1 /* Intel(R) Gigabit Ethernet Linux driver
2  * Copyright(c) 2007-2014 Intel Corporation.
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms and conditions of the GNU General Public License,
6  * version 2, as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope it will be useful, but WITHOUT
9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
11  * more details.
12  *
13  * You should have received a copy of the GNU General Public License along with
14  * this program; if not, see <http://www.gnu.org/licenses/>.
15  *
16  * The full GNU General Public License is included in this distribution in
17  * the file called "COPYING".
18  *
19  * Contact Information:
20  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
21  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
22  */
23 
24 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
25 
26 #include <linux/module.h>
27 #include <linux/types.h>
28 #include <linux/init.h>
29 #include <linux/bitops.h>
30 #include <linux/vmalloc.h>
31 #include <linux/pagemap.h>
32 #include <linux/netdevice.h>
33 #include <linux/ipv6.h>
34 #include <linux/slab.h>
35 #include <net/checksum.h>
36 #include <net/ip6_checksum.h>
37 #include <linux/net_tstamp.h>
38 #include <linux/mii.h>
39 #include <linux/ethtool.h>
40 #include <linux/if.h>
41 #include <linux/if_vlan.h>
42 #include <linux/pci.h>
43 #include <linux/pci-aspm.h>
44 #include <linux/delay.h>
45 #include <linux/interrupt.h>
46 #include <linux/ip.h>
47 #include <linux/tcp.h>
48 #include <linux/sctp.h>
49 #include <linux/if_ether.h>
50 #include <linux/aer.h>
51 #include <linux/prefetch.h>
52 #include <linux/pm_runtime.h>
53 #ifdef CONFIG_IGB_DCA
54 #include <linux/dca.h>
55 #endif
56 #include <linux/i2c.h>
57 #include "igb.h"
58 
59 #define MAJ 5
60 #define MIN 2
61 #define BUILD 15
62 #define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
63 __stringify(BUILD) "-k"
64 char igb_driver_name[] = "igb";
65 char igb_driver_version[] = DRV_VERSION;
66 static const char igb_driver_string[] =
67 				"Intel(R) Gigabit Ethernet Network Driver";
68 static const char igb_copyright[] =
69 				"Copyright (c) 2007-2014 Intel Corporation.";
70 
71 static const struct e1000_info *igb_info_tbl[] = {
72 	[board_82575] = &e1000_82575_info,
73 };
74 
75 static const struct pci_device_id igb_pci_tbl[] = {
76 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
77 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_SGMII) },
78 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
79 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I211_COPPER), board_82575 },
80 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER), board_82575 },
81 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_FIBER), board_82575 },
82 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES), board_82575 },
83 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SGMII), board_82575 },
84 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER_FLASHLESS), board_82575 },
85 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES_FLASHLESS), board_82575 },
86 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_COPPER), board_82575 },
87 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_FIBER), board_82575 },
88 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SERDES), board_82575 },
89 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SGMII), board_82575 },
90 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER), board_82575 },
91 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_FIBER), board_82575 },
92 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_QUAD_FIBER), board_82575 },
93 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SERDES), board_82575 },
94 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SGMII), board_82575 },
95 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER_DUAL), board_82575 },
96 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SGMII), board_82575 },
97 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SERDES), board_82575 },
98 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_BACKPLANE), board_82575 },
99 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SFP), board_82575 },
100 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
101 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
102 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
103 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
104 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
105 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
106 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER_ET2), board_82575 },
107 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
108 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
109 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
110 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
111 	/* required last entry */
112 	{0, }
113 };
114 
115 MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
116 
117 static int igb_setup_all_tx_resources(struct igb_adapter *);
118 static int igb_setup_all_rx_resources(struct igb_adapter *);
119 static void igb_free_all_tx_resources(struct igb_adapter *);
120 static void igb_free_all_rx_resources(struct igb_adapter *);
121 static void igb_setup_mrqc(struct igb_adapter *);
122 static int igb_probe(struct pci_dev *, const struct pci_device_id *);
123 static void igb_remove(struct pci_dev *pdev);
124 static int igb_sw_init(struct igb_adapter *);
125 static int igb_open(struct net_device *);
126 static int igb_close(struct net_device *);
127 static void igb_configure(struct igb_adapter *);
128 static void igb_configure_tx(struct igb_adapter *);
129 static void igb_configure_rx(struct igb_adapter *);
130 static void igb_clean_all_tx_rings(struct igb_adapter *);
131 static void igb_clean_all_rx_rings(struct igb_adapter *);
132 static void igb_clean_tx_ring(struct igb_ring *);
133 static void igb_clean_rx_ring(struct igb_ring *);
134 static void igb_set_rx_mode(struct net_device *);
135 static void igb_update_phy_info(unsigned long);
136 static void igb_watchdog(unsigned long);
137 static void igb_watchdog_task(struct work_struct *);
138 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb, struct net_device *);
139 static struct rtnl_link_stats64 *igb_get_stats64(struct net_device *dev,
140 					  struct rtnl_link_stats64 *stats);
141 static int igb_change_mtu(struct net_device *, int);
142 static int igb_set_mac(struct net_device *, void *);
143 static void igb_set_uta(struct igb_adapter *adapter);
144 static irqreturn_t igb_intr(int irq, void *);
145 static irqreturn_t igb_intr_msi(int irq, void *);
146 static irqreturn_t igb_msix_other(int irq, void *);
147 static irqreturn_t igb_msix_ring(int irq, void *);
148 #ifdef CONFIG_IGB_DCA
149 static void igb_update_dca(struct igb_q_vector *);
150 static void igb_setup_dca(struct igb_adapter *);
151 #endif /* CONFIG_IGB_DCA */
152 static int igb_poll(struct napi_struct *, int);
153 static bool igb_clean_tx_irq(struct igb_q_vector *);
154 static bool igb_clean_rx_irq(struct igb_q_vector *, int);
155 static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
156 static void igb_tx_timeout(struct net_device *);
157 static void igb_reset_task(struct work_struct *);
158 static void igb_vlan_mode(struct net_device *netdev,
159 			  netdev_features_t features);
160 static int igb_vlan_rx_add_vid(struct net_device *, __be16, u16);
161 static int igb_vlan_rx_kill_vid(struct net_device *, __be16, u16);
162 static void igb_restore_vlan(struct igb_adapter *);
163 static void igb_rar_set_qsel(struct igb_adapter *, u8 *, u32 , u8);
164 static void igb_ping_all_vfs(struct igb_adapter *);
165 static void igb_msg_task(struct igb_adapter *);
166 static void igb_vmm_control(struct igb_adapter *);
167 static int igb_set_vf_mac(struct igb_adapter *, int, unsigned char *);
168 static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
169 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac);
170 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
171 			       int vf, u16 vlan, u8 qos);
172 static int igb_ndo_set_vf_bw(struct net_device *, int, int, int);
173 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
174 				   bool setting);
175 static int igb_ndo_get_vf_config(struct net_device *netdev, int vf,
176 				 struct ifla_vf_info *ivi);
177 static void igb_check_vf_rate_limit(struct igb_adapter *);
178 
179 #ifdef CONFIG_PCI_IOV
180 static int igb_vf_configure(struct igb_adapter *adapter, int vf);
181 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs);
182 #endif
183 
184 #ifdef CONFIG_PM
185 #ifdef CONFIG_PM_SLEEP
186 static int igb_suspend(struct device *);
187 #endif
188 static int igb_resume(struct device *);
189 #ifdef CONFIG_PM_RUNTIME
190 static int igb_runtime_suspend(struct device *dev);
191 static int igb_runtime_resume(struct device *dev);
192 static int igb_runtime_idle(struct device *dev);
193 #endif
194 static const struct dev_pm_ops igb_pm_ops = {
195 	SET_SYSTEM_SLEEP_PM_OPS(igb_suspend, igb_resume)
196 	SET_RUNTIME_PM_OPS(igb_runtime_suspend, igb_runtime_resume,
197 			igb_runtime_idle)
198 };
199 #endif
200 static void igb_shutdown(struct pci_dev *);
201 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs);
202 #ifdef CONFIG_IGB_DCA
203 static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
204 static struct notifier_block dca_notifier = {
205 	.notifier_call	= igb_notify_dca,
206 	.next		= NULL,
207 	.priority	= 0
208 };
209 #endif
210 #ifdef CONFIG_NET_POLL_CONTROLLER
211 /* for netdump / net console */
212 static void igb_netpoll(struct net_device *);
213 #endif
214 #ifdef CONFIG_PCI_IOV
215 static unsigned int max_vfs;
216 module_param(max_vfs, uint, 0);
217 MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate per physical function");
218 #endif /* CONFIG_PCI_IOV */
219 
220 static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
221 		     pci_channel_state_t);
222 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
223 static void igb_io_resume(struct pci_dev *);
224 
225 static const struct pci_error_handlers igb_err_handler = {
226 	.error_detected = igb_io_error_detected,
227 	.slot_reset = igb_io_slot_reset,
228 	.resume = igb_io_resume,
229 };
230 
231 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba);
232 
233 static struct pci_driver igb_driver = {
234 	.name     = igb_driver_name,
235 	.id_table = igb_pci_tbl,
236 	.probe    = igb_probe,
237 	.remove   = igb_remove,
238 #ifdef CONFIG_PM
239 	.driver.pm = &igb_pm_ops,
240 #endif
241 	.shutdown = igb_shutdown,
242 	.sriov_configure = igb_pci_sriov_configure,
243 	.err_handler = &igb_err_handler
244 };
245 
246 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
247 MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
248 MODULE_LICENSE("GPL");
249 MODULE_VERSION(DRV_VERSION);
250 
251 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
252 static int debug = -1;
253 module_param(debug, int, 0);
254 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
255 
256 struct igb_reg_info {
257 	u32 ofs;
258 	char *name;
259 };
260 
261 static const struct igb_reg_info igb_reg_info_tbl[] = {
262 
263 	/* General Registers */
264 	{E1000_CTRL, "CTRL"},
265 	{E1000_STATUS, "STATUS"},
266 	{E1000_CTRL_EXT, "CTRL_EXT"},
267 
268 	/* Interrupt Registers */
269 	{E1000_ICR, "ICR"},
270 
271 	/* RX Registers */
272 	{E1000_RCTL, "RCTL"},
273 	{E1000_RDLEN(0), "RDLEN"},
274 	{E1000_RDH(0), "RDH"},
275 	{E1000_RDT(0), "RDT"},
276 	{E1000_RXDCTL(0), "RXDCTL"},
277 	{E1000_RDBAL(0), "RDBAL"},
278 	{E1000_RDBAH(0), "RDBAH"},
279 
280 	/* TX Registers */
281 	{E1000_TCTL, "TCTL"},
282 	{E1000_TDBAL(0), "TDBAL"},
283 	{E1000_TDBAH(0), "TDBAH"},
284 	{E1000_TDLEN(0), "TDLEN"},
285 	{E1000_TDH(0), "TDH"},
286 	{E1000_TDT(0), "TDT"},
287 	{E1000_TXDCTL(0), "TXDCTL"},
288 	{E1000_TDFH, "TDFH"},
289 	{E1000_TDFT, "TDFT"},
290 	{E1000_TDFHS, "TDFHS"},
291 	{E1000_TDFPC, "TDFPC"},
292 
293 	/* List Terminator */
294 	{}
295 };
296 
297 /* igb_regdump - register printout routine */
298 static void igb_regdump(struct e1000_hw *hw, struct igb_reg_info *reginfo)
299 {
300 	int n = 0;
301 	char rname[16];
302 	u32 regs[8];
303 
304 	switch (reginfo->ofs) {
305 	case E1000_RDLEN(0):
306 		for (n = 0; n < 4; n++)
307 			regs[n] = rd32(E1000_RDLEN(n));
308 		break;
309 	case E1000_RDH(0):
310 		for (n = 0; n < 4; n++)
311 			regs[n] = rd32(E1000_RDH(n));
312 		break;
313 	case E1000_RDT(0):
314 		for (n = 0; n < 4; n++)
315 			regs[n] = rd32(E1000_RDT(n));
316 		break;
317 	case E1000_RXDCTL(0):
318 		for (n = 0; n < 4; n++)
319 			regs[n] = rd32(E1000_RXDCTL(n));
320 		break;
321 	case E1000_RDBAL(0):
322 		for (n = 0; n < 4; n++)
323 			regs[n] = rd32(E1000_RDBAL(n));
324 		break;
325 	case E1000_RDBAH(0):
326 		for (n = 0; n < 4; n++)
327 			regs[n] = rd32(E1000_RDBAH(n));
328 		break;
329 	case E1000_TDBAL(0):
330 		for (n = 0; n < 4; n++)
331 			regs[n] = rd32(E1000_RDBAL(n));
332 		break;
333 	case E1000_TDBAH(0):
334 		for (n = 0; n < 4; n++)
335 			regs[n] = rd32(E1000_TDBAH(n));
336 		break;
337 	case E1000_TDLEN(0):
338 		for (n = 0; n < 4; n++)
339 			regs[n] = rd32(E1000_TDLEN(n));
340 		break;
341 	case E1000_TDH(0):
342 		for (n = 0; n < 4; n++)
343 			regs[n] = rd32(E1000_TDH(n));
344 		break;
345 	case E1000_TDT(0):
346 		for (n = 0; n < 4; n++)
347 			regs[n] = rd32(E1000_TDT(n));
348 		break;
349 	case E1000_TXDCTL(0):
350 		for (n = 0; n < 4; n++)
351 			regs[n] = rd32(E1000_TXDCTL(n));
352 		break;
353 	default:
354 		pr_info("%-15s %08x\n", reginfo->name, rd32(reginfo->ofs));
355 		return;
356 	}
357 
358 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-3]");
359 	pr_info("%-15s %08x %08x %08x %08x\n", rname, regs[0], regs[1],
360 		regs[2], regs[3]);
361 }
362 
363 /* igb_dump - Print registers, Tx-rings and Rx-rings */
364 static void igb_dump(struct igb_adapter *adapter)
365 {
366 	struct net_device *netdev = adapter->netdev;
367 	struct e1000_hw *hw = &adapter->hw;
368 	struct igb_reg_info *reginfo;
369 	struct igb_ring *tx_ring;
370 	union e1000_adv_tx_desc *tx_desc;
371 	struct my_u0 { u64 a; u64 b; } *u0;
372 	struct igb_ring *rx_ring;
373 	union e1000_adv_rx_desc *rx_desc;
374 	u32 staterr;
375 	u16 i, n;
376 
377 	if (!netif_msg_hw(adapter))
378 		return;
379 
380 	/* Print netdevice Info */
381 	if (netdev) {
382 		dev_info(&adapter->pdev->dev, "Net device Info\n");
383 		pr_info("Device Name     state            trans_start      last_rx\n");
384 		pr_info("%-15s %016lX %016lX %016lX\n", netdev->name,
385 			netdev->state, netdev->trans_start, netdev->last_rx);
386 	}
387 
388 	/* Print Registers */
389 	dev_info(&adapter->pdev->dev, "Register Dump\n");
390 	pr_info(" Register Name   Value\n");
391 	for (reginfo = (struct igb_reg_info *)igb_reg_info_tbl;
392 	     reginfo->name; reginfo++) {
393 		igb_regdump(hw, reginfo);
394 	}
395 
396 	/* Print TX Ring Summary */
397 	if (!netdev || !netif_running(netdev))
398 		goto exit;
399 
400 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
401 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
402 	for (n = 0; n < adapter->num_tx_queues; n++) {
403 		struct igb_tx_buffer *buffer_info;
404 		tx_ring = adapter->tx_ring[n];
405 		buffer_info = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
406 		pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
407 			n, tx_ring->next_to_use, tx_ring->next_to_clean,
408 			(u64)dma_unmap_addr(buffer_info, dma),
409 			dma_unmap_len(buffer_info, len),
410 			buffer_info->next_to_watch,
411 			(u64)buffer_info->time_stamp);
412 	}
413 
414 	/* Print TX Rings */
415 	if (!netif_msg_tx_done(adapter))
416 		goto rx_ring_summary;
417 
418 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
419 
420 	/* Transmit Descriptor Formats
421 	 *
422 	 * Advanced Transmit Descriptor
423 	 *   +--------------------------------------------------------------+
424 	 * 0 |         Buffer Address [63:0]                                |
425 	 *   +--------------------------------------------------------------+
426 	 * 8 | PAYLEN  | PORTS  |CC|IDX | STA | DCMD  |DTYP|MAC|RSV| DTALEN |
427 	 *   +--------------------------------------------------------------+
428 	 *   63      46 45    40 39 38 36 35 32 31   24             15       0
429 	 */
430 
431 	for (n = 0; n < adapter->num_tx_queues; n++) {
432 		tx_ring = adapter->tx_ring[n];
433 		pr_info("------------------------------------\n");
434 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
435 		pr_info("------------------------------------\n");
436 		pr_info("T [desc]     [address 63:0  ] [PlPOCIStDDM Ln] [bi->dma       ] leng  ntw timestamp        bi->skb\n");
437 
438 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
439 			const char *next_desc;
440 			struct igb_tx_buffer *buffer_info;
441 			tx_desc = IGB_TX_DESC(tx_ring, i);
442 			buffer_info = &tx_ring->tx_buffer_info[i];
443 			u0 = (struct my_u0 *)tx_desc;
444 			if (i == tx_ring->next_to_use &&
445 			    i == tx_ring->next_to_clean)
446 				next_desc = " NTC/U";
447 			else if (i == tx_ring->next_to_use)
448 				next_desc = " NTU";
449 			else if (i == tx_ring->next_to_clean)
450 				next_desc = " NTC";
451 			else
452 				next_desc = "";
453 
454 			pr_info("T [0x%03X]    %016llX %016llX %016llX %04X  %p %016llX %p%s\n",
455 				i, le64_to_cpu(u0->a),
456 				le64_to_cpu(u0->b),
457 				(u64)dma_unmap_addr(buffer_info, dma),
458 				dma_unmap_len(buffer_info, len),
459 				buffer_info->next_to_watch,
460 				(u64)buffer_info->time_stamp,
461 				buffer_info->skb, next_desc);
462 
463 			if (netif_msg_pktdata(adapter) && buffer_info->skb)
464 				print_hex_dump(KERN_INFO, "",
465 					DUMP_PREFIX_ADDRESS,
466 					16, 1, buffer_info->skb->data,
467 					dma_unmap_len(buffer_info, len),
468 					true);
469 		}
470 	}
471 
472 	/* Print RX Rings Summary */
473 rx_ring_summary:
474 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
475 	pr_info("Queue [NTU] [NTC]\n");
476 	for (n = 0; n < adapter->num_rx_queues; n++) {
477 		rx_ring = adapter->rx_ring[n];
478 		pr_info(" %5d %5X %5X\n",
479 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
480 	}
481 
482 	/* Print RX Rings */
483 	if (!netif_msg_rx_status(adapter))
484 		goto exit;
485 
486 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
487 
488 	/* Advanced Receive Descriptor (Read) Format
489 	 *    63                                           1        0
490 	 *    +-----------------------------------------------------+
491 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
492 	 *    +----------------------------------------------+------+
493 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
494 	 *    +-----------------------------------------------------+
495 	 *
496 	 *
497 	 * Advanced Receive Descriptor (Write-Back) Format
498 	 *
499 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
500 	 *   +------------------------------------------------------+
501 	 * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
502 	 *   | Checksum   Ident  |   |           |    | Type | Type |
503 	 *   +------------------------------------------------------+
504 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
505 	 *   +------------------------------------------------------+
506 	 *   63       48 47    32 31            20 19               0
507 	 */
508 
509 	for (n = 0; n < adapter->num_rx_queues; n++) {
510 		rx_ring = adapter->rx_ring[n];
511 		pr_info("------------------------------------\n");
512 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
513 		pr_info("------------------------------------\n");
514 		pr_info("R  [desc]      [ PktBuf     A0] [  HeadBuf   DD] [bi->dma       ] [bi->skb] <-- Adv Rx Read format\n");
515 		pr_info("RWB[desc]      [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
516 
517 		for (i = 0; i < rx_ring->count; i++) {
518 			const char *next_desc;
519 			struct igb_rx_buffer *buffer_info;
520 			buffer_info = &rx_ring->rx_buffer_info[i];
521 			rx_desc = IGB_RX_DESC(rx_ring, i);
522 			u0 = (struct my_u0 *)rx_desc;
523 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
524 
525 			if (i == rx_ring->next_to_use)
526 				next_desc = " NTU";
527 			else if (i == rx_ring->next_to_clean)
528 				next_desc = " NTC";
529 			else
530 				next_desc = "";
531 
532 			if (staterr & E1000_RXD_STAT_DD) {
533 				/* Descriptor Done */
534 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %s\n",
535 					"RWB", i,
536 					le64_to_cpu(u0->a),
537 					le64_to_cpu(u0->b),
538 					next_desc);
539 			} else {
540 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %s\n",
541 					"R  ", i,
542 					le64_to_cpu(u0->a),
543 					le64_to_cpu(u0->b),
544 					(u64)buffer_info->dma,
545 					next_desc);
546 
547 				if (netif_msg_pktdata(adapter) &&
548 				    buffer_info->dma && buffer_info->page) {
549 					print_hex_dump(KERN_INFO, "",
550 					  DUMP_PREFIX_ADDRESS,
551 					  16, 1,
552 					  page_address(buffer_info->page) +
553 						      buffer_info->page_offset,
554 					  IGB_RX_BUFSZ, true);
555 				}
556 			}
557 		}
558 	}
559 
560 exit:
561 	return;
562 }
563 
564 /**
565  *  igb_get_i2c_data - Reads the I2C SDA data bit
566  *  @hw: pointer to hardware structure
567  *  @i2cctl: Current value of I2CCTL register
568  *
569  *  Returns the I2C data bit value
570  **/
571 static int igb_get_i2c_data(void *data)
572 {
573 	struct igb_adapter *adapter = (struct igb_adapter *)data;
574 	struct e1000_hw *hw = &adapter->hw;
575 	s32 i2cctl = rd32(E1000_I2CPARAMS);
576 
577 	return !!(i2cctl & E1000_I2C_DATA_IN);
578 }
579 
580 /**
581  *  igb_set_i2c_data - Sets the I2C data bit
582  *  @data: pointer to hardware structure
583  *  @state: I2C data value (0 or 1) to set
584  *
585  *  Sets the I2C data bit
586  **/
587 static void igb_set_i2c_data(void *data, int state)
588 {
589 	struct igb_adapter *adapter = (struct igb_adapter *)data;
590 	struct e1000_hw *hw = &adapter->hw;
591 	s32 i2cctl = rd32(E1000_I2CPARAMS);
592 
593 	if (state)
594 		i2cctl |= E1000_I2C_DATA_OUT;
595 	else
596 		i2cctl &= ~E1000_I2C_DATA_OUT;
597 
598 	i2cctl &= ~E1000_I2C_DATA_OE_N;
599 	i2cctl |= E1000_I2C_CLK_OE_N;
600 	wr32(E1000_I2CPARAMS, i2cctl);
601 	wrfl();
602 
603 }
604 
605 /**
606  *  igb_set_i2c_clk - Sets the I2C SCL clock
607  *  @data: pointer to hardware structure
608  *  @state: state to set clock
609  *
610  *  Sets the I2C clock line to state
611  **/
612 static void igb_set_i2c_clk(void *data, int state)
613 {
614 	struct igb_adapter *adapter = (struct igb_adapter *)data;
615 	struct e1000_hw *hw = &adapter->hw;
616 	s32 i2cctl = rd32(E1000_I2CPARAMS);
617 
618 	if (state) {
619 		i2cctl |= E1000_I2C_CLK_OUT;
620 		i2cctl &= ~E1000_I2C_CLK_OE_N;
621 	} else {
622 		i2cctl &= ~E1000_I2C_CLK_OUT;
623 		i2cctl &= ~E1000_I2C_CLK_OE_N;
624 	}
625 	wr32(E1000_I2CPARAMS, i2cctl);
626 	wrfl();
627 }
628 
629 /**
630  *  igb_get_i2c_clk - Gets the I2C SCL clock state
631  *  @data: pointer to hardware structure
632  *
633  *  Gets the I2C clock state
634  **/
635 static int igb_get_i2c_clk(void *data)
636 {
637 	struct igb_adapter *adapter = (struct igb_adapter *)data;
638 	struct e1000_hw *hw = &adapter->hw;
639 	s32 i2cctl = rd32(E1000_I2CPARAMS);
640 
641 	return !!(i2cctl & E1000_I2C_CLK_IN);
642 }
643 
644 static const struct i2c_algo_bit_data igb_i2c_algo = {
645 	.setsda		= igb_set_i2c_data,
646 	.setscl		= igb_set_i2c_clk,
647 	.getsda		= igb_get_i2c_data,
648 	.getscl		= igb_get_i2c_clk,
649 	.udelay		= 5,
650 	.timeout	= 20,
651 };
652 
653 /**
654  *  igb_get_hw_dev - return device
655  *  @hw: pointer to hardware structure
656  *
657  *  used by hardware layer to print debugging information
658  **/
659 struct net_device *igb_get_hw_dev(struct e1000_hw *hw)
660 {
661 	struct igb_adapter *adapter = hw->back;
662 	return adapter->netdev;
663 }
664 
665 /**
666  *  igb_init_module - Driver Registration Routine
667  *
668  *  igb_init_module is the first routine called when the driver is
669  *  loaded. All it does is register with the PCI subsystem.
670  **/
671 static int __init igb_init_module(void)
672 {
673 	int ret;
674 
675 	pr_info("%s - version %s\n",
676 	       igb_driver_string, igb_driver_version);
677 	pr_info("%s\n", igb_copyright);
678 
679 #ifdef CONFIG_IGB_DCA
680 	dca_register_notify(&dca_notifier);
681 #endif
682 	ret = pci_register_driver(&igb_driver);
683 	return ret;
684 }
685 
686 module_init(igb_init_module);
687 
688 /**
689  *  igb_exit_module - Driver Exit Cleanup Routine
690  *
691  *  igb_exit_module is called just before the driver is removed
692  *  from memory.
693  **/
694 static void __exit igb_exit_module(void)
695 {
696 #ifdef CONFIG_IGB_DCA
697 	dca_unregister_notify(&dca_notifier);
698 #endif
699 	pci_unregister_driver(&igb_driver);
700 }
701 
702 module_exit(igb_exit_module);
703 
704 #define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
705 /**
706  *  igb_cache_ring_register - Descriptor ring to register mapping
707  *  @adapter: board private structure to initialize
708  *
709  *  Once we know the feature-set enabled for the device, we'll cache
710  *  the register offset the descriptor ring is assigned to.
711  **/
712 static void igb_cache_ring_register(struct igb_adapter *adapter)
713 {
714 	int i = 0, j = 0;
715 	u32 rbase_offset = adapter->vfs_allocated_count;
716 
717 	switch (adapter->hw.mac.type) {
718 	case e1000_82576:
719 		/* The queues are allocated for virtualization such that VF 0
720 		 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
721 		 * In order to avoid collision we start at the first free queue
722 		 * and continue consuming queues in the same sequence
723 		 */
724 		if (adapter->vfs_allocated_count) {
725 			for (; i < adapter->rss_queues; i++)
726 				adapter->rx_ring[i]->reg_idx = rbase_offset +
727 							       Q_IDX_82576(i);
728 		}
729 		/* Fall through */
730 	case e1000_82575:
731 	case e1000_82580:
732 	case e1000_i350:
733 	case e1000_i354:
734 	case e1000_i210:
735 	case e1000_i211:
736 		/* Fall through */
737 	default:
738 		for (; i < adapter->num_rx_queues; i++)
739 			adapter->rx_ring[i]->reg_idx = rbase_offset + i;
740 		for (; j < adapter->num_tx_queues; j++)
741 			adapter->tx_ring[j]->reg_idx = rbase_offset + j;
742 		break;
743 	}
744 }
745 
746 u32 igb_rd32(struct e1000_hw *hw, u32 reg)
747 {
748 	struct igb_adapter *igb = container_of(hw, struct igb_adapter, hw);
749 	u8 __iomem *hw_addr = ACCESS_ONCE(hw->hw_addr);
750 	u32 value = 0;
751 
752 	if (E1000_REMOVED(hw_addr))
753 		return ~value;
754 
755 	value = readl(&hw_addr[reg]);
756 
757 	/* reads should not return all F's */
758 	if (!(~value) && (!reg || !(~readl(hw_addr)))) {
759 		struct net_device *netdev = igb->netdev;
760 		hw->hw_addr = NULL;
761 		netif_device_detach(netdev);
762 		netdev_err(netdev, "PCIe link lost, device now detached\n");
763 	}
764 
765 	return value;
766 }
767 
768 /**
769  *  igb_write_ivar - configure ivar for given MSI-X vector
770  *  @hw: pointer to the HW structure
771  *  @msix_vector: vector number we are allocating to a given ring
772  *  @index: row index of IVAR register to write within IVAR table
773  *  @offset: column offset of in IVAR, should be multiple of 8
774  *
775  *  This function is intended to handle the writing of the IVAR register
776  *  for adapters 82576 and newer.  The IVAR table consists of 2 columns,
777  *  each containing an cause allocation for an Rx and Tx ring, and a
778  *  variable number of rows depending on the number of queues supported.
779  **/
780 static void igb_write_ivar(struct e1000_hw *hw, int msix_vector,
781 			   int index, int offset)
782 {
783 	u32 ivar = array_rd32(E1000_IVAR0, index);
784 
785 	/* clear any bits that are currently set */
786 	ivar &= ~((u32)0xFF << offset);
787 
788 	/* write vector and valid bit */
789 	ivar |= (msix_vector | E1000_IVAR_VALID) << offset;
790 
791 	array_wr32(E1000_IVAR0, index, ivar);
792 }
793 
794 #define IGB_N0_QUEUE -1
795 static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
796 {
797 	struct igb_adapter *adapter = q_vector->adapter;
798 	struct e1000_hw *hw = &adapter->hw;
799 	int rx_queue = IGB_N0_QUEUE;
800 	int tx_queue = IGB_N0_QUEUE;
801 	u32 msixbm = 0;
802 
803 	if (q_vector->rx.ring)
804 		rx_queue = q_vector->rx.ring->reg_idx;
805 	if (q_vector->tx.ring)
806 		tx_queue = q_vector->tx.ring->reg_idx;
807 
808 	switch (hw->mac.type) {
809 	case e1000_82575:
810 		/* The 82575 assigns vectors using a bitmask, which matches the
811 		 * bitmask for the EICR/EIMS/EIMC registers.  To assign one
812 		 * or more queues to a vector, we write the appropriate bits
813 		 * into the MSIXBM register for that vector.
814 		 */
815 		if (rx_queue > IGB_N0_QUEUE)
816 			msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
817 		if (tx_queue > IGB_N0_QUEUE)
818 			msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
819 		if (!(adapter->flags & IGB_FLAG_HAS_MSIX) && msix_vector == 0)
820 			msixbm |= E1000_EIMS_OTHER;
821 		array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
822 		q_vector->eims_value = msixbm;
823 		break;
824 	case e1000_82576:
825 		/* 82576 uses a table that essentially consists of 2 columns
826 		 * with 8 rows.  The ordering is column-major so we use the
827 		 * lower 3 bits as the row index, and the 4th bit as the
828 		 * column offset.
829 		 */
830 		if (rx_queue > IGB_N0_QUEUE)
831 			igb_write_ivar(hw, msix_vector,
832 				       rx_queue & 0x7,
833 				       (rx_queue & 0x8) << 1);
834 		if (tx_queue > IGB_N0_QUEUE)
835 			igb_write_ivar(hw, msix_vector,
836 				       tx_queue & 0x7,
837 				       ((tx_queue & 0x8) << 1) + 8);
838 		q_vector->eims_value = 1 << msix_vector;
839 		break;
840 	case e1000_82580:
841 	case e1000_i350:
842 	case e1000_i354:
843 	case e1000_i210:
844 	case e1000_i211:
845 		/* On 82580 and newer adapters the scheme is similar to 82576
846 		 * however instead of ordering column-major we have things
847 		 * ordered row-major.  So we traverse the table by using
848 		 * bit 0 as the column offset, and the remaining bits as the
849 		 * row index.
850 		 */
851 		if (rx_queue > IGB_N0_QUEUE)
852 			igb_write_ivar(hw, msix_vector,
853 				       rx_queue >> 1,
854 				       (rx_queue & 0x1) << 4);
855 		if (tx_queue > IGB_N0_QUEUE)
856 			igb_write_ivar(hw, msix_vector,
857 				       tx_queue >> 1,
858 				       ((tx_queue & 0x1) << 4) + 8);
859 		q_vector->eims_value = 1 << msix_vector;
860 		break;
861 	default:
862 		BUG();
863 		break;
864 	}
865 
866 	/* add q_vector eims value to global eims_enable_mask */
867 	adapter->eims_enable_mask |= q_vector->eims_value;
868 
869 	/* configure q_vector to set itr on first interrupt */
870 	q_vector->set_itr = 1;
871 }
872 
873 /**
874  *  igb_configure_msix - Configure MSI-X hardware
875  *  @adapter: board private structure to initialize
876  *
877  *  igb_configure_msix sets up the hardware to properly
878  *  generate MSI-X interrupts.
879  **/
880 static void igb_configure_msix(struct igb_adapter *adapter)
881 {
882 	u32 tmp;
883 	int i, vector = 0;
884 	struct e1000_hw *hw = &adapter->hw;
885 
886 	adapter->eims_enable_mask = 0;
887 
888 	/* set vector for other causes, i.e. link changes */
889 	switch (hw->mac.type) {
890 	case e1000_82575:
891 		tmp = rd32(E1000_CTRL_EXT);
892 		/* enable MSI-X PBA support*/
893 		tmp |= E1000_CTRL_EXT_PBA_CLR;
894 
895 		/* Auto-Mask interrupts upon ICR read. */
896 		tmp |= E1000_CTRL_EXT_EIAME;
897 		tmp |= E1000_CTRL_EXT_IRCA;
898 
899 		wr32(E1000_CTRL_EXT, tmp);
900 
901 		/* enable msix_other interrupt */
902 		array_wr32(E1000_MSIXBM(0), vector++, E1000_EIMS_OTHER);
903 		adapter->eims_other = E1000_EIMS_OTHER;
904 
905 		break;
906 
907 	case e1000_82576:
908 	case e1000_82580:
909 	case e1000_i350:
910 	case e1000_i354:
911 	case e1000_i210:
912 	case e1000_i211:
913 		/* Turn on MSI-X capability first, or our settings
914 		 * won't stick.  And it will take days to debug.
915 		 */
916 		wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
917 		     E1000_GPIE_PBA | E1000_GPIE_EIAME |
918 		     E1000_GPIE_NSICR);
919 
920 		/* enable msix_other interrupt */
921 		adapter->eims_other = 1 << vector;
922 		tmp = (vector++ | E1000_IVAR_VALID) << 8;
923 
924 		wr32(E1000_IVAR_MISC, tmp);
925 		break;
926 	default:
927 		/* do nothing, since nothing else supports MSI-X */
928 		break;
929 	} /* switch (hw->mac.type) */
930 
931 	adapter->eims_enable_mask |= adapter->eims_other;
932 
933 	for (i = 0; i < adapter->num_q_vectors; i++)
934 		igb_assign_vector(adapter->q_vector[i], vector++);
935 
936 	wrfl();
937 }
938 
939 /**
940  *  igb_request_msix - Initialize MSI-X interrupts
941  *  @adapter: board private structure to initialize
942  *
943  *  igb_request_msix allocates MSI-X vectors and requests interrupts from the
944  *  kernel.
945  **/
946 static int igb_request_msix(struct igb_adapter *adapter)
947 {
948 	struct net_device *netdev = adapter->netdev;
949 	struct e1000_hw *hw = &adapter->hw;
950 	int i, err = 0, vector = 0, free_vector = 0;
951 
952 	err = request_irq(adapter->msix_entries[vector].vector,
953 			  igb_msix_other, 0, netdev->name, adapter);
954 	if (err)
955 		goto err_out;
956 
957 	for (i = 0; i < adapter->num_q_vectors; i++) {
958 		struct igb_q_vector *q_vector = adapter->q_vector[i];
959 
960 		vector++;
961 
962 		q_vector->itr_register = hw->hw_addr + E1000_EITR(vector);
963 
964 		if (q_vector->rx.ring && q_vector->tx.ring)
965 			sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
966 				q_vector->rx.ring->queue_index);
967 		else if (q_vector->tx.ring)
968 			sprintf(q_vector->name, "%s-tx-%u", netdev->name,
969 				q_vector->tx.ring->queue_index);
970 		else if (q_vector->rx.ring)
971 			sprintf(q_vector->name, "%s-rx-%u", netdev->name,
972 				q_vector->rx.ring->queue_index);
973 		else
974 			sprintf(q_vector->name, "%s-unused", netdev->name);
975 
976 		err = request_irq(adapter->msix_entries[vector].vector,
977 				  igb_msix_ring, 0, q_vector->name,
978 				  q_vector);
979 		if (err)
980 			goto err_free;
981 	}
982 
983 	igb_configure_msix(adapter);
984 	return 0;
985 
986 err_free:
987 	/* free already assigned IRQs */
988 	free_irq(adapter->msix_entries[free_vector++].vector, adapter);
989 
990 	vector--;
991 	for (i = 0; i < vector; i++) {
992 		free_irq(adapter->msix_entries[free_vector++].vector,
993 			 adapter->q_vector[i]);
994 	}
995 err_out:
996 	return err;
997 }
998 
999 /**
1000  *  igb_free_q_vector - Free memory allocated for specific interrupt vector
1001  *  @adapter: board private structure to initialize
1002  *  @v_idx: Index of vector to be freed
1003  *
1004  *  This function frees the memory allocated to the q_vector.
1005  **/
1006 static void igb_free_q_vector(struct igb_adapter *adapter, int v_idx)
1007 {
1008 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1009 
1010 	adapter->q_vector[v_idx] = NULL;
1011 
1012 	/* igb_get_stats64() might access the rings on this vector,
1013 	 * we must wait a grace period before freeing it.
1014 	 */
1015 	if (q_vector)
1016 		kfree_rcu(q_vector, rcu);
1017 }
1018 
1019 /**
1020  *  igb_reset_q_vector - Reset config for interrupt vector
1021  *  @adapter: board private structure to initialize
1022  *  @v_idx: Index of vector to be reset
1023  *
1024  *  If NAPI is enabled it will delete any references to the
1025  *  NAPI struct. This is preparation for igb_free_q_vector.
1026  **/
1027 static void igb_reset_q_vector(struct igb_adapter *adapter, int v_idx)
1028 {
1029 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1030 
1031 	/* Coming from igb_set_interrupt_capability, the vectors are not yet
1032 	 * allocated. So, q_vector is NULL so we should stop here.
1033 	 */
1034 	if (!q_vector)
1035 		return;
1036 
1037 	if (q_vector->tx.ring)
1038 		adapter->tx_ring[q_vector->tx.ring->queue_index] = NULL;
1039 
1040 	if (q_vector->rx.ring)
1041 		adapter->tx_ring[q_vector->rx.ring->queue_index] = NULL;
1042 
1043 	netif_napi_del(&q_vector->napi);
1044 
1045 }
1046 
1047 static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
1048 {
1049 	int v_idx = adapter->num_q_vectors;
1050 
1051 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
1052 		pci_disable_msix(adapter->pdev);
1053 	else if (adapter->flags & IGB_FLAG_HAS_MSI)
1054 		pci_disable_msi(adapter->pdev);
1055 
1056 	while (v_idx--)
1057 		igb_reset_q_vector(adapter, v_idx);
1058 }
1059 
1060 /**
1061  *  igb_free_q_vectors - Free memory allocated for interrupt vectors
1062  *  @adapter: board private structure to initialize
1063  *
1064  *  This function frees the memory allocated to the q_vectors.  In addition if
1065  *  NAPI is enabled it will delete any references to the NAPI struct prior
1066  *  to freeing the q_vector.
1067  **/
1068 static void igb_free_q_vectors(struct igb_adapter *adapter)
1069 {
1070 	int v_idx = adapter->num_q_vectors;
1071 
1072 	adapter->num_tx_queues = 0;
1073 	adapter->num_rx_queues = 0;
1074 	adapter->num_q_vectors = 0;
1075 
1076 	while (v_idx--) {
1077 		igb_reset_q_vector(adapter, v_idx);
1078 		igb_free_q_vector(adapter, v_idx);
1079 	}
1080 }
1081 
1082 /**
1083  *  igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1084  *  @adapter: board private structure to initialize
1085  *
1086  *  This function resets the device so that it has 0 Rx queues, Tx queues, and
1087  *  MSI-X interrupts allocated.
1088  */
1089 static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
1090 {
1091 	igb_free_q_vectors(adapter);
1092 	igb_reset_interrupt_capability(adapter);
1093 }
1094 
1095 /**
1096  *  igb_set_interrupt_capability - set MSI or MSI-X if supported
1097  *  @adapter: board private structure to initialize
1098  *  @msix: boolean value of MSIX capability
1099  *
1100  *  Attempt to configure interrupts using the best available
1101  *  capabilities of the hardware and kernel.
1102  **/
1103 static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
1104 {
1105 	int err;
1106 	int numvecs, i;
1107 
1108 	if (!msix)
1109 		goto msi_only;
1110 	adapter->flags |= IGB_FLAG_HAS_MSIX;
1111 
1112 	/* Number of supported queues. */
1113 	adapter->num_rx_queues = adapter->rss_queues;
1114 	if (adapter->vfs_allocated_count)
1115 		adapter->num_tx_queues = 1;
1116 	else
1117 		adapter->num_tx_queues = adapter->rss_queues;
1118 
1119 	/* start with one vector for every Rx queue */
1120 	numvecs = adapter->num_rx_queues;
1121 
1122 	/* if Tx handler is separate add 1 for every Tx queue */
1123 	if (!(adapter->flags & IGB_FLAG_QUEUE_PAIRS))
1124 		numvecs += adapter->num_tx_queues;
1125 
1126 	/* store the number of vectors reserved for queues */
1127 	adapter->num_q_vectors = numvecs;
1128 
1129 	/* add 1 vector for link status interrupts */
1130 	numvecs++;
1131 	for (i = 0; i < numvecs; i++)
1132 		adapter->msix_entries[i].entry = i;
1133 
1134 	err = pci_enable_msix_range(adapter->pdev,
1135 				    adapter->msix_entries,
1136 				    numvecs,
1137 				    numvecs);
1138 	if (err > 0)
1139 		return;
1140 
1141 	igb_reset_interrupt_capability(adapter);
1142 
1143 	/* If we can't do MSI-X, try MSI */
1144 msi_only:
1145 	adapter->flags &= ~IGB_FLAG_HAS_MSIX;
1146 #ifdef CONFIG_PCI_IOV
1147 	/* disable SR-IOV for non MSI-X configurations */
1148 	if (adapter->vf_data) {
1149 		struct e1000_hw *hw = &adapter->hw;
1150 		/* disable iov and allow time for transactions to clear */
1151 		pci_disable_sriov(adapter->pdev);
1152 		msleep(500);
1153 
1154 		kfree(adapter->vf_data);
1155 		adapter->vf_data = NULL;
1156 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1157 		wrfl();
1158 		msleep(100);
1159 		dev_info(&adapter->pdev->dev, "IOV Disabled\n");
1160 	}
1161 #endif
1162 	adapter->vfs_allocated_count = 0;
1163 	adapter->rss_queues = 1;
1164 	adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
1165 	adapter->num_rx_queues = 1;
1166 	adapter->num_tx_queues = 1;
1167 	adapter->num_q_vectors = 1;
1168 	if (!pci_enable_msi(adapter->pdev))
1169 		adapter->flags |= IGB_FLAG_HAS_MSI;
1170 }
1171 
1172 static void igb_add_ring(struct igb_ring *ring,
1173 			 struct igb_ring_container *head)
1174 {
1175 	head->ring = ring;
1176 	head->count++;
1177 }
1178 
1179 /**
1180  *  igb_alloc_q_vector - Allocate memory for a single interrupt vector
1181  *  @adapter: board private structure to initialize
1182  *  @v_count: q_vectors allocated on adapter, used for ring interleaving
1183  *  @v_idx: index of vector in adapter struct
1184  *  @txr_count: total number of Tx rings to allocate
1185  *  @txr_idx: index of first Tx ring to allocate
1186  *  @rxr_count: total number of Rx rings to allocate
1187  *  @rxr_idx: index of first Rx ring to allocate
1188  *
1189  *  We allocate one q_vector.  If allocation fails we return -ENOMEM.
1190  **/
1191 static int igb_alloc_q_vector(struct igb_adapter *adapter,
1192 			      int v_count, int v_idx,
1193 			      int txr_count, int txr_idx,
1194 			      int rxr_count, int rxr_idx)
1195 {
1196 	struct igb_q_vector *q_vector;
1197 	struct igb_ring *ring;
1198 	int ring_count, size;
1199 
1200 	/* igb only supports 1 Tx and/or 1 Rx queue per vector */
1201 	if (txr_count > 1 || rxr_count > 1)
1202 		return -ENOMEM;
1203 
1204 	ring_count = txr_count + rxr_count;
1205 	size = sizeof(struct igb_q_vector) +
1206 	       (sizeof(struct igb_ring) * ring_count);
1207 
1208 	/* allocate q_vector and rings */
1209 	q_vector = adapter->q_vector[v_idx];
1210 	if (!q_vector)
1211 		q_vector = kzalloc(size, GFP_KERNEL);
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_tx_stop_all_queues(netdev);
1782 
1783 	/* disable transmits in the hardware */
1784 	tctl = rd32(E1000_TCTL);
1785 	tctl &= ~E1000_TCTL_EN;
1786 	wr32(E1000_TCTL, tctl);
1787 	/* flush both disables and wait for them to finish */
1788 	wrfl();
1789 	usleep_range(10000, 11000);
1790 
1791 	igb_irq_disable(adapter);
1792 
1793 	adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
1794 
1795 	for (i = 0; i < adapter->num_q_vectors; i++) {
1796 		if (adapter->q_vector[i]) {
1797 			napi_synchronize(&adapter->q_vector[i]->napi);
1798 			napi_disable(&adapter->q_vector[i]->napi);
1799 		}
1800 	}
1801 
1802 
1803 	del_timer_sync(&adapter->watchdog_timer);
1804 	del_timer_sync(&adapter->phy_info_timer);
1805 
1806 	netif_carrier_off(netdev);
1807 
1808 	/* record the stats before reset*/
1809 	spin_lock(&adapter->stats64_lock);
1810 	igb_update_stats(adapter, &adapter->stats64);
1811 	spin_unlock(&adapter->stats64_lock);
1812 
1813 	adapter->link_speed = 0;
1814 	adapter->link_duplex = 0;
1815 
1816 	if (!pci_channel_offline(adapter->pdev))
1817 		igb_reset(adapter);
1818 	igb_clean_all_tx_rings(adapter);
1819 	igb_clean_all_rx_rings(adapter);
1820 #ifdef CONFIG_IGB_DCA
1821 
1822 	/* since we reset the hardware DCA settings were cleared */
1823 	igb_setup_dca(adapter);
1824 #endif
1825 }
1826 
1827 void igb_reinit_locked(struct igb_adapter *adapter)
1828 {
1829 	WARN_ON(in_interrupt());
1830 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
1831 		usleep_range(1000, 2000);
1832 	igb_down(adapter);
1833 	igb_up(adapter);
1834 	clear_bit(__IGB_RESETTING, &adapter->state);
1835 }
1836 
1837 /** igb_enable_mas - Media Autosense re-enable after swap
1838  *
1839  * @adapter: adapter struct
1840  **/
1841 static s32 igb_enable_mas(struct igb_adapter *adapter)
1842 {
1843 	struct e1000_hw *hw = &adapter->hw;
1844 	u32 connsw;
1845 	s32 ret_val = 0;
1846 
1847 	connsw = rd32(E1000_CONNSW);
1848 	if (!(hw->phy.media_type == e1000_media_type_copper))
1849 		return ret_val;
1850 
1851 	/* configure for SerDes media detect */
1852 	if (!(connsw & E1000_CONNSW_SERDESD)) {
1853 		connsw |= E1000_CONNSW_ENRGSRC;
1854 		connsw |= E1000_CONNSW_AUTOSENSE_EN;
1855 		wr32(E1000_CONNSW, connsw);
1856 		wrfl();
1857 	} else if (connsw & E1000_CONNSW_SERDESD) {
1858 		/* already SerDes, no need to enable anything */
1859 		return ret_val;
1860 	} else {
1861 		netdev_info(adapter->netdev,
1862 			"MAS: Unable to configure feature, disabling..\n");
1863 		adapter->flags &= ~IGB_FLAG_MAS_ENABLE;
1864 	}
1865 	return ret_val;
1866 }
1867 
1868 void igb_reset(struct igb_adapter *adapter)
1869 {
1870 	struct pci_dev *pdev = adapter->pdev;
1871 	struct e1000_hw *hw = &adapter->hw;
1872 	struct e1000_mac_info *mac = &hw->mac;
1873 	struct e1000_fc_info *fc = &hw->fc;
1874 	u32 pba = 0, tx_space, min_tx_space, min_rx_space, hwm;
1875 
1876 	/* Repartition Pba for greater than 9k mtu
1877 	 * To take effect CTRL.RST is required.
1878 	 */
1879 	switch (mac->type) {
1880 	case e1000_i350:
1881 	case e1000_i354:
1882 	case e1000_82580:
1883 		pba = rd32(E1000_RXPBS);
1884 		pba = igb_rxpbs_adjust_82580(pba);
1885 		break;
1886 	case e1000_82576:
1887 		pba = rd32(E1000_RXPBS);
1888 		pba &= E1000_RXPBS_SIZE_MASK_82576;
1889 		break;
1890 	case e1000_82575:
1891 	case e1000_i210:
1892 	case e1000_i211:
1893 	default:
1894 		pba = E1000_PBA_34K;
1895 		break;
1896 	}
1897 
1898 	if ((adapter->max_frame_size > ETH_FRAME_LEN + ETH_FCS_LEN) &&
1899 	    (mac->type < e1000_82576)) {
1900 		/* adjust PBA for jumbo frames */
1901 		wr32(E1000_PBA, pba);
1902 
1903 		/* To maintain wire speed transmits, the Tx FIFO should be
1904 		 * large enough to accommodate two full transmit packets,
1905 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
1906 		 * the Rx FIFO should be large enough to accommodate at least
1907 		 * one full receive packet and is similarly rounded up and
1908 		 * expressed in KB.
1909 		 */
1910 		pba = rd32(E1000_PBA);
1911 		/* upper 16 bits has Tx packet buffer allocation size in KB */
1912 		tx_space = pba >> 16;
1913 		/* lower 16 bits has Rx packet buffer allocation size in KB */
1914 		pba &= 0xffff;
1915 		/* the Tx fifo also stores 16 bytes of information about the Tx
1916 		 * but don't include ethernet FCS because hardware appends it
1917 		 */
1918 		min_tx_space = (adapter->max_frame_size +
1919 				sizeof(union e1000_adv_tx_desc) -
1920 				ETH_FCS_LEN) * 2;
1921 		min_tx_space = ALIGN(min_tx_space, 1024);
1922 		min_tx_space >>= 10;
1923 		/* software strips receive CRC, so leave room for it */
1924 		min_rx_space = adapter->max_frame_size;
1925 		min_rx_space = ALIGN(min_rx_space, 1024);
1926 		min_rx_space >>= 10;
1927 
1928 		/* If current Tx allocation is less than the min Tx FIFO size,
1929 		 * and the min Tx FIFO size is less than the current Rx FIFO
1930 		 * allocation, take space away from current Rx allocation
1931 		 */
1932 		if (tx_space < min_tx_space &&
1933 		    ((min_tx_space - tx_space) < pba)) {
1934 			pba = pba - (min_tx_space - tx_space);
1935 
1936 			/* if short on Rx space, Rx wins and must trump Tx
1937 			 * adjustment
1938 			 */
1939 			if (pba < min_rx_space)
1940 				pba = min_rx_space;
1941 		}
1942 		wr32(E1000_PBA, pba);
1943 	}
1944 
1945 	/* flow control settings */
1946 	/* The high water mark must be low enough to fit one full frame
1947 	 * (or the size used for early receive) above it in the Rx FIFO.
1948 	 * Set it to the lower of:
1949 	 * - 90% of the Rx FIFO size, or
1950 	 * - the full Rx FIFO size minus one full frame
1951 	 */
1952 	hwm = min(((pba << 10) * 9 / 10),
1953 			((pba << 10) - 2 * adapter->max_frame_size));
1954 
1955 	fc->high_water = hwm & 0xFFFFFFF0;	/* 16-byte granularity */
1956 	fc->low_water = fc->high_water - 16;
1957 	fc->pause_time = 0xFFFF;
1958 	fc->send_xon = 1;
1959 	fc->current_mode = fc->requested_mode;
1960 
1961 	/* disable receive for all VFs and wait one second */
1962 	if (adapter->vfs_allocated_count) {
1963 		int i;
1964 
1965 		for (i = 0 ; i < adapter->vfs_allocated_count; i++)
1966 			adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
1967 
1968 		/* ping all the active vfs to let them know we are going down */
1969 		igb_ping_all_vfs(adapter);
1970 
1971 		/* disable transmits and receives */
1972 		wr32(E1000_VFRE, 0);
1973 		wr32(E1000_VFTE, 0);
1974 	}
1975 
1976 	/* Allow time for pending master requests to run */
1977 	hw->mac.ops.reset_hw(hw);
1978 	wr32(E1000_WUC, 0);
1979 
1980 	if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
1981 		/* need to resetup here after media swap */
1982 		adapter->ei.get_invariants(hw);
1983 		adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
1984 	}
1985 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
1986 		if (igb_enable_mas(adapter))
1987 			dev_err(&pdev->dev,
1988 				"Error enabling Media Auto Sense\n");
1989 	}
1990 	if (hw->mac.ops.init_hw(hw))
1991 		dev_err(&pdev->dev, "Hardware Error\n");
1992 
1993 	/* Flow control settings reset on hardware reset, so guarantee flow
1994 	 * control is off when forcing speed.
1995 	 */
1996 	if (!hw->mac.autoneg)
1997 		igb_force_mac_fc(hw);
1998 
1999 	igb_init_dmac(adapter, pba);
2000 #ifdef CONFIG_IGB_HWMON
2001 	/* Re-initialize the thermal sensor on i350 devices. */
2002 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
2003 		if (mac->type == e1000_i350 && hw->bus.func == 0) {
2004 			/* If present, re-initialize the external thermal sensor
2005 			 * interface.
2006 			 */
2007 			if (adapter->ets)
2008 				mac->ops.init_thermal_sensor_thresh(hw);
2009 		}
2010 	}
2011 #endif
2012 	/* Re-establish EEE setting */
2013 	if (hw->phy.media_type == e1000_media_type_copper) {
2014 		switch (mac->type) {
2015 		case e1000_i350:
2016 		case e1000_i210:
2017 		case e1000_i211:
2018 			igb_set_eee_i350(hw, true, true);
2019 			break;
2020 		case e1000_i354:
2021 			igb_set_eee_i354(hw, true, true);
2022 			break;
2023 		default:
2024 			break;
2025 		}
2026 	}
2027 	if (!netif_running(adapter->netdev))
2028 		igb_power_down_link(adapter);
2029 
2030 	igb_update_mng_vlan(adapter);
2031 
2032 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2033 	wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2034 
2035 	/* Re-enable PTP, where applicable. */
2036 	igb_ptp_reset(adapter);
2037 
2038 	igb_get_phy_info(hw);
2039 }
2040 
2041 static netdev_features_t igb_fix_features(struct net_device *netdev,
2042 	netdev_features_t features)
2043 {
2044 	/* Since there is no support for separate Rx/Tx vlan accel
2045 	 * enable/disable make sure Tx flag is always in same state as Rx.
2046 	 */
2047 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2048 		features |= NETIF_F_HW_VLAN_CTAG_TX;
2049 	else
2050 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2051 
2052 	return features;
2053 }
2054 
2055 static int igb_set_features(struct net_device *netdev,
2056 	netdev_features_t features)
2057 {
2058 	netdev_features_t changed = netdev->features ^ features;
2059 	struct igb_adapter *adapter = netdev_priv(netdev);
2060 
2061 	if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2062 		igb_vlan_mode(netdev, features);
2063 
2064 	if (!(changed & NETIF_F_RXALL))
2065 		return 0;
2066 
2067 	netdev->features = features;
2068 
2069 	if (netif_running(netdev))
2070 		igb_reinit_locked(adapter);
2071 	else
2072 		igb_reset(adapter);
2073 
2074 	return 0;
2075 }
2076 
2077 static const struct net_device_ops igb_netdev_ops = {
2078 	.ndo_open		= igb_open,
2079 	.ndo_stop		= igb_close,
2080 	.ndo_start_xmit		= igb_xmit_frame,
2081 	.ndo_get_stats64	= igb_get_stats64,
2082 	.ndo_set_rx_mode	= igb_set_rx_mode,
2083 	.ndo_set_mac_address	= igb_set_mac,
2084 	.ndo_change_mtu		= igb_change_mtu,
2085 	.ndo_do_ioctl		= igb_ioctl,
2086 	.ndo_tx_timeout		= igb_tx_timeout,
2087 	.ndo_validate_addr	= eth_validate_addr,
2088 	.ndo_vlan_rx_add_vid	= igb_vlan_rx_add_vid,
2089 	.ndo_vlan_rx_kill_vid	= igb_vlan_rx_kill_vid,
2090 	.ndo_set_vf_mac		= igb_ndo_set_vf_mac,
2091 	.ndo_set_vf_vlan	= igb_ndo_set_vf_vlan,
2092 	.ndo_set_vf_rate	= igb_ndo_set_vf_bw,
2093 	.ndo_set_vf_spoofchk	= igb_ndo_set_vf_spoofchk,
2094 	.ndo_get_vf_config	= igb_ndo_get_vf_config,
2095 #ifdef CONFIG_NET_POLL_CONTROLLER
2096 	.ndo_poll_controller	= igb_netpoll,
2097 #endif
2098 	.ndo_fix_features	= igb_fix_features,
2099 	.ndo_set_features	= igb_set_features,
2100 };
2101 
2102 /**
2103  * igb_set_fw_version - Configure version string for ethtool
2104  * @adapter: adapter struct
2105  **/
2106 void igb_set_fw_version(struct igb_adapter *adapter)
2107 {
2108 	struct e1000_hw *hw = &adapter->hw;
2109 	struct e1000_fw_version fw;
2110 
2111 	igb_get_fw_version(hw, &fw);
2112 
2113 	switch (hw->mac.type) {
2114 	case e1000_i210:
2115 	case e1000_i211:
2116 		if (!(igb_get_flash_presence_i210(hw))) {
2117 			snprintf(adapter->fw_version,
2118 				 sizeof(adapter->fw_version),
2119 				 "%2d.%2d-%d",
2120 				 fw.invm_major, fw.invm_minor,
2121 				 fw.invm_img_type);
2122 			break;
2123 		}
2124 		/* fall through */
2125 	default:
2126 		/* if option is rom valid, display its version too */
2127 		if (fw.or_valid) {
2128 			snprintf(adapter->fw_version,
2129 				 sizeof(adapter->fw_version),
2130 				 "%d.%d, 0x%08x, %d.%d.%d",
2131 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2132 				 fw.or_major, fw.or_build, fw.or_patch);
2133 		/* no option rom */
2134 		} else if (fw.etrack_id != 0X0000) {
2135 			snprintf(adapter->fw_version,
2136 			    sizeof(adapter->fw_version),
2137 			    "%d.%d, 0x%08x",
2138 			    fw.eep_major, fw.eep_minor, fw.etrack_id);
2139 		} else {
2140 		snprintf(adapter->fw_version,
2141 		    sizeof(adapter->fw_version),
2142 		    "%d.%d.%d",
2143 		    fw.eep_major, fw.eep_minor, fw.eep_build);
2144 		}
2145 		break;
2146 	}
2147 }
2148 
2149 /**
2150  * igb_init_mas - init Media Autosense feature if enabled in the NVM
2151  *
2152  * @adapter: adapter struct
2153  **/
2154 static void igb_init_mas(struct igb_adapter *adapter)
2155 {
2156 	struct e1000_hw *hw = &adapter->hw;
2157 	u16 eeprom_data;
2158 
2159 	hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
2160 	switch (hw->bus.func) {
2161 	case E1000_FUNC_0:
2162 		if (eeprom_data & IGB_MAS_ENABLE_0) {
2163 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2164 			netdev_info(adapter->netdev,
2165 				"MAS: Enabling Media Autosense for port %d\n",
2166 				hw->bus.func);
2167 		}
2168 		break;
2169 	case E1000_FUNC_1:
2170 		if (eeprom_data & IGB_MAS_ENABLE_1) {
2171 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2172 			netdev_info(adapter->netdev,
2173 				"MAS: Enabling Media Autosense for port %d\n",
2174 				hw->bus.func);
2175 		}
2176 		break;
2177 	case E1000_FUNC_2:
2178 		if (eeprom_data & IGB_MAS_ENABLE_2) {
2179 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2180 			netdev_info(adapter->netdev,
2181 				"MAS: Enabling Media Autosense for port %d\n",
2182 				hw->bus.func);
2183 		}
2184 		break;
2185 	case E1000_FUNC_3:
2186 		if (eeprom_data & IGB_MAS_ENABLE_3) {
2187 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2188 			netdev_info(adapter->netdev,
2189 				"MAS: Enabling Media Autosense for port %d\n",
2190 				hw->bus.func);
2191 		}
2192 		break;
2193 	default:
2194 		/* Shouldn't get here */
2195 		netdev_err(adapter->netdev,
2196 			"MAS: Invalid port configuration, returning\n");
2197 		break;
2198 	}
2199 }
2200 
2201 /**
2202  *  igb_init_i2c - Init I2C interface
2203  *  @adapter: pointer to adapter structure
2204  **/
2205 static s32 igb_init_i2c(struct igb_adapter *adapter)
2206 {
2207 	s32 status = 0;
2208 
2209 	/* I2C interface supported on i350 devices */
2210 	if (adapter->hw.mac.type != e1000_i350)
2211 		return 0;
2212 
2213 	/* Initialize the i2c bus which is controlled by the registers.
2214 	 * This bus will use the i2c_algo_bit structue that implements
2215 	 * the protocol through toggling of the 4 bits in the register.
2216 	 */
2217 	adapter->i2c_adap.owner = THIS_MODULE;
2218 	adapter->i2c_algo = igb_i2c_algo;
2219 	adapter->i2c_algo.data = adapter;
2220 	adapter->i2c_adap.algo_data = &adapter->i2c_algo;
2221 	adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
2222 	strlcpy(adapter->i2c_adap.name, "igb BB",
2223 		sizeof(adapter->i2c_adap.name));
2224 	status = i2c_bit_add_bus(&adapter->i2c_adap);
2225 	return status;
2226 }
2227 
2228 /**
2229  *  igb_probe - Device Initialization Routine
2230  *  @pdev: PCI device information struct
2231  *  @ent: entry in igb_pci_tbl
2232  *
2233  *  Returns 0 on success, negative on failure
2234  *
2235  *  igb_probe initializes an adapter identified by a pci_dev structure.
2236  *  The OS initialization, configuring of the adapter private structure,
2237  *  and a hardware reset occur.
2238  **/
2239 static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2240 {
2241 	struct net_device *netdev;
2242 	struct igb_adapter *adapter;
2243 	struct e1000_hw *hw;
2244 	u16 eeprom_data = 0;
2245 	s32 ret_val;
2246 	static int global_quad_port_a; /* global quad port a indication */
2247 	const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
2248 	int err, pci_using_dac;
2249 	u8 part_str[E1000_PBANUM_LENGTH];
2250 
2251 	/* Catch broken hardware that put the wrong VF device ID in
2252 	 * the PCIe SR-IOV capability.
2253 	 */
2254 	if (pdev->is_virtfn) {
2255 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
2256 			pci_name(pdev), pdev->vendor, pdev->device);
2257 		return -EINVAL;
2258 	}
2259 
2260 	err = pci_enable_device_mem(pdev);
2261 	if (err)
2262 		return err;
2263 
2264 	pci_using_dac = 0;
2265 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2266 	if (!err) {
2267 		pci_using_dac = 1;
2268 	} else {
2269 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
2270 		if (err) {
2271 			dev_err(&pdev->dev,
2272 				"No usable DMA configuration, aborting\n");
2273 			goto err_dma;
2274 		}
2275 	}
2276 
2277 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
2278 					   IORESOURCE_MEM),
2279 					   igb_driver_name);
2280 	if (err)
2281 		goto err_pci_reg;
2282 
2283 	pci_enable_pcie_error_reporting(pdev);
2284 
2285 	pci_set_master(pdev);
2286 	pci_save_state(pdev);
2287 
2288 	err = -ENOMEM;
2289 	netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
2290 				   IGB_MAX_TX_QUEUES);
2291 	if (!netdev)
2292 		goto err_alloc_etherdev;
2293 
2294 	SET_NETDEV_DEV(netdev, &pdev->dev);
2295 
2296 	pci_set_drvdata(pdev, netdev);
2297 	adapter = netdev_priv(netdev);
2298 	adapter->netdev = netdev;
2299 	adapter->pdev = pdev;
2300 	hw = &adapter->hw;
2301 	hw->back = adapter;
2302 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
2303 
2304 	err = -EIO;
2305 	hw->hw_addr = pci_iomap(pdev, 0, 0);
2306 	if (!hw->hw_addr)
2307 		goto err_ioremap;
2308 
2309 	netdev->netdev_ops = &igb_netdev_ops;
2310 	igb_set_ethtool_ops(netdev);
2311 	netdev->watchdog_timeo = 5 * HZ;
2312 
2313 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
2314 
2315 	netdev->mem_start = pci_resource_start(pdev, 0);
2316 	netdev->mem_end = pci_resource_end(pdev, 0);
2317 
2318 	/* PCI config space info */
2319 	hw->vendor_id = pdev->vendor;
2320 	hw->device_id = pdev->device;
2321 	hw->revision_id = pdev->revision;
2322 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
2323 	hw->subsystem_device_id = pdev->subsystem_device;
2324 
2325 	/* Copy the default MAC, PHY and NVM function pointers */
2326 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
2327 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
2328 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
2329 	/* Initialize skew-specific constants */
2330 	err = ei->get_invariants(hw);
2331 	if (err)
2332 		goto err_sw_init;
2333 
2334 	/* setup the private structure */
2335 	err = igb_sw_init(adapter);
2336 	if (err)
2337 		goto err_sw_init;
2338 
2339 	igb_get_bus_info_pcie(hw);
2340 
2341 	hw->phy.autoneg_wait_to_complete = false;
2342 
2343 	/* Copper options */
2344 	if (hw->phy.media_type == e1000_media_type_copper) {
2345 		hw->phy.mdix = AUTO_ALL_MODES;
2346 		hw->phy.disable_polarity_correction = false;
2347 		hw->phy.ms_type = e1000_ms_hw_default;
2348 	}
2349 
2350 	if (igb_check_reset_block(hw))
2351 		dev_info(&pdev->dev,
2352 			"PHY reset is blocked due to SOL/IDER session.\n");
2353 
2354 	/* features is initialized to 0 in allocation, it might have bits
2355 	 * set by igb_sw_init so we should use an or instead of an
2356 	 * assignment.
2357 	 */
2358 	netdev->features |= NETIF_F_SG |
2359 			    NETIF_F_IP_CSUM |
2360 			    NETIF_F_IPV6_CSUM |
2361 			    NETIF_F_TSO |
2362 			    NETIF_F_TSO6 |
2363 			    NETIF_F_RXHASH |
2364 			    NETIF_F_RXCSUM |
2365 			    NETIF_F_HW_VLAN_CTAG_RX |
2366 			    NETIF_F_HW_VLAN_CTAG_TX;
2367 
2368 	/* copy netdev features into list of user selectable features */
2369 	netdev->hw_features |= netdev->features;
2370 	netdev->hw_features |= NETIF_F_RXALL;
2371 
2372 	/* set this bit last since it cannot be part of hw_features */
2373 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
2374 
2375 	netdev->vlan_features |= NETIF_F_TSO |
2376 				 NETIF_F_TSO6 |
2377 				 NETIF_F_IP_CSUM |
2378 				 NETIF_F_IPV6_CSUM |
2379 				 NETIF_F_SG;
2380 
2381 	netdev->priv_flags |= IFF_SUPP_NOFCS;
2382 
2383 	if (pci_using_dac) {
2384 		netdev->features |= NETIF_F_HIGHDMA;
2385 		netdev->vlan_features |= NETIF_F_HIGHDMA;
2386 	}
2387 
2388 	if (hw->mac.type >= e1000_82576) {
2389 		netdev->hw_features |= NETIF_F_SCTP_CSUM;
2390 		netdev->features |= NETIF_F_SCTP_CSUM;
2391 	}
2392 
2393 	netdev->priv_flags |= IFF_UNICAST_FLT;
2394 
2395 	adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
2396 
2397 	/* before reading the NVM, reset the controller to put the device in a
2398 	 * known good starting state
2399 	 */
2400 	hw->mac.ops.reset_hw(hw);
2401 
2402 	/* make sure the NVM is good , i211/i210 parts can have special NVM
2403 	 * that doesn't contain a checksum
2404 	 */
2405 	switch (hw->mac.type) {
2406 	case e1000_i210:
2407 	case e1000_i211:
2408 		if (igb_get_flash_presence_i210(hw)) {
2409 			if (hw->nvm.ops.validate(hw) < 0) {
2410 				dev_err(&pdev->dev,
2411 					"The NVM Checksum Is Not Valid\n");
2412 				err = -EIO;
2413 				goto err_eeprom;
2414 			}
2415 		}
2416 		break;
2417 	default:
2418 		if (hw->nvm.ops.validate(hw) < 0) {
2419 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
2420 			err = -EIO;
2421 			goto err_eeprom;
2422 		}
2423 		break;
2424 	}
2425 
2426 	/* copy the MAC address out of the NVM */
2427 	if (hw->mac.ops.read_mac_addr(hw))
2428 		dev_err(&pdev->dev, "NVM Read Error\n");
2429 
2430 	memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
2431 
2432 	if (!is_valid_ether_addr(netdev->dev_addr)) {
2433 		dev_err(&pdev->dev, "Invalid MAC Address\n");
2434 		err = -EIO;
2435 		goto err_eeprom;
2436 	}
2437 
2438 	/* get firmware version for ethtool -i */
2439 	igb_set_fw_version(adapter);
2440 
2441 	/* configure RXPBSIZE and TXPBSIZE */
2442 	if (hw->mac.type == e1000_i210) {
2443 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
2444 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
2445 	}
2446 
2447 	setup_timer(&adapter->watchdog_timer, igb_watchdog,
2448 		    (unsigned long) adapter);
2449 	setup_timer(&adapter->phy_info_timer, igb_update_phy_info,
2450 		    (unsigned long) adapter);
2451 
2452 	INIT_WORK(&adapter->reset_task, igb_reset_task);
2453 	INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
2454 
2455 	/* Initialize link properties that are user-changeable */
2456 	adapter->fc_autoneg = true;
2457 	hw->mac.autoneg = true;
2458 	hw->phy.autoneg_advertised = 0x2f;
2459 
2460 	hw->fc.requested_mode = e1000_fc_default;
2461 	hw->fc.current_mode = e1000_fc_default;
2462 
2463 	igb_validate_mdi_setting(hw);
2464 
2465 	/* By default, support wake on port A */
2466 	if (hw->bus.func == 0)
2467 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2468 
2469 	/* Check the NVM for wake support on non-port A ports */
2470 	if (hw->mac.type >= e1000_82580)
2471 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
2472 				 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
2473 				 &eeprom_data);
2474 	else if (hw->bus.func == 1)
2475 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
2476 
2477 	if (eeprom_data & IGB_EEPROM_APME)
2478 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2479 
2480 	/* now that we have the eeprom settings, apply the special cases where
2481 	 * the eeprom may be wrong or the board simply won't support wake on
2482 	 * lan on a particular port
2483 	 */
2484 	switch (pdev->device) {
2485 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
2486 		adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2487 		break;
2488 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
2489 	case E1000_DEV_ID_82576_FIBER:
2490 	case E1000_DEV_ID_82576_SERDES:
2491 		/* Wake events only supported on port A for dual fiber
2492 		 * regardless of eeprom setting
2493 		 */
2494 		if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
2495 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2496 		break;
2497 	case E1000_DEV_ID_82576_QUAD_COPPER:
2498 	case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
2499 		/* if quad port adapter, disable WoL on all but port A */
2500 		if (global_quad_port_a != 0)
2501 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2502 		else
2503 			adapter->flags |= IGB_FLAG_QUAD_PORT_A;
2504 		/* Reset for multiple quad port adapters */
2505 		if (++global_quad_port_a == 4)
2506 			global_quad_port_a = 0;
2507 		break;
2508 	default:
2509 		/* If the device can't wake, don't set software support */
2510 		if (!device_can_wakeup(&adapter->pdev->dev))
2511 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2512 	}
2513 
2514 	/* initialize the wol settings based on the eeprom settings */
2515 	if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
2516 		adapter->wol |= E1000_WUFC_MAG;
2517 
2518 	/* Some vendors want WoL disabled by default, but still supported */
2519 	if ((hw->mac.type == e1000_i350) &&
2520 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
2521 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2522 		adapter->wol = 0;
2523 	}
2524 
2525 	device_set_wakeup_enable(&adapter->pdev->dev,
2526 				 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
2527 
2528 	/* reset the hardware with the new settings */
2529 	igb_reset(adapter);
2530 
2531 	/* Init the I2C interface */
2532 	err = igb_init_i2c(adapter);
2533 	if (err) {
2534 		dev_err(&pdev->dev, "failed to init i2c interface\n");
2535 		goto err_eeprom;
2536 	}
2537 
2538 	/* let the f/w know that the h/w is now under the control of the
2539 	 * driver.
2540 	 */
2541 	igb_get_hw_control(adapter);
2542 
2543 	strcpy(netdev->name, "eth%d");
2544 	err = register_netdev(netdev);
2545 	if (err)
2546 		goto err_register;
2547 
2548 	/* carrier off reporting is important to ethtool even BEFORE open */
2549 	netif_carrier_off(netdev);
2550 
2551 #ifdef CONFIG_IGB_DCA
2552 	if (dca_add_requester(&pdev->dev) == 0) {
2553 		adapter->flags |= IGB_FLAG_DCA_ENABLED;
2554 		dev_info(&pdev->dev, "DCA enabled\n");
2555 		igb_setup_dca(adapter);
2556 	}
2557 
2558 #endif
2559 #ifdef CONFIG_IGB_HWMON
2560 	/* Initialize the thermal sensor on i350 devices. */
2561 	if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
2562 		u16 ets_word;
2563 
2564 		/* Read the NVM to determine if this i350 device supports an
2565 		 * external thermal sensor.
2566 		 */
2567 		hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
2568 		if (ets_word != 0x0000 && ets_word != 0xFFFF)
2569 			adapter->ets = true;
2570 		else
2571 			adapter->ets = false;
2572 		if (igb_sysfs_init(adapter))
2573 			dev_err(&pdev->dev,
2574 				"failed to allocate sysfs resources\n");
2575 	} else {
2576 		adapter->ets = false;
2577 	}
2578 #endif
2579 	/* Check if Media Autosense is enabled */
2580 	adapter->ei = *ei;
2581 	if (hw->dev_spec._82575.mas_capable)
2582 		igb_init_mas(adapter);
2583 
2584 	/* do hw tstamp init after resetting */
2585 	igb_ptp_init(adapter);
2586 
2587 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
2588 	/* print bus type/speed/width info, not applicable to i354 */
2589 	if (hw->mac.type != e1000_i354) {
2590 		dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
2591 			 netdev->name,
2592 			 ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
2593 			  (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
2594 			   "unknown"),
2595 			 ((hw->bus.width == e1000_bus_width_pcie_x4) ?
2596 			  "Width x4" :
2597 			  (hw->bus.width == e1000_bus_width_pcie_x2) ?
2598 			  "Width x2" :
2599 			  (hw->bus.width == e1000_bus_width_pcie_x1) ?
2600 			  "Width x1" : "unknown"), netdev->dev_addr);
2601 	}
2602 
2603 	if ((hw->mac.type >= e1000_i210 ||
2604 	     igb_get_flash_presence_i210(hw))) {
2605 		ret_val = igb_read_part_string(hw, part_str,
2606 					       E1000_PBANUM_LENGTH);
2607 	} else {
2608 		ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
2609 	}
2610 
2611 	if (ret_val)
2612 		strcpy(part_str, "Unknown");
2613 	dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
2614 	dev_info(&pdev->dev,
2615 		"Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
2616 		(adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
2617 		(adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
2618 		adapter->num_rx_queues, adapter->num_tx_queues);
2619 	if (hw->phy.media_type == e1000_media_type_copper) {
2620 		switch (hw->mac.type) {
2621 		case e1000_i350:
2622 		case e1000_i210:
2623 		case e1000_i211:
2624 			/* Enable EEE for internal copper PHY devices */
2625 			err = igb_set_eee_i350(hw, true, true);
2626 			if ((!err) &&
2627 			    (!hw->dev_spec._82575.eee_disable)) {
2628 				adapter->eee_advert =
2629 					MDIO_EEE_100TX | MDIO_EEE_1000T;
2630 				adapter->flags |= IGB_FLAG_EEE;
2631 			}
2632 			break;
2633 		case e1000_i354:
2634 			if ((rd32(E1000_CTRL_EXT) &
2635 			    E1000_CTRL_EXT_LINK_MODE_SGMII)) {
2636 				err = igb_set_eee_i354(hw, true, true);
2637 				if ((!err) &&
2638 					(!hw->dev_spec._82575.eee_disable)) {
2639 					adapter->eee_advert =
2640 					   MDIO_EEE_100TX | MDIO_EEE_1000T;
2641 					adapter->flags |= IGB_FLAG_EEE;
2642 				}
2643 			}
2644 			break;
2645 		default:
2646 			break;
2647 		}
2648 	}
2649 	pm_runtime_put_noidle(&pdev->dev);
2650 	return 0;
2651 
2652 err_register:
2653 	igb_release_hw_control(adapter);
2654 	memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
2655 err_eeprom:
2656 	if (!igb_check_reset_block(hw))
2657 		igb_reset_phy(hw);
2658 
2659 	if (hw->flash_address)
2660 		iounmap(hw->flash_address);
2661 err_sw_init:
2662 	igb_clear_interrupt_scheme(adapter);
2663 	pci_iounmap(pdev, hw->hw_addr);
2664 err_ioremap:
2665 	free_netdev(netdev);
2666 err_alloc_etherdev:
2667 	pci_release_selected_regions(pdev,
2668 				     pci_select_bars(pdev, IORESOURCE_MEM));
2669 err_pci_reg:
2670 err_dma:
2671 	pci_disable_device(pdev);
2672 	return err;
2673 }
2674 
2675 #ifdef CONFIG_PCI_IOV
2676 static int igb_disable_sriov(struct pci_dev *pdev)
2677 {
2678 	struct net_device *netdev = pci_get_drvdata(pdev);
2679 	struct igb_adapter *adapter = netdev_priv(netdev);
2680 	struct e1000_hw *hw = &adapter->hw;
2681 
2682 	/* reclaim resources allocated to VFs */
2683 	if (adapter->vf_data) {
2684 		/* disable iov and allow time for transactions to clear */
2685 		if (pci_vfs_assigned(pdev)) {
2686 			dev_warn(&pdev->dev,
2687 				 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
2688 			return -EPERM;
2689 		} else {
2690 			pci_disable_sriov(pdev);
2691 			msleep(500);
2692 		}
2693 
2694 		kfree(adapter->vf_data);
2695 		adapter->vf_data = NULL;
2696 		adapter->vfs_allocated_count = 0;
2697 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
2698 		wrfl();
2699 		msleep(100);
2700 		dev_info(&pdev->dev, "IOV Disabled\n");
2701 
2702 		/* Re-enable DMA Coalescing flag since IOV is turned off */
2703 		adapter->flags |= IGB_FLAG_DMAC;
2704 	}
2705 
2706 	return 0;
2707 }
2708 
2709 static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
2710 {
2711 	struct net_device *netdev = pci_get_drvdata(pdev);
2712 	struct igb_adapter *adapter = netdev_priv(netdev);
2713 	int old_vfs = pci_num_vf(pdev);
2714 	int err = 0;
2715 	int i;
2716 
2717 	if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
2718 		err = -EPERM;
2719 		goto out;
2720 	}
2721 	if (!num_vfs)
2722 		goto out;
2723 
2724 	if (old_vfs) {
2725 		dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
2726 			 old_vfs, max_vfs);
2727 		adapter->vfs_allocated_count = old_vfs;
2728 	} else
2729 		adapter->vfs_allocated_count = num_vfs;
2730 
2731 	adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
2732 				sizeof(struct vf_data_storage), GFP_KERNEL);
2733 
2734 	/* if allocation failed then we do not support SR-IOV */
2735 	if (!adapter->vf_data) {
2736 		adapter->vfs_allocated_count = 0;
2737 		dev_err(&pdev->dev,
2738 			"Unable to allocate memory for VF Data Storage\n");
2739 		err = -ENOMEM;
2740 		goto out;
2741 	}
2742 
2743 	/* only call pci_enable_sriov() if no VFs are allocated already */
2744 	if (!old_vfs) {
2745 		err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
2746 		if (err)
2747 			goto err_out;
2748 	}
2749 	dev_info(&pdev->dev, "%d VFs allocated\n",
2750 		 adapter->vfs_allocated_count);
2751 	for (i = 0; i < adapter->vfs_allocated_count; i++)
2752 		igb_vf_configure(adapter, i);
2753 
2754 	/* DMA Coalescing is not supported in IOV mode. */
2755 	adapter->flags &= ~IGB_FLAG_DMAC;
2756 	goto out;
2757 
2758 err_out:
2759 	kfree(adapter->vf_data);
2760 	adapter->vf_data = NULL;
2761 	adapter->vfs_allocated_count = 0;
2762 out:
2763 	return err;
2764 }
2765 
2766 #endif
2767 /**
2768  *  igb_remove_i2c - Cleanup  I2C interface
2769  *  @adapter: pointer to adapter structure
2770  **/
2771 static void igb_remove_i2c(struct igb_adapter *adapter)
2772 {
2773 	/* free the adapter bus structure */
2774 	i2c_del_adapter(&adapter->i2c_adap);
2775 }
2776 
2777 /**
2778  *  igb_remove - Device Removal Routine
2779  *  @pdev: PCI device information struct
2780  *
2781  *  igb_remove is called by the PCI subsystem to alert the driver
2782  *  that it should release a PCI device.  The could be caused by a
2783  *  Hot-Plug event, or because the driver is going to be removed from
2784  *  memory.
2785  **/
2786 static void igb_remove(struct pci_dev *pdev)
2787 {
2788 	struct net_device *netdev = pci_get_drvdata(pdev);
2789 	struct igb_adapter *adapter = netdev_priv(netdev);
2790 	struct e1000_hw *hw = &adapter->hw;
2791 
2792 	pm_runtime_get_noresume(&pdev->dev);
2793 #ifdef CONFIG_IGB_HWMON
2794 	igb_sysfs_exit(adapter);
2795 #endif
2796 	igb_remove_i2c(adapter);
2797 	igb_ptp_stop(adapter);
2798 	/* The watchdog timer may be rescheduled, so explicitly
2799 	 * disable watchdog from being rescheduled.
2800 	 */
2801 	set_bit(__IGB_DOWN, &adapter->state);
2802 	del_timer_sync(&adapter->watchdog_timer);
2803 	del_timer_sync(&adapter->phy_info_timer);
2804 
2805 	cancel_work_sync(&adapter->reset_task);
2806 	cancel_work_sync(&adapter->watchdog_task);
2807 
2808 #ifdef CONFIG_IGB_DCA
2809 	if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
2810 		dev_info(&pdev->dev, "DCA disabled\n");
2811 		dca_remove_requester(&pdev->dev);
2812 		adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
2813 		wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
2814 	}
2815 #endif
2816 
2817 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
2818 	 * would have already happened in close and is redundant.
2819 	 */
2820 	igb_release_hw_control(adapter);
2821 
2822 	unregister_netdev(netdev);
2823 
2824 	igb_clear_interrupt_scheme(adapter);
2825 
2826 #ifdef CONFIG_PCI_IOV
2827 	igb_disable_sriov(pdev);
2828 #endif
2829 
2830 	pci_iounmap(pdev, hw->hw_addr);
2831 	if (hw->flash_address)
2832 		iounmap(hw->flash_address);
2833 	pci_release_selected_regions(pdev,
2834 				     pci_select_bars(pdev, IORESOURCE_MEM));
2835 
2836 	kfree(adapter->shadow_vfta);
2837 	free_netdev(netdev);
2838 
2839 	pci_disable_pcie_error_reporting(pdev);
2840 
2841 	pci_disable_device(pdev);
2842 }
2843 
2844 /**
2845  *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
2846  *  @adapter: board private structure to initialize
2847  *
2848  *  This function initializes the vf specific data storage and then attempts to
2849  *  allocate the VFs.  The reason for ordering it this way is because it is much
2850  *  mor expensive time wise to disable SR-IOV than it is to allocate and free
2851  *  the memory for the VFs.
2852  **/
2853 static void igb_probe_vfs(struct igb_adapter *adapter)
2854 {
2855 #ifdef CONFIG_PCI_IOV
2856 	struct pci_dev *pdev = adapter->pdev;
2857 	struct e1000_hw *hw = &adapter->hw;
2858 
2859 	/* Virtualization features not supported on i210 family. */
2860 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
2861 		return;
2862 
2863 	pci_sriov_set_totalvfs(pdev, 7);
2864 	igb_pci_enable_sriov(pdev, max_vfs);
2865 
2866 #endif /* CONFIG_PCI_IOV */
2867 }
2868 
2869 static void igb_init_queue_configuration(struct igb_adapter *adapter)
2870 {
2871 	struct e1000_hw *hw = &adapter->hw;
2872 	u32 max_rss_queues;
2873 
2874 	/* Determine the maximum number of RSS queues supported. */
2875 	switch (hw->mac.type) {
2876 	case e1000_i211:
2877 		max_rss_queues = IGB_MAX_RX_QUEUES_I211;
2878 		break;
2879 	case e1000_82575:
2880 	case e1000_i210:
2881 		max_rss_queues = IGB_MAX_RX_QUEUES_82575;
2882 		break;
2883 	case e1000_i350:
2884 		/* I350 cannot do RSS and SR-IOV at the same time */
2885 		if (!!adapter->vfs_allocated_count) {
2886 			max_rss_queues = 1;
2887 			break;
2888 		}
2889 		/* fall through */
2890 	case e1000_82576:
2891 		if (!!adapter->vfs_allocated_count) {
2892 			max_rss_queues = 2;
2893 			break;
2894 		}
2895 		/* fall through */
2896 	case e1000_82580:
2897 	case e1000_i354:
2898 	default:
2899 		max_rss_queues = IGB_MAX_RX_QUEUES;
2900 		break;
2901 	}
2902 
2903 	adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
2904 
2905 	/* Determine if we need to pair queues. */
2906 	switch (hw->mac.type) {
2907 	case e1000_82575:
2908 	case e1000_i211:
2909 		/* Device supports enough interrupts without queue pairing. */
2910 		break;
2911 	case e1000_82576:
2912 		/* If VFs are going to be allocated with RSS queues then we
2913 		 * should pair the queues in order to conserve interrupts due
2914 		 * to limited supply.
2915 		 */
2916 		if ((adapter->rss_queues > 1) &&
2917 		    (adapter->vfs_allocated_count > 6))
2918 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
2919 		/* fall through */
2920 	case e1000_82580:
2921 	case e1000_i350:
2922 	case e1000_i354:
2923 	case e1000_i210:
2924 	default:
2925 		/* If rss_queues > half of max_rss_queues, pair the queues in
2926 		 * order to conserve interrupts due to limited supply.
2927 		 */
2928 		if (adapter->rss_queues > (max_rss_queues / 2))
2929 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
2930 		break;
2931 	}
2932 }
2933 
2934 /**
2935  *  igb_sw_init - Initialize general software structures (struct igb_adapter)
2936  *  @adapter: board private structure to initialize
2937  *
2938  *  igb_sw_init initializes the Adapter private data structure.
2939  *  Fields are initialized based on PCI device information and
2940  *  OS network device settings (MTU size).
2941  **/
2942 static int igb_sw_init(struct igb_adapter *adapter)
2943 {
2944 	struct e1000_hw *hw = &adapter->hw;
2945 	struct net_device *netdev = adapter->netdev;
2946 	struct pci_dev *pdev = adapter->pdev;
2947 
2948 	pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
2949 
2950 	/* set default ring sizes */
2951 	adapter->tx_ring_count = IGB_DEFAULT_TXD;
2952 	adapter->rx_ring_count = IGB_DEFAULT_RXD;
2953 
2954 	/* set default ITR values */
2955 	adapter->rx_itr_setting = IGB_DEFAULT_ITR;
2956 	adapter->tx_itr_setting = IGB_DEFAULT_ITR;
2957 
2958 	/* set default work limits */
2959 	adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
2960 
2961 	adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN +
2962 				  VLAN_HLEN;
2963 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
2964 
2965 	spin_lock_init(&adapter->stats64_lock);
2966 #ifdef CONFIG_PCI_IOV
2967 	switch (hw->mac.type) {
2968 	case e1000_82576:
2969 	case e1000_i350:
2970 		if (max_vfs > 7) {
2971 			dev_warn(&pdev->dev,
2972 				 "Maximum of 7 VFs per PF, using max\n");
2973 			max_vfs = adapter->vfs_allocated_count = 7;
2974 		} else
2975 			adapter->vfs_allocated_count = max_vfs;
2976 		if (adapter->vfs_allocated_count)
2977 			dev_warn(&pdev->dev,
2978 				 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
2979 		break;
2980 	default:
2981 		break;
2982 	}
2983 #endif /* CONFIG_PCI_IOV */
2984 
2985 	igb_init_queue_configuration(adapter);
2986 
2987 	/* Setup and initialize a copy of the hw vlan table array */
2988 	adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
2989 				       GFP_ATOMIC);
2990 
2991 	/* This call may decrease the number of queues */
2992 	if (igb_init_interrupt_scheme(adapter, true)) {
2993 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
2994 		return -ENOMEM;
2995 	}
2996 
2997 	igb_probe_vfs(adapter);
2998 
2999 	/* Explicitly disable IRQ since the NIC can be in any state. */
3000 	igb_irq_disable(adapter);
3001 
3002 	if (hw->mac.type >= e1000_i350)
3003 		adapter->flags &= ~IGB_FLAG_DMAC;
3004 
3005 	set_bit(__IGB_DOWN, &adapter->state);
3006 	return 0;
3007 }
3008 
3009 /**
3010  *  igb_open - Called when a network interface is made active
3011  *  @netdev: network interface device structure
3012  *
3013  *  Returns 0 on success, negative value on failure
3014  *
3015  *  The open entry point is called when a network interface is made
3016  *  active by the system (IFF_UP).  At this point all resources needed
3017  *  for transmit and receive operations are allocated, the interrupt
3018  *  handler is registered with the OS, the watchdog timer is started,
3019  *  and the stack is notified that the interface is ready.
3020  **/
3021 static int __igb_open(struct net_device *netdev, bool resuming)
3022 {
3023 	struct igb_adapter *adapter = netdev_priv(netdev);
3024 	struct e1000_hw *hw = &adapter->hw;
3025 	struct pci_dev *pdev = adapter->pdev;
3026 	int err;
3027 	int i;
3028 
3029 	/* disallow open during test */
3030 	if (test_bit(__IGB_TESTING, &adapter->state)) {
3031 		WARN_ON(resuming);
3032 		return -EBUSY;
3033 	}
3034 
3035 	if (!resuming)
3036 		pm_runtime_get_sync(&pdev->dev);
3037 
3038 	netif_carrier_off(netdev);
3039 
3040 	/* allocate transmit descriptors */
3041 	err = igb_setup_all_tx_resources(adapter);
3042 	if (err)
3043 		goto err_setup_tx;
3044 
3045 	/* allocate receive descriptors */
3046 	err = igb_setup_all_rx_resources(adapter);
3047 	if (err)
3048 		goto err_setup_rx;
3049 
3050 	igb_power_up_link(adapter);
3051 
3052 	/* before we allocate an interrupt, we must be ready to handle it.
3053 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3054 	 * as soon as we call pci_request_irq, so we have to setup our
3055 	 * clean_rx handler before we do so.
3056 	 */
3057 	igb_configure(adapter);
3058 
3059 	err = igb_request_irq(adapter);
3060 	if (err)
3061 		goto err_req_irq;
3062 
3063 	/* Notify the stack of the actual queue counts. */
3064 	err = netif_set_real_num_tx_queues(adapter->netdev,
3065 					   adapter->num_tx_queues);
3066 	if (err)
3067 		goto err_set_queues;
3068 
3069 	err = netif_set_real_num_rx_queues(adapter->netdev,
3070 					   adapter->num_rx_queues);
3071 	if (err)
3072 		goto err_set_queues;
3073 
3074 	/* From here on the code is the same as igb_up() */
3075 	clear_bit(__IGB_DOWN, &adapter->state);
3076 
3077 	for (i = 0; i < adapter->num_q_vectors; i++)
3078 		napi_enable(&(adapter->q_vector[i]->napi));
3079 
3080 	/* Clear any pending interrupts. */
3081 	rd32(E1000_ICR);
3082 
3083 	igb_irq_enable(adapter);
3084 
3085 	/* notify VFs that reset has been completed */
3086 	if (adapter->vfs_allocated_count) {
3087 		u32 reg_data = rd32(E1000_CTRL_EXT);
3088 
3089 		reg_data |= E1000_CTRL_EXT_PFRSTD;
3090 		wr32(E1000_CTRL_EXT, reg_data);
3091 	}
3092 
3093 	netif_tx_start_all_queues(netdev);
3094 
3095 	if (!resuming)
3096 		pm_runtime_put(&pdev->dev);
3097 
3098 	/* start the watchdog. */
3099 	hw->mac.get_link_status = 1;
3100 	schedule_work(&adapter->watchdog_task);
3101 
3102 	return 0;
3103 
3104 err_set_queues:
3105 	igb_free_irq(adapter);
3106 err_req_irq:
3107 	igb_release_hw_control(adapter);
3108 	igb_power_down_link(adapter);
3109 	igb_free_all_rx_resources(adapter);
3110 err_setup_rx:
3111 	igb_free_all_tx_resources(adapter);
3112 err_setup_tx:
3113 	igb_reset(adapter);
3114 	if (!resuming)
3115 		pm_runtime_put(&pdev->dev);
3116 
3117 	return err;
3118 }
3119 
3120 static int igb_open(struct net_device *netdev)
3121 {
3122 	return __igb_open(netdev, false);
3123 }
3124 
3125 /**
3126  *  igb_close - Disables a network interface
3127  *  @netdev: network interface device structure
3128  *
3129  *  Returns 0, this is not allowed to fail
3130  *
3131  *  The close entry point is called when an interface is de-activated
3132  *  by the OS.  The hardware is still under the driver's control, but
3133  *  needs to be disabled.  A global MAC reset is issued to stop the
3134  *  hardware, and all transmit and receive resources are freed.
3135  **/
3136 static int __igb_close(struct net_device *netdev, bool suspending)
3137 {
3138 	struct igb_adapter *adapter = netdev_priv(netdev);
3139 	struct pci_dev *pdev = adapter->pdev;
3140 
3141 	WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
3142 
3143 	if (!suspending)
3144 		pm_runtime_get_sync(&pdev->dev);
3145 
3146 	igb_down(adapter);
3147 	igb_free_irq(adapter);
3148 
3149 	igb_free_all_tx_resources(adapter);
3150 	igb_free_all_rx_resources(adapter);
3151 
3152 	if (!suspending)
3153 		pm_runtime_put_sync(&pdev->dev);
3154 	return 0;
3155 }
3156 
3157 static int igb_close(struct net_device *netdev)
3158 {
3159 	return __igb_close(netdev, false);
3160 }
3161 
3162 /**
3163  *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
3164  *  @tx_ring: tx descriptor ring (for a specific queue) to setup
3165  *
3166  *  Return 0 on success, negative on failure
3167  **/
3168 int igb_setup_tx_resources(struct igb_ring *tx_ring)
3169 {
3170 	struct device *dev = tx_ring->dev;
3171 	int size;
3172 
3173 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3174 
3175 	tx_ring->tx_buffer_info = vzalloc(size);
3176 	if (!tx_ring->tx_buffer_info)
3177 		goto err;
3178 
3179 	/* round up to nearest 4K */
3180 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
3181 	tx_ring->size = ALIGN(tx_ring->size, 4096);
3182 
3183 	tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
3184 					   &tx_ring->dma, GFP_KERNEL);
3185 	if (!tx_ring->desc)
3186 		goto err;
3187 
3188 	tx_ring->next_to_use = 0;
3189 	tx_ring->next_to_clean = 0;
3190 
3191 	return 0;
3192 
3193 err:
3194 	vfree(tx_ring->tx_buffer_info);
3195 	tx_ring->tx_buffer_info = NULL;
3196 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
3197 	return -ENOMEM;
3198 }
3199 
3200 /**
3201  *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
3202  *				 (Descriptors) for all queues
3203  *  @adapter: board private structure
3204  *
3205  *  Return 0 on success, negative on failure
3206  **/
3207 static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
3208 {
3209 	struct pci_dev *pdev = adapter->pdev;
3210 	int i, err = 0;
3211 
3212 	for (i = 0; i < adapter->num_tx_queues; i++) {
3213 		err = igb_setup_tx_resources(adapter->tx_ring[i]);
3214 		if (err) {
3215 			dev_err(&pdev->dev,
3216 				"Allocation for Tx Queue %u failed\n", i);
3217 			for (i--; i >= 0; i--)
3218 				igb_free_tx_resources(adapter->tx_ring[i]);
3219 			break;
3220 		}
3221 	}
3222 
3223 	return err;
3224 }
3225 
3226 /**
3227  *  igb_setup_tctl - configure the transmit control registers
3228  *  @adapter: Board private structure
3229  **/
3230 void igb_setup_tctl(struct igb_adapter *adapter)
3231 {
3232 	struct e1000_hw *hw = &adapter->hw;
3233 	u32 tctl;
3234 
3235 	/* disable queue 0 which is enabled by default on 82575 and 82576 */
3236 	wr32(E1000_TXDCTL(0), 0);
3237 
3238 	/* Program the Transmit Control Register */
3239 	tctl = rd32(E1000_TCTL);
3240 	tctl &= ~E1000_TCTL_CT;
3241 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
3242 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
3243 
3244 	igb_config_collision_dist(hw);
3245 
3246 	/* Enable transmits */
3247 	tctl |= E1000_TCTL_EN;
3248 
3249 	wr32(E1000_TCTL, tctl);
3250 }
3251 
3252 /**
3253  *  igb_configure_tx_ring - Configure transmit ring after Reset
3254  *  @adapter: board private structure
3255  *  @ring: tx ring to configure
3256  *
3257  *  Configure a transmit ring after a reset.
3258  **/
3259 void igb_configure_tx_ring(struct igb_adapter *adapter,
3260 			   struct igb_ring *ring)
3261 {
3262 	struct e1000_hw *hw = &adapter->hw;
3263 	u32 txdctl = 0;
3264 	u64 tdba = ring->dma;
3265 	int reg_idx = ring->reg_idx;
3266 
3267 	/* disable the queue */
3268 	wr32(E1000_TXDCTL(reg_idx), 0);
3269 	wrfl();
3270 	mdelay(10);
3271 
3272 	wr32(E1000_TDLEN(reg_idx),
3273 	     ring->count * sizeof(union e1000_adv_tx_desc));
3274 	wr32(E1000_TDBAL(reg_idx),
3275 	     tdba & 0x00000000ffffffffULL);
3276 	wr32(E1000_TDBAH(reg_idx), tdba >> 32);
3277 
3278 	ring->tail = hw->hw_addr + E1000_TDT(reg_idx);
3279 	wr32(E1000_TDH(reg_idx), 0);
3280 	writel(0, ring->tail);
3281 
3282 	txdctl |= IGB_TX_PTHRESH;
3283 	txdctl |= IGB_TX_HTHRESH << 8;
3284 	txdctl |= IGB_TX_WTHRESH << 16;
3285 
3286 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
3287 	wr32(E1000_TXDCTL(reg_idx), txdctl);
3288 }
3289 
3290 /**
3291  *  igb_configure_tx - Configure transmit Unit after Reset
3292  *  @adapter: board private structure
3293  *
3294  *  Configure the Tx unit of the MAC after a reset.
3295  **/
3296 static void igb_configure_tx(struct igb_adapter *adapter)
3297 {
3298 	int i;
3299 
3300 	for (i = 0; i < adapter->num_tx_queues; i++)
3301 		igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
3302 }
3303 
3304 /**
3305  *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
3306  *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
3307  *
3308  *  Returns 0 on success, negative on failure
3309  **/
3310 int igb_setup_rx_resources(struct igb_ring *rx_ring)
3311 {
3312 	struct device *dev = rx_ring->dev;
3313 	int size;
3314 
3315 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3316 
3317 	rx_ring->rx_buffer_info = vzalloc(size);
3318 	if (!rx_ring->rx_buffer_info)
3319 		goto err;
3320 
3321 	/* Round up to nearest 4K */
3322 	rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
3323 	rx_ring->size = ALIGN(rx_ring->size, 4096);
3324 
3325 	rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
3326 					   &rx_ring->dma, GFP_KERNEL);
3327 	if (!rx_ring->desc)
3328 		goto err;
3329 
3330 	rx_ring->next_to_alloc = 0;
3331 	rx_ring->next_to_clean = 0;
3332 	rx_ring->next_to_use = 0;
3333 
3334 	return 0;
3335 
3336 err:
3337 	vfree(rx_ring->rx_buffer_info);
3338 	rx_ring->rx_buffer_info = NULL;
3339 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
3340 	return -ENOMEM;
3341 }
3342 
3343 /**
3344  *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
3345  *				 (Descriptors) for all queues
3346  *  @adapter: board private structure
3347  *
3348  *  Return 0 on success, negative on failure
3349  **/
3350 static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
3351 {
3352 	struct pci_dev *pdev = adapter->pdev;
3353 	int i, err = 0;
3354 
3355 	for (i = 0; i < adapter->num_rx_queues; i++) {
3356 		err = igb_setup_rx_resources(adapter->rx_ring[i]);
3357 		if (err) {
3358 			dev_err(&pdev->dev,
3359 				"Allocation for Rx Queue %u failed\n", i);
3360 			for (i--; i >= 0; i--)
3361 				igb_free_rx_resources(adapter->rx_ring[i]);
3362 			break;
3363 		}
3364 	}
3365 
3366 	return err;
3367 }
3368 
3369 /**
3370  *  igb_setup_mrqc - configure the multiple receive queue control registers
3371  *  @adapter: Board private structure
3372  **/
3373 static void igb_setup_mrqc(struct igb_adapter *adapter)
3374 {
3375 	struct e1000_hw *hw = &adapter->hw;
3376 	u32 mrqc, rxcsum;
3377 	u32 j, num_rx_queues;
3378 	static const u32 rsskey[10] = { 0xDA565A6D, 0xC20E5B25, 0x3D256741,
3379 					0xB08FA343, 0xCB2BCAD0, 0xB4307BAE,
3380 					0xA32DCB77, 0x0CF23080, 0x3BB7426A,
3381 					0xFA01ACBE };
3382 
3383 	/* Fill out hash function seeds */
3384 	for (j = 0; j < 10; j++)
3385 		wr32(E1000_RSSRK(j), rsskey[j]);
3386 
3387 	num_rx_queues = adapter->rss_queues;
3388 
3389 	switch (hw->mac.type) {
3390 	case e1000_82576:
3391 		/* 82576 supports 2 RSS queues for SR-IOV */
3392 		if (adapter->vfs_allocated_count)
3393 			num_rx_queues = 2;
3394 		break;
3395 	default:
3396 		break;
3397 	}
3398 
3399 	if (adapter->rss_indir_tbl_init != num_rx_queues) {
3400 		for (j = 0; j < IGB_RETA_SIZE; j++)
3401 			adapter->rss_indir_tbl[j] =
3402 			(j * num_rx_queues) / IGB_RETA_SIZE;
3403 		adapter->rss_indir_tbl_init = num_rx_queues;
3404 	}
3405 	igb_write_rss_indir_tbl(adapter);
3406 
3407 	/* Disable raw packet checksumming so that RSS hash is placed in
3408 	 * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
3409 	 * offloads as they are enabled by default
3410 	 */
3411 	rxcsum = rd32(E1000_RXCSUM);
3412 	rxcsum |= E1000_RXCSUM_PCSD;
3413 
3414 	if (adapter->hw.mac.type >= e1000_82576)
3415 		/* Enable Receive Checksum Offload for SCTP */
3416 		rxcsum |= E1000_RXCSUM_CRCOFL;
3417 
3418 	/* Don't need to set TUOFL or IPOFL, they default to 1 */
3419 	wr32(E1000_RXCSUM, rxcsum);
3420 
3421 	/* Generate RSS hash based on packet types, TCP/UDP
3422 	 * port numbers and/or IPv4/v6 src and dst addresses
3423 	 */
3424 	mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
3425 	       E1000_MRQC_RSS_FIELD_IPV4_TCP |
3426 	       E1000_MRQC_RSS_FIELD_IPV6 |
3427 	       E1000_MRQC_RSS_FIELD_IPV6_TCP |
3428 	       E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
3429 
3430 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
3431 		mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
3432 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
3433 		mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
3434 
3435 	/* If VMDq is enabled then we set the appropriate mode for that, else
3436 	 * we default to RSS so that an RSS hash is calculated per packet even
3437 	 * if we are only using one queue
3438 	 */
3439 	if (adapter->vfs_allocated_count) {
3440 		if (hw->mac.type > e1000_82575) {
3441 			/* Set the default pool for the PF's first queue */
3442 			u32 vtctl = rd32(E1000_VT_CTL);
3443 
3444 			vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
3445 				   E1000_VT_CTL_DISABLE_DEF_POOL);
3446 			vtctl |= adapter->vfs_allocated_count <<
3447 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
3448 			wr32(E1000_VT_CTL, vtctl);
3449 		}
3450 		if (adapter->rss_queues > 1)
3451 			mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_2Q;
3452 		else
3453 			mrqc |= E1000_MRQC_ENABLE_VMDQ;
3454 	} else {
3455 		if (hw->mac.type != e1000_i211)
3456 			mrqc |= E1000_MRQC_ENABLE_RSS_4Q;
3457 	}
3458 	igb_vmm_control(adapter);
3459 
3460 	wr32(E1000_MRQC, mrqc);
3461 }
3462 
3463 /**
3464  *  igb_setup_rctl - configure the receive control registers
3465  *  @adapter: Board private structure
3466  **/
3467 void igb_setup_rctl(struct igb_adapter *adapter)
3468 {
3469 	struct e1000_hw *hw = &adapter->hw;
3470 	u32 rctl;
3471 
3472 	rctl = rd32(E1000_RCTL);
3473 
3474 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3475 	rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
3476 
3477 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
3478 		(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3479 
3480 	/* enable stripping of CRC. It's unlikely this will break BMC
3481 	 * redirection as it did with e1000. Newer features require
3482 	 * that the HW strips the CRC.
3483 	 */
3484 	rctl |= E1000_RCTL_SECRC;
3485 
3486 	/* disable store bad packets and clear size bits. */
3487 	rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
3488 
3489 	/* enable LPE to prevent packets larger than max_frame_size */
3490 	rctl |= E1000_RCTL_LPE;
3491 
3492 	/* disable queue 0 to prevent tail write w/o re-config */
3493 	wr32(E1000_RXDCTL(0), 0);
3494 
3495 	/* Attention!!!  For SR-IOV PF driver operations you must enable
3496 	 * queue drop for all VF and PF queues to prevent head of line blocking
3497 	 * if an un-trusted VF does not provide descriptors to hardware.
3498 	 */
3499 	if (adapter->vfs_allocated_count) {
3500 		/* set all queue drop enable bits */
3501 		wr32(E1000_QDE, ALL_QUEUES);
3502 	}
3503 
3504 	/* This is useful for sniffing bad packets. */
3505 	if (adapter->netdev->features & NETIF_F_RXALL) {
3506 		/* UPE and MPE will be handled by normal PROMISC logic
3507 		 * in e1000e_set_rx_mode
3508 		 */
3509 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
3510 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
3511 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
3512 
3513 		rctl &= ~(E1000_RCTL_VFE | /* Disable VLAN filter */
3514 			  E1000_RCTL_DPF | /* Allow filtered pause */
3515 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
3516 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3517 		 * and that breaks VLANs.
3518 		 */
3519 	}
3520 
3521 	wr32(E1000_RCTL, rctl);
3522 }
3523 
3524 static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
3525 				   int vfn)
3526 {
3527 	struct e1000_hw *hw = &adapter->hw;
3528 	u32 vmolr;
3529 
3530 	/* if it isn't the PF check to see if VFs are enabled and
3531 	 * increase the size to support vlan tags
3532 	 */
3533 	if (vfn < adapter->vfs_allocated_count &&
3534 	    adapter->vf_data[vfn].vlans_enabled)
3535 		size += VLAN_TAG_SIZE;
3536 
3537 	vmolr = rd32(E1000_VMOLR(vfn));
3538 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
3539 	vmolr |= size | E1000_VMOLR_LPE;
3540 	wr32(E1000_VMOLR(vfn), vmolr);
3541 
3542 	return 0;
3543 }
3544 
3545 /**
3546  *  igb_rlpml_set - set maximum receive packet size
3547  *  @adapter: board private structure
3548  *
3549  *  Configure maximum receivable packet size.
3550  **/
3551 static void igb_rlpml_set(struct igb_adapter *adapter)
3552 {
3553 	u32 max_frame_size = adapter->max_frame_size;
3554 	struct e1000_hw *hw = &adapter->hw;
3555 	u16 pf_id = adapter->vfs_allocated_count;
3556 
3557 	if (pf_id) {
3558 		igb_set_vf_rlpml(adapter, max_frame_size, pf_id);
3559 		/* If we're in VMDQ or SR-IOV mode, then set global RLPML
3560 		 * to our max jumbo frame size, in case we need to enable
3561 		 * jumbo frames on one of the rings later.
3562 		 * This will not pass over-length frames into the default
3563 		 * queue because it's gated by the VMOLR.RLPML.
3564 		 */
3565 		max_frame_size = MAX_JUMBO_FRAME_SIZE;
3566 	}
3567 
3568 	wr32(E1000_RLPML, max_frame_size);
3569 }
3570 
3571 static inline void igb_set_vmolr(struct igb_adapter *adapter,
3572 				 int vfn, bool aupe)
3573 {
3574 	struct e1000_hw *hw = &adapter->hw;
3575 	u32 vmolr;
3576 
3577 	/* This register exists only on 82576 and newer so if we are older then
3578 	 * we should exit and do nothing
3579 	 */
3580 	if (hw->mac.type < e1000_82576)
3581 		return;
3582 
3583 	vmolr = rd32(E1000_VMOLR(vfn));
3584 	vmolr |= E1000_VMOLR_STRVLAN; /* Strip vlan tags */
3585 	if (hw->mac.type == e1000_i350) {
3586 		u32 dvmolr;
3587 
3588 		dvmolr = rd32(E1000_DVMOLR(vfn));
3589 		dvmolr |= E1000_DVMOLR_STRVLAN;
3590 		wr32(E1000_DVMOLR(vfn), dvmolr);
3591 	}
3592 	if (aupe)
3593 		vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
3594 	else
3595 		vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
3596 
3597 	/* clear all bits that might not be set */
3598 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
3599 
3600 	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
3601 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
3602 	/* for VMDq only allow the VFs and pool 0 to accept broadcast and
3603 	 * multicast packets
3604 	 */
3605 	if (vfn <= adapter->vfs_allocated_count)
3606 		vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
3607 
3608 	wr32(E1000_VMOLR(vfn), vmolr);
3609 }
3610 
3611 /**
3612  *  igb_configure_rx_ring - Configure a receive ring after Reset
3613  *  @adapter: board private structure
3614  *  @ring: receive ring to be configured
3615  *
3616  *  Configure the Rx unit of the MAC after a reset.
3617  **/
3618 void igb_configure_rx_ring(struct igb_adapter *adapter,
3619 			   struct igb_ring *ring)
3620 {
3621 	struct e1000_hw *hw = &adapter->hw;
3622 	u64 rdba = ring->dma;
3623 	int reg_idx = ring->reg_idx;
3624 	u32 srrctl = 0, rxdctl = 0;
3625 
3626 	/* disable the queue */
3627 	wr32(E1000_RXDCTL(reg_idx), 0);
3628 
3629 	/* Set DMA base address registers */
3630 	wr32(E1000_RDBAL(reg_idx),
3631 	     rdba & 0x00000000ffffffffULL);
3632 	wr32(E1000_RDBAH(reg_idx), rdba >> 32);
3633 	wr32(E1000_RDLEN(reg_idx),
3634 	     ring->count * sizeof(union e1000_adv_rx_desc));
3635 
3636 	/* initialize head and tail */
3637 	ring->tail = hw->hw_addr + E1000_RDT(reg_idx);
3638 	wr32(E1000_RDH(reg_idx), 0);
3639 	writel(0, ring->tail);
3640 
3641 	/* set descriptor configuration */
3642 	srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
3643 	srrctl |= IGB_RX_BUFSZ >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3644 	srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3645 	if (hw->mac.type >= e1000_82580)
3646 		srrctl |= E1000_SRRCTL_TIMESTAMP;
3647 	/* Only set Drop Enable if we are supporting multiple queues */
3648 	if (adapter->vfs_allocated_count || adapter->num_rx_queues > 1)
3649 		srrctl |= E1000_SRRCTL_DROP_EN;
3650 
3651 	wr32(E1000_SRRCTL(reg_idx), srrctl);
3652 
3653 	/* set filtering for VMDQ pools */
3654 	igb_set_vmolr(adapter, reg_idx & 0x7, true);
3655 
3656 	rxdctl |= IGB_RX_PTHRESH;
3657 	rxdctl |= IGB_RX_HTHRESH << 8;
3658 	rxdctl |= IGB_RX_WTHRESH << 16;
3659 
3660 	/* enable receive descriptor fetching */
3661 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3662 	wr32(E1000_RXDCTL(reg_idx), rxdctl);
3663 }
3664 
3665 /**
3666  *  igb_configure_rx - Configure receive Unit after Reset
3667  *  @adapter: board private structure
3668  *
3669  *  Configure the Rx unit of the MAC after a reset.
3670  **/
3671 static void igb_configure_rx(struct igb_adapter *adapter)
3672 {
3673 	int i;
3674 
3675 	/* set UTA to appropriate mode */
3676 	igb_set_uta(adapter);
3677 
3678 	/* set the correct pool for the PF default MAC address in entry 0 */
3679 	igb_rar_set_qsel(adapter, adapter->hw.mac.addr, 0,
3680 			 adapter->vfs_allocated_count);
3681 
3682 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3683 	 * the Base and Length of the Rx Descriptor Ring
3684 	 */
3685 	for (i = 0; i < adapter->num_rx_queues; i++)
3686 		igb_configure_rx_ring(adapter, adapter->rx_ring[i]);
3687 }
3688 
3689 /**
3690  *  igb_free_tx_resources - Free Tx Resources per Queue
3691  *  @tx_ring: Tx descriptor ring for a specific queue
3692  *
3693  *  Free all transmit software resources
3694  **/
3695 void igb_free_tx_resources(struct igb_ring *tx_ring)
3696 {
3697 	igb_clean_tx_ring(tx_ring);
3698 
3699 	vfree(tx_ring->tx_buffer_info);
3700 	tx_ring->tx_buffer_info = NULL;
3701 
3702 	/* if not set, then don't free */
3703 	if (!tx_ring->desc)
3704 		return;
3705 
3706 	dma_free_coherent(tx_ring->dev, tx_ring->size,
3707 			  tx_ring->desc, tx_ring->dma);
3708 
3709 	tx_ring->desc = NULL;
3710 }
3711 
3712 /**
3713  *  igb_free_all_tx_resources - Free Tx Resources for All Queues
3714  *  @adapter: board private structure
3715  *
3716  *  Free all transmit software resources
3717  **/
3718 static void igb_free_all_tx_resources(struct igb_adapter *adapter)
3719 {
3720 	int i;
3721 
3722 	for (i = 0; i < adapter->num_tx_queues; i++)
3723 		if (adapter->tx_ring[i])
3724 			igb_free_tx_resources(adapter->tx_ring[i]);
3725 }
3726 
3727 void igb_unmap_and_free_tx_resource(struct igb_ring *ring,
3728 				    struct igb_tx_buffer *tx_buffer)
3729 {
3730 	if (tx_buffer->skb) {
3731 		dev_kfree_skb_any(tx_buffer->skb);
3732 		if (dma_unmap_len(tx_buffer, len))
3733 			dma_unmap_single(ring->dev,
3734 					 dma_unmap_addr(tx_buffer, dma),
3735 					 dma_unmap_len(tx_buffer, len),
3736 					 DMA_TO_DEVICE);
3737 	} else if (dma_unmap_len(tx_buffer, len)) {
3738 		dma_unmap_page(ring->dev,
3739 			       dma_unmap_addr(tx_buffer, dma),
3740 			       dma_unmap_len(tx_buffer, len),
3741 			       DMA_TO_DEVICE);
3742 	}
3743 	tx_buffer->next_to_watch = NULL;
3744 	tx_buffer->skb = NULL;
3745 	dma_unmap_len_set(tx_buffer, len, 0);
3746 	/* buffer_info must be completely set up in the transmit path */
3747 }
3748 
3749 /**
3750  *  igb_clean_tx_ring - Free Tx Buffers
3751  *  @tx_ring: ring to be cleaned
3752  **/
3753 static void igb_clean_tx_ring(struct igb_ring *tx_ring)
3754 {
3755 	struct igb_tx_buffer *buffer_info;
3756 	unsigned long size;
3757 	u16 i;
3758 
3759 	if (!tx_ring->tx_buffer_info)
3760 		return;
3761 	/* Free all the Tx ring sk_buffs */
3762 
3763 	for (i = 0; i < tx_ring->count; i++) {
3764 		buffer_info = &tx_ring->tx_buffer_info[i];
3765 		igb_unmap_and_free_tx_resource(tx_ring, buffer_info);
3766 	}
3767 
3768 	netdev_tx_reset_queue(txring_txq(tx_ring));
3769 
3770 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3771 	memset(tx_ring->tx_buffer_info, 0, size);
3772 
3773 	/* Zero out the descriptor ring */
3774 	memset(tx_ring->desc, 0, tx_ring->size);
3775 
3776 	tx_ring->next_to_use = 0;
3777 	tx_ring->next_to_clean = 0;
3778 }
3779 
3780 /**
3781  *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
3782  *  @adapter: board private structure
3783  **/
3784 static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
3785 {
3786 	int i;
3787 
3788 	for (i = 0; i < adapter->num_tx_queues; i++)
3789 		if (adapter->tx_ring[i])
3790 			igb_clean_tx_ring(adapter->tx_ring[i]);
3791 }
3792 
3793 /**
3794  *  igb_free_rx_resources - Free Rx Resources
3795  *  @rx_ring: ring to clean the resources from
3796  *
3797  *  Free all receive software resources
3798  **/
3799 void igb_free_rx_resources(struct igb_ring *rx_ring)
3800 {
3801 	igb_clean_rx_ring(rx_ring);
3802 
3803 	vfree(rx_ring->rx_buffer_info);
3804 	rx_ring->rx_buffer_info = NULL;
3805 
3806 	/* if not set, then don't free */
3807 	if (!rx_ring->desc)
3808 		return;
3809 
3810 	dma_free_coherent(rx_ring->dev, rx_ring->size,
3811 			  rx_ring->desc, rx_ring->dma);
3812 
3813 	rx_ring->desc = NULL;
3814 }
3815 
3816 /**
3817  *  igb_free_all_rx_resources - Free Rx Resources for All Queues
3818  *  @adapter: board private structure
3819  *
3820  *  Free all receive software resources
3821  **/
3822 static void igb_free_all_rx_resources(struct igb_adapter *adapter)
3823 {
3824 	int i;
3825 
3826 	for (i = 0; i < adapter->num_rx_queues; i++)
3827 		if (adapter->rx_ring[i])
3828 			igb_free_rx_resources(adapter->rx_ring[i]);
3829 }
3830 
3831 /**
3832  *  igb_clean_rx_ring - Free Rx Buffers per Queue
3833  *  @rx_ring: ring to free buffers from
3834  **/
3835 static void igb_clean_rx_ring(struct igb_ring *rx_ring)
3836 {
3837 	unsigned long size;
3838 	u16 i;
3839 
3840 	if (rx_ring->skb)
3841 		dev_kfree_skb(rx_ring->skb);
3842 	rx_ring->skb = NULL;
3843 
3844 	if (!rx_ring->rx_buffer_info)
3845 		return;
3846 
3847 	/* Free all the Rx ring sk_buffs */
3848 	for (i = 0; i < rx_ring->count; i++) {
3849 		struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
3850 
3851 		if (!buffer_info->page)
3852 			continue;
3853 
3854 		dma_unmap_page(rx_ring->dev,
3855 			       buffer_info->dma,
3856 			       PAGE_SIZE,
3857 			       DMA_FROM_DEVICE);
3858 		__free_page(buffer_info->page);
3859 
3860 		buffer_info->page = NULL;
3861 	}
3862 
3863 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3864 	memset(rx_ring->rx_buffer_info, 0, size);
3865 
3866 	/* Zero out the descriptor ring */
3867 	memset(rx_ring->desc, 0, rx_ring->size);
3868 
3869 	rx_ring->next_to_alloc = 0;
3870 	rx_ring->next_to_clean = 0;
3871 	rx_ring->next_to_use = 0;
3872 }
3873 
3874 /**
3875  *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
3876  *  @adapter: board private structure
3877  **/
3878 static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
3879 {
3880 	int i;
3881 
3882 	for (i = 0; i < adapter->num_rx_queues; i++)
3883 		if (adapter->rx_ring[i])
3884 			igb_clean_rx_ring(adapter->rx_ring[i]);
3885 }
3886 
3887 /**
3888  *  igb_set_mac - Change the Ethernet Address of the NIC
3889  *  @netdev: network interface device structure
3890  *  @p: pointer to an address structure
3891  *
3892  *  Returns 0 on success, negative on failure
3893  **/
3894 static int igb_set_mac(struct net_device *netdev, void *p)
3895 {
3896 	struct igb_adapter *adapter = netdev_priv(netdev);
3897 	struct e1000_hw *hw = &adapter->hw;
3898 	struct sockaddr *addr = p;
3899 
3900 	if (!is_valid_ether_addr(addr->sa_data))
3901 		return -EADDRNOTAVAIL;
3902 
3903 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
3904 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
3905 
3906 	/* set the correct pool for the new PF MAC address in entry 0 */
3907 	igb_rar_set_qsel(adapter, hw->mac.addr, 0,
3908 			 adapter->vfs_allocated_count);
3909 
3910 	return 0;
3911 }
3912 
3913 /**
3914  *  igb_write_mc_addr_list - write multicast addresses to MTA
3915  *  @netdev: network interface device structure
3916  *
3917  *  Writes multicast address list to the MTA hash table.
3918  *  Returns: -ENOMEM on failure
3919  *           0 on no addresses written
3920  *           X on writing X addresses to MTA
3921  **/
3922 static int igb_write_mc_addr_list(struct net_device *netdev)
3923 {
3924 	struct igb_adapter *adapter = netdev_priv(netdev);
3925 	struct e1000_hw *hw = &adapter->hw;
3926 	struct netdev_hw_addr *ha;
3927 	u8  *mta_list;
3928 	int i;
3929 
3930 	if (netdev_mc_empty(netdev)) {
3931 		/* nothing to program, so clear mc list */
3932 		igb_update_mc_addr_list(hw, NULL, 0);
3933 		igb_restore_vf_multicasts(adapter);
3934 		return 0;
3935 	}
3936 
3937 	mta_list = kzalloc(netdev_mc_count(netdev) * 6, GFP_ATOMIC);
3938 	if (!mta_list)
3939 		return -ENOMEM;
3940 
3941 	/* The shared function expects a packed array of only addresses. */
3942 	i = 0;
3943 	netdev_for_each_mc_addr(ha, netdev)
3944 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3945 
3946 	igb_update_mc_addr_list(hw, mta_list, i);
3947 	kfree(mta_list);
3948 
3949 	return netdev_mc_count(netdev);
3950 }
3951 
3952 /**
3953  *  igb_write_uc_addr_list - write unicast addresses to RAR table
3954  *  @netdev: network interface device structure
3955  *
3956  *  Writes unicast address list to the RAR table.
3957  *  Returns: -ENOMEM on failure/insufficient address space
3958  *           0 on no addresses written
3959  *           X on writing X addresses to the RAR table
3960  **/
3961 static int igb_write_uc_addr_list(struct net_device *netdev)
3962 {
3963 	struct igb_adapter *adapter = netdev_priv(netdev);
3964 	struct e1000_hw *hw = &adapter->hw;
3965 	unsigned int vfn = adapter->vfs_allocated_count;
3966 	unsigned int rar_entries = hw->mac.rar_entry_count - (vfn + 1);
3967 	int count = 0;
3968 
3969 	/* return ENOMEM indicating insufficient memory for addresses */
3970 	if (netdev_uc_count(netdev) > rar_entries)
3971 		return -ENOMEM;
3972 
3973 	if (!netdev_uc_empty(netdev) && rar_entries) {
3974 		struct netdev_hw_addr *ha;
3975 
3976 		netdev_for_each_uc_addr(ha, netdev) {
3977 			if (!rar_entries)
3978 				break;
3979 			igb_rar_set_qsel(adapter, ha->addr,
3980 					 rar_entries--,
3981 					 vfn);
3982 			count++;
3983 		}
3984 	}
3985 	/* write the addresses in reverse order to avoid write combining */
3986 	for (; rar_entries > 0 ; rar_entries--) {
3987 		wr32(E1000_RAH(rar_entries), 0);
3988 		wr32(E1000_RAL(rar_entries), 0);
3989 	}
3990 	wrfl();
3991 
3992 	return count;
3993 }
3994 
3995 /**
3996  *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
3997  *  @netdev: network interface device structure
3998  *
3999  *  The set_rx_mode entry point is called whenever the unicast or multicast
4000  *  address lists or the network interface flags are updated.  This routine is
4001  *  responsible for configuring the hardware for proper unicast, multicast,
4002  *  promiscuous mode, and all-multi behavior.
4003  **/
4004 static void igb_set_rx_mode(struct net_device *netdev)
4005 {
4006 	struct igb_adapter *adapter = netdev_priv(netdev);
4007 	struct e1000_hw *hw = &adapter->hw;
4008 	unsigned int vfn = adapter->vfs_allocated_count;
4009 	u32 rctl, vmolr = 0;
4010 	int count;
4011 
4012 	/* Check for Promiscuous and All Multicast modes */
4013 	rctl = rd32(E1000_RCTL);
4014 
4015 	/* clear the effected bits */
4016 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE | E1000_RCTL_VFE);
4017 
4018 	if (netdev->flags & IFF_PROMISC) {
4019 		/* retain VLAN HW filtering if in VT mode */
4020 		if (adapter->vfs_allocated_count)
4021 			rctl |= E1000_RCTL_VFE;
4022 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
4023 		vmolr |= (E1000_VMOLR_ROPE | E1000_VMOLR_MPME);
4024 	} else {
4025 		if (netdev->flags & IFF_ALLMULTI) {
4026 			rctl |= E1000_RCTL_MPE;
4027 			vmolr |= E1000_VMOLR_MPME;
4028 		} else {
4029 			/* Write addresses to the MTA, if the attempt fails
4030 			 * then we should just turn on promiscuous mode so
4031 			 * that we can at least receive multicast traffic
4032 			 */
4033 			count = igb_write_mc_addr_list(netdev);
4034 			if (count < 0) {
4035 				rctl |= E1000_RCTL_MPE;
4036 				vmolr |= E1000_VMOLR_MPME;
4037 			} else if (count) {
4038 				vmolr |= E1000_VMOLR_ROMPE;
4039 			}
4040 		}
4041 		/* Write addresses to available RAR registers, if there is not
4042 		 * sufficient space to store all the addresses then enable
4043 		 * unicast promiscuous mode
4044 		 */
4045 		count = igb_write_uc_addr_list(netdev);
4046 		if (count < 0) {
4047 			rctl |= E1000_RCTL_UPE;
4048 			vmolr |= E1000_VMOLR_ROPE;
4049 		}
4050 		rctl |= E1000_RCTL_VFE;
4051 	}
4052 	wr32(E1000_RCTL, rctl);
4053 
4054 	/* In order to support SR-IOV and eventually VMDq it is necessary to set
4055 	 * the VMOLR to enable the appropriate modes.  Without this workaround
4056 	 * we will have issues with VLAN tag stripping not being done for frames
4057 	 * that are only arriving because we are the default pool
4058 	 */
4059 	if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
4060 		return;
4061 
4062 	vmolr |= rd32(E1000_VMOLR(vfn)) &
4063 		 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
4064 	wr32(E1000_VMOLR(vfn), vmolr);
4065 	igb_restore_vf_multicasts(adapter);
4066 }
4067 
4068 static void igb_check_wvbr(struct igb_adapter *adapter)
4069 {
4070 	struct e1000_hw *hw = &adapter->hw;
4071 	u32 wvbr = 0;
4072 
4073 	switch (hw->mac.type) {
4074 	case e1000_82576:
4075 	case e1000_i350:
4076 		wvbr = rd32(E1000_WVBR);
4077 		if (!wvbr)
4078 			return;
4079 		break;
4080 	default:
4081 		break;
4082 	}
4083 
4084 	adapter->wvbr |= wvbr;
4085 }
4086 
4087 #define IGB_STAGGERED_QUEUE_OFFSET 8
4088 
4089 static void igb_spoof_check(struct igb_adapter *adapter)
4090 {
4091 	int j;
4092 
4093 	if (!adapter->wvbr)
4094 		return;
4095 
4096 	for (j = 0; j < adapter->vfs_allocated_count; j++) {
4097 		if (adapter->wvbr & (1 << j) ||
4098 		    adapter->wvbr & (1 << (j + IGB_STAGGERED_QUEUE_OFFSET))) {
4099 			dev_warn(&adapter->pdev->dev,
4100 				"Spoof event(s) detected on VF %d\n", j);
4101 			adapter->wvbr &=
4102 				~((1 << j) |
4103 				  (1 << (j + IGB_STAGGERED_QUEUE_OFFSET)));
4104 		}
4105 	}
4106 }
4107 
4108 /* Need to wait a few seconds after link up to get diagnostic information from
4109  * the phy
4110  */
4111 static void igb_update_phy_info(unsigned long data)
4112 {
4113 	struct igb_adapter *adapter = (struct igb_adapter *) data;
4114 	igb_get_phy_info(&adapter->hw);
4115 }
4116 
4117 /**
4118  *  igb_has_link - check shared code for link and determine up/down
4119  *  @adapter: pointer to driver private info
4120  **/
4121 bool igb_has_link(struct igb_adapter *adapter)
4122 {
4123 	struct e1000_hw *hw = &adapter->hw;
4124 	bool link_active = false;
4125 
4126 	/* get_link_status is set on LSC (link status) interrupt or
4127 	 * rx sequence error interrupt.  get_link_status will stay
4128 	 * false until the e1000_check_for_link establishes link
4129 	 * for copper adapters ONLY
4130 	 */
4131 	switch (hw->phy.media_type) {
4132 	case e1000_media_type_copper:
4133 		if (!hw->mac.get_link_status)
4134 			return true;
4135 	case e1000_media_type_internal_serdes:
4136 		hw->mac.ops.check_for_link(hw);
4137 		link_active = !hw->mac.get_link_status;
4138 		break;
4139 	default:
4140 	case e1000_media_type_unknown:
4141 		break;
4142 	}
4143 
4144 	if (((hw->mac.type == e1000_i210) ||
4145 	     (hw->mac.type == e1000_i211)) &&
4146 	     (hw->phy.id == I210_I_PHY_ID)) {
4147 		if (!netif_carrier_ok(adapter->netdev)) {
4148 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4149 		} else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
4150 			adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
4151 			adapter->link_check_timeout = jiffies;
4152 		}
4153 	}
4154 
4155 	return link_active;
4156 }
4157 
4158 static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
4159 {
4160 	bool ret = false;
4161 	u32 ctrl_ext, thstat;
4162 
4163 	/* check for thermal sensor event on i350 copper only */
4164 	if (hw->mac.type == e1000_i350) {
4165 		thstat = rd32(E1000_THSTAT);
4166 		ctrl_ext = rd32(E1000_CTRL_EXT);
4167 
4168 		if ((hw->phy.media_type == e1000_media_type_copper) &&
4169 		    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
4170 			ret = !!(thstat & event);
4171 	}
4172 
4173 	return ret;
4174 }
4175 
4176 /**
4177  *  igb_check_lvmmc - check for malformed packets received
4178  *  and indicated in LVMMC register
4179  *  @adapter: pointer to adapter
4180  **/
4181 static void igb_check_lvmmc(struct igb_adapter *adapter)
4182 {
4183 	struct e1000_hw *hw = &adapter->hw;
4184 	u32 lvmmc;
4185 
4186 	lvmmc = rd32(E1000_LVMMC);
4187 	if (lvmmc) {
4188 		if (unlikely(net_ratelimit())) {
4189 			netdev_warn(adapter->netdev,
4190 				    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
4191 				    lvmmc);
4192 		}
4193 	}
4194 }
4195 
4196 /**
4197  *  igb_watchdog - Timer Call-back
4198  *  @data: pointer to adapter cast into an unsigned long
4199  **/
4200 static void igb_watchdog(unsigned long data)
4201 {
4202 	struct igb_adapter *adapter = (struct igb_adapter *)data;
4203 	/* Do the rest outside of interrupt context */
4204 	schedule_work(&adapter->watchdog_task);
4205 }
4206 
4207 static void igb_watchdog_task(struct work_struct *work)
4208 {
4209 	struct igb_adapter *adapter = container_of(work,
4210 						   struct igb_adapter,
4211 						   watchdog_task);
4212 	struct e1000_hw *hw = &adapter->hw;
4213 	struct e1000_phy_info *phy = &hw->phy;
4214 	struct net_device *netdev = adapter->netdev;
4215 	u32 link;
4216 	int i;
4217 	u32 connsw;
4218 
4219 	link = igb_has_link(adapter);
4220 
4221 	if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
4222 		if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
4223 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4224 		else
4225 			link = false;
4226 	}
4227 
4228 	/* Force link down if we have fiber to swap to */
4229 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4230 		if (hw->phy.media_type == e1000_media_type_copper) {
4231 			connsw = rd32(E1000_CONNSW);
4232 			if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
4233 				link = 0;
4234 		}
4235 	}
4236 	if (link) {
4237 		/* Perform a reset if the media type changed. */
4238 		if (hw->dev_spec._82575.media_changed) {
4239 			hw->dev_spec._82575.media_changed = false;
4240 			adapter->flags |= IGB_FLAG_MEDIA_RESET;
4241 			igb_reset(adapter);
4242 		}
4243 		/* Cancel scheduled suspend requests. */
4244 		pm_runtime_resume(netdev->dev.parent);
4245 
4246 		if (!netif_carrier_ok(netdev)) {
4247 			u32 ctrl;
4248 
4249 			hw->mac.ops.get_speed_and_duplex(hw,
4250 							 &adapter->link_speed,
4251 							 &adapter->link_duplex);
4252 
4253 			ctrl = rd32(E1000_CTRL);
4254 			/* Links status message must follow this format */
4255 			netdev_info(netdev,
4256 			       "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4257 			       netdev->name,
4258 			       adapter->link_speed,
4259 			       adapter->link_duplex == FULL_DUPLEX ?
4260 			       "Full" : "Half",
4261 			       (ctrl & E1000_CTRL_TFCE) &&
4262 			       (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
4263 			       (ctrl & E1000_CTRL_RFCE) ?  "RX" :
4264 			       (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
4265 
4266 			/* disable EEE if enabled */
4267 			if ((adapter->flags & IGB_FLAG_EEE) &&
4268 				(adapter->link_duplex == HALF_DUPLEX)) {
4269 				dev_info(&adapter->pdev->dev,
4270 				"EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
4271 				adapter->hw.dev_spec._82575.eee_disable = true;
4272 				adapter->flags &= ~IGB_FLAG_EEE;
4273 			}
4274 
4275 			/* check if SmartSpeed worked */
4276 			igb_check_downshift(hw);
4277 			if (phy->speed_downgraded)
4278 				netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
4279 
4280 			/* check for thermal sensor event */
4281 			if (igb_thermal_sensor_event(hw,
4282 			    E1000_THSTAT_LINK_THROTTLE))
4283 				netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
4284 
4285 			/* adjust timeout factor according to speed/duplex */
4286 			adapter->tx_timeout_factor = 1;
4287 			switch (adapter->link_speed) {
4288 			case SPEED_10:
4289 				adapter->tx_timeout_factor = 14;
4290 				break;
4291 			case SPEED_100:
4292 				/* maybe add some timeout factor ? */
4293 				break;
4294 			}
4295 
4296 			netif_carrier_on(netdev);
4297 
4298 			igb_ping_all_vfs(adapter);
4299 			igb_check_vf_rate_limit(adapter);
4300 
4301 			/* link state has changed, schedule phy info update */
4302 			if (!test_bit(__IGB_DOWN, &adapter->state))
4303 				mod_timer(&adapter->phy_info_timer,
4304 					  round_jiffies(jiffies + 2 * HZ));
4305 		}
4306 	} else {
4307 		if (netif_carrier_ok(netdev)) {
4308 			adapter->link_speed = 0;
4309 			adapter->link_duplex = 0;
4310 
4311 			/* check for thermal sensor event */
4312 			if (igb_thermal_sensor_event(hw,
4313 			    E1000_THSTAT_PWR_DOWN)) {
4314 				netdev_err(netdev, "The network adapter was stopped because it overheated\n");
4315 			}
4316 
4317 			/* Links status message must follow this format */
4318 			netdev_info(netdev, "igb: %s NIC Link is Down\n",
4319 			       netdev->name);
4320 			netif_carrier_off(netdev);
4321 
4322 			igb_ping_all_vfs(adapter);
4323 
4324 			/* link state has changed, schedule phy info update */
4325 			if (!test_bit(__IGB_DOWN, &adapter->state))
4326 				mod_timer(&adapter->phy_info_timer,
4327 					  round_jiffies(jiffies + 2 * HZ));
4328 
4329 			/* link is down, time to check for alternate media */
4330 			if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4331 				igb_check_swap_media(adapter);
4332 				if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
4333 					schedule_work(&adapter->reset_task);
4334 					/* return immediately */
4335 					return;
4336 				}
4337 			}
4338 			pm_schedule_suspend(netdev->dev.parent,
4339 					    MSEC_PER_SEC * 5);
4340 
4341 		/* also check for alternate media here */
4342 		} else if (!netif_carrier_ok(netdev) &&
4343 			   (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
4344 			igb_check_swap_media(adapter);
4345 			if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
4346 				schedule_work(&adapter->reset_task);
4347 				/* return immediately */
4348 				return;
4349 			}
4350 		}
4351 	}
4352 
4353 	spin_lock(&adapter->stats64_lock);
4354 	igb_update_stats(adapter, &adapter->stats64);
4355 	spin_unlock(&adapter->stats64_lock);
4356 
4357 	for (i = 0; i < adapter->num_tx_queues; i++) {
4358 		struct igb_ring *tx_ring = adapter->tx_ring[i];
4359 		if (!netif_carrier_ok(netdev)) {
4360 			/* We've lost link, so the controller stops DMA,
4361 			 * but we've got queued Tx work that's never going
4362 			 * to get done, so reset controller to flush Tx.
4363 			 * (Do the reset outside of interrupt context).
4364 			 */
4365 			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
4366 				adapter->tx_timeout_count++;
4367 				schedule_work(&adapter->reset_task);
4368 				/* return immediately since reset is imminent */
4369 				return;
4370 			}
4371 		}
4372 
4373 		/* Force detection of hung controller every watchdog period */
4374 		set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
4375 	}
4376 
4377 	/* Cause software interrupt to ensure Rx ring is cleaned */
4378 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
4379 		u32 eics = 0;
4380 
4381 		for (i = 0; i < adapter->num_q_vectors; i++)
4382 			eics |= adapter->q_vector[i]->eims_value;
4383 		wr32(E1000_EICS, eics);
4384 	} else {
4385 		wr32(E1000_ICS, E1000_ICS_RXDMT0);
4386 	}
4387 
4388 	igb_spoof_check(adapter);
4389 	igb_ptp_rx_hang(adapter);
4390 
4391 	/* Check LVMMC register on i350/i354 only */
4392 	if ((adapter->hw.mac.type == e1000_i350) ||
4393 	    (adapter->hw.mac.type == e1000_i354))
4394 		igb_check_lvmmc(adapter);
4395 
4396 	/* Reset the timer */
4397 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
4398 		if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
4399 			mod_timer(&adapter->watchdog_timer,
4400 				  round_jiffies(jiffies +  HZ));
4401 		else
4402 			mod_timer(&adapter->watchdog_timer,
4403 				  round_jiffies(jiffies + 2 * HZ));
4404 	}
4405 }
4406 
4407 enum latency_range {
4408 	lowest_latency = 0,
4409 	low_latency = 1,
4410 	bulk_latency = 2,
4411 	latency_invalid = 255
4412 };
4413 
4414 /**
4415  *  igb_update_ring_itr - update the dynamic ITR value based on packet size
4416  *  @q_vector: pointer to q_vector
4417  *
4418  *  Stores a new ITR value based on strictly on packet size.  This
4419  *  algorithm is less sophisticated than that used in igb_update_itr,
4420  *  due to the difficulty of synchronizing statistics across multiple
4421  *  receive rings.  The divisors and thresholds used by this function
4422  *  were determined based on theoretical maximum wire speed and testing
4423  *  data, in order to minimize response time while increasing bulk
4424  *  throughput.
4425  *  This functionality is controlled by ethtool's coalescing settings.
4426  *  NOTE:  This function is called only when operating in a multiqueue
4427  *         receive environment.
4428  **/
4429 static void igb_update_ring_itr(struct igb_q_vector *q_vector)
4430 {
4431 	int new_val = q_vector->itr_val;
4432 	int avg_wire_size = 0;
4433 	struct igb_adapter *adapter = q_vector->adapter;
4434 	unsigned int packets;
4435 
4436 	/* For non-gigabit speeds, just fix the interrupt rate at 4000
4437 	 * ints/sec - ITR timer value of 120 ticks.
4438 	 */
4439 	if (adapter->link_speed != SPEED_1000) {
4440 		new_val = IGB_4K_ITR;
4441 		goto set_itr_val;
4442 	}
4443 
4444 	packets = q_vector->rx.total_packets;
4445 	if (packets)
4446 		avg_wire_size = q_vector->rx.total_bytes / packets;
4447 
4448 	packets = q_vector->tx.total_packets;
4449 	if (packets)
4450 		avg_wire_size = max_t(u32, avg_wire_size,
4451 				      q_vector->tx.total_bytes / packets);
4452 
4453 	/* if avg_wire_size isn't set no work was done */
4454 	if (!avg_wire_size)
4455 		goto clear_counts;
4456 
4457 	/* Add 24 bytes to size to account for CRC, preamble, and gap */
4458 	avg_wire_size += 24;
4459 
4460 	/* Don't starve jumbo frames */
4461 	avg_wire_size = min(avg_wire_size, 3000);
4462 
4463 	/* Give a little boost to mid-size frames */
4464 	if ((avg_wire_size > 300) && (avg_wire_size < 1200))
4465 		new_val = avg_wire_size / 3;
4466 	else
4467 		new_val = avg_wire_size / 2;
4468 
4469 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
4470 	if (new_val < IGB_20K_ITR &&
4471 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
4472 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
4473 		new_val = IGB_20K_ITR;
4474 
4475 set_itr_val:
4476 	if (new_val != q_vector->itr_val) {
4477 		q_vector->itr_val = new_val;
4478 		q_vector->set_itr = 1;
4479 	}
4480 clear_counts:
4481 	q_vector->rx.total_bytes = 0;
4482 	q_vector->rx.total_packets = 0;
4483 	q_vector->tx.total_bytes = 0;
4484 	q_vector->tx.total_packets = 0;
4485 }
4486 
4487 /**
4488  *  igb_update_itr - update the dynamic ITR value based on statistics
4489  *  @q_vector: pointer to q_vector
4490  *  @ring_container: ring info to update the itr for
4491  *
4492  *  Stores a new ITR value based on packets and byte
4493  *  counts during the last interrupt.  The advantage of per interrupt
4494  *  computation is faster updates and more accurate ITR for the current
4495  *  traffic pattern.  Constants in this function were computed
4496  *  based on theoretical maximum wire speed and thresholds were set based
4497  *  on testing data as well as attempting to minimize response time
4498  *  while increasing bulk throughput.
4499  *  This functionality is controlled by ethtool's coalescing settings.
4500  *  NOTE:  These calculations are only valid when operating in a single-
4501  *         queue environment.
4502  **/
4503 static void igb_update_itr(struct igb_q_vector *q_vector,
4504 			   struct igb_ring_container *ring_container)
4505 {
4506 	unsigned int packets = ring_container->total_packets;
4507 	unsigned int bytes = ring_container->total_bytes;
4508 	u8 itrval = ring_container->itr;
4509 
4510 	/* no packets, exit with status unchanged */
4511 	if (packets == 0)
4512 		return;
4513 
4514 	switch (itrval) {
4515 	case lowest_latency:
4516 		/* handle TSO and jumbo frames */
4517 		if (bytes/packets > 8000)
4518 			itrval = bulk_latency;
4519 		else if ((packets < 5) && (bytes > 512))
4520 			itrval = low_latency;
4521 		break;
4522 	case low_latency:  /* 50 usec aka 20000 ints/s */
4523 		if (bytes > 10000) {
4524 			/* this if handles the TSO accounting */
4525 			if (bytes/packets > 8000)
4526 				itrval = bulk_latency;
4527 			else if ((packets < 10) || ((bytes/packets) > 1200))
4528 				itrval = bulk_latency;
4529 			else if ((packets > 35))
4530 				itrval = lowest_latency;
4531 		} else if (bytes/packets > 2000) {
4532 			itrval = bulk_latency;
4533 		} else if (packets <= 2 && bytes < 512) {
4534 			itrval = lowest_latency;
4535 		}
4536 		break;
4537 	case bulk_latency: /* 250 usec aka 4000 ints/s */
4538 		if (bytes > 25000) {
4539 			if (packets > 35)
4540 				itrval = low_latency;
4541 		} else if (bytes < 1500) {
4542 			itrval = low_latency;
4543 		}
4544 		break;
4545 	}
4546 
4547 	/* clear work counters since we have the values we need */
4548 	ring_container->total_bytes = 0;
4549 	ring_container->total_packets = 0;
4550 
4551 	/* write updated itr to ring container */
4552 	ring_container->itr = itrval;
4553 }
4554 
4555 static void igb_set_itr(struct igb_q_vector *q_vector)
4556 {
4557 	struct igb_adapter *adapter = q_vector->adapter;
4558 	u32 new_itr = q_vector->itr_val;
4559 	u8 current_itr = 0;
4560 
4561 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
4562 	if (adapter->link_speed != SPEED_1000) {
4563 		current_itr = 0;
4564 		new_itr = IGB_4K_ITR;
4565 		goto set_itr_now;
4566 	}
4567 
4568 	igb_update_itr(q_vector, &q_vector->tx);
4569 	igb_update_itr(q_vector, &q_vector->rx);
4570 
4571 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
4572 
4573 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
4574 	if (current_itr == lowest_latency &&
4575 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
4576 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
4577 		current_itr = low_latency;
4578 
4579 	switch (current_itr) {
4580 	/* counts and packets in update_itr are dependent on these numbers */
4581 	case lowest_latency:
4582 		new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
4583 		break;
4584 	case low_latency:
4585 		new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
4586 		break;
4587 	case bulk_latency:
4588 		new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
4589 		break;
4590 	default:
4591 		break;
4592 	}
4593 
4594 set_itr_now:
4595 	if (new_itr != q_vector->itr_val) {
4596 		/* this attempts to bias the interrupt rate towards Bulk
4597 		 * by adding intermediate steps when interrupt rate is
4598 		 * increasing
4599 		 */
4600 		new_itr = new_itr > q_vector->itr_val ?
4601 			  max((new_itr * q_vector->itr_val) /
4602 			  (new_itr + (q_vector->itr_val >> 2)),
4603 			  new_itr) : new_itr;
4604 		/* Don't write the value here; it resets the adapter's
4605 		 * internal timer, and causes us to delay far longer than
4606 		 * we should between interrupts.  Instead, we write the ITR
4607 		 * value at the beginning of the next interrupt so the timing
4608 		 * ends up being correct.
4609 		 */
4610 		q_vector->itr_val = new_itr;
4611 		q_vector->set_itr = 1;
4612 	}
4613 }
4614 
4615 static void igb_tx_ctxtdesc(struct igb_ring *tx_ring, u32 vlan_macip_lens,
4616 			    u32 type_tucmd, u32 mss_l4len_idx)
4617 {
4618 	struct e1000_adv_tx_context_desc *context_desc;
4619 	u16 i = tx_ring->next_to_use;
4620 
4621 	context_desc = IGB_TX_CTXTDESC(tx_ring, i);
4622 
4623 	i++;
4624 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
4625 
4626 	/* set bits to identify this as an advanced context descriptor */
4627 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
4628 
4629 	/* For 82575, context index must be unique per ring. */
4630 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
4631 		mss_l4len_idx |= tx_ring->reg_idx << 4;
4632 
4633 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
4634 	context_desc->seqnum_seed	= 0;
4635 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
4636 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
4637 }
4638 
4639 static int igb_tso(struct igb_ring *tx_ring,
4640 		   struct igb_tx_buffer *first,
4641 		   u8 *hdr_len)
4642 {
4643 	struct sk_buff *skb = first->skb;
4644 	u32 vlan_macip_lens, type_tucmd;
4645 	u32 mss_l4len_idx, l4len;
4646 	int err;
4647 
4648 	if (skb->ip_summed != CHECKSUM_PARTIAL)
4649 		return 0;
4650 
4651 	if (!skb_is_gso(skb))
4652 		return 0;
4653 
4654 	err = skb_cow_head(skb, 0);
4655 	if (err < 0)
4656 		return err;
4657 
4658 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
4659 	type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
4660 
4661 	if (first->protocol == htons(ETH_P_IP)) {
4662 		struct iphdr *iph = ip_hdr(skb);
4663 		iph->tot_len = 0;
4664 		iph->check = 0;
4665 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
4666 							 iph->daddr, 0,
4667 							 IPPROTO_TCP,
4668 							 0);
4669 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
4670 		first->tx_flags |= IGB_TX_FLAGS_TSO |
4671 				   IGB_TX_FLAGS_CSUM |
4672 				   IGB_TX_FLAGS_IPV4;
4673 	} else if (skb_is_gso_v6(skb)) {
4674 		ipv6_hdr(skb)->payload_len = 0;
4675 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
4676 						       &ipv6_hdr(skb)->daddr,
4677 						       0, IPPROTO_TCP, 0);
4678 		first->tx_flags |= IGB_TX_FLAGS_TSO |
4679 				   IGB_TX_FLAGS_CSUM;
4680 	}
4681 
4682 	/* compute header lengths */
4683 	l4len = tcp_hdrlen(skb);
4684 	*hdr_len = skb_transport_offset(skb) + l4len;
4685 
4686 	/* update gso size and bytecount with header size */
4687 	first->gso_segs = skb_shinfo(skb)->gso_segs;
4688 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
4689 
4690 	/* MSS L4LEN IDX */
4691 	mss_l4len_idx = l4len << E1000_ADVTXD_L4LEN_SHIFT;
4692 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
4693 
4694 	/* VLAN MACLEN IPLEN */
4695 	vlan_macip_lens = skb_network_header_len(skb);
4696 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
4697 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
4698 
4699 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
4700 
4701 	return 1;
4702 }
4703 
4704 static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
4705 {
4706 	struct sk_buff *skb = first->skb;
4707 	u32 vlan_macip_lens = 0;
4708 	u32 mss_l4len_idx = 0;
4709 	u32 type_tucmd = 0;
4710 
4711 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
4712 		if (!(first->tx_flags & IGB_TX_FLAGS_VLAN))
4713 			return;
4714 	} else {
4715 		u8 l4_hdr = 0;
4716 
4717 		switch (first->protocol) {
4718 		case htons(ETH_P_IP):
4719 			vlan_macip_lens |= skb_network_header_len(skb);
4720 			type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
4721 			l4_hdr = ip_hdr(skb)->protocol;
4722 			break;
4723 		case htons(ETH_P_IPV6):
4724 			vlan_macip_lens |= skb_network_header_len(skb);
4725 			l4_hdr = ipv6_hdr(skb)->nexthdr;
4726 			break;
4727 		default:
4728 			if (unlikely(net_ratelimit())) {
4729 				dev_warn(tx_ring->dev,
4730 					 "partial checksum but proto=%x!\n",
4731 					 first->protocol);
4732 			}
4733 			break;
4734 		}
4735 
4736 		switch (l4_hdr) {
4737 		case IPPROTO_TCP:
4738 			type_tucmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
4739 			mss_l4len_idx = tcp_hdrlen(skb) <<
4740 					E1000_ADVTXD_L4LEN_SHIFT;
4741 			break;
4742 		case IPPROTO_SCTP:
4743 			type_tucmd |= E1000_ADVTXD_TUCMD_L4T_SCTP;
4744 			mss_l4len_idx = sizeof(struct sctphdr) <<
4745 					E1000_ADVTXD_L4LEN_SHIFT;
4746 			break;
4747 		case IPPROTO_UDP:
4748 			mss_l4len_idx = sizeof(struct udphdr) <<
4749 					E1000_ADVTXD_L4LEN_SHIFT;
4750 			break;
4751 		default:
4752 			if (unlikely(net_ratelimit())) {
4753 				dev_warn(tx_ring->dev,
4754 					 "partial checksum but l4 proto=%x!\n",
4755 					 l4_hdr);
4756 			}
4757 			break;
4758 		}
4759 
4760 		/* update TX checksum flag */
4761 		first->tx_flags |= IGB_TX_FLAGS_CSUM;
4762 	}
4763 
4764 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
4765 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
4766 
4767 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
4768 }
4769 
4770 #define IGB_SET_FLAG(_input, _flag, _result) \
4771 	((_flag <= _result) ? \
4772 	 ((u32)(_input & _flag) * (_result / _flag)) : \
4773 	 ((u32)(_input & _flag) / (_flag / _result)))
4774 
4775 static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
4776 {
4777 	/* set type for advanced descriptor with frame checksum insertion */
4778 	u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
4779 		       E1000_ADVTXD_DCMD_DEXT |
4780 		       E1000_ADVTXD_DCMD_IFCS;
4781 
4782 	/* set HW vlan bit if vlan is present */
4783 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
4784 				 (E1000_ADVTXD_DCMD_VLE));
4785 
4786 	/* set segmentation bits for TSO */
4787 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
4788 				 (E1000_ADVTXD_DCMD_TSE));
4789 
4790 	/* set timestamp bit if present */
4791 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
4792 				 (E1000_ADVTXD_MAC_TSTAMP));
4793 
4794 	/* insert frame checksum */
4795 	cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
4796 
4797 	return cmd_type;
4798 }
4799 
4800 static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
4801 				 union e1000_adv_tx_desc *tx_desc,
4802 				 u32 tx_flags, unsigned int paylen)
4803 {
4804 	u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
4805 
4806 	/* 82575 requires a unique index per ring */
4807 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
4808 		olinfo_status |= tx_ring->reg_idx << 4;
4809 
4810 	/* insert L4 checksum */
4811 	olinfo_status |= IGB_SET_FLAG(tx_flags,
4812 				      IGB_TX_FLAGS_CSUM,
4813 				      (E1000_TXD_POPTS_TXSM << 8));
4814 
4815 	/* insert IPv4 checksum */
4816 	olinfo_status |= IGB_SET_FLAG(tx_flags,
4817 				      IGB_TX_FLAGS_IPV4,
4818 				      (E1000_TXD_POPTS_IXSM << 8));
4819 
4820 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
4821 }
4822 
4823 static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
4824 {
4825 	struct net_device *netdev = tx_ring->netdev;
4826 
4827 	netif_stop_subqueue(netdev, tx_ring->queue_index);
4828 
4829 	/* Herbert's original patch had:
4830 	 *  smp_mb__after_netif_stop_queue();
4831 	 * but since that doesn't exist yet, just open code it.
4832 	 */
4833 	smp_mb();
4834 
4835 	/* We need to check again in a case another CPU has just
4836 	 * made room available.
4837 	 */
4838 	if (igb_desc_unused(tx_ring) < size)
4839 		return -EBUSY;
4840 
4841 	/* A reprieve! */
4842 	netif_wake_subqueue(netdev, tx_ring->queue_index);
4843 
4844 	u64_stats_update_begin(&tx_ring->tx_syncp2);
4845 	tx_ring->tx_stats.restart_queue2++;
4846 	u64_stats_update_end(&tx_ring->tx_syncp2);
4847 
4848 	return 0;
4849 }
4850 
4851 static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
4852 {
4853 	if (igb_desc_unused(tx_ring) >= size)
4854 		return 0;
4855 	return __igb_maybe_stop_tx(tx_ring, size);
4856 }
4857 
4858 static void igb_tx_map(struct igb_ring *tx_ring,
4859 		       struct igb_tx_buffer *first,
4860 		       const u8 hdr_len)
4861 {
4862 	struct sk_buff *skb = first->skb;
4863 	struct igb_tx_buffer *tx_buffer;
4864 	union e1000_adv_tx_desc *tx_desc;
4865 	struct skb_frag_struct *frag;
4866 	dma_addr_t dma;
4867 	unsigned int data_len, size;
4868 	u32 tx_flags = first->tx_flags;
4869 	u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
4870 	u16 i = tx_ring->next_to_use;
4871 
4872 	tx_desc = IGB_TX_DESC(tx_ring, i);
4873 
4874 	igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
4875 
4876 	size = skb_headlen(skb);
4877 	data_len = skb->data_len;
4878 
4879 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
4880 
4881 	tx_buffer = first;
4882 
4883 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
4884 		if (dma_mapping_error(tx_ring->dev, dma))
4885 			goto dma_error;
4886 
4887 		/* record length, and DMA address */
4888 		dma_unmap_len_set(tx_buffer, len, size);
4889 		dma_unmap_addr_set(tx_buffer, dma, dma);
4890 
4891 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
4892 
4893 		while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
4894 			tx_desc->read.cmd_type_len =
4895 				cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
4896 
4897 			i++;
4898 			tx_desc++;
4899 			if (i == tx_ring->count) {
4900 				tx_desc = IGB_TX_DESC(tx_ring, 0);
4901 				i = 0;
4902 			}
4903 			tx_desc->read.olinfo_status = 0;
4904 
4905 			dma += IGB_MAX_DATA_PER_TXD;
4906 			size -= IGB_MAX_DATA_PER_TXD;
4907 
4908 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
4909 		}
4910 
4911 		if (likely(!data_len))
4912 			break;
4913 
4914 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
4915 
4916 		i++;
4917 		tx_desc++;
4918 		if (i == tx_ring->count) {
4919 			tx_desc = IGB_TX_DESC(tx_ring, 0);
4920 			i = 0;
4921 		}
4922 		tx_desc->read.olinfo_status = 0;
4923 
4924 		size = skb_frag_size(frag);
4925 		data_len -= size;
4926 
4927 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
4928 				       size, DMA_TO_DEVICE);
4929 
4930 		tx_buffer = &tx_ring->tx_buffer_info[i];
4931 	}
4932 
4933 	/* write last descriptor with RS and EOP bits */
4934 	cmd_type |= size | IGB_TXD_DCMD;
4935 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
4936 
4937 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
4938 
4939 	/* set the timestamp */
4940 	first->time_stamp = jiffies;
4941 
4942 	/* Force memory writes to complete before letting h/w know there
4943 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
4944 	 * memory model archs, such as IA-64).
4945 	 *
4946 	 * We also need this memory barrier to make certain all of the
4947 	 * status bits have been updated before next_to_watch is written.
4948 	 */
4949 	wmb();
4950 
4951 	/* set next_to_watch value indicating a packet is present */
4952 	first->next_to_watch = tx_desc;
4953 
4954 	i++;
4955 	if (i == tx_ring->count)
4956 		i = 0;
4957 
4958 	tx_ring->next_to_use = i;
4959 
4960 	/* Make sure there is space in the ring for the next send. */
4961 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
4962 
4963 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
4964 		writel(i, tx_ring->tail);
4965 
4966 		/* we need this if more than one processor can write to our tail
4967 		 * at a time, it synchronizes IO on IA64/Altix systems
4968 		 */
4969 		mmiowb();
4970 	}
4971 	return;
4972 
4973 dma_error:
4974 	dev_err(tx_ring->dev, "TX DMA map failed\n");
4975 
4976 	/* clear dma mappings for failed tx_buffer_info map */
4977 	for (;;) {
4978 		tx_buffer = &tx_ring->tx_buffer_info[i];
4979 		igb_unmap_and_free_tx_resource(tx_ring, tx_buffer);
4980 		if (tx_buffer == first)
4981 			break;
4982 		if (i == 0)
4983 			i = tx_ring->count;
4984 		i--;
4985 	}
4986 
4987 	tx_ring->next_to_use = i;
4988 }
4989 
4990 netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
4991 				struct igb_ring *tx_ring)
4992 {
4993 	struct igb_tx_buffer *first;
4994 	int tso;
4995 	u32 tx_flags = 0;
4996 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
4997 	__be16 protocol = vlan_get_protocol(skb);
4998 	u8 hdr_len = 0;
4999 
5000 	/* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
5001 	 *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
5002 	 *       + 2 desc gap to keep tail from touching head,
5003 	 *       + 1 desc for context descriptor,
5004 	 * otherwise try next time
5005 	 */
5006 	if (NETDEV_FRAG_PAGE_MAX_SIZE > IGB_MAX_DATA_PER_TXD) {
5007 		unsigned short f;
5008 
5009 		for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
5010 			count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
5011 	} else {
5012 		count += skb_shinfo(skb)->nr_frags;
5013 	}
5014 
5015 	if (igb_maybe_stop_tx(tx_ring, count + 3)) {
5016 		/* this is a hard error */
5017 		return NETDEV_TX_BUSY;
5018 	}
5019 
5020 	/* record the location of the first descriptor for this packet */
5021 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
5022 	first->skb = skb;
5023 	first->bytecount = skb->len;
5024 	first->gso_segs = 1;
5025 
5026 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
5027 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
5028 
5029 		if (!test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
5030 					   &adapter->state)) {
5031 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5032 			tx_flags |= IGB_TX_FLAGS_TSTAMP;
5033 
5034 			adapter->ptp_tx_skb = skb_get(skb);
5035 			adapter->ptp_tx_start = jiffies;
5036 			if (adapter->hw.mac.type == e1000_82576)
5037 				schedule_work(&adapter->ptp_tx_work);
5038 		}
5039 	}
5040 
5041 	skb_tx_timestamp(skb);
5042 
5043 	if (vlan_tx_tag_present(skb)) {
5044 		tx_flags |= IGB_TX_FLAGS_VLAN;
5045 		tx_flags |= (vlan_tx_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
5046 	}
5047 
5048 	/* record initial flags and protocol */
5049 	first->tx_flags = tx_flags;
5050 	first->protocol = protocol;
5051 
5052 	tso = igb_tso(tx_ring, first, &hdr_len);
5053 	if (tso < 0)
5054 		goto out_drop;
5055 	else if (!tso)
5056 		igb_tx_csum(tx_ring, first);
5057 
5058 	igb_tx_map(tx_ring, first, hdr_len);
5059 
5060 	return NETDEV_TX_OK;
5061 
5062 out_drop:
5063 	igb_unmap_and_free_tx_resource(tx_ring, first);
5064 
5065 	return NETDEV_TX_OK;
5066 }
5067 
5068 static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
5069 						    struct sk_buff *skb)
5070 {
5071 	unsigned int r_idx = skb->queue_mapping;
5072 
5073 	if (r_idx >= adapter->num_tx_queues)
5074 		r_idx = r_idx % adapter->num_tx_queues;
5075 
5076 	return adapter->tx_ring[r_idx];
5077 }
5078 
5079 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
5080 				  struct net_device *netdev)
5081 {
5082 	struct igb_adapter *adapter = netdev_priv(netdev);
5083 
5084 	if (test_bit(__IGB_DOWN, &adapter->state)) {
5085 		dev_kfree_skb_any(skb);
5086 		return NETDEV_TX_OK;
5087 	}
5088 
5089 	if (skb->len <= 0) {
5090 		dev_kfree_skb_any(skb);
5091 		return NETDEV_TX_OK;
5092 	}
5093 
5094 	/* The minimum packet size with TCTL.PSP set is 17 so pad the skb
5095 	 * in order to meet this minimum size requirement.
5096 	 */
5097 	if (unlikely(skb->len < 17)) {
5098 		if (skb_pad(skb, 17 - skb->len))
5099 			return NETDEV_TX_OK;
5100 		skb->len = 17;
5101 		skb_set_tail_pointer(skb, 17);
5102 	}
5103 
5104 	return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
5105 }
5106 
5107 /**
5108  *  igb_tx_timeout - Respond to a Tx Hang
5109  *  @netdev: network interface device structure
5110  **/
5111 static void igb_tx_timeout(struct net_device *netdev)
5112 {
5113 	struct igb_adapter *adapter = netdev_priv(netdev);
5114 	struct e1000_hw *hw = &adapter->hw;
5115 
5116 	/* Do the reset outside of interrupt context */
5117 	adapter->tx_timeout_count++;
5118 
5119 	if (hw->mac.type >= e1000_82580)
5120 		hw->dev_spec._82575.global_device_reset = true;
5121 
5122 	schedule_work(&adapter->reset_task);
5123 	wr32(E1000_EICS,
5124 	     (adapter->eims_enable_mask & ~adapter->eims_other));
5125 }
5126 
5127 static void igb_reset_task(struct work_struct *work)
5128 {
5129 	struct igb_adapter *adapter;
5130 	adapter = container_of(work, struct igb_adapter, reset_task);
5131 
5132 	igb_dump(adapter);
5133 	netdev_err(adapter->netdev, "Reset adapter\n");
5134 	igb_reinit_locked(adapter);
5135 }
5136 
5137 /**
5138  *  igb_get_stats64 - Get System Network Statistics
5139  *  @netdev: network interface device structure
5140  *  @stats: rtnl_link_stats64 pointer
5141  **/
5142 static struct rtnl_link_stats64 *igb_get_stats64(struct net_device *netdev,
5143 						struct rtnl_link_stats64 *stats)
5144 {
5145 	struct igb_adapter *adapter = netdev_priv(netdev);
5146 
5147 	spin_lock(&adapter->stats64_lock);
5148 	igb_update_stats(adapter, &adapter->stats64);
5149 	memcpy(stats, &adapter->stats64, sizeof(*stats));
5150 	spin_unlock(&adapter->stats64_lock);
5151 
5152 	return stats;
5153 }
5154 
5155 /**
5156  *  igb_change_mtu - Change the Maximum Transfer Unit
5157  *  @netdev: network interface device structure
5158  *  @new_mtu: new value for maximum frame size
5159  *
5160  *  Returns 0 on success, negative on failure
5161  **/
5162 static int igb_change_mtu(struct net_device *netdev, int new_mtu)
5163 {
5164 	struct igb_adapter *adapter = netdev_priv(netdev);
5165 	struct pci_dev *pdev = adapter->pdev;
5166 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
5167 
5168 	if ((new_mtu < 68) || (max_frame > MAX_JUMBO_FRAME_SIZE)) {
5169 		dev_err(&pdev->dev, "Invalid MTU setting\n");
5170 		return -EINVAL;
5171 	}
5172 
5173 #define MAX_STD_JUMBO_FRAME_SIZE 9238
5174 	if (max_frame > MAX_STD_JUMBO_FRAME_SIZE) {
5175 		dev_err(&pdev->dev, "MTU > 9216 not supported.\n");
5176 		return -EINVAL;
5177 	}
5178 
5179 	/* adjust max frame to be at least the size of a standard frame */
5180 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
5181 		max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
5182 
5183 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
5184 		usleep_range(1000, 2000);
5185 
5186 	/* igb_down has a dependency on max_frame_size */
5187 	adapter->max_frame_size = max_frame;
5188 
5189 	if (netif_running(netdev))
5190 		igb_down(adapter);
5191 
5192 	dev_info(&pdev->dev, "changing MTU from %d to %d\n",
5193 		 netdev->mtu, new_mtu);
5194 	netdev->mtu = new_mtu;
5195 
5196 	if (netif_running(netdev))
5197 		igb_up(adapter);
5198 	else
5199 		igb_reset(adapter);
5200 
5201 	clear_bit(__IGB_RESETTING, &adapter->state);
5202 
5203 	return 0;
5204 }
5205 
5206 /**
5207  *  igb_update_stats - Update the board statistics counters
5208  *  @adapter: board private structure
5209  **/
5210 void igb_update_stats(struct igb_adapter *adapter,
5211 		      struct rtnl_link_stats64 *net_stats)
5212 {
5213 	struct e1000_hw *hw = &adapter->hw;
5214 	struct pci_dev *pdev = adapter->pdev;
5215 	u32 reg, mpc;
5216 	int i;
5217 	u64 bytes, packets;
5218 	unsigned int start;
5219 	u64 _bytes, _packets;
5220 
5221 	/* Prevent stats update while adapter is being reset, or if the pci
5222 	 * connection is down.
5223 	 */
5224 	if (adapter->link_speed == 0)
5225 		return;
5226 	if (pci_channel_offline(pdev))
5227 		return;
5228 
5229 	bytes = 0;
5230 	packets = 0;
5231 
5232 	rcu_read_lock();
5233 	for (i = 0; i < adapter->num_rx_queues; i++) {
5234 		struct igb_ring *ring = adapter->rx_ring[i];
5235 		u32 rqdpc = rd32(E1000_RQDPC(i));
5236 		if (hw->mac.type >= e1000_i210)
5237 			wr32(E1000_RQDPC(i), 0);
5238 
5239 		if (rqdpc) {
5240 			ring->rx_stats.drops += rqdpc;
5241 			net_stats->rx_fifo_errors += rqdpc;
5242 		}
5243 
5244 		do {
5245 			start = u64_stats_fetch_begin_irq(&ring->rx_syncp);
5246 			_bytes = ring->rx_stats.bytes;
5247 			_packets = ring->rx_stats.packets;
5248 		} while (u64_stats_fetch_retry_irq(&ring->rx_syncp, start));
5249 		bytes += _bytes;
5250 		packets += _packets;
5251 	}
5252 
5253 	net_stats->rx_bytes = bytes;
5254 	net_stats->rx_packets = packets;
5255 
5256 	bytes = 0;
5257 	packets = 0;
5258 	for (i = 0; i < adapter->num_tx_queues; i++) {
5259 		struct igb_ring *ring = adapter->tx_ring[i];
5260 		do {
5261 			start = u64_stats_fetch_begin_irq(&ring->tx_syncp);
5262 			_bytes = ring->tx_stats.bytes;
5263 			_packets = ring->tx_stats.packets;
5264 		} while (u64_stats_fetch_retry_irq(&ring->tx_syncp, start));
5265 		bytes += _bytes;
5266 		packets += _packets;
5267 	}
5268 	net_stats->tx_bytes = bytes;
5269 	net_stats->tx_packets = packets;
5270 	rcu_read_unlock();
5271 
5272 	/* read stats registers */
5273 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
5274 	adapter->stats.gprc += rd32(E1000_GPRC);
5275 	adapter->stats.gorc += rd32(E1000_GORCL);
5276 	rd32(E1000_GORCH); /* clear GORCL */
5277 	adapter->stats.bprc += rd32(E1000_BPRC);
5278 	adapter->stats.mprc += rd32(E1000_MPRC);
5279 	adapter->stats.roc += rd32(E1000_ROC);
5280 
5281 	adapter->stats.prc64 += rd32(E1000_PRC64);
5282 	adapter->stats.prc127 += rd32(E1000_PRC127);
5283 	adapter->stats.prc255 += rd32(E1000_PRC255);
5284 	adapter->stats.prc511 += rd32(E1000_PRC511);
5285 	adapter->stats.prc1023 += rd32(E1000_PRC1023);
5286 	adapter->stats.prc1522 += rd32(E1000_PRC1522);
5287 	adapter->stats.symerrs += rd32(E1000_SYMERRS);
5288 	adapter->stats.sec += rd32(E1000_SEC);
5289 
5290 	mpc = rd32(E1000_MPC);
5291 	adapter->stats.mpc += mpc;
5292 	net_stats->rx_fifo_errors += mpc;
5293 	adapter->stats.scc += rd32(E1000_SCC);
5294 	adapter->stats.ecol += rd32(E1000_ECOL);
5295 	adapter->stats.mcc += rd32(E1000_MCC);
5296 	adapter->stats.latecol += rd32(E1000_LATECOL);
5297 	adapter->stats.dc += rd32(E1000_DC);
5298 	adapter->stats.rlec += rd32(E1000_RLEC);
5299 	adapter->stats.xonrxc += rd32(E1000_XONRXC);
5300 	adapter->stats.xontxc += rd32(E1000_XONTXC);
5301 	adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
5302 	adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
5303 	adapter->stats.fcruc += rd32(E1000_FCRUC);
5304 	adapter->stats.gptc += rd32(E1000_GPTC);
5305 	adapter->stats.gotc += rd32(E1000_GOTCL);
5306 	rd32(E1000_GOTCH); /* clear GOTCL */
5307 	adapter->stats.rnbc += rd32(E1000_RNBC);
5308 	adapter->stats.ruc += rd32(E1000_RUC);
5309 	adapter->stats.rfc += rd32(E1000_RFC);
5310 	adapter->stats.rjc += rd32(E1000_RJC);
5311 	adapter->stats.tor += rd32(E1000_TORH);
5312 	adapter->stats.tot += rd32(E1000_TOTH);
5313 	adapter->stats.tpr += rd32(E1000_TPR);
5314 
5315 	adapter->stats.ptc64 += rd32(E1000_PTC64);
5316 	adapter->stats.ptc127 += rd32(E1000_PTC127);
5317 	adapter->stats.ptc255 += rd32(E1000_PTC255);
5318 	adapter->stats.ptc511 += rd32(E1000_PTC511);
5319 	adapter->stats.ptc1023 += rd32(E1000_PTC1023);
5320 	adapter->stats.ptc1522 += rd32(E1000_PTC1522);
5321 
5322 	adapter->stats.mptc += rd32(E1000_MPTC);
5323 	adapter->stats.bptc += rd32(E1000_BPTC);
5324 
5325 	adapter->stats.tpt += rd32(E1000_TPT);
5326 	adapter->stats.colc += rd32(E1000_COLC);
5327 
5328 	adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
5329 	/* read internal phy specific stats */
5330 	reg = rd32(E1000_CTRL_EXT);
5331 	if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
5332 		adapter->stats.rxerrc += rd32(E1000_RXERRC);
5333 
5334 		/* this stat has invalid values on i210/i211 */
5335 		if ((hw->mac.type != e1000_i210) &&
5336 		    (hw->mac.type != e1000_i211))
5337 			adapter->stats.tncrs += rd32(E1000_TNCRS);
5338 	}
5339 
5340 	adapter->stats.tsctc += rd32(E1000_TSCTC);
5341 	adapter->stats.tsctfc += rd32(E1000_TSCTFC);
5342 
5343 	adapter->stats.iac += rd32(E1000_IAC);
5344 	adapter->stats.icrxoc += rd32(E1000_ICRXOC);
5345 	adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
5346 	adapter->stats.icrxatc += rd32(E1000_ICRXATC);
5347 	adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
5348 	adapter->stats.ictxatc += rd32(E1000_ICTXATC);
5349 	adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
5350 	adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
5351 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
5352 
5353 	/* Fill out the OS statistics structure */
5354 	net_stats->multicast = adapter->stats.mprc;
5355 	net_stats->collisions = adapter->stats.colc;
5356 
5357 	/* Rx Errors */
5358 
5359 	/* RLEC on some newer hardware can be incorrect so build
5360 	 * our own version based on RUC and ROC
5361 	 */
5362 	net_stats->rx_errors = adapter->stats.rxerrc +
5363 		adapter->stats.crcerrs + adapter->stats.algnerrc +
5364 		adapter->stats.ruc + adapter->stats.roc +
5365 		adapter->stats.cexterr;
5366 	net_stats->rx_length_errors = adapter->stats.ruc +
5367 				      adapter->stats.roc;
5368 	net_stats->rx_crc_errors = adapter->stats.crcerrs;
5369 	net_stats->rx_frame_errors = adapter->stats.algnerrc;
5370 	net_stats->rx_missed_errors = adapter->stats.mpc;
5371 
5372 	/* Tx Errors */
5373 	net_stats->tx_errors = adapter->stats.ecol +
5374 			       adapter->stats.latecol;
5375 	net_stats->tx_aborted_errors = adapter->stats.ecol;
5376 	net_stats->tx_window_errors = adapter->stats.latecol;
5377 	net_stats->tx_carrier_errors = adapter->stats.tncrs;
5378 
5379 	/* Tx Dropped needs to be maintained elsewhere */
5380 
5381 	/* Management Stats */
5382 	adapter->stats.mgptc += rd32(E1000_MGTPTC);
5383 	adapter->stats.mgprc += rd32(E1000_MGTPRC);
5384 	adapter->stats.mgpdc += rd32(E1000_MGTPDC);
5385 
5386 	/* OS2BMC Stats */
5387 	reg = rd32(E1000_MANC);
5388 	if (reg & E1000_MANC_EN_BMC2OS) {
5389 		adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
5390 		adapter->stats.o2bspc += rd32(E1000_O2BSPC);
5391 		adapter->stats.b2ospc += rd32(E1000_B2OSPC);
5392 		adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
5393 	}
5394 }
5395 
5396 static irqreturn_t igb_msix_other(int irq, void *data)
5397 {
5398 	struct igb_adapter *adapter = data;
5399 	struct e1000_hw *hw = &adapter->hw;
5400 	u32 icr = rd32(E1000_ICR);
5401 	/* reading ICR causes bit 31 of EICR to be cleared */
5402 
5403 	if (icr & E1000_ICR_DRSTA)
5404 		schedule_work(&adapter->reset_task);
5405 
5406 	if (icr & E1000_ICR_DOUTSYNC) {
5407 		/* HW is reporting DMA is out of sync */
5408 		adapter->stats.doosync++;
5409 		/* The DMA Out of Sync is also indication of a spoof event
5410 		 * in IOV mode. Check the Wrong VM Behavior register to
5411 		 * see if it is really a spoof event.
5412 		 */
5413 		igb_check_wvbr(adapter);
5414 	}
5415 
5416 	/* Check for a mailbox event */
5417 	if (icr & E1000_ICR_VMMB)
5418 		igb_msg_task(adapter);
5419 
5420 	if (icr & E1000_ICR_LSC) {
5421 		hw->mac.get_link_status = 1;
5422 		/* guard against interrupt when we're going down */
5423 		if (!test_bit(__IGB_DOWN, &adapter->state))
5424 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
5425 	}
5426 
5427 	if (icr & E1000_ICR_TS) {
5428 		u32 tsicr = rd32(E1000_TSICR);
5429 
5430 		if (tsicr & E1000_TSICR_TXTS) {
5431 			/* acknowledge the interrupt */
5432 			wr32(E1000_TSICR, E1000_TSICR_TXTS);
5433 			/* retrieve hardware timestamp */
5434 			schedule_work(&adapter->ptp_tx_work);
5435 		}
5436 	}
5437 
5438 	wr32(E1000_EIMS, adapter->eims_other);
5439 
5440 	return IRQ_HANDLED;
5441 }
5442 
5443 static void igb_write_itr(struct igb_q_vector *q_vector)
5444 {
5445 	struct igb_adapter *adapter = q_vector->adapter;
5446 	u32 itr_val = q_vector->itr_val & 0x7FFC;
5447 
5448 	if (!q_vector->set_itr)
5449 		return;
5450 
5451 	if (!itr_val)
5452 		itr_val = 0x4;
5453 
5454 	if (adapter->hw.mac.type == e1000_82575)
5455 		itr_val |= itr_val << 16;
5456 	else
5457 		itr_val |= E1000_EITR_CNT_IGNR;
5458 
5459 	writel(itr_val, q_vector->itr_register);
5460 	q_vector->set_itr = 0;
5461 }
5462 
5463 static irqreturn_t igb_msix_ring(int irq, void *data)
5464 {
5465 	struct igb_q_vector *q_vector = data;
5466 
5467 	/* Write the ITR value calculated from the previous interrupt. */
5468 	igb_write_itr(q_vector);
5469 
5470 	napi_schedule(&q_vector->napi);
5471 
5472 	return IRQ_HANDLED;
5473 }
5474 
5475 #ifdef CONFIG_IGB_DCA
5476 static void igb_update_tx_dca(struct igb_adapter *adapter,
5477 			      struct igb_ring *tx_ring,
5478 			      int cpu)
5479 {
5480 	struct e1000_hw *hw = &adapter->hw;
5481 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
5482 
5483 	if (hw->mac.type != e1000_82575)
5484 		txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
5485 
5486 	/* We can enable relaxed ordering for reads, but not writes when
5487 	 * DCA is enabled.  This is due to a known issue in some chipsets
5488 	 * which will cause the DCA tag to be cleared.
5489 	 */
5490 	txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
5491 		  E1000_DCA_TXCTRL_DATA_RRO_EN |
5492 		  E1000_DCA_TXCTRL_DESC_DCA_EN;
5493 
5494 	wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
5495 }
5496 
5497 static void igb_update_rx_dca(struct igb_adapter *adapter,
5498 			      struct igb_ring *rx_ring,
5499 			      int cpu)
5500 {
5501 	struct e1000_hw *hw = &adapter->hw;
5502 	u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
5503 
5504 	if (hw->mac.type != e1000_82575)
5505 		rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
5506 
5507 	/* We can enable relaxed ordering for reads, but not writes when
5508 	 * DCA is enabled.  This is due to a known issue in some chipsets
5509 	 * which will cause the DCA tag to be cleared.
5510 	 */
5511 	rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
5512 		  E1000_DCA_RXCTRL_DESC_DCA_EN;
5513 
5514 	wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
5515 }
5516 
5517 static void igb_update_dca(struct igb_q_vector *q_vector)
5518 {
5519 	struct igb_adapter *adapter = q_vector->adapter;
5520 	int cpu = get_cpu();
5521 
5522 	if (q_vector->cpu == cpu)
5523 		goto out_no_update;
5524 
5525 	if (q_vector->tx.ring)
5526 		igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
5527 
5528 	if (q_vector->rx.ring)
5529 		igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
5530 
5531 	q_vector->cpu = cpu;
5532 out_no_update:
5533 	put_cpu();
5534 }
5535 
5536 static void igb_setup_dca(struct igb_adapter *adapter)
5537 {
5538 	struct e1000_hw *hw = &adapter->hw;
5539 	int i;
5540 
5541 	if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
5542 		return;
5543 
5544 	/* Always use CB2 mode, difference is masked in the CB driver. */
5545 	wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
5546 
5547 	for (i = 0; i < adapter->num_q_vectors; i++) {
5548 		adapter->q_vector[i]->cpu = -1;
5549 		igb_update_dca(adapter->q_vector[i]);
5550 	}
5551 }
5552 
5553 static int __igb_notify_dca(struct device *dev, void *data)
5554 {
5555 	struct net_device *netdev = dev_get_drvdata(dev);
5556 	struct igb_adapter *adapter = netdev_priv(netdev);
5557 	struct pci_dev *pdev = adapter->pdev;
5558 	struct e1000_hw *hw = &adapter->hw;
5559 	unsigned long event = *(unsigned long *)data;
5560 
5561 	switch (event) {
5562 	case DCA_PROVIDER_ADD:
5563 		/* if already enabled, don't do it again */
5564 		if (adapter->flags & IGB_FLAG_DCA_ENABLED)
5565 			break;
5566 		if (dca_add_requester(dev) == 0) {
5567 			adapter->flags |= IGB_FLAG_DCA_ENABLED;
5568 			dev_info(&pdev->dev, "DCA enabled\n");
5569 			igb_setup_dca(adapter);
5570 			break;
5571 		}
5572 		/* Fall Through since DCA is disabled. */
5573 	case DCA_PROVIDER_REMOVE:
5574 		if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
5575 			/* without this a class_device is left
5576 			 * hanging around in the sysfs model
5577 			 */
5578 			dca_remove_requester(dev);
5579 			dev_info(&pdev->dev, "DCA disabled\n");
5580 			adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
5581 			wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
5582 		}
5583 		break;
5584 	}
5585 
5586 	return 0;
5587 }
5588 
5589 static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
5590 			  void *p)
5591 {
5592 	int ret_val;
5593 
5594 	ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
5595 					 __igb_notify_dca);
5596 
5597 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
5598 }
5599 #endif /* CONFIG_IGB_DCA */
5600 
5601 #ifdef CONFIG_PCI_IOV
5602 static int igb_vf_configure(struct igb_adapter *adapter, int vf)
5603 {
5604 	unsigned char mac_addr[ETH_ALEN];
5605 
5606 	eth_zero_addr(mac_addr);
5607 	igb_set_vf_mac(adapter, vf, mac_addr);
5608 
5609 	/* By default spoof check is enabled for all VFs */
5610 	adapter->vf_data[vf].spoofchk_enabled = true;
5611 
5612 	return 0;
5613 }
5614 
5615 #endif
5616 static void igb_ping_all_vfs(struct igb_adapter *adapter)
5617 {
5618 	struct e1000_hw *hw = &adapter->hw;
5619 	u32 ping;
5620 	int i;
5621 
5622 	for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
5623 		ping = E1000_PF_CONTROL_MSG;
5624 		if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
5625 			ping |= E1000_VT_MSGTYPE_CTS;
5626 		igb_write_mbx(hw, &ping, 1, i);
5627 	}
5628 }
5629 
5630 static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
5631 {
5632 	struct e1000_hw *hw = &adapter->hw;
5633 	u32 vmolr = rd32(E1000_VMOLR(vf));
5634 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
5635 
5636 	vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
5637 			    IGB_VF_FLAG_MULTI_PROMISC);
5638 	vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
5639 
5640 	if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
5641 		vmolr |= E1000_VMOLR_MPME;
5642 		vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
5643 		*msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
5644 	} else {
5645 		/* if we have hashes and we are clearing a multicast promisc
5646 		 * flag we need to write the hashes to the MTA as this step
5647 		 * was previously skipped
5648 		 */
5649 		if (vf_data->num_vf_mc_hashes > 30) {
5650 			vmolr |= E1000_VMOLR_MPME;
5651 		} else if (vf_data->num_vf_mc_hashes) {
5652 			int j;
5653 
5654 			vmolr |= E1000_VMOLR_ROMPE;
5655 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
5656 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
5657 		}
5658 	}
5659 
5660 	wr32(E1000_VMOLR(vf), vmolr);
5661 
5662 	/* there are flags left unprocessed, likely not supported */
5663 	if (*msgbuf & E1000_VT_MSGINFO_MASK)
5664 		return -EINVAL;
5665 
5666 	return 0;
5667 }
5668 
5669 static int igb_set_vf_multicasts(struct igb_adapter *adapter,
5670 				  u32 *msgbuf, u32 vf)
5671 {
5672 	int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
5673 	u16 *hash_list = (u16 *)&msgbuf[1];
5674 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
5675 	int i;
5676 
5677 	/* salt away the number of multicast addresses assigned
5678 	 * to this VF for later use to restore when the PF multi cast
5679 	 * list changes
5680 	 */
5681 	vf_data->num_vf_mc_hashes = n;
5682 
5683 	/* only up to 30 hash values supported */
5684 	if (n > 30)
5685 		n = 30;
5686 
5687 	/* store the hashes for later use */
5688 	for (i = 0; i < n; i++)
5689 		vf_data->vf_mc_hashes[i] = hash_list[i];
5690 
5691 	/* Flush and reset the mta with the new values */
5692 	igb_set_rx_mode(adapter->netdev);
5693 
5694 	return 0;
5695 }
5696 
5697 static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
5698 {
5699 	struct e1000_hw *hw = &adapter->hw;
5700 	struct vf_data_storage *vf_data;
5701 	int i, j;
5702 
5703 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
5704 		u32 vmolr = rd32(E1000_VMOLR(i));
5705 
5706 		vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
5707 
5708 		vf_data = &adapter->vf_data[i];
5709 
5710 		if ((vf_data->num_vf_mc_hashes > 30) ||
5711 		    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
5712 			vmolr |= E1000_VMOLR_MPME;
5713 		} else if (vf_data->num_vf_mc_hashes) {
5714 			vmolr |= E1000_VMOLR_ROMPE;
5715 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
5716 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
5717 		}
5718 		wr32(E1000_VMOLR(i), vmolr);
5719 	}
5720 }
5721 
5722 static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
5723 {
5724 	struct e1000_hw *hw = &adapter->hw;
5725 	u32 pool_mask, reg, vid;
5726 	int i;
5727 
5728 	pool_mask = 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
5729 
5730 	/* Find the vlan filter for this id */
5731 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5732 		reg = rd32(E1000_VLVF(i));
5733 
5734 		/* remove the vf from the pool */
5735 		reg &= ~pool_mask;
5736 
5737 		/* if pool is empty then remove entry from vfta */
5738 		if (!(reg & E1000_VLVF_POOLSEL_MASK) &&
5739 		    (reg & E1000_VLVF_VLANID_ENABLE)) {
5740 			reg = 0;
5741 			vid = reg & E1000_VLVF_VLANID_MASK;
5742 			igb_vfta_set(hw, vid, false);
5743 		}
5744 
5745 		wr32(E1000_VLVF(i), reg);
5746 	}
5747 
5748 	adapter->vf_data[vf].vlans_enabled = 0;
5749 }
5750 
5751 static s32 igb_vlvf_set(struct igb_adapter *adapter, u32 vid, bool add, u32 vf)
5752 {
5753 	struct e1000_hw *hw = &adapter->hw;
5754 	u32 reg, i;
5755 
5756 	/* The vlvf table only exists on 82576 hardware and newer */
5757 	if (hw->mac.type < e1000_82576)
5758 		return -1;
5759 
5760 	/* we only need to do this if VMDq is enabled */
5761 	if (!adapter->vfs_allocated_count)
5762 		return -1;
5763 
5764 	/* Find the vlan filter for this id */
5765 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5766 		reg = rd32(E1000_VLVF(i));
5767 		if ((reg & E1000_VLVF_VLANID_ENABLE) &&
5768 		    vid == (reg & E1000_VLVF_VLANID_MASK))
5769 			break;
5770 	}
5771 
5772 	if (add) {
5773 		if (i == E1000_VLVF_ARRAY_SIZE) {
5774 			/* Did not find a matching VLAN ID entry that was
5775 			 * enabled.  Search for a free filter entry, i.e.
5776 			 * one without the enable bit set
5777 			 */
5778 			for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5779 				reg = rd32(E1000_VLVF(i));
5780 				if (!(reg & E1000_VLVF_VLANID_ENABLE))
5781 					break;
5782 			}
5783 		}
5784 		if (i < E1000_VLVF_ARRAY_SIZE) {
5785 			/* Found an enabled/available entry */
5786 			reg |= 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
5787 
5788 			/* if !enabled we need to set this up in vfta */
5789 			if (!(reg & E1000_VLVF_VLANID_ENABLE)) {
5790 				/* add VID to filter table */
5791 				igb_vfta_set(hw, vid, true);
5792 				reg |= E1000_VLVF_VLANID_ENABLE;
5793 			}
5794 			reg &= ~E1000_VLVF_VLANID_MASK;
5795 			reg |= vid;
5796 			wr32(E1000_VLVF(i), reg);
5797 
5798 			/* do not modify RLPML for PF devices */
5799 			if (vf >= adapter->vfs_allocated_count)
5800 				return 0;
5801 
5802 			if (!adapter->vf_data[vf].vlans_enabled) {
5803 				u32 size;
5804 
5805 				reg = rd32(E1000_VMOLR(vf));
5806 				size = reg & E1000_VMOLR_RLPML_MASK;
5807 				size += 4;
5808 				reg &= ~E1000_VMOLR_RLPML_MASK;
5809 				reg |= size;
5810 				wr32(E1000_VMOLR(vf), reg);
5811 			}
5812 
5813 			adapter->vf_data[vf].vlans_enabled++;
5814 		}
5815 	} else {
5816 		if (i < E1000_VLVF_ARRAY_SIZE) {
5817 			/* remove vf from the pool */
5818 			reg &= ~(1 << (E1000_VLVF_POOLSEL_SHIFT + vf));
5819 			/* if pool is empty then remove entry from vfta */
5820 			if (!(reg & E1000_VLVF_POOLSEL_MASK)) {
5821 				reg = 0;
5822 				igb_vfta_set(hw, vid, false);
5823 			}
5824 			wr32(E1000_VLVF(i), reg);
5825 
5826 			/* do not modify RLPML for PF devices */
5827 			if (vf >= adapter->vfs_allocated_count)
5828 				return 0;
5829 
5830 			adapter->vf_data[vf].vlans_enabled--;
5831 			if (!adapter->vf_data[vf].vlans_enabled) {
5832 				u32 size;
5833 
5834 				reg = rd32(E1000_VMOLR(vf));
5835 				size = reg & E1000_VMOLR_RLPML_MASK;
5836 				size -= 4;
5837 				reg &= ~E1000_VMOLR_RLPML_MASK;
5838 				reg |= size;
5839 				wr32(E1000_VMOLR(vf), reg);
5840 			}
5841 		}
5842 	}
5843 	return 0;
5844 }
5845 
5846 static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
5847 {
5848 	struct e1000_hw *hw = &adapter->hw;
5849 
5850 	if (vid)
5851 		wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
5852 	else
5853 		wr32(E1000_VMVIR(vf), 0);
5854 }
5855 
5856 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
5857 			       int vf, u16 vlan, u8 qos)
5858 {
5859 	int err = 0;
5860 	struct igb_adapter *adapter = netdev_priv(netdev);
5861 
5862 	if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
5863 		return -EINVAL;
5864 	if (vlan || qos) {
5865 		err = igb_vlvf_set(adapter, vlan, !!vlan, vf);
5866 		if (err)
5867 			goto out;
5868 		igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
5869 		igb_set_vmolr(adapter, vf, !vlan);
5870 		adapter->vf_data[vf].pf_vlan = vlan;
5871 		adapter->vf_data[vf].pf_qos = qos;
5872 		dev_info(&adapter->pdev->dev,
5873 			 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
5874 		if (test_bit(__IGB_DOWN, &adapter->state)) {
5875 			dev_warn(&adapter->pdev->dev,
5876 				 "The VF VLAN has been set, but the PF device is not up.\n");
5877 			dev_warn(&adapter->pdev->dev,
5878 				 "Bring the PF device up before attempting to use the VF device.\n");
5879 		}
5880 	} else {
5881 		igb_vlvf_set(adapter, adapter->vf_data[vf].pf_vlan,
5882 			     false, vf);
5883 		igb_set_vmvir(adapter, vlan, vf);
5884 		igb_set_vmolr(adapter, vf, true);
5885 		adapter->vf_data[vf].pf_vlan = 0;
5886 		adapter->vf_data[vf].pf_qos = 0;
5887 	}
5888 out:
5889 	return err;
5890 }
5891 
5892 static int igb_find_vlvf_entry(struct igb_adapter *adapter, int vid)
5893 {
5894 	struct e1000_hw *hw = &adapter->hw;
5895 	int i;
5896 	u32 reg;
5897 
5898 	/* Find the vlan filter for this id */
5899 	for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
5900 		reg = rd32(E1000_VLVF(i));
5901 		if ((reg & E1000_VLVF_VLANID_ENABLE) &&
5902 		    vid == (reg & E1000_VLVF_VLANID_MASK))
5903 			break;
5904 	}
5905 
5906 	if (i >= E1000_VLVF_ARRAY_SIZE)
5907 		i = -1;
5908 
5909 	return i;
5910 }
5911 
5912 static int igb_set_vf_vlan(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
5913 {
5914 	struct e1000_hw *hw = &adapter->hw;
5915 	int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
5916 	int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
5917 	int err = 0;
5918 
5919 	/* If in promiscuous mode we need to make sure the PF also has
5920 	 * the VLAN filter set.
5921 	 */
5922 	if (add && (adapter->netdev->flags & IFF_PROMISC))
5923 		err = igb_vlvf_set(adapter, vid, add,
5924 				   adapter->vfs_allocated_count);
5925 	if (err)
5926 		goto out;
5927 
5928 	err = igb_vlvf_set(adapter, vid, add, vf);
5929 
5930 	if (err)
5931 		goto out;
5932 
5933 	/* Go through all the checks to see if the VLAN filter should
5934 	 * be wiped completely.
5935 	 */
5936 	if (!add && (adapter->netdev->flags & IFF_PROMISC)) {
5937 		u32 vlvf, bits;
5938 		int regndx = igb_find_vlvf_entry(adapter, vid);
5939 
5940 		if (regndx < 0)
5941 			goto out;
5942 		/* See if any other pools are set for this VLAN filter
5943 		 * entry other than the PF.
5944 		 */
5945 		vlvf = bits = rd32(E1000_VLVF(regndx));
5946 		bits &= 1 << (E1000_VLVF_POOLSEL_SHIFT +
5947 			      adapter->vfs_allocated_count);
5948 		/* If the filter was removed then ensure PF pool bit
5949 		 * is cleared if the PF only added itself to the pool
5950 		 * because the PF is in promiscuous mode.
5951 		 */
5952 		if ((vlvf & VLAN_VID_MASK) == vid &&
5953 		    !test_bit(vid, adapter->active_vlans) &&
5954 		    !bits)
5955 			igb_vlvf_set(adapter, vid, add,
5956 				     adapter->vfs_allocated_count);
5957 	}
5958 
5959 out:
5960 	return err;
5961 }
5962 
5963 static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
5964 {
5965 	/* clear flags - except flag that indicates PF has set the MAC */
5966 	adapter->vf_data[vf].flags &= IGB_VF_FLAG_PF_SET_MAC;
5967 	adapter->vf_data[vf].last_nack = jiffies;
5968 
5969 	/* reset offloads to defaults */
5970 	igb_set_vmolr(adapter, vf, true);
5971 
5972 	/* reset vlans for device */
5973 	igb_clear_vf_vfta(adapter, vf);
5974 	if (adapter->vf_data[vf].pf_vlan)
5975 		igb_ndo_set_vf_vlan(adapter->netdev, vf,
5976 				    adapter->vf_data[vf].pf_vlan,
5977 				    adapter->vf_data[vf].pf_qos);
5978 	else
5979 		igb_clear_vf_vfta(adapter, vf);
5980 
5981 	/* reset multicast table array for vf */
5982 	adapter->vf_data[vf].num_vf_mc_hashes = 0;
5983 
5984 	/* Flush and reset the mta with the new values */
5985 	igb_set_rx_mode(adapter->netdev);
5986 }
5987 
5988 static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
5989 {
5990 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
5991 
5992 	/* clear mac address as we were hotplug removed/added */
5993 	if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
5994 		eth_zero_addr(vf_mac);
5995 
5996 	/* process remaining reset events */
5997 	igb_vf_reset(adapter, vf);
5998 }
5999 
6000 static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
6001 {
6002 	struct e1000_hw *hw = &adapter->hw;
6003 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6004 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
6005 	u32 reg, msgbuf[3];
6006 	u8 *addr = (u8 *)(&msgbuf[1]);
6007 
6008 	/* process all the same items cleared in a function level reset */
6009 	igb_vf_reset(adapter, vf);
6010 
6011 	/* set vf mac address */
6012 	igb_rar_set_qsel(adapter, vf_mac, rar_entry, vf);
6013 
6014 	/* enable transmit and receive for vf */
6015 	reg = rd32(E1000_VFTE);
6016 	wr32(E1000_VFTE, reg | (1 << vf));
6017 	reg = rd32(E1000_VFRE);
6018 	wr32(E1000_VFRE, reg | (1 << vf));
6019 
6020 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
6021 
6022 	/* reply to reset with ack and vf mac address */
6023 	msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
6024 	memcpy(addr, vf_mac, ETH_ALEN);
6025 	igb_write_mbx(hw, msgbuf, 3, vf);
6026 }
6027 
6028 static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
6029 {
6030 	/* The VF MAC Address is stored in a packed array of bytes
6031 	 * starting at the second 32 bit word of the msg array
6032 	 */
6033 	unsigned char *addr = (char *)&msg[1];
6034 	int err = -1;
6035 
6036 	if (is_valid_ether_addr(addr))
6037 		err = igb_set_vf_mac(adapter, vf, addr);
6038 
6039 	return err;
6040 }
6041 
6042 static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
6043 {
6044 	struct e1000_hw *hw = &adapter->hw;
6045 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6046 	u32 msg = E1000_VT_MSGTYPE_NACK;
6047 
6048 	/* if device isn't clear to send it shouldn't be reading either */
6049 	if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
6050 	    time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
6051 		igb_write_mbx(hw, &msg, 1, vf);
6052 		vf_data->last_nack = jiffies;
6053 	}
6054 }
6055 
6056 static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
6057 {
6058 	struct pci_dev *pdev = adapter->pdev;
6059 	u32 msgbuf[E1000_VFMAILBOX_SIZE];
6060 	struct e1000_hw *hw = &adapter->hw;
6061 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6062 	s32 retval;
6063 
6064 	retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf);
6065 
6066 	if (retval) {
6067 		/* if receive failed revoke VF CTS stats and restart init */
6068 		dev_err(&pdev->dev, "Error receiving message from VF\n");
6069 		vf_data->flags &= ~IGB_VF_FLAG_CTS;
6070 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
6071 			return;
6072 		goto out;
6073 	}
6074 
6075 	/* this is a message we already processed, do nothing */
6076 	if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
6077 		return;
6078 
6079 	/* until the vf completes a reset it should not be
6080 	 * allowed to start any configuration.
6081 	 */
6082 	if (msgbuf[0] == E1000_VF_RESET) {
6083 		igb_vf_reset_msg(adapter, vf);
6084 		return;
6085 	}
6086 
6087 	if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
6088 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
6089 			return;
6090 		retval = -1;
6091 		goto out;
6092 	}
6093 
6094 	switch ((msgbuf[0] & 0xFFFF)) {
6095 	case E1000_VF_SET_MAC_ADDR:
6096 		retval = -EINVAL;
6097 		if (!(vf_data->flags & IGB_VF_FLAG_PF_SET_MAC))
6098 			retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
6099 		else
6100 			dev_warn(&pdev->dev,
6101 				 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
6102 				 vf);
6103 		break;
6104 	case E1000_VF_SET_PROMISC:
6105 		retval = igb_set_vf_promisc(adapter, msgbuf, vf);
6106 		break;
6107 	case E1000_VF_SET_MULTICAST:
6108 		retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
6109 		break;
6110 	case E1000_VF_SET_LPE:
6111 		retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
6112 		break;
6113 	case E1000_VF_SET_VLAN:
6114 		retval = -1;
6115 		if (vf_data->pf_vlan)
6116 			dev_warn(&pdev->dev,
6117 				 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
6118 				 vf);
6119 		else
6120 			retval = igb_set_vf_vlan(adapter, msgbuf, vf);
6121 		break;
6122 	default:
6123 		dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
6124 		retval = -1;
6125 		break;
6126 	}
6127 
6128 	msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
6129 out:
6130 	/* notify the VF of the results of what it sent us */
6131 	if (retval)
6132 		msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
6133 	else
6134 		msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
6135 
6136 	igb_write_mbx(hw, msgbuf, 1, vf);
6137 }
6138 
6139 static void igb_msg_task(struct igb_adapter *adapter)
6140 {
6141 	struct e1000_hw *hw = &adapter->hw;
6142 	u32 vf;
6143 
6144 	for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
6145 		/* process any reset requests */
6146 		if (!igb_check_for_rst(hw, vf))
6147 			igb_vf_reset_event(adapter, vf);
6148 
6149 		/* process any messages pending */
6150 		if (!igb_check_for_msg(hw, vf))
6151 			igb_rcv_msg_from_vf(adapter, vf);
6152 
6153 		/* process any acks */
6154 		if (!igb_check_for_ack(hw, vf))
6155 			igb_rcv_ack_from_vf(adapter, vf);
6156 	}
6157 }
6158 
6159 /**
6160  *  igb_set_uta - Set unicast filter table address
6161  *  @adapter: board private structure
6162  *
6163  *  The unicast table address is a register array of 32-bit registers.
6164  *  The table is meant to be used in a way similar to how the MTA is used
6165  *  however due to certain limitations in the hardware it is necessary to
6166  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
6167  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
6168  **/
6169 static void igb_set_uta(struct igb_adapter *adapter)
6170 {
6171 	struct e1000_hw *hw = &adapter->hw;
6172 	int i;
6173 
6174 	/* The UTA table only exists on 82576 hardware and newer */
6175 	if (hw->mac.type < e1000_82576)
6176 		return;
6177 
6178 	/* we only need to do this if VMDq is enabled */
6179 	if (!adapter->vfs_allocated_count)
6180 		return;
6181 
6182 	for (i = 0; i < hw->mac.uta_reg_count; i++)
6183 		array_wr32(E1000_UTA, i, ~0);
6184 }
6185 
6186 /**
6187  *  igb_intr_msi - Interrupt Handler
6188  *  @irq: interrupt number
6189  *  @data: pointer to a network interface device structure
6190  **/
6191 static irqreturn_t igb_intr_msi(int irq, void *data)
6192 {
6193 	struct igb_adapter *adapter = data;
6194 	struct igb_q_vector *q_vector = adapter->q_vector[0];
6195 	struct e1000_hw *hw = &adapter->hw;
6196 	/* read ICR disables interrupts using IAM */
6197 	u32 icr = rd32(E1000_ICR);
6198 
6199 	igb_write_itr(q_vector);
6200 
6201 	if (icr & E1000_ICR_DRSTA)
6202 		schedule_work(&adapter->reset_task);
6203 
6204 	if (icr & E1000_ICR_DOUTSYNC) {
6205 		/* HW is reporting DMA is out of sync */
6206 		adapter->stats.doosync++;
6207 	}
6208 
6209 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
6210 		hw->mac.get_link_status = 1;
6211 		if (!test_bit(__IGB_DOWN, &adapter->state))
6212 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6213 	}
6214 
6215 	if (icr & E1000_ICR_TS) {
6216 		u32 tsicr = rd32(E1000_TSICR);
6217 
6218 		if (tsicr & E1000_TSICR_TXTS) {
6219 			/* acknowledge the interrupt */
6220 			wr32(E1000_TSICR, E1000_TSICR_TXTS);
6221 			/* retrieve hardware timestamp */
6222 			schedule_work(&adapter->ptp_tx_work);
6223 		}
6224 	}
6225 
6226 	napi_schedule(&q_vector->napi);
6227 
6228 	return IRQ_HANDLED;
6229 }
6230 
6231 /**
6232  *  igb_intr - Legacy Interrupt Handler
6233  *  @irq: interrupt number
6234  *  @data: pointer to a network interface device structure
6235  **/
6236 static irqreturn_t igb_intr(int irq, void *data)
6237 {
6238 	struct igb_adapter *adapter = data;
6239 	struct igb_q_vector *q_vector = adapter->q_vector[0];
6240 	struct e1000_hw *hw = &adapter->hw;
6241 	/* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
6242 	 * need for the IMC write
6243 	 */
6244 	u32 icr = rd32(E1000_ICR);
6245 
6246 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
6247 	 * not set, then the adapter didn't send an interrupt
6248 	 */
6249 	if (!(icr & E1000_ICR_INT_ASSERTED))
6250 		return IRQ_NONE;
6251 
6252 	igb_write_itr(q_vector);
6253 
6254 	if (icr & E1000_ICR_DRSTA)
6255 		schedule_work(&adapter->reset_task);
6256 
6257 	if (icr & E1000_ICR_DOUTSYNC) {
6258 		/* HW is reporting DMA is out of sync */
6259 		adapter->stats.doosync++;
6260 	}
6261 
6262 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
6263 		hw->mac.get_link_status = 1;
6264 		/* guard against interrupt when we're going down */
6265 		if (!test_bit(__IGB_DOWN, &adapter->state))
6266 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6267 	}
6268 
6269 	if (icr & E1000_ICR_TS) {
6270 		u32 tsicr = rd32(E1000_TSICR);
6271 
6272 		if (tsicr & E1000_TSICR_TXTS) {
6273 			/* acknowledge the interrupt */
6274 			wr32(E1000_TSICR, E1000_TSICR_TXTS);
6275 			/* retrieve hardware timestamp */
6276 			schedule_work(&adapter->ptp_tx_work);
6277 		}
6278 	}
6279 
6280 	napi_schedule(&q_vector->napi);
6281 
6282 	return IRQ_HANDLED;
6283 }
6284 
6285 static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
6286 {
6287 	struct igb_adapter *adapter = q_vector->adapter;
6288 	struct e1000_hw *hw = &adapter->hw;
6289 
6290 	if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
6291 	    (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
6292 		if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
6293 			igb_set_itr(q_vector);
6294 		else
6295 			igb_update_ring_itr(q_vector);
6296 	}
6297 
6298 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
6299 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
6300 			wr32(E1000_EIMS, q_vector->eims_value);
6301 		else
6302 			igb_irq_enable(adapter);
6303 	}
6304 }
6305 
6306 /**
6307  *  igb_poll - NAPI Rx polling callback
6308  *  @napi: napi polling structure
6309  *  @budget: count of how many packets we should handle
6310  **/
6311 static int igb_poll(struct napi_struct *napi, int budget)
6312 {
6313 	struct igb_q_vector *q_vector = container_of(napi,
6314 						     struct igb_q_vector,
6315 						     napi);
6316 	bool clean_complete = true;
6317 
6318 #ifdef CONFIG_IGB_DCA
6319 	if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
6320 		igb_update_dca(q_vector);
6321 #endif
6322 	if (q_vector->tx.ring)
6323 		clean_complete = igb_clean_tx_irq(q_vector);
6324 
6325 	if (q_vector->rx.ring)
6326 		clean_complete &= igb_clean_rx_irq(q_vector, budget);
6327 
6328 	/* If all work not completed, return budget and keep polling */
6329 	if (!clean_complete)
6330 		return budget;
6331 
6332 	/* If not enough Rx work done, exit the polling mode */
6333 	napi_complete(napi);
6334 	igb_ring_irq_enable(q_vector);
6335 
6336 	return 0;
6337 }
6338 
6339 /**
6340  *  igb_clean_tx_irq - Reclaim resources after transmit completes
6341  *  @q_vector: pointer to q_vector containing needed info
6342  *
6343  *  returns true if ring is completely cleaned
6344  **/
6345 static bool igb_clean_tx_irq(struct igb_q_vector *q_vector)
6346 {
6347 	struct igb_adapter *adapter = q_vector->adapter;
6348 	struct igb_ring *tx_ring = q_vector->tx.ring;
6349 	struct igb_tx_buffer *tx_buffer;
6350 	union e1000_adv_tx_desc *tx_desc;
6351 	unsigned int total_bytes = 0, total_packets = 0;
6352 	unsigned int budget = q_vector->tx.work_limit;
6353 	unsigned int i = tx_ring->next_to_clean;
6354 
6355 	if (test_bit(__IGB_DOWN, &adapter->state))
6356 		return true;
6357 
6358 	tx_buffer = &tx_ring->tx_buffer_info[i];
6359 	tx_desc = IGB_TX_DESC(tx_ring, i);
6360 	i -= tx_ring->count;
6361 
6362 	do {
6363 		union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
6364 
6365 		/* if next_to_watch is not set then there is no work pending */
6366 		if (!eop_desc)
6367 			break;
6368 
6369 		/* prevent any other reads prior to eop_desc */
6370 		read_barrier_depends();
6371 
6372 		/* if DD is not set pending work has not been completed */
6373 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
6374 			break;
6375 
6376 		/* clear next_to_watch to prevent false hangs */
6377 		tx_buffer->next_to_watch = NULL;
6378 
6379 		/* update the statistics for this packet */
6380 		total_bytes += tx_buffer->bytecount;
6381 		total_packets += tx_buffer->gso_segs;
6382 
6383 		/* free the skb */
6384 		dev_consume_skb_any(tx_buffer->skb);
6385 
6386 		/* unmap skb header data */
6387 		dma_unmap_single(tx_ring->dev,
6388 				 dma_unmap_addr(tx_buffer, dma),
6389 				 dma_unmap_len(tx_buffer, len),
6390 				 DMA_TO_DEVICE);
6391 
6392 		/* clear tx_buffer data */
6393 		tx_buffer->skb = NULL;
6394 		dma_unmap_len_set(tx_buffer, len, 0);
6395 
6396 		/* clear last DMA location and unmap remaining buffers */
6397 		while (tx_desc != eop_desc) {
6398 			tx_buffer++;
6399 			tx_desc++;
6400 			i++;
6401 			if (unlikely(!i)) {
6402 				i -= tx_ring->count;
6403 				tx_buffer = tx_ring->tx_buffer_info;
6404 				tx_desc = IGB_TX_DESC(tx_ring, 0);
6405 			}
6406 
6407 			/* unmap any remaining paged data */
6408 			if (dma_unmap_len(tx_buffer, len)) {
6409 				dma_unmap_page(tx_ring->dev,
6410 					       dma_unmap_addr(tx_buffer, dma),
6411 					       dma_unmap_len(tx_buffer, len),
6412 					       DMA_TO_DEVICE);
6413 				dma_unmap_len_set(tx_buffer, len, 0);
6414 			}
6415 		}
6416 
6417 		/* move us one more past the eop_desc for start of next pkt */
6418 		tx_buffer++;
6419 		tx_desc++;
6420 		i++;
6421 		if (unlikely(!i)) {
6422 			i -= tx_ring->count;
6423 			tx_buffer = tx_ring->tx_buffer_info;
6424 			tx_desc = IGB_TX_DESC(tx_ring, 0);
6425 		}
6426 
6427 		/* issue prefetch for next Tx descriptor */
6428 		prefetch(tx_desc);
6429 
6430 		/* update budget accounting */
6431 		budget--;
6432 	} while (likely(budget));
6433 
6434 	netdev_tx_completed_queue(txring_txq(tx_ring),
6435 				  total_packets, total_bytes);
6436 	i += tx_ring->count;
6437 	tx_ring->next_to_clean = i;
6438 	u64_stats_update_begin(&tx_ring->tx_syncp);
6439 	tx_ring->tx_stats.bytes += total_bytes;
6440 	tx_ring->tx_stats.packets += total_packets;
6441 	u64_stats_update_end(&tx_ring->tx_syncp);
6442 	q_vector->tx.total_bytes += total_bytes;
6443 	q_vector->tx.total_packets += total_packets;
6444 
6445 	if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
6446 		struct e1000_hw *hw = &adapter->hw;
6447 
6448 		/* Detect a transmit hang in hardware, this serializes the
6449 		 * check with the clearing of time_stamp and movement of i
6450 		 */
6451 		clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
6452 		if (tx_buffer->next_to_watch &&
6453 		    time_after(jiffies, tx_buffer->time_stamp +
6454 			       (adapter->tx_timeout_factor * HZ)) &&
6455 		    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
6456 
6457 			/* detected Tx unit hang */
6458 			dev_err(tx_ring->dev,
6459 				"Detected Tx Unit Hang\n"
6460 				"  Tx Queue             <%d>\n"
6461 				"  TDH                  <%x>\n"
6462 				"  TDT                  <%x>\n"
6463 				"  next_to_use          <%x>\n"
6464 				"  next_to_clean        <%x>\n"
6465 				"buffer_info[next_to_clean]\n"
6466 				"  time_stamp           <%lx>\n"
6467 				"  next_to_watch        <%p>\n"
6468 				"  jiffies              <%lx>\n"
6469 				"  desc.status          <%x>\n",
6470 				tx_ring->queue_index,
6471 				rd32(E1000_TDH(tx_ring->reg_idx)),
6472 				readl(tx_ring->tail),
6473 				tx_ring->next_to_use,
6474 				tx_ring->next_to_clean,
6475 				tx_buffer->time_stamp,
6476 				tx_buffer->next_to_watch,
6477 				jiffies,
6478 				tx_buffer->next_to_watch->wb.status);
6479 			netif_stop_subqueue(tx_ring->netdev,
6480 					    tx_ring->queue_index);
6481 
6482 			/* we are about to reset, no point in enabling stuff */
6483 			return true;
6484 		}
6485 	}
6486 
6487 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
6488 	if (unlikely(total_packets &&
6489 	    netif_carrier_ok(tx_ring->netdev) &&
6490 	    igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
6491 		/* Make sure that anybody stopping the queue after this
6492 		 * sees the new next_to_clean.
6493 		 */
6494 		smp_mb();
6495 		if (__netif_subqueue_stopped(tx_ring->netdev,
6496 					     tx_ring->queue_index) &&
6497 		    !(test_bit(__IGB_DOWN, &adapter->state))) {
6498 			netif_wake_subqueue(tx_ring->netdev,
6499 					    tx_ring->queue_index);
6500 
6501 			u64_stats_update_begin(&tx_ring->tx_syncp);
6502 			tx_ring->tx_stats.restart_queue++;
6503 			u64_stats_update_end(&tx_ring->tx_syncp);
6504 		}
6505 	}
6506 
6507 	return !!budget;
6508 }
6509 
6510 /**
6511  *  igb_reuse_rx_page - page flip buffer and store it back on the ring
6512  *  @rx_ring: rx descriptor ring to store buffers on
6513  *  @old_buff: donor buffer to have page reused
6514  *
6515  *  Synchronizes page for reuse by the adapter
6516  **/
6517 static void igb_reuse_rx_page(struct igb_ring *rx_ring,
6518 			      struct igb_rx_buffer *old_buff)
6519 {
6520 	struct igb_rx_buffer *new_buff;
6521 	u16 nta = rx_ring->next_to_alloc;
6522 
6523 	new_buff = &rx_ring->rx_buffer_info[nta];
6524 
6525 	/* update, and store next to alloc */
6526 	nta++;
6527 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
6528 
6529 	/* transfer page from old buffer to new buffer */
6530 	*new_buff = *old_buff;
6531 
6532 	/* sync the buffer for use by the device */
6533 	dma_sync_single_range_for_device(rx_ring->dev, old_buff->dma,
6534 					 old_buff->page_offset,
6535 					 IGB_RX_BUFSZ,
6536 					 DMA_FROM_DEVICE);
6537 }
6538 
6539 static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer,
6540 				  struct page *page,
6541 				  unsigned int truesize)
6542 {
6543 	/* avoid re-using remote pages */
6544 	if (unlikely(page_to_nid(page) != numa_node_id()))
6545 		return false;
6546 
6547 	if (unlikely(page->pfmemalloc))
6548 		return false;
6549 
6550 #if (PAGE_SIZE < 8192)
6551 	/* if we are only owner of page we can reuse it */
6552 	if (unlikely(page_count(page) != 1))
6553 		return false;
6554 
6555 	/* flip page offset to other buffer */
6556 	rx_buffer->page_offset ^= IGB_RX_BUFSZ;
6557 
6558 	/* Even if we own the page, we are not allowed to use atomic_set()
6559 	 * This would break get_page_unless_zero() users.
6560 	 */
6561 	atomic_inc(&page->_count);
6562 #else
6563 	/* move offset up to the next cache line */
6564 	rx_buffer->page_offset += truesize;
6565 
6566 	if (rx_buffer->page_offset > (PAGE_SIZE - IGB_RX_BUFSZ))
6567 		return false;
6568 
6569 	/* bump ref count on page before it is given to the stack */
6570 	get_page(page);
6571 #endif
6572 
6573 	return true;
6574 }
6575 
6576 /**
6577  *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
6578  *  @rx_ring: rx descriptor ring to transact packets on
6579  *  @rx_buffer: buffer containing page to add
6580  *  @rx_desc: descriptor containing length of buffer written by hardware
6581  *  @skb: sk_buff to place the data into
6582  *
6583  *  This function will add the data contained in rx_buffer->page to the skb.
6584  *  This is done either through a direct copy if the data in the buffer is
6585  *  less than the skb header size, otherwise it will just attach the page as
6586  *  a frag to the skb.
6587  *
6588  *  The function will then update the page offset if necessary and return
6589  *  true if the buffer can be reused by the adapter.
6590  **/
6591 static bool igb_add_rx_frag(struct igb_ring *rx_ring,
6592 			    struct igb_rx_buffer *rx_buffer,
6593 			    union e1000_adv_rx_desc *rx_desc,
6594 			    struct sk_buff *skb)
6595 {
6596 	struct page *page = rx_buffer->page;
6597 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
6598 #if (PAGE_SIZE < 8192)
6599 	unsigned int truesize = IGB_RX_BUFSZ;
6600 #else
6601 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
6602 #endif
6603 
6604 	if ((size <= IGB_RX_HDR_LEN) && !skb_is_nonlinear(skb)) {
6605 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
6606 
6607 		if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
6608 			igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
6609 			va += IGB_TS_HDR_LEN;
6610 			size -= IGB_TS_HDR_LEN;
6611 		}
6612 
6613 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
6614 
6615 		/* we can reuse buffer as-is, just make sure it is local */
6616 		if (likely((page_to_nid(page) == numa_node_id()) &&
6617 			   !page->pfmemalloc))
6618 			return true;
6619 
6620 		/* this page cannot be reused so discard it */
6621 		put_page(page);
6622 		return false;
6623 	}
6624 
6625 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
6626 			rx_buffer->page_offset, size, truesize);
6627 
6628 	return igb_can_reuse_rx_page(rx_buffer, page, truesize);
6629 }
6630 
6631 static struct sk_buff *igb_fetch_rx_buffer(struct igb_ring *rx_ring,
6632 					   union e1000_adv_rx_desc *rx_desc,
6633 					   struct sk_buff *skb)
6634 {
6635 	struct igb_rx_buffer *rx_buffer;
6636 	struct page *page;
6637 
6638 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
6639 
6640 	page = rx_buffer->page;
6641 	prefetchw(page);
6642 
6643 	if (likely(!skb)) {
6644 		void *page_addr = page_address(page) +
6645 				  rx_buffer->page_offset;
6646 
6647 		/* prefetch first cache line of first page */
6648 		prefetch(page_addr);
6649 #if L1_CACHE_BYTES < 128
6650 		prefetch(page_addr + L1_CACHE_BYTES);
6651 #endif
6652 
6653 		/* allocate a skb to store the frags */
6654 		skb = netdev_alloc_skb_ip_align(rx_ring->netdev,
6655 						IGB_RX_HDR_LEN);
6656 		if (unlikely(!skb)) {
6657 			rx_ring->rx_stats.alloc_failed++;
6658 			return NULL;
6659 		}
6660 
6661 		/* we will be copying header into skb->data in
6662 		 * pskb_may_pull so it is in our interest to prefetch
6663 		 * it now to avoid a possible cache miss
6664 		 */
6665 		prefetchw(skb->data);
6666 	}
6667 
6668 	/* we are reusing so sync this buffer for CPU use */
6669 	dma_sync_single_range_for_cpu(rx_ring->dev,
6670 				      rx_buffer->dma,
6671 				      rx_buffer->page_offset,
6672 				      IGB_RX_BUFSZ,
6673 				      DMA_FROM_DEVICE);
6674 
6675 	/* pull page into skb */
6676 	if (igb_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
6677 		/* hand second half of page back to the ring */
6678 		igb_reuse_rx_page(rx_ring, rx_buffer);
6679 	} else {
6680 		/* we are not reusing the buffer so unmap it */
6681 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
6682 			       PAGE_SIZE, DMA_FROM_DEVICE);
6683 	}
6684 
6685 	/* clear contents of rx_buffer */
6686 	rx_buffer->page = NULL;
6687 
6688 	return skb;
6689 }
6690 
6691 static inline void igb_rx_checksum(struct igb_ring *ring,
6692 				   union e1000_adv_rx_desc *rx_desc,
6693 				   struct sk_buff *skb)
6694 {
6695 	skb_checksum_none_assert(skb);
6696 
6697 	/* Ignore Checksum bit is set */
6698 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
6699 		return;
6700 
6701 	/* Rx checksum disabled via ethtool */
6702 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
6703 		return;
6704 
6705 	/* TCP/UDP checksum error bit is set */
6706 	if (igb_test_staterr(rx_desc,
6707 			     E1000_RXDEXT_STATERR_TCPE |
6708 			     E1000_RXDEXT_STATERR_IPE)) {
6709 		/* work around errata with sctp packets where the TCPE aka
6710 		 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
6711 		 * packets, (aka let the stack check the crc32c)
6712 		 */
6713 		if (!((skb->len == 60) &&
6714 		      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
6715 			u64_stats_update_begin(&ring->rx_syncp);
6716 			ring->rx_stats.csum_err++;
6717 			u64_stats_update_end(&ring->rx_syncp);
6718 		}
6719 		/* let the stack verify checksum errors */
6720 		return;
6721 	}
6722 	/* It must be a TCP or UDP packet with a valid checksum */
6723 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
6724 				      E1000_RXD_STAT_UDPCS))
6725 		skb->ip_summed = CHECKSUM_UNNECESSARY;
6726 
6727 	dev_dbg(ring->dev, "cksum success: bits %08X\n",
6728 		le32_to_cpu(rx_desc->wb.upper.status_error));
6729 }
6730 
6731 static inline void igb_rx_hash(struct igb_ring *ring,
6732 			       union e1000_adv_rx_desc *rx_desc,
6733 			       struct sk_buff *skb)
6734 {
6735 	if (ring->netdev->features & NETIF_F_RXHASH)
6736 		skb_set_hash(skb,
6737 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
6738 			     PKT_HASH_TYPE_L3);
6739 }
6740 
6741 /**
6742  *  igb_is_non_eop - process handling of non-EOP buffers
6743  *  @rx_ring: Rx ring being processed
6744  *  @rx_desc: Rx descriptor for current buffer
6745  *  @skb: current socket buffer containing buffer in progress
6746  *
6747  *  This function updates next to clean.  If the buffer is an EOP buffer
6748  *  this function exits returning false, otherwise it will place the
6749  *  sk_buff in the next buffer to be chained and return true indicating
6750  *  that this is in fact a non-EOP buffer.
6751  **/
6752 static bool igb_is_non_eop(struct igb_ring *rx_ring,
6753 			   union e1000_adv_rx_desc *rx_desc)
6754 {
6755 	u32 ntc = rx_ring->next_to_clean + 1;
6756 
6757 	/* fetch, update, and store next to clean */
6758 	ntc = (ntc < rx_ring->count) ? ntc : 0;
6759 	rx_ring->next_to_clean = ntc;
6760 
6761 	prefetch(IGB_RX_DESC(rx_ring, ntc));
6762 
6763 	if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
6764 		return false;
6765 
6766 	return true;
6767 }
6768 
6769 /**
6770  *  igb_pull_tail - igb specific version of skb_pull_tail
6771  *  @rx_ring: rx descriptor ring packet is being transacted on
6772  *  @rx_desc: pointer to the EOP Rx descriptor
6773  *  @skb: pointer to current skb being adjusted
6774  *
6775  *  This function is an igb specific version of __pskb_pull_tail.  The
6776  *  main difference between this version and the original function is that
6777  *  this function can make several assumptions about the state of things
6778  *  that allow for significant optimizations versus the standard function.
6779  *  As a result we can do things like drop a frag and maintain an accurate
6780  *  truesize for the skb.
6781  */
6782 static void igb_pull_tail(struct igb_ring *rx_ring,
6783 			  union e1000_adv_rx_desc *rx_desc,
6784 			  struct sk_buff *skb)
6785 {
6786 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
6787 	unsigned char *va;
6788 	unsigned int pull_len;
6789 
6790 	/* it is valid to use page_address instead of kmap since we are
6791 	 * working with pages allocated out of the lomem pool per
6792 	 * alloc_page(GFP_ATOMIC)
6793 	 */
6794 	va = skb_frag_address(frag);
6795 
6796 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
6797 		/* retrieve timestamp from buffer */
6798 		igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
6799 
6800 		/* update pointers to remove timestamp header */
6801 		skb_frag_size_sub(frag, IGB_TS_HDR_LEN);
6802 		frag->page_offset += IGB_TS_HDR_LEN;
6803 		skb->data_len -= IGB_TS_HDR_LEN;
6804 		skb->len -= IGB_TS_HDR_LEN;
6805 
6806 		/* move va to start of packet data */
6807 		va += IGB_TS_HDR_LEN;
6808 	}
6809 
6810 	/* we need the header to contain the greater of either ETH_HLEN or
6811 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
6812 	 */
6813 	pull_len = eth_get_headlen(va, IGB_RX_HDR_LEN);
6814 
6815 	/* align pull length to size of long to optimize memcpy performance */
6816 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
6817 
6818 	/* update all of the pointers */
6819 	skb_frag_size_sub(frag, pull_len);
6820 	frag->page_offset += pull_len;
6821 	skb->data_len -= pull_len;
6822 	skb->tail += pull_len;
6823 }
6824 
6825 /**
6826  *  igb_cleanup_headers - Correct corrupted or empty headers
6827  *  @rx_ring: rx descriptor ring packet is being transacted on
6828  *  @rx_desc: pointer to the EOP Rx descriptor
6829  *  @skb: pointer to current skb being fixed
6830  *
6831  *  Address the case where we are pulling data in on pages only
6832  *  and as such no data is present in the skb header.
6833  *
6834  *  In addition if skb is not at least 60 bytes we need to pad it so that
6835  *  it is large enough to qualify as a valid Ethernet frame.
6836  *
6837  *  Returns true if an error was encountered and skb was freed.
6838  **/
6839 static bool igb_cleanup_headers(struct igb_ring *rx_ring,
6840 				union e1000_adv_rx_desc *rx_desc,
6841 				struct sk_buff *skb)
6842 {
6843 	if (unlikely((igb_test_staterr(rx_desc,
6844 				       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
6845 		struct net_device *netdev = rx_ring->netdev;
6846 		if (!(netdev->features & NETIF_F_RXALL)) {
6847 			dev_kfree_skb_any(skb);
6848 			return true;
6849 		}
6850 	}
6851 
6852 	/* place header in linear portion of buffer */
6853 	if (skb_is_nonlinear(skb))
6854 		igb_pull_tail(rx_ring, rx_desc, skb);
6855 
6856 	/* if skb_pad returns an error the skb was freed */
6857 	if (unlikely(skb->len < 60)) {
6858 		int pad_len = 60 - skb->len;
6859 
6860 		if (skb_pad(skb, pad_len))
6861 			return true;
6862 		__skb_put(skb, pad_len);
6863 	}
6864 
6865 	return false;
6866 }
6867 
6868 /**
6869  *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
6870  *  @rx_ring: rx descriptor ring packet is being transacted on
6871  *  @rx_desc: pointer to the EOP Rx descriptor
6872  *  @skb: pointer to current skb being populated
6873  *
6874  *  This function checks the ring, descriptor, and packet information in
6875  *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
6876  *  other fields within the skb.
6877  **/
6878 static void igb_process_skb_fields(struct igb_ring *rx_ring,
6879 				   union e1000_adv_rx_desc *rx_desc,
6880 				   struct sk_buff *skb)
6881 {
6882 	struct net_device *dev = rx_ring->netdev;
6883 
6884 	igb_rx_hash(rx_ring, rx_desc, skb);
6885 
6886 	igb_rx_checksum(rx_ring, rx_desc, skb);
6887 
6888 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
6889 	    !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
6890 		igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
6891 
6892 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
6893 	    igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
6894 		u16 vid;
6895 
6896 		if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
6897 		    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
6898 			vid = be16_to_cpu(rx_desc->wb.upper.vlan);
6899 		else
6900 			vid = le16_to_cpu(rx_desc->wb.upper.vlan);
6901 
6902 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
6903 	}
6904 
6905 	skb_record_rx_queue(skb, rx_ring->queue_index);
6906 
6907 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
6908 }
6909 
6910 static bool igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
6911 {
6912 	struct igb_ring *rx_ring = q_vector->rx.ring;
6913 	struct sk_buff *skb = rx_ring->skb;
6914 	unsigned int total_bytes = 0, total_packets = 0;
6915 	u16 cleaned_count = igb_desc_unused(rx_ring);
6916 
6917 	while (likely(total_packets < budget)) {
6918 		union e1000_adv_rx_desc *rx_desc;
6919 
6920 		/* return some buffers to hardware, one at a time is too slow */
6921 		if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
6922 			igb_alloc_rx_buffers(rx_ring, cleaned_count);
6923 			cleaned_count = 0;
6924 		}
6925 
6926 		rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
6927 
6928 		if (!igb_test_staterr(rx_desc, E1000_RXD_STAT_DD))
6929 			break;
6930 
6931 		/* This memory barrier is needed to keep us from reading
6932 		 * any other fields out of the rx_desc until we know the
6933 		 * RXD_STAT_DD bit is set
6934 		 */
6935 		rmb();
6936 
6937 		/* retrieve a buffer from the ring */
6938 		skb = igb_fetch_rx_buffer(rx_ring, rx_desc, skb);
6939 
6940 		/* exit if we failed to retrieve a buffer */
6941 		if (!skb)
6942 			break;
6943 
6944 		cleaned_count++;
6945 
6946 		/* fetch next buffer in frame if non-eop */
6947 		if (igb_is_non_eop(rx_ring, rx_desc))
6948 			continue;
6949 
6950 		/* verify the packet layout is correct */
6951 		if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
6952 			skb = NULL;
6953 			continue;
6954 		}
6955 
6956 		/* probably a little skewed due to removing CRC */
6957 		total_bytes += skb->len;
6958 
6959 		/* populate checksum, timestamp, VLAN, and protocol */
6960 		igb_process_skb_fields(rx_ring, rx_desc, skb);
6961 
6962 		napi_gro_receive(&q_vector->napi, skb);
6963 
6964 		/* reset skb pointer */
6965 		skb = NULL;
6966 
6967 		/* update budget accounting */
6968 		total_packets++;
6969 	}
6970 
6971 	/* place incomplete frames back on ring for completion */
6972 	rx_ring->skb = skb;
6973 
6974 	u64_stats_update_begin(&rx_ring->rx_syncp);
6975 	rx_ring->rx_stats.packets += total_packets;
6976 	rx_ring->rx_stats.bytes += total_bytes;
6977 	u64_stats_update_end(&rx_ring->rx_syncp);
6978 	q_vector->rx.total_packets += total_packets;
6979 	q_vector->rx.total_bytes += total_bytes;
6980 
6981 	if (cleaned_count)
6982 		igb_alloc_rx_buffers(rx_ring, cleaned_count);
6983 
6984 	return total_packets < budget;
6985 }
6986 
6987 static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
6988 				  struct igb_rx_buffer *bi)
6989 {
6990 	struct page *page = bi->page;
6991 	dma_addr_t dma;
6992 
6993 	/* since we are recycling buffers we should seldom need to alloc */
6994 	if (likely(page))
6995 		return true;
6996 
6997 	/* alloc new page for storage */
6998 	page = __skb_alloc_page(GFP_ATOMIC | __GFP_COLD, NULL);
6999 	if (unlikely(!page)) {
7000 		rx_ring->rx_stats.alloc_failed++;
7001 		return false;
7002 	}
7003 
7004 	/* map page for use */
7005 	dma = dma_map_page(rx_ring->dev, page, 0, PAGE_SIZE, DMA_FROM_DEVICE);
7006 
7007 	/* if mapping failed free memory back to system since
7008 	 * there isn't much point in holding memory we can't use
7009 	 */
7010 	if (dma_mapping_error(rx_ring->dev, dma)) {
7011 		__free_page(page);
7012 
7013 		rx_ring->rx_stats.alloc_failed++;
7014 		return false;
7015 	}
7016 
7017 	bi->dma = dma;
7018 	bi->page = page;
7019 	bi->page_offset = 0;
7020 
7021 	return true;
7022 }
7023 
7024 /**
7025  *  igb_alloc_rx_buffers - Replace used receive buffers; packet split
7026  *  @adapter: address of board private structure
7027  **/
7028 void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
7029 {
7030 	union e1000_adv_rx_desc *rx_desc;
7031 	struct igb_rx_buffer *bi;
7032 	u16 i = rx_ring->next_to_use;
7033 
7034 	/* nothing to do */
7035 	if (!cleaned_count)
7036 		return;
7037 
7038 	rx_desc = IGB_RX_DESC(rx_ring, i);
7039 	bi = &rx_ring->rx_buffer_info[i];
7040 	i -= rx_ring->count;
7041 
7042 	do {
7043 		if (!igb_alloc_mapped_page(rx_ring, bi))
7044 			break;
7045 
7046 		/* Refresh the desc even if buffer_addrs didn't change
7047 		 * because each write-back erases this info.
7048 		 */
7049 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
7050 
7051 		rx_desc++;
7052 		bi++;
7053 		i++;
7054 		if (unlikely(!i)) {
7055 			rx_desc = IGB_RX_DESC(rx_ring, 0);
7056 			bi = rx_ring->rx_buffer_info;
7057 			i -= rx_ring->count;
7058 		}
7059 
7060 		/* clear the hdr_addr for the next_to_use descriptor */
7061 		rx_desc->read.hdr_addr = 0;
7062 
7063 		cleaned_count--;
7064 	} while (cleaned_count);
7065 
7066 	i += rx_ring->count;
7067 
7068 	if (rx_ring->next_to_use != i) {
7069 		/* record the next descriptor to use */
7070 		rx_ring->next_to_use = i;
7071 
7072 		/* update next to alloc since we have filled the ring */
7073 		rx_ring->next_to_alloc = i;
7074 
7075 		/* Force memory writes to complete before letting h/w
7076 		 * know there are new descriptors to fetch.  (Only
7077 		 * applicable for weak-ordered memory model archs,
7078 		 * such as IA-64).
7079 		 */
7080 		wmb();
7081 		writel(i, rx_ring->tail);
7082 	}
7083 }
7084 
7085 /**
7086  * igb_mii_ioctl -
7087  * @netdev:
7088  * @ifreq:
7089  * @cmd:
7090  **/
7091 static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7092 {
7093 	struct igb_adapter *adapter = netdev_priv(netdev);
7094 	struct mii_ioctl_data *data = if_mii(ifr);
7095 
7096 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
7097 		return -EOPNOTSUPP;
7098 
7099 	switch (cmd) {
7100 	case SIOCGMIIPHY:
7101 		data->phy_id = adapter->hw.phy.addr;
7102 		break;
7103 	case SIOCGMIIREG:
7104 		if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
7105 				     &data->val_out))
7106 			return -EIO;
7107 		break;
7108 	case SIOCSMIIREG:
7109 	default:
7110 		return -EOPNOTSUPP;
7111 	}
7112 	return 0;
7113 }
7114 
7115 /**
7116  * igb_ioctl -
7117  * @netdev:
7118  * @ifreq:
7119  * @cmd:
7120  **/
7121 static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7122 {
7123 	switch (cmd) {
7124 	case SIOCGMIIPHY:
7125 	case SIOCGMIIREG:
7126 	case SIOCSMIIREG:
7127 		return igb_mii_ioctl(netdev, ifr, cmd);
7128 	case SIOCGHWTSTAMP:
7129 		return igb_ptp_get_ts_config(netdev, ifr);
7130 	case SIOCSHWTSTAMP:
7131 		return igb_ptp_set_ts_config(netdev, ifr);
7132 	default:
7133 		return -EOPNOTSUPP;
7134 	}
7135 }
7136 
7137 void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
7138 {
7139 	struct igb_adapter *adapter = hw->back;
7140 
7141 	pci_read_config_word(adapter->pdev, reg, value);
7142 }
7143 
7144 void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
7145 {
7146 	struct igb_adapter *adapter = hw->back;
7147 
7148 	pci_write_config_word(adapter->pdev, reg, *value);
7149 }
7150 
7151 s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
7152 {
7153 	struct igb_adapter *adapter = hw->back;
7154 
7155 	if (pcie_capability_read_word(adapter->pdev, reg, value))
7156 		return -E1000_ERR_CONFIG;
7157 
7158 	return 0;
7159 }
7160 
7161 s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
7162 {
7163 	struct igb_adapter *adapter = hw->back;
7164 
7165 	if (pcie_capability_write_word(adapter->pdev, reg, *value))
7166 		return -E1000_ERR_CONFIG;
7167 
7168 	return 0;
7169 }
7170 
7171 static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
7172 {
7173 	struct igb_adapter *adapter = netdev_priv(netdev);
7174 	struct e1000_hw *hw = &adapter->hw;
7175 	u32 ctrl, rctl;
7176 	bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
7177 
7178 	if (enable) {
7179 		/* enable VLAN tag insert/strip */
7180 		ctrl = rd32(E1000_CTRL);
7181 		ctrl |= E1000_CTRL_VME;
7182 		wr32(E1000_CTRL, ctrl);
7183 
7184 		/* Disable CFI check */
7185 		rctl = rd32(E1000_RCTL);
7186 		rctl &= ~E1000_RCTL_CFIEN;
7187 		wr32(E1000_RCTL, rctl);
7188 	} else {
7189 		/* disable VLAN tag insert/strip */
7190 		ctrl = rd32(E1000_CTRL);
7191 		ctrl &= ~E1000_CTRL_VME;
7192 		wr32(E1000_CTRL, ctrl);
7193 	}
7194 
7195 	igb_rlpml_set(adapter);
7196 }
7197 
7198 static int igb_vlan_rx_add_vid(struct net_device *netdev,
7199 			       __be16 proto, u16 vid)
7200 {
7201 	struct igb_adapter *adapter = netdev_priv(netdev);
7202 	struct e1000_hw *hw = &adapter->hw;
7203 	int pf_id = adapter->vfs_allocated_count;
7204 
7205 	/* attempt to add filter to vlvf array */
7206 	igb_vlvf_set(adapter, vid, true, pf_id);
7207 
7208 	/* add the filter since PF can receive vlans w/o entry in vlvf */
7209 	igb_vfta_set(hw, vid, true);
7210 
7211 	set_bit(vid, adapter->active_vlans);
7212 
7213 	return 0;
7214 }
7215 
7216 static int igb_vlan_rx_kill_vid(struct net_device *netdev,
7217 				__be16 proto, u16 vid)
7218 {
7219 	struct igb_adapter *adapter = netdev_priv(netdev);
7220 	struct e1000_hw *hw = &adapter->hw;
7221 	int pf_id = adapter->vfs_allocated_count;
7222 	s32 err;
7223 
7224 	/* remove vlan from VLVF table array */
7225 	err = igb_vlvf_set(adapter, vid, false, pf_id);
7226 
7227 	/* if vid was not present in VLVF just remove it from table */
7228 	if (err)
7229 		igb_vfta_set(hw, vid, false);
7230 
7231 	clear_bit(vid, adapter->active_vlans);
7232 
7233 	return 0;
7234 }
7235 
7236 static void igb_restore_vlan(struct igb_adapter *adapter)
7237 {
7238 	u16 vid;
7239 
7240 	igb_vlan_mode(adapter->netdev, adapter->netdev->features);
7241 
7242 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
7243 		igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
7244 }
7245 
7246 int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
7247 {
7248 	struct pci_dev *pdev = adapter->pdev;
7249 	struct e1000_mac_info *mac = &adapter->hw.mac;
7250 
7251 	mac->autoneg = 0;
7252 
7253 	/* Make sure dplx is at most 1 bit and lsb of speed is not set
7254 	 * for the switch() below to work
7255 	 */
7256 	if ((spd & 1) || (dplx & ~1))
7257 		goto err_inval;
7258 
7259 	/* Fiber NIC's only allow 1000 gbps Full duplex
7260 	 * and 100Mbps Full duplex for 100baseFx sfp
7261 	 */
7262 	if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
7263 		switch (spd + dplx) {
7264 		case SPEED_10 + DUPLEX_HALF:
7265 		case SPEED_10 + DUPLEX_FULL:
7266 		case SPEED_100 + DUPLEX_HALF:
7267 			goto err_inval;
7268 		default:
7269 			break;
7270 		}
7271 	}
7272 
7273 	switch (spd + dplx) {
7274 	case SPEED_10 + DUPLEX_HALF:
7275 		mac->forced_speed_duplex = ADVERTISE_10_HALF;
7276 		break;
7277 	case SPEED_10 + DUPLEX_FULL:
7278 		mac->forced_speed_duplex = ADVERTISE_10_FULL;
7279 		break;
7280 	case SPEED_100 + DUPLEX_HALF:
7281 		mac->forced_speed_duplex = ADVERTISE_100_HALF;
7282 		break;
7283 	case SPEED_100 + DUPLEX_FULL:
7284 		mac->forced_speed_duplex = ADVERTISE_100_FULL;
7285 		break;
7286 	case SPEED_1000 + DUPLEX_FULL:
7287 		mac->autoneg = 1;
7288 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
7289 		break;
7290 	case SPEED_1000 + DUPLEX_HALF: /* not supported */
7291 	default:
7292 		goto err_inval;
7293 	}
7294 
7295 	/* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
7296 	adapter->hw.phy.mdix = AUTO_ALL_MODES;
7297 
7298 	return 0;
7299 
7300 err_inval:
7301 	dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
7302 	return -EINVAL;
7303 }
7304 
7305 static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
7306 			  bool runtime)
7307 {
7308 	struct net_device *netdev = pci_get_drvdata(pdev);
7309 	struct igb_adapter *adapter = netdev_priv(netdev);
7310 	struct e1000_hw *hw = &adapter->hw;
7311 	u32 ctrl, rctl, status;
7312 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
7313 #ifdef CONFIG_PM
7314 	int retval = 0;
7315 #endif
7316 
7317 	netif_device_detach(netdev);
7318 
7319 	if (netif_running(netdev))
7320 		__igb_close(netdev, true);
7321 
7322 	igb_clear_interrupt_scheme(adapter);
7323 
7324 #ifdef CONFIG_PM
7325 	retval = pci_save_state(pdev);
7326 	if (retval)
7327 		return retval;
7328 #endif
7329 
7330 	status = rd32(E1000_STATUS);
7331 	if (status & E1000_STATUS_LU)
7332 		wufc &= ~E1000_WUFC_LNKC;
7333 
7334 	if (wufc) {
7335 		igb_setup_rctl(adapter);
7336 		igb_set_rx_mode(netdev);
7337 
7338 		/* turn on all-multi mode if wake on multicast is enabled */
7339 		if (wufc & E1000_WUFC_MC) {
7340 			rctl = rd32(E1000_RCTL);
7341 			rctl |= E1000_RCTL_MPE;
7342 			wr32(E1000_RCTL, rctl);
7343 		}
7344 
7345 		ctrl = rd32(E1000_CTRL);
7346 		/* advertise wake from D3Cold */
7347 		#define E1000_CTRL_ADVD3WUC 0x00100000
7348 		/* phy power management enable */
7349 		#define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
7350 		ctrl |= E1000_CTRL_ADVD3WUC;
7351 		wr32(E1000_CTRL, ctrl);
7352 
7353 		/* Allow time for pending master requests to run */
7354 		igb_disable_pcie_master(hw);
7355 
7356 		wr32(E1000_WUC, E1000_WUC_PME_EN);
7357 		wr32(E1000_WUFC, wufc);
7358 	} else {
7359 		wr32(E1000_WUC, 0);
7360 		wr32(E1000_WUFC, 0);
7361 	}
7362 
7363 	*enable_wake = wufc || adapter->en_mng_pt;
7364 	if (!*enable_wake)
7365 		igb_power_down_link(adapter);
7366 	else
7367 		igb_power_up_link(adapter);
7368 
7369 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7370 	 * would have already happened in close and is redundant.
7371 	 */
7372 	igb_release_hw_control(adapter);
7373 
7374 	pci_disable_device(pdev);
7375 
7376 	return 0;
7377 }
7378 
7379 #ifdef CONFIG_PM
7380 #ifdef CONFIG_PM_SLEEP
7381 static int igb_suspend(struct device *dev)
7382 {
7383 	int retval;
7384 	bool wake;
7385 	struct pci_dev *pdev = to_pci_dev(dev);
7386 
7387 	retval = __igb_shutdown(pdev, &wake, 0);
7388 	if (retval)
7389 		return retval;
7390 
7391 	if (wake) {
7392 		pci_prepare_to_sleep(pdev);
7393 	} else {
7394 		pci_wake_from_d3(pdev, false);
7395 		pci_set_power_state(pdev, PCI_D3hot);
7396 	}
7397 
7398 	return 0;
7399 }
7400 #endif /* CONFIG_PM_SLEEP */
7401 
7402 static int igb_resume(struct device *dev)
7403 {
7404 	struct pci_dev *pdev = to_pci_dev(dev);
7405 	struct net_device *netdev = pci_get_drvdata(pdev);
7406 	struct igb_adapter *adapter = netdev_priv(netdev);
7407 	struct e1000_hw *hw = &adapter->hw;
7408 	u32 err;
7409 
7410 	pci_set_power_state(pdev, PCI_D0);
7411 	pci_restore_state(pdev);
7412 	pci_save_state(pdev);
7413 
7414 	if (!pci_device_is_present(pdev))
7415 		return -ENODEV;
7416 	err = pci_enable_device_mem(pdev);
7417 	if (err) {
7418 		dev_err(&pdev->dev,
7419 			"igb: Cannot enable PCI device from suspend\n");
7420 		return err;
7421 	}
7422 	pci_set_master(pdev);
7423 
7424 	pci_enable_wake(pdev, PCI_D3hot, 0);
7425 	pci_enable_wake(pdev, PCI_D3cold, 0);
7426 
7427 	if (igb_init_interrupt_scheme(adapter, true)) {
7428 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
7429 		return -ENOMEM;
7430 	}
7431 
7432 	igb_reset(adapter);
7433 
7434 	/* let the f/w know that the h/w is now under the control of the
7435 	 * driver.
7436 	 */
7437 	igb_get_hw_control(adapter);
7438 
7439 	wr32(E1000_WUS, ~0);
7440 
7441 	if (netdev->flags & IFF_UP) {
7442 		rtnl_lock();
7443 		err = __igb_open(netdev, true);
7444 		rtnl_unlock();
7445 		if (err)
7446 			return err;
7447 	}
7448 
7449 	netif_device_attach(netdev);
7450 	return 0;
7451 }
7452 
7453 #ifdef CONFIG_PM_RUNTIME
7454 static int igb_runtime_idle(struct device *dev)
7455 {
7456 	struct pci_dev *pdev = to_pci_dev(dev);
7457 	struct net_device *netdev = pci_get_drvdata(pdev);
7458 	struct igb_adapter *adapter = netdev_priv(netdev);
7459 
7460 	if (!igb_has_link(adapter))
7461 		pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
7462 
7463 	return -EBUSY;
7464 }
7465 
7466 static int igb_runtime_suspend(struct device *dev)
7467 {
7468 	struct pci_dev *pdev = to_pci_dev(dev);
7469 	int retval;
7470 	bool wake;
7471 
7472 	retval = __igb_shutdown(pdev, &wake, 1);
7473 	if (retval)
7474 		return retval;
7475 
7476 	if (wake) {
7477 		pci_prepare_to_sleep(pdev);
7478 	} else {
7479 		pci_wake_from_d3(pdev, false);
7480 		pci_set_power_state(pdev, PCI_D3hot);
7481 	}
7482 
7483 	return 0;
7484 }
7485 
7486 static int igb_runtime_resume(struct device *dev)
7487 {
7488 	return igb_resume(dev);
7489 }
7490 #endif /* CONFIG_PM_RUNTIME */
7491 #endif
7492 
7493 static void igb_shutdown(struct pci_dev *pdev)
7494 {
7495 	bool wake;
7496 
7497 	__igb_shutdown(pdev, &wake, 0);
7498 
7499 	if (system_state == SYSTEM_POWER_OFF) {
7500 		pci_wake_from_d3(pdev, wake);
7501 		pci_set_power_state(pdev, PCI_D3hot);
7502 	}
7503 }
7504 
7505 #ifdef CONFIG_PCI_IOV
7506 static int igb_sriov_reinit(struct pci_dev *dev)
7507 {
7508 	struct net_device *netdev = pci_get_drvdata(dev);
7509 	struct igb_adapter *adapter = netdev_priv(netdev);
7510 	struct pci_dev *pdev = adapter->pdev;
7511 
7512 	rtnl_lock();
7513 
7514 	if (netif_running(netdev))
7515 		igb_close(netdev);
7516 	else
7517 		igb_reset(adapter);
7518 
7519 	igb_clear_interrupt_scheme(adapter);
7520 
7521 	igb_init_queue_configuration(adapter);
7522 
7523 	if (igb_init_interrupt_scheme(adapter, true)) {
7524 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
7525 		return -ENOMEM;
7526 	}
7527 
7528 	if (netif_running(netdev))
7529 		igb_open(netdev);
7530 
7531 	rtnl_unlock();
7532 
7533 	return 0;
7534 }
7535 
7536 static int igb_pci_disable_sriov(struct pci_dev *dev)
7537 {
7538 	int err = igb_disable_sriov(dev);
7539 
7540 	if (!err)
7541 		err = igb_sriov_reinit(dev);
7542 
7543 	return err;
7544 }
7545 
7546 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
7547 {
7548 	int err = igb_enable_sriov(dev, num_vfs);
7549 
7550 	if (err)
7551 		goto out;
7552 
7553 	err = igb_sriov_reinit(dev);
7554 	if (!err)
7555 		return num_vfs;
7556 
7557 out:
7558 	return err;
7559 }
7560 
7561 #endif
7562 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
7563 {
7564 #ifdef CONFIG_PCI_IOV
7565 	if (num_vfs == 0)
7566 		return igb_pci_disable_sriov(dev);
7567 	else
7568 		return igb_pci_enable_sriov(dev, num_vfs);
7569 #endif
7570 	return 0;
7571 }
7572 
7573 #ifdef CONFIG_NET_POLL_CONTROLLER
7574 /* Polling 'interrupt' - used by things like netconsole to send skbs
7575  * without having to re-enable interrupts. It's not called while
7576  * the interrupt routine is executing.
7577  */
7578 static void igb_netpoll(struct net_device *netdev)
7579 {
7580 	struct igb_adapter *adapter = netdev_priv(netdev);
7581 	struct e1000_hw *hw = &adapter->hw;
7582 	struct igb_q_vector *q_vector;
7583 	int i;
7584 
7585 	for (i = 0; i < adapter->num_q_vectors; i++) {
7586 		q_vector = adapter->q_vector[i];
7587 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
7588 			wr32(E1000_EIMC, q_vector->eims_value);
7589 		else
7590 			igb_irq_disable(adapter);
7591 		napi_schedule(&q_vector->napi);
7592 	}
7593 }
7594 #endif /* CONFIG_NET_POLL_CONTROLLER */
7595 
7596 /**
7597  *  igb_io_error_detected - called when PCI error is detected
7598  *  @pdev: Pointer to PCI device
7599  *  @state: The current pci connection state
7600  *
7601  *  This function is called after a PCI bus error affecting
7602  *  this device has been detected.
7603  **/
7604 static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
7605 					      pci_channel_state_t state)
7606 {
7607 	struct net_device *netdev = pci_get_drvdata(pdev);
7608 	struct igb_adapter *adapter = netdev_priv(netdev);
7609 
7610 	netif_device_detach(netdev);
7611 
7612 	if (state == pci_channel_io_perm_failure)
7613 		return PCI_ERS_RESULT_DISCONNECT;
7614 
7615 	if (netif_running(netdev))
7616 		igb_down(adapter);
7617 	pci_disable_device(pdev);
7618 
7619 	/* Request a slot slot reset. */
7620 	return PCI_ERS_RESULT_NEED_RESET;
7621 }
7622 
7623 /**
7624  *  igb_io_slot_reset - called after the pci bus has been reset.
7625  *  @pdev: Pointer to PCI device
7626  *
7627  *  Restart the card from scratch, as if from a cold-boot. Implementation
7628  *  resembles the first-half of the igb_resume routine.
7629  **/
7630 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
7631 {
7632 	struct net_device *netdev = pci_get_drvdata(pdev);
7633 	struct igb_adapter *adapter = netdev_priv(netdev);
7634 	struct e1000_hw *hw = &adapter->hw;
7635 	pci_ers_result_t result;
7636 	int err;
7637 
7638 	if (pci_enable_device_mem(pdev)) {
7639 		dev_err(&pdev->dev,
7640 			"Cannot re-enable PCI device after reset.\n");
7641 		result = PCI_ERS_RESULT_DISCONNECT;
7642 	} else {
7643 		pci_set_master(pdev);
7644 		pci_restore_state(pdev);
7645 		pci_save_state(pdev);
7646 
7647 		pci_enable_wake(pdev, PCI_D3hot, 0);
7648 		pci_enable_wake(pdev, PCI_D3cold, 0);
7649 
7650 		igb_reset(adapter);
7651 		wr32(E1000_WUS, ~0);
7652 		result = PCI_ERS_RESULT_RECOVERED;
7653 	}
7654 
7655 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7656 	if (err) {
7657 		dev_err(&pdev->dev,
7658 			"pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7659 			err);
7660 		/* non-fatal, continue */
7661 	}
7662 
7663 	return result;
7664 }
7665 
7666 /**
7667  *  igb_io_resume - called when traffic can start flowing again.
7668  *  @pdev: Pointer to PCI device
7669  *
7670  *  This callback is called when the error recovery driver tells us that
7671  *  its OK to resume normal operation. Implementation resembles the
7672  *  second-half of the igb_resume routine.
7673  */
7674 static void igb_io_resume(struct pci_dev *pdev)
7675 {
7676 	struct net_device *netdev = pci_get_drvdata(pdev);
7677 	struct igb_adapter *adapter = netdev_priv(netdev);
7678 
7679 	if (netif_running(netdev)) {
7680 		if (igb_up(adapter)) {
7681 			dev_err(&pdev->dev, "igb_up failed after reset\n");
7682 			return;
7683 		}
7684 	}
7685 
7686 	netif_device_attach(netdev);
7687 
7688 	/* let the f/w know that the h/w is now under the control of the
7689 	 * driver.
7690 	 */
7691 	igb_get_hw_control(adapter);
7692 }
7693 
7694 static void igb_rar_set_qsel(struct igb_adapter *adapter, u8 *addr, u32 index,
7695 			     u8 qsel)
7696 {
7697 	u32 rar_low, rar_high;
7698 	struct e1000_hw *hw = &adapter->hw;
7699 
7700 	/* HW expects these in little endian so we reverse the byte order
7701 	 * from network order (big endian) to little endian
7702 	 */
7703 	rar_low = ((u32) addr[0] | ((u32) addr[1] << 8) |
7704 		   ((u32) addr[2] << 16) | ((u32) addr[3] << 24));
7705 	rar_high = ((u32) addr[4] | ((u32) addr[5] << 8));
7706 
7707 	/* Indicate to hardware the Address is Valid. */
7708 	rar_high |= E1000_RAH_AV;
7709 
7710 	if (hw->mac.type == e1000_82575)
7711 		rar_high |= E1000_RAH_POOL_1 * qsel;
7712 	else
7713 		rar_high |= E1000_RAH_POOL_1 << qsel;
7714 
7715 	wr32(E1000_RAL(index), rar_low);
7716 	wrfl();
7717 	wr32(E1000_RAH(index), rar_high);
7718 	wrfl();
7719 }
7720 
7721 static int igb_set_vf_mac(struct igb_adapter *adapter,
7722 			  int vf, unsigned char *mac_addr)
7723 {
7724 	struct e1000_hw *hw = &adapter->hw;
7725 	/* VF MAC addresses start at end of receive addresses and moves
7726 	 * towards the first, as a result a collision should not be possible
7727 	 */
7728 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
7729 
7730 	memcpy(adapter->vf_data[vf].vf_mac_addresses, mac_addr, ETH_ALEN);
7731 
7732 	igb_rar_set_qsel(adapter, mac_addr, rar_entry, vf);
7733 
7734 	return 0;
7735 }
7736 
7737 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
7738 {
7739 	struct igb_adapter *adapter = netdev_priv(netdev);
7740 	if (!is_valid_ether_addr(mac) || (vf >= adapter->vfs_allocated_count))
7741 		return -EINVAL;
7742 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
7743 	dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n", mac, vf);
7744 	dev_info(&adapter->pdev->dev,
7745 		 "Reload the VF driver to make this change effective.");
7746 	if (test_bit(__IGB_DOWN, &adapter->state)) {
7747 		dev_warn(&adapter->pdev->dev,
7748 			 "The VF MAC address has been set, but the PF device is not up.\n");
7749 		dev_warn(&adapter->pdev->dev,
7750 			 "Bring the PF device up before attempting to use the VF device.\n");
7751 	}
7752 	return igb_set_vf_mac(adapter, vf, mac);
7753 }
7754 
7755 static int igb_link_mbps(int internal_link_speed)
7756 {
7757 	switch (internal_link_speed) {
7758 	case SPEED_100:
7759 		return 100;
7760 	case SPEED_1000:
7761 		return 1000;
7762 	default:
7763 		return 0;
7764 	}
7765 }
7766 
7767 static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
7768 				  int link_speed)
7769 {
7770 	int rf_dec, rf_int;
7771 	u32 bcnrc_val;
7772 
7773 	if (tx_rate != 0) {
7774 		/* Calculate the rate factor values to set */
7775 		rf_int = link_speed / tx_rate;
7776 		rf_dec = (link_speed - (rf_int * tx_rate));
7777 		rf_dec = (rf_dec * (1 << E1000_RTTBCNRC_RF_INT_SHIFT)) /
7778 			 tx_rate;
7779 
7780 		bcnrc_val = E1000_RTTBCNRC_RS_ENA;
7781 		bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
7782 			      E1000_RTTBCNRC_RF_INT_MASK);
7783 		bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
7784 	} else {
7785 		bcnrc_val = 0;
7786 	}
7787 
7788 	wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
7789 	/* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
7790 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
7791 	 */
7792 	wr32(E1000_RTTBCNRM, 0x14);
7793 	wr32(E1000_RTTBCNRC, bcnrc_val);
7794 }
7795 
7796 static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
7797 {
7798 	int actual_link_speed, i;
7799 	bool reset_rate = false;
7800 
7801 	/* VF TX rate limit was not set or not supported */
7802 	if ((adapter->vf_rate_link_speed == 0) ||
7803 	    (adapter->hw.mac.type != e1000_82576))
7804 		return;
7805 
7806 	actual_link_speed = igb_link_mbps(adapter->link_speed);
7807 	if (actual_link_speed != adapter->vf_rate_link_speed) {
7808 		reset_rate = true;
7809 		adapter->vf_rate_link_speed = 0;
7810 		dev_info(&adapter->pdev->dev,
7811 			 "Link speed has been changed. VF Transmit rate is disabled\n");
7812 	}
7813 
7814 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
7815 		if (reset_rate)
7816 			adapter->vf_data[i].tx_rate = 0;
7817 
7818 		igb_set_vf_rate_limit(&adapter->hw, i,
7819 				      adapter->vf_data[i].tx_rate,
7820 				      actual_link_speed);
7821 	}
7822 }
7823 
7824 static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
7825 			     int min_tx_rate, int max_tx_rate)
7826 {
7827 	struct igb_adapter *adapter = netdev_priv(netdev);
7828 	struct e1000_hw *hw = &adapter->hw;
7829 	int actual_link_speed;
7830 
7831 	if (hw->mac.type != e1000_82576)
7832 		return -EOPNOTSUPP;
7833 
7834 	if (min_tx_rate)
7835 		return -EINVAL;
7836 
7837 	actual_link_speed = igb_link_mbps(adapter->link_speed);
7838 	if ((vf >= adapter->vfs_allocated_count) ||
7839 	    (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
7840 	    (max_tx_rate < 0) ||
7841 	    (max_tx_rate > actual_link_speed))
7842 		return -EINVAL;
7843 
7844 	adapter->vf_rate_link_speed = actual_link_speed;
7845 	adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
7846 	igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
7847 
7848 	return 0;
7849 }
7850 
7851 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
7852 				   bool setting)
7853 {
7854 	struct igb_adapter *adapter = netdev_priv(netdev);
7855 	struct e1000_hw *hw = &adapter->hw;
7856 	u32 reg_val, reg_offset;
7857 
7858 	if (!adapter->vfs_allocated_count)
7859 		return -EOPNOTSUPP;
7860 
7861 	if (vf >= adapter->vfs_allocated_count)
7862 		return -EINVAL;
7863 
7864 	reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
7865 	reg_val = rd32(reg_offset);
7866 	if (setting)
7867 		reg_val |= ((1 << vf) |
7868 			    (1 << (vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT)));
7869 	else
7870 		reg_val &= ~((1 << vf) |
7871 			     (1 << (vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT)));
7872 	wr32(reg_offset, reg_val);
7873 
7874 	adapter->vf_data[vf].spoofchk_enabled = setting;
7875 	return 0;
7876 }
7877 
7878 static int igb_ndo_get_vf_config(struct net_device *netdev,
7879 				 int vf, struct ifla_vf_info *ivi)
7880 {
7881 	struct igb_adapter *adapter = netdev_priv(netdev);
7882 	if (vf >= adapter->vfs_allocated_count)
7883 		return -EINVAL;
7884 	ivi->vf = vf;
7885 	memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
7886 	ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
7887 	ivi->min_tx_rate = 0;
7888 	ivi->vlan = adapter->vf_data[vf].pf_vlan;
7889 	ivi->qos = adapter->vf_data[vf].pf_qos;
7890 	ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
7891 	return 0;
7892 }
7893 
7894 static void igb_vmm_control(struct igb_adapter *adapter)
7895 {
7896 	struct e1000_hw *hw = &adapter->hw;
7897 	u32 reg;
7898 
7899 	switch (hw->mac.type) {
7900 	case e1000_82575:
7901 	case e1000_i210:
7902 	case e1000_i211:
7903 	case e1000_i354:
7904 	default:
7905 		/* replication is not supported for 82575 */
7906 		return;
7907 	case e1000_82576:
7908 		/* notify HW that the MAC is adding vlan tags */
7909 		reg = rd32(E1000_DTXCTL);
7910 		reg |= E1000_DTXCTL_VLAN_ADDED;
7911 		wr32(E1000_DTXCTL, reg);
7912 		/* Fall through */
7913 	case e1000_82580:
7914 		/* enable replication vlan tag stripping */
7915 		reg = rd32(E1000_RPLOLR);
7916 		reg |= E1000_RPLOLR_STRVLAN;
7917 		wr32(E1000_RPLOLR, reg);
7918 		/* Fall through */
7919 	case e1000_i350:
7920 		/* none of the above registers are supported by i350 */
7921 		break;
7922 	}
7923 
7924 	if (adapter->vfs_allocated_count) {
7925 		igb_vmdq_set_loopback_pf(hw, true);
7926 		igb_vmdq_set_replication_pf(hw, true);
7927 		igb_vmdq_set_anti_spoofing_pf(hw, true,
7928 					      adapter->vfs_allocated_count);
7929 	} else {
7930 		igb_vmdq_set_loopback_pf(hw, false);
7931 		igb_vmdq_set_replication_pf(hw, false);
7932 	}
7933 }
7934 
7935 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
7936 {
7937 	struct e1000_hw *hw = &adapter->hw;
7938 	u32 dmac_thr;
7939 	u16 hwm;
7940 
7941 	if (hw->mac.type > e1000_82580) {
7942 		if (adapter->flags & IGB_FLAG_DMAC) {
7943 			u32 reg;
7944 
7945 			/* force threshold to 0. */
7946 			wr32(E1000_DMCTXTH, 0);
7947 
7948 			/* DMA Coalescing high water mark needs to be greater
7949 			 * than the Rx threshold. Set hwm to PBA - max frame
7950 			 * size in 16B units, capping it at PBA - 6KB.
7951 			 */
7952 			hwm = 64 * pba - adapter->max_frame_size / 16;
7953 			if (hwm < 64 * (pba - 6))
7954 				hwm = 64 * (pba - 6);
7955 			reg = rd32(E1000_FCRTC);
7956 			reg &= ~E1000_FCRTC_RTH_COAL_MASK;
7957 			reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
7958 				& E1000_FCRTC_RTH_COAL_MASK);
7959 			wr32(E1000_FCRTC, reg);
7960 
7961 			/* Set the DMA Coalescing Rx threshold to PBA - 2 * max
7962 			 * frame size, capping it at PBA - 10KB.
7963 			 */
7964 			dmac_thr = pba - adapter->max_frame_size / 512;
7965 			if (dmac_thr < pba - 10)
7966 				dmac_thr = pba - 10;
7967 			reg = rd32(E1000_DMACR);
7968 			reg &= ~E1000_DMACR_DMACTHR_MASK;
7969 			reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
7970 				& E1000_DMACR_DMACTHR_MASK);
7971 
7972 			/* transition to L0x or L1 if available..*/
7973 			reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
7974 
7975 			/* watchdog timer= +-1000 usec in 32usec intervals */
7976 			reg |= (1000 >> 5);
7977 
7978 			/* Disable BMC-to-OS Watchdog Enable */
7979 			if (hw->mac.type != e1000_i354)
7980 				reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
7981 
7982 			wr32(E1000_DMACR, reg);
7983 
7984 			/* no lower threshold to disable
7985 			 * coalescing(smart fifb)-UTRESH=0
7986 			 */
7987 			wr32(E1000_DMCRTRH, 0);
7988 
7989 			reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
7990 
7991 			wr32(E1000_DMCTLX, reg);
7992 
7993 			/* free space in tx packet buffer to wake from
7994 			 * DMA coal
7995 			 */
7996 			wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
7997 			     (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
7998 
7999 			/* make low power state decision controlled
8000 			 * by DMA coal
8001 			 */
8002 			reg = rd32(E1000_PCIEMISC);
8003 			reg &= ~E1000_PCIEMISC_LX_DECISION;
8004 			wr32(E1000_PCIEMISC, reg);
8005 		} /* endif adapter->dmac is not disabled */
8006 	} else if (hw->mac.type == e1000_82580) {
8007 		u32 reg = rd32(E1000_PCIEMISC);
8008 
8009 		wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
8010 		wr32(E1000_DMACR, 0);
8011 	}
8012 }
8013 
8014 /**
8015  *  igb_read_i2c_byte - Reads 8 bit word over I2C
8016  *  @hw: pointer to hardware structure
8017  *  @byte_offset: byte offset to read
8018  *  @dev_addr: device address
8019  *  @data: value read
8020  *
8021  *  Performs byte read operation over I2C interface at
8022  *  a specified device address.
8023  **/
8024 s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
8025 		      u8 dev_addr, u8 *data)
8026 {
8027 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
8028 	struct i2c_client *this_client = adapter->i2c_client;
8029 	s32 status;
8030 	u16 swfw_mask = 0;
8031 
8032 	if (!this_client)
8033 		return E1000_ERR_I2C;
8034 
8035 	swfw_mask = E1000_SWFW_PHY0_SM;
8036 
8037 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
8038 		return E1000_ERR_SWFW_SYNC;
8039 
8040 	status = i2c_smbus_read_byte_data(this_client, byte_offset);
8041 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
8042 
8043 	if (status < 0)
8044 		return E1000_ERR_I2C;
8045 	else {
8046 		*data = status;
8047 		return 0;
8048 	}
8049 }
8050 
8051 /**
8052  *  igb_write_i2c_byte - Writes 8 bit word over I2C
8053  *  @hw: pointer to hardware structure
8054  *  @byte_offset: byte offset to write
8055  *  @dev_addr: device address
8056  *  @data: value to write
8057  *
8058  *  Performs byte write operation over I2C interface at
8059  *  a specified device address.
8060  **/
8061 s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
8062 		       u8 dev_addr, u8 data)
8063 {
8064 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
8065 	struct i2c_client *this_client = adapter->i2c_client;
8066 	s32 status;
8067 	u16 swfw_mask = E1000_SWFW_PHY0_SM;
8068 
8069 	if (!this_client)
8070 		return E1000_ERR_I2C;
8071 
8072 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
8073 		return E1000_ERR_SWFW_SYNC;
8074 	status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
8075 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
8076 
8077 	if (status)
8078 		return E1000_ERR_I2C;
8079 	else
8080 		return 0;
8081 
8082 }
8083 
8084 int igb_reinit_queues(struct igb_adapter *adapter)
8085 {
8086 	struct net_device *netdev = adapter->netdev;
8087 	struct pci_dev *pdev = adapter->pdev;
8088 	int err = 0;
8089 
8090 	if (netif_running(netdev))
8091 		igb_close(netdev);
8092 
8093 	igb_reset_interrupt_capability(adapter);
8094 
8095 	if (igb_init_interrupt_scheme(adapter, true)) {
8096 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8097 		return -ENOMEM;
8098 	}
8099 
8100 	if (netif_running(netdev))
8101 		err = igb_open(netdev);
8102 
8103 	return err;
8104 }
8105 /* igb_main.c */
8106