xref: /linux/drivers/net/ethernet/intel/igb/igb_main.c (revision e7d759f31ca295d589f7420719c311870bb3166f)
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/prefetch.h>
32 #include <linux/bpf.h>
33 #include <linux/bpf_trace.h>
34 #include <linux/pm_runtime.h>
35 #include <linux/etherdevice.h>
36 #ifdef CONFIG_IGB_DCA
37 #include <linux/dca.h>
38 #endif
39 #include <linux/i2c.h>
40 #include "igb.h"
41 
42 enum queue_mode {
43 	QUEUE_MODE_STRICT_PRIORITY,
44 	QUEUE_MODE_STREAM_RESERVATION,
45 };
46 
47 enum tx_queue_prio {
48 	TX_QUEUE_PRIO_HIGH,
49 	TX_QUEUE_PRIO_LOW,
50 };
51 
52 char igb_driver_name[] = "igb";
53 static const char igb_driver_string[] =
54 				"Intel(R) Gigabit Ethernet Network Driver";
55 static const char igb_copyright[] =
56 				"Copyright (c) 2007-2014 Intel Corporation.";
57 
58 static const struct e1000_info *igb_info_tbl[] = {
59 	[board_82575] = &e1000_82575_info,
60 };
61 
62 static const struct pci_device_id igb_pci_tbl[] = {
63 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
64 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_SGMII) },
65 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
66 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I211_COPPER), board_82575 },
67 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER), board_82575 },
68 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_FIBER), board_82575 },
69 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES), board_82575 },
70 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SGMII), board_82575 },
71 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER_FLASHLESS), board_82575 },
72 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES_FLASHLESS), board_82575 },
73 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_COPPER), board_82575 },
74 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_FIBER), board_82575 },
75 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SERDES), board_82575 },
76 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SGMII), board_82575 },
77 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER), board_82575 },
78 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_FIBER), board_82575 },
79 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_QUAD_FIBER), board_82575 },
80 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SERDES), board_82575 },
81 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SGMII), board_82575 },
82 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER_DUAL), board_82575 },
83 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SGMII), board_82575 },
84 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SERDES), board_82575 },
85 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_BACKPLANE), board_82575 },
86 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SFP), board_82575 },
87 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
88 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
89 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
90 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
91 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
92 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
93 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER_ET2), board_82575 },
94 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
95 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
96 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
97 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
98 	/* required last entry */
99 	{0, }
100 };
101 
102 MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
103 
104 static int igb_setup_all_tx_resources(struct igb_adapter *);
105 static int igb_setup_all_rx_resources(struct igb_adapter *);
106 static void igb_free_all_tx_resources(struct igb_adapter *);
107 static void igb_free_all_rx_resources(struct igb_adapter *);
108 static void igb_setup_mrqc(struct igb_adapter *);
109 static int igb_probe(struct pci_dev *, const struct pci_device_id *);
110 static void igb_remove(struct pci_dev *pdev);
111 static void igb_init_queue_configuration(struct igb_adapter *adapter);
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_disable_sriov(struct pci_dev *dev, bool reinit);
179 #endif
180 
181 static int igb_suspend(struct device *);
182 static int igb_resume(struct device *);
183 static int igb_runtime_suspend(struct device *dev);
184 static int igb_runtime_resume(struct device *dev);
185 static int igb_runtime_idle(struct device *dev);
186 #ifdef CONFIG_PM
187 static const struct dev_pm_ops igb_pm_ops = {
188 	SET_SYSTEM_SLEEP_PM_OPS(igb_suspend, igb_resume)
189 	SET_RUNTIME_PM_OPS(igb_runtime_suspend, igb_runtime_resume,
190 			igb_runtime_idle)
191 };
192 #endif
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 #ifdef CONFIG_IGB_HWMON
2260 /**
2261  *  igb_set_i2c_bb - Init I2C interface
2262  *  @hw: pointer to hardware structure
2263  **/
2264 static void igb_set_i2c_bb(struct e1000_hw *hw)
2265 {
2266 	u32 ctrl_ext;
2267 	s32 i2cctl;
2268 
2269 	ctrl_ext = rd32(E1000_CTRL_EXT);
2270 	ctrl_ext |= E1000_CTRL_I2C_ENA;
2271 	wr32(E1000_CTRL_EXT, ctrl_ext);
2272 	wrfl();
2273 
2274 	i2cctl = rd32(E1000_I2CPARAMS);
2275 	i2cctl |= E1000_I2CBB_EN
2276 		| E1000_I2C_CLK_OE_N
2277 		| E1000_I2C_DATA_OE_N;
2278 	wr32(E1000_I2CPARAMS, i2cctl);
2279 	wrfl();
2280 }
2281 #endif
2282 
2283 void igb_reset(struct igb_adapter *adapter)
2284 {
2285 	struct pci_dev *pdev = adapter->pdev;
2286 	struct e1000_hw *hw = &adapter->hw;
2287 	struct e1000_mac_info *mac = &hw->mac;
2288 	struct e1000_fc_info *fc = &hw->fc;
2289 	u32 pba, hwm;
2290 
2291 	/* Repartition Pba for greater than 9k mtu
2292 	 * To take effect CTRL.RST is required.
2293 	 */
2294 	switch (mac->type) {
2295 	case e1000_i350:
2296 	case e1000_i354:
2297 	case e1000_82580:
2298 		pba = rd32(E1000_RXPBS);
2299 		pba = igb_rxpbs_adjust_82580(pba);
2300 		break;
2301 	case e1000_82576:
2302 		pba = rd32(E1000_RXPBS);
2303 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2304 		break;
2305 	case e1000_82575:
2306 	case e1000_i210:
2307 	case e1000_i211:
2308 	default:
2309 		pba = E1000_PBA_34K;
2310 		break;
2311 	}
2312 
2313 	if (mac->type == e1000_82575) {
2314 		u32 min_rx_space, min_tx_space, needed_tx_space;
2315 
2316 		/* write Rx PBA so that hardware can report correct Tx PBA */
2317 		wr32(E1000_PBA, pba);
2318 
2319 		/* To maintain wire speed transmits, the Tx FIFO should be
2320 		 * large enough to accommodate two full transmit packets,
2321 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
2322 		 * the Rx FIFO should be large enough to accommodate at least
2323 		 * one full receive packet and is similarly rounded up and
2324 		 * expressed in KB.
2325 		 */
2326 		min_rx_space = DIV_ROUND_UP(MAX_JUMBO_FRAME_SIZE, 1024);
2327 
2328 		/* The Tx FIFO also stores 16 bytes of information about the Tx
2329 		 * but don't include Ethernet FCS because hardware appends it.
2330 		 * We only need to round down to the nearest 512 byte block
2331 		 * count since the value we care about is 2 frames, not 1.
2332 		 */
2333 		min_tx_space = adapter->max_frame_size;
2334 		min_tx_space += sizeof(union e1000_adv_tx_desc) - ETH_FCS_LEN;
2335 		min_tx_space = DIV_ROUND_UP(min_tx_space, 512);
2336 
2337 		/* upper 16 bits has Tx packet buffer allocation size in KB */
2338 		needed_tx_space = min_tx_space - (rd32(E1000_PBA) >> 16);
2339 
2340 		/* If current Tx allocation is less than the min Tx FIFO size,
2341 		 * and the min Tx FIFO size is less than the current Rx FIFO
2342 		 * allocation, take space away from current Rx allocation.
2343 		 */
2344 		if (needed_tx_space < pba) {
2345 			pba -= needed_tx_space;
2346 
2347 			/* if short on Rx space, Rx wins and must trump Tx
2348 			 * adjustment
2349 			 */
2350 			if (pba < min_rx_space)
2351 				pba = min_rx_space;
2352 		}
2353 
2354 		/* adjust PBA for jumbo frames */
2355 		wr32(E1000_PBA, pba);
2356 	}
2357 
2358 	/* flow control settings
2359 	 * The high water mark must be low enough to fit one full frame
2360 	 * after transmitting the pause frame.  As such we must have enough
2361 	 * space to allow for us to complete our current transmit and then
2362 	 * receive the frame that is in progress from the link partner.
2363 	 * Set it to:
2364 	 * - the full Rx FIFO size minus one full Tx plus one full Rx frame
2365 	 */
2366 	hwm = (pba << 10) - (adapter->max_frame_size + MAX_JUMBO_FRAME_SIZE);
2367 
2368 	fc->high_water = hwm & 0xFFFFFFF0;	/* 16-byte granularity */
2369 	fc->low_water = fc->high_water - 16;
2370 	fc->pause_time = 0xFFFF;
2371 	fc->send_xon = 1;
2372 	fc->current_mode = fc->requested_mode;
2373 
2374 	/* disable receive for all VFs and wait one second */
2375 	if (adapter->vfs_allocated_count) {
2376 		int i;
2377 
2378 		for (i = 0 ; i < adapter->vfs_allocated_count; i++)
2379 			adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
2380 
2381 		/* ping all the active vfs to let them know we are going down */
2382 		igb_ping_all_vfs(adapter);
2383 
2384 		/* disable transmits and receives */
2385 		wr32(E1000_VFRE, 0);
2386 		wr32(E1000_VFTE, 0);
2387 	}
2388 
2389 	/* Allow time for pending master requests to run */
2390 	hw->mac.ops.reset_hw(hw);
2391 	wr32(E1000_WUC, 0);
2392 
2393 	if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
2394 		/* need to resetup here after media swap */
2395 		adapter->ei.get_invariants(hw);
2396 		adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
2397 	}
2398 	if ((mac->type == e1000_82575 || mac->type == e1000_i350) &&
2399 	    (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
2400 		igb_enable_mas(adapter);
2401 	}
2402 	if (hw->mac.ops.init_hw(hw))
2403 		dev_err(&pdev->dev, "Hardware Error\n");
2404 
2405 	/* RAR registers were cleared during init_hw, clear mac table */
2406 	igb_flush_mac_table(adapter);
2407 	__dev_uc_unsync(adapter->netdev, NULL);
2408 
2409 	/* Recover default RAR entry */
2410 	igb_set_default_mac_filter(adapter);
2411 
2412 	/* Flow control settings reset on hardware reset, so guarantee flow
2413 	 * control is off when forcing speed.
2414 	 */
2415 	if (!hw->mac.autoneg)
2416 		igb_force_mac_fc(hw);
2417 
2418 	igb_init_dmac(adapter, pba);
2419 #ifdef CONFIG_IGB_HWMON
2420 	/* Re-initialize the thermal sensor on i350 devices. */
2421 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
2422 		if (mac->type == e1000_i350 && hw->bus.func == 0) {
2423 			/* If present, re-initialize the external thermal sensor
2424 			 * interface.
2425 			 */
2426 			if (adapter->ets)
2427 				igb_set_i2c_bb(hw);
2428 			mac->ops.init_thermal_sensor_thresh(hw);
2429 		}
2430 	}
2431 #endif
2432 	/* Re-establish EEE setting */
2433 	if (hw->phy.media_type == e1000_media_type_copper) {
2434 		switch (mac->type) {
2435 		case e1000_i350:
2436 		case e1000_i210:
2437 		case e1000_i211:
2438 			igb_set_eee_i350(hw, true, true);
2439 			break;
2440 		case e1000_i354:
2441 			igb_set_eee_i354(hw, true, true);
2442 			break;
2443 		default:
2444 			break;
2445 		}
2446 	}
2447 	if (!netif_running(adapter->netdev))
2448 		igb_power_down_link(adapter);
2449 
2450 	igb_update_mng_vlan(adapter);
2451 
2452 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2453 	wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2454 
2455 	/* Re-enable PTP, where applicable. */
2456 	if (adapter->ptp_flags & IGB_PTP_ENABLED)
2457 		igb_ptp_reset(adapter);
2458 
2459 	igb_get_phy_info(hw);
2460 }
2461 
2462 static netdev_features_t igb_fix_features(struct net_device *netdev,
2463 	netdev_features_t features)
2464 {
2465 	/* Since there is no support for separate Rx/Tx vlan accel
2466 	 * enable/disable make sure Tx flag is always in same state as Rx.
2467 	 */
2468 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2469 		features |= NETIF_F_HW_VLAN_CTAG_TX;
2470 	else
2471 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2472 
2473 	return features;
2474 }
2475 
2476 static int igb_set_features(struct net_device *netdev,
2477 	netdev_features_t features)
2478 {
2479 	netdev_features_t changed = netdev->features ^ features;
2480 	struct igb_adapter *adapter = netdev_priv(netdev);
2481 
2482 	if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2483 		igb_vlan_mode(netdev, features);
2484 
2485 	if (!(changed & (NETIF_F_RXALL | NETIF_F_NTUPLE)))
2486 		return 0;
2487 
2488 	if (!(features & NETIF_F_NTUPLE)) {
2489 		struct hlist_node *node2;
2490 		struct igb_nfc_filter *rule;
2491 
2492 		spin_lock(&adapter->nfc_lock);
2493 		hlist_for_each_entry_safe(rule, node2,
2494 					  &adapter->nfc_filter_list, nfc_node) {
2495 			igb_erase_filter(adapter, rule);
2496 			hlist_del(&rule->nfc_node);
2497 			kfree(rule);
2498 		}
2499 		spin_unlock(&adapter->nfc_lock);
2500 		adapter->nfc_filter_count = 0;
2501 	}
2502 
2503 	netdev->features = features;
2504 
2505 	if (netif_running(netdev))
2506 		igb_reinit_locked(adapter);
2507 	else
2508 		igb_reset(adapter);
2509 
2510 	return 1;
2511 }
2512 
2513 static int igb_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
2514 			   struct net_device *dev,
2515 			   const unsigned char *addr, u16 vid,
2516 			   u16 flags,
2517 			   struct netlink_ext_ack *extack)
2518 {
2519 	/* guarantee we can provide a unique filter for the unicast address */
2520 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
2521 		struct igb_adapter *adapter = netdev_priv(dev);
2522 		int vfn = adapter->vfs_allocated_count;
2523 
2524 		if (netdev_uc_count(dev) >= igb_available_rars(adapter, vfn))
2525 			return -ENOMEM;
2526 	}
2527 
2528 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
2529 }
2530 
2531 #define IGB_MAX_MAC_HDR_LEN	127
2532 #define IGB_MAX_NETWORK_HDR_LEN	511
2533 
2534 static netdev_features_t
2535 igb_features_check(struct sk_buff *skb, struct net_device *dev,
2536 		   netdev_features_t features)
2537 {
2538 	unsigned int network_hdr_len, mac_hdr_len;
2539 
2540 	/* Make certain the headers can be described by a context descriptor */
2541 	mac_hdr_len = skb_network_header(skb) - skb->data;
2542 	if (unlikely(mac_hdr_len > IGB_MAX_MAC_HDR_LEN))
2543 		return features & ~(NETIF_F_HW_CSUM |
2544 				    NETIF_F_SCTP_CRC |
2545 				    NETIF_F_GSO_UDP_L4 |
2546 				    NETIF_F_HW_VLAN_CTAG_TX |
2547 				    NETIF_F_TSO |
2548 				    NETIF_F_TSO6);
2549 
2550 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
2551 	if (unlikely(network_hdr_len >  IGB_MAX_NETWORK_HDR_LEN))
2552 		return features & ~(NETIF_F_HW_CSUM |
2553 				    NETIF_F_SCTP_CRC |
2554 				    NETIF_F_GSO_UDP_L4 |
2555 				    NETIF_F_TSO |
2556 				    NETIF_F_TSO6);
2557 
2558 	/* We can only support IPV4 TSO in tunnels if we can mangle the
2559 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
2560 	 */
2561 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
2562 		features &= ~NETIF_F_TSO;
2563 
2564 	return features;
2565 }
2566 
2567 static void igb_offload_apply(struct igb_adapter *adapter, s32 queue)
2568 {
2569 	if (!is_fqtss_enabled(adapter)) {
2570 		enable_fqtss(adapter, true);
2571 		return;
2572 	}
2573 
2574 	igb_config_tx_modes(adapter, queue);
2575 
2576 	if (!is_any_cbs_enabled(adapter) && !is_any_txtime_enabled(adapter))
2577 		enable_fqtss(adapter, false);
2578 }
2579 
2580 static int igb_offload_cbs(struct igb_adapter *adapter,
2581 			   struct tc_cbs_qopt_offload *qopt)
2582 {
2583 	struct e1000_hw *hw = &adapter->hw;
2584 	int err;
2585 
2586 	/* CBS offloading is only supported by i210 controller. */
2587 	if (hw->mac.type != e1000_i210)
2588 		return -EOPNOTSUPP;
2589 
2590 	/* CBS offloading is only supported by queue 0 and queue 1. */
2591 	if (qopt->queue < 0 || qopt->queue > 1)
2592 		return -EINVAL;
2593 
2594 	err = igb_save_cbs_params(adapter, qopt->queue, qopt->enable,
2595 				  qopt->idleslope, qopt->sendslope,
2596 				  qopt->hicredit, qopt->locredit);
2597 	if (err)
2598 		return err;
2599 
2600 	igb_offload_apply(adapter, qopt->queue);
2601 
2602 	return 0;
2603 }
2604 
2605 #define ETHER_TYPE_FULL_MASK ((__force __be16)~0)
2606 #define VLAN_PRIO_FULL_MASK (0x07)
2607 
2608 static int igb_parse_cls_flower(struct igb_adapter *adapter,
2609 				struct flow_cls_offload *f,
2610 				int traffic_class,
2611 				struct igb_nfc_filter *input)
2612 {
2613 	struct flow_rule *rule = flow_cls_offload_flow_rule(f);
2614 	struct flow_dissector *dissector = rule->match.dissector;
2615 	struct netlink_ext_ack *extack = f->common.extack;
2616 
2617 	if (dissector->used_keys &
2618 	    ~(BIT_ULL(FLOW_DISSECTOR_KEY_BASIC) |
2619 	      BIT_ULL(FLOW_DISSECTOR_KEY_CONTROL) |
2620 	      BIT_ULL(FLOW_DISSECTOR_KEY_ETH_ADDRS) |
2621 	      BIT_ULL(FLOW_DISSECTOR_KEY_VLAN))) {
2622 		NL_SET_ERR_MSG_MOD(extack,
2623 				   "Unsupported key used, only BASIC, CONTROL, ETH_ADDRS and VLAN are supported");
2624 		return -EOPNOTSUPP;
2625 	}
2626 
2627 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
2628 		struct flow_match_eth_addrs match;
2629 
2630 		flow_rule_match_eth_addrs(rule, &match);
2631 		if (!is_zero_ether_addr(match.mask->dst)) {
2632 			if (!is_broadcast_ether_addr(match.mask->dst)) {
2633 				NL_SET_ERR_MSG_MOD(extack, "Only full masks are supported for destination MAC address");
2634 				return -EINVAL;
2635 			}
2636 
2637 			input->filter.match_flags |=
2638 				IGB_FILTER_FLAG_DST_MAC_ADDR;
2639 			ether_addr_copy(input->filter.dst_addr, match.key->dst);
2640 		}
2641 
2642 		if (!is_zero_ether_addr(match.mask->src)) {
2643 			if (!is_broadcast_ether_addr(match.mask->src)) {
2644 				NL_SET_ERR_MSG_MOD(extack, "Only full masks are supported for source MAC address");
2645 				return -EINVAL;
2646 			}
2647 
2648 			input->filter.match_flags |=
2649 				IGB_FILTER_FLAG_SRC_MAC_ADDR;
2650 			ether_addr_copy(input->filter.src_addr, match.key->src);
2651 		}
2652 	}
2653 
2654 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
2655 		struct flow_match_basic match;
2656 
2657 		flow_rule_match_basic(rule, &match);
2658 		if (match.mask->n_proto) {
2659 			if (match.mask->n_proto != ETHER_TYPE_FULL_MASK) {
2660 				NL_SET_ERR_MSG_MOD(extack, "Only full mask is supported for EtherType filter");
2661 				return -EINVAL;
2662 			}
2663 
2664 			input->filter.match_flags |= IGB_FILTER_FLAG_ETHER_TYPE;
2665 			input->filter.etype = match.key->n_proto;
2666 		}
2667 	}
2668 
2669 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
2670 		struct flow_match_vlan match;
2671 
2672 		flow_rule_match_vlan(rule, &match);
2673 		if (match.mask->vlan_priority) {
2674 			if (match.mask->vlan_priority != VLAN_PRIO_FULL_MASK) {
2675 				NL_SET_ERR_MSG_MOD(extack, "Only full mask is supported for VLAN priority");
2676 				return -EINVAL;
2677 			}
2678 
2679 			input->filter.match_flags |= IGB_FILTER_FLAG_VLAN_TCI;
2680 			input->filter.vlan_tci =
2681 				(__force __be16)match.key->vlan_priority;
2682 		}
2683 	}
2684 
2685 	input->action = traffic_class;
2686 	input->cookie = f->cookie;
2687 
2688 	return 0;
2689 }
2690 
2691 static int igb_configure_clsflower(struct igb_adapter *adapter,
2692 				   struct flow_cls_offload *cls_flower)
2693 {
2694 	struct netlink_ext_ack *extack = cls_flower->common.extack;
2695 	struct igb_nfc_filter *filter, *f;
2696 	int err, tc;
2697 
2698 	tc = tc_classid_to_hwtc(adapter->netdev, cls_flower->classid);
2699 	if (tc < 0) {
2700 		NL_SET_ERR_MSG_MOD(extack, "Invalid traffic class");
2701 		return -EINVAL;
2702 	}
2703 
2704 	filter = kzalloc(sizeof(*filter), GFP_KERNEL);
2705 	if (!filter)
2706 		return -ENOMEM;
2707 
2708 	err = igb_parse_cls_flower(adapter, cls_flower, tc, filter);
2709 	if (err < 0)
2710 		goto err_parse;
2711 
2712 	spin_lock(&adapter->nfc_lock);
2713 
2714 	hlist_for_each_entry(f, &adapter->nfc_filter_list, nfc_node) {
2715 		if (!memcmp(&f->filter, &filter->filter, sizeof(f->filter))) {
2716 			err = -EEXIST;
2717 			NL_SET_ERR_MSG_MOD(extack,
2718 					   "This filter is already set in ethtool");
2719 			goto err_locked;
2720 		}
2721 	}
2722 
2723 	hlist_for_each_entry(f, &adapter->cls_flower_list, nfc_node) {
2724 		if (!memcmp(&f->filter, &filter->filter, sizeof(f->filter))) {
2725 			err = -EEXIST;
2726 			NL_SET_ERR_MSG_MOD(extack,
2727 					   "This filter is already set in cls_flower");
2728 			goto err_locked;
2729 		}
2730 	}
2731 
2732 	err = igb_add_filter(adapter, filter);
2733 	if (err < 0) {
2734 		NL_SET_ERR_MSG_MOD(extack, "Could not add filter to the adapter");
2735 		goto err_locked;
2736 	}
2737 
2738 	hlist_add_head(&filter->nfc_node, &adapter->cls_flower_list);
2739 
2740 	spin_unlock(&adapter->nfc_lock);
2741 
2742 	return 0;
2743 
2744 err_locked:
2745 	spin_unlock(&adapter->nfc_lock);
2746 
2747 err_parse:
2748 	kfree(filter);
2749 
2750 	return err;
2751 }
2752 
2753 static int igb_delete_clsflower(struct igb_adapter *adapter,
2754 				struct flow_cls_offload *cls_flower)
2755 {
2756 	struct igb_nfc_filter *filter;
2757 	int err;
2758 
2759 	spin_lock(&adapter->nfc_lock);
2760 
2761 	hlist_for_each_entry(filter, &adapter->cls_flower_list, nfc_node)
2762 		if (filter->cookie == cls_flower->cookie)
2763 			break;
2764 
2765 	if (!filter) {
2766 		err = -ENOENT;
2767 		goto out;
2768 	}
2769 
2770 	err = igb_erase_filter(adapter, filter);
2771 	if (err < 0)
2772 		goto out;
2773 
2774 	hlist_del(&filter->nfc_node);
2775 	kfree(filter);
2776 
2777 out:
2778 	spin_unlock(&adapter->nfc_lock);
2779 
2780 	return err;
2781 }
2782 
2783 static int igb_setup_tc_cls_flower(struct igb_adapter *adapter,
2784 				   struct flow_cls_offload *cls_flower)
2785 {
2786 	switch (cls_flower->command) {
2787 	case FLOW_CLS_REPLACE:
2788 		return igb_configure_clsflower(adapter, cls_flower);
2789 	case FLOW_CLS_DESTROY:
2790 		return igb_delete_clsflower(adapter, cls_flower);
2791 	case FLOW_CLS_STATS:
2792 		return -EOPNOTSUPP;
2793 	default:
2794 		return -EOPNOTSUPP;
2795 	}
2796 }
2797 
2798 static int igb_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
2799 				 void *cb_priv)
2800 {
2801 	struct igb_adapter *adapter = cb_priv;
2802 
2803 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
2804 		return -EOPNOTSUPP;
2805 
2806 	switch (type) {
2807 	case TC_SETUP_CLSFLOWER:
2808 		return igb_setup_tc_cls_flower(adapter, type_data);
2809 
2810 	default:
2811 		return -EOPNOTSUPP;
2812 	}
2813 }
2814 
2815 static int igb_offload_txtime(struct igb_adapter *adapter,
2816 			      struct tc_etf_qopt_offload *qopt)
2817 {
2818 	struct e1000_hw *hw = &adapter->hw;
2819 	int err;
2820 
2821 	/* Launchtime offloading is only supported by i210 controller. */
2822 	if (hw->mac.type != e1000_i210)
2823 		return -EOPNOTSUPP;
2824 
2825 	/* Launchtime offloading is only supported by queues 0 and 1. */
2826 	if (qopt->queue < 0 || qopt->queue > 1)
2827 		return -EINVAL;
2828 
2829 	err = igb_save_txtime_params(adapter, qopt->queue, qopt->enable);
2830 	if (err)
2831 		return err;
2832 
2833 	igb_offload_apply(adapter, qopt->queue);
2834 
2835 	return 0;
2836 }
2837 
2838 static int igb_tc_query_caps(struct igb_adapter *adapter,
2839 			     struct tc_query_caps_base *base)
2840 {
2841 	switch (base->type) {
2842 	case TC_SETUP_QDISC_TAPRIO: {
2843 		struct tc_taprio_caps *caps = base->caps;
2844 
2845 		caps->broken_mqprio = true;
2846 
2847 		return 0;
2848 	}
2849 	default:
2850 		return -EOPNOTSUPP;
2851 	}
2852 }
2853 
2854 static LIST_HEAD(igb_block_cb_list);
2855 
2856 static int igb_setup_tc(struct net_device *dev, enum tc_setup_type type,
2857 			void *type_data)
2858 {
2859 	struct igb_adapter *adapter = netdev_priv(dev);
2860 
2861 	switch (type) {
2862 	case TC_QUERY_CAPS:
2863 		return igb_tc_query_caps(adapter, type_data);
2864 	case TC_SETUP_QDISC_CBS:
2865 		return igb_offload_cbs(adapter, type_data);
2866 	case TC_SETUP_BLOCK:
2867 		return flow_block_cb_setup_simple(type_data,
2868 						  &igb_block_cb_list,
2869 						  igb_setup_tc_block_cb,
2870 						  adapter, adapter, true);
2871 
2872 	case TC_SETUP_QDISC_ETF:
2873 		return igb_offload_txtime(adapter, type_data);
2874 
2875 	default:
2876 		return -EOPNOTSUPP;
2877 	}
2878 }
2879 
2880 static int igb_xdp_setup(struct net_device *dev, struct netdev_bpf *bpf)
2881 {
2882 	int i, frame_size = dev->mtu + IGB_ETH_PKT_HDR_PAD;
2883 	struct igb_adapter *adapter = netdev_priv(dev);
2884 	struct bpf_prog *prog = bpf->prog, *old_prog;
2885 	bool running = netif_running(dev);
2886 	bool need_reset;
2887 
2888 	/* verify igb ring attributes are sufficient for XDP */
2889 	for (i = 0; i < adapter->num_rx_queues; i++) {
2890 		struct igb_ring *ring = adapter->rx_ring[i];
2891 
2892 		if (frame_size > igb_rx_bufsz(ring)) {
2893 			NL_SET_ERR_MSG_MOD(bpf->extack,
2894 					   "The RX buffer size is too small for the frame size");
2895 			netdev_warn(dev, "XDP RX buffer size %d is too small for the frame size %d\n",
2896 				    igb_rx_bufsz(ring), frame_size);
2897 			return -EINVAL;
2898 		}
2899 	}
2900 
2901 	old_prog = xchg(&adapter->xdp_prog, prog);
2902 	need_reset = (!!prog != !!old_prog);
2903 
2904 	/* device is up and bpf is added/removed, must setup the RX queues */
2905 	if (need_reset && running) {
2906 		igb_close(dev);
2907 	} else {
2908 		for (i = 0; i < adapter->num_rx_queues; i++)
2909 			(void)xchg(&adapter->rx_ring[i]->xdp_prog,
2910 			    adapter->xdp_prog);
2911 	}
2912 
2913 	if (old_prog)
2914 		bpf_prog_put(old_prog);
2915 
2916 	/* bpf is just replaced, RXQ and MTU are already setup */
2917 	if (!need_reset) {
2918 		return 0;
2919 	} else {
2920 		if (prog)
2921 			xdp_features_set_redirect_target(dev, true);
2922 		else
2923 			xdp_features_clear_redirect_target(dev);
2924 	}
2925 
2926 	if (running)
2927 		igb_open(dev);
2928 
2929 	return 0;
2930 }
2931 
2932 static int igb_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2933 {
2934 	switch (xdp->command) {
2935 	case XDP_SETUP_PROG:
2936 		return igb_xdp_setup(dev, xdp);
2937 	default:
2938 		return -EINVAL;
2939 	}
2940 }
2941 
2942 static void igb_xdp_ring_update_tail(struct igb_ring *ring)
2943 {
2944 	/* Force memory writes to complete before letting h/w know there
2945 	 * are new descriptors to fetch.
2946 	 */
2947 	wmb();
2948 	writel(ring->next_to_use, ring->tail);
2949 }
2950 
2951 static struct igb_ring *igb_xdp_tx_queue_mapping(struct igb_adapter *adapter)
2952 {
2953 	unsigned int r_idx = smp_processor_id();
2954 
2955 	if (r_idx >= adapter->num_tx_queues)
2956 		r_idx = r_idx % adapter->num_tx_queues;
2957 
2958 	return adapter->tx_ring[r_idx];
2959 }
2960 
2961 static int igb_xdp_xmit_back(struct igb_adapter *adapter, struct xdp_buff *xdp)
2962 {
2963 	struct xdp_frame *xdpf = xdp_convert_buff_to_frame(xdp);
2964 	int cpu = smp_processor_id();
2965 	struct igb_ring *tx_ring;
2966 	struct netdev_queue *nq;
2967 	u32 ret;
2968 
2969 	if (unlikely(!xdpf))
2970 		return IGB_XDP_CONSUMED;
2971 
2972 	/* During program transitions its possible adapter->xdp_prog is assigned
2973 	 * but ring has not been configured yet. In this case simply abort xmit.
2974 	 */
2975 	tx_ring = adapter->xdp_prog ? igb_xdp_tx_queue_mapping(adapter) : NULL;
2976 	if (unlikely(!tx_ring))
2977 		return IGB_XDP_CONSUMED;
2978 
2979 	nq = txring_txq(tx_ring);
2980 	__netif_tx_lock(nq, cpu);
2981 	/* Avoid transmit queue timeout since we share it with the slow path */
2982 	txq_trans_cond_update(nq);
2983 	ret = igb_xmit_xdp_ring(adapter, tx_ring, xdpf);
2984 	__netif_tx_unlock(nq);
2985 
2986 	return ret;
2987 }
2988 
2989 static int igb_xdp_xmit(struct net_device *dev, int n,
2990 			struct xdp_frame **frames, u32 flags)
2991 {
2992 	struct igb_adapter *adapter = netdev_priv(dev);
2993 	int cpu = smp_processor_id();
2994 	struct igb_ring *tx_ring;
2995 	struct netdev_queue *nq;
2996 	int nxmit = 0;
2997 	int i;
2998 
2999 	if (unlikely(test_bit(__IGB_DOWN, &adapter->state)))
3000 		return -ENETDOWN;
3001 
3002 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
3003 		return -EINVAL;
3004 
3005 	/* During program transitions its possible adapter->xdp_prog is assigned
3006 	 * but ring has not been configured yet. In this case simply abort xmit.
3007 	 */
3008 	tx_ring = adapter->xdp_prog ? igb_xdp_tx_queue_mapping(adapter) : NULL;
3009 	if (unlikely(!tx_ring))
3010 		return -ENXIO;
3011 
3012 	nq = txring_txq(tx_ring);
3013 	__netif_tx_lock(nq, cpu);
3014 
3015 	/* Avoid transmit queue timeout since we share it with the slow path */
3016 	txq_trans_cond_update(nq);
3017 
3018 	for (i = 0; i < n; i++) {
3019 		struct xdp_frame *xdpf = frames[i];
3020 		int err;
3021 
3022 		err = igb_xmit_xdp_ring(adapter, tx_ring, xdpf);
3023 		if (err != IGB_XDP_TX)
3024 			break;
3025 		nxmit++;
3026 	}
3027 
3028 	__netif_tx_unlock(nq);
3029 
3030 	if (unlikely(flags & XDP_XMIT_FLUSH))
3031 		igb_xdp_ring_update_tail(tx_ring);
3032 
3033 	return nxmit;
3034 }
3035 
3036 static const struct net_device_ops igb_netdev_ops = {
3037 	.ndo_open		= igb_open,
3038 	.ndo_stop		= igb_close,
3039 	.ndo_start_xmit		= igb_xmit_frame,
3040 	.ndo_get_stats64	= igb_get_stats64,
3041 	.ndo_set_rx_mode	= igb_set_rx_mode,
3042 	.ndo_set_mac_address	= igb_set_mac,
3043 	.ndo_change_mtu		= igb_change_mtu,
3044 	.ndo_eth_ioctl		= igb_ioctl,
3045 	.ndo_tx_timeout		= igb_tx_timeout,
3046 	.ndo_validate_addr	= eth_validate_addr,
3047 	.ndo_vlan_rx_add_vid	= igb_vlan_rx_add_vid,
3048 	.ndo_vlan_rx_kill_vid	= igb_vlan_rx_kill_vid,
3049 	.ndo_set_vf_mac		= igb_ndo_set_vf_mac,
3050 	.ndo_set_vf_vlan	= igb_ndo_set_vf_vlan,
3051 	.ndo_set_vf_rate	= igb_ndo_set_vf_bw,
3052 	.ndo_set_vf_spoofchk	= igb_ndo_set_vf_spoofchk,
3053 	.ndo_set_vf_trust	= igb_ndo_set_vf_trust,
3054 	.ndo_get_vf_config	= igb_ndo_get_vf_config,
3055 	.ndo_fix_features	= igb_fix_features,
3056 	.ndo_set_features	= igb_set_features,
3057 	.ndo_fdb_add		= igb_ndo_fdb_add,
3058 	.ndo_features_check	= igb_features_check,
3059 	.ndo_setup_tc		= igb_setup_tc,
3060 	.ndo_bpf		= igb_xdp,
3061 	.ndo_xdp_xmit		= igb_xdp_xmit,
3062 };
3063 
3064 /**
3065  * igb_set_fw_version - Configure version string for ethtool
3066  * @adapter: adapter struct
3067  **/
3068 void igb_set_fw_version(struct igb_adapter *adapter)
3069 {
3070 	struct e1000_hw *hw = &adapter->hw;
3071 	struct e1000_fw_version fw;
3072 	char *lbuf;
3073 
3074 	igb_get_fw_version(hw, &fw);
3075 
3076 	switch (hw->mac.type) {
3077 	case e1000_i210:
3078 	case e1000_i211:
3079 		if (!(igb_get_flash_presence_i210(hw))) {
3080 			lbuf = kasprintf(GFP_KERNEL, "%2d.%2d-%d",
3081 					 fw.invm_major, fw.invm_minor,
3082 					 fw.invm_img_type);
3083 			break;
3084 		}
3085 		fallthrough;
3086 	default:
3087 		/* if option rom is valid, display its version too */
3088 		if (fw.or_valid) {
3089 			lbuf = kasprintf(GFP_KERNEL, "%d.%d, 0x%08x, %d.%d.%d",
3090 					 fw.eep_major, fw.eep_minor,
3091 					 fw.etrack_id, fw.or_major, fw.or_build,
3092 					 fw.or_patch);
3093 		/* no option rom */
3094 		} else if (fw.etrack_id != 0X0000) {
3095 			lbuf = kasprintf(GFP_KERNEL, "%d.%d, 0x%08x",
3096 					 fw.eep_major, fw.eep_minor,
3097 					 fw.etrack_id);
3098 		} else {
3099 			lbuf = kasprintf(GFP_KERNEL, "%d.%d.%d", fw.eep_major,
3100 					 fw.eep_minor, fw.eep_build);
3101 		}
3102 		break;
3103 	}
3104 
3105 	/* the truncate happens here if it doesn't fit */
3106 	strscpy(adapter->fw_version, lbuf, sizeof(adapter->fw_version));
3107 	kfree(lbuf);
3108 }
3109 
3110 /**
3111  * igb_init_mas - init Media Autosense feature if enabled in the NVM
3112  *
3113  * @adapter: adapter struct
3114  **/
3115 static void igb_init_mas(struct igb_adapter *adapter)
3116 {
3117 	struct e1000_hw *hw = &adapter->hw;
3118 	u16 eeprom_data;
3119 
3120 	hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
3121 	switch (hw->bus.func) {
3122 	case E1000_FUNC_0:
3123 		if (eeprom_data & IGB_MAS_ENABLE_0) {
3124 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3125 			netdev_info(adapter->netdev,
3126 				"MAS: Enabling Media Autosense for port %d\n",
3127 				hw->bus.func);
3128 		}
3129 		break;
3130 	case E1000_FUNC_1:
3131 		if (eeprom_data & IGB_MAS_ENABLE_1) {
3132 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3133 			netdev_info(adapter->netdev,
3134 				"MAS: Enabling Media Autosense for port %d\n",
3135 				hw->bus.func);
3136 		}
3137 		break;
3138 	case E1000_FUNC_2:
3139 		if (eeprom_data & IGB_MAS_ENABLE_2) {
3140 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3141 			netdev_info(adapter->netdev,
3142 				"MAS: Enabling Media Autosense for port %d\n",
3143 				hw->bus.func);
3144 		}
3145 		break;
3146 	case E1000_FUNC_3:
3147 		if (eeprom_data & IGB_MAS_ENABLE_3) {
3148 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3149 			netdev_info(adapter->netdev,
3150 				"MAS: Enabling Media Autosense for port %d\n",
3151 				hw->bus.func);
3152 		}
3153 		break;
3154 	default:
3155 		/* Shouldn't get here */
3156 		netdev_err(adapter->netdev,
3157 			"MAS: Invalid port configuration, returning\n");
3158 		break;
3159 	}
3160 }
3161 
3162 /**
3163  *  igb_init_i2c - Init I2C interface
3164  *  @adapter: pointer to adapter structure
3165  **/
3166 static s32 igb_init_i2c(struct igb_adapter *adapter)
3167 {
3168 	s32 status = 0;
3169 
3170 	/* I2C interface supported on i350 devices */
3171 	if (adapter->hw.mac.type != e1000_i350)
3172 		return 0;
3173 
3174 	/* Initialize the i2c bus which is controlled by the registers.
3175 	 * This bus will use the i2c_algo_bit structure that implements
3176 	 * the protocol through toggling of the 4 bits in the register.
3177 	 */
3178 	adapter->i2c_adap.owner = THIS_MODULE;
3179 	adapter->i2c_algo = igb_i2c_algo;
3180 	adapter->i2c_algo.data = adapter;
3181 	adapter->i2c_adap.algo_data = &adapter->i2c_algo;
3182 	adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
3183 	strscpy(adapter->i2c_adap.name, "igb BB",
3184 		sizeof(adapter->i2c_adap.name));
3185 	status = i2c_bit_add_bus(&adapter->i2c_adap);
3186 	return status;
3187 }
3188 
3189 /**
3190  *  igb_probe - Device Initialization Routine
3191  *  @pdev: PCI device information struct
3192  *  @ent: entry in igb_pci_tbl
3193  *
3194  *  Returns 0 on success, negative on failure
3195  *
3196  *  igb_probe initializes an adapter identified by a pci_dev structure.
3197  *  The OS initialization, configuring of the adapter private structure,
3198  *  and a hardware reset occur.
3199  **/
3200 static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
3201 {
3202 	struct net_device *netdev;
3203 	struct igb_adapter *adapter;
3204 	struct e1000_hw *hw;
3205 	u16 eeprom_data = 0;
3206 	s32 ret_val;
3207 	static int global_quad_port_a; /* global quad port a indication */
3208 	const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
3209 	u8 part_str[E1000_PBANUM_LENGTH];
3210 	int err;
3211 
3212 	/* Catch broken hardware that put the wrong VF device ID in
3213 	 * the PCIe SR-IOV capability.
3214 	 */
3215 	if (pdev->is_virtfn) {
3216 		WARN(1, KERN_ERR "%s (%x:%x) should not be a VF!\n",
3217 			pci_name(pdev), pdev->vendor, pdev->device);
3218 		return -EINVAL;
3219 	}
3220 
3221 	err = pci_enable_device_mem(pdev);
3222 	if (err)
3223 		return err;
3224 
3225 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
3226 	if (err) {
3227 		dev_err(&pdev->dev,
3228 			"No usable DMA configuration, aborting\n");
3229 		goto err_dma;
3230 	}
3231 
3232 	err = pci_request_mem_regions(pdev, igb_driver_name);
3233 	if (err)
3234 		goto err_pci_reg;
3235 
3236 	pci_set_master(pdev);
3237 	pci_save_state(pdev);
3238 
3239 	err = -ENOMEM;
3240 	netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
3241 				   IGB_MAX_TX_QUEUES);
3242 	if (!netdev)
3243 		goto err_alloc_etherdev;
3244 
3245 	SET_NETDEV_DEV(netdev, &pdev->dev);
3246 
3247 	pci_set_drvdata(pdev, netdev);
3248 	adapter = netdev_priv(netdev);
3249 	adapter->netdev = netdev;
3250 	adapter->pdev = pdev;
3251 	hw = &adapter->hw;
3252 	hw->back = adapter;
3253 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
3254 
3255 	err = -EIO;
3256 	adapter->io_addr = pci_iomap(pdev, 0, 0);
3257 	if (!adapter->io_addr)
3258 		goto err_ioremap;
3259 	/* hw->hw_addr can be altered, we'll use adapter->io_addr for unmap */
3260 	hw->hw_addr = adapter->io_addr;
3261 
3262 	netdev->netdev_ops = &igb_netdev_ops;
3263 	igb_set_ethtool_ops(netdev);
3264 	netdev->watchdog_timeo = 5 * HZ;
3265 
3266 	strscpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
3267 
3268 	netdev->mem_start = pci_resource_start(pdev, 0);
3269 	netdev->mem_end = pci_resource_end(pdev, 0);
3270 
3271 	/* PCI config space info */
3272 	hw->vendor_id = pdev->vendor;
3273 	hw->device_id = pdev->device;
3274 	hw->revision_id = pdev->revision;
3275 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
3276 	hw->subsystem_device_id = pdev->subsystem_device;
3277 
3278 	/* Copy the default MAC, PHY and NVM function pointers */
3279 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
3280 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
3281 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
3282 	/* Initialize skew-specific constants */
3283 	err = ei->get_invariants(hw);
3284 	if (err)
3285 		goto err_sw_init;
3286 
3287 	/* setup the private structure */
3288 	err = igb_sw_init(adapter);
3289 	if (err)
3290 		goto err_sw_init;
3291 
3292 	igb_get_bus_info_pcie(hw);
3293 
3294 	hw->phy.autoneg_wait_to_complete = false;
3295 
3296 	/* Copper options */
3297 	if (hw->phy.media_type == e1000_media_type_copper) {
3298 		hw->phy.mdix = AUTO_ALL_MODES;
3299 		hw->phy.disable_polarity_correction = false;
3300 		hw->phy.ms_type = e1000_ms_hw_default;
3301 	}
3302 
3303 	if (igb_check_reset_block(hw))
3304 		dev_info(&pdev->dev,
3305 			"PHY reset is blocked due to SOL/IDER session.\n");
3306 
3307 	/* features is initialized to 0 in allocation, it might have bits
3308 	 * set by igb_sw_init so we should use an or instead of an
3309 	 * assignment.
3310 	 */
3311 	netdev->features |= NETIF_F_SG |
3312 			    NETIF_F_TSO |
3313 			    NETIF_F_TSO6 |
3314 			    NETIF_F_RXHASH |
3315 			    NETIF_F_RXCSUM |
3316 			    NETIF_F_HW_CSUM;
3317 
3318 	if (hw->mac.type >= e1000_82576)
3319 		netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_GSO_UDP_L4;
3320 
3321 	if (hw->mac.type >= e1000_i350)
3322 		netdev->features |= NETIF_F_HW_TC;
3323 
3324 #define IGB_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
3325 				  NETIF_F_GSO_GRE_CSUM | \
3326 				  NETIF_F_GSO_IPXIP4 | \
3327 				  NETIF_F_GSO_IPXIP6 | \
3328 				  NETIF_F_GSO_UDP_TUNNEL | \
3329 				  NETIF_F_GSO_UDP_TUNNEL_CSUM)
3330 
3331 	netdev->gso_partial_features = IGB_GSO_PARTIAL_FEATURES;
3332 	netdev->features |= NETIF_F_GSO_PARTIAL | IGB_GSO_PARTIAL_FEATURES;
3333 
3334 	/* copy netdev features into list of user selectable features */
3335 	netdev->hw_features |= netdev->features |
3336 			       NETIF_F_HW_VLAN_CTAG_RX |
3337 			       NETIF_F_HW_VLAN_CTAG_TX |
3338 			       NETIF_F_RXALL;
3339 
3340 	if (hw->mac.type >= e1000_i350)
3341 		netdev->hw_features |= NETIF_F_NTUPLE;
3342 
3343 	netdev->features |= NETIF_F_HIGHDMA;
3344 
3345 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
3346 	netdev->mpls_features |= NETIF_F_HW_CSUM;
3347 	netdev->hw_enc_features |= netdev->vlan_features;
3348 
3349 	/* set this bit last since it cannot be part of vlan_features */
3350 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
3351 			    NETIF_F_HW_VLAN_CTAG_RX |
3352 			    NETIF_F_HW_VLAN_CTAG_TX;
3353 
3354 	netdev->priv_flags |= IFF_SUPP_NOFCS;
3355 
3356 	netdev->priv_flags |= IFF_UNICAST_FLT;
3357 	netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT;
3358 
3359 	/* MTU range: 68 - 9216 */
3360 	netdev->min_mtu = ETH_MIN_MTU;
3361 	netdev->max_mtu = MAX_STD_JUMBO_FRAME_SIZE;
3362 
3363 	adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
3364 
3365 	/* before reading the NVM, reset the controller to put the device in a
3366 	 * known good starting state
3367 	 */
3368 	hw->mac.ops.reset_hw(hw);
3369 
3370 	/* make sure the NVM is good , i211/i210 parts can have special NVM
3371 	 * that doesn't contain a checksum
3372 	 */
3373 	switch (hw->mac.type) {
3374 	case e1000_i210:
3375 	case e1000_i211:
3376 		if (igb_get_flash_presence_i210(hw)) {
3377 			if (hw->nvm.ops.validate(hw) < 0) {
3378 				dev_err(&pdev->dev,
3379 					"The NVM Checksum Is Not Valid\n");
3380 				err = -EIO;
3381 				goto err_eeprom;
3382 			}
3383 		}
3384 		break;
3385 	default:
3386 		if (hw->nvm.ops.validate(hw) < 0) {
3387 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
3388 			err = -EIO;
3389 			goto err_eeprom;
3390 		}
3391 		break;
3392 	}
3393 
3394 	if (eth_platform_get_mac_address(&pdev->dev, hw->mac.addr)) {
3395 		/* copy the MAC address out of the NVM */
3396 		if (hw->mac.ops.read_mac_addr(hw))
3397 			dev_err(&pdev->dev, "NVM Read Error\n");
3398 	}
3399 
3400 	eth_hw_addr_set(netdev, hw->mac.addr);
3401 
3402 	if (!is_valid_ether_addr(netdev->dev_addr)) {
3403 		dev_err(&pdev->dev, "Invalid MAC Address\n");
3404 		err = -EIO;
3405 		goto err_eeprom;
3406 	}
3407 
3408 	igb_set_default_mac_filter(adapter);
3409 
3410 	/* get firmware version for ethtool -i */
3411 	igb_set_fw_version(adapter);
3412 
3413 	/* configure RXPBSIZE and TXPBSIZE */
3414 	if (hw->mac.type == e1000_i210) {
3415 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
3416 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
3417 	}
3418 
3419 	timer_setup(&adapter->watchdog_timer, igb_watchdog, 0);
3420 	timer_setup(&adapter->phy_info_timer, igb_update_phy_info, 0);
3421 
3422 	INIT_WORK(&adapter->reset_task, igb_reset_task);
3423 	INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
3424 
3425 	/* Initialize link properties that are user-changeable */
3426 	adapter->fc_autoneg = true;
3427 	hw->mac.autoneg = true;
3428 	hw->phy.autoneg_advertised = 0x2f;
3429 
3430 	hw->fc.requested_mode = e1000_fc_default;
3431 	hw->fc.current_mode = e1000_fc_default;
3432 
3433 	igb_validate_mdi_setting(hw);
3434 
3435 	/* By default, support wake on port A */
3436 	if (hw->bus.func == 0)
3437 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3438 
3439 	/* Check the NVM for wake support on non-port A ports */
3440 	if (hw->mac.type >= e1000_82580)
3441 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
3442 				 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
3443 				 &eeprom_data);
3444 	else if (hw->bus.func == 1)
3445 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3446 
3447 	if (eeprom_data & IGB_EEPROM_APME)
3448 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3449 
3450 	/* now that we have the eeprom settings, apply the special cases where
3451 	 * the eeprom may be wrong or the board simply won't support wake on
3452 	 * lan on a particular port
3453 	 */
3454 	switch (pdev->device) {
3455 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
3456 		adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3457 		break;
3458 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
3459 	case E1000_DEV_ID_82576_FIBER:
3460 	case E1000_DEV_ID_82576_SERDES:
3461 		/* Wake events only supported on port A for dual fiber
3462 		 * regardless of eeprom setting
3463 		 */
3464 		if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
3465 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3466 		break;
3467 	case E1000_DEV_ID_82576_QUAD_COPPER:
3468 	case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
3469 		/* if quad port adapter, disable WoL on all but port A */
3470 		if (global_quad_port_a != 0)
3471 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3472 		else
3473 			adapter->flags |= IGB_FLAG_QUAD_PORT_A;
3474 		/* Reset for multiple quad port adapters */
3475 		if (++global_quad_port_a == 4)
3476 			global_quad_port_a = 0;
3477 		break;
3478 	default:
3479 		/* If the device can't wake, don't set software support */
3480 		if (!device_can_wakeup(&adapter->pdev->dev))
3481 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3482 	}
3483 
3484 	/* initialize the wol settings based on the eeprom settings */
3485 	if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
3486 		adapter->wol |= E1000_WUFC_MAG;
3487 
3488 	/* Some vendors want WoL disabled by default, but still supported */
3489 	if ((hw->mac.type == e1000_i350) &&
3490 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
3491 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3492 		adapter->wol = 0;
3493 	}
3494 
3495 	/* Some vendors want the ability to Use the EEPROM setting as
3496 	 * enable/disable only, and not for capability
3497 	 */
3498 	if (((hw->mac.type == e1000_i350) ||
3499 	     (hw->mac.type == e1000_i354)) &&
3500 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_DELL)) {
3501 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3502 		adapter->wol = 0;
3503 	}
3504 	if (hw->mac.type == e1000_i350) {
3505 		if (((pdev->subsystem_device == 0x5001) ||
3506 		     (pdev->subsystem_device == 0x5002)) &&
3507 				(hw->bus.func == 0)) {
3508 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3509 			adapter->wol = 0;
3510 		}
3511 		if (pdev->subsystem_device == 0x1F52)
3512 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3513 	}
3514 
3515 	device_set_wakeup_enable(&adapter->pdev->dev,
3516 				 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
3517 
3518 	/* reset the hardware with the new settings */
3519 	igb_reset(adapter);
3520 
3521 	/* Init the I2C interface */
3522 	err = igb_init_i2c(adapter);
3523 	if (err) {
3524 		dev_err(&pdev->dev, "failed to init i2c interface\n");
3525 		goto err_eeprom;
3526 	}
3527 
3528 	/* let the f/w know that the h/w is now under the control of the
3529 	 * driver.
3530 	 */
3531 	igb_get_hw_control(adapter);
3532 
3533 	strcpy(netdev->name, "eth%d");
3534 	err = register_netdev(netdev);
3535 	if (err)
3536 		goto err_register;
3537 
3538 	/* carrier off reporting is important to ethtool even BEFORE open */
3539 	netif_carrier_off(netdev);
3540 
3541 #ifdef CONFIG_IGB_DCA
3542 	if (dca_add_requester(&pdev->dev) == 0) {
3543 		adapter->flags |= IGB_FLAG_DCA_ENABLED;
3544 		dev_info(&pdev->dev, "DCA enabled\n");
3545 		igb_setup_dca(adapter);
3546 	}
3547 
3548 #endif
3549 #ifdef CONFIG_IGB_HWMON
3550 	/* Initialize the thermal sensor on i350 devices. */
3551 	if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
3552 		u16 ets_word;
3553 
3554 		/* Read the NVM to determine if this i350 device supports an
3555 		 * external thermal sensor.
3556 		 */
3557 		hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
3558 		if (ets_word != 0x0000 && ets_word != 0xFFFF)
3559 			adapter->ets = true;
3560 		else
3561 			adapter->ets = false;
3562 		/* Only enable I2C bit banging if an external thermal
3563 		 * sensor is supported.
3564 		 */
3565 		if (adapter->ets)
3566 			igb_set_i2c_bb(hw);
3567 		hw->mac.ops.init_thermal_sensor_thresh(hw);
3568 		if (igb_sysfs_init(adapter))
3569 			dev_err(&pdev->dev,
3570 				"failed to allocate sysfs resources\n");
3571 	} else {
3572 		adapter->ets = false;
3573 	}
3574 #endif
3575 	/* Check if Media Autosense is enabled */
3576 	adapter->ei = *ei;
3577 	if (hw->dev_spec._82575.mas_capable)
3578 		igb_init_mas(adapter);
3579 
3580 	/* do hw tstamp init after resetting */
3581 	igb_ptp_init(adapter);
3582 
3583 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
3584 	/* print bus type/speed/width info, not applicable to i354 */
3585 	if (hw->mac.type != e1000_i354) {
3586 		dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
3587 			 netdev->name,
3588 			 ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
3589 			  (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
3590 			   "unknown"),
3591 			 ((hw->bus.width == e1000_bus_width_pcie_x4) ?
3592 			  "Width x4" :
3593 			  (hw->bus.width == e1000_bus_width_pcie_x2) ?
3594 			  "Width x2" :
3595 			  (hw->bus.width == e1000_bus_width_pcie_x1) ?
3596 			  "Width x1" : "unknown"), netdev->dev_addr);
3597 	}
3598 
3599 	if ((hw->mac.type == e1000_82576 &&
3600 	     rd32(E1000_EECD) & E1000_EECD_PRES) ||
3601 	    (hw->mac.type >= e1000_i210 ||
3602 	     igb_get_flash_presence_i210(hw))) {
3603 		ret_val = igb_read_part_string(hw, part_str,
3604 					       E1000_PBANUM_LENGTH);
3605 	} else {
3606 		ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
3607 	}
3608 
3609 	if (ret_val)
3610 		strcpy(part_str, "Unknown");
3611 	dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
3612 	dev_info(&pdev->dev,
3613 		"Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
3614 		(adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
3615 		(adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
3616 		adapter->num_rx_queues, adapter->num_tx_queues);
3617 	if (hw->phy.media_type == e1000_media_type_copper) {
3618 		switch (hw->mac.type) {
3619 		case e1000_i350:
3620 		case e1000_i210:
3621 		case e1000_i211:
3622 			/* Enable EEE for internal copper PHY devices */
3623 			err = igb_set_eee_i350(hw, true, true);
3624 			if ((!err) &&
3625 			    (!hw->dev_spec._82575.eee_disable)) {
3626 				adapter->eee_advert =
3627 					MDIO_EEE_100TX | MDIO_EEE_1000T;
3628 				adapter->flags |= IGB_FLAG_EEE;
3629 			}
3630 			break;
3631 		case e1000_i354:
3632 			if ((rd32(E1000_CTRL_EXT) &
3633 			    E1000_CTRL_EXT_LINK_MODE_SGMII)) {
3634 				err = igb_set_eee_i354(hw, true, true);
3635 				if ((!err) &&
3636 					(!hw->dev_spec._82575.eee_disable)) {
3637 					adapter->eee_advert =
3638 					   MDIO_EEE_100TX | MDIO_EEE_1000T;
3639 					adapter->flags |= IGB_FLAG_EEE;
3640 				}
3641 			}
3642 			break;
3643 		default:
3644 			break;
3645 		}
3646 	}
3647 
3648 	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_NO_DIRECT_COMPLETE);
3649 
3650 	pm_runtime_put_noidle(&pdev->dev);
3651 	return 0;
3652 
3653 err_register:
3654 	igb_release_hw_control(adapter);
3655 	memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
3656 err_eeprom:
3657 	if (!igb_check_reset_block(hw))
3658 		igb_reset_phy(hw);
3659 
3660 	if (hw->flash_address)
3661 		iounmap(hw->flash_address);
3662 err_sw_init:
3663 	kfree(adapter->mac_table);
3664 	kfree(adapter->shadow_vfta);
3665 	igb_clear_interrupt_scheme(adapter);
3666 #ifdef CONFIG_PCI_IOV
3667 	igb_disable_sriov(pdev, false);
3668 #endif
3669 	pci_iounmap(pdev, adapter->io_addr);
3670 err_ioremap:
3671 	free_netdev(netdev);
3672 err_alloc_etherdev:
3673 	pci_release_mem_regions(pdev);
3674 err_pci_reg:
3675 err_dma:
3676 	pci_disable_device(pdev);
3677 	return err;
3678 }
3679 
3680 #ifdef CONFIG_PCI_IOV
3681 static int igb_sriov_reinit(struct pci_dev *dev)
3682 {
3683 	struct net_device *netdev = pci_get_drvdata(dev);
3684 	struct igb_adapter *adapter = netdev_priv(netdev);
3685 	struct pci_dev *pdev = adapter->pdev;
3686 
3687 	rtnl_lock();
3688 
3689 	if (netif_running(netdev))
3690 		igb_close(netdev);
3691 	else
3692 		igb_reset(adapter);
3693 
3694 	igb_clear_interrupt_scheme(adapter);
3695 
3696 	igb_init_queue_configuration(adapter);
3697 
3698 	if (igb_init_interrupt_scheme(adapter, true)) {
3699 		rtnl_unlock();
3700 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
3701 		return -ENOMEM;
3702 	}
3703 
3704 	if (netif_running(netdev))
3705 		igb_open(netdev);
3706 
3707 	rtnl_unlock();
3708 
3709 	return 0;
3710 }
3711 
3712 static int igb_disable_sriov(struct pci_dev *pdev, bool reinit)
3713 {
3714 	struct net_device *netdev = pci_get_drvdata(pdev);
3715 	struct igb_adapter *adapter = netdev_priv(netdev);
3716 	struct e1000_hw *hw = &adapter->hw;
3717 	unsigned long flags;
3718 
3719 	/* reclaim resources allocated to VFs */
3720 	if (adapter->vf_data) {
3721 		/* disable iov and allow time for transactions to clear */
3722 		if (pci_vfs_assigned(pdev)) {
3723 			dev_warn(&pdev->dev,
3724 				 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
3725 			return -EPERM;
3726 		} else {
3727 			pci_disable_sriov(pdev);
3728 			msleep(500);
3729 		}
3730 		spin_lock_irqsave(&adapter->vfs_lock, flags);
3731 		kfree(adapter->vf_mac_list);
3732 		adapter->vf_mac_list = NULL;
3733 		kfree(adapter->vf_data);
3734 		adapter->vf_data = NULL;
3735 		adapter->vfs_allocated_count = 0;
3736 		spin_unlock_irqrestore(&adapter->vfs_lock, flags);
3737 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
3738 		wrfl();
3739 		msleep(100);
3740 		dev_info(&pdev->dev, "IOV Disabled\n");
3741 
3742 		/* Re-enable DMA Coalescing flag since IOV is turned off */
3743 		adapter->flags |= IGB_FLAG_DMAC;
3744 	}
3745 
3746 	return reinit ? igb_sriov_reinit(pdev) : 0;
3747 }
3748 
3749 static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs, bool reinit)
3750 {
3751 	struct net_device *netdev = pci_get_drvdata(pdev);
3752 	struct igb_adapter *adapter = netdev_priv(netdev);
3753 	int old_vfs = pci_num_vf(pdev);
3754 	struct vf_mac_filter *mac_list;
3755 	int err = 0;
3756 	int num_vf_mac_filters, i;
3757 
3758 	if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
3759 		err = -EPERM;
3760 		goto out;
3761 	}
3762 	if (!num_vfs)
3763 		goto out;
3764 
3765 	if (old_vfs) {
3766 		dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
3767 			 old_vfs, max_vfs);
3768 		adapter->vfs_allocated_count = old_vfs;
3769 	} else
3770 		adapter->vfs_allocated_count = num_vfs;
3771 
3772 	adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
3773 				sizeof(struct vf_data_storage), GFP_KERNEL);
3774 
3775 	/* if allocation failed then we do not support SR-IOV */
3776 	if (!adapter->vf_data) {
3777 		adapter->vfs_allocated_count = 0;
3778 		err = -ENOMEM;
3779 		goto out;
3780 	}
3781 
3782 	/* Due to the limited number of RAR entries calculate potential
3783 	 * number of MAC filters available for the VFs. Reserve entries
3784 	 * for PF default MAC, PF MAC filters and at least one RAR entry
3785 	 * for each VF for VF MAC.
3786 	 */
3787 	num_vf_mac_filters = adapter->hw.mac.rar_entry_count -
3788 			     (1 + IGB_PF_MAC_FILTERS_RESERVED +
3789 			      adapter->vfs_allocated_count);
3790 
3791 	adapter->vf_mac_list = kcalloc(num_vf_mac_filters,
3792 				       sizeof(struct vf_mac_filter),
3793 				       GFP_KERNEL);
3794 
3795 	mac_list = adapter->vf_mac_list;
3796 	INIT_LIST_HEAD(&adapter->vf_macs.l);
3797 
3798 	if (adapter->vf_mac_list) {
3799 		/* Initialize list of VF MAC filters */
3800 		for (i = 0; i < num_vf_mac_filters; i++) {
3801 			mac_list->vf = -1;
3802 			mac_list->free = true;
3803 			list_add(&mac_list->l, &adapter->vf_macs.l);
3804 			mac_list++;
3805 		}
3806 	} else {
3807 		/* If we could not allocate memory for the VF MAC filters
3808 		 * we can continue without this feature but warn user.
3809 		 */
3810 		dev_err(&pdev->dev,
3811 			"Unable to allocate memory for VF MAC filter list\n");
3812 	}
3813 
3814 	dev_info(&pdev->dev, "%d VFs allocated\n",
3815 		 adapter->vfs_allocated_count);
3816 	for (i = 0; i < adapter->vfs_allocated_count; i++)
3817 		igb_vf_configure(adapter, i);
3818 
3819 	/* DMA Coalescing is not supported in IOV mode. */
3820 	adapter->flags &= ~IGB_FLAG_DMAC;
3821 
3822 	if (reinit) {
3823 		err = igb_sriov_reinit(pdev);
3824 		if (err)
3825 			goto err_out;
3826 	}
3827 
3828 	/* only call pci_enable_sriov() if no VFs are allocated already */
3829 	if (!old_vfs) {
3830 		err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
3831 		if (err)
3832 			goto err_out;
3833 	}
3834 
3835 	goto out;
3836 
3837 err_out:
3838 	kfree(adapter->vf_mac_list);
3839 	adapter->vf_mac_list = NULL;
3840 	kfree(adapter->vf_data);
3841 	adapter->vf_data = NULL;
3842 	adapter->vfs_allocated_count = 0;
3843 out:
3844 	return err;
3845 }
3846 
3847 #endif
3848 /**
3849  *  igb_remove_i2c - Cleanup  I2C interface
3850  *  @adapter: pointer to adapter structure
3851  **/
3852 static void igb_remove_i2c(struct igb_adapter *adapter)
3853 {
3854 	/* free the adapter bus structure */
3855 	i2c_del_adapter(&adapter->i2c_adap);
3856 }
3857 
3858 /**
3859  *  igb_remove - Device Removal Routine
3860  *  @pdev: PCI device information struct
3861  *
3862  *  igb_remove is called by the PCI subsystem to alert the driver
3863  *  that it should release a PCI device.  The could be caused by a
3864  *  Hot-Plug event, or because the driver is going to be removed from
3865  *  memory.
3866  **/
3867 static void igb_remove(struct pci_dev *pdev)
3868 {
3869 	struct net_device *netdev = pci_get_drvdata(pdev);
3870 	struct igb_adapter *adapter = netdev_priv(netdev);
3871 	struct e1000_hw *hw = &adapter->hw;
3872 
3873 	pm_runtime_get_noresume(&pdev->dev);
3874 #ifdef CONFIG_IGB_HWMON
3875 	igb_sysfs_exit(adapter);
3876 #endif
3877 	igb_remove_i2c(adapter);
3878 	igb_ptp_stop(adapter);
3879 	/* The watchdog timer may be rescheduled, so explicitly
3880 	 * disable watchdog from being rescheduled.
3881 	 */
3882 	set_bit(__IGB_DOWN, &adapter->state);
3883 	del_timer_sync(&adapter->watchdog_timer);
3884 	del_timer_sync(&adapter->phy_info_timer);
3885 
3886 	cancel_work_sync(&adapter->reset_task);
3887 	cancel_work_sync(&adapter->watchdog_task);
3888 
3889 #ifdef CONFIG_IGB_DCA
3890 	if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
3891 		dev_info(&pdev->dev, "DCA disabled\n");
3892 		dca_remove_requester(&pdev->dev);
3893 		adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
3894 		wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
3895 	}
3896 #endif
3897 
3898 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
3899 	 * would have already happened in close and is redundant.
3900 	 */
3901 	igb_release_hw_control(adapter);
3902 
3903 #ifdef CONFIG_PCI_IOV
3904 	igb_disable_sriov(pdev, false);
3905 #endif
3906 
3907 	unregister_netdev(netdev);
3908 
3909 	igb_clear_interrupt_scheme(adapter);
3910 
3911 	pci_iounmap(pdev, adapter->io_addr);
3912 	if (hw->flash_address)
3913 		iounmap(hw->flash_address);
3914 	pci_release_mem_regions(pdev);
3915 
3916 	kfree(adapter->mac_table);
3917 	kfree(adapter->shadow_vfta);
3918 	free_netdev(netdev);
3919 
3920 	pci_disable_device(pdev);
3921 }
3922 
3923 /**
3924  *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
3925  *  @adapter: board private structure to initialize
3926  *
3927  *  This function initializes the vf specific data storage and then attempts to
3928  *  allocate the VFs.  The reason for ordering it this way is because it is much
3929  *  mor expensive time wise to disable SR-IOV than it is to allocate and free
3930  *  the memory for the VFs.
3931  **/
3932 static void igb_probe_vfs(struct igb_adapter *adapter)
3933 {
3934 #ifdef CONFIG_PCI_IOV
3935 	struct pci_dev *pdev = adapter->pdev;
3936 	struct e1000_hw *hw = &adapter->hw;
3937 
3938 	/* Virtualization features not supported on i210 and 82580 family. */
3939 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211) ||
3940 	    (hw->mac.type == e1000_82580))
3941 		return;
3942 
3943 	/* Of the below we really only want the effect of getting
3944 	 * IGB_FLAG_HAS_MSIX set (if available), without which
3945 	 * igb_enable_sriov() has no effect.
3946 	 */
3947 	igb_set_interrupt_capability(adapter, true);
3948 	igb_reset_interrupt_capability(adapter);
3949 
3950 	pci_sriov_set_totalvfs(pdev, 7);
3951 	igb_enable_sriov(pdev, max_vfs, false);
3952 
3953 #endif /* CONFIG_PCI_IOV */
3954 }
3955 
3956 unsigned int igb_get_max_rss_queues(struct igb_adapter *adapter)
3957 {
3958 	struct e1000_hw *hw = &adapter->hw;
3959 	unsigned int max_rss_queues;
3960 
3961 	/* Determine the maximum number of RSS queues supported. */
3962 	switch (hw->mac.type) {
3963 	case e1000_i211:
3964 		max_rss_queues = IGB_MAX_RX_QUEUES_I211;
3965 		break;
3966 	case e1000_82575:
3967 	case e1000_i210:
3968 		max_rss_queues = IGB_MAX_RX_QUEUES_82575;
3969 		break;
3970 	case e1000_i350:
3971 		/* I350 cannot do RSS and SR-IOV at the same time */
3972 		if (!!adapter->vfs_allocated_count) {
3973 			max_rss_queues = 1;
3974 			break;
3975 		}
3976 		fallthrough;
3977 	case e1000_82576:
3978 		if (!!adapter->vfs_allocated_count) {
3979 			max_rss_queues = 2;
3980 			break;
3981 		}
3982 		fallthrough;
3983 	case e1000_82580:
3984 	case e1000_i354:
3985 	default:
3986 		max_rss_queues = IGB_MAX_RX_QUEUES;
3987 		break;
3988 	}
3989 
3990 	return max_rss_queues;
3991 }
3992 
3993 static void igb_init_queue_configuration(struct igb_adapter *adapter)
3994 {
3995 	u32 max_rss_queues;
3996 
3997 	max_rss_queues = igb_get_max_rss_queues(adapter);
3998 	adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
3999 
4000 	igb_set_flag_queue_pairs(adapter, max_rss_queues);
4001 }
4002 
4003 void igb_set_flag_queue_pairs(struct igb_adapter *adapter,
4004 			      const u32 max_rss_queues)
4005 {
4006 	struct e1000_hw *hw = &adapter->hw;
4007 
4008 	/* Determine if we need to pair queues. */
4009 	switch (hw->mac.type) {
4010 	case e1000_82575:
4011 	case e1000_i211:
4012 		/* Device supports enough interrupts without queue pairing. */
4013 		break;
4014 	case e1000_82576:
4015 	case e1000_82580:
4016 	case e1000_i350:
4017 	case e1000_i354:
4018 	case e1000_i210:
4019 	default:
4020 		/* If rss_queues > half of max_rss_queues, pair the queues in
4021 		 * order to conserve interrupts due to limited supply.
4022 		 */
4023 		if (adapter->rss_queues > (max_rss_queues / 2))
4024 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
4025 		else
4026 			adapter->flags &= ~IGB_FLAG_QUEUE_PAIRS;
4027 		break;
4028 	}
4029 }
4030 
4031 /**
4032  *  igb_sw_init - Initialize general software structures (struct igb_adapter)
4033  *  @adapter: board private structure to initialize
4034  *
4035  *  igb_sw_init initializes the Adapter private data structure.
4036  *  Fields are initialized based on PCI device information and
4037  *  OS network device settings (MTU size).
4038  **/
4039 static int igb_sw_init(struct igb_adapter *adapter)
4040 {
4041 	struct e1000_hw *hw = &adapter->hw;
4042 	struct net_device *netdev = adapter->netdev;
4043 	struct pci_dev *pdev = adapter->pdev;
4044 
4045 	pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
4046 
4047 	/* set default ring sizes */
4048 	adapter->tx_ring_count = IGB_DEFAULT_TXD;
4049 	adapter->rx_ring_count = IGB_DEFAULT_RXD;
4050 
4051 	/* set default ITR values */
4052 	adapter->rx_itr_setting = IGB_DEFAULT_ITR;
4053 	adapter->tx_itr_setting = IGB_DEFAULT_ITR;
4054 
4055 	/* set default work limits */
4056 	adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
4057 
4058 	adapter->max_frame_size = netdev->mtu + IGB_ETH_PKT_HDR_PAD;
4059 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4060 
4061 	spin_lock_init(&adapter->nfc_lock);
4062 	spin_lock_init(&adapter->stats64_lock);
4063 
4064 	/* init spinlock to avoid concurrency of VF resources */
4065 	spin_lock_init(&adapter->vfs_lock);
4066 #ifdef CONFIG_PCI_IOV
4067 	switch (hw->mac.type) {
4068 	case e1000_82576:
4069 	case e1000_i350:
4070 		if (max_vfs > 7) {
4071 			dev_warn(&pdev->dev,
4072 				 "Maximum of 7 VFs per PF, using max\n");
4073 			max_vfs = adapter->vfs_allocated_count = 7;
4074 		} else
4075 			adapter->vfs_allocated_count = max_vfs;
4076 		if (adapter->vfs_allocated_count)
4077 			dev_warn(&pdev->dev,
4078 				 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
4079 		break;
4080 	default:
4081 		break;
4082 	}
4083 #endif /* CONFIG_PCI_IOV */
4084 
4085 	/* Assume MSI-X interrupts, will be checked during IRQ allocation */
4086 	adapter->flags |= IGB_FLAG_HAS_MSIX;
4087 
4088 	adapter->mac_table = kcalloc(hw->mac.rar_entry_count,
4089 				     sizeof(struct igb_mac_addr),
4090 				     GFP_KERNEL);
4091 	if (!adapter->mac_table)
4092 		return -ENOMEM;
4093 
4094 	igb_probe_vfs(adapter);
4095 
4096 	igb_init_queue_configuration(adapter);
4097 
4098 	/* Setup and initialize a copy of the hw vlan table array */
4099 	adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
4100 				       GFP_KERNEL);
4101 	if (!adapter->shadow_vfta)
4102 		return -ENOMEM;
4103 
4104 	/* This call may decrease the number of queues */
4105 	if (igb_init_interrupt_scheme(adapter, true)) {
4106 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
4107 		return -ENOMEM;
4108 	}
4109 
4110 	/* Explicitly disable IRQ since the NIC can be in any state. */
4111 	igb_irq_disable(adapter);
4112 
4113 	if (hw->mac.type >= e1000_i350)
4114 		adapter->flags &= ~IGB_FLAG_DMAC;
4115 
4116 	set_bit(__IGB_DOWN, &adapter->state);
4117 	return 0;
4118 }
4119 
4120 /**
4121  *  __igb_open - Called when a network interface is made active
4122  *  @netdev: network interface device structure
4123  *  @resuming: indicates whether we are in a resume call
4124  *
4125  *  Returns 0 on success, negative value on failure
4126  *
4127  *  The open entry point is called when a network interface is made
4128  *  active by the system (IFF_UP).  At this point all resources needed
4129  *  for transmit and receive operations are allocated, the interrupt
4130  *  handler is registered with the OS, the watchdog timer is started,
4131  *  and the stack is notified that the interface is ready.
4132  **/
4133 static int __igb_open(struct net_device *netdev, bool resuming)
4134 {
4135 	struct igb_adapter *adapter = netdev_priv(netdev);
4136 	struct e1000_hw *hw = &adapter->hw;
4137 	struct pci_dev *pdev = adapter->pdev;
4138 	int err;
4139 	int i;
4140 
4141 	/* disallow open during test */
4142 	if (test_bit(__IGB_TESTING, &adapter->state)) {
4143 		WARN_ON(resuming);
4144 		return -EBUSY;
4145 	}
4146 
4147 	if (!resuming)
4148 		pm_runtime_get_sync(&pdev->dev);
4149 
4150 	netif_carrier_off(netdev);
4151 
4152 	/* allocate transmit descriptors */
4153 	err = igb_setup_all_tx_resources(adapter);
4154 	if (err)
4155 		goto err_setup_tx;
4156 
4157 	/* allocate receive descriptors */
4158 	err = igb_setup_all_rx_resources(adapter);
4159 	if (err)
4160 		goto err_setup_rx;
4161 
4162 	igb_power_up_link(adapter);
4163 
4164 	/* before we allocate an interrupt, we must be ready to handle it.
4165 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4166 	 * as soon as we call pci_request_irq, so we have to setup our
4167 	 * clean_rx handler before we do so.
4168 	 */
4169 	igb_configure(adapter);
4170 
4171 	err = igb_request_irq(adapter);
4172 	if (err)
4173 		goto err_req_irq;
4174 
4175 	/* Notify the stack of the actual queue counts. */
4176 	err = netif_set_real_num_tx_queues(adapter->netdev,
4177 					   adapter->num_tx_queues);
4178 	if (err)
4179 		goto err_set_queues;
4180 
4181 	err = netif_set_real_num_rx_queues(adapter->netdev,
4182 					   adapter->num_rx_queues);
4183 	if (err)
4184 		goto err_set_queues;
4185 
4186 	/* From here on the code is the same as igb_up() */
4187 	clear_bit(__IGB_DOWN, &adapter->state);
4188 
4189 	for (i = 0; i < adapter->num_q_vectors; i++)
4190 		napi_enable(&(adapter->q_vector[i]->napi));
4191 
4192 	/* Clear any pending interrupts. */
4193 	rd32(E1000_TSICR);
4194 	rd32(E1000_ICR);
4195 
4196 	igb_irq_enable(adapter);
4197 
4198 	/* notify VFs that reset has been completed */
4199 	if (adapter->vfs_allocated_count) {
4200 		u32 reg_data = rd32(E1000_CTRL_EXT);
4201 
4202 		reg_data |= E1000_CTRL_EXT_PFRSTD;
4203 		wr32(E1000_CTRL_EXT, reg_data);
4204 	}
4205 
4206 	netif_tx_start_all_queues(netdev);
4207 
4208 	if (!resuming)
4209 		pm_runtime_put(&pdev->dev);
4210 
4211 	/* start the watchdog. */
4212 	hw->mac.get_link_status = 1;
4213 	schedule_work(&adapter->watchdog_task);
4214 
4215 	return 0;
4216 
4217 err_set_queues:
4218 	igb_free_irq(adapter);
4219 err_req_irq:
4220 	igb_release_hw_control(adapter);
4221 	igb_power_down_link(adapter);
4222 	igb_free_all_rx_resources(adapter);
4223 err_setup_rx:
4224 	igb_free_all_tx_resources(adapter);
4225 err_setup_tx:
4226 	igb_reset(adapter);
4227 	if (!resuming)
4228 		pm_runtime_put(&pdev->dev);
4229 
4230 	return err;
4231 }
4232 
4233 int igb_open(struct net_device *netdev)
4234 {
4235 	return __igb_open(netdev, false);
4236 }
4237 
4238 /**
4239  *  __igb_close - Disables a network interface
4240  *  @netdev: network interface device structure
4241  *  @suspending: indicates we are in a suspend call
4242  *
4243  *  Returns 0, this is not allowed to fail
4244  *
4245  *  The close entry point is called when an interface is de-activated
4246  *  by the OS.  The hardware is still under the driver's control, but
4247  *  needs to be disabled.  A global MAC reset is issued to stop the
4248  *  hardware, and all transmit and receive resources are freed.
4249  **/
4250 static int __igb_close(struct net_device *netdev, bool suspending)
4251 {
4252 	struct igb_adapter *adapter = netdev_priv(netdev);
4253 	struct pci_dev *pdev = adapter->pdev;
4254 
4255 	WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
4256 
4257 	if (!suspending)
4258 		pm_runtime_get_sync(&pdev->dev);
4259 
4260 	igb_down(adapter);
4261 	igb_free_irq(adapter);
4262 
4263 	igb_free_all_tx_resources(adapter);
4264 	igb_free_all_rx_resources(adapter);
4265 
4266 	if (!suspending)
4267 		pm_runtime_put_sync(&pdev->dev);
4268 	return 0;
4269 }
4270 
4271 int igb_close(struct net_device *netdev)
4272 {
4273 	if (netif_device_present(netdev) || netdev->dismantle)
4274 		return __igb_close(netdev, false);
4275 	return 0;
4276 }
4277 
4278 /**
4279  *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
4280  *  @tx_ring: tx descriptor ring (for a specific queue) to setup
4281  *
4282  *  Return 0 on success, negative on failure
4283  **/
4284 int igb_setup_tx_resources(struct igb_ring *tx_ring)
4285 {
4286 	struct device *dev = tx_ring->dev;
4287 	int size;
4288 
4289 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
4290 
4291 	tx_ring->tx_buffer_info = vmalloc(size);
4292 	if (!tx_ring->tx_buffer_info)
4293 		goto err;
4294 
4295 	/* round up to nearest 4K */
4296 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
4297 	tx_ring->size = ALIGN(tx_ring->size, 4096);
4298 
4299 	tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4300 					   &tx_ring->dma, GFP_KERNEL);
4301 	if (!tx_ring->desc)
4302 		goto err;
4303 
4304 	tx_ring->next_to_use = 0;
4305 	tx_ring->next_to_clean = 0;
4306 
4307 	return 0;
4308 
4309 err:
4310 	vfree(tx_ring->tx_buffer_info);
4311 	tx_ring->tx_buffer_info = NULL;
4312 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4313 	return -ENOMEM;
4314 }
4315 
4316 /**
4317  *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
4318  *				 (Descriptors) for all queues
4319  *  @adapter: board private structure
4320  *
4321  *  Return 0 on success, negative on failure
4322  **/
4323 static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
4324 {
4325 	struct pci_dev *pdev = adapter->pdev;
4326 	int i, err = 0;
4327 
4328 	for (i = 0; i < adapter->num_tx_queues; i++) {
4329 		err = igb_setup_tx_resources(adapter->tx_ring[i]);
4330 		if (err) {
4331 			dev_err(&pdev->dev,
4332 				"Allocation for Tx Queue %u failed\n", i);
4333 			for (i--; i >= 0; i--)
4334 				igb_free_tx_resources(adapter->tx_ring[i]);
4335 			break;
4336 		}
4337 	}
4338 
4339 	return err;
4340 }
4341 
4342 /**
4343  *  igb_setup_tctl - configure the transmit control registers
4344  *  @adapter: Board private structure
4345  **/
4346 void igb_setup_tctl(struct igb_adapter *adapter)
4347 {
4348 	struct e1000_hw *hw = &adapter->hw;
4349 	u32 tctl;
4350 
4351 	/* disable queue 0 which is enabled by default on 82575 and 82576 */
4352 	wr32(E1000_TXDCTL(0), 0);
4353 
4354 	/* Program the Transmit Control Register */
4355 	tctl = rd32(E1000_TCTL);
4356 	tctl &= ~E1000_TCTL_CT;
4357 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
4358 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
4359 
4360 	igb_config_collision_dist(hw);
4361 
4362 	/* Enable transmits */
4363 	tctl |= E1000_TCTL_EN;
4364 
4365 	wr32(E1000_TCTL, tctl);
4366 }
4367 
4368 /**
4369  *  igb_configure_tx_ring - Configure transmit ring after Reset
4370  *  @adapter: board private structure
4371  *  @ring: tx ring to configure
4372  *
4373  *  Configure a transmit ring after a reset.
4374  **/
4375 void igb_configure_tx_ring(struct igb_adapter *adapter,
4376 			   struct igb_ring *ring)
4377 {
4378 	struct e1000_hw *hw = &adapter->hw;
4379 	u32 txdctl = 0;
4380 	u64 tdba = ring->dma;
4381 	int reg_idx = ring->reg_idx;
4382 
4383 	wr32(E1000_TDLEN(reg_idx),
4384 	     ring->count * sizeof(union e1000_adv_tx_desc));
4385 	wr32(E1000_TDBAL(reg_idx),
4386 	     tdba & 0x00000000ffffffffULL);
4387 	wr32(E1000_TDBAH(reg_idx), tdba >> 32);
4388 
4389 	ring->tail = adapter->io_addr + E1000_TDT(reg_idx);
4390 	wr32(E1000_TDH(reg_idx), 0);
4391 	writel(0, ring->tail);
4392 
4393 	txdctl |= IGB_TX_PTHRESH;
4394 	txdctl |= IGB_TX_HTHRESH << 8;
4395 	txdctl |= IGB_TX_WTHRESH << 16;
4396 
4397 	/* reinitialize tx_buffer_info */
4398 	memset(ring->tx_buffer_info, 0,
4399 	       sizeof(struct igb_tx_buffer) * ring->count);
4400 
4401 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
4402 	wr32(E1000_TXDCTL(reg_idx), txdctl);
4403 }
4404 
4405 /**
4406  *  igb_configure_tx - Configure transmit Unit after Reset
4407  *  @adapter: board private structure
4408  *
4409  *  Configure the Tx unit of the MAC after a reset.
4410  **/
4411 static void igb_configure_tx(struct igb_adapter *adapter)
4412 {
4413 	struct e1000_hw *hw = &adapter->hw;
4414 	int i;
4415 
4416 	/* disable the queues */
4417 	for (i = 0; i < adapter->num_tx_queues; i++)
4418 		wr32(E1000_TXDCTL(adapter->tx_ring[i]->reg_idx), 0);
4419 
4420 	wrfl();
4421 	usleep_range(10000, 20000);
4422 
4423 	for (i = 0; i < adapter->num_tx_queues; i++)
4424 		igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
4425 }
4426 
4427 /**
4428  *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
4429  *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
4430  *
4431  *  Returns 0 on success, negative on failure
4432  **/
4433 int igb_setup_rx_resources(struct igb_ring *rx_ring)
4434 {
4435 	struct igb_adapter *adapter = netdev_priv(rx_ring->netdev);
4436 	struct device *dev = rx_ring->dev;
4437 	int size, res;
4438 
4439 	/* XDP RX-queue info */
4440 	if (xdp_rxq_info_is_reg(&rx_ring->xdp_rxq))
4441 		xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4442 	res = xdp_rxq_info_reg(&rx_ring->xdp_rxq, rx_ring->netdev,
4443 			       rx_ring->queue_index, 0);
4444 	if (res < 0) {
4445 		dev_err(dev, "Failed to register xdp_rxq index %u\n",
4446 			rx_ring->queue_index);
4447 		return res;
4448 	}
4449 
4450 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
4451 
4452 	rx_ring->rx_buffer_info = vmalloc(size);
4453 	if (!rx_ring->rx_buffer_info)
4454 		goto err;
4455 
4456 	/* Round up to nearest 4K */
4457 	rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
4458 	rx_ring->size = ALIGN(rx_ring->size, 4096);
4459 
4460 	rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4461 					   &rx_ring->dma, GFP_KERNEL);
4462 	if (!rx_ring->desc)
4463 		goto err;
4464 
4465 	rx_ring->next_to_alloc = 0;
4466 	rx_ring->next_to_clean = 0;
4467 	rx_ring->next_to_use = 0;
4468 
4469 	rx_ring->xdp_prog = adapter->xdp_prog;
4470 
4471 	return 0;
4472 
4473 err:
4474 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4475 	vfree(rx_ring->rx_buffer_info);
4476 	rx_ring->rx_buffer_info = NULL;
4477 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4478 	return -ENOMEM;
4479 }
4480 
4481 /**
4482  *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
4483  *				 (Descriptors) for all queues
4484  *  @adapter: board private structure
4485  *
4486  *  Return 0 on success, negative on failure
4487  **/
4488 static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
4489 {
4490 	struct pci_dev *pdev = adapter->pdev;
4491 	int i, err = 0;
4492 
4493 	for (i = 0; i < adapter->num_rx_queues; i++) {
4494 		err = igb_setup_rx_resources(adapter->rx_ring[i]);
4495 		if (err) {
4496 			dev_err(&pdev->dev,
4497 				"Allocation for Rx Queue %u failed\n", i);
4498 			for (i--; i >= 0; i--)
4499 				igb_free_rx_resources(adapter->rx_ring[i]);
4500 			break;
4501 		}
4502 	}
4503 
4504 	return err;
4505 }
4506 
4507 /**
4508  *  igb_setup_mrqc - configure the multiple receive queue control registers
4509  *  @adapter: Board private structure
4510  **/
4511 static void igb_setup_mrqc(struct igb_adapter *adapter)
4512 {
4513 	struct e1000_hw *hw = &adapter->hw;
4514 	u32 mrqc, rxcsum;
4515 	u32 j, num_rx_queues;
4516 	u32 rss_key[10];
4517 
4518 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
4519 	for (j = 0; j < 10; j++)
4520 		wr32(E1000_RSSRK(j), rss_key[j]);
4521 
4522 	num_rx_queues = adapter->rss_queues;
4523 
4524 	switch (hw->mac.type) {
4525 	case e1000_82576:
4526 		/* 82576 supports 2 RSS queues for SR-IOV */
4527 		if (adapter->vfs_allocated_count)
4528 			num_rx_queues = 2;
4529 		break;
4530 	default:
4531 		break;
4532 	}
4533 
4534 	if (adapter->rss_indir_tbl_init != num_rx_queues) {
4535 		for (j = 0; j < IGB_RETA_SIZE; j++)
4536 			adapter->rss_indir_tbl[j] =
4537 			(j * num_rx_queues) / IGB_RETA_SIZE;
4538 		adapter->rss_indir_tbl_init = num_rx_queues;
4539 	}
4540 	igb_write_rss_indir_tbl(adapter);
4541 
4542 	/* Disable raw packet checksumming so that RSS hash is placed in
4543 	 * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
4544 	 * offloads as they are enabled by default
4545 	 */
4546 	rxcsum = rd32(E1000_RXCSUM);
4547 	rxcsum |= E1000_RXCSUM_PCSD;
4548 
4549 	if (adapter->hw.mac.type >= e1000_82576)
4550 		/* Enable Receive Checksum Offload for SCTP */
4551 		rxcsum |= E1000_RXCSUM_CRCOFL;
4552 
4553 	/* Don't need to set TUOFL or IPOFL, they default to 1 */
4554 	wr32(E1000_RXCSUM, rxcsum);
4555 
4556 	/* Generate RSS hash based on packet types, TCP/UDP
4557 	 * port numbers and/or IPv4/v6 src and dst addresses
4558 	 */
4559 	mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
4560 	       E1000_MRQC_RSS_FIELD_IPV4_TCP |
4561 	       E1000_MRQC_RSS_FIELD_IPV6 |
4562 	       E1000_MRQC_RSS_FIELD_IPV6_TCP |
4563 	       E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
4564 
4565 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
4566 		mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
4567 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
4568 		mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
4569 
4570 	/* If VMDq is enabled then we set the appropriate mode for that, else
4571 	 * we default to RSS so that an RSS hash is calculated per packet even
4572 	 * if we are only using one queue
4573 	 */
4574 	if (adapter->vfs_allocated_count) {
4575 		if (hw->mac.type > e1000_82575) {
4576 			/* Set the default pool for the PF's first queue */
4577 			u32 vtctl = rd32(E1000_VT_CTL);
4578 
4579 			vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
4580 				   E1000_VT_CTL_DISABLE_DEF_POOL);
4581 			vtctl |= adapter->vfs_allocated_count <<
4582 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
4583 			wr32(E1000_VT_CTL, vtctl);
4584 		}
4585 		if (adapter->rss_queues > 1)
4586 			mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_MQ;
4587 		else
4588 			mrqc |= E1000_MRQC_ENABLE_VMDQ;
4589 	} else {
4590 		mrqc |= E1000_MRQC_ENABLE_RSS_MQ;
4591 	}
4592 	igb_vmm_control(adapter);
4593 
4594 	wr32(E1000_MRQC, mrqc);
4595 }
4596 
4597 /**
4598  *  igb_setup_rctl - configure the receive control registers
4599  *  @adapter: Board private structure
4600  **/
4601 void igb_setup_rctl(struct igb_adapter *adapter)
4602 {
4603 	struct e1000_hw *hw = &adapter->hw;
4604 	u32 rctl;
4605 
4606 	rctl = rd32(E1000_RCTL);
4607 
4608 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
4609 	rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
4610 
4611 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
4612 		(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
4613 
4614 	/* enable stripping of CRC. It's unlikely this will break BMC
4615 	 * redirection as it did with e1000. Newer features require
4616 	 * that the HW strips the CRC.
4617 	 */
4618 	rctl |= E1000_RCTL_SECRC;
4619 
4620 	/* disable store bad packets and clear size bits. */
4621 	rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
4622 
4623 	/* enable LPE to allow for reception of jumbo frames */
4624 	rctl |= E1000_RCTL_LPE;
4625 
4626 	/* disable queue 0 to prevent tail write w/o re-config */
4627 	wr32(E1000_RXDCTL(0), 0);
4628 
4629 	/* Attention!!!  For SR-IOV PF driver operations you must enable
4630 	 * queue drop for all VF and PF queues to prevent head of line blocking
4631 	 * if an un-trusted VF does not provide descriptors to hardware.
4632 	 */
4633 	if (adapter->vfs_allocated_count) {
4634 		/* set all queue drop enable bits */
4635 		wr32(E1000_QDE, ALL_QUEUES);
4636 	}
4637 
4638 	/* This is useful for sniffing bad packets. */
4639 	if (adapter->netdev->features & NETIF_F_RXALL) {
4640 		/* UPE and MPE will be handled by normal PROMISC logic
4641 		 * in e1000e_set_rx_mode
4642 		 */
4643 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
4644 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
4645 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
4646 
4647 		rctl &= ~(E1000_RCTL_DPF | /* Allow filtered pause */
4648 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
4649 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
4650 		 * and that breaks VLANs.
4651 		 */
4652 	}
4653 
4654 	wr32(E1000_RCTL, rctl);
4655 }
4656 
4657 static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
4658 				   int vfn)
4659 {
4660 	struct e1000_hw *hw = &adapter->hw;
4661 	u32 vmolr;
4662 
4663 	if (size > MAX_JUMBO_FRAME_SIZE)
4664 		size = MAX_JUMBO_FRAME_SIZE;
4665 
4666 	vmolr = rd32(E1000_VMOLR(vfn));
4667 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
4668 	vmolr |= size | E1000_VMOLR_LPE;
4669 	wr32(E1000_VMOLR(vfn), vmolr);
4670 
4671 	return 0;
4672 }
4673 
4674 static inline void igb_set_vf_vlan_strip(struct igb_adapter *adapter,
4675 					 int vfn, bool enable)
4676 {
4677 	struct e1000_hw *hw = &adapter->hw;
4678 	u32 val, reg;
4679 
4680 	if (hw->mac.type < e1000_82576)
4681 		return;
4682 
4683 	if (hw->mac.type == e1000_i350)
4684 		reg = E1000_DVMOLR(vfn);
4685 	else
4686 		reg = E1000_VMOLR(vfn);
4687 
4688 	val = rd32(reg);
4689 	if (enable)
4690 		val |= E1000_VMOLR_STRVLAN;
4691 	else
4692 		val &= ~(E1000_VMOLR_STRVLAN);
4693 	wr32(reg, val);
4694 }
4695 
4696 static inline void igb_set_vmolr(struct igb_adapter *adapter,
4697 				 int vfn, bool aupe)
4698 {
4699 	struct e1000_hw *hw = &adapter->hw;
4700 	u32 vmolr;
4701 
4702 	/* This register exists only on 82576 and newer so if we are older then
4703 	 * we should exit and do nothing
4704 	 */
4705 	if (hw->mac.type < e1000_82576)
4706 		return;
4707 
4708 	vmolr = rd32(E1000_VMOLR(vfn));
4709 	if (aupe)
4710 		vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
4711 	else
4712 		vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
4713 
4714 	/* clear all bits that might not be set */
4715 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
4716 
4717 	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
4718 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
4719 	/* for VMDq only allow the VFs and pool 0 to accept broadcast and
4720 	 * multicast packets
4721 	 */
4722 	if (vfn <= adapter->vfs_allocated_count)
4723 		vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
4724 
4725 	wr32(E1000_VMOLR(vfn), vmolr);
4726 }
4727 
4728 /**
4729  *  igb_setup_srrctl - configure the split and replication receive control
4730  *                     registers
4731  *  @adapter: Board private structure
4732  *  @ring: receive ring to be configured
4733  **/
4734 void igb_setup_srrctl(struct igb_adapter *adapter, struct igb_ring *ring)
4735 {
4736 	struct e1000_hw *hw = &adapter->hw;
4737 	int reg_idx = ring->reg_idx;
4738 	u32 srrctl = 0;
4739 
4740 	srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
4741 	if (ring_uses_large_buffer(ring))
4742 		srrctl |= IGB_RXBUFFER_3072 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4743 	else
4744 		srrctl |= IGB_RXBUFFER_2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4745 	srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
4746 	if (hw->mac.type >= e1000_82580)
4747 		srrctl |= E1000_SRRCTL_TIMESTAMP;
4748 	/* Only set Drop Enable if VFs allocated, or we are supporting multiple
4749 	 * queues and rx flow control is disabled
4750 	 */
4751 	if (adapter->vfs_allocated_count ||
4752 	    (!(hw->fc.current_mode & e1000_fc_rx_pause) &&
4753 	     adapter->num_rx_queues > 1))
4754 		srrctl |= E1000_SRRCTL_DROP_EN;
4755 
4756 	wr32(E1000_SRRCTL(reg_idx), srrctl);
4757 }
4758 
4759 /**
4760  *  igb_configure_rx_ring - Configure a receive ring after Reset
4761  *  @adapter: board private structure
4762  *  @ring: receive ring to be configured
4763  *
4764  *  Configure the Rx unit of the MAC after a reset.
4765  **/
4766 void igb_configure_rx_ring(struct igb_adapter *adapter,
4767 			   struct igb_ring *ring)
4768 {
4769 	struct e1000_hw *hw = &adapter->hw;
4770 	union e1000_adv_rx_desc *rx_desc;
4771 	u64 rdba = ring->dma;
4772 	int reg_idx = ring->reg_idx;
4773 	u32 rxdctl = 0;
4774 
4775 	xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4776 	WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4777 					   MEM_TYPE_PAGE_SHARED, NULL));
4778 
4779 	/* disable the queue */
4780 	wr32(E1000_RXDCTL(reg_idx), 0);
4781 
4782 	/* Set DMA base address registers */
4783 	wr32(E1000_RDBAL(reg_idx),
4784 	     rdba & 0x00000000ffffffffULL);
4785 	wr32(E1000_RDBAH(reg_idx), rdba >> 32);
4786 	wr32(E1000_RDLEN(reg_idx),
4787 	     ring->count * sizeof(union e1000_adv_rx_desc));
4788 
4789 	/* initialize head and tail */
4790 	ring->tail = adapter->io_addr + E1000_RDT(reg_idx);
4791 	wr32(E1000_RDH(reg_idx), 0);
4792 	writel(0, ring->tail);
4793 
4794 	/* set descriptor configuration */
4795 	igb_setup_srrctl(adapter, ring);
4796 
4797 	/* set filtering for VMDQ pools */
4798 	igb_set_vmolr(adapter, reg_idx & 0x7, true);
4799 
4800 	rxdctl |= IGB_RX_PTHRESH;
4801 	rxdctl |= IGB_RX_HTHRESH << 8;
4802 	rxdctl |= IGB_RX_WTHRESH << 16;
4803 
4804 	/* initialize rx_buffer_info */
4805 	memset(ring->rx_buffer_info, 0,
4806 	       sizeof(struct igb_rx_buffer) * ring->count);
4807 
4808 	/* initialize Rx descriptor 0 */
4809 	rx_desc = IGB_RX_DESC(ring, 0);
4810 	rx_desc->wb.upper.length = 0;
4811 
4812 	/* enable receive descriptor fetching */
4813 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
4814 	wr32(E1000_RXDCTL(reg_idx), rxdctl);
4815 }
4816 
4817 static void igb_set_rx_buffer_len(struct igb_adapter *adapter,
4818 				  struct igb_ring *rx_ring)
4819 {
4820 #if (PAGE_SIZE < 8192)
4821 	struct e1000_hw *hw = &adapter->hw;
4822 #endif
4823 
4824 	/* set build_skb and buffer size flags */
4825 	clear_ring_build_skb_enabled(rx_ring);
4826 	clear_ring_uses_large_buffer(rx_ring);
4827 
4828 	if (adapter->flags & IGB_FLAG_RX_LEGACY)
4829 		return;
4830 
4831 	set_ring_build_skb_enabled(rx_ring);
4832 
4833 #if (PAGE_SIZE < 8192)
4834 	if (adapter->max_frame_size > IGB_MAX_FRAME_BUILD_SKB ||
4835 	    rd32(E1000_RCTL) & E1000_RCTL_SBP)
4836 		set_ring_uses_large_buffer(rx_ring);
4837 #endif
4838 }
4839 
4840 /**
4841  *  igb_configure_rx - Configure receive Unit after Reset
4842  *  @adapter: board private structure
4843  *
4844  *  Configure the Rx unit of the MAC after a reset.
4845  **/
4846 static void igb_configure_rx(struct igb_adapter *adapter)
4847 {
4848 	int i;
4849 
4850 	/* set the correct pool for the PF default MAC address in entry 0 */
4851 	igb_set_default_mac_filter(adapter);
4852 
4853 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
4854 	 * the Base and Length of the Rx Descriptor Ring
4855 	 */
4856 	for (i = 0; i < adapter->num_rx_queues; i++) {
4857 		struct igb_ring *rx_ring = adapter->rx_ring[i];
4858 
4859 		igb_set_rx_buffer_len(adapter, rx_ring);
4860 		igb_configure_rx_ring(adapter, rx_ring);
4861 	}
4862 }
4863 
4864 /**
4865  *  igb_free_tx_resources - Free Tx Resources per Queue
4866  *  @tx_ring: Tx descriptor ring for a specific queue
4867  *
4868  *  Free all transmit software resources
4869  **/
4870 void igb_free_tx_resources(struct igb_ring *tx_ring)
4871 {
4872 	igb_clean_tx_ring(tx_ring);
4873 
4874 	vfree(tx_ring->tx_buffer_info);
4875 	tx_ring->tx_buffer_info = NULL;
4876 
4877 	/* if not set, then don't free */
4878 	if (!tx_ring->desc)
4879 		return;
4880 
4881 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4882 			  tx_ring->desc, tx_ring->dma);
4883 
4884 	tx_ring->desc = NULL;
4885 }
4886 
4887 /**
4888  *  igb_free_all_tx_resources - Free Tx Resources for All Queues
4889  *  @adapter: board private structure
4890  *
4891  *  Free all transmit software resources
4892  **/
4893 static void igb_free_all_tx_resources(struct igb_adapter *adapter)
4894 {
4895 	int i;
4896 
4897 	for (i = 0; i < adapter->num_tx_queues; i++)
4898 		if (adapter->tx_ring[i])
4899 			igb_free_tx_resources(adapter->tx_ring[i]);
4900 }
4901 
4902 /**
4903  *  igb_clean_tx_ring - Free Tx Buffers
4904  *  @tx_ring: ring to be cleaned
4905  **/
4906 static void igb_clean_tx_ring(struct igb_ring *tx_ring)
4907 {
4908 	u16 i = tx_ring->next_to_clean;
4909 	struct igb_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
4910 
4911 	while (i != tx_ring->next_to_use) {
4912 		union e1000_adv_tx_desc *eop_desc, *tx_desc;
4913 
4914 		/* Free all the Tx ring sk_buffs or xdp frames */
4915 		if (tx_buffer->type == IGB_TYPE_SKB)
4916 			dev_kfree_skb_any(tx_buffer->skb);
4917 		else
4918 			xdp_return_frame(tx_buffer->xdpf);
4919 
4920 		/* unmap skb header data */
4921 		dma_unmap_single(tx_ring->dev,
4922 				 dma_unmap_addr(tx_buffer, dma),
4923 				 dma_unmap_len(tx_buffer, len),
4924 				 DMA_TO_DEVICE);
4925 
4926 		/* check for eop_desc to determine the end of the packet */
4927 		eop_desc = tx_buffer->next_to_watch;
4928 		tx_desc = IGB_TX_DESC(tx_ring, i);
4929 
4930 		/* unmap remaining buffers */
4931 		while (tx_desc != eop_desc) {
4932 			tx_buffer++;
4933 			tx_desc++;
4934 			i++;
4935 			if (unlikely(i == tx_ring->count)) {
4936 				i = 0;
4937 				tx_buffer = tx_ring->tx_buffer_info;
4938 				tx_desc = IGB_TX_DESC(tx_ring, 0);
4939 			}
4940 
4941 			/* unmap any remaining paged data */
4942 			if (dma_unmap_len(tx_buffer, len))
4943 				dma_unmap_page(tx_ring->dev,
4944 					       dma_unmap_addr(tx_buffer, dma),
4945 					       dma_unmap_len(tx_buffer, len),
4946 					       DMA_TO_DEVICE);
4947 		}
4948 
4949 		tx_buffer->next_to_watch = NULL;
4950 
4951 		/* move us one more past the eop_desc for start of next pkt */
4952 		tx_buffer++;
4953 		i++;
4954 		if (unlikely(i == tx_ring->count)) {
4955 			i = 0;
4956 			tx_buffer = tx_ring->tx_buffer_info;
4957 		}
4958 	}
4959 
4960 	/* reset BQL for queue */
4961 	netdev_tx_reset_queue(txring_txq(tx_ring));
4962 
4963 	/* reset next_to_use and next_to_clean */
4964 	tx_ring->next_to_use = 0;
4965 	tx_ring->next_to_clean = 0;
4966 }
4967 
4968 /**
4969  *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
4970  *  @adapter: board private structure
4971  **/
4972 static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
4973 {
4974 	int i;
4975 
4976 	for (i = 0; i < adapter->num_tx_queues; i++)
4977 		if (adapter->tx_ring[i])
4978 			igb_clean_tx_ring(adapter->tx_ring[i]);
4979 }
4980 
4981 /**
4982  *  igb_free_rx_resources - Free Rx Resources
4983  *  @rx_ring: ring to clean the resources from
4984  *
4985  *  Free all receive software resources
4986  **/
4987 void igb_free_rx_resources(struct igb_ring *rx_ring)
4988 {
4989 	igb_clean_rx_ring(rx_ring);
4990 
4991 	rx_ring->xdp_prog = NULL;
4992 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4993 	vfree(rx_ring->rx_buffer_info);
4994 	rx_ring->rx_buffer_info = NULL;
4995 
4996 	/* if not set, then don't free */
4997 	if (!rx_ring->desc)
4998 		return;
4999 
5000 	dma_free_coherent(rx_ring->dev, rx_ring->size,
5001 			  rx_ring->desc, rx_ring->dma);
5002 
5003 	rx_ring->desc = NULL;
5004 }
5005 
5006 /**
5007  *  igb_free_all_rx_resources - Free Rx Resources for All Queues
5008  *  @adapter: board private structure
5009  *
5010  *  Free all receive software resources
5011  **/
5012 static void igb_free_all_rx_resources(struct igb_adapter *adapter)
5013 {
5014 	int i;
5015 
5016 	for (i = 0; i < adapter->num_rx_queues; i++)
5017 		if (adapter->rx_ring[i])
5018 			igb_free_rx_resources(adapter->rx_ring[i]);
5019 }
5020 
5021 /**
5022  *  igb_clean_rx_ring - Free Rx Buffers per Queue
5023  *  @rx_ring: ring to free buffers from
5024  **/
5025 static void igb_clean_rx_ring(struct igb_ring *rx_ring)
5026 {
5027 	u16 i = rx_ring->next_to_clean;
5028 
5029 	dev_kfree_skb(rx_ring->skb);
5030 	rx_ring->skb = NULL;
5031 
5032 	/* Free all the Rx ring sk_buffs */
5033 	while (i != rx_ring->next_to_alloc) {
5034 		struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
5035 
5036 		/* Invalidate cache lines that may have been written to by
5037 		 * device so that we avoid corrupting memory.
5038 		 */
5039 		dma_sync_single_range_for_cpu(rx_ring->dev,
5040 					      buffer_info->dma,
5041 					      buffer_info->page_offset,
5042 					      igb_rx_bufsz(rx_ring),
5043 					      DMA_FROM_DEVICE);
5044 
5045 		/* free resources associated with mapping */
5046 		dma_unmap_page_attrs(rx_ring->dev,
5047 				     buffer_info->dma,
5048 				     igb_rx_pg_size(rx_ring),
5049 				     DMA_FROM_DEVICE,
5050 				     IGB_RX_DMA_ATTR);
5051 		__page_frag_cache_drain(buffer_info->page,
5052 					buffer_info->pagecnt_bias);
5053 
5054 		i++;
5055 		if (i == rx_ring->count)
5056 			i = 0;
5057 	}
5058 
5059 	rx_ring->next_to_alloc = 0;
5060 	rx_ring->next_to_clean = 0;
5061 	rx_ring->next_to_use = 0;
5062 }
5063 
5064 /**
5065  *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
5066  *  @adapter: board private structure
5067  **/
5068 static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
5069 {
5070 	int i;
5071 
5072 	for (i = 0; i < adapter->num_rx_queues; i++)
5073 		if (adapter->rx_ring[i])
5074 			igb_clean_rx_ring(adapter->rx_ring[i]);
5075 }
5076 
5077 /**
5078  *  igb_set_mac - Change the Ethernet Address of the NIC
5079  *  @netdev: network interface device structure
5080  *  @p: pointer to an address structure
5081  *
5082  *  Returns 0 on success, negative on failure
5083  **/
5084 static int igb_set_mac(struct net_device *netdev, void *p)
5085 {
5086 	struct igb_adapter *adapter = netdev_priv(netdev);
5087 	struct e1000_hw *hw = &adapter->hw;
5088 	struct sockaddr *addr = p;
5089 
5090 	if (!is_valid_ether_addr(addr->sa_data))
5091 		return -EADDRNOTAVAIL;
5092 
5093 	eth_hw_addr_set(netdev, addr->sa_data);
5094 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
5095 
5096 	/* set the correct pool for the new PF MAC address in entry 0 */
5097 	igb_set_default_mac_filter(adapter);
5098 
5099 	return 0;
5100 }
5101 
5102 /**
5103  *  igb_write_mc_addr_list - write multicast addresses to MTA
5104  *  @netdev: network interface device structure
5105  *
5106  *  Writes multicast address list to the MTA hash table.
5107  *  Returns: -ENOMEM on failure
5108  *           0 on no addresses written
5109  *           X on writing X addresses to MTA
5110  **/
5111 static int igb_write_mc_addr_list(struct net_device *netdev)
5112 {
5113 	struct igb_adapter *adapter = netdev_priv(netdev);
5114 	struct e1000_hw *hw = &adapter->hw;
5115 	struct netdev_hw_addr *ha;
5116 	u8  *mta_list;
5117 	int i;
5118 
5119 	if (netdev_mc_empty(netdev)) {
5120 		/* nothing to program, so clear mc list */
5121 		igb_update_mc_addr_list(hw, NULL, 0);
5122 		igb_restore_vf_multicasts(adapter);
5123 		return 0;
5124 	}
5125 
5126 	mta_list = kcalloc(netdev_mc_count(netdev), 6, GFP_ATOMIC);
5127 	if (!mta_list)
5128 		return -ENOMEM;
5129 
5130 	/* The shared function expects a packed array of only addresses. */
5131 	i = 0;
5132 	netdev_for_each_mc_addr(ha, netdev)
5133 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
5134 
5135 	igb_update_mc_addr_list(hw, mta_list, i);
5136 	kfree(mta_list);
5137 
5138 	return netdev_mc_count(netdev);
5139 }
5140 
5141 static int igb_vlan_promisc_enable(struct igb_adapter *adapter)
5142 {
5143 	struct e1000_hw *hw = &adapter->hw;
5144 	u32 i, pf_id;
5145 
5146 	switch (hw->mac.type) {
5147 	case e1000_i210:
5148 	case e1000_i211:
5149 	case e1000_i350:
5150 		/* VLAN filtering needed for VLAN prio filter */
5151 		if (adapter->netdev->features & NETIF_F_NTUPLE)
5152 			break;
5153 		fallthrough;
5154 	case e1000_82576:
5155 	case e1000_82580:
5156 	case e1000_i354:
5157 		/* VLAN filtering needed for pool filtering */
5158 		if (adapter->vfs_allocated_count)
5159 			break;
5160 		fallthrough;
5161 	default:
5162 		return 1;
5163 	}
5164 
5165 	/* We are already in VLAN promisc, nothing to do */
5166 	if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
5167 		return 0;
5168 
5169 	if (!adapter->vfs_allocated_count)
5170 		goto set_vfta;
5171 
5172 	/* Add PF to all active pools */
5173 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
5174 
5175 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
5176 		u32 vlvf = rd32(E1000_VLVF(i));
5177 
5178 		vlvf |= BIT(pf_id);
5179 		wr32(E1000_VLVF(i), vlvf);
5180 	}
5181 
5182 set_vfta:
5183 	/* Set all bits in the VLAN filter table array */
5184 	for (i = E1000_VLAN_FILTER_TBL_SIZE; i--;)
5185 		hw->mac.ops.write_vfta(hw, i, ~0U);
5186 
5187 	/* Set flag so we don't redo unnecessary work */
5188 	adapter->flags |= IGB_FLAG_VLAN_PROMISC;
5189 
5190 	return 0;
5191 }
5192 
5193 #define VFTA_BLOCK_SIZE 8
5194 static void igb_scrub_vfta(struct igb_adapter *adapter, u32 vfta_offset)
5195 {
5196 	struct e1000_hw *hw = &adapter->hw;
5197 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
5198 	u32 vid_start = vfta_offset * 32;
5199 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
5200 	u32 i, vid, word, bits, pf_id;
5201 
5202 	/* guarantee that we don't scrub out management VLAN */
5203 	vid = adapter->mng_vlan_id;
5204 	if (vid >= vid_start && vid < vid_end)
5205 		vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
5206 
5207 	if (!adapter->vfs_allocated_count)
5208 		goto set_vfta;
5209 
5210 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
5211 
5212 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
5213 		u32 vlvf = rd32(E1000_VLVF(i));
5214 
5215 		/* pull VLAN ID from VLVF */
5216 		vid = vlvf & VLAN_VID_MASK;
5217 
5218 		/* only concern ourselves with a certain range */
5219 		if (vid < vid_start || vid >= vid_end)
5220 			continue;
5221 
5222 		if (vlvf & E1000_VLVF_VLANID_ENABLE) {
5223 			/* record VLAN ID in VFTA */
5224 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
5225 
5226 			/* if PF is part of this then continue */
5227 			if (test_bit(vid, adapter->active_vlans))
5228 				continue;
5229 		}
5230 
5231 		/* remove PF from the pool */
5232 		bits = ~BIT(pf_id);
5233 		bits &= rd32(E1000_VLVF(i));
5234 		wr32(E1000_VLVF(i), bits);
5235 	}
5236 
5237 set_vfta:
5238 	/* extract values from active_vlans and write back to VFTA */
5239 	for (i = VFTA_BLOCK_SIZE; i--;) {
5240 		vid = (vfta_offset + i) * 32;
5241 		word = vid / BITS_PER_LONG;
5242 		bits = vid % BITS_PER_LONG;
5243 
5244 		vfta[i] |= adapter->active_vlans[word] >> bits;
5245 
5246 		hw->mac.ops.write_vfta(hw, vfta_offset + i, vfta[i]);
5247 	}
5248 }
5249 
5250 static void igb_vlan_promisc_disable(struct igb_adapter *adapter)
5251 {
5252 	u32 i;
5253 
5254 	/* We are not in VLAN promisc, nothing to do */
5255 	if (!(adapter->flags & IGB_FLAG_VLAN_PROMISC))
5256 		return;
5257 
5258 	/* Set flag so we don't redo unnecessary work */
5259 	adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
5260 
5261 	for (i = 0; i < E1000_VLAN_FILTER_TBL_SIZE; i += VFTA_BLOCK_SIZE)
5262 		igb_scrub_vfta(adapter, i);
5263 }
5264 
5265 /**
5266  *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
5267  *  @netdev: network interface device structure
5268  *
5269  *  The set_rx_mode entry point is called whenever the unicast or multicast
5270  *  address lists or the network interface flags are updated.  This routine is
5271  *  responsible for configuring the hardware for proper unicast, multicast,
5272  *  promiscuous mode, and all-multi behavior.
5273  **/
5274 static void igb_set_rx_mode(struct net_device *netdev)
5275 {
5276 	struct igb_adapter *adapter = netdev_priv(netdev);
5277 	struct e1000_hw *hw = &adapter->hw;
5278 	unsigned int vfn = adapter->vfs_allocated_count;
5279 	u32 rctl = 0, vmolr = 0, rlpml = MAX_JUMBO_FRAME_SIZE;
5280 	int count;
5281 
5282 	/* Check for Promiscuous and All Multicast modes */
5283 	if (netdev->flags & IFF_PROMISC) {
5284 		rctl |= E1000_RCTL_UPE | E1000_RCTL_MPE;
5285 		vmolr |= E1000_VMOLR_MPME;
5286 
5287 		/* enable use of UTA filter to force packets to default pool */
5288 		if (hw->mac.type == e1000_82576)
5289 			vmolr |= E1000_VMOLR_ROPE;
5290 	} else {
5291 		if (netdev->flags & IFF_ALLMULTI) {
5292 			rctl |= E1000_RCTL_MPE;
5293 			vmolr |= E1000_VMOLR_MPME;
5294 		} else {
5295 			/* Write addresses to the MTA, if the attempt fails
5296 			 * then we should just turn on promiscuous mode so
5297 			 * that we can at least receive multicast traffic
5298 			 */
5299 			count = igb_write_mc_addr_list(netdev);
5300 			if (count < 0) {
5301 				rctl |= E1000_RCTL_MPE;
5302 				vmolr |= E1000_VMOLR_MPME;
5303 			} else if (count) {
5304 				vmolr |= E1000_VMOLR_ROMPE;
5305 			}
5306 		}
5307 	}
5308 
5309 	/* Write addresses to available RAR registers, if there is not
5310 	 * sufficient space to store all the addresses then enable
5311 	 * unicast promiscuous mode
5312 	 */
5313 	if (__dev_uc_sync(netdev, igb_uc_sync, igb_uc_unsync)) {
5314 		rctl |= E1000_RCTL_UPE;
5315 		vmolr |= E1000_VMOLR_ROPE;
5316 	}
5317 
5318 	/* enable VLAN filtering by default */
5319 	rctl |= E1000_RCTL_VFE;
5320 
5321 	/* disable VLAN filtering for modes that require it */
5322 	if ((netdev->flags & IFF_PROMISC) ||
5323 	    (netdev->features & NETIF_F_RXALL)) {
5324 		/* if we fail to set all rules then just clear VFE */
5325 		if (igb_vlan_promisc_enable(adapter))
5326 			rctl &= ~E1000_RCTL_VFE;
5327 	} else {
5328 		igb_vlan_promisc_disable(adapter);
5329 	}
5330 
5331 	/* update state of unicast, multicast, and VLAN filtering modes */
5332 	rctl |= rd32(E1000_RCTL) & ~(E1000_RCTL_UPE | E1000_RCTL_MPE |
5333 				     E1000_RCTL_VFE);
5334 	wr32(E1000_RCTL, rctl);
5335 
5336 #if (PAGE_SIZE < 8192)
5337 	if (!adapter->vfs_allocated_count) {
5338 		if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5339 			rlpml = IGB_MAX_FRAME_BUILD_SKB;
5340 	}
5341 #endif
5342 	wr32(E1000_RLPML, rlpml);
5343 
5344 	/* In order to support SR-IOV and eventually VMDq it is necessary to set
5345 	 * the VMOLR to enable the appropriate modes.  Without this workaround
5346 	 * we will have issues with VLAN tag stripping not being done for frames
5347 	 * that are only arriving because we are the default pool
5348 	 */
5349 	if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
5350 		return;
5351 
5352 	/* set UTA to appropriate mode */
5353 	igb_set_uta(adapter, !!(vmolr & E1000_VMOLR_ROPE));
5354 
5355 	vmolr |= rd32(E1000_VMOLR(vfn)) &
5356 		 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
5357 
5358 	/* enable Rx jumbo frames, restrict as needed to support build_skb */
5359 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
5360 #if (PAGE_SIZE < 8192)
5361 	if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5362 		vmolr |= IGB_MAX_FRAME_BUILD_SKB;
5363 	else
5364 #endif
5365 		vmolr |= MAX_JUMBO_FRAME_SIZE;
5366 	vmolr |= E1000_VMOLR_LPE;
5367 
5368 	wr32(E1000_VMOLR(vfn), vmolr);
5369 
5370 	igb_restore_vf_multicasts(adapter);
5371 }
5372 
5373 static void igb_check_wvbr(struct igb_adapter *adapter)
5374 {
5375 	struct e1000_hw *hw = &adapter->hw;
5376 	u32 wvbr = 0;
5377 
5378 	switch (hw->mac.type) {
5379 	case e1000_82576:
5380 	case e1000_i350:
5381 		wvbr = rd32(E1000_WVBR);
5382 		if (!wvbr)
5383 			return;
5384 		break;
5385 	default:
5386 		break;
5387 	}
5388 
5389 	adapter->wvbr |= wvbr;
5390 }
5391 
5392 #define IGB_STAGGERED_QUEUE_OFFSET 8
5393 
5394 static void igb_spoof_check(struct igb_adapter *adapter)
5395 {
5396 	int j;
5397 
5398 	if (!adapter->wvbr)
5399 		return;
5400 
5401 	for (j = 0; j < adapter->vfs_allocated_count; j++) {
5402 		if (adapter->wvbr & BIT(j) ||
5403 		    adapter->wvbr & BIT(j + IGB_STAGGERED_QUEUE_OFFSET)) {
5404 			dev_warn(&adapter->pdev->dev,
5405 				"Spoof event(s) detected on VF %d\n", j);
5406 			adapter->wvbr &=
5407 				~(BIT(j) |
5408 				  BIT(j + IGB_STAGGERED_QUEUE_OFFSET));
5409 		}
5410 	}
5411 }
5412 
5413 /* Need to wait a few seconds after link up to get diagnostic information from
5414  * the phy
5415  */
5416 static void igb_update_phy_info(struct timer_list *t)
5417 {
5418 	struct igb_adapter *adapter = from_timer(adapter, t, phy_info_timer);
5419 	igb_get_phy_info(&adapter->hw);
5420 }
5421 
5422 /**
5423  *  igb_has_link - check shared code for link and determine up/down
5424  *  @adapter: pointer to driver private info
5425  **/
5426 bool igb_has_link(struct igb_adapter *adapter)
5427 {
5428 	struct e1000_hw *hw = &adapter->hw;
5429 	bool link_active = false;
5430 
5431 	/* get_link_status is set on LSC (link status) interrupt or
5432 	 * rx sequence error interrupt.  get_link_status will stay
5433 	 * false until the e1000_check_for_link establishes link
5434 	 * for copper adapters ONLY
5435 	 */
5436 	switch (hw->phy.media_type) {
5437 	case e1000_media_type_copper:
5438 		if (!hw->mac.get_link_status)
5439 			return true;
5440 		fallthrough;
5441 	case e1000_media_type_internal_serdes:
5442 		hw->mac.ops.check_for_link(hw);
5443 		link_active = !hw->mac.get_link_status;
5444 		break;
5445 	default:
5446 	case e1000_media_type_unknown:
5447 		break;
5448 	}
5449 
5450 	if (((hw->mac.type == e1000_i210) ||
5451 	     (hw->mac.type == e1000_i211)) &&
5452 	     (hw->phy.id == I210_I_PHY_ID)) {
5453 		if (!netif_carrier_ok(adapter->netdev)) {
5454 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5455 		} else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
5456 			adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
5457 			adapter->link_check_timeout = jiffies;
5458 		}
5459 	}
5460 
5461 	return link_active;
5462 }
5463 
5464 static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
5465 {
5466 	bool ret = false;
5467 	u32 ctrl_ext, thstat;
5468 
5469 	/* check for thermal sensor event on i350 copper only */
5470 	if (hw->mac.type == e1000_i350) {
5471 		thstat = rd32(E1000_THSTAT);
5472 		ctrl_ext = rd32(E1000_CTRL_EXT);
5473 
5474 		if ((hw->phy.media_type == e1000_media_type_copper) &&
5475 		    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
5476 			ret = !!(thstat & event);
5477 	}
5478 
5479 	return ret;
5480 }
5481 
5482 /**
5483  *  igb_check_lvmmc - check for malformed packets received
5484  *  and indicated in LVMMC register
5485  *  @adapter: pointer to adapter
5486  **/
5487 static void igb_check_lvmmc(struct igb_adapter *adapter)
5488 {
5489 	struct e1000_hw *hw = &adapter->hw;
5490 	u32 lvmmc;
5491 
5492 	lvmmc = rd32(E1000_LVMMC);
5493 	if (lvmmc) {
5494 		if (unlikely(net_ratelimit())) {
5495 			netdev_warn(adapter->netdev,
5496 				    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
5497 				    lvmmc);
5498 		}
5499 	}
5500 }
5501 
5502 /**
5503  *  igb_watchdog - Timer Call-back
5504  *  @t: pointer to timer_list containing our private info pointer
5505  **/
5506 static void igb_watchdog(struct timer_list *t)
5507 {
5508 	struct igb_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5509 	/* Do the rest outside of interrupt context */
5510 	schedule_work(&adapter->watchdog_task);
5511 }
5512 
5513 static void igb_watchdog_task(struct work_struct *work)
5514 {
5515 	struct igb_adapter *adapter = container_of(work,
5516 						   struct igb_adapter,
5517 						   watchdog_task);
5518 	struct e1000_hw *hw = &adapter->hw;
5519 	struct e1000_phy_info *phy = &hw->phy;
5520 	struct net_device *netdev = adapter->netdev;
5521 	u32 link;
5522 	int i;
5523 	u32 connsw;
5524 	u16 phy_data, retry_count = 20;
5525 
5526 	link = igb_has_link(adapter);
5527 
5528 	if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
5529 		if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
5530 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5531 		else
5532 			link = false;
5533 	}
5534 
5535 	/* Force link down if we have fiber to swap to */
5536 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5537 		if (hw->phy.media_type == e1000_media_type_copper) {
5538 			connsw = rd32(E1000_CONNSW);
5539 			if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
5540 				link = 0;
5541 		}
5542 	}
5543 	if (link) {
5544 		/* Perform a reset if the media type changed. */
5545 		if (hw->dev_spec._82575.media_changed) {
5546 			hw->dev_spec._82575.media_changed = false;
5547 			adapter->flags |= IGB_FLAG_MEDIA_RESET;
5548 			igb_reset(adapter);
5549 		}
5550 		/* Cancel scheduled suspend requests. */
5551 		pm_runtime_resume(netdev->dev.parent);
5552 
5553 		if (!netif_carrier_ok(netdev)) {
5554 			u32 ctrl;
5555 
5556 			hw->mac.ops.get_speed_and_duplex(hw,
5557 							 &adapter->link_speed,
5558 							 &adapter->link_duplex);
5559 
5560 			ctrl = rd32(E1000_CTRL);
5561 			/* Links status message must follow this format */
5562 			netdev_info(netdev,
5563 			       "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5564 			       netdev->name,
5565 			       adapter->link_speed,
5566 			       adapter->link_duplex == FULL_DUPLEX ?
5567 			       "Full" : "Half",
5568 			       (ctrl & E1000_CTRL_TFCE) &&
5569 			       (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
5570 			       (ctrl & E1000_CTRL_RFCE) ?  "RX" :
5571 			       (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
5572 
5573 			/* disable EEE if enabled */
5574 			if ((adapter->flags & IGB_FLAG_EEE) &&
5575 				(adapter->link_duplex == HALF_DUPLEX)) {
5576 				dev_info(&adapter->pdev->dev,
5577 				"EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
5578 				adapter->hw.dev_spec._82575.eee_disable = true;
5579 				adapter->flags &= ~IGB_FLAG_EEE;
5580 			}
5581 
5582 			/* check if SmartSpeed worked */
5583 			igb_check_downshift(hw);
5584 			if (phy->speed_downgraded)
5585 				netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
5586 
5587 			/* check for thermal sensor event */
5588 			if (igb_thermal_sensor_event(hw,
5589 			    E1000_THSTAT_LINK_THROTTLE))
5590 				netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
5591 
5592 			/* adjust timeout factor according to speed/duplex */
5593 			adapter->tx_timeout_factor = 1;
5594 			switch (adapter->link_speed) {
5595 			case SPEED_10:
5596 				adapter->tx_timeout_factor = 14;
5597 				break;
5598 			case SPEED_100:
5599 				/* maybe add some timeout factor ? */
5600 				break;
5601 			}
5602 
5603 			if (adapter->link_speed != SPEED_1000 ||
5604 			    !hw->phy.ops.read_reg)
5605 				goto no_wait;
5606 
5607 			/* wait for Remote receiver status OK */
5608 retry_read_status:
5609 			if (!igb_read_phy_reg(hw, PHY_1000T_STATUS,
5610 					      &phy_data)) {
5611 				if (!(phy_data & SR_1000T_REMOTE_RX_STATUS) &&
5612 				    retry_count) {
5613 					msleep(100);
5614 					retry_count--;
5615 					goto retry_read_status;
5616 				} else if (!retry_count) {
5617 					dev_err(&adapter->pdev->dev, "exceed max 2 second\n");
5618 				}
5619 			} else {
5620 				dev_err(&adapter->pdev->dev, "read 1000Base-T Status Reg\n");
5621 			}
5622 no_wait:
5623 			netif_carrier_on(netdev);
5624 
5625 			igb_ping_all_vfs(adapter);
5626 			igb_check_vf_rate_limit(adapter);
5627 
5628 			/* link state has changed, schedule phy info update */
5629 			if (!test_bit(__IGB_DOWN, &adapter->state))
5630 				mod_timer(&adapter->phy_info_timer,
5631 					  round_jiffies(jiffies + 2 * HZ));
5632 		}
5633 	} else {
5634 		if (netif_carrier_ok(netdev)) {
5635 			adapter->link_speed = 0;
5636 			adapter->link_duplex = 0;
5637 
5638 			/* check for thermal sensor event */
5639 			if (igb_thermal_sensor_event(hw,
5640 			    E1000_THSTAT_PWR_DOWN)) {
5641 				netdev_err(netdev, "The network adapter was stopped because it overheated\n");
5642 			}
5643 
5644 			/* Links status message must follow this format */
5645 			netdev_info(netdev, "igb: %s NIC Link is Down\n",
5646 			       netdev->name);
5647 			netif_carrier_off(netdev);
5648 
5649 			igb_ping_all_vfs(adapter);
5650 
5651 			/* link state has changed, schedule phy info update */
5652 			if (!test_bit(__IGB_DOWN, &adapter->state))
5653 				mod_timer(&adapter->phy_info_timer,
5654 					  round_jiffies(jiffies + 2 * HZ));
5655 
5656 			/* link is down, time to check for alternate media */
5657 			if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5658 				igb_check_swap_media(adapter);
5659 				if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5660 					schedule_work(&adapter->reset_task);
5661 					/* return immediately */
5662 					return;
5663 				}
5664 			}
5665 			pm_schedule_suspend(netdev->dev.parent,
5666 					    MSEC_PER_SEC * 5);
5667 
5668 		/* also check for alternate media here */
5669 		} else if (!netif_carrier_ok(netdev) &&
5670 			   (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
5671 			igb_check_swap_media(adapter);
5672 			if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5673 				schedule_work(&adapter->reset_task);
5674 				/* return immediately */
5675 				return;
5676 			}
5677 		}
5678 	}
5679 
5680 	spin_lock(&adapter->stats64_lock);
5681 	igb_update_stats(adapter);
5682 	spin_unlock(&adapter->stats64_lock);
5683 
5684 	for (i = 0; i < adapter->num_tx_queues; i++) {
5685 		struct igb_ring *tx_ring = adapter->tx_ring[i];
5686 		if (!netif_carrier_ok(netdev)) {
5687 			/* We've lost link, so the controller stops DMA,
5688 			 * but we've got queued Tx work that's never going
5689 			 * to get done, so reset controller to flush Tx.
5690 			 * (Do the reset outside of interrupt context).
5691 			 */
5692 			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
5693 				adapter->tx_timeout_count++;
5694 				schedule_work(&adapter->reset_task);
5695 				/* return immediately since reset is imminent */
5696 				return;
5697 			}
5698 		}
5699 
5700 		/* Force detection of hung controller every watchdog period */
5701 		set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
5702 	}
5703 
5704 	/* Cause software interrupt to ensure Rx ring is cleaned */
5705 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
5706 		u32 eics = 0;
5707 
5708 		for (i = 0; i < adapter->num_q_vectors; i++)
5709 			eics |= adapter->q_vector[i]->eims_value;
5710 		wr32(E1000_EICS, eics);
5711 	} else {
5712 		wr32(E1000_ICS, E1000_ICS_RXDMT0);
5713 	}
5714 
5715 	igb_spoof_check(adapter);
5716 	igb_ptp_rx_hang(adapter);
5717 	igb_ptp_tx_hang(adapter);
5718 
5719 	/* Check LVMMC register on i350/i354 only */
5720 	if ((adapter->hw.mac.type == e1000_i350) ||
5721 	    (adapter->hw.mac.type == e1000_i354))
5722 		igb_check_lvmmc(adapter);
5723 
5724 	/* Reset the timer */
5725 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
5726 		if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
5727 			mod_timer(&adapter->watchdog_timer,
5728 				  round_jiffies(jiffies +  HZ));
5729 		else
5730 			mod_timer(&adapter->watchdog_timer,
5731 				  round_jiffies(jiffies + 2 * HZ));
5732 	}
5733 }
5734 
5735 enum latency_range {
5736 	lowest_latency = 0,
5737 	low_latency = 1,
5738 	bulk_latency = 2,
5739 	latency_invalid = 255
5740 };
5741 
5742 /**
5743  *  igb_update_ring_itr - update the dynamic ITR value based on packet size
5744  *  @q_vector: pointer to q_vector
5745  *
5746  *  Stores a new ITR value based on strictly on packet size.  This
5747  *  algorithm is less sophisticated than that used in igb_update_itr,
5748  *  due to the difficulty of synchronizing statistics across multiple
5749  *  receive rings.  The divisors and thresholds used by this function
5750  *  were determined based on theoretical maximum wire speed and testing
5751  *  data, in order to minimize response time while increasing bulk
5752  *  throughput.
5753  *  This functionality is controlled by ethtool's coalescing settings.
5754  *  NOTE:  This function is called only when operating in a multiqueue
5755  *         receive environment.
5756  **/
5757 static void igb_update_ring_itr(struct igb_q_vector *q_vector)
5758 {
5759 	int new_val = q_vector->itr_val;
5760 	int avg_wire_size = 0;
5761 	struct igb_adapter *adapter = q_vector->adapter;
5762 	unsigned int packets;
5763 
5764 	/* For non-gigabit speeds, just fix the interrupt rate at 4000
5765 	 * ints/sec - ITR timer value of 120 ticks.
5766 	 */
5767 	if (adapter->link_speed != SPEED_1000) {
5768 		new_val = IGB_4K_ITR;
5769 		goto set_itr_val;
5770 	}
5771 
5772 	packets = q_vector->rx.total_packets;
5773 	if (packets)
5774 		avg_wire_size = q_vector->rx.total_bytes / packets;
5775 
5776 	packets = q_vector->tx.total_packets;
5777 	if (packets)
5778 		avg_wire_size = max_t(u32, avg_wire_size,
5779 				      q_vector->tx.total_bytes / packets);
5780 
5781 	/* if avg_wire_size isn't set no work was done */
5782 	if (!avg_wire_size)
5783 		goto clear_counts;
5784 
5785 	/* Add 24 bytes to size to account for CRC, preamble, and gap */
5786 	avg_wire_size += 24;
5787 
5788 	/* Don't starve jumbo frames */
5789 	avg_wire_size = min(avg_wire_size, 3000);
5790 
5791 	/* Give a little boost to mid-size frames */
5792 	if ((avg_wire_size > 300) && (avg_wire_size < 1200))
5793 		new_val = avg_wire_size / 3;
5794 	else
5795 		new_val = avg_wire_size / 2;
5796 
5797 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5798 	if (new_val < IGB_20K_ITR &&
5799 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5800 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5801 		new_val = IGB_20K_ITR;
5802 
5803 set_itr_val:
5804 	if (new_val != q_vector->itr_val) {
5805 		q_vector->itr_val = new_val;
5806 		q_vector->set_itr = 1;
5807 	}
5808 clear_counts:
5809 	q_vector->rx.total_bytes = 0;
5810 	q_vector->rx.total_packets = 0;
5811 	q_vector->tx.total_bytes = 0;
5812 	q_vector->tx.total_packets = 0;
5813 }
5814 
5815 /**
5816  *  igb_update_itr - update the dynamic ITR value based on statistics
5817  *  @q_vector: pointer to q_vector
5818  *  @ring_container: ring info to update the itr for
5819  *
5820  *  Stores a new ITR value based on packets and byte
5821  *  counts during the last interrupt.  The advantage of per interrupt
5822  *  computation is faster updates and more accurate ITR for the current
5823  *  traffic pattern.  Constants in this function were computed
5824  *  based on theoretical maximum wire speed and thresholds were set based
5825  *  on testing data as well as attempting to minimize response time
5826  *  while increasing bulk throughput.
5827  *  This functionality is controlled by ethtool's coalescing settings.
5828  *  NOTE:  These calculations are only valid when operating in a single-
5829  *         queue environment.
5830  **/
5831 static void igb_update_itr(struct igb_q_vector *q_vector,
5832 			   struct igb_ring_container *ring_container)
5833 {
5834 	unsigned int packets = ring_container->total_packets;
5835 	unsigned int bytes = ring_container->total_bytes;
5836 	u8 itrval = ring_container->itr;
5837 
5838 	/* no packets, exit with status unchanged */
5839 	if (packets == 0)
5840 		return;
5841 
5842 	switch (itrval) {
5843 	case lowest_latency:
5844 		/* handle TSO and jumbo frames */
5845 		if (bytes/packets > 8000)
5846 			itrval = bulk_latency;
5847 		else if ((packets < 5) && (bytes > 512))
5848 			itrval = low_latency;
5849 		break;
5850 	case low_latency:  /* 50 usec aka 20000 ints/s */
5851 		if (bytes > 10000) {
5852 			/* this if handles the TSO accounting */
5853 			if (bytes/packets > 8000)
5854 				itrval = bulk_latency;
5855 			else if ((packets < 10) || ((bytes/packets) > 1200))
5856 				itrval = bulk_latency;
5857 			else if ((packets > 35))
5858 				itrval = lowest_latency;
5859 		} else if (bytes/packets > 2000) {
5860 			itrval = bulk_latency;
5861 		} else if (packets <= 2 && bytes < 512) {
5862 			itrval = lowest_latency;
5863 		}
5864 		break;
5865 	case bulk_latency: /* 250 usec aka 4000 ints/s */
5866 		if (bytes > 25000) {
5867 			if (packets > 35)
5868 				itrval = low_latency;
5869 		} else if (bytes < 1500) {
5870 			itrval = low_latency;
5871 		}
5872 		break;
5873 	}
5874 
5875 	/* clear work counters since we have the values we need */
5876 	ring_container->total_bytes = 0;
5877 	ring_container->total_packets = 0;
5878 
5879 	/* write updated itr to ring container */
5880 	ring_container->itr = itrval;
5881 }
5882 
5883 static void igb_set_itr(struct igb_q_vector *q_vector)
5884 {
5885 	struct igb_adapter *adapter = q_vector->adapter;
5886 	u32 new_itr = q_vector->itr_val;
5887 	u8 current_itr = 0;
5888 
5889 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
5890 	if (adapter->link_speed != SPEED_1000) {
5891 		current_itr = 0;
5892 		new_itr = IGB_4K_ITR;
5893 		goto set_itr_now;
5894 	}
5895 
5896 	igb_update_itr(q_vector, &q_vector->tx);
5897 	igb_update_itr(q_vector, &q_vector->rx);
5898 
5899 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
5900 
5901 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5902 	if (current_itr == lowest_latency &&
5903 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5904 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5905 		current_itr = low_latency;
5906 
5907 	switch (current_itr) {
5908 	/* counts and packets in update_itr are dependent on these numbers */
5909 	case lowest_latency:
5910 		new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
5911 		break;
5912 	case low_latency:
5913 		new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
5914 		break;
5915 	case bulk_latency:
5916 		new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
5917 		break;
5918 	default:
5919 		break;
5920 	}
5921 
5922 set_itr_now:
5923 	if (new_itr != q_vector->itr_val) {
5924 		/* this attempts to bias the interrupt rate towards Bulk
5925 		 * by adding intermediate steps when interrupt rate is
5926 		 * increasing
5927 		 */
5928 		new_itr = new_itr > q_vector->itr_val ?
5929 			  max((new_itr * q_vector->itr_val) /
5930 			  (new_itr + (q_vector->itr_val >> 2)),
5931 			  new_itr) : new_itr;
5932 		/* Don't write the value here; it resets the adapter's
5933 		 * internal timer, and causes us to delay far longer than
5934 		 * we should between interrupts.  Instead, we write the ITR
5935 		 * value at the beginning of the next interrupt so the timing
5936 		 * ends up being correct.
5937 		 */
5938 		q_vector->itr_val = new_itr;
5939 		q_vector->set_itr = 1;
5940 	}
5941 }
5942 
5943 static void igb_tx_ctxtdesc(struct igb_ring *tx_ring,
5944 			    struct igb_tx_buffer *first,
5945 			    u32 vlan_macip_lens, u32 type_tucmd,
5946 			    u32 mss_l4len_idx)
5947 {
5948 	struct e1000_adv_tx_context_desc *context_desc;
5949 	u16 i = tx_ring->next_to_use;
5950 	struct timespec64 ts;
5951 
5952 	context_desc = IGB_TX_CTXTDESC(tx_ring, i);
5953 
5954 	i++;
5955 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
5956 
5957 	/* set bits to identify this as an advanced context descriptor */
5958 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
5959 
5960 	/* For 82575, context index must be unique per ring. */
5961 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5962 		mss_l4len_idx |= tx_ring->reg_idx << 4;
5963 
5964 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
5965 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
5966 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
5967 
5968 	/* We assume there is always a valid tx time available. Invalid times
5969 	 * should have been handled by the upper layers.
5970 	 */
5971 	if (tx_ring->launchtime_enable) {
5972 		ts = ktime_to_timespec64(first->skb->tstamp);
5973 		skb_txtime_consumed(first->skb);
5974 		context_desc->seqnum_seed = cpu_to_le32(ts.tv_nsec / 32);
5975 	} else {
5976 		context_desc->seqnum_seed = 0;
5977 	}
5978 }
5979 
5980 static int igb_tso(struct igb_ring *tx_ring,
5981 		   struct igb_tx_buffer *first,
5982 		   u8 *hdr_len)
5983 {
5984 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
5985 	struct sk_buff *skb = first->skb;
5986 	union {
5987 		struct iphdr *v4;
5988 		struct ipv6hdr *v6;
5989 		unsigned char *hdr;
5990 	} ip;
5991 	union {
5992 		struct tcphdr *tcp;
5993 		struct udphdr *udp;
5994 		unsigned char *hdr;
5995 	} l4;
5996 	u32 paylen, l4_offset;
5997 	int err;
5998 
5999 	if (skb->ip_summed != CHECKSUM_PARTIAL)
6000 		return 0;
6001 
6002 	if (!skb_is_gso(skb))
6003 		return 0;
6004 
6005 	err = skb_cow_head(skb, 0);
6006 	if (err < 0)
6007 		return err;
6008 
6009 	ip.hdr = skb_network_header(skb);
6010 	l4.hdr = skb_checksum_start(skb);
6011 
6012 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6013 	type_tucmd = (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ?
6014 		      E1000_ADVTXD_TUCMD_L4T_UDP : E1000_ADVTXD_TUCMD_L4T_TCP;
6015 
6016 	/* initialize outer IP header fields */
6017 	if (ip.v4->version == 4) {
6018 		unsigned char *csum_start = skb_checksum_start(skb);
6019 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
6020 
6021 		/* IP header will have to cancel out any data that
6022 		 * is not a part of the outer IP header
6023 		 */
6024 		ip.v4->check = csum_fold(csum_partial(trans_start,
6025 						      csum_start - trans_start,
6026 						      0));
6027 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
6028 
6029 		ip.v4->tot_len = 0;
6030 		first->tx_flags |= IGB_TX_FLAGS_TSO |
6031 				   IGB_TX_FLAGS_CSUM |
6032 				   IGB_TX_FLAGS_IPV4;
6033 	} else {
6034 		ip.v6->payload_len = 0;
6035 		first->tx_flags |= IGB_TX_FLAGS_TSO |
6036 				   IGB_TX_FLAGS_CSUM;
6037 	}
6038 
6039 	/* determine offset of inner transport header */
6040 	l4_offset = l4.hdr - skb->data;
6041 
6042 	/* remove payload length from inner checksum */
6043 	paylen = skb->len - l4_offset;
6044 	if (type_tucmd & E1000_ADVTXD_TUCMD_L4T_TCP) {
6045 		/* compute length of segmentation header */
6046 		*hdr_len = (l4.tcp->doff * 4) + l4_offset;
6047 		csum_replace_by_diff(&l4.tcp->check,
6048 			(__force __wsum)htonl(paylen));
6049 	} else {
6050 		/* compute length of segmentation header */
6051 		*hdr_len = sizeof(*l4.udp) + l4_offset;
6052 		csum_replace_by_diff(&l4.udp->check,
6053 				     (__force __wsum)htonl(paylen));
6054 	}
6055 
6056 	/* update gso size and bytecount with header size */
6057 	first->gso_segs = skb_shinfo(skb)->gso_segs;
6058 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
6059 
6060 	/* MSS L4LEN IDX */
6061 	mss_l4len_idx = (*hdr_len - l4_offset) << E1000_ADVTXD_L4LEN_SHIFT;
6062 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
6063 
6064 	/* VLAN MACLEN IPLEN */
6065 	vlan_macip_lens = l4.hdr - ip.hdr;
6066 	vlan_macip_lens |= (ip.hdr - skb->data) << E1000_ADVTXD_MACLEN_SHIFT;
6067 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
6068 
6069 	igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens,
6070 			type_tucmd, mss_l4len_idx);
6071 
6072 	return 1;
6073 }
6074 
6075 static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
6076 {
6077 	struct sk_buff *skb = first->skb;
6078 	u32 vlan_macip_lens = 0;
6079 	u32 type_tucmd = 0;
6080 
6081 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
6082 csum_failed:
6083 		if (!(first->tx_flags & IGB_TX_FLAGS_VLAN) &&
6084 		    !tx_ring->launchtime_enable)
6085 			return;
6086 		goto no_csum;
6087 	}
6088 
6089 	switch (skb->csum_offset) {
6090 	case offsetof(struct tcphdr, check):
6091 		type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
6092 		fallthrough;
6093 	case offsetof(struct udphdr, check):
6094 		break;
6095 	case offsetof(struct sctphdr, checksum):
6096 		/* validate that this is actually an SCTP request */
6097 		if (skb_csum_is_sctp(skb)) {
6098 			type_tucmd = E1000_ADVTXD_TUCMD_L4T_SCTP;
6099 			break;
6100 		}
6101 		fallthrough;
6102 	default:
6103 		skb_checksum_help(skb);
6104 		goto csum_failed;
6105 	}
6106 
6107 	/* update TX checksum flag */
6108 	first->tx_flags |= IGB_TX_FLAGS_CSUM;
6109 	vlan_macip_lens = skb_checksum_start_offset(skb) -
6110 			  skb_network_offset(skb);
6111 no_csum:
6112 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
6113 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
6114 
6115 	igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens, type_tucmd, 0);
6116 }
6117 
6118 #define IGB_SET_FLAG(_input, _flag, _result) \
6119 	((_flag <= _result) ? \
6120 	 ((u32)(_input & _flag) * (_result / _flag)) : \
6121 	 ((u32)(_input & _flag) / (_flag / _result)))
6122 
6123 static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
6124 {
6125 	/* set type for advanced descriptor with frame checksum insertion */
6126 	u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
6127 		       E1000_ADVTXD_DCMD_DEXT |
6128 		       E1000_ADVTXD_DCMD_IFCS;
6129 
6130 	/* set HW vlan bit if vlan is present */
6131 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
6132 				 (E1000_ADVTXD_DCMD_VLE));
6133 
6134 	/* set segmentation bits for TSO */
6135 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
6136 				 (E1000_ADVTXD_DCMD_TSE));
6137 
6138 	/* set timestamp bit if present */
6139 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
6140 				 (E1000_ADVTXD_MAC_TSTAMP));
6141 
6142 	/* insert frame checksum */
6143 	cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
6144 
6145 	return cmd_type;
6146 }
6147 
6148 static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
6149 				 union e1000_adv_tx_desc *tx_desc,
6150 				 u32 tx_flags, unsigned int paylen)
6151 {
6152 	u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
6153 
6154 	/* 82575 requires a unique index per ring */
6155 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
6156 		olinfo_status |= tx_ring->reg_idx << 4;
6157 
6158 	/* insert L4 checksum */
6159 	olinfo_status |= IGB_SET_FLAG(tx_flags,
6160 				      IGB_TX_FLAGS_CSUM,
6161 				      (E1000_TXD_POPTS_TXSM << 8));
6162 
6163 	/* insert IPv4 checksum */
6164 	olinfo_status |= IGB_SET_FLAG(tx_flags,
6165 				      IGB_TX_FLAGS_IPV4,
6166 				      (E1000_TXD_POPTS_IXSM << 8));
6167 
6168 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6169 }
6170 
6171 static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
6172 {
6173 	struct net_device *netdev = tx_ring->netdev;
6174 
6175 	netif_stop_subqueue(netdev, tx_ring->queue_index);
6176 
6177 	/* Herbert's original patch had:
6178 	 *  smp_mb__after_netif_stop_queue();
6179 	 * but since that doesn't exist yet, just open code it.
6180 	 */
6181 	smp_mb();
6182 
6183 	/* We need to check again in a case another CPU has just
6184 	 * made room available.
6185 	 */
6186 	if (igb_desc_unused(tx_ring) < size)
6187 		return -EBUSY;
6188 
6189 	/* A reprieve! */
6190 	netif_wake_subqueue(netdev, tx_ring->queue_index);
6191 
6192 	u64_stats_update_begin(&tx_ring->tx_syncp2);
6193 	tx_ring->tx_stats.restart_queue2++;
6194 	u64_stats_update_end(&tx_ring->tx_syncp2);
6195 
6196 	return 0;
6197 }
6198 
6199 static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
6200 {
6201 	if (igb_desc_unused(tx_ring) >= size)
6202 		return 0;
6203 	return __igb_maybe_stop_tx(tx_ring, size);
6204 }
6205 
6206 static int igb_tx_map(struct igb_ring *tx_ring,
6207 		      struct igb_tx_buffer *first,
6208 		      const u8 hdr_len)
6209 {
6210 	struct sk_buff *skb = first->skb;
6211 	struct igb_tx_buffer *tx_buffer;
6212 	union e1000_adv_tx_desc *tx_desc;
6213 	skb_frag_t *frag;
6214 	dma_addr_t dma;
6215 	unsigned int data_len, size;
6216 	u32 tx_flags = first->tx_flags;
6217 	u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
6218 	u16 i = tx_ring->next_to_use;
6219 
6220 	tx_desc = IGB_TX_DESC(tx_ring, i);
6221 
6222 	igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
6223 
6224 	size = skb_headlen(skb);
6225 	data_len = skb->data_len;
6226 
6227 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6228 
6229 	tx_buffer = first;
6230 
6231 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
6232 		if (dma_mapping_error(tx_ring->dev, dma))
6233 			goto dma_error;
6234 
6235 		/* record length, and DMA address */
6236 		dma_unmap_len_set(tx_buffer, len, size);
6237 		dma_unmap_addr_set(tx_buffer, dma, dma);
6238 
6239 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6240 
6241 		while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
6242 			tx_desc->read.cmd_type_len =
6243 				cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
6244 
6245 			i++;
6246 			tx_desc++;
6247 			if (i == tx_ring->count) {
6248 				tx_desc = IGB_TX_DESC(tx_ring, 0);
6249 				i = 0;
6250 			}
6251 			tx_desc->read.olinfo_status = 0;
6252 
6253 			dma += IGB_MAX_DATA_PER_TXD;
6254 			size -= IGB_MAX_DATA_PER_TXD;
6255 
6256 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
6257 		}
6258 
6259 		if (likely(!data_len))
6260 			break;
6261 
6262 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
6263 
6264 		i++;
6265 		tx_desc++;
6266 		if (i == tx_ring->count) {
6267 			tx_desc = IGB_TX_DESC(tx_ring, 0);
6268 			i = 0;
6269 		}
6270 		tx_desc->read.olinfo_status = 0;
6271 
6272 		size = skb_frag_size(frag);
6273 		data_len -= size;
6274 
6275 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
6276 				       size, DMA_TO_DEVICE);
6277 
6278 		tx_buffer = &tx_ring->tx_buffer_info[i];
6279 	}
6280 
6281 	/* write last descriptor with RS and EOP bits */
6282 	cmd_type |= size | IGB_TXD_DCMD;
6283 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6284 
6285 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
6286 
6287 	/* set the timestamp */
6288 	first->time_stamp = jiffies;
6289 
6290 	skb_tx_timestamp(skb);
6291 
6292 	/* Force memory writes to complete before letting h/w know there
6293 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
6294 	 * memory model archs, such as IA-64).
6295 	 *
6296 	 * We also need this memory barrier to make certain all of the
6297 	 * status bits have been updated before next_to_watch is written.
6298 	 */
6299 	dma_wmb();
6300 
6301 	/* set next_to_watch value indicating a packet is present */
6302 	first->next_to_watch = tx_desc;
6303 
6304 	i++;
6305 	if (i == tx_ring->count)
6306 		i = 0;
6307 
6308 	tx_ring->next_to_use = i;
6309 
6310 	/* Make sure there is space in the ring for the next send. */
6311 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
6312 
6313 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
6314 		writel(i, tx_ring->tail);
6315 	}
6316 	return 0;
6317 
6318 dma_error:
6319 	dev_err(tx_ring->dev, "TX DMA map failed\n");
6320 	tx_buffer = &tx_ring->tx_buffer_info[i];
6321 
6322 	/* clear dma mappings for failed tx_buffer_info map */
6323 	while (tx_buffer != first) {
6324 		if (dma_unmap_len(tx_buffer, len))
6325 			dma_unmap_page(tx_ring->dev,
6326 				       dma_unmap_addr(tx_buffer, dma),
6327 				       dma_unmap_len(tx_buffer, len),
6328 				       DMA_TO_DEVICE);
6329 		dma_unmap_len_set(tx_buffer, len, 0);
6330 
6331 		if (i-- == 0)
6332 			i += tx_ring->count;
6333 		tx_buffer = &tx_ring->tx_buffer_info[i];
6334 	}
6335 
6336 	if (dma_unmap_len(tx_buffer, len))
6337 		dma_unmap_single(tx_ring->dev,
6338 				 dma_unmap_addr(tx_buffer, dma),
6339 				 dma_unmap_len(tx_buffer, len),
6340 				 DMA_TO_DEVICE);
6341 	dma_unmap_len_set(tx_buffer, len, 0);
6342 
6343 	dev_kfree_skb_any(tx_buffer->skb);
6344 	tx_buffer->skb = NULL;
6345 
6346 	tx_ring->next_to_use = i;
6347 
6348 	return -1;
6349 }
6350 
6351 int igb_xmit_xdp_ring(struct igb_adapter *adapter,
6352 		      struct igb_ring *tx_ring,
6353 		      struct xdp_frame *xdpf)
6354 {
6355 	struct skb_shared_info *sinfo = xdp_get_shared_info_from_frame(xdpf);
6356 	u8 nr_frags = unlikely(xdp_frame_has_frags(xdpf)) ? sinfo->nr_frags : 0;
6357 	u16 count, i, index = tx_ring->next_to_use;
6358 	struct igb_tx_buffer *tx_head = &tx_ring->tx_buffer_info[index];
6359 	struct igb_tx_buffer *tx_buffer = tx_head;
6360 	union e1000_adv_tx_desc *tx_desc = IGB_TX_DESC(tx_ring, index);
6361 	u32 len = xdpf->len, cmd_type, olinfo_status;
6362 	void *data = xdpf->data;
6363 
6364 	count = TXD_USE_COUNT(len);
6365 	for (i = 0; i < nr_frags; i++)
6366 		count += TXD_USE_COUNT(skb_frag_size(&sinfo->frags[i]));
6367 
6368 	if (igb_maybe_stop_tx(tx_ring, count + 3))
6369 		return IGB_XDP_CONSUMED;
6370 
6371 	i = 0;
6372 	/* record the location of the first descriptor for this packet */
6373 	tx_head->bytecount = xdp_get_frame_len(xdpf);
6374 	tx_head->type = IGB_TYPE_XDP;
6375 	tx_head->gso_segs = 1;
6376 	tx_head->xdpf = xdpf;
6377 
6378 	olinfo_status = tx_head->bytecount << E1000_ADVTXD_PAYLEN_SHIFT;
6379 	/* 82575 requires a unique index per ring */
6380 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
6381 		olinfo_status |= tx_ring->reg_idx << 4;
6382 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6383 
6384 	for (;;) {
6385 		dma_addr_t dma;
6386 
6387 		dma = dma_map_single(tx_ring->dev, data, len, DMA_TO_DEVICE);
6388 		if (dma_mapping_error(tx_ring->dev, dma))
6389 			goto unmap;
6390 
6391 		/* record length, and DMA address */
6392 		dma_unmap_len_set(tx_buffer, len, len);
6393 		dma_unmap_addr_set(tx_buffer, dma, dma);
6394 
6395 		/* put descriptor type bits */
6396 		cmd_type = E1000_ADVTXD_DTYP_DATA | E1000_ADVTXD_DCMD_DEXT |
6397 			   E1000_ADVTXD_DCMD_IFCS | len;
6398 
6399 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6400 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6401 
6402 		tx_buffer->protocol = 0;
6403 
6404 		if (++index == tx_ring->count)
6405 			index = 0;
6406 
6407 		if (i == nr_frags)
6408 			break;
6409 
6410 		tx_buffer = &tx_ring->tx_buffer_info[index];
6411 		tx_desc = IGB_TX_DESC(tx_ring, index);
6412 		tx_desc->read.olinfo_status = 0;
6413 
6414 		data = skb_frag_address(&sinfo->frags[i]);
6415 		len = skb_frag_size(&sinfo->frags[i]);
6416 		i++;
6417 	}
6418 	tx_desc->read.cmd_type_len |= cpu_to_le32(IGB_TXD_DCMD);
6419 
6420 	netdev_tx_sent_queue(txring_txq(tx_ring), tx_head->bytecount);
6421 	/* set the timestamp */
6422 	tx_head->time_stamp = jiffies;
6423 
6424 	/* Avoid any potential race with xdp_xmit and cleanup */
6425 	smp_wmb();
6426 
6427 	/* set next_to_watch value indicating a packet is present */
6428 	tx_head->next_to_watch = tx_desc;
6429 	tx_ring->next_to_use = index;
6430 
6431 	/* Make sure there is space in the ring for the next send. */
6432 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
6433 
6434 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more())
6435 		writel(index, tx_ring->tail);
6436 
6437 	return IGB_XDP_TX;
6438 
6439 unmap:
6440 	for (;;) {
6441 		tx_buffer = &tx_ring->tx_buffer_info[index];
6442 		if (dma_unmap_len(tx_buffer, len))
6443 			dma_unmap_page(tx_ring->dev,
6444 				       dma_unmap_addr(tx_buffer, dma),
6445 				       dma_unmap_len(tx_buffer, len),
6446 				       DMA_TO_DEVICE);
6447 		dma_unmap_len_set(tx_buffer, len, 0);
6448 		if (tx_buffer == tx_head)
6449 			break;
6450 
6451 		if (!index)
6452 			index += tx_ring->count;
6453 		index--;
6454 	}
6455 
6456 	return IGB_XDP_CONSUMED;
6457 }
6458 
6459 netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
6460 				struct igb_ring *tx_ring)
6461 {
6462 	struct igb_tx_buffer *first;
6463 	int tso;
6464 	u32 tx_flags = 0;
6465 	unsigned short f;
6466 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
6467 	__be16 protocol = vlan_get_protocol(skb);
6468 	u8 hdr_len = 0;
6469 
6470 	/* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
6471 	 *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
6472 	 *       + 2 desc gap to keep tail from touching head,
6473 	 *       + 1 desc for context descriptor,
6474 	 * otherwise try next time
6475 	 */
6476 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6477 		count += TXD_USE_COUNT(skb_frag_size(
6478 						&skb_shinfo(skb)->frags[f]));
6479 
6480 	if (igb_maybe_stop_tx(tx_ring, count + 3)) {
6481 		/* this is a hard error */
6482 		return NETDEV_TX_BUSY;
6483 	}
6484 
6485 	/* record the location of the first descriptor for this packet */
6486 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6487 	first->type = IGB_TYPE_SKB;
6488 	first->skb = skb;
6489 	first->bytecount = skb->len;
6490 	first->gso_segs = 1;
6491 
6492 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
6493 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6494 
6495 		if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
6496 		    !test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
6497 					   &adapter->state)) {
6498 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
6499 			tx_flags |= IGB_TX_FLAGS_TSTAMP;
6500 
6501 			adapter->ptp_tx_skb = skb_get(skb);
6502 			adapter->ptp_tx_start = jiffies;
6503 			if (adapter->hw.mac.type == e1000_82576)
6504 				schedule_work(&adapter->ptp_tx_work);
6505 		} else {
6506 			adapter->tx_hwtstamp_skipped++;
6507 		}
6508 	}
6509 
6510 	if (skb_vlan_tag_present(skb)) {
6511 		tx_flags |= IGB_TX_FLAGS_VLAN;
6512 		tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
6513 	}
6514 
6515 	/* record initial flags and protocol */
6516 	first->tx_flags = tx_flags;
6517 	first->protocol = protocol;
6518 
6519 	tso = igb_tso(tx_ring, first, &hdr_len);
6520 	if (tso < 0)
6521 		goto out_drop;
6522 	else if (!tso)
6523 		igb_tx_csum(tx_ring, first);
6524 
6525 	if (igb_tx_map(tx_ring, first, hdr_len))
6526 		goto cleanup_tx_tstamp;
6527 
6528 	return NETDEV_TX_OK;
6529 
6530 out_drop:
6531 	dev_kfree_skb_any(first->skb);
6532 	first->skb = NULL;
6533 cleanup_tx_tstamp:
6534 	if (unlikely(tx_flags & IGB_TX_FLAGS_TSTAMP)) {
6535 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6536 
6537 		dev_kfree_skb_any(adapter->ptp_tx_skb);
6538 		adapter->ptp_tx_skb = NULL;
6539 		if (adapter->hw.mac.type == e1000_82576)
6540 			cancel_work_sync(&adapter->ptp_tx_work);
6541 		clear_bit_unlock(__IGB_PTP_TX_IN_PROGRESS, &adapter->state);
6542 	}
6543 
6544 	return NETDEV_TX_OK;
6545 }
6546 
6547 static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
6548 						    struct sk_buff *skb)
6549 {
6550 	unsigned int r_idx = skb->queue_mapping;
6551 
6552 	if (r_idx >= adapter->num_tx_queues)
6553 		r_idx = r_idx % adapter->num_tx_queues;
6554 
6555 	return adapter->tx_ring[r_idx];
6556 }
6557 
6558 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
6559 				  struct net_device *netdev)
6560 {
6561 	struct igb_adapter *adapter = netdev_priv(netdev);
6562 
6563 	/* The minimum packet size with TCTL.PSP set is 17 so pad the skb
6564 	 * in order to meet this minimum size requirement.
6565 	 */
6566 	if (skb_put_padto(skb, 17))
6567 		return NETDEV_TX_OK;
6568 
6569 	return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
6570 }
6571 
6572 /**
6573  *  igb_tx_timeout - Respond to a Tx Hang
6574  *  @netdev: network interface device structure
6575  *  @txqueue: number of the Tx queue that hung (unused)
6576  **/
6577 static void igb_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
6578 {
6579 	struct igb_adapter *adapter = netdev_priv(netdev);
6580 	struct e1000_hw *hw = &adapter->hw;
6581 
6582 	/* Do the reset outside of interrupt context */
6583 	adapter->tx_timeout_count++;
6584 
6585 	if (hw->mac.type >= e1000_82580)
6586 		hw->dev_spec._82575.global_device_reset = true;
6587 
6588 	schedule_work(&adapter->reset_task);
6589 	wr32(E1000_EICS,
6590 	     (adapter->eims_enable_mask & ~adapter->eims_other));
6591 }
6592 
6593 static void igb_reset_task(struct work_struct *work)
6594 {
6595 	struct igb_adapter *adapter;
6596 	adapter = container_of(work, struct igb_adapter, reset_task);
6597 
6598 	rtnl_lock();
6599 	/* If we're already down or resetting, just bail */
6600 	if (test_bit(__IGB_DOWN, &adapter->state) ||
6601 	    test_bit(__IGB_RESETTING, &adapter->state)) {
6602 		rtnl_unlock();
6603 		return;
6604 	}
6605 
6606 	igb_dump(adapter);
6607 	netdev_err(adapter->netdev, "Reset adapter\n");
6608 	igb_reinit_locked(adapter);
6609 	rtnl_unlock();
6610 }
6611 
6612 /**
6613  *  igb_get_stats64 - Get System Network Statistics
6614  *  @netdev: network interface device structure
6615  *  @stats: rtnl_link_stats64 pointer
6616  **/
6617 static void igb_get_stats64(struct net_device *netdev,
6618 			    struct rtnl_link_stats64 *stats)
6619 {
6620 	struct igb_adapter *adapter = netdev_priv(netdev);
6621 
6622 	spin_lock(&adapter->stats64_lock);
6623 	igb_update_stats(adapter);
6624 	memcpy(stats, &adapter->stats64, sizeof(*stats));
6625 	spin_unlock(&adapter->stats64_lock);
6626 }
6627 
6628 /**
6629  *  igb_change_mtu - Change the Maximum Transfer Unit
6630  *  @netdev: network interface device structure
6631  *  @new_mtu: new value for maximum frame size
6632  *
6633  *  Returns 0 on success, negative on failure
6634  **/
6635 static int igb_change_mtu(struct net_device *netdev, int new_mtu)
6636 {
6637 	struct igb_adapter *adapter = netdev_priv(netdev);
6638 	int max_frame = new_mtu + IGB_ETH_PKT_HDR_PAD;
6639 
6640 	if (adapter->xdp_prog) {
6641 		int i;
6642 
6643 		for (i = 0; i < adapter->num_rx_queues; i++) {
6644 			struct igb_ring *ring = adapter->rx_ring[i];
6645 
6646 			if (max_frame > igb_rx_bufsz(ring)) {
6647 				netdev_warn(adapter->netdev,
6648 					    "Requested MTU size is not supported with XDP. Max frame size is %d\n",
6649 					    max_frame);
6650 				return -EINVAL;
6651 			}
6652 		}
6653 	}
6654 
6655 	/* adjust max frame to be at least the size of a standard frame */
6656 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
6657 		max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
6658 
6659 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
6660 		usleep_range(1000, 2000);
6661 
6662 	/* igb_down has a dependency on max_frame_size */
6663 	adapter->max_frame_size = max_frame;
6664 
6665 	if (netif_running(netdev))
6666 		igb_down(adapter);
6667 
6668 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6669 		   netdev->mtu, new_mtu);
6670 	netdev->mtu = new_mtu;
6671 
6672 	if (netif_running(netdev))
6673 		igb_up(adapter);
6674 	else
6675 		igb_reset(adapter);
6676 
6677 	clear_bit(__IGB_RESETTING, &adapter->state);
6678 
6679 	return 0;
6680 }
6681 
6682 /**
6683  *  igb_update_stats - Update the board statistics counters
6684  *  @adapter: board private structure
6685  **/
6686 void igb_update_stats(struct igb_adapter *adapter)
6687 {
6688 	struct rtnl_link_stats64 *net_stats = &adapter->stats64;
6689 	struct e1000_hw *hw = &adapter->hw;
6690 	struct pci_dev *pdev = adapter->pdev;
6691 	u32 reg, mpc;
6692 	int i;
6693 	u64 bytes, packets;
6694 	unsigned int start;
6695 	u64 _bytes, _packets;
6696 
6697 	/* Prevent stats update while adapter is being reset, or if the pci
6698 	 * connection is down.
6699 	 */
6700 	if (adapter->link_speed == 0)
6701 		return;
6702 	if (pci_channel_offline(pdev))
6703 		return;
6704 
6705 	bytes = 0;
6706 	packets = 0;
6707 
6708 	rcu_read_lock();
6709 	for (i = 0; i < adapter->num_rx_queues; i++) {
6710 		struct igb_ring *ring = adapter->rx_ring[i];
6711 		u32 rqdpc = rd32(E1000_RQDPC(i));
6712 		if (hw->mac.type >= e1000_i210)
6713 			wr32(E1000_RQDPC(i), 0);
6714 
6715 		if (rqdpc) {
6716 			ring->rx_stats.drops += rqdpc;
6717 			net_stats->rx_fifo_errors += rqdpc;
6718 		}
6719 
6720 		do {
6721 			start = u64_stats_fetch_begin(&ring->rx_syncp);
6722 			_bytes = ring->rx_stats.bytes;
6723 			_packets = ring->rx_stats.packets;
6724 		} while (u64_stats_fetch_retry(&ring->rx_syncp, start));
6725 		bytes += _bytes;
6726 		packets += _packets;
6727 	}
6728 
6729 	net_stats->rx_bytes = bytes;
6730 	net_stats->rx_packets = packets;
6731 
6732 	bytes = 0;
6733 	packets = 0;
6734 	for (i = 0; i < adapter->num_tx_queues; i++) {
6735 		struct igb_ring *ring = adapter->tx_ring[i];
6736 		do {
6737 			start = u64_stats_fetch_begin(&ring->tx_syncp);
6738 			_bytes = ring->tx_stats.bytes;
6739 			_packets = ring->tx_stats.packets;
6740 		} while (u64_stats_fetch_retry(&ring->tx_syncp, start));
6741 		bytes += _bytes;
6742 		packets += _packets;
6743 	}
6744 	net_stats->tx_bytes = bytes;
6745 	net_stats->tx_packets = packets;
6746 	rcu_read_unlock();
6747 
6748 	/* read stats registers */
6749 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
6750 	adapter->stats.gprc += rd32(E1000_GPRC);
6751 	adapter->stats.gorc += rd32(E1000_GORCL);
6752 	rd32(E1000_GORCH); /* clear GORCL */
6753 	adapter->stats.bprc += rd32(E1000_BPRC);
6754 	adapter->stats.mprc += rd32(E1000_MPRC);
6755 	adapter->stats.roc += rd32(E1000_ROC);
6756 
6757 	adapter->stats.prc64 += rd32(E1000_PRC64);
6758 	adapter->stats.prc127 += rd32(E1000_PRC127);
6759 	adapter->stats.prc255 += rd32(E1000_PRC255);
6760 	adapter->stats.prc511 += rd32(E1000_PRC511);
6761 	adapter->stats.prc1023 += rd32(E1000_PRC1023);
6762 	adapter->stats.prc1522 += rd32(E1000_PRC1522);
6763 	adapter->stats.symerrs += rd32(E1000_SYMERRS);
6764 	adapter->stats.sec += rd32(E1000_SEC);
6765 
6766 	mpc = rd32(E1000_MPC);
6767 	adapter->stats.mpc += mpc;
6768 	net_stats->rx_fifo_errors += mpc;
6769 	adapter->stats.scc += rd32(E1000_SCC);
6770 	adapter->stats.ecol += rd32(E1000_ECOL);
6771 	adapter->stats.mcc += rd32(E1000_MCC);
6772 	adapter->stats.latecol += rd32(E1000_LATECOL);
6773 	adapter->stats.dc += rd32(E1000_DC);
6774 	adapter->stats.rlec += rd32(E1000_RLEC);
6775 	adapter->stats.xonrxc += rd32(E1000_XONRXC);
6776 	adapter->stats.xontxc += rd32(E1000_XONTXC);
6777 	adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
6778 	adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
6779 	adapter->stats.fcruc += rd32(E1000_FCRUC);
6780 	adapter->stats.gptc += rd32(E1000_GPTC);
6781 	adapter->stats.gotc += rd32(E1000_GOTCL);
6782 	rd32(E1000_GOTCH); /* clear GOTCL */
6783 	adapter->stats.rnbc += rd32(E1000_RNBC);
6784 	adapter->stats.ruc += rd32(E1000_RUC);
6785 	adapter->stats.rfc += rd32(E1000_RFC);
6786 	adapter->stats.rjc += rd32(E1000_RJC);
6787 	adapter->stats.tor += rd32(E1000_TORH);
6788 	adapter->stats.tot += rd32(E1000_TOTH);
6789 	adapter->stats.tpr += rd32(E1000_TPR);
6790 
6791 	adapter->stats.ptc64 += rd32(E1000_PTC64);
6792 	adapter->stats.ptc127 += rd32(E1000_PTC127);
6793 	adapter->stats.ptc255 += rd32(E1000_PTC255);
6794 	adapter->stats.ptc511 += rd32(E1000_PTC511);
6795 	adapter->stats.ptc1023 += rd32(E1000_PTC1023);
6796 	adapter->stats.ptc1522 += rd32(E1000_PTC1522);
6797 
6798 	adapter->stats.mptc += rd32(E1000_MPTC);
6799 	adapter->stats.bptc += rd32(E1000_BPTC);
6800 
6801 	adapter->stats.tpt += rd32(E1000_TPT);
6802 	adapter->stats.colc += rd32(E1000_COLC);
6803 
6804 	adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
6805 	/* read internal phy specific stats */
6806 	reg = rd32(E1000_CTRL_EXT);
6807 	if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
6808 		adapter->stats.rxerrc += rd32(E1000_RXERRC);
6809 
6810 		/* this stat has invalid values on i210/i211 */
6811 		if ((hw->mac.type != e1000_i210) &&
6812 		    (hw->mac.type != e1000_i211))
6813 			adapter->stats.tncrs += rd32(E1000_TNCRS);
6814 	}
6815 
6816 	adapter->stats.tsctc += rd32(E1000_TSCTC);
6817 	adapter->stats.tsctfc += rd32(E1000_TSCTFC);
6818 
6819 	adapter->stats.iac += rd32(E1000_IAC);
6820 	adapter->stats.icrxoc += rd32(E1000_ICRXOC);
6821 	adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
6822 	adapter->stats.icrxatc += rd32(E1000_ICRXATC);
6823 	adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
6824 	adapter->stats.ictxatc += rd32(E1000_ICTXATC);
6825 	adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
6826 	adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
6827 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
6828 
6829 	/* Fill out the OS statistics structure */
6830 	net_stats->multicast = adapter->stats.mprc;
6831 	net_stats->collisions = adapter->stats.colc;
6832 
6833 	/* Rx Errors */
6834 
6835 	/* RLEC on some newer hardware can be incorrect so build
6836 	 * our own version based on RUC and ROC
6837 	 */
6838 	net_stats->rx_errors = adapter->stats.rxerrc +
6839 		adapter->stats.crcerrs + adapter->stats.algnerrc +
6840 		adapter->stats.ruc + adapter->stats.roc +
6841 		adapter->stats.cexterr;
6842 	net_stats->rx_length_errors = adapter->stats.ruc +
6843 				      adapter->stats.roc;
6844 	net_stats->rx_crc_errors = adapter->stats.crcerrs;
6845 	net_stats->rx_frame_errors = adapter->stats.algnerrc;
6846 	net_stats->rx_missed_errors = adapter->stats.mpc;
6847 
6848 	/* Tx Errors */
6849 	net_stats->tx_errors = adapter->stats.ecol +
6850 			       adapter->stats.latecol;
6851 	net_stats->tx_aborted_errors = adapter->stats.ecol;
6852 	net_stats->tx_window_errors = adapter->stats.latecol;
6853 	net_stats->tx_carrier_errors = adapter->stats.tncrs;
6854 
6855 	/* Tx Dropped needs to be maintained elsewhere */
6856 
6857 	/* Management Stats */
6858 	adapter->stats.mgptc += rd32(E1000_MGTPTC);
6859 	adapter->stats.mgprc += rd32(E1000_MGTPRC);
6860 	adapter->stats.mgpdc += rd32(E1000_MGTPDC);
6861 
6862 	/* OS2BMC Stats */
6863 	reg = rd32(E1000_MANC);
6864 	if (reg & E1000_MANC_EN_BMC2OS) {
6865 		adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
6866 		adapter->stats.o2bspc += rd32(E1000_O2BSPC);
6867 		adapter->stats.b2ospc += rd32(E1000_B2OSPC);
6868 		adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
6869 	}
6870 }
6871 
6872 static void igb_perout(struct igb_adapter *adapter, int tsintr_tt)
6873 {
6874 	int pin = ptp_find_pin(adapter->ptp_clock, PTP_PF_PEROUT, tsintr_tt);
6875 	struct e1000_hw *hw = &adapter->hw;
6876 	struct timespec64 ts;
6877 	u32 tsauxc;
6878 
6879 	if (pin < 0 || pin >= IGB_N_SDP)
6880 		return;
6881 
6882 	spin_lock(&adapter->tmreg_lock);
6883 
6884 	if (hw->mac.type == e1000_82580 ||
6885 	    hw->mac.type == e1000_i354 ||
6886 	    hw->mac.type == e1000_i350) {
6887 		s64 ns = timespec64_to_ns(&adapter->perout[tsintr_tt].period);
6888 		u32 systiml, systimh, level_mask, level, rem;
6889 		u64 systim, now;
6890 
6891 		/* read systim registers in sequence */
6892 		rd32(E1000_SYSTIMR);
6893 		systiml = rd32(E1000_SYSTIML);
6894 		systimh = rd32(E1000_SYSTIMH);
6895 		systim = (((u64)(systimh & 0xFF)) << 32) | ((u64)systiml);
6896 		now = timecounter_cyc2time(&adapter->tc, systim);
6897 
6898 		if (pin < 2) {
6899 			level_mask = (tsintr_tt == 1) ? 0x80000 : 0x40000;
6900 			level = (rd32(E1000_CTRL) & level_mask) ? 1 : 0;
6901 		} else {
6902 			level_mask = (tsintr_tt == 1) ? 0x80 : 0x40;
6903 			level = (rd32(E1000_CTRL_EXT) & level_mask) ? 1 : 0;
6904 		}
6905 
6906 		div_u64_rem(now, ns, &rem);
6907 		systim = systim + (ns - rem);
6908 
6909 		/* synchronize pin level with rising/falling edges */
6910 		div_u64_rem(now, ns << 1, &rem);
6911 		if (rem < ns) {
6912 			/* first half of period */
6913 			if (level == 0) {
6914 				/* output is already low, skip this period */
6915 				systim += ns;
6916 				pr_notice("igb: periodic output on %s missed falling edge\n",
6917 					  adapter->sdp_config[pin].name);
6918 			}
6919 		} else {
6920 			/* second half of period */
6921 			if (level == 1) {
6922 				/* output is already high, skip this period */
6923 				systim += ns;
6924 				pr_notice("igb: periodic output on %s missed rising edge\n",
6925 					  adapter->sdp_config[pin].name);
6926 			}
6927 		}
6928 
6929 		/* for this chip family tv_sec is the upper part of the binary value,
6930 		 * so not seconds
6931 		 */
6932 		ts.tv_nsec = (u32)systim;
6933 		ts.tv_sec  = ((u32)(systim >> 32)) & 0xFF;
6934 	} else {
6935 		ts = timespec64_add(adapter->perout[tsintr_tt].start,
6936 				    adapter->perout[tsintr_tt].period);
6937 	}
6938 
6939 	/* u32 conversion of tv_sec is safe until y2106 */
6940 	wr32((tsintr_tt == 1) ? E1000_TRGTTIML1 : E1000_TRGTTIML0, ts.tv_nsec);
6941 	wr32((tsintr_tt == 1) ? E1000_TRGTTIMH1 : E1000_TRGTTIMH0, (u32)ts.tv_sec);
6942 	tsauxc = rd32(E1000_TSAUXC);
6943 	tsauxc |= TSAUXC_EN_TT0;
6944 	wr32(E1000_TSAUXC, tsauxc);
6945 	adapter->perout[tsintr_tt].start = ts;
6946 
6947 	spin_unlock(&adapter->tmreg_lock);
6948 }
6949 
6950 static void igb_extts(struct igb_adapter *adapter, int tsintr_tt)
6951 {
6952 	int pin = ptp_find_pin(adapter->ptp_clock, PTP_PF_EXTTS, tsintr_tt);
6953 	int auxstmpl = (tsintr_tt == 1) ? E1000_AUXSTMPL1 : E1000_AUXSTMPL0;
6954 	int auxstmph = (tsintr_tt == 1) ? E1000_AUXSTMPH1 : E1000_AUXSTMPH0;
6955 	struct e1000_hw *hw = &adapter->hw;
6956 	struct ptp_clock_event event;
6957 	struct timespec64 ts;
6958 	unsigned long flags;
6959 
6960 	if (pin < 0 || pin >= IGB_N_SDP)
6961 		return;
6962 
6963 	if (hw->mac.type == e1000_82580 ||
6964 	    hw->mac.type == e1000_i354 ||
6965 	    hw->mac.type == e1000_i350) {
6966 		u64 ns = rd32(auxstmpl);
6967 
6968 		ns += ((u64)(rd32(auxstmph) & 0xFF)) << 32;
6969 		spin_lock_irqsave(&adapter->tmreg_lock, flags);
6970 		ns = timecounter_cyc2time(&adapter->tc, ns);
6971 		spin_unlock_irqrestore(&adapter->tmreg_lock, flags);
6972 		ts = ns_to_timespec64(ns);
6973 	} else {
6974 		ts.tv_nsec = rd32(auxstmpl);
6975 		ts.tv_sec  = rd32(auxstmph);
6976 	}
6977 
6978 	event.type = PTP_CLOCK_EXTTS;
6979 	event.index = tsintr_tt;
6980 	event.timestamp = ts.tv_sec * 1000000000ULL + ts.tv_nsec;
6981 	ptp_clock_event(adapter->ptp_clock, &event);
6982 }
6983 
6984 static void igb_tsync_interrupt(struct igb_adapter *adapter)
6985 {
6986 	struct e1000_hw *hw = &adapter->hw;
6987 	u32 ack = 0, tsicr = rd32(E1000_TSICR);
6988 	struct ptp_clock_event event;
6989 
6990 	if (tsicr & TSINTR_SYS_WRAP) {
6991 		event.type = PTP_CLOCK_PPS;
6992 		if (adapter->ptp_caps.pps)
6993 			ptp_clock_event(adapter->ptp_clock, &event);
6994 		ack |= TSINTR_SYS_WRAP;
6995 	}
6996 
6997 	if (tsicr & E1000_TSICR_TXTS) {
6998 		/* retrieve hardware timestamp */
6999 		schedule_work(&adapter->ptp_tx_work);
7000 		ack |= E1000_TSICR_TXTS;
7001 	}
7002 
7003 	if (tsicr & TSINTR_TT0) {
7004 		igb_perout(adapter, 0);
7005 		ack |= TSINTR_TT0;
7006 	}
7007 
7008 	if (tsicr & TSINTR_TT1) {
7009 		igb_perout(adapter, 1);
7010 		ack |= TSINTR_TT1;
7011 	}
7012 
7013 	if (tsicr & TSINTR_AUTT0) {
7014 		igb_extts(adapter, 0);
7015 		ack |= TSINTR_AUTT0;
7016 	}
7017 
7018 	if (tsicr & TSINTR_AUTT1) {
7019 		igb_extts(adapter, 1);
7020 		ack |= TSINTR_AUTT1;
7021 	}
7022 
7023 	/* acknowledge the interrupts */
7024 	wr32(E1000_TSICR, ack);
7025 }
7026 
7027 static irqreturn_t igb_msix_other(int irq, void *data)
7028 {
7029 	struct igb_adapter *adapter = data;
7030 	struct e1000_hw *hw = &adapter->hw;
7031 	u32 icr = rd32(E1000_ICR);
7032 	/* reading ICR causes bit 31 of EICR to be cleared */
7033 
7034 	if (icr & E1000_ICR_DRSTA)
7035 		schedule_work(&adapter->reset_task);
7036 
7037 	if (icr & E1000_ICR_DOUTSYNC) {
7038 		/* HW is reporting DMA is out of sync */
7039 		adapter->stats.doosync++;
7040 		/* The DMA Out of Sync is also indication of a spoof event
7041 		 * in IOV mode. Check the Wrong VM Behavior register to
7042 		 * see if it is really a spoof event.
7043 		 */
7044 		igb_check_wvbr(adapter);
7045 	}
7046 
7047 	/* Check for a mailbox event */
7048 	if (icr & E1000_ICR_VMMB)
7049 		igb_msg_task(adapter);
7050 
7051 	if (icr & E1000_ICR_LSC) {
7052 		hw->mac.get_link_status = 1;
7053 		/* guard against interrupt when we're going down */
7054 		if (!test_bit(__IGB_DOWN, &adapter->state))
7055 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
7056 	}
7057 
7058 	if (icr & E1000_ICR_TS)
7059 		igb_tsync_interrupt(adapter);
7060 
7061 	wr32(E1000_EIMS, adapter->eims_other);
7062 
7063 	return IRQ_HANDLED;
7064 }
7065 
7066 static void igb_write_itr(struct igb_q_vector *q_vector)
7067 {
7068 	struct igb_adapter *adapter = q_vector->adapter;
7069 	u32 itr_val = q_vector->itr_val & 0x7FFC;
7070 
7071 	if (!q_vector->set_itr)
7072 		return;
7073 
7074 	if (!itr_val)
7075 		itr_val = 0x4;
7076 
7077 	if (adapter->hw.mac.type == e1000_82575)
7078 		itr_val |= itr_val << 16;
7079 	else
7080 		itr_val |= E1000_EITR_CNT_IGNR;
7081 
7082 	writel(itr_val, q_vector->itr_register);
7083 	q_vector->set_itr = 0;
7084 }
7085 
7086 static irqreturn_t igb_msix_ring(int irq, void *data)
7087 {
7088 	struct igb_q_vector *q_vector = data;
7089 
7090 	/* Write the ITR value calculated from the previous interrupt. */
7091 	igb_write_itr(q_vector);
7092 
7093 	napi_schedule(&q_vector->napi);
7094 
7095 	return IRQ_HANDLED;
7096 }
7097 
7098 #ifdef CONFIG_IGB_DCA
7099 static void igb_update_tx_dca(struct igb_adapter *adapter,
7100 			      struct igb_ring *tx_ring,
7101 			      int cpu)
7102 {
7103 	struct e1000_hw *hw = &adapter->hw;
7104 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
7105 
7106 	if (hw->mac.type != e1000_82575)
7107 		txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
7108 
7109 	/* We can enable relaxed ordering for reads, but not writes when
7110 	 * DCA is enabled.  This is due to a known issue in some chipsets
7111 	 * which will cause the DCA tag to be cleared.
7112 	 */
7113 	txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
7114 		  E1000_DCA_TXCTRL_DATA_RRO_EN |
7115 		  E1000_DCA_TXCTRL_DESC_DCA_EN;
7116 
7117 	wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
7118 }
7119 
7120 static void igb_update_rx_dca(struct igb_adapter *adapter,
7121 			      struct igb_ring *rx_ring,
7122 			      int cpu)
7123 {
7124 	struct e1000_hw *hw = &adapter->hw;
7125 	u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
7126 
7127 	if (hw->mac.type != e1000_82575)
7128 		rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
7129 
7130 	/* We can enable relaxed ordering for reads, but not writes when
7131 	 * DCA is enabled.  This is due to a known issue in some chipsets
7132 	 * which will cause the DCA tag to be cleared.
7133 	 */
7134 	rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
7135 		  E1000_DCA_RXCTRL_DESC_DCA_EN;
7136 
7137 	wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
7138 }
7139 
7140 static void igb_update_dca(struct igb_q_vector *q_vector)
7141 {
7142 	struct igb_adapter *adapter = q_vector->adapter;
7143 	int cpu = get_cpu();
7144 
7145 	if (q_vector->cpu == cpu)
7146 		goto out_no_update;
7147 
7148 	if (q_vector->tx.ring)
7149 		igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
7150 
7151 	if (q_vector->rx.ring)
7152 		igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
7153 
7154 	q_vector->cpu = cpu;
7155 out_no_update:
7156 	put_cpu();
7157 }
7158 
7159 static void igb_setup_dca(struct igb_adapter *adapter)
7160 {
7161 	struct e1000_hw *hw = &adapter->hw;
7162 	int i;
7163 
7164 	if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
7165 		return;
7166 
7167 	/* Always use CB2 mode, difference is masked in the CB driver. */
7168 	wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
7169 
7170 	for (i = 0; i < adapter->num_q_vectors; i++) {
7171 		adapter->q_vector[i]->cpu = -1;
7172 		igb_update_dca(adapter->q_vector[i]);
7173 	}
7174 }
7175 
7176 static int __igb_notify_dca(struct device *dev, void *data)
7177 {
7178 	struct net_device *netdev = dev_get_drvdata(dev);
7179 	struct igb_adapter *adapter = netdev_priv(netdev);
7180 	struct pci_dev *pdev = adapter->pdev;
7181 	struct e1000_hw *hw = &adapter->hw;
7182 	unsigned long event = *(unsigned long *)data;
7183 
7184 	switch (event) {
7185 	case DCA_PROVIDER_ADD:
7186 		/* if already enabled, don't do it again */
7187 		if (adapter->flags & IGB_FLAG_DCA_ENABLED)
7188 			break;
7189 		if (dca_add_requester(dev) == 0) {
7190 			adapter->flags |= IGB_FLAG_DCA_ENABLED;
7191 			dev_info(&pdev->dev, "DCA enabled\n");
7192 			igb_setup_dca(adapter);
7193 			break;
7194 		}
7195 		fallthrough; /* since DCA is disabled. */
7196 	case DCA_PROVIDER_REMOVE:
7197 		if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
7198 			/* without this a class_device is left
7199 			 * hanging around in the sysfs model
7200 			 */
7201 			dca_remove_requester(dev);
7202 			dev_info(&pdev->dev, "DCA disabled\n");
7203 			adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
7204 			wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
7205 		}
7206 		break;
7207 	}
7208 
7209 	return 0;
7210 }
7211 
7212 static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
7213 			  void *p)
7214 {
7215 	int ret_val;
7216 
7217 	ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
7218 					 __igb_notify_dca);
7219 
7220 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
7221 }
7222 #endif /* CONFIG_IGB_DCA */
7223 
7224 #ifdef CONFIG_PCI_IOV
7225 static int igb_vf_configure(struct igb_adapter *adapter, int vf)
7226 {
7227 	unsigned char mac_addr[ETH_ALEN];
7228 
7229 	eth_zero_addr(mac_addr);
7230 	igb_set_vf_mac(adapter, vf, mac_addr);
7231 
7232 	/* By default spoof check is enabled for all VFs */
7233 	adapter->vf_data[vf].spoofchk_enabled = true;
7234 
7235 	/* By default VFs are not trusted */
7236 	adapter->vf_data[vf].trusted = false;
7237 
7238 	return 0;
7239 }
7240 
7241 #endif
7242 static void igb_ping_all_vfs(struct igb_adapter *adapter)
7243 {
7244 	struct e1000_hw *hw = &adapter->hw;
7245 	u32 ping;
7246 	int i;
7247 
7248 	for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
7249 		ping = E1000_PF_CONTROL_MSG;
7250 		if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
7251 			ping |= E1000_VT_MSGTYPE_CTS;
7252 		igb_write_mbx(hw, &ping, 1, i);
7253 	}
7254 }
7255 
7256 static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
7257 {
7258 	struct e1000_hw *hw = &adapter->hw;
7259 	u32 vmolr = rd32(E1000_VMOLR(vf));
7260 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7261 
7262 	vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
7263 			    IGB_VF_FLAG_MULTI_PROMISC);
7264 	vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
7265 
7266 	if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
7267 		vmolr |= E1000_VMOLR_MPME;
7268 		vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
7269 		*msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
7270 	} else {
7271 		/* if we have hashes and we are clearing a multicast promisc
7272 		 * flag we need to write the hashes to the MTA as this step
7273 		 * was previously skipped
7274 		 */
7275 		if (vf_data->num_vf_mc_hashes > 30) {
7276 			vmolr |= E1000_VMOLR_MPME;
7277 		} else if (vf_data->num_vf_mc_hashes) {
7278 			int j;
7279 
7280 			vmolr |= E1000_VMOLR_ROMPE;
7281 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
7282 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
7283 		}
7284 	}
7285 
7286 	wr32(E1000_VMOLR(vf), vmolr);
7287 
7288 	/* there are flags left unprocessed, likely not supported */
7289 	if (*msgbuf & E1000_VT_MSGINFO_MASK)
7290 		return -EINVAL;
7291 
7292 	return 0;
7293 }
7294 
7295 static int igb_set_vf_multicasts(struct igb_adapter *adapter,
7296 				  u32 *msgbuf, u32 vf)
7297 {
7298 	int n = FIELD_GET(E1000_VT_MSGINFO_MASK, msgbuf[0]);
7299 	u16 *hash_list = (u16 *)&msgbuf[1];
7300 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7301 	int i;
7302 
7303 	/* salt away the number of multicast addresses assigned
7304 	 * to this VF for later use to restore when the PF multi cast
7305 	 * list changes
7306 	 */
7307 	vf_data->num_vf_mc_hashes = n;
7308 
7309 	/* only up to 30 hash values supported */
7310 	if (n > 30)
7311 		n = 30;
7312 
7313 	/* store the hashes for later use */
7314 	for (i = 0; i < n; i++)
7315 		vf_data->vf_mc_hashes[i] = hash_list[i];
7316 
7317 	/* Flush and reset the mta with the new values */
7318 	igb_set_rx_mode(adapter->netdev);
7319 
7320 	return 0;
7321 }
7322 
7323 static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
7324 {
7325 	struct e1000_hw *hw = &adapter->hw;
7326 	struct vf_data_storage *vf_data;
7327 	int i, j;
7328 
7329 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
7330 		u32 vmolr = rd32(E1000_VMOLR(i));
7331 
7332 		vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
7333 
7334 		vf_data = &adapter->vf_data[i];
7335 
7336 		if ((vf_data->num_vf_mc_hashes > 30) ||
7337 		    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
7338 			vmolr |= E1000_VMOLR_MPME;
7339 		} else if (vf_data->num_vf_mc_hashes) {
7340 			vmolr |= E1000_VMOLR_ROMPE;
7341 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
7342 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
7343 		}
7344 		wr32(E1000_VMOLR(i), vmolr);
7345 	}
7346 }
7347 
7348 static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
7349 {
7350 	struct e1000_hw *hw = &adapter->hw;
7351 	u32 pool_mask, vlvf_mask, i;
7352 
7353 	/* create mask for VF and other pools */
7354 	pool_mask = E1000_VLVF_POOLSEL_MASK;
7355 	vlvf_mask = BIT(E1000_VLVF_POOLSEL_SHIFT + vf);
7356 
7357 	/* drop PF from pool bits */
7358 	pool_mask &= ~BIT(E1000_VLVF_POOLSEL_SHIFT +
7359 			     adapter->vfs_allocated_count);
7360 
7361 	/* Find the vlan filter for this id */
7362 	for (i = E1000_VLVF_ARRAY_SIZE; i--;) {
7363 		u32 vlvf = rd32(E1000_VLVF(i));
7364 		u32 vfta_mask, vid, vfta;
7365 
7366 		/* remove the vf from the pool */
7367 		if (!(vlvf & vlvf_mask))
7368 			continue;
7369 
7370 		/* clear out bit from VLVF */
7371 		vlvf ^= vlvf_mask;
7372 
7373 		/* if other pools are present, just remove ourselves */
7374 		if (vlvf & pool_mask)
7375 			goto update_vlvfb;
7376 
7377 		/* if PF is present, leave VFTA */
7378 		if (vlvf & E1000_VLVF_POOLSEL_MASK)
7379 			goto update_vlvf;
7380 
7381 		vid = vlvf & E1000_VLVF_VLANID_MASK;
7382 		vfta_mask = BIT(vid % 32);
7383 
7384 		/* clear bit from VFTA */
7385 		vfta = adapter->shadow_vfta[vid / 32];
7386 		if (vfta & vfta_mask)
7387 			hw->mac.ops.write_vfta(hw, vid / 32, vfta ^ vfta_mask);
7388 update_vlvf:
7389 		/* clear pool selection enable */
7390 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
7391 			vlvf &= E1000_VLVF_POOLSEL_MASK;
7392 		else
7393 			vlvf = 0;
7394 update_vlvfb:
7395 		/* clear pool bits */
7396 		wr32(E1000_VLVF(i), vlvf);
7397 	}
7398 }
7399 
7400 static int igb_find_vlvf_entry(struct e1000_hw *hw, u32 vlan)
7401 {
7402 	u32 vlvf;
7403 	int idx;
7404 
7405 	/* short cut the special case */
7406 	if (vlan == 0)
7407 		return 0;
7408 
7409 	/* Search for the VLAN id in the VLVF entries */
7410 	for (idx = E1000_VLVF_ARRAY_SIZE; --idx;) {
7411 		vlvf = rd32(E1000_VLVF(idx));
7412 		if ((vlvf & VLAN_VID_MASK) == vlan)
7413 			break;
7414 	}
7415 
7416 	return idx;
7417 }
7418 
7419 static void igb_update_pf_vlvf(struct igb_adapter *adapter, u32 vid)
7420 {
7421 	struct e1000_hw *hw = &adapter->hw;
7422 	u32 bits, pf_id;
7423 	int idx;
7424 
7425 	idx = igb_find_vlvf_entry(hw, vid);
7426 	if (!idx)
7427 		return;
7428 
7429 	/* See if any other pools are set for this VLAN filter
7430 	 * entry other than the PF.
7431 	 */
7432 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
7433 	bits = ~BIT(pf_id) & E1000_VLVF_POOLSEL_MASK;
7434 	bits &= rd32(E1000_VLVF(idx));
7435 
7436 	/* Disable the filter so this falls into the default pool. */
7437 	if (!bits) {
7438 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
7439 			wr32(E1000_VLVF(idx), BIT(pf_id));
7440 		else
7441 			wr32(E1000_VLVF(idx), 0);
7442 	}
7443 }
7444 
7445 static s32 igb_set_vf_vlan(struct igb_adapter *adapter, u32 vid,
7446 			   bool add, u32 vf)
7447 {
7448 	int pf_id = adapter->vfs_allocated_count;
7449 	struct e1000_hw *hw = &adapter->hw;
7450 	int err;
7451 
7452 	/* If VLAN overlaps with one the PF is currently monitoring make
7453 	 * sure that we are able to allocate a VLVF entry.  This may be
7454 	 * redundant but it guarantees PF will maintain visibility to
7455 	 * the VLAN.
7456 	 */
7457 	if (add && test_bit(vid, adapter->active_vlans)) {
7458 		err = igb_vfta_set(hw, vid, pf_id, true, false);
7459 		if (err)
7460 			return err;
7461 	}
7462 
7463 	err = igb_vfta_set(hw, vid, vf, add, false);
7464 
7465 	if (add && !err)
7466 		return err;
7467 
7468 	/* If we failed to add the VF VLAN or we are removing the VF VLAN
7469 	 * we may need to drop the PF pool bit in order to allow us to free
7470 	 * up the VLVF resources.
7471 	 */
7472 	if (test_bit(vid, adapter->active_vlans) ||
7473 	    (adapter->flags & IGB_FLAG_VLAN_PROMISC))
7474 		igb_update_pf_vlvf(adapter, vid);
7475 
7476 	return err;
7477 }
7478 
7479 static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
7480 {
7481 	struct e1000_hw *hw = &adapter->hw;
7482 
7483 	if (vid)
7484 		wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
7485 	else
7486 		wr32(E1000_VMVIR(vf), 0);
7487 }
7488 
7489 static int igb_enable_port_vlan(struct igb_adapter *adapter, int vf,
7490 				u16 vlan, u8 qos)
7491 {
7492 	int err;
7493 
7494 	err = igb_set_vf_vlan(adapter, vlan, true, vf);
7495 	if (err)
7496 		return err;
7497 
7498 	igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
7499 	igb_set_vmolr(adapter, vf, !vlan);
7500 
7501 	/* revoke access to previous VLAN */
7502 	if (vlan != adapter->vf_data[vf].pf_vlan)
7503 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
7504 				false, vf);
7505 
7506 	adapter->vf_data[vf].pf_vlan = vlan;
7507 	adapter->vf_data[vf].pf_qos = qos;
7508 	igb_set_vf_vlan_strip(adapter, vf, true);
7509 	dev_info(&adapter->pdev->dev,
7510 		 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
7511 	if (test_bit(__IGB_DOWN, &adapter->state)) {
7512 		dev_warn(&adapter->pdev->dev,
7513 			 "The VF VLAN has been set, but the PF device is not up.\n");
7514 		dev_warn(&adapter->pdev->dev,
7515 			 "Bring the PF device up before attempting to use the VF device.\n");
7516 	}
7517 
7518 	return err;
7519 }
7520 
7521 static int igb_disable_port_vlan(struct igb_adapter *adapter, int vf)
7522 {
7523 	/* Restore tagless access via VLAN 0 */
7524 	igb_set_vf_vlan(adapter, 0, true, vf);
7525 
7526 	igb_set_vmvir(adapter, 0, vf);
7527 	igb_set_vmolr(adapter, vf, true);
7528 
7529 	/* Remove any PF assigned VLAN */
7530 	if (adapter->vf_data[vf].pf_vlan)
7531 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
7532 				false, vf);
7533 
7534 	adapter->vf_data[vf].pf_vlan = 0;
7535 	adapter->vf_data[vf].pf_qos = 0;
7536 	igb_set_vf_vlan_strip(adapter, vf, false);
7537 
7538 	return 0;
7539 }
7540 
7541 static int igb_ndo_set_vf_vlan(struct net_device *netdev, int vf,
7542 			       u16 vlan, u8 qos, __be16 vlan_proto)
7543 {
7544 	struct igb_adapter *adapter = netdev_priv(netdev);
7545 
7546 	if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
7547 		return -EINVAL;
7548 
7549 	if (vlan_proto != htons(ETH_P_8021Q))
7550 		return -EPROTONOSUPPORT;
7551 
7552 	return (vlan || qos) ? igb_enable_port_vlan(adapter, vf, vlan, qos) :
7553 			       igb_disable_port_vlan(adapter, vf);
7554 }
7555 
7556 static int igb_set_vf_vlan_msg(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
7557 {
7558 	int add = FIELD_GET(E1000_VT_MSGINFO_MASK, msgbuf[0]);
7559 	int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
7560 	int ret;
7561 
7562 	if (adapter->vf_data[vf].pf_vlan)
7563 		return -1;
7564 
7565 	/* VLAN 0 is a special case, don't allow it to be removed */
7566 	if (!vid && !add)
7567 		return 0;
7568 
7569 	ret = igb_set_vf_vlan(adapter, vid, !!add, vf);
7570 	if (!ret)
7571 		igb_set_vf_vlan_strip(adapter, vf, !!vid);
7572 	return ret;
7573 }
7574 
7575 static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
7576 {
7577 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7578 
7579 	/* clear flags - except flag that indicates PF has set the MAC */
7580 	vf_data->flags &= IGB_VF_FLAG_PF_SET_MAC;
7581 	vf_data->last_nack = jiffies;
7582 
7583 	/* reset vlans for device */
7584 	igb_clear_vf_vfta(adapter, vf);
7585 	igb_set_vf_vlan(adapter, vf_data->pf_vlan, true, vf);
7586 	igb_set_vmvir(adapter, vf_data->pf_vlan |
7587 			       (vf_data->pf_qos << VLAN_PRIO_SHIFT), vf);
7588 	igb_set_vmolr(adapter, vf, !vf_data->pf_vlan);
7589 	igb_set_vf_vlan_strip(adapter, vf, !!(vf_data->pf_vlan));
7590 
7591 	/* reset multicast table array for vf */
7592 	adapter->vf_data[vf].num_vf_mc_hashes = 0;
7593 
7594 	/* Flush and reset the mta with the new values */
7595 	igb_set_rx_mode(adapter->netdev);
7596 }
7597 
7598 static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
7599 {
7600 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7601 
7602 	/* clear mac address as we were hotplug removed/added */
7603 	if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
7604 		eth_zero_addr(vf_mac);
7605 
7606 	/* process remaining reset events */
7607 	igb_vf_reset(adapter, vf);
7608 }
7609 
7610 static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
7611 {
7612 	struct e1000_hw *hw = &adapter->hw;
7613 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7614 	u32 reg, msgbuf[3] = {};
7615 	u8 *addr = (u8 *)(&msgbuf[1]);
7616 
7617 	/* process all the same items cleared in a function level reset */
7618 	igb_vf_reset(adapter, vf);
7619 
7620 	/* set vf mac address */
7621 	igb_set_vf_mac(adapter, vf, vf_mac);
7622 
7623 	/* enable transmit and receive for vf */
7624 	reg = rd32(E1000_VFTE);
7625 	wr32(E1000_VFTE, reg | BIT(vf));
7626 	reg = rd32(E1000_VFRE);
7627 	wr32(E1000_VFRE, reg | BIT(vf));
7628 
7629 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
7630 
7631 	/* reply to reset with ack and vf mac address */
7632 	if (!is_zero_ether_addr(vf_mac)) {
7633 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
7634 		memcpy(addr, vf_mac, ETH_ALEN);
7635 	} else {
7636 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
7637 	}
7638 	igb_write_mbx(hw, msgbuf, 3, vf);
7639 }
7640 
7641 static void igb_flush_mac_table(struct igb_adapter *adapter)
7642 {
7643 	struct e1000_hw *hw = &adapter->hw;
7644 	int i;
7645 
7646 	for (i = 0; i < hw->mac.rar_entry_count; i++) {
7647 		adapter->mac_table[i].state &= ~IGB_MAC_STATE_IN_USE;
7648 		eth_zero_addr(adapter->mac_table[i].addr);
7649 		adapter->mac_table[i].queue = 0;
7650 		igb_rar_set_index(adapter, i);
7651 	}
7652 }
7653 
7654 static int igb_available_rars(struct igb_adapter *adapter, u8 queue)
7655 {
7656 	struct e1000_hw *hw = &adapter->hw;
7657 	/* do not count rar entries reserved for VFs MAC addresses */
7658 	int rar_entries = hw->mac.rar_entry_count -
7659 			  adapter->vfs_allocated_count;
7660 	int i, count = 0;
7661 
7662 	for (i = 0; i < rar_entries; i++) {
7663 		/* do not count default entries */
7664 		if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT)
7665 			continue;
7666 
7667 		/* do not count "in use" entries for different queues */
7668 		if ((adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE) &&
7669 		    (adapter->mac_table[i].queue != queue))
7670 			continue;
7671 
7672 		count++;
7673 	}
7674 
7675 	return count;
7676 }
7677 
7678 /* Set default MAC address for the PF in the first RAR entry */
7679 static void igb_set_default_mac_filter(struct igb_adapter *adapter)
7680 {
7681 	struct igb_mac_addr *mac_table = &adapter->mac_table[0];
7682 
7683 	ether_addr_copy(mac_table->addr, adapter->hw.mac.addr);
7684 	mac_table->queue = adapter->vfs_allocated_count;
7685 	mac_table->state = IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7686 
7687 	igb_rar_set_index(adapter, 0);
7688 }
7689 
7690 /* If the filter to be added and an already existing filter express
7691  * the same address and address type, it should be possible to only
7692  * override the other configurations, for example the queue to steer
7693  * traffic.
7694  */
7695 static bool igb_mac_entry_can_be_used(const struct igb_mac_addr *entry,
7696 				      const u8 *addr, const u8 flags)
7697 {
7698 	if (!(entry->state & IGB_MAC_STATE_IN_USE))
7699 		return true;
7700 
7701 	if ((entry->state & IGB_MAC_STATE_SRC_ADDR) !=
7702 	    (flags & IGB_MAC_STATE_SRC_ADDR))
7703 		return false;
7704 
7705 	if (!ether_addr_equal(addr, entry->addr))
7706 		return false;
7707 
7708 	return true;
7709 }
7710 
7711 /* Add a MAC filter for 'addr' directing matching traffic to 'queue',
7712  * 'flags' is used to indicate what kind of match is made, match is by
7713  * default for the destination address, if matching by source address
7714  * is desired the flag IGB_MAC_STATE_SRC_ADDR can be used.
7715  */
7716 static int igb_add_mac_filter_flags(struct igb_adapter *adapter,
7717 				    const u8 *addr, const u8 queue,
7718 				    const u8 flags)
7719 {
7720 	struct e1000_hw *hw = &adapter->hw;
7721 	int rar_entries = hw->mac.rar_entry_count -
7722 			  adapter->vfs_allocated_count;
7723 	int i;
7724 
7725 	if (is_zero_ether_addr(addr))
7726 		return -EINVAL;
7727 
7728 	/* Search for the first empty entry in the MAC table.
7729 	 * Do not touch entries at the end of the table reserved for the VF MAC
7730 	 * addresses.
7731 	 */
7732 	for (i = 0; i < rar_entries; i++) {
7733 		if (!igb_mac_entry_can_be_used(&adapter->mac_table[i],
7734 					       addr, flags))
7735 			continue;
7736 
7737 		ether_addr_copy(adapter->mac_table[i].addr, addr);
7738 		adapter->mac_table[i].queue = queue;
7739 		adapter->mac_table[i].state |= IGB_MAC_STATE_IN_USE | flags;
7740 
7741 		igb_rar_set_index(adapter, i);
7742 		return i;
7743 	}
7744 
7745 	return -ENOSPC;
7746 }
7747 
7748 static int igb_add_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7749 			      const u8 queue)
7750 {
7751 	return igb_add_mac_filter_flags(adapter, addr, queue, 0);
7752 }
7753 
7754 /* Remove a MAC filter for 'addr' directing matching traffic to
7755  * 'queue', 'flags' is used to indicate what kind of match need to be
7756  * removed, match is by default for the destination address, if
7757  * matching by source address is to be removed the flag
7758  * IGB_MAC_STATE_SRC_ADDR can be used.
7759  */
7760 static int igb_del_mac_filter_flags(struct igb_adapter *adapter,
7761 				    const u8 *addr, const u8 queue,
7762 				    const u8 flags)
7763 {
7764 	struct e1000_hw *hw = &adapter->hw;
7765 	int rar_entries = hw->mac.rar_entry_count -
7766 			  adapter->vfs_allocated_count;
7767 	int i;
7768 
7769 	if (is_zero_ether_addr(addr))
7770 		return -EINVAL;
7771 
7772 	/* Search for matching entry in the MAC table based on given address
7773 	 * and queue. Do not touch entries at the end of the table reserved
7774 	 * for the VF MAC addresses.
7775 	 */
7776 	for (i = 0; i < rar_entries; i++) {
7777 		if (!(adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE))
7778 			continue;
7779 		if ((adapter->mac_table[i].state & flags) != flags)
7780 			continue;
7781 		if (adapter->mac_table[i].queue != queue)
7782 			continue;
7783 		if (!ether_addr_equal(adapter->mac_table[i].addr, addr))
7784 			continue;
7785 
7786 		/* When a filter for the default address is "deleted",
7787 		 * we return it to its initial configuration
7788 		 */
7789 		if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT) {
7790 			adapter->mac_table[i].state =
7791 				IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7792 			adapter->mac_table[i].queue =
7793 				adapter->vfs_allocated_count;
7794 		} else {
7795 			adapter->mac_table[i].state = 0;
7796 			adapter->mac_table[i].queue = 0;
7797 			eth_zero_addr(adapter->mac_table[i].addr);
7798 		}
7799 
7800 		igb_rar_set_index(adapter, i);
7801 		return 0;
7802 	}
7803 
7804 	return -ENOENT;
7805 }
7806 
7807 static int igb_del_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7808 			      const u8 queue)
7809 {
7810 	return igb_del_mac_filter_flags(adapter, addr, queue, 0);
7811 }
7812 
7813 int igb_add_mac_steering_filter(struct igb_adapter *adapter,
7814 				const u8 *addr, u8 queue, u8 flags)
7815 {
7816 	struct e1000_hw *hw = &adapter->hw;
7817 
7818 	/* In theory, this should be supported on 82575 as well, but
7819 	 * that part wasn't easily accessible during development.
7820 	 */
7821 	if (hw->mac.type != e1000_i210)
7822 		return -EOPNOTSUPP;
7823 
7824 	return igb_add_mac_filter_flags(adapter, addr, queue,
7825 					IGB_MAC_STATE_QUEUE_STEERING | flags);
7826 }
7827 
7828 int igb_del_mac_steering_filter(struct igb_adapter *adapter,
7829 				const u8 *addr, u8 queue, u8 flags)
7830 {
7831 	return igb_del_mac_filter_flags(adapter, addr, queue,
7832 					IGB_MAC_STATE_QUEUE_STEERING | flags);
7833 }
7834 
7835 static int igb_uc_sync(struct net_device *netdev, const unsigned char *addr)
7836 {
7837 	struct igb_adapter *adapter = netdev_priv(netdev);
7838 	int ret;
7839 
7840 	ret = igb_add_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7841 
7842 	return min_t(int, ret, 0);
7843 }
7844 
7845 static int igb_uc_unsync(struct net_device *netdev, const unsigned char *addr)
7846 {
7847 	struct igb_adapter *adapter = netdev_priv(netdev);
7848 
7849 	igb_del_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7850 
7851 	return 0;
7852 }
7853 
7854 static int igb_set_vf_mac_filter(struct igb_adapter *adapter, const int vf,
7855 				 const u32 info, const u8 *addr)
7856 {
7857 	struct pci_dev *pdev = adapter->pdev;
7858 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7859 	struct vf_mac_filter *entry;
7860 	bool found = false;
7861 	int ret = 0;
7862 
7863 	if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7864 	    !vf_data->trusted) {
7865 		dev_warn(&pdev->dev,
7866 			 "VF %d requested MAC filter but is administratively denied\n",
7867 			  vf);
7868 		return -EINVAL;
7869 	}
7870 	if (!is_valid_ether_addr(addr)) {
7871 		dev_warn(&pdev->dev,
7872 			 "VF %d attempted to set invalid MAC filter\n",
7873 			  vf);
7874 		return -EINVAL;
7875 	}
7876 
7877 	switch (info) {
7878 	case E1000_VF_MAC_FILTER_CLR:
7879 		/* remove all unicast MAC filters related to the current VF */
7880 		list_for_each_entry(entry, &adapter->vf_macs.l, l) {
7881 			if (entry->vf == vf) {
7882 				entry->vf = -1;
7883 				entry->free = true;
7884 				igb_del_mac_filter(adapter, entry->vf_mac, vf);
7885 			}
7886 		}
7887 		break;
7888 	case E1000_VF_MAC_FILTER_ADD:
7889 		/* try to find empty slot in the list */
7890 		list_for_each_entry(entry, &adapter->vf_macs.l, l) {
7891 			if (entry->free) {
7892 				found = true;
7893 				break;
7894 			}
7895 		}
7896 
7897 		if (found) {
7898 			entry->free = false;
7899 			entry->vf = vf;
7900 			ether_addr_copy(entry->vf_mac, addr);
7901 
7902 			ret = igb_add_mac_filter(adapter, addr, vf);
7903 			ret = min_t(int, ret, 0);
7904 		} else {
7905 			ret = -ENOSPC;
7906 		}
7907 
7908 		if (ret == -ENOSPC)
7909 			dev_warn(&pdev->dev,
7910 				 "VF %d has requested MAC filter but there is no space for it\n",
7911 				 vf);
7912 		break;
7913 	default:
7914 		ret = -EINVAL;
7915 		break;
7916 	}
7917 
7918 	return ret;
7919 }
7920 
7921 static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
7922 {
7923 	struct pci_dev *pdev = adapter->pdev;
7924 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7925 	u32 info = msg[0] & E1000_VT_MSGINFO_MASK;
7926 
7927 	/* The VF MAC Address is stored in a packed array of bytes
7928 	 * starting at the second 32 bit word of the msg array
7929 	 */
7930 	unsigned char *addr = (unsigned char *)&msg[1];
7931 	int ret = 0;
7932 
7933 	if (!info) {
7934 		if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7935 		    !vf_data->trusted) {
7936 			dev_warn(&pdev->dev,
7937 				 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
7938 				 vf);
7939 			return -EINVAL;
7940 		}
7941 
7942 		if (!is_valid_ether_addr(addr)) {
7943 			dev_warn(&pdev->dev,
7944 				 "VF %d attempted to set invalid MAC\n",
7945 				 vf);
7946 			return -EINVAL;
7947 		}
7948 
7949 		ret = igb_set_vf_mac(adapter, vf, addr);
7950 	} else {
7951 		ret = igb_set_vf_mac_filter(adapter, vf, info, addr);
7952 	}
7953 
7954 	return ret;
7955 }
7956 
7957 static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
7958 {
7959 	struct e1000_hw *hw = &adapter->hw;
7960 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7961 	u32 msg = E1000_VT_MSGTYPE_NACK;
7962 
7963 	/* if device isn't clear to send it shouldn't be reading either */
7964 	if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
7965 	    time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
7966 		igb_write_mbx(hw, &msg, 1, vf);
7967 		vf_data->last_nack = jiffies;
7968 	}
7969 }
7970 
7971 static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
7972 {
7973 	struct pci_dev *pdev = adapter->pdev;
7974 	u32 msgbuf[E1000_VFMAILBOX_SIZE];
7975 	struct e1000_hw *hw = &adapter->hw;
7976 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7977 	s32 retval;
7978 
7979 	retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf, false);
7980 
7981 	if (retval) {
7982 		/* if receive failed revoke VF CTS stats and restart init */
7983 		dev_err(&pdev->dev, "Error receiving message from VF\n");
7984 		vf_data->flags &= ~IGB_VF_FLAG_CTS;
7985 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7986 			goto unlock;
7987 		goto out;
7988 	}
7989 
7990 	/* this is a message we already processed, do nothing */
7991 	if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
7992 		goto unlock;
7993 
7994 	/* until the vf completes a reset it should not be
7995 	 * allowed to start any configuration.
7996 	 */
7997 	if (msgbuf[0] == E1000_VF_RESET) {
7998 		/* unlocks mailbox */
7999 		igb_vf_reset_msg(adapter, vf);
8000 		return;
8001 	}
8002 
8003 	if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
8004 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
8005 			goto unlock;
8006 		retval = -1;
8007 		goto out;
8008 	}
8009 
8010 	switch ((msgbuf[0] & 0xFFFF)) {
8011 	case E1000_VF_SET_MAC_ADDR:
8012 		retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
8013 		break;
8014 	case E1000_VF_SET_PROMISC:
8015 		retval = igb_set_vf_promisc(adapter, msgbuf, vf);
8016 		break;
8017 	case E1000_VF_SET_MULTICAST:
8018 		retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
8019 		break;
8020 	case E1000_VF_SET_LPE:
8021 		retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
8022 		break;
8023 	case E1000_VF_SET_VLAN:
8024 		retval = -1;
8025 		if (vf_data->pf_vlan)
8026 			dev_warn(&pdev->dev,
8027 				 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
8028 				 vf);
8029 		else
8030 			retval = igb_set_vf_vlan_msg(adapter, msgbuf, vf);
8031 		break;
8032 	default:
8033 		dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
8034 		retval = -1;
8035 		break;
8036 	}
8037 
8038 	msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
8039 out:
8040 	/* notify the VF of the results of what it sent us */
8041 	if (retval)
8042 		msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
8043 	else
8044 		msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
8045 
8046 	/* unlocks mailbox */
8047 	igb_write_mbx(hw, msgbuf, 1, vf);
8048 	return;
8049 
8050 unlock:
8051 	igb_unlock_mbx(hw, vf);
8052 }
8053 
8054 static void igb_msg_task(struct igb_adapter *adapter)
8055 {
8056 	struct e1000_hw *hw = &adapter->hw;
8057 	unsigned long flags;
8058 	u32 vf;
8059 
8060 	spin_lock_irqsave(&adapter->vfs_lock, flags);
8061 	for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
8062 		/* process any reset requests */
8063 		if (!igb_check_for_rst(hw, vf))
8064 			igb_vf_reset_event(adapter, vf);
8065 
8066 		/* process any messages pending */
8067 		if (!igb_check_for_msg(hw, vf))
8068 			igb_rcv_msg_from_vf(adapter, vf);
8069 
8070 		/* process any acks */
8071 		if (!igb_check_for_ack(hw, vf))
8072 			igb_rcv_ack_from_vf(adapter, vf);
8073 	}
8074 	spin_unlock_irqrestore(&adapter->vfs_lock, flags);
8075 }
8076 
8077 /**
8078  *  igb_set_uta - Set unicast filter table address
8079  *  @adapter: board private structure
8080  *  @set: boolean indicating if we are setting or clearing bits
8081  *
8082  *  The unicast table address is a register array of 32-bit registers.
8083  *  The table is meant to be used in a way similar to how the MTA is used
8084  *  however due to certain limitations in the hardware it is necessary to
8085  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
8086  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
8087  **/
8088 static void igb_set_uta(struct igb_adapter *adapter, bool set)
8089 {
8090 	struct e1000_hw *hw = &adapter->hw;
8091 	u32 uta = set ? ~0 : 0;
8092 	int i;
8093 
8094 	/* we only need to do this if VMDq is enabled */
8095 	if (!adapter->vfs_allocated_count)
8096 		return;
8097 
8098 	for (i = hw->mac.uta_reg_count; i--;)
8099 		array_wr32(E1000_UTA, i, uta);
8100 }
8101 
8102 /**
8103  *  igb_intr_msi - Interrupt Handler
8104  *  @irq: interrupt number
8105  *  @data: pointer to a network interface device structure
8106  **/
8107 static irqreturn_t igb_intr_msi(int irq, void *data)
8108 {
8109 	struct igb_adapter *adapter = data;
8110 	struct igb_q_vector *q_vector = adapter->q_vector[0];
8111 	struct e1000_hw *hw = &adapter->hw;
8112 	/* read ICR disables interrupts using IAM */
8113 	u32 icr = rd32(E1000_ICR);
8114 
8115 	igb_write_itr(q_vector);
8116 
8117 	if (icr & E1000_ICR_DRSTA)
8118 		schedule_work(&adapter->reset_task);
8119 
8120 	if (icr & E1000_ICR_DOUTSYNC) {
8121 		/* HW is reporting DMA is out of sync */
8122 		adapter->stats.doosync++;
8123 	}
8124 
8125 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
8126 		hw->mac.get_link_status = 1;
8127 		if (!test_bit(__IGB_DOWN, &adapter->state))
8128 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
8129 	}
8130 
8131 	if (icr & E1000_ICR_TS)
8132 		igb_tsync_interrupt(adapter);
8133 
8134 	napi_schedule(&q_vector->napi);
8135 
8136 	return IRQ_HANDLED;
8137 }
8138 
8139 /**
8140  *  igb_intr - Legacy Interrupt Handler
8141  *  @irq: interrupt number
8142  *  @data: pointer to a network interface device structure
8143  **/
8144 static irqreturn_t igb_intr(int irq, void *data)
8145 {
8146 	struct igb_adapter *adapter = data;
8147 	struct igb_q_vector *q_vector = adapter->q_vector[0];
8148 	struct e1000_hw *hw = &adapter->hw;
8149 	/* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
8150 	 * need for the IMC write
8151 	 */
8152 	u32 icr = rd32(E1000_ICR);
8153 
8154 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
8155 	 * not set, then the adapter didn't send an interrupt
8156 	 */
8157 	if (!(icr & E1000_ICR_INT_ASSERTED))
8158 		return IRQ_NONE;
8159 
8160 	igb_write_itr(q_vector);
8161 
8162 	if (icr & E1000_ICR_DRSTA)
8163 		schedule_work(&adapter->reset_task);
8164 
8165 	if (icr & E1000_ICR_DOUTSYNC) {
8166 		/* HW is reporting DMA is out of sync */
8167 		adapter->stats.doosync++;
8168 	}
8169 
8170 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
8171 		hw->mac.get_link_status = 1;
8172 		/* guard against interrupt when we're going down */
8173 		if (!test_bit(__IGB_DOWN, &adapter->state))
8174 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
8175 	}
8176 
8177 	if (icr & E1000_ICR_TS)
8178 		igb_tsync_interrupt(adapter);
8179 
8180 	napi_schedule(&q_vector->napi);
8181 
8182 	return IRQ_HANDLED;
8183 }
8184 
8185 static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
8186 {
8187 	struct igb_adapter *adapter = q_vector->adapter;
8188 	struct e1000_hw *hw = &adapter->hw;
8189 
8190 	if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
8191 	    (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
8192 		if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
8193 			igb_set_itr(q_vector);
8194 		else
8195 			igb_update_ring_itr(q_vector);
8196 	}
8197 
8198 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
8199 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
8200 			wr32(E1000_EIMS, q_vector->eims_value);
8201 		else
8202 			igb_irq_enable(adapter);
8203 	}
8204 }
8205 
8206 /**
8207  *  igb_poll - NAPI Rx polling callback
8208  *  @napi: napi polling structure
8209  *  @budget: count of how many packets we should handle
8210  **/
8211 static int igb_poll(struct napi_struct *napi, int budget)
8212 {
8213 	struct igb_q_vector *q_vector = container_of(napi,
8214 						     struct igb_q_vector,
8215 						     napi);
8216 	bool clean_complete = true;
8217 	int work_done = 0;
8218 
8219 #ifdef CONFIG_IGB_DCA
8220 	if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
8221 		igb_update_dca(q_vector);
8222 #endif
8223 	if (q_vector->tx.ring)
8224 		clean_complete = igb_clean_tx_irq(q_vector, budget);
8225 
8226 	if (q_vector->rx.ring) {
8227 		int cleaned = igb_clean_rx_irq(q_vector, budget);
8228 
8229 		work_done += cleaned;
8230 		if (cleaned >= budget)
8231 			clean_complete = false;
8232 	}
8233 
8234 	/* If all work not completed, return budget and keep polling */
8235 	if (!clean_complete)
8236 		return budget;
8237 
8238 	/* Exit the polling mode, but don't re-enable interrupts if stack might
8239 	 * poll us due to busy-polling
8240 	 */
8241 	if (likely(napi_complete_done(napi, work_done)))
8242 		igb_ring_irq_enable(q_vector);
8243 
8244 	return work_done;
8245 }
8246 
8247 /**
8248  *  igb_clean_tx_irq - Reclaim resources after transmit completes
8249  *  @q_vector: pointer to q_vector containing needed info
8250  *  @napi_budget: Used to determine if we are in netpoll
8251  *
8252  *  returns true if ring is completely cleaned
8253  **/
8254 static bool igb_clean_tx_irq(struct igb_q_vector *q_vector, int napi_budget)
8255 {
8256 	struct igb_adapter *adapter = q_vector->adapter;
8257 	struct igb_ring *tx_ring = q_vector->tx.ring;
8258 	struct igb_tx_buffer *tx_buffer;
8259 	union e1000_adv_tx_desc *tx_desc;
8260 	unsigned int total_bytes = 0, total_packets = 0;
8261 	unsigned int budget = q_vector->tx.work_limit;
8262 	unsigned int i = tx_ring->next_to_clean;
8263 
8264 	if (test_bit(__IGB_DOWN, &adapter->state))
8265 		return true;
8266 
8267 	tx_buffer = &tx_ring->tx_buffer_info[i];
8268 	tx_desc = IGB_TX_DESC(tx_ring, i);
8269 	i -= tx_ring->count;
8270 
8271 	do {
8272 		union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
8273 
8274 		/* if next_to_watch is not set then there is no work pending */
8275 		if (!eop_desc)
8276 			break;
8277 
8278 		/* prevent any other reads prior to eop_desc */
8279 		smp_rmb();
8280 
8281 		/* if DD is not set pending work has not been completed */
8282 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
8283 			break;
8284 
8285 		/* clear next_to_watch to prevent false hangs */
8286 		tx_buffer->next_to_watch = NULL;
8287 
8288 		/* update the statistics for this packet */
8289 		total_bytes += tx_buffer->bytecount;
8290 		total_packets += tx_buffer->gso_segs;
8291 
8292 		/* free the skb */
8293 		if (tx_buffer->type == IGB_TYPE_SKB)
8294 			napi_consume_skb(tx_buffer->skb, napi_budget);
8295 		else
8296 			xdp_return_frame(tx_buffer->xdpf);
8297 
8298 		/* unmap skb header data */
8299 		dma_unmap_single(tx_ring->dev,
8300 				 dma_unmap_addr(tx_buffer, dma),
8301 				 dma_unmap_len(tx_buffer, len),
8302 				 DMA_TO_DEVICE);
8303 
8304 		/* clear tx_buffer data */
8305 		dma_unmap_len_set(tx_buffer, len, 0);
8306 
8307 		/* clear last DMA location and unmap remaining buffers */
8308 		while (tx_desc != eop_desc) {
8309 			tx_buffer++;
8310 			tx_desc++;
8311 			i++;
8312 			if (unlikely(!i)) {
8313 				i -= tx_ring->count;
8314 				tx_buffer = tx_ring->tx_buffer_info;
8315 				tx_desc = IGB_TX_DESC(tx_ring, 0);
8316 			}
8317 
8318 			/* unmap any remaining paged data */
8319 			if (dma_unmap_len(tx_buffer, len)) {
8320 				dma_unmap_page(tx_ring->dev,
8321 					       dma_unmap_addr(tx_buffer, dma),
8322 					       dma_unmap_len(tx_buffer, len),
8323 					       DMA_TO_DEVICE);
8324 				dma_unmap_len_set(tx_buffer, len, 0);
8325 			}
8326 		}
8327 
8328 		/* move us one more past the eop_desc for start of next pkt */
8329 		tx_buffer++;
8330 		tx_desc++;
8331 		i++;
8332 		if (unlikely(!i)) {
8333 			i -= tx_ring->count;
8334 			tx_buffer = tx_ring->tx_buffer_info;
8335 			tx_desc = IGB_TX_DESC(tx_ring, 0);
8336 		}
8337 
8338 		/* issue prefetch for next Tx descriptor */
8339 		prefetch(tx_desc);
8340 
8341 		/* update budget accounting */
8342 		budget--;
8343 	} while (likely(budget));
8344 
8345 	netdev_tx_completed_queue(txring_txq(tx_ring),
8346 				  total_packets, total_bytes);
8347 	i += tx_ring->count;
8348 	tx_ring->next_to_clean = i;
8349 	u64_stats_update_begin(&tx_ring->tx_syncp);
8350 	tx_ring->tx_stats.bytes += total_bytes;
8351 	tx_ring->tx_stats.packets += total_packets;
8352 	u64_stats_update_end(&tx_ring->tx_syncp);
8353 	q_vector->tx.total_bytes += total_bytes;
8354 	q_vector->tx.total_packets += total_packets;
8355 
8356 	if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
8357 		struct e1000_hw *hw = &adapter->hw;
8358 
8359 		/* Detect a transmit hang in hardware, this serializes the
8360 		 * check with the clearing of time_stamp and movement of i
8361 		 */
8362 		clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
8363 		if (tx_buffer->next_to_watch &&
8364 		    time_after(jiffies, tx_buffer->time_stamp +
8365 			       (adapter->tx_timeout_factor * HZ)) &&
8366 		    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
8367 
8368 			/* detected Tx unit hang */
8369 			dev_err(tx_ring->dev,
8370 				"Detected Tx Unit Hang\n"
8371 				"  Tx Queue             <%d>\n"
8372 				"  TDH                  <%x>\n"
8373 				"  TDT                  <%x>\n"
8374 				"  next_to_use          <%x>\n"
8375 				"  next_to_clean        <%x>\n"
8376 				"buffer_info[next_to_clean]\n"
8377 				"  time_stamp           <%lx>\n"
8378 				"  next_to_watch        <%p>\n"
8379 				"  jiffies              <%lx>\n"
8380 				"  desc.status          <%x>\n",
8381 				tx_ring->queue_index,
8382 				rd32(E1000_TDH(tx_ring->reg_idx)),
8383 				readl(tx_ring->tail),
8384 				tx_ring->next_to_use,
8385 				tx_ring->next_to_clean,
8386 				tx_buffer->time_stamp,
8387 				tx_buffer->next_to_watch,
8388 				jiffies,
8389 				tx_buffer->next_to_watch->wb.status);
8390 			netif_stop_subqueue(tx_ring->netdev,
8391 					    tx_ring->queue_index);
8392 
8393 			/* we are about to reset, no point in enabling stuff */
8394 			return true;
8395 		}
8396 	}
8397 
8398 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
8399 	if (unlikely(total_packets &&
8400 	    netif_carrier_ok(tx_ring->netdev) &&
8401 	    igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
8402 		/* Make sure that anybody stopping the queue after this
8403 		 * sees the new next_to_clean.
8404 		 */
8405 		smp_mb();
8406 		if (__netif_subqueue_stopped(tx_ring->netdev,
8407 					     tx_ring->queue_index) &&
8408 		    !(test_bit(__IGB_DOWN, &adapter->state))) {
8409 			netif_wake_subqueue(tx_ring->netdev,
8410 					    tx_ring->queue_index);
8411 
8412 			u64_stats_update_begin(&tx_ring->tx_syncp);
8413 			tx_ring->tx_stats.restart_queue++;
8414 			u64_stats_update_end(&tx_ring->tx_syncp);
8415 		}
8416 	}
8417 
8418 	return !!budget;
8419 }
8420 
8421 /**
8422  *  igb_reuse_rx_page - page flip buffer and store it back on the ring
8423  *  @rx_ring: rx descriptor ring to store buffers on
8424  *  @old_buff: donor buffer to have page reused
8425  *
8426  *  Synchronizes page for reuse by the adapter
8427  **/
8428 static void igb_reuse_rx_page(struct igb_ring *rx_ring,
8429 			      struct igb_rx_buffer *old_buff)
8430 {
8431 	struct igb_rx_buffer *new_buff;
8432 	u16 nta = rx_ring->next_to_alloc;
8433 
8434 	new_buff = &rx_ring->rx_buffer_info[nta];
8435 
8436 	/* update, and store next to alloc */
8437 	nta++;
8438 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
8439 
8440 	/* Transfer page from old buffer to new buffer.
8441 	 * Move each member individually to avoid possible store
8442 	 * forwarding stalls.
8443 	 */
8444 	new_buff->dma		= old_buff->dma;
8445 	new_buff->page		= old_buff->page;
8446 	new_buff->page_offset	= old_buff->page_offset;
8447 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
8448 }
8449 
8450 static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer,
8451 				  int rx_buf_pgcnt)
8452 {
8453 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
8454 	struct page *page = rx_buffer->page;
8455 
8456 	/* avoid re-using remote and pfmemalloc pages */
8457 	if (!dev_page_is_reusable(page))
8458 		return false;
8459 
8460 #if (PAGE_SIZE < 8192)
8461 	/* if we are only owner of page we can reuse it */
8462 	if (unlikely((rx_buf_pgcnt - pagecnt_bias) > 1))
8463 		return false;
8464 #else
8465 #define IGB_LAST_OFFSET \
8466 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IGB_RXBUFFER_2048)
8467 
8468 	if (rx_buffer->page_offset > IGB_LAST_OFFSET)
8469 		return false;
8470 #endif
8471 
8472 	/* If we have drained the page fragment pool we need to update
8473 	 * the pagecnt_bias and page count so that we fully restock the
8474 	 * number of references the driver holds.
8475 	 */
8476 	if (unlikely(pagecnt_bias == 1)) {
8477 		page_ref_add(page, USHRT_MAX - 1);
8478 		rx_buffer->pagecnt_bias = USHRT_MAX;
8479 	}
8480 
8481 	return true;
8482 }
8483 
8484 /**
8485  *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
8486  *  @rx_ring: rx descriptor ring to transact packets on
8487  *  @rx_buffer: buffer containing page to add
8488  *  @skb: sk_buff to place the data into
8489  *  @size: size of buffer to be added
8490  *
8491  *  This function will add the data contained in rx_buffer->page to the skb.
8492  **/
8493 static void igb_add_rx_frag(struct igb_ring *rx_ring,
8494 			    struct igb_rx_buffer *rx_buffer,
8495 			    struct sk_buff *skb,
8496 			    unsigned int size)
8497 {
8498 #if (PAGE_SIZE < 8192)
8499 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8500 #else
8501 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
8502 				SKB_DATA_ALIGN(IGB_SKB_PAD + size) :
8503 				SKB_DATA_ALIGN(size);
8504 #endif
8505 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
8506 			rx_buffer->page_offset, size, truesize);
8507 #if (PAGE_SIZE < 8192)
8508 	rx_buffer->page_offset ^= truesize;
8509 #else
8510 	rx_buffer->page_offset += truesize;
8511 #endif
8512 }
8513 
8514 static struct sk_buff *igb_construct_skb(struct igb_ring *rx_ring,
8515 					 struct igb_rx_buffer *rx_buffer,
8516 					 struct xdp_buff *xdp,
8517 					 ktime_t timestamp)
8518 {
8519 #if (PAGE_SIZE < 8192)
8520 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8521 #else
8522 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
8523 					       xdp->data_hard_start);
8524 #endif
8525 	unsigned int size = xdp->data_end - xdp->data;
8526 	unsigned int headlen;
8527 	struct sk_buff *skb;
8528 
8529 	/* prefetch first cache line of first page */
8530 	net_prefetch(xdp->data);
8531 
8532 	/* allocate a skb to store the frags */
8533 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
8534 	if (unlikely(!skb))
8535 		return NULL;
8536 
8537 	if (timestamp)
8538 		skb_hwtstamps(skb)->hwtstamp = timestamp;
8539 
8540 	/* Determine available headroom for copy */
8541 	headlen = size;
8542 	if (headlen > IGB_RX_HDR_LEN)
8543 		headlen = eth_get_headlen(skb->dev, xdp->data, IGB_RX_HDR_LEN);
8544 
8545 	/* align pull length to size of long to optimize memcpy performance */
8546 	memcpy(__skb_put(skb, headlen), xdp->data, ALIGN(headlen, sizeof(long)));
8547 
8548 	/* update all of the pointers */
8549 	size -= headlen;
8550 	if (size) {
8551 		skb_add_rx_frag(skb, 0, rx_buffer->page,
8552 				(xdp->data + headlen) - page_address(rx_buffer->page),
8553 				size, truesize);
8554 #if (PAGE_SIZE < 8192)
8555 		rx_buffer->page_offset ^= truesize;
8556 #else
8557 		rx_buffer->page_offset += truesize;
8558 #endif
8559 	} else {
8560 		rx_buffer->pagecnt_bias++;
8561 	}
8562 
8563 	return skb;
8564 }
8565 
8566 static struct sk_buff *igb_build_skb(struct igb_ring *rx_ring,
8567 				     struct igb_rx_buffer *rx_buffer,
8568 				     struct xdp_buff *xdp,
8569 				     ktime_t timestamp)
8570 {
8571 #if (PAGE_SIZE < 8192)
8572 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8573 #else
8574 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
8575 				SKB_DATA_ALIGN(xdp->data_end -
8576 					       xdp->data_hard_start);
8577 #endif
8578 	unsigned int metasize = xdp->data - xdp->data_meta;
8579 	struct sk_buff *skb;
8580 
8581 	/* prefetch first cache line of first page */
8582 	net_prefetch(xdp->data_meta);
8583 
8584 	/* build an skb around the page buffer */
8585 	skb = napi_build_skb(xdp->data_hard_start, truesize);
8586 	if (unlikely(!skb))
8587 		return NULL;
8588 
8589 	/* update pointers within the skb to store the data */
8590 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
8591 	__skb_put(skb, xdp->data_end - xdp->data);
8592 
8593 	if (metasize)
8594 		skb_metadata_set(skb, metasize);
8595 
8596 	if (timestamp)
8597 		skb_hwtstamps(skb)->hwtstamp = timestamp;
8598 
8599 	/* update buffer offset */
8600 #if (PAGE_SIZE < 8192)
8601 	rx_buffer->page_offset ^= truesize;
8602 #else
8603 	rx_buffer->page_offset += truesize;
8604 #endif
8605 
8606 	return skb;
8607 }
8608 
8609 static struct sk_buff *igb_run_xdp(struct igb_adapter *adapter,
8610 				   struct igb_ring *rx_ring,
8611 				   struct xdp_buff *xdp)
8612 {
8613 	int err, result = IGB_XDP_PASS;
8614 	struct bpf_prog *xdp_prog;
8615 	u32 act;
8616 
8617 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
8618 
8619 	if (!xdp_prog)
8620 		goto xdp_out;
8621 
8622 	prefetchw(xdp->data_hard_start); /* xdp_frame write */
8623 
8624 	act = bpf_prog_run_xdp(xdp_prog, xdp);
8625 	switch (act) {
8626 	case XDP_PASS:
8627 		break;
8628 	case XDP_TX:
8629 		result = igb_xdp_xmit_back(adapter, xdp);
8630 		if (result == IGB_XDP_CONSUMED)
8631 			goto out_failure;
8632 		break;
8633 	case XDP_REDIRECT:
8634 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
8635 		if (err)
8636 			goto out_failure;
8637 		result = IGB_XDP_REDIR;
8638 		break;
8639 	default:
8640 		bpf_warn_invalid_xdp_action(adapter->netdev, xdp_prog, act);
8641 		fallthrough;
8642 	case XDP_ABORTED:
8643 out_failure:
8644 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
8645 		fallthrough;
8646 	case XDP_DROP:
8647 		result = IGB_XDP_CONSUMED;
8648 		break;
8649 	}
8650 xdp_out:
8651 	return ERR_PTR(-result);
8652 }
8653 
8654 static unsigned int igb_rx_frame_truesize(struct igb_ring *rx_ring,
8655 					  unsigned int size)
8656 {
8657 	unsigned int truesize;
8658 
8659 #if (PAGE_SIZE < 8192)
8660 	truesize = igb_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
8661 #else
8662 	truesize = ring_uses_build_skb(rx_ring) ?
8663 		SKB_DATA_ALIGN(IGB_SKB_PAD + size) +
8664 		SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
8665 		SKB_DATA_ALIGN(size);
8666 #endif
8667 	return truesize;
8668 }
8669 
8670 static void igb_rx_buffer_flip(struct igb_ring *rx_ring,
8671 			       struct igb_rx_buffer *rx_buffer,
8672 			       unsigned int size)
8673 {
8674 	unsigned int truesize = igb_rx_frame_truesize(rx_ring, size);
8675 #if (PAGE_SIZE < 8192)
8676 	rx_buffer->page_offset ^= truesize;
8677 #else
8678 	rx_buffer->page_offset += truesize;
8679 #endif
8680 }
8681 
8682 static inline void igb_rx_checksum(struct igb_ring *ring,
8683 				   union e1000_adv_rx_desc *rx_desc,
8684 				   struct sk_buff *skb)
8685 {
8686 	skb_checksum_none_assert(skb);
8687 
8688 	/* Ignore Checksum bit is set */
8689 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
8690 		return;
8691 
8692 	/* Rx checksum disabled via ethtool */
8693 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
8694 		return;
8695 
8696 	/* TCP/UDP checksum error bit is set */
8697 	if (igb_test_staterr(rx_desc,
8698 			     E1000_RXDEXT_STATERR_TCPE |
8699 			     E1000_RXDEXT_STATERR_IPE)) {
8700 		/* work around errata with sctp packets where the TCPE aka
8701 		 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
8702 		 * packets, (aka let the stack check the crc32c)
8703 		 */
8704 		if (!((skb->len == 60) &&
8705 		      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
8706 			u64_stats_update_begin(&ring->rx_syncp);
8707 			ring->rx_stats.csum_err++;
8708 			u64_stats_update_end(&ring->rx_syncp);
8709 		}
8710 		/* let the stack verify checksum errors */
8711 		return;
8712 	}
8713 	/* It must be a TCP or UDP packet with a valid checksum */
8714 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
8715 				      E1000_RXD_STAT_UDPCS))
8716 		skb->ip_summed = CHECKSUM_UNNECESSARY;
8717 
8718 	dev_dbg(ring->dev, "cksum success: bits %08X\n",
8719 		le32_to_cpu(rx_desc->wb.upper.status_error));
8720 }
8721 
8722 static inline void igb_rx_hash(struct igb_ring *ring,
8723 			       union e1000_adv_rx_desc *rx_desc,
8724 			       struct sk_buff *skb)
8725 {
8726 	if (ring->netdev->features & NETIF_F_RXHASH)
8727 		skb_set_hash(skb,
8728 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
8729 			     PKT_HASH_TYPE_L3);
8730 }
8731 
8732 /**
8733  *  igb_is_non_eop - process handling of non-EOP buffers
8734  *  @rx_ring: Rx ring being processed
8735  *  @rx_desc: Rx descriptor for current buffer
8736  *
8737  *  This function updates next to clean.  If the buffer is an EOP buffer
8738  *  this function exits returning false, otherwise it will place the
8739  *  sk_buff in the next buffer to be chained and return true indicating
8740  *  that this is in fact a non-EOP buffer.
8741  **/
8742 static bool igb_is_non_eop(struct igb_ring *rx_ring,
8743 			   union e1000_adv_rx_desc *rx_desc)
8744 {
8745 	u32 ntc = rx_ring->next_to_clean + 1;
8746 
8747 	/* fetch, update, and store next to clean */
8748 	ntc = (ntc < rx_ring->count) ? ntc : 0;
8749 	rx_ring->next_to_clean = ntc;
8750 
8751 	prefetch(IGB_RX_DESC(rx_ring, ntc));
8752 
8753 	if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
8754 		return false;
8755 
8756 	return true;
8757 }
8758 
8759 /**
8760  *  igb_cleanup_headers - Correct corrupted or empty headers
8761  *  @rx_ring: rx descriptor ring packet is being transacted on
8762  *  @rx_desc: pointer to the EOP Rx descriptor
8763  *  @skb: pointer to current skb being fixed
8764  *
8765  *  Address the case where we are pulling data in on pages only
8766  *  and as such no data is present in the skb header.
8767  *
8768  *  In addition if skb is not at least 60 bytes we need to pad it so that
8769  *  it is large enough to qualify as a valid Ethernet frame.
8770  *
8771  *  Returns true if an error was encountered and skb was freed.
8772  **/
8773 static bool igb_cleanup_headers(struct igb_ring *rx_ring,
8774 				union e1000_adv_rx_desc *rx_desc,
8775 				struct sk_buff *skb)
8776 {
8777 	/* XDP packets use error pointer so abort at this point */
8778 	if (IS_ERR(skb))
8779 		return true;
8780 
8781 	if (unlikely((igb_test_staterr(rx_desc,
8782 				       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
8783 		struct net_device *netdev = rx_ring->netdev;
8784 		if (!(netdev->features & NETIF_F_RXALL)) {
8785 			dev_kfree_skb_any(skb);
8786 			return true;
8787 		}
8788 	}
8789 
8790 	/* if eth_skb_pad returns an error the skb was freed */
8791 	if (eth_skb_pad(skb))
8792 		return true;
8793 
8794 	return false;
8795 }
8796 
8797 /**
8798  *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
8799  *  @rx_ring: rx descriptor ring packet is being transacted on
8800  *  @rx_desc: pointer to the EOP Rx descriptor
8801  *  @skb: pointer to current skb being populated
8802  *
8803  *  This function checks the ring, descriptor, and packet information in
8804  *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
8805  *  other fields within the skb.
8806  **/
8807 static void igb_process_skb_fields(struct igb_ring *rx_ring,
8808 				   union e1000_adv_rx_desc *rx_desc,
8809 				   struct sk_buff *skb)
8810 {
8811 	struct net_device *dev = rx_ring->netdev;
8812 
8813 	igb_rx_hash(rx_ring, rx_desc, skb);
8814 
8815 	igb_rx_checksum(rx_ring, rx_desc, skb);
8816 
8817 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
8818 	    !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
8819 		igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
8820 
8821 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
8822 	    igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
8823 		u16 vid;
8824 
8825 		if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
8826 		    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
8827 			vid = be16_to_cpu((__force __be16)rx_desc->wb.upper.vlan);
8828 		else
8829 			vid = le16_to_cpu(rx_desc->wb.upper.vlan);
8830 
8831 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
8832 	}
8833 
8834 	skb_record_rx_queue(skb, rx_ring->queue_index);
8835 
8836 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
8837 }
8838 
8839 static unsigned int igb_rx_offset(struct igb_ring *rx_ring)
8840 {
8841 	return ring_uses_build_skb(rx_ring) ? IGB_SKB_PAD : 0;
8842 }
8843 
8844 static struct igb_rx_buffer *igb_get_rx_buffer(struct igb_ring *rx_ring,
8845 					       const unsigned int size, int *rx_buf_pgcnt)
8846 {
8847 	struct igb_rx_buffer *rx_buffer;
8848 
8849 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
8850 	*rx_buf_pgcnt =
8851 #if (PAGE_SIZE < 8192)
8852 		page_count(rx_buffer->page);
8853 #else
8854 		0;
8855 #endif
8856 	prefetchw(rx_buffer->page);
8857 
8858 	/* we are reusing so sync this buffer for CPU use */
8859 	dma_sync_single_range_for_cpu(rx_ring->dev,
8860 				      rx_buffer->dma,
8861 				      rx_buffer->page_offset,
8862 				      size,
8863 				      DMA_FROM_DEVICE);
8864 
8865 	rx_buffer->pagecnt_bias--;
8866 
8867 	return rx_buffer;
8868 }
8869 
8870 static void igb_put_rx_buffer(struct igb_ring *rx_ring,
8871 			      struct igb_rx_buffer *rx_buffer, int rx_buf_pgcnt)
8872 {
8873 	if (igb_can_reuse_rx_page(rx_buffer, rx_buf_pgcnt)) {
8874 		/* hand second half of page back to the ring */
8875 		igb_reuse_rx_page(rx_ring, rx_buffer);
8876 	} else {
8877 		/* We are not reusing the buffer so unmap it and free
8878 		 * any references we are holding to it
8879 		 */
8880 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
8881 				     igb_rx_pg_size(rx_ring), DMA_FROM_DEVICE,
8882 				     IGB_RX_DMA_ATTR);
8883 		__page_frag_cache_drain(rx_buffer->page,
8884 					rx_buffer->pagecnt_bias);
8885 	}
8886 
8887 	/* clear contents of rx_buffer */
8888 	rx_buffer->page = NULL;
8889 }
8890 
8891 static int igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
8892 {
8893 	struct igb_adapter *adapter = q_vector->adapter;
8894 	struct igb_ring *rx_ring = q_vector->rx.ring;
8895 	struct sk_buff *skb = rx_ring->skb;
8896 	unsigned int total_bytes = 0, total_packets = 0;
8897 	u16 cleaned_count = igb_desc_unused(rx_ring);
8898 	unsigned int xdp_xmit = 0;
8899 	struct xdp_buff xdp;
8900 	u32 frame_sz = 0;
8901 	int rx_buf_pgcnt;
8902 
8903 	/* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
8904 #if (PAGE_SIZE < 8192)
8905 	frame_sz = igb_rx_frame_truesize(rx_ring, 0);
8906 #endif
8907 	xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq);
8908 
8909 	while (likely(total_packets < budget)) {
8910 		union e1000_adv_rx_desc *rx_desc;
8911 		struct igb_rx_buffer *rx_buffer;
8912 		ktime_t timestamp = 0;
8913 		int pkt_offset = 0;
8914 		unsigned int size;
8915 		void *pktbuf;
8916 
8917 		/* return some buffers to hardware, one at a time is too slow */
8918 		if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
8919 			igb_alloc_rx_buffers(rx_ring, cleaned_count);
8920 			cleaned_count = 0;
8921 		}
8922 
8923 		rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
8924 		size = le16_to_cpu(rx_desc->wb.upper.length);
8925 		if (!size)
8926 			break;
8927 
8928 		/* This memory barrier is needed to keep us from reading
8929 		 * any other fields out of the rx_desc until we know the
8930 		 * descriptor has been written back
8931 		 */
8932 		dma_rmb();
8933 
8934 		rx_buffer = igb_get_rx_buffer(rx_ring, size, &rx_buf_pgcnt);
8935 		pktbuf = page_address(rx_buffer->page) + rx_buffer->page_offset;
8936 
8937 		/* pull rx packet timestamp if available and valid */
8938 		if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
8939 			int ts_hdr_len;
8940 
8941 			ts_hdr_len = igb_ptp_rx_pktstamp(rx_ring->q_vector,
8942 							 pktbuf, &timestamp);
8943 
8944 			pkt_offset += ts_hdr_len;
8945 			size -= ts_hdr_len;
8946 		}
8947 
8948 		/* retrieve a buffer from the ring */
8949 		if (!skb) {
8950 			unsigned char *hard_start = pktbuf - igb_rx_offset(rx_ring);
8951 			unsigned int offset = pkt_offset + igb_rx_offset(rx_ring);
8952 
8953 			xdp_prepare_buff(&xdp, hard_start, offset, size, true);
8954 			xdp_buff_clear_frags_flag(&xdp);
8955 #if (PAGE_SIZE > 4096)
8956 			/* At larger PAGE_SIZE, frame_sz depend on len size */
8957 			xdp.frame_sz = igb_rx_frame_truesize(rx_ring, size);
8958 #endif
8959 			skb = igb_run_xdp(adapter, rx_ring, &xdp);
8960 		}
8961 
8962 		if (IS_ERR(skb)) {
8963 			unsigned int xdp_res = -PTR_ERR(skb);
8964 
8965 			if (xdp_res & (IGB_XDP_TX | IGB_XDP_REDIR)) {
8966 				xdp_xmit |= xdp_res;
8967 				igb_rx_buffer_flip(rx_ring, rx_buffer, size);
8968 			} else {
8969 				rx_buffer->pagecnt_bias++;
8970 			}
8971 			total_packets++;
8972 			total_bytes += size;
8973 		} else if (skb)
8974 			igb_add_rx_frag(rx_ring, rx_buffer, skb, size);
8975 		else if (ring_uses_build_skb(rx_ring))
8976 			skb = igb_build_skb(rx_ring, rx_buffer, &xdp,
8977 					    timestamp);
8978 		else
8979 			skb = igb_construct_skb(rx_ring, rx_buffer,
8980 						&xdp, timestamp);
8981 
8982 		/* exit if we failed to retrieve a buffer */
8983 		if (!skb) {
8984 			rx_ring->rx_stats.alloc_failed++;
8985 			rx_buffer->pagecnt_bias++;
8986 			break;
8987 		}
8988 
8989 		igb_put_rx_buffer(rx_ring, rx_buffer, rx_buf_pgcnt);
8990 		cleaned_count++;
8991 
8992 		/* fetch next buffer in frame if non-eop */
8993 		if (igb_is_non_eop(rx_ring, rx_desc))
8994 			continue;
8995 
8996 		/* verify the packet layout is correct */
8997 		if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
8998 			skb = NULL;
8999 			continue;
9000 		}
9001 
9002 		/* probably a little skewed due to removing CRC */
9003 		total_bytes += skb->len;
9004 
9005 		/* populate checksum, timestamp, VLAN, and protocol */
9006 		igb_process_skb_fields(rx_ring, rx_desc, skb);
9007 
9008 		napi_gro_receive(&q_vector->napi, skb);
9009 
9010 		/* reset skb pointer */
9011 		skb = NULL;
9012 
9013 		/* update budget accounting */
9014 		total_packets++;
9015 	}
9016 
9017 	/* place incomplete frames back on ring for completion */
9018 	rx_ring->skb = skb;
9019 
9020 	if (xdp_xmit & IGB_XDP_REDIR)
9021 		xdp_do_flush();
9022 
9023 	if (xdp_xmit & IGB_XDP_TX) {
9024 		struct igb_ring *tx_ring = igb_xdp_tx_queue_mapping(adapter);
9025 
9026 		igb_xdp_ring_update_tail(tx_ring);
9027 	}
9028 
9029 	u64_stats_update_begin(&rx_ring->rx_syncp);
9030 	rx_ring->rx_stats.packets += total_packets;
9031 	rx_ring->rx_stats.bytes += total_bytes;
9032 	u64_stats_update_end(&rx_ring->rx_syncp);
9033 	q_vector->rx.total_packets += total_packets;
9034 	q_vector->rx.total_bytes += total_bytes;
9035 
9036 	if (cleaned_count)
9037 		igb_alloc_rx_buffers(rx_ring, cleaned_count);
9038 
9039 	return total_packets;
9040 }
9041 
9042 static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
9043 				  struct igb_rx_buffer *bi)
9044 {
9045 	struct page *page = bi->page;
9046 	dma_addr_t dma;
9047 
9048 	/* since we are recycling buffers we should seldom need to alloc */
9049 	if (likely(page))
9050 		return true;
9051 
9052 	/* alloc new page for storage */
9053 	page = dev_alloc_pages(igb_rx_pg_order(rx_ring));
9054 	if (unlikely(!page)) {
9055 		rx_ring->rx_stats.alloc_failed++;
9056 		return false;
9057 	}
9058 
9059 	/* map page for use */
9060 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
9061 				 igb_rx_pg_size(rx_ring),
9062 				 DMA_FROM_DEVICE,
9063 				 IGB_RX_DMA_ATTR);
9064 
9065 	/* if mapping failed free memory back to system since
9066 	 * there isn't much point in holding memory we can't use
9067 	 */
9068 	if (dma_mapping_error(rx_ring->dev, dma)) {
9069 		__free_pages(page, igb_rx_pg_order(rx_ring));
9070 
9071 		rx_ring->rx_stats.alloc_failed++;
9072 		return false;
9073 	}
9074 
9075 	bi->dma = dma;
9076 	bi->page = page;
9077 	bi->page_offset = igb_rx_offset(rx_ring);
9078 	page_ref_add(page, USHRT_MAX - 1);
9079 	bi->pagecnt_bias = USHRT_MAX;
9080 
9081 	return true;
9082 }
9083 
9084 /**
9085  *  igb_alloc_rx_buffers - Replace used receive buffers
9086  *  @rx_ring: rx descriptor ring to allocate new receive buffers
9087  *  @cleaned_count: count of buffers to allocate
9088  **/
9089 void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
9090 {
9091 	union e1000_adv_rx_desc *rx_desc;
9092 	struct igb_rx_buffer *bi;
9093 	u16 i = rx_ring->next_to_use;
9094 	u16 bufsz;
9095 
9096 	/* nothing to do */
9097 	if (!cleaned_count)
9098 		return;
9099 
9100 	rx_desc = IGB_RX_DESC(rx_ring, i);
9101 	bi = &rx_ring->rx_buffer_info[i];
9102 	i -= rx_ring->count;
9103 
9104 	bufsz = igb_rx_bufsz(rx_ring);
9105 
9106 	do {
9107 		if (!igb_alloc_mapped_page(rx_ring, bi))
9108 			break;
9109 
9110 		/* sync the buffer for use by the device */
9111 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
9112 						 bi->page_offset, bufsz,
9113 						 DMA_FROM_DEVICE);
9114 
9115 		/* Refresh the desc even if buffer_addrs didn't change
9116 		 * because each write-back erases this info.
9117 		 */
9118 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
9119 
9120 		rx_desc++;
9121 		bi++;
9122 		i++;
9123 		if (unlikely(!i)) {
9124 			rx_desc = IGB_RX_DESC(rx_ring, 0);
9125 			bi = rx_ring->rx_buffer_info;
9126 			i -= rx_ring->count;
9127 		}
9128 
9129 		/* clear the length for the next_to_use descriptor */
9130 		rx_desc->wb.upper.length = 0;
9131 
9132 		cleaned_count--;
9133 	} while (cleaned_count);
9134 
9135 	i += rx_ring->count;
9136 
9137 	if (rx_ring->next_to_use != i) {
9138 		/* record the next descriptor to use */
9139 		rx_ring->next_to_use = i;
9140 
9141 		/* update next to alloc since we have filled the ring */
9142 		rx_ring->next_to_alloc = i;
9143 
9144 		/* Force memory writes to complete before letting h/w
9145 		 * know there are new descriptors to fetch.  (Only
9146 		 * applicable for weak-ordered memory model archs,
9147 		 * such as IA-64).
9148 		 */
9149 		dma_wmb();
9150 		writel(i, rx_ring->tail);
9151 	}
9152 }
9153 
9154 /**
9155  * igb_mii_ioctl -
9156  * @netdev: pointer to netdev struct
9157  * @ifr: interface structure
9158  * @cmd: ioctl command to execute
9159  **/
9160 static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
9161 {
9162 	struct igb_adapter *adapter = netdev_priv(netdev);
9163 	struct mii_ioctl_data *data = if_mii(ifr);
9164 
9165 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
9166 		return -EOPNOTSUPP;
9167 
9168 	switch (cmd) {
9169 	case SIOCGMIIPHY:
9170 		data->phy_id = adapter->hw.phy.addr;
9171 		break;
9172 	case SIOCGMIIREG:
9173 		if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
9174 				     &data->val_out))
9175 			return -EIO;
9176 		break;
9177 	case SIOCSMIIREG:
9178 	default:
9179 		return -EOPNOTSUPP;
9180 	}
9181 	return 0;
9182 }
9183 
9184 /**
9185  * igb_ioctl -
9186  * @netdev: pointer to netdev struct
9187  * @ifr: interface structure
9188  * @cmd: ioctl command to execute
9189  **/
9190 static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
9191 {
9192 	switch (cmd) {
9193 	case SIOCGMIIPHY:
9194 	case SIOCGMIIREG:
9195 	case SIOCSMIIREG:
9196 		return igb_mii_ioctl(netdev, ifr, cmd);
9197 	case SIOCGHWTSTAMP:
9198 		return igb_ptp_get_ts_config(netdev, ifr);
9199 	case SIOCSHWTSTAMP:
9200 		return igb_ptp_set_ts_config(netdev, ifr);
9201 	default:
9202 		return -EOPNOTSUPP;
9203 	}
9204 }
9205 
9206 void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
9207 {
9208 	struct igb_adapter *adapter = hw->back;
9209 
9210 	pci_read_config_word(adapter->pdev, reg, value);
9211 }
9212 
9213 void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
9214 {
9215 	struct igb_adapter *adapter = hw->back;
9216 
9217 	pci_write_config_word(adapter->pdev, reg, *value);
9218 }
9219 
9220 s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
9221 {
9222 	struct igb_adapter *adapter = hw->back;
9223 
9224 	if (pcie_capability_read_word(adapter->pdev, reg, value))
9225 		return -E1000_ERR_CONFIG;
9226 
9227 	return 0;
9228 }
9229 
9230 s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
9231 {
9232 	struct igb_adapter *adapter = hw->back;
9233 
9234 	if (pcie_capability_write_word(adapter->pdev, reg, *value))
9235 		return -E1000_ERR_CONFIG;
9236 
9237 	return 0;
9238 }
9239 
9240 static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
9241 {
9242 	struct igb_adapter *adapter = netdev_priv(netdev);
9243 	struct e1000_hw *hw = &adapter->hw;
9244 	u32 ctrl, rctl;
9245 	bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
9246 
9247 	if (enable) {
9248 		/* enable VLAN tag insert/strip */
9249 		ctrl = rd32(E1000_CTRL);
9250 		ctrl |= E1000_CTRL_VME;
9251 		wr32(E1000_CTRL, ctrl);
9252 
9253 		/* Disable CFI check */
9254 		rctl = rd32(E1000_RCTL);
9255 		rctl &= ~E1000_RCTL_CFIEN;
9256 		wr32(E1000_RCTL, rctl);
9257 	} else {
9258 		/* disable VLAN tag insert/strip */
9259 		ctrl = rd32(E1000_CTRL);
9260 		ctrl &= ~E1000_CTRL_VME;
9261 		wr32(E1000_CTRL, ctrl);
9262 	}
9263 
9264 	igb_set_vf_vlan_strip(adapter, adapter->vfs_allocated_count, enable);
9265 }
9266 
9267 static int igb_vlan_rx_add_vid(struct net_device *netdev,
9268 			       __be16 proto, u16 vid)
9269 {
9270 	struct igb_adapter *adapter = netdev_priv(netdev);
9271 	struct e1000_hw *hw = &adapter->hw;
9272 	int pf_id = adapter->vfs_allocated_count;
9273 
9274 	/* add the filter since PF can receive vlans w/o entry in vlvf */
9275 	if (!vid || !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
9276 		igb_vfta_set(hw, vid, pf_id, true, !!vid);
9277 
9278 	set_bit(vid, adapter->active_vlans);
9279 
9280 	return 0;
9281 }
9282 
9283 static int igb_vlan_rx_kill_vid(struct net_device *netdev,
9284 				__be16 proto, u16 vid)
9285 {
9286 	struct igb_adapter *adapter = netdev_priv(netdev);
9287 	int pf_id = adapter->vfs_allocated_count;
9288 	struct e1000_hw *hw = &adapter->hw;
9289 
9290 	/* remove VID from filter table */
9291 	if (vid && !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
9292 		igb_vfta_set(hw, vid, pf_id, false, true);
9293 
9294 	clear_bit(vid, adapter->active_vlans);
9295 
9296 	return 0;
9297 }
9298 
9299 static void igb_restore_vlan(struct igb_adapter *adapter)
9300 {
9301 	u16 vid = 1;
9302 
9303 	igb_vlan_mode(adapter->netdev, adapter->netdev->features);
9304 	igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
9305 
9306 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
9307 		igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
9308 }
9309 
9310 int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
9311 {
9312 	struct pci_dev *pdev = adapter->pdev;
9313 	struct e1000_mac_info *mac = &adapter->hw.mac;
9314 
9315 	mac->autoneg = 0;
9316 
9317 	/* Make sure dplx is at most 1 bit and lsb of speed is not set
9318 	 * for the switch() below to work
9319 	 */
9320 	if ((spd & 1) || (dplx & ~1))
9321 		goto err_inval;
9322 
9323 	/* Fiber NIC's only allow 1000 gbps Full duplex
9324 	 * and 100Mbps Full duplex for 100baseFx sfp
9325 	 */
9326 	if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
9327 		switch (spd + dplx) {
9328 		case SPEED_10 + DUPLEX_HALF:
9329 		case SPEED_10 + DUPLEX_FULL:
9330 		case SPEED_100 + DUPLEX_HALF:
9331 			goto err_inval;
9332 		default:
9333 			break;
9334 		}
9335 	}
9336 
9337 	switch (spd + dplx) {
9338 	case SPEED_10 + DUPLEX_HALF:
9339 		mac->forced_speed_duplex = ADVERTISE_10_HALF;
9340 		break;
9341 	case SPEED_10 + DUPLEX_FULL:
9342 		mac->forced_speed_duplex = ADVERTISE_10_FULL;
9343 		break;
9344 	case SPEED_100 + DUPLEX_HALF:
9345 		mac->forced_speed_duplex = ADVERTISE_100_HALF;
9346 		break;
9347 	case SPEED_100 + DUPLEX_FULL:
9348 		mac->forced_speed_duplex = ADVERTISE_100_FULL;
9349 		break;
9350 	case SPEED_1000 + DUPLEX_FULL:
9351 		mac->autoneg = 1;
9352 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
9353 		break;
9354 	case SPEED_1000 + DUPLEX_HALF: /* not supported */
9355 	default:
9356 		goto err_inval;
9357 	}
9358 
9359 	/* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
9360 	adapter->hw.phy.mdix = AUTO_ALL_MODES;
9361 
9362 	return 0;
9363 
9364 err_inval:
9365 	dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
9366 	return -EINVAL;
9367 }
9368 
9369 static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
9370 			  bool runtime)
9371 {
9372 	struct net_device *netdev = pci_get_drvdata(pdev);
9373 	struct igb_adapter *adapter = netdev_priv(netdev);
9374 	struct e1000_hw *hw = &adapter->hw;
9375 	u32 ctrl, rctl, status;
9376 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
9377 	bool wake;
9378 
9379 	rtnl_lock();
9380 	netif_device_detach(netdev);
9381 
9382 	if (netif_running(netdev))
9383 		__igb_close(netdev, true);
9384 
9385 	igb_ptp_suspend(adapter);
9386 
9387 	igb_clear_interrupt_scheme(adapter);
9388 	rtnl_unlock();
9389 
9390 	status = rd32(E1000_STATUS);
9391 	if (status & E1000_STATUS_LU)
9392 		wufc &= ~E1000_WUFC_LNKC;
9393 
9394 	if (wufc) {
9395 		igb_setup_rctl(adapter);
9396 		igb_set_rx_mode(netdev);
9397 
9398 		/* turn on all-multi mode if wake on multicast is enabled */
9399 		if (wufc & E1000_WUFC_MC) {
9400 			rctl = rd32(E1000_RCTL);
9401 			rctl |= E1000_RCTL_MPE;
9402 			wr32(E1000_RCTL, rctl);
9403 		}
9404 
9405 		ctrl = rd32(E1000_CTRL);
9406 		ctrl |= E1000_CTRL_ADVD3WUC;
9407 		wr32(E1000_CTRL, ctrl);
9408 
9409 		/* Allow time for pending master requests to run */
9410 		igb_disable_pcie_master(hw);
9411 
9412 		wr32(E1000_WUC, E1000_WUC_PME_EN);
9413 		wr32(E1000_WUFC, wufc);
9414 	} else {
9415 		wr32(E1000_WUC, 0);
9416 		wr32(E1000_WUFC, 0);
9417 	}
9418 
9419 	wake = wufc || adapter->en_mng_pt;
9420 	if (!wake)
9421 		igb_power_down_link(adapter);
9422 	else
9423 		igb_power_up_link(adapter);
9424 
9425 	if (enable_wake)
9426 		*enable_wake = wake;
9427 
9428 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
9429 	 * would have already happened in close and is redundant.
9430 	 */
9431 	igb_release_hw_control(adapter);
9432 
9433 	pci_disable_device(pdev);
9434 
9435 	return 0;
9436 }
9437 
9438 static void igb_deliver_wake_packet(struct net_device *netdev)
9439 {
9440 	struct igb_adapter *adapter = netdev_priv(netdev);
9441 	struct e1000_hw *hw = &adapter->hw;
9442 	struct sk_buff *skb;
9443 	u32 wupl;
9444 
9445 	wupl = rd32(E1000_WUPL) & E1000_WUPL_MASK;
9446 
9447 	/* WUPM stores only the first 128 bytes of the wake packet.
9448 	 * Read the packet only if we have the whole thing.
9449 	 */
9450 	if ((wupl == 0) || (wupl > E1000_WUPM_BYTES))
9451 		return;
9452 
9453 	skb = netdev_alloc_skb_ip_align(netdev, E1000_WUPM_BYTES);
9454 	if (!skb)
9455 		return;
9456 
9457 	skb_put(skb, wupl);
9458 
9459 	/* Ensure reads are 32-bit aligned */
9460 	wupl = roundup(wupl, 4);
9461 
9462 	memcpy_fromio(skb->data, hw->hw_addr + E1000_WUPM_REG(0), wupl);
9463 
9464 	skb->protocol = eth_type_trans(skb, netdev);
9465 	netif_rx(skb);
9466 }
9467 
9468 static int __maybe_unused igb_suspend(struct device *dev)
9469 {
9470 	return __igb_shutdown(to_pci_dev(dev), NULL, 0);
9471 }
9472 
9473 static int __maybe_unused __igb_resume(struct device *dev, bool rpm)
9474 {
9475 	struct pci_dev *pdev = to_pci_dev(dev);
9476 	struct net_device *netdev = pci_get_drvdata(pdev);
9477 	struct igb_adapter *adapter = netdev_priv(netdev);
9478 	struct e1000_hw *hw = &adapter->hw;
9479 	u32 err, val;
9480 
9481 	pci_set_power_state(pdev, PCI_D0);
9482 	pci_restore_state(pdev);
9483 	pci_save_state(pdev);
9484 
9485 	if (!pci_device_is_present(pdev))
9486 		return -ENODEV;
9487 	err = pci_enable_device_mem(pdev);
9488 	if (err) {
9489 		dev_err(&pdev->dev,
9490 			"igb: Cannot enable PCI device from suspend\n");
9491 		return err;
9492 	}
9493 	pci_set_master(pdev);
9494 
9495 	pci_enable_wake(pdev, PCI_D3hot, 0);
9496 	pci_enable_wake(pdev, PCI_D3cold, 0);
9497 
9498 	if (igb_init_interrupt_scheme(adapter, true)) {
9499 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9500 		return -ENOMEM;
9501 	}
9502 
9503 	igb_reset(adapter);
9504 
9505 	/* let the f/w know that the h/w is now under the control of the
9506 	 * driver.
9507 	 */
9508 	igb_get_hw_control(adapter);
9509 
9510 	val = rd32(E1000_WUS);
9511 	if (val & WAKE_PKT_WUS)
9512 		igb_deliver_wake_packet(netdev);
9513 
9514 	wr32(E1000_WUS, ~0);
9515 
9516 	if (!rpm)
9517 		rtnl_lock();
9518 	if (!err && netif_running(netdev))
9519 		err = __igb_open(netdev, true);
9520 
9521 	if (!err)
9522 		netif_device_attach(netdev);
9523 	if (!rpm)
9524 		rtnl_unlock();
9525 
9526 	return err;
9527 }
9528 
9529 static int __maybe_unused igb_resume(struct device *dev)
9530 {
9531 	return __igb_resume(dev, false);
9532 }
9533 
9534 static int __maybe_unused igb_runtime_idle(struct device *dev)
9535 {
9536 	struct net_device *netdev = dev_get_drvdata(dev);
9537 	struct igb_adapter *adapter = netdev_priv(netdev);
9538 
9539 	if (!igb_has_link(adapter))
9540 		pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
9541 
9542 	return -EBUSY;
9543 }
9544 
9545 static int __maybe_unused igb_runtime_suspend(struct device *dev)
9546 {
9547 	return __igb_shutdown(to_pci_dev(dev), NULL, 1);
9548 }
9549 
9550 static int __maybe_unused igb_runtime_resume(struct device *dev)
9551 {
9552 	return __igb_resume(dev, true);
9553 }
9554 
9555 static void igb_shutdown(struct pci_dev *pdev)
9556 {
9557 	bool wake;
9558 
9559 	__igb_shutdown(pdev, &wake, 0);
9560 
9561 	if (system_state == SYSTEM_POWER_OFF) {
9562 		pci_wake_from_d3(pdev, wake);
9563 		pci_set_power_state(pdev, PCI_D3hot);
9564 	}
9565 }
9566 
9567 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
9568 {
9569 #ifdef CONFIG_PCI_IOV
9570 	int err;
9571 
9572 	if (num_vfs == 0) {
9573 		return igb_disable_sriov(dev, true);
9574 	} else {
9575 		err = igb_enable_sriov(dev, num_vfs, true);
9576 		return err ? err : num_vfs;
9577 	}
9578 #endif
9579 	return 0;
9580 }
9581 
9582 /**
9583  *  igb_io_error_detected - called when PCI error is detected
9584  *  @pdev: Pointer to PCI device
9585  *  @state: The current pci connection state
9586  *
9587  *  This function is called after a PCI bus error affecting
9588  *  this device has been detected.
9589  **/
9590 static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
9591 					      pci_channel_state_t state)
9592 {
9593 	struct net_device *netdev = pci_get_drvdata(pdev);
9594 	struct igb_adapter *adapter = netdev_priv(netdev);
9595 
9596 	if (state == pci_channel_io_normal) {
9597 		dev_warn(&pdev->dev, "Non-correctable non-fatal error reported.\n");
9598 		return PCI_ERS_RESULT_CAN_RECOVER;
9599 	}
9600 
9601 	netif_device_detach(netdev);
9602 
9603 	if (state == pci_channel_io_perm_failure)
9604 		return PCI_ERS_RESULT_DISCONNECT;
9605 
9606 	if (netif_running(netdev))
9607 		igb_down(adapter);
9608 	pci_disable_device(pdev);
9609 
9610 	/* Request a slot reset. */
9611 	return PCI_ERS_RESULT_NEED_RESET;
9612 }
9613 
9614 /**
9615  *  igb_io_slot_reset - called after the pci bus has been reset.
9616  *  @pdev: Pointer to PCI device
9617  *
9618  *  Restart the card from scratch, as if from a cold-boot. Implementation
9619  *  resembles the first-half of the __igb_resume routine.
9620  **/
9621 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
9622 {
9623 	struct net_device *netdev = pci_get_drvdata(pdev);
9624 	struct igb_adapter *adapter = netdev_priv(netdev);
9625 	struct e1000_hw *hw = &adapter->hw;
9626 	pci_ers_result_t result;
9627 
9628 	if (pci_enable_device_mem(pdev)) {
9629 		dev_err(&pdev->dev,
9630 			"Cannot re-enable PCI device after reset.\n");
9631 		result = PCI_ERS_RESULT_DISCONNECT;
9632 	} else {
9633 		pci_set_master(pdev);
9634 		pci_restore_state(pdev);
9635 		pci_save_state(pdev);
9636 
9637 		pci_enable_wake(pdev, PCI_D3hot, 0);
9638 		pci_enable_wake(pdev, PCI_D3cold, 0);
9639 
9640 		/* In case of PCI error, adapter lose its HW address
9641 		 * so we should re-assign it here.
9642 		 */
9643 		hw->hw_addr = adapter->io_addr;
9644 
9645 		igb_reset(adapter);
9646 		wr32(E1000_WUS, ~0);
9647 		result = PCI_ERS_RESULT_RECOVERED;
9648 	}
9649 
9650 	return result;
9651 }
9652 
9653 /**
9654  *  igb_io_resume - called when traffic can start flowing again.
9655  *  @pdev: Pointer to PCI device
9656  *
9657  *  This callback is called when the error recovery driver tells us that
9658  *  its OK to resume normal operation. Implementation resembles the
9659  *  second-half of the __igb_resume routine.
9660  */
9661 static void igb_io_resume(struct pci_dev *pdev)
9662 {
9663 	struct net_device *netdev = pci_get_drvdata(pdev);
9664 	struct igb_adapter *adapter = netdev_priv(netdev);
9665 
9666 	if (netif_running(netdev)) {
9667 		if (igb_up(adapter)) {
9668 			dev_err(&pdev->dev, "igb_up failed after reset\n");
9669 			return;
9670 		}
9671 	}
9672 
9673 	netif_device_attach(netdev);
9674 
9675 	/* let the f/w know that the h/w is now under the control of the
9676 	 * driver.
9677 	 */
9678 	igb_get_hw_control(adapter);
9679 }
9680 
9681 /**
9682  *  igb_rar_set_index - Sync RAL[index] and RAH[index] registers with MAC table
9683  *  @adapter: Pointer to adapter structure
9684  *  @index: Index of the RAR entry which need to be synced with MAC table
9685  **/
9686 static void igb_rar_set_index(struct igb_adapter *adapter, u32 index)
9687 {
9688 	struct e1000_hw *hw = &adapter->hw;
9689 	u32 rar_low, rar_high;
9690 	u8 *addr = adapter->mac_table[index].addr;
9691 
9692 	/* HW expects these to be in network order when they are plugged
9693 	 * into the registers which are little endian.  In order to guarantee
9694 	 * that ordering we need to do an leXX_to_cpup here in order to be
9695 	 * ready for the byteswap that occurs with writel
9696 	 */
9697 	rar_low = le32_to_cpup((__le32 *)(addr));
9698 	rar_high = le16_to_cpup((__le16 *)(addr + 4));
9699 
9700 	/* Indicate to hardware the Address is Valid. */
9701 	if (adapter->mac_table[index].state & IGB_MAC_STATE_IN_USE) {
9702 		if (is_valid_ether_addr(addr))
9703 			rar_high |= E1000_RAH_AV;
9704 
9705 		if (adapter->mac_table[index].state & IGB_MAC_STATE_SRC_ADDR)
9706 			rar_high |= E1000_RAH_ASEL_SRC_ADDR;
9707 
9708 		switch (hw->mac.type) {
9709 		case e1000_82575:
9710 		case e1000_i210:
9711 			if (adapter->mac_table[index].state &
9712 			    IGB_MAC_STATE_QUEUE_STEERING)
9713 				rar_high |= E1000_RAH_QSEL_ENABLE;
9714 
9715 			rar_high |= E1000_RAH_POOL_1 *
9716 				    adapter->mac_table[index].queue;
9717 			break;
9718 		default:
9719 			rar_high |= E1000_RAH_POOL_1 <<
9720 				    adapter->mac_table[index].queue;
9721 			break;
9722 		}
9723 	}
9724 
9725 	wr32(E1000_RAL(index), rar_low);
9726 	wrfl();
9727 	wr32(E1000_RAH(index), rar_high);
9728 	wrfl();
9729 }
9730 
9731 static int igb_set_vf_mac(struct igb_adapter *adapter,
9732 			  int vf, unsigned char *mac_addr)
9733 {
9734 	struct e1000_hw *hw = &adapter->hw;
9735 	/* VF MAC addresses start at end of receive addresses and moves
9736 	 * towards the first, as a result a collision should not be possible
9737 	 */
9738 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
9739 	unsigned char *vf_mac_addr = adapter->vf_data[vf].vf_mac_addresses;
9740 
9741 	ether_addr_copy(vf_mac_addr, mac_addr);
9742 	ether_addr_copy(adapter->mac_table[rar_entry].addr, mac_addr);
9743 	adapter->mac_table[rar_entry].queue = vf;
9744 	adapter->mac_table[rar_entry].state |= IGB_MAC_STATE_IN_USE;
9745 	igb_rar_set_index(adapter, rar_entry);
9746 
9747 	return 0;
9748 }
9749 
9750 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
9751 {
9752 	struct igb_adapter *adapter = netdev_priv(netdev);
9753 
9754 	if (vf >= adapter->vfs_allocated_count)
9755 		return -EINVAL;
9756 
9757 	/* Setting the VF MAC to 0 reverts the IGB_VF_FLAG_PF_SET_MAC
9758 	 * flag and allows to overwrite the MAC via VF netdev.  This
9759 	 * is necessary to allow libvirt a way to restore the original
9760 	 * MAC after unbinding vfio-pci and reloading igbvf after shutting
9761 	 * down a VM.
9762 	 */
9763 	if (is_zero_ether_addr(mac)) {
9764 		adapter->vf_data[vf].flags &= ~IGB_VF_FLAG_PF_SET_MAC;
9765 		dev_info(&adapter->pdev->dev,
9766 			 "remove administratively set MAC on VF %d\n",
9767 			 vf);
9768 	} else if (is_valid_ether_addr(mac)) {
9769 		adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
9770 		dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n",
9771 			 mac, vf);
9772 		dev_info(&adapter->pdev->dev,
9773 			 "Reload the VF driver to make this change effective.");
9774 		/* Generate additional warning if PF is down */
9775 		if (test_bit(__IGB_DOWN, &adapter->state)) {
9776 			dev_warn(&adapter->pdev->dev,
9777 				 "The VF MAC address has been set, but the PF device is not up.\n");
9778 			dev_warn(&adapter->pdev->dev,
9779 				 "Bring the PF device up before attempting to use the VF device.\n");
9780 		}
9781 	} else {
9782 		return -EINVAL;
9783 	}
9784 	return igb_set_vf_mac(adapter, vf, mac);
9785 }
9786 
9787 static int igb_link_mbps(int internal_link_speed)
9788 {
9789 	switch (internal_link_speed) {
9790 	case SPEED_100:
9791 		return 100;
9792 	case SPEED_1000:
9793 		return 1000;
9794 	default:
9795 		return 0;
9796 	}
9797 }
9798 
9799 static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
9800 				  int link_speed)
9801 {
9802 	int rf_dec, rf_int;
9803 	u32 bcnrc_val;
9804 
9805 	if (tx_rate != 0) {
9806 		/* Calculate the rate factor values to set */
9807 		rf_int = link_speed / tx_rate;
9808 		rf_dec = (link_speed - (rf_int * tx_rate));
9809 		rf_dec = (rf_dec * BIT(E1000_RTTBCNRC_RF_INT_SHIFT)) /
9810 			 tx_rate;
9811 
9812 		bcnrc_val = E1000_RTTBCNRC_RS_ENA;
9813 		bcnrc_val |= FIELD_PREP(E1000_RTTBCNRC_RF_INT_MASK, rf_int);
9814 		bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
9815 	} else {
9816 		bcnrc_val = 0;
9817 	}
9818 
9819 	wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
9820 	/* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
9821 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
9822 	 */
9823 	wr32(E1000_RTTBCNRM, 0x14);
9824 	wr32(E1000_RTTBCNRC, bcnrc_val);
9825 }
9826 
9827 static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
9828 {
9829 	int actual_link_speed, i;
9830 	bool reset_rate = false;
9831 
9832 	/* VF TX rate limit was not set or not supported */
9833 	if ((adapter->vf_rate_link_speed == 0) ||
9834 	    (adapter->hw.mac.type != e1000_82576))
9835 		return;
9836 
9837 	actual_link_speed = igb_link_mbps(adapter->link_speed);
9838 	if (actual_link_speed != adapter->vf_rate_link_speed) {
9839 		reset_rate = true;
9840 		adapter->vf_rate_link_speed = 0;
9841 		dev_info(&adapter->pdev->dev,
9842 			 "Link speed has been changed. VF Transmit rate is disabled\n");
9843 	}
9844 
9845 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
9846 		if (reset_rate)
9847 			adapter->vf_data[i].tx_rate = 0;
9848 
9849 		igb_set_vf_rate_limit(&adapter->hw, i,
9850 				      adapter->vf_data[i].tx_rate,
9851 				      actual_link_speed);
9852 	}
9853 }
9854 
9855 static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
9856 			     int min_tx_rate, int max_tx_rate)
9857 {
9858 	struct igb_adapter *adapter = netdev_priv(netdev);
9859 	struct e1000_hw *hw = &adapter->hw;
9860 	int actual_link_speed;
9861 
9862 	if (hw->mac.type != e1000_82576)
9863 		return -EOPNOTSUPP;
9864 
9865 	if (min_tx_rate)
9866 		return -EINVAL;
9867 
9868 	actual_link_speed = igb_link_mbps(adapter->link_speed);
9869 	if ((vf >= adapter->vfs_allocated_count) ||
9870 	    (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
9871 	    (max_tx_rate < 0) ||
9872 	    (max_tx_rate > actual_link_speed))
9873 		return -EINVAL;
9874 
9875 	adapter->vf_rate_link_speed = actual_link_speed;
9876 	adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
9877 	igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
9878 
9879 	return 0;
9880 }
9881 
9882 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
9883 				   bool setting)
9884 {
9885 	struct igb_adapter *adapter = netdev_priv(netdev);
9886 	struct e1000_hw *hw = &adapter->hw;
9887 	u32 reg_val, reg_offset;
9888 
9889 	if (!adapter->vfs_allocated_count)
9890 		return -EOPNOTSUPP;
9891 
9892 	if (vf >= adapter->vfs_allocated_count)
9893 		return -EINVAL;
9894 
9895 	reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
9896 	reg_val = rd32(reg_offset);
9897 	if (setting)
9898 		reg_val |= (BIT(vf) |
9899 			    BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9900 	else
9901 		reg_val &= ~(BIT(vf) |
9902 			     BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9903 	wr32(reg_offset, reg_val);
9904 
9905 	adapter->vf_data[vf].spoofchk_enabled = setting;
9906 	return 0;
9907 }
9908 
9909 static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf, bool setting)
9910 {
9911 	struct igb_adapter *adapter = netdev_priv(netdev);
9912 
9913 	if (vf >= adapter->vfs_allocated_count)
9914 		return -EINVAL;
9915 	if (adapter->vf_data[vf].trusted == setting)
9916 		return 0;
9917 
9918 	adapter->vf_data[vf].trusted = setting;
9919 
9920 	dev_info(&adapter->pdev->dev, "VF %u is %strusted\n",
9921 		 vf, setting ? "" : "not ");
9922 	return 0;
9923 }
9924 
9925 static int igb_ndo_get_vf_config(struct net_device *netdev,
9926 				 int vf, struct ifla_vf_info *ivi)
9927 {
9928 	struct igb_adapter *adapter = netdev_priv(netdev);
9929 	if (vf >= adapter->vfs_allocated_count)
9930 		return -EINVAL;
9931 	ivi->vf = vf;
9932 	memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
9933 	ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
9934 	ivi->min_tx_rate = 0;
9935 	ivi->vlan = adapter->vf_data[vf].pf_vlan;
9936 	ivi->qos = adapter->vf_data[vf].pf_qos;
9937 	ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
9938 	ivi->trusted = adapter->vf_data[vf].trusted;
9939 	return 0;
9940 }
9941 
9942 static void igb_vmm_control(struct igb_adapter *adapter)
9943 {
9944 	struct e1000_hw *hw = &adapter->hw;
9945 	u32 reg;
9946 
9947 	switch (hw->mac.type) {
9948 	case e1000_82575:
9949 	case e1000_i210:
9950 	case e1000_i211:
9951 	case e1000_i354:
9952 	default:
9953 		/* replication is not supported for 82575 */
9954 		return;
9955 	case e1000_82576:
9956 		/* notify HW that the MAC is adding vlan tags */
9957 		reg = rd32(E1000_DTXCTL);
9958 		reg |= E1000_DTXCTL_VLAN_ADDED;
9959 		wr32(E1000_DTXCTL, reg);
9960 		fallthrough;
9961 	case e1000_82580:
9962 		/* enable replication vlan tag stripping */
9963 		reg = rd32(E1000_RPLOLR);
9964 		reg |= E1000_RPLOLR_STRVLAN;
9965 		wr32(E1000_RPLOLR, reg);
9966 		fallthrough;
9967 	case e1000_i350:
9968 		/* none of the above registers are supported by i350 */
9969 		break;
9970 	}
9971 
9972 	if (adapter->vfs_allocated_count) {
9973 		igb_vmdq_set_loopback_pf(hw, true);
9974 		igb_vmdq_set_replication_pf(hw, true);
9975 		igb_vmdq_set_anti_spoofing_pf(hw, true,
9976 					      adapter->vfs_allocated_count);
9977 	} else {
9978 		igb_vmdq_set_loopback_pf(hw, false);
9979 		igb_vmdq_set_replication_pf(hw, false);
9980 	}
9981 }
9982 
9983 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
9984 {
9985 	struct e1000_hw *hw = &adapter->hw;
9986 	u32 dmac_thr;
9987 	u16 hwm;
9988 	u32 reg;
9989 
9990 	if (hw->mac.type > e1000_82580) {
9991 		if (adapter->flags & IGB_FLAG_DMAC) {
9992 			/* force threshold to 0. */
9993 			wr32(E1000_DMCTXTH, 0);
9994 
9995 			/* DMA Coalescing high water mark needs to be greater
9996 			 * than the Rx threshold. Set hwm to PBA - max frame
9997 			 * size in 16B units, capping it at PBA - 6KB.
9998 			 */
9999 			hwm = 64 * (pba - 6);
10000 			reg = rd32(E1000_FCRTC);
10001 			reg &= ~E1000_FCRTC_RTH_COAL_MASK;
10002 			reg |= FIELD_PREP(E1000_FCRTC_RTH_COAL_MASK, hwm);
10003 			wr32(E1000_FCRTC, reg);
10004 
10005 			/* Set the DMA Coalescing Rx threshold to PBA - 2 * max
10006 			 * frame size, capping it at PBA - 10KB.
10007 			 */
10008 			dmac_thr = pba - 10;
10009 			reg = rd32(E1000_DMACR);
10010 			reg &= ~E1000_DMACR_DMACTHR_MASK;
10011 			reg |= FIELD_PREP(E1000_DMACR_DMACTHR_MASK, dmac_thr);
10012 
10013 			/* transition to L0x or L1 if available..*/
10014 			reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
10015 
10016 			/* watchdog timer= +-1000 usec in 32usec intervals */
10017 			reg |= (1000 >> 5);
10018 
10019 			/* Disable BMC-to-OS Watchdog Enable */
10020 			if (hw->mac.type != e1000_i354)
10021 				reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
10022 			wr32(E1000_DMACR, reg);
10023 
10024 			/* no lower threshold to disable
10025 			 * coalescing(smart fifb)-UTRESH=0
10026 			 */
10027 			wr32(E1000_DMCRTRH, 0);
10028 
10029 			reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
10030 
10031 			wr32(E1000_DMCTLX, reg);
10032 
10033 			/* free space in tx packet buffer to wake from
10034 			 * DMA coal
10035 			 */
10036 			wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
10037 			     (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
10038 		}
10039 
10040 		if (hw->mac.type >= e1000_i210 ||
10041 		    (adapter->flags & IGB_FLAG_DMAC)) {
10042 			reg = rd32(E1000_PCIEMISC);
10043 			reg |= E1000_PCIEMISC_LX_DECISION;
10044 			wr32(E1000_PCIEMISC, reg);
10045 		} /* endif adapter->dmac is not disabled */
10046 	} else if (hw->mac.type == e1000_82580) {
10047 		u32 reg = rd32(E1000_PCIEMISC);
10048 
10049 		wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
10050 		wr32(E1000_DMACR, 0);
10051 	}
10052 }
10053 
10054 /**
10055  *  igb_read_i2c_byte - Reads 8 bit word over I2C
10056  *  @hw: pointer to hardware structure
10057  *  @byte_offset: byte offset to read
10058  *  @dev_addr: device address
10059  *  @data: value read
10060  *
10061  *  Performs byte read operation over I2C interface at
10062  *  a specified device address.
10063  **/
10064 s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
10065 		      u8 dev_addr, u8 *data)
10066 {
10067 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
10068 	struct i2c_client *this_client = adapter->i2c_client;
10069 	s32 status;
10070 	u16 swfw_mask = 0;
10071 
10072 	if (!this_client)
10073 		return E1000_ERR_I2C;
10074 
10075 	swfw_mask = E1000_SWFW_PHY0_SM;
10076 
10077 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
10078 		return E1000_ERR_SWFW_SYNC;
10079 
10080 	status = i2c_smbus_read_byte_data(this_client, byte_offset);
10081 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
10082 
10083 	if (status < 0)
10084 		return E1000_ERR_I2C;
10085 	else {
10086 		*data = status;
10087 		return 0;
10088 	}
10089 }
10090 
10091 /**
10092  *  igb_write_i2c_byte - Writes 8 bit word over I2C
10093  *  @hw: pointer to hardware structure
10094  *  @byte_offset: byte offset to write
10095  *  @dev_addr: device address
10096  *  @data: value to write
10097  *
10098  *  Performs byte write operation over I2C interface at
10099  *  a specified device address.
10100  **/
10101 s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
10102 		       u8 dev_addr, u8 data)
10103 {
10104 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
10105 	struct i2c_client *this_client = adapter->i2c_client;
10106 	s32 status;
10107 	u16 swfw_mask = E1000_SWFW_PHY0_SM;
10108 
10109 	if (!this_client)
10110 		return E1000_ERR_I2C;
10111 
10112 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
10113 		return E1000_ERR_SWFW_SYNC;
10114 	status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
10115 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
10116 
10117 	if (status)
10118 		return E1000_ERR_I2C;
10119 	else
10120 		return 0;
10121 
10122 }
10123 
10124 int igb_reinit_queues(struct igb_adapter *adapter)
10125 {
10126 	struct net_device *netdev = adapter->netdev;
10127 	struct pci_dev *pdev = adapter->pdev;
10128 	int err = 0;
10129 
10130 	if (netif_running(netdev))
10131 		igb_close(netdev);
10132 
10133 	igb_reset_interrupt_capability(adapter);
10134 
10135 	if (igb_init_interrupt_scheme(adapter, true)) {
10136 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
10137 		return -ENOMEM;
10138 	}
10139 
10140 	if (netif_running(netdev))
10141 		err = igb_open(netdev);
10142 
10143 	return err;
10144 }
10145 
10146 static void igb_nfc_filter_exit(struct igb_adapter *adapter)
10147 {
10148 	struct igb_nfc_filter *rule;
10149 
10150 	spin_lock(&adapter->nfc_lock);
10151 
10152 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
10153 		igb_erase_filter(adapter, rule);
10154 
10155 	hlist_for_each_entry(rule, &adapter->cls_flower_list, nfc_node)
10156 		igb_erase_filter(adapter, rule);
10157 
10158 	spin_unlock(&adapter->nfc_lock);
10159 }
10160 
10161 static void igb_nfc_filter_restore(struct igb_adapter *adapter)
10162 {
10163 	struct igb_nfc_filter *rule;
10164 
10165 	spin_lock(&adapter->nfc_lock);
10166 
10167 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
10168 		igb_add_filter(adapter, rule);
10169 
10170 	spin_unlock(&adapter->nfc_lock);
10171 }
10172 /* igb_main.c */
10173