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 <linux/net_tstamp.h>
38 #include <linux/mii.h>
39 #include <linux/ethtool.h>
40 #include <linux/if.h>
41 #include <linux/if_vlan.h>
42 #include <linux/pci.h>
43 #include <linux/pci-aspm.h>
44 #include <linux/delay.h>
45 #include <linux/interrupt.h>
46 #include <linux/ip.h>
47 #include <linux/tcp.h>
48 #include <linux/sctp.h>
49 #include <linux/if_ether.h>
50 #include <linux/aer.h>
51 #include <linux/prefetch.h>
52 #include <linux/pm_runtime.h>
53 #ifdef CONFIG_IGB_DCA
54 #include <linux/dca.h>
55 #endif
56 #include <linux/i2c.h>
57 #include "igb.h"
58 
59 #define MAJ 5
60 #define MIN 2
61 #define BUILD 15
62 #define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
63 __stringify(BUILD) "-k"
64 char igb_driver_name[] = "igb";
65 char igb_driver_version[] = DRV_VERSION;
66 static const char igb_driver_string[] =
67 				"Intel(R) Gigabit Ethernet Network Driver";
68 static const char igb_copyright[] =
69 				"Copyright (c) 2007-2014 Intel Corporation.";
70 
71 static const struct e1000_info *igb_info_tbl[] = {
72 	[board_82575] = &e1000_82575_info,
73 };
74 
75 static const struct pci_device_id igb_pci_tbl[] = {
76 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
77 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_SGMII) },
78 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
79 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I211_COPPER), board_82575 },
80 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER), board_82575 },
81 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_FIBER), board_82575 },
82 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES), board_82575 },
83 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SGMII), board_82575 },
84 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER_FLASHLESS), board_82575 },
85 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES_FLASHLESS), board_82575 },
86 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_COPPER), board_82575 },
87 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_FIBER), board_82575 },
88 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SERDES), board_82575 },
89 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SGMII), board_82575 },
90 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER), board_82575 },
91 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_FIBER), board_82575 },
92 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_QUAD_FIBER), board_82575 },
93 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SERDES), board_82575 },
94 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SGMII), board_82575 },
95 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER_DUAL), board_82575 },
96 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SGMII), board_82575 },
97 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SERDES), board_82575 },
98 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_BACKPLANE), board_82575 },
99 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SFP), board_82575 },
100 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
101 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
102 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
103 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
104 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
105 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
106 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER_ET2), board_82575 },
107 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
108 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
109 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
110 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
111 	/* required last entry */
112 	{0, }
113 };
114 
115 MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
116 
117 static int igb_setup_all_tx_resources(struct igb_adapter *);
118 static int igb_setup_all_rx_resources(struct igb_adapter *);
119 static void igb_free_all_tx_resources(struct igb_adapter *);
120 static void igb_free_all_rx_resources(struct igb_adapter *);
121 static void igb_setup_mrqc(struct igb_adapter *);
122 static int igb_probe(struct pci_dev *, const struct pci_device_id *);
123 static void igb_remove(struct pci_dev *pdev);
124 static int igb_sw_init(struct igb_adapter *);
125 static int igb_open(struct net_device *);
126 static int igb_close(struct net_device *);
127 static void igb_configure(struct igb_adapter *);
128 static void igb_configure_tx(struct igb_adapter *);
129 static void igb_configure_rx(struct igb_adapter *);
130 static void igb_clean_all_tx_rings(struct igb_adapter *);
131 static void igb_clean_all_rx_rings(struct igb_adapter *);
132 static void igb_clean_tx_ring(struct igb_ring *);
133 static void igb_clean_rx_ring(struct igb_ring *);
134 static void igb_set_rx_mode(struct net_device *);
135 static void igb_update_phy_info(unsigned long);
136 static void igb_watchdog(unsigned long);
137 static void igb_watchdog_task(struct work_struct *);
138 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb, struct net_device *);
139 static struct rtnl_link_stats64 *igb_get_stats64(struct net_device *dev,
140 					  struct rtnl_link_stats64 *stats);
141 static int igb_change_mtu(struct net_device *, int);
142 static int igb_set_mac(struct net_device *, void *);
143 static void igb_set_uta(struct igb_adapter *adapter);
144 static irqreturn_t igb_intr(int irq, void *);
145 static irqreturn_t igb_intr_msi(int irq, void *);
146 static irqreturn_t igb_msix_other(int irq, void *);
147 static irqreturn_t igb_msix_ring(int irq, void *);
148 #ifdef CONFIG_IGB_DCA
149 static void igb_update_dca(struct igb_q_vector *);
150 static void igb_setup_dca(struct igb_adapter *);
151 #endif /* CONFIG_IGB_DCA */
152 static int igb_poll(struct napi_struct *, int);
153 static bool igb_clean_tx_irq(struct igb_q_vector *);
154 static bool igb_clean_rx_irq(struct igb_q_vector *, int);
155 static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
156 static void igb_tx_timeout(struct net_device *);
157 static void igb_reset_task(struct work_struct *);
158 static void igb_vlan_mode(struct net_device *netdev,
159 			  netdev_features_t features);
160 static int igb_vlan_rx_add_vid(struct net_device *, __be16, u16);
161 static int igb_vlan_rx_kill_vid(struct net_device *, __be16, u16);
162 static void igb_restore_vlan(struct igb_adapter *);
163 static void igb_rar_set_qsel(struct igb_adapter *, u8 *, u32 , u8);
164 static void igb_ping_all_vfs(struct igb_adapter *);
165 static void igb_msg_task(struct igb_adapter *);
166 static void igb_vmm_control(struct igb_adapter *);
167 static int igb_set_vf_mac(struct igb_adapter *, int, unsigned char *);
168 static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
169 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac);
170 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
171 			       int vf, u16 vlan, u8 qos);
172 static int igb_ndo_set_vf_bw(struct net_device *, int, int, int);
173 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
174 				   bool setting);
175 static int igb_ndo_get_vf_config(struct net_device *netdev, int vf,
176 				 struct ifla_vf_info *ivi);
177 static void igb_check_vf_rate_limit(struct igb_adapter *);
178 
179 #ifdef CONFIG_PCI_IOV
180 static int igb_vf_configure(struct igb_adapter *adapter, int vf);
181 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs);
182 #endif
183 
184 #ifdef CONFIG_PM
185 #ifdef CONFIG_PM_SLEEP
186 static int igb_suspend(struct device *);
187 #endif
188 static int igb_resume(struct device *);
189 static int igb_runtime_suspend(struct device *dev);
190 static int igb_runtime_resume(struct device *dev);
191 static int igb_runtime_idle(struct device *dev);
192 static const struct dev_pm_ops igb_pm_ops = {
193 	SET_SYSTEM_SLEEP_PM_OPS(igb_suspend, igb_resume)
194 	SET_RUNTIME_PM_OPS(igb_runtime_suspend, igb_runtime_resume,
195 			igb_runtime_idle)
196 };
197 #endif
198 static void igb_shutdown(struct pci_dev *);
199 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs);
200 #ifdef CONFIG_IGB_DCA
201 static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
202 static struct notifier_block dca_notifier = {
203 	.notifier_call	= igb_notify_dca,
204 	.next		= NULL,
205 	.priority	= 0
206 };
207 #endif
208 #ifdef CONFIG_NET_POLL_CONTROLLER
209 /* for netdump / net console */
210 static void igb_netpoll(struct net_device *);
211 #endif
212 #ifdef CONFIG_PCI_IOV
213 static unsigned int max_vfs;
214 module_param(max_vfs, uint, 0);
215 MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate per physical function");
216 #endif /* CONFIG_PCI_IOV */
217 
218 static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
219 		     pci_channel_state_t);
220 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
221 static void igb_io_resume(struct pci_dev *);
222 
223 static const struct pci_error_handlers igb_err_handler = {
224 	.error_detected = igb_io_error_detected,
225 	.slot_reset = igb_io_slot_reset,
226 	.resume = igb_io_resume,
227 };
228 
229 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba);
230 
231 static struct pci_driver igb_driver = {
232 	.name     = igb_driver_name,
233 	.id_table = igb_pci_tbl,
234 	.probe    = igb_probe,
235 	.remove   = igb_remove,
236 #ifdef CONFIG_PM
237 	.driver.pm = &igb_pm_ops,
238 #endif
239 	.shutdown = igb_shutdown,
240 	.sriov_configure = igb_pci_sriov_configure,
241 	.err_handler = &igb_err_handler
242 };
243 
244 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
245 MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
246 MODULE_LICENSE("GPL");
247 MODULE_VERSION(DRV_VERSION);
248 
249 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
250 static int debug = -1;
251 module_param(debug, int, 0);
252 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
253 
254 struct igb_reg_info {
255 	u32 ofs;
256 	char *name;
257 };
258 
259 static const struct igb_reg_info igb_reg_info_tbl[] = {
260 
261 	/* General Registers */
262 	{E1000_CTRL, "CTRL"},
263 	{E1000_STATUS, "STATUS"},
264 	{E1000_CTRL_EXT, "CTRL_EXT"},
265 
266 	/* Interrupt Registers */
267 	{E1000_ICR, "ICR"},
268 
269 	/* RX Registers */
270 	{E1000_RCTL, "RCTL"},
271 	{E1000_RDLEN(0), "RDLEN"},
272 	{E1000_RDH(0), "RDH"},
273 	{E1000_RDT(0), "RDT"},
274 	{E1000_RXDCTL(0), "RXDCTL"},
275 	{E1000_RDBAL(0), "RDBAL"},
276 	{E1000_RDBAH(0), "RDBAH"},
277 
278 	/* TX Registers */
279 	{E1000_TCTL, "TCTL"},
280 	{E1000_TDBAL(0), "TDBAL"},
281 	{E1000_TDBAH(0), "TDBAH"},
282 	{E1000_TDLEN(0), "TDLEN"},
283 	{E1000_TDH(0), "TDH"},
284 	{E1000_TDT(0), "TDT"},
285 	{E1000_TXDCTL(0), "TXDCTL"},
286 	{E1000_TDFH, "TDFH"},
287 	{E1000_TDFT, "TDFT"},
288 	{E1000_TDFHS, "TDFHS"},
289 	{E1000_TDFPC, "TDFPC"},
290 
291 	/* List Terminator */
292 	{}
293 };
294 
295 /* igb_regdump - register printout routine */
296 static void igb_regdump(struct e1000_hw *hw, struct igb_reg_info *reginfo)
297 {
298 	int n = 0;
299 	char rname[16];
300 	u32 regs[8];
301 
302 	switch (reginfo->ofs) {
303 	case E1000_RDLEN(0):
304 		for (n = 0; n < 4; n++)
305 			regs[n] = rd32(E1000_RDLEN(n));
306 		break;
307 	case E1000_RDH(0):
308 		for (n = 0; n < 4; n++)
309 			regs[n] = rd32(E1000_RDH(n));
310 		break;
311 	case E1000_RDT(0):
312 		for (n = 0; n < 4; n++)
313 			regs[n] = rd32(E1000_RDT(n));
314 		break;
315 	case E1000_RXDCTL(0):
316 		for (n = 0; n < 4; n++)
317 			regs[n] = rd32(E1000_RXDCTL(n));
318 		break;
319 	case E1000_RDBAL(0):
320 		for (n = 0; n < 4; n++)
321 			regs[n] = rd32(E1000_RDBAL(n));
322 		break;
323 	case E1000_RDBAH(0):
324 		for (n = 0; n < 4; n++)
325 			regs[n] = rd32(E1000_RDBAH(n));
326 		break;
327 	case E1000_TDBAL(0):
328 		for (n = 0; n < 4; n++)
329 			regs[n] = rd32(E1000_RDBAL(n));
330 		break;
331 	case E1000_TDBAH(0):
332 		for (n = 0; n < 4; n++)
333 			regs[n] = rd32(E1000_TDBAH(n));
334 		break;
335 	case E1000_TDLEN(0):
336 		for (n = 0; n < 4; n++)
337 			regs[n] = rd32(E1000_TDLEN(n));
338 		break;
339 	case E1000_TDH(0):
340 		for (n = 0; n < 4; n++)
341 			regs[n] = rd32(E1000_TDH(n));
342 		break;
343 	case E1000_TDT(0):
344 		for (n = 0; n < 4; n++)
345 			regs[n] = rd32(E1000_TDT(n));
346 		break;
347 	case E1000_TXDCTL(0):
348 		for (n = 0; n < 4; n++)
349 			regs[n] = rd32(E1000_TXDCTL(n));
350 		break;
351 	default:
352 		pr_info("%-15s %08x\n", reginfo->name, rd32(reginfo->ofs));
353 		return;
354 	}
355 
356 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-3]");
357 	pr_info("%-15s %08x %08x %08x %08x\n", rname, regs[0], regs[1],
358 		regs[2], regs[3]);
359 }
360 
361 /* igb_dump - Print registers, Tx-rings and Rx-rings */
362 static void igb_dump(struct igb_adapter *adapter)
363 {
364 	struct net_device *netdev = adapter->netdev;
365 	struct e1000_hw *hw = &adapter->hw;
366 	struct igb_reg_info *reginfo;
367 	struct igb_ring *tx_ring;
368 	union e1000_adv_tx_desc *tx_desc;
369 	struct my_u0 { u64 a; u64 b; } *u0;
370 	struct igb_ring *rx_ring;
371 	union e1000_adv_rx_desc *rx_desc;
372 	u32 staterr;
373 	u16 i, n;
374 
375 	if (!netif_msg_hw(adapter))
376 		return;
377 
378 	/* Print netdevice Info */
379 	if (netdev) {
380 		dev_info(&adapter->pdev->dev, "Net device Info\n");
381 		pr_info("Device Name     state            trans_start      last_rx\n");
382 		pr_info("%-15s %016lX %016lX %016lX\n", netdev->name,
383 			netdev->state, netdev->trans_start, netdev->last_rx);
384 	}
385 
386 	/* Print Registers */
387 	dev_info(&adapter->pdev->dev, "Register Dump\n");
388 	pr_info(" Register Name   Value\n");
389 	for (reginfo = (struct igb_reg_info *)igb_reg_info_tbl;
390 	     reginfo->name; reginfo++) {
391 		igb_regdump(hw, reginfo);
392 	}
393 
394 	/* Print TX Ring Summary */
395 	if (!netdev || !netif_running(netdev))
396 		goto exit;
397 
398 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
399 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
400 	for (n = 0; n < adapter->num_tx_queues; n++) {
401 		struct igb_tx_buffer *buffer_info;
402 		tx_ring = adapter->tx_ring[n];
403 		buffer_info = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
404 		pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
405 			n, tx_ring->next_to_use, tx_ring->next_to_clean,
406 			(u64)dma_unmap_addr(buffer_info, dma),
407 			dma_unmap_len(buffer_info, len),
408 			buffer_info->next_to_watch,
409 			(u64)buffer_info->time_stamp);
410 	}
411 
412 	/* Print TX Rings */
413 	if (!netif_msg_tx_done(adapter))
414 		goto rx_ring_summary;
415 
416 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
417 
418 	/* Transmit Descriptor Formats
419 	 *
420 	 * Advanced Transmit Descriptor
421 	 *   +--------------------------------------------------------------+
422 	 * 0 |         Buffer Address [63:0]                                |
423 	 *   +--------------------------------------------------------------+
424 	 * 8 | PAYLEN  | PORTS  |CC|IDX | STA | DCMD  |DTYP|MAC|RSV| DTALEN |
425 	 *   +--------------------------------------------------------------+
426 	 *   63      46 45    40 39 38 36 35 32 31   24             15       0
427 	 */
428 
429 	for (n = 0; n < adapter->num_tx_queues; n++) {
430 		tx_ring = adapter->tx_ring[n];
431 		pr_info("------------------------------------\n");
432 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
433 		pr_info("------------------------------------\n");
434 		pr_info("T [desc]     [address 63:0  ] [PlPOCIStDDM Ln] [bi->dma       ] leng  ntw timestamp        bi->skb\n");
435 
436 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
437 			const char *next_desc;
438 			struct igb_tx_buffer *buffer_info;
439 			tx_desc = IGB_TX_DESC(tx_ring, i);
440 			buffer_info = &tx_ring->tx_buffer_info[i];
441 			u0 = (struct my_u0 *)tx_desc;
442 			if (i == tx_ring->next_to_use &&
443 			    i == tx_ring->next_to_clean)
444 				next_desc = " NTC/U";
445 			else if (i == tx_ring->next_to_use)
446 				next_desc = " NTU";
447 			else if (i == tx_ring->next_to_clean)
448 				next_desc = " NTC";
449 			else
450 				next_desc = "";
451 
452 			pr_info("T [0x%03X]    %016llX %016llX %016llX %04X  %p %016llX %p%s\n",
453 				i, le64_to_cpu(u0->a),
454 				le64_to_cpu(u0->b),
455 				(u64)dma_unmap_addr(buffer_info, dma),
456 				dma_unmap_len(buffer_info, len),
457 				buffer_info->next_to_watch,
458 				(u64)buffer_info->time_stamp,
459 				buffer_info->skb, next_desc);
460 
461 			if (netif_msg_pktdata(adapter) && buffer_info->skb)
462 				print_hex_dump(KERN_INFO, "",
463 					DUMP_PREFIX_ADDRESS,
464 					16, 1, buffer_info->skb->data,
465 					dma_unmap_len(buffer_info, len),
466 					true);
467 		}
468 	}
469 
470 	/* Print RX Rings Summary */
471 rx_ring_summary:
472 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
473 	pr_info("Queue [NTU] [NTC]\n");
474 	for (n = 0; n < adapter->num_rx_queues; n++) {
475 		rx_ring = adapter->rx_ring[n];
476 		pr_info(" %5d %5X %5X\n",
477 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
478 	}
479 
480 	/* Print RX Rings */
481 	if (!netif_msg_rx_status(adapter))
482 		goto exit;
483 
484 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
485 
486 	/* Advanced Receive Descriptor (Read) Format
487 	 *    63                                           1        0
488 	 *    +-----------------------------------------------------+
489 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
490 	 *    +----------------------------------------------+------+
491 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
492 	 *    +-----------------------------------------------------+
493 	 *
494 	 *
495 	 * Advanced Receive Descriptor (Write-Back) Format
496 	 *
497 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
498 	 *   +------------------------------------------------------+
499 	 * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
500 	 *   | Checksum   Ident  |   |           |    | Type | Type |
501 	 *   +------------------------------------------------------+
502 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
503 	 *   +------------------------------------------------------+
504 	 *   63       48 47    32 31            20 19               0
505 	 */
506 
507 	for (n = 0; n < adapter->num_rx_queues; n++) {
508 		rx_ring = adapter->rx_ring[n];
509 		pr_info("------------------------------------\n");
510 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
511 		pr_info("------------------------------------\n");
512 		pr_info("R  [desc]      [ PktBuf     A0] [  HeadBuf   DD] [bi->dma       ] [bi->skb] <-- Adv Rx Read format\n");
513 		pr_info("RWB[desc]      [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
514 
515 		for (i = 0; i < rx_ring->count; i++) {
516 			const char *next_desc;
517 			struct igb_rx_buffer *buffer_info;
518 			buffer_info = &rx_ring->rx_buffer_info[i];
519 			rx_desc = IGB_RX_DESC(rx_ring, i);
520 			u0 = (struct my_u0 *)rx_desc;
521 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
522 
523 			if (i == rx_ring->next_to_use)
524 				next_desc = " NTU";
525 			else if (i == rx_ring->next_to_clean)
526 				next_desc = " NTC";
527 			else
528 				next_desc = "";
529 
530 			if (staterr & E1000_RXD_STAT_DD) {
531 				/* Descriptor Done */
532 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %s\n",
533 					"RWB", i,
534 					le64_to_cpu(u0->a),
535 					le64_to_cpu(u0->b),
536 					next_desc);
537 			} else {
538 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %s\n",
539 					"R  ", i,
540 					le64_to_cpu(u0->a),
541 					le64_to_cpu(u0->b),
542 					(u64)buffer_info->dma,
543 					next_desc);
544 
545 				if (netif_msg_pktdata(adapter) &&
546 				    buffer_info->dma && buffer_info->page) {
547 					print_hex_dump(KERN_INFO, "",
548 					  DUMP_PREFIX_ADDRESS,
549 					  16, 1,
550 					  page_address(buffer_info->page) +
551 						      buffer_info->page_offset,
552 					  IGB_RX_BUFSZ, true);
553 				}
554 			}
555 		}
556 	}
557 
558 exit:
559 	return;
560 }
561 
562 /**
563  *  igb_get_i2c_data - Reads the I2C SDA data bit
564  *  @hw: pointer to hardware structure
565  *  @i2cctl: Current value of I2CCTL register
566  *
567  *  Returns the I2C data bit value
568  **/
569 static int igb_get_i2c_data(void *data)
570 {
571 	struct igb_adapter *adapter = (struct igb_adapter *)data;
572 	struct e1000_hw *hw = &adapter->hw;
573 	s32 i2cctl = rd32(E1000_I2CPARAMS);
574 
575 	return !!(i2cctl & E1000_I2C_DATA_IN);
576 }
577 
578 /**
579  *  igb_set_i2c_data - Sets the I2C data bit
580  *  @data: pointer to hardware structure
581  *  @state: I2C data value (0 or 1) to set
582  *
583  *  Sets the I2C data bit
584  **/
585 static void igb_set_i2c_data(void *data, int state)
586 {
587 	struct igb_adapter *adapter = (struct igb_adapter *)data;
588 	struct e1000_hw *hw = &adapter->hw;
589 	s32 i2cctl = rd32(E1000_I2CPARAMS);
590 
591 	if (state)
592 		i2cctl |= E1000_I2C_DATA_OUT;
593 	else
594 		i2cctl &= ~E1000_I2C_DATA_OUT;
595 
596 	i2cctl &= ~E1000_I2C_DATA_OE_N;
597 	i2cctl |= E1000_I2C_CLK_OE_N;
598 	wr32(E1000_I2CPARAMS, i2cctl);
599 	wrfl();
600 
601 }
602 
603 /**
604  *  igb_set_i2c_clk - Sets the I2C SCL clock
605  *  @data: pointer to hardware structure
606  *  @state: state to set clock
607  *
608  *  Sets the I2C clock line to state
609  **/
610 static void igb_set_i2c_clk(void *data, int state)
611 {
612 	struct igb_adapter *adapter = (struct igb_adapter *)data;
613 	struct e1000_hw *hw = &adapter->hw;
614 	s32 i2cctl = rd32(E1000_I2CPARAMS);
615 
616 	if (state) {
617 		i2cctl |= E1000_I2C_CLK_OUT;
618 		i2cctl &= ~E1000_I2C_CLK_OE_N;
619 	} else {
620 		i2cctl &= ~E1000_I2C_CLK_OUT;
621 		i2cctl &= ~E1000_I2C_CLK_OE_N;
622 	}
623 	wr32(E1000_I2CPARAMS, i2cctl);
624 	wrfl();
625 }
626 
627 /**
628  *  igb_get_i2c_clk - Gets the I2C SCL clock state
629  *  @data: pointer to hardware structure
630  *
631  *  Gets the I2C clock state
632  **/
633 static int igb_get_i2c_clk(void *data)
634 {
635 	struct igb_adapter *adapter = (struct igb_adapter *)data;
636 	struct e1000_hw *hw = &adapter->hw;
637 	s32 i2cctl = rd32(E1000_I2CPARAMS);
638 
639 	return !!(i2cctl & E1000_I2C_CLK_IN);
640 }
641 
642 static const struct i2c_algo_bit_data igb_i2c_algo = {
643 	.setsda		= igb_set_i2c_data,
644 	.setscl		= igb_set_i2c_clk,
645 	.getsda		= igb_get_i2c_data,
646 	.getscl		= igb_get_i2c_clk,
647 	.udelay		= 5,
648 	.timeout	= 20,
649 };
650 
651 /**
652  *  igb_get_hw_dev - return device
653  *  @hw: pointer to hardware structure
654  *
655  *  used by hardware layer to print debugging information
656  **/
657 struct net_device *igb_get_hw_dev(struct e1000_hw *hw)
658 {
659 	struct igb_adapter *adapter = hw->back;
660 	return adapter->netdev;
661 }
662 
663 /**
664  *  igb_init_module - Driver Registration Routine
665  *
666  *  igb_init_module is the first routine called when the driver is
667  *  loaded. All it does is register with the PCI subsystem.
668  **/
669 static int __init igb_init_module(void)
670 {
671 	int ret;
672 
673 	pr_info("%s - version %s\n",
674 	       igb_driver_string, igb_driver_version);
675 	pr_info("%s\n", igb_copyright);
676 
677 #ifdef CONFIG_IGB_DCA
678 	dca_register_notify(&dca_notifier);
679 #endif
680 	ret = pci_register_driver(&igb_driver);
681 	return ret;
682 }
683 
684 module_init(igb_init_module);
685 
686 /**
687  *  igb_exit_module - Driver Exit Cleanup Routine
688  *
689  *  igb_exit_module is called just before the driver is removed
690  *  from memory.
691  **/
692 static void __exit igb_exit_module(void)
693 {
694 #ifdef CONFIG_IGB_DCA
695 	dca_unregister_notify(&dca_notifier);
696 #endif
697 	pci_unregister_driver(&igb_driver);
698 }
699 
700 module_exit(igb_exit_module);
701 
702 #define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
703 /**
704  *  igb_cache_ring_register - Descriptor ring to register mapping
705  *  @adapter: board private structure to initialize
706  *
707  *  Once we know the feature-set enabled for the device, we'll cache
708  *  the register offset the descriptor ring is assigned to.
709  **/
710 static void igb_cache_ring_register(struct igb_adapter *adapter)
711 {
712 	int i = 0, j = 0;
713 	u32 rbase_offset = adapter->vfs_allocated_count;
714 
715 	switch (adapter->hw.mac.type) {
716 	case e1000_82576:
717 		/* The queues are allocated for virtualization such that VF 0
718 		 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
719 		 * In order to avoid collision we start at the first free queue
720 		 * and continue consuming queues in the same sequence
721 		 */
722 		if (adapter->vfs_allocated_count) {
723 			for (; i < adapter->rss_queues; i++)
724 				adapter->rx_ring[i]->reg_idx = rbase_offset +
725 							       Q_IDX_82576(i);
726 		}
727 		/* Fall through */
728 	case e1000_82575:
729 	case e1000_82580:
730 	case e1000_i350:
731 	case e1000_i354:
732 	case e1000_i210:
733 	case e1000_i211:
734 		/* Fall through */
735 	default:
736 		for (; i < adapter->num_rx_queues; i++)
737 			adapter->rx_ring[i]->reg_idx = rbase_offset + i;
738 		for (; j < adapter->num_tx_queues; j++)
739 			adapter->tx_ring[j]->reg_idx = rbase_offset + j;
740 		break;
741 	}
742 }
743 
744 u32 igb_rd32(struct e1000_hw *hw, u32 reg)
745 {
746 	struct igb_adapter *igb = container_of(hw, struct igb_adapter, hw);
747 	u8 __iomem *hw_addr = ACCESS_ONCE(hw->hw_addr);
748 	u32 value = 0;
749 
750 	if (E1000_REMOVED(hw_addr))
751 		return ~value;
752 
753 	value = readl(&hw_addr[reg]);
754 
755 	/* reads should not return all F's */
756 	if (!(~value) && (!reg || !(~readl(hw_addr)))) {
757 		struct net_device *netdev = igb->netdev;
758 		hw->hw_addr = NULL;
759 		netif_device_detach(netdev);
760 		netdev_err(netdev, "PCIe link lost, device now detached\n");
761 	}
762 
763 	return value;
764 }
765 
766 /**
767  *  igb_write_ivar - configure ivar for given MSI-X vector
768  *  @hw: pointer to the HW structure
769  *  @msix_vector: vector number we are allocating to a given ring
770  *  @index: row index of IVAR register to write within IVAR table
771  *  @offset: column offset of in IVAR, should be multiple of 8
772  *
773  *  This function is intended to handle the writing of the IVAR register
774  *  for adapters 82576 and newer.  The IVAR table consists of 2 columns,
775  *  each containing an cause allocation for an Rx and Tx ring, and a
776  *  variable number of rows depending on the number of queues supported.
777  **/
778 static void igb_write_ivar(struct e1000_hw *hw, int msix_vector,
779 			   int index, int offset)
780 {
781 	u32 ivar = array_rd32(E1000_IVAR0, index);
782 
783 	/* clear any bits that are currently set */
784 	ivar &= ~((u32)0xFF << offset);
785 
786 	/* write vector and valid bit */
787 	ivar |= (msix_vector | E1000_IVAR_VALID) << offset;
788 
789 	array_wr32(E1000_IVAR0, index, ivar);
790 }
791 
792 #define IGB_N0_QUEUE -1
793 static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
794 {
795 	struct igb_adapter *adapter = q_vector->adapter;
796 	struct e1000_hw *hw = &adapter->hw;
797 	int rx_queue = IGB_N0_QUEUE;
798 	int tx_queue = IGB_N0_QUEUE;
799 	u32 msixbm = 0;
800 
801 	if (q_vector->rx.ring)
802 		rx_queue = q_vector->rx.ring->reg_idx;
803 	if (q_vector->tx.ring)
804 		tx_queue = q_vector->tx.ring->reg_idx;
805 
806 	switch (hw->mac.type) {
807 	case e1000_82575:
808 		/* The 82575 assigns vectors using a bitmask, which matches the
809 		 * bitmask for the EICR/EIMS/EIMC registers.  To assign one
810 		 * or more queues to a vector, we write the appropriate bits
811 		 * into the MSIXBM register for that vector.
812 		 */
813 		if (rx_queue > IGB_N0_QUEUE)
814 			msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
815 		if (tx_queue > IGB_N0_QUEUE)
816 			msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
817 		if (!(adapter->flags & IGB_FLAG_HAS_MSIX) && msix_vector == 0)
818 			msixbm |= E1000_EIMS_OTHER;
819 		array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
820 		q_vector->eims_value = msixbm;
821 		break;
822 	case e1000_82576:
823 		/* 82576 uses a table that essentially consists of 2 columns
824 		 * with 8 rows.  The ordering is column-major so we use the
825 		 * lower 3 bits as the row index, and the 4th bit as the
826 		 * column offset.
827 		 */
828 		if (rx_queue > IGB_N0_QUEUE)
829 			igb_write_ivar(hw, msix_vector,
830 				       rx_queue & 0x7,
831 				       (rx_queue & 0x8) << 1);
832 		if (tx_queue > IGB_N0_QUEUE)
833 			igb_write_ivar(hw, msix_vector,
834 				       tx_queue & 0x7,
835 				       ((tx_queue & 0x8) << 1) + 8);
836 		q_vector->eims_value = 1 << msix_vector;
837 		break;
838 	case e1000_82580:
839 	case e1000_i350:
840 	case e1000_i354:
841 	case e1000_i210:
842 	case e1000_i211:
843 		/* On 82580 and newer adapters the scheme is similar to 82576
844 		 * however instead of ordering column-major we have things
845 		 * ordered row-major.  So we traverse the table by using
846 		 * bit 0 as the column offset, and the remaining bits as the
847 		 * row index.
848 		 */
849 		if (rx_queue > IGB_N0_QUEUE)
850 			igb_write_ivar(hw, msix_vector,
851 				       rx_queue >> 1,
852 				       (rx_queue & 0x1) << 4);
853 		if (tx_queue > IGB_N0_QUEUE)
854 			igb_write_ivar(hw, msix_vector,
855 				       tx_queue >> 1,
856 				       ((tx_queue & 0x1) << 4) + 8);
857 		q_vector->eims_value = 1 << msix_vector;
858 		break;
859 	default:
860 		BUG();
861 		break;
862 	}
863 
864 	/* add q_vector eims value to global eims_enable_mask */
865 	adapter->eims_enable_mask |= q_vector->eims_value;
866 
867 	/* configure q_vector to set itr on first interrupt */
868 	q_vector->set_itr = 1;
869 }
870 
871 /**
872  *  igb_configure_msix - Configure MSI-X hardware
873  *  @adapter: board private structure to initialize
874  *
875  *  igb_configure_msix sets up the hardware to properly
876  *  generate MSI-X interrupts.
877  **/
878 static void igb_configure_msix(struct igb_adapter *adapter)
879 {
880 	u32 tmp;
881 	int i, vector = 0;
882 	struct e1000_hw *hw = &adapter->hw;
883 
884 	adapter->eims_enable_mask = 0;
885 
886 	/* set vector for other causes, i.e. link changes */
887 	switch (hw->mac.type) {
888 	case e1000_82575:
889 		tmp = rd32(E1000_CTRL_EXT);
890 		/* enable MSI-X PBA support*/
891 		tmp |= E1000_CTRL_EXT_PBA_CLR;
892 
893 		/* Auto-Mask interrupts upon ICR read. */
894 		tmp |= E1000_CTRL_EXT_EIAME;
895 		tmp |= E1000_CTRL_EXT_IRCA;
896 
897 		wr32(E1000_CTRL_EXT, tmp);
898 
899 		/* enable msix_other interrupt */
900 		array_wr32(E1000_MSIXBM(0), vector++, E1000_EIMS_OTHER);
901 		adapter->eims_other = E1000_EIMS_OTHER;
902 
903 		break;
904 
905 	case e1000_82576:
906 	case e1000_82580:
907 	case e1000_i350:
908 	case e1000_i354:
909 	case e1000_i210:
910 	case e1000_i211:
911 		/* Turn on MSI-X capability first, or our settings
912 		 * won't stick.  And it will take days to debug.
913 		 */
914 		wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
915 		     E1000_GPIE_PBA | E1000_GPIE_EIAME |
916 		     E1000_GPIE_NSICR);
917 
918 		/* enable msix_other interrupt */
919 		adapter->eims_other = 1 << vector;
920 		tmp = (vector++ | E1000_IVAR_VALID) << 8;
921 
922 		wr32(E1000_IVAR_MISC, tmp);
923 		break;
924 	default:
925 		/* do nothing, since nothing else supports MSI-X */
926 		break;
927 	} /* switch (hw->mac.type) */
928 
929 	adapter->eims_enable_mask |= adapter->eims_other;
930 
931 	for (i = 0; i < adapter->num_q_vectors; i++)
932 		igb_assign_vector(adapter->q_vector[i], vector++);
933 
934 	wrfl();
935 }
936 
937 /**
938  *  igb_request_msix - Initialize MSI-X interrupts
939  *  @adapter: board private structure to initialize
940  *
941  *  igb_request_msix allocates MSI-X vectors and requests interrupts from the
942  *  kernel.
943  **/
944 static int igb_request_msix(struct igb_adapter *adapter)
945 {
946 	struct net_device *netdev = adapter->netdev;
947 	struct e1000_hw *hw = &adapter->hw;
948 	int i, err = 0, vector = 0, free_vector = 0;
949 
950 	err = request_irq(adapter->msix_entries[vector].vector,
951 			  igb_msix_other, 0, netdev->name, adapter);
952 	if (err)
953 		goto err_out;
954 
955 	for (i = 0; i < adapter->num_q_vectors; i++) {
956 		struct igb_q_vector *q_vector = adapter->q_vector[i];
957 
958 		vector++;
959 
960 		q_vector->itr_register = hw->hw_addr + E1000_EITR(vector);
961 
962 		if (q_vector->rx.ring && q_vector->tx.ring)
963 			sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
964 				q_vector->rx.ring->queue_index);
965 		else if (q_vector->tx.ring)
966 			sprintf(q_vector->name, "%s-tx-%u", netdev->name,
967 				q_vector->tx.ring->queue_index);
968 		else if (q_vector->rx.ring)
969 			sprintf(q_vector->name, "%s-rx-%u", netdev->name,
970 				q_vector->rx.ring->queue_index);
971 		else
972 			sprintf(q_vector->name, "%s-unused", netdev->name);
973 
974 		err = request_irq(adapter->msix_entries[vector].vector,
975 				  igb_msix_ring, 0, q_vector->name,
976 				  q_vector);
977 		if (err)
978 			goto err_free;
979 	}
980 
981 	igb_configure_msix(adapter);
982 	return 0;
983 
984 err_free:
985 	/* free already assigned IRQs */
986 	free_irq(adapter->msix_entries[free_vector++].vector, adapter);
987 
988 	vector--;
989 	for (i = 0; i < vector; i++) {
990 		free_irq(adapter->msix_entries[free_vector++].vector,
991 			 adapter->q_vector[i]);
992 	}
993 err_out:
994 	return err;
995 }
996 
997 /**
998  *  igb_free_q_vector - Free memory allocated for specific interrupt vector
999  *  @adapter: board private structure to initialize
1000  *  @v_idx: Index of vector to be freed
1001  *
1002  *  This function frees the memory allocated to the q_vector.
1003  **/
1004 static void igb_free_q_vector(struct igb_adapter *adapter, int v_idx)
1005 {
1006 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1007 
1008 	adapter->q_vector[v_idx] = NULL;
1009 
1010 	/* igb_get_stats64() might access the rings on this vector,
1011 	 * we must wait a grace period before freeing it.
1012 	 */
1013 	if (q_vector)
1014 		kfree_rcu(q_vector, rcu);
1015 }
1016 
1017 /**
1018  *  igb_reset_q_vector - Reset config for interrupt vector
1019  *  @adapter: board private structure to initialize
1020  *  @v_idx: Index of vector to be reset
1021  *
1022  *  If NAPI is enabled it will delete any references to the
1023  *  NAPI struct. This is preparation for igb_free_q_vector.
1024  **/
1025 static void igb_reset_q_vector(struct igb_adapter *adapter, int v_idx)
1026 {
1027 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1028 
1029 	/* Coming from igb_set_interrupt_capability, the vectors are not yet
1030 	 * allocated. So, q_vector is NULL so we should stop here.
1031 	 */
1032 	if (!q_vector)
1033 		return;
1034 
1035 	if (q_vector->tx.ring)
1036 		adapter->tx_ring[q_vector->tx.ring->queue_index] = NULL;
1037 
1038 	if (q_vector->rx.ring)
1039 		adapter->tx_ring[q_vector->rx.ring->queue_index] = NULL;
1040 
1041 	netif_napi_del(&q_vector->napi);
1042 
1043 }
1044 
1045 static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
1046 {
1047 	int v_idx = adapter->num_q_vectors;
1048 
1049 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
1050 		pci_disable_msix(adapter->pdev);
1051 	else if (adapter->flags & IGB_FLAG_HAS_MSI)
1052 		pci_disable_msi(adapter->pdev);
1053 
1054 	while (v_idx--)
1055 		igb_reset_q_vector(adapter, v_idx);
1056 }
1057 
1058 /**
1059  *  igb_free_q_vectors - Free memory allocated for interrupt vectors
1060  *  @adapter: board private structure to initialize
1061  *
1062  *  This function frees the memory allocated to the q_vectors.  In addition if
1063  *  NAPI is enabled it will delete any references to the NAPI struct prior
1064  *  to freeing the q_vector.
1065  **/
1066 static void igb_free_q_vectors(struct igb_adapter *adapter)
1067 {
1068 	int v_idx = adapter->num_q_vectors;
1069 
1070 	adapter->num_tx_queues = 0;
1071 	adapter->num_rx_queues = 0;
1072 	adapter->num_q_vectors = 0;
1073 
1074 	while (v_idx--) {
1075 		igb_reset_q_vector(adapter, v_idx);
1076 		igb_free_q_vector(adapter, v_idx);
1077 	}
1078 }
1079 
1080 /**
1081  *  igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1082  *  @adapter: board private structure to initialize
1083  *
1084  *  This function resets the device so that it has 0 Rx queues, Tx queues, and
1085  *  MSI-X interrupts allocated.
1086  */
1087 static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
1088 {
1089 	igb_free_q_vectors(adapter);
1090 	igb_reset_interrupt_capability(adapter);
1091 }
1092 
1093 /**
1094  *  igb_set_interrupt_capability - set MSI or MSI-X if supported
1095  *  @adapter: board private structure to initialize
1096  *  @msix: boolean value of MSIX capability
1097  *
1098  *  Attempt to configure interrupts using the best available
1099  *  capabilities of the hardware and kernel.
1100  **/
1101 static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
1102 {
1103 	int err;
1104 	int numvecs, i;
1105 
1106 	if (!msix)
1107 		goto msi_only;
1108 	adapter->flags |= IGB_FLAG_HAS_MSIX;
1109 
1110 	/* Number of supported queues. */
1111 	adapter->num_rx_queues = adapter->rss_queues;
1112 	if (adapter->vfs_allocated_count)
1113 		adapter->num_tx_queues = 1;
1114 	else
1115 		adapter->num_tx_queues = adapter->rss_queues;
1116 
1117 	/* start with one vector for every Rx queue */
1118 	numvecs = adapter->num_rx_queues;
1119 
1120 	/* if Tx handler is separate add 1 for every Tx queue */
1121 	if (!(adapter->flags & IGB_FLAG_QUEUE_PAIRS))
1122 		numvecs += adapter->num_tx_queues;
1123 
1124 	/* store the number of vectors reserved for queues */
1125 	adapter->num_q_vectors = numvecs;
1126 
1127 	/* add 1 vector for link status interrupts */
1128 	numvecs++;
1129 	for (i = 0; i < numvecs; i++)
1130 		adapter->msix_entries[i].entry = i;
1131 
1132 	err = pci_enable_msix_range(adapter->pdev,
1133 				    adapter->msix_entries,
1134 				    numvecs,
1135 				    numvecs);
1136 	if (err > 0)
1137 		return;
1138 
1139 	igb_reset_interrupt_capability(adapter);
1140 
1141 	/* If we can't do MSI-X, try MSI */
1142 msi_only:
1143 	adapter->flags &= ~IGB_FLAG_HAS_MSIX;
1144 #ifdef CONFIG_PCI_IOV
1145 	/* disable SR-IOV for non MSI-X configurations */
1146 	if (adapter->vf_data) {
1147 		struct e1000_hw *hw = &adapter->hw;
1148 		/* disable iov and allow time for transactions to clear */
1149 		pci_disable_sriov(adapter->pdev);
1150 		msleep(500);
1151 
1152 		kfree(adapter->vf_data);
1153 		adapter->vf_data = NULL;
1154 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1155 		wrfl();
1156 		msleep(100);
1157 		dev_info(&adapter->pdev->dev, "IOV Disabled\n");
1158 	}
1159 #endif
1160 	adapter->vfs_allocated_count = 0;
1161 	adapter->rss_queues = 1;
1162 	adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
1163 	adapter->num_rx_queues = 1;
1164 	adapter->num_tx_queues = 1;
1165 	adapter->num_q_vectors = 1;
1166 	if (!pci_enable_msi(adapter->pdev))
1167 		adapter->flags |= IGB_FLAG_HAS_MSI;
1168 }
1169 
1170 static void igb_add_ring(struct igb_ring *ring,
1171 			 struct igb_ring_container *head)
1172 {
1173 	head->ring = ring;
1174 	head->count++;
1175 }
1176 
1177 /**
1178  *  igb_alloc_q_vector - Allocate memory for a single interrupt vector
1179  *  @adapter: board private structure to initialize
1180  *  @v_count: q_vectors allocated on adapter, used for ring interleaving
1181  *  @v_idx: index of vector in adapter struct
1182  *  @txr_count: total number of Tx rings to allocate
1183  *  @txr_idx: index of first Tx ring to allocate
1184  *  @rxr_count: total number of Rx rings to allocate
1185  *  @rxr_idx: index of first Rx ring to allocate
1186  *
1187  *  We allocate one q_vector.  If allocation fails we return -ENOMEM.
1188  **/
1189 static int igb_alloc_q_vector(struct igb_adapter *adapter,
1190 			      int v_count, int v_idx,
1191 			      int txr_count, int txr_idx,
1192 			      int rxr_count, int rxr_idx)
1193 {
1194 	struct igb_q_vector *q_vector;
1195 	struct igb_ring *ring;
1196 	int ring_count, size;
1197 
1198 	/* igb only supports 1 Tx and/or 1 Rx queue per vector */
1199 	if (txr_count > 1 || rxr_count > 1)
1200 		return -ENOMEM;
1201 
1202 	ring_count = txr_count + rxr_count;
1203 	size = sizeof(struct igb_q_vector) +
1204 	       (sizeof(struct igb_ring) * ring_count);
1205 
1206 	/* allocate q_vector and rings */
1207 	q_vector = adapter->q_vector[v_idx];
1208 	if (!q_vector)
1209 		q_vector = kzalloc(size, GFP_KERNEL);
1210 	if (!q_vector)
1211 		return -ENOMEM;
1212 
1213 	/* initialize NAPI */
1214 	netif_napi_add(adapter->netdev, &q_vector->napi,
1215 		       igb_poll, 64);
1216 
1217 	/* tie q_vector and adapter together */
1218 	adapter->q_vector[v_idx] = q_vector;
1219 	q_vector->adapter = adapter;
1220 
1221 	/* initialize work limits */
1222 	q_vector->tx.work_limit = adapter->tx_work_limit;
1223 
1224 	/* initialize ITR configuration */
1225 	q_vector->itr_register = adapter->hw.hw_addr + E1000_EITR(0);
1226 	q_vector->itr_val = IGB_START_ITR;
1227 
1228 	/* initialize pointer to rings */
1229 	ring = q_vector->ring;
1230 
1231 	/* intialize ITR */
1232 	if (rxr_count) {
1233 		/* rx or rx/tx vector */
1234 		if (!adapter->rx_itr_setting || adapter->rx_itr_setting > 3)
1235 			q_vector->itr_val = adapter->rx_itr_setting;
1236 	} else {
1237 		/* tx only vector */
1238 		if (!adapter->tx_itr_setting || adapter->tx_itr_setting > 3)
1239 			q_vector->itr_val = adapter->tx_itr_setting;
1240 	}
1241 
1242 	if (txr_count) {
1243 		/* assign generic ring traits */
1244 		ring->dev = &adapter->pdev->dev;
1245 		ring->netdev = adapter->netdev;
1246 
1247 		/* configure backlink on ring */
1248 		ring->q_vector = q_vector;
1249 
1250 		/* update q_vector Tx values */
1251 		igb_add_ring(ring, &q_vector->tx);
1252 
1253 		/* For 82575, context index must be unique per ring. */
1254 		if (adapter->hw.mac.type == e1000_82575)
1255 			set_bit(IGB_RING_FLAG_TX_CTX_IDX, &ring->flags);
1256 
1257 		/* apply Tx specific ring traits */
1258 		ring->count = adapter->tx_ring_count;
1259 		ring->queue_index = txr_idx;
1260 
1261 		u64_stats_init(&ring->tx_syncp);
1262 		u64_stats_init(&ring->tx_syncp2);
1263 
1264 		/* assign ring to adapter */
1265 		adapter->tx_ring[txr_idx] = ring;
1266 
1267 		/* push pointer to next ring */
1268 		ring++;
1269 	}
1270 
1271 	if (rxr_count) {
1272 		/* assign generic ring traits */
1273 		ring->dev = &adapter->pdev->dev;
1274 		ring->netdev = adapter->netdev;
1275 
1276 		/* configure backlink on ring */
1277 		ring->q_vector = q_vector;
1278 
1279 		/* update q_vector Rx values */
1280 		igb_add_ring(ring, &q_vector->rx);
1281 
1282 		/* set flag indicating ring supports SCTP checksum offload */
1283 		if (adapter->hw.mac.type >= e1000_82576)
1284 			set_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags);
1285 
1286 		/* On i350, i354, i210, and i211, loopback VLAN packets
1287 		 * have the tag byte-swapped.
1288 		 */
1289 		if (adapter->hw.mac.type >= e1000_i350)
1290 			set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &ring->flags);
1291 
1292 		/* apply Rx specific ring traits */
1293 		ring->count = adapter->rx_ring_count;
1294 		ring->queue_index = rxr_idx;
1295 
1296 		u64_stats_init(&ring->rx_syncp);
1297 
1298 		/* assign ring to adapter */
1299 		adapter->rx_ring[rxr_idx] = ring;
1300 	}
1301 
1302 	return 0;
1303 }
1304 
1305 
1306 /**
1307  *  igb_alloc_q_vectors - Allocate memory for interrupt vectors
1308  *  @adapter: board private structure to initialize
1309  *
1310  *  We allocate one q_vector per queue interrupt.  If allocation fails we
1311  *  return -ENOMEM.
1312  **/
1313 static int igb_alloc_q_vectors(struct igb_adapter *adapter)
1314 {
1315 	int q_vectors = adapter->num_q_vectors;
1316 	int rxr_remaining = adapter->num_rx_queues;
1317 	int txr_remaining = adapter->num_tx_queues;
1318 	int rxr_idx = 0, txr_idx = 0, v_idx = 0;
1319 	int err;
1320 
1321 	if (q_vectors >= (rxr_remaining + txr_remaining)) {
1322 		for (; rxr_remaining; v_idx++) {
1323 			err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1324 						 0, 0, 1, rxr_idx);
1325 
1326 			if (err)
1327 				goto err_out;
1328 
1329 			/* update counts and index */
1330 			rxr_remaining--;
1331 			rxr_idx++;
1332 		}
1333 	}
1334 
1335 	for (; v_idx < q_vectors; v_idx++) {
1336 		int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
1337 		int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);
1338 
1339 		err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1340 					 tqpv, txr_idx, rqpv, rxr_idx);
1341 
1342 		if (err)
1343 			goto err_out;
1344 
1345 		/* update counts and index */
1346 		rxr_remaining -= rqpv;
1347 		txr_remaining -= tqpv;
1348 		rxr_idx++;
1349 		txr_idx++;
1350 	}
1351 
1352 	return 0;
1353 
1354 err_out:
1355 	adapter->num_tx_queues = 0;
1356 	adapter->num_rx_queues = 0;
1357 	adapter->num_q_vectors = 0;
1358 
1359 	while (v_idx--)
1360 		igb_free_q_vector(adapter, v_idx);
1361 
1362 	return -ENOMEM;
1363 }
1364 
1365 /**
1366  *  igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1367  *  @adapter: board private structure to initialize
1368  *  @msix: boolean value of MSIX capability
1369  *
1370  *  This function initializes the interrupts and allocates all of the queues.
1371  **/
1372 static int igb_init_interrupt_scheme(struct igb_adapter *adapter, bool msix)
1373 {
1374 	struct pci_dev *pdev = adapter->pdev;
1375 	int err;
1376 
1377 	igb_set_interrupt_capability(adapter, msix);
1378 
1379 	err = igb_alloc_q_vectors(adapter);
1380 	if (err) {
1381 		dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
1382 		goto err_alloc_q_vectors;
1383 	}
1384 
1385 	igb_cache_ring_register(adapter);
1386 
1387 	return 0;
1388 
1389 err_alloc_q_vectors:
1390 	igb_reset_interrupt_capability(adapter);
1391 	return err;
1392 }
1393 
1394 /**
1395  *  igb_request_irq - initialize interrupts
1396  *  @adapter: board private structure to initialize
1397  *
1398  *  Attempts to configure interrupts using the best available
1399  *  capabilities of the hardware and kernel.
1400  **/
1401 static int igb_request_irq(struct igb_adapter *adapter)
1402 {
1403 	struct net_device *netdev = adapter->netdev;
1404 	struct pci_dev *pdev = adapter->pdev;
1405 	int err = 0;
1406 
1407 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1408 		err = igb_request_msix(adapter);
1409 		if (!err)
1410 			goto request_done;
1411 		/* fall back to MSI */
1412 		igb_free_all_tx_resources(adapter);
1413 		igb_free_all_rx_resources(adapter);
1414 
1415 		igb_clear_interrupt_scheme(adapter);
1416 		err = igb_init_interrupt_scheme(adapter, false);
1417 		if (err)
1418 			goto request_done;
1419 
1420 		igb_setup_all_tx_resources(adapter);
1421 		igb_setup_all_rx_resources(adapter);
1422 		igb_configure(adapter);
1423 	}
1424 
1425 	igb_assign_vector(adapter->q_vector[0], 0);
1426 
1427 	if (adapter->flags & IGB_FLAG_HAS_MSI) {
1428 		err = request_irq(pdev->irq, igb_intr_msi, 0,
1429 				  netdev->name, adapter);
1430 		if (!err)
1431 			goto request_done;
1432 
1433 		/* fall back to legacy interrupts */
1434 		igb_reset_interrupt_capability(adapter);
1435 		adapter->flags &= ~IGB_FLAG_HAS_MSI;
1436 	}
1437 
1438 	err = request_irq(pdev->irq, igb_intr, IRQF_SHARED,
1439 			  netdev->name, adapter);
1440 
1441 	if (err)
1442 		dev_err(&pdev->dev, "Error %d getting interrupt\n",
1443 			err);
1444 
1445 request_done:
1446 	return err;
1447 }
1448 
1449 static void igb_free_irq(struct igb_adapter *adapter)
1450 {
1451 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1452 		int vector = 0, i;
1453 
1454 		free_irq(adapter->msix_entries[vector++].vector, adapter);
1455 
1456 		for (i = 0; i < adapter->num_q_vectors; i++)
1457 			free_irq(adapter->msix_entries[vector++].vector,
1458 				 adapter->q_vector[i]);
1459 	} else {
1460 		free_irq(adapter->pdev->irq, adapter);
1461 	}
1462 }
1463 
1464 /**
1465  *  igb_irq_disable - Mask off interrupt generation on the NIC
1466  *  @adapter: board private structure
1467  **/
1468 static void igb_irq_disable(struct igb_adapter *adapter)
1469 {
1470 	struct e1000_hw *hw = &adapter->hw;
1471 
1472 	/* we need to be careful when disabling interrupts.  The VFs are also
1473 	 * mapped into these registers and so clearing the bits can cause
1474 	 * issues on the VF drivers so we only need to clear what we set
1475 	 */
1476 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1477 		u32 regval = rd32(E1000_EIAM);
1478 
1479 		wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
1480 		wr32(E1000_EIMC, adapter->eims_enable_mask);
1481 		regval = rd32(E1000_EIAC);
1482 		wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
1483 	}
1484 
1485 	wr32(E1000_IAM, 0);
1486 	wr32(E1000_IMC, ~0);
1487 	wrfl();
1488 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1489 		int i;
1490 
1491 		for (i = 0; i < adapter->num_q_vectors; i++)
1492 			synchronize_irq(adapter->msix_entries[i].vector);
1493 	} else {
1494 		synchronize_irq(adapter->pdev->irq);
1495 	}
1496 }
1497 
1498 /**
1499  *  igb_irq_enable - Enable default interrupt generation settings
1500  *  @adapter: board private structure
1501  **/
1502 static void igb_irq_enable(struct igb_adapter *adapter)
1503 {
1504 	struct e1000_hw *hw = &adapter->hw;
1505 
1506 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1507 		u32 ims = E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_DRSTA;
1508 		u32 regval = rd32(E1000_EIAC);
1509 
1510 		wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
1511 		regval = rd32(E1000_EIAM);
1512 		wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
1513 		wr32(E1000_EIMS, adapter->eims_enable_mask);
1514 		if (adapter->vfs_allocated_count) {
1515 			wr32(E1000_MBVFIMR, 0xFF);
1516 			ims |= E1000_IMS_VMMB;
1517 		}
1518 		wr32(E1000_IMS, ims);
1519 	} else {
1520 		wr32(E1000_IMS, IMS_ENABLE_MASK |
1521 				E1000_IMS_DRSTA);
1522 		wr32(E1000_IAM, IMS_ENABLE_MASK |
1523 				E1000_IMS_DRSTA);
1524 	}
1525 }
1526 
1527 static void igb_update_mng_vlan(struct igb_adapter *adapter)
1528 {
1529 	struct e1000_hw *hw = &adapter->hw;
1530 	u16 vid = adapter->hw.mng_cookie.vlan_id;
1531 	u16 old_vid = adapter->mng_vlan_id;
1532 
1533 	if (hw->mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1534 		/* add VID to filter table */
1535 		igb_vfta_set(hw, vid, true);
1536 		adapter->mng_vlan_id = vid;
1537 	} else {
1538 		adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1539 	}
1540 
1541 	if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1542 	    (vid != old_vid) &&
1543 	    !test_bit(old_vid, adapter->active_vlans)) {
1544 		/* remove VID from filter table */
1545 		igb_vfta_set(hw, old_vid, false);
1546 	}
1547 }
1548 
1549 /**
1550  *  igb_release_hw_control - release control of the h/w to f/w
1551  *  @adapter: address of board private structure
1552  *
1553  *  igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1554  *  For ASF and Pass Through versions of f/w this means that the
1555  *  driver is no longer loaded.
1556  **/
1557 static void igb_release_hw_control(struct igb_adapter *adapter)
1558 {
1559 	struct e1000_hw *hw = &adapter->hw;
1560 	u32 ctrl_ext;
1561 
1562 	/* Let firmware take over control of h/w */
1563 	ctrl_ext = rd32(E1000_CTRL_EXT);
1564 	wr32(E1000_CTRL_EXT,
1565 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1566 }
1567 
1568 /**
1569  *  igb_get_hw_control - get control of the h/w from f/w
1570  *  @adapter: address of board private structure
1571  *
1572  *  igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1573  *  For ASF and Pass Through versions of f/w this means that
1574  *  the driver is loaded.
1575  **/
1576 static void igb_get_hw_control(struct igb_adapter *adapter)
1577 {
1578 	struct e1000_hw *hw = &adapter->hw;
1579 	u32 ctrl_ext;
1580 
1581 	/* Let firmware know the driver has taken over */
1582 	ctrl_ext = rd32(E1000_CTRL_EXT);
1583 	wr32(E1000_CTRL_EXT,
1584 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1585 }
1586 
1587 /**
1588  *  igb_configure - configure the hardware for RX and TX
1589  *  @adapter: private board structure
1590  **/
1591 static void igb_configure(struct igb_adapter *adapter)
1592 {
1593 	struct net_device *netdev = adapter->netdev;
1594 	int i;
1595 
1596 	igb_get_hw_control(adapter);
1597 	igb_set_rx_mode(netdev);
1598 
1599 	igb_restore_vlan(adapter);
1600 
1601 	igb_setup_tctl(adapter);
1602 	igb_setup_mrqc(adapter);
1603 	igb_setup_rctl(adapter);
1604 
1605 	igb_configure_tx(adapter);
1606 	igb_configure_rx(adapter);
1607 
1608 	igb_rx_fifo_flush_82575(&adapter->hw);
1609 
1610 	/* call igb_desc_unused which always leaves
1611 	 * at least 1 descriptor unused to make sure
1612 	 * next_to_use != next_to_clean
1613 	 */
1614 	for (i = 0; i < adapter->num_rx_queues; i++) {
1615 		struct igb_ring *ring = adapter->rx_ring[i];
1616 		igb_alloc_rx_buffers(ring, igb_desc_unused(ring));
1617 	}
1618 }
1619 
1620 /**
1621  *  igb_power_up_link - Power up the phy/serdes link
1622  *  @adapter: address of board private structure
1623  **/
1624 void igb_power_up_link(struct igb_adapter *adapter)
1625 {
1626 	igb_reset_phy(&adapter->hw);
1627 
1628 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
1629 		igb_power_up_phy_copper(&adapter->hw);
1630 	else
1631 		igb_power_up_serdes_link_82575(&adapter->hw);
1632 
1633 	igb_setup_link(&adapter->hw);
1634 }
1635 
1636 /**
1637  *  igb_power_down_link - Power down the phy/serdes link
1638  *  @adapter: address of board private structure
1639  */
1640 static void igb_power_down_link(struct igb_adapter *adapter)
1641 {
1642 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
1643 		igb_power_down_phy_copper_82575(&adapter->hw);
1644 	else
1645 		igb_shutdown_serdes_link_82575(&adapter->hw);
1646 }
1647 
1648 /**
1649  * Detect and switch function for Media Auto Sense
1650  * @adapter: address of the board private structure
1651  **/
1652 static void igb_check_swap_media(struct igb_adapter *adapter)
1653 {
1654 	struct e1000_hw *hw = &adapter->hw;
1655 	u32 ctrl_ext, connsw;
1656 	bool swap_now = false;
1657 
1658 	ctrl_ext = rd32(E1000_CTRL_EXT);
1659 	connsw = rd32(E1000_CONNSW);
1660 
1661 	/* need to live swap if current media is copper and we have fiber/serdes
1662 	 * to go to.
1663 	 */
1664 
1665 	if ((hw->phy.media_type == e1000_media_type_copper) &&
1666 	    (!(connsw & E1000_CONNSW_AUTOSENSE_EN))) {
1667 		swap_now = true;
1668 	} else if (!(connsw & E1000_CONNSW_SERDESD)) {
1669 		/* copper signal takes time to appear */
1670 		if (adapter->copper_tries < 4) {
1671 			adapter->copper_tries++;
1672 			connsw |= E1000_CONNSW_AUTOSENSE_CONF;
1673 			wr32(E1000_CONNSW, connsw);
1674 			return;
1675 		} else {
1676 			adapter->copper_tries = 0;
1677 			if ((connsw & E1000_CONNSW_PHYSD) &&
1678 			    (!(connsw & E1000_CONNSW_PHY_PDN))) {
1679 				swap_now = true;
1680 				connsw &= ~E1000_CONNSW_AUTOSENSE_CONF;
1681 				wr32(E1000_CONNSW, connsw);
1682 			}
1683 		}
1684 	}
1685 
1686 	if (!swap_now)
1687 		return;
1688 
1689 	switch (hw->phy.media_type) {
1690 	case e1000_media_type_copper:
1691 		netdev_info(adapter->netdev,
1692 			"MAS: changing media to fiber/serdes\n");
1693 		ctrl_ext |=
1694 			E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
1695 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
1696 		adapter->copper_tries = 0;
1697 		break;
1698 	case e1000_media_type_internal_serdes:
1699 	case e1000_media_type_fiber:
1700 		netdev_info(adapter->netdev,
1701 			"MAS: changing media to copper\n");
1702 		ctrl_ext &=
1703 			~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
1704 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
1705 		break;
1706 	default:
1707 		/* shouldn't get here during regular operation */
1708 		netdev_err(adapter->netdev,
1709 			"AMS: Invalid media type found, returning\n");
1710 		break;
1711 	}
1712 	wr32(E1000_CTRL_EXT, ctrl_ext);
1713 }
1714 
1715 /**
1716  *  igb_up - Open the interface and prepare it to handle traffic
1717  *  @adapter: board private structure
1718  **/
1719 int igb_up(struct igb_adapter *adapter)
1720 {
1721 	struct e1000_hw *hw = &adapter->hw;
1722 	int i;
1723 
1724 	/* hardware has been reset, we need to reload some things */
1725 	igb_configure(adapter);
1726 
1727 	clear_bit(__IGB_DOWN, &adapter->state);
1728 
1729 	for (i = 0; i < adapter->num_q_vectors; i++)
1730 		napi_enable(&(adapter->q_vector[i]->napi));
1731 
1732 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
1733 		igb_configure_msix(adapter);
1734 	else
1735 		igb_assign_vector(adapter->q_vector[0], 0);
1736 
1737 	/* Clear any pending interrupts. */
1738 	rd32(E1000_ICR);
1739 	igb_irq_enable(adapter);
1740 
1741 	/* notify VFs that reset has been completed */
1742 	if (adapter->vfs_allocated_count) {
1743 		u32 reg_data = rd32(E1000_CTRL_EXT);
1744 
1745 		reg_data |= E1000_CTRL_EXT_PFRSTD;
1746 		wr32(E1000_CTRL_EXT, reg_data);
1747 	}
1748 
1749 	netif_tx_start_all_queues(adapter->netdev);
1750 
1751 	/* start the watchdog. */
1752 	hw->mac.get_link_status = 1;
1753 	schedule_work(&adapter->watchdog_task);
1754 
1755 	if ((adapter->flags & IGB_FLAG_EEE) &&
1756 	    (!hw->dev_spec._82575.eee_disable))
1757 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
1758 
1759 	return 0;
1760 }
1761 
1762 void igb_down(struct igb_adapter *adapter)
1763 {
1764 	struct net_device *netdev = adapter->netdev;
1765 	struct e1000_hw *hw = &adapter->hw;
1766 	u32 tctl, rctl;
1767 	int i;
1768 
1769 	/* signal that we're down so the interrupt handler does not
1770 	 * reschedule our watchdog timer
1771 	 */
1772 	set_bit(__IGB_DOWN, &adapter->state);
1773 
1774 	/* disable receives in the hardware */
1775 	rctl = rd32(E1000_RCTL);
1776 	wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
1777 	/* flush and sleep below */
1778 
1779 	netif_carrier_off(netdev);
1780 	netif_tx_stop_all_queues(netdev);
1781 
1782 	/* disable transmits in the hardware */
1783 	tctl = rd32(E1000_TCTL);
1784 	tctl &= ~E1000_TCTL_EN;
1785 	wr32(E1000_TCTL, tctl);
1786 	/* flush both disables and wait for them to finish */
1787 	wrfl();
1788 	usleep_range(10000, 11000);
1789 
1790 	igb_irq_disable(adapter);
1791 
1792 	adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
1793 
1794 	for (i = 0; i < adapter->num_q_vectors; i++) {
1795 		if (adapter->q_vector[i]) {
1796 			napi_synchronize(&adapter->q_vector[i]->napi);
1797 			napi_disable(&adapter->q_vector[i]->napi);
1798 		}
1799 	}
1800 
1801 	del_timer_sync(&adapter->watchdog_timer);
1802 	del_timer_sync(&adapter->phy_info_timer);
1803 
1804 	/* record the stats before reset*/
1805 	spin_lock(&adapter->stats64_lock);
1806 	igb_update_stats(adapter, &adapter->stats64);
1807 	spin_unlock(&adapter->stats64_lock);
1808 
1809 	adapter->link_speed = 0;
1810 	adapter->link_duplex = 0;
1811 
1812 	if (!pci_channel_offline(adapter->pdev))
1813 		igb_reset(adapter);
1814 	igb_clean_all_tx_rings(adapter);
1815 	igb_clean_all_rx_rings(adapter);
1816 #ifdef CONFIG_IGB_DCA
1817 
1818 	/* since we reset the hardware DCA settings were cleared */
1819 	igb_setup_dca(adapter);
1820 #endif
1821 }
1822 
1823 void igb_reinit_locked(struct igb_adapter *adapter)
1824 {
1825 	WARN_ON(in_interrupt());
1826 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
1827 		usleep_range(1000, 2000);
1828 	igb_down(adapter);
1829 	igb_up(adapter);
1830 	clear_bit(__IGB_RESETTING, &adapter->state);
1831 }
1832 
1833 /** igb_enable_mas - Media Autosense re-enable after swap
1834  *
1835  * @adapter: adapter struct
1836  **/
1837 static s32 igb_enable_mas(struct igb_adapter *adapter)
1838 {
1839 	struct e1000_hw *hw = &adapter->hw;
1840 	u32 connsw;
1841 	s32 ret_val = 0;
1842 
1843 	connsw = rd32(E1000_CONNSW);
1844 	if (!(hw->phy.media_type == e1000_media_type_copper))
1845 		return ret_val;
1846 
1847 	/* configure for SerDes media detect */
1848 	if (!(connsw & E1000_CONNSW_SERDESD)) {
1849 		connsw |= E1000_CONNSW_ENRGSRC;
1850 		connsw |= E1000_CONNSW_AUTOSENSE_EN;
1851 		wr32(E1000_CONNSW, connsw);
1852 		wrfl();
1853 	} else if (connsw & E1000_CONNSW_SERDESD) {
1854 		/* already SerDes, no need to enable anything */
1855 		return ret_val;
1856 	} else {
1857 		netdev_info(adapter->netdev,
1858 			"MAS: Unable to configure feature, disabling..\n");
1859 		adapter->flags &= ~IGB_FLAG_MAS_ENABLE;
1860 	}
1861 	return ret_val;
1862 }
1863 
1864 void igb_reset(struct igb_adapter *adapter)
1865 {
1866 	struct pci_dev *pdev = adapter->pdev;
1867 	struct e1000_hw *hw = &adapter->hw;
1868 	struct e1000_mac_info *mac = &hw->mac;
1869 	struct e1000_fc_info *fc = &hw->fc;
1870 	u32 pba = 0, tx_space, min_tx_space, min_rx_space, hwm;
1871 
1872 	/* Repartition Pba for greater than 9k mtu
1873 	 * To take effect CTRL.RST is required.
1874 	 */
1875 	switch (mac->type) {
1876 	case e1000_i350:
1877 	case e1000_i354:
1878 	case e1000_82580:
1879 		pba = rd32(E1000_RXPBS);
1880 		pba = igb_rxpbs_adjust_82580(pba);
1881 		break;
1882 	case e1000_82576:
1883 		pba = rd32(E1000_RXPBS);
1884 		pba &= E1000_RXPBS_SIZE_MASK_82576;
1885 		break;
1886 	case e1000_82575:
1887 	case e1000_i210:
1888 	case e1000_i211:
1889 	default:
1890 		pba = E1000_PBA_34K;
1891 		break;
1892 	}
1893 
1894 	if ((adapter->max_frame_size > ETH_FRAME_LEN + ETH_FCS_LEN) &&
1895 	    (mac->type < e1000_82576)) {
1896 		/* adjust PBA for jumbo frames */
1897 		wr32(E1000_PBA, pba);
1898 
1899 		/* To maintain wire speed transmits, the Tx FIFO should be
1900 		 * large enough to accommodate two full transmit packets,
1901 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
1902 		 * the Rx FIFO should be large enough to accommodate at least
1903 		 * one full receive packet and is similarly rounded up and
1904 		 * expressed in KB.
1905 		 */
1906 		pba = rd32(E1000_PBA);
1907 		/* upper 16 bits has Tx packet buffer allocation size in KB */
1908 		tx_space = pba >> 16;
1909 		/* lower 16 bits has Rx packet buffer allocation size in KB */
1910 		pba &= 0xffff;
1911 		/* the Tx fifo also stores 16 bytes of information about the Tx
1912 		 * but don't include ethernet FCS because hardware appends it
1913 		 */
1914 		min_tx_space = (adapter->max_frame_size +
1915 				sizeof(union e1000_adv_tx_desc) -
1916 				ETH_FCS_LEN) * 2;
1917 		min_tx_space = ALIGN(min_tx_space, 1024);
1918 		min_tx_space >>= 10;
1919 		/* software strips receive CRC, so leave room for it */
1920 		min_rx_space = adapter->max_frame_size;
1921 		min_rx_space = ALIGN(min_rx_space, 1024);
1922 		min_rx_space >>= 10;
1923 
1924 		/* If current Tx allocation is less than the min Tx FIFO size,
1925 		 * and the min Tx FIFO size is less than the current Rx FIFO
1926 		 * allocation, take space away from current Rx allocation
1927 		 */
1928 		if (tx_space < min_tx_space &&
1929 		    ((min_tx_space - tx_space) < pba)) {
1930 			pba = pba - (min_tx_space - tx_space);
1931 
1932 			/* if short on Rx space, Rx wins and must trump Tx
1933 			 * adjustment
1934 			 */
1935 			if (pba < min_rx_space)
1936 				pba = min_rx_space;
1937 		}
1938 		wr32(E1000_PBA, pba);
1939 	}
1940 
1941 	/* flow control settings */
1942 	/* The high water mark must be low enough to fit one full frame
1943 	 * (or the size used for early receive) above it in the Rx FIFO.
1944 	 * Set it to the lower of:
1945 	 * - 90% of the Rx FIFO size, or
1946 	 * - the full Rx FIFO size minus one full frame
1947 	 */
1948 	hwm = min(((pba << 10) * 9 / 10),
1949 			((pba << 10) - 2 * adapter->max_frame_size));
1950 
1951 	fc->high_water = hwm & 0xFFFFFFF0;	/* 16-byte granularity */
1952 	fc->low_water = fc->high_water - 16;
1953 	fc->pause_time = 0xFFFF;
1954 	fc->send_xon = 1;
1955 	fc->current_mode = fc->requested_mode;
1956 
1957 	/* disable receive for all VFs and wait one second */
1958 	if (adapter->vfs_allocated_count) {
1959 		int i;
1960 
1961 		for (i = 0 ; i < adapter->vfs_allocated_count; i++)
1962 			adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
1963 
1964 		/* ping all the active vfs to let them know we are going down */
1965 		igb_ping_all_vfs(adapter);
1966 
1967 		/* disable transmits and receives */
1968 		wr32(E1000_VFRE, 0);
1969 		wr32(E1000_VFTE, 0);
1970 	}
1971 
1972 	/* Allow time for pending master requests to run */
1973 	hw->mac.ops.reset_hw(hw);
1974 	wr32(E1000_WUC, 0);
1975 
1976 	if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
1977 		/* need to resetup here after media swap */
1978 		adapter->ei.get_invariants(hw);
1979 		adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
1980 	}
1981 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
1982 		if (igb_enable_mas(adapter))
1983 			dev_err(&pdev->dev,
1984 				"Error enabling Media Auto Sense\n");
1985 	}
1986 	if (hw->mac.ops.init_hw(hw))
1987 		dev_err(&pdev->dev, "Hardware Error\n");
1988 
1989 	/* Flow control settings reset on hardware reset, so guarantee flow
1990 	 * control is off when forcing speed.
1991 	 */
1992 	if (!hw->mac.autoneg)
1993 		igb_force_mac_fc(hw);
1994 
1995 	igb_init_dmac(adapter, pba);
1996 #ifdef CONFIG_IGB_HWMON
1997 	/* Re-initialize the thermal sensor on i350 devices. */
1998 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
1999 		if (mac->type == e1000_i350 && hw->bus.func == 0) {
2000 			/* If present, re-initialize the external thermal sensor
2001 			 * interface.
2002 			 */
2003 			if (adapter->ets)
2004 				mac->ops.init_thermal_sensor_thresh(hw);
2005 		}
2006 	}
2007 #endif
2008 	/* Re-establish EEE setting */
2009 	if (hw->phy.media_type == e1000_media_type_copper) {
2010 		switch (mac->type) {
2011 		case e1000_i350:
2012 		case e1000_i210:
2013 		case e1000_i211:
2014 			igb_set_eee_i350(hw, true, true);
2015 			break;
2016 		case e1000_i354:
2017 			igb_set_eee_i354(hw, true, true);
2018 			break;
2019 		default:
2020 			break;
2021 		}
2022 	}
2023 	if (!netif_running(adapter->netdev))
2024 		igb_power_down_link(adapter);
2025 
2026 	igb_update_mng_vlan(adapter);
2027 
2028 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2029 	wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2030 
2031 	/* Re-enable PTP, where applicable. */
2032 	igb_ptp_reset(adapter);
2033 
2034 	igb_get_phy_info(hw);
2035 }
2036 
2037 static netdev_features_t igb_fix_features(struct net_device *netdev,
2038 	netdev_features_t features)
2039 {
2040 	/* Since there is no support for separate Rx/Tx vlan accel
2041 	 * enable/disable make sure Tx flag is always in same state as Rx.
2042 	 */
2043 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2044 		features |= NETIF_F_HW_VLAN_CTAG_TX;
2045 	else
2046 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2047 
2048 	return features;
2049 }
2050 
2051 static int igb_set_features(struct net_device *netdev,
2052 	netdev_features_t features)
2053 {
2054 	netdev_features_t changed = netdev->features ^ features;
2055 	struct igb_adapter *adapter = netdev_priv(netdev);
2056 
2057 	if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2058 		igb_vlan_mode(netdev, features);
2059 
2060 	if (!(changed & NETIF_F_RXALL))
2061 		return 0;
2062 
2063 	netdev->features = features;
2064 
2065 	if (netif_running(netdev))
2066 		igb_reinit_locked(adapter);
2067 	else
2068 		igb_reset(adapter);
2069 
2070 	return 0;
2071 }
2072 
2073 static const struct net_device_ops igb_netdev_ops = {
2074 	.ndo_open		= igb_open,
2075 	.ndo_stop		= igb_close,
2076 	.ndo_start_xmit		= igb_xmit_frame,
2077 	.ndo_get_stats64	= igb_get_stats64,
2078 	.ndo_set_rx_mode	= igb_set_rx_mode,
2079 	.ndo_set_mac_address	= igb_set_mac,
2080 	.ndo_change_mtu		= igb_change_mtu,
2081 	.ndo_do_ioctl		= igb_ioctl,
2082 	.ndo_tx_timeout		= igb_tx_timeout,
2083 	.ndo_validate_addr	= eth_validate_addr,
2084 	.ndo_vlan_rx_add_vid	= igb_vlan_rx_add_vid,
2085 	.ndo_vlan_rx_kill_vid	= igb_vlan_rx_kill_vid,
2086 	.ndo_set_vf_mac		= igb_ndo_set_vf_mac,
2087 	.ndo_set_vf_vlan	= igb_ndo_set_vf_vlan,
2088 	.ndo_set_vf_rate	= igb_ndo_set_vf_bw,
2089 	.ndo_set_vf_spoofchk	= igb_ndo_set_vf_spoofchk,
2090 	.ndo_get_vf_config	= igb_ndo_get_vf_config,
2091 #ifdef CONFIG_NET_POLL_CONTROLLER
2092 	.ndo_poll_controller	= igb_netpoll,
2093 #endif
2094 	.ndo_fix_features	= igb_fix_features,
2095 	.ndo_set_features	= igb_set_features,
2096 	.ndo_features_check	= passthru_features_check,
2097 };
2098 
2099 /**
2100  * igb_set_fw_version - Configure version string for ethtool
2101  * @adapter: adapter struct
2102  **/
2103 void igb_set_fw_version(struct igb_adapter *adapter)
2104 {
2105 	struct e1000_hw *hw = &adapter->hw;
2106 	struct e1000_fw_version fw;
2107 
2108 	igb_get_fw_version(hw, &fw);
2109 
2110 	switch (hw->mac.type) {
2111 	case e1000_i210:
2112 	case e1000_i211:
2113 		if (!(igb_get_flash_presence_i210(hw))) {
2114 			snprintf(adapter->fw_version,
2115 				 sizeof(adapter->fw_version),
2116 				 "%2d.%2d-%d",
2117 				 fw.invm_major, fw.invm_minor,
2118 				 fw.invm_img_type);
2119 			break;
2120 		}
2121 		/* fall through */
2122 	default:
2123 		/* if option is rom valid, display its version too */
2124 		if (fw.or_valid) {
2125 			snprintf(adapter->fw_version,
2126 				 sizeof(adapter->fw_version),
2127 				 "%d.%d, 0x%08x, %d.%d.%d",
2128 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2129 				 fw.or_major, fw.or_build, fw.or_patch);
2130 		/* no option rom */
2131 		} else if (fw.etrack_id != 0X0000) {
2132 			snprintf(adapter->fw_version,
2133 			    sizeof(adapter->fw_version),
2134 			    "%d.%d, 0x%08x",
2135 			    fw.eep_major, fw.eep_minor, fw.etrack_id);
2136 		} else {
2137 		snprintf(adapter->fw_version,
2138 		    sizeof(adapter->fw_version),
2139 		    "%d.%d.%d",
2140 		    fw.eep_major, fw.eep_minor, fw.eep_build);
2141 		}
2142 		break;
2143 	}
2144 }
2145 
2146 /**
2147  * igb_init_mas - init Media Autosense feature if enabled in the NVM
2148  *
2149  * @adapter: adapter struct
2150  **/
2151 static void igb_init_mas(struct igb_adapter *adapter)
2152 {
2153 	struct e1000_hw *hw = &adapter->hw;
2154 	u16 eeprom_data;
2155 
2156 	hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
2157 	switch (hw->bus.func) {
2158 	case E1000_FUNC_0:
2159 		if (eeprom_data & IGB_MAS_ENABLE_0) {
2160 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2161 			netdev_info(adapter->netdev,
2162 				"MAS: Enabling Media Autosense for port %d\n",
2163 				hw->bus.func);
2164 		}
2165 		break;
2166 	case E1000_FUNC_1:
2167 		if (eeprom_data & IGB_MAS_ENABLE_1) {
2168 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2169 			netdev_info(adapter->netdev,
2170 				"MAS: Enabling Media Autosense for port %d\n",
2171 				hw->bus.func);
2172 		}
2173 		break;
2174 	case E1000_FUNC_2:
2175 		if (eeprom_data & IGB_MAS_ENABLE_2) {
2176 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2177 			netdev_info(adapter->netdev,
2178 				"MAS: Enabling Media Autosense for port %d\n",
2179 				hw->bus.func);
2180 		}
2181 		break;
2182 	case E1000_FUNC_3:
2183 		if (eeprom_data & IGB_MAS_ENABLE_3) {
2184 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2185 			netdev_info(adapter->netdev,
2186 				"MAS: Enabling Media Autosense for port %d\n",
2187 				hw->bus.func);
2188 		}
2189 		break;
2190 	default:
2191 		/* Shouldn't get here */
2192 		netdev_err(adapter->netdev,
2193 			"MAS: Invalid port configuration, returning\n");
2194 		break;
2195 	}
2196 }
2197 
2198 /**
2199  *  igb_init_i2c - Init I2C interface
2200  *  @adapter: pointer to adapter structure
2201  **/
2202 static s32 igb_init_i2c(struct igb_adapter *adapter)
2203 {
2204 	s32 status = 0;
2205 
2206 	/* I2C interface supported on i350 devices */
2207 	if (adapter->hw.mac.type != e1000_i350)
2208 		return 0;
2209 
2210 	/* Initialize the i2c bus which is controlled by the registers.
2211 	 * This bus will use the i2c_algo_bit structue that implements
2212 	 * the protocol through toggling of the 4 bits in the register.
2213 	 */
2214 	adapter->i2c_adap.owner = THIS_MODULE;
2215 	adapter->i2c_algo = igb_i2c_algo;
2216 	adapter->i2c_algo.data = adapter;
2217 	adapter->i2c_adap.algo_data = &adapter->i2c_algo;
2218 	adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
2219 	strlcpy(adapter->i2c_adap.name, "igb BB",
2220 		sizeof(adapter->i2c_adap.name));
2221 	status = i2c_bit_add_bus(&adapter->i2c_adap);
2222 	return status;
2223 }
2224 
2225 /**
2226  *  igb_probe - Device Initialization Routine
2227  *  @pdev: PCI device information struct
2228  *  @ent: entry in igb_pci_tbl
2229  *
2230  *  Returns 0 on success, negative on failure
2231  *
2232  *  igb_probe initializes an adapter identified by a pci_dev structure.
2233  *  The OS initialization, configuring of the adapter private structure,
2234  *  and a hardware reset occur.
2235  **/
2236 static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2237 {
2238 	struct net_device *netdev;
2239 	struct igb_adapter *adapter;
2240 	struct e1000_hw *hw;
2241 	u16 eeprom_data = 0;
2242 	s32 ret_val;
2243 	static int global_quad_port_a; /* global quad port a indication */
2244 	const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
2245 	int err, pci_using_dac;
2246 	u8 part_str[E1000_PBANUM_LENGTH];
2247 
2248 	/* Catch broken hardware that put the wrong VF device ID in
2249 	 * the PCIe SR-IOV capability.
2250 	 */
2251 	if (pdev->is_virtfn) {
2252 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
2253 			pci_name(pdev), pdev->vendor, pdev->device);
2254 		return -EINVAL;
2255 	}
2256 
2257 	err = pci_enable_device_mem(pdev);
2258 	if (err)
2259 		return err;
2260 
2261 	pci_using_dac = 0;
2262 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2263 	if (!err) {
2264 		pci_using_dac = 1;
2265 	} else {
2266 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
2267 		if (err) {
2268 			dev_err(&pdev->dev,
2269 				"No usable DMA configuration, aborting\n");
2270 			goto err_dma;
2271 		}
2272 	}
2273 
2274 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
2275 					   IORESOURCE_MEM),
2276 					   igb_driver_name);
2277 	if (err)
2278 		goto err_pci_reg;
2279 
2280 	pci_enable_pcie_error_reporting(pdev);
2281 
2282 	pci_set_master(pdev);
2283 	pci_save_state(pdev);
2284 
2285 	err = -ENOMEM;
2286 	netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
2287 				   IGB_MAX_TX_QUEUES);
2288 	if (!netdev)
2289 		goto err_alloc_etherdev;
2290 
2291 	SET_NETDEV_DEV(netdev, &pdev->dev);
2292 
2293 	pci_set_drvdata(pdev, netdev);
2294 	adapter = netdev_priv(netdev);
2295 	adapter->netdev = netdev;
2296 	adapter->pdev = pdev;
2297 	hw = &adapter->hw;
2298 	hw->back = adapter;
2299 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
2300 
2301 	err = -EIO;
2302 	hw->hw_addr = pci_iomap(pdev, 0, 0);
2303 	if (!hw->hw_addr)
2304 		goto err_ioremap;
2305 
2306 	netdev->netdev_ops = &igb_netdev_ops;
2307 	igb_set_ethtool_ops(netdev);
2308 	netdev->watchdog_timeo = 5 * HZ;
2309 
2310 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
2311 
2312 	netdev->mem_start = pci_resource_start(pdev, 0);
2313 	netdev->mem_end = pci_resource_end(pdev, 0);
2314 
2315 	/* PCI config space info */
2316 	hw->vendor_id = pdev->vendor;
2317 	hw->device_id = pdev->device;
2318 	hw->revision_id = pdev->revision;
2319 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
2320 	hw->subsystem_device_id = pdev->subsystem_device;
2321 
2322 	/* Copy the default MAC, PHY and NVM function pointers */
2323 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
2324 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
2325 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
2326 	/* Initialize skew-specific constants */
2327 	err = ei->get_invariants(hw);
2328 	if (err)
2329 		goto err_sw_init;
2330 
2331 	/* setup the private structure */
2332 	err = igb_sw_init(adapter);
2333 	if (err)
2334 		goto err_sw_init;
2335 
2336 	igb_get_bus_info_pcie(hw);
2337 
2338 	hw->phy.autoneg_wait_to_complete = false;
2339 
2340 	/* Copper options */
2341 	if (hw->phy.media_type == e1000_media_type_copper) {
2342 		hw->phy.mdix = AUTO_ALL_MODES;
2343 		hw->phy.disable_polarity_correction = false;
2344 		hw->phy.ms_type = e1000_ms_hw_default;
2345 	}
2346 
2347 	if (igb_check_reset_block(hw))
2348 		dev_info(&pdev->dev,
2349 			"PHY reset is blocked due to SOL/IDER session.\n");
2350 
2351 	/* features is initialized to 0 in allocation, it might have bits
2352 	 * set by igb_sw_init so we should use an or instead of an
2353 	 * assignment.
2354 	 */
2355 	netdev->features |= NETIF_F_SG |
2356 			    NETIF_F_IP_CSUM |
2357 			    NETIF_F_IPV6_CSUM |
2358 			    NETIF_F_TSO |
2359 			    NETIF_F_TSO6 |
2360 			    NETIF_F_RXHASH |
2361 			    NETIF_F_RXCSUM |
2362 			    NETIF_F_HW_VLAN_CTAG_RX |
2363 			    NETIF_F_HW_VLAN_CTAG_TX;
2364 
2365 	/* copy netdev features into list of user selectable features */
2366 	netdev->hw_features |= netdev->features;
2367 	netdev->hw_features |= NETIF_F_RXALL;
2368 
2369 	/* set this bit last since it cannot be part of hw_features */
2370 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
2371 
2372 	netdev->vlan_features |= NETIF_F_TSO |
2373 				 NETIF_F_TSO6 |
2374 				 NETIF_F_IP_CSUM |
2375 				 NETIF_F_IPV6_CSUM |
2376 				 NETIF_F_SG;
2377 
2378 	netdev->priv_flags |= IFF_SUPP_NOFCS;
2379 
2380 	if (pci_using_dac) {
2381 		netdev->features |= NETIF_F_HIGHDMA;
2382 		netdev->vlan_features |= NETIF_F_HIGHDMA;
2383 	}
2384 
2385 	if (hw->mac.type >= e1000_82576) {
2386 		netdev->hw_features |= NETIF_F_SCTP_CSUM;
2387 		netdev->features |= NETIF_F_SCTP_CSUM;
2388 	}
2389 
2390 	netdev->priv_flags |= IFF_UNICAST_FLT;
2391 
2392 	adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
2393 
2394 	/* before reading the NVM, reset the controller to put the device in a
2395 	 * known good starting state
2396 	 */
2397 	hw->mac.ops.reset_hw(hw);
2398 
2399 	/* make sure the NVM is good , i211/i210 parts can have special NVM
2400 	 * that doesn't contain a checksum
2401 	 */
2402 	switch (hw->mac.type) {
2403 	case e1000_i210:
2404 	case e1000_i211:
2405 		if (igb_get_flash_presence_i210(hw)) {
2406 			if (hw->nvm.ops.validate(hw) < 0) {
2407 				dev_err(&pdev->dev,
2408 					"The NVM Checksum Is Not Valid\n");
2409 				err = -EIO;
2410 				goto err_eeprom;
2411 			}
2412 		}
2413 		break;
2414 	default:
2415 		if (hw->nvm.ops.validate(hw) < 0) {
2416 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
2417 			err = -EIO;
2418 			goto err_eeprom;
2419 		}
2420 		break;
2421 	}
2422 
2423 	/* copy the MAC address out of the NVM */
2424 	if (hw->mac.ops.read_mac_addr(hw))
2425 		dev_err(&pdev->dev, "NVM Read Error\n");
2426 
2427 	memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
2428 
2429 	if (!is_valid_ether_addr(netdev->dev_addr)) {
2430 		dev_err(&pdev->dev, "Invalid MAC Address\n");
2431 		err = -EIO;
2432 		goto err_eeprom;
2433 	}
2434 
2435 	/* get firmware version for ethtool -i */
2436 	igb_set_fw_version(adapter);
2437 
2438 	/* configure RXPBSIZE and TXPBSIZE */
2439 	if (hw->mac.type == e1000_i210) {
2440 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
2441 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
2442 	}
2443 
2444 	setup_timer(&adapter->watchdog_timer, igb_watchdog,
2445 		    (unsigned long) adapter);
2446 	setup_timer(&adapter->phy_info_timer, igb_update_phy_info,
2447 		    (unsigned long) adapter);
2448 
2449 	INIT_WORK(&adapter->reset_task, igb_reset_task);
2450 	INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
2451 
2452 	/* Initialize link properties that are user-changeable */
2453 	adapter->fc_autoneg = true;
2454 	hw->mac.autoneg = true;
2455 	hw->phy.autoneg_advertised = 0x2f;
2456 
2457 	hw->fc.requested_mode = e1000_fc_default;
2458 	hw->fc.current_mode = e1000_fc_default;
2459 
2460 	igb_validate_mdi_setting(hw);
2461 
2462 	/* By default, support wake on port A */
2463 	if (hw->bus.func == 0)
2464 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2465 
2466 	/* Check the NVM for wake support on non-port A ports */
2467 	if (hw->mac.type >= e1000_82580)
2468 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
2469 				 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
2470 				 &eeprom_data);
2471 	else if (hw->bus.func == 1)
2472 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
2473 
2474 	if (eeprom_data & IGB_EEPROM_APME)
2475 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2476 
2477 	/* now that we have the eeprom settings, apply the special cases where
2478 	 * the eeprom may be wrong or the board simply won't support wake on
2479 	 * lan on a particular port
2480 	 */
2481 	switch (pdev->device) {
2482 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
2483 		adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2484 		break;
2485 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
2486 	case E1000_DEV_ID_82576_FIBER:
2487 	case E1000_DEV_ID_82576_SERDES:
2488 		/* Wake events only supported on port A for dual fiber
2489 		 * regardless of eeprom setting
2490 		 */
2491 		if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
2492 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2493 		break;
2494 	case E1000_DEV_ID_82576_QUAD_COPPER:
2495 	case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
2496 		/* if quad port adapter, disable WoL on all but port A */
2497 		if (global_quad_port_a != 0)
2498 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2499 		else
2500 			adapter->flags |= IGB_FLAG_QUAD_PORT_A;
2501 		/* Reset for multiple quad port adapters */
2502 		if (++global_quad_port_a == 4)
2503 			global_quad_port_a = 0;
2504 		break;
2505 	default:
2506 		/* If the device can't wake, don't set software support */
2507 		if (!device_can_wakeup(&adapter->pdev->dev))
2508 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2509 	}
2510 
2511 	/* initialize the wol settings based on the eeprom settings */
2512 	if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
2513 		adapter->wol |= E1000_WUFC_MAG;
2514 
2515 	/* Some vendors want WoL disabled by default, but still supported */
2516 	if ((hw->mac.type == e1000_i350) &&
2517 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
2518 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2519 		adapter->wol = 0;
2520 	}
2521 
2522 	device_set_wakeup_enable(&adapter->pdev->dev,
2523 				 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
2524 
2525 	/* reset the hardware with the new settings */
2526 	igb_reset(adapter);
2527 
2528 	/* Init the I2C interface */
2529 	err = igb_init_i2c(adapter);
2530 	if (err) {
2531 		dev_err(&pdev->dev, "failed to init i2c interface\n");
2532 		goto err_eeprom;
2533 	}
2534 
2535 	/* let the f/w know that the h/w is now under the control of the
2536 	 * driver.
2537 	 */
2538 	igb_get_hw_control(adapter);
2539 
2540 	strcpy(netdev->name, "eth%d");
2541 	err = register_netdev(netdev);
2542 	if (err)
2543 		goto err_register;
2544 
2545 	/* carrier off reporting is important to ethtool even BEFORE open */
2546 	netif_carrier_off(netdev);
2547 
2548 #ifdef CONFIG_IGB_DCA
2549 	if (dca_add_requester(&pdev->dev) == 0) {
2550 		adapter->flags |= IGB_FLAG_DCA_ENABLED;
2551 		dev_info(&pdev->dev, "DCA enabled\n");
2552 		igb_setup_dca(adapter);
2553 	}
2554 
2555 #endif
2556 #ifdef CONFIG_IGB_HWMON
2557 	/* Initialize the thermal sensor on i350 devices. */
2558 	if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
2559 		u16 ets_word;
2560 
2561 		/* Read the NVM to determine if this i350 device supports an
2562 		 * external thermal sensor.
2563 		 */
2564 		hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
2565 		if (ets_word != 0x0000 && ets_word != 0xFFFF)
2566 			adapter->ets = true;
2567 		else
2568 			adapter->ets = false;
2569 		if (igb_sysfs_init(adapter))
2570 			dev_err(&pdev->dev,
2571 				"failed to allocate sysfs resources\n");
2572 	} else {
2573 		adapter->ets = false;
2574 	}
2575 #endif
2576 	/* Check if Media Autosense is enabled */
2577 	adapter->ei = *ei;
2578 	if (hw->dev_spec._82575.mas_capable)
2579 		igb_init_mas(adapter);
2580 
2581 	/* do hw tstamp init after resetting */
2582 	igb_ptp_init(adapter);
2583 
2584 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
2585 	/* print bus type/speed/width info, not applicable to i354 */
2586 	if (hw->mac.type != e1000_i354) {
2587 		dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
2588 			 netdev->name,
2589 			 ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
2590 			  (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
2591 			   "unknown"),
2592 			 ((hw->bus.width == e1000_bus_width_pcie_x4) ?
2593 			  "Width x4" :
2594 			  (hw->bus.width == e1000_bus_width_pcie_x2) ?
2595 			  "Width x2" :
2596 			  (hw->bus.width == e1000_bus_width_pcie_x1) ?
2597 			  "Width x1" : "unknown"), netdev->dev_addr);
2598 	}
2599 
2600 	if ((hw->mac.type >= e1000_i210 ||
2601 	     igb_get_flash_presence_i210(hw))) {
2602 		ret_val = igb_read_part_string(hw, part_str,
2603 					       E1000_PBANUM_LENGTH);
2604 	} else {
2605 		ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
2606 	}
2607 
2608 	if (ret_val)
2609 		strcpy(part_str, "Unknown");
2610 	dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
2611 	dev_info(&pdev->dev,
2612 		"Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
2613 		(adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
2614 		(adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
2615 		adapter->num_rx_queues, adapter->num_tx_queues);
2616 	if (hw->phy.media_type == e1000_media_type_copper) {
2617 		switch (hw->mac.type) {
2618 		case e1000_i350:
2619 		case e1000_i210:
2620 		case e1000_i211:
2621 			/* Enable EEE for internal copper PHY devices */
2622 			err = igb_set_eee_i350(hw, true, true);
2623 			if ((!err) &&
2624 			    (!hw->dev_spec._82575.eee_disable)) {
2625 				adapter->eee_advert =
2626 					MDIO_EEE_100TX | MDIO_EEE_1000T;
2627 				adapter->flags |= IGB_FLAG_EEE;
2628 			}
2629 			break;
2630 		case e1000_i354:
2631 			if ((rd32(E1000_CTRL_EXT) &
2632 			    E1000_CTRL_EXT_LINK_MODE_SGMII)) {
2633 				err = igb_set_eee_i354(hw, true, true);
2634 				if ((!err) &&
2635 					(!hw->dev_spec._82575.eee_disable)) {
2636 					adapter->eee_advert =
2637 					   MDIO_EEE_100TX | MDIO_EEE_1000T;
2638 					adapter->flags |= IGB_FLAG_EEE;
2639 				}
2640 			}
2641 			break;
2642 		default:
2643 			break;
2644 		}
2645 	}
2646 	pm_runtime_put_noidle(&pdev->dev);
2647 	return 0;
2648 
2649 err_register:
2650 	igb_release_hw_control(adapter);
2651 	memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
2652 err_eeprom:
2653 	if (!igb_check_reset_block(hw))
2654 		igb_reset_phy(hw);
2655 
2656 	if (hw->flash_address)
2657 		iounmap(hw->flash_address);
2658 err_sw_init:
2659 	igb_clear_interrupt_scheme(adapter);
2660 	pci_iounmap(pdev, hw->hw_addr);
2661 err_ioremap:
2662 	free_netdev(netdev);
2663 err_alloc_etherdev:
2664 	pci_release_selected_regions(pdev,
2665 				     pci_select_bars(pdev, IORESOURCE_MEM));
2666 err_pci_reg:
2667 err_dma:
2668 	pci_disable_device(pdev);
2669 	return err;
2670 }
2671 
2672 #ifdef CONFIG_PCI_IOV
2673 static int igb_disable_sriov(struct pci_dev *pdev)
2674 {
2675 	struct net_device *netdev = pci_get_drvdata(pdev);
2676 	struct igb_adapter *adapter = netdev_priv(netdev);
2677 	struct e1000_hw *hw = &adapter->hw;
2678 
2679 	/* reclaim resources allocated to VFs */
2680 	if (adapter->vf_data) {
2681 		/* disable iov and allow time for transactions to clear */
2682 		if (pci_vfs_assigned(pdev)) {
2683 			dev_warn(&pdev->dev,
2684 				 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
2685 			return -EPERM;
2686 		} else {
2687 			pci_disable_sriov(pdev);
2688 			msleep(500);
2689 		}
2690 
2691 		kfree(adapter->vf_data);
2692 		adapter->vf_data = NULL;
2693 		adapter->vfs_allocated_count = 0;
2694 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
2695 		wrfl();
2696 		msleep(100);
2697 		dev_info(&pdev->dev, "IOV Disabled\n");
2698 
2699 		/* Re-enable DMA Coalescing flag since IOV is turned off */
2700 		adapter->flags |= IGB_FLAG_DMAC;
2701 	}
2702 
2703 	return 0;
2704 }
2705 
2706 static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
2707 {
2708 	struct net_device *netdev = pci_get_drvdata(pdev);
2709 	struct igb_adapter *adapter = netdev_priv(netdev);
2710 	int old_vfs = pci_num_vf(pdev);
2711 	int err = 0;
2712 	int i;
2713 
2714 	if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
2715 		err = -EPERM;
2716 		goto out;
2717 	}
2718 	if (!num_vfs)
2719 		goto out;
2720 
2721 	if (old_vfs) {
2722 		dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
2723 			 old_vfs, max_vfs);
2724 		adapter->vfs_allocated_count = old_vfs;
2725 	} else
2726 		adapter->vfs_allocated_count = num_vfs;
2727 
2728 	adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
2729 				sizeof(struct vf_data_storage), GFP_KERNEL);
2730 
2731 	/* if allocation failed then we do not support SR-IOV */
2732 	if (!adapter->vf_data) {
2733 		adapter->vfs_allocated_count = 0;
2734 		dev_err(&pdev->dev,
2735 			"Unable to allocate memory for VF Data Storage\n");
2736 		err = -ENOMEM;
2737 		goto out;
2738 	}
2739 
2740 	/* only call pci_enable_sriov() if no VFs are allocated already */
2741 	if (!old_vfs) {
2742 		err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
2743 		if (err)
2744 			goto err_out;
2745 	}
2746 	dev_info(&pdev->dev, "%d VFs allocated\n",
2747 		 adapter->vfs_allocated_count);
2748 	for (i = 0; i < adapter->vfs_allocated_count; i++)
2749 		igb_vf_configure(adapter, i);
2750 
2751 	/* DMA Coalescing is not supported in IOV mode. */
2752 	adapter->flags &= ~IGB_FLAG_DMAC;
2753 	goto out;
2754 
2755 err_out:
2756 	kfree(adapter->vf_data);
2757 	adapter->vf_data = NULL;
2758 	adapter->vfs_allocated_count = 0;
2759 out:
2760 	return err;
2761 }
2762 
2763 #endif
2764 /**
2765  *  igb_remove_i2c - Cleanup  I2C interface
2766  *  @adapter: pointer to adapter structure
2767  **/
2768 static void igb_remove_i2c(struct igb_adapter *adapter)
2769 {
2770 	/* free the adapter bus structure */
2771 	i2c_del_adapter(&adapter->i2c_adap);
2772 }
2773 
2774 /**
2775  *  igb_remove - Device Removal Routine
2776  *  @pdev: PCI device information struct
2777  *
2778  *  igb_remove is called by the PCI subsystem to alert the driver
2779  *  that it should release a PCI device.  The could be caused by a
2780  *  Hot-Plug event, or because the driver is going to be removed from
2781  *  memory.
2782  **/
2783 static void igb_remove(struct pci_dev *pdev)
2784 {
2785 	struct net_device *netdev = pci_get_drvdata(pdev);
2786 	struct igb_adapter *adapter = netdev_priv(netdev);
2787 	struct e1000_hw *hw = &adapter->hw;
2788 
2789 	pm_runtime_get_noresume(&pdev->dev);
2790 #ifdef CONFIG_IGB_HWMON
2791 	igb_sysfs_exit(adapter);
2792 #endif
2793 	igb_remove_i2c(adapter);
2794 	igb_ptp_stop(adapter);
2795 	/* The watchdog timer may be rescheduled, so explicitly
2796 	 * disable watchdog from being rescheduled.
2797 	 */
2798 	set_bit(__IGB_DOWN, &adapter->state);
2799 	del_timer_sync(&adapter->watchdog_timer);
2800 	del_timer_sync(&adapter->phy_info_timer);
2801 
2802 	cancel_work_sync(&adapter->reset_task);
2803 	cancel_work_sync(&adapter->watchdog_task);
2804 
2805 #ifdef CONFIG_IGB_DCA
2806 	if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
2807 		dev_info(&pdev->dev, "DCA disabled\n");
2808 		dca_remove_requester(&pdev->dev);
2809 		adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
2810 		wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
2811 	}
2812 #endif
2813 
2814 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
2815 	 * would have already happened in close and is redundant.
2816 	 */
2817 	igb_release_hw_control(adapter);
2818 
2819 	unregister_netdev(netdev);
2820 
2821 	igb_clear_interrupt_scheme(adapter);
2822 
2823 #ifdef CONFIG_PCI_IOV
2824 	igb_disable_sriov(pdev);
2825 #endif
2826 
2827 	pci_iounmap(pdev, hw->hw_addr);
2828 	if (hw->flash_address)
2829 		iounmap(hw->flash_address);
2830 	pci_release_selected_regions(pdev,
2831 				     pci_select_bars(pdev, IORESOURCE_MEM));
2832 
2833 	kfree(adapter->shadow_vfta);
2834 	free_netdev(netdev);
2835 
2836 	pci_disable_pcie_error_reporting(pdev);
2837 
2838 	pci_disable_device(pdev);
2839 }
2840 
2841 /**
2842  *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
2843  *  @adapter: board private structure to initialize
2844  *
2845  *  This function initializes the vf specific data storage and then attempts to
2846  *  allocate the VFs.  The reason for ordering it this way is because it is much
2847  *  mor expensive time wise to disable SR-IOV than it is to allocate and free
2848  *  the memory for the VFs.
2849  **/
2850 static void igb_probe_vfs(struct igb_adapter *adapter)
2851 {
2852 #ifdef CONFIG_PCI_IOV
2853 	struct pci_dev *pdev = adapter->pdev;
2854 	struct e1000_hw *hw = &adapter->hw;
2855 
2856 	/* Virtualization features not supported on i210 family. */
2857 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
2858 		return;
2859 
2860 	pci_sriov_set_totalvfs(pdev, 7);
2861 	igb_pci_enable_sriov(pdev, max_vfs);
2862 
2863 #endif /* CONFIG_PCI_IOV */
2864 }
2865 
2866 static void igb_init_queue_configuration(struct igb_adapter *adapter)
2867 {
2868 	struct e1000_hw *hw = &adapter->hw;
2869 	u32 max_rss_queues;
2870 
2871 	/* Determine the maximum number of RSS queues supported. */
2872 	switch (hw->mac.type) {
2873 	case e1000_i211:
2874 		max_rss_queues = IGB_MAX_RX_QUEUES_I211;
2875 		break;
2876 	case e1000_82575:
2877 	case e1000_i210:
2878 		max_rss_queues = IGB_MAX_RX_QUEUES_82575;
2879 		break;
2880 	case e1000_i350:
2881 		/* I350 cannot do RSS and SR-IOV at the same time */
2882 		if (!!adapter->vfs_allocated_count) {
2883 			max_rss_queues = 1;
2884 			break;
2885 		}
2886 		/* fall through */
2887 	case e1000_82576:
2888 		if (!!adapter->vfs_allocated_count) {
2889 			max_rss_queues = 2;
2890 			break;
2891 		}
2892 		/* fall through */
2893 	case e1000_82580:
2894 	case e1000_i354:
2895 	default:
2896 		max_rss_queues = IGB_MAX_RX_QUEUES;
2897 		break;
2898 	}
2899 
2900 	adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
2901 
2902 	/* Determine if we need to pair queues. */
2903 	switch (hw->mac.type) {
2904 	case e1000_82575:
2905 	case e1000_i211:
2906 		/* Device supports enough interrupts without queue pairing. */
2907 		break;
2908 	case e1000_82576:
2909 		/* If VFs are going to be allocated with RSS queues then we
2910 		 * should pair the queues in order to conserve interrupts due
2911 		 * to limited supply.
2912 		 */
2913 		if ((adapter->rss_queues > 1) &&
2914 		    (adapter->vfs_allocated_count > 6))
2915 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
2916 		/* fall through */
2917 	case e1000_82580:
2918 	case e1000_i350:
2919 	case e1000_i354:
2920 	case e1000_i210:
2921 	default:
2922 		/* If rss_queues > half of max_rss_queues, pair the queues in
2923 		 * order to conserve interrupts due to limited supply.
2924 		 */
2925 		if (adapter->rss_queues > (max_rss_queues / 2))
2926 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
2927 		break;
2928 	}
2929 }
2930 
2931 /**
2932  *  igb_sw_init - Initialize general software structures (struct igb_adapter)
2933  *  @adapter: board private structure to initialize
2934  *
2935  *  igb_sw_init initializes the Adapter private data structure.
2936  *  Fields are initialized based on PCI device information and
2937  *  OS network device settings (MTU size).
2938  **/
2939 static int igb_sw_init(struct igb_adapter *adapter)
2940 {
2941 	struct e1000_hw *hw = &adapter->hw;
2942 	struct net_device *netdev = adapter->netdev;
2943 	struct pci_dev *pdev = adapter->pdev;
2944 
2945 	pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
2946 
2947 	/* set default ring sizes */
2948 	adapter->tx_ring_count = IGB_DEFAULT_TXD;
2949 	adapter->rx_ring_count = IGB_DEFAULT_RXD;
2950 
2951 	/* set default ITR values */
2952 	adapter->rx_itr_setting = IGB_DEFAULT_ITR;
2953 	adapter->tx_itr_setting = IGB_DEFAULT_ITR;
2954 
2955 	/* set default work limits */
2956 	adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
2957 
2958 	adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN +
2959 				  VLAN_HLEN;
2960 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
2961 
2962 	spin_lock_init(&adapter->stats64_lock);
2963 #ifdef CONFIG_PCI_IOV
2964 	switch (hw->mac.type) {
2965 	case e1000_82576:
2966 	case e1000_i350:
2967 		if (max_vfs > 7) {
2968 			dev_warn(&pdev->dev,
2969 				 "Maximum of 7 VFs per PF, using max\n");
2970 			max_vfs = adapter->vfs_allocated_count = 7;
2971 		} else
2972 			adapter->vfs_allocated_count = max_vfs;
2973 		if (adapter->vfs_allocated_count)
2974 			dev_warn(&pdev->dev,
2975 				 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
2976 		break;
2977 	default:
2978 		break;
2979 	}
2980 #endif /* CONFIG_PCI_IOV */
2981 
2982 	igb_init_queue_configuration(adapter);
2983 
2984 	/* Setup and initialize a copy of the hw vlan table array */
2985 	adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
2986 				       GFP_ATOMIC);
2987 
2988 	/* This call may decrease the number of queues */
2989 	if (igb_init_interrupt_scheme(adapter, true)) {
2990 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
2991 		return -ENOMEM;
2992 	}
2993 
2994 	igb_probe_vfs(adapter);
2995 
2996 	/* Explicitly disable IRQ since the NIC can be in any state. */
2997 	igb_irq_disable(adapter);
2998 
2999 	if (hw->mac.type >= e1000_i350)
3000 		adapter->flags &= ~IGB_FLAG_DMAC;
3001 
3002 	set_bit(__IGB_DOWN, &adapter->state);
3003 	return 0;
3004 }
3005 
3006 /**
3007  *  igb_open - Called when a network interface is made active
3008  *  @netdev: network interface device structure
3009  *
3010  *  Returns 0 on success, negative value on failure
3011  *
3012  *  The open entry point is called when a network interface is made
3013  *  active by the system (IFF_UP).  At this point all resources needed
3014  *  for transmit and receive operations are allocated, the interrupt
3015  *  handler is registered with the OS, the watchdog timer is started,
3016  *  and the stack is notified that the interface is ready.
3017  **/
3018 static int __igb_open(struct net_device *netdev, bool resuming)
3019 {
3020 	struct igb_adapter *adapter = netdev_priv(netdev);
3021 	struct e1000_hw *hw = &adapter->hw;
3022 	struct pci_dev *pdev = adapter->pdev;
3023 	int err;
3024 	int i;
3025 
3026 	/* disallow open during test */
3027 	if (test_bit(__IGB_TESTING, &adapter->state)) {
3028 		WARN_ON(resuming);
3029 		return -EBUSY;
3030 	}
3031 
3032 	if (!resuming)
3033 		pm_runtime_get_sync(&pdev->dev);
3034 
3035 	netif_carrier_off(netdev);
3036 
3037 	/* allocate transmit descriptors */
3038 	err = igb_setup_all_tx_resources(adapter);
3039 	if (err)
3040 		goto err_setup_tx;
3041 
3042 	/* allocate receive descriptors */
3043 	err = igb_setup_all_rx_resources(adapter);
3044 	if (err)
3045 		goto err_setup_rx;
3046 
3047 	igb_power_up_link(adapter);
3048 
3049 	/* before we allocate an interrupt, we must be ready to handle it.
3050 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3051 	 * as soon as we call pci_request_irq, so we have to setup our
3052 	 * clean_rx handler before we do so.
3053 	 */
3054 	igb_configure(adapter);
3055 
3056 	err = igb_request_irq(adapter);
3057 	if (err)
3058 		goto err_req_irq;
3059 
3060 	/* Notify the stack of the actual queue counts. */
3061 	err = netif_set_real_num_tx_queues(adapter->netdev,
3062 					   adapter->num_tx_queues);
3063 	if (err)
3064 		goto err_set_queues;
3065 
3066 	err = netif_set_real_num_rx_queues(adapter->netdev,
3067 					   adapter->num_rx_queues);
3068 	if (err)
3069 		goto err_set_queues;
3070 
3071 	/* From here on the code is the same as igb_up() */
3072 	clear_bit(__IGB_DOWN, &adapter->state);
3073 
3074 	for (i = 0; i < adapter->num_q_vectors; i++)
3075 		napi_enable(&(adapter->q_vector[i]->napi));
3076 
3077 	/* Clear any pending interrupts. */
3078 	rd32(E1000_ICR);
3079 
3080 	igb_irq_enable(adapter);
3081 
3082 	/* notify VFs that reset has been completed */
3083 	if (adapter->vfs_allocated_count) {
3084 		u32 reg_data = rd32(E1000_CTRL_EXT);
3085 
3086 		reg_data |= E1000_CTRL_EXT_PFRSTD;
3087 		wr32(E1000_CTRL_EXT, reg_data);
3088 	}
3089 
3090 	netif_tx_start_all_queues(netdev);
3091 
3092 	if (!resuming)
3093 		pm_runtime_put(&pdev->dev);
3094 
3095 	/* start the watchdog. */
3096 	hw->mac.get_link_status = 1;
3097 	schedule_work(&adapter->watchdog_task);
3098 
3099 	return 0;
3100 
3101 err_set_queues:
3102 	igb_free_irq(adapter);
3103 err_req_irq:
3104 	igb_release_hw_control(adapter);
3105 	igb_power_down_link(adapter);
3106 	igb_free_all_rx_resources(adapter);
3107 err_setup_rx:
3108 	igb_free_all_tx_resources(adapter);
3109 err_setup_tx:
3110 	igb_reset(adapter);
3111 	if (!resuming)
3112 		pm_runtime_put(&pdev->dev);
3113 
3114 	return err;
3115 }
3116 
3117 static int igb_open(struct net_device *netdev)
3118 {
3119 	return __igb_open(netdev, false);
3120 }
3121 
3122 /**
3123  *  igb_close - Disables a network interface
3124  *  @netdev: network interface device structure
3125  *
3126  *  Returns 0, this is not allowed to fail
3127  *
3128  *  The close entry point is called when an interface is de-activated
3129  *  by the OS.  The hardware is still under the driver's control, but
3130  *  needs to be disabled.  A global MAC reset is issued to stop the
3131  *  hardware, and all transmit and receive resources are freed.
3132  **/
3133 static int __igb_close(struct net_device *netdev, bool suspending)
3134 {
3135 	struct igb_adapter *adapter = netdev_priv(netdev);
3136 	struct pci_dev *pdev = adapter->pdev;
3137 
3138 	WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
3139 
3140 	if (!suspending)
3141 		pm_runtime_get_sync(&pdev->dev);
3142 
3143 	igb_down(adapter);
3144 	igb_free_irq(adapter);
3145 
3146 	igb_free_all_tx_resources(adapter);
3147 	igb_free_all_rx_resources(adapter);
3148 
3149 	if (!suspending)
3150 		pm_runtime_put_sync(&pdev->dev);
3151 	return 0;
3152 }
3153 
3154 static int igb_close(struct net_device *netdev)
3155 {
3156 	return __igb_close(netdev, false);
3157 }
3158 
3159 /**
3160  *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
3161  *  @tx_ring: tx descriptor ring (for a specific queue) to setup
3162  *
3163  *  Return 0 on success, negative on failure
3164  **/
3165 int igb_setup_tx_resources(struct igb_ring *tx_ring)
3166 {
3167 	struct device *dev = tx_ring->dev;
3168 	int size;
3169 
3170 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3171 
3172 	tx_ring->tx_buffer_info = vzalloc(size);
3173 	if (!tx_ring->tx_buffer_info)
3174 		goto err;
3175 
3176 	/* round up to nearest 4K */
3177 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
3178 	tx_ring->size = ALIGN(tx_ring->size, 4096);
3179 
3180 	tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
3181 					   &tx_ring->dma, GFP_KERNEL);
3182 	if (!tx_ring->desc)
3183 		goto err;
3184 
3185 	tx_ring->next_to_use = 0;
3186 	tx_ring->next_to_clean = 0;
3187 
3188 	return 0;
3189 
3190 err:
3191 	vfree(tx_ring->tx_buffer_info);
3192 	tx_ring->tx_buffer_info = NULL;
3193 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
3194 	return -ENOMEM;
3195 }
3196 
3197 /**
3198  *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
3199  *				 (Descriptors) for all queues
3200  *  @adapter: board private structure
3201  *
3202  *  Return 0 on success, negative on failure
3203  **/
3204 static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
3205 {
3206 	struct pci_dev *pdev = adapter->pdev;
3207 	int i, err = 0;
3208 
3209 	for (i = 0; i < adapter->num_tx_queues; i++) {
3210 		err = igb_setup_tx_resources(adapter->tx_ring[i]);
3211 		if (err) {
3212 			dev_err(&pdev->dev,
3213 				"Allocation for Tx Queue %u failed\n", i);
3214 			for (i--; i >= 0; i--)
3215 				igb_free_tx_resources(adapter->tx_ring[i]);
3216 			break;
3217 		}
3218 	}
3219 
3220 	return err;
3221 }
3222 
3223 /**
3224  *  igb_setup_tctl - configure the transmit control registers
3225  *  @adapter: Board private structure
3226  **/
3227 void igb_setup_tctl(struct igb_adapter *adapter)
3228 {
3229 	struct e1000_hw *hw = &adapter->hw;
3230 	u32 tctl;
3231 
3232 	/* disable queue 0 which is enabled by default on 82575 and 82576 */
3233 	wr32(E1000_TXDCTL(0), 0);
3234 
3235 	/* Program the Transmit Control Register */
3236 	tctl = rd32(E1000_TCTL);
3237 	tctl &= ~E1000_TCTL_CT;
3238 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
3239 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
3240 
3241 	igb_config_collision_dist(hw);
3242 
3243 	/* Enable transmits */
3244 	tctl |= E1000_TCTL_EN;
3245 
3246 	wr32(E1000_TCTL, tctl);
3247 }
3248 
3249 /**
3250  *  igb_configure_tx_ring - Configure transmit ring after Reset
3251  *  @adapter: board private structure
3252  *  @ring: tx ring to configure
3253  *
3254  *  Configure a transmit ring after a reset.
3255  **/
3256 void igb_configure_tx_ring(struct igb_adapter *adapter,
3257 			   struct igb_ring *ring)
3258 {
3259 	struct e1000_hw *hw = &adapter->hw;
3260 	u32 txdctl = 0;
3261 	u64 tdba = ring->dma;
3262 	int reg_idx = ring->reg_idx;
3263 
3264 	/* disable the queue */
3265 	wr32(E1000_TXDCTL(reg_idx), 0);
3266 	wrfl();
3267 	mdelay(10);
3268 
3269 	wr32(E1000_TDLEN(reg_idx),
3270 	     ring->count * sizeof(union e1000_adv_tx_desc));
3271 	wr32(E1000_TDBAL(reg_idx),
3272 	     tdba & 0x00000000ffffffffULL);
3273 	wr32(E1000_TDBAH(reg_idx), tdba >> 32);
3274 
3275 	ring->tail = hw->hw_addr + E1000_TDT(reg_idx);
3276 	wr32(E1000_TDH(reg_idx), 0);
3277 	writel(0, ring->tail);
3278 
3279 	txdctl |= IGB_TX_PTHRESH;
3280 	txdctl |= IGB_TX_HTHRESH << 8;
3281 	txdctl |= IGB_TX_WTHRESH << 16;
3282 
3283 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
3284 	wr32(E1000_TXDCTL(reg_idx), txdctl);
3285 }
3286 
3287 /**
3288  *  igb_configure_tx - Configure transmit Unit after Reset
3289  *  @adapter: board private structure
3290  *
3291  *  Configure the Tx unit of the MAC after a reset.
3292  **/
3293 static void igb_configure_tx(struct igb_adapter *adapter)
3294 {
3295 	int i;
3296 
3297 	for (i = 0; i < adapter->num_tx_queues; i++)
3298 		igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
3299 }
3300 
3301 /**
3302  *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
3303  *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
3304  *
3305  *  Returns 0 on success, negative on failure
3306  **/
3307 int igb_setup_rx_resources(struct igb_ring *rx_ring)
3308 {
3309 	struct device *dev = rx_ring->dev;
3310 	int size;
3311 
3312 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3313 
3314 	rx_ring->rx_buffer_info = vzalloc(size);
3315 	if (!rx_ring->rx_buffer_info)
3316 		goto err;
3317 
3318 	/* Round up to nearest 4K */
3319 	rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
3320 	rx_ring->size = ALIGN(rx_ring->size, 4096);
3321 
3322 	rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
3323 					   &rx_ring->dma, GFP_KERNEL);
3324 	if (!rx_ring->desc)
3325 		goto err;
3326 
3327 	rx_ring->next_to_alloc = 0;
3328 	rx_ring->next_to_clean = 0;
3329 	rx_ring->next_to_use = 0;
3330 
3331 	return 0;
3332 
3333 err:
3334 	vfree(rx_ring->rx_buffer_info);
3335 	rx_ring->rx_buffer_info = NULL;
3336 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
3337 	return -ENOMEM;
3338 }
3339 
3340 /**
3341  *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
3342  *				 (Descriptors) for all queues
3343  *  @adapter: board private structure
3344  *
3345  *  Return 0 on success, negative on failure
3346  **/
3347 static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
3348 {
3349 	struct pci_dev *pdev = adapter->pdev;
3350 	int i, err = 0;
3351 
3352 	for (i = 0; i < adapter->num_rx_queues; i++) {
3353 		err = igb_setup_rx_resources(adapter->rx_ring[i]);
3354 		if (err) {
3355 			dev_err(&pdev->dev,
3356 				"Allocation for Rx Queue %u failed\n", i);
3357 			for (i--; i >= 0; i--)
3358 				igb_free_rx_resources(adapter->rx_ring[i]);
3359 			break;
3360 		}
3361 	}
3362 
3363 	return err;
3364 }
3365 
3366 /**
3367  *  igb_setup_mrqc - configure the multiple receive queue control registers
3368  *  @adapter: Board private structure
3369  **/
3370 static void igb_setup_mrqc(struct igb_adapter *adapter)
3371 {
3372 	struct e1000_hw *hw = &adapter->hw;
3373 	u32 mrqc, rxcsum;
3374 	u32 j, num_rx_queues;
3375 	u32 rss_key[10];
3376 
3377 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3378 	for (j = 0; j < 10; j++)
3379 		wr32(E1000_RSSRK(j), rss_key[j]);
3380 
3381 	num_rx_queues = adapter->rss_queues;
3382 
3383 	switch (hw->mac.type) {
3384 	case e1000_82576:
3385 		/* 82576 supports 2 RSS queues for SR-IOV */
3386 		if (adapter->vfs_allocated_count)
3387 			num_rx_queues = 2;
3388 		break;
3389 	default:
3390 		break;
3391 	}
3392 
3393 	if (adapter->rss_indir_tbl_init != num_rx_queues) {
3394 		for (j = 0; j < IGB_RETA_SIZE; j++)
3395 			adapter->rss_indir_tbl[j] =
3396 			(j * num_rx_queues) / IGB_RETA_SIZE;
3397 		adapter->rss_indir_tbl_init = num_rx_queues;
3398 	}
3399 	igb_write_rss_indir_tbl(adapter);
3400 
3401 	/* Disable raw packet checksumming so that RSS hash is placed in
3402 	 * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
3403 	 * offloads as they are enabled by default
3404 	 */
3405 	rxcsum = rd32(E1000_RXCSUM);
3406 	rxcsum |= E1000_RXCSUM_PCSD;
3407 
3408 	if (adapter->hw.mac.type >= e1000_82576)
3409 		/* Enable Receive Checksum Offload for SCTP */
3410 		rxcsum |= E1000_RXCSUM_CRCOFL;
3411 
3412 	/* Don't need to set TUOFL or IPOFL, they default to 1 */
3413 	wr32(E1000_RXCSUM, rxcsum);
3414 
3415 	/* Generate RSS hash based on packet types, TCP/UDP
3416 	 * port numbers and/or IPv4/v6 src and dst addresses
3417 	 */
3418 	mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
3419 	       E1000_MRQC_RSS_FIELD_IPV4_TCP |
3420 	       E1000_MRQC_RSS_FIELD_IPV6 |
3421 	       E1000_MRQC_RSS_FIELD_IPV6_TCP |
3422 	       E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
3423 
3424 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
3425 		mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
3426 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
3427 		mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
3428 
3429 	/* If VMDq is enabled then we set the appropriate mode for that, else
3430 	 * we default to RSS so that an RSS hash is calculated per packet even
3431 	 * if we are only using one queue
3432 	 */
3433 	if (adapter->vfs_allocated_count) {
3434 		if (hw->mac.type > e1000_82575) {
3435 			/* Set the default pool for the PF's first queue */
3436 			u32 vtctl = rd32(E1000_VT_CTL);
3437 
3438 			vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
3439 				   E1000_VT_CTL_DISABLE_DEF_POOL);
3440 			vtctl |= adapter->vfs_allocated_count <<
3441 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
3442 			wr32(E1000_VT_CTL, vtctl);
3443 		}
3444 		if (adapter->rss_queues > 1)
3445 			mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_2Q;
3446 		else
3447 			mrqc |= E1000_MRQC_ENABLE_VMDQ;
3448 	} else {
3449 		if (hw->mac.type != e1000_i211)
3450 			mrqc |= E1000_MRQC_ENABLE_RSS_4Q;
3451 	}
3452 	igb_vmm_control(adapter);
3453 
3454 	wr32(E1000_MRQC, mrqc);
3455 }
3456 
3457 /**
3458  *  igb_setup_rctl - configure the receive control registers
3459  *  @adapter: Board private structure
3460  **/
3461 void igb_setup_rctl(struct igb_adapter *adapter)
3462 {
3463 	struct e1000_hw *hw = &adapter->hw;
3464 	u32 rctl;
3465 
3466 	rctl = rd32(E1000_RCTL);
3467 
3468 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3469 	rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
3470 
3471 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
3472 		(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3473 
3474 	/* enable stripping of CRC. It's unlikely this will break BMC
3475 	 * redirection as it did with e1000. Newer features require
3476 	 * that the HW strips the CRC.
3477 	 */
3478 	rctl |= E1000_RCTL_SECRC;
3479 
3480 	/* disable store bad packets and clear size bits. */
3481 	rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
3482 
3483 	/* enable LPE to prevent packets larger than max_frame_size */
3484 	rctl |= E1000_RCTL_LPE;
3485 
3486 	/* disable queue 0 to prevent tail write w/o re-config */
3487 	wr32(E1000_RXDCTL(0), 0);
3488 
3489 	/* Attention!!!  For SR-IOV PF driver operations you must enable
3490 	 * queue drop for all VF and PF queues to prevent head of line blocking
3491 	 * if an un-trusted VF does not provide descriptors to hardware.
3492 	 */
3493 	if (adapter->vfs_allocated_count) {
3494 		/* set all queue drop enable bits */
3495 		wr32(E1000_QDE, ALL_QUEUES);
3496 	}
3497 
3498 	/* This is useful for sniffing bad packets. */
3499 	if (adapter->netdev->features & NETIF_F_RXALL) {
3500 		/* UPE and MPE will be handled by normal PROMISC logic
3501 		 * in e1000e_set_rx_mode
3502 		 */
3503 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
3504 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
3505 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
3506 
3507 		rctl &= ~(E1000_RCTL_VFE | /* Disable VLAN filter */
3508 			  E1000_RCTL_DPF | /* Allow filtered pause */
3509 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
3510 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3511 		 * and that breaks VLANs.
3512 		 */
3513 	}
3514 
3515 	wr32(E1000_RCTL, rctl);
3516 }
3517 
3518 static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
3519 				   int vfn)
3520 {
3521 	struct e1000_hw *hw = &adapter->hw;
3522 	u32 vmolr;
3523 
3524 	/* if it isn't the PF check to see if VFs are enabled and
3525 	 * increase the size to support vlan tags
3526 	 */
3527 	if (vfn < adapter->vfs_allocated_count &&
3528 	    adapter->vf_data[vfn].vlans_enabled)
3529 		size += VLAN_TAG_SIZE;
3530 
3531 	vmolr = rd32(E1000_VMOLR(vfn));
3532 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
3533 	vmolr |= size | E1000_VMOLR_LPE;
3534 	wr32(E1000_VMOLR(vfn), vmolr);
3535 
3536 	return 0;
3537 }
3538 
3539 /**
3540  *  igb_rlpml_set - set maximum receive packet size
3541  *  @adapter: board private structure
3542  *
3543  *  Configure maximum receivable packet size.
3544  **/
3545 static void igb_rlpml_set(struct igb_adapter *adapter)
3546 {
3547 	u32 max_frame_size = adapter->max_frame_size;
3548 	struct e1000_hw *hw = &adapter->hw;
3549 	u16 pf_id = adapter->vfs_allocated_count;
3550 
3551 	if (pf_id) {
3552 		igb_set_vf_rlpml(adapter, max_frame_size, pf_id);
3553 		/* If we're in VMDQ or SR-IOV mode, then set global RLPML
3554 		 * to our max jumbo frame size, in case we need to enable
3555 		 * jumbo frames on one of the rings later.
3556 		 * This will not pass over-length frames into the default
3557 		 * queue because it's gated by the VMOLR.RLPML.
3558 		 */
3559 		max_frame_size = MAX_JUMBO_FRAME_SIZE;
3560 	}
3561 
3562 	wr32(E1000_RLPML, max_frame_size);
3563 }
3564 
3565 static inline void igb_set_vmolr(struct igb_adapter *adapter,
3566 				 int vfn, bool aupe)
3567 {
3568 	struct e1000_hw *hw = &adapter->hw;
3569 	u32 vmolr;
3570 
3571 	/* This register exists only on 82576 and newer so if we are older then
3572 	 * we should exit and do nothing
3573 	 */
3574 	if (hw->mac.type < e1000_82576)
3575 		return;
3576 
3577 	vmolr = rd32(E1000_VMOLR(vfn));
3578 	vmolr |= E1000_VMOLR_STRVLAN; /* Strip vlan tags */
3579 	if (hw->mac.type == e1000_i350) {
3580 		u32 dvmolr;
3581 
3582 		dvmolr = rd32(E1000_DVMOLR(vfn));
3583 		dvmolr |= E1000_DVMOLR_STRVLAN;
3584 		wr32(E1000_DVMOLR(vfn), dvmolr);
3585 	}
3586 	if (aupe)
3587 		vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
3588 	else
3589 		vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
3590 
3591 	/* clear all bits that might not be set */
3592 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
3593 
3594 	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
3595 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
3596 	/* for VMDq only allow the VFs and pool 0 to accept broadcast and
3597 	 * multicast packets
3598 	 */
3599 	if (vfn <= adapter->vfs_allocated_count)
3600 		vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
3601 
3602 	wr32(E1000_VMOLR(vfn), vmolr);
3603 }
3604 
3605 /**
3606  *  igb_configure_rx_ring - Configure a receive ring after Reset
3607  *  @adapter: board private structure
3608  *  @ring: receive ring to be configured
3609  *
3610  *  Configure the Rx unit of the MAC after a reset.
3611  **/
3612 void igb_configure_rx_ring(struct igb_adapter *adapter,
3613 			   struct igb_ring *ring)
3614 {
3615 	struct e1000_hw *hw = &adapter->hw;
3616 	u64 rdba = ring->dma;
3617 	int reg_idx = ring->reg_idx;
3618 	u32 srrctl = 0, rxdctl = 0;
3619 
3620 	/* disable the queue */
3621 	wr32(E1000_RXDCTL(reg_idx), 0);
3622 
3623 	/* Set DMA base address registers */
3624 	wr32(E1000_RDBAL(reg_idx),
3625 	     rdba & 0x00000000ffffffffULL);
3626 	wr32(E1000_RDBAH(reg_idx), rdba >> 32);
3627 	wr32(E1000_RDLEN(reg_idx),
3628 	     ring->count * sizeof(union e1000_adv_rx_desc));
3629 
3630 	/* initialize head and tail */
3631 	ring->tail = hw->hw_addr + E1000_RDT(reg_idx);
3632 	wr32(E1000_RDH(reg_idx), 0);
3633 	writel(0, ring->tail);
3634 
3635 	/* set descriptor configuration */
3636 	srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
3637 	srrctl |= IGB_RX_BUFSZ >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3638 	srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3639 	if (hw->mac.type >= e1000_82580)
3640 		srrctl |= E1000_SRRCTL_TIMESTAMP;
3641 	/* Only set Drop Enable if we are supporting multiple queues */
3642 	if (adapter->vfs_allocated_count || adapter->num_rx_queues > 1)
3643 		srrctl |= E1000_SRRCTL_DROP_EN;
3644 
3645 	wr32(E1000_SRRCTL(reg_idx), srrctl);
3646 
3647 	/* set filtering for VMDQ pools */
3648 	igb_set_vmolr(adapter, reg_idx & 0x7, true);
3649 
3650 	rxdctl |= IGB_RX_PTHRESH;
3651 	rxdctl |= IGB_RX_HTHRESH << 8;
3652 	rxdctl |= IGB_RX_WTHRESH << 16;
3653 
3654 	/* enable receive descriptor fetching */
3655 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3656 	wr32(E1000_RXDCTL(reg_idx), rxdctl);
3657 }
3658 
3659 /**
3660  *  igb_configure_rx - Configure receive Unit after Reset
3661  *  @adapter: board private structure
3662  *
3663  *  Configure the Rx unit of the MAC after a reset.
3664  **/
3665 static void igb_configure_rx(struct igb_adapter *adapter)
3666 {
3667 	int i;
3668 
3669 	/* set UTA to appropriate mode */
3670 	igb_set_uta(adapter);
3671 
3672 	/* set the correct pool for the PF default MAC address in entry 0 */
3673 	igb_rar_set_qsel(adapter, adapter->hw.mac.addr, 0,
3674 			 adapter->vfs_allocated_count);
3675 
3676 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3677 	 * the Base and Length of the Rx Descriptor Ring
3678 	 */
3679 	for (i = 0; i < adapter->num_rx_queues; i++)
3680 		igb_configure_rx_ring(adapter, adapter->rx_ring[i]);
3681 }
3682 
3683 /**
3684  *  igb_free_tx_resources - Free Tx Resources per Queue
3685  *  @tx_ring: Tx descriptor ring for a specific queue
3686  *
3687  *  Free all transmit software resources
3688  **/
3689 void igb_free_tx_resources(struct igb_ring *tx_ring)
3690 {
3691 	igb_clean_tx_ring(tx_ring);
3692 
3693 	vfree(tx_ring->tx_buffer_info);
3694 	tx_ring->tx_buffer_info = NULL;
3695 
3696 	/* if not set, then don't free */
3697 	if (!tx_ring->desc)
3698 		return;
3699 
3700 	dma_free_coherent(tx_ring->dev, tx_ring->size,
3701 			  tx_ring->desc, tx_ring->dma);
3702 
3703 	tx_ring->desc = NULL;
3704 }
3705 
3706 /**
3707  *  igb_free_all_tx_resources - Free Tx Resources for All Queues
3708  *  @adapter: board private structure
3709  *
3710  *  Free all transmit software resources
3711  **/
3712 static void igb_free_all_tx_resources(struct igb_adapter *adapter)
3713 {
3714 	int i;
3715 
3716 	for (i = 0; i < adapter->num_tx_queues; i++)
3717 		if (adapter->tx_ring[i])
3718 			igb_free_tx_resources(adapter->tx_ring[i]);
3719 }
3720 
3721 void igb_unmap_and_free_tx_resource(struct igb_ring *ring,
3722 				    struct igb_tx_buffer *tx_buffer)
3723 {
3724 	if (tx_buffer->skb) {
3725 		dev_kfree_skb_any(tx_buffer->skb);
3726 		if (dma_unmap_len(tx_buffer, len))
3727 			dma_unmap_single(ring->dev,
3728 					 dma_unmap_addr(tx_buffer, dma),
3729 					 dma_unmap_len(tx_buffer, len),
3730 					 DMA_TO_DEVICE);
3731 	} else if (dma_unmap_len(tx_buffer, len)) {
3732 		dma_unmap_page(ring->dev,
3733 			       dma_unmap_addr(tx_buffer, dma),
3734 			       dma_unmap_len(tx_buffer, len),
3735 			       DMA_TO_DEVICE);
3736 	}
3737 	tx_buffer->next_to_watch = NULL;
3738 	tx_buffer->skb = NULL;
3739 	dma_unmap_len_set(tx_buffer, len, 0);
3740 	/* buffer_info must be completely set up in the transmit path */
3741 }
3742 
3743 /**
3744  *  igb_clean_tx_ring - Free Tx Buffers
3745  *  @tx_ring: ring to be cleaned
3746  **/
3747 static void igb_clean_tx_ring(struct igb_ring *tx_ring)
3748 {
3749 	struct igb_tx_buffer *buffer_info;
3750 	unsigned long size;
3751 	u16 i;
3752 
3753 	if (!tx_ring->tx_buffer_info)
3754 		return;
3755 	/* Free all the Tx ring sk_buffs */
3756 
3757 	for (i = 0; i < tx_ring->count; i++) {
3758 		buffer_info = &tx_ring->tx_buffer_info[i];
3759 		igb_unmap_and_free_tx_resource(tx_ring, buffer_info);
3760 	}
3761 
3762 	netdev_tx_reset_queue(txring_txq(tx_ring));
3763 
3764 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3765 	memset(tx_ring->tx_buffer_info, 0, size);
3766 
3767 	/* Zero out the descriptor ring */
3768 	memset(tx_ring->desc, 0, tx_ring->size);
3769 
3770 	tx_ring->next_to_use = 0;
3771 	tx_ring->next_to_clean = 0;
3772 }
3773 
3774 /**
3775  *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
3776  *  @adapter: board private structure
3777  **/
3778 static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
3779 {
3780 	int i;
3781 
3782 	for (i = 0; i < adapter->num_tx_queues; i++)
3783 		if (adapter->tx_ring[i])
3784 			igb_clean_tx_ring(adapter->tx_ring[i]);
3785 }
3786 
3787 /**
3788  *  igb_free_rx_resources - Free Rx Resources
3789  *  @rx_ring: ring to clean the resources from
3790  *
3791  *  Free all receive software resources
3792  **/
3793 void igb_free_rx_resources(struct igb_ring *rx_ring)
3794 {
3795 	igb_clean_rx_ring(rx_ring);
3796 
3797 	vfree(rx_ring->rx_buffer_info);
3798 	rx_ring->rx_buffer_info = NULL;
3799 
3800 	/* if not set, then don't free */
3801 	if (!rx_ring->desc)
3802 		return;
3803 
3804 	dma_free_coherent(rx_ring->dev, rx_ring->size,
3805 			  rx_ring->desc, rx_ring->dma);
3806 
3807 	rx_ring->desc = NULL;
3808 }
3809 
3810 /**
3811  *  igb_free_all_rx_resources - Free Rx Resources for All Queues
3812  *  @adapter: board private structure
3813  *
3814  *  Free all receive software resources
3815  **/
3816 static void igb_free_all_rx_resources(struct igb_adapter *adapter)
3817 {
3818 	int i;
3819 
3820 	for (i = 0; i < adapter->num_rx_queues; i++)
3821 		if (adapter->rx_ring[i])
3822 			igb_free_rx_resources(adapter->rx_ring[i]);
3823 }
3824 
3825 /**
3826  *  igb_clean_rx_ring - Free Rx Buffers per Queue
3827  *  @rx_ring: ring to free buffers from
3828  **/
3829 static void igb_clean_rx_ring(struct igb_ring *rx_ring)
3830 {
3831 	unsigned long size;
3832 	u16 i;
3833 
3834 	if (rx_ring->skb)
3835 		dev_kfree_skb(rx_ring->skb);
3836 	rx_ring->skb = NULL;
3837 
3838 	if (!rx_ring->rx_buffer_info)
3839 		return;
3840 
3841 	/* Free all the Rx ring sk_buffs */
3842 	for (i = 0; i < rx_ring->count; i++) {
3843 		struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
3844 
3845 		if (!buffer_info->page)
3846 			continue;
3847 
3848 		dma_unmap_page(rx_ring->dev,
3849 			       buffer_info->dma,
3850 			       PAGE_SIZE,
3851 			       DMA_FROM_DEVICE);
3852 		__free_page(buffer_info->page);
3853 
3854 		buffer_info->page = NULL;
3855 	}
3856 
3857 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3858 	memset(rx_ring->rx_buffer_info, 0, size);
3859 
3860 	/* Zero out the descriptor ring */
3861 	memset(rx_ring->desc, 0, rx_ring->size);
3862 
3863 	rx_ring->next_to_alloc = 0;
3864 	rx_ring->next_to_clean = 0;
3865 	rx_ring->next_to_use = 0;
3866 }
3867 
3868 /**
3869  *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
3870  *  @adapter: board private structure
3871  **/
3872 static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
3873 {
3874 	int i;
3875 
3876 	for (i = 0; i < adapter->num_rx_queues; i++)
3877 		if (adapter->rx_ring[i])
3878 			igb_clean_rx_ring(adapter->rx_ring[i]);
3879 }
3880 
3881 /**
3882  *  igb_set_mac - Change the Ethernet Address of the NIC
3883  *  @netdev: network interface device structure
3884  *  @p: pointer to an address structure
3885  *
3886  *  Returns 0 on success, negative on failure
3887  **/
3888 static int igb_set_mac(struct net_device *netdev, void *p)
3889 {
3890 	struct igb_adapter *adapter = netdev_priv(netdev);
3891 	struct e1000_hw *hw = &adapter->hw;
3892 	struct sockaddr *addr = p;
3893 
3894 	if (!is_valid_ether_addr(addr->sa_data))
3895 		return -EADDRNOTAVAIL;
3896 
3897 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
3898 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
3899 
3900 	/* set the correct pool for the new PF MAC address in entry 0 */
3901 	igb_rar_set_qsel(adapter, hw->mac.addr, 0,
3902 			 adapter->vfs_allocated_count);
3903 
3904 	return 0;
3905 }
3906 
3907 /**
3908  *  igb_write_mc_addr_list - write multicast addresses to MTA
3909  *  @netdev: network interface device structure
3910  *
3911  *  Writes multicast address list to the MTA hash table.
3912  *  Returns: -ENOMEM on failure
3913  *           0 on no addresses written
3914  *           X on writing X addresses to MTA
3915  **/
3916 static int igb_write_mc_addr_list(struct net_device *netdev)
3917 {
3918 	struct igb_adapter *adapter = netdev_priv(netdev);
3919 	struct e1000_hw *hw = &adapter->hw;
3920 	struct netdev_hw_addr *ha;
3921 	u8  *mta_list;
3922 	int i;
3923 
3924 	if (netdev_mc_empty(netdev)) {
3925 		/* nothing to program, so clear mc list */
3926 		igb_update_mc_addr_list(hw, NULL, 0);
3927 		igb_restore_vf_multicasts(adapter);
3928 		return 0;
3929 	}
3930 
3931 	mta_list = kzalloc(netdev_mc_count(netdev) * 6, GFP_ATOMIC);
3932 	if (!mta_list)
3933 		return -ENOMEM;
3934 
3935 	/* The shared function expects a packed array of only addresses. */
3936 	i = 0;
3937 	netdev_for_each_mc_addr(ha, netdev)
3938 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3939 
3940 	igb_update_mc_addr_list(hw, mta_list, i);
3941 	kfree(mta_list);
3942 
3943 	return netdev_mc_count(netdev);
3944 }
3945 
3946 /**
3947  *  igb_write_uc_addr_list - write unicast addresses to RAR table
3948  *  @netdev: network interface device structure
3949  *
3950  *  Writes unicast address list to the RAR table.
3951  *  Returns: -ENOMEM on failure/insufficient address space
3952  *           0 on no addresses written
3953  *           X on writing X addresses to the RAR table
3954  **/
3955 static int igb_write_uc_addr_list(struct net_device *netdev)
3956 {
3957 	struct igb_adapter *adapter = netdev_priv(netdev);
3958 	struct e1000_hw *hw = &adapter->hw;
3959 	unsigned int vfn = adapter->vfs_allocated_count;
3960 	unsigned int rar_entries = hw->mac.rar_entry_count - (vfn + 1);
3961 	int count = 0;
3962 
3963 	/* return ENOMEM indicating insufficient memory for addresses */
3964 	if (netdev_uc_count(netdev) > rar_entries)
3965 		return -ENOMEM;
3966 
3967 	if (!netdev_uc_empty(netdev) && rar_entries) {
3968 		struct netdev_hw_addr *ha;
3969 
3970 		netdev_for_each_uc_addr(ha, netdev) {
3971 			if (!rar_entries)
3972 				break;
3973 			igb_rar_set_qsel(adapter, ha->addr,
3974 					 rar_entries--,
3975 					 vfn);
3976 			count++;
3977 		}
3978 	}
3979 	/* write the addresses in reverse order to avoid write combining */
3980 	for (; rar_entries > 0 ; rar_entries--) {
3981 		wr32(E1000_RAH(rar_entries), 0);
3982 		wr32(E1000_RAL(rar_entries), 0);
3983 	}
3984 	wrfl();
3985 
3986 	return count;
3987 }
3988 
3989 /**
3990  *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
3991  *  @netdev: network interface device structure
3992  *
3993  *  The set_rx_mode entry point is called whenever the unicast or multicast
3994  *  address lists or the network interface flags are updated.  This routine is
3995  *  responsible for configuring the hardware for proper unicast, multicast,
3996  *  promiscuous mode, and all-multi behavior.
3997  **/
3998 static void igb_set_rx_mode(struct net_device *netdev)
3999 {
4000 	struct igb_adapter *adapter = netdev_priv(netdev);
4001 	struct e1000_hw *hw = &adapter->hw;
4002 	unsigned int vfn = adapter->vfs_allocated_count;
4003 	u32 rctl, vmolr = 0;
4004 	int count;
4005 
4006 	/* Check for Promiscuous and All Multicast modes */
4007 	rctl = rd32(E1000_RCTL);
4008 
4009 	/* clear the effected bits */
4010 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE | E1000_RCTL_VFE);
4011 
4012 	if (netdev->flags & IFF_PROMISC) {
4013 		/* retain VLAN HW filtering if in VT mode */
4014 		if (adapter->vfs_allocated_count)
4015 			rctl |= E1000_RCTL_VFE;
4016 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
4017 		vmolr |= (E1000_VMOLR_ROPE | E1000_VMOLR_MPME);
4018 	} else {
4019 		if (netdev->flags & IFF_ALLMULTI) {
4020 			rctl |= E1000_RCTL_MPE;
4021 			vmolr |= E1000_VMOLR_MPME;
4022 		} else {
4023 			/* Write addresses to the MTA, if the attempt fails
4024 			 * then we should just turn on promiscuous mode so
4025 			 * that we can at least receive multicast traffic
4026 			 */
4027 			count = igb_write_mc_addr_list(netdev);
4028 			if (count < 0) {
4029 				rctl |= E1000_RCTL_MPE;
4030 				vmolr |= E1000_VMOLR_MPME;
4031 			} else if (count) {
4032 				vmolr |= E1000_VMOLR_ROMPE;
4033 			}
4034 		}
4035 		/* Write addresses to available RAR registers, if there is not
4036 		 * sufficient space to store all the addresses then enable
4037 		 * unicast promiscuous mode
4038 		 */
4039 		count = igb_write_uc_addr_list(netdev);
4040 		if (count < 0) {
4041 			rctl |= E1000_RCTL_UPE;
4042 			vmolr |= E1000_VMOLR_ROPE;
4043 		}
4044 		rctl |= E1000_RCTL_VFE;
4045 	}
4046 	wr32(E1000_RCTL, rctl);
4047 
4048 	/* In order to support SR-IOV and eventually VMDq it is necessary to set
4049 	 * the VMOLR to enable the appropriate modes.  Without this workaround
4050 	 * we will have issues with VLAN tag stripping not being done for frames
4051 	 * that are only arriving because we are the default pool
4052 	 */
4053 	if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
4054 		return;
4055 
4056 	vmolr |= rd32(E1000_VMOLR(vfn)) &
4057 		 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
4058 	wr32(E1000_VMOLR(vfn), vmolr);
4059 	igb_restore_vf_multicasts(adapter);
4060 }
4061 
4062 static void igb_check_wvbr(struct igb_adapter *adapter)
4063 {
4064 	struct e1000_hw *hw = &adapter->hw;
4065 	u32 wvbr = 0;
4066 
4067 	switch (hw->mac.type) {
4068 	case e1000_82576:
4069 	case e1000_i350:
4070 		wvbr = rd32(E1000_WVBR);
4071 		if (!wvbr)
4072 			return;
4073 		break;
4074 	default:
4075 		break;
4076 	}
4077 
4078 	adapter->wvbr |= wvbr;
4079 }
4080 
4081 #define IGB_STAGGERED_QUEUE_OFFSET 8
4082 
4083 static void igb_spoof_check(struct igb_adapter *adapter)
4084 {
4085 	int j;
4086 
4087 	if (!adapter->wvbr)
4088 		return;
4089 
4090 	for (j = 0; j < adapter->vfs_allocated_count; j++) {
4091 		if (adapter->wvbr & (1 << j) ||
4092 		    adapter->wvbr & (1 << (j + IGB_STAGGERED_QUEUE_OFFSET))) {
4093 			dev_warn(&adapter->pdev->dev,
4094 				"Spoof event(s) detected on VF %d\n", j);
4095 			adapter->wvbr &=
4096 				~((1 << j) |
4097 				  (1 << (j + IGB_STAGGERED_QUEUE_OFFSET)));
4098 		}
4099 	}
4100 }
4101 
4102 /* Need to wait a few seconds after link up to get diagnostic information from
4103  * the phy
4104  */
4105 static void igb_update_phy_info(unsigned long data)
4106 {
4107 	struct igb_adapter *adapter = (struct igb_adapter *) data;
4108 	igb_get_phy_info(&adapter->hw);
4109 }
4110 
4111 /**
4112  *  igb_has_link - check shared code for link and determine up/down
4113  *  @adapter: pointer to driver private info
4114  **/
4115 bool igb_has_link(struct igb_adapter *adapter)
4116 {
4117 	struct e1000_hw *hw = &adapter->hw;
4118 	bool link_active = false;
4119 
4120 	/* get_link_status is set on LSC (link status) interrupt or
4121 	 * rx sequence error interrupt.  get_link_status will stay
4122 	 * false until the e1000_check_for_link establishes link
4123 	 * for copper adapters ONLY
4124 	 */
4125 	switch (hw->phy.media_type) {
4126 	case e1000_media_type_copper:
4127 		if (!hw->mac.get_link_status)
4128 			return true;
4129 	case e1000_media_type_internal_serdes:
4130 		hw->mac.ops.check_for_link(hw);
4131 		link_active = !hw->mac.get_link_status;
4132 		break;
4133 	default:
4134 	case e1000_media_type_unknown:
4135 		break;
4136 	}
4137 
4138 	if (((hw->mac.type == e1000_i210) ||
4139 	     (hw->mac.type == e1000_i211)) &&
4140 	     (hw->phy.id == I210_I_PHY_ID)) {
4141 		if (!netif_carrier_ok(adapter->netdev)) {
4142 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4143 		} else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
4144 			adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
4145 			adapter->link_check_timeout = jiffies;
4146 		}
4147 	}
4148 
4149 	return link_active;
4150 }
4151 
4152 static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
4153 {
4154 	bool ret = false;
4155 	u32 ctrl_ext, thstat;
4156 
4157 	/* check for thermal sensor event on i350 copper only */
4158 	if (hw->mac.type == e1000_i350) {
4159 		thstat = rd32(E1000_THSTAT);
4160 		ctrl_ext = rd32(E1000_CTRL_EXT);
4161 
4162 		if ((hw->phy.media_type == e1000_media_type_copper) &&
4163 		    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
4164 			ret = !!(thstat & event);
4165 	}
4166 
4167 	return ret;
4168 }
4169 
4170 /**
4171  *  igb_check_lvmmc - check for malformed packets received
4172  *  and indicated in LVMMC register
4173  *  @adapter: pointer to adapter
4174  **/
4175 static void igb_check_lvmmc(struct igb_adapter *adapter)
4176 {
4177 	struct e1000_hw *hw = &adapter->hw;
4178 	u32 lvmmc;
4179 
4180 	lvmmc = rd32(E1000_LVMMC);
4181 	if (lvmmc) {
4182 		if (unlikely(net_ratelimit())) {
4183 			netdev_warn(adapter->netdev,
4184 				    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
4185 				    lvmmc);
4186 		}
4187 	}
4188 }
4189 
4190 /**
4191  *  igb_watchdog - Timer Call-back
4192  *  @data: pointer to adapter cast into an unsigned long
4193  **/
4194 static void igb_watchdog(unsigned long data)
4195 {
4196 	struct igb_adapter *adapter = (struct igb_adapter *)data;
4197 	/* Do the rest outside of interrupt context */
4198 	schedule_work(&adapter->watchdog_task);
4199 }
4200 
4201 static void igb_watchdog_task(struct work_struct *work)
4202 {
4203 	struct igb_adapter *adapter = container_of(work,
4204 						   struct igb_adapter,
4205 						   watchdog_task);
4206 	struct e1000_hw *hw = &adapter->hw;
4207 	struct e1000_phy_info *phy = &hw->phy;
4208 	struct net_device *netdev = adapter->netdev;
4209 	u32 link;
4210 	int i;
4211 	u32 connsw;
4212 
4213 	link = igb_has_link(adapter);
4214 
4215 	if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
4216 		if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
4217 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4218 		else
4219 			link = false;
4220 	}
4221 
4222 	/* Force link down if we have fiber to swap to */
4223 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4224 		if (hw->phy.media_type == e1000_media_type_copper) {
4225 			connsw = rd32(E1000_CONNSW);
4226 			if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
4227 				link = 0;
4228 		}
4229 	}
4230 	if (link) {
4231 		/* Perform a reset if the media type changed. */
4232 		if (hw->dev_spec._82575.media_changed) {
4233 			hw->dev_spec._82575.media_changed = false;
4234 			adapter->flags |= IGB_FLAG_MEDIA_RESET;
4235 			igb_reset(adapter);
4236 		}
4237 		/* Cancel scheduled suspend requests. */
4238 		pm_runtime_resume(netdev->dev.parent);
4239 
4240 		if (!netif_carrier_ok(netdev)) {
4241 			u32 ctrl;
4242 
4243 			hw->mac.ops.get_speed_and_duplex(hw,
4244 							 &adapter->link_speed,
4245 							 &adapter->link_duplex);
4246 
4247 			ctrl = rd32(E1000_CTRL);
4248 			/* Links status message must follow this format */
4249 			netdev_info(netdev,
4250 			       "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4251 			       netdev->name,
4252 			       adapter->link_speed,
4253 			       adapter->link_duplex == FULL_DUPLEX ?
4254 			       "Full" : "Half",
4255 			       (ctrl & E1000_CTRL_TFCE) &&
4256 			       (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
4257 			       (ctrl & E1000_CTRL_RFCE) ?  "RX" :
4258 			       (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
4259 
4260 			/* disable EEE if enabled */
4261 			if ((adapter->flags & IGB_FLAG_EEE) &&
4262 				(adapter->link_duplex == HALF_DUPLEX)) {
4263 				dev_info(&adapter->pdev->dev,
4264 				"EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
4265 				adapter->hw.dev_spec._82575.eee_disable = true;
4266 				adapter->flags &= ~IGB_FLAG_EEE;
4267 			}
4268 
4269 			/* check if SmartSpeed worked */
4270 			igb_check_downshift(hw);
4271 			if (phy->speed_downgraded)
4272 				netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
4273 
4274 			/* check for thermal sensor event */
4275 			if (igb_thermal_sensor_event(hw,
4276 			    E1000_THSTAT_LINK_THROTTLE))
4277 				netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
4278 
4279 			/* adjust timeout factor according to speed/duplex */
4280 			adapter->tx_timeout_factor = 1;
4281 			switch (adapter->link_speed) {
4282 			case SPEED_10:
4283 				adapter->tx_timeout_factor = 14;
4284 				break;
4285 			case SPEED_100:
4286 				/* maybe add some timeout factor ? */
4287 				break;
4288 			}
4289 
4290 			netif_carrier_on(netdev);
4291 
4292 			igb_ping_all_vfs(adapter);
4293 			igb_check_vf_rate_limit(adapter);
4294 
4295 			/* link state has changed, schedule phy info update */
4296 			if (!test_bit(__IGB_DOWN, &adapter->state))
4297 				mod_timer(&adapter->phy_info_timer,
4298 					  round_jiffies(jiffies + 2 * HZ));
4299 		}
4300 	} else {
4301 		if (netif_carrier_ok(netdev)) {
4302 			adapter->link_speed = 0;
4303 			adapter->link_duplex = 0;
4304 
4305 			/* check for thermal sensor event */
4306 			if (igb_thermal_sensor_event(hw,
4307 			    E1000_THSTAT_PWR_DOWN)) {
4308 				netdev_err(netdev, "The network adapter was stopped because it overheated\n");
4309 			}
4310 
4311 			/* Links status message must follow this format */
4312 			netdev_info(netdev, "igb: %s NIC Link is Down\n",
4313 			       netdev->name);
4314 			netif_carrier_off(netdev);
4315 
4316 			igb_ping_all_vfs(adapter);
4317 
4318 			/* link state has changed, schedule phy info update */
4319 			if (!test_bit(__IGB_DOWN, &adapter->state))
4320 				mod_timer(&adapter->phy_info_timer,
4321 					  round_jiffies(jiffies + 2 * HZ));
4322 
4323 			/* link is down, time to check for alternate media */
4324 			if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4325 				igb_check_swap_media(adapter);
4326 				if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
4327 					schedule_work(&adapter->reset_task);
4328 					/* return immediately */
4329 					return;
4330 				}
4331 			}
4332 			pm_schedule_suspend(netdev->dev.parent,
4333 					    MSEC_PER_SEC * 5);
4334 
4335 		/* also check for alternate media here */
4336 		} else if (!netif_carrier_ok(netdev) &&
4337 			   (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
4338 			igb_check_swap_media(adapter);
4339 			if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
4340 				schedule_work(&adapter->reset_task);
4341 				/* return immediately */
4342 				return;
4343 			}
4344 		}
4345 	}
4346 
4347 	spin_lock(&adapter->stats64_lock);
4348 	igb_update_stats(adapter, &adapter->stats64);
4349 	spin_unlock(&adapter->stats64_lock);
4350 
4351 	for (i = 0; i < adapter->num_tx_queues; i++) {
4352 		struct igb_ring *tx_ring = adapter->tx_ring[i];
4353 		if (!netif_carrier_ok(netdev)) {
4354 			/* We've lost link, so the controller stops DMA,
4355 			 * but we've got queued Tx work that's never going
4356 			 * to get done, so reset controller to flush Tx.
4357 			 * (Do the reset outside of interrupt context).
4358 			 */
4359 			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
4360 				adapter->tx_timeout_count++;
4361 				schedule_work(&adapter->reset_task);
4362 				/* return immediately since reset is imminent */
4363 				return;
4364 			}
4365 		}
4366 
4367 		/* Force detection of hung controller every watchdog period */
4368 		set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
4369 	}
4370 
4371 	/* Cause software interrupt to ensure Rx ring is cleaned */
4372 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
4373 		u32 eics = 0;
4374 
4375 		for (i = 0; i < adapter->num_q_vectors; i++)
4376 			eics |= adapter->q_vector[i]->eims_value;
4377 		wr32(E1000_EICS, eics);
4378 	} else {
4379 		wr32(E1000_ICS, E1000_ICS_RXDMT0);
4380 	}
4381 
4382 	igb_spoof_check(adapter);
4383 	igb_ptp_rx_hang(adapter);
4384 
4385 	/* Check LVMMC register on i350/i354 only */
4386 	if ((adapter->hw.mac.type == e1000_i350) ||
4387 	    (adapter->hw.mac.type == e1000_i354))
4388 		igb_check_lvmmc(adapter);
4389 
4390 	/* Reset the timer */
4391 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
4392 		if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
4393 			mod_timer(&adapter->watchdog_timer,
4394 				  round_jiffies(jiffies +  HZ));
4395 		else
4396 			mod_timer(&adapter->watchdog_timer,
4397 				  round_jiffies(jiffies + 2 * HZ));
4398 	}
4399 }
4400 
4401 enum latency_range {
4402 	lowest_latency = 0,
4403 	low_latency = 1,
4404 	bulk_latency = 2,
4405 	latency_invalid = 255
4406 };
4407 
4408 /**
4409  *  igb_update_ring_itr - update the dynamic ITR value based on packet size
4410  *  @q_vector: pointer to q_vector
4411  *
4412  *  Stores a new ITR value based on strictly on packet size.  This
4413  *  algorithm is less sophisticated than that used in igb_update_itr,
4414  *  due to the difficulty of synchronizing statistics across multiple
4415  *  receive rings.  The divisors and thresholds used by this function
4416  *  were determined based on theoretical maximum wire speed and testing
4417  *  data, in order to minimize response time while increasing bulk
4418  *  throughput.
4419  *  This functionality is controlled by ethtool's coalescing settings.
4420  *  NOTE:  This function is called only when operating in a multiqueue
4421  *         receive environment.
4422  **/
4423 static void igb_update_ring_itr(struct igb_q_vector *q_vector)
4424 {
4425 	int new_val = q_vector->itr_val;
4426 	int avg_wire_size = 0;
4427 	struct igb_adapter *adapter = q_vector->adapter;
4428 	unsigned int packets;
4429 
4430 	/* For non-gigabit speeds, just fix the interrupt rate at 4000
4431 	 * ints/sec - ITR timer value of 120 ticks.
4432 	 */
4433 	if (adapter->link_speed != SPEED_1000) {
4434 		new_val = IGB_4K_ITR;
4435 		goto set_itr_val;
4436 	}
4437 
4438 	packets = q_vector->rx.total_packets;
4439 	if (packets)
4440 		avg_wire_size = q_vector->rx.total_bytes / packets;
4441 
4442 	packets = q_vector->tx.total_packets;
4443 	if (packets)
4444 		avg_wire_size = max_t(u32, avg_wire_size,
4445 				      q_vector->tx.total_bytes / packets);
4446 
4447 	/* if avg_wire_size isn't set no work was done */
4448 	if (!avg_wire_size)
4449 		goto clear_counts;
4450 
4451 	/* Add 24 bytes to size to account for CRC, preamble, and gap */
4452 	avg_wire_size += 24;
4453 
4454 	/* Don't starve jumbo frames */
4455 	avg_wire_size = min(avg_wire_size, 3000);
4456 
4457 	/* Give a little boost to mid-size frames */
4458 	if ((avg_wire_size > 300) && (avg_wire_size < 1200))
4459 		new_val = avg_wire_size / 3;
4460 	else
4461 		new_val = avg_wire_size / 2;
4462 
4463 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
4464 	if (new_val < IGB_20K_ITR &&
4465 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
4466 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
4467 		new_val = IGB_20K_ITR;
4468 
4469 set_itr_val:
4470 	if (new_val != q_vector->itr_val) {
4471 		q_vector->itr_val = new_val;
4472 		q_vector->set_itr = 1;
4473 	}
4474 clear_counts:
4475 	q_vector->rx.total_bytes = 0;
4476 	q_vector->rx.total_packets = 0;
4477 	q_vector->tx.total_bytes = 0;
4478 	q_vector->tx.total_packets = 0;
4479 }
4480 
4481 /**
4482  *  igb_update_itr - update the dynamic ITR value based on statistics
4483  *  @q_vector: pointer to q_vector
4484  *  @ring_container: ring info to update the itr for
4485  *
4486  *  Stores a new ITR value based on packets and byte
4487  *  counts during the last interrupt.  The advantage of per interrupt
4488  *  computation is faster updates and more accurate ITR for the current
4489  *  traffic pattern.  Constants in this function were computed
4490  *  based on theoretical maximum wire speed and thresholds were set based
4491  *  on testing data as well as attempting to minimize response time
4492  *  while increasing bulk throughput.
4493  *  This functionality is controlled by ethtool's coalescing settings.
4494  *  NOTE:  These calculations are only valid when operating in a single-
4495  *         queue environment.
4496  **/
4497 static void igb_update_itr(struct igb_q_vector *q_vector,
4498 			   struct igb_ring_container *ring_container)
4499 {
4500 	unsigned int packets = ring_container->total_packets;
4501 	unsigned int bytes = ring_container->total_bytes;
4502 	u8 itrval = ring_container->itr;
4503 
4504 	/* no packets, exit with status unchanged */
4505 	if (packets == 0)
4506 		return;
4507 
4508 	switch (itrval) {
4509 	case lowest_latency:
4510 		/* handle TSO and jumbo frames */
4511 		if (bytes/packets > 8000)
4512 			itrval = bulk_latency;
4513 		else if ((packets < 5) && (bytes > 512))
4514 			itrval = low_latency;
4515 		break;
4516 	case low_latency:  /* 50 usec aka 20000 ints/s */
4517 		if (bytes > 10000) {
4518 			/* this if handles the TSO accounting */
4519 			if (bytes/packets > 8000)
4520 				itrval = bulk_latency;
4521 			else if ((packets < 10) || ((bytes/packets) > 1200))
4522 				itrval = bulk_latency;
4523 			else if ((packets > 35))
4524 				itrval = lowest_latency;
4525 		} else if (bytes/packets > 2000) {
4526 			itrval = bulk_latency;
4527 		} else if (packets <= 2 && bytes < 512) {
4528 			itrval = lowest_latency;
4529 		}
4530 		break;
4531 	case bulk_latency: /* 250 usec aka 4000 ints/s */
4532 		if (bytes > 25000) {
4533 			if (packets > 35)
4534 				itrval = low_latency;
4535 		} else if (bytes < 1500) {
4536 			itrval = low_latency;
4537 		}
4538 		break;
4539 	}
4540 
4541 	/* clear work counters since we have the values we need */
4542 	ring_container->total_bytes = 0;
4543 	ring_container->total_packets = 0;
4544 
4545 	/* write updated itr to ring container */
4546 	ring_container->itr = itrval;
4547 }
4548 
4549 static void igb_set_itr(struct igb_q_vector *q_vector)
4550 {
4551 	struct igb_adapter *adapter = q_vector->adapter;
4552 	u32 new_itr = q_vector->itr_val;
4553 	u8 current_itr = 0;
4554 
4555 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
4556 	if (adapter->link_speed != SPEED_1000) {
4557 		current_itr = 0;
4558 		new_itr = IGB_4K_ITR;
4559 		goto set_itr_now;
4560 	}
4561 
4562 	igb_update_itr(q_vector, &q_vector->tx);
4563 	igb_update_itr(q_vector, &q_vector->rx);
4564 
4565 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
4566 
4567 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
4568 	if (current_itr == lowest_latency &&
4569 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
4570 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
4571 		current_itr = low_latency;
4572 
4573 	switch (current_itr) {
4574 	/* counts and packets in update_itr are dependent on these numbers */
4575 	case lowest_latency:
4576 		new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
4577 		break;
4578 	case low_latency:
4579 		new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
4580 		break;
4581 	case bulk_latency:
4582 		new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
4583 		break;
4584 	default:
4585 		break;
4586 	}
4587 
4588 set_itr_now:
4589 	if (new_itr != q_vector->itr_val) {
4590 		/* this attempts to bias the interrupt rate towards Bulk
4591 		 * by adding intermediate steps when interrupt rate is
4592 		 * increasing
4593 		 */
4594 		new_itr = new_itr > q_vector->itr_val ?
4595 			  max((new_itr * q_vector->itr_val) /
4596 			  (new_itr + (q_vector->itr_val >> 2)),
4597 			  new_itr) : new_itr;
4598 		/* Don't write the value here; it resets the adapter's
4599 		 * internal timer, and causes us to delay far longer than
4600 		 * we should between interrupts.  Instead, we write the ITR
4601 		 * value at the beginning of the next interrupt so the timing
4602 		 * ends up being correct.
4603 		 */
4604 		q_vector->itr_val = new_itr;
4605 		q_vector->set_itr = 1;
4606 	}
4607 }
4608 
4609 static void igb_tx_ctxtdesc(struct igb_ring *tx_ring, u32 vlan_macip_lens,
4610 			    u32 type_tucmd, u32 mss_l4len_idx)
4611 {
4612 	struct e1000_adv_tx_context_desc *context_desc;
4613 	u16 i = tx_ring->next_to_use;
4614 
4615 	context_desc = IGB_TX_CTXTDESC(tx_ring, i);
4616 
4617 	i++;
4618 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
4619 
4620 	/* set bits to identify this as an advanced context descriptor */
4621 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
4622 
4623 	/* For 82575, context index must be unique per ring. */
4624 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
4625 		mss_l4len_idx |= tx_ring->reg_idx << 4;
4626 
4627 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
4628 	context_desc->seqnum_seed	= 0;
4629 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
4630 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
4631 }
4632 
4633 static int igb_tso(struct igb_ring *tx_ring,
4634 		   struct igb_tx_buffer *first,
4635 		   u8 *hdr_len)
4636 {
4637 	struct sk_buff *skb = first->skb;
4638 	u32 vlan_macip_lens, type_tucmd;
4639 	u32 mss_l4len_idx, l4len;
4640 	int err;
4641 
4642 	if (skb->ip_summed != CHECKSUM_PARTIAL)
4643 		return 0;
4644 
4645 	if (!skb_is_gso(skb))
4646 		return 0;
4647 
4648 	err = skb_cow_head(skb, 0);
4649 	if (err < 0)
4650 		return err;
4651 
4652 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
4653 	type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
4654 
4655 	if (first->protocol == htons(ETH_P_IP)) {
4656 		struct iphdr *iph = ip_hdr(skb);
4657 		iph->tot_len = 0;
4658 		iph->check = 0;
4659 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
4660 							 iph->daddr, 0,
4661 							 IPPROTO_TCP,
4662 							 0);
4663 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
4664 		first->tx_flags |= IGB_TX_FLAGS_TSO |
4665 				   IGB_TX_FLAGS_CSUM |
4666 				   IGB_TX_FLAGS_IPV4;
4667 	} else if (skb_is_gso_v6(skb)) {
4668 		ipv6_hdr(skb)->payload_len = 0;
4669 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
4670 						       &ipv6_hdr(skb)->daddr,
4671 						       0, IPPROTO_TCP, 0);
4672 		first->tx_flags |= IGB_TX_FLAGS_TSO |
4673 				   IGB_TX_FLAGS_CSUM;
4674 	}
4675 
4676 	/* compute header lengths */
4677 	l4len = tcp_hdrlen(skb);
4678 	*hdr_len = skb_transport_offset(skb) + l4len;
4679 
4680 	/* update gso size and bytecount with header size */
4681 	first->gso_segs = skb_shinfo(skb)->gso_segs;
4682 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
4683 
4684 	/* MSS L4LEN IDX */
4685 	mss_l4len_idx = l4len << E1000_ADVTXD_L4LEN_SHIFT;
4686 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
4687 
4688 	/* VLAN MACLEN IPLEN */
4689 	vlan_macip_lens = skb_network_header_len(skb);
4690 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
4691 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
4692 
4693 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
4694 
4695 	return 1;
4696 }
4697 
4698 static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
4699 {
4700 	struct sk_buff *skb = first->skb;
4701 	u32 vlan_macip_lens = 0;
4702 	u32 mss_l4len_idx = 0;
4703 	u32 type_tucmd = 0;
4704 
4705 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
4706 		if (!(first->tx_flags & IGB_TX_FLAGS_VLAN))
4707 			return;
4708 	} else {
4709 		u8 l4_hdr = 0;
4710 
4711 		switch (first->protocol) {
4712 		case htons(ETH_P_IP):
4713 			vlan_macip_lens |= skb_network_header_len(skb);
4714 			type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
4715 			l4_hdr = ip_hdr(skb)->protocol;
4716 			break;
4717 		case htons(ETH_P_IPV6):
4718 			vlan_macip_lens |= skb_network_header_len(skb);
4719 			l4_hdr = ipv6_hdr(skb)->nexthdr;
4720 			break;
4721 		default:
4722 			if (unlikely(net_ratelimit())) {
4723 				dev_warn(tx_ring->dev,
4724 					 "partial checksum but proto=%x!\n",
4725 					 first->protocol);
4726 			}
4727 			break;
4728 		}
4729 
4730 		switch (l4_hdr) {
4731 		case IPPROTO_TCP:
4732 			type_tucmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
4733 			mss_l4len_idx = tcp_hdrlen(skb) <<
4734 					E1000_ADVTXD_L4LEN_SHIFT;
4735 			break;
4736 		case IPPROTO_SCTP:
4737 			type_tucmd |= E1000_ADVTXD_TUCMD_L4T_SCTP;
4738 			mss_l4len_idx = sizeof(struct sctphdr) <<
4739 					E1000_ADVTXD_L4LEN_SHIFT;
4740 			break;
4741 		case IPPROTO_UDP:
4742 			mss_l4len_idx = sizeof(struct udphdr) <<
4743 					E1000_ADVTXD_L4LEN_SHIFT;
4744 			break;
4745 		default:
4746 			if (unlikely(net_ratelimit())) {
4747 				dev_warn(tx_ring->dev,
4748 					 "partial checksum but l4 proto=%x!\n",
4749 					 l4_hdr);
4750 			}
4751 			break;
4752 		}
4753 
4754 		/* update TX checksum flag */
4755 		first->tx_flags |= IGB_TX_FLAGS_CSUM;
4756 	}
4757 
4758 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
4759 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
4760 
4761 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
4762 }
4763 
4764 #define IGB_SET_FLAG(_input, _flag, _result) \
4765 	((_flag <= _result) ? \
4766 	 ((u32)(_input & _flag) * (_result / _flag)) : \
4767 	 ((u32)(_input & _flag) / (_flag / _result)))
4768 
4769 static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
4770 {
4771 	/* set type for advanced descriptor with frame checksum insertion */
4772 	u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
4773 		       E1000_ADVTXD_DCMD_DEXT |
4774 		       E1000_ADVTXD_DCMD_IFCS;
4775 
4776 	/* set HW vlan bit if vlan is present */
4777 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
4778 				 (E1000_ADVTXD_DCMD_VLE));
4779 
4780 	/* set segmentation bits for TSO */
4781 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
4782 				 (E1000_ADVTXD_DCMD_TSE));
4783 
4784 	/* set timestamp bit if present */
4785 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
4786 				 (E1000_ADVTXD_MAC_TSTAMP));
4787 
4788 	/* insert frame checksum */
4789 	cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
4790 
4791 	return cmd_type;
4792 }
4793 
4794 static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
4795 				 union e1000_adv_tx_desc *tx_desc,
4796 				 u32 tx_flags, unsigned int paylen)
4797 {
4798 	u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
4799 
4800 	/* 82575 requires a unique index per ring */
4801 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
4802 		olinfo_status |= tx_ring->reg_idx << 4;
4803 
4804 	/* insert L4 checksum */
4805 	olinfo_status |= IGB_SET_FLAG(tx_flags,
4806 				      IGB_TX_FLAGS_CSUM,
4807 				      (E1000_TXD_POPTS_TXSM << 8));
4808 
4809 	/* insert IPv4 checksum */
4810 	olinfo_status |= IGB_SET_FLAG(tx_flags,
4811 				      IGB_TX_FLAGS_IPV4,
4812 				      (E1000_TXD_POPTS_IXSM << 8));
4813 
4814 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
4815 }
4816 
4817 static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
4818 {
4819 	struct net_device *netdev = tx_ring->netdev;
4820 
4821 	netif_stop_subqueue(netdev, tx_ring->queue_index);
4822 
4823 	/* Herbert's original patch had:
4824 	 *  smp_mb__after_netif_stop_queue();
4825 	 * but since that doesn't exist yet, just open code it.
4826 	 */
4827 	smp_mb();
4828 
4829 	/* We need to check again in a case another CPU has just
4830 	 * made room available.
4831 	 */
4832 	if (igb_desc_unused(tx_ring) < size)
4833 		return -EBUSY;
4834 
4835 	/* A reprieve! */
4836 	netif_wake_subqueue(netdev, tx_ring->queue_index);
4837 
4838 	u64_stats_update_begin(&tx_ring->tx_syncp2);
4839 	tx_ring->tx_stats.restart_queue2++;
4840 	u64_stats_update_end(&tx_ring->tx_syncp2);
4841 
4842 	return 0;
4843 }
4844 
4845 static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
4846 {
4847 	if (igb_desc_unused(tx_ring) >= size)
4848 		return 0;
4849 	return __igb_maybe_stop_tx(tx_ring, size);
4850 }
4851 
4852 static void igb_tx_map(struct igb_ring *tx_ring,
4853 		       struct igb_tx_buffer *first,
4854 		       const u8 hdr_len)
4855 {
4856 	struct sk_buff *skb = first->skb;
4857 	struct igb_tx_buffer *tx_buffer;
4858 	union e1000_adv_tx_desc *tx_desc;
4859 	struct skb_frag_struct *frag;
4860 	dma_addr_t dma;
4861 	unsigned int data_len, size;
4862 	u32 tx_flags = first->tx_flags;
4863 	u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
4864 	u16 i = tx_ring->next_to_use;
4865 
4866 	tx_desc = IGB_TX_DESC(tx_ring, i);
4867 
4868 	igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
4869 
4870 	size = skb_headlen(skb);
4871 	data_len = skb->data_len;
4872 
4873 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
4874 
4875 	tx_buffer = first;
4876 
4877 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
4878 		if (dma_mapping_error(tx_ring->dev, dma))
4879 			goto dma_error;
4880 
4881 		/* record length, and DMA address */
4882 		dma_unmap_len_set(tx_buffer, len, size);
4883 		dma_unmap_addr_set(tx_buffer, dma, dma);
4884 
4885 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
4886 
4887 		while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
4888 			tx_desc->read.cmd_type_len =
4889 				cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
4890 
4891 			i++;
4892 			tx_desc++;
4893 			if (i == tx_ring->count) {
4894 				tx_desc = IGB_TX_DESC(tx_ring, 0);
4895 				i = 0;
4896 			}
4897 			tx_desc->read.olinfo_status = 0;
4898 
4899 			dma += IGB_MAX_DATA_PER_TXD;
4900 			size -= IGB_MAX_DATA_PER_TXD;
4901 
4902 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
4903 		}
4904 
4905 		if (likely(!data_len))
4906 			break;
4907 
4908 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
4909 
4910 		i++;
4911 		tx_desc++;
4912 		if (i == tx_ring->count) {
4913 			tx_desc = IGB_TX_DESC(tx_ring, 0);
4914 			i = 0;
4915 		}
4916 		tx_desc->read.olinfo_status = 0;
4917 
4918 		size = skb_frag_size(frag);
4919 		data_len -= size;
4920 
4921 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
4922 				       size, DMA_TO_DEVICE);
4923 
4924 		tx_buffer = &tx_ring->tx_buffer_info[i];
4925 	}
4926 
4927 	/* write last descriptor with RS and EOP bits */
4928 	cmd_type |= size | IGB_TXD_DCMD;
4929 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
4930 
4931 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
4932 
4933 	/* set the timestamp */
4934 	first->time_stamp = jiffies;
4935 
4936 	/* Force memory writes to complete before letting h/w know there
4937 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
4938 	 * memory model archs, such as IA-64).
4939 	 *
4940 	 * We also need this memory barrier to make certain all of the
4941 	 * status bits have been updated before next_to_watch is written.
4942 	 */
4943 	wmb();
4944 
4945 	/* set next_to_watch value indicating a packet is present */
4946 	first->next_to_watch = tx_desc;
4947 
4948 	i++;
4949 	if (i == tx_ring->count)
4950 		i = 0;
4951 
4952 	tx_ring->next_to_use = i;
4953 
4954 	/* Make sure there is space in the ring for the next send. */
4955 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
4956 
4957 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
4958 		writel(i, tx_ring->tail);
4959 
4960 		/* we need this if more than one processor can write to our tail
4961 		 * at a time, it synchronizes IO on IA64/Altix systems
4962 		 */
4963 		mmiowb();
4964 	}
4965 	return;
4966 
4967 dma_error:
4968 	dev_err(tx_ring->dev, "TX DMA map failed\n");
4969 
4970 	/* clear dma mappings for failed tx_buffer_info map */
4971 	for (;;) {
4972 		tx_buffer = &tx_ring->tx_buffer_info[i];
4973 		igb_unmap_and_free_tx_resource(tx_ring, tx_buffer);
4974 		if (tx_buffer == first)
4975 			break;
4976 		if (i == 0)
4977 			i = tx_ring->count;
4978 		i--;
4979 	}
4980 
4981 	tx_ring->next_to_use = i;
4982 }
4983 
4984 netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
4985 				struct igb_ring *tx_ring)
4986 {
4987 	struct igb_tx_buffer *first;
4988 	int tso;
4989 	u32 tx_flags = 0;
4990 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
4991 	__be16 protocol = vlan_get_protocol(skb);
4992 	u8 hdr_len = 0;
4993 
4994 	/* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
4995 	 *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
4996 	 *       + 2 desc gap to keep tail from touching head,
4997 	 *       + 1 desc for context descriptor,
4998 	 * otherwise try next time
4999 	 */
5000 	if (NETDEV_FRAG_PAGE_MAX_SIZE > IGB_MAX_DATA_PER_TXD) {
5001 		unsigned short f;
5002 
5003 		for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
5004 			count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
5005 	} else {
5006 		count += skb_shinfo(skb)->nr_frags;
5007 	}
5008 
5009 	if (igb_maybe_stop_tx(tx_ring, count + 3)) {
5010 		/* this is a hard error */
5011 		return NETDEV_TX_BUSY;
5012 	}
5013 
5014 	/* record the location of the first descriptor for this packet */
5015 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
5016 	first->skb = skb;
5017 	first->bytecount = skb->len;
5018 	first->gso_segs = 1;
5019 
5020 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
5021 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
5022 
5023 		if (!test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
5024 					   &adapter->state)) {
5025 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5026 			tx_flags |= IGB_TX_FLAGS_TSTAMP;
5027 
5028 			adapter->ptp_tx_skb = skb_get(skb);
5029 			adapter->ptp_tx_start = jiffies;
5030 			if (adapter->hw.mac.type == e1000_82576)
5031 				schedule_work(&adapter->ptp_tx_work);
5032 		}
5033 	}
5034 
5035 	skb_tx_timestamp(skb);
5036 
5037 	if (skb_vlan_tag_present(skb)) {
5038 		tx_flags |= IGB_TX_FLAGS_VLAN;
5039 		tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
5040 	}
5041 
5042 	/* record initial flags and protocol */
5043 	first->tx_flags = tx_flags;
5044 	first->protocol = protocol;
5045 
5046 	tso = igb_tso(tx_ring, first, &hdr_len);
5047 	if (tso < 0)
5048 		goto out_drop;
5049 	else if (!tso)
5050 		igb_tx_csum(tx_ring, first);
5051 
5052 	igb_tx_map(tx_ring, first, hdr_len);
5053 
5054 	return NETDEV_TX_OK;
5055 
5056 out_drop:
5057 	igb_unmap_and_free_tx_resource(tx_ring, first);
5058 
5059 	return NETDEV_TX_OK;
5060 }
5061 
5062 static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
5063 						    struct sk_buff *skb)
5064 {
5065 	unsigned int r_idx = skb->queue_mapping;
5066 
5067 	if (r_idx >= adapter->num_tx_queues)
5068 		r_idx = r_idx % adapter->num_tx_queues;
5069 
5070 	return adapter->tx_ring[r_idx];
5071 }
5072 
5073 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
5074 				  struct net_device *netdev)
5075 {
5076 	struct igb_adapter *adapter = netdev_priv(netdev);
5077 
5078 	if (test_bit(__IGB_DOWN, &adapter->state)) {
5079 		dev_kfree_skb_any(skb);
5080 		return NETDEV_TX_OK;
5081 	}
5082 
5083 	if (skb->len <= 0) {
5084 		dev_kfree_skb_any(skb);
5085 		return NETDEV_TX_OK;
5086 	}
5087 
5088 	/* The minimum packet size with TCTL.PSP set is 17 so pad the skb
5089 	 * in order to meet this minimum size requirement.
5090 	 */
5091 	if (skb_put_padto(skb, 17))
5092 		return NETDEV_TX_OK;
5093 
5094 	return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
5095 }
5096 
5097 /**
5098  *  igb_tx_timeout - Respond to a Tx Hang
5099  *  @netdev: network interface device structure
5100  **/
5101 static void igb_tx_timeout(struct net_device *netdev)
5102 {
5103 	struct igb_adapter *adapter = netdev_priv(netdev);
5104 	struct e1000_hw *hw = &adapter->hw;
5105 
5106 	/* Do the reset outside of interrupt context */
5107 	adapter->tx_timeout_count++;
5108 
5109 	if (hw->mac.type >= e1000_82580)
5110 		hw->dev_spec._82575.global_device_reset = true;
5111 
5112 	schedule_work(&adapter->reset_task);
5113 	wr32(E1000_EICS,
5114 	     (adapter->eims_enable_mask & ~adapter->eims_other));
5115 }
5116 
5117 static void igb_reset_task(struct work_struct *work)
5118 {
5119 	struct igb_adapter *adapter;
5120 	adapter = container_of(work, struct igb_adapter, reset_task);
5121 
5122 	igb_dump(adapter);
5123 	netdev_err(adapter->netdev, "Reset adapter\n");
5124 	igb_reinit_locked(adapter);
5125 }
5126 
5127 /**
5128  *  igb_get_stats64 - Get System Network Statistics
5129  *  @netdev: network interface device structure
5130  *  @stats: rtnl_link_stats64 pointer
5131  **/
5132 static struct rtnl_link_stats64 *igb_get_stats64(struct net_device *netdev,
5133 						struct rtnl_link_stats64 *stats)
5134 {
5135 	struct igb_adapter *adapter = netdev_priv(netdev);
5136 
5137 	spin_lock(&adapter->stats64_lock);
5138 	igb_update_stats(adapter, &adapter->stats64);
5139 	memcpy(stats, &adapter->stats64, sizeof(*stats));
5140 	spin_unlock(&adapter->stats64_lock);
5141 
5142 	return stats;
5143 }
5144 
5145 /**
5146  *  igb_change_mtu - Change the Maximum Transfer Unit
5147  *  @netdev: network interface device structure
5148  *  @new_mtu: new value for maximum frame size
5149  *
5150  *  Returns 0 on success, negative on failure
5151  **/
5152 static int igb_change_mtu(struct net_device *netdev, int new_mtu)
5153 {
5154 	struct igb_adapter *adapter = netdev_priv(netdev);
5155 	struct pci_dev *pdev = adapter->pdev;
5156 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
5157 
5158 	if ((new_mtu < 68) || (max_frame > MAX_JUMBO_FRAME_SIZE)) {
5159 		dev_err(&pdev->dev, "Invalid MTU setting\n");
5160 		return -EINVAL;
5161 	}
5162 
5163 #define MAX_STD_JUMBO_FRAME_SIZE 9238
5164 	if (max_frame > MAX_STD_JUMBO_FRAME_SIZE) {
5165 		dev_err(&pdev->dev, "MTU > 9216 not supported.\n");
5166 		return -EINVAL;
5167 	}
5168 
5169 	/* adjust max frame to be at least the size of a standard frame */
5170 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
5171 		max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
5172 
5173 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
5174 		usleep_range(1000, 2000);
5175 
5176 	/* igb_down has a dependency on max_frame_size */
5177 	adapter->max_frame_size = max_frame;
5178 
5179 	if (netif_running(netdev))
5180 		igb_down(adapter);
5181 
5182 	dev_info(&pdev->dev, "changing MTU from %d to %d\n",
5183 		 netdev->mtu, new_mtu);
5184 	netdev->mtu = new_mtu;
5185 
5186 	if (netif_running(netdev))
5187 		igb_up(adapter);
5188 	else
5189 		igb_reset(adapter);
5190 
5191 	clear_bit(__IGB_RESETTING, &adapter->state);
5192 
5193 	return 0;
5194 }
5195 
5196 /**
5197  *  igb_update_stats - Update the board statistics counters
5198  *  @adapter: board private structure
5199  **/
5200 void igb_update_stats(struct igb_adapter *adapter,
5201 		      struct rtnl_link_stats64 *net_stats)
5202 {
5203 	struct e1000_hw *hw = &adapter->hw;
5204 	struct pci_dev *pdev = adapter->pdev;
5205 	u32 reg, mpc;
5206 	int i;
5207 	u64 bytes, packets;
5208 	unsigned int start;
5209 	u64 _bytes, _packets;
5210 
5211 	/* Prevent stats update while adapter is being reset, or if the pci
5212 	 * connection is down.
5213 	 */
5214 	if (adapter->link_speed == 0)
5215 		return;
5216 	if (pci_channel_offline(pdev))
5217 		return;
5218 
5219 	bytes = 0;
5220 	packets = 0;
5221 
5222 	rcu_read_lock();
5223 	for (i = 0; i < adapter->num_rx_queues; i++) {
5224 		struct igb_ring *ring = adapter->rx_ring[i];
5225 		u32 rqdpc = rd32(E1000_RQDPC(i));
5226 		if (hw->mac.type >= e1000_i210)
5227 			wr32(E1000_RQDPC(i), 0);
5228 
5229 		if (rqdpc) {
5230 			ring->rx_stats.drops += rqdpc;
5231 			net_stats->rx_fifo_errors += rqdpc;
5232 		}
5233 
5234 		do {
5235 			start = u64_stats_fetch_begin_irq(&ring->rx_syncp);
5236 			_bytes = ring->rx_stats.bytes;
5237 			_packets = ring->rx_stats.packets;
5238 		} while (u64_stats_fetch_retry_irq(&ring->rx_syncp, start));
5239 		bytes += _bytes;
5240 		packets += _packets;
5241 	}
5242 
5243 	net_stats->rx_bytes = bytes;
5244 	net_stats->rx_packets = packets;
5245 
5246 	bytes = 0;
5247 	packets = 0;
5248 	for (i = 0; i < adapter->num_tx_queues; i++) {
5249 		struct igb_ring *ring = adapter->tx_ring[i];
5250 		do {
5251 			start = u64_stats_fetch_begin_irq(&ring->tx_syncp);
5252 			_bytes = ring->tx_stats.bytes;
5253 			_packets = ring->tx_stats.packets;
5254 		} while (u64_stats_fetch_retry_irq(&ring->tx_syncp, start));
5255 		bytes += _bytes;
5256 		packets += _packets;
5257 	}
5258 	net_stats->tx_bytes = bytes;
5259 	net_stats->tx_packets = packets;
5260 	rcu_read_unlock();
5261 
5262 	/* read stats registers */
5263 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
5264 	adapter->stats.gprc += rd32(E1000_GPRC);
5265 	adapter->stats.gorc += rd32(E1000_GORCL);
5266 	rd32(E1000_GORCH); /* clear GORCL */
5267 	adapter->stats.bprc += rd32(E1000_BPRC);
5268 	adapter->stats.mprc += rd32(E1000_MPRC);
5269 	adapter->stats.roc += rd32(E1000_ROC);
5270 
5271 	adapter->stats.prc64 += rd32(E1000_PRC64);
5272 	adapter->stats.prc127 += rd32(E1000_PRC127);
5273 	adapter->stats.prc255 += rd32(E1000_PRC255);
5274 	adapter->stats.prc511 += rd32(E1000_PRC511);
5275 	adapter->stats.prc1023 += rd32(E1000_PRC1023);
5276 	adapter->stats.prc1522 += rd32(E1000_PRC1522);
5277 	adapter->stats.symerrs += rd32(E1000_SYMERRS);
5278 	adapter->stats.sec += rd32(E1000_SEC);
5279 
5280 	mpc = rd32(E1000_MPC);
5281 	adapter->stats.mpc += mpc;
5282 	net_stats->rx_fifo_errors += mpc;
5283 	adapter->stats.scc += rd32(E1000_SCC);
5284 	adapter->stats.ecol += rd32(E1000_ECOL);
5285 	adapter->stats.mcc += rd32(E1000_MCC);
5286 	adapter->stats.latecol += rd32(E1000_LATECOL);
5287 	adapter->stats.dc += rd32(E1000_DC);
5288 	adapter->stats.rlec += rd32(E1000_RLEC);
5289 	adapter->stats.xonrxc += rd32(E1000_XONRXC);
5290 	adapter->stats.xontxc += rd32(E1000_XONTXC);
5291 	adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
5292 	adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
5293 	adapter->stats.fcruc += rd32(E1000_FCRUC);
5294 	adapter->stats.gptc += rd32(E1000_GPTC);
5295 	adapter->stats.gotc += rd32(E1000_GOTCL);
5296 	rd32(E1000_GOTCH); /* clear GOTCL */
5297 	adapter->stats.rnbc += rd32(E1000_RNBC);
5298 	adapter->stats.ruc += rd32(E1000_RUC);
5299 	adapter->stats.rfc += rd32(E1000_RFC);
5300 	adapter->stats.rjc += rd32(E1000_RJC);
5301 	adapter->stats.tor += rd32(E1000_TORH);
5302 	adapter->stats.tot += rd32(E1000_TOTH);
5303 	adapter->stats.tpr += rd32(E1000_TPR);
5304 
5305 	adapter->stats.ptc64 += rd32(E1000_PTC64);
5306 	adapter->stats.ptc127 += rd32(E1000_PTC127);
5307 	adapter->stats.ptc255 += rd32(E1000_PTC255);
5308 	adapter->stats.ptc511 += rd32(E1000_PTC511);
5309 	adapter->stats.ptc1023 += rd32(E1000_PTC1023);
5310 	adapter->stats.ptc1522 += rd32(E1000_PTC1522);
5311 
5312 	adapter->stats.mptc += rd32(E1000_MPTC);
5313 	adapter->stats.bptc += rd32(E1000_BPTC);
5314 
5315 	adapter->stats.tpt += rd32(E1000_TPT);
5316 	adapter->stats.colc += rd32(E1000_COLC);
5317 
5318 	adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
5319 	/* read internal phy specific stats */
5320 	reg = rd32(E1000_CTRL_EXT);
5321 	if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
5322 		adapter->stats.rxerrc += rd32(E1000_RXERRC);
5323 
5324 		/* this stat has invalid values on i210/i211 */
5325 		if ((hw->mac.type != e1000_i210) &&
5326 		    (hw->mac.type != e1000_i211))
5327 			adapter->stats.tncrs += rd32(E1000_TNCRS);
5328 	}
5329 
5330 	adapter->stats.tsctc += rd32(E1000_TSCTC);
5331 	adapter->stats.tsctfc += rd32(E1000_TSCTFC);
5332 
5333 	adapter->stats.iac += rd32(E1000_IAC);
5334 	adapter->stats.icrxoc += rd32(E1000_ICRXOC);
5335 	adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
5336 	adapter->stats.icrxatc += rd32(E1000_ICRXATC);
5337 	adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
5338 	adapter->stats.ictxatc += rd32(E1000_ICTXATC);
5339 	adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
5340 	adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
5341 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
5342 
5343 	/* Fill out the OS statistics structure */
5344 	net_stats->multicast = adapter->stats.mprc;
5345 	net_stats->collisions = adapter->stats.colc;
5346 
5347 	/* Rx Errors */
5348 
5349 	/* RLEC on some newer hardware can be incorrect so build
5350 	 * our own version based on RUC and ROC
5351 	 */
5352 	net_stats->rx_errors = adapter->stats.rxerrc +
5353 		adapter->stats.crcerrs + adapter->stats.algnerrc +
5354 		adapter->stats.ruc + adapter->stats.roc +
5355 		adapter->stats.cexterr;
5356 	net_stats->rx_length_errors = adapter->stats.ruc +
5357 				      adapter->stats.roc;
5358 	net_stats->rx_crc_errors = adapter->stats.crcerrs;
5359 	net_stats->rx_frame_errors = adapter->stats.algnerrc;
5360 	net_stats->rx_missed_errors = adapter->stats.mpc;
5361 
5362 	/* Tx Errors */
5363 	net_stats->tx_errors = adapter->stats.ecol +
5364 			       adapter->stats.latecol;
5365 	net_stats->tx_aborted_errors = adapter->stats.ecol;
5366 	net_stats->tx_window_errors = adapter->stats.latecol;
5367 	net_stats->tx_carrier_errors = adapter->stats.tncrs;
5368 
5369 	/* Tx Dropped needs to be maintained elsewhere */
5370 
5371 	/* Management Stats */
5372 	adapter->stats.mgptc += rd32(E1000_MGTPTC);
5373 	adapter->stats.mgprc += rd32(E1000_MGTPRC);
5374 	adapter->stats.mgpdc += rd32(E1000_MGTPDC);
5375 
5376 	/* OS2BMC Stats */
5377 	reg = rd32(E1000_MANC);
5378 	if (reg & E1000_MANC_EN_BMC2OS) {
5379 		adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
5380 		adapter->stats.o2bspc += rd32(E1000_O2BSPC);
5381 		adapter->stats.b2ospc += rd32(E1000_B2OSPC);
5382 		adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
5383 	}
5384 }
5385 
5386 static void igb_tsync_interrupt(struct igb_adapter *adapter)
5387 {
5388 	struct e1000_hw *hw = &adapter->hw;
5389 	struct ptp_clock_event event;
5390 	struct timespec ts;
5391 	u32 ack = 0, tsauxc, sec, nsec, tsicr = rd32(E1000_TSICR);
5392 
5393 	if (tsicr & TSINTR_SYS_WRAP) {
5394 		event.type = PTP_CLOCK_PPS;
5395 		if (adapter->ptp_caps.pps)
5396 			ptp_clock_event(adapter->ptp_clock, &event);
5397 		else
5398 			dev_err(&adapter->pdev->dev, "unexpected SYS WRAP");
5399 		ack |= TSINTR_SYS_WRAP;
5400 	}
5401 
5402 	if (tsicr & E1000_TSICR_TXTS) {
5403 		/* retrieve hardware timestamp */
5404 		schedule_work(&adapter->ptp_tx_work);
5405 		ack |= E1000_TSICR_TXTS;
5406 	}
5407 
5408 	if (tsicr & TSINTR_TT0) {
5409 		spin_lock(&adapter->tmreg_lock);
5410 		ts = timespec_add(adapter->perout[0].start,
5411 				  adapter->perout[0].period);
5412 		wr32(E1000_TRGTTIML0, ts.tv_nsec);
5413 		wr32(E1000_TRGTTIMH0, ts.tv_sec);
5414 		tsauxc = rd32(E1000_TSAUXC);
5415 		tsauxc |= TSAUXC_EN_TT0;
5416 		wr32(E1000_TSAUXC, tsauxc);
5417 		adapter->perout[0].start = ts;
5418 		spin_unlock(&adapter->tmreg_lock);
5419 		ack |= TSINTR_TT0;
5420 	}
5421 
5422 	if (tsicr & TSINTR_TT1) {
5423 		spin_lock(&adapter->tmreg_lock);
5424 		ts = timespec_add(adapter->perout[1].start,
5425 				  adapter->perout[1].period);
5426 		wr32(E1000_TRGTTIML1, ts.tv_nsec);
5427 		wr32(E1000_TRGTTIMH1, ts.tv_sec);
5428 		tsauxc = rd32(E1000_TSAUXC);
5429 		tsauxc |= TSAUXC_EN_TT1;
5430 		wr32(E1000_TSAUXC, tsauxc);
5431 		adapter->perout[1].start = ts;
5432 		spin_unlock(&adapter->tmreg_lock);
5433 		ack |= TSINTR_TT1;
5434 	}
5435 
5436 	if (tsicr & TSINTR_AUTT0) {
5437 		nsec = rd32(E1000_AUXSTMPL0);
5438 		sec  = rd32(E1000_AUXSTMPH0);
5439 		event.type = PTP_CLOCK_EXTTS;
5440 		event.index = 0;
5441 		event.timestamp = sec * 1000000000ULL + nsec;
5442 		ptp_clock_event(adapter->ptp_clock, &event);
5443 		ack |= TSINTR_AUTT0;
5444 	}
5445 
5446 	if (tsicr & TSINTR_AUTT1) {
5447 		nsec = rd32(E1000_AUXSTMPL1);
5448 		sec  = rd32(E1000_AUXSTMPH1);
5449 		event.type = PTP_CLOCK_EXTTS;
5450 		event.index = 1;
5451 		event.timestamp = sec * 1000000000ULL + nsec;
5452 		ptp_clock_event(adapter->ptp_clock, &event);
5453 		ack |= TSINTR_AUTT1;
5454 	}
5455 
5456 	/* acknowledge the interrupts */
5457 	wr32(E1000_TSICR, ack);
5458 }
5459 
5460 static irqreturn_t igb_msix_other(int irq, void *data)
5461 {
5462 	struct igb_adapter *adapter = data;
5463 	struct e1000_hw *hw = &adapter->hw;
5464 	u32 icr = rd32(E1000_ICR);
5465 	/* reading ICR causes bit 31 of EICR to be cleared */
5466 
5467 	if (icr & E1000_ICR_DRSTA)
5468 		schedule_work(&adapter->reset_task);
5469 
5470 	if (icr & E1000_ICR_DOUTSYNC) {
5471 		/* HW is reporting DMA is out of sync */
5472 		adapter->stats.doosync++;
5473 		/* The DMA Out of Sync is also indication of a spoof event
5474 		 * in IOV mode. Check the Wrong VM Behavior register to
5475 		 * see if it is really a spoof event.
5476 		 */
5477 		igb_check_wvbr(adapter);
5478 	}
5479 
5480 	/* Check for a mailbox event */
5481 	if (icr & E1000_ICR_VMMB)
5482 		igb_msg_task(adapter);
5483 
5484 	if (icr & E1000_ICR_LSC) {
5485 		hw->mac.get_link_status = 1;
5486 		/* guard against interrupt when we're going down */
5487 		if (!test_bit(__IGB_DOWN, &adapter->state))
5488 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
5489 	}
5490 
5491 	if (icr & E1000_ICR_TS)
5492 		igb_tsync_interrupt(adapter);
5493 
5494 	wr32(E1000_EIMS, adapter->eims_other);
5495 
5496 	return IRQ_HANDLED;
5497 }
5498 
5499 static void igb_write_itr(struct igb_q_vector *q_vector)
5500 {
5501 	struct igb_adapter *adapter = q_vector->adapter;
5502 	u32 itr_val = q_vector->itr_val & 0x7FFC;
5503 
5504 	if (!q_vector->set_itr)
5505 		return;
5506 
5507 	if (!itr_val)
5508 		itr_val = 0x4;
5509 
5510 	if (adapter->hw.mac.type == e1000_82575)
5511 		itr_val |= itr_val << 16;
5512 	else
5513 		itr_val |= E1000_EITR_CNT_IGNR;
5514 
5515 	writel(itr_val, q_vector->itr_register);
5516 	q_vector->set_itr = 0;
5517 }
5518 
5519 static irqreturn_t igb_msix_ring(int irq, void *data)
5520 {
5521 	struct igb_q_vector *q_vector = data;
5522 
5523 	/* Write the ITR value calculated from the previous interrupt. */
5524 	igb_write_itr(q_vector);
5525 
5526 	napi_schedule(&q_vector->napi);
5527 
5528 	return IRQ_HANDLED;
5529 }
5530 
5531 #ifdef CONFIG_IGB_DCA
5532 static void igb_update_tx_dca(struct igb_adapter *adapter,
5533 			      struct igb_ring *tx_ring,
5534 			      int cpu)
5535 {
5536 	struct e1000_hw *hw = &adapter->hw;
5537 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
5538 
5539 	if (hw->mac.type != e1000_82575)
5540 		txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
5541 
5542 	/* We can enable relaxed ordering for reads, but not writes when
5543 	 * DCA is enabled.  This is due to a known issue in some chipsets
5544 	 * which will cause the DCA tag to be cleared.
5545 	 */
5546 	txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
5547 		  E1000_DCA_TXCTRL_DATA_RRO_EN |
5548 		  E1000_DCA_TXCTRL_DESC_DCA_EN;
5549 
5550 	wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
5551 }
5552 
5553 static void igb_update_rx_dca(struct igb_adapter *adapter,
5554 			      struct igb_ring *rx_ring,
5555 			      int cpu)
5556 {
5557 	struct e1000_hw *hw = &adapter->hw;
5558 	u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
5559 
5560 	if (hw->mac.type != e1000_82575)
5561 		rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
5562 
5563 	/* We can enable relaxed ordering for reads, but not writes when
5564 	 * DCA is enabled.  This is due to a known issue in some chipsets
5565 	 * which will cause the DCA tag to be cleared.
5566 	 */
5567 	rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
5568 		  E1000_DCA_RXCTRL_DESC_DCA_EN;
5569 
5570 	wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
5571 }
5572 
5573 static void igb_update_dca(struct igb_q_vector *q_vector)
5574 {
5575 	struct igb_adapter *adapter = q_vector->adapter;
5576 	int cpu = get_cpu();
5577 
5578 	if (q_vector->cpu == cpu)
5579 		goto out_no_update;
5580 
5581 	if (q_vector->tx.ring)
5582 		igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
5583 
5584 	if (q_vector->rx.ring)
5585 		igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
5586 
5587 	q_vector->cpu = cpu;
5588 out_no_update:
5589 	put_cpu();
5590 }
5591 
5592 static void igb_setup_dca(struct igb_adapter *adapter)
5593 {
5594 	struct e1000_hw *hw = &adapter->hw;
5595 	int i;
5596 
5597 	if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
5598 		return;
5599 
5600 	/* Always use CB2 mode, difference is masked in the CB driver. */
5601 	wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
5602 
5603 	for (i = 0; i < adapter->num_q_vectors; i++) {
5604 		adapter->q_vector[i]->cpu = -1;
5605 		igb_update_dca(adapter->q_vector[i]);
5606 	}
5607 }
5608 
5609 static int __igb_notify_dca(struct device *dev, void *data)
5610 {
5611 	struct net_device *netdev = dev_get_drvdata(dev);
5612 	struct igb_adapter *adapter = netdev_priv(netdev);
5613 	struct pci_dev *pdev = adapter->pdev;
5614 	struct e1000_hw *hw = &adapter->hw;
5615 	unsigned long event = *(unsigned long *)data;
5616 
5617 	switch (event) {
5618 	case DCA_PROVIDER_ADD:
5619 		/* if already enabled, don't do it again */
5620 		if (adapter->flags & IGB_FLAG_DCA_ENABLED)
5621 			break;
5622 		if (dca_add_requester(dev) == 0) {
5623 			adapter->flags |= IGB_FLAG_DCA_ENABLED;
5624 			dev_info(&pdev->dev, "DCA enabled\n");
5625 			igb_setup_dca(adapter);
5626 			break;
5627 		}
5628 		/* Fall Through since DCA is disabled. */
5629 	case DCA_PROVIDER_REMOVE:
5630 		if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
5631 			/* without this a class_device is left
5632 			 * hanging around in the sysfs model
5633 			 */
5634 			dca_remove_requester(dev);
5635 			dev_info(&pdev->dev, "DCA disabled\n");
5636 			adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
5637 			wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
5638 		}
5639 		break;
5640 	}
5641 
5642 	return 0;
5643 }
5644 
5645 static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
5646 			  void *p)
5647 {
5648 	int ret_val;
5649 
5650 	ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
5651 					 __igb_notify_dca);
5652 
5653 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
5654 }
5655 #endif /* CONFIG_IGB_DCA */
5656 
5657 #ifdef CONFIG_PCI_IOV
5658 static int igb_vf_configure(struct igb_adapter *adapter, int vf)
5659 {
5660 	unsigned char mac_addr[ETH_ALEN];
5661 
5662 	eth_zero_addr(mac_addr);
5663 	igb_set_vf_mac(adapter, vf, mac_addr);
5664 
5665 	/* By default spoof check is enabled for all VFs */
5666 	adapter->vf_data[vf].spoofchk_enabled = true;
5667 
5668 	return 0;
5669 }
5670 
5671 #endif
5672 static void igb_ping_all_vfs(struct igb_adapter *adapter)
5673 {
5674 	struct e1000_hw *hw = &adapter->hw;
5675 	u32 ping;
5676 	int i;
5677 
5678 	for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
5679 		ping = E1000_PF_CONTROL_MSG;
5680 		if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
5681 			ping |= E1000_VT_MSGTYPE_CTS;
5682 		igb_write_mbx(hw, &ping, 1, i);
5683 	}
5684 }
5685 
5686 static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
5687 {
5688 	struct e1000_hw *hw = &adapter->hw;
5689 	u32 vmolr = rd32(E1000_VMOLR(vf));
5690 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
5691 
5692 	vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
5693 			    IGB_VF_FLAG_MULTI_PROMISC);
5694 	vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
5695 
5696 	if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
5697 		vmolr |= E1000_VMOLR_MPME;
5698 		vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
5699 		*msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
5700 	} else {
5701 		/* if we have hashes and we are clearing a multicast promisc
5702 		 * flag we need to write the hashes to the MTA as this step
5703 		 * was previously skipped
5704 		 */
5705 		if (vf_data->num_vf_mc_hashes > 30) {
5706 			vmolr |= E1000_VMOLR_MPME;
5707 		} else if (vf_data->num_vf_mc_hashes) {
5708 			int j;
5709 
5710 			vmolr |= E1000_VMOLR_ROMPE;
5711 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
5712 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
5713 		}
5714 	}
5715 
5716 	wr32(E1000_VMOLR(vf), vmolr);
5717 
5718 	/* there are flags left unprocessed, likely not supported */
5719 	if (*msgbuf & E1000_VT_MSGINFO_MASK)
5720 		return -EINVAL;
5721 
5722 	return 0;
5723 }
5724 
5725 static int igb_set_vf_multicasts(struct igb_adapter *adapter,
5726 				  u32 *msgbuf, u32 vf)
5727 {
5728 	int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
5729 	u16 *hash_list = (u16 *)&msgbuf[1];
5730 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
5731 	int i;
5732 
5733 	/* salt away the number of multicast addresses assigned
5734 	 * to this VF for later use to restore when the PF multi cast
5735 	 * list changes
5736 	 */
5737 	vf_data->num_vf_mc_hashes = n;
5738 
5739 	/* only up to 30 hash values supported */
5740 	if (n > 30)
5741 		n = 30;
5742 
5743 	/* store the hashes for later use */
5744 	for (i = 0; i < n; i++)
5745 		vf_data->vf_mc_hashes[i] = hash_list[i];
5746 
5747 	/* Flush and reset the mta with the new values */
5748 	igb_set_rx_mode(adapter->netdev);
5749 
5750 	return 0;
5751 }
5752 
5753 static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
5754 {
5755 	struct e1000_hw *hw = &adapter->hw;
5756 	struct vf_data_storage *vf_data;
5757 	int i, j;
5758 
5759 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
5760 		u32 vmolr = rd32(E1000_VMOLR(i));
5761 
5762 		vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
5763 
5764 		vf_data = &adapter->vf_data[i];
5765 
5766 		if ((vf_data->num_vf_mc_hashes > 30) ||
5767 		    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
5768 			vmolr |= E1000_VMOLR_MPME;
5769 		} else if (vf_data->num_vf_mc_hashes) {
5770 			vmolr |= E1000_VMOLR_ROMPE;
5771 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
5772 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
5773 		}
5774 		wr32(E1000_VMOLR(i), vmolr);
5775 	}
5776 }
5777 
5778 static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
5779 {
5780 	struct e1000_hw *hw = &adapter->hw;
5781 	u32 pool_mask, reg, vid;
5782 	int i;
5783 
5784 	pool_mask = 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
5785 
5786 	/* Find the vlan filter for this id */
5787 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5788 		reg = rd32(E1000_VLVF(i));
5789 
5790 		/* remove the vf from the pool */
5791 		reg &= ~pool_mask;
5792 
5793 		/* if pool is empty then remove entry from vfta */
5794 		if (!(reg & E1000_VLVF_POOLSEL_MASK) &&
5795 		    (reg & E1000_VLVF_VLANID_ENABLE)) {
5796 			reg = 0;
5797 			vid = reg & E1000_VLVF_VLANID_MASK;
5798 			igb_vfta_set(hw, vid, false);
5799 		}
5800 
5801 		wr32(E1000_VLVF(i), reg);
5802 	}
5803 
5804 	adapter->vf_data[vf].vlans_enabled = 0;
5805 }
5806 
5807 static s32 igb_vlvf_set(struct igb_adapter *adapter, u32 vid, bool add, u32 vf)
5808 {
5809 	struct e1000_hw *hw = &adapter->hw;
5810 	u32 reg, i;
5811 
5812 	/* The vlvf table only exists on 82576 hardware and newer */
5813 	if (hw->mac.type < e1000_82576)
5814 		return -1;
5815 
5816 	/* we only need to do this if VMDq is enabled */
5817 	if (!adapter->vfs_allocated_count)
5818 		return -1;
5819 
5820 	/* Find the vlan filter for this id */
5821 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5822 		reg = rd32(E1000_VLVF(i));
5823 		if ((reg & E1000_VLVF_VLANID_ENABLE) &&
5824 		    vid == (reg & E1000_VLVF_VLANID_MASK))
5825 			break;
5826 	}
5827 
5828 	if (add) {
5829 		if (i == E1000_VLVF_ARRAY_SIZE) {
5830 			/* Did not find a matching VLAN ID entry that was
5831 			 * enabled.  Search for a free filter entry, i.e.
5832 			 * one without the enable bit set
5833 			 */
5834 			for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5835 				reg = rd32(E1000_VLVF(i));
5836 				if (!(reg & E1000_VLVF_VLANID_ENABLE))
5837 					break;
5838 			}
5839 		}
5840 		if (i < E1000_VLVF_ARRAY_SIZE) {
5841 			/* Found an enabled/available entry */
5842 			reg |= 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
5843 
5844 			/* if !enabled we need to set this up in vfta */
5845 			if (!(reg & E1000_VLVF_VLANID_ENABLE)) {
5846 				/* add VID to filter table */
5847 				igb_vfta_set(hw, vid, true);
5848 				reg |= E1000_VLVF_VLANID_ENABLE;
5849 			}
5850 			reg &= ~E1000_VLVF_VLANID_MASK;
5851 			reg |= vid;
5852 			wr32(E1000_VLVF(i), reg);
5853 
5854 			/* do not modify RLPML for PF devices */
5855 			if (vf >= adapter->vfs_allocated_count)
5856 				return 0;
5857 
5858 			if (!adapter->vf_data[vf].vlans_enabled) {
5859 				u32 size;
5860 
5861 				reg = rd32(E1000_VMOLR(vf));
5862 				size = reg & E1000_VMOLR_RLPML_MASK;
5863 				size += 4;
5864 				reg &= ~E1000_VMOLR_RLPML_MASK;
5865 				reg |= size;
5866 				wr32(E1000_VMOLR(vf), reg);
5867 			}
5868 
5869 			adapter->vf_data[vf].vlans_enabled++;
5870 		}
5871 	} else {
5872 		if (i < E1000_VLVF_ARRAY_SIZE) {
5873 			/* remove vf from the pool */
5874 			reg &= ~(1 << (E1000_VLVF_POOLSEL_SHIFT + vf));
5875 			/* if pool is empty then remove entry from vfta */
5876 			if (!(reg & E1000_VLVF_POOLSEL_MASK)) {
5877 				reg = 0;
5878 				igb_vfta_set(hw, vid, false);
5879 			}
5880 			wr32(E1000_VLVF(i), reg);
5881 
5882 			/* do not modify RLPML for PF devices */
5883 			if (vf >= adapter->vfs_allocated_count)
5884 				return 0;
5885 
5886 			adapter->vf_data[vf].vlans_enabled--;
5887 			if (!adapter->vf_data[vf].vlans_enabled) {
5888 				u32 size;
5889 
5890 				reg = rd32(E1000_VMOLR(vf));
5891 				size = reg & E1000_VMOLR_RLPML_MASK;
5892 				size -= 4;
5893 				reg &= ~E1000_VMOLR_RLPML_MASK;
5894 				reg |= size;
5895 				wr32(E1000_VMOLR(vf), reg);
5896 			}
5897 		}
5898 	}
5899 	return 0;
5900 }
5901 
5902 static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
5903 {
5904 	struct e1000_hw *hw = &adapter->hw;
5905 
5906 	if (vid)
5907 		wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
5908 	else
5909 		wr32(E1000_VMVIR(vf), 0);
5910 }
5911 
5912 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
5913 			       int vf, u16 vlan, u8 qos)
5914 {
5915 	int err = 0;
5916 	struct igb_adapter *adapter = netdev_priv(netdev);
5917 
5918 	if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
5919 		return -EINVAL;
5920 	if (vlan || qos) {
5921 		err = igb_vlvf_set(adapter, vlan, !!vlan, vf);
5922 		if (err)
5923 			goto out;
5924 		igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
5925 		igb_set_vmolr(adapter, vf, !vlan);
5926 		adapter->vf_data[vf].pf_vlan = vlan;
5927 		adapter->vf_data[vf].pf_qos = qos;
5928 		dev_info(&adapter->pdev->dev,
5929 			 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
5930 		if (test_bit(__IGB_DOWN, &adapter->state)) {
5931 			dev_warn(&adapter->pdev->dev,
5932 				 "The VF VLAN has been set, but the PF device is not up.\n");
5933 			dev_warn(&adapter->pdev->dev,
5934 				 "Bring the PF device up before attempting to use the VF device.\n");
5935 		}
5936 	} else {
5937 		igb_vlvf_set(adapter, adapter->vf_data[vf].pf_vlan,
5938 			     false, vf);
5939 		igb_set_vmvir(adapter, vlan, vf);
5940 		igb_set_vmolr(adapter, vf, true);
5941 		adapter->vf_data[vf].pf_vlan = 0;
5942 		adapter->vf_data[vf].pf_qos = 0;
5943 	}
5944 out:
5945 	return err;
5946 }
5947 
5948 static int igb_find_vlvf_entry(struct igb_adapter *adapter, int vid)
5949 {
5950 	struct e1000_hw *hw = &adapter->hw;
5951 	int i;
5952 	u32 reg;
5953 
5954 	/* Find the vlan filter for this id */
5955 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5956 		reg = rd32(E1000_VLVF(i));
5957 		if ((reg & E1000_VLVF_VLANID_ENABLE) &&
5958 		    vid == (reg & E1000_VLVF_VLANID_MASK))
5959 			break;
5960 	}
5961 
5962 	if (i >= E1000_VLVF_ARRAY_SIZE)
5963 		i = -1;
5964 
5965 	return i;
5966 }
5967 
5968 static int igb_set_vf_vlan(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
5969 {
5970 	struct e1000_hw *hw = &adapter->hw;
5971 	int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
5972 	int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
5973 	int err = 0;
5974 
5975 	/* If in promiscuous mode we need to make sure the PF also has
5976 	 * the VLAN filter set.
5977 	 */
5978 	if (add && (adapter->netdev->flags & IFF_PROMISC))
5979 		err = igb_vlvf_set(adapter, vid, add,
5980 				   adapter->vfs_allocated_count);
5981 	if (err)
5982 		goto out;
5983 
5984 	err = igb_vlvf_set(adapter, vid, add, vf);
5985 
5986 	if (err)
5987 		goto out;
5988 
5989 	/* Go through all the checks to see if the VLAN filter should
5990 	 * be wiped completely.
5991 	 */
5992 	if (!add && (adapter->netdev->flags & IFF_PROMISC)) {
5993 		u32 vlvf, bits;
5994 		int regndx = igb_find_vlvf_entry(adapter, vid);
5995 
5996 		if (regndx < 0)
5997 			goto out;
5998 		/* See if any other pools are set for this VLAN filter
5999 		 * entry other than the PF.
6000 		 */
6001 		vlvf = bits = rd32(E1000_VLVF(regndx));
6002 		bits &= 1 << (E1000_VLVF_POOLSEL_SHIFT +
6003 			      adapter->vfs_allocated_count);
6004 		/* If the filter was removed then ensure PF pool bit
6005 		 * is cleared if the PF only added itself to the pool
6006 		 * because the PF is in promiscuous mode.
6007 		 */
6008 		if ((vlvf & VLAN_VID_MASK) == vid &&
6009 		    !test_bit(vid, adapter->active_vlans) &&
6010 		    !bits)
6011 			igb_vlvf_set(adapter, vid, add,
6012 				     adapter->vfs_allocated_count);
6013 	}
6014 
6015 out:
6016 	return err;
6017 }
6018 
6019 static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
6020 {
6021 	/* clear flags - except flag that indicates PF has set the MAC */
6022 	adapter->vf_data[vf].flags &= IGB_VF_FLAG_PF_SET_MAC;
6023 	adapter->vf_data[vf].last_nack = jiffies;
6024 
6025 	/* reset offloads to defaults */
6026 	igb_set_vmolr(adapter, vf, true);
6027 
6028 	/* reset vlans for device */
6029 	igb_clear_vf_vfta(adapter, vf);
6030 	if (adapter->vf_data[vf].pf_vlan)
6031 		igb_ndo_set_vf_vlan(adapter->netdev, vf,
6032 				    adapter->vf_data[vf].pf_vlan,
6033 				    adapter->vf_data[vf].pf_qos);
6034 	else
6035 		igb_clear_vf_vfta(adapter, vf);
6036 
6037 	/* reset multicast table array for vf */
6038 	adapter->vf_data[vf].num_vf_mc_hashes = 0;
6039 
6040 	/* Flush and reset the mta with the new values */
6041 	igb_set_rx_mode(adapter->netdev);
6042 }
6043 
6044 static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
6045 {
6046 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6047 
6048 	/* clear mac address as we were hotplug removed/added */
6049 	if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
6050 		eth_zero_addr(vf_mac);
6051 
6052 	/* process remaining reset events */
6053 	igb_vf_reset(adapter, vf);
6054 }
6055 
6056 static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
6057 {
6058 	struct e1000_hw *hw = &adapter->hw;
6059 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6060 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
6061 	u32 reg, msgbuf[3];
6062 	u8 *addr = (u8 *)(&msgbuf[1]);
6063 
6064 	/* process all the same items cleared in a function level reset */
6065 	igb_vf_reset(adapter, vf);
6066 
6067 	/* set vf mac address */
6068 	igb_rar_set_qsel(adapter, vf_mac, rar_entry, vf);
6069 
6070 	/* enable transmit and receive for vf */
6071 	reg = rd32(E1000_VFTE);
6072 	wr32(E1000_VFTE, reg | (1 << vf));
6073 	reg = rd32(E1000_VFRE);
6074 	wr32(E1000_VFRE, reg | (1 << vf));
6075 
6076 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
6077 
6078 	/* reply to reset with ack and vf mac address */
6079 	if (!is_zero_ether_addr(vf_mac)) {
6080 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
6081 		memcpy(addr, vf_mac, ETH_ALEN);
6082 	} else {
6083 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
6084 	}
6085 	igb_write_mbx(hw, msgbuf, 3, vf);
6086 }
6087 
6088 static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
6089 {
6090 	/* The VF MAC Address is stored in a packed array of bytes
6091 	 * starting at the second 32 bit word of the msg array
6092 	 */
6093 	unsigned char *addr = (char *)&msg[1];
6094 	int err = -1;
6095 
6096 	if (is_valid_ether_addr(addr))
6097 		err = igb_set_vf_mac(adapter, vf, addr);
6098 
6099 	return err;
6100 }
6101 
6102 static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
6103 {
6104 	struct e1000_hw *hw = &adapter->hw;
6105 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6106 	u32 msg = E1000_VT_MSGTYPE_NACK;
6107 
6108 	/* if device isn't clear to send it shouldn't be reading either */
6109 	if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
6110 	    time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
6111 		igb_write_mbx(hw, &msg, 1, vf);
6112 		vf_data->last_nack = jiffies;
6113 	}
6114 }
6115 
6116 static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
6117 {
6118 	struct pci_dev *pdev = adapter->pdev;
6119 	u32 msgbuf[E1000_VFMAILBOX_SIZE];
6120 	struct e1000_hw *hw = &adapter->hw;
6121 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6122 	s32 retval;
6123 
6124 	retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf);
6125 
6126 	if (retval) {
6127 		/* if receive failed revoke VF CTS stats and restart init */
6128 		dev_err(&pdev->dev, "Error receiving message from VF\n");
6129 		vf_data->flags &= ~IGB_VF_FLAG_CTS;
6130 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
6131 			return;
6132 		goto out;
6133 	}
6134 
6135 	/* this is a message we already processed, do nothing */
6136 	if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
6137 		return;
6138 
6139 	/* until the vf completes a reset it should not be
6140 	 * allowed to start any configuration.
6141 	 */
6142 	if (msgbuf[0] == E1000_VF_RESET) {
6143 		igb_vf_reset_msg(adapter, vf);
6144 		return;
6145 	}
6146 
6147 	if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
6148 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
6149 			return;
6150 		retval = -1;
6151 		goto out;
6152 	}
6153 
6154 	switch ((msgbuf[0] & 0xFFFF)) {
6155 	case E1000_VF_SET_MAC_ADDR:
6156 		retval = -EINVAL;
6157 		if (!(vf_data->flags & IGB_VF_FLAG_PF_SET_MAC))
6158 			retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
6159 		else
6160 			dev_warn(&pdev->dev,
6161 				 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
6162 				 vf);
6163 		break;
6164 	case E1000_VF_SET_PROMISC:
6165 		retval = igb_set_vf_promisc(adapter, msgbuf, vf);
6166 		break;
6167 	case E1000_VF_SET_MULTICAST:
6168 		retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
6169 		break;
6170 	case E1000_VF_SET_LPE:
6171 		retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
6172 		break;
6173 	case E1000_VF_SET_VLAN:
6174 		retval = -1;
6175 		if (vf_data->pf_vlan)
6176 			dev_warn(&pdev->dev,
6177 				 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
6178 				 vf);
6179 		else
6180 			retval = igb_set_vf_vlan(adapter, msgbuf, vf);
6181 		break;
6182 	default:
6183 		dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
6184 		retval = -1;
6185 		break;
6186 	}
6187 
6188 	msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
6189 out:
6190 	/* notify the VF of the results of what it sent us */
6191 	if (retval)
6192 		msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
6193 	else
6194 		msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
6195 
6196 	igb_write_mbx(hw, msgbuf, 1, vf);
6197 }
6198 
6199 static void igb_msg_task(struct igb_adapter *adapter)
6200 {
6201 	struct e1000_hw *hw = &adapter->hw;
6202 	u32 vf;
6203 
6204 	for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
6205 		/* process any reset requests */
6206 		if (!igb_check_for_rst(hw, vf))
6207 			igb_vf_reset_event(adapter, vf);
6208 
6209 		/* process any messages pending */
6210 		if (!igb_check_for_msg(hw, vf))
6211 			igb_rcv_msg_from_vf(adapter, vf);
6212 
6213 		/* process any acks */
6214 		if (!igb_check_for_ack(hw, vf))
6215 			igb_rcv_ack_from_vf(adapter, vf);
6216 	}
6217 }
6218 
6219 /**
6220  *  igb_set_uta - Set unicast filter table address
6221  *  @adapter: board private structure
6222  *
6223  *  The unicast table address is a register array of 32-bit registers.
6224  *  The table is meant to be used in a way similar to how the MTA is used
6225  *  however due to certain limitations in the hardware it is necessary to
6226  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
6227  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
6228  **/
6229 static void igb_set_uta(struct igb_adapter *adapter)
6230 {
6231 	struct e1000_hw *hw = &adapter->hw;
6232 	int i;
6233 
6234 	/* The UTA table only exists on 82576 hardware and newer */
6235 	if (hw->mac.type < e1000_82576)
6236 		return;
6237 
6238 	/* we only need to do this if VMDq is enabled */
6239 	if (!adapter->vfs_allocated_count)
6240 		return;
6241 
6242 	for (i = 0; i < hw->mac.uta_reg_count; i++)
6243 		array_wr32(E1000_UTA, i, ~0);
6244 }
6245 
6246 /**
6247  *  igb_intr_msi - Interrupt Handler
6248  *  @irq: interrupt number
6249  *  @data: pointer to a network interface device structure
6250  **/
6251 static irqreturn_t igb_intr_msi(int irq, void *data)
6252 {
6253 	struct igb_adapter *adapter = data;
6254 	struct igb_q_vector *q_vector = adapter->q_vector[0];
6255 	struct e1000_hw *hw = &adapter->hw;
6256 	/* read ICR disables interrupts using IAM */
6257 	u32 icr = rd32(E1000_ICR);
6258 
6259 	igb_write_itr(q_vector);
6260 
6261 	if (icr & E1000_ICR_DRSTA)
6262 		schedule_work(&adapter->reset_task);
6263 
6264 	if (icr & E1000_ICR_DOUTSYNC) {
6265 		/* HW is reporting DMA is out of sync */
6266 		adapter->stats.doosync++;
6267 	}
6268 
6269 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
6270 		hw->mac.get_link_status = 1;
6271 		if (!test_bit(__IGB_DOWN, &adapter->state))
6272 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6273 	}
6274 
6275 	if (icr & E1000_ICR_TS)
6276 		igb_tsync_interrupt(adapter);
6277 
6278 	napi_schedule(&q_vector->napi);
6279 
6280 	return IRQ_HANDLED;
6281 }
6282 
6283 /**
6284  *  igb_intr - Legacy Interrupt Handler
6285  *  @irq: interrupt number
6286  *  @data: pointer to a network interface device structure
6287  **/
6288 static irqreturn_t igb_intr(int irq, void *data)
6289 {
6290 	struct igb_adapter *adapter = data;
6291 	struct igb_q_vector *q_vector = adapter->q_vector[0];
6292 	struct e1000_hw *hw = &adapter->hw;
6293 	/* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
6294 	 * need for the IMC write
6295 	 */
6296 	u32 icr = rd32(E1000_ICR);
6297 
6298 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
6299 	 * not set, then the adapter didn't send an interrupt
6300 	 */
6301 	if (!(icr & E1000_ICR_INT_ASSERTED))
6302 		return IRQ_NONE;
6303 
6304 	igb_write_itr(q_vector);
6305 
6306 	if (icr & E1000_ICR_DRSTA)
6307 		schedule_work(&adapter->reset_task);
6308 
6309 	if (icr & E1000_ICR_DOUTSYNC) {
6310 		/* HW is reporting DMA is out of sync */
6311 		adapter->stats.doosync++;
6312 	}
6313 
6314 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
6315 		hw->mac.get_link_status = 1;
6316 		/* guard against interrupt when we're going down */
6317 		if (!test_bit(__IGB_DOWN, &adapter->state))
6318 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6319 	}
6320 
6321 	if (icr & E1000_ICR_TS)
6322 		igb_tsync_interrupt(adapter);
6323 
6324 	napi_schedule(&q_vector->napi);
6325 
6326 	return IRQ_HANDLED;
6327 }
6328 
6329 static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
6330 {
6331 	struct igb_adapter *adapter = q_vector->adapter;
6332 	struct e1000_hw *hw = &adapter->hw;
6333 
6334 	if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
6335 	    (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
6336 		if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
6337 			igb_set_itr(q_vector);
6338 		else
6339 			igb_update_ring_itr(q_vector);
6340 	}
6341 
6342 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
6343 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
6344 			wr32(E1000_EIMS, q_vector->eims_value);
6345 		else
6346 			igb_irq_enable(adapter);
6347 	}
6348 }
6349 
6350 /**
6351  *  igb_poll - NAPI Rx polling callback
6352  *  @napi: napi polling structure
6353  *  @budget: count of how many packets we should handle
6354  **/
6355 static int igb_poll(struct napi_struct *napi, int budget)
6356 {
6357 	struct igb_q_vector *q_vector = container_of(napi,
6358 						     struct igb_q_vector,
6359 						     napi);
6360 	bool clean_complete = true;
6361 
6362 #ifdef CONFIG_IGB_DCA
6363 	if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
6364 		igb_update_dca(q_vector);
6365 #endif
6366 	if (q_vector->tx.ring)
6367 		clean_complete = igb_clean_tx_irq(q_vector);
6368 
6369 	if (q_vector->rx.ring)
6370 		clean_complete &= igb_clean_rx_irq(q_vector, budget);
6371 
6372 	/* If all work not completed, return budget and keep polling */
6373 	if (!clean_complete)
6374 		return budget;
6375 
6376 	/* If not enough Rx work done, exit the polling mode */
6377 	napi_complete(napi);
6378 	igb_ring_irq_enable(q_vector);
6379 
6380 	return 0;
6381 }
6382 
6383 /**
6384  *  igb_clean_tx_irq - Reclaim resources after transmit completes
6385  *  @q_vector: pointer to q_vector containing needed info
6386  *
6387  *  returns true if ring is completely cleaned
6388  **/
6389 static bool igb_clean_tx_irq(struct igb_q_vector *q_vector)
6390 {
6391 	struct igb_adapter *adapter = q_vector->adapter;
6392 	struct igb_ring *tx_ring = q_vector->tx.ring;
6393 	struct igb_tx_buffer *tx_buffer;
6394 	union e1000_adv_tx_desc *tx_desc;
6395 	unsigned int total_bytes = 0, total_packets = 0;
6396 	unsigned int budget = q_vector->tx.work_limit;
6397 	unsigned int i = tx_ring->next_to_clean;
6398 
6399 	if (test_bit(__IGB_DOWN, &adapter->state))
6400 		return true;
6401 
6402 	tx_buffer = &tx_ring->tx_buffer_info[i];
6403 	tx_desc = IGB_TX_DESC(tx_ring, i);
6404 	i -= tx_ring->count;
6405 
6406 	do {
6407 		union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
6408 
6409 		/* if next_to_watch is not set then there is no work pending */
6410 		if (!eop_desc)
6411 			break;
6412 
6413 		/* prevent any other reads prior to eop_desc */
6414 		read_barrier_depends();
6415 
6416 		/* if DD is not set pending work has not been completed */
6417 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
6418 			break;
6419 
6420 		/* clear next_to_watch to prevent false hangs */
6421 		tx_buffer->next_to_watch = NULL;
6422 
6423 		/* update the statistics for this packet */
6424 		total_bytes += tx_buffer->bytecount;
6425 		total_packets += tx_buffer->gso_segs;
6426 
6427 		/* free the skb */
6428 		dev_consume_skb_any(tx_buffer->skb);
6429 
6430 		/* unmap skb header data */
6431 		dma_unmap_single(tx_ring->dev,
6432 				 dma_unmap_addr(tx_buffer, dma),
6433 				 dma_unmap_len(tx_buffer, len),
6434 				 DMA_TO_DEVICE);
6435 
6436 		/* clear tx_buffer data */
6437 		tx_buffer->skb = NULL;
6438 		dma_unmap_len_set(tx_buffer, len, 0);
6439 
6440 		/* clear last DMA location and unmap remaining buffers */
6441 		while (tx_desc != eop_desc) {
6442 			tx_buffer++;
6443 			tx_desc++;
6444 			i++;
6445 			if (unlikely(!i)) {
6446 				i -= tx_ring->count;
6447 				tx_buffer = tx_ring->tx_buffer_info;
6448 				tx_desc = IGB_TX_DESC(tx_ring, 0);
6449 			}
6450 
6451 			/* unmap any remaining paged data */
6452 			if (dma_unmap_len(tx_buffer, len)) {
6453 				dma_unmap_page(tx_ring->dev,
6454 					       dma_unmap_addr(tx_buffer, dma),
6455 					       dma_unmap_len(tx_buffer, len),
6456 					       DMA_TO_DEVICE);
6457 				dma_unmap_len_set(tx_buffer, len, 0);
6458 			}
6459 		}
6460 
6461 		/* move us one more past the eop_desc for start of next pkt */
6462 		tx_buffer++;
6463 		tx_desc++;
6464 		i++;
6465 		if (unlikely(!i)) {
6466 			i -= tx_ring->count;
6467 			tx_buffer = tx_ring->tx_buffer_info;
6468 			tx_desc = IGB_TX_DESC(tx_ring, 0);
6469 		}
6470 
6471 		/* issue prefetch for next Tx descriptor */
6472 		prefetch(tx_desc);
6473 
6474 		/* update budget accounting */
6475 		budget--;
6476 	} while (likely(budget));
6477 
6478 	netdev_tx_completed_queue(txring_txq(tx_ring),
6479 				  total_packets, total_bytes);
6480 	i += tx_ring->count;
6481 	tx_ring->next_to_clean = i;
6482 	u64_stats_update_begin(&tx_ring->tx_syncp);
6483 	tx_ring->tx_stats.bytes += total_bytes;
6484 	tx_ring->tx_stats.packets += total_packets;
6485 	u64_stats_update_end(&tx_ring->tx_syncp);
6486 	q_vector->tx.total_bytes += total_bytes;
6487 	q_vector->tx.total_packets += total_packets;
6488 
6489 	if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
6490 		struct e1000_hw *hw = &adapter->hw;
6491 
6492 		/* Detect a transmit hang in hardware, this serializes the
6493 		 * check with the clearing of time_stamp and movement of i
6494 		 */
6495 		clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
6496 		if (tx_buffer->next_to_watch &&
6497 		    time_after(jiffies, tx_buffer->time_stamp +
6498 			       (adapter->tx_timeout_factor * HZ)) &&
6499 		    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
6500 
6501 			/* detected Tx unit hang */
6502 			dev_err(tx_ring->dev,
6503 				"Detected Tx Unit Hang\n"
6504 				"  Tx Queue             <%d>\n"
6505 				"  TDH                  <%x>\n"
6506 				"  TDT                  <%x>\n"
6507 				"  next_to_use          <%x>\n"
6508 				"  next_to_clean        <%x>\n"
6509 				"buffer_info[next_to_clean]\n"
6510 				"  time_stamp           <%lx>\n"
6511 				"  next_to_watch        <%p>\n"
6512 				"  jiffies              <%lx>\n"
6513 				"  desc.status          <%x>\n",
6514 				tx_ring->queue_index,
6515 				rd32(E1000_TDH(tx_ring->reg_idx)),
6516 				readl(tx_ring->tail),
6517 				tx_ring->next_to_use,
6518 				tx_ring->next_to_clean,
6519 				tx_buffer->time_stamp,
6520 				tx_buffer->next_to_watch,
6521 				jiffies,
6522 				tx_buffer->next_to_watch->wb.status);
6523 			netif_stop_subqueue(tx_ring->netdev,
6524 					    tx_ring->queue_index);
6525 
6526 			/* we are about to reset, no point in enabling stuff */
6527 			return true;
6528 		}
6529 	}
6530 
6531 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
6532 	if (unlikely(total_packets &&
6533 	    netif_carrier_ok(tx_ring->netdev) &&
6534 	    igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
6535 		/* Make sure that anybody stopping the queue after this
6536 		 * sees the new next_to_clean.
6537 		 */
6538 		smp_mb();
6539 		if (__netif_subqueue_stopped(tx_ring->netdev,
6540 					     tx_ring->queue_index) &&
6541 		    !(test_bit(__IGB_DOWN, &adapter->state))) {
6542 			netif_wake_subqueue(tx_ring->netdev,
6543 					    tx_ring->queue_index);
6544 
6545 			u64_stats_update_begin(&tx_ring->tx_syncp);
6546 			tx_ring->tx_stats.restart_queue++;
6547 			u64_stats_update_end(&tx_ring->tx_syncp);
6548 		}
6549 	}
6550 
6551 	return !!budget;
6552 }
6553 
6554 /**
6555  *  igb_reuse_rx_page - page flip buffer and store it back on the ring
6556  *  @rx_ring: rx descriptor ring to store buffers on
6557  *  @old_buff: donor buffer to have page reused
6558  *
6559  *  Synchronizes page for reuse by the adapter
6560  **/
6561 static void igb_reuse_rx_page(struct igb_ring *rx_ring,
6562 			      struct igb_rx_buffer *old_buff)
6563 {
6564 	struct igb_rx_buffer *new_buff;
6565 	u16 nta = rx_ring->next_to_alloc;
6566 
6567 	new_buff = &rx_ring->rx_buffer_info[nta];
6568 
6569 	/* update, and store next to alloc */
6570 	nta++;
6571 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
6572 
6573 	/* transfer page from old buffer to new buffer */
6574 	*new_buff = *old_buff;
6575 
6576 	/* sync the buffer for use by the device */
6577 	dma_sync_single_range_for_device(rx_ring->dev, old_buff->dma,
6578 					 old_buff->page_offset,
6579 					 IGB_RX_BUFSZ,
6580 					 DMA_FROM_DEVICE);
6581 }
6582 
6583 static inline bool igb_page_is_reserved(struct page *page)
6584 {
6585 	return (page_to_nid(page) != numa_mem_id()) || page->pfmemalloc;
6586 }
6587 
6588 static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer,
6589 				  struct page *page,
6590 				  unsigned int truesize)
6591 {
6592 	/* avoid re-using remote pages */
6593 	if (unlikely(igb_page_is_reserved(page)))
6594 		return false;
6595 
6596 #if (PAGE_SIZE < 8192)
6597 	/* if we are only owner of page we can reuse it */
6598 	if (unlikely(page_count(page) != 1))
6599 		return false;
6600 
6601 	/* flip page offset to other buffer */
6602 	rx_buffer->page_offset ^= IGB_RX_BUFSZ;
6603 #else
6604 	/* move offset up to the next cache line */
6605 	rx_buffer->page_offset += truesize;
6606 
6607 	if (rx_buffer->page_offset > (PAGE_SIZE - IGB_RX_BUFSZ))
6608 		return false;
6609 #endif
6610 
6611 	/* Even if we own the page, we are not allowed to use atomic_set()
6612 	 * This would break get_page_unless_zero() users.
6613 	 */
6614 	atomic_inc(&page->_count);
6615 
6616 	return true;
6617 }
6618 
6619 /**
6620  *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
6621  *  @rx_ring: rx descriptor ring to transact packets on
6622  *  @rx_buffer: buffer containing page to add
6623  *  @rx_desc: descriptor containing length of buffer written by hardware
6624  *  @skb: sk_buff to place the data into
6625  *
6626  *  This function will add the data contained in rx_buffer->page to the skb.
6627  *  This is done either through a direct copy if the data in the buffer is
6628  *  less than the skb header size, otherwise it will just attach the page as
6629  *  a frag to the skb.
6630  *
6631  *  The function will then update the page offset if necessary and return
6632  *  true if the buffer can be reused by the adapter.
6633  **/
6634 static bool igb_add_rx_frag(struct igb_ring *rx_ring,
6635 			    struct igb_rx_buffer *rx_buffer,
6636 			    union e1000_adv_rx_desc *rx_desc,
6637 			    struct sk_buff *skb)
6638 {
6639 	struct page *page = rx_buffer->page;
6640 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
6641 #if (PAGE_SIZE < 8192)
6642 	unsigned int truesize = IGB_RX_BUFSZ;
6643 #else
6644 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
6645 #endif
6646 
6647 	if ((size <= IGB_RX_HDR_LEN) && !skb_is_nonlinear(skb)) {
6648 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
6649 
6650 		if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
6651 			igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
6652 			va += IGB_TS_HDR_LEN;
6653 			size -= IGB_TS_HDR_LEN;
6654 		}
6655 
6656 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
6657 
6658 		/* page is not reserved, we can reuse buffer as-is */
6659 		if (likely(!igb_page_is_reserved(page)))
6660 			return true;
6661 
6662 		/* this page cannot be reused so discard it */
6663 		__free_page(page);
6664 		return false;
6665 	}
6666 
6667 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
6668 			rx_buffer->page_offset, size, truesize);
6669 
6670 	return igb_can_reuse_rx_page(rx_buffer, page, truesize);
6671 }
6672 
6673 static struct sk_buff *igb_fetch_rx_buffer(struct igb_ring *rx_ring,
6674 					   union e1000_adv_rx_desc *rx_desc,
6675 					   struct sk_buff *skb)
6676 {
6677 	struct igb_rx_buffer *rx_buffer;
6678 	struct page *page;
6679 
6680 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
6681 	page = rx_buffer->page;
6682 	prefetchw(page);
6683 
6684 	if (likely(!skb)) {
6685 		void *page_addr = page_address(page) +
6686 				  rx_buffer->page_offset;
6687 
6688 		/* prefetch first cache line of first page */
6689 		prefetch(page_addr);
6690 #if L1_CACHE_BYTES < 128
6691 		prefetch(page_addr + L1_CACHE_BYTES);
6692 #endif
6693 
6694 		/* allocate a skb to store the frags */
6695 		skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
6696 		if (unlikely(!skb)) {
6697 			rx_ring->rx_stats.alloc_failed++;
6698 			return NULL;
6699 		}
6700 
6701 		/* we will be copying header into skb->data in
6702 		 * pskb_may_pull so it is in our interest to prefetch
6703 		 * it now to avoid a possible cache miss
6704 		 */
6705 		prefetchw(skb->data);
6706 	}
6707 
6708 	/* we are reusing so sync this buffer for CPU use */
6709 	dma_sync_single_range_for_cpu(rx_ring->dev,
6710 				      rx_buffer->dma,
6711 				      rx_buffer->page_offset,
6712 				      IGB_RX_BUFSZ,
6713 				      DMA_FROM_DEVICE);
6714 
6715 	/* pull page into skb */
6716 	if (igb_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
6717 		/* hand second half of page back to the ring */
6718 		igb_reuse_rx_page(rx_ring, rx_buffer);
6719 	} else {
6720 		/* we are not reusing the buffer so unmap it */
6721 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
6722 			       PAGE_SIZE, DMA_FROM_DEVICE);
6723 	}
6724 
6725 	/* clear contents of rx_buffer */
6726 	rx_buffer->page = NULL;
6727 
6728 	return skb;
6729 }
6730 
6731 static inline void igb_rx_checksum(struct igb_ring *ring,
6732 				   union e1000_adv_rx_desc *rx_desc,
6733 				   struct sk_buff *skb)
6734 {
6735 	skb_checksum_none_assert(skb);
6736 
6737 	/* Ignore Checksum bit is set */
6738 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
6739 		return;
6740 
6741 	/* Rx checksum disabled via ethtool */
6742 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
6743 		return;
6744 
6745 	/* TCP/UDP checksum error bit is set */
6746 	if (igb_test_staterr(rx_desc,
6747 			     E1000_RXDEXT_STATERR_TCPE |
6748 			     E1000_RXDEXT_STATERR_IPE)) {
6749 		/* work around errata with sctp packets where the TCPE aka
6750 		 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
6751 		 * packets, (aka let the stack check the crc32c)
6752 		 */
6753 		if (!((skb->len == 60) &&
6754 		      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
6755 			u64_stats_update_begin(&ring->rx_syncp);
6756 			ring->rx_stats.csum_err++;
6757 			u64_stats_update_end(&ring->rx_syncp);
6758 		}
6759 		/* let the stack verify checksum errors */
6760 		return;
6761 	}
6762 	/* It must be a TCP or UDP packet with a valid checksum */
6763 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
6764 				      E1000_RXD_STAT_UDPCS))
6765 		skb->ip_summed = CHECKSUM_UNNECESSARY;
6766 
6767 	dev_dbg(ring->dev, "cksum success: bits %08X\n",
6768 		le32_to_cpu(rx_desc->wb.upper.status_error));
6769 }
6770 
6771 static inline void igb_rx_hash(struct igb_ring *ring,
6772 			       union e1000_adv_rx_desc *rx_desc,
6773 			       struct sk_buff *skb)
6774 {
6775 	if (ring->netdev->features & NETIF_F_RXHASH)
6776 		skb_set_hash(skb,
6777 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
6778 			     PKT_HASH_TYPE_L3);
6779 }
6780 
6781 /**
6782  *  igb_is_non_eop - process handling of non-EOP buffers
6783  *  @rx_ring: Rx ring being processed
6784  *  @rx_desc: Rx descriptor for current buffer
6785  *  @skb: current socket buffer containing buffer in progress
6786  *
6787  *  This function updates next to clean.  If the buffer is an EOP buffer
6788  *  this function exits returning false, otherwise it will place the
6789  *  sk_buff in the next buffer to be chained and return true indicating
6790  *  that this is in fact a non-EOP buffer.
6791  **/
6792 static bool igb_is_non_eop(struct igb_ring *rx_ring,
6793 			   union e1000_adv_rx_desc *rx_desc)
6794 {
6795 	u32 ntc = rx_ring->next_to_clean + 1;
6796 
6797 	/* fetch, update, and store next to clean */
6798 	ntc = (ntc < rx_ring->count) ? ntc : 0;
6799 	rx_ring->next_to_clean = ntc;
6800 
6801 	prefetch(IGB_RX_DESC(rx_ring, ntc));
6802 
6803 	if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
6804 		return false;
6805 
6806 	return true;
6807 }
6808 
6809 /**
6810  *  igb_pull_tail - igb specific version of skb_pull_tail
6811  *  @rx_ring: rx descriptor ring packet is being transacted on
6812  *  @rx_desc: pointer to the EOP Rx descriptor
6813  *  @skb: pointer to current skb being adjusted
6814  *
6815  *  This function is an igb specific version of __pskb_pull_tail.  The
6816  *  main difference between this version and the original function is that
6817  *  this function can make several assumptions about the state of things
6818  *  that allow for significant optimizations versus the standard function.
6819  *  As a result we can do things like drop a frag and maintain an accurate
6820  *  truesize for the skb.
6821  */
6822 static void igb_pull_tail(struct igb_ring *rx_ring,
6823 			  union e1000_adv_rx_desc *rx_desc,
6824 			  struct sk_buff *skb)
6825 {
6826 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
6827 	unsigned char *va;
6828 	unsigned int pull_len;
6829 
6830 	/* it is valid to use page_address instead of kmap since we are
6831 	 * working with pages allocated out of the lomem pool per
6832 	 * alloc_page(GFP_ATOMIC)
6833 	 */
6834 	va = skb_frag_address(frag);
6835 
6836 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
6837 		/* retrieve timestamp from buffer */
6838 		igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
6839 
6840 		/* update pointers to remove timestamp header */
6841 		skb_frag_size_sub(frag, IGB_TS_HDR_LEN);
6842 		frag->page_offset += IGB_TS_HDR_LEN;
6843 		skb->data_len -= IGB_TS_HDR_LEN;
6844 		skb->len -= IGB_TS_HDR_LEN;
6845 
6846 		/* move va to start of packet data */
6847 		va += IGB_TS_HDR_LEN;
6848 	}
6849 
6850 	/* we need the header to contain the greater of either ETH_HLEN or
6851 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
6852 	 */
6853 	pull_len = eth_get_headlen(va, IGB_RX_HDR_LEN);
6854 
6855 	/* align pull length to size of long to optimize memcpy performance */
6856 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
6857 
6858 	/* update all of the pointers */
6859 	skb_frag_size_sub(frag, pull_len);
6860 	frag->page_offset += pull_len;
6861 	skb->data_len -= pull_len;
6862 	skb->tail += pull_len;
6863 }
6864 
6865 /**
6866  *  igb_cleanup_headers - Correct corrupted or empty headers
6867  *  @rx_ring: rx descriptor ring packet is being transacted on
6868  *  @rx_desc: pointer to the EOP Rx descriptor
6869  *  @skb: pointer to current skb being fixed
6870  *
6871  *  Address the case where we are pulling data in on pages only
6872  *  and as such no data is present in the skb header.
6873  *
6874  *  In addition if skb is not at least 60 bytes we need to pad it so that
6875  *  it is large enough to qualify as a valid Ethernet frame.
6876  *
6877  *  Returns true if an error was encountered and skb was freed.
6878  **/
6879 static bool igb_cleanup_headers(struct igb_ring *rx_ring,
6880 				union e1000_adv_rx_desc *rx_desc,
6881 				struct sk_buff *skb)
6882 {
6883 	if (unlikely((igb_test_staterr(rx_desc,
6884 				       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
6885 		struct net_device *netdev = rx_ring->netdev;
6886 		if (!(netdev->features & NETIF_F_RXALL)) {
6887 			dev_kfree_skb_any(skb);
6888 			return true;
6889 		}
6890 	}
6891 
6892 	/* place header in linear portion of buffer */
6893 	if (skb_is_nonlinear(skb))
6894 		igb_pull_tail(rx_ring, rx_desc, skb);
6895 
6896 	/* if eth_skb_pad returns an error the skb was freed */
6897 	if (eth_skb_pad(skb))
6898 		return true;
6899 
6900 	return false;
6901 }
6902 
6903 /**
6904  *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
6905  *  @rx_ring: rx descriptor ring packet is being transacted on
6906  *  @rx_desc: pointer to the EOP Rx descriptor
6907  *  @skb: pointer to current skb being populated
6908  *
6909  *  This function checks the ring, descriptor, and packet information in
6910  *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
6911  *  other fields within the skb.
6912  **/
6913 static void igb_process_skb_fields(struct igb_ring *rx_ring,
6914 				   union e1000_adv_rx_desc *rx_desc,
6915 				   struct sk_buff *skb)
6916 {
6917 	struct net_device *dev = rx_ring->netdev;
6918 
6919 	igb_rx_hash(rx_ring, rx_desc, skb);
6920 
6921 	igb_rx_checksum(rx_ring, rx_desc, skb);
6922 
6923 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
6924 	    !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
6925 		igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
6926 
6927 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
6928 	    igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
6929 		u16 vid;
6930 
6931 		if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
6932 		    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
6933 			vid = be16_to_cpu(rx_desc->wb.upper.vlan);
6934 		else
6935 			vid = le16_to_cpu(rx_desc->wb.upper.vlan);
6936 
6937 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
6938 	}
6939 
6940 	skb_record_rx_queue(skb, rx_ring->queue_index);
6941 
6942 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
6943 }
6944 
6945 static bool igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
6946 {
6947 	struct igb_ring *rx_ring = q_vector->rx.ring;
6948 	struct sk_buff *skb = rx_ring->skb;
6949 	unsigned int total_bytes = 0, total_packets = 0;
6950 	u16 cleaned_count = igb_desc_unused(rx_ring);
6951 
6952 	while (likely(total_packets < budget)) {
6953 		union e1000_adv_rx_desc *rx_desc;
6954 
6955 		/* return some buffers to hardware, one at a time is too slow */
6956 		if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
6957 			igb_alloc_rx_buffers(rx_ring, cleaned_count);
6958 			cleaned_count = 0;
6959 		}
6960 
6961 		rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
6962 
6963 		if (!rx_desc->wb.upper.status_error)
6964 			break;
6965 
6966 		/* This memory barrier is needed to keep us from reading
6967 		 * any other fields out of the rx_desc until we know the
6968 		 * descriptor has been written back
6969 		 */
6970 		dma_rmb();
6971 
6972 		/* retrieve a buffer from the ring */
6973 		skb = igb_fetch_rx_buffer(rx_ring, rx_desc, skb);
6974 
6975 		/* exit if we failed to retrieve a buffer */
6976 		if (!skb)
6977 			break;
6978 
6979 		cleaned_count++;
6980 
6981 		/* fetch next buffer in frame if non-eop */
6982 		if (igb_is_non_eop(rx_ring, rx_desc))
6983 			continue;
6984 
6985 		/* verify the packet layout is correct */
6986 		if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
6987 			skb = NULL;
6988 			continue;
6989 		}
6990 
6991 		/* probably a little skewed due to removing CRC */
6992 		total_bytes += skb->len;
6993 
6994 		/* populate checksum, timestamp, VLAN, and protocol */
6995 		igb_process_skb_fields(rx_ring, rx_desc, skb);
6996 
6997 		napi_gro_receive(&q_vector->napi, skb);
6998 
6999 		/* reset skb pointer */
7000 		skb = NULL;
7001 
7002 		/* update budget accounting */
7003 		total_packets++;
7004 	}
7005 
7006 	/* place incomplete frames back on ring for completion */
7007 	rx_ring->skb = skb;
7008 
7009 	u64_stats_update_begin(&rx_ring->rx_syncp);
7010 	rx_ring->rx_stats.packets += total_packets;
7011 	rx_ring->rx_stats.bytes += total_bytes;
7012 	u64_stats_update_end(&rx_ring->rx_syncp);
7013 	q_vector->rx.total_packets += total_packets;
7014 	q_vector->rx.total_bytes += total_bytes;
7015 
7016 	if (cleaned_count)
7017 		igb_alloc_rx_buffers(rx_ring, cleaned_count);
7018 
7019 	return total_packets < budget;
7020 }
7021 
7022 static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
7023 				  struct igb_rx_buffer *bi)
7024 {
7025 	struct page *page = bi->page;
7026 	dma_addr_t dma;
7027 
7028 	/* since we are recycling buffers we should seldom need to alloc */
7029 	if (likely(page))
7030 		return true;
7031 
7032 	/* alloc new page for storage */
7033 	page = dev_alloc_page();
7034 	if (unlikely(!page)) {
7035 		rx_ring->rx_stats.alloc_failed++;
7036 		return false;
7037 	}
7038 
7039 	/* map page for use */
7040 	dma = dma_map_page(rx_ring->dev, page, 0, PAGE_SIZE, DMA_FROM_DEVICE);
7041 
7042 	/* if mapping failed free memory back to system since
7043 	 * there isn't much point in holding memory we can't use
7044 	 */
7045 	if (dma_mapping_error(rx_ring->dev, dma)) {
7046 		__free_page(page);
7047 
7048 		rx_ring->rx_stats.alloc_failed++;
7049 		return false;
7050 	}
7051 
7052 	bi->dma = dma;
7053 	bi->page = page;
7054 	bi->page_offset = 0;
7055 
7056 	return true;
7057 }
7058 
7059 /**
7060  *  igb_alloc_rx_buffers - Replace used receive buffers; packet split
7061  *  @adapter: address of board private structure
7062  **/
7063 void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
7064 {
7065 	union e1000_adv_rx_desc *rx_desc;
7066 	struct igb_rx_buffer *bi;
7067 	u16 i = rx_ring->next_to_use;
7068 
7069 	/* nothing to do */
7070 	if (!cleaned_count)
7071 		return;
7072 
7073 	rx_desc = IGB_RX_DESC(rx_ring, i);
7074 	bi = &rx_ring->rx_buffer_info[i];
7075 	i -= rx_ring->count;
7076 
7077 	do {
7078 		if (!igb_alloc_mapped_page(rx_ring, bi))
7079 			break;
7080 
7081 		/* Refresh the desc even if buffer_addrs didn't change
7082 		 * because each write-back erases this info.
7083 		 */
7084 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
7085 
7086 		rx_desc++;
7087 		bi++;
7088 		i++;
7089 		if (unlikely(!i)) {
7090 			rx_desc = IGB_RX_DESC(rx_ring, 0);
7091 			bi = rx_ring->rx_buffer_info;
7092 			i -= rx_ring->count;
7093 		}
7094 
7095 		/* clear the status bits for the next_to_use descriptor */
7096 		rx_desc->wb.upper.status_error = 0;
7097 
7098 		cleaned_count--;
7099 	} while (cleaned_count);
7100 
7101 	i += rx_ring->count;
7102 
7103 	if (rx_ring->next_to_use != i) {
7104 		/* record the next descriptor to use */
7105 		rx_ring->next_to_use = i;
7106 
7107 		/* update next to alloc since we have filled the ring */
7108 		rx_ring->next_to_alloc = i;
7109 
7110 		/* Force memory writes to complete before letting h/w
7111 		 * know there are new descriptors to fetch.  (Only
7112 		 * applicable for weak-ordered memory model archs,
7113 		 * such as IA-64).
7114 		 */
7115 		wmb();
7116 		writel(i, rx_ring->tail);
7117 	}
7118 }
7119 
7120 /**
7121  * igb_mii_ioctl -
7122  * @netdev:
7123  * @ifreq:
7124  * @cmd:
7125  **/
7126 static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7127 {
7128 	struct igb_adapter *adapter = netdev_priv(netdev);
7129 	struct mii_ioctl_data *data = if_mii(ifr);
7130 
7131 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
7132 		return -EOPNOTSUPP;
7133 
7134 	switch (cmd) {
7135 	case SIOCGMIIPHY:
7136 		data->phy_id = adapter->hw.phy.addr;
7137 		break;
7138 	case SIOCGMIIREG:
7139 		if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
7140 				     &data->val_out))
7141 			return -EIO;
7142 		break;
7143 	case SIOCSMIIREG:
7144 	default:
7145 		return -EOPNOTSUPP;
7146 	}
7147 	return 0;
7148 }
7149 
7150 /**
7151  * igb_ioctl -
7152  * @netdev:
7153  * @ifreq:
7154  * @cmd:
7155  **/
7156 static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7157 {
7158 	switch (cmd) {
7159 	case SIOCGMIIPHY:
7160 	case SIOCGMIIREG:
7161 	case SIOCSMIIREG:
7162 		return igb_mii_ioctl(netdev, ifr, cmd);
7163 	case SIOCGHWTSTAMP:
7164 		return igb_ptp_get_ts_config(netdev, ifr);
7165 	case SIOCSHWTSTAMP:
7166 		return igb_ptp_set_ts_config(netdev, ifr);
7167 	default:
7168 		return -EOPNOTSUPP;
7169 	}
7170 }
7171 
7172 void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
7173 {
7174 	struct igb_adapter *adapter = hw->back;
7175 
7176 	pci_read_config_word(adapter->pdev, reg, value);
7177 }
7178 
7179 void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
7180 {
7181 	struct igb_adapter *adapter = hw->back;
7182 
7183 	pci_write_config_word(adapter->pdev, reg, *value);
7184 }
7185 
7186 s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
7187 {
7188 	struct igb_adapter *adapter = hw->back;
7189 
7190 	if (pcie_capability_read_word(adapter->pdev, reg, value))
7191 		return -E1000_ERR_CONFIG;
7192 
7193 	return 0;
7194 }
7195 
7196 s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
7197 {
7198 	struct igb_adapter *adapter = hw->back;
7199 
7200 	if (pcie_capability_write_word(adapter->pdev, reg, *value))
7201 		return -E1000_ERR_CONFIG;
7202 
7203 	return 0;
7204 }
7205 
7206 static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
7207 {
7208 	struct igb_adapter *adapter = netdev_priv(netdev);
7209 	struct e1000_hw *hw = &adapter->hw;
7210 	u32 ctrl, rctl;
7211 	bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
7212 
7213 	if (enable) {
7214 		/* enable VLAN tag insert/strip */
7215 		ctrl = rd32(E1000_CTRL);
7216 		ctrl |= E1000_CTRL_VME;
7217 		wr32(E1000_CTRL, ctrl);
7218 
7219 		/* Disable CFI check */
7220 		rctl = rd32(E1000_RCTL);
7221 		rctl &= ~E1000_RCTL_CFIEN;
7222 		wr32(E1000_RCTL, rctl);
7223 	} else {
7224 		/* disable VLAN tag insert/strip */
7225 		ctrl = rd32(E1000_CTRL);
7226 		ctrl &= ~E1000_CTRL_VME;
7227 		wr32(E1000_CTRL, ctrl);
7228 	}
7229 
7230 	igb_rlpml_set(adapter);
7231 }
7232 
7233 static int igb_vlan_rx_add_vid(struct net_device *netdev,
7234 			       __be16 proto, u16 vid)
7235 {
7236 	struct igb_adapter *adapter = netdev_priv(netdev);
7237 	struct e1000_hw *hw = &adapter->hw;
7238 	int pf_id = adapter->vfs_allocated_count;
7239 
7240 	/* attempt to add filter to vlvf array */
7241 	igb_vlvf_set(adapter, vid, true, pf_id);
7242 
7243 	/* add the filter since PF can receive vlans w/o entry in vlvf */
7244 	igb_vfta_set(hw, vid, true);
7245 
7246 	set_bit(vid, adapter->active_vlans);
7247 
7248 	return 0;
7249 }
7250 
7251 static int igb_vlan_rx_kill_vid(struct net_device *netdev,
7252 				__be16 proto, u16 vid)
7253 {
7254 	struct igb_adapter *adapter = netdev_priv(netdev);
7255 	struct e1000_hw *hw = &adapter->hw;
7256 	int pf_id = adapter->vfs_allocated_count;
7257 	s32 err;
7258 
7259 	/* remove vlan from VLVF table array */
7260 	err = igb_vlvf_set(adapter, vid, false, pf_id);
7261 
7262 	/* if vid was not present in VLVF just remove it from table */
7263 	if (err)
7264 		igb_vfta_set(hw, vid, false);
7265 
7266 	clear_bit(vid, adapter->active_vlans);
7267 
7268 	return 0;
7269 }
7270 
7271 static void igb_restore_vlan(struct igb_adapter *adapter)
7272 {
7273 	u16 vid;
7274 
7275 	igb_vlan_mode(adapter->netdev, adapter->netdev->features);
7276 
7277 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
7278 		igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
7279 }
7280 
7281 int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
7282 {
7283 	struct pci_dev *pdev = adapter->pdev;
7284 	struct e1000_mac_info *mac = &adapter->hw.mac;
7285 
7286 	mac->autoneg = 0;
7287 
7288 	/* Make sure dplx is at most 1 bit and lsb of speed is not set
7289 	 * for the switch() below to work
7290 	 */
7291 	if ((spd & 1) || (dplx & ~1))
7292 		goto err_inval;
7293 
7294 	/* Fiber NIC's only allow 1000 gbps Full duplex
7295 	 * and 100Mbps Full duplex for 100baseFx sfp
7296 	 */
7297 	if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
7298 		switch (spd + dplx) {
7299 		case SPEED_10 + DUPLEX_HALF:
7300 		case SPEED_10 + DUPLEX_FULL:
7301 		case SPEED_100 + DUPLEX_HALF:
7302 			goto err_inval;
7303 		default:
7304 			break;
7305 		}
7306 	}
7307 
7308 	switch (spd + dplx) {
7309 	case SPEED_10 + DUPLEX_HALF:
7310 		mac->forced_speed_duplex = ADVERTISE_10_HALF;
7311 		break;
7312 	case SPEED_10 + DUPLEX_FULL:
7313 		mac->forced_speed_duplex = ADVERTISE_10_FULL;
7314 		break;
7315 	case SPEED_100 + DUPLEX_HALF:
7316 		mac->forced_speed_duplex = ADVERTISE_100_HALF;
7317 		break;
7318 	case SPEED_100 + DUPLEX_FULL:
7319 		mac->forced_speed_duplex = ADVERTISE_100_FULL;
7320 		break;
7321 	case SPEED_1000 + DUPLEX_FULL:
7322 		mac->autoneg = 1;
7323 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
7324 		break;
7325 	case SPEED_1000 + DUPLEX_HALF: /* not supported */
7326 	default:
7327 		goto err_inval;
7328 	}
7329 
7330 	/* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
7331 	adapter->hw.phy.mdix = AUTO_ALL_MODES;
7332 
7333 	return 0;
7334 
7335 err_inval:
7336 	dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
7337 	return -EINVAL;
7338 }
7339 
7340 static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
7341 			  bool runtime)
7342 {
7343 	struct net_device *netdev = pci_get_drvdata(pdev);
7344 	struct igb_adapter *adapter = netdev_priv(netdev);
7345 	struct e1000_hw *hw = &adapter->hw;
7346 	u32 ctrl, rctl, status;
7347 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
7348 #ifdef CONFIG_PM
7349 	int retval = 0;
7350 #endif
7351 
7352 	netif_device_detach(netdev);
7353 
7354 	if (netif_running(netdev))
7355 		__igb_close(netdev, true);
7356 
7357 	igb_clear_interrupt_scheme(adapter);
7358 
7359 #ifdef CONFIG_PM
7360 	retval = pci_save_state(pdev);
7361 	if (retval)
7362 		return retval;
7363 #endif
7364 
7365 	status = rd32(E1000_STATUS);
7366 	if (status & E1000_STATUS_LU)
7367 		wufc &= ~E1000_WUFC_LNKC;
7368 
7369 	if (wufc) {
7370 		igb_setup_rctl(adapter);
7371 		igb_set_rx_mode(netdev);
7372 
7373 		/* turn on all-multi mode if wake on multicast is enabled */
7374 		if (wufc & E1000_WUFC_MC) {
7375 			rctl = rd32(E1000_RCTL);
7376 			rctl |= E1000_RCTL_MPE;
7377 			wr32(E1000_RCTL, rctl);
7378 		}
7379 
7380 		ctrl = rd32(E1000_CTRL);
7381 		/* advertise wake from D3Cold */
7382 		#define E1000_CTRL_ADVD3WUC 0x00100000
7383 		/* phy power management enable */
7384 		#define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
7385 		ctrl |= E1000_CTRL_ADVD3WUC;
7386 		wr32(E1000_CTRL, ctrl);
7387 
7388 		/* Allow time for pending master requests to run */
7389 		igb_disable_pcie_master(hw);
7390 
7391 		wr32(E1000_WUC, E1000_WUC_PME_EN);
7392 		wr32(E1000_WUFC, wufc);
7393 	} else {
7394 		wr32(E1000_WUC, 0);
7395 		wr32(E1000_WUFC, 0);
7396 	}
7397 
7398 	*enable_wake = wufc || adapter->en_mng_pt;
7399 	if (!*enable_wake)
7400 		igb_power_down_link(adapter);
7401 	else
7402 		igb_power_up_link(adapter);
7403 
7404 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7405 	 * would have already happened in close and is redundant.
7406 	 */
7407 	igb_release_hw_control(adapter);
7408 
7409 	pci_disable_device(pdev);
7410 
7411 	return 0;
7412 }
7413 
7414 #ifdef CONFIG_PM
7415 #ifdef CONFIG_PM_SLEEP
7416 static int igb_suspend(struct device *dev)
7417 {
7418 	int retval;
7419 	bool wake;
7420 	struct pci_dev *pdev = to_pci_dev(dev);
7421 
7422 	retval = __igb_shutdown(pdev, &wake, 0);
7423 	if (retval)
7424 		return retval;
7425 
7426 	if (wake) {
7427 		pci_prepare_to_sleep(pdev);
7428 	} else {
7429 		pci_wake_from_d3(pdev, false);
7430 		pci_set_power_state(pdev, PCI_D3hot);
7431 	}
7432 
7433 	return 0;
7434 }
7435 #endif /* CONFIG_PM_SLEEP */
7436 
7437 static int igb_resume(struct device *dev)
7438 {
7439 	struct pci_dev *pdev = to_pci_dev(dev);
7440 	struct net_device *netdev = pci_get_drvdata(pdev);
7441 	struct igb_adapter *adapter = netdev_priv(netdev);
7442 	struct e1000_hw *hw = &adapter->hw;
7443 	u32 err;
7444 
7445 	pci_set_power_state(pdev, PCI_D0);
7446 	pci_restore_state(pdev);
7447 	pci_save_state(pdev);
7448 
7449 	if (!pci_device_is_present(pdev))
7450 		return -ENODEV;
7451 	err = pci_enable_device_mem(pdev);
7452 	if (err) {
7453 		dev_err(&pdev->dev,
7454 			"igb: Cannot enable PCI device from suspend\n");
7455 		return err;
7456 	}
7457 	pci_set_master(pdev);
7458 
7459 	pci_enable_wake(pdev, PCI_D3hot, 0);
7460 	pci_enable_wake(pdev, PCI_D3cold, 0);
7461 
7462 	if (igb_init_interrupt_scheme(adapter, true)) {
7463 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
7464 		return -ENOMEM;
7465 	}
7466 
7467 	igb_reset(adapter);
7468 
7469 	/* let the f/w know that the h/w is now under the control of the
7470 	 * driver.
7471 	 */
7472 	igb_get_hw_control(adapter);
7473 
7474 	wr32(E1000_WUS, ~0);
7475 
7476 	if (netdev->flags & IFF_UP) {
7477 		rtnl_lock();
7478 		err = __igb_open(netdev, true);
7479 		rtnl_unlock();
7480 		if (err)
7481 			return err;
7482 	}
7483 
7484 	netif_device_attach(netdev);
7485 	return 0;
7486 }
7487 
7488 static int igb_runtime_idle(struct device *dev)
7489 {
7490 	struct pci_dev *pdev = to_pci_dev(dev);
7491 	struct net_device *netdev = pci_get_drvdata(pdev);
7492 	struct igb_adapter *adapter = netdev_priv(netdev);
7493 
7494 	if (!igb_has_link(adapter))
7495 		pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
7496 
7497 	return -EBUSY;
7498 }
7499 
7500 static int igb_runtime_suspend(struct device *dev)
7501 {
7502 	struct pci_dev *pdev = to_pci_dev(dev);
7503 	int retval;
7504 	bool wake;
7505 
7506 	retval = __igb_shutdown(pdev, &wake, 1);
7507 	if (retval)
7508 		return retval;
7509 
7510 	if (wake) {
7511 		pci_prepare_to_sleep(pdev);
7512 	} else {
7513 		pci_wake_from_d3(pdev, false);
7514 		pci_set_power_state(pdev, PCI_D3hot);
7515 	}
7516 
7517 	return 0;
7518 }
7519 
7520 static int igb_runtime_resume(struct device *dev)
7521 {
7522 	return igb_resume(dev);
7523 }
7524 #endif /* CONFIG_PM */
7525 
7526 static void igb_shutdown(struct pci_dev *pdev)
7527 {
7528 	bool wake;
7529 
7530 	__igb_shutdown(pdev, &wake, 0);
7531 
7532 	if (system_state == SYSTEM_POWER_OFF) {
7533 		pci_wake_from_d3(pdev, wake);
7534 		pci_set_power_state(pdev, PCI_D3hot);
7535 	}
7536 }
7537 
7538 #ifdef CONFIG_PCI_IOV
7539 static int igb_sriov_reinit(struct pci_dev *dev)
7540 {
7541 	struct net_device *netdev = pci_get_drvdata(dev);
7542 	struct igb_adapter *adapter = netdev_priv(netdev);
7543 	struct pci_dev *pdev = adapter->pdev;
7544 
7545 	rtnl_lock();
7546 
7547 	if (netif_running(netdev))
7548 		igb_close(netdev);
7549 	else
7550 		igb_reset(adapter);
7551 
7552 	igb_clear_interrupt_scheme(adapter);
7553 
7554 	igb_init_queue_configuration(adapter);
7555 
7556 	if (igb_init_interrupt_scheme(adapter, true)) {
7557 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
7558 		return -ENOMEM;
7559 	}
7560 
7561 	if (netif_running(netdev))
7562 		igb_open(netdev);
7563 
7564 	rtnl_unlock();
7565 
7566 	return 0;
7567 }
7568 
7569 static int igb_pci_disable_sriov(struct pci_dev *dev)
7570 {
7571 	int err = igb_disable_sriov(dev);
7572 
7573 	if (!err)
7574 		err = igb_sriov_reinit(dev);
7575 
7576 	return err;
7577 }
7578 
7579 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
7580 {
7581 	int err = igb_enable_sriov(dev, num_vfs);
7582 
7583 	if (err)
7584 		goto out;
7585 
7586 	err = igb_sriov_reinit(dev);
7587 	if (!err)
7588 		return num_vfs;
7589 
7590 out:
7591 	return err;
7592 }
7593 
7594 #endif
7595 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
7596 {
7597 #ifdef CONFIG_PCI_IOV
7598 	if (num_vfs == 0)
7599 		return igb_pci_disable_sriov(dev);
7600 	else
7601 		return igb_pci_enable_sriov(dev, num_vfs);
7602 #endif
7603 	return 0;
7604 }
7605 
7606 #ifdef CONFIG_NET_POLL_CONTROLLER
7607 /* Polling 'interrupt' - used by things like netconsole to send skbs
7608  * without having to re-enable interrupts. It's not called while
7609  * the interrupt routine is executing.
7610  */
7611 static void igb_netpoll(struct net_device *netdev)
7612 {
7613 	struct igb_adapter *adapter = netdev_priv(netdev);
7614 	struct e1000_hw *hw = &adapter->hw;
7615 	struct igb_q_vector *q_vector;
7616 	int i;
7617 
7618 	for (i = 0; i < adapter->num_q_vectors; i++) {
7619 		q_vector = adapter->q_vector[i];
7620 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
7621 			wr32(E1000_EIMC, q_vector->eims_value);
7622 		else
7623 			igb_irq_disable(adapter);
7624 		napi_schedule(&q_vector->napi);
7625 	}
7626 }
7627 #endif /* CONFIG_NET_POLL_CONTROLLER */
7628 
7629 /**
7630  *  igb_io_error_detected - called when PCI error is detected
7631  *  @pdev: Pointer to PCI device
7632  *  @state: The current pci connection state
7633  *
7634  *  This function is called after a PCI bus error affecting
7635  *  this device has been detected.
7636  **/
7637 static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
7638 					      pci_channel_state_t state)
7639 {
7640 	struct net_device *netdev = pci_get_drvdata(pdev);
7641 	struct igb_adapter *adapter = netdev_priv(netdev);
7642 
7643 	netif_device_detach(netdev);
7644 
7645 	if (state == pci_channel_io_perm_failure)
7646 		return PCI_ERS_RESULT_DISCONNECT;
7647 
7648 	if (netif_running(netdev))
7649 		igb_down(adapter);
7650 	pci_disable_device(pdev);
7651 
7652 	/* Request a slot slot reset. */
7653 	return PCI_ERS_RESULT_NEED_RESET;
7654 }
7655 
7656 /**
7657  *  igb_io_slot_reset - called after the pci bus has been reset.
7658  *  @pdev: Pointer to PCI device
7659  *
7660  *  Restart the card from scratch, as if from a cold-boot. Implementation
7661  *  resembles the first-half of the igb_resume routine.
7662  **/
7663 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
7664 {
7665 	struct net_device *netdev = pci_get_drvdata(pdev);
7666 	struct igb_adapter *adapter = netdev_priv(netdev);
7667 	struct e1000_hw *hw = &adapter->hw;
7668 	pci_ers_result_t result;
7669 	int err;
7670 
7671 	if (pci_enable_device_mem(pdev)) {
7672 		dev_err(&pdev->dev,
7673 			"Cannot re-enable PCI device after reset.\n");
7674 		result = PCI_ERS_RESULT_DISCONNECT;
7675 	} else {
7676 		pci_set_master(pdev);
7677 		pci_restore_state(pdev);
7678 		pci_save_state(pdev);
7679 
7680 		pci_enable_wake(pdev, PCI_D3hot, 0);
7681 		pci_enable_wake(pdev, PCI_D3cold, 0);
7682 
7683 		igb_reset(adapter);
7684 		wr32(E1000_WUS, ~0);
7685 		result = PCI_ERS_RESULT_RECOVERED;
7686 	}
7687 
7688 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7689 	if (err) {
7690 		dev_err(&pdev->dev,
7691 			"pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7692 			err);
7693 		/* non-fatal, continue */
7694 	}
7695 
7696 	return result;
7697 }
7698 
7699 /**
7700  *  igb_io_resume - called when traffic can start flowing again.
7701  *  @pdev: Pointer to PCI device
7702  *
7703  *  This callback is called when the error recovery driver tells us that
7704  *  its OK to resume normal operation. Implementation resembles the
7705  *  second-half of the igb_resume routine.
7706  */
7707 static void igb_io_resume(struct pci_dev *pdev)
7708 {
7709 	struct net_device *netdev = pci_get_drvdata(pdev);
7710 	struct igb_adapter *adapter = netdev_priv(netdev);
7711 
7712 	if (netif_running(netdev)) {
7713 		if (igb_up(adapter)) {
7714 			dev_err(&pdev->dev, "igb_up failed after reset\n");
7715 			return;
7716 		}
7717 	}
7718 
7719 	netif_device_attach(netdev);
7720 
7721 	/* let the f/w know that the h/w is now under the control of the
7722 	 * driver.
7723 	 */
7724 	igb_get_hw_control(adapter);
7725 }
7726 
7727 static void igb_rar_set_qsel(struct igb_adapter *adapter, u8 *addr, u32 index,
7728 			     u8 qsel)
7729 {
7730 	u32 rar_low, rar_high;
7731 	struct e1000_hw *hw = &adapter->hw;
7732 
7733 	/* HW expects these in little endian so we reverse the byte order
7734 	 * from network order (big endian) to little endian
7735 	 */
7736 	rar_low = ((u32) addr[0] | ((u32) addr[1] << 8) |
7737 		   ((u32) addr[2] << 16) | ((u32) addr[3] << 24));
7738 	rar_high = ((u32) addr[4] | ((u32) addr[5] << 8));
7739 
7740 	/* Indicate to hardware the Address is Valid. */
7741 	rar_high |= E1000_RAH_AV;
7742 
7743 	if (hw->mac.type == e1000_82575)
7744 		rar_high |= E1000_RAH_POOL_1 * qsel;
7745 	else
7746 		rar_high |= E1000_RAH_POOL_1 << qsel;
7747 
7748 	wr32(E1000_RAL(index), rar_low);
7749 	wrfl();
7750 	wr32(E1000_RAH(index), rar_high);
7751 	wrfl();
7752 }
7753 
7754 static int igb_set_vf_mac(struct igb_adapter *adapter,
7755 			  int vf, unsigned char *mac_addr)
7756 {
7757 	struct e1000_hw *hw = &adapter->hw;
7758 	/* VF MAC addresses start at end of receive addresses and moves
7759 	 * towards the first, as a result a collision should not be possible
7760 	 */
7761 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
7762 
7763 	memcpy(adapter->vf_data[vf].vf_mac_addresses, mac_addr, ETH_ALEN);
7764 
7765 	igb_rar_set_qsel(adapter, mac_addr, rar_entry, vf);
7766 
7767 	return 0;
7768 }
7769 
7770 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
7771 {
7772 	struct igb_adapter *adapter = netdev_priv(netdev);
7773 	if (!is_valid_ether_addr(mac) || (vf >= adapter->vfs_allocated_count))
7774 		return -EINVAL;
7775 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
7776 	dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n", mac, vf);
7777 	dev_info(&adapter->pdev->dev,
7778 		 "Reload the VF driver to make this change effective.");
7779 	if (test_bit(__IGB_DOWN, &adapter->state)) {
7780 		dev_warn(&adapter->pdev->dev,
7781 			 "The VF MAC address has been set, but the PF device is not up.\n");
7782 		dev_warn(&adapter->pdev->dev,
7783 			 "Bring the PF device up before attempting to use the VF device.\n");
7784 	}
7785 	return igb_set_vf_mac(adapter, vf, mac);
7786 }
7787 
7788 static int igb_link_mbps(int internal_link_speed)
7789 {
7790 	switch (internal_link_speed) {
7791 	case SPEED_100:
7792 		return 100;
7793 	case SPEED_1000:
7794 		return 1000;
7795 	default:
7796 		return 0;
7797 	}
7798 }
7799 
7800 static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
7801 				  int link_speed)
7802 {
7803 	int rf_dec, rf_int;
7804 	u32 bcnrc_val;
7805 
7806 	if (tx_rate != 0) {
7807 		/* Calculate the rate factor values to set */
7808 		rf_int = link_speed / tx_rate;
7809 		rf_dec = (link_speed - (rf_int * tx_rate));
7810 		rf_dec = (rf_dec * (1 << E1000_RTTBCNRC_RF_INT_SHIFT)) /
7811 			 tx_rate;
7812 
7813 		bcnrc_val = E1000_RTTBCNRC_RS_ENA;
7814 		bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
7815 			      E1000_RTTBCNRC_RF_INT_MASK);
7816 		bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
7817 	} else {
7818 		bcnrc_val = 0;
7819 	}
7820 
7821 	wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
7822 	/* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
7823 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
7824 	 */
7825 	wr32(E1000_RTTBCNRM, 0x14);
7826 	wr32(E1000_RTTBCNRC, bcnrc_val);
7827 }
7828 
7829 static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
7830 {
7831 	int actual_link_speed, i;
7832 	bool reset_rate = false;
7833 
7834 	/* VF TX rate limit was not set or not supported */
7835 	if ((adapter->vf_rate_link_speed == 0) ||
7836 	    (adapter->hw.mac.type != e1000_82576))
7837 		return;
7838 
7839 	actual_link_speed = igb_link_mbps(adapter->link_speed);
7840 	if (actual_link_speed != adapter->vf_rate_link_speed) {
7841 		reset_rate = true;
7842 		adapter->vf_rate_link_speed = 0;
7843 		dev_info(&adapter->pdev->dev,
7844 			 "Link speed has been changed. VF Transmit rate is disabled\n");
7845 	}
7846 
7847 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
7848 		if (reset_rate)
7849 			adapter->vf_data[i].tx_rate = 0;
7850 
7851 		igb_set_vf_rate_limit(&adapter->hw, i,
7852 				      adapter->vf_data[i].tx_rate,
7853 				      actual_link_speed);
7854 	}
7855 }
7856 
7857 static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
7858 			     int min_tx_rate, int max_tx_rate)
7859 {
7860 	struct igb_adapter *adapter = netdev_priv(netdev);
7861 	struct e1000_hw *hw = &adapter->hw;
7862 	int actual_link_speed;
7863 
7864 	if (hw->mac.type != e1000_82576)
7865 		return -EOPNOTSUPP;
7866 
7867 	if (min_tx_rate)
7868 		return -EINVAL;
7869 
7870 	actual_link_speed = igb_link_mbps(adapter->link_speed);
7871 	if ((vf >= adapter->vfs_allocated_count) ||
7872 	    (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
7873 	    (max_tx_rate < 0) ||
7874 	    (max_tx_rate > actual_link_speed))
7875 		return -EINVAL;
7876 
7877 	adapter->vf_rate_link_speed = actual_link_speed;
7878 	adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
7879 	igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
7880 
7881 	return 0;
7882 }
7883 
7884 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
7885 				   bool setting)
7886 {
7887 	struct igb_adapter *adapter = netdev_priv(netdev);
7888 	struct e1000_hw *hw = &adapter->hw;
7889 	u32 reg_val, reg_offset;
7890 
7891 	if (!adapter->vfs_allocated_count)
7892 		return -EOPNOTSUPP;
7893 
7894 	if (vf >= adapter->vfs_allocated_count)
7895 		return -EINVAL;
7896 
7897 	reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
7898 	reg_val = rd32(reg_offset);
7899 	if (setting)
7900 		reg_val |= ((1 << vf) |
7901 			    (1 << (vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT)));
7902 	else
7903 		reg_val &= ~((1 << vf) |
7904 			     (1 << (vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT)));
7905 	wr32(reg_offset, reg_val);
7906 
7907 	adapter->vf_data[vf].spoofchk_enabled = setting;
7908 	return 0;
7909 }
7910 
7911 static int igb_ndo_get_vf_config(struct net_device *netdev,
7912 				 int vf, struct ifla_vf_info *ivi)
7913 {
7914 	struct igb_adapter *adapter = netdev_priv(netdev);
7915 	if (vf >= adapter->vfs_allocated_count)
7916 		return -EINVAL;
7917 	ivi->vf = vf;
7918 	memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
7919 	ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
7920 	ivi->min_tx_rate = 0;
7921 	ivi->vlan = adapter->vf_data[vf].pf_vlan;
7922 	ivi->qos = adapter->vf_data[vf].pf_qos;
7923 	ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
7924 	return 0;
7925 }
7926 
7927 static void igb_vmm_control(struct igb_adapter *adapter)
7928 {
7929 	struct e1000_hw *hw = &adapter->hw;
7930 	u32 reg;
7931 
7932 	switch (hw->mac.type) {
7933 	case e1000_82575:
7934 	case e1000_i210:
7935 	case e1000_i211:
7936 	case e1000_i354:
7937 	default:
7938 		/* replication is not supported for 82575 */
7939 		return;
7940 	case e1000_82576:
7941 		/* notify HW that the MAC is adding vlan tags */
7942 		reg = rd32(E1000_DTXCTL);
7943 		reg |= E1000_DTXCTL_VLAN_ADDED;
7944 		wr32(E1000_DTXCTL, reg);
7945 		/* Fall through */
7946 	case e1000_82580:
7947 		/* enable replication vlan tag stripping */
7948 		reg = rd32(E1000_RPLOLR);
7949 		reg |= E1000_RPLOLR_STRVLAN;
7950 		wr32(E1000_RPLOLR, reg);
7951 		/* Fall through */
7952 	case e1000_i350:
7953 		/* none of the above registers are supported by i350 */
7954 		break;
7955 	}
7956 
7957 	if (adapter->vfs_allocated_count) {
7958 		igb_vmdq_set_loopback_pf(hw, true);
7959 		igb_vmdq_set_replication_pf(hw, true);
7960 		igb_vmdq_set_anti_spoofing_pf(hw, true,
7961 					      adapter->vfs_allocated_count);
7962 	} else {
7963 		igb_vmdq_set_loopback_pf(hw, false);
7964 		igb_vmdq_set_replication_pf(hw, false);
7965 	}
7966 }
7967 
7968 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
7969 {
7970 	struct e1000_hw *hw = &adapter->hw;
7971 	u32 dmac_thr;
7972 	u16 hwm;
7973 
7974 	if (hw->mac.type > e1000_82580) {
7975 		if (adapter->flags & IGB_FLAG_DMAC) {
7976 			u32 reg;
7977 
7978 			/* force threshold to 0. */
7979 			wr32(E1000_DMCTXTH, 0);
7980 
7981 			/* DMA Coalescing high water mark needs to be greater
7982 			 * than the Rx threshold. Set hwm to PBA - max frame
7983 			 * size in 16B units, capping it at PBA - 6KB.
7984 			 */
7985 			hwm = 64 * pba - adapter->max_frame_size / 16;
7986 			if (hwm < 64 * (pba - 6))
7987 				hwm = 64 * (pba - 6);
7988 			reg = rd32(E1000_FCRTC);
7989 			reg &= ~E1000_FCRTC_RTH_COAL_MASK;
7990 			reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
7991 				& E1000_FCRTC_RTH_COAL_MASK);
7992 			wr32(E1000_FCRTC, reg);
7993 
7994 			/* Set the DMA Coalescing Rx threshold to PBA - 2 * max
7995 			 * frame size, capping it at PBA - 10KB.
7996 			 */
7997 			dmac_thr = pba - adapter->max_frame_size / 512;
7998 			if (dmac_thr < pba - 10)
7999 				dmac_thr = pba - 10;
8000 			reg = rd32(E1000_DMACR);
8001 			reg &= ~E1000_DMACR_DMACTHR_MASK;
8002 			reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
8003 				& E1000_DMACR_DMACTHR_MASK);
8004 
8005 			/* transition to L0x or L1 if available..*/
8006 			reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
8007 
8008 			/* watchdog timer= +-1000 usec in 32usec intervals */
8009 			reg |= (1000 >> 5);
8010 
8011 			/* Disable BMC-to-OS Watchdog Enable */
8012 			if (hw->mac.type != e1000_i354)
8013 				reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
8014 
8015 			wr32(E1000_DMACR, reg);
8016 
8017 			/* no lower threshold to disable
8018 			 * coalescing(smart fifb)-UTRESH=0
8019 			 */
8020 			wr32(E1000_DMCRTRH, 0);
8021 
8022 			reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
8023 
8024 			wr32(E1000_DMCTLX, reg);
8025 
8026 			/* free space in tx packet buffer to wake from
8027 			 * DMA coal
8028 			 */
8029 			wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
8030 			     (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
8031 
8032 			/* make low power state decision controlled
8033 			 * by DMA coal
8034 			 */
8035 			reg = rd32(E1000_PCIEMISC);
8036 			reg &= ~E1000_PCIEMISC_LX_DECISION;
8037 			wr32(E1000_PCIEMISC, reg);
8038 		} /* endif adapter->dmac is not disabled */
8039 	} else if (hw->mac.type == e1000_82580) {
8040 		u32 reg = rd32(E1000_PCIEMISC);
8041 
8042 		wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
8043 		wr32(E1000_DMACR, 0);
8044 	}
8045 }
8046 
8047 /**
8048  *  igb_read_i2c_byte - Reads 8 bit word over I2C
8049  *  @hw: pointer to hardware structure
8050  *  @byte_offset: byte offset to read
8051  *  @dev_addr: device address
8052  *  @data: value read
8053  *
8054  *  Performs byte read operation over I2C interface at
8055  *  a specified device address.
8056  **/
8057 s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
8058 		      u8 dev_addr, u8 *data)
8059 {
8060 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
8061 	struct i2c_client *this_client = adapter->i2c_client;
8062 	s32 status;
8063 	u16 swfw_mask = 0;
8064 
8065 	if (!this_client)
8066 		return E1000_ERR_I2C;
8067 
8068 	swfw_mask = E1000_SWFW_PHY0_SM;
8069 
8070 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
8071 		return E1000_ERR_SWFW_SYNC;
8072 
8073 	status = i2c_smbus_read_byte_data(this_client, byte_offset);
8074 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
8075 
8076 	if (status < 0)
8077 		return E1000_ERR_I2C;
8078 	else {
8079 		*data = status;
8080 		return 0;
8081 	}
8082 }
8083 
8084 /**
8085  *  igb_write_i2c_byte - Writes 8 bit word over I2C
8086  *  @hw: pointer to hardware structure
8087  *  @byte_offset: byte offset to write
8088  *  @dev_addr: device address
8089  *  @data: value to write
8090  *
8091  *  Performs byte write operation over I2C interface at
8092  *  a specified device address.
8093  **/
8094 s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
8095 		       u8 dev_addr, u8 data)
8096 {
8097 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
8098 	struct i2c_client *this_client = adapter->i2c_client;
8099 	s32 status;
8100 	u16 swfw_mask = E1000_SWFW_PHY0_SM;
8101 
8102 	if (!this_client)
8103 		return E1000_ERR_I2C;
8104 
8105 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
8106 		return E1000_ERR_SWFW_SYNC;
8107 	status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
8108 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
8109 
8110 	if (status)
8111 		return E1000_ERR_I2C;
8112 	else
8113 		return 0;
8114 
8115 }
8116 
8117 int igb_reinit_queues(struct igb_adapter *adapter)
8118 {
8119 	struct net_device *netdev = adapter->netdev;
8120 	struct pci_dev *pdev = adapter->pdev;
8121 	int err = 0;
8122 
8123 	if (netif_running(netdev))
8124 		igb_close(netdev);
8125 
8126 	igb_reset_interrupt_capability(adapter);
8127 
8128 	if (igb_init_interrupt_scheme(adapter, true)) {
8129 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8130 		return -ENOMEM;
8131 	}
8132 
8133 	if (netif_running(netdev))
8134 		err = igb_open(netdev);
8135 
8136 	return err;
8137 }
8138 /* igb_main.c */
8139