ixgb_main.c revision b3f4d5990bfc8b060e5010c1464789fca1f4c5b4
1/*******************************************************************************
2
3  Intel PRO/10GbE Linux driver
4  Copyright(c) 1999 - 2008 Intel Corporation.
5
6  This program is free software; you can redistribute it and/or modify it
7  under the terms and conditions of the GNU General Public License,
8  version 2, as published by the Free Software Foundation.
9
10  This program is distributed in the hope it will be useful, but WITHOUT
11  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  more details.
14
15  You should have received a copy of the GNU General Public License along with
16  this program; if not, write to the Free Software Foundation, Inc.,
17  51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18
19  The full GNU General Public License is included in this distribution in
20  the file called "COPYING".
21
22  Contact Information:
23  Linux NICS <linux.nics@intel.com>
24  e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25  Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26
27*******************************************************************************/
28
29#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
30
31#include <linux/prefetch.h>
32#include "ixgb.h"
33
34char ixgb_driver_name[] = "ixgb";
35static char ixgb_driver_string[] = "Intel(R) PRO/10GbE Network Driver";
36
37#define DRIVERNAPI "-NAPI"
38#define DRV_VERSION "1.0.135-k2" DRIVERNAPI
39const char ixgb_driver_version[] = DRV_VERSION;
40static const char ixgb_copyright[] = "Copyright (c) 1999-2008 Intel Corporation.";
41
42#define IXGB_CB_LENGTH 256
43static unsigned int copybreak __read_mostly = IXGB_CB_LENGTH;
44module_param(copybreak, uint, 0644);
45MODULE_PARM_DESC(copybreak,
46	"Maximum size of packet that is copied to a new buffer on receive");
47
48/* ixgb_pci_tbl - PCI Device ID Table
49 *
50 * Wildcard entries (PCI_ANY_ID) should come last
51 * Last entry must be all 0s
52 *
53 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
54 *   Class, Class Mask, private data (not used) }
55 */
56static DEFINE_PCI_DEVICE_TABLE(ixgb_pci_tbl) = {
57	{INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX,
58	 PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
59	{INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_CX4,
60	 PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
61	{INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_SR,
62	 PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
63	{INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_LR,
64	 PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
65
66	/* required last entry */
67	{0,}
68};
69
70MODULE_DEVICE_TABLE(pci, ixgb_pci_tbl);
71
72/* Local Function Prototypes */
73static int ixgb_init_module(void);
74static void ixgb_exit_module(void);
75static int ixgb_probe(struct pci_dev *pdev, const struct pci_device_id *ent);
76static void __devexit ixgb_remove(struct pci_dev *pdev);
77static int ixgb_sw_init(struct ixgb_adapter *adapter);
78static int ixgb_open(struct net_device *netdev);
79static int ixgb_close(struct net_device *netdev);
80static void ixgb_configure_tx(struct ixgb_adapter *adapter);
81static void ixgb_configure_rx(struct ixgb_adapter *adapter);
82static void ixgb_setup_rctl(struct ixgb_adapter *adapter);
83static void ixgb_clean_tx_ring(struct ixgb_adapter *adapter);
84static void ixgb_clean_rx_ring(struct ixgb_adapter *adapter);
85static void ixgb_set_multi(struct net_device *netdev);
86static void ixgb_watchdog(unsigned long data);
87static netdev_tx_t ixgb_xmit_frame(struct sk_buff *skb,
88				   struct net_device *netdev);
89static struct net_device_stats *ixgb_get_stats(struct net_device *netdev);
90static int ixgb_change_mtu(struct net_device *netdev, int new_mtu);
91static int ixgb_set_mac(struct net_device *netdev, void *p);
92static irqreturn_t ixgb_intr(int irq, void *data);
93static bool ixgb_clean_tx_irq(struct ixgb_adapter *adapter);
94
95static int ixgb_clean(struct napi_struct *, int);
96static bool ixgb_clean_rx_irq(struct ixgb_adapter *, int *, int);
97static void ixgb_alloc_rx_buffers(struct ixgb_adapter *, int);
98
99static void ixgb_tx_timeout(struct net_device *dev);
100static void ixgb_tx_timeout_task(struct work_struct *work);
101
102static void ixgb_vlan_strip_enable(struct ixgb_adapter *adapter);
103static void ixgb_vlan_strip_disable(struct ixgb_adapter *adapter);
104static int ixgb_vlan_rx_add_vid(struct net_device *netdev, u16 vid);
105static int ixgb_vlan_rx_kill_vid(struct net_device *netdev, u16 vid);
106static void ixgb_restore_vlan(struct ixgb_adapter *adapter);
107
108#ifdef CONFIG_NET_POLL_CONTROLLER
109/* for netdump / net console */
110static void ixgb_netpoll(struct net_device *dev);
111#endif
112
113static pci_ers_result_t ixgb_io_error_detected (struct pci_dev *pdev,
114                             enum pci_channel_state state);
115static pci_ers_result_t ixgb_io_slot_reset (struct pci_dev *pdev);
116static void ixgb_io_resume (struct pci_dev *pdev);
117
118static struct pci_error_handlers ixgb_err_handler = {
119	.error_detected = ixgb_io_error_detected,
120	.slot_reset = ixgb_io_slot_reset,
121	.resume = ixgb_io_resume,
122};
123
124static struct pci_driver ixgb_driver = {
125	.name     = ixgb_driver_name,
126	.id_table = ixgb_pci_tbl,
127	.probe    = ixgb_probe,
128	.remove   = __devexit_p(ixgb_remove),
129	.err_handler = &ixgb_err_handler
130};
131
132MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
133MODULE_DESCRIPTION("Intel(R) PRO/10GbE Network Driver");
134MODULE_LICENSE("GPL");
135MODULE_VERSION(DRV_VERSION);
136
137#define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
138static int debug = -1;
139module_param(debug, int, 0);
140MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
141
142/**
143 * ixgb_init_module - Driver Registration Routine
144 *
145 * ixgb_init_module is the first routine called when the driver is
146 * loaded. All it does is register with the PCI subsystem.
147 **/
148
149static int __init
150ixgb_init_module(void)
151{
152	pr_info("%s - version %s\n", ixgb_driver_string, ixgb_driver_version);
153	pr_info("%s\n", ixgb_copyright);
154
155	return pci_register_driver(&ixgb_driver);
156}
157
158module_init(ixgb_init_module);
159
160/**
161 * ixgb_exit_module - Driver Exit Cleanup Routine
162 *
163 * ixgb_exit_module is called just before the driver is removed
164 * from memory.
165 **/
166
167static void __exit
168ixgb_exit_module(void)
169{
170	pci_unregister_driver(&ixgb_driver);
171}
172
173module_exit(ixgb_exit_module);
174
175/**
176 * ixgb_irq_disable - Mask off interrupt generation on the NIC
177 * @adapter: board private structure
178 **/
179
180static void
181ixgb_irq_disable(struct ixgb_adapter *adapter)
182{
183	IXGB_WRITE_REG(&adapter->hw, IMC, ~0);
184	IXGB_WRITE_FLUSH(&adapter->hw);
185	synchronize_irq(adapter->pdev->irq);
186}
187
188/**
189 * ixgb_irq_enable - Enable default interrupt generation settings
190 * @adapter: board private structure
191 **/
192
193static void
194ixgb_irq_enable(struct ixgb_adapter *adapter)
195{
196	u32 val = IXGB_INT_RXT0 | IXGB_INT_RXDMT0 |
197		  IXGB_INT_TXDW | IXGB_INT_LSC;
198	if (adapter->hw.subsystem_vendor_id == SUN_SUBVENDOR_ID)
199		val |= IXGB_INT_GPI0;
200	IXGB_WRITE_REG(&adapter->hw, IMS, val);
201	IXGB_WRITE_FLUSH(&adapter->hw);
202}
203
204int
205ixgb_up(struct ixgb_adapter *adapter)
206{
207	struct net_device *netdev = adapter->netdev;
208	int err, irq_flags = IRQF_SHARED;
209	int max_frame = netdev->mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
210	struct ixgb_hw *hw = &adapter->hw;
211
212	/* hardware has been reset, we need to reload some things */
213
214	ixgb_rar_set(hw, netdev->dev_addr, 0);
215	ixgb_set_multi(netdev);
216
217	ixgb_restore_vlan(adapter);
218
219	ixgb_configure_tx(adapter);
220	ixgb_setup_rctl(adapter);
221	ixgb_configure_rx(adapter);
222	ixgb_alloc_rx_buffers(adapter, IXGB_DESC_UNUSED(&adapter->rx_ring));
223
224	/* disable interrupts and get the hardware into a known state */
225	IXGB_WRITE_REG(&adapter->hw, IMC, 0xffffffff);
226
227	/* only enable MSI if bus is in PCI-X mode */
228	if (IXGB_READ_REG(&adapter->hw, STATUS) & IXGB_STATUS_PCIX_MODE) {
229		err = pci_enable_msi(adapter->pdev);
230		if (!err) {
231			adapter->have_msi = true;
232			irq_flags = 0;
233		}
234		/* proceed to try to request regular interrupt */
235	}
236
237	err = request_irq(adapter->pdev->irq, ixgb_intr, irq_flags,
238	                  netdev->name, netdev);
239	if (err) {
240		if (adapter->have_msi)
241			pci_disable_msi(adapter->pdev);
242		netif_err(adapter, probe, adapter->netdev,
243			  "Unable to allocate interrupt Error: %d\n", err);
244		return err;
245	}
246
247	if ((hw->max_frame_size != max_frame) ||
248		(hw->max_frame_size !=
249		(IXGB_READ_REG(hw, MFS) >> IXGB_MFS_SHIFT))) {
250
251		hw->max_frame_size = max_frame;
252
253		IXGB_WRITE_REG(hw, MFS, hw->max_frame_size << IXGB_MFS_SHIFT);
254
255		if (hw->max_frame_size >
256		   IXGB_MAX_ENET_FRAME_SIZE_WITHOUT_FCS + ENET_FCS_LENGTH) {
257			u32 ctrl0 = IXGB_READ_REG(hw, CTRL0);
258
259			if (!(ctrl0 & IXGB_CTRL0_JFE)) {
260				ctrl0 |= IXGB_CTRL0_JFE;
261				IXGB_WRITE_REG(hw, CTRL0, ctrl0);
262			}
263		}
264	}
265
266	clear_bit(__IXGB_DOWN, &adapter->flags);
267
268	napi_enable(&adapter->napi);
269	ixgb_irq_enable(adapter);
270
271	netif_wake_queue(netdev);
272
273	mod_timer(&adapter->watchdog_timer, jiffies);
274
275	return 0;
276}
277
278void
279ixgb_down(struct ixgb_adapter *adapter, bool kill_watchdog)
280{
281	struct net_device *netdev = adapter->netdev;
282
283	/* prevent the interrupt handler from restarting watchdog */
284	set_bit(__IXGB_DOWN, &adapter->flags);
285
286	napi_disable(&adapter->napi);
287	/* waiting for NAPI to complete can re-enable interrupts */
288	ixgb_irq_disable(adapter);
289	free_irq(adapter->pdev->irq, netdev);
290
291	if (adapter->have_msi)
292		pci_disable_msi(adapter->pdev);
293
294	if (kill_watchdog)
295		del_timer_sync(&adapter->watchdog_timer);
296
297	adapter->link_speed = 0;
298	adapter->link_duplex = 0;
299	netif_carrier_off(netdev);
300	netif_stop_queue(netdev);
301
302	ixgb_reset(adapter);
303	ixgb_clean_tx_ring(adapter);
304	ixgb_clean_rx_ring(adapter);
305}
306
307void
308ixgb_reset(struct ixgb_adapter *adapter)
309{
310	struct ixgb_hw *hw = &adapter->hw;
311
312	ixgb_adapter_stop(hw);
313	if (!ixgb_init_hw(hw))
314		netif_err(adapter, probe, adapter->netdev, "ixgb_init_hw failed\n");
315
316	/* restore frame size information */
317	IXGB_WRITE_REG(hw, MFS, hw->max_frame_size << IXGB_MFS_SHIFT);
318	if (hw->max_frame_size >
319	    IXGB_MAX_ENET_FRAME_SIZE_WITHOUT_FCS + ENET_FCS_LENGTH) {
320		u32 ctrl0 = IXGB_READ_REG(hw, CTRL0);
321		if (!(ctrl0 & IXGB_CTRL0_JFE)) {
322			ctrl0 |= IXGB_CTRL0_JFE;
323			IXGB_WRITE_REG(hw, CTRL0, ctrl0);
324		}
325	}
326}
327
328static netdev_features_t
329ixgb_fix_features(struct net_device *netdev, netdev_features_t features)
330{
331	/*
332	 * Tx VLAN insertion does not work per HW design when Rx stripping is
333	 * disabled.
334	 */
335	if (!(features & NETIF_F_HW_VLAN_RX))
336		features &= ~NETIF_F_HW_VLAN_TX;
337
338	return features;
339}
340
341static int
342ixgb_set_features(struct net_device *netdev, netdev_features_t features)
343{
344	struct ixgb_adapter *adapter = netdev_priv(netdev);
345	netdev_features_t changed = features ^ netdev->features;
346
347	if (!(changed & (NETIF_F_RXCSUM|NETIF_F_HW_VLAN_RX)))
348		return 0;
349
350	adapter->rx_csum = !!(features & NETIF_F_RXCSUM);
351
352	if (netif_running(netdev)) {
353		ixgb_down(adapter, true);
354		ixgb_up(adapter);
355		ixgb_set_speed_duplex(netdev);
356	} else
357		ixgb_reset(adapter);
358
359	return 0;
360}
361
362
363static const struct net_device_ops ixgb_netdev_ops = {
364	.ndo_open 		= ixgb_open,
365	.ndo_stop		= ixgb_close,
366	.ndo_start_xmit		= ixgb_xmit_frame,
367	.ndo_get_stats		= ixgb_get_stats,
368	.ndo_set_rx_mode	= ixgb_set_multi,
369	.ndo_validate_addr	= eth_validate_addr,
370	.ndo_set_mac_address	= ixgb_set_mac,
371	.ndo_change_mtu		= ixgb_change_mtu,
372	.ndo_tx_timeout		= ixgb_tx_timeout,
373	.ndo_vlan_rx_add_vid	= ixgb_vlan_rx_add_vid,
374	.ndo_vlan_rx_kill_vid	= ixgb_vlan_rx_kill_vid,
375#ifdef CONFIG_NET_POLL_CONTROLLER
376	.ndo_poll_controller	= ixgb_netpoll,
377#endif
378	.ndo_fix_features       = ixgb_fix_features,
379	.ndo_set_features       = ixgb_set_features,
380};
381
382/**
383 * ixgb_probe - Device Initialization Routine
384 * @pdev: PCI device information struct
385 * @ent: entry in ixgb_pci_tbl
386 *
387 * Returns 0 on success, negative on failure
388 *
389 * ixgb_probe initializes an adapter identified by a pci_dev structure.
390 * The OS initialization, configuring of the adapter private structure,
391 * and a hardware reset occur.
392 **/
393
394static int __devinit
395ixgb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
396{
397	struct net_device *netdev = NULL;
398	struct ixgb_adapter *adapter;
399	static int cards_found = 0;
400	int pci_using_dac;
401	int i;
402	int err;
403
404	err = pci_enable_device(pdev);
405	if (err)
406		return err;
407
408	pci_using_dac = 0;
409	err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(64));
410	if (!err) {
411		err = dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
412		if (!err)
413			pci_using_dac = 1;
414	} else {
415		err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
416		if (err) {
417			err = dma_set_coherent_mask(&pdev->dev,
418						    DMA_BIT_MASK(32));
419			if (err) {
420				pr_err("No usable DMA configuration, aborting\n");
421				goto err_dma_mask;
422			}
423		}
424	}
425
426	err = pci_request_regions(pdev, ixgb_driver_name);
427	if (err)
428		goto err_request_regions;
429
430	pci_set_master(pdev);
431
432	netdev = alloc_etherdev(sizeof(struct ixgb_adapter));
433	if (!netdev) {
434		err = -ENOMEM;
435		goto err_alloc_etherdev;
436	}
437
438	SET_NETDEV_DEV(netdev, &pdev->dev);
439
440	pci_set_drvdata(pdev, netdev);
441	adapter = netdev_priv(netdev);
442	adapter->netdev = netdev;
443	adapter->pdev = pdev;
444	adapter->hw.back = adapter;
445	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
446
447	adapter->hw.hw_addr = pci_ioremap_bar(pdev, BAR_0);
448	if (!adapter->hw.hw_addr) {
449		err = -EIO;
450		goto err_ioremap;
451	}
452
453	for (i = BAR_1; i <= BAR_5; i++) {
454		if (pci_resource_len(pdev, i) == 0)
455			continue;
456		if (pci_resource_flags(pdev, i) & IORESOURCE_IO) {
457			adapter->hw.io_base = pci_resource_start(pdev, i);
458			break;
459		}
460	}
461
462	netdev->netdev_ops = &ixgb_netdev_ops;
463	ixgb_set_ethtool_ops(netdev);
464	netdev->watchdog_timeo = 5 * HZ;
465	netif_napi_add(netdev, &adapter->napi, ixgb_clean, 64);
466
467	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
468
469	adapter->bd_number = cards_found;
470	adapter->link_speed = 0;
471	adapter->link_duplex = 0;
472
473	/* setup the private structure */
474
475	err = ixgb_sw_init(adapter);
476	if (err)
477		goto err_sw_init;
478
479	netdev->hw_features = NETIF_F_SG |
480			   NETIF_F_TSO |
481			   NETIF_F_HW_CSUM |
482			   NETIF_F_HW_VLAN_TX |
483			   NETIF_F_HW_VLAN_RX;
484	netdev->features = netdev->hw_features |
485			   NETIF_F_HW_VLAN_FILTER;
486	netdev->hw_features |= NETIF_F_RXCSUM;
487
488	if (pci_using_dac) {
489		netdev->features |= NETIF_F_HIGHDMA;
490		netdev->vlan_features |= NETIF_F_HIGHDMA;
491	}
492
493	/* make sure the EEPROM is good */
494
495	if (!ixgb_validate_eeprom_checksum(&adapter->hw)) {
496		netif_err(adapter, probe, adapter->netdev,
497			  "The EEPROM Checksum Is Not Valid\n");
498		err = -EIO;
499		goto err_eeprom;
500	}
501
502	ixgb_get_ee_mac_addr(&adapter->hw, netdev->dev_addr);
503	memcpy(netdev->perm_addr, netdev->dev_addr, netdev->addr_len);
504
505	if (!is_valid_ether_addr(netdev->perm_addr)) {
506		netif_err(adapter, probe, adapter->netdev, "Invalid MAC Address\n");
507		err = -EIO;
508		goto err_eeprom;
509	}
510
511	adapter->part_num = ixgb_get_ee_pba_number(&adapter->hw);
512
513	init_timer(&adapter->watchdog_timer);
514	adapter->watchdog_timer.function = ixgb_watchdog;
515	adapter->watchdog_timer.data = (unsigned long)adapter;
516
517	INIT_WORK(&adapter->tx_timeout_task, ixgb_tx_timeout_task);
518
519	strcpy(netdev->name, "eth%d");
520	err = register_netdev(netdev);
521	if (err)
522		goto err_register;
523
524	/* carrier off reporting is important to ethtool even BEFORE open */
525	netif_carrier_off(netdev);
526
527	netif_info(adapter, probe, adapter->netdev,
528		   "Intel(R) PRO/10GbE Network Connection\n");
529	ixgb_check_options(adapter);
530	/* reset the hardware with the new settings */
531
532	ixgb_reset(adapter);
533
534	cards_found++;
535	return 0;
536
537err_register:
538err_sw_init:
539err_eeprom:
540	iounmap(adapter->hw.hw_addr);
541err_ioremap:
542	free_netdev(netdev);
543err_alloc_etherdev:
544	pci_release_regions(pdev);
545err_request_regions:
546err_dma_mask:
547	pci_disable_device(pdev);
548	return err;
549}
550
551/**
552 * ixgb_remove - Device Removal Routine
553 * @pdev: PCI device information struct
554 *
555 * ixgb_remove is called by the PCI subsystem to alert the driver
556 * that it should release a PCI device.  The could be caused by a
557 * Hot-Plug event, or because the driver is going to be removed from
558 * memory.
559 **/
560
561static void __devexit
562ixgb_remove(struct pci_dev *pdev)
563{
564	struct net_device *netdev = pci_get_drvdata(pdev);
565	struct ixgb_adapter *adapter = netdev_priv(netdev);
566
567	cancel_work_sync(&adapter->tx_timeout_task);
568
569	unregister_netdev(netdev);
570
571	iounmap(adapter->hw.hw_addr);
572	pci_release_regions(pdev);
573
574	free_netdev(netdev);
575	pci_disable_device(pdev);
576}
577
578/**
579 * ixgb_sw_init - Initialize general software structures (struct ixgb_adapter)
580 * @adapter: board private structure to initialize
581 *
582 * ixgb_sw_init initializes the Adapter private data structure.
583 * Fields are initialized based on PCI device information and
584 * OS network device settings (MTU size).
585 **/
586
587static int __devinit
588ixgb_sw_init(struct ixgb_adapter *adapter)
589{
590	struct ixgb_hw *hw = &adapter->hw;
591	struct net_device *netdev = adapter->netdev;
592	struct pci_dev *pdev = adapter->pdev;
593
594	/* PCI config space info */
595
596	hw->vendor_id = pdev->vendor;
597	hw->device_id = pdev->device;
598	hw->subsystem_vendor_id = pdev->subsystem_vendor;
599	hw->subsystem_id = pdev->subsystem_device;
600
601	hw->max_frame_size = netdev->mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
602	adapter->rx_buffer_len = hw->max_frame_size + 8; /* + 8 for errata */
603
604	if ((hw->device_id == IXGB_DEVICE_ID_82597EX) ||
605	    (hw->device_id == IXGB_DEVICE_ID_82597EX_CX4) ||
606	    (hw->device_id == IXGB_DEVICE_ID_82597EX_LR) ||
607	    (hw->device_id == IXGB_DEVICE_ID_82597EX_SR))
608		hw->mac_type = ixgb_82597;
609	else {
610		/* should never have loaded on this device */
611		netif_err(adapter, probe, adapter->netdev, "unsupported device id\n");
612	}
613
614	/* enable flow control to be programmed */
615	hw->fc.send_xon = 1;
616
617	set_bit(__IXGB_DOWN, &adapter->flags);
618	return 0;
619}
620
621/**
622 * ixgb_open - Called when a network interface is made active
623 * @netdev: network interface device structure
624 *
625 * Returns 0 on success, negative value on failure
626 *
627 * The open entry point is called when a network interface is made
628 * active by the system (IFF_UP).  At this point all resources needed
629 * for transmit and receive operations are allocated, the interrupt
630 * handler is registered with the OS, the watchdog timer is started,
631 * and the stack is notified that the interface is ready.
632 **/
633
634static int
635ixgb_open(struct net_device *netdev)
636{
637	struct ixgb_adapter *adapter = netdev_priv(netdev);
638	int err;
639
640	/* allocate transmit descriptors */
641	err = ixgb_setup_tx_resources(adapter);
642	if (err)
643		goto err_setup_tx;
644
645	netif_carrier_off(netdev);
646
647	/* allocate receive descriptors */
648
649	err = ixgb_setup_rx_resources(adapter);
650	if (err)
651		goto err_setup_rx;
652
653	err = ixgb_up(adapter);
654	if (err)
655		goto err_up;
656
657	netif_start_queue(netdev);
658
659	return 0;
660
661err_up:
662	ixgb_free_rx_resources(adapter);
663err_setup_rx:
664	ixgb_free_tx_resources(adapter);
665err_setup_tx:
666	ixgb_reset(adapter);
667
668	return err;
669}
670
671/**
672 * ixgb_close - Disables a network interface
673 * @netdev: network interface device structure
674 *
675 * Returns 0, this is not allowed to fail
676 *
677 * The close entry point is called when an interface is de-activated
678 * by the OS.  The hardware is still under the drivers control, but
679 * needs to be disabled.  A global MAC reset is issued to stop the
680 * hardware, and all transmit and receive resources are freed.
681 **/
682
683static int
684ixgb_close(struct net_device *netdev)
685{
686	struct ixgb_adapter *adapter = netdev_priv(netdev);
687
688	ixgb_down(adapter, true);
689
690	ixgb_free_tx_resources(adapter);
691	ixgb_free_rx_resources(adapter);
692
693	return 0;
694}
695
696/**
697 * ixgb_setup_tx_resources - allocate Tx resources (Descriptors)
698 * @adapter: board private structure
699 *
700 * Return 0 on success, negative on failure
701 **/
702
703int
704ixgb_setup_tx_resources(struct ixgb_adapter *adapter)
705{
706	struct ixgb_desc_ring *txdr = &adapter->tx_ring;
707	struct pci_dev *pdev = adapter->pdev;
708	int size;
709
710	size = sizeof(struct ixgb_buffer) * txdr->count;
711	txdr->buffer_info = vzalloc(size);
712	if (!txdr->buffer_info) {
713		netif_err(adapter, probe, adapter->netdev,
714			  "Unable to allocate transmit descriptor ring memory\n");
715		return -ENOMEM;
716	}
717
718	/* round up to nearest 4K */
719
720	txdr->size = txdr->count * sizeof(struct ixgb_tx_desc);
721	txdr->size = ALIGN(txdr->size, 4096);
722
723	txdr->desc = dma_alloc_coherent(&pdev->dev, txdr->size, &txdr->dma,
724					GFP_KERNEL);
725	if (!txdr->desc) {
726		vfree(txdr->buffer_info);
727		netif_err(adapter, probe, adapter->netdev,
728			  "Unable to allocate transmit descriptor memory\n");
729		return -ENOMEM;
730	}
731	memset(txdr->desc, 0, txdr->size);
732
733	txdr->next_to_use = 0;
734	txdr->next_to_clean = 0;
735
736	return 0;
737}
738
739/**
740 * ixgb_configure_tx - Configure 82597 Transmit Unit after Reset.
741 * @adapter: board private structure
742 *
743 * Configure the Tx unit of the MAC after a reset.
744 **/
745
746static void
747ixgb_configure_tx(struct ixgb_adapter *adapter)
748{
749	u64 tdba = adapter->tx_ring.dma;
750	u32 tdlen = adapter->tx_ring.count * sizeof(struct ixgb_tx_desc);
751	u32 tctl;
752	struct ixgb_hw *hw = &adapter->hw;
753
754	/* Setup the Base and Length of the Tx Descriptor Ring
755	 * tx_ring.dma can be either a 32 or 64 bit value
756	 */
757
758	IXGB_WRITE_REG(hw, TDBAL, (tdba & 0x00000000ffffffffULL));
759	IXGB_WRITE_REG(hw, TDBAH, (tdba >> 32));
760
761	IXGB_WRITE_REG(hw, TDLEN, tdlen);
762
763	/* Setup the HW Tx Head and Tail descriptor pointers */
764
765	IXGB_WRITE_REG(hw, TDH, 0);
766	IXGB_WRITE_REG(hw, TDT, 0);
767
768	/* don't set up txdctl, it induces performance problems if configured
769	 * incorrectly */
770	/* Set the Tx Interrupt Delay register */
771
772	IXGB_WRITE_REG(hw, TIDV, adapter->tx_int_delay);
773
774	/* Program the Transmit Control Register */
775
776	tctl = IXGB_TCTL_TCE | IXGB_TCTL_TXEN | IXGB_TCTL_TPDE;
777	IXGB_WRITE_REG(hw, TCTL, tctl);
778
779	/* Setup Transmit Descriptor Settings for this adapter */
780	adapter->tx_cmd_type =
781		IXGB_TX_DESC_TYPE |
782		(adapter->tx_int_delay_enable ? IXGB_TX_DESC_CMD_IDE : 0);
783}
784
785/**
786 * ixgb_setup_rx_resources - allocate Rx resources (Descriptors)
787 * @adapter: board private structure
788 *
789 * Returns 0 on success, negative on failure
790 **/
791
792int
793ixgb_setup_rx_resources(struct ixgb_adapter *adapter)
794{
795	struct ixgb_desc_ring *rxdr = &adapter->rx_ring;
796	struct pci_dev *pdev = adapter->pdev;
797	int size;
798
799	size = sizeof(struct ixgb_buffer) * rxdr->count;
800	rxdr->buffer_info = vzalloc(size);
801	if (!rxdr->buffer_info) {
802		netif_err(adapter, probe, adapter->netdev,
803			  "Unable to allocate receive descriptor ring\n");
804		return -ENOMEM;
805	}
806
807	/* Round up to nearest 4K */
808
809	rxdr->size = rxdr->count * sizeof(struct ixgb_rx_desc);
810	rxdr->size = ALIGN(rxdr->size, 4096);
811
812	rxdr->desc = dma_alloc_coherent(&pdev->dev, rxdr->size, &rxdr->dma,
813					GFP_KERNEL);
814
815	if (!rxdr->desc) {
816		vfree(rxdr->buffer_info);
817		netif_err(adapter, probe, adapter->netdev,
818			  "Unable to allocate receive descriptors\n");
819		return -ENOMEM;
820	}
821	memset(rxdr->desc, 0, rxdr->size);
822
823	rxdr->next_to_clean = 0;
824	rxdr->next_to_use = 0;
825
826	return 0;
827}
828
829/**
830 * ixgb_setup_rctl - configure the receive control register
831 * @adapter: Board private structure
832 **/
833
834static void
835ixgb_setup_rctl(struct ixgb_adapter *adapter)
836{
837	u32 rctl;
838
839	rctl = IXGB_READ_REG(&adapter->hw, RCTL);
840
841	rctl &= ~(3 << IXGB_RCTL_MO_SHIFT);
842
843	rctl |=
844		IXGB_RCTL_BAM | IXGB_RCTL_RDMTS_1_2 |
845		IXGB_RCTL_RXEN | IXGB_RCTL_CFF |
846		(adapter->hw.mc_filter_type << IXGB_RCTL_MO_SHIFT);
847
848	rctl |= IXGB_RCTL_SECRC;
849
850	if (adapter->rx_buffer_len <= IXGB_RXBUFFER_2048)
851		rctl |= IXGB_RCTL_BSIZE_2048;
852	else if (adapter->rx_buffer_len <= IXGB_RXBUFFER_4096)
853		rctl |= IXGB_RCTL_BSIZE_4096;
854	else if (adapter->rx_buffer_len <= IXGB_RXBUFFER_8192)
855		rctl |= IXGB_RCTL_BSIZE_8192;
856	else if (adapter->rx_buffer_len <= IXGB_RXBUFFER_16384)
857		rctl |= IXGB_RCTL_BSIZE_16384;
858
859	IXGB_WRITE_REG(&adapter->hw, RCTL, rctl);
860}
861
862/**
863 * ixgb_configure_rx - Configure 82597 Receive Unit after Reset.
864 * @adapter: board private structure
865 *
866 * Configure the Rx unit of the MAC after a reset.
867 **/
868
869static void
870ixgb_configure_rx(struct ixgb_adapter *adapter)
871{
872	u64 rdba = adapter->rx_ring.dma;
873	u32 rdlen = adapter->rx_ring.count * sizeof(struct ixgb_rx_desc);
874	struct ixgb_hw *hw = &adapter->hw;
875	u32 rctl;
876	u32 rxcsum;
877
878	/* make sure receives are disabled while setting up the descriptors */
879
880	rctl = IXGB_READ_REG(hw, RCTL);
881	IXGB_WRITE_REG(hw, RCTL, rctl & ~IXGB_RCTL_RXEN);
882
883	/* set the Receive Delay Timer Register */
884
885	IXGB_WRITE_REG(hw, RDTR, adapter->rx_int_delay);
886
887	/* Setup the Base and Length of the Rx Descriptor Ring */
888
889	IXGB_WRITE_REG(hw, RDBAL, (rdba & 0x00000000ffffffffULL));
890	IXGB_WRITE_REG(hw, RDBAH, (rdba >> 32));
891
892	IXGB_WRITE_REG(hw, RDLEN, rdlen);
893
894	/* Setup the HW Rx Head and Tail Descriptor Pointers */
895	IXGB_WRITE_REG(hw, RDH, 0);
896	IXGB_WRITE_REG(hw, RDT, 0);
897
898	/* due to the hardware errata with RXDCTL, we are unable to use any of
899	 * the performance enhancing features of it without causing other
900	 * subtle bugs, some of the bugs could include receive length
901	 * corruption at high data rates (WTHRESH > 0) and/or receive
902	 * descriptor ring irregularites (particularly in hardware cache) */
903	IXGB_WRITE_REG(hw, RXDCTL, 0);
904
905	/* Enable Receive Checksum Offload for TCP and UDP */
906	if (adapter->rx_csum) {
907		rxcsum = IXGB_READ_REG(hw, RXCSUM);
908		rxcsum |= IXGB_RXCSUM_TUOFL;
909		IXGB_WRITE_REG(hw, RXCSUM, rxcsum);
910	}
911
912	/* Enable Receives */
913
914	IXGB_WRITE_REG(hw, RCTL, rctl);
915}
916
917/**
918 * ixgb_free_tx_resources - Free Tx Resources
919 * @adapter: board private structure
920 *
921 * Free all transmit software resources
922 **/
923
924void
925ixgb_free_tx_resources(struct ixgb_adapter *adapter)
926{
927	struct pci_dev *pdev = adapter->pdev;
928
929	ixgb_clean_tx_ring(adapter);
930
931	vfree(adapter->tx_ring.buffer_info);
932	adapter->tx_ring.buffer_info = NULL;
933
934	dma_free_coherent(&pdev->dev, adapter->tx_ring.size,
935			  adapter->tx_ring.desc, adapter->tx_ring.dma);
936
937	adapter->tx_ring.desc = NULL;
938}
939
940static void
941ixgb_unmap_and_free_tx_resource(struct ixgb_adapter *adapter,
942                                struct ixgb_buffer *buffer_info)
943{
944	if (buffer_info->dma) {
945		if (buffer_info->mapped_as_page)
946			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
947				       buffer_info->length, DMA_TO_DEVICE);
948		else
949			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
950					 buffer_info->length, DMA_TO_DEVICE);
951		buffer_info->dma = 0;
952	}
953
954	if (buffer_info->skb) {
955		dev_kfree_skb_any(buffer_info->skb);
956		buffer_info->skb = NULL;
957	}
958	buffer_info->time_stamp = 0;
959	/* these fields must always be initialized in tx
960	 * buffer_info->length = 0;
961	 * buffer_info->next_to_watch = 0; */
962}
963
964/**
965 * ixgb_clean_tx_ring - Free Tx Buffers
966 * @adapter: board private structure
967 **/
968
969static void
970ixgb_clean_tx_ring(struct ixgb_adapter *adapter)
971{
972	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
973	struct ixgb_buffer *buffer_info;
974	unsigned long size;
975	unsigned int i;
976
977	/* Free all the Tx ring sk_buffs */
978
979	for (i = 0; i < tx_ring->count; i++) {
980		buffer_info = &tx_ring->buffer_info[i];
981		ixgb_unmap_and_free_tx_resource(adapter, buffer_info);
982	}
983
984	size = sizeof(struct ixgb_buffer) * tx_ring->count;
985	memset(tx_ring->buffer_info, 0, size);
986
987	/* Zero out the descriptor ring */
988
989	memset(tx_ring->desc, 0, tx_ring->size);
990
991	tx_ring->next_to_use = 0;
992	tx_ring->next_to_clean = 0;
993
994	IXGB_WRITE_REG(&adapter->hw, TDH, 0);
995	IXGB_WRITE_REG(&adapter->hw, TDT, 0);
996}
997
998/**
999 * ixgb_free_rx_resources - Free Rx Resources
1000 * @adapter: board private structure
1001 *
1002 * Free all receive software resources
1003 **/
1004
1005void
1006ixgb_free_rx_resources(struct ixgb_adapter *adapter)
1007{
1008	struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
1009	struct pci_dev *pdev = adapter->pdev;
1010
1011	ixgb_clean_rx_ring(adapter);
1012
1013	vfree(rx_ring->buffer_info);
1014	rx_ring->buffer_info = NULL;
1015
1016	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
1017			  rx_ring->dma);
1018
1019	rx_ring->desc = NULL;
1020}
1021
1022/**
1023 * ixgb_clean_rx_ring - Free Rx Buffers
1024 * @adapter: board private structure
1025 **/
1026
1027static void
1028ixgb_clean_rx_ring(struct ixgb_adapter *adapter)
1029{
1030	struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
1031	struct ixgb_buffer *buffer_info;
1032	struct pci_dev *pdev = adapter->pdev;
1033	unsigned long size;
1034	unsigned int i;
1035
1036	/* Free all the Rx ring sk_buffs */
1037
1038	for (i = 0; i < rx_ring->count; i++) {
1039		buffer_info = &rx_ring->buffer_info[i];
1040		if (buffer_info->dma) {
1041			dma_unmap_single(&pdev->dev,
1042					 buffer_info->dma,
1043					 buffer_info->length,
1044					 DMA_FROM_DEVICE);
1045			buffer_info->dma = 0;
1046			buffer_info->length = 0;
1047		}
1048
1049		if (buffer_info->skb) {
1050			dev_kfree_skb(buffer_info->skb);
1051			buffer_info->skb = NULL;
1052		}
1053	}
1054
1055	size = sizeof(struct ixgb_buffer) * rx_ring->count;
1056	memset(rx_ring->buffer_info, 0, size);
1057
1058	/* Zero out the descriptor ring */
1059
1060	memset(rx_ring->desc, 0, rx_ring->size);
1061
1062	rx_ring->next_to_clean = 0;
1063	rx_ring->next_to_use = 0;
1064
1065	IXGB_WRITE_REG(&adapter->hw, RDH, 0);
1066	IXGB_WRITE_REG(&adapter->hw, RDT, 0);
1067}
1068
1069/**
1070 * ixgb_set_mac - Change the Ethernet Address of the NIC
1071 * @netdev: network interface device structure
1072 * @p: pointer to an address structure
1073 *
1074 * Returns 0 on success, negative on failure
1075 **/
1076
1077static int
1078ixgb_set_mac(struct net_device *netdev, void *p)
1079{
1080	struct ixgb_adapter *adapter = netdev_priv(netdev);
1081	struct sockaddr *addr = p;
1082
1083	if (!is_valid_ether_addr(addr->sa_data))
1084		return -EADDRNOTAVAIL;
1085
1086	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
1087
1088	ixgb_rar_set(&adapter->hw, addr->sa_data, 0);
1089
1090	return 0;
1091}
1092
1093/**
1094 * ixgb_set_multi - Multicast and Promiscuous mode set
1095 * @netdev: network interface device structure
1096 *
1097 * The set_multi entry point is called whenever the multicast address
1098 * list or the network interface flags are updated.  This routine is
1099 * responsible for configuring the hardware for proper multicast,
1100 * promiscuous mode, and all-multi behavior.
1101 **/
1102
1103static void
1104ixgb_set_multi(struct net_device *netdev)
1105{
1106	struct ixgb_adapter *adapter = netdev_priv(netdev);
1107	struct ixgb_hw *hw = &adapter->hw;
1108	struct netdev_hw_addr *ha;
1109	u32 rctl;
1110
1111	/* Check for Promiscuous and All Multicast modes */
1112
1113	rctl = IXGB_READ_REG(hw, RCTL);
1114
1115	if (netdev->flags & IFF_PROMISC) {
1116		rctl |= (IXGB_RCTL_UPE | IXGB_RCTL_MPE);
1117		/* disable VLAN filtering */
1118		rctl &= ~IXGB_RCTL_CFIEN;
1119		rctl &= ~IXGB_RCTL_VFE;
1120	} else {
1121		if (netdev->flags & IFF_ALLMULTI) {
1122			rctl |= IXGB_RCTL_MPE;
1123			rctl &= ~IXGB_RCTL_UPE;
1124		} else {
1125			rctl &= ~(IXGB_RCTL_UPE | IXGB_RCTL_MPE);
1126		}
1127		/* enable VLAN filtering */
1128		rctl |= IXGB_RCTL_VFE;
1129		rctl &= ~IXGB_RCTL_CFIEN;
1130	}
1131
1132	if (netdev_mc_count(netdev) > IXGB_MAX_NUM_MULTICAST_ADDRESSES) {
1133		rctl |= IXGB_RCTL_MPE;
1134		IXGB_WRITE_REG(hw, RCTL, rctl);
1135	} else {
1136		u8 *mta = kmalloc(IXGB_MAX_NUM_MULTICAST_ADDRESSES *
1137			      ETH_ALEN, GFP_ATOMIC);
1138		u8 *addr;
1139		if (!mta)
1140			goto alloc_failed;
1141
1142		IXGB_WRITE_REG(hw, RCTL, rctl);
1143
1144		addr = mta;
1145		netdev_for_each_mc_addr(ha, netdev) {
1146			memcpy(addr, ha->addr, ETH_ALEN);
1147			addr += ETH_ALEN;
1148		}
1149
1150		ixgb_mc_addr_list_update(hw, mta, netdev_mc_count(netdev), 0);
1151		kfree(mta);
1152	}
1153
1154alloc_failed:
1155	if (netdev->features & NETIF_F_HW_VLAN_RX)
1156		ixgb_vlan_strip_enable(adapter);
1157	else
1158		ixgb_vlan_strip_disable(adapter);
1159
1160}
1161
1162/**
1163 * ixgb_watchdog - Timer Call-back
1164 * @data: pointer to netdev cast into an unsigned long
1165 **/
1166
1167static void
1168ixgb_watchdog(unsigned long data)
1169{
1170	struct ixgb_adapter *adapter = (struct ixgb_adapter *)data;
1171	struct net_device *netdev = adapter->netdev;
1172	struct ixgb_desc_ring *txdr = &adapter->tx_ring;
1173
1174	ixgb_check_for_link(&adapter->hw);
1175
1176	if (ixgb_check_for_bad_link(&adapter->hw)) {
1177		/* force the reset path */
1178		netif_stop_queue(netdev);
1179	}
1180
1181	if (adapter->hw.link_up) {
1182		if (!netif_carrier_ok(netdev)) {
1183			netdev_info(netdev,
1184				    "NIC Link is Up 10 Gbps Full Duplex, Flow Control: %s\n",
1185				    (adapter->hw.fc.type == ixgb_fc_full) ?
1186				    "RX/TX" :
1187				    (adapter->hw.fc.type == ixgb_fc_rx_pause) ?
1188				     "RX" :
1189				    (adapter->hw.fc.type == ixgb_fc_tx_pause) ?
1190				    "TX" : "None");
1191			adapter->link_speed = 10000;
1192			adapter->link_duplex = FULL_DUPLEX;
1193			netif_carrier_on(netdev);
1194		}
1195	} else {
1196		if (netif_carrier_ok(netdev)) {
1197			adapter->link_speed = 0;
1198			adapter->link_duplex = 0;
1199			netdev_info(netdev, "NIC Link is Down\n");
1200			netif_carrier_off(netdev);
1201		}
1202	}
1203
1204	ixgb_update_stats(adapter);
1205
1206	if (!netif_carrier_ok(netdev)) {
1207		if (IXGB_DESC_UNUSED(txdr) + 1 < txdr->count) {
1208			/* We've lost link, so the controller stops DMA,
1209			 * but we've got queued Tx work that's never going
1210			 * to get done, so reset controller to flush Tx.
1211			 * (Do the reset outside of interrupt context). */
1212			schedule_work(&adapter->tx_timeout_task);
1213			/* return immediately since reset is imminent */
1214			return;
1215		}
1216	}
1217
1218	/* Force detection of hung controller every watchdog period */
1219	adapter->detect_tx_hung = true;
1220
1221	/* generate an interrupt to force clean up of any stragglers */
1222	IXGB_WRITE_REG(&adapter->hw, ICS, IXGB_INT_TXDW);
1223
1224	/* Reset the timer */
1225	mod_timer(&adapter->watchdog_timer, jiffies + 2 * HZ);
1226}
1227
1228#define IXGB_TX_FLAGS_CSUM		0x00000001
1229#define IXGB_TX_FLAGS_VLAN		0x00000002
1230#define IXGB_TX_FLAGS_TSO		0x00000004
1231
1232static int
1233ixgb_tso(struct ixgb_adapter *adapter, struct sk_buff *skb)
1234{
1235	struct ixgb_context_desc *context_desc;
1236	unsigned int i;
1237	u8 ipcss, ipcso, tucss, tucso, hdr_len;
1238	u16 ipcse, tucse, mss;
1239	int err;
1240
1241	if (likely(skb_is_gso(skb))) {
1242		struct ixgb_buffer *buffer_info;
1243		struct iphdr *iph;
1244
1245		if (skb_header_cloned(skb)) {
1246			err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
1247			if (err)
1248				return err;
1249		}
1250
1251		hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
1252		mss = skb_shinfo(skb)->gso_size;
1253		iph = ip_hdr(skb);
1254		iph->tot_len = 0;
1255		iph->check = 0;
1256		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
1257							 iph->daddr, 0,
1258							 IPPROTO_TCP, 0);
1259		ipcss = skb_network_offset(skb);
1260		ipcso = (void *)&(iph->check) - (void *)skb->data;
1261		ipcse = skb_transport_offset(skb) - 1;
1262		tucss = skb_transport_offset(skb);
1263		tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
1264		tucse = 0;
1265
1266		i = adapter->tx_ring.next_to_use;
1267		context_desc = IXGB_CONTEXT_DESC(adapter->tx_ring, i);
1268		buffer_info = &adapter->tx_ring.buffer_info[i];
1269		WARN_ON(buffer_info->dma != 0);
1270
1271		context_desc->ipcss = ipcss;
1272		context_desc->ipcso = ipcso;
1273		context_desc->ipcse = cpu_to_le16(ipcse);
1274		context_desc->tucss = tucss;
1275		context_desc->tucso = tucso;
1276		context_desc->tucse = cpu_to_le16(tucse);
1277		context_desc->mss = cpu_to_le16(mss);
1278		context_desc->hdr_len = hdr_len;
1279		context_desc->status = 0;
1280		context_desc->cmd_type_len = cpu_to_le32(
1281						  IXGB_CONTEXT_DESC_TYPE
1282						| IXGB_CONTEXT_DESC_CMD_TSE
1283						| IXGB_CONTEXT_DESC_CMD_IP
1284						| IXGB_CONTEXT_DESC_CMD_TCP
1285						| IXGB_CONTEXT_DESC_CMD_IDE
1286						| (skb->len - (hdr_len)));
1287
1288
1289		if (++i == adapter->tx_ring.count) i = 0;
1290		adapter->tx_ring.next_to_use = i;
1291
1292		return 1;
1293	}
1294
1295	return 0;
1296}
1297
1298static bool
1299ixgb_tx_csum(struct ixgb_adapter *adapter, struct sk_buff *skb)
1300{
1301	struct ixgb_context_desc *context_desc;
1302	unsigned int i;
1303	u8 css, cso;
1304
1305	if (likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
1306		struct ixgb_buffer *buffer_info;
1307		css = skb_checksum_start_offset(skb);
1308		cso = css + skb->csum_offset;
1309
1310		i = adapter->tx_ring.next_to_use;
1311		context_desc = IXGB_CONTEXT_DESC(adapter->tx_ring, i);
1312		buffer_info = &adapter->tx_ring.buffer_info[i];
1313		WARN_ON(buffer_info->dma != 0);
1314
1315		context_desc->tucss = css;
1316		context_desc->tucso = cso;
1317		context_desc->tucse = 0;
1318		/* zero out any previously existing data in one instruction */
1319		*(u32 *)&(context_desc->ipcss) = 0;
1320		context_desc->status = 0;
1321		context_desc->hdr_len = 0;
1322		context_desc->mss = 0;
1323		context_desc->cmd_type_len =
1324			cpu_to_le32(IXGB_CONTEXT_DESC_TYPE
1325				    | IXGB_TX_DESC_CMD_IDE);
1326
1327		if (++i == adapter->tx_ring.count) i = 0;
1328		adapter->tx_ring.next_to_use = i;
1329
1330		return true;
1331	}
1332
1333	return false;
1334}
1335
1336#define IXGB_MAX_TXD_PWR	14
1337#define IXGB_MAX_DATA_PER_TXD	(1<<IXGB_MAX_TXD_PWR)
1338
1339static int
1340ixgb_tx_map(struct ixgb_adapter *adapter, struct sk_buff *skb,
1341	    unsigned int first)
1342{
1343	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
1344	struct pci_dev *pdev = adapter->pdev;
1345	struct ixgb_buffer *buffer_info;
1346	int len = skb_headlen(skb);
1347	unsigned int offset = 0, size, count = 0, i;
1348	unsigned int mss = skb_shinfo(skb)->gso_size;
1349	unsigned int nr_frags = skb_shinfo(skb)->nr_frags;
1350	unsigned int f;
1351
1352	i = tx_ring->next_to_use;
1353
1354	while (len) {
1355		buffer_info = &tx_ring->buffer_info[i];
1356		size = min(len, IXGB_MAX_DATA_PER_TXD);
1357		/* Workaround for premature desc write-backs
1358		 * in TSO mode.  Append 4-byte sentinel desc */
1359		if (unlikely(mss && !nr_frags && size == len && size > 8))
1360			size -= 4;
1361
1362		buffer_info->length = size;
1363		WARN_ON(buffer_info->dma != 0);
1364		buffer_info->time_stamp = jiffies;
1365		buffer_info->mapped_as_page = false;
1366		buffer_info->dma = dma_map_single(&pdev->dev,
1367						  skb->data + offset,
1368						  size, DMA_TO_DEVICE);
1369		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
1370			goto dma_error;
1371		buffer_info->next_to_watch = 0;
1372
1373		len -= size;
1374		offset += size;
1375		count++;
1376		if (len) {
1377			i++;
1378			if (i == tx_ring->count)
1379				i = 0;
1380		}
1381	}
1382
1383	for (f = 0; f < nr_frags; f++) {
1384		const struct skb_frag_struct *frag;
1385
1386		frag = &skb_shinfo(skb)->frags[f];
1387		len = skb_frag_size(frag);
1388		offset = 0;
1389
1390		while (len) {
1391			i++;
1392			if (i == tx_ring->count)
1393				i = 0;
1394
1395			buffer_info = &tx_ring->buffer_info[i];
1396			size = min(len, IXGB_MAX_DATA_PER_TXD);
1397
1398			/* Workaround for premature desc write-backs
1399			 * in TSO mode.  Append 4-byte sentinel desc */
1400			if (unlikely(mss && (f == (nr_frags - 1))
1401				     && size == len && size > 8))
1402				size -= 4;
1403
1404			buffer_info->length = size;
1405			buffer_info->time_stamp = jiffies;
1406			buffer_info->mapped_as_page = true;
1407			buffer_info->dma =
1408				skb_frag_dma_map(&pdev->dev, frag, offset, size,
1409						 DMA_TO_DEVICE);
1410			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
1411				goto dma_error;
1412			buffer_info->next_to_watch = 0;
1413
1414			len -= size;
1415			offset += size;
1416			count++;
1417		}
1418	}
1419	tx_ring->buffer_info[i].skb = skb;
1420	tx_ring->buffer_info[first].next_to_watch = i;
1421
1422	return count;
1423
1424dma_error:
1425	dev_err(&pdev->dev, "TX DMA map failed\n");
1426	buffer_info->dma = 0;
1427	if (count)
1428		count--;
1429
1430	while (count--) {
1431		if (i==0)
1432			i += tx_ring->count;
1433		i--;
1434		buffer_info = &tx_ring->buffer_info[i];
1435		ixgb_unmap_and_free_tx_resource(adapter, buffer_info);
1436	}
1437
1438	return 0;
1439}
1440
1441static void
1442ixgb_tx_queue(struct ixgb_adapter *adapter, int count, int vlan_id,int tx_flags)
1443{
1444	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
1445	struct ixgb_tx_desc *tx_desc = NULL;
1446	struct ixgb_buffer *buffer_info;
1447	u32 cmd_type_len = adapter->tx_cmd_type;
1448	u8 status = 0;
1449	u8 popts = 0;
1450	unsigned int i;
1451
1452	if (tx_flags & IXGB_TX_FLAGS_TSO) {
1453		cmd_type_len |= IXGB_TX_DESC_CMD_TSE;
1454		popts |= (IXGB_TX_DESC_POPTS_IXSM | IXGB_TX_DESC_POPTS_TXSM);
1455	}
1456
1457	if (tx_flags & IXGB_TX_FLAGS_CSUM)
1458		popts |= IXGB_TX_DESC_POPTS_TXSM;
1459
1460	if (tx_flags & IXGB_TX_FLAGS_VLAN)
1461		cmd_type_len |= IXGB_TX_DESC_CMD_VLE;
1462
1463	i = tx_ring->next_to_use;
1464
1465	while (count--) {
1466		buffer_info = &tx_ring->buffer_info[i];
1467		tx_desc = IXGB_TX_DESC(*tx_ring, i);
1468		tx_desc->buff_addr = cpu_to_le64(buffer_info->dma);
1469		tx_desc->cmd_type_len =
1470			cpu_to_le32(cmd_type_len | buffer_info->length);
1471		tx_desc->status = status;
1472		tx_desc->popts = popts;
1473		tx_desc->vlan = cpu_to_le16(vlan_id);
1474
1475		if (++i == tx_ring->count) i = 0;
1476	}
1477
1478	tx_desc->cmd_type_len |=
1479		cpu_to_le32(IXGB_TX_DESC_CMD_EOP | IXGB_TX_DESC_CMD_RS);
1480
1481	/* Force memory writes to complete before letting h/w
1482	 * know there are new descriptors to fetch.  (Only
1483	 * applicable for weak-ordered memory model archs,
1484	 * such as IA-64). */
1485	wmb();
1486
1487	tx_ring->next_to_use = i;
1488	IXGB_WRITE_REG(&adapter->hw, TDT, i);
1489}
1490
1491static int __ixgb_maybe_stop_tx(struct net_device *netdev, int size)
1492{
1493	struct ixgb_adapter *adapter = netdev_priv(netdev);
1494	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
1495
1496	netif_stop_queue(netdev);
1497	/* Herbert's original patch had:
1498	 *  smp_mb__after_netif_stop_queue();
1499	 * but since that doesn't exist yet, just open code it. */
1500	smp_mb();
1501
1502	/* We need to check again in a case another CPU has just
1503	 * made room available. */
1504	if (likely(IXGB_DESC_UNUSED(tx_ring) < size))
1505		return -EBUSY;
1506
1507	/* A reprieve! */
1508	netif_start_queue(netdev);
1509	++adapter->restart_queue;
1510	return 0;
1511}
1512
1513static int ixgb_maybe_stop_tx(struct net_device *netdev,
1514                              struct ixgb_desc_ring *tx_ring, int size)
1515{
1516	if (likely(IXGB_DESC_UNUSED(tx_ring) >= size))
1517		return 0;
1518	return __ixgb_maybe_stop_tx(netdev, size);
1519}
1520
1521
1522/* Tx Descriptors needed, worst case */
1523#define TXD_USE_COUNT(S) (((S) >> IXGB_MAX_TXD_PWR) + \
1524			 (((S) & (IXGB_MAX_DATA_PER_TXD - 1)) ? 1 : 0))
1525#define DESC_NEEDED TXD_USE_COUNT(IXGB_MAX_DATA_PER_TXD) /* skb->date */ + \
1526	MAX_SKB_FRAGS * TXD_USE_COUNT(PAGE_SIZE) + 1 /* for context */ \
1527	+ 1 /* one more needed for sentinel TSO workaround */
1528
1529static netdev_tx_t
1530ixgb_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
1531{
1532	struct ixgb_adapter *adapter = netdev_priv(netdev);
1533	unsigned int first;
1534	unsigned int tx_flags = 0;
1535	int vlan_id = 0;
1536	int count = 0;
1537	int tso;
1538
1539	if (test_bit(__IXGB_DOWN, &adapter->flags)) {
1540		dev_kfree_skb(skb);
1541		return NETDEV_TX_OK;
1542	}
1543
1544	if (skb->len <= 0) {
1545		dev_kfree_skb(skb);
1546		return NETDEV_TX_OK;
1547	}
1548
1549	if (unlikely(ixgb_maybe_stop_tx(netdev, &adapter->tx_ring,
1550                     DESC_NEEDED)))
1551		return NETDEV_TX_BUSY;
1552
1553	if (vlan_tx_tag_present(skb)) {
1554		tx_flags |= IXGB_TX_FLAGS_VLAN;
1555		vlan_id = vlan_tx_tag_get(skb);
1556	}
1557
1558	first = adapter->tx_ring.next_to_use;
1559
1560	tso = ixgb_tso(adapter, skb);
1561	if (tso < 0) {
1562		dev_kfree_skb(skb);
1563		return NETDEV_TX_OK;
1564	}
1565
1566	if (likely(tso))
1567		tx_flags |= IXGB_TX_FLAGS_TSO;
1568	else if (ixgb_tx_csum(adapter, skb))
1569		tx_flags |= IXGB_TX_FLAGS_CSUM;
1570
1571	count = ixgb_tx_map(adapter, skb, first);
1572
1573	if (count) {
1574		ixgb_tx_queue(adapter, count, vlan_id, tx_flags);
1575		/* Make sure there is space in the ring for the next send. */
1576		ixgb_maybe_stop_tx(netdev, &adapter->tx_ring, DESC_NEEDED);
1577
1578	} else {
1579		dev_kfree_skb_any(skb);
1580		adapter->tx_ring.buffer_info[first].time_stamp = 0;
1581		adapter->tx_ring.next_to_use = first;
1582	}
1583
1584	return NETDEV_TX_OK;
1585}
1586
1587/**
1588 * ixgb_tx_timeout - Respond to a Tx Hang
1589 * @netdev: network interface device structure
1590 **/
1591
1592static void
1593ixgb_tx_timeout(struct net_device *netdev)
1594{
1595	struct ixgb_adapter *adapter = netdev_priv(netdev);
1596
1597	/* Do the reset outside of interrupt context */
1598	schedule_work(&adapter->tx_timeout_task);
1599}
1600
1601static void
1602ixgb_tx_timeout_task(struct work_struct *work)
1603{
1604	struct ixgb_adapter *adapter =
1605		container_of(work, struct ixgb_adapter, tx_timeout_task);
1606
1607	adapter->tx_timeout_count++;
1608	ixgb_down(adapter, true);
1609	ixgb_up(adapter);
1610}
1611
1612/**
1613 * ixgb_get_stats - Get System Network Statistics
1614 * @netdev: network interface device structure
1615 *
1616 * Returns the address of the device statistics structure.
1617 * The statistics are actually updated from the timer callback.
1618 **/
1619
1620static struct net_device_stats *
1621ixgb_get_stats(struct net_device *netdev)
1622{
1623	return &netdev->stats;
1624}
1625
1626/**
1627 * ixgb_change_mtu - Change the Maximum Transfer Unit
1628 * @netdev: network interface device structure
1629 * @new_mtu: new value for maximum frame size
1630 *
1631 * Returns 0 on success, negative on failure
1632 **/
1633
1634static int
1635ixgb_change_mtu(struct net_device *netdev, int new_mtu)
1636{
1637	struct ixgb_adapter *adapter = netdev_priv(netdev);
1638	int max_frame = new_mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
1639	int old_max_frame = netdev->mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
1640
1641	/* MTU < 68 is an error for IPv4 traffic, just don't allow it */
1642	if ((new_mtu < 68) ||
1643	    (max_frame > IXGB_MAX_JUMBO_FRAME_SIZE + ENET_FCS_LENGTH)) {
1644		netif_err(adapter, probe, adapter->netdev,
1645			  "Invalid MTU setting %d\n", new_mtu);
1646		return -EINVAL;
1647	}
1648
1649	if (old_max_frame == max_frame)
1650		return 0;
1651
1652	if (netif_running(netdev))
1653		ixgb_down(adapter, true);
1654
1655	adapter->rx_buffer_len = max_frame + 8; /* + 8 for errata */
1656
1657	netdev->mtu = new_mtu;
1658
1659	if (netif_running(netdev))
1660		ixgb_up(adapter);
1661
1662	return 0;
1663}
1664
1665/**
1666 * ixgb_update_stats - Update the board statistics counters.
1667 * @adapter: board private structure
1668 **/
1669
1670void
1671ixgb_update_stats(struct ixgb_adapter *adapter)
1672{
1673	struct net_device *netdev = adapter->netdev;
1674	struct pci_dev *pdev = adapter->pdev;
1675
1676	/* Prevent stats update while adapter is being reset */
1677	if (pci_channel_offline(pdev))
1678		return;
1679
1680	if ((netdev->flags & IFF_PROMISC) || (netdev->flags & IFF_ALLMULTI) ||
1681	   (netdev_mc_count(netdev) > IXGB_MAX_NUM_MULTICAST_ADDRESSES)) {
1682		u64 multi = IXGB_READ_REG(&adapter->hw, MPRCL);
1683		u32 bcast_l = IXGB_READ_REG(&adapter->hw, BPRCL);
1684		u32 bcast_h = IXGB_READ_REG(&adapter->hw, BPRCH);
1685		u64 bcast = ((u64)bcast_h << 32) | bcast_l;
1686
1687		multi |= ((u64)IXGB_READ_REG(&adapter->hw, MPRCH) << 32);
1688		/* fix up multicast stats by removing broadcasts */
1689		if (multi >= bcast)
1690			multi -= bcast;
1691
1692		adapter->stats.mprcl += (multi & 0xFFFFFFFF);
1693		adapter->stats.mprch += (multi >> 32);
1694		adapter->stats.bprcl += bcast_l;
1695		adapter->stats.bprch += bcast_h;
1696	} else {
1697		adapter->stats.mprcl += IXGB_READ_REG(&adapter->hw, MPRCL);
1698		adapter->stats.mprch += IXGB_READ_REG(&adapter->hw, MPRCH);
1699		adapter->stats.bprcl += IXGB_READ_REG(&adapter->hw, BPRCL);
1700		adapter->stats.bprch += IXGB_READ_REG(&adapter->hw, BPRCH);
1701	}
1702	adapter->stats.tprl += IXGB_READ_REG(&adapter->hw, TPRL);
1703	adapter->stats.tprh += IXGB_READ_REG(&adapter->hw, TPRH);
1704	adapter->stats.gprcl += IXGB_READ_REG(&adapter->hw, GPRCL);
1705	adapter->stats.gprch += IXGB_READ_REG(&adapter->hw, GPRCH);
1706	adapter->stats.uprcl += IXGB_READ_REG(&adapter->hw, UPRCL);
1707	adapter->stats.uprch += IXGB_READ_REG(&adapter->hw, UPRCH);
1708	adapter->stats.vprcl += IXGB_READ_REG(&adapter->hw, VPRCL);
1709	adapter->stats.vprch += IXGB_READ_REG(&adapter->hw, VPRCH);
1710	adapter->stats.jprcl += IXGB_READ_REG(&adapter->hw, JPRCL);
1711	adapter->stats.jprch += IXGB_READ_REG(&adapter->hw, JPRCH);
1712	adapter->stats.gorcl += IXGB_READ_REG(&adapter->hw, GORCL);
1713	adapter->stats.gorch += IXGB_READ_REG(&adapter->hw, GORCH);
1714	adapter->stats.torl += IXGB_READ_REG(&adapter->hw, TORL);
1715	adapter->stats.torh += IXGB_READ_REG(&adapter->hw, TORH);
1716	adapter->stats.rnbc += IXGB_READ_REG(&adapter->hw, RNBC);
1717	adapter->stats.ruc += IXGB_READ_REG(&adapter->hw, RUC);
1718	adapter->stats.roc += IXGB_READ_REG(&adapter->hw, ROC);
1719	adapter->stats.rlec += IXGB_READ_REG(&adapter->hw, RLEC);
1720	adapter->stats.crcerrs += IXGB_READ_REG(&adapter->hw, CRCERRS);
1721	adapter->stats.icbc += IXGB_READ_REG(&adapter->hw, ICBC);
1722	adapter->stats.ecbc += IXGB_READ_REG(&adapter->hw, ECBC);
1723	adapter->stats.mpc += IXGB_READ_REG(&adapter->hw, MPC);
1724	adapter->stats.tptl += IXGB_READ_REG(&adapter->hw, TPTL);
1725	adapter->stats.tpth += IXGB_READ_REG(&adapter->hw, TPTH);
1726	adapter->stats.gptcl += IXGB_READ_REG(&adapter->hw, GPTCL);
1727	adapter->stats.gptch += IXGB_READ_REG(&adapter->hw, GPTCH);
1728	adapter->stats.bptcl += IXGB_READ_REG(&adapter->hw, BPTCL);
1729	adapter->stats.bptch += IXGB_READ_REG(&adapter->hw, BPTCH);
1730	adapter->stats.mptcl += IXGB_READ_REG(&adapter->hw, MPTCL);
1731	adapter->stats.mptch += IXGB_READ_REG(&adapter->hw, MPTCH);
1732	adapter->stats.uptcl += IXGB_READ_REG(&adapter->hw, UPTCL);
1733	adapter->stats.uptch += IXGB_READ_REG(&adapter->hw, UPTCH);
1734	adapter->stats.vptcl += IXGB_READ_REG(&adapter->hw, VPTCL);
1735	adapter->stats.vptch += IXGB_READ_REG(&adapter->hw, VPTCH);
1736	adapter->stats.jptcl += IXGB_READ_REG(&adapter->hw, JPTCL);
1737	adapter->stats.jptch += IXGB_READ_REG(&adapter->hw, JPTCH);
1738	adapter->stats.gotcl += IXGB_READ_REG(&adapter->hw, GOTCL);
1739	adapter->stats.gotch += IXGB_READ_REG(&adapter->hw, GOTCH);
1740	adapter->stats.totl += IXGB_READ_REG(&adapter->hw, TOTL);
1741	adapter->stats.toth += IXGB_READ_REG(&adapter->hw, TOTH);
1742	adapter->stats.dc += IXGB_READ_REG(&adapter->hw, DC);
1743	adapter->stats.plt64c += IXGB_READ_REG(&adapter->hw, PLT64C);
1744	adapter->stats.tsctc += IXGB_READ_REG(&adapter->hw, TSCTC);
1745	adapter->stats.tsctfc += IXGB_READ_REG(&adapter->hw, TSCTFC);
1746	adapter->stats.ibic += IXGB_READ_REG(&adapter->hw, IBIC);
1747	adapter->stats.rfc += IXGB_READ_REG(&adapter->hw, RFC);
1748	adapter->stats.lfc += IXGB_READ_REG(&adapter->hw, LFC);
1749	adapter->stats.pfrc += IXGB_READ_REG(&adapter->hw, PFRC);
1750	adapter->stats.pftc += IXGB_READ_REG(&adapter->hw, PFTC);
1751	adapter->stats.mcfrc += IXGB_READ_REG(&adapter->hw, MCFRC);
1752	adapter->stats.mcftc += IXGB_READ_REG(&adapter->hw, MCFTC);
1753	adapter->stats.xonrxc += IXGB_READ_REG(&adapter->hw, XONRXC);
1754	adapter->stats.xontxc += IXGB_READ_REG(&adapter->hw, XONTXC);
1755	adapter->stats.xoffrxc += IXGB_READ_REG(&adapter->hw, XOFFRXC);
1756	adapter->stats.xofftxc += IXGB_READ_REG(&adapter->hw, XOFFTXC);
1757	adapter->stats.rjc += IXGB_READ_REG(&adapter->hw, RJC);
1758
1759	/* Fill out the OS statistics structure */
1760
1761	netdev->stats.rx_packets = adapter->stats.gprcl;
1762	netdev->stats.tx_packets = adapter->stats.gptcl;
1763	netdev->stats.rx_bytes = adapter->stats.gorcl;
1764	netdev->stats.tx_bytes = adapter->stats.gotcl;
1765	netdev->stats.multicast = adapter->stats.mprcl;
1766	netdev->stats.collisions = 0;
1767
1768	/* ignore RLEC as it reports errors for padded (<64bytes) frames
1769	 * with a length in the type/len field */
1770	netdev->stats.rx_errors =
1771	    /* adapter->stats.rnbc + */ adapter->stats.crcerrs +
1772	    adapter->stats.ruc +
1773	    adapter->stats.roc /*+ adapter->stats.rlec */  +
1774	    adapter->stats.icbc +
1775	    adapter->stats.ecbc + adapter->stats.mpc;
1776
1777	/* see above
1778	 * netdev->stats.rx_length_errors = adapter->stats.rlec;
1779	 */
1780
1781	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
1782	netdev->stats.rx_fifo_errors = adapter->stats.mpc;
1783	netdev->stats.rx_missed_errors = adapter->stats.mpc;
1784	netdev->stats.rx_over_errors = adapter->stats.mpc;
1785
1786	netdev->stats.tx_errors = 0;
1787	netdev->stats.rx_frame_errors = 0;
1788	netdev->stats.tx_aborted_errors = 0;
1789	netdev->stats.tx_carrier_errors = 0;
1790	netdev->stats.tx_fifo_errors = 0;
1791	netdev->stats.tx_heartbeat_errors = 0;
1792	netdev->stats.tx_window_errors = 0;
1793}
1794
1795#define IXGB_MAX_INTR 10
1796/**
1797 * ixgb_intr - Interrupt Handler
1798 * @irq: interrupt number
1799 * @data: pointer to a network interface device structure
1800 **/
1801
1802static irqreturn_t
1803ixgb_intr(int irq, void *data)
1804{
1805	struct net_device *netdev = data;
1806	struct ixgb_adapter *adapter = netdev_priv(netdev);
1807	struct ixgb_hw *hw = &adapter->hw;
1808	u32 icr = IXGB_READ_REG(hw, ICR);
1809
1810	if (unlikely(!icr))
1811		return IRQ_NONE;  /* Not our interrupt */
1812
1813	if (unlikely(icr & (IXGB_INT_RXSEQ | IXGB_INT_LSC)))
1814		if (!test_bit(__IXGB_DOWN, &adapter->flags))
1815			mod_timer(&adapter->watchdog_timer, jiffies);
1816
1817	if (napi_schedule_prep(&adapter->napi)) {
1818
1819		/* Disable interrupts and register for poll. The flush
1820		  of the posted write is intentionally left out.
1821		*/
1822
1823		IXGB_WRITE_REG(&adapter->hw, IMC, ~0);
1824		__napi_schedule(&adapter->napi);
1825	}
1826	return IRQ_HANDLED;
1827}
1828
1829/**
1830 * ixgb_clean - NAPI Rx polling callback
1831 * @adapter: board private structure
1832 **/
1833
1834static int
1835ixgb_clean(struct napi_struct *napi, int budget)
1836{
1837	struct ixgb_adapter *adapter = container_of(napi, struct ixgb_adapter, napi);
1838	int work_done = 0;
1839
1840	ixgb_clean_tx_irq(adapter);
1841	ixgb_clean_rx_irq(adapter, &work_done, budget);
1842
1843	/* If budget not fully consumed, exit the polling mode */
1844	if (work_done < budget) {
1845		napi_complete(napi);
1846		if (!test_bit(__IXGB_DOWN, &adapter->flags))
1847			ixgb_irq_enable(adapter);
1848	}
1849
1850	return work_done;
1851}
1852
1853/**
1854 * ixgb_clean_tx_irq - Reclaim resources after transmit completes
1855 * @adapter: board private structure
1856 **/
1857
1858static bool
1859ixgb_clean_tx_irq(struct ixgb_adapter *adapter)
1860{
1861	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
1862	struct net_device *netdev = adapter->netdev;
1863	struct ixgb_tx_desc *tx_desc, *eop_desc;
1864	struct ixgb_buffer *buffer_info;
1865	unsigned int i, eop;
1866	bool cleaned = false;
1867
1868	i = tx_ring->next_to_clean;
1869	eop = tx_ring->buffer_info[i].next_to_watch;
1870	eop_desc = IXGB_TX_DESC(*tx_ring, eop);
1871
1872	while (eop_desc->status & IXGB_TX_DESC_STATUS_DD) {
1873
1874		rmb(); /* read buffer_info after eop_desc */
1875		for (cleaned = false; !cleaned; ) {
1876			tx_desc = IXGB_TX_DESC(*tx_ring, i);
1877			buffer_info = &tx_ring->buffer_info[i];
1878
1879			if (tx_desc->popts &
1880			   (IXGB_TX_DESC_POPTS_TXSM |
1881			    IXGB_TX_DESC_POPTS_IXSM))
1882				adapter->hw_csum_tx_good++;
1883
1884			ixgb_unmap_and_free_tx_resource(adapter, buffer_info);
1885
1886			*(u32 *)&(tx_desc->status) = 0;
1887
1888			cleaned = (i == eop);
1889			if (++i == tx_ring->count) i = 0;
1890		}
1891
1892		eop = tx_ring->buffer_info[i].next_to_watch;
1893		eop_desc = IXGB_TX_DESC(*tx_ring, eop);
1894	}
1895
1896	tx_ring->next_to_clean = i;
1897
1898	if (unlikely(cleaned && netif_carrier_ok(netdev) &&
1899		     IXGB_DESC_UNUSED(tx_ring) >= DESC_NEEDED)) {
1900		/* Make sure that anybody stopping the queue after this
1901		 * sees the new next_to_clean. */
1902		smp_mb();
1903
1904		if (netif_queue_stopped(netdev) &&
1905		    !(test_bit(__IXGB_DOWN, &adapter->flags))) {
1906			netif_wake_queue(netdev);
1907			++adapter->restart_queue;
1908		}
1909	}
1910
1911	if (adapter->detect_tx_hung) {
1912		/* detect a transmit hang in hardware, this serializes the
1913		 * check with the clearing of time_stamp and movement of i */
1914		adapter->detect_tx_hung = false;
1915		if (tx_ring->buffer_info[eop].time_stamp &&
1916		   time_after(jiffies, tx_ring->buffer_info[eop].time_stamp + HZ)
1917		   && !(IXGB_READ_REG(&adapter->hw, STATUS) &
1918		        IXGB_STATUS_TXOFF)) {
1919			/* detected Tx unit hang */
1920			netif_err(adapter, drv, adapter->netdev,
1921				  "Detected Tx Unit Hang\n"
1922				  "  TDH                  <%x>\n"
1923				  "  TDT                  <%x>\n"
1924				  "  next_to_use          <%x>\n"
1925				  "  next_to_clean        <%x>\n"
1926				  "buffer_info[next_to_clean]\n"
1927				  "  time_stamp           <%lx>\n"
1928				  "  next_to_watch        <%x>\n"
1929				  "  jiffies              <%lx>\n"
1930				  "  next_to_watch.status <%x>\n",
1931				  IXGB_READ_REG(&adapter->hw, TDH),
1932				  IXGB_READ_REG(&adapter->hw, TDT),
1933				  tx_ring->next_to_use,
1934				  tx_ring->next_to_clean,
1935				  tx_ring->buffer_info[eop].time_stamp,
1936				  eop,
1937				  jiffies,
1938				  eop_desc->status);
1939			netif_stop_queue(netdev);
1940		}
1941	}
1942
1943	return cleaned;
1944}
1945
1946/**
1947 * ixgb_rx_checksum - Receive Checksum Offload for 82597.
1948 * @adapter: board private structure
1949 * @rx_desc: receive descriptor
1950 * @sk_buff: socket buffer with received data
1951 **/
1952
1953static void
1954ixgb_rx_checksum(struct ixgb_adapter *adapter,
1955                 struct ixgb_rx_desc *rx_desc,
1956                 struct sk_buff *skb)
1957{
1958	/* Ignore Checksum bit is set OR
1959	 * TCP Checksum has not been calculated
1960	 */
1961	if ((rx_desc->status & IXGB_RX_DESC_STATUS_IXSM) ||
1962	   (!(rx_desc->status & IXGB_RX_DESC_STATUS_TCPCS))) {
1963		skb_checksum_none_assert(skb);
1964		return;
1965	}
1966
1967	/* At this point we know the hardware did the TCP checksum */
1968	/* now look at the TCP checksum error bit */
1969	if (rx_desc->errors & IXGB_RX_DESC_ERRORS_TCPE) {
1970		/* let the stack verify checksum errors */
1971		skb_checksum_none_assert(skb);
1972		adapter->hw_csum_rx_error++;
1973	} else {
1974		/* TCP checksum is good */
1975		skb->ip_summed = CHECKSUM_UNNECESSARY;
1976		adapter->hw_csum_rx_good++;
1977	}
1978}
1979
1980/*
1981 * this should improve performance for small packets with large amounts
1982 * of reassembly being done in the stack
1983 */
1984static void ixgb_check_copybreak(struct net_device *netdev,
1985				 struct ixgb_buffer *buffer_info,
1986				 u32 length, struct sk_buff **skb)
1987{
1988	struct sk_buff *new_skb;
1989
1990	if (length > copybreak)
1991		return;
1992
1993	new_skb = netdev_alloc_skb_ip_align(netdev, length);
1994	if (!new_skb)
1995		return;
1996
1997	skb_copy_to_linear_data_offset(new_skb, -NET_IP_ALIGN,
1998				       (*skb)->data - NET_IP_ALIGN,
1999				       length + NET_IP_ALIGN);
2000	/* save the skb in buffer_info as good */
2001	buffer_info->skb = *skb;
2002	*skb = new_skb;
2003}
2004
2005/**
2006 * ixgb_clean_rx_irq - Send received data up the network stack,
2007 * @adapter: board private structure
2008 **/
2009
2010static bool
2011ixgb_clean_rx_irq(struct ixgb_adapter *adapter, int *work_done, int work_to_do)
2012{
2013	struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
2014	struct net_device *netdev = adapter->netdev;
2015	struct pci_dev *pdev = adapter->pdev;
2016	struct ixgb_rx_desc *rx_desc, *next_rxd;
2017	struct ixgb_buffer *buffer_info, *next_buffer, *next2_buffer;
2018	u32 length;
2019	unsigned int i, j;
2020	int cleaned_count = 0;
2021	bool cleaned = false;
2022
2023	i = rx_ring->next_to_clean;
2024	rx_desc = IXGB_RX_DESC(*rx_ring, i);
2025	buffer_info = &rx_ring->buffer_info[i];
2026
2027	while (rx_desc->status & IXGB_RX_DESC_STATUS_DD) {
2028		struct sk_buff *skb;
2029		u8 status;
2030
2031		if (*work_done >= work_to_do)
2032			break;
2033
2034		(*work_done)++;
2035		rmb();	/* read descriptor and rx_buffer_info after status DD */
2036		status = rx_desc->status;
2037		skb = buffer_info->skb;
2038		buffer_info->skb = NULL;
2039
2040		prefetch(skb->data - NET_IP_ALIGN);
2041
2042		if (++i == rx_ring->count)
2043			i = 0;
2044		next_rxd = IXGB_RX_DESC(*rx_ring, i);
2045		prefetch(next_rxd);
2046
2047		j = i + 1;
2048		if (j == rx_ring->count)
2049			j = 0;
2050		next2_buffer = &rx_ring->buffer_info[j];
2051		prefetch(next2_buffer);
2052
2053		next_buffer = &rx_ring->buffer_info[i];
2054
2055		cleaned = true;
2056		cleaned_count++;
2057
2058		dma_unmap_single(&pdev->dev,
2059				 buffer_info->dma,
2060				 buffer_info->length,
2061				 DMA_FROM_DEVICE);
2062		buffer_info->dma = 0;
2063
2064		length = le16_to_cpu(rx_desc->length);
2065		rx_desc->length = 0;
2066
2067		if (unlikely(!(status & IXGB_RX_DESC_STATUS_EOP))) {
2068
2069			/* All receives must fit into a single buffer */
2070
2071			pr_debug("Receive packet consumed multiple buffers length<%x>\n",
2072				 length);
2073
2074			dev_kfree_skb_irq(skb);
2075			goto rxdesc_done;
2076		}
2077
2078		if (unlikely(rx_desc->errors &
2079		    (IXGB_RX_DESC_ERRORS_CE | IXGB_RX_DESC_ERRORS_SE |
2080		     IXGB_RX_DESC_ERRORS_P | IXGB_RX_DESC_ERRORS_RXE))) {
2081			dev_kfree_skb_irq(skb);
2082			goto rxdesc_done;
2083		}
2084
2085		ixgb_check_copybreak(netdev, buffer_info, length, &skb);
2086
2087		/* Good Receive */
2088		skb_put(skb, length);
2089
2090		/* Receive Checksum Offload */
2091		ixgb_rx_checksum(adapter, rx_desc, skb);
2092
2093		skb->protocol = eth_type_trans(skb, netdev);
2094		if (status & IXGB_RX_DESC_STATUS_VP)
2095			__vlan_hwaccel_put_tag(skb,
2096					       le16_to_cpu(rx_desc->special));
2097
2098		netif_receive_skb(skb);
2099
2100rxdesc_done:
2101		/* clean up descriptor, might be written over by hw */
2102		rx_desc->status = 0;
2103
2104		/* return some buffers to hardware, one at a time is too slow */
2105		if (unlikely(cleaned_count >= IXGB_RX_BUFFER_WRITE)) {
2106			ixgb_alloc_rx_buffers(adapter, cleaned_count);
2107			cleaned_count = 0;
2108		}
2109
2110		/* use prefetched values */
2111		rx_desc = next_rxd;
2112		buffer_info = next_buffer;
2113	}
2114
2115	rx_ring->next_to_clean = i;
2116
2117	cleaned_count = IXGB_DESC_UNUSED(rx_ring);
2118	if (cleaned_count)
2119		ixgb_alloc_rx_buffers(adapter, cleaned_count);
2120
2121	return cleaned;
2122}
2123
2124/**
2125 * ixgb_alloc_rx_buffers - Replace used receive buffers
2126 * @adapter: address of board private structure
2127 **/
2128
2129static void
2130ixgb_alloc_rx_buffers(struct ixgb_adapter *adapter, int cleaned_count)
2131{
2132	struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
2133	struct net_device *netdev = adapter->netdev;
2134	struct pci_dev *pdev = adapter->pdev;
2135	struct ixgb_rx_desc *rx_desc;
2136	struct ixgb_buffer *buffer_info;
2137	struct sk_buff *skb;
2138	unsigned int i;
2139	long cleancount;
2140
2141	i = rx_ring->next_to_use;
2142	buffer_info = &rx_ring->buffer_info[i];
2143	cleancount = IXGB_DESC_UNUSED(rx_ring);
2144
2145
2146	/* leave three descriptors unused */
2147	while (--cleancount > 2 && cleaned_count--) {
2148		/* recycle! its good for you */
2149		skb = buffer_info->skb;
2150		if (skb) {
2151			skb_trim(skb, 0);
2152			goto map_skb;
2153		}
2154
2155		skb = netdev_alloc_skb_ip_align(netdev, adapter->rx_buffer_len);
2156		if (unlikely(!skb)) {
2157			/* Better luck next round */
2158			adapter->alloc_rx_buff_failed++;
2159			break;
2160		}
2161
2162		buffer_info->skb = skb;
2163		buffer_info->length = adapter->rx_buffer_len;
2164map_skb:
2165		buffer_info->dma = dma_map_single(&pdev->dev,
2166		                                  skb->data,
2167		                                  adapter->rx_buffer_len,
2168						  DMA_FROM_DEVICE);
2169
2170		rx_desc = IXGB_RX_DESC(*rx_ring, i);
2171		rx_desc->buff_addr = cpu_to_le64(buffer_info->dma);
2172		/* guarantee DD bit not set now before h/w gets descriptor
2173		 * this is the rest of the workaround for h/w double
2174		 * writeback. */
2175		rx_desc->status = 0;
2176
2177
2178		if (++i == rx_ring->count) i = 0;
2179		buffer_info = &rx_ring->buffer_info[i];
2180	}
2181
2182	if (likely(rx_ring->next_to_use != i)) {
2183		rx_ring->next_to_use = i;
2184		if (unlikely(i-- == 0))
2185			i = (rx_ring->count - 1);
2186
2187		/* Force memory writes to complete before letting h/w
2188		 * know there are new descriptors to fetch.  (Only
2189		 * applicable for weak-ordered memory model archs, such
2190		 * as IA-64). */
2191		wmb();
2192		IXGB_WRITE_REG(&adapter->hw, RDT, i);
2193	}
2194}
2195
2196static void
2197ixgb_vlan_strip_enable(struct ixgb_adapter *adapter)
2198{
2199	u32 ctrl;
2200
2201	/* enable VLAN tag insert/strip */
2202	ctrl = IXGB_READ_REG(&adapter->hw, CTRL0);
2203	ctrl |= IXGB_CTRL0_VME;
2204	IXGB_WRITE_REG(&adapter->hw, CTRL0, ctrl);
2205}
2206
2207static void
2208ixgb_vlan_strip_disable(struct ixgb_adapter *adapter)
2209{
2210	u32 ctrl;
2211
2212	/* disable VLAN tag insert/strip */
2213	ctrl = IXGB_READ_REG(&adapter->hw, CTRL0);
2214	ctrl &= ~IXGB_CTRL0_VME;
2215	IXGB_WRITE_REG(&adapter->hw, CTRL0, ctrl);
2216}
2217
2218static int
2219ixgb_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
2220{
2221	struct ixgb_adapter *adapter = netdev_priv(netdev);
2222	u32 vfta, index;
2223
2224	/* add VID to filter table */
2225
2226	index = (vid >> 5) & 0x7F;
2227	vfta = IXGB_READ_REG_ARRAY(&adapter->hw, VFTA, index);
2228	vfta |= (1 << (vid & 0x1F));
2229	ixgb_write_vfta(&adapter->hw, index, vfta);
2230	set_bit(vid, adapter->active_vlans);
2231
2232	return 0;
2233}
2234
2235static int
2236ixgb_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
2237{
2238	struct ixgb_adapter *adapter = netdev_priv(netdev);
2239	u32 vfta, index;
2240
2241	/* remove VID from filter table */
2242
2243	index = (vid >> 5) & 0x7F;
2244	vfta = IXGB_READ_REG_ARRAY(&adapter->hw, VFTA, index);
2245	vfta &= ~(1 << (vid & 0x1F));
2246	ixgb_write_vfta(&adapter->hw, index, vfta);
2247	clear_bit(vid, adapter->active_vlans);
2248
2249	return 0;
2250}
2251
2252static void
2253ixgb_restore_vlan(struct ixgb_adapter *adapter)
2254{
2255	u16 vid;
2256
2257	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2258		ixgb_vlan_rx_add_vid(adapter->netdev, vid);
2259}
2260
2261#ifdef CONFIG_NET_POLL_CONTROLLER
2262/*
2263 * Polling 'interrupt' - used by things like netconsole to send skbs
2264 * without having to re-enable interrupts. It's not called while
2265 * the interrupt routine is executing.
2266 */
2267
2268static void ixgb_netpoll(struct net_device *dev)
2269{
2270	struct ixgb_adapter *adapter = netdev_priv(dev);
2271
2272	disable_irq(adapter->pdev->irq);
2273	ixgb_intr(adapter->pdev->irq, dev);
2274	enable_irq(adapter->pdev->irq);
2275}
2276#endif
2277
2278/**
2279 * ixgb_io_error_detected() - called when PCI error is detected
2280 * @pdev    pointer to pci device with error
2281 * @state   pci channel state after error
2282 *
2283 * This callback is called by the PCI subsystem whenever
2284 * a PCI bus error is detected.
2285 */
2286static pci_ers_result_t ixgb_io_error_detected(struct pci_dev *pdev,
2287                                               enum pci_channel_state state)
2288{
2289	struct net_device *netdev = pci_get_drvdata(pdev);
2290	struct ixgb_adapter *adapter = netdev_priv(netdev);
2291
2292	netif_device_detach(netdev);
2293
2294	if (state == pci_channel_io_perm_failure)
2295		return PCI_ERS_RESULT_DISCONNECT;
2296
2297	if (netif_running(netdev))
2298		ixgb_down(adapter, true);
2299
2300	pci_disable_device(pdev);
2301
2302	/* Request a slot reset. */
2303	return PCI_ERS_RESULT_NEED_RESET;
2304}
2305
2306/**
2307 * ixgb_io_slot_reset - called after the pci bus has been reset.
2308 * @pdev    pointer to pci device with error
2309 *
2310 * This callback is called after the PCI bus has been reset.
2311 * Basically, this tries to restart the card from scratch.
2312 * This is a shortened version of the device probe/discovery code,
2313 * it resembles the first-half of the ixgb_probe() routine.
2314 */
2315static pci_ers_result_t ixgb_io_slot_reset(struct pci_dev *pdev)
2316{
2317	struct net_device *netdev = pci_get_drvdata(pdev);
2318	struct ixgb_adapter *adapter = netdev_priv(netdev);
2319
2320	if (pci_enable_device(pdev)) {
2321		netif_err(adapter, probe, adapter->netdev,
2322			  "Cannot re-enable PCI device after reset\n");
2323		return PCI_ERS_RESULT_DISCONNECT;
2324	}
2325
2326	/* Perform card reset only on one instance of the card */
2327	if (0 != PCI_FUNC (pdev->devfn))
2328		return PCI_ERS_RESULT_RECOVERED;
2329
2330	pci_set_master(pdev);
2331
2332	netif_carrier_off(netdev);
2333	netif_stop_queue(netdev);
2334	ixgb_reset(adapter);
2335
2336	/* Make sure the EEPROM is good */
2337	if (!ixgb_validate_eeprom_checksum(&adapter->hw)) {
2338		netif_err(adapter, probe, adapter->netdev,
2339			  "After reset, the EEPROM checksum is not valid\n");
2340		return PCI_ERS_RESULT_DISCONNECT;
2341	}
2342	ixgb_get_ee_mac_addr(&adapter->hw, netdev->dev_addr);
2343	memcpy(netdev->perm_addr, netdev->dev_addr, netdev->addr_len);
2344
2345	if (!is_valid_ether_addr(netdev->perm_addr)) {
2346		netif_err(adapter, probe, adapter->netdev,
2347			  "After reset, invalid MAC address\n");
2348		return PCI_ERS_RESULT_DISCONNECT;
2349	}
2350
2351	return PCI_ERS_RESULT_RECOVERED;
2352}
2353
2354/**
2355 * ixgb_io_resume - called when its OK to resume normal operations
2356 * @pdev    pointer to pci device with error
2357 *
2358 * The error recovery driver tells us that its OK to resume
2359 * normal operation. Implementation resembles the second-half
2360 * of the ixgb_probe() routine.
2361 */
2362static void ixgb_io_resume(struct pci_dev *pdev)
2363{
2364	struct net_device *netdev = pci_get_drvdata(pdev);
2365	struct ixgb_adapter *adapter = netdev_priv(netdev);
2366
2367	pci_set_master(pdev);
2368
2369	if (netif_running(netdev)) {
2370		if (ixgb_up(adapter)) {
2371			pr_err("can't bring device back up after reset\n");
2372			return;
2373		}
2374	}
2375
2376	netif_device_attach(netdev);
2377	mod_timer(&adapter->watchdog_timer, jiffies);
2378}
2379
2380/* ixgb_main.c */
2381