xref: /linux/drivers/net/ethernet/intel/igb/igb_main.c (revision 780f6a9afe8b0e303406a39f6968cf1daa6c3d51)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright(c) 2007 - 2018 Intel Corporation. */
3 
4 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
5 
6 #include <linux/module.h>
7 #include <linux/types.h>
8 #include <linux/init.h>
9 #include <linux/bitops.h>
10 #include <linux/vmalloc.h>
11 #include <linux/pagemap.h>
12 #include <linux/netdevice.h>
13 #include <linux/ipv6.h>
14 #include <linux/slab.h>
15 #include <net/checksum.h>
16 #include <net/ip6_checksum.h>
17 #include <net/pkt_sched.h>
18 #include <net/pkt_cls.h>
19 #include <linux/net_tstamp.h>
20 #include <linux/mii.h>
21 #include <linux/ethtool.h>
22 #include <linux/if.h>
23 #include <linux/if_vlan.h>
24 #include <linux/pci.h>
25 #include <linux/delay.h>
26 #include <linux/interrupt.h>
27 #include <linux/ip.h>
28 #include <linux/tcp.h>
29 #include <linux/sctp.h>
30 #include <linux/if_ether.h>
31 #include <linux/aer.h>
32 #include <linux/prefetch.h>
33 #include <linux/bpf.h>
34 #include <linux/bpf_trace.h>
35 #include <linux/pm_runtime.h>
36 #include <linux/etherdevice.h>
37 #ifdef CONFIG_IGB_DCA
38 #include <linux/dca.h>
39 #endif
40 #include <linux/i2c.h>
41 #include "igb.h"
42 
43 enum queue_mode {
44 	QUEUE_MODE_STRICT_PRIORITY,
45 	QUEUE_MODE_STREAM_RESERVATION,
46 };
47 
48 enum tx_queue_prio {
49 	TX_QUEUE_PRIO_HIGH,
50 	TX_QUEUE_PRIO_LOW,
51 };
52 
53 char igb_driver_name[] = "igb";
54 static const char igb_driver_string[] =
55 				"Intel(R) Gigabit Ethernet Network Driver";
56 static const char igb_copyright[] =
57 				"Copyright (c) 2007-2014 Intel Corporation.";
58 
59 static const struct e1000_info *igb_info_tbl[] = {
60 	[board_82575] = &e1000_82575_info,
61 };
62 
63 static const struct pci_device_id igb_pci_tbl[] = {
64 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
65 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_SGMII) },
66 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
67 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I211_COPPER), board_82575 },
68 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER), board_82575 },
69 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_FIBER), board_82575 },
70 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES), board_82575 },
71 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SGMII), board_82575 },
72 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER_FLASHLESS), board_82575 },
73 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES_FLASHLESS), board_82575 },
74 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_COPPER), board_82575 },
75 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_FIBER), board_82575 },
76 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SERDES), board_82575 },
77 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SGMII), board_82575 },
78 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER), board_82575 },
79 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_FIBER), board_82575 },
80 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_QUAD_FIBER), board_82575 },
81 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SERDES), board_82575 },
82 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SGMII), board_82575 },
83 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER_DUAL), board_82575 },
84 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SGMII), board_82575 },
85 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SERDES), board_82575 },
86 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_BACKPLANE), board_82575 },
87 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SFP), board_82575 },
88 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
89 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
90 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
91 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
92 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
93 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
94 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER_ET2), board_82575 },
95 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
96 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
97 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
98 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
99 	/* required last entry */
100 	{0, }
101 };
102 
103 MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
104 
105 static int igb_setup_all_tx_resources(struct igb_adapter *);
106 static int igb_setup_all_rx_resources(struct igb_adapter *);
107 static void igb_free_all_tx_resources(struct igb_adapter *);
108 static void igb_free_all_rx_resources(struct igb_adapter *);
109 static void igb_setup_mrqc(struct igb_adapter *);
110 static int igb_probe(struct pci_dev *, const struct pci_device_id *);
111 static void igb_remove(struct pci_dev *pdev);
112 static int igb_sw_init(struct igb_adapter *);
113 int igb_open(struct net_device *);
114 int igb_close(struct net_device *);
115 static void igb_configure(struct igb_adapter *);
116 static void igb_configure_tx(struct igb_adapter *);
117 static void igb_configure_rx(struct igb_adapter *);
118 static void igb_clean_all_tx_rings(struct igb_adapter *);
119 static void igb_clean_all_rx_rings(struct igb_adapter *);
120 static void igb_clean_tx_ring(struct igb_ring *);
121 static void igb_clean_rx_ring(struct igb_ring *);
122 static void igb_set_rx_mode(struct net_device *);
123 static void igb_update_phy_info(struct timer_list *);
124 static void igb_watchdog(struct timer_list *);
125 static void igb_watchdog_task(struct work_struct *);
126 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb, struct net_device *);
127 static void igb_get_stats64(struct net_device *dev,
128 			    struct rtnl_link_stats64 *stats);
129 static int igb_change_mtu(struct net_device *, int);
130 static int igb_set_mac(struct net_device *, void *);
131 static void igb_set_uta(struct igb_adapter *adapter, bool set);
132 static irqreturn_t igb_intr(int irq, void *);
133 static irqreturn_t igb_intr_msi(int irq, void *);
134 static irqreturn_t igb_msix_other(int irq, void *);
135 static irqreturn_t igb_msix_ring(int irq, void *);
136 #ifdef CONFIG_IGB_DCA
137 static void igb_update_dca(struct igb_q_vector *);
138 static void igb_setup_dca(struct igb_adapter *);
139 #endif /* CONFIG_IGB_DCA */
140 static int igb_poll(struct napi_struct *, int);
141 static bool igb_clean_tx_irq(struct igb_q_vector *, int);
142 static int igb_clean_rx_irq(struct igb_q_vector *, int);
143 static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
144 static void igb_tx_timeout(struct net_device *, unsigned int txqueue);
145 static void igb_reset_task(struct work_struct *);
146 static void igb_vlan_mode(struct net_device *netdev,
147 			  netdev_features_t features);
148 static int igb_vlan_rx_add_vid(struct net_device *, __be16, u16);
149 static int igb_vlan_rx_kill_vid(struct net_device *, __be16, u16);
150 static void igb_restore_vlan(struct igb_adapter *);
151 static void igb_rar_set_index(struct igb_adapter *, u32);
152 static void igb_ping_all_vfs(struct igb_adapter *);
153 static void igb_msg_task(struct igb_adapter *);
154 static void igb_vmm_control(struct igb_adapter *);
155 static int igb_set_vf_mac(struct igb_adapter *, int, unsigned char *);
156 static void igb_flush_mac_table(struct igb_adapter *);
157 static int igb_available_rars(struct igb_adapter *, u8);
158 static void igb_set_default_mac_filter(struct igb_adapter *);
159 static int igb_uc_sync(struct net_device *, const unsigned char *);
160 static int igb_uc_unsync(struct net_device *, const unsigned char *);
161 static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
162 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac);
163 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
164 			       int vf, u16 vlan, u8 qos, __be16 vlan_proto);
165 static int igb_ndo_set_vf_bw(struct net_device *, int, int, int);
166 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
167 				   bool setting);
168 static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf,
169 				bool setting);
170 static int igb_ndo_get_vf_config(struct net_device *netdev, int vf,
171 				 struct ifla_vf_info *ivi);
172 static void igb_check_vf_rate_limit(struct igb_adapter *);
173 static void igb_nfc_filter_exit(struct igb_adapter *adapter);
174 static void igb_nfc_filter_restore(struct igb_adapter *adapter);
175 
176 #ifdef CONFIG_PCI_IOV
177 static int igb_vf_configure(struct igb_adapter *adapter, int vf);
178 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs);
179 static int igb_disable_sriov(struct pci_dev *dev);
180 static int igb_pci_disable_sriov(struct pci_dev *dev);
181 #endif
182 
183 static int igb_suspend(struct device *);
184 static int igb_resume(struct device *);
185 static int igb_runtime_suspend(struct device *dev);
186 static int igb_runtime_resume(struct device *dev);
187 static int igb_runtime_idle(struct device *dev);
188 static const struct dev_pm_ops igb_pm_ops = {
189 	SET_SYSTEM_SLEEP_PM_OPS(igb_suspend, igb_resume)
190 	SET_RUNTIME_PM_OPS(igb_runtime_suspend, igb_runtime_resume,
191 			igb_runtime_idle)
192 };
193 static void igb_shutdown(struct pci_dev *);
194 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs);
195 #ifdef CONFIG_IGB_DCA
196 static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
197 static struct notifier_block dca_notifier = {
198 	.notifier_call	= igb_notify_dca,
199 	.next		= NULL,
200 	.priority	= 0
201 };
202 #endif
203 #ifdef CONFIG_PCI_IOV
204 static unsigned int max_vfs;
205 module_param(max_vfs, uint, 0);
206 MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate per physical function");
207 #endif /* CONFIG_PCI_IOV */
208 
209 static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
210 		     pci_channel_state_t);
211 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
212 static void igb_io_resume(struct pci_dev *);
213 
214 static const struct pci_error_handlers igb_err_handler = {
215 	.error_detected = igb_io_error_detected,
216 	.slot_reset = igb_io_slot_reset,
217 	.resume = igb_io_resume,
218 };
219 
220 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba);
221 
222 static struct pci_driver igb_driver = {
223 	.name     = igb_driver_name,
224 	.id_table = igb_pci_tbl,
225 	.probe    = igb_probe,
226 	.remove   = igb_remove,
227 #ifdef CONFIG_PM
228 	.driver.pm = &igb_pm_ops,
229 #endif
230 	.shutdown = igb_shutdown,
231 	.sriov_configure = igb_pci_sriov_configure,
232 	.err_handler = &igb_err_handler
233 };
234 
235 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
236 MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
237 MODULE_LICENSE("GPL v2");
238 
239 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
240 static int debug = -1;
241 module_param(debug, int, 0);
242 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
243 
244 struct igb_reg_info {
245 	u32 ofs;
246 	char *name;
247 };
248 
249 static const struct igb_reg_info igb_reg_info_tbl[] = {
250 
251 	/* General Registers */
252 	{E1000_CTRL, "CTRL"},
253 	{E1000_STATUS, "STATUS"},
254 	{E1000_CTRL_EXT, "CTRL_EXT"},
255 
256 	/* Interrupt Registers */
257 	{E1000_ICR, "ICR"},
258 
259 	/* RX Registers */
260 	{E1000_RCTL, "RCTL"},
261 	{E1000_RDLEN(0), "RDLEN"},
262 	{E1000_RDH(0), "RDH"},
263 	{E1000_RDT(0), "RDT"},
264 	{E1000_RXDCTL(0), "RXDCTL"},
265 	{E1000_RDBAL(0), "RDBAL"},
266 	{E1000_RDBAH(0), "RDBAH"},
267 
268 	/* TX Registers */
269 	{E1000_TCTL, "TCTL"},
270 	{E1000_TDBAL(0), "TDBAL"},
271 	{E1000_TDBAH(0), "TDBAH"},
272 	{E1000_TDLEN(0), "TDLEN"},
273 	{E1000_TDH(0), "TDH"},
274 	{E1000_TDT(0), "TDT"},
275 	{E1000_TXDCTL(0), "TXDCTL"},
276 	{E1000_TDFH, "TDFH"},
277 	{E1000_TDFT, "TDFT"},
278 	{E1000_TDFHS, "TDFHS"},
279 	{E1000_TDFPC, "TDFPC"},
280 
281 	/* List Terminator */
282 	{}
283 };
284 
285 /* igb_regdump - register printout routine */
286 static void igb_regdump(struct e1000_hw *hw, struct igb_reg_info *reginfo)
287 {
288 	int n = 0;
289 	char rname[16];
290 	u32 regs[8];
291 
292 	switch (reginfo->ofs) {
293 	case E1000_RDLEN(0):
294 		for (n = 0; n < 4; n++)
295 			regs[n] = rd32(E1000_RDLEN(n));
296 		break;
297 	case E1000_RDH(0):
298 		for (n = 0; n < 4; n++)
299 			regs[n] = rd32(E1000_RDH(n));
300 		break;
301 	case E1000_RDT(0):
302 		for (n = 0; n < 4; n++)
303 			regs[n] = rd32(E1000_RDT(n));
304 		break;
305 	case E1000_RXDCTL(0):
306 		for (n = 0; n < 4; n++)
307 			regs[n] = rd32(E1000_RXDCTL(n));
308 		break;
309 	case E1000_RDBAL(0):
310 		for (n = 0; n < 4; n++)
311 			regs[n] = rd32(E1000_RDBAL(n));
312 		break;
313 	case E1000_RDBAH(0):
314 		for (n = 0; n < 4; n++)
315 			regs[n] = rd32(E1000_RDBAH(n));
316 		break;
317 	case E1000_TDBAL(0):
318 		for (n = 0; n < 4; n++)
319 			regs[n] = rd32(E1000_TDBAL(n));
320 		break;
321 	case E1000_TDBAH(0):
322 		for (n = 0; n < 4; n++)
323 			regs[n] = rd32(E1000_TDBAH(n));
324 		break;
325 	case E1000_TDLEN(0):
326 		for (n = 0; n < 4; n++)
327 			regs[n] = rd32(E1000_TDLEN(n));
328 		break;
329 	case E1000_TDH(0):
330 		for (n = 0; n < 4; n++)
331 			regs[n] = rd32(E1000_TDH(n));
332 		break;
333 	case E1000_TDT(0):
334 		for (n = 0; n < 4; n++)
335 			regs[n] = rd32(E1000_TDT(n));
336 		break;
337 	case E1000_TXDCTL(0):
338 		for (n = 0; n < 4; n++)
339 			regs[n] = rd32(E1000_TXDCTL(n));
340 		break;
341 	default:
342 		pr_info("%-15s %08x\n", reginfo->name, rd32(reginfo->ofs));
343 		return;
344 	}
345 
346 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-3]");
347 	pr_info("%-15s %08x %08x %08x %08x\n", rname, regs[0], regs[1],
348 		regs[2], regs[3]);
349 }
350 
351 /* igb_dump - Print registers, Tx-rings and Rx-rings */
352 static void igb_dump(struct igb_adapter *adapter)
353 {
354 	struct net_device *netdev = adapter->netdev;
355 	struct e1000_hw *hw = &adapter->hw;
356 	struct igb_reg_info *reginfo;
357 	struct igb_ring *tx_ring;
358 	union e1000_adv_tx_desc *tx_desc;
359 	struct my_u0 { __le64 a; __le64 b; } *u0;
360 	struct igb_ring *rx_ring;
361 	union e1000_adv_rx_desc *rx_desc;
362 	u32 staterr;
363 	u16 i, n;
364 
365 	if (!netif_msg_hw(adapter))
366 		return;
367 
368 	/* Print netdevice Info */
369 	if (netdev) {
370 		dev_info(&adapter->pdev->dev, "Net device Info\n");
371 		pr_info("Device Name     state            trans_start\n");
372 		pr_info("%-15s %016lX %016lX\n", netdev->name,
373 			netdev->state, dev_trans_start(netdev));
374 	}
375 
376 	/* Print Registers */
377 	dev_info(&adapter->pdev->dev, "Register Dump\n");
378 	pr_info(" Register Name   Value\n");
379 	for (reginfo = (struct igb_reg_info *)igb_reg_info_tbl;
380 	     reginfo->name; reginfo++) {
381 		igb_regdump(hw, reginfo);
382 	}
383 
384 	/* Print TX Ring Summary */
385 	if (!netdev || !netif_running(netdev))
386 		goto exit;
387 
388 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
389 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
390 	for (n = 0; n < adapter->num_tx_queues; n++) {
391 		struct igb_tx_buffer *buffer_info;
392 		tx_ring = adapter->tx_ring[n];
393 		buffer_info = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
394 		pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
395 			n, tx_ring->next_to_use, tx_ring->next_to_clean,
396 			(u64)dma_unmap_addr(buffer_info, dma),
397 			dma_unmap_len(buffer_info, len),
398 			buffer_info->next_to_watch,
399 			(u64)buffer_info->time_stamp);
400 	}
401 
402 	/* Print TX Rings */
403 	if (!netif_msg_tx_done(adapter))
404 		goto rx_ring_summary;
405 
406 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
407 
408 	/* Transmit Descriptor Formats
409 	 *
410 	 * Advanced Transmit Descriptor
411 	 *   +--------------------------------------------------------------+
412 	 * 0 |         Buffer Address [63:0]                                |
413 	 *   +--------------------------------------------------------------+
414 	 * 8 | PAYLEN  | PORTS  |CC|IDX | STA | DCMD  |DTYP|MAC|RSV| DTALEN |
415 	 *   +--------------------------------------------------------------+
416 	 *   63      46 45    40 39 38 36 35 32 31   24             15       0
417 	 */
418 
419 	for (n = 0; n < adapter->num_tx_queues; n++) {
420 		tx_ring = adapter->tx_ring[n];
421 		pr_info("------------------------------------\n");
422 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
423 		pr_info("------------------------------------\n");
424 		pr_info("T [desc]     [address 63:0  ] [PlPOCIStDDM Ln] [bi->dma       ] leng  ntw timestamp        bi->skb\n");
425 
426 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
427 			const char *next_desc;
428 			struct igb_tx_buffer *buffer_info;
429 			tx_desc = IGB_TX_DESC(tx_ring, i);
430 			buffer_info = &tx_ring->tx_buffer_info[i];
431 			u0 = (struct my_u0 *)tx_desc;
432 			if (i == tx_ring->next_to_use &&
433 			    i == tx_ring->next_to_clean)
434 				next_desc = " NTC/U";
435 			else if (i == tx_ring->next_to_use)
436 				next_desc = " NTU";
437 			else if (i == tx_ring->next_to_clean)
438 				next_desc = " NTC";
439 			else
440 				next_desc = "";
441 
442 			pr_info("T [0x%03X]    %016llX %016llX %016llX %04X  %p %016llX %p%s\n",
443 				i, le64_to_cpu(u0->a),
444 				le64_to_cpu(u0->b),
445 				(u64)dma_unmap_addr(buffer_info, dma),
446 				dma_unmap_len(buffer_info, len),
447 				buffer_info->next_to_watch,
448 				(u64)buffer_info->time_stamp,
449 				buffer_info->skb, next_desc);
450 
451 			if (netif_msg_pktdata(adapter) && buffer_info->skb)
452 				print_hex_dump(KERN_INFO, "",
453 					DUMP_PREFIX_ADDRESS,
454 					16, 1, buffer_info->skb->data,
455 					dma_unmap_len(buffer_info, len),
456 					true);
457 		}
458 	}
459 
460 	/* Print RX Rings Summary */
461 rx_ring_summary:
462 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
463 	pr_info("Queue [NTU] [NTC]\n");
464 	for (n = 0; n < adapter->num_rx_queues; n++) {
465 		rx_ring = adapter->rx_ring[n];
466 		pr_info(" %5d %5X %5X\n",
467 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
468 	}
469 
470 	/* Print RX Rings */
471 	if (!netif_msg_rx_status(adapter))
472 		goto exit;
473 
474 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
475 
476 	/* Advanced Receive Descriptor (Read) Format
477 	 *    63                                           1        0
478 	 *    +-----------------------------------------------------+
479 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
480 	 *    +----------------------------------------------+------+
481 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
482 	 *    +-----------------------------------------------------+
483 	 *
484 	 *
485 	 * Advanced Receive Descriptor (Write-Back) Format
486 	 *
487 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
488 	 *   +------------------------------------------------------+
489 	 * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
490 	 *   | Checksum   Ident  |   |           |    | Type | Type |
491 	 *   +------------------------------------------------------+
492 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
493 	 *   +------------------------------------------------------+
494 	 *   63       48 47    32 31            20 19               0
495 	 */
496 
497 	for (n = 0; n < adapter->num_rx_queues; n++) {
498 		rx_ring = adapter->rx_ring[n];
499 		pr_info("------------------------------------\n");
500 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
501 		pr_info("------------------------------------\n");
502 		pr_info("R  [desc]      [ PktBuf     A0] [  HeadBuf   DD] [bi->dma       ] [bi->skb] <-- Adv Rx Read format\n");
503 		pr_info("RWB[desc]      [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
504 
505 		for (i = 0; i < rx_ring->count; i++) {
506 			const char *next_desc;
507 			struct igb_rx_buffer *buffer_info;
508 			buffer_info = &rx_ring->rx_buffer_info[i];
509 			rx_desc = IGB_RX_DESC(rx_ring, i);
510 			u0 = (struct my_u0 *)rx_desc;
511 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
512 
513 			if (i == rx_ring->next_to_use)
514 				next_desc = " NTU";
515 			else if (i == rx_ring->next_to_clean)
516 				next_desc = " NTC";
517 			else
518 				next_desc = "";
519 
520 			if (staterr & E1000_RXD_STAT_DD) {
521 				/* Descriptor Done */
522 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %s\n",
523 					"RWB", i,
524 					le64_to_cpu(u0->a),
525 					le64_to_cpu(u0->b),
526 					next_desc);
527 			} else {
528 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %s\n",
529 					"R  ", i,
530 					le64_to_cpu(u0->a),
531 					le64_to_cpu(u0->b),
532 					(u64)buffer_info->dma,
533 					next_desc);
534 
535 				if (netif_msg_pktdata(adapter) &&
536 				    buffer_info->dma && buffer_info->page) {
537 					print_hex_dump(KERN_INFO, "",
538 					  DUMP_PREFIX_ADDRESS,
539 					  16, 1,
540 					  page_address(buffer_info->page) +
541 						      buffer_info->page_offset,
542 					  igb_rx_bufsz(rx_ring), true);
543 				}
544 			}
545 		}
546 	}
547 
548 exit:
549 	return;
550 }
551 
552 /**
553  *  igb_get_i2c_data - Reads the I2C SDA data bit
554  *  @data: opaque pointer to adapter struct
555  *
556  *  Returns the I2C data bit value
557  **/
558 static int igb_get_i2c_data(void *data)
559 {
560 	struct igb_adapter *adapter = (struct igb_adapter *)data;
561 	struct e1000_hw *hw = &adapter->hw;
562 	s32 i2cctl = rd32(E1000_I2CPARAMS);
563 
564 	return !!(i2cctl & E1000_I2C_DATA_IN);
565 }
566 
567 /**
568  *  igb_set_i2c_data - Sets the I2C data bit
569  *  @data: pointer to hardware structure
570  *  @state: I2C data value (0 or 1) to set
571  *
572  *  Sets the I2C data bit
573  **/
574 static void igb_set_i2c_data(void *data, int state)
575 {
576 	struct igb_adapter *adapter = (struct igb_adapter *)data;
577 	struct e1000_hw *hw = &adapter->hw;
578 	s32 i2cctl = rd32(E1000_I2CPARAMS);
579 
580 	if (state) {
581 		i2cctl |= E1000_I2C_DATA_OUT | E1000_I2C_DATA_OE_N;
582 	} else {
583 		i2cctl &= ~E1000_I2C_DATA_OE_N;
584 		i2cctl &= ~E1000_I2C_DATA_OUT;
585 	}
586 
587 	wr32(E1000_I2CPARAMS, i2cctl);
588 	wrfl();
589 }
590 
591 /**
592  *  igb_set_i2c_clk - Sets the I2C SCL clock
593  *  @data: pointer to hardware structure
594  *  @state: state to set clock
595  *
596  *  Sets the I2C clock line to state
597  **/
598 static void igb_set_i2c_clk(void *data, int state)
599 {
600 	struct igb_adapter *adapter = (struct igb_adapter *)data;
601 	struct e1000_hw *hw = &adapter->hw;
602 	s32 i2cctl = rd32(E1000_I2CPARAMS);
603 
604 	if (state) {
605 		i2cctl |= E1000_I2C_CLK_OUT | E1000_I2C_CLK_OE_N;
606 	} else {
607 		i2cctl &= ~E1000_I2C_CLK_OUT;
608 		i2cctl &= ~E1000_I2C_CLK_OE_N;
609 	}
610 	wr32(E1000_I2CPARAMS, i2cctl);
611 	wrfl();
612 }
613 
614 /**
615  *  igb_get_i2c_clk - Gets the I2C SCL clock state
616  *  @data: pointer to hardware structure
617  *
618  *  Gets the I2C clock state
619  **/
620 static int igb_get_i2c_clk(void *data)
621 {
622 	struct igb_adapter *adapter = (struct igb_adapter *)data;
623 	struct e1000_hw *hw = &adapter->hw;
624 	s32 i2cctl = rd32(E1000_I2CPARAMS);
625 
626 	return !!(i2cctl & E1000_I2C_CLK_IN);
627 }
628 
629 static const struct i2c_algo_bit_data igb_i2c_algo = {
630 	.setsda		= igb_set_i2c_data,
631 	.setscl		= igb_set_i2c_clk,
632 	.getsda		= igb_get_i2c_data,
633 	.getscl		= igb_get_i2c_clk,
634 	.udelay		= 5,
635 	.timeout	= 20,
636 };
637 
638 /**
639  *  igb_get_hw_dev - return device
640  *  @hw: pointer to hardware structure
641  *
642  *  used by hardware layer to print debugging information
643  **/
644 struct net_device *igb_get_hw_dev(struct e1000_hw *hw)
645 {
646 	struct igb_adapter *adapter = hw->back;
647 	return adapter->netdev;
648 }
649 
650 /**
651  *  igb_init_module - Driver Registration Routine
652  *
653  *  igb_init_module is the first routine called when the driver is
654  *  loaded. All it does is register with the PCI subsystem.
655  **/
656 static int __init igb_init_module(void)
657 {
658 	int ret;
659 
660 	pr_info("%s\n", igb_driver_string);
661 	pr_info("%s\n", igb_copyright);
662 
663 #ifdef CONFIG_IGB_DCA
664 	dca_register_notify(&dca_notifier);
665 #endif
666 	ret = pci_register_driver(&igb_driver);
667 	return ret;
668 }
669 
670 module_init(igb_init_module);
671 
672 /**
673  *  igb_exit_module - Driver Exit Cleanup Routine
674  *
675  *  igb_exit_module is called just before the driver is removed
676  *  from memory.
677  **/
678 static void __exit igb_exit_module(void)
679 {
680 #ifdef CONFIG_IGB_DCA
681 	dca_unregister_notify(&dca_notifier);
682 #endif
683 	pci_unregister_driver(&igb_driver);
684 }
685 
686 module_exit(igb_exit_module);
687 
688 #define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
689 /**
690  *  igb_cache_ring_register - Descriptor ring to register mapping
691  *  @adapter: board private structure to initialize
692  *
693  *  Once we know the feature-set enabled for the device, we'll cache
694  *  the register offset the descriptor ring is assigned to.
695  **/
696 static void igb_cache_ring_register(struct igb_adapter *adapter)
697 {
698 	int i = 0, j = 0;
699 	u32 rbase_offset = adapter->vfs_allocated_count;
700 
701 	switch (adapter->hw.mac.type) {
702 	case e1000_82576:
703 		/* The queues are allocated for virtualization such that VF 0
704 		 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
705 		 * In order to avoid collision we start at the first free queue
706 		 * and continue consuming queues in the same sequence
707 		 */
708 		if (adapter->vfs_allocated_count) {
709 			for (; i < adapter->rss_queues; i++)
710 				adapter->rx_ring[i]->reg_idx = rbase_offset +
711 							       Q_IDX_82576(i);
712 		}
713 		fallthrough;
714 	case e1000_82575:
715 	case e1000_82580:
716 	case e1000_i350:
717 	case e1000_i354:
718 	case e1000_i210:
719 	case e1000_i211:
720 	default:
721 		for (; i < adapter->num_rx_queues; i++)
722 			adapter->rx_ring[i]->reg_idx = rbase_offset + i;
723 		for (; j < adapter->num_tx_queues; j++)
724 			adapter->tx_ring[j]->reg_idx = rbase_offset + j;
725 		break;
726 	}
727 }
728 
729 u32 igb_rd32(struct e1000_hw *hw, u32 reg)
730 {
731 	struct igb_adapter *igb = container_of(hw, struct igb_adapter, hw);
732 	u8 __iomem *hw_addr = READ_ONCE(hw->hw_addr);
733 	u32 value = 0;
734 
735 	if (E1000_REMOVED(hw_addr))
736 		return ~value;
737 
738 	value = readl(&hw_addr[reg]);
739 
740 	/* reads should not return all F's */
741 	if (!(~value) && (!reg || !(~readl(hw_addr)))) {
742 		struct net_device *netdev = igb->netdev;
743 		hw->hw_addr = NULL;
744 		netdev_err(netdev, "PCIe link lost\n");
745 		WARN(pci_device_is_present(igb->pdev),
746 		     "igb: Failed to read reg 0x%x!\n", reg);
747 	}
748 
749 	return value;
750 }
751 
752 /**
753  *  igb_write_ivar - configure ivar for given MSI-X vector
754  *  @hw: pointer to the HW structure
755  *  @msix_vector: vector number we are allocating to a given ring
756  *  @index: row index of IVAR register to write within IVAR table
757  *  @offset: column offset of in IVAR, should be multiple of 8
758  *
759  *  This function is intended to handle the writing of the IVAR register
760  *  for adapters 82576 and newer.  The IVAR table consists of 2 columns,
761  *  each containing an cause allocation for an Rx and Tx ring, and a
762  *  variable number of rows depending on the number of queues supported.
763  **/
764 static void igb_write_ivar(struct e1000_hw *hw, int msix_vector,
765 			   int index, int offset)
766 {
767 	u32 ivar = array_rd32(E1000_IVAR0, index);
768 
769 	/* clear any bits that are currently set */
770 	ivar &= ~((u32)0xFF << offset);
771 
772 	/* write vector and valid bit */
773 	ivar |= (msix_vector | E1000_IVAR_VALID) << offset;
774 
775 	array_wr32(E1000_IVAR0, index, ivar);
776 }
777 
778 #define IGB_N0_QUEUE -1
779 static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
780 {
781 	struct igb_adapter *adapter = q_vector->adapter;
782 	struct e1000_hw *hw = &adapter->hw;
783 	int rx_queue = IGB_N0_QUEUE;
784 	int tx_queue = IGB_N0_QUEUE;
785 	u32 msixbm = 0;
786 
787 	if (q_vector->rx.ring)
788 		rx_queue = q_vector->rx.ring->reg_idx;
789 	if (q_vector->tx.ring)
790 		tx_queue = q_vector->tx.ring->reg_idx;
791 
792 	switch (hw->mac.type) {
793 	case e1000_82575:
794 		/* The 82575 assigns vectors using a bitmask, which matches the
795 		 * bitmask for the EICR/EIMS/EIMC registers.  To assign one
796 		 * or more queues to a vector, we write the appropriate bits
797 		 * into the MSIXBM register for that vector.
798 		 */
799 		if (rx_queue > IGB_N0_QUEUE)
800 			msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
801 		if (tx_queue > IGB_N0_QUEUE)
802 			msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
803 		if (!(adapter->flags & IGB_FLAG_HAS_MSIX) && msix_vector == 0)
804 			msixbm |= E1000_EIMS_OTHER;
805 		array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
806 		q_vector->eims_value = msixbm;
807 		break;
808 	case e1000_82576:
809 		/* 82576 uses a table that essentially consists of 2 columns
810 		 * with 8 rows.  The ordering is column-major so we use the
811 		 * lower 3 bits as the row index, and the 4th bit as the
812 		 * column offset.
813 		 */
814 		if (rx_queue > IGB_N0_QUEUE)
815 			igb_write_ivar(hw, msix_vector,
816 				       rx_queue & 0x7,
817 				       (rx_queue & 0x8) << 1);
818 		if (tx_queue > IGB_N0_QUEUE)
819 			igb_write_ivar(hw, msix_vector,
820 				       tx_queue & 0x7,
821 				       ((tx_queue & 0x8) << 1) + 8);
822 		q_vector->eims_value = BIT(msix_vector);
823 		break;
824 	case e1000_82580:
825 	case e1000_i350:
826 	case e1000_i354:
827 	case e1000_i210:
828 	case e1000_i211:
829 		/* On 82580 and newer adapters the scheme is similar to 82576
830 		 * however instead of ordering column-major we have things
831 		 * ordered row-major.  So we traverse the table by using
832 		 * bit 0 as the column offset, and the remaining bits as the
833 		 * row index.
834 		 */
835 		if (rx_queue > IGB_N0_QUEUE)
836 			igb_write_ivar(hw, msix_vector,
837 				       rx_queue >> 1,
838 				       (rx_queue & 0x1) << 4);
839 		if (tx_queue > IGB_N0_QUEUE)
840 			igb_write_ivar(hw, msix_vector,
841 				       tx_queue >> 1,
842 				       ((tx_queue & 0x1) << 4) + 8);
843 		q_vector->eims_value = BIT(msix_vector);
844 		break;
845 	default:
846 		BUG();
847 		break;
848 	}
849 
850 	/* add q_vector eims value to global eims_enable_mask */
851 	adapter->eims_enable_mask |= q_vector->eims_value;
852 
853 	/* configure q_vector to set itr on first interrupt */
854 	q_vector->set_itr = 1;
855 }
856 
857 /**
858  *  igb_configure_msix - Configure MSI-X hardware
859  *  @adapter: board private structure to initialize
860  *
861  *  igb_configure_msix sets up the hardware to properly
862  *  generate MSI-X interrupts.
863  **/
864 static void igb_configure_msix(struct igb_adapter *adapter)
865 {
866 	u32 tmp;
867 	int i, vector = 0;
868 	struct e1000_hw *hw = &adapter->hw;
869 
870 	adapter->eims_enable_mask = 0;
871 
872 	/* set vector for other causes, i.e. link changes */
873 	switch (hw->mac.type) {
874 	case e1000_82575:
875 		tmp = rd32(E1000_CTRL_EXT);
876 		/* enable MSI-X PBA support*/
877 		tmp |= E1000_CTRL_EXT_PBA_CLR;
878 
879 		/* Auto-Mask interrupts upon ICR read. */
880 		tmp |= E1000_CTRL_EXT_EIAME;
881 		tmp |= E1000_CTRL_EXT_IRCA;
882 
883 		wr32(E1000_CTRL_EXT, tmp);
884 
885 		/* enable msix_other interrupt */
886 		array_wr32(E1000_MSIXBM(0), vector++, E1000_EIMS_OTHER);
887 		adapter->eims_other = E1000_EIMS_OTHER;
888 
889 		break;
890 
891 	case e1000_82576:
892 	case e1000_82580:
893 	case e1000_i350:
894 	case e1000_i354:
895 	case e1000_i210:
896 	case e1000_i211:
897 		/* Turn on MSI-X capability first, or our settings
898 		 * won't stick.  And it will take days to debug.
899 		 */
900 		wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
901 		     E1000_GPIE_PBA | E1000_GPIE_EIAME |
902 		     E1000_GPIE_NSICR);
903 
904 		/* enable msix_other interrupt */
905 		adapter->eims_other = BIT(vector);
906 		tmp = (vector++ | E1000_IVAR_VALID) << 8;
907 
908 		wr32(E1000_IVAR_MISC, tmp);
909 		break;
910 	default:
911 		/* do nothing, since nothing else supports MSI-X */
912 		break;
913 	} /* switch (hw->mac.type) */
914 
915 	adapter->eims_enable_mask |= adapter->eims_other;
916 
917 	for (i = 0; i < adapter->num_q_vectors; i++)
918 		igb_assign_vector(adapter->q_vector[i], vector++);
919 
920 	wrfl();
921 }
922 
923 /**
924  *  igb_request_msix - Initialize MSI-X interrupts
925  *  @adapter: board private structure to initialize
926  *
927  *  igb_request_msix allocates MSI-X vectors and requests interrupts from the
928  *  kernel.
929  **/
930 static int igb_request_msix(struct igb_adapter *adapter)
931 {
932 	unsigned int num_q_vectors = adapter->num_q_vectors;
933 	struct net_device *netdev = adapter->netdev;
934 	int i, err = 0, vector = 0, free_vector = 0;
935 
936 	err = request_irq(adapter->msix_entries[vector].vector,
937 			  igb_msix_other, 0, netdev->name, adapter);
938 	if (err)
939 		goto err_out;
940 
941 	if (num_q_vectors > MAX_Q_VECTORS) {
942 		num_q_vectors = MAX_Q_VECTORS;
943 		dev_warn(&adapter->pdev->dev,
944 			 "The number of queue vectors (%d) is higher than max allowed (%d)\n",
945 			 adapter->num_q_vectors, MAX_Q_VECTORS);
946 	}
947 	for (i = 0; i < num_q_vectors; i++) {
948 		struct igb_q_vector *q_vector = adapter->q_vector[i];
949 
950 		vector++;
951 
952 		q_vector->itr_register = adapter->io_addr + E1000_EITR(vector);
953 
954 		if (q_vector->rx.ring && q_vector->tx.ring)
955 			sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
956 				q_vector->rx.ring->queue_index);
957 		else if (q_vector->tx.ring)
958 			sprintf(q_vector->name, "%s-tx-%u", netdev->name,
959 				q_vector->tx.ring->queue_index);
960 		else if (q_vector->rx.ring)
961 			sprintf(q_vector->name, "%s-rx-%u", netdev->name,
962 				q_vector->rx.ring->queue_index);
963 		else
964 			sprintf(q_vector->name, "%s-unused", netdev->name);
965 
966 		err = request_irq(adapter->msix_entries[vector].vector,
967 				  igb_msix_ring, 0, q_vector->name,
968 				  q_vector);
969 		if (err)
970 			goto err_free;
971 	}
972 
973 	igb_configure_msix(adapter);
974 	return 0;
975 
976 err_free:
977 	/* free already assigned IRQs */
978 	free_irq(adapter->msix_entries[free_vector++].vector, adapter);
979 
980 	vector--;
981 	for (i = 0; i < vector; i++) {
982 		free_irq(adapter->msix_entries[free_vector++].vector,
983 			 adapter->q_vector[i]);
984 	}
985 err_out:
986 	return err;
987 }
988 
989 /**
990  *  igb_free_q_vector - Free memory allocated for specific interrupt vector
991  *  @adapter: board private structure to initialize
992  *  @v_idx: Index of vector to be freed
993  *
994  *  This function frees the memory allocated to the q_vector.
995  **/
996 static void igb_free_q_vector(struct igb_adapter *adapter, int v_idx)
997 {
998 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
999 
1000 	adapter->q_vector[v_idx] = NULL;
1001 
1002 	/* igb_get_stats64() might access the rings on this vector,
1003 	 * we must wait a grace period before freeing it.
1004 	 */
1005 	if (q_vector)
1006 		kfree_rcu(q_vector, rcu);
1007 }
1008 
1009 /**
1010  *  igb_reset_q_vector - Reset config for interrupt vector
1011  *  @adapter: board private structure to initialize
1012  *  @v_idx: Index of vector to be reset
1013  *
1014  *  If NAPI is enabled it will delete any references to the
1015  *  NAPI struct. This is preparation for igb_free_q_vector.
1016  **/
1017 static void igb_reset_q_vector(struct igb_adapter *adapter, int v_idx)
1018 {
1019 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1020 
1021 	/* Coming from igb_set_interrupt_capability, the vectors are not yet
1022 	 * allocated. So, q_vector is NULL so we should stop here.
1023 	 */
1024 	if (!q_vector)
1025 		return;
1026 
1027 	if (q_vector->tx.ring)
1028 		adapter->tx_ring[q_vector->tx.ring->queue_index] = NULL;
1029 
1030 	if (q_vector->rx.ring)
1031 		adapter->rx_ring[q_vector->rx.ring->queue_index] = NULL;
1032 
1033 	netif_napi_del(&q_vector->napi);
1034 
1035 }
1036 
1037 static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
1038 {
1039 	int v_idx = adapter->num_q_vectors;
1040 
1041 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
1042 		pci_disable_msix(adapter->pdev);
1043 	else if (adapter->flags & IGB_FLAG_HAS_MSI)
1044 		pci_disable_msi(adapter->pdev);
1045 
1046 	while (v_idx--)
1047 		igb_reset_q_vector(adapter, v_idx);
1048 }
1049 
1050 /**
1051  *  igb_free_q_vectors - Free memory allocated for interrupt vectors
1052  *  @adapter: board private structure to initialize
1053  *
1054  *  This function frees the memory allocated to the q_vectors.  In addition if
1055  *  NAPI is enabled it will delete any references to the NAPI struct prior
1056  *  to freeing the q_vector.
1057  **/
1058 static void igb_free_q_vectors(struct igb_adapter *adapter)
1059 {
1060 	int v_idx = adapter->num_q_vectors;
1061 
1062 	adapter->num_tx_queues = 0;
1063 	adapter->num_rx_queues = 0;
1064 	adapter->num_q_vectors = 0;
1065 
1066 	while (v_idx--) {
1067 		igb_reset_q_vector(adapter, v_idx);
1068 		igb_free_q_vector(adapter, v_idx);
1069 	}
1070 }
1071 
1072 /**
1073  *  igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1074  *  @adapter: board private structure to initialize
1075  *
1076  *  This function resets the device so that it has 0 Rx queues, Tx queues, and
1077  *  MSI-X interrupts allocated.
1078  */
1079 static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
1080 {
1081 	igb_free_q_vectors(adapter);
1082 	igb_reset_interrupt_capability(adapter);
1083 }
1084 
1085 /**
1086  *  igb_set_interrupt_capability - set MSI or MSI-X if supported
1087  *  @adapter: board private structure to initialize
1088  *  @msix: boolean value of MSIX capability
1089  *
1090  *  Attempt to configure interrupts using the best available
1091  *  capabilities of the hardware and kernel.
1092  **/
1093 static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
1094 {
1095 	int err;
1096 	int numvecs, i;
1097 
1098 	if (!msix)
1099 		goto msi_only;
1100 	adapter->flags |= IGB_FLAG_HAS_MSIX;
1101 
1102 	/* Number of supported queues. */
1103 	adapter->num_rx_queues = adapter->rss_queues;
1104 	if (adapter->vfs_allocated_count)
1105 		adapter->num_tx_queues = 1;
1106 	else
1107 		adapter->num_tx_queues = adapter->rss_queues;
1108 
1109 	/* start with one vector for every Rx queue */
1110 	numvecs = adapter->num_rx_queues;
1111 
1112 	/* if Tx handler is separate add 1 for every Tx queue */
1113 	if (!(adapter->flags & IGB_FLAG_QUEUE_PAIRS))
1114 		numvecs += adapter->num_tx_queues;
1115 
1116 	/* store the number of vectors reserved for queues */
1117 	adapter->num_q_vectors = numvecs;
1118 
1119 	/* add 1 vector for link status interrupts */
1120 	numvecs++;
1121 	for (i = 0; i < numvecs; i++)
1122 		adapter->msix_entries[i].entry = i;
1123 
1124 	err = pci_enable_msix_range(adapter->pdev,
1125 				    adapter->msix_entries,
1126 				    numvecs,
1127 				    numvecs);
1128 	if (err > 0)
1129 		return;
1130 
1131 	igb_reset_interrupt_capability(adapter);
1132 
1133 	/* If we can't do MSI-X, try MSI */
1134 msi_only:
1135 	adapter->flags &= ~IGB_FLAG_HAS_MSIX;
1136 #ifdef CONFIG_PCI_IOV
1137 	/* disable SR-IOV for non MSI-X configurations */
1138 	if (adapter->vf_data) {
1139 		struct e1000_hw *hw = &adapter->hw;
1140 		/* disable iov and allow time for transactions to clear */
1141 		pci_disable_sriov(adapter->pdev);
1142 		msleep(500);
1143 
1144 		kfree(adapter->vf_mac_list);
1145 		adapter->vf_mac_list = NULL;
1146 		kfree(adapter->vf_data);
1147 		adapter->vf_data = NULL;
1148 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1149 		wrfl();
1150 		msleep(100);
1151 		dev_info(&adapter->pdev->dev, "IOV Disabled\n");
1152 	}
1153 #endif
1154 	adapter->vfs_allocated_count = 0;
1155 	adapter->rss_queues = 1;
1156 	adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
1157 	adapter->num_rx_queues = 1;
1158 	adapter->num_tx_queues = 1;
1159 	adapter->num_q_vectors = 1;
1160 	if (!pci_enable_msi(adapter->pdev))
1161 		adapter->flags |= IGB_FLAG_HAS_MSI;
1162 }
1163 
1164 static void igb_add_ring(struct igb_ring *ring,
1165 			 struct igb_ring_container *head)
1166 {
1167 	head->ring = ring;
1168 	head->count++;
1169 }
1170 
1171 /**
1172  *  igb_alloc_q_vector - Allocate memory for a single interrupt vector
1173  *  @adapter: board private structure to initialize
1174  *  @v_count: q_vectors allocated on adapter, used for ring interleaving
1175  *  @v_idx: index of vector in adapter struct
1176  *  @txr_count: total number of Tx rings to allocate
1177  *  @txr_idx: index of first Tx ring to allocate
1178  *  @rxr_count: total number of Rx rings to allocate
1179  *  @rxr_idx: index of first Rx ring to allocate
1180  *
1181  *  We allocate one q_vector.  If allocation fails we return -ENOMEM.
1182  **/
1183 static int igb_alloc_q_vector(struct igb_adapter *adapter,
1184 			      int v_count, int v_idx,
1185 			      int txr_count, int txr_idx,
1186 			      int rxr_count, int rxr_idx)
1187 {
1188 	struct igb_q_vector *q_vector;
1189 	struct igb_ring *ring;
1190 	int ring_count;
1191 	size_t size;
1192 
1193 	/* igb only supports 1 Tx and/or 1 Rx queue per vector */
1194 	if (txr_count > 1 || rxr_count > 1)
1195 		return -ENOMEM;
1196 
1197 	ring_count = txr_count + rxr_count;
1198 	size = kmalloc_size_roundup(struct_size(q_vector, ring, ring_count));
1199 
1200 	/* allocate q_vector and rings */
1201 	q_vector = adapter->q_vector[v_idx];
1202 	if (!q_vector) {
1203 		q_vector = kzalloc(size, GFP_KERNEL);
1204 	} else if (size > ksize(q_vector)) {
1205 		struct igb_q_vector *new_q_vector;
1206 
1207 		new_q_vector = kzalloc(size, GFP_KERNEL);
1208 		if (new_q_vector)
1209 			kfree_rcu(q_vector, rcu);
1210 		q_vector = new_q_vector;
1211 	} else {
1212 		memset(q_vector, 0, size);
1213 	}
1214 	if (!q_vector)
1215 		return -ENOMEM;
1216 
1217 	/* initialize NAPI */
1218 	netif_napi_add(adapter->netdev, &q_vector->napi, igb_poll);
1219 
1220 	/* tie q_vector and adapter together */
1221 	adapter->q_vector[v_idx] = q_vector;
1222 	q_vector->adapter = adapter;
1223 
1224 	/* initialize work limits */
1225 	q_vector->tx.work_limit = adapter->tx_work_limit;
1226 
1227 	/* initialize ITR configuration */
1228 	q_vector->itr_register = adapter->io_addr + E1000_EITR(0);
1229 	q_vector->itr_val = IGB_START_ITR;
1230 
1231 	/* initialize pointer to rings */
1232 	ring = q_vector->ring;
1233 
1234 	/* intialize ITR */
1235 	if (rxr_count) {
1236 		/* rx or rx/tx vector */
1237 		if (!adapter->rx_itr_setting || adapter->rx_itr_setting > 3)
1238 			q_vector->itr_val = adapter->rx_itr_setting;
1239 	} else {
1240 		/* tx only vector */
1241 		if (!adapter->tx_itr_setting || adapter->tx_itr_setting > 3)
1242 			q_vector->itr_val = adapter->tx_itr_setting;
1243 	}
1244 
1245 	if (txr_count) {
1246 		/* assign generic ring traits */
1247 		ring->dev = &adapter->pdev->dev;
1248 		ring->netdev = adapter->netdev;
1249 
1250 		/* configure backlink on ring */
1251 		ring->q_vector = q_vector;
1252 
1253 		/* update q_vector Tx values */
1254 		igb_add_ring(ring, &q_vector->tx);
1255 
1256 		/* For 82575, context index must be unique per ring. */
1257 		if (adapter->hw.mac.type == e1000_82575)
1258 			set_bit(IGB_RING_FLAG_TX_CTX_IDX, &ring->flags);
1259 
1260 		/* apply Tx specific ring traits */
1261 		ring->count = adapter->tx_ring_count;
1262 		ring->queue_index = txr_idx;
1263 
1264 		ring->cbs_enable = false;
1265 		ring->idleslope = 0;
1266 		ring->sendslope = 0;
1267 		ring->hicredit = 0;
1268 		ring->locredit = 0;
1269 
1270 		u64_stats_init(&ring->tx_syncp);
1271 		u64_stats_init(&ring->tx_syncp2);
1272 
1273 		/* assign ring to adapter */
1274 		adapter->tx_ring[txr_idx] = ring;
1275 
1276 		/* push pointer to next ring */
1277 		ring++;
1278 	}
1279 
1280 	if (rxr_count) {
1281 		/* assign generic ring traits */
1282 		ring->dev = &adapter->pdev->dev;
1283 		ring->netdev = adapter->netdev;
1284 
1285 		/* configure backlink on ring */
1286 		ring->q_vector = q_vector;
1287 
1288 		/* update q_vector Rx values */
1289 		igb_add_ring(ring, &q_vector->rx);
1290 
1291 		/* set flag indicating ring supports SCTP checksum offload */
1292 		if (adapter->hw.mac.type >= e1000_82576)
1293 			set_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags);
1294 
1295 		/* On i350, i354, i210, and i211, loopback VLAN packets
1296 		 * have the tag byte-swapped.
1297 		 */
1298 		if (adapter->hw.mac.type >= e1000_i350)
1299 			set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &ring->flags);
1300 
1301 		/* apply Rx specific ring traits */
1302 		ring->count = adapter->rx_ring_count;
1303 		ring->queue_index = rxr_idx;
1304 
1305 		u64_stats_init(&ring->rx_syncp);
1306 
1307 		/* assign ring to adapter */
1308 		adapter->rx_ring[rxr_idx] = ring;
1309 	}
1310 
1311 	return 0;
1312 }
1313 
1314 
1315 /**
1316  *  igb_alloc_q_vectors - Allocate memory for interrupt vectors
1317  *  @adapter: board private structure to initialize
1318  *
1319  *  We allocate one q_vector per queue interrupt.  If allocation fails we
1320  *  return -ENOMEM.
1321  **/
1322 static int igb_alloc_q_vectors(struct igb_adapter *adapter)
1323 {
1324 	int q_vectors = adapter->num_q_vectors;
1325 	int rxr_remaining = adapter->num_rx_queues;
1326 	int txr_remaining = adapter->num_tx_queues;
1327 	int rxr_idx = 0, txr_idx = 0, v_idx = 0;
1328 	int err;
1329 
1330 	if (q_vectors >= (rxr_remaining + txr_remaining)) {
1331 		for (; rxr_remaining; v_idx++) {
1332 			err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1333 						 0, 0, 1, rxr_idx);
1334 
1335 			if (err)
1336 				goto err_out;
1337 
1338 			/* update counts and index */
1339 			rxr_remaining--;
1340 			rxr_idx++;
1341 		}
1342 	}
1343 
1344 	for (; v_idx < q_vectors; v_idx++) {
1345 		int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
1346 		int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);
1347 
1348 		err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1349 					 tqpv, txr_idx, rqpv, rxr_idx);
1350 
1351 		if (err)
1352 			goto err_out;
1353 
1354 		/* update counts and index */
1355 		rxr_remaining -= rqpv;
1356 		txr_remaining -= tqpv;
1357 		rxr_idx++;
1358 		txr_idx++;
1359 	}
1360 
1361 	return 0;
1362 
1363 err_out:
1364 	adapter->num_tx_queues = 0;
1365 	adapter->num_rx_queues = 0;
1366 	adapter->num_q_vectors = 0;
1367 
1368 	while (v_idx--)
1369 		igb_free_q_vector(adapter, v_idx);
1370 
1371 	return -ENOMEM;
1372 }
1373 
1374 /**
1375  *  igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1376  *  @adapter: board private structure to initialize
1377  *  @msix: boolean value of MSIX capability
1378  *
1379  *  This function initializes the interrupts and allocates all of the queues.
1380  **/
1381 static int igb_init_interrupt_scheme(struct igb_adapter *adapter, bool msix)
1382 {
1383 	struct pci_dev *pdev = adapter->pdev;
1384 	int err;
1385 
1386 	igb_set_interrupt_capability(adapter, msix);
1387 
1388 	err = igb_alloc_q_vectors(adapter);
1389 	if (err) {
1390 		dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
1391 		goto err_alloc_q_vectors;
1392 	}
1393 
1394 	igb_cache_ring_register(adapter);
1395 
1396 	return 0;
1397 
1398 err_alloc_q_vectors:
1399 	igb_reset_interrupt_capability(adapter);
1400 	return err;
1401 }
1402 
1403 /**
1404  *  igb_request_irq - initialize interrupts
1405  *  @adapter: board private structure to initialize
1406  *
1407  *  Attempts to configure interrupts using the best available
1408  *  capabilities of the hardware and kernel.
1409  **/
1410 static int igb_request_irq(struct igb_adapter *adapter)
1411 {
1412 	struct net_device *netdev = adapter->netdev;
1413 	struct pci_dev *pdev = adapter->pdev;
1414 	int err = 0;
1415 
1416 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1417 		err = igb_request_msix(adapter);
1418 		if (!err)
1419 			goto request_done;
1420 		/* fall back to MSI */
1421 		igb_free_all_tx_resources(adapter);
1422 		igb_free_all_rx_resources(adapter);
1423 
1424 		igb_clear_interrupt_scheme(adapter);
1425 		err = igb_init_interrupt_scheme(adapter, false);
1426 		if (err)
1427 			goto request_done;
1428 
1429 		igb_setup_all_tx_resources(adapter);
1430 		igb_setup_all_rx_resources(adapter);
1431 		igb_configure(adapter);
1432 	}
1433 
1434 	igb_assign_vector(adapter->q_vector[0], 0);
1435 
1436 	if (adapter->flags & IGB_FLAG_HAS_MSI) {
1437 		err = request_irq(pdev->irq, igb_intr_msi, 0,
1438 				  netdev->name, adapter);
1439 		if (!err)
1440 			goto request_done;
1441 
1442 		/* fall back to legacy interrupts */
1443 		igb_reset_interrupt_capability(adapter);
1444 		adapter->flags &= ~IGB_FLAG_HAS_MSI;
1445 	}
1446 
1447 	err = request_irq(pdev->irq, igb_intr, IRQF_SHARED,
1448 			  netdev->name, adapter);
1449 
1450 	if (err)
1451 		dev_err(&pdev->dev, "Error %d getting interrupt\n",
1452 			err);
1453 
1454 request_done:
1455 	return err;
1456 }
1457 
1458 static void igb_free_irq(struct igb_adapter *adapter)
1459 {
1460 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1461 		int vector = 0, i;
1462 
1463 		free_irq(adapter->msix_entries[vector++].vector, adapter);
1464 
1465 		for (i = 0; i < adapter->num_q_vectors; i++)
1466 			free_irq(adapter->msix_entries[vector++].vector,
1467 				 adapter->q_vector[i]);
1468 	} else {
1469 		free_irq(adapter->pdev->irq, adapter);
1470 	}
1471 }
1472 
1473 /**
1474  *  igb_irq_disable - Mask off interrupt generation on the NIC
1475  *  @adapter: board private structure
1476  **/
1477 static void igb_irq_disable(struct igb_adapter *adapter)
1478 {
1479 	struct e1000_hw *hw = &adapter->hw;
1480 
1481 	/* we need to be careful when disabling interrupts.  The VFs are also
1482 	 * mapped into these registers and so clearing the bits can cause
1483 	 * issues on the VF drivers so we only need to clear what we set
1484 	 */
1485 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1486 		u32 regval = rd32(E1000_EIAM);
1487 
1488 		wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
1489 		wr32(E1000_EIMC, adapter->eims_enable_mask);
1490 		regval = rd32(E1000_EIAC);
1491 		wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
1492 	}
1493 
1494 	wr32(E1000_IAM, 0);
1495 	wr32(E1000_IMC, ~0);
1496 	wrfl();
1497 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1498 		int i;
1499 
1500 		for (i = 0; i < adapter->num_q_vectors; i++)
1501 			synchronize_irq(adapter->msix_entries[i].vector);
1502 	} else {
1503 		synchronize_irq(adapter->pdev->irq);
1504 	}
1505 }
1506 
1507 /**
1508  *  igb_irq_enable - Enable default interrupt generation settings
1509  *  @adapter: board private structure
1510  **/
1511 static void igb_irq_enable(struct igb_adapter *adapter)
1512 {
1513 	struct e1000_hw *hw = &adapter->hw;
1514 
1515 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1516 		u32 ims = E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_DRSTA;
1517 		u32 regval = rd32(E1000_EIAC);
1518 
1519 		wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
1520 		regval = rd32(E1000_EIAM);
1521 		wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
1522 		wr32(E1000_EIMS, adapter->eims_enable_mask);
1523 		if (adapter->vfs_allocated_count) {
1524 			wr32(E1000_MBVFIMR, 0xFF);
1525 			ims |= E1000_IMS_VMMB;
1526 		}
1527 		wr32(E1000_IMS, ims);
1528 	} else {
1529 		wr32(E1000_IMS, IMS_ENABLE_MASK |
1530 				E1000_IMS_DRSTA);
1531 		wr32(E1000_IAM, IMS_ENABLE_MASK |
1532 				E1000_IMS_DRSTA);
1533 	}
1534 }
1535 
1536 static void igb_update_mng_vlan(struct igb_adapter *adapter)
1537 {
1538 	struct e1000_hw *hw = &adapter->hw;
1539 	u16 pf_id = adapter->vfs_allocated_count;
1540 	u16 vid = adapter->hw.mng_cookie.vlan_id;
1541 	u16 old_vid = adapter->mng_vlan_id;
1542 
1543 	if (hw->mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1544 		/* add VID to filter table */
1545 		igb_vfta_set(hw, vid, pf_id, true, true);
1546 		adapter->mng_vlan_id = vid;
1547 	} else {
1548 		adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1549 	}
1550 
1551 	if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1552 	    (vid != old_vid) &&
1553 	    !test_bit(old_vid, adapter->active_vlans)) {
1554 		/* remove VID from filter table */
1555 		igb_vfta_set(hw, vid, pf_id, false, true);
1556 	}
1557 }
1558 
1559 /**
1560  *  igb_release_hw_control - release control of the h/w to f/w
1561  *  @adapter: address of board private structure
1562  *
1563  *  igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1564  *  For ASF and Pass Through versions of f/w this means that the
1565  *  driver is no longer loaded.
1566  **/
1567 static void igb_release_hw_control(struct igb_adapter *adapter)
1568 {
1569 	struct e1000_hw *hw = &adapter->hw;
1570 	u32 ctrl_ext;
1571 
1572 	/* Let firmware take over control of h/w */
1573 	ctrl_ext = rd32(E1000_CTRL_EXT);
1574 	wr32(E1000_CTRL_EXT,
1575 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1576 }
1577 
1578 /**
1579  *  igb_get_hw_control - get control of the h/w from f/w
1580  *  @adapter: address of board private structure
1581  *
1582  *  igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1583  *  For ASF and Pass Through versions of f/w this means that
1584  *  the driver is loaded.
1585  **/
1586 static void igb_get_hw_control(struct igb_adapter *adapter)
1587 {
1588 	struct e1000_hw *hw = &adapter->hw;
1589 	u32 ctrl_ext;
1590 
1591 	/* Let firmware know the driver has taken over */
1592 	ctrl_ext = rd32(E1000_CTRL_EXT);
1593 	wr32(E1000_CTRL_EXT,
1594 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1595 }
1596 
1597 static void enable_fqtss(struct igb_adapter *adapter, bool enable)
1598 {
1599 	struct net_device *netdev = adapter->netdev;
1600 	struct e1000_hw *hw = &adapter->hw;
1601 
1602 	WARN_ON(hw->mac.type != e1000_i210);
1603 
1604 	if (enable)
1605 		adapter->flags |= IGB_FLAG_FQTSS;
1606 	else
1607 		adapter->flags &= ~IGB_FLAG_FQTSS;
1608 
1609 	if (netif_running(netdev))
1610 		schedule_work(&adapter->reset_task);
1611 }
1612 
1613 static bool is_fqtss_enabled(struct igb_adapter *adapter)
1614 {
1615 	return (adapter->flags & IGB_FLAG_FQTSS) ? true : false;
1616 }
1617 
1618 static void set_tx_desc_fetch_prio(struct e1000_hw *hw, int queue,
1619 				   enum tx_queue_prio prio)
1620 {
1621 	u32 val;
1622 
1623 	WARN_ON(hw->mac.type != e1000_i210);
1624 	WARN_ON(queue < 0 || queue > 4);
1625 
1626 	val = rd32(E1000_I210_TXDCTL(queue));
1627 
1628 	if (prio == TX_QUEUE_PRIO_HIGH)
1629 		val |= E1000_TXDCTL_PRIORITY;
1630 	else
1631 		val &= ~E1000_TXDCTL_PRIORITY;
1632 
1633 	wr32(E1000_I210_TXDCTL(queue), val);
1634 }
1635 
1636 static void set_queue_mode(struct e1000_hw *hw, int queue, enum queue_mode mode)
1637 {
1638 	u32 val;
1639 
1640 	WARN_ON(hw->mac.type != e1000_i210);
1641 	WARN_ON(queue < 0 || queue > 1);
1642 
1643 	val = rd32(E1000_I210_TQAVCC(queue));
1644 
1645 	if (mode == QUEUE_MODE_STREAM_RESERVATION)
1646 		val |= E1000_TQAVCC_QUEUEMODE;
1647 	else
1648 		val &= ~E1000_TQAVCC_QUEUEMODE;
1649 
1650 	wr32(E1000_I210_TQAVCC(queue), val);
1651 }
1652 
1653 static bool is_any_cbs_enabled(struct igb_adapter *adapter)
1654 {
1655 	int i;
1656 
1657 	for (i = 0; i < adapter->num_tx_queues; i++) {
1658 		if (adapter->tx_ring[i]->cbs_enable)
1659 			return true;
1660 	}
1661 
1662 	return false;
1663 }
1664 
1665 static bool is_any_txtime_enabled(struct igb_adapter *adapter)
1666 {
1667 	int i;
1668 
1669 	for (i = 0; i < adapter->num_tx_queues; i++) {
1670 		if (adapter->tx_ring[i]->launchtime_enable)
1671 			return true;
1672 	}
1673 
1674 	return false;
1675 }
1676 
1677 /**
1678  *  igb_config_tx_modes - Configure "Qav Tx mode" features on igb
1679  *  @adapter: pointer to adapter struct
1680  *  @queue: queue number
1681  *
1682  *  Configure CBS and Launchtime for a given hardware queue.
1683  *  Parameters are retrieved from the correct Tx ring, so
1684  *  igb_save_cbs_params() and igb_save_txtime_params() should be used
1685  *  for setting those correctly prior to this function being called.
1686  **/
1687 static void igb_config_tx_modes(struct igb_adapter *adapter, int queue)
1688 {
1689 	struct net_device *netdev = adapter->netdev;
1690 	struct e1000_hw *hw = &adapter->hw;
1691 	struct igb_ring *ring;
1692 	u32 tqavcc, tqavctrl;
1693 	u16 value;
1694 
1695 	WARN_ON(hw->mac.type != e1000_i210);
1696 	WARN_ON(queue < 0 || queue > 1);
1697 	ring = adapter->tx_ring[queue];
1698 
1699 	/* If any of the Qav features is enabled, configure queues as SR and
1700 	 * with HIGH PRIO. If none is, then configure them with LOW PRIO and
1701 	 * as SP.
1702 	 */
1703 	if (ring->cbs_enable || ring->launchtime_enable) {
1704 		set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_HIGH);
1705 		set_queue_mode(hw, queue, QUEUE_MODE_STREAM_RESERVATION);
1706 	} else {
1707 		set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_LOW);
1708 		set_queue_mode(hw, queue, QUEUE_MODE_STRICT_PRIORITY);
1709 	}
1710 
1711 	/* If CBS is enabled, set DataTranARB and config its parameters. */
1712 	if (ring->cbs_enable || queue == 0) {
1713 		/* i210 does not allow the queue 0 to be in the Strict
1714 		 * Priority mode while the Qav mode is enabled, so,
1715 		 * instead of disabling strict priority mode, we give
1716 		 * queue 0 the maximum of credits possible.
1717 		 *
1718 		 * See section 8.12.19 of the i210 datasheet, "Note:
1719 		 * Queue0 QueueMode must be set to 1b when
1720 		 * TransmitMode is set to Qav."
1721 		 */
1722 		if (queue == 0 && !ring->cbs_enable) {
1723 			/* max "linkspeed" idleslope in kbps */
1724 			ring->idleslope = 1000000;
1725 			ring->hicredit = ETH_FRAME_LEN;
1726 		}
1727 
1728 		/* Always set data transfer arbitration to credit-based
1729 		 * shaper algorithm on TQAVCTRL if CBS is enabled for any of
1730 		 * the queues.
1731 		 */
1732 		tqavctrl = rd32(E1000_I210_TQAVCTRL);
1733 		tqavctrl |= E1000_TQAVCTRL_DATATRANARB;
1734 		wr32(E1000_I210_TQAVCTRL, tqavctrl);
1735 
1736 		/* According to i210 datasheet section 7.2.7.7, we should set
1737 		 * the 'idleSlope' field from TQAVCC register following the
1738 		 * equation:
1739 		 *
1740 		 * For 100 Mbps link speed:
1741 		 *
1742 		 *     value = BW * 0x7735 * 0.2                          (E1)
1743 		 *
1744 		 * For 1000Mbps link speed:
1745 		 *
1746 		 *     value = BW * 0x7735 * 2                            (E2)
1747 		 *
1748 		 * E1 and E2 can be merged into one equation as shown below.
1749 		 * Note that 'link-speed' is in Mbps.
1750 		 *
1751 		 *     value = BW * 0x7735 * 2 * link-speed
1752 		 *                           --------------               (E3)
1753 		 *                                1000
1754 		 *
1755 		 * 'BW' is the percentage bandwidth out of full link speed
1756 		 * which can be found with the following equation. Note that
1757 		 * idleSlope here is the parameter from this function which
1758 		 * is in kbps.
1759 		 *
1760 		 *     BW =     idleSlope
1761 		 *          -----------------                             (E4)
1762 		 *          link-speed * 1000
1763 		 *
1764 		 * That said, we can come up with a generic equation to
1765 		 * calculate the value we should set it TQAVCC register by
1766 		 * replacing 'BW' in E3 by E4. The resulting equation is:
1767 		 *
1768 		 * value =     idleSlope     * 0x7735 * 2 * link-speed
1769 		 *         -----------------            --------------    (E5)
1770 		 *         link-speed * 1000                 1000
1771 		 *
1772 		 * 'link-speed' is present in both sides of the fraction so
1773 		 * it is canceled out. The final equation is the following:
1774 		 *
1775 		 *     value = idleSlope * 61034
1776 		 *             -----------------                          (E6)
1777 		 *                  1000000
1778 		 *
1779 		 * NOTE: For i210, given the above, we can see that idleslope
1780 		 *       is represented in 16.38431 kbps units by the value at
1781 		 *       the TQAVCC register (1Gbps / 61034), which reduces
1782 		 *       the granularity for idleslope increments.
1783 		 *       For instance, if you want to configure a 2576kbps
1784 		 *       idleslope, the value to be written on the register
1785 		 *       would have to be 157.23. If rounded down, you end
1786 		 *       up with less bandwidth available than originally
1787 		 *       required (~2572 kbps). If rounded up, you end up
1788 		 *       with a higher bandwidth (~2589 kbps). Below the
1789 		 *       approach we take is to always round up the
1790 		 *       calculated value, so the resulting bandwidth might
1791 		 *       be slightly higher for some configurations.
1792 		 */
1793 		value = DIV_ROUND_UP_ULL(ring->idleslope * 61034ULL, 1000000);
1794 
1795 		tqavcc = rd32(E1000_I210_TQAVCC(queue));
1796 		tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1797 		tqavcc |= value;
1798 		wr32(E1000_I210_TQAVCC(queue), tqavcc);
1799 
1800 		wr32(E1000_I210_TQAVHC(queue),
1801 		     0x80000000 + ring->hicredit * 0x7735);
1802 	} else {
1803 
1804 		/* Set idleSlope to zero. */
1805 		tqavcc = rd32(E1000_I210_TQAVCC(queue));
1806 		tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1807 		wr32(E1000_I210_TQAVCC(queue), tqavcc);
1808 
1809 		/* Set hiCredit to zero. */
1810 		wr32(E1000_I210_TQAVHC(queue), 0);
1811 
1812 		/* If CBS is not enabled for any queues anymore, then return to
1813 		 * the default state of Data Transmission Arbitration on
1814 		 * TQAVCTRL.
1815 		 */
1816 		if (!is_any_cbs_enabled(adapter)) {
1817 			tqavctrl = rd32(E1000_I210_TQAVCTRL);
1818 			tqavctrl &= ~E1000_TQAVCTRL_DATATRANARB;
1819 			wr32(E1000_I210_TQAVCTRL, tqavctrl);
1820 		}
1821 	}
1822 
1823 	/* If LaunchTime is enabled, set DataTranTIM. */
1824 	if (ring->launchtime_enable) {
1825 		/* Always set DataTranTIM on TQAVCTRL if LaunchTime is enabled
1826 		 * for any of the SR queues, and configure fetchtime delta.
1827 		 * XXX NOTE:
1828 		 *     - LaunchTime will be enabled for all SR queues.
1829 		 *     - A fixed offset can be added relative to the launch
1830 		 *       time of all packets if configured at reg LAUNCH_OS0.
1831 		 *       We are keeping it as 0 for now (default value).
1832 		 */
1833 		tqavctrl = rd32(E1000_I210_TQAVCTRL);
1834 		tqavctrl |= E1000_TQAVCTRL_DATATRANTIM |
1835 		       E1000_TQAVCTRL_FETCHTIME_DELTA;
1836 		wr32(E1000_I210_TQAVCTRL, tqavctrl);
1837 	} else {
1838 		/* If Launchtime is not enabled for any SR queues anymore,
1839 		 * then clear DataTranTIM on TQAVCTRL and clear fetchtime delta,
1840 		 * effectively disabling Launchtime.
1841 		 */
1842 		if (!is_any_txtime_enabled(adapter)) {
1843 			tqavctrl = rd32(E1000_I210_TQAVCTRL);
1844 			tqavctrl &= ~E1000_TQAVCTRL_DATATRANTIM;
1845 			tqavctrl &= ~E1000_TQAVCTRL_FETCHTIME_DELTA;
1846 			wr32(E1000_I210_TQAVCTRL, tqavctrl);
1847 		}
1848 	}
1849 
1850 	/* XXX: In i210 controller the sendSlope and loCredit parameters from
1851 	 * CBS are not configurable by software so we don't do any 'controller
1852 	 * configuration' in respect to these parameters.
1853 	 */
1854 
1855 	netdev_dbg(netdev, "Qav Tx mode: cbs %s, launchtime %s, queue %d idleslope %d sendslope %d hiCredit %d locredit %d\n",
1856 		   ring->cbs_enable ? "enabled" : "disabled",
1857 		   ring->launchtime_enable ? "enabled" : "disabled",
1858 		   queue,
1859 		   ring->idleslope, ring->sendslope,
1860 		   ring->hicredit, ring->locredit);
1861 }
1862 
1863 static int igb_save_txtime_params(struct igb_adapter *adapter, int queue,
1864 				  bool enable)
1865 {
1866 	struct igb_ring *ring;
1867 
1868 	if (queue < 0 || queue > adapter->num_tx_queues)
1869 		return -EINVAL;
1870 
1871 	ring = adapter->tx_ring[queue];
1872 	ring->launchtime_enable = enable;
1873 
1874 	return 0;
1875 }
1876 
1877 static int igb_save_cbs_params(struct igb_adapter *adapter, int queue,
1878 			       bool enable, int idleslope, int sendslope,
1879 			       int hicredit, int locredit)
1880 {
1881 	struct igb_ring *ring;
1882 
1883 	if (queue < 0 || queue > adapter->num_tx_queues)
1884 		return -EINVAL;
1885 
1886 	ring = adapter->tx_ring[queue];
1887 
1888 	ring->cbs_enable = enable;
1889 	ring->idleslope = idleslope;
1890 	ring->sendslope = sendslope;
1891 	ring->hicredit = hicredit;
1892 	ring->locredit = locredit;
1893 
1894 	return 0;
1895 }
1896 
1897 /**
1898  *  igb_setup_tx_mode - Switch to/from Qav Tx mode when applicable
1899  *  @adapter: pointer to adapter struct
1900  *
1901  *  Configure TQAVCTRL register switching the controller's Tx mode
1902  *  if FQTSS mode is enabled or disabled. Additionally, will issue
1903  *  a call to igb_config_tx_modes() per queue so any previously saved
1904  *  Tx parameters are applied.
1905  **/
1906 static void igb_setup_tx_mode(struct igb_adapter *adapter)
1907 {
1908 	struct net_device *netdev = adapter->netdev;
1909 	struct e1000_hw *hw = &adapter->hw;
1910 	u32 val;
1911 
1912 	/* Only i210 controller supports changing the transmission mode. */
1913 	if (hw->mac.type != e1000_i210)
1914 		return;
1915 
1916 	if (is_fqtss_enabled(adapter)) {
1917 		int i, max_queue;
1918 
1919 		/* Configure TQAVCTRL register: set transmit mode to 'Qav',
1920 		 * set data fetch arbitration to 'round robin', set SP_WAIT_SR
1921 		 * so SP queues wait for SR ones.
1922 		 */
1923 		val = rd32(E1000_I210_TQAVCTRL);
1924 		val |= E1000_TQAVCTRL_XMIT_MODE | E1000_TQAVCTRL_SP_WAIT_SR;
1925 		val &= ~E1000_TQAVCTRL_DATAFETCHARB;
1926 		wr32(E1000_I210_TQAVCTRL, val);
1927 
1928 		/* Configure Tx and Rx packet buffers sizes as described in
1929 		 * i210 datasheet section 7.2.7.7.
1930 		 */
1931 		val = rd32(E1000_TXPBS);
1932 		val &= ~I210_TXPBSIZE_MASK;
1933 		val |= I210_TXPBSIZE_PB0_6KB | I210_TXPBSIZE_PB1_6KB |
1934 			I210_TXPBSIZE_PB2_6KB | I210_TXPBSIZE_PB3_6KB;
1935 		wr32(E1000_TXPBS, val);
1936 
1937 		val = rd32(E1000_RXPBS);
1938 		val &= ~I210_RXPBSIZE_MASK;
1939 		val |= I210_RXPBSIZE_PB_30KB;
1940 		wr32(E1000_RXPBS, val);
1941 
1942 		/* Section 8.12.9 states that MAX_TPKT_SIZE from DTXMXPKTSZ
1943 		 * register should not exceed the buffer size programmed in
1944 		 * TXPBS. The smallest buffer size programmed in TXPBS is 4kB
1945 		 * so according to the datasheet we should set MAX_TPKT_SIZE to
1946 		 * 4kB / 64.
1947 		 *
1948 		 * However, when we do so, no frame from queue 2 and 3 are
1949 		 * transmitted.  It seems the MAX_TPKT_SIZE should not be great
1950 		 * or _equal_ to the buffer size programmed in TXPBS. For this
1951 		 * reason, we set MAX_ TPKT_SIZE to (4kB - 1) / 64.
1952 		 */
1953 		val = (4096 - 1) / 64;
1954 		wr32(E1000_I210_DTXMXPKTSZ, val);
1955 
1956 		/* Since FQTSS mode is enabled, apply any CBS configuration
1957 		 * previously set. If no previous CBS configuration has been
1958 		 * done, then the initial configuration is applied, which means
1959 		 * CBS is disabled.
1960 		 */
1961 		max_queue = (adapter->num_tx_queues < I210_SR_QUEUES_NUM) ?
1962 			    adapter->num_tx_queues : I210_SR_QUEUES_NUM;
1963 
1964 		for (i = 0; i < max_queue; i++) {
1965 			igb_config_tx_modes(adapter, i);
1966 		}
1967 	} else {
1968 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
1969 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
1970 		wr32(E1000_I210_DTXMXPKTSZ, I210_DTXMXPKTSZ_DEFAULT);
1971 
1972 		val = rd32(E1000_I210_TQAVCTRL);
1973 		/* According to Section 8.12.21, the other flags we've set when
1974 		 * enabling FQTSS are not relevant when disabling FQTSS so we
1975 		 * don't set they here.
1976 		 */
1977 		val &= ~E1000_TQAVCTRL_XMIT_MODE;
1978 		wr32(E1000_I210_TQAVCTRL, val);
1979 	}
1980 
1981 	netdev_dbg(netdev, "FQTSS %s\n", (is_fqtss_enabled(adapter)) ?
1982 		   "enabled" : "disabled");
1983 }
1984 
1985 /**
1986  *  igb_configure - configure the hardware for RX and TX
1987  *  @adapter: private board structure
1988  **/
1989 static void igb_configure(struct igb_adapter *adapter)
1990 {
1991 	struct net_device *netdev = adapter->netdev;
1992 	int i;
1993 
1994 	igb_get_hw_control(adapter);
1995 	igb_set_rx_mode(netdev);
1996 	igb_setup_tx_mode(adapter);
1997 
1998 	igb_restore_vlan(adapter);
1999 
2000 	igb_setup_tctl(adapter);
2001 	igb_setup_mrqc(adapter);
2002 	igb_setup_rctl(adapter);
2003 
2004 	igb_nfc_filter_restore(adapter);
2005 	igb_configure_tx(adapter);
2006 	igb_configure_rx(adapter);
2007 
2008 	igb_rx_fifo_flush_82575(&adapter->hw);
2009 
2010 	/* call igb_desc_unused which always leaves
2011 	 * at least 1 descriptor unused to make sure
2012 	 * next_to_use != next_to_clean
2013 	 */
2014 	for (i = 0; i < adapter->num_rx_queues; i++) {
2015 		struct igb_ring *ring = adapter->rx_ring[i];
2016 		igb_alloc_rx_buffers(ring, igb_desc_unused(ring));
2017 	}
2018 }
2019 
2020 /**
2021  *  igb_power_up_link - Power up the phy/serdes link
2022  *  @adapter: address of board private structure
2023  **/
2024 void igb_power_up_link(struct igb_adapter *adapter)
2025 {
2026 	igb_reset_phy(&adapter->hw);
2027 
2028 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
2029 		igb_power_up_phy_copper(&adapter->hw);
2030 	else
2031 		igb_power_up_serdes_link_82575(&adapter->hw);
2032 
2033 	igb_setup_link(&adapter->hw);
2034 }
2035 
2036 /**
2037  *  igb_power_down_link - Power down the phy/serdes link
2038  *  @adapter: address of board private structure
2039  */
2040 static void igb_power_down_link(struct igb_adapter *adapter)
2041 {
2042 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
2043 		igb_power_down_phy_copper_82575(&adapter->hw);
2044 	else
2045 		igb_shutdown_serdes_link_82575(&adapter->hw);
2046 }
2047 
2048 /**
2049  * igb_check_swap_media -  Detect and switch function for Media Auto Sense
2050  * @adapter: address of the board private structure
2051  **/
2052 static void igb_check_swap_media(struct igb_adapter *adapter)
2053 {
2054 	struct e1000_hw *hw = &adapter->hw;
2055 	u32 ctrl_ext, connsw;
2056 	bool swap_now = false;
2057 
2058 	ctrl_ext = rd32(E1000_CTRL_EXT);
2059 	connsw = rd32(E1000_CONNSW);
2060 
2061 	/* need to live swap if current media is copper and we have fiber/serdes
2062 	 * to go to.
2063 	 */
2064 
2065 	if ((hw->phy.media_type == e1000_media_type_copper) &&
2066 	    (!(connsw & E1000_CONNSW_AUTOSENSE_EN))) {
2067 		swap_now = true;
2068 	} else if ((hw->phy.media_type != e1000_media_type_copper) &&
2069 		   !(connsw & E1000_CONNSW_SERDESD)) {
2070 		/* copper signal takes time to appear */
2071 		if (adapter->copper_tries < 4) {
2072 			adapter->copper_tries++;
2073 			connsw |= E1000_CONNSW_AUTOSENSE_CONF;
2074 			wr32(E1000_CONNSW, connsw);
2075 			return;
2076 		} else {
2077 			adapter->copper_tries = 0;
2078 			if ((connsw & E1000_CONNSW_PHYSD) &&
2079 			    (!(connsw & E1000_CONNSW_PHY_PDN))) {
2080 				swap_now = true;
2081 				connsw &= ~E1000_CONNSW_AUTOSENSE_CONF;
2082 				wr32(E1000_CONNSW, connsw);
2083 			}
2084 		}
2085 	}
2086 
2087 	if (!swap_now)
2088 		return;
2089 
2090 	switch (hw->phy.media_type) {
2091 	case e1000_media_type_copper:
2092 		netdev_info(adapter->netdev,
2093 			"MAS: changing media to fiber/serdes\n");
2094 		ctrl_ext |=
2095 			E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2096 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
2097 		adapter->copper_tries = 0;
2098 		break;
2099 	case e1000_media_type_internal_serdes:
2100 	case e1000_media_type_fiber:
2101 		netdev_info(adapter->netdev,
2102 			"MAS: changing media to copper\n");
2103 		ctrl_ext &=
2104 			~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2105 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
2106 		break;
2107 	default:
2108 		/* shouldn't get here during regular operation */
2109 		netdev_err(adapter->netdev,
2110 			"AMS: Invalid media type found, returning\n");
2111 		break;
2112 	}
2113 	wr32(E1000_CTRL_EXT, ctrl_ext);
2114 }
2115 
2116 /**
2117  *  igb_up - Open the interface and prepare it to handle traffic
2118  *  @adapter: board private structure
2119  **/
2120 int igb_up(struct igb_adapter *adapter)
2121 {
2122 	struct e1000_hw *hw = &adapter->hw;
2123 	int i;
2124 
2125 	/* hardware has been reset, we need to reload some things */
2126 	igb_configure(adapter);
2127 
2128 	clear_bit(__IGB_DOWN, &adapter->state);
2129 
2130 	for (i = 0; i < adapter->num_q_vectors; i++)
2131 		napi_enable(&(adapter->q_vector[i]->napi));
2132 
2133 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
2134 		igb_configure_msix(adapter);
2135 	else
2136 		igb_assign_vector(adapter->q_vector[0], 0);
2137 
2138 	/* Clear any pending interrupts. */
2139 	rd32(E1000_TSICR);
2140 	rd32(E1000_ICR);
2141 	igb_irq_enable(adapter);
2142 
2143 	/* notify VFs that reset has been completed */
2144 	if (adapter->vfs_allocated_count) {
2145 		u32 reg_data = rd32(E1000_CTRL_EXT);
2146 
2147 		reg_data |= E1000_CTRL_EXT_PFRSTD;
2148 		wr32(E1000_CTRL_EXT, reg_data);
2149 	}
2150 
2151 	netif_tx_start_all_queues(adapter->netdev);
2152 
2153 	/* start the watchdog. */
2154 	hw->mac.get_link_status = 1;
2155 	schedule_work(&adapter->watchdog_task);
2156 
2157 	if ((adapter->flags & IGB_FLAG_EEE) &&
2158 	    (!hw->dev_spec._82575.eee_disable))
2159 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
2160 
2161 	return 0;
2162 }
2163 
2164 void igb_down(struct igb_adapter *adapter)
2165 {
2166 	struct net_device *netdev = adapter->netdev;
2167 	struct e1000_hw *hw = &adapter->hw;
2168 	u32 tctl, rctl;
2169 	int i;
2170 
2171 	/* signal that we're down so the interrupt handler does not
2172 	 * reschedule our watchdog timer
2173 	 */
2174 	set_bit(__IGB_DOWN, &adapter->state);
2175 
2176 	/* disable receives in the hardware */
2177 	rctl = rd32(E1000_RCTL);
2178 	wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
2179 	/* flush and sleep below */
2180 
2181 	igb_nfc_filter_exit(adapter);
2182 
2183 	netif_carrier_off(netdev);
2184 	netif_tx_stop_all_queues(netdev);
2185 
2186 	/* disable transmits in the hardware */
2187 	tctl = rd32(E1000_TCTL);
2188 	tctl &= ~E1000_TCTL_EN;
2189 	wr32(E1000_TCTL, tctl);
2190 	/* flush both disables and wait for them to finish */
2191 	wrfl();
2192 	usleep_range(10000, 11000);
2193 
2194 	igb_irq_disable(adapter);
2195 
2196 	adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
2197 
2198 	for (i = 0; i < adapter->num_q_vectors; i++) {
2199 		if (adapter->q_vector[i]) {
2200 			napi_synchronize(&adapter->q_vector[i]->napi);
2201 			napi_disable(&adapter->q_vector[i]->napi);
2202 		}
2203 	}
2204 
2205 	del_timer_sync(&adapter->watchdog_timer);
2206 	del_timer_sync(&adapter->phy_info_timer);
2207 
2208 	/* record the stats before reset*/
2209 	spin_lock(&adapter->stats64_lock);
2210 	igb_update_stats(adapter);
2211 	spin_unlock(&adapter->stats64_lock);
2212 
2213 	adapter->link_speed = 0;
2214 	adapter->link_duplex = 0;
2215 
2216 	if (!pci_channel_offline(adapter->pdev))
2217 		igb_reset(adapter);
2218 
2219 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
2220 	adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
2221 
2222 	igb_clean_all_tx_rings(adapter);
2223 	igb_clean_all_rx_rings(adapter);
2224 #ifdef CONFIG_IGB_DCA
2225 
2226 	/* since we reset the hardware DCA settings were cleared */
2227 	igb_setup_dca(adapter);
2228 #endif
2229 }
2230 
2231 void igb_reinit_locked(struct igb_adapter *adapter)
2232 {
2233 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
2234 		usleep_range(1000, 2000);
2235 	igb_down(adapter);
2236 	igb_up(adapter);
2237 	clear_bit(__IGB_RESETTING, &adapter->state);
2238 }
2239 
2240 /** igb_enable_mas - Media Autosense re-enable after swap
2241  *
2242  * @adapter: adapter struct
2243  **/
2244 static void igb_enable_mas(struct igb_adapter *adapter)
2245 {
2246 	struct e1000_hw *hw = &adapter->hw;
2247 	u32 connsw = rd32(E1000_CONNSW);
2248 
2249 	/* configure for SerDes media detect */
2250 	if ((hw->phy.media_type == e1000_media_type_copper) &&
2251 	    (!(connsw & E1000_CONNSW_SERDESD))) {
2252 		connsw |= E1000_CONNSW_ENRGSRC;
2253 		connsw |= E1000_CONNSW_AUTOSENSE_EN;
2254 		wr32(E1000_CONNSW, connsw);
2255 		wrfl();
2256 	}
2257 }
2258 
2259 #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(FLOW_DISSECTOR_KEY_BASIC) |
2619 	      BIT(FLOW_DISSECTOR_KEY_CONTROL) |
2620 	      BIT(FLOW_DISSECTOR_KEY_ETH_ADDRS) |
2621 	      BIT(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 LIST_HEAD(igb_block_cb_list);
2839 
2840 static int igb_setup_tc(struct net_device *dev, enum tc_setup_type type,
2841 			void *type_data)
2842 {
2843 	struct igb_adapter *adapter = netdev_priv(dev);
2844 
2845 	switch (type) {
2846 	case TC_SETUP_QDISC_CBS:
2847 		return igb_offload_cbs(adapter, type_data);
2848 	case TC_SETUP_BLOCK:
2849 		return flow_block_cb_setup_simple(type_data,
2850 						  &igb_block_cb_list,
2851 						  igb_setup_tc_block_cb,
2852 						  adapter, adapter, true);
2853 
2854 	case TC_SETUP_QDISC_ETF:
2855 		return igb_offload_txtime(adapter, type_data);
2856 
2857 	default:
2858 		return -EOPNOTSUPP;
2859 	}
2860 }
2861 
2862 static int igb_xdp_setup(struct net_device *dev, struct netdev_bpf *bpf)
2863 {
2864 	int i, frame_size = dev->mtu + IGB_ETH_PKT_HDR_PAD;
2865 	struct igb_adapter *adapter = netdev_priv(dev);
2866 	struct bpf_prog *prog = bpf->prog, *old_prog;
2867 	bool running = netif_running(dev);
2868 	bool need_reset;
2869 
2870 	/* verify igb ring attributes are sufficient for XDP */
2871 	for (i = 0; i < adapter->num_rx_queues; i++) {
2872 		struct igb_ring *ring = adapter->rx_ring[i];
2873 
2874 		if (frame_size > igb_rx_bufsz(ring)) {
2875 			NL_SET_ERR_MSG_MOD(bpf->extack,
2876 					   "The RX buffer size is too small for the frame size");
2877 			netdev_warn(dev, "XDP RX buffer size %d is too small for the frame size %d\n",
2878 				    igb_rx_bufsz(ring), frame_size);
2879 			return -EINVAL;
2880 		}
2881 	}
2882 
2883 	old_prog = xchg(&adapter->xdp_prog, prog);
2884 	need_reset = (!!prog != !!old_prog);
2885 
2886 	/* device is up and bpf is added/removed, must setup the RX queues */
2887 	if (need_reset && running) {
2888 		igb_close(dev);
2889 	} else {
2890 		for (i = 0; i < adapter->num_rx_queues; i++)
2891 			(void)xchg(&adapter->rx_ring[i]->xdp_prog,
2892 			    adapter->xdp_prog);
2893 	}
2894 
2895 	if (old_prog)
2896 		bpf_prog_put(old_prog);
2897 
2898 	/* bpf is just replaced, RXQ and MTU are already setup */
2899 	if (!need_reset)
2900 		return 0;
2901 
2902 	if (running)
2903 		igb_open(dev);
2904 
2905 	return 0;
2906 }
2907 
2908 static int igb_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2909 {
2910 	switch (xdp->command) {
2911 	case XDP_SETUP_PROG:
2912 		return igb_xdp_setup(dev, xdp);
2913 	default:
2914 		return -EINVAL;
2915 	}
2916 }
2917 
2918 static void igb_xdp_ring_update_tail(struct igb_ring *ring)
2919 {
2920 	/* Force memory writes to complete before letting h/w know there
2921 	 * are new descriptors to fetch.
2922 	 */
2923 	wmb();
2924 	writel(ring->next_to_use, ring->tail);
2925 }
2926 
2927 static struct igb_ring *igb_xdp_tx_queue_mapping(struct igb_adapter *adapter)
2928 {
2929 	unsigned int r_idx = smp_processor_id();
2930 
2931 	if (r_idx >= adapter->num_tx_queues)
2932 		r_idx = r_idx % adapter->num_tx_queues;
2933 
2934 	return adapter->tx_ring[r_idx];
2935 }
2936 
2937 static int igb_xdp_xmit_back(struct igb_adapter *adapter, struct xdp_buff *xdp)
2938 {
2939 	struct xdp_frame *xdpf = xdp_convert_buff_to_frame(xdp);
2940 	int cpu = smp_processor_id();
2941 	struct igb_ring *tx_ring;
2942 	struct netdev_queue *nq;
2943 	u32 ret;
2944 
2945 	if (unlikely(!xdpf))
2946 		return IGB_XDP_CONSUMED;
2947 
2948 	/* During program transitions its possible adapter->xdp_prog is assigned
2949 	 * but ring has not been configured yet. In this case simply abort xmit.
2950 	 */
2951 	tx_ring = adapter->xdp_prog ? igb_xdp_tx_queue_mapping(adapter) : NULL;
2952 	if (unlikely(!tx_ring))
2953 		return IGB_XDP_CONSUMED;
2954 
2955 	nq = txring_txq(tx_ring);
2956 	__netif_tx_lock(nq, cpu);
2957 	/* Avoid transmit queue timeout since we share it with the slow path */
2958 	txq_trans_cond_update(nq);
2959 	ret = igb_xmit_xdp_ring(adapter, tx_ring, xdpf);
2960 	__netif_tx_unlock(nq);
2961 
2962 	return ret;
2963 }
2964 
2965 static int igb_xdp_xmit(struct net_device *dev, int n,
2966 			struct xdp_frame **frames, u32 flags)
2967 {
2968 	struct igb_adapter *adapter = netdev_priv(dev);
2969 	int cpu = smp_processor_id();
2970 	struct igb_ring *tx_ring;
2971 	struct netdev_queue *nq;
2972 	int nxmit = 0;
2973 	int i;
2974 
2975 	if (unlikely(test_bit(__IGB_DOWN, &adapter->state)))
2976 		return -ENETDOWN;
2977 
2978 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
2979 		return -EINVAL;
2980 
2981 	/* During program transitions its possible adapter->xdp_prog is assigned
2982 	 * but ring has not been configured yet. In this case simply abort xmit.
2983 	 */
2984 	tx_ring = adapter->xdp_prog ? igb_xdp_tx_queue_mapping(adapter) : NULL;
2985 	if (unlikely(!tx_ring))
2986 		return -ENXIO;
2987 
2988 	nq = txring_txq(tx_ring);
2989 	__netif_tx_lock(nq, cpu);
2990 
2991 	/* Avoid transmit queue timeout since we share it with the slow path */
2992 	txq_trans_cond_update(nq);
2993 
2994 	for (i = 0; i < n; i++) {
2995 		struct xdp_frame *xdpf = frames[i];
2996 		int err;
2997 
2998 		err = igb_xmit_xdp_ring(adapter, tx_ring, xdpf);
2999 		if (err != IGB_XDP_TX)
3000 			break;
3001 		nxmit++;
3002 	}
3003 
3004 	__netif_tx_unlock(nq);
3005 
3006 	if (unlikely(flags & XDP_XMIT_FLUSH))
3007 		igb_xdp_ring_update_tail(tx_ring);
3008 
3009 	return nxmit;
3010 }
3011 
3012 static const struct net_device_ops igb_netdev_ops = {
3013 	.ndo_open		= igb_open,
3014 	.ndo_stop		= igb_close,
3015 	.ndo_start_xmit		= igb_xmit_frame,
3016 	.ndo_get_stats64	= igb_get_stats64,
3017 	.ndo_set_rx_mode	= igb_set_rx_mode,
3018 	.ndo_set_mac_address	= igb_set_mac,
3019 	.ndo_change_mtu		= igb_change_mtu,
3020 	.ndo_eth_ioctl		= igb_ioctl,
3021 	.ndo_tx_timeout		= igb_tx_timeout,
3022 	.ndo_validate_addr	= eth_validate_addr,
3023 	.ndo_vlan_rx_add_vid	= igb_vlan_rx_add_vid,
3024 	.ndo_vlan_rx_kill_vid	= igb_vlan_rx_kill_vid,
3025 	.ndo_set_vf_mac		= igb_ndo_set_vf_mac,
3026 	.ndo_set_vf_vlan	= igb_ndo_set_vf_vlan,
3027 	.ndo_set_vf_rate	= igb_ndo_set_vf_bw,
3028 	.ndo_set_vf_spoofchk	= igb_ndo_set_vf_spoofchk,
3029 	.ndo_set_vf_trust	= igb_ndo_set_vf_trust,
3030 	.ndo_get_vf_config	= igb_ndo_get_vf_config,
3031 	.ndo_fix_features	= igb_fix_features,
3032 	.ndo_set_features	= igb_set_features,
3033 	.ndo_fdb_add		= igb_ndo_fdb_add,
3034 	.ndo_features_check	= igb_features_check,
3035 	.ndo_setup_tc		= igb_setup_tc,
3036 	.ndo_bpf		= igb_xdp,
3037 	.ndo_xdp_xmit		= igb_xdp_xmit,
3038 };
3039 
3040 /**
3041  * igb_set_fw_version - Configure version string for ethtool
3042  * @adapter: adapter struct
3043  **/
3044 void igb_set_fw_version(struct igb_adapter *adapter)
3045 {
3046 	struct e1000_hw *hw = &adapter->hw;
3047 	struct e1000_fw_version fw;
3048 
3049 	igb_get_fw_version(hw, &fw);
3050 
3051 	switch (hw->mac.type) {
3052 	case e1000_i210:
3053 	case e1000_i211:
3054 		if (!(igb_get_flash_presence_i210(hw))) {
3055 			snprintf(adapter->fw_version,
3056 				 sizeof(adapter->fw_version),
3057 				 "%2d.%2d-%d",
3058 				 fw.invm_major, fw.invm_minor,
3059 				 fw.invm_img_type);
3060 			break;
3061 		}
3062 		fallthrough;
3063 	default:
3064 		/* if option is rom valid, display its version too */
3065 		if (fw.or_valid) {
3066 			snprintf(adapter->fw_version,
3067 				 sizeof(adapter->fw_version),
3068 				 "%d.%d, 0x%08x, %d.%d.%d",
3069 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
3070 				 fw.or_major, fw.or_build, fw.or_patch);
3071 		/* no option rom */
3072 		} else if (fw.etrack_id != 0X0000) {
3073 			snprintf(adapter->fw_version,
3074 			    sizeof(adapter->fw_version),
3075 			    "%d.%d, 0x%08x",
3076 			    fw.eep_major, fw.eep_minor, fw.etrack_id);
3077 		} else {
3078 		snprintf(adapter->fw_version,
3079 		    sizeof(adapter->fw_version),
3080 		    "%d.%d.%d",
3081 		    fw.eep_major, fw.eep_minor, fw.eep_build);
3082 		}
3083 		break;
3084 	}
3085 }
3086 
3087 /**
3088  * igb_init_mas - init Media Autosense feature if enabled in the NVM
3089  *
3090  * @adapter: adapter struct
3091  **/
3092 static void igb_init_mas(struct igb_adapter *adapter)
3093 {
3094 	struct e1000_hw *hw = &adapter->hw;
3095 	u16 eeprom_data;
3096 
3097 	hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
3098 	switch (hw->bus.func) {
3099 	case E1000_FUNC_0:
3100 		if (eeprom_data & IGB_MAS_ENABLE_0) {
3101 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3102 			netdev_info(adapter->netdev,
3103 				"MAS: Enabling Media Autosense for port %d\n",
3104 				hw->bus.func);
3105 		}
3106 		break;
3107 	case E1000_FUNC_1:
3108 		if (eeprom_data & IGB_MAS_ENABLE_1) {
3109 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3110 			netdev_info(adapter->netdev,
3111 				"MAS: Enabling Media Autosense for port %d\n",
3112 				hw->bus.func);
3113 		}
3114 		break;
3115 	case E1000_FUNC_2:
3116 		if (eeprom_data & IGB_MAS_ENABLE_2) {
3117 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3118 			netdev_info(adapter->netdev,
3119 				"MAS: Enabling Media Autosense for port %d\n",
3120 				hw->bus.func);
3121 		}
3122 		break;
3123 	case E1000_FUNC_3:
3124 		if (eeprom_data & IGB_MAS_ENABLE_3) {
3125 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
3126 			netdev_info(adapter->netdev,
3127 				"MAS: Enabling Media Autosense for port %d\n",
3128 				hw->bus.func);
3129 		}
3130 		break;
3131 	default:
3132 		/* Shouldn't get here */
3133 		netdev_err(adapter->netdev,
3134 			"MAS: Invalid port configuration, returning\n");
3135 		break;
3136 	}
3137 }
3138 
3139 /**
3140  *  igb_init_i2c - Init I2C interface
3141  *  @adapter: pointer to adapter structure
3142  **/
3143 static s32 igb_init_i2c(struct igb_adapter *adapter)
3144 {
3145 	s32 status = 0;
3146 
3147 	/* I2C interface supported on i350 devices */
3148 	if (adapter->hw.mac.type != e1000_i350)
3149 		return 0;
3150 
3151 	/* Initialize the i2c bus which is controlled by the registers.
3152 	 * This bus will use the i2c_algo_bit structure that implements
3153 	 * the protocol through toggling of the 4 bits in the register.
3154 	 */
3155 	adapter->i2c_adap.owner = THIS_MODULE;
3156 	adapter->i2c_algo = igb_i2c_algo;
3157 	adapter->i2c_algo.data = adapter;
3158 	adapter->i2c_adap.algo_data = &adapter->i2c_algo;
3159 	adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
3160 	strscpy(adapter->i2c_adap.name, "igb BB",
3161 		sizeof(adapter->i2c_adap.name));
3162 	status = i2c_bit_add_bus(&adapter->i2c_adap);
3163 	return status;
3164 }
3165 
3166 /**
3167  *  igb_probe - Device Initialization Routine
3168  *  @pdev: PCI device information struct
3169  *  @ent: entry in igb_pci_tbl
3170  *
3171  *  Returns 0 on success, negative on failure
3172  *
3173  *  igb_probe initializes an adapter identified by a pci_dev structure.
3174  *  The OS initialization, configuring of the adapter private structure,
3175  *  and a hardware reset occur.
3176  **/
3177 static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
3178 {
3179 	struct net_device *netdev;
3180 	struct igb_adapter *adapter;
3181 	struct e1000_hw *hw;
3182 	u16 eeprom_data = 0;
3183 	s32 ret_val;
3184 	static int global_quad_port_a; /* global quad port a indication */
3185 	const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
3186 	u8 part_str[E1000_PBANUM_LENGTH];
3187 	int err;
3188 
3189 	/* Catch broken hardware that put the wrong VF device ID in
3190 	 * the PCIe SR-IOV capability.
3191 	 */
3192 	if (pdev->is_virtfn) {
3193 		WARN(1, KERN_ERR "%s (%x:%x) should not be a VF!\n",
3194 			pci_name(pdev), pdev->vendor, pdev->device);
3195 		return -EINVAL;
3196 	}
3197 
3198 	err = pci_enable_device_mem(pdev);
3199 	if (err)
3200 		return err;
3201 
3202 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
3203 	if (err) {
3204 		dev_err(&pdev->dev,
3205 			"No usable DMA configuration, aborting\n");
3206 		goto err_dma;
3207 	}
3208 
3209 	err = pci_request_mem_regions(pdev, igb_driver_name);
3210 	if (err)
3211 		goto err_pci_reg;
3212 
3213 	pci_enable_pcie_error_reporting(pdev);
3214 
3215 	pci_set_master(pdev);
3216 	pci_save_state(pdev);
3217 
3218 	err = -ENOMEM;
3219 	netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
3220 				   IGB_MAX_TX_QUEUES);
3221 	if (!netdev)
3222 		goto err_alloc_etherdev;
3223 
3224 	SET_NETDEV_DEV(netdev, &pdev->dev);
3225 
3226 	pci_set_drvdata(pdev, netdev);
3227 	adapter = netdev_priv(netdev);
3228 	adapter->netdev = netdev;
3229 	adapter->pdev = pdev;
3230 	hw = &adapter->hw;
3231 	hw->back = adapter;
3232 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
3233 
3234 	err = -EIO;
3235 	adapter->io_addr = pci_iomap(pdev, 0, 0);
3236 	if (!adapter->io_addr)
3237 		goto err_ioremap;
3238 	/* hw->hw_addr can be altered, we'll use adapter->io_addr for unmap */
3239 	hw->hw_addr = adapter->io_addr;
3240 
3241 	netdev->netdev_ops = &igb_netdev_ops;
3242 	igb_set_ethtool_ops(netdev);
3243 	netdev->watchdog_timeo = 5 * HZ;
3244 
3245 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
3246 
3247 	netdev->mem_start = pci_resource_start(pdev, 0);
3248 	netdev->mem_end = pci_resource_end(pdev, 0);
3249 
3250 	/* PCI config space info */
3251 	hw->vendor_id = pdev->vendor;
3252 	hw->device_id = pdev->device;
3253 	hw->revision_id = pdev->revision;
3254 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
3255 	hw->subsystem_device_id = pdev->subsystem_device;
3256 
3257 	/* Copy the default MAC, PHY and NVM function pointers */
3258 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
3259 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
3260 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
3261 	/* Initialize skew-specific constants */
3262 	err = ei->get_invariants(hw);
3263 	if (err)
3264 		goto err_sw_init;
3265 
3266 	/* setup the private structure */
3267 	err = igb_sw_init(adapter);
3268 	if (err)
3269 		goto err_sw_init;
3270 
3271 	igb_get_bus_info_pcie(hw);
3272 
3273 	hw->phy.autoneg_wait_to_complete = false;
3274 
3275 	/* Copper options */
3276 	if (hw->phy.media_type == e1000_media_type_copper) {
3277 		hw->phy.mdix = AUTO_ALL_MODES;
3278 		hw->phy.disable_polarity_correction = false;
3279 		hw->phy.ms_type = e1000_ms_hw_default;
3280 	}
3281 
3282 	if (igb_check_reset_block(hw))
3283 		dev_info(&pdev->dev,
3284 			"PHY reset is blocked due to SOL/IDER session.\n");
3285 
3286 	/* features is initialized to 0 in allocation, it might have bits
3287 	 * set by igb_sw_init so we should use an or instead of an
3288 	 * assignment.
3289 	 */
3290 	netdev->features |= NETIF_F_SG |
3291 			    NETIF_F_TSO |
3292 			    NETIF_F_TSO6 |
3293 			    NETIF_F_RXHASH |
3294 			    NETIF_F_RXCSUM |
3295 			    NETIF_F_HW_CSUM;
3296 
3297 	if (hw->mac.type >= e1000_82576)
3298 		netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_GSO_UDP_L4;
3299 
3300 	if (hw->mac.type >= e1000_i350)
3301 		netdev->features |= NETIF_F_HW_TC;
3302 
3303 #define IGB_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
3304 				  NETIF_F_GSO_GRE_CSUM | \
3305 				  NETIF_F_GSO_IPXIP4 | \
3306 				  NETIF_F_GSO_IPXIP6 | \
3307 				  NETIF_F_GSO_UDP_TUNNEL | \
3308 				  NETIF_F_GSO_UDP_TUNNEL_CSUM)
3309 
3310 	netdev->gso_partial_features = IGB_GSO_PARTIAL_FEATURES;
3311 	netdev->features |= NETIF_F_GSO_PARTIAL | IGB_GSO_PARTIAL_FEATURES;
3312 
3313 	/* copy netdev features into list of user selectable features */
3314 	netdev->hw_features |= netdev->features |
3315 			       NETIF_F_HW_VLAN_CTAG_RX |
3316 			       NETIF_F_HW_VLAN_CTAG_TX |
3317 			       NETIF_F_RXALL;
3318 
3319 	if (hw->mac.type >= e1000_i350)
3320 		netdev->hw_features |= NETIF_F_NTUPLE;
3321 
3322 	netdev->features |= NETIF_F_HIGHDMA;
3323 
3324 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
3325 	netdev->mpls_features |= NETIF_F_HW_CSUM;
3326 	netdev->hw_enc_features |= netdev->vlan_features;
3327 
3328 	/* set this bit last since it cannot be part of vlan_features */
3329 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
3330 			    NETIF_F_HW_VLAN_CTAG_RX |
3331 			    NETIF_F_HW_VLAN_CTAG_TX;
3332 
3333 	netdev->priv_flags |= IFF_SUPP_NOFCS;
3334 
3335 	netdev->priv_flags |= IFF_UNICAST_FLT;
3336 
3337 	/* MTU range: 68 - 9216 */
3338 	netdev->min_mtu = ETH_MIN_MTU;
3339 	netdev->max_mtu = MAX_STD_JUMBO_FRAME_SIZE;
3340 
3341 	adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
3342 
3343 	/* before reading the NVM, reset the controller to put the device in a
3344 	 * known good starting state
3345 	 */
3346 	hw->mac.ops.reset_hw(hw);
3347 
3348 	/* make sure the NVM is good , i211/i210 parts can have special NVM
3349 	 * that doesn't contain a checksum
3350 	 */
3351 	switch (hw->mac.type) {
3352 	case e1000_i210:
3353 	case e1000_i211:
3354 		if (igb_get_flash_presence_i210(hw)) {
3355 			if (hw->nvm.ops.validate(hw) < 0) {
3356 				dev_err(&pdev->dev,
3357 					"The NVM Checksum Is Not Valid\n");
3358 				err = -EIO;
3359 				goto err_eeprom;
3360 			}
3361 		}
3362 		break;
3363 	default:
3364 		if (hw->nvm.ops.validate(hw) < 0) {
3365 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
3366 			err = -EIO;
3367 			goto err_eeprom;
3368 		}
3369 		break;
3370 	}
3371 
3372 	if (eth_platform_get_mac_address(&pdev->dev, hw->mac.addr)) {
3373 		/* copy the MAC address out of the NVM */
3374 		if (hw->mac.ops.read_mac_addr(hw))
3375 			dev_err(&pdev->dev, "NVM Read Error\n");
3376 	}
3377 
3378 	eth_hw_addr_set(netdev, hw->mac.addr);
3379 
3380 	if (!is_valid_ether_addr(netdev->dev_addr)) {
3381 		dev_err(&pdev->dev, "Invalid MAC Address\n");
3382 		err = -EIO;
3383 		goto err_eeprom;
3384 	}
3385 
3386 	igb_set_default_mac_filter(adapter);
3387 
3388 	/* get firmware version for ethtool -i */
3389 	igb_set_fw_version(adapter);
3390 
3391 	/* configure RXPBSIZE and TXPBSIZE */
3392 	if (hw->mac.type == e1000_i210) {
3393 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
3394 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
3395 	}
3396 
3397 	timer_setup(&adapter->watchdog_timer, igb_watchdog, 0);
3398 	timer_setup(&adapter->phy_info_timer, igb_update_phy_info, 0);
3399 
3400 	INIT_WORK(&adapter->reset_task, igb_reset_task);
3401 	INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
3402 
3403 	/* Initialize link properties that are user-changeable */
3404 	adapter->fc_autoneg = true;
3405 	hw->mac.autoneg = true;
3406 	hw->phy.autoneg_advertised = 0x2f;
3407 
3408 	hw->fc.requested_mode = e1000_fc_default;
3409 	hw->fc.current_mode = e1000_fc_default;
3410 
3411 	igb_validate_mdi_setting(hw);
3412 
3413 	/* By default, support wake on port A */
3414 	if (hw->bus.func == 0)
3415 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3416 
3417 	/* Check the NVM for wake support on non-port A ports */
3418 	if (hw->mac.type >= e1000_82580)
3419 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
3420 				 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
3421 				 &eeprom_data);
3422 	else if (hw->bus.func == 1)
3423 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3424 
3425 	if (eeprom_data & IGB_EEPROM_APME)
3426 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3427 
3428 	/* now that we have the eeprom settings, apply the special cases where
3429 	 * the eeprom may be wrong or the board simply won't support wake on
3430 	 * lan on a particular port
3431 	 */
3432 	switch (pdev->device) {
3433 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
3434 		adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3435 		break;
3436 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
3437 	case E1000_DEV_ID_82576_FIBER:
3438 	case E1000_DEV_ID_82576_SERDES:
3439 		/* Wake events only supported on port A for dual fiber
3440 		 * regardless of eeprom setting
3441 		 */
3442 		if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
3443 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3444 		break;
3445 	case E1000_DEV_ID_82576_QUAD_COPPER:
3446 	case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
3447 		/* if quad port adapter, disable WoL on all but port A */
3448 		if (global_quad_port_a != 0)
3449 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3450 		else
3451 			adapter->flags |= IGB_FLAG_QUAD_PORT_A;
3452 		/* Reset for multiple quad port adapters */
3453 		if (++global_quad_port_a == 4)
3454 			global_quad_port_a = 0;
3455 		break;
3456 	default:
3457 		/* If the device can't wake, don't set software support */
3458 		if (!device_can_wakeup(&adapter->pdev->dev))
3459 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3460 	}
3461 
3462 	/* initialize the wol settings based on the eeprom settings */
3463 	if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
3464 		adapter->wol |= E1000_WUFC_MAG;
3465 
3466 	/* Some vendors want WoL disabled by default, but still supported */
3467 	if ((hw->mac.type == e1000_i350) &&
3468 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
3469 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3470 		adapter->wol = 0;
3471 	}
3472 
3473 	/* Some vendors want the ability to Use the EEPROM setting as
3474 	 * enable/disable only, and not for capability
3475 	 */
3476 	if (((hw->mac.type == e1000_i350) ||
3477 	     (hw->mac.type == e1000_i354)) &&
3478 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_DELL)) {
3479 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3480 		adapter->wol = 0;
3481 	}
3482 	if (hw->mac.type == e1000_i350) {
3483 		if (((pdev->subsystem_device == 0x5001) ||
3484 		     (pdev->subsystem_device == 0x5002)) &&
3485 				(hw->bus.func == 0)) {
3486 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3487 			adapter->wol = 0;
3488 		}
3489 		if (pdev->subsystem_device == 0x1F52)
3490 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3491 	}
3492 
3493 	device_set_wakeup_enable(&adapter->pdev->dev,
3494 				 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
3495 
3496 	/* reset the hardware with the new settings */
3497 	igb_reset(adapter);
3498 
3499 	/* Init the I2C interface */
3500 	err = igb_init_i2c(adapter);
3501 	if (err) {
3502 		dev_err(&pdev->dev, "failed to init i2c interface\n");
3503 		goto err_eeprom;
3504 	}
3505 
3506 	/* let the f/w know that the h/w is now under the control of the
3507 	 * driver.
3508 	 */
3509 	igb_get_hw_control(adapter);
3510 
3511 	strcpy(netdev->name, "eth%d");
3512 	err = register_netdev(netdev);
3513 	if (err)
3514 		goto err_register;
3515 
3516 	/* carrier off reporting is important to ethtool even BEFORE open */
3517 	netif_carrier_off(netdev);
3518 
3519 #ifdef CONFIG_IGB_DCA
3520 	if (dca_add_requester(&pdev->dev) == 0) {
3521 		adapter->flags |= IGB_FLAG_DCA_ENABLED;
3522 		dev_info(&pdev->dev, "DCA enabled\n");
3523 		igb_setup_dca(adapter);
3524 	}
3525 
3526 #endif
3527 #ifdef CONFIG_IGB_HWMON
3528 	/* Initialize the thermal sensor on i350 devices. */
3529 	if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
3530 		u16 ets_word;
3531 
3532 		/* Read the NVM to determine if this i350 device supports an
3533 		 * external thermal sensor.
3534 		 */
3535 		hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
3536 		if (ets_word != 0x0000 && ets_word != 0xFFFF)
3537 			adapter->ets = true;
3538 		else
3539 			adapter->ets = false;
3540 		/* Only enable I2C bit banging if an external thermal
3541 		 * sensor is supported.
3542 		 */
3543 		if (adapter->ets)
3544 			igb_set_i2c_bb(hw);
3545 		hw->mac.ops.init_thermal_sensor_thresh(hw);
3546 		if (igb_sysfs_init(adapter))
3547 			dev_err(&pdev->dev,
3548 				"failed to allocate sysfs resources\n");
3549 	} else {
3550 		adapter->ets = false;
3551 	}
3552 #endif
3553 	/* Check if Media Autosense is enabled */
3554 	adapter->ei = *ei;
3555 	if (hw->dev_spec._82575.mas_capable)
3556 		igb_init_mas(adapter);
3557 
3558 	/* do hw tstamp init after resetting */
3559 	igb_ptp_init(adapter);
3560 
3561 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
3562 	/* print bus type/speed/width info, not applicable to i354 */
3563 	if (hw->mac.type != e1000_i354) {
3564 		dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
3565 			 netdev->name,
3566 			 ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
3567 			  (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
3568 			   "unknown"),
3569 			 ((hw->bus.width == e1000_bus_width_pcie_x4) ?
3570 			  "Width x4" :
3571 			  (hw->bus.width == e1000_bus_width_pcie_x2) ?
3572 			  "Width x2" :
3573 			  (hw->bus.width == e1000_bus_width_pcie_x1) ?
3574 			  "Width x1" : "unknown"), netdev->dev_addr);
3575 	}
3576 
3577 	if ((hw->mac.type == e1000_82576 &&
3578 	     rd32(E1000_EECD) & E1000_EECD_PRES) ||
3579 	    (hw->mac.type >= e1000_i210 ||
3580 	     igb_get_flash_presence_i210(hw))) {
3581 		ret_val = igb_read_part_string(hw, part_str,
3582 					       E1000_PBANUM_LENGTH);
3583 	} else {
3584 		ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
3585 	}
3586 
3587 	if (ret_val)
3588 		strcpy(part_str, "Unknown");
3589 	dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
3590 	dev_info(&pdev->dev,
3591 		"Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
3592 		(adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
3593 		(adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
3594 		adapter->num_rx_queues, adapter->num_tx_queues);
3595 	if (hw->phy.media_type == e1000_media_type_copper) {
3596 		switch (hw->mac.type) {
3597 		case e1000_i350:
3598 		case e1000_i210:
3599 		case e1000_i211:
3600 			/* Enable EEE for internal copper PHY devices */
3601 			err = igb_set_eee_i350(hw, true, true);
3602 			if ((!err) &&
3603 			    (!hw->dev_spec._82575.eee_disable)) {
3604 				adapter->eee_advert =
3605 					MDIO_EEE_100TX | MDIO_EEE_1000T;
3606 				adapter->flags |= IGB_FLAG_EEE;
3607 			}
3608 			break;
3609 		case e1000_i354:
3610 			if ((rd32(E1000_CTRL_EXT) &
3611 			    E1000_CTRL_EXT_LINK_MODE_SGMII)) {
3612 				err = igb_set_eee_i354(hw, true, true);
3613 				if ((!err) &&
3614 					(!hw->dev_spec._82575.eee_disable)) {
3615 					adapter->eee_advert =
3616 					   MDIO_EEE_100TX | MDIO_EEE_1000T;
3617 					adapter->flags |= IGB_FLAG_EEE;
3618 				}
3619 			}
3620 			break;
3621 		default:
3622 			break;
3623 		}
3624 	}
3625 
3626 	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_NO_DIRECT_COMPLETE);
3627 
3628 	pm_runtime_put_noidle(&pdev->dev);
3629 	return 0;
3630 
3631 err_register:
3632 	igb_release_hw_control(adapter);
3633 	memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
3634 err_eeprom:
3635 	if (!igb_check_reset_block(hw))
3636 		igb_reset_phy(hw);
3637 
3638 	if (hw->flash_address)
3639 		iounmap(hw->flash_address);
3640 err_sw_init:
3641 	kfree(adapter->mac_table);
3642 	kfree(adapter->shadow_vfta);
3643 	igb_clear_interrupt_scheme(adapter);
3644 #ifdef CONFIG_PCI_IOV
3645 	igb_disable_sriov(pdev);
3646 #endif
3647 	pci_iounmap(pdev, adapter->io_addr);
3648 err_ioremap:
3649 	free_netdev(netdev);
3650 err_alloc_etherdev:
3651 	pci_disable_pcie_error_reporting(pdev);
3652 	pci_release_mem_regions(pdev);
3653 err_pci_reg:
3654 err_dma:
3655 	pci_disable_device(pdev);
3656 	return err;
3657 }
3658 
3659 #ifdef CONFIG_PCI_IOV
3660 static int igb_disable_sriov(struct pci_dev *pdev)
3661 {
3662 	struct net_device *netdev = pci_get_drvdata(pdev);
3663 	struct igb_adapter *adapter = netdev_priv(netdev);
3664 	struct e1000_hw *hw = &adapter->hw;
3665 	unsigned long flags;
3666 
3667 	/* reclaim resources allocated to VFs */
3668 	if (adapter->vf_data) {
3669 		/* disable iov and allow time for transactions to clear */
3670 		if (pci_vfs_assigned(pdev)) {
3671 			dev_warn(&pdev->dev,
3672 				 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
3673 			return -EPERM;
3674 		} else {
3675 			pci_disable_sriov(pdev);
3676 			msleep(500);
3677 		}
3678 		spin_lock_irqsave(&adapter->vfs_lock, flags);
3679 		kfree(adapter->vf_mac_list);
3680 		adapter->vf_mac_list = NULL;
3681 		kfree(adapter->vf_data);
3682 		adapter->vf_data = NULL;
3683 		adapter->vfs_allocated_count = 0;
3684 		spin_unlock_irqrestore(&adapter->vfs_lock, flags);
3685 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
3686 		wrfl();
3687 		msleep(100);
3688 		dev_info(&pdev->dev, "IOV Disabled\n");
3689 
3690 		/* Re-enable DMA Coalescing flag since IOV is turned off */
3691 		adapter->flags |= IGB_FLAG_DMAC;
3692 	}
3693 
3694 	return 0;
3695 }
3696 
3697 static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
3698 {
3699 	struct net_device *netdev = pci_get_drvdata(pdev);
3700 	struct igb_adapter *adapter = netdev_priv(netdev);
3701 	int old_vfs = pci_num_vf(pdev);
3702 	struct vf_mac_filter *mac_list;
3703 	int err = 0;
3704 	int num_vf_mac_filters, i;
3705 
3706 	if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
3707 		err = -EPERM;
3708 		goto out;
3709 	}
3710 	if (!num_vfs)
3711 		goto out;
3712 
3713 	if (old_vfs) {
3714 		dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
3715 			 old_vfs, max_vfs);
3716 		adapter->vfs_allocated_count = old_vfs;
3717 	} else
3718 		adapter->vfs_allocated_count = num_vfs;
3719 
3720 	adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
3721 				sizeof(struct vf_data_storage), GFP_KERNEL);
3722 
3723 	/* if allocation failed then we do not support SR-IOV */
3724 	if (!adapter->vf_data) {
3725 		adapter->vfs_allocated_count = 0;
3726 		err = -ENOMEM;
3727 		goto out;
3728 	}
3729 
3730 	/* Due to the limited number of RAR entries calculate potential
3731 	 * number of MAC filters available for the VFs. Reserve entries
3732 	 * for PF default MAC, PF MAC filters and at least one RAR entry
3733 	 * for each VF for VF MAC.
3734 	 */
3735 	num_vf_mac_filters = adapter->hw.mac.rar_entry_count -
3736 			     (1 + IGB_PF_MAC_FILTERS_RESERVED +
3737 			      adapter->vfs_allocated_count);
3738 
3739 	adapter->vf_mac_list = kcalloc(num_vf_mac_filters,
3740 				       sizeof(struct vf_mac_filter),
3741 				       GFP_KERNEL);
3742 
3743 	mac_list = adapter->vf_mac_list;
3744 	INIT_LIST_HEAD(&adapter->vf_macs.l);
3745 
3746 	if (adapter->vf_mac_list) {
3747 		/* Initialize list of VF MAC filters */
3748 		for (i = 0; i < num_vf_mac_filters; i++) {
3749 			mac_list->vf = -1;
3750 			mac_list->free = true;
3751 			list_add(&mac_list->l, &adapter->vf_macs.l);
3752 			mac_list++;
3753 		}
3754 	} else {
3755 		/* If we could not allocate memory for the VF MAC filters
3756 		 * we can continue without this feature but warn user.
3757 		 */
3758 		dev_err(&pdev->dev,
3759 			"Unable to allocate memory for VF MAC filter list\n");
3760 	}
3761 
3762 	/* only call pci_enable_sriov() if no VFs are allocated already */
3763 	if (!old_vfs) {
3764 		err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
3765 		if (err)
3766 			goto err_out;
3767 	}
3768 	dev_info(&pdev->dev, "%d VFs allocated\n",
3769 		 adapter->vfs_allocated_count);
3770 	for (i = 0; i < adapter->vfs_allocated_count; i++)
3771 		igb_vf_configure(adapter, i);
3772 
3773 	/* DMA Coalescing is not supported in IOV mode. */
3774 	adapter->flags &= ~IGB_FLAG_DMAC;
3775 	goto out;
3776 
3777 err_out:
3778 	kfree(adapter->vf_mac_list);
3779 	adapter->vf_mac_list = NULL;
3780 	kfree(adapter->vf_data);
3781 	adapter->vf_data = NULL;
3782 	adapter->vfs_allocated_count = 0;
3783 out:
3784 	return err;
3785 }
3786 
3787 #endif
3788 /**
3789  *  igb_remove_i2c - Cleanup  I2C interface
3790  *  @adapter: pointer to adapter structure
3791  **/
3792 static void igb_remove_i2c(struct igb_adapter *adapter)
3793 {
3794 	/* free the adapter bus structure */
3795 	i2c_del_adapter(&adapter->i2c_adap);
3796 }
3797 
3798 /**
3799  *  igb_remove - Device Removal Routine
3800  *  @pdev: PCI device information struct
3801  *
3802  *  igb_remove is called by the PCI subsystem to alert the driver
3803  *  that it should release a PCI device.  The could be caused by a
3804  *  Hot-Plug event, or because the driver is going to be removed from
3805  *  memory.
3806  **/
3807 static void igb_remove(struct pci_dev *pdev)
3808 {
3809 	struct net_device *netdev = pci_get_drvdata(pdev);
3810 	struct igb_adapter *adapter = netdev_priv(netdev);
3811 	struct e1000_hw *hw = &adapter->hw;
3812 
3813 	pm_runtime_get_noresume(&pdev->dev);
3814 #ifdef CONFIG_IGB_HWMON
3815 	igb_sysfs_exit(adapter);
3816 #endif
3817 	igb_remove_i2c(adapter);
3818 	igb_ptp_stop(adapter);
3819 	/* The watchdog timer may be rescheduled, so explicitly
3820 	 * disable watchdog from being rescheduled.
3821 	 */
3822 	set_bit(__IGB_DOWN, &adapter->state);
3823 	del_timer_sync(&adapter->watchdog_timer);
3824 	del_timer_sync(&adapter->phy_info_timer);
3825 
3826 	cancel_work_sync(&adapter->reset_task);
3827 	cancel_work_sync(&adapter->watchdog_task);
3828 
3829 #ifdef CONFIG_IGB_DCA
3830 	if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
3831 		dev_info(&pdev->dev, "DCA disabled\n");
3832 		dca_remove_requester(&pdev->dev);
3833 		adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
3834 		wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
3835 	}
3836 #endif
3837 
3838 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
3839 	 * would have already happened in close and is redundant.
3840 	 */
3841 	igb_release_hw_control(adapter);
3842 
3843 #ifdef CONFIG_PCI_IOV
3844 	rtnl_lock();
3845 	igb_disable_sriov(pdev);
3846 	rtnl_unlock();
3847 #endif
3848 
3849 	unregister_netdev(netdev);
3850 
3851 	igb_clear_interrupt_scheme(adapter);
3852 
3853 	pci_iounmap(pdev, adapter->io_addr);
3854 	if (hw->flash_address)
3855 		iounmap(hw->flash_address);
3856 	pci_release_mem_regions(pdev);
3857 
3858 	kfree(adapter->mac_table);
3859 	kfree(adapter->shadow_vfta);
3860 	free_netdev(netdev);
3861 
3862 	pci_disable_pcie_error_reporting(pdev);
3863 
3864 	pci_disable_device(pdev);
3865 }
3866 
3867 /**
3868  *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
3869  *  @adapter: board private structure to initialize
3870  *
3871  *  This function initializes the vf specific data storage and then attempts to
3872  *  allocate the VFs.  The reason for ordering it this way is because it is much
3873  *  mor expensive time wise to disable SR-IOV than it is to allocate and free
3874  *  the memory for the VFs.
3875  **/
3876 static void igb_probe_vfs(struct igb_adapter *adapter)
3877 {
3878 #ifdef CONFIG_PCI_IOV
3879 	struct pci_dev *pdev = adapter->pdev;
3880 	struct e1000_hw *hw = &adapter->hw;
3881 
3882 	/* Virtualization features not supported on i210 family. */
3883 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
3884 		return;
3885 
3886 	/* Of the below we really only want the effect of getting
3887 	 * IGB_FLAG_HAS_MSIX set (if available), without which
3888 	 * igb_enable_sriov() has no effect.
3889 	 */
3890 	igb_set_interrupt_capability(adapter, true);
3891 	igb_reset_interrupt_capability(adapter);
3892 
3893 	pci_sriov_set_totalvfs(pdev, 7);
3894 	igb_enable_sriov(pdev, max_vfs);
3895 
3896 #endif /* CONFIG_PCI_IOV */
3897 }
3898 
3899 unsigned int igb_get_max_rss_queues(struct igb_adapter *adapter)
3900 {
3901 	struct e1000_hw *hw = &adapter->hw;
3902 	unsigned int max_rss_queues;
3903 
3904 	/* Determine the maximum number of RSS queues supported. */
3905 	switch (hw->mac.type) {
3906 	case e1000_i211:
3907 		max_rss_queues = IGB_MAX_RX_QUEUES_I211;
3908 		break;
3909 	case e1000_82575:
3910 	case e1000_i210:
3911 		max_rss_queues = IGB_MAX_RX_QUEUES_82575;
3912 		break;
3913 	case e1000_i350:
3914 		/* I350 cannot do RSS and SR-IOV at the same time */
3915 		if (!!adapter->vfs_allocated_count) {
3916 			max_rss_queues = 1;
3917 			break;
3918 		}
3919 		fallthrough;
3920 	case e1000_82576:
3921 		if (!!adapter->vfs_allocated_count) {
3922 			max_rss_queues = 2;
3923 			break;
3924 		}
3925 		fallthrough;
3926 	case e1000_82580:
3927 	case e1000_i354:
3928 	default:
3929 		max_rss_queues = IGB_MAX_RX_QUEUES;
3930 		break;
3931 	}
3932 
3933 	return max_rss_queues;
3934 }
3935 
3936 static void igb_init_queue_configuration(struct igb_adapter *adapter)
3937 {
3938 	u32 max_rss_queues;
3939 
3940 	max_rss_queues = igb_get_max_rss_queues(adapter);
3941 	adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
3942 
3943 	igb_set_flag_queue_pairs(adapter, max_rss_queues);
3944 }
3945 
3946 void igb_set_flag_queue_pairs(struct igb_adapter *adapter,
3947 			      const u32 max_rss_queues)
3948 {
3949 	struct e1000_hw *hw = &adapter->hw;
3950 
3951 	/* Determine if we need to pair queues. */
3952 	switch (hw->mac.type) {
3953 	case e1000_82575:
3954 	case e1000_i211:
3955 		/* Device supports enough interrupts without queue pairing. */
3956 		break;
3957 	case e1000_82576:
3958 	case e1000_82580:
3959 	case e1000_i350:
3960 	case e1000_i354:
3961 	case e1000_i210:
3962 	default:
3963 		/* If rss_queues > half of max_rss_queues, pair the queues in
3964 		 * order to conserve interrupts due to limited supply.
3965 		 */
3966 		if (adapter->rss_queues > (max_rss_queues / 2))
3967 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
3968 		else
3969 			adapter->flags &= ~IGB_FLAG_QUEUE_PAIRS;
3970 		break;
3971 	}
3972 }
3973 
3974 /**
3975  *  igb_sw_init - Initialize general software structures (struct igb_adapter)
3976  *  @adapter: board private structure to initialize
3977  *
3978  *  igb_sw_init initializes the Adapter private data structure.
3979  *  Fields are initialized based on PCI device information and
3980  *  OS network device settings (MTU size).
3981  **/
3982 static int igb_sw_init(struct igb_adapter *adapter)
3983 {
3984 	struct e1000_hw *hw = &adapter->hw;
3985 	struct net_device *netdev = adapter->netdev;
3986 	struct pci_dev *pdev = adapter->pdev;
3987 
3988 	pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
3989 
3990 	/* set default ring sizes */
3991 	adapter->tx_ring_count = IGB_DEFAULT_TXD;
3992 	adapter->rx_ring_count = IGB_DEFAULT_RXD;
3993 
3994 	/* set default ITR values */
3995 	adapter->rx_itr_setting = IGB_DEFAULT_ITR;
3996 	adapter->tx_itr_setting = IGB_DEFAULT_ITR;
3997 
3998 	/* set default work limits */
3999 	adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
4000 
4001 	adapter->max_frame_size = netdev->mtu + IGB_ETH_PKT_HDR_PAD;
4002 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4003 
4004 	spin_lock_init(&adapter->nfc_lock);
4005 	spin_lock_init(&adapter->stats64_lock);
4006 
4007 	/* init spinlock to avoid concurrency of VF resources */
4008 	spin_lock_init(&adapter->vfs_lock);
4009 #ifdef CONFIG_PCI_IOV
4010 	switch (hw->mac.type) {
4011 	case e1000_82576:
4012 	case e1000_i350:
4013 		if (max_vfs > 7) {
4014 			dev_warn(&pdev->dev,
4015 				 "Maximum of 7 VFs per PF, using max\n");
4016 			max_vfs = adapter->vfs_allocated_count = 7;
4017 		} else
4018 			adapter->vfs_allocated_count = max_vfs;
4019 		if (adapter->vfs_allocated_count)
4020 			dev_warn(&pdev->dev,
4021 				 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
4022 		break;
4023 	default:
4024 		break;
4025 	}
4026 #endif /* CONFIG_PCI_IOV */
4027 
4028 	/* Assume MSI-X interrupts, will be checked during IRQ allocation */
4029 	adapter->flags |= IGB_FLAG_HAS_MSIX;
4030 
4031 	adapter->mac_table = kcalloc(hw->mac.rar_entry_count,
4032 				     sizeof(struct igb_mac_addr),
4033 				     GFP_KERNEL);
4034 	if (!adapter->mac_table)
4035 		return -ENOMEM;
4036 
4037 	igb_probe_vfs(adapter);
4038 
4039 	igb_init_queue_configuration(adapter);
4040 
4041 	/* Setup and initialize a copy of the hw vlan table array */
4042 	adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
4043 				       GFP_KERNEL);
4044 	if (!adapter->shadow_vfta)
4045 		return -ENOMEM;
4046 
4047 	/* This call may decrease the number of queues */
4048 	if (igb_init_interrupt_scheme(adapter, true)) {
4049 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
4050 		return -ENOMEM;
4051 	}
4052 
4053 	/* Explicitly disable IRQ since the NIC can be in any state. */
4054 	igb_irq_disable(adapter);
4055 
4056 	if (hw->mac.type >= e1000_i350)
4057 		adapter->flags &= ~IGB_FLAG_DMAC;
4058 
4059 	set_bit(__IGB_DOWN, &adapter->state);
4060 	return 0;
4061 }
4062 
4063 /**
4064  *  __igb_open - Called when a network interface is made active
4065  *  @netdev: network interface device structure
4066  *  @resuming: indicates whether we are in a resume call
4067  *
4068  *  Returns 0 on success, negative value on failure
4069  *
4070  *  The open entry point is called when a network interface is made
4071  *  active by the system (IFF_UP).  At this point all resources needed
4072  *  for transmit and receive operations are allocated, the interrupt
4073  *  handler is registered with the OS, the watchdog timer is started,
4074  *  and the stack is notified that the interface is ready.
4075  **/
4076 static int __igb_open(struct net_device *netdev, bool resuming)
4077 {
4078 	struct igb_adapter *adapter = netdev_priv(netdev);
4079 	struct e1000_hw *hw = &adapter->hw;
4080 	struct pci_dev *pdev = adapter->pdev;
4081 	int err;
4082 	int i;
4083 
4084 	/* disallow open during test */
4085 	if (test_bit(__IGB_TESTING, &adapter->state)) {
4086 		WARN_ON(resuming);
4087 		return -EBUSY;
4088 	}
4089 
4090 	if (!resuming)
4091 		pm_runtime_get_sync(&pdev->dev);
4092 
4093 	netif_carrier_off(netdev);
4094 
4095 	/* allocate transmit descriptors */
4096 	err = igb_setup_all_tx_resources(adapter);
4097 	if (err)
4098 		goto err_setup_tx;
4099 
4100 	/* allocate receive descriptors */
4101 	err = igb_setup_all_rx_resources(adapter);
4102 	if (err)
4103 		goto err_setup_rx;
4104 
4105 	igb_power_up_link(adapter);
4106 
4107 	/* before we allocate an interrupt, we must be ready to handle it.
4108 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4109 	 * as soon as we call pci_request_irq, so we have to setup our
4110 	 * clean_rx handler before we do so.
4111 	 */
4112 	igb_configure(adapter);
4113 
4114 	err = igb_request_irq(adapter);
4115 	if (err)
4116 		goto err_req_irq;
4117 
4118 	/* Notify the stack of the actual queue counts. */
4119 	err = netif_set_real_num_tx_queues(adapter->netdev,
4120 					   adapter->num_tx_queues);
4121 	if (err)
4122 		goto err_set_queues;
4123 
4124 	err = netif_set_real_num_rx_queues(adapter->netdev,
4125 					   adapter->num_rx_queues);
4126 	if (err)
4127 		goto err_set_queues;
4128 
4129 	/* From here on the code is the same as igb_up() */
4130 	clear_bit(__IGB_DOWN, &adapter->state);
4131 
4132 	for (i = 0; i < adapter->num_q_vectors; i++)
4133 		napi_enable(&(adapter->q_vector[i]->napi));
4134 
4135 	/* Clear any pending interrupts. */
4136 	rd32(E1000_TSICR);
4137 	rd32(E1000_ICR);
4138 
4139 	igb_irq_enable(adapter);
4140 
4141 	/* notify VFs that reset has been completed */
4142 	if (adapter->vfs_allocated_count) {
4143 		u32 reg_data = rd32(E1000_CTRL_EXT);
4144 
4145 		reg_data |= E1000_CTRL_EXT_PFRSTD;
4146 		wr32(E1000_CTRL_EXT, reg_data);
4147 	}
4148 
4149 	netif_tx_start_all_queues(netdev);
4150 
4151 	if (!resuming)
4152 		pm_runtime_put(&pdev->dev);
4153 
4154 	/* start the watchdog. */
4155 	hw->mac.get_link_status = 1;
4156 	schedule_work(&adapter->watchdog_task);
4157 
4158 	return 0;
4159 
4160 err_set_queues:
4161 	igb_free_irq(adapter);
4162 err_req_irq:
4163 	igb_release_hw_control(adapter);
4164 	igb_power_down_link(adapter);
4165 	igb_free_all_rx_resources(adapter);
4166 err_setup_rx:
4167 	igb_free_all_tx_resources(adapter);
4168 err_setup_tx:
4169 	igb_reset(adapter);
4170 	if (!resuming)
4171 		pm_runtime_put(&pdev->dev);
4172 
4173 	return err;
4174 }
4175 
4176 int igb_open(struct net_device *netdev)
4177 {
4178 	return __igb_open(netdev, false);
4179 }
4180 
4181 /**
4182  *  __igb_close - Disables a network interface
4183  *  @netdev: network interface device structure
4184  *  @suspending: indicates we are in a suspend call
4185  *
4186  *  Returns 0, this is not allowed to fail
4187  *
4188  *  The close entry point is called when an interface is de-activated
4189  *  by the OS.  The hardware is still under the driver's control, but
4190  *  needs to be disabled.  A global MAC reset is issued to stop the
4191  *  hardware, and all transmit and receive resources are freed.
4192  **/
4193 static int __igb_close(struct net_device *netdev, bool suspending)
4194 {
4195 	struct igb_adapter *adapter = netdev_priv(netdev);
4196 	struct pci_dev *pdev = adapter->pdev;
4197 
4198 	WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
4199 
4200 	if (!suspending)
4201 		pm_runtime_get_sync(&pdev->dev);
4202 
4203 	igb_down(adapter);
4204 	igb_free_irq(adapter);
4205 
4206 	igb_free_all_tx_resources(adapter);
4207 	igb_free_all_rx_resources(adapter);
4208 
4209 	if (!suspending)
4210 		pm_runtime_put_sync(&pdev->dev);
4211 	return 0;
4212 }
4213 
4214 int igb_close(struct net_device *netdev)
4215 {
4216 	if (netif_device_present(netdev) || netdev->dismantle)
4217 		return __igb_close(netdev, false);
4218 	return 0;
4219 }
4220 
4221 /**
4222  *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
4223  *  @tx_ring: tx descriptor ring (for a specific queue) to setup
4224  *
4225  *  Return 0 on success, negative on failure
4226  **/
4227 int igb_setup_tx_resources(struct igb_ring *tx_ring)
4228 {
4229 	struct device *dev = tx_ring->dev;
4230 	int size;
4231 
4232 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
4233 
4234 	tx_ring->tx_buffer_info = vmalloc(size);
4235 	if (!tx_ring->tx_buffer_info)
4236 		goto err;
4237 
4238 	/* round up to nearest 4K */
4239 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
4240 	tx_ring->size = ALIGN(tx_ring->size, 4096);
4241 
4242 	tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4243 					   &tx_ring->dma, GFP_KERNEL);
4244 	if (!tx_ring->desc)
4245 		goto err;
4246 
4247 	tx_ring->next_to_use = 0;
4248 	tx_ring->next_to_clean = 0;
4249 
4250 	return 0;
4251 
4252 err:
4253 	vfree(tx_ring->tx_buffer_info);
4254 	tx_ring->tx_buffer_info = NULL;
4255 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4256 	return -ENOMEM;
4257 }
4258 
4259 /**
4260  *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
4261  *				 (Descriptors) for all queues
4262  *  @adapter: board private structure
4263  *
4264  *  Return 0 on success, negative on failure
4265  **/
4266 static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
4267 {
4268 	struct pci_dev *pdev = adapter->pdev;
4269 	int i, err = 0;
4270 
4271 	for (i = 0; i < adapter->num_tx_queues; i++) {
4272 		err = igb_setup_tx_resources(adapter->tx_ring[i]);
4273 		if (err) {
4274 			dev_err(&pdev->dev,
4275 				"Allocation for Tx Queue %u failed\n", i);
4276 			for (i--; i >= 0; i--)
4277 				igb_free_tx_resources(adapter->tx_ring[i]);
4278 			break;
4279 		}
4280 	}
4281 
4282 	return err;
4283 }
4284 
4285 /**
4286  *  igb_setup_tctl - configure the transmit control registers
4287  *  @adapter: Board private structure
4288  **/
4289 void igb_setup_tctl(struct igb_adapter *adapter)
4290 {
4291 	struct e1000_hw *hw = &adapter->hw;
4292 	u32 tctl;
4293 
4294 	/* disable queue 0 which is enabled by default on 82575 and 82576 */
4295 	wr32(E1000_TXDCTL(0), 0);
4296 
4297 	/* Program the Transmit Control Register */
4298 	tctl = rd32(E1000_TCTL);
4299 	tctl &= ~E1000_TCTL_CT;
4300 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
4301 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
4302 
4303 	igb_config_collision_dist(hw);
4304 
4305 	/* Enable transmits */
4306 	tctl |= E1000_TCTL_EN;
4307 
4308 	wr32(E1000_TCTL, tctl);
4309 }
4310 
4311 /**
4312  *  igb_configure_tx_ring - Configure transmit ring after Reset
4313  *  @adapter: board private structure
4314  *  @ring: tx ring to configure
4315  *
4316  *  Configure a transmit ring after a reset.
4317  **/
4318 void igb_configure_tx_ring(struct igb_adapter *adapter,
4319 			   struct igb_ring *ring)
4320 {
4321 	struct e1000_hw *hw = &adapter->hw;
4322 	u32 txdctl = 0;
4323 	u64 tdba = ring->dma;
4324 	int reg_idx = ring->reg_idx;
4325 
4326 	wr32(E1000_TDLEN(reg_idx),
4327 	     ring->count * sizeof(union e1000_adv_tx_desc));
4328 	wr32(E1000_TDBAL(reg_idx),
4329 	     tdba & 0x00000000ffffffffULL);
4330 	wr32(E1000_TDBAH(reg_idx), tdba >> 32);
4331 
4332 	ring->tail = adapter->io_addr + E1000_TDT(reg_idx);
4333 	wr32(E1000_TDH(reg_idx), 0);
4334 	writel(0, ring->tail);
4335 
4336 	txdctl |= IGB_TX_PTHRESH;
4337 	txdctl |= IGB_TX_HTHRESH << 8;
4338 	txdctl |= IGB_TX_WTHRESH << 16;
4339 
4340 	/* reinitialize tx_buffer_info */
4341 	memset(ring->tx_buffer_info, 0,
4342 	       sizeof(struct igb_tx_buffer) * ring->count);
4343 
4344 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
4345 	wr32(E1000_TXDCTL(reg_idx), txdctl);
4346 }
4347 
4348 /**
4349  *  igb_configure_tx - Configure transmit Unit after Reset
4350  *  @adapter: board private structure
4351  *
4352  *  Configure the Tx unit of the MAC after a reset.
4353  **/
4354 static void igb_configure_tx(struct igb_adapter *adapter)
4355 {
4356 	struct e1000_hw *hw = &adapter->hw;
4357 	int i;
4358 
4359 	/* disable the queues */
4360 	for (i = 0; i < adapter->num_tx_queues; i++)
4361 		wr32(E1000_TXDCTL(adapter->tx_ring[i]->reg_idx), 0);
4362 
4363 	wrfl();
4364 	usleep_range(10000, 20000);
4365 
4366 	for (i = 0; i < adapter->num_tx_queues; i++)
4367 		igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
4368 }
4369 
4370 /**
4371  *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
4372  *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
4373  *
4374  *  Returns 0 on success, negative on failure
4375  **/
4376 int igb_setup_rx_resources(struct igb_ring *rx_ring)
4377 {
4378 	struct igb_adapter *adapter = netdev_priv(rx_ring->netdev);
4379 	struct device *dev = rx_ring->dev;
4380 	int size, res;
4381 
4382 	/* XDP RX-queue info */
4383 	if (xdp_rxq_info_is_reg(&rx_ring->xdp_rxq))
4384 		xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4385 	res = xdp_rxq_info_reg(&rx_ring->xdp_rxq, rx_ring->netdev,
4386 			       rx_ring->queue_index, 0);
4387 	if (res < 0) {
4388 		dev_err(dev, "Failed to register xdp_rxq index %u\n",
4389 			rx_ring->queue_index);
4390 		return res;
4391 	}
4392 
4393 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
4394 
4395 	rx_ring->rx_buffer_info = vmalloc(size);
4396 	if (!rx_ring->rx_buffer_info)
4397 		goto err;
4398 
4399 	/* Round up to nearest 4K */
4400 	rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
4401 	rx_ring->size = ALIGN(rx_ring->size, 4096);
4402 
4403 	rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4404 					   &rx_ring->dma, GFP_KERNEL);
4405 	if (!rx_ring->desc)
4406 		goto err;
4407 
4408 	rx_ring->next_to_alloc = 0;
4409 	rx_ring->next_to_clean = 0;
4410 	rx_ring->next_to_use = 0;
4411 
4412 	rx_ring->xdp_prog = adapter->xdp_prog;
4413 
4414 	return 0;
4415 
4416 err:
4417 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4418 	vfree(rx_ring->rx_buffer_info);
4419 	rx_ring->rx_buffer_info = NULL;
4420 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4421 	return -ENOMEM;
4422 }
4423 
4424 /**
4425  *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
4426  *				 (Descriptors) for all queues
4427  *  @adapter: board private structure
4428  *
4429  *  Return 0 on success, negative on failure
4430  **/
4431 static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
4432 {
4433 	struct pci_dev *pdev = adapter->pdev;
4434 	int i, err = 0;
4435 
4436 	for (i = 0; i < adapter->num_rx_queues; i++) {
4437 		err = igb_setup_rx_resources(adapter->rx_ring[i]);
4438 		if (err) {
4439 			dev_err(&pdev->dev,
4440 				"Allocation for Rx Queue %u failed\n", i);
4441 			for (i--; i >= 0; i--)
4442 				igb_free_rx_resources(adapter->rx_ring[i]);
4443 			break;
4444 		}
4445 	}
4446 
4447 	return err;
4448 }
4449 
4450 /**
4451  *  igb_setup_mrqc - configure the multiple receive queue control registers
4452  *  @adapter: Board private structure
4453  **/
4454 static void igb_setup_mrqc(struct igb_adapter *adapter)
4455 {
4456 	struct e1000_hw *hw = &adapter->hw;
4457 	u32 mrqc, rxcsum;
4458 	u32 j, num_rx_queues;
4459 	u32 rss_key[10];
4460 
4461 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
4462 	for (j = 0; j < 10; j++)
4463 		wr32(E1000_RSSRK(j), rss_key[j]);
4464 
4465 	num_rx_queues = adapter->rss_queues;
4466 
4467 	switch (hw->mac.type) {
4468 	case e1000_82576:
4469 		/* 82576 supports 2 RSS queues for SR-IOV */
4470 		if (adapter->vfs_allocated_count)
4471 			num_rx_queues = 2;
4472 		break;
4473 	default:
4474 		break;
4475 	}
4476 
4477 	if (adapter->rss_indir_tbl_init != num_rx_queues) {
4478 		for (j = 0; j < IGB_RETA_SIZE; j++)
4479 			adapter->rss_indir_tbl[j] =
4480 			(j * num_rx_queues) / IGB_RETA_SIZE;
4481 		adapter->rss_indir_tbl_init = num_rx_queues;
4482 	}
4483 	igb_write_rss_indir_tbl(adapter);
4484 
4485 	/* Disable raw packet checksumming so that RSS hash is placed in
4486 	 * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
4487 	 * offloads as they are enabled by default
4488 	 */
4489 	rxcsum = rd32(E1000_RXCSUM);
4490 	rxcsum |= E1000_RXCSUM_PCSD;
4491 
4492 	if (adapter->hw.mac.type >= e1000_82576)
4493 		/* Enable Receive Checksum Offload for SCTP */
4494 		rxcsum |= E1000_RXCSUM_CRCOFL;
4495 
4496 	/* Don't need to set TUOFL or IPOFL, they default to 1 */
4497 	wr32(E1000_RXCSUM, rxcsum);
4498 
4499 	/* Generate RSS hash based on packet types, TCP/UDP
4500 	 * port numbers and/or IPv4/v6 src and dst addresses
4501 	 */
4502 	mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
4503 	       E1000_MRQC_RSS_FIELD_IPV4_TCP |
4504 	       E1000_MRQC_RSS_FIELD_IPV6 |
4505 	       E1000_MRQC_RSS_FIELD_IPV6_TCP |
4506 	       E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
4507 
4508 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
4509 		mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
4510 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
4511 		mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
4512 
4513 	/* If VMDq is enabled then we set the appropriate mode for that, else
4514 	 * we default to RSS so that an RSS hash is calculated per packet even
4515 	 * if we are only using one queue
4516 	 */
4517 	if (adapter->vfs_allocated_count) {
4518 		if (hw->mac.type > e1000_82575) {
4519 			/* Set the default pool for the PF's first queue */
4520 			u32 vtctl = rd32(E1000_VT_CTL);
4521 
4522 			vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
4523 				   E1000_VT_CTL_DISABLE_DEF_POOL);
4524 			vtctl |= adapter->vfs_allocated_count <<
4525 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
4526 			wr32(E1000_VT_CTL, vtctl);
4527 		}
4528 		if (adapter->rss_queues > 1)
4529 			mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_MQ;
4530 		else
4531 			mrqc |= E1000_MRQC_ENABLE_VMDQ;
4532 	} else {
4533 		mrqc |= E1000_MRQC_ENABLE_RSS_MQ;
4534 	}
4535 	igb_vmm_control(adapter);
4536 
4537 	wr32(E1000_MRQC, mrqc);
4538 }
4539 
4540 /**
4541  *  igb_setup_rctl - configure the receive control registers
4542  *  @adapter: Board private structure
4543  **/
4544 void igb_setup_rctl(struct igb_adapter *adapter)
4545 {
4546 	struct e1000_hw *hw = &adapter->hw;
4547 	u32 rctl;
4548 
4549 	rctl = rd32(E1000_RCTL);
4550 
4551 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
4552 	rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
4553 
4554 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
4555 		(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
4556 
4557 	/* enable stripping of CRC. It's unlikely this will break BMC
4558 	 * redirection as it did with e1000. Newer features require
4559 	 * that the HW strips the CRC.
4560 	 */
4561 	rctl |= E1000_RCTL_SECRC;
4562 
4563 	/* disable store bad packets and clear size bits. */
4564 	rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
4565 
4566 	/* enable LPE to allow for reception of jumbo frames */
4567 	rctl |= E1000_RCTL_LPE;
4568 
4569 	/* disable queue 0 to prevent tail write w/o re-config */
4570 	wr32(E1000_RXDCTL(0), 0);
4571 
4572 	/* Attention!!!  For SR-IOV PF driver operations you must enable
4573 	 * queue drop for all VF and PF queues to prevent head of line blocking
4574 	 * if an un-trusted VF does not provide descriptors to hardware.
4575 	 */
4576 	if (adapter->vfs_allocated_count) {
4577 		/* set all queue drop enable bits */
4578 		wr32(E1000_QDE, ALL_QUEUES);
4579 	}
4580 
4581 	/* This is useful for sniffing bad packets. */
4582 	if (adapter->netdev->features & NETIF_F_RXALL) {
4583 		/* UPE and MPE will be handled by normal PROMISC logic
4584 		 * in e1000e_set_rx_mode
4585 		 */
4586 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
4587 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
4588 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
4589 
4590 		rctl &= ~(E1000_RCTL_DPF | /* Allow filtered pause */
4591 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
4592 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
4593 		 * and that breaks VLANs.
4594 		 */
4595 	}
4596 
4597 	wr32(E1000_RCTL, rctl);
4598 }
4599 
4600 static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
4601 				   int vfn)
4602 {
4603 	struct e1000_hw *hw = &adapter->hw;
4604 	u32 vmolr;
4605 
4606 	if (size > MAX_JUMBO_FRAME_SIZE)
4607 		size = MAX_JUMBO_FRAME_SIZE;
4608 
4609 	vmolr = rd32(E1000_VMOLR(vfn));
4610 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
4611 	vmolr |= size | E1000_VMOLR_LPE;
4612 	wr32(E1000_VMOLR(vfn), vmolr);
4613 
4614 	return 0;
4615 }
4616 
4617 static inline void igb_set_vf_vlan_strip(struct igb_adapter *adapter,
4618 					 int vfn, bool enable)
4619 {
4620 	struct e1000_hw *hw = &adapter->hw;
4621 	u32 val, reg;
4622 
4623 	if (hw->mac.type < e1000_82576)
4624 		return;
4625 
4626 	if (hw->mac.type == e1000_i350)
4627 		reg = E1000_DVMOLR(vfn);
4628 	else
4629 		reg = E1000_VMOLR(vfn);
4630 
4631 	val = rd32(reg);
4632 	if (enable)
4633 		val |= E1000_VMOLR_STRVLAN;
4634 	else
4635 		val &= ~(E1000_VMOLR_STRVLAN);
4636 	wr32(reg, val);
4637 }
4638 
4639 static inline void igb_set_vmolr(struct igb_adapter *adapter,
4640 				 int vfn, bool aupe)
4641 {
4642 	struct e1000_hw *hw = &adapter->hw;
4643 	u32 vmolr;
4644 
4645 	/* This register exists only on 82576 and newer so if we are older then
4646 	 * we should exit and do nothing
4647 	 */
4648 	if (hw->mac.type < e1000_82576)
4649 		return;
4650 
4651 	vmolr = rd32(E1000_VMOLR(vfn));
4652 	if (aupe)
4653 		vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
4654 	else
4655 		vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
4656 
4657 	/* clear all bits that might not be set */
4658 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
4659 
4660 	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
4661 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
4662 	/* for VMDq only allow the VFs and pool 0 to accept broadcast and
4663 	 * multicast packets
4664 	 */
4665 	if (vfn <= adapter->vfs_allocated_count)
4666 		vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
4667 
4668 	wr32(E1000_VMOLR(vfn), vmolr);
4669 }
4670 
4671 /**
4672  *  igb_setup_srrctl - configure the split and replication receive control
4673  *                     registers
4674  *  @adapter: Board private structure
4675  *  @ring: receive ring to be configured
4676  **/
4677 void igb_setup_srrctl(struct igb_adapter *adapter, struct igb_ring *ring)
4678 {
4679 	struct e1000_hw *hw = &adapter->hw;
4680 	int reg_idx = ring->reg_idx;
4681 	u32 srrctl = 0;
4682 
4683 	srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
4684 	if (ring_uses_large_buffer(ring))
4685 		srrctl |= IGB_RXBUFFER_3072 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4686 	else
4687 		srrctl |= IGB_RXBUFFER_2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4688 	srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
4689 	if (hw->mac.type >= e1000_82580)
4690 		srrctl |= E1000_SRRCTL_TIMESTAMP;
4691 	/* Only set Drop Enable if VFs allocated, or we are supporting multiple
4692 	 * queues and rx flow control is disabled
4693 	 */
4694 	if (adapter->vfs_allocated_count ||
4695 	    (!(hw->fc.current_mode & e1000_fc_rx_pause) &&
4696 	     adapter->num_rx_queues > 1))
4697 		srrctl |= E1000_SRRCTL_DROP_EN;
4698 
4699 	wr32(E1000_SRRCTL(reg_idx), srrctl);
4700 }
4701 
4702 /**
4703  *  igb_configure_rx_ring - Configure a receive ring after Reset
4704  *  @adapter: board private structure
4705  *  @ring: receive ring to be configured
4706  *
4707  *  Configure the Rx unit of the MAC after a reset.
4708  **/
4709 void igb_configure_rx_ring(struct igb_adapter *adapter,
4710 			   struct igb_ring *ring)
4711 {
4712 	struct e1000_hw *hw = &adapter->hw;
4713 	union e1000_adv_rx_desc *rx_desc;
4714 	u64 rdba = ring->dma;
4715 	int reg_idx = ring->reg_idx;
4716 	u32 rxdctl = 0;
4717 
4718 	xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4719 	WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4720 					   MEM_TYPE_PAGE_SHARED, NULL));
4721 
4722 	/* disable the queue */
4723 	wr32(E1000_RXDCTL(reg_idx), 0);
4724 
4725 	/* Set DMA base address registers */
4726 	wr32(E1000_RDBAL(reg_idx),
4727 	     rdba & 0x00000000ffffffffULL);
4728 	wr32(E1000_RDBAH(reg_idx), rdba >> 32);
4729 	wr32(E1000_RDLEN(reg_idx),
4730 	     ring->count * sizeof(union e1000_adv_rx_desc));
4731 
4732 	/* initialize head and tail */
4733 	ring->tail = adapter->io_addr + E1000_RDT(reg_idx);
4734 	wr32(E1000_RDH(reg_idx), 0);
4735 	writel(0, ring->tail);
4736 
4737 	/* set descriptor configuration */
4738 	igb_setup_srrctl(adapter, ring);
4739 
4740 	/* set filtering for VMDQ pools */
4741 	igb_set_vmolr(adapter, reg_idx & 0x7, true);
4742 
4743 	rxdctl |= IGB_RX_PTHRESH;
4744 	rxdctl |= IGB_RX_HTHRESH << 8;
4745 	rxdctl |= IGB_RX_WTHRESH << 16;
4746 
4747 	/* initialize rx_buffer_info */
4748 	memset(ring->rx_buffer_info, 0,
4749 	       sizeof(struct igb_rx_buffer) * ring->count);
4750 
4751 	/* initialize Rx descriptor 0 */
4752 	rx_desc = IGB_RX_DESC(ring, 0);
4753 	rx_desc->wb.upper.length = 0;
4754 
4755 	/* enable receive descriptor fetching */
4756 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
4757 	wr32(E1000_RXDCTL(reg_idx), rxdctl);
4758 }
4759 
4760 static void igb_set_rx_buffer_len(struct igb_adapter *adapter,
4761 				  struct igb_ring *rx_ring)
4762 {
4763 	/* set build_skb and buffer size flags */
4764 	clear_ring_build_skb_enabled(rx_ring);
4765 	clear_ring_uses_large_buffer(rx_ring);
4766 
4767 	if (adapter->flags & IGB_FLAG_RX_LEGACY)
4768 		return;
4769 
4770 	set_ring_build_skb_enabled(rx_ring);
4771 
4772 #if (PAGE_SIZE < 8192)
4773 	if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4774 		return;
4775 
4776 	set_ring_uses_large_buffer(rx_ring);
4777 #endif
4778 }
4779 
4780 /**
4781  *  igb_configure_rx - Configure receive Unit after Reset
4782  *  @adapter: board private structure
4783  *
4784  *  Configure the Rx unit of the MAC after a reset.
4785  **/
4786 static void igb_configure_rx(struct igb_adapter *adapter)
4787 {
4788 	int i;
4789 
4790 	/* set the correct pool for the PF default MAC address in entry 0 */
4791 	igb_set_default_mac_filter(adapter);
4792 
4793 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
4794 	 * the Base and Length of the Rx Descriptor Ring
4795 	 */
4796 	for (i = 0; i < adapter->num_rx_queues; i++) {
4797 		struct igb_ring *rx_ring = adapter->rx_ring[i];
4798 
4799 		igb_set_rx_buffer_len(adapter, rx_ring);
4800 		igb_configure_rx_ring(adapter, rx_ring);
4801 	}
4802 }
4803 
4804 /**
4805  *  igb_free_tx_resources - Free Tx Resources per Queue
4806  *  @tx_ring: Tx descriptor ring for a specific queue
4807  *
4808  *  Free all transmit software resources
4809  **/
4810 void igb_free_tx_resources(struct igb_ring *tx_ring)
4811 {
4812 	igb_clean_tx_ring(tx_ring);
4813 
4814 	vfree(tx_ring->tx_buffer_info);
4815 	tx_ring->tx_buffer_info = NULL;
4816 
4817 	/* if not set, then don't free */
4818 	if (!tx_ring->desc)
4819 		return;
4820 
4821 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4822 			  tx_ring->desc, tx_ring->dma);
4823 
4824 	tx_ring->desc = NULL;
4825 }
4826 
4827 /**
4828  *  igb_free_all_tx_resources - Free Tx Resources for All Queues
4829  *  @adapter: board private structure
4830  *
4831  *  Free all transmit software resources
4832  **/
4833 static void igb_free_all_tx_resources(struct igb_adapter *adapter)
4834 {
4835 	int i;
4836 
4837 	for (i = 0; i < adapter->num_tx_queues; i++)
4838 		if (adapter->tx_ring[i])
4839 			igb_free_tx_resources(adapter->tx_ring[i]);
4840 }
4841 
4842 /**
4843  *  igb_clean_tx_ring - Free Tx Buffers
4844  *  @tx_ring: ring to be cleaned
4845  **/
4846 static void igb_clean_tx_ring(struct igb_ring *tx_ring)
4847 {
4848 	u16 i = tx_ring->next_to_clean;
4849 	struct igb_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
4850 
4851 	while (i != tx_ring->next_to_use) {
4852 		union e1000_adv_tx_desc *eop_desc, *tx_desc;
4853 
4854 		/* Free all the Tx ring sk_buffs or xdp frames */
4855 		if (tx_buffer->type == IGB_TYPE_SKB)
4856 			dev_kfree_skb_any(tx_buffer->skb);
4857 		else
4858 			xdp_return_frame(tx_buffer->xdpf);
4859 
4860 		/* unmap skb header data */
4861 		dma_unmap_single(tx_ring->dev,
4862 				 dma_unmap_addr(tx_buffer, dma),
4863 				 dma_unmap_len(tx_buffer, len),
4864 				 DMA_TO_DEVICE);
4865 
4866 		/* check for eop_desc to determine the end of the packet */
4867 		eop_desc = tx_buffer->next_to_watch;
4868 		tx_desc = IGB_TX_DESC(tx_ring, i);
4869 
4870 		/* unmap remaining buffers */
4871 		while (tx_desc != eop_desc) {
4872 			tx_buffer++;
4873 			tx_desc++;
4874 			i++;
4875 			if (unlikely(i == tx_ring->count)) {
4876 				i = 0;
4877 				tx_buffer = tx_ring->tx_buffer_info;
4878 				tx_desc = IGB_TX_DESC(tx_ring, 0);
4879 			}
4880 
4881 			/* unmap any remaining paged data */
4882 			if (dma_unmap_len(tx_buffer, len))
4883 				dma_unmap_page(tx_ring->dev,
4884 					       dma_unmap_addr(tx_buffer, dma),
4885 					       dma_unmap_len(tx_buffer, len),
4886 					       DMA_TO_DEVICE);
4887 		}
4888 
4889 		tx_buffer->next_to_watch = NULL;
4890 
4891 		/* move us one more past the eop_desc for start of next pkt */
4892 		tx_buffer++;
4893 		i++;
4894 		if (unlikely(i == tx_ring->count)) {
4895 			i = 0;
4896 			tx_buffer = tx_ring->tx_buffer_info;
4897 		}
4898 	}
4899 
4900 	/* reset BQL for queue */
4901 	netdev_tx_reset_queue(txring_txq(tx_ring));
4902 
4903 	/* reset next_to_use and next_to_clean */
4904 	tx_ring->next_to_use = 0;
4905 	tx_ring->next_to_clean = 0;
4906 }
4907 
4908 /**
4909  *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
4910  *  @adapter: board private structure
4911  **/
4912 static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
4913 {
4914 	int i;
4915 
4916 	for (i = 0; i < adapter->num_tx_queues; i++)
4917 		if (adapter->tx_ring[i])
4918 			igb_clean_tx_ring(adapter->tx_ring[i]);
4919 }
4920 
4921 /**
4922  *  igb_free_rx_resources - Free Rx Resources
4923  *  @rx_ring: ring to clean the resources from
4924  *
4925  *  Free all receive software resources
4926  **/
4927 void igb_free_rx_resources(struct igb_ring *rx_ring)
4928 {
4929 	igb_clean_rx_ring(rx_ring);
4930 
4931 	rx_ring->xdp_prog = NULL;
4932 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4933 	vfree(rx_ring->rx_buffer_info);
4934 	rx_ring->rx_buffer_info = NULL;
4935 
4936 	/* if not set, then don't free */
4937 	if (!rx_ring->desc)
4938 		return;
4939 
4940 	dma_free_coherent(rx_ring->dev, rx_ring->size,
4941 			  rx_ring->desc, rx_ring->dma);
4942 
4943 	rx_ring->desc = NULL;
4944 }
4945 
4946 /**
4947  *  igb_free_all_rx_resources - Free Rx Resources for All Queues
4948  *  @adapter: board private structure
4949  *
4950  *  Free all receive software resources
4951  **/
4952 static void igb_free_all_rx_resources(struct igb_adapter *adapter)
4953 {
4954 	int i;
4955 
4956 	for (i = 0; i < adapter->num_rx_queues; i++)
4957 		if (adapter->rx_ring[i])
4958 			igb_free_rx_resources(adapter->rx_ring[i]);
4959 }
4960 
4961 /**
4962  *  igb_clean_rx_ring - Free Rx Buffers per Queue
4963  *  @rx_ring: ring to free buffers from
4964  **/
4965 static void igb_clean_rx_ring(struct igb_ring *rx_ring)
4966 {
4967 	u16 i = rx_ring->next_to_clean;
4968 
4969 	dev_kfree_skb(rx_ring->skb);
4970 	rx_ring->skb = NULL;
4971 
4972 	/* Free all the Rx ring sk_buffs */
4973 	while (i != rx_ring->next_to_alloc) {
4974 		struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
4975 
4976 		/* Invalidate cache lines that may have been written to by
4977 		 * device so that we avoid corrupting memory.
4978 		 */
4979 		dma_sync_single_range_for_cpu(rx_ring->dev,
4980 					      buffer_info->dma,
4981 					      buffer_info->page_offset,
4982 					      igb_rx_bufsz(rx_ring),
4983 					      DMA_FROM_DEVICE);
4984 
4985 		/* free resources associated with mapping */
4986 		dma_unmap_page_attrs(rx_ring->dev,
4987 				     buffer_info->dma,
4988 				     igb_rx_pg_size(rx_ring),
4989 				     DMA_FROM_DEVICE,
4990 				     IGB_RX_DMA_ATTR);
4991 		__page_frag_cache_drain(buffer_info->page,
4992 					buffer_info->pagecnt_bias);
4993 
4994 		i++;
4995 		if (i == rx_ring->count)
4996 			i = 0;
4997 	}
4998 
4999 	rx_ring->next_to_alloc = 0;
5000 	rx_ring->next_to_clean = 0;
5001 	rx_ring->next_to_use = 0;
5002 }
5003 
5004 /**
5005  *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
5006  *  @adapter: board private structure
5007  **/
5008 static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
5009 {
5010 	int i;
5011 
5012 	for (i = 0; i < adapter->num_rx_queues; i++)
5013 		if (adapter->rx_ring[i])
5014 			igb_clean_rx_ring(adapter->rx_ring[i]);
5015 }
5016 
5017 /**
5018  *  igb_set_mac - Change the Ethernet Address of the NIC
5019  *  @netdev: network interface device structure
5020  *  @p: pointer to an address structure
5021  *
5022  *  Returns 0 on success, negative on failure
5023  **/
5024 static int igb_set_mac(struct net_device *netdev, void *p)
5025 {
5026 	struct igb_adapter *adapter = netdev_priv(netdev);
5027 	struct e1000_hw *hw = &adapter->hw;
5028 	struct sockaddr *addr = p;
5029 
5030 	if (!is_valid_ether_addr(addr->sa_data))
5031 		return -EADDRNOTAVAIL;
5032 
5033 	eth_hw_addr_set(netdev, addr->sa_data);
5034 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
5035 
5036 	/* set the correct pool for the new PF MAC address in entry 0 */
5037 	igb_set_default_mac_filter(adapter);
5038 
5039 	return 0;
5040 }
5041 
5042 /**
5043  *  igb_write_mc_addr_list - write multicast addresses to MTA
5044  *  @netdev: network interface device structure
5045  *
5046  *  Writes multicast address list to the MTA hash table.
5047  *  Returns: -ENOMEM on failure
5048  *           0 on no addresses written
5049  *           X on writing X addresses to MTA
5050  **/
5051 static int igb_write_mc_addr_list(struct net_device *netdev)
5052 {
5053 	struct igb_adapter *adapter = netdev_priv(netdev);
5054 	struct e1000_hw *hw = &adapter->hw;
5055 	struct netdev_hw_addr *ha;
5056 	u8  *mta_list;
5057 	int i;
5058 
5059 	if (netdev_mc_empty(netdev)) {
5060 		/* nothing to program, so clear mc list */
5061 		igb_update_mc_addr_list(hw, NULL, 0);
5062 		igb_restore_vf_multicasts(adapter);
5063 		return 0;
5064 	}
5065 
5066 	mta_list = kcalloc(netdev_mc_count(netdev), 6, GFP_ATOMIC);
5067 	if (!mta_list)
5068 		return -ENOMEM;
5069 
5070 	/* The shared function expects a packed array of only addresses. */
5071 	i = 0;
5072 	netdev_for_each_mc_addr(ha, netdev)
5073 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
5074 
5075 	igb_update_mc_addr_list(hw, mta_list, i);
5076 	kfree(mta_list);
5077 
5078 	return netdev_mc_count(netdev);
5079 }
5080 
5081 static int igb_vlan_promisc_enable(struct igb_adapter *adapter)
5082 {
5083 	struct e1000_hw *hw = &adapter->hw;
5084 	u32 i, pf_id;
5085 
5086 	switch (hw->mac.type) {
5087 	case e1000_i210:
5088 	case e1000_i211:
5089 	case e1000_i350:
5090 		/* VLAN filtering needed for VLAN prio filter */
5091 		if (adapter->netdev->features & NETIF_F_NTUPLE)
5092 			break;
5093 		fallthrough;
5094 	case e1000_82576:
5095 	case e1000_82580:
5096 	case e1000_i354:
5097 		/* VLAN filtering needed for pool filtering */
5098 		if (adapter->vfs_allocated_count)
5099 			break;
5100 		fallthrough;
5101 	default:
5102 		return 1;
5103 	}
5104 
5105 	/* We are already in VLAN promisc, nothing to do */
5106 	if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
5107 		return 0;
5108 
5109 	if (!adapter->vfs_allocated_count)
5110 		goto set_vfta;
5111 
5112 	/* Add PF to all active pools */
5113 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
5114 
5115 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
5116 		u32 vlvf = rd32(E1000_VLVF(i));
5117 
5118 		vlvf |= BIT(pf_id);
5119 		wr32(E1000_VLVF(i), vlvf);
5120 	}
5121 
5122 set_vfta:
5123 	/* Set all bits in the VLAN filter table array */
5124 	for (i = E1000_VLAN_FILTER_TBL_SIZE; i--;)
5125 		hw->mac.ops.write_vfta(hw, i, ~0U);
5126 
5127 	/* Set flag so we don't redo unnecessary work */
5128 	adapter->flags |= IGB_FLAG_VLAN_PROMISC;
5129 
5130 	return 0;
5131 }
5132 
5133 #define VFTA_BLOCK_SIZE 8
5134 static void igb_scrub_vfta(struct igb_adapter *adapter, u32 vfta_offset)
5135 {
5136 	struct e1000_hw *hw = &adapter->hw;
5137 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
5138 	u32 vid_start = vfta_offset * 32;
5139 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
5140 	u32 i, vid, word, bits, pf_id;
5141 
5142 	/* guarantee that we don't scrub out management VLAN */
5143 	vid = adapter->mng_vlan_id;
5144 	if (vid >= vid_start && vid < vid_end)
5145 		vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
5146 
5147 	if (!adapter->vfs_allocated_count)
5148 		goto set_vfta;
5149 
5150 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
5151 
5152 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
5153 		u32 vlvf = rd32(E1000_VLVF(i));
5154 
5155 		/* pull VLAN ID from VLVF */
5156 		vid = vlvf & VLAN_VID_MASK;
5157 
5158 		/* only concern ourselves with a certain range */
5159 		if (vid < vid_start || vid >= vid_end)
5160 			continue;
5161 
5162 		if (vlvf & E1000_VLVF_VLANID_ENABLE) {
5163 			/* record VLAN ID in VFTA */
5164 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
5165 
5166 			/* if PF is part of this then continue */
5167 			if (test_bit(vid, adapter->active_vlans))
5168 				continue;
5169 		}
5170 
5171 		/* remove PF from the pool */
5172 		bits = ~BIT(pf_id);
5173 		bits &= rd32(E1000_VLVF(i));
5174 		wr32(E1000_VLVF(i), bits);
5175 	}
5176 
5177 set_vfta:
5178 	/* extract values from active_vlans and write back to VFTA */
5179 	for (i = VFTA_BLOCK_SIZE; i--;) {
5180 		vid = (vfta_offset + i) * 32;
5181 		word = vid / BITS_PER_LONG;
5182 		bits = vid % BITS_PER_LONG;
5183 
5184 		vfta[i] |= adapter->active_vlans[word] >> bits;
5185 
5186 		hw->mac.ops.write_vfta(hw, vfta_offset + i, vfta[i]);
5187 	}
5188 }
5189 
5190 static void igb_vlan_promisc_disable(struct igb_adapter *adapter)
5191 {
5192 	u32 i;
5193 
5194 	/* We are not in VLAN promisc, nothing to do */
5195 	if (!(adapter->flags & IGB_FLAG_VLAN_PROMISC))
5196 		return;
5197 
5198 	/* Set flag so we don't redo unnecessary work */
5199 	adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
5200 
5201 	for (i = 0; i < E1000_VLAN_FILTER_TBL_SIZE; i += VFTA_BLOCK_SIZE)
5202 		igb_scrub_vfta(adapter, i);
5203 }
5204 
5205 /**
5206  *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
5207  *  @netdev: network interface device structure
5208  *
5209  *  The set_rx_mode entry point is called whenever the unicast or multicast
5210  *  address lists or the network interface flags are updated.  This routine is
5211  *  responsible for configuring the hardware for proper unicast, multicast,
5212  *  promiscuous mode, and all-multi behavior.
5213  **/
5214 static void igb_set_rx_mode(struct net_device *netdev)
5215 {
5216 	struct igb_adapter *adapter = netdev_priv(netdev);
5217 	struct e1000_hw *hw = &adapter->hw;
5218 	unsigned int vfn = adapter->vfs_allocated_count;
5219 	u32 rctl = 0, vmolr = 0, rlpml = MAX_JUMBO_FRAME_SIZE;
5220 	int count;
5221 
5222 	/* Check for Promiscuous and All Multicast modes */
5223 	if (netdev->flags & IFF_PROMISC) {
5224 		rctl |= E1000_RCTL_UPE | E1000_RCTL_MPE;
5225 		vmolr |= E1000_VMOLR_MPME;
5226 
5227 		/* enable use of UTA filter to force packets to default pool */
5228 		if (hw->mac.type == e1000_82576)
5229 			vmolr |= E1000_VMOLR_ROPE;
5230 	} else {
5231 		if (netdev->flags & IFF_ALLMULTI) {
5232 			rctl |= E1000_RCTL_MPE;
5233 			vmolr |= E1000_VMOLR_MPME;
5234 		} else {
5235 			/* Write addresses to the MTA, if the attempt fails
5236 			 * then we should just turn on promiscuous mode so
5237 			 * that we can at least receive multicast traffic
5238 			 */
5239 			count = igb_write_mc_addr_list(netdev);
5240 			if (count < 0) {
5241 				rctl |= E1000_RCTL_MPE;
5242 				vmolr |= E1000_VMOLR_MPME;
5243 			} else if (count) {
5244 				vmolr |= E1000_VMOLR_ROMPE;
5245 			}
5246 		}
5247 	}
5248 
5249 	/* Write addresses to available RAR registers, if there is not
5250 	 * sufficient space to store all the addresses then enable
5251 	 * unicast promiscuous mode
5252 	 */
5253 	if (__dev_uc_sync(netdev, igb_uc_sync, igb_uc_unsync)) {
5254 		rctl |= E1000_RCTL_UPE;
5255 		vmolr |= E1000_VMOLR_ROPE;
5256 	}
5257 
5258 	/* enable VLAN filtering by default */
5259 	rctl |= E1000_RCTL_VFE;
5260 
5261 	/* disable VLAN filtering for modes that require it */
5262 	if ((netdev->flags & IFF_PROMISC) ||
5263 	    (netdev->features & NETIF_F_RXALL)) {
5264 		/* if we fail to set all rules then just clear VFE */
5265 		if (igb_vlan_promisc_enable(adapter))
5266 			rctl &= ~E1000_RCTL_VFE;
5267 	} else {
5268 		igb_vlan_promisc_disable(adapter);
5269 	}
5270 
5271 	/* update state of unicast, multicast, and VLAN filtering modes */
5272 	rctl |= rd32(E1000_RCTL) & ~(E1000_RCTL_UPE | E1000_RCTL_MPE |
5273 				     E1000_RCTL_VFE);
5274 	wr32(E1000_RCTL, rctl);
5275 
5276 #if (PAGE_SIZE < 8192)
5277 	if (!adapter->vfs_allocated_count) {
5278 		if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5279 			rlpml = IGB_MAX_FRAME_BUILD_SKB;
5280 	}
5281 #endif
5282 	wr32(E1000_RLPML, rlpml);
5283 
5284 	/* In order to support SR-IOV and eventually VMDq it is necessary to set
5285 	 * the VMOLR to enable the appropriate modes.  Without this workaround
5286 	 * we will have issues with VLAN tag stripping not being done for frames
5287 	 * that are only arriving because we are the default pool
5288 	 */
5289 	if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
5290 		return;
5291 
5292 	/* set UTA to appropriate mode */
5293 	igb_set_uta(adapter, !!(vmolr & E1000_VMOLR_ROPE));
5294 
5295 	vmolr |= rd32(E1000_VMOLR(vfn)) &
5296 		 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
5297 
5298 	/* enable Rx jumbo frames, restrict as needed to support build_skb */
5299 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
5300 #if (PAGE_SIZE < 8192)
5301 	if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5302 		vmolr |= IGB_MAX_FRAME_BUILD_SKB;
5303 	else
5304 #endif
5305 		vmolr |= MAX_JUMBO_FRAME_SIZE;
5306 	vmolr |= E1000_VMOLR_LPE;
5307 
5308 	wr32(E1000_VMOLR(vfn), vmolr);
5309 
5310 	igb_restore_vf_multicasts(adapter);
5311 }
5312 
5313 static void igb_check_wvbr(struct igb_adapter *adapter)
5314 {
5315 	struct e1000_hw *hw = &adapter->hw;
5316 	u32 wvbr = 0;
5317 
5318 	switch (hw->mac.type) {
5319 	case e1000_82576:
5320 	case e1000_i350:
5321 		wvbr = rd32(E1000_WVBR);
5322 		if (!wvbr)
5323 			return;
5324 		break;
5325 	default:
5326 		break;
5327 	}
5328 
5329 	adapter->wvbr |= wvbr;
5330 }
5331 
5332 #define IGB_STAGGERED_QUEUE_OFFSET 8
5333 
5334 static void igb_spoof_check(struct igb_adapter *adapter)
5335 {
5336 	int j;
5337 
5338 	if (!adapter->wvbr)
5339 		return;
5340 
5341 	for (j = 0; j < adapter->vfs_allocated_count; j++) {
5342 		if (adapter->wvbr & BIT(j) ||
5343 		    adapter->wvbr & BIT(j + IGB_STAGGERED_QUEUE_OFFSET)) {
5344 			dev_warn(&adapter->pdev->dev,
5345 				"Spoof event(s) detected on VF %d\n", j);
5346 			adapter->wvbr &=
5347 				~(BIT(j) |
5348 				  BIT(j + IGB_STAGGERED_QUEUE_OFFSET));
5349 		}
5350 	}
5351 }
5352 
5353 /* Need to wait a few seconds after link up to get diagnostic information from
5354  * the phy
5355  */
5356 static void igb_update_phy_info(struct timer_list *t)
5357 {
5358 	struct igb_adapter *adapter = from_timer(adapter, t, phy_info_timer);
5359 	igb_get_phy_info(&adapter->hw);
5360 }
5361 
5362 /**
5363  *  igb_has_link - check shared code for link and determine up/down
5364  *  @adapter: pointer to driver private info
5365  **/
5366 bool igb_has_link(struct igb_adapter *adapter)
5367 {
5368 	struct e1000_hw *hw = &adapter->hw;
5369 	bool link_active = false;
5370 
5371 	/* get_link_status is set on LSC (link status) interrupt or
5372 	 * rx sequence error interrupt.  get_link_status will stay
5373 	 * false until the e1000_check_for_link establishes link
5374 	 * for copper adapters ONLY
5375 	 */
5376 	switch (hw->phy.media_type) {
5377 	case e1000_media_type_copper:
5378 		if (!hw->mac.get_link_status)
5379 			return true;
5380 		fallthrough;
5381 	case e1000_media_type_internal_serdes:
5382 		hw->mac.ops.check_for_link(hw);
5383 		link_active = !hw->mac.get_link_status;
5384 		break;
5385 	default:
5386 	case e1000_media_type_unknown:
5387 		break;
5388 	}
5389 
5390 	if (((hw->mac.type == e1000_i210) ||
5391 	     (hw->mac.type == e1000_i211)) &&
5392 	     (hw->phy.id == I210_I_PHY_ID)) {
5393 		if (!netif_carrier_ok(adapter->netdev)) {
5394 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5395 		} else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
5396 			adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
5397 			adapter->link_check_timeout = jiffies;
5398 		}
5399 	}
5400 
5401 	return link_active;
5402 }
5403 
5404 static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
5405 {
5406 	bool ret = false;
5407 	u32 ctrl_ext, thstat;
5408 
5409 	/* check for thermal sensor event on i350 copper only */
5410 	if (hw->mac.type == e1000_i350) {
5411 		thstat = rd32(E1000_THSTAT);
5412 		ctrl_ext = rd32(E1000_CTRL_EXT);
5413 
5414 		if ((hw->phy.media_type == e1000_media_type_copper) &&
5415 		    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
5416 			ret = !!(thstat & event);
5417 	}
5418 
5419 	return ret;
5420 }
5421 
5422 /**
5423  *  igb_check_lvmmc - check for malformed packets received
5424  *  and indicated in LVMMC register
5425  *  @adapter: pointer to adapter
5426  **/
5427 static void igb_check_lvmmc(struct igb_adapter *adapter)
5428 {
5429 	struct e1000_hw *hw = &adapter->hw;
5430 	u32 lvmmc;
5431 
5432 	lvmmc = rd32(E1000_LVMMC);
5433 	if (lvmmc) {
5434 		if (unlikely(net_ratelimit())) {
5435 			netdev_warn(adapter->netdev,
5436 				    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
5437 				    lvmmc);
5438 		}
5439 	}
5440 }
5441 
5442 /**
5443  *  igb_watchdog - Timer Call-back
5444  *  @t: pointer to timer_list containing our private info pointer
5445  **/
5446 static void igb_watchdog(struct timer_list *t)
5447 {
5448 	struct igb_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5449 	/* Do the rest outside of interrupt context */
5450 	schedule_work(&adapter->watchdog_task);
5451 }
5452 
5453 static void igb_watchdog_task(struct work_struct *work)
5454 {
5455 	struct igb_adapter *adapter = container_of(work,
5456 						   struct igb_adapter,
5457 						   watchdog_task);
5458 	struct e1000_hw *hw = &adapter->hw;
5459 	struct e1000_phy_info *phy = &hw->phy;
5460 	struct net_device *netdev = adapter->netdev;
5461 	u32 link;
5462 	int i;
5463 	u32 connsw;
5464 	u16 phy_data, retry_count = 20;
5465 
5466 	link = igb_has_link(adapter);
5467 
5468 	if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
5469 		if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
5470 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5471 		else
5472 			link = false;
5473 	}
5474 
5475 	/* Force link down if we have fiber to swap to */
5476 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5477 		if (hw->phy.media_type == e1000_media_type_copper) {
5478 			connsw = rd32(E1000_CONNSW);
5479 			if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
5480 				link = 0;
5481 		}
5482 	}
5483 	if (link) {
5484 		/* Perform a reset if the media type changed. */
5485 		if (hw->dev_spec._82575.media_changed) {
5486 			hw->dev_spec._82575.media_changed = false;
5487 			adapter->flags |= IGB_FLAG_MEDIA_RESET;
5488 			igb_reset(adapter);
5489 		}
5490 		/* Cancel scheduled suspend requests. */
5491 		pm_runtime_resume(netdev->dev.parent);
5492 
5493 		if (!netif_carrier_ok(netdev)) {
5494 			u32 ctrl;
5495 
5496 			hw->mac.ops.get_speed_and_duplex(hw,
5497 							 &adapter->link_speed,
5498 							 &adapter->link_duplex);
5499 
5500 			ctrl = rd32(E1000_CTRL);
5501 			/* Links status message must follow this format */
5502 			netdev_info(netdev,
5503 			       "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5504 			       netdev->name,
5505 			       adapter->link_speed,
5506 			       adapter->link_duplex == FULL_DUPLEX ?
5507 			       "Full" : "Half",
5508 			       (ctrl & E1000_CTRL_TFCE) &&
5509 			       (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
5510 			       (ctrl & E1000_CTRL_RFCE) ?  "RX" :
5511 			       (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
5512 
5513 			/* disable EEE if enabled */
5514 			if ((adapter->flags & IGB_FLAG_EEE) &&
5515 				(adapter->link_duplex == HALF_DUPLEX)) {
5516 				dev_info(&adapter->pdev->dev,
5517 				"EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
5518 				adapter->hw.dev_spec._82575.eee_disable = true;
5519 				adapter->flags &= ~IGB_FLAG_EEE;
5520 			}
5521 
5522 			/* check if SmartSpeed worked */
5523 			igb_check_downshift(hw);
5524 			if (phy->speed_downgraded)
5525 				netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
5526 
5527 			/* check for thermal sensor event */
5528 			if (igb_thermal_sensor_event(hw,
5529 			    E1000_THSTAT_LINK_THROTTLE))
5530 				netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
5531 
5532 			/* adjust timeout factor according to speed/duplex */
5533 			adapter->tx_timeout_factor = 1;
5534 			switch (adapter->link_speed) {
5535 			case SPEED_10:
5536 				adapter->tx_timeout_factor = 14;
5537 				break;
5538 			case SPEED_100:
5539 				/* maybe add some timeout factor ? */
5540 				break;
5541 			}
5542 
5543 			if (adapter->link_speed != SPEED_1000 ||
5544 			    !hw->phy.ops.read_reg)
5545 				goto no_wait;
5546 
5547 			/* wait for Remote receiver status OK */
5548 retry_read_status:
5549 			if (!igb_read_phy_reg(hw, PHY_1000T_STATUS,
5550 					      &phy_data)) {
5551 				if (!(phy_data & SR_1000T_REMOTE_RX_STATUS) &&
5552 				    retry_count) {
5553 					msleep(100);
5554 					retry_count--;
5555 					goto retry_read_status;
5556 				} else if (!retry_count) {
5557 					dev_err(&adapter->pdev->dev, "exceed max 2 second\n");
5558 				}
5559 			} else {
5560 				dev_err(&adapter->pdev->dev, "read 1000Base-T Status Reg\n");
5561 			}
5562 no_wait:
5563 			netif_carrier_on(netdev);
5564 
5565 			igb_ping_all_vfs(adapter);
5566 			igb_check_vf_rate_limit(adapter);
5567 
5568 			/* link state has changed, schedule phy info update */
5569 			if (!test_bit(__IGB_DOWN, &adapter->state))
5570 				mod_timer(&adapter->phy_info_timer,
5571 					  round_jiffies(jiffies + 2 * HZ));
5572 		}
5573 	} else {
5574 		if (netif_carrier_ok(netdev)) {
5575 			adapter->link_speed = 0;
5576 			adapter->link_duplex = 0;
5577 
5578 			/* check for thermal sensor event */
5579 			if (igb_thermal_sensor_event(hw,
5580 			    E1000_THSTAT_PWR_DOWN)) {
5581 				netdev_err(netdev, "The network adapter was stopped because it overheated\n");
5582 			}
5583 
5584 			/* Links status message must follow this format */
5585 			netdev_info(netdev, "igb: %s NIC Link is Down\n",
5586 			       netdev->name);
5587 			netif_carrier_off(netdev);
5588 
5589 			igb_ping_all_vfs(adapter);
5590 
5591 			/* link state has changed, schedule phy info update */
5592 			if (!test_bit(__IGB_DOWN, &adapter->state))
5593 				mod_timer(&adapter->phy_info_timer,
5594 					  round_jiffies(jiffies + 2 * HZ));
5595 
5596 			/* link is down, time to check for alternate media */
5597 			if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5598 				igb_check_swap_media(adapter);
5599 				if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5600 					schedule_work(&adapter->reset_task);
5601 					/* return immediately */
5602 					return;
5603 				}
5604 			}
5605 			pm_schedule_suspend(netdev->dev.parent,
5606 					    MSEC_PER_SEC * 5);
5607 
5608 		/* also check for alternate media here */
5609 		} else if (!netif_carrier_ok(netdev) &&
5610 			   (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
5611 			igb_check_swap_media(adapter);
5612 			if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5613 				schedule_work(&adapter->reset_task);
5614 				/* return immediately */
5615 				return;
5616 			}
5617 		}
5618 	}
5619 
5620 	spin_lock(&adapter->stats64_lock);
5621 	igb_update_stats(adapter);
5622 	spin_unlock(&adapter->stats64_lock);
5623 
5624 	for (i = 0; i < adapter->num_tx_queues; i++) {
5625 		struct igb_ring *tx_ring = adapter->tx_ring[i];
5626 		if (!netif_carrier_ok(netdev)) {
5627 			/* We've lost link, so the controller stops DMA,
5628 			 * but we've got queued Tx work that's never going
5629 			 * to get done, so reset controller to flush Tx.
5630 			 * (Do the reset outside of interrupt context).
5631 			 */
5632 			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
5633 				adapter->tx_timeout_count++;
5634 				schedule_work(&adapter->reset_task);
5635 				/* return immediately since reset is imminent */
5636 				return;
5637 			}
5638 		}
5639 
5640 		/* Force detection of hung controller every watchdog period */
5641 		set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
5642 	}
5643 
5644 	/* Cause software interrupt to ensure Rx ring is cleaned */
5645 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
5646 		u32 eics = 0;
5647 
5648 		for (i = 0; i < adapter->num_q_vectors; i++)
5649 			eics |= adapter->q_vector[i]->eims_value;
5650 		wr32(E1000_EICS, eics);
5651 	} else {
5652 		wr32(E1000_ICS, E1000_ICS_RXDMT0);
5653 	}
5654 
5655 	igb_spoof_check(adapter);
5656 	igb_ptp_rx_hang(adapter);
5657 	igb_ptp_tx_hang(adapter);
5658 
5659 	/* Check LVMMC register on i350/i354 only */
5660 	if ((adapter->hw.mac.type == e1000_i350) ||
5661 	    (adapter->hw.mac.type == e1000_i354))
5662 		igb_check_lvmmc(adapter);
5663 
5664 	/* Reset the timer */
5665 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
5666 		if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
5667 			mod_timer(&adapter->watchdog_timer,
5668 				  round_jiffies(jiffies +  HZ));
5669 		else
5670 			mod_timer(&adapter->watchdog_timer,
5671 				  round_jiffies(jiffies + 2 * HZ));
5672 	}
5673 }
5674 
5675 enum latency_range {
5676 	lowest_latency = 0,
5677 	low_latency = 1,
5678 	bulk_latency = 2,
5679 	latency_invalid = 255
5680 };
5681 
5682 /**
5683  *  igb_update_ring_itr - update the dynamic ITR value based on packet size
5684  *  @q_vector: pointer to q_vector
5685  *
5686  *  Stores a new ITR value based on strictly on packet size.  This
5687  *  algorithm is less sophisticated than that used in igb_update_itr,
5688  *  due to the difficulty of synchronizing statistics across multiple
5689  *  receive rings.  The divisors and thresholds used by this function
5690  *  were determined based on theoretical maximum wire speed and testing
5691  *  data, in order to minimize response time while increasing bulk
5692  *  throughput.
5693  *  This functionality is controlled by ethtool's coalescing settings.
5694  *  NOTE:  This function is called only when operating in a multiqueue
5695  *         receive environment.
5696  **/
5697 static void igb_update_ring_itr(struct igb_q_vector *q_vector)
5698 {
5699 	int new_val = q_vector->itr_val;
5700 	int avg_wire_size = 0;
5701 	struct igb_adapter *adapter = q_vector->adapter;
5702 	unsigned int packets;
5703 
5704 	/* For non-gigabit speeds, just fix the interrupt rate at 4000
5705 	 * ints/sec - ITR timer value of 120 ticks.
5706 	 */
5707 	if (adapter->link_speed != SPEED_1000) {
5708 		new_val = IGB_4K_ITR;
5709 		goto set_itr_val;
5710 	}
5711 
5712 	packets = q_vector->rx.total_packets;
5713 	if (packets)
5714 		avg_wire_size = q_vector->rx.total_bytes / packets;
5715 
5716 	packets = q_vector->tx.total_packets;
5717 	if (packets)
5718 		avg_wire_size = max_t(u32, avg_wire_size,
5719 				      q_vector->tx.total_bytes / packets);
5720 
5721 	/* if avg_wire_size isn't set no work was done */
5722 	if (!avg_wire_size)
5723 		goto clear_counts;
5724 
5725 	/* Add 24 bytes to size to account for CRC, preamble, and gap */
5726 	avg_wire_size += 24;
5727 
5728 	/* Don't starve jumbo frames */
5729 	avg_wire_size = min(avg_wire_size, 3000);
5730 
5731 	/* Give a little boost to mid-size frames */
5732 	if ((avg_wire_size > 300) && (avg_wire_size < 1200))
5733 		new_val = avg_wire_size / 3;
5734 	else
5735 		new_val = avg_wire_size / 2;
5736 
5737 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5738 	if (new_val < IGB_20K_ITR &&
5739 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5740 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5741 		new_val = IGB_20K_ITR;
5742 
5743 set_itr_val:
5744 	if (new_val != q_vector->itr_val) {
5745 		q_vector->itr_val = new_val;
5746 		q_vector->set_itr = 1;
5747 	}
5748 clear_counts:
5749 	q_vector->rx.total_bytes = 0;
5750 	q_vector->rx.total_packets = 0;
5751 	q_vector->tx.total_bytes = 0;
5752 	q_vector->tx.total_packets = 0;
5753 }
5754 
5755 /**
5756  *  igb_update_itr - update the dynamic ITR value based on statistics
5757  *  @q_vector: pointer to q_vector
5758  *  @ring_container: ring info to update the itr for
5759  *
5760  *  Stores a new ITR value based on packets and byte
5761  *  counts during the last interrupt.  The advantage of per interrupt
5762  *  computation is faster updates and more accurate ITR for the current
5763  *  traffic pattern.  Constants in this function were computed
5764  *  based on theoretical maximum wire speed and thresholds were set based
5765  *  on testing data as well as attempting to minimize response time
5766  *  while increasing bulk throughput.
5767  *  This functionality is controlled by ethtool's coalescing settings.
5768  *  NOTE:  These calculations are only valid when operating in a single-
5769  *         queue environment.
5770  **/
5771 static void igb_update_itr(struct igb_q_vector *q_vector,
5772 			   struct igb_ring_container *ring_container)
5773 {
5774 	unsigned int packets = ring_container->total_packets;
5775 	unsigned int bytes = ring_container->total_bytes;
5776 	u8 itrval = ring_container->itr;
5777 
5778 	/* no packets, exit with status unchanged */
5779 	if (packets == 0)
5780 		return;
5781 
5782 	switch (itrval) {
5783 	case lowest_latency:
5784 		/* handle TSO and jumbo frames */
5785 		if (bytes/packets > 8000)
5786 			itrval = bulk_latency;
5787 		else if ((packets < 5) && (bytes > 512))
5788 			itrval = low_latency;
5789 		break;
5790 	case low_latency:  /* 50 usec aka 20000 ints/s */
5791 		if (bytes > 10000) {
5792 			/* this if handles the TSO accounting */
5793 			if (bytes/packets > 8000)
5794 				itrval = bulk_latency;
5795 			else if ((packets < 10) || ((bytes/packets) > 1200))
5796 				itrval = bulk_latency;
5797 			else if ((packets > 35))
5798 				itrval = lowest_latency;
5799 		} else if (bytes/packets > 2000) {
5800 			itrval = bulk_latency;
5801 		} else if (packets <= 2 && bytes < 512) {
5802 			itrval = lowest_latency;
5803 		}
5804 		break;
5805 	case bulk_latency: /* 250 usec aka 4000 ints/s */
5806 		if (bytes > 25000) {
5807 			if (packets > 35)
5808 				itrval = low_latency;
5809 		} else if (bytes < 1500) {
5810 			itrval = low_latency;
5811 		}
5812 		break;
5813 	}
5814 
5815 	/* clear work counters since we have the values we need */
5816 	ring_container->total_bytes = 0;
5817 	ring_container->total_packets = 0;
5818 
5819 	/* write updated itr to ring container */
5820 	ring_container->itr = itrval;
5821 }
5822 
5823 static void igb_set_itr(struct igb_q_vector *q_vector)
5824 {
5825 	struct igb_adapter *adapter = q_vector->adapter;
5826 	u32 new_itr = q_vector->itr_val;
5827 	u8 current_itr = 0;
5828 
5829 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
5830 	if (adapter->link_speed != SPEED_1000) {
5831 		current_itr = 0;
5832 		new_itr = IGB_4K_ITR;
5833 		goto set_itr_now;
5834 	}
5835 
5836 	igb_update_itr(q_vector, &q_vector->tx);
5837 	igb_update_itr(q_vector, &q_vector->rx);
5838 
5839 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
5840 
5841 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5842 	if (current_itr == lowest_latency &&
5843 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5844 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5845 		current_itr = low_latency;
5846 
5847 	switch (current_itr) {
5848 	/* counts and packets in update_itr are dependent on these numbers */
5849 	case lowest_latency:
5850 		new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
5851 		break;
5852 	case low_latency:
5853 		new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
5854 		break;
5855 	case bulk_latency:
5856 		new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
5857 		break;
5858 	default:
5859 		break;
5860 	}
5861 
5862 set_itr_now:
5863 	if (new_itr != q_vector->itr_val) {
5864 		/* this attempts to bias the interrupt rate towards Bulk
5865 		 * by adding intermediate steps when interrupt rate is
5866 		 * increasing
5867 		 */
5868 		new_itr = new_itr > q_vector->itr_val ?
5869 			  max((new_itr * q_vector->itr_val) /
5870 			  (new_itr + (q_vector->itr_val >> 2)),
5871 			  new_itr) : new_itr;
5872 		/* Don't write the value here; it resets the adapter's
5873 		 * internal timer, and causes us to delay far longer than
5874 		 * we should between interrupts.  Instead, we write the ITR
5875 		 * value at the beginning of the next interrupt so the timing
5876 		 * ends up being correct.
5877 		 */
5878 		q_vector->itr_val = new_itr;
5879 		q_vector->set_itr = 1;
5880 	}
5881 }
5882 
5883 static void igb_tx_ctxtdesc(struct igb_ring *tx_ring,
5884 			    struct igb_tx_buffer *first,
5885 			    u32 vlan_macip_lens, u32 type_tucmd,
5886 			    u32 mss_l4len_idx)
5887 {
5888 	struct e1000_adv_tx_context_desc *context_desc;
5889 	u16 i = tx_ring->next_to_use;
5890 	struct timespec64 ts;
5891 
5892 	context_desc = IGB_TX_CTXTDESC(tx_ring, i);
5893 
5894 	i++;
5895 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
5896 
5897 	/* set bits to identify this as an advanced context descriptor */
5898 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
5899 
5900 	/* For 82575, context index must be unique per ring. */
5901 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5902 		mss_l4len_idx |= tx_ring->reg_idx << 4;
5903 
5904 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
5905 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
5906 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
5907 
5908 	/* We assume there is always a valid tx time available. Invalid times
5909 	 * should have been handled by the upper layers.
5910 	 */
5911 	if (tx_ring->launchtime_enable) {
5912 		ts = ktime_to_timespec64(first->skb->tstamp);
5913 		skb_txtime_consumed(first->skb);
5914 		context_desc->seqnum_seed = cpu_to_le32(ts.tv_nsec / 32);
5915 	} else {
5916 		context_desc->seqnum_seed = 0;
5917 	}
5918 }
5919 
5920 static int igb_tso(struct igb_ring *tx_ring,
5921 		   struct igb_tx_buffer *first,
5922 		   u8 *hdr_len)
5923 {
5924 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
5925 	struct sk_buff *skb = first->skb;
5926 	union {
5927 		struct iphdr *v4;
5928 		struct ipv6hdr *v6;
5929 		unsigned char *hdr;
5930 	} ip;
5931 	union {
5932 		struct tcphdr *tcp;
5933 		struct udphdr *udp;
5934 		unsigned char *hdr;
5935 	} l4;
5936 	u32 paylen, l4_offset;
5937 	int err;
5938 
5939 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5940 		return 0;
5941 
5942 	if (!skb_is_gso(skb))
5943 		return 0;
5944 
5945 	err = skb_cow_head(skb, 0);
5946 	if (err < 0)
5947 		return err;
5948 
5949 	ip.hdr = skb_network_header(skb);
5950 	l4.hdr = skb_checksum_start(skb);
5951 
5952 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5953 	type_tucmd = (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ?
5954 		      E1000_ADVTXD_TUCMD_L4T_UDP : E1000_ADVTXD_TUCMD_L4T_TCP;
5955 
5956 	/* initialize outer IP header fields */
5957 	if (ip.v4->version == 4) {
5958 		unsigned char *csum_start = skb_checksum_start(skb);
5959 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
5960 
5961 		/* IP header will have to cancel out any data that
5962 		 * is not a part of the outer IP header
5963 		 */
5964 		ip.v4->check = csum_fold(csum_partial(trans_start,
5965 						      csum_start - trans_start,
5966 						      0));
5967 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
5968 
5969 		ip.v4->tot_len = 0;
5970 		first->tx_flags |= IGB_TX_FLAGS_TSO |
5971 				   IGB_TX_FLAGS_CSUM |
5972 				   IGB_TX_FLAGS_IPV4;
5973 	} else {
5974 		ip.v6->payload_len = 0;
5975 		first->tx_flags |= IGB_TX_FLAGS_TSO |
5976 				   IGB_TX_FLAGS_CSUM;
5977 	}
5978 
5979 	/* determine offset of inner transport header */
5980 	l4_offset = l4.hdr - skb->data;
5981 
5982 	/* remove payload length from inner checksum */
5983 	paylen = skb->len - l4_offset;
5984 	if (type_tucmd & E1000_ADVTXD_TUCMD_L4T_TCP) {
5985 		/* compute length of segmentation header */
5986 		*hdr_len = (l4.tcp->doff * 4) + l4_offset;
5987 		csum_replace_by_diff(&l4.tcp->check,
5988 			(__force __wsum)htonl(paylen));
5989 	} else {
5990 		/* compute length of segmentation header */
5991 		*hdr_len = sizeof(*l4.udp) + l4_offset;
5992 		csum_replace_by_diff(&l4.udp->check,
5993 				     (__force __wsum)htonl(paylen));
5994 	}
5995 
5996 	/* update gso size and bytecount with header size */
5997 	first->gso_segs = skb_shinfo(skb)->gso_segs;
5998 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
5999 
6000 	/* MSS L4LEN IDX */
6001 	mss_l4len_idx = (*hdr_len - l4_offset) << E1000_ADVTXD_L4LEN_SHIFT;
6002 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
6003 
6004 	/* VLAN MACLEN IPLEN */
6005 	vlan_macip_lens = l4.hdr - ip.hdr;
6006 	vlan_macip_lens |= (ip.hdr - skb->data) << E1000_ADVTXD_MACLEN_SHIFT;
6007 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
6008 
6009 	igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens,
6010 			type_tucmd, mss_l4len_idx);
6011 
6012 	return 1;
6013 }
6014 
6015 static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
6016 {
6017 	struct sk_buff *skb = first->skb;
6018 	u32 vlan_macip_lens = 0;
6019 	u32 type_tucmd = 0;
6020 
6021 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
6022 csum_failed:
6023 		if (!(first->tx_flags & IGB_TX_FLAGS_VLAN) &&
6024 		    !tx_ring->launchtime_enable)
6025 			return;
6026 		goto no_csum;
6027 	}
6028 
6029 	switch (skb->csum_offset) {
6030 	case offsetof(struct tcphdr, check):
6031 		type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
6032 		fallthrough;
6033 	case offsetof(struct udphdr, check):
6034 		break;
6035 	case offsetof(struct sctphdr, checksum):
6036 		/* validate that this is actually an SCTP request */
6037 		if (skb_csum_is_sctp(skb)) {
6038 			type_tucmd = E1000_ADVTXD_TUCMD_L4T_SCTP;
6039 			break;
6040 		}
6041 		fallthrough;
6042 	default:
6043 		skb_checksum_help(skb);
6044 		goto csum_failed;
6045 	}
6046 
6047 	/* update TX checksum flag */
6048 	first->tx_flags |= IGB_TX_FLAGS_CSUM;
6049 	vlan_macip_lens = skb_checksum_start_offset(skb) -
6050 			  skb_network_offset(skb);
6051 no_csum:
6052 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
6053 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
6054 
6055 	igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens, type_tucmd, 0);
6056 }
6057 
6058 #define IGB_SET_FLAG(_input, _flag, _result) \
6059 	((_flag <= _result) ? \
6060 	 ((u32)(_input & _flag) * (_result / _flag)) : \
6061 	 ((u32)(_input & _flag) / (_flag / _result)))
6062 
6063 static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
6064 {
6065 	/* set type for advanced descriptor with frame checksum insertion */
6066 	u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
6067 		       E1000_ADVTXD_DCMD_DEXT |
6068 		       E1000_ADVTXD_DCMD_IFCS;
6069 
6070 	/* set HW vlan bit if vlan is present */
6071 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
6072 				 (E1000_ADVTXD_DCMD_VLE));
6073 
6074 	/* set segmentation bits for TSO */
6075 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
6076 				 (E1000_ADVTXD_DCMD_TSE));
6077 
6078 	/* set timestamp bit if present */
6079 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
6080 				 (E1000_ADVTXD_MAC_TSTAMP));
6081 
6082 	/* insert frame checksum */
6083 	cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
6084 
6085 	return cmd_type;
6086 }
6087 
6088 static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
6089 				 union e1000_adv_tx_desc *tx_desc,
6090 				 u32 tx_flags, unsigned int paylen)
6091 {
6092 	u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
6093 
6094 	/* 82575 requires a unique index per ring */
6095 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
6096 		olinfo_status |= tx_ring->reg_idx << 4;
6097 
6098 	/* insert L4 checksum */
6099 	olinfo_status |= IGB_SET_FLAG(tx_flags,
6100 				      IGB_TX_FLAGS_CSUM,
6101 				      (E1000_TXD_POPTS_TXSM << 8));
6102 
6103 	/* insert IPv4 checksum */
6104 	olinfo_status |= IGB_SET_FLAG(tx_flags,
6105 				      IGB_TX_FLAGS_IPV4,
6106 				      (E1000_TXD_POPTS_IXSM << 8));
6107 
6108 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6109 }
6110 
6111 static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
6112 {
6113 	struct net_device *netdev = tx_ring->netdev;
6114 
6115 	netif_stop_subqueue(netdev, tx_ring->queue_index);
6116 
6117 	/* Herbert's original patch had:
6118 	 *  smp_mb__after_netif_stop_queue();
6119 	 * but since that doesn't exist yet, just open code it.
6120 	 */
6121 	smp_mb();
6122 
6123 	/* We need to check again in a case another CPU has just
6124 	 * made room available.
6125 	 */
6126 	if (igb_desc_unused(tx_ring) < size)
6127 		return -EBUSY;
6128 
6129 	/* A reprieve! */
6130 	netif_wake_subqueue(netdev, tx_ring->queue_index);
6131 
6132 	u64_stats_update_begin(&tx_ring->tx_syncp2);
6133 	tx_ring->tx_stats.restart_queue2++;
6134 	u64_stats_update_end(&tx_ring->tx_syncp2);
6135 
6136 	return 0;
6137 }
6138 
6139 static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
6140 {
6141 	if (igb_desc_unused(tx_ring) >= size)
6142 		return 0;
6143 	return __igb_maybe_stop_tx(tx_ring, size);
6144 }
6145 
6146 static int igb_tx_map(struct igb_ring *tx_ring,
6147 		      struct igb_tx_buffer *first,
6148 		      const u8 hdr_len)
6149 {
6150 	struct sk_buff *skb = first->skb;
6151 	struct igb_tx_buffer *tx_buffer;
6152 	union e1000_adv_tx_desc *tx_desc;
6153 	skb_frag_t *frag;
6154 	dma_addr_t dma;
6155 	unsigned int data_len, size;
6156 	u32 tx_flags = first->tx_flags;
6157 	u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
6158 	u16 i = tx_ring->next_to_use;
6159 
6160 	tx_desc = IGB_TX_DESC(tx_ring, i);
6161 
6162 	igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
6163 
6164 	size = skb_headlen(skb);
6165 	data_len = skb->data_len;
6166 
6167 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6168 
6169 	tx_buffer = first;
6170 
6171 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
6172 		if (dma_mapping_error(tx_ring->dev, dma))
6173 			goto dma_error;
6174 
6175 		/* record length, and DMA address */
6176 		dma_unmap_len_set(tx_buffer, len, size);
6177 		dma_unmap_addr_set(tx_buffer, dma, dma);
6178 
6179 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6180 
6181 		while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
6182 			tx_desc->read.cmd_type_len =
6183 				cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
6184 
6185 			i++;
6186 			tx_desc++;
6187 			if (i == tx_ring->count) {
6188 				tx_desc = IGB_TX_DESC(tx_ring, 0);
6189 				i = 0;
6190 			}
6191 			tx_desc->read.olinfo_status = 0;
6192 
6193 			dma += IGB_MAX_DATA_PER_TXD;
6194 			size -= IGB_MAX_DATA_PER_TXD;
6195 
6196 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
6197 		}
6198 
6199 		if (likely(!data_len))
6200 			break;
6201 
6202 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
6203 
6204 		i++;
6205 		tx_desc++;
6206 		if (i == tx_ring->count) {
6207 			tx_desc = IGB_TX_DESC(tx_ring, 0);
6208 			i = 0;
6209 		}
6210 		tx_desc->read.olinfo_status = 0;
6211 
6212 		size = skb_frag_size(frag);
6213 		data_len -= size;
6214 
6215 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
6216 				       size, DMA_TO_DEVICE);
6217 
6218 		tx_buffer = &tx_ring->tx_buffer_info[i];
6219 	}
6220 
6221 	/* write last descriptor with RS and EOP bits */
6222 	cmd_type |= size | IGB_TXD_DCMD;
6223 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6224 
6225 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
6226 
6227 	/* set the timestamp */
6228 	first->time_stamp = jiffies;
6229 
6230 	skb_tx_timestamp(skb);
6231 
6232 	/* Force memory writes to complete before letting h/w know there
6233 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
6234 	 * memory model archs, such as IA-64).
6235 	 *
6236 	 * We also need this memory barrier to make certain all of the
6237 	 * status bits have been updated before next_to_watch is written.
6238 	 */
6239 	dma_wmb();
6240 
6241 	/* set next_to_watch value indicating a packet is present */
6242 	first->next_to_watch = tx_desc;
6243 
6244 	i++;
6245 	if (i == tx_ring->count)
6246 		i = 0;
6247 
6248 	tx_ring->next_to_use = i;
6249 
6250 	/* Make sure there is space in the ring for the next send. */
6251 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
6252 
6253 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
6254 		writel(i, tx_ring->tail);
6255 	}
6256 	return 0;
6257 
6258 dma_error:
6259 	dev_err(tx_ring->dev, "TX DMA map failed\n");
6260 	tx_buffer = &tx_ring->tx_buffer_info[i];
6261 
6262 	/* clear dma mappings for failed tx_buffer_info map */
6263 	while (tx_buffer != first) {
6264 		if (dma_unmap_len(tx_buffer, len))
6265 			dma_unmap_page(tx_ring->dev,
6266 				       dma_unmap_addr(tx_buffer, dma),
6267 				       dma_unmap_len(tx_buffer, len),
6268 				       DMA_TO_DEVICE);
6269 		dma_unmap_len_set(tx_buffer, len, 0);
6270 
6271 		if (i-- == 0)
6272 			i += tx_ring->count;
6273 		tx_buffer = &tx_ring->tx_buffer_info[i];
6274 	}
6275 
6276 	if (dma_unmap_len(tx_buffer, len))
6277 		dma_unmap_single(tx_ring->dev,
6278 				 dma_unmap_addr(tx_buffer, dma),
6279 				 dma_unmap_len(tx_buffer, len),
6280 				 DMA_TO_DEVICE);
6281 	dma_unmap_len_set(tx_buffer, len, 0);
6282 
6283 	dev_kfree_skb_any(tx_buffer->skb);
6284 	tx_buffer->skb = NULL;
6285 
6286 	tx_ring->next_to_use = i;
6287 
6288 	return -1;
6289 }
6290 
6291 int igb_xmit_xdp_ring(struct igb_adapter *adapter,
6292 		      struct igb_ring *tx_ring,
6293 		      struct xdp_frame *xdpf)
6294 {
6295 	struct skb_shared_info *sinfo = xdp_get_shared_info_from_frame(xdpf);
6296 	u8 nr_frags = unlikely(xdp_frame_has_frags(xdpf)) ? sinfo->nr_frags : 0;
6297 	u16 count, i, index = tx_ring->next_to_use;
6298 	struct igb_tx_buffer *tx_head = &tx_ring->tx_buffer_info[index];
6299 	struct igb_tx_buffer *tx_buffer = tx_head;
6300 	union e1000_adv_tx_desc *tx_desc = IGB_TX_DESC(tx_ring, index);
6301 	u32 len = xdpf->len, cmd_type, olinfo_status;
6302 	void *data = xdpf->data;
6303 
6304 	count = TXD_USE_COUNT(len);
6305 	for (i = 0; i < nr_frags; i++)
6306 		count += TXD_USE_COUNT(skb_frag_size(&sinfo->frags[i]));
6307 
6308 	if (igb_maybe_stop_tx(tx_ring, count + 3))
6309 		return IGB_XDP_CONSUMED;
6310 
6311 	i = 0;
6312 	/* record the location of the first descriptor for this packet */
6313 	tx_head->bytecount = xdp_get_frame_len(xdpf);
6314 	tx_head->type = IGB_TYPE_XDP;
6315 	tx_head->gso_segs = 1;
6316 	tx_head->xdpf = xdpf;
6317 
6318 	olinfo_status = tx_head->bytecount << E1000_ADVTXD_PAYLEN_SHIFT;
6319 	/* 82575 requires a unique index per ring */
6320 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
6321 		olinfo_status |= tx_ring->reg_idx << 4;
6322 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6323 
6324 	for (;;) {
6325 		dma_addr_t dma;
6326 
6327 		dma = dma_map_single(tx_ring->dev, data, len, DMA_TO_DEVICE);
6328 		if (dma_mapping_error(tx_ring->dev, dma))
6329 			goto unmap;
6330 
6331 		/* record length, and DMA address */
6332 		dma_unmap_len_set(tx_buffer, len, len);
6333 		dma_unmap_addr_set(tx_buffer, dma, dma);
6334 
6335 		/* put descriptor type bits */
6336 		cmd_type = E1000_ADVTXD_DTYP_DATA | E1000_ADVTXD_DCMD_DEXT |
6337 			   E1000_ADVTXD_DCMD_IFCS | len;
6338 
6339 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6340 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6341 
6342 		tx_buffer->protocol = 0;
6343 
6344 		if (++index == tx_ring->count)
6345 			index = 0;
6346 
6347 		if (i == nr_frags)
6348 			break;
6349 
6350 		tx_buffer = &tx_ring->tx_buffer_info[index];
6351 		tx_desc = IGB_TX_DESC(tx_ring, index);
6352 		tx_desc->read.olinfo_status = 0;
6353 
6354 		data = skb_frag_address(&sinfo->frags[i]);
6355 		len = skb_frag_size(&sinfo->frags[i]);
6356 		i++;
6357 	}
6358 	tx_desc->read.cmd_type_len |= cpu_to_le32(IGB_TXD_DCMD);
6359 
6360 	netdev_tx_sent_queue(txring_txq(tx_ring), tx_head->bytecount);
6361 	/* set the timestamp */
6362 	tx_head->time_stamp = jiffies;
6363 
6364 	/* Avoid any potential race with xdp_xmit and cleanup */
6365 	smp_wmb();
6366 
6367 	/* set next_to_watch value indicating a packet is present */
6368 	tx_head->next_to_watch = tx_desc;
6369 	tx_ring->next_to_use = index;
6370 
6371 	/* Make sure there is space in the ring for the next send. */
6372 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
6373 
6374 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more())
6375 		writel(index, tx_ring->tail);
6376 
6377 	return IGB_XDP_TX;
6378 
6379 unmap:
6380 	for (;;) {
6381 		tx_buffer = &tx_ring->tx_buffer_info[index];
6382 		if (dma_unmap_len(tx_buffer, len))
6383 			dma_unmap_page(tx_ring->dev,
6384 				       dma_unmap_addr(tx_buffer, dma),
6385 				       dma_unmap_len(tx_buffer, len),
6386 				       DMA_TO_DEVICE);
6387 		dma_unmap_len_set(tx_buffer, len, 0);
6388 		if (tx_buffer == tx_head)
6389 			break;
6390 
6391 		if (!index)
6392 			index += tx_ring->count;
6393 		index--;
6394 	}
6395 
6396 	return IGB_XDP_CONSUMED;
6397 }
6398 
6399 netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
6400 				struct igb_ring *tx_ring)
6401 {
6402 	struct igb_tx_buffer *first;
6403 	int tso;
6404 	u32 tx_flags = 0;
6405 	unsigned short f;
6406 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
6407 	__be16 protocol = vlan_get_protocol(skb);
6408 	u8 hdr_len = 0;
6409 
6410 	/* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
6411 	 *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
6412 	 *       + 2 desc gap to keep tail from touching head,
6413 	 *       + 1 desc for context descriptor,
6414 	 * otherwise try next time
6415 	 */
6416 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6417 		count += TXD_USE_COUNT(skb_frag_size(
6418 						&skb_shinfo(skb)->frags[f]));
6419 
6420 	if (igb_maybe_stop_tx(tx_ring, count + 3)) {
6421 		/* this is a hard error */
6422 		return NETDEV_TX_BUSY;
6423 	}
6424 
6425 	/* record the location of the first descriptor for this packet */
6426 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6427 	first->type = IGB_TYPE_SKB;
6428 	first->skb = skb;
6429 	first->bytecount = skb->len;
6430 	first->gso_segs = 1;
6431 
6432 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
6433 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6434 
6435 		if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
6436 		    !test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
6437 					   &adapter->state)) {
6438 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
6439 			tx_flags |= IGB_TX_FLAGS_TSTAMP;
6440 
6441 			adapter->ptp_tx_skb = skb_get(skb);
6442 			adapter->ptp_tx_start = jiffies;
6443 			if (adapter->hw.mac.type == e1000_82576)
6444 				schedule_work(&adapter->ptp_tx_work);
6445 		} else {
6446 			adapter->tx_hwtstamp_skipped++;
6447 		}
6448 	}
6449 
6450 	if (skb_vlan_tag_present(skb)) {
6451 		tx_flags |= IGB_TX_FLAGS_VLAN;
6452 		tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
6453 	}
6454 
6455 	/* record initial flags and protocol */
6456 	first->tx_flags = tx_flags;
6457 	first->protocol = protocol;
6458 
6459 	tso = igb_tso(tx_ring, first, &hdr_len);
6460 	if (tso < 0)
6461 		goto out_drop;
6462 	else if (!tso)
6463 		igb_tx_csum(tx_ring, first);
6464 
6465 	if (igb_tx_map(tx_ring, first, hdr_len))
6466 		goto cleanup_tx_tstamp;
6467 
6468 	return NETDEV_TX_OK;
6469 
6470 out_drop:
6471 	dev_kfree_skb_any(first->skb);
6472 	first->skb = NULL;
6473 cleanup_tx_tstamp:
6474 	if (unlikely(tx_flags & IGB_TX_FLAGS_TSTAMP)) {
6475 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6476 
6477 		dev_kfree_skb_any(adapter->ptp_tx_skb);
6478 		adapter->ptp_tx_skb = NULL;
6479 		if (adapter->hw.mac.type == e1000_82576)
6480 			cancel_work_sync(&adapter->ptp_tx_work);
6481 		clear_bit_unlock(__IGB_PTP_TX_IN_PROGRESS, &adapter->state);
6482 	}
6483 
6484 	return NETDEV_TX_OK;
6485 }
6486 
6487 static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
6488 						    struct sk_buff *skb)
6489 {
6490 	unsigned int r_idx = skb->queue_mapping;
6491 
6492 	if (r_idx >= adapter->num_tx_queues)
6493 		r_idx = r_idx % adapter->num_tx_queues;
6494 
6495 	return adapter->tx_ring[r_idx];
6496 }
6497 
6498 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
6499 				  struct net_device *netdev)
6500 {
6501 	struct igb_adapter *adapter = netdev_priv(netdev);
6502 
6503 	/* The minimum packet size with TCTL.PSP set is 17 so pad the skb
6504 	 * in order to meet this minimum size requirement.
6505 	 */
6506 	if (skb_put_padto(skb, 17))
6507 		return NETDEV_TX_OK;
6508 
6509 	return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
6510 }
6511 
6512 /**
6513  *  igb_tx_timeout - Respond to a Tx Hang
6514  *  @netdev: network interface device structure
6515  *  @txqueue: number of the Tx queue that hung (unused)
6516  **/
6517 static void igb_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
6518 {
6519 	struct igb_adapter *adapter = netdev_priv(netdev);
6520 	struct e1000_hw *hw = &adapter->hw;
6521 
6522 	/* Do the reset outside of interrupt context */
6523 	adapter->tx_timeout_count++;
6524 
6525 	if (hw->mac.type >= e1000_82580)
6526 		hw->dev_spec._82575.global_device_reset = true;
6527 
6528 	schedule_work(&adapter->reset_task);
6529 	wr32(E1000_EICS,
6530 	     (adapter->eims_enable_mask & ~adapter->eims_other));
6531 }
6532 
6533 static void igb_reset_task(struct work_struct *work)
6534 {
6535 	struct igb_adapter *adapter;
6536 	adapter = container_of(work, struct igb_adapter, reset_task);
6537 
6538 	rtnl_lock();
6539 	/* If we're already down or resetting, just bail */
6540 	if (test_bit(__IGB_DOWN, &adapter->state) ||
6541 	    test_bit(__IGB_RESETTING, &adapter->state)) {
6542 		rtnl_unlock();
6543 		return;
6544 	}
6545 
6546 	igb_dump(adapter);
6547 	netdev_err(adapter->netdev, "Reset adapter\n");
6548 	igb_reinit_locked(adapter);
6549 	rtnl_unlock();
6550 }
6551 
6552 /**
6553  *  igb_get_stats64 - Get System Network Statistics
6554  *  @netdev: network interface device structure
6555  *  @stats: rtnl_link_stats64 pointer
6556  **/
6557 static void igb_get_stats64(struct net_device *netdev,
6558 			    struct rtnl_link_stats64 *stats)
6559 {
6560 	struct igb_adapter *adapter = netdev_priv(netdev);
6561 
6562 	spin_lock(&adapter->stats64_lock);
6563 	igb_update_stats(adapter);
6564 	memcpy(stats, &adapter->stats64, sizeof(*stats));
6565 	spin_unlock(&adapter->stats64_lock);
6566 }
6567 
6568 /**
6569  *  igb_change_mtu - Change the Maximum Transfer Unit
6570  *  @netdev: network interface device structure
6571  *  @new_mtu: new value for maximum frame size
6572  *
6573  *  Returns 0 on success, negative on failure
6574  **/
6575 static int igb_change_mtu(struct net_device *netdev, int new_mtu)
6576 {
6577 	struct igb_adapter *adapter = netdev_priv(netdev);
6578 	int max_frame = new_mtu + IGB_ETH_PKT_HDR_PAD;
6579 
6580 	if (adapter->xdp_prog) {
6581 		int i;
6582 
6583 		for (i = 0; i < adapter->num_rx_queues; i++) {
6584 			struct igb_ring *ring = adapter->rx_ring[i];
6585 
6586 			if (max_frame > igb_rx_bufsz(ring)) {
6587 				netdev_warn(adapter->netdev,
6588 					    "Requested MTU size is not supported with XDP. Max frame size is %d\n",
6589 					    max_frame);
6590 				return -EINVAL;
6591 			}
6592 		}
6593 	}
6594 
6595 	/* adjust max frame to be at least the size of a standard frame */
6596 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
6597 		max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
6598 
6599 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
6600 		usleep_range(1000, 2000);
6601 
6602 	/* igb_down has a dependency on max_frame_size */
6603 	adapter->max_frame_size = max_frame;
6604 
6605 	if (netif_running(netdev))
6606 		igb_down(adapter);
6607 
6608 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6609 		   netdev->mtu, new_mtu);
6610 	netdev->mtu = new_mtu;
6611 
6612 	if (netif_running(netdev))
6613 		igb_up(adapter);
6614 	else
6615 		igb_reset(adapter);
6616 
6617 	clear_bit(__IGB_RESETTING, &adapter->state);
6618 
6619 	return 0;
6620 }
6621 
6622 /**
6623  *  igb_update_stats - Update the board statistics counters
6624  *  @adapter: board private structure
6625  **/
6626 void igb_update_stats(struct igb_adapter *adapter)
6627 {
6628 	struct rtnl_link_stats64 *net_stats = &adapter->stats64;
6629 	struct e1000_hw *hw = &adapter->hw;
6630 	struct pci_dev *pdev = adapter->pdev;
6631 	u32 reg, mpc;
6632 	int i;
6633 	u64 bytes, packets;
6634 	unsigned int start;
6635 	u64 _bytes, _packets;
6636 
6637 	/* Prevent stats update while adapter is being reset, or if the pci
6638 	 * connection is down.
6639 	 */
6640 	if (adapter->link_speed == 0)
6641 		return;
6642 	if (pci_channel_offline(pdev))
6643 		return;
6644 
6645 	bytes = 0;
6646 	packets = 0;
6647 
6648 	rcu_read_lock();
6649 	for (i = 0; i < adapter->num_rx_queues; i++) {
6650 		struct igb_ring *ring = adapter->rx_ring[i];
6651 		u32 rqdpc = rd32(E1000_RQDPC(i));
6652 		if (hw->mac.type >= e1000_i210)
6653 			wr32(E1000_RQDPC(i), 0);
6654 
6655 		if (rqdpc) {
6656 			ring->rx_stats.drops += rqdpc;
6657 			net_stats->rx_fifo_errors += rqdpc;
6658 		}
6659 
6660 		do {
6661 			start = u64_stats_fetch_begin(&ring->rx_syncp);
6662 			_bytes = ring->rx_stats.bytes;
6663 			_packets = ring->rx_stats.packets;
6664 		} while (u64_stats_fetch_retry(&ring->rx_syncp, start));
6665 		bytes += _bytes;
6666 		packets += _packets;
6667 	}
6668 
6669 	net_stats->rx_bytes = bytes;
6670 	net_stats->rx_packets = packets;
6671 
6672 	bytes = 0;
6673 	packets = 0;
6674 	for (i = 0; i < adapter->num_tx_queues; i++) {
6675 		struct igb_ring *ring = adapter->tx_ring[i];
6676 		do {
6677 			start = u64_stats_fetch_begin(&ring->tx_syncp);
6678 			_bytes = ring->tx_stats.bytes;
6679 			_packets = ring->tx_stats.packets;
6680 		} while (u64_stats_fetch_retry(&ring->tx_syncp, start));
6681 		bytes += _bytes;
6682 		packets += _packets;
6683 	}
6684 	net_stats->tx_bytes = bytes;
6685 	net_stats->tx_packets = packets;
6686 	rcu_read_unlock();
6687 
6688 	/* read stats registers */
6689 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
6690 	adapter->stats.gprc += rd32(E1000_GPRC);
6691 	adapter->stats.gorc += rd32(E1000_GORCL);
6692 	rd32(E1000_GORCH); /* clear GORCL */
6693 	adapter->stats.bprc += rd32(E1000_BPRC);
6694 	adapter->stats.mprc += rd32(E1000_MPRC);
6695 	adapter->stats.roc += rd32(E1000_ROC);
6696 
6697 	adapter->stats.prc64 += rd32(E1000_PRC64);
6698 	adapter->stats.prc127 += rd32(E1000_PRC127);
6699 	adapter->stats.prc255 += rd32(E1000_PRC255);
6700 	adapter->stats.prc511 += rd32(E1000_PRC511);
6701 	adapter->stats.prc1023 += rd32(E1000_PRC1023);
6702 	adapter->stats.prc1522 += rd32(E1000_PRC1522);
6703 	adapter->stats.symerrs += rd32(E1000_SYMERRS);
6704 	adapter->stats.sec += rd32(E1000_SEC);
6705 
6706 	mpc = rd32(E1000_MPC);
6707 	adapter->stats.mpc += mpc;
6708 	net_stats->rx_fifo_errors += mpc;
6709 	adapter->stats.scc += rd32(E1000_SCC);
6710 	adapter->stats.ecol += rd32(E1000_ECOL);
6711 	adapter->stats.mcc += rd32(E1000_MCC);
6712 	adapter->stats.latecol += rd32(E1000_LATECOL);
6713 	adapter->stats.dc += rd32(E1000_DC);
6714 	adapter->stats.rlec += rd32(E1000_RLEC);
6715 	adapter->stats.xonrxc += rd32(E1000_XONRXC);
6716 	adapter->stats.xontxc += rd32(E1000_XONTXC);
6717 	adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
6718 	adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
6719 	adapter->stats.fcruc += rd32(E1000_FCRUC);
6720 	adapter->stats.gptc += rd32(E1000_GPTC);
6721 	adapter->stats.gotc += rd32(E1000_GOTCL);
6722 	rd32(E1000_GOTCH); /* clear GOTCL */
6723 	adapter->stats.rnbc += rd32(E1000_RNBC);
6724 	adapter->stats.ruc += rd32(E1000_RUC);
6725 	adapter->stats.rfc += rd32(E1000_RFC);
6726 	adapter->stats.rjc += rd32(E1000_RJC);
6727 	adapter->stats.tor += rd32(E1000_TORH);
6728 	adapter->stats.tot += rd32(E1000_TOTH);
6729 	adapter->stats.tpr += rd32(E1000_TPR);
6730 
6731 	adapter->stats.ptc64 += rd32(E1000_PTC64);
6732 	adapter->stats.ptc127 += rd32(E1000_PTC127);
6733 	adapter->stats.ptc255 += rd32(E1000_PTC255);
6734 	adapter->stats.ptc511 += rd32(E1000_PTC511);
6735 	adapter->stats.ptc1023 += rd32(E1000_PTC1023);
6736 	adapter->stats.ptc1522 += rd32(E1000_PTC1522);
6737 
6738 	adapter->stats.mptc += rd32(E1000_MPTC);
6739 	adapter->stats.bptc += rd32(E1000_BPTC);
6740 
6741 	adapter->stats.tpt += rd32(E1000_TPT);
6742 	adapter->stats.colc += rd32(E1000_COLC);
6743 
6744 	adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
6745 	/* read internal phy specific stats */
6746 	reg = rd32(E1000_CTRL_EXT);
6747 	if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
6748 		adapter->stats.rxerrc += rd32(E1000_RXERRC);
6749 
6750 		/* this stat has invalid values on i210/i211 */
6751 		if ((hw->mac.type != e1000_i210) &&
6752 		    (hw->mac.type != e1000_i211))
6753 			adapter->stats.tncrs += rd32(E1000_TNCRS);
6754 	}
6755 
6756 	adapter->stats.tsctc += rd32(E1000_TSCTC);
6757 	adapter->stats.tsctfc += rd32(E1000_TSCTFC);
6758 
6759 	adapter->stats.iac += rd32(E1000_IAC);
6760 	adapter->stats.icrxoc += rd32(E1000_ICRXOC);
6761 	adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
6762 	adapter->stats.icrxatc += rd32(E1000_ICRXATC);
6763 	adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
6764 	adapter->stats.ictxatc += rd32(E1000_ICTXATC);
6765 	adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
6766 	adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
6767 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
6768 
6769 	/* Fill out the OS statistics structure */
6770 	net_stats->multicast = adapter->stats.mprc;
6771 	net_stats->collisions = adapter->stats.colc;
6772 
6773 	/* Rx Errors */
6774 
6775 	/* RLEC on some newer hardware can be incorrect so build
6776 	 * our own version based on RUC and ROC
6777 	 */
6778 	net_stats->rx_errors = adapter->stats.rxerrc +
6779 		adapter->stats.crcerrs + adapter->stats.algnerrc +
6780 		adapter->stats.ruc + adapter->stats.roc +
6781 		adapter->stats.cexterr;
6782 	net_stats->rx_length_errors = adapter->stats.ruc +
6783 				      adapter->stats.roc;
6784 	net_stats->rx_crc_errors = adapter->stats.crcerrs;
6785 	net_stats->rx_frame_errors = adapter->stats.algnerrc;
6786 	net_stats->rx_missed_errors = adapter->stats.mpc;
6787 
6788 	/* Tx Errors */
6789 	net_stats->tx_errors = adapter->stats.ecol +
6790 			       adapter->stats.latecol;
6791 	net_stats->tx_aborted_errors = adapter->stats.ecol;
6792 	net_stats->tx_window_errors = adapter->stats.latecol;
6793 	net_stats->tx_carrier_errors = adapter->stats.tncrs;
6794 
6795 	/* Tx Dropped needs to be maintained elsewhere */
6796 
6797 	/* Management Stats */
6798 	adapter->stats.mgptc += rd32(E1000_MGTPTC);
6799 	adapter->stats.mgprc += rd32(E1000_MGTPRC);
6800 	adapter->stats.mgpdc += rd32(E1000_MGTPDC);
6801 
6802 	/* OS2BMC Stats */
6803 	reg = rd32(E1000_MANC);
6804 	if (reg & E1000_MANC_EN_BMC2OS) {
6805 		adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
6806 		adapter->stats.o2bspc += rd32(E1000_O2BSPC);
6807 		adapter->stats.b2ospc += rd32(E1000_B2OSPC);
6808 		adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
6809 	}
6810 }
6811 
6812 static void igb_perout(struct igb_adapter *adapter, int tsintr_tt)
6813 {
6814 	int pin = ptp_find_pin(adapter->ptp_clock, PTP_PF_PEROUT, tsintr_tt);
6815 	struct e1000_hw *hw = &adapter->hw;
6816 	struct timespec64 ts;
6817 	u32 tsauxc;
6818 
6819 	if (pin < 0 || pin >= IGB_N_SDP)
6820 		return;
6821 
6822 	spin_lock(&adapter->tmreg_lock);
6823 
6824 	if (hw->mac.type == e1000_82580 ||
6825 	    hw->mac.type == e1000_i354 ||
6826 	    hw->mac.type == e1000_i350) {
6827 		s64 ns = timespec64_to_ns(&adapter->perout[tsintr_tt].period);
6828 		u32 systiml, systimh, level_mask, level, rem;
6829 		u64 systim, now;
6830 
6831 		/* read systim registers in sequence */
6832 		rd32(E1000_SYSTIMR);
6833 		systiml = rd32(E1000_SYSTIML);
6834 		systimh = rd32(E1000_SYSTIMH);
6835 		systim = (((u64)(systimh & 0xFF)) << 32) | ((u64)systiml);
6836 		now = timecounter_cyc2time(&adapter->tc, systim);
6837 
6838 		if (pin < 2) {
6839 			level_mask = (tsintr_tt == 1) ? 0x80000 : 0x40000;
6840 			level = (rd32(E1000_CTRL) & level_mask) ? 1 : 0;
6841 		} else {
6842 			level_mask = (tsintr_tt == 1) ? 0x80 : 0x40;
6843 			level = (rd32(E1000_CTRL_EXT) & level_mask) ? 1 : 0;
6844 		}
6845 
6846 		div_u64_rem(now, ns, &rem);
6847 		systim = systim + (ns - rem);
6848 
6849 		/* synchronize pin level with rising/falling edges */
6850 		div_u64_rem(now, ns << 1, &rem);
6851 		if (rem < ns) {
6852 			/* first half of period */
6853 			if (level == 0) {
6854 				/* output is already low, skip this period */
6855 				systim += ns;
6856 				pr_notice("igb: periodic output on %s missed falling edge\n",
6857 					  adapter->sdp_config[pin].name);
6858 			}
6859 		} else {
6860 			/* second half of period */
6861 			if (level == 1) {
6862 				/* output is already high, skip this period */
6863 				systim += ns;
6864 				pr_notice("igb: periodic output on %s missed rising edge\n",
6865 					  adapter->sdp_config[pin].name);
6866 			}
6867 		}
6868 
6869 		/* for this chip family tv_sec is the upper part of the binary value,
6870 		 * so not seconds
6871 		 */
6872 		ts.tv_nsec = (u32)systim;
6873 		ts.tv_sec  = ((u32)(systim >> 32)) & 0xFF;
6874 	} else {
6875 		ts = timespec64_add(adapter->perout[tsintr_tt].start,
6876 				    adapter->perout[tsintr_tt].period);
6877 	}
6878 
6879 	/* u32 conversion of tv_sec is safe until y2106 */
6880 	wr32((tsintr_tt == 1) ? E1000_TRGTTIML1 : E1000_TRGTTIML0, ts.tv_nsec);
6881 	wr32((tsintr_tt == 1) ? E1000_TRGTTIMH1 : E1000_TRGTTIMH0, (u32)ts.tv_sec);
6882 	tsauxc = rd32(E1000_TSAUXC);
6883 	tsauxc |= TSAUXC_EN_TT0;
6884 	wr32(E1000_TSAUXC, tsauxc);
6885 	adapter->perout[tsintr_tt].start = ts;
6886 
6887 	spin_unlock(&adapter->tmreg_lock);
6888 }
6889 
6890 static void igb_extts(struct igb_adapter *adapter, int tsintr_tt)
6891 {
6892 	int pin = ptp_find_pin(adapter->ptp_clock, PTP_PF_EXTTS, tsintr_tt);
6893 	int auxstmpl = (tsintr_tt == 1) ? E1000_AUXSTMPL1 : E1000_AUXSTMPL0;
6894 	int auxstmph = (tsintr_tt == 1) ? E1000_AUXSTMPH1 : E1000_AUXSTMPH0;
6895 	struct e1000_hw *hw = &adapter->hw;
6896 	struct ptp_clock_event event;
6897 	struct timespec64 ts;
6898 
6899 	if (pin < 0 || pin >= IGB_N_SDP)
6900 		return;
6901 
6902 	if (hw->mac.type == e1000_82580 ||
6903 	    hw->mac.type == e1000_i354 ||
6904 	    hw->mac.type == e1000_i350) {
6905 		s64 ns = rd32(auxstmpl);
6906 
6907 		ns += ((s64)(rd32(auxstmph) & 0xFF)) << 32;
6908 		ts = ns_to_timespec64(ns);
6909 	} else {
6910 		ts.tv_nsec = rd32(auxstmpl);
6911 		ts.tv_sec  = rd32(auxstmph);
6912 	}
6913 
6914 	event.type = PTP_CLOCK_EXTTS;
6915 	event.index = tsintr_tt;
6916 	event.timestamp = ts.tv_sec * 1000000000ULL + ts.tv_nsec;
6917 	ptp_clock_event(adapter->ptp_clock, &event);
6918 }
6919 
6920 static void igb_tsync_interrupt(struct igb_adapter *adapter)
6921 {
6922 	struct e1000_hw *hw = &adapter->hw;
6923 	u32 ack = 0, tsicr = rd32(E1000_TSICR);
6924 	struct ptp_clock_event event;
6925 
6926 	if (tsicr & TSINTR_SYS_WRAP) {
6927 		event.type = PTP_CLOCK_PPS;
6928 		if (adapter->ptp_caps.pps)
6929 			ptp_clock_event(adapter->ptp_clock, &event);
6930 		ack |= TSINTR_SYS_WRAP;
6931 	}
6932 
6933 	if (tsicr & E1000_TSICR_TXTS) {
6934 		/* retrieve hardware timestamp */
6935 		schedule_work(&adapter->ptp_tx_work);
6936 		ack |= E1000_TSICR_TXTS;
6937 	}
6938 
6939 	if (tsicr & TSINTR_TT0) {
6940 		igb_perout(adapter, 0);
6941 		ack |= TSINTR_TT0;
6942 	}
6943 
6944 	if (tsicr & TSINTR_TT1) {
6945 		igb_perout(adapter, 1);
6946 		ack |= TSINTR_TT1;
6947 	}
6948 
6949 	if (tsicr & TSINTR_AUTT0) {
6950 		igb_extts(adapter, 0);
6951 		ack |= TSINTR_AUTT0;
6952 	}
6953 
6954 	if (tsicr & TSINTR_AUTT1) {
6955 		igb_extts(adapter, 1);
6956 		ack |= TSINTR_AUTT1;
6957 	}
6958 
6959 	/* acknowledge the interrupts */
6960 	wr32(E1000_TSICR, ack);
6961 }
6962 
6963 static irqreturn_t igb_msix_other(int irq, void *data)
6964 {
6965 	struct igb_adapter *adapter = data;
6966 	struct e1000_hw *hw = &adapter->hw;
6967 	u32 icr = rd32(E1000_ICR);
6968 	/* reading ICR causes bit 31 of EICR to be cleared */
6969 
6970 	if (icr & E1000_ICR_DRSTA)
6971 		schedule_work(&adapter->reset_task);
6972 
6973 	if (icr & E1000_ICR_DOUTSYNC) {
6974 		/* HW is reporting DMA is out of sync */
6975 		adapter->stats.doosync++;
6976 		/* The DMA Out of Sync is also indication of a spoof event
6977 		 * in IOV mode. Check the Wrong VM Behavior register to
6978 		 * see if it is really a spoof event.
6979 		 */
6980 		igb_check_wvbr(adapter);
6981 	}
6982 
6983 	/* Check for a mailbox event */
6984 	if (icr & E1000_ICR_VMMB)
6985 		igb_msg_task(adapter);
6986 
6987 	if (icr & E1000_ICR_LSC) {
6988 		hw->mac.get_link_status = 1;
6989 		/* guard against interrupt when we're going down */
6990 		if (!test_bit(__IGB_DOWN, &adapter->state))
6991 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6992 	}
6993 
6994 	if (icr & E1000_ICR_TS)
6995 		igb_tsync_interrupt(adapter);
6996 
6997 	wr32(E1000_EIMS, adapter->eims_other);
6998 
6999 	return IRQ_HANDLED;
7000 }
7001 
7002 static void igb_write_itr(struct igb_q_vector *q_vector)
7003 {
7004 	struct igb_adapter *adapter = q_vector->adapter;
7005 	u32 itr_val = q_vector->itr_val & 0x7FFC;
7006 
7007 	if (!q_vector->set_itr)
7008 		return;
7009 
7010 	if (!itr_val)
7011 		itr_val = 0x4;
7012 
7013 	if (adapter->hw.mac.type == e1000_82575)
7014 		itr_val |= itr_val << 16;
7015 	else
7016 		itr_val |= E1000_EITR_CNT_IGNR;
7017 
7018 	writel(itr_val, q_vector->itr_register);
7019 	q_vector->set_itr = 0;
7020 }
7021 
7022 static irqreturn_t igb_msix_ring(int irq, void *data)
7023 {
7024 	struct igb_q_vector *q_vector = data;
7025 
7026 	/* Write the ITR value calculated from the previous interrupt. */
7027 	igb_write_itr(q_vector);
7028 
7029 	napi_schedule(&q_vector->napi);
7030 
7031 	return IRQ_HANDLED;
7032 }
7033 
7034 #ifdef CONFIG_IGB_DCA
7035 static void igb_update_tx_dca(struct igb_adapter *adapter,
7036 			      struct igb_ring *tx_ring,
7037 			      int cpu)
7038 {
7039 	struct e1000_hw *hw = &adapter->hw;
7040 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
7041 
7042 	if (hw->mac.type != e1000_82575)
7043 		txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
7044 
7045 	/* We can enable relaxed ordering for reads, but not writes when
7046 	 * DCA is enabled.  This is due to a known issue in some chipsets
7047 	 * which will cause the DCA tag to be cleared.
7048 	 */
7049 	txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
7050 		  E1000_DCA_TXCTRL_DATA_RRO_EN |
7051 		  E1000_DCA_TXCTRL_DESC_DCA_EN;
7052 
7053 	wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
7054 }
7055 
7056 static void igb_update_rx_dca(struct igb_adapter *adapter,
7057 			      struct igb_ring *rx_ring,
7058 			      int cpu)
7059 {
7060 	struct e1000_hw *hw = &adapter->hw;
7061 	u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
7062 
7063 	if (hw->mac.type != e1000_82575)
7064 		rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
7065 
7066 	/* We can enable relaxed ordering for reads, but not writes when
7067 	 * DCA is enabled.  This is due to a known issue in some chipsets
7068 	 * which will cause the DCA tag to be cleared.
7069 	 */
7070 	rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
7071 		  E1000_DCA_RXCTRL_DESC_DCA_EN;
7072 
7073 	wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
7074 }
7075 
7076 static void igb_update_dca(struct igb_q_vector *q_vector)
7077 {
7078 	struct igb_adapter *adapter = q_vector->adapter;
7079 	int cpu = get_cpu();
7080 
7081 	if (q_vector->cpu == cpu)
7082 		goto out_no_update;
7083 
7084 	if (q_vector->tx.ring)
7085 		igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
7086 
7087 	if (q_vector->rx.ring)
7088 		igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
7089 
7090 	q_vector->cpu = cpu;
7091 out_no_update:
7092 	put_cpu();
7093 }
7094 
7095 static void igb_setup_dca(struct igb_adapter *adapter)
7096 {
7097 	struct e1000_hw *hw = &adapter->hw;
7098 	int i;
7099 
7100 	if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
7101 		return;
7102 
7103 	/* Always use CB2 mode, difference is masked in the CB driver. */
7104 	wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
7105 
7106 	for (i = 0; i < adapter->num_q_vectors; i++) {
7107 		adapter->q_vector[i]->cpu = -1;
7108 		igb_update_dca(adapter->q_vector[i]);
7109 	}
7110 }
7111 
7112 static int __igb_notify_dca(struct device *dev, void *data)
7113 {
7114 	struct net_device *netdev = dev_get_drvdata(dev);
7115 	struct igb_adapter *adapter = netdev_priv(netdev);
7116 	struct pci_dev *pdev = adapter->pdev;
7117 	struct e1000_hw *hw = &adapter->hw;
7118 	unsigned long event = *(unsigned long *)data;
7119 
7120 	switch (event) {
7121 	case DCA_PROVIDER_ADD:
7122 		/* if already enabled, don't do it again */
7123 		if (adapter->flags & IGB_FLAG_DCA_ENABLED)
7124 			break;
7125 		if (dca_add_requester(dev) == 0) {
7126 			adapter->flags |= IGB_FLAG_DCA_ENABLED;
7127 			dev_info(&pdev->dev, "DCA enabled\n");
7128 			igb_setup_dca(adapter);
7129 			break;
7130 		}
7131 		fallthrough; /* since DCA is disabled. */
7132 	case DCA_PROVIDER_REMOVE:
7133 		if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
7134 			/* without this a class_device is left
7135 			 * hanging around in the sysfs model
7136 			 */
7137 			dca_remove_requester(dev);
7138 			dev_info(&pdev->dev, "DCA disabled\n");
7139 			adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
7140 			wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
7141 		}
7142 		break;
7143 	}
7144 
7145 	return 0;
7146 }
7147 
7148 static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
7149 			  void *p)
7150 {
7151 	int ret_val;
7152 
7153 	ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
7154 					 __igb_notify_dca);
7155 
7156 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
7157 }
7158 #endif /* CONFIG_IGB_DCA */
7159 
7160 #ifdef CONFIG_PCI_IOV
7161 static int igb_vf_configure(struct igb_adapter *adapter, int vf)
7162 {
7163 	unsigned char mac_addr[ETH_ALEN];
7164 
7165 	eth_zero_addr(mac_addr);
7166 	igb_set_vf_mac(adapter, vf, mac_addr);
7167 
7168 	/* By default spoof check is enabled for all VFs */
7169 	adapter->vf_data[vf].spoofchk_enabled = true;
7170 
7171 	/* By default VFs are not trusted */
7172 	adapter->vf_data[vf].trusted = false;
7173 
7174 	return 0;
7175 }
7176 
7177 #endif
7178 static void igb_ping_all_vfs(struct igb_adapter *adapter)
7179 {
7180 	struct e1000_hw *hw = &adapter->hw;
7181 	u32 ping;
7182 	int i;
7183 
7184 	for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
7185 		ping = E1000_PF_CONTROL_MSG;
7186 		if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
7187 			ping |= E1000_VT_MSGTYPE_CTS;
7188 		igb_write_mbx(hw, &ping, 1, i);
7189 	}
7190 }
7191 
7192 static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
7193 {
7194 	struct e1000_hw *hw = &adapter->hw;
7195 	u32 vmolr = rd32(E1000_VMOLR(vf));
7196 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7197 
7198 	vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
7199 			    IGB_VF_FLAG_MULTI_PROMISC);
7200 	vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
7201 
7202 	if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
7203 		vmolr |= E1000_VMOLR_MPME;
7204 		vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
7205 		*msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
7206 	} else {
7207 		/* if we have hashes and we are clearing a multicast promisc
7208 		 * flag we need to write the hashes to the MTA as this step
7209 		 * was previously skipped
7210 		 */
7211 		if (vf_data->num_vf_mc_hashes > 30) {
7212 			vmolr |= E1000_VMOLR_MPME;
7213 		} else if (vf_data->num_vf_mc_hashes) {
7214 			int j;
7215 
7216 			vmolr |= E1000_VMOLR_ROMPE;
7217 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
7218 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
7219 		}
7220 	}
7221 
7222 	wr32(E1000_VMOLR(vf), vmolr);
7223 
7224 	/* there are flags left unprocessed, likely not supported */
7225 	if (*msgbuf & E1000_VT_MSGINFO_MASK)
7226 		return -EINVAL;
7227 
7228 	return 0;
7229 }
7230 
7231 static int igb_set_vf_multicasts(struct igb_adapter *adapter,
7232 				  u32 *msgbuf, u32 vf)
7233 {
7234 	int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
7235 	u16 *hash_list = (u16 *)&msgbuf[1];
7236 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7237 	int i;
7238 
7239 	/* salt away the number of multicast addresses assigned
7240 	 * to this VF for later use to restore when the PF multi cast
7241 	 * list changes
7242 	 */
7243 	vf_data->num_vf_mc_hashes = n;
7244 
7245 	/* only up to 30 hash values supported */
7246 	if (n > 30)
7247 		n = 30;
7248 
7249 	/* store the hashes for later use */
7250 	for (i = 0; i < n; i++)
7251 		vf_data->vf_mc_hashes[i] = hash_list[i];
7252 
7253 	/* Flush and reset the mta with the new values */
7254 	igb_set_rx_mode(adapter->netdev);
7255 
7256 	return 0;
7257 }
7258 
7259 static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
7260 {
7261 	struct e1000_hw *hw = &adapter->hw;
7262 	struct vf_data_storage *vf_data;
7263 	int i, j;
7264 
7265 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
7266 		u32 vmolr = rd32(E1000_VMOLR(i));
7267 
7268 		vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
7269 
7270 		vf_data = &adapter->vf_data[i];
7271 
7272 		if ((vf_data->num_vf_mc_hashes > 30) ||
7273 		    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
7274 			vmolr |= E1000_VMOLR_MPME;
7275 		} else if (vf_data->num_vf_mc_hashes) {
7276 			vmolr |= E1000_VMOLR_ROMPE;
7277 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
7278 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
7279 		}
7280 		wr32(E1000_VMOLR(i), vmolr);
7281 	}
7282 }
7283 
7284 static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
7285 {
7286 	struct e1000_hw *hw = &adapter->hw;
7287 	u32 pool_mask, vlvf_mask, i;
7288 
7289 	/* create mask for VF and other pools */
7290 	pool_mask = E1000_VLVF_POOLSEL_MASK;
7291 	vlvf_mask = BIT(E1000_VLVF_POOLSEL_SHIFT + vf);
7292 
7293 	/* drop PF from pool bits */
7294 	pool_mask &= ~BIT(E1000_VLVF_POOLSEL_SHIFT +
7295 			     adapter->vfs_allocated_count);
7296 
7297 	/* Find the vlan filter for this id */
7298 	for (i = E1000_VLVF_ARRAY_SIZE; i--;) {
7299 		u32 vlvf = rd32(E1000_VLVF(i));
7300 		u32 vfta_mask, vid, vfta;
7301 
7302 		/* remove the vf from the pool */
7303 		if (!(vlvf & vlvf_mask))
7304 			continue;
7305 
7306 		/* clear out bit from VLVF */
7307 		vlvf ^= vlvf_mask;
7308 
7309 		/* if other pools are present, just remove ourselves */
7310 		if (vlvf & pool_mask)
7311 			goto update_vlvfb;
7312 
7313 		/* if PF is present, leave VFTA */
7314 		if (vlvf & E1000_VLVF_POOLSEL_MASK)
7315 			goto update_vlvf;
7316 
7317 		vid = vlvf & E1000_VLVF_VLANID_MASK;
7318 		vfta_mask = BIT(vid % 32);
7319 
7320 		/* clear bit from VFTA */
7321 		vfta = adapter->shadow_vfta[vid / 32];
7322 		if (vfta & vfta_mask)
7323 			hw->mac.ops.write_vfta(hw, vid / 32, vfta ^ vfta_mask);
7324 update_vlvf:
7325 		/* clear pool selection enable */
7326 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
7327 			vlvf &= E1000_VLVF_POOLSEL_MASK;
7328 		else
7329 			vlvf = 0;
7330 update_vlvfb:
7331 		/* clear pool bits */
7332 		wr32(E1000_VLVF(i), vlvf);
7333 	}
7334 }
7335 
7336 static int igb_find_vlvf_entry(struct e1000_hw *hw, u32 vlan)
7337 {
7338 	u32 vlvf;
7339 	int idx;
7340 
7341 	/* short cut the special case */
7342 	if (vlan == 0)
7343 		return 0;
7344 
7345 	/* Search for the VLAN id in the VLVF entries */
7346 	for (idx = E1000_VLVF_ARRAY_SIZE; --idx;) {
7347 		vlvf = rd32(E1000_VLVF(idx));
7348 		if ((vlvf & VLAN_VID_MASK) == vlan)
7349 			break;
7350 	}
7351 
7352 	return idx;
7353 }
7354 
7355 static void igb_update_pf_vlvf(struct igb_adapter *adapter, u32 vid)
7356 {
7357 	struct e1000_hw *hw = &adapter->hw;
7358 	u32 bits, pf_id;
7359 	int idx;
7360 
7361 	idx = igb_find_vlvf_entry(hw, vid);
7362 	if (!idx)
7363 		return;
7364 
7365 	/* See if any other pools are set for this VLAN filter
7366 	 * entry other than the PF.
7367 	 */
7368 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
7369 	bits = ~BIT(pf_id) & E1000_VLVF_POOLSEL_MASK;
7370 	bits &= rd32(E1000_VLVF(idx));
7371 
7372 	/* Disable the filter so this falls into the default pool. */
7373 	if (!bits) {
7374 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
7375 			wr32(E1000_VLVF(idx), BIT(pf_id));
7376 		else
7377 			wr32(E1000_VLVF(idx), 0);
7378 	}
7379 }
7380 
7381 static s32 igb_set_vf_vlan(struct igb_adapter *adapter, u32 vid,
7382 			   bool add, u32 vf)
7383 {
7384 	int pf_id = adapter->vfs_allocated_count;
7385 	struct e1000_hw *hw = &adapter->hw;
7386 	int err;
7387 
7388 	/* If VLAN overlaps with one the PF is currently monitoring make
7389 	 * sure that we are able to allocate a VLVF entry.  This may be
7390 	 * redundant but it guarantees PF will maintain visibility to
7391 	 * the VLAN.
7392 	 */
7393 	if (add && test_bit(vid, adapter->active_vlans)) {
7394 		err = igb_vfta_set(hw, vid, pf_id, true, false);
7395 		if (err)
7396 			return err;
7397 	}
7398 
7399 	err = igb_vfta_set(hw, vid, vf, add, false);
7400 
7401 	if (add && !err)
7402 		return err;
7403 
7404 	/* If we failed to add the VF VLAN or we are removing the VF VLAN
7405 	 * we may need to drop the PF pool bit in order to allow us to free
7406 	 * up the VLVF resources.
7407 	 */
7408 	if (test_bit(vid, adapter->active_vlans) ||
7409 	    (adapter->flags & IGB_FLAG_VLAN_PROMISC))
7410 		igb_update_pf_vlvf(adapter, vid);
7411 
7412 	return err;
7413 }
7414 
7415 static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
7416 {
7417 	struct e1000_hw *hw = &adapter->hw;
7418 
7419 	if (vid)
7420 		wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
7421 	else
7422 		wr32(E1000_VMVIR(vf), 0);
7423 }
7424 
7425 static int igb_enable_port_vlan(struct igb_adapter *adapter, int vf,
7426 				u16 vlan, u8 qos)
7427 {
7428 	int err;
7429 
7430 	err = igb_set_vf_vlan(adapter, vlan, true, vf);
7431 	if (err)
7432 		return err;
7433 
7434 	igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
7435 	igb_set_vmolr(adapter, vf, !vlan);
7436 
7437 	/* revoke access to previous VLAN */
7438 	if (vlan != adapter->vf_data[vf].pf_vlan)
7439 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
7440 				false, vf);
7441 
7442 	adapter->vf_data[vf].pf_vlan = vlan;
7443 	adapter->vf_data[vf].pf_qos = qos;
7444 	igb_set_vf_vlan_strip(adapter, vf, true);
7445 	dev_info(&adapter->pdev->dev,
7446 		 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
7447 	if (test_bit(__IGB_DOWN, &adapter->state)) {
7448 		dev_warn(&adapter->pdev->dev,
7449 			 "The VF VLAN has been set, but the PF device is not up.\n");
7450 		dev_warn(&adapter->pdev->dev,
7451 			 "Bring the PF device up before attempting to use the VF device.\n");
7452 	}
7453 
7454 	return err;
7455 }
7456 
7457 static int igb_disable_port_vlan(struct igb_adapter *adapter, int vf)
7458 {
7459 	/* Restore tagless access via VLAN 0 */
7460 	igb_set_vf_vlan(adapter, 0, true, vf);
7461 
7462 	igb_set_vmvir(adapter, 0, vf);
7463 	igb_set_vmolr(adapter, vf, true);
7464 
7465 	/* Remove any PF assigned VLAN */
7466 	if (adapter->vf_data[vf].pf_vlan)
7467 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
7468 				false, vf);
7469 
7470 	adapter->vf_data[vf].pf_vlan = 0;
7471 	adapter->vf_data[vf].pf_qos = 0;
7472 	igb_set_vf_vlan_strip(adapter, vf, false);
7473 
7474 	return 0;
7475 }
7476 
7477 static int igb_ndo_set_vf_vlan(struct net_device *netdev, int vf,
7478 			       u16 vlan, u8 qos, __be16 vlan_proto)
7479 {
7480 	struct igb_adapter *adapter = netdev_priv(netdev);
7481 
7482 	if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
7483 		return -EINVAL;
7484 
7485 	if (vlan_proto != htons(ETH_P_8021Q))
7486 		return -EPROTONOSUPPORT;
7487 
7488 	return (vlan || qos) ? igb_enable_port_vlan(adapter, vf, vlan, qos) :
7489 			       igb_disable_port_vlan(adapter, vf);
7490 }
7491 
7492 static int igb_set_vf_vlan_msg(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
7493 {
7494 	int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
7495 	int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
7496 	int ret;
7497 
7498 	if (adapter->vf_data[vf].pf_vlan)
7499 		return -1;
7500 
7501 	/* VLAN 0 is a special case, don't allow it to be removed */
7502 	if (!vid && !add)
7503 		return 0;
7504 
7505 	ret = igb_set_vf_vlan(adapter, vid, !!add, vf);
7506 	if (!ret)
7507 		igb_set_vf_vlan_strip(adapter, vf, !!vid);
7508 	return ret;
7509 }
7510 
7511 static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
7512 {
7513 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7514 
7515 	/* clear flags - except flag that indicates PF has set the MAC */
7516 	vf_data->flags &= IGB_VF_FLAG_PF_SET_MAC;
7517 	vf_data->last_nack = jiffies;
7518 
7519 	/* reset vlans for device */
7520 	igb_clear_vf_vfta(adapter, vf);
7521 	igb_set_vf_vlan(adapter, vf_data->pf_vlan, true, vf);
7522 	igb_set_vmvir(adapter, vf_data->pf_vlan |
7523 			       (vf_data->pf_qos << VLAN_PRIO_SHIFT), vf);
7524 	igb_set_vmolr(adapter, vf, !vf_data->pf_vlan);
7525 	igb_set_vf_vlan_strip(adapter, vf, !!(vf_data->pf_vlan));
7526 
7527 	/* reset multicast table array for vf */
7528 	adapter->vf_data[vf].num_vf_mc_hashes = 0;
7529 
7530 	/* Flush and reset the mta with the new values */
7531 	igb_set_rx_mode(adapter->netdev);
7532 }
7533 
7534 static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
7535 {
7536 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7537 
7538 	/* clear mac address as we were hotplug removed/added */
7539 	if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
7540 		eth_zero_addr(vf_mac);
7541 
7542 	/* process remaining reset events */
7543 	igb_vf_reset(adapter, vf);
7544 }
7545 
7546 static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
7547 {
7548 	struct e1000_hw *hw = &adapter->hw;
7549 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7550 	u32 reg, msgbuf[3] = {};
7551 	u8 *addr = (u8 *)(&msgbuf[1]);
7552 
7553 	/* process all the same items cleared in a function level reset */
7554 	igb_vf_reset(adapter, vf);
7555 
7556 	/* set vf mac address */
7557 	igb_set_vf_mac(adapter, vf, vf_mac);
7558 
7559 	/* enable transmit and receive for vf */
7560 	reg = rd32(E1000_VFTE);
7561 	wr32(E1000_VFTE, reg | BIT(vf));
7562 	reg = rd32(E1000_VFRE);
7563 	wr32(E1000_VFRE, reg | BIT(vf));
7564 
7565 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
7566 
7567 	/* reply to reset with ack and vf mac address */
7568 	if (!is_zero_ether_addr(vf_mac)) {
7569 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
7570 		memcpy(addr, vf_mac, ETH_ALEN);
7571 	} else {
7572 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
7573 	}
7574 	igb_write_mbx(hw, msgbuf, 3, vf);
7575 }
7576 
7577 static void igb_flush_mac_table(struct igb_adapter *adapter)
7578 {
7579 	struct e1000_hw *hw = &adapter->hw;
7580 	int i;
7581 
7582 	for (i = 0; i < hw->mac.rar_entry_count; i++) {
7583 		adapter->mac_table[i].state &= ~IGB_MAC_STATE_IN_USE;
7584 		eth_zero_addr(adapter->mac_table[i].addr);
7585 		adapter->mac_table[i].queue = 0;
7586 		igb_rar_set_index(adapter, i);
7587 	}
7588 }
7589 
7590 static int igb_available_rars(struct igb_adapter *adapter, u8 queue)
7591 {
7592 	struct e1000_hw *hw = &adapter->hw;
7593 	/* do not count rar entries reserved for VFs MAC addresses */
7594 	int rar_entries = hw->mac.rar_entry_count -
7595 			  adapter->vfs_allocated_count;
7596 	int i, count = 0;
7597 
7598 	for (i = 0; i < rar_entries; i++) {
7599 		/* do not count default entries */
7600 		if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT)
7601 			continue;
7602 
7603 		/* do not count "in use" entries for different queues */
7604 		if ((adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE) &&
7605 		    (adapter->mac_table[i].queue != queue))
7606 			continue;
7607 
7608 		count++;
7609 	}
7610 
7611 	return count;
7612 }
7613 
7614 /* Set default MAC address for the PF in the first RAR entry */
7615 static void igb_set_default_mac_filter(struct igb_adapter *adapter)
7616 {
7617 	struct igb_mac_addr *mac_table = &adapter->mac_table[0];
7618 
7619 	ether_addr_copy(mac_table->addr, adapter->hw.mac.addr);
7620 	mac_table->queue = adapter->vfs_allocated_count;
7621 	mac_table->state = IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7622 
7623 	igb_rar_set_index(adapter, 0);
7624 }
7625 
7626 /* If the filter to be added and an already existing filter express
7627  * the same address and address type, it should be possible to only
7628  * override the other configurations, for example the queue to steer
7629  * traffic.
7630  */
7631 static bool igb_mac_entry_can_be_used(const struct igb_mac_addr *entry,
7632 				      const u8 *addr, const u8 flags)
7633 {
7634 	if (!(entry->state & IGB_MAC_STATE_IN_USE))
7635 		return true;
7636 
7637 	if ((entry->state & IGB_MAC_STATE_SRC_ADDR) !=
7638 	    (flags & IGB_MAC_STATE_SRC_ADDR))
7639 		return false;
7640 
7641 	if (!ether_addr_equal(addr, entry->addr))
7642 		return false;
7643 
7644 	return true;
7645 }
7646 
7647 /* Add a MAC filter for 'addr' directing matching traffic to 'queue',
7648  * 'flags' is used to indicate what kind of match is made, match is by
7649  * default for the destination address, if matching by source address
7650  * is desired the flag IGB_MAC_STATE_SRC_ADDR can be used.
7651  */
7652 static int igb_add_mac_filter_flags(struct igb_adapter *adapter,
7653 				    const u8 *addr, const u8 queue,
7654 				    const u8 flags)
7655 {
7656 	struct e1000_hw *hw = &adapter->hw;
7657 	int rar_entries = hw->mac.rar_entry_count -
7658 			  adapter->vfs_allocated_count;
7659 	int i;
7660 
7661 	if (is_zero_ether_addr(addr))
7662 		return -EINVAL;
7663 
7664 	/* Search for the first empty entry in the MAC table.
7665 	 * Do not touch entries at the end of the table reserved for the VF MAC
7666 	 * addresses.
7667 	 */
7668 	for (i = 0; i < rar_entries; i++) {
7669 		if (!igb_mac_entry_can_be_used(&adapter->mac_table[i],
7670 					       addr, flags))
7671 			continue;
7672 
7673 		ether_addr_copy(adapter->mac_table[i].addr, addr);
7674 		adapter->mac_table[i].queue = queue;
7675 		adapter->mac_table[i].state |= IGB_MAC_STATE_IN_USE | flags;
7676 
7677 		igb_rar_set_index(adapter, i);
7678 		return i;
7679 	}
7680 
7681 	return -ENOSPC;
7682 }
7683 
7684 static int igb_add_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7685 			      const u8 queue)
7686 {
7687 	return igb_add_mac_filter_flags(adapter, addr, queue, 0);
7688 }
7689 
7690 /* Remove a MAC filter for 'addr' directing matching traffic to
7691  * 'queue', 'flags' is used to indicate what kind of match need to be
7692  * removed, match is by default for the destination address, if
7693  * matching by source address is to be removed the flag
7694  * IGB_MAC_STATE_SRC_ADDR can be used.
7695  */
7696 static int igb_del_mac_filter_flags(struct igb_adapter *adapter,
7697 				    const u8 *addr, const u8 queue,
7698 				    const u8 flags)
7699 {
7700 	struct e1000_hw *hw = &adapter->hw;
7701 	int rar_entries = hw->mac.rar_entry_count -
7702 			  adapter->vfs_allocated_count;
7703 	int i;
7704 
7705 	if (is_zero_ether_addr(addr))
7706 		return -EINVAL;
7707 
7708 	/* Search for matching entry in the MAC table based on given address
7709 	 * and queue. Do not touch entries at the end of the table reserved
7710 	 * for the VF MAC addresses.
7711 	 */
7712 	for (i = 0; i < rar_entries; i++) {
7713 		if (!(adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE))
7714 			continue;
7715 		if ((adapter->mac_table[i].state & flags) != flags)
7716 			continue;
7717 		if (adapter->mac_table[i].queue != queue)
7718 			continue;
7719 		if (!ether_addr_equal(adapter->mac_table[i].addr, addr))
7720 			continue;
7721 
7722 		/* When a filter for the default address is "deleted",
7723 		 * we return it to its initial configuration
7724 		 */
7725 		if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT) {
7726 			adapter->mac_table[i].state =
7727 				IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7728 			adapter->mac_table[i].queue =
7729 				adapter->vfs_allocated_count;
7730 		} else {
7731 			adapter->mac_table[i].state = 0;
7732 			adapter->mac_table[i].queue = 0;
7733 			eth_zero_addr(adapter->mac_table[i].addr);
7734 		}
7735 
7736 		igb_rar_set_index(adapter, i);
7737 		return 0;
7738 	}
7739 
7740 	return -ENOENT;
7741 }
7742 
7743 static int igb_del_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7744 			      const u8 queue)
7745 {
7746 	return igb_del_mac_filter_flags(adapter, addr, queue, 0);
7747 }
7748 
7749 int igb_add_mac_steering_filter(struct igb_adapter *adapter,
7750 				const u8 *addr, u8 queue, u8 flags)
7751 {
7752 	struct e1000_hw *hw = &adapter->hw;
7753 
7754 	/* In theory, this should be supported on 82575 as well, but
7755 	 * that part wasn't easily accessible during development.
7756 	 */
7757 	if (hw->mac.type != e1000_i210)
7758 		return -EOPNOTSUPP;
7759 
7760 	return igb_add_mac_filter_flags(adapter, addr, queue,
7761 					IGB_MAC_STATE_QUEUE_STEERING | flags);
7762 }
7763 
7764 int igb_del_mac_steering_filter(struct igb_adapter *adapter,
7765 				const u8 *addr, u8 queue, u8 flags)
7766 {
7767 	return igb_del_mac_filter_flags(adapter, addr, queue,
7768 					IGB_MAC_STATE_QUEUE_STEERING | flags);
7769 }
7770 
7771 static int igb_uc_sync(struct net_device *netdev, const unsigned char *addr)
7772 {
7773 	struct igb_adapter *adapter = netdev_priv(netdev);
7774 	int ret;
7775 
7776 	ret = igb_add_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7777 
7778 	return min_t(int, ret, 0);
7779 }
7780 
7781 static int igb_uc_unsync(struct net_device *netdev, const unsigned char *addr)
7782 {
7783 	struct igb_adapter *adapter = netdev_priv(netdev);
7784 
7785 	igb_del_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7786 
7787 	return 0;
7788 }
7789 
7790 static int igb_set_vf_mac_filter(struct igb_adapter *adapter, const int vf,
7791 				 const u32 info, const u8 *addr)
7792 {
7793 	struct pci_dev *pdev = adapter->pdev;
7794 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7795 	struct list_head *pos;
7796 	struct vf_mac_filter *entry = NULL;
7797 	int ret = 0;
7798 
7799 	if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7800 	    !vf_data->trusted) {
7801 		dev_warn(&pdev->dev,
7802 			 "VF %d requested MAC filter but is administratively denied\n",
7803 			  vf);
7804 		return -EINVAL;
7805 	}
7806 	if (!is_valid_ether_addr(addr)) {
7807 		dev_warn(&pdev->dev,
7808 			 "VF %d attempted to set invalid MAC filter\n",
7809 			  vf);
7810 		return -EINVAL;
7811 	}
7812 
7813 	switch (info) {
7814 	case E1000_VF_MAC_FILTER_CLR:
7815 		/* remove all unicast MAC filters related to the current VF */
7816 		list_for_each(pos, &adapter->vf_macs.l) {
7817 			entry = list_entry(pos, struct vf_mac_filter, l);
7818 			if (entry->vf == vf) {
7819 				entry->vf = -1;
7820 				entry->free = true;
7821 				igb_del_mac_filter(adapter, entry->vf_mac, vf);
7822 			}
7823 		}
7824 		break;
7825 	case E1000_VF_MAC_FILTER_ADD:
7826 		/* try to find empty slot in the list */
7827 		list_for_each(pos, &adapter->vf_macs.l) {
7828 			entry = list_entry(pos, struct vf_mac_filter, l);
7829 			if (entry->free)
7830 				break;
7831 		}
7832 
7833 		if (entry && entry->free) {
7834 			entry->free = false;
7835 			entry->vf = vf;
7836 			ether_addr_copy(entry->vf_mac, addr);
7837 
7838 			ret = igb_add_mac_filter(adapter, addr, vf);
7839 			ret = min_t(int, ret, 0);
7840 		} else {
7841 			ret = -ENOSPC;
7842 		}
7843 
7844 		if (ret == -ENOSPC)
7845 			dev_warn(&pdev->dev,
7846 				 "VF %d has requested MAC filter but there is no space for it\n",
7847 				 vf);
7848 		break;
7849 	default:
7850 		ret = -EINVAL;
7851 		break;
7852 	}
7853 
7854 	return ret;
7855 }
7856 
7857 static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
7858 {
7859 	struct pci_dev *pdev = adapter->pdev;
7860 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7861 	u32 info = msg[0] & E1000_VT_MSGINFO_MASK;
7862 
7863 	/* The VF MAC Address is stored in a packed array of bytes
7864 	 * starting at the second 32 bit word of the msg array
7865 	 */
7866 	unsigned char *addr = (unsigned char *)&msg[1];
7867 	int ret = 0;
7868 
7869 	if (!info) {
7870 		if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7871 		    !vf_data->trusted) {
7872 			dev_warn(&pdev->dev,
7873 				 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
7874 				 vf);
7875 			return -EINVAL;
7876 		}
7877 
7878 		if (!is_valid_ether_addr(addr)) {
7879 			dev_warn(&pdev->dev,
7880 				 "VF %d attempted to set invalid MAC\n",
7881 				 vf);
7882 			return -EINVAL;
7883 		}
7884 
7885 		ret = igb_set_vf_mac(adapter, vf, addr);
7886 	} else {
7887 		ret = igb_set_vf_mac_filter(adapter, vf, info, addr);
7888 	}
7889 
7890 	return ret;
7891 }
7892 
7893 static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
7894 {
7895 	struct e1000_hw *hw = &adapter->hw;
7896 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7897 	u32 msg = E1000_VT_MSGTYPE_NACK;
7898 
7899 	/* if device isn't clear to send it shouldn't be reading either */
7900 	if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
7901 	    time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
7902 		igb_write_mbx(hw, &msg, 1, vf);
7903 		vf_data->last_nack = jiffies;
7904 	}
7905 }
7906 
7907 static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
7908 {
7909 	struct pci_dev *pdev = adapter->pdev;
7910 	u32 msgbuf[E1000_VFMAILBOX_SIZE];
7911 	struct e1000_hw *hw = &adapter->hw;
7912 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7913 	s32 retval;
7914 
7915 	retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf, false);
7916 
7917 	if (retval) {
7918 		/* if receive failed revoke VF CTS stats and restart init */
7919 		dev_err(&pdev->dev, "Error receiving message from VF\n");
7920 		vf_data->flags &= ~IGB_VF_FLAG_CTS;
7921 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7922 			goto unlock;
7923 		goto out;
7924 	}
7925 
7926 	/* this is a message we already processed, do nothing */
7927 	if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
7928 		goto unlock;
7929 
7930 	/* until the vf completes a reset it should not be
7931 	 * allowed to start any configuration.
7932 	 */
7933 	if (msgbuf[0] == E1000_VF_RESET) {
7934 		/* unlocks mailbox */
7935 		igb_vf_reset_msg(adapter, vf);
7936 		return;
7937 	}
7938 
7939 	if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
7940 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7941 			goto unlock;
7942 		retval = -1;
7943 		goto out;
7944 	}
7945 
7946 	switch ((msgbuf[0] & 0xFFFF)) {
7947 	case E1000_VF_SET_MAC_ADDR:
7948 		retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
7949 		break;
7950 	case E1000_VF_SET_PROMISC:
7951 		retval = igb_set_vf_promisc(adapter, msgbuf, vf);
7952 		break;
7953 	case E1000_VF_SET_MULTICAST:
7954 		retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
7955 		break;
7956 	case E1000_VF_SET_LPE:
7957 		retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
7958 		break;
7959 	case E1000_VF_SET_VLAN:
7960 		retval = -1;
7961 		if (vf_data->pf_vlan)
7962 			dev_warn(&pdev->dev,
7963 				 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
7964 				 vf);
7965 		else
7966 			retval = igb_set_vf_vlan_msg(adapter, msgbuf, vf);
7967 		break;
7968 	default:
7969 		dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
7970 		retval = -1;
7971 		break;
7972 	}
7973 
7974 	msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
7975 out:
7976 	/* notify the VF of the results of what it sent us */
7977 	if (retval)
7978 		msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
7979 	else
7980 		msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
7981 
7982 	/* unlocks mailbox */
7983 	igb_write_mbx(hw, msgbuf, 1, vf);
7984 	return;
7985 
7986 unlock:
7987 	igb_unlock_mbx(hw, vf);
7988 }
7989 
7990 static void igb_msg_task(struct igb_adapter *adapter)
7991 {
7992 	struct e1000_hw *hw = &adapter->hw;
7993 	unsigned long flags;
7994 	u32 vf;
7995 
7996 	spin_lock_irqsave(&adapter->vfs_lock, flags);
7997 	for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
7998 		/* process any reset requests */
7999 		if (!igb_check_for_rst(hw, vf))
8000 			igb_vf_reset_event(adapter, vf);
8001 
8002 		/* process any messages pending */
8003 		if (!igb_check_for_msg(hw, vf))
8004 			igb_rcv_msg_from_vf(adapter, vf);
8005 
8006 		/* process any acks */
8007 		if (!igb_check_for_ack(hw, vf))
8008 			igb_rcv_ack_from_vf(adapter, vf);
8009 	}
8010 	spin_unlock_irqrestore(&adapter->vfs_lock, flags);
8011 }
8012 
8013 /**
8014  *  igb_set_uta - Set unicast filter table address
8015  *  @adapter: board private structure
8016  *  @set: boolean indicating if we are setting or clearing bits
8017  *
8018  *  The unicast table address is a register array of 32-bit registers.
8019  *  The table is meant to be used in a way similar to how the MTA is used
8020  *  however due to certain limitations in the hardware it is necessary to
8021  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
8022  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
8023  **/
8024 static void igb_set_uta(struct igb_adapter *adapter, bool set)
8025 {
8026 	struct e1000_hw *hw = &adapter->hw;
8027 	u32 uta = set ? ~0 : 0;
8028 	int i;
8029 
8030 	/* we only need to do this if VMDq is enabled */
8031 	if (!adapter->vfs_allocated_count)
8032 		return;
8033 
8034 	for (i = hw->mac.uta_reg_count; i--;)
8035 		array_wr32(E1000_UTA, i, uta);
8036 }
8037 
8038 /**
8039  *  igb_intr_msi - Interrupt Handler
8040  *  @irq: interrupt number
8041  *  @data: pointer to a network interface device structure
8042  **/
8043 static irqreturn_t igb_intr_msi(int irq, void *data)
8044 {
8045 	struct igb_adapter *adapter = data;
8046 	struct igb_q_vector *q_vector = adapter->q_vector[0];
8047 	struct e1000_hw *hw = &adapter->hw;
8048 	/* read ICR disables interrupts using IAM */
8049 	u32 icr = rd32(E1000_ICR);
8050 
8051 	igb_write_itr(q_vector);
8052 
8053 	if (icr & E1000_ICR_DRSTA)
8054 		schedule_work(&adapter->reset_task);
8055 
8056 	if (icr & E1000_ICR_DOUTSYNC) {
8057 		/* HW is reporting DMA is out of sync */
8058 		adapter->stats.doosync++;
8059 	}
8060 
8061 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
8062 		hw->mac.get_link_status = 1;
8063 		if (!test_bit(__IGB_DOWN, &adapter->state))
8064 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
8065 	}
8066 
8067 	if (icr & E1000_ICR_TS)
8068 		igb_tsync_interrupt(adapter);
8069 
8070 	napi_schedule(&q_vector->napi);
8071 
8072 	return IRQ_HANDLED;
8073 }
8074 
8075 /**
8076  *  igb_intr - Legacy Interrupt Handler
8077  *  @irq: interrupt number
8078  *  @data: pointer to a network interface device structure
8079  **/
8080 static irqreturn_t igb_intr(int irq, void *data)
8081 {
8082 	struct igb_adapter *adapter = data;
8083 	struct igb_q_vector *q_vector = adapter->q_vector[0];
8084 	struct e1000_hw *hw = &adapter->hw;
8085 	/* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
8086 	 * need for the IMC write
8087 	 */
8088 	u32 icr = rd32(E1000_ICR);
8089 
8090 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
8091 	 * not set, then the adapter didn't send an interrupt
8092 	 */
8093 	if (!(icr & E1000_ICR_INT_ASSERTED))
8094 		return IRQ_NONE;
8095 
8096 	igb_write_itr(q_vector);
8097 
8098 	if (icr & E1000_ICR_DRSTA)
8099 		schedule_work(&adapter->reset_task);
8100 
8101 	if (icr & E1000_ICR_DOUTSYNC) {
8102 		/* HW is reporting DMA is out of sync */
8103 		adapter->stats.doosync++;
8104 	}
8105 
8106 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
8107 		hw->mac.get_link_status = 1;
8108 		/* guard against interrupt when we're going down */
8109 		if (!test_bit(__IGB_DOWN, &adapter->state))
8110 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
8111 	}
8112 
8113 	if (icr & E1000_ICR_TS)
8114 		igb_tsync_interrupt(adapter);
8115 
8116 	napi_schedule(&q_vector->napi);
8117 
8118 	return IRQ_HANDLED;
8119 }
8120 
8121 static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
8122 {
8123 	struct igb_adapter *adapter = q_vector->adapter;
8124 	struct e1000_hw *hw = &adapter->hw;
8125 
8126 	if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
8127 	    (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
8128 		if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
8129 			igb_set_itr(q_vector);
8130 		else
8131 			igb_update_ring_itr(q_vector);
8132 	}
8133 
8134 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
8135 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
8136 			wr32(E1000_EIMS, q_vector->eims_value);
8137 		else
8138 			igb_irq_enable(adapter);
8139 	}
8140 }
8141 
8142 /**
8143  *  igb_poll - NAPI Rx polling callback
8144  *  @napi: napi polling structure
8145  *  @budget: count of how many packets we should handle
8146  **/
8147 static int igb_poll(struct napi_struct *napi, int budget)
8148 {
8149 	struct igb_q_vector *q_vector = container_of(napi,
8150 						     struct igb_q_vector,
8151 						     napi);
8152 	bool clean_complete = true;
8153 	int work_done = 0;
8154 
8155 #ifdef CONFIG_IGB_DCA
8156 	if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
8157 		igb_update_dca(q_vector);
8158 #endif
8159 	if (q_vector->tx.ring)
8160 		clean_complete = igb_clean_tx_irq(q_vector, budget);
8161 
8162 	if (q_vector->rx.ring) {
8163 		int cleaned = igb_clean_rx_irq(q_vector, budget);
8164 
8165 		work_done += cleaned;
8166 		if (cleaned >= budget)
8167 			clean_complete = false;
8168 	}
8169 
8170 	/* If all work not completed, return budget and keep polling */
8171 	if (!clean_complete)
8172 		return budget;
8173 
8174 	/* Exit the polling mode, but don't re-enable interrupts if stack might
8175 	 * poll us due to busy-polling
8176 	 */
8177 	if (likely(napi_complete_done(napi, work_done)))
8178 		igb_ring_irq_enable(q_vector);
8179 
8180 	return work_done;
8181 }
8182 
8183 /**
8184  *  igb_clean_tx_irq - Reclaim resources after transmit completes
8185  *  @q_vector: pointer to q_vector containing needed info
8186  *  @napi_budget: Used to determine if we are in netpoll
8187  *
8188  *  returns true if ring is completely cleaned
8189  **/
8190 static bool igb_clean_tx_irq(struct igb_q_vector *q_vector, int napi_budget)
8191 {
8192 	struct igb_adapter *adapter = q_vector->adapter;
8193 	struct igb_ring *tx_ring = q_vector->tx.ring;
8194 	struct igb_tx_buffer *tx_buffer;
8195 	union e1000_adv_tx_desc *tx_desc;
8196 	unsigned int total_bytes = 0, total_packets = 0;
8197 	unsigned int budget = q_vector->tx.work_limit;
8198 	unsigned int i = tx_ring->next_to_clean;
8199 
8200 	if (test_bit(__IGB_DOWN, &adapter->state))
8201 		return true;
8202 
8203 	tx_buffer = &tx_ring->tx_buffer_info[i];
8204 	tx_desc = IGB_TX_DESC(tx_ring, i);
8205 	i -= tx_ring->count;
8206 
8207 	do {
8208 		union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
8209 
8210 		/* if next_to_watch is not set then there is no work pending */
8211 		if (!eop_desc)
8212 			break;
8213 
8214 		/* prevent any other reads prior to eop_desc */
8215 		smp_rmb();
8216 
8217 		/* if DD is not set pending work has not been completed */
8218 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
8219 			break;
8220 
8221 		/* clear next_to_watch to prevent false hangs */
8222 		tx_buffer->next_to_watch = NULL;
8223 
8224 		/* update the statistics for this packet */
8225 		total_bytes += tx_buffer->bytecount;
8226 		total_packets += tx_buffer->gso_segs;
8227 
8228 		/* free the skb */
8229 		if (tx_buffer->type == IGB_TYPE_SKB)
8230 			napi_consume_skb(tx_buffer->skb, napi_budget);
8231 		else
8232 			xdp_return_frame(tx_buffer->xdpf);
8233 
8234 		/* unmap skb header data */
8235 		dma_unmap_single(tx_ring->dev,
8236 				 dma_unmap_addr(tx_buffer, dma),
8237 				 dma_unmap_len(tx_buffer, len),
8238 				 DMA_TO_DEVICE);
8239 
8240 		/* clear tx_buffer data */
8241 		dma_unmap_len_set(tx_buffer, len, 0);
8242 
8243 		/* clear last DMA location and unmap remaining buffers */
8244 		while (tx_desc != eop_desc) {
8245 			tx_buffer++;
8246 			tx_desc++;
8247 			i++;
8248 			if (unlikely(!i)) {
8249 				i -= tx_ring->count;
8250 				tx_buffer = tx_ring->tx_buffer_info;
8251 				tx_desc = IGB_TX_DESC(tx_ring, 0);
8252 			}
8253 
8254 			/* unmap any remaining paged data */
8255 			if (dma_unmap_len(tx_buffer, len)) {
8256 				dma_unmap_page(tx_ring->dev,
8257 					       dma_unmap_addr(tx_buffer, dma),
8258 					       dma_unmap_len(tx_buffer, len),
8259 					       DMA_TO_DEVICE);
8260 				dma_unmap_len_set(tx_buffer, len, 0);
8261 			}
8262 		}
8263 
8264 		/* move us one more past the eop_desc for start of next pkt */
8265 		tx_buffer++;
8266 		tx_desc++;
8267 		i++;
8268 		if (unlikely(!i)) {
8269 			i -= tx_ring->count;
8270 			tx_buffer = tx_ring->tx_buffer_info;
8271 			tx_desc = IGB_TX_DESC(tx_ring, 0);
8272 		}
8273 
8274 		/* issue prefetch for next Tx descriptor */
8275 		prefetch(tx_desc);
8276 
8277 		/* update budget accounting */
8278 		budget--;
8279 	} while (likely(budget));
8280 
8281 	netdev_tx_completed_queue(txring_txq(tx_ring),
8282 				  total_packets, total_bytes);
8283 	i += tx_ring->count;
8284 	tx_ring->next_to_clean = i;
8285 	u64_stats_update_begin(&tx_ring->tx_syncp);
8286 	tx_ring->tx_stats.bytes += total_bytes;
8287 	tx_ring->tx_stats.packets += total_packets;
8288 	u64_stats_update_end(&tx_ring->tx_syncp);
8289 	q_vector->tx.total_bytes += total_bytes;
8290 	q_vector->tx.total_packets += total_packets;
8291 
8292 	if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
8293 		struct e1000_hw *hw = &adapter->hw;
8294 
8295 		/* Detect a transmit hang in hardware, this serializes the
8296 		 * check with the clearing of time_stamp and movement of i
8297 		 */
8298 		clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
8299 		if (tx_buffer->next_to_watch &&
8300 		    time_after(jiffies, tx_buffer->time_stamp +
8301 			       (adapter->tx_timeout_factor * HZ)) &&
8302 		    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
8303 
8304 			/* detected Tx unit hang */
8305 			dev_err(tx_ring->dev,
8306 				"Detected Tx Unit Hang\n"
8307 				"  Tx Queue             <%d>\n"
8308 				"  TDH                  <%x>\n"
8309 				"  TDT                  <%x>\n"
8310 				"  next_to_use          <%x>\n"
8311 				"  next_to_clean        <%x>\n"
8312 				"buffer_info[next_to_clean]\n"
8313 				"  time_stamp           <%lx>\n"
8314 				"  next_to_watch        <%p>\n"
8315 				"  jiffies              <%lx>\n"
8316 				"  desc.status          <%x>\n",
8317 				tx_ring->queue_index,
8318 				rd32(E1000_TDH(tx_ring->reg_idx)),
8319 				readl(tx_ring->tail),
8320 				tx_ring->next_to_use,
8321 				tx_ring->next_to_clean,
8322 				tx_buffer->time_stamp,
8323 				tx_buffer->next_to_watch,
8324 				jiffies,
8325 				tx_buffer->next_to_watch->wb.status);
8326 			netif_stop_subqueue(tx_ring->netdev,
8327 					    tx_ring->queue_index);
8328 
8329 			/* we are about to reset, no point in enabling stuff */
8330 			return true;
8331 		}
8332 	}
8333 
8334 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
8335 	if (unlikely(total_packets &&
8336 	    netif_carrier_ok(tx_ring->netdev) &&
8337 	    igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
8338 		/* Make sure that anybody stopping the queue after this
8339 		 * sees the new next_to_clean.
8340 		 */
8341 		smp_mb();
8342 		if (__netif_subqueue_stopped(tx_ring->netdev,
8343 					     tx_ring->queue_index) &&
8344 		    !(test_bit(__IGB_DOWN, &adapter->state))) {
8345 			netif_wake_subqueue(tx_ring->netdev,
8346 					    tx_ring->queue_index);
8347 
8348 			u64_stats_update_begin(&tx_ring->tx_syncp);
8349 			tx_ring->tx_stats.restart_queue++;
8350 			u64_stats_update_end(&tx_ring->tx_syncp);
8351 		}
8352 	}
8353 
8354 	return !!budget;
8355 }
8356 
8357 /**
8358  *  igb_reuse_rx_page - page flip buffer and store it back on the ring
8359  *  @rx_ring: rx descriptor ring to store buffers on
8360  *  @old_buff: donor buffer to have page reused
8361  *
8362  *  Synchronizes page for reuse by the adapter
8363  **/
8364 static void igb_reuse_rx_page(struct igb_ring *rx_ring,
8365 			      struct igb_rx_buffer *old_buff)
8366 {
8367 	struct igb_rx_buffer *new_buff;
8368 	u16 nta = rx_ring->next_to_alloc;
8369 
8370 	new_buff = &rx_ring->rx_buffer_info[nta];
8371 
8372 	/* update, and store next to alloc */
8373 	nta++;
8374 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
8375 
8376 	/* Transfer page from old buffer to new buffer.
8377 	 * Move each member individually to avoid possible store
8378 	 * forwarding stalls.
8379 	 */
8380 	new_buff->dma		= old_buff->dma;
8381 	new_buff->page		= old_buff->page;
8382 	new_buff->page_offset	= old_buff->page_offset;
8383 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
8384 }
8385 
8386 static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer,
8387 				  int rx_buf_pgcnt)
8388 {
8389 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
8390 	struct page *page = rx_buffer->page;
8391 
8392 	/* avoid re-using remote and pfmemalloc pages */
8393 	if (!dev_page_is_reusable(page))
8394 		return false;
8395 
8396 #if (PAGE_SIZE < 8192)
8397 	/* if we are only owner of page we can reuse it */
8398 	if (unlikely((rx_buf_pgcnt - pagecnt_bias) > 1))
8399 		return false;
8400 #else
8401 #define IGB_LAST_OFFSET \
8402 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IGB_RXBUFFER_2048)
8403 
8404 	if (rx_buffer->page_offset > IGB_LAST_OFFSET)
8405 		return false;
8406 #endif
8407 
8408 	/* If we have drained the page fragment pool we need to update
8409 	 * the pagecnt_bias and page count so that we fully restock the
8410 	 * number of references the driver holds.
8411 	 */
8412 	if (unlikely(pagecnt_bias == 1)) {
8413 		page_ref_add(page, USHRT_MAX - 1);
8414 		rx_buffer->pagecnt_bias = USHRT_MAX;
8415 	}
8416 
8417 	return true;
8418 }
8419 
8420 /**
8421  *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
8422  *  @rx_ring: rx descriptor ring to transact packets on
8423  *  @rx_buffer: buffer containing page to add
8424  *  @skb: sk_buff to place the data into
8425  *  @size: size of buffer to be added
8426  *
8427  *  This function will add the data contained in rx_buffer->page to the skb.
8428  **/
8429 static void igb_add_rx_frag(struct igb_ring *rx_ring,
8430 			    struct igb_rx_buffer *rx_buffer,
8431 			    struct sk_buff *skb,
8432 			    unsigned int size)
8433 {
8434 #if (PAGE_SIZE < 8192)
8435 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8436 #else
8437 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
8438 				SKB_DATA_ALIGN(IGB_SKB_PAD + size) :
8439 				SKB_DATA_ALIGN(size);
8440 #endif
8441 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
8442 			rx_buffer->page_offset, size, truesize);
8443 #if (PAGE_SIZE < 8192)
8444 	rx_buffer->page_offset ^= truesize;
8445 #else
8446 	rx_buffer->page_offset += truesize;
8447 #endif
8448 }
8449 
8450 static struct sk_buff *igb_construct_skb(struct igb_ring *rx_ring,
8451 					 struct igb_rx_buffer *rx_buffer,
8452 					 struct xdp_buff *xdp,
8453 					 ktime_t timestamp)
8454 {
8455 #if (PAGE_SIZE < 8192)
8456 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8457 #else
8458 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
8459 					       xdp->data_hard_start);
8460 #endif
8461 	unsigned int size = xdp->data_end - xdp->data;
8462 	unsigned int headlen;
8463 	struct sk_buff *skb;
8464 
8465 	/* prefetch first cache line of first page */
8466 	net_prefetch(xdp->data);
8467 
8468 	/* allocate a skb to store the frags */
8469 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
8470 	if (unlikely(!skb))
8471 		return NULL;
8472 
8473 	if (timestamp)
8474 		skb_hwtstamps(skb)->hwtstamp = timestamp;
8475 
8476 	/* Determine available headroom for copy */
8477 	headlen = size;
8478 	if (headlen > IGB_RX_HDR_LEN)
8479 		headlen = eth_get_headlen(skb->dev, xdp->data, IGB_RX_HDR_LEN);
8480 
8481 	/* align pull length to size of long to optimize memcpy performance */
8482 	memcpy(__skb_put(skb, headlen), xdp->data, ALIGN(headlen, sizeof(long)));
8483 
8484 	/* update all of the pointers */
8485 	size -= headlen;
8486 	if (size) {
8487 		skb_add_rx_frag(skb, 0, rx_buffer->page,
8488 				(xdp->data + headlen) - page_address(rx_buffer->page),
8489 				size, truesize);
8490 #if (PAGE_SIZE < 8192)
8491 		rx_buffer->page_offset ^= truesize;
8492 #else
8493 		rx_buffer->page_offset += truesize;
8494 #endif
8495 	} else {
8496 		rx_buffer->pagecnt_bias++;
8497 	}
8498 
8499 	return skb;
8500 }
8501 
8502 static struct sk_buff *igb_build_skb(struct igb_ring *rx_ring,
8503 				     struct igb_rx_buffer *rx_buffer,
8504 				     struct xdp_buff *xdp,
8505 				     ktime_t timestamp)
8506 {
8507 #if (PAGE_SIZE < 8192)
8508 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8509 #else
8510 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
8511 				SKB_DATA_ALIGN(xdp->data_end -
8512 					       xdp->data_hard_start);
8513 #endif
8514 	unsigned int metasize = xdp->data - xdp->data_meta;
8515 	struct sk_buff *skb;
8516 
8517 	/* prefetch first cache line of first page */
8518 	net_prefetch(xdp->data_meta);
8519 
8520 	/* build an skb around the page buffer */
8521 	skb = napi_build_skb(xdp->data_hard_start, truesize);
8522 	if (unlikely(!skb))
8523 		return NULL;
8524 
8525 	/* update pointers within the skb to store the data */
8526 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
8527 	__skb_put(skb, xdp->data_end - xdp->data);
8528 
8529 	if (metasize)
8530 		skb_metadata_set(skb, metasize);
8531 
8532 	if (timestamp)
8533 		skb_hwtstamps(skb)->hwtstamp = timestamp;
8534 
8535 	/* update buffer offset */
8536 #if (PAGE_SIZE < 8192)
8537 	rx_buffer->page_offset ^= truesize;
8538 #else
8539 	rx_buffer->page_offset += truesize;
8540 #endif
8541 
8542 	return skb;
8543 }
8544 
8545 static struct sk_buff *igb_run_xdp(struct igb_adapter *adapter,
8546 				   struct igb_ring *rx_ring,
8547 				   struct xdp_buff *xdp)
8548 {
8549 	int err, result = IGB_XDP_PASS;
8550 	struct bpf_prog *xdp_prog;
8551 	u32 act;
8552 
8553 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
8554 
8555 	if (!xdp_prog)
8556 		goto xdp_out;
8557 
8558 	prefetchw(xdp->data_hard_start); /* xdp_frame write */
8559 
8560 	act = bpf_prog_run_xdp(xdp_prog, xdp);
8561 	switch (act) {
8562 	case XDP_PASS:
8563 		break;
8564 	case XDP_TX:
8565 		result = igb_xdp_xmit_back(adapter, xdp);
8566 		if (result == IGB_XDP_CONSUMED)
8567 			goto out_failure;
8568 		break;
8569 	case XDP_REDIRECT:
8570 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
8571 		if (err)
8572 			goto out_failure;
8573 		result = IGB_XDP_REDIR;
8574 		break;
8575 	default:
8576 		bpf_warn_invalid_xdp_action(adapter->netdev, xdp_prog, act);
8577 		fallthrough;
8578 	case XDP_ABORTED:
8579 out_failure:
8580 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
8581 		fallthrough;
8582 	case XDP_DROP:
8583 		result = IGB_XDP_CONSUMED;
8584 		break;
8585 	}
8586 xdp_out:
8587 	return ERR_PTR(-result);
8588 }
8589 
8590 static unsigned int igb_rx_frame_truesize(struct igb_ring *rx_ring,
8591 					  unsigned int size)
8592 {
8593 	unsigned int truesize;
8594 
8595 #if (PAGE_SIZE < 8192)
8596 	truesize = igb_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
8597 #else
8598 	truesize = ring_uses_build_skb(rx_ring) ?
8599 		SKB_DATA_ALIGN(IGB_SKB_PAD + size) +
8600 		SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
8601 		SKB_DATA_ALIGN(size);
8602 #endif
8603 	return truesize;
8604 }
8605 
8606 static void igb_rx_buffer_flip(struct igb_ring *rx_ring,
8607 			       struct igb_rx_buffer *rx_buffer,
8608 			       unsigned int size)
8609 {
8610 	unsigned int truesize = igb_rx_frame_truesize(rx_ring, size);
8611 #if (PAGE_SIZE < 8192)
8612 	rx_buffer->page_offset ^= truesize;
8613 #else
8614 	rx_buffer->page_offset += truesize;
8615 #endif
8616 }
8617 
8618 static inline void igb_rx_checksum(struct igb_ring *ring,
8619 				   union e1000_adv_rx_desc *rx_desc,
8620 				   struct sk_buff *skb)
8621 {
8622 	skb_checksum_none_assert(skb);
8623 
8624 	/* Ignore Checksum bit is set */
8625 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
8626 		return;
8627 
8628 	/* Rx checksum disabled via ethtool */
8629 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
8630 		return;
8631 
8632 	/* TCP/UDP checksum error bit is set */
8633 	if (igb_test_staterr(rx_desc,
8634 			     E1000_RXDEXT_STATERR_TCPE |
8635 			     E1000_RXDEXT_STATERR_IPE)) {
8636 		/* work around errata with sctp packets where the TCPE aka
8637 		 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
8638 		 * packets, (aka let the stack check the crc32c)
8639 		 */
8640 		if (!((skb->len == 60) &&
8641 		      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
8642 			u64_stats_update_begin(&ring->rx_syncp);
8643 			ring->rx_stats.csum_err++;
8644 			u64_stats_update_end(&ring->rx_syncp);
8645 		}
8646 		/* let the stack verify checksum errors */
8647 		return;
8648 	}
8649 	/* It must be a TCP or UDP packet with a valid checksum */
8650 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
8651 				      E1000_RXD_STAT_UDPCS))
8652 		skb->ip_summed = CHECKSUM_UNNECESSARY;
8653 
8654 	dev_dbg(ring->dev, "cksum success: bits %08X\n",
8655 		le32_to_cpu(rx_desc->wb.upper.status_error));
8656 }
8657 
8658 static inline void igb_rx_hash(struct igb_ring *ring,
8659 			       union e1000_adv_rx_desc *rx_desc,
8660 			       struct sk_buff *skb)
8661 {
8662 	if (ring->netdev->features & NETIF_F_RXHASH)
8663 		skb_set_hash(skb,
8664 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
8665 			     PKT_HASH_TYPE_L3);
8666 }
8667 
8668 /**
8669  *  igb_is_non_eop - process handling of non-EOP buffers
8670  *  @rx_ring: Rx ring being processed
8671  *  @rx_desc: Rx descriptor for current buffer
8672  *
8673  *  This function updates next to clean.  If the buffer is an EOP buffer
8674  *  this function exits returning false, otherwise it will place the
8675  *  sk_buff in the next buffer to be chained and return true indicating
8676  *  that this is in fact a non-EOP buffer.
8677  **/
8678 static bool igb_is_non_eop(struct igb_ring *rx_ring,
8679 			   union e1000_adv_rx_desc *rx_desc)
8680 {
8681 	u32 ntc = rx_ring->next_to_clean + 1;
8682 
8683 	/* fetch, update, and store next to clean */
8684 	ntc = (ntc < rx_ring->count) ? ntc : 0;
8685 	rx_ring->next_to_clean = ntc;
8686 
8687 	prefetch(IGB_RX_DESC(rx_ring, ntc));
8688 
8689 	if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
8690 		return false;
8691 
8692 	return true;
8693 }
8694 
8695 /**
8696  *  igb_cleanup_headers - Correct corrupted or empty headers
8697  *  @rx_ring: rx descriptor ring packet is being transacted on
8698  *  @rx_desc: pointer to the EOP Rx descriptor
8699  *  @skb: pointer to current skb being fixed
8700  *
8701  *  Address the case where we are pulling data in on pages only
8702  *  and as such no data is present in the skb header.
8703  *
8704  *  In addition if skb is not at least 60 bytes we need to pad it so that
8705  *  it is large enough to qualify as a valid Ethernet frame.
8706  *
8707  *  Returns true if an error was encountered and skb was freed.
8708  **/
8709 static bool igb_cleanup_headers(struct igb_ring *rx_ring,
8710 				union e1000_adv_rx_desc *rx_desc,
8711 				struct sk_buff *skb)
8712 {
8713 	/* XDP packets use error pointer so abort at this point */
8714 	if (IS_ERR(skb))
8715 		return true;
8716 
8717 	if (unlikely((igb_test_staterr(rx_desc,
8718 				       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
8719 		struct net_device *netdev = rx_ring->netdev;
8720 		if (!(netdev->features & NETIF_F_RXALL)) {
8721 			dev_kfree_skb_any(skb);
8722 			return true;
8723 		}
8724 	}
8725 
8726 	/* if eth_skb_pad returns an error the skb was freed */
8727 	if (eth_skb_pad(skb))
8728 		return true;
8729 
8730 	return false;
8731 }
8732 
8733 /**
8734  *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
8735  *  @rx_ring: rx descriptor ring packet is being transacted on
8736  *  @rx_desc: pointer to the EOP Rx descriptor
8737  *  @skb: pointer to current skb being populated
8738  *
8739  *  This function checks the ring, descriptor, and packet information in
8740  *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
8741  *  other fields within the skb.
8742  **/
8743 static void igb_process_skb_fields(struct igb_ring *rx_ring,
8744 				   union e1000_adv_rx_desc *rx_desc,
8745 				   struct sk_buff *skb)
8746 {
8747 	struct net_device *dev = rx_ring->netdev;
8748 
8749 	igb_rx_hash(rx_ring, rx_desc, skb);
8750 
8751 	igb_rx_checksum(rx_ring, rx_desc, skb);
8752 
8753 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
8754 	    !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
8755 		igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
8756 
8757 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
8758 	    igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
8759 		u16 vid;
8760 
8761 		if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
8762 		    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
8763 			vid = be16_to_cpu((__force __be16)rx_desc->wb.upper.vlan);
8764 		else
8765 			vid = le16_to_cpu(rx_desc->wb.upper.vlan);
8766 
8767 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
8768 	}
8769 
8770 	skb_record_rx_queue(skb, rx_ring->queue_index);
8771 
8772 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
8773 }
8774 
8775 static unsigned int igb_rx_offset(struct igb_ring *rx_ring)
8776 {
8777 	return ring_uses_build_skb(rx_ring) ? IGB_SKB_PAD : 0;
8778 }
8779 
8780 static struct igb_rx_buffer *igb_get_rx_buffer(struct igb_ring *rx_ring,
8781 					       const unsigned int size, int *rx_buf_pgcnt)
8782 {
8783 	struct igb_rx_buffer *rx_buffer;
8784 
8785 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
8786 	*rx_buf_pgcnt =
8787 #if (PAGE_SIZE < 8192)
8788 		page_count(rx_buffer->page);
8789 #else
8790 		0;
8791 #endif
8792 	prefetchw(rx_buffer->page);
8793 
8794 	/* we are reusing so sync this buffer for CPU use */
8795 	dma_sync_single_range_for_cpu(rx_ring->dev,
8796 				      rx_buffer->dma,
8797 				      rx_buffer->page_offset,
8798 				      size,
8799 				      DMA_FROM_DEVICE);
8800 
8801 	rx_buffer->pagecnt_bias--;
8802 
8803 	return rx_buffer;
8804 }
8805 
8806 static void igb_put_rx_buffer(struct igb_ring *rx_ring,
8807 			      struct igb_rx_buffer *rx_buffer, int rx_buf_pgcnt)
8808 {
8809 	if (igb_can_reuse_rx_page(rx_buffer, rx_buf_pgcnt)) {
8810 		/* hand second half of page back to the ring */
8811 		igb_reuse_rx_page(rx_ring, rx_buffer);
8812 	} else {
8813 		/* We are not reusing the buffer so unmap it and free
8814 		 * any references we are holding to it
8815 		 */
8816 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
8817 				     igb_rx_pg_size(rx_ring), DMA_FROM_DEVICE,
8818 				     IGB_RX_DMA_ATTR);
8819 		__page_frag_cache_drain(rx_buffer->page,
8820 					rx_buffer->pagecnt_bias);
8821 	}
8822 
8823 	/* clear contents of rx_buffer */
8824 	rx_buffer->page = NULL;
8825 }
8826 
8827 static int igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
8828 {
8829 	struct igb_adapter *adapter = q_vector->adapter;
8830 	struct igb_ring *rx_ring = q_vector->rx.ring;
8831 	struct sk_buff *skb = rx_ring->skb;
8832 	unsigned int total_bytes = 0, total_packets = 0;
8833 	u16 cleaned_count = igb_desc_unused(rx_ring);
8834 	unsigned int xdp_xmit = 0;
8835 	struct xdp_buff xdp;
8836 	u32 frame_sz = 0;
8837 	int rx_buf_pgcnt;
8838 
8839 	/* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
8840 #if (PAGE_SIZE < 8192)
8841 	frame_sz = igb_rx_frame_truesize(rx_ring, 0);
8842 #endif
8843 	xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq);
8844 
8845 	while (likely(total_packets < budget)) {
8846 		union e1000_adv_rx_desc *rx_desc;
8847 		struct igb_rx_buffer *rx_buffer;
8848 		ktime_t timestamp = 0;
8849 		int pkt_offset = 0;
8850 		unsigned int size;
8851 		void *pktbuf;
8852 
8853 		/* return some buffers to hardware, one at a time is too slow */
8854 		if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
8855 			igb_alloc_rx_buffers(rx_ring, cleaned_count);
8856 			cleaned_count = 0;
8857 		}
8858 
8859 		rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
8860 		size = le16_to_cpu(rx_desc->wb.upper.length);
8861 		if (!size)
8862 			break;
8863 
8864 		/* This memory barrier is needed to keep us from reading
8865 		 * any other fields out of the rx_desc until we know the
8866 		 * descriptor has been written back
8867 		 */
8868 		dma_rmb();
8869 
8870 		rx_buffer = igb_get_rx_buffer(rx_ring, size, &rx_buf_pgcnt);
8871 		pktbuf = page_address(rx_buffer->page) + rx_buffer->page_offset;
8872 
8873 		/* pull rx packet timestamp if available and valid */
8874 		if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
8875 			int ts_hdr_len;
8876 
8877 			ts_hdr_len = igb_ptp_rx_pktstamp(rx_ring->q_vector,
8878 							 pktbuf, &timestamp);
8879 
8880 			pkt_offset += ts_hdr_len;
8881 			size -= ts_hdr_len;
8882 		}
8883 
8884 		/* retrieve a buffer from the ring */
8885 		if (!skb) {
8886 			unsigned char *hard_start = pktbuf - igb_rx_offset(rx_ring);
8887 			unsigned int offset = pkt_offset + igb_rx_offset(rx_ring);
8888 
8889 			xdp_prepare_buff(&xdp, hard_start, offset, size, true);
8890 			xdp_buff_clear_frags_flag(&xdp);
8891 #if (PAGE_SIZE > 4096)
8892 			/* At larger PAGE_SIZE, frame_sz depend on len size */
8893 			xdp.frame_sz = igb_rx_frame_truesize(rx_ring, size);
8894 #endif
8895 			skb = igb_run_xdp(adapter, rx_ring, &xdp);
8896 		}
8897 
8898 		if (IS_ERR(skb)) {
8899 			unsigned int xdp_res = -PTR_ERR(skb);
8900 
8901 			if (xdp_res & (IGB_XDP_TX | IGB_XDP_REDIR)) {
8902 				xdp_xmit |= xdp_res;
8903 				igb_rx_buffer_flip(rx_ring, rx_buffer, size);
8904 			} else {
8905 				rx_buffer->pagecnt_bias++;
8906 			}
8907 			total_packets++;
8908 			total_bytes += size;
8909 		} else if (skb)
8910 			igb_add_rx_frag(rx_ring, rx_buffer, skb, size);
8911 		else if (ring_uses_build_skb(rx_ring))
8912 			skb = igb_build_skb(rx_ring, rx_buffer, &xdp,
8913 					    timestamp);
8914 		else
8915 			skb = igb_construct_skb(rx_ring, rx_buffer,
8916 						&xdp, timestamp);
8917 
8918 		/* exit if we failed to retrieve a buffer */
8919 		if (!skb) {
8920 			rx_ring->rx_stats.alloc_failed++;
8921 			rx_buffer->pagecnt_bias++;
8922 			break;
8923 		}
8924 
8925 		igb_put_rx_buffer(rx_ring, rx_buffer, rx_buf_pgcnt);
8926 		cleaned_count++;
8927 
8928 		/* fetch next buffer in frame if non-eop */
8929 		if (igb_is_non_eop(rx_ring, rx_desc))
8930 			continue;
8931 
8932 		/* verify the packet layout is correct */
8933 		if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
8934 			skb = NULL;
8935 			continue;
8936 		}
8937 
8938 		/* probably a little skewed due to removing CRC */
8939 		total_bytes += skb->len;
8940 
8941 		/* populate checksum, timestamp, VLAN, and protocol */
8942 		igb_process_skb_fields(rx_ring, rx_desc, skb);
8943 
8944 		napi_gro_receive(&q_vector->napi, skb);
8945 
8946 		/* reset skb pointer */
8947 		skb = NULL;
8948 
8949 		/* update budget accounting */
8950 		total_packets++;
8951 	}
8952 
8953 	/* place incomplete frames back on ring for completion */
8954 	rx_ring->skb = skb;
8955 
8956 	if (xdp_xmit & IGB_XDP_REDIR)
8957 		xdp_do_flush();
8958 
8959 	if (xdp_xmit & IGB_XDP_TX) {
8960 		struct igb_ring *tx_ring = igb_xdp_tx_queue_mapping(adapter);
8961 
8962 		igb_xdp_ring_update_tail(tx_ring);
8963 	}
8964 
8965 	u64_stats_update_begin(&rx_ring->rx_syncp);
8966 	rx_ring->rx_stats.packets += total_packets;
8967 	rx_ring->rx_stats.bytes += total_bytes;
8968 	u64_stats_update_end(&rx_ring->rx_syncp);
8969 	q_vector->rx.total_packets += total_packets;
8970 	q_vector->rx.total_bytes += total_bytes;
8971 
8972 	if (cleaned_count)
8973 		igb_alloc_rx_buffers(rx_ring, cleaned_count);
8974 
8975 	return total_packets;
8976 }
8977 
8978 static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
8979 				  struct igb_rx_buffer *bi)
8980 {
8981 	struct page *page = bi->page;
8982 	dma_addr_t dma;
8983 
8984 	/* since we are recycling buffers we should seldom need to alloc */
8985 	if (likely(page))
8986 		return true;
8987 
8988 	/* alloc new page for storage */
8989 	page = dev_alloc_pages(igb_rx_pg_order(rx_ring));
8990 	if (unlikely(!page)) {
8991 		rx_ring->rx_stats.alloc_failed++;
8992 		return false;
8993 	}
8994 
8995 	/* map page for use */
8996 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
8997 				 igb_rx_pg_size(rx_ring),
8998 				 DMA_FROM_DEVICE,
8999 				 IGB_RX_DMA_ATTR);
9000 
9001 	/* if mapping failed free memory back to system since
9002 	 * there isn't much point in holding memory we can't use
9003 	 */
9004 	if (dma_mapping_error(rx_ring->dev, dma)) {
9005 		__free_pages(page, igb_rx_pg_order(rx_ring));
9006 
9007 		rx_ring->rx_stats.alloc_failed++;
9008 		return false;
9009 	}
9010 
9011 	bi->dma = dma;
9012 	bi->page = page;
9013 	bi->page_offset = igb_rx_offset(rx_ring);
9014 	page_ref_add(page, USHRT_MAX - 1);
9015 	bi->pagecnt_bias = USHRT_MAX;
9016 
9017 	return true;
9018 }
9019 
9020 /**
9021  *  igb_alloc_rx_buffers - Replace used receive buffers
9022  *  @rx_ring: rx descriptor ring to allocate new receive buffers
9023  *  @cleaned_count: count of buffers to allocate
9024  **/
9025 void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
9026 {
9027 	union e1000_adv_rx_desc *rx_desc;
9028 	struct igb_rx_buffer *bi;
9029 	u16 i = rx_ring->next_to_use;
9030 	u16 bufsz;
9031 
9032 	/* nothing to do */
9033 	if (!cleaned_count)
9034 		return;
9035 
9036 	rx_desc = IGB_RX_DESC(rx_ring, i);
9037 	bi = &rx_ring->rx_buffer_info[i];
9038 	i -= rx_ring->count;
9039 
9040 	bufsz = igb_rx_bufsz(rx_ring);
9041 
9042 	do {
9043 		if (!igb_alloc_mapped_page(rx_ring, bi))
9044 			break;
9045 
9046 		/* sync the buffer for use by the device */
9047 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
9048 						 bi->page_offset, bufsz,
9049 						 DMA_FROM_DEVICE);
9050 
9051 		/* Refresh the desc even if buffer_addrs didn't change
9052 		 * because each write-back erases this info.
9053 		 */
9054 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
9055 
9056 		rx_desc++;
9057 		bi++;
9058 		i++;
9059 		if (unlikely(!i)) {
9060 			rx_desc = IGB_RX_DESC(rx_ring, 0);
9061 			bi = rx_ring->rx_buffer_info;
9062 			i -= rx_ring->count;
9063 		}
9064 
9065 		/* clear the length for the next_to_use descriptor */
9066 		rx_desc->wb.upper.length = 0;
9067 
9068 		cleaned_count--;
9069 	} while (cleaned_count);
9070 
9071 	i += rx_ring->count;
9072 
9073 	if (rx_ring->next_to_use != i) {
9074 		/* record the next descriptor to use */
9075 		rx_ring->next_to_use = i;
9076 
9077 		/* update next to alloc since we have filled the ring */
9078 		rx_ring->next_to_alloc = i;
9079 
9080 		/* Force memory writes to complete before letting h/w
9081 		 * know there are new descriptors to fetch.  (Only
9082 		 * applicable for weak-ordered memory model archs,
9083 		 * such as IA-64).
9084 		 */
9085 		dma_wmb();
9086 		writel(i, rx_ring->tail);
9087 	}
9088 }
9089 
9090 /**
9091  * igb_mii_ioctl -
9092  * @netdev: pointer to netdev struct
9093  * @ifr: interface structure
9094  * @cmd: ioctl command to execute
9095  **/
9096 static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
9097 {
9098 	struct igb_adapter *adapter = netdev_priv(netdev);
9099 	struct mii_ioctl_data *data = if_mii(ifr);
9100 
9101 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
9102 		return -EOPNOTSUPP;
9103 
9104 	switch (cmd) {
9105 	case SIOCGMIIPHY:
9106 		data->phy_id = adapter->hw.phy.addr;
9107 		break;
9108 	case SIOCGMIIREG:
9109 		if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
9110 				     &data->val_out))
9111 			return -EIO;
9112 		break;
9113 	case SIOCSMIIREG:
9114 	default:
9115 		return -EOPNOTSUPP;
9116 	}
9117 	return 0;
9118 }
9119 
9120 /**
9121  * igb_ioctl -
9122  * @netdev: pointer to netdev struct
9123  * @ifr: interface structure
9124  * @cmd: ioctl command to execute
9125  **/
9126 static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
9127 {
9128 	switch (cmd) {
9129 	case SIOCGMIIPHY:
9130 	case SIOCGMIIREG:
9131 	case SIOCSMIIREG:
9132 		return igb_mii_ioctl(netdev, ifr, cmd);
9133 	case SIOCGHWTSTAMP:
9134 		return igb_ptp_get_ts_config(netdev, ifr);
9135 	case SIOCSHWTSTAMP:
9136 		return igb_ptp_set_ts_config(netdev, ifr);
9137 	default:
9138 		return -EOPNOTSUPP;
9139 	}
9140 }
9141 
9142 void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
9143 {
9144 	struct igb_adapter *adapter = hw->back;
9145 
9146 	pci_read_config_word(adapter->pdev, reg, value);
9147 }
9148 
9149 void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
9150 {
9151 	struct igb_adapter *adapter = hw->back;
9152 
9153 	pci_write_config_word(adapter->pdev, reg, *value);
9154 }
9155 
9156 s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
9157 {
9158 	struct igb_adapter *adapter = hw->back;
9159 
9160 	if (pcie_capability_read_word(adapter->pdev, reg, value))
9161 		return -E1000_ERR_CONFIG;
9162 
9163 	return 0;
9164 }
9165 
9166 s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
9167 {
9168 	struct igb_adapter *adapter = hw->back;
9169 
9170 	if (pcie_capability_write_word(adapter->pdev, reg, *value))
9171 		return -E1000_ERR_CONFIG;
9172 
9173 	return 0;
9174 }
9175 
9176 static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
9177 {
9178 	struct igb_adapter *adapter = netdev_priv(netdev);
9179 	struct e1000_hw *hw = &adapter->hw;
9180 	u32 ctrl, rctl;
9181 	bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
9182 
9183 	if (enable) {
9184 		/* enable VLAN tag insert/strip */
9185 		ctrl = rd32(E1000_CTRL);
9186 		ctrl |= E1000_CTRL_VME;
9187 		wr32(E1000_CTRL, ctrl);
9188 
9189 		/* Disable CFI check */
9190 		rctl = rd32(E1000_RCTL);
9191 		rctl &= ~E1000_RCTL_CFIEN;
9192 		wr32(E1000_RCTL, rctl);
9193 	} else {
9194 		/* disable VLAN tag insert/strip */
9195 		ctrl = rd32(E1000_CTRL);
9196 		ctrl &= ~E1000_CTRL_VME;
9197 		wr32(E1000_CTRL, ctrl);
9198 	}
9199 
9200 	igb_set_vf_vlan_strip(adapter, adapter->vfs_allocated_count, enable);
9201 }
9202 
9203 static int igb_vlan_rx_add_vid(struct net_device *netdev,
9204 			       __be16 proto, u16 vid)
9205 {
9206 	struct igb_adapter *adapter = netdev_priv(netdev);
9207 	struct e1000_hw *hw = &adapter->hw;
9208 	int pf_id = adapter->vfs_allocated_count;
9209 
9210 	/* add the filter since PF can receive vlans w/o entry in vlvf */
9211 	if (!vid || !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
9212 		igb_vfta_set(hw, vid, pf_id, true, !!vid);
9213 
9214 	set_bit(vid, adapter->active_vlans);
9215 
9216 	return 0;
9217 }
9218 
9219 static int igb_vlan_rx_kill_vid(struct net_device *netdev,
9220 				__be16 proto, u16 vid)
9221 {
9222 	struct igb_adapter *adapter = netdev_priv(netdev);
9223 	int pf_id = adapter->vfs_allocated_count;
9224 	struct e1000_hw *hw = &adapter->hw;
9225 
9226 	/* remove VID from filter table */
9227 	if (vid && !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
9228 		igb_vfta_set(hw, vid, pf_id, false, true);
9229 
9230 	clear_bit(vid, adapter->active_vlans);
9231 
9232 	return 0;
9233 }
9234 
9235 static void igb_restore_vlan(struct igb_adapter *adapter)
9236 {
9237 	u16 vid = 1;
9238 
9239 	igb_vlan_mode(adapter->netdev, adapter->netdev->features);
9240 	igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
9241 
9242 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
9243 		igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
9244 }
9245 
9246 int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
9247 {
9248 	struct pci_dev *pdev = adapter->pdev;
9249 	struct e1000_mac_info *mac = &adapter->hw.mac;
9250 
9251 	mac->autoneg = 0;
9252 
9253 	/* Make sure dplx is at most 1 bit and lsb of speed is not set
9254 	 * for the switch() below to work
9255 	 */
9256 	if ((spd & 1) || (dplx & ~1))
9257 		goto err_inval;
9258 
9259 	/* Fiber NIC's only allow 1000 gbps Full duplex
9260 	 * and 100Mbps Full duplex for 100baseFx sfp
9261 	 */
9262 	if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
9263 		switch (spd + dplx) {
9264 		case SPEED_10 + DUPLEX_HALF:
9265 		case SPEED_10 + DUPLEX_FULL:
9266 		case SPEED_100 + DUPLEX_HALF:
9267 			goto err_inval;
9268 		default:
9269 			break;
9270 		}
9271 	}
9272 
9273 	switch (spd + dplx) {
9274 	case SPEED_10 + DUPLEX_HALF:
9275 		mac->forced_speed_duplex = ADVERTISE_10_HALF;
9276 		break;
9277 	case SPEED_10 + DUPLEX_FULL:
9278 		mac->forced_speed_duplex = ADVERTISE_10_FULL;
9279 		break;
9280 	case SPEED_100 + DUPLEX_HALF:
9281 		mac->forced_speed_duplex = ADVERTISE_100_HALF;
9282 		break;
9283 	case SPEED_100 + DUPLEX_FULL:
9284 		mac->forced_speed_duplex = ADVERTISE_100_FULL;
9285 		break;
9286 	case SPEED_1000 + DUPLEX_FULL:
9287 		mac->autoneg = 1;
9288 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
9289 		break;
9290 	case SPEED_1000 + DUPLEX_HALF: /* not supported */
9291 	default:
9292 		goto err_inval;
9293 	}
9294 
9295 	/* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
9296 	adapter->hw.phy.mdix = AUTO_ALL_MODES;
9297 
9298 	return 0;
9299 
9300 err_inval:
9301 	dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
9302 	return -EINVAL;
9303 }
9304 
9305 static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
9306 			  bool runtime)
9307 {
9308 	struct net_device *netdev = pci_get_drvdata(pdev);
9309 	struct igb_adapter *adapter = netdev_priv(netdev);
9310 	struct e1000_hw *hw = &adapter->hw;
9311 	u32 ctrl, rctl, status;
9312 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
9313 	bool wake;
9314 
9315 	rtnl_lock();
9316 	netif_device_detach(netdev);
9317 
9318 	if (netif_running(netdev))
9319 		__igb_close(netdev, true);
9320 
9321 	igb_ptp_suspend(adapter);
9322 
9323 	igb_clear_interrupt_scheme(adapter);
9324 	rtnl_unlock();
9325 
9326 	status = rd32(E1000_STATUS);
9327 	if (status & E1000_STATUS_LU)
9328 		wufc &= ~E1000_WUFC_LNKC;
9329 
9330 	if (wufc) {
9331 		igb_setup_rctl(adapter);
9332 		igb_set_rx_mode(netdev);
9333 
9334 		/* turn on all-multi mode if wake on multicast is enabled */
9335 		if (wufc & E1000_WUFC_MC) {
9336 			rctl = rd32(E1000_RCTL);
9337 			rctl |= E1000_RCTL_MPE;
9338 			wr32(E1000_RCTL, rctl);
9339 		}
9340 
9341 		ctrl = rd32(E1000_CTRL);
9342 		ctrl |= E1000_CTRL_ADVD3WUC;
9343 		wr32(E1000_CTRL, ctrl);
9344 
9345 		/* Allow time for pending master requests to run */
9346 		igb_disable_pcie_master(hw);
9347 
9348 		wr32(E1000_WUC, E1000_WUC_PME_EN);
9349 		wr32(E1000_WUFC, wufc);
9350 	} else {
9351 		wr32(E1000_WUC, 0);
9352 		wr32(E1000_WUFC, 0);
9353 	}
9354 
9355 	wake = wufc || adapter->en_mng_pt;
9356 	if (!wake)
9357 		igb_power_down_link(adapter);
9358 	else
9359 		igb_power_up_link(adapter);
9360 
9361 	if (enable_wake)
9362 		*enable_wake = wake;
9363 
9364 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
9365 	 * would have already happened in close and is redundant.
9366 	 */
9367 	igb_release_hw_control(adapter);
9368 
9369 	pci_disable_device(pdev);
9370 
9371 	return 0;
9372 }
9373 
9374 static void igb_deliver_wake_packet(struct net_device *netdev)
9375 {
9376 	struct igb_adapter *adapter = netdev_priv(netdev);
9377 	struct e1000_hw *hw = &adapter->hw;
9378 	struct sk_buff *skb;
9379 	u32 wupl;
9380 
9381 	wupl = rd32(E1000_WUPL) & E1000_WUPL_MASK;
9382 
9383 	/* WUPM stores only the first 128 bytes of the wake packet.
9384 	 * Read the packet only if we have the whole thing.
9385 	 */
9386 	if ((wupl == 0) || (wupl > E1000_WUPM_BYTES))
9387 		return;
9388 
9389 	skb = netdev_alloc_skb_ip_align(netdev, E1000_WUPM_BYTES);
9390 	if (!skb)
9391 		return;
9392 
9393 	skb_put(skb, wupl);
9394 
9395 	/* Ensure reads are 32-bit aligned */
9396 	wupl = roundup(wupl, 4);
9397 
9398 	memcpy_fromio(skb->data, hw->hw_addr + E1000_WUPM_REG(0), wupl);
9399 
9400 	skb->protocol = eth_type_trans(skb, netdev);
9401 	netif_rx(skb);
9402 }
9403 
9404 static int __maybe_unused igb_suspend(struct device *dev)
9405 {
9406 	return __igb_shutdown(to_pci_dev(dev), NULL, 0);
9407 }
9408 
9409 static int __maybe_unused __igb_resume(struct device *dev, bool rpm)
9410 {
9411 	struct pci_dev *pdev = to_pci_dev(dev);
9412 	struct net_device *netdev = pci_get_drvdata(pdev);
9413 	struct igb_adapter *adapter = netdev_priv(netdev);
9414 	struct e1000_hw *hw = &adapter->hw;
9415 	u32 err, val;
9416 
9417 	pci_set_power_state(pdev, PCI_D0);
9418 	pci_restore_state(pdev);
9419 	pci_save_state(pdev);
9420 
9421 	if (!pci_device_is_present(pdev))
9422 		return -ENODEV;
9423 	err = pci_enable_device_mem(pdev);
9424 	if (err) {
9425 		dev_err(&pdev->dev,
9426 			"igb: Cannot enable PCI device from suspend\n");
9427 		return err;
9428 	}
9429 	pci_set_master(pdev);
9430 
9431 	pci_enable_wake(pdev, PCI_D3hot, 0);
9432 	pci_enable_wake(pdev, PCI_D3cold, 0);
9433 
9434 	if (igb_init_interrupt_scheme(adapter, true)) {
9435 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9436 		return -ENOMEM;
9437 	}
9438 
9439 	igb_reset(adapter);
9440 
9441 	/* let the f/w know that the h/w is now under the control of the
9442 	 * driver.
9443 	 */
9444 	igb_get_hw_control(adapter);
9445 
9446 	val = rd32(E1000_WUS);
9447 	if (val & WAKE_PKT_WUS)
9448 		igb_deliver_wake_packet(netdev);
9449 
9450 	wr32(E1000_WUS, ~0);
9451 
9452 	if (!rpm)
9453 		rtnl_lock();
9454 	if (!err && netif_running(netdev))
9455 		err = __igb_open(netdev, true);
9456 
9457 	if (!err)
9458 		netif_device_attach(netdev);
9459 	if (!rpm)
9460 		rtnl_unlock();
9461 
9462 	return err;
9463 }
9464 
9465 static int __maybe_unused igb_resume(struct device *dev)
9466 {
9467 	return __igb_resume(dev, false);
9468 }
9469 
9470 static int __maybe_unused igb_runtime_idle(struct device *dev)
9471 {
9472 	struct net_device *netdev = dev_get_drvdata(dev);
9473 	struct igb_adapter *adapter = netdev_priv(netdev);
9474 
9475 	if (!igb_has_link(adapter))
9476 		pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
9477 
9478 	return -EBUSY;
9479 }
9480 
9481 static int __maybe_unused igb_runtime_suspend(struct device *dev)
9482 {
9483 	return __igb_shutdown(to_pci_dev(dev), NULL, 1);
9484 }
9485 
9486 static int __maybe_unused igb_runtime_resume(struct device *dev)
9487 {
9488 	return __igb_resume(dev, true);
9489 }
9490 
9491 static void igb_shutdown(struct pci_dev *pdev)
9492 {
9493 	bool wake;
9494 
9495 	__igb_shutdown(pdev, &wake, 0);
9496 
9497 	if (system_state == SYSTEM_POWER_OFF) {
9498 		pci_wake_from_d3(pdev, wake);
9499 		pci_set_power_state(pdev, PCI_D3hot);
9500 	}
9501 }
9502 
9503 #ifdef CONFIG_PCI_IOV
9504 static int igb_sriov_reinit(struct pci_dev *dev)
9505 {
9506 	struct net_device *netdev = pci_get_drvdata(dev);
9507 	struct igb_adapter *adapter = netdev_priv(netdev);
9508 	struct pci_dev *pdev = adapter->pdev;
9509 
9510 	rtnl_lock();
9511 
9512 	if (netif_running(netdev))
9513 		igb_close(netdev);
9514 	else
9515 		igb_reset(adapter);
9516 
9517 	igb_clear_interrupt_scheme(adapter);
9518 
9519 	igb_init_queue_configuration(adapter);
9520 
9521 	if (igb_init_interrupt_scheme(adapter, true)) {
9522 		rtnl_unlock();
9523 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9524 		return -ENOMEM;
9525 	}
9526 
9527 	if (netif_running(netdev))
9528 		igb_open(netdev);
9529 
9530 	rtnl_unlock();
9531 
9532 	return 0;
9533 }
9534 
9535 static int igb_pci_disable_sriov(struct pci_dev *dev)
9536 {
9537 	int err = igb_disable_sriov(dev);
9538 
9539 	if (!err)
9540 		err = igb_sriov_reinit(dev);
9541 
9542 	return err;
9543 }
9544 
9545 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
9546 {
9547 	int err = igb_enable_sriov(dev, num_vfs);
9548 
9549 	if (err)
9550 		goto out;
9551 
9552 	err = igb_sriov_reinit(dev);
9553 	if (!err)
9554 		return num_vfs;
9555 
9556 out:
9557 	return err;
9558 }
9559 
9560 #endif
9561 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
9562 {
9563 #ifdef CONFIG_PCI_IOV
9564 	if (num_vfs == 0)
9565 		return igb_pci_disable_sriov(dev);
9566 	else
9567 		return igb_pci_enable_sriov(dev, num_vfs);
9568 #endif
9569 	return 0;
9570 }
9571 
9572 /**
9573  *  igb_io_error_detected - called when PCI error is detected
9574  *  @pdev: Pointer to PCI device
9575  *  @state: The current pci connection state
9576  *
9577  *  This function is called after a PCI bus error affecting
9578  *  this device has been detected.
9579  **/
9580 static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
9581 					      pci_channel_state_t state)
9582 {
9583 	struct net_device *netdev = pci_get_drvdata(pdev);
9584 	struct igb_adapter *adapter = netdev_priv(netdev);
9585 
9586 	netif_device_detach(netdev);
9587 
9588 	if (state == pci_channel_io_perm_failure)
9589 		return PCI_ERS_RESULT_DISCONNECT;
9590 
9591 	if (netif_running(netdev))
9592 		igb_down(adapter);
9593 	pci_disable_device(pdev);
9594 
9595 	/* Request a slot reset. */
9596 	return PCI_ERS_RESULT_NEED_RESET;
9597 }
9598 
9599 /**
9600  *  igb_io_slot_reset - called after the pci bus has been reset.
9601  *  @pdev: Pointer to PCI device
9602  *
9603  *  Restart the card from scratch, as if from a cold-boot. Implementation
9604  *  resembles the first-half of the __igb_resume routine.
9605  **/
9606 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
9607 {
9608 	struct net_device *netdev = pci_get_drvdata(pdev);
9609 	struct igb_adapter *adapter = netdev_priv(netdev);
9610 	struct e1000_hw *hw = &adapter->hw;
9611 	pci_ers_result_t result;
9612 
9613 	if (pci_enable_device_mem(pdev)) {
9614 		dev_err(&pdev->dev,
9615 			"Cannot re-enable PCI device after reset.\n");
9616 		result = PCI_ERS_RESULT_DISCONNECT;
9617 	} else {
9618 		pci_set_master(pdev);
9619 		pci_restore_state(pdev);
9620 		pci_save_state(pdev);
9621 
9622 		pci_enable_wake(pdev, PCI_D3hot, 0);
9623 		pci_enable_wake(pdev, PCI_D3cold, 0);
9624 
9625 		/* In case of PCI error, adapter lose its HW address
9626 		 * so we should re-assign it here.
9627 		 */
9628 		hw->hw_addr = adapter->io_addr;
9629 
9630 		igb_reset(adapter);
9631 		wr32(E1000_WUS, ~0);
9632 		result = PCI_ERS_RESULT_RECOVERED;
9633 	}
9634 
9635 	return result;
9636 }
9637 
9638 /**
9639  *  igb_io_resume - called when traffic can start flowing again.
9640  *  @pdev: Pointer to PCI device
9641  *
9642  *  This callback is called when the error recovery driver tells us that
9643  *  its OK to resume normal operation. Implementation resembles the
9644  *  second-half of the __igb_resume routine.
9645  */
9646 static void igb_io_resume(struct pci_dev *pdev)
9647 {
9648 	struct net_device *netdev = pci_get_drvdata(pdev);
9649 	struct igb_adapter *adapter = netdev_priv(netdev);
9650 
9651 	if (netif_running(netdev)) {
9652 		if (igb_up(adapter)) {
9653 			dev_err(&pdev->dev, "igb_up failed after reset\n");
9654 			return;
9655 		}
9656 	}
9657 
9658 	netif_device_attach(netdev);
9659 
9660 	/* let the f/w know that the h/w is now under the control of the
9661 	 * driver.
9662 	 */
9663 	igb_get_hw_control(adapter);
9664 }
9665 
9666 /**
9667  *  igb_rar_set_index - Sync RAL[index] and RAH[index] registers with MAC table
9668  *  @adapter: Pointer to adapter structure
9669  *  @index: Index of the RAR entry which need to be synced with MAC table
9670  **/
9671 static void igb_rar_set_index(struct igb_adapter *adapter, u32 index)
9672 {
9673 	struct e1000_hw *hw = &adapter->hw;
9674 	u32 rar_low, rar_high;
9675 	u8 *addr = adapter->mac_table[index].addr;
9676 
9677 	/* HW expects these to be in network order when they are plugged
9678 	 * into the registers which are little endian.  In order to guarantee
9679 	 * that ordering we need to do an leXX_to_cpup here in order to be
9680 	 * ready for the byteswap that occurs with writel
9681 	 */
9682 	rar_low = le32_to_cpup((__le32 *)(addr));
9683 	rar_high = le16_to_cpup((__le16 *)(addr + 4));
9684 
9685 	/* Indicate to hardware the Address is Valid. */
9686 	if (adapter->mac_table[index].state & IGB_MAC_STATE_IN_USE) {
9687 		if (is_valid_ether_addr(addr))
9688 			rar_high |= E1000_RAH_AV;
9689 
9690 		if (adapter->mac_table[index].state & IGB_MAC_STATE_SRC_ADDR)
9691 			rar_high |= E1000_RAH_ASEL_SRC_ADDR;
9692 
9693 		switch (hw->mac.type) {
9694 		case e1000_82575:
9695 		case e1000_i210:
9696 			if (adapter->mac_table[index].state &
9697 			    IGB_MAC_STATE_QUEUE_STEERING)
9698 				rar_high |= E1000_RAH_QSEL_ENABLE;
9699 
9700 			rar_high |= E1000_RAH_POOL_1 *
9701 				    adapter->mac_table[index].queue;
9702 			break;
9703 		default:
9704 			rar_high |= E1000_RAH_POOL_1 <<
9705 				    adapter->mac_table[index].queue;
9706 			break;
9707 		}
9708 	}
9709 
9710 	wr32(E1000_RAL(index), rar_low);
9711 	wrfl();
9712 	wr32(E1000_RAH(index), rar_high);
9713 	wrfl();
9714 }
9715 
9716 static int igb_set_vf_mac(struct igb_adapter *adapter,
9717 			  int vf, unsigned char *mac_addr)
9718 {
9719 	struct e1000_hw *hw = &adapter->hw;
9720 	/* VF MAC addresses start at end of receive addresses and moves
9721 	 * towards the first, as a result a collision should not be possible
9722 	 */
9723 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
9724 	unsigned char *vf_mac_addr = adapter->vf_data[vf].vf_mac_addresses;
9725 
9726 	ether_addr_copy(vf_mac_addr, mac_addr);
9727 	ether_addr_copy(adapter->mac_table[rar_entry].addr, mac_addr);
9728 	adapter->mac_table[rar_entry].queue = vf;
9729 	adapter->mac_table[rar_entry].state |= IGB_MAC_STATE_IN_USE;
9730 	igb_rar_set_index(adapter, rar_entry);
9731 
9732 	return 0;
9733 }
9734 
9735 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
9736 {
9737 	struct igb_adapter *adapter = netdev_priv(netdev);
9738 
9739 	if (vf >= adapter->vfs_allocated_count)
9740 		return -EINVAL;
9741 
9742 	/* Setting the VF MAC to 0 reverts the IGB_VF_FLAG_PF_SET_MAC
9743 	 * flag and allows to overwrite the MAC via VF netdev.  This
9744 	 * is necessary to allow libvirt a way to restore the original
9745 	 * MAC after unbinding vfio-pci and reloading igbvf after shutting
9746 	 * down a VM.
9747 	 */
9748 	if (is_zero_ether_addr(mac)) {
9749 		adapter->vf_data[vf].flags &= ~IGB_VF_FLAG_PF_SET_MAC;
9750 		dev_info(&adapter->pdev->dev,
9751 			 "remove administratively set MAC on VF %d\n",
9752 			 vf);
9753 	} else if (is_valid_ether_addr(mac)) {
9754 		adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
9755 		dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n",
9756 			 mac, vf);
9757 		dev_info(&adapter->pdev->dev,
9758 			 "Reload the VF driver to make this change effective.");
9759 		/* Generate additional warning if PF is down */
9760 		if (test_bit(__IGB_DOWN, &adapter->state)) {
9761 			dev_warn(&adapter->pdev->dev,
9762 				 "The VF MAC address has been set, but the PF device is not up.\n");
9763 			dev_warn(&adapter->pdev->dev,
9764 				 "Bring the PF device up before attempting to use the VF device.\n");
9765 		}
9766 	} else {
9767 		return -EINVAL;
9768 	}
9769 	return igb_set_vf_mac(adapter, vf, mac);
9770 }
9771 
9772 static int igb_link_mbps(int internal_link_speed)
9773 {
9774 	switch (internal_link_speed) {
9775 	case SPEED_100:
9776 		return 100;
9777 	case SPEED_1000:
9778 		return 1000;
9779 	default:
9780 		return 0;
9781 	}
9782 }
9783 
9784 static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
9785 				  int link_speed)
9786 {
9787 	int rf_dec, rf_int;
9788 	u32 bcnrc_val;
9789 
9790 	if (tx_rate != 0) {
9791 		/* Calculate the rate factor values to set */
9792 		rf_int = link_speed / tx_rate;
9793 		rf_dec = (link_speed - (rf_int * tx_rate));
9794 		rf_dec = (rf_dec * BIT(E1000_RTTBCNRC_RF_INT_SHIFT)) /
9795 			 tx_rate;
9796 
9797 		bcnrc_val = E1000_RTTBCNRC_RS_ENA;
9798 		bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
9799 			      E1000_RTTBCNRC_RF_INT_MASK);
9800 		bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
9801 	} else {
9802 		bcnrc_val = 0;
9803 	}
9804 
9805 	wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
9806 	/* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
9807 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
9808 	 */
9809 	wr32(E1000_RTTBCNRM, 0x14);
9810 	wr32(E1000_RTTBCNRC, bcnrc_val);
9811 }
9812 
9813 static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
9814 {
9815 	int actual_link_speed, i;
9816 	bool reset_rate = false;
9817 
9818 	/* VF TX rate limit was not set or not supported */
9819 	if ((adapter->vf_rate_link_speed == 0) ||
9820 	    (adapter->hw.mac.type != e1000_82576))
9821 		return;
9822 
9823 	actual_link_speed = igb_link_mbps(adapter->link_speed);
9824 	if (actual_link_speed != adapter->vf_rate_link_speed) {
9825 		reset_rate = true;
9826 		adapter->vf_rate_link_speed = 0;
9827 		dev_info(&adapter->pdev->dev,
9828 			 "Link speed has been changed. VF Transmit rate is disabled\n");
9829 	}
9830 
9831 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
9832 		if (reset_rate)
9833 			adapter->vf_data[i].tx_rate = 0;
9834 
9835 		igb_set_vf_rate_limit(&adapter->hw, i,
9836 				      adapter->vf_data[i].tx_rate,
9837 				      actual_link_speed);
9838 	}
9839 }
9840 
9841 static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
9842 			     int min_tx_rate, int max_tx_rate)
9843 {
9844 	struct igb_adapter *adapter = netdev_priv(netdev);
9845 	struct e1000_hw *hw = &adapter->hw;
9846 	int actual_link_speed;
9847 
9848 	if (hw->mac.type != e1000_82576)
9849 		return -EOPNOTSUPP;
9850 
9851 	if (min_tx_rate)
9852 		return -EINVAL;
9853 
9854 	actual_link_speed = igb_link_mbps(adapter->link_speed);
9855 	if ((vf >= adapter->vfs_allocated_count) ||
9856 	    (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
9857 	    (max_tx_rate < 0) ||
9858 	    (max_tx_rate > actual_link_speed))
9859 		return -EINVAL;
9860 
9861 	adapter->vf_rate_link_speed = actual_link_speed;
9862 	adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
9863 	igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
9864 
9865 	return 0;
9866 }
9867 
9868 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
9869 				   bool setting)
9870 {
9871 	struct igb_adapter *adapter = netdev_priv(netdev);
9872 	struct e1000_hw *hw = &adapter->hw;
9873 	u32 reg_val, reg_offset;
9874 
9875 	if (!adapter->vfs_allocated_count)
9876 		return -EOPNOTSUPP;
9877 
9878 	if (vf >= adapter->vfs_allocated_count)
9879 		return -EINVAL;
9880 
9881 	reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
9882 	reg_val = rd32(reg_offset);
9883 	if (setting)
9884 		reg_val |= (BIT(vf) |
9885 			    BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9886 	else
9887 		reg_val &= ~(BIT(vf) |
9888 			     BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9889 	wr32(reg_offset, reg_val);
9890 
9891 	adapter->vf_data[vf].spoofchk_enabled = setting;
9892 	return 0;
9893 }
9894 
9895 static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf, bool setting)
9896 {
9897 	struct igb_adapter *adapter = netdev_priv(netdev);
9898 
9899 	if (vf >= adapter->vfs_allocated_count)
9900 		return -EINVAL;
9901 	if (adapter->vf_data[vf].trusted == setting)
9902 		return 0;
9903 
9904 	adapter->vf_data[vf].trusted = setting;
9905 
9906 	dev_info(&adapter->pdev->dev, "VF %u is %strusted\n",
9907 		 vf, setting ? "" : "not ");
9908 	return 0;
9909 }
9910 
9911 static int igb_ndo_get_vf_config(struct net_device *netdev,
9912 				 int vf, struct ifla_vf_info *ivi)
9913 {
9914 	struct igb_adapter *adapter = netdev_priv(netdev);
9915 	if (vf >= adapter->vfs_allocated_count)
9916 		return -EINVAL;
9917 	ivi->vf = vf;
9918 	memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
9919 	ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
9920 	ivi->min_tx_rate = 0;
9921 	ivi->vlan = adapter->vf_data[vf].pf_vlan;
9922 	ivi->qos = adapter->vf_data[vf].pf_qos;
9923 	ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
9924 	ivi->trusted = adapter->vf_data[vf].trusted;
9925 	return 0;
9926 }
9927 
9928 static void igb_vmm_control(struct igb_adapter *adapter)
9929 {
9930 	struct e1000_hw *hw = &adapter->hw;
9931 	u32 reg;
9932 
9933 	switch (hw->mac.type) {
9934 	case e1000_82575:
9935 	case e1000_i210:
9936 	case e1000_i211:
9937 	case e1000_i354:
9938 	default:
9939 		/* replication is not supported for 82575 */
9940 		return;
9941 	case e1000_82576:
9942 		/* notify HW that the MAC is adding vlan tags */
9943 		reg = rd32(E1000_DTXCTL);
9944 		reg |= E1000_DTXCTL_VLAN_ADDED;
9945 		wr32(E1000_DTXCTL, reg);
9946 		fallthrough;
9947 	case e1000_82580:
9948 		/* enable replication vlan tag stripping */
9949 		reg = rd32(E1000_RPLOLR);
9950 		reg |= E1000_RPLOLR_STRVLAN;
9951 		wr32(E1000_RPLOLR, reg);
9952 		fallthrough;
9953 	case e1000_i350:
9954 		/* none of the above registers are supported by i350 */
9955 		break;
9956 	}
9957 
9958 	if (adapter->vfs_allocated_count) {
9959 		igb_vmdq_set_loopback_pf(hw, true);
9960 		igb_vmdq_set_replication_pf(hw, true);
9961 		igb_vmdq_set_anti_spoofing_pf(hw, true,
9962 					      adapter->vfs_allocated_count);
9963 	} else {
9964 		igb_vmdq_set_loopback_pf(hw, false);
9965 		igb_vmdq_set_replication_pf(hw, false);
9966 	}
9967 }
9968 
9969 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
9970 {
9971 	struct e1000_hw *hw = &adapter->hw;
9972 	u32 dmac_thr;
9973 	u16 hwm;
9974 	u32 reg;
9975 
9976 	if (hw->mac.type > e1000_82580) {
9977 		if (adapter->flags & IGB_FLAG_DMAC) {
9978 			/* force threshold to 0. */
9979 			wr32(E1000_DMCTXTH, 0);
9980 
9981 			/* DMA Coalescing high water mark needs to be greater
9982 			 * than the Rx threshold. Set hwm to PBA - max frame
9983 			 * size in 16B units, capping it at PBA - 6KB.
9984 			 */
9985 			hwm = 64 * (pba - 6);
9986 			reg = rd32(E1000_FCRTC);
9987 			reg &= ~E1000_FCRTC_RTH_COAL_MASK;
9988 			reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
9989 				& E1000_FCRTC_RTH_COAL_MASK);
9990 			wr32(E1000_FCRTC, reg);
9991 
9992 			/* Set the DMA Coalescing Rx threshold to PBA - 2 * max
9993 			 * frame size, capping it at PBA - 10KB.
9994 			 */
9995 			dmac_thr = pba - 10;
9996 			reg = rd32(E1000_DMACR);
9997 			reg &= ~E1000_DMACR_DMACTHR_MASK;
9998 			reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
9999 				& E1000_DMACR_DMACTHR_MASK);
10000 
10001 			/* transition to L0x or L1 if available..*/
10002 			reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
10003 
10004 			/* watchdog timer= +-1000 usec in 32usec intervals */
10005 			reg |= (1000 >> 5);
10006 
10007 			/* Disable BMC-to-OS Watchdog Enable */
10008 			if (hw->mac.type != e1000_i354)
10009 				reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
10010 			wr32(E1000_DMACR, reg);
10011 
10012 			/* no lower threshold to disable
10013 			 * coalescing(smart fifb)-UTRESH=0
10014 			 */
10015 			wr32(E1000_DMCRTRH, 0);
10016 
10017 			reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
10018 
10019 			wr32(E1000_DMCTLX, reg);
10020 
10021 			/* free space in tx packet buffer to wake from
10022 			 * DMA coal
10023 			 */
10024 			wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
10025 			     (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
10026 		}
10027 
10028 		if (hw->mac.type >= e1000_i210 ||
10029 		    (adapter->flags & IGB_FLAG_DMAC)) {
10030 			reg = rd32(E1000_PCIEMISC);
10031 			reg |= E1000_PCIEMISC_LX_DECISION;
10032 			wr32(E1000_PCIEMISC, reg);
10033 		} /* endif adapter->dmac is not disabled */
10034 	} else if (hw->mac.type == e1000_82580) {
10035 		u32 reg = rd32(E1000_PCIEMISC);
10036 
10037 		wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
10038 		wr32(E1000_DMACR, 0);
10039 	}
10040 }
10041 
10042 /**
10043  *  igb_read_i2c_byte - Reads 8 bit word over I2C
10044  *  @hw: pointer to hardware structure
10045  *  @byte_offset: byte offset to read
10046  *  @dev_addr: device address
10047  *  @data: value read
10048  *
10049  *  Performs byte read operation over I2C interface at
10050  *  a specified device address.
10051  **/
10052 s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
10053 		      u8 dev_addr, u8 *data)
10054 {
10055 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
10056 	struct i2c_client *this_client = adapter->i2c_client;
10057 	s32 status;
10058 	u16 swfw_mask = 0;
10059 
10060 	if (!this_client)
10061 		return E1000_ERR_I2C;
10062 
10063 	swfw_mask = E1000_SWFW_PHY0_SM;
10064 
10065 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
10066 		return E1000_ERR_SWFW_SYNC;
10067 
10068 	status = i2c_smbus_read_byte_data(this_client, byte_offset);
10069 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
10070 
10071 	if (status < 0)
10072 		return E1000_ERR_I2C;
10073 	else {
10074 		*data = status;
10075 		return 0;
10076 	}
10077 }
10078 
10079 /**
10080  *  igb_write_i2c_byte - Writes 8 bit word over I2C
10081  *  @hw: pointer to hardware structure
10082  *  @byte_offset: byte offset to write
10083  *  @dev_addr: device address
10084  *  @data: value to write
10085  *
10086  *  Performs byte write operation over I2C interface at
10087  *  a specified device address.
10088  **/
10089 s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
10090 		       u8 dev_addr, u8 data)
10091 {
10092 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
10093 	struct i2c_client *this_client = adapter->i2c_client;
10094 	s32 status;
10095 	u16 swfw_mask = E1000_SWFW_PHY0_SM;
10096 
10097 	if (!this_client)
10098 		return E1000_ERR_I2C;
10099 
10100 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
10101 		return E1000_ERR_SWFW_SYNC;
10102 	status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
10103 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
10104 
10105 	if (status)
10106 		return E1000_ERR_I2C;
10107 	else
10108 		return 0;
10109 
10110 }
10111 
10112 int igb_reinit_queues(struct igb_adapter *adapter)
10113 {
10114 	struct net_device *netdev = adapter->netdev;
10115 	struct pci_dev *pdev = adapter->pdev;
10116 	int err = 0;
10117 
10118 	if (netif_running(netdev))
10119 		igb_close(netdev);
10120 
10121 	igb_reset_interrupt_capability(adapter);
10122 
10123 	if (igb_init_interrupt_scheme(adapter, true)) {
10124 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
10125 		return -ENOMEM;
10126 	}
10127 
10128 	if (netif_running(netdev))
10129 		err = igb_open(netdev);
10130 
10131 	return err;
10132 }
10133 
10134 static void igb_nfc_filter_exit(struct igb_adapter *adapter)
10135 {
10136 	struct igb_nfc_filter *rule;
10137 
10138 	spin_lock(&adapter->nfc_lock);
10139 
10140 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
10141 		igb_erase_filter(adapter, rule);
10142 
10143 	hlist_for_each_entry(rule, &adapter->cls_flower_list, nfc_node)
10144 		igb_erase_filter(adapter, rule);
10145 
10146 	spin_unlock(&adapter->nfc_lock);
10147 }
10148 
10149 static void igb_nfc_filter_restore(struct igb_adapter *adapter)
10150 {
10151 	struct igb_nfc_filter *rule;
10152 
10153 	spin_lock(&adapter->nfc_lock);
10154 
10155 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
10156 		igb_add_filter(adapter, rule);
10157 
10158 	spin_unlock(&adapter->nfc_lock);
10159 }
10160 /* igb_main.c */
10161