pci_dev structure.
3158  *  The OS initialization, configuring of the adapter private structure,
3159  *  and a hardware reset occur.
3160  **/
3161 static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
3162 {
3163 	struct net_device *netdev;
3164 	struct igb_adapter *adapter;
3165 	struct e1000_hw *hw;
3166 	u16 eeprom_data = 0;
3167 	s32 ret_val;
3168 	static int global_quad_port_a; /* global quad port a indication */
3169 	const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
3170 	u8 part_str[E1000_PBANUM_LENGTH];
3171 	int err;
3172 
3173 	/* Catch broken hardware that put the wrong VF device ID in
3174 	 * the PCIe SR-IOV capability.
3175 	 */
3176 	if (pdev->is_virtfn) {
3177 		WARN(1, KERN_ERR "%s (%x:%x) should not be a VF!\n",
3178 			pci_name(pdev), pdev->vendor, pdev->device);
3179 		return -EINVAL;
3180 	}
3181 
3182 	err = pci_enable_device_mem(pdev);
3183 	if (err)
3184 		return err;
3185 
3186 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
3187 	if (err) {
3188 		dev_err(&pdev->dev,
3189 			"No usable DMA configuration, aborting\n");
3190 		goto err_dma;
3191 	}
3192 
3193 	err = pci_request_mem_regions(pdev, igb_driver_name);
3194 	if (err)
3195 		goto err_pci_reg;
3196 
3197 	pci_set_master(pdev);
3198 	pci_save_state(pdev);
3199 
3200 	err = -ENOMEM;
3201 	netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
3202 				   IGB_MAX_TX_QUEUES);
3203 	if (!netdev)
3204 		goto err_alloc_etherdev;
3205 
3206 	SET_NETDEV_DEV(netdev, &pdev->dev);
3207 
3208 	pci_set_drvdata(pdev, netdev);
3209 	adapter = netdev_priv(netdev);
3210 	adapter->netdev = netdev;
3211 	adapter->pdev = pdev;
3212 	hw = &adapter->hw;
3213 	hw->back = adapter;
3214 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
3215 
3216 	err = -EIO;
3217 	adapter->io_addr = pci_iomap(pdev, 0, 0);
3218 	if (!adapter->io_addr)
3219 		goto err_ioremap;
3220 	/* hw->hw_addr can be altered, we'll use adapter->io_addr for unmap */
3221 	hw->hw_addr = adapter->io_addr;
3222 
3223 	netdev->netdev_ops = &igb_netdev_ops;
3224 	igb_set_ethtool_ops(netdev);
3225 	netdev->watchdog_timeo = 5 * HZ;
3226 
3227 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
3228 
3229 	netdev->mem_start = pci_resource_start(pdev, 0);
3230 	netdev->mem_end = pci_resource_end(pdev, 0);
3231 
3232 	/* PCI config space info */
3233 	hw->vendor_id = pdev->vendor;
3234 	hw->device_id = pdev->device;
3235 	hw->revision_id = pdev->revision;
3236 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
3237 	hw->subsystem_device_id = pdev->subsystem_device;
3238 
3239 	/* Copy the default MAC, PHY and NVM function pointers */
3240 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
3241 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
3242 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
3243 	/* Initialize skew-specific constants */
3244 	err = ei->get_invariants(hw);
3245 	if (err)
3246 		goto err_sw_init;
3247 
3248 	/* setup the private structure */
3249 	err = igb_sw_init(adapter);
3250 	if (err)
3251 		goto err_sw_init;
3252 
3253 	igb_get_bus_info_pcie(hw);
3254 
3255 	hw->phy.autoneg_wait_to_complete = false;
3256 
3257 	/* Copper options */
3258 	if (hw->phy.media_type == e1000_media_type_copper) {
3259 		hw->phy.mdix = AUTO_ALL_MODES;
3260 		hw->phy.disable_polarity_correction = false;
3261 		hw->phy.ms_type = e1000_ms_hw_default;
3262 	}
3263 
3264 	if (igb_check_reset_block(hw))
3265 		dev_info(&pdev->dev,
3266 			"PHY reset is blocked due to SOL/IDER session.\n");
3267 
3268 	/* features is initialized to 0 in allocation, it might have bits
3269 	 * set by igb_sw_init so we should use an or instead of an
3270 	 * assignment.
3271 	 */
3272 	netdev->features |= NETIF_F_SG |
3273 			    NETIF_F_TSO |
3274 			    NETIF_F_TSO6 |
3275 			    NETIF_F_RXHASH |
3276 			    NETIF_F_RXCSUM |
3277 			    NETIF_F_HW_CSUM;
3278 
3279 	if (hw->mac.type >= e1000_82576)
3280 		netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_GSO_UDP_L4;
3281 
3282 	if (hw->mac.type >= e1000_i350)
3283 		netdev->features |= NETIF_F_HW_TC;
3284 
3285 #define IGB_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
3286 				  NETIF_F_GSO_GRE_CSUM | \
3287 				  NETIF_F_GSO_IPXIP4 | \
3288 				  NETIF_F_GSO_IPXIP6 | \
3289 				  NETIF_F_GSO_UDP_TUNNEL | \
3290 				  NETIF_F_GSO_UDP_TUNNEL_CSUM)
3291 
3292 	netdev->gso_partial_features = IGB_GSO_PARTIAL_FEATURES;
3293 	netdev->features |= NETIF_F_GSO_PARTIAL | IGB_GSO_PARTIAL_FEATURES;
3294 
3295 	/* copy netdev features into list of user selectable features */
3296 	netdev->hw_features |= netdev->features |
3297 			       NETIF_F_HW_VLAN_CTAG_RX |
3298 			       NETIF_F_HW_VLAN_CTAG_TX |
3299 			       NETIF_F_RXALL;
3300 
3301 	if (hw->mac.type >= e1000_i350)
3302 		netdev->hw_features |= NETIF_F_NTUPLE;
3303 
3304 	netdev->features |= NETIF_F_HIGHDMA;
3305 
3306 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
3307 	netdev->mpls_features |= NETIF_F_HW_CSUM;
3308 	netdev->hw_enc_features |= netdev->vlan_features;
3309 
3310 	/* set this bit last since it cannot be part of vlan_features */
3311 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
3312 			    NETIF_F_HW_VLAN_CTAG_RX |
3313 			    NETIF_F_HW_VLAN_CTAG_TX;
3314 
3315 	netdev->priv_flags |= IFF_SUPP_NOFCS;
3316 
3317 	netdev->priv_flags |= IFF_UNICAST_FLT;
3318 
3319 	/* MTU range: 68 - 9216 */
3320 	netdev->min_mtu = ETH_MIN_MTU;
3321 	netdev->max_mtu = MAX_STD_JUMBO_FRAME_SIZE;
3322 
3323 	adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
3324 
3325 	/* before reading the NVM, reset the controller to put the device in a
3326 	 * known good starting state
3327 	 */
3328 	hw->mac.ops.reset_hw(hw);
3329 
3330 	/* make sure the NVM is good , i211/i210 parts can have special NVM
3331 	 * that doesn't contain a checksum
3332 	 */
3333 	switch (hw->mac.type) {
3334 	case e1000_i210:
3335 	case e1000_i211:
3336 		if (igb_get_flash_presence_i210(hw)) {
3337 			if (hw->nvm.ops.validate(hw) < 0) {
3338 				dev_err(&pdev->dev,
3339 					"The NVM Checksum Is Not Valid\n");
3340 				err = -EIO;
3341 				goto err_eeprom;
3342 			}
3343 		}
3344 		break;
3345 	default:
3346 		if (hw->nvm.ops.validate(hw) < 0) {
3347 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
3348 			err = -EIO;
3349 			goto err_eeprom;
3350 		}
3351 		break;
3352 	}
3353 
3354 	if (eth_platform_get_mac_address(&pdev->dev, hw->mac.addr)) {
3355 		/* copy the MAC address out of the NVM */
3356 		if (hw->mac.ops.read_mac_addr(hw))
3357 			dev_err(&pdev->dev, "NVM Read Error\n");
3358 	}
3359 
3360 	eth_hw_addr_set(netdev, hw->mac.addr);
3361 
3362 	if (!is_valid_ether_addr(netdev->dev_addr)) {
3363 		dev_err(&pdev->dev, "Invalid MAC Address\n");
3364 		err = -EIO;
3365 		goto err_eeprom;
3366 	}
3367 
3368 	igb_set_default_mac_filter(adapter);
3369 
3370 	/* get firmware version for ethtool -i */
3371 	igb_set_fw_version(adapter);
3372 
3373 	/* configure RXPBSIZE and TXPBSIZE */
3374 	if (hw->mac.type == e1000_i210) {
3375 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
3376 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
3377 	}
3378 
3379 	timer_setup(&adapter->watchdog_timer, igb_watchdog, 0);
3380 	timer_setup(&adapter->phy_info_timer, igb_update_phy_info, 0);
3381 
3382 	INIT_WORK(&adapter->reset_task, igb_reset_task);
3383 	INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
3384 
3385 	/* Initialize link properties that are user-changeable */
3386 	adapter->fc_autoneg = true;
3387 	hw->mac.autoneg = true;
3388 	hw->phy.autoneg_advertised = 0x2f;
3389 
3390 	hw->fc.requested_mode = e1000_fc_default;
3391 	hw->fc.current_mode = e1000_fc_default;
3392 
3393 	igb_validate_mdi_setting(hw);
3394 
3395 	/* By default, support wake on port A */
3396 	if (hw->bus.func == 0)
3397 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3398 
3399 	/* Check the NVM for wake support on non-port A ports */
3400 	if (hw->mac.type >= e1000_82580)
3401 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
3402 				 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
3403 				 &eeprom_data);
3404 	else if (hw->bus.func == 1)
3405 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3406 
3407 	if (eeprom_data & IGB_EEPROM_APME)
3408 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3409 
3410 	/* now that we have the eeprom settings, apply the special cases where
3411 	 * the eeprom may be wrong or the board simply won't support wake on
3412 	 * lan on a particular port
3413 	 */
3414 	switch (pdev->device) {
3415 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
3416 		adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3417 		break;
3418 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
3419 	case E1000_DEV_ID_82576_FIBER:
3420 	case E1000_DEV_ID_82576_SERDES:
3421 		/* Wake events only supported on port A for dual fiber
3422 		 * regardless of eeprom setting
3423 		 */
3424 		if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
3425 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3426 		break;
3427 	case E1000_DEV_ID_82576_QUAD_COPPER:
3428 	case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
3429 		/* if quad port adapter, disable WoL on all but port A */
3430 		if (global_quad_port_a != 0)
3431 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3432 		else
3433 			adapter->flags |= IGB_FLAG_QUAD_PORT_A;
3434 		/* Reset for multiple quad port adapters */
3435 		if (++global_quad_port_a == 4)
3436 			global_quad_port_a = 0;
3437 		break;
3438 	default:
3439 		/* If the device can't wake, don't set software support */
3440 		if (!device_can_wakeup(&adapter->pdev->dev))
3441 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3442 	}
3443 
3444 	/* initialize the wol settings based on the eeprom settings */
3445 	if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
3446 		adapter->wol |= E1000_WUFC_MAG;
3447 
3448 	/* Some vendors want WoL disabled by default, but still supported */
3449 	if ((hw->mac.type == e1000_i350) &&
3450 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
3451 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3452 		adapter->wol = 0;
3453 	}
3454 
3455 	/* Some vendors want the ability to Use the EEPROM setting as
3456 	 * enable/disable only, and not for capability
3457 	 */
3458 	if (((hw->mac.type == e1000_i350) ||
3459 	     (hw->mac.type == e1000_i354)) &&
3460 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_DELL)) {
3461 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3462 		adapter->wol = 0;
3463 	}
3464 	if (hw->mac.type == e1000_i350) {
3465 		if (((pdev->subsystem_device == 0x5001) ||
3466 		     (pdev->subsystem_device == 0x5002)) &&
3467 				(hw->bus.func == 0)) {
3468 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3469 			adapter->wol = 0;
3470 		}
3471 		if (pdev->subsystem_device == 0x1F52)
3472 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3473 	}
3474 
3475 	device_set_wakeup_enable(&adapter->pdev->dev,
3476 				 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
3477 
3478 	/* reset the hardware with the new settings */
3479 	igb_reset(adapter);
3480 
3481 	/* Init the I2C interface */
3482 	err = igb_init_i2c(adapter);
3483 	if (err) {
3484 		dev_err(&pdev->dev, "failed to init i2c interface\n");
3485 		goto err_eeprom;
3486 	}
3487 
3488 	/* let the f/w know that the h/w is now under the control of the
3489 	 * driver.
3490 	 */
3491 	igb_get_hw_control(adapter);
3492 
3493 	strcpy(netdev->name, "eth%d");
3494 	err = register_netdev(netdev);
3495 	if (err)
3496 		goto err_register;
3497 
3498 	/* carrier off reporting is important to ethtool even BEFORE open */
3499 	netif_carrier_off(netdev);
3500 
3501 #ifdef CONFIG_IGB_DCA
3502 	if (dca_add_requester(&pdev->dev) == 0) {
3503 		adapter->flags |= IGB_FLAG_DCA_ENABLED;
3504 		dev_info(&pdev->dev, "DCA enabled\n");
3505 		igb_setup_dca(adapter);
3506 	}
3507 
3508 #endif
3509 #ifdef CONFIG_IGB_HWMON
3510 	/* Initialize the thermal sensor on i350 devices. */
3511 	if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
3512 		u16 ets_word;
3513 
3514 		/* Read the NVM to determine if this i350 device supports an
3515 		 * external thermal sensor.
3516 		 */
3517 		hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
3518 		if (ets_word != 0x0000 && ets_word != 0xFFFF)
3519 			adapter->ets = true;
3520 		else
3521 			adapter->ets = false;
3522 		if (igb_sysfs_init(adapter))
3523 			dev_err(&pdev->dev,
3524 				"failed to allocate sysfs resources\n");
3525 	} else {
3526 		adapter->ets = false;
3527 	}
3528 #endif
3529 	/* Check if Media Autosense is enabled */
3530 	adapter->ei = *ei;
3531 	if (hw->dev_spec._82575.mas_capable)
3532 		igb_init_mas(adapter);
3533 
3534 	/* do hw tstamp init after resetting */
3535 	igb_ptp_init(adapter);
3536 
3537 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
3538 	/* print bus type/speed/width info, not applicable to i354 */
3539 	if (hw->mac.type != e1000_i354) {
3540 		dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
3541 			 netdev->name,
3542 			 ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
3543 			  (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
3544 			   "unknown"),
3545 			 ((hw->bus.width == e1000_bus_width_pcie_x4) ?
3546 			  "Width x4" :
3547 			  (hw->bus.width == e1000_bus_width_pcie_x2) ?
3548 			  "Width x2" :
3549 			  (hw->bus.width == e1000_bus_width_pcie_x1) ?
3550 			  "Width x1" : "unknown"), netdev->dev_addr);
3551 	}
3552 
3553 	if ((hw->mac.type == e1000_82576 &&
3554 	     rd32(E1000_EECD) & E1000_EECD_PRES) ||
3555 	    (hw->mac.type >= e1000_i210 ||
3556 	     igb_get_flash_presence_i210(hw))) {
3557 		ret_val = igb_read_part_string(hw, part_str,
3558 					       E1000_PBANUM_LENGTH);
3559 	} else {
3560 		ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
3561 	}
3562 
3563 	if (ret_val)
3564 		strcpy(part_str, "Unknown");
3565 	dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
3566 	dev_info(&pdev->dev,
3567 		"Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
3568 		(adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
3569 		(adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
3570 		adapter->num_rx_queues, adapter->num_tx_queues);
3571 	if (hw->phy.media_type == e1000_media_type_copper) {
3572 		switch (hw->mac.type) {
3573 		case e1000_i350:
3574 		case e1000_i210:
3575 		case e1000_i211:
3576 			/* Enable EEE for internal copper PHY devices */
3577 			err = igb_set_eee_i350(hw, true, true);
3578 			if ((!err) &&
3579 			    (!hw->dev_spec._82575.eee_disable)) {
3580 				adapter->eee_advert =
3581 					MDIO_EEE_100TX | MDIO_EEE_1000T;
3582 				adapter->flags |= IGB_FLAG_EEE;
3583 			}
3584 			break;
3585 		case e1000_i354:
3586 			if ((rd32(E1000_CTRL_EXT) &
3587 			    E1000_CTRL_EXT_LINK_MODE_SGMII)) {
3588 				err = igb_set_eee_i354(hw, true, true);
3589 				if ((!err) &&
3590 					(!hw->dev_spec._82575.eee_disable)) {
3591 					adapter->eee_advert =
3592 					   MDIO_EEE_100TX | MDIO_EEE_1000T;
3593 					adapter->flags |= IGB_FLAG_EEE;
3594 				}
3595 			}
3596 			break;
3597 		default:
3598 			break;
3599 		}
3600 	}
3601 
3602 	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_NO_DIRECT_COMPLETE);
3603 
3604 	pm_runtime_put_noidle(&pdev->dev);
3605 	return 0;
3606 
3607 err_register:
3608 	igb_release_hw_control(adapter);
3609 	memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
3610 err_eeprom:
3611 	if (!igb_check_reset_block(hw))
3612 		igb_reset_phy(hw);
3613 
3614 	if (hw->flash_address)
3615 		iounmap(hw->flash_address);
3616 err_sw_init:
3617 	kfree(adapter->mac_table);
3618 	kfree(adapter->shadow_vfta);
3619 	igb_clear_interrupt_scheme(adapter);
3620 #ifdef CONFIG_PCI_IOV
3621 	igb_disable_sriov(pdev);
3622 #endif
3623 	pci_iounmap(pdev, adapter->io_addr);
3624 err_ioremap:
3625 	free_netdev(netdev);
3626 err_alloc_etherdev:
3627 	pci_release_mem_regions(pdev);
3628 err_pci_reg:
3629 err_dma:
3630 	pci_disable_device(pdev);
3631 	return err;
3632 }
3633 
3634 #ifdef CONFIG_PCI_IOV
3635 static int igb_disable_sriov(struct pci_dev *pdev)
3636 {
3637 	struct net_device *netdev = pci_get_drvdata(pdev);
3638 	struct igb_adapter *adapter = netdev_priv(netdev);
3639 	struct e1000_hw *hw = &adapter->hw;
3640 	unsigned long flags;
3641 
3642 	/* reclaim resources allocated to VFs */
3643 	if (adapter->vf_data) {
3644 		/* disable iov and allow time for transactions to clear */
3645 		if (pci_vfs_assigned(pdev)) {
3646 			dev_warn(&pdev->dev,
3647 				 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
3648 			return -EPERM;
3649 		} else {
3650 			pci_disable_sriov(pdev);
3651 			msleep(500);
3652 		}
3653 		spin_lock_irqsave(&adapter->vfs_lock, flags);
3654 		kfree(adapter->vf_mac_list);
3655 		adapter->vf_mac_list = NULL;
3656 		kfree(adapter->vf_data);
3657 		adapter->vf_data = NULL;
3658 		adapter->vfs_allocated_count = 0;
3659 		spin_unlock_irqrestore(&adapter->vfs_lock, flags);
3660 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
3661 		wrfl();
3662 		msleep(100);
3663 		dev_info(&pdev->dev, "IOV Disabled\n");
3664 
3665 		/* Re-enable DMA Coalescing flag since IOV is turned off */
3666 		adapter->flags |= IGB_FLAG_DMAC;
3667 	}
3668 
3669 	return 0;
3670 }
3671 
3672 static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
3673 {
3674 	struct net_device *netdev = pci_get_drvdata(pdev);
3675 	struct igb_adapter *adapter = netdev_priv(netdev);
3676 	int old_vfs = pci_num_vf(pdev);
3677 	struct vf_mac_filter *mac_list;
3678 	int err = 0;
3679 	int num_vf_mac_filters, i;
3680 
3681 	if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
3682 		err = -EPERM;
3683 		goto out;
3684 	}
3685 	if (!num_vfs)
3686 		goto out;
3687 
3688 	if (old_vfs) {
3689 		dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
3690 			 old_vfs, max_vfs);
3691 		adapter->vfs_allocated_count = old_vfs;
3692 	} else
3693 		adapter->vfs_allocated_count = num_vfs;
3694 
3695 	adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
3696 				sizeof(struct vf_data_storage), GFP_KERNEL);
3697 
3698 	/* if allocation failed then we do not support SR-IOV */
3699 	if (!adapter->vf_data) {
3700 		adapter->vfs_allocated_count = 0;
3701 		err = -ENOMEM;
3702 		goto out;
3703 	}
3704 
3705 	/* Due to the limited number of RAR entries calculate potential
3706 	 * number of MAC filters available for the VFs. Reserve entries
3707 	 * for PF default MAC, PF MAC filters and at least one RAR entry
3708 	 * for each VF for VF MAC.
3709 	 */
3710 	num_vf_mac_filters = adapter->hw.mac.rar_entry_count -
3711 			     (1 + IGB_PF_MAC_FILTERS_RESERVED +
3712 			      adapter->vfs_allocated_count);
3713 
3714 	adapter->vf_mac_list = kcalloc(num_vf_mac_filters,
3715 				       sizeof(struct vf_mac_filter),
3716 				       GFP_KERNEL);
3717 
3718 	mac_list = adapter->vf_mac_list;
3719 	INIT_LIST_HEAD(&adapter->vf_macs.l);
3720 
3721 	if (adapter->vf_mac_list) {
3722 		/* Initialize list of VF MAC filters */
3723 		for (i = 0; i < num_vf_mac_filters; i++) {
3724 			mac_list->vf = -1;
3725 			mac_list->free = true;
3726 			list_add(&mac_list->l, &adapter->vf_macs.l);
3727 			mac_list++;
3728 		}
3729 	} else {
3730 		/* If we could not allocate memory for the VF MAC filters
3731 		 * we can continue without this feature but warn user.
3732 		 */
3733 		dev_err(&pdev->dev,
3734 			"Unable to allocate memory for VF MAC filter list\n");
3735 	}
3736 
3737 	/* only call pci_enable_sriov() if no VFs are allocated already */
3738 	if (!old_vfs) {
3739 		err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
3740 		if (err)
3741 			goto err_out;
3742 	}
3743 	dev_info(&pdev->dev, "%d VFs allocated\n",
3744 		 adapter->vfs_allocated_count);
3745 	for (i = 0; i < adapter->vfs_allocated_count; i++)
3746 		igb_vf_configure(adapter, i);
3747 
3748 	/* DMA Coalescing is not supported in IOV mode. */
3749 	adapter->flags &= ~IGB_FLAG_DMAC;
3750 	goto out;
3751 
3752 err_out:
3753 	kfree(adapter->vf_mac_list);
3754 	adapter->vf_mac_list = NULL;
3755 	kfree(adapter->vf_data);
3756 	adapter->vf_data = NULL;
3757 	adapter->vfs_allocated_count = 0;
3758 out:
3759 	return err;
3760 }
3761 
3762 #endif
3763 /**
3764  *  igb_remove_i2c - Cleanup  I2C interface
3765  *  @adapter: pointer to adapter structure
3766  **/
3767 static void igb_remove_i2c(struct igb_adapter *adapter)
3768 {
3769 	/* free the adapter bus structure */
3770 	i2c_del_adapter(&adapter->i2c_adap);
3771 }
3772 
3773 /**
3774  *  igb_remove - Device Removal Routine
3775  *  @pdev: PCI device information struct
3776  *
3777  *  igb_remove is called by the PCI subsystem to alert the driver
3778  *  that it should release a PCI device.  The could be caused by a
3779  *  Hot-Plug event, or because the driver is going to be removed from
3780  *  memory.
3781  **/
3782 static void igb_remove(struct pci_dev *pdev)
3783 {
3784 	struct net_device *netdev = pci_get_drvdata(pdev);
3785 	struct igb_adapter *adapter = netdev_priv(netdev);
3786 	struct e1000_hw *hw = &adapter->hw;
3787 
3788 	pm_runtime_get_noresume(&pdev->dev);
3789 #ifdef CONFIG_IGB_HWMON
3790 	igb_sysfs_exit(adapter);
3791 #endif
3792 	igb_remove_i2c(adapter);
3793 	igb_ptp_stop(adapter);
3794 	/* The watchdog timer may be rescheduled, so explicitly
3795 	 * disable watchdog from being rescheduled.
3796 	 */
3797 	set_bit(__IGB_DOWN, &adapter->state);
3798 	del_timer_sync(&adapter->watchdog_timer);
3799 	del_timer_sync(&adapter->phy_info_timer);
3800 
3801 	cancel_work_sync(&adapter->reset_task);
3802 	cancel_work_sync(&adapter->watchdog_task);
3803 
3804 #ifdef CONFIG_IGB_DCA
3805 	if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
3806 		dev_info(&pdev->dev, "DCA disabled\n");
3807 		dca_remove_requester(&pdev->dev);
3808 		adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
3809 		wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
3810 	}
3811 #endif
3812 
3813 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
3814 	 * would have already happened in close and is redundant.
3815 	 */
3816 	igb_release_hw_control(adapter);
3817 
3818 #ifdef CONFIG_PCI_IOV
3819 	rtnl_lock();
3820 	igb_disable_sriov(pdev);
3821 	rtnl_unlock();
3822 #endif
3823 
3824 	unregister_netdev(netdev);
3825 
3826 	igb_clear_interrupt_scheme(adapter);
3827 
3828 	pci_iounmap(pdev, adapter->io_addr);
3829 	if (hw->flash_address)
3830 		iounmap(hw->flash_address);
3831 	pci_release_mem_regions(pdev);
3832 
3833 	kfree(adapter->mac_table);
3834 	kfree(adapter->shadow_vfta);
3835 	free_netdev(netdev);
3836 
3837 	pci_disable_device(pdev);
3838 }
3839 
3840 /**
3841  *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
3842  *  @adapter: board private structure to initialize
3843  *
3844  *  This function initializes the vf specific data storage and then attempts to
3845  *  allocate the VFs.  The reason for ordering it this way is because it is much
3846  *  mor expensive time wise to disable SR-IOV than it is to allocate and free
3847  *  the memory for the VFs.
3848  **/
3849 static void igb_probe_vfs(struct igb_adapter *adapter)
3850 {
3851 #ifdef CONFIG_PCI_IOV
3852 	struct pci_dev *pdev = adapter->pdev;
3853 	struct e1000_hw *hw = &adapter->hw;
3854 
3855 	/* Virtualization features not supported on i210 family. */
3856 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
3857 		return;
3858 
3859 	/* Of the below we really only want the effect of getting
3860 	 * IGB_FLAG_HAS_MSIX set (if available), without which
3861 	 * igb_enable_sriov() has no effect.
3862 	 */
3863 	igb_set_interrupt_capability(adapter, true);
3864 	igb_reset_interrupt_capability(adapter);
3865 
3866 	pci_sriov_set_totalvfs(pdev, 7);
3867 	igb_enable_sriov(pdev, max_vfs);
3868 
3869 #endif /* CONFIG_PCI_IOV */
3870 }
3871 
3872 unsigned int igb_get_max_rss_queues(struct igb_adapter *adapter)
3873 {
3874 	struct e1000_hw *hw = &adapter->hw;
3875 	unsigned int max_rss_queues;
3876 
3877 	/* Determine the maximum number of RSS queues supported. */
3878 	switch (hw->mac.type) {
3879 	case e1000_i211:
3880 		max_rss_queues = IGB_MAX_RX_QUEUES_I211;
3881 		break;
3882 	case e1000_82575:
3883 	case e1000_i210:
3884 		max_rss_queues = IGB_MAX_RX_QUEUES_82575;
3885 		break;
3886 	case e1000_i350:
3887 		/* I350 cannot do RSS and SR-IOV at the same time */
3888 		if (!!adapter->vfs_allocated_count) {
3889 			max_rss_queues = 1;
3890 			break;
3891 		}
3892 		fallthrough;
3893 	case e1000_82576:
3894 		if (!!adapter->vfs_allocated_count) {
3895 			max_rss_queues = 2;
3896 			break;
3897 		}
3898 		fallthrough;
3899 	case e1000_82580:
3900 	case e1000_i354:
3901 	default:
3902 		max_rss_queues = IGB_MAX_RX_QUEUES;
3903 		break;
3904 	}
3905 
3906 	return max_rss_queues;
3907 }
3908 
3909 static void igb_init_queue_configuration(struct igb_adapter *adapter)
3910 {
3911 	u32 max_rss_queues;
3912 
3913 	max_rss_queues = igb_get_max_rss_queues(adapter);
3914 	adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
3915 
3916 	igb_set_flag_queue_pairs(adapter, max_rss_queues);
3917 }
3918 
3919 void igb_set_flag_queue_pairs(struct igb_adapter *adapter,
3920 			      const u32 max_rss_queues)
3921 {
3922 	struct e1000_hw *hw = &adapter->hw;
3923 
3924 	/* Determine if we need to pair queues. */
3925 	switch (hw->mac.type) {
3926 	case e1000_82575:
3927 	case e1000_i211:
3928 		/* Device supports enough interrupts without queue pairing. */
3929 		break;
3930 	case e1000_82576:
3931 	case e1000_82580:
3932 	case e1000_i350:
3933 	case e1000_i354:
3934 	case e1000_i210:
3935 	default:
3936 		/* If rss_queues > half of max_rss_queues, pair the queues in
3937 		 * order to conserve interrupts due to limited supply.
3938 		 */
3939 		if (adapter->rss_queues > (max_rss_queues / 2))
3940 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
3941 		else
3942 			adapter->flags &= ~IGB_FLAG_QUEUE_PAIRS;
3943 		break;
3944 	}
3945 }
3946 
3947 /**
3948  *  igb_sw_init - Initialize general software structures (struct igb_adapter)
3949  *  @adapter: board private structure to initialize
3950  *
3951  *  igb_sw_init initializes the Adapter private data structure.
3952  *  Fields are initialized based on PCI device information and
3953  *  OS network device settings (MTU size).
3954  **/
3955 static int igb_sw_init(struct igb_adapter *adapter)
3956 {
3957 	struct e1000_hw *hw = &adapter->hw;
3958 	struct net_device *netdev = adapter->netdev;
3959 	struct pci_dev *pdev = adapter->pdev;
3960 
3961 	pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
3962 
3963 	/* set default ring sizes */
3964 	adapter->tx_ring_count = IGB_DEFAULT_TXD;
3965 	adapter->rx_ring_count = IGB_DEFAULT_RXD;
3966 
3967 	/* set default ITR values */
3968 	adapter->rx_itr_setting = IGB_DEFAULT_ITR;
3969 	adapter->tx_itr_setting = IGB_DEFAULT_ITR;
3970 
3971 	/* set default work limits */
3972 	adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
3973 
3974 	adapter->max_frame_size = netdev->mtu + IGB_ETH_PKT_HDR_PAD;
3975 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
3976 
3977 	spin_lock_init(&adapter->nfc_lock);
3978 	spin_lock_init(&adapter->stats64_lock);
3979 
3980 	/* init spinlock to avoid concurrency of VF resources */
3981 	spin_lock_init(&adapter->vfs_lock);
3982 #ifdef CONFIG_PCI_IOV
3983 	switch (hw->mac.type) {
3984 	case e1000_82576:
3985 	case e1000_i350:
3986 		if (max_vfs > 7) {
3987 			dev_warn(&pdev->dev,
3988 				 "Maximum of 7 VFs per PF, using max\n");
3989 			max_vfs = adapter->vfs_allocated_count = 7;
3990 		} else
3991 			adapter->vfs_allocated_count = max_vfs;
3992 		if (adapter->vfs_allocated_count)
3993 			dev_warn(&pdev->dev,
3994 				 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
3995 		break;
3996 	default:
3997 		break;
3998 	}
3999 #endif /* CONFIG_PCI_IOV */
4000 
4001 	/* Assume MSI-X interrupts, will be checked during IRQ allocation */
4002 	adapter->flags |= IGB_FLAG_HAS_MSIX;
4003 
4004 	adapter->mac_table = kcalloc(hw->mac.rar_entry_count,
4005 				     sizeof(struct igb_mac_addr),
4006 				     GFP_KERNEL);
4007 	if (!adapter->mac_table)
4008 		return -ENOMEM;
4009 
4010 	igb_probe_vfs(adapter);
4011 
4012 	igb_init_queue_configuration(adapter);
4013 
4014 	/* Setup and initialize a copy of the hw vlan table array */
4015 	adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
4016 				       GFP_KERNEL);
4017 	if (!adapter->shadow_vfta)
4018 		return -ENOMEM;
4019 
4020 	/* This call may decrease the number of queues */
4021 	if (igb_init_interrupt_scheme(adapter, true)) {
4022 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
4023 		return -ENOMEM;
4024 	}
4025 
4026 	/* Explicitly disable IRQ since the NIC can be in any state. */
4027 	igb_irq_disable(adapter);
4028 
4029 	if (hw->mac.type >= e1000_i350)
4030 		adapter->flags &= ~IGB_FLAG_DMAC;
4031 
4032 	set_bit(__IGB_DOWN, &adapter->state);
4033 	return 0;
4034 }
4035 
4036 /**
4037  *  __igb_open - Called when a network interface is made active
4038  *  @netdev: network interface device structure
4039  *  @resuming: indicates whether we are in a resume call
4040  *
4041  *  Returns 0 on success, negative value on failure
4042  *
4043  *  The open entry point is called when a network interface is made
4044  *  active by the system (IFF_UP).  At this point all resources needed
4045  *  for transmit and receive operations are allocated, the interrupt
4046  *  handler is registered with the OS, the watchdog timer is started,
4047  *  and the stack is notified that the interface is ready.
4048  **/
4049 static int __igb_open(struct net_device *netdev, bool resuming)
4050 {
4051 	struct igb_adapter *adapter = netdev_priv(netdev);
4052 	struct e1000_hw *hw = &adapter->hw;
4053 	struct pci_dev *pdev = adapter->pdev;
4054 	int err;
4055 	int i;
4056 
4057 	/* disallow open during test */
4058 	if (test_bit(__IGB_TESTING, &adapter->state)) {
4059 		WARN_ON(resuming);
4060 		return -EBUSY;
4061 	}
4062 
4063 	if (!resuming)
4064 		pm_runtime_get_sync(&pdev->dev);
4065 
4066 	netif_carrier_off(netdev);
4067 
4068 	/* allocate transmit descriptors */
4069 	err = igb_setup_all_tx_resources(adapter);
4070 	if (err)
4071 		goto err_setup_tx;
4072 
4073 	/* allocate receive descriptors */
4074 	err = igb_setup_all_rx_resources(adapter);
4075 	if (err)
4076 		goto err_setup_rx;
4077 
4078 	igb_power_up_link(adapter);
4079 
4080 	/* before we allocate an interrupt, we must be ready to handle it.
4081 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4082 	 * as soon as we call pci_request_irq, so we have to setup our
4083 	 * clean_rx handler before we do so.
4084 	 */
4085 	igb_configure(adapter);
4086 
4087 	err = igb_request_irq(adapter);
4088 	if (err)
4089 		goto err_req_irq;
4090 
4091 	/* Notify the stack of the actual queue counts. */
4092 	err = netif_set_real_num_tx_queues(adapter->netdev,
4093 					   adapter->num_tx_queues);
4094 	if (err)
4095 		goto err_set_queues;
4096 
4097 	err = netif_set_real_num_rx_queues(adapter->netdev,
4098 					   adapter->num_rx_queues);
4099 	if (err)
4100 		goto err_set_queues;
4101 
4102 	/* From here on the code is the same as igb_up() */
4103 	clear_bit(__IGB_DOWN, &adapter->state);
4104 
4105 	for (i = 0; i < adapter->num_q_vectors; i++)
4106 		napi_enable(&(adapter->q_vector[i]->napi));
4107 
4108 	/* Clear any pending interrupts. */
4109 	rd32(E1000_TSICR);
4110 	rd32(E1000_ICR);
4111 
4112 	igb_irq_enable(adapter);
4113 
4114 	/* notify VFs that reset has been completed */
4115 	if (adapter->vfs_allocated_count) {
4116 		u32 reg_data = rd32(E1000_CTRL_EXT);
4117 
4118 		reg_data |= E1000_CTRL_EXT_PFRSTD;
4119 		wr32(E1000_CTRL_EXT, reg_data);
4120 	}
4121 
4122 	netif_tx_start_all_queues(netdev);
4123 
4124 	if (!resuming)
4125 		pm_runtime_put(&pdev->dev);
4126 
4127 	/* start the watchdog. */
4128 	hw->mac.get_link_status = 1;
4129 	schedule_work(&adapter->watchdog_task);
4130 
4131 	return 0;
4132 
4133 err_set_queues:
4134 	igb_free_irq(adapter);
4135 err_req_irq:
4136 	igb_release_hw_control(adapter);
4137 	igb_power_down_link(adapter);
4138 	igb_free_all_rx_resources(adapter);
4139 err_setup_rx:
4140 	igb_free_all_tx_resources(adapter);
4141 err_setup_tx:
4142 	igb_reset(adapter);
4143 	if (!resuming)
4144 		pm_runtime_put(&pdev->dev);
4145 
4146 	return err;
4147 }
4148 
4149 int igb_open(struct net_device *netdev)
4150 {
4151 	return __igb_open(netdev, false);
4152 }
4153 
4154 /**
4155  *  __igb_close - Disables a network interface
4156  *  @netdev: network interface device structure
4157  *  @suspending: indicates we are in a suspend call
4158  *
4159  *  Returns 0, this is not allowed to fail
4160  *
4161  *  The close entry point is called when an interface is de-activated
4162  *  by the OS.  The hardware is still under the driver's control, but
4163  *  needs to be disabled.  A global MAC reset is issued to stop the
4164  *  hardware, and all transmit and receive resources are freed.
4165  **/
4166 static int __igb_close(struct net_device *netdev, bool suspending)
4167 {
4168 	struct igb_adapter *adapter = netdev_priv(netdev);
4169 	struct pci_dev *pdev = adapter->pdev;
4170 
4171 	WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
4172 
4173 	if (!suspending)
4174 		pm_runtime_get_sync(&pdev->dev);
4175 
4176 	igb_down(adapter);
4177 	igb_free_irq(adapter);
4178 
4179 	igb_free_all_tx_resources(adapter);
4180 	igb_free_all_rx_resources(adapter);
4181 
4182 	if (!suspending)
4183 		pm_runtime_put_sync(&pdev->dev);
4184 	return 0;
4185 }
4186 
4187 int igb_close(struct net_device *netdev)
4188 {
4189 	if (netif_device_present(netdev) || netdev->dismantle)
4190 		return __igb_close(netdev, false);
4191 	return 0;
4192 }
4193 
4194 /**
4195  *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
4196  *  @tx_ring: tx descriptor ring (for a specific queue) to setup
4197  *
4198  *  Return 0 on success, negative on failure
4199  **/
4200 int igb_setup_tx_resources(struct igb_ring *tx_ring)
4201 {
4202 	struct device *dev = tx_ring->dev;
4203 	int size;
4204 
4205 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
4206 
4207 	tx_ring->tx_buffer_info = vmalloc(size);
4208 	if (!tx_ring->tx_buffer_info)
4209 		goto err;
4210 
4211 	/* round up to nearest 4K */
4212 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
4213 	tx_ring->size = ALIGN(tx_ring->size, 4096);
4214 
4215 	tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4216 					   &tx_ring->dma, GFP_KERNEL);
4217 	if (!tx_ring->desc)
4218 		goto err;
4219 
4220 	tx_ring->next_to_use = 0;
4221 	tx_ring->next_to_clean = 0;
4222 
4223 	return 0;
4224 
4225 err:
4226 	vfree(tx_ring->tx_buffer_info);
4227 	tx_ring->tx_buffer_info = NULL;
4228 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4229 	return -ENOMEM;
4230 }
4231 
4232 /**
4233  *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
4234  *				 (Descriptors) for all queues
4235  *  @adapter: board private structure
4236  *
4237  *  Return 0 on success, negative on failure
4238  **/
4239 static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
4240 {
4241 	struct pci_dev *pdev = adapter->pdev;
4242 	int i, err = 0;
4243 
4244 	for (i = 0; i < adapter->num_tx_queues; i++) {
4245 		err = igb_setup_tx_resources(adapter->tx_ring[i]);
4246 		if (err) {
4247 			dev_err(&pdev->dev,
4248 				"Allocation for Tx Queue %u failed\n", i);
4249 			for (i--; i >= 0; i--)
4250 				igb_free_tx_resources(adapter->tx_ring[i]);
4251 			break;
4252 		}
4253 	}
4254 
4255 	return err;
4256 }
4257 
4258 /**
4259  *  igb_setup_tctl - configure the transmit control registers
4260  *  @adapter: Board private structure
4261  **/
4262 void igb_setup_tctl(struct igb_adapter *adapter)
4263 {
4264 	struct e1000_hw *hw = &adapter->hw;
4265 	u32 tctl;
4266 
4267 	/* disable queue 0 which is enabled by default on 82575 and 82576 */
4268 	wr32(E1000_TXDCTL(0), 0);
4269 
4270 	/* Program the Transmit Control Register */
4271 	tctl = rd32(E1000_TCTL);
4272 	tctl &= ~E1000_TCTL_CT;
4273 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
4274 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
4275 
4276 	igb_config_collision_dist(hw);
4277 
4278 	/* Enable transmits */
4279 	tctl |= E1000_TCTL_EN;
4280 
4281 	wr32(E1000_TCTL, tctl);
4282 }
4283 
4284 /**
4285  *  igb_configure_tx_ring - Configure transmit ring after Reset
4286  *  @adapter: board private structure
4287  *  @ring: tx ring to configure
4288  *
4289  *  Configure a transmit ring after a reset.
4290  **/
4291 void igb_configure_tx_ring(struct igb_adapter *adapter,
4292 			   struct igb_ring *ring)
4293 {
4294 	struct e1000_hw *hw = &adapter->hw;
4295 	u32 txdctl = 0;
4296 	u64 tdba = ring->dma;
4297 	int reg_idx = ring->reg_idx;
4298 
4299 	wr32(E1000_TDLEN(reg_idx),
4300 	     ring->count * sizeof(union e1000_adv_tx_desc));
4301 	wr32(E1000_TDBAL(reg_idx),
4302 	     tdba & 0x00000000ffffffffULL);
4303 	wr32(E1000_TDBAH(reg_idx), tdba >> 32);
4304 
4305 	ring->tail = adapter->io_addr + E1000_TDT(reg_idx);
4306 	wr32(E1000_TDH(reg_idx), 0);
4307 	writel(0, ring->tail);
4308 
4309 	txdctl |= IGB_TX_PTHRESH;
4310 	txdctl |= IGB_TX_HTHRESH << 8;
4311 	txdctl |= IGB_TX_WTHRESH << 16;
4312 
4313 	/* reinitialize tx_buffer_info */
4314 	memset(ring->tx_buffer_info, 0,
4315 	       sizeof(struct igb_tx_buffer) * ring->count);
4316 
4317 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
4318 	wr32(E1000_TXDCTL(reg_idx), txdctl);
4319 }
4320 
4321 /**
4322  *  igb_configure_tx - Configure transmit Unit after Reset
4323  *  @adapter: board private structure
4324  *
4325  *  Configure the Tx unit of the MAC after a reset.
4326  **/
4327 static void igb_configure_tx(struct igb_adapter *adapter)
4328 {
4329 	struct e1000_hw *hw = &adapter->hw;
4330 	int i;
4331 
4332 	/* disable the queues */
4333 	for (i = 0; i < adapter->num_tx_queues; i++)
4334 		wr32(E1000_TXDCTL(adapter->tx_ring[i]->reg_idx), 0);
4335 
4336 	wrfl();
4337 	usleep_range(10000, 20000);
4338 
4339 	for (i = 0; i < adapter->num_tx_queues; i++)
4340 		igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
4341 }
4342 
4343 /**
4344  *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
4345  *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
4346  *
4347  *  Returns 0 on success, negative on failure
4348  **/
4349 int igb_setup_rx_resources(struct igb_ring *rx_ring)
4350 {
4351 	struct igb_adapter *adapter = netdev_priv(rx_ring->netdev);
4352 	struct device *dev = rx_ring->dev;
4353 	int size, res;
4354 
4355 	/* XDP RX-queue info */
4356 	if (xdp_rxq_info_is_reg(&rx_ring->xdp_rxq))
4357 		xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4358 	res = xdp_rxq_info_reg(&rx_ring->xdp_rxq, rx_ring->netdev,
4359 			       rx_ring->queue_index, 0);
4360 	if (res < 0) {
4361 		dev_err(dev, "Failed to register xdp_rxq index %u\n",
4362 			rx_ring->queue_index);
4363 		return res;
4364 	}
4365 
4366 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
4367 
4368 	rx_ring->rx_buffer_info = vmalloc(size);
4369 	if (!rx_ring->rx_buffer_info)
4370 		goto err;
4371 
4372 	/* Round up to nearest 4K */
4373 	rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
4374 	rx_ring->size = ALIGN(rx_ring->size, 4096);
4375 
4376 	rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4377 					   &rx_ring->dma, GFP_KERNEL);
4378 	if (!rx_ring->desc)
4379 		goto err;
4380 
4381 	rx_ring->next_to_alloc = 0;
4382 	rx_ring->next_to_clean = 0;
4383 	rx_ring->next_to_use = 0;
4384 
4385 	rx_ring->xdp_prog = adapter->xdp_prog;
4386 
4387 	return 0;
4388 
4389 err:
4390 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4391 	vfree(rx_ring->rx_buffer_info);
4392 	rx_ring->rx_buffer_info = NULL;
4393 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4394 	return -ENOMEM;
4395 }
4396 
4397 /**
4398  *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
4399  *				 (Descriptors) for all queues
4400  *  @adapter: board private structure
4401  *
4402  *  Return 0 on success, negative on failure
4403  **/
4404 static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
4405 {
4406 	struct pci_dev *pdev = adapter->pdev;
4407 	int i, err = 0;
4408 
4409 	for (i = 0; i < adapter->num_rx_queues; i++) {
4410 		err = igb_setup_rx_resources(adapter->rx_ring[i]);
4411 		if (err) {
4412 			dev_err(&pdev->dev,
4413 				"Allocation for Rx Queue %u failed\n", i);
4414 			for (i--; i >= 0; i--)
4415 				igb_free_rx_resources(adapter->rx_ring[i]);
4416 			break;
4417 		}
4418 	}
4419 
4420 	return err;
4421 }
4422 
4423 /**
4424  *  igb_setup_mrqc - configure the multiple receive queue control registers
4425  *  @adapter: Board private structure
4426  **/
4427 static void igb_setup_mrqc(struct igb_adapter *adapter)
4428 {
4429 	struct e1000_hw *hw = &adapter->hw;
4430 	u32 mrqc, rxcsum;
4431 	u32 j, num_rx_queues;
4432 	u32 rss_key[10];
4433 
4434 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
4435 	for (j = 0; j < 10; j++)
4436 		wr32(E1000_RSSRK(j), rss_key[j]);
4437 
4438 	num_rx_queues = adapter->rss_queues;
4439 
4440 	switch (hw->mac.type) {
4441 	case e1000_82576:
4442 		/* 82576 supports 2 RSS queues for SR-IOV */
4443 		if (adapter->vfs_allocated_count)
4444 			num_rx_queues = 2;
4445 		break;
4446 	default:
4447 		break;
4448 	}
4449 
4450 	if (adapter->rss_indir_tbl_init != num_rx_queues) {
4451 		for (j = 0; j < IGB_RETA_SIZE; j++)
4452 			adapter->rss_indir_tbl[j] =
4453 			(j * num_rx_queues) / IGB_RETA_SIZE;
4454 		adapter->rss_indir_tbl_init = num_rx_queues;
4455 	}
4456 	igb_write_rss_indir_tbl(adapter);
4457 
4458 	/* Disable raw packet checksumming so that RSS hash is placed in
4459 	 * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
4460 	 * offloads as they are enabled by default
4461 	 */
4462 	rxcsum = rd32(E1000_RXCSUM);
4463 	rxcsum |= E1000_RXCSUM_PCSD;
4464 
4465 	if (adapter->hw.mac.type >= e1000_82576)
4466 		/* Enable Receive Checksum Offload for SCTP */
4467 		rxcsum |= E1000_RXCSUM_CRCOFL;
4468 
4469 	/* Don't need to set TUOFL or IPOFL, they default to 1 */
4470 	wr32(E1000_RXCSUM, rxcsum);
4471 
4472 	/* Generate RSS hash based on packet types, TCP/UDP
4473 	 * port numbers and/or IPv4/v6 src and dst addresses
4474 	 */
4475 	mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
4476 	       E1000_MRQC_RSS_FIELD_IPV4_TCP |
4477 	       E1000_MRQC_RSS_FIELD_IPV6 |
4478 	       E1000_MRQC_RSS_FIELD_IPV6_TCP |
4479 	       E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
4480 
4481 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
4482 		mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
4483 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
4484 		mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
4485 
4486 	/* If VMDq is enabled then we set the appropriate mode for that, else
4487 	 * we default to RSS so that an RSS hash is calculated per packet even
4488 	 * if we are only using one queue
4489 	 */
4490 	if (adapter->vfs_allocated_count) {
4491 		if (hw->mac.type > e1000_82575) {
4492 			/* Set the default pool for the PF's first queue */
4493 			u32 vtctl = rd32(E1000_VT_CTL);
4494 
4495 			vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
4496 				   E1000_VT_CTL_DISABLE_DEF_POOL);
4497 			vtctl |= adapter->vfs_allocated_count <<
4498 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
4499 			wr32(E1000_VT_CTL, vtctl);
4500 		}
4501 		if (adapter->rss_queues > 1)
4502 			mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_MQ;
4503 		else
4504 			mrqc |= E1000_MRQC_ENABLE_VMDQ;
4505 	} else {
4506 		mrqc |= E1000_MRQC_ENABLE_RSS_MQ;
4507 	}
4508 	igb_vmm_control(adapter);
4509 
4510 	wr32(E1000_MRQC, mrqc);
4511 }
4512 
4513 /**
4514  *  igb_setup_rctl - configure the receive control registers
4515  *  @adapter: Board private structure
4516  **/
4517 void igb_setup_rctl(struct igb_adapter *adapter)
4518 {
4519 	struct e1000_hw *hw = &adapter->hw;
4520 	u32 rctl;
4521 
4522 	rctl = rd32(E1000_RCTL);
4523 
4524 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
4525 	rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
4526 
4527 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
4528 		(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
4529 
4530 	/* enable stripping of CRC. It's unlikely this will break BMC
4531 	 * redirection as it did with e1000. Newer features require
4532 	 * that the HW strips the CRC.
4533 	 */
4534 	rctl |= E1000_RCTL_SECRC;
4535 
4536 	/* disable store bad packets and clear size bits. */
4537 	rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
4538 
4539 	/* enable LPE to allow for reception of jumbo frames */
4540 	rctl |= E1000_RCTL_LPE;
4541 
4542 	/* disable queue 0 to prevent tail write w/o re-config */
4543 	wr32(E1000_RXDCTL(0), 0);
4544 
4545 	/* Attention!!!  For SR-IOV PF driver operations you must enable
4546 	 * queue drop for all VF and PF queues to prevent head of line blocking
4547 	 * if an un-trusted VF does not provide descriptors to hardware.
4548 	 */
4549 	if (adapter->vfs_allocated_count) {
4550 		/* set all queue drop enable bits */
4551 		wr32(E1000_QDE, ALL_QUEUES);
4552 	}
4553 
4554 	/* This is useful for sniffing bad packets. */
4555 	if (adapter->netdev->features & NETIF_F_RXALL) {
4556 		/* UPE and MPE will be handled by normal PROMISC logic
4557 		 * in e1000e_set_rx_mode
4558 		 */
4559 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
4560 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
4561 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
4562 
4563 		rctl &= ~(E1000_RCTL_DPF | /* Allow filtered pause */
4564 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
4565 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
4566 		 * and that breaks VLANs.
4567 		 */
4568 	}
4569 
4570 	wr32(E1000_RCTL, rctl);
4571 }
4572 
4573 static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
4574 				   int vfn)
4575 {
4576 	struct e1000_hw *hw = &adapter->hw;
4577 	u32 vmolr;
4578 
4579 	if (size > MAX_JUMBO_FRAME_SIZE)
4580 		size = MAX_JUMBO_FRAME_SIZE;
4581 
4582 	vmolr = rd32(E1000_VMOLR(vfn));
4583 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
4584 	vmolr |= size | E1000_VMOLR_LPE;
4585 	wr32(E1000_VMOLR(vfn), vmolr);
4586 
4587 	return 0;
4588 }
4589 
4590 static inline void igb_set_vf_vlan_strip(struct igb_adapter *adapter,
4591 					 int vfn, bool enable)
4592 {
4593 	struct e1000_hw *hw = &adapter->hw;
4594 	u32 val, reg;
4595 
4596 	if (hw->mac.type < e1000_82576)
4597 		return;
4598 
4599 	if (hw->mac.type == e1000_i350)
4600 		reg = E1000_DVMOLR(vfn);
4601 	else
4602 		reg = E1000_VMOLR(vfn);
4603 
4604 	val = rd32(reg);
4605 	if (enable)
4606 		val |= E1000_VMOLR_STRVLAN;
4607 	else
4608 		val &= ~(E1000_VMOLR_STRVLAN);
4609 	wr32(reg, val);
4610 }
4611 
4612 static inline void igb_set_vmolr(struct igb_adapter *adapter,
4613 				 int vfn, bool aupe)
4614 {
4615 	struct e1000_hw *hw = &adapter->hw;
4616 	u32 vmolr;
4617 
4618 	/* This register exists only on 82576 and newer so if we are older then
4619 	 * we should exit and do nothing
4620 	 */
4621 	if (hw->mac.type < e1000_82576)
4622 		return;
4623 
4624 	vmolr = rd32(E1000_VMOLR(vfn));
4625 	if (aupe)
4626 		vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
4627 	else
4628 		vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
4629 
4630 	/* clear all bits that might not be set */
4631 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
4632 
4633 	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
4634 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
4635 	/* for VMDq only allow the VFs and pool 0 to accept broadcast and
4636 	 * multicast packets
4637 	 */
4638 	if (vfn <= adapter->vfs_allocated_count)
4639 		vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
4640 
4641 	wr32(E1000_VMOLR(vfn), vmolr);
4642 }
4643 
4644 /**
4645  *  igb_setup_srrctl - configure the split and replication receive control
4646  *                     registers
4647  *  @adapter: Board private structure
4648  *  @ring: receive ring to be configured
4649  **/
4650 void igb_setup_srrctl(struct igb_adapter *adapter, struct igb_ring *ring)
4651 {
4652 	struct e1000_hw *hw = &adapter->hw;
4653 	int reg_idx = ring->reg_idx;
4654 	u32 srrctl = 0;
4655 
4656 	srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
4657 	if (ring_uses_large_buffer(ring))
4658 		srrctl |= IGB_RXBUFFER_3072 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4659 	else
4660 		srrctl |= IGB_RXBUFFER_2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4661 	srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
4662 	if (hw->mac.type >= e1000_82580)
4663 		srrctl |= E1000_SRRCTL_TIMESTAMP;
4664 	/* Only set Drop Enable if VFs allocated, or we are supporting multiple
4665 	 * queues and rx flow control is disabled
4666 	 */
4667 	if (adapter->vfs_allocated_count ||
4668 	    (!(hw->fc.current_mode & e1000_fc_rx_pause) &&
4669 	     adapter->num_rx_queues > 1))
4670 		srrctl |= E1000_SRRCTL_DROP_EN;
4671 
4672 	wr32(E1000_SRRCTL(reg_idx), srrctl);
4673 }
4674 
4675 /**
4676  *  igb_configure_rx_ring - Configure a receive ring after Reset
4677  *  @adapter: board private structure
4678  *  @ring: receive ring to be configured
4679  *
4680  *  Configure the Rx unit of the MAC after a reset.
4681  **/
4682 void igb_configure_rx_ring(struct igb_adapter *adapter,
4683 			   struct igb_ring *ring)
4684 {
4685 	struct e1000_hw *hw = &adapter->hw;
4686 	union e1000_adv_rx_desc *rx_desc;
4687 	u64 rdba = ring->dma;
4688 	int reg_idx = ring->reg_idx;
4689 	u32 rxdctl = 0;
4690 
4691 	xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4692 	WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4693 					   MEM_TYPE_PAGE_SHARED, NULL));
4694 
4695 	/* disable the queue */
4696 	wr32(E1000_RXDCTL(reg_idx), 0);
4697 
4698 	/* Set DMA base address registers */
4699 	wr32(E1000_RDBAL(reg_idx),
4700 	     rdba & 0x00000000ffffffffULL);
4701 	wr32(E1000_RDBAH(reg_idx), rdba >> 32);
4702 	wr32(E1000_RDLEN(reg_idx),
4703 	     ring->count * sizeof(union e1000_adv_rx_desc));
4704 
4705 	/* initialize head and tail */
4706 	ring->tail = adapter->io_addr + E1000_RDT(reg_idx);
4707 	wr32(E1000_RDH(reg_idx), 0);
4708 	writel(0, ring->tail);
4709 
4710 	/* set descriptor configuration */
4711 	igb_setup_srrctl(adapter, ring);
4712 
4713 	/* set filtering for VMDQ pools */
4714 	igb_set_vmolr(adapter, reg_idx & 0x7, true);
4715 
4716 	rxdctl |= IGB_RX_PTHRESH;
4717 	rxdctl |= IGB_RX_HTHRESH << 8;
4718 	rxdctl |= IGB_RX_WTHRESH << 16;
4719 
4720 	/* initialize rx_buffer_info */
4721 	memset(ring->rx_buffer_info, 0,
4722 	       sizeof(struct igb_rx_buffer) * ring->count);
4723 
4724 	/* initialize Rx descriptor 0 */
4725 	rx_desc = IGB_RX_DESC(ring, 0);
4726 	rx_desc->wb.upper.length = 0;
4727 
4728 	/* enable receive descriptor fetching */
4729 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
4730 	wr32(E1000_RXDCTL(reg_idx), rxdctl);
4731 }
4732 
4733 static void igb_set_rx_buffer_len(struct igb_adapter *adapter,
4734 				  struct igb_ring *rx_ring)
4735 {
4736 	/* set build_skb and buffer size flags */
4737 	clear_ring_build_skb_enabled(rx_ring);
4738 	clear_ring_uses_large_buffer(rx_ring);
4739 
4740 	if (adapter->flags & IGB_FLAG_RX_LEGACY)
4741 		return;
4742 
4743 	set_ring_build_skb_enabled(rx_ring);
4744 
4745 #if (PAGE_SIZE < 8192)
4746 	if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4747 		return;
4748 
4749 	set_ring_uses_large_buffer(rx_ring);
4750 #endif
4751 }
4752 
4753 /**
4754  *  igb_configure_rx - Configure receive Unit after Reset
4755  *  @adapter: board private structure
4756  *
4757  *  Configure the Rx unit of the MAC after a reset.
4758  **/
4759 static void igb_configure_rx(struct igb_adapter *adapter)
4760 {
4761 	int i;
4762 
4763 	/* set the correct pool for the PF default MAC address in entry 0 */
4764 	igb_set_default_mac_filter(adapter);
4765 
4766 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
4767 	 * the Base and Length of the Rx Descriptor Ring
4768 	 */
4769 	for (i = 0; i < adapter->num_rx_queues; i++) {
4770 		struct igb_ring *rx_ring = adapter->rx_ring[i];
4771 
4772 		igb_set_rx_buffer_len(adapter, rx_ring);
4773 		igb_configure_rx_ring(adapter, rx_ring);
4774 	}
4775 }
4776 
4777 /**
4778  *  igb_free_tx_resources - Free Tx Resources per Queue
4779  *  @tx_ring: Tx descriptor ring for a specific queue
4780  *
4781  *  Free all transmit software resources
4782  **/
4783 void igb_free_tx_resources(struct igb_ring *tx_ring)
4784 {
4785 	igb_clean_tx_ring(tx_ring);
4786 
4787 	vfree(tx_ring->tx_buffer_info);
4788 	tx_ring->tx_buffer_info = NULL;
4789 
4790 	/* if not set, then don't free */
4791 	if (!tx_ring->desc)
4792 		return;
4793 
4794 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4795 			  tx_ring->desc, tx_ring->dma);
4796 
4797 	tx_ring->desc = NULL;
4798 }
4799 
4800 /**
4801  *  igb_free_all_tx_resources - Free Tx Resources for All Queues
4802  *  @adapter: board private structure
4803  *
4804  *  Free all transmit software resources
4805  **/
4806 static void igb_free_all_tx_resources(struct igb_adapter *adapter)
4807 {
4808 	int i;
4809 
4810 	for (i = 0; i < adapter->num_tx_queues; i++)
4811 		if (adapter->tx_ring[i])
4812 			igb_free_tx_resources(adapter->tx_ring[i]);
4813 }
4814 
4815 /**
4816  *  igb_clean_tx_ring - Free Tx Buffers
4817  *  @tx_ring: ring to be cleaned
4818  **/
4819 static void igb_clean_tx_ring(struct igb_ring *tx_ring)
4820 {
4821 	u16 i = tx_ring->next_to_clean;
4822 	struct igb_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
4823 
4824 	while (i != tx_ring->next_to_use) {
4825 		union e1000_adv_tx_desc *eop_desc, *tx_desc;
4826 
4827 		/* Free all the Tx ring sk_buffs or xdp frames */
4828 		if (tx_buffer->type == IGB_TYPE_SKB)
4829 			dev_kfree_skb_any(tx_buffer->skb);
4830 		else
4831 			xdp_return_frame(tx_buffer->xdpf);
4832 
4833 		/* unmap skb header data */
4834 		dma_unmap_single(tx_ring->dev,
4835 				 dma_unmap_addr(tx_buffer, dma),
4836 				 dma_unmap_len(tx_buffer, len),
4837 				 DMA_TO_DEVICE);
4838 
4839 		/* check for eop_desc to determine the end of the packet */
4840 		eop_desc = tx_buffer->next_to_watch;
4841 		tx_desc = IGB_TX_DESC(tx_ring, i);
4842 
4843 		/* unmap remaining buffers */
4844 		while (tx_desc != eop_desc) {
4845 			tx_buffer++;
4846 			tx_desc++;
4847 			i++;
4848 			if (unlikely(i == tx_ring->count)) {
4849 				i = 0;
4850 				tx_buffer = tx_ring->tx_buffer_info;
4851 				tx_desc = IGB_TX_DESC(tx_ring, 0);
4852 			}
4853 
4854 			/* unmap any remaining paged data */
4855 			if (dma_unmap_len(tx_buffer, len))
4856 				dma_unmap_page(tx_ring->dev,
4857 					       dma_unmap_addr(tx_buffer, dma),
4858 					       dma_unmap_len(tx_buffer, len),
4859 					       DMA_TO_DEVICE);
4860 		}
4861 
4862 		tx_buffer->next_to_watch = NULL;
4863 
4864 		/* move us one more past the eop_desc for start of next pkt */
4865 		tx_buffer++;
4866 		i++;
4867 		if (unlikely(i == tx_ring->count)) {
4868 			i = 0;
4869 			tx_buffer = tx_ring->tx_buffer_info;
4870 		}
4871 	}
4872 
4873 	/* reset BQL for queue */
4874 	netdev_tx_reset_queue(txring_txq(tx_ring));
4875 
4876 	/* reset next_to_use and next_to_clean */
4877 	tx_ring->next_to_use = 0;
4878 	tx_ring->next_to_clean = 0;
4879 }
4880 
4881 /**
4882  *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
4883  *  @adapter: board private structure
4884  **/
4885 static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
4886 {
4887 	int i;
4888 
4889 	for (i = 0; i < adapter->num_tx_queues; i++)
4890 		if (adapter->tx_ring[i])
4891 			igb_clean_tx_ring(adapter->tx_ring[i]);
4892 }
4893 
4894 /**
4895  *  igb_free_rx_resources - Free Rx Resources
4896  *  @rx_ring: ring to clean the resources from
4897  *
4898  *  Free all receive software resources
4899  **/
4900 void igb_free_rx_resources(struct igb_ring *rx_ring)
4901 {
4902 	igb_clean_rx_ring(rx_ring);
4903 
4904 	rx_ring->xdp_prog = NULL;
4905 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4906 	vfree(rx_ring->rx_buffer_info);
4907 	rx_ring->rx_buffer_info = NULL;
4908 
4909 	/* if not set, then don't free */
4910 	if (!rx_ring->desc)
4911 		return;
4912 
4913 	dma_free_coherent(rx_ring->dev, rx_ring->size,
4914 			  rx_ring->desc, rx_ring->dma);
4915 
4916 	rx_ring->desc = NULL;
4917 }
4918 
4919 /**
4920  *  igb_free_all_rx_resources - Free Rx Resources for All Queues
4921  *  @adapter: board private structure
4922  *
4923  *  Free all receive software resources
4924  **/
4925 static void igb_free_all_rx_resources(struct igb_adapter *adapter)
4926 {
4927 	int i;
4928 
4929 	for (i = 0; i < adapter->num_rx_queues; i++)
4930 		if (adapter->rx_ring[i])
4931 			igb_free_rx_resources(adapter->rx_ring[i]);
4932 }
4933 
4934 /**
4935  *  igb_clean_rx_ring - Free Rx Buffers per Queue
4936  *  @rx_ring: ring to free buffers from
4937  **/
4938 static void igb_clean_rx_ring(struct igb_ring *rx_ring)
4939 {
4940 	u16 i = rx_ring->next_to_clean;
4941 
4942 	dev_kfree_skb(rx_ring->skb);
4943 	rx_ring->skb = NULL;
4944 
4945 	/* Free all the Rx ring sk_buffs */
4946 	while (i != rx_ring->next_to_alloc) {
4947 		struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
4948 
4949 		/* Invalidate cache lines that may have been written to by
4950 		 * device so that we avoid corrupting memory.
4951 		 */
4952 		dma_sync_single_range_for_cpu(rx_ring->dev,
4953 					      buffer_info->dma,
4954 					      buffer_info->page_offset,
4955 					      igb_rx_bufsz(rx_ring),
4956 					      DMA_FROM_DEVICE);
4957 
4958 		/* free resources associated with mapping */
4959 		dma_unmap_page_attrs(rx_ring->dev,
4960 				     buffer_info->dma,
4961 				     igb_rx_pg_size(rx_ring),
4962 				     DMA_FROM_DEVICE,
4963 				     IGB_RX_DMA_ATTR);
4964 		__page_frag_cache_drain(buffer_info->page,
4965 					buffer_info->pagecnt_bias);
4966 
4967 		i++;
4968 		if (i == rx_ring->count)
4969 			i = 0;
4970 	}
4971 
4972 	rx_ring->next_to_alloc = 0;
4973 	rx_ring->next_to_clean = 0;
4974 	rx_ring->next_to_use = 0;
4975 }
4976 
4977 /**
4978  *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
4979  *  @adapter: board private structure
4980  **/
4981 static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
4982 {
4983 	int i;
4984 
4985 	for (i = 0; i < adapter->num_rx_queues; i++)
4986 		if (adapter->rx_ring[i])
4987 			igb_clean_rx_ring(adapter->rx_ring[i]);
4988 }
4989 
4990 /**
4991  *  igb_set_mac - Change the Ethernet Address of the NIC
4992  *  @netdev: network interface device structure
4993  *  @p: pointer to an address structure
4994  *
4995  *  Returns 0 on success, negative on failure
4996  **/
4997 static int igb_set_mac(struct net_device *netdev, void *p)
4998 {
4999 	struct igb_adapter *adapter = netdev_priv(netdev);
5000 	struct e1000_hw *hw = &adapter->hw;
5001 	struct sockaddr *addr = p;
5002 
5003 	if (!is_valid_ether_addr(addr->sa_data))
5004 		return -EADDRNOTAVAIL;
5005 
5006 	eth_hw_addr_set(netdev, addr->sa_data);
5007 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
5008 
5009 	/* set the correct pool for the new PF MAC address in entry 0 */
5010 	igb_set_default_mac_filter(adapter);
5011 
5012 	return 0;
5013 }
5014 
5015 /**
5016  *  igb_write_mc_addr_list - write multicast addresses to MTA
5017  *  @netdev: network interface device structure
5018  *
5019  *  Writes multicast address list to the MTA hash table.
5020  *  Returns: -ENOMEM on failure
5021  *           0 on no addresses written
5022  *           X on writing X addresses to MTA
5023  **/
5024 static int igb_write_mc_addr_list(struct net_device *netdev)
5025 {
5026 	struct igb_adapter *adapter = netdev_priv(netdev);
5027 	struct e1000_hw *hw = &adapter->hw;
5028 	struct netdev_hw_addr *ha;
5029 	u8  *mta_list;
5030 	int i;
5031 
5032 	if (netdev_mc_empty(netdev)) {
5033 		/* nothing to program, so clear mc list */
5034 		igb_update_mc_addr_list(hw, NULL, 0);
5035 		igb_restore_vf_multicasts(adapter);
5036 		return 0;
5037 	}
5038 
5039 	mta_list = kcalloc(netdev_mc_count(netdev), 6, GFP_ATOMIC);
5040 	if (!mta_list)
5041 		return -ENOMEM;
5042 
5043 	/* The shared function expects a packed array of only addresses. */
5044 	i = 0;
5045 	netdev_for_each_mc_addr(ha, netdev)
5046 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
5047 
5048 	igb_update_mc_addr_list(hw, mta_list, i);
5049 	kfree(mta_list);
5050 
5051 	return netdev_mc_count(netdev);
5052 }
5053 
5054 static int igb_vlan_promisc_enable(struct igb_adapter *adapter)
5055 {
5056 	struct e1000_hw *hw = &adapter->hw;
5057 	u32 i, pf_id;
5058 
5059 	switch (hw->mac.type) {
5060 	case e1000_i210:
5061 	case e1000_i211:
5062 	case e1000_i350:
5063 		/* VLAN filtering needed for VLAN prio filter */
5064 		if (adapter->netdev->features & NETIF_F_NTUPLE)
5065 			break;
5066 		fallthrough;
5067 	case e1000_82576:
5068 	case e1000_82580:
5069 	case e1000_i354:
5070 		/* VLAN filtering needed for pool filtering */
5071 		if (adapter->vfs_allocated_count)
5072 			break;
5073 		fallthrough;
5074 	default:
5075 		return 1;
5076 	}
5077 
5078 	/* We are already in VLAN promisc, nothing to do */
5079 	if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
5080 		return 0;
5081 
5082 	if (!adapter->vfs_allocated_count)
5083 		goto set_vfta;
5084 
5085 	/* Add PF to all active pools */
5086 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
5087 
5088 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
5089 		u32 vlvf = rd32(E1000_VLVF(i));
5090 
5091 		vlvf |= BIT(pf_id);
5092 		wr32(E1000_VLVF(i), vlvf);
5093 	}
5094 
5095 set_vfta:
5096 	/* Set all bits in the VLAN filter table array */
5097 	for (i = E1000_VLAN_FILTER_TBL_SIZE; i--;)
5098 		hw->mac.ops.write_vfta(hw, i, ~0U);
5099 
5100 	/* Set flag so we don't redo unnecessary work */
5101 	adapter->flags |= IGB_FLAG_VLAN_PROMISC;
5102 
5103 	return 0;
5104 }
5105 
5106 #define VFTA_BLOCK_SIZE 8
5107 static void igb_scrub_vfta(struct igb_adapter *adapter, u32 vfta_offset)
5108 {
5109 	struct e1000_hw *hw = &adapter->hw;
5110 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
5111 	u32 vid_start = vfta_offset * 32;
5112 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
5113 	u32 i, vid, word, bits, pf_id;
5114 
5115 	/* guarantee that we don't scrub out management VLAN */
5116 	vid = adapter->mng_vlan_id;
5117 	if (vid >= vid_start && vid < vid_end)
5118 		vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
5119 
5120 	if (!adapter->vfs_allocated_count)
5121 		goto set_vfta;
5122 
5123 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
5124 
5125 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
5126 		u32 vlvf = rd32(E1000_VLVF(i));
5127 
5128 		/* pull VLAN ID from VLVF */
5129 		vid = vlvf & VLAN_VID_MASK;
5130 
5131 		/* only concern ourselves with a certain range */
5132 		if (vid < vid_start || vid >= vid_end)
5133 			continue;
5134 
5135 		if (vlvf & E1000_VLVF_VLANID_ENABLE) {
5136 			/* record VLAN ID in VFTA */
5137 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
5138 
5139 			/* if PF is part of this then continue */
5140 			if (test_bit(vid, adapter->active_vlans))
5141 				continue;
5142 		}
5143 
5144 		/* remove PF from the pool */
5145 		bits = ~BIT(pf_id);
5146 		bits &= rd32(E1000_VLVF(i));
5147 		wr32(E1000_VLVF(i), bits);
5148 	}
5149 
5150 set_vfta:
5151 	/* extract values from active_vlans and write back to VFTA */
5152 	for (i = VFTA_BLOCK_SIZE; i--;) {
5153 		vid = (vfta_offset + i) * 32;
5154 		word = vid / BITS_PER_LONG;
5155 		bits = vid % BITS_PER_LONG;
5156 
5157 		vfta[i] |= adapter->active_vlans[word] >> bits;
5158 
5159 		hw->mac.ops.write_vfta(hw, vfta_offset + i, vfta[i]);
5160 	}
5161 }
5162 
5163 static void igb_vlan_promisc_disable(struct igb_adapter *adapter)
5164 {
5165 	u32 i;
5166 
5167 	/* We are not in VLAN promisc, nothing to do */
5168 	if (!(adapter->flags & IGB_FLAG_VLAN_PROMISC))
5169 		return;
5170 
5171 	/* Set flag so we don't redo unnecessary work */
5172 	adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
5173 
5174 	for (i = 0; i < E1000_VLAN_FILTER_TBL_SIZE; i += VFTA_BLOCK_SIZE)
5175 		igb_scrub_vfta(adapter, i);
5176 }
5177 
5178 /**
5179  *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
5180  *  @netdev: network interface device structure
5181  *
5182  *  The set_rx_mode entry point is called whenever the unicast or multicast
5183  *  address lists or the network interface flags are updated.  This routine is
5184  *  responsible for configuring the hardware for proper unicast, multicast,
5185  *  promiscuous mode, and all-multi behavior.
5186  **/
5187 static void igb_set_rx_mode(struct net_device *netdev)
5188 {
5189 	struct igb_adapter *adapter = netdev_priv(netdev);
5190 	struct e1000_hw *hw = &adapter->hw;
5191 	unsigned int vfn = adapter->vfs_allocated_count;
5192 	u32 rctl = 0, vmolr = 0, rlpml = MAX_JUMBO_FRAME_SIZE;
5193 	int count;
5194 
5195 	/* Check for Promiscuous and All Multicast modes */
5196 	if (netdev->flags & IFF_PROMISC) {
5197 		rctl |= E1000_RCTL_UPE | E1000_RCTL_MPE;
5198 		vmolr |= E1000_VMOLR_MPME;
5199 
5200 		/* enable use of UTA filter to force packets to default pool */
5201 		if (hw->mac.type == e1000_82576)
5202 			vmolr |= E1000_VMOLR_ROPE;
5203 	} else {
5204 		if (netdev->flags & IFF_ALLMULTI) {
5205 			rctl |= E1000_RCTL_MPE;
5206 			vmolr |= E1000_VMOLR_MPME;
5207 		} else {
5208 			/* Write addresses to the MTA, if the attempt fails
5209 			 * then we should just turn on promiscuous mode so
5210 			 * that we can at least receive multicast traffic
5211 			 */
5212 			count = igb_write_mc_addr_list(netdev);
5213 			if (count < 0) {
5214 				rctl |= E1000_RCTL_MPE;
5215 				vmolr |= E1000_VMOLR_MPME;
5216 			} else if (count) {
5217 				vmolr |= E1000_VMOLR_ROMPE;
5218 			}
5219 		}
5220 	}
5221 
5222 	/* Write addresses to available RAR registers, if there is not
5223 	 * sufficient space to store all the addresses then enable
5224 	 * unicast promiscuous mode
5225 	 */
5226 	if (__dev_uc_sync(netdev, igb_uc_sync, igb_uc_unsync)) {
5227 		rctl |= E1000_RCTL_UPE;
5228 		vmolr |= E1000_VMOLR_ROPE;
5229 	}
5230 
5231 	/* enable VLAN filtering by default */
5232 	rctl |= E1000_RCTL_VFE;
5233 
5234 	/* disable VLAN filtering for modes that require it */
5235 	if ((netdev->flags & IFF_PROMISC) ||
5236 	    (netdev->features & NETIF_F_RXALL)) {
5237 		/* if we fail to set all rules then just clear VFE */
5238 		if (igb_vlan_promisc_enable(adapter))
5239 			rctl &= ~E1000_RCTL_VFE;
5240 	} else {
5241 		igb_vlan_promisc_disable(adapter);
5242 	}
5243 
5244 	/* update state of unicast, multicast, and VLAN filtering modes */
5245 	rctl |= rd32(E1000_RCTL) & ~(E1000_RCTL_UPE | E1000_RCTL_MPE |
5246 				     E1000_RCTL_VFE);
5247 	wr32(E1000_RCTL, rctl);
5248 
5249 #if (PAGE_SIZE < 8192)
5250 	if (!adapter->vfs_allocated_count) {
5251 		if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5252 			rlpml = IGB_MAX_FRAME_BUILD_SKB;
5253 	}
5254 #endif
5255 	wr32(E1000_RLPML, rlpml);
5256 
5257 	/* In order to support SR-IOV and eventually VMDq it is necessary to set
5258 	 * the VMOLR to enable the appropriate modes.  Without this workaround
5259 	 * we will have issues with VLAN tag stripping not being done for frames
5260 	 * that are only arriving because we are the default pool
5261 	 */
5262 	if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
5263 		return;
5264 
5265 	/* set UTA to appropriate mode */
5266 	igb_set_uta(adapter, !!(vmolr & E1000_VMOLR_ROPE));
5267 
5268 	vmolr |= rd32(E1000_VMOLR(vfn)) &
5269 		 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
5270 
5271 	/* enable Rx jumbo frames, restrict as needed to support build_skb */
5272 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
5273 #if (PAGE_SIZE < 8192)
5274 	if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5275 		vmolr |= IGB_MAX_FRAME_BUILD_SKB;
5276 	else
5277 #endif
5278 		vmolr |= MAX_JUMBO_FRAME_SIZE;
5279 	vmolr |= E1000_VMOLR_LPE;
5280 
5281 	wr32(E1000_VMOLR(vfn), vmolr);
5282 
5283 	igb_restore_vf_multicasts(adapter);
5284 }
5285 
5286 static void igb_check_wvbr(struct igb_adapter *adapter)
5287 {
5288 	struct e1000_hw *hw = &adapter->hw;
5289 	u32 wvbr = 0;
5290 
5291 	switch (hw->mac.type) {
5292 	case e1000_82576:
5293 	case e1000_i350:
5294 		wvbr = rd32(E1000_WVBR);
5295 		if (!wvbr)
5296 			return;
5297 		break;
5298 	default:
5299 		break;
5300 	}
5301 
5302 	adapter->wvbr |= wvbr;
5303 }
5304 
5305 #define IGB_STAGGERED_QUEUE_OFFSET 8
5306 
5307 static void igb_spoof_check(struct igb_adapter *adapter)
5308 {
5309 	int j;
5310 
5311 	if (!adapter->wvbr)
5312 		return;
5313 
5314 	for (j = 0; j < adapter->vfs_allocated_count; j++) {
5315 		if (adapter->wvbr & BIT(j) ||
5316 		    adapter->wvbr & BIT(j + IGB_STAGGERED_QUEUE_OFFSET)) {
5317 			dev_warn(&adapter->pdev->dev,
5318 				"Spoof event(s) detected on VF %d\n", j);
5319 			adapter->wvbr &=
5320 				~(BIT(j) |
5321 				  BIT(j + IGB_STAGGERED_QUEUE_OFFSET));
5322 		}
5323 	}
5324 }
5325 
5326 /* Need to wait a few seconds after link up to get diagnostic information from
5327  * the phy
5328  */
5329 static void igb_update_phy_info(struct timer_list *t)
5330 {
5331 	struct igb_adapter *adapter = from_timer(adapter, t, phy_info_timer);
5332 	igb_get_phy_info(&adapter->hw);
5333 }
5334 
5335 /**
5336  *  igb_has_link - check shared code for link and determine up/down
5337  *  @adapter: pointer to driver private info
5338  **/
5339 bool igb_has_link(struct igb_adapter *adapter)
5340 {
5341 	struct e1000_hw *hw = &adapter->hw;
5342 	bool link_active = false;
5343 
5344 	/* get_link_status is set on LSC (link status) interrupt or
5345 	 * rx sequence error interrupt.  get_link_status will stay
5346 	 * false until the e1000_check_for_link establishes link
5347 	 * for copper adapters ONLY
5348 	 */
5349 	switch (hw->phy.media_type) {
5350 	case e1000_media_type_copper:
5351 		if (!hw->mac.get_link_status)
5352 			return true;
5353 		fallthrough;
5354 	case e1000_media_type_internal_serdes:
5355 		hw->mac.ops.check_for_link(hw);
5356 		link_active = !hw->mac.get_link_status;
5357 		break;
5358 	default:
5359 	case e1000_media_type_unknown:
5360 		break;
5361 	}
5362 
5363 	if (((hw->mac.type == e1000_i210) ||
5364 	     (hw->mac.type == e1000_i211)) &&
5365 	     (hw->phy.id == I210_I_PHY_ID)) {
5366 		if (!netif_carrier_ok(adapter->netdev)) {
5367 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5368 		} else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
5369 			adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
5370 			adapter->link_check_timeout = jiffies;
5371 		}
5372 	}
5373 
5374 	return link_active;
5375 }
5376 
5377 static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
5378 {
5379 	bool ret = false;
5380 	u32 ctrl_ext, thstat;
5381 
5382 	/* check for thermal sensor event on i350 copper only */
5383 	if (hw->mac.type == e1000_i350) {
5384 		thstat = rd32(E1000_THSTAT);
5385 		ctrl_ext = rd32(E1000_CTRL_EXT);
5386 
5387 		if ((hw->phy.media_type == e1000_media_type_copper) &&
5388 		    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
5389 			ret = !!(thstat & event);
5390 	}
5391 
5392 	return ret;
5393 }
5394 
5395 /**
5396  *  igb_check_lvmmc - check for malformed packets received
5397  *  and indicated in LVMMC register
5398  *  @adapter: pointer to adapter
5399  **/
5400 static void igb_check_lvmmc(struct igb_adapter *adapter)
5401 {
5402 	struct e1000_hw *hw = &adapter->hw;
5403 	u32 lvmmc;
5404 
5405 	lvmmc = rd32(E1000_LVMMC);
5406 	if (lvmmc) {
5407 		if (unlikely(net_ratelimit())) {
5408 			netdev_warn(adapter->netdev,
5409 				    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
5410 				    lvmmc);
5411 		}
5412 	}
5413 }
5414 
5415 /**
5416  *  igb_watchdog - Timer Call-back
5417  *  @t: pointer to timer_list containing our private info pointer
5418  **/
5419 static void igb_watchdog(struct timer_list *t)
5420 {
5421 	struct igb_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5422 	/* Do the rest outside of interrupt context */
5423 	schedule_work(&adapter->watchdog_task);
5424 }
5425 
5426 static void igb_watchdog_task(struct work_struct *work)
5427 {
5428 	struct igb_adapter *adapter = container_of(work,
5429 						   struct igb_adapter,
5430 						   watchdog_task);
5431 	struct e1000_hw *hw = &adapter->hw;
5432 	struct e1000_phy_info *phy = &hw->phy;
5433 	struct net_device *netdev = adapter->netdev;
5434 	u32 link;
5435 	int i;
5436 	u32 connsw;
5437 	u16 phy_data, retry_count = 20;
5438 
5439 	link = igb_has_link(adapter);
5440 
5441 	if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
5442 		if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
5443 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5444 		else
5445 			link = false;
5446 	}
5447 
5448 	/* Force link down if we have fiber to swap to */
5449 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5450 		if (hw->phy.media_type == e1000_media_type_copper) {
5451 			connsw = rd32(E1000_CONNSW);
5452 			if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
5453 				link = 0;
5454 		}
5455 	}
5456 	if (link) {
5457 		/* Perform a reset if the media type changed. */
5458 		if (hw->dev_spec._82575.media_changed) {
5459 			hw->dev_spec._82575.media_changed = false;
5460 			adapter->flags |= IGB_FLAG_MEDIA_RESET;
5461 			igb_reset(adapter);
5462 		}
5463 		/* Cancel scheduled suspend requests. */
5464 		pm_runtime_resume(netdev->dev.parent);
5465 
5466 		if (!netif_carrier_ok(netdev)) {
5467 			u32 ctrl;
5468 
5469 			hw->mac.ops.get_speed_and_duplex(hw,
5470 							 &adapter->link_speed,
5471 							 &adapter->link_duplex);
5472 
5473 			ctrl = rd32(E1000_CTRL);
5474 			/* Links status message must follow this format */
5475 			netdev_info(netdev,
5476 			       "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5477 			       netdev->name,
5478 			       adapter->link_speed,
5479 			       adapter->link_duplex == FULL_DUPLEX ?
5480 			       "Full" : "Half",
5481 			       (ctrl & E1000_CTRL_TFCE) &&
5482 			       (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
5483 			       (ctrl & E1000_CTRL_RFCE) ?  "RX" :
5484 			       (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
5485 
5486 			/* disable EEE if enabled */
5487 			if ((adapter->flags & IGB_FLAG_EEE) &&
5488 				(adapter->link_duplex == HALF_DUPLEX)) {
5489 				dev_info(&adapter->pdev->dev,
5490 				"EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
5491 				adapter->hw.dev_spec._82575.eee_disable = true;
5492 				adapter->flags &= ~IGB_FLAG_EEE;
5493 			}
5494 
5495 			/* check if SmartSpeed worked */
5496 			igb_check_downshift(hw);
5497 			if (phy->speed_downgraded)
5498 				netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
5499 
5500 			/* check for thermal sensor event */
5501 			if (igb_thermal_sensor_event(hw,
5502 			    E1000_THSTAT_LINK_THROTTLE))
5503 				netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
5504 
5505 			/* adjust timeout factor according to speed/duplex */
5506 			adapter->tx_timeout_factor = 1;
5507 			switch (adapter->link_speed) {
5508 			case SPEED_10:
5509 				adapter->tx_timeout_factor = 14;
5510 				break;
5511 			case SPEED_100:
5512 				/* maybe add some timeout factor ? */
5513 				break;
5514 			}
5515 
5516 			if (adapter->link_speed != SPEED_1000 ||
5517 			    !hw->phy.ops.read_reg)
5518 				goto no_wait;
5519 
5520 			/* wait for Remote receiver status OK */
5521 retry_read_status:
5522 			if (!igb_read_phy_reg(hw, PHY_1000T_STATUS,
5523 					      &phy_data)) {
5524 				if (!(phy_data & SR_1000T_REMOTE_RX_STATUS) &&
5525 				    retry_count) {
5526 					msleep(100);
5527 					retry_count--;
5528 					goto retry_read_status;
5529 				} else if (!retry_count) {
5530 					dev_err(&adapter->pdev->dev, "exceed max 2 second\n");
5531 				}
5532 			} else {
5533 				dev_err(&adapter->pdev->dev, "read 1000Base-T Status Reg\n");
5534 			}
5535 no_wait:
5536 			netif_carrier_on(netdev);
5537 
5538 			igb_ping_all_vfs(adapter);
5539 			igb_check_vf_rate_limit(adapter);
5540 
5541 			/* link state has changed, schedule phy info update */
5542 			if (!test_bit(__IGB_DOWN, &adapter->state))
5543 				mod_timer(&adapter->phy_info_timer,
5544 					  round_jiffies(jiffies + 2 * HZ));
5545 		}
5546 	} else {
5547 		if (netif_carrier_ok(netdev)) {
5548 			adapter->link_speed = 0;
5549 			adapter->link_duplex = 0;
5550 
5551 			/* check for thermal sensor event */
5552 			if (igb_thermal_sensor_event(hw,
5553 			    E1000_THSTAT_PWR_DOWN)) {
5554 				netdev_err(netdev, "The network adapter was stopped because it overheated\n");
5555 			}
5556 
5557 			/* Links status message must follow this format */
5558 			netdev_info(netdev, "igb: %s NIC Link is Down\n",
5559 			       netdev->name);
5560 			netif_carrier_off(netdev);
5561 
5562 			igb_ping_all_vfs(adapter);
5563 
5564 			/* link state has changed, schedule phy info update */
5565 			if (!test_bit(__IGB_DOWN, &adapter->state))
5566 				mod_timer(&adapter->phy_info_timer,
5567 					  round_jiffies(jiffies + 2 * HZ));
5568 
5569 			/* link is down, time to check for alternate media */
5570 			if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5571 				igb_check_swap_media(adapter);
5572 				if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5573 					schedule_work(&adapter->reset_task);
5574 					/* return immediately */
5575 					return;
5576 				}
5577 			}
5578 			pm_schedule_suspend(netdev->dev.parent,
5579 					    MSEC_PER_SEC * 5);
5580 
5581 		/* also check for alternate media here */
5582 		} else if (!netif_carrier_ok(netdev) &&
5583 			   (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
5584 			igb_check_swap_media(adapter);
5585 			if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5586 				schedule_work(&adapter->reset_task);
5587 				/* return immediately */
5588 				return;
5589 			}
5590 		}
5591 	}
5592 
5593 	spin_lock(&adapter->stats64_lock);
5594 	igb_update_stats(adapter);
5595 	spin_unlock(&adapter->stats64_lock);
5596 
5597 	for (i = 0; i < adapter->num_tx_queues; i++) {
5598 		struct igb_ring *tx_ring = adapter->tx_ring[i];
5599 		if (!netif_carrier_ok(netdev)) {
5600 			/* We've lost link, so the controller stops DMA,
5601 			 * but we've got queued Tx work that's never going
5602 			 * to get done, so reset controller to flush Tx.
5603 			 * (Do the reset outside of interrupt context).
5604 			 */
5605 			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
5606 				adapter->tx_timeout_count++;
5607 				schedule_work(&adapter->reset_task);
5608 				/* return immediately since reset is imminent */
5609 				return;
5610 			}
5611 		}
5612 
5613 		/* Force detection of hung controller every watchdog period */
5614 		set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
5615 	}
5616 
5617 	/* Cause software interrupt to ensure Rx ring is cleaned */
5618 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
5619 		u32 eics = 0;
5620 
5621 		for (i = 0; i < adapter->num_q_vectors; i++)
5622 			eics |= adapter->q_vector[i]->eims_value;
5623 		wr32(E1000_EICS, eics);
5624 	} else {
5625 		wr32(E1000_ICS, E1000_ICS_RXDMT0);
5626 	}
5627 
5628 	igb_spoof_check(adapter);
5629 	igb_ptp_rx_hang(adapter);
5630 	igb_ptp_tx_hang(adapter);
5631 
5632 	/* Check LVMMC register on i350/i354 only */
5633 	if ((adapter->hw.mac.type == e1000_i350) ||
5634 	    (adapter->hw.mac.type == e1000_i354))
5635 		igb_check_lvmmc(adapter);
5636 
5637 	/* Reset the timer */
5638 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
5639 		if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
5640 			mod_timer(&adapter->watchdog_timer,
5641 				  round_jiffies(jiffies +  HZ));
5642 		else
5643 			mod_timer(&adapter->watchdog_timer,
5644 				  round_jiffies(jiffies + 2 * HZ));
5645 	}
5646 }
5647 
5648 enum latency_range {
5649 	lowest_latency = 0,
5650 	low_latency = 1,
5651 	bulk_latency = 2,
5652 	latency_invalid = 255
5653 };
5654 
5655 /**
5656  *  igb_update_ring_itr - update the dynamic ITR value based on packet size
5657  *  @q_vector: pointer to q_vector
5658  *
5659  *  Stores a new ITR value based on strictly on packet size.  This
5660  *  algorithm is less sophisticated than that used in igb_update_itr,
5661  *  due to the difficulty of synchronizing statistics across multiple
5662  *  receive rings.  The divisors and thresholds used by this function
5663  *  were determined based on theoretical maximum wire speed and testing
5664  *  data, in order to minimize response time while increasing bulk
5665  *  throughput.
5666  *  This functionality is controlled by ethtool's coalescing settings.
5667  *  NOTE:  This function is called only when operating in a multiqueue
5668  *         receive environment.
5669  **/
5670 static void igb_update_ring_itr(struct igb_q_vector *q_vector)
5671 {
5672 	int new_val = q_vector->itr_val;
5673 	int avg_wire_size = 0;
5674 	struct igb_adapter *adapter = q_vector->adapter;
5675 	unsigned int packets;
5676 
5677 	/* For non-gigabit speeds, just fix the interrupt rate at 4000
5678 	 * ints/sec - ITR timer value of 120 ticks.
5679 	 */
5680 	if (adapter->link_speed != SPEED_1000) {
5681 		new_val = IGB_4K_ITR;
5682 		goto set_itr_val;
5683 	}
5684 
5685 	packets = q_vector->rx.total_packets;
5686 	if (packets)
5687 		avg_wire_size = q_vector->rx.total_bytes / packets;
5688 
5689 	packets = q_vector->tx.total_packets;
5690 	if (packets)
5691 		avg_wire_size = max_t(u32, avg_wire_size,
5692 				      q_vector->tx.total_bytes / packets);
5693 
5694 	/* if avg_wire_size isn't set no work was done */
5695 	if (!avg_wire_size)
5696 		goto clear_counts;
5697 
5698 	/* Add 24 bytes to size to account for CRC, preamble, and gap */
5699 	avg_wire_size += 24;
5700 
5701 	/* Don't starve jumbo frames */
5702 	avg_wire_size = min(avg_wire_size, 3000);
5703 
5704 	/* Give a little boost to mid-size frames */
5705 	if ((avg_wire_size > 300) && (avg_wire_size < 1200))
5706 		new_val = avg_wire_size / 3;
5707 	else
5708 		new_val = avg_wire_size / 2;
5709 
5710 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5711 	if (new_val < IGB_20K_ITR &&
5712 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5713 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5714 		new_val = IGB_20K_ITR;
5715 
5716 set_itr_val:
5717 	if (new_val != q_vector->itr_val) {
5718 		q_vector->itr_val = new_val;
5719 		q_vector->set_itr = 1;
5720 	}
5721 clear_counts:
5722 	q_vector->rx.total_bytes = 0;
5723 	q_vector->rx.total_packets = 0;
5724 	q_vector->tx.total_bytes = 0;
5725 	q_vector->tx.total_packets = 0;
5726 }
5727 
5728 /**
5729  *  igb_update_itr - update the dynamic ITR value based on statistics
5730  *  @q_vector: pointer to q_vector
5731  *  @ring_container: ring info to update the itr for
5732  *
5733  *  Stores a new ITR value based on packets and byte
5734  *  counts during the last interrupt.  The advantage of per interrupt
5735  *  computation is faster updates and more accurate ITR for the current
5736  *  traffic pattern.  Constants in this function were computed
5737  *  based on theoretical maximum wire speed and thresholds were set based
5738  *  on testing data as well as attempting to minimize response time
5739  *  while increasing bulk throughput.
5740  *  This functionality is controlled by ethtool's coalescing settings.
5741  *  NOTE:  These calculations are only valid when operating in a single-
5742  *         queue environment.
5743  **/
5744 static void igb_update_itr(struct igb_q_vector *q_vector,
5745 			   struct igb_ring_container *ring_container)
5746 {
5747 	unsigned int packets = ring_container->total_packets;
5748 	unsigned int bytes = ring_container->total_bytes;
5749 	u8 itrval = ring_container->itr;
5750 
5751 	/* no packets, exit with status unchanged */
5752 	if (packets == 0)
5753 		return;
5754 
5755 	switch (itrval) {
5756 	case lowest_latency:
5757 		/* handle TSO and jumbo frames */
5758 		if (bytes/packets > 8000)
5759 			itrval = bulk_latency;
5760 		else if ((packets < 5) && (bytes > 512))
5761 			itrval = low_latency;
5762 		break;
5763 	case low_latency:  /* 50 usec aka 20000 ints/s */
5764 		if (bytes > 10000) {
5765 			/* this if handles the TSO accounting */
5766 			if (bytes/packets > 8000)
5767 				itrval = bulk_latency;
5768 			else if ((packets < 10) || ((bytes/packets) > 1200))
5769 				itrval = bulk_latency;
5770 			else if ((packets > 35))
5771 				itrval = lowest_latency;
5772 		} else if (bytes/packets > 2000) {
5773 			itrval = bulk_latency;
5774 		} else if (packets <= 2 && bytes < 512) {
5775 			itrval = lowest_latency;
5776 		}
5777 		break;
5778 	case bulk_latency: /* 250 usec aka 4000 ints/s */
5779 		if (bytes > 25000) {
5780 			if (packets > 35)
5781 				itrval = low_latency;
5782 		} else if (bytes < 1500) {
5783 			itrval = low_latency;
5784 		}
5785 		break;
5786 	}
5787 
5788 	/* clear work counters since we have the values we need */
5789 	ring_container->total_bytes = 0;
5790 	ring_container->total_packets = 0;
5791 
5792 	/* write updated itr to ring container */
5793 	ring_container->itr = itrval;
5794 }
5795 
5796 static void igb_set_itr(struct igb_q_vector *q_vector)
5797 {
5798 	struct igb_adapter *adapter = q_vector->adapter;
5799 	u32 new_itr = q_vector->itr_val;
5800 	u8 current_itr = 0;
5801 
5802 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
5803 	if (adapter->link_speed != SPEED_1000) {
5804 		current_itr = 0;
5805 		new_itr = IGB_4K_ITR;
5806 		goto set_itr_now;
5807 	}
5808 
5809 	igb_update_itr(q_vector, &q_vector->tx);
5810 	igb_update_itr(q_vector, &q_vector->rx);
5811 
5812 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
5813 
5814 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5815 	if (current_itr == lowest_latency &&
5816 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5817 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5818 		current_itr = low_latency;
5819 
5820 	switch (current_itr) {
5821 	/* counts and packets in update_itr are dependent on these numbers */
5822 	case lowest_latency:
5823 		new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
5824 		break;
5825 	case low_latency:
5826 		new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
5827 		break;
5828 	case bulk_latency:
5829 		new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
5830 		break;
5831 	default:
5832 		break;
5833 	}
5834 
5835 set_itr_now:
5836 	if (new_itr != q_vector->itr_val) {
5837 		/* this attempts to bias the interrupt rate towards Bulk
5838 		 * by adding intermediate steps when interrupt rate is
5839 		 * increasing
5840 		 */
5841 		new_itr = new_itr > q_vector->itr_val ?
5842 			  max((new_itr * q_vector->itr_val) /
5843 			  (new_itr + (q_vector->itr_val >> 2)),
5844 			  new_itr) : new_itr;
5845 		/* Don't write the value here; it resets the adapter's
5846 		 * internal timer, and causes us to delay far longer than
5847 		 * we should between interrupts.  Instead, we write the ITR
5848 		 * value at the beginning of the next interrupt so the timing
5849 		 * ends up being correct.
5850 		 */
5851 		q_vector->itr_val = new_itr;
5852 		q_vector->set_itr = 1;
5853 	}
5854 }
5855 
5856 static void igb_tx_ctxtdesc(struct igb_ring *tx_ring,
5857 			    struct igb_tx_buffer *first,
5858 			    u32 vlan_macip_lens, u32 type_tucmd,
5859 			    u32 mss_l4len_idx)
5860 {
5861 	struct e1000_adv_tx_context_desc *context_desc;
5862 	u16 i = tx_ring->next_to_use;
5863 	struct timespec64 ts;
5864 
5865 	context_desc = IGB_TX_CTXTDESC(tx_ring, i);
5866 
5867 	i++;
5868 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
5869 
5870 	/* set bits to identify this as an advanced context descriptor */
5871 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
5872 
5873 	/* For 82575, context index must be unique per ring. */
5874 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5875 		mss_l4len_idx |= tx_ring->reg_idx << 4;
5876 
5877 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
5878 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
5879 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
5880 
5881 	/* We assume there is always a valid tx time available. Invalid times
5882 	 * should have been handled by the upper layers.
5883 	 */
5884 	if (tx_ring->launchtime_enable) {
5885 		ts = ktime_to_timespec64(first->skb->tstamp);
5886 		skb_txtime_consumed(first->skb);
5887 		context_desc->seqnum_seed = cpu_to_le32(ts.tv_nsec / 32);
5888 	} else {
5889 		context_desc->seqnum_seed = 0;
5890 	}
5891 }
5892 
5893 static int igb_tso(struct igb_ring *tx_ring,
5894 		   struct igb_tx_buffer *first,
5895 		   u8 *hdr_len)
5896 {
5897 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
5898 	struct sk_buff *skb = first->skb;
5899 	union {
5900 		struct iphdr *v4;
5901 		struct ipv6hdr *v6;
5902 		unsigned char *hdr;
5903 	} ip;
5904 	union {
5905 		struct tcphdr *tcp;
5906 		struct udphdr *udp;
5907 		unsigned char *hdr;
5908 	} l4;
5909 	u32 paylen, l4_offset;
5910 	int err;
5911 
5912 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5913 		return 0;
5914 
5915 	if (!skb_is_gso(skb))
5916 		return 0;
5917 
5918 	err = skb_cow_head(skb, 0);
5919 	if (err < 0)
5920 		return err;
5921 
5922 	ip.hdr = skb_network_header(skb);
5923 	l4.hdr = skb_checksum_start(skb);
5924 
5925 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5926 	type_tucmd = (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ?
5927 		      E1000_ADVTXD_TUCMD_L4T_UDP : E1000_ADVTXD_TUCMD_L4T_TCP;
5928 
5929 	/* initialize outer IP header fields */
5930 	if (ip.v4->version == 4) {
5931 		unsigned char *csum_start = skb_checksum_start(skb);
5932 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
5933 
5934 		/* IP header will have to cancel out any data that
5935 		 * is not a part of the outer IP header
5936 		 */
5937 		ip.v4->check = csum_fold(csum_partial(trans_start,
5938 						      csum_start - trans_start,
5939 						      0));
5940 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
5941 
5942 		ip.v4->tot_len = 0;
5943 		first->tx_flags |= IGB_TX_FLAGS_TSO |
5944 				   IGB_TX_FLAGS_CSUM |
5945 				   IGB_TX_FLAGS_IPV4;
5946 	} else {
5947 		ip.v6->payload_len = 0;
5948 		first->tx_flags |= IGB_TX_FLAGS_TSO |
5949 				   IGB_TX_FLAGS_CSUM;
5950 	}
5951 
5952 	/* determine offset of inner transport header */
5953 	l4_offset = l4.hdr - skb->data;
5954 
5955 	/* remove payload length from inner checksum */
5956 	paylen = skb->len - l4_offset;
5957 	if (type_tucmd & E1000_ADVTXD_TUCMD_L4T_TCP) {
5958 		/* compute length of segmentation header */
5959 		*hdr_len = (l4.tcp->doff * 4) + l4_offset;
5960 		csum_replace_by_diff(&l4.tcp->check,
5961 			(__force __wsum)htonl(paylen));
5962 	} else {
5963 		/* compute length of segmentation header */
5964 		*hdr_len = sizeof(*l4.udp) + l4_offset;
5965 		csum_replace_by_diff(&l4.udp->check,
5966 				     (__force __wsum)htonl(paylen));
5967 	}
5968 
5969 	/* update gso size and bytecount with header size */
5970 	first->gso_segs = skb_shinfo(skb)->gso_segs;
5971 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
5972 
5973 	/* MSS L4LEN IDX */
5974 	mss_l4len_idx = (*hdr_len - l4_offset) << E1000_ADVTXD_L4LEN_SHIFT;
5975 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
5976 
5977 	/* VLAN MACLEN IPLEN */
5978 	vlan_macip_lens = l4.hdr - ip.hdr;
5979 	vlan_macip_lens |= (ip.hdr - skb->data) << E1000_ADVTXD_MACLEN_SHIFT;
5980 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
5981 
5982 	igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens,
5983 			type_tucmd, mss_l4len_idx);
5984 
5985 	return 1;
5986 }
5987 
5988 static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
5989 {
5990 	struct sk_buff *skb = first->skb;
5991 	u32 vlan_macip_lens = 0;
5992 	u32 type_tucmd = 0;
5993 
5994 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
5995 csum_failed:
5996 		if (!(first->tx_flags & IGB_TX_FLAGS_VLAN) &&
5997 		    !tx_ring->launchtime_enable)
5998 			return;
5999 		goto no_csum;
6000 	}
6001 
6002 	switch (skb->csum_offset) {
6003 	case offsetof(struct tcphdr, check):
6004 		type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
6005 		fallthrough;
6006 	case offsetof(struct udphdr, check):
6007 		break;
6008 	case offsetof(struct sctphdr, checksum):
6009 		/* validate that this is actually an SCTP request */
6010 		if (skb_csum_is_sctp(skb)) {
6011 			type_tucmd = E1000_ADVTXD_TUCMD_L4T_SCTP;
6012 			break;
6013 		}
6014 		fallthrough;
6015 	default:
6016 		skb_checksum_help(skb);
6017 		goto csum_failed;
6018 	}
6019 
6020 	/* update TX checksum flag */
6021 	first->tx_flags |= IGB_TX_FLAGS_CSUM;
6022 	vlan_macip_lens = skb_checksum_start_offset(skb) -
6023 			  skb_network_offset(skb);
6024 no_csum:
6025 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
6026 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
6027 
6028 	igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens, type_tucmd, 0);
6029 }
6030 
6031 #define IGB_SET_FLAG(_input, _flag, _result) \
6032 	((_flag <= _result) ? \
6033 	 ((u32)(_input & _flag) * (_result / _flag)) : \
6034 	 ((u32)(_input & _flag) / (_flag / _result)))
6035 
6036 static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
6037 {
6038 	/* set type for advanced descriptor with frame checksum insertion */
6039 	u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
6040 		       E1000_ADVTXD_DCMD_DEXT |
6041 		       E1000_ADVTXD_DCMD_IFCS;
6042 
6043 	/* set HW vlan bit if vlan is present */
6044 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
6045 				 (E1000_ADVTXD_DCMD_VLE));
6046 
6047 	/* set segmentation bits for TSO */
6048 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
6049 				 (E1000_ADVTXD_DCMD_TSE));
6050 
6051 	/* set timestamp bit if present */
6052 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
6053 				 (E1000_ADVTXD_MAC_TSTAMP));
6054 
6055 	/* insert frame checksum */
6056 	cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
6057 
6058 	return cmd_type;
6059 }
6060 
6061 static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
6062 				 union e1000_adv_tx_desc *tx_desc,
6063 				 u32 tx_flags, unsigned int paylen)
6064 {
6065 	u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
6066 
6067 	/* 82575 requires a unique index per ring */
6068 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
6069 		olinfo_status |= tx_ring->reg_idx << 4;
6070 
6071 	/* insert L4 checksum */
6072 	olinfo_status |= IGB_SET_FLAG(tx_flags,
6073 				      IGB_TX_FLAGS_CSUM,
6074 				      (E1000_TXD_POPTS_TXSM << 8));
6075 
6076 	/* insert IPv4 checksum */
6077 	olinfo_status |= IGB_SET_FLAG(tx_flags,
6078 				      IGB_TX_FLAGS_IPV4,
6079 				      (E1000_TXD_POPTS_IXSM << 8));
6080 
6081 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6082 }
6083 
6084 static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
6085 {
6086 	struct net_device *netdev = tx_ring->netdev;
6087 
6088 	netif_stop_subqueue(netdev, tx_ring->queue_index);
6089 
6090 	/* Herbert's original patch had:
6091 	 *  smp_mb__after_netif_stop_queue();
6092 	 * but since that doesn't exist yet, just open code it.
6093 	 */
6094 	smp_mb();
6095 
6096 	/* We need to check again in a case another CPU has just
6097 	 * made room available.
6098 	 */
6099 	if (igb_desc_unused(tx_ring) < size)
6100 		return -EBUSY;
6101 
6102 	/* A reprieve! */
6103 	netif_wake_subqueue(netdev, tx_ring->queue_index);
6104 
6105 	u64_stats_update_begin(&tx_ring->tx_syncp2);
6106 	tx_ring->tx_stats.restart_queue2++;
6107 	u64_stats_update_end(&tx_ring->tx_syncp2);
6108 
6109 	return 0;
6110 }
6111 
6112 static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
6113 {
6114 	if (igb_desc_unused(tx_ring) >= size)
6115 		return 0;
6116 	return __igb_maybe_stop_tx(tx_ring, size);
6117 }
6118 
6119 static int igb_tx_map(struct igb_ring *tx_ring,
6120 		      struct igb_tx_buffer *first,
6121 		      const u8 hdr_len)
6122 {
6123 	struct sk_buff *skb = first->skb;
6124 	struct igb_tx_buffer *tx_buffer;
6125 	union e1000_adv_tx_desc *tx_desc;
6126 	skb_frag_t *frag;
6127 	dma_addr_t dma;
6128 	unsigned int data_len, size;
6129 	u32 tx_flags = first->tx_flags;
6130 	u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
6131 	u16 i = tx_ring->next_to_use;
6132 
6133 	tx_desc = IGB_TX_DESC(tx_ring, i);
6134 
6135 	igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
6136 
6137 	size = skb_headlen(skb);
6138 	data_len = skb->data_len;
6139 
6140 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6141 
6142 	tx_buffer = first;
6143 
6144 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
6145 		if (dma_mapping_error(tx_ring->dev, dma))
6146 			goto dma_error;
6147 
6148 		/* record length, and DMA address */
6149 		dma_unmap_len_set(tx_buffer, len, size);
6150 		dma_unmap_addr_set(tx_buffer, dma, dma);
6151 
6152 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6153 
6154 		while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
6155 			tx_desc->read.cmd_type_len =
6156 				cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
6157 
6158 			i++;
6159 			tx_desc++;
6160 			if (i == tx_ring->count) {
6161 				tx_desc = IGB_TX_DESC(tx_ring, 0);
6162 				i = 0;
6163 			}
6164 			tx_desc->read.olinfo_status = 0;
6165 
6166 			dma += IGB_MAX_DATA_PER_TXD;
6167 			size -= IGB_MAX_DATA_PER_TXD;
6168 
6169 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
6170 		}
6171 
6172 		if (likely(!data_len))
6173 			break;
6174 
6175 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
6176 
6177 		i++;
6178 		tx_desc++;
6179 		if (i == tx_ring->count) {
6180 			tx_desc = IGB_TX_DESC(tx_ring, 0);
6181 			i = 0;
6182 		}
6183 		tx_desc->read.olinfo_status = 0;
6184 
6185 		size = skb_frag_size(frag);
6186 		data_len -= size;
6187 
6188 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
6189 				       size, DMA_TO_DEVICE);
6190 
6191 		tx_buffer = &tx_ring->tx_buffer_info[i];
6192 	}
6193 
6194 	/* write last descriptor with RS and EOP bits */
6195 	cmd_type |= size | IGB_TXD_DCMD;
6196 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6197 
6198 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
6199 
6200 	/* set the timestamp */
6201 	first->time_stamp = jiffies;
6202 
6203 	skb_tx_timestamp(skb);
6204 
6205 	/* Force memory writes to complete before letting h/w know there
6206 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
6207 	 * memory model archs, such as IA-64).
6208 	 *
6209 	 * We also need this memory barrier to make certain all of the
6210 	 * status bits have been updated before next_to_watch is written.
6211 	 */
6212 	dma_wmb();
6213 
6214 	/* set next_to_watch value indicating a packet is present */
6215 	first->next_to_watch = tx_desc;
6216 
6217 	i++;
6218 	if (i == tx_ring->count)
6219 		i = 0;
6220 
6221 	tx_ring->next_to_use = i;
6222 
6223 	/* Make sure there is space in the ring for the next send. */
6224 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
6225 
6226 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
6227 		writel(i, tx_ring->tail);
6228 	}
6229 	return 0;
6230 
6231 dma_error:
6232 	dev_err(tx_ring->dev, "TX DMA map failed\n");
6233 	tx_buffer = &tx_ring->tx_buffer_info[i];
6234 
6235 	/* clear dma mappings for failed tx_buffer_info map */
6236 	while (tx_buffer != first) {
6237 		if (dma_unmap_len(tx_buffer, len))
6238 			dma_unmap_page(tx_ring->dev,
6239 				       dma_unmap_addr(tx_buffer, dma),
6240 				       dma_unmap_len(tx_buffer, len),
6241 				       DMA_TO_DEVICE);
6242 		dma_unmap_len_set(tx_buffer, len, 0);
6243 
6244 		if (i-- == 0)
6245 			i += tx_ring->count;
6246 		tx_buffer = &tx_ring->tx_buffer_info[i];
6247 	}
6248 
6249 	if (dma_unmap_len(tx_buffer, len))
6250 		dma_unmap_single(tx_ring->dev,
6251 				 dma_unmap_addr(tx_buffer, dma),
6252 				 dma_unmap_len(tx_buffer, len),
6253 				 DMA_TO_DEVICE);
6254 	dma_unmap_len_set(tx_buffer, len, 0);
6255 
6256 	dev_kfree_skb_any(tx_buffer->skb);
6257 	tx_buffer->skb = NULL;
6258 
6259 	tx_ring->next_to_use = i;
6260 
6261 	return -1;
6262 }
6263 
6264 int igb_xmit_xdp_ring(struct igb_adapter *adapter,
6265 		      struct igb_ring *tx_ring,
6266 		      struct xdp_frame *xdpf)
6267 {
6268 	struct skb_shared_info *sinfo = xdp_get_shared_info_from_frame(xdpf);
6269 	u8 nr_frags = unlikely(xdp_frame_has_frags(xdpf)) ? sinfo->nr_frags : 0;
6270 	u16 count, i, index = tx_ring->next_to_use;
6271 	struct igb_tx_buffer *tx_head = &tx_ring->tx_buffer_info[index];
6272 	struct igb_tx_buffer *tx_buffer = tx_head;
6273 	union e1000_adv_tx_desc *tx_desc = IGB_TX_DESC(tx_ring, index);
6274 	u32 len = xdpf->len, cmd_type, olinfo_status;
6275 	void *data = xdpf->data;
6276 
6277 	count = TXD_USE_COUNT(len);
6278 	for (i = 0; i < nr_frags; i++)
6279 		count += TXD_USE_COUNT(skb_frag_size(&sinfo->frags[i]));
6280 
6281 	if (igb_maybe_stop_tx(tx_ring, count + 3))
6282 		return IGB_XDP_CONSUMED;
6283 
6284 	i = 0;
6285 	/* record the location of the first descriptor for this packet */
6286 	tx_head->bytecount = xdp_get_frame_len(xdpf);
6287 	tx_head->type = IGB_TYPE_XDP;
6288 	tx_head->gso_segs = 1;
6289 	tx_head->xdpf = xdpf;
6290 
6291 	olinfo_status = tx_head->bytecount << E1000_ADVTXD_PAYLEN_SHIFT;
6292 	/* 82575 requires a unique index per ring */
6293 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
6294 		olinfo_status |= tx_ring->reg_idx << 4;
6295 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6296 
6297 	for (;;) {
6298 		dma_addr_t dma;
6299 
6300 		dma = dma_map_single(tx_ring->dev, data, len, DMA_TO_DEVICE);
6301 		if (dma_mapping_error(tx_ring->dev, dma))
6302 			goto unmap;
6303 
6304 		/* record length, and DMA address */
6305 		dma_unmap_len_set(tx_buffer, len, len);
6306 		dma_unmap_addr_set(tx_buffer, dma, dma);
6307 
6308 		/* put descriptor type bits */
6309 		cmd_type = E1000_ADVTXD_DTYP_DATA | E1000_ADVTXD_DCMD_DEXT |
6310 			   E1000_ADVTXD_DCMD_IFCS | len;
6311 
6312 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6313 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6314 
6315 		tx_buffer->protocol = 0;
6316 
6317 		if (++index == tx_ring->count)
6318 			index = 0;
6319 
6320 		if (i == nr_frags)
6321 			break;
6322 
6323 		tx_buffer = &tx_ring->tx_buffer_info[index];
6324 		tx_desc = IGB_TX_DESC(tx_ring, index);
6325 		tx_desc->read.olinfo_status = 0;
6326 
6327 		data = skb_frag_address(&sinfo->frags[i]);
6328 		len = skb_frag_size(&sinfo->frags[i]);
6329 		i++;
6330 	}
6331 	tx_desc->read.cmd_type_len |= cpu_to_le32(IGB_TXD_DCMD);
6332 
6333 	netdev_tx_sent_queue(txring_txq(tx_ring), tx_head->bytecount);
6334 	/* set the timestamp */
6335 	tx_head->time_stamp = jiffies;
6336 
6337 	/* Avoid any potential race with xdp_xmit and cleanup */
6338 	smp_wmb();
6339 
6340 	/* set next_to_watch value indicating a packet is present */
6341 	tx_head->next_to_watch = tx_desc;
6342 	tx_ring->next_to_use = index;
6343 
6344 	/* Make sure there is space in the ring for the next send. */
6345 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
6346 
6347 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more())
6348 		writel(index, tx_ring->tail);
6349 
6350 	return IGB_XDP_TX;
6351 
6352 unmap:
6353 	for (;;) {
6354 		tx_buffer = &tx_ring->tx_buffer_info[index];
6355 		if (dma_unmap_len(tx_buffer, len))
6356 			dma_unmap_page(tx_ring->dev,
6357 				       dma_unmap_addr(tx_buffer, dma),
6358 				       dma_unmap_len(tx_buffer, len),
6359 				       DMA_TO_DEVICE);
6360 		dma_unmap_len_set(tx_buffer, len, 0);
6361 		if (tx_buffer == tx_head)
6362 			break;
6363 
6364 		if (!index)
6365 			index += tx_ring->count;
6366 		index--;
6367 	}
6368 
6369 	return IGB_XDP_CONSUMED;
6370 }
6371 
6372 netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
6373 				struct igb_ring *tx_ring)
6374 {
6375 	struct igb_tx_buffer *first;
6376 	int tso;
6377 	u32 tx_flags = 0;
6378 	unsigned short f;
6379 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
6380 	__be16 protocol = vlan_get_protocol(skb);
6381 	u8 hdr_len = 0;
6382 
6383 	/* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
6384 	 *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
6385 	 *       + 2 desc gap to keep tail from touching head,
6386 	 *       + 1 desc for context descriptor,
6387 	 * otherwise try next time
6388 	 */
6389 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6390 		count += TXD_USE_COUNT(skb_frag_size(
6391 						&skb_shinfo(skb)->frags[f]));
6392 
6393 	if (igb_maybe_stop_tx(tx_ring, count + 3)) {
6394 		/* this is a hard error */
6395 		return NETDEV_TX_BUSY;
6396 	}
6397 
6398 	/* record the location of the first descriptor for this packet */
6399 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6400 	first->type = IGB_TYPE_SKB;
6401 	first->skb = skb;
6402 	first->bytecount = skb->len;
6403 	first->gso_segs = 1;
6404 
6405 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
6406 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6407 
6408 		if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
6409 		    !test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
6410 					   &adapter->state)) {
6411 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
6412 			tx_flags |= IGB_TX_FLAGS_TSTAMP;
6413 
6414 			adapter->ptp_tx_skb = skb_get(skb);
6415 			adapter->ptp_tx_start = jiffies;
6416 			if (adapter->hw.mac.type == e1000_82576)
6417 				schedule_work(&adapter->ptp_tx_work);
6418 		} else {
6419 			adapter->tx_hwtstamp_skipped++;
6420 		}
6421 	}
6422 
6423 	if (skb_vlan_tag_present(skb)) {
6424 		tx_flags |= IGB_TX_FLAGS_VLAN;
6425 		tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
6426 	}
6427 
6428 	/* record initial flags and protocol */
6429 	first->tx_flags = tx_flags;
6430 	first->protocol = protocol;
6431 
6432 	tso = igb_tso(tx_ring, first, &hdr_len);
6433 	if (tso < 0)
6434 		goto out_drop;
6435 	else if (!tso)
6436 		igb_tx_csum(tx_ring, first);
6437 
6438 	if (igb_tx_map(tx_ring, first, hdr_len))
6439 		goto cleanup_tx_tstamp;
6440 
6441 	return NETDEV_TX_OK;
6442 
6443 out_drop:
6444 	dev_kfree_skb_any(first->skb);
6445 	first->skb = NULL;
6446 cleanup_tx_tstamp:
6447 	if (unlikely(tx_flags & IGB_TX_FLAGS_TSTAMP)) {
6448 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6449 
6450 		dev_kfree_skb_any(adapter->ptp_tx_skb);
6451 		adapter->ptp_tx_skb = NULL;
6452 		if (adapter->hw.mac.type == e1000_82576)
6453 			cancel_work_sync(&adapter->ptp_tx_work);
6454 		clear_bit_unlock(__IGB_PTP_TX_IN_PROGRESS, &adapter->state);
6455 	}
6456 
6457 	return NETDEV_TX_OK;
6458 }
6459 
6460 static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
6461 						    struct sk_buff *skb)
6462 {
6463 	unsigned int r_idx = skb->queue_mapping;
6464 
6465 	if (r_idx >= adapter->num_tx_queues)
6466 		r_idx = r_idx % adapter->num_tx_queues;
6467 
6468 	return adapter->tx_ring[r_idx];
6469 }
6470 
6471 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
6472 				  struct net_device *netdev)
6473 {
6474 	struct igb_adapter *adapter = netdev_priv(netdev);
6475 
6476 	/* The minimum packet size with TCTL.PSP set is 17 so pad the skb
6477 	 * in order to meet this minimum size requirement.
6478 	 */
6479 	if (skb_put_padto(skb, 17))
6480 		return NETDEV_TX_OK;
6481 
6482 	return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
6483 }
6484 
6485 /**
6486  *  igb_tx_timeout - Respond to a Tx Hang
6487  *  @netdev: network interface device structure
6488  *  @txqueue: number of the Tx queue that hung (unused)
6489  **/
6490 static void igb_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
6491 {
6492 	struct igb_adapter *adapter = netdev_priv(netdev);
6493 	struct e1000_hw *hw = &adapter->hw;
6494 
6495 	/* Do the reset outside of interrupt context */
6496 	adapter->tx_timeout_count++;
6497 
6498 	if (hw->mac.type >= e1000_82580)
6499 		hw->dev_spec._82575.global_device_reset = true;
6500 
6501 	schedule_work(&adapter->reset_task);
6502 	wr32(E1000_EICS,
6503 	     (adapter->eims_enable_mask & ~adapter->eims_other));
6504 }
6505 
6506 static void igb_reset_task(struct work_struct *work)
6507 {
6508 	struct igb_adapter *adapter;
6509 	adapter = container_of(work, struct igb_adapter, reset_task);
6510 
6511 	rtnl_lock();
6512 	/* If we're already down or resetting, just bail */
6513 	if (test_bit(__IGB_DOWN, &adapter->state) ||
6514 	    test_bit(__IGB_RESETTING, &adapter->state)) {
6515 		rtnl_unlock();
6516 		return;
6517 	}
6518 
6519 	igb_dump(adapter);
6520 	netdev_err(adapter->netdev, "Reset adapter\n");
6521 	igb_reinit_locked(adapter);
6522 	rtnl_unlock();
6523 }
6524 
6525 /**
6526  *  igb_get_stats64 - Get System Network Statistics
6527  *  @netdev: network interface device structure
6528  *  @stats: rtnl_link_stats64 pointer
6529  **/
6530 static void igb_get_stats64(struct net_device *netdev,
6531 			    struct rtnl_link_stats64 *stats)
6532 {
6533 	struct igb_adapter *adapter = netdev_priv(netdev);
6534 
6535 	spin_lock(&adapter->stats64_lock);
6536 	igb_update_stats(adapter);
6537 	memcpy(stats, &adapter->stats64, sizeof(*stats));
6538 	spin_unlock(&adapter->stats64_lock);
6539 }
6540 
6541 /**
6542  *  igb_change_mtu - Change the Maximum Transfer Unit
6543  *  @netdev: network interface device structure
6544  *  @new_mtu: new value for maximum frame size
6545  *
6546  *  Returns 0 on success, negative on failure
6547  **/
6548 static int igb_change_mtu(struct net_device *netdev, int new_mtu)
6549 {
6550 	struct igb_adapter *adapter = netdev_priv(netdev);
6551 	int max_frame = new_mtu + IGB_ETH_PKT_HDR_PAD;
6552 
6553 	if (adapter->xdp_prog) {
6554 		int i;
6555 
6556 		for (i = 0; i < adapter->num_rx_queues; i++) {
6557 			struct igb_ring *ring = adapter->rx_ring[i];
6558 
6559 			if (max_frame > igb_rx_bufsz(ring)) {
6560 				netdev_warn(adapter->netdev,
6561 					    "Requested MTU size is not supported with XDP. Max frame size is %d\n",
6562 					    max_frame);
6563 				return -EINVAL;
6564 			}
6565 		}
6566 	}
6567 
6568 	/* adjust max frame to be at least the size of a standard frame */
6569 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
6570 		max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
6571 
6572 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
6573 		usleep_range(1000, 2000);
6574 
6575 	/* igb_down has a dependency on max_frame_size */
6576 	adapter->max_frame_size = max_frame;
6577 
6578 	if (netif_running(netdev))
6579 		igb_down(adapter);
6580 
6581 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6582 		   netdev->mtu, new_mtu);
6583 	netdev->mtu = new_mtu;
6584 
6585 	if (netif_running(netdev))
6586 		igb_up(adapter);
6587 	else
6588 		igb_reset(adapter);
6589 
6590 	clear_bit(__IGB_RESETTING, &adapter->state);
6591 
6592 	return 0;
6593 }
6594 
6595 /**
6596  *  igb_update_stats - Update the board statistics counters
6597  *  @adapter: board private structure
6598  **/
6599 void igb_update_stats(struct igb_adapter *adapter)
6600 {
6601 	struct rtnl_link_stats64 *net_stats = &adapter->stats64;
6602 	struct e1000_hw *hw = &adapter->hw;
6603 	struct pci_dev *pdev = adapter->pdev;
6604 	u32 reg, mpc;
6605 	int i;
6606 	u64 bytes, packets;
6607 	unsigned int start;
6608 	u64 _bytes, _packets;
6609 
6610 	/* Prevent stats update while adapter is being reset, or if the pci
6611 	 * connection is down.
6612 	 */
6613 	if (adapter->link_speed == 0)
6614 		return;
6615 	if (pci_channel_offline(pdev))
6616 		return;
6617 
6618 	bytes = 0;
6619 	packets = 0;
6620 
6621 	rcu_read_lock();
6622 	for (i = 0; i < adapter->num_rx_queues; i++) {
6623 		struct igb_ring *ring = adapter->rx_ring[i];
6624 		u32 rqdpc = rd32(E1000_RQDPC(i));
6625 		if (hw->mac.type >= e1000_i210)
6626 			wr32(E1000_RQDPC(i), 0);
6627 
6628 		if (rqdpc) {
6629 			ring->rx_stats.drops += rqdpc;
6630 			net_stats->rx_fifo_errors += rqdpc;
6631 		}
6632 
6633 		do {
6634 			start = u64_stats_fetch_begin(&ring->rx_syncp);
6635 			_bytes = ring->rx_stats.bytes;
6636 			_packets = ring->rx_stats.packets;
6637 		} while (u64_stats_fetch_retry(&ring->rx_syncp, start));
6638 		bytes += _bytes;
6639 		packets += _packets;
6640 	}
6641 
6642 	net_stats->rx_bytes = bytes;
6643 	net_stats->rx_packets = packets;
6644 
6645 	bytes = 0;
6646 	packets = 0;
6647 	for (i = 0; i < adapter->num_tx_queues; i++) {
6648 		struct igb_ring *ring = adapter->tx_ring[i];
6649 		do {
6650 			start = u64_stats_fetch_begin(&ring->tx_syncp);
6651 			_bytes = ring->tx_stats.bytes;
6652 			_packets = ring->tx_stats.packets;
6653 		} while (u64_stats_fetch_retry(&ring->tx_syncp, start));
6654 		bytes += _bytes;
6655 		packets += _packets;
6656 	}
6657 	net_stats->tx_bytes = bytes;
6658 	net_stats->tx_packets = packets;
6659 	rcu_read_unlock();
6660 
6661 	/* read stats registers */
6662 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
6663 	adapter->stats.gprc += rd32(E1000_GPRC);
6664 	adapter->stats.gorc += rd32(E1000_GORCL);
6665 	rd32(E1000_GORCH); /* clear GORCL */
6666 	adapter->stats.bprc += rd32(E1000_BPRC);
6667 	adapter->stats.mprc += rd32(E1000_MPRC);
6668 	adapter->stats.roc += rd32(E1000_ROC);
6669 
6670 	adapter->stats.prc64 += rd32(E1000_PRC64);
6671 	adapter->stats.prc127 += rd32(E1000_PRC127);
6672 	adapter->stats.prc255 += rd32(E1000_PRC255);
6673 	adapter->stats.prc511 += rd32(E1000_PRC511);
6674 	adapter->stats.prc1023 += rd32(E1000_PRC1023);
6675 	adapter->stats.prc1522 += rd32(E1000_PRC1522);
6676 	adapter->stats.symerrs += rd32(E1000_SYMERRS);
6677 	adapter->stats.sec += rd32(E1000_SEC);
6678 
6679 	mpc = rd32(E1000_MPC);
6680 	adapter->stats.mpc += mpc;
6681 	net_stats->rx_fifo_errors += mpc;
6682 	adapter->stats.scc += rd32(E1000_SCC);
6683 	adapter->stats.ecol += rd32(E1000_ECOL);
6684 	adapter->stats.mcc += rd32(E1000_MCC);
6685 	adapter->stats.latecol += rd32(E1000_LATECOL);
6686 	adapter->stats.dc += rd32(E1000_DC);
6687 	adapter->stats.rlec += rd32(E1000_RLEC);
6688 	adapter->stats.xonrxc += rd32(E1000_XONRXC);
6689 	adapter->stats.xontxc += rd32(E1000_XONTXC);
6690 	adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
6691 	adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
6692 	adapter->stats.fcruc += rd32(E1000_FCRUC);
6693 	adapter->stats.gptc += rd32(E1000_GPTC);
6694 	adapter->stats.gotc += rd32(E1000_GOTCL);
6695 	rd32(E1000_GOTCH); /* clear GOTCL */
6696 	adapter->stats.rnbc += rd32(E1000_RNBC);
6697 	adapter->stats.ruc += rd32(E1000_RUC);
6698 	adapter->stats.rfc += rd32(E1000_RFC);
6699 	adapter->stats.rjc += rd32(E1000_RJC);
6700 	adapter->stats.tor += rd32(E1000_TORH);
6701 	adapter->stats.tot += rd32(E1000_TOTH);
6702 	adapter->stats.tpr += rd32(E1000_TPR);
6703 
6704 	adapter->stats.ptc64 += rd32(E1000_PTC64);
6705 	adapter->stats.ptc127 += rd32(E1000_PTC127);
6706 	adapter->stats.ptc255 += rd32(E1000_PTC255);
6707 	adapter->stats.ptc511 += rd32(E1000_PTC511);
6708 	adapter->stats.ptc1023 += rd32(E1000_PTC1023);
6709 	adapter->stats.ptc1522 += rd32(E1000_PTC1522);
6710 
6711 	adapter->stats.mptc += rd32(E1000_MPTC);
6712 	adapter->stats.bptc += rd32(E1000_BPTC);
6713 
6714 	adapter->stats.tpt += rd32(E1000_TPT);
6715 	adapter->stats.colc += rd32(E1000_COLC);
6716 
6717 	adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
6718 	/* read internal phy specific stats */
6719 	reg = rd32(E1000_CTRL_EXT);
6720 	if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
6721 		adapter->stats.rxerrc += rd32(E1000_RXERRC);
6722 
6723 		/* this stat has invalid values on i210/i211 */
6724 		if ((hw->mac.type != e1000_i210) &&
6725 		    (hw->mac.type != e1000_i211))
6726 			adapter->stats.tncrs += rd32(E1000_TNCRS);
6727 	}
6728 
6729 	adapter->stats.tsctc += rd32(E1000_TSCTC);
6730 	adapter->stats.tsctfc += rd32(E1000_TSCTFC);
6731 
6732 	adapter->stats.iac += rd32(E1000_IAC);
6733 	adapter->stats.icrxoc += rd32(E1000_ICRXOC);
6734 	adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
6735 	adapter->stats.icrxatc += rd32(E1000_ICRXATC);
6736 	adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
6737 	adapter->stats.ictxatc += rd32(E1000_ICTXATC);
6738 	adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
6739 	adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
6740 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
6741 
6742 	/* Fill out the OS statistics structure */
6743 	net_stats->multicast = adapter->stats.mprc;
6744 	net_stats->collisions = adapter->stats.colc;
6745 
6746 	/* Rx Errors */
6747 
6748 	/* RLEC on some newer hardware can be incorrect so build
6749 	 * our own version based on RUC and ROC
6750 	 */
6751 	net_stats->rx_errors = adapter->stats.rxerrc +
6752 		adapter->stats.crcerrs + adapter->stats.algnerrc +
6753 		adapter->stats.ruc + adapter->stats.roc +
6754 		adapter->stats.cexterr;
6755 	net_stats->rx_length_errors = adapter->stats.ruc +
6756 				      adapter->stats.roc;
6757 	net_stats->rx_crc_errors = adapter->stats.crcerrs;
6758 	net_stats->rx_frame_errors = adapter->stats.algnerrc;
6759 	net_stats->rx_missed_errors = adapter->stats.mpc;
6760 
6761 	/* Tx Errors */
6762 	net_stats->tx_errors = adapter->stats.ecol +
6763 			       adapter->stats.latecol;
6764 	net_stats->tx_aborted_errors = adapter->stats.ecol;
6765 	net_stats->tx_window_errors = adapter->stats.latecol;
6766 	net_stats->tx_carrier_errors = adapter->stats.tncrs;
6767 
6768 	/* Tx Dropped needs to be maintained elsewhere */
6769 
6770 	/* Management Stats */
6771 	adapter->stats.mgptc += rd32(E1000_MGTPTC);
6772 	adapter->stats.mgprc += rd32(E1000_MGTPRC);
6773 	adapter->stats.mgpdc += rd32(E1000_MGTPDC);
6774 
6775 	/* OS2BMC Stats */
6776 	reg = rd32(E1000_MANC);
6777 	if (reg & E1000_MANC_EN_BMC2OS) {
6778 		adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
6779 		adapter->stats.o2bspc += rd32(E1000_O2BSPC);
6780 		adapter->stats.b2ospc += rd32(E1000_B2OSPC);
6781 		adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
6782 	}
6783 }
6784 
6785 static void igb_perout(struct igb_adapter *adapter, int tsintr_tt)
6786 {
6787 	int pin = ptp_find_pin(adapter->ptp_clock, PTP_PF_PEROUT, tsintr_tt);
6788 	struct e1000_hw *hw = &adapter->hw;
6789 	struct timespec64 ts;
6790 	u32 tsauxc;
6791 
6792 	if (pin < 0 || pin >= IGB_N_PEROUT)
6793 		return;
6794 
6795 	spin_lock(&adapter->tmreg_lock);
6796 
6797 	if (hw->mac.type == e1000_82580 ||
6798 	    hw->mac.type == e1000_i354 ||
6799 	    hw->mac.type == e1000_i350) {
6800 		s64 ns = timespec64_to_ns(&adapter->perout[pin].period);
6801 		u32 systiml, systimh, level_mask, level, rem;
6802 		u64 systim, now;
6803 
6804 		/* read systim registers in sequence */
6805 		rd32(E1000_SYSTIMR);
6806 		systiml = rd32(E1000_SYSTIML);
6807 		systimh = rd32(E1000_SYSTIMH);
6808 		systim = (((u64)(systimh & 0xFF)) << 32) | ((u64)systiml);
6809 		now = timecounter_cyc2time(&adapter->tc, systim);
6810 
6811 		if (pin < 2) {
6812 			level_mask = (tsintr_tt == 1) ? 0x80000 : 0x40000;
6813 			level = (rd32(E1000_CTRL) & level_mask) ? 1 : 0;
6814 		} else {
6815 			level_mask = (tsintr_tt == 1) ? 0x80 : 0x40;
6816 			level = (rd32(E1000_CTRL_EXT) & level_mask) ? 1 : 0;
6817 		}
6818 
6819 		div_u64_rem(now, ns, &rem);
6820 		systim = systim + (ns - rem);
6821 
6822 		/* synchronize pin level with rising/falling edges */
6823 		div_u64_rem(now, ns << 1, &rem);
6824 		if (rem < ns) {
6825 			/* first half of period */
6826 			if (level == 0) {
6827 				/* output is already low, skip this period */
6828 				systim += ns;
6829 				pr_notice("igb: periodic output on %s missed falling edge\n",
6830 					  adapter->sdp_config[pin].name);
6831 			}
6832 		} else {
6833 			/* second half of period */
6834 			if (level == 1) {
6835 				/* output is already high, skip this period */
6836 				systim += ns;
6837 				pr_notice("igb: periodic output on %s missed rising edge\n",
6838 					  adapter->sdp_config[pin].name);
6839 			}
6840 		}
6841 
6842 		/* for this chip family tv_sec is the upper part of the binary value,
6843 		 * so not seconds
6844 		 */
6845 		ts.tv_nsec = (u32)systim;
6846 		ts.tv_sec  = ((u32)(systim >> 32)) & 0xFF;
6847 	} else {
6848 		ts = timespec64_add(adapter->perout[pin].start,
6849 				    adapter->perout[pin].period);
6850 	}
6851 
6852 	/* u32 conversion of tv_sec is safe until y2106 */
6853 	wr32((tsintr_tt == 1) ? E1000_TRGTTIML1 : E1000_TRGTTIML0, ts.tv_nsec);
6854 	wr32((tsintr_tt == 1) ? E1000_TRGTTIMH1 : E1000_TRGTTIMH0, (u32)ts.tv_sec);
6855 	tsauxc = rd32(E1000_TSAUXC);
6856 	tsauxc |= TSAUXC_EN_TT0;
6857 	wr32(E1000_TSAUXC, tsauxc);
6858 	adapter->perout[pin].start = ts;
6859 
6860 	spin_unlock(&adapter->tmreg_lock);
6861 }
6862 
6863 static void igb_extts(struct igb_adapter *adapter, int tsintr_tt)
6864 {
6865 	int pin = ptp_find_pin(adapter->ptp_clock, PTP_PF_EXTTS, tsintr_tt);
6866 	int auxstmpl = (tsintr_tt == 1) ? E1000_AUXSTMPL1 : E1000_AUXSTMPL0;
6867 	int auxstmph = (tsintr_tt == 1) ? E1000_AUXSTMPH1 : E1000_AUXSTMPH0;
6868 	struct e1000_hw *hw = &adapter->hw;
6869 	struct ptp_clock_event event;
6870 	struct timespec64 ts;
6871 
6872 	if (pin < 0 || pin >= IGB_N_EXTTS)
6873 		return;
6874 
6875 	if (hw->mac.type == e1000_82580 ||
6876 	    hw->mac.type == e1000_i354 ||
6877 	    hw->mac.type == e1000_i350) {
6878 		s64 ns = rd32(auxstmpl);
6879 
6880 		ns += ((s64)(rd32(auxstmph) & 0xFF)) << 32;
6881 		ts = ns_to_timespec64(ns);
6882 	} else {
6883 		ts.tv_nsec = rd32(auxstmpl);
6884 		ts.tv_sec  = rd32(auxstmph);
6885 	}
6886 
6887 	event.type = PTP_CLOCK_EXTTS;
6888 	event.index = tsintr_tt;
6889 	event.timestamp = ts.tv_sec * 1000000000ULL + ts.tv_nsec;
6890 	ptp_clock_event(adapter->ptp_clock, &event);
6891 }
6892 
6893 static void igb_tsync_interrupt(struct igb_adapter *adapter)
6894 {
6895 	struct e1000_hw *hw = &adapter->hw;
6896 	u32 ack = 0, tsicr = rd32(E1000_TSICR);
6897 	struct ptp_clock_event event;
6898 
6899 	if (tsicr & TSINTR_SYS_WRAP) {
6900 		event.type = PTP_CLOCK_PPS;
6901 		if (adapter->ptp_caps.pps)
6902 			ptp_clock_event(adapter->ptp_clock, &event);
6903 		ack |= TSINTR_SYS_WRAP;
6904 	}
6905 
6906 	if (tsicr & E1000_TSICR_TXTS) {
6907 		/* retrieve hardware timestamp */
6908 		schedule_work(&adapter->ptp_tx_work);
6909 		ack |= E1000_TSICR_TXTS;
6910 	}
6911 
6912 	if (tsicr & TSINTR_TT0) {
6913 		igb_perout(adapter, 0);
6914 		ack |= TSINTR_TT0;
6915 	}
6916 
6917 	if (tsicr & TSINTR_TT1) {
6918 		igb_perout(adapter, 1);
6919 		ack |= TSINTR_TT1;
6920 	}
6921 
6922 	if (tsicr & TSINTR_AUTT0) {
6923 		igb_extts(adapter, 0);
6924 		ack |= TSINTR_AUTT0;
6925 	}
6926 
6927 	if (tsicr & TSINTR_AUTT1) {
6928 		igb_extts(adapter, 1);
6929 		ack |= TSINTR_AUTT1;
6930 	}
6931 
6932 	/* acknowledge the interrupts */
6933 	wr32(E1000_TSICR, ack);
6934 }
6935 
6936 static irqreturn_t igb_msix_other(int irq, void *data)
6937 {
6938 	struct igb_adapter *adapter = data;
6939 	struct e1000_hw *hw = &adapter->hw;
6940 	u32 icr = rd32(E1000_ICR);
6941 	/* reading ICR causes bit 31 of EICR to be cleared */
6942 
6943 	if (icr & E1000_ICR_DRSTA)
6944 		schedule_work(&adapter->reset_task);
6945 
6946 	if (icr & E1000_ICR_DOUTSYNC) {
6947 		/* HW is reporting DMA is out of sync */
6948 		adapter->stats.doosync++;
6949 		/* The DMA Out of Sync is also indication of a spoof event
6950 		 * in IOV mode. Check the Wrong VM Behavior register to
6951 		 * see if it is really a spoof event.
6952 		 */
6953 		igb_check_wvbr(adapter);
6954 	}
6955 
6956 	/* Check for a mailbox event */
6957 	if (icr & E1000_ICR_VMMB)
6958 		igb_msg_task(adapter);
6959 
6960 	if (icr & E1000_ICR_LSC) {
6961 		hw->mac.get_link_status = 1;
6962 		/* guard against interrupt when we're going down */
6963 		if (!test_bit(__IGB_DOWN, &adapter->state))
6964 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6965 	}
6966 
6967 	if (icr & E1000_ICR_TS)
6968 		igb_tsync_interrupt(adapter);
6969 
6970 	wr32(E1000_EIMS, adapter->eims_other);
6971 
6972 	return IRQ_HANDLED;
6973 }
6974 
6975 static void igb_write_itr(struct igb_q_vector *q_vector)
6976 {
6977 	struct igb_adapter *adapter = q_vector->adapter;
6978 	u32 itr_val = q_vector->itr_val & 0x7FFC;
6979 
6980 	if (!q_vector->set_itr)
6981 		return;
6982 
6983 	if (!itr_val)
6984 		itr_val = 0x4;
6985 
6986 	if (adapter->hw.mac.type == e1000_82575)
6987 		itr_val |= itr_val << 16;
6988 	else
6989 		itr_val |= E1000_EITR_CNT_IGNR;
6990 
6991 	writel(itr_val, q_vector->itr_register);
6992 	q_vector->set_itr = 0;
6993 }
6994 
6995 static irqreturn_t igb_msix_ring(int irq, void *data)
6996 {
6997 	struct igb_q_vector *q_vector = data;
6998 
6999 	/* Write the ITR value calculated from the previous interrupt. */
7000 	igb_write_itr(q_vector);
7001 
7002 	napi_schedule(&q_vector->napi);
7003 
7004 	return IRQ_HANDLED;
7005 }
7006 
7007 #ifdef CONFIG_IGB_DCA
7008 static void igb_update_tx_dca(struct igb_adapter *adapter,
7009 			      struct igb_ring *tx_ring,
7010 			      int cpu)
7011 {
7012 	struct e1000_hw *hw = &adapter->hw;
7013 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
7014 
7015 	if (hw->mac.type != e1000_82575)
7016 		txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
7017 
7018 	/* We can enable relaxed ordering for reads, but not writes when
7019 	 * DCA is enabled.  This is due to a known issue in some chipsets
7020 	 * which will cause the DCA tag to be cleared.
7021 	 */
7022 	txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
7023 		  E1000_DCA_TXCTRL_DATA_RRO_EN |
7024 		  E1000_DCA_TXCTRL_DESC_DCA_EN;
7025 
7026 	wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
7027 }
7028 
7029 static void igb_update_rx_dca(struct igb_adapter *adapter,
7030 			      struct igb_ring *rx_ring,
7031 			      int cpu)
7032 {
7033 	struct e1000_hw *hw = &adapter->hw;
7034 	u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
7035 
7036 	if (hw->mac.type != e1000_82575)
7037 		rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
7038 
7039 	/* We can enable relaxed ordering for reads, but not writes when
7040 	 * DCA is enabled.  This is due to a known issue in some chipsets
7041 	 * which will cause the DCA tag to be cleared.
7042 	 */
7043 	rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
7044 		  E1000_DCA_RXCTRL_DESC_DCA_EN;
7045 
7046 	wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
7047 }
7048 
7049 static void igb_update_dca(struct igb_q_vector *q_vector)
7050 {
7051 	struct igb_adapter *adapter = q_vector->adapter;
7052 	int cpu = get_cpu();
7053 
7054 	if (q_vector->cpu == cpu)
7055 		goto out_no_update;
7056 
7057 	if (q_vector->tx.ring)
7058 		igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
7059 
7060 	if (q_vector->rx.ring)
7061 		igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
7062 
7063 	q_vector->cpu = cpu;
7064 out_no_update:
7065 	put_cpu();
7066 }
7067 
7068 static void igb_setup_dca(struct igb_adapter *adapter)
7069 {
7070 	struct e1000_hw *hw = &adapter->hw;
7071 	int i;
7072 
7073 	if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
7074 		return;
7075 
7076 	/* Always use CB2 mode, difference is masked in the CB driver. */
7077 	wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
7078 
7079 	for (i = 0; i < adapter->num_q_vectors; i++) {
7080 		adapter->q_vector[i]->cpu = -1;
7081 		igb_update_dca(adapter->q_vector[i]);
7082 	}
7083 }
7084 
7085 static int __igb_notify_dca(struct device *dev, void *data)
7086 {
7087 	struct net_device *netdev = dev_get_drvdata(dev);
7088 	struct igb_adapter *adapter = netdev_priv(netdev);
7089 	struct pci_dev *pdev = adapter->pdev;
7090 	struct e1000_hw *hw = &adapter->hw;
7091 	unsigned long event = *(unsigned long *)data;
7092 
7093 	switch (event) {
7094 	case DCA_PROVIDER_ADD:
7095 		/* if already enabled, don't do it again */
7096 		if (adapter->flags & IGB_FLAG_DCA_ENABLED)
7097 			break;
7098 		if (dca_add_requester(dev) == 0) {
7099 			adapter->flags |= IGB_FLAG_DCA_ENABLED;
7100 			dev_info(&pdev->dev, "DCA enabled\n");
7101 			igb_setup_dca(adapter);
7102 			break;
7103 		}
7104 		fallthrough; /* since DCA is disabled. */
7105 	case DCA_PROVIDER_REMOVE:
7106 		if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
7107 			/* without this a class_device is left
7108 			 * hanging around in the sysfs model
7109 			 */
7110 			dca_remove_requester(dev);
7111 			dev_info(&pdev->dev, "DCA disabled\n");
7112 			adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
7113 			wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
7114 		}
7115 		break;
7116 	}
7117 
7118 	return 0;
7119 }
7120 
7121 static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
7122 			  void *p)
7123 {
7124 	int ret_val;
7125 
7126 	ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
7127 					 __igb_notify_dca);
7128 
7129 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
7130 }
7131 #endif /* CONFIG_IGB_DCA */
7132 
7133 #ifdef CONFIG_PCI_IOV
7134 static int igb_vf_configure(struct igb_adapter *adapter, int vf)
7135 {
7136 	unsigned char mac_addr[ETH_ALEN];
7137 
7138 	eth_zero_addr(mac_addr);
7139 	igb_set_vf_mac(adapter, vf, mac_addr);
7140 
7141 	/* By default spoof check is enabled for all VFs */
7142 	adapter->vf_data[vf].spoofchk_enabled = true;
7143 
7144 	/* By default VFs are not trusted */
7145 	adapter->vf_data[vf].trusted = false;
7146 
7147 	return 0;
7148 }
7149 
7150 #endif
7151 static void igb_ping_all_vfs(struct igb_adapter *adapter)
7152 {
7153 	struct e1000_hw *hw = &adapter->hw;
7154 	u32 ping;
7155 	int i;
7156 
7157 	for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
7158 		ping = E1000_PF_CONTROL_MSG;
7159 		if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
7160 			ping |= E1000_VT_MSGTYPE_CTS;
7161 		igb_write_mbx(hw, &ping, 1, i);
7162 	}
7163 }
7164 
7165 static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
7166 {
7167 	struct e1000_hw *hw = &adapter->hw;
7168 	u32 vmolr = rd32(E1000_VMOLR(vf));
7169 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7170 
7171 	vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
7172 			    IGB_VF_FLAG_MULTI_PROMISC);
7173 	vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
7174 
7175 	if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
7176 		vmolr |= E1000_VMOLR_MPME;
7177 		vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
7178 		*msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
7179 	} else {
7180 		/* if we have hashes and we are clearing a multicast promisc
7181 		 * flag we need to write the hashes to the MTA as this step
7182 		 * was previously skipped
7183 		 */
7184 		if (vf_data->num_vf_mc_hashes > 30) {
7185 			vmolr |= E1000_VMOLR_MPME;
7186 		} else if (vf_data->num_vf_mc_hashes) {
7187 			int j;
7188 
7189 			vmolr |= E1000_VMOLR_ROMPE;
7190 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
7191 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
7192 		}
7193 	}
7194 
7195 	wr32(E1000_VMOLR(vf), vmolr);
7196 
7197 	/* there are flags left unprocessed, likely not supported */
7198 	if (*msgbuf & E1000_VT_MSGINFO_MASK)
7199 		return -EINVAL;
7200 
7201 	return 0;
7202 }
7203 
7204 static int igb_set_vf_multicasts(struct igb_adapter *adapter,
7205 				  u32 *msgbuf, u32 vf)
7206 {
7207 	int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
7208 	u16 *hash_list = (u16 *)&msgbuf[1];
7209 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7210 	int i;
7211 
7212 	/* salt away the number of multicast addresses assigned
7213 	 * to this VF for later use to restore when the PF multi cast
7214 	 * list changes
7215 	 */
7216 	vf_data->num_vf_mc_hashes = n;
7217 
7218 	/* only up to 30 hash values supported */
7219 	if (n > 30)
7220 		n = 30;
7221 
7222 	/* store the hashes for later use */
7223 	for (i = 0; i < n; i++)
7224 		vf_data->vf_mc_hashes[i] = hash_list[i];
7225 
7226 	/* Flush and reset the mta with the new values */
7227 	igb_set_rx_mode(adapter->netdev);
7228 
7229 	return 0;
7230 }
7231 
7232 static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
7233 {
7234 	struct e1000_hw *hw = &adapter->hw;
7235 	struct vf_data_storage *vf_data;
7236 	int i, j;
7237 
7238 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
7239 		u32 vmolr = rd32(E1000_VMOLR(i));
7240 
7241 		vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
7242 
7243 		vf_data = &adapter->vf_data[i];
7244 
7245 		if ((vf_data->num_vf_mc_hashes > 30) ||
7246 		    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
7247 			vmolr |= E1000_VMOLR_MPME;
7248 		} else if (vf_data->num_vf_mc_hashes) {
7249 			vmolr |= E1000_VMOLR_ROMPE;
7250 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
7251 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
7252 		}
7253 		wr32(E1000_VMOLR(i), vmolr);
7254 	}
7255 }
7256 
7257 static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
7258 {
7259 	struct e1000_hw *hw = &adapter->hw;
7260 	u32 pool_mask, vlvf_mask, i;
7261 
7262 	/* create mask for VF and other pools */
7263 	pool_mask = E1000_VLVF_POOLSEL_MASK;
7264 	vlvf_mask = BIT(E1000_VLVF_POOLSEL_SHIFT + vf);
7265 
7266 	/* drop PF from pool bits */
7267 	pool_mask &= ~BIT(E1000_VLVF_POOLSEL_SHIFT +
7268 			     adapter->vfs_allocated_count);
7269 
7270 	/* Find the vlan filter for this id */
7271 	for (i = E1000_VLVF_ARRAY_SIZE; i--;) {
7272 		u32 vlvf = rd32(E1000_VLVF(i));
7273 		u32 vfta_mask, vid, vfta;
7274 
7275 		/* remove the vf from the pool */
7276 		if (!(vlvf & vlvf_mask))
7277 			continue;
7278 
7279 		/* clear out bit from VLVF */
7280 		vlvf ^= vlvf_mask;
7281 
7282 		/* if other pools are present, just remove ourselves */
7283 		if (vlvf & pool_mask)
7284 			goto update_vlvfb;
7285 
7286 		/* if PF is present, leave VFTA */
7287 		if (vlvf & E1000_VLVF_POOLSEL_MASK)
7288 			goto update_vlvf;
7289 
7290 		vid = vlvf & E1000_VLVF_VLANID_MASK;
7291 		vfta_mask = BIT(vid % 32);
7292 
7293 		/* clear bit from VFTA */
7294 		vfta = adapter->shadow_vfta[vid / 32];
7295 		if (vfta & vfta_mask)
7296 			hw->mac.ops.write_vfta(hw, vid / 32, vfta ^ vfta_mask);
7297 update_vlvf:
7298 		/* clear pool selection enable */
7299 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
7300 			vlvf &= E1000_VLVF_POOLSEL_MASK;
7301 		else
7302 			vlvf = 0;
7303 update_vlvfb:
7304 		/* clear pool bits */
7305 		wr32(E1000_VLVF(i), vlvf);
7306 	}
7307 }
7308 
7309 static int igb_find_vlvf_entry(struct e1000_hw *hw, u32 vlan)
7310 {
7311 	u32 vlvf;
7312 	int idx;
7313 
7314 	/* short cut the special case */
7315 	if (vlan == 0)
7316 		return 0;
7317 
7318 	/* Search for the VLAN id in the VLVF entries */
7319 	for (idx = E1000_VLVF_ARRAY_SIZE; --idx;) {
7320 		vlvf = rd32(E1000_VLVF(idx));
7321 		if ((vlvf & VLAN_VID_MASK) == vlan)
7322 			break;
7323 	}
7324 
7325 	return idx;
7326 }
7327 
7328 static void igb_update_pf_vlvf(struct igb_adapter *adapter, u32 vid)
7329 {
7330 	struct e1000_hw *hw = &adapter->hw;
7331 	u32 bits, pf_id;
7332 	int idx;
7333 
7334 	idx = igb_find_vlvf_entry(hw, vid);
7335 	if (!idx)
7336 		return;
7337 
7338 	/* See if any other pools are set for this VLAN filter
7339 	 * entry other than the PF.
7340 	 */
7341 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
7342 	bits = ~BIT(pf_id) & E1000_VLVF_POOLSEL_MASK;
7343 	bits &= rd32(E1000_VLVF(idx));
7344 
7345 	/* Disable the filter so this falls into the default pool. */
7346 	if (!bits) {
7347 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
7348 			wr32(E1000_VLVF(idx), BIT(pf_id));
7349 		else
7350 			wr32(E1000_VLVF(idx), 0);
7351 	}
7352 }
7353 
7354 static s32 igb_set_vf_vlan(struct igb_adapter *adapter, u32 vid,
7355 			   bool add, u32 vf)
7356 {
7357 	int pf_id = adapter->vfs_allocated_count;
7358 	struct e1000_hw *hw = &adapter->hw;
7359 	int err;
7360 
7361 	/* If VLAN overlaps with one the PF is currently monitoring make
7362 	 * sure that we are able to allocate a VLVF entry.  This may be
7363 	 * redundant but it guarantees PF will maintain visibility to
7364 	 * the VLAN.
7365 	 */
7366 	if (add && test_bit(vid, adapter->active_vlans)) {
7367 		err = igb_vfta_set(hw, vid, pf_id, true, false);
7368 		if (err)
7369 			return err;
7370 	}
7371 
7372 	err = igb_vfta_set(hw, vid, vf, add, false);
7373 
7374 	if (add && !err)
7375 		return err;
7376 
7377 	/* If we failed to add the VF VLAN or we are removing the VF VLAN
7378 	 * we may need to drop the PF pool bit in order to allow us to free
7379 	 * up the VLVF resources.
7380 	 */
7381 	if (test_bit(vid, adapter->active_vlans) ||
7382 	    (adapter->flags & IGB_FLAG_VLAN_PROMISC))
7383 		igb_update_pf_vlvf(adapter, vid);
7384 
7385 	return err;
7386 }
7387 
7388 static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
7389 {
7390 	struct e1000_hw *hw = &adapter->hw;
7391 
7392 	if (vid)
7393 		wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
7394 	else
7395 		wr32(E1000_VMVIR(vf), 0);
7396 }
7397 
7398 static int igb_enable_port_vlan(struct igb_adapter *adapter, int vf,
7399 				u16 vlan, u8 qos)
7400 {
7401 	int err;
7402 
7403 	err = igb_set_vf_vlan(adapter, vlan, true, vf);
7404 	if (err)
7405 		return err;
7406 
7407 	igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
7408 	igb_set_vmolr(adapter, vf, !vlan);
7409 
7410 	/* revoke access to previous VLAN */
7411 	if (vlan != adapter->vf_data[vf].pf_vlan)
7412 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
7413 				false, vf);
7414 
7415 	adapter->vf_data[vf].pf_vlan = vlan;
7416 	adapter->vf_data[vf].pf_qos = qos;
7417 	igb_set_vf_vlan_strip(adapter, vf, true);
7418 	dev_info(&adapter->pdev->dev,
7419 		 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
7420 	if (test_bit(__IGB_DOWN, &adapter->state)) {
7421 		dev_warn(&adapter->pdev->dev,
7422 			 "The VF VLAN has been set, but the PF device is not up.\n");
7423 		dev_warn(&adapter->pdev->dev,
7424 			 "Bring the PF device up before attempting to use the VF device.\n");
7425 	}
7426 
7427 	return err;
7428 }
7429 
7430 static int igb_disable_port_vlan(struct igb_adapter *adapter, int vf)
7431 {
7432 	/* Restore tagless access via VLAN 0 */
7433 	igb_set_vf_vlan(adapter, 0, true, vf);
7434 
7435 	igb_set_vmvir(adapter, 0, vf);
7436 	igb_set_vmolr(adapter, vf, true);
7437 
7438 	/* Remove any PF assigned VLAN */
7439 	if (adapter->vf_data[vf].pf_vlan)
7440 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
7441 				false, vf);
7442 
7443 	adapter->vf_data[vf].pf_vlan = 0;
7444 	adapter->vf_data[vf].pf_qos = 0;
7445 	igb_set_vf_vlan_strip(adapter, vf, false);
7446 
7447 	return 0;
7448 }
7449 
7450 static int igb_ndo_set_vf_vlan(struct net_device *netdev, int vf,
7451 			       u16 vlan, u8 qos, __be16 vlan_proto)
7452 {
7453 	struct igb_adapter *adapter = netdev_priv(netdev);
7454 
7455 	if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
7456 		return -EINVAL;
7457 
7458 	if (vlan_proto != htons(ETH_P_8021Q))
7459 		return -EPROTONOSUPPORT;
7460 
7461 	return (vlan || qos) ? igb_enable_port_vlan(adapter, vf, vlan, qos) :
7462 			       igb_disable_port_vlan(adapter, vf);
7463 }
7464 
7465 static int igb_set_vf_vlan_msg(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
7466 {
7467 	int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
7468 	int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
7469 	int ret;
7470 
7471 	if (adapter->vf_data[vf].pf_vlan)
7472 		return -1;
7473 
7474 	/* VLAN 0 is a special case, don't allow it to be removed */
7475 	if (!vid && !add)
7476 		return 0;
7477 
7478 	ret = igb_set_vf_vlan(adapter, vid, !!add, vf);
7479 	if (!ret)
7480 		igb_set_vf_vlan_strip(adapter, vf, !!vid);
7481 	return ret;
7482 }
7483 
7484 static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
7485 {
7486 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7487 
7488 	/* clear flags - except flag that indicates PF has set the MAC */
7489 	vf_data->flags &= IGB_VF_FLAG_PF_SET_MAC;
7490 	vf_data->last_nack = jiffies;
7491 
7492 	/* reset vlans for device */
7493 	igb_clear_vf_vfta(adapter, vf);
7494 	igb_set_vf_vlan(adapter, vf_data->pf_vlan, true, vf);
7495 	igb_set_vmvir(adapter, vf_data->pf_vlan |
7496 			       (vf_data->pf_qos << VLAN_PRIO_SHIFT), vf);
7497 	igb_set_vmolr(adapter, vf, !vf_data->pf_vlan);
7498 	igb_set_vf_vlan_strip(adapter, vf, !!(vf_data->pf_vlan));
7499 
7500 	/* reset multicast table array for vf */
7501 	adapter->vf_data[vf].num_vf_mc_hashes = 0;
7502 
7503 	/* Flush and reset the mta with the new values */
7504 	igb_set_rx_mode(adapter->netdev);
7505 }
7506 
7507 static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
7508 {
7509 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7510 
7511 	/* clear mac address as we were hotplug removed/added */
7512 	if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
7513 		eth_zero_addr(vf_mac);
7514 
7515 	/* process remaining reset events */
7516 	igb_vf_reset(adapter, vf);
7517 }
7518 
7519 static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
7520 {
7521 	struct e1000_hw *hw = &adapter->hw;
7522 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7523 	u32 reg, msgbuf[3] = {};
7524 	u8 *addr = (u8 *)(&msgbuf[1]);
7525 
7526 	/* process all the same items cleared in a function level reset */
7527 	igb_vf_reset(adapter, vf);
7528 
7529 	/* set vf mac address */
7530 	igb_set_vf_mac(adapter, vf, vf_mac);
7531 
7532 	/* enable transmit and receive for vf */
7533 	reg = rd32(E1000_VFTE);
7534 	wr32(E1000_VFTE, reg | BIT(vf));
7535 	reg = rd32(E1000_VFRE);
7536 	wr32(E1000_VFRE, reg | BIT(vf));
7537 
7538 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
7539 
7540 	/* reply to reset with ack and vf mac address */
7541 	if (!is_zero_ether_addr(vf_mac)) {
7542 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
7543 		memcpy(addr, vf_mac, ETH_ALEN);
7544 	} else {
7545 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
7546 	}
7547 	igb_write_mbx(hw, msgbuf, 3, vf);
7548 }
7549 
7550 static void igb_flush_mac_table(struct igb_adapter *adapter)
7551 {
7552 	struct e1000_hw *hw = &adapter->hw;
7553 	int i;
7554 
7555 	for (i = 0; i < hw->mac.rar_entry_count; i++) {
7556 		adapter->mac_table[i].state &= ~IGB_MAC_STATE_IN_USE;
7557 		eth_zero_addr(adapter->mac_table[i].addr);
7558 		adapter->mac_table[i].queue = 0;
7559 		igb_rar_set_index(adapter, i);
7560 	}
7561 }
7562 
7563 static int igb_available_rars(struct igb_adapter *adapter, u8 queue)
7564 {
7565 	struct e1000_hw *hw = &adapter->hw;
7566 	/* do not count rar entries reserved for VFs MAC addresses */
7567 	int rar_entries = hw->mac.rar_entry_count -
7568 			  adapter->vfs_allocated_count;
7569 	int i, count = 0;
7570 
7571 	for (i = 0; i < rar_entries; i++) {
7572 		/* do not count default entries */
7573 		if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT)
7574 			continue;
7575 
7576 		/* do not count "in use" entries for different queues */
7577 		if ((adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE) &&
7578 		    (adapter->mac_table[i].queue != queue))
7579 			continue;
7580 
7581 		count++;
7582 	}
7583 
7584 	return count;
7585 }
7586 
7587 /* Set default MAC address for the PF in the first RAR entry */
7588 static void igb_set_default_mac_filter(struct igb_adapter *adapter)
7589 {
7590 	struct igb_mac_addr *mac_table = &adapter->mac_table[0];
7591 
7592 	ether_addr_copy(mac_table->addr, adapter->hw.mac.addr);
7593 	mac_table->queue = adapter->vfs_allocated_count;
7594 	mac_table->state = IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7595 
7596 	igb_rar_set_index(adapter, 0);
7597 }
7598 
7599 /* If the filter to be added and an already existing filter express
7600  * the same address and address type, it should be possible to only
7601  * override the other configurations, for example the queue to steer
7602  * traffic.
7603  */
7604 static bool igb_mac_entry_can_be_used(const struct igb_mac_addr *entry,
7605 				      const u8 *addr, const u8 flags)
7606 {
7607 	if (!(entry->state & IGB_MAC_STATE_IN_USE))
7608 		return true;
7609 
7610 	if ((entry->state & IGB_MAC_STATE_SRC_ADDR) !=
7611 	    (flags & IGB_MAC_STATE_SRC_ADDR))
7612 		return false;
7613 
7614 	if (!ether_addr_equal(addr, entry->addr))
7615 		return false;
7616 
7617 	return true;
7618 }
7619 
7620 /* Add a MAC filter for 'addr' directing matching traffic to 'queue',
7621  * 'flags' is used to indicate what kind of match is made, match is by
7622  * default for the destination address, if matching by source address
7623  * is desired the flag IGB_MAC_STATE_SRC_ADDR can be used.
7624  */
7625 static int igb_add_mac_filter_flags(struct igb_adapter *adapter,
7626 				    const u8 *addr, const u8 queue,
7627 				    const u8 flags)
7628 {
7629 	struct e1000_hw *hw = &adapter->hw;
7630 	int rar_entries = hw->mac.rar_entry_count -
7631 			  adapter->vfs_allocated_count;
7632 	int i;
7633 
7634 	if (is_zero_ether_addr(addr))
7635 		return -EINVAL;
7636 
7637 	/* Search for the first empty entry in the MAC table.
7638 	 * Do not touch entries at the end of the table reserved for the VF MAC
7639 	 * addresses.
7640 	 */
7641 	for (i = 0; i < rar_entries; i++) {
7642 		if (!igb_mac_entry_can_be_used(&adapter->mac_table[i],
7643 					       addr, flags))
7644 			continue;
7645 
7646 		ether_addr_copy(adapter->mac_table[i].addr, addr);
7647 		adapter->mac_table[i].queue = queue;
7648 		adapter->mac_table[i].state |= IGB_MAC_STATE_IN_USE | flags;
7649 
7650 		igb_rar_set_index(adapter, i);
7651 		return i;
7652 	}
7653 
7654 	return -ENOSPC;
7655 }
7656 
7657 static int igb_add_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7658 			      const u8 queue)
7659 {
7660 	return igb_add_mac_filter_flags(adapter, addr, queue, 0);
7661 }
7662 
7663 /* Remove a MAC filter for 'addr' directing matching traffic to
7664  * 'queue', 'flags' is used to indicate what kind of match need to be
7665  * removed, match is by default for the destination address, if
7666  * matching by source address is to be removed the flag
7667  * IGB_MAC_STATE_SRC_ADDR can be used.
7668  */
7669 static int igb_del_mac_filter_flags(struct igb_adapter *adapter,
7670 				    const u8 *addr, const u8 queue,
7671 				    const u8 flags)
7672 {
7673 	struct e1000_hw *hw = &adapter->hw;
7674 	int rar_entries = hw->mac.rar_entry_count -
7675 			  adapter->vfs_allocated_count;
7676 	int i;
7677 
7678 	if (is_zero_ether_addr(addr))
7679 		return -EINVAL;
7680 
7681 	/* Search for matching entry in the MAC table based on given address
7682 	 * and queue. Do not touch entries at the end of the table reserved
7683 	 * for the VF MAC addresses.
7684 	 */
7685 	for (i = 0; i < rar_entries; i++) {
7686 		if (!(adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE))
7687 			continue;
7688 		if ((adapter->mac_table[i].state & flags) != flags)
7689 			continue;
7690 		if (adapter->mac_table[i].queue != queue)
7691 			continue;
7692 		if (!ether_addr_equal(adapter->mac_table[i].addr, addr))
7693 			continue;
7694 
7695 		/* When a filter for the default address is "deleted",
7696 		 * we return it to its initial configuration
7697 		 */
7698 		if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT) {
7699 			adapter->mac_table[i].state =
7700 				IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7701 			adapter->mac_table[i].queue =
7702 				adapter->vfs_allocated_count;
7703 		} else {
7704 			adapter->mac_table[i].state = 0;
7705 			adapter->mac_table[i].queue = 0;
7706 			eth_zero_addr(adapter->mac_table[i].addr);
7707 		}
7708 
7709 		igb_rar_set_index(adapter, i);
7710 		return 0;
7711 	}
7712 
7713 	return -ENOENT;
7714 }
7715 
7716 static int igb_del_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7717 			      const u8 queue)
7718 {
7719 	return igb_del_mac_filter_flags(adapter, addr, queue, 0);
7720 }
7721 
7722 int igb_add_mac_steering_filter(struct igb_adapter *adapter,
7723 				const u8 *addr, u8 queue, u8 flags)
7724 {
7725 	struct e1000_hw *hw = &adapter->hw;
7726 
7727 	/* In theory, this should be supported on 82575 as well, but
7728 	 * that part wasn't easily accessible during development.
7729 	 */
7730 	if (hw->mac.type != e1000_i210)
7731 		return -EOPNOTSUPP;
7732 
7733 	return igb_add_mac_filter_flags(adapter, addr, queue,
7734 					IGB_MAC_STATE_QUEUE_STEERING | flags);
7735 }
7736 
7737 int igb_del_mac_steering_filter(struct igb_adapter *adapter,
7738 				const u8 *addr, u8 queue, u8 flags)
7739 {
7740 	return igb_del_mac_filter_flags(adapter, addr, queue,
7741 					IGB_MAC_STATE_QUEUE_STEERING | flags);
7742 }
7743 
7744 static int igb_uc_sync(struct net_device *netdev, const unsigned char *addr)
7745 {
7746 	struct igb_adapter *adapter = netdev_priv(netdev);
7747 	int ret;
7748 
7749 	ret = igb_add_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7750 
7751 	return min_t(int, ret, 0);
7752 }
7753 
7754 static int igb_uc_unsync(struct net_device *netdev, const unsigned char *addr)
7755 {
7756 	struct igb_adapter *adapter = netdev_priv(netdev);
7757 
7758 	igb_del_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7759 
7760 	return 0;
7761 }
7762 
7763 static int igb_set_vf_mac_filter(struct igb_adapter *adapter, const int vf,
7764 				 const u32 info, const u8 *addr)
7765 {
7766 	struct pci_dev *pdev = adapter->pdev;
7767 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7768 	struct list_head *pos;
7769 	struct vf_mac_filter *entry = NULL;
7770 	int ret = 0;
7771 
7772 	if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7773 	    !vf_data->trusted) {
7774 		dev_warn(&pdev->dev,
7775 			 "VF %d requested MAC filter but is administratively denied\n",
7776 			  vf);
7777 		return -EINVAL;
7778 	}
7779 	if (!is_valid_ether_addr(addr)) {
7780 		dev_warn(&pdev->dev,
7781 			 "VF %d attempted to set invalid MAC filter\n",
7782 			  vf);
7783 		return -EINVAL;
7784 	}
7785 
7786 	switch (info) {
7787 	case E1000_VF_MAC_FILTER_CLR:
7788 		/* remove all unicast MAC filters related to the current VF */
7789 		list_for_each(pos, &adapter->vf_macs.l) {
7790 			entry = list_entry(pos, struct vf_mac_filter, l);
7791 			if (entry->vf == vf) {
7792 				entry->vf = -1;
7793 				entry->free = true;
7794 				igb_del_mac_filter(adapter, entry->vf_mac, vf);
7795 			}
7796 		}
7797 		break;
7798 	case E1000_VF_MAC_FILTER_ADD:
7799 		/* try to find empty slot in the list */
7800 		list_for_each(pos, &adapter->vf_macs.l) {
7801 			entry = list_entry(pos, struct vf_mac_filter, l);
7802 			if (entry->free)
7803 				break;
7804 		}
7805 
7806 		if (entry && entry->free) {
7807 			entry->free = false;
7808 			entry->vf = vf;
7809 			ether_addr_copy(entry->vf_mac, addr);
7810 
7811 			ret = igb_add_mac_filter(adapter, addr, vf);
7812 			ret = min_t(int, ret, 0);
7813 		} else {
7814 			ret = -ENOSPC;
7815 		}
7816 
7817 		if (ret == -ENOSPC)
7818 			dev_warn(&pdev->dev,
7819 				 "VF %d has requested MAC filter but there is no space for it\n",
7820 				 vf);
7821 		break;
7822 	default:
7823 		ret = -EINVAL;
7824 		break;
7825 	}
7826 
7827 	return ret;
7828 }
7829 
7830 static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
7831 {
7832 	struct pci_dev *pdev = adapter->pdev;
7833 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7834 	u32 info = msg[0] & E1000_VT_MSGINFO_MASK;
7835 
7836 	/* The VF MAC Address is stored in a packed array of bytes
7837 	 * starting at the second 32 bit word of the msg array
7838 	 */
7839 	unsigned char *addr = (unsigned char *)&msg[1];
7840 	int ret = 0;
7841 
7842 	if (!info) {
7843 		if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7844 		    !vf_data->trusted) {
7845 			dev_warn(&pdev->dev,
7846 				 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
7847 				 vf);
7848 			return -EINVAL;
7849 		}
7850 
7851 		if (!is_valid_ether_addr(addr)) {
7852 			dev_warn(&pdev->dev,
7853 				 "VF %d attempted to set invalid MAC\n",
7854 				 vf);
7855 			return -EINVAL;
7856 		}
7857 
7858 		ret = igb_set_vf_mac(adapter, vf, addr);
7859 	} else {
7860 		ret = igb_set_vf_mac_filter(adapter, vf, info, addr);
7861 	}
7862 
7863 	return ret;
7864 }
7865 
7866 static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
7867 {
7868 	struct e1000_hw *hw = &adapter->hw;
7869 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7870 	u32 msg = E1000_VT_MSGTYPE_NACK;
7871 
7872 	/* if device isn't clear to send it shouldn't be reading either */
7873 	if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
7874 	    time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
7875 		igb_write_mbx(hw, &msg, 1, vf);
7876 		vf_data->last_nack = jiffies;
7877 	}
7878 }
7879 
7880 static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
7881 {
7882 	struct pci_dev *pdev = adapter->pdev;
7883 	u32 msgbuf[E1000_VFMAILBOX_SIZE];
7884 	struct e1000_hw *hw = &adapter->hw;
7885 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7886 	s32 retval;
7887 
7888 	retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf, false);
7889 
7890 	if (retval) {
7891 		/* if receive failed revoke VF CTS stats and restart init */
7892 		dev_err(&pdev->dev, "Error receiving message from VF\n");
7893 		vf_data->flags &= ~IGB_VF_FLAG_CTS;
7894 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7895 			goto unlock;
7896 		goto out;
7897 	}
7898 
7899 	/* this is a message we already processed, do nothing */
7900 	if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
7901 		goto unlock;
7902 
7903 	/* until the vf completes a reset it should not be
7904 	 * allowed to start any configuration.
7905 	 */
7906 	if (msgbuf[0] == E1000_VF_RESET) {
7907 		/* unlocks mailbox */
7908 		igb_vf_reset_msg(adapter, vf);
7909 		return;
7910 	}
7911 
7912 	if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
7913 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7914 			goto unlock;
7915 		retval = -1;
7916 		goto out;
7917 	}
7918 
7919 	switch ((msgbuf[0] & 0xFFFF)) {
7920 	case E1000_VF_SET_MAC_ADDR:
7921 		retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
7922 		break;
7923 	case E1000_VF_SET_PROMISC:
7924 		retval = igb_set_vf_promisc(adapter, msgbuf, vf);
7925 		break;
7926 	case E1000_VF_SET_MULTICAST:
7927 		retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
7928 		break;
7929 	case E1000_VF_SET_LPE:
7930 		retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
7931 		break;
7932 	case E1000_VF_SET_VLAN:
7933 		retval = -1;
7934 		if (vf_data->pf_vlan)
7935 			dev_warn(&pdev->dev,
7936 				 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
7937 				 vf);
7938 		else
7939 			retval = igb_set_vf_vlan_msg(adapter, msgbuf, vf);
7940 		break;
7941 	default:
7942 		dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
7943 		retval = -1;
7944 		break;
7945 	}
7946 
7947 	msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
7948 out:
7949 	/* notify the VF of the results of what it sent us */
7950 	if (retval)
7951 		msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
7952 	else
7953 		msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
7954 
7955 	/* unlocks mailbox */
7956 	igb_write_mbx(hw, msgbuf, 1, vf);
7957 	return;
7958 
7959 unlock:
7960 	igb_unlock_mbx(hw, vf);
7961 }
7962 
7963 static void igb_msg_task(struct igb_adapter *adapter)
7964 {
7965 	struct e1000_hw *hw = &adapter->hw;
7966 	unsigned long flags;
7967 	u32 vf;
7968 
7969 	spin_lock_irqsave(&adapter->vfs_lock, flags);
7970 	for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
7971 		/* process any reset requests */
7972 		if (!igb_check_for_rst(hw, vf))
7973 			igb_vf_reset_event(adapter, vf);
7974 
7975 		/* process any messages pending */
7976 		if (!igb_check_for_msg(hw, vf))
7977 			igb_rcv_msg_from_vf(adapter, vf);
7978 
7979 		/* process any acks */
7980 		if (!igb_check_for_ack(hw, vf))
7981 			igb_rcv_ack_from_vf(adapter, vf);
7982 	}
7983 	spin_unlock_irqrestore(&adapter->vfs_lock, flags);
7984 }
7985 
7986 /**
7987  *  igb_set_uta - Set unicast filter table address
7988  *  @adapter: board private structure
7989  *  @set: boolean indicating if we are setting or clearing bits
7990  *
7991  *  The unicast table address is a register array of 32-bit registers.
7992  *  The table is meant to be used in a way similar to how the MTA is used
7993  *  however due to certain limitations in the hardware it is necessary to
7994  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
7995  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
7996  **/
7997 static void igb_set_uta(struct igb_adapter *adapter, bool set)
7998 {
7999 	struct e1000_hw *hw = &adapter->hw;
8000 	u32 uta = set ? ~0 : 0;
8001 	int i;
8002 
8003 	/* we only need to do this if VMDq is enabled */
8004 	if (!adapter->vfs_allocated_count)
8005 		return;
8006 
8007 	for (i = hw->mac.uta_reg_count; i--;)
8008 		array_wr32(E1000_UTA, i, uta);
8009 }
8010 
8011 /**
8012  *  igb_intr_msi - Interrupt Handler
8013  *  @irq: interrupt number
8014  *  @data: pointer to a network interface device structure
8015  **/
8016 static irqreturn_t igb_intr_msi(int irq, void *data)
8017 {
8018 	struct igb_adapter *adapter = data;
8019 	struct igb_q_vector *q_vector = adapter->q_vector[0];
8020 	struct e1000_hw *hw = &adapter->hw;
8021 	/* read ICR disables interrupts using IAM */
8022 	u32 icr = rd32(E1000_ICR);
8023 
8024 	igb_write_itr(q_vector);
8025 
8026 	if (icr & E1000_ICR_DRSTA)
8027 		schedule_work(&adapter->reset_task);
8028 
8029 	if (icr & E1000_ICR_DOUTSYNC) {
8030 		/* HW is reporting DMA is out of sync */
8031 		adapter->stats.doosync++;
8032 	}
8033 
8034 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
8035 		hw->mac.get_link_status = 1;
8036 		if (!test_bit(__IGB_DOWN, &adapter->state))
8037 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
8038 	}
8039 
8040 	if (icr & E1000_ICR_TS)
8041 		igb_tsync_interrupt(adapter);
8042 
8043 	napi_schedule(&q_vector->napi);
8044 
8045 	return IRQ_HANDLED;
8046 }
8047 
8048 /**
8049  *  igb_intr - Legacy Interrupt Handler
8050  *  @irq: interrupt number
8051  *  @data: pointer to a network interface device structure
8052  **/
8053 static irqreturn_t igb_intr(int irq, void *data)
8054 {
8055 	struct igb_adapter *adapter = data;
8056 	struct igb_q_vector *q_vector = adapter->q_vector[0];
8057 	struct e1000_hw *hw = &adapter->hw;
8058 	/* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
8059 	 * need for the IMC write
8060 	 */
8061 	u32 icr = rd32(E1000_ICR);
8062 
8063 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
8064 	 * not set, then the adapter didn't send an interrupt
8065 	 */
8066 	if (!(icr & E1000_ICR_INT_ASSERTED))
8067 		return IRQ_NONE;
8068 
8069 	igb_write_itr(q_vector);
8070 
8071 	if (icr & E1000_ICR_DRSTA)
8072 		schedule_work(&adapter->reset_task);
8073 
8074 	if (icr & E1000_ICR_DOUTSYNC) {
8075 		/* HW is reporting DMA is out of sync */
8076 		adapter->stats.doosync++;
8077 	}
8078 
8079 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
8080 		hw->mac.get_link_status = 1;
8081 		/* guard against interrupt when we're going down */
8082 		if (!test_bit(__IGB_DOWN, &adapter->state))
8083 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
8084 	}
8085 
8086 	if (icr & E1000_ICR_TS)
8087 		igb_tsync_interrupt(adapter);
8088 
8089 	napi_schedule(&q_vector->napi);
8090 
8091 	return IRQ_HANDLED;
8092 }
8093 
8094 static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
8095 {
8096 	struct igb_adapter *adapter = q_vector->adapter;
8097 	struct e1000_hw *hw = &adapter->hw;
8098 
8099 	if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
8100 	    (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
8101 		if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
8102 			igb_set_itr(q_vector);
8103 		else
8104 			igb_update_ring_itr(q_vector);
8105 	}
8106 
8107 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
8108 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
8109 			wr32(E1000_EIMS, q_vector->eims_value);
8110 		else
8111 			igb_irq_enable(adapter);
8112 	}
8113 }
8114 
8115 /**
8116  *  igb_poll - NAPI Rx polling callback
8117  *  @napi: napi polling structure
8118  *  @budget: count of how many packets we should handle
8119  **/
8120 static int igb_poll(struct napi_struct *napi, int budget)
8121 {
8122 	struct igb_q_vector *q_vector = container_of(napi,
8123 						     struct igb_q_vector,
8124 						     napi);
8125 	bool clean_complete = true;
8126 	int work_done = 0;
8127 
8128 #ifdef CONFIG_IGB_DCA
8129 	if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
8130 		igb_update_dca(q_vector);
8131 #endif
8132 	if (q_vector->tx.ring)
8133 		clean_complete = igb_clean_tx_irq(q_vector, budget);
8134 
8135 	if (q_vector->rx.ring) {
8136 		int cleaned = igb_clean_rx_irq(q_vector, budget);
8137 
8138 		work_done += cleaned;
8139 		if (cleaned >= budget)
8140 			clean_complete = false;
8141 	}
8142 
8143 	/* If all work not completed, return budget and keep polling */
8144 	if (!clean_complete)
8145 		return budget;
8146 
8147 	/* Exit the polling mode, but don't re-enable interrupts if stack might
8148 	 * poll us due to busy-polling
8149 	 */
8150 	if (likely(napi_complete_done(napi, work_done)))
8151 		igb_ring_irq_enable(q_vector);
8152 
8153 	return work_done;
8154 }
8155 
8156 /**
8157  *  igb_clean_tx_irq - Reclaim resources after transmit completes
8158  *  @q_vector: pointer to q_vector containing needed info
8159  *  @napi_budget: Used to determine if we are in netpoll
8160  *
8161  *  returns true if ring is completely cleaned
8162  **/
8163 static bool igb_clean_tx_irq(struct igb_q_vector *q_vector, int napi_budget)
8164 {
8165 	struct igb_adapter *adapter = q_vector->adapter;
8166 	struct igb_ring *tx_ring = q_vector->tx.ring;
8167 	struct igb_tx_buffer *tx_buffer;
8168 	union e1000_adv_tx_desc *tx_desc;
8169 	unsigned int total_bytes = 0, total_packets = 0;
8170 	unsigned int budget = q_vector->tx.work_limit;
8171 	unsigned int i = tx_ring->next_to_clean;
8172 
8173 	if (test_bit(__IGB_DOWN, &adapter->state))
8174 		return true;
8175 
8176 	tx_buffer = &tx_ring->tx_buffer_info[i];
8177 	tx_desc = IGB_TX_DESC(tx_ring, i);
8178 	i -= tx_ring->count;
8179 
8180 	do {
8181 		union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
8182 
8183 		/* if next_to_watch is not set then there is no work pending */
8184 		if (!eop_desc)
8185 			break;
8186 
8187 		/* prevent any other reads prior to eop_desc */
8188 		smp_rmb();
8189 
8190 		/* if DD is not set pending work has not been completed */
8191 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
8192 			break;
8193 
8194 		/* clear next_to_watch to prevent false hangs */
8195 		tx_buffer->next_to_watch = NULL;
8196 
8197 		/* update the statistics for this packet */
8198 		total_bytes += tx_buffer->bytecount;
8199 		total_packets += tx_buffer->gso_segs;
8200 
8201 		/* free the skb */
8202 		if (tx_buffer->type == IGB_TYPE_SKB)
8203 			napi_consume_skb(tx_buffer->skb, napi_budget);
8204 		else
8205 			xdp_return_frame(tx_buffer->xdpf);
8206 
8207 		/* unmap skb header data */
8208 		dma_unmap_single(tx_ring->dev,
8209 				 dma_unmap_addr(tx_buffer, dma),
8210 				 dma_unmap_len(tx_buffer, len),
8211 				 DMA_TO_DEVICE);
8212 
8213 		/* clear tx_buffer data */
8214 		dma_unmap_len_set(tx_buffer, len, 0);
8215 
8216 		/* clear last DMA location and unmap remaining buffers */
8217 		while (tx_desc != eop_desc) {
8218 			tx_buffer++;
8219 			tx_desc++;
8220 			i++;
8221 			if (unlikely(!i)) {
8222 				i -= tx_ring->count;
8223 				tx_buffer = tx_ring->tx_buffer_info;
8224 				tx_desc = IGB_TX_DESC(tx_ring, 0);
8225 			}
8226 
8227 			/* unmap any remaining paged data */
8228 			if (dma_unmap_len(tx_buffer, len)) {
8229 				dma_unmap_page(tx_ring->dev,
8230 					       dma_unmap_addr(tx_buffer, dma),
8231 					       dma_unmap_len(tx_buffer, len),
8232 					       DMA_TO_DEVICE);
8233 				dma_unmap_len_set(tx_buffer, len, 0);
8234 			}
8235 		}
8236 
8237 		/* move us one more past the eop_desc for start of next pkt */
8238 		tx_buffer++;
8239 		tx_desc++;
8240 		i++;
8241 		if (unlikely(!i)) {
8242 			i -= tx_ring->count;
8243 			tx_buffer = tx_ring->tx_buffer_info;
8244 			tx_desc = IGB_TX_DESC(tx_ring, 0);
8245 		}
8246 
8247 		/* issue prefetch for next Tx descriptor */
8248 		prefetch(tx_desc);
8249 
8250 		/* update budget accounting */
8251 		budget--;
8252 	} while (likely(budget));
8253 
8254 	netdev_tx_completed_queue(txring_txq(tx_ring),
8255 				  total_packets, total_bytes);
8256 	i += tx_ring->count;
8257 	tx_ring->next_to_clean = i;
8258 	u64_stats_update_begin(&tx_ring->tx_syncp);
8259 	tx_ring->tx_stats.bytes += total_bytes;
8260 	tx_ring->tx_stats.packets += total_packets;
8261 	u64_stats_update_end(&tx_ring->tx_syncp);
8262 	q_vector->tx.total_bytes += total_bytes;
8263 	q_vector->tx.total_packets += total_packets;
8264 
8265 	if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
8266 		struct e1000_hw *hw = &adapter->hw;
8267 
8268 		/* Detect a transmit hang in hardware, this serializes the
8269 		 * check with the clearing of time_stamp and movement of i
8270 		 */
8271 		clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
8272 		if (tx_buffer->next_to_watch &&
8273 		    time_after(jiffies, tx_buffer->time_stamp +
8274 			       (adapter->tx_timeout_factor * HZ)) &&
8275 		    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
8276 
8277 			/* detected Tx unit hang */
8278 			dev_err(tx_ring->dev,
8279 				"Detected Tx Unit Hang\n"
8280 				"  Tx Queue             <%d>\n"
8281 				"  TDH                  <%x>\n"
8282 				"  TDT                  <%x>\n"
8283 				"  next_to_use          <%x>\n"
8284 				"  next_to_clean        <%x>\n"
8285 				"buffer_info[next_to_clean]\n"
8286 				"  time_stamp           <%lx>\n"
8287 				"  next_to_watch        <%p>\n"
8288 				"  jiffies              <%lx>\n"
8289 				"  desc.status          <%x>\n",
8290 				tx_ring->queue_index,
8291 				rd32(E1000_TDH(tx_ring->reg_idx)),
8292 				readl(tx_ring->tail),
8293 				tx_ring->next_to_use,
8294 				tx_ring->next_to_clean,
8295 				tx_buffer->time_stamp,
8296 				tx_buffer->next_to_watch,
8297 				jiffies,
8298 				tx_buffer->next_to_watch->wb.status);
8299 			netif_stop_subqueue(tx_ring->netdev,
8300 					    tx_ring->queue_index);
8301 
8302 			/* we are about to reset, no point in enabling stuff */
8303 			return true;
8304 		}
8305 	}
8306 
8307 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
8308 	if (unlikely(total_packets &&
8309 	    netif_carrier_ok(tx_ring->netdev) &&
8310 	    igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
8311 		/* Make sure that anybody stopping the queue after this
8312 		 * sees the new next_to_clean.
8313 		 */
8314 		smp_mb();
8315 		if (__netif_subqueue_stopped(tx_ring->netdev,
8316 					     tx_ring->queue_index) &&
8317 		    !(test_bit(__IGB_DOWN, &adapter->state))) {
8318 			netif_wake_subqueue(tx_ring->netdev,
8319 					    tx_ring->queue_index);
8320 
8321 			u64_stats_update_begin(&tx_ring->tx_syncp);
8322 			tx_ring->tx_stats.restart_queue++;
8323 			u64_stats_update_end(&tx_ring->tx_syncp);
8324 		}
8325 	}
8326 
8327 	return !!budget;
8328 }
8329 
8330 /**
8331  *  igb_reuse_rx_page - page flip buffer and store it back on the ring
8332  *  @rx_ring: rx descriptor ring to store buffers on
8333  *  @old_buff: donor buffer to have page reused
8334  *
8335  *  Synchronizes page for reuse by the adapter
8336  **/
8337 static void igb_reuse_rx_page(struct igb_ring *rx_ring,
8338 			      struct igb_rx_buffer *old_buff)
8339 {
8340 	struct igb_rx_buffer *new_buff;
8341 	u16 nta = rx_ring->next_to_alloc;
8342 
8343 	new_buff = &rx_ring->rx_buffer_info[nta];
8344 
8345 	/* update, and store next to alloc */
8346 	nta++;
8347 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
8348 
8349 	/* Transfer page from old buffer to new buffer.
8350 	 * Move each member individually to avoid possible store
8351 	 * forwarding stalls.
8352 	 */
8353 	new_buff->dma		= old_buff->dma;
8354 	new_buff->page		= old_buff->page;
8355 	new_buff->page_offset	= old_buff->page_offset;
8356 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
8357 }
8358 
8359 static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer,
8360 				  int rx_buf_pgcnt)
8361 {
8362 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
8363 	struct page *page = rx_buffer->page;
8364 
8365 	/* avoid re-using remote and pfmemalloc pages */
8366 	if (!dev_page_is_reusable(page))
8367 		return false;
8368 
8369 #if (PAGE_SIZE < 8192)
8370 	/* if we are only owner of page we can reuse it */
8371 	if (unlikely((rx_buf_pgcnt - pagecnt_bias) > 1))
8372 		return false;
8373 #else
8374 #define IGB_LAST_OFFSET \
8375 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IGB_RXBUFFER_2048)
8376 
8377 	if (rx_buffer->page_offset > IGB_LAST_OFFSET)
8378 		return false;
8379 #endif
8380 
8381 	/* If we have drained the page fragment pool we need to update
8382 	 * the pagecnt_bias and page count so that we fully restock the
8383 	 * number of references the driver holds.
8384 	 */
8385 	if (unlikely(pagecnt_bias == 1)) {
8386 		page_ref_add(page, USHRT_MAX - 1);
8387 		rx_buffer->pagecnt_bias = USHRT_MAX;
8388 	}
8389 
8390 	return true;
8391 }
8392 
8393 /**
8394  *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
8395  *  @rx_ring: rx descriptor ring to transact packets on
8396  *  @rx_buffer: buffer containing page to add
8397  *  @skb: sk_buff to place the data into
8398  *  @size: size of buffer to be added
8399  *
8400  *  This function will add the data contained in rx_buffer->page to the skb.
8401  **/
8402 static void igb_add_rx_frag(struct igb_ring *rx_ring,
8403 			    struct igb_rx_buffer *rx_buffer,
8404 			    struct sk_buff *skb,
8405 			    unsigned int size)
8406 {
8407 #if (PAGE_SIZE < 8192)
8408 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8409 #else
8410 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
8411 				SKB_DATA_ALIGN(IGB_SKB_PAD + size) :
8412 				SKB_DATA_ALIGN(size);
8413 #endif
8414 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
8415 			rx_buffer->page_offset, size, truesize);
8416 #if (PAGE_SIZE < 8192)
8417 	rx_buffer->page_offset ^= truesize;
8418 #else
8419 	rx_buffer->page_offset += truesize;
8420 #endif
8421 }
8422 
8423 static struct sk_buff *igb_construct_skb(struct igb_ring *rx_ring,
8424 					 struct igb_rx_buffer *rx_buffer,
8425 					 struct xdp_buff *xdp,
8426 					 ktime_t timestamp)
8427 {
8428 #if (PAGE_SIZE < 8192)
8429 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8430 #else
8431 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
8432 					       xdp->data_hard_start);
8433 #endif
8434 	unsigned int size = xdp->data_end - xdp->data;
8435 	unsigned int headlen;
8436 	struct sk_buff *skb;
8437 
8438 	/* prefetch first cache line of first page */
8439 	net_prefetch(xdp->data);
8440 
8441 	/* allocate a skb to store the frags */
8442 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
8443 	if (unlikely(!skb))
8444 		return NULL;
8445 
8446 	if (timestamp)
8447 		skb_hwtstamps(skb)->hwtstamp = timestamp;
8448 
8449 	/* Determine available headroom for copy */
8450 	headlen = size;
8451 	if (headlen > IGB_RX_HDR_LEN)
8452 		headlen = eth_get_headlen(skb->dev, xdp->data, IGB_RX_HDR_LEN);
8453 
8454 	/* align pull length to size of long to optimize memcpy performance */
8455 	memcpy(__skb_put(skb, headlen), xdp->data, ALIGN(headlen, sizeof(long)));
8456 
8457 	/* update all of the pointers */
8458 	size -= headlen;
8459 	if (size) {
8460 		skb_add_rx_frag(skb, 0, rx_buffer->page,
8461 				(xdp->data + headlen) - page_address(rx_buffer->page),
8462 				size, truesize);
8463 #if (PAGE_SIZE < 8192)
8464 		rx_buffer->page_offset ^= truesize;
8465 #else
8466 		rx_buffer->page_offset += truesize;
8467 #endif
8468 	} else {
8469 		rx_buffer->pagecnt_bias++;
8470 	}
8471 
8472 	return skb;
8473 }
8474 
8475 static struct sk_buff *igb_build_skb(struct igb_ring *rx_ring,
8476 				     struct igb_rx_buffer *rx_buffer,
8477 				     struct xdp_buff *xdp,
8478 				     ktime_t timestamp)
8479 {
8480 #if (PAGE_SIZE < 8192)
8481 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8482 #else
8483 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
8484 				SKB_DATA_ALIGN(xdp->data_end -
8485 					       xdp->data_hard_start);
8486 #endif
8487 	unsigned int metasize = xdp->data - xdp->data_meta;
8488 	struct sk_buff *skb;
8489 
8490 	/* prefetch first cache line of first page */
8491 	net_prefetch(xdp->data_meta);
8492 
8493 	/* build an skb around the page buffer */
8494 	skb = napi_build_skb(xdp->data_hard_start, truesize);
8495 	if (unlikely(!skb))
8496 		return NULL;
8497 
8498 	/* update pointers within the skb to store the data */
8499 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
8500 	__skb_put(skb, xdp->data_end - xdp->data);
8501 
8502 	if (metasize)
8503 		skb_metadata_set(skb, metasize);
8504 
8505 	if (timestamp)
8506 		skb_hwtstamps(skb)->hwtstamp = timestamp;
8507 
8508 	/* update buffer offset */
8509 #if (PAGE_SIZE < 8192)
8510 	rx_buffer->page_offset ^= truesize;
8511 #else
8512 	rx_buffer->page_offset += truesize;
8513 #endif
8514 
8515 	return skb;
8516 }
8517 
8518 static struct sk_buff *igb_run_xdp(struct igb_adapter *adapter,
8519 				   struct igb_ring *rx_ring,
8520 				   struct xdp_buff *xdp)
8521 {
8522 	int err, result = IGB_XDP_PASS;
8523 	struct bpf_prog *xdp_prog;
8524 	u32 act;
8525 
8526 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
8527 
8528 	if (!xdp_prog)
8529 		goto xdp_out;
8530 
8531 	prefetchw(xdp->data_hard_start); /* xdp_frame write */
8532 
8533 	act = bpf_prog_run_xdp(xdp_prog, xdp);
8534 	switch (act) {
8535 	case XDP_PASS:
8536 		break;
8537 	case XDP_TX:
8538 		result = igb_xdp_xmit_back(adapter, xdp);
8539 		if (result == IGB_XDP_CONSUMED)
8540 			goto out_failure;
8541 		break;
8542 	case XDP_REDIRECT:
8543 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
8544 		if (err)
8545 			goto out_failure;
8546 		result = IGB_XDP_REDIR;
8547 		break;
8548 	default:
8549 		bpf_warn_invalid_xdp_action(adapter->netdev, xdp_prog, act);
8550 		fallthrough;
8551 	case XDP_ABORTED:
8552 out_failure:
8553 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
8554 		fallthrough;
8555 	case XDP_DROP:
8556 		result = IGB_XDP_CONSUMED;
8557 		break;
8558 	}
8559 xdp_out:
8560 	return ERR_PTR(-result);
8561 }
8562 
8563 static unsigned int igb_rx_frame_truesize(struct igb_ring *rx_ring,
8564 					  unsigned int size)
8565 {
8566 	unsigned int truesize;
8567 
8568 #if (PAGE_SIZE < 8192)
8569 	truesize = igb_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
8570 #else
8571 	truesize = ring_uses_build_skb(rx_ring) ?
8572 		SKB_DATA_ALIGN(IGB_SKB_PAD + size) +
8573 		SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
8574 		SKB_DATA_ALIGN(size);
8575 #endif
8576 	return truesize;
8577 }
8578 
8579 static void igb_rx_buffer_flip(struct igb_ring *rx_ring,
8580 			       struct igb_rx_buffer *rx_buffer,
8581 			       unsigned int size)
8582 {
8583 	unsigned int truesize = igb_rx_frame_truesize(rx_ring, size);
8584 #if (PAGE_SIZE < 8192)
8585 	rx_buffer->page_offset ^= truesize;
8586 #else
8587 	rx_buffer->page_offset += truesize;
8588 #endif
8589 }
8590 
8591 static inline void igb_rx_checksum(struct igb_ring *ring,
8592 				   union e1000_adv_rx_desc *rx_desc,
8593 				   struct sk_buff *skb)
8594 {
8595 	skb_checksum_none_assert(skb);
8596 
8597 	/* Ignore Checksum bit is set */
8598 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
8599 		return;
8600 
8601 	/* Rx checksum disabled via ethtool */
8602 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
8603 		return;
8604 
8605 	/* TCP/UDP checksum error bit is set */
8606 	if (igb_test_staterr(rx_desc,
8607 			     E1000_RXDEXT_STATERR_TCPE |
8608 			     E1000_RXDEXT_STATERR_IPE)) {
8609 		/* work around errata with sctp packets where the TCPE aka
8610 		 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
8611 		 * packets, (aka let the stack check the crc32c)
8612 		 */
8613 		if (!((skb->len == 60) &&
8614 		      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
8615 			u64_stats_update_begin(&ring->rx_syncp);
8616 			ring->rx_stats.csum_err++;
8617 			u64_stats_update_end(&ring->rx_syncp);
8618 		}
8619 		/* let the stack verify checksum errors */
8620 		return;
8621 	}
8622 	/* It must be a TCP or UDP packet with a valid checksum */
8623 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
8624 				      E1000_RXD_STAT_UDPCS))
8625 		skb->ip_summed = CHECKSUM_UNNECESSARY;
8626 
8627 	dev_dbg(ring->dev, "cksum success: bits %08X\n",
8628 		le32_to_cpu(rx_desc->wb.upper.status_error));
8629 }
8630 
8631 static inline void igb_rx_hash(struct igb_ring *ring,
8632 			       union e1000_adv_rx_desc *rx_desc,
8633 			       struct sk_buff *skb)
8634 {
8635 	if (ring->netdev->features & NETIF_F_RXHASH)
8636 		skb_set_hash(skb,
8637 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
8638 			     PKT_HASH_TYPE_L3);
8639 }
8640 
8641 /**
8642  *  igb_is_non_eop - process handling of non-EOP buffers
8643  *  @rx_ring: Rx ring being processed
8644  *  @rx_desc: Rx descriptor for current buffer
8645  *
8646  *  This function updates next to clean.  If the buffer is an EOP buffer
8647  *  this function exits returning false, otherwise it will place the
8648  *  sk_buff in the next buffer to be chained and return true indicating
8649  *  that this is in fact a non-EOP buffer.
8650  **/
8651 static bool igb_is_non_eop(struct igb_ring *rx_ring,
8652 			   union e1000_adv_rx_desc *rx_desc)
8653 {
8654 	u32 ntc = rx_ring->next_to_clean + 1;
8655 
8656 	/* fetch, update, and store next to clean */
8657 	ntc = (ntc < rx_ring->count) ? ntc : 0;
8658 	rx_ring->next_to_clean = ntc;
8659 
8660 	prefetch(IGB_RX_DESC(rx_ring, ntc));
8661 
8662 	if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
8663 		return false;
8664 
8665 	return true;
8666 }
8667 
8668 /**
8669  *  igb_cleanup_headers - Correct corrupted or empty headers
8670  *  @rx_ring: rx descriptor ring packet is being transacted on
8671  *  @rx_desc: pointer to the EOP Rx descriptor
8672  *  @skb: pointer to current skb being fixed
8673  *
8674  *  Address the case where we are pulling data in on pages only
8675  *  and as such no data is present in the skb header.
8676  *
8677  *  In addition if skb is not at least 60 bytes we need to pad it so that
8678  *  it is large enough to qualify as a valid Ethernet frame.
8679  *
8680  *  Returns true if an error was encountered and skb was freed.
8681  **/
8682 static bool igb_cleanup_headers(struct igb_ring *rx_ring,
8683 				union e1000_adv_rx_desc *rx_desc,
8684 				struct sk_buff *skb)
8685 {
8686 	/* XDP packets use error pointer so abort at this point */
8687 	if (IS_ERR(skb))
8688 		return true;
8689 
8690 	if (unlikely((igb_test_staterr(rx_desc,
8691 				       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
8692 		struct net_device *netdev = rx_ring->netdev;
8693 		if (!(netdev->features & NETIF_F_RXALL)) {
8694 			dev_kfree_skb_any(skb);
8695 			return true;
8696 		}
8697 	}
8698 
8699 	/* if eth_skb_pad returns an error the skb was freed */
8700 	if (eth_skb_pad(skb))
8701 		return true;
8702 
8703 	return false;
8704 }
8705 
8706 /**
8707  *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
8708  *  @rx_ring: rx descriptor ring packet is being transacted on
8709  *  @rx_desc: pointer to the EOP Rx descriptor
8710  *  @skb: pointer to current skb being populated
8711  *
8712  *  This function checks the ring, descriptor, and packet information in
8713  *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
8714  *  other fields within the skb.
8715  **/
8716 static void igb_process_skb_fields(struct igb_ring *rx_ring,
8717 				   union e1000_adv_rx_desc *rx_desc,
8718 				   struct sk_buff *skb)
8719 {
8720 	struct net_device *dev = rx_ring->netdev;
8721 
8722 	igb_rx_hash(rx_ring, rx_desc, skb);
8723 
8724 	igb_rx_checksum(rx_ring, rx_desc, skb);
8725 
8726 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
8727 	    !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
8728 		igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
8729 
8730 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
8731 	    igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
8732 		u16 vid;
8733 
8734 		if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
8735 		    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
8736 			vid = be16_to_cpu((__force __be16)rx_desc->wb.upper.vlan);
8737 		else
8738 			vid = le16_to_cpu(rx_desc->wb.upper.vlan);
8739 
8740 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
8741 	}
8742 
8743 	skb_record_rx_queue(skb, rx_ring->queue_index);
8744 
8745 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
8746 }
8747 
8748 static unsigned int igb_rx_offset(struct igb_ring *rx_ring)
8749 {
8750 	return ring_uses_build_skb(rx_ring) ? IGB_SKB_PAD : 0;
8751 }
8752 
8753 static struct igb_rx_buffer *igb_get_rx_buffer(struct igb_ring *rx_ring,
8754 					       const unsigned int size, int *rx_buf_pgcnt)
8755 {
8756 	struct igb_rx_buffer *rx_buffer;
8757 
8758 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
8759 	*rx_buf_pgcnt =
8760 #if (PAGE_SIZE < 8192)
8761 		page_count(rx_buffer->page);
8762 #else
8763 		0;
8764 #endif
8765 	prefetchw(rx_buffer->page);
8766 
8767 	/* we are reusing so sync this buffer for CPU use */
8768 	dma_sync_single_range_for_cpu(rx_ring->dev,
8769 				      rx_buffer->dma,
8770 				      rx_buffer->page_offset,
8771 				      size,
8772 				      DMA_FROM_DEVICE);
8773 
8774 	rx_buffer->pagecnt_bias--;
8775 
8776 	return rx_buffer;
8777 }
8778 
8779 static void igb_put_rx_buffer(struct igb_ring *rx_ring,
8780 			      struct igb_rx_buffer *rx_buffer, int rx_buf_pgcnt)
8781 {
8782 	if (igb_can_reuse_rx_page(rx_buffer, rx_buf_pgcnt)) {
8783 		/* hand second half of page back to the ring */
8784 		igb_reuse_rx_page(rx_ring, rx_buffer);
8785 	} else {
8786 		/* We are not reusing the buffer so unmap it and free
8787 		 * any references we are holding to it
8788 		 */
8789 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
8790 				     igb_rx_pg_size(rx_ring), DMA_FROM_DEVICE,
8791 				     IGB_RX_DMA_ATTR);
8792 		__page_frag_cache_drain(rx_buffer->page,
8793 					rx_buffer->pagecnt_bias);
8794 	}
8795 
8796 	/* clear contents of rx_buffer */
8797 	rx_buffer->page = NULL;
8798 }
8799 
8800 static int igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
8801 {
8802 	struct igb_adapter *adapter = q_vector->adapter;
8803 	struct igb_ring *rx_ring = q_vector->rx.ring;
8804 	struct sk_buff *skb = rx_ring->skb;
8805 	unsigned int total_bytes = 0, total_packets = 0;
8806 	u16 cleaned_count = igb_desc_unused(rx_ring);
8807 	unsigned int xdp_xmit = 0;
8808 	struct xdp_buff xdp;
8809 	u32 frame_sz = 0;
8810 	int rx_buf_pgcnt;
8811 
8812 	/* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
8813 #if (PAGE_SIZE < 8192)
8814 	frame_sz = igb_rx_frame_truesize(rx_ring, 0);
8815 #endif
8816 	xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq);
8817 
8818 	while (likely(total_packets < budget)) {
8819 		union e1000_adv_rx_desc *rx_desc;
8820 		struct igb_rx_buffer *rx_buffer;
8821 		ktime_t timestamp = 0;
8822 		int pkt_offset = 0;
8823 		unsigned int size;
8824 		void *pktbuf;
8825 
8826 		/* return some buffers to hardware, one at a time is too slow */
8827 		if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
8828 			igb_alloc_rx_buffers(rx_ring, cleaned_count);
8829 			cleaned_count = 0;
8830 		}
8831 
8832 		rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
8833 		size = le16_to_cpu(rx_desc->wb.upper.length);
8834 		if (!size)
8835 			break;
8836 
8837 		/* This memory barrier is needed to keep us from reading
8838 		 * any other fields out of the rx_desc until we know the
8839 		 * descriptor has been written back
8840 		 */
8841 		dma_rmb();
8842 
8843 		rx_buffer = igb_get_rx_buffer(rx_ring, size, &rx_buf_pgcnt);
8844 		pktbuf = page_address(rx_buffer->page) + rx_buffer->page_offset;
8845 
8846 		/* pull rx packet timestamp if available and valid */
8847 		if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
8848 			int ts_hdr_len;
8849 
8850 			ts_hdr_len = igb_ptp_rx_pktstamp(rx_ring->q_vector,
8851 							 pktbuf, &timestamp);
8852 
8853 			pkt_offset += ts_hdr_len;
8854 			size -= ts_hdr_len;
8855 		}
8856 
8857 		/* retrieve a buffer from the ring */
8858 		if (!skb) {
8859 			unsigned char *hard_start = pktbuf - igb_rx_offset(rx_ring);
8860 			unsigned int offset = pkt_offset + igb_rx_offset(rx_ring);
8861 
8862 			xdp_prepare_buff(&xdp, hard_start, offset, size, true);
8863 			xdp_buff_clear_frags_flag(&xdp);
8864 #if (PAGE_SIZE > 4096)
8865 			/* At larger PAGE_SIZE, frame_sz depend on len size */
8866 			xdp.frame_sz = igb_rx_frame_truesize(rx_ring, size);
8867 #endif
8868 			skb = igb_run_xdp(adapter, rx_ring, &xdp);
8869 		}
8870 
8871 		if (IS_ERR(skb)) {
8872 			unsigned int xdp_res = -PTR_ERR(skb);
8873 
8874 			if (xdp_res & (IGB_XDP_TX | IGB_XDP_REDIR)) {
8875 				xdp_xmit |= xdp_res;
8876 				igb_rx_buffer_flip(rx_ring, rx_buffer, size);
8877 			} else {
8878 				rx_buffer->pagecnt_bias++;
8879 			}
8880 			total_packets++;
8881 			total_bytes += size;
8882 		} else if (skb)
8883 			igb_add_rx_frag(rx_ring, rx_buffer, skb, size);
8884 		else if (ring_uses_build_skb(rx_ring))
8885 			skb = igb_build_skb(rx_ring, rx_buffer, &xdp,
8886 					    timestamp);
8887 		else
8888 			skb = igb_construct_skb(rx_ring, rx_buffer,
8889 						&xdp, timestamp);
8890 
8891 		/* exit if we failed to retrieve a buffer */
8892 		if (!skb) {
8893 			rx_ring->rx_stats.alloc_failed++;
8894 			rx_buffer->pagecnt_bias++;
8895 			break;
8896 		}
8897 
8898 		igb_put_rx_buffer(rx_ring, rx_buffer, rx_buf_pgcnt);
8899 		cleaned_count++;
8900 
8901 		/* fetch next buffer in frame if non-eop */
8902 		if (igb_is_non_eop(rx_ring, rx_desc))
8903 			continue;
8904 
8905 		/* verify the packet layout is correct */
8906 		if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
8907 			skb = NULL;
8908 			continue;
8909 		}
8910 
8911 		/* probably a little skewed due to removing CRC */
8912 		total_bytes += skb->len;
8913 
8914 		/* populate checksum, timestamp, VLAN, and protocol */
8915 		igb_process_skb_fields(rx_ring, rx_desc, skb);
8916 
8917 		napi_gro_receive(&q_vector->napi, skb);
8918 
8919 		/* reset skb pointer */
8920 		skb = NULL;
8921 
8922 		/* update budget accounting */
8923 		total_packets++;
8924 	}
8925 
8926 	/* place incomplete frames back on ring for completion */
8927 	rx_ring->skb = skb;
8928 
8929 	if (xdp_xmit & IGB_XDP_REDIR)
8930 		xdp_do_flush();
8931 
8932 	if (xdp_xmit & IGB_XDP_TX) {
8933 		struct igb_ring *tx_ring = igb_xdp_tx_queue_mapping(adapter);
8934 
8935 		igb_xdp_ring_update_tail(tx_ring);
8936 	}
8937 
8938 	u64_stats_update_begin(&rx_ring->rx_syncp);
8939 	rx_ring->rx_stats.packets += total_packets;
8940 	rx_ring->rx_stats.bytes += total_bytes;
8941 	u64_stats_update_end(&rx_ring->rx_syncp);
8942 	q_vector->rx.total_packets += total_packets;
8943 	q_vector->rx.total_bytes += total_bytes;
8944 
8945 	if (cleaned_count)
8946 		igb_alloc_rx_buffers(rx_ring, cleaned_count);
8947 
8948 	return total_packets;
8949 }
8950 
8951 static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
8952 				  struct igb_rx_buffer *bi)
8953 {
8954 	struct page *page = bi->page;
8955 	dma_addr_t dma;
8956 
8957 	/* since we are recycling buffers we should seldom need to alloc */
8958 	if (likely(page))
8959 		return true;
8960 
8961 	/* alloc new page for storage */
8962 	page = dev_alloc_pages(igb_rx_pg_order(rx_ring));
8963 	if (unlikely(!page)) {
8964 		rx_ring->rx_stats.alloc_failed++;
8965 		return false;
8966 	}
8967 
8968 	/* map page for use */
8969 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
8970 				 igb_rx_pg_size(rx_ring),
8971 				 DMA_FROM_DEVICE,
8972 				 IGB_RX_DMA_ATTR);
8973 
8974 	/* if mapping failed free memory back to system since
8975 	 * there isn't much point in holding memory we can't use
8976 	 */
8977 	if (dma_mapping_error(rx_ring->dev, dma)) {
8978 		__free_pages(page, igb_rx_pg_order(rx_ring));
8979 
8980 		rx_ring->rx_stats.alloc_failed++;
8981 		return false;
8982 	}
8983 
8984 	bi->dma = dma;
8985 	bi->page = page;
8986 	bi->page_offset = igb_rx_offset(rx_ring);
8987 	page_ref_add(page, USHRT_MAX - 1);
8988 	bi->pagecnt_bias = USHRT_MAX;
8989 
8990 	return true;
8991 }
8992 
8993 /**
8994  *  igb_alloc_rx_buffers - Replace used receive buffers
8995  *  @rx_ring: rx descriptor ring to allocate new receive buffers
8996  *  @cleaned_count: count of buffers to allocate
8997  **/
8998 void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
8999 {
9000 	union e1000_adv_rx_desc *rx_desc;
9001 	struct igb_rx_buffer *bi;
9002 	u16 i = rx_ring->next_to_use;
9003 	u16 bufsz;
9004 
9005 	/* nothing to do */
9006 	if (!cleaned_count)
9007 		return;
9008 
9009 	rx_desc = IGB_RX_DESC(rx_ring, i);
9010 	bi = &rx_ring->rx_buffer_info[i];
9011 	i -= rx_ring->count;
9012 
9013 	bufsz = igb_rx_bufsz(rx_ring);
9014 
9015 	do {
9016 		if (!igb_alloc_mapped_page(rx_ring, bi))
9017 			break;
9018 
9019 		/* sync the buffer for use by the device */
9020 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
9021 						 bi->page_offset, bufsz,
9022 						 DMA_FROM_DEVICE);
9023 
9024 		/* Refresh the desc even if buffer_addrs didn't change
9025 		 * because each write-back erases this info.
9026 		 */
9027 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
9028 
9029 		rx_desc++;
9030 		bi++;
9031 		i++;
9032 		if (unlikely(!i)) {
9033 			rx_desc = IGB_RX_DESC(rx_ring, 0);
9034 			bi = rx_ring->rx_buffer_info;
9035 			i -= rx_ring->count;
9036 		}
9037 
9038 		/* clear the length for the next_to_use descriptor */
9039 		rx_desc->wb.upper.length = 0;
9040 
9041 		cleaned_count--;
9042 	} while (cleaned_count);
9043 
9044 	i += rx_ring->count;
9045 
9046 	if (rx_ring->next_to_use != i) {
9047 		/* record the next descriptor to use */
9048 		rx_ring->next_to_use = i;
9049 
9050 		/* update next to alloc since we have filled the ring */
9051 		rx_ring->next_to_alloc = i;
9052 
9053 		/* Force memory writes to complete before letting h/w
9054 		 * know there are new descriptors to fetch.  (Only
9055 		 * applicable for weak-ordered memory model archs,
9056 		 * such as IA-64).
9057 		 */
9058 		dma_wmb();
9059 		writel(i, rx_ring->tail);
9060 	}
9061 }
9062 
9063 /**
9064  * igb_mii_ioctl -
9065  * @netdev: pointer to netdev struct
9066  * @ifr: interface structure
9067  * @cmd: ioctl command to execute
9068  **/
9069 static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
9070 {
9071 	struct igb_adapter *adapter = netdev_priv(netdev);
9072 	struct mii_ioctl_data *data = if_mii(ifr);
9073 
9074 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
9075 		return -EOPNOTSUPP;
9076 
9077 	switch (cmd) {
9078 	case SIOCGMIIPHY:
9079 		data->phy_id = adapter->hw.phy.addr;
9080 		break;
9081 	case SIOCGMIIREG:
9082 		if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
9083 				     &data->val_out))
9084 			return -EIO;
9085 		break;
9086 	case SIOCSMIIREG:
9087 	default:
9088 		return -EOPNOTSUPP;
9089 	}
9090 	return 0;
9091 }
9092 
9093 /**
9094  * igb_ioctl -
9095  * @netdev: pointer to netdev struct
9096  * @ifr: interface structure
9097  * @cmd: ioctl command to execute
9098  **/
9099 static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
9100 {
9101 	switch (cmd) {
9102 	case SIOCGMIIPHY:
9103 	case SIOCGMIIREG:
9104 	case SIOCSMIIREG:
9105 		return igb_mii_ioctl(netdev, ifr, cmd);
9106 	case SIOCGHWTSTAMP:
9107 		return igb_ptp_get_ts_config(netdev, ifr);
9108 	case SIOCSHWTSTAMP:
9109 		return igb_ptp_set_ts_config(netdev, ifr);
9110 	default:
9111 		return -EOPNOTSUPP;
9112 	}
9113 }
9114 
9115 void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
9116 {
9117 	struct igb_adapter *adapter = hw->back;
9118 
9119 	pci_read_config_word(adapter->pdev, reg, value);
9120 }
9121 
9122 void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
9123 {
9124 	struct igb_adapter *adapter = hw->back;
9125 
9126 	pci_write_config_word(adapter->pdev, reg, *value);
9127 }
9128 
9129 s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
9130 {
9131 	struct igb_adapter *adapter = hw->back;
9132 
9133 	if (pcie_capability_read_word(adapter->pdev, reg, value))
9134 		return -E1000_ERR_CONFIG;
9135 
9136 	return 0;
9137 }
9138 
9139 s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
9140 {
9141 	struct igb_adapter *adapter = hw->back;
9142 
9143 	if (pcie_capability_write_word(adapter->pdev, reg, *value))
9144 		return -E1000_ERR_CONFIG;
9145 
9146 	return 0;
9147 }
9148 
9149 static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
9150 {
9151 	struct igb_adapter *adapter = netdev_priv(netdev);
9152 	struct e1000_hw *hw = &adapter->hw;
9153 	u32 ctrl, rctl;
9154 	bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
9155 
9156 	if (enable) {
9157 		/* enable VLAN tag insert/strip */
9158 		ctrl = rd32(E1000_CTRL);
9159 		ctrl |= E1000_CTRL_VME;
9160 		wr32(E1000_CTRL, ctrl);
9161 
9162 		/* Disable CFI check */
9163 		rctl = rd32(E1000_RCTL);
9164 		rctl &= ~E1000_RCTL_CFIEN;
9165 		wr32(E1000_RCTL, rctl);
9166 	} else {
9167 		/* disable VLAN tag insert/strip */
9168 		ctrl = rd32(E1000_CTRL);
9169 		ctrl &= ~E1000_CTRL_VME;
9170 		wr32(E1000_CTRL, ctrl);
9171 	}
9172 
9173 	igb_set_vf_vlan_strip(adapter, adapter->vfs_allocated_count, enable);
9174 }
9175 
9176 static int igb_vlan_rx_add_vid(struct net_device *netdev,
9177 			       __be16 proto, u16 vid)
9178 {
9179 	struct igb_adapter *adapter = netdev_priv(netdev);
9180 	struct e1000_hw *hw = &adapter->hw;
9181 	int pf_id = adapter->vfs_allocated_count;
9182 
9183 	/* add the filter since PF can receive vlans w/o entry in vlvf */
9184 	if (!vid || !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
9185 		igb_vfta_set(hw, vid, pf_id, true, !!vid);
9186 
9187 	set_bit(vid, adapter->active_vlans);
9188 
9189 	return 0;
9190 }
9191 
9192 static int igb_vlan_rx_kill_vid(struct net_device *netdev,
9193 				__be16 proto, u16 vid)
9194 {
9195 	struct igb_adapter *adapter = netdev_priv(netdev);
9196 	int pf_id = adapter->vfs_allocated_count;
9197 	struct e1000_hw *hw = &adapter->hw;
9198 
9199 	/* remove VID from filter table */
9200 	if (vid && !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
9201 		igb_vfta_set(hw, vid, pf_id, false, true);
9202 
9203 	clear_bit(vid, adapter->active_vlans);
9204 
9205 	return 0;
9206 }
9207 
9208 static void igb_restore_vlan(struct igb_adapter *adapter)
9209 {
9210 	u16 vid = 1;
9211 
9212 	igb_vlan_mode(adapter->netdev, adapter->netdev->features);
9213 	igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
9214 
9215 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
9216 		igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
9217 }
9218 
9219 int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
9220 {
9221 	struct pci_dev *pdev = adapter->pdev;
9222 	struct e1000_mac_info *mac = &adapter->hw.mac;
9223 
9224 	mac->autoneg = 0;
9225 
9226 	/* Make sure dplx is at most 1 bit and lsb of speed is not set
9227 	 * for the switch() below to work
9228 	 */
9229 	if ((spd & 1) || (dplx & ~1))
9230 		goto err_inval;
9231 
9232 	/* Fiber NIC's only allow 1000 gbps Full duplex
9233 	 * and 100Mbps Full duplex for 100baseFx sfp
9234 	 */
9235 	if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
9236 		switch (spd + dplx) {
9237 		case SPEED_10 + DUPLEX_HALF:
9238 		case SPEED_10 + DUPLEX_FULL:
9239 		case SPEED_100 + DUPLEX_HALF:
9240 			goto err_inval;
9241 		default:
9242 			break;
9243 		}
9244 	}
9245 
9246 	switch (spd + dplx) {
9247 	case SPEED_10 + DUPLEX_HALF:
9248 		mac->forced_speed_duplex = ADVERTISE_10_HALF;
9249 		break;
9250 	case SPEED_10 + DUPLEX_FULL:
9251 		mac->forced_speed_duplex = ADVERTISE_10_FULL;
9252 		break;
9253 	case SPEED_100 + DUPLEX_HALF:
9254 		mac->forced_speed_duplex = ADVERTISE_100_HALF;
9255 		break;
9256 	case SPEED_100 + DUPLEX_FULL:
9257 		mac->forced_speed_duplex = ADVERTISE_100_FULL;
9258 		break;
9259 	case SPEED_1000 + DUPLEX_FULL:
9260 		mac->autoneg = 1;
9261 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
9262 		break;
9263 	case SPEED_1000 + DUPLEX_HALF: /* not supported */
9264 	default:
9265 		goto err_inval;
9266 	}
9267 
9268 	/* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
9269 	adapter->hw.phy.mdix = AUTO_ALL_MODES;
9270 
9271 	return 0;
9272 
9273 err_inval:
9274 	dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
9275 	return -EINVAL;
9276 }
9277 
9278 static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
9279 			  bool runtime)
9280 {
9281 	struct net_device *netdev = pci_get_drvdata(pdev);
9282 	struct igb_adapter *adapter = netdev_priv(netdev);
9283 	struct e1000_hw *hw = &adapter->hw;
9284 	u32 ctrl, rctl, status;
9285 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
9286 	bool wake;
9287 
9288 	rtnl_lock();
9289 	netif_device_detach(netdev);
9290 
9291 	if (netif_running(netdev))
9292 		__igb_close(netdev, true);
9293 
9294 	igb_ptp_suspend(adapter);
9295 
9296 	igb_clear_interrupt_scheme(adapter);
9297 	rtnl_unlock();
9298 
9299 	status = rd32(E1000_STATUS);
9300 	if (status & E1000_STATUS_LU)
9301 		wufc &= ~E1000_WUFC_LNKC;
9302 
9303 	if (wufc) {
9304 		igb_setup_rctl(adapter);
9305 		igb_set_rx_mode(netdev);
9306 
9307 		/* turn on all-multi mode if wake on multicast is enabled */
9308 		if (wufc & E1000_WUFC_MC) {
9309 			rctl = rd32(E1000_RCTL);
9310 			rctl |= E1000_RCTL_MPE;
9311 			wr32(E1000_RCTL, rctl);
9312 		}
9313 
9314 		ctrl = rd32(E1000_CTRL);
9315 		ctrl |= E1000_CTRL_ADVD3WUC;
9316 		wr32(E1000_CTRL, ctrl);
9317 
9318 		/* Allow time for pending master requests to run */
9319 		igb_disable_pcie_master(hw);
9320 
9321 		wr32(E1000_WUC, E1000_WUC_PME_EN);
9322 		wr32(E1000_WUFC, wufc);
9323 	} else {
9324 		wr32(E1000_WUC, 0);
9325 		wr32(E1000_WUFC, 0);
9326 	}
9327 
9328 	wake = wufc || adapter->en_mng_pt;
9329 	if (!wake)
9330 		igb_power_down_link(adapter);
9331 	else
9332 		igb_power_up_link(adapter);
9333 
9334 	if (enable_wake)
9335 		*enable_wake = wake;
9336 
9337 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
9338 	 * would have already happened in close and is redundant.
9339 	 */
9340 	igb_release_hw_control(adapter);
9341 
9342 	pci_disable_device(pdev);
9343 
9344 	return 0;
9345 }
9346 
9347 static void igb_deliver_wake_packet(struct net_device *netdev)
9348 {
9349 	struct igb_adapter *adapter = netdev_priv(netdev);
9350 	struct e1000_hw *hw = &adapter->hw;
9351 	struct sk_buff *skb;
9352 	u32 wupl;
9353 
9354 	wupl = rd32(E1000_WUPL) & E1000_WUPL_MASK;
9355 
9356 	/* WUPM stores only the first 128 bytes of the wake packet.
9357 	 * Read the packet only if we have the whole thing.
9358 	 */
9359 	if ((wupl == 0) || (wupl > E1000_WUPM_BYTES))
9360 		return;
9361 
9362 	skb = netdev_alloc_skb_ip_align(netdev, E1000_WUPM_BYTES);
9363 	if (!skb)
9364 		return;
9365 
9366 	skb_put(skb, wupl);
9367 
9368 	/* Ensure reads are 32-bit aligned */
9369 	wupl = roundup(wupl, 4);
9370 
9371 	memcpy_fromio(skb->data, hw->hw_addr + E1000_WUPM_REG(0), wupl);
9372 
9373 	skb->protocol = eth_type_trans(skb, netdev);
9374 	netif_rx(skb);
9375 }
9376 
9377 static int __maybe_unused igb_suspend(struct device *dev)
9378 {
9379 	return __igb_shutdown(to_pci_dev(dev), NULL, 0);
9380 }
9381 
9382 static int __maybe_unused __igb_resume(struct device *dev, bool rpm)
9383 {
9384 	struct pci_dev *pdev = to_pci_dev(dev);
9385 	struct net_device *netdev = pci_get_drvdata(pdev);
9386 	struct igb_adapter *adapter = netdev_priv(netdev);
9387 	struct e1000_hw *hw = &adapter->hw;
9388 	u32 err, val;
9389 
9390 	pci_set_power_state(pdev, PCI_D0);
9391 	pci_restore_state(pdev);
9392 	pci_save_state(pdev);
9393 
9394 	if (!pci_device_is_present(pdev))
9395 		return -ENODEV;
9396 	err = pci_enable_device_mem(pdev);
9397 	if (err) {
9398 		dev_err(&pdev->dev,
9399 			"igb: Cannot enable PCI device from suspend\n");
9400 		return err;
9401 	}
9402 	pci_set_master(pdev);
9403 
9404 	pci_enable_wake(pdev, PCI_D3hot, 0);
9405 	pci_enable_wake(pdev, PCI_D3cold, 0);
9406 
9407 	if (igb_init_interrupt_scheme(adapter, true)) {
9408 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9409 		return -ENOMEM;
9410 	}
9411 
9412 	igb_reset(adapter);
9413 
9414 	/* let the f/w know that the h/w is now under the control of the
9415 	 * driver.
9416 	 */
9417 	igb_get_hw_control(adapter);
9418 
9419 	val = rd32(E1000_WUS);
9420 	if (val & WAKE_PKT_WUS)
9421 		igb_deliver_wake_packet(netdev);
9422 
9423 	wr32(E1000_WUS, ~0);
9424 
9425 	if (!rpm)
9426 		rtnl_lock();
9427 	if (!err && netif_running(netdev))
9428 		err = __igb_open(netdev, true);
9429 
9430 	if (!err)
9431 		netif_device_attach(netdev);
9432 	if (!rpm)
9433 		rtnl_unlock();
9434 
9435 	return err;
9436 }
9437 
9438 static int __maybe_unused igb_resume(struct device *dev)
9439 {
9440 	return __igb_resume(dev, false);
9441 }
9442 
9443 static int __maybe_unused igb_runtime_idle(struct device *dev)
9444 {
9445 	struct net_device *netdev = dev_get_drvdata(dev);
9446 	struct igb_adapter *adapter = netdev_priv(netdev);
9447 
9448 	if (!igb_has_link(adapter))
9449 		pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
9450 
9451 	return -EBUSY;
9452 }
9453 
9454 static int __maybe_unused igb_runtime_suspend(struct device *dev)
9455 {
9456 	return __igb_shutdown(to_pci_dev(dev), NULL, 1);
9457 }
9458 
9459 static int __maybe_unused igb_runtime_resume(struct device *dev)
9460 {
9461 	return __igb_resume(dev, true);
9462 }
9463 
9464 static void igb_shutdown(struct pci_dev *pdev)
9465 {
9466 	bool wake;
9467 
9468 	__igb_shutdown(pdev, &wake, 0);
9469 
9470 	if (system_state == SYSTEM_POWER_OFF) {
9471 		pci_wake_from_d3(pdev, wake);
9472 		pci_set_power_state(pdev, PCI_D3hot);
9473 	}
9474 }
9475 
9476 #ifdef CONFIG_PCI_IOV
9477 static int igb_sriov_reinit(struct pci_dev *dev)
9478 {
9479 	struct net_device *netdev = pci_get_drvdata(dev);
9480 	struct igb_adapter *adapter = netdev_priv(netdev);
9481 	struct pci_dev *pdev = adapter->pdev;
9482 
9483 	rtnl_lock();
9484 
9485 	if (netif_running(netdev))
9486 		igb_close(netdev);
9487 	else
9488 		igb_reset(adapter);
9489 
9490 	igb_clear_interrupt_scheme(adapter);
9491 
9492 	igb_init_queue_configuration(adapter);
9493 
9494 	if (igb_init_interrupt_scheme(adapter, true)) {
9495 		rtnl_unlock();
9496 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9497 		return -ENOMEM;
9498 	}
9499 
9500 	if (netif_running(netdev))
9501 		igb_open(netdev);
9502 
9503 	rtnl_unlock();
9504 
9505 	return 0;
9506 }
9507 
9508 static int igb_pci_disable_sriov(struct pci_dev *dev)
9509 {
9510 	int err = igb_disable_sriov(dev);
9511 
9512 	if (!err)
9513 		err = igb_sriov_reinit(dev);
9514 
9515 	return err;
9516 }
9517 
9518 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
9519 {
9520 	int err = igb_enable_sriov(dev, num_vfs);
9521 
9522 	if (err)
9523 		goto out;
9524 
9525 	err = igb_sriov_reinit(dev);
9526 	if (!err)
9527 		return num_vfs;
9528 
9529 out:
9530 	return err;
9531 }
9532 
9533 #endif
9534 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
9535 {
9536 #ifdef CONFIG_PCI_IOV
9537 	if (num_vfs == 0)
9538 		return igb_pci_disable_sriov(dev);
9539 	else
9540 		return igb_pci_enable_sriov(dev, num_vfs);
9541 #endif
9542 	return 0;
9543 }
9544 
9545 /**
9546  *  igb_io_error_detected - called when PCI error is detected
9547  *  @pdev: Pointer to PCI device
9548  *  @state: The current pci connection state
9549  *
9550  *  This function is called after a PCI bus error affecting
9551  *  this device has been detected.
9552  **/
9553 static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
9554 					      pci_channel_state_t state)
9555 {
9556 	struct net_device *netdev = pci_get_drvdata(pdev);
9557 	struct igb_adapter *adapter = netdev_priv(netdev);
9558 
9559 	netif_device_detach(netdev);
9560 
9561 	if (state == pci_channel_io_perm_failure)
9562 		return PCI_ERS_RESULT_DISCONNECT;
9563 
9564 	if (netif_running(netdev))
9565 		igb_down(adapter);
9566 	pci_disable_device(pdev);
9567 
9568 	/* Request a slot reset. */
9569 	return PCI_ERS_RESULT_NEED_RESET;
9570 }
9571 
9572 /**
9573  *  igb_io_slot_reset - called after the pci bus has been reset.
9574  *  @pdev: Pointer to PCI device
9575  *
9576  *  Restart the card from scratch, as if from a cold-boot. Implementation
9577  *  resembles the first-half of the __igb_resume routine.
9578  **/
9579 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
9580 {
9581 	struct net_device *netdev = pci_get_drvdata(pdev);
9582 	struct igb_adapter *adapter = netdev_priv(netdev);
9583 	struct e1000_hw *hw = &adapter->hw;
9584 	pci_ers_result_t result;
9585 
9586 	if (pci_enable_device_mem(pdev)) {
9587 		dev_err(&pdev->dev,
9588 			"Cannot re-enable PCI device after reset.\n");
9589 		result = PCI_ERS_RESULT_DISCONNECT;
9590 	} else {
9591 		pci_set_master(pdev);
9592 		pci_restore_state(pdev);
9593 		pci_save_state(pdev);
9594 
9595 		pci_enable_wake(pdev, PCI_D3hot, 0);
9596 		pci_enable_wake(pdev, PCI_D3cold, 0);
9597 
9598 		/* In case of PCI error, adapter lose its HW address
9599 		 * so we should re-assign it here.
9600 		 */
9601 		hw->hw_addr = adapter->io_addr;
9602 
9603 		igb_reset(adapter);
9604 		wr32(E1000_WUS, ~0);
9605 		result = PCI_ERS_RESULT_RECOVERED;
9606 	}
9607 
9608 	return result;
9609 }
9610 
9611 /**
9612  *  igb_io_resume - called when traffic can start flowing again.
9613  *  @pdev: Pointer to PCI device
9614  *
9615  *  This callback is called when the error recovery driver tells us that
9616  *  its OK to resume normal operation. Implementation resembles the
9617  *  second-half of the __igb_resume routine.
9618  */
9619 static void igb_io_resume(struct pci_dev *pdev)
9620 {
9621 	struct net_device *netdev = pci_get_drvdata(pdev);
9622 	struct igb_adapter *adapter = netdev_priv(netdev);
9623 
9624 	if (netif_running(netdev)) {
9625 		if (igb_up(adapter)) {
9626 			dev_err(&pdev->dev, "igb_up failed after reset\n");
9627 			return;
9628 		}
9629 	}
9630 
9631 	netif_device_attach(netdev);
9632 
9633 	/* let the f/w know that the h/w is now under the control of the
9634 	 * driver.
9635 	 */
9636 	igb_get_hw_control(adapter);
9637 }
9638 
9639 /**
9640  *  igb_rar_set_index - Sync RAL[index] and RAH[index] registers with MAC table
9641  *  @adapter: Pointer to adapter structure
9642  *  @index: Index of the RAR entry which need to be synced with MAC table
9643  **/
9644 static void igb_rar_set_index(struct igb_adapter *adapter, u32 index)
9645 {
9646 	struct e1000_hw *hw = &adapter->hw;
9647 	u32 rar_low, rar_high;
9648 	u8 *addr = adapter->mac_table[index].addr;
9649 
9650 	/* HW expects these to be in network order when they are plugged
9651 	 * into the registers which are little endian.  In order to guarantee
9652 	 * that ordering we need to do an leXX_to_cpup here in order to be
9653 	 * ready for the byteswap that occurs with writel
9654 	 */
9655 	rar_low = le32_to_cpup((__le32 *)(addr));
9656 	rar_high = le16_to_cpup((__le16 *)(addr + 4));
9657 
9658 	/* Indicate to hardware the Address is Valid. */
9659 	if (adapter->mac_table[index].state & IGB_MAC_STATE_IN_USE) {
9660 		if (is_valid_ether_addr(addr))
9661 			rar_high |= E1000_RAH_AV;
9662 
9663 		if (adapter->mac_table[index].state & IGB_MAC_STATE_SRC_ADDR)
9664 			rar_high |= E1000_RAH_ASEL_SRC_ADDR;
9665 
9666 		switch (hw->mac.type) {
9667 		case e1000_82575:
9668 		case e1000_i210:
9669 			if (adapter->mac_table[index].state &
9670 			    IGB_MAC_STATE_QUEUE_STEERING)
9671 				rar_high |= E1000_RAH_QSEL_ENABLE;
9672 
9673 			rar_high |= E1000_RAH_POOL_1 *
9674 				    adapter->mac_table[index].queue;
9675 			break;
9676 		default:
9677 			rar_high |= E1000_RAH_POOL_1 <<
9678 				    adapter->mac_table[index].queue;
9679 			break;
9680 		}
9681 	}
9682 
9683 	wr32(E1000_RAL(index), rar_low);
9684 	wrfl();
9685 	wr32(E1000_RAH(index), rar_high);
9686 	wrfl();
9687 }
9688 
9689 static int igb_set_vf_mac(struct igb_adapter *adapter,
9690 			  int vf, unsigned char *mac_addr)
9691 {
9692 	struct e1000_hw *hw = &adapter->hw;
9693 	/* VF MAC addresses start at end of receive addresses and moves
9694 	 * towards the first, as a result a collision should not be possible
9695 	 */
9696 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
9697 	unsigned char *vf_mac_addr = adapter->vf_data[vf].vf_mac_addresses;
9698 
9699 	ether_addr_copy(vf_mac_addr, mac_addr);
9700 	ether_addr_copy(adapter->mac_table[rar_entry].addr, mac_addr);
9701 	adapter->mac_table[rar_entry].queue = vf;
9702 	adapter->mac_table[rar_entry].state |= IGB_MAC_STATE_IN_USE;
9703 	igb_rar_set_index(adapter, rar_entry);
9704 
9705 	return 0;
9706 }
9707 
9708 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
9709 {
9710 	struct igb_adapter *adapter = netdev_priv(netdev);
9711 
9712 	if (vf >= adapter->vfs_allocated_count)
9713 		return -EINVAL;
9714 
9715 	/* Setting the VF MAC to 0 reverts the IGB_VF_FLAG_PF_SET_MAC
9716 	 * flag and allows to overwrite the MAC via VF netdev.  This
9717 	 * is necessary to allow libvirt a way to restore the original
9718 	 * MAC after unbinding vfio-pci and reloading igbvf after shutting
9719 	 * down a VM.
9720 	 */
9721 	if (is_zero_ether_addr(mac)) {
9722 		adapter->vf_data[vf].flags &= ~IGB_VF_FLAG_PF_SET_MAC;
9723 		dev_info(&adapter->pdev->dev,
9724 			 "remove administratively set MAC on VF %d\n",
9725 			 vf);
9726 	} else if (is_valid_ether_addr(mac)) {
9727 		adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
9728 		dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n",
9729 			 mac, vf);
9730 		dev_info(&adapter->pdev->dev,
9731 			 "Reload the VF driver to make this change effective.");
9732 		/* Generate additional warning if PF is down */
9733 		if (test_bit(__IGB_DOWN, &adapter->state)) {
9734 			dev_warn(&adapter->pdev->dev,
9735 				 "The VF MAC address has been set, but the PF device is not up.\n");
9736 			dev_warn(&adapter->pdev->dev,
9737 				 "Bring the PF device up before attempting to use the VF device.\n");
9738 		}
9739 	} else {
9740 		return -EINVAL;
9741 	}
9742 	return igb_set_vf_mac(adapter, vf, mac);
9743 }
9744 
9745 static int igb_link_mbps(int internal_link_speed)
9746 {
9747 	switch (internal_link_speed) {
9748 	case SPEED_100:
9749 		return 100;
9750 	case SPEED_1000:
9751 		return 1000;
9752 	default:
9753 		return 0;
9754 	}
9755 }
9756 
9757 static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
9758 				  int link_speed)
9759 {
9760 	int rf_dec, rf_int;
9761 	u32 bcnrc_val;
9762 
9763 	if (tx_rate != 0) {
9764 		/* Calculate the rate factor values to set */
9765 		rf_int = link_speed / tx_rate;
9766 		rf_dec = (link_speed - (rf_int * tx_rate));
9767 		rf_dec = (rf_dec * BIT(E1000_RTTBCNRC_RF_INT_SHIFT)) /
9768 			 tx_rate;
9769 
9770 		bcnrc_val = E1000_RTTBCNRC_RS_ENA;
9771 		bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
9772 			      E1000_RTTBCNRC_RF_INT_MASK);
9773 		bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
9774 	} else {
9775 		bcnrc_val = 0;
9776 	}
9777 
9778 	wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
9779 	/* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
9780 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
9781 	 */
9782 	wr32(E1000_RTTBCNRM, 0x14);
9783 	wr32(E1000_RTTBCNRC, bcnrc_val);
9784 }
9785 
9786 static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
9787 {
9788 	int actual_link_speed, i;
9789 	bool reset_rate = false;
9790 
9791 	/* VF TX rate limit was not set or not supported */
9792 	if ((adapter->vf_rate_link_speed == 0) ||
9793 	    (adapter->hw.mac.type != e1000_82576))
9794 		return;
9795 
9796 	actual_link_speed = igb_link_mbps(adapter->link_speed);
9797 	if (actual_link_speed != adapter->vf_rate_link_speed) {
9798 		reset_rate = true;
9799 		adapter->vf_rate_link_speed = 0;
9800 		dev_info(&adapter->pdev->dev,
9801 			 "Link speed has been changed. VF Transmit rate is disabled\n");
9802 	}
9803 
9804 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
9805 		if (reset_rate)
9806 			adapter->vf_data[i].tx_rate = 0;
9807 
9808 		igb_set_vf_rate_limit(&adapter->hw, i,
9809 				      adapter->vf_data[i].tx_rate,
9810 				      actual_link_speed);
9811 	}
9812 }
9813 
9814 static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
9815 			     int min_tx_rate, int max_tx_rate)
9816 {
9817 	struct igb_adapter *adapter = netdev_priv(netdev);
9818 	struct e1000_hw *hw = &adapter->hw;
9819 	int actual_link_speed;
9820 
9821 	if (hw->mac.type != e1000_82576)
9822 		return -EOPNOTSUPP;
9823 
9824 	if (min_tx_rate)
9825 		return -EINVAL;
9826 
9827 	actual_link_speed = igb_link_mbps(adapter->link_speed);
9828 	if ((vf >= adapter->vfs_allocated_count) ||
9829 	    (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
9830 	    (max_tx_rate < 0) ||
9831 	    (max_tx_rate > actual_link_speed))
9832 		return -EINVAL;
9833 
9834 	adapter->vf_rate_link_speed = actual_link_speed;
9835 	adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
9836 	igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
9837 
9838 	return 0;
9839 }
9840 
9841 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
9842 				   bool setting)
9843 {
9844 	struct igb_adapter *adapter = netdev_priv(netdev);
9845 	struct e1000_hw *hw = &adapter->hw;
9846 	u32 reg_val, reg_offset;
9847 
9848 	if (!adapter->vfs_allocated_count)
9849 		return -EOPNOTSUPP;
9850 
9851 	if (vf >= adapter->vfs_allocated_count)
9852 		return -EINVAL;
9853 
9854 	reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
9855 	reg_val = rd32(reg_offset);
9856 	if (setting)
9857 		reg_val |= (BIT(vf) |
9858 			    BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9859 	else
9860 		reg_val &= ~(BIT(vf) |
9861 			     BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9862 	wr32(reg_offset, reg_val);
9863 
9864 	adapter->vf_data[vf].spoofchk_enabled = setting;
9865 	return 0;
9866 }
9867 
9868 static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf, bool setting)
9869 {
9870 	struct igb_adapter *adapter = netdev_priv(netdev);
9871 
9872 	if (vf >= adapter->vfs_allocated_count)
9873 		return -EINVAL;
9874 	if (adapter->vf_data[vf].trusted == setting)
9875 		return 0;
9876 
9877 	adapter->vf_data[vf].trusted = setting;
9878 
9879 	dev_info(&adapter->pdev->dev, "VF %u is %strusted\n",
9880 		 vf, setting ? "" : "not ");
9881 	return 0;
9882 }
9883 
9884 static int igb_ndo_get_vf_config(struct net_device *netdev,
9885 				 int vf, struct ifla_vf_info *ivi)
9886 {
9887 	struct igb_adapter *adapter = netdev_priv(netdev);
9888 	if (vf >= adapter->vfs_allocated_count)
9889 		return -EINVAL;
9890 	ivi->vf = vf;
9891 	memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
9892 	ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
9893 	ivi->min_tx_rate = 0;
9894 	ivi->vlan = adapter->vf_data[vf].pf_vlan;
9895 	ivi->qos = adapter->vf_data[vf].pf_qos;
9896 	ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
9897 	ivi->trusted = adapter->vf_data[vf].trusted;
9898 	return 0;
9899 }
9900 
9901 static void igb_vmm_control(struct igb_adapter *adapter)
9902 {
9903 	struct e1000_hw *hw = &adapter->hw;
9904 	u32 reg;
9905 
9906 	switch (hw->mac.type) {
9907 	case e1000_82575:
9908 	case e1000_i210:
9909 	case e1000_i211:
9910 	case e1000_i354:
9911 	default:
9912 		/* replication is not supported for 82575 */
9913 		return;
9914 	case e1000_82576:
9915 		/* notify HW that the MAC is adding vlan tags */
9916 		reg = rd32(E1000_DTXCTL);
9917 		reg |= E1000_DTXCTL_VLAN_ADDED;
9918 		wr32(E1000_DTXCTL, reg);
9919 		fallthrough;
9920 	case e1000_82580:
9921 		/* enable replication vlan tag stripping */
9922 		reg = rd32(E1000_RPLOLR);
9923 		reg |= E1000_RPLOLR_STRVLAN;
9924 		wr32(E1000_RPLOLR, reg);
9925 		fallthrough;
9926 	case e1000_i350:
9927 		/* none of the above registers are supported by i350 */
9928 		break;
9929 	}
9930 
9931 	if (adapter->vfs_allocated_count) {
9932 		igb_vmdq_set_loopback_pf(hw, true);
9933 		igb_vmdq_set_replication_pf(hw, true);
9934 		igb_vmdq_set_anti_spoofing_pf(hw, true,
9935 					      adapter->vfs_allocated_count);
9936 	} else {
9937 		igb_vmdq_set_loopback_pf(hw, false);
9938 		igb_vmdq_set_replication_pf(hw, false);
9939 	}
9940 }
9941 
9942 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
9943 {
9944 	struct e1000_hw *hw = &adapter->hw;
9945 	u32 dmac_thr;
9946 	u16 hwm;
9947 	u32 reg;
9948 
9949 	if (hw->mac.type > e1000_82580) {
9950 		if (adapter->flags & IGB_FLAG_DMAC) {
9951 			/* force threshold to 0. */
9952 			wr32(E1000_DMCTXTH, 0);
9953 
9954 			/* DMA Coalescing high water mark needs to be greater
9955 			 * than the Rx threshold. Set hwm to PBA - max frame
9956 			 * size in 16B units, capping it at PBA - 6KB.
9957 			 */
9958 			hwm = 64 * (pba - 6);
9959 			reg = rd32(E1000_FCRTC);
9960 			reg &= ~E1000_FCRTC_RTH_COAL_MASK;
9961 			reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
9962 				& E1000_FCRTC_RTH_COAL_MASK);
9963 			wr32(E1000_FCRTC, reg);
9964 
9965 			/* Set the DMA Coalescing Rx threshold to PBA - 2 * max
9966 			 * frame size, capping it at PBA - 10KB.
9967 			 */
9968 			dmac_thr = pba - 10;
9969 			reg = rd32(E1000_DMACR);
9970 			reg &= ~E1000_DMACR_DMACTHR_MASK;
9971 			reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
9972 				& E1000_DMACR_DMACTHR_MASK);
9973 
9974 			/* transition to L0x or L1 if available..*/
9975 			reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
9976 
9977 			/* watchdog timer= +-1000 usec in 32usec intervals */
9978 			reg |= (1000 >> 5);
9979 
9980 			/* Disable BMC-to-OS Watchdog Enable */
9981 			if (hw->mac.type != e1000_i354)
9982 				reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
9983 			wr32(E1000_DMACR, reg);
9984 
9985 			/* no lower threshold to disable
9986 			 * coalescing(smart fifb)-UTRESH=0
9987 			 */
9988 			wr32(E1000_DMCRTRH, 0);
9989 
9990 			reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
9991 
9992 			wr32(E1000_DMCTLX, reg);
9993 
9994 			/* free space in tx packet buffer to wake from
9995 			 * DMA coal
9996 			 */
9997 			wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
9998 			     (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
9999 		}
10000 
10001 		if (hw->mac.type >= e1000_i210 ||
10002 		    (adapter->flags & IGB_FLAG_DMAC)) {
10003 			reg = rd32(E1000_PCIEMISC);
10004 			reg |= E1000_PCIEMISC_LX_DECISION;
10005 			wr32(E1000_PCIEMISC, reg);
10006 		} /* endif adapter->dmac is not disabled */
10007 	} else if (hw->mac.type == e1000_82580) {
10008 		u32 reg = rd32(E1000_PCIEMISC);
10009 
10010 		wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
10011 		wr32(E1000_DMACR, 0);
10012 	}
10013 }
10014 
10015 /**
10016  *  igb_read_i2c_byte - Reads 8 bit word over I2C
10017  *  @hw: pointer to hardware structure
10018  *  @byte_offset: byte offset to read
10019  *  @dev_addr: device address
10020  *  @data: value read
10021  *
10022  *  Performs byte read operation over I2C interface at
10023  *  a specified device address.
10024  **/
10025 s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
10026 		      u8 dev_addr, u8 *data)
10027 {
10028 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
10029 	struct i2c_client *this_client = adapter->i2c_client;
10030 	s32 status;
10031 	u16 swfw_mask = 0;
10032 
10033 	if (!this_client)
10034 		return E1000_ERR_I2C;
10035 
10036 	swfw_mask = E1000_SWFW_PHY0_SM;
10037 
10038 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
10039 		return E1000_ERR_SWFW_SYNC;
10040 
10041 	status = i2c_smbus_read_byte_data(this_client, byte_offset);
10042 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
10043 
10044 	if (status < 0)
10045 		return E1000_ERR_I2C;
10046 	else {
10047 		*data = status;
10048 		return 0;
10049 	}
10050 }
10051 
10052 /**
10053  *  igb_write_i2c_byte - Writes 8 bit word over I2C
10054  *  @hw: pointer to hardware structure
10055  *  @byte_offset: byte offset to write
10056  *  @dev_addr: device address
10057  *  @data: value to write
10058  *
10059  *  Performs byte write operation over I2C interface at
10060  *  a specified device address.
10061  **/
10062 s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
10063 		       u8 dev_addr, u8 data)
10064 {
10065 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
10066 	struct i2c_client *this_client = adapter->i2c_client;
10067 	s32 status;
10068 	u16 swfw_mask = E1000_SWFW_PHY0_SM;
10069 
10070 	if (!this_client)
10071 		return E1000_ERR_I2C;
10072 
10073 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
10074 		return E1000_ERR_SWFW_SYNC;
10075 	status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
10076 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
10077 
10078 	if (status)
10079 		return E1000_ERR_I2C;
10080 	else
10081 		return 0;
10082 
10083 }
10084 
10085 int igb_reinit_queues(struct igb_adapter *adapter)
10086 {
10087 	struct net_device *netdev = adapter->netdev;
10088 	struct pci_dev *pdev = adapter->pdev;
10089 	int err = 0;
10090 
10091 	if (netif_running(netdev))
10092 		igb_close(netdev);
10093 
10094 	igb_reset_interrupt_capability(adapter);
10095 
10096 	if (igb_init_interrupt_scheme(adapter, true)) {
10097 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
10098 		return -ENOMEM;
10099 	}
10100 
10101 	if (netif_running(netdev))
10102 		err = igb_open(netdev);
10103 
10104 	return err;
10105 }
10106 
10107 static void igb_nfc_filter_exit(struct igb_adapter *adapter)
10108 {
10109 	struct igb_nfc_filter *rule;
10110 
10111 	spin_lock(&adapter->nfc_lock);
10112 
10113 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
10114 		igb_erase_filter(adapter, rule);
10115 
10116 	hlist_for_each_entry(rule, &adapter->cls_flower_list, nfc_node)
10117 		igb_erase_filter(adapter, rule);
10118 
10119 	spin_unlock(&adapter->nfc_lock);
10120 }
10121 
10122 static void igb_nfc_filter_restore(struct igb_adapter *adapter)
10123 {
10124 	struct igb_nfc_filter *rule;
10125 
10126 	spin_lock(&adapter->nfc_lock);
10127 
10128 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
10129 		igb_add_filter(adapter, rule);
10130 
10131 	spin_unlock(&adapter->nfc_lock);
10132 }
10133 /* igb_main.c */
10134