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_tx_stop_all_queues(netdev);
1780 
1781 	/* disable transmits in the hardware */
1782 	tctl = rd32(E1000_TCTL);
1783 	tctl &= ~E1000_TCTL_EN;
1784 	wr32(E1000_TCTL, tctl);
1785 	/* flush both disables and wait for them to finish */
1786 	wrfl();
1787 	usleep_range(10000, 11000);
1788 
1789 	igb_irq_disable(adapter);
1790 
1791 	adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
1792 
1793 	for (i = 0; i < adapter->num_q_vectors; i++) {
1794 		if (adapter->q_vector[i]) {
1795 			napi_synchronize(&adapter->q_vector[i]->napi);
1796 			napi_disable(&adapter->q_vector[i]->napi);
1797 		}
1798 	}
1799 
1800 
1801 	del_timer_sync(&adapter->watchdog_timer);
1802 	del_timer_sync(&adapter->phy_info_timer);
1803 
1804 	netif_carrier_off(netdev);
1805 
1806 	/* record the stats before reset*/
1807 	spin_lock(&adapter->stats64_lock);
1808 	igb_update_stats(adapter, &adapter->stats64);
1809 	spin_unlock(&adapter->stats64_lock);
1810 
1811 	adapter->link_speed = 0;
1812 	adapter->link_duplex = 0;
1813 
1814 	if (!pci_channel_offline(adapter->pdev))
1815 		igb_reset(adapter);
1816 	igb_clean_all_tx_rings(adapter);
1817 	igb_clean_all_rx_rings(adapter);
1818 #ifdef CONFIG_IGB_DCA
1819 
1820 	/* since we reset the hardware DCA settings were cleared */
1821 	igb_setup_dca(adapter);
1822 #endif
1823 }
1824 
1825 void igb_reinit_locked(struct igb_adapter *adapter)
1826 {
1827 	WARN_ON(in_interrupt());
1828 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
1829 		usleep_range(1000, 2000);
1830 	igb_down(adapter);
1831 	igb_up(adapter);
1832 	clear_bit(__IGB_RESETTING, &adapter->state);
1833 }
1834 
1835 /** igb_enable_mas - Media Autosense re-enable after swap
1836  *
1837  * @adapter: adapter struct
1838  **/
1839 static s32 igb_enable_mas(struct igb_adapter *adapter)
1840 {
1841 	struct e1000_hw *hw = &adapter->hw;
1842 	u32 connsw;
1843 	s32 ret_val = 0;
1844 
1845 	connsw = rd32(E1000_CONNSW);
1846 	if (!(hw->phy.media_type == e1000_media_type_copper))
1847 		return ret_val;
1848 
1849 	/* configure for SerDes media detect */
1850 	if (!(connsw & E1000_CONNSW_SERDESD)) {
1851 		connsw |= E1000_CONNSW_ENRGSRC;
1852 		connsw |= E1000_CONNSW_AUTOSENSE_EN;
1853 		wr32(E1000_CONNSW, connsw);
1854 		wrfl();
1855 	} else if (connsw & E1000_CONNSW_SERDESD) {
1856 		/* already SerDes, no need to enable anything */
1857 		return ret_val;
1858 	} else {
1859 		netdev_info(adapter->netdev,
1860 			"MAS: Unable to configure feature, disabling..\n");
1861 		adapter->flags &= ~IGB_FLAG_MAS_ENABLE;
1862 	}
1863 	return ret_val;
1864 }
1865 
1866 void igb_reset(struct igb_adapter *adapter)
1867 {
1868 	struct pci_dev *pdev = adapter->pdev;
1869 	struct e1000_hw *hw = &adapter->hw;
1870 	struct e1000_mac_info *mac = &hw->mac;
1871 	struct e1000_fc_info *fc = &hw->fc;
1872 	u32 pba = 0, tx_space, min_tx_space, min_rx_space, hwm;
1873 
1874 	/* Repartition Pba for greater than 9k mtu
1875 	 * To take effect CTRL.RST is required.
1876 	 */
1877 	switch (mac->type) {
1878 	case e1000_i350:
1879 	case e1000_i354:
1880 	case e1000_82580:
1881 		pba = rd32(E1000_RXPBS);
1882 		pba = igb_rxpbs_adjust_82580(pba);
1883 		break;
1884 	case e1000_82576:
1885 		pba = rd32(E1000_RXPBS);
1886 		pba &= E1000_RXPBS_SIZE_MASK_82576;
1887 		break;
1888 	case e1000_82575:
1889 	case e1000_i210:
1890 	case e1000_i211:
1891 	default:
1892 		pba = E1000_PBA_34K;
1893 		break;
1894 	}
1895 
1896 	if ((adapter->max_frame_size > ETH_FRAME_LEN + ETH_FCS_LEN) &&
1897 	    (mac->type < e1000_82576)) {
1898 		/* adjust PBA for jumbo frames */
1899 		wr32(E1000_PBA, pba);
1900 
1901 		/* To maintain wire speed transmits, the Tx FIFO should be
1902 		 * large enough to accommodate two full transmit packets,
1903 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
1904 		 * the Rx FIFO should be large enough to accommodate at least
1905 		 * one full receive packet and is similarly rounded up and
1906 		 * expressed in KB.
1907 		 */
1908 		pba = rd32(E1000_PBA);
1909 		/* upper 16 bits has Tx packet buffer allocation size in KB */
1910 		tx_space = pba >> 16;
1911 		/* lower 16 bits has Rx packet buffer allocation size in KB */
1912 		pba &= 0xffff;
1913 		/* the Tx fifo also stores 16 bytes of information about the Tx
1914 		 * but don't include ethernet FCS because hardware appends it
1915 		 */
1916 		min_tx_space = (adapter->max_frame_size +
1917 				sizeof(union e1000_adv_tx_desc) -
1918 				ETH_FCS_LEN) * 2;
1919 		min_tx_space = ALIGN(min_tx_space, 1024);
1920 		min_tx_space >>= 10;
1921 		/* software strips receive CRC, so leave room for it */
1922 		min_rx_space = adapter->max_frame_size;
1923 		min_rx_space = ALIGN(min_rx_space, 1024);
1924 		min_rx_space >>= 10;
1925 
1926 		/* If current Tx allocation is less than the min Tx FIFO size,
1927 		 * and the min Tx FIFO size is less than the current Rx FIFO
1928 		 * allocation, take space away from current Rx allocation
1929 		 */
1930 		if (tx_space < min_tx_space &&
1931 		    ((min_tx_space - tx_space) < pba)) {
1932 			pba = pba - (min_tx_space - tx_space);
1933 
1934 			/* if short on Rx space, Rx wins and must trump Tx
1935 			 * adjustment
1936 			 */
1937 			if (pba < min_rx_space)
1938 				pba = min_rx_space;
1939 		}
1940 		wr32(E1000_PBA, pba);
1941 	}
1942 
1943 	/* flow control settings */
1944 	/* The high water mark must be low enough to fit one full frame
1945 	 * (or the size used for early receive) above it in the Rx FIFO.
1946 	 * Set it to the lower of:
1947 	 * - 90% of the Rx FIFO size, or
1948 	 * - the full Rx FIFO size minus one full frame
1949 	 */
1950 	hwm = min(((pba << 10) * 9 / 10),
1951 			((pba << 10) - 2 * adapter->max_frame_size));
1952 
1953 	fc->high_water = hwm & 0xFFFFFFF0;	/* 16-byte granularity */
1954 	fc->low_water = fc->high_water - 16;
1955 	fc->pause_time = 0xFFFF;
1956 	fc->send_xon = 1;
1957 	fc->current_mode = fc->requested_mode;
1958 
1959 	/* disable receive for all VFs and wait one second */
1960 	if (adapter->vfs_allocated_count) {
1961 		int i;
1962 
1963 		for (i = 0 ; i < adapter->vfs_allocated_count; i++)
1964 			adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
1965 
1966 		/* ping all the active vfs to let them know we are going down */
1967 		igb_ping_all_vfs(adapter);
1968 
1969 		/* disable transmits and receives */
1970 		wr32(E1000_VFRE, 0);
1971 		wr32(E1000_VFTE, 0);
1972 	}
1973 
1974 	/* Allow time for pending master requests to run */
1975 	hw->mac.ops.reset_hw(hw);
1976 	wr32(E1000_WUC, 0);
1977 
1978 	if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
1979 		/* need to resetup here after media swap */
1980 		adapter->ei.get_invariants(hw);
1981 		adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
1982 	}
1983 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
1984 		if (igb_enable_mas(adapter))
1985 			dev_err(&pdev->dev,
1986 				"Error enabling Media Auto Sense\n");
1987 	}
1988 	if (hw->mac.ops.init_hw(hw))
1989 		dev_err(&pdev->dev, "Hardware Error\n");
1990 
1991 	/* Flow control settings reset on hardware reset, so guarantee flow
1992 	 * control is off when forcing speed.
1993 	 */
1994 	if (!hw->mac.autoneg)
1995 		igb_force_mac_fc(hw);
1996 
1997 	igb_init_dmac(adapter, pba);
1998 #ifdef CONFIG_IGB_HWMON
1999 	/* Re-initialize the thermal sensor on i350 devices. */
2000 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
2001 		if (mac->type == e1000_i350 && hw->bus.func == 0) {
2002 			/* If present, re-initialize the external thermal sensor
2003 			 * interface.
2004 			 */
2005 			if (adapter->ets)
2006 				mac->ops.init_thermal_sensor_thresh(hw);
2007 		}
2008 	}
2009 #endif
2010 	/* Re-establish EEE setting */
2011 	if (hw->phy.media_type == e1000_media_type_copper) {
2012 		switch (mac->type) {
2013 		case e1000_i350:
2014 		case e1000_i210:
2015 		case e1000_i211:
2016 			igb_set_eee_i350(hw, true, true);
2017 			break;
2018 		case e1000_i354:
2019 			igb_set_eee_i354(hw, true, true);
2020 			break;
2021 		default:
2022 			break;
2023 		}
2024 	}
2025 	if (!netif_running(adapter->netdev))
2026 		igb_power_down_link(adapter);
2027 
2028 	igb_update_mng_vlan(adapter);
2029 
2030 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2031 	wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2032 
2033 	/* Re-enable PTP, where applicable. */
2034 	igb_ptp_reset(adapter);
2035 
2036 	igb_get_phy_info(hw);
2037 }
2038 
2039 static netdev_features_t igb_fix_features(struct net_device *netdev,
2040 	netdev_features_t features)
2041 {
2042 	/* Since there is no support for separate Rx/Tx vlan accel
2043 	 * enable/disable make sure Tx flag is always in same state as Rx.
2044 	 */
2045 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2046 		features |= NETIF_F_HW_VLAN_CTAG_TX;
2047 	else
2048 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2049 
2050 	return features;
2051 }
2052 
2053 static int igb_set_features(struct net_device *netdev,
2054 	netdev_features_t features)
2055 {
2056 	netdev_features_t changed = netdev->features ^ features;
2057 	struct igb_adapter *adapter = netdev_priv(netdev);
2058 
2059 	if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2060 		igb_vlan_mode(netdev, features);
2061 
2062 	if (!(changed & NETIF_F_RXALL))
2063 		return 0;
2064 
2065 	netdev->features = features;
2066 
2067 	if (netif_running(netdev))
2068 		igb_reinit_locked(adapter);
2069 	else
2070 		igb_reset(adapter);
2071 
2072 	return 0;
2073 }
2074 
2075 static const struct net_device_ops igb_netdev_ops = {
2076 	.ndo_open		= igb_open,
2077 	.ndo_stop		= igb_close,
2078 	.ndo_start_xmit		= igb_xmit_frame,
2079 	.ndo_get_stats64	= igb_get_stats64,
2080 	.ndo_set_rx_mode	= igb_set_rx_mode,
2081 	.ndo_set_mac_address	= igb_set_mac,
2082 	.ndo_change_mtu		= igb_change_mtu,
2083 	.ndo_do_ioctl		= igb_ioctl,
2084 	.ndo_tx_timeout		= igb_tx_timeout,
2085 	.ndo_validate_addr	= eth_validate_addr,
2086 	.ndo_vlan_rx_add_vid	= igb_vlan_rx_add_vid,
2087 	.ndo_vlan_rx_kill_vid	= igb_vlan_rx_kill_vid,
2088 	.ndo_set_vf_mac		= igb_ndo_set_vf_mac,
2089 	.ndo_set_vf_vlan	= igb_ndo_set_vf_vlan,
2090 	.ndo_set_vf_rate	= igb_ndo_set_vf_bw,
2091 	.ndo_set_vf_spoofchk	= igb_ndo_set_vf_spoofchk,
2092 	.ndo_get_vf_config	= igb_ndo_get_vf_config,
2093 #ifdef CONFIG_NET_POLL_CONTROLLER
2094 	.ndo_poll_controller	= igb_netpoll,
2095 #endif
2096 	.ndo_fix_features	= igb_fix_features,
2097 	.ndo_set_features	= igb_set_features,
2098 };
2099 
2100 /**
2101  * igb_set_fw_version - Configure version string for ethtool
2102  * @adapter: adapter struct
2103  **/
2104 void igb_set_fw_version(struct igb_adapter *adapter)
2105 {
2106 	struct e1000_hw *hw = &adapter->hw;
2107 	struct e1000_fw_version fw;
2108 
2109 	igb_get_fw_version(hw, &fw);
2110 
2111 	switch (hw->mac.type) {
2112 	case e1000_i210:
2113 	case e1000_i211:
2114 		if (!(igb_get_flash_presence_i210(hw))) {
2115 			snprintf(adapter->fw_version,
2116 				 sizeof(adapter->fw_version),
2117 				 "%2d.%2d-%d",
2118 				 fw.invm_major, fw.invm_minor,
2119 				 fw.invm_img_type);
2120 			break;
2121 		}
2122 		/* fall through */
2123 	default:
2124 		/* if option is rom valid, display its version too */
2125 		if (fw.or_valid) {
2126 			snprintf(adapter->fw_version,
2127 				 sizeof(adapter->fw_version),
2128 				 "%d.%d, 0x%08x, %d.%d.%d",
2129 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2130 				 fw.or_major, fw.or_build, fw.or_patch);
2131 		/* no option rom */
2132 		} else if (fw.etrack_id != 0X0000) {
2133 			snprintf(adapter->fw_version,
2134 			    sizeof(adapter->fw_version),
2135 			    "%d.%d, 0x%08x",
2136 			    fw.eep_major, fw.eep_minor, fw.etrack_id);
2137 		} else {
2138 		snprintf(adapter->fw_version,
2139 		    sizeof(adapter->fw_version),
2140 		    "%d.%d.%d",
2141 		    fw.eep_major, fw.eep_minor, fw.eep_build);
2142 		}
2143 		break;
2144 	}
2145 }
2146 
2147 /**
2148  * igb_init_mas - init Media Autosense feature if enabled in the NVM
2149  *
2150  * @adapter: adapter struct
2151  **/
2152 static void igb_init_mas(struct igb_adapter *adapter)
2153 {
2154 	struct e1000_hw *hw = &adapter->hw;
2155 	u16 eeprom_data;
2156 
2157 	hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
2158 	switch (hw->bus.func) {
2159 	case E1000_FUNC_0:
2160 		if (eeprom_data & IGB_MAS_ENABLE_0) {
2161 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2162 			netdev_info(adapter->netdev,
2163 				"MAS: Enabling Media Autosense for port %d\n",
2164 				hw->bus.func);
2165 		}
2166 		break;
2167 	case E1000_FUNC_1:
2168 		if (eeprom_data & IGB_MAS_ENABLE_1) {
2169 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2170 			netdev_info(adapter->netdev,
2171 				"MAS: Enabling Media Autosense for port %d\n",
2172 				hw->bus.func);
2173 		}
2174 		break;
2175 	case E1000_FUNC_2:
2176 		if (eeprom_data & IGB_MAS_ENABLE_2) {
2177 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2178 			netdev_info(adapter->netdev,
2179 				"MAS: Enabling Media Autosense for port %d\n",
2180 				hw->bus.func);
2181 		}
2182 		break;
2183 	case E1000_FUNC_3:
2184 		if (eeprom_data & IGB_MAS_ENABLE_3) {
2185 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2186 			netdev_info(adapter->netdev,
2187 				"MAS: Enabling Media Autosense for port %d\n",
2188 				hw->bus.func);
2189 		}
2190 		break;
2191 	default:
2192 		/* Shouldn't get here */
2193 		netdev_err(adapter->netdev,
2194 			"MAS: Invalid port configuration, returning\n");
2195 		break;
2196 	}
2197 }
2198 
2199 /**
2200  *  igb_init_i2c - Init I2C interface
2201  *  @adapter: pointer to adapter structure
2202  **/
2203 static s32 igb_init_i2c(struct igb_adapter *adapter)
2204 {
2205 	s32 status = 0;
2206 
2207 	/* I2C interface supported on i350 devices */
2208 	if (adapter->hw.mac.type != e1000_i350)
2209 		return 0;
2210 
2211 	/* Initialize the i2c bus which is controlled by the registers.
2212 	 * This bus will use the i2c_algo_bit structue that implements
2213 	 * the protocol through toggling of the 4 bits in the register.
2214 	 */
2215 	adapter->i2c_adap.owner = THIS_MODULE;
2216 	adapter->i2c_algo = igb_i2c_algo;
2217 	adapter->i2c_algo.data = adapter;
2218 	adapter->i2c_adap.algo_data = &adapter->i2c_algo;
2219 	adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
2220 	strlcpy(adapter->i2c_adap.name, "igb BB",
2221 		sizeof(adapter->i2c_adap.name));
2222 	status = i2c_bit_add_bus(&adapter->i2c_adap);
2223 	return status;
2224 }
2225 
2226 /**
2227  *  igb_probe - Device Initialization Routine
2228  *  @pdev: PCI device information struct
2229  *  @ent: entry in igb_pci_tbl
2230  *
2231  *  Returns 0 on success, negative on failure
2232  *
2233  *  igb_probe initializes an adapter identified by a pci_dev structure.
2234  *  The OS initialization, configuring of the adapter private structure,
2235  *  and a hardware reset occur.
2236  **/
2237 static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2238 {
2239 	struct net_device *netdev;
2240 	struct igb_adapter *adapter;
2241 	struct e1000_hw *hw;
2242 	u16 eeprom_data = 0;
2243 	s32 ret_val;
2244 	static int global_quad_port_a; /* global quad port a indication */
2245 	const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
2246 	int err, pci_using_dac;
2247 	u8 part_str[E1000_PBANUM_LENGTH];
2248 
2249 	/* Catch broken hardware that put the wrong VF device ID in
2250 	 * the PCIe SR-IOV capability.
2251 	 */
2252 	if (pdev->is_virtfn) {
2253 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
2254 			pci_name(pdev), pdev->vendor, pdev->device);
2255 		return -EINVAL;
2256 	}
2257 
2258 	err = pci_enable_device_mem(pdev);
2259 	if (err)
2260 		return err;
2261 
2262 	pci_using_dac = 0;
2263 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2264 	if (!err) {
2265 		pci_using_dac = 1;
2266 	} else {
2267 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
2268 		if (err) {
2269 			dev_err(&pdev->dev,
2270 				"No usable DMA configuration, aborting\n");
2271 			goto err_dma;
2272 		}
2273 	}
2274 
2275 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
2276 					   IORESOURCE_MEM),
2277 					   igb_driver_name);
2278 	if (err)
2279 		goto err_pci_reg;
2280 
2281 	pci_enable_pcie_error_reporting(pdev);
2282 
2283 	pci_set_master(pdev);
2284 	pci_save_state(pdev);
2285 
2286 	err = -ENOMEM;
2287 	netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
2288 				   IGB_MAX_TX_QUEUES);
2289 	if (!netdev)
2290 		goto err_alloc_etherdev;
2291 
2292 	SET_NETDEV_DEV(netdev, &pdev->dev);
2293 
2294 	pci_set_drvdata(pdev, netdev);
2295 	adapter = netdev_priv(netdev);
2296 	adapter->netdev = netdev;
2297 	adapter->pdev = pdev;
2298 	hw = &adapter->hw;
2299 	hw->back = adapter;
2300 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
2301 
2302 	err = -EIO;
2303 	hw->hw_addr = pci_iomap(pdev, 0, 0);
2304 	if (!hw->hw_addr)
2305 		goto err_ioremap;
2306 
2307 	netdev->netdev_ops = &igb_netdev_ops;
2308 	igb_set_ethtool_ops(netdev);
2309 	netdev->watchdog_timeo = 5 * HZ;
2310 
2311 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
2312 
2313 	netdev->mem_start = pci_resource_start(pdev, 0);
2314 	netdev->mem_end = pci_resource_end(pdev, 0);
2315 
2316 	/* PCI config space info */
2317 	hw->vendor_id = pdev->vendor;
2318 	hw->device_id = pdev->device;
2319 	hw->revision_id = pdev->revision;
2320 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
2321 	hw->subsystem_device_id = pdev->subsystem_device;
2322 
2323 	/* Copy the default MAC, PHY and NVM function pointers */
2324 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
2325 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
2326 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
2327 	/* Initialize skew-specific constants */
2328 	err = ei->get_invariants(hw);
2329 	if (err)
2330 		goto err_sw_init;
2331 
2332 	/* setup the private structure */
2333 	err = igb_sw_init(adapter);
2334 	if (err)
2335 		goto err_sw_init;
2336 
2337 	igb_get_bus_info_pcie(hw);
2338 
2339 	hw->phy.autoneg_wait_to_complete = false;
2340 
2341 	/* Copper options */
2342 	if (hw->phy.media_type == e1000_media_type_copper) {
2343 		hw->phy.mdix = AUTO_ALL_MODES;
2344 		hw->phy.disable_polarity_correction = false;
2345 		hw->phy.ms_type = e1000_ms_hw_default;
2346 	}
2347 
2348 	if (igb_check_reset_block(hw))
2349 		dev_info(&pdev->dev,
2350 			"PHY reset is blocked due to SOL/IDER session.\n");
2351 
2352 	/* features is initialized to 0 in allocation, it might have bits
2353 	 * set by igb_sw_init so we should use an or instead of an
2354 	 * assignment.
2355 	 */
2356 	netdev->features |= NETIF_F_SG |
2357 			    NETIF_F_IP_CSUM |
2358 			    NETIF_F_IPV6_CSUM |
2359 			    NETIF_F_TSO |
2360 			    NETIF_F_TSO6 |
2361 			    NETIF_F_RXHASH |
2362 			    NETIF_F_RXCSUM |
2363 			    NETIF_F_HW_VLAN_CTAG_RX |
2364 			    NETIF_F_HW_VLAN_CTAG_TX;
2365 
2366 	/* copy netdev features into list of user selectable features */
2367 	netdev->hw_features |= netdev->features;
2368 	netdev->hw_features |= NETIF_F_RXALL;
2369 
2370 	/* set this bit last since it cannot be part of hw_features */
2371 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
2372 
2373 	netdev->vlan_features |= NETIF_F_TSO |
2374 				 NETIF_F_TSO6 |
2375 				 NETIF_F_IP_CSUM |
2376 				 NETIF_F_IPV6_CSUM |
2377 				 NETIF_F_SG;
2378 
2379 	netdev->priv_flags |= IFF_SUPP_NOFCS;
2380 
2381 	if (pci_using_dac) {
2382 		netdev->features |= NETIF_F_HIGHDMA;
2383 		netdev->vlan_features |= NETIF_F_HIGHDMA;
2384 	}
2385 
2386 	if (hw->mac.type >= e1000_82576) {
2387 		netdev->hw_features |= NETIF_F_SCTP_CSUM;
2388 		netdev->features |= NETIF_F_SCTP_CSUM;
2389 	}
2390 
2391 	netdev->priv_flags |= IFF_UNICAST_FLT;
2392 
2393 	adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
2394 
2395 	/* before reading the NVM, reset the controller to put the device in a
2396 	 * known good starting state
2397 	 */
2398 	hw->mac.ops.reset_hw(hw);
2399 
2400 	/* make sure the NVM is good , i211/i210 parts can have special NVM
2401 	 * that doesn't contain a checksum
2402 	 */
2403 	switch (hw->mac.type) {
2404 	case e1000_i210:
2405 	case e1000_i211:
2406 		if (igb_get_flash_presence_i210(hw)) {
2407 			if (hw->nvm.ops.validate(hw) < 0) {
2408 				dev_err(&pdev->dev,
2409 					"The NVM Checksum Is Not Valid\n");
2410 				err = -EIO;
2411 				goto err_eeprom;
2412 			}
2413 		}
2414 		break;
2415 	default:
2416 		if (hw->nvm.ops.validate(hw) < 0) {
2417 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
2418 			err = -EIO;
2419 			goto err_eeprom;
2420 		}
2421 		break;
2422 	}
2423 
2424 	/* copy the MAC address out of the NVM */
2425 	if (hw->mac.ops.read_mac_addr(hw))
2426 		dev_err(&pdev->dev, "NVM Read Error\n");
2427 
2428 	memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
2429 
2430 	if (!is_valid_ether_addr(netdev->dev_addr)) {
2431 		dev_err(&pdev->dev, "Invalid MAC Address\n");
2432 		err = -EIO;
2433 		goto err_eeprom;
2434 	}
2435 
2436 	/* get firmware version for ethtool -i */
2437 	igb_set_fw_version(adapter);
2438 
2439 	/* configure RXPBSIZE and TXPBSIZE */
2440 	if (hw->mac.type == e1000_i210) {
2441 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
2442 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
2443 	}
2444 
2445 	setup_timer(&adapter->watchdog_timer, igb_watchdog,
2446 		    (unsigned long) adapter);
2447 	setup_timer(&adapter->phy_info_timer, igb_update_phy_info,
2448 		    (unsigned long) adapter);
2449 
2450 	INIT_WORK(&adapter->reset_task, igb_reset_task);
2451 	INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
2452 
2453 	/* Initialize link properties that are user-changeable */
2454 	adapter->fc_autoneg = true;
2455 	hw->mac.autoneg = true;
2456 	hw->phy.autoneg_advertised = 0x2f;
2457 
2458 	hw->fc.requested_mode = e1000_fc_default;
2459 	hw->fc.current_mode = e1000_fc_default;
2460 
2461 	igb_validate_mdi_setting(hw);
2462 
2463 	/* By default, support wake on port A */
2464 	if (hw->bus.func == 0)
2465 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2466 
2467 	/* Check the NVM for wake support on non-port A ports */
2468 	if (hw->mac.type >= e1000_82580)
2469 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
2470 				 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
2471 				 &eeprom_data);
2472 	else if (hw->bus.func == 1)
2473 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
2474 
2475 	if (eeprom_data & IGB_EEPROM_APME)
2476 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2477 
2478 	/* now that we have the eeprom settings, apply the special cases where
2479 	 * the eeprom may be wrong or the board simply won't support wake on
2480 	 * lan on a particular port
2481 	 */
2482 	switch (pdev->device) {
2483 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
2484 		adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2485 		break;
2486 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
2487 	case E1000_DEV_ID_82576_FIBER:
2488 	case E1000_DEV_ID_82576_SERDES:
2489 		/* Wake events only supported on port A for dual fiber
2490 		 * regardless of eeprom setting
2491 		 */
2492 		if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
2493 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2494 		break;
2495 	case E1000_DEV_ID_82576_QUAD_COPPER:
2496 	case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
2497 		/* if quad port adapter, disable WoL on all but port A */
2498 		if (global_quad_port_a != 0)
2499 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2500 		else
2501 			adapter->flags |= IGB_FLAG_QUAD_PORT_A;
2502 		/* Reset for multiple quad port adapters */
2503 		if (++global_quad_port_a == 4)
2504 			global_quad_port_a = 0;
2505 		break;
2506 	default:
2507 		/* If the device can't wake, don't set software support */
2508 		if (!device_can_wakeup(&adapter->pdev->dev))
2509 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2510 	}
2511 
2512 	/* initialize the wol settings based on the eeprom settings */
2513 	if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
2514 		adapter->wol |= E1000_WUFC_MAG;
2515 
2516 	/* Some vendors want WoL disabled by default, but still supported */
2517 	if ((hw->mac.type == e1000_i350) &&
2518 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
2519 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2520 		adapter->wol = 0;
2521 	}
2522 
2523 	device_set_wakeup_enable(&adapter->pdev->dev,
2524 				 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
2525 
2526 	/* reset the hardware with the new settings */
2527 	igb_reset(adapter);
2528 
2529 	/* Init the I2C interface */
2530 	err = igb_init_i2c(adapter);
2531 	if (err) {
2532 		dev_err(&pdev->dev, "failed to init i2c interface\n");
2533 		goto err_eeprom;
2534 	}
2535 
2536 	/* let the f/w know that the h/w is now under the control of the
2537 	 * driver.
2538 	 */
2539 	igb_get_hw_control(adapter);
2540 
2541 	strcpy(netdev->name, "eth%d");
2542 	err = register_netdev(netdev);
2543 	if (err)
2544 		goto err_register;
2545 
2546 	/* carrier off reporting is important to ethtool even BEFORE open */
2547 	netif_carrier_off(netdev);
2548 
2549 #ifdef CONFIG_IGB_DCA
2550 	if (dca_add_requester(&pdev->dev) == 0) {
2551 		adapter->flags |= IGB_FLAG_DCA_ENABLED;
2552 		dev_info(&pdev->dev, "DCA enabled\n");
2553 		igb_setup_dca(adapter);
2554 	}
2555 
2556 #endif
2557 #ifdef CONFIG_IGB_HWMON
2558 	/* Initialize the thermal sensor on i350 devices. */
2559 	if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
2560 		u16 ets_word;
2561 
2562 		/* Read the NVM to determine if this i350 device supports an
2563 		 * external thermal sensor.
2564 		 */
2565 		hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
2566 		if (ets_word != 0x0000 && ets_word != 0xFFFF)
2567 			adapter->ets = true;
2568 		else
2569 			adapter->ets = false;
2570 		if (igb_sysfs_init(adapter))
2571 			dev_err(&pdev->dev,
2572 				"failed to allocate sysfs resources\n");
2573 	} else {
2574 		adapter->ets = false;
2575 	}
2576 #endif
2577 	/* Check if Media Autosense is enabled */
2578 	adapter->ei = *ei;
2579 	if (hw->dev_spec._82575.mas_capable)
2580 		igb_init_mas(adapter);
2581 
2582 	/* do hw tstamp init after resetting */
2583 	igb_ptp_init(adapter);
2584 
2585 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
2586 	/* print bus type/speed/width info, not applicable to i354 */
2587 	if (hw->mac.type != e1000_i354) {
2588 		dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
2589 			 netdev->name,
2590 			 ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
2591 			  (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
2592 			   "unknown"),
2593 			 ((hw->bus.width == e1000_bus_width_pcie_x4) ?
2594 			  "Width x4" :
2595 			  (hw->bus.width == e1000_bus_width_pcie_x2) ?
2596 			  "Width x2" :
2597 			  (hw->bus.width == e1000_bus_width_pcie_x1) ?
2598 			  "Width x1" : "unknown"), netdev->dev_addr);
2599 	}
2600 
2601 	if ((hw->mac.type >= e1000_i210 ||
2602 	     igb_get_flash_presence_i210(hw))) {
2603 		ret_val = igb_read_part_string(hw, part_str,
2604 					       E1000_PBANUM_LENGTH);
2605 	} else {
2606 		ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
2607 	}
2608 
2609 	if (ret_val)
2610 		strcpy(part_str, "Unknown");
2611 	dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
2612 	dev_info(&pdev->dev,
2613 		"Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
2614 		(adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
2615 		(adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
2616 		adapter->num_rx_queues, adapter->num_tx_queues);
2617 	if (hw->phy.media_type == e1000_media_type_copper) {
2618 		switch (hw->mac.type) {
2619 		case e1000_i350:
2620 		case e1000_i210:
2621 		case e1000_i211:
2622 			/* Enable EEE for internal copper PHY devices */
2623 			err = igb_set_eee_i350(hw, true, true);
2624 			if ((!err) &&
2625 			    (!hw->dev_spec._82575.eee_disable)) {
2626 				adapter->eee_advert =
2627 					MDIO_EEE_100TX | MDIO_EEE_1000T;
2628 				adapter->flags |= IGB_FLAG_EEE;
2629 			}
2630 			break;
2631 		case e1000_i354:
2632 			if ((rd32(E1000_CTRL_EXT) &
2633 			    E1000_CTRL_EXT_LINK_MODE_SGMII)) {
2634 				err = igb_set_eee_i354(hw, true, true);
2635 				if ((!err) &&
2636 					(!hw->dev_spec._82575.eee_disable)) {
2637 					adapter->eee_advert =
2638 					   MDIO_EEE_100TX | MDIO_EEE_1000T;
2639 					adapter->flags |= IGB_FLAG_EEE;
2640 				}
2641 			}
2642 			break;
2643 		default:
2644 			break;
2645 		}
2646 	}
2647 	pm_runtime_put_noidle(&pdev->dev);
2648 	return 0;
2649 
2650 err_register:
2651 	igb_release_hw_control(adapter);
2652 	memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
2653 err_eeprom:
2654 	if (!igb_check_reset_block(hw))
2655 		igb_reset_phy(hw);
2656 
2657 	if (hw->flash_address)
2658 		iounmap(hw->flash_address);
2659 err_sw_init:
2660 	igb_clear_interrupt_scheme(adapter);
2661 	pci_iounmap(pdev, hw->hw_addr);
2662 err_ioremap:
2663 	free_netdev(netdev);
2664 err_alloc_etherdev:
2665 	pci_release_selected_regions(pdev,
2666 				     pci_select_bars(pdev, IORESOURCE_MEM));
2667 err_pci_reg:
2668 err_dma:
2669 	pci_disable_device(pdev);
2670 	return err;
2671 }
2672 
2673 #ifdef CONFIG_PCI_IOV
2674 static int igb_disable_sriov(struct pci_dev *pdev)
2675 {
2676 	struct net_device *netdev = pci_get_drvdata(pdev);
2677 	struct igb_adapter *adapter = netdev_priv(netdev);
2678 	struct e1000_hw *hw = &adapter->hw;
2679 
2680 	/* reclaim resources allocated to VFs */
2681 	if (adapter->vf_data) {
2682 		/* disable iov and allow time for transactions to clear */
2683 		if (pci_vfs_assigned(pdev)) {
2684 			dev_warn(&pdev->dev,
2685 				 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
2686 			return -EPERM;
2687 		} else {
2688 			pci_disable_sriov(pdev);
2689 			msleep(500);
2690 		}
2691 
2692 		kfree(adapter->vf_data);
2693 		adapter->vf_data = NULL;
2694 		adapter->vfs_allocated_count = 0;
2695 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
2696 		wrfl();
2697 		msleep(100);
2698 		dev_info(&pdev->dev, "IOV Disabled\n");
2699 
2700 		/* Re-enable DMA Coalescing flag since IOV is turned off */
2701 		adapter->flags |= IGB_FLAG_DMAC;
2702 	}
2703 
2704 	return 0;
2705 }
2706 
2707 static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
2708 {
2709 	struct net_device *netdev = pci_get_drvdata(pdev);
2710 	struct igb_adapter *adapter = netdev_priv(netdev);
2711 	int old_vfs = pci_num_vf(pdev);
2712 	int err = 0;
2713 	int i;
2714 
2715 	if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
2716 		err = -EPERM;
2717 		goto out;
2718 	}
2719 	if (!num_vfs)
2720 		goto out;
2721 
2722 	if (old_vfs) {
2723 		dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
2724 			 old_vfs, max_vfs);
2725 		adapter->vfs_allocated_count = old_vfs;
2726 	} else
2727 		adapter->vfs_allocated_count = num_vfs;
2728 
2729 	adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
2730 				sizeof(struct vf_data_storage), GFP_KERNEL);
2731 
2732 	/* if allocation failed then we do not support SR-IOV */
2733 	if (!adapter->vf_data) {
2734 		adapter->vfs_allocated_count = 0;
2735 		dev_err(&pdev->dev,
2736 			"Unable to allocate memory for VF Data Storage\n");
2737 		err = -ENOMEM;
2738 		goto out;
2739 	}
2740 
2741 	/* only call pci_enable_sriov() if no VFs are allocated already */
2742 	if (!old_vfs) {
2743 		err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
2744 		if (err)
2745 			goto err_out;
2746 	}
2747 	dev_info(&pdev->dev, "%d VFs allocated\n",
2748 		 adapter->vfs_allocated_count);
2749 	for (i = 0; i < adapter->vfs_allocated_count; i++)
2750 		igb_vf_configure(adapter, i);
2751 
2752 	/* DMA Coalescing is not supported in IOV mode. */
2753 	adapter->flags &= ~IGB_FLAG_DMAC;
2754 	goto out;
2755 
2756 err_out:
2757 	kfree(adapter->vf_data);
2758 	adapter->vf_data = NULL;
2759 	adapter->vfs_allocated_count = 0;
2760 out:
2761 	return err;
2762 }
2763 
2764 #endif
2765 /**
2766  *  igb_remove_i2c - Cleanup  I2C interface
2767  *  @adapter: pointer to adapter structure
2768  **/
2769 static void igb_remove_i2c(struct igb_adapter *adapter)
2770 {
2771 	/* free the adapter bus structure */
2772 	i2c_del_adapter(&adapter->i2c_adap);
2773 }
2774 
2775 /**
2776  *  igb_remove - Device Removal Routine
2777  *  @pdev: PCI device information struct
2778  *
2779  *  igb_remove is called by the PCI subsystem to alert the driver
2780  *  that it should release a PCI device.  The could be caused by a
2781  *  Hot-Plug event, or because the driver is going to be removed from
2782  *  memory.
2783  **/
2784 static void igb_remove(struct pci_dev *pdev)
2785 {
2786 	struct net_device *netdev = pci_get_drvdata(pdev);
2787 	struct igb_adapter *adapter = netdev_priv(netdev);
2788 	struct e1000_hw *hw = &adapter->hw;
2789 
2790 	pm_runtime_get_noresume(&pdev->dev);
2791 #ifdef CONFIG_IGB_HWMON
2792 	igb_sysfs_exit(adapter);
2793 #endif
2794 	igb_remove_i2c(adapter);
2795 	igb_ptp_stop(adapter);
2796 	/* The watchdog timer may be rescheduled, so explicitly
2797 	 * disable watchdog from being rescheduled.
2798 	 */
2799 	set_bit(__IGB_DOWN, &adapter->state);
2800 	del_timer_sync(&adapter->watchdog_timer);
2801 	del_timer_sync(&adapter->phy_info_timer);
2802 
2803 	cancel_work_sync(&adapter->reset_task);
2804 	cancel_work_sync(&adapter->watchdog_task);
2805 
2806 #ifdef CONFIG_IGB_DCA
2807 	if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
2808 		dev_info(&pdev->dev, "DCA disabled\n");
2809 		dca_remove_requester(&pdev->dev);
2810 		adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
2811 		wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
2812 	}
2813 #endif
2814 
2815 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
2816 	 * would have already happened in close and is redundant.
2817 	 */
2818 	igb_release_hw_control(adapter);
2819 
2820 	unregister_netdev(netdev);
2821 
2822 	igb_clear_interrupt_scheme(adapter);
2823 
2824 #ifdef CONFIG_PCI_IOV
2825 	igb_disable_sriov(pdev);
2826 #endif
2827 
2828 	pci_iounmap(pdev, hw->hw_addr);
2829 	if (hw->flash_address)
2830 		iounmap(hw->flash_address);
2831 	pci_release_selected_regions(pdev,
2832 				     pci_select_bars(pdev, IORESOURCE_MEM));
2833 
2834 	kfree(adapter->shadow_vfta);
2835 	free_netdev(netdev);
2836 
2837 	pci_disable_pcie_error_reporting(pdev);
2838 
2839 	pci_disable_device(pdev);
2840 }
2841 
2842 /**
2843  *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
2844  *  @adapter: board private structure to initialize
2845  *
2846  *  This function initializes the vf specific data storage and then attempts to
2847  *  allocate the VFs.  The reason for ordering it this way is because it is much
2848  *  mor expensive time wise to disable SR-IOV than it is to allocate and free
2849  *  the memory for the VFs.
2850  **/
2851 static void igb_probe_vfs(struct igb_adapter *adapter)
2852 {
2853 #ifdef CONFIG_PCI_IOV
2854 	struct pci_dev *pdev = adapter->pdev;
2855 	struct e1000_hw *hw = &adapter->hw;
2856 
2857 	/* Virtualization features not supported on i210 family. */
2858 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
2859 		return;
2860 
2861 	pci_sriov_set_totalvfs(pdev, 7);
2862 	igb_pci_enable_sriov(pdev, max_vfs);
2863 
2864 #endif /* CONFIG_PCI_IOV */
2865 }
2866 
2867 static void igb_init_queue_configuration(struct igb_adapter *adapter)
2868 {
2869 	struct e1000_hw *hw = &adapter->hw;
2870 	u32 max_rss_queues;
2871 
2872 	/* Determine the maximum number of RSS queues supported. */
2873 	switch (hw->mac.type) {
2874 	case e1000_i211:
2875 		max_rss_queues = IGB_MAX_RX_QUEUES_I211;
2876 		break;
2877 	case e1000_82575:
2878 	case e1000_i210:
2879 		max_rss_queues = IGB_MAX_RX_QUEUES_82575;
2880 		break;
2881 	case e1000_i350:
2882 		/* I350 cannot do RSS and SR-IOV at the same time */
2883 		if (!!adapter->vfs_allocated_count) {
2884 			max_rss_queues = 1;
2885 			break;
2886 		}
2887 		/* fall through */
2888 	case e1000_82576:
2889 		if (!!adapter->vfs_allocated_count) {
2890 			max_rss_queues = 2;
2891 			break;
2892 		}
2893 		/* fall through */
2894 	case e1000_82580:
2895 	case e1000_i354:
2896 	default:
2897 		max_rss_queues = IGB_MAX_RX_QUEUES;
2898 		break;
2899 	}
2900 
2901 	adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
2902 
2903 	/* Determine if we need to pair queues. */
2904 	switch (hw->mac.type) {
2905 	case e1000_82575:
2906 	case e1000_i211:
2907 		/* Device supports enough interrupts without queue pairing. */
2908 		break;
2909 	case e1000_82576:
2910 		/* If VFs are going to be allocated with RSS queues then we
2911 		 * should pair the queues in order to conserve interrupts due
2912 		 * to limited supply.
2913 		 */
2914 		if ((adapter->rss_queues > 1) &&
2915 		    (adapter->vfs_allocated_count > 6))
2916 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
2917 		/* fall through */
2918 	case e1000_82580:
2919 	case e1000_i350:
2920 	case e1000_i354:
2921 	case e1000_i210:
2922 	default:
2923 		/* If rss_queues > half of max_rss_queues, pair the queues in
2924 		 * order to conserve interrupts due to limited supply.
2925 		 */
2926 		if (adapter->rss_queues > (max_rss_queues / 2))
2927 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
2928 		break;
2929 	}
2930 }
2931 
2932 /**
2933  *  igb_sw_init - Initialize general software structures (struct igb_adapter)
2934  *  @adapter: board private structure to initialize
2935  *
2936  *  igb_sw_init initializes the Adapter private data structure.
2937  *  Fields are initialized based on PCI device information and
2938  *  OS network device settings (MTU size).
2939  **/
2940 static int igb_sw_init(struct igb_adapter *adapter)
2941 {
2942 	struct e1000_hw *hw = &adapter->hw;
2943 	struct net_device *netdev = adapter->netdev;
2944 	struct pci_dev *pdev = adapter->pdev;
2945 
2946 	pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
2947 
2948 	/* set default ring sizes */
2949 	adapter->tx_ring_count = IGB_DEFAULT_TXD;
2950 	adapter->rx_ring_count = IGB_DEFAULT_RXD;
2951 
2952 	/* set default ITR values */
2953 	adapter->rx_itr_setting = IGB_DEFAULT_ITR;
2954 	adapter->tx_itr_setting = IGB_DEFAULT_ITR;
2955 
2956 	/* set default work limits */
2957 	adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
2958 
2959 	adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN +
2960 				  VLAN_HLEN;
2961 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
2962 
2963 	spin_lock_init(&adapter->stats64_lock);
2964 #ifdef CONFIG_PCI_IOV
2965 	switch (hw->mac.type) {
2966 	case e1000_82576:
2967 	case e1000_i350:
2968 		if (max_vfs > 7) {
2969 			dev_warn(&pdev->dev,
2970 				 "Maximum of 7 VFs per PF, using max\n");
2971 			max_vfs = adapter->vfs_allocated_count = 7;
2972 		} else
2973 			adapter->vfs_allocated_count = max_vfs;
2974 		if (adapter->vfs_allocated_count)
2975 			dev_warn(&pdev->dev,
2976 				 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
2977 		break;
2978 	default:
2979 		break;
2980 	}
2981 #endif /* CONFIG_PCI_IOV */
2982 
2983 	igb_init_queue_configuration(adapter);
2984 
2985 	/* Setup and initialize a copy of the hw vlan table array */
2986 	adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
2987 				       GFP_ATOMIC);
2988 
2989 	/* This call may decrease the number of queues */
2990 	if (igb_init_interrupt_scheme(adapter, true)) {
2991 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
2992 		return -ENOMEM;
2993 	}
2994 
2995 	igb_probe_vfs(adapter);
2996 
2997 	/* Explicitly disable IRQ since the NIC can be in any state. */
2998 	igb_irq_disable(adapter);
2999 
3000 	if (hw->mac.type >= e1000_i350)
3001 		adapter->flags &= ~IGB_FLAG_DMAC;
3002 
3003 	set_bit(__IGB_DOWN, &adapter->state);
3004 	return 0;
3005 }
3006 
3007 /**
3008  *  igb_open - Called when a network interface is made active
3009  *  @netdev: network interface device structure
3010  *
3011  *  Returns 0 on success, negative value on failure
3012  *
3013  *  The open entry point is called when a network interface is made
3014  *  active by the system (IFF_UP).  At this point all resources needed
3015  *  for transmit and receive operations are allocated, the interrupt
3016  *  handler is registered with the OS, the watchdog timer is started,
3017  *  and the stack is notified that the interface is ready.
3018  **/
3019 static int __igb_open(struct net_device *netdev, bool resuming)
3020 {
3021 	struct igb_adapter *adapter = netdev_priv(netdev);
3022 	struct e1000_hw *hw = &adapter->hw;
3023 	struct pci_dev *pdev = adapter->pdev;
3024 	int err;
3025 	int i;
3026 
3027 	/* disallow open during test */
3028 	if (test_bit(__IGB_TESTING, &adapter->state)) {
3029 		WARN_ON(resuming);
3030 		return -EBUSY;
3031 	}
3032 
3033 	if (!resuming)
3034 		pm_runtime_get_sync(&pdev->dev);
3035 
3036 	netif_carrier_off(netdev);
3037 
3038 	/* allocate transmit descriptors */
3039 	err = igb_setup_all_tx_resources(adapter);
3040 	if (err)
3041 		goto err_setup_tx;
3042 
3043 	/* allocate receive descriptors */
3044 	err = igb_setup_all_rx_resources(adapter);
3045 	if (err)
3046 		goto err_setup_rx;
3047 
3048 	igb_power_up_link(adapter);
3049 
3050 	/* before we allocate an interrupt, we must be ready to handle it.
3051 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3052 	 * as soon as we call pci_request_irq, so we have to setup our
3053 	 * clean_rx handler before we do so.
3054 	 */
3055 	igb_configure(adapter);
3056 
3057 	err = igb_request_irq(adapter);
3058 	if (err)
3059 		goto err_req_irq;
3060 
3061 	/* Notify the stack of the actual queue counts. */
3062 	err = netif_set_real_num_tx_queues(adapter->netdev,
3063 					   adapter->num_tx_queues);
3064 	if (err)
3065 		goto err_set_queues;
3066 
3067 	err = netif_set_real_num_rx_queues(adapter->netdev,
3068 					   adapter->num_rx_queues);
3069 	if (err)
3070 		goto err_set_queues;
3071 
3072 	/* From here on the code is the same as igb_up() */
3073 	clear_bit(__IGB_DOWN, &adapter->state);
3074 
3075 	for (i = 0; i < adapter->num_q_vectors; i++)
3076 		napi_enable(&(adapter->q_vector[i]->napi));
3077 
3078 	/* Clear any pending interrupts. */
3079 	rd32(E1000_ICR);
3080 
3081 	igb_irq_enable(adapter);
3082 
3083 	/* notify VFs that reset has been completed */
3084 	if (adapter->vfs_allocated_count) {
3085 		u32 reg_data = rd32(E1000_CTRL_EXT);
3086 
3087 		reg_data |= E1000_CTRL_EXT_PFRSTD;
3088 		wr32(E1000_CTRL_EXT, reg_data);
3089 	}
3090 
3091 	netif_tx_start_all_queues(netdev);
3092 
3093 	if (!resuming)
3094 		pm_runtime_put(&pdev->dev);
3095 
3096 	/* start the watchdog. */
3097 	hw->mac.get_link_status = 1;
3098 	schedule_work(&adapter->watchdog_task);
3099 
3100 	return 0;
3101 
3102 err_set_queues:
3103 	igb_free_irq(adapter);
3104 err_req_irq:
3105 	igb_release_hw_control(adapter);
3106 	igb_power_down_link(adapter);
3107 	igb_free_all_rx_resources(adapter);
3108 err_setup_rx:
3109 	igb_free_all_tx_resources(adapter);
3110 err_setup_tx:
3111 	igb_reset(adapter);
3112 	if (!resuming)
3113 		pm_runtime_put(&pdev->dev);
3114 
3115 	return err;
3116 }
3117 
3118 static int igb_open(struct net_device *netdev)
3119 {
3120 	return __igb_open(netdev, false);
3121 }
3122 
3123 /**
3124  *  igb_close - Disables a network interface
3125  *  @netdev: network interface device structure
3126  *
3127  *  Returns 0, this is not allowed to fail
3128  *
3129  *  The close entry point is called when an interface is de-activated
3130  *  by the OS.  The hardware is still under the driver's control, but
3131  *  needs to be disabled.  A global MAC reset is issued to stop the
3132  *  hardware, and all transmit and receive resources are freed.
3133  **/
3134 static int __igb_close(struct net_device *netdev, bool suspending)
3135 {
3136 	struct igb_adapter *adapter = netdev_priv(netdev);
3137 	struct pci_dev *pdev = adapter->pdev;
3138 
3139 	WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
3140 
3141 	if (!suspending)
3142 		pm_runtime_get_sync(&pdev->dev);
3143 
3144 	igb_down(adapter);
3145 	igb_free_irq(adapter);
3146 
3147 	igb_free_all_tx_resources(adapter);
3148 	igb_free_all_rx_resources(adapter);
3149 
3150 	if (!suspending)
3151 		pm_runtime_put_sync(&pdev->dev);
3152 	return 0;
3153 }
3154 
3155 static int igb_close(struct net_device *netdev)
3156 {
3157 	return __igb_close(netdev, false);
3158 }
3159 
3160 /**
3161  *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
3162  *  @tx_ring: tx descriptor ring (for a specific queue) to setup
3163  *
3164  *  Return 0 on success, negative on failure
3165  **/
3166 int igb_setup_tx_resources(struct igb_ring *tx_ring)
3167 {
3168 	struct device *dev = tx_ring->dev;
3169 	int size;
3170 
3171 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3172 
3173 	tx_ring->tx_buffer_info = vzalloc(size);
3174 	if (!tx_ring->tx_buffer_info)
3175 		goto err;
3176 
3177 	/* round up to nearest 4K */
3178 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
3179 	tx_ring->size = ALIGN(tx_ring->size, 4096);
3180 
3181 	tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
3182 					   &tx_ring->dma, GFP_KERNEL);
3183 	if (!tx_ring->desc)
3184 		goto err;
3185 
3186 	tx_ring->next_to_use = 0;
3187 	tx_ring->next_to_clean = 0;
3188 
3189 	return 0;
3190 
3191 err:
3192 	vfree(tx_ring->tx_buffer_info);
3193 	tx_ring->tx_buffer_info = NULL;
3194 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
3195 	return -ENOMEM;
3196 }
3197 
3198 /**
3199  *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
3200  *				 (Descriptors) for all queues
3201  *  @adapter: board private structure
3202  *
3203  *  Return 0 on success, negative on failure
3204  **/
3205 static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
3206 {
3207 	struct pci_dev *pdev = adapter->pdev;
3208 	int i, err = 0;
3209 
3210 	for (i = 0; i < adapter->num_tx_queues; i++) {
3211 		err = igb_setup_tx_resources(adapter->tx_ring[i]);
3212 		if (err) {
3213 			dev_err(&pdev->dev,
3214 				"Allocation for Tx Queue %u failed\n", i);
3215 			for (i--; i >= 0; i--)
3216 				igb_free_tx_resources(adapter->tx_ring[i]);
3217 			break;
3218 		}
3219 	}
3220 
3221 	return err;
3222 }
3223 
3224 /**
3225  *  igb_setup_tctl - configure the transmit control registers
3226  *  @adapter: Board private structure
3227  **/
3228 void igb_setup_tctl(struct igb_adapter *adapter)
3229 {
3230 	struct e1000_hw *hw = &adapter->hw;
3231 	u32 tctl;
3232 
3233 	/* disable queue 0 which is enabled by default on 82575 and 82576 */
3234 	wr32(E1000_TXDCTL(0), 0);
3235 
3236 	/* Program the Transmit Control Register */
3237 	tctl = rd32(E1000_TCTL);
3238 	tctl &= ~E1000_TCTL_CT;
3239 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
3240 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
3241 
3242 	igb_config_collision_dist(hw);
3243 
3244 	/* Enable transmits */
3245 	tctl |= E1000_TCTL_EN;
3246 
3247 	wr32(E1000_TCTL, tctl);
3248 }
3249 
3250 /**
3251  *  igb_configure_tx_ring - Configure transmit ring after Reset
3252  *  @adapter: board private structure
3253  *  @ring: tx ring to configure
3254  *
3255  *  Configure a transmit ring after a reset.
3256  **/
3257 void igb_configure_tx_ring(struct igb_adapter *adapter,
3258 			   struct igb_ring *ring)
3259 {
3260 	struct e1000_hw *hw = &adapter->hw;
3261 	u32 txdctl = 0;
3262 	u64 tdba = ring->dma;
3263 	int reg_idx = ring->reg_idx;
3264 
3265 	/* disable the queue */
3266 	wr32(E1000_TXDCTL(reg_idx), 0);
3267 	wrfl();
3268 	mdelay(10);
3269 
3270 	wr32(E1000_TDLEN(reg_idx),
3271 	     ring->count * sizeof(union e1000_adv_tx_desc));
3272 	wr32(E1000_TDBAL(reg_idx),
3273 	     tdba & 0x00000000ffffffffULL);
3274 	wr32(E1000_TDBAH(reg_idx), tdba >> 32);
3275 
3276 	ring->tail = hw->hw_addr + E1000_TDT(reg_idx);
3277 	wr32(E1000_TDH(reg_idx), 0);
3278 	writel(0, ring->tail);
3279 
3280 	txdctl |= IGB_TX_PTHRESH;
3281 	txdctl |= IGB_TX_HTHRESH << 8;
3282 	txdctl |= IGB_TX_WTHRESH << 16;
3283 
3284 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
3285 	wr32(E1000_TXDCTL(reg_idx), txdctl);
3286 }
3287 
3288 /**
3289  *  igb_configure_tx - Configure transmit Unit after Reset
3290  *  @adapter: board private structure
3291  *
3292  *  Configure the Tx unit of the MAC after a reset.
3293  **/
3294 static void igb_configure_tx(struct igb_adapter *adapter)
3295 {
3296 	int i;
3297 
3298 	for (i = 0; i < adapter->num_tx_queues; i++)
3299 		igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
3300 }
3301 
3302 /**
3303  *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
3304  *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
3305  *
3306  *  Returns 0 on success, negative on failure
3307  **/
3308 int igb_setup_rx_resources(struct igb_ring *rx_ring)
3309 {
3310 	struct device *dev = rx_ring->dev;
3311 	int size;
3312 
3313 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3314 
3315 	rx_ring->rx_buffer_info = vzalloc(size);
3316 	if (!rx_ring->rx_buffer_info)
3317 		goto err;
3318 
3319 	/* Round up to nearest 4K */
3320 	rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
3321 	rx_ring->size = ALIGN(rx_ring->size, 4096);
3322 
3323 	rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
3324 					   &rx_ring->dma, GFP_KERNEL);
3325 	if (!rx_ring->desc)
3326 		goto err;
3327 
3328 	rx_ring->next_to_alloc = 0;
3329 	rx_ring->next_to_clean = 0;
3330 	rx_ring->next_to_use = 0;
3331 
3332 	return 0;
3333 
3334 err:
3335 	vfree(rx_ring->rx_buffer_info);
3336 	rx_ring->rx_buffer_info = NULL;
3337 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
3338 	return -ENOMEM;
3339 }
3340 
3341 /**
3342  *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
3343  *				 (Descriptors) for all queues
3344  *  @adapter: board private structure
3345  *
3346  *  Return 0 on success, negative on failure
3347  **/
3348 static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
3349 {
3350 	struct pci_dev *pdev = adapter->pdev;
3351 	int i, err = 0;
3352 
3353 	for (i = 0; i < adapter->num_rx_queues; i++) {
3354 		err = igb_setup_rx_resources(adapter->rx_ring[i]);
3355 		if (err) {
3356 			dev_err(&pdev->dev,
3357 				"Allocation for Rx Queue %u failed\n", i);
3358 			for (i--; i >= 0; i--)
3359 				igb_free_rx_resources(adapter->rx_ring[i]);
3360 			break;
3361 		}
3362 	}
3363 
3364 	return err;
3365 }
3366 
3367 /**
3368  *  igb_setup_mrqc - configure the multiple receive queue control registers
3369  *  @adapter: Board private structure
3370  **/
3371 static void igb_setup_mrqc(struct igb_adapter *adapter)
3372 {
3373 	struct e1000_hw *hw = &adapter->hw;
3374 	u32 mrqc, rxcsum;
3375 	u32 j, num_rx_queues;
3376 	u32 rss_key[10];
3377 
3378 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3379 	for (j = 0; j < 10; j++)
3380 		wr32(E1000_RSSRK(j), rss_key[j]);
3381 
3382 	num_rx_queues = adapter->rss_queues;
3383 
3384 	switch (hw->mac.type) {
3385 	case e1000_82576:
3386 		/* 82576 supports 2 RSS queues for SR-IOV */
3387 		if (adapter->vfs_allocated_count)
3388 			num_rx_queues = 2;
3389 		break;
3390 	default:
3391 		break;
3392 	}
3393 
3394 	if (adapter->rss_indir_tbl_init != num_rx_queues) {
3395 		for (j = 0; j < IGB_RETA_SIZE; j++)
3396 			adapter->rss_indir_tbl[j] =
3397 			(j * num_rx_queues) / IGB_RETA_SIZE;
3398 		adapter->rss_indir_tbl_init = num_rx_queues;
3399 	}
3400 	igb_write_rss_indir_tbl(adapter);
3401 
3402 	/* Disable raw packet checksumming so that RSS hash is placed in
3403 	 * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
3404 	 * offloads as they are enabled by default
3405 	 */
3406 	rxcsum = rd32(E1000_RXCSUM);
3407 	rxcsum |= E1000_RXCSUM_PCSD;
3408 
3409 	if (adapter->hw.mac.type >= e1000_82576)
3410 		/* Enable Receive Checksum Offload for SCTP */
3411 		rxcsum |= E1000_RXCSUM_CRCOFL;
3412 
3413 	/* Don't need to set TUOFL or IPOFL, they default to 1 */
3414 	wr32(E1000_RXCSUM, rxcsum);
3415 
3416 	/* Generate RSS hash based on packet types, TCP/UDP
3417 	 * port numbers and/or IPv4/v6 src and dst addresses
3418 	 */
3419 	mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
3420 	       E1000_MRQC_RSS_FIELD_IPV4_TCP |
3421 	       E1000_MRQC_RSS_FIELD_IPV6 |
3422 	       E1000_MRQC_RSS_FIELD_IPV6_TCP |
3423 	       E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
3424 
3425 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
3426 		mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
3427 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
3428 		mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
3429 
3430 	/* If VMDq is enabled then we set the appropriate mode for that, else
3431 	 * we default to RSS so that an RSS hash is calculated per packet even
3432 	 * if we are only using one queue
3433 	 */
3434 	if (adapter->vfs_allocated_count) {
3435 		if (hw->mac.type > e1000_82575) {
3436 			/* Set the default pool for the PF's first queue */
3437 			u32 vtctl = rd32(E1000_VT_CTL);
3438 
3439 			vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
3440 				   E1000_VT_CTL_DISABLE_DEF_POOL);
3441 			vtctl |= adapter->vfs_allocated_count <<
3442 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
3443 			wr32(E1000_VT_CTL, vtctl);
3444 		}
3445 		if (adapter->rss_queues > 1)
3446 			mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_2Q;
3447 		else
3448 			mrqc |= E1000_MRQC_ENABLE_VMDQ;
3449 	} else {
3450 		if (hw->mac.type != e1000_i211)
3451 			mrqc |= E1000_MRQC_ENABLE_RSS_4Q;
3452 	}
3453 	igb_vmm_control(adapter);
3454 
3455 	wr32(E1000_MRQC, mrqc);
3456 }
3457 
3458 /**
3459  *  igb_setup_rctl - configure the receive control registers
3460  *  @adapter: Board private structure
3461  **/
3462 void igb_setup_rctl(struct igb_adapter *adapter)
3463 {
3464 	struct e1000_hw *hw = &adapter->hw;
3465 	u32 rctl;
3466 
3467 	rctl = rd32(E1000_RCTL);
3468 
3469 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3470 	rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
3471 
3472 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
3473 		(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3474 
3475 	/* enable stripping of CRC. It's unlikely this will break BMC
3476 	 * redirection as it did with e1000. Newer features require
3477 	 * that the HW strips the CRC.
3478 	 */
3479 	rctl |= E1000_RCTL_SECRC;
3480 
3481 	/* disable store bad packets and clear size bits. */
3482 	rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
3483 
3484 	/* enable LPE to prevent packets larger than max_frame_size */
3485 	rctl |= E1000_RCTL_LPE;
3486 
3487 	/* disable queue 0 to prevent tail write w/o re-config */
3488 	wr32(E1000_RXDCTL(0), 0);
3489 
3490 	/* Attention!!!  For SR-IOV PF driver operations you must enable
3491 	 * queue drop for all VF and PF queues to prevent head of line blocking
3492 	 * if an un-trusted VF does not provide descriptors to hardware.
3493 	 */
3494 	if (adapter->vfs_allocated_count) {
3495 		/* set all queue drop enable bits */
3496 		wr32(E1000_QDE, ALL_QUEUES);
3497 	}
3498 
3499 	/* This is useful for sniffing bad packets. */
3500 	if (adapter->netdev->features & NETIF_F_RXALL) {
3501 		/* UPE and MPE will be handled by normal PROMISC logic
3502 		 * in e1000e_set_rx_mode
3503 		 */
3504 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
3505 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
3506 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
3507 
3508 		rctl &= ~(E1000_RCTL_VFE | /* Disable VLAN filter */
3509 			  E1000_RCTL_DPF | /* Allow filtered pause */
3510 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
3511 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3512 		 * and that breaks VLANs.
3513 		 */
3514 	}
3515 
3516 	wr32(E1000_RCTL, rctl);
3517 }
3518 
3519 static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
3520 				   int vfn)
3521 {
3522 	struct e1000_hw *hw = &adapter->hw;
3523 	u32 vmolr;
3524 
3525 	/* if it isn't the PF check to see if VFs are enabled and
3526 	 * increase the size to support vlan tags
3527 	 */
3528 	if (vfn < adapter->vfs_allocated_count &&
3529 	    adapter->vf_data[vfn].vlans_enabled)
3530 		size += VLAN_TAG_SIZE;
3531 
3532 	vmolr = rd32(E1000_VMOLR(vfn));
3533 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
3534 	vmolr |= size | E1000_VMOLR_LPE;
3535 	wr32(E1000_VMOLR(vfn), vmolr);
3536 
3537 	return 0;
3538 }
3539 
3540 /**
3541  *  igb_rlpml_set - set maximum receive packet size
3542  *  @adapter: board private structure
3543  *
3544  *  Configure maximum receivable packet size.
3545  **/
3546 static void igb_rlpml_set(struct igb_adapter *adapter)
3547 {
3548 	u32 max_frame_size = adapter->max_frame_size;
3549 	struct e1000_hw *hw = &adapter->hw;
3550 	u16 pf_id = adapter->vfs_allocated_count;
3551 
3552 	if (pf_id) {
3553 		igb_set_vf_rlpml(adapter, max_frame_size, pf_id);
3554 		/* If we're in VMDQ or SR-IOV mode, then set global RLPML
3555 		 * to our max jumbo frame size, in case we need to enable
3556 		 * jumbo frames on one of the rings later.
3557 		 * This will not pass over-length frames into the default
3558 		 * queue because it's gated by the VMOLR.RLPML.
3559 		 */
3560 		max_frame_size = MAX_JUMBO_FRAME_SIZE;
3561 	}
3562 
3563 	wr32(E1000_RLPML, max_frame_size);
3564 }
3565 
3566 static inline void igb_set_vmolr(struct igb_adapter *adapter,
3567 				 int vfn, bool aupe)
3568 {
3569 	struct e1000_hw *hw = &adapter->hw;
3570 	u32 vmolr;
3571 
3572 	/* This register exists only on 82576 and newer so if we are older then
3573 	 * we should exit and do nothing
3574 	 */
3575 	if (hw->mac.type < e1000_82576)
3576 		return;
3577 
3578 	vmolr = rd32(E1000_VMOLR(vfn));
3579 	vmolr |= E1000_VMOLR_STRVLAN; /* Strip vlan tags */
3580 	if (hw->mac.type == e1000_i350) {
3581 		u32 dvmolr;
3582 
3583 		dvmolr = rd32(E1000_DVMOLR(vfn));
3584 		dvmolr |= E1000_DVMOLR_STRVLAN;
3585 		wr32(E1000_DVMOLR(vfn), dvmolr);
3586 	}
3587 	if (aupe)
3588 		vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
3589 	else
3590 		vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
3591 
3592 	/* clear all bits that might not be set */
3593 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
3594 
3595 	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
3596 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
3597 	/* for VMDq only allow the VFs and pool 0 to accept broadcast and
3598 	 * multicast packets
3599 	 */
3600 	if (vfn <= adapter->vfs_allocated_count)
3601 		vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
3602 
3603 	wr32(E1000_VMOLR(vfn), vmolr);
3604 }
3605 
3606 /**
3607  *  igb_configure_rx_ring - Configure a receive ring after Reset
3608  *  @adapter: board private structure
3609  *  @ring: receive ring to be configured
3610  *
3611  *  Configure the Rx unit of the MAC after a reset.
3612  **/
3613 void igb_configure_rx_ring(struct igb_adapter *adapter,
3614 			   struct igb_ring *ring)
3615 {
3616 	struct e1000_hw *hw = &adapter->hw;
3617 	u64 rdba = ring->dma;
3618 	int reg_idx = ring->reg_idx;
3619 	u32 srrctl = 0, rxdctl = 0;
3620 
3621 	/* disable the queue */
3622 	wr32(E1000_RXDCTL(reg_idx), 0);
3623 
3624 	/* Set DMA base address registers */
3625 	wr32(E1000_RDBAL(reg_idx),
3626 	     rdba & 0x00000000ffffffffULL);
3627 	wr32(E1000_RDBAH(reg_idx), rdba >> 32);
3628 	wr32(E1000_RDLEN(reg_idx),
3629 	     ring->count * sizeof(union e1000_adv_rx_desc));
3630 
3631 	/* initialize head and tail */
3632 	ring->tail = hw->hw_addr + E1000_RDT(reg_idx);
3633 	wr32(E1000_RDH(reg_idx), 0);
3634 	writel(0, ring->tail);
3635 
3636 	/* set descriptor configuration */
3637 	srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
3638 	srrctl |= IGB_RX_BUFSZ >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3639 	srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3640 	if (hw->mac.type >= e1000_82580)
3641 		srrctl |= E1000_SRRCTL_TIMESTAMP;
3642 	/* Only set Drop Enable if we are supporting multiple queues */
3643 	if (adapter->vfs_allocated_count || adapter->num_rx_queues > 1)
3644 		srrctl |= E1000_SRRCTL_DROP_EN;
3645 
3646 	wr32(E1000_SRRCTL(reg_idx), srrctl);
3647 
3648 	/* set filtering for VMDQ pools */
3649 	igb_set_vmolr(adapter, reg_idx & 0x7, true);
3650 
3651 	rxdctl |= IGB_RX_PTHRESH;
3652 	rxdctl |= IGB_RX_HTHRESH << 8;
3653 	rxdctl |= IGB_RX_WTHRESH << 16;
3654 
3655 	/* enable receive descriptor fetching */
3656 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3657 	wr32(E1000_RXDCTL(reg_idx), rxdctl);
3658 }
3659 
3660 /**
3661  *  igb_configure_rx - Configure receive Unit after Reset
3662  *  @adapter: board private structure
3663  *
3664  *  Configure the Rx unit of the MAC after a reset.
3665  **/
3666 static void igb_configure_rx(struct igb_adapter *adapter)
3667 {
3668 	int i;
3669 
3670 	/* set UTA to appropriate mode */
3671 	igb_set_uta(adapter);
3672 
3673 	/* set the correct pool for the PF default MAC address in entry 0 */
3674 	igb_rar_set_qsel(adapter, adapter->hw.mac.addr, 0,
3675 			 adapter->vfs_allocated_count);
3676 
3677 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3678 	 * the Base and Length of the Rx Descriptor Ring
3679 	 */
3680 	for (i = 0; i < adapter->num_rx_queues; i++)
3681 		igb_configure_rx_ring(adapter, adapter->rx_ring[i]);
3682 }
3683 
3684 /**
3685  *  igb_free_tx_resources - Free Tx Resources per Queue
3686  *  @tx_ring: Tx descriptor ring for a specific queue
3687  *
3688  *  Free all transmit software resources
3689  **/
3690 void igb_free_tx_resources(struct igb_ring *tx_ring)
3691 {
3692 	igb_clean_tx_ring(tx_ring);
3693 
3694 	vfree(tx_ring->tx_buffer_info);
3695 	tx_ring->tx_buffer_info = NULL;
3696 
3697 	/* if not set, then don't free */
3698 	if (!tx_ring->desc)
3699 		return;
3700 
3701 	dma_free_coherent(tx_ring->dev, tx_ring->size,
3702 			  tx_ring->desc, tx_ring->dma);
3703 
3704 	tx_ring->desc = NULL;
3705 }
3706 
3707 /**
3708  *  igb_free_all_tx_resources - Free Tx Resources for All Queues
3709  *  @adapter: board private structure
3710  *
3711  *  Free all transmit software resources
3712  **/
3713 static void igb_free_all_tx_resources(struct igb_adapter *adapter)
3714 {
3715 	int i;
3716 
3717 	for (i = 0; i < adapter->num_tx_queues; i++)
3718 		if (adapter->tx_ring[i])
3719 			igb_free_tx_resources(adapter->tx_ring[i]);
3720 }
3721 
3722 void igb_unmap_and_free_tx_resource(struct igb_ring *ring,
3723 				    struct igb_tx_buffer *tx_buffer)
3724 {
3725 	if (tx_buffer->skb) {
3726 		dev_kfree_skb_any(tx_buffer->skb);
3727 		if (dma_unmap_len(tx_buffer, len))
3728 			dma_unmap_single(ring->dev,
3729 					 dma_unmap_addr(tx_buffer, dma),
3730 					 dma_unmap_len(tx_buffer, len),
3731 					 DMA_TO_DEVICE);
3732 	} else if (dma_unmap_len(tx_buffer, len)) {
3733 		dma_unmap_page(ring->dev,
3734 			       dma_unmap_addr(tx_buffer, dma),
3735 			       dma_unmap_len(tx_buffer, len),
3736 			       DMA_TO_DEVICE);
3737 	}
3738 	tx_buffer->next_to_watch = NULL;
3739 	tx_buffer->skb = NULL;
3740 	dma_unmap_len_set(tx_buffer, len, 0);
3741 	/* buffer_info must be completely set up in the transmit path */
3742 }
3743 
3744 /**
3745  *  igb_clean_tx_ring - Free Tx Buffers
3746  *  @tx_ring: ring to be cleaned
3747  **/
3748 static void igb_clean_tx_ring(struct igb_ring *tx_ring)
3749 {
3750 	struct igb_tx_buffer *buffer_info;
3751 	unsigned long size;
3752 	u16 i;
3753 
3754 	if (!tx_ring->tx_buffer_info)
3755 		return;
3756 	/* Free all the Tx ring sk_buffs */
3757 
3758 	for (i = 0; i < tx_ring->count; i++) {
3759 		buffer_info = &tx_ring->tx_buffer_info[i];
3760 		igb_unmap_and_free_tx_resource(tx_ring, buffer_info);
3761 	}
3762 
3763 	netdev_tx_reset_queue(txring_txq(tx_ring));
3764 
3765 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3766 	memset(tx_ring->tx_buffer_info, 0, size);
3767 
3768 	/* Zero out the descriptor ring */
3769 	memset(tx_ring->desc, 0, tx_ring->size);
3770 
3771 	tx_ring->next_to_use = 0;
3772 	tx_ring->next_to_clean = 0;
3773 }
3774 
3775 /**
3776  *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
3777  *  @adapter: board private structure
3778  **/
3779 static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
3780 {
3781 	int i;
3782 
3783 	for (i = 0; i < adapter->num_tx_queues; i++)
3784 		if (adapter->tx_ring[i])
3785 			igb_clean_tx_ring(adapter->tx_ring[i]);
3786 }
3787 
3788 /**
3789  *  igb_free_rx_resources - Free Rx Resources
3790  *  @rx_ring: ring to clean the resources from
3791  *
3792  *  Free all receive software resources
3793  **/
3794 void igb_free_rx_resources(struct igb_ring *rx_ring)
3795 {
3796 	igb_clean_rx_ring(rx_ring);
3797 
3798 	vfree(rx_ring->rx_buffer_info);
3799 	rx_ring->rx_buffer_info = NULL;
3800 
3801 	/* if not set, then don't free */
3802 	if (!rx_ring->desc)
3803 		return;
3804 
3805 	dma_free_coherent(rx_ring->dev, rx_ring->size,
3806 			  rx_ring->desc, rx_ring->dma);
3807 
3808 	rx_ring->desc = NULL;
3809 }
3810 
3811 /**
3812  *  igb_free_all_rx_resources - Free Rx Resources for All Queues
3813  *  @adapter: board private structure
3814  *
3815  *  Free all receive software resources
3816  **/
3817 static void igb_free_all_rx_resources(struct igb_adapter *adapter)
3818 {
3819 	int i;
3820 
3821 	for (i = 0; i < adapter->num_rx_queues; i++)
3822 		if (adapter->rx_ring[i])
3823 			igb_free_rx_resources(adapter->rx_ring[i]);
3824 }
3825 
3826 /**
3827  *  igb_clean_rx_ring - Free Rx Buffers per Queue
3828  *  @rx_ring: ring to free buffers from
3829  **/
3830 static void igb_clean_rx_ring(struct igb_ring *rx_ring)
3831 {
3832 	unsigned long size;
3833 	u16 i;
3834 
3835 	if (rx_ring->skb)
3836 		dev_kfree_skb(rx_ring->skb);
3837 	rx_ring->skb = NULL;
3838 
3839 	if (!rx_ring->rx_buffer_info)
3840 		return;
3841 
3842 	/* Free all the Rx ring sk_buffs */
3843 	for (i = 0; i < rx_ring->count; i++) {
3844 		struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
3845 
3846 		if (!buffer_info->page)
3847 			continue;
3848 
3849 		dma_unmap_page(rx_ring->dev,
3850 			       buffer_info->dma,
3851 			       PAGE_SIZE,
3852 			       DMA_FROM_DEVICE);
3853 		__free_page(buffer_info->page);
3854 
3855 		buffer_info->page = NULL;
3856 	}
3857 
3858 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3859 	memset(rx_ring->rx_buffer_info, 0, size);
3860 
3861 	/* Zero out the descriptor ring */
3862 	memset(rx_ring->desc, 0, rx_ring->size);
3863 
3864 	rx_ring->next_to_alloc = 0;
3865 	rx_ring->next_to_clean = 0;
3866 	rx_ring->next_to_use = 0;
3867 }
3868 
3869 /**
3870  *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
3871  *  @adapter: board private structure
3872  **/
3873 static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
3874 {
3875 	int i;
3876 
3877 	for (i = 0; i < adapter->num_rx_queues; i++)
3878 		if (adapter->rx_ring[i])
3879 			igb_clean_rx_ring(adapter->rx_ring[i]);
3880 }
3881 
3882 /**
3883  *  igb_set_mac - Change the Ethernet Address of the NIC
3884  *  @netdev: network interface device structure
3885  *  @p: pointer to an address structure
3886  *
3887  *  Returns 0 on success, negative on failure
3888  **/
3889 static int igb_set_mac(struct net_device *netdev, void *p)
3890 {
3891 	struct igb_adapter *adapter = netdev_priv(netdev);
3892 	struct e1000_hw *hw = &adapter->hw;
3893 	struct sockaddr *addr = p;
3894 
3895 	if (!is_valid_ether_addr(addr->sa_data))
3896 		return -EADDRNOTAVAIL;
3897 
3898 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
3899 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
3900 
3901 	/* set the correct pool for the new PF MAC address in entry 0 */
3902 	igb_rar_set_qsel(adapter, hw->mac.addr, 0,
3903 			 adapter->vfs_allocated_count);
3904 
3905 	return 0;
3906 }
3907 
3908 /**
3909  *  igb_write_mc_addr_list - write multicast addresses to MTA
3910  *  @netdev: network interface device structure
3911  *
3912  *  Writes multicast address list to the MTA hash table.
3913  *  Returns: -ENOMEM on failure
3914  *           0 on no addresses written
3915  *           X on writing X addresses to MTA
3916  **/
3917 static int igb_write_mc_addr_list(struct net_device *netdev)
3918 {
3919 	struct igb_adapter *adapter = netdev_priv(netdev);
3920 	struct e1000_hw *hw = &adapter->hw;
3921 	struct netdev_hw_addr *ha;
3922 	u8  *mta_list;
3923 	int i;
3924 
3925 	if (netdev_mc_empty(netdev)) {
3926 		/* nothing to program, so clear mc list */
3927 		igb_update_mc_addr_list(hw, NULL, 0);
3928 		igb_restore_vf_multicasts(adapter);
3929 		return 0;
3930 	}
3931 
3932 	mta_list = kzalloc(netdev_mc_count(netdev) * 6, GFP_ATOMIC);
3933 	if (!mta_list)
3934 		return -ENOMEM;
3935 
3936 	/* The shared function expects a packed array of only addresses. */
3937 	i = 0;
3938 	netdev_for_each_mc_addr(ha, netdev)
3939 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3940 
3941 	igb_update_mc_addr_list(hw, mta_list, i);
3942 	kfree(mta_list);
3943 
3944 	return netdev_mc_count(netdev);
3945 }
3946 
3947 /**
3948  *  igb_write_uc_addr_list - write unicast addresses to RAR table
3949  *  @netdev: network interface device structure
3950  *
3951  *  Writes unicast address list to the RAR table.
3952  *  Returns: -ENOMEM on failure/insufficient address space
3953  *           0 on no addresses written
3954  *           X on writing X addresses to the RAR table
3955  **/
3956 static int igb_write_uc_addr_list(struct net_device *netdev)
3957 {
3958 	struct igb_adapter *adapter = netdev_priv(netdev);
3959 	struct e1000_hw *hw = &adapter->hw;
3960 	unsigned int vfn = adapter->vfs_allocated_count;
3961 	unsigned int rar_entries = hw->mac.rar_entry_count - (vfn + 1);
3962 	int count = 0;
3963 
3964 	/* return ENOMEM indicating insufficient memory for addresses */
3965 	if (netdev_uc_count(netdev) > rar_entries)
3966 		return -ENOMEM;
3967 
3968 	if (!netdev_uc_empty(netdev) && rar_entries) {
3969 		struct netdev_hw_addr *ha;
3970 
3971 		netdev_for_each_uc_addr(ha, netdev) {
3972 			if (!rar_entries)
3973 				break;
3974 			igb_rar_set_qsel(adapter, ha->addr,
3975 					 rar_entries--,
3976 					 vfn);
3977 			count++;
3978 		}
3979 	}
3980 	/* write the addresses in reverse order to avoid write combining */
3981 	for (; rar_entries > 0 ; rar_entries--) {
3982 		wr32(E1000_RAH(rar_entries), 0);
3983 		wr32(E1000_RAL(rar_entries), 0);
3984 	}
3985 	wrfl();
3986 
3987 	return count;
3988 }
3989 
3990 /**
3991  *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
3992  *  @netdev: network interface device structure
3993  *
3994  *  The set_rx_mode entry point is called whenever the unicast or multicast
3995  *  address lists or the network interface flags are updated.  This routine is
3996  *  responsible for configuring the hardware for proper unicast, multicast,
3997  *  promiscuous mode, and all-multi behavior.
3998  **/
3999 static void igb_set_rx_mode(struct net_device *netdev)
4000 {
4001 	struct igb_adapter *adapter = netdev_priv(netdev);
4002 	struct e1000_hw *hw = &adapter->hw;
4003 	unsigned int vfn = adapter->vfs_allocated_count;
4004 	u32 rctl, vmolr = 0;
4005 	int count;
4006 
4007 	/* Check for Promiscuous and All Multicast modes */
4008 	rctl = rd32(E1000_RCTL);
4009 
4010 	/* clear the effected bits */
4011 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE | E1000_RCTL_VFE);
4012 
4013 	if (netdev->flags & IFF_PROMISC) {
4014 		/* retain VLAN HW filtering if in VT mode */
4015 		if (adapter->vfs_allocated_count)
4016 			rctl |= E1000_RCTL_VFE;
4017 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
4018 		vmolr |= (E1000_VMOLR_ROPE | E1000_VMOLR_MPME);
4019 	} else {
4020 		if (netdev->flags & IFF_ALLMULTI) {
4021 			rctl |= E1000_RCTL_MPE;
4022 			vmolr |= E1000_VMOLR_MPME;
4023 		} else {
4024 			/* Write addresses to the MTA, if the attempt fails
4025 			 * then we should just turn on promiscuous mode so
4026 			 * that we can at least receive multicast traffic
4027 			 */
4028 			count = igb_write_mc_addr_list(netdev);
4029 			if (count < 0) {
4030 				rctl |= E1000_RCTL_MPE;
4031 				vmolr |= E1000_VMOLR_MPME;
4032 			} else if (count) {
4033 				vmolr |= E1000_VMOLR_ROMPE;
4034 			}
4035 		}
4036 		/* Write addresses to available RAR registers, if there is not
4037 		 * sufficient space to store all the addresses then enable
4038 		 * unicast promiscuous mode
4039 		 */
4040 		count = igb_write_uc_addr_list(netdev);
4041 		if (count < 0) {
4042 			rctl |= E1000_RCTL_UPE;
4043 			vmolr |= E1000_VMOLR_ROPE;
4044 		}
4045 		rctl |= E1000_RCTL_VFE;
4046 	}
4047 	wr32(E1000_RCTL, rctl);
4048 
4049 	/* In order to support SR-IOV and eventually VMDq it is necessary to set
4050 	 * the VMOLR to enable the appropriate modes.  Without this workaround
4051 	 * we will have issues with VLAN tag stripping not being done for frames
4052 	 * that are only arriving because we are the default pool
4053 	 */
4054 	if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
4055 		return;
4056 
4057 	vmolr |= rd32(E1000_VMOLR(vfn)) &
4058 		 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
4059 	wr32(E1000_VMOLR(vfn), vmolr);
4060 	igb_restore_vf_multicasts(adapter);
4061 }
4062 
4063 static void igb_check_wvbr(struct igb_adapter *adapter)
4064 {
4065 	struct e1000_hw *hw = &adapter->hw;
4066 	u32 wvbr = 0;
4067 
4068 	switch (hw->mac.type) {
4069 	case e1000_82576:
4070 	case e1000_i350:
4071 		wvbr = rd32(E1000_WVBR);
4072 		if (!wvbr)
4073 			return;
4074 		break;
4075 	default:
4076 		break;
4077 	}
4078 
4079 	adapter->wvbr |= wvbr;
4080 }
4081 
4082 #define IGB_STAGGERED_QUEUE_OFFSET 8
4083 
4084 static void igb_spoof_check(struct igb_adapter *adapter)
4085 {
4086 	int j;
4087 
4088 	if (!adapter->wvbr)
4089 		return;
4090 
4091 	for (j = 0; j < adapter->vfs_allocated_count; j++) {
4092 		if (adapter->wvbr & (1 << j) ||
4093 		    adapter->wvbr & (1 << (j + IGB_STAGGERED_QUEUE_OFFSET))) {
4094 			dev_warn(&adapter->pdev->dev,
4095 				"Spoof event(s) detected on VF %d\n", j);
4096 			adapter->wvbr &=
4097 				~((1 << j) |
4098 				  (1 << (j + IGB_STAGGERED_QUEUE_OFFSET)));
4099 		}
4100 	}
4101 }
4102 
4103 /* Need to wait a few seconds after link up to get diagnostic information from
4104  * the phy
4105  */
4106 static void igb_update_phy_info(unsigned long data)
4107 {
4108 	struct igb_adapter *adapter = (struct igb_adapter *) data;
4109 	igb_get_phy_info(&adapter->hw);
4110 }
4111 
4112 /**
4113  *  igb_has_link - check shared code for link and determine up/down
4114  *  @adapter: pointer to driver private info
4115  **/
4116 bool igb_has_link(struct igb_adapter *adapter)
4117 {
4118 	struct e1000_hw *hw = &adapter->hw;
4119 	bool link_active = false;
4120 
4121 	/* get_link_status is set on LSC (link status) interrupt or
4122 	 * rx sequence error interrupt.  get_link_status will stay
4123 	 * false until the e1000_check_for_link establishes link
4124 	 * for copper adapters ONLY
4125 	 */
4126 	switch (hw->phy.media_type) {
4127 	case e1000_media_type_copper:
4128 		if (!hw->mac.get_link_status)
4129 			return true;
4130 	case e1000_media_type_internal_serdes:
4131 		hw->mac.ops.check_for_link(hw);
4132 		link_active = !hw->mac.get_link_status;
4133 		break;
4134 	default:
4135 	case e1000_media_type_unknown:
4136 		break;
4137 	}
4138 
4139 	if (((hw->mac.type == e1000_i210) ||
4140 	     (hw->mac.type == e1000_i211)) &&
4141 	     (hw->phy.id == I210_I_PHY_ID)) {
4142 		if (!netif_carrier_ok(adapter->netdev)) {
4143 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4144 		} else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
4145 			adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
4146 			adapter->link_check_timeout = jiffies;
4147 		}
4148 	}
4149 
4150 	return link_active;
4151 }
4152 
4153 static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
4154 {
4155 	bool ret = false;
4156 	u32 ctrl_ext, thstat;
4157 
4158 	/* check for thermal sensor event on i350 copper only */
4159 	if (hw->mac.type == e1000_i350) {
4160 		thstat = rd32(E1000_THSTAT);
4161 		ctrl_ext = rd32(E1000_CTRL_EXT);
4162 
4163 		if ((hw->phy.media_type == e1000_media_type_copper) &&
4164 		    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
4165 			ret = !!(thstat & event);
4166 	}
4167 
4168 	return ret;
4169 }
4170 
4171 /**
4172  *  igb_check_lvmmc - check for malformed packets received
4173  *  and indicated in LVMMC register
4174  *  @adapter: pointer to adapter
4175  **/
4176 static void igb_check_lvmmc(struct igb_adapter *adapter)
4177 {
4178 	struct e1000_hw *hw = &adapter->hw;
4179 	u32 lvmmc;
4180 
4181 	lvmmc = rd32(E1000_LVMMC);
4182 	if (lvmmc) {
4183 		if (unlikely(net_ratelimit())) {
4184 			netdev_warn(adapter->netdev,
4185 				    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
4186 				    lvmmc);
4187 		}
4188 	}
4189 }
4190 
4191 /**
4192  *  igb_watchdog - Timer Call-back
4193  *  @data: pointer to adapter cast into an unsigned long
4194  **/
4195 static void igb_watchdog(unsigned long data)
4196 {
4197 	struct igb_adapter *adapter = (struct igb_adapter *)data;
4198 	/* Do the rest outside of interrupt context */
4199 	schedule_work(&adapter->watchdog_task);
4200 }
4201 
4202 static void igb_watchdog_task(struct work_struct *work)
4203 {
4204 	struct igb_adapter *adapter = container_of(work,
4205 						   struct igb_adapter,
4206 						   watchdog_task);
4207 	struct e1000_hw *hw = &adapter->hw;
4208 	struct e1000_phy_info *phy = &hw->phy;
4209 	struct net_device *netdev = adapter->netdev;
4210 	u32 link;
4211 	int i;
4212 	u32 connsw;
4213 
4214 	link = igb_has_link(adapter);
4215 
4216 	if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
4217 		if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
4218 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4219 		else
4220 			link = false;
4221 	}
4222 
4223 	/* Force link down if we have fiber to swap to */
4224 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4225 		if (hw->phy.media_type == e1000_media_type_copper) {
4226 			connsw = rd32(E1000_CONNSW);
4227 			if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
4228 				link = 0;
4229 		}
4230 	}
4231 	if (link) {
4232 		/* Perform a reset if the media type changed. */
4233 		if (hw->dev_spec._82575.media_changed) {
4234 			hw->dev_spec._82575.media_changed = false;
4235 			adapter->flags |= IGB_FLAG_MEDIA_RESET;
4236 			igb_reset(adapter);
4237 		}
4238 		/* Cancel scheduled suspend requests. */
4239 		pm_runtime_resume(netdev->dev.parent);
4240 
4241 		if (!netif_carrier_ok(netdev)) {
4242 			u32 ctrl;
4243 
4244 			hw->mac.ops.get_speed_and_duplex(hw,
4245 							 &adapter->link_speed,
4246 							 &adapter->link_duplex);
4247 
4248 			ctrl = rd32(E1000_CTRL);
4249 			/* Links status message must follow this format */
4250 			netdev_info(netdev,
4251 			       "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4252 			       netdev->name,
4253 			       adapter->link_speed,
4254 			       adapter->link_duplex == FULL_DUPLEX ?
4255 			       "Full" : "Half",
4256 			       (ctrl & E1000_CTRL_TFCE) &&
4257 			       (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
4258 			       (ctrl & E1000_CTRL_RFCE) ?  "RX" :
4259 			       (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
4260 
4261 			/* disable EEE if enabled */
4262 			if ((adapter->flags & IGB_FLAG_EEE) &&
4263 				(adapter->link_duplex == HALF_DUPLEX)) {
4264 				dev_info(&adapter->pdev->dev,
4265 				"EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
4266 				adapter->hw.dev_spec._82575.eee_disable = true;
4267 				adapter->flags &= ~IGB_FLAG_EEE;
4268 			}
4269 
4270 			/* check if SmartSpeed worked */
4271 			igb_check_downshift(hw);
4272 			if (phy->speed_downgraded)
4273 				netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
4274 
4275 			/* check for thermal sensor event */
4276 			if (igb_thermal_sensor_event(hw,
4277 			    E1000_THSTAT_LINK_THROTTLE))
4278 				netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
4279 
4280 			/* adjust timeout factor according to speed/duplex */
4281 			adapter->tx_timeout_factor = 1;
4282 			switch (adapter->link_speed) {
4283 			case SPEED_10:
4284 				adapter->tx_timeout_factor = 14;
4285 				break;
4286 			case SPEED_100:
4287 				/* maybe add some timeout factor ? */
4288 				break;
4289 			}
4290 
4291 			netif_carrier_on(netdev);
4292 
4293 			igb_ping_all_vfs(adapter);
4294 			igb_check_vf_rate_limit(adapter);
4295 
4296 			/* link state has changed, schedule phy info update */
4297 			if (!test_bit(__IGB_DOWN, &adapter->state))
4298 				mod_timer(&adapter->phy_info_timer,
4299 					  round_jiffies(jiffies + 2 * HZ));
4300 		}
4301 	} else {
4302 		if (netif_carrier_ok(netdev)) {
4303 			adapter->link_speed = 0;
4304 			adapter->link_duplex = 0;
4305 
4306 			/* check for thermal sensor event */
4307 			if (igb_thermal_sensor_event(hw,
4308 			    E1000_THSTAT_PWR_DOWN)) {
4309 				netdev_err(netdev, "The network adapter was stopped because it overheated\n");
4310 			}
4311 
4312 			/* Links status message must follow this format */
4313 			netdev_info(netdev, "igb: %s NIC Link is Down\n",
4314 			       netdev->name);
4315 			netif_carrier_off(netdev);
4316 
4317 			igb_ping_all_vfs(adapter);
4318 
4319 			/* link state has changed, schedule phy info update */
4320 			if (!test_bit(__IGB_DOWN, &adapter->state))
4321 				mod_timer(&adapter->phy_info_timer,
4322 					  round_jiffies(jiffies + 2 * HZ));
4323 
4324 			/* link is down, time to check for alternate media */
4325 			if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4326 				igb_check_swap_media(adapter);
4327 				if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
4328 					schedule_work(&adapter->reset_task);
4329 					/* return immediately */
4330 					return;
4331 				}
4332 			}
4333 			pm_schedule_suspend(netdev->dev.parent,
4334 					    MSEC_PER_SEC * 5);
4335 
4336 		/* also check for alternate media here */
4337 		} else if (!netif_carrier_ok(netdev) &&
4338 			   (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
4339 			igb_check_swap_media(adapter);
4340 			if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
4341 				schedule_work(&adapter->reset_task);
4342 				/* return immediately */
4343 				return;
4344 			}
4345 		}
4346 	}
4347 
4348 	spin_lock(&adapter->stats64_lock);
4349 	igb_update_stats(adapter, &adapter->stats64);
4350 	spin_unlock(&adapter->stats64_lock);
4351 
4352 	for (i = 0; i < adapter->num_tx_queues; i++) {
4353 		struct igb_ring *tx_ring = adapter->tx_ring[i];
4354 		if (!netif_carrier_ok(netdev)) {
4355 			/* We've lost link, so the controller stops DMA,
4356 			 * but we've got queued Tx work that's never going
4357 			 * to get done, so reset controller to flush Tx.
4358 			 * (Do the reset outside of interrupt context).
4359 			 */
4360 			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
4361 				adapter->tx_timeout_count++;
4362 				schedule_work(&adapter->reset_task);
4363 				/* return immediately since reset is imminent */
4364 				return;
4365 			}
4366 		}
4367 
4368 		/* Force detection of hung controller every watchdog period */
4369 		set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
4370 	}
4371 
4372 	/* Cause software interrupt to ensure Rx ring is cleaned */
4373 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
4374 		u32 eics = 0;
4375 
4376 		for (i = 0; i < adapter->num_q_vectors; i++)
4377 			eics |= adapter->q_vector[i]->eims_value;
4378 		wr32(E1000_EICS, eics);
4379 	} else {
4380 		wr32(E1000_ICS, E1000_ICS_RXDMT0);
4381 	}
4382 
4383 	igb_spoof_check(adapter);
4384 	igb_ptp_rx_hang(adapter);
4385 
4386 	/* Check LVMMC register on i350/i354 only */
4387 	if ((adapter->hw.mac.type == e1000_i350) ||
4388 	    (adapter->hw.mac.type == e1000_i354))
4389 		igb_check_lvmmc(adapter);
4390 
4391 	/* Reset the timer */
4392 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
4393 		if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
4394 			mod_timer(&adapter->watchdog_timer,
4395 				  round_jiffies(jiffies +  HZ));
4396 		else
4397 			mod_timer(&adapter->watchdog_timer,
4398 				  round_jiffies(jiffies + 2 * HZ));
4399 	}
4400 }
4401 
4402 enum latency_range {
4403 	lowest_latency = 0,
4404 	low_latency = 1,
4405 	bulk_latency = 2,
4406 	latency_invalid = 255
4407 };
4408 
4409 /**
4410  *  igb_update_ring_itr - update the dynamic ITR value based on packet size
4411  *  @q_vector: pointer to q_vector
4412  *
4413  *  Stores a new ITR value based on strictly on packet size.  This
4414  *  algorithm is less sophisticated than that used in igb_update_itr,
4415  *  due to the difficulty of synchronizing statistics across multiple
4416  *  receive rings.  The divisors and thresholds used by this function
4417  *  were determined based on theoretical maximum wire speed and testing
4418  *  data, in order to minimize response time while increasing bulk
4419  *  throughput.
4420  *  This functionality is controlled by ethtool's coalescing settings.
4421  *  NOTE:  This function is called only when operating in a multiqueue
4422  *         receive environment.
4423  **/
4424 static void igb_update_ring_itr(struct igb_q_vector *q_vector)
4425 {
4426 	int new_val = q_vector->itr_val;
4427 	int avg_wire_size = 0;
4428 	struct igb_adapter *adapter = q_vector->adapter;
4429 	unsigned int packets;
4430 
4431 	/* For non-gigabit speeds, just fix the interrupt rate at 4000
4432 	 * ints/sec - ITR timer value of 120 ticks.
4433 	 */
4434 	if (adapter->link_speed != SPEED_1000) {
4435 		new_val = IGB_4K_ITR;
4436 		goto set_itr_val;
4437 	}
4438 
4439 	packets = q_vector->rx.total_packets;
4440 	if (packets)
4441 		avg_wire_size = q_vector->rx.total_bytes / packets;
4442 
4443 	packets = q_vector->tx.total_packets;
4444 	if (packets)
4445 		avg_wire_size = max_t(u32, avg_wire_size,
4446 				      q_vector->tx.total_bytes / packets);
4447 
4448 	/* if avg_wire_size isn't set no work was done */
4449 	if (!avg_wire_size)
4450 		goto clear_counts;
4451 
4452 	/* Add 24 bytes to size to account for CRC, preamble, and gap */
4453 	avg_wire_size += 24;
4454 
4455 	/* Don't starve jumbo frames */
4456 	avg_wire_size = min(avg_wire_size, 3000);
4457 
4458 	/* Give a little boost to mid-size frames */
4459 	if ((avg_wire_size > 300) && (avg_wire_size < 1200))
4460 		new_val = avg_wire_size / 3;
4461 	else
4462 		new_val = avg_wire_size / 2;
4463 
4464 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
4465 	if (new_val < IGB_20K_ITR &&
4466 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
4467 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
4468 		new_val = IGB_20K_ITR;
4469 
4470 set_itr_val:
4471 	if (new_val != q_vector->itr_val) {
4472 		q_vector->itr_val = new_val;
4473 		q_vector->set_itr = 1;
4474 	}
4475 clear_counts:
4476 	q_vector->rx.total_bytes = 0;
4477 	q_vector->rx.total_packets = 0;
4478 	q_vector->tx.total_bytes = 0;
4479 	q_vector->tx.total_packets = 0;
4480 }
4481 
4482 /**
4483  *  igb_update_itr - update the dynamic ITR value based on statistics
4484  *  @q_vector: pointer to q_vector
4485  *  @ring_container: ring info to update the itr for
4486  *
4487  *  Stores a new ITR value based on packets and byte
4488  *  counts during the last interrupt.  The advantage of per interrupt
4489  *  computation is faster updates and more accurate ITR for the current
4490  *  traffic pattern.  Constants in this function were computed
4491  *  based on theoretical maximum wire speed and thresholds were set based
4492  *  on testing data as well as attempting to minimize response time
4493  *  while increasing bulk throughput.
4494  *  This functionality is controlled by ethtool's coalescing settings.
4495  *  NOTE:  These calculations are only valid when operating in a single-
4496  *         queue environment.
4497  **/
4498 static void igb_update_itr(struct igb_q_vector *q_vector,
4499 			   struct igb_ring_container *ring_container)
4500 {
4501 	unsigned int packets = ring_container->total_packets;
4502 	unsigned int bytes = ring_container->total_bytes;
4503 	u8 itrval = ring_container->itr;
4504 
4505 	/* no packets, exit with status unchanged */
4506 	if (packets == 0)
4507 		return;
4508 
4509 	switch (itrval) {
4510 	case lowest_latency:
4511 		/* handle TSO and jumbo frames */
4512 		if (bytes/packets > 8000)
4513 			itrval = bulk_latency;
4514 		else if ((packets < 5) && (bytes > 512))
4515 			itrval = low_latency;
4516 		break;
4517 	case low_latency:  /* 50 usec aka 20000 ints/s */
4518 		if (bytes > 10000) {
4519 			/* this if handles the TSO accounting */
4520 			if (bytes/packets > 8000)
4521 				itrval = bulk_latency;
4522 			else if ((packets < 10) || ((bytes/packets) > 1200))
4523 				itrval = bulk_latency;
4524 			else if ((packets > 35))
4525 				itrval = lowest_latency;
4526 		} else if (bytes/packets > 2000) {
4527 			itrval = bulk_latency;
4528 		} else if (packets <= 2 && bytes < 512) {
4529 			itrval = lowest_latency;
4530 		}
4531 		break;
4532 	case bulk_latency: /* 250 usec aka 4000 ints/s */
4533 		if (bytes > 25000) {
4534 			if (packets > 35)
4535 				itrval = low_latency;
4536 		} else if (bytes < 1500) {
4537 			itrval = low_latency;
4538 		}
4539 		break;
4540 	}
4541 
4542 	/* clear work counters since we have the values we need */
4543 	ring_container->total_bytes = 0;
4544 	ring_container->total_packets = 0;
4545 
4546 	/* write updated itr to ring container */
4547 	ring_container->itr = itrval;
4548 }
4549 
4550 static void igb_set_itr(struct igb_q_vector *q_vector)
4551 {
4552 	struct igb_adapter *adapter = q_vector->adapter;
4553 	u32 new_itr = q_vector->itr_val;
4554 	u8 current_itr = 0;
4555 
4556 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
4557 	if (adapter->link_speed != SPEED_1000) {
4558 		current_itr = 0;
4559 		new_itr = IGB_4K_ITR;
4560 		goto set_itr_now;
4561 	}
4562 
4563 	igb_update_itr(q_vector, &q_vector->tx);
4564 	igb_update_itr(q_vector, &q_vector->rx);
4565 
4566 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
4567 
4568 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
4569 	if (current_itr == lowest_latency &&
4570 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
4571 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
4572 		current_itr = low_latency;
4573 
4574 	switch (current_itr) {
4575 	/* counts and packets in update_itr are dependent on these numbers */
4576 	case lowest_latency:
4577 		new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
4578 		break;
4579 	case low_latency:
4580 		new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
4581 		break;
4582 	case bulk_latency:
4583 		new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
4584 		break;
4585 	default:
4586 		break;
4587 	}
4588 
4589 set_itr_now:
4590 	if (new_itr != q_vector->itr_val) {
4591 		/* this attempts to bias the interrupt rate towards Bulk
4592 		 * by adding intermediate steps when interrupt rate is
4593 		 * increasing
4594 		 */
4595 		new_itr = new_itr > q_vector->itr_val ?
4596 			  max((new_itr * q_vector->itr_val) /
4597 			  (new_itr + (q_vector->itr_val >> 2)),
4598 			  new_itr) : new_itr;
4599 		/* Don't write the value here; it resets the adapter's
4600 		 * internal timer, and causes us to delay far longer than
4601 		 * we should between interrupts.  Instead, we write the ITR
4602 		 * value at the beginning of the next interrupt so the timing
4603 		 * ends up being correct.
4604 		 */
4605 		q_vector->itr_val = new_itr;
4606 		q_vector->set_itr = 1;
4607 	}
4608 }
4609 
4610 static void igb_tx_ctxtdesc(struct igb_ring *tx_ring, u32 vlan_macip_lens,
4611 			    u32 type_tucmd, u32 mss_l4len_idx)
4612 {
4613 	struct e1000_adv_tx_context_desc *context_desc;
4614 	u16 i = tx_ring->next_to_use;
4615 
4616 	context_desc = IGB_TX_CTXTDESC(tx_ring, i);
4617 
4618 	i++;
4619 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
4620 
4621 	/* set bits to identify this as an advanced context descriptor */
4622 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
4623 
4624 	/* For 82575, context index must be unique per ring. */
4625 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
4626 		mss_l4len_idx |= tx_ring->reg_idx << 4;
4627 
4628 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
4629 	context_desc->seqnum_seed	= 0;
4630 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
4631 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
4632 }
4633 
4634 static int igb_tso(struct igb_ring *tx_ring,
4635 		   struct igb_tx_buffer *first,
4636 		   u8 *hdr_len)
4637 {
4638 	struct sk_buff *skb = first->skb;
4639 	u32 vlan_macip_lens, type_tucmd;
4640 	u32 mss_l4len_idx, l4len;
4641 	int err;
4642 
4643 	if (skb->ip_summed != CHECKSUM_PARTIAL)
4644 		return 0;
4645 
4646 	if (!skb_is_gso(skb))
4647 		return 0;
4648 
4649 	err = skb_cow_head(skb, 0);
4650 	if (err < 0)
4651 		return err;
4652 
4653 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
4654 	type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
4655 
4656 	if (first->protocol == htons(ETH_P_IP)) {
4657 		struct iphdr *iph = ip_hdr(skb);
4658 		iph->tot_len = 0;
4659 		iph->check = 0;
4660 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
4661 							 iph->daddr, 0,
4662 							 IPPROTO_TCP,
4663 							 0);
4664 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
4665 		first->tx_flags |= IGB_TX_FLAGS_TSO |
4666 				   IGB_TX_FLAGS_CSUM |
4667 				   IGB_TX_FLAGS_IPV4;
4668 	} else if (skb_is_gso_v6(skb)) {
4669 		ipv6_hdr(skb)->payload_len = 0;
4670 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
4671 						       &ipv6_hdr(skb)->daddr,
4672 						       0, IPPROTO_TCP, 0);
4673 		first->tx_flags |= IGB_TX_FLAGS_TSO |
4674 				   IGB_TX_FLAGS_CSUM;
4675 	}
4676 
4677 	/* compute header lengths */
4678 	l4len = tcp_hdrlen(skb);
4679 	*hdr_len = skb_transport_offset(skb) + l4len;
4680 
4681 	/* update gso size and bytecount with header size */
4682 	first->gso_segs = skb_shinfo(skb)->gso_segs;
4683 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
4684 
4685 	/* MSS L4LEN IDX */
4686 	mss_l4len_idx = l4len << E1000_ADVTXD_L4LEN_SHIFT;
4687 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
4688 
4689 	/* VLAN MACLEN IPLEN */
4690 	vlan_macip_lens = skb_network_header_len(skb);
4691 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
4692 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
4693 
4694 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
4695 
4696 	return 1;
4697 }
4698 
4699 static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
4700 {
4701 	struct sk_buff *skb = first->skb;
4702 	u32 vlan_macip_lens = 0;
4703 	u32 mss_l4len_idx = 0;
4704 	u32 type_tucmd = 0;
4705 
4706 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
4707 		if (!(first->tx_flags & IGB_TX_FLAGS_VLAN))
4708 			return;
4709 	} else {
4710 		u8 l4_hdr = 0;
4711 
4712 		switch (first->protocol) {
4713 		case htons(ETH_P_IP):
4714 			vlan_macip_lens |= skb_network_header_len(skb);
4715 			type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
4716 			l4_hdr = ip_hdr(skb)->protocol;
4717 			break;
4718 		case htons(ETH_P_IPV6):
4719 			vlan_macip_lens |= skb_network_header_len(skb);
4720 			l4_hdr = ipv6_hdr(skb)->nexthdr;
4721 			break;
4722 		default:
4723 			if (unlikely(net_ratelimit())) {
4724 				dev_warn(tx_ring->dev,
4725 					 "partial checksum but proto=%x!\n",
4726 					 first->protocol);
4727 			}
4728 			break;
4729 		}
4730 
4731 		switch (l4_hdr) {
4732 		case IPPROTO_TCP:
4733 			type_tucmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
4734 			mss_l4len_idx = tcp_hdrlen(skb) <<
4735 					E1000_ADVTXD_L4LEN_SHIFT;
4736 			break;
4737 		case IPPROTO_SCTP:
4738 			type_tucmd |= E1000_ADVTXD_TUCMD_L4T_SCTP;
4739 			mss_l4len_idx = sizeof(struct sctphdr) <<
4740 					E1000_ADVTXD_L4LEN_SHIFT;
4741 			break;
4742 		case IPPROTO_UDP:
4743 			mss_l4len_idx = sizeof(struct udphdr) <<
4744 					E1000_ADVTXD_L4LEN_SHIFT;
4745 			break;
4746 		default:
4747 			if (unlikely(net_ratelimit())) {
4748 				dev_warn(tx_ring->dev,
4749 					 "partial checksum but l4 proto=%x!\n",
4750 					 l4_hdr);
4751 			}
4752 			break;
4753 		}
4754 
4755 		/* update TX checksum flag */
4756 		first->tx_flags |= IGB_TX_FLAGS_CSUM;
4757 	}
4758 
4759 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
4760 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
4761 
4762 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
4763 }
4764 
4765 #define IGB_SET_FLAG(_input, _flag, _result) \
4766 	((_flag <= _result) ? \
4767 	 ((u32)(_input & _flag) * (_result / _flag)) : \
4768 	 ((u32)(_input & _flag) / (_flag / _result)))
4769 
4770 static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
4771 {
4772 	/* set type for advanced descriptor with frame checksum insertion */
4773 	u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
4774 		       E1000_ADVTXD_DCMD_DEXT |
4775 		       E1000_ADVTXD_DCMD_IFCS;
4776 
4777 	/* set HW vlan bit if vlan is present */
4778 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
4779 				 (E1000_ADVTXD_DCMD_VLE));
4780 
4781 	/* set segmentation bits for TSO */
4782 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
4783 				 (E1000_ADVTXD_DCMD_TSE));
4784 
4785 	/* set timestamp bit if present */
4786 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
4787 				 (E1000_ADVTXD_MAC_TSTAMP));
4788 
4789 	/* insert frame checksum */
4790 	cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
4791 
4792 	return cmd_type;
4793 }
4794 
4795 static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
4796 				 union e1000_adv_tx_desc *tx_desc,
4797 				 u32 tx_flags, unsigned int paylen)
4798 {
4799 	u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
4800 
4801 	/* 82575 requires a unique index per ring */
4802 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
4803 		olinfo_status |= tx_ring->reg_idx << 4;
4804 
4805 	/* insert L4 checksum */
4806 	olinfo_status |= IGB_SET_FLAG(tx_flags,
4807 				      IGB_TX_FLAGS_CSUM,
4808 				      (E1000_TXD_POPTS_TXSM << 8));
4809 
4810 	/* insert IPv4 checksum */
4811 	olinfo_status |= IGB_SET_FLAG(tx_flags,
4812 				      IGB_TX_FLAGS_IPV4,
4813 				      (E1000_TXD_POPTS_IXSM << 8));
4814 
4815 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
4816 }
4817 
4818 static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
4819 {
4820 	struct net_device *netdev = tx_ring->netdev;
4821 
4822 	netif_stop_subqueue(netdev, tx_ring->queue_index);
4823 
4824 	/* Herbert's original patch had:
4825 	 *  smp_mb__after_netif_stop_queue();
4826 	 * but since that doesn't exist yet, just open code it.
4827 	 */
4828 	smp_mb();
4829 
4830 	/* We need to check again in a case another CPU has just
4831 	 * made room available.
4832 	 */
4833 	if (igb_desc_unused(tx_ring) < size)
4834 		return -EBUSY;
4835 
4836 	/* A reprieve! */
4837 	netif_wake_subqueue(netdev, tx_ring->queue_index);
4838 
4839 	u64_stats_update_begin(&tx_ring->tx_syncp2);
4840 	tx_ring->tx_stats.restart_queue2++;
4841 	u64_stats_update_end(&tx_ring->tx_syncp2);
4842 
4843 	return 0;
4844 }
4845 
4846 static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
4847 {
4848 	if (igb_desc_unused(tx_ring) >= size)
4849 		return 0;
4850 	return __igb_maybe_stop_tx(tx_ring, size);
4851 }
4852 
4853 static void igb_tx_map(struct igb_ring *tx_ring,
4854 		       struct igb_tx_buffer *first,
4855 		       const u8 hdr_len)
4856 {
4857 	struct sk_buff *skb = first->skb;
4858 	struct igb_tx_buffer *tx_buffer;
4859 	union e1000_adv_tx_desc *tx_desc;
4860 	struct skb_frag_struct *frag;
4861 	dma_addr_t dma;
4862 	unsigned int data_len, size;
4863 	u32 tx_flags = first->tx_flags;
4864 	u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
4865 	u16 i = tx_ring->next_to_use;
4866 
4867 	tx_desc = IGB_TX_DESC(tx_ring, i);
4868 
4869 	igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
4870 
4871 	size = skb_headlen(skb);
4872 	data_len = skb->data_len;
4873 
4874 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
4875 
4876 	tx_buffer = first;
4877 
4878 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
4879 		if (dma_mapping_error(tx_ring->dev, dma))
4880 			goto dma_error;
4881 
4882 		/* record length, and DMA address */
4883 		dma_unmap_len_set(tx_buffer, len, size);
4884 		dma_unmap_addr_set(tx_buffer, dma, dma);
4885 
4886 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
4887 
4888 		while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
4889 			tx_desc->read.cmd_type_len =
4890 				cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
4891 
4892 			i++;
4893 			tx_desc++;
4894 			if (i == tx_ring->count) {
4895 				tx_desc = IGB_TX_DESC(tx_ring, 0);
4896 				i = 0;
4897 			}
4898 			tx_desc->read.olinfo_status = 0;
4899 
4900 			dma += IGB_MAX_DATA_PER_TXD;
4901 			size -= IGB_MAX_DATA_PER_TXD;
4902 
4903 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
4904 		}
4905 
4906 		if (likely(!data_len))
4907 			break;
4908 
4909 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
4910 
4911 		i++;
4912 		tx_desc++;
4913 		if (i == tx_ring->count) {
4914 			tx_desc = IGB_TX_DESC(tx_ring, 0);
4915 			i = 0;
4916 		}
4917 		tx_desc->read.olinfo_status = 0;
4918 
4919 		size = skb_frag_size(frag);
4920 		data_len -= size;
4921 
4922 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
4923 				       size, DMA_TO_DEVICE);
4924 
4925 		tx_buffer = &tx_ring->tx_buffer_info[i];
4926 	}
4927 
4928 	/* write last descriptor with RS and EOP bits */
4929 	cmd_type |= size | IGB_TXD_DCMD;
4930 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
4931 
4932 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
4933 
4934 	/* set the timestamp */
4935 	first->time_stamp = jiffies;
4936 
4937 	/* Force memory writes to complete before letting h/w know there
4938 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
4939 	 * memory model archs, such as IA-64).
4940 	 *
4941 	 * We also need this memory barrier to make certain all of the
4942 	 * status bits have been updated before next_to_watch is written.
4943 	 */
4944 	wmb();
4945 
4946 	/* set next_to_watch value indicating a packet is present */
4947 	first->next_to_watch = tx_desc;
4948 
4949 	i++;
4950 	if (i == tx_ring->count)
4951 		i = 0;
4952 
4953 	tx_ring->next_to_use = i;
4954 
4955 	/* Make sure there is space in the ring for the next send. */
4956 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
4957 
4958 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
4959 		writel(i, tx_ring->tail);
4960 
4961 		/* we need this if more than one processor can write to our tail
4962 		 * at a time, it synchronizes IO on IA64/Altix systems
4963 		 */
4964 		mmiowb();
4965 	}
4966 	return;
4967 
4968 dma_error:
4969 	dev_err(tx_ring->dev, "TX DMA map failed\n");
4970 
4971 	/* clear dma mappings for failed tx_buffer_info map */
4972 	for (;;) {
4973 		tx_buffer = &tx_ring->tx_buffer_info[i];
4974 		igb_unmap_and_free_tx_resource(tx_ring, tx_buffer);
4975 		if (tx_buffer == first)
4976 			break;
4977 		if (i == 0)
4978 			i = tx_ring->count;
4979 		i--;
4980 	}
4981 
4982 	tx_ring->next_to_use = i;
4983 }
4984 
4985 netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
4986 				struct igb_ring *tx_ring)
4987 {
4988 	struct igb_tx_buffer *first;
4989 	int tso;
4990 	u32 tx_flags = 0;
4991 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
4992 	__be16 protocol = vlan_get_protocol(skb);
4993 	u8 hdr_len = 0;
4994 
4995 	/* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
4996 	 *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
4997 	 *       + 2 desc gap to keep tail from touching head,
4998 	 *       + 1 desc for context descriptor,
4999 	 * otherwise try next time
5000 	 */
5001 	if (NETDEV_FRAG_PAGE_MAX_SIZE > IGB_MAX_DATA_PER_TXD) {
5002 		unsigned short f;
5003 
5004 		for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
5005 			count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
5006 	} else {
5007 		count += skb_shinfo(skb)->nr_frags;
5008 	}
5009 
5010 	if (igb_maybe_stop_tx(tx_ring, count + 3)) {
5011 		/* this is a hard error */
5012 		return NETDEV_TX_BUSY;
5013 	}
5014 
5015 	/* record the location of the first descriptor for this packet */
5016 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
5017 	first->skb = skb;
5018 	first->bytecount = skb->len;
5019 	first->gso_segs = 1;
5020 
5021 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
5022 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
5023 
5024 		if (!test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
5025 					   &adapter->state)) {
5026 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5027 			tx_flags |= IGB_TX_FLAGS_TSTAMP;
5028 
5029 			adapter->ptp_tx_skb = skb_get(skb);
5030 			adapter->ptp_tx_start = jiffies;
5031 			if (adapter->hw.mac.type == e1000_82576)
5032 				schedule_work(&adapter->ptp_tx_work);
5033 		}
5034 	}
5035 
5036 	skb_tx_timestamp(skb);
5037 
5038 	if (skb_vlan_tag_present(skb)) {
5039 		tx_flags |= IGB_TX_FLAGS_VLAN;
5040 		tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
5041 	}
5042 
5043 	/* record initial flags and protocol */
5044 	first->tx_flags = tx_flags;
5045 	first->protocol = protocol;
5046 
5047 	tso = igb_tso(tx_ring, first, &hdr_len);
5048 	if (tso < 0)
5049 		goto out_drop;
5050 	else if (!tso)
5051 		igb_tx_csum(tx_ring, first);
5052 
5053 	igb_tx_map(tx_ring, first, hdr_len);
5054 
5055 	return NETDEV_TX_OK;
5056 
5057 out_drop:
5058 	igb_unmap_and_free_tx_resource(tx_ring, first);
5059 
5060 	return NETDEV_TX_OK;
5061 }
5062 
5063 static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
5064 						    struct sk_buff *skb)
5065 {
5066 	unsigned int r_idx = skb->queue_mapping;
5067 
5068 	if (r_idx >= adapter->num_tx_queues)
5069 		r_idx = r_idx % adapter->num_tx_queues;
5070 
5071 	return adapter->tx_ring[r_idx];
5072 }
5073 
5074 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
5075 				  struct net_device *netdev)
5076 {
5077 	struct igb_adapter *adapter = netdev_priv(netdev);
5078 
5079 	if (test_bit(__IGB_DOWN, &adapter->state)) {
5080 		dev_kfree_skb_any(skb);
5081 		return NETDEV_TX_OK;
5082 	}
5083 
5084 	if (skb->len <= 0) {
5085 		dev_kfree_skb_any(skb);
5086 		return NETDEV_TX_OK;
5087 	}
5088 
5089 	/* The minimum packet size with TCTL.PSP set is 17 so pad the skb
5090 	 * in order to meet this minimum size requirement.
5091 	 */
5092 	if (skb_put_padto(skb, 17))
5093 		return NETDEV_TX_OK;
5094 
5095 	return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
5096 }
5097 
5098 /**
5099  *  igb_tx_timeout - Respond to a Tx Hang
5100  *  @netdev: network interface device structure
5101  **/
5102 static void igb_tx_timeout(struct net_device *netdev)
5103 {
5104 	struct igb_adapter *adapter = netdev_priv(netdev);
5105 	struct e1000_hw *hw = &adapter->hw;
5106 
5107 	/* Do the reset outside of interrupt context */
5108 	adapter->tx_timeout_count++;
5109 
5110 	if (hw->mac.type >= e1000_82580)
5111 		hw->dev_spec._82575.global_device_reset = true;
5112 
5113 	schedule_work(&adapter->reset_task);
5114 	wr32(E1000_EICS,
5115 	     (adapter->eims_enable_mask & ~adapter->eims_other));
5116 }
5117 
5118 static void igb_reset_task(struct work_struct *work)
5119 {
5120 	struct igb_adapter *adapter;
5121 	adapter = container_of(work, struct igb_adapter, reset_task);
5122 
5123 	igb_dump(adapter);
5124 	netdev_err(adapter->netdev, "Reset adapter\n");
5125 	igb_reinit_locked(adapter);
5126 }
5127 
5128 /**
5129  *  igb_get_stats64 - Get System Network Statistics
5130  *  @netdev: network interface device structure
5131  *  @stats: rtnl_link_stats64 pointer
5132  **/
5133 static struct rtnl_link_stats64 *igb_get_stats64(struct net_device *netdev,
5134 						struct rtnl_link_stats64 *stats)
5135 {
5136 	struct igb_adapter *adapter = netdev_priv(netdev);
5137 
5138 	spin_lock(&adapter->stats64_lock);
5139 	igb_update_stats(adapter, &adapter->stats64);
5140 	memcpy(stats, &adapter->stats64, sizeof(*stats));
5141 	spin_unlock(&adapter->stats64_lock);
5142 
5143 	return stats;
5144 }
5145 
5146 /**
5147  *  igb_change_mtu - Change the Maximum Transfer Unit
5148  *  @netdev: network interface device structure
5149  *  @new_mtu: new value for maximum frame size
5150  *
5151  *  Returns 0 on success, negative on failure
5152  **/
5153 static int igb_change_mtu(struct net_device *netdev, int new_mtu)
5154 {
5155 	struct igb_adapter *adapter = netdev_priv(netdev);
5156 	struct pci_dev *pdev = adapter->pdev;
5157 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
5158 
5159 	if ((new_mtu < 68) || (max_frame > MAX_JUMBO_FRAME_SIZE)) {
5160 		dev_err(&pdev->dev, "Invalid MTU setting\n");
5161 		return -EINVAL;
5162 	}
5163 
5164 #define MAX_STD_JUMBO_FRAME_SIZE 9238
5165 	if (max_frame > MAX_STD_JUMBO_FRAME_SIZE) {
5166 		dev_err(&pdev->dev, "MTU > 9216 not supported.\n");
5167 		return -EINVAL;
5168 	}
5169 
5170 	/* adjust max frame to be at least the size of a standard frame */
5171 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
5172 		max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
5173 
5174 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
5175 		usleep_range(1000, 2000);
5176 
5177 	/* igb_down has a dependency on max_frame_size */
5178 	adapter->max_frame_size = max_frame;
5179 
5180 	if (netif_running(netdev))
5181 		igb_down(adapter);
5182 
5183 	dev_info(&pdev->dev, "changing MTU from %d to %d\n",
5184 		 netdev->mtu, new_mtu);
5185 	netdev->mtu = new_mtu;
5186 
5187 	if (netif_running(netdev))
5188 		igb_up(adapter);
5189 	else
5190 		igb_reset(adapter);
5191 
5192 	clear_bit(__IGB_RESETTING, &adapter->state);
5193 
5194 	return 0;
5195 }
5196 
5197 /**
5198  *  igb_update_stats - Update the board statistics counters
5199  *  @adapter: board private structure
5200  **/
5201 void igb_update_stats(struct igb_adapter *adapter,
5202 		      struct rtnl_link_stats64 *net_stats)
5203 {
5204 	struct e1000_hw *hw = &adapter->hw;
5205 	struct pci_dev *pdev = adapter->pdev;
5206 	u32 reg, mpc;
5207 	int i;
5208 	u64 bytes, packets;
5209 	unsigned int start;
5210 	u64 _bytes, _packets;
5211 
5212 	/* Prevent stats update while adapter is being reset, or if the pci
5213 	 * connection is down.
5214 	 */
5215 	if (adapter->link_speed == 0)
5216 		return;
5217 	if (pci_channel_offline(pdev))
5218 		return;
5219 
5220 	bytes = 0;
5221 	packets = 0;
5222 
5223 	rcu_read_lock();
5224 	for (i = 0; i < adapter->num_rx_queues; i++) {
5225 		struct igb_ring *ring = adapter->rx_ring[i];
5226 		u32 rqdpc = rd32(E1000_RQDPC(i));
5227 		if (hw->mac.type >= e1000_i210)
5228 			wr32(E1000_RQDPC(i), 0);
5229 
5230 		if (rqdpc) {
5231 			ring->rx_stats.drops += rqdpc;
5232 			net_stats->rx_fifo_errors += rqdpc;
5233 		}
5234 
5235 		do {
5236 			start = u64_stats_fetch_begin_irq(&ring->rx_syncp);
5237 			_bytes = ring->rx_stats.bytes;
5238 			_packets = ring->rx_stats.packets;
5239 		} while (u64_stats_fetch_retry_irq(&ring->rx_syncp, start));
5240 		bytes += _bytes;
5241 		packets += _packets;
5242 	}
5243 
5244 	net_stats->rx_bytes = bytes;
5245 	net_stats->rx_packets = packets;
5246 
5247 	bytes = 0;
5248 	packets = 0;
5249 	for (i = 0; i < adapter->num_tx_queues; i++) {
5250 		struct igb_ring *ring = adapter->tx_ring[i];
5251 		do {
5252 			start = u64_stats_fetch_begin_irq(&ring->tx_syncp);
5253 			_bytes = ring->tx_stats.bytes;
5254 			_packets = ring->tx_stats.packets;
5255 		} while (u64_stats_fetch_retry_irq(&ring->tx_syncp, start));
5256 		bytes += _bytes;
5257 		packets += _packets;
5258 	}
5259 	net_stats->tx_bytes = bytes;
5260 	net_stats->tx_packets = packets;
5261 	rcu_read_unlock();
5262 
5263 	/* read stats registers */
5264 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
5265 	adapter->stats.gprc += rd32(E1000_GPRC);
5266 	adapter->stats.gorc += rd32(E1000_GORCL);
5267 	rd32(E1000_GORCH); /* clear GORCL */
5268 	adapter->stats.bprc += rd32(E1000_BPRC);
5269 	adapter->stats.mprc += rd32(E1000_MPRC);
5270 	adapter->stats.roc += rd32(E1000_ROC);
5271 
5272 	adapter->stats.prc64 += rd32(E1000_PRC64);
5273 	adapter->stats.prc127 += rd32(E1000_PRC127);
5274 	adapter->stats.prc255 += rd32(E1000_PRC255);
5275 	adapter->stats.prc511 += rd32(E1000_PRC511);
5276 	adapter->stats.prc1023 += rd32(E1000_PRC1023);
5277 	adapter->stats.prc1522 += rd32(E1000_PRC1522);
5278 	adapter->stats.symerrs += rd32(E1000_SYMERRS);
5279 	adapter->stats.sec += rd32(E1000_SEC);
5280 
5281 	mpc = rd32(E1000_MPC);
5282 	adapter->stats.mpc += mpc;
5283 	net_stats->rx_fifo_errors += mpc;
5284 	adapter->stats.scc += rd32(E1000_SCC);
5285 	adapter->stats.ecol += rd32(E1000_ECOL);
5286 	adapter->stats.mcc += rd32(E1000_MCC);
5287 	adapter->stats.latecol += rd32(E1000_LATECOL);
5288 	adapter->stats.dc += rd32(E1000_DC);
5289 	adapter->stats.rlec += rd32(E1000_RLEC);
5290 	adapter->stats.xonrxc += rd32(E1000_XONRXC);
5291 	adapter->stats.xontxc += rd32(E1000_XONTXC);
5292 	adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
5293 	adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
5294 	adapter->stats.fcruc += rd32(E1000_FCRUC);
5295 	adapter->stats.gptc += rd32(E1000_GPTC);
5296 	adapter->stats.gotc += rd32(E1000_GOTCL);
5297 	rd32(E1000_GOTCH); /* clear GOTCL */
5298 	adapter->stats.rnbc += rd32(E1000_RNBC);
5299 	adapter->stats.ruc += rd32(E1000_RUC);
5300 	adapter->stats.rfc += rd32(E1000_RFC);
5301 	adapter->stats.rjc += rd32(E1000_RJC);
5302 	adapter->stats.tor += rd32(E1000_TORH);
5303 	adapter->stats.tot += rd32(E1000_TOTH);
5304 	adapter->stats.tpr += rd32(E1000_TPR);
5305 
5306 	adapter->stats.ptc64 += rd32(E1000_PTC64);
5307 	adapter->stats.ptc127 += rd32(E1000_PTC127);
5308 	adapter->stats.ptc255 += rd32(E1000_PTC255);
5309 	adapter->stats.ptc511 += rd32(E1000_PTC511);
5310 	adapter->stats.ptc1023 += rd32(E1000_PTC1023);
5311 	adapter->stats.ptc1522 += rd32(E1000_PTC1522);
5312 
5313 	adapter->stats.mptc += rd32(E1000_MPTC);
5314 	adapter->stats.bptc += rd32(E1000_BPTC);
5315 
5316 	adapter->stats.tpt += rd32(E1000_TPT);
5317 	adapter->stats.colc += rd32(E1000_COLC);
5318 
5319 	adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
5320 	/* read internal phy specific stats */
5321 	reg = rd32(E1000_CTRL_EXT);
5322 	if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
5323 		adapter->stats.rxerrc += rd32(E1000_RXERRC);
5324 
5325 		/* this stat has invalid values on i210/i211 */
5326 		if ((hw->mac.type != e1000_i210) &&
5327 		    (hw->mac.type != e1000_i211))
5328 			adapter->stats.tncrs += rd32(E1000_TNCRS);
5329 	}
5330 
5331 	adapter->stats.tsctc += rd32(E1000_TSCTC);
5332 	adapter->stats.tsctfc += rd32(E1000_TSCTFC);
5333 
5334 	adapter->stats.iac += rd32(E1000_IAC);
5335 	adapter->stats.icrxoc += rd32(E1000_ICRXOC);
5336 	adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
5337 	adapter->stats.icrxatc += rd32(E1000_ICRXATC);
5338 	adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
5339 	adapter->stats.ictxatc += rd32(E1000_ICTXATC);
5340 	adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
5341 	adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
5342 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
5343 
5344 	/* Fill out the OS statistics structure */
5345 	net_stats->multicast = adapter->stats.mprc;
5346 	net_stats->collisions = adapter->stats.colc;
5347 
5348 	/* Rx Errors */
5349 
5350 	/* RLEC on some newer hardware can be incorrect so build
5351 	 * our own version based on RUC and ROC
5352 	 */
5353 	net_stats->rx_errors = adapter->stats.rxerrc +
5354 		adapter->stats.crcerrs + adapter->stats.algnerrc +
5355 		adapter->stats.ruc + adapter->stats.roc +
5356 		adapter->stats.cexterr;
5357 	net_stats->rx_length_errors = adapter->stats.ruc +
5358 				      adapter->stats.roc;
5359 	net_stats->rx_crc_errors = adapter->stats.crcerrs;
5360 	net_stats->rx_frame_errors = adapter->stats.algnerrc;
5361 	net_stats->rx_missed_errors = adapter->stats.mpc;
5362 
5363 	/* Tx Errors */
5364 	net_stats->tx_errors = adapter->stats.ecol +
5365 			       adapter->stats.latecol;
5366 	net_stats->tx_aborted_errors = adapter->stats.ecol;
5367 	net_stats->tx_window_errors = adapter->stats.latecol;
5368 	net_stats->tx_carrier_errors = adapter->stats.tncrs;
5369 
5370 	/* Tx Dropped needs to be maintained elsewhere */
5371 
5372 	/* Management Stats */
5373 	adapter->stats.mgptc += rd32(E1000_MGTPTC);
5374 	adapter->stats.mgprc += rd32(E1000_MGTPRC);
5375 	adapter->stats.mgpdc += rd32(E1000_MGTPDC);
5376 
5377 	/* OS2BMC Stats */
5378 	reg = rd32(E1000_MANC);
5379 	if (reg & E1000_MANC_EN_BMC2OS) {
5380 		adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
5381 		adapter->stats.o2bspc += rd32(E1000_O2BSPC);
5382 		adapter->stats.b2ospc += rd32(E1000_B2OSPC);
5383 		adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
5384 	}
5385 }
5386 
5387 static void igb_tsync_interrupt(struct igb_adapter *adapter)
5388 {
5389 	struct e1000_hw *hw = &adapter->hw;
5390 	struct ptp_clock_event event;
5391 	struct timespec ts;
5392 	u32 ack = 0, tsauxc, sec, nsec, tsicr = rd32(E1000_TSICR);
5393 
5394 	if (tsicr & TSINTR_SYS_WRAP) {
5395 		event.type = PTP_CLOCK_PPS;
5396 		if (adapter->ptp_caps.pps)
5397 			ptp_clock_event(adapter->ptp_clock, &event);
5398 		else
5399 			dev_err(&adapter->pdev->dev, "unexpected SYS WRAP");
5400 		ack |= TSINTR_SYS_WRAP;
5401 	}
5402 
5403 	if (tsicr & E1000_TSICR_TXTS) {
5404 		/* retrieve hardware timestamp */
5405 		schedule_work(&adapter->ptp_tx_work);
5406 		ack |= E1000_TSICR_TXTS;
5407 	}
5408 
5409 	if (tsicr & TSINTR_TT0) {
5410 		spin_lock(&adapter->tmreg_lock);
5411 		ts = timespec_add(adapter->perout[0].start,
5412 				  adapter->perout[0].period);
5413 		wr32(E1000_TRGTTIML0, ts.tv_nsec);
5414 		wr32(E1000_TRGTTIMH0, ts.tv_sec);
5415 		tsauxc = rd32(E1000_TSAUXC);
5416 		tsauxc |= TSAUXC_EN_TT0;
5417 		wr32(E1000_TSAUXC, tsauxc);
5418 		adapter->perout[0].start = ts;
5419 		spin_unlock(&adapter->tmreg_lock);
5420 		ack |= TSINTR_TT0;
5421 	}
5422 
5423 	if (tsicr & TSINTR_TT1) {
5424 		spin_lock(&adapter->tmreg_lock);
5425 		ts = timespec_add(adapter->perout[1].start,
5426 				  adapter->perout[1].period);
5427 		wr32(E1000_TRGTTIML1, ts.tv_nsec);
5428 		wr32(E1000_TRGTTIMH1, ts.tv_sec);
5429 		tsauxc = rd32(E1000_TSAUXC);
5430 		tsauxc |= TSAUXC_EN_TT1;
5431 		wr32(E1000_TSAUXC, tsauxc);
5432 		adapter->perout[1].start = ts;
5433 		spin_unlock(&adapter->tmreg_lock);
5434 		ack |= TSINTR_TT1;
5435 	}
5436 
5437 	if (tsicr & TSINTR_AUTT0) {
5438 		nsec = rd32(E1000_AUXSTMPL0);
5439 		sec  = rd32(E1000_AUXSTMPH0);
5440 		event.type = PTP_CLOCK_EXTTS;
5441 		event.index = 0;
5442 		event.timestamp = sec * 1000000000ULL + nsec;
5443 		ptp_clock_event(adapter->ptp_clock, &event);
5444 		ack |= TSINTR_AUTT0;
5445 	}
5446 
5447 	if (tsicr & TSINTR_AUTT1) {
5448 		nsec = rd32(E1000_AUXSTMPL1);
5449 		sec  = rd32(E1000_AUXSTMPH1);
5450 		event.type = PTP_CLOCK_EXTTS;
5451 		event.index = 1;
5452 		event.timestamp = sec * 1000000000ULL + nsec;
5453 		ptp_clock_event(adapter->ptp_clock, &event);
5454 		ack |= TSINTR_AUTT1;
5455 	}
5456 
5457 	/* acknowledge the interrupts */
5458 	wr32(E1000_TSICR, ack);
5459 }
5460 
5461 static irqreturn_t igb_msix_other(int irq, void *data)
5462 {
5463 	struct igb_adapter *adapter = data;
5464 	struct e1000_hw *hw = &adapter->hw;
5465 	u32 icr = rd32(E1000_ICR);
5466 	/* reading ICR causes bit 31 of EICR to be cleared */
5467 
5468 	if (icr & E1000_ICR_DRSTA)
5469 		schedule_work(&adapter->reset_task);
5470 
5471 	if (icr & E1000_ICR_DOUTSYNC) {
5472 		/* HW is reporting DMA is out of sync */
5473 		adapter->stats.doosync++;
5474 		/* The DMA Out of Sync is also indication of a spoof event
5475 		 * in IOV mode. Check the Wrong VM Behavior register to
5476 		 * see if it is really a spoof event.
5477 		 */
5478 		igb_check_wvbr(adapter);
5479 	}
5480 
5481 	/* Check for a mailbox event */
5482 	if (icr & E1000_ICR_VMMB)
5483 		igb_msg_task(adapter);
5484 
5485 	if (icr & E1000_ICR_LSC) {
5486 		hw->mac.get_link_status = 1;
5487 		/* guard against interrupt when we're going down */
5488 		if (!test_bit(__IGB_DOWN, &adapter->state))
5489 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
5490 	}
5491 
5492 	if (icr & E1000_ICR_TS)
5493 		igb_tsync_interrupt(adapter);
5494 
5495 	wr32(E1000_EIMS, adapter->eims_other);
5496 
5497 	return IRQ_HANDLED;
5498 }
5499 
5500 static void igb_write_itr(struct igb_q_vector *q_vector)
5501 {
5502 	struct igb_adapter *adapter = q_vector->adapter;
5503 	u32 itr_val = q_vector->itr_val & 0x7FFC;
5504 
5505 	if (!q_vector->set_itr)
5506 		return;
5507 
5508 	if (!itr_val)
5509 		itr_val = 0x4;
5510 
5511 	if (adapter->hw.mac.type == e1000_82575)
5512 		itr_val |= itr_val << 16;
5513 	else
5514 		itr_val |= E1000_EITR_CNT_IGNR;
5515 
5516 	writel(itr_val, q_vector->itr_register);
5517 	q_vector->set_itr = 0;
5518 }
5519 
5520 static irqreturn_t igb_msix_ring(int irq, void *data)
5521 {
5522 	struct igb_q_vector *q_vector = data;
5523 
5524 	/* Write the ITR value calculated from the previous interrupt. */
5525 	igb_write_itr(q_vector);
5526 
5527 	napi_schedule(&q_vector->napi);
5528 
5529 	return IRQ_HANDLED;
5530 }
5531 
5532 #ifdef CONFIG_IGB_DCA
5533 static void igb_update_tx_dca(struct igb_adapter *adapter,
5534 			      struct igb_ring *tx_ring,
5535 			      int cpu)
5536 {
5537 	struct e1000_hw *hw = &adapter->hw;
5538 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
5539 
5540 	if (hw->mac.type != e1000_82575)
5541 		txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
5542 
5543 	/* We can enable relaxed ordering for reads, but not writes when
5544 	 * DCA is enabled.  This is due to a known issue in some chipsets
5545 	 * which will cause the DCA tag to be cleared.
5546 	 */
5547 	txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
5548 		  E1000_DCA_TXCTRL_DATA_RRO_EN |
5549 		  E1000_DCA_TXCTRL_DESC_DCA_EN;
5550 
5551 	wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
5552 }
5553 
5554 static void igb_update_rx_dca(struct igb_adapter *adapter,
5555 			      struct igb_ring *rx_ring,
5556 			      int cpu)
5557 {
5558 	struct e1000_hw *hw = &adapter->hw;
5559 	u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
5560 
5561 	if (hw->mac.type != e1000_82575)
5562 		rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
5563 
5564 	/* We can enable relaxed ordering for reads, but not writes when
5565 	 * DCA is enabled.  This is due to a known issue in some chipsets
5566 	 * which will cause the DCA tag to be cleared.
5567 	 */
5568 	rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
5569 		  E1000_DCA_RXCTRL_DESC_DCA_EN;
5570 
5571 	wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
5572 }
5573 
5574 static void igb_update_dca(struct igb_q_vector *q_vector)
5575 {
5576 	struct igb_adapter *adapter = q_vector->adapter;
5577 	int cpu = get_cpu();
5578 
5579 	if (q_vector->cpu == cpu)
5580 		goto out_no_update;
5581 
5582 	if (q_vector->tx.ring)
5583 		igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
5584 
5585 	if (q_vector->rx.ring)
5586 		igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
5587 
5588 	q_vector->cpu = cpu;
5589 out_no_update:
5590 	put_cpu();
5591 }
5592 
5593 static void igb_setup_dca(struct igb_adapter *adapter)
5594 {
5595 	struct e1000_hw *hw = &adapter->hw;
5596 	int i;
5597 
5598 	if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
5599 		return;
5600 
5601 	/* Always use CB2 mode, difference is masked in the CB driver. */
5602 	wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
5603 
5604 	for (i = 0; i < adapter->num_q_vectors; i++) {
5605 		adapter->q_vector[i]->cpu = -1;
5606 		igb_update_dca(adapter->q_vector[i]);
5607 	}
5608 }
5609 
5610 static int __igb_notify_dca(struct device *dev, void *data)
5611 {
5612 	struct net_device *netdev = dev_get_drvdata(dev);
5613 	struct igb_adapter *adapter = netdev_priv(netdev);
5614 	struct pci_dev *pdev = adapter->pdev;
5615 	struct e1000_hw *hw = &adapter->hw;
5616 	unsigned long event = *(unsigned long *)data;
5617 
5618 	switch (event) {
5619 	case DCA_PROVIDER_ADD:
5620 		/* if already enabled, don't do it again */
5621 		if (adapter->flags & IGB_FLAG_DCA_ENABLED)
5622 			break;
5623 		if (dca_add_requester(dev) == 0) {
5624 			adapter->flags |= IGB_FLAG_DCA_ENABLED;
5625 			dev_info(&pdev->dev, "DCA enabled\n");
5626 			igb_setup_dca(adapter);
5627 			break;
5628 		}
5629 		/* Fall Through since DCA is disabled. */
5630 	case DCA_PROVIDER_REMOVE:
5631 		if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
5632 			/* without this a class_device is left
5633 			 * hanging around in the sysfs model
5634 			 */
5635 			dca_remove_requester(dev);
5636 			dev_info(&pdev->dev, "DCA disabled\n");
5637 			adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
5638 			wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
5639 		}
5640 		break;
5641 	}
5642 
5643 	return 0;
5644 }
5645 
5646 static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
5647 			  void *p)
5648 {
5649 	int ret_val;
5650 
5651 	ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
5652 					 __igb_notify_dca);
5653 
5654 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
5655 }
5656 #endif /* CONFIG_IGB_DCA */
5657 
5658 #ifdef CONFIG_PCI_IOV
5659 static int igb_vf_configure(struct igb_adapter *adapter, int vf)
5660 {
5661 	unsigned char mac_addr[ETH_ALEN];
5662 
5663 	eth_zero_addr(mac_addr);
5664 	igb_set_vf_mac(adapter, vf, mac_addr);
5665 
5666 	/* By default spoof check is enabled for all VFs */
5667 	adapter->vf_data[vf].spoofchk_enabled = true;
5668 
5669 	return 0;
5670 }
5671 
5672 #endif
5673 static void igb_ping_all_vfs(struct igb_adapter *adapter)
5674 {
5675 	struct e1000_hw *hw = &adapter->hw;
5676 	u32 ping;
5677 	int i;
5678 
5679 	for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
5680 		ping = E1000_PF_CONTROL_MSG;
5681 		if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
5682 			ping |= E1000_VT_MSGTYPE_CTS;
5683 		igb_write_mbx(hw, &ping, 1, i);
5684 	}
5685 }
5686 
5687 static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
5688 {
5689 	struct e1000_hw *hw = &adapter->hw;
5690 	u32 vmolr = rd32(E1000_VMOLR(vf));
5691 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
5692 
5693 	vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
5694 			    IGB_VF_FLAG_MULTI_PROMISC);
5695 	vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
5696 
5697 	if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
5698 		vmolr |= E1000_VMOLR_MPME;
5699 		vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
5700 		*msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
5701 	} else {
5702 		/* if we have hashes and we are clearing a multicast promisc
5703 		 * flag we need to write the hashes to the MTA as this step
5704 		 * was previously skipped
5705 		 */
5706 		if (vf_data->num_vf_mc_hashes > 30) {
5707 			vmolr |= E1000_VMOLR_MPME;
5708 		} else if (vf_data->num_vf_mc_hashes) {
5709 			int j;
5710 
5711 			vmolr |= E1000_VMOLR_ROMPE;
5712 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
5713 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
5714 		}
5715 	}
5716 
5717 	wr32(E1000_VMOLR(vf), vmolr);
5718 
5719 	/* there are flags left unprocessed, likely not supported */
5720 	if (*msgbuf & E1000_VT_MSGINFO_MASK)
5721 		return -EINVAL;
5722 
5723 	return 0;
5724 }
5725 
5726 static int igb_set_vf_multicasts(struct igb_adapter *adapter,
5727 				  u32 *msgbuf, u32 vf)
5728 {
5729 	int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
5730 	u16 *hash_list = (u16 *)&msgbuf[1];
5731 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
5732 	int i;
5733 
5734 	/* salt away the number of multicast addresses assigned
5735 	 * to this VF for later use to restore when the PF multi cast
5736 	 * list changes
5737 	 */
5738 	vf_data->num_vf_mc_hashes = n;
5739 
5740 	/* only up to 30 hash values supported */
5741 	if (n > 30)
5742 		n = 30;
5743 
5744 	/* store the hashes for later use */
5745 	for (i = 0; i < n; i++)
5746 		vf_data->vf_mc_hashes[i] = hash_list[i];
5747 
5748 	/* Flush and reset the mta with the new values */
5749 	igb_set_rx_mode(adapter->netdev);
5750 
5751 	return 0;
5752 }
5753 
5754 static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
5755 {
5756 	struct e1000_hw *hw = &adapter->hw;
5757 	struct vf_data_storage *vf_data;
5758 	int i, j;
5759 
5760 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
5761 		u32 vmolr = rd32(E1000_VMOLR(i));
5762 
5763 		vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
5764 
5765 		vf_data = &adapter->vf_data[i];
5766 
5767 		if ((vf_data->num_vf_mc_hashes > 30) ||
5768 		    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
5769 			vmolr |= E1000_VMOLR_MPME;
5770 		} else if (vf_data->num_vf_mc_hashes) {
5771 			vmolr |= E1000_VMOLR_ROMPE;
5772 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
5773 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
5774 		}
5775 		wr32(E1000_VMOLR(i), vmolr);
5776 	}
5777 }
5778 
5779 static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
5780 {
5781 	struct e1000_hw *hw = &adapter->hw;
5782 	u32 pool_mask, reg, vid;
5783 	int i;
5784 
5785 	pool_mask = 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
5786 
5787 	/* Find the vlan filter for this id */
5788 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5789 		reg = rd32(E1000_VLVF(i));
5790 
5791 		/* remove the vf from the pool */
5792 		reg &= ~pool_mask;
5793 
5794 		/* if pool is empty then remove entry from vfta */
5795 		if (!(reg & E1000_VLVF_POOLSEL_MASK) &&
5796 		    (reg & E1000_VLVF_VLANID_ENABLE)) {
5797 			reg = 0;
5798 			vid = reg & E1000_VLVF_VLANID_MASK;
5799 			igb_vfta_set(hw, vid, false);
5800 		}
5801 
5802 		wr32(E1000_VLVF(i), reg);
5803 	}
5804 
5805 	adapter->vf_data[vf].vlans_enabled = 0;
5806 }
5807 
5808 static s32 igb_vlvf_set(struct igb_adapter *adapter, u32 vid, bool add, u32 vf)
5809 {
5810 	struct e1000_hw *hw = &adapter->hw;
5811 	u32 reg, i;
5812 
5813 	/* The vlvf table only exists on 82576 hardware and newer */
5814 	if (hw->mac.type < e1000_82576)
5815 		return -1;
5816 
5817 	/* we only need to do this if VMDq is enabled */
5818 	if (!adapter->vfs_allocated_count)
5819 		return -1;
5820 
5821 	/* Find the vlan filter for this id */
5822 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5823 		reg = rd32(E1000_VLVF(i));
5824 		if ((reg & E1000_VLVF_VLANID_ENABLE) &&
5825 		    vid == (reg & E1000_VLVF_VLANID_MASK))
5826 			break;
5827 	}
5828 
5829 	if (add) {
5830 		if (i == E1000_VLVF_ARRAY_SIZE) {
5831 			/* Did not find a matching VLAN ID entry that was
5832 			 * enabled.  Search for a free filter entry, i.e.
5833 			 * one without the enable bit set
5834 			 */
5835 			for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5836 				reg = rd32(E1000_VLVF(i));
5837 				if (!(reg & E1000_VLVF_VLANID_ENABLE))
5838 					break;
5839 			}
5840 		}
5841 		if (i < E1000_VLVF_ARRAY_SIZE) {
5842 			/* Found an enabled/available entry */
5843 			reg |= 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
5844 
5845 			/* if !enabled we need to set this up in vfta */
5846 			if (!(reg & E1000_VLVF_VLANID_ENABLE)) {
5847 				/* add VID to filter table */
5848 				igb_vfta_set(hw, vid, true);
5849 				reg |= E1000_VLVF_VLANID_ENABLE;
5850 			}
5851 			reg &= ~E1000_VLVF_VLANID_MASK;
5852 			reg |= vid;
5853 			wr32(E1000_VLVF(i), reg);
5854 
5855 			/* do not modify RLPML for PF devices */
5856 			if (vf >= adapter->vfs_allocated_count)
5857 				return 0;
5858 
5859 			if (!adapter->vf_data[vf].vlans_enabled) {
5860 				u32 size;
5861 
5862 				reg = rd32(E1000_VMOLR(vf));
5863 				size = reg & E1000_VMOLR_RLPML_MASK;
5864 				size += 4;
5865 				reg &= ~E1000_VMOLR_RLPML_MASK;
5866 				reg |= size;
5867 				wr32(E1000_VMOLR(vf), reg);
5868 			}
5869 
5870 			adapter->vf_data[vf].vlans_enabled++;
5871 		}
5872 	} else {
5873 		if (i < E1000_VLVF_ARRAY_SIZE) {
5874 			/* remove vf from the pool */
5875 			reg &= ~(1 << (E1000_VLVF_POOLSEL_SHIFT + vf));
5876 			/* if pool is empty then remove entry from vfta */
5877 			if (!(reg & E1000_VLVF_POOLSEL_MASK)) {
5878 				reg = 0;
5879 				igb_vfta_set(hw, vid, false);
5880 			}
5881 			wr32(E1000_VLVF(i), reg);
5882 
5883 			/* do not modify RLPML for PF devices */
5884 			if (vf >= adapter->vfs_allocated_count)
5885 				return 0;
5886 
5887 			adapter->vf_data[vf].vlans_enabled--;
5888 			if (!adapter->vf_data[vf].vlans_enabled) {
5889 				u32 size;
5890 
5891 				reg = rd32(E1000_VMOLR(vf));
5892 				size = reg & E1000_VMOLR_RLPML_MASK;
5893 				size -= 4;
5894 				reg &= ~E1000_VMOLR_RLPML_MASK;
5895 				reg |= size;
5896 				wr32(E1000_VMOLR(vf), reg);
5897 			}
5898 		}
5899 	}
5900 	return 0;
5901 }
5902 
5903 static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
5904 {
5905 	struct e1000_hw *hw = &adapter->hw;
5906 
5907 	if (vid)
5908 		wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
5909 	else
5910 		wr32(E1000_VMVIR(vf), 0);
5911 }
5912 
5913 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
5914 			       int vf, u16 vlan, u8 qos)
5915 {
5916 	int err = 0;
5917 	struct igb_adapter *adapter = netdev_priv(netdev);
5918 
5919 	if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
5920 		return -EINVAL;
5921 	if (vlan || qos) {
5922 		err = igb_vlvf_set(adapter, vlan, !!vlan, vf);
5923 		if (err)
5924 			goto out;
5925 		igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
5926 		igb_set_vmolr(adapter, vf, !vlan);
5927 		adapter->vf_data[vf].pf_vlan = vlan;
5928 		adapter->vf_data[vf].pf_qos = qos;
5929 		dev_info(&adapter->pdev->dev,
5930 			 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
5931 		if (test_bit(__IGB_DOWN, &adapter->state)) {
5932 			dev_warn(&adapter->pdev->dev,
5933 				 "The VF VLAN has been set, but the PF device is not up.\n");
5934 			dev_warn(&adapter->pdev->dev,
5935 				 "Bring the PF device up before attempting to use the VF device.\n");
5936 		}
5937 	} else {
5938 		igb_vlvf_set(adapter, adapter->vf_data[vf].pf_vlan,
5939 			     false, vf);
5940 		igb_set_vmvir(adapter, vlan, vf);
5941 		igb_set_vmolr(adapter, vf, true);
5942 		adapter->vf_data[vf].pf_vlan = 0;
5943 		adapter->vf_data[vf].pf_qos = 0;
5944 	}
5945 out:
5946 	return err;
5947 }
5948 
5949 static int igb_find_vlvf_entry(struct igb_adapter *adapter, int vid)
5950 {
5951 	struct e1000_hw *hw = &adapter->hw;
5952 	int i;
5953 	u32 reg;
5954 
5955 	/* Find the vlan filter for this id */
5956 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5957 		reg = rd32(E1000_VLVF(i));
5958 		if ((reg & E1000_VLVF_VLANID_ENABLE) &&
5959 		    vid == (reg & E1000_VLVF_VLANID_MASK))
5960 			break;
5961 	}
5962 
5963 	if (i >= E1000_VLVF_ARRAY_SIZE)
5964 		i = -1;
5965 
5966 	return i;
5967 }
5968 
5969 static int igb_set_vf_vlan(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
5970 {
5971 	struct e1000_hw *hw = &adapter->hw;
5972 	int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
5973 	int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
5974 	int err = 0;
5975 
5976 	/* If in promiscuous mode we need to make sure the PF also has
5977 	 * the VLAN filter set.
5978 	 */
5979 	if (add && (adapter->netdev->flags & IFF_PROMISC))
5980 		err = igb_vlvf_set(adapter, vid, add,
5981 				   adapter->vfs_allocated_count);
5982 	if (err)
5983 		goto out;
5984 
5985 	err = igb_vlvf_set(adapter, vid, add, vf);
5986 
5987 	if (err)
5988 		goto out;
5989 
5990 	/* Go through all the checks to see if the VLAN filter should
5991 	 * be wiped completely.
5992 	 */
5993 	if (!add && (adapter->netdev->flags & IFF_PROMISC)) {
5994 		u32 vlvf, bits;
5995 		int regndx = igb_find_vlvf_entry(adapter, vid);
5996 
5997 		if (regndx < 0)
5998 			goto out;
5999 		/* See if any other pools are set for this VLAN filter
6000 		 * entry other than the PF.
6001 		 */
6002 		vlvf = bits = rd32(E1000_VLVF(regndx));
6003 		bits &= 1 << (E1000_VLVF_POOLSEL_SHIFT +
6004 			      adapter->vfs_allocated_count);
6005 		/* If the filter was removed then ensure PF pool bit
6006 		 * is cleared if the PF only added itself to the pool
6007 		 * because the PF is in promiscuous mode.
6008 		 */
6009 		if ((vlvf & VLAN_VID_MASK) == vid &&
6010 		    !test_bit(vid, adapter->active_vlans) &&
6011 		    !bits)
6012 			igb_vlvf_set(adapter, vid, add,
6013 				     adapter->vfs_allocated_count);
6014 	}
6015 
6016 out:
6017 	return err;
6018 }
6019 
6020 static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
6021 {
6022 	/* clear flags - except flag that indicates PF has set the MAC */
6023 	adapter->vf_data[vf].flags &= IGB_VF_FLAG_PF_SET_MAC;
6024 	adapter->vf_data[vf].last_nack = jiffies;
6025 
6026 	/* reset offloads to defaults */
6027 	igb_set_vmolr(adapter, vf, true);
6028 
6029 	/* reset vlans for device */
6030 	igb_clear_vf_vfta(adapter, vf);
6031 	if (adapter->vf_data[vf].pf_vlan)
6032 		igb_ndo_set_vf_vlan(adapter->netdev, vf,
6033 				    adapter->vf_data[vf].pf_vlan,
6034 				    adapter->vf_data[vf].pf_qos);
6035 	else
6036 		igb_clear_vf_vfta(adapter, vf);
6037 
6038 	/* reset multicast table array for vf */
6039 	adapter->vf_data[vf].num_vf_mc_hashes = 0;
6040 
6041 	/* Flush and reset the mta with the new values */
6042 	igb_set_rx_mode(adapter->netdev);
6043 }
6044 
6045 static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
6046 {
6047 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6048 
6049 	/* clear mac address as we were hotplug removed/added */
6050 	if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
6051 		eth_zero_addr(vf_mac);
6052 
6053 	/* process remaining reset events */
6054 	igb_vf_reset(adapter, vf);
6055 }
6056 
6057 static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
6058 {
6059 	struct e1000_hw *hw = &adapter->hw;
6060 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6061 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
6062 	u32 reg, msgbuf[3];
6063 	u8 *addr = (u8 *)(&msgbuf[1]);
6064 
6065 	/* process all the same items cleared in a function level reset */
6066 	igb_vf_reset(adapter, vf);
6067 
6068 	/* set vf mac address */
6069 	igb_rar_set_qsel(adapter, vf_mac, rar_entry, vf);
6070 
6071 	/* enable transmit and receive for vf */
6072 	reg = rd32(E1000_VFTE);
6073 	wr32(E1000_VFTE, reg | (1 << vf));
6074 	reg = rd32(E1000_VFRE);
6075 	wr32(E1000_VFRE, reg | (1 << vf));
6076 
6077 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
6078 
6079 	/* reply to reset with ack and vf mac address */
6080 	if (!is_zero_ether_addr(vf_mac)) {
6081 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
6082 		memcpy(addr, vf_mac, ETH_ALEN);
6083 	} else {
6084 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
6085 	}
6086 	igb_write_mbx(hw, msgbuf, 3, vf);
6087 }
6088 
6089 static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
6090 {
6091 	/* The VF MAC Address is stored in a packed array of bytes
6092 	 * starting at the second 32 bit word of the msg array
6093 	 */
6094 	unsigned char *addr = (char *)&msg[1];
6095 	int err = -1;
6096 
6097 	if (is_valid_ether_addr(addr))
6098 		err = igb_set_vf_mac(adapter, vf, addr);
6099 
6100 	return err;
6101 }
6102 
6103 static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
6104 {
6105 	struct e1000_hw *hw = &adapter->hw;
6106 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6107 	u32 msg = E1000_VT_MSGTYPE_NACK;
6108 
6109 	/* if device isn't clear to send it shouldn't be reading either */
6110 	if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
6111 	    time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
6112 		igb_write_mbx(hw, &msg, 1, vf);
6113 		vf_data->last_nack = jiffies;
6114 	}
6115 }
6116 
6117 static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
6118 {
6119 	struct pci_dev *pdev = adapter->pdev;
6120 	u32 msgbuf[E1000_VFMAILBOX_SIZE];
6121 	struct e1000_hw *hw = &adapter->hw;
6122 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6123 	s32 retval;
6124 
6125 	retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf);
6126 
6127 	if (retval) {
6128 		/* if receive failed revoke VF CTS stats and restart init */
6129 		dev_err(&pdev->dev, "Error receiving message from VF\n");
6130 		vf_data->flags &= ~IGB_VF_FLAG_CTS;
6131 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
6132 			return;
6133 		goto out;
6134 	}
6135 
6136 	/* this is a message we already processed, do nothing */
6137 	if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
6138 		return;
6139 
6140 	/* until the vf completes a reset it should not be
6141 	 * allowed to start any configuration.
6142 	 */
6143 	if (msgbuf[0] == E1000_VF_RESET) {
6144 		igb_vf_reset_msg(adapter, vf);
6145 		return;
6146 	}
6147 
6148 	if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
6149 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
6150 			return;
6151 		retval = -1;
6152 		goto out;
6153 	}
6154 
6155 	switch ((msgbuf[0] & 0xFFFF)) {
6156 	case E1000_VF_SET_MAC_ADDR:
6157 		retval = -EINVAL;
6158 		if (!(vf_data->flags & IGB_VF_FLAG_PF_SET_MAC))
6159 			retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
6160 		else
6161 			dev_warn(&pdev->dev,
6162 				 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
6163 				 vf);
6164 		break;
6165 	case E1000_VF_SET_PROMISC:
6166 		retval = igb_set_vf_promisc(adapter, msgbuf, vf);
6167 		break;
6168 	case E1000_VF_SET_MULTICAST:
6169 		retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
6170 		break;
6171 	case E1000_VF_SET_LPE:
6172 		retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
6173 		break;
6174 	case E1000_VF_SET_VLAN:
6175 		retval = -1;
6176 		if (vf_data->pf_vlan)
6177 			dev_warn(&pdev->dev,
6178 				 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
6179 				 vf);
6180 		else
6181 			retval = igb_set_vf_vlan(adapter, msgbuf, vf);
6182 		break;
6183 	default:
6184 		dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
6185 		retval = -1;
6186 		break;
6187 	}
6188 
6189 	msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
6190 out:
6191 	/* notify the VF of the results of what it sent us */
6192 	if (retval)
6193 		msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
6194 	else
6195 		msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
6196 
6197 	igb_write_mbx(hw, msgbuf, 1, vf);
6198 }
6199 
6200 static void igb_msg_task(struct igb_adapter *adapter)
6201 {
6202 	struct e1000_hw *hw = &adapter->hw;
6203 	u32 vf;
6204 
6205 	for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
6206 		/* process any reset requests */
6207 		if (!igb_check_for_rst(hw, vf))
6208 			igb_vf_reset_event(adapter, vf);
6209 
6210 		/* process any messages pending */
6211 		if (!igb_check_for_msg(hw, vf))
6212 			igb_rcv_msg_from_vf(adapter, vf);
6213 
6214 		/* process any acks */
6215 		if (!igb_check_for_ack(hw, vf))
6216 			igb_rcv_ack_from_vf(adapter, vf);
6217 	}
6218 }
6219 
6220 /**
6221  *  igb_set_uta - Set unicast filter table address
6222  *  @adapter: board private structure
6223  *
6224  *  The unicast table address is a register array of 32-bit registers.
6225  *  The table is meant to be used in a way similar to how the MTA is used
6226  *  however due to certain limitations in the hardware it is necessary to
6227  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
6228  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
6229  **/
6230 static void igb_set_uta(struct igb_adapter *adapter)
6231 {
6232 	struct e1000_hw *hw = &adapter->hw;
6233 	int i;
6234 
6235 	/* The UTA table only exists on 82576 hardware and newer */
6236 	if (hw->mac.type < e1000_82576)
6237 		return;
6238 
6239 	/* we only need to do this if VMDq is enabled */
6240 	if (!adapter->vfs_allocated_count)
6241 		return;
6242 
6243 	for (i = 0; i < hw->mac.uta_reg_count; i++)
6244 		array_wr32(E1000_UTA, i, ~0);
6245 }
6246 
6247 /**
6248  *  igb_intr_msi - Interrupt Handler
6249  *  @irq: interrupt number
6250  *  @data: pointer to a network interface device structure
6251  **/
6252 static irqreturn_t igb_intr_msi(int irq, void *data)
6253 {
6254 	struct igb_adapter *adapter = data;
6255 	struct igb_q_vector *q_vector = adapter->q_vector[0];
6256 	struct e1000_hw *hw = &adapter->hw;
6257 	/* read ICR disables interrupts using IAM */
6258 	u32 icr = rd32(E1000_ICR);
6259 
6260 	igb_write_itr(q_vector);
6261 
6262 	if (icr & E1000_ICR_DRSTA)
6263 		schedule_work(&adapter->reset_task);
6264 
6265 	if (icr & E1000_ICR_DOUTSYNC) {
6266 		/* HW is reporting DMA is out of sync */
6267 		adapter->stats.doosync++;
6268 	}
6269 
6270 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
6271 		hw->mac.get_link_status = 1;
6272 		if (!test_bit(__IGB_DOWN, &adapter->state))
6273 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6274 	}
6275 
6276 	if (icr & E1000_ICR_TS)
6277 		igb_tsync_interrupt(adapter);
6278 
6279 	napi_schedule(&q_vector->napi);
6280 
6281 	return IRQ_HANDLED;
6282 }
6283 
6284 /**
6285  *  igb_intr - Legacy Interrupt Handler
6286  *  @irq: interrupt number
6287  *  @data: pointer to a network interface device structure
6288  **/
6289 static irqreturn_t igb_intr(int irq, void *data)
6290 {
6291 	struct igb_adapter *adapter = data;
6292 	struct igb_q_vector *q_vector = adapter->q_vector[0];
6293 	struct e1000_hw *hw = &adapter->hw;
6294 	/* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
6295 	 * need for the IMC write
6296 	 */
6297 	u32 icr = rd32(E1000_ICR);
6298 
6299 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
6300 	 * not set, then the adapter didn't send an interrupt
6301 	 */
6302 	if (!(icr & E1000_ICR_INT_ASSERTED))
6303 		return IRQ_NONE;
6304 
6305 	igb_write_itr(q_vector);
6306 
6307 	if (icr & E1000_ICR_DRSTA)
6308 		schedule_work(&adapter->reset_task);
6309 
6310 	if (icr & E1000_ICR_DOUTSYNC) {
6311 		/* HW is reporting DMA is out of sync */
6312 		adapter->stats.doosync++;
6313 	}
6314 
6315 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
6316 		hw->mac.get_link_status = 1;
6317 		/* guard against interrupt when we're going down */
6318 		if (!test_bit(__IGB_DOWN, &adapter->state))
6319 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6320 	}
6321 
6322 	if (icr & E1000_ICR_TS)
6323 		igb_tsync_interrupt(adapter);
6324 
6325 	napi_schedule(&q_vector->napi);
6326 
6327 	return IRQ_HANDLED;
6328 }
6329 
6330 static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
6331 {
6332 	struct igb_adapter *adapter = q_vector->adapter;
6333 	struct e1000_hw *hw = &adapter->hw;
6334 
6335 	if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
6336 	    (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
6337 		if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
6338 			igb_set_itr(q_vector);
6339 		else
6340 			igb_update_ring_itr(q_vector);
6341 	}
6342 
6343 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
6344 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
6345 			wr32(E1000_EIMS, q_vector->eims_value);
6346 		else
6347 			igb_irq_enable(adapter);
6348 	}
6349 }
6350 
6351 /**
6352  *  igb_poll - NAPI Rx polling callback
6353  *  @napi: napi polling structure
6354  *  @budget: count of how many packets we should handle
6355  **/
6356 static int igb_poll(struct napi_struct *napi, int budget)
6357 {
6358 	struct igb_q_vector *q_vector = container_of(napi,
6359 						     struct igb_q_vector,
6360 						     napi);
6361 	bool clean_complete = true;
6362 
6363 #ifdef CONFIG_IGB_DCA
6364 	if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
6365 		igb_update_dca(q_vector);
6366 #endif
6367 	if (q_vector->tx.ring)
6368 		clean_complete = igb_clean_tx_irq(q_vector);
6369 
6370 	if (q_vector->rx.ring)
6371 		clean_complete &= igb_clean_rx_irq(q_vector, budget);
6372 
6373 	/* If all work not completed, return budget and keep polling */
6374 	if (!clean_complete)
6375 		return budget;
6376 
6377 	/* If not enough Rx work done, exit the polling mode */
6378 	napi_complete(napi);
6379 	igb_ring_irq_enable(q_vector);
6380 
6381 	return 0;
6382 }
6383 
6384 /**
6385  *  igb_clean_tx_irq - Reclaim resources after transmit completes
6386  *  @q_vector: pointer to q_vector containing needed info
6387  *
6388  *  returns true if ring is completely cleaned
6389  **/
6390 static bool igb_clean_tx_irq(struct igb_q_vector *q_vector)
6391 {
6392 	struct igb_adapter *adapter = q_vector->adapter;
6393 	struct igb_ring *tx_ring = q_vector->tx.ring;
6394 	struct igb_tx_buffer *tx_buffer;
6395 	union e1000_adv_tx_desc *tx_desc;
6396 	unsigned int total_bytes = 0, total_packets = 0;
6397 	unsigned int budget = q_vector->tx.work_limit;
6398 	unsigned int i = tx_ring->next_to_clean;
6399 
6400 	if (test_bit(__IGB_DOWN, &adapter->state))
6401 		return true;
6402 
6403 	tx_buffer = &tx_ring->tx_buffer_info[i];
6404 	tx_desc = IGB_TX_DESC(tx_ring, i);
6405 	i -= tx_ring->count;
6406 
6407 	do {
6408 		union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
6409 
6410 		/* if next_to_watch is not set then there is no work pending */
6411 		if (!eop_desc)
6412 			break;
6413 
6414 		/* prevent any other reads prior to eop_desc */
6415 		read_barrier_depends();
6416 
6417 		/* if DD is not set pending work has not been completed */
6418 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
6419 			break;
6420 
6421 		/* clear next_to_watch to prevent false hangs */
6422 		tx_buffer->next_to_watch = NULL;
6423 
6424 		/* update the statistics for this packet */
6425 		total_bytes += tx_buffer->bytecount;
6426 		total_packets += tx_buffer->gso_segs;
6427 
6428 		/* free the skb */
6429 		dev_consume_skb_any(tx_buffer->skb);
6430 
6431 		/* unmap skb header data */
6432 		dma_unmap_single(tx_ring->dev,
6433 				 dma_unmap_addr(tx_buffer, dma),
6434 				 dma_unmap_len(tx_buffer, len),
6435 				 DMA_TO_DEVICE);
6436 
6437 		/* clear tx_buffer data */
6438 		tx_buffer->skb = NULL;
6439 		dma_unmap_len_set(tx_buffer, len, 0);
6440 
6441 		/* clear last DMA location and unmap remaining buffers */
6442 		while (tx_desc != eop_desc) {
6443 			tx_buffer++;
6444 			tx_desc++;
6445 			i++;
6446 			if (unlikely(!i)) {
6447 				i -= tx_ring->count;
6448 				tx_buffer = tx_ring->tx_buffer_info;
6449 				tx_desc = IGB_TX_DESC(tx_ring, 0);
6450 			}
6451 
6452 			/* unmap any remaining paged data */
6453 			if (dma_unmap_len(tx_buffer, len)) {
6454 				dma_unmap_page(tx_ring->dev,
6455 					       dma_unmap_addr(tx_buffer, dma),
6456 					       dma_unmap_len(tx_buffer, len),
6457 					       DMA_TO_DEVICE);
6458 				dma_unmap_len_set(tx_buffer, len, 0);
6459 			}
6460 		}
6461 
6462 		/* move us one more past the eop_desc for start of next pkt */
6463 		tx_buffer++;
6464 		tx_desc++;
6465 		i++;
6466 		if (unlikely(!i)) {
6467 			i -= tx_ring->count;
6468 			tx_buffer = tx_ring->tx_buffer_info;
6469 			tx_desc = IGB_TX_DESC(tx_ring, 0);
6470 		}
6471 
6472 		/* issue prefetch for next Tx descriptor */
6473 		prefetch(tx_desc);
6474 
6475 		/* update budget accounting */
6476 		budget--;
6477 	} while (likely(budget));
6478 
6479 	netdev_tx_completed_queue(txring_txq(tx_ring),
6480 				  total_packets, total_bytes);
6481 	i += tx_ring->count;
6482 	tx_ring->next_to_clean = i;
6483 	u64_stats_update_begin(&tx_ring->tx_syncp);
6484 	tx_ring->tx_stats.bytes += total_bytes;
6485 	tx_ring->tx_stats.packets += total_packets;
6486 	u64_stats_update_end(&tx_ring->tx_syncp);
6487 	q_vector->tx.total_bytes += total_bytes;
6488 	q_vector->tx.total_packets += total_packets;
6489 
6490 	if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
6491 		struct e1000_hw *hw = &adapter->hw;
6492 
6493 		/* Detect a transmit hang in hardware, this serializes the
6494 		 * check with the clearing of time_stamp and movement of i
6495 		 */
6496 		clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
6497 		if (tx_buffer->next_to_watch &&
6498 		    time_after(jiffies, tx_buffer->time_stamp +
6499 			       (adapter->tx_timeout_factor * HZ)) &&
6500 		    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
6501 
6502 			/* detected Tx unit hang */
6503 			dev_err(tx_ring->dev,
6504 				"Detected Tx Unit Hang\n"
6505 				"  Tx Queue             <%d>\n"
6506 				"  TDH                  <%x>\n"
6507 				"  TDT                  <%x>\n"
6508 				"  next_to_use          <%x>\n"
6509 				"  next_to_clean        <%x>\n"
6510 				"buffer_info[next_to_clean]\n"
6511 				"  time_stamp           <%lx>\n"
6512 				"  next_to_watch        <%p>\n"
6513 				"  jiffies              <%lx>\n"
6514 				"  desc.status          <%x>\n",
6515 				tx_ring->queue_index,
6516 				rd32(E1000_TDH(tx_ring->reg_idx)),
6517 				readl(tx_ring->tail),
6518 				tx_ring->next_to_use,
6519 				tx_ring->next_to_clean,
6520 				tx_buffer->time_stamp,
6521 				tx_buffer->next_to_watch,
6522 				jiffies,
6523 				tx_buffer->next_to_watch->wb.status);
6524 			netif_stop_subqueue(tx_ring->netdev,
6525 					    tx_ring->queue_index);
6526 
6527 			/* we are about to reset, no point in enabling stuff */
6528 			return true;
6529 		}
6530 	}
6531 
6532 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
6533 	if (unlikely(total_packets &&
6534 	    netif_carrier_ok(tx_ring->netdev) &&
6535 	    igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
6536 		/* Make sure that anybody stopping the queue after this
6537 		 * sees the new next_to_clean.
6538 		 */
6539 		smp_mb();
6540 		if (__netif_subqueue_stopped(tx_ring->netdev,
6541 					     tx_ring->queue_index) &&
6542 		    !(test_bit(__IGB_DOWN, &adapter->state))) {
6543 			netif_wake_subqueue(tx_ring->netdev,
6544 					    tx_ring->queue_index);
6545 
6546 			u64_stats_update_begin(&tx_ring->tx_syncp);
6547 			tx_ring->tx_stats.restart_queue++;
6548 			u64_stats_update_end(&tx_ring->tx_syncp);
6549 		}
6550 	}
6551 
6552 	return !!budget;
6553 }
6554 
6555 /**
6556  *  igb_reuse_rx_page - page flip buffer and store it back on the ring
6557  *  @rx_ring: rx descriptor ring to store buffers on
6558  *  @old_buff: donor buffer to have page reused
6559  *
6560  *  Synchronizes page for reuse by the adapter
6561  **/
6562 static void igb_reuse_rx_page(struct igb_ring *rx_ring,
6563 			      struct igb_rx_buffer *old_buff)
6564 {
6565 	struct igb_rx_buffer *new_buff;
6566 	u16 nta = rx_ring->next_to_alloc;
6567 
6568 	new_buff = &rx_ring->rx_buffer_info[nta];
6569 
6570 	/* update, and store next to alloc */
6571 	nta++;
6572 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
6573 
6574 	/* transfer page from old buffer to new buffer */
6575 	*new_buff = *old_buff;
6576 
6577 	/* sync the buffer for use by the device */
6578 	dma_sync_single_range_for_device(rx_ring->dev, old_buff->dma,
6579 					 old_buff->page_offset,
6580 					 IGB_RX_BUFSZ,
6581 					 DMA_FROM_DEVICE);
6582 }
6583 
6584 static inline bool igb_page_is_reserved(struct page *page)
6585 {
6586 	return (page_to_nid(page) != numa_mem_id()) || page->pfmemalloc;
6587 }
6588 
6589 static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer,
6590 				  struct page *page,
6591 				  unsigned int truesize)
6592 {
6593 	/* avoid re-using remote pages */
6594 	if (unlikely(igb_page_is_reserved(page)))
6595 		return false;
6596 
6597 #if (PAGE_SIZE < 8192)
6598 	/* if we are only owner of page we can reuse it */
6599 	if (unlikely(page_count(page) != 1))
6600 		return false;
6601 
6602 	/* flip page offset to other buffer */
6603 	rx_buffer->page_offset ^= IGB_RX_BUFSZ;
6604 #else
6605 	/* move offset up to the next cache line */
6606 	rx_buffer->page_offset += truesize;
6607 
6608 	if (rx_buffer->page_offset > (PAGE_SIZE - IGB_RX_BUFSZ))
6609 		return false;
6610 #endif
6611 
6612 	/* Even if we own the page, we are not allowed to use atomic_set()
6613 	 * This would break get_page_unless_zero() users.
6614 	 */
6615 	atomic_inc(&page->_count);
6616 
6617 	return true;
6618 }
6619 
6620 /**
6621  *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
6622  *  @rx_ring: rx descriptor ring to transact packets on
6623  *  @rx_buffer: buffer containing page to add
6624  *  @rx_desc: descriptor containing length of buffer written by hardware
6625  *  @skb: sk_buff to place the data into
6626  *
6627  *  This function will add the data contained in rx_buffer->page to the skb.
6628  *  This is done either through a direct copy if the data in the buffer is
6629  *  less than the skb header size, otherwise it will just attach the page as
6630  *  a frag to the skb.
6631  *
6632  *  The function will then update the page offset if necessary and return
6633  *  true if the buffer can be reused by the adapter.
6634  **/
6635 static bool igb_add_rx_frag(struct igb_ring *rx_ring,
6636 			    struct igb_rx_buffer *rx_buffer,
6637 			    union e1000_adv_rx_desc *rx_desc,
6638 			    struct sk_buff *skb)
6639 {
6640 	struct page *page = rx_buffer->page;
6641 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
6642 #if (PAGE_SIZE < 8192)
6643 	unsigned int truesize = IGB_RX_BUFSZ;
6644 #else
6645 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
6646 #endif
6647 
6648 	if ((size <= IGB_RX_HDR_LEN) && !skb_is_nonlinear(skb)) {
6649 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
6650 
6651 		if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
6652 			igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
6653 			va += IGB_TS_HDR_LEN;
6654 			size -= IGB_TS_HDR_LEN;
6655 		}
6656 
6657 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
6658 
6659 		/* page is not reserved, we can reuse buffer as-is */
6660 		if (likely(!igb_page_is_reserved(page)))
6661 			return true;
6662 
6663 		/* this page cannot be reused so discard it */
6664 		__free_page(page);
6665 		return false;
6666 	}
6667 
6668 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
6669 			rx_buffer->page_offset, size, truesize);
6670 
6671 	return igb_can_reuse_rx_page(rx_buffer, page, truesize);
6672 }
6673 
6674 static struct sk_buff *igb_fetch_rx_buffer(struct igb_ring *rx_ring,
6675 					   union e1000_adv_rx_desc *rx_desc,
6676 					   struct sk_buff *skb)
6677 {
6678 	struct igb_rx_buffer *rx_buffer;
6679 	struct page *page;
6680 
6681 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
6682 	page = rx_buffer->page;
6683 	prefetchw(page);
6684 
6685 	if (likely(!skb)) {
6686 		void *page_addr = page_address(page) +
6687 				  rx_buffer->page_offset;
6688 
6689 		/* prefetch first cache line of first page */
6690 		prefetch(page_addr);
6691 #if L1_CACHE_BYTES < 128
6692 		prefetch(page_addr + L1_CACHE_BYTES);
6693 #endif
6694 
6695 		/* allocate a skb to store the frags */
6696 		skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
6697 		if (unlikely(!skb)) {
6698 			rx_ring->rx_stats.alloc_failed++;
6699 			return NULL;
6700 		}
6701 
6702 		/* we will be copying header into skb->data in
6703 		 * pskb_may_pull so it is in our interest to prefetch
6704 		 * it now to avoid a possible cache miss
6705 		 */
6706 		prefetchw(skb->data);
6707 	}
6708 
6709 	/* we are reusing so sync this buffer for CPU use */
6710 	dma_sync_single_range_for_cpu(rx_ring->dev,
6711 				      rx_buffer->dma,
6712 				      rx_buffer->page_offset,
6713 				      IGB_RX_BUFSZ,
6714 				      DMA_FROM_DEVICE);
6715 
6716 	/* pull page into skb */
6717 	if (igb_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
6718 		/* hand second half of page back to the ring */
6719 		igb_reuse_rx_page(rx_ring, rx_buffer);
6720 	} else {
6721 		/* we are not reusing the buffer so unmap it */
6722 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
6723 			       PAGE_SIZE, DMA_FROM_DEVICE);
6724 	}
6725 
6726 	/* clear contents of rx_buffer */
6727 	rx_buffer->page = NULL;
6728 
6729 	return skb;
6730 }
6731 
6732 static inline void igb_rx_checksum(struct igb_ring *ring,
6733 				   union e1000_adv_rx_desc *rx_desc,
6734 				   struct sk_buff *skb)
6735 {
6736 	skb_checksum_none_assert(skb);
6737 
6738 	/* Ignore Checksum bit is set */
6739 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
6740 		return;
6741 
6742 	/* Rx checksum disabled via ethtool */
6743 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
6744 		return;
6745 
6746 	/* TCP/UDP checksum error bit is set */
6747 	if (igb_test_staterr(rx_desc,
6748 			     E1000_RXDEXT_STATERR_TCPE |
6749 			     E1000_RXDEXT_STATERR_IPE)) {
6750 		/* work around errata with sctp packets where the TCPE aka
6751 		 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
6752 		 * packets, (aka let the stack check the crc32c)
6753 		 */
6754 		if (!((skb->len == 60) &&
6755 		      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
6756 			u64_stats_update_begin(&ring->rx_syncp);
6757 			ring->rx_stats.csum_err++;
6758 			u64_stats_update_end(&ring->rx_syncp);
6759 		}
6760 		/* let the stack verify checksum errors */
6761 		return;
6762 	}
6763 	/* It must be a TCP or UDP packet with a valid checksum */
6764 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
6765 				      E1000_RXD_STAT_UDPCS))
6766 		skb->ip_summed = CHECKSUM_UNNECESSARY;
6767 
6768 	dev_dbg(ring->dev, "cksum success: bits %08X\n",
6769 		le32_to_cpu(rx_desc->wb.upper.status_error));
6770 }
6771 
6772 static inline void igb_rx_hash(struct igb_ring *ring,
6773 			       union e1000_adv_rx_desc *rx_desc,
6774 			       struct sk_buff *skb)
6775 {
6776 	if (ring->netdev->features & NETIF_F_RXHASH)
6777 		skb_set_hash(skb,
6778 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
6779 			     PKT_HASH_TYPE_L3);
6780 }
6781 
6782 /**
6783  *  igb_is_non_eop - process handling of non-EOP buffers
6784  *  @rx_ring: Rx ring being processed
6785  *  @rx_desc: Rx descriptor for current buffer
6786  *  @skb: current socket buffer containing buffer in progress
6787  *
6788  *  This function updates next to clean.  If the buffer is an EOP buffer
6789  *  this function exits returning false, otherwise it will place the
6790  *  sk_buff in the next buffer to be chained and return true indicating
6791  *  that this is in fact a non-EOP buffer.
6792  **/
6793 static bool igb_is_non_eop(struct igb_ring *rx_ring,
6794 			   union e1000_adv_rx_desc *rx_desc)
6795 {
6796 	u32 ntc = rx_ring->next_to_clean + 1;
6797 
6798 	/* fetch, update, and store next to clean */
6799 	ntc = (ntc < rx_ring->count) ? ntc : 0;
6800 	rx_ring->next_to_clean = ntc;
6801 
6802 	prefetch(IGB_RX_DESC(rx_ring, ntc));
6803 
6804 	if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
6805 		return false;
6806 
6807 	return true;
6808 }
6809 
6810 /**
6811  *  igb_pull_tail - igb specific version of skb_pull_tail
6812  *  @rx_ring: rx descriptor ring packet is being transacted on
6813  *  @rx_desc: pointer to the EOP Rx descriptor
6814  *  @skb: pointer to current skb being adjusted
6815  *
6816  *  This function is an igb specific version of __pskb_pull_tail.  The
6817  *  main difference between this version and the original function is that
6818  *  this function can make several assumptions about the state of things
6819  *  that allow for significant optimizations versus the standard function.
6820  *  As a result we can do things like drop a frag and maintain an accurate
6821  *  truesize for the skb.
6822  */
6823 static void igb_pull_tail(struct igb_ring *rx_ring,
6824 			  union e1000_adv_rx_desc *rx_desc,
6825 			  struct sk_buff *skb)
6826 {
6827 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
6828 	unsigned char *va;
6829 	unsigned int pull_len;
6830 
6831 	/* it is valid to use page_address instead of kmap since we are
6832 	 * working with pages allocated out of the lomem pool per
6833 	 * alloc_page(GFP_ATOMIC)
6834 	 */
6835 	va = skb_frag_address(frag);
6836 
6837 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
6838 		/* retrieve timestamp from buffer */
6839 		igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
6840 
6841 		/* update pointers to remove timestamp header */
6842 		skb_frag_size_sub(frag, IGB_TS_HDR_LEN);
6843 		frag->page_offset += IGB_TS_HDR_LEN;
6844 		skb->data_len -= IGB_TS_HDR_LEN;
6845 		skb->len -= IGB_TS_HDR_LEN;
6846 
6847 		/* move va to start of packet data */
6848 		va += IGB_TS_HDR_LEN;
6849 	}
6850 
6851 	/* we need the header to contain the greater of either ETH_HLEN or
6852 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
6853 	 */
6854 	pull_len = eth_get_headlen(va, IGB_RX_HDR_LEN);
6855 
6856 	/* align pull length to size of long to optimize memcpy performance */
6857 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
6858 
6859 	/* update all of the pointers */
6860 	skb_frag_size_sub(frag, pull_len);
6861 	frag->page_offset += pull_len;
6862 	skb->data_len -= pull_len;
6863 	skb->tail += pull_len;
6864 }
6865 
6866 /**
6867  *  igb_cleanup_headers - Correct corrupted or empty headers
6868  *  @rx_ring: rx descriptor ring packet is being transacted on
6869  *  @rx_desc: pointer to the EOP Rx descriptor
6870  *  @skb: pointer to current skb being fixed
6871  *
6872  *  Address the case where we are pulling data in on pages only
6873  *  and as such no data is present in the skb header.
6874  *
6875  *  In addition if skb is not at least 60 bytes we need to pad it so that
6876  *  it is large enough to qualify as a valid Ethernet frame.
6877  *
6878  *  Returns true if an error was encountered and skb was freed.
6879  **/
6880 static bool igb_cleanup_headers(struct igb_ring *rx_ring,
6881 				union e1000_adv_rx_desc *rx_desc,
6882 				struct sk_buff *skb)
6883 {
6884 	if (unlikely((igb_test_staterr(rx_desc,
6885 				       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
6886 		struct net_device *netdev = rx_ring->netdev;
6887 		if (!(netdev->features & NETIF_F_RXALL)) {
6888 			dev_kfree_skb_any(skb);
6889 			return true;
6890 		}
6891 	}
6892 
6893 	/* place header in linear portion of buffer */
6894 	if (skb_is_nonlinear(skb))
6895 		igb_pull_tail(rx_ring, rx_desc, skb);
6896 
6897 	/* if eth_skb_pad returns an error the skb was freed */
6898 	if (eth_skb_pad(skb))
6899 		return true;
6900 
6901 	return false;
6902 }
6903 
6904 /**
6905  *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
6906  *  @rx_ring: rx descriptor ring packet is being transacted on
6907  *  @rx_desc: pointer to the EOP Rx descriptor
6908  *  @skb: pointer to current skb being populated
6909  *
6910  *  This function checks the ring, descriptor, and packet information in
6911  *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
6912  *  other fields within the skb.
6913  **/
6914 static void igb_process_skb_fields(struct igb_ring *rx_ring,
6915 				   union e1000_adv_rx_desc *rx_desc,
6916 				   struct sk_buff *skb)
6917 {
6918 	struct net_device *dev = rx_ring->netdev;
6919 
6920 	igb_rx_hash(rx_ring, rx_desc, skb);
6921 
6922 	igb_rx_checksum(rx_ring, rx_desc, skb);
6923 
6924 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
6925 	    !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
6926 		igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
6927 
6928 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
6929 	    igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
6930 		u16 vid;
6931 
6932 		if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
6933 		    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
6934 			vid = be16_to_cpu(rx_desc->wb.upper.vlan);
6935 		else
6936 			vid = le16_to_cpu(rx_desc->wb.upper.vlan);
6937 
6938 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
6939 	}
6940 
6941 	skb_record_rx_queue(skb, rx_ring->queue_index);
6942 
6943 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
6944 }
6945 
6946 static bool igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
6947 {
6948 	struct igb_ring *rx_ring = q_vector->rx.ring;
6949 	struct sk_buff *skb = rx_ring->skb;
6950 	unsigned int total_bytes = 0, total_packets = 0;
6951 	u16 cleaned_count = igb_desc_unused(rx_ring);
6952 
6953 	while (likely(total_packets < budget)) {
6954 		union e1000_adv_rx_desc *rx_desc;
6955 
6956 		/* return some buffers to hardware, one at a time is too slow */
6957 		if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
6958 			igb_alloc_rx_buffers(rx_ring, cleaned_count);
6959 			cleaned_count = 0;
6960 		}
6961 
6962 		rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
6963 
6964 		if (!rx_desc->wb.upper.status_error)
6965 			break;
6966 
6967 		/* This memory barrier is needed to keep us from reading
6968 		 * any other fields out of the rx_desc until we know the
6969 		 * descriptor has been written back
6970 		 */
6971 		dma_rmb();
6972 
6973 		/* retrieve a buffer from the ring */
6974 		skb = igb_fetch_rx_buffer(rx_ring, rx_desc, skb);
6975 
6976 		/* exit if we failed to retrieve a buffer */
6977 		if (!skb)
6978 			break;
6979 
6980 		cleaned_count++;
6981 
6982 		/* fetch next buffer in frame if non-eop */
6983 		if (igb_is_non_eop(rx_ring, rx_desc))
6984 			continue;
6985 
6986 		/* verify the packet layout is correct */
6987 		if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
6988 			skb = NULL;
6989 			continue;
6990 		}
6991 
6992 		/* probably a little skewed due to removing CRC */
6993 		total_bytes += skb->len;
6994 
6995 		/* populate checksum, timestamp, VLAN, and protocol */
6996 		igb_process_skb_fields(rx_ring, rx_desc, skb);
6997 
6998 		napi_gro_receive(&q_vector->napi, skb);
6999 
7000 		/* reset skb pointer */
7001 		skb = NULL;
7002 
7003 		/* update budget accounting */
7004 		total_packets++;
7005 	}
7006 
7007 	/* place incomplete frames back on ring for completion */
7008 	rx_ring->skb = skb;
7009 
7010 	u64_stats_update_begin(&rx_ring->rx_syncp);
7011 	rx_ring->rx_stats.packets += total_packets;
7012 	rx_ring->rx_stats.bytes += total_bytes;
7013 	u64_stats_update_end(&rx_ring->rx_syncp);
7014 	q_vector->rx.total_packets += total_packets;
7015 	q_vector->rx.total_bytes += total_bytes;
7016 
7017 	if (cleaned_count)
7018 		igb_alloc_rx_buffers(rx_ring, cleaned_count);
7019 
7020 	return total_packets < budget;
7021 }
7022 
7023 static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
7024 				  struct igb_rx_buffer *bi)
7025 {
7026 	struct page *page = bi->page;
7027 	dma_addr_t dma;
7028 
7029 	/* since we are recycling buffers we should seldom need to alloc */
7030 	if (likely(page))
7031 		return true;
7032 
7033 	/* alloc new page for storage */
7034 	page = dev_alloc_page();
7035 	if (unlikely(!page)) {
7036 		rx_ring->rx_stats.alloc_failed++;
7037 		return false;
7038 	}
7039 
7040 	/* map page for use */
7041 	dma = dma_map_page(rx_ring->dev, page, 0, PAGE_SIZE, DMA_FROM_DEVICE);
7042 
7043 	/* if mapping failed free memory back to system since
7044 	 * there isn't much point in holding memory we can't use
7045 	 */
7046 	if (dma_mapping_error(rx_ring->dev, dma)) {
7047 		__free_page(page);
7048 
7049 		rx_ring->rx_stats.alloc_failed++;
7050 		return false;
7051 	}
7052 
7053 	bi->dma = dma;
7054 	bi->page = page;
7055 	bi->page_offset = 0;
7056 
7057 	return true;
7058 }
7059 
7060 /**
7061  *  igb_alloc_rx_buffers - Replace used receive buffers; packet split
7062  *  @adapter: address of board private structure
7063  **/
7064 void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
7065 {
7066 	union e1000_adv_rx_desc *rx_desc;
7067 	struct igb_rx_buffer *bi;
7068 	u16 i = rx_ring->next_to_use;
7069 
7070 	/* nothing to do */
7071 	if (!cleaned_count)
7072 		return;
7073 
7074 	rx_desc = IGB_RX_DESC(rx_ring, i);
7075 	bi = &rx_ring->rx_buffer_info[i];
7076 	i -= rx_ring->count;
7077 
7078 	do {
7079 		if (!igb_alloc_mapped_page(rx_ring, bi))
7080 			break;
7081 
7082 		/* Refresh the desc even if buffer_addrs didn't change
7083 		 * because each write-back erases this info.
7084 		 */
7085 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
7086 
7087 		rx_desc++;
7088 		bi++;
7089 		i++;
7090 		if (unlikely(!i)) {
7091 			rx_desc = IGB_RX_DESC(rx_ring, 0);
7092 			bi = rx_ring->rx_buffer_info;
7093 			i -= rx_ring->count;
7094 		}
7095 
7096 		/* clear the status bits for the next_to_use descriptor */
7097 		rx_desc->wb.upper.status_error = 0;
7098 
7099 		cleaned_count--;
7100 	} while (cleaned_count);
7101 
7102 	i += rx_ring->count;
7103 
7104 	if (rx_ring->next_to_use != i) {
7105 		/* record the next descriptor to use */
7106 		rx_ring->next_to_use = i;
7107 
7108 		/* update next to alloc since we have filled the ring */
7109 		rx_ring->next_to_alloc = i;
7110 
7111 		/* Force memory writes to complete before letting h/w
7112 		 * know there are new descriptors to fetch.  (Only
7113 		 * applicable for weak-ordered memory model archs,
7114 		 * such as IA-64).
7115 		 */
7116 		wmb();
7117 		writel(i, rx_ring->tail);
7118 	}
7119 }
7120 
7121 /**
7122  * igb_mii_ioctl -
7123  * @netdev:
7124  * @ifreq:
7125  * @cmd:
7126  **/
7127 static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7128 {
7129 	struct igb_adapter *adapter = netdev_priv(netdev);
7130 	struct mii_ioctl_data *data = if_mii(ifr);
7131 
7132 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
7133 		return -EOPNOTSUPP;
7134 
7135 	switch (cmd) {
7136 	case SIOCGMIIPHY:
7137 		data->phy_id = adapter->hw.phy.addr;
7138 		break;
7139 	case SIOCGMIIREG:
7140 		if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
7141 				     &data->val_out))
7142 			return -EIO;
7143 		break;
7144 	case SIOCSMIIREG:
7145 	default:
7146 		return -EOPNOTSUPP;
7147 	}
7148 	return 0;
7149 }
7150 
7151 /**
7152  * igb_ioctl -
7153  * @netdev:
7154  * @ifreq:
7155  * @cmd:
7156  **/
7157 static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7158 {
7159 	switch (cmd) {
7160 	case SIOCGMIIPHY:
7161 	case SIOCGMIIREG:
7162 	case SIOCSMIIREG:
7163 		return igb_mii_ioctl(netdev, ifr, cmd);
7164 	case SIOCGHWTSTAMP:
7165 		return igb_ptp_get_ts_config(netdev, ifr);
7166 	case SIOCSHWTSTAMP:
7167 		return igb_ptp_set_ts_config(netdev, ifr);
7168 	default:
7169 		return -EOPNOTSUPP;
7170 	}
7171 }
7172 
7173 void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
7174 {
7175 	struct igb_adapter *adapter = hw->back;
7176 
7177 	pci_read_config_word(adapter->pdev, reg, value);
7178 }
7179 
7180 void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
7181 {
7182 	struct igb_adapter *adapter = hw->back;
7183 
7184 	pci_write_config_word(adapter->pdev, reg, *value);
7185 }
7186 
7187 s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
7188 {
7189 	struct igb_adapter *adapter = hw->back;
7190 
7191 	if (pcie_capability_read_word(adapter->pdev, reg, value))
7192 		return -E1000_ERR_CONFIG;
7193 
7194 	return 0;
7195 }
7196 
7197 s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
7198 {
7199 	struct igb_adapter *adapter = hw->back;
7200 
7201 	if (pcie_capability_write_word(adapter->pdev, reg, *value))
7202 		return -E1000_ERR_CONFIG;
7203 
7204 	return 0;
7205 }
7206 
7207 static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
7208 {
7209 	struct igb_adapter *adapter = netdev_priv(netdev);
7210 	struct e1000_hw *hw = &adapter->hw;
7211 	u32 ctrl, rctl;
7212 	bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
7213 
7214 	if (enable) {
7215 		/* enable VLAN tag insert/strip */
7216 		ctrl = rd32(E1000_CTRL);
7217 		ctrl |= E1000_CTRL_VME;
7218 		wr32(E1000_CTRL, ctrl);
7219 
7220 		/* Disable CFI check */
7221 		rctl = rd32(E1000_RCTL);
7222 		rctl &= ~E1000_RCTL_CFIEN;
7223 		wr32(E1000_RCTL, rctl);
7224 	} else {
7225 		/* disable VLAN tag insert/strip */
7226 		ctrl = rd32(E1000_CTRL);
7227 		ctrl &= ~E1000_CTRL_VME;
7228 		wr32(E1000_CTRL, ctrl);
7229 	}
7230 
7231 	igb_rlpml_set(adapter);
7232 }
7233 
7234 static int igb_vlan_rx_add_vid(struct net_device *netdev,
7235 			       __be16 proto, u16 vid)
7236 {
7237 	struct igb_adapter *adapter = netdev_priv(netdev);
7238 	struct e1000_hw *hw = &adapter->hw;
7239 	int pf_id = adapter->vfs_allocated_count;
7240 
7241 	/* attempt to add filter to vlvf array */
7242 	igb_vlvf_set(adapter, vid, true, pf_id);
7243 
7244 	/* add the filter since PF can receive vlans w/o entry in vlvf */
7245 	igb_vfta_set(hw, vid, true);
7246 
7247 	set_bit(vid, adapter->active_vlans);
7248 
7249 	return 0;
7250 }
7251 
7252 static int igb_vlan_rx_kill_vid(struct net_device *netdev,
7253 				__be16 proto, u16 vid)
7254 {
7255 	struct igb_adapter *adapter = netdev_priv(netdev);
7256 	struct e1000_hw *hw = &adapter->hw;
7257 	int pf_id = adapter->vfs_allocated_count;
7258 	s32 err;
7259 
7260 	/* remove vlan from VLVF table array */
7261 	err = igb_vlvf_set(adapter, vid, false, pf_id);
7262 
7263 	/* if vid was not present in VLVF just remove it from table */
7264 	if (err)
7265 		igb_vfta_set(hw, vid, false);
7266 
7267 	clear_bit(vid, adapter->active_vlans);
7268 
7269 	return 0;
7270 }
7271 
7272 static void igb_restore_vlan(struct igb_adapter *adapter)
7273 {
7274 	u16 vid;
7275 
7276 	igb_vlan_mode(adapter->netdev, adapter->netdev->features);
7277 
7278 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
7279 		igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
7280 }
7281 
7282 int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
7283 {
7284 	struct pci_dev *pdev = adapter->pdev;
7285 	struct e1000_mac_info *mac = &adapter->hw.mac;
7286 
7287 	mac->autoneg = 0;
7288 
7289 	/* Make sure dplx is at most 1 bit and lsb of speed is not set
7290 	 * for the switch() below to work
7291 	 */
7292 	if ((spd & 1) || (dplx & ~1))
7293 		goto err_inval;
7294 
7295 	/* Fiber NIC's only allow 1000 gbps Full duplex
7296 	 * and 100Mbps Full duplex for 100baseFx sfp
7297 	 */
7298 	if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
7299 		switch (spd + dplx) {
7300 		case SPEED_10 + DUPLEX_HALF:
7301 		case SPEED_10 + DUPLEX_FULL:
7302 		case SPEED_100 + DUPLEX_HALF:
7303 			goto err_inval;
7304 		default:
7305 			break;
7306 		}
7307 	}
7308 
7309 	switch (spd + dplx) {
7310 	case SPEED_10 + DUPLEX_HALF:
7311 		mac->forced_speed_duplex = ADVERTISE_10_HALF;
7312 		break;
7313 	case SPEED_10 + DUPLEX_FULL:
7314 		mac->forced_speed_duplex = ADVERTISE_10_FULL;
7315 		break;
7316 	case SPEED_100 + DUPLEX_HALF:
7317 		mac->forced_speed_duplex = ADVERTISE_100_HALF;
7318 		break;
7319 	case SPEED_100 + DUPLEX_FULL:
7320 		mac->forced_speed_duplex = ADVERTISE_100_FULL;
7321 		break;
7322 	case SPEED_1000 + DUPLEX_FULL:
7323 		mac->autoneg = 1;
7324 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
7325 		break;
7326 	case SPEED_1000 + DUPLEX_HALF: /* not supported */
7327 	default:
7328 		goto err_inval;
7329 	}
7330 
7331 	/* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
7332 	adapter->hw.phy.mdix = AUTO_ALL_MODES;
7333 
7334 	return 0;
7335 
7336 err_inval:
7337 	dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
7338 	return -EINVAL;
7339 }
7340 
7341 static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
7342 			  bool runtime)
7343 {
7344 	struct net_device *netdev = pci_get_drvdata(pdev);
7345 	struct igb_adapter *adapter = netdev_priv(netdev);
7346 	struct e1000_hw *hw = &adapter->hw;
7347 	u32 ctrl, rctl, status;
7348 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
7349 #ifdef CONFIG_PM
7350 	int retval = 0;
7351 #endif
7352 
7353 	netif_device_detach(netdev);
7354 
7355 	if (netif_running(netdev))
7356 		__igb_close(netdev, true);
7357 
7358 	igb_clear_interrupt_scheme(adapter);
7359 
7360 #ifdef CONFIG_PM
7361 	retval = pci_save_state(pdev);
7362 	if (retval)
7363 		return retval;
7364 #endif
7365 
7366 	status = rd32(E1000_STATUS);
7367 	if (status & E1000_STATUS_LU)
7368 		wufc &= ~E1000_WUFC_LNKC;
7369 
7370 	if (wufc) {
7371 		igb_setup_rctl(adapter);
7372 		igb_set_rx_mode(netdev);
7373 
7374 		/* turn on all-multi mode if wake on multicast is enabled */
7375 		if (wufc & E1000_WUFC_MC) {
7376 			rctl = rd32(E1000_RCTL);
7377 			rctl |= E1000_RCTL_MPE;
7378 			wr32(E1000_RCTL, rctl);
7379 		}
7380 
7381 		ctrl = rd32(E1000_CTRL);
7382 		/* advertise wake from D3Cold */
7383 		#define E1000_CTRL_ADVD3WUC 0x00100000
7384 		/* phy power management enable */
7385 		#define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
7386 		ctrl |= E1000_CTRL_ADVD3WUC;
7387 		wr32(E1000_CTRL, ctrl);
7388 
7389 		/* Allow time for pending master requests to run */
7390 		igb_disable_pcie_master(hw);
7391 
7392 		wr32(E1000_WUC, E1000_WUC_PME_EN);
7393 		wr32(E1000_WUFC, wufc);
7394 	} else {
7395 		wr32(E1000_WUC, 0);
7396 		wr32(E1000_WUFC, 0);
7397 	}
7398 
7399 	*enable_wake = wufc || adapter->en_mng_pt;
7400 	if (!*enable_wake)
7401 		igb_power_down_link(adapter);
7402 	else
7403 		igb_power_up_link(adapter);
7404 
7405 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7406 	 * would have already happened in close and is redundant.
7407 	 */
7408 	igb_release_hw_control(adapter);
7409 
7410 	pci_disable_device(pdev);
7411 
7412 	return 0;
7413 }
7414 
7415 #ifdef CONFIG_PM
7416 #ifdef CONFIG_PM_SLEEP
7417 static int igb_suspend(struct device *dev)
7418 {
7419 	int retval;
7420 	bool wake;
7421 	struct pci_dev *pdev = to_pci_dev(dev);
7422 
7423 	retval = __igb_shutdown(pdev, &wake, 0);
7424 	if (retval)
7425 		return retval;
7426 
7427 	if (wake) {
7428 		pci_prepare_to_sleep(pdev);
7429 	} else {
7430 		pci_wake_from_d3(pdev, false);
7431 		pci_set_power_state(pdev, PCI_D3hot);
7432 	}
7433 
7434 	return 0;
7435 }
7436 #endif /* CONFIG_PM_SLEEP */
7437 
7438 static int igb_resume(struct device *dev)
7439 {
7440 	struct pci_dev *pdev = to_pci_dev(dev);
7441 	struct net_device *netdev = pci_get_drvdata(pdev);
7442 	struct igb_adapter *adapter = netdev_priv(netdev);
7443 	struct e1000_hw *hw = &adapter->hw;
7444 	u32 err;
7445 
7446 	pci_set_power_state(pdev, PCI_D0);
7447 	pci_restore_state(pdev);
7448 	pci_save_state(pdev);
7449 
7450 	if (!pci_device_is_present(pdev))
7451 		return -ENODEV;
7452 	err = pci_enable_device_mem(pdev);
7453 	if (err) {
7454 		dev_err(&pdev->dev,
7455 			"igb: Cannot enable PCI device from suspend\n");
7456 		return err;
7457 	}
7458 	pci_set_master(pdev);
7459 
7460 	pci_enable_wake(pdev, PCI_D3hot, 0);
7461 	pci_enable_wake(pdev, PCI_D3cold, 0);
7462 
7463 	if (igb_init_interrupt_scheme(adapter, true)) {
7464 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
7465 		return -ENOMEM;
7466 	}
7467 
7468 	igb_reset(adapter);
7469 
7470 	/* let the f/w know that the h/w is now under the control of the
7471 	 * driver.
7472 	 */
7473 	igb_get_hw_control(adapter);
7474 
7475 	wr32(E1000_WUS, ~0);
7476 
7477 	if (netdev->flags & IFF_UP) {
7478 		rtnl_lock();
7479 		err = __igb_open(netdev, true);
7480 		rtnl_unlock();
7481 		if (err)
7482 			return err;
7483 	}
7484 
7485 	netif_device_attach(netdev);
7486 	return 0;
7487 }
7488 
7489 static int igb_runtime_idle(struct device *dev)
7490 {
7491 	struct pci_dev *pdev = to_pci_dev(dev);
7492 	struct net_device *netdev = pci_get_drvdata(pdev);
7493 	struct igb_adapter *adapter = netdev_priv(netdev);
7494 
7495 	if (!igb_has_link(adapter))
7496 		pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
7497 
7498 	return -EBUSY;
7499 }
7500 
7501 static int igb_runtime_suspend(struct device *dev)
7502 {
7503 	struct pci_dev *pdev = to_pci_dev(dev);
7504 	int retval;
7505 	bool wake;
7506 
7507 	retval = __igb_shutdown(pdev, &wake, 1);
7508 	if (retval)
7509 		return retval;
7510 
7511 	if (wake) {
7512 		pci_prepare_to_sleep(pdev);
7513 	} else {
7514 		pci_wake_from_d3(pdev, false);
7515 		pci_set_power_state(pdev, PCI_D3hot);
7516 	}
7517 
7518 	return 0;
7519 }
7520 
7521 static int igb_runtime_resume(struct device *dev)
7522 {
7523 	return igb_resume(dev);
7524 }
7525 #endif /* CONFIG_PM */
7526 
7527 static void igb_shutdown(struct pci_dev *pdev)
7528 {
7529 	bool wake;
7530 
7531 	__igb_shutdown(pdev, &wake, 0);
7532 
7533 	if (system_state == SYSTEM_POWER_OFF) {
7534 		pci_wake_from_d3(pdev, wake);
7535 		pci_set_power_state(pdev, PCI_D3hot);
7536 	}
7537 }
7538 
7539 #ifdef CONFIG_PCI_IOV
7540 static int igb_sriov_reinit(struct pci_dev *dev)
7541 {
7542 	struct net_device *netdev = pci_get_drvdata(dev);
7543 	struct igb_adapter *adapter = netdev_priv(netdev);
7544 	struct pci_dev *pdev = adapter->pdev;
7545 
7546 	rtnl_lock();
7547 
7548 	if (netif_running(netdev))
7549 		igb_close(netdev);
7550 	else
7551 		igb_reset(adapter);
7552 
7553 	igb_clear_interrupt_scheme(adapter);
7554 
7555 	igb_init_queue_configuration(adapter);
7556 
7557 	if (igb_init_interrupt_scheme(adapter, true)) {
7558 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
7559 		return -ENOMEM;
7560 	}
7561 
7562 	if (netif_running(netdev))
7563 		igb_open(netdev);
7564 
7565 	rtnl_unlock();
7566 
7567 	return 0;
7568 }
7569 
7570 static int igb_pci_disable_sriov(struct pci_dev *dev)
7571 {
7572 	int err = igb_disable_sriov(dev);
7573 
7574 	if (!err)
7575 		err = igb_sriov_reinit(dev);
7576 
7577 	return err;
7578 }
7579 
7580 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
7581 {
7582 	int err = igb_enable_sriov(dev, num_vfs);
7583 
7584 	if (err)
7585 		goto out;
7586 
7587 	err = igb_sriov_reinit(dev);
7588 	if (!err)
7589 		return num_vfs;
7590 
7591 out:
7592 	return err;
7593 }
7594 
7595 #endif
7596 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
7597 {
7598 #ifdef CONFIG_PCI_IOV
7599 	if (num_vfs == 0)
7600 		return igb_pci_disable_sriov(dev);
7601 	else
7602 		return igb_pci_enable_sriov(dev, num_vfs);
7603 #endif
7604 	return 0;
7605 }
7606 
7607 #ifdef CONFIG_NET_POLL_CONTROLLER
7608 /* Polling 'interrupt' - used by things like netconsole to send skbs
7609  * without having to re-enable interrupts. It's not called while
7610  * the interrupt routine is executing.
7611  */
7612 static void igb_netpoll(struct net_device *netdev)
7613 {
7614 	struct igb_adapter *adapter = netdev_priv(netdev);
7615 	struct e1000_hw *hw = &adapter->hw;
7616 	struct igb_q_vector *q_vector;
7617 	int i;
7618 
7619 	for (i = 0; i < adapter->num_q_vectors; i++) {
7620 		q_vector = adapter->q_vector[i];
7621 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
7622 			wr32(E1000_EIMC, q_vector->eims_value);
7623 		else
7624 			igb_irq_disable(adapter);
7625 		napi_schedule(&q_vector->napi);
7626 	}
7627 }
7628 #endif /* CONFIG_NET_POLL_CONTROLLER */
7629 
7630 /**
7631  *  igb_io_error_detected - called when PCI error is detected
7632  *  @pdev: Pointer to PCI device
7633  *  @state: The current pci connection state
7634  *
7635  *  This function is called after a PCI bus error affecting
7636  *  this device has been detected.
7637  **/
7638 static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
7639 					      pci_channel_state_t state)
7640 {
7641 	struct net_device *netdev = pci_get_drvdata(pdev);
7642 	struct igb_adapter *adapter = netdev_priv(netdev);
7643 
7644 	netif_device_detach(netdev);
7645 
7646 	if (state == pci_channel_io_perm_failure)
7647 		return PCI_ERS_RESULT_DISCONNECT;
7648 
7649 	if (netif_running(netdev))
7650 		igb_down(adapter);
7651 	pci_disable_device(pdev);
7652 
7653 	/* Request a slot slot reset. */
7654 	return PCI_ERS_RESULT_NEED_RESET;
7655 }
7656 
7657 /**
7658  *  igb_io_slot_reset - called after the pci bus has been reset.
7659  *  @pdev: Pointer to PCI device
7660  *
7661  *  Restart the card from scratch, as if from a cold-boot. Implementation
7662  *  resembles the first-half of the igb_resume routine.
7663  **/
7664 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
7665 {
7666 	struct net_device *netdev = pci_get_drvdata(pdev);
7667 	struct igb_adapter *adapter = netdev_priv(netdev);
7668 	struct e1000_hw *hw = &adapter->hw;
7669 	pci_ers_result_t result;
7670 	int err;
7671 
7672 	if (pci_enable_device_mem(pdev)) {
7673 		dev_err(&pdev->dev,
7674 			"Cannot re-enable PCI device after reset.\n");
7675 		result = PCI_ERS_RESULT_DISCONNECT;
7676 	} else {
7677 		pci_set_master(pdev);
7678 		pci_restore_state(pdev);
7679 		pci_save_state(pdev);
7680 
7681 		pci_enable_wake(pdev, PCI_D3hot, 0);
7682 		pci_enable_wake(pdev, PCI_D3cold, 0);
7683 
7684 		igb_reset(adapter);
7685 		wr32(E1000_WUS, ~0);
7686 		result = PCI_ERS_RESULT_RECOVERED;
7687 	}
7688 
7689 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7690 	if (err) {
7691 		dev_err(&pdev->dev,
7692 			"pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7693 			err);
7694 		/* non-fatal, continue */
7695 	}
7696 
7697 	return result;
7698 }
7699 
7700 /**
7701  *  igb_io_resume - called when traffic can start flowing again.
7702  *  @pdev: Pointer to PCI device
7703  *
7704  *  This callback is called when the error recovery driver tells us that
7705  *  its OK to resume normal operation. Implementation resembles the
7706  *  second-half of the igb_resume routine.
7707  */
7708 static void igb_io_resume(struct pci_dev *pdev)
7709 {
7710 	struct net_device *netdev = pci_get_drvdata(pdev);
7711 	struct igb_adapter *adapter = netdev_priv(netdev);
7712 
7713 	if (netif_running(netdev)) {
7714 		if (igb_up(adapter)) {
7715 			dev_err(&pdev->dev, "igb_up failed after reset\n");
7716 			return;
7717 		}
7718 	}
7719 
7720 	netif_device_attach(netdev);
7721 
7722 	/* let the f/w know that the h/w is now under the control of the
7723 	 * driver.
7724 	 */
7725 	igb_get_hw_control(adapter);
7726 }
7727 
7728 static void igb_rar_set_qsel(struct igb_adapter *adapter, u8 *addr, u32 index,
7729 			     u8 qsel)
7730 {
7731 	u32 rar_low, rar_high;
7732 	struct e1000_hw *hw = &adapter->hw;
7733 
7734 	/* HW expects these in little endian so we reverse the byte order
7735 	 * from network order (big endian) to little endian
7736 	 */
7737 	rar_low = ((u32) addr[0] | ((u32) addr[1] << 8) |
7738 		   ((u32) addr[2] << 16) | ((u32) addr[3] << 24));
7739 	rar_high = ((u32) addr[4] | ((u32) addr[5] << 8));
7740 
7741 	/* Indicate to hardware the Address is Valid. */
7742 	rar_high |= E1000_RAH_AV;
7743 
7744 	if (hw->mac.type == e1000_82575)
7745 		rar_high |= E1000_RAH_POOL_1 * qsel;
7746 	else
7747 		rar_high |= E1000_RAH_POOL_1 << qsel;
7748 
7749 	wr32(E1000_RAL(index), rar_low);
7750 	wrfl();
7751 	wr32(E1000_RAH(index), rar_high);
7752 	wrfl();
7753 }
7754 
7755 static int igb_set_vf_mac(struct igb_adapter *adapter,
7756 			  int vf, unsigned char *mac_addr)
7757 {
7758 	struct e1000_hw *hw = &adapter->hw;
7759 	/* VF MAC addresses start at end of receive addresses and moves
7760 	 * towards the first, as a result a collision should not be possible
7761 	 */
7762 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
7763 
7764 	memcpy(adapter->vf_data[vf].vf_mac_addresses, mac_addr, ETH_ALEN);
7765 
7766 	igb_rar_set_qsel(adapter, mac_addr, rar_entry, vf);
7767 
7768 	return 0;
7769 }
7770 
7771 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
7772 {
7773 	struct igb_adapter *adapter = netdev_priv(netdev);
7774 	if (!is_valid_ether_addr(mac) || (vf >= adapter->vfs_allocated_count))
7775 		return -EINVAL;
7776 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
7777 	dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n", mac, vf);
7778 	dev_info(&adapter->pdev->dev,
7779 		 "Reload the VF driver to make this change effective.");
7780 	if (test_bit(__IGB_DOWN, &adapter->state)) {
7781 		dev_warn(&adapter->pdev->dev,
7782 			 "The VF MAC address has been set, but the PF device is not up.\n");
7783 		dev_warn(&adapter->pdev->dev,
7784 			 "Bring the PF device up before attempting to use the VF device.\n");
7785 	}
7786 	return igb_set_vf_mac(adapter, vf, mac);
7787 }
7788 
7789 static int igb_link_mbps(int internal_link_speed)
7790 {
7791 	switch (internal_link_speed) {
7792 	case SPEED_100:
7793 		return 100;
7794 	case SPEED_1000:
7795 		return 1000;
7796 	default:
7797 		return 0;
7798 	}
7799 }
7800 
7801 static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
7802 				  int link_speed)
7803 {
7804 	int rf_dec, rf_int;
7805 	u32 bcnrc_val;
7806 
7807 	if (tx_rate != 0) {
7808 		/* Calculate the rate factor values to set */
7809 		rf_int = link_speed / tx_rate;
7810 		rf_dec = (link_speed - (rf_int * tx_rate));
7811 		rf_dec = (rf_dec * (1 << E1000_RTTBCNRC_RF_INT_SHIFT)) /
7812 			 tx_rate;
7813 
7814 		bcnrc_val = E1000_RTTBCNRC_RS_ENA;
7815 		bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
7816 			      E1000_RTTBCNRC_RF_INT_MASK);
7817 		bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
7818 	} else {
7819 		bcnrc_val = 0;
7820 	}
7821 
7822 	wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
7823 	/* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
7824 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
7825 	 */
7826 	wr32(E1000_RTTBCNRM, 0x14);
7827 	wr32(E1000_RTTBCNRC, bcnrc_val);
7828 }
7829 
7830 static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
7831 {
7832 	int actual_link_speed, i;
7833 	bool reset_rate = false;
7834 
7835 	/* VF TX rate limit was not set or not supported */
7836 	if ((adapter->vf_rate_link_speed == 0) ||
7837 	    (adapter->hw.mac.type != e1000_82576))
7838 		return;
7839 
7840 	actual_link_speed = igb_link_mbps(adapter->link_speed);
7841 	if (actual_link_speed != adapter->vf_rate_link_speed) {
7842 		reset_rate = true;
7843 		adapter->vf_rate_link_speed = 0;
7844 		dev_info(&adapter->pdev->dev,
7845 			 "Link speed has been changed. VF Transmit rate is disabled\n");
7846 	}
7847 
7848 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
7849 		if (reset_rate)
7850 			adapter->vf_data[i].tx_rate = 0;
7851 
7852 		igb_set_vf_rate_limit(&adapter->hw, i,
7853 				      adapter->vf_data[i].tx_rate,
7854 				      actual_link_speed);
7855 	}
7856 }
7857 
7858 static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
7859 			     int min_tx_rate, int max_tx_rate)
7860 {
7861 	struct igb_adapter *adapter = netdev_priv(netdev);
7862 	struct e1000_hw *hw = &adapter->hw;
7863 	int actual_link_speed;
7864 
7865 	if (hw->mac.type != e1000_82576)
7866 		return -EOPNOTSUPP;
7867 
7868 	if (min_tx_rate)
7869 		return -EINVAL;
7870 
7871 	actual_link_speed = igb_link_mbps(adapter->link_speed);
7872 	if ((vf >= adapter->vfs_allocated_count) ||
7873 	    (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
7874 	    (max_tx_rate < 0) ||
7875 	    (max_tx_rate > actual_link_speed))
7876 		return -EINVAL;
7877 
7878 	adapter->vf_rate_link_speed = actual_link_speed;
7879 	adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
7880 	igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
7881 
7882 	return 0;
7883 }
7884 
7885 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
7886 				   bool setting)
7887 {
7888 	struct igb_adapter *adapter = netdev_priv(netdev);
7889 	struct e1000_hw *hw = &adapter->hw;
7890 	u32 reg_val, reg_offset;
7891 
7892 	if (!adapter->vfs_allocated_count)
7893 		return -EOPNOTSUPP;
7894 
7895 	if (vf >= adapter->vfs_allocated_count)
7896 		return -EINVAL;
7897 
7898 	reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
7899 	reg_val = rd32(reg_offset);
7900 	if (setting)
7901 		reg_val |= ((1 << vf) |
7902 			    (1 << (vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT)));
7903 	else
7904 		reg_val &= ~((1 << vf) |
7905 			     (1 << (vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT)));
7906 	wr32(reg_offset, reg_val);
7907 
7908 	adapter->vf_data[vf].spoofchk_enabled = setting;
7909 	return 0;
7910 }
7911 
7912 static int igb_ndo_get_vf_config(struct net_device *netdev,
7913 				 int vf, struct ifla_vf_info *ivi)
7914 {
7915 	struct igb_adapter *adapter = netdev_priv(netdev);
7916 	if (vf >= adapter->vfs_allocated_count)
7917 		return -EINVAL;
7918 	ivi->vf = vf;
7919 	memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
7920 	ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
7921 	ivi->min_tx_rate = 0;
7922 	ivi->vlan = adapter->vf_data[vf].pf_vlan;
7923 	ivi->qos = adapter->vf_data[vf].pf_qos;
7924 	ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
7925 	return 0;
7926 }
7927 
7928 static void igb_vmm_control(struct igb_adapter *adapter)
7929 {
7930 	struct e1000_hw *hw = &adapter->hw;
7931 	u32 reg;
7932 
7933 	switch (hw->mac.type) {
7934 	case e1000_82575:
7935 	case e1000_i210:
7936 	case e1000_i211:
7937 	case e1000_i354:
7938 	default:
7939 		/* replication is not supported for 82575 */
7940 		return;
7941 	case e1000_82576:
7942 		/* notify HW that the MAC is adding vlan tags */
7943 		reg = rd32(E1000_DTXCTL);
7944 		reg |= E1000_DTXCTL_VLAN_ADDED;
7945 		wr32(E1000_DTXCTL, reg);
7946 		/* Fall through */
7947 	case e1000_82580:
7948 		/* enable replication vlan tag stripping */
7949 		reg = rd32(E1000_RPLOLR);
7950 		reg |= E1000_RPLOLR_STRVLAN;
7951 		wr32(E1000_RPLOLR, reg);
7952 		/* Fall through */
7953 	case e1000_i350:
7954 		/* none of the above registers are supported by i350 */
7955 		break;
7956 	}
7957 
7958 	if (adapter->vfs_allocated_count) {
7959 		igb_vmdq_set_loopback_pf(hw, true);
7960 		igb_vmdq_set_replication_pf(hw, true);
7961 		igb_vmdq_set_anti_spoofing_pf(hw, true,
7962 					      adapter->vfs_allocated_count);
7963 	} else {
7964 		igb_vmdq_set_loopback_pf(hw, false);
7965 		igb_vmdq_set_replication_pf(hw, false);
7966 	}
7967 }
7968 
7969 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
7970 {
7971 	struct e1000_hw *hw = &adapter->hw;
7972 	u32 dmac_thr;
7973 	u16 hwm;
7974 
7975 	if (hw->mac.type > e1000_82580) {
7976 		if (adapter->flags & IGB_FLAG_DMAC) {
7977 			u32 reg;
7978 
7979 			/* force threshold to 0. */
7980 			wr32(E1000_DMCTXTH, 0);
7981 
7982 			/* DMA Coalescing high water mark needs to be greater
7983 			 * than the Rx threshold. Set hwm to PBA - max frame
7984 			 * size in 16B units, capping it at PBA - 6KB.
7985 			 */
7986 			hwm = 64 * pba - adapter->max_frame_size / 16;
7987 			if (hwm < 64 * (pba - 6))
7988 				hwm = 64 * (pba - 6);
7989 			reg = rd32(E1000_FCRTC);
7990 			reg &= ~E1000_FCRTC_RTH_COAL_MASK;
7991 			reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
7992 				& E1000_FCRTC_RTH_COAL_MASK);
7993 			wr32(E1000_FCRTC, reg);
7994 
7995 			/* Set the DMA Coalescing Rx threshold to PBA - 2 * max
7996 			 * frame size, capping it at PBA - 10KB.
7997 			 */
7998 			dmac_thr = pba - adapter->max_frame_size / 512;
7999 			if (dmac_thr < pba - 10)
8000 				dmac_thr = pba - 10;
8001 			reg = rd32(E1000_DMACR);
8002 			reg &= ~E1000_DMACR_DMACTHR_MASK;
8003 			reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
8004 				& E1000_DMACR_DMACTHR_MASK);
8005 
8006 			/* transition to L0x or L1 if available..*/
8007 			reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
8008 
8009 			/* watchdog timer= +-1000 usec in 32usec intervals */
8010 			reg |= (1000 >> 5);
8011 
8012 			/* Disable BMC-to-OS Watchdog Enable */
8013 			if (hw->mac.type != e1000_i354)
8014 				reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
8015 
8016 			wr32(E1000_DMACR, reg);
8017 
8018 			/* no lower threshold to disable
8019 			 * coalescing(smart fifb)-UTRESH=0
8020 			 */
8021 			wr32(E1000_DMCRTRH, 0);
8022 
8023 			reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
8024 
8025 			wr32(E1000_DMCTLX, reg);
8026 
8027 			/* free space in tx packet buffer to wake from
8028 			 * DMA coal
8029 			 */
8030 			wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
8031 			     (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
8032 
8033 			/* make low power state decision controlled
8034 			 * by DMA coal
8035 			 */
8036 			reg = rd32(E1000_PCIEMISC);
8037 			reg &= ~E1000_PCIEMISC_LX_DECISION;
8038 			wr32(E1000_PCIEMISC, reg);
8039 		} /* endif adapter->dmac is not disabled */
8040 	} else if (hw->mac.type == e1000_82580) {
8041 		u32 reg = rd32(E1000_PCIEMISC);
8042 
8043 		wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
8044 		wr32(E1000_DMACR, 0);
8045 	}
8046 }
8047 
8048 /**
8049  *  igb_read_i2c_byte - Reads 8 bit word over I2C
8050  *  @hw: pointer to hardware structure
8051  *  @byte_offset: byte offset to read
8052  *  @dev_addr: device address
8053  *  @data: value read
8054  *
8055  *  Performs byte read operation over I2C interface at
8056  *  a specified device address.
8057  **/
8058 s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
8059 		      u8 dev_addr, u8 *data)
8060 {
8061 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
8062 	struct i2c_client *this_client = adapter->i2c_client;
8063 	s32 status;
8064 	u16 swfw_mask = 0;
8065 
8066 	if (!this_client)
8067 		return E1000_ERR_I2C;
8068 
8069 	swfw_mask = E1000_SWFW_PHY0_SM;
8070 
8071 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
8072 		return E1000_ERR_SWFW_SYNC;
8073 
8074 	status = i2c_smbus_read_byte_data(this_client, byte_offset);
8075 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
8076 
8077 	if (status < 0)
8078 		return E1000_ERR_I2C;
8079 	else {
8080 		*data = status;
8081 		return 0;
8082 	}
8083 }
8084 
8085 /**
8086  *  igb_write_i2c_byte - Writes 8 bit word over I2C
8087  *  @hw: pointer to hardware structure
8088  *  @byte_offset: byte offset to write
8089  *  @dev_addr: device address
8090  *  @data: value to write
8091  *
8092  *  Performs byte write operation over I2C interface at
8093  *  a specified device address.
8094  **/
8095 s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
8096 		       u8 dev_addr, u8 data)
8097 {
8098 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
8099 	struct i2c_client *this_client = adapter->i2c_client;
8100 	s32 status;
8101 	u16 swfw_mask = E1000_SWFW_PHY0_SM;
8102 
8103 	if (!this_client)
8104 		return E1000_ERR_I2C;
8105 
8106 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
8107 		return E1000_ERR_SWFW_SYNC;
8108 	status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
8109 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
8110 
8111 	if (status)
8112 		return E1000_ERR_I2C;
8113 	else
8114 		return 0;
8115 
8116 }
8117 
8118 int igb_reinit_queues(struct igb_adapter *adapter)
8119 {
8120 	struct net_device *netdev = adapter->netdev;
8121 	struct pci_dev *pdev = adapter->pdev;
8122 	int err = 0;
8123 
8124 	if (netif_running(netdev))
8125 		igb_close(netdev);
8126 
8127 	igb_reset_interrupt_capability(adapter);
8128 
8129 	if (igb_init_interrupt_scheme(adapter, true)) {
8130 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8131 		return -ENOMEM;
8132 	}
8133 
8134 	if (netif_running(netdev))
8135 		err = igb_open(netdev);
8136 
8137 	return err;
8138 }
8139 /* igb_main.c */
8140