]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - drivers/net/igb/igb_main.c
igb: misc cleanups within igb_ethtool.c
[mirror_ubuntu-artful-kernel.git] / drivers / net / igb / igb_main.c
CommitLineData
9d5c8243
AK
1/*******************************************************************************
2
3 Intel(R) Gigabit Ethernet Linux driver
86d5d38f 4 Copyright(c) 2007-2009 Intel Corporation.
9d5c8243
AK
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 e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
24 Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25
26*******************************************************************************/
27
28#include <linux/module.h>
29#include <linux/types.h>
30#include <linux/init.h>
31#include <linux/vmalloc.h>
32#include <linux/pagemap.h>
33#include <linux/netdevice.h>
9d5c8243
AK
34#include <linux/ipv6.h>
35#include <net/checksum.h>
36#include <net/ip6_checksum.h>
c6cb090b 37#include <linux/net_tstamp.h>
9d5c8243
AK
38#include <linux/mii.h>
39#include <linux/ethtool.h>
40#include <linux/if_vlan.h>
41#include <linux/pci.h>
c54106bb 42#include <linux/pci-aspm.h>
9d5c8243
AK
43#include <linux/delay.h>
44#include <linux/interrupt.h>
45#include <linux/if_ether.h>
40a914fa 46#include <linux/aer.h>
421e02f0 47#ifdef CONFIG_IGB_DCA
fe4506b6
JC
48#include <linux/dca.h>
49#endif
9d5c8243
AK
50#include "igb.h"
51
86d5d38f 52#define DRV_VERSION "1.3.16-k2"
9d5c8243
AK
53char igb_driver_name[] = "igb";
54char igb_driver_version[] = DRV_VERSION;
55static const char igb_driver_string[] =
56 "Intel(R) Gigabit Ethernet Network Driver";
86d5d38f 57static const char igb_copyright[] = "Copyright (c) 2007-2009 Intel Corporation.";
9d5c8243 58
9d5c8243
AK
59static const struct e1000_info *igb_info_tbl[] = {
60 [board_82575] = &e1000_82575_info,
61};
62
63static struct pci_device_id igb_pci_tbl[] = {
2d064c06 64 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
9eb2341d 65 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
747d49ba 66 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
2d064c06
AD
67 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
68 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
4703bf73 69 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
c8ea5ea9 70 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
9d5c8243
AK
71 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
72 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
73 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
74 /* required last entry */
75 {0, }
76};
77
78MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
79
80void igb_reset(struct igb_adapter *);
81static int igb_setup_all_tx_resources(struct igb_adapter *);
82static int igb_setup_all_rx_resources(struct igb_adapter *);
83static void igb_free_all_tx_resources(struct igb_adapter *);
84static void igb_free_all_rx_resources(struct igb_adapter *);
06cf2666 85static void igb_setup_mrqc(struct igb_adapter *);
9d5c8243
AK
86void igb_update_stats(struct igb_adapter *);
87static int igb_probe(struct pci_dev *, const struct pci_device_id *);
88static void __devexit igb_remove(struct pci_dev *pdev);
89static int igb_sw_init(struct igb_adapter *);
90static int igb_open(struct net_device *);
91static int igb_close(struct net_device *);
92static void igb_configure_tx(struct igb_adapter *);
93static void igb_configure_rx(struct igb_adapter *);
9d5c8243
AK
94static void igb_clean_all_tx_rings(struct igb_adapter *);
95static void igb_clean_all_rx_rings(struct igb_adapter *);
3b644cf6
MW
96static void igb_clean_tx_ring(struct igb_ring *);
97static void igb_clean_rx_ring(struct igb_ring *);
ff41f8dc 98static void igb_set_rx_mode(struct net_device *);
9d5c8243
AK
99static void igb_update_phy_info(unsigned long);
100static void igb_watchdog(unsigned long);
101static void igb_watchdog_task(struct work_struct *);
b1a436c3 102static netdev_tx_t igb_xmit_frame_adv(struct sk_buff *skb, struct net_device *);
9d5c8243
AK
103static struct net_device_stats *igb_get_stats(struct net_device *);
104static int igb_change_mtu(struct net_device *, int);
105static int igb_set_mac(struct net_device *, void *);
68d480c4 106static void igb_set_uta(struct igb_adapter *adapter);
9d5c8243
AK
107static irqreturn_t igb_intr(int irq, void *);
108static irqreturn_t igb_intr_msi(int irq, void *);
109static irqreturn_t igb_msix_other(int irq, void *);
047e0030 110static irqreturn_t igb_msix_ring(int irq, void *);
421e02f0 111#ifdef CONFIG_IGB_DCA
047e0030 112static void igb_update_dca(struct igb_q_vector *);
fe4506b6 113static void igb_setup_dca(struct igb_adapter *);
421e02f0 114#endif /* CONFIG_IGB_DCA */
047e0030 115static bool igb_clean_tx_irq(struct igb_q_vector *);
661086df 116static int igb_poll(struct napi_struct *, int);
047e0030 117static bool igb_clean_rx_irq_adv(struct igb_q_vector *, int *, int);
9d5c8243
AK
118static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
119static void igb_tx_timeout(struct net_device *);
120static void igb_reset_task(struct work_struct *);
121static void igb_vlan_rx_register(struct net_device *, struct vlan_group *);
122static void igb_vlan_rx_add_vid(struct net_device *, u16);
123static void igb_vlan_rx_kill_vid(struct net_device *, u16);
124static void igb_restore_vlan(struct igb_adapter *);
26ad9178 125static void igb_rar_set_qsel(struct igb_adapter *, u8 *, u32 , u8);
4ae196df
AD
126static void igb_ping_all_vfs(struct igb_adapter *);
127static void igb_msg_task(struct igb_adapter *);
128static int igb_rcv_msg_from_vf(struct igb_adapter *, u32);
4ae196df 129static void igb_vmm_control(struct igb_adapter *);
4ae196df
AD
130static int igb_set_vf_mac(struct igb_adapter *adapter, int, unsigned char *);
131static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
9d5c8243 132
c8159b2d
ED
133static inline void igb_set_vmolr(struct e1000_hw *hw, int vfn)
134{
135 u32 reg_data;
136
137 reg_data = rd32(E1000_VMOLR(vfn));
138 reg_data |= E1000_VMOLR_BAM | /* Accept broadcast */
c8159b2d
ED
139 E1000_VMOLR_ROMPE | /* Accept packets matched in MTA */
140 E1000_VMOLR_AUPE | /* Accept untagged packets */
141 E1000_VMOLR_STRVLAN; /* Strip vlan tags */
142 wr32(E1000_VMOLR(vfn), reg_data);
143}
144
145static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
146 int vfn)
147{
148 struct e1000_hw *hw = &adapter->hw;
149 u32 vmolr;
150
ae641bdc
AD
151 /* if it isn't the PF check to see if VFs are enabled and
152 * increase the size to support vlan tags */
153 if (vfn < adapter->vfs_allocated_count &&
154 adapter->vf_data[vfn].vlans_enabled)
155 size += VLAN_TAG_SIZE;
156
c8159b2d
ED
157 vmolr = rd32(E1000_VMOLR(vfn));
158 vmolr &= ~E1000_VMOLR_RLPML_MASK;
159 vmolr |= size | E1000_VMOLR_LPE;
160 wr32(E1000_VMOLR(vfn), vmolr);
161
162 return 0;
163}
164
9d5c8243 165#ifdef CONFIG_PM
3fe7c4c9 166static int igb_suspend(struct pci_dev *, pm_message_t);
9d5c8243
AK
167static int igb_resume(struct pci_dev *);
168#endif
169static void igb_shutdown(struct pci_dev *);
421e02f0 170#ifdef CONFIG_IGB_DCA
fe4506b6
JC
171static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
172static struct notifier_block dca_notifier = {
173 .notifier_call = igb_notify_dca,
174 .next = NULL,
175 .priority = 0
176};
177#endif
9d5c8243
AK
178#ifdef CONFIG_NET_POLL_CONTROLLER
179/* for netdump / net console */
180static void igb_netpoll(struct net_device *);
181#endif
37680117 182#ifdef CONFIG_PCI_IOV
2a3abf6d
AD
183static unsigned int max_vfs = 0;
184module_param(max_vfs, uint, 0);
185MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate "
186 "per physical function");
187#endif /* CONFIG_PCI_IOV */
188
9d5c8243
AK
189static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
190 pci_channel_state_t);
191static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
192static void igb_io_resume(struct pci_dev *);
193
194static struct pci_error_handlers igb_err_handler = {
195 .error_detected = igb_io_error_detected,
196 .slot_reset = igb_io_slot_reset,
197 .resume = igb_io_resume,
198};
199
200
201static struct pci_driver igb_driver = {
202 .name = igb_driver_name,
203 .id_table = igb_pci_tbl,
204 .probe = igb_probe,
205 .remove = __devexit_p(igb_remove),
206#ifdef CONFIG_PM
207 /* Power Managment Hooks */
208 .suspend = igb_suspend,
209 .resume = igb_resume,
210#endif
211 .shutdown = igb_shutdown,
212 .err_handler = &igb_err_handler
213};
214
7dfc16fa
AD
215static int global_quad_port_a; /* global quad port a indication */
216
9d5c8243
AK
217MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
218MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
219MODULE_LICENSE("GPL");
220MODULE_VERSION(DRV_VERSION);
221
38c845c7
PO
222/**
223 * igb_read_clock - read raw cycle counter (to be used by time counter)
224 */
225static cycle_t igb_read_clock(const struct cyclecounter *tc)
226{
227 struct igb_adapter *adapter =
228 container_of(tc, struct igb_adapter, cycles);
229 struct e1000_hw *hw = &adapter->hw;
c5b9bd5e
AD
230 u64 stamp = 0;
231 int shift = 0;
38c845c7 232
c5b9bd5e
AD
233 stamp |= (u64)rd32(E1000_SYSTIML) << shift;
234 stamp |= (u64)rd32(E1000_SYSTIMH) << (shift + 32);
38c845c7
PO
235 return stamp;
236}
237
9d5c8243
AK
238#ifdef DEBUG
239/**
240 * igb_get_hw_dev_name - return device name string
241 * used by hardware layer to print debugging information
242 **/
243char *igb_get_hw_dev_name(struct e1000_hw *hw)
244{
245 struct igb_adapter *adapter = hw->back;
246 return adapter->netdev->name;
247}
38c845c7
PO
248
249/**
250 * igb_get_time_str - format current NIC and system time as string
251 */
252static char *igb_get_time_str(struct igb_adapter *adapter,
253 char buffer[160])
254{
255 cycle_t hw = adapter->cycles.read(&adapter->cycles);
256 struct timespec nic = ns_to_timespec(timecounter_read(&adapter->clock));
257 struct timespec sys;
258 struct timespec delta;
259 getnstimeofday(&sys);
260
261 delta = timespec_sub(nic, sys);
262
263 sprintf(buffer,
33af6bcc
PO
264 "HW %llu, NIC %ld.%09lus, SYS %ld.%09lus, NIC-SYS %lds + %09luns",
265 hw,
38c845c7
PO
266 (long)nic.tv_sec, nic.tv_nsec,
267 (long)sys.tv_sec, sys.tv_nsec,
268 (long)delta.tv_sec, delta.tv_nsec);
269
270 return buffer;
271}
9d5c8243
AK
272#endif
273
274/**
275 * igb_init_module - Driver Registration Routine
276 *
277 * igb_init_module is the first routine called when the driver is
278 * loaded. All it does is register with the PCI subsystem.
279 **/
280static int __init igb_init_module(void)
281{
282 int ret;
283 printk(KERN_INFO "%s - version %s\n",
284 igb_driver_string, igb_driver_version);
285
286 printk(KERN_INFO "%s\n", igb_copyright);
287
7dfc16fa
AD
288 global_quad_port_a = 0;
289
421e02f0 290#ifdef CONFIG_IGB_DCA
fe4506b6
JC
291 dca_register_notify(&dca_notifier);
292#endif
bbd98fe4
AD
293
294 ret = pci_register_driver(&igb_driver);
9d5c8243
AK
295 return ret;
296}
297
298module_init(igb_init_module);
299
300/**
301 * igb_exit_module - Driver Exit Cleanup Routine
302 *
303 * igb_exit_module is called just before the driver is removed
304 * from memory.
305 **/
306static void __exit igb_exit_module(void)
307{
421e02f0 308#ifdef CONFIG_IGB_DCA
fe4506b6
JC
309 dca_unregister_notify(&dca_notifier);
310#endif
9d5c8243
AK
311 pci_unregister_driver(&igb_driver);
312}
313
314module_exit(igb_exit_module);
315
26bc19ec
AD
316#define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
317/**
318 * igb_cache_ring_register - Descriptor ring to register mapping
319 * @adapter: board private structure to initialize
320 *
321 * Once we know the feature-set enabled for the device, we'll cache
322 * the register offset the descriptor ring is assigned to.
323 **/
324static void igb_cache_ring_register(struct igb_adapter *adapter)
325{
326 int i;
047e0030 327 u32 rbase_offset = adapter->vfs_allocated_count;
26bc19ec
AD
328
329 switch (adapter->hw.mac.type) {
330 case e1000_82576:
331 /* The queues are allocated for virtualization such that VF 0
332 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
333 * In order to avoid collision we start at the first free queue
334 * and continue consuming queues in the same sequence
335 */
336 for (i = 0; i < adapter->num_rx_queues; i++)
1bfaf07b
AD
337 adapter->rx_ring[i].reg_idx = rbase_offset +
338 Q_IDX_82576(i);
26bc19ec 339 for (i = 0; i < adapter->num_tx_queues; i++)
1bfaf07b
AD
340 adapter->tx_ring[i].reg_idx = rbase_offset +
341 Q_IDX_82576(i);
26bc19ec
AD
342 break;
343 case e1000_82575:
344 default:
345 for (i = 0; i < adapter->num_rx_queues; i++)
346 adapter->rx_ring[i].reg_idx = i;
347 for (i = 0; i < adapter->num_tx_queues; i++)
348 adapter->tx_ring[i].reg_idx = i;
349 break;
350 }
351}
352
047e0030
AD
353static void igb_free_queues(struct igb_adapter *adapter)
354{
355 kfree(adapter->tx_ring);
356 kfree(adapter->rx_ring);
357
358 adapter->tx_ring = NULL;
359 adapter->rx_ring = NULL;
360
361 adapter->num_rx_queues = 0;
362 adapter->num_tx_queues = 0;
363}
364
9d5c8243
AK
365/**
366 * igb_alloc_queues - Allocate memory for all rings
367 * @adapter: board private structure to initialize
368 *
369 * We allocate one ring per queue at run-time since we don't know the
370 * number of queues at compile-time.
371 **/
372static int igb_alloc_queues(struct igb_adapter *adapter)
373{
374 int i;
375
376 adapter->tx_ring = kcalloc(adapter->num_tx_queues,
377 sizeof(struct igb_ring), GFP_KERNEL);
378 if (!adapter->tx_ring)
047e0030 379 goto err;
9d5c8243
AK
380
381 adapter->rx_ring = kcalloc(adapter->num_rx_queues,
382 sizeof(struct igb_ring), GFP_KERNEL);
047e0030
AD
383 if (!adapter->rx_ring)
384 goto err;
6eb5a7f1 385
661086df
PWJ
386 for (i = 0; i < adapter->num_tx_queues; i++) {
387 struct igb_ring *ring = &(adapter->tx_ring[i]);
68fd9910 388 ring->count = adapter->tx_ring_count;
661086df 389 ring->queue_index = i;
80785298 390 ring->pdev = adapter->pdev;
e694e964 391 ring->netdev = adapter->netdev;
85ad76b2
AD
392 /* For 82575, context index must be unique per ring. */
393 if (adapter->hw.mac.type == e1000_82575)
394 ring->flags = IGB_RING_FLAG_TX_CTX_IDX;
661086df 395 }
85ad76b2 396
9d5c8243
AK
397 for (i = 0; i < adapter->num_rx_queues; i++) {
398 struct igb_ring *ring = &(adapter->rx_ring[i]);
68fd9910 399 ring->count = adapter->rx_ring_count;
844290e5 400 ring->queue_index = i;
80785298 401 ring->pdev = adapter->pdev;
e694e964 402 ring->netdev = adapter->netdev;
4c844851 403 ring->rx_buffer_len = MAXIMUM_ETHERNET_VLAN_SIZE;
85ad76b2
AD
404 ring->flags = IGB_RING_FLAG_RX_CSUM; /* enable rx checksum */
405 /* set flag indicating ring supports SCTP checksum offload */
406 if (adapter->hw.mac.type >= e1000_82576)
407 ring->flags |= IGB_RING_FLAG_RX_SCTP_CSUM;
9d5c8243 408 }
26bc19ec
AD
409
410 igb_cache_ring_register(adapter);
9d5c8243 411
047e0030 412 return 0;
a88f10ec 413
047e0030
AD
414err:
415 igb_free_queues(adapter);
d1a8c9e1 416
047e0030 417 return -ENOMEM;
a88f10ec
AD
418}
419
9d5c8243 420#define IGB_N0_QUEUE -1
047e0030 421static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
9d5c8243
AK
422{
423 u32 msixbm = 0;
047e0030 424 struct igb_adapter *adapter = q_vector->adapter;
9d5c8243 425 struct e1000_hw *hw = &adapter->hw;
2d064c06 426 u32 ivar, index;
047e0030
AD
427 int rx_queue = IGB_N0_QUEUE;
428 int tx_queue = IGB_N0_QUEUE;
429
430 if (q_vector->rx_ring)
431 rx_queue = q_vector->rx_ring->reg_idx;
432 if (q_vector->tx_ring)
433 tx_queue = q_vector->tx_ring->reg_idx;
2d064c06
AD
434
435 switch (hw->mac.type) {
436 case e1000_82575:
9d5c8243
AK
437 /* The 82575 assigns vectors using a bitmask, which matches the
438 bitmask for the EICR/EIMS/EIMC registers. To assign one
439 or more queues to a vector, we write the appropriate bits
440 into the MSIXBM register for that vector. */
047e0030 441 if (rx_queue > IGB_N0_QUEUE)
9d5c8243 442 msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
047e0030 443 if (tx_queue > IGB_N0_QUEUE)
9d5c8243 444 msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
9d5c8243 445 array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
047e0030 446 q_vector->eims_value = msixbm;
2d064c06
AD
447 break;
448 case e1000_82576:
26bc19ec 449 /* 82576 uses a table-based method for assigning vectors.
2d064c06
AD
450 Each queue has a single entry in the table to which we write
451 a vector number along with a "valid" bit. Sadly, the layout
452 of the table is somewhat counterintuitive. */
453 if (rx_queue > IGB_N0_QUEUE) {
047e0030 454 index = (rx_queue & 0x7);
2d064c06 455 ivar = array_rd32(E1000_IVAR0, index);
047e0030 456 if (rx_queue < 8) {
26bc19ec
AD
457 /* vector goes into low byte of register */
458 ivar = ivar & 0xFFFFFF00;
459 ivar |= msix_vector | E1000_IVAR_VALID;
047e0030
AD
460 } else {
461 /* vector goes into third byte of register */
462 ivar = ivar & 0xFF00FFFF;
463 ivar |= (msix_vector | E1000_IVAR_VALID) << 16;
2d064c06 464 }
2d064c06
AD
465 array_wr32(E1000_IVAR0, index, ivar);
466 }
467 if (tx_queue > IGB_N0_QUEUE) {
047e0030 468 index = (tx_queue & 0x7);
2d064c06 469 ivar = array_rd32(E1000_IVAR0, index);
047e0030 470 if (tx_queue < 8) {
26bc19ec
AD
471 /* vector goes into second byte of register */
472 ivar = ivar & 0xFFFF00FF;
473 ivar |= (msix_vector | E1000_IVAR_VALID) << 8;
047e0030
AD
474 } else {
475 /* vector goes into high byte of register */
476 ivar = ivar & 0x00FFFFFF;
477 ivar |= (msix_vector | E1000_IVAR_VALID) << 24;
2d064c06 478 }
2d064c06
AD
479 array_wr32(E1000_IVAR0, index, ivar);
480 }
047e0030 481 q_vector->eims_value = 1 << msix_vector;
2d064c06
AD
482 break;
483 default:
484 BUG();
485 break;
486 }
9d5c8243
AK
487}
488
489/**
490 * igb_configure_msix - Configure MSI-X hardware
491 *
492 * igb_configure_msix sets up the hardware to properly
493 * generate MSI-X interrupts.
494 **/
495static void igb_configure_msix(struct igb_adapter *adapter)
496{
497 u32 tmp;
498 int i, vector = 0;
499 struct e1000_hw *hw = &adapter->hw;
500
501 adapter->eims_enable_mask = 0;
9d5c8243
AK
502
503 /* set vector for other causes, i.e. link changes */
2d064c06
AD
504 switch (hw->mac.type) {
505 case e1000_82575:
9d5c8243
AK
506 tmp = rd32(E1000_CTRL_EXT);
507 /* enable MSI-X PBA support*/
508 tmp |= E1000_CTRL_EXT_PBA_CLR;
509
510 /* Auto-Mask interrupts upon ICR read. */
511 tmp |= E1000_CTRL_EXT_EIAME;
512 tmp |= E1000_CTRL_EXT_IRCA;
513
514 wr32(E1000_CTRL_EXT, tmp);
047e0030
AD
515
516 /* enable msix_other interrupt */
517 array_wr32(E1000_MSIXBM(0), vector++,
518 E1000_EIMS_OTHER);
844290e5 519 adapter->eims_other = E1000_EIMS_OTHER;
9d5c8243 520
2d064c06
AD
521 break;
522
523 case e1000_82576:
047e0030
AD
524 /* Turn on MSI-X capability first, or our settings
525 * won't stick. And it will take days to debug. */
526 wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
527 E1000_GPIE_PBA | E1000_GPIE_EIAME |
528 E1000_GPIE_NSICR);
529
530 /* enable msix_other interrupt */
531 adapter->eims_other = 1 << vector;
2d064c06 532 tmp = (vector++ | E1000_IVAR_VALID) << 8;
2d064c06 533
047e0030 534 wr32(E1000_IVAR_MISC, tmp);
2d064c06
AD
535 break;
536 default:
537 /* do nothing, since nothing else supports MSI-X */
538 break;
539 } /* switch (hw->mac.type) */
047e0030
AD
540
541 adapter->eims_enable_mask |= adapter->eims_other;
542
543 for (i = 0; i < adapter->num_q_vectors; i++) {
544 struct igb_q_vector *q_vector = adapter->q_vector[i];
545 igb_assign_vector(q_vector, vector++);
546 adapter->eims_enable_mask |= q_vector->eims_value;
547 }
548
9d5c8243
AK
549 wrfl();
550}
551
552/**
553 * igb_request_msix - Initialize MSI-X interrupts
554 *
555 * igb_request_msix allocates MSI-X vectors and requests interrupts from the
556 * kernel.
557 **/
558static int igb_request_msix(struct igb_adapter *adapter)
559{
560 struct net_device *netdev = adapter->netdev;
047e0030 561 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK
562 int i, err = 0, vector = 0;
563
047e0030
AD
564 err = request_irq(adapter->msix_entries[vector].vector,
565 &igb_msix_other, 0, netdev->name, adapter);
566 if (err)
567 goto out;
568 vector++;
569
570 for (i = 0; i < adapter->num_q_vectors; i++) {
571 struct igb_q_vector *q_vector = adapter->q_vector[i];
572
573 q_vector->itr_register = hw->hw_addr + E1000_EITR(vector);
574
575 if (q_vector->rx_ring && q_vector->tx_ring)
576 sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
577 q_vector->rx_ring->queue_index);
578 else if (q_vector->tx_ring)
579 sprintf(q_vector->name, "%s-tx-%u", netdev->name,
580 q_vector->tx_ring->queue_index);
581 else if (q_vector->rx_ring)
582 sprintf(q_vector->name, "%s-rx-%u", netdev->name,
583 q_vector->rx_ring->queue_index);
9d5c8243 584 else
047e0030
AD
585 sprintf(q_vector->name, "%s-unused", netdev->name);
586
9d5c8243 587 err = request_irq(adapter->msix_entries[vector].vector,
047e0030
AD
588 &igb_msix_ring, 0, q_vector->name,
589 q_vector);
9d5c8243
AK
590 if (err)
591 goto out;
9d5c8243
AK
592 vector++;
593 }
594
9d5c8243
AK
595 igb_configure_msix(adapter);
596 return 0;
597out:
598 return err;
599}
600
601static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
602{
603 if (adapter->msix_entries) {
604 pci_disable_msix(adapter->pdev);
605 kfree(adapter->msix_entries);
606 adapter->msix_entries = NULL;
047e0030 607 } else if (adapter->flags & IGB_FLAG_HAS_MSI) {
9d5c8243 608 pci_disable_msi(adapter->pdev);
047e0030 609 }
9d5c8243
AK
610}
611
047e0030
AD
612/**
613 * igb_free_q_vectors - Free memory allocated for interrupt vectors
614 * @adapter: board private structure to initialize
615 *
616 * This function frees the memory allocated to the q_vectors. In addition if
617 * NAPI is enabled it will delete any references to the NAPI struct prior
618 * to freeing the q_vector.
619 **/
620static void igb_free_q_vectors(struct igb_adapter *adapter)
621{
622 int v_idx;
623
624 for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
625 struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
626 adapter->q_vector[v_idx] = NULL;
627 netif_napi_del(&q_vector->napi);
628 kfree(q_vector);
629 }
630 adapter->num_q_vectors = 0;
631}
632
633/**
634 * igb_clear_interrupt_scheme - reset the device to a state of no interrupts
635 *
636 * This function resets the device so that it has 0 rx queues, tx queues, and
637 * MSI-X interrupts allocated.
638 */
639static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
640{
641 igb_free_queues(adapter);
642 igb_free_q_vectors(adapter);
643 igb_reset_interrupt_capability(adapter);
644}
9d5c8243
AK
645
646/**
647 * igb_set_interrupt_capability - set MSI or MSI-X if supported
648 *
649 * Attempt to configure interrupts using the best available
650 * capabilities of the hardware and kernel.
651 **/
652static void igb_set_interrupt_capability(struct igb_adapter *adapter)
653{
654 int err;
655 int numvecs, i;
656
83b7180d 657 /* Number of supported queues. */
83b7180d
AD
658 adapter->num_rx_queues = min_t(u32, IGB_MAX_RX_QUEUES, num_online_cpus());
659 adapter->num_tx_queues = min_t(u32, IGB_MAX_TX_QUEUES, num_online_cpus());
660
047e0030
AD
661 /* start with one vector for every rx queue */
662 numvecs = adapter->num_rx_queues;
663
664 /* if tx handler is seperate add 1 for every tx queue */
665 numvecs += adapter->num_tx_queues;
666
667 /* store the number of vectors reserved for queues */
668 adapter->num_q_vectors = numvecs;
669
670 /* add 1 vector for link status interrupts */
671 numvecs++;
9d5c8243
AK
672 adapter->msix_entries = kcalloc(numvecs, sizeof(struct msix_entry),
673 GFP_KERNEL);
674 if (!adapter->msix_entries)
675 goto msi_only;
676
677 for (i = 0; i < numvecs; i++)
678 adapter->msix_entries[i].entry = i;
679
680 err = pci_enable_msix(adapter->pdev,
681 adapter->msix_entries,
682 numvecs);
683 if (err == 0)
34a20e89 684 goto out;
9d5c8243
AK
685
686 igb_reset_interrupt_capability(adapter);
687
688 /* If we can't do MSI-X, try MSI */
689msi_only:
2a3abf6d
AD
690#ifdef CONFIG_PCI_IOV
691 /* disable SR-IOV for non MSI-X configurations */
692 if (adapter->vf_data) {
693 struct e1000_hw *hw = &adapter->hw;
694 /* disable iov and allow time for transactions to clear */
695 pci_disable_sriov(adapter->pdev);
696 msleep(500);
697
698 kfree(adapter->vf_data);
699 adapter->vf_data = NULL;
700 wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
701 msleep(100);
702 dev_info(&adapter->pdev->dev, "IOV Disabled\n");
703 }
704#endif
4fc82adf
AD
705 adapter->vfs_allocated_count = 0;
706 adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
9d5c8243 707 adapter->num_rx_queues = 1;
661086df 708 adapter->num_tx_queues = 1;
047e0030 709 adapter->num_q_vectors = 1;
9d5c8243 710 if (!pci_enable_msi(adapter->pdev))
7dfc16fa 711 adapter->flags |= IGB_FLAG_HAS_MSI;
34a20e89 712out:
661086df 713 /* Notify the stack of the (possibly) reduced Tx Queue count. */
fd2ea0a7 714 adapter->netdev->real_num_tx_queues = adapter->num_tx_queues;
9d5c8243
AK
715 return;
716}
717
047e0030
AD
718/**
719 * igb_alloc_q_vectors - Allocate memory for interrupt vectors
720 * @adapter: board private structure to initialize
721 *
722 * We allocate one q_vector per queue interrupt. If allocation fails we
723 * return -ENOMEM.
724 **/
725static int igb_alloc_q_vectors(struct igb_adapter *adapter)
726{
727 struct igb_q_vector *q_vector;
728 struct e1000_hw *hw = &adapter->hw;
729 int v_idx;
730
731 for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
732 q_vector = kzalloc(sizeof(struct igb_q_vector), GFP_KERNEL);
733 if (!q_vector)
734 goto err_out;
735 q_vector->adapter = adapter;
736 q_vector->itr_shift = (hw->mac.type == e1000_82575) ? 16 : 0;
737 q_vector->itr_register = hw->hw_addr + E1000_EITR(0);
738 q_vector->itr_val = IGB_START_ITR;
739 q_vector->set_itr = 1;
740 netif_napi_add(adapter->netdev, &q_vector->napi, igb_poll, 64);
741 adapter->q_vector[v_idx] = q_vector;
742 }
743 return 0;
744
745err_out:
746 while (v_idx) {
747 v_idx--;
748 q_vector = adapter->q_vector[v_idx];
749 netif_napi_del(&q_vector->napi);
750 kfree(q_vector);
751 adapter->q_vector[v_idx] = NULL;
752 }
753 return -ENOMEM;
754}
755
756static void igb_map_rx_ring_to_vector(struct igb_adapter *adapter,
757 int ring_idx, int v_idx)
758{
759 struct igb_q_vector *q_vector;
760
761 q_vector = adapter->q_vector[v_idx];
762 q_vector->rx_ring = &adapter->rx_ring[ring_idx];
763 q_vector->rx_ring->q_vector = q_vector;
4fc82adf
AD
764 q_vector->itr_val = adapter->rx_itr_setting;
765 if (q_vector->itr_val && q_vector->itr_val <= 3)
766 q_vector->itr_val = IGB_START_ITR;
047e0030
AD
767}
768
769static void igb_map_tx_ring_to_vector(struct igb_adapter *adapter,
770 int ring_idx, int v_idx)
771{
772 struct igb_q_vector *q_vector;
773
774 q_vector = adapter->q_vector[v_idx];
775 q_vector->tx_ring = &adapter->tx_ring[ring_idx];
776 q_vector->tx_ring->q_vector = q_vector;
4fc82adf
AD
777 q_vector->itr_val = adapter->tx_itr_setting;
778 if (q_vector->itr_val && q_vector->itr_val <= 3)
779 q_vector->itr_val = IGB_START_ITR;
047e0030
AD
780}
781
782/**
783 * igb_map_ring_to_vector - maps allocated queues to vectors
784 *
785 * This function maps the recently allocated queues to vectors.
786 **/
787static int igb_map_ring_to_vector(struct igb_adapter *adapter)
788{
789 int i;
790 int v_idx = 0;
791
792 if ((adapter->num_q_vectors < adapter->num_rx_queues) ||
793 (adapter->num_q_vectors < adapter->num_tx_queues))
794 return -ENOMEM;
795
796 if (adapter->num_q_vectors >=
797 (adapter->num_rx_queues + adapter->num_tx_queues)) {
798 for (i = 0; i < adapter->num_rx_queues; i++)
799 igb_map_rx_ring_to_vector(adapter, i, v_idx++);
800 for (i = 0; i < adapter->num_tx_queues; i++)
801 igb_map_tx_ring_to_vector(adapter, i, v_idx++);
802 } else {
803 for (i = 0; i < adapter->num_rx_queues; i++) {
804 if (i < adapter->num_tx_queues)
805 igb_map_tx_ring_to_vector(adapter, i, v_idx);
806 igb_map_rx_ring_to_vector(adapter, i, v_idx++);
807 }
808 for (; i < adapter->num_tx_queues; i++)
809 igb_map_tx_ring_to_vector(adapter, i, v_idx++);
810 }
811 return 0;
812}
813
814/**
815 * igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
816 *
817 * This function initializes the interrupts and allocates all of the queues.
818 **/
819static int igb_init_interrupt_scheme(struct igb_adapter *adapter)
820{
821 struct pci_dev *pdev = adapter->pdev;
822 int err;
823
824 igb_set_interrupt_capability(adapter);
825
826 err = igb_alloc_q_vectors(adapter);
827 if (err) {
828 dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
829 goto err_alloc_q_vectors;
830 }
831
832 err = igb_alloc_queues(adapter);
833 if (err) {
834 dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
835 goto err_alloc_queues;
836 }
837
838 err = igb_map_ring_to_vector(adapter);
839 if (err) {
840 dev_err(&pdev->dev, "Invalid q_vector to ring mapping\n");
841 goto err_map_queues;
842 }
843
844
845 return 0;
846err_map_queues:
847 igb_free_queues(adapter);
848err_alloc_queues:
849 igb_free_q_vectors(adapter);
850err_alloc_q_vectors:
851 igb_reset_interrupt_capability(adapter);
852 return err;
853}
854
9d5c8243
AK
855/**
856 * igb_request_irq - initialize interrupts
857 *
858 * Attempts to configure interrupts using the best available
859 * capabilities of the hardware and kernel.
860 **/
861static int igb_request_irq(struct igb_adapter *adapter)
862{
863 struct net_device *netdev = adapter->netdev;
047e0030 864 struct pci_dev *pdev = adapter->pdev;
9d5c8243
AK
865 struct e1000_hw *hw = &adapter->hw;
866 int err = 0;
867
868 if (adapter->msix_entries) {
869 err = igb_request_msix(adapter);
844290e5 870 if (!err)
9d5c8243 871 goto request_done;
9d5c8243 872 /* fall back to MSI */
047e0030 873 igb_clear_interrupt_scheme(adapter);
9d5c8243 874 if (!pci_enable_msi(adapter->pdev))
7dfc16fa 875 adapter->flags |= IGB_FLAG_HAS_MSI;
9d5c8243
AK
876 igb_free_all_tx_resources(adapter);
877 igb_free_all_rx_resources(adapter);
047e0030 878 adapter->num_tx_queues = 1;
9d5c8243 879 adapter->num_rx_queues = 1;
047e0030
AD
880 adapter->num_q_vectors = 1;
881 err = igb_alloc_q_vectors(adapter);
882 if (err) {
883 dev_err(&pdev->dev,
884 "Unable to allocate memory for vectors\n");
885 goto request_done;
886 }
887 err = igb_alloc_queues(adapter);
888 if (err) {
889 dev_err(&pdev->dev,
890 "Unable to allocate memory for queues\n");
891 igb_free_q_vectors(adapter);
892 goto request_done;
893 }
894 igb_setup_all_tx_resources(adapter);
895 igb_setup_all_rx_resources(adapter);
844290e5 896 } else {
2d064c06
AD
897 switch (hw->mac.type) {
898 case e1000_82575:
899 wr32(E1000_MSIXBM(0),
047e0030
AD
900 (E1000_EICR_RX_QUEUE0 |
901 E1000_EICR_TX_QUEUE0 |
902 E1000_EIMS_OTHER));
2d064c06
AD
903 break;
904 case e1000_82576:
905 wr32(E1000_IVAR0, E1000_IVAR_VALID);
906 break;
907 default:
908 break;
909 }
9d5c8243 910 }
844290e5 911
7dfc16fa 912 if (adapter->flags & IGB_FLAG_HAS_MSI) {
9d5c8243 913 err = request_irq(adapter->pdev->irq, &igb_intr_msi, 0,
047e0030 914 netdev->name, adapter);
9d5c8243
AK
915 if (!err)
916 goto request_done;
047e0030 917
9d5c8243
AK
918 /* fall back to legacy interrupts */
919 igb_reset_interrupt_capability(adapter);
7dfc16fa 920 adapter->flags &= ~IGB_FLAG_HAS_MSI;
9d5c8243
AK
921 }
922
923 err = request_irq(adapter->pdev->irq, &igb_intr, IRQF_SHARED,
047e0030 924 netdev->name, adapter);
9d5c8243 925
6cb5e577 926 if (err)
9d5c8243
AK
927 dev_err(&adapter->pdev->dev, "Error %d getting interrupt\n",
928 err);
9d5c8243
AK
929
930request_done:
931 return err;
932}
933
934static void igb_free_irq(struct igb_adapter *adapter)
935{
9d5c8243
AK
936 if (adapter->msix_entries) {
937 int vector = 0, i;
938
047e0030 939 free_irq(adapter->msix_entries[vector++].vector, adapter);
9d5c8243 940
047e0030
AD
941 for (i = 0; i < adapter->num_q_vectors; i++) {
942 struct igb_q_vector *q_vector = adapter->q_vector[i];
943 free_irq(adapter->msix_entries[vector++].vector,
944 q_vector);
945 }
946 } else {
947 free_irq(adapter->pdev->irq, adapter);
9d5c8243 948 }
9d5c8243
AK
949}
950
951/**
952 * igb_irq_disable - Mask off interrupt generation on the NIC
953 * @adapter: board private structure
954 **/
955static void igb_irq_disable(struct igb_adapter *adapter)
956{
957 struct e1000_hw *hw = &adapter->hw;
958
959 if (adapter->msix_entries) {
2dfd1212
AD
960 u32 regval = rd32(E1000_EIAM);
961 wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
962 wr32(E1000_EIMC, adapter->eims_enable_mask);
963 regval = rd32(E1000_EIAC);
964 wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
9d5c8243 965 }
844290e5
PW
966
967 wr32(E1000_IAM, 0);
9d5c8243
AK
968 wr32(E1000_IMC, ~0);
969 wrfl();
970 synchronize_irq(adapter->pdev->irq);
971}
972
973/**
974 * igb_irq_enable - Enable default interrupt generation settings
975 * @adapter: board private structure
976 **/
977static void igb_irq_enable(struct igb_adapter *adapter)
978{
979 struct e1000_hw *hw = &adapter->hw;
980
981 if (adapter->msix_entries) {
2dfd1212
AD
982 u32 regval = rd32(E1000_EIAC);
983 wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
984 regval = rd32(E1000_EIAM);
985 wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
844290e5 986 wr32(E1000_EIMS, adapter->eims_enable_mask);
4ae196df
AD
987 if (adapter->vfs_allocated_count)
988 wr32(E1000_MBVFIMR, 0xFF);
989 wr32(E1000_IMS, (E1000_IMS_LSC | E1000_IMS_VMMB |
990 E1000_IMS_DOUTSYNC));
844290e5
PW
991 } else {
992 wr32(E1000_IMS, IMS_ENABLE_MASK);
993 wr32(E1000_IAM, IMS_ENABLE_MASK);
994 }
9d5c8243
AK
995}
996
997static void igb_update_mng_vlan(struct igb_adapter *adapter)
998{
999 struct net_device *netdev = adapter->netdev;
1000 u16 vid = adapter->hw.mng_cookie.vlan_id;
1001 u16 old_vid = adapter->mng_vlan_id;
1002 if (adapter->vlgrp) {
1003 if (!vlan_group_get_device(adapter->vlgrp, vid)) {
1004 if (adapter->hw.mng_cookie.status &
1005 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1006 igb_vlan_rx_add_vid(netdev, vid);
1007 adapter->mng_vlan_id = vid;
1008 } else
1009 adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1010
1011 if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1012 (vid != old_vid) &&
1013 !vlan_group_get_device(adapter->vlgrp, old_vid))
1014 igb_vlan_rx_kill_vid(netdev, old_vid);
1015 } else
1016 adapter->mng_vlan_id = vid;
1017 }
1018}
1019
1020/**
1021 * igb_release_hw_control - release control of the h/w to f/w
1022 * @adapter: address of board private structure
1023 *
1024 * igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1025 * For ASF and Pass Through versions of f/w this means that the
1026 * driver is no longer loaded.
1027 *
1028 **/
1029static void igb_release_hw_control(struct igb_adapter *adapter)
1030{
1031 struct e1000_hw *hw = &adapter->hw;
1032 u32 ctrl_ext;
1033
1034 /* Let firmware take over control of h/w */
1035 ctrl_ext = rd32(E1000_CTRL_EXT);
1036 wr32(E1000_CTRL_EXT,
1037 ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1038}
1039
1040
1041/**
1042 * igb_get_hw_control - get control of the h/w from f/w
1043 * @adapter: address of board private structure
1044 *
1045 * igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1046 * For ASF and Pass Through versions of f/w this means that
1047 * the driver is loaded.
1048 *
1049 **/
1050static void igb_get_hw_control(struct igb_adapter *adapter)
1051{
1052 struct e1000_hw *hw = &adapter->hw;
1053 u32 ctrl_ext;
1054
1055 /* Let firmware know the driver has taken over */
1056 ctrl_ext = rd32(E1000_CTRL_EXT);
1057 wr32(E1000_CTRL_EXT,
1058 ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1059}
1060
9d5c8243
AK
1061/**
1062 * igb_configure - configure the hardware for RX and TX
1063 * @adapter: private board structure
1064 **/
1065static void igb_configure(struct igb_adapter *adapter)
1066{
1067 struct net_device *netdev = adapter->netdev;
1068 int i;
1069
1070 igb_get_hw_control(adapter);
ff41f8dc 1071 igb_set_rx_mode(netdev);
9d5c8243
AK
1072
1073 igb_restore_vlan(adapter);
9d5c8243 1074
85b430b4 1075 igb_setup_tctl(adapter);
06cf2666 1076 igb_setup_mrqc(adapter);
9d5c8243 1077 igb_setup_rctl(adapter);
85b430b4
AD
1078
1079 igb_configure_tx(adapter);
9d5c8243 1080 igb_configure_rx(adapter);
662d7205
AD
1081
1082 igb_rx_fifo_flush_82575(&adapter->hw);
1083
c493ea45 1084 /* call igb_desc_unused which always leaves
9d5c8243
AK
1085 * at least 1 descriptor unused to make sure
1086 * next_to_use != next_to_clean */
1087 for (i = 0; i < adapter->num_rx_queues; i++) {
1088 struct igb_ring *ring = &adapter->rx_ring[i];
c493ea45 1089 igb_alloc_rx_buffers_adv(ring, igb_desc_unused(ring));
9d5c8243
AK
1090 }
1091
1092
1093 adapter->tx_queue_len = netdev->tx_queue_len;
1094}
1095
1096
1097/**
1098 * igb_up - Open the interface and prepare it to handle traffic
1099 * @adapter: board private structure
1100 **/
1101
1102int igb_up(struct igb_adapter *adapter)
1103{
1104 struct e1000_hw *hw = &adapter->hw;
1105 int i;
1106
1107 /* hardware has been reset, we need to reload some things */
1108 igb_configure(adapter);
1109
1110 clear_bit(__IGB_DOWN, &adapter->state);
1111
047e0030
AD
1112 for (i = 0; i < adapter->num_q_vectors; i++) {
1113 struct igb_q_vector *q_vector = adapter->q_vector[i];
1114 napi_enable(&q_vector->napi);
1115 }
844290e5 1116 if (adapter->msix_entries)
9d5c8243 1117 igb_configure_msix(adapter);
9d5c8243 1118
e1739522
AD
1119 igb_set_vmolr(hw, adapter->vfs_allocated_count);
1120
9d5c8243
AK
1121 /* Clear any pending interrupts. */
1122 rd32(E1000_ICR);
1123 igb_irq_enable(adapter);
1124
d4960307
AD
1125 /* notify VFs that reset has been completed */
1126 if (adapter->vfs_allocated_count) {
1127 u32 reg_data = rd32(E1000_CTRL_EXT);
1128 reg_data |= E1000_CTRL_EXT_PFRSTD;
1129 wr32(E1000_CTRL_EXT, reg_data);
1130 }
1131
4cb9be7a
JB
1132 netif_tx_start_all_queues(adapter->netdev);
1133
9d5c8243
AK
1134 /* Fire a link change interrupt to start the watchdog. */
1135 wr32(E1000_ICS, E1000_ICS_LSC);
1136 return 0;
1137}
1138
1139void igb_down(struct igb_adapter *adapter)
1140{
1141 struct e1000_hw *hw = &adapter->hw;
1142 struct net_device *netdev = adapter->netdev;
1143 u32 tctl, rctl;
1144 int i;
1145
1146 /* signal that we're down so the interrupt handler does not
1147 * reschedule our watchdog timer */
1148 set_bit(__IGB_DOWN, &adapter->state);
1149
1150 /* disable receives in the hardware */
1151 rctl = rd32(E1000_RCTL);
1152 wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
1153 /* flush and sleep below */
1154
fd2ea0a7 1155 netif_tx_stop_all_queues(netdev);
9d5c8243
AK
1156
1157 /* disable transmits in the hardware */
1158 tctl = rd32(E1000_TCTL);
1159 tctl &= ~E1000_TCTL_EN;
1160 wr32(E1000_TCTL, tctl);
1161 /* flush both disables and wait for them to finish */
1162 wrfl();
1163 msleep(10);
1164
047e0030
AD
1165 for (i = 0; i < adapter->num_q_vectors; i++) {
1166 struct igb_q_vector *q_vector = adapter->q_vector[i];
1167 napi_disable(&q_vector->napi);
1168 }
9d5c8243 1169
9d5c8243
AK
1170 igb_irq_disable(adapter);
1171
1172 del_timer_sync(&adapter->watchdog_timer);
1173 del_timer_sync(&adapter->phy_info_timer);
1174
1175 netdev->tx_queue_len = adapter->tx_queue_len;
1176 netif_carrier_off(netdev);
04fe6358
AD
1177
1178 /* record the stats before reset*/
1179 igb_update_stats(adapter);
1180
9d5c8243
AK
1181 adapter->link_speed = 0;
1182 adapter->link_duplex = 0;
1183
3023682e
JK
1184 if (!pci_channel_offline(adapter->pdev))
1185 igb_reset(adapter);
9d5c8243
AK
1186 igb_clean_all_tx_rings(adapter);
1187 igb_clean_all_rx_rings(adapter);
7e0e99ef
AD
1188#ifdef CONFIG_IGB_DCA
1189
1190 /* since we reset the hardware DCA settings were cleared */
1191 igb_setup_dca(adapter);
1192#endif
9d5c8243
AK
1193}
1194
1195void igb_reinit_locked(struct igb_adapter *adapter)
1196{
1197 WARN_ON(in_interrupt());
1198 while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
1199 msleep(1);
1200 igb_down(adapter);
1201 igb_up(adapter);
1202 clear_bit(__IGB_RESETTING, &adapter->state);
1203}
1204
1205void igb_reset(struct igb_adapter *adapter)
1206{
1207 struct e1000_hw *hw = &adapter->hw;
2d064c06
AD
1208 struct e1000_mac_info *mac = &hw->mac;
1209 struct e1000_fc_info *fc = &hw->fc;
9d5c8243
AK
1210 u32 pba = 0, tx_space, min_tx_space, min_rx_space;
1211 u16 hwm;
1212
1213 /* Repartition Pba for greater than 9k mtu
1214 * To take effect CTRL.RST is required.
1215 */
fa4dfae0
AD
1216 switch (mac->type) {
1217 case e1000_82576:
2d064c06 1218 pba = E1000_PBA_64K;
fa4dfae0
AD
1219 break;
1220 case e1000_82575:
1221 default:
1222 pba = E1000_PBA_34K;
1223 break;
2d064c06 1224 }
9d5c8243 1225
2d064c06
AD
1226 if ((adapter->max_frame_size > ETH_FRAME_LEN + ETH_FCS_LEN) &&
1227 (mac->type < e1000_82576)) {
9d5c8243
AK
1228 /* adjust PBA for jumbo frames */
1229 wr32(E1000_PBA, pba);
1230
1231 /* To maintain wire speed transmits, the Tx FIFO should be
1232 * large enough to accommodate two full transmit packets,
1233 * rounded up to the next 1KB and expressed in KB. Likewise,
1234 * the Rx FIFO should be large enough to accommodate at least
1235 * one full receive packet and is similarly rounded up and
1236 * expressed in KB. */
1237 pba = rd32(E1000_PBA);
1238 /* upper 16 bits has Tx packet buffer allocation size in KB */
1239 tx_space = pba >> 16;
1240 /* lower 16 bits has Rx packet buffer allocation size in KB */
1241 pba &= 0xffff;
1242 /* the tx fifo also stores 16 bytes of information about the tx
1243 * but don't include ethernet FCS because hardware appends it */
1244 min_tx_space = (adapter->max_frame_size +
85e8d004 1245 sizeof(union e1000_adv_tx_desc) -
9d5c8243
AK
1246 ETH_FCS_LEN) * 2;
1247 min_tx_space = ALIGN(min_tx_space, 1024);
1248 min_tx_space >>= 10;
1249 /* software strips receive CRC, so leave room for it */
1250 min_rx_space = adapter->max_frame_size;
1251 min_rx_space = ALIGN(min_rx_space, 1024);
1252 min_rx_space >>= 10;
1253
1254 /* If current Tx allocation is less than the min Tx FIFO size,
1255 * and the min Tx FIFO size is less than the current Rx FIFO
1256 * allocation, take space away from current Rx allocation */
1257 if (tx_space < min_tx_space &&
1258 ((min_tx_space - tx_space) < pba)) {
1259 pba = pba - (min_tx_space - tx_space);
1260
1261 /* if short on rx space, rx wins and must trump tx
1262 * adjustment */
1263 if (pba < min_rx_space)
1264 pba = min_rx_space;
1265 }
2d064c06 1266 wr32(E1000_PBA, pba);
9d5c8243 1267 }
9d5c8243
AK
1268
1269 /* flow control settings */
1270 /* The high water mark must be low enough to fit one full frame
1271 * (or the size used for early receive) above it in the Rx FIFO.
1272 * Set it to the lower of:
1273 * - 90% of the Rx FIFO size, or
1274 * - the full Rx FIFO size minus one full frame */
1275 hwm = min(((pba << 10) * 9 / 10),
2d064c06 1276 ((pba << 10) - 2 * adapter->max_frame_size));
9d5c8243 1277
2d064c06
AD
1278 if (mac->type < e1000_82576) {
1279 fc->high_water = hwm & 0xFFF8; /* 8-byte granularity */
1280 fc->low_water = fc->high_water - 8;
1281 } else {
1282 fc->high_water = hwm & 0xFFF0; /* 16-byte granularity */
1283 fc->low_water = fc->high_water - 16;
1284 }
9d5c8243
AK
1285 fc->pause_time = 0xFFFF;
1286 fc->send_xon = 1;
0cce119a 1287 fc->current_mode = fc->requested_mode;
9d5c8243 1288
4ae196df
AD
1289 /* disable receive for all VFs and wait one second */
1290 if (adapter->vfs_allocated_count) {
1291 int i;
1292 for (i = 0 ; i < adapter->vfs_allocated_count; i++)
1293 adapter->vf_data[i].clear_to_send = false;
1294
1295 /* ping all the active vfs to let them know we are going down */
1296 igb_ping_all_vfs(adapter);
1297
1298 /* disable transmits and receives */
1299 wr32(E1000_VFRE, 0);
1300 wr32(E1000_VFTE, 0);
1301 }
1302
9d5c8243
AK
1303 /* Allow time for pending master requests to run */
1304 adapter->hw.mac.ops.reset_hw(&adapter->hw);
1305 wr32(E1000_WUC, 0);
1306
1307 if (adapter->hw.mac.ops.init_hw(&adapter->hw))
1308 dev_err(&adapter->pdev->dev, "Hardware Error\n");
1309
1310 igb_update_mng_vlan(adapter);
1311
1312 /* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
1313 wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
1314
1315 igb_reset_adaptive(&adapter->hw);
f5f4cf08 1316 igb_get_phy_info(&adapter->hw);
9d5c8243
AK
1317}
1318
2e5c6922
SH
1319static const struct net_device_ops igb_netdev_ops = {
1320 .ndo_open = igb_open,
1321 .ndo_stop = igb_close,
00829823 1322 .ndo_start_xmit = igb_xmit_frame_adv,
2e5c6922 1323 .ndo_get_stats = igb_get_stats,
ff41f8dc
AD
1324 .ndo_set_rx_mode = igb_set_rx_mode,
1325 .ndo_set_multicast_list = igb_set_rx_mode,
2e5c6922
SH
1326 .ndo_set_mac_address = igb_set_mac,
1327 .ndo_change_mtu = igb_change_mtu,
1328 .ndo_do_ioctl = igb_ioctl,
1329 .ndo_tx_timeout = igb_tx_timeout,
1330 .ndo_validate_addr = eth_validate_addr,
1331 .ndo_vlan_rx_register = igb_vlan_rx_register,
1332 .ndo_vlan_rx_add_vid = igb_vlan_rx_add_vid,
1333 .ndo_vlan_rx_kill_vid = igb_vlan_rx_kill_vid,
1334#ifdef CONFIG_NET_POLL_CONTROLLER
1335 .ndo_poll_controller = igb_netpoll,
1336#endif
1337};
1338
9d5c8243
AK
1339/**
1340 * igb_probe - Device Initialization Routine
1341 * @pdev: PCI device information struct
1342 * @ent: entry in igb_pci_tbl
1343 *
1344 * Returns 0 on success, negative on failure
1345 *
1346 * igb_probe initializes an adapter identified by a pci_dev structure.
1347 * The OS initialization, configuring of the adapter private structure,
1348 * and a hardware reset occur.
1349 **/
1350static int __devinit igb_probe(struct pci_dev *pdev,
1351 const struct pci_device_id *ent)
1352{
1353 struct net_device *netdev;
1354 struct igb_adapter *adapter;
1355 struct e1000_hw *hw;
1356 const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
1357 unsigned long mmio_start, mmio_len;
2d6a5e95 1358 int err, pci_using_dac;
682337fe 1359 u16 eeprom_data = 0;
9d5c8243
AK
1360 u16 eeprom_apme_mask = IGB_EEPROM_APME;
1361 u32 part_num;
1362
aed5dec3 1363 err = pci_enable_device_mem(pdev);
9d5c8243
AK
1364 if (err)
1365 return err;
1366
1367 pci_using_dac = 0;
6a35528a 1368 err = pci_set_dma_mask(pdev, DMA_BIT_MASK(64));
9d5c8243 1369 if (!err) {
6a35528a 1370 err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
9d5c8243
AK
1371 if (!err)
1372 pci_using_dac = 1;
1373 } else {
284901a9 1374 err = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
9d5c8243 1375 if (err) {
284901a9 1376 err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
9d5c8243
AK
1377 if (err) {
1378 dev_err(&pdev->dev, "No usable DMA "
1379 "configuration, aborting\n");
1380 goto err_dma;
1381 }
1382 }
1383 }
1384
aed5dec3
AD
1385 err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
1386 IORESOURCE_MEM),
1387 igb_driver_name);
9d5c8243
AK
1388 if (err)
1389 goto err_pci_reg;
1390
19d5afd4 1391 pci_enable_pcie_error_reporting(pdev);
40a914fa 1392
9d5c8243 1393 pci_set_master(pdev);
c682fc23 1394 pci_save_state(pdev);
9d5c8243
AK
1395
1396 err = -ENOMEM;
1bfaf07b
AD
1397 netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
1398 IGB_ABS_MAX_TX_QUEUES);
9d5c8243
AK
1399 if (!netdev)
1400 goto err_alloc_etherdev;
1401
1402 SET_NETDEV_DEV(netdev, &pdev->dev);
1403
1404 pci_set_drvdata(pdev, netdev);
1405 adapter = netdev_priv(netdev);
1406 adapter->netdev = netdev;
1407 adapter->pdev = pdev;
1408 hw = &adapter->hw;
1409 hw->back = adapter;
1410 adapter->msg_enable = NETIF_MSG_DRV | NETIF_MSG_PROBE;
1411
1412 mmio_start = pci_resource_start(pdev, 0);
1413 mmio_len = pci_resource_len(pdev, 0);
1414
1415 err = -EIO;
28b0759c
AD
1416 hw->hw_addr = ioremap(mmio_start, mmio_len);
1417 if (!hw->hw_addr)
9d5c8243
AK
1418 goto err_ioremap;
1419
2e5c6922 1420 netdev->netdev_ops = &igb_netdev_ops;
9d5c8243 1421 igb_set_ethtool_ops(netdev);
9d5c8243 1422 netdev->watchdog_timeo = 5 * HZ;
9d5c8243
AK
1423
1424 strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
1425
1426 netdev->mem_start = mmio_start;
1427 netdev->mem_end = mmio_start + mmio_len;
1428
9d5c8243
AK
1429 /* PCI config space info */
1430 hw->vendor_id = pdev->vendor;
1431 hw->device_id = pdev->device;
1432 hw->revision_id = pdev->revision;
1433 hw->subsystem_vendor_id = pdev->subsystem_vendor;
1434 hw->subsystem_device_id = pdev->subsystem_device;
1435
1436 /* setup the private structure */
1437 hw->back = adapter;
1438 /* Copy the default MAC, PHY and NVM function pointers */
1439 memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
1440 memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
1441 memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
1442 /* Initialize skew-specific constants */
1443 err = ei->get_invariants(hw);
1444 if (err)
450c87c8 1445 goto err_sw_init;
9d5c8243 1446
2a3abf6d
AD
1447#ifdef CONFIG_PCI_IOV
1448 /* since iov functionality isn't critical to base device function we
1449 * can accept failure. If it fails we don't allow iov to be enabled */
1450 if (hw->mac.type == e1000_82576) {
1451 /* 82576 supports a maximum of 7 VFs in addition to the PF */
1452 unsigned int num_vfs = (max_vfs > 7) ? 7 : max_vfs;
1453 int i;
1454 unsigned char mac_addr[ETH_ALEN];
1455
9ca046d5 1456 if (num_vfs) {
2a3abf6d
AD
1457 adapter->vf_data = kcalloc(num_vfs,
1458 sizeof(struct vf_data_storage),
1459 GFP_KERNEL);
9ca046d5
AD
1460 if (!adapter->vf_data) {
1461 dev_err(&pdev->dev,
1462 "Could not allocate VF private data - "
1463 "IOV enable failed\n");
2a3abf6d 1464 } else {
9ca046d5
AD
1465 err = pci_enable_sriov(pdev, num_vfs);
1466 if (!err) {
1467 adapter->vfs_allocated_count = num_vfs;
1468 dev_info(&pdev->dev,
1469 "%d vfs allocated\n",
1470 num_vfs);
1471 for (i = 0;
1472 i < adapter->vfs_allocated_count;
1473 i++) {
1474 random_ether_addr(mac_addr);
1475 igb_set_vf_mac(adapter, i,
1476 mac_addr);
1477 }
1478 } else {
1479 kfree(adapter->vf_data);
1480 adapter->vf_data = NULL;
1481 }
2a3abf6d
AD
1482 }
1483 }
1484 }
1485
1486#endif
450c87c8 1487 /* setup the private structure */
9d5c8243
AK
1488 err = igb_sw_init(adapter);
1489 if (err)
1490 goto err_sw_init;
1491
1492 igb_get_bus_info_pcie(hw);
1493
1494 hw->phy.autoneg_wait_to_complete = false;
1495 hw->mac.adaptive_ifs = true;
1496
1497 /* Copper options */
1498 if (hw->phy.media_type == e1000_media_type_copper) {
1499 hw->phy.mdix = AUTO_ALL_MODES;
1500 hw->phy.disable_polarity_correction = false;
1501 hw->phy.ms_type = e1000_ms_hw_default;
1502 }
1503
1504 if (igb_check_reset_block(hw))
1505 dev_info(&pdev->dev,
1506 "PHY reset is blocked due to SOL/IDER session.\n");
1507
1508 netdev->features = NETIF_F_SG |
7d8eb29e 1509 NETIF_F_IP_CSUM |
9d5c8243
AK
1510 NETIF_F_HW_VLAN_TX |
1511 NETIF_F_HW_VLAN_RX |
1512 NETIF_F_HW_VLAN_FILTER;
1513
7d8eb29e 1514 netdev->features |= NETIF_F_IPV6_CSUM;
9d5c8243 1515 netdev->features |= NETIF_F_TSO;
9d5c8243 1516 netdev->features |= NETIF_F_TSO6;
48f29ffc 1517
5c0999b7 1518 netdev->features |= NETIF_F_GRO;
d3352520 1519
48f29ffc
JK
1520 netdev->vlan_features |= NETIF_F_TSO;
1521 netdev->vlan_features |= NETIF_F_TSO6;
7d8eb29e 1522 netdev->vlan_features |= NETIF_F_IP_CSUM;
cd1da503 1523 netdev->vlan_features |= NETIF_F_IPV6_CSUM;
48f29ffc
JK
1524 netdev->vlan_features |= NETIF_F_SG;
1525
9d5c8243
AK
1526 if (pci_using_dac)
1527 netdev->features |= NETIF_F_HIGHDMA;
1528
b9473560
JB
1529 if (adapter->hw.mac.type == e1000_82576)
1530 netdev->features |= NETIF_F_SCTP_CSUM;
1531
9d5c8243
AK
1532 adapter->en_mng_pt = igb_enable_mng_pass_thru(&adapter->hw);
1533
1534 /* before reading the NVM, reset the controller to put the device in a
1535 * known good starting state */
1536 hw->mac.ops.reset_hw(hw);
1537
1538 /* make sure the NVM is good */
1539 if (igb_validate_nvm_checksum(hw) < 0) {
1540 dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
1541 err = -EIO;
1542 goto err_eeprom;
1543 }
1544
1545 /* copy the MAC address out of the NVM */
1546 if (hw->mac.ops.read_mac_addr(hw))
1547 dev_err(&pdev->dev, "NVM Read Error\n");
1548
1549 memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
1550 memcpy(netdev->perm_addr, hw->mac.addr, netdev->addr_len);
1551
1552 if (!is_valid_ether_addr(netdev->perm_addr)) {
1553 dev_err(&pdev->dev, "Invalid MAC Address\n");
1554 err = -EIO;
1555 goto err_eeprom;
1556 }
1557
0e340485
AD
1558 setup_timer(&adapter->watchdog_timer, &igb_watchdog,
1559 (unsigned long) adapter);
1560 setup_timer(&adapter->phy_info_timer, &igb_update_phy_info,
1561 (unsigned long) adapter);
9d5c8243
AK
1562
1563 INIT_WORK(&adapter->reset_task, igb_reset_task);
1564 INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
1565
450c87c8 1566 /* Initialize link properties that are user-changeable */
9d5c8243
AK
1567 adapter->fc_autoneg = true;
1568 hw->mac.autoneg = true;
1569 hw->phy.autoneg_advertised = 0x2f;
1570
0cce119a
AD
1571 hw->fc.requested_mode = e1000_fc_default;
1572 hw->fc.current_mode = e1000_fc_default;
9d5c8243 1573
9d5c8243
AK
1574 igb_validate_mdi_setting(hw);
1575
9d5c8243
AK
1576 /* Initial Wake on LAN setting If APM wake is enabled in the EEPROM,
1577 * enable the ACPI Magic Packet filter
1578 */
1579
a2cf8b6c 1580 if (hw->bus.func == 0)
312c75ae 1581 hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
a2cf8b6c
AD
1582 else if (hw->bus.func == 1)
1583 hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
9d5c8243
AK
1584
1585 if (eeprom_data & eeprom_apme_mask)
1586 adapter->eeprom_wol |= E1000_WUFC_MAG;
1587
1588 /* now that we have the eeprom settings, apply the special cases where
1589 * the eeprom may be wrong or the board simply won't support wake on
1590 * lan on a particular port */
1591 switch (pdev->device) {
1592 case E1000_DEV_ID_82575GB_QUAD_COPPER:
1593 adapter->eeprom_wol = 0;
1594 break;
1595 case E1000_DEV_ID_82575EB_FIBER_SERDES:
2d064c06
AD
1596 case E1000_DEV_ID_82576_FIBER:
1597 case E1000_DEV_ID_82576_SERDES:
9d5c8243
AK
1598 /* Wake events only supported on port A for dual fiber
1599 * regardless of eeprom setting */
1600 if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
1601 adapter->eeprom_wol = 0;
1602 break;
c8ea5ea9
AD
1603 case E1000_DEV_ID_82576_QUAD_COPPER:
1604 /* if quad port adapter, disable WoL on all but port A */
1605 if (global_quad_port_a != 0)
1606 adapter->eeprom_wol = 0;
1607 else
1608 adapter->flags |= IGB_FLAG_QUAD_PORT_A;
1609 /* Reset for multiple quad port adapters */
1610 if (++global_quad_port_a == 4)
1611 global_quad_port_a = 0;
1612 break;
9d5c8243
AK
1613 }
1614
1615 /* initialize the wol settings based on the eeprom settings */
1616 adapter->wol = adapter->eeprom_wol;
e1b86d84 1617 device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
9d5c8243
AK
1618
1619 /* reset the hardware with the new settings */
1620 igb_reset(adapter);
1621
1622 /* let the f/w know that the h/w is now under the control of the
1623 * driver. */
1624 igb_get_hw_control(adapter);
1625
9d5c8243
AK
1626 strcpy(netdev->name, "eth%d");
1627 err = register_netdev(netdev);
1628 if (err)
1629 goto err_register;
1630
b168dfc5
JB
1631 /* carrier off reporting is important to ethtool even BEFORE open */
1632 netif_carrier_off(netdev);
1633
421e02f0 1634#ifdef CONFIG_IGB_DCA
bbd98fe4 1635 if (dca_add_requester(&pdev->dev) == 0) {
7dfc16fa 1636 adapter->flags |= IGB_FLAG_DCA_ENABLED;
fe4506b6 1637 dev_info(&pdev->dev, "DCA enabled\n");
fe4506b6
JC
1638 igb_setup_dca(adapter);
1639 }
fe4506b6 1640
38c845c7 1641#endif
38c845c7 1642
c5b9bd5e
AD
1643 switch (hw->mac.type) {
1644 case e1000_82576:
1645 /*
1646 * Initialize hardware timer: we keep it running just in case
1647 * that some program needs it later on.
1648 */
1649 memset(&adapter->cycles, 0, sizeof(adapter->cycles));
1650 adapter->cycles.read = igb_read_clock;
1651 adapter->cycles.mask = CLOCKSOURCE_MASK(64);
1652 adapter->cycles.mult = 1;
1653 /**
1654 * Scale the NIC clock cycle by a large factor so that
1655 * relatively small clock corrections can be added or
1656 * substracted at each clock tick. The drawbacks of a large
1657 * factor are a) that the clock register overflows more quickly
1658 * (not such a big deal) and b) that the increment per tick has
1659 * to fit into 24 bits. As a result we need to use a shift of
1660 * 19 so we can fit a value of 16 into the TIMINCA register.
1661 */
1662 adapter->cycles.shift = IGB_82576_TSYNC_SHIFT;
1663 wr32(E1000_TIMINCA,
1664 (1 << E1000_TIMINCA_16NS_SHIFT) |
1665 (16 << IGB_82576_TSYNC_SHIFT));
1666
1667 /* Set registers so that rollover occurs soon to test this. */
1668 wr32(E1000_SYSTIML, 0x00000000);
1669 wr32(E1000_SYSTIMH, 0xFF800000);
1670 wrfl();
1671
1672 timecounter_init(&adapter->clock,
1673 &adapter->cycles,
1674 ktime_to_ns(ktime_get_real()));
1675 /*
1676 * Synchronize our NIC clock against system wall clock. NIC
1677 * time stamp reading requires ~3us per sample, each sample
1678 * was pretty stable even under load => only require 10
1679 * samples for each offset comparison.
1680 */
1681 memset(&adapter->compare, 0, sizeof(adapter->compare));
1682 adapter->compare.source = &adapter->clock;
1683 adapter->compare.target = ktime_get_real;
1684 adapter->compare.num_samples = 10;
1685 timecompare_update(&adapter->compare, 0);
1686 break;
1687 case e1000_82575:
1688 /* 82575 does not support timesync */
1689 default:
1690 break;
38c845c7 1691 }
38c845c7 1692
9d5c8243
AK
1693 dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
1694 /* print bus type/speed/width info */
7c510e4b 1695 dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
9d5c8243
AK
1696 netdev->name,
1697 ((hw->bus.speed == e1000_bus_speed_2500)
1698 ? "2.5Gb/s" : "unknown"),
59c3de89
AD
1699 ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
1700 (hw->bus.width == e1000_bus_width_pcie_x2) ? "Width x2" :
1701 (hw->bus.width == e1000_bus_width_pcie_x1) ? "Width x1" :
1702 "unknown"),
7c510e4b 1703 netdev->dev_addr);
9d5c8243
AK
1704
1705 igb_read_part_num(hw, &part_num);
1706 dev_info(&pdev->dev, "%s: PBA No: %06x-%03x\n", netdev->name,
1707 (part_num >> 8), (part_num & 0xff));
1708
1709 dev_info(&pdev->dev,
1710 "Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
1711 adapter->msix_entries ? "MSI-X" :
7dfc16fa 1712 (adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
9d5c8243
AK
1713 adapter->num_rx_queues, adapter->num_tx_queues);
1714
9d5c8243
AK
1715 return 0;
1716
1717err_register:
1718 igb_release_hw_control(adapter);
1719err_eeprom:
1720 if (!igb_check_reset_block(hw))
f5f4cf08 1721 igb_reset_phy(hw);
9d5c8243
AK
1722
1723 if (hw->flash_address)
1724 iounmap(hw->flash_address);
9d5c8243 1725err_sw_init:
047e0030 1726 igb_clear_interrupt_scheme(adapter);
9d5c8243
AK
1727 iounmap(hw->hw_addr);
1728err_ioremap:
1729 free_netdev(netdev);
1730err_alloc_etherdev:
aed5dec3
AD
1731 pci_release_selected_regions(pdev, pci_select_bars(pdev,
1732 IORESOURCE_MEM));
9d5c8243
AK
1733err_pci_reg:
1734err_dma:
1735 pci_disable_device(pdev);
1736 return err;
1737}
1738
1739/**
1740 * igb_remove - Device Removal Routine
1741 * @pdev: PCI device information struct
1742 *
1743 * igb_remove is called by the PCI subsystem to alert the driver
1744 * that it should release a PCI device. The could be caused by a
1745 * Hot-Plug event, or because the driver is going to be removed from
1746 * memory.
1747 **/
1748static void __devexit igb_remove(struct pci_dev *pdev)
1749{
1750 struct net_device *netdev = pci_get_drvdata(pdev);
1751 struct igb_adapter *adapter = netdev_priv(netdev);
fe4506b6 1752 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK
1753
1754 /* flush_scheduled work may reschedule our watchdog task, so
1755 * explicitly disable watchdog tasks from being rescheduled */
1756 set_bit(__IGB_DOWN, &adapter->state);
1757 del_timer_sync(&adapter->watchdog_timer);
1758 del_timer_sync(&adapter->phy_info_timer);
1759
1760 flush_scheduled_work();
1761
421e02f0 1762#ifdef CONFIG_IGB_DCA
7dfc16fa 1763 if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
fe4506b6
JC
1764 dev_info(&pdev->dev, "DCA disabled\n");
1765 dca_remove_requester(&pdev->dev);
7dfc16fa 1766 adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
cbd347ad 1767 wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
fe4506b6
JC
1768 }
1769#endif
1770
9d5c8243
AK
1771 /* Release control of h/w to f/w. If f/w is AMT enabled, this
1772 * would have already happened in close and is redundant. */
1773 igb_release_hw_control(adapter);
1774
1775 unregister_netdev(netdev);
1776
f5f4cf08
AD
1777 if (!igb_check_reset_block(&adapter->hw))
1778 igb_reset_phy(&adapter->hw);
9d5c8243 1779
047e0030 1780 igb_clear_interrupt_scheme(adapter);
9d5c8243 1781
37680117
AD
1782#ifdef CONFIG_PCI_IOV
1783 /* reclaim resources allocated to VFs */
1784 if (adapter->vf_data) {
1785 /* disable iov and allow time for transactions to clear */
1786 pci_disable_sriov(pdev);
1787 msleep(500);
1788
1789 kfree(adapter->vf_data);
1790 adapter->vf_data = NULL;
1791 wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1792 msleep(100);
1793 dev_info(&pdev->dev, "IOV Disabled\n");
1794 }
1795#endif
28b0759c
AD
1796 iounmap(hw->hw_addr);
1797 if (hw->flash_address)
1798 iounmap(hw->flash_address);
aed5dec3
AD
1799 pci_release_selected_regions(pdev, pci_select_bars(pdev,
1800 IORESOURCE_MEM));
9d5c8243
AK
1801
1802 free_netdev(netdev);
1803
19d5afd4 1804 pci_disable_pcie_error_reporting(pdev);
40a914fa 1805
9d5c8243
AK
1806 pci_disable_device(pdev);
1807}
1808
1809/**
1810 * igb_sw_init - Initialize general software structures (struct igb_adapter)
1811 * @adapter: board private structure to initialize
1812 *
1813 * igb_sw_init initializes the Adapter private data structure.
1814 * Fields are initialized based on PCI device information and
1815 * OS network device settings (MTU size).
1816 **/
1817static int __devinit igb_sw_init(struct igb_adapter *adapter)
1818{
1819 struct e1000_hw *hw = &adapter->hw;
1820 struct net_device *netdev = adapter->netdev;
1821 struct pci_dev *pdev = adapter->pdev;
1822
1823 pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
1824
68fd9910
AD
1825 adapter->tx_ring_count = IGB_DEFAULT_TXD;
1826 adapter->rx_ring_count = IGB_DEFAULT_RXD;
4fc82adf
AD
1827 adapter->rx_itr_setting = IGB_DEFAULT_ITR;
1828 adapter->tx_itr_setting = IGB_DEFAULT_ITR;
1829
9d5c8243
AK
1830 adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
1831 adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
1832
661086df
PWJ
1833 /* This call may decrease the number of queues depending on
1834 * interrupt mode. */
047e0030 1835 if (igb_init_interrupt_scheme(adapter)) {
9d5c8243
AK
1836 dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
1837 return -ENOMEM;
1838 }
1839
1840 /* Explicitly disable IRQ since the NIC can be in any state. */
1841 igb_irq_disable(adapter);
1842
1843 set_bit(__IGB_DOWN, &adapter->state);
1844 return 0;
1845}
1846
1847/**
1848 * igb_open - Called when a network interface is made active
1849 * @netdev: network interface device structure
1850 *
1851 * Returns 0 on success, negative value on failure
1852 *
1853 * The open entry point is called when a network interface is made
1854 * active by the system (IFF_UP). At this point all resources needed
1855 * for transmit and receive operations are allocated, the interrupt
1856 * handler is registered with the OS, the watchdog timer is started,
1857 * and the stack is notified that the interface is ready.
1858 **/
1859static int igb_open(struct net_device *netdev)
1860{
1861 struct igb_adapter *adapter = netdev_priv(netdev);
1862 struct e1000_hw *hw = &adapter->hw;
1863 int err;
1864 int i;
1865
1866 /* disallow open during test */
1867 if (test_bit(__IGB_TESTING, &adapter->state))
1868 return -EBUSY;
1869
b168dfc5
JB
1870 netif_carrier_off(netdev);
1871
9d5c8243
AK
1872 /* allocate transmit descriptors */
1873 err = igb_setup_all_tx_resources(adapter);
1874 if (err)
1875 goto err_setup_tx;
1876
1877 /* allocate receive descriptors */
1878 err = igb_setup_all_rx_resources(adapter);
1879 if (err)
1880 goto err_setup_rx;
1881
1882 /* e1000_power_up_phy(adapter); */
1883
1884 adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1885 if ((adapter->hw.mng_cookie.status &
1886 E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
1887 igb_update_mng_vlan(adapter);
1888
1889 /* before we allocate an interrupt, we must be ready to handle it.
1890 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
1891 * as soon as we call pci_request_irq, so we have to setup our
1892 * clean_rx handler before we do so. */
1893 igb_configure(adapter);
1894
e1739522
AD
1895 igb_set_vmolr(hw, adapter->vfs_allocated_count);
1896
9d5c8243
AK
1897 err = igb_request_irq(adapter);
1898 if (err)
1899 goto err_req_irq;
1900
1901 /* From here on the code is the same as igb_up() */
1902 clear_bit(__IGB_DOWN, &adapter->state);
1903
047e0030
AD
1904 for (i = 0; i < adapter->num_q_vectors; i++) {
1905 struct igb_q_vector *q_vector = adapter->q_vector[i];
1906 napi_enable(&q_vector->napi);
1907 }
9d5c8243
AK
1908
1909 /* Clear any pending interrupts. */
1910 rd32(E1000_ICR);
844290e5
PW
1911
1912 igb_irq_enable(adapter);
1913
d4960307
AD
1914 /* notify VFs that reset has been completed */
1915 if (adapter->vfs_allocated_count) {
1916 u32 reg_data = rd32(E1000_CTRL_EXT);
1917 reg_data |= E1000_CTRL_EXT_PFRSTD;
1918 wr32(E1000_CTRL_EXT, reg_data);
1919 }
1920
d55b53ff
JK
1921 netif_tx_start_all_queues(netdev);
1922
9d5c8243
AK
1923 /* Fire a link status change interrupt to start the watchdog. */
1924 wr32(E1000_ICS, E1000_ICS_LSC);
1925
1926 return 0;
1927
1928err_req_irq:
1929 igb_release_hw_control(adapter);
1930 /* e1000_power_down_phy(adapter); */
1931 igb_free_all_rx_resources(adapter);
1932err_setup_rx:
1933 igb_free_all_tx_resources(adapter);
1934err_setup_tx:
1935 igb_reset(adapter);
1936
1937 return err;
1938}
1939
1940/**
1941 * igb_close - Disables a network interface
1942 * @netdev: network interface device structure
1943 *
1944 * Returns 0, this is not allowed to fail
1945 *
1946 * The close entry point is called when an interface is de-activated
1947 * by the OS. The hardware is still under the driver's control, but
1948 * needs to be disabled. A global MAC reset is issued to stop the
1949 * hardware, and all transmit and receive resources are freed.
1950 **/
1951static int igb_close(struct net_device *netdev)
1952{
1953 struct igb_adapter *adapter = netdev_priv(netdev);
1954
1955 WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
1956 igb_down(adapter);
1957
1958 igb_free_irq(adapter);
1959
1960 igb_free_all_tx_resources(adapter);
1961 igb_free_all_rx_resources(adapter);
1962
1963 /* kill manageability vlan ID if supported, but not if a vlan with
1964 * the same ID is registered on the host OS (let 8021q kill it) */
1965 if ((adapter->hw.mng_cookie.status &
1966 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
1967 !(adapter->vlgrp &&
1968 vlan_group_get_device(adapter->vlgrp, adapter->mng_vlan_id)))
1969 igb_vlan_rx_kill_vid(netdev, adapter->mng_vlan_id);
1970
1971 return 0;
1972}
1973
1974/**
1975 * igb_setup_tx_resources - allocate Tx resources (Descriptors)
9d5c8243
AK
1976 * @tx_ring: tx descriptor ring (for a specific queue) to setup
1977 *
1978 * Return 0 on success, negative on failure
1979 **/
80785298 1980int igb_setup_tx_resources(struct igb_ring *tx_ring)
9d5c8243 1981{
80785298 1982 struct pci_dev *pdev = tx_ring->pdev;
9d5c8243
AK
1983 int size;
1984
1985 size = sizeof(struct igb_buffer) * tx_ring->count;
1986 tx_ring->buffer_info = vmalloc(size);
1987 if (!tx_ring->buffer_info)
1988 goto err;
1989 memset(tx_ring->buffer_info, 0, size);
1990
1991 /* round up to nearest 4K */
85e8d004 1992 tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
9d5c8243
AK
1993 tx_ring->size = ALIGN(tx_ring->size, 4096);
1994
1995 tx_ring->desc = pci_alloc_consistent(pdev, tx_ring->size,
1996 &tx_ring->dma);
1997
1998 if (!tx_ring->desc)
1999 goto err;
2000
9d5c8243
AK
2001 tx_ring->next_to_use = 0;
2002 tx_ring->next_to_clean = 0;
9d5c8243
AK
2003 return 0;
2004
2005err:
2006 vfree(tx_ring->buffer_info);
047e0030 2007 dev_err(&pdev->dev,
9d5c8243
AK
2008 "Unable to allocate memory for the transmit descriptor ring\n");
2009 return -ENOMEM;
2010}
2011
2012/**
2013 * igb_setup_all_tx_resources - wrapper to allocate Tx resources
2014 * (Descriptors) for all queues
2015 * @adapter: board private structure
2016 *
2017 * Return 0 on success, negative on failure
2018 **/
2019static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
2020{
2021 int i, err = 0;
661086df 2022 int r_idx;
9d5c8243
AK
2023
2024 for (i = 0; i < adapter->num_tx_queues; i++) {
80785298 2025 err = igb_setup_tx_resources(&adapter->tx_ring[i]);
9d5c8243
AK
2026 if (err) {
2027 dev_err(&adapter->pdev->dev,
2028 "Allocation for Tx Queue %u failed\n", i);
2029 for (i--; i >= 0; i--)
3b644cf6 2030 igb_free_tx_resources(&adapter->tx_ring[i]);
9d5c8243
AK
2031 break;
2032 }
2033 }
2034
661086df
PWJ
2035 for (i = 0; i < IGB_MAX_TX_QUEUES; i++) {
2036 r_idx = i % adapter->num_tx_queues;
2037 adapter->multi_tx_table[i] = &adapter->tx_ring[r_idx];
eebbbdba 2038 }
9d5c8243
AK
2039 return err;
2040}
2041
2042/**
85b430b4
AD
2043 * igb_setup_tctl - configure the transmit control registers
2044 * @adapter: Board private structure
9d5c8243 2045 **/
d7ee5b3a 2046void igb_setup_tctl(struct igb_adapter *adapter)
9d5c8243 2047{
9d5c8243
AK
2048 struct e1000_hw *hw = &adapter->hw;
2049 u32 tctl;
9d5c8243 2050
85b430b4
AD
2051 /* disable queue 0 which is enabled by default on 82575 and 82576 */
2052 wr32(E1000_TXDCTL(0), 0);
9d5c8243
AK
2053
2054 /* Program the Transmit Control Register */
9d5c8243
AK
2055 tctl = rd32(E1000_TCTL);
2056 tctl &= ~E1000_TCTL_CT;
2057 tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2058 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2059
2060 igb_config_collision_dist(hw);
2061
9d5c8243
AK
2062 /* Enable transmits */
2063 tctl |= E1000_TCTL_EN;
2064
2065 wr32(E1000_TCTL, tctl);
2066}
2067
85b430b4
AD
2068/**
2069 * igb_configure_tx_ring - Configure transmit ring after Reset
2070 * @adapter: board private structure
2071 * @ring: tx ring to configure
2072 *
2073 * Configure a transmit ring after a reset.
2074 **/
d7ee5b3a
AD
2075void igb_configure_tx_ring(struct igb_adapter *adapter,
2076 struct igb_ring *ring)
85b430b4
AD
2077{
2078 struct e1000_hw *hw = &adapter->hw;
2079 u32 txdctl;
2080 u64 tdba = ring->dma;
2081 int reg_idx = ring->reg_idx;
2082
2083 /* disable the queue */
2084 txdctl = rd32(E1000_TXDCTL(reg_idx));
2085 wr32(E1000_TXDCTL(reg_idx),
2086 txdctl & ~E1000_TXDCTL_QUEUE_ENABLE);
2087 wrfl();
2088 mdelay(10);
2089
2090 wr32(E1000_TDLEN(reg_idx),
2091 ring->count * sizeof(union e1000_adv_tx_desc));
2092 wr32(E1000_TDBAL(reg_idx),
2093 tdba & 0x00000000ffffffffULL);
2094 wr32(E1000_TDBAH(reg_idx), tdba >> 32);
2095
fce99e34
AD
2096 ring->head = hw->hw_addr + E1000_TDH(reg_idx);
2097 ring->tail = hw->hw_addr + E1000_TDT(reg_idx);
2098 writel(0, ring->head);
2099 writel(0, ring->tail);
85b430b4
AD
2100
2101 txdctl |= IGB_TX_PTHRESH;
2102 txdctl |= IGB_TX_HTHRESH << 8;
2103 txdctl |= IGB_TX_WTHRESH << 16;
2104
2105 txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
2106 wr32(E1000_TXDCTL(reg_idx), txdctl);
2107}
2108
2109/**
2110 * igb_configure_tx - Configure transmit Unit after Reset
2111 * @adapter: board private structure
2112 *
2113 * Configure the Tx unit of the MAC after a reset.
2114 **/
2115static void igb_configure_tx(struct igb_adapter *adapter)
2116{
2117 int i;
2118
2119 for (i = 0; i < adapter->num_tx_queues; i++)
2120 igb_configure_tx_ring(adapter, &adapter->tx_ring[i]);
85b430b4
AD
2121}
2122
9d5c8243
AK
2123/**
2124 * igb_setup_rx_resources - allocate Rx resources (Descriptors)
9d5c8243
AK
2125 * @rx_ring: rx descriptor ring (for a specific queue) to setup
2126 *
2127 * Returns 0 on success, negative on failure
2128 **/
80785298 2129int igb_setup_rx_resources(struct igb_ring *rx_ring)
9d5c8243 2130{
80785298 2131 struct pci_dev *pdev = rx_ring->pdev;
9d5c8243
AK
2132 int size, desc_len;
2133
2134 size = sizeof(struct igb_buffer) * rx_ring->count;
2135 rx_ring->buffer_info = vmalloc(size);
2136 if (!rx_ring->buffer_info)
2137 goto err;
2138 memset(rx_ring->buffer_info, 0, size);
2139
2140 desc_len = sizeof(union e1000_adv_rx_desc);
2141
2142 /* Round up to nearest 4K */
2143 rx_ring->size = rx_ring->count * desc_len;
2144 rx_ring->size = ALIGN(rx_ring->size, 4096);
2145
2146 rx_ring->desc = pci_alloc_consistent(pdev, rx_ring->size,
2147 &rx_ring->dma);
2148
2149 if (!rx_ring->desc)
2150 goto err;
2151
2152 rx_ring->next_to_clean = 0;
2153 rx_ring->next_to_use = 0;
9d5c8243 2154
9d5c8243
AK
2155 return 0;
2156
2157err:
2158 vfree(rx_ring->buffer_info);
80785298 2159 dev_err(&pdev->dev, "Unable to allocate memory for "
9d5c8243
AK
2160 "the receive descriptor ring\n");
2161 return -ENOMEM;
2162}
2163
2164/**
2165 * igb_setup_all_rx_resources - wrapper to allocate Rx resources
2166 * (Descriptors) for all queues
2167 * @adapter: board private structure
2168 *
2169 * Return 0 on success, negative on failure
2170 **/
2171static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
2172{
2173 int i, err = 0;
2174
2175 for (i = 0; i < adapter->num_rx_queues; i++) {
80785298 2176 err = igb_setup_rx_resources(&adapter->rx_ring[i]);
9d5c8243
AK
2177 if (err) {
2178 dev_err(&adapter->pdev->dev,
2179 "Allocation for Rx Queue %u failed\n", i);
2180 for (i--; i >= 0; i--)
3b644cf6 2181 igb_free_rx_resources(&adapter->rx_ring[i]);
9d5c8243
AK
2182 break;
2183 }
2184 }
2185
2186 return err;
2187}
2188
06cf2666
AD
2189/**
2190 * igb_setup_mrqc - configure the multiple receive queue control registers
2191 * @adapter: Board private structure
2192 **/
2193static void igb_setup_mrqc(struct igb_adapter *adapter)
2194{
2195 struct e1000_hw *hw = &adapter->hw;
2196 u32 mrqc, rxcsum;
2197 u32 j, num_rx_queues, shift = 0, shift2 = 0;
2198 union e1000_reta {
2199 u32 dword;
2200 u8 bytes[4];
2201 } reta;
2202 static const u8 rsshash[40] = {
2203 0x6d, 0x5a, 0x56, 0xda, 0x25, 0x5b, 0x0e, 0xc2, 0x41, 0x67,
2204 0x25, 0x3d, 0x43, 0xa3, 0x8f, 0xb0, 0xd0, 0xca, 0x2b, 0xcb,
2205 0xae, 0x7b, 0x30, 0xb4, 0x77, 0xcb, 0x2d, 0xa3, 0x80, 0x30,
2206 0xf2, 0x0c, 0x6a, 0x42, 0xb7, 0x3b, 0xbe, 0xac, 0x01, 0xfa };
2207
2208 /* Fill out hash function seeds */
2209 for (j = 0; j < 10; j++) {
2210 u32 rsskey = rsshash[(j * 4)];
2211 rsskey |= rsshash[(j * 4) + 1] << 8;
2212 rsskey |= rsshash[(j * 4) + 2] << 16;
2213 rsskey |= rsshash[(j * 4) + 3] << 24;
2214 array_wr32(E1000_RSSRK(0), j, rsskey);
2215 }
2216
2217 num_rx_queues = adapter->num_rx_queues;
2218
2219 if (adapter->vfs_allocated_count) {
2220 /* 82575 and 82576 supports 2 RSS queues for VMDq */
2221 switch (hw->mac.type) {
2222 case e1000_82576:
2223 shift = 3;
2224 num_rx_queues = 2;
2225 break;
2226 case e1000_82575:
2227 shift = 2;
2228 shift2 = 6;
2229 default:
2230 break;
2231 }
2232 } else {
2233 if (hw->mac.type == e1000_82575)
2234 shift = 6;
2235 }
2236
2237 for (j = 0; j < (32 * 4); j++) {
2238 reta.bytes[j & 3] = (j % num_rx_queues) << shift;
2239 if (shift2)
2240 reta.bytes[j & 3] |= num_rx_queues << shift2;
2241 if ((j & 3) == 3)
2242 wr32(E1000_RETA(j >> 2), reta.dword);
2243 }
2244
2245 /*
2246 * Disable raw packet checksumming so that RSS hash is placed in
2247 * descriptor on writeback. No need to enable TCP/UDP/IP checksum
2248 * offloads as they are enabled by default
2249 */
2250 rxcsum = rd32(E1000_RXCSUM);
2251 rxcsum |= E1000_RXCSUM_PCSD;
2252
2253 if (adapter->hw.mac.type >= e1000_82576)
2254 /* Enable Receive Checksum Offload for SCTP */
2255 rxcsum |= E1000_RXCSUM_CRCOFL;
2256
2257 /* Don't need to set TUOFL or IPOFL, they default to 1 */
2258 wr32(E1000_RXCSUM, rxcsum);
2259
2260 /* If VMDq is enabled then we set the appropriate mode for that, else
2261 * we default to RSS so that an RSS hash is calculated per packet even
2262 * if we are only using one queue */
2263 if (adapter->vfs_allocated_count) {
2264 if (hw->mac.type > e1000_82575) {
2265 /* Set the default pool for the PF's first queue */
2266 u32 vtctl = rd32(E1000_VT_CTL);
2267 vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
2268 E1000_VT_CTL_DISABLE_DEF_POOL);
2269 vtctl |= adapter->vfs_allocated_count <<
2270 E1000_VT_CTL_DEFAULT_POOL_SHIFT;
2271 wr32(E1000_VT_CTL, vtctl);
2272 }
2273 if (adapter->num_rx_queues > 1)
2274 mrqc = E1000_MRQC_ENABLE_VMDQ_RSS_2Q;
2275 else
2276 mrqc = E1000_MRQC_ENABLE_VMDQ;
2277 } else {
2278 mrqc = E1000_MRQC_ENABLE_RSS_4Q;
2279 }
2280 igb_vmm_control(adapter);
2281
2282 mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2283 E1000_MRQC_RSS_FIELD_IPV4_TCP);
2284 mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2285 E1000_MRQC_RSS_FIELD_IPV6_TCP);
2286 mrqc |= (E1000_MRQC_RSS_FIELD_IPV4_UDP |
2287 E1000_MRQC_RSS_FIELD_IPV6_UDP);
2288 mrqc |= (E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2289 E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2290
2291 wr32(E1000_MRQC, mrqc);
2292}
2293
9d5c8243
AK
2294/**
2295 * igb_setup_rctl - configure the receive control registers
2296 * @adapter: Board private structure
2297 **/
d7ee5b3a 2298void igb_setup_rctl(struct igb_adapter *adapter)
9d5c8243
AK
2299{
2300 struct e1000_hw *hw = &adapter->hw;
2301 u32 rctl;
9d5c8243
AK
2302
2303 rctl = rd32(E1000_RCTL);
2304
2305 rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
69d728ba 2306 rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
9d5c8243 2307
69d728ba 2308 rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
28b0759c 2309 (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
9d5c8243 2310
87cb7e8c
AK
2311 /*
2312 * enable stripping of CRC. It's unlikely this will break BMC
2313 * redirection as it did with e1000. Newer features require
2314 * that the HW strips the CRC.
73cd78f1 2315 */
87cb7e8c 2316 rctl |= E1000_RCTL_SECRC;
9d5c8243 2317
9b07f3d3 2318 /*
ec54d7d6 2319 * disable store bad packets and clear size bits.
9b07f3d3 2320 */
ec54d7d6 2321 rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
9d5c8243 2322
6ec43fe6
AD
2323 /* enable LPE to prevent packets larger than max_frame_size */
2324 rctl |= E1000_RCTL_LPE;
9d5c8243 2325
952f72a8
AD
2326 /* disable queue 0 to prevent tail write w/o re-config */
2327 wr32(E1000_RXDCTL(0), 0);
9d5c8243 2328
e1739522
AD
2329 /* Attention!!! For SR-IOV PF driver operations you must enable
2330 * queue drop for all VF and PF queues to prevent head of line blocking
2331 * if an un-trusted VF does not provide descriptors to hardware.
2332 */
2333 if (adapter->vfs_allocated_count) {
2334 u32 vmolr;
2335
e1739522
AD
2336 /* set all queue drop enable bits */
2337 wr32(E1000_QDE, ALL_QUEUES);
e1739522 2338
77a22941 2339 vmolr = rd32(E1000_VMOLR(adapter->vfs_allocated_count));
e1739522
AD
2340 if (rctl & E1000_RCTL_LPE)
2341 vmolr |= E1000_VMOLR_LPE;
77a22941 2342 if (adapter->num_rx_queues > 1)
e1739522 2343 vmolr |= E1000_VMOLR_RSSE;
77a22941 2344 wr32(E1000_VMOLR(adapter->vfs_allocated_count), vmolr);
e1739522
AD
2345 }
2346
9d5c8243
AK
2347 wr32(E1000_RCTL, rctl);
2348}
2349
e1739522
AD
2350/**
2351 * igb_rlpml_set - set maximum receive packet size
2352 * @adapter: board private structure
2353 *
2354 * Configure maximum receivable packet size.
2355 **/
2356static void igb_rlpml_set(struct igb_adapter *adapter)
2357{
2358 u32 max_frame_size = adapter->max_frame_size;
2359 struct e1000_hw *hw = &adapter->hw;
2360 u16 pf_id = adapter->vfs_allocated_count;
2361
2362 if (adapter->vlgrp)
2363 max_frame_size += VLAN_TAG_SIZE;
2364
2365 /* if vfs are enabled we set RLPML to the largest possible request
2366 * size and set the VMOLR RLPML to the size we need */
2367 if (pf_id) {
2368 igb_set_vf_rlpml(adapter, max_frame_size, pf_id);
2369 max_frame_size = MAX_STD_JUMBO_FRAME_SIZE + VLAN_TAG_SIZE;
2370 }
2371
2372 wr32(E1000_RLPML, max_frame_size);
2373}
2374
85b430b4
AD
2375/**
2376 * igb_configure_rx_ring - Configure a receive ring after Reset
2377 * @adapter: board private structure
2378 * @ring: receive ring to be configured
2379 *
2380 * Configure the Rx unit of the MAC after a reset.
2381 **/
d7ee5b3a
AD
2382void igb_configure_rx_ring(struct igb_adapter *adapter,
2383 struct igb_ring *ring)
85b430b4
AD
2384{
2385 struct e1000_hw *hw = &adapter->hw;
2386 u64 rdba = ring->dma;
2387 int reg_idx = ring->reg_idx;
952f72a8 2388 u32 srrctl, rxdctl;
85b430b4
AD
2389
2390 /* disable the queue */
2391 rxdctl = rd32(E1000_RXDCTL(reg_idx));
2392 wr32(E1000_RXDCTL(reg_idx),
2393 rxdctl & ~E1000_RXDCTL_QUEUE_ENABLE);
2394
2395 /* Set DMA base address registers */
2396 wr32(E1000_RDBAL(reg_idx),
2397 rdba & 0x00000000ffffffffULL);
2398 wr32(E1000_RDBAH(reg_idx), rdba >> 32);
2399 wr32(E1000_RDLEN(reg_idx),
2400 ring->count * sizeof(union e1000_adv_rx_desc));
2401
2402 /* initialize head and tail */
fce99e34
AD
2403 ring->head = hw->hw_addr + E1000_RDH(reg_idx);
2404 ring->tail = hw->hw_addr + E1000_RDT(reg_idx);
2405 writel(0, ring->head);
2406 writel(0, ring->tail);
85b430b4 2407
952f72a8 2408 /* set descriptor configuration */
4c844851
AD
2409 if (ring->rx_buffer_len < IGB_RXBUFFER_1024) {
2410 srrctl = ALIGN(ring->rx_buffer_len, 64) <<
952f72a8
AD
2411 E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
2412#if (PAGE_SIZE / 2) > IGB_RXBUFFER_16384
2413 srrctl |= IGB_RXBUFFER_16384 >>
2414 E1000_SRRCTL_BSIZEPKT_SHIFT;
2415#else
2416 srrctl |= (PAGE_SIZE / 2) >>
2417 E1000_SRRCTL_BSIZEPKT_SHIFT;
2418#endif
2419 srrctl |= E1000_SRRCTL_DESCTYPE_HDR_SPLIT_ALWAYS;
2420 } else {
4c844851 2421 srrctl = ALIGN(ring->rx_buffer_len, 1024) >>
952f72a8
AD
2422 E1000_SRRCTL_BSIZEPKT_SHIFT;
2423 srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
2424 }
2425
2426 wr32(E1000_SRRCTL(reg_idx), srrctl);
2427
85b430b4
AD
2428 /* enable receive descriptor fetching */
2429 rxdctl = rd32(E1000_RXDCTL(reg_idx));
2430 rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
2431 rxdctl &= 0xFFF00000;
2432 rxdctl |= IGB_RX_PTHRESH;
2433 rxdctl |= IGB_RX_HTHRESH << 8;
2434 rxdctl |= IGB_RX_WTHRESH << 16;
2435 wr32(E1000_RXDCTL(reg_idx), rxdctl);
2436}
2437
9d5c8243
AK
2438/**
2439 * igb_configure_rx - Configure receive Unit after Reset
2440 * @adapter: board private structure
2441 *
2442 * Configure the Rx unit of the MAC after a reset.
2443 **/
2444static void igb_configure_rx(struct igb_adapter *adapter)
2445{
9107584e 2446 int i;
9d5c8243 2447
68d480c4
AD
2448 /* set UTA to appropriate mode */
2449 igb_set_uta(adapter);
2450
26ad9178
AD
2451 /* set the correct pool for the PF default MAC address in entry 0 */
2452 igb_rar_set_qsel(adapter, adapter->hw.mac.addr, 0,
2453 adapter->vfs_allocated_count);
2454
06cf2666
AD
2455 /* Setup the HW Rx Head and Tail Descriptor Pointers and
2456 * the Base and Length of the Rx Descriptor Ring */
2457 for (i = 0; i < adapter->num_rx_queues; i++)
2458 igb_configure_rx_ring(adapter, &adapter->rx_ring[i]);
9d5c8243
AK
2459}
2460
2461/**
2462 * igb_free_tx_resources - Free Tx Resources per Queue
9d5c8243
AK
2463 * @tx_ring: Tx descriptor ring for a specific queue
2464 *
2465 * Free all transmit software resources
2466 **/
68fd9910 2467void igb_free_tx_resources(struct igb_ring *tx_ring)
9d5c8243 2468{
3b644cf6 2469 igb_clean_tx_ring(tx_ring);
9d5c8243
AK
2470
2471 vfree(tx_ring->buffer_info);
2472 tx_ring->buffer_info = NULL;
2473
80785298
AD
2474 pci_free_consistent(tx_ring->pdev, tx_ring->size,
2475 tx_ring->desc, tx_ring->dma);
9d5c8243
AK
2476
2477 tx_ring->desc = NULL;
2478}
2479
2480/**
2481 * igb_free_all_tx_resources - Free Tx Resources for All Queues
2482 * @adapter: board private structure
2483 *
2484 * Free all transmit software resources
2485 **/
2486static void igb_free_all_tx_resources(struct igb_adapter *adapter)
2487{
2488 int i;
2489
2490 for (i = 0; i < adapter->num_tx_queues; i++)
3b644cf6 2491 igb_free_tx_resources(&adapter->tx_ring[i]);
9d5c8243
AK
2492}
2493
b1a436c3
AD
2494void igb_unmap_and_free_tx_resource(struct igb_ring *tx_ring,
2495 struct igb_buffer *buffer_info)
9d5c8243 2496{
65689fef 2497 buffer_info->dma = 0;
9d5c8243 2498 if (buffer_info->skb) {
80785298
AD
2499 skb_dma_unmap(&tx_ring->pdev->dev,
2500 buffer_info->skb,
65689fef 2501 DMA_TO_DEVICE);
9d5c8243
AK
2502 dev_kfree_skb_any(buffer_info->skb);
2503 buffer_info->skb = NULL;
2504 }
2505 buffer_info->time_stamp = 0;
2506 /* buffer_info must be completely set up in the transmit path */
2507}
2508
2509/**
2510 * igb_clean_tx_ring - Free Tx Buffers
9d5c8243
AK
2511 * @tx_ring: ring to be cleaned
2512 **/
3b644cf6 2513static void igb_clean_tx_ring(struct igb_ring *tx_ring)
9d5c8243
AK
2514{
2515 struct igb_buffer *buffer_info;
2516 unsigned long size;
2517 unsigned int i;
2518
2519 if (!tx_ring->buffer_info)
2520 return;
2521 /* Free all the Tx ring sk_buffs */
2522
2523 for (i = 0; i < tx_ring->count; i++) {
2524 buffer_info = &tx_ring->buffer_info[i];
80785298 2525 igb_unmap_and_free_tx_resource(tx_ring, buffer_info);
9d5c8243
AK
2526 }
2527
2528 size = sizeof(struct igb_buffer) * tx_ring->count;
2529 memset(tx_ring->buffer_info, 0, size);
2530
2531 /* Zero out the descriptor ring */
2532
2533 memset(tx_ring->desc, 0, tx_ring->size);
2534
2535 tx_ring->next_to_use = 0;
2536 tx_ring->next_to_clean = 0;
2537
fce99e34
AD
2538 writel(0, tx_ring->head);
2539 writel(0, tx_ring->tail);
9d5c8243
AK
2540}
2541
2542/**
2543 * igb_clean_all_tx_rings - Free Tx Buffers for all queues
2544 * @adapter: board private structure
2545 **/
2546static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
2547{
2548 int i;
2549
2550 for (i = 0; i < adapter->num_tx_queues; i++)
3b644cf6 2551 igb_clean_tx_ring(&adapter->tx_ring[i]);
9d5c8243
AK
2552}
2553
2554/**
2555 * igb_free_rx_resources - Free Rx Resources
9d5c8243
AK
2556 * @rx_ring: ring to clean the resources from
2557 *
2558 * Free all receive software resources
2559 **/
68fd9910 2560void igb_free_rx_resources(struct igb_ring *rx_ring)
9d5c8243 2561{
3b644cf6 2562 igb_clean_rx_ring(rx_ring);
9d5c8243
AK
2563
2564 vfree(rx_ring->buffer_info);
2565 rx_ring->buffer_info = NULL;
2566
80785298
AD
2567 pci_free_consistent(rx_ring->pdev, rx_ring->size,
2568 rx_ring->desc, rx_ring->dma);
9d5c8243
AK
2569
2570 rx_ring->desc = NULL;
2571}
2572
2573/**
2574 * igb_free_all_rx_resources - Free Rx Resources for All Queues
2575 * @adapter: board private structure
2576 *
2577 * Free all receive software resources
2578 **/
2579static void igb_free_all_rx_resources(struct igb_adapter *adapter)
2580{
2581 int i;
2582
2583 for (i = 0; i < adapter->num_rx_queues; i++)
3b644cf6 2584 igb_free_rx_resources(&adapter->rx_ring[i]);
9d5c8243
AK
2585}
2586
2587/**
2588 * igb_clean_rx_ring - Free Rx Buffers per Queue
9d5c8243
AK
2589 * @rx_ring: ring to free buffers from
2590 **/
3b644cf6 2591static void igb_clean_rx_ring(struct igb_ring *rx_ring)
9d5c8243
AK
2592{
2593 struct igb_buffer *buffer_info;
9d5c8243
AK
2594 unsigned long size;
2595 unsigned int i;
2596
2597 if (!rx_ring->buffer_info)
2598 return;
2599 /* Free all the Rx ring sk_buffs */
2600 for (i = 0; i < rx_ring->count; i++) {
2601 buffer_info = &rx_ring->buffer_info[i];
2602 if (buffer_info->dma) {
80785298
AD
2603 pci_unmap_single(rx_ring->pdev,
2604 buffer_info->dma,
4c844851 2605 rx_ring->rx_buffer_len,
6ec43fe6 2606 PCI_DMA_FROMDEVICE);
9d5c8243
AK
2607 buffer_info->dma = 0;
2608 }
2609
2610 if (buffer_info->skb) {
2611 dev_kfree_skb(buffer_info->skb);
2612 buffer_info->skb = NULL;
2613 }
6ec43fe6 2614 if (buffer_info->page_dma) {
80785298
AD
2615 pci_unmap_page(rx_ring->pdev,
2616 buffer_info->page_dma,
6ec43fe6
AD
2617 PAGE_SIZE / 2,
2618 PCI_DMA_FROMDEVICE);
2619 buffer_info->page_dma = 0;
2620 }
9d5c8243 2621 if (buffer_info->page) {
9d5c8243
AK
2622 put_page(buffer_info->page);
2623 buffer_info->page = NULL;
bf36c1a0 2624 buffer_info->page_offset = 0;
9d5c8243
AK
2625 }
2626 }
2627
9d5c8243
AK
2628 size = sizeof(struct igb_buffer) * rx_ring->count;
2629 memset(rx_ring->buffer_info, 0, size);
2630
2631 /* Zero out the descriptor ring */
2632 memset(rx_ring->desc, 0, rx_ring->size);
2633
2634 rx_ring->next_to_clean = 0;
2635 rx_ring->next_to_use = 0;
2636
fce99e34
AD
2637 writel(0, rx_ring->head);
2638 writel(0, rx_ring->tail);
9d5c8243
AK
2639}
2640
2641/**
2642 * igb_clean_all_rx_rings - Free Rx Buffers for all queues
2643 * @adapter: board private structure
2644 **/
2645static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
2646{
2647 int i;
2648
2649 for (i = 0; i < adapter->num_rx_queues; i++)
3b644cf6 2650 igb_clean_rx_ring(&adapter->rx_ring[i]);
9d5c8243
AK
2651}
2652
2653/**
2654 * igb_set_mac - Change the Ethernet Address of the NIC
2655 * @netdev: network interface device structure
2656 * @p: pointer to an address structure
2657 *
2658 * Returns 0 on success, negative on failure
2659 **/
2660static int igb_set_mac(struct net_device *netdev, void *p)
2661{
2662 struct igb_adapter *adapter = netdev_priv(netdev);
28b0759c 2663 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK
2664 struct sockaddr *addr = p;
2665
2666 if (!is_valid_ether_addr(addr->sa_data))
2667 return -EADDRNOTAVAIL;
2668
2669 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
28b0759c 2670 memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
9d5c8243 2671
26ad9178
AD
2672 /* set the correct pool for the new PF MAC address in entry 0 */
2673 igb_rar_set_qsel(adapter, hw->mac.addr, 0,
2674 adapter->vfs_allocated_count);
e1739522 2675
9d5c8243
AK
2676 return 0;
2677}
2678
2679/**
68d480c4 2680 * igb_write_mc_addr_list - write multicast addresses to MTA
9d5c8243
AK
2681 * @netdev: network interface device structure
2682 *
68d480c4
AD
2683 * Writes multicast address list to the MTA hash table.
2684 * Returns: -ENOMEM on failure
2685 * 0 on no addresses written
2686 * X on writing X addresses to MTA
9d5c8243 2687 **/
68d480c4 2688static int igb_write_mc_addr_list(struct net_device *netdev)
9d5c8243
AK
2689{
2690 struct igb_adapter *adapter = netdev_priv(netdev);
2691 struct e1000_hw *hw = &adapter->hw;
ff41f8dc 2692 struct dev_mc_list *mc_ptr = netdev->mc_list;
68d480c4
AD
2693 u8 *mta_list;
2694 u32 vmolr = 0;
9d5c8243
AK
2695 int i;
2696
68d480c4
AD
2697 if (!netdev->mc_count) {
2698 /* nothing to program, so clear mc list */
2699 igb_update_mc_addr_list(hw, NULL, 0);
2700 igb_restore_vf_multicasts(adapter);
2701 return 0;
2702 }
9d5c8243 2703
68d480c4
AD
2704 mta_list = kzalloc(netdev->mc_count * 6, GFP_ATOMIC);
2705 if (!mta_list)
2706 return -ENOMEM;
ff41f8dc 2707
68d480c4
AD
2708 /* set vmolr receive overflow multicast bit */
2709 vmolr |= E1000_VMOLR_ROMPE;
2710
2711 /* The shared function expects a packed array of only addresses. */
2712 mc_ptr = netdev->mc_list;
2713
2714 for (i = 0; i < netdev->mc_count; i++) {
2715 if (!mc_ptr)
2716 break;
2717 memcpy(mta_list + (i*ETH_ALEN), mc_ptr->dmi_addr, ETH_ALEN);
2718 mc_ptr = mc_ptr->next;
746b9f02 2719 }
68d480c4
AD
2720 igb_update_mc_addr_list(hw, mta_list, i);
2721 kfree(mta_list);
2722
2723 return netdev->mc_count;
2724}
2725
2726/**
2727 * igb_write_uc_addr_list - write unicast addresses to RAR table
2728 * @netdev: network interface device structure
2729 *
2730 * Writes unicast address list to the RAR table.
2731 * Returns: -ENOMEM on failure/insufficient address space
2732 * 0 on no addresses written
2733 * X on writing X addresses to the RAR table
2734 **/
2735static int igb_write_uc_addr_list(struct net_device *netdev)
2736{
2737 struct igb_adapter *adapter = netdev_priv(netdev);
2738 struct e1000_hw *hw = &adapter->hw;
2739 unsigned int vfn = adapter->vfs_allocated_count;
2740 unsigned int rar_entries = hw->mac.rar_entry_count - (vfn + 1);
2741 int count = 0;
2742
2743 /* return ENOMEM indicating insufficient memory for addresses */
2744 if (netdev->uc.count > rar_entries)
2745 return -ENOMEM;
9d5c8243 2746
ff41f8dc
AD
2747 if (netdev->uc.count && rar_entries) {
2748 struct netdev_hw_addr *ha;
2749 list_for_each_entry(ha, &netdev->uc.list, list) {
2750 if (!rar_entries)
2751 break;
26ad9178
AD
2752 igb_rar_set_qsel(adapter, ha->addr,
2753 rar_entries--,
68d480c4
AD
2754 vfn);
2755 count++;
ff41f8dc
AD
2756 }
2757 }
2758 /* write the addresses in reverse order to avoid write combining */
2759 for (; rar_entries > 0 ; rar_entries--) {
2760 wr32(E1000_RAH(rar_entries), 0);
2761 wr32(E1000_RAL(rar_entries), 0);
2762 }
2763 wrfl();
2764
68d480c4
AD
2765 return count;
2766}
2767
2768/**
2769 * igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
2770 * @netdev: network interface device structure
2771 *
2772 * The set_rx_mode entry point is called whenever the unicast or multicast
2773 * address lists or the network interface flags are updated. This routine is
2774 * responsible for configuring the hardware for proper unicast, multicast,
2775 * promiscuous mode, and all-multi behavior.
2776 **/
2777static void igb_set_rx_mode(struct net_device *netdev)
2778{
2779 struct igb_adapter *adapter = netdev_priv(netdev);
2780 struct e1000_hw *hw = &adapter->hw;
2781 unsigned int vfn = adapter->vfs_allocated_count;
2782 u32 rctl, vmolr = 0;
2783 int count;
2784
2785 /* Check for Promiscuous and All Multicast modes */
2786 rctl = rd32(E1000_RCTL);
2787
2788 /* clear the effected bits */
2789 rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE | E1000_RCTL_VFE);
2790
2791 if (netdev->flags & IFF_PROMISC) {
2792 rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2793 vmolr |= (E1000_VMOLR_ROPE | E1000_VMOLR_MPME);
2794 } else {
2795 if (netdev->flags & IFF_ALLMULTI) {
2796 rctl |= E1000_RCTL_MPE;
2797 vmolr |= E1000_VMOLR_MPME;
2798 } else {
2799 /*
2800 * Write addresses to the MTA, if the attempt fails
2801 * then we should just turn on promiscous mode so
2802 * that we can at least receive multicast traffic
2803 */
2804 count = igb_write_mc_addr_list(netdev);
2805 if (count < 0) {
2806 rctl |= E1000_RCTL_MPE;
2807 vmolr |= E1000_VMOLR_MPME;
2808 } else if (count) {
2809 vmolr |= E1000_VMOLR_ROMPE;
2810 }
2811 }
2812 /*
2813 * Write addresses to available RAR registers, if there is not
2814 * sufficient space to store all the addresses then enable
2815 * unicast promiscous mode
2816 */
2817 count = igb_write_uc_addr_list(netdev);
2818 if (count < 0) {
2819 rctl |= E1000_RCTL_UPE;
2820 vmolr |= E1000_VMOLR_ROPE;
2821 }
2822 rctl |= E1000_RCTL_VFE;
28fc06f5 2823 }
68d480c4 2824 wr32(E1000_RCTL, rctl);
28fc06f5 2825
68d480c4
AD
2826 /*
2827 * In order to support SR-IOV and eventually VMDq it is necessary to set
2828 * the VMOLR to enable the appropriate modes. Without this workaround
2829 * we will have issues with VLAN tag stripping not being done for frames
2830 * that are only arriving because we are the default pool
2831 */
2832 if (hw->mac.type < e1000_82576)
28fc06f5 2833 return;
9d5c8243 2834
68d480c4
AD
2835 vmolr |= rd32(E1000_VMOLR(vfn)) &
2836 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
2837 wr32(E1000_VMOLR(vfn), vmolr);
28fc06f5 2838 igb_restore_vf_multicasts(adapter);
9d5c8243
AK
2839}
2840
2841/* Need to wait a few seconds after link up to get diagnostic information from
2842 * the phy */
2843static void igb_update_phy_info(unsigned long data)
2844{
2845 struct igb_adapter *adapter = (struct igb_adapter *) data;
f5f4cf08 2846 igb_get_phy_info(&adapter->hw);
9d5c8243
AK
2847}
2848
4d6b725e
AD
2849/**
2850 * igb_has_link - check shared code for link and determine up/down
2851 * @adapter: pointer to driver private info
2852 **/
2853static bool igb_has_link(struct igb_adapter *adapter)
2854{
2855 struct e1000_hw *hw = &adapter->hw;
2856 bool link_active = false;
2857 s32 ret_val = 0;
2858
2859 /* get_link_status is set on LSC (link status) interrupt or
2860 * rx sequence error interrupt. get_link_status will stay
2861 * false until the e1000_check_for_link establishes link
2862 * for copper adapters ONLY
2863 */
2864 switch (hw->phy.media_type) {
2865 case e1000_media_type_copper:
2866 if (hw->mac.get_link_status) {
2867 ret_val = hw->mac.ops.check_for_link(hw);
2868 link_active = !hw->mac.get_link_status;
2869 } else {
2870 link_active = true;
2871 }
2872 break;
4d6b725e
AD
2873 case e1000_media_type_internal_serdes:
2874 ret_val = hw->mac.ops.check_for_link(hw);
2875 link_active = hw->mac.serdes_has_link;
2876 break;
2877 default:
2878 case e1000_media_type_unknown:
2879 break;
2880 }
2881
2882 return link_active;
2883}
2884
9d5c8243
AK
2885/**
2886 * igb_watchdog - Timer Call-back
2887 * @data: pointer to adapter cast into an unsigned long
2888 **/
2889static void igb_watchdog(unsigned long data)
2890{
2891 struct igb_adapter *adapter = (struct igb_adapter *)data;
2892 /* Do the rest outside of interrupt context */
2893 schedule_work(&adapter->watchdog_task);
2894}
2895
2896static void igb_watchdog_task(struct work_struct *work)
2897{
2898 struct igb_adapter *adapter = container_of(work,
2899 struct igb_adapter, watchdog_task);
2900 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK
2901 struct net_device *netdev = adapter->netdev;
2902 struct igb_ring *tx_ring = adapter->tx_ring;
9d5c8243 2903 u32 link;
7a6ea550 2904 int i;
9d5c8243 2905
4d6b725e
AD
2906 link = igb_has_link(adapter);
2907 if ((netif_carrier_ok(netdev)) && link)
9d5c8243
AK
2908 goto link_up;
2909
9d5c8243
AK
2910 if (link) {
2911 if (!netif_carrier_ok(netdev)) {
2912 u32 ctrl;
2913 hw->mac.ops.get_speed_and_duplex(&adapter->hw,
2914 &adapter->link_speed,
2915 &adapter->link_duplex);
2916
2917 ctrl = rd32(E1000_CTRL);
527d47c1
AD
2918 /* Links status message must follow this format */
2919 printk(KERN_INFO "igb: %s NIC Link is Up %d Mbps %s, "
9d5c8243 2920 "Flow Control: %s\n",
527d47c1 2921 netdev->name,
9d5c8243
AK
2922 adapter->link_speed,
2923 adapter->link_duplex == FULL_DUPLEX ?
2924 "Full Duplex" : "Half Duplex",
2925 ((ctrl & E1000_CTRL_TFCE) && (ctrl &
2926 E1000_CTRL_RFCE)) ? "RX/TX" : ((ctrl &
2927 E1000_CTRL_RFCE) ? "RX" : ((ctrl &
2928 E1000_CTRL_TFCE) ? "TX" : "None")));
2929
2930 /* tweak tx_queue_len according to speed/duplex and
2931 * adjust the timeout factor */
2932 netdev->tx_queue_len = adapter->tx_queue_len;
2933 adapter->tx_timeout_factor = 1;
2934 switch (adapter->link_speed) {
2935 case SPEED_10:
2936 netdev->tx_queue_len = 10;
2937 adapter->tx_timeout_factor = 14;
2938 break;
2939 case SPEED_100:
2940 netdev->tx_queue_len = 100;
2941 /* maybe add some timeout factor ? */
2942 break;
2943 }
2944
2945 netif_carrier_on(netdev);
9d5c8243 2946
4ae196df
AD
2947 igb_ping_all_vfs(adapter);
2948
4b1a9877 2949 /* link state has changed, schedule phy info update */
9d5c8243
AK
2950 if (!test_bit(__IGB_DOWN, &adapter->state))
2951 mod_timer(&adapter->phy_info_timer,
2952 round_jiffies(jiffies + 2 * HZ));
2953 }
2954 } else {
2955 if (netif_carrier_ok(netdev)) {
2956 adapter->link_speed = 0;
2957 adapter->link_duplex = 0;
527d47c1
AD
2958 /* Links status message must follow this format */
2959 printk(KERN_INFO "igb: %s NIC Link is Down\n",
2960 netdev->name);
9d5c8243 2961 netif_carrier_off(netdev);
4b1a9877 2962
4ae196df
AD
2963 igb_ping_all_vfs(adapter);
2964
4b1a9877 2965 /* link state has changed, schedule phy info update */
9d5c8243
AK
2966 if (!test_bit(__IGB_DOWN, &adapter->state))
2967 mod_timer(&adapter->phy_info_timer,
2968 round_jiffies(jiffies + 2 * HZ));
2969 }
2970 }
2971
2972link_up:
2973 igb_update_stats(adapter);
2974
4b1a9877 2975 hw->mac.tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
9d5c8243 2976 adapter->tpt_old = adapter->stats.tpt;
4b1a9877 2977 hw->mac.collision_delta = adapter->stats.colc - adapter->colc_old;
9d5c8243
AK
2978 adapter->colc_old = adapter->stats.colc;
2979
2980 adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
2981 adapter->gorc_old = adapter->stats.gorc;
2982 adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
2983 adapter->gotc_old = adapter->stats.gotc;
2984
2985 igb_update_adaptive(&adapter->hw);
2986
2987 if (!netif_carrier_ok(netdev)) {
c493ea45 2988 if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
9d5c8243
AK
2989 /* We've lost link, so the controller stops DMA,
2990 * but we've got queued Tx work that's never going
2991 * to get done, so reset controller to flush Tx.
2992 * (Do the reset outside of interrupt context). */
2993 adapter->tx_timeout_count++;
2994 schedule_work(&adapter->reset_task);
c2d5ab49
JB
2995 /* return immediately since reset is imminent */
2996 return;
9d5c8243
AK
2997 }
2998 }
2999
3000 /* Cause software interrupt to ensure rx ring is cleaned */
7a6ea550 3001 if (adapter->msix_entries) {
047e0030
AD
3002 u32 eics = 0;
3003 for (i = 0; i < adapter->num_q_vectors; i++) {
3004 struct igb_q_vector *q_vector = adapter->q_vector[i];
3005 eics |= q_vector->eims_value;
3006 }
7a6ea550
AD
3007 wr32(E1000_EICS, eics);
3008 } else {
3009 wr32(E1000_ICS, E1000_ICS_RXDMT0);
3010 }
9d5c8243
AK
3011
3012 /* Force detection of hung controller every watchdog period */
3013 tx_ring->detect_tx_hung = true;
3014
3015 /* Reset the timer */
3016 if (!test_bit(__IGB_DOWN, &adapter->state))
3017 mod_timer(&adapter->watchdog_timer,
3018 round_jiffies(jiffies + 2 * HZ));
3019}
3020
3021enum latency_range {
3022 lowest_latency = 0,
3023 low_latency = 1,
3024 bulk_latency = 2,
3025 latency_invalid = 255
3026};
3027
6eb5a7f1
AD
3028/**
3029 * igb_update_ring_itr - update the dynamic ITR value based on packet size
3030 *
3031 * Stores a new ITR value based on strictly on packet size. This
3032 * algorithm is less sophisticated than that used in igb_update_itr,
3033 * due to the difficulty of synchronizing statistics across multiple
3034 * receive rings. The divisors and thresholds used by this fuction
3035 * were determined based on theoretical maximum wire speed and testing
3036 * data, in order to minimize response time while increasing bulk
3037 * throughput.
3038 * This functionality is controlled by the InterruptThrottleRate module
3039 * parameter (see igb_param.c)
3040 * NOTE: This function is called only when operating in a multiqueue
3041 * receive environment.
047e0030 3042 * @q_vector: pointer to q_vector
6eb5a7f1 3043 **/
047e0030 3044static void igb_update_ring_itr(struct igb_q_vector *q_vector)
9d5c8243 3045{
047e0030 3046 int new_val = q_vector->itr_val;
6eb5a7f1 3047 int avg_wire_size = 0;
047e0030 3048 struct igb_adapter *adapter = q_vector->adapter;
9d5c8243 3049
6eb5a7f1
AD
3050 /* For non-gigabit speeds, just fix the interrupt rate at 4000
3051 * ints/sec - ITR timer value of 120 ticks.
3052 */
3053 if (adapter->link_speed != SPEED_1000) {
047e0030 3054 new_val = 976;
6eb5a7f1 3055 goto set_itr_val;
9d5c8243 3056 }
047e0030
AD
3057
3058 if (q_vector->rx_ring && q_vector->rx_ring->total_packets) {
3059 struct igb_ring *ring = q_vector->rx_ring;
3060 avg_wire_size = ring->total_bytes / ring->total_packets;
3061 }
3062
3063 if (q_vector->tx_ring && q_vector->tx_ring->total_packets) {
3064 struct igb_ring *ring = q_vector->tx_ring;
3065 avg_wire_size = max_t(u32, avg_wire_size,
3066 (ring->total_bytes /
3067 ring->total_packets));
3068 }
3069
3070 /* if avg_wire_size isn't set no work was done */
3071 if (!avg_wire_size)
3072 goto clear_counts;
9d5c8243 3073
6eb5a7f1
AD
3074 /* Add 24 bytes to size to account for CRC, preamble, and gap */
3075 avg_wire_size += 24;
3076
3077 /* Don't starve jumbo frames */
3078 avg_wire_size = min(avg_wire_size, 3000);
9d5c8243 3079
6eb5a7f1
AD
3080 /* Give a little boost to mid-size frames */
3081 if ((avg_wire_size > 300) && (avg_wire_size < 1200))
3082 new_val = avg_wire_size / 3;
3083 else
3084 new_val = avg_wire_size / 2;
9d5c8243 3085
6eb5a7f1 3086set_itr_val:
047e0030
AD
3087 if (new_val != q_vector->itr_val) {
3088 q_vector->itr_val = new_val;
3089 q_vector->set_itr = 1;
9d5c8243 3090 }
6eb5a7f1 3091clear_counts:
047e0030
AD
3092 if (q_vector->rx_ring) {
3093 q_vector->rx_ring->total_bytes = 0;
3094 q_vector->rx_ring->total_packets = 0;
3095 }
3096 if (q_vector->tx_ring) {
3097 q_vector->tx_ring->total_bytes = 0;
3098 q_vector->tx_ring->total_packets = 0;
3099 }
9d5c8243
AK
3100}
3101
3102/**
3103 * igb_update_itr - update the dynamic ITR value based on statistics
3104 * Stores a new ITR value based on packets and byte
3105 * counts during the last interrupt. The advantage of per interrupt
3106 * computation is faster updates and more accurate ITR for the current
3107 * traffic pattern. Constants in this function were computed
3108 * based on theoretical maximum wire speed and thresholds were set based
3109 * on testing data as well as attempting to minimize response time
3110 * while increasing bulk throughput.
3111 * this functionality is controlled by the InterruptThrottleRate module
3112 * parameter (see igb_param.c)
3113 * NOTE: These calculations are only valid when operating in a single-
3114 * queue environment.
3115 * @adapter: pointer to adapter
047e0030 3116 * @itr_setting: current q_vector->itr_val
9d5c8243
AK
3117 * @packets: the number of packets during this measurement interval
3118 * @bytes: the number of bytes during this measurement interval
3119 **/
3120static unsigned int igb_update_itr(struct igb_adapter *adapter, u16 itr_setting,
3121 int packets, int bytes)
3122{
3123 unsigned int retval = itr_setting;
3124
3125 if (packets == 0)
3126 goto update_itr_done;
3127
3128 switch (itr_setting) {
3129 case lowest_latency:
3130 /* handle TSO and jumbo frames */
3131 if (bytes/packets > 8000)
3132 retval = bulk_latency;
3133 else if ((packets < 5) && (bytes > 512))
3134 retval = low_latency;
3135 break;
3136 case low_latency: /* 50 usec aka 20000 ints/s */
3137 if (bytes > 10000) {
3138 /* this if handles the TSO accounting */
3139 if (bytes/packets > 8000) {
3140 retval = bulk_latency;
3141 } else if ((packets < 10) || ((bytes/packets) > 1200)) {
3142 retval = bulk_latency;
3143 } else if ((packets > 35)) {
3144 retval = lowest_latency;
3145 }
3146 } else if (bytes/packets > 2000) {
3147 retval = bulk_latency;
3148 } else if (packets <= 2 && bytes < 512) {
3149 retval = lowest_latency;
3150 }
3151 break;
3152 case bulk_latency: /* 250 usec aka 4000 ints/s */
3153 if (bytes > 25000) {
3154 if (packets > 35)
3155 retval = low_latency;
1e5c3d21 3156 } else if (bytes < 1500) {
9d5c8243
AK
3157 retval = low_latency;
3158 }
3159 break;
3160 }
3161
3162update_itr_done:
3163 return retval;
3164}
3165
6eb5a7f1 3166static void igb_set_itr(struct igb_adapter *adapter)
9d5c8243 3167{
047e0030 3168 struct igb_q_vector *q_vector = adapter->q_vector[0];
9d5c8243 3169 u16 current_itr;
047e0030 3170 u32 new_itr = q_vector->itr_val;
9d5c8243
AK
3171
3172 /* for non-gigabit speeds, just fix the interrupt rate at 4000 */
3173 if (adapter->link_speed != SPEED_1000) {
3174 current_itr = 0;
3175 new_itr = 4000;
3176 goto set_itr_now;
3177 }
3178
3179 adapter->rx_itr = igb_update_itr(adapter,
3180 adapter->rx_itr,
3181 adapter->rx_ring->total_packets,
3182 adapter->rx_ring->total_bytes);
9d5c8243 3183
047e0030
AD
3184 adapter->tx_itr = igb_update_itr(adapter,
3185 adapter->tx_itr,
3186 adapter->tx_ring->total_packets,
3187 adapter->tx_ring->total_bytes);
3188 current_itr = max(adapter->rx_itr, adapter->tx_itr);
9d5c8243 3189
6eb5a7f1 3190 /* conservative mode (itr 3) eliminates the lowest_latency setting */
4fc82adf 3191 if (adapter->rx_itr_setting == 3 && current_itr == lowest_latency)
6eb5a7f1
AD
3192 current_itr = low_latency;
3193
9d5c8243
AK
3194 switch (current_itr) {
3195 /* counts and packets in update_itr are dependent on these numbers */
3196 case lowest_latency:
78b1f607 3197 new_itr = 56; /* aka 70,000 ints/sec */
9d5c8243
AK
3198 break;
3199 case low_latency:
78b1f607 3200 new_itr = 196; /* aka 20,000 ints/sec */
9d5c8243
AK
3201 break;
3202 case bulk_latency:
78b1f607 3203 new_itr = 980; /* aka 4,000 ints/sec */
9d5c8243
AK
3204 break;
3205 default:
3206 break;
3207 }
3208
3209set_itr_now:
6eb5a7f1
AD
3210 adapter->rx_ring->total_bytes = 0;
3211 adapter->rx_ring->total_packets = 0;
047e0030
AD
3212 adapter->tx_ring->total_bytes = 0;
3213 adapter->tx_ring->total_packets = 0;
6eb5a7f1 3214
047e0030 3215 if (new_itr != q_vector->itr_val) {
9d5c8243
AK
3216 /* this attempts to bias the interrupt rate towards Bulk
3217 * by adding intermediate steps when interrupt rate is
3218 * increasing */
047e0030
AD
3219 new_itr = new_itr > q_vector->itr_val ?
3220 max((new_itr * q_vector->itr_val) /
3221 (new_itr + (q_vector->itr_val >> 2)),
3222 new_itr) :
9d5c8243
AK
3223 new_itr;
3224 /* Don't write the value here; it resets the adapter's
3225 * internal timer, and causes us to delay far longer than
3226 * we should between interrupts. Instead, we write the ITR
3227 * value at the beginning of the next interrupt so the timing
3228 * ends up being correct.
3229 */
047e0030
AD
3230 q_vector->itr_val = new_itr;
3231 q_vector->set_itr = 1;
9d5c8243
AK
3232 }
3233
3234 return;
3235}
3236
9d5c8243
AK
3237#define IGB_TX_FLAGS_CSUM 0x00000001
3238#define IGB_TX_FLAGS_VLAN 0x00000002
3239#define IGB_TX_FLAGS_TSO 0x00000004
3240#define IGB_TX_FLAGS_IPV4 0x00000008
33af6bcc 3241#define IGB_TX_FLAGS_TSTAMP 0x00000010
9d5c8243
AK
3242#define IGB_TX_FLAGS_VLAN_MASK 0xffff0000
3243#define IGB_TX_FLAGS_VLAN_SHIFT 16
3244
85ad76b2 3245static inline int igb_tso_adv(struct igb_ring *tx_ring,
9d5c8243
AK
3246 struct sk_buff *skb, u32 tx_flags, u8 *hdr_len)
3247{
3248 struct e1000_adv_tx_context_desc *context_desc;
3249 unsigned int i;
3250 int err;
3251 struct igb_buffer *buffer_info;
3252 u32 info = 0, tu_cmd = 0;
3253 u32 mss_l4len_idx, l4len;
3254 *hdr_len = 0;
3255
3256 if (skb_header_cloned(skb)) {
3257 err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3258 if (err)
3259 return err;
3260 }
3261
3262 l4len = tcp_hdrlen(skb);
3263 *hdr_len += l4len;
3264
3265 if (skb->protocol == htons(ETH_P_IP)) {
3266 struct iphdr *iph = ip_hdr(skb);
3267 iph->tot_len = 0;
3268 iph->check = 0;
3269 tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
3270 iph->daddr, 0,
3271 IPPROTO_TCP,
3272 0);
3273 } else if (skb_shinfo(skb)->gso_type == SKB_GSO_TCPV6) {
3274 ipv6_hdr(skb)->payload_len = 0;
3275 tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
3276 &ipv6_hdr(skb)->daddr,
3277 0, IPPROTO_TCP, 0);
3278 }
3279
3280 i = tx_ring->next_to_use;
3281
3282 buffer_info = &tx_ring->buffer_info[i];
3283 context_desc = E1000_TX_CTXTDESC_ADV(*tx_ring, i);
3284 /* VLAN MACLEN IPLEN */
3285 if (tx_flags & IGB_TX_FLAGS_VLAN)
3286 info |= (tx_flags & IGB_TX_FLAGS_VLAN_MASK);
3287 info |= (skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT);
3288 *hdr_len += skb_network_offset(skb);
3289 info |= skb_network_header_len(skb);
3290 *hdr_len += skb_network_header_len(skb);
3291 context_desc->vlan_macip_lens = cpu_to_le32(info);
3292
3293 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
3294 tu_cmd |= (E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT);
3295
3296 if (skb->protocol == htons(ETH_P_IP))
3297 tu_cmd |= E1000_ADVTXD_TUCMD_IPV4;
3298 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
3299
3300 context_desc->type_tucmd_mlhl = cpu_to_le32(tu_cmd);
3301
3302 /* MSS L4LEN IDX */
3303 mss_l4len_idx = (skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT);
3304 mss_l4len_idx |= (l4len << E1000_ADVTXD_L4LEN_SHIFT);
3305
73cd78f1 3306 /* For 82575, context index must be unique per ring. */
85ad76b2
AD
3307 if (tx_ring->flags & IGB_RING_FLAG_TX_CTX_IDX)
3308 mss_l4len_idx |= tx_ring->reg_idx << 4;
9d5c8243
AK
3309
3310 context_desc->mss_l4len_idx = cpu_to_le32(mss_l4len_idx);
3311 context_desc->seqnum_seed = 0;
3312
3313 buffer_info->time_stamp = jiffies;
0e014cb1 3314 buffer_info->next_to_watch = i;
9d5c8243
AK
3315 buffer_info->dma = 0;
3316 i++;
3317 if (i == tx_ring->count)
3318 i = 0;
3319
3320 tx_ring->next_to_use = i;
3321
3322 return true;
3323}
3324
85ad76b2
AD
3325static inline bool igb_tx_csum_adv(struct igb_ring *tx_ring,
3326 struct sk_buff *skb, u32 tx_flags)
9d5c8243
AK
3327{
3328 struct e1000_adv_tx_context_desc *context_desc;
80785298 3329 struct pci_dev *pdev = tx_ring->pdev;
9d5c8243
AK
3330 struct igb_buffer *buffer_info;
3331 u32 info = 0, tu_cmd = 0;
80785298 3332 unsigned int i;
9d5c8243
AK
3333
3334 if ((skb->ip_summed == CHECKSUM_PARTIAL) ||
3335 (tx_flags & IGB_TX_FLAGS_VLAN)) {
3336 i = tx_ring->next_to_use;
3337 buffer_info = &tx_ring->buffer_info[i];
3338 context_desc = E1000_TX_CTXTDESC_ADV(*tx_ring, i);
3339
3340 if (tx_flags & IGB_TX_FLAGS_VLAN)
3341 info |= (tx_flags & IGB_TX_FLAGS_VLAN_MASK);
3342 info |= (skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT);
3343 if (skb->ip_summed == CHECKSUM_PARTIAL)
3344 info |= skb_network_header_len(skb);
3345
3346 context_desc->vlan_macip_lens = cpu_to_le32(info);
3347
3348 tu_cmd |= (E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT);
3349
3350 if (skb->ip_summed == CHECKSUM_PARTIAL) {
fa4a7ef3
AJ
3351 __be16 protocol;
3352
3353 if (skb->protocol == cpu_to_be16(ETH_P_8021Q)) {
3354 const struct vlan_ethhdr *vhdr =
3355 (const struct vlan_ethhdr*)skb->data;
3356
3357 protocol = vhdr->h_vlan_encapsulated_proto;
3358 } else {
3359 protocol = skb->protocol;
3360 }
3361
3362 switch (protocol) {
09640e63 3363 case cpu_to_be16(ETH_P_IP):
9d5c8243 3364 tu_cmd |= E1000_ADVTXD_TUCMD_IPV4;
44b0cda3
MW
3365 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
3366 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
b9473560
JB
3367 else if (ip_hdr(skb)->protocol == IPPROTO_SCTP)
3368 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_SCTP;
44b0cda3 3369 break;
09640e63 3370 case cpu_to_be16(ETH_P_IPV6):
44b0cda3
MW
3371 /* XXX what about other V6 headers?? */
3372 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
3373 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
b9473560
JB
3374 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_SCTP)
3375 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_SCTP;
44b0cda3
MW
3376 break;
3377 default:
3378 if (unlikely(net_ratelimit()))
80785298 3379 dev_warn(&pdev->dev,
44b0cda3
MW
3380 "partial checksum but proto=%x!\n",
3381 skb->protocol);
3382 break;
3383 }
9d5c8243
AK
3384 }
3385
3386 context_desc->type_tucmd_mlhl = cpu_to_le32(tu_cmd);
3387 context_desc->seqnum_seed = 0;
85ad76b2 3388 if (tx_ring->flags & IGB_RING_FLAG_TX_CTX_IDX)
7dfc16fa 3389 context_desc->mss_l4len_idx =
85ad76b2 3390 cpu_to_le32(tx_ring->reg_idx << 4);
9d5c8243
AK
3391
3392 buffer_info->time_stamp = jiffies;
0e014cb1 3393 buffer_info->next_to_watch = i;
9d5c8243
AK
3394 buffer_info->dma = 0;
3395
3396 i++;
3397 if (i == tx_ring->count)
3398 i = 0;
3399 tx_ring->next_to_use = i;
3400
3401 return true;
3402 }
9d5c8243
AK
3403 return false;
3404}
3405
3406#define IGB_MAX_TXD_PWR 16
3407#define IGB_MAX_DATA_PER_TXD (1<<IGB_MAX_TXD_PWR)
3408
80785298 3409static inline int igb_tx_map_adv(struct igb_ring *tx_ring, struct sk_buff *skb,
0e014cb1 3410 unsigned int first)
9d5c8243
AK
3411{
3412 struct igb_buffer *buffer_info;
80785298 3413 struct pci_dev *pdev = tx_ring->pdev;
9d5c8243
AK
3414 unsigned int len = skb_headlen(skb);
3415 unsigned int count = 0, i;
3416 unsigned int f;
65689fef 3417 dma_addr_t *map;
9d5c8243
AK
3418
3419 i = tx_ring->next_to_use;
3420
80785298
AD
3421 if (skb_dma_map(&pdev->dev, skb, DMA_TO_DEVICE)) {
3422 dev_err(&pdev->dev, "TX DMA map failed\n");
65689fef
AD
3423 return 0;
3424 }
3425
3426 map = skb_shinfo(skb)->dma_maps;
3427
9d5c8243
AK
3428 buffer_info = &tx_ring->buffer_info[i];
3429 BUG_ON(len >= IGB_MAX_DATA_PER_TXD);
3430 buffer_info->length = len;
3431 /* set time_stamp *before* dma to help avoid a possible race */
3432 buffer_info->time_stamp = jiffies;
0e014cb1 3433 buffer_info->next_to_watch = i;
042a53a9 3434 buffer_info->dma = skb_shinfo(skb)->dma_head;
9d5c8243
AK
3435
3436 for (f = 0; f < skb_shinfo(skb)->nr_frags; f++) {
3437 struct skb_frag_struct *frag;
3438
65689fef
AD
3439 i++;
3440 if (i == tx_ring->count)
3441 i = 0;
3442
9d5c8243
AK
3443 frag = &skb_shinfo(skb)->frags[f];
3444 len = frag->size;
3445
3446 buffer_info = &tx_ring->buffer_info[i];
3447 BUG_ON(len >= IGB_MAX_DATA_PER_TXD);
3448 buffer_info->length = len;
3449 buffer_info->time_stamp = jiffies;
0e014cb1 3450 buffer_info->next_to_watch = i;
65689fef 3451 buffer_info->dma = map[count];
9d5c8243 3452 count++;
9d5c8243
AK
3453 }
3454
9d5c8243 3455 tx_ring->buffer_info[i].skb = skb;
0e014cb1 3456 tx_ring->buffer_info[first].next_to_watch = i;
9d5c8243 3457
042a53a9 3458 return count + 1;
9d5c8243
AK
3459}
3460
85ad76b2 3461static inline void igb_tx_queue_adv(struct igb_ring *tx_ring,
9d5c8243
AK
3462 int tx_flags, int count, u32 paylen,
3463 u8 hdr_len)
3464{
3465 union e1000_adv_tx_desc *tx_desc = NULL;
3466 struct igb_buffer *buffer_info;
3467 u32 olinfo_status = 0, cmd_type_len;
3468 unsigned int i;
3469
3470 cmd_type_len = (E1000_ADVTXD_DTYP_DATA | E1000_ADVTXD_DCMD_IFCS |
3471 E1000_ADVTXD_DCMD_DEXT);
3472
3473 if (tx_flags & IGB_TX_FLAGS_VLAN)
3474 cmd_type_len |= E1000_ADVTXD_DCMD_VLE;
3475
33af6bcc
PO
3476 if (tx_flags & IGB_TX_FLAGS_TSTAMP)
3477 cmd_type_len |= E1000_ADVTXD_MAC_TSTAMP;
3478
9d5c8243
AK
3479 if (tx_flags & IGB_TX_FLAGS_TSO) {
3480 cmd_type_len |= E1000_ADVTXD_DCMD_TSE;
3481
3482 /* insert tcp checksum */
3483 olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
3484
3485 /* insert ip checksum */
3486 if (tx_flags & IGB_TX_FLAGS_IPV4)
3487 olinfo_status |= E1000_TXD_POPTS_IXSM << 8;
3488
3489 } else if (tx_flags & IGB_TX_FLAGS_CSUM) {
3490 olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
3491 }
3492
85ad76b2
AD
3493 if ((tx_ring->flags & IGB_RING_FLAG_TX_CTX_IDX) &&
3494 (tx_flags & (IGB_TX_FLAGS_CSUM |
3495 IGB_TX_FLAGS_TSO |
7dfc16fa 3496 IGB_TX_FLAGS_VLAN)))
85ad76b2 3497 olinfo_status |= tx_ring->reg_idx << 4;
9d5c8243
AK
3498
3499 olinfo_status |= ((paylen - hdr_len) << E1000_ADVTXD_PAYLEN_SHIFT);
3500
3501 i = tx_ring->next_to_use;
3502 while (count--) {
3503 buffer_info = &tx_ring->buffer_info[i];
3504 tx_desc = E1000_TX_DESC_ADV(*tx_ring, i);
3505 tx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
3506 tx_desc->read.cmd_type_len =
3507 cpu_to_le32(cmd_type_len | buffer_info->length);
3508 tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
3509 i++;
3510 if (i == tx_ring->count)
3511 i = 0;
3512 }
3513
85ad76b2 3514 tx_desc->read.cmd_type_len |= cpu_to_le32(IGB_ADVTXD_DCMD);
9d5c8243
AK
3515 /* Force memory writes to complete before letting h/w
3516 * know there are new descriptors to fetch. (Only
3517 * applicable for weak-ordered memory model archs,
3518 * such as IA-64). */
3519 wmb();
3520
3521 tx_ring->next_to_use = i;
fce99e34 3522 writel(i, tx_ring->tail);
9d5c8243
AK
3523 /* we need this if more than one processor can write to our tail
3524 * at a time, it syncronizes IO on IA64/Altix systems */
3525 mmiowb();
3526}
3527
e694e964 3528static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, int size)
9d5c8243 3529{
e694e964
AD
3530 struct net_device *netdev = tx_ring->netdev;
3531
661086df 3532 netif_stop_subqueue(netdev, tx_ring->queue_index);
661086df 3533
9d5c8243
AK
3534 /* Herbert's original patch had:
3535 * smp_mb__after_netif_stop_queue();
3536 * but since that doesn't exist yet, just open code it. */
3537 smp_mb();
3538
3539 /* We need to check again in a case another CPU has just
3540 * made room available. */
c493ea45 3541 if (igb_desc_unused(tx_ring) < size)
9d5c8243
AK
3542 return -EBUSY;
3543
3544 /* A reprieve! */
661086df 3545 netif_wake_subqueue(netdev, tx_ring->queue_index);
04a5fcaa 3546 tx_ring->tx_stats.restart_queue++;
9d5c8243
AK
3547 return 0;
3548}
3549
e694e964 3550static int igb_maybe_stop_tx(struct igb_ring *tx_ring, int size)
9d5c8243 3551{
c493ea45 3552 if (igb_desc_unused(tx_ring) >= size)
9d5c8243 3553 return 0;
e694e964 3554 return __igb_maybe_stop_tx(tx_ring, size);
9d5c8243
AK
3555}
3556
b1a436c3
AD
3557netdev_tx_t igb_xmit_frame_ring_adv(struct sk_buff *skb,
3558 struct igb_ring *tx_ring)
9d5c8243 3559{
e694e964 3560 struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
0e014cb1 3561 unsigned int first;
9d5c8243 3562 unsigned int tx_flags = 0;
9d5c8243 3563 u8 hdr_len = 0;
65689fef 3564 int count = 0;
9d5c8243 3565 int tso = 0;
c5b9bd5e 3566 union skb_shared_tx *shtx = skb_tx(skb);
9d5c8243 3567
9d5c8243
AK
3568 /* need: 1 descriptor per page,
3569 * + 2 desc gap to keep tail from touching head,
3570 * + 1 desc for skb->data,
3571 * + 1 desc for context descriptor,
3572 * otherwise try next time */
e694e964 3573 if (igb_maybe_stop_tx(tx_ring, skb_shinfo(skb)->nr_frags + 4)) {
9d5c8243 3574 /* this is a hard error */
9d5c8243
AK
3575 return NETDEV_TX_BUSY;
3576 }
33af6bcc 3577
33af6bcc
PO
3578 if (unlikely(shtx->hardware)) {
3579 shtx->in_progress = 1;
3580 tx_flags |= IGB_TX_FLAGS_TSTAMP;
33af6bcc 3581 }
9d5c8243
AK
3582
3583 if (adapter->vlgrp && vlan_tx_tag_present(skb)) {
3584 tx_flags |= IGB_TX_FLAGS_VLAN;
3585 tx_flags |= (vlan_tx_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
3586 }
3587
661086df
PWJ
3588 if (skb->protocol == htons(ETH_P_IP))
3589 tx_flags |= IGB_TX_FLAGS_IPV4;
3590
0e014cb1 3591 first = tx_ring->next_to_use;
85ad76b2
AD
3592 if (skb_is_gso(skb)) {
3593 tso = igb_tso_adv(tx_ring, skb, tx_flags, &hdr_len);
3594 if (tso < 0) {
3595 dev_kfree_skb_any(skb);
3596 return NETDEV_TX_OK;
3597 }
9d5c8243
AK
3598 }
3599
3600 if (tso)
3601 tx_flags |= IGB_TX_FLAGS_TSO;
85ad76b2 3602 else if (igb_tx_csum_adv(tx_ring, skb, tx_flags) &&
bc1cbd34
AD
3603 (skb->ip_summed == CHECKSUM_PARTIAL))
3604 tx_flags |= IGB_TX_FLAGS_CSUM;
9d5c8243 3605
65689fef
AD
3606 /*
3607 * count reflects descriptors mapped, if 0 then mapping error
3608 * has occured and we need to rewind the descriptor queue
3609 */
80785298 3610 count = igb_tx_map_adv(tx_ring, skb, first);
65689fef 3611
85ad76b2 3612 if (!count) {
65689fef
AD
3613 dev_kfree_skb_any(skb);
3614 tx_ring->buffer_info[first].time_stamp = 0;
3615 tx_ring->next_to_use = first;
85ad76b2 3616 return NETDEV_TX_OK;
65689fef 3617 }
9d5c8243 3618
85ad76b2
AD
3619 igb_tx_queue_adv(tx_ring, tx_flags, count, skb->len, hdr_len);
3620
3621 /* Make sure there is space in the ring for the next send. */
e694e964 3622 igb_maybe_stop_tx(tx_ring, MAX_SKB_FRAGS + 4);
85ad76b2 3623
9d5c8243
AK
3624 return NETDEV_TX_OK;
3625}
3626
3b29a56d
SH
3627static netdev_tx_t igb_xmit_frame_adv(struct sk_buff *skb,
3628 struct net_device *netdev)
9d5c8243
AK
3629{
3630 struct igb_adapter *adapter = netdev_priv(netdev);
661086df 3631 struct igb_ring *tx_ring;
661086df 3632 int r_idx = 0;
b1a436c3
AD
3633
3634 if (test_bit(__IGB_DOWN, &adapter->state)) {
3635 dev_kfree_skb_any(skb);
3636 return NETDEV_TX_OK;
3637 }
3638
3639 if (skb->len <= 0) {
3640 dev_kfree_skb_any(skb);
3641 return NETDEV_TX_OK;
3642 }
3643
1bfaf07b 3644 r_idx = skb->queue_mapping & (IGB_ABS_MAX_TX_QUEUES - 1);
661086df 3645 tx_ring = adapter->multi_tx_table[r_idx];
9d5c8243
AK
3646
3647 /* This goes back to the question of how to logically map a tx queue
3648 * to a flow. Right now, performance is impacted slightly negatively
3649 * if using multiple tx queues. If the stack breaks away from a
3650 * single qdisc implementation, we can look at this again. */
e694e964 3651 return igb_xmit_frame_ring_adv(skb, tx_ring);
9d5c8243
AK
3652}
3653
3654/**
3655 * igb_tx_timeout - Respond to a Tx Hang
3656 * @netdev: network interface device structure
3657 **/
3658static void igb_tx_timeout(struct net_device *netdev)
3659{
3660 struct igb_adapter *adapter = netdev_priv(netdev);
3661 struct e1000_hw *hw = &adapter->hw;
3662
3663 /* Do the reset outside of interrupt context */
3664 adapter->tx_timeout_count++;
3665 schedule_work(&adapter->reset_task);
265de409
AD
3666 wr32(E1000_EICS,
3667 (adapter->eims_enable_mask & ~adapter->eims_other));
9d5c8243
AK
3668}
3669
3670static void igb_reset_task(struct work_struct *work)
3671{
3672 struct igb_adapter *adapter;
3673 adapter = container_of(work, struct igb_adapter, reset_task);
3674
3675 igb_reinit_locked(adapter);
3676}
3677
3678/**
3679 * igb_get_stats - Get System Network Statistics
3680 * @netdev: network interface device structure
3681 *
3682 * Returns the address of the device statistics structure.
3683 * The statistics are actually updated from the timer callback.
3684 **/
73cd78f1 3685static struct net_device_stats *igb_get_stats(struct net_device *netdev)
9d5c8243 3686{
9d5c8243 3687 /* only return the current stats */
8d24e933 3688 return &netdev->stats;
9d5c8243
AK
3689}
3690
3691/**
3692 * igb_change_mtu - Change the Maximum Transfer Unit
3693 * @netdev: network interface device structure
3694 * @new_mtu: new value for maximum frame size
3695 *
3696 * Returns 0 on success, negative on failure
3697 **/
3698static int igb_change_mtu(struct net_device *netdev, int new_mtu)
3699{
3700 struct igb_adapter *adapter = netdev_priv(netdev);
3701 int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
4c844851 3702 u32 rx_buffer_len, i;
9d5c8243
AK
3703
3704 if ((max_frame < ETH_ZLEN + ETH_FCS_LEN) ||
3705 (max_frame > MAX_JUMBO_FRAME_SIZE)) {
3706 dev_err(&adapter->pdev->dev, "Invalid MTU setting\n");
3707 return -EINVAL;
3708 }
3709
9d5c8243
AK
3710 if (max_frame > MAX_STD_JUMBO_FRAME_SIZE) {
3711 dev_err(&adapter->pdev->dev, "MTU > 9216 not supported.\n");
3712 return -EINVAL;
3713 }
3714
3715 while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
3716 msleep(1);
73cd78f1 3717
9d5c8243
AK
3718 /* igb_down has a dependency on max_frame_size */
3719 adapter->max_frame_size = max_frame;
9d5c8243
AK
3720 /* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
3721 * means we reserve 2 more, this pushes us to allocate from the next
3722 * larger slab size.
3723 * i.e. RXBUFFER_2048 --> size-4096 slab
3724 */
3725
7d95b717 3726 if (max_frame <= IGB_RXBUFFER_1024)
4c844851 3727 rx_buffer_len = IGB_RXBUFFER_1024;
6ec43fe6 3728 else if (max_frame <= MAXIMUM_ETHERNET_VLAN_SIZE)
4c844851 3729 rx_buffer_len = MAXIMUM_ETHERNET_VLAN_SIZE;
6ec43fe6 3730 else
4c844851
AD
3731 rx_buffer_len = IGB_RXBUFFER_128;
3732
3733 if (netif_running(netdev))
3734 igb_down(adapter);
9d5c8243
AK
3735
3736 dev_info(&adapter->pdev->dev, "changing MTU from %d to %d\n",
3737 netdev->mtu, new_mtu);
3738 netdev->mtu = new_mtu;
3739
4c844851
AD
3740 for (i = 0; i < adapter->num_rx_queues; i++)
3741 adapter->rx_ring[i].rx_buffer_len = rx_buffer_len;
3742
9d5c8243
AK
3743 if (netif_running(netdev))
3744 igb_up(adapter);
3745 else
3746 igb_reset(adapter);
3747
3748 clear_bit(__IGB_RESETTING, &adapter->state);
3749
3750 return 0;
3751}
3752
3753/**
3754 * igb_update_stats - Update the board statistics counters
3755 * @adapter: board private structure
3756 **/
3757
3758void igb_update_stats(struct igb_adapter *adapter)
3759{
8d24e933 3760 struct net_device *netdev = adapter->netdev;
9d5c8243
AK
3761 struct e1000_hw *hw = &adapter->hw;
3762 struct pci_dev *pdev = adapter->pdev;
3763 u16 phy_tmp;
3764
3765#define PHY_IDLE_ERROR_COUNT_MASK 0x00FF
3766
3767 /*
3768 * Prevent stats update while adapter is being reset, or if the pci
3769 * connection is down.
3770 */
3771 if (adapter->link_speed == 0)
3772 return;
3773 if (pci_channel_offline(pdev))
3774 return;
3775
3776 adapter->stats.crcerrs += rd32(E1000_CRCERRS);
3777 adapter->stats.gprc += rd32(E1000_GPRC);
3778 adapter->stats.gorc += rd32(E1000_GORCL);
3779 rd32(E1000_GORCH); /* clear GORCL */
3780 adapter->stats.bprc += rd32(E1000_BPRC);
3781 adapter->stats.mprc += rd32(E1000_MPRC);
3782 adapter->stats.roc += rd32(E1000_ROC);
3783
3784 adapter->stats.prc64 += rd32(E1000_PRC64);
3785 adapter->stats.prc127 += rd32(E1000_PRC127);
3786 adapter->stats.prc255 += rd32(E1000_PRC255);
3787 adapter->stats.prc511 += rd32(E1000_PRC511);
3788 adapter->stats.prc1023 += rd32(E1000_PRC1023);
3789 adapter->stats.prc1522 += rd32(E1000_PRC1522);
3790 adapter->stats.symerrs += rd32(E1000_SYMERRS);
3791 adapter->stats.sec += rd32(E1000_SEC);
3792
3793 adapter->stats.mpc += rd32(E1000_MPC);
3794 adapter->stats.scc += rd32(E1000_SCC);
3795 adapter->stats.ecol += rd32(E1000_ECOL);
3796 adapter->stats.mcc += rd32(E1000_MCC);
3797 adapter->stats.latecol += rd32(E1000_LATECOL);
3798 adapter->stats.dc += rd32(E1000_DC);
3799 adapter->stats.rlec += rd32(E1000_RLEC);
3800 adapter->stats.xonrxc += rd32(E1000_XONRXC);
3801 adapter->stats.xontxc += rd32(E1000_XONTXC);
3802 adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
3803 adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
3804 adapter->stats.fcruc += rd32(E1000_FCRUC);
3805 adapter->stats.gptc += rd32(E1000_GPTC);
3806 adapter->stats.gotc += rd32(E1000_GOTCL);
3807 rd32(E1000_GOTCH); /* clear GOTCL */
3808 adapter->stats.rnbc += rd32(E1000_RNBC);
3809 adapter->stats.ruc += rd32(E1000_RUC);
3810 adapter->stats.rfc += rd32(E1000_RFC);
3811 adapter->stats.rjc += rd32(E1000_RJC);
3812 adapter->stats.tor += rd32(E1000_TORH);
3813 adapter->stats.tot += rd32(E1000_TOTH);
3814 adapter->stats.tpr += rd32(E1000_TPR);
3815
3816 adapter->stats.ptc64 += rd32(E1000_PTC64);
3817 adapter->stats.ptc127 += rd32(E1000_PTC127);
3818 adapter->stats.ptc255 += rd32(E1000_PTC255);
3819 adapter->stats.ptc511 += rd32(E1000_PTC511);
3820 adapter->stats.ptc1023 += rd32(E1000_PTC1023);
3821 adapter->stats.ptc1522 += rd32(E1000_PTC1522);
3822
3823 adapter->stats.mptc += rd32(E1000_MPTC);
3824 adapter->stats.bptc += rd32(E1000_BPTC);
3825
3826 /* used for adaptive IFS */
3827
3828 hw->mac.tx_packet_delta = rd32(E1000_TPT);
3829 adapter->stats.tpt += hw->mac.tx_packet_delta;
3830 hw->mac.collision_delta = rd32(E1000_COLC);
3831 adapter->stats.colc += hw->mac.collision_delta;
3832
3833 adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
3834 adapter->stats.rxerrc += rd32(E1000_RXERRC);
3835 adapter->stats.tncrs += rd32(E1000_TNCRS);
3836 adapter->stats.tsctc += rd32(E1000_TSCTC);
3837 adapter->stats.tsctfc += rd32(E1000_TSCTFC);
3838
3839 adapter->stats.iac += rd32(E1000_IAC);
3840 adapter->stats.icrxoc += rd32(E1000_ICRXOC);
3841 adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
3842 adapter->stats.icrxatc += rd32(E1000_ICRXATC);
3843 adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
3844 adapter->stats.ictxatc += rd32(E1000_ICTXATC);
3845 adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
3846 adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
3847 adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
3848
3849 /* Fill out the OS statistics structure */
8d24e933
AK
3850 netdev->stats.multicast = adapter->stats.mprc;
3851 netdev->stats.collisions = adapter->stats.colc;
9d5c8243
AK
3852
3853 /* Rx Errors */
3854
8c0ab70a
JDB
3855 if (hw->mac.type != e1000_82575) {
3856 u32 rqdpc_tmp;
3ea73afa 3857 u64 rqdpc_total = 0;
8c0ab70a
JDB
3858 int i;
3859 /* Read out drops stats per RX queue. Notice RQDPC (Receive
3860 * Queue Drop Packet Count) stats only gets incremented, if
3861 * the DROP_EN but it set (in the SRRCTL register for that
3862 * queue). If DROP_EN bit is NOT set, then the some what
3863 * equivalent count is stored in RNBC (not per queue basis).
3864 * Also note the drop count is due to lack of available
3865 * descriptors.
3866 */
3867 for (i = 0; i < adapter->num_rx_queues; i++) {
3868 rqdpc_tmp = rd32(E1000_RQDPC(i)) & 0xFFF;
3869 adapter->rx_ring[i].rx_stats.drops += rqdpc_tmp;
3ea73afa 3870 rqdpc_total += adapter->rx_ring[i].rx_stats.drops;
8c0ab70a 3871 }
8d24e933 3872 netdev->stats.rx_fifo_errors = rqdpc_total;
8c0ab70a
JDB
3873 }
3874
3ea73afa
JDB
3875 /* Note RNBC (Receive No Buffers Count) is an not an exact
3876 * drop count as the hardware FIFO might save the day. Thats
3877 * one of the reason for saving it in rx_fifo_errors, as its
3878 * potentially not a true drop.
3879 */
8d24e933 3880 netdev->stats.rx_fifo_errors += adapter->stats.rnbc;
3ea73afa 3881
9d5c8243 3882 /* RLEC on some newer hardware can be incorrect so build
8c0ab70a 3883 * our own version based on RUC and ROC */
8d24e933 3884 netdev->stats.rx_errors = adapter->stats.rxerrc +
9d5c8243
AK
3885 adapter->stats.crcerrs + adapter->stats.algnerrc +
3886 adapter->stats.ruc + adapter->stats.roc +
3887 adapter->stats.cexterr;
8d24e933 3888 netdev->stats.rx_length_errors = adapter->stats.ruc +
9d5c8243 3889 adapter->stats.roc;
8d24e933
AK
3890 netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
3891 netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
3892 netdev->stats.rx_missed_errors = adapter->stats.mpc;
9d5c8243
AK
3893
3894 /* Tx Errors */
8d24e933 3895 netdev->stats.tx_errors = adapter->stats.ecol +
9d5c8243 3896 adapter->stats.latecol;
8d24e933
AK
3897 netdev->stats.tx_aborted_errors = adapter->stats.ecol;
3898 netdev->stats.tx_window_errors = adapter->stats.latecol;
3899 netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
9d5c8243
AK
3900
3901 /* Tx Dropped needs to be maintained elsewhere */
3902
3903 /* Phy Stats */
3904 if (hw->phy.media_type == e1000_media_type_copper) {
3905 if ((adapter->link_speed == SPEED_1000) &&
73cd78f1 3906 (!igb_read_phy_reg(hw, PHY_1000T_STATUS, &phy_tmp))) {
9d5c8243
AK
3907 phy_tmp &= PHY_IDLE_ERROR_COUNT_MASK;
3908 adapter->phy_stats.idle_errors += phy_tmp;
3909 }
3910 }
3911
3912 /* Management Stats */
3913 adapter->stats.mgptc += rd32(E1000_MGTPTC);
3914 adapter->stats.mgprc += rd32(E1000_MGTPRC);
3915 adapter->stats.mgpdc += rd32(E1000_MGTPDC);
3916}
3917
9d5c8243
AK
3918static irqreturn_t igb_msix_other(int irq, void *data)
3919{
047e0030 3920 struct igb_adapter *adapter = data;
9d5c8243 3921 struct e1000_hw *hw = &adapter->hw;
844290e5 3922 u32 icr = rd32(E1000_ICR);
844290e5 3923 /* reading ICR causes bit 31 of EICR to be cleared */
dda0e083 3924
047e0030 3925 if (icr & E1000_ICR_DOUTSYNC) {
dda0e083
AD
3926 /* HW is reporting DMA is out of sync */
3927 adapter->stats.doosync++;
3928 }
eebbbdba 3929
4ae196df
AD
3930 /* Check for a mailbox event */
3931 if (icr & E1000_ICR_VMMB)
3932 igb_msg_task(adapter);
3933
3934 if (icr & E1000_ICR_LSC) {
3935 hw->mac.get_link_status = 1;
3936 /* guard against interrupt when we're going down */
3937 if (!test_bit(__IGB_DOWN, &adapter->state))
3938 mod_timer(&adapter->watchdog_timer, jiffies + 1);
3939 }
3940
3941 wr32(E1000_IMS, E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_VMMB);
844290e5 3942 wr32(E1000_EIMS, adapter->eims_other);
9d5c8243
AK
3943
3944 return IRQ_HANDLED;
3945}
3946
047e0030 3947static void igb_write_itr(struct igb_q_vector *q_vector)
9d5c8243 3948{
047e0030 3949 u32 itr_val = q_vector->itr_val & 0x7FFC;
9d5c8243 3950
047e0030
AD
3951 if (!q_vector->set_itr)
3952 return;
73cd78f1 3953
047e0030
AD
3954 if (!itr_val)
3955 itr_val = 0x4;
661086df 3956
047e0030
AD
3957 if (q_vector->itr_shift)
3958 itr_val |= itr_val << q_vector->itr_shift;
661086df 3959 else
047e0030 3960 itr_val |= 0x8000000;
661086df 3961
047e0030
AD
3962 writel(itr_val, q_vector->itr_register);
3963 q_vector->set_itr = 0;
6eb5a7f1
AD
3964}
3965
047e0030 3966static irqreturn_t igb_msix_ring(int irq, void *data)
9d5c8243 3967{
047e0030 3968 struct igb_q_vector *q_vector = data;
9d5c8243 3969
047e0030
AD
3970 /* Write the ITR value calculated from the previous interrupt. */
3971 igb_write_itr(q_vector);
9d5c8243 3972
047e0030 3973 napi_schedule(&q_vector->napi);
844290e5 3974
047e0030 3975 return IRQ_HANDLED;
fe4506b6
JC
3976}
3977
421e02f0 3978#ifdef CONFIG_IGB_DCA
047e0030 3979static void igb_update_dca(struct igb_q_vector *q_vector)
fe4506b6 3980{
047e0030 3981 struct igb_adapter *adapter = q_vector->adapter;
fe4506b6
JC
3982 struct e1000_hw *hw = &adapter->hw;
3983 int cpu = get_cpu();
fe4506b6 3984
047e0030
AD
3985 if (q_vector->cpu == cpu)
3986 goto out_no_update;
3987
3988 if (q_vector->tx_ring) {
3989 int q = q_vector->tx_ring->reg_idx;
3990 u32 dca_txctrl = rd32(E1000_DCA_TXCTRL(q));
3991 if (hw->mac.type == e1000_82575) {
3992 dca_txctrl &= ~E1000_DCA_TXCTRL_CPUID_MASK;
3993 dca_txctrl |= dca3_get_tag(&adapter->pdev->dev, cpu);
2d064c06 3994 } else {
047e0030
AD
3995 dca_txctrl &= ~E1000_DCA_TXCTRL_CPUID_MASK_82576;
3996 dca_txctrl |= dca3_get_tag(&adapter->pdev->dev, cpu) <<
3997 E1000_DCA_TXCTRL_CPUID_SHIFT;
3998 }
3999 dca_txctrl |= E1000_DCA_TXCTRL_DESC_DCA_EN;
4000 wr32(E1000_DCA_TXCTRL(q), dca_txctrl);
4001 }
4002 if (q_vector->rx_ring) {
4003 int q = q_vector->rx_ring->reg_idx;
4004 u32 dca_rxctrl = rd32(E1000_DCA_RXCTRL(q));
4005 if (hw->mac.type == e1000_82575) {
2d064c06 4006 dca_rxctrl &= ~E1000_DCA_RXCTRL_CPUID_MASK;
92be7917 4007 dca_rxctrl |= dca3_get_tag(&adapter->pdev->dev, cpu);
047e0030
AD
4008 } else {
4009 dca_rxctrl &= ~E1000_DCA_RXCTRL_CPUID_MASK_82576;
4010 dca_rxctrl |= dca3_get_tag(&adapter->pdev->dev, cpu) <<
4011 E1000_DCA_RXCTRL_CPUID_SHIFT;
2d064c06 4012 }
fe4506b6
JC
4013 dca_rxctrl |= E1000_DCA_RXCTRL_DESC_DCA_EN;
4014 dca_rxctrl |= E1000_DCA_RXCTRL_HEAD_DCA_EN;
4015 dca_rxctrl |= E1000_DCA_RXCTRL_DATA_DCA_EN;
4016 wr32(E1000_DCA_RXCTRL(q), dca_rxctrl);
fe4506b6 4017 }
047e0030
AD
4018 q_vector->cpu = cpu;
4019out_no_update:
fe4506b6
JC
4020 put_cpu();
4021}
4022
4023static void igb_setup_dca(struct igb_adapter *adapter)
4024{
7e0e99ef 4025 struct e1000_hw *hw = &adapter->hw;
fe4506b6
JC
4026 int i;
4027
7dfc16fa 4028 if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
fe4506b6
JC
4029 return;
4030
7e0e99ef
AD
4031 /* Always use CB2 mode, difference is masked in the CB driver. */
4032 wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
4033
047e0030
AD
4034 for (i = 0; i < adapter->num_q_vectors; i++) {
4035 struct igb_q_vector *q_vector = adapter->q_vector[i];
4036 q_vector->cpu = -1;
4037 igb_update_dca(q_vector);
fe4506b6
JC
4038 }
4039}
4040
4041static int __igb_notify_dca(struct device *dev, void *data)
4042{
4043 struct net_device *netdev = dev_get_drvdata(dev);
4044 struct igb_adapter *adapter = netdev_priv(netdev);
4045 struct e1000_hw *hw = &adapter->hw;
4046 unsigned long event = *(unsigned long *)data;
4047
4048 switch (event) {
4049 case DCA_PROVIDER_ADD:
4050 /* if already enabled, don't do it again */
7dfc16fa 4051 if (adapter->flags & IGB_FLAG_DCA_ENABLED)
fe4506b6 4052 break;
fe4506b6
JC
4053 /* Always use CB2 mode, difference is masked
4054 * in the CB driver. */
cbd347ad 4055 wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
fe4506b6 4056 if (dca_add_requester(dev) == 0) {
bbd98fe4 4057 adapter->flags |= IGB_FLAG_DCA_ENABLED;
fe4506b6
JC
4058 dev_info(&adapter->pdev->dev, "DCA enabled\n");
4059 igb_setup_dca(adapter);
4060 break;
4061 }
4062 /* Fall Through since DCA is disabled. */
4063 case DCA_PROVIDER_REMOVE:
7dfc16fa 4064 if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
fe4506b6 4065 /* without this a class_device is left
047e0030 4066 * hanging around in the sysfs model */
fe4506b6
JC
4067 dca_remove_requester(dev);
4068 dev_info(&adapter->pdev->dev, "DCA disabled\n");
7dfc16fa 4069 adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
cbd347ad 4070 wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
fe4506b6
JC
4071 }
4072 break;
4073 }
bbd98fe4 4074
fe4506b6 4075 return 0;
9d5c8243
AK
4076}
4077
fe4506b6
JC
4078static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
4079 void *p)
4080{
4081 int ret_val;
4082
4083 ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
4084 __igb_notify_dca);
4085
4086 return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
4087}
421e02f0 4088#endif /* CONFIG_IGB_DCA */
9d5c8243 4089
4ae196df
AD
4090static void igb_ping_all_vfs(struct igb_adapter *adapter)
4091{
4092 struct e1000_hw *hw = &adapter->hw;
4093 u32 ping;
4094 int i;
4095
4096 for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
4097 ping = E1000_PF_CONTROL_MSG;
4098 if (adapter->vf_data[i].clear_to_send)
4099 ping |= E1000_VT_MSGTYPE_CTS;
4100 igb_write_mbx(hw, &ping, 1, i);
4101 }
4102}
4103
4104static int igb_set_vf_multicasts(struct igb_adapter *adapter,
4105 u32 *msgbuf, u32 vf)
4106{
4107 int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
4108 u16 *hash_list = (u16 *)&msgbuf[1];
4109 struct vf_data_storage *vf_data = &adapter->vf_data[vf];
4110 int i;
4111
4112 /* only up to 30 hash values supported */
4113 if (n > 30)
4114 n = 30;
4115
4116 /* salt away the number of multi cast addresses assigned
4117 * to this VF for later use to restore when the PF multi cast
4118 * list changes
4119 */
4120 vf_data->num_vf_mc_hashes = n;
4121
4122 /* VFs are limited to using the MTA hash table for their multicast
4123 * addresses */
4124 for (i = 0; i < n; i++)
a419aef8 4125 vf_data->vf_mc_hashes[i] = hash_list[i];
4ae196df
AD
4126
4127 /* Flush and reset the mta with the new values */
ff41f8dc 4128 igb_set_rx_mode(adapter->netdev);
4ae196df
AD
4129
4130 return 0;
4131}
4132
4133static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
4134{
4135 struct e1000_hw *hw = &adapter->hw;
4136 struct vf_data_storage *vf_data;
4137 int i, j;
4138
4139 for (i = 0; i < adapter->vfs_allocated_count; i++) {
4140 vf_data = &adapter->vf_data[i];
75f4f382 4141 for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
4ae196df
AD
4142 igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
4143 }
4144}
4145
4146static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
4147{
4148 struct e1000_hw *hw = &adapter->hw;
4149 u32 pool_mask, reg, vid;
4150 int i;
4151
4152 pool_mask = 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
4153
4154 /* Find the vlan filter for this id */
4155 for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
4156 reg = rd32(E1000_VLVF(i));
4157
4158 /* remove the vf from the pool */
4159 reg &= ~pool_mask;
4160
4161 /* if pool is empty then remove entry from vfta */
4162 if (!(reg & E1000_VLVF_POOLSEL_MASK) &&
4163 (reg & E1000_VLVF_VLANID_ENABLE)) {
4164 reg = 0;
4165 vid = reg & E1000_VLVF_VLANID_MASK;
4166 igb_vfta_set(hw, vid, false);
4167 }
4168
4169 wr32(E1000_VLVF(i), reg);
4170 }
ae641bdc
AD
4171
4172 adapter->vf_data[vf].vlans_enabled = 0;
4ae196df
AD
4173}
4174
4175static s32 igb_vlvf_set(struct igb_adapter *adapter, u32 vid, bool add, u32 vf)
4176{
4177 struct e1000_hw *hw = &adapter->hw;
4178 u32 reg, i;
4179
4180 /* It is an error to call this function when VFs are not enabled */
4181 if (!adapter->vfs_allocated_count)
4182 return -1;
4183
4184 /* Find the vlan filter for this id */
4185 for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
4186 reg = rd32(E1000_VLVF(i));
4187 if ((reg & E1000_VLVF_VLANID_ENABLE) &&
4188 vid == (reg & E1000_VLVF_VLANID_MASK))
4189 break;
4190 }
4191
4192 if (add) {
4193 if (i == E1000_VLVF_ARRAY_SIZE) {
4194 /* Did not find a matching VLAN ID entry that was
4195 * enabled. Search for a free filter entry, i.e.
4196 * one without the enable bit set
4197 */
4198 for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
4199 reg = rd32(E1000_VLVF(i));
4200 if (!(reg & E1000_VLVF_VLANID_ENABLE))
4201 break;
4202 }
4203 }
4204 if (i < E1000_VLVF_ARRAY_SIZE) {
4205 /* Found an enabled/available entry */
4206 reg |= 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
4207
4208 /* if !enabled we need to set this up in vfta */
4209 if (!(reg & E1000_VLVF_VLANID_ENABLE)) {
cad6d05f
AD
4210 /* add VID to filter table, if bit already set
4211 * PF must have added it outside of table */
4212 if (igb_vfta_set(hw, vid, true))
4213 reg |= 1 << (E1000_VLVF_POOLSEL_SHIFT +
4214 adapter->vfs_allocated_count);
4ae196df
AD
4215 reg |= E1000_VLVF_VLANID_ENABLE;
4216 }
cad6d05f
AD
4217 reg &= ~E1000_VLVF_VLANID_MASK;
4218 reg |= vid;
4ae196df
AD
4219
4220 wr32(E1000_VLVF(i), reg);
ae641bdc
AD
4221
4222 /* do not modify RLPML for PF devices */
4223 if (vf >= adapter->vfs_allocated_count)
4224 return 0;
4225
4226 if (!adapter->vf_data[vf].vlans_enabled) {
4227 u32 size;
4228 reg = rd32(E1000_VMOLR(vf));
4229 size = reg & E1000_VMOLR_RLPML_MASK;
4230 size += 4;
4231 reg &= ~E1000_VMOLR_RLPML_MASK;
4232 reg |= size;
4233 wr32(E1000_VMOLR(vf), reg);
4234 }
4235 adapter->vf_data[vf].vlans_enabled++;
4236
4ae196df
AD
4237 return 0;
4238 }
4239 } else {
4240 if (i < E1000_VLVF_ARRAY_SIZE) {
4241 /* remove vf from the pool */
4242 reg &= ~(1 << (E1000_VLVF_POOLSEL_SHIFT + vf));
4243 /* if pool is empty then remove entry from vfta */
4244 if (!(reg & E1000_VLVF_POOLSEL_MASK)) {
4245 reg = 0;
4246 igb_vfta_set(hw, vid, false);
4247 }
4248 wr32(E1000_VLVF(i), reg);
ae641bdc
AD
4249
4250 /* do not modify RLPML for PF devices */
4251 if (vf >= adapter->vfs_allocated_count)
4252 return 0;
4253
4254 adapter->vf_data[vf].vlans_enabled--;
4255 if (!adapter->vf_data[vf].vlans_enabled) {
4256 u32 size;
4257 reg = rd32(E1000_VMOLR(vf));
4258 size = reg & E1000_VMOLR_RLPML_MASK;
4259 size -= 4;
4260 reg &= ~E1000_VMOLR_RLPML_MASK;
4261 reg |= size;
4262 wr32(E1000_VMOLR(vf), reg);
4263 }
4ae196df
AD
4264 return 0;
4265 }
4266 }
4267 return -1;
4268}
4269
4270static int igb_set_vf_vlan(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
4271{
4272 int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
4273 int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
4274
4275 return igb_vlvf_set(adapter, vid, add, vf);
4276}
4277
4278static inline void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
4279{
4280 struct e1000_hw *hw = &adapter->hw;
4281
4282 /* disable mailbox functionality for vf */
4283 adapter->vf_data[vf].clear_to_send = false;
4284
4285 /* reset offloads to defaults */
4286 igb_set_vmolr(hw, vf);
4287
4288 /* reset vlans for device */
4289 igb_clear_vf_vfta(adapter, vf);
4290
4291 /* reset multicast table array for vf */
4292 adapter->vf_data[vf].num_vf_mc_hashes = 0;
4293
4294 /* Flush and reset the mta with the new values */
ff41f8dc 4295 igb_set_rx_mode(adapter->netdev);
4ae196df
AD
4296}
4297
4298static inline void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
4299{
4300 struct e1000_hw *hw = &adapter->hw;
4301 unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
ff41f8dc 4302 int rar_entry = hw->mac.rar_entry_count - (vf + 1);
4ae196df
AD
4303 u32 reg, msgbuf[3];
4304 u8 *addr = (u8 *)(&msgbuf[1]);
4305
4306 /* process all the same items cleared in a function level reset */
4307 igb_vf_reset_event(adapter, vf);
4308
4309 /* set vf mac address */
26ad9178 4310 igb_rar_set_qsel(adapter, vf_mac, rar_entry, vf);
4ae196df
AD
4311
4312 /* enable transmit and receive for vf */
4313 reg = rd32(E1000_VFTE);
4314 wr32(E1000_VFTE, reg | (1 << vf));
4315 reg = rd32(E1000_VFRE);
4316 wr32(E1000_VFRE, reg | (1 << vf));
4317
4318 /* enable mailbox functionality for vf */
4319 adapter->vf_data[vf].clear_to_send = true;
4320
4321 /* reply to reset with ack and vf mac address */
4322 msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
4323 memcpy(addr, vf_mac, 6);
4324 igb_write_mbx(hw, msgbuf, 3, vf);
4325}
4326
4327static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
4328{
4329 unsigned char *addr = (char *)&msg[1];
4330 int err = -1;
4331
4332 if (is_valid_ether_addr(addr))
4333 err = igb_set_vf_mac(adapter, vf, addr);
4334
4335 return err;
4336
4337}
4338
4339static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
4340{
4341 struct e1000_hw *hw = &adapter->hw;
4342 u32 msg = E1000_VT_MSGTYPE_NACK;
4343
4344 /* if device isn't clear to send it shouldn't be reading either */
4345 if (!adapter->vf_data[vf].clear_to_send)
4346 igb_write_mbx(hw, &msg, 1, vf);
4347}
4348
4349
4350static void igb_msg_task(struct igb_adapter *adapter)
4351{
4352 struct e1000_hw *hw = &adapter->hw;
4353 u32 vf;
4354
4355 for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
4356 /* process any reset requests */
4357 if (!igb_check_for_rst(hw, vf)) {
4358 adapter->vf_data[vf].clear_to_send = false;
4359 igb_vf_reset_event(adapter, vf);
4360 }
4361
4362 /* process any messages pending */
4363 if (!igb_check_for_msg(hw, vf))
4364 igb_rcv_msg_from_vf(adapter, vf);
4365
4366 /* process any acks */
4367 if (!igb_check_for_ack(hw, vf))
4368 igb_rcv_ack_from_vf(adapter, vf);
4369
4370 }
4371}
4372
4373static int igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
4374{
4375 u32 mbx_size = E1000_VFMAILBOX_SIZE;
4376 u32 msgbuf[mbx_size];
4377 struct e1000_hw *hw = &adapter->hw;
4378 s32 retval;
4379
4380 retval = igb_read_mbx(hw, msgbuf, mbx_size, vf);
4381
4382 if (retval)
4383 dev_err(&adapter->pdev->dev,
4384 "Error receiving message from VF\n");
4385
4386 /* this is a message we already processed, do nothing */
4387 if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
4388 return retval;
4389
4390 /*
4391 * until the vf completes a reset it should not be
4392 * allowed to start any configuration.
4393 */
4394
4395 if (msgbuf[0] == E1000_VF_RESET) {
4396 igb_vf_reset_msg(adapter, vf);
4397
4398 return retval;
4399 }
4400
4401 if (!adapter->vf_data[vf].clear_to_send) {
4402 msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
4403 igb_write_mbx(hw, msgbuf, 1, vf);
4404 return retval;
4405 }
4406
4407 switch ((msgbuf[0] & 0xFFFF)) {
4408 case E1000_VF_SET_MAC_ADDR:
4409 retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
4410 break;
4411 case E1000_VF_SET_MULTICAST:
4412 retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
4413 break;
4414 case E1000_VF_SET_LPE:
4415 retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
4416 break;
4417 case E1000_VF_SET_VLAN:
4418 retval = igb_set_vf_vlan(adapter, msgbuf, vf);
4419 break;
4420 default:
4421 dev_err(&adapter->pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
4422 retval = -1;
4423 break;
4424 }
4425
4426 /* notify the VF of the results of what it sent us */
4427 if (retval)
4428 msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
4429 else
4430 msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
4431
4432 msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
4433
4434 igb_write_mbx(hw, msgbuf, 1, vf);
4435
4436 return retval;
4437}
4438
68d480c4
AD
4439/**
4440 * igb_set_uta - Set unicast filter table address
4441 * @adapter: board private structure
4442 *
4443 * The unicast table address is a register array of 32-bit registers.
4444 * The table is meant to be used in a way similar to how the MTA is used
4445 * however due to certain limitations in the hardware it is necessary to
4446 * set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscous
4447 * enable bit to allow vlan tag stripping when promiscous mode is enabled
4448 **/
4449static void igb_set_uta(struct igb_adapter *adapter)
4450{
4451 struct e1000_hw *hw = &adapter->hw;
4452 int i;
4453
4454 /* The UTA table only exists on 82576 hardware and newer */
4455 if (hw->mac.type < e1000_82576)
4456 return;
4457
4458 /* we only need to do this if VMDq is enabled */
4459 if (!adapter->vfs_allocated_count)
4460 return;
4461
4462 for (i = 0; i < hw->mac.uta_reg_count; i++)
4463 array_wr32(E1000_UTA, i, ~0);
4464}
4465
9d5c8243
AK
4466/**
4467 * igb_intr_msi - Interrupt Handler
4468 * @irq: interrupt number
4469 * @data: pointer to a network interface device structure
4470 **/
4471static irqreturn_t igb_intr_msi(int irq, void *data)
4472{
047e0030
AD
4473 struct igb_adapter *adapter = data;
4474 struct igb_q_vector *q_vector = adapter->q_vector[0];
9d5c8243
AK
4475 struct e1000_hw *hw = &adapter->hw;
4476 /* read ICR disables interrupts using IAM */
4477 u32 icr = rd32(E1000_ICR);
4478
047e0030 4479 igb_write_itr(q_vector);
9d5c8243 4480
047e0030 4481 if (icr & E1000_ICR_DOUTSYNC) {
dda0e083
AD
4482 /* HW is reporting DMA is out of sync */
4483 adapter->stats.doosync++;
4484 }
4485
9d5c8243
AK
4486 if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
4487 hw->mac.get_link_status = 1;
4488 if (!test_bit(__IGB_DOWN, &adapter->state))
4489 mod_timer(&adapter->watchdog_timer, jiffies + 1);
4490 }
4491
047e0030 4492 napi_schedule(&q_vector->napi);
9d5c8243
AK
4493
4494 return IRQ_HANDLED;
4495}
4496
4497/**
4a3c6433 4498 * igb_intr - Legacy Interrupt Handler
9d5c8243
AK
4499 * @irq: interrupt number
4500 * @data: pointer to a network interface device structure
4501 **/
4502static irqreturn_t igb_intr(int irq, void *data)
4503{
047e0030
AD
4504 struct igb_adapter *adapter = data;
4505 struct igb_q_vector *q_vector = adapter->q_vector[0];
9d5c8243
AK
4506 struct e1000_hw *hw = &adapter->hw;
4507 /* Interrupt Auto-Mask...upon reading ICR, interrupts are masked. No
4508 * need for the IMC write */
4509 u32 icr = rd32(E1000_ICR);
9d5c8243
AK
4510 if (!icr)
4511 return IRQ_NONE; /* Not our interrupt */
4512
047e0030 4513 igb_write_itr(q_vector);
9d5c8243
AK
4514
4515 /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
4516 * not set, then the adapter didn't send an interrupt */
4517 if (!(icr & E1000_ICR_INT_ASSERTED))
4518 return IRQ_NONE;
4519
047e0030 4520 if (icr & E1000_ICR_DOUTSYNC) {
dda0e083
AD
4521 /* HW is reporting DMA is out of sync */
4522 adapter->stats.doosync++;
4523 }
4524
9d5c8243
AK
4525 if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
4526 hw->mac.get_link_status = 1;
4527 /* guard against interrupt when we're going down */
4528 if (!test_bit(__IGB_DOWN, &adapter->state))
4529 mod_timer(&adapter->watchdog_timer, jiffies + 1);
4530 }
4531
047e0030 4532 napi_schedule(&q_vector->napi);
9d5c8243
AK
4533
4534 return IRQ_HANDLED;
4535}
4536
047e0030 4537static inline void igb_ring_irq_enable(struct igb_q_vector *q_vector)
9d5c8243 4538{
047e0030 4539 struct igb_adapter *adapter = q_vector->adapter;
46544258 4540 struct e1000_hw *hw = &adapter->hw;
9d5c8243 4541
4fc82adf
AD
4542 if ((q_vector->rx_ring && (adapter->rx_itr_setting & 3)) ||
4543 (!q_vector->rx_ring && (adapter->tx_itr_setting & 3))) {
047e0030 4544 if (!adapter->msix_entries)
6eb5a7f1 4545 igb_set_itr(adapter);
46544258 4546 else
047e0030 4547 igb_update_ring_itr(q_vector);
9d5c8243
AK
4548 }
4549
46544258
AD
4550 if (!test_bit(__IGB_DOWN, &adapter->state)) {
4551 if (adapter->msix_entries)
047e0030 4552 wr32(E1000_EIMS, q_vector->eims_value);
46544258
AD
4553 else
4554 igb_irq_enable(adapter);
4555 }
9d5c8243
AK
4556}
4557
46544258
AD
4558/**
4559 * igb_poll - NAPI Rx polling callback
4560 * @napi: napi polling structure
4561 * @budget: count of how many packets we should handle
4562 **/
4563static int igb_poll(struct napi_struct *napi, int budget)
9d5c8243 4564{
047e0030
AD
4565 struct igb_q_vector *q_vector = container_of(napi,
4566 struct igb_q_vector,
4567 napi);
4568 int tx_clean_complete = 1, work_done = 0;
9d5c8243 4569
421e02f0 4570#ifdef CONFIG_IGB_DCA
047e0030
AD
4571 if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
4572 igb_update_dca(q_vector);
fe4506b6 4573#endif
047e0030
AD
4574 if (q_vector->tx_ring)
4575 tx_clean_complete = igb_clean_tx_irq(q_vector);
9d5c8243 4576
047e0030
AD
4577 if (q_vector->rx_ring)
4578 igb_clean_rx_irq_adv(q_vector, &work_done, budget);
4579
4580 if (!tx_clean_complete)
4581 work_done = budget;
46544258 4582
9d5c8243 4583 /* If not enough Rx work done, exit the polling mode */
5e6d5b17 4584 if (work_done < budget) {
288379f0 4585 napi_complete(napi);
047e0030 4586 igb_ring_irq_enable(q_vector);
9d5c8243
AK
4587 }
4588
46544258 4589 return work_done;
9d5c8243 4590}
6d8126f9 4591
33af6bcc 4592/**
c5b9bd5e 4593 * igb_systim_to_hwtstamp - convert system time value to hw timestamp
33af6bcc 4594 * @adapter: board private structure
c5b9bd5e
AD
4595 * @shhwtstamps: timestamp structure to update
4596 * @regval: unsigned 64bit system time value.
4597 *
4598 * We need to convert the system time value stored in the RX/TXSTMP registers
4599 * into a hwtstamp which can be used by the upper level timestamping functions
4600 */
4601static void igb_systim_to_hwtstamp(struct igb_adapter *adapter,
4602 struct skb_shared_hwtstamps *shhwtstamps,
4603 u64 regval)
4604{
4605 u64 ns;
4606
4607 ns = timecounter_cyc2time(&adapter->clock, regval);
4608 timecompare_update(&adapter->compare, ns);
4609 memset(shhwtstamps, 0, sizeof(struct skb_shared_hwtstamps));
4610 shhwtstamps->hwtstamp = ns_to_ktime(ns);
4611 shhwtstamps->syststamp = timecompare_transform(&adapter->compare, ns);
4612}
4613
4614/**
4615 * igb_tx_hwtstamp - utility function which checks for TX time stamp
4616 * @q_vector: pointer to q_vector containing needed info
33af6bcc
PO
4617 * @skb: packet that was just sent
4618 *
4619 * If we were asked to do hardware stamping and such a time stamp is
4620 * available, then it must have been for this skb here because we only
4621 * allow only one such packet into the queue.
4622 */
c5b9bd5e 4623static void igb_tx_hwtstamp(struct igb_q_vector *q_vector, struct sk_buff *skb)
33af6bcc 4624{
c5b9bd5e 4625 struct igb_adapter *adapter = q_vector->adapter;
33af6bcc
PO
4626 union skb_shared_tx *shtx = skb_tx(skb);
4627 struct e1000_hw *hw = &adapter->hw;
c5b9bd5e
AD
4628 struct skb_shared_hwtstamps shhwtstamps;
4629 u64 regval;
33af6bcc 4630
c5b9bd5e
AD
4631 /* if skb does not support hw timestamp or TX stamp not valid exit */
4632 if (likely(!shtx->hardware) ||
4633 !(rd32(E1000_TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID))
4634 return;
4635
4636 regval = rd32(E1000_TXSTMPL);
4637 regval |= (u64)rd32(E1000_TXSTMPH) << 32;
4638
4639 igb_systim_to_hwtstamp(adapter, &shhwtstamps, regval);
4640 skb_tstamp_tx(skb, &shhwtstamps);
33af6bcc
PO
4641}
4642
9d5c8243
AK
4643/**
4644 * igb_clean_tx_irq - Reclaim resources after transmit completes
047e0030 4645 * @q_vector: pointer to q_vector containing needed info
9d5c8243
AK
4646 * returns true if ring is completely cleaned
4647 **/
047e0030 4648static bool igb_clean_tx_irq(struct igb_q_vector *q_vector)
9d5c8243 4649{
047e0030
AD
4650 struct igb_adapter *adapter = q_vector->adapter;
4651 struct igb_ring *tx_ring = q_vector->tx_ring;
e694e964 4652 struct net_device *netdev = tx_ring->netdev;
0e014cb1 4653 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK
4654 struct igb_buffer *buffer_info;
4655 struct sk_buff *skb;
0e014cb1 4656 union e1000_adv_tx_desc *tx_desc, *eop_desc;
9d5c8243 4657 unsigned int total_bytes = 0, total_packets = 0;
0e014cb1
AD
4658 unsigned int i, eop, count = 0;
4659 bool cleaned = false;
9d5c8243 4660
9d5c8243 4661 i = tx_ring->next_to_clean;
0e014cb1
AD
4662 eop = tx_ring->buffer_info[i].next_to_watch;
4663 eop_desc = E1000_TX_DESC_ADV(*tx_ring, eop);
4664
4665 while ((eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)) &&
4666 (count < tx_ring->count)) {
4667 for (cleaned = false; !cleaned; count++) {
4668 tx_desc = E1000_TX_DESC_ADV(*tx_ring, i);
9d5c8243 4669 buffer_info = &tx_ring->buffer_info[i];
0e014cb1 4670 cleaned = (i == eop);
9d5c8243
AK
4671 skb = buffer_info->skb;
4672
4673 if (skb) {
4674 unsigned int segs, bytecount;
4675 /* gso_segs is currently only valid for tcp */
4676 segs = skb_shinfo(skb)->gso_segs ?: 1;
4677 /* multiply data chunks by size of headers */
4678 bytecount = ((segs - 1) * skb_headlen(skb)) +
4679 skb->len;
4680 total_packets += segs;
4681 total_bytes += bytecount;
33af6bcc 4682
c5b9bd5e 4683 igb_tx_hwtstamp(q_vector, skb);
9d5c8243
AK
4684 }
4685
80785298 4686 igb_unmap_and_free_tx_resource(tx_ring, buffer_info);
0e014cb1 4687 tx_desc->wb.status = 0;
9d5c8243
AK
4688
4689 i++;
4690 if (i == tx_ring->count)
4691 i = 0;
9d5c8243 4692 }
0e014cb1
AD
4693 eop = tx_ring->buffer_info[i].next_to_watch;
4694 eop_desc = E1000_TX_DESC_ADV(*tx_ring, eop);
4695 }
4696
9d5c8243
AK
4697 tx_ring->next_to_clean = i;
4698
fc7d345d 4699 if (unlikely(count &&
9d5c8243 4700 netif_carrier_ok(netdev) &&
c493ea45 4701 igb_desc_unused(tx_ring) >= IGB_TX_QUEUE_WAKE)) {
9d5c8243
AK
4702 /* Make sure that anybody stopping the queue after this
4703 * sees the new next_to_clean.
4704 */
4705 smp_mb();
661086df
PWJ
4706 if (__netif_subqueue_stopped(netdev, tx_ring->queue_index) &&
4707 !(test_bit(__IGB_DOWN, &adapter->state))) {
4708 netif_wake_subqueue(netdev, tx_ring->queue_index);
04a5fcaa 4709 tx_ring->tx_stats.restart_queue++;
661086df 4710 }
9d5c8243
AK
4711 }
4712
4713 if (tx_ring->detect_tx_hung) {
4714 /* Detect a transmit hang in hardware, this serializes the
4715 * check with the clearing of time_stamp and movement of i */
4716 tx_ring->detect_tx_hung = false;
4717 if (tx_ring->buffer_info[i].time_stamp &&
4718 time_after(jiffies, tx_ring->buffer_info[i].time_stamp +
4719 (adapter->tx_timeout_factor * HZ))
4720 && !(rd32(E1000_STATUS) &
4721 E1000_STATUS_TXOFF)) {
4722
9d5c8243 4723 /* detected Tx unit hang */
80785298 4724 dev_err(&tx_ring->pdev->dev,
9d5c8243 4725 "Detected Tx Unit Hang\n"
2d064c06 4726 " Tx Queue <%d>\n"
9d5c8243
AK
4727 " TDH <%x>\n"
4728 " TDT <%x>\n"
4729 " next_to_use <%x>\n"
4730 " next_to_clean <%x>\n"
9d5c8243
AK
4731 "buffer_info[next_to_clean]\n"
4732 " time_stamp <%lx>\n"
0e014cb1 4733 " next_to_watch <%x>\n"
9d5c8243
AK
4734 " jiffies <%lx>\n"
4735 " desc.status <%x>\n",
2d064c06 4736 tx_ring->queue_index,
fce99e34
AD
4737 readl(tx_ring->head),
4738 readl(tx_ring->tail),
9d5c8243
AK
4739 tx_ring->next_to_use,
4740 tx_ring->next_to_clean,
9d5c8243 4741 tx_ring->buffer_info[i].time_stamp,
0e014cb1 4742 eop,
9d5c8243 4743 jiffies,
0e014cb1 4744 eop_desc->wb.status);
661086df 4745 netif_stop_subqueue(netdev, tx_ring->queue_index);
9d5c8243
AK
4746 }
4747 }
4748 tx_ring->total_bytes += total_bytes;
4749 tx_ring->total_packets += total_packets;
e21ed353
AD
4750 tx_ring->tx_stats.bytes += total_bytes;
4751 tx_ring->tx_stats.packets += total_packets;
8d24e933
AK
4752 netdev->stats.tx_bytes += total_bytes;
4753 netdev->stats.tx_packets += total_packets;
0e014cb1 4754 return (count < tx_ring->count);
9d5c8243
AK
4755}
4756
9d5c8243
AK
4757/**
4758 * igb_receive_skb - helper function to handle rx indications
047e0030
AD
4759 * @q_vector: structure containing interrupt and ring information
4760 * @skb: packet to send up
4761 * @vlan_tag: vlan tag for packet
9d5c8243 4762 **/
047e0030
AD
4763static void igb_receive_skb(struct igb_q_vector *q_vector,
4764 struct sk_buff *skb,
4765 u16 vlan_tag)
4766{
4767 struct igb_adapter *adapter = q_vector->adapter;
4768
4769 if (vlan_tag)
4770 vlan_gro_receive(&q_vector->napi, adapter->vlgrp,
4771 vlan_tag, skb);
182ff8df 4772 else
047e0030 4773 napi_gro_receive(&q_vector->napi, skb);
9d5c8243
AK
4774}
4775
04a5fcaa 4776static inline void igb_rx_checksum_adv(struct igb_ring *ring,
9d5c8243
AK
4777 u32 status_err, struct sk_buff *skb)
4778{
4779 skb->ip_summed = CHECKSUM_NONE;
4780
4781 /* Ignore Checksum bit is set or checksum is disabled through ethtool */
85ad76b2
AD
4782 if (!(ring->flags & IGB_RING_FLAG_RX_CSUM) ||
4783 (status_err & E1000_RXD_STAT_IXSM))
9d5c8243 4784 return;
85ad76b2 4785
9d5c8243
AK
4786 /* TCP/UDP checksum error bit is set */
4787 if (status_err &
4788 (E1000_RXDEXT_STATERR_TCPE | E1000_RXDEXT_STATERR_IPE)) {
b9473560
JB
4789 /*
4790 * work around errata with sctp packets where the TCPE aka
4791 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
4792 * packets, (aka let the stack check the crc32c)
4793 */
85ad76b2
AD
4794 if ((skb->len == 60) &&
4795 (ring->flags & IGB_RING_FLAG_RX_SCTP_CSUM))
04a5fcaa 4796 ring->rx_stats.csum_err++;
85ad76b2 4797
9d5c8243 4798 /* let the stack verify checksum errors */
9d5c8243
AK
4799 return;
4800 }
4801 /* It must be a TCP or UDP packet with a valid checksum */
4802 if (status_err & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS))
4803 skb->ip_summed = CHECKSUM_UNNECESSARY;
4804
85ad76b2 4805 dev_dbg(&ring->pdev->dev, "cksum success: bits %08X\n", status_err);
9d5c8243
AK
4806}
4807
c5b9bd5e
AD
4808static inline void igb_rx_hwtstamp(struct igb_q_vector *q_vector, u32 staterr,
4809 struct sk_buff *skb)
4810{
4811 struct igb_adapter *adapter = q_vector->adapter;
4812 struct e1000_hw *hw = &adapter->hw;
4813 u64 regval;
4814
4815 /*
4816 * If this bit is set, then the RX registers contain the time stamp. No
4817 * other packet will be time stamped until we read these registers, so
4818 * read the registers to make them available again. Because only one
4819 * packet can be time stamped at a time, we know that the register
4820 * values must belong to this one here and therefore we don't need to
4821 * compare any of the additional attributes stored for it.
4822 *
4823 * If nothing went wrong, then it should have a skb_shared_tx that we
4824 * can turn into a skb_shared_hwtstamps.
4825 */
4826 if (likely(!(staterr & E1000_RXDADV_STAT_TS)))
4827 return;
4828 if (!(rd32(E1000_TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
4829 return;
4830
4831 regval = rd32(E1000_RXSTMPL);
4832 regval |= (u64)rd32(E1000_RXSTMPH) << 32;
4833
4834 igb_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), regval);
4835}
4c844851 4836static inline u16 igb_get_hlen(struct igb_ring *rx_ring,
2d94d8ab
AD
4837 union e1000_adv_rx_desc *rx_desc)
4838{
4839 /* HW will not DMA in data larger than the given buffer, even if it
4840 * parses the (NFS, of course) header to be larger. In that case, it
4841 * fills the header buffer and spills the rest into the page.
4842 */
4843 u16 hlen = (le16_to_cpu(rx_desc->wb.lower.lo_dword.hdr_info) &
4844 E1000_RXDADV_HDRBUFLEN_MASK) >> E1000_RXDADV_HDRBUFLEN_SHIFT;
4c844851
AD
4845 if (hlen > rx_ring->rx_buffer_len)
4846 hlen = rx_ring->rx_buffer_len;
2d94d8ab
AD
4847 return hlen;
4848}
4849
047e0030
AD
4850static bool igb_clean_rx_irq_adv(struct igb_q_vector *q_vector,
4851 int *work_done, int budget)
9d5c8243 4852{
047e0030 4853 struct igb_ring *rx_ring = q_vector->rx_ring;
e694e964 4854 struct net_device *netdev = rx_ring->netdev;
80785298 4855 struct pci_dev *pdev = rx_ring->pdev;
9d5c8243
AK
4856 union e1000_adv_rx_desc *rx_desc , *next_rxd;
4857 struct igb_buffer *buffer_info , *next_buffer;
4858 struct sk_buff *skb;
9d5c8243
AK
4859 bool cleaned = false;
4860 int cleaned_count = 0;
4861 unsigned int total_bytes = 0, total_packets = 0;
73cd78f1 4862 unsigned int i;
2d94d8ab
AD
4863 u32 staterr;
4864 u16 length;
047e0030 4865 u16 vlan_tag;
9d5c8243
AK
4866
4867 i = rx_ring->next_to_clean;
69d3ca53 4868 buffer_info = &rx_ring->buffer_info[i];
9d5c8243
AK
4869 rx_desc = E1000_RX_DESC_ADV(*rx_ring, i);
4870 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
4871
4872 while (staterr & E1000_RXD_STAT_DD) {
4873 if (*work_done >= budget)
4874 break;
4875 (*work_done)++;
9d5c8243 4876
69d3ca53
AD
4877 skb = buffer_info->skb;
4878 prefetch(skb->data - NET_IP_ALIGN);
4879 buffer_info->skb = NULL;
4880
4881 i++;
4882 if (i == rx_ring->count)
4883 i = 0;
4884 next_rxd = E1000_RX_DESC_ADV(*rx_ring, i);
4885 prefetch(next_rxd);
4886 next_buffer = &rx_ring->buffer_info[i];
9d5c8243
AK
4887
4888 length = le16_to_cpu(rx_desc->wb.upper.length);
4889 cleaned = true;
4890 cleaned_count++;
4891
2d94d8ab 4892 if (buffer_info->dma) {
bf36c1a0 4893 pci_unmap_single(pdev, buffer_info->dma,
4c844851 4894 rx_ring->rx_buffer_len,
bf36c1a0 4895 PCI_DMA_FROMDEVICE);
91615f76 4896 buffer_info->dma = 0;
4c844851 4897 if (rx_ring->rx_buffer_len >= IGB_RXBUFFER_1024) {
6ec43fe6
AD
4898 skb_put(skb, length);
4899 goto send_up;
4900 }
4c844851 4901 skb_put(skb, igb_get_hlen(rx_ring, rx_desc));
bf36c1a0
AD
4902 }
4903
4904 if (length) {
9d5c8243 4905 pci_unmap_page(pdev, buffer_info->page_dma,
bf36c1a0 4906 PAGE_SIZE / 2, PCI_DMA_FROMDEVICE);
9d5c8243 4907 buffer_info->page_dma = 0;
bf36c1a0
AD
4908
4909 skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags++,
4910 buffer_info->page,
4911 buffer_info->page_offset,
4912 length);
4913
6ec43fe6 4914 if (page_count(buffer_info->page) != 1)
bf36c1a0
AD
4915 buffer_info->page = NULL;
4916 else
4917 get_page(buffer_info->page);
9d5c8243
AK
4918
4919 skb->len += length;
4920 skb->data_len += length;
9d5c8243 4921
bf36c1a0 4922 skb->truesize += length;
9d5c8243 4923 }
9d5c8243 4924
bf36c1a0 4925 if (!(staterr & E1000_RXD_STAT_EOP)) {
b2d56536
AD
4926 buffer_info->skb = next_buffer->skb;
4927 buffer_info->dma = next_buffer->dma;
4928 next_buffer->skb = skb;
4929 next_buffer->dma = 0;
bf36c1a0
AD
4930 goto next_desc;
4931 }
69d3ca53 4932send_up:
9d5c8243
AK
4933 if (staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) {
4934 dev_kfree_skb_irq(skb);
4935 goto next_desc;
4936 }
9d5c8243 4937
c5b9bd5e 4938 igb_rx_hwtstamp(q_vector, staterr, skb);
9d5c8243
AK
4939 total_bytes += skb->len;
4940 total_packets++;
4941
85ad76b2 4942 igb_rx_checksum_adv(rx_ring, staterr, skb);
9d5c8243
AK
4943
4944 skb->protocol = eth_type_trans(skb, netdev);
047e0030
AD
4945 skb_record_rx_queue(skb, rx_ring->queue_index);
4946
4947 vlan_tag = ((staterr & E1000_RXD_STAT_VP) ?
4948 le16_to_cpu(rx_desc->wb.upper.vlan) : 0);
9d5c8243 4949
047e0030 4950 igb_receive_skb(q_vector, skb, vlan_tag);
9d5c8243 4951
9d5c8243
AK
4952next_desc:
4953 rx_desc->wb.upper.status_error = 0;
4954
4955 /* return some buffers to hardware, one at a time is too slow */
4956 if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
3b644cf6 4957 igb_alloc_rx_buffers_adv(rx_ring, cleaned_count);
9d5c8243
AK
4958 cleaned_count = 0;
4959 }
4960
4961 /* use prefetched values */
4962 rx_desc = next_rxd;
4963 buffer_info = next_buffer;
9d5c8243
AK
4964 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
4965 }
bf36c1a0 4966
9d5c8243 4967 rx_ring->next_to_clean = i;
c493ea45 4968 cleaned_count = igb_desc_unused(rx_ring);
9d5c8243
AK
4969
4970 if (cleaned_count)
3b644cf6 4971 igb_alloc_rx_buffers_adv(rx_ring, cleaned_count);
9d5c8243
AK
4972
4973 rx_ring->total_packets += total_packets;
4974 rx_ring->total_bytes += total_bytes;
4975 rx_ring->rx_stats.packets += total_packets;
4976 rx_ring->rx_stats.bytes += total_bytes;
8d24e933
AK
4977 netdev->stats.rx_bytes += total_bytes;
4978 netdev->stats.rx_packets += total_packets;
9d5c8243
AK
4979 return cleaned;
4980}
4981
9d5c8243
AK
4982/**
4983 * igb_alloc_rx_buffers_adv - Replace used receive buffers; packet split
4984 * @adapter: address of board private structure
4985 **/
d7ee5b3a 4986void igb_alloc_rx_buffers_adv(struct igb_ring *rx_ring, int cleaned_count)
9d5c8243 4987{
e694e964 4988 struct net_device *netdev = rx_ring->netdev;
9d5c8243
AK
4989 union e1000_adv_rx_desc *rx_desc;
4990 struct igb_buffer *buffer_info;
4991 struct sk_buff *skb;
4992 unsigned int i;
db761762 4993 int bufsz;
9d5c8243
AK
4994
4995 i = rx_ring->next_to_use;
4996 buffer_info = &rx_ring->buffer_info[i];
4997
4c844851 4998 bufsz = rx_ring->rx_buffer_len;
db761762 4999
9d5c8243
AK
5000 while (cleaned_count--) {
5001 rx_desc = E1000_RX_DESC_ADV(*rx_ring, i);
5002
6ec43fe6 5003 if ((bufsz < IGB_RXBUFFER_1024) && !buffer_info->page_dma) {
9d5c8243 5004 if (!buffer_info->page) {
bf36c1a0
AD
5005 buffer_info->page = alloc_page(GFP_ATOMIC);
5006 if (!buffer_info->page) {
04a5fcaa 5007 rx_ring->rx_stats.alloc_failed++;
bf36c1a0
AD
5008 goto no_buffers;
5009 }
5010 buffer_info->page_offset = 0;
5011 } else {
5012 buffer_info->page_offset ^= PAGE_SIZE / 2;
9d5c8243
AK
5013 }
5014 buffer_info->page_dma =
80785298 5015 pci_map_page(rx_ring->pdev, buffer_info->page,
bf36c1a0
AD
5016 buffer_info->page_offset,
5017 PAGE_SIZE / 2,
9d5c8243
AK
5018 PCI_DMA_FROMDEVICE);
5019 }
5020
5021 if (!buffer_info->skb) {
89d71a66 5022 skb = netdev_alloc_skb_ip_align(netdev, bufsz);
9d5c8243 5023 if (!skb) {
04a5fcaa 5024 rx_ring->rx_stats.alloc_failed++;
9d5c8243
AK
5025 goto no_buffers;
5026 }
5027
9d5c8243 5028 buffer_info->skb = skb;
80785298
AD
5029 buffer_info->dma = pci_map_single(rx_ring->pdev,
5030 skb->data,
9d5c8243
AK
5031 bufsz,
5032 PCI_DMA_FROMDEVICE);
9d5c8243
AK
5033 }
5034 /* Refresh the desc even if buffer_addrs didn't change because
5035 * each write-back erases this info. */
6ec43fe6 5036 if (bufsz < IGB_RXBUFFER_1024) {
9d5c8243
AK
5037 rx_desc->read.pkt_addr =
5038 cpu_to_le64(buffer_info->page_dma);
5039 rx_desc->read.hdr_addr = cpu_to_le64(buffer_info->dma);
5040 } else {
5041 rx_desc->read.pkt_addr =
5042 cpu_to_le64(buffer_info->dma);
5043 rx_desc->read.hdr_addr = 0;
5044 }
5045
5046 i++;
5047 if (i == rx_ring->count)
5048 i = 0;
5049 buffer_info = &rx_ring->buffer_info[i];
5050 }
5051
5052no_buffers:
5053 if (rx_ring->next_to_use != i) {
5054 rx_ring->next_to_use = i;
5055 if (i == 0)
5056 i = (rx_ring->count - 1);
5057 else
5058 i--;
5059
5060 /* Force memory writes to complete before letting h/w
5061 * know there are new descriptors to fetch. (Only
5062 * applicable for weak-ordered memory model archs,
5063 * such as IA-64). */
5064 wmb();
fce99e34 5065 writel(i, rx_ring->tail);
9d5c8243
AK
5066 }
5067}
5068
5069/**
5070 * igb_mii_ioctl -
5071 * @netdev:
5072 * @ifreq:
5073 * @cmd:
5074 **/
5075static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
5076{
5077 struct igb_adapter *adapter = netdev_priv(netdev);
5078 struct mii_ioctl_data *data = if_mii(ifr);
5079
5080 if (adapter->hw.phy.media_type != e1000_media_type_copper)
5081 return -EOPNOTSUPP;
5082
5083 switch (cmd) {
5084 case SIOCGMIIPHY:
5085 data->phy_id = adapter->hw.phy.addr;
5086 break;
5087 case SIOCGMIIREG:
f5f4cf08
AD
5088 if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
5089 &data->val_out))
9d5c8243
AK
5090 return -EIO;
5091 break;
5092 case SIOCSMIIREG:
5093 default:
5094 return -EOPNOTSUPP;
5095 }
5096 return 0;
5097}
5098
c6cb090b
PO
5099/**
5100 * igb_hwtstamp_ioctl - control hardware time stamping
5101 * @netdev:
5102 * @ifreq:
5103 * @cmd:
5104 *
33af6bcc
PO
5105 * Outgoing time stamping can be enabled and disabled. Play nice and
5106 * disable it when requested, although it shouldn't case any overhead
5107 * when no packet needs it. At most one packet in the queue may be
5108 * marked for time stamping, otherwise it would be impossible to tell
5109 * for sure to which packet the hardware time stamp belongs.
5110 *
5111 * Incoming time stamping has to be configured via the hardware
5112 * filters. Not all combinations are supported, in particular event
5113 * type has to be specified. Matching the kind of event packet is
5114 * not supported, with the exception of "all V2 events regardless of
5115 * level 2 or 4".
5116 *
c6cb090b
PO
5117 **/
5118static int igb_hwtstamp_ioctl(struct net_device *netdev,
5119 struct ifreq *ifr, int cmd)
5120{
33af6bcc
PO
5121 struct igb_adapter *adapter = netdev_priv(netdev);
5122 struct e1000_hw *hw = &adapter->hw;
c6cb090b 5123 struct hwtstamp_config config;
c5b9bd5e
AD
5124 u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
5125 u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
33af6bcc 5126 u32 tsync_rx_cfg = 0;
c5b9bd5e
AD
5127 bool is_l4 = false;
5128 bool is_l2 = false;
33af6bcc 5129 u32 regval;
c6cb090b
PO
5130
5131 if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
5132 return -EFAULT;
5133
5134 /* reserved for future extensions */
5135 if (config.flags)
5136 return -EINVAL;
5137
33af6bcc
PO
5138 switch (config.tx_type) {
5139 case HWTSTAMP_TX_OFF:
c5b9bd5e 5140 tsync_tx_ctl = 0;
33af6bcc 5141 case HWTSTAMP_TX_ON:
33af6bcc
PO
5142 break;
5143 default:
5144 return -ERANGE;
5145 }
5146
5147 switch (config.rx_filter) {
5148 case HWTSTAMP_FILTER_NONE:
c5b9bd5e 5149 tsync_rx_ctl = 0;
33af6bcc
PO
5150 break;
5151 case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
5152 case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
5153 case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
5154 case HWTSTAMP_FILTER_ALL:
5155 /*
5156 * register TSYNCRXCFG must be set, therefore it is not
5157 * possible to time stamp both Sync and Delay_Req messages
5158 * => fall back to time stamping all packets
5159 */
c5b9bd5e 5160 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
33af6bcc
PO
5161 config.rx_filter = HWTSTAMP_FILTER_ALL;
5162 break;
5163 case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
c5b9bd5e 5164 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
33af6bcc 5165 tsync_rx_cfg = E1000_TSYNCRXCFG_PTP_V1_SYNC_MESSAGE;
c5b9bd5e 5166 is_l4 = true;
33af6bcc
PO
5167 break;
5168 case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
c5b9bd5e 5169 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
33af6bcc 5170 tsync_rx_cfg = E1000_TSYNCRXCFG_PTP_V1_DELAY_REQ_MESSAGE;
c5b9bd5e 5171 is_l4 = true;
33af6bcc
PO
5172 break;
5173 case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
5174 case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
c5b9bd5e 5175 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
33af6bcc 5176 tsync_rx_cfg = E1000_TSYNCRXCFG_PTP_V2_SYNC_MESSAGE;
c5b9bd5e
AD
5177 is_l2 = true;
5178 is_l4 = true;
33af6bcc
PO
5179 config.rx_filter = HWTSTAMP_FILTER_SOME;
5180 break;
5181 case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
5182 case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
c5b9bd5e 5183 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
33af6bcc 5184 tsync_rx_cfg = E1000_TSYNCRXCFG_PTP_V2_DELAY_REQ_MESSAGE;
c5b9bd5e
AD
5185 is_l2 = true;
5186 is_l4 = true;
33af6bcc
PO
5187 config.rx_filter = HWTSTAMP_FILTER_SOME;
5188 break;
5189 case HWTSTAMP_FILTER_PTP_V2_EVENT:
5190 case HWTSTAMP_FILTER_PTP_V2_SYNC:
5191 case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
c5b9bd5e 5192 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
33af6bcc 5193 config.rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
c5b9bd5e 5194 is_l2 = true;
33af6bcc
PO
5195 break;
5196 default:
5197 return -ERANGE;
5198 }
5199
c5b9bd5e
AD
5200 if (hw->mac.type == e1000_82575) {
5201 if (tsync_rx_ctl | tsync_tx_ctl)
5202 return -EINVAL;
5203 return 0;
5204 }
5205
33af6bcc
PO
5206 /* enable/disable TX */
5207 regval = rd32(E1000_TSYNCTXCTL);
c5b9bd5e
AD
5208 regval &= ~E1000_TSYNCTXCTL_ENABLED;
5209 regval |= tsync_tx_ctl;
33af6bcc
PO
5210 wr32(E1000_TSYNCTXCTL, regval);
5211
c5b9bd5e 5212 /* enable/disable RX */
33af6bcc 5213 regval = rd32(E1000_TSYNCRXCTL);
c5b9bd5e
AD
5214 regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
5215 regval |= tsync_rx_ctl;
33af6bcc 5216 wr32(E1000_TSYNCRXCTL, regval);
33af6bcc 5217
c5b9bd5e
AD
5218 /* define which PTP packets are time stamped */
5219 wr32(E1000_TSYNCRXCFG, tsync_rx_cfg);
33af6bcc 5220
c5b9bd5e
AD
5221 /* define ethertype filter for timestamped packets */
5222 if (is_l2)
5223 wr32(E1000_ETQF(3),
5224 (E1000_ETQF_FILTER_ENABLE | /* enable filter */
5225 E1000_ETQF_1588 | /* enable timestamping */
5226 ETH_P_1588)); /* 1588 eth protocol type */
5227 else
5228 wr32(E1000_ETQF(3), 0);
5229
5230#define PTP_PORT 319
5231 /* L4 Queue Filter[3]: filter by destination port and protocol */
5232 if (is_l4) {
5233 u32 ftqf = (IPPROTO_UDP /* UDP */
5234 | E1000_FTQF_VF_BP /* VF not compared */
5235 | E1000_FTQF_1588_TIME_STAMP /* Enable Timestamping */
5236 | E1000_FTQF_MASK); /* mask all inputs */
5237 ftqf &= ~E1000_FTQF_MASK_PROTO_BP; /* enable protocol check */
5238
5239 wr32(E1000_IMIR(3), htons(PTP_PORT));
5240 wr32(E1000_IMIREXT(3),
5241 (E1000_IMIREXT_SIZE_BP | E1000_IMIREXT_CTRL_BP));
5242 if (hw->mac.type == e1000_82576) {
5243 /* enable source port check */
5244 wr32(E1000_SPQF(3), htons(PTP_PORT));
5245 ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
5246 }
5247 wr32(E1000_FTQF(3), ftqf);
5248 } else {
5249 wr32(E1000_FTQF(3), E1000_FTQF_MASK);
5250 }
33af6bcc
PO
5251 wrfl();
5252
5253 adapter->hwtstamp_config = config;
5254
5255 /* clear TX/RX time stamp registers, just to be sure */
5256 regval = rd32(E1000_TXSTMPH);
5257 regval = rd32(E1000_RXSTMPH);
c6cb090b 5258
33af6bcc
PO
5259 return copy_to_user(ifr->ifr_data, &config, sizeof(config)) ?
5260 -EFAULT : 0;
c6cb090b
PO
5261}
5262
9d5c8243
AK
5263/**
5264 * igb_ioctl -
5265 * @netdev:
5266 * @ifreq:
5267 * @cmd:
5268 **/
5269static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
5270{
5271 switch (cmd) {
5272 case SIOCGMIIPHY:
5273 case SIOCGMIIREG:
5274 case SIOCSMIIREG:
5275 return igb_mii_ioctl(netdev, ifr, cmd);
c6cb090b
PO
5276 case SIOCSHWTSTAMP:
5277 return igb_hwtstamp_ioctl(netdev, ifr, cmd);
9d5c8243
AK
5278 default:
5279 return -EOPNOTSUPP;
5280 }
5281}
5282
009bc06e
AD
5283s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
5284{
5285 struct igb_adapter *adapter = hw->back;
5286 u16 cap_offset;
5287
5288 cap_offset = pci_find_capability(adapter->pdev, PCI_CAP_ID_EXP);
5289 if (!cap_offset)
5290 return -E1000_ERR_CONFIG;
5291
5292 pci_read_config_word(adapter->pdev, cap_offset + reg, value);
5293
5294 return 0;
5295}
5296
5297s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
5298{
5299 struct igb_adapter *adapter = hw->back;
5300 u16 cap_offset;
5301
5302 cap_offset = pci_find_capability(adapter->pdev, PCI_CAP_ID_EXP);
5303 if (!cap_offset)
5304 return -E1000_ERR_CONFIG;
5305
5306 pci_write_config_word(adapter->pdev, cap_offset + reg, *value);
5307
5308 return 0;
5309}
5310
9d5c8243
AK
5311static void igb_vlan_rx_register(struct net_device *netdev,
5312 struct vlan_group *grp)
5313{
5314 struct igb_adapter *adapter = netdev_priv(netdev);
5315 struct e1000_hw *hw = &adapter->hw;
5316 u32 ctrl, rctl;
5317
5318 igb_irq_disable(adapter);
5319 adapter->vlgrp = grp;
5320
5321 if (grp) {
5322 /* enable VLAN tag insert/strip */
5323 ctrl = rd32(E1000_CTRL);
5324 ctrl |= E1000_CTRL_VME;
5325 wr32(E1000_CTRL, ctrl);
5326
5327 /* enable VLAN receive filtering */
5328 rctl = rd32(E1000_RCTL);
9d5c8243
AK
5329 rctl &= ~E1000_RCTL_CFIEN;
5330 wr32(E1000_RCTL, rctl);
5331 igb_update_mng_vlan(adapter);
9d5c8243
AK
5332 } else {
5333 /* disable VLAN tag insert/strip */
5334 ctrl = rd32(E1000_CTRL);
5335 ctrl &= ~E1000_CTRL_VME;
5336 wr32(E1000_CTRL, ctrl);
5337
9d5c8243
AK
5338 if (adapter->mng_vlan_id != (u16)IGB_MNG_VLAN_NONE) {
5339 igb_vlan_rx_kill_vid(netdev, adapter->mng_vlan_id);
5340 adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
5341 }
9d5c8243
AK
5342 }
5343
e1739522
AD
5344 igb_rlpml_set(adapter);
5345
9d5c8243
AK
5346 if (!test_bit(__IGB_DOWN, &adapter->state))
5347 igb_irq_enable(adapter);
5348}
5349
5350static void igb_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
5351{
5352 struct igb_adapter *adapter = netdev_priv(netdev);
5353 struct e1000_hw *hw = &adapter->hw;
4ae196df 5354 int pf_id = adapter->vfs_allocated_count;
9d5c8243 5355
28b0759c 5356 if ((hw->mng_cookie.status &
9d5c8243
AK
5357 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
5358 (vid == adapter->mng_vlan_id))
5359 return;
4ae196df
AD
5360
5361 /* add vid to vlvf if sr-iov is enabled,
5362 * if that fails add directly to filter table */
5363 if (igb_vlvf_set(adapter, vid, true, pf_id))
5364 igb_vfta_set(hw, vid, true);
5365
9d5c8243
AK
5366}
5367
5368static void igb_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
5369{
5370 struct igb_adapter *adapter = netdev_priv(netdev);
5371 struct e1000_hw *hw = &adapter->hw;
4ae196df 5372 int pf_id = adapter->vfs_allocated_count;
9d5c8243
AK
5373
5374 igb_irq_disable(adapter);
5375 vlan_group_set_device(adapter->vlgrp, vid, NULL);
5376
5377 if (!test_bit(__IGB_DOWN, &adapter->state))
5378 igb_irq_enable(adapter);
5379
5380 if ((adapter->hw.mng_cookie.status &
5381 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
5382 (vid == adapter->mng_vlan_id)) {
5383 /* release control to f/w */
5384 igb_release_hw_control(adapter);
5385 return;
5386 }
5387
4ae196df
AD
5388 /* remove vid from vlvf if sr-iov is enabled,
5389 * if not in vlvf remove from vfta */
5390 if (igb_vlvf_set(adapter, vid, false, pf_id))
5391 igb_vfta_set(hw, vid, false);
9d5c8243
AK
5392}
5393
5394static void igb_restore_vlan(struct igb_adapter *adapter)
5395{
5396 igb_vlan_rx_register(adapter->netdev, adapter->vlgrp);
5397
5398 if (adapter->vlgrp) {
5399 u16 vid;
5400 for (vid = 0; vid < VLAN_GROUP_ARRAY_LEN; vid++) {
5401 if (!vlan_group_get_device(adapter->vlgrp, vid))
5402 continue;
5403 igb_vlan_rx_add_vid(adapter->netdev, vid);
5404 }
5405 }
5406}
5407
5408int igb_set_spd_dplx(struct igb_adapter *adapter, u16 spddplx)
5409{
5410 struct e1000_mac_info *mac = &adapter->hw.mac;
5411
5412 mac->autoneg = 0;
5413
9d5c8243
AK
5414 switch (spddplx) {
5415 case SPEED_10 + DUPLEX_HALF:
5416 mac->forced_speed_duplex = ADVERTISE_10_HALF;
5417 break;
5418 case SPEED_10 + DUPLEX_FULL:
5419 mac->forced_speed_duplex = ADVERTISE_10_FULL;
5420 break;
5421 case SPEED_100 + DUPLEX_HALF:
5422 mac->forced_speed_duplex = ADVERTISE_100_HALF;
5423 break;
5424 case SPEED_100 + DUPLEX_FULL:
5425 mac->forced_speed_duplex = ADVERTISE_100_FULL;
5426 break;
5427 case SPEED_1000 + DUPLEX_FULL:
5428 mac->autoneg = 1;
5429 adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
5430 break;
5431 case SPEED_1000 + DUPLEX_HALF: /* not supported */
5432 default:
5433 dev_err(&adapter->pdev->dev,
5434 "Unsupported Speed/Duplex configuration\n");
5435 return -EINVAL;
5436 }
5437 return 0;
5438}
5439
3fe7c4c9 5440static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake)
9d5c8243
AK
5441{
5442 struct net_device *netdev = pci_get_drvdata(pdev);
5443 struct igb_adapter *adapter = netdev_priv(netdev);
5444 struct e1000_hw *hw = &adapter->hw;
2d064c06 5445 u32 ctrl, rctl, status;
9d5c8243
AK
5446 u32 wufc = adapter->wol;
5447#ifdef CONFIG_PM
5448 int retval = 0;
5449#endif
5450
5451 netif_device_detach(netdev);
5452
a88f10ec
AD
5453 if (netif_running(netdev))
5454 igb_close(netdev);
5455
047e0030 5456 igb_clear_interrupt_scheme(adapter);
9d5c8243
AK
5457
5458#ifdef CONFIG_PM
5459 retval = pci_save_state(pdev);
5460 if (retval)
5461 return retval;
5462#endif
5463
5464 status = rd32(E1000_STATUS);
5465 if (status & E1000_STATUS_LU)
5466 wufc &= ~E1000_WUFC_LNKC;
5467
5468 if (wufc) {
5469 igb_setup_rctl(adapter);
ff41f8dc 5470 igb_set_rx_mode(netdev);
9d5c8243
AK
5471
5472 /* turn on all-multi mode if wake on multicast is enabled */
5473 if (wufc & E1000_WUFC_MC) {
5474 rctl = rd32(E1000_RCTL);
5475 rctl |= E1000_RCTL_MPE;
5476 wr32(E1000_RCTL, rctl);
5477 }
5478
5479 ctrl = rd32(E1000_CTRL);
5480 /* advertise wake from D3Cold */
5481 #define E1000_CTRL_ADVD3WUC 0x00100000
5482 /* phy power management enable */
5483 #define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
5484 ctrl |= E1000_CTRL_ADVD3WUC;
5485 wr32(E1000_CTRL, ctrl);
5486
9d5c8243
AK
5487 /* Allow time for pending master requests to run */
5488 igb_disable_pcie_master(&adapter->hw);
5489
5490 wr32(E1000_WUC, E1000_WUC_PME_EN);
5491 wr32(E1000_WUFC, wufc);
9d5c8243
AK
5492 } else {
5493 wr32(E1000_WUC, 0);
5494 wr32(E1000_WUFC, 0);
9d5c8243
AK
5495 }
5496
3fe7c4c9
RW
5497 *enable_wake = wufc || adapter->en_mng_pt;
5498 if (!*enable_wake)
2fb02a26 5499 igb_shutdown_serdes_link_82575(hw);
9d5c8243
AK
5500
5501 /* Release control of h/w to f/w. If f/w is AMT enabled, this
5502 * would have already happened in close and is redundant. */
5503 igb_release_hw_control(adapter);
5504
5505 pci_disable_device(pdev);
5506
9d5c8243
AK
5507 return 0;
5508}
5509
5510#ifdef CONFIG_PM
3fe7c4c9
RW
5511static int igb_suspend(struct pci_dev *pdev, pm_message_t state)
5512{
5513 int retval;
5514 bool wake;
5515
5516 retval = __igb_shutdown(pdev, &wake);
5517 if (retval)
5518 return retval;
5519
5520 if (wake) {
5521 pci_prepare_to_sleep(pdev);
5522 } else {
5523 pci_wake_from_d3(pdev, false);
5524 pci_set_power_state(pdev, PCI_D3hot);
5525 }
5526
5527 return 0;
5528}
5529
9d5c8243
AK
5530static int igb_resume(struct pci_dev *pdev)
5531{
5532 struct net_device *netdev = pci_get_drvdata(pdev);
5533 struct igb_adapter *adapter = netdev_priv(netdev);
5534 struct e1000_hw *hw = &adapter->hw;
5535 u32 err;
5536
5537 pci_set_power_state(pdev, PCI_D0);
5538 pci_restore_state(pdev);
42bfd33a 5539
aed5dec3 5540 err = pci_enable_device_mem(pdev);
9d5c8243
AK
5541 if (err) {
5542 dev_err(&pdev->dev,
5543 "igb: Cannot enable PCI device from suspend\n");
5544 return err;
5545 }
5546 pci_set_master(pdev);
5547
5548 pci_enable_wake(pdev, PCI_D3hot, 0);
5549 pci_enable_wake(pdev, PCI_D3cold, 0);
5550
047e0030 5551 if (igb_init_interrupt_scheme(adapter)) {
a88f10ec
AD
5552 dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
5553 return -ENOMEM;
9d5c8243
AK
5554 }
5555
5556 /* e1000_power_up_phy(adapter); */
5557
5558 igb_reset(adapter);
a8564f03
AD
5559
5560 /* let the f/w know that the h/w is now under the control of the
5561 * driver. */
5562 igb_get_hw_control(adapter);
5563
9d5c8243
AK
5564 wr32(E1000_WUS, ~0);
5565
a88f10ec
AD
5566 if (netif_running(netdev)) {
5567 err = igb_open(netdev);
5568 if (err)
5569 return err;
5570 }
9d5c8243
AK
5571
5572 netif_device_attach(netdev);
5573
9d5c8243
AK
5574 return 0;
5575}
5576#endif
5577
5578static void igb_shutdown(struct pci_dev *pdev)
5579{
3fe7c4c9
RW
5580 bool wake;
5581
5582 __igb_shutdown(pdev, &wake);
5583
5584 if (system_state == SYSTEM_POWER_OFF) {
5585 pci_wake_from_d3(pdev, wake);
5586 pci_set_power_state(pdev, PCI_D3hot);
5587 }
9d5c8243
AK
5588}
5589
5590#ifdef CONFIG_NET_POLL_CONTROLLER
5591/*
5592 * Polling 'interrupt' - used by things like netconsole to send skbs
5593 * without having to re-enable interrupts. It's not called while
5594 * the interrupt routine is executing.
5595 */
5596static void igb_netpoll(struct net_device *netdev)
5597{
5598 struct igb_adapter *adapter = netdev_priv(netdev);
eebbbdba 5599 struct e1000_hw *hw = &adapter->hw;
9d5c8243 5600 int i;
9d5c8243 5601
eebbbdba 5602 if (!adapter->msix_entries) {
047e0030 5603 struct igb_q_vector *q_vector = adapter->q_vector[0];
eebbbdba 5604 igb_irq_disable(adapter);
047e0030 5605 napi_schedule(&q_vector->napi);
eebbbdba
AD
5606 return;
5607 }
9d5c8243 5608
047e0030
AD
5609 for (i = 0; i < adapter->num_q_vectors; i++) {
5610 struct igb_q_vector *q_vector = adapter->q_vector[i];
5611 wr32(E1000_EIMC, q_vector->eims_value);
5612 napi_schedule(&q_vector->napi);
eebbbdba 5613 }
9d5c8243
AK
5614}
5615#endif /* CONFIG_NET_POLL_CONTROLLER */
5616
5617/**
5618 * igb_io_error_detected - called when PCI error is detected
5619 * @pdev: Pointer to PCI device
5620 * @state: The current pci connection state
5621 *
5622 * This function is called after a PCI bus error affecting
5623 * this device has been detected.
5624 */
5625static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
5626 pci_channel_state_t state)
5627{
5628 struct net_device *netdev = pci_get_drvdata(pdev);
5629 struct igb_adapter *adapter = netdev_priv(netdev);
5630
5631 netif_device_detach(netdev);
5632
59ed6eec
AD
5633 if (state == pci_channel_io_perm_failure)
5634 return PCI_ERS_RESULT_DISCONNECT;
5635
9d5c8243
AK
5636 if (netif_running(netdev))
5637 igb_down(adapter);
5638 pci_disable_device(pdev);
5639
5640 /* Request a slot slot reset. */
5641 return PCI_ERS_RESULT_NEED_RESET;
5642}
5643
5644/**
5645 * igb_io_slot_reset - called after the pci bus has been reset.
5646 * @pdev: Pointer to PCI device
5647 *
5648 * Restart the card from scratch, as if from a cold-boot. Implementation
5649 * resembles the first-half of the igb_resume routine.
5650 */
5651static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
5652{
5653 struct net_device *netdev = pci_get_drvdata(pdev);
5654 struct igb_adapter *adapter = netdev_priv(netdev);
5655 struct e1000_hw *hw = &adapter->hw;
40a914fa 5656 pci_ers_result_t result;
42bfd33a 5657 int err;
9d5c8243 5658
aed5dec3 5659 if (pci_enable_device_mem(pdev)) {
9d5c8243
AK
5660 dev_err(&pdev->dev,
5661 "Cannot re-enable PCI device after reset.\n");
40a914fa
AD
5662 result = PCI_ERS_RESULT_DISCONNECT;
5663 } else {
5664 pci_set_master(pdev);
5665 pci_restore_state(pdev);
9d5c8243 5666
40a914fa
AD
5667 pci_enable_wake(pdev, PCI_D3hot, 0);
5668 pci_enable_wake(pdev, PCI_D3cold, 0);
9d5c8243 5669
40a914fa
AD
5670 igb_reset(adapter);
5671 wr32(E1000_WUS, ~0);
5672 result = PCI_ERS_RESULT_RECOVERED;
5673 }
9d5c8243 5674
ea943d41
JK
5675 err = pci_cleanup_aer_uncorrect_error_status(pdev);
5676 if (err) {
5677 dev_err(&pdev->dev, "pci_cleanup_aer_uncorrect_error_status "
5678 "failed 0x%0x\n", err);
5679 /* non-fatal, continue */
5680 }
40a914fa
AD
5681
5682 return result;
9d5c8243
AK
5683}
5684
5685/**
5686 * igb_io_resume - called when traffic can start flowing again.
5687 * @pdev: Pointer to PCI device
5688 *
5689 * This callback is called when the error recovery driver tells us that
5690 * its OK to resume normal operation. Implementation resembles the
5691 * second-half of the igb_resume routine.
5692 */
5693static void igb_io_resume(struct pci_dev *pdev)
5694{
5695 struct net_device *netdev = pci_get_drvdata(pdev);
5696 struct igb_adapter *adapter = netdev_priv(netdev);
5697
9d5c8243
AK
5698 if (netif_running(netdev)) {
5699 if (igb_up(adapter)) {
5700 dev_err(&pdev->dev, "igb_up failed after reset\n");
5701 return;
5702 }
5703 }
5704
5705 netif_device_attach(netdev);
5706
5707 /* let the f/w know that the h/w is now under the control of the
5708 * driver. */
5709 igb_get_hw_control(adapter);
9d5c8243
AK
5710}
5711
26ad9178
AD
5712static void igb_rar_set_qsel(struct igb_adapter *adapter, u8 *addr, u32 index,
5713 u8 qsel)
5714{
5715 u32 rar_low, rar_high;
5716 struct e1000_hw *hw = &adapter->hw;
5717
5718 /* HW expects these in little endian so we reverse the byte order
5719 * from network order (big endian) to little endian
5720 */
5721 rar_low = ((u32) addr[0] | ((u32) addr[1] << 8) |
5722 ((u32) addr[2] << 16) | ((u32) addr[3] << 24));
5723 rar_high = ((u32) addr[4] | ((u32) addr[5] << 8));
5724
5725 /* Indicate to hardware the Address is Valid. */
5726 rar_high |= E1000_RAH_AV;
5727
5728 if (hw->mac.type == e1000_82575)
5729 rar_high |= E1000_RAH_POOL_1 * qsel;
5730 else
5731 rar_high |= E1000_RAH_POOL_1 << qsel;
5732
5733 wr32(E1000_RAL(index), rar_low);
5734 wrfl();
5735 wr32(E1000_RAH(index), rar_high);
5736 wrfl();
5737}
5738
4ae196df
AD
5739static int igb_set_vf_mac(struct igb_adapter *adapter,
5740 int vf, unsigned char *mac_addr)
5741{
5742 struct e1000_hw *hw = &adapter->hw;
ff41f8dc
AD
5743 /* VF MAC addresses start at end of receive addresses and moves
5744 * torwards the first, as a result a collision should not be possible */
5745 int rar_entry = hw->mac.rar_entry_count - (vf + 1);
4ae196df 5746
37680117 5747 memcpy(adapter->vf_data[vf].vf_mac_addresses, mac_addr, ETH_ALEN);
4ae196df 5748
26ad9178 5749 igb_rar_set_qsel(adapter, mac_addr, rar_entry, vf);
4ae196df
AD
5750
5751 return 0;
5752}
5753
5754static void igb_vmm_control(struct igb_adapter *adapter)
5755{
5756 struct e1000_hw *hw = &adapter->hw;
10d8e907 5757 u32 reg;
4ae196df 5758
d4960307
AD
5759 /* replication is not supported for 82575 */
5760 if (hw->mac.type == e1000_82575)
4ae196df
AD
5761 return;
5762
10d8e907
AD
5763 /* enable replication vlan tag stripping */
5764 reg = rd32(E1000_RPLOLR);
5765 reg |= E1000_RPLOLR_STRVLAN;
5766 wr32(E1000_RPLOLR, reg);
5767
5768 /* notify HW that the MAC is adding vlan tags */
5769 reg = rd32(E1000_DTXCTL);
5770 reg |= E1000_DTXCTL_VLAN_ADDED;
5771 wr32(E1000_DTXCTL, reg);
5772
d4960307
AD
5773 if (adapter->vfs_allocated_count) {
5774 igb_vmdq_set_loopback_pf(hw, true);
5775 igb_vmdq_set_replication_pf(hw, true);
5776 } else {
5777 igb_vmdq_set_loopback_pf(hw, false);
5778 igb_vmdq_set_replication_pf(hw, false);
5779 }
4ae196df
AD
5780}
5781
9d5c8243 5782/* igb_main.c */