1 // SPDX-License-Identifier: GPL-2.0
2 /* Intel(R) Gigabit Ethernet Linux driver
3  * Copyright(c) 2007-2014 Intel Corporation.
4  *
5  * This program is free software; you can redistribute it and/or modify it
6  * under the terms and conditions of the GNU General Public License,
7  * version 2, as published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope it will be useful, but WITHOUT
10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
12  * more details.
13  *
14  * You should have received a copy of the GNU General Public License along with
15  * this program; if not, see <http://www.gnu.org/licenses/>.
16  *
17  * The full GNU General Public License is included in this distribution in
18  * the file called "COPYING".
19  *
20  * Contact Information:
21  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
22  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
23  */
24 
25 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
26 
27 #include <linux/module.h>
28 #include <linux/types.h>
29 #include <linux/init.h>
30 #include <linux/bitops.h>
31 #include <linux/vmalloc.h>
32 #include <linux/pagemap.h>
33 #include <linux/netdevice.h>
34 #include <linux/ipv6.h>
35 #include <linux/slab.h>
36 #include <net/checksum.h>
37 #include <net/ip6_checksum.h>
38 #include <net/pkt_sched.h>
39 #include <linux/net_tstamp.h>
40 #include <linux/mii.h>
41 #include <linux/ethtool.h>
42 #include <linux/if.h>
43 #include <linux/if_vlan.h>
44 #include <linux/pci.h>
45 #include <linux/pci-aspm.h>
46 #include <linux/delay.h>
47 #include <linux/interrupt.h>
48 #include <linux/ip.h>
49 #include <linux/tcp.h>
50 #include <linux/sctp.h>
51 #include <linux/if_ether.h>
52 #include <linux/aer.h>
53 #include <linux/prefetch.h>
54 #include <linux/pm_runtime.h>
55 #include <linux/etherdevice.h>
56 #ifdef CONFIG_IGB_DCA
57 #include <linux/dca.h>
58 #endif
59 #include <linux/i2c.h>
60 #include "igb.h"
61 
62 #define MAJ 5
63 #define MIN 4
64 #define BUILD 0
65 #define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
66 __stringify(BUILD) "-k"
67 
68 enum queue_mode {
69 	QUEUE_MODE_STRICT_PRIORITY,
70 	QUEUE_MODE_STREAM_RESERVATION,
71 };
72 
73 enum tx_queue_prio {
74 	TX_QUEUE_PRIO_HIGH,
75 	TX_QUEUE_PRIO_LOW,
76 };
77 
78 char igb_driver_name[] = "igb";
79 char igb_driver_version[] = DRV_VERSION;
80 static const char igb_driver_string[] =
81 				"Intel(R) Gigabit Ethernet Network Driver";
82 static const char igb_copyright[] =
83 				"Copyright (c) 2007-2014 Intel Corporation.";
84 
85 static const struct e1000_info *igb_info_tbl[] = {
86 	[board_82575] = &e1000_82575_info,
87 };
88 
89 static const struct pci_device_id igb_pci_tbl[] = {
90 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
91 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_SGMII) },
92 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
93 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I211_COPPER), board_82575 },
94 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER), board_82575 },
95 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_FIBER), board_82575 },
96 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES), board_82575 },
97 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SGMII), board_82575 },
98 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER_FLASHLESS), board_82575 },
99 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES_FLASHLESS), board_82575 },
100 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_COPPER), board_82575 },
101 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_FIBER), board_82575 },
102 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SERDES), board_82575 },
103 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SGMII), board_82575 },
104 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER), board_82575 },
105 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_FIBER), board_82575 },
106 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_QUAD_FIBER), board_82575 },
107 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SERDES), board_82575 },
108 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SGMII), board_82575 },
109 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER_DUAL), board_82575 },
110 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SGMII), board_82575 },
111 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SERDES), board_82575 },
112 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_BACKPLANE), board_82575 },
113 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SFP), board_82575 },
114 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
115 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
116 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
117 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
118 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
119 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
120 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER_ET2), board_82575 },
121 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
122 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
123 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
124 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
125 	/* required last entry */
126 	{0, }
127 };
128 
129 MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
130 
131 static int igb_setup_all_tx_resources(struct igb_adapter *);
132 static int igb_setup_all_rx_resources(struct igb_adapter *);
133 static void igb_free_all_tx_resources(struct igb_adapter *);
134 static void igb_free_all_rx_resources(struct igb_adapter *);
135 static void igb_setup_mrqc(struct igb_adapter *);
136 static int igb_probe(struct pci_dev *, const struct pci_device_id *);
137 static void igb_remove(struct pci_dev *pdev);
138 static int igb_sw_init(struct igb_adapter *);
139 int igb_open(struct net_device *);
140 int igb_close(struct net_device *);
141 static void igb_configure(struct igb_adapter *);
142 static void igb_configure_tx(struct igb_adapter *);
143 static void igb_configure_rx(struct igb_adapter *);
144 static void igb_clean_all_tx_rings(struct igb_adapter *);
145 static void igb_clean_all_rx_rings(struct igb_adapter *);
146 static void igb_clean_tx_ring(struct igb_ring *);
147 static void igb_clean_rx_ring(struct igb_ring *);
148 static void igb_set_rx_mode(struct net_device *);
149 static void igb_update_phy_info(struct timer_list *);
150 static void igb_watchdog(struct timer_list *);
151 static void igb_watchdog_task(struct work_struct *);
152 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb, struct net_device *);
153 static void igb_get_stats64(struct net_device *dev,
154 			    struct rtnl_link_stats64 *stats);
155 static int igb_change_mtu(struct net_device *, int);
156 static int igb_set_mac(struct net_device *, void *);
157 static void igb_set_uta(struct igb_adapter *adapter, bool set);
158 static irqreturn_t igb_intr(int irq, void *);
159 static irqreturn_t igb_intr_msi(int irq, void *);
160 static irqreturn_t igb_msix_other(int irq, void *);
161 static irqreturn_t igb_msix_ring(int irq, void *);
162 #ifdef CONFIG_IGB_DCA
163 static void igb_update_dca(struct igb_q_vector *);
164 static void igb_setup_dca(struct igb_adapter *);
165 #endif /* CONFIG_IGB_DCA */
166 static int igb_poll(struct napi_struct *, int);
167 static bool igb_clean_tx_irq(struct igb_q_vector *, int);
168 static int igb_clean_rx_irq(struct igb_q_vector *, int);
169 static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
170 static void igb_tx_timeout(struct net_device *);
171 static void igb_reset_task(struct work_struct *);
172 static void igb_vlan_mode(struct net_device *netdev,
173 			  netdev_features_t features);
174 static int igb_vlan_rx_add_vid(struct net_device *, __be16, u16);
175 static int igb_vlan_rx_kill_vid(struct net_device *, __be16, u16);
176 static void igb_restore_vlan(struct igb_adapter *);
177 static void igb_rar_set_index(struct igb_adapter *, u32);
178 static void igb_ping_all_vfs(struct igb_adapter *);
179 static void igb_msg_task(struct igb_adapter *);
180 static void igb_vmm_control(struct igb_adapter *);
181 static int igb_set_vf_mac(struct igb_adapter *, int, unsigned char *);
182 static void igb_flush_mac_table(struct igb_adapter *);
183 static int igb_available_rars(struct igb_adapter *, u8);
184 static void igb_set_default_mac_filter(struct igb_adapter *);
185 static int igb_uc_sync(struct net_device *, const unsigned char *);
186 static int igb_uc_unsync(struct net_device *, const unsigned char *);
187 static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
188 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac);
189 static int igb_ndo_set_vf_vlan(struct net_device *netdev,
190 			       int vf, u16 vlan, u8 qos, __be16 vlan_proto);
191 static int igb_ndo_set_vf_bw(struct net_device *, int, int, int);
192 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
193 				   bool setting);
194 static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf,
195 				bool setting);
196 static int igb_ndo_get_vf_config(struct net_device *netdev, int vf,
197 				 struct ifla_vf_info *ivi);
198 static void igb_check_vf_rate_limit(struct igb_adapter *);
199 static void igb_nfc_filter_exit(struct igb_adapter *adapter);
200 static void igb_nfc_filter_restore(struct igb_adapter *adapter);
201 
202 #ifdef CONFIG_PCI_IOV
203 static int igb_vf_configure(struct igb_adapter *adapter, int vf);
204 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs);
205 static int igb_disable_sriov(struct pci_dev *dev);
206 static int igb_pci_disable_sriov(struct pci_dev *dev);
207 #endif
208 
209 static int igb_suspend(struct device *);
210 static int igb_resume(struct device *);
211 static int igb_runtime_suspend(struct device *dev);
212 static int igb_runtime_resume(struct device *dev);
213 static int igb_runtime_idle(struct device *dev);
214 static const struct dev_pm_ops igb_pm_ops = {
215 	SET_SYSTEM_SLEEP_PM_OPS(igb_suspend, igb_resume)
216 	SET_RUNTIME_PM_OPS(igb_runtime_suspend, igb_runtime_resume,
217 			igb_runtime_idle)
218 };
219 static void igb_shutdown(struct pci_dev *);
220 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs);
221 #ifdef CONFIG_IGB_DCA
222 static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
223 static struct notifier_block dca_notifier = {
224 	.notifier_call	= igb_notify_dca,
225 	.next		= NULL,
226 	.priority	= 0
227 };
228 #endif
229 #ifdef CONFIG_NET_POLL_CONTROLLER
230 /* for netdump / net console */
231 static void igb_netpoll(struct net_device *);
232 #endif
233 #ifdef CONFIG_PCI_IOV
234 static unsigned int max_vfs;
235 module_param(max_vfs, uint, 0);
236 MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate per physical function");
237 #endif /* CONFIG_PCI_IOV */
238 
239 static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
240 		     pci_channel_state_t);
241 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
242 static void igb_io_resume(struct pci_dev *);
243 
244 static const struct pci_error_handlers igb_err_handler = {
245 	.error_detected = igb_io_error_detected,
246 	.slot_reset = igb_io_slot_reset,
247 	.resume = igb_io_resume,
248 };
249 
250 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba);
251 
252 static struct pci_driver igb_driver = {
253 	.name     = igb_driver_name,
254 	.id_table = igb_pci_tbl,
255 	.probe    = igb_probe,
256 	.remove   = igb_remove,
257 #ifdef CONFIG_PM
258 	.driver.pm = &igb_pm_ops,
259 #endif
260 	.shutdown = igb_shutdown,
261 	.sriov_configure = igb_pci_sriov_configure,
262 	.err_handler = &igb_err_handler
263 };
264 
265 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
266 MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
267 MODULE_LICENSE("GPL");
268 MODULE_VERSION(DRV_VERSION);
269 
270 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
271 static int debug = -1;
272 module_param(debug, int, 0);
273 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
274 
275 struct igb_reg_info {
276 	u32 ofs;
277 	char *name;
278 };
279 
280 static const struct igb_reg_info igb_reg_info_tbl[] = {
281 
282 	/* General Registers */
283 	{E1000_CTRL, "CTRL"},
284 	{E1000_STATUS, "STATUS"},
285 	{E1000_CTRL_EXT, "CTRL_EXT"},
286 
287 	/* Interrupt Registers */
288 	{E1000_ICR, "ICR"},
289 
290 	/* RX Registers */
291 	{E1000_RCTL, "RCTL"},
292 	{E1000_RDLEN(0), "RDLEN"},
293 	{E1000_RDH(0), "RDH"},
294 	{E1000_RDT(0), "RDT"},
295 	{E1000_RXDCTL(0), "RXDCTL"},
296 	{E1000_RDBAL(0), "RDBAL"},
297 	{E1000_RDBAH(0), "RDBAH"},
298 
299 	/* TX Registers */
300 	{E1000_TCTL, "TCTL"},
301 	{E1000_TDBAL(0), "TDBAL"},
302 	{E1000_TDBAH(0), "TDBAH"},
303 	{E1000_TDLEN(0), "TDLEN"},
304 	{E1000_TDH(0), "TDH"},
305 	{E1000_TDT(0), "TDT"},
306 	{E1000_TXDCTL(0), "TXDCTL"},
307 	{E1000_TDFH, "TDFH"},
308 	{E1000_TDFT, "TDFT"},
309 	{E1000_TDFHS, "TDFHS"},
310 	{E1000_TDFPC, "TDFPC"},
311 
312 	/* List Terminator */
313 	{}
314 };
315 
316 /* igb_regdump - register printout routine */
317 static void igb_regdump(struct e1000_hw *hw, struct igb_reg_info *reginfo)
318 {
319 	int n = 0;
320 	char rname[16];
321 	u32 regs[8];
322 
323 	switch (reginfo->ofs) {
324 	case E1000_RDLEN(0):
325 		for (n = 0; n < 4; n++)
326 			regs[n] = rd32(E1000_RDLEN(n));
327 		break;
328 	case E1000_RDH(0):
329 		for (n = 0; n < 4; n++)
330 			regs[n] = rd32(E1000_RDH(n));
331 		break;
332 	case E1000_RDT(0):
333 		for (n = 0; n < 4; n++)
334 			regs[n] = rd32(E1000_RDT(n));
335 		break;
336 	case E1000_RXDCTL(0):
337 		for (n = 0; n < 4; n++)
338 			regs[n] = rd32(E1000_RXDCTL(n));
339 		break;
340 	case E1000_RDBAL(0):
341 		for (n = 0; n < 4; n++)
342 			regs[n] = rd32(E1000_RDBAL(n));
343 		break;
344 	case E1000_RDBAH(0):
345 		for (n = 0; n < 4; n++)
346 			regs[n] = rd32(E1000_RDBAH(n));
347 		break;
348 	case E1000_TDBAL(0):
349 		for (n = 0; n < 4; n++)
350 			regs[n] = rd32(E1000_RDBAL(n));
351 		break;
352 	case E1000_TDBAH(0):
353 		for (n = 0; n < 4; n++)
354 			regs[n] = rd32(E1000_TDBAH(n));
355 		break;
356 	case E1000_TDLEN(0):
357 		for (n = 0; n < 4; n++)
358 			regs[n] = rd32(E1000_TDLEN(n));
359 		break;
360 	case E1000_TDH(0):
361 		for (n = 0; n < 4; n++)
362 			regs[n] = rd32(E1000_TDH(n));
363 		break;
364 	case E1000_TDT(0):
365 		for (n = 0; n < 4; n++)
366 			regs[n] = rd32(E1000_TDT(n));
367 		break;
368 	case E1000_TXDCTL(0):
369 		for (n = 0; n < 4; n++)
370 			regs[n] = rd32(E1000_TXDCTL(n));
371 		break;
372 	default:
373 		pr_info("%-15s %08x\n", reginfo->name, rd32(reginfo->ofs));
374 		return;
375 	}
376 
377 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-3]");
378 	pr_info("%-15s %08x %08x %08x %08x\n", rname, regs[0], regs[1],
379 		regs[2], regs[3]);
380 }
381 
382 /* igb_dump - Print registers, Tx-rings and Rx-rings */
383 static void igb_dump(struct igb_adapter *adapter)
384 {
385 	struct net_device *netdev = adapter->netdev;
386 	struct e1000_hw *hw = &adapter->hw;
387 	struct igb_reg_info *reginfo;
388 	struct igb_ring *tx_ring;
389 	union e1000_adv_tx_desc *tx_desc;
390 	struct my_u0 { u64 a; u64 b; } *u0;
391 	struct igb_ring *rx_ring;
392 	union e1000_adv_rx_desc *rx_desc;
393 	u32 staterr;
394 	u16 i, n;
395 
396 	if (!netif_msg_hw(adapter))
397 		return;
398 
399 	/* Print netdevice Info */
400 	if (netdev) {
401 		dev_info(&adapter->pdev->dev, "Net device Info\n");
402 		pr_info("Device Name     state            trans_start\n");
403 		pr_info("%-15s %016lX %016lX\n", netdev->name,
404 			netdev->state, dev_trans_start(netdev));
405 	}
406 
407 	/* Print Registers */
408 	dev_info(&adapter->pdev->dev, "Register Dump\n");
409 	pr_info(" Register Name   Value\n");
410 	for (reginfo = (struct igb_reg_info *)igb_reg_info_tbl;
411 	     reginfo->name; reginfo++) {
412 		igb_regdump(hw, reginfo);
413 	}
414 
415 	/* Print TX Ring Summary */
416 	if (!netdev || !netif_running(netdev))
417 		goto exit;
418 
419 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
420 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
421 	for (n = 0; n < adapter->num_tx_queues; n++) {
422 		struct igb_tx_buffer *buffer_info;
423 		tx_ring = adapter->tx_ring[n];
424 		buffer_info = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
425 		pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
426 			n, tx_ring->next_to_use, tx_ring->next_to_clean,
427 			(u64)dma_unmap_addr(buffer_info, dma),
428 			dma_unmap_len(buffer_info, len),
429 			buffer_info->next_to_watch,
430 			(u64)buffer_info->time_stamp);
431 	}
432 
433 	/* Print TX Rings */
434 	if (!netif_msg_tx_done(adapter))
435 		goto rx_ring_summary;
436 
437 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
438 
439 	/* Transmit Descriptor Formats
440 	 *
441 	 * Advanced Transmit Descriptor
442 	 *   +--------------------------------------------------------------+
443 	 * 0 |         Buffer Address [63:0]                                |
444 	 *   +--------------------------------------------------------------+
445 	 * 8 | PAYLEN  | PORTS  |CC|IDX | STA | DCMD  |DTYP|MAC|RSV| DTALEN |
446 	 *   +--------------------------------------------------------------+
447 	 *   63      46 45    40 39 38 36 35 32 31   24             15       0
448 	 */
449 
450 	for (n = 0; n < adapter->num_tx_queues; n++) {
451 		tx_ring = adapter->tx_ring[n];
452 		pr_info("------------------------------------\n");
453 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
454 		pr_info("------------------------------------\n");
455 		pr_info("T [desc]     [address 63:0  ] [PlPOCIStDDM Ln] [bi->dma       ] leng  ntw timestamp        bi->skb\n");
456 
457 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
458 			const char *next_desc;
459 			struct igb_tx_buffer *buffer_info;
460 			tx_desc = IGB_TX_DESC(tx_ring, i);
461 			buffer_info = &tx_ring->tx_buffer_info[i];
462 			u0 = (struct my_u0 *)tx_desc;
463 			if (i == tx_ring->next_to_use &&
464 			    i == tx_ring->next_to_clean)
465 				next_desc = " NTC/U";
466 			else if (i == tx_ring->next_to_use)
467 				next_desc = " NTU";
468 			else if (i == tx_ring->next_to_clean)
469 				next_desc = " NTC";
470 			else
471 				next_desc = "";
472 
473 			pr_info("T [0x%03X]    %016llX %016llX %016llX %04X  %p %016llX %p%s\n",
474 				i, le64_to_cpu(u0->a),
475 				le64_to_cpu(u0->b),
476 				(u64)dma_unmap_addr(buffer_info, dma),
477 				dma_unmap_len(buffer_info, len),
478 				buffer_info->next_to_watch,
479 				(u64)buffer_info->time_stamp,
480 				buffer_info->skb, next_desc);
481 
482 			if (netif_msg_pktdata(adapter) && buffer_info->skb)
483 				print_hex_dump(KERN_INFO, "",
484 					DUMP_PREFIX_ADDRESS,
485 					16, 1, buffer_info->skb->data,
486 					dma_unmap_len(buffer_info, len),
487 					true);
488 		}
489 	}
490 
491 	/* Print RX Rings Summary */
492 rx_ring_summary:
493 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
494 	pr_info("Queue [NTU] [NTC]\n");
495 	for (n = 0; n < adapter->num_rx_queues; n++) {
496 		rx_ring = adapter->rx_ring[n];
497 		pr_info(" %5d %5X %5X\n",
498 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
499 	}
500 
501 	/* Print RX Rings */
502 	if (!netif_msg_rx_status(adapter))
503 		goto exit;
504 
505 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
506 
507 	/* Advanced Receive Descriptor (Read) Format
508 	 *    63                                           1        0
509 	 *    +-----------------------------------------------------+
510 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
511 	 *    +----------------------------------------------+------+
512 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
513 	 *    +-----------------------------------------------------+
514 	 *
515 	 *
516 	 * Advanced Receive Descriptor (Write-Back) Format
517 	 *
518 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
519 	 *   +------------------------------------------------------+
520 	 * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
521 	 *   | Checksum   Ident  |   |           |    | Type | Type |
522 	 *   +------------------------------------------------------+
523 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
524 	 *   +------------------------------------------------------+
525 	 *   63       48 47    32 31            20 19               0
526 	 */
527 
528 	for (n = 0; n < adapter->num_rx_queues; n++) {
529 		rx_ring = adapter->rx_ring[n];
530 		pr_info("------------------------------------\n");
531 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
532 		pr_info("------------------------------------\n");
533 		pr_info("R  [desc]      [ PktBuf     A0] [  HeadBuf   DD] [bi->dma       ] [bi->skb] <-- Adv Rx Read format\n");
534 		pr_info("RWB[desc]      [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
535 
536 		for (i = 0; i < rx_ring->count; i++) {
537 			const char *next_desc;
538 			struct igb_rx_buffer *buffer_info;
539 			buffer_info = &rx_ring->rx_buffer_info[i];
540 			rx_desc = IGB_RX_DESC(rx_ring, i);
541 			u0 = (struct my_u0 *)rx_desc;
542 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
543 
544 			if (i == rx_ring->next_to_use)
545 				next_desc = " NTU";
546 			else if (i == rx_ring->next_to_clean)
547 				next_desc = " NTC";
548 			else
549 				next_desc = "";
550 
551 			if (staterr & E1000_RXD_STAT_DD) {
552 				/* Descriptor Done */
553 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %s\n",
554 					"RWB", i,
555 					le64_to_cpu(u0->a),
556 					le64_to_cpu(u0->b),
557 					next_desc);
558 			} else {
559 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %s\n",
560 					"R  ", i,
561 					le64_to_cpu(u0->a),
562 					le64_to_cpu(u0->b),
563 					(u64)buffer_info->dma,
564 					next_desc);
565 
566 				if (netif_msg_pktdata(adapter) &&
567 				    buffer_info->dma && buffer_info->page) {
568 					print_hex_dump(KERN_INFO, "",
569 					  DUMP_PREFIX_ADDRESS,
570 					  16, 1,
571 					  page_address(buffer_info->page) +
572 						      buffer_info->page_offset,
573 					  igb_rx_bufsz(rx_ring), true);
574 				}
575 			}
576 		}
577 	}
578 
579 exit:
580 	return;
581 }
582 
583 /**
584  *  igb_get_i2c_data - Reads the I2C SDA data bit
585  *  @hw: pointer to hardware structure
586  *  @i2cctl: Current value of I2CCTL register
587  *
588  *  Returns the I2C data bit value
589  **/
590 static int igb_get_i2c_data(void *data)
591 {
592 	struct igb_adapter *adapter = (struct igb_adapter *)data;
593 	struct e1000_hw *hw = &adapter->hw;
594 	s32 i2cctl = rd32(E1000_I2CPARAMS);
595 
596 	return !!(i2cctl & E1000_I2C_DATA_IN);
597 }
598 
599 /**
600  *  igb_set_i2c_data - Sets the I2C data bit
601  *  @data: pointer to hardware structure
602  *  @state: I2C data value (0 or 1) to set
603  *
604  *  Sets the I2C data bit
605  **/
606 static void igb_set_i2c_data(void *data, int state)
607 {
608 	struct igb_adapter *adapter = (struct igb_adapter *)data;
609 	struct e1000_hw *hw = &adapter->hw;
610 	s32 i2cctl = rd32(E1000_I2CPARAMS);
611 
612 	if (state)
613 		i2cctl |= E1000_I2C_DATA_OUT;
614 	else
615 		i2cctl &= ~E1000_I2C_DATA_OUT;
616 
617 	i2cctl &= ~E1000_I2C_DATA_OE_N;
618 	i2cctl |= E1000_I2C_CLK_OE_N;
619 	wr32(E1000_I2CPARAMS, i2cctl);
620 	wrfl();
621 
622 }
623 
624 /**
625  *  igb_set_i2c_clk - Sets the I2C SCL clock
626  *  @data: pointer to hardware structure
627  *  @state: state to set clock
628  *
629  *  Sets the I2C clock line to state
630  **/
631 static void igb_set_i2c_clk(void *data, int state)
632 {
633 	struct igb_adapter *adapter = (struct igb_adapter *)data;
634 	struct e1000_hw *hw = &adapter->hw;
635 	s32 i2cctl = rd32(E1000_I2CPARAMS);
636 
637 	if (state) {
638 		i2cctl |= E1000_I2C_CLK_OUT;
639 		i2cctl &= ~E1000_I2C_CLK_OE_N;
640 	} else {
641 		i2cctl &= ~E1000_I2C_CLK_OUT;
642 		i2cctl &= ~E1000_I2C_CLK_OE_N;
643 	}
644 	wr32(E1000_I2CPARAMS, i2cctl);
645 	wrfl();
646 }
647 
648 /**
649  *  igb_get_i2c_clk - Gets the I2C SCL clock state
650  *  @data: pointer to hardware structure
651  *
652  *  Gets the I2C clock state
653  **/
654 static int igb_get_i2c_clk(void *data)
655 {
656 	struct igb_adapter *adapter = (struct igb_adapter *)data;
657 	struct e1000_hw *hw = &adapter->hw;
658 	s32 i2cctl = rd32(E1000_I2CPARAMS);
659 
660 	return !!(i2cctl & E1000_I2C_CLK_IN);
661 }
662 
663 static const struct i2c_algo_bit_data igb_i2c_algo = {
664 	.setsda		= igb_set_i2c_data,
665 	.setscl		= igb_set_i2c_clk,
666 	.getsda		= igb_get_i2c_data,
667 	.getscl		= igb_get_i2c_clk,
668 	.udelay		= 5,
669 	.timeout	= 20,
670 };
671 
672 /**
673  *  igb_get_hw_dev - return device
674  *  @hw: pointer to hardware structure
675  *
676  *  used by hardware layer to print debugging information
677  **/
678 struct net_device *igb_get_hw_dev(struct e1000_hw *hw)
679 {
680 	struct igb_adapter *adapter = hw->back;
681 	return adapter->netdev;
682 }
683 
684 /**
685  *  igb_init_module - Driver Registration Routine
686  *
687  *  igb_init_module is the first routine called when the driver is
688  *  loaded. All it does is register with the PCI subsystem.
689  **/
690 static int __init igb_init_module(void)
691 {
692 	int ret;
693 
694 	pr_info("%s - version %s\n",
695 	       igb_driver_string, igb_driver_version);
696 	pr_info("%s\n", igb_copyright);
697 
698 #ifdef CONFIG_IGB_DCA
699 	dca_register_notify(&dca_notifier);
700 #endif
701 	ret = pci_register_driver(&igb_driver);
702 	return ret;
703 }
704 
705 module_init(igb_init_module);
706 
707 /**
708  *  igb_exit_module - Driver Exit Cleanup Routine
709  *
710  *  igb_exit_module is called just before the driver is removed
711  *  from memory.
712  **/
713 static void __exit igb_exit_module(void)
714 {
715 #ifdef CONFIG_IGB_DCA
716 	dca_unregister_notify(&dca_notifier);
717 #endif
718 	pci_unregister_driver(&igb_driver);
719 }
720 
721 module_exit(igb_exit_module);
722 
723 #define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
724 /**
725  *  igb_cache_ring_register - Descriptor ring to register mapping
726  *  @adapter: board private structure to initialize
727  *
728  *  Once we know the feature-set enabled for the device, we'll cache
729  *  the register offset the descriptor ring is assigned to.
730  **/
731 static void igb_cache_ring_register(struct igb_adapter *adapter)
732 {
733 	int i = 0, j = 0;
734 	u32 rbase_offset = adapter->vfs_allocated_count;
735 
736 	switch (adapter->hw.mac.type) {
737 	case e1000_82576:
738 		/* The queues are allocated for virtualization such that VF 0
739 		 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
740 		 * In order to avoid collision we start at the first free queue
741 		 * and continue consuming queues in the same sequence
742 		 */
743 		if (adapter->vfs_allocated_count) {
744 			for (; i < adapter->rss_queues; i++)
745 				adapter->rx_ring[i]->reg_idx = rbase_offset +
746 							       Q_IDX_82576(i);
747 		}
748 		/* Fall through */
749 	case e1000_82575:
750 	case e1000_82580:
751 	case e1000_i350:
752 	case e1000_i354:
753 	case e1000_i210:
754 	case e1000_i211:
755 		/* Fall through */
756 	default:
757 		for (; i < adapter->num_rx_queues; i++)
758 			adapter->rx_ring[i]->reg_idx = rbase_offset + i;
759 		for (; j < adapter->num_tx_queues; j++)
760 			adapter->tx_ring[j]->reg_idx = rbase_offset + j;
761 		break;
762 	}
763 }
764 
765 u32 igb_rd32(struct e1000_hw *hw, u32 reg)
766 {
767 	struct igb_adapter *igb = container_of(hw, struct igb_adapter, hw);
768 	u8 __iomem *hw_addr = READ_ONCE(hw->hw_addr);
769 	u32 value = 0;
770 
771 	if (E1000_REMOVED(hw_addr))
772 		return ~value;
773 
774 	value = readl(&hw_addr[reg]);
775 
776 	/* reads should not return all F's */
777 	if (!(~value) && (!reg || !(~readl(hw_addr)))) {
778 		struct net_device *netdev = igb->netdev;
779 		hw->hw_addr = NULL;
780 		netdev_err(netdev, "PCIe link lost\n");
781 	}
782 
783 	return value;
784 }
785 
786 /**
787  *  igb_write_ivar - configure ivar for given MSI-X vector
788  *  @hw: pointer to the HW structure
789  *  @msix_vector: vector number we are allocating to a given ring
790  *  @index: row index of IVAR register to write within IVAR table
791  *  @offset: column offset of in IVAR, should be multiple of 8
792  *
793  *  This function is intended to handle the writing of the IVAR register
794  *  for adapters 82576 and newer.  The IVAR table consists of 2 columns,
795  *  each containing an cause allocation for an Rx and Tx ring, and a
796  *  variable number of rows depending on the number of queues supported.
797  **/
798 static void igb_write_ivar(struct e1000_hw *hw, int msix_vector,
799 			   int index, int offset)
800 {
801 	u32 ivar = array_rd32(E1000_IVAR0, index);
802 
803 	/* clear any bits that are currently set */
804 	ivar &= ~((u32)0xFF << offset);
805 
806 	/* write vector and valid bit */
807 	ivar |= (msix_vector | E1000_IVAR_VALID) << offset;
808 
809 	array_wr32(E1000_IVAR0, index, ivar);
810 }
811 
812 #define IGB_N0_QUEUE -1
813 static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
814 {
815 	struct igb_adapter *adapter = q_vector->adapter;
816 	struct e1000_hw *hw = &adapter->hw;
817 	int rx_queue = IGB_N0_QUEUE;
818 	int tx_queue = IGB_N0_QUEUE;
819 	u32 msixbm = 0;
820 
821 	if (q_vector->rx.ring)
822 		rx_queue = q_vector->rx.ring->reg_idx;
823 	if (q_vector->tx.ring)
824 		tx_queue = q_vector->tx.ring->reg_idx;
825 
826 	switch (hw->mac.type) {
827 	case e1000_82575:
828 		/* The 82575 assigns vectors using a bitmask, which matches the
829 		 * bitmask for the EICR/EIMS/EIMC registers.  To assign one
830 		 * or more queues to a vector, we write the appropriate bits
831 		 * into the MSIXBM register for that vector.
832 		 */
833 		if (rx_queue > IGB_N0_QUEUE)
834 			msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
835 		if (tx_queue > IGB_N0_QUEUE)
836 			msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
837 		if (!(adapter->flags & IGB_FLAG_HAS_MSIX) && msix_vector == 0)
838 			msixbm |= E1000_EIMS_OTHER;
839 		array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
840 		q_vector->eims_value = msixbm;
841 		break;
842 	case e1000_82576:
843 		/* 82576 uses a table that essentially consists of 2 columns
844 		 * with 8 rows.  The ordering is column-major so we use the
845 		 * lower 3 bits as the row index, and the 4th bit as the
846 		 * column offset.
847 		 */
848 		if (rx_queue > IGB_N0_QUEUE)
849 			igb_write_ivar(hw, msix_vector,
850 				       rx_queue & 0x7,
851 				       (rx_queue & 0x8) << 1);
852 		if (tx_queue > IGB_N0_QUEUE)
853 			igb_write_ivar(hw, msix_vector,
854 				       tx_queue & 0x7,
855 				       ((tx_queue & 0x8) << 1) + 8);
856 		q_vector->eims_value = BIT(msix_vector);
857 		break;
858 	case e1000_82580:
859 	case e1000_i350:
860 	case e1000_i354:
861 	case e1000_i210:
862 	case e1000_i211:
863 		/* On 82580 and newer adapters the scheme is similar to 82576
864 		 * however instead of ordering column-major we have things
865 		 * ordered row-major.  So we traverse the table by using
866 		 * bit 0 as the column offset, and the remaining bits as the
867 		 * row index.
868 		 */
869 		if (rx_queue > IGB_N0_QUEUE)
870 			igb_write_ivar(hw, msix_vector,
871 				       rx_queue >> 1,
872 				       (rx_queue & 0x1) << 4);
873 		if (tx_queue > IGB_N0_QUEUE)
874 			igb_write_ivar(hw, msix_vector,
875 				       tx_queue >> 1,
876 				       ((tx_queue & 0x1) << 4) + 8);
877 		q_vector->eims_value = BIT(msix_vector);
878 		break;
879 	default:
880 		BUG();
881 		break;
882 	}
883 
884 	/* add q_vector eims value to global eims_enable_mask */
885 	adapter->eims_enable_mask |= q_vector->eims_value;
886 
887 	/* configure q_vector to set itr on first interrupt */
888 	q_vector->set_itr = 1;
889 }
890 
891 /**
892  *  igb_configure_msix - Configure MSI-X hardware
893  *  @adapter: board private structure to initialize
894  *
895  *  igb_configure_msix sets up the hardware to properly
896  *  generate MSI-X interrupts.
897  **/
898 static void igb_configure_msix(struct igb_adapter *adapter)
899 {
900 	u32 tmp;
901 	int i, vector = 0;
902 	struct e1000_hw *hw = &adapter->hw;
903 
904 	adapter->eims_enable_mask = 0;
905 
906 	/* set vector for other causes, i.e. link changes */
907 	switch (hw->mac.type) {
908 	case e1000_82575:
909 		tmp = rd32(E1000_CTRL_EXT);
910 		/* enable MSI-X PBA support*/
911 		tmp |= E1000_CTRL_EXT_PBA_CLR;
912 
913 		/* Auto-Mask interrupts upon ICR read. */
914 		tmp |= E1000_CTRL_EXT_EIAME;
915 		tmp |= E1000_CTRL_EXT_IRCA;
916 
917 		wr32(E1000_CTRL_EXT, tmp);
918 
919 		/* enable msix_other interrupt */
920 		array_wr32(E1000_MSIXBM(0), vector++, E1000_EIMS_OTHER);
921 		adapter->eims_other = E1000_EIMS_OTHER;
922 
923 		break;
924 
925 	case e1000_82576:
926 	case e1000_82580:
927 	case e1000_i350:
928 	case e1000_i354:
929 	case e1000_i210:
930 	case e1000_i211:
931 		/* Turn on MSI-X capability first, or our settings
932 		 * won't stick.  And it will take days to debug.
933 		 */
934 		wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
935 		     E1000_GPIE_PBA | E1000_GPIE_EIAME |
936 		     E1000_GPIE_NSICR);
937 
938 		/* enable msix_other interrupt */
939 		adapter->eims_other = BIT(vector);
940 		tmp = (vector++ | E1000_IVAR_VALID) << 8;
941 
942 		wr32(E1000_IVAR_MISC, tmp);
943 		break;
944 	default:
945 		/* do nothing, since nothing else supports MSI-X */
946 		break;
947 	} /* switch (hw->mac.type) */
948 
949 	adapter->eims_enable_mask |= adapter->eims_other;
950 
951 	for (i = 0; i < adapter->num_q_vectors; i++)
952 		igb_assign_vector(adapter->q_vector[i], vector++);
953 
954 	wrfl();
955 }
956 
957 /**
958  *  igb_request_msix - Initialize MSI-X interrupts
959  *  @adapter: board private structure to initialize
960  *
961  *  igb_request_msix allocates MSI-X vectors and requests interrupts from the
962  *  kernel.
963  **/
964 static int igb_request_msix(struct igb_adapter *adapter)
965 {
966 	struct net_device *netdev = adapter->netdev;
967 	int i, err = 0, vector = 0, free_vector = 0;
968 
969 	err = request_irq(adapter->msix_entries[vector].vector,
970 			  igb_msix_other, 0, netdev->name, adapter);
971 	if (err)
972 		goto err_out;
973 
974 	for (i = 0; i < adapter->num_q_vectors; i++) {
975 		struct igb_q_vector *q_vector = adapter->q_vector[i];
976 
977 		vector++;
978 
979 		q_vector->itr_register = adapter->io_addr + E1000_EITR(vector);
980 
981 		if (q_vector->rx.ring && q_vector->tx.ring)
982 			sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
983 				q_vector->rx.ring->queue_index);
984 		else if (q_vector->tx.ring)
985 			sprintf(q_vector->name, "%s-tx-%u", netdev->name,
986 				q_vector->tx.ring->queue_index);
987 		else if (q_vector->rx.ring)
988 			sprintf(q_vector->name, "%s-rx-%u", netdev->name,
989 				q_vector->rx.ring->queue_index);
990 		else
991 			sprintf(q_vector->name, "%s-unused", netdev->name);
992 
993 		err = request_irq(adapter->msix_entries[vector].vector,
994 				  igb_msix_ring, 0, q_vector->name,
995 				  q_vector);
996 		if (err)
997 			goto err_free;
998 	}
999 
1000 	igb_configure_msix(adapter);
1001 	return 0;
1002 
1003 err_free:
1004 	/* free already assigned IRQs */
1005 	free_irq(adapter->msix_entries[free_vector++].vector, adapter);
1006 
1007 	vector--;
1008 	for (i = 0; i < vector; i++) {
1009 		free_irq(adapter->msix_entries[free_vector++].vector,
1010 			 adapter->q_vector[i]);
1011 	}
1012 err_out:
1013 	return err;
1014 }
1015 
1016 /**
1017  *  igb_free_q_vector - Free memory allocated for specific interrupt vector
1018  *  @adapter: board private structure to initialize
1019  *  @v_idx: Index of vector to be freed
1020  *
1021  *  This function frees the memory allocated to the q_vector.
1022  **/
1023 static void igb_free_q_vector(struct igb_adapter *adapter, int v_idx)
1024 {
1025 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1026 
1027 	adapter->q_vector[v_idx] = NULL;
1028 
1029 	/* igb_get_stats64() might access the rings on this vector,
1030 	 * we must wait a grace period before freeing it.
1031 	 */
1032 	if (q_vector)
1033 		kfree_rcu(q_vector, rcu);
1034 }
1035 
1036 /**
1037  *  igb_reset_q_vector - Reset config for interrupt vector
1038  *  @adapter: board private structure to initialize
1039  *  @v_idx: Index of vector to be reset
1040  *
1041  *  If NAPI is enabled it will delete any references to the
1042  *  NAPI struct. This is preparation for igb_free_q_vector.
1043  **/
1044 static void igb_reset_q_vector(struct igb_adapter *adapter, int v_idx)
1045 {
1046 	struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1047 
1048 	/* Coming from igb_set_interrupt_capability, the vectors are not yet
1049 	 * allocated. So, q_vector is NULL so we should stop here.
1050 	 */
1051 	if (!q_vector)
1052 		return;
1053 
1054 	if (q_vector->tx.ring)
1055 		adapter->tx_ring[q_vector->tx.ring->queue_index] = NULL;
1056 
1057 	if (q_vector->rx.ring)
1058 		adapter->rx_ring[q_vector->rx.ring->queue_index] = NULL;
1059 
1060 	netif_napi_del(&q_vector->napi);
1061 
1062 }
1063 
1064 static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
1065 {
1066 	int v_idx = adapter->num_q_vectors;
1067 
1068 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
1069 		pci_disable_msix(adapter->pdev);
1070 	else if (adapter->flags & IGB_FLAG_HAS_MSI)
1071 		pci_disable_msi(adapter->pdev);
1072 
1073 	while (v_idx--)
1074 		igb_reset_q_vector(adapter, v_idx);
1075 }
1076 
1077 /**
1078  *  igb_free_q_vectors - Free memory allocated for interrupt vectors
1079  *  @adapter: board private structure to initialize
1080  *
1081  *  This function frees the memory allocated to the q_vectors.  In addition if
1082  *  NAPI is enabled it will delete any references to the NAPI struct prior
1083  *  to freeing the q_vector.
1084  **/
1085 static void igb_free_q_vectors(struct igb_adapter *adapter)
1086 {
1087 	int v_idx = adapter->num_q_vectors;
1088 
1089 	adapter->num_tx_queues = 0;
1090 	adapter->num_rx_queues = 0;
1091 	adapter->num_q_vectors = 0;
1092 
1093 	while (v_idx--) {
1094 		igb_reset_q_vector(adapter, v_idx);
1095 		igb_free_q_vector(adapter, v_idx);
1096 	}
1097 }
1098 
1099 /**
1100  *  igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1101  *  @adapter: board private structure to initialize
1102  *
1103  *  This function resets the device so that it has 0 Rx queues, Tx queues, and
1104  *  MSI-X interrupts allocated.
1105  */
1106 static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
1107 {
1108 	igb_free_q_vectors(adapter);
1109 	igb_reset_interrupt_capability(adapter);
1110 }
1111 
1112 /**
1113  *  igb_set_interrupt_capability - set MSI or MSI-X if supported
1114  *  @adapter: board private structure to initialize
1115  *  @msix: boolean value of MSIX capability
1116  *
1117  *  Attempt to configure interrupts using the best available
1118  *  capabilities of the hardware and kernel.
1119  **/
1120 static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
1121 {
1122 	int err;
1123 	int numvecs, i;
1124 
1125 	if (!msix)
1126 		goto msi_only;
1127 	adapter->flags |= IGB_FLAG_HAS_MSIX;
1128 
1129 	/* Number of supported queues. */
1130 	adapter->num_rx_queues = adapter->rss_queues;
1131 	if (adapter->vfs_allocated_count)
1132 		adapter->num_tx_queues = 1;
1133 	else
1134 		adapter->num_tx_queues = adapter->rss_queues;
1135 
1136 	/* start with one vector for every Rx queue */
1137 	numvecs = adapter->num_rx_queues;
1138 
1139 	/* if Tx handler is separate add 1 for every Tx queue */
1140 	if (!(adapter->flags & IGB_FLAG_QUEUE_PAIRS))
1141 		numvecs += adapter->num_tx_queues;
1142 
1143 	/* store the number of vectors reserved for queues */
1144 	adapter->num_q_vectors = numvecs;
1145 
1146 	/* add 1 vector for link status interrupts */
1147 	numvecs++;
1148 	for (i = 0; i < numvecs; i++)
1149 		adapter->msix_entries[i].entry = i;
1150 
1151 	err = pci_enable_msix_range(adapter->pdev,
1152 				    adapter->msix_entries,
1153 				    numvecs,
1154 				    numvecs);
1155 	if (err > 0)
1156 		return;
1157 
1158 	igb_reset_interrupt_capability(adapter);
1159 
1160 	/* If we can't do MSI-X, try MSI */
1161 msi_only:
1162 	adapter->flags &= ~IGB_FLAG_HAS_MSIX;
1163 #ifdef CONFIG_PCI_IOV
1164 	/* disable SR-IOV for non MSI-X configurations */
1165 	if (adapter->vf_data) {
1166 		struct e1000_hw *hw = &adapter->hw;
1167 		/* disable iov and allow time for transactions to clear */
1168 		pci_disable_sriov(adapter->pdev);
1169 		msleep(500);
1170 
1171 		kfree(adapter->vf_mac_list);
1172 		adapter->vf_mac_list = NULL;
1173 		kfree(adapter->vf_data);
1174 		adapter->vf_data = NULL;
1175 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1176 		wrfl();
1177 		msleep(100);
1178 		dev_info(&adapter->pdev->dev, "IOV Disabled\n");
1179 	}
1180 #endif
1181 	adapter->vfs_allocated_count = 0;
1182 	adapter->rss_queues = 1;
1183 	adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
1184 	adapter->num_rx_queues = 1;
1185 	adapter->num_tx_queues = 1;
1186 	adapter->num_q_vectors = 1;
1187 	if (!pci_enable_msi(adapter->pdev))
1188 		adapter->flags |= IGB_FLAG_HAS_MSI;
1189 }
1190 
1191 static void igb_add_ring(struct igb_ring *ring,
1192 			 struct igb_ring_container *head)
1193 {
1194 	head->ring = ring;
1195 	head->count++;
1196 }
1197 
1198 /**
1199  *  igb_alloc_q_vector - Allocate memory for a single interrupt vector
1200  *  @adapter: board private structure to initialize
1201  *  @v_count: q_vectors allocated on adapter, used for ring interleaving
1202  *  @v_idx: index of vector in adapter struct
1203  *  @txr_count: total number of Tx rings to allocate
1204  *  @txr_idx: index of first Tx ring to allocate
1205  *  @rxr_count: total number of Rx rings to allocate
1206  *  @rxr_idx: index of first Rx ring to allocate
1207  *
1208  *  We allocate one q_vector.  If allocation fails we return -ENOMEM.
1209  **/
1210 static int igb_alloc_q_vector(struct igb_adapter *adapter,
1211 			      int v_count, int v_idx,
1212 			      int txr_count, int txr_idx,
1213 			      int rxr_count, int rxr_idx)
1214 {
1215 	struct igb_q_vector *q_vector;
1216 	struct igb_ring *ring;
1217 	int ring_count, size;
1218 
1219 	/* igb only supports 1 Tx and/or 1 Rx queue per vector */
1220 	if (txr_count > 1 || rxr_count > 1)
1221 		return -ENOMEM;
1222 
1223 	ring_count = txr_count + rxr_count;
1224 	size = sizeof(struct igb_q_vector) +
1225 	       (sizeof(struct igb_ring) * ring_count);
1226 
1227 	/* allocate q_vector and rings */
1228 	q_vector = adapter->q_vector[v_idx];
1229 	if (!q_vector) {
1230 		q_vector = kzalloc(size, GFP_KERNEL);
1231 	} else if (size > ksize(q_vector)) {
1232 		kfree_rcu(q_vector, rcu);
1233 		q_vector = kzalloc(size, GFP_KERNEL);
1234 	} else {
1235 		memset(q_vector, 0, size);
1236 	}
1237 	if (!q_vector)
1238 		return -ENOMEM;
1239 
1240 	/* initialize NAPI */
1241 	netif_napi_add(adapter->netdev, &q_vector->napi,
1242 		       igb_poll, 64);
1243 
1244 	/* tie q_vector and adapter together */
1245 	adapter->q_vector[v_idx] = q_vector;
1246 	q_vector->adapter = adapter;
1247 
1248 	/* initialize work limits */
1249 	q_vector->tx.work_limit = adapter->tx_work_limit;
1250 
1251 	/* initialize ITR configuration */
1252 	q_vector->itr_register = adapter->io_addr + E1000_EITR(0);
1253 	q_vector->itr_val = IGB_START_ITR;
1254 
1255 	/* initialize pointer to rings */
1256 	ring = q_vector->ring;
1257 
1258 	/* intialize ITR */
1259 	if (rxr_count) {
1260 		/* rx or rx/tx vector */
1261 		if (!adapter->rx_itr_setting || adapter->rx_itr_setting > 3)
1262 			q_vector->itr_val = adapter->rx_itr_setting;
1263 	} else {
1264 		/* tx only vector */
1265 		if (!adapter->tx_itr_setting || adapter->tx_itr_setting > 3)
1266 			q_vector->itr_val = adapter->tx_itr_setting;
1267 	}
1268 
1269 	if (txr_count) {
1270 		/* assign generic ring traits */
1271 		ring->dev = &adapter->pdev->dev;
1272 		ring->netdev = adapter->netdev;
1273 
1274 		/* configure backlink on ring */
1275 		ring->q_vector = q_vector;
1276 
1277 		/* update q_vector Tx values */
1278 		igb_add_ring(ring, &q_vector->tx);
1279 
1280 		/* For 82575, context index must be unique per ring. */
1281 		if (adapter->hw.mac.type == e1000_82575)
1282 			set_bit(IGB_RING_FLAG_TX_CTX_IDX, &ring->flags);
1283 
1284 		/* apply Tx specific ring traits */
1285 		ring->count = adapter->tx_ring_count;
1286 		ring->queue_index = txr_idx;
1287 
1288 		ring->cbs_enable = false;
1289 		ring->idleslope = 0;
1290 		ring->sendslope = 0;
1291 		ring->hicredit = 0;
1292 		ring->locredit = 0;
1293 
1294 		u64_stats_init(&ring->tx_syncp);
1295 		u64_stats_init(&ring->tx_syncp2);
1296 
1297 		/* assign ring to adapter */
1298 		adapter->tx_ring[txr_idx] = ring;
1299 
1300 		/* push pointer to next ring */
1301 		ring++;
1302 	}
1303 
1304 	if (rxr_count) {
1305 		/* assign generic ring traits */
1306 		ring->dev = &adapter->pdev->dev;
1307 		ring->netdev = adapter->netdev;
1308 
1309 		/* configure backlink on ring */
1310 		ring->q_vector = q_vector;
1311 
1312 		/* update q_vector Rx values */
1313 		igb_add_ring(ring, &q_vector->rx);
1314 
1315 		/* set flag indicating ring supports SCTP checksum offload */
1316 		if (adapter->hw.mac.type >= e1000_82576)
1317 			set_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags);
1318 
1319 		/* On i350, i354, i210, and i211, loopback VLAN packets
1320 		 * have the tag byte-swapped.
1321 		 */
1322 		if (adapter->hw.mac.type >= e1000_i350)
1323 			set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &ring->flags);
1324 
1325 		/* apply Rx specific ring traits */
1326 		ring->count = adapter->rx_ring_count;
1327 		ring->queue_index = rxr_idx;
1328 
1329 		u64_stats_init(&ring->rx_syncp);
1330 
1331 		/* assign ring to adapter */
1332 		adapter->rx_ring[rxr_idx] = ring;
1333 	}
1334 
1335 	return 0;
1336 }
1337 
1338 
1339 /**
1340  *  igb_alloc_q_vectors - Allocate memory for interrupt vectors
1341  *  @adapter: board private structure to initialize
1342  *
1343  *  We allocate one q_vector per queue interrupt.  If allocation fails we
1344  *  return -ENOMEM.
1345  **/
1346 static int igb_alloc_q_vectors(struct igb_adapter *adapter)
1347 {
1348 	int q_vectors = adapter->num_q_vectors;
1349 	int rxr_remaining = adapter->num_rx_queues;
1350 	int txr_remaining = adapter->num_tx_queues;
1351 	int rxr_idx = 0, txr_idx = 0, v_idx = 0;
1352 	int err;
1353 
1354 	if (q_vectors >= (rxr_remaining + txr_remaining)) {
1355 		for (; rxr_remaining; v_idx++) {
1356 			err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1357 						 0, 0, 1, rxr_idx);
1358 
1359 			if (err)
1360 				goto err_out;
1361 
1362 			/* update counts and index */
1363 			rxr_remaining--;
1364 			rxr_idx++;
1365 		}
1366 	}
1367 
1368 	for (; v_idx < q_vectors; v_idx++) {
1369 		int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
1370 		int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);
1371 
1372 		err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1373 					 tqpv, txr_idx, rqpv, rxr_idx);
1374 
1375 		if (err)
1376 			goto err_out;
1377 
1378 		/* update counts and index */
1379 		rxr_remaining -= rqpv;
1380 		txr_remaining -= tqpv;
1381 		rxr_idx++;
1382 		txr_idx++;
1383 	}
1384 
1385 	return 0;
1386 
1387 err_out:
1388 	adapter->num_tx_queues = 0;
1389 	adapter->num_rx_queues = 0;
1390 	adapter->num_q_vectors = 0;
1391 
1392 	while (v_idx--)
1393 		igb_free_q_vector(adapter, v_idx);
1394 
1395 	return -ENOMEM;
1396 }
1397 
1398 /**
1399  *  igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1400  *  @adapter: board private structure to initialize
1401  *  @msix: boolean value of MSIX capability
1402  *
1403  *  This function initializes the interrupts and allocates all of the queues.
1404  **/
1405 static int igb_init_interrupt_scheme(struct igb_adapter *adapter, bool msix)
1406 {
1407 	struct pci_dev *pdev = adapter->pdev;
1408 	int err;
1409 
1410 	igb_set_interrupt_capability(adapter, msix);
1411 
1412 	err = igb_alloc_q_vectors(adapter);
1413 	if (err) {
1414 		dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
1415 		goto err_alloc_q_vectors;
1416 	}
1417 
1418 	igb_cache_ring_register(adapter);
1419 
1420 	return 0;
1421 
1422 err_alloc_q_vectors:
1423 	igb_reset_interrupt_capability(adapter);
1424 	return err;
1425 }
1426 
1427 /**
1428  *  igb_request_irq - initialize interrupts
1429  *  @adapter: board private structure to initialize
1430  *
1431  *  Attempts to configure interrupts using the best available
1432  *  capabilities of the hardware and kernel.
1433  **/
1434 static int igb_request_irq(struct igb_adapter *adapter)
1435 {
1436 	struct net_device *netdev = adapter->netdev;
1437 	struct pci_dev *pdev = adapter->pdev;
1438 	int err = 0;
1439 
1440 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1441 		err = igb_request_msix(adapter);
1442 		if (!err)
1443 			goto request_done;
1444 		/* fall back to MSI */
1445 		igb_free_all_tx_resources(adapter);
1446 		igb_free_all_rx_resources(adapter);
1447 
1448 		igb_clear_interrupt_scheme(adapter);
1449 		err = igb_init_interrupt_scheme(adapter, false);
1450 		if (err)
1451 			goto request_done;
1452 
1453 		igb_setup_all_tx_resources(adapter);
1454 		igb_setup_all_rx_resources(adapter);
1455 		igb_configure(adapter);
1456 	}
1457 
1458 	igb_assign_vector(adapter->q_vector[0], 0);
1459 
1460 	if (adapter->flags & IGB_FLAG_HAS_MSI) {
1461 		err = request_irq(pdev->irq, igb_intr_msi, 0,
1462 				  netdev->name, adapter);
1463 		if (!err)
1464 			goto request_done;
1465 
1466 		/* fall back to legacy interrupts */
1467 		igb_reset_interrupt_capability(adapter);
1468 		adapter->flags &= ~IGB_FLAG_HAS_MSI;
1469 	}
1470 
1471 	err = request_irq(pdev->irq, igb_intr, IRQF_SHARED,
1472 			  netdev->name, adapter);
1473 
1474 	if (err)
1475 		dev_err(&pdev->dev, "Error %d getting interrupt\n",
1476 			err);
1477 
1478 request_done:
1479 	return err;
1480 }
1481 
1482 static void igb_free_irq(struct igb_adapter *adapter)
1483 {
1484 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1485 		int vector = 0, i;
1486 
1487 		free_irq(adapter->msix_entries[vector++].vector, adapter);
1488 
1489 		for (i = 0; i < adapter->num_q_vectors; i++)
1490 			free_irq(adapter->msix_entries[vector++].vector,
1491 				 adapter->q_vector[i]);
1492 	} else {
1493 		free_irq(adapter->pdev->irq, adapter);
1494 	}
1495 }
1496 
1497 /**
1498  *  igb_irq_disable - Mask off interrupt generation on the NIC
1499  *  @adapter: board private structure
1500  **/
1501 static void igb_irq_disable(struct igb_adapter *adapter)
1502 {
1503 	struct e1000_hw *hw = &adapter->hw;
1504 
1505 	/* we need to be careful when disabling interrupts.  The VFs are also
1506 	 * mapped into these registers and so clearing the bits can cause
1507 	 * issues on the VF drivers so we only need to clear what we set
1508 	 */
1509 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1510 		u32 regval = rd32(E1000_EIAM);
1511 
1512 		wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
1513 		wr32(E1000_EIMC, adapter->eims_enable_mask);
1514 		regval = rd32(E1000_EIAC);
1515 		wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
1516 	}
1517 
1518 	wr32(E1000_IAM, 0);
1519 	wr32(E1000_IMC, ~0);
1520 	wrfl();
1521 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1522 		int i;
1523 
1524 		for (i = 0; i < adapter->num_q_vectors; i++)
1525 			synchronize_irq(adapter->msix_entries[i].vector);
1526 	} else {
1527 		synchronize_irq(adapter->pdev->irq);
1528 	}
1529 }
1530 
1531 /**
1532  *  igb_irq_enable - Enable default interrupt generation settings
1533  *  @adapter: board private structure
1534  **/
1535 static void igb_irq_enable(struct igb_adapter *adapter)
1536 {
1537 	struct e1000_hw *hw = &adapter->hw;
1538 
1539 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1540 		u32 ims = E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_DRSTA;
1541 		u32 regval = rd32(E1000_EIAC);
1542 
1543 		wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
1544 		regval = rd32(E1000_EIAM);
1545 		wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
1546 		wr32(E1000_EIMS, adapter->eims_enable_mask);
1547 		if (adapter->vfs_allocated_count) {
1548 			wr32(E1000_MBVFIMR, 0xFF);
1549 			ims |= E1000_IMS_VMMB;
1550 		}
1551 		wr32(E1000_IMS, ims);
1552 	} else {
1553 		wr32(E1000_IMS, IMS_ENABLE_MASK |
1554 				E1000_IMS_DRSTA);
1555 		wr32(E1000_IAM, IMS_ENABLE_MASK |
1556 				E1000_IMS_DRSTA);
1557 	}
1558 }
1559 
1560 static void igb_update_mng_vlan(struct igb_adapter *adapter)
1561 {
1562 	struct e1000_hw *hw = &adapter->hw;
1563 	u16 pf_id = adapter->vfs_allocated_count;
1564 	u16 vid = adapter->hw.mng_cookie.vlan_id;
1565 	u16 old_vid = adapter->mng_vlan_id;
1566 
1567 	if (hw->mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1568 		/* add VID to filter table */
1569 		igb_vfta_set(hw, vid, pf_id, true, true);
1570 		adapter->mng_vlan_id = vid;
1571 	} else {
1572 		adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1573 	}
1574 
1575 	if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1576 	    (vid != old_vid) &&
1577 	    !test_bit(old_vid, adapter->active_vlans)) {
1578 		/* remove VID from filter table */
1579 		igb_vfta_set(hw, vid, pf_id, false, true);
1580 	}
1581 }
1582 
1583 /**
1584  *  igb_release_hw_control - release control of the h/w to f/w
1585  *  @adapter: address of board private structure
1586  *
1587  *  igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1588  *  For ASF and Pass Through versions of f/w this means that the
1589  *  driver is no longer loaded.
1590  **/
1591 static void igb_release_hw_control(struct igb_adapter *adapter)
1592 {
1593 	struct e1000_hw *hw = &adapter->hw;
1594 	u32 ctrl_ext;
1595 
1596 	/* Let firmware take over control of h/w */
1597 	ctrl_ext = rd32(E1000_CTRL_EXT);
1598 	wr32(E1000_CTRL_EXT,
1599 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1600 }
1601 
1602 /**
1603  *  igb_get_hw_control - get control of the h/w from f/w
1604  *  @adapter: address of board private structure
1605  *
1606  *  igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1607  *  For ASF and Pass Through versions of f/w this means that
1608  *  the driver is loaded.
1609  **/
1610 static void igb_get_hw_control(struct igb_adapter *adapter)
1611 {
1612 	struct e1000_hw *hw = &adapter->hw;
1613 	u32 ctrl_ext;
1614 
1615 	/* Let firmware know the driver has taken over */
1616 	ctrl_ext = rd32(E1000_CTRL_EXT);
1617 	wr32(E1000_CTRL_EXT,
1618 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1619 }
1620 
1621 static void enable_fqtss(struct igb_adapter *adapter, bool enable)
1622 {
1623 	struct net_device *netdev = adapter->netdev;
1624 	struct e1000_hw *hw = &adapter->hw;
1625 
1626 	WARN_ON(hw->mac.type != e1000_i210);
1627 
1628 	if (enable)
1629 		adapter->flags |= IGB_FLAG_FQTSS;
1630 	else
1631 		adapter->flags &= ~IGB_FLAG_FQTSS;
1632 
1633 	if (netif_running(netdev))
1634 		schedule_work(&adapter->reset_task);
1635 }
1636 
1637 static bool is_fqtss_enabled(struct igb_adapter *adapter)
1638 {
1639 	return (adapter->flags & IGB_FLAG_FQTSS) ? true : false;
1640 }
1641 
1642 static void set_tx_desc_fetch_prio(struct e1000_hw *hw, int queue,
1643 				   enum tx_queue_prio prio)
1644 {
1645 	u32 val;
1646 
1647 	WARN_ON(hw->mac.type != e1000_i210);
1648 	WARN_ON(queue < 0 || queue > 4);
1649 
1650 	val = rd32(E1000_I210_TXDCTL(queue));
1651 
1652 	if (prio == TX_QUEUE_PRIO_HIGH)
1653 		val |= E1000_TXDCTL_PRIORITY;
1654 	else
1655 		val &= ~E1000_TXDCTL_PRIORITY;
1656 
1657 	wr32(E1000_I210_TXDCTL(queue), val);
1658 }
1659 
1660 static void set_queue_mode(struct e1000_hw *hw, int queue, enum queue_mode mode)
1661 {
1662 	u32 val;
1663 
1664 	WARN_ON(hw->mac.type != e1000_i210);
1665 	WARN_ON(queue < 0 || queue > 1);
1666 
1667 	val = rd32(E1000_I210_TQAVCC(queue));
1668 
1669 	if (mode == QUEUE_MODE_STREAM_RESERVATION)
1670 		val |= E1000_TQAVCC_QUEUEMODE;
1671 	else
1672 		val &= ~E1000_TQAVCC_QUEUEMODE;
1673 
1674 	wr32(E1000_I210_TQAVCC(queue), val);
1675 }
1676 
1677 /**
1678  *  igb_configure_cbs - Configure Credit-Based Shaper (CBS)
1679  *  @adapter: pointer to adapter struct
1680  *  @queue: queue number
1681  *  @enable: true = enable CBS, false = disable CBS
1682  *  @idleslope: idleSlope in kbps
1683  *  @sendslope: sendSlope in kbps
1684  *  @hicredit: hiCredit in bytes
1685  *  @locredit: loCredit in bytes
1686  *
1687  *  Configure CBS for a given hardware queue. When disabling, idleslope,
1688  *  sendslope, hicredit, locredit arguments are ignored. Returns 0 if
1689  *  success. Negative otherwise.
1690  **/
1691 static void igb_configure_cbs(struct igb_adapter *adapter, int queue,
1692 			      bool enable, int idleslope, int sendslope,
1693 			      int hicredit, int locredit)
1694 {
1695 	struct net_device *netdev = adapter->netdev;
1696 	struct e1000_hw *hw = &adapter->hw;
1697 	u32 tqavcc;
1698 	u16 value;
1699 
1700 	WARN_ON(hw->mac.type != e1000_i210);
1701 	WARN_ON(queue < 0 || queue > 1);
1702 
1703 	if (enable) {
1704 		set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_HIGH);
1705 		set_queue_mode(hw, queue, QUEUE_MODE_STREAM_RESERVATION);
1706 
1707 		/* According to i210 datasheet section 7.2.7.7, we should set
1708 		 * the 'idleSlope' field from TQAVCC register following the
1709 		 * equation:
1710 		 *
1711 		 * For 100 Mbps link speed:
1712 		 *
1713 		 *     value = BW * 0x7735 * 0.2                          (E1)
1714 		 *
1715 		 * For 1000Mbps link speed:
1716 		 *
1717 		 *     value = BW * 0x7735 * 2                            (E2)
1718 		 *
1719 		 * E1 and E2 can be merged into one equation as shown below.
1720 		 * Note that 'link-speed' is in Mbps.
1721 		 *
1722 		 *     value = BW * 0x7735 * 2 * link-speed
1723 		 *                           --------------               (E3)
1724 		 *                                1000
1725 		 *
1726 		 * 'BW' is the percentage bandwidth out of full link speed
1727 		 * which can be found with the following equation. Note that
1728 		 * idleSlope here is the parameter from this function which
1729 		 * is in kbps.
1730 		 *
1731 		 *     BW =     idleSlope
1732 		 *          -----------------                             (E4)
1733 		 *          link-speed * 1000
1734 		 *
1735 		 * That said, we can come up with a generic equation to
1736 		 * calculate the value we should set it TQAVCC register by
1737 		 * replacing 'BW' in E3 by E4. The resulting equation is:
1738 		 *
1739 		 * value =     idleSlope     * 0x7735 * 2 * link-speed
1740 		 *         -----------------            --------------    (E5)
1741 		 *         link-speed * 1000                 1000
1742 		 *
1743 		 * 'link-speed' is present in both sides of the fraction so
1744 		 * it is canceled out. The final equation is the following:
1745 		 *
1746 		 *     value = idleSlope * 61034
1747 		 *             -----------------                          (E6)
1748 		 *                  1000000
1749 		 *
1750 		 * NOTE: For i210, given the above, we can see that idleslope
1751 		 *       is represented in 16.38431 kbps units by the value at
1752 		 *       the TQAVCC register (1Gbps / 61034), which reduces
1753 		 *       the granularity for idleslope increments.
1754 		 *       For instance, if you want to configure a 2576kbps
1755 		 *       idleslope, the value to be written on the register
1756 		 *       would have to be 157.23. If rounded down, you end
1757 		 *       up with less bandwidth available than originally
1758 		 *       required (~2572 kbps). If rounded up, you end up
1759 		 *       with a higher bandwidth (~2589 kbps). Below the
1760 		 *       approach we take is to always round up the
1761 		 *       calculated value, so the resulting bandwidth might
1762 		 *       be slightly higher for some configurations.
1763 		 */
1764 		value = DIV_ROUND_UP_ULL(idleslope * 61034ULL, 1000000);
1765 
1766 		tqavcc = rd32(E1000_I210_TQAVCC(queue));
1767 		tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1768 		tqavcc |= value;
1769 		wr32(E1000_I210_TQAVCC(queue), tqavcc);
1770 
1771 		wr32(E1000_I210_TQAVHC(queue), 0x80000000 + hicredit * 0x7735);
1772 	} else {
1773 		set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_LOW);
1774 		set_queue_mode(hw, queue, QUEUE_MODE_STRICT_PRIORITY);
1775 
1776 		/* Set idleSlope to zero. */
1777 		tqavcc = rd32(E1000_I210_TQAVCC(queue));
1778 		tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1779 		wr32(E1000_I210_TQAVCC(queue), tqavcc);
1780 
1781 		/* Set hiCredit to zero. */
1782 		wr32(E1000_I210_TQAVHC(queue), 0);
1783 	}
1784 
1785 	/* XXX: In i210 controller the sendSlope and loCredit parameters from
1786 	 * CBS are not configurable by software so we don't do any 'controller
1787 	 * configuration' in respect to these parameters.
1788 	 */
1789 
1790 	netdev_dbg(netdev, "CBS %s: queue %d idleslope %d sendslope %d hiCredit %d locredit %d\n",
1791 		   (enable) ? "enabled" : "disabled", queue,
1792 		   idleslope, sendslope, hicredit, locredit);
1793 }
1794 
1795 static int igb_save_cbs_params(struct igb_adapter *adapter, int queue,
1796 			       bool enable, int idleslope, int sendslope,
1797 			       int hicredit, int locredit)
1798 {
1799 	struct igb_ring *ring;
1800 
1801 	if (queue < 0 || queue > adapter->num_tx_queues)
1802 		return -EINVAL;
1803 
1804 	ring = adapter->tx_ring[queue];
1805 
1806 	ring->cbs_enable = enable;
1807 	ring->idleslope = idleslope;
1808 	ring->sendslope = sendslope;
1809 	ring->hicredit = hicredit;
1810 	ring->locredit = locredit;
1811 
1812 	return 0;
1813 }
1814 
1815 static bool is_any_cbs_enabled(struct igb_adapter *adapter)
1816 {
1817 	struct igb_ring *ring;
1818 	int i;
1819 
1820 	for (i = 0; i < adapter->num_tx_queues; i++) {
1821 		ring = adapter->tx_ring[i];
1822 
1823 		if (ring->cbs_enable)
1824 			return true;
1825 	}
1826 
1827 	return false;
1828 }
1829 
1830 static void igb_setup_tx_mode(struct igb_adapter *adapter)
1831 {
1832 	struct net_device *netdev = adapter->netdev;
1833 	struct e1000_hw *hw = &adapter->hw;
1834 	u32 val;
1835 
1836 	/* Only i210 controller supports changing the transmission mode. */
1837 	if (hw->mac.type != e1000_i210)
1838 		return;
1839 
1840 	if (is_fqtss_enabled(adapter)) {
1841 		int i, max_queue;
1842 
1843 		/* Configure TQAVCTRL register: set transmit mode to 'Qav',
1844 		 * set data fetch arbitration to 'round robin' and set data
1845 		 * transfer arbitration to 'credit shaper algorithm.
1846 		 */
1847 		val = rd32(E1000_I210_TQAVCTRL);
1848 		val |= E1000_TQAVCTRL_XMIT_MODE | E1000_TQAVCTRL_DATATRANARB;
1849 		val &= ~E1000_TQAVCTRL_DATAFETCHARB;
1850 		wr32(E1000_I210_TQAVCTRL, val);
1851 
1852 		/* Configure Tx and Rx packet buffers sizes as described in
1853 		 * i210 datasheet section 7.2.7.7.
1854 		 */
1855 		val = rd32(E1000_TXPBS);
1856 		val &= ~I210_TXPBSIZE_MASK;
1857 		val |= I210_TXPBSIZE_PB0_8KB | I210_TXPBSIZE_PB1_8KB |
1858 			I210_TXPBSIZE_PB2_4KB | I210_TXPBSIZE_PB3_4KB;
1859 		wr32(E1000_TXPBS, val);
1860 
1861 		val = rd32(E1000_RXPBS);
1862 		val &= ~I210_RXPBSIZE_MASK;
1863 		val |= I210_RXPBSIZE_PB_32KB;
1864 		wr32(E1000_RXPBS, val);
1865 
1866 		/* Section 8.12.9 states that MAX_TPKT_SIZE from DTXMXPKTSZ
1867 		 * register should not exceed the buffer size programmed in
1868 		 * TXPBS. The smallest buffer size programmed in TXPBS is 4kB
1869 		 * so according to the datasheet we should set MAX_TPKT_SIZE to
1870 		 * 4kB / 64.
1871 		 *
1872 		 * However, when we do so, no frame from queue 2 and 3 are
1873 		 * transmitted.  It seems the MAX_TPKT_SIZE should not be great
1874 		 * or _equal_ to the buffer size programmed in TXPBS. For this
1875 		 * reason, we set set MAX_ TPKT_SIZE to (4kB - 1) / 64.
1876 		 */
1877 		val = (4096 - 1) / 64;
1878 		wr32(E1000_I210_DTXMXPKTSZ, val);
1879 
1880 		/* Since FQTSS mode is enabled, apply any CBS configuration
1881 		 * previously set. If no previous CBS configuration has been
1882 		 * done, then the initial configuration is applied, which means
1883 		 * CBS is disabled.
1884 		 */
1885 		max_queue = (adapter->num_tx_queues < I210_SR_QUEUES_NUM) ?
1886 			    adapter->num_tx_queues : I210_SR_QUEUES_NUM;
1887 
1888 		for (i = 0; i < max_queue; i++) {
1889 			struct igb_ring *ring = adapter->tx_ring[i];
1890 
1891 			igb_configure_cbs(adapter, i, ring->cbs_enable,
1892 					  ring->idleslope, ring->sendslope,
1893 					  ring->hicredit, ring->locredit);
1894 		}
1895 	} else {
1896 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
1897 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
1898 		wr32(E1000_I210_DTXMXPKTSZ, I210_DTXMXPKTSZ_DEFAULT);
1899 
1900 		val = rd32(E1000_I210_TQAVCTRL);
1901 		/* According to Section 8.12.21, the other flags we've set when
1902 		 * enabling FQTSS are not relevant when disabling FQTSS so we
1903 		 * don't set they here.
1904 		 */
1905 		val &= ~E1000_TQAVCTRL_XMIT_MODE;
1906 		wr32(E1000_I210_TQAVCTRL, val);
1907 	}
1908 
1909 	netdev_dbg(netdev, "FQTSS %s\n", (is_fqtss_enabled(adapter)) ?
1910 		   "enabled" : "disabled");
1911 }
1912 
1913 /**
1914  *  igb_configure - configure the hardware for RX and TX
1915  *  @adapter: private board structure
1916  **/
1917 static void igb_configure(struct igb_adapter *adapter)
1918 {
1919 	struct net_device *netdev = adapter->netdev;
1920 	int i;
1921 
1922 	igb_get_hw_control(adapter);
1923 	igb_set_rx_mode(netdev);
1924 	igb_setup_tx_mode(adapter);
1925 
1926 	igb_restore_vlan(adapter);
1927 
1928 	igb_setup_tctl(adapter);
1929 	igb_setup_mrqc(adapter);
1930 	igb_setup_rctl(adapter);
1931 
1932 	igb_nfc_filter_restore(adapter);
1933 	igb_configure_tx(adapter);
1934 	igb_configure_rx(adapter);
1935 
1936 	igb_rx_fifo_flush_82575(&adapter->hw);
1937 
1938 	/* call igb_desc_unused which always leaves
1939 	 * at least 1 descriptor unused to make sure
1940 	 * next_to_use != next_to_clean
1941 	 */
1942 	for (i = 0; i < adapter->num_rx_queues; i++) {
1943 		struct igb_ring *ring = adapter->rx_ring[i];
1944 		igb_alloc_rx_buffers(ring, igb_desc_unused(ring));
1945 	}
1946 }
1947 
1948 /**
1949  *  igb_power_up_link - Power up the phy/serdes link
1950  *  @adapter: address of board private structure
1951  **/
1952 void igb_power_up_link(struct igb_adapter *adapter)
1953 {
1954 	igb_reset_phy(&adapter->hw);
1955 
1956 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
1957 		igb_power_up_phy_copper(&adapter->hw);
1958 	else
1959 		igb_power_up_serdes_link_82575(&adapter->hw);
1960 
1961 	igb_setup_link(&adapter->hw);
1962 }
1963 
1964 /**
1965  *  igb_power_down_link - Power down the phy/serdes link
1966  *  @adapter: address of board private structure
1967  */
1968 static void igb_power_down_link(struct igb_adapter *adapter)
1969 {
1970 	if (adapter->hw.phy.media_type == e1000_media_type_copper)
1971 		igb_power_down_phy_copper_82575(&adapter->hw);
1972 	else
1973 		igb_shutdown_serdes_link_82575(&adapter->hw);
1974 }
1975 
1976 /**
1977  * Detect and switch function for Media Auto Sense
1978  * @adapter: address of the board private structure
1979  **/
1980 static void igb_check_swap_media(struct igb_adapter *adapter)
1981 {
1982 	struct e1000_hw *hw = &adapter->hw;
1983 	u32 ctrl_ext, connsw;
1984 	bool swap_now = false;
1985 
1986 	ctrl_ext = rd32(E1000_CTRL_EXT);
1987 	connsw = rd32(E1000_CONNSW);
1988 
1989 	/* need to live swap if current media is copper and we have fiber/serdes
1990 	 * to go to.
1991 	 */
1992 
1993 	if ((hw->phy.media_type == e1000_media_type_copper) &&
1994 	    (!(connsw & E1000_CONNSW_AUTOSENSE_EN))) {
1995 		swap_now = true;
1996 	} else if (!(connsw & E1000_CONNSW_SERDESD)) {
1997 		/* copper signal takes time to appear */
1998 		if (adapter->copper_tries < 4) {
1999 			adapter->copper_tries++;
2000 			connsw |= E1000_CONNSW_AUTOSENSE_CONF;
2001 			wr32(E1000_CONNSW, connsw);
2002 			return;
2003 		} else {
2004 			adapter->copper_tries = 0;
2005 			if ((connsw & E1000_CONNSW_PHYSD) &&
2006 			    (!(connsw & E1000_CONNSW_PHY_PDN))) {
2007 				swap_now = true;
2008 				connsw &= ~E1000_CONNSW_AUTOSENSE_CONF;
2009 				wr32(E1000_CONNSW, connsw);
2010 			}
2011 		}
2012 	}
2013 
2014 	if (!swap_now)
2015 		return;
2016 
2017 	switch (hw->phy.media_type) {
2018 	case e1000_media_type_copper:
2019 		netdev_info(adapter->netdev,
2020 			"MAS: changing media to fiber/serdes\n");
2021 		ctrl_ext |=
2022 			E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2023 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
2024 		adapter->copper_tries = 0;
2025 		break;
2026 	case e1000_media_type_internal_serdes:
2027 	case e1000_media_type_fiber:
2028 		netdev_info(adapter->netdev,
2029 			"MAS: changing media to copper\n");
2030 		ctrl_ext &=
2031 			~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2032 		adapter->flags |= IGB_FLAG_MEDIA_RESET;
2033 		break;
2034 	default:
2035 		/* shouldn't get here during regular operation */
2036 		netdev_err(adapter->netdev,
2037 			"AMS: Invalid media type found, returning\n");
2038 		break;
2039 	}
2040 	wr32(E1000_CTRL_EXT, ctrl_ext);
2041 }
2042 
2043 /**
2044  *  igb_up - Open the interface and prepare it to handle traffic
2045  *  @adapter: board private structure
2046  **/
2047 int igb_up(struct igb_adapter *adapter)
2048 {
2049 	struct e1000_hw *hw = &adapter->hw;
2050 	int i;
2051 
2052 	/* hardware has been reset, we need to reload some things */
2053 	igb_configure(adapter);
2054 
2055 	clear_bit(__IGB_DOWN, &adapter->state);
2056 
2057 	for (i = 0; i < adapter->num_q_vectors; i++)
2058 		napi_enable(&(adapter->q_vector[i]->napi));
2059 
2060 	if (adapter->flags & IGB_FLAG_HAS_MSIX)
2061 		igb_configure_msix(adapter);
2062 	else
2063 		igb_assign_vector(adapter->q_vector[0], 0);
2064 
2065 	/* Clear any pending interrupts. */
2066 	rd32(E1000_ICR);
2067 	igb_irq_enable(adapter);
2068 
2069 	/* notify VFs that reset has been completed */
2070 	if (adapter->vfs_allocated_count) {
2071 		u32 reg_data = rd32(E1000_CTRL_EXT);
2072 
2073 		reg_data |= E1000_CTRL_EXT_PFRSTD;
2074 		wr32(E1000_CTRL_EXT, reg_data);
2075 	}
2076 
2077 	netif_tx_start_all_queues(adapter->netdev);
2078 
2079 	/* start the watchdog. */
2080 	hw->mac.get_link_status = 1;
2081 	schedule_work(&adapter->watchdog_task);
2082 
2083 	if ((adapter->flags & IGB_FLAG_EEE) &&
2084 	    (!hw->dev_spec._82575.eee_disable))
2085 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
2086 
2087 	return 0;
2088 }
2089 
2090 void igb_down(struct igb_adapter *adapter)
2091 {
2092 	struct net_device *netdev = adapter->netdev;
2093 	struct e1000_hw *hw = &adapter->hw;
2094 	u32 tctl, rctl;
2095 	int i;
2096 
2097 	/* signal that we're down so the interrupt handler does not
2098 	 * reschedule our watchdog timer
2099 	 */
2100 	set_bit(__IGB_DOWN, &adapter->state);
2101 
2102 	/* disable receives in the hardware */
2103 	rctl = rd32(E1000_RCTL);
2104 	wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
2105 	/* flush and sleep below */
2106 
2107 	igb_nfc_filter_exit(adapter);
2108 
2109 	netif_carrier_off(netdev);
2110 	netif_tx_stop_all_queues(netdev);
2111 
2112 	/* disable transmits in the hardware */
2113 	tctl = rd32(E1000_TCTL);
2114 	tctl &= ~E1000_TCTL_EN;
2115 	wr32(E1000_TCTL, tctl);
2116 	/* flush both disables and wait for them to finish */
2117 	wrfl();
2118 	usleep_range(10000, 11000);
2119 
2120 	igb_irq_disable(adapter);
2121 
2122 	adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
2123 
2124 	for (i = 0; i < adapter->num_q_vectors; i++) {
2125 		if (adapter->q_vector[i]) {
2126 			napi_synchronize(&adapter->q_vector[i]->napi);
2127 			napi_disable(&adapter->q_vector[i]->napi);
2128 		}
2129 	}
2130 
2131 	del_timer_sync(&adapter->watchdog_timer);
2132 	del_timer_sync(&adapter->phy_info_timer);
2133 
2134 	/* record the stats before reset*/
2135 	spin_lock(&adapter->stats64_lock);
2136 	igb_update_stats(adapter);
2137 	spin_unlock(&adapter->stats64_lock);
2138 
2139 	adapter->link_speed = 0;
2140 	adapter->link_duplex = 0;
2141 
2142 	if (!pci_channel_offline(adapter->pdev))
2143 		igb_reset(adapter);
2144 
2145 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
2146 	adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
2147 
2148 	igb_clean_all_tx_rings(adapter);
2149 	igb_clean_all_rx_rings(adapter);
2150 #ifdef CONFIG_IGB_DCA
2151 
2152 	/* since we reset the hardware DCA settings were cleared */
2153 	igb_setup_dca(adapter);
2154 #endif
2155 }
2156 
2157 void igb_reinit_locked(struct igb_adapter *adapter)
2158 {
2159 	WARN_ON(in_interrupt());
2160 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
2161 		usleep_range(1000, 2000);
2162 	igb_down(adapter);
2163 	igb_up(adapter);
2164 	clear_bit(__IGB_RESETTING, &adapter->state);
2165 }
2166 
2167 /** igb_enable_mas - Media Autosense re-enable after swap
2168  *
2169  * @adapter: adapter struct
2170  **/
2171 static void igb_enable_mas(struct igb_adapter *adapter)
2172 {
2173 	struct e1000_hw *hw = &adapter->hw;
2174 	u32 connsw = rd32(E1000_CONNSW);
2175 
2176 	/* configure for SerDes media detect */
2177 	if ((hw->phy.media_type == e1000_media_type_copper) &&
2178 	    (!(connsw & E1000_CONNSW_SERDESD))) {
2179 		connsw |= E1000_CONNSW_ENRGSRC;
2180 		connsw |= E1000_CONNSW_AUTOSENSE_EN;
2181 		wr32(E1000_CONNSW, connsw);
2182 		wrfl();
2183 	}
2184 }
2185 
2186 void igb_reset(struct igb_adapter *adapter)
2187 {
2188 	struct pci_dev *pdev = adapter->pdev;
2189 	struct e1000_hw *hw = &adapter->hw;
2190 	struct e1000_mac_info *mac = &hw->mac;
2191 	struct e1000_fc_info *fc = &hw->fc;
2192 	u32 pba, hwm;
2193 
2194 	/* Repartition Pba for greater than 9k mtu
2195 	 * To take effect CTRL.RST is required.
2196 	 */
2197 	switch (mac->type) {
2198 	case e1000_i350:
2199 	case e1000_i354:
2200 	case e1000_82580:
2201 		pba = rd32(E1000_RXPBS);
2202 		pba = igb_rxpbs_adjust_82580(pba);
2203 		break;
2204 	case e1000_82576:
2205 		pba = rd32(E1000_RXPBS);
2206 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2207 		break;
2208 	case e1000_82575:
2209 	case e1000_i210:
2210 	case e1000_i211:
2211 	default:
2212 		pba = E1000_PBA_34K;
2213 		break;
2214 	}
2215 
2216 	if (mac->type == e1000_82575) {
2217 		u32 min_rx_space, min_tx_space, needed_tx_space;
2218 
2219 		/* write Rx PBA so that hardware can report correct Tx PBA */
2220 		wr32(E1000_PBA, pba);
2221 
2222 		/* To maintain wire speed transmits, the Tx FIFO should be
2223 		 * large enough to accommodate two full transmit packets,
2224 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
2225 		 * the Rx FIFO should be large enough to accommodate at least
2226 		 * one full receive packet and is similarly rounded up and
2227 		 * expressed in KB.
2228 		 */
2229 		min_rx_space = DIV_ROUND_UP(MAX_JUMBO_FRAME_SIZE, 1024);
2230 
2231 		/* The Tx FIFO also stores 16 bytes of information about the Tx
2232 		 * but don't include Ethernet FCS because hardware appends it.
2233 		 * We only need to round down to the nearest 512 byte block
2234 		 * count since the value we care about is 2 frames, not 1.
2235 		 */
2236 		min_tx_space = adapter->max_frame_size;
2237 		min_tx_space += sizeof(union e1000_adv_tx_desc) - ETH_FCS_LEN;
2238 		min_tx_space = DIV_ROUND_UP(min_tx_space, 512);
2239 
2240 		/* upper 16 bits has Tx packet buffer allocation size in KB */
2241 		needed_tx_space = min_tx_space - (rd32(E1000_PBA) >> 16);
2242 
2243 		/* If current Tx allocation is less than the min Tx FIFO size,
2244 		 * and the min Tx FIFO size is less than the current Rx FIFO
2245 		 * allocation, take space away from current Rx allocation.
2246 		 */
2247 		if (needed_tx_space < pba) {
2248 			pba -= needed_tx_space;
2249 
2250 			/* if short on Rx space, Rx wins and must trump Tx
2251 			 * adjustment
2252 			 */
2253 			if (pba < min_rx_space)
2254 				pba = min_rx_space;
2255 		}
2256 
2257 		/* adjust PBA for jumbo frames */
2258 		wr32(E1000_PBA, pba);
2259 	}
2260 
2261 	/* flow control settings
2262 	 * The high water mark must be low enough to fit one full frame
2263 	 * after transmitting the pause frame.  As such we must have enough
2264 	 * space to allow for us to complete our current transmit and then
2265 	 * receive the frame that is in progress from the link partner.
2266 	 * Set it to:
2267 	 * - the full Rx FIFO size minus one full Tx plus one full Rx frame
2268 	 */
2269 	hwm = (pba << 10) - (adapter->max_frame_size + MAX_JUMBO_FRAME_SIZE);
2270 
2271 	fc->high_water = hwm & 0xFFFFFFF0;	/* 16-byte granularity */
2272 	fc->low_water = fc->high_water - 16;
2273 	fc->pause_time = 0xFFFF;
2274 	fc->send_xon = 1;
2275 	fc->current_mode = fc->requested_mode;
2276 
2277 	/* disable receive for all VFs and wait one second */
2278 	if (adapter->vfs_allocated_count) {
2279 		int i;
2280 
2281 		for (i = 0 ; i < adapter->vfs_allocated_count; i++)
2282 			adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
2283 
2284 		/* ping all the active vfs to let them know we are going down */
2285 		igb_ping_all_vfs(adapter);
2286 
2287 		/* disable transmits and receives */
2288 		wr32(E1000_VFRE, 0);
2289 		wr32(E1000_VFTE, 0);
2290 	}
2291 
2292 	/* Allow time for pending master requests to run */
2293 	hw->mac.ops.reset_hw(hw);
2294 	wr32(E1000_WUC, 0);
2295 
2296 	if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
2297 		/* need to resetup here after media swap */
2298 		adapter->ei.get_invariants(hw);
2299 		adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
2300 	}
2301 	if ((mac->type == e1000_82575) &&
2302 	    (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
2303 		igb_enable_mas(adapter);
2304 	}
2305 	if (hw->mac.ops.init_hw(hw))
2306 		dev_err(&pdev->dev, "Hardware Error\n");
2307 
2308 	/* RAR registers were cleared during init_hw, clear mac table */
2309 	igb_flush_mac_table(adapter);
2310 	__dev_uc_unsync(adapter->netdev, NULL);
2311 
2312 	/* Recover default RAR entry */
2313 	igb_set_default_mac_filter(adapter);
2314 
2315 	/* Flow control settings reset on hardware reset, so guarantee flow
2316 	 * control is off when forcing speed.
2317 	 */
2318 	if (!hw->mac.autoneg)
2319 		igb_force_mac_fc(hw);
2320 
2321 	igb_init_dmac(adapter, pba);
2322 #ifdef CONFIG_IGB_HWMON
2323 	/* Re-initialize the thermal sensor on i350 devices. */
2324 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
2325 		if (mac->type == e1000_i350 && hw->bus.func == 0) {
2326 			/* If present, re-initialize the external thermal sensor
2327 			 * interface.
2328 			 */
2329 			if (adapter->ets)
2330 				mac->ops.init_thermal_sensor_thresh(hw);
2331 		}
2332 	}
2333 #endif
2334 	/* Re-establish EEE setting */
2335 	if (hw->phy.media_type == e1000_media_type_copper) {
2336 		switch (mac->type) {
2337 		case e1000_i350:
2338 		case e1000_i210:
2339 		case e1000_i211:
2340 			igb_set_eee_i350(hw, true, true);
2341 			break;
2342 		case e1000_i354:
2343 			igb_set_eee_i354(hw, true, true);
2344 			break;
2345 		default:
2346 			break;
2347 		}
2348 	}
2349 	if (!netif_running(adapter->netdev))
2350 		igb_power_down_link(adapter);
2351 
2352 	igb_update_mng_vlan(adapter);
2353 
2354 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2355 	wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2356 
2357 	/* Re-enable PTP, where applicable. */
2358 	if (adapter->ptp_flags & IGB_PTP_ENABLED)
2359 		igb_ptp_reset(adapter);
2360 
2361 	igb_get_phy_info(hw);
2362 }
2363 
2364 static netdev_features_t igb_fix_features(struct net_device *netdev,
2365 	netdev_features_t features)
2366 {
2367 	/* Since there is no support for separate Rx/Tx vlan accel
2368 	 * enable/disable make sure Tx flag is always in same state as Rx.
2369 	 */
2370 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2371 		features |= NETIF_F_HW_VLAN_CTAG_TX;
2372 	else
2373 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2374 
2375 	return features;
2376 }
2377 
2378 static int igb_set_features(struct net_device *netdev,
2379 	netdev_features_t features)
2380 {
2381 	netdev_features_t changed = netdev->features ^ features;
2382 	struct igb_adapter *adapter = netdev_priv(netdev);
2383 
2384 	if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2385 		igb_vlan_mode(netdev, features);
2386 
2387 	if (!(changed & (NETIF_F_RXALL | NETIF_F_NTUPLE)))
2388 		return 0;
2389 
2390 	if (!(features & NETIF_F_NTUPLE)) {
2391 		struct hlist_node *node2;
2392 		struct igb_nfc_filter *rule;
2393 
2394 		spin_lock(&adapter->nfc_lock);
2395 		hlist_for_each_entry_safe(rule, node2,
2396 					  &adapter->nfc_filter_list, nfc_node) {
2397 			igb_erase_filter(adapter, rule);
2398 			hlist_del(&rule->nfc_node);
2399 			kfree(rule);
2400 		}
2401 		spin_unlock(&adapter->nfc_lock);
2402 		adapter->nfc_filter_count = 0;
2403 	}
2404 
2405 	netdev->features = features;
2406 
2407 	if (netif_running(netdev))
2408 		igb_reinit_locked(adapter);
2409 	else
2410 		igb_reset(adapter);
2411 
2412 	return 0;
2413 }
2414 
2415 static int igb_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
2416 			   struct net_device *dev,
2417 			   const unsigned char *addr, u16 vid,
2418 			   u16 flags)
2419 {
2420 	/* guarantee we can provide a unique filter for the unicast address */
2421 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
2422 		struct igb_adapter *adapter = netdev_priv(dev);
2423 		int vfn = adapter->vfs_allocated_count;
2424 
2425 		if (netdev_uc_count(dev) >= igb_available_rars(adapter, vfn))
2426 			return -ENOMEM;
2427 	}
2428 
2429 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
2430 }
2431 
2432 #define IGB_MAX_MAC_HDR_LEN	127
2433 #define IGB_MAX_NETWORK_HDR_LEN	511
2434 
2435 static netdev_features_t
2436 igb_features_check(struct sk_buff *skb, struct net_device *dev,
2437 		   netdev_features_t features)
2438 {
2439 	unsigned int network_hdr_len, mac_hdr_len;
2440 
2441 	/* Make certain the headers can be described by a context descriptor */
2442 	mac_hdr_len = skb_network_header(skb) - skb->data;
2443 	if (unlikely(mac_hdr_len > IGB_MAX_MAC_HDR_LEN))
2444 		return features & ~(NETIF_F_HW_CSUM |
2445 				    NETIF_F_SCTP_CRC |
2446 				    NETIF_F_HW_VLAN_CTAG_TX |
2447 				    NETIF_F_TSO |
2448 				    NETIF_F_TSO6);
2449 
2450 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
2451 	if (unlikely(network_hdr_len >  IGB_MAX_NETWORK_HDR_LEN))
2452 		return features & ~(NETIF_F_HW_CSUM |
2453 				    NETIF_F_SCTP_CRC |
2454 				    NETIF_F_TSO |
2455 				    NETIF_F_TSO6);
2456 
2457 	/* We can only support IPV4 TSO in tunnels if we can mangle the
2458 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
2459 	 */
2460 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
2461 		features &= ~NETIF_F_TSO;
2462 
2463 	return features;
2464 }
2465 
2466 static int igb_offload_cbs(struct igb_adapter *adapter,
2467 			   struct tc_cbs_qopt_offload *qopt)
2468 {
2469 	struct e1000_hw *hw = &adapter->hw;
2470 	int err;
2471 
2472 	/* CBS offloading is only supported by i210 controller. */
2473 	if (hw->mac.type != e1000_i210)
2474 		return -EOPNOTSUPP;
2475 
2476 	/* CBS offloading is only supported by queue 0 and queue 1. */
2477 	if (qopt->queue < 0 || qopt->queue > 1)
2478 		return -EINVAL;
2479 
2480 	err = igb_save_cbs_params(adapter, qopt->queue, qopt->enable,
2481 				  qopt->idleslope, qopt->sendslope,
2482 				  qopt->hicredit, qopt->locredit);
2483 	if (err)
2484 		return err;
2485 
2486 	if (is_fqtss_enabled(adapter)) {
2487 		igb_configure_cbs(adapter, qopt->queue, qopt->enable,
2488 				  qopt->idleslope, qopt->sendslope,
2489 				  qopt->hicredit, qopt->locredit);
2490 
2491 		if (!is_any_cbs_enabled(adapter))
2492 			enable_fqtss(adapter, false);
2493 
2494 	} else {
2495 		enable_fqtss(adapter, true);
2496 	}
2497 
2498 	return 0;
2499 }
2500 
2501 static int igb_setup_tc(struct net_device *dev, enum tc_setup_type type,
2502 			void *type_data)
2503 {
2504 	struct igb_adapter *adapter = netdev_priv(dev);
2505 
2506 	switch (type) {
2507 	case TC_SETUP_QDISC_CBS:
2508 		return igb_offload_cbs(adapter, type_data);
2509 
2510 	default:
2511 		return -EOPNOTSUPP;
2512 	}
2513 }
2514 
2515 static const struct net_device_ops igb_netdev_ops = {
2516 	.ndo_open		= igb_open,
2517 	.ndo_stop		= igb_close,
2518 	.ndo_start_xmit		= igb_xmit_frame,
2519 	.ndo_get_stats64	= igb_get_stats64,
2520 	.ndo_set_rx_mode	= igb_set_rx_mode,
2521 	.ndo_set_mac_address	= igb_set_mac,
2522 	.ndo_change_mtu		= igb_change_mtu,
2523 	.ndo_do_ioctl		= igb_ioctl,
2524 	.ndo_tx_timeout		= igb_tx_timeout,
2525 	.ndo_validate_addr	= eth_validate_addr,
2526 	.ndo_vlan_rx_add_vid	= igb_vlan_rx_add_vid,
2527 	.ndo_vlan_rx_kill_vid	= igb_vlan_rx_kill_vid,
2528 	.ndo_set_vf_mac		= igb_ndo_set_vf_mac,
2529 	.ndo_set_vf_vlan	= igb_ndo_set_vf_vlan,
2530 	.ndo_set_vf_rate	= igb_ndo_set_vf_bw,
2531 	.ndo_set_vf_spoofchk	= igb_ndo_set_vf_spoofchk,
2532 	.ndo_set_vf_trust	= igb_ndo_set_vf_trust,
2533 	.ndo_get_vf_config	= igb_ndo_get_vf_config,
2534 #ifdef CONFIG_NET_POLL_CONTROLLER
2535 	.ndo_poll_controller	= igb_netpoll,
2536 #endif
2537 	.ndo_fix_features	= igb_fix_features,
2538 	.ndo_set_features	= igb_set_features,
2539 	.ndo_fdb_add		= igb_ndo_fdb_add,
2540 	.ndo_features_check	= igb_features_check,
2541 	.ndo_setup_tc		= igb_setup_tc,
2542 };
2543 
2544 /**
2545  * igb_set_fw_version - Configure version string for ethtool
2546  * @adapter: adapter struct
2547  **/
2548 void igb_set_fw_version(struct igb_adapter *adapter)
2549 {
2550 	struct e1000_hw *hw = &adapter->hw;
2551 	struct e1000_fw_version fw;
2552 
2553 	igb_get_fw_version(hw, &fw);
2554 
2555 	switch (hw->mac.type) {
2556 	case e1000_i210:
2557 	case e1000_i211:
2558 		if (!(igb_get_flash_presence_i210(hw))) {
2559 			snprintf(adapter->fw_version,
2560 				 sizeof(adapter->fw_version),
2561 				 "%2d.%2d-%d",
2562 				 fw.invm_major, fw.invm_minor,
2563 				 fw.invm_img_type);
2564 			break;
2565 		}
2566 		/* fall through */
2567 	default:
2568 		/* if option is rom valid, display its version too */
2569 		if (fw.or_valid) {
2570 			snprintf(adapter->fw_version,
2571 				 sizeof(adapter->fw_version),
2572 				 "%d.%d, 0x%08x, %d.%d.%d",
2573 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2574 				 fw.or_major, fw.or_build, fw.or_patch);
2575 		/* no option rom */
2576 		} else if (fw.etrack_id != 0X0000) {
2577 			snprintf(adapter->fw_version,
2578 			    sizeof(adapter->fw_version),
2579 			    "%d.%d, 0x%08x",
2580 			    fw.eep_major, fw.eep_minor, fw.etrack_id);
2581 		} else {
2582 		snprintf(adapter->fw_version,
2583 		    sizeof(adapter->fw_version),
2584 		    "%d.%d.%d",
2585 		    fw.eep_major, fw.eep_minor, fw.eep_build);
2586 		}
2587 		break;
2588 	}
2589 }
2590 
2591 /**
2592  * igb_init_mas - init Media Autosense feature if enabled in the NVM
2593  *
2594  * @adapter: adapter struct
2595  **/
2596 static void igb_init_mas(struct igb_adapter *adapter)
2597 {
2598 	struct e1000_hw *hw = &adapter->hw;
2599 	u16 eeprom_data;
2600 
2601 	hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
2602 	switch (hw->bus.func) {
2603 	case E1000_FUNC_0:
2604 		if (eeprom_data & IGB_MAS_ENABLE_0) {
2605 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2606 			netdev_info(adapter->netdev,
2607 				"MAS: Enabling Media Autosense for port %d\n",
2608 				hw->bus.func);
2609 		}
2610 		break;
2611 	case E1000_FUNC_1:
2612 		if (eeprom_data & IGB_MAS_ENABLE_1) {
2613 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2614 			netdev_info(adapter->netdev,
2615 				"MAS: Enabling Media Autosense for port %d\n",
2616 				hw->bus.func);
2617 		}
2618 		break;
2619 	case E1000_FUNC_2:
2620 		if (eeprom_data & IGB_MAS_ENABLE_2) {
2621 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2622 			netdev_info(adapter->netdev,
2623 				"MAS: Enabling Media Autosense for port %d\n",
2624 				hw->bus.func);
2625 		}
2626 		break;
2627 	case E1000_FUNC_3:
2628 		if (eeprom_data & IGB_MAS_ENABLE_3) {
2629 			adapter->flags |= IGB_FLAG_MAS_ENABLE;
2630 			netdev_info(adapter->netdev,
2631 				"MAS: Enabling Media Autosense for port %d\n",
2632 				hw->bus.func);
2633 		}
2634 		break;
2635 	default:
2636 		/* Shouldn't get here */
2637 		netdev_err(adapter->netdev,
2638 			"MAS: Invalid port configuration, returning\n");
2639 		break;
2640 	}
2641 }
2642 
2643 /**
2644  *  igb_init_i2c - Init I2C interface
2645  *  @adapter: pointer to adapter structure
2646  **/
2647 static s32 igb_init_i2c(struct igb_adapter *adapter)
2648 {
2649 	s32 status = 0;
2650 
2651 	/* I2C interface supported on i350 devices */
2652 	if (adapter->hw.mac.type != e1000_i350)
2653 		return 0;
2654 
2655 	/* Initialize the i2c bus which is controlled by the registers.
2656 	 * This bus will use the i2c_algo_bit structue that implements
2657 	 * the protocol through toggling of the 4 bits in the register.
2658 	 */
2659 	adapter->i2c_adap.owner = THIS_MODULE;
2660 	adapter->i2c_algo = igb_i2c_algo;
2661 	adapter->i2c_algo.data = adapter;
2662 	adapter->i2c_adap.algo_data = &adapter->i2c_algo;
2663 	adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
2664 	strlcpy(adapter->i2c_adap.name, "igb BB",
2665 		sizeof(adapter->i2c_adap.name));
2666 	status = i2c_bit_add_bus(&adapter->i2c_adap);
2667 	return status;
2668 }
2669 
2670 /**
2671  *  igb_probe - Device Initialization Routine
2672  *  @pdev: PCI device information struct
2673  *  @ent: entry in igb_pci_tbl
2674  *
2675  *  Returns 0 on success, negative on failure
2676  *
2677  *  igb_probe initializes an adapter identified by a pci_dev structure.
2678  *  The OS initialization, configuring of the adapter private structure,
2679  *  and a hardware reset occur.
2680  **/
2681 static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2682 {
2683 	struct net_device *netdev;
2684 	struct igb_adapter *adapter;
2685 	struct e1000_hw *hw;
2686 	u16 eeprom_data = 0;
2687 	s32 ret_val;
2688 	static int global_quad_port_a; /* global quad port a indication */
2689 	const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
2690 	int err, pci_using_dac;
2691 	u8 part_str[E1000_PBANUM_LENGTH];
2692 
2693 	/* Catch broken hardware that put the wrong VF device ID in
2694 	 * the PCIe SR-IOV capability.
2695 	 */
2696 	if (pdev->is_virtfn) {
2697 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
2698 			pci_name(pdev), pdev->vendor, pdev->device);
2699 		return -EINVAL;
2700 	}
2701 
2702 	err = pci_enable_device_mem(pdev);
2703 	if (err)
2704 		return err;
2705 
2706 	pci_using_dac = 0;
2707 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2708 	if (!err) {
2709 		pci_using_dac = 1;
2710 	} else {
2711 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
2712 		if (err) {
2713 			dev_err(&pdev->dev,
2714 				"No usable DMA configuration, aborting\n");
2715 			goto err_dma;
2716 		}
2717 	}
2718 
2719 	err = pci_request_mem_regions(pdev, igb_driver_name);
2720 	if (err)
2721 		goto err_pci_reg;
2722 
2723 	pci_enable_pcie_error_reporting(pdev);
2724 
2725 	pci_set_master(pdev);
2726 	pci_save_state(pdev);
2727 
2728 	err = -ENOMEM;
2729 	netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
2730 				   IGB_MAX_TX_QUEUES);
2731 	if (!netdev)
2732 		goto err_alloc_etherdev;
2733 
2734 	SET_NETDEV_DEV(netdev, &pdev->dev);
2735 
2736 	pci_set_drvdata(pdev, netdev);
2737 	adapter = netdev_priv(netdev);
2738 	adapter->netdev = netdev;
2739 	adapter->pdev = pdev;
2740 	hw = &adapter->hw;
2741 	hw->back = adapter;
2742 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
2743 
2744 	err = -EIO;
2745 	adapter->io_addr = pci_iomap(pdev, 0, 0);
2746 	if (!adapter->io_addr)
2747 		goto err_ioremap;
2748 	/* hw->hw_addr can be altered, we'll use adapter->io_addr for unmap */
2749 	hw->hw_addr = adapter->io_addr;
2750 
2751 	netdev->netdev_ops = &igb_netdev_ops;
2752 	igb_set_ethtool_ops(netdev);
2753 	netdev->watchdog_timeo = 5 * HZ;
2754 
2755 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
2756 
2757 	netdev->mem_start = pci_resource_start(pdev, 0);
2758 	netdev->mem_end = pci_resource_end(pdev, 0);
2759 
2760 	/* PCI config space info */
2761 	hw->vendor_id = pdev->vendor;
2762 	hw->device_id = pdev->device;
2763 	hw->revision_id = pdev->revision;
2764 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
2765 	hw->subsystem_device_id = pdev->subsystem_device;
2766 
2767 	/* Copy the default MAC, PHY and NVM function pointers */
2768 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
2769 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
2770 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
2771 	/* Initialize skew-specific constants */
2772 	err = ei->get_invariants(hw);
2773 	if (err)
2774 		goto err_sw_init;
2775 
2776 	/* setup the private structure */
2777 	err = igb_sw_init(adapter);
2778 	if (err)
2779 		goto err_sw_init;
2780 
2781 	igb_get_bus_info_pcie(hw);
2782 
2783 	hw->phy.autoneg_wait_to_complete = false;
2784 
2785 	/* Copper options */
2786 	if (hw->phy.media_type == e1000_media_type_copper) {
2787 		hw->phy.mdix = AUTO_ALL_MODES;
2788 		hw->phy.disable_polarity_correction = false;
2789 		hw->phy.ms_type = e1000_ms_hw_default;
2790 	}
2791 
2792 	if (igb_check_reset_block(hw))
2793 		dev_info(&pdev->dev,
2794 			"PHY reset is blocked due to SOL/IDER session.\n");
2795 
2796 	/* features is initialized to 0 in allocation, it might have bits
2797 	 * set by igb_sw_init so we should use an or instead of an
2798 	 * assignment.
2799 	 */
2800 	netdev->features |= NETIF_F_SG |
2801 			    NETIF_F_TSO |
2802 			    NETIF_F_TSO6 |
2803 			    NETIF_F_RXHASH |
2804 			    NETIF_F_RXCSUM |
2805 			    NETIF_F_HW_CSUM;
2806 
2807 	if (hw->mac.type >= e1000_82576)
2808 		netdev->features |= NETIF_F_SCTP_CRC;
2809 
2810 #define IGB_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
2811 				  NETIF_F_GSO_GRE_CSUM | \
2812 				  NETIF_F_GSO_IPXIP4 | \
2813 				  NETIF_F_GSO_IPXIP6 | \
2814 				  NETIF_F_GSO_UDP_TUNNEL | \
2815 				  NETIF_F_GSO_UDP_TUNNEL_CSUM)
2816 
2817 	netdev->gso_partial_features = IGB_GSO_PARTIAL_FEATURES;
2818 	netdev->features |= NETIF_F_GSO_PARTIAL | IGB_GSO_PARTIAL_FEATURES;
2819 
2820 	/* copy netdev features into list of user selectable features */
2821 	netdev->hw_features |= netdev->features |
2822 			       NETIF_F_HW_VLAN_CTAG_RX |
2823 			       NETIF_F_HW_VLAN_CTAG_TX |
2824 			       NETIF_F_RXALL;
2825 
2826 	if (hw->mac.type >= e1000_i350)
2827 		netdev->hw_features |= NETIF_F_NTUPLE;
2828 
2829 	if (pci_using_dac)
2830 		netdev->features |= NETIF_F_HIGHDMA;
2831 
2832 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
2833 	netdev->mpls_features |= NETIF_F_HW_CSUM;
2834 	netdev->hw_enc_features |= netdev->vlan_features;
2835 
2836 	/* set this bit last since it cannot be part of vlan_features */
2837 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
2838 			    NETIF_F_HW_VLAN_CTAG_RX |
2839 			    NETIF_F_HW_VLAN_CTAG_TX;
2840 
2841 	netdev->priv_flags |= IFF_SUPP_NOFCS;
2842 
2843 	netdev->priv_flags |= IFF_UNICAST_FLT;
2844 
2845 	/* MTU range: 68 - 9216 */
2846 	netdev->min_mtu = ETH_MIN_MTU;
2847 	netdev->max_mtu = MAX_STD_JUMBO_FRAME_SIZE;
2848 
2849 	adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
2850 
2851 	/* before reading the NVM, reset the controller to put the device in a
2852 	 * known good starting state
2853 	 */
2854 	hw->mac.ops.reset_hw(hw);
2855 
2856 	/* make sure the NVM is good , i211/i210 parts can have special NVM
2857 	 * that doesn't contain a checksum
2858 	 */
2859 	switch (hw->mac.type) {
2860 	case e1000_i210:
2861 	case e1000_i211:
2862 		if (igb_get_flash_presence_i210(hw)) {
2863 			if (hw->nvm.ops.validate(hw) < 0) {
2864 				dev_err(&pdev->dev,
2865 					"The NVM Checksum Is Not Valid\n");
2866 				err = -EIO;
2867 				goto err_eeprom;
2868 			}
2869 		}
2870 		break;
2871 	default:
2872 		if (hw->nvm.ops.validate(hw) < 0) {
2873 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
2874 			err = -EIO;
2875 			goto err_eeprom;
2876 		}
2877 		break;
2878 	}
2879 
2880 	if (eth_platform_get_mac_address(&pdev->dev, hw->mac.addr)) {
2881 		/* copy the MAC address out of the NVM */
2882 		if (hw->mac.ops.read_mac_addr(hw))
2883 			dev_err(&pdev->dev, "NVM Read Error\n");
2884 	}
2885 
2886 	memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
2887 
2888 	if (!is_valid_ether_addr(netdev->dev_addr)) {
2889 		dev_err(&pdev->dev, "Invalid MAC Address\n");
2890 		err = -EIO;
2891 		goto err_eeprom;
2892 	}
2893 
2894 	igb_set_default_mac_filter(adapter);
2895 
2896 	/* get firmware version for ethtool -i */
2897 	igb_set_fw_version(adapter);
2898 
2899 	/* configure RXPBSIZE and TXPBSIZE */
2900 	if (hw->mac.type == e1000_i210) {
2901 		wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
2902 		wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
2903 	}
2904 
2905 	timer_setup(&adapter->watchdog_timer, igb_watchdog, 0);
2906 	timer_setup(&adapter->phy_info_timer, igb_update_phy_info, 0);
2907 
2908 	INIT_WORK(&adapter->reset_task, igb_reset_task);
2909 	INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
2910 
2911 	/* Initialize link properties that are user-changeable */
2912 	adapter->fc_autoneg = true;
2913 	hw->mac.autoneg = true;
2914 	hw->phy.autoneg_advertised = 0x2f;
2915 
2916 	hw->fc.requested_mode = e1000_fc_default;
2917 	hw->fc.current_mode = e1000_fc_default;
2918 
2919 	igb_validate_mdi_setting(hw);
2920 
2921 	/* By default, support wake on port A */
2922 	if (hw->bus.func == 0)
2923 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2924 
2925 	/* Check the NVM for wake support on non-port A ports */
2926 	if (hw->mac.type >= e1000_82580)
2927 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
2928 				 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
2929 				 &eeprom_data);
2930 	else if (hw->bus.func == 1)
2931 		hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
2932 
2933 	if (eeprom_data & IGB_EEPROM_APME)
2934 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2935 
2936 	/* now that we have the eeprom settings, apply the special cases where
2937 	 * the eeprom may be wrong or the board simply won't support wake on
2938 	 * lan on a particular port
2939 	 */
2940 	switch (pdev->device) {
2941 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
2942 		adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2943 		break;
2944 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
2945 	case E1000_DEV_ID_82576_FIBER:
2946 	case E1000_DEV_ID_82576_SERDES:
2947 		/* Wake events only supported on port A for dual fiber
2948 		 * regardless of eeprom setting
2949 		 */
2950 		if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
2951 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2952 		break;
2953 	case E1000_DEV_ID_82576_QUAD_COPPER:
2954 	case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
2955 		/* if quad port adapter, disable WoL on all but port A */
2956 		if (global_quad_port_a != 0)
2957 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2958 		else
2959 			adapter->flags |= IGB_FLAG_QUAD_PORT_A;
2960 		/* Reset for multiple quad port adapters */
2961 		if (++global_quad_port_a == 4)
2962 			global_quad_port_a = 0;
2963 		break;
2964 	default:
2965 		/* If the device can't wake, don't set software support */
2966 		if (!device_can_wakeup(&adapter->pdev->dev))
2967 			adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
2968 	}
2969 
2970 	/* initialize the wol settings based on the eeprom settings */
2971 	if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
2972 		adapter->wol |= E1000_WUFC_MAG;
2973 
2974 	/* Some vendors want WoL disabled by default, but still supported */
2975 	if ((hw->mac.type == e1000_i350) &&
2976 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
2977 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2978 		adapter->wol = 0;
2979 	}
2980 
2981 	/* Some vendors want the ability to Use the EEPROM setting as
2982 	 * enable/disable only, and not for capability
2983 	 */
2984 	if (((hw->mac.type == e1000_i350) ||
2985 	     (hw->mac.type == e1000_i354)) &&
2986 	    (pdev->subsystem_vendor == PCI_VENDOR_ID_DELL)) {
2987 		adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2988 		adapter->wol = 0;
2989 	}
2990 	if (hw->mac.type == e1000_i350) {
2991 		if (((pdev->subsystem_device == 0x5001) ||
2992 		     (pdev->subsystem_device == 0x5002)) &&
2993 				(hw->bus.func == 0)) {
2994 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2995 			adapter->wol = 0;
2996 		}
2997 		if (pdev->subsystem_device == 0x1F52)
2998 			adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
2999 	}
3000 
3001 	device_set_wakeup_enable(&adapter->pdev->dev,
3002 				 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
3003 
3004 	/* reset the hardware with the new settings */
3005 	igb_reset(adapter);
3006 
3007 	/* Init the I2C interface */
3008 	err = igb_init_i2c(adapter);
3009 	if (err) {
3010 		dev_err(&pdev->dev, "failed to init i2c interface\n");
3011 		goto err_eeprom;
3012 	}
3013 
3014 	/* let the f/w know that the h/w is now under the control of the
3015 	 * driver.
3016 	 */
3017 	igb_get_hw_control(adapter);
3018 
3019 	strcpy(netdev->name, "eth%d");
3020 	err = register_netdev(netdev);
3021 	if (err)
3022 		goto err_register;
3023 
3024 	/* carrier off reporting is important to ethtool even BEFORE open */
3025 	netif_carrier_off(netdev);
3026 
3027 #ifdef CONFIG_IGB_DCA
3028 	if (dca_add_requester(&pdev->dev) == 0) {
3029 		adapter->flags |= IGB_FLAG_DCA_ENABLED;
3030 		dev_info(&pdev->dev, "DCA enabled\n");
3031 		igb_setup_dca(adapter);
3032 	}
3033 
3034 #endif
3035 #ifdef CONFIG_IGB_HWMON
3036 	/* Initialize the thermal sensor on i350 devices. */
3037 	if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
3038 		u16 ets_word;
3039 
3040 		/* Read the NVM to determine if this i350 device supports an
3041 		 * external thermal sensor.
3042 		 */
3043 		hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
3044 		if (ets_word != 0x0000 && ets_word != 0xFFFF)
3045 			adapter->ets = true;
3046 		else
3047 			adapter->ets = false;
3048 		if (igb_sysfs_init(adapter))
3049 			dev_err(&pdev->dev,
3050 				"failed to allocate sysfs resources\n");
3051 	} else {
3052 		adapter->ets = false;
3053 	}
3054 #endif
3055 	/* Check if Media Autosense is enabled */
3056 	adapter->ei = *ei;
3057 	if (hw->dev_spec._82575.mas_capable)
3058 		igb_init_mas(adapter);
3059 
3060 	/* do hw tstamp init after resetting */
3061 	igb_ptp_init(adapter);
3062 
3063 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
3064 	/* print bus type/speed/width info, not applicable to i354 */
3065 	if (hw->mac.type != e1000_i354) {
3066 		dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
3067 			 netdev->name,
3068 			 ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
3069 			  (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
3070 			   "unknown"),
3071 			 ((hw->bus.width == e1000_bus_width_pcie_x4) ?
3072 			  "Width x4" :
3073 			  (hw->bus.width == e1000_bus_width_pcie_x2) ?
3074 			  "Width x2" :
3075 			  (hw->bus.width == e1000_bus_width_pcie_x1) ?
3076 			  "Width x1" : "unknown"), netdev->dev_addr);
3077 	}
3078 
3079 	if ((hw->mac.type >= e1000_i210 ||
3080 	     igb_get_flash_presence_i210(hw))) {
3081 		ret_val = igb_read_part_string(hw, part_str,
3082 					       E1000_PBANUM_LENGTH);
3083 	} else {
3084 		ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
3085 	}
3086 
3087 	if (ret_val)
3088 		strcpy(part_str, "Unknown");
3089 	dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
3090 	dev_info(&pdev->dev,
3091 		"Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
3092 		(adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
3093 		(adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
3094 		adapter->num_rx_queues, adapter->num_tx_queues);
3095 	if (hw->phy.media_type == e1000_media_type_copper) {
3096 		switch (hw->mac.type) {
3097 		case e1000_i350:
3098 		case e1000_i210:
3099 		case e1000_i211:
3100 			/* Enable EEE for internal copper PHY devices */
3101 			err = igb_set_eee_i350(hw, true, true);
3102 			if ((!err) &&
3103 			    (!hw->dev_spec._82575.eee_disable)) {
3104 				adapter->eee_advert =
3105 					MDIO_EEE_100TX | MDIO_EEE_1000T;
3106 				adapter->flags |= IGB_FLAG_EEE;
3107 			}
3108 			break;
3109 		case e1000_i354:
3110 			if ((rd32(E1000_CTRL_EXT) &
3111 			    E1000_CTRL_EXT_LINK_MODE_SGMII)) {
3112 				err = igb_set_eee_i354(hw, true, true);
3113 				if ((!err) &&
3114 					(!hw->dev_spec._82575.eee_disable)) {
3115 					adapter->eee_advert =
3116 					   MDIO_EEE_100TX | MDIO_EEE_1000T;
3117 					adapter->flags |= IGB_FLAG_EEE;
3118 				}
3119 			}
3120 			break;
3121 		default:
3122 			break;
3123 		}
3124 	}
3125 	pm_runtime_put_noidle(&pdev->dev);
3126 	return 0;
3127 
3128 err_register:
3129 	igb_release_hw_control(adapter);
3130 	memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
3131 err_eeprom:
3132 	if (!igb_check_reset_block(hw))
3133 		igb_reset_phy(hw);
3134 
3135 	if (hw->flash_address)
3136 		iounmap(hw->flash_address);
3137 err_sw_init:
3138 	kfree(adapter->mac_table);
3139 	kfree(adapter->shadow_vfta);
3140 	igb_clear_interrupt_scheme(adapter);
3141 #ifdef CONFIG_PCI_IOV
3142 	igb_disable_sriov(pdev);
3143 #endif
3144 	pci_iounmap(pdev, adapter->io_addr);
3145 err_ioremap:
3146 	free_netdev(netdev);
3147 err_alloc_etherdev:
3148 	pci_release_mem_regions(pdev);
3149 err_pci_reg:
3150 err_dma:
3151 	pci_disable_device(pdev);
3152 	return err;
3153 }
3154 
3155 #ifdef CONFIG_PCI_IOV
3156 static int igb_disable_sriov(struct pci_dev *pdev)
3157 {
3158 	struct net_device *netdev = pci_get_drvdata(pdev);
3159 	struct igb_adapter *adapter = netdev_priv(netdev);
3160 	struct e1000_hw *hw = &adapter->hw;
3161 
3162 	/* reclaim resources allocated to VFs */
3163 	if (adapter->vf_data) {
3164 		/* disable iov and allow time for transactions to clear */
3165 		if (pci_vfs_assigned(pdev)) {
3166 			dev_warn(&pdev->dev,
3167 				 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
3168 			return -EPERM;
3169 		} else {
3170 			pci_disable_sriov(pdev);
3171 			msleep(500);
3172 		}
3173 
3174 		kfree(adapter->vf_mac_list);
3175 		adapter->vf_mac_list = NULL;
3176 		kfree(adapter->vf_data);
3177 		adapter->vf_data = NULL;
3178 		adapter->vfs_allocated_count = 0;
3179 		wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
3180 		wrfl();
3181 		msleep(100);
3182 		dev_info(&pdev->dev, "IOV Disabled\n");
3183 
3184 		/* Re-enable DMA Coalescing flag since IOV is turned off */
3185 		adapter->flags |= IGB_FLAG_DMAC;
3186 	}
3187 
3188 	return 0;
3189 }
3190 
3191 static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
3192 {
3193 	struct net_device *netdev = pci_get_drvdata(pdev);
3194 	struct igb_adapter *adapter = netdev_priv(netdev);
3195 	int old_vfs = pci_num_vf(pdev);
3196 	struct vf_mac_filter *mac_list;
3197 	int err = 0;
3198 	int num_vf_mac_filters, i;
3199 
3200 	if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
3201 		err = -EPERM;
3202 		goto out;
3203 	}
3204 	if (!num_vfs)
3205 		goto out;
3206 
3207 	if (old_vfs) {
3208 		dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
3209 			 old_vfs, max_vfs);
3210 		adapter->vfs_allocated_count = old_vfs;
3211 	} else
3212 		adapter->vfs_allocated_count = num_vfs;
3213 
3214 	adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
3215 				sizeof(struct vf_data_storage), GFP_KERNEL);
3216 
3217 	/* if allocation failed then we do not support SR-IOV */
3218 	if (!adapter->vf_data) {
3219 		adapter->vfs_allocated_count = 0;
3220 		err = -ENOMEM;
3221 		goto out;
3222 	}
3223 
3224 	/* Due to the limited number of RAR entries calculate potential
3225 	 * number of MAC filters available for the VFs. Reserve entries
3226 	 * for PF default MAC, PF MAC filters and at least one RAR entry
3227 	 * for each VF for VF MAC.
3228 	 */
3229 	num_vf_mac_filters = adapter->hw.mac.rar_entry_count -
3230 			     (1 + IGB_PF_MAC_FILTERS_RESERVED +
3231 			      adapter->vfs_allocated_count);
3232 
3233 	adapter->vf_mac_list = kcalloc(num_vf_mac_filters,
3234 				       sizeof(struct vf_mac_filter),
3235 				       GFP_KERNEL);
3236 
3237 	mac_list = adapter->vf_mac_list;
3238 	INIT_LIST_HEAD(&adapter->vf_macs.l);
3239 
3240 	if (adapter->vf_mac_list) {
3241 		/* Initialize list of VF MAC filters */
3242 		for (i = 0; i < num_vf_mac_filters; i++) {
3243 			mac_list->vf = -1;
3244 			mac_list->free = true;
3245 			list_add(&mac_list->l, &adapter->vf_macs.l);
3246 			mac_list++;
3247 		}
3248 	} else {
3249 		/* If we could not allocate memory for the VF MAC filters
3250 		 * we can continue without this feature but warn user.
3251 		 */
3252 		dev_err(&pdev->dev,
3253 			"Unable to allocate memory for VF MAC filter list\n");
3254 	}
3255 
3256 	/* only call pci_enable_sriov() if no VFs are allocated already */
3257 	if (!old_vfs) {
3258 		err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
3259 		if (err)
3260 			goto err_out;
3261 	}
3262 	dev_info(&pdev->dev, "%d VFs allocated\n",
3263 		 adapter->vfs_allocated_count);
3264 	for (i = 0; i < adapter->vfs_allocated_count; i++)
3265 		igb_vf_configure(adapter, i);
3266 
3267 	/* DMA Coalescing is not supported in IOV mode. */
3268 	adapter->flags &= ~IGB_FLAG_DMAC;
3269 	goto out;
3270 
3271 err_out:
3272 	kfree(adapter->vf_mac_list);
3273 	adapter->vf_mac_list = NULL;
3274 	kfree(adapter->vf_data);
3275 	adapter->vf_data = NULL;
3276 	adapter->vfs_allocated_count = 0;
3277 out:
3278 	return err;
3279 }
3280 
3281 #endif
3282 /**
3283  *  igb_remove_i2c - Cleanup  I2C interface
3284  *  @adapter: pointer to adapter structure
3285  **/
3286 static void igb_remove_i2c(struct igb_adapter *adapter)
3287 {
3288 	/* free the adapter bus structure */
3289 	i2c_del_adapter(&adapter->i2c_adap);
3290 }
3291 
3292 /**
3293  *  igb_remove - Device Removal Routine
3294  *  @pdev: PCI device information struct
3295  *
3296  *  igb_remove is called by the PCI subsystem to alert the driver
3297  *  that it should release a PCI device.  The could be caused by a
3298  *  Hot-Plug event, or because the driver is going to be removed from
3299  *  memory.
3300  **/
3301 static void igb_remove(struct pci_dev *pdev)
3302 {
3303 	struct net_device *netdev = pci_get_drvdata(pdev);
3304 	struct igb_adapter *adapter = netdev_priv(netdev);
3305 	struct e1000_hw *hw = &adapter->hw;
3306 
3307 	pm_runtime_get_noresume(&pdev->dev);
3308 #ifdef CONFIG_IGB_HWMON
3309 	igb_sysfs_exit(adapter);
3310 #endif
3311 	igb_remove_i2c(adapter);
3312 	igb_ptp_stop(adapter);
3313 	/* The watchdog timer may be rescheduled, so explicitly
3314 	 * disable watchdog from being rescheduled.
3315 	 */
3316 	set_bit(__IGB_DOWN, &adapter->state);
3317 	del_timer_sync(&adapter->watchdog_timer);
3318 	del_timer_sync(&adapter->phy_info_timer);
3319 
3320 	cancel_work_sync(&adapter->reset_task);
3321 	cancel_work_sync(&adapter->watchdog_task);
3322 
3323 #ifdef CONFIG_IGB_DCA
3324 	if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
3325 		dev_info(&pdev->dev, "DCA disabled\n");
3326 		dca_remove_requester(&pdev->dev);
3327 		adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
3328 		wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
3329 	}
3330 #endif
3331 
3332 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
3333 	 * would have already happened in close and is redundant.
3334 	 */
3335 	igb_release_hw_control(adapter);
3336 
3337 #ifdef CONFIG_PCI_IOV
3338 	igb_disable_sriov(pdev);
3339 #endif
3340 
3341 	unregister_netdev(netdev);
3342 
3343 	igb_clear_interrupt_scheme(adapter);
3344 
3345 	pci_iounmap(pdev, adapter->io_addr);
3346 	if (hw->flash_address)
3347 		iounmap(hw->flash_address);
3348 	pci_release_mem_regions(pdev);
3349 
3350 	kfree(adapter->mac_table);
3351 	kfree(adapter->shadow_vfta);
3352 	free_netdev(netdev);
3353 
3354 	pci_disable_pcie_error_reporting(pdev);
3355 
3356 	pci_disable_device(pdev);
3357 }
3358 
3359 /**
3360  *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
3361  *  @adapter: board private structure to initialize
3362  *
3363  *  This function initializes the vf specific data storage and then attempts to
3364  *  allocate the VFs.  The reason for ordering it this way is because it is much
3365  *  mor expensive time wise to disable SR-IOV than it is to allocate and free
3366  *  the memory for the VFs.
3367  **/
3368 static void igb_probe_vfs(struct igb_adapter *adapter)
3369 {
3370 #ifdef CONFIG_PCI_IOV
3371 	struct pci_dev *pdev = adapter->pdev;
3372 	struct e1000_hw *hw = &adapter->hw;
3373 
3374 	/* Virtualization features not supported on i210 family. */
3375 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
3376 		return;
3377 
3378 	/* Of the below we really only want the effect of getting
3379 	 * IGB_FLAG_HAS_MSIX set (if available), without which
3380 	 * igb_enable_sriov() has no effect.
3381 	 */
3382 	igb_set_interrupt_capability(adapter, true);
3383 	igb_reset_interrupt_capability(adapter);
3384 
3385 	pci_sriov_set_totalvfs(pdev, 7);
3386 	igb_enable_sriov(pdev, max_vfs);
3387 
3388 #endif /* CONFIG_PCI_IOV */
3389 }
3390 
3391 unsigned int igb_get_max_rss_queues(struct igb_adapter *adapter)
3392 {
3393 	struct e1000_hw *hw = &adapter->hw;
3394 	unsigned int max_rss_queues;
3395 
3396 	/* Determine the maximum number of RSS queues supported. */
3397 	switch (hw->mac.type) {
3398 	case e1000_i211:
3399 		max_rss_queues = IGB_MAX_RX_QUEUES_I211;
3400 		break;
3401 	case e1000_82575:
3402 	case e1000_i210:
3403 		max_rss_queues = IGB_MAX_RX_QUEUES_82575;
3404 		break;
3405 	case e1000_i350:
3406 		/* I350 cannot do RSS and SR-IOV at the same time */
3407 		if (!!adapter->vfs_allocated_count) {
3408 			max_rss_queues = 1;
3409 			break;
3410 		}
3411 		/* fall through */
3412 	case e1000_82576:
3413 		if (!!adapter->vfs_allocated_count) {
3414 			max_rss_queues = 2;
3415 			break;
3416 		}
3417 		/* fall through */
3418 	case e1000_82580:
3419 	case e1000_i354:
3420 	default:
3421 		max_rss_queues = IGB_MAX_RX_QUEUES;
3422 		break;
3423 	}
3424 
3425 	return max_rss_queues;
3426 }
3427 
3428 static void igb_init_queue_configuration(struct igb_adapter *adapter)
3429 {
3430 	u32 max_rss_queues;
3431 
3432 	max_rss_queues = igb_get_max_rss_queues(adapter);
3433 	adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
3434 
3435 	igb_set_flag_queue_pairs(adapter, max_rss_queues);
3436 }
3437 
3438 void igb_set_flag_queue_pairs(struct igb_adapter *adapter,
3439 			      const u32 max_rss_queues)
3440 {
3441 	struct e1000_hw *hw = &adapter->hw;
3442 
3443 	/* Determine if we need to pair queues. */
3444 	switch (hw->mac.type) {
3445 	case e1000_82575:
3446 	case e1000_i211:
3447 		/* Device supports enough interrupts without queue pairing. */
3448 		break;
3449 	case e1000_82576:
3450 	case e1000_82580:
3451 	case e1000_i350:
3452 	case e1000_i354:
3453 	case e1000_i210:
3454 	default:
3455 		/* If rss_queues > half of max_rss_queues, pair the queues in
3456 		 * order to conserve interrupts due to limited supply.
3457 		 */
3458 		if (adapter->rss_queues > (max_rss_queues / 2))
3459 			adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
3460 		else
3461 			adapter->flags &= ~IGB_FLAG_QUEUE_PAIRS;
3462 		break;
3463 	}
3464 }
3465 
3466 /**
3467  *  igb_sw_init - Initialize general software structures (struct igb_adapter)
3468  *  @adapter: board private structure to initialize
3469  *
3470  *  igb_sw_init initializes the Adapter private data structure.
3471  *  Fields are initialized based on PCI device information and
3472  *  OS network device settings (MTU size).
3473  **/
3474 static int igb_sw_init(struct igb_adapter *adapter)
3475 {
3476 	struct e1000_hw *hw = &adapter->hw;
3477 	struct net_device *netdev = adapter->netdev;
3478 	struct pci_dev *pdev = adapter->pdev;
3479 
3480 	pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
3481 
3482 	/* set default ring sizes */
3483 	adapter->tx_ring_count = IGB_DEFAULT_TXD;
3484 	adapter->rx_ring_count = IGB_DEFAULT_RXD;
3485 
3486 	/* set default ITR values */
3487 	adapter->rx_itr_setting = IGB_DEFAULT_ITR;
3488 	adapter->tx_itr_setting = IGB_DEFAULT_ITR;
3489 
3490 	/* set default work limits */
3491 	adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
3492 
3493 	adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN +
3494 				  VLAN_HLEN;
3495 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
3496 
3497 	spin_lock_init(&adapter->nfc_lock);
3498 	spin_lock_init(&adapter->stats64_lock);
3499 #ifdef CONFIG_PCI_IOV
3500 	switch (hw->mac.type) {
3501 	case e1000_82576:
3502 	case e1000_i350:
3503 		if (max_vfs > 7) {
3504 			dev_warn(&pdev->dev,
3505 				 "Maximum of 7 VFs per PF, using max\n");
3506 			max_vfs = adapter->vfs_allocated_count = 7;
3507 		} else
3508 			adapter->vfs_allocated_count = max_vfs;
3509 		if (adapter->vfs_allocated_count)
3510 			dev_warn(&pdev->dev,
3511 				 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
3512 		break;
3513 	default:
3514 		break;
3515 	}
3516 #endif /* CONFIG_PCI_IOV */
3517 
3518 	/* Assume MSI-X interrupts, will be checked during IRQ allocation */
3519 	adapter->flags |= IGB_FLAG_HAS_MSIX;
3520 
3521 	adapter->mac_table = kzalloc(sizeof(struct igb_mac_addr) *
3522 				     hw->mac.rar_entry_count, GFP_ATOMIC);
3523 	if (!adapter->mac_table)
3524 		return -ENOMEM;
3525 
3526 	igb_probe_vfs(adapter);
3527 
3528 	igb_init_queue_configuration(adapter);
3529 
3530 	/* Setup and initialize a copy of the hw vlan table array */
3531 	adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
3532 				       GFP_ATOMIC);
3533 	if (!adapter->shadow_vfta)
3534 		return -ENOMEM;
3535 
3536 	/* This call may decrease the number of queues */
3537 	if (igb_init_interrupt_scheme(adapter, true)) {
3538 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
3539 		return -ENOMEM;
3540 	}
3541 
3542 	/* Explicitly disable IRQ since the NIC can be in any state. */
3543 	igb_irq_disable(adapter);
3544 
3545 	if (hw->mac.type >= e1000_i350)
3546 		adapter->flags &= ~IGB_FLAG_DMAC;
3547 
3548 	set_bit(__IGB_DOWN, &adapter->state);
3549 	return 0;
3550 }
3551 
3552 /**
3553  *  igb_open - Called when a network interface is made active
3554  *  @netdev: network interface device structure
3555  *
3556  *  Returns 0 on success, negative value on failure
3557  *
3558  *  The open entry point is called when a network interface is made
3559  *  active by the system (IFF_UP).  At this point all resources needed
3560  *  for transmit and receive operations are allocated, the interrupt
3561  *  handler is registered with the OS, the watchdog timer is started,
3562  *  and the stack is notified that the interface is ready.
3563  **/
3564 static int __igb_open(struct net_device *netdev, bool resuming)
3565 {
3566 	struct igb_adapter *adapter = netdev_priv(netdev);
3567 	struct e1000_hw *hw = &adapter->hw;
3568 	struct pci_dev *pdev = adapter->pdev;
3569 	int err;
3570 	int i;
3571 
3572 	/* disallow open during test */
3573 	if (test_bit(__IGB_TESTING, &adapter->state)) {
3574 		WARN_ON(resuming);
3575 		return -EBUSY;
3576 	}
3577 
3578 	if (!resuming)
3579 		pm_runtime_get_sync(&pdev->dev);
3580 
3581 	netif_carrier_off(netdev);
3582 
3583 	/* allocate transmit descriptors */
3584 	err = igb_setup_all_tx_resources(adapter);
3585 	if (err)
3586 		goto err_setup_tx;
3587 
3588 	/* allocate receive descriptors */
3589 	err = igb_setup_all_rx_resources(adapter);
3590 	if (err)
3591 		goto err_setup_rx;
3592 
3593 	igb_power_up_link(adapter);
3594 
3595 	/* before we allocate an interrupt, we must be ready to handle it.
3596 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3597 	 * as soon as we call pci_request_irq, so we have to setup our
3598 	 * clean_rx handler before we do so.
3599 	 */
3600 	igb_configure(adapter);
3601 
3602 	err = igb_request_irq(adapter);
3603 	if (err)
3604 		goto err_req_irq;
3605 
3606 	/* Notify the stack of the actual queue counts. */
3607 	err = netif_set_real_num_tx_queues(adapter->netdev,
3608 					   adapter->num_tx_queues);
3609 	if (err)
3610 		goto err_set_queues;
3611 
3612 	err = netif_set_real_num_rx_queues(adapter->netdev,
3613 					   adapter->num_rx_queues);
3614 	if (err)
3615 		goto err_set_queues;
3616 
3617 	/* From here on the code is the same as igb_up() */
3618 	clear_bit(__IGB_DOWN, &adapter->state);
3619 
3620 	for (i = 0; i < adapter->num_q_vectors; i++)
3621 		napi_enable(&(adapter->q_vector[i]->napi));
3622 
3623 	/* Clear any pending interrupts. */
3624 	rd32(E1000_ICR);
3625 
3626 	igb_irq_enable(adapter);
3627 
3628 	/* notify VFs that reset has been completed */
3629 	if (adapter->vfs_allocated_count) {
3630 		u32 reg_data = rd32(E1000_CTRL_EXT);
3631 
3632 		reg_data |= E1000_CTRL_EXT_PFRSTD;
3633 		wr32(E1000_CTRL_EXT, reg_data);
3634 	}
3635 
3636 	netif_tx_start_all_queues(netdev);
3637 
3638 	if (!resuming)
3639 		pm_runtime_put(&pdev->dev);
3640 
3641 	/* start the watchdog. */
3642 	hw->mac.get_link_status = 1;
3643 	schedule_work(&adapter->watchdog_task);
3644 
3645 	return 0;
3646 
3647 err_set_queues:
3648 	igb_free_irq(adapter);
3649 err_req_irq:
3650 	igb_release_hw_control(adapter);
3651 	igb_power_down_link(adapter);
3652 	igb_free_all_rx_resources(adapter);
3653 err_setup_rx:
3654 	igb_free_all_tx_resources(adapter);
3655 err_setup_tx:
3656 	igb_reset(adapter);
3657 	if (!resuming)
3658 		pm_runtime_put(&pdev->dev);
3659 
3660 	return err;
3661 }
3662 
3663 int igb_open(struct net_device *netdev)
3664 {
3665 	return __igb_open(netdev, false);
3666 }
3667 
3668 /**
3669  *  igb_close - Disables a network interface
3670  *  @netdev: network interface device structure
3671  *
3672  *  Returns 0, this is not allowed to fail
3673  *
3674  *  The close entry point is called when an interface is de-activated
3675  *  by the OS.  The hardware is still under the driver's control, but
3676  *  needs to be disabled.  A global MAC reset is issued to stop the
3677  *  hardware, and all transmit and receive resources are freed.
3678  **/
3679 static int __igb_close(struct net_device *netdev, bool suspending)
3680 {
3681 	struct igb_adapter *adapter = netdev_priv(netdev);
3682 	struct pci_dev *pdev = adapter->pdev;
3683 
3684 	WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
3685 
3686 	if (!suspending)
3687 		pm_runtime_get_sync(&pdev->dev);
3688 
3689 	igb_down(adapter);
3690 	igb_free_irq(adapter);
3691 
3692 	igb_free_all_tx_resources(adapter);
3693 	igb_free_all_rx_resources(adapter);
3694 
3695 	if (!suspending)
3696 		pm_runtime_put_sync(&pdev->dev);
3697 	return 0;
3698 }
3699 
3700 int igb_close(struct net_device *netdev)
3701 {
3702 	if (netif_device_present(netdev) || netdev->dismantle)
3703 		return __igb_close(netdev, false);
3704 	return 0;
3705 }
3706 
3707 /**
3708  *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
3709  *  @tx_ring: tx descriptor ring (for a specific queue) to setup
3710  *
3711  *  Return 0 on success, negative on failure
3712  **/
3713 int igb_setup_tx_resources(struct igb_ring *tx_ring)
3714 {
3715 	struct device *dev = tx_ring->dev;
3716 	int size;
3717 
3718 	size = sizeof(struct igb_tx_buffer) * tx_ring->count;
3719 
3720 	tx_ring->tx_buffer_info = vmalloc(size);
3721 	if (!tx_ring->tx_buffer_info)
3722 		goto err;
3723 
3724 	/* round up to nearest 4K */
3725 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
3726 	tx_ring->size = ALIGN(tx_ring->size, 4096);
3727 
3728 	tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
3729 					   &tx_ring->dma, GFP_KERNEL);
3730 	if (!tx_ring->desc)
3731 		goto err;
3732 
3733 	tx_ring->next_to_use = 0;
3734 	tx_ring->next_to_clean = 0;
3735 
3736 	return 0;
3737 
3738 err:
3739 	vfree(tx_ring->tx_buffer_info);
3740 	tx_ring->tx_buffer_info = NULL;
3741 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
3742 	return -ENOMEM;
3743 }
3744 
3745 /**
3746  *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
3747  *				 (Descriptors) for all queues
3748  *  @adapter: board private structure
3749  *
3750  *  Return 0 on success, negative on failure
3751  **/
3752 static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
3753 {
3754 	struct pci_dev *pdev = adapter->pdev;
3755 	int i, err = 0;
3756 
3757 	for (i = 0; i < adapter->num_tx_queues; i++) {
3758 		err = igb_setup_tx_resources(adapter->tx_ring[i]);
3759 		if (err) {
3760 			dev_err(&pdev->dev,
3761 				"Allocation for Tx Queue %u failed\n", i);
3762 			for (i--; i >= 0; i--)
3763 				igb_free_tx_resources(adapter->tx_ring[i]);
3764 			break;
3765 		}
3766 	}
3767 
3768 	return err;
3769 }
3770 
3771 /**
3772  *  igb_setup_tctl - configure the transmit control registers
3773  *  @adapter: Board private structure
3774  **/
3775 void igb_setup_tctl(struct igb_adapter *adapter)
3776 {
3777 	struct e1000_hw *hw = &adapter->hw;
3778 	u32 tctl;
3779 
3780 	/* disable queue 0 which is enabled by default on 82575 and 82576 */
3781 	wr32(E1000_TXDCTL(0), 0);
3782 
3783 	/* Program the Transmit Control Register */
3784 	tctl = rd32(E1000_TCTL);
3785 	tctl &= ~E1000_TCTL_CT;
3786 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
3787 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
3788 
3789 	igb_config_collision_dist(hw);
3790 
3791 	/* Enable transmits */
3792 	tctl |= E1000_TCTL_EN;
3793 
3794 	wr32(E1000_TCTL, tctl);
3795 }
3796 
3797 /**
3798  *  igb_configure_tx_ring - Configure transmit ring after Reset
3799  *  @adapter: board private structure
3800  *  @ring: tx ring to configure
3801  *
3802  *  Configure a transmit ring after a reset.
3803  **/
3804 void igb_configure_tx_ring(struct igb_adapter *adapter,
3805 			   struct igb_ring *ring)
3806 {
3807 	struct e1000_hw *hw = &adapter->hw;
3808 	u32 txdctl = 0;
3809 	u64 tdba = ring->dma;
3810 	int reg_idx = ring->reg_idx;
3811 
3812 	/* disable the queue */
3813 	wr32(E1000_TXDCTL(reg_idx), 0);
3814 	wrfl();
3815 	mdelay(10);
3816 
3817 	wr32(E1000_TDLEN(reg_idx),
3818 	     ring->count * sizeof(union e1000_adv_tx_desc));
3819 	wr32(E1000_TDBAL(reg_idx),
3820 	     tdba & 0x00000000ffffffffULL);
3821 	wr32(E1000_TDBAH(reg_idx), tdba >> 32);
3822 
3823 	ring->tail = adapter->io_addr + E1000_TDT(reg_idx);
3824 	wr32(E1000_TDH(reg_idx), 0);
3825 	writel(0, ring->tail);
3826 
3827 	txdctl |= IGB_TX_PTHRESH;
3828 	txdctl |= IGB_TX_HTHRESH << 8;
3829 	txdctl |= IGB_TX_WTHRESH << 16;
3830 
3831 	/* reinitialize tx_buffer_info */
3832 	memset(ring->tx_buffer_info, 0,
3833 	       sizeof(struct igb_tx_buffer) * ring->count);
3834 
3835 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
3836 	wr32(E1000_TXDCTL(reg_idx), txdctl);
3837 }
3838 
3839 /**
3840  *  igb_configure_tx - Configure transmit Unit after Reset
3841  *  @adapter: board private structure
3842  *
3843  *  Configure the Tx unit of the MAC after a reset.
3844  **/
3845 static void igb_configure_tx(struct igb_adapter *adapter)
3846 {
3847 	int i;
3848 
3849 	for (i = 0; i < adapter->num_tx_queues; i++)
3850 		igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
3851 }
3852 
3853 /**
3854  *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
3855  *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
3856  *
3857  *  Returns 0 on success, negative on failure
3858  **/
3859 int igb_setup_rx_resources(struct igb_ring *rx_ring)
3860 {
3861 	struct device *dev = rx_ring->dev;
3862 	int size;
3863 
3864 	size = sizeof(struct igb_rx_buffer) * rx_ring->count;
3865 
3866 	rx_ring->rx_buffer_info = vmalloc(size);
3867 	if (!rx_ring->rx_buffer_info)
3868 		goto err;
3869 
3870 	/* Round up to nearest 4K */
3871 	rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
3872 	rx_ring->size = ALIGN(rx_ring->size, 4096);
3873 
3874 	rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
3875 					   &rx_ring->dma, GFP_KERNEL);
3876 	if (!rx_ring->desc)
3877 		goto err;
3878 
3879 	rx_ring->next_to_alloc = 0;
3880 	rx_ring->next_to_clean = 0;
3881 	rx_ring->next_to_use = 0;
3882 
3883 	return 0;
3884 
3885 err:
3886 	vfree(rx_ring->rx_buffer_info);
3887 	rx_ring->rx_buffer_info = NULL;
3888 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
3889 	return -ENOMEM;
3890 }
3891 
3892 /**
3893  *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
3894  *				 (Descriptors) for all queues
3895  *  @adapter: board private structure
3896  *
3897  *  Return 0 on success, negative on failure
3898  **/
3899 static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
3900 {
3901 	struct pci_dev *pdev = adapter->pdev;
3902 	int i, err = 0;
3903 
3904 	for (i = 0; i < adapter->num_rx_queues; i++) {
3905 		err = igb_setup_rx_resources(adapter->rx_ring[i]);
3906 		if (err) {
3907 			dev_err(&pdev->dev,
3908 				"Allocation for Rx Queue %u failed\n", i);
3909 			for (i--; i >= 0; i--)
3910 				igb_free_rx_resources(adapter->rx_ring[i]);
3911 			break;
3912 		}
3913 	}
3914 
3915 	return err;
3916 }
3917 
3918 /**
3919  *  igb_setup_mrqc - configure the multiple receive queue control registers
3920  *  @adapter: Board private structure
3921  **/
3922 static void igb_setup_mrqc(struct igb_adapter *adapter)
3923 {
3924 	struct e1000_hw *hw = &adapter->hw;
3925 	u32 mrqc, rxcsum;
3926 	u32 j, num_rx_queues;
3927 	u32 rss_key[10];
3928 
3929 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3930 	for (j = 0; j < 10; j++)
3931 		wr32(E1000_RSSRK(j), rss_key[j]);
3932 
3933 	num_rx_queues = adapter->rss_queues;
3934 
3935 	switch (hw->mac.type) {
3936 	case e1000_82576:
3937 		/* 82576 supports 2 RSS queues for SR-IOV */
3938 		if (adapter->vfs_allocated_count)
3939 			num_rx_queues = 2;
3940 		break;
3941 	default:
3942 		break;
3943 	}
3944 
3945 	if (adapter->rss_indir_tbl_init != num_rx_queues) {
3946 		for (j = 0; j < IGB_RETA_SIZE; j++)
3947 			adapter->rss_indir_tbl[j] =
3948 			(j * num_rx_queues) / IGB_RETA_SIZE;
3949 		adapter->rss_indir_tbl_init = num_rx_queues;
3950 	}
3951 	igb_write_rss_indir_tbl(adapter);
3952 
3953 	/* Disable raw packet checksumming so that RSS hash is placed in
3954 	 * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
3955 	 * offloads as they are enabled by default
3956 	 */
3957 	rxcsum = rd32(E1000_RXCSUM);
3958 	rxcsum |= E1000_RXCSUM_PCSD;
3959 
3960 	if (adapter->hw.mac.type >= e1000_82576)
3961 		/* Enable Receive Checksum Offload for SCTP */
3962 		rxcsum |= E1000_RXCSUM_CRCOFL;
3963 
3964 	/* Don't need to set TUOFL or IPOFL, they default to 1 */
3965 	wr32(E1000_RXCSUM, rxcsum);
3966 
3967 	/* Generate RSS hash based on packet types, TCP/UDP
3968 	 * port numbers and/or IPv4/v6 src and dst addresses
3969 	 */
3970 	mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
3971 	       E1000_MRQC_RSS_FIELD_IPV4_TCP |
3972 	       E1000_MRQC_RSS_FIELD_IPV6 |
3973 	       E1000_MRQC_RSS_FIELD_IPV6_TCP |
3974 	       E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
3975 
3976 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
3977 		mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
3978 	if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
3979 		mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
3980 
3981 	/* If VMDq is enabled then we set the appropriate mode for that, else
3982 	 * we default to RSS so that an RSS hash is calculated per packet even
3983 	 * if we are only using one queue
3984 	 */
3985 	if (adapter->vfs_allocated_count) {
3986 		if (hw->mac.type > e1000_82575) {
3987 			/* Set the default pool for the PF's first queue */
3988 			u32 vtctl = rd32(E1000_VT_CTL);
3989 
3990 			vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
3991 				   E1000_VT_CTL_DISABLE_DEF_POOL);
3992 			vtctl |= adapter->vfs_allocated_count <<
3993 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
3994 			wr32(E1000_VT_CTL, vtctl);
3995 		}
3996 		if (adapter->rss_queues > 1)
3997 			mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_MQ;
3998 		else
3999 			mrqc |= E1000_MRQC_ENABLE_VMDQ;
4000 	} else {
4001 		if (hw->mac.type != e1000_i211)
4002 			mrqc |= E1000_MRQC_ENABLE_RSS_MQ;
4003 	}
4004 	igb_vmm_control(adapter);
4005 
4006 	wr32(E1000_MRQC, mrqc);
4007 }
4008 
4009 /**
4010  *  igb_setup_rctl - configure the receive control registers
4011  *  @adapter: Board private structure
4012  **/
4013 void igb_setup_rctl(struct igb_adapter *adapter)
4014 {
4015 	struct e1000_hw *hw = &adapter->hw;
4016 	u32 rctl;
4017 
4018 	rctl = rd32(E1000_RCTL);
4019 
4020 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
4021 	rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
4022 
4023 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
4024 		(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
4025 
4026 	/* enable stripping of CRC. It's unlikely this will break BMC
4027 	 * redirection as it did with e1000. Newer features require
4028 	 * that the HW strips the CRC.
4029 	 */
4030 	rctl |= E1000_RCTL_SECRC;
4031 
4032 	/* disable store bad packets and clear size bits. */
4033 	rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
4034 
4035 	/* enable LPE to allow for reception of jumbo frames */
4036 	rctl |= E1000_RCTL_LPE;
4037 
4038 	/* disable queue 0 to prevent tail write w/o re-config */
4039 	wr32(E1000_RXDCTL(0), 0);
4040 
4041 	/* Attention!!!  For SR-IOV PF driver operations you must enable
4042 	 * queue drop for all VF and PF queues to prevent head of line blocking
4043 	 * if an un-trusted VF does not provide descriptors to hardware.
4044 	 */
4045 	if (adapter->vfs_allocated_count) {
4046 		/* set all queue drop enable bits */
4047 		wr32(E1000_QDE, ALL_QUEUES);
4048 	}
4049 
4050 	/* This is useful for sniffing bad packets. */
4051 	if (adapter->netdev->features & NETIF_F_RXALL) {
4052 		/* UPE and MPE will be handled by normal PROMISC logic
4053 		 * in e1000e_set_rx_mode
4054 		 */
4055 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
4056 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
4057 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
4058 
4059 		rctl &= ~(E1000_RCTL_DPF | /* Allow filtered pause */
4060 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
4061 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
4062 		 * and that breaks VLANs.
4063 		 */
4064 	}
4065 
4066 	wr32(E1000_RCTL, rctl);
4067 }
4068 
4069 static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
4070 				   int vfn)
4071 {
4072 	struct e1000_hw *hw = &adapter->hw;
4073 	u32 vmolr;
4074 
4075 	if (size > MAX_JUMBO_FRAME_SIZE)
4076 		size = MAX_JUMBO_FRAME_SIZE;
4077 
4078 	vmolr = rd32(E1000_VMOLR(vfn));
4079 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
4080 	vmolr |= size | E1000_VMOLR_LPE;
4081 	wr32(E1000_VMOLR(vfn), vmolr);
4082 
4083 	return 0;
4084 }
4085 
4086 static inline void igb_set_vf_vlan_strip(struct igb_adapter *adapter,
4087 					 int vfn, bool enable)
4088 {
4089 	struct e1000_hw *hw = &adapter->hw;
4090 	u32 val, reg;
4091 
4092 	if (hw->mac.type < e1000_82576)
4093 		return;
4094 
4095 	if (hw->mac.type == e1000_i350)
4096 		reg = E1000_DVMOLR(vfn);
4097 	else
4098 		reg = E1000_VMOLR(vfn);
4099 
4100 	val = rd32(reg);
4101 	if (enable)
4102 		val |= E1000_VMOLR_STRVLAN;
4103 	else
4104 		val &= ~(E1000_VMOLR_STRVLAN);
4105 	wr32(reg, val);
4106 }
4107 
4108 static inline void igb_set_vmolr(struct igb_adapter *adapter,
4109 				 int vfn, bool aupe)
4110 {
4111 	struct e1000_hw *hw = &adapter->hw;
4112 	u32 vmolr;
4113 
4114 	/* This register exists only on 82576 and newer so if we are older then
4115 	 * we should exit and do nothing
4116 	 */
4117 	if (hw->mac.type < e1000_82576)
4118 		return;
4119 
4120 	vmolr = rd32(E1000_VMOLR(vfn));
4121 	if (aupe)
4122 		vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
4123 	else
4124 		vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
4125 
4126 	/* clear all bits that might not be set */
4127 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
4128 
4129 	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
4130 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
4131 	/* for VMDq only allow the VFs and pool 0 to accept broadcast and
4132 	 * multicast packets
4133 	 */
4134 	if (vfn <= adapter->vfs_allocated_count)
4135 		vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
4136 
4137 	wr32(E1000_VMOLR(vfn), vmolr);
4138 }
4139 
4140 /**
4141  *  igb_configure_rx_ring - Configure a receive ring after Reset
4142  *  @adapter: board private structure
4143  *  @ring: receive ring to be configured
4144  *
4145  *  Configure the Rx unit of the MAC after a reset.
4146  **/
4147 void igb_configure_rx_ring(struct igb_adapter *adapter,
4148 			   struct igb_ring *ring)
4149 {
4150 	struct e1000_hw *hw = &adapter->hw;
4151 	union e1000_adv_rx_desc *rx_desc;
4152 	u64 rdba = ring->dma;
4153 	int reg_idx = ring->reg_idx;
4154 	u32 srrctl = 0, rxdctl = 0;
4155 
4156 	/* disable the queue */
4157 	wr32(E1000_RXDCTL(reg_idx), 0);
4158 
4159 	/* Set DMA base address registers */
4160 	wr32(E1000_RDBAL(reg_idx),
4161 	     rdba & 0x00000000ffffffffULL);
4162 	wr32(E1000_RDBAH(reg_idx), rdba >> 32);
4163 	wr32(E1000_RDLEN(reg_idx),
4164 	     ring->count * sizeof(union e1000_adv_rx_desc));
4165 
4166 	/* initialize head and tail */
4167 	ring->tail = adapter->io_addr + E1000_RDT(reg_idx);
4168 	wr32(E1000_RDH(reg_idx), 0);
4169 	writel(0, ring->tail);
4170 
4171 	/* set descriptor configuration */
4172 	srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
4173 	if (ring_uses_large_buffer(ring))
4174 		srrctl |= IGB_RXBUFFER_3072 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4175 	else
4176 		srrctl |= IGB_RXBUFFER_2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4177 	srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
4178 	if (hw->mac.type >= e1000_82580)
4179 		srrctl |= E1000_SRRCTL_TIMESTAMP;
4180 	/* Only set Drop Enable if we are supporting multiple queues */
4181 	if (adapter->vfs_allocated_count || adapter->num_rx_queues > 1)
4182 		srrctl |= E1000_SRRCTL_DROP_EN;
4183 
4184 	wr32(E1000_SRRCTL(reg_idx), srrctl);
4185 
4186 	/* set filtering for VMDQ pools */
4187 	igb_set_vmolr(adapter, reg_idx & 0x7, true);
4188 
4189 	rxdctl |= IGB_RX_PTHRESH;
4190 	rxdctl |= IGB_RX_HTHRESH << 8;
4191 	rxdctl |= IGB_RX_WTHRESH << 16;
4192 
4193 	/* initialize rx_buffer_info */
4194 	memset(ring->rx_buffer_info, 0,
4195 	       sizeof(struct igb_rx_buffer) * ring->count);
4196 
4197 	/* initialize Rx descriptor 0 */
4198 	rx_desc = IGB_RX_DESC(ring, 0);
4199 	rx_desc->wb.upper.length = 0;
4200 
4201 	/* enable receive descriptor fetching */
4202 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
4203 	wr32(E1000_RXDCTL(reg_idx), rxdctl);
4204 }
4205 
4206 static void igb_set_rx_buffer_len(struct igb_adapter *adapter,
4207 				  struct igb_ring *rx_ring)
4208 {
4209 	/* set build_skb and buffer size flags */
4210 	clear_ring_build_skb_enabled(rx_ring);
4211 	clear_ring_uses_large_buffer(rx_ring);
4212 
4213 	if (adapter->flags & IGB_FLAG_RX_LEGACY)
4214 		return;
4215 
4216 	set_ring_build_skb_enabled(rx_ring);
4217 
4218 #if (PAGE_SIZE < 8192)
4219 	if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4220 		return;
4221 
4222 	set_ring_uses_large_buffer(rx_ring);
4223 #endif
4224 }
4225 
4226 /**
4227  *  igb_configure_rx - Configure receive Unit after Reset
4228  *  @adapter: board private structure
4229  *
4230  *  Configure the Rx unit of the MAC after a reset.
4231  **/
4232 static void igb_configure_rx(struct igb_adapter *adapter)
4233 {
4234 	int i;
4235 
4236 	/* set the correct pool for the PF default MAC address in entry 0 */
4237 	igb_set_default_mac_filter(adapter);
4238 
4239 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
4240 	 * the Base and Length of the Rx Descriptor Ring
4241 	 */
4242 	for (i = 0; i < adapter->num_rx_queues; i++) {
4243 		struct igb_ring *rx_ring = adapter->rx_ring[i];
4244 
4245 		igb_set_rx_buffer_len(adapter, rx_ring);
4246 		igb_configure_rx_ring(adapter, rx_ring);
4247 	}
4248 }
4249 
4250 /**
4251  *  igb_free_tx_resources - Free Tx Resources per Queue
4252  *  @tx_ring: Tx descriptor ring for a specific queue
4253  *
4254  *  Free all transmit software resources
4255  **/
4256 void igb_free_tx_resources(struct igb_ring *tx_ring)
4257 {
4258 	igb_clean_tx_ring(tx_ring);
4259 
4260 	vfree(tx_ring->tx_buffer_info);
4261 	tx_ring->tx_buffer_info = NULL;
4262 
4263 	/* if not set, then don't free */
4264 	if (!tx_ring->desc)
4265 		return;
4266 
4267 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4268 			  tx_ring->desc, tx_ring->dma);
4269 
4270 	tx_ring->desc = NULL;
4271 }
4272 
4273 /**
4274  *  igb_free_all_tx_resources - Free Tx Resources for All Queues
4275  *  @adapter: board private structure
4276  *
4277  *  Free all transmit software resources
4278  **/
4279 static void igb_free_all_tx_resources(struct igb_adapter *adapter)
4280 {
4281 	int i;
4282 
4283 	for (i = 0; i < adapter->num_tx_queues; i++)
4284 		if (adapter->tx_ring[i])
4285 			igb_free_tx_resources(adapter->tx_ring[i]);
4286 }
4287 
4288 /**
4289  *  igb_clean_tx_ring - Free Tx Buffers
4290  *  @tx_ring: ring to be cleaned
4291  **/
4292 static void igb_clean_tx_ring(struct igb_ring *tx_ring)
4293 {
4294 	u16 i = tx_ring->next_to_clean;
4295 	struct igb_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
4296 
4297 	while (i != tx_ring->next_to_use) {
4298 		union e1000_adv_tx_desc *eop_desc, *tx_desc;
4299 
4300 		/* Free all the Tx ring sk_buffs */
4301 		dev_kfree_skb_any(tx_buffer->skb);
4302 
4303 		/* unmap skb header data */
4304 		dma_unmap_single(tx_ring->dev,
4305 				 dma_unmap_addr(tx_buffer, dma),
4306 				 dma_unmap_len(tx_buffer, len),
4307 				 DMA_TO_DEVICE);
4308 
4309 		/* check for eop_desc to determine the end of the packet */
4310 		eop_desc = tx_buffer->next_to_watch;
4311 		tx_desc = IGB_TX_DESC(tx_ring, i);
4312 
4313 		/* unmap remaining buffers */
4314 		while (tx_desc != eop_desc) {
4315 			tx_buffer++;
4316 			tx_desc++;
4317 			i++;
4318 			if (unlikely(i == tx_ring->count)) {
4319 				i = 0;
4320 				tx_buffer = tx_ring->tx_buffer_info;
4321 				tx_desc = IGB_TX_DESC(tx_ring, 0);
4322 			}
4323 
4324 			/* unmap any remaining paged data */
4325 			if (dma_unmap_len(tx_buffer, len))
4326 				dma_unmap_page(tx_ring->dev,
4327 					       dma_unmap_addr(tx_buffer, dma),
4328 					       dma_unmap_len(tx_buffer, len),
4329 					       DMA_TO_DEVICE);
4330 		}
4331 
4332 		/* move us one more past the eop_desc for start of next pkt */
4333 		tx_buffer++;
4334 		i++;
4335 		if (unlikely(i == tx_ring->count)) {
4336 			i = 0;
4337 			tx_buffer = tx_ring->tx_buffer_info;
4338 		}
4339 	}
4340 
4341 	/* reset BQL for queue */
4342 	netdev_tx_reset_queue(txring_txq(tx_ring));
4343 
4344 	/* reset next_to_use and next_to_clean */
4345 	tx_ring->next_to_use = 0;
4346 	tx_ring->next_to_clean = 0;
4347 }
4348 
4349 /**
4350  *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
4351  *  @adapter: board private structure
4352  **/
4353 static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
4354 {
4355 	int i;
4356 
4357 	for (i = 0; i < adapter->num_tx_queues; i++)
4358 		if (adapter->tx_ring[i])
4359 			igb_clean_tx_ring(adapter->tx_ring[i]);
4360 }
4361 
4362 /**
4363  *  igb_free_rx_resources - Free Rx Resources
4364  *  @rx_ring: ring to clean the resources from
4365  *
4366  *  Free all receive software resources
4367  **/
4368 void igb_free_rx_resources(struct igb_ring *rx_ring)
4369 {
4370 	igb_clean_rx_ring(rx_ring);
4371 
4372 	vfree(rx_ring->rx_buffer_info);
4373 	rx_ring->rx_buffer_info = NULL;
4374 
4375 	/* if not set, then don't free */
4376 	if (!rx_ring->desc)
4377 		return;
4378 
4379 	dma_free_coherent(rx_ring->dev, rx_ring->size,
4380 			  rx_ring->desc, rx_ring->dma);
4381 
4382 	rx_ring->desc = NULL;
4383 }
4384 
4385 /**
4386  *  igb_free_all_rx_resources - Free Rx Resources for All Queues
4387  *  @adapter: board private structure
4388  *
4389  *  Free all receive software resources
4390  **/
4391 static void igb_free_all_rx_resources(struct igb_adapter *adapter)
4392 {
4393 	int i;
4394 
4395 	for (i = 0; i < adapter->num_rx_queues; i++)
4396 		if (adapter->rx_ring[i])
4397 			igb_free_rx_resources(adapter->rx_ring[i]);
4398 }
4399 
4400 /**
4401  *  igb_clean_rx_ring - Free Rx Buffers per Queue
4402  *  @rx_ring: ring to free buffers from
4403  **/
4404 static void igb_clean_rx_ring(struct igb_ring *rx_ring)
4405 {
4406 	u16 i = rx_ring->next_to_clean;
4407 
4408 	if (rx_ring->skb)
4409 		dev_kfree_skb(rx_ring->skb);
4410 	rx_ring->skb = NULL;
4411 
4412 	/* Free all the Rx ring sk_buffs */
4413 	while (i != rx_ring->next_to_alloc) {
4414 		struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
4415 
4416 		/* Invalidate cache lines that may have been written to by
4417 		 * device so that we avoid corrupting memory.
4418 		 */
4419 		dma_sync_single_range_for_cpu(rx_ring->dev,
4420 					      buffer_info->dma,
4421 					      buffer_info->page_offset,
4422 					      igb_rx_bufsz(rx_ring),
4423 					      DMA_FROM_DEVICE);
4424 
4425 		/* free resources associated with mapping */
4426 		dma_unmap_page_attrs(rx_ring->dev,
4427 				     buffer_info->dma,
4428 				     igb_rx_pg_size(rx_ring),
4429 				     DMA_FROM_DEVICE,
4430 				     IGB_RX_DMA_ATTR);
4431 		__page_frag_cache_drain(buffer_info->page,
4432 					buffer_info->pagecnt_bias);
4433 
4434 		i++;
4435 		if (i == rx_ring->count)
4436 			i = 0;
4437 	}
4438 
4439 	rx_ring->next_to_alloc = 0;
4440 	rx_ring->next_to_clean = 0;
4441 	rx_ring->next_to_use = 0;
4442 }
4443 
4444 /**
4445  *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
4446  *  @adapter: board private structure
4447  **/
4448 static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
4449 {
4450 	int i;
4451 
4452 	for (i = 0; i < adapter->num_rx_queues; i++)
4453 		if (adapter->rx_ring[i])
4454 			igb_clean_rx_ring(adapter->rx_ring[i]);
4455 }
4456 
4457 /**
4458  *  igb_set_mac - Change the Ethernet Address of the NIC
4459  *  @netdev: network interface device structure
4460  *  @p: pointer to an address structure
4461  *
4462  *  Returns 0 on success, negative on failure
4463  **/
4464 static int igb_set_mac(struct net_device *netdev, void *p)
4465 {
4466 	struct igb_adapter *adapter = netdev_priv(netdev);
4467 	struct e1000_hw *hw = &adapter->hw;
4468 	struct sockaddr *addr = p;
4469 
4470 	if (!is_valid_ether_addr(addr->sa_data))
4471 		return -EADDRNOTAVAIL;
4472 
4473 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4474 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
4475 
4476 	/* set the correct pool for the new PF MAC address in entry 0 */
4477 	igb_set_default_mac_filter(adapter);
4478 
4479 	return 0;
4480 }
4481 
4482 /**
4483  *  igb_write_mc_addr_list - write multicast addresses to MTA
4484  *  @netdev: network interface device structure
4485  *
4486  *  Writes multicast address list to the MTA hash table.
4487  *  Returns: -ENOMEM on failure
4488  *           0 on no addresses written
4489  *           X on writing X addresses to MTA
4490  **/
4491 static int igb_write_mc_addr_list(struct net_device *netdev)
4492 {
4493 	struct igb_adapter *adapter = netdev_priv(netdev);
4494 	struct e1000_hw *hw = &adapter->hw;
4495 	struct netdev_hw_addr *ha;
4496 	u8  *mta_list;
4497 	int i;
4498 
4499 	if (netdev_mc_empty(netdev)) {
4500 		/* nothing to program, so clear mc list */
4501 		igb_update_mc_addr_list(hw, NULL, 0);
4502 		igb_restore_vf_multicasts(adapter);
4503 		return 0;
4504 	}
4505 
4506 	mta_list = kzalloc(netdev_mc_count(netdev) * 6, GFP_ATOMIC);
4507 	if (!mta_list)
4508 		return -ENOMEM;
4509 
4510 	/* The shared function expects a packed array of only addresses. */
4511 	i = 0;
4512 	netdev_for_each_mc_addr(ha, netdev)
4513 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
4514 
4515 	igb_update_mc_addr_list(hw, mta_list, i);
4516 	kfree(mta_list);
4517 
4518 	return netdev_mc_count(netdev);
4519 }
4520 
4521 static int igb_vlan_promisc_enable(struct igb_adapter *adapter)
4522 {
4523 	struct e1000_hw *hw = &adapter->hw;
4524 	u32 i, pf_id;
4525 
4526 	switch (hw->mac.type) {
4527 	case e1000_i210:
4528 	case e1000_i211:
4529 	case e1000_i350:
4530 		/* VLAN filtering needed for VLAN prio filter */
4531 		if (adapter->netdev->features & NETIF_F_NTUPLE)
4532 			break;
4533 		/* fall through */
4534 	case e1000_82576:
4535 	case e1000_82580:
4536 	case e1000_i354:
4537 		/* VLAN filtering needed for pool filtering */
4538 		if (adapter->vfs_allocated_count)
4539 			break;
4540 		/* fall through */
4541 	default:
4542 		return 1;
4543 	}
4544 
4545 	/* We are already in VLAN promisc, nothing to do */
4546 	if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
4547 		return 0;
4548 
4549 	if (!adapter->vfs_allocated_count)
4550 		goto set_vfta;
4551 
4552 	/* Add PF to all active pools */
4553 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
4554 
4555 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
4556 		u32 vlvf = rd32(E1000_VLVF(i));
4557 
4558 		vlvf |= BIT(pf_id);
4559 		wr32(E1000_VLVF(i), vlvf);
4560 	}
4561 
4562 set_vfta:
4563 	/* Set all bits in the VLAN filter table array */
4564 	for (i = E1000_VLAN_FILTER_TBL_SIZE; i--;)
4565 		hw->mac.ops.write_vfta(hw, i, ~0U);
4566 
4567 	/* Set flag so we don't redo unnecessary work */
4568 	adapter->flags |= IGB_FLAG_VLAN_PROMISC;
4569 
4570 	return 0;
4571 }
4572 
4573 #define VFTA_BLOCK_SIZE 8
4574 static void igb_scrub_vfta(struct igb_adapter *adapter, u32 vfta_offset)
4575 {
4576 	struct e1000_hw *hw = &adapter->hw;
4577 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4578 	u32 vid_start = vfta_offset * 32;
4579 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4580 	u32 i, vid, word, bits, pf_id;
4581 
4582 	/* guarantee that we don't scrub out management VLAN */
4583 	vid = adapter->mng_vlan_id;
4584 	if (vid >= vid_start && vid < vid_end)
4585 		vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4586 
4587 	if (!adapter->vfs_allocated_count)
4588 		goto set_vfta;
4589 
4590 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
4591 
4592 	for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
4593 		u32 vlvf = rd32(E1000_VLVF(i));
4594 
4595 		/* pull VLAN ID from VLVF */
4596 		vid = vlvf & VLAN_VID_MASK;
4597 
4598 		/* only concern ourselves with a certain range */
4599 		if (vid < vid_start || vid >= vid_end)
4600 			continue;
4601 
4602 		if (vlvf & E1000_VLVF_VLANID_ENABLE) {
4603 			/* record VLAN ID in VFTA */
4604 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4605 
4606 			/* if PF is part of this then continue */
4607 			if (test_bit(vid, adapter->active_vlans))
4608 				continue;
4609 		}
4610 
4611 		/* remove PF from the pool */
4612 		bits = ~BIT(pf_id);
4613 		bits &= rd32(E1000_VLVF(i));
4614 		wr32(E1000_VLVF(i), bits);
4615 	}
4616 
4617 set_vfta:
4618 	/* extract values from active_vlans and write back to VFTA */
4619 	for (i = VFTA_BLOCK_SIZE; i--;) {
4620 		vid = (vfta_offset + i) * 32;
4621 		word = vid / BITS_PER_LONG;
4622 		bits = vid % BITS_PER_LONG;
4623 
4624 		vfta[i] |= adapter->active_vlans[word] >> bits;
4625 
4626 		hw->mac.ops.write_vfta(hw, vfta_offset + i, vfta[i]);
4627 	}
4628 }
4629 
4630 static void igb_vlan_promisc_disable(struct igb_adapter *adapter)
4631 {
4632 	u32 i;
4633 
4634 	/* We are not in VLAN promisc, nothing to do */
4635 	if (!(adapter->flags & IGB_FLAG_VLAN_PROMISC))
4636 		return;
4637 
4638 	/* Set flag so we don't redo unnecessary work */
4639 	adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
4640 
4641 	for (i = 0; i < E1000_VLAN_FILTER_TBL_SIZE; i += VFTA_BLOCK_SIZE)
4642 		igb_scrub_vfta(adapter, i);
4643 }
4644 
4645 /**
4646  *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
4647  *  @netdev: network interface device structure
4648  *
4649  *  The set_rx_mode entry point is called whenever the unicast or multicast
4650  *  address lists or the network interface flags are updated.  This routine is
4651  *  responsible for configuring the hardware for proper unicast, multicast,
4652  *  promiscuous mode, and all-multi behavior.
4653  **/
4654 static void igb_set_rx_mode(struct net_device *netdev)
4655 {
4656 	struct igb_adapter *adapter = netdev_priv(netdev);
4657 	struct e1000_hw *hw = &adapter->hw;
4658 	unsigned int vfn = adapter->vfs_allocated_count;
4659 	u32 rctl = 0, vmolr = 0, rlpml = MAX_JUMBO_FRAME_SIZE;
4660 	int count;
4661 
4662 	/* Check for Promiscuous and All Multicast modes */
4663 	if (netdev->flags & IFF_PROMISC) {
4664 		rctl |= E1000_RCTL_UPE | E1000_RCTL_MPE;
4665 		vmolr |= E1000_VMOLR_MPME;
4666 
4667 		/* enable use of UTA filter to force packets to default pool */
4668 		if (hw->mac.type == e1000_82576)
4669 			vmolr |= E1000_VMOLR_ROPE;
4670 	} else {
4671 		if (netdev->flags & IFF_ALLMULTI) {
4672 			rctl |= E1000_RCTL_MPE;
4673 			vmolr |= E1000_VMOLR_MPME;
4674 		} else {
4675 			/* Write addresses to the MTA, if the attempt fails
4676 			 * then we should just turn on promiscuous mode so
4677 			 * that we can at least receive multicast traffic
4678 			 */
4679 			count = igb_write_mc_addr_list(netdev);
4680 			if (count < 0) {
4681 				rctl |= E1000_RCTL_MPE;
4682 				vmolr |= E1000_VMOLR_MPME;
4683 			} else if (count) {
4684 				vmolr |= E1000_VMOLR_ROMPE;
4685 			}
4686 		}
4687 	}
4688 
4689 	/* Write addresses to available RAR registers, if there is not
4690 	 * sufficient space to store all the addresses then enable
4691 	 * unicast promiscuous mode
4692 	 */
4693 	if (__dev_uc_sync(netdev, igb_uc_sync, igb_uc_unsync)) {
4694 		rctl |= E1000_RCTL_UPE;
4695 		vmolr |= E1000_VMOLR_ROPE;
4696 	}
4697 
4698 	/* enable VLAN filtering by default */
4699 	rctl |= E1000_RCTL_VFE;
4700 
4701 	/* disable VLAN filtering for modes that require it */
4702 	if ((netdev->flags & IFF_PROMISC) ||
4703 	    (netdev->features & NETIF_F_RXALL)) {
4704 		/* if we fail to set all rules then just clear VFE */
4705 		if (igb_vlan_promisc_enable(adapter))
4706 			rctl &= ~E1000_RCTL_VFE;
4707 	} else {
4708 		igb_vlan_promisc_disable(adapter);
4709 	}
4710 
4711 	/* update state of unicast, multicast, and VLAN filtering modes */
4712 	rctl |= rd32(E1000_RCTL) & ~(E1000_RCTL_UPE | E1000_RCTL_MPE |
4713 				     E1000_RCTL_VFE);
4714 	wr32(E1000_RCTL, rctl);
4715 
4716 #if (PAGE_SIZE < 8192)
4717 	if (!adapter->vfs_allocated_count) {
4718 		if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4719 			rlpml = IGB_MAX_FRAME_BUILD_SKB;
4720 	}
4721 #endif
4722 	wr32(E1000_RLPML, rlpml);
4723 
4724 	/* In order to support SR-IOV and eventually VMDq it is necessary to set
4725 	 * the VMOLR to enable the appropriate modes.  Without this workaround
4726 	 * we will have issues with VLAN tag stripping not being done for frames
4727 	 * that are only arriving because we are the default pool
4728 	 */
4729 	if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
4730 		return;
4731 
4732 	/* set UTA to appropriate mode */
4733 	igb_set_uta(adapter, !!(vmolr & E1000_VMOLR_ROPE));
4734 
4735 	vmolr |= rd32(E1000_VMOLR(vfn)) &
4736 		 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
4737 
4738 	/* enable Rx jumbo frames, restrict as needed to support build_skb */
4739 	vmolr &= ~E1000_VMOLR_RLPML_MASK;
4740 #if (PAGE_SIZE < 8192)
4741 	if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4742 		vmolr |= IGB_MAX_FRAME_BUILD_SKB;
4743 	else
4744 #endif
4745 		vmolr |= MAX_JUMBO_FRAME_SIZE;
4746 	vmolr |= E1000_VMOLR_LPE;
4747 
4748 	wr32(E1000_VMOLR(vfn), vmolr);
4749 
4750 	igb_restore_vf_multicasts(adapter);
4751 }
4752 
4753 static void igb_check_wvbr(struct igb_adapter *adapter)
4754 {
4755 	struct e1000_hw *hw = &adapter->hw;
4756 	u32 wvbr = 0;
4757 
4758 	switch (hw->mac.type) {
4759 	case e1000_82576:
4760 	case e1000_i350:
4761 		wvbr = rd32(E1000_WVBR);
4762 		if (!wvbr)
4763 			return;
4764 		break;
4765 	default:
4766 		break;
4767 	}
4768 
4769 	adapter->wvbr |= wvbr;
4770 }
4771 
4772 #define IGB_STAGGERED_QUEUE_OFFSET 8
4773 
4774 static void igb_spoof_check(struct igb_adapter *adapter)
4775 {
4776 	int j;
4777 
4778 	if (!adapter->wvbr)
4779 		return;
4780 
4781 	for (j = 0; j < adapter->vfs_allocated_count; j++) {
4782 		if (adapter->wvbr & BIT(j) ||
4783 		    adapter->wvbr & BIT(j + IGB_STAGGERED_QUEUE_OFFSET)) {
4784 			dev_warn(&adapter->pdev->dev,
4785 				"Spoof event(s) detected on VF %d\n", j);
4786 			adapter->wvbr &=
4787 				~(BIT(j) |
4788 				  BIT(j + IGB_STAGGERED_QUEUE_OFFSET));
4789 		}
4790 	}
4791 }
4792 
4793 /* Need to wait a few seconds after link up to get diagnostic information from
4794  * the phy
4795  */
4796 static void igb_update_phy_info(struct timer_list *t)
4797 {
4798 	struct igb_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4799 	igb_get_phy_info(&adapter->hw);
4800 }
4801 
4802 /**
4803  *  igb_has_link - check shared code for link and determine up/down
4804  *  @adapter: pointer to driver private info
4805  **/
4806 bool igb_has_link(struct igb_adapter *adapter)
4807 {
4808 	struct e1000_hw *hw = &adapter->hw;
4809 	bool link_active = false;
4810 
4811 	/* get_link_status is set on LSC (link status) interrupt or
4812 	 * rx sequence error interrupt.  get_link_status will stay
4813 	 * false until the e1000_check_for_link establishes link
4814 	 * for copper adapters ONLY
4815 	 */
4816 	switch (hw->phy.media_type) {
4817 	case e1000_media_type_copper:
4818 		if (!hw->mac.get_link_status)
4819 			return true;
4820 	case e1000_media_type_internal_serdes:
4821 		hw->mac.ops.check_for_link(hw);
4822 		link_active = !hw->mac.get_link_status;
4823 		break;
4824 	default:
4825 	case e1000_media_type_unknown:
4826 		break;
4827 	}
4828 
4829 	if (((hw->mac.type == e1000_i210) ||
4830 	     (hw->mac.type == e1000_i211)) &&
4831 	     (hw->phy.id == I210_I_PHY_ID)) {
4832 		if (!netif_carrier_ok(adapter->netdev)) {
4833 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4834 		} else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
4835 			adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
4836 			adapter->link_check_timeout = jiffies;
4837 		}
4838 	}
4839 
4840 	return link_active;
4841 }
4842 
4843 static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
4844 {
4845 	bool ret = false;
4846 	u32 ctrl_ext, thstat;
4847 
4848 	/* check for thermal sensor event on i350 copper only */
4849 	if (hw->mac.type == e1000_i350) {
4850 		thstat = rd32(E1000_THSTAT);
4851 		ctrl_ext = rd32(E1000_CTRL_EXT);
4852 
4853 		if ((hw->phy.media_type == e1000_media_type_copper) &&
4854 		    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
4855 			ret = !!(thstat & event);
4856 	}
4857 
4858 	return ret;
4859 }
4860 
4861 /**
4862  *  igb_check_lvmmc - check for malformed packets received
4863  *  and indicated in LVMMC register
4864  *  @adapter: pointer to adapter
4865  **/
4866 static void igb_check_lvmmc(struct igb_adapter *adapter)
4867 {
4868 	struct e1000_hw *hw = &adapter->hw;
4869 	u32 lvmmc;
4870 
4871 	lvmmc = rd32(E1000_LVMMC);
4872 	if (lvmmc) {
4873 		if (unlikely(net_ratelimit())) {
4874 			netdev_warn(adapter->netdev,
4875 				    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
4876 				    lvmmc);
4877 		}
4878 	}
4879 }
4880 
4881 /**
4882  *  igb_watchdog - Timer Call-back
4883  *  @data: pointer to adapter cast into an unsigned long
4884  **/
4885 static void igb_watchdog(struct timer_list *t)
4886 {
4887 	struct igb_adapter *adapter = from_timer(adapter, t, watchdog_timer);
4888 	/* Do the rest outside of interrupt context */
4889 	schedule_work(&adapter->watchdog_task);
4890 }
4891 
4892 static void igb_watchdog_task(struct work_struct *work)
4893 {
4894 	struct igb_adapter *adapter = container_of(work,
4895 						   struct igb_adapter,
4896 						   watchdog_task);
4897 	struct e1000_hw *hw = &adapter->hw;
4898 	struct e1000_phy_info *phy = &hw->phy;
4899 	struct net_device *netdev = adapter->netdev;
4900 	u32 link;
4901 	int i;
4902 	u32 connsw;
4903 	u16 phy_data, retry_count = 20;
4904 
4905 	link = igb_has_link(adapter);
4906 
4907 	if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
4908 		if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
4909 			adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
4910 		else
4911 			link = false;
4912 	}
4913 
4914 	/* Force link down if we have fiber to swap to */
4915 	if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
4916 		if (hw->phy.media_type == e1000_media_type_copper) {
4917 			connsw = rd32(E1000_CONNSW);
4918 			if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
4919 				link = 0;
4920 		}
4921 	}
4922 	if (link) {
4923 		/* Perform a reset if the media type changed. */
4924 		if (hw->dev_spec._82575.media_changed) {
4925 			hw->dev_spec._82575.media_changed = false;
4926 			adapter->flags |= IGB_FLAG_MEDIA_RESET;
4927 			igb_reset(adapter);
4928 		}
4929 		/* Cancel scheduled suspend requests. */
4930 		pm_runtime_resume(netdev->dev.parent);
4931 
4932 		if (!netif_carrier_ok(netdev)) {
4933 			u32 ctrl;
4934 
4935 			hw->mac.ops.get_speed_and_duplex(hw,
4936 							 &adapter->link_speed,
4937 							 &adapter->link_duplex);
4938 
4939 			ctrl = rd32(E1000_CTRL);
4940 			/* Links status message must follow this format */
4941 			netdev_info(netdev,
4942 			       "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4943 			       netdev->name,
4944 			       adapter->link_speed,
4945 			       adapter->link_duplex == FULL_DUPLEX ?
4946 			       "Full" : "Half",
4947 			       (ctrl & E1000_CTRL_TFCE) &&
4948 			       (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
4949 			       (ctrl & E1000_CTRL_RFCE) ?  "RX" :
4950 			       (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
4951 
4952 			/* disable EEE if enabled */
4953 			if ((adapter->flags & IGB_FLAG_EEE) &&
4954 				(adapter->link_duplex == HALF_DUPLEX)) {
4955 				dev_info(&adapter->pdev->dev,
4956 				"EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
4957 				adapter->hw.dev_spec._82575.eee_disable = true;
4958 				adapter->flags &= ~IGB_FLAG_EEE;
4959 			}
4960 
4961 			/* check if SmartSpeed worked */
4962 			igb_check_downshift(hw);
4963 			if (phy->speed_downgraded)
4964 				netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
4965 
4966 			/* check for thermal sensor event */
4967 			if (igb_thermal_sensor_event(hw,
4968 			    E1000_THSTAT_LINK_THROTTLE))
4969 				netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
4970 
4971 			/* adjust timeout factor according to speed/duplex */
4972 			adapter->tx_timeout_factor = 1;
4973 			switch (adapter->link_speed) {
4974 			case SPEED_10:
4975 				adapter->tx_timeout_factor = 14;
4976 				break;
4977 			case SPEED_100:
4978 				/* maybe add some timeout factor ? */
4979 				break;
4980 			}
4981 
4982 			if (adapter->link_speed != SPEED_1000)
4983 				goto no_wait;
4984 
4985 			/* wait for Remote receiver status OK */
4986 retry_read_status:
4987 			if (!igb_read_phy_reg(hw, PHY_1000T_STATUS,
4988 					      &phy_data)) {
4989 				if (!(phy_data & SR_1000T_REMOTE_RX_STATUS) &&
4990 				    retry_count) {
4991 					msleep(100);
4992 					retry_count--;
4993 					goto retry_read_status;
4994 				} else if (!retry_count) {
4995 					dev_err(&adapter->pdev->dev, "exceed max 2 second\n");
4996 				}
4997 			} else {
4998 				dev_err(&adapter->pdev->dev, "read 1000Base-T Status Reg\n");
4999 			}
5000 no_wait:
5001 			netif_carrier_on(netdev);
5002 
5003 			igb_ping_all_vfs(adapter);
5004 			igb_check_vf_rate_limit(adapter);
5005 
5006 			/* link state has changed, schedule phy info update */
5007 			if (!test_bit(__IGB_DOWN, &adapter->state))
5008 				mod_timer(&adapter->phy_info_timer,
5009 					  round_jiffies(jiffies + 2 * HZ));
5010 		}
5011 	} else {
5012 		if (netif_carrier_ok(netdev)) {
5013 			adapter->link_speed = 0;
5014 			adapter->link_duplex = 0;
5015 
5016 			/* check for thermal sensor event */
5017 			if (igb_thermal_sensor_event(hw,
5018 			    E1000_THSTAT_PWR_DOWN)) {
5019 				netdev_err(netdev, "The network adapter was stopped because it overheated\n");
5020 			}
5021 
5022 			/* Links status message must follow this format */
5023 			netdev_info(netdev, "igb: %s NIC Link is Down\n",
5024 			       netdev->name);
5025 			netif_carrier_off(netdev);
5026 
5027 			igb_ping_all_vfs(adapter);
5028 
5029 			/* link state has changed, schedule phy info update */
5030 			if (!test_bit(__IGB_DOWN, &adapter->state))
5031 				mod_timer(&adapter->phy_info_timer,
5032 					  round_jiffies(jiffies + 2 * HZ));
5033 
5034 			/* link is down, time to check for alternate media */
5035 			if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5036 				igb_check_swap_media(adapter);
5037 				if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5038 					schedule_work(&adapter->reset_task);
5039 					/* return immediately */
5040 					return;
5041 				}
5042 			}
5043 			pm_schedule_suspend(netdev->dev.parent,
5044 					    MSEC_PER_SEC * 5);
5045 
5046 		/* also check for alternate media here */
5047 		} else if (!netif_carrier_ok(netdev) &&
5048 			   (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
5049 			igb_check_swap_media(adapter);
5050 			if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5051 				schedule_work(&adapter->reset_task);
5052 				/* return immediately */
5053 				return;
5054 			}
5055 		}
5056 	}
5057 
5058 	spin_lock(&adapter->stats64_lock);
5059 	igb_update_stats(adapter);
5060 	spin_unlock(&adapter->stats64_lock);
5061 
5062 	for (i = 0; i < adapter->num_tx_queues; i++) {
5063 		struct igb_ring *tx_ring = adapter->tx_ring[i];
5064 		if (!netif_carrier_ok(netdev)) {
5065 			/* We've lost link, so the controller stops DMA,
5066 			 * but we've got queued Tx work that's never going
5067 			 * to get done, so reset controller to flush Tx.
5068 			 * (Do the reset outside of interrupt context).
5069 			 */
5070 			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
5071 				adapter->tx_timeout_count++;
5072 				schedule_work(&adapter->reset_task);
5073 				/* return immediately since reset is imminent */
5074 				return;
5075 			}
5076 		}
5077 
5078 		/* Force detection of hung controller every watchdog period */
5079 		set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
5080 	}
5081 
5082 	/* Cause software interrupt to ensure Rx ring is cleaned */
5083 	if (adapter->flags & IGB_FLAG_HAS_MSIX) {
5084 		u32 eics = 0;
5085 
5086 		for (i = 0; i < adapter->num_q_vectors; i++)
5087 			eics |= adapter->q_vector[i]->eims_value;
5088 		wr32(E1000_EICS, eics);
5089 	} else {
5090 		wr32(E1000_ICS, E1000_ICS_RXDMT0);
5091 	}
5092 
5093 	igb_spoof_check(adapter);
5094 	igb_ptp_rx_hang(adapter);
5095 	igb_ptp_tx_hang(adapter);
5096 
5097 	/* Check LVMMC register on i350/i354 only */
5098 	if ((adapter->hw.mac.type == e1000_i350) ||
5099 	    (adapter->hw.mac.type == e1000_i354))
5100 		igb_check_lvmmc(adapter);
5101 
5102 	/* Reset the timer */
5103 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
5104 		if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
5105 			mod_timer(&adapter->watchdog_timer,
5106 				  round_jiffies(jiffies +  HZ));
5107 		else
5108 			mod_timer(&adapter->watchdog_timer,
5109 				  round_jiffies(jiffies + 2 * HZ));
5110 	}
5111 }
5112 
5113 enum latency_range {
5114 	lowest_latency = 0,
5115 	low_latency = 1,
5116 	bulk_latency = 2,
5117 	latency_invalid = 255
5118 };
5119 
5120 /**
5121  *  igb_update_ring_itr - update the dynamic ITR value based on packet size
5122  *  @q_vector: pointer to q_vector
5123  *
5124  *  Stores a new ITR value based on strictly on packet size.  This
5125  *  algorithm is less sophisticated than that used in igb_update_itr,
5126  *  due to the difficulty of synchronizing statistics across multiple
5127  *  receive rings.  The divisors and thresholds used by this function
5128  *  were determined based on theoretical maximum wire speed and testing
5129  *  data, in order to minimize response time while increasing bulk
5130  *  throughput.
5131  *  This functionality is controlled by ethtool's coalescing settings.
5132  *  NOTE:  This function is called only when operating in a multiqueue
5133  *         receive environment.
5134  **/
5135 static void igb_update_ring_itr(struct igb_q_vector *q_vector)
5136 {
5137 	int new_val = q_vector->itr_val;
5138 	int avg_wire_size = 0;
5139 	struct igb_adapter *adapter = q_vector->adapter;
5140 	unsigned int packets;
5141 
5142 	/* For non-gigabit speeds, just fix the interrupt rate at 4000
5143 	 * ints/sec - ITR timer value of 120 ticks.
5144 	 */
5145 	if (adapter->link_speed != SPEED_1000) {
5146 		new_val = IGB_4K_ITR;
5147 		goto set_itr_val;
5148 	}
5149 
5150 	packets = q_vector->rx.total_packets;
5151 	if (packets)
5152 		avg_wire_size = q_vector->rx.total_bytes / packets;
5153 
5154 	packets = q_vector->tx.total_packets;
5155 	if (packets)
5156 		avg_wire_size = max_t(u32, avg_wire_size,
5157 				      q_vector->tx.total_bytes / packets);
5158 
5159 	/* if avg_wire_size isn't set no work was done */
5160 	if (!avg_wire_size)
5161 		goto clear_counts;
5162 
5163 	/* Add 24 bytes to size to account for CRC, preamble, and gap */
5164 	avg_wire_size += 24;
5165 
5166 	/* Don't starve jumbo frames */
5167 	avg_wire_size = min(avg_wire_size, 3000);
5168 
5169 	/* Give a little boost to mid-size frames */
5170 	if ((avg_wire_size > 300) && (avg_wire_size < 1200))
5171 		new_val = avg_wire_size / 3;
5172 	else
5173 		new_val = avg_wire_size / 2;
5174 
5175 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5176 	if (new_val < IGB_20K_ITR &&
5177 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5178 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5179 		new_val = IGB_20K_ITR;
5180 
5181 set_itr_val:
5182 	if (new_val != q_vector->itr_val) {
5183 		q_vector->itr_val = new_val;
5184 		q_vector->set_itr = 1;
5185 	}
5186 clear_counts:
5187 	q_vector->rx.total_bytes = 0;
5188 	q_vector->rx.total_packets = 0;
5189 	q_vector->tx.total_bytes = 0;
5190 	q_vector->tx.total_packets = 0;
5191 }
5192 
5193 /**
5194  *  igb_update_itr - update the dynamic ITR value based on statistics
5195  *  @q_vector: pointer to q_vector
5196  *  @ring_container: ring info to update the itr for
5197  *
5198  *  Stores a new ITR value based on packets and byte
5199  *  counts during the last interrupt.  The advantage of per interrupt
5200  *  computation is faster updates and more accurate ITR for the current
5201  *  traffic pattern.  Constants in this function were computed
5202  *  based on theoretical maximum wire speed and thresholds were set based
5203  *  on testing data as well as attempting to minimize response time
5204  *  while increasing bulk throughput.
5205  *  This functionality is controlled by ethtool's coalescing settings.
5206  *  NOTE:  These calculations are only valid when operating in a single-
5207  *         queue environment.
5208  **/
5209 static void igb_update_itr(struct igb_q_vector *q_vector,
5210 			   struct igb_ring_container *ring_container)
5211 {
5212 	unsigned int packets = ring_container->total_packets;
5213 	unsigned int bytes = ring_container->total_bytes;
5214 	u8 itrval = ring_container->itr;
5215 
5216 	/* no packets, exit with status unchanged */
5217 	if (packets == 0)
5218 		return;
5219 
5220 	switch (itrval) {
5221 	case lowest_latency:
5222 		/* handle TSO and jumbo frames */
5223 		if (bytes/packets > 8000)
5224 			itrval = bulk_latency;
5225 		else if ((packets < 5) && (bytes > 512))
5226 			itrval = low_latency;
5227 		break;
5228 	case low_latency:  /* 50 usec aka 20000 ints/s */
5229 		if (bytes > 10000) {
5230 			/* this if handles the TSO accounting */
5231 			if (bytes/packets > 8000)
5232 				itrval = bulk_latency;
5233 			else if ((packets < 10) || ((bytes/packets) > 1200))
5234 				itrval = bulk_latency;
5235 			else if ((packets > 35))
5236 				itrval = lowest_latency;
5237 		} else if (bytes/packets > 2000) {
5238 			itrval = bulk_latency;
5239 		} else if (packets <= 2 && bytes < 512) {
5240 			itrval = lowest_latency;
5241 		}
5242 		break;
5243 	case bulk_latency: /* 250 usec aka 4000 ints/s */
5244 		if (bytes > 25000) {
5245 			if (packets > 35)
5246 				itrval = low_latency;
5247 		} else if (bytes < 1500) {
5248 			itrval = low_latency;
5249 		}
5250 		break;
5251 	}
5252 
5253 	/* clear work counters since we have the values we need */
5254 	ring_container->total_bytes = 0;
5255 	ring_container->total_packets = 0;
5256 
5257 	/* write updated itr to ring container */
5258 	ring_container->itr = itrval;
5259 }
5260 
5261 static void igb_set_itr(struct igb_q_vector *q_vector)
5262 {
5263 	struct igb_adapter *adapter = q_vector->adapter;
5264 	u32 new_itr = q_vector->itr_val;
5265 	u8 current_itr = 0;
5266 
5267 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
5268 	if (adapter->link_speed != SPEED_1000) {
5269 		current_itr = 0;
5270 		new_itr = IGB_4K_ITR;
5271 		goto set_itr_now;
5272 	}
5273 
5274 	igb_update_itr(q_vector, &q_vector->tx);
5275 	igb_update_itr(q_vector, &q_vector->rx);
5276 
5277 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
5278 
5279 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
5280 	if (current_itr == lowest_latency &&
5281 	    ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5282 	     (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5283 		current_itr = low_latency;
5284 
5285 	switch (current_itr) {
5286 	/* counts and packets in update_itr are dependent on these numbers */
5287 	case lowest_latency:
5288 		new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
5289 		break;
5290 	case low_latency:
5291 		new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
5292 		break;
5293 	case bulk_latency:
5294 		new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
5295 		break;
5296 	default:
5297 		break;
5298 	}
5299 
5300 set_itr_now:
5301 	if (new_itr != q_vector->itr_val) {
5302 		/* this attempts to bias the interrupt rate towards Bulk
5303 		 * by adding intermediate steps when interrupt rate is
5304 		 * increasing
5305 		 */
5306 		new_itr = new_itr > q_vector->itr_val ?
5307 			  max((new_itr * q_vector->itr_val) /
5308 			  (new_itr + (q_vector->itr_val >> 2)),
5309 			  new_itr) : new_itr;
5310 		/* Don't write the value here; it resets the adapter's
5311 		 * internal timer, and causes us to delay far longer than
5312 		 * we should between interrupts.  Instead, we write the ITR
5313 		 * value at the beginning of the next interrupt so the timing
5314 		 * ends up being correct.
5315 		 */
5316 		q_vector->itr_val = new_itr;
5317 		q_vector->set_itr = 1;
5318 	}
5319 }
5320 
5321 static void igb_tx_ctxtdesc(struct igb_ring *tx_ring, u32 vlan_macip_lens,
5322 			    u32 type_tucmd, u32 mss_l4len_idx)
5323 {
5324 	struct e1000_adv_tx_context_desc *context_desc;
5325 	u16 i = tx_ring->next_to_use;
5326 
5327 	context_desc = IGB_TX_CTXTDESC(tx_ring, i);
5328 
5329 	i++;
5330 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
5331 
5332 	/* set bits to identify this as an advanced context descriptor */
5333 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
5334 
5335 	/* For 82575, context index must be unique per ring. */
5336 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5337 		mss_l4len_idx |= tx_ring->reg_idx << 4;
5338 
5339 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
5340 	context_desc->seqnum_seed	= 0;
5341 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
5342 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
5343 }
5344 
5345 static int igb_tso(struct igb_ring *tx_ring,
5346 		   struct igb_tx_buffer *first,
5347 		   u8 *hdr_len)
5348 {
5349 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
5350 	struct sk_buff *skb = first->skb;
5351 	union {
5352 		struct iphdr *v4;
5353 		struct ipv6hdr *v6;
5354 		unsigned char *hdr;
5355 	} ip;
5356 	union {
5357 		struct tcphdr *tcp;
5358 		unsigned char *hdr;
5359 	} l4;
5360 	u32 paylen, l4_offset;
5361 	int err;
5362 
5363 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5364 		return 0;
5365 
5366 	if (!skb_is_gso(skb))
5367 		return 0;
5368 
5369 	err = skb_cow_head(skb, 0);
5370 	if (err < 0)
5371 		return err;
5372 
5373 	ip.hdr = skb_network_header(skb);
5374 	l4.hdr = skb_checksum_start(skb);
5375 
5376 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5377 	type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
5378 
5379 	/* initialize outer IP header fields */
5380 	if (ip.v4->version == 4) {
5381 		unsigned char *csum_start = skb_checksum_start(skb);
5382 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
5383 
5384 		/* IP header will have to cancel out any data that
5385 		 * is not a part of the outer IP header
5386 		 */
5387 		ip.v4->check = csum_fold(csum_partial(trans_start,
5388 						      csum_start - trans_start,
5389 						      0));
5390 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
5391 
5392 		ip.v4->tot_len = 0;
5393 		first->tx_flags |= IGB_TX_FLAGS_TSO |
5394 				   IGB_TX_FLAGS_CSUM |
5395 				   IGB_TX_FLAGS_IPV4;
5396 	} else {
5397 		ip.v6->payload_len = 0;
5398 		first->tx_flags |= IGB_TX_FLAGS_TSO |
5399 				   IGB_TX_FLAGS_CSUM;
5400 	}
5401 
5402 	/* determine offset of inner transport header */
5403 	l4_offset = l4.hdr - skb->data;
5404 
5405 	/* compute length of segmentation header */
5406 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
5407 
5408 	/* remove payload length from inner checksum */
5409 	paylen = skb->len - l4_offset;
5410 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
5411 
5412 	/* update gso size and bytecount with header size */
5413 	first->gso_segs = skb_shinfo(skb)->gso_segs;
5414 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
5415 
5416 	/* MSS L4LEN IDX */
5417 	mss_l4len_idx = (*hdr_len - l4_offset) << E1000_ADVTXD_L4LEN_SHIFT;
5418 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
5419 
5420 	/* VLAN MACLEN IPLEN */
5421 	vlan_macip_lens = l4.hdr - ip.hdr;
5422 	vlan_macip_lens |= (ip.hdr - skb->data) << E1000_ADVTXD_MACLEN_SHIFT;
5423 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
5424 
5425 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
5426 
5427 	return 1;
5428 }
5429 
5430 static inline bool igb_ipv6_csum_is_sctp(struct sk_buff *skb)
5431 {
5432 	unsigned int offset = 0;
5433 
5434 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
5435 
5436 	return offset == skb_checksum_start_offset(skb);
5437 }
5438 
5439 static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
5440 {
5441 	struct sk_buff *skb = first->skb;
5442 	u32 vlan_macip_lens = 0;
5443 	u32 type_tucmd = 0;
5444 
5445 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
5446 csum_failed:
5447 		if (!(first->tx_flags & IGB_TX_FLAGS_VLAN))
5448 			return;
5449 		goto no_csum;
5450 	}
5451 
5452 	switch (skb->csum_offset) {
5453 	case offsetof(struct tcphdr, check):
5454 		type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
5455 		/* fall through */
5456 	case offsetof(struct udphdr, check):
5457 		break;
5458 	case offsetof(struct sctphdr, checksum):
5459 		/* validate that this is actually an SCTP request */
5460 		if (((first->protocol == htons(ETH_P_IP)) &&
5461 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
5462 		    ((first->protocol == htons(ETH_P_IPV6)) &&
5463 		     igb_ipv6_csum_is_sctp(skb))) {
5464 			type_tucmd = E1000_ADVTXD_TUCMD_L4T_SCTP;
5465 			break;
5466 		}
5467 	default:
5468 		skb_checksum_help(skb);
5469 		goto csum_failed;
5470 	}
5471 
5472 	/* update TX checksum flag */
5473 	first->tx_flags |= IGB_TX_FLAGS_CSUM;
5474 	vlan_macip_lens = skb_checksum_start_offset(skb) -
5475 			  skb_network_offset(skb);
5476 no_csum:
5477 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
5478 	vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
5479 
5480 	igb_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, 0);
5481 }
5482 
5483 #define IGB_SET_FLAG(_input, _flag, _result) \
5484 	((_flag <= _result) ? \
5485 	 ((u32)(_input & _flag) * (_result / _flag)) : \
5486 	 ((u32)(_input & _flag) / (_flag / _result)))
5487 
5488 static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
5489 {
5490 	/* set type for advanced descriptor with frame checksum insertion */
5491 	u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
5492 		       E1000_ADVTXD_DCMD_DEXT |
5493 		       E1000_ADVTXD_DCMD_IFCS;
5494 
5495 	/* set HW vlan bit if vlan is present */
5496 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
5497 				 (E1000_ADVTXD_DCMD_VLE));
5498 
5499 	/* set segmentation bits for TSO */
5500 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
5501 				 (E1000_ADVTXD_DCMD_TSE));
5502 
5503 	/* set timestamp bit if present */
5504 	cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
5505 				 (E1000_ADVTXD_MAC_TSTAMP));
5506 
5507 	/* insert frame checksum */
5508 	cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
5509 
5510 	return cmd_type;
5511 }
5512 
5513 static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
5514 				 union e1000_adv_tx_desc *tx_desc,
5515 				 u32 tx_flags, unsigned int paylen)
5516 {
5517 	u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
5518 
5519 	/* 82575 requires a unique index per ring */
5520 	if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5521 		olinfo_status |= tx_ring->reg_idx << 4;
5522 
5523 	/* insert L4 checksum */
5524 	olinfo_status |= IGB_SET_FLAG(tx_flags,
5525 				      IGB_TX_FLAGS_CSUM,
5526 				      (E1000_TXD_POPTS_TXSM << 8));
5527 
5528 	/* insert IPv4 checksum */
5529 	olinfo_status |= IGB_SET_FLAG(tx_flags,
5530 				      IGB_TX_FLAGS_IPV4,
5531 				      (E1000_TXD_POPTS_IXSM << 8));
5532 
5533 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
5534 }
5535 
5536 static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
5537 {
5538 	struct net_device *netdev = tx_ring->netdev;
5539 
5540 	netif_stop_subqueue(netdev, tx_ring->queue_index);
5541 
5542 	/* Herbert's original patch had:
5543 	 *  smp_mb__after_netif_stop_queue();
5544 	 * but since that doesn't exist yet, just open code it.
5545 	 */
5546 	smp_mb();
5547 
5548 	/* We need to check again in a case another CPU has just
5549 	 * made room available.
5550 	 */
5551 	if (igb_desc_unused(tx_ring) < size)
5552 		return -EBUSY;
5553 
5554 	/* A reprieve! */
5555 	netif_wake_subqueue(netdev, tx_ring->queue_index);
5556 
5557 	u64_stats_update_begin(&tx_ring->tx_syncp2);
5558 	tx_ring->tx_stats.restart_queue2++;
5559 	u64_stats_update_end(&tx_ring->tx_syncp2);
5560 
5561 	return 0;
5562 }
5563 
5564 static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
5565 {
5566 	if (igb_desc_unused(tx_ring) >= size)
5567 		return 0;
5568 	return __igb_maybe_stop_tx(tx_ring, size);
5569 }
5570 
5571 static int igb_tx_map(struct igb_ring *tx_ring,
5572 		      struct igb_tx_buffer *first,
5573 		      const u8 hdr_len)
5574 {
5575 	struct sk_buff *skb = first->skb;
5576 	struct igb_tx_buffer *tx_buffer;
5577 	union e1000_adv_tx_desc *tx_desc;
5578 	struct skb_frag_struct *frag;
5579 	dma_addr_t dma;
5580 	unsigned int data_len, size;
5581 	u32 tx_flags = first->tx_flags;
5582 	u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
5583 	u16 i = tx_ring->next_to_use;
5584 
5585 	tx_desc = IGB_TX_DESC(tx_ring, i);
5586 
5587 	igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
5588 
5589 	size = skb_headlen(skb);
5590 	data_len = skb->data_len;
5591 
5592 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
5593 
5594 	tx_buffer = first;
5595 
5596 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
5597 		if (dma_mapping_error(tx_ring->dev, dma))
5598 			goto dma_error;
5599 
5600 		/* record length, and DMA address */
5601 		dma_unmap_len_set(tx_buffer, len, size);
5602 		dma_unmap_addr_set(tx_buffer, dma, dma);
5603 
5604 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
5605 
5606 		while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
5607 			tx_desc->read.cmd_type_len =
5608 				cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
5609 
5610 			i++;
5611 			tx_desc++;
5612 			if (i == tx_ring->count) {
5613 				tx_desc = IGB_TX_DESC(tx_ring, 0);
5614 				i = 0;
5615 			}
5616 			tx_desc->read.olinfo_status = 0;
5617 
5618 			dma += IGB_MAX_DATA_PER_TXD;
5619 			size -= IGB_MAX_DATA_PER_TXD;
5620 
5621 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
5622 		}
5623 
5624 		if (likely(!data_len))
5625 			break;
5626 
5627 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
5628 
5629 		i++;
5630 		tx_desc++;
5631 		if (i == tx_ring->count) {
5632 			tx_desc = IGB_TX_DESC(tx_ring, 0);
5633 			i = 0;
5634 		}
5635 		tx_desc->read.olinfo_status = 0;
5636 
5637 		size = skb_frag_size(frag);
5638 		data_len -= size;
5639 
5640 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
5641 				       size, DMA_TO_DEVICE);
5642 
5643 		tx_buffer = &tx_ring->tx_buffer_info[i];
5644 	}
5645 
5646 	/* write last descriptor with RS and EOP bits */
5647 	cmd_type |= size | IGB_TXD_DCMD;
5648 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
5649 
5650 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
5651 
5652 	/* set the timestamp */
5653 	first->time_stamp = jiffies;
5654 
5655 	/* Force memory writes to complete before letting h/w know there
5656 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
5657 	 * memory model archs, such as IA-64).
5658 	 *
5659 	 * We also need this memory barrier to make certain all of the
5660 	 * status bits have been updated before next_to_watch is written.
5661 	 */
5662 	wmb();
5663 
5664 	/* set next_to_watch value indicating a packet is present */
5665 	first->next_to_watch = tx_desc;
5666 
5667 	i++;
5668 	if (i == tx_ring->count)
5669 		i = 0;
5670 
5671 	tx_ring->next_to_use = i;
5672 
5673 	/* Make sure there is space in the ring for the next send. */
5674 	igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
5675 
5676 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
5677 		writel(i, tx_ring->tail);
5678 
5679 		/* we need this if more than one processor can write to our tail
5680 		 * at a time, it synchronizes IO on IA64/Altix systems
5681 		 */
5682 		mmiowb();
5683 	}
5684 	return 0;
5685 
5686 dma_error:
5687 	dev_err(tx_ring->dev, "TX DMA map failed\n");
5688 	tx_buffer = &tx_ring->tx_buffer_info[i];
5689 
5690 	/* clear dma mappings for failed tx_buffer_info map */
5691 	while (tx_buffer != first) {
5692 		if (dma_unmap_len(tx_buffer, len))
5693 			dma_unmap_page(tx_ring->dev,
5694 				       dma_unmap_addr(tx_buffer, dma),
5695 				       dma_unmap_len(tx_buffer, len),
5696 				       DMA_TO_DEVICE);
5697 		dma_unmap_len_set(tx_buffer, len, 0);
5698 
5699 		if (i-- == 0)
5700 			i += tx_ring->count;
5701 		tx_buffer = &tx_ring->tx_buffer_info[i];
5702 	}
5703 
5704 	if (dma_unmap_len(tx_buffer, len))
5705 		dma_unmap_single(tx_ring->dev,
5706 				 dma_unmap_addr(tx_buffer, dma),
5707 				 dma_unmap_len(tx_buffer, len),
5708 				 DMA_TO_DEVICE);
5709 	dma_unmap_len_set(tx_buffer, len, 0);
5710 
5711 	dev_kfree_skb_any(tx_buffer->skb);
5712 	tx_buffer->skb = NULL;
5713 
5714 	tx_ring->next_to_use = i;
5715 
5716 	return -1;
5717 }
5718 
5719 netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
5720 				struct igb_ring *tx_ring)
5721 {
5722 	struct igb_tx_buffer *first;
5723 	int tso;
5724 	u32 tx_flags = 0;
5725 	unsigned short f;
5726 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
5727 	__be16 protocol = vlan_get_protocol(skb);
5728 	u8 hdr_len = 0;
5729 
5730 	/* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
5731 	 *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
5732 	 *       + 2 desc gap to keep tail from touching head,
5733 	 *       + 1 desc for context descriptor,
5734 	 * otherwise try next time
5735 	 */
5736 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
5737 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
5738 
5739 	if (igb_maybe_stop_tx(tx_ring, count + 3)) {
5740 		/* this is a hard error */
5741 		return NETDEV_TX_BUSY;
5742 	}
5743 
5744 	/* record the location of the first descriptor for this packet */
5745 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
5746 	first->skb = skb;
5747 	first->bytecount = skb->len;
5748 	first->gso_segs = 1;
5749 
5750 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
5751 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
5752 
5753 		if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
5754 		    !test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
5755 					   &adapter->state)) {
5756 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5757 			tx_flags |= IGB_TX_FLAGS_TSTAMP;
5758 
5759 			adapter->ptp_tx_skb = skb_get(skb);
5760 			adapter->ptp_tx_start = jiffies;
5761 			if (adapter->hw.mac.type == e1000_82576)
5762 				schedule_work(&adapter->ptp_tx_work);
5763 		} else {
5764 			adapter->tx_hwtstamp_skipped++;
5765 		}
5766 	}
5767 
5768 	skb_tx_timestamp(skb);
5769 
5770 	if (skb_vlan_tag_present(skb)) {
5771 		tx_flags |= IGB_TX_FLAGS_VLAN;
5772 		tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
5773 	}
5774 
5775 	/* record initial flags and protocol */
5776 	first->tx_flags = tx_flags;
5777 	first->protocol = protocol;
5778 
5779 	tso = igb_tso(tx_ring, first, &hdr_len);
5780 	if (tso < 0)
5781 		goto out_drop;
5782 	else if (!tso)
5783 		igb_tx_csum(tx_ring, first);
5784 
5785 	if (igb_tx_map(tx_ring, first, hdr_len))
5786 		goto cleanup_tx_tstamp;
5787 
5788 	return NETDEV_TX_OK;
5789 
5790 out_drop:
5791 	dev_kfree_skb_any(first->skb);
5792 	first->skb = NULL;
5793 cleanup_tx_tstamp:
5794 	if (unlikely(tx_flags & IGB_TX_FLAGS_TSTAMP)) {
5795 		struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
5796 
5797 		dev_kfree_skb_any(adapter->ptp_tx_skb);
5798 		adapter->ptp_tx_skb = NULL;
5799 		if (adapter->hw.mac.type == e1000_82576)
5800 			cancel_work_sync(&adapter->ptp_tx_work);
5801 		clear_bit_unlock(__IGB_PTP_TX_IN_PROGRESS, &adapter->state);
5802 	}
5803 
5804 	return NETDEV_TX_OK;
5805 }
5806 
5807 static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
5808 						    struct sk_buff *skb)
5809 {
5810 	unsigned int r_idx = skb->queue_mapping;
5811 
5812 	if (r_idx >= adapter->num_tx_queues)
5813 		r_idx = r_idx % adapter->num_tx_queues;
5814 
5815 	return adapter->tx_ring[r_idx];
5816 }
5817 
5818 static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
5819 				  struct net_device *netdev)
5820 {
5821 	struct igb_adapter *adapter = netdev_priv(netdev);
5822 
5823 	/* The minimum packet size with TCTL.PSP set is 17 so pad the skb
5824 	 * in order to meet this minimum size requirement.
5825 	 */
5826 	if (skb_put_padto(skb, 17))
5827 		return NETDEV_TX_OK;
5828 
5829 	return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
5830 }
5831 
5832 /**
5833  *  igb_tx_timeout - Respond to a Tx Hang
5834  *  @netdev: network interface device structure
5835  **/
5836 static void igb_tx_timeout(struct net_device *netdev)
5837 {
5838 	struct igb_adapter *adapter = netdev_priv(netdev);
5839 	struct e1000_hw *hw = &adapter->hw;
5840 
5841 	/* Do the reset outside of interrupt context */
5842 	adapter->tx_timeout_count++;
5843 
5844 	if (hw->mac.type >= e1000_82580)
5845 		hw->dev_spec._82575.global_device_reset = true;
5846 
5847 	schedule_work(&adapter->reset_task);
5848 	wr32(E1000_EICS,
5849 	     (adapter->eims_enable_mask & ~adapter->eims_other));
5850 }
5851 
5852 static void igb_reset_task(struct work_struct *work)
5853 {
5854 	struct igb_adapter *adapter;
5855 	adapter = container_of(work, struct igb_adapter, reset_task);
5856 
5857 	igb_dump(adapter);
5858 	netdev_err(adapter->netdev, "Reset adapter\n");
5859 	igb_reinit_locked(adapter);
5860 }
5861 
5862 /**
5863  *  igb_get_stats64 - Get System Network Statistics
5864  *  @netdev: network interface device structure
5865  *  @stats: rtnl_link_stats64 pointer
5866  **/
5867 static void igb_get_stats64(struct net_device *netdev,
5868 			    struct rtnl_link_stats64 *stats)
5869 {
5870 	struct igb_adapter *adapter = netdev_priv(netdev);
5871 
5872 	spin_lock(&adapter->stats64_lock);
5873 	igb_update_stats(adapter);
5874 	memcpy(stats, &adapter->stats64, sizeof(*stats));
5875 	spin_unlock(&adapter->stats64_lock);
5876 }
5877 
5878 /**
5879  *  igb_change_mtu - Change the Maximum Transfer Unit
5880  *  @netdev: network interface device structure
5881  *  @new_mtu: new value for maximum frame size
5882  *
5883  *  Returns 0 on success, negative on failure
5884  **/
5885 static int igb_change_mtu(struct net_device *netdev, int new_mtu)
5886 {
5887 	struct igb_adapter *adapter = netdev_priv(netdev);
5888 	struct pci_dev *pdev = adapter->pdev;
5889 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
5890 
5891 	/* adjust max frame to be at least the size of a standard frame */
5892 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
5893 		max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
5894 
5895 	while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
5896 		usleep_range(1000, 2000);
5897 
5898 	/* igb_down has a dependency on max_frame_size */
5899 	adapter->max_frame_size = max_frame;
5900 
5901 	if (netif_running(netdev))
5902 		igb_down(adapter);
5903 
5904 	dev_info(&pdev->dev, "changing MTU from %d to %d\n",
5905 		 netdev->mtu, new_mtu);
5906 	netdev->mtu = new_mtu;
5907 
5908 	if (netif_running(netdev))
5909 		igb_up(adapter);
5910 	else
5911 		igb_reset(adapter);
5912 
5913 	clear_bit(__IGB_RESETTING, &adapter->state);
5914 
5915 	return 0;
5916 }
5917 
5918 /**
5919  *  igb_update_stats - Update the board statistics counters
5920  *  @adapter: board private structure
5921  **/
5922 void igb_update_stats(struct igb_adapter *adapter)
5923 {
5924 	struct rtnl_link_stats64 *net_stats = &adapter->stats64;
5925 	struct e1000_hw *hw = &adapter->hw;
5926 	struct pci_dev *pdev = adapter->pdev;
5927 	u32 reg, mpc;
5928 	int i;
5929 	u64 bytes, packets;
5930 	unsigned int start;
5931 	u64 _bytes, _packets;
5932 
5933 	/* Prevent stats update while adapter is being reset, or if the pci
5934 	 * connection is down.
5935 	 */
5936 	if (adapter->link_speed == 0)
5937 		return;
5938 	if (pci_channel_offline(pdev))
5939 		return;
5940 
5941 	bytes = 0;
5942 	packets = 0;
5943 
5944 	rcu_read_lock();
5945 	for (i = 0; i < adapter->num_rx_queues; i++) {
5946 		struct igb_ring *ring = adapter->rx_ring[i];
5947 		u32 rqdpc = rd32(E1000_RQDPC(i));
5948 		if (hw->mac.type >= e1000_i210)
5949 			wr32(E1000_RQDPC(i), 0);
5950 
5951 		if (rqdpc) {
5952 			ring->rx_stats.drops += rqdpc;
5953 			net_stats->rx_fifo_errors += rqdpc;
5954 		}
5955 
5956 		do {
5957 			start = u64_stats_fetch_begin_irq(&ring->rx_syncp);
5958 			_bytes = ring->rx_stats.bytes;
5959 			_packets = ring->rx_stats.packets;
5960 		} while (u64_stats_fetch_retry_irq(&ring->rx_syncp, start));
5961 		bytes += _bytes;
5962 		packets += _packets;
5963 	}
5964 
5965 	net_stats->rx_bytes = bytes;
5966 	net_stats->rx_packets = packets;
5967 
5968 	bytes = 0;
5969 	packets = 0;
5970 	for (i = 0; i < adapter->num_tx_queues; i++) {
5971 		struct igb_ring *ring = adapter->tx_ring[i];
5972 		do {
5973 			start = u64_stats_fetch_begin_irq(&ring->tx_syncp);
5974 			_bytes = ring->tx_stats.bytes;
5975 			_packets = ring->tx_stats.packets;
5976 		} while (u64_stats_fetch_retry_irq(&ring->tx_syncp, start));
5977 		bytes += _bytes;
5978 		packets += _packets;
5979 	}
5980 	net_stats->tx_bytes = bytes;
5981 	net_stats->tx_packets = packets;
5982 	rcu_read_unlock();
5983 
5984 	/* read stats registers */
5985 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
5986 	adapter->stats.gprc += rd32(E1000_GPRC);
5987 	adapter->stats.gorc += rd32(E1000_GORCL);
5988 	rd32(E1000_GORCH); /* clear GORCL */
5989 	adapter->stats.bprc += rd32(E1000_BPRC);
5990 	adapter->stats.mprc += rd32(E1000_MPRC);
5991 	adapter->stats.roc += rd32(E1000_ROC);
5992 
5993 	adapter->stats.prc64 += rd32(E1000_PRC64);
5994 	adapter->stats.prc127 += rd32(E1000_PRC127);
5995 	adapter->stats.prc255 += rd32(E1000_PRC255);
5996 	adapter->stats.prc511 += rd32(E1000_PRC511);
5997 	adapter->stats.prc1023 += rd32(E1000_PRC1023);
5998 	adapter->stats.prc1522 += rd32(E1000_PRC1522);
5999 	adapter->stats.symerrs += rd32(E1000_SYMERRS);
6000 	adapter->stats.sec += rd32(E1000_SEC);
6001 
6002 	mpc = rd32(E1000_MPC);
6003 	adapter->stats.mpc += mpc;
6004 	net_stats->rx_fifo_errors += mpc;
6005 	adapter->stats.scc += rd32(E1000_SCC);
6006 	adapter->stats.ecol += rd32(E1000_ECOL);
6007 	adapter->stats.mcc += rd32(E1000_MCC);
6008 	adapter->stats.latecol += rd32(E1000_LATECOL);
6009 	adapter->stats.dc += rd32(E1000_DC);
6010 	adapter->stats.rlec += rd32(E1000_RLEC);
6011 	adapter->stats.xonrxc += rd32(E1000_XONRXC);
6012 	adapter->stats.xontxc += rd32(E1000_XONTXC);
6013 	adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
6014 	adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
6015 	adapter->stats.fcruc += rd32(E1000_FCRUC);
6016 	adapter->stats.gptc += rd32(E1000_GPTC);
6017 	adapter->stats.gotc += rd32(E1000_GOTCL);
6018 	rd32(E1000_GOTCH); /* clear GOTCL */
6019 	adapter->stats.rnbc += rd32(E1000_RNBC);
6020 	adapter->stats.ruc += rd32(E1000_RUC);
6021 	adapter->stats.rfc += rd32(E1000_RFC);
6022 	adapter->stats.rjc += rd32(E1000_RJC);
6023 	adapter->stats.tor += rd32(E1000_TORH);
6024 	adapter->stats.tot += rd32(E1000_TOTH);
6025 	adapter->stats.tpr += rd32(E1000_TPR);
6026 
6027 	adapter->stats.ptc64 += rd32(E1000_PTC64);
6028 	adapter->stats.ptc127 += rd32(E1000_PTC127);
6029 	adapter->stats.ptc255 += rd32(E1000_PTC255);
6030 	adapter->stats.ptc511 += rd32(E1000_PTC511);
6031 	adapter->stats.ptc1023 += rd32(E1000_PTC1023);
6032 	adapter->stats.ptc1522 += rd32(E1000_PTC1522);
6033 
6034 	adapter->stats.mptc += rd32(E1000_MPTC);
6035 	adapter->stats.bptc += rd32(E1000_BPTC);
6036 
6037 	adapter->stats.tpt += rd32(E1000_TPT);
6038 	adapter->stats.colc += rd32(E1000_COLC);
6039 
6040 	adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
6041 	/* read internal phy specific stats */
6042 	reg = rd32(E1000_CTRL_EXT);
6043 	if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
6044 		adapter->stats.rxerrc += rd32(E1000_RXERRC);
6045 
6046 		/* this stat has invalid values on i210/i211 */
6047 		if ((hw->mac.type != e1000_i210) &&
6048 		    (hw->mac.type != e1000_i211))
6049 			adapter->stats.tncrs += rd32(E1000_TNCRS);
6050 	}
6051 
6052 	adapter->stats.tsctc += rd32(E1000_TSCTC);
6053 	adapter->stats.tsctfc += rd32(E1000_TSCTFC);
6054 
6055 	adapter->stats.iac += rd32(E1000_IAC);
6056 	adapter->stats.icrxoc += rd32(E1000_ICRXOC);
6057 	adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
6058 	adapter->stats.icrxatc += rd32(E1000_ICRXATC);
6059 	adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
6060 	adapter->stats.ictxatc += rd32(E1000_ICTXATC);
6061 	adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
6062 	adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
6063 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
6064 
6065 	/* Fill out the OS statistics structure */
6066 	net_stats->multicast = adapter->stats.mprc;
6067 	net_stats->collisions = adapter->stats.colc;
6068 
6069 	/* Rx Errors */
6070 
6071 	/* RLEC on some newer hardware can be incorrect so build
6072 	 * our own version based on RUC and ROC
6073 	 */
6074 	net_stats->rx_errors = adapter->stats.rxerrc +
6075 		adapter->stats.crcerrs + adapter->stats.algnerrc +
6076 		adapter->stats.ruc + adapter->stats.roc +
6077 		adapter->stats.cexterr;
6078 	net_stats->rx_length_errors = adapter->stats.ruc +
6079 				      adapter->stats.roc;
6080 	net_stats->rx_crc_errors = adapter->stats.crcerrs;
6081 	net_stats->rx_frame_errors = adapter->stats.algnerrc;
6082 	net_stats->rx_missed_errors = adapter->stats.mpc;
6083 
6084 	/* Tx Errors */
6085 	net_stats->tx_errors = adapter->stats.ecol +
6086 			       adapter->stats.latecol;
6087 	net_stats->tx_aborted_errors = adapter->stats.ecol;
6088 	net_stats->tx_window_errors = adapter->stats.latecol;
6089 	net_stats->tx_carrier_errors = adapter->stats.tncrs;
6090 
6091 	/* Tx Dropped needs to be maintained elsewhere */
6092 
6093 	/* Management Stats */
6094 	adapter->stats.mgptc += rd32(E1000_MGTPTC);
6095 	adapter->stats.mgprc += rd32(E1000_MGTPRC);
6096 	adapter->stats.mgpdc += rd32(E1000_MGTPDC);
6097 
6098 	/* OS2BMC Stats */
6099 	reg = rd32(E1000_MANC);
6100 	if (reg & E1000_MANC_EN_BMC2OS) {
6101 		adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
6102 		adapter->stats.o2bspc += rd32(E1000_O2BSPC);
6103 		adapter->stats.b2ospc += rd32(E1000_B2OSPC);
6104 		adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
6105 	}
6106 }
6107 
6108 static void igb_tsync_interrupt(struct igb_adapter *adapter)
6109 {
6110 	struct e1000_hw *hw = &adapter->hw;
6111 	struct ptp_clock_event event;
6112 	struct timespec64 ts;
6113 	u32 ack = 0, tsauxc, sec, nsec, tsicr = rd32(E1000_TSICR);
6114 
6115 	if (tsicr & TSINTR_SYS_WRAP) {
6116 		event.type = PTP_CLOCK_PPS;
6117 		if (adapter->ptp_caps.pps)
6118 			ptp_clock_event(adapter->ptp_clock, &event);
6119 		ack |= TSINTR_SYS_WRAP;
6120 	}
6121 
6122 	if (tsicr & E1000_TSICR_TXTS) {
6123 		/* retrieve hardware timestamp */
6124 		schedule_work(&adapter->ptp_tx_work);
6125 		ack |= E1000_TSICR_TXTS;
6126 	}
6127 
6128 	if (tsicr & TSINTR_TT0) {
6129 		spin_lock(&adapter->tmreg_lock);
6130 		ts = timespec64_add(adapter->perout[0].start,
6131 				    adapter->perout[0].period);
6132 		/* u32 conversion of tv_sec is safe until y2106 */
6133 		wr32(E1000_TRGTTIML0, ts.tv_nsec);
6134 		wr32(E1000_TRGTTIMH0, (u32)ts.tv_sec);
6135 		tsauxc = rd32(E1000_TSAUXC);
6136 		tsauxc |= TSAUXC_EN_TT0;
6137 		wr32(E1000_TSAUXC, tsauxc);
6138 		adapter->perout[0].start = ts;
6139 		spin_unlock(&adapter->tmreg_lock);
6140 		ack |= TSINTR_TT0;
6141 	}
6142 
6143 	if (tsicr & TSINTR_TT1) {
6144 		spin_lock(&adapter->tmreg_lock);
6145 		ts = timespec64_add(adapter->perout[1].start,
6146 				    adapter->perout[1].period);
6147 		wr32(E1000_TRGTTIML1, ts.tv_nsec);
6148 		wr32(E1000_TRGTTIMH1, (u32)ts.tv_sec);
6149 		tsauxc = rd32(E1000_TSAUXC);
6150 		tsauxc |= TSAUXC_EN_TT1;
6151 		wr32(E1000_TSAUXC, tsauxc);
6152 		adapter->perout[1].start = ts;
6153 		spin_unlock(&adapter->tmreg_lock);
6154 		ack |= TSINTR_TT1;
6155 	}
6156 
6157 	if (tsicr & TSINTR_AUTT0) {
6158 		nsec = rd32(E1000_AUXSTMPL0);
6159 		sec  = rd32(E1000_AUXSTMPH0);
6160 		event.type = PTP_CLOCK_EXTTS;
6161 		event.index = 0;
6162 		event.timestamp = sec * 1000000000ULL + nsec;
6163 		ptp_clock_event(adapter->ptp_clock, &event);
6164 		ack |= TSINTR_AUTT0;
6165 	}
6166 
6167 	if (tsicr & TSINTR_AUTT1) {
6168 		nsec = rd32(E1000_AUXSTMPL1);
6169 		sec  = rd32(E1000_AUXSTMPH1);
6170 		event.type = PTP_CLOCK_EXTTS;
6171 		event.index = 1;
6172 		event.timestamp = sec * 1000000000ULL + nsec;
6173 		ptp_clock_event(adapter->ptp_clock, &event);
6174 		ack |= TSINTR_AUTT1;
6175 	}
6176 
6177 	/* acknowledge the interrupts */
6178 	wr32(E1000_TSICR, ack);
6179 }
6180 
6181 static irqreturn_t igb_msix_other(int irq, void *data)
6182 {
6183 	struct igb_adapter *adapter = data;
6184 	struct e1000_hw *hw = &adapter->hw;
6185 	u32 icr = rd32(E1000_ICR);
6186 	/* reading ICR causes bit 31 of EICR to be cleared */
6187 
6188 	if (icr & E1000_ICR_DRSTA)
6189 		schedule_work(&adapter->reset_task);
6190 
6191 	if (icr & E1000_ICR_DOUTSYNC) {
6192 		/* HW is reporting DMA is out of sync */
6193 		adapter->stats.doosync++;
6194 		/* The DMA Out of Sync is also indication of a spoof event
6195 		 * in IOV mode. Check the Wrong VM Behavior register to
6196 		 * see if it is really a spoof event.
6197 		 */
6198 		igb_check_wvbr(adapter);
6199 	}
6200 
6201 	/* Check for a mailbox event */
6202 	if (icr & E1000_ICR_VMMB)
6203 		igb_msg_task(adapter);
6204 
6205 	if (icr & E1000_ICR_LSC) {
6206 		hw->mac.get_link_status = 1;
6207 		/* guard against interrupt when we're going down */
6208 		if (!test_bit(__IGB_DOWN, &adapter->state))
6209 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
6210 	}
6211 
6212 	if (icr & E1000_ICR_TS)
6213 		igb_tsync_interrupt(adapter);
6214 
6215 	wr32(E1000_EIMS, adapter->eims_other);
6216 
6217 	return IRQ_HANDLED;
6218 }
6219 
6220 static void igb_write_itr(struct igb_q_vector *q_vector)
6221 {
6222 	struct igb_adapter *adapter = q_vector->adapter;
6223 	u32 itr_val = q_vector->itr_val & 0x7FFC;
6224 
6225 	if (!q_vector->set_itr)
6226 		return;
6227 
6228 	if (!itr_val)
6229 		itr_val = 0x4;
6230 
6231 	if (adapter->hw.mac.type == e1000_82575)
6232 		itr_val |= itr_val << 16;
6233 	else
6234 		itr_val |= E1000_EITR_CNT_IGNR;
6235 
6236 	writel(itr_val, q_vector->itr_register);
6237 	q_vector->set_itr = 0;
6238 }
6239 
6240 static irqreturn_t igb_msix_ring(int irq, void *data)
6241 {
6242 	struct igb_q_vector *q_vector = data;
6243 
6244 	/* Write the ITR value calculated from the previous interrupt. */
6245 	igb_write_itr(q_vector);
6246 
6247 	napi_schedule(&q_vector->napi);
6248 
6249 	return IRQ_HANDLED;
6250 }
6251 
6252 #ifdef CONFIG_IGB_DCA
6253 static void igb_update_tx_dca(struct igb_adapter *adapter,
6254 			      struct igb_ring *tx_ring,
6255 			      int cpu)
6256 {
6257 	struct e1000_hw *hw = &adapter->hw;
6258 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
6259 
6260 	if (hw->mac.type != e1000_82575)
6261 		txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
6262 
6263 	/* We can enable relaxed ordering for reads, but not writes when
6264 	 * DCA is enabled.  This is due to a known issue in some chipsets
6265 	 * which will cause the DCA tag to be cleared.
6266 	 */
6267 	txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
6268 		  E1000_DCA_TXCTRL_DATA_RRO_EN |
6269 		  E1000_DCA_TXCTRL_DESC_DCA_EN;
6270 
6271 	wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
6272 }
6273 
6274 static void igb_update_rx_dca(struct igb_adapter *adapter,
6275 			      struct igb_ring *rx_ring,
6276 			      int cpu)
6277 {
6278 	struct e1000_hw *hw = &adapter->hw;
6279 	u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
6280 
6281 	if (hw->mac.type != e1000_82575)
6282 		rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
6283 
6284 	/* We can enable relaxed ordering for reads, but not writes when
6285 	 * DCA is enabled.  This is due to a known issue in some chipsets
6286 	 * which will cause the DCA tag to be cleared.
6287 	 */
6288 	rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
6289 		  E1000_DCA_RXCTRL_DESC_DCA_EN;
6290 
6291 	wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
6292 }
6293 
6294 static void igb_update_dca(struct igb_q_vector *q_vector)
6295 {
6296 	struct igb_adapter *adapter = q_vector->adapter;
6297 	int cpu = get_cpu();
6298 
6299 	if (q_vector->cpu == cpu)
6300 		goto out_no_update;
6301 
6302 	if (q_vector->tx.ring)
6303 		igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
6304 
6305 	if (q_vector->rx.ring)
6306 		igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
6307 
6308 	q_vector->cpu = cpu;
6309 out_no_update:
6310 	put_cpu();
6311 }
6312 
6313 static void igb_setup_dca(struct igb_adapter *adapter)
6314 {
6315 	struct e1000_hw *hw = &adapter->hw;
6316 	int i;
6317 
6318 	if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
6319 		return;
6320 
6321 	/* Always use CB2 mode, difference is masked in the CB driver. */
6322 	wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
6323 
6324 	for (i = 0; i < adapter->num_q_vectors; i++) {
6325 		adapter->q_vector[i]->cpu = -1;
6326 		igb_update_dca(adapter->q_vector[i]);
6327 	}
6328 }
6329 
6330 static int __igb_notify_dca(struct device *dev, void *data)
6331 {
6332 	struct net_device *netdev = dev_get_drvdata(dev);
6333 	struct igb_adapter *adapter = netdev_priv(netdev);
6334 	struct pci_dev *pdev = adapter->pdev;
6335 	struct e1000_hw *hw = &adapter->hw;
6336 	unsigned long event = *(unsigned long *)data;
6337 
6338 	switch (event) {
6339 	case DCA_PROVIDER_ADD:
6340 		/* if already enabled, don't do it again */
6341 		if (adapter->flags & IGB_FLAG_DCA_ENABLED)
6342 			break;
6343 		if (dca_add_requester(dev) == 0) {
6344 			adapter->flags |= IGB_FLAG_DCA_ENABLED;
6345 			dev_info(&pdev->dev, "DCA enabled\n");
6346 			igb_setup_dca(adapter);
6347 			break;
6348 		}
6349 		/* Fall Through since DCA is disabled. */
6350 	case DCA_PROVIDER_REMOVE:
6351 		if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
6352 			/* without this a class_device is left
6353 			 * hanging around in the sysfs model
6354 			 */
6355 			dca_remove_requester(dev);
6356 			dev_info(&pdev->dev, "DCA disabled\n");
6357 			adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
6358 			wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
6359 		}
6360 		break;
6361 	}
6362 
6363 	return 0;
6364 }
6365 
6366 static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
6367 			  void *p)
6368 {
6369 	int ret_val;
6370 
6371 	ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
6372 					 __igb_notify_dca);
6373 
6374 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
6375 }
6376 #endif /* CONFIG_IGB_DCA */
6377 
6378 #ifdef CONFIG_PCI_IOV
6379 static int igb_vf_configure(struct igb_adapter *adapter, int vf)
6380 {
6381 	unsigned char mac_addr[ETH_ALEN];
6382 
6383 	eth_zero_addr(mac_addr);
6384 	igb_set_vf_mac(adapter, vf, mac_addr);
6385 
6386 	/* By default spoof check is enabled for all VFs */
6387 	adapter->vf_data[vf].spoofchk_enabled = true;
6388 
6389 	/* By default VFs are not trusted */
6390 	adapter->vf_data[vf].trusted = false;
6391 
6392 	return 0;
6393 }
6394 
6395 #endif
6396 static void igb_ping_all_vfs(struct igb_adapter *adapter)
6397 {
6398 	struct e1000_hw *hw = &adapter->hw;
6399 	u32 ping;
6400 	int i;
6401 
6402 	for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
6403 		ping = E1000_PF_CONTROL_MSG;
6404 		if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
6405 			ping |= E1000_VT_MSGTYPE_CTS;
6406 		igb_write_mbx(hw, &ping, 1, i);
6407 	}
6408 }
6409 
6410 static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
6411 {
6412 	struct e1000_hw *hw = &adapter->hw;
6413 	u32 vmolr = rd32(E1000_VMOLR(vf));
6414 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6415 
6416 	vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
6417 			    IGB_VF_FLAG_MULTI_PROMISC);
6418 	vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
6419 
6420 	if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
6421 		vmolr |= E1000_VMOLR_MPME;
6422 		vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
6423 		*msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
6424 	} else {
6425 		/* if we have hashes and we are clearing a multicast promisc
6426 		 * flag we need to write the hashes to the MTA as this step
6427 		 * was previously skipped
6428 		 */
6429 		if (vf_data->num_vf_mc_hashes > 30) {
6430 			vmolr |= E1000_VMOLR_MPME;
6431 		} else if (vf_data->num_vf_mc_hashes) {
6432 			int j;
6433 
6434 			vmolr |= E1000_VMOLR_ROMPE;
6435 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
6436 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
6437 		}
6438 	}
6439 
6440 	wr32(E1000_VMOLR(vf), vmolr);
6441 
6442 	/* there are flags left unprocessed, likely not supported */
6443 	if (*msgbuf & E1000_VT_MSGINFO_MASK)
6444 		return -EINVAL;
6445 
6446 	return 0;
6447 }
6448 
6449 static int igb_set_vf_multicasts(struct igb_adapter *adapter,
6450 				  u32 *msgbuf, u32 vf)
6451 {
6452 	int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
6453 	u16 *hash_list = (u16 *)&msgbuf[1];
6454 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6455 	int i;
6456 
6457 	/* salt away the number of multicast addresses assigned
6458 	 * to this VF for later use to restore when the PF multi cast
6459 	 * list changes
6460 	 */
6461 	vf_data->num_vf_mc_hashes = n;
6462 
6463 	/* only up to 30 hash values supported */
6464 	if (n > 30)
6465 		n = 30;
6466 
6467 	/* store the hashes for later use */
6468 	for (i = 0; i < n; i++)
6469 		vf_data->vf_mc_hashes[i] = hash_list[i];
6470 
6471 	/* Flush and reset the mta with the new values */
6472 	igb_set_rx_mode(adapter->netdev);
6473 
6474 	return 0;
6475 }
6476 
6477 static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
6478 {
6479 	struct e1000_hw *hw = &adapter->hw;
6480 	struct vf_data_storage *vf_data;
6481 	int i, j;
6482 
6483 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
6484 		u32 vmolr = rd32(E1000_VMOLR(i));
6485 
6486 		vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
6487 
6488 		vf_data = &adapter->vf_data[i];
6489 
6490 		if ((vf_data->num_vf_mc_hashes > 30) ||
6491 		    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
6492 			vmolr |= E1000_VMOLR_MPME;
6493 		} else if (vf_data->num_vf_mc_hashes) {
6494 			vmolr |= E1000_VMOLR_ROMPE;
6495 			for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
6496 				igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
6497 		}
6498 		wr32(E1000_VMOLR(i), vmolr);
6499 	}
6500 }
6501 
6502 static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
6503 {
6504 	struct e1000_hw *hw = &adapter->hw;
6505 	u32 pool_mask, vlvf_mask, i;
6506 
6507 	/* create mask for VF and other pools */
6508 	pool_mask = E1000_VLVF_POOLSEL_MASK;
6509 	vlvf_mask = BIT(E1000_VLVF_POOLSEL_SHIFT + vf);
6510 
6511 	/* drop PF from pool bits */
6512 	pool_mask &= ~BIT(E1000_VLVF_POOLSEL_SHIFT +
6513 			     adapter->vfs_allocated_count);
6514 
6515 	/* Find the vlan filter for this id */
6516 	for (i = E1000_VLVF_ARRAY_SIZE; i--;) {
6517 		u32 vlvf = rd32(E1000_VLVF(i));
6518 		u32 vfta_mask, vid, vfta;
6519 
6520 		/* remove the vf from the pool */
6521 		if (!(vlvf & vlvf_mask))
6522 			continue;
6523 
6524 		/* clear out bit from VLVF */
6525 		vlvf ^= vlvf_mask;
6526 
6527 		/* if other pools are present, just remove ourselves */
6528 		if (vlvf & pool_mask)
6529 			goto update_vlvfb;
6530 
6531 		/* if PF is present, leave VFTA */
6532 		if (vlvf & E1000_VLVF_POOLSEL_MASK)
6533 			goto update_vlvf;
6534 
6535 		vid = vlvf & E1000_VLVF_VLANID_MASK;
6536 		vfta_mask = BIT(vid % 32);
6537 
6538 		/* clear bit from VFTA */
6539 		vfta = adapter->shadow_vfta[vid / 32];
6540 		if (vfta & vfta_mask)
6541 			hw->mac.ops.write_vfta(hw, vid / 32, vfta ^ vfta_mask);
6542 update_vlvf:
6543 		/* clear pool selection enable */
6544 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
6545 			vlvf &= E1000_VLVF_POOLSEL_MASK;
6546 		else
6547 			vlvf = 0;
6548 update_vlvfb:
6549 		/* clear pool bits */
6550 		wr32(E1000_VLVF(i), vlvf);
6551 	}
6552 }
6553 
6554 static int igb_find_vlvf_entry(struct e1000_hw *hw, u32 vlan)
6555 {
6556 	u32 vlvf;
6557 	int idx;
6558 
6559 	/* short cut the special case */
6560 	if (vlan == 0)
6561 		return 0;
6562 
6563 	/* Search for the VLAN id in the VLVF entries */
6564 	for (idx = E1000_VLVF_ARRAY_SIZE; --idx;) {
6565 		vlvf = rd32(E1000_VLVF(idx));
6566 		if ((vlvf & VLAN_VID_MASK) == vlan)
6567 			break;
6568 	}
6569 
6570 	return idx;
6571 }
6572 
6573 static void igb_update_pf_vlvf(struct igb_adapter *adapter, u32 vid)
6574 {
6575 	struct e1000_hw *hw = &adapter->hw;
6576 	u32 bits, pf_id;
6577 	int idx;
6578 
6579 	idx = igb_find_vlvf_entry(hw, vid);
6580 	if (!idx)
6581 		return;
6582 
6583 	/* See if any other pools are set for this VLAN filter
6584 	 * entry other than the PF.
6585 	 */
6586 	pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
6587 	bits = ~BIT(pf_id) & E1000_VLVF_POOLSEL_MASK;
6588 	bits &= rd32(E1000_VLVF(idx));
6589 
6590 	/* Disable the filter so this falls into the default pool. */
6591 	if (!bits) {
6592 		if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
6593 			wr32(E1000_VLVF(idx), BIT(pf_id));
6594 		else
6595 			wr32(E1000_VLVF(idx), 0);
6596 	}
6597 }
6598 
6599 static s32 igb_set_vf_vlan(struct igb_adapter *adapter, u32 vid,
6600 			   bool add, u32 vf)
6601 {
6602 	int pf_id = adapter->vfs_allocated_count;
6603 	struct e1000_hw *hw = &adapter->hw;
6604 	int err;
6605 
6606 	/* If VLAN overlaps with one the PF is currently monitoring make
6607 	 * sure that we are able to allocate a VLVF entry.  This may be
6608 	 * redundant but it guarantees PF will maintain visibility to
6609 	 * the VLAN.
6610 	 */
6611 	if (add && test_bit(vid, adapter->active_vlans)) {
6612 		err = igb_vfta_set(hw, vid, pf_id, true, false);
6613 		if (err)
6614 			return err;
6615 	}
6616 
6617 	err = igb_vfta_set(hw, vid, vf, add, false);
6618 
6619 	if (add && !err)
6620 		return err;
6621 
6622 	/* If we failed to add the VF VLAN or we are removing the VF VLAN
6623 	 * we may need to drop the PF pool bit in order to allow us to free
6624 	 * up the VLVF resources.
6625 	 */
6626 	if (test_bit(vid, adapter->active_vlans) ||
6627 	    (adapter->flags & IGB_FLAG_VLAN_PROMISC))
6628 		igb_update_pf_vlvf(adapter, vid);
6629 
6630 	return err;
6631 }
6632 
6633 static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
6634 {
6635 	struct e1000_hw *hw = &adapter->hw;
6636 
6637 	if (vid)
6638 		wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
6639 	else
6640 		wr32(E1000_VMVIR(vf), 0);
6641 }
6642 
6643 static int igb_enable_port_vlan(struct igb_adapter *adapter, int vf,
6644 				u16 vlan, u8 qos)
6645 {
6646 	int err;
6647 
6648 	err = igb_set_vf_vlan(adapter, vlan, true, vf);
6649 	if (err)
6650 		return err;
6651 
6652 	igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
6653 	igb_set_vmolr(adapter, vf, !vlan);
6654 
6655 	/* revoke access to previous VLAN */
6656 	if (vlan != adapter->vf_data[vf].pf_vlan)
6657 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
6658 				false, vf);
6659 
6660 	adapter->vf_data[vf].pf_vlan = vlan;
6661 	adapter->vf_data[vf].pf_qos = qos;
6662 	igb_set_vf_vlan_strip(adapter, vf, true);
6663 	dev_info(&adapter->pdev->dev,
6664 		 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
6665 	if (test_bit(__IGB_DOWN, &adapter->state)) {
6666 		dev_warn(&adapter->pdev->dev,
6667 			 "The VF VLAN has been set, but the PF device is not up.\n");
6668 		dev_warn(&adapter->pdev->dev,
6669 			 "Bring the PF device up before attempting to use the VF device.\n");
6670 	}
6671 
6672 	return err;
6673 }
6674 
6675 static int igb_disable_port_vlan(struct igb_adapter *adapter, int vf)
6676 {
6677 	/* Restore tagless access via VLAN 0 */
6678 	igb_set_vf_vlan(adapter, 0, true, vf);
6679 
6680 	igb_set_vmvir(adapter, 0, vf);
6681 	igb_set_vmolr(adapter, vf, true);
6682 
6683 	/* Remove any PF assigned VLAN */
6684 	if (adapter->vf_data[vf].pf_vlan)
6685 		igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
6686 				false, vf);
6687 
6688 	adapter->vf_data[vf].pf_vlan = 0;
6689 	adapter->vf_data[vf].pf_qos = 0;
6690 	igb_set_vf_vlan_strip(adapter, vf, false);
6691 
6692 	return 0;
6693 }
6694 
6695 static int igb_ndo_set_vf_vlan(struct net_device *netdev, int vf,
6696 			       u16 vlan, u8 qos, __be16 vlan_proto)
6697 {
6698 	struct igb_adapter *adapter = netdev_priv(netdev);
6699 
6700 	if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
6701 		return -EINVAL;
6702 
6703 	if (vlan_proto != htons(ETH_P_8021Q))
6704 		return -EPROTONOSUPPORT;
6705 
6706 	return (vlan || qos) ? igb_enable_port_vlan(adapter, vf, vlan, qos) :
6707 			       igb_disable_port_vlan(adapter, vf);
6708 }
6709 
6710 static int igb_set_vf_vlan_msg(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
6711 {
6712 	int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
6713 	int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
6714 	int ret;
6715 
6716 	if (adapter->vf_data[vf].pf_vlan)
6717 		return -1;
6718 
6719 	/* VLAN 0 is a special case, don't allow it to be removed */
6720 	if (!vid && !add)
6721 		return 0;
6722 
6723 	ret = igb_set_vf_vlan(adapter, vid, !!add, vf);
6724 	if (!ret)
6725 		igb_set_vf_vlan_strip(adapter, vf, !!vid);
6726 	return ret;
6727 }
6728 
6729 static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
6730 {
6731 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6732 
6733 	/* clear flags - except flag that indicates PF has set the MAC */
6734 	vf_data->flags &= IGB_VF_FLAG_PF_SET_MAC;
6735 	vf_data->last_nack = jiffies;
6736 
6737 	/* reset vlans for device */
6738 	igb_clear_vf_vfta(adapter, vf);
6739 	igb_set_vf_vlan(adapter, vf_data->pf_vlan, true, vf);
6740 	igb_set_vmvir(adapter, vf_data->pf_vlan |
6741 			       (vf_data->pf_qos << VLAN_PRIO_SHIFT), vf);
6742 	igb_set_vmolr(adapter, vf, !vf_data->pf_vlan);
6743 	igb_set_vf_vlan_strip(adapter, vf, !!(vf_data->pf_vlan));
6744 
6745 	/* reset multicast table array for vf */
6746 	adapter->vf_data[vf].num_vf_mc_hashes = 0;
6747 
6748 	/* Flush and reset the mta with the new values */
6749 	igb_set_rx_mode(adapter->netdev);
6750 }
6751 
6752 static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
6753 {
6754 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6755 
6756 	/* clear mac address as we were hotplug removed/added */
6757 	if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
6758 		eth_zero_addr(vf_mac);
6759 
6760 	/* process remaining reset events */
6761 	igb_vf_reset(adapter, vf);
6762 }
6763 
6764 static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
6765 {
6766 	struct e1000_hw *hw = &adapter->hw;
6767 	unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
6768 	u32 reg, msgbuf[3];
6769 	u8 *addr = (u8 *)(&msgbuf[1]);
6770 
6771 	/* process all the same items cleared in a function level reset */
6772 	igb_vf_reset(adapter, vf);
6773 
6774 	/* set vf mac address */
6775 	igb_set_vf_mac(adapter, vf, vf_mac);
6776 
6777 	/* enable transmit and receive for vf */
6778 	reg = rd32(E1000_VFTE);
6779 	wr32(E1000_VFTE, reg | BIT(vf));
6780 	reg = rd32(E1000_VFRE);
6781 	wr32(E1000_VFRE, reg | BIT(vf));
6782 
6783 	adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
6784 
6785 	/* reply to reset with ack and vf mac address */
6786 	if (!is_zero_ether_addr(vf_mac)) {
6787 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
6788 		memcpy(addr, vf_mac, ETH_ALEN);
6789 	} else {
6790 		msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
6791 	}
6792 	igb_write_mbx(hw, msgbuf, 3, vf);
6793 }
6794 
6795 static void igb_flush_mac_table(struct igb_adapter *adapter)
6796 {
6797 	struct e1000_hw *hw = &adapter->hw;
6798 	int i;
6799 
6800 	for (i = 0; i < hw->mac.rar_entry_count; i++) {
6801 		adapter->mac_table[i].state &= ~IGB_MAC_STATE_IN_USE;
6802 		memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
6803 		adapter->mac_table[i].queue = 0;
6804 		igb_rar_set_index(adapter, i);
6805 	}
6806 }
6807 
6808 static int igb_available_rars(struct igb_adapter *adapter, u8 queue)
6809 {
6810 	struct e1000_hw *hw = &adapter->hw;
6811 	/* do not count rar entries reserved for VFs MAC addresses */
6812 	int rar_entries = hw->mac.rar_entry_count -
6813 			  adapter->vfs_allocated_count;
6814 	int i, count = 0;
6815 
6816 	for (i = 0; i < rar_entries; i++) {
6817 		/* do not count default entries */
6818 		if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT)
6819 			continue;
6820 
6821 		/* do not count "in use" entries for different queues */
6822 		if ((adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE) &&
6823 		    (adapter->mac_table[i].queue != queue))
6824 			continue;
6825 
6826 		count++;
6827 	}
6828 
6829 	return count;
6830 }
6831 
6832 /* Set default MAC address for the PF in the first RAR entry */
6833 static void igb_set_default_mac_filter(struct igb_adapter *adapter)
6834 {
6835 	struct igb_mac_addr *mac_table = &adapter->mac_table[0];
6836 
6837 	ether_addr_copy(mac_table->addr, adapter->hw.mac.addr);
6838 	mac_table->queue = adapter->vfs_allocated_count;
6839 	mac_table->state = IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
6840 
6841 	igb_rar_set_index(adapter, 0);
6842 }
6843 
6844 static int igb_add_mac_filter(struct igb_adapter *adapter, const u8 *addr,
6845 			      const u8 queue)
6846 {
6847 	struct e1000_hw *hw = &adapter->hw;
6848 	int rar_entries = hw->mac.rar_entry_count -
6849 			  adapter->vfs_allocated_count;
6850 	int i;
6851 
6852 	if (is_zero_ether_addr(addr))
6853 		return -EINVAL;
6854 
6855 	/* Search for the first empty entry in the MAC table.
6856 	 * Do not touch entries at the end of the table reserved for the VF MAC
6857 	 * addresses.
6858 	 */
6859 	for (i = 0; i < rar_entries; i++) {
6860 		if (adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE)
6861 			continue;
6862 
6863 		ether_addr_copy(adapter->mac_table[i].addr, addr);
6864 		adapter->mac_table[i].queue = queue;
6865 		adapter->mac_table[i].state |= IGB_MAC_STATE_IN_USE;
6866 
6867 		igb_rar_set_index(adapter, i);
6868 		return i;
6869 	}
6870 
6871 	return -ENOSPC;
6872 }
6873 
6874 static int igb_del_mac_filter(struct igb_adapter *adapter, const u8 *addr,
6875 			      const u8 queue)
6876 {
6877 	struct e1000_hw *hw = &adapter->hw;
6878 	int rar_entries = hw->mac.rar_entry_count -
6879 			  adapter->vfs_allocated_count;
6880 	int i;
6881 
6882 	if (is_zero_ether_addr(addr))
6883 		return -EINVAL;
6884 
6885 	/* Search for matching entry in the MAC table based on given address
6886 	 * and queue. Do not touch entries at the end of the table reserved
6887 	 * for the VF MAC addresses.
6888 	 */
6889 	for (i = 0; i < rar_entries; i++) {
6890 		if (!(adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE))
6891 			continue;
6892 		if (adapter->mac_table[i].queue != queue)
6893 			continue;
6894 		if (!ether_addr_equal(adapter->mac_table[i].addr, addr))
6895 			continue;
6896 
6897 		adapter->mac_table[i].state &= ~IGB_MAC_STATE_IN_USE;
6898 		memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
6899 		adapter->mac_table[i].queue = 0;
6900 
6901 		igb_rar_set_index(adapter, i);
6902 		return 0;
6903 	}
6904 
6905 	return -ENOENT;
6906 }
6907 
6908 static int igb_uc_sync(struct net_device *netdev, const unsigned char *addr)
6909 {
6910 	struct igb_adapter *adapter = netdev_priv(netdev);
6911 	int ret;
6912 
6913 	ret = igb_add_mac_filter(adapter, addr, adapter->vfs_allocated_count);
6914 
6915 	return min_t(int, ret, 0);
6916 }
6917 
6918 static int igb_uc_unsync(struct net_device *netdev, const unsigned char *addr)
6919 {
6920 	struct igb_adapter *adapter = netdev_priv(netdev);
6921 
6922 	igb_del_mac_filter(adapter, addr, adapter->vfs_allocated_count);
6923 
6924 	return 0;
6925 }
6926 
6927 static int igb_set_vf_mac_filter(struct igb_adapter *adapter, const int vf,
6928 				 const u32 info, const u8 *addr)
6929 {
6930 	struct pci_dev *pdev = adapter->pdev;
6931 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6932 	struct list_head *pos;
6933 	struct vf_mac_filter *entry = NULL;
6934 	int ret = 0;
6935 
6936 	switch (info) {
6937 	case E1000_VF_MAC_FILTER_CLR:
6938 		/* remove all unicast MAC filters related to the current VF */
6939 		list_for_each(pos, &adapter->vf_macs.l) {
6940 			entry = list_entry(pos, struct vf_mac_filter, l);
6941 			if (entry->vf == vf) {
6942 				entry->vf = -1;
6943 				entry->free = true;
6944 				igb_del_mac_filter(adapter, entry->vf_mac, vf);
6945 			}
6946 		}
6947 		break;
6948 	case E1000_VF_MAC_FILTER_ADD:
6949 		if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
6950 		    !vf_data->trusted) {
6951 			dev_warn(&pdev->dev,
6952 				 "VF %d requested MAC filter but is administratively denied\n",
6953 				 vf);
6954 			return -EINVAL;
6955 		}
6956 		if (!is_valid_ether_addr(addr)) {
6957 			dev_warn(&pdev->dev,
6958 				 "VF %d attempted to set invalid MAC filter\n",
6959 				 vf);
6960 			return -EINVAL;
6961 		}
6962 
6963 		/* try to find empty slot in the list */
6964 		list_for_each(pos, &adapter->vf_macs.l) {
6965 			entry = list_entry(pos, struct vf_mac_filter, l);
6966 			if (entry->free)
6967 				break;
6968 		}
6969 
6970 		if (entry && entry->free) {
6971 			entry->free = false;
6972 			entry->vf = vf;
6973 			ether_addr_copy(entry->vf_mac, addr);
6974 
6975 			ret = igb_add_mac_filter(adapter, addr, vf);
6976 			ret = min_t(int, ret, 0);
6977 		} else {
6978 			ret = -ENOSPC;
6979 		}
6980 
6981 		if (ret == -ENOSPC)
6982 			dev_warn(&pdev->dev,
6983 				 "VF %d has requested MAC filter but there is no space for it\n",
6984 				 vf);
6985 		break;
6986 	default:
6987 		ret = -EINVAL;
6988 		break;
6989 	}
6990 
6991 	return ret;
6992 }
6993 
6994 static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
6995 {
6996 	struct pci_dev *pdev = adapter->pdev;
6997 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6998 	u32 info = msg[0] & E1000_VT_MSGINFO_MASK;
6999 
7000 	/* The VF MAC Address is stored in a packed array of bytes
7001 	 * starting at the second 32 bit word of the msg array
7002 	 */
7003 	unsigned char *addr = (unsigned char *)&msg[1];
7004 	int ret = 0;
7005 
7006 	if (!info) {
7007 		if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7008 		    !vf_data->trusted) {
7009 			dev_warn(&pdev->dev,
7010 				 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
7011 				 vf);
7012 			return -EINVAL;
7013 		}
7014 
7015 		if (!is_valid_ether_addr(addr)) {
7016 			dev_warn(&pdev->dev,
7017 				 "VF %d attempted to set invalid MAC\n",
7018 				 vf);
7019 			return -EINVAL;
7020 		}
7021 
7022 		ret = igb_set_vf_mac(adapter, vf, addr);
7023 	} else {
7024 		ret = igb_set_vf_mac_filter(adapter, vf, info, addr);
7025 	}
7026 
7027 	return ret;
7028 }
7029 
7030 static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
7031 {
7032 	struct e1000_hw *hw = &adapter->hw;
7033 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7034 	u32 msg = E1000_VT_MSGTYPE_NACK;
7035 
7036 	/* if device isn't clear to send it shouldn't be reading either */
7037 	if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
7038 	    time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
7039 		igb_write_mbx(hw, &msg, 1, vf);
7040 		vf_data->last_nack = jiffies;
7041 	}
7042 }
7043 
7044 static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
7045 {
7046 	struct pci_dev *pdev = adapter->pdev;
7047 	u32 msgbuf[E1000_VFMAILBOX_SIZE];
7048 	struct e1000_hw *hw = &adapter->hw;
7049 	struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7050 	s32 retval;
7051 
7052 	retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf, false);
7053 
7054 	if (retval) {
7055 		/* if receive failed revoke VF CTS stats and restart init */
7056 		dev_err(&pdev->dev, "Error receiving message from VF\n");
7057 		vf_data->flags &= ~IGB_VF_FLAG_CTS;
7058 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7059 			goto unlock;
7060 		goto out;
7061 	}
7062 
7063 	/* this is a message we already processed, do nothing */
7064 	if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
7065 		goto unlock;
7066 
7067 	/* until the vf completes a reset it should not be
7068 	 * allowed to start any configuration.
7069 	 */
7070 	if (msgbuf[0] == E1000_VF_RESET) {
7071 		/* unlocks mailbox */
7072 		igb_vf_reset_msg(adapter, vf);
7073 		return;
7074 	}
7075 
7076 	if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
7077 		if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7078 			goto unlock;
7079 		retval = -1;
7080 		goto out;
7081 	}
7082 
7083 	switch ((msgbuf[0] & 0xFFFF)) {
7084 	case E1000_VF_SET_MAC_ADDR:
7085 		retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
7086 		break;
7087 	case E1000_VF_SET_PROMISC:
7088 		retval = igb_set_vf_promisc(adapter, msgbuf, vf);
7089 		break;
7090 	case E1000_VF_SET_MULTICAST:
7091 		retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
7092 		break;
7093 	case E1000_VF_SET_LPE:
7094 		retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
7095 		break;
7096 	case E1000_VF_SET_VLAN:
7097 		retval = -1;
7098 		if (vf_data->pf_vlan)
7099 			dev_warn(&pdev->dev,
7100 				 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
7101 				 vf);
7102 		else
7103 			retval = igb_set_vf_vlan_msg(adapter, msgbuf, vf);
7104 		break;
7105 	default:
7106 		dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
7107 		retval = -1;
7108 		break;
7109 	}
7110 
7111 	msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
7112 out:
7113 	/* notify the VF of the results of what it sent us */
7114 	if (retval)
7115 		msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
7116 	else
7117 		msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
7118 
7119 	/* unlocks mailbox */
7120 	igb_write_mbx(hw, msgbuf, 1, vf);
7121 	return;
7122 
7123 unlock:
7124 	igb_unlock_mbx(hw, vf);
7125 }
7126 
7127 static void igb_msg_task(struct igb_adapter *adapter)
7128 {
7129 	struct e1000_hw *hw = &adapter->hw;
7130 	u32 vf;
7131 
7132 	for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
7133 		/* process any reset requests */
7134 		if (!igb_check_for_rst(hw, vf))
7135 			igb_vf_reset_event(adapter, vf);
7136 
7137 		/* process any messages pending */
7138 		if (!igb_check_for_msg(hw, vf))
7139 			igb_rcv_msg_from_vf(adapter, vf);
7140 
7141 		/* process any acks */
7142 		if (!igb_check_for_ack(hw, vf))
7143 			igb_rcv_ack_from_vf(adapter, vf);
7144 	}
7145 }
7146 
7147 /**
7148  *  igb_set_uta - Set unicast filter table address
7149  *  @adapter: board private structure
7150  *  @set: boolean indicating if we are setting or clearing bits
7151  *
7152  *  The unicast table address is a register array of 32-bit registers.
7153  *  The table is meant to be used in a way similar to how the MTA is used
7154  *  however due to certain limitations in the hardware it is necessary to
7155  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
7156  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
7157  **/
7158 static void igb_set_uta(struct igb_adapter *adapter, bool set)
7159 {
7160 	struct e1000_hw *hw = &adapter->hw;
7161 	u32 uta = set ? ~0 : 0;
7162 	int i;
7163 
7164 	/* we only need to do this if VMDq is enabled */
7165 	if (!adapter->vfs_allocated_count)
7166 		return;
7167 
7168 	for (i = hw->mac.uta_reg_count; i--;)
7169 		array_wr32(E1000_UTA, i, uta);
7170 }
7171 
7172 /**
7173  *  igb_intr_msi - Interrupt Handler
7174  *  @irq: interrupt number
7175  *  @data: pointer to a network interface device structure
7176  **/
7177 static irqreturn_t igb_intr_msi(int irq, void *data)
7178 {
7179 	struct igb_adapter *adapter = data;
7180 	struct igb_q_vector *q_vector = adapter->q_vector[0];
7181 	struct e1000_hw *hw = &adapter->hw;
7182 	/* read ICR disables interrupts using IAM */
7183 	u32 icr = rd32(E1000_ICR);
7184 
7185 	igb_write_itr(q_vector);
7186 
7187 	if (icr & E1000_ICR_DRSTA)
7188 		schedule_work(&adapter->reset_task);
7189 
7190 	if (icr & E1000_ICR_DOUTSYNC) {
7191 		/* HW is reporting DMA is out of sync */
7192 		adapter->stats.doosync++;
7193 	}
7194 
7195 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
7196 		hw->mac.get_link_status = 1;
7197 		if (!test_bit(__IGB_DOWN, &adapter->state))
7198 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
7199 	}
7200 
7201 	if (icr & E1000_ICR_TS)
7202 		igb_tsync_interrupt(adapter);
7203 
7204 	napi_schedule(&q_vector->napi);
7205 
7206 	return IRQ_HANDLED;
7207 }
7208 
7209 /**
7210  *  igb_intr - Legacy Interrupt Handler
7211  *  @irq: interrupt number
7212  *  @data: pointer to a network interface device structure
7213  **/
7214 static irqreturn_t igb_intr(int irq, void *data)
7215 {
7216 	struct igb_adapter *adapter = data;
7217 	struct igb_q_vector *q_vector = adapter->q_vector[0];
7218 	struct e1000_hw *hw = &adapter->hw;
7219 	/* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
7220 	 * need for the IMC write
7221 	 */
7222 	u32 icr = rd32(E1000_ICR);
7223 
7224 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
7225 	 * not set, then the adapter didn't send an interrupt
7226 	 */
7227 	if (!(icr & E1000_ICR_INT_ASSERTED))
7228 		return IRQ_NONE;
7229 
7230 	igb_write_itr(q_vector);
7231 
7232 	if (icr & E1000_ICR_DRSTA)
7233 		schedule_work(&adapter->reset_task);
7234 
7235 	if (icr & E1000_ICR_DOUTSYNC) {
7236 		/* HW is reporting DMA is out of sync */
7237 		adapter->stats.doosync++;
7238 	}
7239 
7240 	if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
7241 		hw->mac.get_link_status = 1;
7242 		/* guard against interrupt when we're going down */
7243 		if (!test_bit(__IGB_DOWN, &adapter->state))
7244 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
7245 	}
7246 
7247 	if (icr & E1000_ICR_TS)
7248 		igb_tsync_interrupt(adapter);
7249 
7250 	napi_schedule(&q_vector->napi);
7251 
7252 	return IRQ_HANDLED;
7253 }
7254 
7255 static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
7256 {
7257 	struct igb_adapter *adapter = q_vector->adapter;
7258 	struct e1000_hw *hw = &adapter->hw;
7259 
7260 	if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
7261 	    (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
7262 		if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
7263 			igb_set_itr(q_vector);
7264 		else
7265 			igb_update_ring_itr(q_vector);
7266 	}
7267 
7268 	if (!test_bit(__IGB_DOWN, &adapter->state)) {
7269 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
7270 			wr32(E1000_EIMS, q_vector->eims_value);
7271 		else
7272 			igb_irq_enable(adapter);
7273 	}
7274 }
7275 
7276 /**
7277  *  igb_poll - NAPI Rx polling callback
7278  *  @napi: napi polling structure
7279  *  @budget: count of how many packets we should handle
7280  **/
7281 static int igb_poll(struct napi_struct *napi, int budget)
7282 {
7283 	struct igb_q_vector *q_vector = container_of(napi,
7284 						     struct igb_q_vector,
7285 						     napi);
7286 	bool clean_complete = true;
7287 	int work_done = 0;
7288 
7289 #ifdef CONFIG_IGB_DCA
7290 	if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
7291 		igb_update_dca(q_vector);
7292 #endif
7293 	if (q_vector->tx.ring)
7294 		clean_complete = igb_clean_tx_irq(q_vector, budget);
7295 
7296 	if (q_vector->rx.ring) {
7297 		int cleaned = igb_clean_rx_irq(q_vector, budget);
7298 
7299 		work_done += cleaned;
7300 		if (cleaned >= budget)
7301 			clean_complete = false;
7302 	}
7303 
7304 	/* If all work not completed, return budget and keep polling */
7305 	if (!clean_complete)
7306 		return budget;
7307 
7308 	/* If not enough Rx work done, exit the polling mode */
7309 	napi_complete_done(napi, work_done);
7310 	igb_ring_irq_enable(q_vector);
7311 
7312 	return 0;
7313 }
7314 
7315 /**
7316  *  igb_clean_tx_irq - Reclaim resources after transmit completes
7317  *  @q_vector: pointer to q_vector containing needed info
7318  *  @napi_budget: Used to determine if we are in netpoll
7319  *
7320  *  returns true if ring is completely cleaned
7321  **/
7322 static bool igb_clean_tx_irq(struct igb_q_vector *q_vector, int napi_budget)
7323 {
7324 	struct igb_adapter *adapter = q_vector->adapter;
7325 	struct igb_ring *tx_ring = q_vector->tx.ring;
7326 	struct igb_tx_buffer *tx_buffer;
7327 	union e1000_adv_tx_desc *tx_desc;
7328 	unsigned int total_bytes = 0, total_packets = 0;
7329 	unsigned int budget = q_vector->tx.work_limit;
7330 	unsigned int i = tx_ring->next_to_clean;
7331 
7332 	if (test_bit(__IGB_DOWN, &adapter->state))
7333 		return true;
7334 
7335 	tx_buffer = &tx_ring->tx_buffer_info[i];
7336 	tx_desc = IGB_TX_DESC(tx_ring, i);
7337 	i -= tx_ring->count;
7338 
7339 	do {
7340 		union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
7341 
7342 		/* if next_to_watch is not set then there is no work pending */
7343 		if (!eop_desc)
7344 			break;
7345 
7346 		/* prevent any other reads prior to eop_desc */
7347 		smp_rmb();
7348 
7349 		/* if DD is not set pending work has not been completed */
7350 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
7351 			break;
7352 
7353 		/* clear next_to_watch to prevent false hangs */
7354 		tx_buffer->next_to_watch = NULL;
7355 
7356 		/* update the statistics for this packet */
7357 		total_bytes += tx_buffer->bytecount;
7358 		total_packets += tx_buffer->gso_segs;
7359 
7360 		/* free the skb */
7361 		napi_consume_skb(tx_buffer->skb, napi_budget);
7362 
7363 		/* unmap skb header data */
7364 		dma_unmap_single(tx_ring->dev,
7365 				 dma_unmap_addr(tx_buffer, dma),
7366 				 dma_unmap_len(tx_buffer, len),
7367 				 DMA_TO_DEVICE);
7368 
7369 		/* clear tx_buffer data */
7370 		dma_unmap_len_set(tx_buffer, len, 0);
7371 
7372 		/* clear last DMA location and unmap remaining buffers */
7373 		while (tx_desc != eop_desc) {
7374 			tx_buffer++;
7375 			tx_desc++;
7376 			i++;
7377 			if (unlikely(!i)) {
7378 				i -= tx_ring->count;
7379 				tx_buffer = tx_ring->tx_buffer_info;
7380 				tx_desc = IGB_TX_DESC(tx_ring, 0);
7381 			}
7382 
7383 			/* unmap any remaining paged data */
7384 			if (dma_unmap_len(tx_buffer, len)) {
7385 				dma_unmap_page(tx_ring->dev,
7386 					       dma_unmap_addr(tx_buffer, dma),
7387 					       dma_unmap_len(tx_buffer, len),
7388 					       DMA_TO_DEVICE);
7389 				dma_unmap_len_set(tx_buffer, len, 0);
7390 			}
7391 		}
7392 
7393 		/* move us one more past the eop_desc for start of next pkt */
7394 		tx_buffer++;
7395 		tx_desc++;
7396 		i++;
7397 		if (unlikely(!i)) {
7398 			i -= tx_ring->count;
7399 			tx_buffer = tx_ring->tx_buffer_info;
7400 			tx_desc = IGB_TX_DESC(tx_ring, 0);
7401 		}
7402 
7403 		/* issue prefetch for next Tx descriptor */
7404 		prefetch(tx_desc);
7405 
7406 		/* update budget accounting */
7407 		budget--;
7408 	} while (likely(budget));
7409 
7410 	netdev_tx_completed_queue(txring_txq(tx_ring),
7411 				  total_packets, total_bytes);
7412 	i += tx_ring->count;
7413 	tx_ring->next_to_clean = i;
7414 	u64_stats_update_begin(&tx_ring->tx_syncp);
7415 	tx_ring->tx_stats.bytes += total_bytes;
7416 	tx_ring->tx_stats.packets += total_packets;
7417 	u64_stats_update_end(&tx_ring->tx_syncp);
7418 	q_vector->tx.total_bytes += total_bytes;
7419 	q_vector->tx.total_packets += total_packets;
7420 
7421 	if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
7422 		struct e1000_hw *hw = &adapter->hw;
7423 
7424 		/* Detect a transmit hang in hardware, this serializes the
7425 		 * check with the clearing of time_stamp and movement of i
7426 		 */
7427 		clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
7428 		if (tx_buffer->next_to_watch &&
7429 		    time_after(jiffies, tx_buffer->time_stamp +
7430 			       (adapter->tx_timeout_factor * HZ)) &&
7431 		    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
7432 
7433 			/* detected Tx unit hang */
7434 			dev_err(tx_ring->dev,
7435 				"Detected Tx Unit Hang\n"
7436 				"  Tx Queue             <%d>\n"
7437 				"  TDH                  <%x>\n"
7438 				"  TDT                  <%x>\n"
7439 				"  next_to_use          <%x>\n"
7440 				"  next_to_clean        <%x>\n"
7441 				"buffer_info[next_to_clean]\n"
7442 				"  time_stamp           <%lx>\n"
7443 				"  next_to_watch        <%p>\n"
7444 				"  jiffies              <%lx>\n"
7445 				"  desc.status          <%x>\n",
7446 				tx_ring->queue_index,
7447 				rd32(E1000_TDH(tx_ring->reg_idx)),
7448 				readl(tx_ring->tail),
7449 				tx_ring->next_to_use,
7450 				tx_ring->next_to_clean,
7451 				tx_buffer->time_stamp,
7452 				tx_buffer->next_to_watch,
7453 				jiffies,
7454 				tx_buffer->next_to_watch->wb.status);
7455 			netif_stop_subqueue(tx_ring->netdev,
7456 					    tx_ring->queue_index);
7457 
7458 			/* we are about to reset, no point in enabling stuff */
7459 			return true;
7460 		}
7461 	}
7462 
7463 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
7464 	if (unlikely(total_packets &&
7465 	    netif_carrier_ok(tx_ring->netdev) &&
7466 	    igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
7467 		/* Make sure that anybody stopping the queue after this
7468 		 * sees the new next_to_clean.
7469 		 */
7470 		smp_mb();
7471 		if (__netif_subqueue_stopped(tx_ring->netdev,
7472 					     tx_ring->queue_index) &&
7473 		    !(test_bit(__IGB_DOWN, &adapter->state))) {
7474 			netif_wake_subqueue(tx_ring->netdev,
7475 					    tx_ring->queue_index);
7476 
7477 			u64_stats_update_begin(&tx_ring->tx_syncp);
7478 			tx_ring->tx_stats.restart_queue++;
7479 			u64_stats_update_end(&tx_ring->tx_syncp);
7480 		}
7481 	}
7482 
7483 	return !!budget;
7484 }
7485 
7486 /**
7487  *  igb_reuse_rx_page - page flip buffer and store it back on the ring
7488  *  @rx_ring: rx descriptor ring to store buffers on
7489  *  @old_buff: donor buffer to have page reused
7490  *
7491  *  Synchronizes page for reuse by the adapter
7492  **/
7493 static void igb_reuse_rx_page(struct igb_ring *rx_ring,
7494 			      struct igb_rx_buffer *old_buff)
7495 {
7496 	struct igb_rx_buffer *new_buff;
7497 	u16 nta = rx_ring->next_to_alloc;
7498 
7499 	new_buff = &rx_ring->rx_buffer_info[nta];
7500 
7501 	/* update, and store next to alloc */
7502 	nta++;
7503 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
7504 
7505 	/* Transfer page from old buffer to new buffer.
7506 	 * Move each member individually to avoid possible store
7507 	 * forwarding stalls.
7508 	 */
7509 	new_buff->dma		= old_buff->dma;
7510 	new_buff->page		= old_buff->page;
7511 	new_buff->page_offset	= old_buff->page_offset;
7512 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
7513 }
7514 
7515 static inline bool igb_page_is_reserved(struct page *page)
7516 {
7517 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
7518 }
7519 
7520 static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer)
7521 {
7522 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
7523 	struct page *page = rx_buffer->page;
7524 
7525 	/* avoid re-using remote pages */
7526 	if (unlikely(igb_page_is_reserved(page)))
7527 		return false;
7528 
7529 #if (PAGE_SIZE < 8192)
7530 	/* if we are only owner of page we can reuse it */
7531 	if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
7532 		return false;
7533 #else
7534 #define IGB_LAST_OFFSET \
7535 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IGB_RXBUFFER_2048)
7536 
7537 	if (rx_buffer->page_offset > IGB_LAST_OFFSET)
7538 		return false;
7539 #endif
7540 
7541 	/* If we have drained the page fragment pool we need to update
7542 	 * the pagecnt_bias and page count so that we fully restock the
7543 	 * number of references the driver holds.
7544 	 */
7545 	if (unlikely(!pagecnt_bias)) {
7546 		page_ref_add(page, USHRT_MAX);
7547 		rx_buffer->pagecnt_bias = USHRT_MAX;
7548 	}
7549 
7550 	return true;
7551 }
7552 
7553 /**
7554  *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
7555  *  @rx_ring: rx descriptor ring to transact packets on
7556  *  @rx_buffer: buffer containing page to add
7557  *  @skb: sk_buff to place the data into
7558  *  @size: size of buffer to be added
7559  *
7560  *  This function will add the data contained in rx_buffer->page to the skb.
7561  **/
7562 static void igb_add_rx_frag(struct igb_ring *rx_ring,
7563 			    struct igb_rx_buffer *rx_buffer,
7564 			    struct sk_buff *skb,
7565 			    unsigned int size)
7566 {
7567 #if (PAGE_SIZE < 8192)
7568 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
7569 #else
7570 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
7571 				SKB_DATA_ALIGN(IGB_SKB_PAD + size) :
7572 				SKB_DATA_ALIGN(size);
7573 #endif
7574 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
7575 			rx_buffer->page_offset, size, truesize);
7576 #if (PAGE_SIZE < 8192)
7577 	rx_buffer->page_offset ^= truesize;
7578 #else
7579 	rx_buffer->page_offset += truesize;
7580 #endif
7581 }
7582 
7583 static struct sk_buff *igb_construct_skb(struct igb_ring *rx_ring,
7584 					 struct igb_rx_buffer *rx_buffer,
7585 					 union e1000_adv_rx_desc *rx_desc,
7586 					 unsigned int size)
7587 {
7588 	void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
7589 #if (PAGE_SIZE < 8192)
7590 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
7591 #else
7592 	unsigned int truesize = SKB_DATA_ALIGN(size);
7593 #endif
7594 	unsigned int headlen;
7595 	struct sk_buff *skb;
7596 
7597 	/* prefetch first cache line of first page */
7598 	prefetch(va);
7599 #if L1_CACHE_BYTES < 128
7600 	prefetch(va + L1_CACHE_BYTES);
7601 #endif
7602 
7603 	/* allocate a skb to store the frags */
7604 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
7605 	if (unlikely(!skb))
7606 		return NULL;
7607 
7608 	if (unlikely(igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))) {
7609 		igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
7610 		va += IGB_TS_HDR_LEN;
7611 		size -= IGB_TS_HDR_LEN;
7612 	}
7613 
7614 	/* Determine available headroom for copy */
7615 	headlen = size;
7616 	if (headlen > IGB_RX_HDR_LEN)
7617 		headlen = eth_get_headlen(va, IGB_RX_HDR_LEN);
7618 
7619 	/* align pull length to size of long to optimize memcpy performance */
7620 	memcpy(__skb_put(skb, headlen), va, ALIGN(headlen, sizeof(long)));
7621 
7622 	/* update all of the pointers */
7623 	size -= headlen;
7624 	if (size) {
7625 		skb_add_rx_frag(skb, 0, rx_buffer->page,
7626 				(va + headlen) - page_address(rx_buffer->page),
7627 				size, truesize);
7628 #if (PAGE_SIZE < 8192)
7629 		rx_buffer->page_offset ^= truesize;
7630 #else
7631 		rx_buffer->page_offset += truesize;
7632 #endif
7633 	} else {
7634 		rx_buffer->pagecnt_bias++;
7635 	}
7636 
7637 	return skb;
7638 }
7639 
7640 static struct sk_buff *igb_build_skb(struct igb_ring *rx_ring,
7641 				     struct igb_rx_buffer *rx_buffer,
7642 				     union e1000_adv_rx_desc *rx_desc,
7643 				     unsigned int size)
7644 {
7645 	void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
7646 #if (PAGE_SIZE < 8192)
7647 	unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
7648 #else
7649 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
7650 				SKB_DATA_ALIGN(IGB_SKB_PAD + size);
7651 #endif
7652 	struct sk_buff *skb;
7653 
7654 	/* prefetch first cache line of first page */
7655 	prefetch(va);
7656 #if L1_CACHE_BYTES < 128
7657 	prefetch(va + L1_CACHE_BYTES);
7658 #endif
7659 
7660 	/* build an skb around the page buffer */
7661 	skb = build_skb(va - IGB_SKB_PAD, truesize);
7662 	if (unlikely(!skb))
7663 		return NULL;
7664 
7665 	/* update pointers within the skb to store the data */
7666 	skb_reserve(skb, IGB_SKB_PAD);
7667 	__skb_put(skb, size);
7668 
7669 	/* pull timestamp out of packet data */
7670 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
7671 		igb_ptp_rx_pktstamp(rx_ring->q_vector, skb->data, skb);
7672 		__skb_pull(skb, IGB_TS_HDR_LEN);
7673 	}
7674 
7675 	/* update buffer offset */
7676 #if (PAGE_SIZE < 8192)
7677 	rx_buffer->page_offset ^= truesize;
7678 #else
7679 	rx_buffer->page_offset += truesize;
7680 #endif
7681 
7682 	return skb;
7683 }
7684 
7685 static inline void igb_rx_checksum(struct igb_ring *ring,
7686 				   union e1000_adv_rx_desc *rx_desc,
7687 				   struct sk_buff *skb)
7688 {
7689 	skb_checksum_none_assert(skb);
7690 
7691 	/* Ignore Checksum bit is set */
7692 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
7693 		return;
7694 
7695 	/* Rx checksum disabled via ethtool */
7696 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
7697 		return;
7698 
7699 	/* TCP/UDP checksum error bit is set */
7700 	if (igb_test_staterr(rx_desc,
7701 			     E1000_RXDEXT_STATERR_TCPE |
7702 			     E1000_RXDEXT_STATERR_IPE)) {
7703 		/* work around errata with sctp packets where the TCPE aka
7704 		 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
7705 		 * packets, (aka let the stack check the crc32c)
7706 		 */
7707 		if (!((skb->len == 60) &&
7708 		      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
7709 			u64_stats_update_begin(&ring->rx_syncp);
7710 			ring->rx_stats.csum_err++;
7711 			u64_stats_update_end(&ring->rx_syncp);
7712 		}
7713 		/* let the stack verify checksum errors */
7714 		return;
7715 	}
7716 	/* It must be a TCP or UDP packet with a valid checksum */
7717 	if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
7718 				      E1000_RXD_STAT_UDPCS))
7719 		skb->ip_summed = CHECKSUM_UNNECESSARY;
7720 
7721 	dev_dbg(ring->dev, "cksum success: bits %08X\n",
7722 		le32_to_cpu(rx_desc->wb.upper.status_error));
7723 }
7724 
7725 static inline void igb_rx_hash(struct igb_ring *ring,
7726 			       union e1000_adv_rx_desc *rx_desc,
7727 			       struct sk_buff *skb)
7728 {
7729 	if (ring->netdev->features & NETIF_F_RXHASH)
7730 		skb_set_hash(skb,
7731 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
7732 			     PKT_HASH_TYPE_L3);
7733 }
7734 
7735 /**
7736  *  igb_is_non_eop - process handling of non-EOP buffers
7737  *  @rx_ring: Rx ring being processed
7738  *  @rx_desc: Rx descriptor for current buffer
7739  *  @skb: current socket buffer containing buffer in progress
7740  *
7741  *  This function updates next to clean.  If the buffer is an EOP buffer
7742  *  this function exits returning false, otherwise it will place the
7743  *  sk_buff in the next buffer to be chained and return true indicating
7744  *  that this is in fact a non-EOP buffer.
7745  **/
7746 static bool igb_is_non_eop(struct igb_ring *rx_ring,
7747 			   union e1000_adv_rx_desc *rx_desc)
7748 {
7749 	u32 ntc = rx_ring->next_to_clean + 1;
7750 
7751 	/* fetch, update, and store next to clean */
7752 	ntc = (ntc < rx_ring->count) ? ntc : 0;
7753 	rx_ring->next_to_clean = ntc;
7754 
7755 	prefetch(IGB_RX_DESC(rx_ring, ntc));
7756 
7757 	if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
7758 		return false;
7759 
7760 	return true;
7761 }
7762 
7763 /**
7764  *  igb_cleanup_headers - Correct corrupted or empty headers
7765  *  @rx_ring: rx descriptor ring packet is being transacted on
7766  *  @rx_desc: pointer to the EOP Rx descriptor
7767  *  @skb: pointer to current skb being fixed
7768  *
7769  *  Address the case where we are pulling data in on pages only
7770  *  and as such no data is present in the skb header.
7771  *
7772  *  In addition if skb is not at least 60 bytes we need to pad it so that
7773  *  it is large enough to qualify as a valid Ethernet frame.
7774  *
7775  *  Returns true if an error was encountered and skb was freed.
7776  **/
7777 static bool igb_cleanup_headers(struct igb_ring *rx_ring,
7778 				union e1000_adv_rx_desc *rx_desc,
7779 				struct sk_buff *skb)
7780 {
7781 	if (unlikely((igb_test_staterr(rx_desc,
7782 				       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
7783 		struct net_device *netdev = rx_ring->netdev;
7784 		if (!(netdev->features & NETIF_F_RXALL)) {
7785 			dev_kfree_skb_any(skb);
7786 			return true;
7787 		}
7788 	}
7789 
7790 	/* if eth_skb_pad returns an error the skb was freed */
7791 	if (eth_skb_pad(skb))
7792 		return true;
7793 
7794 	return false;
7795 }
7796 
7797 /**
7798  *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
7799  *  @rx_ring: rx descriptor ring packet is being transacted on
7800  *  @rx_desc: pointer to the EOP Rx descriptor
7801  *  @skb: pointer to current skb being populated
7802  *
7803  *  This function checks the ring, descriptor, and packet information in
7804  *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
7805  *  other fields within the skb.
7806  **/
7807 static void igb_process_skb_fields(struct igb_ring *rx_ring,
7808 				   union e1000_adv_rx_desc *rx_desc,
7809 				   struct sk_buff *skb)
7810 {
7811 	struct net_device *dev = rx_ring->netdev;
7812 
7813 	igb_rx_hash(rx_ring, rx_desc, skb);
7814 
7815 	igb_rx_checksum(rx_ring, rx_desc, skb);
7816 
7817 	if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
7818 	    !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
7819 		igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
7820 
7821 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
7822 	    igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
7823 		u16 vid;
7824 
7825 		if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
7826 		    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
7827 			vid = be16_to_cpu(rx_desc->wb.upper.vlan);
7828 		else
7829 			vid = le16_to_cpu(rx_desc->wb.upper.vlan);
7830 
7831 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
7832 	}
7833 
7834 	skb_record_rx_queue(skb, rx_ring->queue_index);
7835 
7836 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
7837 }
7838 
7839 static struct igb_rx_buffer *igb_get_rx_buffer(struct igb_ring *rx_ring,
7840 					       const unsigned int size)
7841 {
7842 	struct igb_rx_buffer *rx_buffer;
7843 
7844 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
7845 	prefetchw(rx_buffer->page);
7846 
7847 	/* we are reusing so sync this buffer for CPU use */
7848 	dma_sync_single_range_for_cpu(rx_ring->dev,
7849 				      rx_buffer->dma,
7850 				      rx_buffer->page_offset,
7851 				      size,
7852 				      DMA_FROM_DEVICE);
7853 
7854 	rx_buffer->pagecnt_bias--;
7855 
7856 	return rx_buffer;
7857 }
7858 
7859 static void igb_put_rx_buffer(struct igb_ring *rx_ring,
7860 			      struct igb_rx_buffer *rx_buffer)
7861 {
7862 	if (igb_can_reuse_rx_page(rx_buffer)) {
7863 		/* hand second half of page back to the ring */
7864 		igb_reuse_rx_page(rx_ring, rx_buffer);
7865 	} else {
7866 		/* We are not reusing the buffer so unmap it and free
7867 		 * any references we are holding to it
7868 		 */
7869 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
7870 				     igb_rx_pg_size(rx_ring), DMA_FROM_DEVICE,
7871 				     IGB_RX_DMA_ATTR);
7872 		__page_frag_cache_drain(rx_buffer->page,
7873 					rx_buffer->pagecnt_bias);
7874 	}
7875 
7876 	/* clear contents of rx_buffer */
7877 	rx_buffer->page = NULL;
7878 }
7879 
7880 static int igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
7881 {
7882 	struct igb_ring *rx_ring = q_vector->rx.ring;
7883 	struct sk_buff *skb = rx_ring->skb;
7884 	unsigned int total_bytes = 0, total_packets = 0;
7885 	u16 cleaned_count = igb_desc_unused(rx_ring);
7886 
7887 	while (likely(total_packets < budget)) {
7888 		union e1000_adv_rx_desc *rx_desc;
7889 		struct igb_rx_buffer *rx_buffer;
7890 		unsigned int size;
7891 
7892 		/* return some buffers to hardware, one at a time is too slow */
7893 		if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
7894 			igb_alloc_rx_buffers(rx_ring, cleaned_count);
7895 			cleaned_count = 0;
7896 		}
7897 
7898 		rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
7899 		size = le16_to_cpu(rx_desc->wb.upper.length);
7900 		if (!size)
7901 			break;
7902 
7903 		/* This memory barrier is needed to keep us from reading
7904 		 * any other fields out of the rx_desc until we know the
7905 		 * descriptor has been written back
7906 		 */
7907 		dma_rmb();
7908 
7909 		rx_buffer = igb_get_rx_buffer(rx_ring, size);
7910 
7911 		/* retrieve a buffer from the ring */
7912 		if (skb)
7913 			igb_add_rx_frag(rx_ring, rx_buffer, skb, size);
7914 		else if (ring_uses_build_skb(rx_ring))
7915 			skb = igb_build_skb(rx_ring, rx_buffer, rx_desc, size);
7916 		else
7917 			skb = igb_construct_skb(rx_ring, rx_buffer,
7918 						rx_desc, size);
7919 
7920 		/* exit if we failed to retrieve a buffer */
7921 		if (!skb) {
7922 			rx_ring->rx_stats.alloc_failed++;
7923 			rx_buffer->pagecnt_bias++;
7924 			break;
7925 		}
7926 
7927 		igb_put_rx_buffer(rx_ring, rx_buffer);
7928 		cleaned_count++;
7929 
7930 		/* fetch next buffer in frame if non-eop */
7931 		if (igb_is_non_eop(rx_ring, rx_desc))
7932 			continue;
7933 
7934 		/* verify the packet layout is correct */
7935 		if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
7936 			skb = NULL;
7937 			continue;
7938 		}
7939 
7940 		/* probably a little skewed due to removing CRC */
7941 		total_bytes += skb->len;
7942 
7943 		/* populate checksum, timestamp, VLAN, and protocol */
7944 		igb_process_skb_fields(rx_ring, rx_desc, skb);
7945 
7946 		napi_gro_receive(&q_vector->napi, skb);
7947 
7948 		/* reset skb pointer */
7949 		skb = NULL;
7950 
7951 		/* update budget accounting */
7952 		total_packets++;
7953 	}
7954 
7955 	/* place incomplete frames back on ring for completion */
7956 	rx_ring->skb = skb;
7957 
7958 	u64_stats_update_begin(&rx_ring->rx_syncp);
7959 	rx_ring->rx_stats.packets += total_packets;
7960 	rx_ring->rx_stats.bytes += total_bytes;
7961 	u64_stats_update_end(&rx_ring->rx_syncp);
7962 	q_vector->rx.total_packets += total_packets;
7963 	q_vector->rx.total_bytes += total_bytes;
7964 
7965 	if (cleaned_count)
7966 		igb_alloc_rx_buffers(rx_ring, cleaned_count);
7967 
7968 	return total_packets;
7969 }
7970 
7971 static inline unsigned int igb_rx_offset(struct igb_ring *rx_ring)
7972 {
7973 	return ring_uses_build_skb(rx_ring) ? IGB_SKB_PAD : 0;
7974 }
7975 
7976 static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
7977 				  struct igb_rx_buffer *bi)
7978 {
7979 	struct page *page = bi->page;
7980 	dma_addr_t dma;
7981 
7982 	/* since we are recycling buffers we should seldom need to alloc */
7983 	if (likely(page))
7984 		return true;
7985 
7986 	/* alloc new page for storage */
7987 	page = dev_alloc_pages(igb_rx_pg_order(rx_ring));
7988 	if (unlikely(!page)) {
7989 		rx_ring->rx_stats.alloc_failed++;
7990 		return false;
7991 	}
7992 
7993 	/* map page for use */
7994 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
7995 				 igb_rx_pg_size(rx_ring),
7996 				 DMA_FROM_DEVICE,
7997 				 IGB_RX_DMA_ATTR);
7998 
7999 	/* if mapping failed free memory back to system since
8000 	 * there isn't much point in holding memory we can't use
8001 	 */
8002 	if (dma_mapping_error(rx_ring->dev, dma)) {
8003 		__free_pages(page, igb_rx_pg_order(rx_ring));
8004 
8005 		rx_ring->rx_stats.alloc_failed++;
8006 		return false;
8007 	}
8008 
8009 	bi->dma = dma;
8010 	bi->page = page;
8011 	bi->page_offset = igb_rx_offset(rx_ring);
8012 	bi->pagecnt_bias = 1;
8013 
8014 	return true;
8015 }
8016 
8017 /**
8018  *  igb_alloc_rx_buffers - Replace used receive buffers; packet split
8019  *  @adapter: address of board private structure
8020  **/
8021 void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
8022 {
8023 	union e1000_adv_rx_desc *rx_desc;
8024 	struct igb_rx_buffer *bi;
8025 	u16 i = rx_ring->next_to_use;
8026 	u16 bufsz;
8027 
8028 	/* nothing to do */
8029 	if (!cleaned_count)
8030 		return;
8031 
8032 	rx_desc = IGB_RX_DESC(rx_ring, i);
8033 	bi = &rx_ring->rx_buffer_info[i];
8034 	i -= rx_ring->count;
8035 
8036 	bufsz = igb_rx_bufsz(rx_ring);
8037 
8038 	do {
8039 		if (!igb_alloc_mapped_page(rx_ring, bi))
8040 			break;
8041 
8042 		/* sync the buffer for use by the device */
8043 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
8044 						 bi->page_offset, bufsz,
8045 						 DMA_FROM_DEVICE);
8046 
8047 		/* Refresh the desc even if buffer_addrs didn't change
8048 		 * because each write-back erases this info.
8049 		 */
8050 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
8051 
8052 		rx_desc++;
8053 		bi++;
8054 		i++;
8055 		if (unlikely(!i)) {
8056 			rx_desc = IGB_RX_DESC(rx_ring, 0);
8057 			bi = rx_ring->rx_buffer_info;
8058 			i -= rx_ring->count;
8059 		}
8060 
8061 		/* clear the length for the next_to_use descriptor */
8062 		rx_desc->wb.upper.length = 0;
8063 
8064 		cleaned_count--;
8065 	} while (cleaned_count);
8066 
8067 	i += rx_ring->count;
8068 
8069 	if (rx_ring->next_to_use != i) {
8070 		/* record the next descriptor to use */
8071 		rx_ring->next_to_use = i;
8072 
8073 		/* update next to alloc since we have filled the ring */
8074 		rx_ring->next_to_alloc = i;
8075 
8076 		/* Force memory writes to complete before letting h/w
8077 		 * know there are new descriptors to fetch.  (Only
8078 		 * applicable for weak-ordered memory model archs,
8079 		 * such as IA-64).
8080 		 */
8081 		wmb();
8082 		writel(i, rx_ring->tail);
8083 	}
8084 }
8085 
8086 /**
8087  * igb_mii_ioctl -
8088  * @netdev:
8089  * @ifreq:
8090  * @cmd:
8091  **/
8092 static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
8093 {
8094 	struct igb_adapter *adapter = netdev_priv(netdev);
8095 	struct mii_ioctl_data *data = if_mii(ifr);
8096 
8097 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
8098 		return -EOPNOTSUPP;
8099 
8100 	switch (cmd) {
8101 	case SIOCGMIIPHY:
8102 		data->phy_id = adapter->hw.phy.addr;
8103 		break;
8104 	case SIOCGMIIREG:
8105 		if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
8106 				     &data->val_out))
8107 			return -EIO;
8108 		break;
8109 	case SIOCSMIIREG:
8110 	default:
8111 		return -EOPNOTSUPP;
8112 	}
8113 	return 0;
8114 }
8115 
8116 /**
8117  * igb_ioctl -
8118  * @netdev:
8119  * @ifreq:
8120  * @cmd:
8121  **/
8122 static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
8123 {
8124 	switch (cmd) {
8125 	case SIOCGMIIPHY:
8126 	case SIOCGMIIREG:
8127 	case SIOCSMIIREG:
8128 		return igb_mii_ioctl(netdev, ifr, cmd);
8129 	case SIOCGHWTSTAMP:
8130 		return igb_ptp_get_ts_config(netdev, ifr);
8131 	case SIOCSHWTSTAMP:
8132 		return igb_ptp_set_ts_config(netdev, ifr);
8133 	default:
8134 		return -EOPNOTSUPP;
8135 	}
8136 }
8137 
8138 void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
8139 {
8140 	struct igb_adapter *adapter = hw->back;
8141 
8142 	pci_read_config_word(adapter->pdev, reg, value);
8143 }
8144 
8145 void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
8146 {
8147 	struct igb_adapter *adapter = hw->back;
8148 
8149 	pci_write_config_word(adapter->pdev, reg, *value);
8150 }
8151 
8152 s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
8153 {
8154 	struct igb_adapter *adapter = hw->back;
8155 
8156 	if (pcie_capability_read_word(adapter->pdev, reg, value))
8157 		return -E1000_ERR_CONFIG;
8158 
8159 	return 0;
8160 }
8161 
8162 s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
8163 {
8164 	struct igb_adapter *adapter = hw->back;
8165 
8166 	if (pcie_capability_write_word(adapter->pdev, reg, *value))
8167 		return -E1000_ERR_CONFIG;
8168 
8169 	return 0;
8170 }
8171 
8172 static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
8173 {
8174 	struct igb_adapter *adapter = netdev_priv(netdev);
8175 	struct e1000_hw *hw = &adapter->hw;
8176 	u32 ctrl, rctl;
8177 	bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
8178 
8179 	if (enable) {
8180 		/* enable VLAN tag insert/strip */
8181 		ctrl = rd32(E1000_CTRL);
8182 		ctrl |= E1000_CTRL_VME;
8183 		wr32(E1000_CTRL, ctrl);
8184 
8185 		/* Disable CFI check */
8186 		rctl = rd32(E1000_RCTL);
8187 		rctl &= ~E1000_RCTL_CFIEN;
8188 		wr32(E1000_RCTL, rctl);
8189 	} else {
8190 		/* disable VLAN tag insert/strip */
8191 		ctrl = rd32(E1000_CTRL);
8192 		ctrl &= ~E1000_CTRL_VME;
8193 		wr32(E1000_CTRL, ctrl);
8194 	}
8195 
8196 	igb_set_vf_vlan_strip(adapter, adapter->vfs_allocated_count, enable);
8197 }
8198 
8199 static int igb_vlan_rx_add_vid(struct net_device *netdev,
8200 			       __be16 proto, u16 vid)
8201 {
8202 	struct igb_adapter *adapter = netdev_priv(netdev);
8203 	struct e1000_hw *hw = &adapter->hw;
8204 	int pf_id = adapter->vfs_allocated_count;
8205 
8206 	/* add the filter since PF can receive vlans w/o entry in vlvf */
8207 	if (!vid || !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
8208 		igb_vfta_set(hw, vid, pf_id, true, !!vid);
8209 
8210 	set_bit(vid, adapter->active_vlans);
8211 
8212 	return 0;
8213 }
8214 
8215 static int igb_vlan_rx_kill_vid(struct net_device *netdev,
8216 				__be16 proto, u16 vid)
8217 {
8218 	struct igb_adapter *adapter = netdev_priv(netdev);
8219 	int pf_id = adapter->vfs_allocated_count;
8220 	struct e1000_hw *hw = &adapter->hw;
8221 
8222 	/* remove VID from filter table */
8223 	if (vid && !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
8224 		igb_vfta_set(hw, vid, pf_id, false, true);
8225 
8226 	clear_bit(vid, adapter->active_vlans);
8227 
8228 	return 0;
8229 }
8230 
8231 static void igb_restore_vlan(struct igb_adapter *adapter)
8232 {
8233 	u16 vid = 1;
8234 
8235 	igb_vlan_mode(adapter->netdev, adapter->netdev->features);
8236 	igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
8237 
8238 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
8239 		igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
8240 }
8241 
8242 int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
8243 {
8244 	struct pci_dev *pdev = adapter->pdev;
8245 	struct e1000_mac_info *mac = &adapter->hw.mac;
8246 
8247 	mac->autoneg = 0;
8248 
8249 	/* Make sure dplx is at most 1 bit and lsb of speed is not set
8250 	 * for the switch() below to work
8251 	 */
8252 	if ((spd & 1) || (dplx & ~1))
8253 		goto err_inval;
8254 
8255 	/* Fiber NIC's only allow 1000 gbps Full duplex
8256 	 * and 100Mbps Full duplex for 100baseFx sfp
8257 	 */
8258 	if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
8259 		switch (spd + dplx) {
8260 		case SPEED_10 + DUPLEX_HALF:
8261 		case SPEED_10 + DUPLEX_FULL:
8262 		case SPEED_100 + DUPLEX_HALF:
8263 			goto err_inval;
8264 		default:
8265 			break;
8266 		}
8267 	}
8268 
8269 	switch (spd + dplx) {
8270 	case SPEED_10 + DUPLEX_HALF:
8271 		mac->forced_speed_duplex = ADVERTISE_10_HALF;
8272 		break;
8273 	case SPEED_10 + DUPLEX_FULL:
8274 		mac->forced_speed_duplex = ADVERTISE_10_FULL;
8275 		break;
8276 	case SPEED_100 + DUPLEX_HALF:
8277 		mac->forced_speed_duplex = ADVERTISE_100_HALF;
8278 		break;
8279 	case SPEED_100 + DUPLEX_FULL:
8280 		mac->forced_speed_duplex = ADVERTISE_100_FULL;
8281 		break;
8282 	case SPEED_1000 + DUPLEX_FULL:
8283 		mac->autoneg = 1;
8284 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
8285 		break;
8286 	case SPEED_1000 + DUPLEX_HALF: /* not supported */
8287 	default:
8288 		goto err_inval;
8289 	}
8290 
8291 	/* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
8292 	adapter->hw.phy.mdix = AUTO_ALL_MODES;
8293 
8294 	return 0;
8295 
8296 err_inval:
8297 	dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
8298 	return -EINVAL;
8299 }
8300 
8301 static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
8302 			  bool runtime)
8303 {
8304 	struct net_device *netdev = pci_get_drvdata(pdev);
8305 	struct igb_adapter *adapter = netdev_priv(netdev);
8306 	struct e1000_hw *hw = &adapter->hw;
8307 	u32 ctrl, rctl, status;
8308 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
8309 #ifdef CONFIG_PM
8310 	int retval = 0;
8311 #endif
8312 
8313 	rtnl_lock();
8314 	netif_device_detach(netdev);
8315 
8316 	if (netif_running(netdev))
8317 		__igb_close(netdev, true);
8318 
8319 	igb_ptp_suspend(adapter);
8320 
8321 	igb_clear_interrupt_scheme(adapter);
8322 	rtnl_unlock();
8323 
8324 #ifdef CONFIG_PM
8325 	retval = pci_save_state(pdev);
8326 	if (retval)
8327 		return retval;
8328 #endif
8329 
8330 	status = rd32(E1000_STATUS);
8331 	if (status & E1000_STATUS_LU)
8332 		wufc &= ~E1000_WUFC_LNKC;
8333 
8334 	if (wufc) {
8335 		igb_setup_rctl(adapter);
8336 		igb_set_rx_mode(netdev);
8337 
8338 		/* turn on all-multi mode if wake on multicast is enabled */
8339 		if (wufc & E1000_WUFC_MC) {
8340 			rctl = rd32(E1000_RCTL);
8341 			rctl |= E1000_RCTL_MPE;
8342 			wr32(E1000_RCTL, rctl);
8343 		}
8344 
8345 		ctrl = rd32(E1000_CTRL);
8346 		/* advertise wake from D3Cold */
8347 		#define E1000_CTRL_ADVD3WUC 0x00100000
8348 		/* phy power management enable */
8349 		#define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
8350 		ctrl |= E1000_CTRL_ADVD3WUC;
8351 		wr32(E1000_CTRL, ctrl);
8352 
8353 		/* Allow time for pending master requests to run */
8354 		igb_disable_pcie_master(hw);
8355 
8356 		wr32(E1000_WUC, E1000_WUC_PME_EN);
8357 		wr32(E1000_WUFC, wufc);
8358 	} else {
8359 		wr32(E1000_WUC, 0);
8360 		wr32(E1000_WUFC, 0);
8361 	}
8362 
8363 	*enable_wake = wufc || adapter->en_mng_pt;
8364 	if (!*enable_wake)
8365 		igb_power_down_link(adapter);
8366 	else
8367 		igb_power_up_link(adapter);
8368 
8369 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
8370 	 * would have already happened in close and is redundant.
8371 	 */
8372 	igb_release_hw_control(adapter);
8373 
8374 	pci_disable_device(pdev);
8375 
8376 	return 0;
8377 }
8378 
8379 static void igb_deliver_wake_packet(struct net_device *netdev)
8380 {
8381 	struct igb_adapter *adapter = netdev_priv(netdev);
8382 	struct e1000_hw *hw = &adapter->hw;
8383 	struct sk_buff *skb;
8384 	u32 wupl;
8385 
8386 	wupl = rd32(E1000_WUPL) & E1000_WUPL_MASK;
8387 
8388 	/* WUPM stores only the first 128 bytes of the wake packet.
8389 	 * Read the packet only if we have the whole thing.
8390 	 */
8391 	if ((wupl == 0) || (wupl > E1000_WUPM_BYTES))
8392 		return;
8393 
8394 	skb = netdev_alloc_skb_ip_align(netdev, E1000_WUPM_BYTES);
8395 	if (!skb)
8396 		return;
8397 
8398 	skb_put(skb, wupl);
8399 
8400 	/* Ensure reads are 32-bit aligned */
8401 	wupl = roundup(wupl, 4);
8402 
8403 	memcpy_fromio(skb->data, hw->hw_addr + E1000_WUPM_REG(0), wupl);
8404 
8405 	skb->protocol = eth_type_trans(skb, netdev);
8406 	netif_rx(skb);
8407 }
8408 
8409 static int __maybe_unused igb_suspend(struct device *dev)
8410 {
8411 	int retval;
8412 	bool wake;
8413 	struct pci_dev *pdev = to_pci_dev(dev);
8414 
8415 	retval = __igb_shutdown(pdev, &wake, 0);
8416 	if (retval)
8417 		return retval;
8418 
8419 	if (wake) {
8420 		pci_prepare_to_sleep(pdev);
8421 	} else {
8422 		pci_wake_from_d3(pdev, false);
8423 		pci_set_power_state(pdev, PCI_D3hot);
8424 	}
8425 
8426 	return 0;
8427 }
8428 
8429 static int __maybe_unused igb_resume(struct device *dev)
8430 {
8431 	struct pci_dev *pdev = to_pci_dev(dev);
8432 	struct net_device *netdev = pci_get_drvdata(pdev);
8433 	struct igb_adapter *adapter = netdev_priv(netdev);
8434 	struct e1000_hw *hw = &adapter->hw;
8435 	u32 err, val;
8436 
8437 	pci_set_power_state(pdev, PCI_D0);
8438 	pci_restore_state(pdev);
8439 	pci_save_state(pdev);
8440 
8441 	if (!pci_device_is_present(pdev))
8442 		return -ENODEV;
8443 	err = pci_enable_device_mem(pdev);
8444 	if (err) {
8445 		dev_err(&pdev->dev,
8446 			"igb: Cannot enable PCI device from suspend\n");
8447 		return err;
8448 	}
8449 	pci_set_master(pdev);
8450 
8451 	pci_enable_wake(pdev, PCI_D3hot, 0);
8452 	pci_enable_wake(pdev, PCI_D3cold, 0);
8453 
8454 	if (igb_init_interrupt_scheme(adapter, true)) {
8455 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8456 		return -ENOMEM;
8457 	}
8458 
8459 	igb_reset(adapter);
8460 
8461 	/* let the f/w know that the h/w is now under the control of the
8462 	 * driver.
8463 	 */
8464 	igb_get_hw_control(adapter);
8465 
8466 	val = rd32(E1000_WUS);
8467 	if (val & WAKE_PKT_WUS)
8468 		igb_deliver_wake_packet(netdev);
8469 
8470 	wr32(E1000_WUS, ~0);
8471 
8472 	rtnl_lock();
8473 	if (!err && netif_running(netdev))
8474 		err = __igb_open(netdev, true);
8475 
8476 	if (!err)
8477 		netif_device_attach(netdev);
8478 	rtnl_unlock();
8479 
8480 	return err;
8481 }
8482 
8483 static int __maybe_unused igb_runtime_idle(struct device *dev)
8484 {
8485 	struct pci_dev *pdev = to_pci_dev(dev);
8486 	struct net_device *netdev = pci_get_drvdata(pdev);
8487 	struct igb_adapter *adapter = netdev_priv(netdev);
8488 
8489 	if (!igb_has_link(adapter))
8490 		pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
8491 
8492 	return -EBUSY;
8493 }
8494 
8495 static int __maybe_unused igb_runtime_suspend(struct device *dev)
8496 {
8497 	struct pci_dev *pdev = to_pci_dev(dev);
8498 	int retval;
8499 	bool wake;
8500 
8501 	retval = __igb_shutdown(pdev, &wake, 1);
8502 	if (retval)
8503 		return retval;
8504 
8505 	if (wake) {
8506 		pci_prepare_to_sleep(pdev);
8507 	} else {
8508 		pci_wake_from_d3(pdev, false);
8509 		pci_set_power_state(pdev, PCI_D3hot);
8510 	}
8511 
8512 	return 0;
8513 }
8514 
8515 static int __maybe_unused igb_runtime_resume(struct device *dev)
8516 {
8517 	return igb_resume(dev);
8518 }
8519 
8520 static void igb_shutdown(struct pci_dev *pdev)
8521 {
8522 	bool wake;
8523 
8524 	__igb_shutdown(pdev, &wake, 0);
8525 
8526 	if (system_state == SYSTEM_POWER_OFF) {
8527 		pci_wake_from_d3(pdev, wake);
8528 		pci_set_power_state(pdev, PCI_D3hot);
8529 	}
8530 }
8531 
8532 #ifdef CONFIG_PCI_IOV
8533 static int igb_sriov_reinit(struct pci_dev *dev)
8534 {
8535 	struct net_device *netdev = pci_get_drvdata(dev);
8536 	struct igb_adapter *adapter = netdev_priv(netdev);
8537 	struct pci_dev *pdev = adapter->pdev;
8538 
8539 	rtnl_lock();
8540 
8541 	if (netif_running(netdev))
8542 		igb_close(netdev);
8543 	else
8544 		igb_reset(adapter);
8545 
8546 	igb_clear_interrupt_scheme(adapter);
8547 
8548 	igb_init_queue_configuration(adapter);
8549 
8550 	if (igb_init_interrupt_scheme(adapter, true)) {
8551 		rtnl_unlock();
8552 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8553 		return -ENOMEM;
8554 	}
8555 
8556 	if (netif_running(netdev))
8557 		igb_open(netdev);
8558 
8559 	rtnl_unlock();
8560 
8561 	return 0;
8562 }
8563 
8564 static int igb_pci_disable_sriov(struct pci_dev *dev)
8565 {
8566 	int err = igb_disable_sriov(dev);
8567 
8568 	if (!err)
8569 		err = igb_sriov_reinit(dev);
8570 
8571 	return err;
8572 }
8573 
8574 static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
8575 {
8576 	int err = igb_enable_sriov(dev, num_vfs);
8577 
8578 	if (err)
8579 		goto out;
8580 
8581 	err = igb_sriov_reinit(dev);
8582 	if (!err)
8583 		return num_vfs;
8584 
8585 out:
8586 	return err;
8587 }
8588 
8589 #endif
8590 static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
8591 {
8592 #ifdef CONFIG_PCI_IOV
8593 	if (num_vfs == 0)
8594 		return igb_pci_disable_sriov(dev);
8595 	else
8596 		return igb_pci_enable_sriov(dev, num_vfs);
8597 #endif
8598 	return 0;
8599 }
8600 
8601 #ifdef CONFIG_NET_POLL_CONTROLLER
8602 /* Polling 'interrupt' - used by things like netconsole to send skbs
8603  * without having to re-enable interrupts. It's not called while
8604  * the interrupt routine is executing.
8605  */
8606 static void igb_netpoll(struct net_device *netdev)
8607 {
8608 	struct igb_adapter *adapter = netdev_priv(netdev);
8609 	struct e1000_hw *hw = &adapter->hw;
8610 	struct igb_q_vector *q_vector;
8611 	int i;
8612 
8613 	for (i = 0; i < adapter->num_q_vectors; i++) {
8614 		q_vector = adapter->q_vector[i];
8615 		if (adapter->flags & IGB_FLAG_HAS_MSIX)
8616 			wr32(E1000_EIMC, q_vector->eims_value);
8617 		else
8618 			igb_irq_disable(adapter);
8619 		napi_schedule(&q_vector->napi);
8620 	}
8621 }
8622 #endif /* CONFIG_NET_POLL_CONTROLLER */
8623 
8624 /**
8625  *  igb_io_error_detected - called when PCI error is detected
8626  *  @pdev: Pointer to PCI device
8627  *  @state: The current pci connection state
8628  *
8629  *  This function is called after a PCI bus error affecting
8630  *  this device has been detected.
8631  **/
8632 static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
8633 					      pci_channel_state_t state)
8634 {
8635 	struct net_device *netdev = pci_get_drvdata(pdev);
8636 	struct igb_adapter *adapter = netdev_priv(netdev);
8637 
8638 	netif_device_detach(netdev);
8639 
8640 	if (state == pci_channel_io_perm_failure)
8641 		return PCI_ERS_RESULT_DISCONNECT;
8642 
8643 	if (netif_running(netdev))
8644 		igb_down(adapter);
8645 	pci_disable_device(pdev);
8646 
8647 	/* Request a slot slot reset. */
8648 	return PCI_ERS_RESULT_NEED_RESET;
8649 }
8650 
8651 /**
8652  *  igb_io_slot_reset - called after the pci bus has been reset.
8653  *  @pdev: Pointer to PCI device
8654  *
8655  *  Restart the card from scratch, as if from a cold-boot. Implementation
8656  *  resembles the first-half of the igb_resume routine.
8657  **/
8658 static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
8659 {
8660 	struct net_device *netdev = pci_get_drvdata(pdev);
8661 	struct igb_adapter *adapter = netdev_priv(netdev);
8662 	struct e1000_hw *hw = &adapter->hw;
8663 	pci_ers_result_t result;
8664 	int err;
8665 
8666 	if (pci_enable_device_mem(pdev)) {
8667 		dev_err(&pdev->dev,
8668 			"Cannot re-enable PCI device after reset.\n");
8669 		result = PCI_ERS_RESULT_DISCONNECT;
8670 	} else {
8671 		pci_set_master(pdev);
8672 		pci_restore_state(pdev);
8673 		pci_save_state(pdev);
8674 
8675 		pci_enable_wake(pdev, PCI_D3hot, 0);
8676 		pci_enable_wake(pdev, PCI_D3cold, 0);
8677 
8678 		/* In case of PCI error, adapter lose its HW address
8679 		 * so we should re-assign it here.
8680 		 */
8681 		hw->hw_addr = adapter->io_addr;
8682 
8683 		igb_reset(adapter);
8684 		wr32(E1000_WUS, ~0);
8685 		result = PCI_ERS_RESULT_RECOVERED;
8686 	}
8687 
8688 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
8689 	if (err) {
8690 		dev_err(&pdev->dev,
8691 			"pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
8692 			err);
8693 		/* non-fatal, continue */
8694 	}
8695 
8696 	return result;
8697 }
8698 
8699 /**
8700  *  igb_io_resume - called when traffic can start flowing again.
8701  *  @pdev: Pointer to PCI device
8702  *
8703  *  This callback is called when the error recovery driver tells us that
8704  *  its OK to resume normal operation. Implementation resembles the
8705  *  second-half of the igb_resume routine.
8706  */
8707 static void igb_io_resume(struct pci_dev *pdev)
8708 {
8709 	struct net_device *netdev = pci_get_drvdata(pdev);
8710 	struct igb_adapter *adapter = netdev_priv(netdev);
8711 
8712 	if (netif_running(netdev)) {
8713 		if (igb_up(adapter)) {
8714 			dev_err(&pdev->dev, "igb_up failed after reset\n");
8715 			return;
8716 		}
8717 	}
8718 
8719 	netif_device_attach(netdev);
8720 
8721 	/* let the f/w know that the h/w is now under the control of the
8722 	 * driver.
8723 	 */
8724 	igb_get_hw_control(adapter);
8725 }
8726 
8727 /**
8728  *  igb_rar_set_index - Sync RAL[index] and RAH[index] registers with MAC table
8729  *  @adapter: Pointer to adapter structure
8730  *  @index: Index of the RAR entry which need to be synced with MAC table
8731  **/
8732 static void igb_rar_set_index(struct igb_adapter *adapter, u32 index)
8733 {
8734 	struct e1000_hw *hw = &adapter->hw;
8735 	u32 rar_low, rar_high;
8736 	u8 *addr = adapter->mac_table[index].addr;
8737 
8738 	/* HW expects these to be in network order when they are plugged
8739 	 * into the registers which are little endian.  In order to guarantee
8740 	 * that ordering we need to do an leXX_to_cpup here in order to be
8741 	 * ready for the byteswap that occurs with writel
8742 	 */
8743 	rar_low = le32_to_cpup((__le32 *)(addr));
8744 	rar_high = le16_to_cpup((__le16 *)(addr + 4));
8745 
8746 	/* Indicate to hardware the Address is Valid. */
8747 	if (adapter->mac_table[index].state & IGB_MAC_STATE_IN_USE) {
8748 		if (is_valid_ether_addr(addr))
8749 			rar_high |= E1000_RAH_AV;
8750 
8751 		if (hw->mac.type == e1000_82575)
8752 			rar_high |= E1000_RAH_POOL_1 *
8753 				    adapter->mac_table[index].queue;
8754 		else
8755 			rar_high |= E1000_RAH_POOL_1 <<
8756 				    adapter->mac_table[index].queue;
8757 	}
8758 
8759 	wr32(E1000_RAL(index), rar_low);
8760 	wrfl();
8761 	wr32(E1000_RAH(index), rar_high);
8762 	wrfl();
8763 }
8764 
8765 static int igb_set_vf_mac(struct igb_adapter *adapter,
8766 			  int vf, unsigned char *mac_addr)
8767 {
8768 	struct e1000_hw *hw = &adapter->hw;
8769 	/* VF MAC addresses start at end of receive addresses and moves
8770 	 * towards the first, as a result a collision should not be possible
8771 	 */
8772 	int rar_entry = hw->mac.rar_entry_count - (vf + 1);
8773 	unsigned char *vf_mac_addr = adapter->vf_data[vf].vf_mac_addresses;
8774 
8775 	ether_addr_copy(vf_mac_addr, mac_addr);
8776 	ether_addr_copy(adapter->mac_table[rar_entry].addr, mac_addr);
8777 	adapter->mac_table[rar_entry].queue = vf;
8778 	adapter->mac_table[rar_entry].state |= IGB_MAC_STATE_IN_USE;
8779 	igb_rar_set_index(adapter, rar_entry);
8780 
8781 	return 0;
8782 }
8783 
8784 static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
8785 {
8786 	struct igb_adapter *adapter = netdev_priv(netdev);
8787 
8788 	if (vf >= adapter->vfs_allocated_count)
8789 		return -EINVAL;
8790 
8791 	/* Setting the VF MAC to 0 reverts the IGB_VF_FLAG_PF_SET_MAC
8792 	 * flag and allows to overwrite the MAC via VF netdev.  This
8793 	 * is necessary to allow libvirt a way to restore the original
8794 	 * MAC after unbinding vfio-pci and reloading igbvf after shutting
8795 	 * down a VM.
8796 	 */
8797 	if (is_zero_ether_addr(mac)) {
8798 		adapter->vf_data[vf].flags &= ~IGB_VF_FLAG_PF_SET_MAC;
8799 		dev_info(&adapter->pdev->dev,
8800 			 "remove administratively set MAC on VF %d\n",
8801 			 vf);
8802 	} else if (is_valid_ether_addr(mac)) {
8803 		adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
8804 		dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n",
8805 			 mac, vf);
8806 		dev_info(&adapter->pdev->dev,
8807 			 "Reload the VF driver to make this change effective.");
8808 		/* Generate additional warning if PF is down */
8809 		if (test_bit(__IGB_DOWN, &adapter->state)) {
8810 			dev_warn(&adapter->pdev->dev,
8811 				 "The VF MAC address has been set, but the PF device is not up.\n");
8812 			dev_warn(&adapter->pdev->dev,
8813 				 "Bring the PF device up before attempting to use the VF device.\n");
8814 		}
8815 	} else {
8816 		return -EINVAL;
8817 	}
8818 	return igb_set_vf_mac(adapter, vf, mac);
8819 }
8820 
8821 static int igb_link_mbps(int internal_link_speed)
8822 {
8823 	switch (internal_link_speed) {
8824 	case SPEED_100:
8825 		return 100;
8826 	case SPEED_1000:
8827 		return 1000;
8828 	default:
8829 		return 0;
8830 	}
8831 }
8832 
8833 static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
8834 				  int link_speed)
8835 {
8836 	int rf_dec, rf_int;
8837 	u32 bcnrc_val;
8838 
8839 	if (tx_rate != 0) {
8840 		/* Calculate the rate factor values to set */
8841 		rf_int = link_speed / tx_rate;
8842 		rf_dec = (link_speed - (rf_int * tx_rate));
8843 		rf_dec = (rf_dec * BIT(E1000_RTTBCNRC_RF_INT_SHIFT)) /
8844 			 tx_rate;
8845 
8846 		bcnrc_val = E1000_RTTBCNRC_RS_ENA;
8847 		bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
8848 			      E1000_RTTBCNRC_RF_INT_MASK);
8849 		bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
8850 	} else {
8851 		bcnrc_val = 0;
8852 	}
8853 
8854 	wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
8855 	/* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
8856 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
8857 	 */
8858 	wr32(E1000_RTTBCNRM, 0x14);
8859 	wr32(E1000_RTTBCNRC, bcnrc_val);
8860 }
8861 
8862 static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
8863 {
8864 	int actual_link_speed, i;
8865 	bool reset_rate = false;
8866 
8867 	/* VF TX rate limit was not set or not supported */
8868 	if ((adapter->vf_rate_link_speed == 0) ||
8869 	    (adapter->hw.mac.type != e1000_82576))
8870 		return;
8871 
8872 	actual_link_speed = igb_link_mbps(adapter->link_speed);
8873 	if (actual_link_speed != adapter->vf_rate_link_speed) {
8874 		reset_rate = true;
8875 		adapter->vf_rate_link_speed = 0;
8876 		dev_info(&adapter->pdev->dev,
8877 			 "Link speed has been changed. VF Transmit rate is disabled\n");
8878 	}
8879 
8880 	for (i = 0; i < adapter->vfs_allocated_count; i++) {
8881 		if (reset_rate)
8882 			adapter->vf_data[i].tx_rate = 0;
8883 
8884 		igb_set_vf_rate_limit(&adapter->hw, i,
8885 				      adapter->vf_data[i].tx_rate,
8886 				      actual_link_speed);
8887 	}
8888 }
8889 
8890 static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
8891 			     int min_tx_rate, int max_tx_rate)
8892 {
8893 	struct igb_adapter *adapter = netdev_priv(netdev);
8894 	struct e1000_hw *hw = &adapter->hw;
8895 	int actual_link_speed;
8896 
8897 	if (hw->mac.type != e1000_82576)
8898 		return -EOPNOTSUPP;
8899 
8900 	if (min_tx_rate)
8901 		return -EINVAL;
8902 
8903 	actual_link_speed = igb_link_mbps(adapter->link_speed);
8904 	if ((vf >= adapter->vfs_allocated_count) ||
8905 	    (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
8906 	    (max_tx_rate < 0) ||
8907 	    (max_tx_rate > actual_link_speed))
8908 		return -EINVAL;
8909 
8910 	adapter->vf_rate_link_speed = actual_link_speed;
8911 	adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
8912 	igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
8913 
8914 	return 0;
8915 }
8916 
8917 static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
8918 				   bool setting)
8919 {
8920 	struct igb_adapter *adapter = netdev_priv(netdev);
8921 	struct e1000_hw *hw = &adapter->hw;
8922 	u32 reg_val, reg_offset;
8923 
8924 	if (!adapter->vfs_allocated_count)
8925 		return -EOPNOTSUPP;
8926 
8927 	if (vf >= adapter->vfs_allocated_count)
8928 		return -EINVAL;
8929 
8930 	reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
8931 	reg_val = rd32(reg_offset);
8932 	if (setting)
8933 		reg_val |= (BIT(vf) |
8934 			    BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
8935 	else
8936 		reg_val &= ~(BIT(vf) |
8937 			     BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
8938 	wr32(reg_offset, reg_val);
8939 
8940 	adapter->vf_data[vf].spoofchk_enabled = setting;
8941 	return 0;
8942 }
8943 
8944 static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf, bool setting)
8945 {
8946 	struct igb_adapter *adapter = netdev_priv(netdev);
8947 
8948 	if (vf >= adapter->vfs_allocated_count)
8949 		return -EINVAL;
8950 	if (adapter->vf_data[vf].trusted == setting)
8951 		return 0;
8952 
8953 	adapter->vf_data[vf].trusted = setting;
8954 
8955 	dev_info(&adapter->pdev->dev, "VF %u is %strusted\n",
8956 		 vf, setting ? "" : "not ");
8957 	return 0;
8958 }
8959 
8960 static int igb_ndo_get_vf_config(struct net_device *netdev,
8961 				 int vf, struct ifla_vf_info *ivi)
8962 {
8963 	struct igb_adapter *adapter = netdev_priv(netdev);
8964 	if (vf >= adapter->vfs_allocated_count)
8965 		return -EINVAL;
8966 	ivi->vf = vf;
8967 	memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
8968 	ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
8969 	ivi->min_tx_rate = 0;
8970 	ivi->vlan = adapter->vf_data[vf].pf_vlan;
8971 	ivi->qos = adapter->vf_data[vf].pf_qos;
8972 	ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
8973 	ivi->trusted = adapter->vf_data[vf].trusted;
8974 	return 0;
8975 }
8976 
8977 static void igb_vmm_control(struct igb_adapter *adapter)
8978 {
8979 	struct e1000_hw *hw = &adapter->hw;
8980 	u32 reg;
8981 
8982 	switch (hw->mac.type) {
8983 	case e1000_82575:
8984 	case e1000_i210:
8985 	case e1000_i211:
8986 	case e1000_i354:
8987 	default:
8988 		/* replication is not supported for 82575 */
8989 		return;
8990 	case e1000_82576:
8991 		/* notify HW that the MAC is adding vlan tags */
8992 		reg = rd32(E1000_DTXCTL);
8993 		reg |= E1000_DTXCTL_VLAN_ADDED;
8994 		wr32(E1000_DTXCTL, reg);
8995 		/* Fall through */
8996 	case e1000_82580:
8997 		/* enable replication vlan tag stripping */
8998 		reg = rd32(E1000_RPLOLR);
8999 		reg |= E1000_RPLOLR_STRVLAN;
9000 		wr32(E1000_RPLOLR, reg);
9001 		/* Fall through */
9002 	case e1000_i350:
9003 		/* none of the above registers are supported by i350 */
9004 		break;
9005 	}
9006 
9007 	if (adapter->vfs_allocated_count) {
9008 		igb_vmdq_set_loopback_pf(hw, true);
9009 		igb_vmdq_set_replication_pf(hw, true);
9010 		igb_vmdq_set_anti_spoofing_pf(hw, true,
9011 					      adapter->vfs_allocated_count);
9012 	} else {
9013 		igb_vmdq_set_loopback_pf(hw, false);
9014 		igb_vmdq_set_replication_pf(hw, false);
9015 	}
9016 }
9017 
9018 static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
9019 {
9020 	struct e1000_hw *hw = &adapter->hw;
9021 	u32 dmac_thr;
9022 	u16 hwm;
9023 
9024 	if (hw->mac.type > e1000_82580) {
9025 		if (adapter->flags & IGB_FLAG_DMAC) {
9026 			u32 reg;
9027 
9028 			/* force threshold to 0. */
9029 			wr32(E1000_DMCTXTH, 0);
9030 
9031 			/* DMA Coalescing high water mark needs to be greater
9032 			 * than the Rx threshold. Set hwm to PBA - max frame
9033 			 * size in 16B units, capping it at PBA - 6KB.
9034 			 */
9035 			hwm = 64 * (pba - 6);
9036 			reg = rd32(E1000_FCRTC);
9037 			reg &= ~E1000_FCRTC_RTH_COAL_MASK;
9038 			reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
9039 				& E1000_FCRTC_RTH_COAL_MASK);
9040 			wr32(E1000_FCRTC, reg);
9041 
9042 			/* Set the DMA Coalescing Rx threshold to PBA - 2 * max
9043 			 * frame size, capping it at PBA - 10KB.
9044 			 */
9045 			dmac_thr = pba - 10;
9046 			reg = rd32(E1000_DMACR);
9047 			reg &= ~E1000_DMACR_DMACTHR_MASK;
9048 			reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
9049 				& E1000_DMACR_DMACTHR_MASK);
9050 
9051 			/* transition to L0x or L1 if available..*/
9052 			reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
9053 
9054 			/* watchdog timer= +-1000 usec in 32usec intervals */
9055 			reg |= (1000 >> 5);
9056 
9057 			/* Disable BMC-to-OS Watchdog Enable */
9058 			if (hw->mac.type != e1000_i354)
9059 				reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
9060 
9061 			wr32(E1000_DMACR, reg);
9062 
9063 			/* no lower threshold to disable
9064 			 * coalescing(smart fifb)-UTRESH=0
9065 			 */
9066 			wr32(E1000_DMCRTRH, 0);
9067 
9068 			reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
9069 
9070 			wr32(E1000_DMCTLX, reg);
9071 
9072 			/* free space in tx packet buffer to wake from
9073 			 * DMA coal
9074 			 */
9075 			wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
9076 			     (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
9077 
9078 			/* make low power state decision controlled
9079 			 * by DMA coal
9080 			 */
9081 			reg = rd32(E1000_PCIEMISC);
9082 			reg &= ~E1000_PCIEMISC_LX_DECISION;
9083 			wr32(E1000_PCIEMISC, reg);
9084 		} /* endif adapter->dmac is not disabled */
9085 	} else if (hw->mac.type == e1000_82580) {
9086 		u32 reg = rd32(E1000_PCIEMISC);
9087 
9088 		wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
9089 		wr32(E1000_DMACR, 0);
9090 	}
9091 }
9092 
9093 /**
9094  *  igb_read_i2c_byte - Reads 8 bit word over I2C
9095  *  @hw: pointer to hardware structure
9096  *  @byte_offset: byte offset to read
9097  *  @dev_addr: device address
9098  *  @data: value read
9099  *
9100  *  Performs byte read operation over I2C interface at
9101  *  a specified device address.
9102  **/
9103 s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
9104 		      u8 dev_addr, u8 *data)
9105 {
9106 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
9107 	struct i2c_client *this_client = adapter->i2c_client;
9108 	s32 status;
9109 	u16 swfw_mask = 0;
9110 
9111 	if (!this_client)
9112 		return E1000_ERR_I2C;
9113 
9114 	swfw_mask = E1000_SWFW_PHY0_SM;
9115 
9116 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
9117 		return E1000_ERR_SWFW_SYNC;
9118 
9119 	status = i2c_smbus_read_byte_data(this_client, byte_offset);
9120 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
9121 
9122 	if (status < 0)
9123 		return E1000_ERR_I2C;
9124 	else {
9125 		*data = status;
9126 		return 0;
9127 	}
9128 }
9129 
9130 /**
9131  *  igb_write_i2c_byte - Writes 8 bit word over I2C
9132  *  @hw: pointer to hardware structure
9133  *  @byte_offset: byte offset to write
9134  *  @dev_addr: device address
9135  *  @data: value to write
9136  *
9137  *  Performs byte write operation over I2C interface at
9138  *  a specified device address.
9139  **/
9140 s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
9141 		       u8 dev_addr, u8 data)
9142 {
9143 	struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
9144 	struct i2c_client *this_client = adapter->i2c_client;
9145 	s32 status;
9146 	u16 swfw_mask = E1000_SWFW_PHY0_SM;
9147 
9148 	if (!this_client)
9149 		return E1000_ERR_I2C;
9150 
9151 	if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
9152 		return E1000_ERR_SWFW_SYNC;
9153 	status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
9154 	hw->mac.ops.release_swfw_sync(hw, swfw_mask);
9155 
9156 	if (status)
9157 		return E1000_ERR_I2C;
9158 	else
9159 		return 0;
9160 
9161 }
9162 
9163 int igb_reinit_queues(struct igb_adapter *adapter)
9164 {
9165 	struct net_device *netdev = adapter->netdev;
9166 	struct pci_dev *pdev = adapter->pdev;
9167 	int err = 0;
9168 
9169 	if (netif_running(netdev))
9170 		igb_close(netdev);
9171 
9172 	igb_reset_interrupt_capability(adapter);
9173 
9174 	if (igb_init_interrupt_scheme(adapter, true)) {
9175 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9176 		return -ENOMEM;
9177 	}
9178 
9179 	if (netif_running(netdev))
9180 		err = igb_open(netdev);
9181 
9182 	return err;
9183 }
9184 
9185 static void igb_nfc_filter_exit(struct igb_adapter *adapter)
9186 {
9187 	struct igb_nfc_filter *rule;
9188 
9189 	spin_lock(&adapter->nfc_lock);
9190 
9191 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
9192 		igb_erase_filter(adapter, rule);
9193 
9194 	spin_unlock(&adapter->nfc_lock);
9195 }
9196 
9197 static void igb_nfc_filter_restore(struct igb_adapter *adapter)
9198 {
9199 	struct igb_nfc_filter *rule;
9200 
9201 	spin_lock(&adapter->nfc_lock);
9202 
9203 	hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
9204 		igb_add_filter(adapter, rule);
9205 
9206 	spin_unlock(&adapter->nfc_lock);
9207 }
9208 /* igb_main.c */
9209