1 /* Intel(R) Gigabit Ethernet Linux driver
2  * Copyright(c) 2007-2014 Intel Corporation.
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms and conditions of the GNU General Public License,
6  * version 2, as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope it will be useful, but WITHOUT
9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
11  * more details.
12  *
13  * You should have received a copy of the GNU General Public License along with
14  * this program; if not, see <http://www.gnu.org/licenses/>.
15  *
16  * The full GNU General Public License is included in this distribution in
17  * the file called "COPYING".
18  *
19  * Contact Information:
20  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
21  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
22  */
23 
24 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
25 
26 #include <linux/module.h>
27 #include <linux/types.h>
28 #include <linux/init.h>
29 #include <linux/bitops.h>
30 #include <linux/vmalloc.h>
31 #include <linux/pagemap.h>
32 #include <linux/netdevice.h>
33 #include <linux/ipv6.h>
34 #include <linux/slab.h>
35 #include <net/checksum.h>
36 #include <net/ip6_checksum.h>
37 #include <net/pkt_sched.h>
38 #include <linux/net_tstamp.h>
39 #include <linux/mii.h>
40 #include <linux/ethtool.h>
41 #include <linux/if.h>
42 #include <linux/if_vlan.h>
43 #include <linux/pci.h>
44 #include <linux/pci-aspm.h>
45 #include <linux/delay.h>
46 #include <linux/interrupt.h>
47 #include <linux/ip.h>
48 #include <linux/tcp.h>
49 #include <linux/sctp.h>
50 #include <linux/if_ether.h>
51 #include <linux/aer.h>
52 #include <linux/prefetch.h>
53 #include <linux/pm_runtime.h>
54 #include <linux/etherdevice.h>
55 #ifdef CONFIG_IGB_DCA
56 #include <linux/dca.h>
57 #endif
58 #include <linux/i2c.h>
59 #include "igb.h"
60 
61 #define MAJ 5
62 #define MIN 4
63 #define BUILD 0
64 #define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
65 __stringify(BUILD) "-k"
66 
67 enum queue_mode {
68 	QUEUE_MODE_STRICT_PRIORITY,
69 	QUEUE_MODE_STREAM_RESERVATION,
70 };
71 
72 enum tx_queue_prio {
73 	TX_QUEUE_PRIO_HIGH,
74 	TX_QUEUE_PRIO_LOW,
75 };
76 
77 char igb_driver_name[] = "igb";
78 char igb_driver_version[] = DRV_VERSION;
79 static const char igb_driver_string[] =
80 				"Intel(R) Gigabit Ethernet Network Driver";
81 static const char igb_copyright[] =
82 				"Copyright (c) 2007-2014 Intel Corporation.";
83 
84 static const struct e1000_info *igb_info_tbl[] = {
85 	[board_82575] = &e1000_82575_info,
86 };
87 
88 static const struct pci_device_id igb_pci_tbl[] = {
89 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
90 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_SGMII) },
91 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
92 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I211_COPPER), board_82575 },
93 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER), board_82575 },
94 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_FIBER), board_82575 },
95 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES), board_82575 },
96 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SGMII), board_82575 },
97 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER_FLASHLESS), board_82575 },
98 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES_FLASHLESS), board_82575 },
99 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_COPPER), board_82575 },
100 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_FIBER), board_82575 },
101 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SERDES), board_82575 },
102 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SGMII), board_82575 },
103 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER), board_82575 },
104 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_FIBER), board_82575 },
105 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_QUAD_FIBER), board_82575 },
106 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SERDES), board_82575 },
107 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SGMII), board_82575 },
108 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER_DUAL), board_82575 },
109 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SGMII), board_82575 },
110 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SERDES), board_82575 },
111 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_BACKPLANE), board_82575 },
112 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SFP), board_82575 },
113 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
114 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
115 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
116 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
117 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
118 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
119 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER_ET2), board_82575 },
120 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
121 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
122 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
123 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
124 	/* required last entry */
125 	{0, }
126 };
127 
128 MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
129 
130 static int igb_setup_all_tx_resources(struct igb_adapter *);
131 static int igb_setup_all_rx_resources(struct igb_adapter *);
132 static void igb_free_all_tx_resources(struct igb_adapter *);
133 static void igb_free_all_rx_resources(struct igb_adapter *);
134 static void igb_setup_mrqc(struct igb_adapter *);
135 static int igb_probe(struct pci_dev *, const struct pci_device_id *);
136 static void igb_remove(struct pci_dev *pdev);
137 static int igb_sw_init(struct igb_adapter *);
138 int igb_open(struct net_device *);
139 int igb_close(struct net_device *);
140 static void igb_configure(struct igb_adapter *);
141 static void igb_configure_tx(struct igb_adapter *);
142 static void igb_configure_rx(struct igb_adapter *);
143 static void igb_clean_all_tx_rings(struct igb_adapter *);
144 static void igb_clean_all_rx_rings(struct igb_adapter *);
145 static void igb_clean_tx_ring(struct igb_ring *);
146 static void igb_clean_rx_ring(struct igb_ring *);
147 static void igb_set_rx_mode(struct net_device *);
148 static void igb_update_phy_info(struct timer_list *);
149 static void igb_watchdog(struct timer_list *);
150 static void igb_watchdog_task(struct work_struct *);
151 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb, struct net_device *);
152 static void igb_get_stats64(struct net_device *dev,
153 			    struct rtnl_link_stats64 *stats);
154 static int igb_change_mtu(struct net_device *, int);
155 static int igb_set_mac(struct net_device *, void *);
156 static void igb_set_uta(struct igb_adapter *adapter, bool set);
157 static irqreturn_t igb_intr(int irq, void *);
158 static irqreturn_t igb_intr_msi(int irq, void *);
159 static irqreturn_t igb_msix_other(int irq, void *);
160 static irqreturn_t igb_msix_ring(int irq, void *);
161 #ifdef CONFIG_IGB_DCA
162 static void igb_update_dca(struct igb_q_vector *);
163 static void igb_setup_dca(struct igb_adapter *);
164 #endif /* CONFIG_IGB_DCA */
165 static int igb_poll(struct napi_struct *, int);
166 static bool igb_clean_tx_irq(struct igb_q_vector *, int);
167 static int igb_clean_rx_irq(struct igb_q_vector *, int);
168 static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
169 static void igb_tx_timeout(struct net_device *);
170 static void igb_reset_task(struct work_struct *);
171 static void igb_vlan_mode(struct net_device *netdev,
172 			  netdev_features_t features);
173 static int igb_vlan_rx_add_vid(struct net_device *, __be16, u16);
174 static int igb_vlan_rx_kill_vid(struct net_device *, __be16, u16);
175 static void igb_restore_vlan(struct igb_adapter *);
176 static void igb_rar_set_index(struct igb_adapter *, u32);
177 static void igb_ping_all_vfs(struct igb_adapter *);
178 static void igb_msg_task(struct igb_adapter *);
179 static void igb_vmm_control(struct igb_adapter *);
180 static int igb_set_vf_mac(struct igb_adapter *, int, unsigned char *);
181 static void igb_flush_mac_table(struct igb_adapter *);
182 static int igb_available_rars(struct igb_adapter *, u8);
183 static void igb_set_default_mac_filter(struct igb_adapter *);
184 static int igb_uc_sync(struct net_device *, const unsigned char *);
185 static int igb_uc_unsync(struct net_device *, const unsigned char *);
186 static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
187 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac);
188 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
189 			       int vf, u16 vlan, u8 qos, __be16 vlan_proto);
190 static int igb_ndo_set_vf_bw(struct net_device *, int, int, int);
191 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
192 				   bool setting);
193 static int igb_ndo_get_vf_config(struct net_device *netdev, int vf,
194 				 struct ifla_vf_info *ivi);
195 static void igb_check_vf_rate_limit(struct igb_adapter *);
196 static void igb_nfc_filter_exit(struct igb_adapter *adapter);
197 static void igb_nfc_filter_restore(struct igb_adapter *adapter);
198 
199 #ifdef CONFIG_PCI_IOV
200 static int igb_vf_configure(struct igb_adapter *adapter, int vf);
201 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs);
202 static int igb_disable_sriov(struct pci_dev *dev);
203 static int igb_pci_disable_sriov(struct pci_dev *dev);
204 #endif
205 
206 static int igb_suspend(struct device *);
207 static int igb_resume(struct device *);
208 static int igb_runtime_suspend(struct device *dev);
209 static int igb_runtime_resume(struct device *dev);
210 static int igb_runtime_idle(struct device *dev);
211 static const struct dev_pm_ops igb_pm_ops = {
212 	SET_SYSTEM_SLEEP_PM_OPS(igb_suspend, igb_resume)
213 	SET_RUNTIME_PM_OPS(igb_runtime_suspend, igb_runtime_resume,
214 			igb_runtime_idle)
215 };
216 static void igb_shutdown(struct pci_dev *);
217 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs);
218 #ifdef CONFIG_IGB_DCA
219 static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
220 static struct notifier_block dca_notifier = {
221 	.notifier_call	= igb_notify_dca,
222 	.next		= NULL,
223 	.priority	= 0
224 };
225 #endif
226 #ifdef CONFIG_NET_POLL_CONTROLLER
227 /* for netdump / net console */
228 static void igb_netpoll(struct net_device *);
229 #endif
230 #ifdef CONFIG_PCI_IOV
231 static unsigned int max_vfs;
232 module_param(max_vfs, uint, 0);
233 MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate per physical function");
234 #endif /* CONFIG_PCI_IOV */
235 
236 static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
237 		     pci_channel_state_t);
238 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
239 static void igb_io_resume(struct pci_dev *);
240 
241 static const struct pci_error_handlers igb_err_handler = {
242 	.error_detected = igb_io_error_detected,
243 	.slot_reset = igb_io_slot_reset,
244 	.resume = igb_io_resume,
245 };
246 
247 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba);
248 
249 static struct pci_driver igb_driver = {
250 	.name     = igb_driver_name,
251 	.id_table = igb_pci_tbl,
252 	.probe    = igb_probe,
253 	.remove   = igb_remove,
254 #ifdef CONFIG_PM
255 	.driver.pm = &igb_pm_ops,
256 #endif
257 	.shutdown = igb_shutdown,
258 	.sriov_configure = igb_pci_sriov_configure,
259 	.err_handler = &igb_err_handler
260 };
261 
262 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
263 MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
264 MODULE_LICENSE("GPL");
265 MODULE_VERSION(DRV_VERSION);
266 
267 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
268 static int debug = -1;
269 module_param(debug, int, 0);
270 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
271 
272 struct igb_reg_info {
273 	u32 ofs;
274 	char *name;
275 };
276 
277 static const struct igb_reg_info igb_reg_info_tbl[] = {
278 
279 	/* General Registers */
280 	{E1000_CTRL, "CTRL"},
281 	{E1000_STATUS, "STATUS"},
282 	{E1000_CTRL_EXT, "CTRL_EXT"},
283 
284 	/* Interrupt Registers */
285 	{E1000_ICR, "ICR"},
286 
287 	/* RX Registers */
288 	{E1000_RCTL, "RCTL"},
289 	{E1000_RDLEN(0), "RDLEN"},
290 	{E1000_RDH(0), "RDH"},
291 	{E1000_RDT(0), "RDT"},
292 	{E1000_RXDCTL(0), "RXDCTL"},
293 	{E1000_RDBAL(0), "RDBAL"},
294 	{E1000_RDBAH(0), "RDBAH"},
295 
296 	/* TX Registers */
297 	{E1000_TCTL, "TCTL"},
298 	{E1000_TDBAL(0), "TDBAL"},
299 	{E1000_TDBAH(0), "TDBAH"},
300 	{E1000_TDLEN(0), "TDLEN"},
301 	{E1000_TDH(0), "TDH"},
302 	{E1000_TDT(0), "TDT"},
303 	{E1000_TXDCTL(0), "TXDCTL"},
304 	{E1000_TDFH, "TDFH"},
305 	{E1000_TDFT, "TDFT"},
306 	{E1000_TDFHS, "TDFHS"},
307 	{E1000_TDFPC, "TDFPC"},
308 
309 	/* List Terminator */
310 	{}
311 };
312 
313 /* igb_regdump - register printout routine */
314 static void igb_regdump(struct e1000_hw *hw, struct igb_reg_info *reginfo)
315 {
316 	int n = 0;
317 	char rname[16];
318 	u32 regs[8];
319 
320 	switch (reginfo->ofs) {
321 	case E1000_RDLEN(0):
322 		for (n = 0; n < 4; n++)
323 			regs[n] = rd32(E1000_RDLEN(n));
324 		break;
325 	case E1000_RDH(0):
326 		for (n = 0; n < 4; n++)
327 			regs[n] = rd32(E1000_RDH(n));
328 		break;
329 	case E1000_RDT(0):
330 		for (n = 0; n < 4; n++)
331 			regs[n] = rd32(E1000_RDT(n));
332 		break;
333 	case E1000_RXDCTL(0):
334 		for (n = 0; n < 4; n++)
335 			regs[n] = rd32(E1000_RXDCTL(n));
336 		break;
337 	case E1000_RDBAL(0):
338 		for (n = 0; n < 4; n++)
339 			regs[n] = rd32(E1000_RDBAL(n));
340 		break;
341 	case E1000_RDBAH(0):
342 		for (n = 0; n < 4; n++)
343 			regs[n] = rd32(E1000_RDBAH(n));
344 		break;
345 	case E1000_TDBAL(0):
346 		for (n = 0; n < 4; n++)
347 			regs[n] = rd32(E1000_RDBAL(n));
348 		break;
349 	case E1000_TDBAH(0):
350 		for (n = 0; n < 4; n++)
351 			regs[n] = rd32(E1000_TDBAH(n));
352 		break;
353 	case E1000_TDLEN(0):
354 		for (n = 0; n < 4; n++)
355 			regs[n] = rd32(E1000_TDLEN(n));
356 		break;
357 	case E1000_TDH(0):
358 		for (n = 0; n < 4; n++)
359 			regs[n] = rd32(E1000_TDH(n));
360 		break;
361 	case E1000_TDT(0):
362 		for (n = 0; n < 4; n++)
363 			regs[n] = rd32(E1000_TDT(n));
364 		break;
365 	case E1000_TXDCTL(0):
366 		for (n = 0; n < 4; n++)
367 			regs[n] = rd32(E1000_TXDCTL(n));
368 		break;
369 	default:
370 		pr_info("%-15s %08x\n", reginfo->name, rd32(reginfo->ofs));
371 		return;
372 	}
373 
374 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-3]");
375 	pr_info("%-15s %08x %08x %08x %08x\n", rname, regs[0], regs[1],
376 		regs[2], regs[3]);
377 }
378 
379 /* igb_dump - Print registers, Tx-rings and Rx-rings */
380 static void igb_dump(struct igb_adapter *adapter)
381 {
382 	struct net_device *netdev = adapter->netdev;
383 	struct e1000_hw *hw = &adapter->hw;
384 	struct igb_reg_info *reginfo;
385 	struct igb_ring *tx_ring;
386 	union e1000_adv_tx_desc *tx_desc;
387 	struct my_u0 { u64 a; u64 b; } *u0;
388 	struct igb_ring *rx_ring;
389 	union e1000_adv_rx_desc *rx_desc;
390 	u32 staterr;
391 	u16 i, n;
392 
393 	if (!netif_msg_hw(adapter))
394 		return;
395 
396 	/* Print netdevice Info */
397 	if (netdev) {
398 		dev_info(&adapter->pdev->dev, "Net device Info\n");
399 		pr_info("Device Name     state            trans_start\n");
400 		pr_info("%-15s %016lX %016lX\n", netdev->name,
401 			netdev->state, dev_trans_start(netdev));
402 	}
403 
404 	/* Print Registers */
405 	dev_info(&adapter->pdev->dev, "Register Dump\n");
406 	pr_info(" Register Name   Value\n");
407 	for (reginfo = (struct igb_reg_info *)igb_reg_info_tbl;
408 	     reginfo->name; reginfo++) {
409 		igb_regdump(hw, reginfo);
410 	}
411 
412 	/* Print TX Ring Summary */
413 	if (!netdev || !netif_running(netdev))
414 		goto exit;
415 
416 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
417 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
418 	for (n = 0; n < adapter->num_tx_queues; n++) {
419 		struct igb_tx_buffer *buffer_info;
420 		tx_ring = adapter->tx_ring[n];
421 		buffer_info = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
422 		pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
423 			n, tx_ring->next_to_use, tx_ring->next_to_clean,
424 			(u64)dma_unmap_addr(buffer_info, dma),
425 			dma_unmap_len(buffer_info, len),
426 			buffer_info->next_to_watch,
427 			(u64)buffer_info->time_stamp);
428 	}
429 
430 	/* Print TX Rings */
431 	if (!netif_msg_tx_done(adapter))
432 		goto rx_ring_summary;
433 
434 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
435 
436 	/* Transmit Descriptor Formats
437 	 *
438 	 * Advanced Transmit Descriptor
439 	 *   +--------------------------------------------------------------+
440 	 * 0 |         Buffer Address [63:0]                                |
441 	 *   +--------------------------------------------------------------+
442 	 * 8 | PAYLEN  | PORTS  |CC|IDX | STA | DCMD  |DTYP|MAC|RSV| DTALEN |
443 	 *   +--------------------------------------------------------------+
444 	 *   63      46 45    40 39 38 36 35 32 31   24             15       0
445 	 */
446 
447 	for (n = 0; n < adapter->num_tx_queues; n++) {
448 		tx_ring = adapter->tx_ring[n];
449 		pr_info("------------------------------------\n");
450 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
451 		pr_info("------------------------------------\n");
452 		pr_info("T [desc]     [address 63:0  ] [PlPOCIStDDM Ln] [bi->dma       ] leng  ntw timestamp        bi->skb\n");
453 
454 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
455 			const char *next_desc;
456 			struct igb_tx_buffer *buffer_info;
457 			tx_desc = IGB_TX_DESC(tx_ring, i);
458 			buffer_info = &tx_ring->tx_buffer_info[i];
459 			u0 = (struct my_u0 *)tx_desc;
460 			if (i == tx_ring->next_to_use &&
461 			    i == tx_ring->next_to_clean)
462 				next_desc = " NTC/U";
463 			else if (i == tx_ring->next_to_use)
464 				next_desc = " NTU";
465 			else if (i == tx_ring->next_to_clean)
466 				next_desc = " NTC";
467 			else
468 				next_desc = "";
469 
470 			pr_info("T [0x%03X]    %016llX %016llX %016llX %04X  %p %016llX %p%s\n",
471 				i, le64_to_cpu(u0->a),
472 				le64_to_cpu(u0->b),
473 				(u64)dma_unmap_addr(buffer_info, dma),
474 				dma_unmap_len(buffer_info, len),
475 				buffer_info->next_to_watch,
476 				(u64)buffer_info->time_stamp,
477 				buffer_info->skb, next_desc);
478 
479 			if (netif_msg_pktdata(adapter) && buffer_info->skb)
480 				print_hex_dump(KERN_INFO, "",
481 					DUMP_PREFIX_ADDRESS,
482 					16, 1, buffer_info->skb->data,
483 					dma_unmap_len(buffer_info, len),
484 					true);
485 		}
486 	}
487 
488 	/* Print RX Rings Summary */
489 rx_ring_summary:
490 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
491 	pr_info("Queue [NTU] [NTC]\n");
492 	for (n = 0; n < adapter->num_rx_queues; n++) {
493 		rx_ring = adapter->rx_ring[n];
494 		pr_info(" %5d %5X %5X\n",
495 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
496 	}
497 
498 	/* Print RX Rings */
499 	if (!netif_msg_rx_status(adapter))
500 		goto exit;
501 
502 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
503 
504 	/* Advanced Receive Descriptor (Read) Format
505 	 *    63                                           1        0
506 	 *    +-----------------------------------------------------+
507 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
508 	 *    +----------------------------------------------+------+
509 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
510 	 *    +-----------------------------------------------------+
511 	 *
512 	 *
513 	 * Advanced Receive Descriptor (Write-Back) Format
514 	 *
515 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
516 	 *   +------------------------------------------------------+
517 	 * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
518 	 *   | Checksum   Ident  |   |           |    | Type | Type |
519 	 *   +------------------------------------------------------+
520 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
521 	 *   +------------------------------------------------------+
522 	 *   63       48 47    32 31            20 19               0
523 	 */
524 
525 	for (n = 0; n < adapter->num_rx_queues; n++) {
526 		rx_ring = adapter->rx_ring[n];
527 		pr_info("------------------------------------\n");
528 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
529 		pr_info("------------------------------------\n");
530 		pr_info("R  [desc]      [ PktBuf     A0] [  HeadBuf   DD] [bi->dma       ] [bi->skb] <-- Adv Rx Read format\n");
531 		pr_info("RWB[desc]      [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
532 
533 		for (i = 0; i < rx_ring->count; i++) {
534 			const char *next_desc;
535 			struct igb_rx_buffer *buffer_info;
536 			buffer_info = &rx_ring->rx_buffer_info[i];
537 			rx_desc = IGB_RX_DESC(rx_ring, i);
538 			u0 = (struct my_u0 *)rx_desc;
539 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
540 
541 			if (i == rx_ring->next_to_use)
542 				next_desc = " NTU";
543 			else if (i == rx_ring->next_to_clean)
544 				next_desc = " NTC";
545 			else
546 				next_desc = "";
547 
548 			if (staterr & E1000_RXD_STAT_DD) {
549 				/* Descriptor Done */
550 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %s\n",
551 					"RWB", i,
552 					le64_to_cpu(u0->a),
553 					le64_to_cpu(u0->b),
554 					next_desc);
555 			} else {
556 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %s\n",
557 					"R  ", i,
558 					le64_to_cpu(u0->a),
559 					le64_to_cpu(u0->b),
560 					(u64)buffer_info->dma,
561 					next_desc);
562 
563 				if (netif_msg_pktdata(adapter) &&
564 				    buffer_info->dma && buffer_info->page) {
565 					print_hex_dump(KERN_INFO, "",
566 					  DUMP_PREFIX_ADDRESS,
567 					  16, 1,
568 					  page_address(buffer_info->page) +
569 						      buffer_info->page_offset,
570 					  igb_rx_bufsz(rx_ring), true);
571 				}
572 			}
573 		}
574 	}
575 
576 exit:
577 	return;
578 }
579 
580 /**
581  *  igb_get_i2c_data - Reads the I2C SDA data bit
582  *  @hw: pointer to hardware structure
583  *  @i2cctl: Current value of I2CCTL register
584  *
585  *  Returns the I2C data bit value
586  **/
587 static int igb_get_i2c_data(void *data)
588 {
589 	struct igb_adapter *adapter = (struct igb_adapter *)data;
590 	struct e1000_hw *hw = &adapter->hw;
591 	s32 i2cctl = rd32(E1000_I2CPARAMS);
592 
593 	return !!(i2cctl & E1000_I2C_DATA_IN);
594 }
595 
596 /**
597  *  igb_set_i2c_data - Sets the I2C data bit
598  *  @data: pointer to hardware structure
599  *  @state: I2C data value (0 or 1) to set
600  *
601  *  Sets the I2C data bit
602  **/
603 static void igb_set_i2c_data(void *data, int state)
604 {
605 	struct igb_adapter *adapter = (struct igb_adapter *)data;
606 	struct e1000_hw *hw = &adapter->hw;
607 	s32 i2cctl = rd32(E1000_I2CPARAMS);
608 
609 	if (state)
610 		i2cctl |= E1000_I2C_DATA_OUT;
611 	else
612 		i2cctl &= ~E1000_I2C_DATA_OUT;
613 
614 	i2cctl &= ~E1000_I2C_DATA_OE_N;
615 	i2cctl |= E1000_I2C_CLK_OE_N;
616 	wr32(E1000_I2CPARAMS, i2cctl);
617 	wrfl();
618 
619 }
620 
621 /**
622  *  igb_set_i2c_clk - Sets the I2C SCL clock
623  *  @data: pointer to hardware structure
624  *  @state: state to set clock
625  *
626  *  Sets the I2C clock line to state
627  **/
628 static void igb_set_i2c_clk(void *data, int state)
629 {
630 	struct igb_adapter *adapter = (struct igb_adapter *)data;
631 	struct e1000_hw *hw = &adapter->hw;
632 	s32 i2cctl = rd32(E1000_I2CPARAMS);
633 
634 	if (state) {
635 		i2cctl |= E1000_I2C_CLK_OUT;
636 		i2cctl &= ~E1000_I2C_CLK_OE_N;
637 	} else {
638 		i2cctl &= ~E1000_I2C_CLK_OUT;
639 		i2cctl &= ~E1000_I2C_CLK_OE_N;
640 	}
641 	wr32(E1000_I2CPARAMS, i2cctl);
642 	wrfl();
643 }
644 
645 /**
646  *  igb_get_i2c_clk - Gets the I2C SCL clock state
647  *  @data: pointer to hardware structure
648  *
649  *  Gets the I2C clock state
650  **/
651 static int igb_get_i2c_clk(void *data)
652 {
653 	struct igb_adapter *adapter = (struct igb_adapter *)data;
654 	struct e1000_hw *hw = &adapter->hw;
655 	s32 i2cctl = rd32(E1000_I2CPARAMS);
656 
657 	return !!(i2cctl & E1000_I2C_CLK_IN);
658 }
659 
660 static const struct i2c_algo_bit_data igb_i2c_algo = {
661 	.setsda		= igb_set_i2c_data,
662 	.setscl		= igb_set_i2c_clk,
663 	.getsda		= igb_get_i2c_data,
664 	.getscl		= igb_get_i2c_clk,
665 	.udelay		= 5,
666 	.timeout	= 20,
667 };
668 
669 /**
670  *  igb_get_hw_dev - return device
671  *  @hw: pointer to hardware structure
672  *
673  *  used by hardware layer to print debugging information
674  **/
675 struct net_device *igb_get_hw_dev(struct e1000_hw *hw)
676 {
677 	struct igb_adapter *adapter = hw->back;
678 	return adapter->netdev;
679 }
680 
681 /**
682  *  igb_init_module - Driver Registration Routine
683  *
684  *  igb_init_module is the first routine called when the driver is
685  *  loaded. All it does is register with the PCI subsystem.
686  **/
687 static int __init igb_init_module(void)
688 {
689 	int ret;
690 
691 	pr_info("%s - version %s\n",
692 	       igb_driver_string, igb_driver_version);
693 	pr_info("%s\n", igb_copyright);
694 
695 #ifdef CONFIG_IGB_DCA
696 	dca_register_notify(&dca_notifier);
697 #endif
698 	ret = pci_register_driver(&igb_driver);
699 	return ret;
700 }
701 
702 module_init(igb_init_module);
703 
704 /**
705  *  igb_exit_module - Driver Exit Cleanup Routine
706  *
707  *  igb_exit_module is called just before the driver is removed
708  *  from memory.
709  **/
710 static void __exit igb_exit_module(void)
711 {
712 #ifdef CONFIG_IGB_DCA
713 	dca_unregister_notify(&dca_notifier);
714 #endif
715 	pci_unregister_driver(&igb_driver);
716 }
717 
718 module_exit(igb_exit_module);
719 
720 #define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
721 /**
722  *  igb_cache_ring_register - Descriptor ring to register mapping
723  *  @adapter: board private structure to initialize
724  *
725  *  Once we know the feature-set enabled for the device, we'll cache
726  *  the register offset the descriptor ring is assigned to.
727  **/
728 static void igb_cache_ring_register(struct igb_adapter *adapter)
729 {
730 	int i = 0, j = 0;
731 	u32 rbase_offset = adapter->vfs_allocated_count;
732 
733 	switch (adapter->hw.mac.type) {
734 	case e1000_82576:
735 		/* The queues are allocated for virtualization such that VF 0
736 		 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
737 		 * In order to avoid collision we start at the first free queue
738 		 * and continue consuming queues in the same sequence
739 		 */
740 		if (adapter->vfs_allocated_count) {
741 			for (; i < adapter->rss_queues; i++)
742 				adapter->rx_ring[i]->reg_idx = rbase_offset +
743 							       Q_IDX_82576(i);
744 		}
745 		/* Fall through */
746 	case e1000_82575:
747 	case e1000_82580:
748 	case e1000_i350:
749 	case e1000_i354:
750 	case e1000_i210:
751 	case e1000_i211:
752 		/* Fall through */
753 	default:
754 		for (; i < adapter->num_rx_queues; i++)
755 			adapter->rx_ring[i]->reg_idx = rbase_offset + i;
756 		for (; j < adapter->num_tx_queues; j++)
757 			adapter->tx_ring[j]->reg_idx = rbase_offset + j;
758 		break;
759 	}
760 }
761 
762 u32 igb_rd32(struct e1000_hw *hw, u32 reg)
763 {
764 	struct igb_adapter *igb = container_of(hw, struct igb_adapter, hw);
765 	u8 __iomem *hw_addr = READ_ONCE(hw->hw_addr);
766 	u32 value = 0;
767 
768 	if (E1000_REMOVED(hw_addr))
769 		return ~value;
770 
771 	value = readl(&hw_addr[reg]);
772 
773 	/* reads should not return all F's */
774 	if (!(~value) && (!reg || !(~readl(hw_addr)))) {
775 		struct net_device *netdev = igb->netdev;
776 		hw->hw_addr = NULL;
777 		netif_device_detach(netdev);
778 		netdev_err(netdev, "PCIe link lost, device now detached\n");
779 	}
780 
781 	return value;
782 }
783 
784 /**
785  *  igb_write_ivar - configure ivar for given MSI-X vector
786  *  @hw: pointer to the HW structure
787  *  @msix_vector: vector number we are allocating to a given ring
788  *  @index: row index of IVAR register to write within IVAR table
789  *  @offset: column offset of in IVAR, should be multiple of 8
790  *
791  *  This function is intended to handle the writing of the IVAR register
792  *  for adapters 82576 and newer.  The IVAR table consists of 2 columns,
793  *  each containing an cause allocation for an Rx and Tx ring, and a
794  *  variable number of rows depending on the number of queues supported.
795  **/
796 static void igb_write_ivar(struct e1000_hw *hw, int msix_vector,
797 			   int index, int offset)
798 {
799 	u32 ivar = array_rd32(E1000_IVAR0, index);
800 
801 	/* clear any bits that are currently set */
802 	ivar &= ~((u32)0xFF << offset);
803 
804 	/* write vector and valid bit */
805 	ivar |= (msix_vector | E1000_IVAR_VALID) << offset;
806 
807 	array_wr32(E1000_IVAR0, index, ivar);
808 }
809 
810 #define IGB_N0_QUEUE -1
811 static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
812 {
813 	struct igb_adapter *adapter = q_vector->adapter;
814 	struct e1000_hw *hw = &adapter->hw;
815 	int rx_queue = IGB_N0_QUEUE;
816 	int tx_queue = IGB_N0_QUEUE;
817 	u32 msixbm = 0;
818 
819 	if (q_vector->rx.ring)
820 		rx_queue = q_vector->rx.ring->reg_idx;
821 	if (q_vector->tx.ring)
822 		tx_queue = q_vector->tx.ring->reg_idx;
823 
824 	switch (hw->mac.type) {
825 	case e1000_82575:
826 		/* The 82575 assigns vectors using a bitmask, which matches the
827 		 * bitmask for the EICR/EIMS/EIMC registers.  To assign one
828 		 * or more queues to a vector, we write the appropriate bits
829 		 * into the MSIXBM register for that vector.
830 		 */
831 		if (rx_queue > IGB_N0_QUEUE)
832 			msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
833 		if (tx_queue > IGB_N0_QUEUE)
834 			msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
835 		if (!(adapter->flags & IGB_FLAG_HAS_MSIX) && msix_vector == 0)
836 			msixbm |= E1000_EIMS_OTHER;
837 		array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
838 		q_vector->eims_value = msixbm;
839 		break;
840 	case e1000_82576:
841 		/* 82576 uses a table that essentially consists of 2 columns
842 		 * with 8 rows.  The ordering is column-major so we use the
843 		 * lower 3 bits as the row index, and the 4th bit as the
844 		 * column offset.
845 		 */
846 		if (rx_queue > IGB_N0_QUEUE)
847 			igb_write_ivar(hw, msix_vector,
848 				       rx_queue & 0x7,
849 				       (rx_queue & 0x8) << 1);
850 		if (tx_queue > IGB_N0_QUEUE)
851 			igb_write_ivar(hw, msix_vector,
852 				       tx_queue & 0x7,
853 				       ((tx_queue & 0x8) << 1) + 8);
854 		q_vector->eims_value = BIT(msix_vector);
855 		break;
856 	case e1000_82580:
857 	case e1000_i350:
858 	case e1000_i354:
859 	case e1000_i210:
860 	case e1000_i211:
861 		/* On 82580 and newer adapters the scheme is similar to 82576
862 		 * however instead of ordering column-major we have things
863 		 * ordered row-major.  So we traverse the table by using
864 		 * bit 0 as the column offset, and the remaining bits as the
865 		 * row index.
866 		 */
867 		if (rx_queue > IGB_N0_QUEUE)
868 			igb_write_ivar(hw, msix_vector,
869 				       rx_queue >> 1,
870 				       (rx_queue & 0x1) << 4);
871 		if (tx_queue > IGB_N0_QUEUE)
872 			igb_write_ivar(hw, msix_vector,
873 				       tx_queue >> 1,
874 				       ((tx_queue & 0x1) << 4) + 8);
875 		q_vector->eims_value = BIT(msix_vector);
876 		break;
877 	default:
878 		BUG();
879 		break;
880 	}
881 
882 	/* add q_vector eims value to global eims_enable_mask */
883 	adapter->eims_enable_mask |= q_vector->eims_value;
884 
885 	/* configure q_vector to set itr on first interrupt */
886 	q_vector->set_itr = 1;
887 }
888 
889 /**
890  *  igb_configure_msix - Configure MSI-X hardware
891  *  @adapter: board private structure to initialize
892  *
893  *  igb_configure_msix sets up the hardware to properly
894  *  generate MSI-X interrupts.
895  **/
896 static void igb_configure_msix(struct igb_adapter *adapter)
897 {
898 	u32 tmp;
899 	int i, vector = 0;
900 	struct e1000_hw *hw = &adapter->hw;
901 
902 	adapter->eims_enable_mask = 0;
903 
904 	/* set vector for other causes, i.e. link changes */
905 	switch (hw->mac.type) {
906 	case e1000_82575:
907 		tmp = rd32(E1000_CTRL_EXT);
908 		/* enable MSI-X PBA support*/
909 		tmp |= E1000_CTRL_EXT_PBA_CLR;
910 
911 		/* Auto-Mask interrupts upon ICR read. */
912 		tmp |= E1000_CTRL_EXT_EIAME;
913 		tmp |= E1000_CTRL_EXT_IRCA;
914 
915 		wr32(E1000_CTRL_EXT, tmp);
916 
917 		/* enable msix_other interrupt */
918 		array_wr32(E1000_MSIXBM(0), vector++, E1000_EIMS_OTHER);
919 		adapter->eims_other = E1000_EIMS_OTHER;
920 
921 		break;
922 
923 	case e1000_82576:
924 	case e1000_82580:
925 	case e1000_i350:
926 	case e1000_i354:
927 	case e1000_i210:
928 	case e1000_i211:
929 		/* Turn on MSI-X capability first, or our settings
930 		 * won't stick.  And it will take days to debug.
931 		 */
932 		wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
933 		     E1000_GPIE_PBA | E1000_GPIE_EIAME |
934 		     E1000_GPIE_NSICR);
935 
936 		/* enable msix_other interrupt */
937 		adapter->eims_other = BIT(vector);
938 		tmp = (vector++ | E1000_IVAR_VALID) << 8;
939 
940 		wr32(E1000_IVAR_MISC, tmp);
941 		break;
942 	default:
943 		/* do nothing, since nothing else supports MSI-X */
944 		break;
945 	} /* switch (hw->mac.type) */
946 
947 	adapter->eims_enable_mask |= adapter->eims_other;
948 
949 	for (i = 0; i < adapter->num_q_vectors; i++)
950 		igb_assign_vector(adapter->q_vector[i], vector++);
951 
952 	wrfl();
953 }
954 
955 /**
956  *  igb_request_msix - Initialize MSI-X interrupts
957  *  @adapter: board private structure to initialize
958  *
959  *  igb_request_msix allocates MSI-X vectors and requests interrupts from the
960  *  kernel.
961  **/
962 static int igb_request_msix(struct igb_adapter *adapter)
963 {
964 	struct net_device *netdev = adapter->netdev;
965 	int i, err = 0, vector = 0, free_vector = 0;
966 
967 	err = request_irq(adapter->msix_entries[vector].vector,
968 			  igb_msix_other, 0, netdev->name, adapter);
969 	if (err)
970 		goto err_out;
971 
972 	for (i = 0; i < adapter->num_q_vectors; i++) {
973 		struct igb_q_vector *q_vector = adapter->q_vector[i];
974 
975 		vector++;
976 
977 		q_vector->itr_register = adapter->io_addr + E1000_EITR(vector);
978 
979 		if (q_vector->rx.ring && q_vector->tx.ring)
980 			sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
981 				q_vector->rx.ring->queue_index);
982 		else if (q_vector->tx.ring)
983 			sprintf(q_vector->name, "%s-tx-%u", netdev->name,
984 				q_vector->tx.ring->queue_index);
985 		else if (q_vector->rx.ring)
986 			sprintf(q_vector->name, "%s-rx-%u", netdev->name,
987 				q_vector->rx.ring->queue_index);
988 		else
989 			sprintf(q_vector->name, "%s-unused", netdev->name);
990 
991 		err = request_irq(adapter->msix_entries[vector].vector,
992 				  igb_msix_ring, 0, q_vector->name,
993 				  q_vector);
994 		if (err)
995 			goto err_free;
996 	}
997 
998 	igb_configure_msix(adapter);
999 	return 0;
1000 
1001 err_free:
1002 	/* free already assigned IRQs */
1003 	free_irq(adapter->msix_entries[free_vector++].vector, adapter);
1004 
1005 	vector--;
1006 	for (i = 0; i < vector; i++) {
1007 		free_irq(adapter->msix_entries[free_vector++].vector,
1008 			 adapter->q_vector[i]);
1009 	}
1010 err_out:
1011 	return err;
1012 }
1013 
1014 /**
1015  *  igb_free_q_vector - Free memory allocated for specific interrupt vector
1016  *  @adapter: board private structure to initialize
1017  *  @v_idx: Index of vector to be freed
1018  *
1019  *  This function frees the memory allocated to the q_vector.
1020  **/
1021 static void igb_free_q_vector(struct igb_adapter *adapter, int v_idx)
1022 {
1023 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1024 
1025 	adapter->q_vector[v_idx] = NULL;
1026 
1027 	/* igb_get_stats64() might access the rings on this vector,
1028 	 * we must wait a grace period before freeing it.
1029 	 */
1030 	if (q_vector)
1031 		kfree_rcu(q_vector, rcu);
1032 }
1033 
1034 /**
1035  *  igb_reset_q_vector - Reset config for interrupt vector
1036  *  @adapter: board private structure to initialize
1037  *  @v_idx: Index of vector to be reset
1038  *
1039  *  If NAPI is enabled it will delete any references to the
1040  *  NAPI struct. This is preparation for igb_free_q_vector.
1041  **/
1042 static void igb_reset_q_vector(struct igb_adapter *adapter, int v_idx)
1043 {
1044 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1045 
1046 	/* Coming from igb_set_interrupt_capability, the vectors are not yet
1047 	 * allocated. So, q_vector is NULL so we should stop here.
1048 	 */
1049 	if (!q_vector)
1050 		return;
1051 
1052 	if (q_vector->tx.ring)
1053 		adapter->tx_ring[q_vector->tx.ring->queue_index] = NULL;
1054 
1055 	if (q_vector->rx.ring)
1056 		adapter->rx_ring[q_vector->rx.ring->queue_index] = NULL;
1057 
1058 	netif_napi_del(&q_vector->napi);
1059 
1060 }
1061 
1062 static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
1063 {
1064 	int v_idx = adapter->num_q_vectors;
1065 
1066 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
1067 		pci_disable_msix(adapter->pdev);
1068 	else if (adapter->flags & IGB_FLAG_HAS_MSI)
1069 		pci_disable_msi(adapter->pdev);
1070 
1071 	while (v_idx--)
1072 		igb_reset_q_vector(adapter, v_idx);
1073 }
1074 
1075 /**
1076  *  igb_free_q_vectors - Free memory allocated for interrupt vectors
1077  *  @adapter: board private structure to initialize
1078  *
1079  *  This function frees the memory allocated to the q_vectors.  In addition if
1080  *  NAPI is enabled it will delete any references to the NAPI struct prior
1081  *  to freeing the q_vector.
1082  **/
1083 static void igb_free_q_vectors(struct igb_adapter *adapter)
1084 {
1085 	int v_idx = adapter->num_q_vectors;
1086 
1087 	adapter->num_tx_queues = 0;
1088 	adapter->num_rx_queues = 0;
1089 	adapter->num_q_vectors = 0;
1090 
1091 	while (v_idx--) {
1092 		igb_reset_q_vector(adapter, v_idx);
1093 		igb_free_q_vector(adapter, v_idx);
1094 	}
1095 }
1096 
1097 /**
1098  *  igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1099  *  @adapter: board private structure to initialize
1100  *
1101  *  This function resets the device so that it has 0 Rx queues, Tx queues, and
1102  *  MSI-X interrupts allocated.
1103  */
1104 static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
1105 {
1106 	igb_free_q_vectors(adapter);
1107 	igb_reset_interrupt_capability(adapter);
1108 }
1109 
1110 /**
1111  *  igb_set_interrupt_capability - set MSI or MSI-X if supported
1112  *  @adapter: board private structure to initialize
1113  *  @msix: boolean value of MSIX capability
1114  *
1115  *  Attempt to configure interrupts using the best available
1116  *  capabilities of the hardware and kernel.
1117  **/
1118 static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
1119 {
1120 	int err;
1121 	int numvecs, i;
1122 
1123 	if (!msix)
1124 		goto msi_only;
1125 	adapter->flags |= IGB_FLAG_HAS_MSIX;
1126 
1127 	/* Number of supported queues. */
1128 	adapter->num_rx_queues = adapter->rss_queues;
1129 	if (adapter->vfs_allocated_count)
1130 		adapter->num_tx_queues = 1;
1131 	else
1132 		adapter->num_tx_queues = adapter->rss_queues;
1133 
1134 	/* start with one vector for every Rx queue */
1135 	numvecs = adapter->num_rx_queues;
1136 
1137 	/* if Tx handler is separate add 1 for every Tx queue */
1138 	if (!(adapter->flags & IGB_FLAG_QUEUE_PAIRS))
1139 		numvecs += adapter->num_tx_queues;
1140 
1141 	/* store the number of vectors reserved for queues */
1142 	adapter->num_q_vectors = numvecs;
1143 
1144 	/* add 1 vector for link status interrupts */
1145 	numvecs++;
1146 	for (i = 0; i < numvecs; i++)
1147 		adapter->msix_entries[i].entry = i;
1148 
1149 	err = pci_enable_msix_range(adapter->pdev,
1150 				    adapter->msix_entries,
1151 				    numvecs,
1152 				    numvecs);
1153 	if (err > 0)
1154 		return;
1155 
1156 	igb_reset_interrupt_capability(adapter);
1157 
1158 	/* If we can't do MSI-X, try MSI */
1159 msi_only:
1160 	adapter->flags &= ~IGB_FLAG_HAS_MSIX;
1161 #ifdef CONFIG_PCI_IOV
1162 	/* disable SR-IOV for non MSI-X configurations */
1163 	if (adapter->vf_data) {
1164 		struct e1000_hw *hw = &adapter->hw;
1165 		/* disable iov and allow time for transactions to clear */
1166 		pci_disable_sriov(adapter->pdev);
1167 		msleep(500);
1168 
1169 		kfree(adapter->vf_mac_list);
1170 		adapter->vf_mac_list = NULL;
1171 		kfree(adapter->vf_data);
1172 		adapter->vf_data = NULL;
1173 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1174 		wrfl();
1175 		msleep(100);
1176 		dev_info(&adapter->pdev->dev, "IOV Disabled\n");
1177 	}
1178 #endif
1179 	adapter->vfs_allocated_count = 0;
1180 	adapter->rss_queues = 1;
1181 	adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
1182 	adapter->num_rx_queues = 1;
1183 	adapter->num_tx_queues = 1;
1184 	adapter->num_q_vectors = 1;
1185 	if (!pci_enable_msi(adapter->pdev))
1186 		adapter->flags |= IGB_FLAG_HAS_MSI;
1187 }
1188 
1189 static void igb_add_ring(struct igb_ring *ring,
1190 			 struct igb_ring_container *head)
1191 {
1192 	head->ring = ring;
1193 	head->count++;
1194 }
1195 
1196 /**
1197  *  igb_alloc_q_vector - Allocate memory for a single interrupt vector
1198  *  @adapter: board private structure to initialize
1199  *  @v_count: q_vectors allocated on adapter, used for ring interleaving
1200  *  @v_idx: index of vector in adapter struct
1201  *  @txr_count: total number of Tx rings to allocate
1202  *  @txr_idx: index of first Tx ring to allocate
1203  *  @rxr_count: total number of Rx rings to allocate
1204  *  @rxr_idx: index of first Rx ring to allocate
1205  *
1206  *  We allocate one q_vector.  If allocation fails we return -ENOMEM.
1207  **/
1208 static int igb_alloc_q_vector(struct igb_adapter *adapter,
1209 			      int v_count, int v_idx,
1210 			      int txr_count, int txr_idx,
1211 			      int rxr_count, int rxr_idx)
1212 {
1213 	struct igb_q_vector *q_vector;
1214 	struct igb_ring *ring;
1215 	int ring_count, size;
1216 
1217 	/* igb only supports 1 Tx and/or 1 Rx queue per vector */
1218 	if (txr_count > 1 || rxr_count > 1)
1219 		return -ENOMEM;
1220 
1221 	ring_count = txr_count + rxr_count;
1222 	size = sizeof(struct igb_q_vector) +
1223 	       (sizeof(struct igb_ring) * ring_count);
1224 
1225 	/* allocate q_vector and rings */
1226 	q_vector = adapter->q_vector[v_idx];
1227 	if (!q_vector) {
1228 		q_vector = kzalloc(size, GFP_KERNEL);
1229 	} else if (size > ksize(q_vector)) {
1230 		kfree_rcu(q_vector, rcu);
1231 		q_vector = kzalloc(size, GFP_KERNEL);
1232 	} else {
1233 		memset(q_vector, 0, size);
1234 	}
1235 	if (!q_vector)
1236 		return -ENOMEM;
1237 
1238 	/* initialize NAPI */
1239 	netif_napi_add(adapter->netdev, &q_vector->napi,
1240 		       igb_poll, 64);
1241 
1242 	/* tie q_vector and adapter together */
1243 	adapter->q_vector[v_idx] = q_vector;
1244 	q_vector->adapter = adapter;
1245 
1246 	/* initialize work limits */
1247 	q_vector->tx.work_limit = adapter->tx_work_limit;
1248 
1249 	/* initialize ITR configuration */
1250 	q_vector->itr_register = adapter->io_addr + E1000_EITR(0);
1251 	q_vector->itr_val = IGB_START_ITR;
1252 
1253 	/* initialize pointer to rings */
1254 	ring = q_vector->ring;
1255 
1256 	/* intialize ITR */
1257 	if (rxr_count) {
1258 		/* rx or rx/tx vector */
1259 		if (!adapter->rx_itr_setting || adapter->rx_itr_setting > 3)
1260 			q_vector->itr_val = adapter->rx_itr_setting;
1261 	} else {
1262 		/* tx only vector */
1263 		if (!adapter->tx_itr_setting || adapter->tx_itr_setting > 3)
1264 			q_vector->itr_val = adapter->tx_itr_setting;
1265 	}
1266 
1267 	if (txr_count) {
1268 		/* assign generic ring traits */
1269 		ring->dev = &adapter->pdev->dev;
1270 		ring->netdev = adapter->netdev;
1271 
1272 		/* configure backlink on ring */
1273 		ring->q_vector = q_vector;
1274 
1275 		/* update q_vector Tx values */
1276 		igb_add_ring(ring, &q_vector->tx);
1277 
1278 		/* For 82575, context index must be unique per ring. */
1279 		if (adapter->hw.mac.type == e1000_82575)
1280 			set_bit(IGB_RING_FLAG_TX_CTX_IDX, &ring->flags);
1281 
1282 		/* apply Tx specific ring traits */
1283 		ring->count = adapter->tx_ring_count;
1284 		ring->queue_index = txr_idx;
1285 
1286 		ring->cbs_enable = false;
1287 		ring->idleslope = 0;
1288 		ring->sendslope = 0;
1289 		ring->hicredit = 0;
1290 		ring->locredit = 0;
1291 
1292 		u64_stats_init(&ring->tx_syncp);
1293 		u64_stats_init(&ring->tx_syncp2);
1294 
1295 		/* assign ring to adapter */
1296 		adapter->tx_ring[txr_idx] = ring;
1297 
1298 		/* push pointer to next ring */
1299 		ring++;
1300 	}
1301 
1302 	if (rxr_count) {
1303 		/* assign generic ring traits */
1304 		ring->dev = &adapter->pdev->dev;
1305 		ring->netdev = adapter->netdev;
1306 
1307 		/* configure backlink on ring */
1308 		ring->q_vector = q_vector;
1309 
1310 		/* update q_vector Rx values */
1311 		igb_add_ring(ring, &q_vector->rx);
1312 
1313 		/* set flag indicating ring supports SCTP checksum offload */
1314 		if (adapter->hw.mac.type >= e1000_82576)
1315 			set_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags);
1316 
1317 		/* On i350, i354, i210, and i211, loopback VLAN packets
1318 		 * have the tag byte-swapped.
1319 		 */
1320 		if (adapter->hw.mac.type >= e1000_i350)
1321 			set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &ring->flags);
1322 
1323 		/* apply Rx specific ring traits */
1324 		ring->count = adapter->rx_ring_count;
1325 		ring->queue_index = rxr_idx;
1326 
1327 		u64_stats_init(&ring->rx_syncp);
1328 
1329 		/* assign ring to adapter */
1330 		adapter->rx_ring[rxr_idx] = ring;
1331 	}
1332 
1333 	return 0;
1334 }
1335 
1336 
1337 /**
1338  *  igb_alloc_q_vectors - Allocate memory for interrupt vectors
1339  *  @adapter: board private structure to initialize
1340  *
1341  *  We allocate one q_vector per queue interrupt.  If allocation fails we
1342  *  return -ENOMEM.
1343  **/
1344 static int igb_alloc_q_vectors(struct igb_adapter *adapter)
1345 {
1346 	int q_vectors = adapter->num_q_vectors;
1347 	int rxr_remaining = adapter->num_rx_queues;
1348 	int txr_remaining = adapter->num_tx_queues;
1349 	int rxr_idx = 0, txr_idx = 0, v_idx = 0;
1350 	int err;
1351 
1352 	if (q_vectors >= (rxr_remaining + txr_remaining)) {
1353 		for (; rxr_remaining; v_idx++) {
1354 			err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1355 						 0, 0, 1, rxr_idx);
1356 
1357 			if (err)
1358 				goto err_out;
1359 
1360 			/* update counts and index */
1361 			rxr_remaining--;
1362 			rxr_idx++;
1363 		}
1364 	}
1365 
1366 	for (; v_idx < q_vectors; v_idx++) {
1367 		int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
1368 		int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);
1369 
1370 		err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1371 					 tqpv, txr_idx, rqpv, rxr_idx);
1372 
1373 		if (err)
1374 			goto err_out;
1375 
1376 		/* update counts and index */
1377 		rxr_remaining -= rqpv;
1378 		txr_remaining -= tqpv;
1379 		rxr_idx++;
1380 		txr_idx++;
1381 	}
1382 
1383 	return 0;
1384 
1385 err_out:
1386 	adapter->num_tx_queues = 0;
1387 	adapter->num_rx_queues = 0;
1388 	adapter->num_q_vectors = 0;
1389 
1390 	while (v_idx--)
1391 		igb_free_q_vector(adapter, v_idx);
1392 
1393 	return -ENOMEM;
1394 }
1395 
1396 /**
1397  *  igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1398  *  @adapter: board private structure to initialize
1399  *  @msix: boolean value of MSIX capability
1400  *
1401  *  This function initializes the interrupts and allocates all of the queues.
1402  **/
1403 static int igb_init_interrupt_scheme(struct igb_adapter *adapter, bool msix)
1404 {
1405 	struct pci_dev *pdev = adapter->pdev;
1406 	int err;
1407 
1408 	igb_set_interrupt_capability(adapter, msix);
1409 
1410 	err = igb_alloc_q_vectors(adapter);
1411 	if (err) {
1412 		dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
1413 		goto err_alloc_q_vectors;
1414 	}
1415 
1416 	igb_cache_ring_register(adapter);
1417 
1418 	return 0;
1419 
1420 err_alloc_q_vectors:
1421 	igb_reset_interrupt_capability(adapter);
1422 	return err;
1423 }
1424 
1425 /**
1426  *  igb_request_irq - initialize interrupts
1427  *  @adapter: board private structure to initialize
1428  *
1429  *  Attempts to configure interrupts using the best available
1430  *  capabilities of the hardware and kernel.
1431  **/
1432 static int igb_request_irq(struct igb_adapter *adapter)
1433 {
1434 	struct net_device *netdev = adapter->netdev;
1435 	struct pci_dev *pdev = adapter->pdev;
1436 	int err = 0;
1437 
1438 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1439 		err = igb_request_msix(adapter);
1440 		if (!err)
1441 			goto request_done;
1442 		/* fall back to MSI */
1443 		igb_free_all_tx_resources(adapter);
1444 		igb_free_all_rx_resources(adapter);
1445 
1446 		igb_clear_interrupt_scheme(adapter);
1447 		err = igb_init_interrupt_scheme(adapter, false);
1448 		if (err)
1449 			goto request_done;
1450 
1451 		igb_setup_all_tx_resources(adapter);
1452 		igb_setup_all_rx_resources(adapter);
1453 		igb_configure(adapter);
1454 	}
1455 
1456 	igb_assign_vector(adapter->q_vector[0], 0);
1457 
1458 	if (adapter->flags & IGB_FLAG_HAS_MSI) {
1459 		err = request_irq(pdev->irq, igb_intr_msi, 0,
1460 				  netdev->name, adapter);
1461 		if (!err)
1462 			goto request_done;
1463 
1464 		/* fall back to legacy interrupts */
1465 		igb_reset_interrupt_capability(adapter);
1466 		adapter->flags &= ~IGB_FLAG_HAS_MSI;
1467 	}
1468 
1469 	err = request_irq(pdev->irq, igb_intr, IRQF_SHARED,
1470 			  netdev->name, adapter);
1471 
1472 	if (err)
1473 		dev_err(&pdev->dev, "Error %d getting interrupt\n",
1474 			err);
1475 
1476 request_done:
1477 	return err;
1478 }
1479 
1480 static void igb_free_irq(struct igb_adapter *adapter)
1481 {
1482 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1483 		int vector = 0, i;
1484 
1485 		free_irq(adapter->msix_entries[vector++].vector, adapter);
1486 
1487 		for (i = 0; i < adapter->num_q_vectors; i++)
1488 			free_irq(adapter->msix_entries[vector++].vector,
1489 				 adapter->q_vector[i]);
1490 	} else {
1491 		free_irq(adapter->pdev->irq, adapter);
1492 	}
1493 }
1494 
1495 /**
1496  *  igb_irq_disable - Mask off interrupt generation on the NIC
1497  *  @adapter: board private structure
1498  **/
1499 static void igb_irq_disable(struct igb_adapter *adapter)
1500 {
1501 	struct e1000_hw *hw = &adapter->hw;
1502 
1503 	/* we need to be careful when disabling interrupts.  The VFs are also
1504 	 * mapped into these registers and so clearing the bits can cause
1505 	 * issues on the VF drivers so we only need to clear what we set
1506 	 */
1507 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1508 		u32 regval = rd32(E1000_EIAM);
1509 
1510 		wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
1511 		wr32(E1000_EIMC, adapter->eims_enable_mask);
1512 		regval = rd32(E1000_EIAC);
1513 		wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
1514 	}
1515 
1516 	wr32(E1000_IAM, 0);
1517 	wr32(E1000_IMC, ~0);
1518 	wrfl();
1519 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1520 		int i;
1521 
1522 		for (i = 0; i < adapter->num_q_vectors; i++)
1523 			synchronize_irq(adapter->msix_entries[i].vector);
1524 	} else {
1525 		synchronize_irq(adapter->pdev->irq);
1526 	}
1527 }
1528 
1529 /**
1530  *  igb_irq_enable - Enable default interrupt generation settings
1531  *  @adapter: board private structure
1532  **/
1533 static void igb_irq_enable(struct igb_adapter *adapter)
1534 {
1535 	struct e1000_hw *hw = &adapter->hw;
1536 
1537 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1538 		u32 ims = E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_DRSTA;
1539 		u32 regval = rd32(E1000_EIAC);
1540 
1541 		wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
1542 		regval = rd32(E1000_EIAM);
1543 		wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
1544 		wr32(E1000_EIMS, adapter->eims_enable_mask);
1545 		if (adapter->vfs_allocated_count) {
1546 			wr32(E1000_MBVFIMR, 0xFF);
1547 			ims |= E1000_IMS_VMMB;
1548 		}
1549 		wr32(E1000_IMS, ims);
1550 	} else {
1551 		wr32(E1000_IMS, IMS_ENABLE_MASK |
1552 				E1000_IMS_DRSTA);
1553 		wr32(E1000_IAM, IMS_ENABLE_MASK |
1554 				E1000_IMS_DRSTA);
1555 	}
1556 }
1557 
1558 static void igb_update_mng_vlan(struct igb_adapter *adapter)
1559 {
1560 	struct e1000_hw *hw = &adapter->hw;
1561 	u16 pf_id = adapter->vfs_allocated_count;
1562 	u16 vid = adapter->hw.mng_cookie.vlan_id;
1563 	u16 old_vid = adapter->mng_vlan_id;
1564 
1565 	if (hw->mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1566 		/* add VID to filter table */
1567 		igb_vfta_set(hw, vid, pf_id, true, true);
1568 		adapter->mng_vlan_id = vid;
1569 	} else {
1570 		adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1571 	}
1572 
1573 	if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1574 	    (vid != old_vid) &&
1575 	    !test_bit(old_vid, adapter->active_vlans)) {
1576 		/* remove VID from filter table */
1577 		igb_vfta_set(hw, vid, pf_id, false, true);
1578 	}
1579 }
1580 
1581 /**
1582  *  igb_release_hw_control - release control of the h/w to f/w
1583  *  @adapter: address of board private structure
1584  *
1585  *  igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1586  *  For ASF and Pass Through versions of f/w this means that the
1587  *  driver is no longer loaded.
1588  **/
1589 static void igb_release_hw_control(struct igb_adapter *adapter)
1590 {
1591 	struct e1000_hw *hw = &adapter->hw;
1592 	u32 ctrl_ext;
1593 
1594 	/* Let firmware take over control of h/w */
1595 	ctrl_ext = rd32(E1000_CTRL_EXT);
1596 	wr32(E1000_CTRL_EXT,
1597 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1598 }
1599 
1600 /**
1601  *  igb_get_hw_control - get control of the h/w from f/w
1602  *  @adapter: address of board private structure
1603  *
1604  *  igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1605  *  For ASF and Pass Through versions of f/w this means that
1606  *  the driver is loaded.
1607  **/
1608 static void igb_get_hw_control(struct igb_adapter *adapter)
1609 {
1610 	struct e1000_hw *hw = &adapter->hw;
1611 	u32 ctrl_ext;
1612 
1613 	/* Let firmware know the driver has taken over */
1614 	ctrl_ext = rd32(E1000_CTRL_EXT);
1615 	wr32(E1000_CTRL_EXT,
1616 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1617 }
1618 
1619 static void enable_fqtss(struct igb_adapter *adapter, bool enable)
1620 {
1621 	struct net_device *netdev = adapter->netdev;
1622 	struct e1000_hw *hw = &adapter->hw;
1623 
1624 	WARN_ON(hw->mac.type != e1000_i210);
1625 
1626 	if (enable)
1627 		adapter->flags |= IGB_FLAG_FQTSS;
1628 	else
1629 		adapter->flags &= ~IGB_FLAG_FQTSS;
1630 
1631 	if (netif_running(netdev))
1632 		schedule_work(&adapter->reset_task);
1633 }
1634 
1635 static bool is_fqtss_enabled(struct igb_adapter *adapter)
1636 {
1637 	return (adapter->flags & IGB_FLAG_FQTSS) ? true : false;
1638 }
1639 
1640 static void set_tx_desc_fetch_prio(struct e1000_hw *hw, int queue,
1641 				   enum tx_queue_prio prio)
1642 {
1643 	u32 val;
1644 
1645 	WARN_ON(hw->mac.type != e1000_i210);
1646 	WARN_ON(queue < 0 || queue > 4);
1647 
1648 	val = rd32(E1000_I210_TXDCTL(queue));
1649 
1650 	if (prio == TX_QUEUE_PRIO_HIGH)
1651 		val |= E1000_TXDCTL_PRIORITY;
1652 	else
1653 		val &= ~E1000_TXDCTL_PRIORITY;
1654 
1655 	wr32(E1000_I210_TXDCTL(queue), val);
1656 }
1657 
1658 static void set_queue_mode(struct e1000_hw *hw, int queue, enum queue_mode mode)
1659 {
1660 	u32 val;
1661 
1662 	WARN_ON(hw->mac.type != e1000_i210);
1663 	WARN_ON(queue < 0 || queue > 1);
1664 
1665 	val = rd32(E1000_I210_TQAVCC(queue));
1666 
1667 	if (mode == QUEUE_MODE_STREAM_RESERVATION)
1668 		val |= E1000_TQAVCC_QUEUEMODE;
1669 	else
1670 		val &= ~E1000_TQAVCC_QUEUEMODE;
1671 
1672 	wr32(E1000_I210_TQAVCC(queue), val);
1673 }
1674 
1675 /**
1676  *  igb_configure_cbs - Configure Credit-Based Shaper (CBS)
1677  *  @adapter: pointer to adapter struct
1678  *  @queue: queue number
1679  *  @enable: true = enable CBS, false = disable CBS
1680  *  @idleslope: idleSlope in kbps
1681  *  @sendslope: sendSlope in kbps
1682  *  @hicredit: hiCredit in bytes
1683  *  @locredit: loCredit in bytes
1684  *
1685  *  Configure CBS for a given hardware queue. When disabling, idleslope,
1686  *  sendslope, hicredit, locredit arguments are ignored. Returns 0 if
1687  *  success. Negative otherwise.
1688  **/
1689 static void igb_configure_cbs(struct igb_adapter *adapter, int queue,
1690 			      bool enable, int idleslope, int sendslope,
1691 			      int hicredit, int locredit)
1692 {
1693 	struct net_device *netdev = adapter->netdev;
1694 	struct e1000_hw *hw = &adapter->hw;
1695 	u32 tqavcc;
1696 	u16 value;
1697 
1698 	WARN_ON(hw->mac.type != e1000_i210);
1699 	WARN_ON(queue < 0 || queue > 1);
1700 
1701 	if (enable) {
1702 		set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_HIGH);
1703 		set_queue_mode(hw, queue, QUEUE_MODE_STREAM_RESERVATION);
1704 
1705 		/* According to i210 datasheet section 7.2.7.7, we should set
1706 		 * the 'idleSlope' field from TQAVCC register following the
1707 		 * equation:
1708 		 *
1709 		 * For 100 Mbps link speed:
1710 		 *
1711 		 *     value = BW * 0x7735 * 0.2                          (E1)
1712 		 *
1713 		 * For 1000Mbps link speed:
1714 		 *
1715 		 *     value = BW * 0x7735 * 2                            (E2)
1716 		 *
1717 		 * E1 and E2 can be merged into one equation as shown below.
1718 		 * Note that 'link-speed' is in Mbps.
1719 		 *
1720 		 *     value = BW * 0x7735 * 2 * link-speed
1721 		 *                           --------------               (E3)
1722 		 *                                1000
1723 		 *
1724 		 * 'BW' is the percentage bandwidth out of full link speed
1725 		 * which can be found with the following equation. Note that
1726 		 * idleSlope here is the parameter from this function which
1727 		 * is in kbps.
1728 		 *
1729 		 *     BW =     idleSlope
1730 		 *          -----------------                             (E4)
1731 		 *          link-speed * 1000
1732 		 *
1733 		 * That said, we can come up with a generic equation to
1734 		 * calculate the value we should set it TQAVCC register by
1735 		 * replacing 'BW' in E3 by E4. The resulting equation is:
1736 		 *
1737 		 * value =     idleSlope     * 0x7735 * 2 * link-speed
1738 		 *         -----------------            --------------    (E5)
1739 		 *         link-speed * 1000                 1000
1740 		 *
1741 		 * 'link-speed' is present in both sides of the fraction so
1742 		 * it is canceled out. The final equation is the following:
1743 		 *
1744 		 *     value = idleSlope * 61034
1745 		 *             -----------------                          (E6)
1746 		 *                  1000000
1747 		 */
1748 		value = DIV_ROUND_UP_ULL(idleslope * 61034ULL, 1000000);
1749 
1750 		tqavcc = rd32(E1000_I210_TQAVCC(queue));
1751 		tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1752 		tqavcc |= value;
1753 		wr32(E1000_I210_TQAVCC(queue), tqavcc);
1754 
1755 		wr32(E1000_I210_TQAVHC(queue), 0x80000000 + hicredit * 0x7735);
1756 	} else {
1757 		set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_LOW);
1758 		set_queue_mode(hw, queue, QUEUE_MODE_STRICT_PRIORITY);
1759 
1760 		/* Set idleSlope to zero. */
1761 		tqavcc = rd32(E1000_I210_TQAVCC(queue));
1762 		tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1763 		wr32(E1000_I210_TQAVCC(queue), tqavcc);
1764 
1765 		/* Set hiCredit to zero. */
1766 		wr32(E1000_I210_TQAVHC(queue), 0);
1767 	}
1768 
1769 	/* XXX: In i210 controller the sendSlope and loCredit parameters from
1770 	 * CBS are not configurable by software so we don't do any 'controller
1771 	 * configuration' in respect to these parameters.
1772 	 */
1773 
1774 	netdev_dbg(netdev, "CBS %s: queue %d idleslope %d sendslope %d hiCredit %d locredit %d\n",
1775 		   (enable) ? "enabled" : "disabled", queue,
1776 		   idleslope, sendslope, hicredit, locredit);
1777 }
1778 
1779 static int igb_save_cbs_params(struct igb_adapter *adapter, int queue,
1780 			       bool enable, int idleslope, int sendslope,
1781 			       int hicredit, int locredit)
1782 {
1783 	struct igb_ring *ring;
1784 
1785 	if (queue < 0 || queue > adapter->num_tx_queues)
1786 		return -EINVAL;
1787 
1788 	ring = adapter->tx_ring[queue];
1789 
1790 	ring->cbs_enable = enable;
1791 	ring->idleslope = idleslope;
1792 	ring->sendslope = sendslope;
1793 	ring->hicredit = hicredit;
1794 	ring->locredit = locredit;
1795 
1796 	return 0;
1797 }
1798 
1799 static bool is_any_cbs_enabled(struct igb_adapter *adapter)
1800 {
1801 	struct igb_ring *ring;
1802 	int i;
1803 
1804 	for (i = 0; i < adapter->num_tx_queues; i++) {
1805 		ring = adapter->tx_ring[i];
1806 
1807 		if (ring->cbs_enable)
1808 			return true;
1809 	}
1810 
1811 	return false;
1812 }
1813 
1814 static void igb_setup_tx_mode(struct igb_adapter *adapter)
1815 {
1816 	struct net_device *netdev = adapter->netdev;
1817 	struct e1000_hw *hw = &adapter->hw;
1818 	u32 val;
1819 
1820 	/* Only i210 controller supports changing the transmission mode. */
1821 	if (hw->mac.type != e1000_i210)
1822 		return;
1823 
1824 	if (is_fqtss_enabled(adapter)) {
1825 		int i, max_queue;
1826 
1827 		/* Configure TQAVCTRL register: set transmit mode to 'Qav',
1828 		 * set data fetch arbitration to 'round robin' and set data
1829 		 * transfer arbitration to 'credit shaper algorithm.
1830 		 */
1831 		val = rd32(E1000_I210_TQAVCTRL);
1832 		val |= E1000_TQAVCTRL_XMIT_MODE | E1000_TQAVCTRL_DATATRANARB;
1833 		val &= ~E1000_TQAVCTRL_DATAFETCHARB;
1834 		wr32(E1000_I210_TQAVCTRL, val);
1835 
1836 		/* Configure Tx and Rx packet buffers sizes as described in
1837 		 * i210 datasheet section 7.2.7.7.
1838 		 */
1839 		val = rd32(E1000_TXPBS);
1840 		val &= ~I210_TXPBSIZE_MASK;
1841 		val |= I210_TXPBSIZE_PB0_8KB | I210_TXPBSIZE_PB1_8KB |
1842 			I210_TXPBSIZE_PB2_4KB | I210_TXPBSIZE_PB3_4KB;
1843 		wr32(E1000_TXPBS, val);
1844 
1845 		val = rd32(E1000_RXPBS);
1846 		val &= ~I210_RXPBSIZE_MASK;
1847 		val |= I210_RXPBSIZE_PB_32KB;
1848 		wr32(E1000_RXPBS, val);
1849 
1850 		/* Section 8.12.9 states that MAX_TPKT_SIZE from DTXMXPKTSZ
1851 		 * register should not exceed the buffer size programmed in
1852 		 * TXPBS. The smallest buffer size programmed in TXPBS is 4kB
1853 		 * so according to the datasheet we should set MAX_TPKT_SIZE to
1854 		 * 4kB / 64.
1855 		 *
1856 		 * However, when we do so, no frame from queue 2 and 3 are
1857 		 * transmitted.  It seems the MAX_TPKT_SIZE should not be great
1858 		 * or _equal_ to the buffer size programmed in TXPBS. For this
1859 		 * reason, we set set MAX_ TPKT_SIZE to (4kB - 1) / 64.
1860 		 */
1861 		val = (4096 - 1) / 64;
1862 		wr32(E1000_I210_DTXMXPKTSZ, val);
1863 
1864 		/* Since FQTSS mode is enabled, apply any CBS configuration
1865 		 * previously set. If no previous CBS configuration has been
1866 		 * done, then the initial configuration is applied, which means
1867 		 * CBS is disabled.
1868 		 */
1869 		max_queue = (adapter->num_tx_queues < I210_SR_QUEUES_NUM) ?
1870 			    adapter->num_tx_queues : I210_SR_QUEUES_NUM;
1871 
1872 		for (i = 0; i < max_queue; i++) {
1873 			struct igb_ring *ring = adapter->tx_ring[i];
1874 
1875 			igb_configure_cbs(adapter, i, ring->cbs_enable,
1876 					  ring->idleslope, ring->sendslope,
1877 					  ring->hicredit, ring->locredit);
1878 		}
1879 	} else {
1880 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
1881 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
1882 		wr32(E1000_I210_DTXMXPKTSZ, I210_DTXMXPKTSZ_DEFAULT);
1883 
1884 		val = rd32(E1000_I210_TQAVCTRL);
1885 		/* According to Section 8.12.21, the other flags we've set when
1886 		 * enabling FQTSS are not relevant when disabling FQTSS so we
1887 		 * don't set they here.
1888 		 */
1889 		val &= ~E1000_TQAVCTRL_XMIT_MODE;
1890 		wr32(E1000_I210_TQAVCTRL, val);
1891 	}
1892 
1893 	netdev_dbg(netdev, "FQTSS %s\n", (is_fqtss_enabled(adapter)) ?
1894 		   "enabled" : "disabled");
1895 }
1896 
1897 /**
1898  *  igb_configure - configure the hardware for RX and TX
1899  *  @adapter: private board structure
1900  **/
1901 static void igb_configure(struct igb_adapter *adapter)
1902 {
1903 	struct net_device *netdev = adapter->netdev;
1904 	int i;
1905 
1906 	igb_get_hw_control(adapter);
1907 	igb_set_rx_mode(netdev);
1908 	igb_setup_tx_mode(adapter);
1909 
1910 	igb_restore_vlan(adapter);
1911 
1912 	igb_setup_tctl(adapter);
1913 	igb_setup_mrqc(adapter);
1914 	igb_setup_rctl(adapter);
1915 
1916 	igb_nfc_filter_restore(adapter);
1917 	igb_configure_tx(adapter);
1918 	igb_configure_rx(adapter);
1919 
1920 	igb_rx_fifo_flush_82575(&adapter->hw);
1921 
1922 	/* call igb_desc_unused which always leaves
1923 	 * at least 1 descriptor unused to make sure
1924 	 * next_to_use != next_to_clean
1925 	 */
1926 	for (i = 0; i < adapter->num_rx_queues; i++) {
1927 		struct igb_ring *ring = adapter->rx_ring[i];
1928 		igb_alloc_rx_buffers(ring, igb_desc_unused(ring));
1929 	}
1930 }
1931 
1932 /**
1933  *  igb_power_up_link - Power up the phy/serdes link
1934  *  @adapter: address of board private structure
1935  **/
1936 void igb_power_up_link(struct igb_adapter *adapter)
1937 {
1938 	igb_reset_phy(&adapter->hw);
1939 
1940 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
1941 		igb_power_up_phy_copper(&adapter->hw);
1942 	else
1943 		igb_power_up_serdes_link_82575(&adapter->hw);
1944 
1945 	igb_setup_link(&adapter->hw);
1946 }
1947 
1948 /**
1949  *  igb_power_down_link - Power down the phy/serdes link
1950  *  @adapter: address of board private structure
1951  */
1952 static void igb_power_down_link(struct igb_adapter *adapter)
1953 {
1954 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
1955 		igb_power_down_phy_copper_82575(&adapter->hw);
1956 	else
1957 		igb_shutdown_serdes_link_82575(&adapter->hw);
1958 }
1959 
1960 /**
1961  * Detect and switch function for Media Auto Sense
1962  * @adapter: address of the board private structure
1963  **/
1964 static void igb_check_swap_media(struct igb_adapter *adapter)
1965 {
1966 	struct e1000_hw *hw = &adapter->hw;
1967 	u32 ctrl_ext, connsw;
1968 	bool swap_now = false;
1969 
1970 	ctrl_ext = rd32(E1000_CTRL_EXT);
1971 	connsw = rd32(E1000_CONNSW);
1972 
1973 	/* need to live swap if current media is copper and we have fiber/serdes
1974 	 * to go to.
1975 	 */
1976 
1977 	if ((hw->phy.media_type == e1000_media_type_copper) &&
1978 	    (!(connsw & E1000_CONNSW_AUTOSENSE_EN))) {
1979 		swap_now = true;
1980 	} else if (!(connsw & E1000_CONNSW_SERDESD)) {
1981 		/* copper signal takes time to appear */
1982 		if (adapter->copper_tries < 4) {
1983 			adapter->copper_tries++;
1984 			connsw |= E1000_CONNSW_AUTOSENSE_CONF;
1985 			wr32(E1000_CONNSW, connsw);
1986 			return;
1987 		} else {
1988 			adapter->copper_tries = 0;
1989 			if ((connsw & E1000_CONNSW_PHYSD) &&
1990 			    (!(connsw & E1000_CONNSW_PHY_PDN))) {
1991 				swap_now = true;
1992 				connsw &= ~E1000_CONNSW_AUTOSENSE_CONF;
1993 				wr32(E1000_CONNSW, connsw);
1994 			}
1995 		}
1996 	}
1997 
1998 	if (!swap_now)
1999 		return;
2000 
2001 	switch (hw->phy.media_type) {
2002 	case e1000_media_type_copper:
2003 		netdev_info(adapter->netdev,
2004 			"MAS: changing media to fiber/serdes\n");
2005 		ctrl_ext |=
2006 			E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2007 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
2008 		adapter->copper_tries = 0;
2009 		break;
2010 	case e1000_media_type_internal_serdes:
2011 	case e1000_media_type_fiber:
2012 		netdev_info(adapter->netdev,
2013 			"MAS: changing media to copper\n");
2014 		ctrl_ext &=
2015 			~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2016 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
2017 		break;
2018 	default:
2019 		/* shouldn't get here during regular operation */
2020 		netdev_err(adapter->netdev,
2021 			"AMS: Invalid media type found, returning\n");
2022 		break;
2023 	}
2024 	wr32(E1000_CTRL_EXT, ctrl_ext);
2025 }
2026 
2027 /**
2028  *  igb_up - Open the interface and prepare it to handle traffic
2029  *  @adapter: board private structure
2030  **/
2031 int igb_up(struct igb_adapter *adapter)
2032 {
2033 	struct e1000_hw *hw = &adapter->hw;
2034 	int i;
2035 
2036 	/* hardware has been reset, we need to reload some things */
2037 	igb_configure(adapter);
2038 
2039 	clear_bit(__IGB_DOWN, &adapter->state);
2040 
2041 	for (i = 0; i < adapter->num_q_vectors; i++)
2042 		napi_enable(&(adapter->q_vector[i]->napi));
2043 
2044 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
2045 		igb_configure_msix(adapter);
2046 	else
2047 		igb_assign_vector(adapter->q_vector[0], 0);
2048 
2049 	/* Clear any pending interrupts. */
2050 	rd32(E1000_ICR);
2051 	igb_irq_enable(adapter);
2052 
2053 	/* notify VFs that reset has been completed */
2054 	if (adapter->vfs_allocated_count) {
2055 		u32 reg_data = rd32(E1000_CTRL_EXT);
2056 
2057 		reg_data |= E1000_CTRL_EXT_PFRSTD;
2058 		wr32(E1000_CTRL_EXT, reg_data);
2059 	}
2060 
2061 	netif_tx_start_all_queues(adapter->netdev);
2062 
2063 	/* start the watchdog. */
2064 	hw->mac.get_link_status = 1;
2065 	schedule_work(&adapter->watchdog_task);
2066 
2067 	if ((adapter->flags & IGB_FLAG_EEE) &&
2068 	    (!hw->dev_spec._82575.eee_disable))
2069 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
2070 
2071 	return 0;
2072 }
2073 
2074 void igb_down(struct igb_adapter *adapter)
2075 {
2076 	struct net_device *netdev = adapter->netdev;
2077 	struct e1000_hw *hw = &adapter->hw;
2078 	u32 tctl, rctl;
2079 	int i;
2080 
2081 	/* signal that we're down so the interrupt handler does not
2082 	 * reschedule our watchdog timer
2083 	 */
2084 	set_bit(__IGB_DOWN, &adapter->state);
2085 
2086 	/* disable receives in the hardware */
2087 	rctl = rd32(E1000_RCTL);
2088 	wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
2089 	/* flush and sleep below */
2090 
2091 	igb_nfc_filter_exit(adapter);
2092 
2093 	netif_carrier_off(netdev);
2094 	netif_tx_stop_all_queues(netdev);
2095 
2096 	/* disable transmits in the hardware */
2097 	tctl = rd32(E1000_TCTL);
2098 	tctl &= ~E1000_TCTL_EN;
2099 	wr32(E1000_TCTL, tctl);
2100 	/* flush both disables and wait for them to finish */
2101 	wrfl();
2102 	usleep_range(10000, 11000);
2103 
2104 	igb_irq_disable(adapter);
2105 
2106 	adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
2107 
2108 	for (i = 0; i < adapter->num_q_vectors; i++) {
2109 		if (adapter->q_vector[i]) {
2110 			napi_synchronize(&adapter->q_vector[i]->napi);
2111 			napi_disable(&adapter->q_vector[i]->napi);
2112 		}
2113 	}
2114 
2115 	del_timer_sync(&adapter->watchdog_timer);
2116 	del_timer_sync(&adapter->phy_info_timer);
2117 
2118 	/* record the stats before reset*/
2119 	spin_lock(&adapter->stats64_lock);
2120 	igb_update_stats(adapter);
2121 	spin_unlock(&adapter->stats64_lock);
2122 
2123 	adapter->link_speed = 0;
2124 	adapter->link_duplex = 0;
2125 
2126 	if (!pci_channel_offline(adapter->pdev))
2127 		igb_reset(adapter);
2128 
2129 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
2130 	adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
2131 
2132 	igb_clean_all_tx_rings(adapter);
2133 	igb_clean_all_rx_rings(adapter);
2134 #ifdef CONFIG_IGB_DCA
2135 
2136 	/* since we reset the hardware DCA settings were cleared */
2137 	igb_setup_dca(adapter);
2138 #endif
2139 }
2140 
2141 void igb_reinit_locked(struct igb_adapter *adapter)
2142 {
2143 	WARN_ON(in_interrupt());
2144 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
2145 		usleep_range(1000, 2000);
2146 	igb_down(adapter);
2147 	igb_up(adapter);
2148 	clear_bit(__IGB_RESETTING, &adapter->state);
2149 }
2150 
2151 /** igb_enable_mas - Media Autosense re-enable after swap
2152  *
2153  * @adapter: adapter struct
2154  **/
2155 static void igb_enable_mas(struct igb_adapter *adapter)
2156 {
2157 	struct e1000_hw *hw = &adapter->hw;
2158 	u32 connsw = rd32(E1000_CONNSW);
2159 
2160 	/* configure for SerDes media detect */
2161 	if ((hw->phy.media_type == e1000_media_type_copper) &&
2162 	    (!(connsw & E1000_CONNSW_SERDESD))) {
2163 		connsw |= E1000_CONNSW_ENRGSRC;
2164 		connsw |= E1000_CONNSW_AUTOSENSE_EN;
2165 		wr32(E1000_CONNSW, connsw);
2166 		wrfl();
2167 	}
2168 }
2169 
2170 void igb_reset(struct igb_adapter *adapter)
2171 {
2172 	struct pci_dev *pdev = adapter->pdev;
2173 	struct e1000_hw *hw = &adapter->hw;
2174 	struct e1000_mac_info *mac = &hw->mac;
2175 	struct e1000_fc_info *fc = &hw->fc;
2176 	u32 pba, hwm;
2177 
2178 	/* Repartition Pba for greater than 9k mtu
2179 	 * To take effect CTRL.RST is required.
2180 	 */
2181 	switch (mac->type) {
2182 	case e1000_i350:
2183 	case e1000_i354:
2184 	case e1000_82580:
2185 		pba = rd32(E1000_RXPBS);
2186 		pba = igb_rxpbs_adjust_82580(pba);
2187 		break;
2188 	case e1000_82576:
2189 		pba = rd32(E1000_RXPBS);
2190 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2191 		break;
2192 	case e1000_82575:
2193 	case e1000_i210:
2194 	case e1000_i211:
2195 	default:
2196 		pba = E1000_PBA_34K;
2197 		break;
2198 	}
2199 
2200 	if (mac->type == e1000_82575) {
2201 		u32 min_rx_space, min_tx_space, needed_tx_space;
2202 
2203 		/* write Rx PBA so that hardware can report correct Tx PBA */
2204 		wr32(E1000_PBA, pba);
2205 
2206 		/* To maintain wire speed transmits, the Tx FIFO should be
2207 		 * large enough to accommodate two full transmit packets,
2208 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
2209 		 * the Rx FIFO should be large enough to accommodate at least
2210 		 * one full receive packet and is similarly rounded up and
2211 		 * expressed in KB.
2212 		 */
2213 		min_rx_space = DIV_ROUND_UP(MAX_JUMBO_FRAME_SIZE, 1024);
2214 
2215 		/* The Tx FIFO also stores 16 bytes of information about the Tx
2216 		 * but don't include Ethernet FCS because hardware appends it.
2217 		 * We only need to round down to the nearest 512 byte block
2218 		 * count since the value we care about is 2 frames, not 1.
2219 		 */
2220 		min_tx_space = adapter->max_frame_size;
2221 		min_tx_space += sizeof(union e1000_adv_tx_desc) - ETH_FCS_LEN;
2222 		min_tx_space = DIV_ROUND_UP(min_tx_space, 512);
2223 
2224 		/* upper 16 bits has Tx packet buffer allocation size in KB */
2225 		needed_tx_space = min_tx_space - (rd32(E1000_PBA) >> 16);
2226 
2227 		/* If current Tx allocation is less than the min Tx FIFO size,
2228 		 * and the min Tx FIFO size is less than the current Rx FIFO
2229 		 * allocation, take space away from current Rx allocation.
2230 		 */
2231 		if (needed_tx_space < pba) {
2232 			pba -= needed_tx_space;
2233 
2234 			/* if short on Rx space, Rx wins and must trump Tx
2235 			 * adjustment
2236 			 */
2237 			if (pba < min_rx_space)
2238 				pba = min_rx_space;
2239 		}
2240 
2241 		/* adjust PBA for jumbo frames */
2242 		wr32(E1000_PBA, pba);
2243 	}
2244 
2245 	/* flow control settings
2246 	 * The high water mark must be low enough to fit one full frame
2247 	 * after transmitting the pause frame.  As such we must have enough
2248 	 * space to allow for us to complete our current transmit and then
2249 	 * receive the frame that is in progress from the link partner.
2250 	 * Set it to:
2251 	 * - the full Rx FIFO size minus one full Tx plus one full Rx frame
2252 	 */
2253 	hwm = (pba << 10) - (adapter->max_frame_size + MAX_JUMBO_FRAME_SIZE);
2254 
2255 	fc->high_water = hwm & 0xFFFFFFF0;	/* 16-byte granularity */
2256 	fc->low_water = fc->high_water - 16;
2257 	fc->pause_time = 0xFFFF;
2258 	fc->send_xon = 1;
2259 	fc->current_mode = fc->requested_mode;
2260 
2261 	/* disable receive for all VFs and wait one second */
2262 	if (adapter->vfs_allocated_count) {
2263 		int i;
2264 
2265 		for (i = 0 ; i < adapter->vfs_allocated_count; i++)
2266 			adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
2267 
2268 		/* ping all the active vfs to let them know we are going down */
2269 		igb_ping_all_vfs(adapter);
2270 
2271 		/* disable transmits and receives */
2272 		wr32(E1000_VFRE, 0);
2273 		wr32(E1000_VFTE, 0);
2274 	}
2275 
2276 	/* Allow time for pending master requests to run */
2277 	hw->mac.ops.reset_hw(hw);
2278 	wr32(E1000_WUC, 0);
2279 
2280 	if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
2281 		/* need to resetup here after media swap */
2282 		adapter->ei.get_invariants(hw);
2283 		adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
2284 	}
2285 	if ((mac->type == e1000_82575) &&
2286 	    (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
2287 		igb_enable_mas(adapter);
2288 	}
2289 	if (hw->mac.ops.init_hw(hw))
2290 		dev_err(&pdev->dev, "Hardware Error\n");
2291 
2292 	/* RAR registers were cleared during init_hw, clear mac table */
2293 	igb_flush_mac_table(adapter);
2294 	__dev_uc_unsync(adapter->netdev, NULL);
2295 
2296 	/* Recover default RAR entry */
2297 	igb_set_default_mac_filter(adapter);
2298 
2299 	/* Flow control settings reset on hardware reset, so guarantee flow
2300 	 * control is off when forcing speed.
2301 	 */
2302 	if (!hw->mac.autoneg)
2303 		igb_force_mac_fc(hw);
2304 
2305 	igb_init_dmac(adapter, pba);
2306 #ifdef CONFIG_IGB_HWMON
2307 	/* Re-initialize the thermal sensor on i350 devices. */
2308 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
2309 		if (mac->type == e1000_i350 && hw->bus.func == 0) {
2310 			/* If present, re-initialize the external thermal sensor
2311 			 * interface.
2312 			 */
2313 			if (adapter->ets)
2314 				mac->ops.init_thermal_sensor_thresh(hw);
2315 		}
2316 	}
2317 #endif
2318 	/* Re-establish EEE setting */
2319 	if (hw->phy.media_type == e1000_media_type_copper) {
2320 		switch (mac->type) {
2321 		case e1000_i350:
2322 		case e1000_i210:
2323 		case e1000_i211:
2324 			igb_set_eee_i350(hw, true, true);
2325 			break;
2326 		case e1000_i354:
2327 			igb_set_eee_i354(hw, true, true);
2328 			break;
2329 		default:
2330 			break;
2331 		}
2332 	}
2333 	if (!netif_running(adapter->netdev))
2334 		igb_power_down_link(adapter);
2335 
2336 	igb_update_mng_vlan(adapter);
2337 
2338 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2339 	wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2340 
2341 	/* Re-enable PTP, where applicable. */
2342 	if (adapter->ptp_flags & IGB_PTP_ENABLED)
2343 		igb_ptp_reset(adapter);
2344 
2345 	igb_get_phy_info(hw);
2346 }
2347 
2348 static netdev_features_t igb_fix_features(struct net_device *netdev,
2349 	netdev_features_t features)
2350 {
2351 	/* Since there is no support for separate Rx/Tx vlan accel
2352 	 * enable/disable make sure Tx flag is always in same state as Rx.
2353 	 */
2354 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2355 		features |= NETIF_F_HW_VLAN_CTAG_TX;
2356 	else
2357 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2358 
2359 	return features;
2360 }
2361 
2362 static int igb_set_features(struct net_device *netdev,
2363 	netdev_features_t features)
2364 {
2365 	netdev_features_t changed = netdev->features ^ features;
2366 	struct igb_adapter *adapter = netdev_priv(netdev);
2367 
2368 	if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2369 		igb_vlan_mode(netdev, features);
2370 
2371 	if (!(changed & (NETIF_F_RXALL | NETIF_F_NTUPLE)))
2372 		return 0;
2373 
2374 	if (!(features & NETIF_F_NTUPLE)) {
2375 		struct hlist_node *node2;
2376 		struct igb_nfc_filter *rule;
2377 
2378 		spin_lock(&adapter->nfc_lock);
2379 		hlist_for_each_entry_safe(rule, node2,
2380 					  &adapter->nfc_filter_list, nfc_node) {
2381 			igb_erase_filter(adapter, rule);
2382 			hlist_del(&rule->nfc_node);
2383 			kfree(rule);
2384 		}
2385 		spin_unlock(&adapter->nfc_lock);
2386 		adapter->nfc_filter_count = 0;
2387 	}
2388 
2389 	netdev->features = features;
2390 
2391 	if (netif_running(netdev))
2392 		igb_reinit_locked(adapter);
2393 	else
2394 		igb_reset(adapter);
2395 
2396 	return 0;
2397 }
2398 
2399 static int igb_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
2400 			   struct net_device *dev,
2401 			   const unsigned char *addr, u16 vid,
2402 			   u16 flags)
2403 {
2404 	/* guarantee we can provide a unique filter for the unicast address */
2405 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
2406 		struct igb_adapter *adapter = netdev_priv(dev);
2407 		int vfn = adapter->vfs_allocated_count;
2408 
2409 		if (netdev_uc_count(dev) >= igb_available_rars(adapter, vfn))
2410 			return -ENOMEM;
2411 	}
2412 
2413 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
2414 }
2415 
2416 #define IGB_MAX_MAC_HDR_LEN	127
2417 #define IGB_MAX_NETWORK_HDR_LEN	511
2418 
2419 static netdev_features_t
2420 igb_features_check(struct sk_buff *skb, struct net_device *dev,
2421 		   netdev_features_t features)
2422 {
2423 	unsigned int network_hdr_len, mac_hdr_len;
2424 
2425 	/* Make certain the headers can be described by a context descriptor */
2426 	mac_hdr_len = skb_network_header(skb) - skb->data;
2427 	if (unlikely(mac_hdr_len > IGB_MAX_MAC_HDR_LEN))
2428 		return features & ~(NETIF_F_HW_CSUM |
2429 				    NETIF_F_SCTP_CRC |
2430 				    NETIF_F_HW_VLAN_CTAG_TX |
2431 				    NETIF_F_TSO |
2432 				    NETIF_F_TSO6);
2433 
2434 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
2435 	if (unlikely(network_hdr_len >  IGB_MAX_NETWORK_HDR_LEN))
2436 		return features & ~(NETIF_F_HW_CSUM |
2437 				    NETIF_F_SCTP_CRC |
2438 				    NETIF_F_TSO |
2439 				    NETIF_F_TSO6);
2440 
2441 	/* We can only support IPV4 TSO in tunnels if we can mangle the
2442 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
2443 	 */
2444 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
2445 		features &= ~NETIF_F_TSO;
2446 
2447 	return features;
2448 }
2449 
2450 static int igb_offload_cbs(struct igb_adapter *adapter,
2451 			   struct tc_cbs_qopt_offload *qopt)
2452 {
2453 	struct e1000_hw *hw = &adapter->hw;
2454 	int err;
2455 
2456 	/* CBS offloading is only supported by i210 controller. */
2457 	if (hw->mac.type != e1000_i210)
2458 		return -EOPNOTSUPP;
2459 
2460 	/* CBS offloading is only supported by queue 0 and queue 1. */
2461 	if (qopt->queue < 0 || qopt->queue > 1)
2462 		return -EINVAL;
2463 
2464 	err = igb_save_cbs_params(adapter, qopt->queue, qopt->enable,
2465 				  qopt->idleslope, qopt->sendslope,
2466 				  qopt->hicredit, qopt->locredit);
2467 	if (err)
2468 		return err;
2469 
2470 	if (is_fqtss_enabled(adapter)) {
2471 		igb_configure_cbs(adapter, qopt->queue, qopt->enable,
2472 				  qopt->idleslope, qopt->sendslope,
2473 				  qopt->hicredit, qopt->locredit);
2474 
2475 		if (!is_any_cbs_enabled(adapter))
2476 			enable_fqtss(adapter, false);
2477 
2478 	} else {
2479 		enable_fqtss(adapter, true);
2480 	}
2481 
2482 	return 0;
2483 }
2484 
2485 static int igb_setup_tc(struct net_device *dev, enum tc_setup_type type,
2486 			void *type_data)
2487 {
2488 	struct igb_adapter *adapter = netdev_priv(dev);
2489 
2490 	switch (type) {
2491 	case TC_SETUP_QDISC_CBS:
2492 		return igb_offload_cbs(adapter, type_data);
2493 
2494 	default:
2495 		return -EOPNOTSUPP;
2496 	}
2497 }
2498 
2499 static const struct net_device_ops igb_netdev_ops = {
2500 	.ndo_open		= igb_open,
2501 	.ndo_stop		= igb_close,
2502 	.ndo_start_xmit		= igb_xmit_frame,
2503 	.ndo_get_stats64	= igb_get_stats64,
2504 	.ndo_set_rx_mode	= igb_set_rx_mode,
2505 	.ndo_set_mac_address	= igb_set_mac,
2506 	.ndo_change_mtu		= igb_change_mtu,
2507 	.ndo_do_ioctl		= igb_ioctl,
2508 	.ndo_tx_timeout		= igb_tx_timeout,
2509 	.ndo_validate_addr	= eth_validate_addr,
2510 	.ndo_vlan_rx_add_vid	= igb_vlan_rx_add_vid,
2511 	.ndo_vlan_rx_kill_vid	= igb_vlan_rx_kill_vid,
2512 	.ndo_set_vf_mac		= igb_ndo_set_vf_mac,
2513 	.ndo_set_vf_vlan	= igb_ndo_set_vf_vlan,
2514 	.ndo_set_vf_rate	= igb_ndo_set_vf_bw,
2515 	.ndo_set_vf_spoofchk	= igb_ndo_set_vf_spoofchk,
2516 	.ndo_get_vf_config	= igb_ndo_get_vf_config,
2517 #ifdef CONFIG_NET_POLL_CONTROLLER
2518 	.ndo_poll_controller	= igb_netpoll,
2519 #endif
2520 	.ndo_fix_features	= igb_fix_features,
2521 	.ndo_set_features	= igb_set_features,
2522 	.ndo_fdb_add		= igb_ndo_fdb_add,
2523 	.ndo_features_check	= igb_features_check,
2524 	.ndo_setup_tc		= igb_setup_tc,
2525 };
2526 
2527 /**
2528  * igb_set_fw_version - Configure version string for ethtool
2529  * @adapter: adapter struct
2530  **/
2531 void igb_set_fw_version(struct igb_adapter *adapter)
2532 {
2533 	struct e1000_hw *hw = &adapter->hw;
2534 	struct e1000_fw_version fw;
2535 
2536 	igb_get_fw_version(hw, &fw);
2537 
2538 	switch (hw->mac.type) {
2539 	case e1000_i210:
2540 	case e1000_i211:
2541 		if (!(igb_get_flash_presence_i210(hw))) {
2542 			snprintf(adapter->fw_version,
2543 				 sizeof(adapter->fw_version),
2544 				 "%2d.%2d-%d",
2545 				 fw.invm_major, fw.invm_minor,
2546 				 fw.invm_img_type);
2547 			break;
2548 		}
2549 		/* fall through */
2550 	default:
2551 		/* if option is rom valid, display its version too */
2552 		if (fw.or_valid) {
2553 			snprintf(adapter->fw_version,
2554 				 sizeof(adapter->fw_version),
2555 				 "%d.%d, 0x%08x, %d.%d.%d",
2556 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2557 				 fw.or_major, fw.or_build, fw.or_patch);
2558 		/* no option rom */
2559 		} else if (fw.etrack_id != 0X0000) {
2560 			snprintf(adapter->fw_version,
2561 			    sizeof(adapter->fw_version),
2562 			    "%d.%d, 0x%08x",
2563 			    fw.eep_major, fw.eep_minor, fw.etrack_id);
2564 		} else {
2565 		snprintf(adapter->fw_version,
2566 		    sizeof(adapter->fw_version),
2567 		    "%d.%d.%d",
2568 		    fw.eep_major, fw.eep_minor, fw.eep_build);
2569 		}
2570 		break;
2571 	}
2572 }
2573 
2574 /**
2575  * igb_init_mas - init Media Autosense feature if enabled in the NVM
2576  *
2577  * @adapter: adapter struct
2578  **/
2579 static void igb_init_mas(struct igb_adapter *adapter)
2580 {
2581 	struct e1000_hw *hw = &adapter->hw;
2582 	u16 eeprom_data;
2583 
2584 	hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
2585 	switch (hw->bus.func) {
2586 	case E1000_FUNC_0:
2587 		if (eeprom_data & IGB_MAS_ENABLE_0) {
2588 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2589 			netdev_info(adapter->netdev,
2590 				"MAS: Enabling Media Autosense for port %d\n",
2591 				hw->bus.func);
2592 		}
2593 		break;
2594 	case E1000_FUNC_1:
2595 		if (eeprom_data & IGB_MAS_ENABLE_1) {
2596 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2597 			netdev_info(adapter->netdev,
2598 				"MAS: Enabling Media Autosense for port %d\n",
2599 				hw->bus.func);
2600 		}
2601 		break;
2602 	case E1000_FUNC_2:
2603 		if (eeprom_data & IGB_MAS_ENABLE_2) {
2604 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2605 			netdev_info(adapter->netdev,
2606 				"MAS: Enabling Media Autosense for port %d\n",
2607 				hw->bus.func);
2608 		}
2609 		break;
2610 	case E1000_FUNC_3:
2611 		if (eeprom_data & IGB_MAS_ENABLE_3) {
2612 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2613 			netdev_info(adapter->netdev,
2614 				"MAS: Enabling Media Autosense for port %d\n",
2615 				hw->bus.func);
2616 		}
2617 		break;
2618 	default:
2619 		/* Shouldn't get here */
2620 		netdev_err(adapter->netdev,
2621 			"MAS: Invalid port configuration, returning\n");
2622 		break;
2623 	}
2624 }
2625 
2626 /**
2627  *  igb_init_i2c - Init I2C interface
2628  *  @adapter: pointer to adapter structure
2629  **/
2630 static s32 igb_init_i2c(struct igb_adapter *adapter)
2631 {
2632 	s32 status = 0;
2633 
2634 	/* I2C interface supported on i350 devices */
2635 	if (adapter->hw.mac.type != e1000_i350)
2636 		return 0;
2637 
2638 	/* Initialize the i2c bus which is controlled by the registers.
2639 	 * This bus will use the i2c_algo_bit structue that implements
2640 	 * the protocol through toggling of the 4 bits in the register.
2641 	 */
2642 	adapter->i2c_adap.owner = THIS_MODULE;
2643 	adapter->i2c_algo = igb_i2c_algo;
2644 	adapter->i2c_algo.data = adapter;
2645 	adapter->i2c_adap.algo_data = &adapter->i2c_algo;
2646 	adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
2647 	strlcpy(adapter->i2c_adap.name, "igb BB",
2648 		sizeof(adapter->i2c_adap.name));
2649 	status = i2c_bit_add_bus(&adapter->i2c_adap);
2650 	return status;
2651 }
2652 
2653 /**
2654  *  igb_probe - Device Initialization Routine
2655  *  @pdev: PCI device information struct
2656  *  @ent: entry in igb_pci_tbl
2657  *
2658  *  Returns 0 on success, negative on failure
2659  *
2660  *  igb_probe initializes an adapter identified by a pci_dev structure.
2661  *  The OS initialization, configuring of the adapter private structure,
2662  *  and a hardware reset occur.
2663  **/
2664 static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2665 {
2666 	struct net_device *netdev;
2667 	struct igb_adapter *adapter;
2668 	struct e1000_hw *hw;
2669 	u16 eeprom_data = 0;
2670 	s32 ret_val;
2671 	static int global_quad_port_a; /* global quad port a indication */
2672 	const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
2673 	int err, pci_using_dac;
2674 	u8 part_str[E1000_PBANUM_LENGTH];
2675 
2676 	/* Catch broken hardware that put the wrong VF device ID in
2677 	 * the PCIe SR-IOV capability.
2678 	 */
2679 	if (pdev->is_virtfn) {
2680 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
2681 			pci_name(pdev), pdev->vendor, pdev->device);
2682 		return -EINVAL;
2683 	}
2684 
2685 	err = pci_enable_device_mem(pdev);
2686 	if (err)
2687 		return err;
2688 
2689 	pci_using_dac = 0;
2690 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2691 	if (!err) {
2692 		pci_using_dac = 1;
2693 	} else {
2694 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
2695 		if (err) {
2696 			dev_err(&pdev->dev,
2697 				"No usable DMA configuration, aborting\n");
2698 			goto err_dma;
2699 		}
2700 	}
2701 
2702 	err = pci_request_mem_regions(pdev, igb_driver_name);
2703 	if (err)
2704 		goto err_pci_reg;
2705 
2706 	pci_enable_pcie_error_reporting(pdev);
2707 
2708 	pci_set_master(pdev);
2709 	pci_save_state(pdev);
2710 
2711 	err = -ENOMEM;
2712 	netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
2713 				   IGB_MAX_TX_QUEUES);
2714 	if (!netdev)
2715 		goto err_alloc_etherdev;
2716 
2717 	SET_NETDEV_DEV(netdev, &pdev->dev);
2718 
2719 	pci_set_drvdata(pdev, netdev);
2720 	adapter = netdev_priv(netdev);
2721 	adapter->netdev = netdev;
2722 	adapter->pdev = pdev;
2723 	hw = &adapter->hw;
2724 	hw->back = adapter;
2725 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
2726 
2727 	err = -EIO;
2728 	adapter->io_addr = pci_iomap(pdev, 0, 0);
2729 	if (!adapter->io_addr)
2730 		goto err_ioremap;
2731 	/* hw->hw_addr can be altered, we'll use adapter->io_addr for unmap */
2732 	hw->hw_addr = adapter->io_addr;
2733 
2734 	netdev->netdev_ops = &igb_netdev_ops;
2735 	igb_set_ethtool_ops(netdev);
2736 	netdev->watchdog_timeo = 5 * HZ;
2737 
2738 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
2739 
2740 	netdev->mem_start = pci_resource_start(pdev, 0);
2741 	netdev->mem_end = pci_resource_end(pdev, 0);
2742 
2743 	/* PCI config space info */
2744 	hw->vendor_id = pdev->vendor;
2745 	hw->device_id = pdev->device;
2746 	hw->revision_id = pdev->revision;
2747 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
2748 	hw->subsystem_device_id = pdev->subsystem_device;
2749 
2750 	/* Copy the default MAC, PHY and NVM function pointers */
2751 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
2752 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
2753 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
2754 	/* Initialize skew-specific constants */
2755 	err = ei->get_invariants(hw);
2756 	if (err)
2757 		goto err_sw_init;
2758 
2759 	/* setup the private structure */
2760 	err = igb_sw_init(adapter);
2761 	if (err)
2762 		goto err_sw_init;
2763 
2764 	igb_get_bus_info_pcie(hw);
2765 
2766 	hw->phy.autoneg_wait_to_complete = false;
2767 
2768 	/* Copper options */
2769 	if (hw->phy.media_type == e1000_media_type_copper) {
2770 		hw->phy.mdix = AUTO_ALL_MODES;
2771 		hw->phy.disable_polarity_correction = false;
2772 		hw->phy.ms_type = e1000_ms_hw_default;
2773 	}
2774 
2775 	if (igb_check_reset_block(hw))
2776 		dev_info(&pdev->dev,
2777 			"PHY reset is blocked due to SOL/IDER session.\n");
2778 
2779 	/* features is initialized to 0 in allocation, it might have bits
2780 	 * set by igb_sw_init so we should use an or instead of an
2781 	 * assignment.
2782 	 */
2783 	netdev->features |= NETIF_F_SG |
2784 			    NETIF_F_TSO |
2785 			    NETIF_F_TSO6 |
2786 			    NETIF_F_RXHASH |
2787 			    NETIF_F_RXCSUM |
2788 			    NETIF_F_HW_CSUM;
2789 
2790 	if (hw->mac.type >= e1000_82576)
2791 		netdev->features |= NETIF_F_SCTP_CRC;
2792 
2793 #define IGB_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
2794 				  NETIF_F_GSO_GRE_CSUM | \
2795 				  NETIF_F_GSO_IPXIP4 | \
2796 				  NETIF_F_GSO_IPXIP6 | \
2797 				  NETIF_F_GSO_UDP_TUNNEL | \
2798 				  NETIF_F_GSO_UDP_TUNNEL_CSUM)
2799 
2800 	netdev->gso_partial_features = IGB_GSO_PARTIAL_FEATURES;
2801 	netdev->features |= NETIF_F_GSO_PARTIAL | IGB_GSO_PARTIAL_FEATURES;
2802 
2803 	/* copy netdev features into list of user selectable features */
2804 	netdev->hw_features |= netdev->features |
2805 			       NETIF_F_HW_VLAN_CTAG_RX |
2806 			       NETIF_F_HW_VLAN_CTAG_TX |
2807 			       NETIF_F_RXALL;
2808 
2809 	if (hw->mac.type >= e1000_i350)
2810 		netdev->hw_features |= NETIF_F_NTUPLE;
2811 
2812 	if (pci_using_dac)
2813 		netdev->features |= NETIF_F_HIGHDMA;
2814 
2815 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
2816 	netdev->mpls_features |= NETIF_F_HW_CSUM;
2817 	netdev->hw_enc_features |= netdev->vlan_features;
2818 
2819 	/* set this bit last since it cannot be part of vlan_features */
2820 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
2821 			    NETIF_F_HW_VLAN_CTAG_RX |
2822 			    NETIF_F_HW_VLAN_CTAG_TX;
2823 
2824 	netdev->priv_flags |= IFF_SUPP_NOFCS;
2825 
2826 	netdev->priv_flags |= IFF_UNICAST_FLT;
2827 
2828 	/* MTU range: 68 - 9216 */
2829 	netdev->min_mtu = ETH_MIN_MTU;
2830 	netdev->max_mtu = MAX_STD_JUMBO_FRAME_SIZE;
2831 
2832 	adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
2833 
2834 	/* before reading the NVM, reset the controller to put the device in a
2835 	 * known good starting state
2836 	 */
2837 	hw->mac.ops.reset_hw(hw);
2838 
2839 	/* make sure the NVM is good , i211/i210 parts can have special NVM
2840 	 * that doesn't contain a checksum
2841 	 */
2842 	switch (hw->mac.type) {
2843 	case e1000_i210:
2844 	case e1000_i211:
2845 		if (igb_get_flash_presence_i210(hw)) {
2846 			if (hw->nvm.ops.validate(hw) < 0) {
2847 				dev_err(&pdev->dev,
2848 					"The NVM Checksum Is Not Valid\n");
2849 				err = -EIO;
2850 				goto err_eeprom;
2851 			}
2852 		}
2853 		break;
2854 	default:
2855 		if (hw->nvm.ops.validate(hw) < 0) {
2856 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
2857 			err = -EIO;
2858 			goto err_eeprom;
2859 		}
2860 		break;
2861 	}
2862 
2863 	if (eth_platform_get_mac_address(&pdev->dev, hw->mac.addr)) {
2864 		/* copy the MAC address out of the NVM */
2865 		if (hw->mac.ops.read_mac_addr(hw))
2866 			dev_err(&pdev->dev, "NVM Read Error\n");
2867 	}
2868 
2869 	memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
2870 
2871 	if (!is_valid_ether_addr(netdev->dev_addr)) {
2872 		dev_err(&pdev->dev, "Invalid MAC Address\n");
2873 		err = -EIO;
2874 		goto err_eeprom;
2875 	}
2876 
2877 	igb_set_default_mac_filter(adapter);
2878 
2879 	/* get firmware version for ethtool -i */
2880 	igb_set_fw_version(adapter);
2881 
2882 	/* configure RXPBSIZE and TXPBSIZE */
2883 	if (hw->mac.type == e1000_i210) {
2884 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
2885 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
2886 	}
2887 
2888 	timer_setup(&adapter->watchdog_timer, igb_watchdog, 0);
2889 	timer_setup(&adapter->phy_info_timer, igb_update_phy_info, 0);
2890 
2891 	INIT_WORK(&adapter->reset_task, igb_reset_task);
2892 	INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
2893 
2894 	/* Initialize link properties that are user-changeable */
2895 	adapter->fc_autoneg = true;
2896 	hw->mac.autoneg = true;
2897 	hw->phy.autoneg_advertised = 0x2f;
2898 
2899 	hw->fc.requested_mode = e1000_fc_default;
2900 	hw->fc.current_mode = e1000_fc_default;
2901 
2902 	igb_validate_mdi_setting(hw);
2903 
2904 	/* By default, support wake on port A */
2905 	if (hw->bus.func == 0)
2906 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2907 
2908 	/* Check the NVM for wake support on non-port A ports */
2909 	if (hw->mac.type >= e1000_82580)
2910 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
2911 				 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
2912 				 &eeprom_data);
2913 	else if (hw->bus.func == 1)
2914 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
2915 
2916 	if (eeprom_data & IGB_EEPROM_APME)
2917 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2918 
2919 	/* now that we have the eeprom settings, apply the special cases where
2920 	 * the eeprom may be wrong or the board simply won't support wake on
2921 	 * lan on a particular port
2922 	 */
2923 	switch (pdev->device) {
2924 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
2925 		adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2926 		break;
2927 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
2928 	case E1000_DEV_ID_82576_FIBER:
2929 	case E1000_DEV_ID_82576_SERDES:
2930 		/* Wake events only supported on port A for dual fiber
2931 		 * regardless of eeprom setting
2932 		 */
2933 		if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
2934 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2935 		break;
2936 	case E1000_DEV_ID_82576_QUAD_COPPER:
2937 	case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
2938 		/* if quad port adapter, disable WoL on all but port A */
2939 		if (global_quad_port_a != 0)
2940 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2941 		else
2942 			adapter->flags |= IGB_FLAG_QUAD_PORT_A;
2943 		/* Reset for multiple quad port adapters */
2944 		if (++global_quad_port_a == 4)
2945 			global_quad_port_a = 0;
2946 		break;
2947 	default:
2948 		/* If the device can't wake, don't set software support */
2949 		if (!device_can_wakeup(&adapter->pdev->dev))
2950 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2951 	}
2952 
2953 	/* initialize the wol settings based on the eeprom settings */
2954 	if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
2955 		adapter->wol |= E1000_WUFC_MAG;
2956 
2957 	/* Some vendors want WoL disabled by default, but still supported */
2958 	if ((hw->mac.type == e1000_i350) &&
2959 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
2960 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2961 		adapter->wol = 0;
2962 	}
2963 
2964 	/* Some vendors want the ability to Use the EEPROM setting as
2965 	 * enable/disable only, and not for capability
2966 	 */
2967 	if (((hw->mac.type == e1000_i350) ||
2968 	     (hw->mac.type == e1000_i354)) &&
2969 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_DELL)) {
2970 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2971 		adapter->wol = 0;
2972 	}
2973 	if (hw->mac.type == e1000_i350) {
2974 		if (((pdev->subsystem_device == 0x5001) ||
2975 		     (pdev->subsystem_device == 0x5002)) &&
2976 				(hw->bus.func == 0)) {
2977 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2978 			adapter->wol = 0;
2979 		}
2980 		if (pdev->subsystem_device == 0x1F52)
2981 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2982 	}
2983 
2984 	device_set_wakeup_enable(&adapter->pdev->dev,
2985 				 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
2986 
2987 	/* reset the hardware with the new settings */
2988 	igb_reset(adapter);
2989 
2990 	/* Init the I2C interface */
2991 	err = igb_init_i2c(adapter);
2992 	if (err) {
2993 		dev_err(&pdev->dev, "failed to init i2c interface\n");
2994 		goto err_eeprom;
2995 	}
2996 
2997 	/* let the f/w know that the h/w is now under the control of the
2998 	 * driver.
2999 	 */
3000 	igb_get_hw_control(adapter);
3001 
3002 	strcpy(netdev->name, "eth%d");
3003 	err = register_netdev(netdev);
3004 	if (err)
3005 		goto err_register;
3006 
3007 	/* carrier off reporting is important to ethtool even BEFORE open */
3008 	netif_carrier_off(netdev);
3009 
3010 #ifdef CONFIG_IGB_DCA
3011 	if (dca_add_requester(&pdev->dev) == 0) {
3012 		adapter->flags |= IGB_FLAG_DCA_ENABLED;
3013 		dev_info(&pdev->dev, "DCA enabled\n");
3014 		igb_setup_dca(adapter);
3015 	}
3016 
3017 #endif
3018 #ifdef CONFIG_IGB_HWMON
3019 	/* Initialize the thermal sensor on i350 devices. */
3020 	if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
3021 		u16 ets_word;
3022 
3023 		/* Read the NVM to determine if this i350 device supports an
3024 		 * external thermal sensor.
3025 		 */
3026 		hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
3027 		if (ets_word != 0x0000 && ets_word != 0xFFFF)
3028 			adapter->ets = true;
3029 		else
3030 			adapter->ets = false;
3031 		if (igb_sysfs_init(adapter))
3032 			dev_err(&pdev->dev,
3033 				"failed to allocate sysfs resources\n");
3034 	} else {
3035 		adapter->ets = false;
3036 	}
3037 #endif
3038 	/* Check if Media Autosense is enabled */
3039 	adapter->ei = *ei;
3040 	if (hw->dev_spec._82575.mas_capable)
3041 		igb_init_mas(adapter);
3042 
3043 	/* do hw tstamp init after resetting */
3044 	igb_ptp_init(adapter);
3045 
3046 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
3047 	/* print bus type/speed/width info, not applicable to i354 */
3048 	if (hw->mac.type != e1000_i354) {
3049 		dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
3050 			 netdev->name,
3051 			 ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
3052 			  (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
3053 			   "unknown"),
3054 			 ((hw->bus.width == e1000_bus_width_pcie_x4) ?
3055 			  "Width x4" :
3056 			  (hw->bus.width == e1000_bus_width_pcie_x2) ?
3057 			  "Width x2" :
3058 			  (hw->bus.width == e1000_bus_width_pcie_x1) ?
3059 			  "Width x1" : "unknown"), netdev->dev_addr);
3060 	}
3061 
3062 	if ((hw->mac.type >= e1000_i210 ||
3063 	     igb_get_flash_presence_i210(hw))) {
3064 		ret_val = igb_read_part_string(hw, part_str,
3065 					       E1000_PBANUM_LENGTH);
3066 	} else {
3067 		ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
3068 	}
3069 
3070 	if (ret_val)
3071 		strcpy(part_str, "Unknown");
3072 	dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
3073 	dev_info(&pdev->dev,
3074 		"Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
3075 		(adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
3076 		(adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
3077 		adapter->num_rx_queues, adapter->num_tx_queues);
3078 	if (hw->phy.media_type == e1000_media_type_copper) {
3079 		switch (hw->mac.type) {
3080 		case e1000_i350:
3081 		case e1000_i210:
3082 		case e1000_i211:
3083 			/* Enable EEE for internal copper PHY devices */
3084 			err = igb_set_eee_i350(hw, true, true);
3085 			if ((!err) &&
3086 			    (!hw->dev_spec._82575.eee_disable)) {
3087 				adapter->eee_advert =
3088 					MDIO_EEE_100TX | MDIO_EEE_1000T;
3089 				adapter->flags |= IGB_FLAG_EEE;
3090 			}
3091 			break;
3092 		case e1000_i354:
3093 			if ((rd32(E1000_CTRL_EXT) &
3094 			    E1000_CTRL_EXT_LINK_MODE_SGMII)) {
3095 				err = igb_set_eee_i354(hw, true, true);
3096 				if ((!err) &&
3097 					(!hw->dev_spec._82575.eee_disable)) {
3098 					adapter->eee_advert =
3099 					   MDIO_EEE_100TX | MDIO_EEE_1000T;
3100 					adapter->flags |= IGB_FLAG_EEE;
3101 				}
3102 			}
3103 			break;
3104 		default:
3105 			break;
3106 		}
3107 	}
3108 	pm_runtime_put_noidle(&pdev->dev);
3109 	return 0;
3110 
3111 err_register:
3112 	igb_release_hw_control(adapter);
3113 	memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
3114 err_eeprom:
3115 	if (!igb_check_reset_block(hw))
3116 		igb_reset_phy(hw);
3117 
3118 	if (hw->flash_address)
3119 		iounmap(hw->flash_address);
3120 err_sw_init:
3121 	kfree(adapter->mac_table);
3122 	kfree(adapter->shadow_vfta);
3123 	igb_clear_interrupt_scheme(adapter);
3124 #ifdef CONFIG_PCI_IOV
3125 	igb_disable_sriov(pdev);
3126 #endif
3127 	pci_iounmap(pdev, adapter->io_addr);
3128 err_ioremap:
3129 	free_netdev(netdev);
3130 err_alloc_etherdev:
3131 	pci_release_mem_regions(pdev);
3132 err_pci_reg:
3133 err_dma:
3134 	pci_disable_device(pdev);
3135 	return err;
3136 }
3137 
3138 #ifdef CONFIG_PCI_IOV
3139 static int igb_disable_sriov(struct pci_dev *pdev)
3140 {
3141 	struct net_device *netdev = pci_get_drvdata(pdev);
3142 	struct igb_adapter *adapter = netdev_priv(netdev);
3143 	struct e1000_hw *hw = &adapter->hw;
3144 
3145 	/* reclaim resources allocated to VFs */
3146 	if (adapter->vf_data) {
3147 		/* disable iov and allow time for transactions to clear */
3148 		if (pci_vfs_assigned(pdev)) {
3149 			dev_warn(&pdev->dev,
3150 				 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
3151 			return -EPERM;
3152 		} else {
3153 			pci_disable_sriov(pdev);
3154 			msleep(500);
3155 		}
3156 
3157 		kfree(adapter->vf_mac_list);
3158 		adapter->vf_mac_list = NULL;
3159 		kfree(adapter->vf_data);
3160 		adapter->vf_data = NULL;
3161 		adapter->vfs_allocated_count = 0;
3162 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
3163 		wrfl();
3164 		msleep(100);
3165 		dev_info(&pdev->dev, "IOV Disabled\n");
3166 
3167 		/* Re-enable DMA Coalescing flag since IOV is turned off */
3168 		adapter->flags |= IGB_FLAG_DMAC;
3169 	}
3170 
3171 	return 0;
3172 }
3173 
3174 static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
3175 {
3176 	struct net_device *netdev = pci_get_drvdata(pdev);
3177 	struct igb_adapter *adapter = netdev_priv(netdev);
3178 	int old_vfs = pci_num_vf(pdev);
3179 	struct vf_mac_filter *mac_list;
3180 	int err = 0;
3181 	int num_vf_mac_filters, i;
3182 
3183 	if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
3184 		err = -EPERM;
3185 		goto out;
3186 	}
3187 	if (!num_vfs)
3188 		goto out;
3189 
3190 	if (old_vfs) {
3191 		dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
3192 			 old_vfs, max_vfs);
3193 		adapter->vfs_allocated_count = old_vfs;
3194 	} else
3195 		adapter->vfs_allocated_count = num_vfs;
3196 
3197 	adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
3198 				sizeof(struct vf_data_storage), GFP_KERNEL);
3199 
3200 	/* if allocation failed then we do not support SR-IOV */
3201 	if (!adapter->vf_data) {
3202 		adapter->vfs_allocated_count = 0;
3203 		dev_err(&pdev->dev,
3204 			"Unable to allocate memory for VF Data Storage\n");
3205 		err = -ENOMEM;
3206 		goto out;
3207 	}
3208 
3209 	/* Due to the limited number of RAR entries calculate potential
3210 	 * number of MAC filters available for the VFs. Reserve entries
3211 	 * for PF default MAC, PF MAC filters and at least one RAR entry
3212 	 * for each VF for VF MAC.
3213 	 */
3214 	num_vf_mac_filters = adapter->hw.mac.rar_entry_count -
3215 			     (1 + IGB_PF_MAC_FILTERS_RESERVED +
3216 			      adapter->vfs_allocated_count);
3217 
3218 	adapter->vf_mac_list = kcalloc(num_vf_mac_filters,
3219 				       sizeof(struct vf_mac_filter),
3220 				       GFP_KERNEL);
3221 
3222 	mac_list = adapter->vf_mac_list;
3223 	INIT_LIST_HEAD(&adapter->vf_macs.l);
3224 
3225 	if (adapter->vf_mac_list) {
3226 		/* Initialize list of VF MAC filters */
3227 		for (i = 0; i < num_vf_mac_filters; i++) {
3228 			mac_list->vf = -1;
3229 			mac_list->free = true;
3230 			list_add(&mac_list->l, &adapter->vf_macs.l);
3231 			mac_list++;
3232 		}
3233 	} else {
3234 		/* If we could not allocate memory for the VF MAC filters
3235 		 * we can continue without this feature but warn user.
3236 		 */
3237 		dev_err(&pdev->dev,
3238 			"Unable to allocate memory for VF MAC filter list\n");
3239 	}
3240 
3241 	/* only call pci_enable_sriov() if no VFs are allocated already */
3242 	if (!old_vfs) {
3243 		err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
3244 		if (err)
3245 			goto err_out;
3246 	}
3247 	dev_info(&pdev->dev, "%d VFs allocated\n",
3248 		 adapter->vfs_allocated_count);
3249 	for (i = 0; i < adapter->vfs_allocated_count; i++)
3250 		igb_vf_configure(adapter, i);
3251 
3252 	/* DMA Coalescing is not supported in IOV mode. */
3253 	adapter->flags &= ~IGB_FLAG_DMAC;
3254 	goto out;
3255 
3256 err_out:
3257 	kfree(adapter->vf_mac_list);
3258 	adapter->vf_mac_list = NULL;
3259 	kfree(adapter->vf_data);
3260 	adapter->vf_data = NULL;
3261 	adapter->vfs_allocated_count = 0;
3262 out:
3263 	return err;
3264 }
3265 
3266 #endif
3267 /**
3268  *  igb_remove_i2c - Cleanup  I2C interface
3269  *  @adapter: pointer to adapter structure
3270  **/
3271 static void igb_remove_i2c(struct igb_adapter *adapter)
3272 {
3273 	/* free the adapter bus structure */
3274 	i2c_del_adapter(&adapter->i2c_adap);
3275 }
3276 
3277 /**
3278  *  igb_remove - Device Removal Routine
3279  *  @pdev: PCI device information struct
3280  *
3281  *  igb_remove is called by the PCI subsystem to alert the driver
3282  *  that it should release a PCI device.  The could be caused by a
3283  *  Hot-Plug event, or because the driver is going to be removed from
3284  *  memory.
3285  **/
3286 static void igb_remove(struct pci_dev *pdev)
3287 {
3288 	struct net_device *netdev = pci_get_drvdata(pdev);
3289 	struct igb_adapter *adapter = netdev_priv(netdev);
3290 	struct e1000_hw *hw = &adapter->hw;
3291 
3292 	pm_runtime_get_noresume(&pdev->dev);
3293 #ifdef CONFIG_IGB_HWMON
3294 	igb_sysfs_exit(adapter);
3295 #endif
3296 	igb_remove_i2c(adapter);
3297 	igb_ptp_stop(adapter);
3298 	/* The watchdog timer may be rescheduled, so explicitly
3299 	 * disable watchdog from being rescheduled.
3300 	 */
3301 	set_bit(__IGB_DOWN, &adapter->state);
3302 	del_timer_sync(&adapter->watchdog_timer);
3303 	del_timer_sync(&adapter->phy_info_timer);
3304 
3305 	cancel_work_sync(&adapter->reset_task);
3306 	cancel_work_sync(&adapter->watchdog_task);
3307 
3308 #ifdef CONFIG_IGB_DCA
3309 	if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
3310 		dev_info(&pdev->dev, "DCA disabled\n");
3311 		dca_remove_requester(&pdev->dev);
3312 		adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
3313 		wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
3314 	}
3315 #endif
3316 
3317 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
3318 	 * would have already happened in close and is redundant.
3319 	 */
3320 	igb_release_hw_control(adapter);
3321 
3322 #ifdef CONFIG_PCI_IOV
3323 	igb_disable_sriov(pdev);
3324 #endif
3325 
3326 	unregister_netdev(netdev);
3327 
3328 	igb_clear_interrupt_scheme(adapter);
3329 
3330 	pci_iounmap(pdev, adapter->io_addr);
3331 	if (hw->flash_address)
3332 		iounmap(hw->flash_address);
3333 	pci_release_mem_regions(pdev);
3334 
3335 	kfree(adapter->mac_table);
3336 	kfree(adapter->shadow_vfta);
3337 	free_netdev(netdev);
3338 
3339 	pci_disable_pcie_error_reporting(pdev);
3340 
3341 	pci_disable_device(pdev);
3342 }
3343 
3344 /**
3345  *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
3346  *  @adapter: board private structure to initialize
3347  *
3348  *  This function initializes the vf specific data storage and then attempts to
3349  *  allocate the VFs.  The reason for ordering it this way is because it is much
3350  *  mor expensive time wise to disable SR-IOV than it is to allocate and free
3351  *  the memory for the VFs.
3352  **/
3353 static void igb_probe_vfs(struct igb_adapter *adapter)
3354 {
3355 #ifdef CONFIG_PCI_IOV
3356 	struct pci_dev *pdev = adapter->pdev;
3357 	struct e1000_hw *hw = &adapter->hw;
3358 
3359 	/* Virtualization features not supported on i210 family. */
3360 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
3361 		return;
3362 
3363 	/* Of the below we really only want the effect of getting
3364 	 * IGB_FLAG_HAS_MSIX set (if available), without which
3365 	 * igb_enable_sriov() has no effect.
3366 	 */
3367 	igb_set_interrupt_capability(adapter, true);
3368 	igb_reset_interrupt_capability(adapter);
3369 
3370 	pci_sriov_set_totalvfs(pdev, 7);
3371 	igb_enable_sriov(pdev, max_vfs);
3372 
3373 #endif /* CONFIG_PCI_IOV */
3374 }
3375 
3376 static void igb_init_queue_configuration(struct igb_adapter *adapter)
3377 {
3378 	struct e1000_hw *hw = &adapter->hw;
3379 	u32 max_rss_queues;
3380 
3381 	/* Determine the maximum number of RSS queues supported. */
3382 	switch (hw->mac.type) {
3383 	case e1000_i211:
3384 		max_rss_queues = IGB_MAX_RX_QUEUES_I211;
3385 		break;
3386 	case e1000_82575:
3387 	case e1000_i210:
3388 		max_rss_queues = IGB_MAX_RX_QUEUES_82575;
3389 		break;
3390 	case e1000_i350:
3391 		/* I350 cannot do RSS and SR-IOV at the same time */
3392 		if (!!adapter->vfs_allocated_count) {
3393 			max_rss_queues = 1;
3394 			break;
3395 		}
3396 		/* fall through */
3397 	case e1000_82576:
3398 		if (!!adapter->vfs_allocated_count) {
3399 			max_rss_queues = 2;
3400 			break;
3401 		}
3402 		/* fall through */
3403 	case e1000_82580:
3404 	case e1000_i354:
3405 	default:
3406 		max_rss_queues = IGB_MAX_RX_QUEUES;
3407 		break;
3408 	}
3409 
3410 	adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
3411 
3412 	igb_set_flag_queue_pairs(adapter, max_rss_queues);
3413 }
3414 
3415 void igb_set_flag_queue_pairs(struct igb_adapter *adapter,
3416 			      const u32 max_rss_queues)
3417 {
3418 	struct e1000_hw *hw = &adapter->hw;
3419 
3420 	/* Determine if we need to pair queues. */
3421 	switch (hw->mac.type) {
3422 	case e1000_82575:
3423 	case e1000_i211:
3424 		/* Device supports enough interrupts without queue pairing. */
3425 		break;
3426 	case e1000_82576:
3427 	case e1000_82580:
3428 	case e1000_i350:
3429 	case e1000_i354:
3430 	case e1000_i210:
3431 	default:
3432 		/* If rss_queues > half of max_rss_queues, pair the queues in
3433 		 * order to conserve interrupts due to limited supply.
3434 		 */
3435 		if (adapter->rss_queues > (max_rss_queues / 2))
3436 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
3437 		else
3438 			adapter->flags &= ~IGB_FLAG_QUEUE_PAIRS;
3439 		break;
3440 	}
3441 }
3442 
3443 /**
3444  *  igb_sw_init - Initialize general software structures (struct igb_adapter)
3445  *  @adapter: board private structure to initialize
3446  *
3447  *  igb_sw_init initializes the Adapter private data structure.
3448  *  Fields are initialized based on PCI device information and
3449  *  OS network device settings (MTU size).
3450  **/
3451 static int igb_sw_init(struct igb_adapter *adapter)
3452 {
3453 	struct e1000_hw *hw = &adapter->hw;
3454 	struct net_device *netdev = adapter->netdev;
3455 	struct pci_dev *pdev = adapter->pdev;
3456 
3457 	pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
3458 
3459 	/* set default ring sizes */
3460 	adapter->tx_ring_count = IGB_DEFAULT_TXD;
3461 	adapter->rx_ring_count = IGB_DEFAULT_RXD;
3462 
3463 	/* set default ITR values */
3464 	adapter->rx_itr_setting = IGB_DEFAULT_ITR;
3465 	adapter->tx_itr_setting = IGB_DEFAULT_ITR;
3466 
3467 	/* set default work limits */
3468 	adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
3469 
3470 	adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN +
3471 				  VLAN_HLEN;
3472 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
3473 
3474 	spin_lock_init(&adapter->nfc_lock);
3475 	spin_lock_init(&adapter->stats64_lock);
3476 #ifdef CONFIG_PCI_IOV
3477 	switch (hw->mac.type) {
3478 	case e1000_82576:
3479 	case e1000_i350:
3480 		if (max_vfs > 7) {
3481 			dev_warn(&pdev->dev,
3482 				 "Maximum of 7 VFs per PF, using max\n");
3483 			max_vfs = adapter->vfs_allocated_count = 7;
3484 		} else
3485 			adapter->vfs_allocated_count = max_vfs;
3486 		if (adapter->vfs_allocated_count)
3487 			dev_warn(&pdev->dev,
3488 				 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
3489 		break;
3490 	default:
3491 		break;
3492 	}
3493 #endif /* CONFIG_PCI_IOV */
3494 
3495 	/* Assume MSI-X interrupts, will be checked during IRQ allocation */
3496 	adapter->flags |= IGB_FLAG_HAS_MSIX;
3497 
3498 	adapter->mac_table = kzalloc(sizeof(struct igb_mac_addr) *
3499 				     hw->mac.rar_entry_count, GFP_ATOMIC);
3500 	if (!adapter->mac_table)
3501 		return -ENOMEM;
3502 
3503 	igb_probe_vfs(adapter);
3504 
3505 	igb_init_queue_configuration(adapter);
3506 
3507 	/* Setup and initialize a copy of the hw vlan table array */
3508 	adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
3509 				       GFP_ATOMIC);
3510 	if (!adapter->shadow_vfta)
3511 		return -ENOMEM;
3512 
3513 	/* This call may decrease the number of queues */
3514 	if (igb_init_interrupt_scheme(adapter, true)) {
3515 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
3516 		return -ENOMEM;
3517 	}
3518 
3519 	/* Explicitly disable IRQ since the NIC can be in any state. */
3520 	igb_irq_disable(adapter);
3521 
3522 	if (hw->mac.type >= e1000_i350)
3523 		adapter->flags &= ~IGB_FLAG_DMAC;
3524 
3525 	set_bit(__IGB_DOWN, &adapter->state);
3526 	return 0;
3527 }
3528 
3529 /**
3530  *  igb_open - Called when a network interface is made active
3531  *  @netdev: network interface device structure
3532  *
3533  *  Returns 0 on success, negative value on failure
3534  *
3535  *  The open entry point is called when a network interface is made
3536  *  active by the system (IFF_UP).  At this point all resources needed
3537  *  for transmit and receive operations are allocated, the interrupt
3538  *  handler is registered with the OS, the watchdog timer is started,
3539  *  and the stack is notified that the interface is ready.
3540  **/
3541 static int __igb_open(struct net_device *netdev, bool resuming)
3542 {
3543 	struct igb_adapter *adapter = netdev_priv(netdev);
3544 	struct e1000_hw *hw = &adapter->hw;
3545 	struct pci_dev *pdev = adapter->pdev;
3546 	int err;
3547 	int i;
3548 
3549 	/* disallow open during test */
3550 	if (test_bit(__IGB_TESTING, &adapter->state)) {
3551 		WARN_ON(resuming);
3552 		return -EBUSY;
3553 	}
3554 
3555 	if (!resuming)
3556 		pm_runtime_get_sync(&pdev->dev);
3557 
3558 	netif_carrier_off(netdev);
3559 
3560 	/* allocate transmit descriptors */
3561 	err = igb_setup_all_tx_resources(adapter);
3562 	if (err)
3563 		goto err_setup_tx;
3564 
3565 	/* allocate receive descriptors */
3566 	err = igb_setup_all_rx_resources(adapter);
3567 	if (err)
3568 		goto err_setup_rx;
3569 
3570 	igb_power_up_link(adapter);
3571 
3572 	/* before we allocate an interrupt, we must be ready to handle it.
3573 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3574 	 * as soon as we call pci_request_irq, so we have to setup our
3575 	 * clean_rx handler before we do so.
3576 	 */
3577 	igb_configure(adapter);
3578 
3579 	err = igb_request_irq(adapter);
3580 	if (err)
3581 		goto err_req_irq;
3582 
3583 	/* Notify the stack of the actual queue counts. */
3584 	err = netif_set_real_num_tx_queues(adapter->netdev,
3585 					   adapter->num_tx_queues);
3586 	if (err)
3587 		goto err_set_queues;
3588 
3589 	err = netif_set_real_num_rx_queues(adapter->netdev,
3590 					   adapter->num_rx_queues);
3591 	if (err)
3592 		goto err_set_queues;
3593 
3594 	/* From here on the code is the same as igb_up() */
3595 	clear_bit(__IGB_DOWN, &adapter->state);
3596 
3597 	for (i = 0; i < adapter->num_q_vectors; i++)
3598 		napi_enable(&(adapter->q_vector[i]->napi));
3599 
3600 	/* Clear any pending interrupts. */
3601 	rd32(E1000_ICR);
3602 
3603 	igb_irq_enable(adapter);
3604 
3605 	/* notify VFs that reset has been completed */
3606 	if (adapter->vfs_allocated_count) {
3607 		u32 reg_data = rd32(E1000_CTRL_EXT);
3608 
3609 		reg_data |= E1000_CTRL_EXT_PFRSTD;
3610 		wr32(E1000_CTRL_EXT, reg_data);
3611 	}
3612 
3613 	netif_tx_start_all_queues(netdev);
3614 
3615 	if (!resuming)
3616 		pm_runtime_put(&pdev->dev);
3617 
3618 	/* start the watchdog. */
3619 	hw->mac.get_link_status = 1;
3620 	schedule_work(&adapter->watchdog_task);
3621 
3622 	return 0;
3623 
3624 err_set_queues:
3625 	igb_free_irq(adapter);
3626 err_req_irq:
3627 	igb_release_hw_control(adapter);
3628 	igb_power_down_link(adapter);
3629 	igb_free_all_rx_resources(adapter);
3630 err_setup_rx:
3631 	igb_free_all_tx_resources(adapter);
3632 err_setup_tx:
3633 	igb_reset(adapter);
3634 	if (!resuming)
3635 		pm_runtime_put(&pdev->dev);
3636 
3637 	return err;
3638 }
3639 
3640 int igb_open(struct net_device *netdev)
3641 {
3642 	return __igb_open(netdev, false);
3643 }
3644 
3645 /**
3646  *  igb_close - Disables a network interface
3647  *  @netdev: network interface device structure
3648  *
3649  *  Returns 0, this is not allowed to fail
3650  *
3651  *  The close entry point is called when an interface is de-activated
3652  *  by the OS.  The hardware is still under the driver's control, but
3653  *  needs to be disabled.  A global MAC reset is issued to stop the
3654  *  hardware, and all transmit and receive resources are freed.
3655  **/
3656 static int __igb_close(struct net_device *netdev, bool suspending)
3657 {
3658 	struct igb_adapter *adapter = netdev_priv(netdev);
3659 	struct pci_dev *pdev = adapter->pdev;
3660 
3661 	WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
3662 
3663 	if (!suspending)
3664 		pm_runtime_get_sync(&pdev->dev);
3665 
3666 	igb_down(adapter);
3667 	igb_free_irq(adapter);
3668 
3669 	igb_free_all_tx_resources(adapter);
3670 	igb_free_all_rx_resources(adapter);
3671 
3672 	if (!suspending)
3673 		pm_runtime_put_sync(&pdev->dev);
3674 	return 0;
3675 }
3676 
3677 int igb_close(struct net_device *netdev)
3678 {
3679 	if (netif_device_present(netdev))
3680 		return __igb_close(netdev, false);
3681 	return 0;
3682 }
3683 
3684 /**
3685  *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
3686  *  @tx_ring: tx descriptor ring (for a specific queue) to setup
3687  *
3688  *  Return 0 on success, negative on failure
3689  **/
3690 int igb_setup_tx_resources(struct igb_ring *tx_ring)
3691 {
3692 	struct device *dev = tx_ring->dev;
3693 	int size;
3694 
3695 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3696 
3697 	tx_ring->tx_buffer_info = vmalloc(size);
3698 	if (!tx_ring->tx_buffer_info)
3699 		goto err;
3700 
3701 	/* round up to nearest 4K */
3702 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
3703 	tx_ring->size = ALIGN(tx_ring->size, 4096);
3704 
3705 	tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
3706 					   &tx_ring->dma, GFP_KERNEL);
3707 	if (!tx_ring->desc)
3708 		goto err;
3709 
3710 	tx_ring->next_to_use = 0;
3711 	tx_ring->next_to_clean = 0;
3712 
3713 	return 0;
3714 
3715 err:
3716 	vfree(tx_ring->tx_buffer_info);
3717 	tx_ring->tx_buffer_info = NULL;
3718 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
3719 	return -ENOMEM;
3720 }
3721 
3722 /**
3723  *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
3724  *				 (Descriptors) for all queues
3725  *  @adapter: board private structure
3726  *
3727  *  Return 0 on success, negative on failure
3728  **/
3729 static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
3730 {
3731 	struct pci_dev *pdev = adapter->pdev;
3732 	int i, err = 0;
3733 
3734 	for (i = 0; i < adapter->num_tx_queues; i++) {
3735 		err = igb_setup_tx_resources(adapter->tx_ring[i]);
3736 		if (err) {
3737 			dev_err(&pdev->dev,
3738 				"Allocation for Tx Queue %u failed\n", i);
3739 			for (i--; i >= 0; i--)
3740 				igb_free_tx_resources(adapter->tx_ring[i]);
3741 			break;
3742 		}
3743 	}
3744 
3745 	return err;
3746 }
3747 
3748 /**
3749  *  igb_setup_tctl - configure the transmit control registers
3750  *  @adapter: Board private structure
3751  **/
3752 void igb_setup_tctl(struct igb_adapter *adapter)
3753 {
3754 	struct e1000_hw *hw = &adapter->hw;
3755 	u32 tctl;
3756 
3757 	/* disable queue 0 which is enabled by default on 82575 and 82576 */
3758 	wr32(E1000_TXDCTL(0), 0);
3759 
3760 	/* Program the Transmit Control Register */
3761 	tctl = rd32(E1000_TCTL);
3762 	tctl &= ~E1000_TCTL_CT;
3763 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
3764 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
3765 
3766 	igb_config_collision_dist(hw);
3767 
3768 	/* Enable transmits */
3769 	tctl |= E1000_TCTL_EN;
3770 
3771 	wr32(E1000_TCTL, tctl);
3772 }
3773 
3774 /**
3775  *  igb_configure_tx_ring - Configure transmit ring after Reset
3776  *  @adapter: board private structure
3777  *  @ring: tx ring to configure
3778  *
3779  *  Configure a transmit ring after a reset.
3780  **/
3781 void igb_configure_tx_ring(struct igb_adapter *adapter,
3782 			   struct igb_ring *ring)
3783 {
3784 	struct e1000_hw *hw = &adapter->hw;
3785 	u32 txdctl = 0;
3786 	u64 tdba = ring->dma;
3787 	int reg_idx = ring->reg_idx;
3788 
3789 	/* disable the queue */
3790 	wr32(E1000_TXDCTL(reg_idx), 0);
3791 	wrfl();
3792 	mdelay(10);
3793 
3794 	wr32(E1000_TDLEN(reg_idx),
3795 	     ring->count * sizeof(union e1000_adv_tx_desc));
3796 	wr32(E1000_TDBAL(reg_idx),
3797 	     tdba & 0x00000000ffffffffULL);
3798 	wr32(E1000_TDBAH(reg_idx), tdba >> 32);
3799 
3800 	ring->tail = adapter->io_addr + E1000_TDT(reg_idx);
3801 	wr32(E1000_TDH(reg_idx), 0);
3802 	writel(0, ring->tail);
3803 
3804 	txdctl |= IGB_TX_PTHRESH;
3805 	txdctl |= IGB_TX_HTHRESH << 8;
3806 	txdctl |= IGB_TX_WTHRESH << 16;
3807 
3808 	/* reinitialize tx_buffer_info */
3809 	memset(ring->tx_buffer_info, 0,
3810 	       sizeof(struct igb_tx_buffer) * ring->count);
3811 
3812 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
3813 	wr32(E1000_TXDCTL(reg_idx), txdctl);
3814 }
3815 
3816 /**
3817  *  igb_configure_tx - Configure transmit Unit after Reset
3818  *  @adapter: board private structure
3819  *
3820  *  Configure the Tx unit of the MAC after a reset.
3821  **/
3822 static void igb_configure_tx(struct igb_adapter *adapter)
3823 {
3824 	int i;
3825 
3826 	for (i = 0; i < adapter->num_tx_queues; i++)
3827 		igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
3828 }
3829 
3830 /**
3831  *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
3832  *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
3833  *
3834  *  Returns 0 on success, negative on failure
3835  **/
3836 int igb_setup_rx_resources(struct igb_ring *rx_ring)
3837 {
3838 	struct device *dev = rx_ring->dev;
3839 	int size;
3840 
3841 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3842 
3843 	rx_ring->rx_buffer_info = vmalloc(size);
3844 	if (!rx_ring->rx_buffer_info)
3845 		goto err;
3846 
3847 	/* Round up to nearest 4K */
3848 	rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
3849 	rx_ring->size = ALIGN(rx_ring->size, 4096);
3850 
3851 	rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
3852 					   &rx_ring->dma, GFP_KERNEL);
3853 	if (!rx_ring->desc)
3854 		goto err;
3855 
3856 	rx_ring->next_to_alloc = 0;
3857 	rx_ring->next_to_clean = 0;
3858 	rx_ring->next_to_use = 0;
3859 
3860 	return 0;
3861 
3862 err:
3863 	vfree(rx_ring->rx_buffer_info);
3864 	rx_ring->rx_buffer_info = NULL;
3865 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
3866 	return -ENOMEM;
3867 }
3868 
3869 /**
3870  *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
3871  *				 (Descriptors) for all queues
3872  *  @adapter: board private structure
3873  *
3874  *  Return 0 on success, negative on failure
3875  **/
3876 static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
3877 {
3878 	struct pci_dev *pdev = adapter->pdev;
3879 	int i, err = 0;
3880 
3881 	for (i = 0; i < adapter->num_rx_queues; i++) {
3882 		err = igb_setup_rx_resources(adapter->rx_ring[i]);
3883 		if (err) {
3884 			dev_err(&pdev->dev,
3885 				"Allocation for Rx Queue %u failed\n", i);
3886 			for (i--; i >= 0; i--)
3887 				igb_free_rx_resources(adapter->rx_ring[i]);
3888 			break;
3889 		}
3890 	}
3891 
3892 	return err;
3893 }
3894 
3895 /**
3896  *  igb_setup_mrqc - configure the multiple receive queue control registers
3897  *  @adapter: Board private structure
3898  **/
3899 static void igb_setup_mrqc(struct igb_adapter *adapter)
3900 {
3901 	struct e1000_hw *hw = &adapter->hw;
3902 	u32 mrqc, rxcsum;
3903 	u32 j, num_rx_queues;
3904 	u32 rss_key[10];
3905 
3906 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3907 	for (j = 0; j < 10; j++)
3908 		wr32(E1000_RSSRK(j), rss_key[j]);
3909 
3910 	num_rx_queues = adapter->rss_queues;
3911 
3912 	switch (hw->mac.type) {
3913 	case e1000_82576:
3914 		/* 82576 supports 2 RSS queues for SR-IOV */
3915 		if (adapter->vfs_allocated_count)
3916 			num_rx_queues = 2;
3917 		break;
3918 	default:
3919 		break;
3920 	}
3921 
3922 	if (adapter->rss_indir_tbl_init != num_rx_queues) {
3923 		for (j = 0; j < IGB_RETA_SIZE; j++)
3924 			adapter->rss_indir_tbl[j] =
3925 			(j * num_rx_queues) / IGB_RETA_SIZE;
3926 		adapter->rss_indir_tbl_init = num_rx_queues;
3927 	}
3928 	igb_write_rss_indir_tbl(adapter);
3929 
3930 	/* Disable raw packet checksumming so that RSS hash is placed in
3931 	 * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
3932 	 * offloads as they are enabled by default
3933 	 */
3934 	rxcsum = rd32(E1000_RXCSUM);
3935 	rxcsum |= E1000_RXCSUM_PCSD;
3936 
3937 	if (adapter->hw.mac.type >= e1000_82576)
3938 		/* Enable Receive Checksum Offload for SCTP */
3939 		rxcsum |= E1000_RXCSUM_CRCOFL;
3940 
3941 	/* Don't need to set TUOFL or IPOFL, they default to 1 */
3942 	wr32(E1000_RXCSUM, rxcsum);
3943 
3944 	/* Generate RSS hash based on packet types, TCP/UDP
3945 	 * port numbers and/or IPv4/v6 src and dst addresses
3946 	 */
3947 	mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
3948 	       E1000_MRQC_RSS_FIELD_IPV4_TCP |
3949 	       E1000_MRQC_RSS_FIELD_IPV6 |
3950 	       E1000_MRQC_RSS_FIELD_IPV6_TCP |
3951 	       E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
3952 
3953 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
3954 		mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
3955 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
3956 		mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
3957 
3958 	/* If VMDq is enabled then we set the appropriate mode for that, else
3959 	 * we default to RSS so that an RSS hash is calculated per packet even
3960 	 * if we are only using one queue
3961 	 */
3962 	if (adapter->vfs_allocated_count) {
3963 		if (hw->mac.type > e1000_82575) {
3964 			/* Set the default pool for the PF's first queue */
3965 			u32 vtctl = rd32(E1000_VT_CTL);
3966 
3967 			vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
3968 				   E1000_VT_CTL_DISABLE_DEF_POOL);
3969 			vtctl |= adapter->vfs_allocated_count <<
3970 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
3971 			wr32(E1000_VT_CTL, vtctl);
3972 		}
3973 		if (adapter->rss_queues > 1)
3974 			mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_MQ;
3975 		else
3976 			mrqc |= E1000_MRQC_ENABLE_VMDQ;
3977 	} else {
3978 		if (hw->mac.type != e1000_i211)
3979 			mrqc |= E1000_MRQC_ENABLE_RSS_MQ;
3980 	}
3981 	igb_vmm_control(adapter);
3982 
3983 	wr32(E1000_MRQC, mrqc);
3984 }
3985 
3986 /**
3987  *  igb_setup_rctl - configure the receive control registers
3988  *  @adapter: Board private structure
3989  **/
3990 void igb_setup_rctl(struct igb_adapter *adapter)
3991 {
3992 	struct e1000_hw *hw = &adapter->hw;
3993 	u32 rctl;
3994 
3995 	rctl = rd32(E1000_RCTL);
3996 
3997 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3998 	rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
3999 
4000 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
4001 		(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
4002 
4003 	/* enable stripping of CRC. It's unlikely this will break BMC
4004 	 * redirection as it did with e1000. Newer features require
4005 	 * that the HW strips the CRC.
4006 	 */
4007 	rctl |= E1000_RCTL_SECRC;
4008 
4009 	/* disable store bad packets and clear size bits. */
4010 	rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
4011 
4012 	/* enable LPE to allow for reception of jumbo frames */
4013 	rctl |= E1000_RCTL_LPE;
4014 
4015 	/* disable queue 0 to prevent tail write w/o re-config */
4016 	wr32(E1000_RXDCTL(0), 0);
4017 
4018 	/* Attention!!!  For SR-IOV PF driver operations you must enable
4019 	 * queue drop for all VF and PF queues to prevent head of line blocking
4020 	 * if an un-trusted VF does not provide descriptors to hardware.
4021 	 */
4022 	if (adapter->vfs_allocated_count) {
4023 		/* set all queue drop enable bits */
4024 		wr32(E1000_QDE, ALL_QUEUES);
4025 	}
4026 
4027 	/* This is useful for sniffing bad packets. */
4028 	if (adapter->netdev->features & NETIF_F_RXALL) {
4029 		/* UPE and MPE will be handled by normal PROMISC logic
4030 		 * in e1000e_set_rx_mode
4031 		 */
4032 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
4033 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
4034 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
4035 
4036 		rctl &= ~(E1000_RCTL_DPF | /* Allow filtered pause */
4037 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
4038 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
4039 		 * and that breaks VLANs.
4040 		 */
4041 	}
4042 
4043 	wr32(E1000_RCTL, rctl);
4044 }
4045 
4046 static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
4047 				   int vfn)
4048 {
4049 	struct e1000_hw *hw = &adapter->hw;
4050 	u32 vmolr;
4051 
4052 	if (size > MAX_JUMBO_FRAME_SIZE)
4053 		size = MAX_JUMBO_FRAME_SIZE;
4054 
4055 	vmolr = rd32(E1000_VMOLR(vfn));
4056 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
4057 	vmolr |= size | E1000_VMOLR_LPE;
4058 	wr32(E1000_VMOLR(vfn), vmolr);
4059 
4060 	return 0;
4061 }
4062 
4063 static inline void igb_set_vf_vlan_strip(struct igb_adapter *adapter,
4064 					 int vfn, bool enable)
4065 {
4066 	struct e1000_hw *hw = &adapter->hw;
4067 	u32 val, reg;
4068 
4069 	if (hw->mac.type < e1000_82576)
4070 		return;
4071 
4072 	if (hw->mac.type == e1000_i350)
4073 		reg = E1000_DVMOLR(vfn);
4074 	else
4075 		reg = E1000_VMOLR(vfn);
4076 
4077 	val = rd32(reg);
4078 	if (enable)
4079 		val |= E1000_VMOLR_STRVLAN;
4080 	else
4081 		val &= ~(E1000_VMOLR_STRVLAN);
4082 	wr32(reg, val);
4083 }
4084 
4085 static inline void igb_set_vmolr(struct igb_adapter *adapter,
4086 				 int vfn, bool aupe)
4087 {
4088 	struct e1000_hw *hw = &adapter->hw;
4089 	u32 vmolr;
4090 
4091 	/* This register exists only on 82576 and newer so if we are older then
4092 	 * we should exit and do nothing
4093 	 */
4094 	if (hw->mac.type < e1000_82576)
4095 		return;
4096 
4097 	vmolr = rd32(E1000_VMOLR(vfn));
4098 	if (aupe)
4099 		vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
4100 	else
4101 		vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
4102 
4103 	/* clear all bits that might not be set */
4104 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
4105 
4106 	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
4107 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
4108 	/* for VMDq only allow the VFs and pool 0 to accept broadcast and
4109 	 * multicast packets
4110 	 */
4111 	if (vfn <= adapter->vfs_allocated_count)
4112 		vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
4113 
4114 	wr32(E1000_VMOLR(vfn), vmolr);
4115 }
4116 
4117 /**
4118  *  igb_configure_rx_ring - Configure a receive ring after Reset
4119  *  @adapter: board private structure
4120  *  @ring: receive ring to be configured
4121  *
4122  *  Configure the Rx unit of the MAC after a reset.
4123  **/
4124 void igb_configure_rx_ring(struct igb_adapter *adapter,
4125 			   struct igb_ring *ring)
4126 {
4127 	struct e1000_hw *hw = &adapter->hw;
4128 	union e1000_adv_rx_desc *rx_desc;
4129 	u64 rdba = ring->dma;
4130 	int reg_idx = ring->reg_idx;
4131 	u32 srrctl = 0, rxdctl = 0;
4132 
4133 	/* disable the queue */
4134 	wr32(E1000_RXDCTL(reg_idx), 0);
4135 
4136 	/* Set DMA base address registers */
4137 	wr32(E1000_RDBAL(reg_idx),
4138 	     rdba & 0x00000000ffffffffULL);
4139 	wr32(E1000_RDBAH(reg_idx), rdba >> 32);
4140 	wr32(E1000_RDLEN(reg_idx),
4141 	     ring->count * sizeof(union e1000_adv_rx_desc));
4142 
4143 	/* initialize head and tail */
4144 	ring->tail = adapter->io_addr + E1000_RDT(reg_idx);
4145 	wr32(E1000_RDH(reg_idx), 0);
4146 	writel(0, ring->tail);
4147 
4148 	/* set descriptor configuration */
4149 	srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
4150 	if (ring_uses_large_buffer(ring))
4151 		srrctl |= IGB_RXBUFFER_3072 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4152 	else
4153 		srrctl |= IGB_RXBUFFER_2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4154 	srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
4155 	if (hw->mac.type >= e1000_82580)
4156 		srrctl |= E1000_SRRCTL_TIMESTAMP;
4157 	/* Only set Drop Enable if we are supporting multiple queues */
4158 	if (adapter->vfs_allocated_count || adapter->num_rx_queues > 1)
4159 		srrctl |= E1000_SRRCTL_DROP_EN;
4160 
4161 	wr32(E1000_SRRCTL(reg_idx), srrctl);
4162 
4163 	/* set filtering for VMDQ pools */
4164 	igb_set_vmolr(adapter, reg_idx & 0x7, true);
4165 
4166 	rxdctl |= IGB_RX_PTHRESH;
4167 	rxdctl |= IGB_RX_HTHRESH << 8;
4168 	rxdctl |= IGB_RX_WTHRESH << 16;
4169 
4170 	/* initialize rx_buffer_info */
4171 	memset(ring->rx_buffer_info, 0,
4172 	       sizeof(struct igb_rx_buffer) * ring->count);
4173 
4174 	/* initialize Rx descriptor 0 */
4175 	rx_desc = IGB_RX_DESC(ring, 0);
4176 	rx_desc->wb.upper.length = 0;
4177 
4178 	/* enable receive descriptor fetching */
4179 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
4180 	wr32(E1000_RXDCTL(reg_idx), rxdctl);
4181 }
4182 
4183 static void igb_set_rx_buffer_len(struct igb_adapter *adapter,
4184 				  struct igb_ring *rx_ring)
4185 {
4186 	/* set build_skb and buffer size flags */
4187 	clear_ring_build_skb_enabled(rx_ring);
4188 	clear_ring_uses_large_buffer(rx_ring);
4189 
4190 	if (adapter->flags & IGB_FLAG_RX_LEGACY)
4191 		return;
4192 
4193 	set_ring_build_skb_enabled(rx_ring);
4194 
4195 #if (PAGE_SIZE < 8192)
4196 	if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4197 		return;
4198 
4199 	set_ring_uses_large_buffer(rx_ring);
4200 #endif
4201 }
4202 
4203 /**
4204  *  igb_configure_rx - Configure receive Unit after Reset
4205  *  @adapter: board private structure
4206  *
4207  *  Configure the Rx unit of the MAC after a reset.
4208  **/
4209 static void igb_configure_rx(struct igb_adapter *adapter)
4210 {
4211 	int i;
4212 
4213 	/* set the correct pool for the PF default MAC address in entry 0 */
4214 	igb_set_default_mac_filter(adapter);
4215 
4216 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
4217 	 * the Base and Length of the Rx Descriptor Ring
4218 	 */
4219 	for (i = 0; i < adapter->num_rx_queues; i++) {
4220 		struct igb_ring *rx_ring = adapter->rx_ring[i];
4221 
4222 		igb_set_rx_buffer_len(adapter, rx_ring);
4223 		igb_configure_rx_ring(adapter, rx_ring);
4224 	}
4225 }
4226 
4227 /**
4228  *  igb_free_tx_resources - Free Tx Resources per Queue
4229  *  @tx_ring: Tx descriptor ring for a specific queue
4230  *
4231  *  Free all transmit software resources
4232  **/
4233 void igb_free_tx_resources(struct igb_ring *tx_ring)
4234 {
4235 	igb_clean_tx_ring(tx_ring);
4236 
4237 	vfree(tx_ring->tx_buffer_info);
4238 	tx_ring->tx_buffer_info = NULL;
4239 
4240 	/* if not set, then don't free */
4241 	if (!tx_ring->desc)
4242 		return;
4243 
4244 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4245 			  tx_ring->desc, tx_ring->dma);
4246 
4247 	tx_ring->desc = NULL;
4248 }
4249 
4250 /**
4251  *  igb_free_all_tx_resources - Free Tx Resources for All Queues
4252  *  @adapter: board private structure
4253  *
4254  *  Free all transmit software resources
4255  **/
4256 static void igb_free_all_tx_resources(struct igb_adapter *adapter)
4257 {
4258 	int i;
4259 
4260 	for (i = 0; i < adapter->num_tx_queues; i++)
4261 		if (adapter->tx_ring[i])
4262 			igb_free_tx_resources(adapter->tx_ring[i]);
4263 }
4264 
4265 /**
4266  *  igb_clean_tx_ring - Free Tx Buffers
4267  *  @tx_ring: ring to be cleaned
4268  **/
4269 static void igb_clean_tx_ring(struct igb_ring *tx_ring)
4270 {
4271 	u16 i = tx_ring->next_to_clean;
4272 	struct igb_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
4273 
4274 	while (i != tx_ring->next_to_use) {
4275 		union e1000_adv_tx_desc *eop_desc, *tx_desc;
4276 
4277 		/* Free all the Tx ring sk_buffs */
4278 		dev_kfree_skb_any(tx_buffer->skb);
4279 
4280 		/* unmap skb header data */
4281 		dma_unmap_single(tx_ring->dev,
4282 				 dma_unmap_addr(tx_buffer, dma),
4283 				 dma_unmap_len(tx_buffer, len),
4284 				 DMA_TO_DEVICE);
4285 
4286 		/* check for eop_desc to determine the end of the packet */
4287 		eop_desc = tx_buffer->next_to_watch;
4288 		tx_desc = IGB_TX_DESC(tx_ring, i);
4289 
4290 		/* unmap remaining buffers */
4291 		while (tx_desc != eop_desc) {
4292 			tx_buffer++;
4293 			tx_desc++;
4294 			i++;
4295 			if (unlikely(i == tx_ring->count)) {
4296 				i = 0;
4297 				tx_buffer = tx_ring->tx_buffer_info;
4298 				tx_desc = IGB_TX_DESC(tx_ring, 0);
4299 			}
4300 
4301 			/* unmap any remaining paged data */
4302 			if (dma_unmap_len(tx_buffer, len))
4303 				dma_unmap_page(tx_ring->dev,
4304 					       dma_unmap_addr(tx_buffer, dma),
4305 					       dma_unmap_len(tx_buffer, len),
4306 					       DMA_TO_DEVICE);
4307 		}
4308 
4309 		/* move us one more past the eop_desc for start of next pkt */
4310 		tx_buffer++;
4311 		i++;
4312 		if (unlikely(i == tx_ring->count)) {
4313 			i = 0;
4314 			tx_buffer = tx_ring->tx_buffer_info;
4315 		}
4316 	}
4317 
4318 	/* reset BQL for queue */
4319 	netdev_tx_reset_queue(txring_txq(tx_ring));
4320 
4321 	/* reset next_to_use and next_to_clean */
4322 	tx_ring->next_to_use = 0;
4323 	tx_ring->next_to_clean = 0;
4324 }
4325 
4326 /**
4327  *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
4328  *  @adapter: board private structure
4329  **/
4330 static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
4331 {
4332 	int i;
4333 
4334 	for (i = 0; i < adapter->num_tx_queues; i++)
4335 		if (adapter->tx_ring[i])
4336 			igb_clean_tx_ring(adapter->tx_ring[i]);
4337 }
4338 
4339 /**
4340  *  igb_free_rx_resources - Free Rx Resources
4341  *  @rx_ring: ring to clean the resources from
4342  *
4343  *  Free all receive software resources
4344  **/
4345 void igb_free_rx_resources(struct igb_ring *rx_ring)
4346 {
4347 	igb_clean_rx_ring(rx_ring);
4348 
4349 	vfree(rx_ring->rx_buffer_info);
4350 	rx_ring->rx_buffer_info = NULL;
4351 
4352 	/* if not set, then don't free */
4353 	if (!rx_ring->desc)
4354 		return;
4355 
4356 	dma_free_coherent(rx_ring->dev, rx_ring->size,
4357 			  rx_ring->desc, rx_ring->dma);
4358 
4359 	rx_ring->desc = NULL;
4360 }
4361 
4362 /**
4363  *  igb_free_all_rx_resources - Free Rx Resources for All Queues
4364  *  @adapter: board private structure
4365  *
4366  *  Free all receive software resources
4367  **/
4368 static void igb_free_all_rx_resources(struct igb_adapter *adapter)
4369 {
4370 	int i;
4371 
4372 	for (i = 0; i < adapter->num_rx_queues; i++)
4373 		if (adapter->rx_ring[i])
4374 			igb_free_rx_resources(adapter->rx_ring[i]);
4375 }
4376 
4377 /**
4378  *  igb_clean_rx_ring - Free Rx Buffers per Queue
4379  *  @rx_ring: ring to free buffers from
4380  **/
4381 static void igb_clean_rx_ring(struct igb_ring *rx_ring)
4382 {
4383 	u16 i = rx_ring->next_to_clean;
4384 
4385 	if (rx_ring->skb)
4386 		dev_kfree_skb(rx_ring->skb);
4387 	rx_ring->skb = NULL;
4388 
4389 	/* Free all the Rx ring sk_buffs */
4390 	while (i != rx_ring->next_to_alloc) {
4391 		struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
4392 
4393 		/* Invalidate cache lines that may have been written to by
4394 		 * device so that we avoid corrupting memory.
4395 		 */
4396 		dma_sync_single_range_for_cpu(rx_ring->dev,
4397 					      buffer_info->dma,
4398 					      buffer_info->page_offset,
4399 					      igb_rx_bufsz(rx_ring),
4400 					      DMA_FROM_DEVICE);
4401 
4402 		/* free resources associated with mapping */
4403 		dma_unmap_page_attrs(rx_ring->dev,
4404 				     buffer_info->dma,
4405 				     igb_rx_pg_size(rx_ring),
4406 				     DMA_FROM_DEVICE,
4407 				     IGB_RX_DMA_ATTR);
4408 		__page_frag_cache_drain(buffer_info->page,
4409 					buffer_info->pagecnt_bias);
4410 
4411 		i++;
4412 		if (i == rx_ring->count)
4413 			i = 0;
4414 	}
4415 
4416 	rx_ring->next_to_alloc = 0;
4417 	rx_ring->next_to_clean = 0;
4418 	rx_ring->next_to_use = 0;
4419 }
4420 
4421 /**
4422  *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
4423  *  @adapter: board private structure
4424  **/
4425 static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
4426 {
4427 	int i;
4428 
4429 	for (i = 0; i < adapter->num_rx_queues; i++)
4430 		if (adapter->rx_ring[i])
4431 			igb_clean_rx_ring(adapter->rx_ring[i]);
4432 }
4433 
4434 /**
4435  *  igb_set_mac - Change the Ethernet Address of the NIC
4436  *  @netdev: network interface device structure
4437  *  @p: pointer to an address structure
4438  *
4439  *  Returns 0 on success, negative on failure
4440  **/
4441 static int igb_set_mac(struct net_device *netdev, void *p)
4442 {
4443 	struct igb_adapter *adapter = netdev_priv(netdev);
4444 	struct e1000_hw *hw = &adapter->hw;
4445 	struct sockaddr *addr = p;
4446 
4447 	if (!is_valid_ether_addr(addr->sa_data))
4448 		return -EADDRNOTAVAIL;
4449 
4450 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4451 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
4452 
4453 	/* set the correct pool for the new PF MAC address in entry 0 */
4454 	igb_set_default_mac_filter(adapter);
4455 
4456 	return 0;
4457 }
4458 
4459 /**
4460  *  igb_write_mc_addr_list - write multicast addresses to MTA
4461  *  @netdev: network interface device structure
4462  *
4463  *  Writes multicast address list to the MTA hash table.
4464  *  Returns: -ENOMEM on failure
4465  *           0 on no addresses written
4466  *           X on writing X addresses to MTA
4467  **/
4468 static int igb_write_mc_addr_list(struct net_device *netdev)
4469 {
4470 	struct igb_adapter *adapter = netdev_priv(netdev);
4471 	struct e1000_hw *hw = &adapter->hw;
4472 	struct netdev_hw_addr *ha;
4473 	u8  *mta_list;
4474 	int i;
4475 
4476 	if (netdev_mc_empty(netdev)) {
4477 		/* nothing to program, so clear mc list */
4478 		igb_update_mc_addr_list(hw, NULL, 0);
4479 		igb_restore_vf_multicasts(adapter);
4480 		return 0;
4481 	}
4482 
4483 	mta_list = kzalloc(netdev_mc_count(netdev) * 6, GFP_ATOMIC);
4484 	if (!mta_list)
4485 		return -ENOMEM;
4486 
4487 	/* The shared function expects a packed array of only addresses. */
4488 	i = 0;
4489 	netdev_for_each_mc_addr(ha, netdev)
4490 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
4491 
4492 	igb_update_mc_addr_list(hw, mta_list, i);
4493 	kfree(mta_list);
4494 
4495 	return netdev_mc_count(netdev);
4496 }
4497 
4498 static int igb_vlan_promisc_enable(struct igb_adapter *adapter)
4499 {
4500 	struct e1000_hw *hw = &adapter->hw;
4501 	u32 i, pf_id;
4502 
4503 	switch (hw->mac.type) {
4504 	case e1000_i210:
4505 	case e1000_i211:
4506 	case e1000_i350:
4507 		/* VLAN filtering needed for VLAN prio filter */
4508 		if (adapter->netdev->features & NETIF_F_NTUPLE)
4509 			break;
4510 		/* fall through */
4511 	case e1000_82576:
4512 	case e1000_82580:
4513 	case e1000_i354:
4514 		/* VLAN filtering needed for pool filtering */
4515 		if (adapter->vfs_allocated_count)
4516 			break;
4517 		/* fall through */
4518 	default:
4519 		return 1;
4520 	}
4521 
4522 	/* We are already in VLAN promisc, nothing to do */
4523 	if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
4524 		return 0;
4525 
4526 	if (!adapter->vfs_allocated_count)
4527 		goto set_vfta;
4528 
4529 	/* Add PF to all active pools */
4530 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
4531 
4532 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
4533 		u32 vlvf = rd32(E1000_VLVF(i));
4534 
4535 		vlvf |= BIT(pf_id);
4536 		wr32(E1000_VLVF(i), vlvf);
4537 	}
4538 
4539 set_vfta:
4540 	/* Set all bits in the VLAN filter table array */
4541 	for (i = E1000_VLAN_FILTER_TBL_SIZE; i--;)
4542 		hw->mac.ops.write_vfta(hw, i, ~0U);
4543 
4544 	/* Set flag so we don't redo unnecessary work */
4545 	adapter->flags |= IGB_FLAG_VLAN_PROMISC;
4546 
4547 	return 0;
4548 }
4549 
4550 #define VFTA_BLOCK_SIZE 8
4551 static void igb_scrub_vfta(struct igb_adapter *adapter, u32 vfta_offset)
4552 {
4553 	struct e1000_hw *hw = &adapter->hw;
4554 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4555 	u32 vid_start = vfta_offset * 32;
4556 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4557 	u32 i, vid, word, bits, pf_id;
4558 
4559 	/* guarantee that we don't scrub out management VLAN */
4560 	vid = adapter->mng_vlan_id;
4561 	if (vid >= vid_start && vid < vid_end)
4562 		vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4563 
4564 	if (!adapter->vfs_allocated_count)
4565 		goto set_vfta;
4566 
4567 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
4568 
4569 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
4570 		u32 vlvf = rd32(E1000_VLVF(i));
4571 
4572 		/* pull VLAN ID from VLVF */
4573 		vid = vlvf & VLAN_VID_MASK;
4574 
4575 		/* only concern ourselves with a certain range */
4576 		if (vid < vid_start || vid >= vid_end)
4577 			continue;
4578 
4579 		if (vlvf & E1000_VLVF_VLANID_ENABLE) {
4580 			/* record VLAN ID in VFTA */
4581 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4582 
4583 			/* if PF is part of this then continue */
4584 			if (test_bit(vid, adapter->active_vlans))
4585 				continue;
4586 		}
4587 
4588 		/* remove PF from the pool */
4589 		bits = ~BIT(pf_id);
4590 		bits &= rd32(E1000_VLVF(i));
4591 		wr32(E1000_VLVF(i), bits);
4592 	}
4593 
4594 set_vfta:
4595 	/* extract values from active_vlans and write back to VFTA */
4596 	for (i = VFTA_BLOCK_SIZE; i--;) {
4597 		vid = (vfta_offset + i) * 32;
4598 		word = vid / BITS_PER_LONG;
4599 		bits = vid % BITS_PER_LONG;
4600 
4601 		vfta[i] |= adapter->active_vlans[word] >> bits;
4602 
4603 		hw->mac.ops.write_vfta(hw, vfta_offset + i, vfta[i]);
4604 	}
4605 }
4606 
4607 static void igb_vlan_promisc_disable(struct igb_adapter *adapter)
4608 {
4609 	u32 i;
4610 
4611 	/* We are not in VLAN promisc, nothing to do */
4612 	if (!(adapter->flags & IGB_FLAG_VLAN_PROMISC))
4613 		return;
4614 
4615 	/* Set flag so we don't redo unnecessary work */
4616 	adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
4617 
4618 	for (i = 0; i < E1000_VLAN_FILTER_TBL_SIZE; i += VFTA_BLOCK_SIZE)
4619 		igb_scrub_vfta(adapter, i);
4620 }
4621 
4622 /**
4623  *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
4624  *  @netdev: network interface device structure
4625  *
4626  *  The set_rx_mode entry point is called whenever the unicast or multicast
4627  *  address lists or the network interface flags are updated.  This routine is
4628  *  responsible for configuring the hardware for proper unicast, multicast,
4629  *  promiscuous mode, and all-multi behavior.
4630  **/
4631 static void igb_set_rx_mode(struct net_device *netdev)
4632 {
4633 	struct igb_adapter *adapter = netdev_priv(netdev);
4634 	struct e1000_hw *hw = &adapter->hw;
4635 	unsigned int vfn = adapter->vfs_allocated_count;
4636 	u32 rctl = 0, vmolr = 0, rlpml = MAX_JUMBO_FRAME_SIZE;
4637 	int count;
4638 
4639 	/* Check for Promiscuous and All Multicast modes */
4640 	if (netdev->flags & IFF_PROMISC) {
4641 		rctl |= E1000_RCTL_UPE | E1000_RCTL_MPE;
4642 		vmolr |= E1000_VMOLR_MPME;
4643 
4644 		/* enable use of UTA filter to force packets to default pool */
4645 		if (hw->mac.type == e1000_82576)
4646 			vmolr |= E1000_VMOLR_ROPE;
4647 	} else {
4648 		if (netdev->flags & IFF_ALLMULTI) {
4649 			rctl |= E1000_RCTL_MPE;
4650 			vmolr |= E1000_VMOLR_MPME;
4651 		} else {
4652 			/* Write addresses to the MTA, if the attempt fails
4653 			 * then we should just turn on promiscuous mode so
4654 			 * that we can at least receive multicast traffic
4655 			 */
4656 			count = igb_write_mc_addr_list(netdev);
4657 			if (count < 0) {
4658 				rctl |= E1000_RCTL_MPE;
4659 				vmolr |= E1000_VMOLR_MPME;
4660 			} else if (count) {
4661 				vmolr |= E1000_VMOLR_ROMPE;
4662 			}
4663 		}
4664 	}
4665 
4666 	/* Write addresses to available RAR registers, if there is not
4667 	 * sufficient space to store all the addresses then enable
4668 	 * unicast promiscuous mode
4669 	 */
4670 	if (__dev_uc_sync(netdev, igb_uc_sync, igb_uc_unsync)) {
4671 		rctl |= E1000_RCTL_UPE;
4672 		vmolr |= E1000_VMOLR_ROPE;
4673 	}
4674 
4675 	/* enable VLAN filtering by default */
4676 	rctl |= E1000_RCTL_VFE;
4677 
4678 	/* disable VLAN filtering for modes that require it */
4679 	if ((netdev->flags & IFF_PROMISC) ||
4680 	    (netdev->features & NETIF_F_RXALL)) {
4681 		/* if we fail to set all rules then just clear VFE */
4682 		if (igb_vlan_promisc_enable(adapter))
4683 			rctl &= ~E1000_RCTL_VFE;
4684 	} else {
4685 		igb_vlan_promisc_disable(adapter);
4686 	}
4687 
4688 	/* update state of unicast, multicast, and VLAN filtering modes */
4689 	rctl |= rd32(E1000_RCTL) & ~(E1000_RCTL_UPE | E1000_RCTL_MPE |
4690 				     E1000_RCTL_VFE);
4691 	wr32(E1000_RCTL, rctl);
4692 
4693 #if (PAGE_SIZE < 8192)
4694 	if (!adapter->vfs_allocated_count) {
4695 		if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4696 			rlpml = IGB_MAX_FRAME_BUILD_SKB;
4697 	}
4698 #endif
4699 	wr32(E1000_RLPML, rlpml);
4700 
4701 	/* In order to support SR-IOV and eventually VMDq it is necessary to set
4702 	 * the VMOLR to enable the appropriate modes.  Without this workaround
4703 	 * we will have issues with VLAN tag stripping not being done for frames
4704 	 * that are only arriving because we are the default pool
4705 	 */
4706 	if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
4707 		return;
4708 
4709 	/* set UTA to appropriate mode */
4710 	igb_set_uta(adapter, !!(vmolr & E1000_VMOLR_ROPE));
4711 
4712 	vmolr |= rd32(E1000_VMOLR(vfn)) &
4713 		 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
4714 
4715 	/* enable Rx jumbo frames, restrict as needed to support build_skb */
4716 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
4717 #if (PAGE_SIZE < 8192)
4718 	if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4719 		vmolr |= IGB_MAX_FRAME_BUILD_SKB;
4720 	else
4721 #endif
4722 		vmolr |= MAX_JUMBO_FRAME_SIZE;
4723 	vmolr |= E1000_VMOLR_LPE;
4724 
4725 	wr32(E1000_VMOLR(vfn), vmolr);
4726 
4727 	igb_restore_vf_multicasts(adapter);
4728 }
4729 
4730 static void igb_check_wvbr(struct igb_adapter *adapter)
4731 {
4732 	struct e1000_hw *hw = &adapter->hw;
4733 	u32 wvbr = 0;
4734 
4735 	switch (hw->mac.type) {
4736 	case e1000_82576:
4737 	case e1000_i350:
4738 		wvbr = rd32(E1000_WVBR);
4739 		if (!wvbr)
4740 			return;
4741 		break;
4742 	default:
4743 		break;
4744 	}
4745 
4746 	adapter->wvbr |= wvbr;
4747 }
4748 
4749 #define IGB_STAGGERED_QUEUE_OFFSET 8
4750 
4751 static void igb_spoof_check(struct igb_adapter *adapter)
4752 {
4753 	int j;
4754 
4755 	if (!adapter->wvbr)
4756 		return;
4757 
4758 	for (j = 0; j < adapter->vfs_allocated_count; j++) {
4759 		if (adapter->wvbr & BIT(j) ||
4760 		    adapter->wvbr & BIT(j + IGB_STAGGERED_QUEUE_OFFSET)) {
4761 			dev_warn(&adapter->pdev->dev,
4762 				"Spoof event(s) detected on VF %d\n", j);
4763 			adapter->wvbr &=
4764 				~(BIT(j) |
4765 				  BIT(j + IGB_STAGGERED_QUEUE_OFFSET));
4766 		}
4767 	}
4768 }
4769 
4770 /* Need to wait a few seconds after link up to get diagnostic information from
4771  * the phy
4772  */
4773 static void igb_update_phy_info(struct timer_list *t)
4774 {
4775 	struct igb_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4776 	igb_get_phy_info(&adapter->hw);
4777 }
4778 
4779 /**
4780  *  igb_has_link - check shared code for link and determine up/down
4781  *  @adapter: pointer to driver private info
4782  **/
4783 bool igb_has_link(struct igb_adapter *adapter)
4784 {
4785 	struct e1000_hw *hw = &adapter->hw;
4786 	bool link_active = false;
4787 
4788 	/* get_link_status is set on LSC (link status) interrupt or
4789 	 * rx sequence error interrupt.  get_link_status will stay
4790 	 * false until the e1000_check_for_link establishes link
4791 	 * for copper adapters ONLY
4792 	 */
4793 	switch (hw->phy.media_type) {
4794 	case e1000_media_type_copper:
4795 		if (!hw->mac.get_link_status)
4796 			return true;
4797 	case e1000_media_type_internal_serdes:
4798 		hw->mac.ops.check_for_link(hw);
4799 		link_active = !hw->mac.get_link_status;
4800 		break;
4801 	default:
4802 	case e1000_media_type_unknown:
4803 		break;
4804 	}
4805 
4806 	if (((hw->mac.type == e1000_i210) ||
4807 	     (hw->mac.type == e1000_i211)) &&
4808 	     (hw->phy.id == I210_I_PHY_ID)) {
4809 		if (!netif_carrier_ok(adapter->netdev)) {
4810 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4811 		} else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
4812 			adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
4813 			adapter->link_check_timeout = jiffies;
4814 		}
4815 	}
4816 
4817 	return link_active;
4818 }
4819 
4820 static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
4821 {
4822 	bool ret = false;
4823 	u32 ctrl_ext, thstat;
4824 
4825 	/* check for thermal sensor event on i350 copper only */
4826 	if (hw->mac.type == e1000_i350) {
4827 		thstat = rd32(E1000_THSTAT);
4828 		ctrl_ext = rd32(E1000_CTRL_EXT);
4829 
4830 		if ((hw->phy.media_type == e1000_media_type_copper) &&
4831 		    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
4832 			ret = !!(thstat & event);
4833 	}
4834 
4835 	return ret;
4836 }
4837 
4838 /**
4839  *  igb_check_lvmmc - check for malformed packets received
4840  *  and indicated in LVMMC register
4841  *  @adapter: pointer to adapter
4842  **/
4843 static void igb_check_lvmmc(struct igb_adapter *adapter)
4844 {
4845 	struct e1000_hw *hw = &adapter->hw;
4846 	u32 lvmmc;
4847 
4848 	lvmmc = rd32(E1000_LVMMC);
4849 	if (lvmmc) {
4850 		if (unlikely(net_ratelimit())) {
4851 			netdev_warn(adapter->netdev,
4852 				    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
4853 				    lvmmc);
4854 		}
4855 	}
4856 }
4857 
4858 /**
4859  *  igb_watchdog - Timer Call-back
4860  *  @data: pointer to adapter cast into an unsigned long
4861  **/
4862 static void igb_watchdog(struct timer_list *t)
4863 {
4864 	struct igb_adapter *adapter = from_timer(adapter, t, watchdog_timer);
4865 	/* Do the rest outside of interrupt context */
4866 	schedule_work(&adapter->watchdog_task);
4867 }
4868 
4869 static void igb_watchdog_task(struct work_struct *work)
4870 {
4871 	struct igb_adapter *adapter = container_of(work,
4872 						   struct igb_adapter,
4873 						   watchdog_task);
4874 	struct e1000_hw *hw = &adapter->hw;
4875 	struct e1000_phy_info *phy = &hw->phy;
4876 	struct net_device *netdev = adapter->netdev;
4877 	u32 link;
4878 	int i;
4879 	u32 connsw;
4880 	u16 phy_data, retry_count = 20;
4881 
4882 	link = igb_has_link(adapter);
4883 
4884 	if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
4885 		if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
4886 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4887 		else
4888 			link = false;
4889 	}
4890 
4891 	/* Force link down if we have fiber to swap to */
4892 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4893 		if (hw->phy.media_type == e1000_media_type_copper) {
4894 			connsw = rd32(E1000_CONNSW);
4895 			if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
4896 				link = 0;
4897 		}
4898 	}
4899 	if (link) {
4900 		/* Perform a reset if the media type changed. */
4901 		if (hw->dev_spec._82575.media_changed) {
4902 			hw->dev_spec._82575.media_changed = false;
4903 			adapter->flags |= IGB_FLAG_MEDIA_RESET;
4904 			igb_reset(adapter);
4905 		}
4906 		/* Cancel scheduled suspend requests. */
4907 		pm_runtime_resume(netdev->dev.parent);
4908 
4909 		if (!netif_carrier_ok(netdev)) {
4910 			u32 ctrl;
4911 
4912 			hw->mac.ops.get_speed_and_duplex(hw,
4913 							 &adapter->link_speed,
4914 							 &adapter->link_duplex);
4915 
4916 			ctrl = rd32(E1000_CTRL);
4917 			/* Links status message must follow this format */
4918 			netdev_info(netdev,
4919 			       "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4920 			       netdev->name,
4921 			       adapter->link_speed,
4922 			       adapter->link_duplex == FULL_DUPLEX ?
4923 			       "Full" : "Half",
4924 			       (ctrl & E1000_CTRL_TFCE) &&
4925 			       (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
4926 			       (ctrl & E1000_CTRL_RFCE) ?  "RX" :
4927 			       (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
4928 
4929 			/* disable EEE if enabled */
4930 			if ((adapter->flags & IGB_FLAG_EEE) &&
4931 				(adapter->link_duplex == HALF_DUPLEX)) {
4932 				dev_info(&adapter->pdev->dev,
4933 				"EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
4934 				adapter->hw.dev_spec._82575.eee_disable = true;
4935 				adapter->flags &= ~IGB_FLAG_EEE;
4936 			}
4937 
4938 			/* check if SmartSpeed worked */
4939 			igb_check_downshift(hw);
4940 			if (phy->speed_downgraded)
4941 				netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
4942 
4943 			/* check for thermal sensor event */
4944 			if (igb_thermal_sensor_event(hw,
4945 			    E1000_THSTAT_LINK_THROTTLE))
4946 				netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
4947 
4948 			/* adjust timeout factor according to speed/duplex */
4949 			adapter->tx_timeout_factor = 1;
4950 			switch (adapter->link_speed) {
4951 			case SPEED_10:
4952 				adapter->tx_timeout_factor = 14;
4953 				break;
4954 			case SPEED_100:
4955 				/* maybe add some timeout factor ? */
4956 				break;
4957 			}
4958 
4959 			if (adapter->link_speed != SPEED_1000)
4960 				goto no_wait;
4961 
4962 			/* wait for Remote receiver status OK */
4963 retry_read_status:
4964 			if (!igb_read_phy_reg(hw, PHY_1000T_STATUS,
4965 					      &phy_data)) {
4966 				if (!(phy_data & SR_1000T_REMOTE_RX_STATUS) &&
4967 				    retry_count) {
4968 					msleep(100);
4969 					retry_count--;
4970 					goto retry_read_status;
4971 				} else if (!retry_count) {
4972 					dev_err(&adapter->pdev->dev, "exceed max 2 second\n");
4973 				}
4974 			} else {
4975 				dev_err(&adapter->pdev->dev, "read 1000Base-T Status Reg\n");
4976 			}
4977 no_wait:
4978 			netif_carrier_on(netdev);
4979 
4980 			igb_ping_all_vfs(adapter);
4981 			igb_check_vf_rate_limit(adapter);
4982 
4983 			/* link state has changed, schedule phy info update */
4984 			if (!test_bit(__IGB_DOWN, &adapter->state))
4985 				mod_timer(&adapter->phy_info_timer,
4986 					  round_jiffies(jiffies + 2 * HZ));
4987 		}
4988 	} else {
4989 		if (netif_carrier_ok(netdev)) {
4990 			adapter->link_speed = 0;
4991 			adapter->link_duplex = 0;
4992 
4993 			/* check for thermal sensor event */
4994 			if (igb_thermal_sensor_event(hw,
4995 			    E1000_THSTAT_PWR_DOWN)) {
4996 				netdev_err(netdev, "The network adapter was stopped because it overheated\n");
4997 			}
4998 
4999 			/* Links status message must follow this format */
5000 			netdev_info(netdev, "igb: %s NIC Link is Down\n",
5001 			       netdev->name);
5002 			netif_carrier_off(netdev);
5003 
5004 			igb_ping_all_vfs(adapter);
5005 
5006 			/* link state has changed, schedule phy info update */
5007 			if (!test_bit(__IGB_DOWN, &adapter->state))
5008 				mod_timer(&adapter->phy_info_timer,
5009 					  round_jiffies(jiffies + 2 * HZ));
5010 
5011 			/* link is down, time to check for alternate media */
5012 			if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5013 				igb_check_swap_media(adapter);
5014 				if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5015 					schedule_work(&adapter->reset_task);
5016 					/* return immediately */
5017 					return;
5018 				}
5019 			}
5020 			pm_schedule_suspend(netdev->dev.parent,
5021 					    MSEC_PER_SEC * 5);
5022 
5023 		/* also check for alternate media here */
5024 		} else if (!netif_carrier_ok(netdev) &&
5025 			   (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
5026 			igb_check_swap_media(adapter);
5027 			if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5028 				schedule_work(&adapter->reset_task);
5029 				/* return immediately */
5030 				return;
5031 			}
5032 		}
5033 	}
5034 
5035 	spin_lock(&adapter->stats64_lock);
5036 	igb_update_stats(adapter);
5037 	spin_unlock(&adapter->stats64_lock);
5038 
5039 	for (i = 0; i < adapter->num_tx_queues; i++) {
5040 		struct igb_ring *tx_ring = adapter->tx_ring[i];
5041 		if (!netif_carrier_ok(netdev)) {
5042 			/* We've lost link, so the controller stops DMA,
5043 			 * but we've got queued Tx work that's never going
5044 			 * to get done, so reset controller to flush Tx.
5045 			 * (Do the reset outside of interrupt context).
5046 			 */
5047 			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
5048 				adapter->tx_timeout_count++;
5049 				schedule_work(&adapter->reset_task);
5050 				/* return immediately since reset is imminent */
5051 				return;
5052 			}
5053 		}
5054 
5055 		/* Force detection of hung controller every watchdog period */
5056 		set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
5057 	}
5058 
5059 	/* Cause software interrupt to ensure Rx ring is cleaned */
5060 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
5061 		u32 eics = 0;
5062 
5063 		for (i = 0; i < adapter->num_q_vectors; i++)
5064 			eics |= adapter->q_vector[i]->eims_value;
5065 		wr32(E1000_EICS, eics);
5066 	} else {
5067 		wr32(E1000_ICS, E1000_ICS_RXDMT0);
5068 	}
5069 
5070 	igb_spoof_check(adapter);
5071 	igb_ptp_rx_hang(adapter);
5072 	igb_ptp_tx_hang(adapter);
5073 
5074 	/* Check LVMMC register on i350/i354 only */
5075 	if ((adapter->hw.mac.type == e1000_i350) ||
5076 	    (adapter->hw.mac.type == e1000_i354))
5077 		igb_check_lvmmc(adapter);
5078 
5079 	/* Reset the timer */
5080 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
5081 		if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
5082 			mod_timer(&adapter->watchdog_timer,
5083 				  round_jiffies(jiffies +  HZ));
5084 		else
5085 			mod_timer(&adapter->watchdog_timer,
5086 				  round_jiffies(jiffies + 2 * HZ));
5087 	}
5088 }
5089 
5090 enum latency_range {
5091 	lowest_latency = 0,
5092 	low_latency = 1,
5093 	bulk_latency = 2,
5094 	latency_invalid = 255
5095 };
5096 
5097 /**
5098  *  igb_update_ring_itr - update the dynamic ITR value based on packet size
5099  *  @q_vector: pointer to q_vector
5100  *
5101  *  Stores a new ITR value based on strictly on packet size.  This
5102  *  algorithm is less sophisticated than that used in igb_update_itr,
5103  *  due to the difficulty of synchronizing statistics across multiple
5104  *  receive rings.  The divisors and thresholds used by this function
5105  *  were determined based on theoretical maximum wire speed and testing
5106  *  data, in order to minimize response time while increasing bulk
5107  *  throughput.
5108  *  This functionality is controlled by ethtool's coalescing settings.
5109  *  NOTE:  This function is called only when operating in a multiqueue
5110  *         receive environment.
5111  **/
5112 static void igb_update_ring_itr(struct igb_q_vector *q_vector)
5113 {
5114 	int new_val = q_vector->itr_val;
5115 	int avg_wire_size = 0;
5116 	struct igb_adapter *adapter = q_vector->adapter;
5117 	unsigned int packets;
5118 
5119 	/* For non-gigabit speeds, just fix the interrupt rate at 4000
5120 	 * ints/sec - ITR timer value of 120 ticks.
5121 	 */
5122 	if (adapter->link_speed != SPEED_1000) {
5123 		new_val = IGB_4K_ITR;
5124 		goto set_itr_val;
5125 	}
5126 
5127 	packets = q_vector->rx.total_packets;
5128 	if (packets)
5129 		avg_wire_size = q_vector->rx.total_bytes / packets;
5130 
5131 	packets = q_vector->tx.total_packets;
5132 	if (packets)
5133 		avg_wire_size = max_t(u32, avg_wire_size,
5134 				      q_vector->tx.total_bytes / packets);
5135 
5136 	/* if avg_wire_size isn't set no work was done */
5137 	if (!avg_wire_size)
5138 		goto clear_counts;
5139 
5140 	/* Add 24 bytes to size to account for CRC, preamble, and gap */
5141 	avg_wire_size += 24;
5142 
5143 	/* Don't starve jumbo frames */
5144 	avg_wire_size = min(avg_wire_size, 3000);
5145 
5146 	/* Give a little boost to mid-size frames */
5147 	if ((avg_wire_size > 300) && (avg_wire_size < 1200))
5148 		new_val = avg_wire_size / 3;
5149 	else
5150 		new_val = avg_wire_size / 2;
5151 
5152 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5153 	if (new_val < IGB_20K_ITR &&
5154 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5155 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5156 		new_val = IGB_20K_ITR;
5157 
5158 set_itr_val:
5159 	if (new_val != q_vector->itr_val) {
5160 		q_vector->itr_val = new_val;
5161 		q_vector->set_itr = 1;
5162 	}
5163 clear_counts:
5164 	q_vector->rx.total_bytes = 0;
5165 	q_vector->rx.total_packets = 0;
5166 	q_vector->tx.total_bytes = 0;
5167 	q_vector->tx.total_packets = 0;
5168 }
5169 
5170 /**
5171  *  igb_update_itr - update the dynamic ITR value based on statistics
5172  *  @q_vector: pointer to q_vector
5173  *  @ring_container: ring info to update the itr for
5174  *
5175  *  Stores a new ITR value based on packets and byte
5176  *  counts during the last interrupt.  The advantage of per interrupt
5177  *  computation is faster updates and more accurate ITR for the current
5178  *  traffic pattern.  Constants in this function were computed
5179  *  based on theoretical maximum wire speed and thresholds were set based
5180  *  on testing data as well as attempting to minimize response time
5181  *  while increasing bulk throughput.
5182  *  This functionality is controlled by ethtool's coalescing settings.
5183  *  NOTE:  These calculations are only valid when operating in a single-
5184  *         queue environment.
5185  **/
5186 static void igb_update_itr(struct igb_q_vector *q_vector,
5187 			   struct igb_ring_container *ring_container)
5188 {
5189 	unsigned int packets = ring_container->total_packets;
5190 	unsigned int bytes = ring_container->total_bytes;
5191 	u8 itrval = ring_container->itr;
5192 
5193 	/* no packets, exit with status unchanged */
5194 	if (packets == 0)
5195 		return;
5196 
5197 	switch (itrval) {
5198 	case lowest_latency:
5199 		/* handle TSO and jumbo frames */
5200 		if (bytes/packets > 8000)
5201 			itrval = bulk_latency;
5202 		else if ((packets < 5) && (bytes > 512))
5203 			itrval = low_latency;
5204 		break;
5205 	case low_latency:  /* 50 usec aka 20000 ints/s */
5206 		if (bytes > 10000) {
5207 			/* this if handles the TSO accounting */
5208 			if (bytes/packets > 8000)
5209 				itrval = bulk_latency;
5210 			else if ((packets < 10) || ((bytes/packets) > 1200))
5211 				itrval = bulk_latency;
5212 			else if ((packets > 35))
5213 				itrval = lowest_latency;
5214 		} else if (bytes/packets > 2000) {
5215 			itrval = bulk_latency;
5216 		} else if (packets <= 2 && bytes < 512) {
5217 			itrval = lowest_latency;
5218 		}
5219 		break;
5220 	case bulk_latency: /* 250 usec aka 4000 ints/s */
5221 		if (bytes > 25000) {
5222 			if (packets > 35)
5223 				itrval = low_latency;
5224 		} else if (bytes < 1500) {
5225 			itrval = low_latency;
5226 		}
5227 		break;
5228 	}
5229 
5230 	/* clear work counters since we have the values we need */
5231 	ring_container->total_bytes = 0;
5232 	ring_container->total_packets = 0;
5233 
5234 	/* write updated itr to ring container */
5235 	ring_container->itr = itrval;
5236 }
5237 
5238 static void igb_set_itr(struct igb_q_vector *q_vector)
5239 {
5240 	struct igb_adapter *adapter = q_vector->adapter;
5241 	u32 new_itr = q_vector->itr_val;
5242 	u8 current_itr = 0;
5243 
5244 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
5245 	if (adapter->link_speed != SPEED_1000) {
5246 		current_itr = 0;
5247 		new_itr = IGB_4K_ITR;
5248 		goto set_itr_now;
5249 	}
5250 
5251 	igb_update_itr(q_vector, &q_vector->tx);
5252 	igb_update_itr(q_vector, &q_vector->rx);
5253 
5254 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
5255 
5256 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5257 	if (current_itr == lowest_latency &&
5258 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5259 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5260 		current_itr = low_latency;
5261 
5262 	switch (current_itr) {
5263 	/* counts and packets in update_itr are dependent on these numbers */
5264 	case lowest_latency:
5265 		new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
5266 		break;
5267 	case low_latency:
5268 		new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
5269 		break;
5270 	case bulk_latency:
5271 		new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
5272 		break;
5273 	default:
5274 		break;
5275 	}
5276 
5277 set_itr_now:
5278 	if (new_itr != q_vector->itr_val) {
5279 		/* this attempts to bias the interrupt rate towards Bulk
5280 		 * by adding intermediate steps when interrupt rate is
5281 		 * increasing
5282 		 */
5283 		new_itr = new_itr > q_vector->itr_val ?
5284 			  max((new_itr * q_vector->itr_val) /
5285 			  (new_itr + (q_vector->itr_val >> 2)),
5286 			  new_itr) : new_itr;
5287 		/* Don't write the value here; it resets the adapter's
5288 		 * internal timer, and causes us to delay far longer than
5289 		 * we should between interrupts.  Instead, we write the ITR
5290 		 * value at the beginning of the next interrupt so the timing
5291 		 * ends up being correct.
5292 		 */
5293 		q_vector->itr_val = new_itr;
5294 		q_vector->set_itr = 1;
5295 	}
5296 }
5297 
5298 static void igb_tx_ctxtdesc(struct igb_ring *tx_ring, u32 vlan_macip_lens,
5299 			    u32 type_tucmd, u32 mss_l4len_idx)
5300 {
5301 	struct e1000_adv_tx_context_desc *context_desc;
5302 	u16 i = tx_ring->next_to_use;
5303 
5304 	context_desc = IGB_TX_CTXTDESC(tx_ring, i);
5305 
5306 	i++;
5307 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
5308 
5309 	/* set bits to identify this as an advanced context descriptor */
5310 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
5311 
5312 	/* For 82575, context index must be unique per ring. */
5313 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5314 		mss_l4len_idx |= tx_ring->reg_idx << 4;
5315 
5316 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
5317 	context_desc->seqnum_seed	= 0;
5318 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
5319 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
5320 }
5321 
5322 static int igb_tso(struct igb_ring *tx_ring,
5323 		   struct igb_tx_buffer *first,
5324 		   u8 *hdr_len)
5325 {
5326 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
5327 	struct sk_buff *skb = first->skb;
5328 	union {
5329 		struct iphdr *v4;
5330 		struct ipv6hdr *v6;
5331 		unsigned char *hdr;
5332 	} ip;
5333 	union {
5334 		struct tcphdr *tcp;
5335 		unsigned char *hdr;
5336 	} l4;
5337 	u32 paylen, l4_offset;
5338 	int err;
5339 
5340 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5341 		return 0;
5342 
5343 	if (!skb_is_gso(skb))
5344 		return 0;
5345 
5346 	err = skb_cow_head(skb, 0);
5347 	if (err < 0)
5348 		return err;
5349 
5350 	ip.hdr = skb_network_header(skb);
5351 	l4.hdr = skb_checksum_start(skb);
5352 
5353 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5354 	type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
5355 
5356 	/* initialize outer IP header fields */
5357 	if (ip.v4->version == 4) {
5358 		unsigned char *csum_start = skb_checksum_start(skb);
5359 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
5360 
5361 		/* IP header will have to cancel out any data that
5362 		 * is not a part of the outer IP header
5363 		 */
5364 		ip.v4->check = csum_fold(csum_partial(trans_start,
5365 						      csum_start - trans_start,
5366 						      0));
5367 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
5368 
5369 		ip.v4->tot_len = 0;
5370 		first->tx_flags |= IGB_TX_FLAGS_TSO |
5371 				   IGB_TX_FLAGS_CSUM |
5372 				   IGB_TX_FLAGS_IPV4;
5373 	} else {
5374 		ip.v6->payload_len = 0;
5375 		first->tx_flags |= IGB_TX_FLAGS_TSO |
5376 				   IGB_TX_FLAGS_CSUM;
5377 	}
5378 
5379 	/* determine offset of inner transport header */
5380 	l4_offset = l4.hdr - skb->data;
5381 
5382 	/* compute length of segmentation header */
5383 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
5384 
5385 	/* remove payload length from inner checksum */
5386 	paylen = skb->len - l4_offset;
5387 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
5388 
5389 	/* update gso size and bytecount with header size */
5390 	first->gso_segs = skb_shinfo(skb)->gso_segs;
5391 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
5392 
5393 	/* MSS L4LEN IDX */
5394 	mss_l4len_idx = (*hdr_len - l4_offset) << E1000_ADVTXD_L4LEN_SHIFT;
5395 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
5396 
5397 	/* VLAN MACLEN IPLEN */
5398 	vlan_macip_lens = l4.hdr - ip.hdr;
5399 	vlan_macip_lens |= (ip.hdr - skb->data) << E1000_ADVTXD_MACLEN_SHIFT;
5400 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
5401 
5402 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
5403 
5404 	return 1;
5405 }
5406 
5407 static inline bool igb_ipv6_csum_is_sctp(struct sk_buff *skb)
5408 {
5409 	unsigned int offset = 0;
5410 
5411 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
5412 
5413 	return offset == skb_checksum_start_offset(skb);
5414 }
5415 
5416 static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
5417 {
5418 	struct sk_buff *skb = first->skb;
5419 	u32 vlan_macip_lens = 0;
5420 	u32 type_tucmd = 0;
5421 
5422 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
5423 csum_failed:
5424 		if (!(first->tx_flags & IGB_TX_FLAGS_VLAN))
5425 			return;
5426 		goto no_csum;
5427 	}
5428 
5429 	switch (skb->csum_offset) {
5430 	case offsetof(struct tcphdr, check):
5431 		type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
5432 		/* fall through */
5433 	case offsetof(struct udphdr, check):
5434 		break;
5435 	case offsetof(struct sctphdr, checksum):
5436 		/* validate that this is actually an SCTP request */
5437 		if (((first->protocol == htons(ETH_P_IP)) &&
5438 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
5439 		    ((first->protocol == htons(ETH_P_IPV6)) &&
5440 		     igb_ipv6_csum_is_sctp(skb))) {
5441 			type_tucmd = E1000_ADVTXD_TUCMD_L4T_SCTP;
5442 			break;
5443 		}
5444 	default:
5445 		skb_checksum_help(skb);
5446 		goto csum_failed;
5447 	}
5448 
5449 	/* update TX checksum flag */
5450 	first->tx_flags |= IGB_TX_FLAGS_CSUM;
5451 	vlan_macip_lens = skb_checksum_start_offset(skb) -
5452 			  skb_network_offset(skb);
5453 no_csum:
5454 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
5455 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
5456 
5457 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, 0);
5458 }
5459 
5460 #define IGB_SET_FLAG(_input, _flag, _result) \
5461 	((_flag <= _result) ? \
5462 	 ((u32)(_input & _flag) * (_result / _flag)) : \
5463 	 ((u32)(_input & _flag) / (_flag / _result)))
5464 
5465 static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
5466 {
5467 	/* set type for advanced descriptor with frame checksum insertion */
5468 	u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
5469 		       E1000_ADVTXD_DCMD_DEXT |
5470 		       E1000_ADVTXD_DCMD_IFCS;
5471 
5472 	/* set HW vlan bit if vlan is present */
5473 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
5474 				 (E1000_ADVTXD_DCMD_VLE));
5475 
5476 	/* set segmentation bits for TSO */
5477 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
5478 				 (E1000_ADVTXD_DCMD_TSE));
5479 
5480 	/* set timestamp bit if present */
5481 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
5482 				 (E1000_ADVTXD_MAC_TSTAMP));
5483 
5484 	/* insert frame checksum */
5485 	cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
5486 
5487 	return cmd_type;
5488 }
5489 
5490 static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
5491 				 union e1000_adv_tx_desc *tx_desc,
5492 				 u32 tx_flags, unsigned int paylen)
5493 {
5494 	u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
5495 
5496 	/* 82575 requires a unique index per ring */
5497 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5498 		olinfo_status |= tx_ring->reg_idx << 4;
5499 
5500 	/* insert L4 checksum */
5501 	olinfo_status |= IGB_SET_FLAG(tx_flags,
5502 				      IGB_TX_FLAGS_CSUM,
5503 				      (E1000_TXD_POPTS_TXSM << 8));
5504 
5505 	/* insert IPv4 checksum */
5506 	olinfo_status |= IGB_SET_FLAG(tx_flags,
5507 				      IGB_TX_FLAGS_IPV4,
5508 				      (E1000_TXD_POPTS_IXSM << 8));
5509 
5510 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
5511 }
5512 
5513 static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
5514 {
5515 	struct net_device *netdev = tx_ring->netdev;
5516 
5517 	netif_stop_subqueue(netdev, tx_ring->queue_index);
5518 
5519 	/* Herbert's original patch had:
5520 	 *  smp_mb__after_netif_stop_queue();
5521 	 * but since that doesn't exist yet, just open code it.
5522 	 */
5523 	smp_mb();
5524 
5525 	/* We need to check again in a case another CPU has just
5526 	 * made room available.
5527 	 */
5528 	if (igb_desc_unused(tx_ring) < size)
5529 		return -EBUSY;
5530 
5531 	/* A reprieve! */
5532 	netif_wake_subqueue(netdev, tx_ring->queue_index);
5533 
5534 	u64_stats_update_begin(&tx_ring->tx_syncp2);
5535 	tx_ring->tx_stats.restart_queue2++;
5536 	u64_stats_update_end(&tx_ring->tx_syncp2);
5537 
5538 	return 0;
5539 }
5540 
5541 static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
5542 {
5543 	if (igb_desc_unused(tx_ring) >= size)
5544 		return 0;
5545 	return __igb_maybe_stop_tx(tx_ring, size);
5546 }
5547 
5548 static int igb_tx_map(struct igb_ring *tx_ring,
5549 		      struct igb_tx_buffer *first,
5550 		      const u8 hdr_len)
5551 {
5552 	struct sk_buff *skb = first->skb;
5553 	struct igb_tx_buffer *tx_buffer;
5554 	union e1000_adv_tx_desc *tx_desc;
5555 	struct skb_frag_struct *frag;
5556 	dma_addr_t dma;
5557 	unsigned int data_len, size;
5558 	u32 tx_flags = first->tx_flags;
5559 	u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
5560 	u16 i = tx_ring->next_to_use;
5561 
5562 	tx_desc = IGB_TX_DESC(tx_ring, i);
5563 
5564 	igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
5565 
5566 	size = skb_headlen(skb);
5567 	data_len = skb->data_len;
5568 
5569 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
5570 
5571 	tx_buffer = first;
5572 
5573 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
5574 		if (dma_mapping_error(tx_ring->dev, dma))
5575 			goto dma_error;
5576 
5577 		/* record length, and DMA address */
5578 		dma_unmap_len_set(tx_buffer, len, size);
5579 		dma_unmap_addr_set(tx_buffer, dma, dma);
5580 
5581 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
5582 
5583 		while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
5584 			tx_desc->read.cmd_type_len =
5585 				cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
5586 
5587 			i++;
5588 			tx_desc++;
5589 			if (i == tx_ring->count) {
5590 				tx_desc = IGB_TX_DESC(tx_ring, 0);
5591 				i = 0;
5592 			}
5593 			tx_desc->read.olinfo_status = 0;
5594 
5595 			dma += IGB_MAX_DATA_PER_TXD;
5596 			size -= IGB_MAX_DATA_PER_TXD;
5597 
5598 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
5599 		}
5600 
5601 		if (likely(!data_len))
5602 			break;
5603 
5604 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
5605 
5606 		i++;
5607 		tx_desc++;
5608 		if (i == tx_ring->count) {
5609 			tx_desc = IGB_TX_DESC(tx_ring, 0);
5610 			i = 0;
5611 		}
5612 		tx_desc->read.olinfo_status = 0;
5613 
5614 		size = skb_frag_size(frag);
5615 		data_len -= size;
5616 
5617 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
5618 				       size, DMA_TO_DEVICE);
5619 
5620 		tx_buffer = &tx_ring->tx_buffer_info[i];
5621 	}
5622 
5623 	/* write last descriptor with RS and EOP bits */
5624 	cmd_type |= size | IGB_TXD_DCMD;
5625 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
5626 
5627 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
5628 
5629 	/* set the timestamp */
5630 	first->time_stamp = jiffies;
5631 
5632 	/* Force memory writes to complete before letting h/w know there
5633 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
5634 	 * memory model archs, such as IA-64).
5635 	 *
5636 	 * We also need this memory barrier to make certain all of the
5637 	 * status bits have been updated before next_to_watch is written.
5638 	 */
5639 	wmb();
5640 
5641 	/* set next_to_watch value indicating a packet is present */
5642 	first->next_to_watch = tx_desc;
5643 
5644 	i++;
5645 	if (i == tx_ring->count)
5646 		i = 0;
5647 
5648 	tx_ring->next_to_use = i;
5649 
5650 	/* Make sure there is space in the ring for the next send. */
5651 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
5652 
5653 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
5654 		writel(i, tx_ring->tail);
5655 
5656 		/* we need this if more than one processor can write to our tail
5657 		 * at a time, it synchronizes IO on IA64/Altix systems
5658 		 */
5659 		mmiowb();
5660 	}
5661 	return 0;
5662 
5663 dma_error:
5664 	dev_err(tx_ring->dev, "TX DMA map failed\n");
5665 	tx_buffer = &tx_ring->tx_buffer_info[i];
5666 
5667 	/* clear dma mappings for failed tx_buffer_info map */
5668 	while (tx_buffer != first) {
5669 		if (dma_unmap_len(tx_buffer, len))
5670 			dma_unmap_page(tx_ring->dev,
5671 				       dma_unmap_addr(tx_buffer, dma),
5672 				       dma_unmap_len(tx_buffer, len),
5673 				       DMA_TO_DEVICE);
5674 		dma_unmap_len_set(tx_buffer, len, 0);
5675 
5676 		if (i-- == 0)
5677 			i += tx_ring->count;
5678 		tx_buffer = &tx_ring->tx_buffer_info[i];
5679 	}
5680 
5681 	if (dma_unmap_len(tx_buffer, len))
5682 		dma_unmap_single(tx_ring->dev,
5683 				 dma_unmap_addr(tx_buffer, dma),
5684 				 dma_unmap_len(tx_buffer, len),
5685 				 DMA_TO_DEVICE);
5686 	dma_unmap_len_set(tx_buffer, len, 0);
5687 
5688 	dev_kfree_skb_any(tx_buffer->skb);
5689 	tx_buffer->skb = NULL;
5690 
5691 	tx_ring->next_to_use = i;
5692 
5693 	return -1;
5694 }
5695 
5696 netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
5697 				struct igb_ring *tx_ring)
5698 {
5699 	struct igb_tx_buffer *first;
5700 	int tso;
5701 	u32 tx_flags = 0;
5702 	unsigned short f;
5703 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
5704 	__be16 protocol = vlan_get_protocol(skb);
5705 	u8 hdr_len = 0;
5706 
5707 	/* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
5708 	 *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
5709 	 *       + 2 desc gap to keep tail from touching head,
5710 	 *       + 1 desc for context descriptor,
5711 	 * otherwise try next time
5712 	 */
5713 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
5714 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
5715 
5716 	if (igb_maybe_stop_tx(tx_ring, count + 3)) {
5717 		/* this is a hard error */
5718 		return NETDEV_TX_BUSY;
5719 	}
5720 
5721 	/* record the location of the first descriptor for this packet */
5722 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
5723 	first->skb = skb;
5724 	first->bytecount = skb->len;
5725 	first->gso_segs = 1;
5726 
5727 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
5728 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
5729 
5730 		if (adapter->tstamp_config.tx_type & HWTSTAMP_TX_ON &&
5731 		    !test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
5732 					   &adapter->state)) {
5733 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5734 			tx_flags |= IGB_TX_FLAGS_TSTAMP;
5735 
5736 			adapter->ptp_tx_skb = skb_get(skb);
5737 			adapter->ptp_tx_start = jiffies;
5738 			if (adapter->hw.mac.type == e1000_82576)
5739 				schedule_work(&adapter->ptp_tx_work);
5740 		} else {
5741 			adapter->tx_hwtstamp_skipped++;
5742 		}
5743 	}
5744 
5745 	skb_tx_timestamp(skb);
5746 
5747 	if (skb_vlan_tag_present(skb)) {
5748 		tx_flags |= IGB_TX_FLAGS_VLAN;
5749 		tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
5750 	}
5751 
5752 	/* record initial flags and protocol */
5753 	first->tx_flags = tx_flags;
5754 	first->protocol = protocol;
5755 
5756 	tso = igb_tso(tx_ring, first, &hdr_len);
5757 	if (tso < 0)
5758 		goto out_drop;
5759 	else if (!tso)
5760 		igb_tx_csum(tx_ring, first);
5761 
5762 	if (igb_tx_map(tx_ring, first, hdr_len))
5763 		goto cleanup_tx_tstamp;
5764 
5765 	return NETDEV_TX_OK;
5766 
5767 out_drop:
5768 	dev_kfree_skb_any(first->skb);
5769 	first->skb = NULL;
5770 cleanup_tx_tstamp:
5771 	if (unlikely(tx_flags & IGB_TX_FLAGS_TSTAMP)) {
5772 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
5773 
5774 		dev_kfree_skb_any(adapter->ptp_tx_skb);
5775 		adapter->ptp_tx_skb = NULL;
5776 		if (adapter->hw.mac.type == e1000_82576)
5777 			cancel_work_sync(&adapter->ptp_tx_work);
5778 		clear_bit_unlock(__IGB_PTP_TX_IN_PROGRESS, &adapter->state);
5779 	}
5780 
5781 	return NETDEV_TX_OK;
5782 }
5783 
5784 static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
5785 						    struct sk_buff *skb)
5786 {
5787 	unsigned int r_idx = skb->queue_mapping;
5788 
5789 	if (r_idx >= adapter->num_tx_queues)
5790 		r_idx = r_idx % adapter->num_tx_queues;
5791 
5792 	return adapter->tx_ring[r_idx];
5793 }
5794 
5795 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
5796 				  struct net_device *netdev)
5797 {
5798 	struct igb_adapter *adapter = netdev_priv(netdev);
5799 
5800 	/* The minimum packet size with TCTL.PSP set is 17 so pad the skb
5801 	 * in order to meet this minimum size requirement.
5802 	 */
5803 	if (skb_put_padto(skb, 17))
5804 		return NETDEV_TX_OK;
5805 
5806 	return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
5807 }
5808 
5809 /**
5810  *  igb_tx_timeout - Respond to a Tx Hang
5811  *  @netdev: network interface device structure
5812  **/
5813 static void igb_tx_timeout(struct net_device *netdev)
5814 {
5815 	struct igb_adapter *adapter = netdev_priv(netdev);
5816 	struct e1000_hw *hw = &adapter->hw;
5817 
5818 	/* Do the reset outside of interrupt context */
5819 	adapter->tx_timeout_count++;
5820 
5821 	if (hw->mac.type >= e1000_82580)
5822 		hw->dev_spec._82575.global_device_reset = true;
5823 
5824 	schedule_work(&adapter->reset_task);
5825 	wr32(E1000_EICS,
5826 	     (adapter->eims_enable_mask & ~adapter->eims_other));
5827 }
5828 
5829 static void igb_reset_task(struct work_struct *work)
5830 {
5831 	struct igb_adapter *adapter;
5832 	adapter = container_of(work, struct igb_adapter, reset_task);
5833 
5834 	igb_dump(adapter);
5835 	netdev_err(adapter->netdev, "Reset adapter\n");
5836 	igb_reinit_locked(adapter);
5837 }
5838 
5839 /**
5840  *  igb_get_stats64 - Get System Network Statistics
5841  *  @netdev: network interface device structure
5842  *  @stats: rtnl_link_stats64 pointer
5843  **/
5844 static void igb_get_stats64(struct net_device *netdev,
5845 			    struct rtnl_link_stats64 *stats)
5846 {
5847 	struct igb_adapter *adapter = netdev_priv(netdev);
5848 
5849 	spin_lock(&adapter->stats64_lock);
5850 	igb_update_stats(adapter);
5851 	memcpy(stats, &adapter->stats64, sizeof(*stats));
5852 	spin_unlock(&adapter->stats64_lock);
5853 }
5854 
5855 /**
5856  *  igb_change_mtu - Change the Maximum Transfer Unit
5857  *  @netdev: network interface device structure
5858  *  @new_mtu: new value for maximum frame size
5859  *
5860  *  Returns 0 on success, negative on failure
5861  **/
5862 static int igb_change_mtu(struct net_device *netdev, int new_mtu)
5863 {
5864 	struct igb_adapter *adapter = netdev_priv(netdev);
5865 	struct pci_dev *pdev = adapter->pdev;
5866 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
5867 
5868 	/* adjust max frame to be at least the size of a standard frame */
5869 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
5870 		max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
5871 
5872 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
5873 		usleep_range(1000, 2000);
5874 
5875 	/* igb_down has a dependency on max_frame_size */
5876 	adapter->max_frame_size = max_frame;
5877 
5878 	if (netif_running(netdev))
5879 		igb_down(adapter);
5880 
5881 	dev_info(&pdev->dev, "changing MTU from %d to %d\n",
5882 		 netdev->mtu, new_mtu);
5883 	netdev->mtu = new_mtu;
5884 
5885 	if (netif_running(netdev))
5886 		igb_up(adapter);
5887 	else
5888 		igb_reset(adapter);
5889 
5890 	clear_bit(__IGB_RESETTING, &adapter->state);
5891 
5892 	return 0;
5893 }
5894 
5895 /**
5896  *  igb_update_stats - Update the board statistics counters
5897  *  @adapter: board private structure
5898  **/
5899 void igb_update_stats(struct igb_adapter *adapter)
5900 {
5901 	struct rtnl_link_stats64 *net_stats = &adapter->stats64;
5902 	struct e1000_hw *hw = &adapter->hw;
5903 	struct pci_dev *pdev = adapter->pdev;
5904 	u32 reg, mpc;
5905 	int i;
5906 	u64 bytes, packets;
5907 	unsigned int start;
5908 	u64 _bytes, _packets;
5909 
5910 	/* Prevent stats update while adapter is being reset, or if the pci
5911 	 * connection is down.
5912 	 */
5913 	if (adapter->link_speed == 0)
5914 		return;
5915 	if (pci_channel_offline(pdev))
5916 		return;
5917 
5918 	bytes = 0;
5919 	packets = 0;
5920 
5921 	rcu_read_lock();
5922 	for (i = 0; i < adapter->num_rx_queues; i++) {
5923 		struct igb_ring *ring = adapter->rx_ring[i];
5924 		u32 rqdpc = rd32(E1000_RQDPC(i));
5925 		if (hw->mac.type >= e1000_i210)
5926 			wr32(E1000_RQDPC(i), 0);
5927 
5928 		if (rqdpc) {
5929 			ring->rx_stats.drops += rqdpc;
5930 			net_stats->rx_fifo_errors += rqdpc;
5931 		}
5932 
5933 		do {
5934 			start = u64_stats_fetch_begin_irq(&ring->rx_syncp);
5935 			_bytes = ring->rx_stats.bytes;
5936 			_packets = ring->rx_stats.packets;
5937 		} while (u64_stats_fetch_retry_irq(&ring->rx_syncp, start));
5938 		bytes += _bytes;
5939 		packets += _packets;
5940 	}
5941 
5942 	net_stats->rx_bytes = bytes;
5943 	net_stats->rx_packets = packets;
5944 
5945 	bytes = 0;
5946 	packets = 0;
5947 	for (i = 0; i < adapter->num_tx_queues; i++) {
5948 		struct igb_ring *ring = adapter->tx_ring[i];
5949 		do {
5950 			start = u64_stats_fetch_begin_irq(&ring->tx_syncp);
5951 			_bytes = ring->tx_stats.bytes;
5952 			_packets = ring->tx_stats.packets;
5953 		} while (u64_stats_fetch_retry_irq(&ring->tx_syncp, start));
5954 		bytes += _bytes;
5955 		packets += _packets;
5956 	}
5957 	net_stats->tx_bytes = bytes;
5958 	net_stats->tx_packets = packets;
5959 	rcu_read_unlock();
5960 
5961 	/* read stats registers */
5962 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
5963 	adapter->stats.gprc += rd32(E1000_GPRC);
5964 	adapter->stats.gorc += rd32(E1000_GORCL);
5965 	rd32(E1000_GORCH); /* clear GORCL */
5966 	adapter->stats.bprc += rd32(E1000_BPRC);
5967 	adapter->stats.mprc += rd32(E1000_MPRC);
5968 	adapter->stats.roc += rd32(E1000_ROC);
5969 
5970 	adapter->stats.prc64 += rd32(E1000_PRC64);
5971 	adapter->stats.prc127 += rd32(E1000_PRC127);
5972 	adapter->stats.prc255 += rd32(E1000_PRC255);
5973 	adapter->stats.prc511 += rd32(E1000_PRC511);
5974 	adapter->stats.prc1023 += rd32(E1000_PRC1023);
5975 	adapter->stats.prc1522 += rd32(E1000_PRC1522);
5976 	adapter->stats.symerrs += rd32(E1000_SYMERRS);
5977 	adapter->stats.sec += rd32(E1000_SEC);
5978 
5979 	mpc = rd32(E1000_MPC);
5980 	adapter->stats.mpc += mpc;
5981 	net_stats->rx_fifo_errors += mpc;
5982 	adapter->stats.scc += rd32(E1000_SCC);
5983 	adapter->stats.ecol += rd32(E1000_ECOL);
5984 	adapter->stats.mcc += rd32(E1000_MCC);
5985 	adapter->stats.latecol += rd32(E1000_LATECOL);
5986 	adapter->stats.dc += rd32(E1000_DC);
5987 	adapter->stats.rlec += rd32(E1000_RLEC);
5988 	adapter->stats.xonrxc += rd32(E1000_XONRXC);
5989 	adapter->stats.xontxc += rd32(E1000_XONTXC);
5990 	adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
5991 	adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
5992 	adapter->stats.fcruc += rd32(E1000_FCRUC);
5993 	adapter->stats.gptc += rd32(E1000_GPTC);
5994 	adapter->stats.gotc += rd32(E1000_GOTCL);
5995 	rd32(E1000_GOTCH); /* clear GOTCL */
5996 	adapter->stats.rnbc += rd32(E1000_RNBC);
5997 	adapter->stats.ruc += rd32(E1000_RUC);
5998 	adapter->stats.rfc += rd32(E1000_RFC);
5999 	adapter->stats.rjc += rd32(E1000_RJC);
6000 	adapter->stats.tor += rd32(E1000_TORH);
6001 	adapter->stats.tot += rd32(E1000_TOTH);
6002 	adapter->stats.tpr += rd32(E1000_TPR);
6003 
6004 	adapter->stats.ptc64 += rd32(E1000_PTC64);
6005 	adapter->stats.ptc127 += rd32(E1000_PTC127);
6006 	adapter->stats.ptc255 += rd32(E1000_PTC255);
6007 	adapter->stats.ptc511 += rd32(E1000_PTC511);
6008 	adapter->stats.ptc1023 += rd32(E1000_PTC1023);
6009 	adapter->stats.ptc1522 += rd32(E1000_PTC1522);
6010 
6011 	adapter->stats.mptc += rd32(E1000_MPTC);
6012 	adapter->stats.bptc += rd32(E1000_BPTC);
6013 
6014 	adapter->stats.tpt += rd32(E1000_TPT);
6015 	adapter->stats.colc += rd32(E1000_COLC);
6016 
6017 	adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
6018 	/* read internal phy specific stats */
6019 	reg = rd32(E1000_CTRL_EXT);
6020 	if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
6021 		adapter->stats.rxerrc += rd32(E1000_RXERRC);
6022 
6023 		/* this stat has invalid values on i210/i211 */
6024 		if ((hw->mac.type != e1000_i210) &&
6025 		    (hw->mac.type != e1000_i211))
6026 			adapter->stats.tncrs += rd32(E1000_TNCRS);
6027 	}
6028 
6029 	adapter->stats.tsctc += rd32(E1000_TSCTC);
6030 	adapter->stats.tsctfc += rd32(E1000_TSCTFC);
6031 
6032 	adapter->stats.iac += rd32(E1000_IAC);
6033 	adapter->stats.icrxoc += rd32(E1000_ICRXOC);
6034 	adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
6035 	adapter->stats.icrxatc += rd32(E1000_ICRXATC);
6036 	adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
6037 	adapter->stats.ictxatc += rd32(E1000_ICTXATC);
6038 	adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
6039 	adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
6040 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
6041 
6042 	/* Fill out the OS statistics structure */
6043 	net_stats->multicast = adapter->stats.mprc;
6044 	net_stats->collisions = adapter->stats.colc;
6045 
6046 	/* Rx Errors */
6047 
6048 	/* RLEC on some newer hardware can be incorrect so build
6049 	 * our own version based on RUC and ROC
6050 	 */
6051 	net_stats->rx_errors = adapter->stats.rxerrc +
6052 		adapter->stats.crcerrs + adapter->stats.algnerrc +
6053 		adapter->stats.ruc + adapter->stats.roc +
6054 		adapter->stats.cexterr;
6055 	net_stats->rx_length_errors = adapter->stats.ruc +
6056 				      adapter->stats.roc;
6057 	net_stats->rx_crc_errors = adapter->stats.crcerrs;
6058 	net_stats->rx_frame_errors = adapter->stats.algnerrc;
6059 	net_stats->rx_missed_errors = adapter->stats.mpc;
6060 
6061 	/* Tx Errors */
6062 	net_stats->tx_errors = adapter->stats.ecol +
6063 			       adapter->stats.latecol;
6064 	net_stats->tx_aborted_errors = adapter->stats.ecol;
6065 	net_stats->tx_window_errors = adapter->stats.latecol;
6066 	net_stats->tx_carrier_errors = adapter->stats.tncrs;
6067 
6068 	/* Tx Dropped needs to be maintained elsewhere */
6069 
6070 	/* Management Stats */
6071 	adapter->stats.mgptc += rd32(E1000_MGTPTC);
6072 	adapter->stats.mgprc += rd32(E1000_MGTPRC);
6073 	adapter->stats.mgpdc += rd32(E1000_MGTPDC);
6074 
6075 	/* OS2BMC Stats */
6076 	reg = rd32(E1000_MANC);
6077 	if (reg & E1000_MANC_EN_BMC2OS) {
6078 		adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
6079 		adapter->stats.o2bspc += rd32(E1000_O2BSPC);
6080 		adapter->stats.b2ospc += rd32(E1000_B2OSPC);
6081 		adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
6082 	}
6083 }
6084 
6085 static void igb_tsync_interrupt(struct igb_adapter *adapter)
6086 {
6087 	struct e1000_hw *hw = &adapter->hw;
6088 	struct ptp_clock_event event;
6089 	struct timespec64 ts;
6090 	u32 ack = 0, tsauxc, sec, nsec, tsicr = rd32(E1000_TSICR);
6091 
6092 	if (tsicr & TSINTR_SYS_WRAP) {
6093 		event.type = PTP_CLOCK_PPS;
6094 		if (adapter->ptp_caps.pps)
6095 			ptp_clock_event(adapter->ptp_clock, &event);
6096 		ack |= TSINTR_SYS_WRAP;
6097 	}
6098 
6099 	if (tsicr & E1000_TSICR_TXTS) {
6100 		/* retrieve hardware timestamp */
6101 		schedule_work(&adapter->ptp_tx_work);
6102 		ack |= E1000_TSICR_TXTS;
6103 	}
6104 
6105 	if (tsicr & TSINTR_TT0) {
6106 		spin_lock(&adapter->tmreg_lock);
6107 		ts = timespec64_add(adapter->perout[0].start,
6108 				    adapter->perout[0].period);
6109 		/* u32 conversion of tv_sec is safe until y2106 */
6110 		wr32(E1000_TRGTTIML0, ts.tv_nsec);
6111 		wr32(E1000_TRGTTIMH0, (u32)ts.tv_sec);
6112 		tsauxc = rd32(E1000_TSAUXC);
6113 		tsauxc |= TSAUXC_EN_TT0;
6114 		wr32(E1000_TSAUXC, tsauxc);
6115 		adapter->perout[0].start = ts;
6116 		spin_unlock(&adapter->tmreg_lock);
6117 		ack |= TSINTR_TT0;
6118 	}
6119 
6120 	if (tsicr & TSINTR_TT1) {
6121 		spin_lock(&adapter->tmreg_lock);
6122 		ts = timespec64_add(adapter->perout[1].start,
6123 				    adapter->perout[1].period);
6124 		wr32(E1000_TRGTTIML1, ts.tv_nsec);
6125 		wr32(E1000_TRGTTIMH1, (u32)ts.tv_sec);
6126 		tsauxc = rd32(E1000_TSAUXC);
6127 		tsauxc |= TSAUXC_EN_TT1;
6128 		wr32(E1000_TSAUXC, tsauxc);
6129 		adapter->perout[1].start = ts;
6130 		spin_unlock(&adapter->tmreg_lock);
6131 		ack |= TSINTR_TT1;
6132 	}
6133 
6134 	if (tsicr & TSINTR_AUTT0) {
6135 		nsec = rd32(E1000_AUXSTMPL0);
6136 		sec  = rd32(E1000_AUXSTMPH0);
6137 		event.type = PTP_CLOCK_EXTTS;
6138 		event.index = 0;
6139 		event.timestamp = sec * 1000000000ULL + nsec;
6140 		ptp_clock_event(adapter->ptp_clock, &event);
6141 		ack |= TSINTR_AUTT0;
6142 	}
6143 
6144 	if (tsicr & TSINTR_AUTT1) {
6145 		nsec = rd32(E1000_AUXSTMPL1);
6146 		sec  = rd32(E1000_AUXSTMPH1);
6147 		event.type = PTP_CLOCK_EXTTS;
6148 		event.index = 1;
6149 		event.timestamp = sec * 1000000000ULL + nsec;
6150 		ptp_clock_event(adapter->ptp_clock, &event);
6151 		ack |= TSINTR_AUTT1;
6152 	}
6153 
6154 	/* acknowledge the interrupts */
6155 	wr32(E1000_TSICR, ack);
6156 }
6157 
6158 static irqreturn_t igb_msix_other(int irq, void *data)
6159 {
6160 	struct igb_adapter *adapter = data;
6161 	struct e1000_hw *hw = &adapter->hw;
6162 	u32 icr = rd32(E1000_ICR);
6163 	/* reading ICR causes bit 31 of EICR to be cleared */
6164 
6165 	if (icr & E1000_ICR_DRSTA)
6166 		schedule_work(&adapter->reset_task);
6167 
6168 	if (icr & E1000_ICR_DOUTSYNC) {
6169 		/* HW is reporting DMA is out of sync */
6170 		adapter->stats.doosync++;
6171 		/* The DMA Out of Sync is also indication of a spoof event
6172 		 * in IOV mode. Check the Wrong VM Behavior register to
6173 		 * see if it is really a spoof event.
6174 		 */
6175 		igb_check_wvbr(adapter);
6176 	}
6177 
6178 	/* Check for a mailbox event */
6179 	if (icr & E1000_ICR_VMMB)
6180 		igb_msg_task(adapter);
6181 
6182 	if (icr & E1000_ICR_LSC) {
6183 		hw->mac.get_link_status = 1;
6184 		/* guard against interrupt when we're going down */
6185 		if (!test_bit(__IGB_DOWN, &adapter->state))
6186 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6187 	}
6188 
6189 	if (icr & E1000_ICR_TS)
6190 		igb_tsync_interrupt(adapter);
6191 
6192 	wr32(E1000_EIMS, adapter->eims_other);
6193 
6194 	return IRQ_HANDLED;
6195 }
6196 
6197 static void igb_write_itr(struct igb_q_vector *q_vector)
6198 {
6199 	struct igb_adapter *adapter = q_vector->adapter;
6200 	u32 itr_val = q_vector->itr_val & 0x7FFC;
6201 
6202 	if (!q_vector->set_itr)
6203 		return;
6204 
6205 	if (!itr_val)
6206 		itr_val = 0x4;
6207 
6208 	if (adapter->hw.mac.type == e1000_82575)
6209 		itr_val |= itr_val << 16;
6210 	else
6211 		itr_val |= E1000_EITR_CNT_IGNR;
6212 
6213 	writel(itr_val, q_vector->itr_register);
6214 	q_vector->set_itr = 0;
6215 }
6216 
6217 static irqreturn_t igb_msix_ring(int irq, void *data)
6218 {
6219 	struct igb_q_vector *q_vector = data;
6220 
6221 	/* Write the ITR value calculated from the previous interrupt. */
6222 	igb_write_itr(q_vector);
6223 
6224 	napi_schedule(&q_vector->napi);
6225 
6226 	return IRQ_HANDLED;
6227 }
6228 
6229 #ifdef CONFIG_IGB_DCA
6230 static void igb_update_tx_dca(struct igb_adapter *adapter,
6231 			      struct igb_ring *tx_ring,
6232 			      int cpu)
6233 {
6234 	struct e1000_hw *hw = &adapter->hw;
6235 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
6236 
6237 	if (hw->mac.type != e1000_82575)
6238 		txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
6239 
6240 	/* We can enable relaxed ordering for reads, but not writes when
6241 	 * DCA is enabled.  This is due to a known issue in some chipsets
6242 	 * which will cause the DCA tag to be cleared.
6243 	 */
6244 	txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
6245 		  E1000_DCA_TXCTRL_DATA_RRO_EN |
6246 		  E1000_DCA_TXCTRL_DESC_DCA_EN;
6247 
6248 	wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
6249 }
6250 
6251 static void igb_update_rx_dca(struct igb_adapter *adapter,
6252 			      struct igb_ring *rx_ring,
6253 			      int cpu)
6254 {
6255 	struct e1000_hw *hw = &adapter->hw;
6256 	u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
6257 
6258 	if (hw->mac.type != e1000_82575)
6259 		rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
6260 
6261 	/* We can enable relaxed ordering for reads, but not writes when
6262 	 * DCA is enabled.  This is due to a known issue in some chipsets
6263 	 * which will cause the DCA tag to be cleared.
6264 	 */
6265 	rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
6266 		  E1000_DCA_RXCTRL_DESC_DCA_EN;
6267 
6268 	wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
6269 }
6270 
6271 static void igb_update_dca(struct igb_q_vector *q_vector)
6272 {
6273 	struct igb_adapter *adapter = q_vector->adapter;
6274 	int cpu = get_cpu();
6275 
6276 	if (q_vector->cpu == cpu)
6277 		goto out_no_update;
6278 
6279 	if (q_vector->tx.ring)
6280 		igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
6281 
6282 	if (q_vector->rx.ring)
6283 		igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
6284 
6285 	q_vector->cpu = cpu;
6286 out_no_update:
6287 	put_cpu();
6288 }
6289 
6290 static void igb_setup_dca(struct igb_adapter *adapter)
6291 {
6292 	struct e1000_hw *hw = &adapter->hw;
6293 	int i;
6294 
6295 	if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
6296 		return;
6297 
6298 	/* Always use CB2 mode, difference is masked in the CB driver. */
6299 	wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
6300 
6301 	for (i = 0; i < adapter->num_q_vectors; i++) {
6302 		adapter->q_vector[i]->cpu = -1;
6303 		igb_update_dca(adapter->q_vector[i]);
6304 	}
6305 }
6306 
6307 static int __igb_notify_dca(struct device *dev, void *data)
6308 {
6309 	struct net_device *netdev = dev_get_drvdata(dev);
6310 	struct igb_adapter *adapter = netdev_priv(netdev);
6311 	struct pci_dev *pdev = adapter->pdev;
6312 	struct e1000_hw *hw = &adapter->hw;
6313 	unsigned long event = *(unsigned long *)data;
6314 
6315 	switch (event) {
6316 	case DCA_PROVIDER_ADD:
6317 		/* if already enabled, don't do it again */
6318 		if (adapter->flags & IGB_FLAG_DCA_ENABLED)
6319 			break;
6320 		if (dca_add_requester(dev) == 0) {
6321 			adapter->flags |= IGB_FLAG_DCA_ENABLED;
6322 			dev_info(&pdev->dev, "DCA enabled\n");
6323 			igb_setup_dca(adapter);
6324 			break;
6325 		}
6326 		/* Fall Through since DCA is disabled. */
6327 	case DCA_PROVIDER_REMOVE:
6328 		if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
6329 			/* without this a class_device is left
6330 			 * hanging around in the sysfs model
6331 			 */
6332 			dca_remove_requester(dev);
6333 			dev_info(&pdev->dev, "DCA disabled\n");
6334 			adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
6335 			wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
6336 		}
6337 		break;
6338 	}
6339 
6340 	return 0;
6341 }
6342 
6343 static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
6344 			  void *p)
6345 {
6346 	int ret_val;
6347 
6348 	ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
6349 					 __igb_notify_dca);
6350 
6351 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
6352 }
6353 #endif /* CONFIG_IGB_DCA */
6354 
6355 #ifdef CONFIG_PCI_IOV
6356 static int igb_vf_configure(struct igb_adapter *adapter, int vf)
6357 {
6358 	unsigned char mac_addr[ETH_ALEN];
6359 
6360 	eth_zero_addr(mac_addr);
6361 	igb_set_vf_mac(adapter, vf, mac_addr);
6362 
6363 	/* By default spoof check is enabled for all VFs */
6364 	adapter->vf_data[vf].spoofchk_enabled = true;
6365 
6366 	return 0;
6367 }
6368 
6369 #endif
6370 static void igb_ping_all_vfs(struct igb_adapter *adapter)
6371 {
6372 	struct e1000_hw *hw = &adapter->hw;
6373 	u32 ping;
6374 	int i;
6375 
6376 	for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
6377 		ping = E1000_PF_CONTROL_MSG;
6378 		if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
6379 			ping |= E1000_VT_MSGTYPE_CTS;
6380 		igb_write_mbx(hw, &ping, 1, i);
6381 	}
6382 }
6383 
6384 static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
6385 {
6386 	struct e1000_hw *hw = &adapter->hw;
6387 	u32 vmolr = rd32(E1000_VMOLR(vf));
6388 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6389 
6390 	vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
6391 			    IGB_VF_FLAG_MULTI_PROMISC);
6392 	vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
6393 
6394 	if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
6395 		vmolr |= E1000_VMOLR_MPME;
6396 		vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
6397 		*msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
6398 	} else {
6399 		/* if we have hashes and we are clearing a multicast promisc
6400 		 * flag we need to write the hashes to the MTA as this step
6401 		 * was previously skipped
6402 		 */
6403 		if (vf_data->num_vf_mc_hashes > 30) {
6404 			vmolr |= E1000_VMOLR_MPME;
6405 		} else if (vf_data->num_vf_mc_hashes) {
6406 			int j;
6407 
6408 			vmolr |= E1000_VMOLR_ROMPE;
6409 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
6410 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
6411 		}
6412 	}
6413 
6414 	wr32(E1000_VMOLR(vf), vmolr);
6415 
6416 	/* there are flags left unprocessed, likely not supported */
6417 	if (*msgbuf & E1000_VT_MSGINFO_MASK)
6418 		return -EINVAL;
6419 
6420 	return 0;
6421 }
6422 
6423 static int igb_set_vf_multicasts(struct igb_adapter *adapter,
6424 				  u32 *msgbuf, u32 vf)
6425 {
6426 	int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
6427 	u16 *hash_list = (u16 *)&msgbuf[1];
6428 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6429 	int i;
6430 
6431 	/* salt away the number of multicast addresses assigned
6432 	 * to this VF for later use to restore when the PF multi cast
6433 	 * list changes
6434 	 */
6435 	vf_data->num_vf_mc_hashes = n;
6436 
6437 	/* only up to 30 hash values supported */
6438 	if (n > 30)
6439 		n = 30;
6440 
6441 	/* store the hashes for later use */
6442 	for (i = 0; i < n; i++)
6443 		vf_data->vf_mc_hashes[i] = hash_list[i];
6444 
6445 	/* Flush and reset the mta with the new values */
6446 	igb_set_rx_mode(adapter->netdev);
6447 
6448 	return 0;
6449 }
6450 
6451 static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
6452 {
6453 	struct e1000_hw *hw = &adapter->hw;
6454 	struct vf_data_storage *vf_data;
6455 	int i, j;
6456 
6457 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
6458 		u32 vmolr = rd32(E1000_VMOLR(i));
6459 
6460 		vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
6461 
6462 		vf_data = &adapter->vf_data[i];
6463 
6464 		if ((vf_data->num_vf_mc_hashes > 30) ||
6465 		    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
6466 			vmolr |= E1000_VMOLR_MPME;
6467 		} else if (vf_data->num_vf_mc_hashes) {
6468 			vmolr |= E1000_VMOLR_ROMPE;
6469 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
6470 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
6471 		}
6472 		wr32(E1000_VMOLR(i), vmolr);
6473 	}
6474 }
6475 
6476 static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
6477 {
6478 	struct e1000_hw *hw = &adapter->hw;
6479 	u32 pool_mask, vlvf_mask, i;
6480 
6481 	/* create mask for VF and other pools */
6482 	pool_mask = E1000_VLVF_POOLSEL_MASK;
6483 	vlvf_mask = BIT(E1000_VLVF_POOLSEL_SHIFT + vf);
6484 
6485 	/* drop PF from pool bits */
6486 	pool_mask &= ~BIT(E1000_VLVF_POOLSEL_SHIFT +
6487 			     adapter->vfs_allocated_count);
6488 
6489 	/* Find the vlan filter for this id */
6490 	for (i = E1000_VLVF_ARRAY_SIZE; i--;) {
6491 		u32 vlvf = rd32(E1000_VLVF(i));
6492 		u32 vfta_mask, vid, vfta;
6493 
6494 		/* remove the vf from the pool */
6495 		if (!(vlvf & vlvf_mask))
6496 			continue;
6497 
6498 		/* clear out bit from VLVF */
6499 		vlvf ^= vlvf_mask;
6500 
6501 		/* if other pools are present, just remove ourselves */
6502 		if (vlvf & pool_mask)
6503 			goto update_vlvfb;
6504 
6505 		/* if PF is present, leave VFTA */
6506 		if (vlvf & E1000_VLVF_POOLSEL_MASK)
6507 			goto update_vlvf;
6508 
6509 		vid = vlvf & E1000_VLVF_VLANID_MASK;
6510 		vfta_mask = BIT(vid % 32);
6511 
6512 		/* clear bit from VFTA */
6513 		vfta = adapter->shadow_vfta[vid / 32];
6514 		if (vfta & vfta_mask)
6515 			hw->mac.ops.write_vfta(hw, vid / 32, vfta ^ vfta_mask);
6516 update_vlvf:
6517 		/* clear pool selection enable */
6518 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
6519 			vlvf &= E1000_VLVF_POOLSEL_MASK;
6520 		else
6521 			vlvf = 0;
6522 update_vlvfb:
6523 		/* clear pool bits */
6524 		wr32(E1000_VLVF(i), vlvf);
6525 	}
6526 }
6527 
6528 static int igb_find_vlvf_entry(struct e1000_hw *hw, u32 vlan)
6529 {
6530 	u32 vlvf;
6531 	int idx;
6532 
6533 	/* short cut the special case */
6534 	if (vlan == 0)
6535 		return 0;
6536 
6537 	/* Search for the VLAN id in the VLVF entries */
6538 	for (idx = E1000_VLVF_ARRAY_SIZE; --idx;) {
6539 		vlvf = rd32(E1000_VLVF(idx));
6540 		if ((vlvf & VLAN_VID_MASK) == vlan)
6541 			break;
6542 	}
6543 
6544 	return idx;
6545 }
6546 
6547 static void igb_update_pf_vlvf(struct igb_adapter *adapter, u32 vid)
6548 {
6549 	struct e1000_hw *hw = &adapter->hw;
6550 	u32 bits, pf_id;
6551 	int idx;
6552 
6553 	idx = igb_find_vlvf_entry(hw, vid);
6554 	if (!idx)
6555 		return;
6556 
6557 	/* See if any other pools are set for this VLAN filter
6558 	 * entry other than the PF.
6559 	 */
6560 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
6561 	bits = ~BIT(pf_id) & E1000_VLVF_POOLSEL_MASK;
6562 	bits &= rd32(E1000_VLVF(idx));
6563 
6564 	/* Disable the filter so this falls into the default pool. */
6565 	if (!bits) {
6566 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
6567 			wr32(E1000_VLVF(idx), BIT(pf_id));
6568 		else
6569 			wr32(E1000_VLVF(idx), 0);
6570 	}
6571 }
6572 
6573 static s32 igb_set_vf_vlan(struct igb_adapter *adapter, u32 vid,
6574 			   bool add, u32 vf)
6575 {
6576 	int pf_id = adapter->vfs_allocated_count;
6577 	struct e1000_hw *hw = &adapter->hw;
6578 	int err;
6579 
6580 	/* If VLAN overlaps with one the PF is currently monitoring make
6581 	 * sure that we are able to allocate a VLVF entry.  This may be
6582 	 * redundant but it guarantees PF will maintain visibility to
6583 	 * the VLAN.
6584 	 */
6585 	if (add && test_bit(vid, adapter->active_vlans)) {
6586 		err = igb_vfta_set(hw, vid, pf_id, true, false);
6587 		if (err)
6588 			return err;
6589 	}
6590 
6591 	err = igb_vfta_set(hw, vid, vf, add, false);
6592 
6593 	if (add && !err)
6594 		return err;
6595 
6596 	/* If we failed to add the VF VLAN or we are removing the VF VLAN
6597 	 * we may need to drop the PF pool bit in order to allow us to free
6598 	 * up the VLVF resources.
6599 	 */
6600 	if (test_bit(vid, adapter->active_vlans) ||
6601 	    (adapter->flags & IGB_FLAG_VLAN_PROMISC))
6602 		igb_update_pf_vlvf(adapter, vid);
6603 
6604 	return err;
6605 }
6606 
6607 static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
6608 {
6609 	struct e1000_hw *hw = &adapter->hw;
6610 
6611 	if (vid)
6612 		wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
6613 	else
6614 		wr32(E1000_VMVIR(vf), 0);
6615 }
6616 
6617 static int igb_enable_port_vlan(struct igb_adapter *adapter, int vf,
6618 				u16 vlan, u8 qos)
6619 {
6620 	int err;
6621 
6622 	err = igb_set_vf_vlan(adapter, vlan, true, vf);
6623 	if (err)
6624 		return err;
6625 
6626 	igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
6627 	igb_set_vmolr(adapter, vf, !vlan);
6628 
6629 	/* revoke access to previous VLAN */
6630 	if (vlan != adapter->vf_data[vf].pf_vlan)
6631 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
6632 				false, vf);
6633 
6634 	adapter->vf_data[vf].pf_vlan = vlan;
6635 	adapter->vf_data[vf].pf_qos = qos;
6636 	igb_set_vf_vlan_strip(adapter, vf, true);
6637 	dev_info(&adapter->pdev->dev,
6638 		 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
6639 	if (test_bit(__IGB_DOWN, &adapter->state)) {
6640 		dev_warn(&adapter->pdev->dev,
6641 			 "The VF VLAN has been set, but the PF device is not up.\n");
6642 		dev_warn(&adapter->pdev->dev,
6643 			 "Bring the PF device up before attempting to use the VF device.\n");
6644 	}
6645 
6646 	return err;
6647 }
6648 
6649 static int igb_disable_port_vlan(struct igb_adapter *adapter, int vf)
6650 {
6651 	/* Restore tagless access via VLAN 0 */
6652 	igb_set_vf_vlan(adapter, 0, true, vf);
6653 
6654 	igb_set_vmvir(adapter, 0, vf);
6655 	igb_set_vmolr(adapter, vf, true);
6656 
6657 	/* Remove any PF assigned VLAN */
6658 	if (adapter->vf_data[vf].pf_vlan)
6659 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
6660 				false, vf);
6661 
6662 	adapter->vf_data[vf].pf_vlan = 0;
6663 	adapter->vf_data[vf].pf_qos = 0;
6664 	igb_set_vf_vlan_strip(adapter, vf, false);
6665 
6666 	return 0;
6667 }
6668 
6669 static int igb_ndo_set_vf_vlan(struct net_device *netdev, int vf,
6670 			       u16 vlan, u8 qos, __be16 vlan_proto)
6671 {
6672 	struct igb_adapter *adapter = netdev_priv(netdev);
6673 
6674 	if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
6675 		return -EINVAL;
6676 
6677 	if (vlan_proto != htons(ETH_P_8021Q))
6678 		return -EPROTONOSUPPORT;
6679 
6680 	return (vlan || qos) ? igb_enable_port_vlan(adapter, vf, vlan, qos) :
6681 			       igb_disable_port_vlan(adapter, vf);
6682 }
6683 
6684 static int igb_set_vf_vlan_msg(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
6685 {
6686 	int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
6687 	int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
6688 	int ret;
6689 
6690 	if (adapter->vf_data[vf].pf_vlan)
6691 		return -1;
6692 
6693 	/* VLAN 0 is a special case, don't allow it to be removed */
6694 	if (!vid && !add)
6695 		return 0;
6696 
6697 	ret = igb_set_vf_vlan(adapter, vid, !!add, vf);
6698 	if (!ret)
6699 		igb_set_vf_vlan_strip(adapter, vf, !!vid);
6700 	return ret;
6701 }
6702 
6703 static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
6704 {
6705 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6706 
6707 	/* clear flags - except flag that indicates PF has set the MAC */
6708 	vf_data->flags &= IGB_VF_FLAG_PF_SET_MAC;
6709 	vf_data->last_nack = jiffies;
6710 
6711 	/* reset vlans for device */
6712 	igb_clear_vf_vfta(adapter, vf);
6713 	igb_set_vf_vlan(adapter, vf_data->pf_vlan, true, vf);
6714 	igb_set_vmvir(adapter, vf_data->pf_vlan |
6715 			       (vf_data->pf_qos << VLAN_PRIO_SHIFT), vf);
6716 	igb_set_vmolr(adapter, vf, !vf_data->pf_vlan);
6717 	igb_set_vf_vlan_strip(adapter, vf, !!(vf_data->pf_vlan));
6718 
6719 	/* reset multicast table array for vf */
6720 	adapter->vf_data[vf].num_vf_mc_hashes = 0;
6721 
6722 	/* Flush and reset the mta with the new values */
6723 	igb_set_rx_mode(adapter->netdev);
6724 }
6725 
6726 static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
6727 {
6728 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6729 
6730 	/* clear mac address as we were hotplug removed/added */
6731 	if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
6732 		eth_zero_addr(vf_mac);
6733 
6734 	/* process remaining reset events */
6735 	igb_vf_reset(adapter, vf);
6736 }
6737 
6738 static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
6739 {
6740 	struct e1000_hw *hw = &adapter->hw;
6741 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6742 	u32 reg, msgbuf[3];
6743 	u8 *addr = (u8 *)(&msgbuf[1]);
6744 
6745 	/* process all the same items cleared in a function level reset */
6746 	igb_vf_reset(adapter, vf);
6747 
6748 	/* set vf mac address */
6749 	igb_set_vf_mac(adapter, vf, vf_mac);
6750 
6751 	/* enable transmit and receive for vf */
6752 	reg = rd32(E1000_VFTE);
6753 	wr32(E1000_VFTE, reg | BIT(vf));
6754 	reg = rd32(E1000_VFRE);
6755 	wr32(E1000_VFRE, reg | BIT(vf));
6756 
6757 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
6758 
6759 	/* reply to reset with ack and vf mac address */
6760 	if (!is_zero_ether_addr(vf_mac)) {
6761 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
6762 		memcpy(addr, vf_mac, ETH_ALEN);
6763 	} else {
6764 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
6765 	}
6766 	igb_write_mbx(hw, msgbuf, 3, vf);
6767 }
6768 
6769 static void igb_flush_mac_table(struct igb_adapter *adapter)
6770 {
6771 	struct e1000_hw *hw = &adapter->hw;
6772 	int i;
6773 
6774 	for (i = 0; i < hw->mac.rar_entry_count; i++) {
6775 		adapter->mac_table[i].state &= ~IGB_MAC_STATE_IN_USE;
6776 		memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
6777 		adapter->mac_table[i].queue = 0;
6778 		igb_rar_set_index(adapter, i);
6779 	}
6780 }
6781 
6782 static int igb_available_rars(struct igb_adapter *adapter, u8 queue)
6783 {
6784 	struct e1000_hw *hw = &adapter->hw;
6785 	/* do not count rar entries reserved for VFs MAC addresses */
6786 	int rar_entries = hw->mac.rar_entry_count -
6787 			  adapter->vfs_allocated_count;
6788 	int i, count = 0;
6789 
6790 	for (i = 0; i < rar_entries; i++) {
6791 		/* do not count default entries */
6792 		if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT)
6793 			continue;
6794 
6795 		/* do not count "in use" entries for different queues */
6796 		if ((adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE) &&
6797 		    (adapter->mac_table[i].queue != queue))
6798 			continue;
6799 
6800 		count++;
6801 	}
6802 
6803 	return count;
6804 }
6805 
6806 /* Set default MAC address for the PF in the first RAR entry */
6807 static void igb_set_default_mac_filter(struct igb_adapter *adapter)
6808 {
6809 	struct igb_mac_addr *mac_table = &adapter->mac_table[0];
6810 
6811 	ether_addr_copy(mac_table->addr, adapter->hw.mac.addr);
6812 	mac_table->queue = adapter->vfs_allocated_count;
6813 	mac_table->state = IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
6814 
6815 	igb_rar_set_index(adapter, 0);
6816 }
6817 
6818 static int igb_add_mac_filter(struct igb_adapter *adapter, const u8 *addr,
6819 			      const u8 queue)
6820 {
6821 	struct e1000_hw *hw = &adapter->hw;
6822 	int rar_entries = hw->mac.rar_entry_count -
6823 			  adapter->vfs_allocated_count;
6824 	int i;
6825 
6826 	if (is_zero_ether_addr(addr))
6827 		return -EINVAL;
6828 
6829 	/* Search for the first empty entry in the MAC table.
6830 	 * Do not touch entries at the end of the table reserved for the VF MAC
6831 	 * addresses.
6832 	 */
6833 	for (i = 0; i < rar_entries; i++) {
6834 		if (adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE)
6835 			continue;
6836 
6837 		ether_addr_copy(adapter->mac_table[i].addr, addr);
6838 		adapter->mac_table[i].queue = queue;
6839 		adapter->mac_table[i].state |= IGB_MAC_STATE_IN_USE;
6840 
6841 		igb_rar_set_index(adapter, i);
6842 		return i;
6843 	}
6844 
6845 	return -ENOSPC;
6846 }
6847 
6848 static int igb_del_mac_filter(struct igb_adapter *adapter, const u8 *addr,
6849 			      const u8 queue)
6850 {
6851 	struct e1000_hw *hw = &adapter->hw;
6852 	int rar_entries = hw->mac.rar_entry_count -
6853 			  adapter->vfs_allocated_count;
6854 	int i;
6855 
6856 	if (is_zero_ether_addr(addr))
6857 		return -EINVAL;
6858 
6859 	/* Search for matching entry in the MAC table based on given address
6860 	 * and queue. Do not touch entries at the end of the table reserved
6861 	 * for the VF MAC addresses.
6862 	 */
6863 	for (i = 0; i < rar_entries; i++) {
6864 		if (!(adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE))
6865 			continue;
6866 		if (adapter->mac_table[i].queue != queue)
6867 			continue;
6868 		if (!ether_addr_equal(adapter->mac_table[i].addr, addr))
6869 			continue;
6870 
6871 		adapter->mac_table[i].state &= ~IGB_MAC_STATE_IN_USE;
6872 		memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
6873 		adapter->mac_table[i].queue = 0;
6874 
6875 		igb_rar_set_index(adapter, i);
6876 		return 0;
6877 	}
6878 
6879 	return -ENOENT;
6880 }
6881 
6882 static int igb_uc_sync(struct net_device *netdev, const unsigned char *addr)
6883 {
6884 	struct igb_adapter *adapter = netdev_priv(netdev);
6885 	int ret;
6886 
6887 	ret = igb_add_mac_filter(adapter, addr, adapter->vfs_allocated_count);
6888 
6889 	return min_t(int, ret, 0);
6890 }
6891 
6892 static int igb_uc_unsync(struct net_device *netdev, const unsigned char *addr)
6893 {
6894 	struct igb_adapter *adapter = netdev_priv(netdev);
6895 
6896 	igb_del_mac_filter(adapter, addr, adapter->vfs_allocated_count);
6897 
6898 	return 0;
6899 }
6900 
6901 static int igb_set_vf_mac_filter(struct igb_adapter *adapter, const int vf,
6902 				 const u32 info, const u8 *addr)
6903 {
6904 	struct pci_dev *pdev = adapter->pdev;
6905 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6906 	struct list_head *pos;
6907 	struct vf_mac_filter *entry = NULL;
6908 	int ret = 0;
6909 
6910 	switch (info) {
6911 	case E1000_VF_MAC_FILTER_CLR:
6912 		/* remove all unicast MAC filters related to the current VF */
6913 		list_for_each(pos, &adapter->vf_macs.l) {
6914 			entry = list_entry(pos, struct vf_mac_filter, l);
6915 			if (entry->vf == vf) {
6916 				entry->vf = -1;
6917 				entry->free = true;
6918 				igb_del_mac_filter(adapter, entry->vf_mac, vf);
6919 			}
6920 		}
6921 		break;
6922 	case E1000_VF_MAC_FILTER_ADD:
6923 		if (vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) {
6924 			dev_warn(&pdev->dev,
6925 				 "VF %d requested MAC filter but is administratively denied\n",
6926 				 vf);
6927 			return -EINVAL;
6928 		}
6929 
6930 		if (!is_valid_ether_addr(addr)) {
6931 			dev_warn(&pdev->dev,
6932 				 "VF %d attempted to set invalid MAC filter\n",
6933 				 vf);
6934 			return -EINVAL;
6935 		}
6936 
6937 		/* try to find empty slot in the list */
6938 		list_for_each(pos, &adapter->vf_macs.l) {
6939 			entry = list_entry(pos, struct vf_mac_filter, l);
6940 			if (entry->free)
6941 				break;
6942 		}
6943 
6944 		if (entry && entry->free) {
6945 			entry->free = false;
6946 			entry->vf = vf;
6947 			ether_addr_copy(entry->vf_mac, addr);
6948 
6949 			ret = igb_add_mac_filter(adapter, addr, vf);
6950 			ret = min_t(int, ret, 0);
6951 		} else {
6952 			ret = -ENOSPC;
6953 		}
6954 
6955 		if (ret == -ENOSPC)
6956 			dev_warn(&pdev->dev,
6957 				 "VF %d has requested MAC filter but there is no space for it\n",
6958 				 vf);
6959 		break;
6960 	default:
6961 		ret = -EINVAL;
6962 		break;
6963 	}
6964 
6965 	return ret;
6966 }
6967 
6968 static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
6969 {
6970 	struct pci_dev *pdev = adapter->pdev;
6971 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6972 	u32 info = msg[0] & E1000_VT_MSGINFO_MASK;
6973 
6974 	/* The VF MAC Address is stored in a packed array of bytes
6975 	 * starting at the second 32 bit word of the msg array
6976 	 */
6977 	unsigned char *addr = (unsigned char *)&msg[1];
6978 	int ret = 0;
6979 
6980 	if (!info) {
6981 		if (vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) {
6982 			dev_warn(&pdev->dev,
6983 				 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
6984 				 vf);
6985 			return -EINVAL;
6986 		}
6987 
6988 		if (!is_valid_ether_addr(addr)) {
6989 			dev_warn(&pdev->dev,
6990 				 "VF %d attempted to set invalid MAC\n",
6991 				 vf);
6992 			return -EINVAL;
6993 		}
6994 
6995 		ret = igb_set_vf_mac(adapter, vf, addr);
6996 	} else {
6997 		ret = igb_set_vf_mac_filter(adapter, vf, info, addr);
6998 	}
6999 
7000 	return ret;
7001 }
7002 
7003 static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
7004 {
7005 	struct e1000_hw *hw = &adapter->hw;
7006 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7007 	u32 msg = E1000_VT_MSGTYPE_NACK;
7008 
7009 	/* if device isn't clear to send it shouldn't be reading either */
7010 	if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
7011 	    time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
7012 		igb_write_mbx(hw, &msg, 1, vf);
7013 		vf_data->last_nack = jiffies;
7014 	}
7015 }
7016 
7017 static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
7018 {
7019 	struct pci_dev *pdev = adapter->pdev;
7020 	u32 msgbuf[E1000_VFMAILBOX_SIZE];
7021 	struct e1000_hw *hw = &adapter->hw;
7022 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7023 	s32 retval;
7024 
7025 	retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf, false);
7026 
7027 	if (retval) {
7028 		/* if receive failed revoke VF CTS stats and restart init */
7029 		dev_err(&pdev->dev, "Error receiving message from VF\n");
7030 		vf_data->flags &= ~IGB_VF_FLAG_CTS;
7031 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7032 			goto unlock;
7033 		goto out;
7034 	}
7035 
7036 	/* this is a message we already processed, do nothing */
7037 	if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
7038 		goto unlock;
7039 
7040 	/* until the vf completes a reset it should not be
7041 	 * allowed to start any configuration.
7042 	 */
7043 	if (msgbuf[0] == E1000_VF_RESET) {
7044 		/* unlocks mailbox */
7045 		igb_vf_reset_msg(adapter, vf);
7046 		return;
7047 	}
7048 
7049 	if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
7050 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7051 			goto unlock;
7052 		retval = -1;
7053 		goto out;
7054 	}
7055 
7056 	switch ((msgbuf[0] & 0xFFFF)) {
7057 	case E1000_VF_SET_MAC_ADDR:
7058 		retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
7059 		break;
7060 	case E1000_VF_SET_PROMISC:
7061 		retval = igb_set_vf_promisc(adapter, msgbuf, vf);
7062 		break;
7063 	case E1000_VF_SET_MULTICAST:
7064 		retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
7065 		break;
7066 	case E1000_VF_SET_LPE:
7067 		retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
7068 		break;
7069 	case E1000_VF_SET_VLAN:
7070 		retval = -1;
7071 		if (vf_data->pf_vlan)
7072 			dev_warn(&pdev->dev,
7073 				 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
7074 				 vf);
7075 		else
7076 			retval = igb_set_vf_vlan_msg(adapter, msgbuf, vf);
7077 		break;
7078 	default:
7079 		dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
7080 		retval = -1;
7081 		break;
7082 	}
7083 
7084 	msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
7085 out:
7086 	/* notify the VF of the results of what it sent us */
7087 	if (retval)
7088 		msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
7089 	else
7090 		msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
7091 
7092 	/* unlocks mailbox */
7093 	igb_write_mbx(hw, msgbuf, 1, vf);
7094 	return;
7095 
7096 unlock:
7097 	igb_unlock_mbx(hw, vf);
7098 }
7099 
7100 static void igb_msg_task(struct igb_adapter *adapter)
7101 {
7102 	struct e1000_hw *hw = &adapter->hw;
7103 	u32 vf;
7104 
7105 	for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
7106 		/* process any reset requests */
7107 		if (!igb_check_for_rst(hw, vf))
7108 			igb_vf_reset_event(adapter, vf);
7109 
7110 		/* process any messages pending */
7111 		if (!igb_check_for_msg(hw, vf))
7112 			igb_rcv_msg_from_vf(adapter, vf);
7113 
7114 		/* process any acks */
7115 		if (!igb_check_for_ack(hw, vf))
7116 			igb_rcv_ack_from_vf(adapter, vf);
7117 	}
7118 }
7119 
7120 /**
7121  *  igb_set_uta - Set unicast filter table address
7122  *  @adapter: board private structure
7123  *  @set: boolean indicating if we are setting or clearing bits
7124  *
7125  *  The unicast table address is a register array of 32-bit registers.
7126  *  The table is meant to be used in a way similar to how the MTA is used
7127  *  however due to certain limitations in the hardware it is necessary to
7128  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
7129  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
7130  **/
7131 static void igb_set_uta(struct igb_adapter *adapter, bool set)
7132 {
7133 	struct e1000_hw *hw = &adapter->hw;
7134 	u32 uta = set ? ~0 : 0;
7135 	int i;
7136 
7137 	/* we only need to do this if VMDq is enabled */
7138 	if (!adapter->vfs_allocated_count)
7139 		return;
7140 
7141 	for (i = hw->mac.uta_reg_count; i--;)
7142 		array_wr32(E1000_UTA, i, uta);
7143 }
7144 
7145 /**
7146  *  igb_intr_msi - Interrupt Handler
7147  *  @irq: interrupt number
7148  *  @data: pointer to a network interface device structure
7149  **/
7150 static irqreturn_t igb_intr_msi(int irq, void *data)
7151 {
7152 	struct igb_adapter *adapter = data;
7153 	struct igb_q_vector *q_vector = adapter->q_vector[0];
7154 	struct e1000_hw *hw = &adapter->hw;
7155 	/* read ICR disables interrupts using IAM */
7156 	u32 icr = rd32(E1000_ICR);
7157 
7158 	igb_write_itr(q_vector);
7159 
7160 	if (icr & E1000_ICR_DRSTA)
7161 		schedule_work(&adapter->reset_task);
7162 
7163 	if (icr & E1000_ICR_DOUTSYNC) {
7164 		/* HW is reporting DMA is out of sync */
7165 		adapter->stats.doosync++;
7166 	}
7167 
7168 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
7169 		hw->mac.get_link_status = 1;
7170 		if (!test_bit(__IGB_DOWN, &adapter->state))
7171 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
7172 	}
7173 
7174 	if (icr & E1000_ICR_TS)
7175 		igb_tsync_interrupt(adapter);
7176 
7177 	napi_schedule(&q_vector->napi);
7178 
7179 	return IRQ_HANDLED;
7180 }
7181 
7182 /**
7183  *  igb_intr - Legacy Interrupt Handler
7184  *  @irq: interrupt number
7185  *  @data: pointer to a network interface device structure
7186  **/
7187 static irqreturn_t igb_intr(int irq, void *data)
7188 {
7189 	struct igb_adapter *adapter = data;
7190 	struct igb_q_vector *q_vector = adapter->q_vector[0];
7191 	struct e1000_hw *hw = &adapter->hw;
7192 	/* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
7193 	 * need for the IMC write
7194 	 */
7195 	u32 icr = rd32(E1000_ICR);
7196 
7197 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
7198 	 * not set, then the adapter didn't send an interrupt
7199 	 */
7200 	if (!(icr & E1000_ICR_INT_ASSERTED))
7201 		return IRQ_NONE;
7202 
7203 	igb_write_itr(q_vector);
7204 
7205 	if (icr & E1000_ICR_DRSTA)
7206 		schedule_work(&adapter->reset_task);
7207 
7208 	if (icr & E1000_ICR_DOUTSYNC) {
7209 		/* HW is reporting DMA is out of sync */
7210 		adapter->stats.doosync++;
7211 	}
7212 
7213 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
7214 		hw->mac.get_link_status = 1;
7215 		/* guard against interrupt when we're going down */
7216 		if (!test_bit(__IGB_DOWN, &adapter->state))
7217 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
7218 	}
7219 
7220 	if (icr & E1000_ICR_TS)
7221 		igb_tsync_interrupt(adapter);
7222 
7223 	napi_schedule(&q_vector->napi);
7224 
7225 	return IRQ_HANDLED;
7226 }
7227 
7228 static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
7229 {
7230 	struct igb_adapter *adapter = q_vector->adapter;
7231 	struct e1000_hw *hw = &adapter->hw;
7232 
7233 	if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
7234 	    (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
7235 		if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
7236 			igb_set_itr(q_vector);
7237 		else
7238 			igb_update_ring_itr(q_vector);
7239 	}
7240 
7241 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
7242 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
7243 			wr32(E1000_EIMS, q_vector->eims_value);
7244 		else
7245 			igb_irq_enable(adapter);
7246 	}
7247 }
7248 
7249 /**
7250  *  igb_poll - NAPI Rx polling callback
7251  *  @napi: napi polling structure
7252  *  @budget: count of how many packets we should handle
7253  **/
7254 static int igb_poll(struct napi_struct *napi, int budget)
7255 {
7256 	struct igb_q_vector *q_vector = container_of(napi,
7257 						     struct igb_q_vector,
7258 						     napi);
7259 	bool clean_complete = true;
7260 	int work_done = 0;
7261 
7262 #ifdef CONFIG_IGB_DCA
7263 	if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
7264 		igb_update_dca(q_vector);
7265 #endif
7266 	if (q_vector->tx.ring)
7267 		clean_complete = igb_clean_tx_irq(q_vector, budget);
7268 
7269 	if (q_vector->rx.ring) {
7270 		int cleaned = igb_clean_rx_irq(q_vector, budget);
7271 
7272 		work_done += cleaned;
7273 		if (cleaned >= budget)
7274 			clean_complete = false;
7275 	}
7276 
7277 	/* If all work not completed, return budget and keep polling */
7278 	if (!clean_complete)
7279 		return budget;
7280 
7281 	/* If not enough Rx work done, exit the polling mode */
7282 	napi_complete_done(napi, work_done);
7283 	igb_ring_irq_enable(q_vector);
7284 
7285 	return 0;
7286 }
7287 
7288 /**
7289  *  igb_clean_tx_irq - Reclaim resources after transmit completes
7290  *  @q_vector: pointer to q_vector containing needed info
7291  *  @napi_budget: Used to determine if we are in netpoll
7292  *
7293  *  returns true if ring is completely cleaned
7294  **/
7295 static bool igb_clean_tx_irq(struct igb_q_vector *q_vector, int napi_budget)
7296 {
7297 	struct igb_adapter *adapter = q_vector->adapter;
7298 	struct igb_ring *tx_ring = q_vector->tx.ring;
7299 	struct igb_tx_buffer *tx_buffer;
7300 	union e1000_adv_tx_desc *tx_desc;
7301 	unsigned int total_bytes = 0, total_packets = 0;
7302 	unsigned int budget = q_vector->tx.work_limit;
7303 	unsigned int i = tx_ring->next_to_clean;
7304 
7305 	if (test_bit(__IGB_DOWN, &adapter->state))
7306 		return true;
7307 
7308 	tx_buffer = &tx_ring->tx_buffer_info[i];
7309 	tx_desc = IGB_TX_DESC(tx_ring, i);
7310 	i -= tx_ring->count;
7311 
7312 	do {
7313 		union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
7314 
7315 		/* if next_to_watch is not set then there is no work pending */
7316 		if (!eop_desc)
7317 			break;
7318 
7319 		/* prevent any other reads prior to eop_desc */
7320 		smp_rmb();
7321 
7322 		/* if DD is not set pending work has not been completed */
7323 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
7324 			break;
7325 
7326 		/* clear next_to_watch to prevent false hangs */
7327 		tx_buffer->next_to_watch = NULL;
7328 
7329 		/* update the statistics for this packet */
7330 		total_bytes += tx_buffer->bytecount;
7331 		total_packets += tx_buffer->gso_segs;
7332 
7333 		/* free the skb */
7334 		napi_consume_skb(tx_buffer->skb, napi_budget);
7335 
7336 		/* unmap skb header data */
7337 		dma_unmap_single(tx_ring->dev,
7338 				 dma_unmap_addr(tx_buffer, dma),
7339 				 dma_unmap_len(tx_buffer, len),
7340 				 DMA_TO_DEVICE);
7341 
7342 		/* clear tx_buffer data */
7343 		dma_unmap_len_set(tx_buffer, len, 0);
7344 
7345 		/* clear last DMA location and unmap remaining buffers */
7346 		while (tx_desc != eop_desc) {
7347 			tx_buffer++;
7348 			tx_desc++;
7349 			i++;
7350 			if (unlikely(!i)) {
7351 				i -= tx_ring->count;
7352 				tx_buffer = tx_ring->tx_buffer_info;
7353 				tx_desc = IGB_TX_DESC(tx_ring, 0);
7354 			}
7355 
7356 			/* unmap any remaining paged data */
7357 			if (dma_unmap_len(tx_buffer, len)) {
7358 				dma_unmap_page(tx_ring->dev,
7359 					       dma_unmap_addr(tx_buffer, dma),
7360 					       dma_unmap_len(tx_buffer, len),
7361 					       DMA_TO_DEVICE);
7362 				dma_unmap_len_set(tx_buffer, len, 0);
7363 			}
7364 		}
7365 
7366 		/* move us one more past the eop_desc for start of next pkt */
7367 		tx_buffer++;
7368 		tx_desc++;
7369 		i++;
7370 		if (unlikely(!i)) {
7371 			i -= tx_ring->count;
7372 			tx_buffer = tx_ring->tx_buffer_info;
7373 			tx_desc = IGB_TX_DESC(tx_ring, 0);
7374 		}
7375 
7376 		/* issue prefetch for next Tx descriptor */
7377 		prefetch(tx_desc);
7378 
7379 		/* update budget accounting */
7380 		budget--;
7381 	} while (likely(budget));
7382 
7383 	netdev_tx_completed_queue(txring_txq(tx_ring),
7384 				  total_packets, total_bytes);
7385 	i += tx_ring->count;
7386 	tx_ring->next_to_clean = i;
7387 	u64_stats_update_begin(&tx_ring->tx_syncp);
7388 	tx_ring->tx_stats.bytes += total_bytes;
7389 	tx_ring->tx_stats.packets += total_packets;
7390 	u64_stats_update_end(&tx_ring->tx_syncp);
7391 	q_vector->tx.total_bytes += total_bytes;
7392 	q_vector->tx.total_packets += total_packets;
7393 
7394 	if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
7395 		struct e1000_hw *hw = &adapter->hw;
7396 
7397 		/* Detect a transmit hang in hardware, this serializes the
7398 		 * check with the clearing of time_stamp and movement of i
7399 		 */
7400 		clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
7401 		if (tx_buffer->next_to_watch &&
7402 		    time_after(jiffies, tx_buffer->time_stamp +
7403 			       (adapter->tx_timeout_factor * HZ)) &&
7404 		    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
7405 
7406 			/* detected Tx unit hang */
7407 			dev_err(tx_ring->dev,
7408 				"Detected Tx Unit Hang\n"
7409 				"  Tx Queue             <%d>\n"
7410 				"  TDH                  <%x>\n"
7411 				"  TDT                  <%x>\n"
7412 				"  next_to_use          <%x>\n"
7413 				"  next_to_clean        <%x>\n"
7414 				"buffer_info[next_to_clean]\n"
7415 				"  time_stamp           <%lx>\n"
7416 				"  next_to_watch        <%p>\n"
7417 				"  jiffies              <%lx>\n"
7418 				"  desc.status          <%x>\n",
7419 				tx_ring->queue_index,
7420 				rd32(E1000_TDH(tx_ring->reg_idx)),
7421 				readl(tx_ring->tail),
7422 				tx_ring->next_to_use,
7423 				tx_ring->next_to_clean,
7424 				tx_buffer->time_stamp,
7425 				tx_buffer->next_to_watch,
7426 				jiffies,
7427 				tx_buffer->next_to_watch->wb.status);
7428 			netif_stop_subqueue(tx_ring->netdev,
7429 					    tx_ring->queue_index);
7430 
7431 			/* we are about to reset, no point in enabling stuff */
7432 			return true;
7433 		}
7434 	}
7435 
7436 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
7437 	if (unlikely(total_packets &&
7438 	    netif_carrier_ok(tx_ring->netdev) &&
7439 	    igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
7440 		/* Make sure that anybody stopping the queue after this
7441 		 * sees the new next_to_clean.
7442 		 */
7443 		smp_mb();
7444 		if (__netif_subqueue_stopped(tx_ring->netdev,
7445 					     tx_ring->queue_index) &&
7446 		    !(test_bit(__IGB_DOWN, &adapter->state))) {
7447 			netif_wake_subqueue(tx_ring->netdev,
7448 					    tx_ring->queue_index);
7449 
7450 			u64_stats_update_begin(&tx_ring->tx_syncp);
7451 			tx_ring->tx_stats.restart_queue++;
7452 			u64_stats_update_end(&tx_ring->tx_syncp);
7453 		}
7454 	}
7455 
7456 	return !!budget;
7457 }
7458 
7459 /**
7460  *  igb_reuse_rx_page - page flip buffer and store it back on the ring
7461  *  @rx_ring: rx descriptor ring to store buffers on
7462  *  @old_buff: donor buffer to have page reused
7463  *
7464  *  Synchronizes page for reuse by the adapter
7465  **/
7466 static void igb_reuse_rx_page(struct igb_ring *rx_ring,
7467 			      struct igb_rx_buffer *old_buff)
7468 {
7469 	struct igb_rx_buffer *new_buff;
7470 	u16 nta = rx_ring->next_to_alloc;
7471 
7472 	new_buff = &rx_ring->rx_buffer_info[nta];
7473 
7474 	/* update, and store next to alloc */
7475 	nta++;
7476 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
7477 
7478 	/* Transfer page from old buffer to new buffer.
7479 	 * Move each member individually to avoid possible store
7480 	 * forwarding stalls.
7481 	 */
7482 	new_buff->dma		= old_buff->dma;
7483 	new_buff->page		= old_buff->page;
7484 	new_buff->page_offset	= old_buff->page_offset;
7485 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
7486 }
7487 
7488 static inline bool igb_page_is_reserved(struct page *page)
7489 {
7490 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
7491 }
7492 
7493 static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer)
7494 {
7495 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
7496 	struct page *page = rx_buffer->page;
7497 
7498 	/* avoid re-using remote pages */
7499 	if (unlikely(igb_page_is_reserved(page)))
7500 		return false;
7501 
7502 #if (PAGE_SIZE < 8192)
7503 	/* if we are only owner of page we can reuse it */
7504 	if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
7505 		return false;
7506 #else
7507 #define IGB_LAST_OFFSET \
7508 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IGB_RXBUFFER_2048)
7509 
7510 	if (rx_buffer->page_offset > IGB_LAST_OFFSET)
7511 		return false;
7512 #endif
7513 
7514 	/* If we have drained the page fragment pool we need to update
7515 	 * the pagecnt_bias and page count so that we fully restock the
7516 	 * number of references the driver holds.
7517 	 */
7518 	if (unlikely(!pagecnt_bias)) {
7519 		page_ref_add(page, USHRT_MAX);
7520 		rx_buffer->pagecnt_bias = USHRT_MAX;
7521 	}
7522 
7523 	return true;
7524 }
7525 
7526 /**
7527  *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
7528  *  @rx_ring: rx descriptor ring to transact packets on
7529  *  @rx_buffer: buffer containing page to add
7530  *  @skb: sk_buff to place the data into
7531  *  @size: size of buffer to be added
7532  *
7533  *  This function will add the data contained in rx_buffer->page to the skb.
7534  **/
7535 static void igb_add_rx_frag(struct igb_ring *rx_ring,
7536 			    struct igb_rx_buffer *rx_buffer,
7537 			    struct sk_buff *skb,
7538 			    unsigned int size)
7539 {
7540 #if (PAGE_SIZE < 8192)
7541 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
7542 #else
7543 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
7544 				SKB_DATA_ALIGN(IGB_SKB_PAD + size) :
7545 				SKB_DATA_ALIGN(size);
7546 #endif
7547 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
7548 			rx_buffer->page_offset, size, truesize);
7549 #if (PAGE_SIZE < 8192)
7550 	rx_buffer->page_offset ^= truesize;
7551 #else
7552 	rx_buffer->page_offset += truesize;
7553 #endif
7554 }
7555 
7556 static struct sk_buff *igb_construct_skb(struct igb_ring *rx_ring,
7557 					 struct igb_rx_buffer *rx_buffer,
7558 					 union e1000_adv_rx_desc *rx_desc,
7559 					 unsigned int size)
7560 {
7561 	void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
7562 #if (PAGE_SIZE < 8192)
7563 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
7564 #else
7565 	unsigned int truesize = SKB_DATA_ALIGN(size);
7566 #endif
7567 	unsigned int headlen;
7568 	struct sk_buff *skb;
7569 
7570 	/* prefetch first cache line of first page */
7571 	prefetch(va);
7572 #if L1_CACHE_BYTES < 128
7573 	prefetch(va + L1_CACHE_BYTES);
7574 #endif
7575 
7576 	/* allocate a skb to store the frags */
7577 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
7578 	if (unlikely(!skb))
7579 		return NULL;
7580 
7581 	if (unlikely(igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))) {
7582 		igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
7583 		va += IGB_TS_HDR_LEN;
7584 		size -= IGB_TS_HDR_LEN;
7585 	}
7586 
7587 	/* Determine available headroom for copy */
7588 	headlen = size;
7589 	if (headlen > IGB_RX_HDR_LEN)
7590 		headlen = eth_get_headlen(va, IGB_RX_HDR_LEN);
7591 
7592 	/* align pull length to size of long to optimize memcpy performance */
7593 	memcpy(__skb_put(skb, headlen), va, ALIGN(headlen, sizeof(long)));
7594 
7595 	/* update all of the pointers */
7596 	size -= headlen;
7597 	if (size) {
7598 		skb_add_rx_frag(skb, 0, rx_buffer->page,
7599 				(va + headlen) - page_address(rx_buffer->page),
7600 				size, truesize);
7601 #if (PAGE_SIZE < 8192)
7602 		rx_buffer->page_offset ^= truesize;
7603 #else
7604 		rx_buffer->page_offset += truesize;
7605 #endif
7606 	} else {
7607 		rx_buffer->pagecnt_bias++;
7608 	}
7609 
7610 	return skb;
7611 }
7612 
7613 static struct sk_buff *igb_build_skb(struct igb_ring *rx_ring,
7614 				     struct igb_rx_buffer *rx_buffer,
7615 				     union e1000_adv_rx_desc *rx_desc,
7616 				     unsigned int size)
7617 {
7618 	void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
7619 #if (PAGE_SIZE < 8192)
7620 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
7621 #else
7622 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
7623 				SKB_DATA_ALIGN(IGB_SKB_PAD + size);
7624 #endif
7625 	struct sk_buff *skb;
7626 
7627 	/* prefetch first cache line of first page */
7628 	prefetch(va);
7629 #if L1_CACHE_BYTES < 128
7630 	prefetch(va + L1_CACHE_BYTES);
7631 #endif
7632 
7633 	/* build an skb around the page buffer */
7634 	skb = build_skb(va - IGB_SKB_PAD, truesize);
7635 	if (unlikely(!skb))
7636 		return NULL;
7637 
7638 	/* update pointers within the skb to store the data */
7639 	skb_reserve(skb, IGB_SKB_PAD);
7640 	__skb_put(skb, size);
7641 
7642 	/* pull timestamp out of packet data */
7643 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
7644 		igb_ptp_rx_pktstamp(rx_ring->q_vector, skb->data, skb);
7645 		__skb_pull(skb, IGB_TS_HDR_LEN);
7646 	}
7647 
7648 	/* update buffer offset */
7649 #if (PAGE_SIZE < 8192)
7650 	rx_buffer->page_offset ^= truesize;
7651 #else
7652 	rx_buffer->page_offset += truesize;
7653 #endif
7654 
7655 	return skb;
7656 }
7657 
7658 static inline void igb_rx_checksum(struct igb_ring *ring,
7659 				   union e1000_adv_rx_desc *rx_desc,
7660 				   struct sk_buff *skb)
7661 {
7662 	skb_checksum_none_assert(skb);
7663 
7664 	/* Ignore Checksum bit is set */
7665 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
7666 		return;
7667 
7668 	/* Rx checksum disabled via ethtool */
7669 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
7670 		return;
7671 
7672 	/* TCP/UDP checksum error bit is set */
7673 	if (igb_test_staterr(rx_desc,
7674 			     E1000_RXDEXT_STATERR_TCPE |
7675 			     E1000_RXDEXT_STATERR_IPE)) {
7676 		/* work around errata with sctp packets where the TCPE aka
7677 		 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
7678 		 * packets, (aka let the stack check the crc32c)
7679 		 */
7680 		if (!((skb->len == 60) &&
7681 		      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
7682 			u64_stats_update_begin(&ring->rx_syncp);
7683 			ring->rx_stats.csum_err++;
7684 			u64_stats_update_end(&ring->rx_syncp);
7685 		}
7686 		/* let the stack verify checksum errors */
7687 		return;
7688 	}
7689 	/* It must be a TCP or UDP packet with a valid checksum */
7690 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
7691 				      E1000_RXD_STAT_UDPCS))
7692 		skb->ip_summed = CHECKSUM_UNNECESSARY;
7693 
7694 	dev_dbg(ring->dev, "cksum success: bits %08X\n",
7695 		le32_to_cpu(rx_desc->wb.upper.status_error));
7696 }
7697 
7698 static inline void igb_rx_hash(struct igb_ring *ring,
7699 			       union e1000_adv_rx_desc *rx_desc,
7700 			       struct sk_buff *skb)
7701 {
7702 	if (ring->netdev->features & NETIF_F_RXHASH)
7703 		skb_set_hash(skb,
7704 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
7705 			     PKT_HASH_TYPE_L3);
7706 }
7707 
7708 /**
7709  *  igb_is_non_eop - process handling of non-EOP buffers
7710  *  @rx_ring: Rx ring being processed
7711  *  @rx_desc: Rx descriptor for current buffer
7712  *  @skb: current socket buffer containing buffer in progress
7713  *
7714  *  This function updates next to clean.  If the buffer is an EOP buffer
7715  *  this function exits returning false, otherwise it will place the
7716  *  sk_buff in the next buffer to be chained and return true indicating
7717  *  that this is in fact a non-EOP buffer.
7718  **/
7719 static bool igb_is_non_eop(struct igb_ring *rx_ring,
7720 			   union e1000_adv_rx_desc *rx_desc)
7721 {
7722 	u32 ntc = rx_ring->next_to_clean + 1;
7723 
7724 	/* fetch, update, and store next to clean */
7725 	ntc = (ntc < rx_ring->count) ? ntc : 0;
7726 	rx_ring->next_to_clean = ntc;
7727 
7728 	prefetch(IGB_RX_DESC(rx_ring, ntc));
7729 
7730 	if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
7731 		return false;
7732 
7733 	return true;
7734 }
7735 
7736 /**
7737  *  igb_cleanup_headers - Correct corrupted or empty headers
7738  *  @rx_ring: rx descriptor ring packet is being transacted on
7739  *  @rx_desc: pointer to the EOP Rx descriptor
7740  *  @skb: pointer to current skb being fixed
7741  *
7742  *  Address the case where we are pulling data in on pages only
7743  *  and as such no data is present in the skb header.
7744  *
7745  *  In addition if skb is not at least 60 bytes we need to pad it so that
7746  *  it is large enough to qualify as a valid Ethernet frame.
7747  *
7748  *  Returns true if an error was encountered and skb was freed.
7749  **/
7750 static bool igb_cleanup_headers(struct igb_ring *rx_ring,
7751 				union e1000_adv_rx_desc *rx_desc,
7752 				struct sk_buff *skb)
7753 {
7754 	if (unlikely((igb_test_staterr(rx_desc,
7755 				       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
7756 		struct net_device *netdev = rx_ring->netdev;
7757 		if (!(netdev->features & NETIF_F_RXALL)) {
7758 			dev_kfree_skb_any(skb);
7759 			return true;
7760 		}
7761 	}
7762 
7763 	/* if eth_skb_pad returns an error the skb was freed */
7764 	if (eth_skb_pad(skb))
7765 		return true;
7766 
7767 	return false;
7768 }
7769 
7770 /**
7771  *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
7772  *  @rx_ring: rx descriptor ring packet is being transacted on
7773  *  @rx_desc: pointer to the EOP Rx descriptor
7774  *  @skb: pointer to current skb being populated
7775  *
7776  *  This function checks the ring, descriptor, and packet information in
7777  *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
7778  *  other fields within the skb.
7779  **/
7780 static void igb_process_skb_fields(struct igb_ring *rx_ring,
7781 				   union e1000_adv_rx_desc *rx_desc,
7782 				   struct sk_buff *skb)
7783 {
7784 	struct net_device *dev = rx_ring->netdev;
7785 
7786 	igb_rx_hash(rx_ring, rx_desc, skb);
7787 
7788 	igb_rx_checksum(rx_ring, rx_desc, skb);
7789 
7790 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
7791 	    !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
7792 		igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
7793 
7794 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
7795 	    igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
7796 		u16 vid;
7797 
7798 		if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
7799 		    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
7800 			vid = be16_to_cpu(rx_desc->wb.upper.vlan);
7801 		else
7802 			vid = le16_to_cpu(rx_desc->wb.upper.vlan);
7803 
7804 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
7805 	}
7806 
7807 	skb_record_rx_queue(skb, rx_ring->queue_index);
7808 
7809 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
7810 }
7811 
7812 static struct igb_rx_buffer *igb_get_rx_buffer(struct igb_ring *rx_ring,
7813 					       const unsigned int size)
7814 {
7815 	struct igb_rx_buffer *rx_buffer;
7816 
7817 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
7818 	prefetchw(rx_buffer->page);
7819 
7820 	/* we are reusing so sync this buffer for CPU use */
7821 	dma_sync_single_range_for_cpu(rx_ring->dev,
7822 				      rx_buffer->dma,
7823 				      rx_buffer->page_offset,
7824 				      size,
7825 				      DMA_FROM_DEVICE);
7826 
7827 	rx_buffer->pagecnt_bias--;
7828 
7829 	return rx_buffer;
7830 }
7831 
7832 static void igb_put_rx_buffer(struct igb_ring *rx_ring,
7833 			      struct igb_rx_buffer *rx_buffer)
7834 {
7835 	if (igb_can_reuse_rx_page(rx_buffer)) {
7836 		/* hand second half of page back to the ring */
7837 		igb_reuse_rx_page(rx_ring, rx_buffer);
7838 	} else {
7839 		/* We are not reusing the buffer so unmap it and free
7840 		 * any references we are holding to it
7841 		 */
7842 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
7843 				     igb_rx_pg_size(rx_ring), DMA_FROM_DEVICE,
7844 				     IGB_RX_DMA_ATTR);
7845 		__page_frag_cache_drain(rx_buffer->page,
7846 					rx_buffer->pagecnt_bias);
7847 	}
7848 
7849 	/* clear contents of rx_buffer */
7850 	rx_buffer->page = NULL;
7851 }
7852 
7853 static int igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
7854 {
7855 	struct igb_ring *rx_ring = q_vector->rx.ring;
7856 	struct sk_buff *skb = rx_ring->skb;
7857 	unsigned int total_bytes = 0, total_packets = 0;
7858 	u16 cleaned_count = igb_desc_unused(rx_ring);
7859 
7860 	while (likely(total_packets < budget)) {
7861 		union e1000_adv_rx_desc *rx_desc;
7862 		struct igb_rx_buffer *rx_buffer;
7863 		unsigned int size;
7864 
7865 		/* return some buffers to hardware, one at a time is too slow */
7866 		if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
7867 			igb_alloc_rx_buffers(rx_ring, cleaned_count);
7868 			cleaned_count = 0;
7869 		}
7870 
7871 		rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
7872 		size = le16_to_cpu(rx_desc->wb.upper.length);
7873 		if (!size)
7874 			break;
7875 
7876 		/* This memory barrier is needed to keep us from reading
7877 		 * any other fields out of the rx_desc until we know the
7878 		 * descriptor has been written back
7879 		 */
7880 		dma_rmb();
7881 
7882 		rx_buffer = igb_get_rx_buffer(rx_ring, size);
7883 
7884 		/* retrieve a buffer from the ring */
7885 		if (skb)
7886 			igb_add_rx_frag(rx_ring, rx_buffer, skb, size);
7887 		else if (ring_uses_build_skb(rx_ring))
7888 			skb = igb_build_skb(rx_ring, rx_buffer, rx_desc, size);
7889 		else
7890 			skb = igb_construct_skb(rx_ring, rx_buffer,
7891 						rx_desc, size);
7892 
7893 		/* exit if we failed to retrieve a buffer */
7894 		if (!skb) {
7895 			rx_ring->rx_stats.alloc_failed++;
7896 			rx_buffer->pagecnt_bias++;
7897 			break;
7898 		}
7899 
7900 		igb_put_rx_buffer(rx_ring, rx_buffer);
7901 		cleaned_count++;
7902 
7903 		/* fetch next buffer in frame if non-eop */
7904 		if (igb_is_non_eop(rx_ring, rx_desc))
7905 			continue;
7906 
7907 		/* verify the packet layout is correct */
7908 		if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
7909 			skb = NULL;
7910 			continue;
7911 		}
7912 
7913 		/* probably a little skewed due to removing CRC */
7914 		total_bytes += skb->len;
7915 
7916 		/* populate checksum, timestamp, VLAN, and protocol */
7917 		igb_process_skb_fields(rx_ring, rx_desc, skb);
7918 
7919 		napi_gro_receive(&q_vector->napi, skb);
7920 
7921 		/* reset skb pointer */
7922 		skb = NULL;
7923 
7924 		/* update budget accounting */
7925 		total_packets++;
7926 	}
7927 
7928 	/* place incomplete frames back on ring for completion */
7929 	rx_ring->skb = skb;
7930 
7931 	u64_stats_update_begin(&rx_ring->rx_syncp);
7932 	rx_ring->rx_stats.packets += total_packets;
7933 	rx_ring->rx_stats.bytes += total_bytes;
7934 	u64_stats_update_end(&rx_ring->rx_syncp);
7935 	q_vector->rx.total_packets += total_packets;
7936 	q_vector->rx.total_bytes += total_bytes;
7937 
7938 	if (cleaned_count)
7939 		igb_alloc_rx_buffers(rx_ring, cleaned_count);
7940 
7941 	return total_packets;
7942 }
7943 
7944 static inline unsigned int igb_rx_offset(struct igb_ring *rx_ring)
7945 {
7946 	return ring_uses_build_skb(rx_ring) ? IGB_SKB_PAD : 0;
7947 }
7948 
7949 static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
7950 				  struct igb_rx_buffer *bi)
7951 {
7952 	struct page *page = bi->page;
7953 	dma_addr_t dma;
7954 
7955 	/* since we are recycling buffers we should seldom need to alloc */
7956 	if (likely(page))
7957 		return true;
7958 
7959 	/* alloc new page for storage */
7960 	page = dev_alloc_pages(igb_rx_pg_order(rx_ring));
7961 	if (unlikely(!page)) {
7962 		rx_ring->rx_stats.alloc_failed++;
7963 		return false;
7964 	}
7965 
7966 	/* map page for use */
7967 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
7968 				 igb_rx_pg_size(rx_ring),
7969 				 DMA_FROM_DEVICE,
7970 				 IGB_RX_DMA_ATTR);
7971 
7972 	/* if mapping failed free memory back to system since
7973 	 * there isn't much point in holding memory we can't use
7974 	 */
7975 	if (dma_mapping_error(rx_ring->dev, dma)) {
7976 		__free_pages(page, igb_rx_pg_order(rx_ring));
7977 
7978 		rx_ring->rx_stats.alloc_failed++;
7979 		return false;
7980 	}
7981 
7982 	bi->dma = dma;
7983 	bi->page = page;
7984 	bi->page_offset = igb_rx_offset(rx_ring);
7985 	bi->pagecnt_bias = 1;
7986 
7987 	return true;
7988 }
7989 
7990 /**
7991  *  igb_alloc_rx_buffers - Replace used receive buffers; packet split
7992  *  @adapter: address of board private structure
7993  **/
7994 void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
7995 {
7996 	union e1000_adv_rx_desc *rx_desc;
7997 	struct igb_rx_buffer *bi;
7998 	u16 i = rx_ring->next_to_use;
7999 	u16 bufsz;
8000 
8001 	/* nothing to do */
8002 	if (!cleaned_count)
8003 		return;
8004 
8005 	rx_desc = IGB_RX_DESC(rx_ring, i);
8006 	bi = &rx_ring->rx_buffer_info[i];
8007 	i -= rx_ring->count;
8008 
8009 	bufsz = igb_rx_bufsz(rx_ring);
8010 
8011 	do {
8012 		if (!igb_alloc_mapped_page(rx_ring, bi))
8013 			break;
8014 
8015 		/* sync the buffer for use by the device */
8016 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
8017 						 bi->page_offset, bufsz,
8018 						 DMA_FROM_DEVICE);
8019 
8020 		/* Refresh the desc even if buffer_addrs didn't change
8021 		 * because each write-back erases this info.
8022 		 */
8023 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
8024 
8025 		rx_desc++;
8026 		bi++;
8027 		i++;
8028 		if (unlikely(!i)) {
8029 			rx_desc = IGB_RX_DESC(rx_ring, 0);
8030 			bi = rx_ring->rx_buffer_info;
8031 			i -= rx_ring->count;
8032 		}
8033 
8034 		/* clear the length for the next_to_use descriptor */
8035 		rx_desc->wb.upper.length = 0;
8036 
8037 		cleaned_count--;
8038 	} while (cleaned_count);
8039 
8040 	i += rx_ring->count;
8041 
8042 	if (rx_ring->next_to_use != i) {
8043 		/* record the next descriptor to use */
8044 		rx_ring->next_to_use = i;
8045 
8046 		/* update next to alloc since we have filled the ring */
8047 		rx_ring->next_to_alloc = i;
8048 
8049 		/* Force memory writes to complete before letting h/w
8050 		 * know there are new descriptors to fetch.  (Only
8051 		 * applicable for weak-ordered memory model archs,
8052 		 * such as IA-64).
8053 		 */
8054 		wmb();
8055 		writel(i, rx_ring->tail);
8056 	}
8057 }
8058 
8059 /**
8060  * igb_mii_ioctl -
8061  * @netdev:
8062  * @ifreq:
8063  * @cmd:
8064  **/
8065 static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
8066 {
8067 	struct igb_adapter *adapter = netdev_priv(netdev);
8068 	struct mii_ioctl_data *data = if_mii(ifr);
8069 
8070 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
8071 		return -EOPNOTSUPP;
8072 
8073 	switch (cmd) {
8074 	case SIOCGMIIPHY:
8075 		data->phy_id = adapter->hw.phy.addr;
8076 		break;
8077 	case SIOCGMIIREG:
8078 		if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
8079 				     &data->val_out))
8080 			return -EIO;
8081 		break;
8082 	case SIOCSMIIREG:
8083 	default:
8084 		return -EOPNOTSUPP;
8085 	}
8086 	return 0;
8087 }
8088 
8089 /**
8090  * igb_ioctl -
8091  * @netdev:
8092  * @ifreq:
8093  * @cmd:
8094  **/
8095 static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
8096 {
8097 	switch (cmd) {
8098 	case SIOCGMIIPHY:
8099 	case SIOCGMIIREG:
8100 	case SIOCSMIIREG:
8101 		return igb_mii_ioctl(netdev, ifr, cmd);
8102 	case SIOCGHWTSTAMP:
8103 		return igb_ptp_get_ts_config(netdev, ifr);
8104 	case SIOCSHWTSTAMP:
8105 		return igb_ptp_set_ts_config(netdev, ifr);
8106 	default:
8107 		return -EOPNOTSUPP;
8108 	}
8109 }
8110 
8111 void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
8112 {
8113 	struct igb_adapter *adapter = hw->back;
8114 
8115 	pci_read_config_word(adapter->pdev, reg, value);
8116 }
8117 
8118 void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
8119 {
8120 	struct igb_adapter *adapter = hw->back;
8121 
8122 	pci_write_config_word(adapter->pdev, reg, *value);
8123 }
8124 
8125 s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
8126 {
8127 	struct igb_adapter *adapter = hw->back;
8128 
8129 	if (pcie_capability_read_word(adapter->pdev, reg, value))
8130 		return -E1000_ERR_CONFIG;
8131 
8132 	return 0;
8133 }
8134 
8135 s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
8136 {
8137 	struct igb_adapter *adapter = hw->back;
8138 
8139 	if (pcie_capability_write_word(adapter->pdev, reg, *value))
8140 		return -E1000_ERR_CONFIG;
8141 
8142 	return 0;
8143 }
8144 
8145 static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
8146 {
8147 	struct igb_adapter *adapter = netdev_priv(netdev);
8148 	struct e1000_hw *hw = &adapter->hw;
8149 	u32 ctrl, rctl;
8150 	bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
8151 
8152 	if (enable) {
8153 		/* enable VLAN tag insert/strip */
8154 		ctrl = rd32(E1000_CTRL);
8155 		ctrl |= E1000_CTRL_VME;
8156 		wr32(E1000_CTRL, ctrl);
8157 
8158 		/* Disable CFI check */
8159 		rctl = rd32(E1000_RCTL);
8160 		rctl &= ~E1000_RCTL_CFIEN;
8161 		wr32(E1000_RCTL, rctl);
8162 	} else {
8163 		/* disable VLAN tag insert/strip */
8164 		ctrl = rd32(E1000_CTRL);
8165 		ctrl &= ~E1000_CTRL_VME;
8166 		wr32(E1000_CTRL, ctrl);
8167 	}
8168 
8169 	igb_set_vf_vlan_strip(adapter, adapter->vfs_allocated_count, enable);
8170 }
8171 
8172 static int igb_vlan_rx_add_vid(struct net_device *netdev,
8173 			       __be16 proto, u16 vid)
8174 {
8175 	struct igb_adapter *adapter = netdev_priv(netdev);
8176 	struct e1000_hw *hw = &adapter->hw;
8177 	int pf_id = adapter->vfs_allocated_count;
8178 
8179 	/* add the filter since PF can receive vlans w/o entry in vlvf */
8180 	if (!vid || !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
8181 		igb_vfta_set(hw, vid, pf_id, true, !!vid);
8182 
8183 	set_bit(vid, adapter->active_vlans);
8184 
8185 	return 0;
8186 }
8187 
8188 static int igb_vlan_rx_kill_vid(struct net_device *netdev,
8189 				__be16 proto, u16 vid)
8190 {
8191 	struct igb_adapter *adapter = netdev_priv(netdev);
8192 	int pf_id = adapter->vfs_allocated_count;
8193 	struct e1000_hw *hw = &adapter->hw;
8194 
8195 	/* remove VID from filter table */
8196 	if (vid && !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
8197 		igb_vfta_set(hw, vid, pf_id, false, true);
8198 
8199 	clear_bit(vid, adapter->active_vlans);
8200 
8201 	return 0;
8202 }
8203 
8204 static void igb_restore_vlan(struct igb_adapter *adapter)
8205 {
8206 	u16 vid = 1;
8207 
8208 	igb_vlan_mode(adapter->netdev, adapter->netdev->features);
8209 	igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
8210 
8211 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
8212 		igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
8213 }
8214 
8215 int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
8216 {
8217 	struct pci_dev *pdev = adapter->pdev;
8218 	struct e1000_mac_info *mac = &adapter->hw.mac;
8219 
8220 	mac->autoneg = 0;
8221 
8222 	/* Make sure dplx is at most 1 bit and lsb of speed is not set
8223 	 * for the switch() below to work
8224 	 */
8225 	if ((spd & 1) || (dplx & ~1))
8226 		goto err_inval;
8227 
8228 	/* Fiber NIC's only allow 1000 gbps Full duplex
8229 	 * and 100Mbps Full duplex for 100baseFx sfp
8230 	 */
8231 	if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
8232 		switch (spd + dplx) {
8233 		case SPEED_10 + DUPLEX_HALF:
8234 		case SPEED_10 + DUPLEX_FULL:
8235 		case SPEED_100 + DUPLEX_HALF:
8236 			goto err_inval;
8237 		default:
8238 			break;
8239 		}
8240 	}
8241 
8242 	switch (spd + dplx) {
8243 	case SPEED_10 + DUPLEX_HALF:
8244 		mac->forced_speed_duplex = ADVERTISE_10_HALF;
8245 		break;
8246 	case SPEED_10 + DUPLEX_FULL:
8247 		mac->forced_speed_duplex = ADVERTISE_10_FULL;
8248 		break;
8249 	case SPEED_100 + DUPLEX_HALF:
8250 		mac->forced_speed_duplex = ADVERTISE_100_HALF;
8251 		break;
8252 	case SPEED_100 + DUPLEX_FULL:
8253 		mac->forced_speed_duplex = ADVERTISE_100_FULL;
8254 		break;
8255 	case SPEED_1000 + DUPLEX_FULL:
8256 		mac->autoneg = 1;
8257 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
8258 		break;
8259 	case SPEED_1000 + DUPLEX_HALF: /* not supported */
8260 	default:
8261 		goto err_inval;
8262 	}
8263 
8264 	/* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
8265 	adapter->hw.phy.mdix = AUTO_ALL_MODES;
8266 
8267 	return 0;
8268 
8269 err_inval:
8270 	dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
8271 	return -EINVAL;
8272 }
8273 
8274 static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
8275 			  bool runtime)
8276 {
8277 	struct net_device *netdev = pci_get_drvdata(pdev);
8278 	struct igb_adapter *adapter = netdev_priv(netdev);
8279 	struct e1000_hw *hw = &adapter->hw;
8280 	u32 ctrl, rctl, status;
8281 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
8282 #ifdef CONFIG_PM
8283 	int retval = 0;
8284 #endif
8285 
8286 	rtnl_lock();
8287 	netif_device_detach(netdev);
8288 
8289 	if (netif_running(netdev))
8290 		__igb_close(netdev, true);
8291 
8292 	igb_ptp_suspend(adapter);
8293 
8294 	igb_clear_interrupt_scheme(adapter);
8295 	rtnl_unlock();
8296 
8297 #ifdef CONFIG_PM
8298 	retval = pci_save_state(pdev);
8299 	if (retval)
8300 		return retval;
8301 #endif
8302 
8303 	status = rd32(E1000_STATUS);
8304 	if (status & E1000_STATUS_LU)
8305 		wufc &= ~E1000_WUFC_LNKC;
8306 
8307 	if (wufc) {
8308 		igb_setup_rctl(adapter);
8309 		igb_set_rx_mode(netdev);
8310 
8311 		/* turn on all-multi mode if wake on multicast is enabled */
8312 		if (wufc & E1000_WUFC_MC) {
8313 			rctl = rd32(E1000_RCTL);
8314 			rctl |= E1000_RCTL_MPE;
8315 			wr32(E1000_RCTL, rctl);
8316 		}
8317 
8318 		ctrl = rd32(E1000_CTRL);
8319 		/* advertise wake from D3Cold */
8320 		#define E1000_CTRL_ADVD3WUC 0x00100000
8321 		/* phy power management enable */
8322 		#define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
8323 		ctrl |= E1000_CTRL_ADVD3WUC;
8324 		wr32(E1000_CTRL, ctrl);
8325 
8326 		/* Allow time for pending master requests to run */
8327 		igb_disable_pcie_master(hw);
8328 
8329 		wr32(E1000_WUC, E1000_WUC_PME_EN);
8330 		wr32(E1000_WUFC, wufc);
8331 	} else {
8332 		wr32(E1000_WUC, 0);
8333 		wr32(E1000_WUFC, 0);
8334 	}
8335 
8336 	*enable_wake = wufc || adapter->en_mng_pt;
8337 	if (!*enable_wake)
8338 		igb_power_down_link(adapter);
8339 	else
8340 		igb_power_up_link(adapter);
8341 
8342 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
8343 	 * would have already happened in close and is redundant.
8344 	 */
8345 	igb_release_hw_control(adapter);
8346 
8347 	pci_disable_device(pdev);
8348 
8349 	return 0;
8350 }
8351 
8352 static void igb_deliver_wake_packet(struct net_device *netdev)
8353 {
8354 	struct igb_adapter *adapter = netdev_priv(netdev);
8355 	struct e1000_hw *hw = &adapter->hw;
8356 	struct sk_buff *skb;
8357 	u32 wupl;
8358 
8359 	wupl = rd32(E1000_WUPL) & E1000_WUPL_MASK;
8360 
8361 	/* WUPM stores only the first 128 bytes of the wake packet.
8362 	 * Read the packet only if we have the whole thing.
8363 	 */
8364 	if ((wupl == 0) || (wupl > E1000_WUPM_BYTES))
8365 		return;
8366 
8367 	skb = netdev_alloc_skb_ip_align(netdev, E1000_WUPM_BYTES);
8368 	if (!skb)
8369 		return;
8370 
8371 	skb_put(skb, wupl);
8372 
8373 	/* Ensure reads are 32-bit aligned */
8374 	wupl = roundup(wupl, 4);
8375 
8376 	memcpy_fromio(skb->data, hw->hw_addr + E1000_WUPM_REG(0), wupl);
8377 
8378 	skb->protocol = eth_type_trans(skb, netdev);
8379 	netif_rx(skb);
8380 }
8381 
8382 static int __maybe_unused igb_suspend(struct device *dev)
8383 {
8384 	int retval;
8385 	bool wake;
8386 	struct pci_dev *pdev = to_pci_dev(dev);
8387 
8388 	retval = __igb_shutdown(pdev, &wake, 0);
8389 	if (retval)
8390 		return retval;
8391 
8392 	if (wake) {
8393 		pci_prepare_to_sleep(pdev);
8394 	} else {
8395 		pci_wake_from_d3(pdev, false);
8396 		pci_set_power_state(pdev, PCI_D3hot);
8397 	}
8398 
8399 	return 0;
8400 }
8401 
8402 static int __maybe_unused igb_resume(struct device *dev)
8403 {
8404 	struct pci_dev *pdev = to_pci_dev(dev);
8405 	struct net_device *netdev = pci_get_drvdata(pdev);
8406 	struct igb_adapter *adapter = netdev_priv(netdev);
8407 	struct e1000_hw *hw = &adapter->hw;
8408 	u32 err, val;
8409 
8410 	pci_set_power_state(pdev, PCI_D0);
8411 	pci_restore_state(pdev);
8412 	pci_save_state(pdev);
8413 
8414 	if (!pci_device_is_present(pdev))
8415 		return -ENODEV;
8416 	err = pci_enable_device_mem(pdev);
8417 	if (err) {
8418 		dev_err(&pdev->dev,
8419 			"igb: Cannot enable PCI device from suspend\n");
8420 		return err;
8421 	}
8422 	pci_set_master(pdev);
8423 
8424 	pci_enable_wake(pdev, PCI_D3hot, 0);
8425 	pci_enable_wake(pdev, PCI_D3cold, 0);
8426 
8427 	if (igb_init_interrupt_scheme(adapter, true)) {
8428 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8429 		return -ENOMEM;
8430 	}
8431 
8432 	igb_reset(adapter);
8433 
8434 	/* let the f/w know that the h/w is now under the control of the
8435 	 * driver.
8436 	 */
8437 	igb_get_hw_control(adapter);
8438 
8439 	val = rd32(E1000_WUS);
8440 	if (val & WAKE_PKT_WUS)
8441 		igb_deliver_wake_packet(netdev);
8442 
8443 	wr32(E1000_WUS, ~0);
8444 
8445 	rtnl_lock();
8446 	if (!err && netif_running(netdev))
8447 		err = __igb_open(netdev, true);
8448 
8449 	if (!err)
8450 		netif_device_attach(netdev);
8451 	rtnl_unlock();
8452 
8453 	return err;
8454 }
8455 
8456 static int __maybe_unused igb_runtime_idle(struct device *dev)
8457 {
8458 	struct pci_dev *pdev = to_pci_dev(dev);
8459 	struct net_device *netdev = pci_get_drvdata(pdev);
8460 	struct igb_adapter *adapter = netdev_priv(netdev);
8461 
8462 	if (!igb_has_link(adapter))
8463 		pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
8464 
8465 	return -EBUSY;
8466 }
8467 
8468 static int __maybe_unused igb_runtime_suspend(struct device *dev)
8469 {
8470 	struct pci_dev *pdev = to_pci_dev(dev);
8471 	int retval;
8472 	bool wake;
8473 
8474 	retval = __igb_shutdown(pdev, &wake, 1);
8475 	if (retval)
8476 		return retval;
8477 
8478 	if (wake) {
8479 		pci_prepare_to_sleep(pdev);
8480 	} else {
8481 		pci_wake_from_d3(pdev, false);
8482 		pci_set_power_state(pdev, PCI_D3hot);
8483 	}
8484 
8485 	return 0;
8486 }
8487 
8488 static int __maybe_unused igb_runtime_resume(struct device *dev)
8489 {
8490 	return igb_resume(dev);
8491 }
8492 
8493 static void igb_shutdown(struct pci_dev *pdev)
8494 {
8495 	bool wake;
8496 
8497 	__igb_shutdown(pdev, &wake, 0);
8498 
8499 	if (system_state == SYSTEM_POWER_OFF) {
8500 		pci_wake_from_d3(pdev, wake);
8501 		pci_set_power_state(pdev, PCI_D3hot);
8502 	}
8503 }
8504 
8505 #ifdef CONFIG_PCI_IOV
8506 static int igb_sriov_reinit(struct pci_dev *dev)
8507 {
8508 	struct net_device *netdev = pci_get_drvdata(dev);
8509 	struct igb_adapter *adapter = netdev_priv(netdev);
8510 	struct pci_dev *pdev = adapter->pdev;
8511 
8512 	rtnl_lock();
8513 
8514 	if (netif_running(netdev))
8515 		igb_close(netdev);
8516 	else
8517 		igb_reset(adapter);
8518 
8519 	igb_clear_interrupt_scheme(adapter);
8520 
8521 	igb_init_queue_configuration(adapter);
8522 
8523 	if (igb_init_interrupt_scheme(adapter, true)) {
8524 		rtnl_unlock();
8525 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8526 		return -ENOMEM;
8527 	}
8528 
8529 	if (netif_running(netdev))
8530 		igb_open(netdev);
8531 
8532 	rtnl_unlock();
8533 
8534 	return 0;
8535 }
8536 
8537 static int igb_pci_disable_sriov(struct pci_dev *dev)
8538 {
8539 	int err = igb_disable_sriov(dev);
8540 
8541 	if (!err)
8542 		err = igb_sriov_reinit(dev);
8543 
8544 	return err;
8545 }
8546 
8547 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
8548 {
8549 	int err = igb_enable_sriov(dev, num_vfs);
8550 
8551 	if (err)
8552 		goto out;
8553 
8554 	err = igb_sriov_reinit(dev);
8555 	if (!err)
8556 		return num_vfs;
8557 
8558 out:
8559 	return err;
8560 }
8561 
8562 #endif
8563 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
8564 {
8565 #ifdef CONFIG_PCI_IOV
8566 	if (num_vfs == 0)
8567 		return igb_pci_disable_sriov(dev);
8568 	else
8569 		return igb_pci_enable_sriov(dev, num_vfs);
8570 #endif
8571 	return 0;
8572 }
8573 
8574 #ifdef CONFIG_NET_POLL_CONTROLLER
8575 /* Polling 'interrupt' - used by things like netconsole to send skbs
8576  * without having to re-enable interrupts. It's not called while
8577  * the interrupt routine is executing.
8578  */
8579 static void igb_netpoll(struct net_device *netdev)
8580 {
8581 	struct igb_adapter *adapter = netdev_priv(netdev);
8582 	struct e1000_hw *hw = &adapter->hw;
8583 	struct igb_q_vector *q_vector;
8584 	int i;
8585 
8586 	for (i = 0; i < adapter->num_q_vectors; i++) {
8587 		q_vector = adapter->q_vector[i];
8588 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
8589 			wr32(E1000_EIMC, q_vector->eims_value);
8590 		else
8591 			igb_irq_disable(adapter);
8592 		napi_schedule(&q_vector->napi);
8593 	}
8594 }
8595 #endif /* CONFIG_NET_POLL_CONTROLLER */
8596 
8597 /**
8598  *  igb_io_error_detected - called when PCI error is detected
8599  *  @pdev: Pointer to PCI device
8600  *  @state: The current pci connection state
8601  *
8602  *  This function is called after a PCI bus error affecting
8603  *  this device has been detected.
8604  **/
8605 static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
8606 					      pci_channel_state_t state)
8607 {
8608 	struct net_device *netdev = pci_get_drvdata(pdev);
8609 	struct igb_adapter *adapter = netdev_priv(netdev);
8610 
8611 	netif_device_detach(netdev);
8612 
8613 	if (state == pci_channel_io_perm_failure)
8614 		return PCI_ERS_RESULT_DISCONNECT;
8615 
8616 	if (netif_running(netdev))
8617 		igb_down(adapter);
8618 	pci_disable_device(pdev);
8619 
8620 	/* Request a slot slot reset. */
8621 	return PCI_ERS_RESULT_NEED_RESET;
8622 }
8623 
8624 /**
8625  *  igb_io_slot_reset - called after the pci bus has been reset.
8626  *  @pdev: Pointer to PCI device
8627  *
8628  *  Restart the card from scratch, as if from a cold-boot. Implementation
8629  *  resembles the first-half of the igb_resume routine.
8630  **/
8631 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
8632 {
8633 	struct net_device *netdev = pci_get_drvdata(pdev);
8634 	struct igb_adapter *adapter = netdev_priv(netdev);
8635 	struct e1000_hw *hw = &adapter->hw;
8636 	pci_ers_result_t result;
8637 	int err;
8638 
8639 	if (pci_enable_device_mem(pdev)) {
8640 		dev_err(&pdev->dev,
8641 			"Cannot re-enable PCI device after reset.\n");
8642 		result = PCI_ERS_RESULT_DISCONNECT;
8643 	} else {
8644 		pci_set_master(pdev);
8645 		pci_restore_state(pdev);
8646 		pci_save_state(pdev);
8647 
8648 		pci_enable_wake(pdev, PCI_D3hot, 0);
8649 		pci_enable_wake(pdev, PCI_D3cold, 0);
8650 
8651 		/* In case of PCI error, adapter lose its HW address
8652 		 * so we should re-assign it here.
8653 		 */
8654 		hw->hw_addr = adapter->io_addr;
8655 
8656 		igb_reset(adapter);
8657 		wr32(E1000_WUS, ~0);
8658 		result = PCI_ERS_RESULT_RECOVERED;
8659 	}
8660 
8661 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
8662 	if (err) {
8663 		dev_err(&pdev->dev,
8664 			"pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
8665 			err);
8666 		/* non-fatal, continue */
8667 	}
8668 
8669 	return result;
8670 }
8671 
8672 /**
8673  *  igb_io_resume - called when traffic can start flowing again.
8674  *  @pdev: Pointer to PCI device
8675  *
8676  *  This callback is called when the error recovery driver tells us that
8677  *  its OK to resume normal operation. Implementation resembles the
8678  *  second-half of the igb_resume routine.
8679  */
8680 static void igb_io_resume(struct pci_dev *pdev)
8681 {
8682 	struct net_device *netdev = pci_get_drvdata(pdev);
8683 	struct igb_adapter *adapter = netdev_priv(netdev);
8684 
8685 	if (netif_running(netdev)) {
8686 		if (igb_up(adapter)) {
8687 			dev_err(&pdev->dev, "igb_up failed after reset\n");
8688 			return;
8689 		}
8690 	}
8691 
8692 	netif_device_attach(netdev);
8693 
8694 	/* let the f/w know that the h/w is now under the control of the
8695 	 * driver.
8696 	 */
8697 	igb_get_hw_control(adapter);
8698 }
8699 
8700 /**
8701  *  igb_rar_set_index - Sync RAL[index] and RAH[index] registers with MAC table
8702  *  @adapter: Pointer to adapter structure
8703  *  @index: Index of the RAR entry which need to be synced with MAC table
8704  **/
8705 static void igb_rar_set_index(struct igb_adapter *adapter, u32 index)
8706 {
8707 	struct e1000_hw *hw = &adapter->hw;
8708 	u32 rar_low, rar_high;
8709 	u8 *addr = adapter->mac_table[index].addr;
8710 
8711 	/* HW expects these to be in network order when they are plugged
8712 	 * into the registers which are little endian.  In order to guarantee
8713 	 * that ordering we need to do an leXX_to_cpup here in order to be
8714 	 * ready for the byteswap that occurs with writel
8715 	 */
8716 	rar_low = le32_to_cpup((__le32 *)(addr));
8717 	rar_high = le16_to_cpup((__le16 *)(addr + 4));
8718 
8719 	/* Indicate to hardware the Address is Valid. */
8720 	if (adapter->mac_table[index].state & IGB_MAC_STATE_IN_USE) {
8721 		rar_high |= E1000_RAH_AV;
8722 
8723 		if (hw->mac.type == e1000_82575)
8724 			rar_high |= E1000_RAH_POOL_1 *
8725 				    adapter->mac_table[index].queue;
8726 		else
8727 			rar_high |= E1000_RAH_POOL_1 <<
8728 				    adapter->mac_table[index].queue;
8729 	}
8730 
8731 	wr32(E1000_RAL(index), rar_low);
8732 	wrfl();
8733 	wr32(E1000_RAH(index), rar_high);
8734 	wrfl();
8735 }
8736 
8737 static int igb_set_vf_mac(struct igb_adapter *adapter,
8738 			  int vf, unsigned char *mac_addr)
8739 {
8740 	struct e1000_hw *hw = &adapter->hw;
8741 	/* VF MAC addresses start at end of receive addresses and moves
8742 	 * towards the first, as a result a collision should not be possible
8743 	 */
8744 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
8745 	unsigned char *vf_mac_addr = adapter->vf_data[vf].vf_mac_addresses;
8746 
8747 	ether_addr_copy(vf_mac_addr, mac_addr);
8748 	ether_addr_copy(adapter->mac_table[rar_entry].addr, mac_addr);
8749 	adapter->mac_table[rar_entry].queue = vf;
8750 	adapter->mac_table[rar_entry].state |= IGB_MAC_STATE_IN_USE;
8751 	igb_rar_set_index(adapter, rar_entry);
8752 
8753 	return 0;
8754 }
8755 
8756 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
8757 {
8758 	struct igb_adapter *adapter = netdev_priv(netdev);
8759 	if (!is_valid_ether_addr(mac) || (vf >= adapter->vfs_allocated_count))
8760 		return -EINVAL;
8761 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
8762 	dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n", mac, vf);
8763 	dev_info(&adapter->pdev->dev,
8764 		 "Reload the VF driver to make this change effective.");
8765 	if (test_bit(__IGB_DOWN, &adapter->state)) {
8766 		dev_warn(&adapter->pdev->dev,
8767 			 "The VF MAC address has been set, but the PF device is not up.\n");
8768 		dev_warn(&adapter->pdev->dev,
8769 			 "Bring the PF device up before attempting to use the VF device.\n");
8770 	}
8771 	return igb_set_vf_mac(adapter, vf, mac);
8772 }
8773 
8774 static int igb_link_mbps(int internal_link_speed)
8775 {
8776 	switch (internal_link_speed) {
8777 	case SPEED_100:
8778 		return 100;
8779 	case SPEED_1000:
8780 		return 1000;
8781 	default:
8782 		return 0;
8783 	}
8784 }
8785 
8786 static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
8787 				  int link_speed)
8788 {
8789 	int rf_dec, rf_int;
8790 	u32 bcnrc_val;
8791 
8792 	if (tx_rate != 0) {
8793 		/* Calculate the rate factor values to set */
8794 		rf_int = link_speed / tx_rate;
8795 		rf_dec = (link_speed - (rf_int * tx_rate));
8796 		rf_dec = (rf_dec * BIT(E1000_RTTBCNRC_RF_INT_SHIFT)) /
8797 			 tx_rate;
8798 
8799 		bcnrc_val = E1000_RTTBCNRC_RS_ENA;
8800 		bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
8801 			      E1000_RTTBCNRC_RF_INT_MASK);
8802 		bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
8803 	} else {
8804 		bcnrc_val = 0;
8805 	}
8806 
8807 	wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
8808 	/* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
8809 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
8810 	 */
8811 	wr32(E1000_RTTBCNRM, 0x14);
8812 	wr32(E1000_RTTBCNRC, bcnrc_val);
8813 }
8814 
8815 static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
8816 {
8817 	int actual_link_speed, i;
8818 	bool reset_rate = false;
8819 
8820 	/* VF TX rate limit was not set or not supported */
8821 	if ((adapter->vf_rate_link_speed == 0) ||
8822 	    (adapter->hw.mac.type != e1000_82576))
8823 		return;
8824 
8825 	actual_link_speed = igb_link_mbps(adapter->link_speed);
8826 	if (actual_link_speed != adapter->vf_rate_link_speed) {
8827 		reset_rate = true;
8828 		adapter->vf_rate_link_speed = 0;
8829 		dev_info(&adapter->pdev->dev,
8830 			 "Link speed has been changed. VF Transmit rate is disabled\n");
8831 	}
8832 
8833 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
8834 		if (reset_rate)
8835 			adapter->vf_data[i].tx_rate = 0;
8836 
8837 		igb_set_vf_rate_limit(&adapter->hw, i,
8838 				      adapter->vf_data[i].tx_rate,
8839 				      actual_link_speed);
8840 	}
8841 }
8842 
8843 static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
8844 			     int min_tx_rate, int max_tx_rate)
8845 {
8846 	struct igb_adapter *adapter = netdev_priv(netdev);
8847 	struct e1000_hw *hw = &adapter->hw;
8848 	int actual_link_speed;
8849 
8850 	if (hw->mac.type != e1000_82576)
8851 		return -EOPNOTSUPP;
8852 
8853 	if (min_tx_rate)
8854 		return -EINVAL;
8855 
8856 	actual_link_speed = igb_link_mbps(adapter->link_speed);
8857 	if ((vf >= adapter->vfs_allocated_count) ||
8858 	    (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
8859 	    (max_tx_rate < 0) ||
8860 	    (max_tx_rate > actual_link_speed))
8861 		return -EINVAL;
8862 
8863 	adapter->vf_rate_link_speed = actual_link_speed;
8864 	adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
8865 	igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
8866 
8867 	return 0;
8868 }
8869 
8870 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
8871 				   bool setting)
8872 {
8873 	struct igb_adapter *adapter = netdev_priv(netdev);
8874 	struct e1000_hw *hw = &adapter->hw;
8875 	u32 reg_val, reg_offset;
8876 
8877 	if (!adapter->vfs_allocated_count)
8878 		return -EOPNOTSUPP;
8879 
8880 	if (vf >= adapter->vfs_allocated_count)
8881 		return -EINVAL;
8882 
8883 	reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
8884 	reg_val = rd32(reg_offset);
8885 	if (setting)
8886 		reg_val |= (BIT(vf) |
8887 			    BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
8888 	else
8889 		reg_val &= ~(BIT(vf) |
8890 			     BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
8891 	wr32(reg_offset, reg_val);
8892 
8893 	adapter->vf_data[vf].spoofchk_enabled = setting;
8894 	return 0;
8895 }
8896 
8897 static int igb_ndo_get_vf_config(struct net_device *netdev,
8898 				 int vf, struct ifla_vf_info *ivi)
8899 {
8900 	struct igb_adapter *adapter = netdev_priv(netdev);
8901 	if (vf >= adapter->vfs_allocated_count)
8902 		return -EINVAL;
8903 	ivi->vf = vf;
8904 	memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
8905 	ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
8906 	ivi->min_tx_rate = 0;
8907 	ivi->vlan = adapter->vf_data[vf].pf_vlan;
8908 	ivi->qos = adapter->vf_data[vf].pf_qos;
8909 	ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
8910 	return 0;
8911 }
8912 
8913 static void igb_vmm_control(struct igb_adapter *adapter)
8914 {
8915 	struct e1000_hw *hw = &adapter->hw;
8916 	u32 reg;
8917 
8918 	switch (hw->mac.type) {
8919 	case e1000_82575:
8920 	case e1000_i210:
8921 	case e1000_i211:
8922 	case e1000_i354:
8923 	default:
8924 		/* replication is not supported for 82575 */
8925 		return;
8926 	case e1000_82576:
8927 		/* notify HW that the MAC is adding vlan tags */
8928 		reg = rd32(E1000_DTXCTL);
8929 		reg |= E1000_DTXCTL_VLAN_ADDED;
8930 		wr32(E1000_DTXCTL, reg);
8931 		/* Fall through */
8932 	case e1000_82580:
8933 		/* enable replication vlan tag stripping */
8934 		reg = rd32(E1000_RPLOLR);
8935 		reg |= E1000_RPLOLR_STRVLAN;
8936 		wr32(E1000_RPLOLR, reg);
8937 		/* Fall through */
8938 	case e1000_i350:
8939 		/* none of the above registers are supported by i350 */
8940 		break;
8941 	}
8942 
8943 	if (adapter->vfs_allocated_count) {
8944 		igb_vmdq_set_loopback_pf(hw, true);
8945 		igb_vmdq_set_replication_pf(hw, true);
8946 		igb_vmdq_set_anti_spoofing_pf(hw, true,
8947 					      adapter->vfs_allocated_count);
8948 	} else {
8949 		igb_vmdq_set_loopback_pf(hw, false);
8950 		igb_vmdq_set_replication_pf(hw, false);
8951 	}
8952 }
8953 
8954 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
8955 {
8956 	struct e1000_hw *hw = &adapter->hw;
8957 	u32 dmac_thr;
8958 	u16 hwm;
8959 
8960 	if (hw->mac.type > e1000_82580) {
8961 		if (adapter->flags & IGB_FLAG_DMAC) {
8962 			u32 reg;
8963 
8964 			/* force threshold to 0. */
8965 			wr32(E1000_DMCTXTH, 0);
8966 
8967 			/* DMA Coalescing high water mark needs to be greater
8968 			 * than the Rx threshold. Set hwm to PBA - max frame
8969 			 * size in 16B units, capping it at PBA - 6KB.
8970 			 */
8971 			hwm = 64 * (pba - 6);
8972 			reg = rd32(E1000_FCRTC);
8973 			reg &= ~E1000_FCRTC_RTH_COAL_MASK;
8974 			reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
8975 				& E1000_FCRTC_RTH_COAL_MASK);
8976 			wr32(E1000_FCRTC, reg);
8977 
8978 			/* Set the DMA Coalescing Rx threshold to PBA - 2 * max
8979 			 * frame size, capping it at PBA - 10KB.
8980 			 */
8981 			dmac_thr = pba - 10;
8982 			reg = rd32(E1000_DMACR);
8983 			reg &= ~E1000_DMACR_DMACTHR_MASK;
8984 			reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
8985 				& E1000_DMACR_DMACTHR_MASK);
8986 
8987 			/* transition to L0x or L1 if available..*/
8988 			reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
8989 
8990 			/* watchdog timer= +-1000 usec in 32usec intervals */
8991 			reg |= (1000 >> 5);
8992 
8993 			/* Disable BMC-to-OS Watchdog Enable */
8994 			if (hw->mac.type != e1000_i354)
8995 				reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
8996 
8997 			wr32(E1000_DMACR, reg);
8998 
8999 			/* no lower threshold to disable
9000 			 * coalescing(smart fifb)-UTRESH=0
9001 			 */
9002 			wr32(E1000_DMCRTRH, 0);
9003 
9004 			reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
9005 
9006 			wr32(E1000_DMCTLX, reg);
9007 
9008 			/* free space in tx packet buffer to wake from
9009 			 * DMA coal
9010 			 */
9011 			wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
9012 			     (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
9013 
9014 			/* make low power state decision controlled
9015 			 * by DMA coal
9016 			 */
9017 			reg = rd32(E1000_PCIEMISC);
9018 			reg &= ~E1000_PCIEMISC_LX_DECISION;
9019 			wr32(E1000_PCIEMISC, reg);
9020 		} /* endif adapter->dmac is not disabled */
9021 	} else if (hw->mac.type == e1000_82580) {
9022 		u32 reg = rd32(E1000_PCIEMISC);
9023 
9024 		wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
9025 		wr32(E1000_DMACR, 0);
9026 	}
9027 }
9028 
9029 /**
9030  *  igb_read_i2c_byte - Reads 8 bit word over I2C
9031  *  @hw: pointer to hardware structure
9032  *  @byte_offset: byte offset to read
9033  *  @dev_addr: device address
9034  *  @data: value read
9035  *
9036  *  Performs byte read operation over I2C interface at
9037  *  a specified device address.
9038  **/
9039 s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
9040 		      u8 dev_addr, u8 *data)
9041 {
9042 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
9043 	struct i2c_client *this_client = adapter->i2c_client;
9044 	s32 status;
9045 	u16 swfw_mask = 0;
9046 
9047 	if (!this_client)
9048 		return E1000_ERR_I2C;
9049 
9050 	swfw_mask = E1000_SWFW_PHY0_SM;
9051 
9052 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
9053 		return E1000_ERR_SWFW_SYNC;
9054 
9055 	status = i2c_smbus_read_byte_data(this_client, byte_offset);
9056 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
9057 
9058 	if (status < 0)
9059 		return E1000_ERR_I2C;
9060 	else {
9061 		*data = status;
9062 		return 0;
9063 	}
9064 }
9065 
9066 /**
9067  *  igb_write_i2c_byte - Writes 8 bit word over I2C
9068  *  @hw: pointer to hardware structure
9069  *  @byte_offset: byte offset to write
9070  *  @dev_addr: device address
9071  *  @data: value to write
9072  *
9073  *  Performs byte write operation over I2C interface at
9074  *  a specified device address.
9075  **/
9076 s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
9077 		       u8 dev_addr, u8 data)
9078 {
9079 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
9080 	struct i2c_client *this_client = adapter->i2c_client;
9081 	s32 status;
9082 	u16 swfw_mask = E1000_SWFW_PHY0_SM;
9083 
9084 	if (!this_client)
9085 		return E1000_ERR_I2C;
9086 
9087 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
9088 		return E1000_ERR_SWFW_SYNC;
9089 	status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
9090 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
9091 
9092 	if (status)
9093 		return E1000_ERR_I2C;
9094 	else
9095 		return 0;
9096 
9097 }
9098 
9099 int igb_reinit_queues(struct igb_adapter *adapter)
9100 {
9101 	struct net_device *netdev = adapter->netdev;
9102 	struct pci_dev *pdev = adapter->pdev;
9103 	int err = 0;
9104 
9105 	if (netif_running(netdev))
9106 		igb_close(netdev);
9107 
9108 	igb_reset_interrupt_capability(adapter);
9109 
9110 	if (igb_init_interrupt_scheme(adapter, true)) {
9111 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9112 		return -ENOMEM;
9113 	}
9114 
9115 	if (netif_running(netdev))
9116 		err = igb_open(netdev);
9117 
9118 	return err;
9119 }
9120 
9121 static void igb_nfc_filter_exit(struct igb_adapter *adapter)
9122 {
9123 	struct igb_nfc_filter *rule;
9124 
9125 	spin_lock(&adapter->nfc_lock);
9126 
9127 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
9128 		igb_erase_filter(adapter, rule);
9129 
9130 	spin_unlock(&adapter->nfc_lock);
9131 }
9132 
9133 static void igb_nfc_filter_restore(struct igb_adapter *adapter)
9134 {
9135 	struct igb_nfc_filter *rule;
9136 
9137 	spin_lock(&adapter->nfc_lock);
9138 
9139 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
9140 		igb_add_filter(adapter, rule);
9141 
9142 	spin_unlock(&adapter->nfc_lock);
9143 }
9144 /* igb_main.c */
9145