xref: /linux/drivers/net/ethernet/intel/igb/igb_main.c (revision 93df8a1ed6231727c5db94a80b1a6bd5ee67cec3)
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 18
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->rx_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 	else
1211 		memset(q_vector, 0, size);
1212 	if (!q_vector)
1213 		return -ENOMEM;
1214 
1215 	/* initialize NAPI */
1216 	netif_napi_add(adapter->netdev, &q_vector->napi,
1217 		       igb_poll, 64);
1218 
1219 	/* tie q_vector and adapter together */
1220 	adapter->q_vector[v_idx] = q_vector;
1221 	q_vector->adapter = adapter;
1222 
1223 	/* initialize work limits */
1224 	q_vector->tx.work_limit = adapter->tx_work_limit;
1225 
1226 	/* initialize ITR configuration */
1227 	q_vector->itr_register = adapter->hw.hw_addr + E1000_EITR(0);
1228 	q_vector->itr_val = IGB_START_ITR;
1229 
1230 	/* initialize pointer to rings */
1231 	ring = q_vector->ring;
1232 
1233 	/* intialize ITR */
1234 	if (rxr_count) {
1235 		/* rx or rx/tx vector */
1236 		if (!adapter->rx_itr_setting || adapter->rx_itr_setting > 3)
1237 			q_vector->itr_val = adapter->rx_itr_setting;
1238 	} else {
1239 		/* tx only vector */
1240 		if (!adapter->tx_itr_setting || adapter->tx_itr_setting > 3)
1241 			q_vector->itr_val = adapter->tx_itr_setting;
1242 	}
1243 
1244 	if (txr_count) {
1245 		/* assign generic ring traits */
1246 		ring->dev = &adapter->pdev->dev;
1247 		ring->netdev = adapter->netdev;
1248 
1249 		/* configure backlink on ring */
1250 		ring->q_vector = q_vector;
1251 
1252 		/* update q_vector Tx values */
1253 		igb_add_ring(ring, &q_vector->tx);
1254 
1255 		/* For 82575, context index must be unique per ring. */
1256 		if (adapter->hw.mac.type == e1000_82575)
1257 			set_bit(IGB_RING_FLAG_TX_CTX_IDX, &ring->flags);
1258 
1259 		/* apply Tx specific ring traits */
1260 		ring->count = adapter->tx_ring_count;
1261 		ring->queue_index = txr_idx;
1262 
1263 		u64_stats_init(&ring->tx_syncp);
1264 		u64_stats_init(&ring->tx_syncp2);
1265 
1266 		/* assign ring to adapter */
1267 		adapter->tx_ring[txr_idx] = ring;
1268 
1269 		/* push pointer to next ring */
1270 		ring++;
1271 	}
1272 
1273 	if (rxr_count) {
1274 		/* assign generic ring traits */
1275 		ring->dev = &adapter->pdev->dev;
1276 		ring->netdev = adapter->netdev;
1277 
1278 		/* configure backlink on ring */
1279 		ring->q_vector = q_vector;
1280 
1281 		/* update q_vector Rx values */
1282 		igb_add_ring(ring, &q_vector->rx);
1283 
1284 		/* set flag indicating ring supports SCTP checksum offload */
1285 		if (adapter->hw.mac.type >= e1000_82576)
1286 			set_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags);
1287 
1288 		/* On i350, i354, i210, and i211, loopback VLAN packets
1289 		 * have the tag byte-swapped.
1290 		 */
1291 		if (adapter->hw.mac.type >= e1000_i350)
1292 			set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &ring->flags);
1293 
1294 		/* apply Rx specific ring traits */
1295 		ring->count = adapter->rx_ring_count;
1296 		ring->queue_index = rxr_idx;
1297 
1298 		u64_stats_init(&ring->rx_syncp);
1299 
1300 		/* assign ring to adapter */
1301 		adapter->rx_ring[rxr_idx] = ring;
1302 	}
1303 
1304 	return 0;
1305 }
1306 
1307 
1308 /**
1309  *  igb_alloc_q_vectors - Allocate memory for interrupt vectors
1310  *  @adapter: board private structure to initialize
1311  *
1312  *  We allocate one q_vector per queue interrupt.  If allocation fails we
1313  *  return -ENOMEM.
1314  **/
1315 static int igb_alloc_q_vectors(struct igb_adapter *adapter)
1316 {
1317 	int q_vectors = adapter->num_q_vectors;
1318 	int rxr_remaining = adapter->num_rx_queues;
1319 	int txr_remaining = adapter->num_tx_queues;
1320 	int rxr_idx = 0, txr_idx = 0, v_idx = 0;
1321 	int err;
1322 
1323 	if (q_vectors >= (rxr_remaining + txr_remaining)) {
1324 		for (; rxr_remaining; v_idx++) {
1325 			err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1326 						 0, 0, 1, rxr_idx);
1327 
1328 			if (err)
1329 				goto err_out;
1330 
1331 			/* update counts and index */
1332 			rxr_remaining--;
1333 			rxr_idx++;
1334 		}
1335 	}
1336 
1337 	for (; v_idx < q_vectors; v_idx++) {
1338 		int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
1339 		int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);
1340 
1341 		err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1342 					 tqpv, txr_idx, rqpv, rxr_idx);
1343 
1344 		if (err)
1345 			goto err_out;
1346 
1347 		/* update counts and index */
1348 		rxr_remaining -= rqpv;
1349 		txr_remaining -= tqpv;
1350 		rxr_idx++;
1351 		txr_idx++;
1352 	}
1353 
1354 	return 0;
1355 
1356 err_out:
1357 	adapter->num_tx_queues = 0;
1358 	adapter->num_rx_queues = 0;
1359 	adapter->num_q_vectors = 0;
1360 
1361 	while (v_idx--)
1362 		igb_free_q_vector(adapter, v_idx);
1363 
1364 	return -ENOMEM;
1365 }
1366 
1367 /**
1368  *  igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1369  *  @adapter: board private structure to initialize
1370  *  @msix: boolean value of MSIX capability
1371  *
1372  *  This function initializes the interrupts and allocates all of the queues.
1373  **/
1374 static int igb_init_interrupt_scheme(struct igb_adapter *adapter, bool msix)
1375 {
1376 	struct pci_dev *pdev = adapter->pdev;
1377 	int err;
1378 
1379 	igb_set_interrupt_capability(adapter, msix);
1380 
1381 	err = igb_alloc_q_vectors(adapter);
1382 	if (err) {
1383 		dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
1384 		goto err_alloc_q_vectors;
1385 	}
1386 
1387 	igb_cache_ring_register(adapter);
1388 
1389 	return 0;
1390 
1391 err_alloc_q_vectors:
1392 	igb_reset_interrupt_capability(adapter);
1393 	return err;
1394 }
1395 
1396 /**
1397  *  igb_request_irq - initialize interrupts
1398  *  @adapter: board private structure to initialize
1399  *
1400  *  Attempts to configure interrupts using the best available
1401  *  capabilities of the hardware and kernel.
1402  **/
1403 static int igb_request_irq(struct igb_adapter *adapter)
1404 {
1405 	struct net_device *netdev = adapter->netdev;
1406 	struct pci_dev *pdev = adapter->pdev;
1407 	int err = 0;
1408 
1409 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1410 		err = igb_request_msix(adapter);
1411 		if (!err)
1412 			goto request_done;
1413 		/* fall back to MSI */
1414 		igb_free_all_tx_resources(adapter);
1415 		igb_free_all_rx_resources(adapter);
1416 
1417 		igb_clear_interrupt_scheme(adapter);
1418 		err = igb_init_interrupt_scheme(adapter, false);
1419 		if (err)
1420 			goto request_done;
1421 
1422 		igb_setup_all_tx_resources(adapter);
1423 		igb_setup_all_rx_resources(adapter);
1424 		igb_configure(adapter);
1425 	}
1426 
1427 	igb_assign_vector(adapter->q_vector[0], 0);
1428 
1429 	if (adapter->flags & IGB_FLAG_HAS_MSI) {
1430 		err = request_irq(pdev->irq, igb_intr_msi, 0,
1431 				  netdev->name, adapter);
1432 		if (!err)
1433 			goto request_done;
1434 
1435 		/* fall back to legacy interrupts */
1436 		igb_reset_interrupt_capability(adapter);
1437 		adapter->flags &= ~IGB_FLAG_HAS_MSI;
1438 	}
1439 
1440 	err = request_irq(pdev->irq, igb_intr, IRQF_SHARED,
1441 			  netdev->name, adapter);
1442 
1443 	if (err)
1444 		dev_err(&pdev->dev, "Error %d getting interrupt\n",
1445 			err);
1446 
1447 request_done:
1448 	return err;
1449 }
1450 
1451 static void igb_free_irq(struct igb_adapter *adapter)
1452 {
1453 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1454 		int vector = 0, i;
1455 
1456 		free_irq(adapter->msix_entries[vector++].vector, adapter);
1457 
1458 		for (i = 0; i < adapter->num_q_vectors; i++)
1459 			free_irq(adapter->msix_entries[vector++].vector,
1460 				 adapter->q_vector[i]);
1461 	} else {
1462 		free_irq(adapter->pdev->irq, adapter);
1463 	}
1464 }
1465 
1466 /**
1467  *  igb_irq_disable - Mask off interrupt generation on the NIC
1468  *  @adapter: board private structure
1469  **/
1470 static void igb_irq_disable(struct igb_adapter *adapter)
1471 {
1472 	struct e1000_hw *hw = &adapter->hw;
1473 
1474 	/* we need to be careful when disabling interrupts.  The VFs are also
1475 	 * mapped into these registers and so clearing the bits can cause
1476 	 * issues on the VF drivers so we only need to clear what we set
1477 	 */
1478 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1479 		u32 regval = rd32(E1000_EIAM);
1480 
1481 		wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
1482 		wr32(E1000_EIMC, adapter->eims_enable_mask);
1483 		regval = rd32(E1000_EIAC);
1484 		wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
1485 	}
1486 
1487 	wr32(E1000_IAM, 0);
1488 	wr32(E1000_IMC, ~0);
1489 	wrfl();
1490 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1491 		int i;
1492 
1493 		for (i = 0; i < adapter->num_q_vectors; i++)
1494 			synchronize_irq(adapter->msix_entries[i].vector);
1495 	} else {
1496 		synchronize_irq(adapter->pdev->irq);
1497 	}
1498 }
1499 
1500 /**
1501  *  igb_irq_enable - Enable default interrupt generation settings
1502  *  @adapter: board private structure
1503  **/
1504 static void igb_irq_enable(struct igb_adapter *adapter)
1505 {
1506 	struct e1000_hw *hw = &adapter->hw;
1507 
1508 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1509 		u32 ims = E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_DRSTA;
1510 		u32 regval = rd32(E1000_EIAC);
1511 
1512 		wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
1513 		regval = rd32(E1000_EIAM);
1514 		wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
1515 		wr32(E1000_EIMS, adapter->eims_enable_mask);
1516 		if (adapter->vfs_allocated_count) {
1517 			wr32(E1000_MBVFIMR, 0xFF);
1518 			ims |= E1000_IMS_VMMB;
1519 		}
1520 		wr32(E1000_IMS, ims);
1521 	} else {
1522 		wr32(E1000_IMS, IMS_ENABLE_MASK |
1523 				E1000_IMS_DRSTA);
1524 		wr32(E1000_IAM, IMS_ENABLE_MASK |
1525 				E1000_IMS_DRSTA);
1526 	}
1527 }
1528 
1529 static void igb_update_mng_vlan(struct igb_adapter *adapter)
1530 {
1531 	struct e1000_hw *hw = &adapter->hw;
1532 	u16 vid = adapter->hw.mng_cookie.vlan_id;
1533 	u16 old_vid = adapter->mng_vlan_id;
1534 
1535 	if (hw->mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1536 		/* add VID to filter table */
1537 		igb_vfta_set(hw, vid, true);
1538 		adapter->mng_vlan_id = vid;
1539 	} else {
1540 		adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1541 	}
1542 
1543 	if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1544 	    (vid != old_vid) &&
1545 	    !test_bit(old_vid, adapter->active_vlans)) {
1546 		/* remove VID from filter table */
1547 		igb_vfta_set(hw, old_vid, false);
1548 	}
1549 }
1550 
1551 /**
1552  *  igb_release_hw_control - release control of the h/w to f/w
1553  *  @adapter: address of board private structure
1554  *
1555  *  igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1556  *  For ASF and Pass Through versions of f/w this means that the
1557  *  driver is no longer loaded.
1558  **/
1559 static void igb_release_hw_control(struct igb_adapter *adapter)
1560 {
1561 	struct e1000_hw *hw = &adapter->hw;
1562 	u32 ctrl_ext;
1563 
1564 	/* Let firmware take over control of h/w */
1565 	ctrl_ext = rd32(E1000_CTRL_EXT);
1566 	wr32(E1000_CTRL_EXT,
1567 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1568 }
1569 
1570 /**
1571  *  igb_get_hw_control - get control of the h/w from f/w
1572  *  @adapter: address of board private structure
1573  *
1574  *  igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1575  *  For ASF and Pass Through versions of f/w this means that
1576  *  the driver is loaded.
1577  **/
1578 static void igb_get_hw_control(struct igb_adapter *adapter)
1579 {
1580 	struct e1000_hw *hw = &adapter->hw;
1581 	u32 ctrl_ext;
1582 
1583 	/* Let firmware know the driver has taken over */
1584 	ctrl_ext = rd32(E1000_CTRL_EXT);
1585 	wr32(E1000_CTRL_EXT,
1586 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1587 }
1588 
1589 /**
1590  *  igb_configure - configure the hardware for RX and TX
1591  *  @adapter: private board structure
1592  **/
1593 static void igb_configure(struct igb_adapter *adapter)
1594 {
1595 	struct net_device *netdev = adapter->netdev;
1596 	int i;
1597 
1598 	igb_get_hw_control(adapter);
1599 	igb_set_rx_mode(netdev);
1600 
1601 	igb_restore_vlan(adapter);
1602 
1603 	igb_setup_tctl(adapter);
1604 	igb_setup_mrqc(adapter);
1605 	igb_setup_rctl(adapter);
1606 
1607 	igb_configure_tx(adapter);
1608 	igb_configure_rx(adapter);
1609 
1610 	igb_rx_fifo_flush_82575(&adapter->hw);
1611 
1612 	/* call igb_desc_unused which always leaves
1613 	 * at least 1 descriptor unused to make sure
1614 	 * next_to_use != next_to_clean
1615 	 */
1616 	for (i = 0; i < adapter->num_rx_queues; i++) {
1617 		struct igb_ring *ring = adapter->rx_ring[i];
1618 		igb_alloc_rx_buffers(ring, igb_desc_unused(ring));
1619 	}
1620 }
1621 
1622 /**
1623  *  igb_power_up_link - Power up the phy/serdes link
1624  *  @adapter: address of board private structure
1625  **/
1626 void igb_power_up_link(struct igb_adapter *adapter)
1627 {
1628 	igb_reset_phy(&adapter->hw);
1629 
1630 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
1631 		igb_power_up_phy_copper(&adapter->hw);
1632 	else
1633 		igb_power_up_serdes_link_82575(&adapter->hw);
1634 
1635 	igb_setup_link(&adapter->hw);
1636 }
1637 
1638 /**
1639  *  igb_power_down_link - Power down the phy/serdes link
1640  *  @adapter: address of board private structure
1641  */
1642 static void igb_power_down_link(struct igb_adapter *adapter)
1643 {
1644 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
1645 		igb_power_down_phy_copper_82575(&adapter->hw);
1646 	else
1647 		igb_shutdown_serdes_link_82575(&adapter->hw);
1648 }
1649 
1650 /**
1651  * Detect and switch function for Media Auto Sense
1652  * @adapter: address of the board private structure
1653  **/
1654 static void igb_check_swap_media(struct igb_adapter *adapter)
1655 {
1656 	struct e1000_hw *hw = &adapter->hw;
1657 	u32 ctrl_ext, connsw;
1658 	bool swap_now = false;
1659 
1660 	ctrl_ext = rd32(E1000_CTRL_EXT);
1661 	connsw = rd32(E1000_CONNSW);
1662 
1663 	/* need to live swap if current media is copper and we have fiber/serdes
1664 	 * to go to.
1665 	 */
1666 
1667 	if ((hw->phy.media_type == e1000_media_type_copper) &&
1668 	    (!(connsw & E1000_CONNSW_AUTOSENSE_EN))) {
1669 		swap_now = true;
1670 	} else if (!(connsw & E1000_CONNSW_SERDESD)) {
1671 		/* copper signal takes time to appear */
1672 		if (adapter->copper_tries < 4) {
1673 			adapter->copper_tries++;
1674 			connsw |= E1000_CONNSW_AUTOSENSE_CONF;
1675 			wr32(E1000_CONNSW, connsw);
1676 			return;
1677 		} else {
1678 			adapter->copper_tries = 0;
1679 			if ((connsw & E1000_CONNSW_PHYSD) &&
1680 			    (!(connsw & E1000_CONNSW_PHY_PDN))) {
1681 				swap_now = true;
1682 				connsw &= ~E1000_CONNSW_AUTOSENSE_CONF;
1683 				wr32(E1000_CONNSW, connsw);
1684 			}
1685 		}
1686 	}
1687 
1688 	if (!swap_now)
1689 		return;
1690 
1691 	switch (hw->phy.media_type) {
1692 	case e1000_media_type_copper:
1693 		netdev_info(adapter->netdev,
1694 			"MAS: changing media to fiber/serdes\n");
1695 		ctrl_ext |=
1696 			E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
1697 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
1698 		adapter->copper_tries = 0;
1699 		break;
1700 	case e1000_media_type_internal_serdes:
1701 	case e1000_media_type_fiber:
1702 		netdev_info(adapter->netdev,
1703 			"MAS: changing media to copper\n");
1704 		ctrl_ext &=
1705 			~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
1706 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
1707 		break;
1708 	default:
1709 		/* shouldn't get here during regular operation */
1710 		netdev_err(adapter->netdev,
1711 			"AMS: Invalid media type found, returning\n");
1712 		break;
1713 	}
1714 	wr32(E1000_CTRL_EXT, ctrl_ext);
1715 }
1716 
1717 /**
1718  *  igb_up - Open the interface and prepare it to handle traffic
1719  *  @adapter: board private structure
1720  **/
1721 int igb_up(struct igb_adapter *adapter)
1722 {
1723 	struct e1000_hw *hw = &adapter->hw;
1724 	int i;
1725 
1726 	/* hardware has been reset, we need to reload some things */
1727 	igb_configure(adapter);
1728 
1729 	clear_bit(__IGB_DOWN, &adapter->state);
1730 
1731 	for (i = 0; i < adapter->num_q_vectors; i++)
1732 		napi_enable(&(adapter->q_vector[i]->napi));
1733 
1734 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
1735 		igb_configure_msix(adapter);
1736 	else
1737 		igb_assign_vector(adapter->q_vector[0], 0);
1738 
1739 	/* Clear any pending interrupts. */
1740 	rd32(E1000_ICR);
1741 	igb_irq_enable(adapter);
1742 
1743 	/* notify VFs that reset has been completed */
1744 	if (adapter->vfs_allocated_count) {
1745 		u32 reg_data = rd32(E1000_CTRL_EXT);
1746 
1747 		reg_data |= E1000_CTRL_EXT_PFRSTD;
1748 		wr32(E1000_CTRL_EXT, reg_data);
1749 	}
1750 
1751 	netif_tx_start_all_queues(adapter->netdev);
1752 
1753 	/* start the watchdog. */
1754 	hw->mac.get_link_status = 1;
1755 	schedule_work(&adapter->watchdog_task);
1756 
1757 	if ((adapter->flags & IGB_FLAG_EEE) &&
1758 	    (!hw->dev_spec._82575.eee_disable))
1759 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
1760 
1761 	return 0;
1762 }
1763 
1764 void igb_down(struct igb_adapter *adapter)
1765 {
1766 	struct net_device *netdev = adapter->netdev;
1767 	struct e1000_hw *hw = &adapter->hw;
1768 	u32 tctl, rctl;
1769 	int i;
1770 
1771 	/* signal that we're down so the interrupt handler does not
1772 	 * reschedule our watchdog timer
1773 	 */
1774 	set_bit(__IGB_DOWN, &adapter->state);
1775 
1776 	/* disable receives in the hardware */
1777 	rctl = rd32(E1000_RCTL);
1778 	wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
1779 	/* flush and sleep below */
1780 
1781 	netif_carrier_off(netdev);
1782 	netif_tx_stop_all_queues(netdev);
1783 
1784 	/* disable transmits in the hardware */
1785 	tctl = rd32(E1000_TCTL);
1786 	tctl &= ~E1000_TCTL_EN;
1787 	wr32(E1000_TCTL, tctl);
1788 	/* flush both disables and wait for them to finish */
1789 	wrfl();
1790 	usleep_range(10000, 11000);
1791 
1792 	igb_irq_disable(adapter);
1793 
1794 	adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
1795 
1796 	for (i = 0; i < adapter->num_q_vectors; i++) {
1797 		if (adapter->q_vector[i]) {
1798 			napi_synchronize(&adapter->q_vector[i]->napi);
1799 			napi_disable(&adapter->q_vector[i]->napi);
1800 		}
1801 	}
1802 
1803 	del_timer_sync(&adapter->watchdog_timer);
1804 	del_timer_sync(&adapter->phy_info_timer);
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 void igb_enable_mas(struct igb_adapter *adapter)
1840 {
1841 	struct e1000_hw *hw = &adapter->hw;
1842 	u32 connsw = rd32(E1000_CONNSW);
1843 
1844 	/* configure for SerDes media detect */
1845 	if ((hw->phy.media_type == e1000_media_type_copper) &&
1846 	    (!(connsw & E1000_CONNSW_SERDESD))) {
1847 		connsw |= E1000_CONNSW_ENRGSRC;
1848 		connsw |= E1000_CONNSW_AUTOSENSE_EN;
1849 		wr32(E1000_CONNSW, connsw);
1850 		wrfl();
1851 	}
1852 }
1853 
1854 void igb_reset(struct igb_adapter *adapter)
1855 {
1856 	struct pci_dev *pdev = adapter->pdev;
1857 	struct e1000_hw *hw = &adapter->hw;
1858 	struct e1000_mac_info *mac = &hw->mac;
1859 	struct e1000_fc_info *fc = &hw->fc;
1860 	u32 pba = 0, tx_space, min_tx_space, min_rx_space, hwm;
1861 
1862 	/* Repartition Pba for greater than 9k mtu
1863 	 * To take effect CTRL.RST is required.
1864 	 */
1865 	switch (mac->type) {
1866 	case e1000_i350:
1867 	case e1000_i354:
1868 	case e1000_82580:
1869 		pba = rd32(E1000_RXPBS);
1870 		pba = igb_rxpbs_adjust_82580(pba);
1871 		break;
1872 	case e1000_82576:
1873 		pba = rd32(E1000_RXPBS);
1874 		pba &= E1000_RXPBS_SIZE_MASK_82576;
1875 		break;
1876 	case e1000_82575:
1877 	case e1000_i210:
1878 	case e1000_i211:
1879 	default:
1880 		pba = E1000_PBA_34K;
1881 		break;
1882 	}
1883 
1884 	if ((adapter->max_frame_size > ETH_FRAME_LEN + ETH_FCS_LEN) &&
1885 	    (mac->type < e1000_82576)) {
1886 		/* adjust PBA for jumbo frames */
1887 		wr32(E1000_PBA, pba);
1888 
1889 		/* To maintain wire speed transmits, the Tx FIFO should be
1890 		 * large enough to accommodate two full transmit packets,
1891 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
1892 		 * the Rx FIFO should be large enough to accommodate at least
1893 		 * one full receive packet and is similarly rounded up and
1894 		 * expressed in KB.
1895 		 */
1896 		pba = rd32(E1000_PBA);
1897 		/* upper 16 bits has Tx packet buffer allocation size in KB */
1898 		tx_space = pba >> 16;
1899 		/* lower 16 bits has Rx packet buffer allocation size in KB */
1900 		pba &= 0xffff;
1901 		/* the Tx fifo also stores 16 bytes of information about the Tx
1902 		 * but don't include ethernet FCS because hardware appends it
1903 		 */
1904 		min_tx_space = (adapter->max_frame_size +
1905 				sizeof(union e1000_adv_tx_desc) -
1906 				ETH_FCS_LEN) * 2;
1907 		min_tx_space = ALIGN(min_tx_space, 1024);
1908 		min_tx_space >>= 10;
1909 		/* software strips receive CRC, so leave room for it */
1910 		min_rx_space = adapter->max_frame_size;
1911 		min_rx_space = ALIGN(min_rx_space, 1024);
1912 		min_rx_space >>= 10;
1913 
1914 		/* If current Tx allocation is less than the min Tx FIFO size,
1915 		 * and the min Tx FIFO size is less than the current Rx FIFO
1916 		 * allocation, take space away from current Rx allocation
1917 		 */
1918 		if (tx_space < min_tx_space &&
1919 		    ((min_tx_space - tx_space) < pba)) {
1920 			pba = pba - (min_tx_space - tx_space);
1921 
1922 			/* if short on Rx space, Rx wins and must trump Tx
1923 			 * adjustment
1924 			 */
1925 			if (pba < min_rx_space)
1926 				pba = min_rx_space;
1927 		}
1928 		wr32(E1000_PBA, pba);
1929 	}
1930 
1931 	/* flow control settings */
1932 	/* The high water mark must be low enough to fit one full frame
1933 	 * (or the size used for early receive) above it in the Rx FIFO.
1934 	 * Set it to the lower of:
1935 	 * - 90% of the Rx FIFO size, or
1936 	 * - the full Rx FIFO size minus one full frame
1937 	 */
1938 	hwm = min(((pba << 10) * 9 / 10),
1939 			((pba << 10) - 2 * adapter->max_frame_size));
1940 
1941 	fc->high_water = hwm & 0xFFFFFFF0;	/* 16-byte granularity */
1942 	fc->low_water = fc->high_water - 16;
1943 	fc->pause_time = 0xFFFF;
1944 	fc->send_xon = 1;
1945 	fc->current_mode = fc->requested_mode;
1946 
1947 	/* disable receive for all VFs and wait one second */
1948 	if (adapter->vfs_allocated_count) {
1949 		int i;
1950 
1951 		for (i = 0 ; i < adapter->vfs_allocated_count; i++)
1952 			adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
1953 
1954 		/* ping all the active vfs to let them know we are going down */
1955 		igb_ping_all_vfs(adapter);
1956 
1957 		/* disable transmits and receives */
1958 		wr32(E1000_VFRE, 0);
1959 		wr32(E1000_VFTE, 0);
1960 	}
1961 
1962 	/* Allow time for pending master requests to run */
1963 	hw->mac.ops.reset_hw(hw);
1964 	wr32(E1000_WUC, 0);
1965 
1966 	if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
1967 		/* need to resetup here after media swap */
1968 		adapter->ei.get_invariants(hw);
1969 		adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
1970 	}
1971 	if ((mac->type == e1000_82575) &&
1972 	    (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
1973 		igb_enable_mas(adapter);
1974 	}
1975 	if (hw->mac.ops.init_hw(hw))
1976 		dev_err(&pdev->dev, "Hardware Error\n");
1977 
1978 	/* Flow control settings reset on hardware reset, so guarantee flow
1979 	 * control is off when forcing speed.
1980 	 */
1981 	if (!hw->mac.autoneg)
1982 		igb_force_mac_fc(hw);
1983 
1984 	igb_init_dmac(adapter, pba);
1985 #ifdef CONFIG_IGB_HWMON
1986 	/* Re-initialize the thermal sensor on i350 devices. */
1987 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
1988 		if (mac->type == e1000_i350 && hw->bus.func == 0) {
1989 			/* If present, re-initialize the external thermal sensor
1990 			 * interface.
1991 			 */
1992 			if (adapter->ets)
1993 				mac->ops.init_thermal_sensor_thresh(hw);
1994 		}
1995 	}
1996 #endif
1997 	/* Re-establish EEE setting */
1998 	if (hw->phy.media_type == e1000_media_type_copper) {
1999 		switch (mac->type) {
2000 		case e1000_i350:
2001 		case e1000_i210:
2002 		case e1000_i211:
2003 			igb_set_eee_i350(hw, true, true);
2004 			break;
2005 		case e1000_i354:
2006 			igb_set_eee_i354(hw, true, true);
2007 			break;
2008 		default:
2009 			break;
2010 		}
2011 	}
2012 	if (!netif_running(adapter->netdev))
2013 		igb_power_down_link(adapter);
2014 
2015 	igb_update_mng_vlan(adapter);
2016 
2017 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2018 	wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2019 
2020 	/* Re-enable PTP, where applicable. */
2021 	igb_ptp_reset(adapter);
2022 
2023 	igb_get_phy_info(hw);
2024 }
2025 
2026 static netdev_features_t igb_fix_features(struct net_device *netdev,
2027 	netdev_features_t features)
2028 {
2029 	/* Since there is no support for separate Rx/Tx vlan accel
2030 	 * enable/disable make sure Tx flag is always in same state as Rx.
2031 	 */
2032 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2033 		features |= NETIF_F_HW_VLAN_CTAG_TX;
2034 	else
2035 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2036 
2037 	return features;
2038 }
2039 
2040 static int igb_set_features(struct net_device *netdev,
2041 	netdev_features_t features)
2042 {
2043 	netdev_features_t changed = netdev->features ^ features;
2044 	struct igb_adapter *adapter = netdev_priv(netdev);
2045 
2046 	if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2047 		igb_vlan_mode(netdev, features);
2048 
2049 	if (!(changed & NETIF_F_RXALL))
2050 		return 0;
2051 
2052 	netdev->features = features;
2053 
2054 	if (netif_running(netdev))
2055 		igb_reinit_locked(adapter);
2056 	else
2057 		igb_reset(adapter);
2058 
2059 	return 0;
2060 }
2061 
2062 static const struct net_device_ops igb_netdev_ops = {
2063 	.ndo_open		= igb_open,
2064 	.ndo_stop		= igb_close,
2065 	.ndo_start_xmit		= igb_xmit_frame,
2066 	.ndo_get_stats64	= igb_get_stats64,
2067 	.ndo_set_rx_mode	= igb_set_rx_mode,
2068 	.ndo_set_mac_address	= igb_set_mac,
2069 	.ndo_change_mtu		= igb_change_mtu,
2070 	.ndo_do_ioctl		= igb_ioctl,
2071 	.ndo_tx_timeout		= igb_tx_timeout,
2072 	.ndo_validate_addr	= eth_validate_addr,
2073 	.ndo_vlan_rx_add_vid	= igb_vlan_rx_add_vid,
2074 	.ndo_vlan_rx_kill_vid	= igb_vlan_rx_kill_vid,
2075 	.ndo_set_vf_mac		= igb_ndo_set_vf_mac,
2076 	.ndo_set_vf_vlan	= igb_ndo_set_vf_vlan,
2077 	.ndo_set_vf_rate	= igb_ndo_set_vf_bw,
2078 	.ndo_set_vf_spoofchk	= igb_ndo_set_vf_spoofchk,
2079 	.ndo_get_vf_config	= igb_ndo_get_vf_config,
2080 #ifdef CONFIG_NET_POLL_CONTROLLER
2081 	.ndo_poll_controller	= igb_netpoll,
2082 #endif
2083 	.ndo_fix_features	= igb_fix_features,
2084 	.ndo_set_features	= igb_set_features,
2085 	.ndo_features_check	= passthru_features_check,
2086 };
2087 
2088 /**
2089  * igb_set_fw_version - Configure version string for ethtool
2090  * @adapter: adapter struct
2091  **/
2092 void igb_set_fw_version(struct igb_adapter *adapter)
2093 {
2094 	struct e1000_hw *hw = &adapter->hw;
2095 	struct e1000_fw_version fw;
2096 
2097 	igb_get_fw_version(hw, &fw);
2098 
2099 	switch (hw->mac.type) {
2100 	case e1000_i210:
2101 	case e1000_i211:
2102 		if (!(igb_get_flash_presence_i210(hw))) {
2103 			snprintf(adapter->fw_version,
2104 				 sizeof(adapter->fw_version),
2105 				 "%2d.%2d-%d",
2106 				 fw.invm_major, fw.invm_minor,
2107 				 fw.invm_img_type);
2108 			break;
2109 		}
2110 		/* fall through */
2111 	default:
2112 		/* if option is rom valid, display its version too */
2113 		if (fw.or_valid) {
2114 			snprintf(adapter->fw_version,
2115 				 sizeof(adapter->fw_version),
2116 				 "%d.%d, 0x%08x, %d.%d.%d",
2117 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2118 				 fw.or_major, fw.or_build, fw.or_patch);
2119 		/* no option rom */
2120 		} else if (fw.etrack_id != 0X0000) {
2121 			snprintf(adapter->fw_version,
2122 			    sizeof(adapter->fw_version),
2123 			    "%d.%d, 0x%08x",
2124 			    fw.eep_major, fw.eep_minor, fw.etrack_id);
2125 		} else {
2126 		snprintf(adapter->fw_version,
2127 		    sizeof(adapter->fw_version),
2128 		    "%d.%d.%d",
2129 		    fw.eep_major, fw.eep_minor, fw.eep_build);
2130 		}
2131 		break;
2132 	}
2133 }
2134 
2135 /**
2136  * igb_init_mas - init Media Autosense feature if enabled in the NVM
2137  *
2138  * @adapter: adapter struct
2139  **/
2140 static void igb_init_mas(struct igb_adapter *adapter)
2141 {
2142 	struct e1000_hw *hw = &adapter->hw;
2143 	u16 eeprom_data;
2144 
2145 	hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
2146 	switch (hw->bus.func) {
2147 	case E1000_FUNC_0:
2148 		if (eeprom_data & IGB_MAS_ENABLE_0) {
2149 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2150 			netdev_info(adapter->netdev,
2151 				"MAS: Enabling Media Autosense for port %d\n",
2152 				hw->bus.func);
2153 		}
2154 		break;
2155 	case E1000_FUNC_1:
2156 		if (eeprom_data & IGB_MAS_ENABLE_1) {
2157 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2158 			netdev_info(adapter->netdev,
2159 				"MAS: Enabling Media Autosense for port %d\n",
2160 				hw->bus.func);
2161 		}
2162 		break;
2163 	case E1000_FUNC_2:
2164 		if (eeprom_data & IGB_MAS_ENABLE_2) {
2165 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2166 			netdev_info(adapter->netdev,
2167 				"MAS: Enabling Media Autosense for port %d\n",
2168 				hw->bus.func);
2169 		}
2170 		break;
2171 	case E1000_FUNC_3:
2172 		if (eeprom_data & IGB_MAS_ENABLE_3) {
2173 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2174 			netdev_info(adapter->netdev,
2175 				"MAS: Enabling Media Autosense for port %d\n",
2176 				hw->bus.func);
2177 		}
2178 		break;
2179 	default:
2180 		/* Shouldn't get here */
2181 		netdev_err(adapter->netdev,
2182 			"MAS: Invalid port configuration, returning\n");
2183 		break;
2184 	}
2185 }
2186 
2187 /**
2188  *  igb_init_i2c - Init I2C interface
2189  *  @adapter: pointer to adapter structure
2190  **/
2191 static s32 igb_init_i2c(struct igb_adapter *adapter)
2192 {
2193 	s32 status = 0;
2194 
2195 	/* I2C interface supported on i350 devices */
2196 	if (adapter->hw.mac.type != e1000_i350)
2197 		return 0;
2198 
2199 	/* Initialize the i2c bus which is controlled by the registers.
2200 	 * This bus will use the i2c_algo_bit structue that implements
2201 	 * the protocol through toggling of the 4 bits in the register.
2202 	 */
2203 	adapter->i2c_adap.owner = THIS_MODULE;
2204 	adapter->i2c_algo = igb_i2c_algo;
2205 	adapter->i2c_algo.data = adapter;
2206 	adapter->i2c_adap.algo_data = &adapter->i2c_algo;
2207 	adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
2208 	strlcpy(adapter->i2c_adap.name, "igb BB",
2209 		sizeof(adapter->i2c_adap.name));
2210 	status = i2c_bit_add_bus(&adapter->i2c_adap);
2211 	return status;
2212 }
2213 
2214 /**
2215  *  igb_probe - Device Initialization Routine
2216  *  @pdev: PCI device information struct
2217  *  @ent: entry in igb_pci_tbl
2218  *
2219  *  Returns 0 on success, negative on failure
2220  *
2221  *  igb_probe initializes an adapter identified by a pci_dev structure.
2222  *  The OS initialization, configuring of the adapter private structure,
2223  *  and a hardware reset occur.
2224  **/
2225 static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2226 {
2227 	struct net_device *netdev;
2228 	struct igb_adapter *adapter;
2229 	struct e1000_hw *hw;
2230 	u16 eeprom_data = 0;
2231 	s32 ret_val;
2232 	static int global_quad_port_a; /* global quad port a indication */
2233 	const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
2234 	int err, pci_using_dac;
2235 	u8 part_str[E1000_PBANUM_LENGTH];
2236 
2237 	/* Catch broken hardware that put the wrong VF device ID in
2238 	 * the PCIe SR-IOV capability.
2239 	 */
2240 	if (pdev->is_virtfn) {
2241 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
2242 			pci_name(pdev), pdev->vendor, pdev->device);
2243 		return -EINVAL;
2244 	}
2245 
2246 	err = pci_enable_device_mem(pdev);
2247 	if (err)
2248 		return err;
2249 
2250 	pci_using_dac = 0;
2251 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2252 	if (!err) {
2253 		pci_using_dac = 1;
2254 	} else {
2255 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
2256 		if (err) {
2257 			dev_err(&pdev->dev,
2258 				"No usable DMA configuration, aborting\n");
2259 			goto err_dma;
2260 		}
2261 	}
2262 
2263 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
2264 					   IORESOURCE_MEM),
2265 					   igb_driver_name);
2266 	if (err)
2267 		goto err_pci_reg;
2268 
2269 	pci_enable_pcie_error_reporting(pdev);
2270 
2271 	pci_set_master(pdev);
2272 	pci_save_state(pdev);
2273 
2274 	err = -ENOMEM;
2275 	netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
2276 				   IGB_MAX_TX_QUEUES);
2277 	if (!netdev)
2278 		goto err_alloc_etherdev;
2279 
2280 	SET_NETDEV_DEV(netdev, &pdev->dev);
2281 
2282 	pci_set_drvdata(pdev, netdev);
2283 	adapter = netdev_priv(netdev);
2284 	adapter->netdev = netdev;
2285 	adapter->pdev = pdev;
2286 	hw = &adapter->hw;
2287 	hw->back = adapter;
2288 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
2289 
2290 	err = -EIO;
2291 	hw->hw_addr = pci_iomap(pdev, 0, 0);
2292 	if (!hw->hw_addr)
2293 		goto err_ioremap;
2294 
2295 	netdev->netdev_ops = &igb_netdev_ops;
2296 	igb_set_ethtool_ops(netdev);
2297 	netdev->watchdog_timeo = 5 * HZ;
2298 
2299 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
2300 
2301 	netdev->mem_start = pci_resource_start(pdev, 0);
2302 	netdev->mem_end = pci_resource_end(pdev, 0);
2303 
2304 	/* PCI config space info */
2305 	hw->vendor_id = pdev->vendor;
2306 	hw->device_id = pdev->device;
2307 	hw->revision_id = pdev->revision;
2308 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
2309 	hw->subsystem_device_id = pdev->subsystem_device;
2310 
2311 	/* Copy the default MAC, PHY and NVM function pointers */
2312 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
2313 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
2314 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
2315 	/* Initialize skew-specific constants */
2316 	err = ei->get_invariants(hw);
2317 	if (err)
2318 		goto err_sw_init;
2319 
2320 	/* setup the private structure */
2321 	err = igb_sw_init(adapter);
2322 	if (err)
2323 		goto err_sw_init;
2324 
2325 	igb_get_bus_info_pcie(hw);
2326 
2327 	hw->phy.autoneg_wait_to_complete = false;
2328 
2329 	/* Copper options */
2330 	if (hw->phy.media_type == e1000_media_type_copper) {
2331 		hw->phy.mdix = AUTO_ALL_MODES;
2332 		hw->phy.disable_polarity_correction = false;
2333 		hw->phy.ms_type = e1000_ms_hw_default;
2334 	}
2335 
2336 	if (igb_check_reset_block(hw))
2337 		dev_info(&pdev->dev,
2338 			"PHY reset is blocked due to SOL/IDER session.\n");
2339 
2340 	/* features is initialized to 0 in allocation, it might have bits
2341 	 * set by igb_sw_init so we should use an or instead of an
2342 	 * assignment.
2343 	 */
2344 	netdev->features |= NETIF_F_SG |
2345 			    NETIF_F_IP_CSUM |
2346 			    NETIF_F_IPV6_CSUM |
2347 			    NETIF_F_TSO |
2348 			    NETIF_F_TSO6 |
2349 			    NETIF_F_RXHASH |
2350 			    NETIF_F_RXCSUM |
2351 			    NETIF_F_HW_VLAN_CTAG_RX |
2352 			    NETIF_F_HW_VLAN_CTAG_TX;
2353 
2354 	/* copy netdev features into list of user selectable features */
2355 	netdev->hw_features |= netdev->features;
2356 	netdev->hw_features |= NETIF_F_RXALL;
2357 
2358 	/* set this bit last since it cannot be part of hw_features */
2359 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
2360 
2361 	netdev->vlan_features |= NETIF_F_TSO |
2362 				 NETIF_F_TSO6 |
2363 				 NETIF_F_IP_CSUM |
2364 				 NETIF_F_IPV6_CSUM |
2365 				 NETIF_F_SG;
2366 
2367 	netdev->priv_flags |= IFF_SUPP_NOFCS;
2368 
2369 	if (pci_using_dac) {
2370 		netdev->features |= NETIF_F_HIGHDMA;
2371 		netdev->vlan_features |= NETIF_F_HIGHDMA;
2372 	}
2373 
2374 	if (hw->mac.type >= e1000_82576) {
2375 		netdev->hw_features |= NETIF_F_SCTP_CSUM;
2376 		netdev->features |= NETIF_F_SCTP_CSUM;
2377 	}
2378 
2379 	netdev->priv_flags |= IFF_UNICAST_FLT;
2380 
2381 	adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
2382 
2383 	/* before reading the NVM, reset the controller to put the device in a
2384 	 * known good starting state
2385 	 */
2386 	hw->mac.ops.reset_hw(hw);
2387 
2388 	/* make sure the NVM is good , i211/i210 parts can have special NVM
2389 	 * that doesn't contain a checksum
2390 	 */
2391 	switch (hw->mac.type) {
2392 	case e1000_i210:
2393 	case e1000_i211:
2394 		if (igb_get_flash_presence_i210(hw)) {
2395 			if (hw->nvm.ops.validate(hw) < 0) {
2396 				dev_err(&pdev->dev,
2397 					"The NVM Checksum Is Not Valid\n");
2398 				err = -EIO;
2399 				goto err_eeprom;
2400 			}
2401 		}
2402 		break;
2403 	default:
2404 		if (hw->nvm.ops.validate(hw) < 0) {
2405 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
2406 			err = -EIO;
2407 			goto err_eeprom;
2408 		}
2409 		break;
2410 	}
2411 
2412 	/* copy the MAC address out of the NVM */
2413 	if (hw->mac.ops.read_mac_addr(hw))
2414 		dev_err(&pdev->dev, "NVM Read Error\n");
2415 
2416 	memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
2417 
2418 	if (!is_valid_ether_addr(netdev->dev_addr)) {
2419 		dev_err(&pdev->dev, "Invalid MAC Address\n");
2420 		err = -EIO;
2421 		goto err_eeprom;
2422 	}
2423 
2424 	/* get firmware version for ethtool -i */
2425 	igb_set_fw_version(adapter);
2426 
2427 	/* configure RXPBSIZE and TXPBSIZE */
2428 	if (hw->mac.type == e1000_i210) {
2429 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
2430 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
2431 	}
2432 
2433 	setup_timer(&adapter->watchdog_timer, igb_watchdog,
2434 		    (unsigned long) adapter);
2435 	setup_timer(&adapter->phy_info_timer, igb_update_phy_info,
2436 		    (unsigned long) adapter);
2437 
2438 	INIT_WORK(&adapter->reset_task, igb_reset_task);
2439 	INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
2440 
2441 	/* Initialize link properties that are user-changeable */
2442 	adapter->fc_autoneg = true;
2443 	hw->mac.autoneg = true;
2444 	hw->phy.autoneg_advertised = 0x2f;
2445 
2446 	hw->fc.requested_mode = e1000_fc_default;
2447 	hw->fc.current_mode = e1000_fc_default;
2448 
2449 	igb_validate_mdi_setting(hw);
2450 
2451 	/* By default, support wake on port A */
2452 	if (hw->bus.func == 0)
2453 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2454 
2455 	/* Check the NVM for wake support on non-port A ports */
2456 	if (hw->mac.type >= e1000_82580)
2457 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
2458 				 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
2459 				 &eeprom_data);
2460 	else if (hw->bus.func == 1)
2461 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
2462 
2463 	if (eeprom_data & IGB_EEPROM_APME)
2464 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2465 
2466 	/* now that we have the eeprom settings, apply the special cases where
2467 	 * the eeprom may be wrong or the board simply won't support wake on
2468 	 * lan on a particular port
2469 	 */
2470 	switch (pdev->device) {
2471 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
2472 		adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2473 		break;
2474 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
2475 	case E1000_DEV_ID_82576_FIBER:
2476 	case E1000_DEV_ID_82576_SERDES:
2477 		/* Wake events only supported on port A for dual fiber
2478 		 * regardless of eeprom setting
2479 		 */
2480 		if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
2481 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2482 		break;
2483 	case E1000_DEV_ID_82576_QUAD_COPPER:
2484 	case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
2485 		/* if quad port adapter, disable WoL on all but port A */
2486 		if (global_quad_port_a != 0)
2487 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2488 		else
2489 			adapter->flags |= IGB_FLAG_QUAD_PORT_A;
2490 		/* Reset for multiple quad port adapters */
2491 		if (++global_quad_port_a == 4)
2492 			global_quad_port_a = 0;
2493 		break;
2494 	default:
2495 		/* If the device can't wake, don't set software support */
2496 		if (!device_can_wakeup(&adapter->pdev->dev))
2497 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2498 	}
2499 
2500 	/* initialize the wol settings based on the eeprom settings */
2501 	if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
2502 		adapter->wol |= E1000_WUFC_MAG;
2503 
2504 	/* Some vendors want WoL disabled by default, but still supported */
2505 	if ((hw->mac.type == e1000_i350) &&
2506 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
2507 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2508 		adapter->wol = 0;
2509 	}
2510 
2511 	device_set_wakeup_enable(&adapter->pdev->dev,
2512 				 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
2513 
2514 	/* reset the hardware with the new settings */
2515 	igb_reset(adapter);
2516 
2517 	/* Init the I2C interface */
2518 	err = igb_init_i2c(adapter);
2519 	if (err) {
2520 		dev_err(&pdev->dev, "failed to init i2c interface\n");
2521 		goto err_eeprom;
2522 	}
2523 
2524 	/* let the f/w know that the h/w is now under the control of the
2525 	 * driver.
2526 	 */
2527 	igb_get_hw_control(adapter);
2528 
2529 	strcpy(netdev->name, "eth%d");
2530 	err = register_netdev(netdev);
2531 	if (err)
2532 		goto err_register;
2533 
2534 	/* carrier off reporting is important to ethtool even BEFORE open */
2535 	netif_carrier_off(netdev);
2536 
2537 #ifdef CONFIG_IGB_DCA
2538 	if (dca_add_requester(&pdev->dev) == 0) {
2539 		adapter->flags |= IGB_FLAG_DCA_ENABLED;
2540 		dev_info(&pdev->dev, "DCA enabled\n");
2541 		igb_setup_dca(adapter);
2542 	}
2543 
2544 #endif
2545 #ifdef CONFIG_IGB_HWMON
2546 	/* Initialize the thermal sensor on i350 devices. */
2547 	if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
2548 		u16 ets_word;
2549 
2550 		/* Read the NVM to determine if this i350 device supports an
2551 		 * external thermal sensor.
2552 		 */
2553 		hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
2554 		if (ets_word != 0x0000 && ets_word != 0xFFFF)
2555 			adapter->ets = true;
2556 		else
2557 			adapter->ets = false;
2558 		if (igb_sysfs_init(adapter))
2559 			dev_err(&pdev->dev,
2560 				"failed to allocate sysfs resources\n");
2561 	} else {
2562 		adapter->ets = false;
2563 	}
2564 #endif
2565 	/* Check if Media Autosense is enabled */
2566 	adapter->ei = *ei;
2567 	if (hw->dev_spec._82575.mas_capable)
2568 		igb_init_mas(adapter);
2569 
2570 	/* do hw tstamp init after resetting */
2571 	igb_ptp_init(adapter);
2572 
2573 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
2574 	/* print bus type/speed/width info, not applicable to i354 */
2575 	if (hw->mac.type != e1000_i354) {
2576 		dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
2577 			 netdev->name,
2578 			 ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
2579 			  (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
2580 			   "unknown"),
2581 			 ((hw->bus.width == e1000_bus_width_pcie_x4) ?
2582 			  "Width x4" :
2583 			  (hw->bus.width == e1000_bus_width_pcie_x2) ?
2584 			  "Width x2" :
2585 			  (hw->bus.width == e1000_bus_width_pcie_x1) ?
2586 			  "Width x1" : "unknown"), netdev->dev_addr);
2587 	}
2588 
2589 	if ((hw->mac.type >= e1000_i210 ||
2590 	     igb_get_flash_presence_i210(hw))) {
2591 		ret_val = igb_read_part_string(hw, part_str,
2592 					       E1000_PBANUM_LENGTH);
2593 	} else {
2594 		ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
2595 	}
2596 
2597 	if (ret_val)
2598 		strcpy(part_str, "Unknown");
2599 	dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
2600 	dev_info(&pdev->dev,
2601 		"Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
2602 		(adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
2603 		(adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
2604 		adapter->num_rx_queues, adapter->num_tx_queues);
2605 	if (hw->phy.media_type == e1000_media_type_copper) {
2606 		switch (hw->mac.type) {
2607 		case e1000_i350:
2608 		case e1000_i210:
2609 		case e1000_i211:
2610 			/* Enable EEE for internal copper PHY devices */
2611 			err = igb_set_eee_i350(hw, true, true);
2612 			if ((!err) &&
2613 			    (!hw->dev_spec._82575.eee_disable)) {
2614 				adapter->eee_advert =
2615 					MDIO_EEE_100TX | MDIO_EEE_1000T;
2616 				adapter->flags |= IGB_FLAG_EEE;
2617 			}
2618 			break;
2619 		case e1000_i354:
2620 			if ((rd32(E1000_CTRL_EXT) &
2621 			    E1000_CTRL_EXT_LINK_MODE_SGMII)) {
2622 				err = igb_set_eee_i354(hw, true, true);
2623 				if ((!err) &&
2624 					(!hw->dev_spec._82575.eee_disable)) {
2625 					adapter->eee_advert =
2626 					   MDIO_EEE_100TX | MDIO_EEE_1000T;
2627 					adapter->flags |= IGB_FLAG_EEE;
2628 				}
2629 			}
2630 			break;
2631 		default:
2632 			break;
2633 		}
2634 	}
2635 	pm_runtime_put_noidle(&pdev->dev);
2636 	return 0;
2637 
2638 err_register:
2639 	igb_release_hw_control(adapter);
2640 	memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
2641 err_eeprom:
2642 	if (!igb_check_reset_block(hw))
2643 		igb_reset_phy(hw);
2644 
2645 	if (hw->flash_address)
2646 		iounmap(hw->flash_address);
2647 err_sw_init:
2648 	igb_clear_interrupt_scheme(adapter);
2649 	pci_iounmap(pdev, hw->hw_addr);
2650 err_ioremap:
2651 	free_netdev(netdev);
2652 err_alloc_etherdev:
2653 	pci_release_selected_regions(pdev,
2654 				     pci_select_bars(pdev, IORESOURCE_MEM));
2655 err_pci_reg:
2656 err_dma:
2657 	pci_disable_device(pdev);
2658 	return err;
2659 }
2660 
2661 #ifdef CONFIG_PCI_IOV
2662 static int igb_disable_sriov(struct pci_dev *pdev)
2663 {
2664 	struct net_device *netdev = pci_get_drvdata(pdev);
2665 	struct igb_adapter *adapter = netdev_priv(netdev);
2666 	struct e1000_hw *hw = &adapter->hw;
2667 
2668 	/* reclaim resources allocated to VFs */
2669 	if (adapter->vf_data) {
2670 		/* disable iov and allow time for transactions to clear */
2671 		if (pci_vfs_assigned(pdev)) {
2672 			dev_warn(&pdev->dev,
2673 				 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
2674 			return -EPERM;
2675 		} else {
2676 			pci_disable_sriov(pdev);
2677 			msleep(500);
2678 		}
2679 
2680 		kfree(adapter->vf_data);
2681 		adapter->vf_data = NULL;
2682 		adapter->vfs_allocated_count = 0;
2683 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
2684 		wrfl();
2685 		msleep(100);
2686 		dev_info(&pdev->dev, "IOV Disabled\n");
2687 
2688 		/* Re-enable DMA Coalescing flag since IOV is turned off */
2689 		adapter->flags |= IGB_FLAG_DMAC;
2690 	}
2691 
2692 	return 0;
2693 }
2694 
2695 static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
2696 {
2697 	struct net_device *netdev = pci_get_drvdata(pdev);
2698 	struct igb_adapter *adapter = netdev_priv(netdev);
2699 	int old_vfs = pci_num_vf(pdev);
2700 	int err = 0;
2701 	int i;
2702 
2703 	if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
2704 		err = -EPERM;
2705 		goto out;
2706 	}
2707 	if (!num_vfs)
2708 		goto out;
2709 
2710 	if (old_vfs) {
2711 		dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
2712 			 old_vfs, max_vfs);
2713 		adapter->vfs_allocated_count = old_vfs;
2714 	} else
2715 		adapter->vfs_allocated_count = num_vfs;
2716 
2717 	adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
2718 				sizeof(struct vf_data_storage), GFP_KERNEL);
2719 
2720 	/* if allocation failed then we do not support SR-IOV */
2721 	if (!adapter->vf_data) {
2722 		adapter->vfs_allocated_count = 0;
2723 		dev_err(&pdev->dev,
2724 			"Unable to allocate memory for VF Data Storage\n");
2725 		err = -ENOMEM;
2726 		goto out;
2727 	}
2728 
2729 	/* only call pci_enable_sriov() if no VFs are allocated already */
2730 	if (!old_vfs) {
2731 		err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
2732 		if (err)
2733 			goto err_out;
2734 	}
2735 	dev_info(&pdev->dev, "%d VFs allocated\n",
2736 		 adapter->vfs_allocated_count);
2737 	for (i = 0; i < adapter->vfs_allocated_count; i++)
2738 		igb_vf_configure(adapter, i);
2739 
2740 	/* DMA Coalescing is not supported in IOV mode. */
2741 	adapter->flags &= ~IGB_FLAG_DMAC;
2742 	goto out;
2743 
2744 err_out:
2745 	kfree(adapter->vf_data);
2746 	adapter->vf_data = NULL;
2747 	adapter->vfs_allocated_count = 0;
2748 out:
2749 	return err;
2750 }
2751 
2752 #endif
2753 /**
2754  *  igb_remove_i2c - Cleanup  I2C interface
2755  *  @adapter: pointer to adapter structure
2756  **/
2757 static void igb_remove_i2c(struct igb_adapter *adapter)
2758 {
2759 	/* free the adapter bus structure */
2760 	i2c_del_adapter(&adapter->i2c_adap);
2761 }
2762 
2763 /**
2764  *  igb_remove - Device Removal Routine
2765  *  @pdev: PCI device information struct
2766  *
2767  *  igb_remove is called by the PCI subsystem to alert the driver
2768  *  that it should release a PCI device.  The could be caused by a
2769  *  Hot-Plug event, or because the driver is going to be removed from
2770  *  memory.
2771  **/
2772 static void igb_remove(struct pci_dev *pdev)
2773 {
2774 	struct net_device *netdev = pci_get_drvdata(pdev);
2775 	struct igb_adapter *adapter = netdev_priv(netdev);
2776 	struct e1000_hw *hw = &adapter->hw;
2777 
2778 	pm_runtime_get_noresume(&pdev->dev);
2779 #ifdef CONFIG_IGB_HWMON
2780 	igb_sysfs_exit(adapter);
2781 #endif
2782 	igb_remove_i2c(adapter);
2783 	igb_ptp_stop(adapter);
2784 	/* The watchdog timer may be rescheduled, so explicitly
2785 	 * disable watchdog from being rescheduled.
2786 	 */
2787 	set_bit(__IGB_DOWN, &adapter->state);
2788 	del_timer_sync(&adapter->watchdog_timer);
2789 	del_timer_sync(&adapter->phy_info_timer);
2790 
2791 	cancel_work_sync(&adapter->reset_task);
2792 	cancel_work_sync(&adapter->watchdog_task);
2793 
2794 #ifdef CONFIG_IGB_DCA
2795 	if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
2796 		dev_info(&pdev->dev, "DCA disabled\n");
2797 		dca_remove_requester(&pdev->dev);
2798 		adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
2799 		wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
2800 	}
2801 #endif
2802 
2803 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
2804 	 * would have already happened in close and is redundant.
2805 	 */
2806 	igb_release_hw_control(adapter);
2807 
2808 	unregister_netdev(netdev);
2809 
2810 	igb_clear_interrupt_scheme(adapter);
2811 
2812 #ifdef CONFIG_PCI_IOV
2813 	igb_disable_sriov(pdev);
2814 #endif
2815 
2816 	pci_iounmap(pdev, hw->hw_addr);
2817 	if (hw->flash_address)
2818 		iounmap(hw->flash_address);
2819 	pci_release_selected_regions(pdev,
2820 				     pci_select_bars(pdev, IORESOURCE_MEM));
2821 
2822 	kfree(adapter->shadow_vfta);
2823 	free_netdev(netdev);
2824 
2825 	pci_disable_pcie_error_reporting(pdev);
2826 
2827 	pci_disable_device(pdev);
2828 }
2829 
2830 /**
2831  *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
2832  *  @adapter: board private structure to initialize
2833  *
2834  *  This function initializes the vf specific data storage and then attempts to
2835  *  allocate the VFs.  The reason for ordering it this way is because it is much
2836  *  mor expensive time wise to disable SR-IOV than it is to allocate and free
2837  *  the memory for the VFs.
2838  **/
2839 static void igb_probe_vfs(struct igb_adapter *adapter)
2840 {
2841 #ifdef CONFIG_PCI_IOV
2842 	struct pci_dev *pdev = adapter->pdev;
2843 	struct e1000_hw *hw = &adapter->hw;
2844 
2845 	/* Virtualization features not supported on i210 family. */
2846 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
2847 		return;
2848 
2849 	pci_sriov_set_totalvfs(pdev, 7);
2850 	igb_pci_enable_sriov(pdev, max_vfs);
2851 
2852 #endif /* CONFIG_PCI_IOV */
2853 }
2854 
2855 static void igb_init_queue_configuration(struct igb_adapter *adapter)
2856 {
2857 	struct e1000_hw *hw = &adapter->hw;
2858 	u32 max_rss_queues;
2859 
2860 	/* Determine the maximum number of RSS queues supported. */
2861 	switch (hw->mac.type) {
2862 	case e1000_i211:
2863 		max_rss_queues = IGB_MAX_RX_QUEUES_I211;
2864 		break;
2865 	case e1000_82575:
2866 	case e1000_i210:
2867 		max_rss_queues = IGB_MAX_RX_QUEUES_82575;
2868 		break;
2869 	case e1000_i350:
2870 		/* I350 cannot do RSS and SR-IOV at the same time */
2871 		if (!!adapter->vfs_allocated_count) {
2872 			max_rss_queues = 1;
2873 			break;
2874 		}
2875 		/* fall through */
2876 	case e1000_82576:
2877 		if (!!adapter->vfs_allocated_count) {
2878 			max_rss_queues = 2;
2879 			break;
2880 		}
2881 		/* fall through */
2882 	case e1000_82580:
2883 	case e1000_i354:
2884 	default:
2885 		max_rss_queues = IGB_MAX_RX_QUEUES;
2886 		break;
2887 	}
2888 
2889 	adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
2890 
2891 	/* Determine if we need to pair queues. */
2892 	switch (hw->mac.type) {
2893 	case e1000_82575:
2894 	case e1000_i211:
2895 		/* Device supports enough interrupts without queue pairing. */
2896 		break;
2897 	case e1000_82576:
2898 		/* If VFs are going to be allocated with RSS queues then we
2899 		 * should pair the queues in order to conserve interrupts due
2900 		 * to limited supply.
2901 		 */
2902 		if ((adapter->rss_queues > 1) &&
2903 		    (adapter->vfs_allocated_count > 6))
2904 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
2905 		/* fall through */
2906 	case e1000_82580:
2907 	case e1000_i350:
2908 	case e1000_i354:
2909 	case e1000_i210:
2910 	default:
2911 		/* If rss_queues > half of max_rss_queues, pair the queues in
2912 		 * order to conserve interrupts due to limited supply.
2913 		 */
2914 		if (adapter->rss_queues > (max_rss_queues / 2))
2915 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
2916 		break;
2917 	}
2918 }
2919 
2920 /**
2921  *  igb_sw_init - Initialize general software structures (struct igb_adapter)
2922  *  @adapter: board private structure to initialize
2923  *
2924  *  igb_sw_init initializes the Adapter private data structure.
2925  *  Fields are initialized based on PCI device information and
2926  *  OS network device settings (MTU size).
2927  **/
2928 static int igb_sw_init(struct igb_adapter *adapter)
2929 {
2930 	struct e1000_hw *hw = &adapter->hw;
2931 	struct net_device *netdev = adapter->netdev;
2932 	struct pci_dev *pdev = adapter->pdev;
2933 
2934 	pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
2935 
2936 	/* set default ring sizes */
2937 	adapter->tx_ring_count = IGB_DEFAULT_TXD;
2938 	adapter->rx_ring_count = IGB_DEFAULT_RXD;
2939 
2940 	/* set default ITR values */
2941 	adapter->rx_itr_setting = IGB_DEFAULT_ITR;
2942 	adapter->tx_itr_setting = IGB_DEFAULT_ITR;
2943 
2944 	/* set default work limits */
2945 	adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
2946 
2947 	adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN +
2948 				  VLAN_HLEN;
2949 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
2950 
2951 	spin_lock_init(&adapter->stats64_lock);
2952 #ifdef CONFIG_PCI_IOV
2953 	switch (hw->mac.type) {
2954 	case e1000_82576:
2955 	case e1000_i350:
2956 		if (max_vfs > 7) {
2957 			dev_warn(&pdev->dev,
2958 				 "Maximum of 7 VFs per PF, using max\n");
2959 			max_vfs = adapter->vfs_allocated_count = 7;
2960 		} else
2961 			adapter->vfs_allocated_count = max_vfs;
2962 		if (adapter->vfs_allocated_count)
2963 			dev_warn(&pdev->dev,
2964 				 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
2965 		break;
2966 	default:
2967 		break;
2968 	}
2969 #endif /* CONFIG_PCI_IOV */
2970 
2971 	igb_init_queue_configuration(adapter);
2972 
2973 	/* Setup and initialize a copy of the hw vlan table array */
2974 	adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
2975 				       GFP_ATOMIC);
2976 
2977 	/* This call may decrease the number of queues */
2978 	if (igb_init_interrupt_scheme(adapter, true)) {
2979 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
2980 		return -ENOMEM;
2981 	}
2982 
2983 	igb_probe_vfs(adapter);
2984 
2985 	/* Explicitly disable IRQ since the NIC can be in any state. */
2986 	igb_irq_disable(adapter);
2987 
2988 	if (hw->mac.type >= e1000_i350)
2989 		adapter->flags &= ~IGB_FLAG_DMAC;
2990 
2991 	set_bit(__IGB_DOWN, &adapter->state);
2992 	return 0;
2993 }
2994 
2995 /**
2996  *  igb_open - Called when a network interface is made active
2997  *  @netdev: network interface device structure
2998  *
2999  *  Returns 0 on success, negative value on failure
3000  *
3001  *  The open entry point is called when a network interface is made
3002  *  active by the system (IFF_UP).  At this point all resources needed
3003  *  for transmit and receive operations are allocated, the interrupt
3004  *  handler is registered with the OS, the watchdog timer is started,
3005  *  and the stack is notified that the interface is ready.
3006  **/
3007 static int __igb_open(struct net_device *netdev, bool resuming)
3008 {
3009 	struct igb_adapter *adapter = netdev_priv(netdev);
3010 	struct e1000_hw *hw = &adapter->hw;
3011 	struct pci_dev *pdev = adapter->pdev;
3012 	int err;
3013 	int i;
3014 
3015 	/* disallow open during test */
3016 	if (test_bit(__IGB_TESTING, &adapter->state)) {
3017 		WARN_ON(resuming);
3018 		return -EBUSY;
3019 	}
3020 
3021 	if (!resuming)
3022 		pm_runtime_get_sync(&pdev->dev);
3023 
3024 	netif_carrier_off(netdev);
3025 
3026 	/* allocate transmit descriptors */
3027 	err = igb_setup_all_tx_resources(adapter);
3028 	if (err)
3029 		goto err_setup_tx;
3030 
3031 	/* allocate receive descriptors */
3032 	err = igb_setup_all_rx_resources(adapter);
3033 	if (err)
3034 		goto err_setup_rx;
3035 
3036 	igb_power_up_link(adapter);
3037 
3038 	/* before we allocate an interrupt, we must be ready to handle it.
3039 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3040 	 * as soon as we call pci_request_irq, so we have to setup our
3041 	 * clean_rx handler before we do so.
3042 	 */
3043 	igb_configure(adapter);
3044 
3045 	err = igb_request_irq(adapter);
3046 	if (err)
3047 		goto err_req_irq;
3048 
3049 	/* Notify the stack of the actual queue counts. */
3050 	err = netif_set_real_num_tx_queues(adapter->netdev,
3051 					   adapter->num_tx_queues);
3052 	if (err)
3053 		goto err_set_queues;
3054 
3055 	err = netif_set_real_num_rx_queues(adapter->netdev,
3056 					   adapter->num_rx_queues);
3057 	if (err)
3058 		goto err_set_queues;
3059 
3060 	/* From here on the code is the same as igb_up() */
3061 	clear_bit(__IGB_DOWN, &adapter->state);
3062 
3063 	for (i = 0; i < adapter->num_q_vectors; i++)
3064 		napi_enable(&(adapter->q_vector[i]->napi));
3065 
3066 	/* Clear any pending interrupts. */
3067 	rd32(E1000_ICR);
3068 
3069 	igb_irq_enable(adapter);
3070 
3071 	/* notify VFs that reset has been completed */
3072 	if (adapter->vfs_allocated_count) {
3073 		u32 reg_data = rd32(E1000_CTRL_EXT);
3074 
3075 		reg_data |= E1000_CTRL_EXT_PFRSTD;
3076 		wr32(E1000_CTRL_EXT, reg_data);
3077 	}
3078 
3079 	netif_tx_start_all_queues(netdev);
3080 
3081 	if (!resuming)
3082 		pm_runtime_put(&pdev->dev);
3083 
3084 	/* start the watchdog. */
3085 	hw->mac.get_link_status = 1;
3086 	schedule_work(&adapter->watchdog_task);
3087 
3088 	return 0;
3089 
3090 err_set_queues:
3091 	igb_free_irq(adapter);
3092 err_req_irq:
3093 	igb_release_hw_control(adapter);
3094 	igb_power_down_link(adapter);
3095 	igb_free_all_rx_resources(adapter);
3096 err_setup_rx:
3097 	igb_free_all_tx_resources(adapter);
3098 err_setup_tx:
3099 	igb_reset(adapter);
3100 	if (!resuming)
3101 		pm_runtime_put(&pdev->dev);
3102 
3103 	return err;
3104 }
3105 
3106 static int igb_open(struct net_device *netdev)
3107 {
3108 	return __igb_open(netdev, false);
3109 }
3110 
3111 /**
3112  *  igb_close - Disables a network interface
3113  *  @netdev: network interface device structure
3114  *
3115  *  Returns 0, this is not allowed to fail
3116  *
3117  *  The close entry point is called when an interface is de-activated
3118  *  by the OS.  The hardware is still under the driver's control, but
3119  *  needs to be disabled.  A global MAC reset is issued to stop the
3120  *  hardware, and all transmit and receive resources are freed.
3121  **/
3122 static int __igb_close(struct net_device *netdev, bool suspending)
3123 {
3124 	struct igb_adapter *adapter = netdev_priv(netdev);
3125 	struct pci_dev *pdev = adapter->pdev;
3126 
3127 	WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
3128 
3129 	if (!suspending)
3130 		pm_runtime_get_sync(&pdev->dev);
3131 
3132 	igb_down(adapter);
3133 	igb_free_irq(adapter);
3134 
3135 	igb_free_all_tx_resources(adapter);
3136 	igb_free_all_rx_resources(adapter);
3137 
3138 	if (!suspending)
3139 		pm_runtime_put_sync(&pdev->dev);
3140 	return 0;
3141 }
3142 
3143 static int igb_close(struct net_device *netdev)
3144 {
3145 	return __igb_close(netdev, false);
3146 }
3147 
3148 /**
3149  *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
3150  *  @tx_ring: tx descriptor ring (for a specific queue) to setup
3151  *
3152  *  Return 0 on success, negative on failure
3153  **/
3154 int igb_setup_tx_resources(struct igb_ring *tx_ring)
3155 {
3156 	struct device *dev = tx_ring->dev;
3157 	int size;
3158 
3159 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3160 
3161 	tx_ring->tx_buffer_info = vzalloc(size);
3162 	if (!tx_ring->tx_buffer_info)
3163 		goto err;
3164 
3165 	/* round up to nearest 4K */
3166 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
3167 	tx_ring->size = ALIGN(tx_ring->size, 4096);
3168 
3169 	tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
3170 					   &tx_ring->dma, GFP_KERNEL);
3171 	if (!tx_ring->desc)
3172 		goto err;
3173 
3174 	tx_ring->next_to_use = 0;
3175 	tx_ring->next_to_clean = 0;
3176 
3177 	return 0;
3178 
3179 err:
3180 	vfree(tx_ring->tx_buffer_info);
3181 	tx_ring->tx_buffer_info = NULL;
3182 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
3183 	return -ENOMEM;
3184 }
3185 
3186 /**
3187  *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
3188  *				 (Descriptors) for all queues
3189  *  @adapter: board private structure
3190  *
3191  *  Return 0 on success, negative on failure
3192  **/
3193 static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
3194 {
3195 	struct pci_dev *pdev = adapter->pdev;
3196 	int i, err = 0;
3197 
3198 	for (i = 0; i < adapter->num_tx_queues; i++) {
3199 		err = igb_setup_tx_resources(adapter->tx_ring[i]);
3200 		if (err) {
3201 			dev_err(&pdev->dev,
3202 				"Allocation for Tx Queue %u failed\n", i);
3203 			for (i--; i >= 0; i--)
3204 				igb_free_tx_resources(adapter->tx_ring[i]);
3205 			break;
3206 		}
3207 	}
3208 
3209 	return err;
3210 }
3211 
3212 /**
3213  *  igb_setup_tctl - configure the transmit control registers
3214  *  @adapter: Board private structure
3215  **/
3216 void igb_setup_tctl(struct igb_adapter *adapter)
3217 {
3218 	struct e1000_hw *hw = &adapter->hw;
3219 	u32 tctl;
3220 
3221 	/* disable queue 0 which is enabled by default on 82575 and 82576 */
3222 	wr32(E1000_TXDCTL(0), 0);
3223 
3224 	/* Program the Transmit Control Register */
3225 	tctl = rd32(E1000_TCTL);
3226 	tctl &= ~E1000_TCTL_CT;
3227 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
3228 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
3229 
3230 	igb_config_collision_dist(hw);
3231 
3232 	/* Enable transmits */
3233 	tctl |= E1000_TCTL_EN;
3234 
3235 	wr32(E1000_TCTL, tctl);
3236 }
3237 
3238 /**
3239  *  igb_configure_tx_ring - Configure transmit ring after Reset
3240  *  @adapter: board private structure
3241  *  @ring: tx ring to configure
3242  *
3243  *  Configure a transmit ring after a reset.
3244  **/
3245 void igb_configure_tx_ring(struct igb_adapter *adapter,
3246 			   struct igb_ring *ring)
3247 {
3248 	struct e1000_hw *hw = &adapter->hw;
3249 	u32 txdctl = 0;
3250 	u64 tdba = ring->dma;
3251 	int reg_idx = ring->reg_idx;
3252 
3253 	/* disable the queue */
3254 	wr32(E1000_TXDCTL(reg_idx), 0);
3255 	wrfl();
3256 	mdelay(10);
3257 
3258 	wr32(E1000_TDLEN(reg_idx),
3259 	     ring->count * sizeof(union e1000_adv_tx_desc));
3260 	wr32(E1000_TDBAL(reg_idx),
3261 	     tdba & 0x00000000ffffffffULL);
3262 	wr32(E1000_TDBAH(reg_idx), tdba >> 32);
3263 
3264 	ring->tail = hw->hw_addr + E1000_TDT(reg_idx);
3265 	wr32(E1000_TDH(reg_idx), 0);
3266 	writel(0, ring->tail);
3267 
3268 	txdctl |= IGB_TX_PTHRESH;
3269 	txdctl |= IGB_TX_HTHRESH << 8;
3270 	txdctl |= IGB_TX_WTHRESH << 16;
3271 
3272 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
3273 	wr32(E1000_TXDCTL(reg_idx), txdctl);
3274 }
3275 
3276 /**
3277  *  igb_configure_tx - Configure transmit Unit after Reset
3278  *  @adapter: board private structure
3279  *
3280  *  Configure the Tx unit of the MAC after a reset.
3281  **/
3282 static void igb_configure_tx(struct igb_adapter *adapter)
3283 {
3284 	int i;
3285 
3286 	for (i = 0; i < adapter->num_tx_queues; i++)
3287 		igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
3288 }
3289 
3290 /**
3291  *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
3292  *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
3293  *
3294  *  Returns 0 on success, negative on failure
3295  **/
3296 int igb_setup_rx_resources(struct igb_ring *rx_ring)
3297 {
3298 	struct device *dev = rx_ring->dev;
3299 	int size;
3300 
3301 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3302 
3303 	rx_ring->rx_buffer_info = vzalloc(size);
3304 	if (!rx_ring->rx_buffer_info)
3305 		goto err;
3306 
3307 	/* Round up to nearest 4K */
3308 	rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
3309 	rx_ring->size = ALIGN(rx_ring->size, 4096);
3310 
3311 	rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
3312 					   &rx_ring->dma, GFP_KERNEL);
3313 	if (!rx_ring->desc)
3314 		goto err;
3315 
3316 	rx_ring->next_to_alloc = 0;
3317 	rx_ring->next_to_clean = 0;
3318 	rx_ring->next_to_use = 0;
3319 
3320 	return 0;
3321 
3322 err:
3323 	vfree(rx_ring->rx_buffer_info);
3324 	rx_ring->rx_buffer_info = NULL;
3325 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
3326 	return -ENOMEM;
3327 }
3328 
3329 /**
3330  *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
3331  *				 (Descriptors) for all queues
3332  *  @adapter: board private structure
3333  *
3334  *  Return 0 on success, negative on failure
3335  **/
3336 static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
3337 {
3338 	struct pci_dev *pdev = adapter->pdev;
3339 	int i, err = 0;
3340 
3341 	for (i = 0; i < adapter->num_rx_queues; i++) {
3342 		err = igb_setup_rx_resources(adapter->rx_ring[i]);
3343 		if (err) {
3344 			dev_err(&pdev->dev,
3345 				"Allocation for Rx Queue %u failed\n", i);
3346 			for (i--; i >= 0; i--)
3347 				igb_free_rx_resources(adapter->rx_ring[i]);
3348 			break;
3349 		}
3350 	}
3351 
3352 	return err;
3353 }
3354 
3355 /**
3356  *  igb_setup_mrqc - configure the multiple receive queue control registers
3357  *  @adapter: Board private structure
3358  **/
3359 static void igb_setup_mrqc(struct igb_adapter *adapter)
3360 {
3361 	struct e1000_hw *hw = &adapter->hw;
3362 	u32 mrqc, rxcsum;
3363 	u32 j, num_rx_queues;
3364 	u32 rss_key[10];
3365 
3366 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3367 	for (j = 0; j < 10; j++)
3368 		wr32(E1000_RSSRK(j), rss_key[j]);
3369 
3370 	num_rx_queues = adapter->rss_queues;
3371 
3372 	switch (hw->mac.type) {
3373 	case e1000_82576:
3374 		/* 82576 supports 2 RSS queues for SR-IOV */
3375 		if (adapter->vfs_allocated_count)
3376 			num_rx_queues = 2;
3377 		break;
3378 	default:
3379 		break;
3380 	}
3381 
3382 	if (adapter->rss_indir_tbl_init != num_rx_queues) {
3383 		for (j = 0; j < IGB_RETA_SIZE; j++)
3384 			adapter->rss_indir_tbl[j] =
3385 			(j * num_rx_queues) / IGB_RETA_SIZE;
3386 		adapter->rss_indir_tbl_init = num_rx_queues;
3387 	}
3388 	igb_write_rss_indir_tbl(adapter);
3389 
3390 	/* Disable raw packet checksumming so that RSS hash is placed in
3391 	 * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
3392 	 * offloads as they are enabled by default
3393 	 */
3394 	rxcsum = rd32(E1000_RXCSUM);
3395 	rxcsum |= E1000_RXCSUM_PCSD;
3396 
3397 	if (adapter->hw.mac.type >= e1000_82576)
3398 		/* Enable Receive Checksum Offload for SCTP */
3399 		rxcsum |= E1000_RXCSUM_CRCOFL;
3400 
3401 	/* Don't need to set TUOFL or IPOFL, they default to 1 */
3402 	wr32(E1000_RXCSUM, rxcsum);
3403 
3404 	/* Generate RSS hash based on packet types, TCP/UDP
3405 	 * port numbers and/or IPv4/v6 src and dst addresses
3406 	 */
3407 	mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
3408 	       E1000_MRQC_RSS_FIELD_IPV4_TCP |
3409 	       E1000_MRQC_RSS_FIELD_IPV6 |
3410 	       E1000_MRQC_RSS_FIELD_IPV6_TCP |
3411 	       E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
3412 
3413 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
3414 		mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
3415 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
3416 		mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
3417 
3418 	/* If VMDq is enabled then we set the appropriate mode for that, else
3419 	 * we default to RSS so that an RSS hash is calculated per packet even
3420 	 * if we are only using one queue
3421 	 */
3422 	if (adapter->vfs_allocated_count) {
3423 		if (hw->mac.type > e1000_82575) {
3424 			/* Set the default pool for the PF's first queue */
3425 			u32 vtctl = rd32(E1000_VT_CTL);
3426 
3427 			vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
3428 				   E1000_VT_CTL_DISABLE_DEF_POOL);
3429 			vtctl |= adapter->vfs_allocated_count <<
3430 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
3431 			wr32(E1000_VT_CTL, vtctl);
3432 		}
3433 		if (adapter->rss_queues > 1)
3434 			mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_2Q;
3435 		else
3436 			mrqc |= E1000_MRQC_ENABLE_VMDQ;
3437 	} else {
3438 		if (hw->mac.type != e1000_i211)
3439 			mrqc |= E1000_MRQC_ENABLE_RSS_4Q;
3440 	}
3441 	igb_vmm_control(adapter);
3442 
3443 	wr32(E1000_MRQC, mrqc);
3444 }
3445 
3446 /**
3447  *  igb_setup_rctl - configure the receive control registers
3448  *  @adapter: Board private structure
3449  **/
3450 void igb_setup_rctl(struct igb_adapter *adapter)
3451 {
3452 	struct e1000_hw *hw = &adapter->hw;
3453 	u32 rctl;
3454 
3455 	rctl = rd32(E1000_RCTL);
3456 
3457 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3458 	rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
3459 
3460 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
3461 		(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3462 
3463 	/* enable stripping of CRC. It's unlikely this will break BMC
3464 	 * redirection as it did with e1000. Newer features require
3465 	 * that the HW strips the CRC.
3466 	 */
3467 	rctl |= E1000_RCTL_SECRC;
3468 
3469 	/* disable store bad packets and clear size bits. */
3470 	rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
3471 
3472 	/* enable LPE to prevent packets larger than max_frame_size */
3473 	rctl |= E1000_RCTL_LPE;
3474 
3475 	/* disable queue 0 to prevent tail write w/o re-config */
3476 	wr32(E1000_RXDCTL(0), 0);
3477 
3478 	/* Attention!!!  For SR-IOV PF driver operations you must enable
3479 	 * queue drop for all VF and PF queues to prevent head of line blocking
3480 	 * if an un-trusted VF does not provide descriptors to hardware.
3481 	 */
3482 	if (adapter->vfs_allocated_count) {
3483 		/* set all queue drop enable bits */
3484 		wr32(E1000_QDE, ALL_QUEUES);
3485 	}
3486 
3487 	/* This is useful for sniffing bad packets. */
3488 	if (adapter->netdev->features & NETIF_F_RXALL) {
3489 		/* UPE and MPE will be handled by normal PROMISC logic
3490 		 * in e1000e_set_rx_mode
3491 		 */
3492 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
3493 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
3494 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
3495 
3496 		rctl &= ~(E1000_RCTL_VFE | /* Disable VLAN filter */
3497 			  E1000_RCTL_DPF | /* Allow filtered pause */
3498 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
3499 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3500 		 * and that breaks VLANs.
3501 		 */
3502 	}
3503 
3504 	wr32(E1000_RCTL, rctl);
3505 }
3506 
3507 static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
3508 				   int vfn)
3509 {
3510 	struct e1000_hw *hw = &adapter->hw;
3511 	u32 vmolr;
3512 
3513 	/* if it isn't the PF check to see if VFs are enabled and
3514 	 * increase the size to support vlan tags
3515 	 */
3516 	if (vfn < adapter->vfs_allocated_count &&
3517 	    adapter->vf_data[vfn].vlans_enabled)
3518 		size += VLAN_TAG_SIZE;
3519 
3520 	vmolr = rd32(E1000_VMOLR(vfn));
3521 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
3522 	vmolr |= size | E1000_VMOLR_LPE;
3523 	wr32(E1000_VMOLR(vfn), vmolr);
3524 
3525 	return 0;
3526 }
3527 
3528 /**
3529  *  igb_rlpml_set - set maximum receive packet size
3530  *  @adapter: board private structure
3531  *
3532  *  Configure maximum receivable packet size.
3533  **/
3534 static void igb_rlpml_set(struct igb_adapter *adapter)
3535 {
3536 	u32 max_frame_size = adapter->max_frame_size;
3537 	struct e1000_hw *hw = &adapter->hw;
3538 	u16 pf_id = adapter->vfs_allocated_count;
3539 
3540 	if (pf_id) {
3541 		igb_set_vf_rlpml(adapter, max_frame_size, pf_id);
3542 		/* If we're in VMDQ or SR-IOV mode, then set global RLPML
3543 		 * to our max jumbo frame size, in case we need to enable
3544 		 * jumbo frames on one of the rings later.
3545 		 * This will not pass over-length frames into the default
3546 		 * queue because it's gated by the VMOLR.RLPML.
3547 		 */
3548 		max_frame_size = MAX_JUMBO_FRAME_SIZE;
3549 	}
3550 
3551 	wr32(E1000_RLPML, max_frame_size);
3552 }
3553 
3554 static inline void igb_set_vmolr(struct igb_adapter *adapter,
3555 				 int vfn, bool aupe)
3556 {
3557 	struct e1000_hw *hw = &adapter->hw;
3558 	u32 vmolr;
3559 
3560 	/* This register exists only on 82576 and newer so if we are older then
3561 	 * we should exit and do nothing
3562 	 */
3563 	if (hw->mac.type < e1000_82576)
3564 		return;
3565 
3566 	vmolr = rd32(E1000_VMOLR(vfn));
3567 	vmolr |= E1000_VMOLR_STRVLAN; /* Strip vlan tags */
3568 	if (hw->mac.type == e1000_i350) {
3569 		u32 dvmolr;
3570 
3571 		dvmolr = rd32(E1000_DVMOLR(vfn));
3572 		dvmolr |= E1000_DVMOLR_STRVLAN;
3573 		wr32(E1000_DVMOLR(vfn), dvmolr);
3574 	}
3575 	if (aupe)
3576 		vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
3577 	else
3578 		vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
3579 
3580 	/* clear all bits that might not be set */
3581 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
3582 
3583 	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
3584 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
3585 	/* for VMDq only allow the VFs and pool 0 to accept broadcast and
3586 	 * multicast packets
3587 	 */
3588 	if (vfn <= adapter->vfs_allocated_count)
3589 		vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
3590 
3591 	wr32(E1000_VMOLR(vfn), vmolr);
3592 }
3593 
3594 /**
3595  *  igb_configure_rx_ring - Configure a receive ring after Reset
3596  *  @adapter: board private structure
3597  *  @ring: receive ring to be configured
3598  *
3599  *  Configure the Rx unit of the MAC after a reset.
3600  **/
3601 void igb_configure_rx_ring(struct igb_adapter *adapter,
3602 			   struct igb_ring *ring)
3603 {
3604 	struct e1000_hw *hw = &adapter->hw;
3605 	u64 rdba = ring->dma;
3606 	int reg_idx = ring->reg_idx;
3607 	u32 srrctl = 0, rxdctl = 0;
3608 
3609 	/* disable the queue */
3610 	wr32(E1000_RXDCTL(reg_idx), 0);
3611 
3612 	/* Set DMA base address registers */
3613 	wr32(E1000_RDBAL(reg_idx),
3614 	     rdba & 0x00000000ffffffffULL);
3615 	wr32(E1000_RDBAH(reg_idx), rdba >> 32);
3616 	wr32(E1000_RDLEN(reg_idx),
3617 	     ring->count * sizeof(union e1000_adv_rx_desc));
3618 
3619 	/* initialize head and tail */
3620 	ring->tail = hw->hw_addr + E1000_RDT(reg_idx);
3621 	wr32(E1000_RDH(reg_idx), 0);
3622 	writel(0, ring->tail);
3623 
3624 	/* set descriptor configuration */
3625 	srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
3626 	srrctl |= IGB_RX_BUFSZ >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3627 	srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3628 	if (hw->mac.type >= e1000_82580)
3629 		srrctl |= E1000_SRRCTL_TIMESTAMP;
3630 	/* Only set Drop Enable if we are supporting multiple queues */
3631 	if (adapter->vfs_allocated_count || adapter->num_rx_queues > 1)
3632 		srrctl |= E1000_SRRCTL_DROP_EN;
3633 
3634 	wr32(E1000_SRRCTL(reg_idx), srrctl);
3635 
3636 	/* set filtering for VMDQ pools */
3637 	igb_set_vmolr(adapter, reg_idx & 0x7, true);
3638 
3639 	rxdctl |= IGB_RX_PTHRESH;
3640 	rxdctl |= IGB_RX_HTHRESH << 8;
3641 	rxdctl |= IGB_RX_WTHRESH << 16;
3642 
3643 	/* enable receive descriptor fetching */
3644 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3645 	wr32(E1000_RXDCTL(reg_idx), rxdctl);
3646 }
3647 
3648 /**
3649  *  igb_configure_rx - Configure receive Unit after Reset
3650  *  @adapter: board private structure
3651  *
3652  *  Configure the Rx unit of the MAC after a reset.
3653  **/
3654 static void igb_configure_rx(struct igb_adapter *adapter)
3655 {
3656 	int i;
3657 
3658 	/* set UTA to appropriate mode */
3659 	igb_set_uta(adapter);
3660 
3661 	/* set the correct pool for the PF default MAC address in entry 0 */
3662 	igb_rar_set_qsel(adapter, adapter->hw.mac.addr, 0,
3663 			 adapter->vfs_allocated_count);
3664 
3665 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3666 	 * the Base and Length of the Rx Descriptor Ring
3667 	 */
3668 	for (i = 0; i < adapter->num_rx_queues; i++)
3669 		igb_configure_rx_ring(adapter, adapter->rx_ring[i]);
3670 }
3671 
3672 /**
3673  *  igb_free_tx_resources - Free Tx Resources per Queue
3674  *  @tx_ring: Tx descriptor ring for a specific queue
3675  *
3676  *  Free all transmit software resources
3677  **/
3678 void igb_free_tx_resources(struct igb_ring *tx_ring)
3679 {
3680 	igb_clean_tx_ring(tx_ring);
3681 
3682 	vfree(tx_ring->tx_buffer_info);
3683 	tx_ring->tx_buffer_info = NULL;
3684 
3685 	/* if not set, then don't free */
3686 	if (!tx_ring->desc)
3687 		return;
3688 
3689 	dma_free_coherent(tx_ring->dev, tx_ring->size,
3690 			  tx_ring->desc, tx_ring->dma);
3691 
3692 	tx_ring->desc = NULL;
3693 }
3694 
3695 /**
3696  *  igb_free_all_tx_resources - Free Tx Resources for All Queues
3697  *  @adapter: board private structure
3698  *
3699  *  Free all transmit software resources
3700  **/
3701 static void igb_free_all_tx_resources(struct igb_adapter *adapter)
3702 {
3703 	int i;
3704 
3705 	for (i = 0; i < adapter->num_tx_queues; i++)
3706 		if (adapter->tx_ring[i])
3707 			igb_free_tx_resources(adapter->tx_ring[i]);
3708 }
3709 
3710 void igb_unmap_and_free_tx_resource(struct igb_ring *ring,
3711 				    struct igb_tx_buffer *tx_buffer)
3712 {
3713 	if (tx_buffer->skb) {
3714 		dev_kfree_skb_any(tx_buffer->skb);
3715 		if (dma_unmap_len(tx_buffer, len))
3716 			dma_unmap_single(ring->dev,
3717 					 dma_unmap_addr(tx_buffer, dma),
3718 					 dma_unmap_len(tx_buffer, len),
3719 					 DMA_TO_DEVICE);
3720 	} else if (dma_unmap_len(tx_buffer, len)) {
3721 		dma_unmap_page(ring->dev,
3722 			       dma_unmap_addr(tx_buffer, dma),
3723 			       dma_unmap_len(tx_buffer, len),
3724 			       DMA_TO_DEVICE);
3725 	}
3726 	tx_buffer->next_to_watch = NULL;
3727 	tx_buffer->skb = NULL;
3728 	dma_unmap_len_set(tx_buffer, len, 0);
3729 	/* buffer_info must be completely set up in the transmit path */
3730 }
3731 
3732 /**
3733  *  igb_clean_tx_ring - Free Tx Buffers
3734  *  @tx_ring: ring to be cleaned
3735  **/
3736 static void igb_clean_tx_ring(struct igb_ring *tx_ring)
3737 {
3738 	struct igb_tx_buffer *buffer_info;
3739 	unsigned long size;
3740 	u16 i;
3741 
3742 	if (!tx_ring->tx_buffer_info)
3743 		return;
3744 	/* Free all the Tx ring sk_buffs */
3745 
3746 	for (i = 0; i < tx_ring->count; i++) {
3747 		buffer_info = &tx_ring->tx_buffer_info[i];
3748 		igb_unmap_and_free_tx_resource(tx_ring, buffer_info);
3749 	}
3750 
3751 	netdev_tx_reset_queue(txring_txq(tx_ring));
3752 
3753 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3754 	memset(tx_ring->tx_buffer_info, 0, size);
3755 
3756 	/* Zero out the descriptor ring */
3757 	memset(tx_ring->desc, 0, tx_ring->size);
3758 
3759 	tx_ring->next_to_use = 0;
3760 	tx_ring->next_to_clean = 0;
3761 }
3762 
3763 /**
3764  *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
3765  *  @adapter: board private structure
3766  **/
3767 static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
3768 {
3769 	int i;
3770 
3771 	for (i = 0; i < adapter->num_tx_queues; i++)
3772 		if (adapter->tx_ring[i])
3773 			igb_clean_tx_ring(adapter->tx_ring[i]);
3774 }
3775 
3776 /**
3777  *  igb_free_rx_resources - Free Rx Resources
3778  *  @rx_ring: ring to clean the resources from
3779  *
3780  *  Free all receive software resources
3781  **/
3782 void igb_free_rx_resources(struct igb_ring *rx_ring)
3783 {
3784 	igb_clean_rx_ring(rx_ring);
3785 
3786 	vfree(rx_ring->rx_buffer_info);
3787 	rx_ring->rx_buffer_info = NULL;
3788 
3789 	/* if not set, then don't free */
3790 	if (!rx_ring->desc)
3791 		return;
3792 
3793 	dma_free_coherent(rx_ring->dev, rx_ring->size,
3794 			  rx_ring->desc, rx_ring->dma);
3795 
3796 	rx_ring->desc = NULL;
3797 }
3798 
3799 /**
3800  *  igb_free_all_rx_resources - Free Rx Resources for All Queues
3801  *  @adapter: board private structure
3802  *
3803  *  Free all receive software resources
3804  **/
3805 static void igb_free_all_rx_resources(struct igb_adapter *adapter)
3806 {
3807 	int i;
3808 
3809 	for (i = 0; i < adapter->num_rx_queues; i++)
3810 		if (adapter->rx_ring[i])
3811 			igb_free_rx_resources(adapter->rx_ring[i]);
3812 }
3813 
3814 /**
3815  *  igb_clean_rx_ring - Free Rx Buffers per Queue
3816  *  @rx_ring: ring to free buffers from
3817  **/
3818 static void igb_clean_rx_ring(struct igb_ring *rx_ring)
3819 {
3820 	unsigned long size;
3821 	u16 i;
3822 
3823 	if (rx_ring->skb)
3824 		dev_kfree_skb(rx_ring->skb);
3825 	rx_ring->skb = NULL;
3826 
3827 	if (!rx_ring->rx_buffer_info)
3828 		return;
3829 
3830 	/* Free all the Rx ring sk_buffs */
3831 	for (i = 0; i < rx_ring->count; i++) {
3832 		struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
3833 
3834 		if (!buffer_info->page)
3835 			continue;
3836 
3837 		dma_unmap_page(rx_ring->dev,
3838 			       buffer_info->dma,
3839 			       PAGE_SIZE,
3840 			       DMA_FROM_DEVICE);
3841 		__free_page(buffer_info->page);
3842 
3843 		buffer_info->page = NULL;
3844 	}
3845 
3846 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3847 	memset(rx_ring->rx_buffer_info, 0, size);
3848 
3849 	/* Zero out the descriptor ring */
3850 	memset(rx_ring->desc, 0, rx_ring->size);
3851 
3852 	rx_ring->next_to_alloc = 0;
3853 	rx_ring->next_to_clean = 0;
3854 	rx_ring->next_to_use = 0;
3855 }
3856 
3857 /**
3858  *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
3859  *  @adapter: board private structure
3860  **/
3861 static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
3862 {
3863 	int i;
3864 
3865 	for (i = 0; i < adapter->num_rx_queues; i++)
3866 		if (adapter->rx_ring[i])
3867 			igb_clean_rx_ring(adapter->rx_ring[i]);
3868 }
3869 
3870 /**
3871  *  igb_set_mac - Change the Ethernet Address of the NIC
3872  *  @netdev: network interface device structure
3873  *  @p: pointer to an address structure
3874  *
3875  *  Returns 0 on success, negative on failure
3876  **/
3877 static int igb_set_mac(struct net_device *netdev, void *p)
3878 {
3879 	struct igb_adapter *adapter = netdev_priv(netdev);
3880 	struct e1000_hw *hw = &adapter->hw;
3881 	struct sockaddr *addr = p;
3882 
3883 	if (!is_valid_ether_addr(addr->sa_data))
3884 		return -EADDRNOTAVAIL;
3885 
3886 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
3887 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
3888 
3889 	/* set the correct pool for the new PF MAC address in entry 0 */
3890 	igb_rar_set_qsel(adapter, hw->mac.addr, 0,
3891 			 adapter->vfs_allocated_count);
3892 
3893 	return 0;
3894 }
3895 
3896 /**
3897  *  igb_write_mc_addr_list - write multicast addresses to MTA
3898  *  @netdev: network interface device structure
3899  *
3900  *  Writes multicast address list to the MTA hash table.
3901  *  Returns: -ENOMEM on failure
3902  *           0 on no addresses written
3903  *           X on writing X addresses to MTA
3904  **/
3905 static int igb_write_mc_addr_list(struct net_device *netdev)
3906 {
3907 	struct igb_adapter *adapter = netdev_priv(netdev);
3908 	struct e1000_hw *hw = &adapter->hw;
3909 	struct netdev_hw_addr *ha;
3910 	u8  *mta_list;
3911 	int i;
3912 
3913 	if (netdev_mc_empty(netdev)) {
3914 		/* nothing to program, so clear mc list */
3915 		igb_update_mc_addr_list(hw, NULL, 0);
3916 		igb_restore_vf_multicasts(adapter);
3917 		return 0;
3918 	}
3919 
3920 	mta_list = kzalloc(netdev_mc_count(netdev) * 6, GFP_ATOMIC);
3921 	if (!mta_list)
3922 		return -ENOMEM;
3923 
3924 	/* The shared function expects a packed array of only addresses. */
3925 	i = 0;
3926 	netdev_for_each_mc_addr(ha, netdev)
3927 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3928 
3929 	igb_update_mc_addr_list(hw, mta_list, i);
3930 	kfree(mta_list);
3931 
3932 	return netdev_mc_count(netdev);
3933 }
3934 
3935 /**
3936  *  igb_write_uc_addr_list - write unicast addresses to RAR table
3937  *  @netdev: network interface device structure
3938  *
3939  *  Writes unicast address list to the RAR table.
3940  *  Returns: -ENOMEM on failure/insufficient address space
3941  *           0 on no addresses written
3942  *           X on writing X addresses to the RAR table
3943  **/
3944 static int igb_write_uc_addr_list(struct net_device *netdev)
3945 {
3946 	struct igb_adapter *adapter = netdev_priv(netdev);
3947 	struct e1000_hw *hw = &adapter->hw;
3948 	unsigned int vfn = adapter->vfs_allocated_count;
3949 	unsigned int rar_entries = hw->mac.rar_entry_count - (vfn + 1);
3950 	int count = 0;
3951 
3952 	/* return ENOMEM indicating insufficient memory for addresses */
3953 	if (netdev_uc_count(netdev) > rar_entries)
3954 		return -ENOMEM;
3955 
3956 	if (!netdev_uc_empty(netdev) && rar_entries) {
3957 		struct netdev_hw_addr *ha;
3958 
3959 		netdev_for_each_uc_addr(ha, netdev) {
3960 			if (!rar_entries)
3961 				break;
3962 			igb_rar_set_qsel(adapter, ha->addr,
3963 					 rar_entries--,
3964 					 vfn);
3965 			count++;
3966 		}
3967 	}
3968 	/* write the addresses in reverse order to avoid write combining */
3969 	for (; rar_entries > 0 ; rar_entries--) {
3970 		wr32(E1000_RAH(rar_entries), 0);
3971 		wr32(E1000_RAL(rar_entries), 0);
3972 	}
3973 	wrfl();
3974 
3975 	return count;
3976 }
3977 
3978 /**
3979  *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
3980  *  @netdev: network interface device structure
3981  *
3982  *  The set_rx_mode entry point is called whenever the unicast or multicast
3983  *  address lists or the network interface flags are updated.  This routine is
3984  *  responsible for configuring the hardware for proper unicast, multicast,
3985  *  promiscuous mode, and all-multi behavior.
3986  **/
3987 static void igb_set_rx_mode(struct net_device *netdev)
3988 {
3989 	struct igb_adapter *adapter = netdev_priv(netdev);
3990 	struct e1000_hw *hw = &adapter->hw;
3991 	unsigned int vfn = adapter->vfs_allocated_count;
3992 	u32 rctl, vmolr = 0;
3993 	int count;
3994 
3995 	/* Check for Promiscuous and All Multicast modes */
3996 	rctl = rd32(E1000_RCTL);
3997 
3998 	/* clear the effected bits */
3999 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE | E1000_RCTL_VFE);
4000 
4001 	if (netdev->flags & IFF_PROMISC) {
4002 		/* retain VLAN HW filtering if in VT mode */
4003 		if (adapter->vfs_allocated_count)
4004 			rctl |= E1000_RCTL_VFE;
4005 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
4006 		vmolr |= (E1000_VMOLR_ROPE | E1000_VMOLR_MPME);
4007 	} else {
4008 		if (netdev->flags & IFF_ALLMULTI) {
4009 			rctl |= E1000_RCTL_MPE;
4010 			vmolr |= E1000_VMOLR_MPME;
4011 		} else {
4012 			/* Write addresses to the MTA, if the attempt fails
4013 			 * then we should just turn on promiscuous mode so
4014 			 * that we can at least receive multicast traffic
4015 			 */
4016 			count = igb_write_mc_addr_list(netdev);
4017 			if (count < 0) {
4018 				rctl |= E1000_RCTL_MPE;
4019 				vmolr |= E1000_VMOLR_MPME;
4020 			} else if (count) {
4021 				vmolr |= E1000_VMOLR_ROMPE;
4022 			}
4023 		}
4024 		/* Write addresses to available RAR registers, if there is not
4025 		 * sufficient space to store all the addresses then enable
4026 		 * unicast promiscuous mode
4027 		 */
4028 		count = igb_write_uc_addr_list(netdev);
4029 		if (count < 0) {
4030 			rctl |= E1000_RCTL_UPE;
4031 			vmolr |= E1000_VMOLR_ROPE;
4032 		}
4033 		rctl |= E1000_RCTL_VFE;
4034 	}
4035 	wr32(E1000_RCTL, rctl);
4036 
4037 	/* In order to support SR-IOV and eventually VMDq it is necessary to set
4038 	 * the VMOLR to enable the appropriate modes.  Without this workaround
4039 	 * we will have issues with VLAN tag stripping not being done for frames
4040 	 * that are only arriving because we are the default pool
4041 	 */
4042 	if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
4043 		return;
4044 
4045 	vmolr |= rd32(E1000_VMOLR(vfn)) &
4046 		 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
4047 	wr32(E1000_VMOLR(vfn), vmolr);
4048 	igb_restore_vf_multicasts(adapter);
4049 }
4050 
4051 static void igb_check_wvbr(struct igb_adapter *adapter)
4052 {
4053 	struct e1000_hw *hw = &adapter->hw;
4054 	u32 wvbr = 0;
4055 
4056 	switch (hw->mac.type) {
4057 	case e1000_82576:
4058 	case e1000_i350:
4059 		wvbr = rd32(E1000_WVBR);
4060 		if (!wvbr)
4061 			return;
4062 		break;
4063 	default:
4064 		break;
4065 	}
4066 
4067 	adapter->wvbr |= wvbr;
4068 }
4069 
4070 #define IGB_STAGGERED_QUEUE_OFFSET 8
4071 
4072 static void igb_spoof_check(struct igb_adapter *adapter)
4073 {
4074 	int j;
4075 
4076 	if (!adapter->wvbr)
4077 		return;
4078 
4079 	for (j = 0; j < adapter->vfs_allocated_count; j++) {
4080 		if (adapter->wvbr & (1 << j) ||
4081 		    adapter->wvbr & (1 << (j + IGB_STAGGERED_QUEUE_OFFSET))) {
4082 			dev_warn(&adapter->pdev->dev,
4083 				"Spoof event(s) detected on VF %d\n", j);
4084 			adapter->wvbr &=
4085 				~((1 << j) |
4086 				  (1 << (j + IGB_STAGGERED_QUEUE_OFFSET)));
4087 		}
4088 	}
4089 }
4090 
4091 /* Need to wait a few seconds after link up to get diagnostic information from
4092  * the phy
4093  */
4094 static void igb_update_phy_info(unsigned long data)
4095 {
4096 	struct igb_adapter *adapter = (struct igb_adapter *) data;
4097 	igb_get_phy_info(&adapter->hw);
4098 }
4099 
4100 /**
4101  *  igb_has_link - check shared code for link and determine up/down
4102  *  @adapter: pointer to driver private info
4103  **/
4104 bool igb_has_link(struct igb_adapter *adapter)
4105 {
4106 	struct e1000_hw *hw = &adapter->hw;
4107 	bool link_active = false;
4108 
4109 	/* get_link_status is set on LSC (link status) interrupt or
4110 	 * rx sequence error interrupt.  get_link_status will stay
4111 	 * false until the e1000_check_for_link establishes link
4112 	 * for copper adapters ONLY
4113 	 */
4114 	switch (hw->phy.media_type) {
4115 	case e1000_media_type_copper:
4116 		if (!hw->mac.get_link_status)
4117 			return true;
4118 	case e1000_media_type_internal_serdes:
4119 		hw->mac.ops.check_for_link(hw);
4120 		link_active = !hw->mac.get_link_status;
4121 		break;
4122 	default:
4123 	case e1000_media_type_unknown:
4124 		break;
4125 	}
4126 
4127 	if (((hw->mac.type == e1000_i210) ||
4128 	     (hw->mac.type == e1000_i211)) &&
4129 	     (hw->phy.id == I210_I_PHY_ID)) {
4130 		if (!netif_carrier_ok(adapter->netdev)) {
4131 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4132 		} else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
4133 			adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
4134 			adapter->link_check_timeout = jiffies;
4135 		}
4136 	}
4137 
4138 	return link_active;
4139 }
4140 
4141 static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
4142 {
4143 	bool ret = false;
4144 	u32 ctrl_ext, thstat;
4145 
4146 	/* check for thermal sensor event on i350 copper only */
4147 	if (hw->mac.type == e1000_i350) {
4148 		thstat = rd32(E1000_THSTAT);
4149 		ctrl_ext = rd32(E1000_CTRL_EXT);
4150 
4151 		if ((hw->phy.media_type == e1000_media_type_copper) &&
4152 		    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
4153 			ret = !!(thstat & event);
4154 	}
4155 
4156 	return ret;
4157 }
4158 
4159 /**
4160  *  igb_check_lvmmc - check for malformed packets received
4161  *  and indicated in LVMMC register
4162  *  @adapter: pointer to adapter
4163  **/
4164 static void igb_check_lvmmc(struct igb_adapter *adapter)
4165 {
4166 	struct e1000_hw *hw = &adapter->hw;
4167 	u32 lvmmc;
4168 
4169 	lvmmc = rd32(E1000_LVMMC);
4170 	if (lvmmc) {
4171 		if (unlikely(net_ratelimit())) {
4172 			netdev_warn(adapter->netdev,
4173 				    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
4174 				    lvmmc);
4175 		}
4176 	}
4177 }
4178 
4179 /**
4180  *  igb_watchdog - Timer Call-back
4181  *  @data: pointer to adapter cast into an unsigned long
4182  **/
4183 static void igb_watchdog(unsigned long data)
4184 {
4185 	struct igb_adapter *adapter = (struct igb_adapter *)data;
4186 	/* Do the rest outside of interrupt context */
4187 	schedule_work(&adapter->watchdog_task);
4188 }
4189 
4190 static void igb_watchdog_task(struct work_struct *work)
4191 {
4192 	struct igb_adapter *adapter = container_of(work,
4193 						   struct igb_adapter,
4194 						   watchdog_task);
4195 	struct e1000_hw *hw = &adapter->hw;
4196 	struct e1000_phy_info *phy = &hw->phy;
4197 	struct net_device *netdev = adapter->netdev;
4198 	u32 link;
4199 	int i;
4200 	u32 connsw;
4201 
4202 	link = igb_has_link(adapter);
4203 
4204 	if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
4205 		if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
4206 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4207 		else
4208 			link = false;
4209 	}
4210 
4211 	/* Force link down if we have fiber to swap to */
4212 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4213 		if (hw->phy.media_type == e1000_media_type_copper) {
4214 			connsw = rd32(E1000_CONNSW);
4215 			if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
4216 				link = 0;
4217 		}
4218 	}
4219 	if (link) {
4220 		/* Perform a reset if the media type changed. */
4221 		if (hw->dev_spec._82575.media_changed) {
4222 			hw->dev_spec._82575.media_changed = false;
4223 			adapter->flags |= IGB_FLAG_MEDIA_RESET;
4224 			igb_reset(adapter);
4225 		}
4226 		/* Cancel scheduled suspend requests. */
4227 		pm_runtime_resume(netdev->dev.parent);
4228 
4229 		if (!netif_carrier_ok(netdev)) {
4230 			u32 ctrl;
4231 
4232 			hw->mac.ops.get_speed_and_duplex(hw,
4233 							 &adapter->link_speed,
4234 							 &adapter->link_duplex);
4235 
4236 			ctrl = rd32(E1000_CTRL);
4237 			/* Links status message must follow this format */
4238 			netdev_info(netdev,
4239 			       "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4240 			       netdev->name,
4241 			       adapter->link_speed,
4242 			       adapter->link_duplex == FULL_DUPLEX ?
4243 			       "Full" : "Half",
4244 			       (ctrl & E1000_CTRL_TFCE) &&
4245 			       (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
4246 			       (ctrl & E1000_CTRL_RFCE) ?  "RX" :
4247 			       (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
4248 
4249 			/* disable EEE if enabled */
4250 			if ((adapter->flags & IGB_FLAG_EEE) &&
4251 				(adapter->link_duplex == HALF_DUPLEX)) {
4252 				dev_info(&adapter->pdev->dev,
4253 				"EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
4254 				adapter->hw.dev_spec._82575.eee_disable = true;
4255 				adapter->flags &= ~IGB_FLAG_EEE;
4256 			}
4257 
4258 			/* check if SmartSpeed worked */
4259 			igb_check_downshift(hw);
4260 			if (phy->speed_downgraded)
4261 				netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
4262 
4263 			/* check for thermal sensor event */
4264 			if (igb_thermal_sensor_event(hw,
4265 			    E1000_THSTAT_LINK_THROTTLE))
4266 				netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
4267 
4268 			/* adjust timeout factor according to speed/duplex */
4269 			adapter->tx_timeout_factor = 1;
4270 			switch (adapter->link_speed) {
4271 			case SPEED_10:
4272 				adapter->tx_timeout_factor = 14;
4273 				break;
4274 			case SPEED_100:
4275 				/* maybe add some timeout factor ? */
4276 				break;
4277 			}
4278 
4279 			netif_carrier_on(netdev);
4280 
4281 			igb_ping_all_vfs(adapter);
4282 			igb_check_vf_rate_limit(adapter);
4283 
4284 			/* link state has changed, schedule phy info update */
4285 			if (!test_bit(__IGB_DOWN, &adapter->state))
4286 				mod_timer(&adapter->phy_info_timer,
4287 					  round_jiffies(jiffies + 2 * HZ));
4288 		}
4289 	} else {
4290 		if (netif_carrier_ok(netdev)) {
4291 			adapter->link_speed = 0;
4292 			adapter->link_duplex = 0;
4293 
4294 			/* check for thermal sensor event */
4295 			if (igb_thermal_sensor_event(hw,
4296 			    E1000_THSTAT_PWR_DOWN)) {
4297 				netdev_err(netdev, "The network adapter was stopped because it overheated\n");
4298 			}
4299 
4300 			/* Links status message must follow this format */
4301 			netdev_info(netdev, "igb: %s NIC Link is Down\n",
4302 			       netdev->name);
4303 			netif_carrier_off(netdev);
4304 
4305 			igb_ping_all_vfs(adapter);
4306 
4307 			/* link state has changed, schedule phy info update */
4308 			if (!test_bit(__IGB_DOWN, &adapter->state))
4309 				mod_timer(&adapter->phy_info_timer,
4310 					  round_jiffies(jiffies + 2 * HZ));
4311 
4312 			/* link is down, time to check for alternate media */
4313 			if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4314 				igb_check_swap_media(adapter);
4315 				if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
4316 					schedule_work(&adapter->reset_task);
4317 					/* return immediately */
4318 					return;
4319 				}
4320 			}
4321 			pm_schedule_suspend(netdev->dev.parent,
4322 					    MSEC_PER_SEC * 5);
4323 
4324 		/* also check for alternate media here */
4325 		} else if (!netif_carrier_ok(netdev) &&
4326 			   (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
4327 			igb_check_swap_media(adapter);
4328 			if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
4329 				schedule_work(&adapter->reset_task);
4330 				/* return immediately */
4331 				return;
4332 			}
4333 		}
4334 	}
4335 
4336 	spin_lock(&adapter->stats64_lock);
4337 	igb_update_stats(adapter, &adapter->stats64);
4338 	spin_unlock(&adapter->stats64_lock);
4339 
4340 	for (i = 0; i < adapter->num_tx_queues; i++) {
4341 		struct igb_ring *tx_ring = adapter->tx_ring[i];
4342 		if (!netif_carrier_ok(netdev)) {
4343 			/* We've lost link, so the controller stops DMA,
4344 			 * but we've got queued Tx work that's never going
4345 			 * to get done, so reset controller to flush Tx.
4346 			 * (Do the reset outside of interrupt context).
4347 			 */
4348 			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
4349 				adapter->tx_timeout_count++;
4350 				schedule_work(&adapter->reset_task);
4351 				/* return immediately since reset is imminent */
4352 				return;
4353 			}
4354 		}
4355 
4356 		/* Force detection of hung controller every watchdog period */
4357 		set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
4358 	}
4359 
4360 	/* Cause software interrupt to ensure Rx ring is cleaned */
4361 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
4362 		u32 eics = 0;
4363 
4364 		for (i = 0; i < adapter->num_q_vectors; i++)
4365 			eics |= adapter->q_vector[i]->eims_value;
4366 		wr32(E1000_EICS, eics);
4367 	} else {
4368 		wr32(E1000_ICS, E1000_ICS_RXDMT0);
4369 	}
4370 
4371 	igb_spoof_check(adapter);
4372 	igb_ptp_rx_hang(adapter);
4373 
4374 	/* Check LVMMC register on i350/i354 only */
4375 	if ((adapter->hw.mac.type == e1000_i350) ||
4376 	    (adapter->hw.mac.type == e1000_i354))
4377 		igb_check_lvmmc(adapter);
4378 
4379 	/* Reset the timer */
4380 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
4381 		if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
4382 			mod_timer(&adapter->watchdog_timer,
4383 				  round_jiffies(jiffies +  HZ));
4384 		else
4385 			mod_timer(&adapter->watchdog_timer,
4386 				  round_jiffies(jiffies + 2 * HZ));
4387 	}
4388 }
4389 
4390 enum latency_range {
4391 	lowest_latency = 0,
4392 	low_latency = 1,
4393 	bulk_latency = 2,
4394 	latency_invalid = 255
4395 };
4396 
4397 /**
4398  *  igb_update_ring_itr - update the dynamic ITR value based on packet size
4399  *  @q_vector: pointer to q_vector
4400  *
4401  *  Stores a new ITR value based on strictly on packet size.  This
4402  *  algorithm is less sophisticated than that used in igb_update_itr,
4403  *  due to the difficulty of synchronizing statistics across multiple
4404  *  receive rings.  The divisors and thresholds used by this function
4405  *  were determined based on theoretical maximum wire speed and testing
4406  *  data, in order to minimize response time while increasing bulk
4407  *  throughput.
4408  *  This functionality is controlled by ethtool's coalescing settings.
4409  *  NOTE:  This function is called only when operating in a multiqueue
4410  *         receive environment.
4411  **/
4412 static void igb_update_ring_itr(struct igb_q_vector *q_vector)
4413 {
4414 	int new_val = q_vector->itr_val;
4415 	int avg_wire_size = 0;
4416 	struct igb_adapter *adapter = q_vector->adapter;
4417 	unsigned int packets;
4418 
4419 	/* For non-gigabit speeds, just fix the interrupt rate at 4000
4420 	 * ints/sec - ITR timer value of 120 ticks.
4421 	 */
4422 	if (adapter->link_speed != SPEED_1000) {
4423 		new_val = IGB_4K_ITR;
4424 		goto set_itr_val;
4425 	}
4426 
4427 	packets = q_vector->rx.total_packets;
4428 	if (packets)
4429 		avg_wire_size = q_vector->rx.total_bytes / packets;
4430 
4431 	packets = q_vector->tx.total_packets;
4432 	if (packets)
4433 		avg_wire_size = max_t(u32, avg_wire_size,
4434 				      q_vector->tx.total_bytes / packets);
4435 
4436 	/* if avg_wire_size isn't set no work was done */
4437 	if (!avg_wire_size)
4438 		goto clear_counts;
4439 
4440 	/* Add 24 bytes to size to account for CRC, preamble, and gap */
4441 	avg_wire_size += 24;
4442 
4443 	/* Don't starve jumbo frames */
4444 	avg_wire_size = min(avg_wire_size, 3000);
4445 
4446 	/* Give a little boost to mid-size frames */
4447 	if ((avg_wire_size > 300) && (avg_wire_size < 1200))
4448 		new_val = avg_wire_size / 3;
4449 	else
4450 		new_val = avg_wire_size / 2;
4451 
4452 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
4453 	if (new_val < IGB_20K_ITR &&
4454 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
4455 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
4456 		new_val = IGB_20K_ITR;
4457 
4458 set_itr_val:
4459 	if (new_val != q_vector->itr_val) {
4460 		q_vector->itr_val = new_val;
4461 		q_vector->set_itr = 1;
4462 	}
4463 clear_counts:
4464 	q_vector->rx.total_bytes = 0;
4465 	q_vector->rx.total_packets = 0;
4466 	q_vector->tx.total_bytes = 0;
4467 	q_vector->tx.total_packets = 0;
4468 }
4469 
4470 /**
4471  *  igb_update_itr - update the dynamic ITR value based on statistics
4472  *  @q_vector: pointer to q_vector
4473  *  @ring_container: ring info to update the itr for
4474  *
4475  *  Stores a new ITR value based on packets and byte
4476  *  counts during the last interrupt.  The advantage of per interrupt
4477  *  computation is faster updates and more accurate ITR for the current
4478  *  traffic pattern.  Constants in this function were computed
4479  *  based on theoretical maximum wire speed and thresholds were set based
4480  *  on testing data as well as attempting to minimize response time
4481  *  while increasing bulk throughput.
4482  *  This functionality is controlled by ethtool's coalescing settings.
4483  *  NOTE:  These calculations are only valid when operating in a single-
4484  *         queue environment.
4485  **/
4486 static void igb_update_itr(struct igb_q_vector *q_vector,
4487 			   struct igb_ring_container *ring_container)
4488 {
4489 	unsigned int packets = ring_container->total_packets;
4490 	unsigned int bytes = ring_container->total_bytes;
4491 	u8 itrval = ring_container->itr;
4492 
4493 	/* no packets, exit with status unchanged */
4494 	if (packets == 0)
4495 		return;
4496 
4497 	switch (itrval) {
4498 	case lowest_latency:
4499 		/* handle TSO and jumbo frames */
4500 		if (bytes/packets > 8000)
4501 			itrval = bulk_latency;
4502 		else if ((packets < 5) && (bytes > 512))
4503 			itrval = low_latency;
4504 		break;
4505 	case low_latency:  /* 50 usec aka 20000 ints/s */
4506 		if (bytes > 10000) {
4507 			/* this if handles the TSO accounting */
4508 			if (bytes/packets > 8000)
4509 				itrval = bulk_latency;
4510 			else if ((packets < 10) || ((bytes/packets) > 1200))
4511 				itrval = bulk_latency;
4512 			else if ((packets > 35))
4513 				itrval = lowest_latency;
4514 		} else if (bytes/packets > 2000) {
4515 			itrval = bulk_latency;
4516 		} else if (packets <= 2 && bytes < 512) {
4517 			itrval = lowest_latency;
4518 		}
4519 		break;
4520 	case bulk_latency: /* 250 usec aka 4000 ints/s */
4521 		if (bytes > 25000) {
4522 			if (packets > 35)
4523 				itrval = low_latency;
4524 		} else if (bytes < 1500) {
4525 			itrval = low_latency;
4526 		}
4527 		break;
4528 	}
4529 
4530 	/* clear work counters since we have the values we need */
4531 	ring_container->total_bytes = 0;
4532 	ring_container->total_packets = 0;
4533 
4534 	/* write updated itr to ring container */
4535 	ring_container->itr = itrval;
4536 }
4537 
4538 static void igb_set_itr(struct igb_q_vector *q_vector)
4539 {
4540 	struct igb_adapter *adapter = q_vector->adapter;
4541 	u32 new_itr = q_vector->itr_val;
4542 	u8 current_itr = 0;
4543 
4544 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
4545 	if (adapter->link_speed != SPEED_1000) {
4546 		current_itr = 0;
4547 		new_itr = IGB_4K_ITR;
4548 		goto set_itr_now;
4549 	}
4550 
4551 	igb_update_itr(q_vector, &q_vector->tx);
4552 	igb_update_itr(q_vector, &q_vector->rx);
4553 
4554 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
4555 
4556 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
4557 	if (current_itr == lowest_latency &&
4558 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
4559 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
4560 		current_itr = low_latency;
4561 
4562 	switch (current_itr) {
4563 	/* counts and packets in update_itr are dependent on these numbers */
4564 	case lowest_latency:
4565 		new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
4566 		break;
4567 	case low_latency:
4568 		new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
4569 		break;
4570 	case bulk_latency:
4571 		new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
4572 		break;
4573 	default:
4574 		break;
4575 	}
4576 
4577 set_itr_now:
4578 	if (new_itr != q_vector->itr_val) {
4579 		/* this attempts to bias the interrupt rate towards Bulk
4580 		 * by adding intermediate steps when interrupt rate is
4581 		 * increasing
4582 		 */
4583 		new_itr = new_itr > q_vector->itr_val ?
4584 			  max((new_itr * q_vector->itr_val) /
4585 			  (new_itr + (q_vector->itr_val >> 2)),
4586 			  new_itr) : new_itr;
4587 		/* Don't write the value here; it resets the adapter's
4588 		 * internal timer, and causes us to delay far longer than
4589 		 * we should between interrupts.  Instead, we write the ITR
4590 		 * value at the beginning of the next interrupt so the timing
4591 		 * ends up being correct.
4592 		 */
4593 		q_vector->itr_val = new_itr;
4594 		q_vector->set_itr = 1;
4595 	}
4596 }
4597 
4598 static void igb_tx_ctxtdesc(struct igb_ring *tx_ring, u32 vlan_macip_lens,
4599 			    u32 type_tucmd, u32 mss_l4len_idx)
4600 {
4601 	struct e1000_adv_tx_context_desc *context_desc;
4602 	u16 i = tx_ring->next_to_use;
4603 
4604 	context_desc = IGB_TX_CTXTDESC(tx_ring, i);
4605 
4606 	i++;
4607 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
4608 
4609 	/* set bits to identify this as an advanced context descriptor */
4610 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
4611 
4612 	/* For 82575, context index must be unique per ring. */
4613 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
4614 		mss_l4len_idx |= tx_ring->reg_idx << 4;
4615 
4616 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
4617 	context_desc->seqnum_seed	= 0;
4618 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
4619 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
4620 }
4621 
4622 static int igb_tso(struct igb_ring *tx_ring,
4623 		   struct igb_tx_buffer *first,
4624 		   u8 *hdr_len)
4625 {
4626 	struct sk_buff *skb = first->skb;
4627 	u32 vlan_macip_lens, type_tucmd;
4628 	u32 mss_l4len_idx, l4len;
4629 	int err;
4630 
4631 	if (skb->ip_summed != CHECKSUM_PARTIAL)
4632 		return 0;
4633 
4634 	if (!skb_is_gso(skb))
4635 		return 0;
4636 
4637 	err = skb_cow_head(skb, 0);
4638 	if (err < 0)
4639 		return err;
4640 
4641 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
4642 	type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
4643 
4644 	if (first->protocol == htons(ETH_P_IP)) {
4645 		struct iphdr *iph = ip_hdr(skb);
4646 		iph->tot_len = 0;
4647 		iph->check = 0;
4648 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
4649 							 iph->daddr, 0,
4650 							 IPPROTO_TCP,
4651 							 0);
4652 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
4653 		first->tx_flags |= IGB_TX_FLAGS_TSO |
4654 				   IGB_TX_FLAGS_CSUM |
4655 				   IGB_TX_FLAGS_IPV4;
4656 	} else if (skb_is_gso_v6(skb)) {
4657 		ipv6_hdr(skb)->payload_len = 0;
4658 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
4659 						       &ipv6_hdr(skb)->daddr,
4660 						       0, IPPROTO_TCP, 0);
4661 		first->tx_flags |= IGB_TX_FLAGS_TSO |
4662 				   IGB_TX_FLAGS_CSUM;
4663 	}
4664 
4665 	/* compute header lengths */
4666 	l4len = tcp_hdrlen(skb);
4667 	*hdr_len = skb_transport_offset(skb) + l4len;
4668 
4669 	/* update gso size and bytecount with header size */
4670 	first->gso_segs = skb_shinfo(skb)->gso_segs;
4671 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
4672 
4673 	/* MSS L4LEN IDX */
4674 	mss_l4len_idx = l4len << E1000_ADVTXD_L4LEN_SHIFT;
4675 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
4676 
4677 	/* VLAN MACLEN IPLEN */
4678 	vlan_macip_lens = skb_network_header_len(skb);
4679 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
4680 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
4681 
4682 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
4683 
4684 	return 1;
4685 }
4686 
4687 static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
4688 {
4689 	struct sk_buff *skb = first->skb;
4690 	u32 vlan_macip_lens = 0;
4691 	u32 mss_l4len_idx = 0;
4692 	u32 type_tucmd = 0;
4693 
4694 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
4695 		if (!(first->tx_flags & IGB_TX_FLAGS_VLAN))
4696 			return;
4697 	} else {
4698 		u8 l4_hdr = 0;
4699 
4700 		switch (first->protocol) {
4701 		case htons(ETH_P_IP):
4702 			vlan_macip_lens |= skb_network_header_len(skb);
4703 			type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
4704 			l4_hdr = ip_hdr(skb)->protocol;
4705 			break;
4706 		case htons(ETH_P_IPV6):
4707 			vlan_macip_lens |= skb_network_header_len(skb);
4708 			l4_hdr = ipv6_hdr(skb)->nexthdr;
4709 			break;
4710 		default:
4711 			if (unlikely(net_ratelimit())) {
4712 				dev_warn(tx_ring->dev,
4713 					 "partial checksum but proto=%x!\n",
4714 					 first->protocol);
4715 			}
4716 			break;
4717 		}
4718 
4719 		switch (l4_hdr) {
4720 		case IPPROTO_TCP:
4721 			type_tucmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
4722 			mss_l4len_idx = tcp_hdrlen(skb) <<
4723 					E1000_ADVTXD_L4LEN_SHIFT;
4724 			break;
4725 		case IPPROTO_SCTP:
4726 			type_tucmd |= E1000_ADVTXD_TUCMD_L4T_SCTP;
4727 			mss_l4len_idx = sizeof(struct sctphdr) <<
4728 					E1000_ADVTXD_L4LEN_SHIFT;
4729 			break;
4730 		case IPPROTO_UDP:
4731 			mss_l4len_idx = sizeof(struct udphdr) <<
4732 					E1000_ADVTXD_L4LEN_SHIFT;
4733 			break;
4734 		default:
4735 			if (unlikely(net_ratelimit())) {
4736 				dev_warn(tx_ring->dev,
4737 					 "partial checksum but l4 proto=%x!\n",
4738 					 l4_hdr);
4739 			}
4740 			break;
4741 		}
4742 
4743 		/* update TX checksum flag */
4744 		first->tx_flags |= IGB_TX_FLAGS_CSUM;
4745 	}
4746 
4747 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
4748 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
4749 
4750 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
4751 }
4752 
4753 #define IGB_SET_FLAG(_input, _flag, _result) \
4754 	((_flag <= _result) ? \
4755 	 ((u32)(_input & _flag) * (_result / _flag)) : \
4756 	 ((u32)(_input & _flag) / (_flag / _result)))
4757 
4758 static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
4759 {
4760 	/* set type for advanced descriptor with frame checksum insertion */
4761 	u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
4762 		       E1000_ADVTXD_DCMD_DEXT |
4763 		       E1000_ADVTXD_DCMD_IFCS;
4764 
4765 	/* set HW vlan bit if vlan is present */
4766 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
4767 				 (E1000_ADVTXD_DCMD_VLE));
4768 
4769 	/* set segmentation bits for TSO */
4770 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
4771 				 (E1000_ADVTXD_DCMD_TSE));
4772 
4773 	/* set timestamp bit if present */
4774 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
4775 				 (E1000_ADVTXD_MAC_TSTAMP));
4776 
4777 	/* insert frame checksum */
4778 	cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
4779 
4780 	return cmd_type;
4781 }
4782 
4783 static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
4784 				 union e1000_adv_tx_desc *tx_desc,
4785 				 u32 tx_flags, unsigned int paylen)
4786 {
4787 	u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
4788 
4789 	/* 82575 requires a unique index per ring */
4790 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
4791 		olinfo_status |= tx_ring->reg_idx << 4;
4792 
4793 	/* insert L4 checksum */
4794 	olinfo_status |= IGB_SET_FLAG(tx_flags,
4795 				      IGB_TX_FLAGS_CSUM,
4796 				      (E1000_TXD_POPTS_TXSM << 8));
4797 
4798 	/* insert IPv4 checksum */
4799 	olinfo_status |= IGB_SET_FLAG(tx_flags,
4800 				      IGB_TX_FLAGS_IPV4,
4801 				      (E1000_TXD_POPTS_IXSM << 8));
4802 
4803 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
4804 }
4805 
4806 static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
4807 {
4808 	struct net_device *netdev = tx_ring->netdev;
4809 
4810 	netif_stop_subqueue(netdev, tx_ring->queue_index);
4811 
4812 	/* Herbert's original patch had:
4813 	 *  smp_mb__after_netif_stop_queue();
4814 	 * but since that doesn't exist yet, just open code it.
4815 	 */
4816 	smp_mb();
4817 
4818 	/* We need to check again in a case another CPU has just
4819 	 * made room available.
4820 	 */
4821 	if (igb_desc_unused(tx_ring) < size)
4822 		return -EBUSY;
4823 
4824 	/* A reprieve! */
4825 	netif_wake_subqueue(netdev, tx_ring->queue_index);
4826 
4827 	u64_stats_update_begin(&tx_ring->tx_syncp2);
4828 	tx_ring->tx_stats.restart_queue2++;
4829 	u64_stats_update_end(&tx_ring->tx_syncp2);
4830 
4831 	return 0;
4832 }
4833 
4834 static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
4835 {
4836 	if (igb_desc_unused(tx_ring) >= size)
4837 		return 0;
4838 	return __igb_maybe_stop_tx(tx_ring, size);
4839 }
4840 
4841 static void igb_tx_map(struct igb_ring *tx_ring,
4842 		       struct igb_tx_buffer *first,
4843 		       const u8 hdr_len)
4844 {
4845 	struct sk_buff *skb = first->skb;
4846 	struct igb_tx_buffer *tx_buffer;
4847 	union e1000_adv_tx_desc *tx_desc;
4848 	struct skb_frag_struct *frag;
4849 	dma_addr_t dma;
4850 	unsigned int data_len, size;
4851 	u32 tx_flags = first->tx_flags;
4852 	u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
4853 	u16 i = tx_ring->next_to_use;
4854 
4855 	tx_desc = IGB_TX_DESC(tx_ring, i);
4856 
4857 	igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
4858 
4859 	size = skb_headlen(skb);
4860 	data_len = skb->data_len;
4861 
4862 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
4863 
4864 	tx_buffer = first;
4865 
4866 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
4867 		if (dma_mapping_error(tx_ring->dev, dma))
4868 			goto dma_error;
4869 
4870 		/* record length, and DMA address */
4871 		dma_unmap_len_set(tx_buffer, len, size);
4872 		dma_unmap_addr_set(tx_buffer, dma, dma);
4873 
4874 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
4875 
4876 		while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
4877 			tx_desc->read.cmd_type_len =
4878 				cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
4879 
4880 			i++;
4881 			tx_desc++;
4882 			if (i == tx_ring->count) {
4883 				tx_desc = IGB_TX_DESC(tx_ring, 0);
4884 				i = 0;
4885 			}
4886 			tx_desc->read.olinfo_status = 0;
4887 
4888 			dma += IGB_MAX_DATA_PER_TXD;
4889 			size -= IGB_MAX_DATA_PER_TXD;
4890 
4891 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
4892 		}
4893 
4894 		if (likely(!data_len))
4895 			break;
4896 
4897 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
4898 
4899 		i++;
4900 		tx_desc++;
4901 		if (i == tx_ring->count) {
4902 			tx_desc = IGB_TX_DESC(tx_ring, 0);
4903 			i = 0;
4904 		}
4905 		tx_desc->read.olinfo_status = 0;
4906 
4907 		size = skb_frag_size(frag);
4908 		data_len -= size;
4909 
4910 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
4911 				       size, DMA_TO_DEVICE);
4912 
4913 		tx_buffer = &tx_ring->tx_buffer_info[i];
4914 	}
4915 
4916 	/* write last descriptor with RS and EOP bits */
4917 	cmd_type |= size | IGB_TXD_DCMD;
4918 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
4919 
4920 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
4921 
4922 	/* set the timestamp */
4923 	first->time_stamp = jiffies;
4924 
4925 	/* Force memory writes to complete before letting h/w know there
4926 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
4927 	 * memory model archs, such as IA-64).
4928 	 *
4929 	 * We also need this memory barrier to make certain all of the
4930 	 * status bits have been updated before next_to_watch is written.
4931 	 */
4932 	wmb();
4933 
4934 	/* set next_to_watch value indicating a packet is present */
4935 	first->next_to_watch = tx_desc;
4936 
4937 	i++;
4938 	if (i == tx_ring->count)
4939 		i = 0;
4940 
4941 	tx_ring->next_to_use = i;
4942 
4943 	/* Make sure there is space in the ring for the next send. */
4944 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
4945 
4946 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
4947 		writel(i, tx_ring->tail);
4948 
4949 		/* we need this if more than one processor can write to our tail
4950 		 * at a time, it synchronizes IO on IA64/Altix systems
4951 		 */
4952 		mmiowb();
4953 	}
4954 	return;
4955 
4956 dma_error:
4957 	dev_err(tx_ring->dev, "TX DMA map failed\n");
4958 
4959 	/* clear dma mappings for failed tx_buffer_info map */
4960 	for (;;) {
4961 		tx_buffer = &tx_ring->tx_buffer_info[i];
4962 		igb_unmap_and_free_tx_resource(tx_ring, tx_buffer);
4963 		if (tx_buffer == first)
4964 			break;
4965 		if (i == 0)
4966 			i = tx_ring->count;
4967 		i--;
4968 	}
4969 
4970 	tx_ring->next_to_use = i;
4971 }
4972 
4973 netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
4974 				struct igb_ring *tx_ring)
4975 {
4976 	struct igb_tx_buffer *first;
4977 	int tso;
4978 	u32 tx_flags = 0;
4979 	unsigned short f;
4980 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
4981 	__be16 protocol = vlan_get_protocol(skb);
4982 	u8 hdr_len = 0;
4983 
4984 	/* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
4985 	 *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
4986 	 *       + 2 desc gap to keep tail from touching head,
4987 	 *       + 1 desc for context descriptor,
4988 	 * otherwise try next time
4989 	 */
4990 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
4991 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
4992 
4993 	if (igb_maybe_stop_tx(tx_ring, count + 3)) {
4994 		/* this is a hard error */
4995 		return NETDEV_TX_BUSY;
4996 	}
4997 
4998 	/* record the location of the first descriptor for this packet */
4999 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
5000 	first->skb = skb;
5001 	first->bytecount = skb->len;
5002 	first->gso_segs = 1;
5003 
5004 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
5005 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
5006 
5007 		if (!test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
5008 					   &adapter->state)) {
5009 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5010 			tx_flags |= IGB_TX_FLAGS_TSTAMP;
5011 
5012 			adapter->ptp_tx_skb = skb_get(skb);
5013 			adapter->ptp_tx_start = jiffies;
5014 			if (adapter->hw.mac.type == e1000_82576)
5015 				schedule_work(&adapter->ptp_tx_work);
5016 		}
5017 	}
5018 
5019 	skb_tx_timestamp(skb);
5020 
5021 	if (skb_vlan_tag_present(skb)) {
5022 		tx_flags |= IGB_TX_FLAGS_VLAN;
5023 		tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
5024 	}
5025 
5026 	/* record initial flags and protocol */
5027 	first->tx_flags = tx_flags;
5028 	first->protocol = protocol;
5029 
5030 	tso = igb_tso(tx_ring, first, &hdr_len);
5031 	if (tso < 0)
5032 		goto out_drop;
5033 	else if (!tso)
5034 		igb_tx_csum(tx_ring, first);
5035 
5036 	igb_tx_map(tx_ring, first, hdr_len);
5037 
5038 	return NETDEV_TX_OK;
5039 
5040 out_drop:
5041 	igb_unmap_and_free_tx_resource(tx_ring, first);
5042 
5043 	return NETDEV_TX_OK;
5044 }
5045 
5046 static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
5047 						    struct sk_buff *skb)
5048 {
5049 	unsigned int r_idx = skb->queue_mapping;
5050 
5051 	if (r_idx >= adapter->num_tx_queues)
5052 		r_idx = r_idx % adapter->num_tx_queues;
5053 
5054 	return adapter->tx_ring[r_idx];
5055 }
5056 
5057 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
5058 				  struct net_device *netdev)
5059 {
5060 	struct igb_adapter *adapter = netdev_priv(netdev);
5061 
5062 	if (test_bit(__IGB_DOWN, &adapter->state)) {
5063 		dev_kfree_skb_any(skb);
5064 		return NETDEV_TX_OK;
5065 	}
5066 
5067 	if (skb->len <= 0) {
5068 		dev_kfree_skb_any(skb);
5069 		return NETDEV_TX_OK;
5070 	}
5071 
5072 	/* The minimum packet size with TCTL.PSP set is 17 so pad the skb
5073 	 * in order to meet this minimum size requirement.
5074 	 */
5075 	if (skb_put_padto(skb, 17))
5076 		return NETDEV_TX_OK;
5077 
5078 	return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
5079 }
5080 
5081 /**
5082  *  igb_tx_timeout - Respond to a Tx Hang
5083  *  @netdev: network interface device structure
5084  **/
5085 static void igb_tx_timeout(struct net_device *netdev)
5086 {
5087 	struct igb_adapter *adapter = netdev_priv(netdev);
5088 	struct e1000_hw *hw = &adapter->hw;
5089 
5090 	/* Do the reset outside of interrupt context */
5091 	adapter->tx_timeout_count++;
5092 
5093 	if (hw->mac.type >= e1000_82580)
5094 		hw->dev_spec._82575.global_device_reset = true;
5095 
5096 	schedule_work(&adapter->reset_task);
5097 	wr32(E1000_EICS,
5098 	     (adapter->eims_enable_mask & ~adapter->eims_other));
5099 }
5100 
5101 static void igb_reset_task(struct work_struct *work)
5102 {
5103 	struct igb_adapter *adapter;
5104 	adapter = container_of(work, struct igb_adapter, reset_task);
5105 
5106 	igb_dump(adapter);
5107 	netdev_err(adapter->netdev, "Reset adapter\n");
5108 	igb_reinit_locked(adapter);
5109 }
5110 
5111 /**
5112  *  igb_get_stats64 - Get System Network Statistics
5113  *  @netdev: network interface device structure
5114  *  @stats: rtnl_link_stats64 pointer
5115  **/
5116 static struct rtnl_link_stats64 *igb_get_stats64(struct net_device *netdev,
5117 						struct rtnl_link_stats64 *stats)
5118 {
5119 	struct igb_adapter *adapter = netdev_priv(netdev);
5120 
5121 	spin_lock(&adapter->stats64_lock);
5122 	igb_update_stats(adapter, &adapter->stats64);
5123 	memcpy(stats, &adapter->stats64, sizeof(*stats));
5124 	spin_unlock(&adapter->stats64_lock);
5125 
5126 	return stats;
5127 }
5128 
5129 /**
5130  *  igb_change_mtu - Change the Maximum Transfer Unit
5131  *  @netdev: network interface device structure
5132  *  @new_mtu: new value for maximum frame size
5133  *
5134  *  Returns 0 on success, negative on failure
5135  **/
5136 static int igb_change_mtu(struct net_device *netdev, int new_mtu)
5137 {
5138 	struct igb_adapter *adapter = netdev_priv(netdev);
5139 	struct pci_dev *pdev = adapter->pdev;
5140 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
5141 
5142 	if ((new_mtu < 68) || (max_frame > MAX_JUMBO_FRAME_SIZE)) {
5143 		dev_err(&pdev->dev, "Invalid MTU setting\n");
5144 		return -EINVAL;
5145 	}
5146 
5147 #define MAX_STD_JUMBO_FRAME_SIZE 9238
5148 	if (max_frame > MAX_STD_JUMBO_FRAME_SIZE) {
5149 		dev_err(&pdev->dev, "MTU > 9216 not supported.\n");
5150 		return -EINVAL;
5151 	}
5152 
5153 	/* adjust max frame to be at least the size of a standard frame */
5154 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
5155 		max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
5156 
5157 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
5158 		usleep_range(1000, 2000);
5159 
5160 	/* igb_down has a dependency on max_frame_size */
5161 	adapter->max_frame_size = max_frame;
5162 
5163 	if (netif_running(netdev))
5164 		igb_down(adapter);
5165 
5166 	dev_info(&pdev->dev, "changing MTU from %d to %d\n",
5167 		 netdev->mtu, new_mtu);
5168 	netdev->mtu = new_mtu;
5169 
5170 	if (netif_running(netdev))
5171 		igb_up(adapter);
5172 	else
5173 		igb_reset(adapter);
5174 
5175 	clear_bit(__IGB_RESETTING, &adapter->state);
5176 
5177 	return 0;
5178 }
5179 
5180 /**
5181  *  igb_update_stats - Update the board statistics counters
5182  *  @adapter: board private structure
5183  **/
5184 void igb_update_stats(struct igb_adapter *adapter,
5185 		      struct rtnl_link_stats64 *net_stats)
5186 {
5187 	struct e1000_hw *hw = &adapter->hw;
5188 	struct pci_dev *pdev = adapter->pdev;
5189 	u32 reg, mpc;
5190 	int i;
5191 	u64 bytes, packets;
5192 	unsigned int start;
5193 	u64 _bytes, _packets;
5194 
5195 	/* Prevent stats update while adapter is being reset, or if the pci
5196 	 * connection is down.
5197 	 */
5198 	if (adapter->link_speed == 0)
5199 		return;
5200 	if (pci_channel_offline(pdev))
5201 		return;
5202 
5203 	bytes = 0;
5204 	packets = 0;
5205 
5206 	rcu_read_lock();
5207 	for (i = 0; i < adapter->num_rx_queues; i++) {
5208 		struct igb_ring *ring = adapter->rx_ring[i];
5209 		u32 rqdpc = rd32(E1000_RQDPC(i));
5210 		if (hw->mac.type >= e1000_i210)
5211 			wr32(E1000_RQDPC(i), 0);
5212 
5213 		if (rqdpc) {
5214 			ring->rx_stats.drops += rqdpc;
5215 			net_stats->rx_fifo_errors += rqdpc;
5216 		}
5217 
5218 		do {
5219 			start = u64_stats_fetch_begin_irq(&ring->rx_syncp);
5220 			_bytes = ring->rx_stats.bytes;
5221 			_packets = ring->rx_stats.packets;
5222 		} while (u64_stats_fetch_retry_irq(&ring->rx_syncp, start));
5223 		bytes += _bytes;
5224 		packets += _packets;
5225 	}
5226 
5227 	net_stats->rx_bytes = bytes;
5228 	net_stats->rx_packets = packets;
5229 
5230 	bytes = 0;
5231 	packets = 0;
5232 	for (i = 0; i < adapter->num_tx_queues; i++) {
5233 		struct igb_ring *ring = adapter->tx_ring[i];
5234 		do {
5235 			start = u64_stats_fetch_begin_irq(&ring->tx_syncp);
5236 			_bytes = ring->tx_stats.bytes;
5237 			_packets = ring->tx_stats.packets;
5238 		} while (u64_stats_fetch_retry_irq(&ring->tx_syncp, start));
5239 		bytes += _bytes;
5240 		packets += _packets;
5241 	}
5242 	net_stats->tx_bytes = bytes;
5243 	net_stats->tx_packets = packets;
5244 	rcu_read_unlock();
5245 
5246 	/* read stats registers */
5247 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
5248 	adapter->stats.gprc += rd32(E1000_GPRC);
5249 	adapter->stats.gorc += rd32(E1000_GORCL);
5250 	rd32(E1000_GORCH); /* clear GORCL */
5251 	adapter->stats.bprc += rd32(E1000_BPRC);
5252 	adapter->stats.mprc += rd32(E1000_MPRC);
5253 	adapter->stats.roc += rd32(E1000_ROC);
5254 
5255 	adapter->stats.prc64 += rd32(E1000_PRC64);
5256 	adapter->stats.prc127 += rd32(E1000_PRC127);
5257 	adapter->stats.prc255 += rd32(E1000_PRC255);
5258 	adapter->stats.prc511 += rd32(E1000_PRC511);
5259 	adapter->stats.prc1023 += rd32(E1000_PRC1023);
5260 	adapter->stats.prc1522 += rd32(E1000_PRC1522);
5261 	adapter->stats.symerrs += rd32(E1000_SYMERRS);
5262 	adapter->stats.sec += rd32(E1000_SEC);
5263 
5264 	mpc = rd32(E1000_MPC);
5265 	adapter->stats.mpc += mpc;
5266 	net_stats->rx_fifo_errors += mpc;
5267 	adapter->stats.scc += rd32(E1000_SCC);
5268 	adapter->stats.ecol += rd32(E1000_ECOL);
5269 	adapter->stats.mcc += rd32(E1000_MCC);
5270 	adapter->stats.latecol += rd32(E1000_LATECOL);
5271 	adapter->stats.dc += rd32(E1000_DC);
5272 	adapter->stats.rlec += rd32(E1000_RLEC);
5273 	adapter->stats.xonrxc += rd32(E1000_XONRXC);
5274 	adapter->stats.xontxc += rd32(E1000_XONTXC);
5275 	adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
5276 	adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
5277 	adapter->stats.fcruc += rd32(E1000_FCRUC);
5278 	adapter->stats.gptc += rd32(E1000_GPTC);
5279 	adapter->stats.gotc += rd32(E1000_GOTCL);
5280 	rd32(E1000_GOTCH); /* clear GOTCL */
5281 	adapter->stats.rnbc += rd32(E1000_RNBC);
5282 	adapter->stats.ruc += rd32(E1000_RUC);
5283 	adapter->stats.rfc += rd32(E1000_RFC);
5284 	adapter->stats.rjc += rd32(E1000_RJC);
5285 	adapter->stats.tor += rd32(E1000_TORH);
5286 	adapter->stats.tot += rd32(E1000_TOTH);
5287 	adapter->stats.tpr += rd32(E1000_TPR);
5288 
5289 	adapter->stats.ptc64 += rd32(E1000_PTC64);
5290 	adapter->stats.ptc127 += rd32(E1000_PTC127);
5291 	adapter->stats.ptc255 += rd32(E1000_PTC255);
5292 	adapter->stats.ptc511 += rd32(E1000_PTC511);
5293 	adapter->stats.ptc1023 += rd32(E1000_PTC1023);
5294 	adapter->stats.ptc1522 += rd32(E1000_PTC1522);
5295 
5296 	adapter->stats.mptc += rd32(E1000_MPTC);
5297 	adapter->stats.bptc += rd32(E1000_BPTC);
5298 
5299 	adapter->stats.tpt += rd32(E1000_TPT);
5300 	adapter->stats.colc += rd32(E1000_COLC);
5301 
5302 	adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
5303 	/* read internal phy specific stats */
5304 	reg = rd32(E1000_CTRL_EXT);
5305 	if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
5306 		adapter->stats.rxerrc += rd32(E1000_RXERRC);
5307 
5308 		/* this stat has invalid values on i210/i211 */
5309 		if ((hw->mac.type != e1000_i210) &&
5310 		    (hw->mac.type != e1000_i211))
5311 			adapter->stats.tncrs += rd32(E1000_TNCRS);
5312 	}
5313 
5314 	adapter->stats.tsctc += rd32(E1000_TSCTC);
5315 	adapter->stats.tsctfc += rd32(E1000_TSCTFC);
5316 
5317 	adapter->stats.iac += rd32(E1000_IAC);
5318 	adapter->stats.icrxoc += rd32(E1000_ICRXOC);
5319 	adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
5320 	adapter->stats.icrxatc += rd32(E1000_ICRXATC);
5321 	adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
5322 	adapter->stats.ictxatc += rd32(E1000_ICTXATC);
5323 	adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
5324 	adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
5325 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
5326 
5327 	/* Fill out the OS statistics structure */
5328 	net_stats->multicast = adapter->stats.mprc;
5329 	net_stats->collisions = adapter->stats.colc;
5330 
5331 	/* Rx Errors */
5332 
5333 	/* RLEC on some newer hardware can be incorrect so build
5334 	 * our own version based on RUC and ROC
5335 	 */
5336 	net_stats->rx_errors = adapter->stats.rxerrc +
5337 		adapter->stats.crcerrs + adapter->stats.algnerrc +
5338 		adapter->stats.ruc + adapter->stats.roc +
5339 		adapter->stats.cexterr;
5340 	net_stats->rx_length_errors = adapter->stats.ruc +
5341 				      adapter->stats.roc;
5342 	net_stats->rx_crc_errors = adapter->stats.crcerrs;
5343 	net_stats->rx_frame_errors = adapter->stats.algnerrc;
5344 	net_stats->rx_missed_errors = adapter->stats.mpc;
5345 
5346 	/* Tx Errors */
5347 	net_stats->tx_errors = adapter->stats.ecol +
5348 			       adapter->stats.latecol;
5349 	net_stats->tx_aborted_errors = adapter->stats.ecol;
5350 	net_stats->tx_window_errors = adapter->stats.latecol;
5351 	net_stats->tx_carrier_errors = adapter->stats.tncrs;
5352 
5353 	/* Tx Dropped needs to be maintained elsewhere */
5354 
5355 	/* Management Stats */
5356 	adapter->stats.mgptc += rd32(E1000_MGTPTC);
5357 	adapter->stats.mgprc += rd32(E1000_MGTPRC);
5358 	adapter->stats.mgpdc += rd32(E1000_MGTPDC);
5359 
5360 	/* OS2BMC Stats */
5361 	reg = rd32(E1000_MANC);
5362 	if (reg & E1000_MANC_EN_BMC2OS) {
5363 		adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
5364 		adapter->stats.o2bspc += rd32(E1000_O2BSPC);
5365 		adapter->stats.b2ospc += rd32(E1000_B2OSPC);
5366 		adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
5367 	}
5368 }
5369 
5370 static void igb_tsync_interrupt(struct igb_adapter *adapter)
5371 {
5372 	struct e1000_hw *hw = &adapter->hw;
5373 	struct ptp_clock_event event;
5374 	struct timespec ts;
5375 	u32 ack = 0, tsauxc, sec, nsec, tsicr = rd32(E1000_TSICR);
5376 
5377 	if (tsicr & TSINTR_SYS_WRAP) {
5378 		event.type = PTP_CLOCK_PPS;
5379 		if (adapter->ptp_caps.pps)
5380 			ptp_clock_event(adapter->ptp_clock, &event);
5381 		else
5382 			dev_err(&adapter->pdev->dev, "unexpected SYS WRAP");
5383 		ack |= TSINTR_SYS_WRAP;
5384 	}
5385 
5386 	if (tsicr & E1000_TSICR_TXTS) {
5387 		/* retrieve hardware timestamp */
5388 		schedule_work(&adapter->ptp_tx_work);
5389 		ack |= E1000_TSICR_TXTS;
5390 	}
5391 
5392 	if (tsicr & TSINTR_TT0) {
5393 		spin_lock(&adapter->tmreg_lock);
5394 		ts = timespec_add(adapter->perout[0].start,
5395 				  adapter->perout[0].period);
5396 		wr32(E1000_TRGTTIML0, ts.tv_nsec);
5397 		wr32(E1000_TRGTTIMH0, ts.tv_sec);
5398 		tsauxc = rd32(E1000_TSAUXC);
5399 		tsauxc |= TSAUXC_EN_TT0;
5400 		wr32(E1000_TSAUXC, tsauxc);
5401 		adapter->perout[0].start = ts;
5402 		spin_unlock(&adapter->tmreg_lock);
5403 		ack |= TSINTR_TT0;
5404 	}
5405 
5406 	if (tsicr & TSINTR_TT1) {
5407 		spin_lock(&adapter->tmreg_lock);
5408 		ts = timespec_add(adapter->perout[1].start,
5409 				  adapter->perout[1].period);
5410 		wr32(E1000_TRGTTIML1, ts.tv_nsec);
5411 		wr32(E1000_TRGTTIMH1, ts.tv_sec);
5412 		tsauxc = rd32(E1000_TSAUXC);
5413 		tsauxc |= TSAUXC_EN_TT1;
5414 		wr32(E1000_TSAUXC, tsauxc);
5415 		adapter->perout[1].start = ts;
5416 		spin_unlock(&adapter->tmreg_lock);
5417 		ack |= TSINTR_TT1;
5418 	}
5419 
5420 	if (tsicr & TSINTR_AUTT0) {
5421 		nsec = rd32(E1000_AUXSTMPL0);
5422 		sec  = rd32(E1000_AUXSTMPH0);
5423 		event.type = PTP_CLOCK_EXTTS;
5424 		event.index = 0;
5425 		event.timestamp = sec * 1000000000ULL + nsec;
5426 		ptp_clock_event(adapter->ptp_clock, &event);
5427 		ack |= TSINTR_AUTT0;
5428 	}
5429 
5430 	if (tsicr & TSINTR_AUTT1) {
5431 		nsec = rd32(E1000_AUXSTMPL1);
5432 		sec  = rd32(E1000_AUXSTMPH1);
5433 		event.type = PTP_CLOCK_EXTTS;
5434 		event.index = 1;
5435 		event.timestamp = sec * 1000000000ULL + nsec;
5436 		ptp_clock_event(adapter->ptp_clock, &event);
5437 		ack |= TSINTR_AUTT1;
5438 	}
5439 
5440 	/* acknowledge the interrupts */
5441 	wr32(E1000_TSICR, ack);
5442 }
5443 
5444 static irqreturn_t igb_msix_other(int irq, void *data)
5445 {
5446 	struct igb_adapter *adapter = data;
5447 	struct e1000_hw *hw = &adapter->hw;
5448 	u32 icr = rd32(E1000_ICR);
5449 	/* reading ICR causes bit 31 of EICR to be cleared */
5450 
5451 	if (icr & E1000_ICR_DRSTA)
5452 		schedule_work(&adapter->reset_task);
5453 
5454 	if (icr & E1000_ICR_DOUTSYNC) {
5455 		/* HW is reporting DMA is out of sync */
5456 		adapter->stats.doosync++;
5457 		/* The DMA Out of Sync is also indication of a spoof event
5458 		 * in IOV mode. Check the Wrong VM Behavior register to
5459 		 * see if it is really a spoof event.
5460 		 */
5461 		igb_check_wvbr(adapter);
5462 	}
5463 
5464 	/* Check for a mailbox event */
5465 	if (icr & E1000_ICR_VMMB)
5466 		igb_msg_task(adapter);
5467 
5468 	if (icr & E1000_ICR_LSC) {
5469 		hw->mac.get_link_status = 1;
5470 		/* guard against interrupt when we're going down */
5471 		if (!test_bit(__IGB_DOWN, &adapter->state))
5472 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
5473 	}
5474 
5475 	if (icr & E1000_ICR_TS)
5476 		igb_tsync_interrupt(adapter);
5477 
5478 	wr32(E1000_EIMS, adapter->eims_other);
5479 
5480 	return IRQ_HANDLED;
5481 }
5482 
5483 static void igb_write_itr(struct igb_q_vector *q_vector)
5484 {
5485 	struct igb_adapter *adapter = q_vector->adapter;
5486 	u32 itr_val = q_vector->itr_val & 0x7FFC;
5487 
5488 	if (!q_vector->set_itr)
5489 		return;
5490 
5491 	if (!itr_val)
5492 		itr_val = 0x4;
5493 
5494 	if (adapter->hw.mac.type == e1000_82575)
5495 		itr_val |= itr_val << 16;
5496 	else
5497 		itr_val |= E1000_EITR_CNT_IGNR;
5498 
5499 	writel(itr_val, q_vector->itr_register);
5500 	q_vector->set_itr = 0;
5501 }
5502 
5503 static irqreturn_t igb_msix_ring(int irq, void *data)
5504 {
5505 	struct igb_q_vector *q_vector = data;
5506 
5507 	/* Write the ITR value calculated from the previous interrupt. */
5508 	igb_write_itr(q_vector);
5509 
5510 	napi_schedule(&q_vector->napi);
5511 
5512 	return IRQ_HANDLED;
5513 }
5514 
5515 #ifdef CONFIG_IGB_DCA
5516 static void igb_update_tx_dca(struct igb_adapter *adapter,
5517 			      struct igb_ring *tx_ring,
5518 			      int cpu)
5519 {
5520 	struct e1000_hw *hw = &adapter->hw;
5521 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
5522 
5523 	if (hw->mac.type != e1000_82575)
5524 		txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
5525 
5526 	/* We can enable relaxed ordering for reads, but not writes when
5527 	 * DCA is enabled.  This is due to a known issue in some chipsets
5528 	 * which will cause the DCA tag to be cleared.
5529 	 */
5530 	txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
5531 		  E1000_DCA_TXCTRL_DATA_RRO_EN |
5532 		  E1000_DCA_TXCTRL_DESC_DCA_EN;
5533 
5534 	wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
5535 }
5536 
5537 static void igb_update_rx_dca(struct igb_adapter *adapter,
5538 			      struct igb_ring *rx_ring,
5539 			      int cpu)
5540 {
5541 	struct e1000_hw *hw = &adapter->hw;
5542 	u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
5543 
5544 	if (hw->mac.type != e1000_82575)
5545 		rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
5546 
5547 	/* We can enable relaxed ordering for reads, but not writes when
5548 	 * DCA is enabled.  This is due to a known issue in some chipsets
5549 	 * which will cause the DCA tag to be cleared.
5550 	 */
5551 	rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
5552 		  E1000_DCA_RXCTRL_DESC_DCA_EN;
5553 
5554 	wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
5555 }
5556 
5557 static void igb_update_dca(struct igb_q_vector *q_vector)
5558 {
5559 	struct igb_adapter *adapter = q_vector->adapter;
5560 	int cpu = get_cpu();
5561 
5562 	if (q_vector->cpu == cpu)
5563 		goto out_no_update;
5564 
5565 	if (q_vector->tx.ring)
5566 		igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
5567 
5568 	if (q_vector->rx.ring)
5569 		igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
5570 
5571 	q_vector->cpu = cpu;
5572 out_no_update:
5573 	put_cpu();
5574 }
5575 
5576 static void igb_setup_dca(struct igb_adapter *adapter)
5577 {
5578 	struct e1000_hw *hw = &adapter->hw;
5579 	int i;
5580 
5581 	if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
5582 		return;
5583 
5584 	/* Always use CB2 mode, difference is masked in the CB driver. */
5585 	wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
5586 
5587 	for (i = 0; i < adapter->num_q_vectors; i++) {
5588 		adapter->q_vector[i]->cpu = -1;
5589 		igb_update_dca(adapter->q_vector[i]);
5590 	}
5591 }
5592 
5593 static int __igb_notify_dca(struct device *dev, void *data)
5594 {
5595 	struct net_device *netdev = dev_get_drvdata(dev);
5596 	struct igb_adapter *adapter = netdev_priv(netdev);
5597 	struct pci_dev *pdev = adapter->pdev;
5598 	struct e1000_hw *hw = &adapter->hw;
5599 	unsigned long event = *(unsigned long *)data;
5600 
5601 	switch (event) {
5602 	case DCA_PROVIDER_ADD:
5603 		/* if already enabled, don't do it again */
5604 		if (adapter->flags & IGB_FLAG_DCA_ENABLED)
5605 			break;
5606 		if (dca_add_requester(dev) == 0) {
5607 			adapter->flags |= IGB_FLAG_DCA_ENABLED;
5608 			dev_info(&pdev->dev, "DCA enabled\n");
5609 			igb_setup_dca(adapter);
5610 			break;
5611 		}
5612 		/* Fall Through since DCA is disabled. */
5613 	case DCA_PROVIDER_REMOVE:
5614 		if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
5615 			/* without this a class_device is left
5616 			 * hanging around in the sysfs model
5617 			 */
5618 			dca_remove_requester(dev);
5619 			dev_info(&pdev->dev, "DCA disabled\n");
5620 			adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
5621 			wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
5622 		}
5623 		break;
5624 	}
5625 
5626 	return 0;
5627 }
5628 
5629 static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
5630 			  void *p)
5631 {
5632 	int ret_val;
5633 
5634 	ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
5635 					 __igb_notify_dca);
5636 
5637 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
5638 }
5639 #endif /* CONFIG_IGB_DCA */
5640 
5641 #ifdef CONFIG_PCI_IOV
5642 static int igb_vf_configure(struct igb_adapter *adapter, int vf)
5643 {
5644 	unsigned char mac_addr[ETH_ALEN];
5645 
5646 	eth_zero_addr(mac_addr);
5647 	igb_set_vf_mac(adapter, vf, mac_addr);
5648 
5649 	/* By default spoof check is enabled for all VFs */
5650 	adapter->vf_data[vf].spoofchk_enabled = true;
5651 
5652 	return 0;
5653 }
5654 
5655 #endif
5656 static void igb_ping_all_vfs(struct igb_adapter *adapter)
5657 {
5658 	struct e1000_hw *hw = &adapter->hw;
5659 	u32 ping;
5660 	int i;
5661 
5662 	for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
5663 		ping = E1000_PF_CONTROL_MSG;
5664 		if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
5665 			ping |= E1000_VT_MSGTYPE_CTS;
5666 		igb_write_mbx(hw, &ping, 1, i);
5667 	}
5668 }
5669 
5670 static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
5671 {
5672 	struct e1000_hw *hw = &adapter->hw;
5673 	u32 vmolr = rd32(E1000_VMOLR(vf));
5674 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
5675 
5676 	vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
5677 			    IGB_VF_FLAG_MULTI_PROMISC);
5678 	vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
5679 
5680 	if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
5681 		vmolr |= E1000_VMOLR_MPME;
5682 		vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
5683 		*msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
5684 	} else {
5685 		/* if we have hashes and we are clearing a multicast promisc
5686 		 * flag we need to write the hashes to the MTA as this step
5687 		 * was previously skipped
5688 		 */
5689 		if (vf_data->num_vf_mc_hashes > 30) {
5690 			vmolr |= E1000_VMOLR_MPME;
5691 		} else if (vf_data->num_vf_mc_hashes) {
5692 			int j;
5693 
5694 			vmolr |= E1000_VMOLR_ROMPE;
5695 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
5696 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
5697 		}
5698 	}
5699 
5700 	wr32(E1000_VMOLR(vf), vmolr);
5701 
5702 	/* there are flags left unprocessed, likely not supported */
5703 	if (*msgbuf & E1000_VT_MSGINFO_MASK)
5704 		return -EINVAL;
5705 
5706 	return 0;
5707 }
5708 
5709 static int igb_set_vf_multicasts(struct igb_adapter *adapter,
5710 				  u32 *msgbuf, u32 vf)
5711 {
5712 	int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
5713 	u16 *hash_list = (u16 *)&msgbuf[1];
5714 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
5715 	int i;
5716 
5717 	/* salt away the number of multicast addresses assigned
5718 	 * to this VF for later use to restore when the PF multi cast
5719 	 * list changes
5720 	 */
5721 	vf_data->num_vf_mc_hashes = n;
5722 
5723 	/* only up to 30 hash values supported */
5724 	if (n > 30)
5725 		n = 30;
5726 
5727 	/* store the hashes for later use */
5728 	for (i = 0; i < n; i++)
5729 		vf_data->vf_mc_hashes[i] = hash_list[i];
5730 
5731 	/* Flush and reset the mta with the new values */
5732 	igb_set_rx_mode(adapter->netdev);
5733 
5734 	return 0;
5735 }
5736 
5737 static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
5738 {
5739 	struct e1000_hw *hw = &adapter->hw;
5740 	struct vf_data_storage *vf_data;
5741 	int i, j;
5742 
5743 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
5744 		u32 vmolr = rd32(E1000_VMOLR(i));
5745 
5746 		vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
5747 
5748 		vf_data = &adapter->vf_data[i];
5749 
5750 		if ((vf_data->num_vf_mc_hashes > 30) ||
5751 		    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
5752 			vmolr |= E1000_VMOLR_MPME;
5753 		} else if (vf_data->num_vf_mc_hashes) {
5754 			vmolr |= E1000_VMOLR_ROMPE;
5755 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
5756 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
5757 		}
5758 		wr32(E1000_VMOLR(i), vmolr);
5759 	}
5760 }
5761 
5762 static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
5763 {
5764 	struct e1000_hw *hw = &adapter->hw;
5765 	u32 pool_mask, reg, vid;
5766 	int i;
5767 
5768 	pool_mask = 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
5769 
5770 	/* Find the vlan filter for this id */
5771 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5772 		reg = rd32(E1000_VLVF(i));
5773 
5774 		/* remove the vf from the pool */
5775 		reg &= ~pool_mask;
5776 
5777 		/* if pool is empty then remove entry from vfta */
5778 		if (!(reg & E1000_VLVF_POOLSEL_MASK) &&
5779 		    (reg & E1000_VLVF_VLANID_ENABLE)) {
5780 			reg = 0;
5781 			vid = reg & E1000_VLVF_VLANID_MASK;
5782 			igb_vfta_set(hw, vid, false);
5783 		}
5784 
5785 		wr32(E1000_VLVF(i), reg);
5786 	}
5787 
5788 	adapter->vf_data[vf].vlans_enabled = 0;
5789 }
5790 
5791 static s32 igb_vlvf_set(struct igb_adapter *adapter, u32 vid, bool add, u32 vf)
5792 {
5793 	struct e1000_hw *hw = &adapter->hw;
5794 	u32 reg, i;
5795 
5796 	/* The vlvf table only exists on 82576 hardware and newer */
5797 	if (hw->mac.type < e1000_82576)
5798 		return -1;
5799 
5800 	/* we only need to do this if VMDq is enabled */
5801 	if (!adapter->vfs_allocated_count)
5802 		return -1;
5803 
5804 	/* Find the vlan filter for this id */
5805 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5806 		reg = rd32(E1000_VLVF(i));
5807 		if ((reg & E1000_VLVF_VLANID_ENABLE) &&
5808 		    vid == (reg & E1000_VLVF_VLANID_MASK))
5809 			break;
5810 	}
5811 
5812 	if (add) {
5813 		if (i == E1000_VLVF_ARRAY_SIZE) {
5814 			/* Did not find a matching VLAN ID entry that was
5815 			 * enabled.  Search for a free filter entry, i.e.
5816 			 * one without the enable bit set
5817 			 */
5818 			for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5819 				reg = rd32(E1000_VLVF(i));
5820 				if (!(reg & E1000_VLVF_VLANID_ENABLE))
5821 					break;
5822 			}
5823 		}
5824 		if (i < E1000_VLVF_ARRAY_SIZE) {
5825 			/* Found an enabled/available entry */
5826 			reg |= 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
5827 
5828 			/* if !enabled we need to set this up in vfta */
5829 			if (!(reg & E1000_VLVF_VLANID_ENABLE)) {
5830 				/* add VID to filter table */
5831 				igb_vfta_set(hw, vid, true);
5832 				reg |= E1000_VLVF_VLANID_ENABLE;
5833 			}
5834 			reg &= ~E1000_VLVF_VLANID_MASK;
5835 			reg |= vid;
5836 			wr32(E1000_VLVF(i), reg);
5837 
5838 			/* do not modify RLPML for PF devices */
5839 			if (vf >= adapter->vfs_allocated_count)
5840 				return 0;
5841 
5842 			if (!adapter->vf_data[vf].vlans_enabled) {
5843 				u32 size;
5844 
5845 				reg = rd32(E1000_VMOLR(vf));
5846 				size = reg & E1000_VMOLR_RLPML_MASK;
5847 				size += 4;
5848 				reg &= ~E1000_VMOLR_RLPML_MASK;
5849 				reg |= size;
5850 				wr32(E1000_VMOLR(vf), reg);
5851 			}
5852 
5853 			adapter->vf_data[vf].vlans_enabled++;
5854 		}
5855 	} else {
5856 		if (i < E1000_VLVF_ARRAY_SIZE) {
5857 			/* remove vf from the pool */
5858 			reg &= ~(1 << (E1000_VLVF_POOLSEL_SHIFT + vf));
5859 			/* if pool is empty then remove entry from vfta */
5860 			if (!(reg & E1000_VLVF_POOLSEL_MASK)) {
5861 				reg = 0;
5862 				igb_vfta_set(hw, vid, false);
5863 			}
5864 			wr32(E1000_VLVF(i), reg);
5865 
5866 			/* do not modify RLPML for PF devices */
5867 			if (vf >= adapter->vfs_allocated_count)
5868 				return 0;
5869 
5870 			adapter->vf_data[vf].vlans_enabled--;
5871 			if (!adapter->vf_data[vf].vlans_enabled) {
5872 				u32 size;
5873 
5874 				reg = rd32(E1000_VMOLR(vf));
5875 				size = reg & E1000_VMOLR_RLPML_MASK;
5876 				size -= 4;
5877 				reg &= ~E1000_VMOLR_RLPML_MASK;
5878 				reg |= size;
5879 				wr32(E1000_VMOLR(vf), reg);
5880 			}
5881 		}
5882 	}
5883 	return 0;
5884 }
5885 
5886 static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
5887 {
5888 	struct e1000_hw *hw = &adapter->hw;
5889 
5890 	if (vid)
5891 		wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
5892 	else
5893 		wr32(E1000_VMVIR(vf), 0);
5894 }
5895 
5896 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
5897 			       int vf, u16 vlan, u8 qos)
5898 {
5899 	int err = 0;
5900 	struct igb_adapter *adapter = netdev_priv(netdev);
5901 
5902 	if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
5903 		return -EINVAL;
5904 	if (vlan || qos) {
5905 		err = igb_vlvf_set(adapter, vlan, !!vlan, vf);
5906 		if (err)
5907 			goto out;
5908 		igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
5909 		igb_set_vmolr(adapter, vf, !vlan);
5910 		adapter->vf_data[vf].pf_vlan = vlan;
5911 		adapter->vf_data[vf].pf_qos = qos;
5912 		dev_info(&adapter->pdev->dev,
5913 			 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
5914 		if (test_bit(__IGB_DOWN, &adapter->state)) {
5915 			dev_warn(&adapter->pdev->dev,
5916 				 "The VF VLAN has been set, but the PF device is not up.\n");
5917 			dev_warn(&adapter->pdev->dev,
5918 				 "Bring the PF device up before attempting to use the VF device.\n");
5919 		}
5920 	} else {
5921 		igb_vlvf_set(adapter, adapter->vf_data[vf].pf_vlan,
5922 			     false, vf);
5923 		igb_set_vmvir(adapter, vlan, vf);
5924 		igb_set_vmolr(adapter, vf, true);
5925 		adapter->vf_data[vf].pf_vlan = 0;
5926 		adapter->vf_data[vf].pf_qos = 0;
5927 	}
5928 out:
5929 	return err;
5930 }
5931 
5932 static int igb_find_vlvf_entry(struct igb_adapter *adapter, int vid)
5933 {
5934 	struct e1000_hw *hw = &adapter->hw;
5935 	int i;
5936 	u32 reg;
5937 
5938 	/* Find the vlan filter for this id */
5939 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5940 		reg = rd32(E1000_VLVF(i));
5941 		if ((reg & E1000_VLVF_VLANID_ENABLE) &&
5942 		    vid == (reg & E1000_VLVF_VLANID_MASK))
5943 			break;
5944 	}
5945 
5946 	if (i >= E1000_VLVF_ARRAY_SIZE)
5947 		i = -1;
5948 
5949 	return i;
5950 }
5951 
5952 static int igb_set_vf_vlan(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
5953 {
5954 	struct e1000_hw *hw = &adapter->hw;
5955 	int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
5956 	int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
5957 	int err = 0;
5958 
5959 	/* If in promiscuous mode we need to make sure the PF also has
5960 	 * the VLAN filter set.
5961 	 */
5962 	if (add && (adapter->netdev->flags & IFF_PROMISC))
5963 		err = igb_vlvf_set(adapter, vid, add,
5964 				   adapter->vfs_allocated_count);
5965 	if (err)
5966 		goto out;
5967 
5968 	err = igb_vlvf_set(adapter, vid, add, vf);
5969 
5970 	if (err)
5971 		goto out;
5972 
5973 	/* Go through all the checks to see if the VLAN filter should
5974 	 * be wiped completely.
5975 	 */
5976 	if (!add && (adapter->netdev->flags & IFF_PROMISC)) {
5977 		u32 vlvf, bits;
5978 		int regndx = igb_find_vlvf_entry(adapter, vid);
5979 
5980 		if (regndx < 0)
5981 			goto out;
5982 		/* See if any other pools are set for this VLAN filter
5983 		 * entry other than the PF.
5984 		 */
5985 		vlvf = bits = rd32(E1000_VLVF(regndx));
5986 		bits &= 1 << (E1000_VLVF_POOLSEL_SHIFT +
5987 			      adapter->vfs_allocated_count);
5988 		/* If the filter was removed then ensure PF pool bit
5989 		 * is cleared if the PF only added itself to the pool
5990 		 * because the PF is in promiscuous mode.
5991 		 */
5992 		if ((vlvf & VLAN_VID_MASK) == vid &&
5993 		    !test_bit(vid, adapter->active_vlans) &&
5994 		    !bits)
5995 			igb_vlvf_set(adapter, vid, add,
5996 				     adapter->vfs_allocated_count);
5997 	}
5998 
5999 out:
6000 	return err;
6001 }
6002 
6003 static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
6004 {
6005 	/* clear flags - except flag that indicates PF has set the MAC */
6006 	adapter->vf_data[vf].flags &= IGB_VF_FLAG_PF_SET_MAC;
6007 	adapter->vf_data[vf].last_nack = jiffies;
6008 
6009 	/* reset offloads to defaults */
6010 	igb_set_vmolr(adapter, vf, true);
6011 
6012 	/* reset vlans for device */
6013 	igb_clear_vf_vfta(adapter, vf);
6014 	if (adapter->vf_data[vf].pf_vlan)
6015 		igb_ndo_set_vf_vlan(adapter->netdev, vf,
6016 				    adapter->vf_data[vf].pf_vlan,
6017 				    adapter->vf_data[vf].pf_qos);
6018 	else
6019 		igb_clear_vf_vfta(adapter, vf);
6020 
6021 	/* reset multicast table array for vf */
6022 	adapter->vf_data[vf].num_vf_mc_hashes = 0;
6023 
6024 	/* Flush and reset the mta with the new values */
6025 	igb_set_rx_mode(adapter->netdev);
6026 }
6027 
6028 static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
6029 {
6030 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6031 
6032 	/* clear mac address as we were hotplug removed/added */
6033 	if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
6034 		eth_zero_addr(vf_mac);
6035 
6036 	/* process remaining reset events */
6037 	igb_vf_reset(adapter, vf);
6038 }
6039 
6040 static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
6041 {
6042 	struct e1000_hw *hw = &adapter->hw;
6043 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6044 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
6045 	u32 reg, msgbuf[3];
6046 	u8 *addr = (u8 *)(&msgbuf[1]);
6047 
6048 	/* process all the same items cleared in a function level reset */
6049 	igb_vf_reset(adapter, vf);
6050 
6051 	/* set vf mac address */
6052 	igb_rar_set_qsel(adapter, vf_mac, rar_entry, vf);
6053 
6054 	/* enable transmit and receive for vf */
6055 	reg = rd32(E1000_VFTE);
6056 	wr32(E1000_VFTE, reg | (1 << vf));
6057 	reg = rd32(E1000_VFRE);
6058 	wr32(E1000_VFRE, reg | (1 << vf));
6059 
6060 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
6061 
6062 	/* reply to reset with ack and vf mac address */
6063 	if (!is_zero_ether_addr(vf_mac)) {
6064 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
6065 		memcpy(addr, vf_mac, ETH_ALEN);
6066 	} else {
6067 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
6068 	}
6069 	igb_write_mbx(hw, msgbuf, 3, vf);
6070 }
6071 
6072 static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
6073 {
6074 	/* The VF MAC Address is stored in a packed array of bytes
6075 	 * starting at the second 32 bit word of the msg array
6076 	 */
6077 	unsigned char *addr = (char *)&msg[1];
6078 	int err = -1;
6079 
6080 	if (is_valid_ether_addr(addr))
6081 		err = igb_set_vf_mac(adapter, vf, addr);
6082 
6083 	return err;
6084 }
6085 
6086 static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
6087 {
6088 	struct e1000_hw *hw = &adapter->hw;
6089 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6090 	u32 msg = E1000_VT_MSGTYPE_NACK;
6091 
6092 	/* if device isn't clear to send it shouldn't be reading either */
6093 	if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
6094 	    time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
6095 		igb_write_mbx(hw, &msg, 1, vf);
6096 		vf_data->last_nack = jiffies;
6097 	}
6098 }
6099 
6100 static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
6101 {
6102 	struct pci_dev *pdev = adapter->pdev;
6103 	u32 msgbuf[E1000_VFMAILBOX_SIZE];
6104 	struct e1000_hw *hw = &adapter->hw;
6105 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6106 	s32 retval;
6107 
6108 	retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf);
6109 
6110 	if (retval) {
6111 		/* if receive failed revoke VF CTS stats and restart init */
6112 		dev_err(&pdev->dev, "Error receiving message from VF\n");
6113 		vf_data->flags &= ~IGB_VF_FLAG_CTS;
6114 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
6115 			return;
6116 		goto out;
6117 	}
6118 
6119 	/* this is a message we already processed, do nothing */
6120 	if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
6121 		return;
6122 
6123 	/* until the vf completes a reset it should not be
6124 	 * allowed to start any configuration.
6125 	 */
6126 	if (msgbuf[0] == E1000_VF_RESET) {
6127 		igb_vf_reset_msg(adapter, vf);
6128 		return;
6129 	}
6130 
6131 	if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
6132 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
6133 			return;
6134 		retval = -1;
6135 		goto out;
6136 	}
6137 
6138 	switch ((msgbuf[0] & 0xFFFF)) {
6139 	case E1000_VF_SET_MAC_ADDR:
6140 		retval = -EINVAL;
6141 		if (!(vf_data->flags & IGB_VF_FLAG_PF_SET_MAC))
6142 			retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
6143 		else
6144 			dev_warn(&pdev->dev,
6145 				 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
6146 				 vf);
6147 		break;
6148 	case E1000_VF_SET_PROMISC:
6149 		retval = igb_set_vf_promisc(adapter, msgbuf, vf);
6150 		break;
6151 	case E1000_VF_SET_MULTICAST:
6152 		retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
6153 		break;
6154 	case E1000_VF_SET_LPE:
6155 		retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
6156 		break;
6157 	case E1000_VF_SET_VLAN:
6158 		retval = -1;
6159 		if (vf_data->pf_vlan)
6160 			dev_warn(&pdev->dev,
6161 				 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
6162 				 vf);
6163 		else
6164 			retval = igb_set_vf_vlan(adapter, msgbuf, vf);
6165 		break;
6166 	default:
6167 		dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
6168 		retval = -1;
6169 		break;
6170 	}
6171 
6172 	msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
6173 out:
6174 	/* notify the VF of the results of what it sent us */
6175 	if (retval)
6176 		msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
6177 	else
6178 		msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
6179 
6180 	igb_write_mbx(hw, msgbuf, 1, vf);
6181 }
6182 
6183 static void igb_msg_task(struct igb_adapter *adapter)
6184 {
6185 	struct e1000_hw *hw = &adapter->hw;
6186 	u32 vf;
6187 
6188 	for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
6189 		/* process any reset requests */
6190 		if (!igb_check_for_rst(hw, vf))
6191 			igb_vf_reset_event(adapter, vf);
6192 
6193 		/* process any messages pending */
6194 		if (!igb_check_for_msg(hw, vf))
6195 			igb_rcv_msg_from_vf(adapter, vf);
6196 
6197 		/* process any acks */
6198 		if (!igb_check_for_ack(hw, vf))
6199 			igb_rcv_ack_from_vf(adapter, vf);
6200 	}
6201 }
6202 
6203 /**
6204  *  igb_set_uta - Set unicast filter table address
6205  *  @adapter: board private structure
6206  *
6207  *  The unicast table address is a register array of 32-bit registers.
6208  *  The table is meant to be used in a way similar to how the MTA is used
6209  *  however due to certain limitations in the hardware it is necessary to
6210  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
6211  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
6212  **/
6213 static void igb_set_uta(struct igb_adapter *adapter)
6214 {
6215 	struct e1000_hw *hw = &adapter->hw;
6216 	int i;
6217 
6218 	/* The UTA table only exists on 82576 hardware and newer */
6219 	if (hw->mac.type < e1000_82576)
6220 		return;
6221 
6222 	/* we only need to do this if VMDq is enabled */
6223 	if (!adapter->vfs_allocated_count)
6224 		return;
6225 
6226 	for (i = 0; i < hw->mac.uta_reg_count; i++)
6227 		array_wr32(E1000_UTA, i, ~0);
6228 }
6229 
6230 /**
6231  *  igb_intr_msi - Interrupt Handler
6232  *  @irq: interrupt number
6233  *  @data: pointer to a network interface device structure
6234  **/
6235 static irqreturn_t igb_intr_msi(int irq, void *data)
6236 {
6237 	struct igb_adapter *adapter = data;
6238 	struct igb_q_vector *q_vector = adapter->q_vector[0];
6239 	struct e1000_hw *hw = &adapter->hw;
6240 	/* read ICR disables interrupts using IAM */
6241 	u32 icr = rd32(E1000_ICR);
6242 
6243 	igb_write_itr(q_vector);
6244 
6245 	if (icr & E1000_ICR_DRSTA)
6246 		schedule_work(&adapter->reset_task);
6247 
6248 	if (icr & E1000_ICR_DOUTSYNC) {
6249 		/* HW is reporting DMA is out of sync */
6250 		adapter->stats.doosync++;
6251 	}
6252 
6253 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
6254 		hw->mac.get_link_status = 1;
6255 		if (!test_bit(__IGB_DOWN, &adapter->state))
6256 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6257 	}
6258 
6259 	if (icr & E1000_ICR_TS)
6260 		igb_tsync_interrupt(adapter);
6261 
6262 	napi_schedule(&q_vector->napi);
6263 
6264 	return IRQ_HANDLED;
6265 }
6266 
6267 /**
6268  *  igb_intr - Legacy Interrupt Handler
6269  *  @irq: interrupt number
6270  *  @data: pointer to a network interface device structure
6271  **/
6272 static irqreturn_t igb_intr(int irq, void *data)
6273 {
6274 	struct igb_adapter *adapter = data;
6275 	struct igb_q_vector *q_vector = adapter->q_vector[0];
6276 	struct e1000_hw *hw = &adapter->hw;
6277 	/* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
6278 	 * need for the IMC write
6279 	 */
6280 	u32 icr = rd32(E1000_ICR);
6281 
6282 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
6283 	 * not set, then the adapter didn't send an interrupt
6284 	 */
6285 	if (!(icr & E1000_ICR_INT_ASSERTED))
6286 		return IRQ_NONE;
6287 
6288 	igb_write_itr(q_vector);
6289 
6290 	if (icr & E1000_ICR_DRSTA)
6291 		schedule_work(&adapter->reset_task);
6292 
6293 	if (icr & E1000_ICR_DOUTSYNC) {
6294 		/* HW is reporting DMA is out of sync */
6295 		adapter->stats.doosync++;
6296 	}
6297 
6298 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
6299 		hw->mac.get_link_status = 1;
6300 		/* guard against interrupt when we're going down */
6301 		if (!test_bit(__IGB_DOWN, &adapter->state))
6302 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6303 	}
6304 
6305 	if (icr & E1000_ICR_TS)
6306 		igb_tsync_interrupt(adapter);
6307 
6308 	napi_schedule(&q_vector->napi);
6309 
6310 	return IRQ_HANDLED;
6311 }
6312 
6313 static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
6314 {
6315 	struct igb_adapter *adapter = q_vector->adapter;
6316 	struct e1000_hw *hw = &adapter->hw;
6317 
6318 	if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
6319 	    (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
6320 		if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
6321 			igb_set_itr(q_vector);
6322 		else
6323 			igb_update_ring_itr(q_vector);
6324 	}
6325 
6326 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
6327 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
6328 			wr32(E1000_EIMS, q_vector->eims_value);
6329 		else
6330 			igb_irq_enable(adapter);
6331 	}
6332 }
6333 
6334 /**
6335  *  igb_poll - NAPI Rx polling callback
6336  *  @napi: napi polling structure
6337  *  @budget: count of how many packets we should handle
6338  **/
6339 static int igb_poll(struct napi_struct *napi, int budget)
6340 {
6341 	struct igb_q_vector *q_vector = container_of(napi,
6342 						     struct igb_q_vector,
6343 						     napi);
6344 	bool clean_complete = true;
6345 
6346 #ifdef CONFIG_IGB_DCA
6347 	if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
6348 		igb_update_dca(q_vector);
6349 #endif
6350 	if (q_vector->tx.ring)
6351 		clean_complete = igb_clean_tx_irq(q_vector);
6352 
6353 	if (q_vector->rx.ring)
6354 		clean_complete &= igb_clean_rx_irq(q_vector, budget);
6355 
6356 	/* If all work not completed, return budget and keep polling */
6357 	if (!clean_complete)
6358 		return budget;
6359 
6360 	/* If not enough Rx work done, exit the polling mode */
6361 	napi_complete(napi);
6362 	igb_ring_irq_enable(q_vector);
6363 
6364 	return 0;
6365 }
6366 
6367 /**
6368  *  igb_clean_tx_irq - Reclaim resources after transmit completes
6369  *  @q_vector: pointer to q_vector containing needed info
6370  *
6371  *  returns true if ring is completely cleaned
6372  **/
6373 static bool igb_clean_tx_irq(struct igb_q_vector *q_vector)
6374 {
6375 	struct igb_adapter *adapter = q_vector->adapter;
6376 	struct igb_ring *tx_ring = q_vector->tx.ring;
6377 	struct igb_tx_buffer *tx_buffer;
6378 	union e1000_adv_tx_desc *tx_desc;
6379 	unsigned int total_bytes = 0, total_packets = 0;
6380 	unsigned int budget = q_vector->tx.work_limit;
6381 	unsigned int i = tx_ring->next_to_clean;
6382 
6383 	if (test_bit(__IGB_DOWN, &adapter->state))
6384 		return true;
6385 
6386 	tx_buffer = &tx_ring->tx_buffer_info[i];
6387 	tx_desc = IGB_TX_DESC(tx_ring, i);
6388 	i -= tx_ring->count;
6389 
6390 	do {
6391 		union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
6392 
6393 		/* if next_to_watch is not set then there is no work pending */
6394 		if (!eop_desc)
6395 			break;
6396 
6397 		/* prevent any other reads prior to eop_desc */
6398 		read_barrier_depends();
6399 
6400 		/* if DD is not set pending work has not been completed */
6401 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
6402 			break;
6403 
6404 		/* clear next_to_watch to prevent false hangs */
6405 		tx_buffer->next_to_watch = NULL;
6406 
6407 		/* update the statistics for this packet */
6408 		total_bytes += tx_buffer->bytecount;
6409 		total_packets += tx_buffer->gso_segs;
6410 
6411 		/* free the skb */
6412 		dev_consume_skb_any(tx_buffer->skb);
6413 
6414 		/* unmap skb header data */
6415 		dma_unmap_single(tx_ring->dev,
6416 				 dma_unmap_addr(tx_buffer, dma),
6417 				 dma_unmap_len(tx_buffer, len),
6418 				 DMA_TO_DEVICE);
6419 
6420 		/* clear tx_buffer data */
6421 		tx_buffer->skb = NULL;
6422 		dma_unmap_len_set(tx_buffer, len, 0);
6423 
6424 		/* clear last DMA location and unmap remaining buffers */
6425 		while (tx_desc != eop_desc) {
6426 			tx_buffer++;
6427 			tx_desc++;
6428 			i++;
6429 			if (unlikely(!i)) {
6430 				i -= tx_ring->count;
6431 				tx_buffer = tx_ring->tx_buffer_info;
6432 				tx_desc = IGB_TX_DESC(tx_ring, 0);
6433 			}
6434 
6435 			/* unmap any remaining paged data */
6436 			if (dma_unmap_len(tx_buffer, len)) {
6437 				dma_unmap_page(tx_ring->dev,
6438 					       dma_unmap_addr(tx_buffer, dma),
6439 					       dma_unmap_len(tx_buffer, len),
6440 					       DMA_TO_DEVICE);
6441 				dma_unmap_len_set(tx_buffer, len, 0);
6442 			}
6443 		}
6444 
6445 		/* move us one more past the eop_desc for start of next pkt */
6446 		tx_buffer++;
6447 		tx_desc++;
6448 		i++;
6449 		if (unlikely(!i)) {
6450 			i -= tx_ring->count;
6451 			tx_buffer = tx_ring->tx_buffer_info;
6452 			tx_desc = IGB_TX_DESC(tx_ring, 0);
6453 		}
6454 
6455 		/* issue prefetch for next Tx descriptor */
6456 		prefetch(tx_desc);
6457 
6458 		/* update budget accounting */
6459 		budget--;
6460 	} while (likely(budget));
6461 
6462 	netdev_tx_completed_queue(txring_txq(tx_ring),
6463 				  total_packets, total_bytes);
6464 	i += tx_ring->count;
6465 	tx_ring->next_to_clean = i;
6466 	u64_stats_update_begin(&tx_ring->tx_syncp);
6467 	tx_ring->tx_stats.bytes += total_bytes;
6468 	tx_ring->tx_stats.packets += total_packets;
6469 	u64_stats_update_end(&tx_ring->tx_syncp);
6470 	q_vector->tx.total_bytes += total_bytes;
6471 	q_vector->tx.total_packets += total_packets;
6472 
6473 	if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
6474 		struct e1000_hw *hw = &adapter->hw;
6475 
6476 		/* Detect a transmit hang in hardware, this serializes the
6477 		 * check with the clearing of time_stamp and movement of i
6478 		 */
6479 		clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
6480 		if (tx_buffer->next_to_watch &&
6481 		    time_after(jiffies, tx_buffer->time_stamp +
6482 			       (adapter->tx_timeout_factor * HZ)) &&
6483 		    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
6484 
6485 			/* detected Tx unit hang */
6486 			dev_err(tx_ring->dev,
6487 				"Detected Tx Unit Hang\n"
6488 				"  Tx Queue             <%d>\n"
6489 				"  TDH                  <%x>\n"
6490 				"  TDT                  <%x>\n"
6491 				"  next_to_use          <%x>\n"
6492 				"  next_to_clean        <%x>\n"
6493 				"buffer_info[next_to_clean]\n"
6494 				"  time_stamp           <%lx>\n"
6495 				"  next_to_watch        <%p>\n"
6496 				"  jiffies              <%lx>\n"
6497 				"  desc.status          <%x>\n",
6498 				tx_ring->queue_index,
6499 				rd32(E1000_TDH(tx_ring->reg_idx)),
6500 				readl(tx_ring->tail),
6501 				tx_ring->next_to_use,
6502 				tx_ring->next_to_clean,
6503 				tx_buffer->time_stamp,
6504 				tx_buffer->next_to_watch,
6505 				jiffies,
6506 				tx_buffer->next_to_watch->wb.status);
6507 			netif_stop_subqueue(tx_ring->netdev,
6508 					    tx_ring->queue_index);
6509 
6510 			/* we are about to reset, no point in enabling stuff */
6511 			return true;
6512 		}
6513 	}
6514 
6515 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
6516 	if (unlikely(total_packets &&
6517 	    netif_carrier_ok(tx_ring->netdev) &&
6518 	    igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
6519 		/* Make sure that anybody stopping the queue after this
6520 		 * sees the new next_to_clean.
6521 		 */
6522 		smp_mb();
6523 		if (__netif_subqueue_stopped(tx_ring->netdev,
6524 					     tx_ring->queue_index) &&
6525 		    !(test_bit(__IGB_DOWN, &adapter->state))) {
6526 			netif_wake_subqueue(tx_ring->netdev,
6527 					    tx_ring->queue_index);
6528 
6529 			u64_stats_update_begin(&tx_ring->tx_syncp);
6530 			tx_ring->tx_stats.restart_queue++;
6531 			u64_stats_update_end(&tx_ring->tx_syncp);
6532 		}
6533 	}
6534 
6535 	return !!budget;
6536 }
6537 
6538 /**
6539  *  igb_reuse_rx_page - page flip buffer and store it back on the ring
6540  *  @rx_ring: rx descriptor ring to store buffers on
6541  *  @old_buff: donor buffer to have page reused
6542  *
6543  *  Synchronizes page for reuse by the adapter
6544  **/
6545 static void igb_reuse_rx_page(struct igb_ring *rx_ring,
6546 			      struct igb_rx_buffer *old_buff)
6547 {
6548 	struct igb_rx_buffer *new_buff;
6549 	u16 nta = rx_ring->next_to_alloc;
6550 
6551 	new_buff = &rx_ring->rx_buffer_info[nta];
6552 
6553 	/* update, and store next to alloc */
6554 	nta++;
6555 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
6556 
6557 	/* transfer page from old buffer to new buffer */
6558 	*new_buff = *old_buff;
6559 
6560 	/* sync the buffer for use by the device */
6561 	dma_sync_single_range_for_device(rx_ring->dev, old_buff->dma,
6562 					 old_buff->page_offset,
6563 					 IGB_RX_BUFSZ,
6564 					 DMA_FROM_DEVICE);
6565 }
6566 
6567 static inline bool igb_page_is_reserved(struct page *page)
6568 {
6569 	return (page_to_nid(page) != numa_mem_id()) || page->pfmemalloc;
6570 }
6571 
6572 static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer,
6573 				  struct page *page,
6574 				  unsigned int truesize)
6575 {
6576 	/* avoid re-using remote pages */
6577 	if (unlikely(igb_page_is_reserved(page)))
6578 		return false;
6579 
6580 #if (PAGE_SIZE < 8192)
6581 	/* if we are only owner of page we can reuse it */
6582 	if (unlikely(page_count(page) != 1))
6583 		return false;
6584 
6585 	/* flip page offset to other buffer */
6586 	rx_buffer->page_offset ^= IGB_RX_BUFSZ;
6587 #else
6588 	/* move offset up to the next cache line */
6589 	rx_buffer->page_offset += truesize;
6590 
6591 	if (rx_buffer->page_offset > (PAGE_SIZE - IGB_RX_BUFSZ))
6592 		return false;
6593 #endif
6594 
6595 	/* Even if we own the page, we are not allowed to use atomic_set()
6596 	 * This would break get_page_unless_zero() users.
6597 	 */
6598 	atomic_inc(&page->_count);
6599 
6600 	return true;
6601 }
6602 
6603 /**
6604  *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
6605  *  @rx_ring: rx descriptor ring to transact packets on
6606  *  @rx_buffer: buffer containing page to add
6607  *  @rx_desc: descriptor containing length of buffer written by hardware
6608  *  @skb: sk_buff to place the data into
6609  *
6610  *  This function will add the data contained in rx_buffer->page to the skb.
6611  *  This is done either through a direct copy if the data in the buffer is
6612  *  less than the skb header size, otherwise it will just attach the page as
6613  *  a frag to the skb.
6614  *
6615  *  The function will then update the page offset if necessary and return
6616  *  true if the buffer can be reused by the adapter.
6617  **/
6618 static bool igb_add_rx_frag(struct igb_ring *rx_ring,
6619 			    struct igb_rx_buffer *rx_buffer,
6620 			    union e1000_adv_rx_desc *rx_desc,
6621 			    struct sk_buff *skb)
6622 {
6623 	struct page *page = rx_buffer->page;
6624 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
6625 #if (PAGE_SIZE < 8192)
6626 	unsigned int truesize = IGB_RX_BUFSZ;
6627 #else
6628 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
6629 #endif
6630 
6631 	if ((size <= IGB_RX_HDR_LEN) && !skb_is_nonlinear(skb)) {
6632 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
6633 
6634 		if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
6635 			igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
6636 			va += IGB_TS_HDR_LEN;
6637 			size -= IGB_TS_HDR_LEN;
6638 		}
6639 
6640 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
6641 
6642 		/* page is not reserved, we can reuse buffer as-is */
6643 		if (likely(!igb_page_is_reserved(page)))
6644 			return true;
6645 
6646 		/* this page cannot be reused so discard it */
6647 		__free_page(page);
6648 		return false;
6649 	}
6650 
6651 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
6652 			rx_buffer->page_offset, size, truesize);
6653 
6654 	return igb_can_reuse_rx_page(rx_buffer, page, truesize);
6655 }
6656 
6657 static struct sk_buff *igb_fetch_rx_buffer(struct igb_ring *rx_ring,
6658 					   union e1000_adv_rx_desc *rx_desc,
6659 					   struct sk_buff *skb)
6660 {
6661 	struct igb_rx_buffer *rx_buffer;
6662 	struct page *page;
6663 
6664 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
6665 	page = rx_buffer->page;
6666 	prefetchw(page);
6667 
6668 	if (likely(!skb)) {
6669 		void *page_addr = page_address(page) +
6670 				  rx_buffer->page_offset;
6671 
6672 		/* prefetch first cache line of first page */
6673 		prefetch(page_addr);
6674 #if L1_CACHE_BYTES < 128
6675 		prefetch(page_addr + L1_CACHE_BYTES);
6676 #endif
6677 
6678 		/* allocate a skb to store the frags */
6679 		skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
6680 		if (unlikely(!skb)) {
6681 			rx_ring->rx_stats.alloc_failed++;
6682 			return NULL;
6683 		}
6684 
6685 		/* we will be copying header into skb->data in
6686 		 * pskb_may_pull so it is in our interest to prefetch
6687 		 * it now to avoid a possible cache miss
6688 		 */
6689 		prefetchw(skb->data);
6690 	}
6691 
6692 	/* we are reusing so sync this buffer for CPU use */
6693 	dma_sync_single_range_for_cpu(rx_ring->dev,
6694 				      rx_buffer->dma,
6695 				      rx_buffer->page_offset,
6696 				      IGB_RX_BUFSZ,
6697 				      DMA_FROM_DEVICE);
6698 
6699 	/* pull page into skb */
6700 	if (igb_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
6701 		/* hand second half of page back to the ring */
6702 		igb_reuse_rx_page(rx_ring, rx_buffer);
6703 	} else {
6704 		/* we are not reusing the buffer so unmap it */
6705 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
6706 			       PAGE_SIZE, DMA_FROM_DEVICE);
6707 	}
6708 
6709 	/* clear contents of rx_buffer */
6710 	rx_buffer->page = NULL;
6711 
6712 	return skb;
6713 }
6714 
6715 static inline void igb_rx_checksum(struct igb_ring *ring,
6716 				   union e1000_adv_rx_desc *rx_desc,
6717 				   struct sk_buff *skb)
6718 {
6719 	skb_checksum_none_assert(skb);
6720 
6721 	/* Ignore Checksum bit is set */
6722 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
6723 		return;
6724 
6725 	/* Rx checksum disabled via ethtool */
6726 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
6727 		return;
6728 
6729 	/* TCP/UDP checksum error bit is set */
6730 	if (igb_test_staterr(rx_desc,
6731 			     E1000_RXDEXT_STATERR_TCPE |
6732 			     E1000_RXDEXT_STATERR_IPE)) {
6733 		/* work around errata with sctp packets where the TCPE aka
6734 		 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
6735 		 * packets, (aka let the stack check the crc32c)
6736 		 */
6737 		if (!((skb->len == 60) &&
6738 		      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
6739 			u64_stats_update_begin(&ring->rx_syncp);
6740 			ring->rx_stats.csum_err++;
6741 			u64_stats_update_end(&ring->rx_syncp);
6742 		}
6743 		/* let the stack verify checksum errors */
6744 		return;
6745 	}
6746 	/* It must be a TCP or UDP packet with a valid checksum */
6747 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
6748 				      E1000_RXD_STAT_UDPCS))
6749 		skb->ip_summed = CHECKSUM_UNNECESSARY;
6750 
6751 	dev_dbg(ring->dev, "cksum success: bits %08X\n",
6752 		le32_to_cpu(rx_desc->wb.upper.status_error));
6753 }
6754 
6755 static inline void igb_rx_hash(struct igb_ring *ring,
6756 			       union e1000_adv_rx_desc *rx_desc,
6757 			       struct sk_buff *skb)
6758 {
6759 	if (ring->netdev->features & NETIF_F_RXHASH)
6760 		skb_set_hash(skb,
6761 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
6762 			     PKT_HASH_TYPE_L3);
6763 }
6764 
6765 /**
6766  *  igb_is_non_eop - process handling of non-EOP buffers
6767  *  @rx_ring: Rx ring being processed
6768  *  @rx_desc: Rx descriptor for current buffer
6769  *  @skb: current socket buffer containing buffer in progress
6770  *
6771  *  This function updates next to clean.  If the buffer is an EOP buffer
6772  *  this function exits returning false, otherwise it will place the
6773  *  sk_buff in the next buffer to be chained and return true indicating
6774  *  that this is in fact a non-EOP buffer.
6775  **/
6776 static bool igb_is_non_eop(struct igb_ring *rx_ring,
6777 			   union e1000_adv_rx_desc *rx_desc)
6778 {
6779 	u32 ntc = rx_ring->next_to_clean + 1;
6780 
6781 	/* fetch, update, and store next to clean */
6782 	ntc = (ntc < rx_ring->count) ? ntc : 0;
6783 	rx_ring->next_to_clean = ntc;
6784 
6785 	prefetch(IGB_RX_DESC(rx_ring, ntc));
6786 
6787 	if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
6788 		return false;
6789 
6790 	return true;
6791 }
6792 
6793 /**
6794  *  igb_pull_tail - igb specific version of skb_pull_tail
6795  *  @rx_ring: rx descriptor ring packet is being transacted on
6796  *  @rx_desc: pointer to the EOP Rx descriptor
6797  *  @skb: pointer to current skb being adjusted
6798  *
6799  *  This function is an igb specific version of __pskb_pull_tail.  The
6800  *  main difference between this version and the original function is that
6801  *  this function can make several assumptions about the state of things
6802  *  that allow for significant optimizations versus the standard function.
6803  *  As a result we can do things like drop a frag and maintain an accurate
6804  *  truesize for the skb.
6805  */
6806 static void igb_pull_tail(struct igb_ring *rx_ring,
6807 			  union e1000_adv_rx_desc *rx_desc,
6808 			  struct sk_buff *skb)
6809 {
6810 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
6811 	unsigned char *va;
6812 	unsigned int pull_len;
6813 
6814 	/* it is valid to use page_address instead of kmap since we are
6815 	 * working with pages allocated out of the lomem pool per
6816 	 * alloc_page(GFP_ATOMIC)
6817 	 */
6818 	va = skb_frag_address(frag);
6819 
6820 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
6821 		/* retrieve timestamp from buffer */
6822 		igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
6823 
6824 		/* update pointers to remove timestamp header */
6825 		skb_frag_size_sub(frag, IGB_TS_HDR_LEN);
6826 		frag->page_offset += IGB_TS_HDR_LEN;
6827 		skb->data_len -= IGB_TS_HDR_LEN;
6828 		skb->len -= IGB_TS_HDR_LEN;
6829 
6830 		/* move va to start of packet data */
6831 		va += IGB_TS_HDR_LEN;
6832 	}
6833 
6834 	/* we need the header to contain the greater of either ETH_HLEN or
6835 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
6836 	 */
6837 	pull_len = eth_get_headlen(va, IGB_RX_HDR_LEN);
6838 
6839 	/* align pull length to size of long to optimize memcpy performance */
6840 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
6841 
6842 	/* update all of the pointers */
6843 	skb_frag_size_sub(frag, pull_len);
6844 	frag->page_offset += pull_len;
6845 	skb->data_len -= pull_len;
6846 	skb->tail += pull_len;
6847 }
6848 
6849 /**
6850  *  igb_cleanup_headers - Correct corrupted or empty headers
6851  *  @rx_ring: rx descriptor ring packet is being transacted on
6852  *  @rx_desc: pointer to the EOP Rx descriptor
6853  *  @skb: pointer to current skb being fixed
6854  *
6855  *  Address the case where we are pulling data in on pages only
6856  *  and as such no data is present in the skb header.
6857  *
6858  *  In addition if skb is not at least 60 bytes we need to pad it so that
6859  *  it is large enough to qualify as a valid Ethernet frame.
6860  *
6861  *  Returns true if an error was encountered and skb was freed.
6862  **/
6863 static bool igb_cleanup_headers(struct igb_ring *rx_ring,
6864 				union e1000_adv_rx_desc *rx_desc,
6865 				struct sk_buff *skb)
6866 {
6867 	if (unlikely((igb_test_staterr(rx_desc,
6868 				       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
6869 		struct net_device *netdev = rx_ring->netdev;
6870 		if (!(netdev->features & NETIF_F_RXALL)) {
6871 			dev_kfree_skb_any(skb);
6872 			return true;
6873 		}
6874 	}
6875 
6876 	/* place header in linear portion of buffer */
6877 	if (skb_is_nonlinear(skb))
6878 		igb_pull_tail(rx_ring, rx_desc, skb);
6879 
6880 	/* if eth_skb_pad returns an error the skb was freed */
6881 	if (eth_skb_pad(skb))
6882 		return true;
6883 
6884 	return false;
6885 }
6886 
6887 /**
6888  *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
6889  *  @rx_ring: rx descriptor ring packet is being transacted on
6890  *  @rx_desc: pointer to the EOP Rx descriptor
6891  *  @skb: pointer to current skb being populated
6892  *
6893  *  This function checks the ring, descriptor, and packet information in
6894  *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
6895  *  other fields within the skb.
6896  **/
6897 static void igb_process_skb_fields(struct igb_ring *rx_ring,
6898 				   union e1000_adv_rx_desc *rx_desc,
6899 				   struct sk_buff *skb)
6900 {
6901 	struct net_device *dev = rx_ring->netdev;
6902 
6903 	igb_rx_hash(rx_ring, rx_desc, skb);
6904 
6905 	igb_rx_checksum(rx_ring, rx_desc, skb);
6906 
6907 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
6908 	    !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
6909 		igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
6910 
6911 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
6912 	    igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
6913 		u16 vid;
6914 
6915 		if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
6916 		    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
6917 			vid = be16_to_cpu(rx_desc->wb.upper.vlan);
6918 		else
6919 			vid = le16_to_cpu(rx_desc->wb.upper.vlan);
6920 
6921 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
6922 	}
6923 
6924 	skb_record_rx_queue(skb, rx_ring->queue_index);
6925 
6926 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
6927 }
6928 
6929 static bool igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
6930 {
6931 	struct igb_ring *rx_ring = q_vector->rx.ring;
6932 	struct sk_buff *skb = rx_ring->skb;
6933 	unsigned int total_bytes = 0, total_packets = 0;
6934 	u16 cleaned_count = igb_desc_unused(rx_ring);
6935 
6936 	while (likely(total_packets < budget)) {
6937 		union e1000_adv_rx_desc *rx_desc;
6938 
6939 		/* return some buffers to hardware, one at a time is too slow */
6940 		if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
6941 			igb_alloc_rx_buffers(rx_ring, cleaned_count);
6942 			cleaned_count = 0;
6943 		}
6944 
6945 		rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
6946 
6947 		if (!rx_desc->wb.upper.status_error)
6948 			break;
6949 
6950 		/* This memory barrier is needed to keep us from reading
6951 		 * any other fields out of the rx_desc until we know the
6952 		 * descriptor has been written back
6953 		 */
6954 		dma_rmb();
6955 
6956 		/* retrieve a buffer from the ring */
6957 		skb = igb_fetch_rx_buffer(rx_ring, rx_desc, skb);
6958 
6959 		/* exit if we failed to retrieve a buffer */
6960 		if (!skb)
6961 			break;
6962 
6963 		cleaned_count++;
6964 
6965 		/* fetch next buffer in frame if non-eop */
6966 		if (igb_is_non_eop(rx_ring, rx_desc))
6967 			continue;
6968 
6969 		/* verify the packet layout is correct */
6970 		if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
6971 			skb = NULL;
6972 			continue;
6973 		}
6974 
6975 		/* probably a little skewed due to removing CRC */
6976 		total_bytes += skb->len;
6977 
6978 		/* populate checksum, timestamp, VLAN, and protocol */
6979 		igb_process_skb_fields(rx_ring, rx_desc, skb);
6980 
6981 		napi_gro_receive(&q_vector->napi, skb);
6982 
6983 		/* reset skb pointer */
6984 		skb = NULL;
6985 
6986 		/* update budget accounting */
6987 		total_packets++;
6988 	}
6989 
6990 	/* place incomplete frames back on ring for completion */
6991 	rx_ring->skb = skb;
6992 
6993 	u64_stats_update_begin(&rx_ring->rx_syncp);
6994 	rx_ring->rx_stats.packets += total_packets;
6995 	rx_ring->rx_stats.bytes += total_bytes;
6996 	u64_stats_update_end(&rx_ring->rx_syncp);
6997 	q_vector->rx.total_packets += total_packets;
6998 	q_vector->rx.total_bytes += total_bytes;
6999 
7000 	if (cleaned_count)
7001 		igb_alloc_rx_buffers(rx_ring, cleaned_count);
7002 
7003 	return total_packets < budget;
7004 }
7005 
7006 static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
7007 				  struct igb_rx_buffer *bi)
7008 {
7009 	struct page *page = bi->page;
7010 	dma_addr_t dma;
7011 
7012 	/* since we are recycling buffers we should seldom need to alloc */
7013 	if (likely(page))
7014 		return true;
7015 
7016 	/* alloc new page for storage */
7017 	page = dev_alloc_page();
7018 	if (unlikely(!page)) {
7019 		rx_ring->rx_stats.alloc_failed++;
7020 		return false;
7021 	}
7022 
7023 	/* map page for use */
7024 	dma = dma_map_page(rx_ring->dev, page, 0, PAGE_SIZE, DMA_FROM_DEVICE);
7025 
7026 	/* if mapping failed free memory back to system since
7027 	 * there isn't much point in holding memory we can't use
7028 	 */
7029 	if (dma_mapping_error(rx_ring->dev, dma)) {
7030 		__free_page(page);
7031 
7032 		rx_ring->rx_stats.alloc_failed++;
7033 		return false;
7034 	}
7035 
7036 	bi->dma = dma;
7037 	bi->page = page;
7038 	bi->page_offset = 0;
7039 
7040 	return true;
7041 }
7042 
7043 /**
7044  *  igb_alloc_rx_buffers - Replace used receive buffers; packet split
7045  *  @adapter: address of board private structure
7046  **/
7047 void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
7048 {
7049 	union e1000_adv_rx_desc *rx_desc;
7050 	struct igb_rx_buffer *bi;
7051 	u16 i = rx_ring->next_to_use;
7052 
7053 	/* nothing to do */
7054 	if (!cleaned_count)
7055 		return;
7056 
7057 	rx_desc = IGB_RX_DESC(rx_ring, i);
7058 	bi = &rx_ring->rx_buffer_info[i];
7059 	i -= rx_ring->count;
7060 
7061 	do {
7062 		if (!igb_alloc_mapped_page(rx_ring, bi))
7063 			break;
7064 
7065 		/* Refresh the desc even if buffer_addrs didn't change
7066 		 * because each write-back erases this info.
7067 		 */
7068 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
7069 
7070 		rx_desc++;
7071 		bi++;
7072 		i++;
7073 		if (unlikely(!i)) {
7074 			rx_desc = IGB_RX_DESC(rx_ring, 0);
7075 			bi = rx_ring->rx_buffer_info;
7076 			i -= rx_ring->count;
7077 		}
7078 
7079 		/* clear the status bits for the next_to_use descriptor */
7080 		rx_desc->wb.upper.status_error = 0;
7081 
7082 		cleaned_count--;
7083 	} while (cleaned_count);
7084 
7085 	i += rx_ring->count;
7086 
7087 	if (rx_ring->next_to_use != i) {
7088 		/* record the next descriptor to use */
7089 		rx_ring->next_to_use = i;
7090 
7091 		/* update next to alloc since we have filled the ring */
7092 		rx_ring->next_to_alloc = i;
7093 
7094 		/* Force memory writes to complete before letting h/w
7095 		 * know there are new descriptors to fetch.  (Only
7096 		 * applicable for weak-ordered memory model archs,
7097 		 * such as IA-64).
7098 		 */
7099 		wmb();
7100 		writel(i, rx_ring->tail);
7101 	}
7102 }
7103 
7104 /**
7105  * igb_mii_ioctl -
7106  * @netdev:
7107  * @ifreq:
7108  * @cmd:
7109  **/
7110 static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7111 {
7112 	struct igb_adapter *adapter = netdev_priv(netdev);
7113 	struct mii_ioctl_data *data = if_mii(ifr);
7114 
7115 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
7116 		return -EOPNOTSUPP;
7117 
7118 	switch (cmd) {
7119 	case SIOCGMIIPHY:
7120 		data->phy_id = adapter->hw.phy.addr;
7121 		break;
7122 	case SIOCGMIIREG:
7123 		if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
7124 				     &data->val_out))
7125 			return -EIO;
7126 		break;
7127 	case SIOCSMIIREG:
7128 	default:
7129 		return -EOPNOTSUPP;
7130 	}
7131 	return 0;
7132 }
7133 
7134 /**
7135  * igb_ioctl -
7136  * @netdev:
7137  * @ifreq:
7138  * @cmd:
7139  **/
7140 static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7141 {
7142 	switch (cmd) {
7143 	case SIOCGMIIPHY:
7144 	case SIOCGMIIREG:
7145 	case SIOCSMIIREG:
7146 		return igb_mii_ioctl(netdev, ifr, cmd);
7147 	case SIOCGHWTSTAMP:
7148 		return igb_ptp_get_ts_config(netdev, ifr);
7149 	case SIOCSHWTSTAMP:
7150 		return igb_ptp_set_ts_config(netdev, ifr);
7151 	default:
7152 		return -EOPNOTSUPP;
7153 	}
7154 }
7155 
7156 void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
7157 {
7158 	struct igb_adapter *adapter = hw->back;
7159 
7160 	pci_read_config_word(adapter->pdev, reg, value);
7161 }
7162 
7163 void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
7164 {
7165 	struct igb_adapter *adapter = hw->back;
7166 
7167 	pci_write_config_word(adapter->pdev, reg, *value);
7168 }
7169 
7170 s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
7171 {
7172 	struct igb_adapter *adapter = hw->back;
7173 
7174 	if (pcie_capability_read_word(adapter->pdev, reg, value))
7175 		return -E1000_ERR_CONFIG;
7176 
7177 	return 0;
7178 }
7179 
7180 s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
7181 {
7182 	struct igb_adapter *adapter = hw->back;
7183 
7184 	if (pcie_capability_write_word(adapter->pdev, reg, *value))
7185 		return -E1000_ERR_CONFIG;
7186 
7187 	return 0;
7188 }
7189 
7190 static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
7191 {
7192 	struct igb_adapter *adapter = netdev_priv(netdev);
7193 	struct e1000_hw *hw = &adapter->hw;
7194 	u32 ctrl, rctl;
7195 	bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
7196 
7197 	if (enable) {
7198 		/* enable VLAN tag insert/strip */
7199 		ctrl = rd32(E1000_CTRL);
7200 		ctrl |= E1000_CTRL_VME;
7201 		wr32(E1000_CTRL, ctrl);
7202 
7203 		/* Disable CFI check */
7204 		rctl = rd32(E1000_RCTL);
7205 		rctl &= ~E1000_RCTL_CFIEN;
7206 		wr32(E1000_RCTL, rctl);
7207 	} else {
7208 		/* disable VLAN tag insert/strip */
7209 		ctrl = rd32(E1000_CTRL);
7210 		ctrl &= ~E1000_CTRL_VME;
7211 		wr32(E1000_CTRL, ctrl);
7212 	}
7213 
7214 	igb_rlpml_set(adapter);
7215 }
7216 
7217 static int igb_vlan_rx_add_vid(struct net_device *netdev,
7218 			       __be16 proto, u16 vid)
7219 {
7220 	struct igb_adapter *adapter = netdev_priv(netdev);
7221 	struct e1000_hw *hw = &adapter->hw;
7222 	int pf_id = adapter->vfs_allocated_count;
7223 
7224 	/* attempt to add filter to vlvf array */
7225 	igb_vlvf_set(adapter, vid, true, pf_id);
7226 
7227 	/* add the filter since PF can receive vlans w/o entry in vlvf */
7228 	igb_vfta_set(hw, vid, true);
7229 
7230 	set_bit(vid, adapter->active_vlans);
7231 
7232 	return 0;
7233 }
7234 
7235 static int igb_vlan_rx_kill_vid(struct net_device *netdev,
7236 				__be16 proto, u16 vid)
7237 {
7238 	struct igb_adapter *adapter = netdev_priv(netdev);
7239 	struct e1000_hw *hw = &adapter->hw;
7240 	int pf_id = adapter->vfs_allocated_count;
7241 	s32 err;
7242 
7243 	/* remove vlan from VLVF table array */
7244 	err = igb_vlvf_set(adapter, vid, false, pf_id);
7245 
7246 	/* if vid was not present in VLVF just remove it from table */
7247 	if (err)
7248 		igb_vfta_set(hw, vid, false);
7249 
7250 	clear_bit(vid, adapter->active_vlans);
7251 
7252 	return 0;
7253 }
7254 
7255 static void igb_restore_vlan(struct igb_adapter *adapter)
7256 {
7257 	u16 vid;
7258 
7259 	igb_vlan_mode(adapter->netdev, adapter->netdev->features);
7260 
7261 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
7262 		igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
7263 }
7264 
7265 int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
7266 {
7267 	struct pci_dev *pdev = adapter->pdev;
7268 	struct e1000_mac_info *mac = &adapter->hw.mac;
7269 
7270 	mac->autoneg = 0;
7271 
7272 	/* Make sure dplx is at most 1 bit and lsb of speed is not set
7273 	 * for the switch() below to work
7274 	 */
7275 	if ((spd & 1) || (dplx & ~1))
7276 		goto err_inval;
7277 
7278 	/* Fiber NIC's only allow 1000 gbps Full duplex
7279 	 * and 100Mbps Full duplex for 100baseFx sfp
7280 	 */
7281 	if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
7282 		switch (spd + dplx) {
7283 		case SPEED_10 + DUPLEX_HALF:
7284 		case SPEED_10 + DUPLEX_FULL:
7285 		case SPEED_100 + DUPLEX_HALF:
7286 			goto err_inval;
7287 		default:
7288 			break;
7289 		}
7290 	}
7291 
7292 	switch (spd + dplx) {
7293 	case SPEED_10 + DUPLEX_HALF:
7294 		mac->forced_speed_duplex = ADVERTISE_10_HALF;
7295 		break;
7296 	case SPEED_10 + DUPLEX_FULL:
7297 		mac->forced_speed_duplex = ADVERTISE_10_FULL;
7298 		break;
7299 	case SPEED_100 + DUPLEX_HALF:
7300 		mac->forced_speed_duplex = ADVERTISE_100_HALF;
7301 		break;
7302 	case SPEED_100 + DUPLEX_FULL:
7303 		mac->forced_speed_duplex = ADVERTISE_100_FULL;
7304 		break;
7305 	case SPEED_1000 + DUPLEX_FULL:
7306 		mac->autoneg = 1;
7307 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
7308 		break;
7309 	case SPEED_1000 + DUPLEX_HALF: /* not supported */
7310 	default:
7311 		goto err_inval;
7312 	}
7313 
7314 	/* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
7315 	adapter->hw.phy.mdix = AUTO_ALL_MODES;
7316 
7317 	return 0;
7318 
7319 err_inval:
7320 	dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
7321 	return -EINVAL;
7322 }
7323 
7324 static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
7325 			  bool runtime)
7326 {
7327 	struct net_device *netdev = pci_get_drvdata(pdev);
7328 	struct igb_adapter *adapter = netdev_priv(netdev);
7329 	struct e1000_hw *hw = &adapter->hw;
7330 	u32 ctrl, rctl, status;
7331 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
7332 #ifdef CONFIG_PM
7333 	int retval = 0;
7334 #endif
7335 
7336 	netif_device_detach(netdev);
7337 
7338 	if (netif_running(netdev))
7339 		__igb_close(netdev, true);
7340 
7341 	igb_clear_interrupt_scheme(adapter);
7342 
7343 #ifdef CONFIG_PM
7344 	retval = pci_save_state(pdev);
7345 	if (retval)
7346 		return retval;
7347 #endif
7348 
7349 	status = rd32(E1000_STATUS);
7350 	if (status & E1000_STATUS_LU)
7351 		wufc &= ~E1000_WUFC_LNKC;
7352 
7353 	if (wufc) {
7354 		igb_setup_rctl(adapter);
7355 		igb_set_rx_mode(netdev);
7356 
7357 		/* turn on all-multi mode if wake on multicast is enabled */
7358 		if (wufc & E1000_WUFC_MC) {
7359 			rctl = rd32(E1000_RCTL);
7360 			rctl |= E1000_RCTL_MPE;
7361 			wr32(E1000_RCTL, rctl);
7362 		}
7363 
7364 		ctrl = rd32(E1000_CTRL);
7365 		/* advertise wake from D3Cold */
7366 		#define E1000_CTRL_ADVD3WUC 0x00100000
7367 		/* phy power management enable */
7368 		#define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
7369 		ctrl |= E1000_CTRL_ADVD3WUC;
7370 		wr32(E1000_CTRL, ctrl);
7371 
7372 		/* Allow time for pending master requests to run */
7373 		igb_disable_pcie_master(hw);
7374 
7375 		wr32(E1000_WUC, E1000_WUC_PME_EN);
7376 		wr32(E1000_WUFC, wufc);
7377 	} else {
7378 		wr32(E1000_WUC, 0);
7379 		wr32(E1000_WUFC, 0);
7380 	}
7381 
7382 	*enable_wake = wufc || adapter->en_mng_pt;
7383 	if (!*enable_wake)
7384 		igb_power_down_link(adapter);
7385 	else
7386 		igb_power_up_link(adapter);
7387 
7388 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7389 	 * would have already happened in close and is redundant.
7390 	 */
7391 	igb_release_hw_control(adapter);
7392 
7393 	pci_disable_device(pdev);
7394 
7395 	return 0;
7396 }
7397 
7398 #ifdef CONFIG_PM
7399 #ifdef CONFIG_PM_SLEEP
7400 static int igb_suspend(struct device *dev)
7401 {
7402 	int retval;
7403 	bool wake;
7404 	struct pci_dev *pdev = to_pci_dev(dev);
7405 
7406 	retval = __igb_shutdown(pdev, &wake, 0);
7407 	if (retval)
7408 		return retval;
7409 
7410 	if (wake) {
7411 		pci_prepare_to_sleep(pdev);
7412 	} else {
7413 		pci_wake_from_d3(pdev, false);
7414 		pci_set_power_state(pdev, PCI_D3hot);
7415 	}
7416 
7417 	return 0;
7418 }
7419 #endif /* CONFIG_PM_SLEEP */
7420 
7421 static int igb_resume(struct device *dev)
7422 {
7423 	struct pci_dev *pdev = to_pci_dev(dev);
7424 	struct net_device *netdev = pci_get_drvdata(pdev);
7425 	struct igb_adapter *adapter = netdev_priv(netdev);
7426 	struct e1000_hw *hw = &adapter->hw;
7427 	u32 err;
7428 
7429 	pci_set_power_state(pdev, PCI_D0);
7430 	pci_restore_state(pdev);
7431 	pci_save_state(pdev);
7432 
7433 	if (!pci_device_is_present(pdev))
7434 		return -ENODEV;
7435 	err = pci_enable_device_mem(pdev);
7436 	if (err) {
7437 		dev_err(&pdev->dev,
7438 			"igb: Cannot enable PCI device from suspend\n");
7439 		return err;
7440 	}
7441 	pci_set_master(pdev);
7442 
7443 	pci_enable_wake(pdev, PCI_D3hot, 0);
7444 	pci_enable_wake(pdev, PCI_D3cold, 0);
7445 
7446 	if (igb_init_interrupt_scheme(adapter, true)) {
7447 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
7448 		return -ENOMEM;
7449 	}
7450 
7451 	igb_reset(adapter);
7452 
7453 	/* let the f/w know that the h/w is now under the control of the
7454 	 * driver.
7455 	 */
7456 	igb_get_hw_control(adapter);
7457 
7458 	wr32(E1000_WUS, ~0);
7459 
7460 	if (netdev->flags & IFF_UP) {
7461 		rtnl_lock();
7462 		err = __igb_open(netdev, true);
7463 		rtnl_unlock();
7464 		if (err)
7465 			return err;
7466 	}
7467 
7468 	netif_device_attach(netdev);
7469 	return 0;
7470 }
7471 
7472 static int igb_runtime_idle(struct device *dev)
7473 {
7474 	struct pci_dev *pdev = to_pci_dev(dev);
7475 	struct net_device *netdev = pci_get_drvdata(pdev);
7476 	struct igb_adapter *adapter = netdev_priv(netdev);
7477 
7478 	if (!igb_has_link(adapter))
7479 		pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
7480 
7481 	return -EBUSY;
7482 }
7483 
7484 static int igb_runtime_suspend(struct device *dev)
7485 {
7486 	struct pci_dev *pdev = to_pci_dev(dev);
7487 	int retval;
7488 	bool wake;
7489 
7490 	retval = __igb_shutdown(pdev, &wake, 1);
7491 	if (retval)
7492 		return retval;
7493 
7494 	if (wake) {
7495 		pci_prepare_to_sleep(pdev);
7496 	} else {
7497 		pci_wake_from_d3(pdev, false);
7498 		pci_set_power_state(pdev, PCI_D3hot);
7499 	}
7500 
7501 	return 0;
7502 }
7503 
7504 static int igb_runtime_resume(struct device *dev)
7505 {
7506 	return igb_resume(dev);
7507 }
7508 #endif /* CONFIG_PM */
7509 
7510 static void igb_shutdown(struct pci_dev *pdev)
7511 {
7512 	bool wake;
7513 
7514 	__igb_shutdown(pdev, &wake, 0);
7515 
7516 	if (system_state == SYSTEM_POWER_OFF) {
7517 		pci_wake_from_d3(pdev, wake);
7518 		pci_set_power_state(pdev, PCI_D3hot);
7519 	}
7520 }
7521 
7522 #ifdef CONFIG_PCI_IOV
7523 static int igb_sriov_reinit(struct pci_dev *dev)
7524 {
7525 	struct net_device *netdev = pci_get_drvdata(dev);
7526 	struct igb_adapter *adapter = netdev_priv(netdev);
7527 	struct pci_dev *pdev = adapter->pdev;
7528 
7529 	rtnl_lock();
7530 
7531 	if (netif_running(netdev))
7532 		igb_close(netdev);
7533 	else
7534 		igb_reset(adapter);
7535 
7536 	igb_clear_interrupt_scheme(adapter);
7537 
7538 	igb_init_queue_configuration(adapter);
7539 
7540 	if (igb_init_interrupt_scheme(adapter, true)) {
7541 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
7542 		return -ENOMEM;
7543 	}
7544 
7545 	if (netif_running(netdev))
7546 		igb_open(netdev);
7547 
7548 	rtnl_unlock();
7549 
7550 	return 0;
7551 }
7552 
7553 static int igb_pci_disable_sriov(struct pci_dev *dev)
7554 {
7555 	int err = igb_disable_sriov(dev);
7556 
7557 	if (!err)
7558 		err = igb_sriov_reinit(dev);
7559 
7560 	return err;
7561 }
7562 
7563 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
7564 {
7565 	int err = igb_enable_sriov(dev, num_vfs);
7566 
7567 	if (err)
7568 		goto out;
7569 
7570 	err = igb_sriov_reinit(dev);
7571 	if (!err)
7572 		return num_vfs;
7573 
7574 out:
7575 	return err;
7576 }
7577 
7578 #endif
7579 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
7580 {
7581 #ifdef CONFIG_PCI_IOV
7582 	if (num_vfs == 0)
7583 		return igb_pci_disable_sriov(dev);
7584 	else
7585 		return igb_pci_enable_sriov(dev, num_vfs);
7586 #endif
7587 	return 0;
7588 }
7589 
7590 #ifdef CONFIG_NET_POLL_CONTROLLER
7591 /* Polling 'interrupt' - used by things like netconsole to send skbs
7592  * without having to re-enable interrupts. It's not called while
7593  * the interrupt routine is executing.
7594  */
7595 static void igb_netpoll(struct net_device *netdev)
7596 {
7597 	struct igb_adapter *adapter = netdev_priv(netdev);
7598 	struct e1000_hw *hw = &adapter->hw;
7599 	struct igb_q_vector *q_vector;
7600 	int i;
7601 
7602 	for (i = 0; i < adapter->num_q_vectors; i++) {
7603 		q_vector = adapter->q_vector[i];
7604 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
7605 			wr32(E1000_EIMC, q_vector->eims_value);
7606 		else
7607 			igb_irq_disable(adapter);
7608 		napi_schedule(&q_vector->napi);
7609 	}
7610 }
7611 #endif /* CONFIG_NET_POLL_CONTROLLER */
7612 
7613 /**
7614  *  igb_io_error_detected - called when PCI error is detected
7615  *  @pdev: Pointer to PCI device
7616  *  @state: The current pci connection state
7617  *
7618  *  This function is called after a PCI bus error affecting
7619  *  this device has been detected.
7620  **/
7621 static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
7622 					      pci_channel_state_t state)
7623 {
7624 	struct net_device *netdev = pci_get_drvdata(pdev);
7625 	struct igb_adapter *adapter = netdev_priv(netdev);
7626 
7627 	netif_device_detach(netdev);
7628 
7629 	if (state == pci_channel_io_perm_failure)
7630 		return PCI_ERS_RESULT_DISCONNECT;
7631 
7632 	if (netif_running(netdev))
7633 		igb_down(adapter);
7634 	pci_disable_device(pdev);
7635 
7636 	/* Request a slot slot reset. */
7637 	return PCI_ERS_RESULT_NEED_RESET;
7638 }
7639 
7640 /**
7641  *  igb_io_slot_reset - called after the pci bus has been reset.
7642  *  @pdev: Pointer to PCI device
7643  *
7644  *  Restart the card from scratch, as if from a cold-boot. Implementation
7645  *  resembles the first-half of the igb_resume routine.
7646  **/
7647 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
7648 {
7649 	struct net_device *netdev = pci_get_drvdata(pdev);
7650 	struct igb_adapter *adapter = netdev_priv(netdev);
7651 	struct e1000_hw *hw = &adapter->hw;
7652 	pci_ers_result_t result;
7653 	int err;
7654 
7655 	if (pci_enable_device_mem(pdev)) {
7656 		dev_err(&pdev->dev,
7657 			"Cannot re-enable PCI device after reset.\n");
7658 		result = PCI_ERS_RESULT_DISCONNECT;
7659 	} else {
7660 		pci_set_master(pdev);
7661 		pci_restore_state(pdev);
7662 		pci_save_state(pdev);
7663 
7664 		pci_enable_wake(pdev, PCI_D3hot, 0);
7665 		pci_enable_wake(pdev, PCI_D3cold, 0);
7666 
7667 		igb_reset(adapter);
7668 		wr32(E1000_WUS, ~0);
7669 		result = PCI_ERS_RESULT_RECOVERED;
7670 	}
7671 
7672 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7673 	if (err) {
7674 		dev_err(&pdev->dev,
7675 			"pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7676 			err);
7677 		/* non-fatal, continue */
7678 	}
7679 
7680 	return result;
7681 }
7682 
7683 /**
7684  *  igb_io_resume - called when traffic can start flowing again.
7685  *  @pdev: Pointer to PCI device
7686  *
7687  *  This callback is called when the error recovery driver tells us that
7688  *  its OK to resume normal operation. Implementation resembles the
7689  *  second-half of the igb_resume routine.
7690  */
7691 static void igb_io_resume(struct pci_dev *pdev)
7692 {
7693 	struct net_device *netdev = pci_get_drvdata(pdev);
7694 	struct igb_adapter *adapter = netdev_priv(netdev);
7695 
7696 	if (netif_running(netdev)) {
7697 		if (igb_up(adapter)) {
7698 			dev_err(&pdev->dev, "igb_up failed after reset\n");
7699 			return;
7700 		}
7701 	}
7702 
7703 	netif_device_attach(netdev);
7704 
7705 	/* let the f/w know that the h/w is now under the control of the
7706 	 * driver.
7707 	 */
7708 	igb_get_hw_control(adapter);
7709 }
7710 
7711 static void igb_rar_set_qsel(struct igb_adapter *adapter, u8 *addr, u32 index,
7712 			     u8 qsel)
7713 {
7714 	u32 rar_low, rar_high;
7715 	struct e1000_hw *hw = &adapter->hw;
7716 
7717 	/* HW expects these in little endian so we reverse the byte order
7718 	 * from network order (big endian) to little endian
7719 	 */
7720 	rar_low = ((u32) addr[0] | ((u32) addr[1] << 8) |
7721 		   ((u32) addr[2] << 16) | ((u32) addr[3] << 24));
7722 	rar_high = ((u32) addr[4] | ((u32) addr[5] << 8));
7723 
7724 	/* Indicate to hardware the Address is Valid. */
7725 	rar_high |= E1000_RAH_AV;
7726 
7727 	if (hw->mac.type == e1000_82575)
7728 		rar_high |= E1000_RAH_POOL_1 * qsel;
7729 	else
7730 		rar_high |= E1000_RAH_POOL_1 << qsel;
7731 
7732 	wr32(E1000_RAL(index), rar_low);
7733 	wrfl();
7734 	wr32(E1000_RAH(index), rar_high);
7735 	wrfl();
7736 }
7737 
7738 static int igb_set_vf_mac(struct igb_adapter *adapter,
7739 			  int vf, unsigned char *mac_addr)
7740 {
7741 	struct e1000_hw *hw = &adapter->hw;
7742 	/* VF MAC addresses start at end of receive addresses and moves
7743 	 * towards the first, as a result a collision should not be possible
7744 	 */
7745 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
7746 
7747 	memcpy(adapter->vf_data[vf].vf_mac_addresses, mac_addr, ETH_ALEN);
7748 
7749 	igb_rar_set_qsel(adapter, mac_addr, rar_entry, vf);
7750 
7751 	return 0;
7752 }
7753 
7754 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
7755 {
7756 	struct igb_adapter *adapter = netdev_priv(netdev);
7757 	if (!is_valid_ether_addr(mac) || (vf >= adapter->vfs_allocated_count))
7758 		return -EINVAL;
7759 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
7760 	dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n", mac, vf);
7761 	dev_info(&adapter->pdev->dev,
7762 		 "Reload the VF driver to make this change effective.");
7763 	if (test_bit(__IGB_DOWN, &adapter->state)) {
7764 		dev_warn(&adapter->pdev->dev,
7765 			 "The VF MAC address has been set, but the PF device is not up.\n");
7766 		dev_warn(&adapter->pdev->dev,
7767 			 "Bring the PF device up before attempting to use the VF device.\n");
7768 	}
7769 	return igb_set_vf_mac(adapter, vf, mac);
7770 }
7771 
7772 static int igb_link_mbps(int internal_link_speed)
7773 {
7774 	switch (internal_link_speed) {
7775 	case SPEED_100:
7776 		return 100;
7777 	case SPEED_1000:
7778 		return 1000;
7779 	default:
7780 		return 0;
7781 	}
7782 }
7783 
7784 static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
7785 				  int link_speed)
7786 {
7787 	int rf_dec, rf_int;
7788 	u32 bcnrc_val;
7789 
7790 	if (tx_rate != 0) {
7791 		/* Calculate the rate factor values to set */
7792 		rf_int = link_speed / tx_rate;
7793 		rf_dec = (link_speed - (rf_int * tx_rate));
7794 		rf_dec = (rf_dec * (1 << E1000_RTTBCNRC_RF_INT_SHIFT)) /
7795 			 tx_rate;
7796 
7797 		bcnrc_val = E1000_RTTBCNRC_RS_ENA;
7798 		bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
7799 			      E1000_RTTBCNRC_RF_INT_MASK);
7800 		bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
7801 	} else {
7802 		bcnrc_val = 0;
7803 	}
7804 
7805 	wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
7806 	/* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
7807 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
7808 	 */
7809 	wr32(E1000_RTTBCNRM, 0x14);
7810 	wr32(E1000_RTTBCNRC, bcnrc_val);
7811 }
7812 
7813 static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
7814 {
7815 	int actual_link_speed, i;
7816 	bool reset_rate = false;
7817 
7818 	/* VF TX rate limit was not set or not supported */
7819 	if ((adapter->vf_rate_link_speed == 0) ||
7820 	    (adapter->hw.mac.type != e1000_82576))
7821 		return;
7822 
7823 	actual_link_speed = igb_link_mbps(adapter->link_speed);
7824 	if (actual_link_speed != adapter->vf_rate_link_speed) {
7825 		reset_rate = true;
7826 		adapter->vf_rate_link_speed = 0;
7827 		dev_info(&adapter->pdev->dev,
7828 			 "Link speed has been changed. VF Transmit rate is disabled\n");
7829 	}
7830 
7831 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
7832 		if (reset_rate)
7833 			adapter->vf_data[i].tx_rate = 0;
7834 
7835 		igb_set_vf_rate_limit(&adapter->hw, i,
7836 				      adapter->vf_data[i].tx_rate,
7837 				      actual_link_speed);
7838 	}
7839 }
7840 
7841 static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
7842 			     int min_tx_rate, int max_tx_rate)
7843 {
7844 	struct igb_adapter *adapter = netdev_priv(netdev);
7845 	struct e1000_hw *hw = &adapter->hw;
7846 	int actual_link_speed;
7847 
7848 	if (hw->mac.type != e1000_82576)
7849 		return -EOPNOTSUPP;
7850 
7851 	if (min_tx_rate)
7852 		return -EINVAL;
7853 
7854 	actual_link_speed = igb_link_mbps(adapter->link_speed);
7855 	if ((vf >= adapter->vfs_allocated_count) ||
7856 	    (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
7857 	    (max_tx_rate < 0) ||
7858 	    (max_tx_rate > actual_link_speed))
7859 		return -EINVAL;
7860 
7861 	adapter->vf_rate_link_speed = actual_link_speed;
7862 	adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
7863 	igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
7864 
7865 	return 0;
7866 }
7867 
7868 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
7869 				   bool setting)
7870 {
7871 	struct igb_adapter *adapter = netdev_priv(netdev);
7872 	struct e1000_hw *hw = &adapter->hw;
7873 	u32 reg_val, reg_offset;
7874 
7875 	if (!adapter->vfs_allocated_count)
7876 		return -EOPNOTSUPP;
7877 
7878 	if (vf >= adapter->vfs_allocated_count)
7879 		return -EINVAL;
7880 
7881 	reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
7882 	reg_val = rd32(reg_offset);
7883 	if (setting)
7884 		reg_val |= ((1 << vf) |
7885 			    (1 << (vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT)));
7886 	else
7887 		reg_val &= ~((1 << vf) |
7888 			     (1 << (vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT)));
7889 	wr32(reg_offset, reg_val);
7890 
7891 	adapter->vf_data[vf].spoofchk_enabled = setting;
7892 	return 0;
7893 }
7894 
7895 static int igb_ndo_get_vf_config(struct net_device *netdev,
7896 				 int vf, struct ifla_vf_info *ivi)
7897 {
7898 	struct igb_adapter *adapter = netdev_priv(netdev);
7899 	if (vf >= adapter->vfs_allocated_count)
7900 		return -EINVAL;
7901 	ivi->vf = vf;
7902 	memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
7903 	ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
7904 	ivi->min_tx_rate = 0;
7905 	ivi->vlan = adapter->vf_data[vf].pf_vlan;
7906 	ivi->qos = adapter->vf_data[vf].pf_qos;
7907 	ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
7908 	return 0;
7909 }
7910 
7911 static void igb_vmm_control(struct igb_adapter *adapter)
7912 {
7913 	struct e1000_hw *hw = &adapter->hw;
7914 	u32 reg;
7915 
7916 	switch (hw->mac.type) {
7917 	case e1000_82575:
7918 	case e1000_i210:
7919 	case e1000_i211:
7920 	case e1000_i354:
7921 	default:
7922 		/* replication is not supported for 82575 */
7923 		return;
7924 	case e1000_82576:
7925 		/* notify HW that the MAC is adding vlan tags */
7926 		reg = rd32(E1000_DTXCTL);
7927 		reg |= E1000_DTXCTL_VLAN_ADDED;
7928 		wr32(E1000_DTXCTL, reg);
7929 		/* Fall through */
7930 	case e1000_82580:
7931 		/* enable replication vlan tag stripping */
7932 		reg = rd32(E1000_RPLOLR);
7933 		reg |= E1000_RPLOLR_STRVLAN;
7934 		wr32(E1000_RPLOLR, reg);
7935 		/* Fall through */
7936 	case e1000_i350:
7937 		/* none of the above registers are supported by i350 */
7938 		break;
7939 	}
7940 
7941 	if (adapter->vfs_allocated_count) {
7942 		igb_vmdq_set_loopback_pf(hw, true);
7943 		igb_vmdq_set_replication_pf(hw, true);
7944 		igb_vmdq_set_anti_spoofing_pf(hw, true,
7945 					      adapter->vfs_allocated_count);
7946 	} else {
7947 		igb_vmdq_set_loopback_pf(hw, false);
7948 		igb_vmdq_set_replication_pf(hw, false);
7949 	}
7950 }
7951 
7952 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
7953 {
7954 	struct e1000_hw *hw = &adapter->hw;
7955 	u32 dmac_thr;
7956 	u16 hwm;
7957 
7958 	if (hw->mac.type > e1000_82580) {
7959 		if (adapter->flags & IGB_FLAG_DMAC) {
7960 			u32 reg;
7961 
7962 			/* force threshold to 0. */
7963 			wr32(E1000_DMCTXTH, 0);
7964 
7965 			/* DMA Coalescing high water mark needs to be greater
7966 			 * than the Rx threshold. Set hwm to PBA - max frame
7967 			 * size in 16B units, capping it at PBA - 6KB.
7968 			 */
7969 			hwm = 64 * pba - adapter->max_frame_size / 16;
7970 			if (hwm < 64 * (pba - 6))
7971 				hwm = 64 * (pba - 6);
7972 			reg = rd32(E1000_FCRTC);
7973 			reg &= ~E1000_FCRTC_RTH_COAL_MASK;
7974 			reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
7975 				& E1000_FCRTC_RTH_COAL_MASK);
7976 			wr32(E1000_FCRTC, reg);
7977 
7978 			/* Set the DMA Coalescing Rx threshold to PBA - 2 * max
7979 			 * frame size, capping it at PBA - 10KB.
7980 			 */
7981 			dmac_thr = pba - adapter->max_frame_size / 512;
7982 			if (dmac_thr < pba - 10)
7983 				dmac_thr = pba - 10;
7984 			reg = rd32(E1000_DMACR);
7985 			reg &= ~E1000_DMACR_DMACTHR_MASK;
7986 			reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
7987 				& E1000_DMACR_DMACTHR_MASK);
7988 
7989 			/* transition to L0x or L1 if available..*/
7990 			reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
7991 
7992 			/* watchdog timer= +-1000 usec in 32usec intervals */
7993 			reg |= (1000 >> 5);
7994 
7995 			/* Disable BMC-to-OS Watchdog Enable */
7996 			if (hw->mac.type != e1000_i354)
7997 				reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
7998 
7999 			wr32(E1000_DMACR, reg);
8000 
8001 			/* no lower threshold to disable
8002 			 * coalescing(smart fifb)-UTRESH=0
8003 			 */
8004 			wr32(E1000_DMCRTRH, 0);
8005 
8006 			reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
8007 
8008 			wr32(E1000_DMCTLX, reg);
8009 
8010 			/* free space in tx packet buffer to wake from
8011 			 * DMA coal
8012 			 */
8013 			wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
8014 			     (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
8015 
8016 			/* make low power state decision controlled
8017 			 * by DMA coal
8018 			 */
8019 			reg = rd32(E1000_PCIEMISC);
8020 			reg &= ~E1000_PCIEMISC_LX_DECISION;
8021 			wr32(E1000_PCIEMISC, reg);
8022 		} /* endif adapter->dmac is not disabled */
8023 	} else if (hw->mac.type == e1000_82580) {
8024 		u32 reg = rd32(E1000_PCIEMISC);
8025 
8026 		wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
8027 		wr32(E1000_DMACR, 0);
8028 	}
8029 }
8030 
8031 /**
8032  *  igb_read_i2c_byte - Reads 8 bit word over I2C
8033  *  @hw: pointer to hardware structure
8034  *  @byte_offset: byte offset to read
8035  *  @dev_addr: device address
8036  *  @data: value read
8037  *
8038  *  Performs byte read operation over I2C interface at
8039  *  a specified device address.
8040  **/
8041 s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
8042 		      u8 dev_addr, u8 *data)
8043 {
8044 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
8045 	struct i2c_client *this_client = adapter->i2c_client;
8046 	s32 status;
8047 	u16 swfw_mask = 0;
8048 
8049 	if (!this_client)
8050 		return E1000_ERR_I2C;
8051 
8052 	swfw_mask = E1000_SWFW_PHY0_SM;
8053 
8054 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
8055 		return E1000_ERR_SWFW_SYNC;
8056 
8057 	status = i2c_smbus_read_byte_data(this_client, byte_offset);
8058 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
8059 
8060 	if (status < 0)
8061 		return E1000_ERR_I2C;
8062 	else {
8063 		*data = status;
8064 		return 0;
8065 	}
8066 }
8067 
8068 /**
8069  *  igb_write_i2c_byte - Writes 8 bit word over I2C
8070  *  @hw: pointer to hardware structure
8071  *  @byte_offset: byte offset to write
8072  *  @dev_addr: device address
8073  *  @data: value to write
8074  *
8075  *  Performs byte write operation over I2C interface at
8076  *  a specified device address.
8077  **/
8078 s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
8079 		       u8 dev_addr, u8 data)
8080 {
8081 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
8082 	struct i2c_client *this_client = adapter->i2c_client;
8083 	s32 status;
8084 	u16 swfw_mask = E1000_SWFW_PHY0_SM;
8085 
8086 	if (!this_client)
8087 		return E1000_ERR_I2C;
8088 
8089 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
8090 		return E1000_ERR_SWFW_SYNC;
8091 	status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
8092 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
8093 
8094 	if (status)
8095 		return E1000_ERR_I2C;
8096 	else
8097 		return 0;
8098 
8099 }
8100 
8101 int igb_reinit_queues(struct igb_adapter *adapter)
8102 {
8103 	struct net_device *netdev = adapter->netdev;
8104 	struct pci_dev *pdev = adapter->pdev;
8105 	int err = 0;
8106 
8107 	if (netif_running(netdev))
8108 		igb_close(netdev);
8109 
8110 	igb_reset_interrupt_capability(adapter);
8111 
8112 	if (igb_init_interrupt_scheme(adapter, true)) {
8113 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8114 		return -ENOMEM;
8115 	}
8116 
8117 	if (netif_running(netdev))
8118 		err = igb_open(netdev);
8119 
8120 	return err;
8121 }
8122 /* igb_main.c */
8123