]> bbs.cooldavid.org Git - net-next-2.6.git/blame - drivers/net/igb/igb_main.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
igb: move alloc_failed and csum_err stats into per rx-ring stat
[net-next-2.6.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 *);
9d5c8243
AK		 85void igb_update_stats(struct igb_adapter *);
86static int igb_probe(struct pci_dev *, const struct pci_device_id *);
87static void __devexit igb_remove(struct pci_dev *pdev);
88static int igb_sw_init(struct igb_adapter *);
89static int igb_open(struct net_device *);
90static int igb_close(struct net_device *);
91static void igb_configure_tx(struct igb_adapter *);
92static void igb_configure_rx(struct igb_adapter *);
85b430b4 93static void igb_setup_tctl(struct igb_adapter *);
9d5c8243
AK		 94static void igb_setup_rctl(struct igb_adapter *);
95static void igb_clean_all_tx_rings(struct igb_adapter *);
96static void igb_clean_all_rx_rings(struct igb_adapter *);
3b644cf6
MW		 97static void igb_clean_tx_ring(struct igb_ring *);
98static void igb_clean_rx_ring(struct igb_ring *);
ff41f8dc 99static void igb_set_rx_mode(struct net_device *);
9d5c8243
AK		 100static void igb_update_phy_info(unsigned long);
101static void igb_watchdog(unsigned long);
102static void igb_watchdog_task(struct work_struct *);
3b29a56d
SH		 103static netdev_tx_t igb_xmit_frame_ring_adv(struct sk_buff *,
104 struct net_device *,
105 struct igb_ring *);
106static netdev_tx_t igb_xmit_frame_adv(struct sk_buff *skb,
107 struct net_device *);
9d5c8243
AK		 108static struct net_device_stats *igb_get_stats(struct net_device *);
109static int igb_change_mtu(struct net_device *, int);
110static int igb_set_mac(struct net_device *, void *);
68d480c4 111static void igb_set_uta(struct igb_adapter *adapter);
9d5c8243
AK		 112static irqreturn_t igb_intr(int irq, void *);
113static irqreturn_t igb_intr_msi(int irq, void *);
114static irqreturn_t igb_msix_other(int irq, void *);
047e0030 115static irqreturn_t igb_msix_ring(int irq, void *);
421e02f0 116#ifdef CONFIG_IGB_DCA
047e0030 117static void igb_update_dca(struct igb_q_vector *);
fe4506b6 118static void igb_setup_dca(struct igb_adapter *);
421e02f0 119#endif /* CONFIG_IGB_DCA */
047e0030 120static bool igb_clean_tx_irq(struct igb_q_vector *);
661086df 121static int igb_poll(struct napi_struct *, int);
047e0030 122static bool igb_clean_rx_irq_adv(struct igb_q_vector *, int *, int);
3b644cf6 123static void igb_alloc_rx_buffers_adv(struct igb_ring *, int);
9d5c8243
AK		 124static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
125static void igb_tx_timeout(struct net_device *);
126static void igb_reset_task(struct work_struct *);
127static void igb_vlan_rx_register(struct net_device *, struct vlan_group *);
128static void igb_vlan_rx_add_vid(struct net_device *, u16);
129static void igb_vlan_rx_kill_vid(struct net_device *, u16);
130static void igb_restore_vlan(struct igb_adapter *);
26ad9178 131static void igb_rar_set_qsel(struct igb_adapter *, u8 *, u32 , u8);
4ae196df
AD		 132static void igb_ping_all_vfs(struct igb_adapter *);
133static void igb_msg_task(struct igb_adapter *);
134static int igb_rcv_msg_from_vf(struct igb_adapter *, u32);
4ae196df 135static void igb_vmm_control(struct igb_adapter *);
4ae196df
AD		 136static int igb_set_vf_mac(struct igb_adapter *adapter, int, unsigned char *);
137static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
9d5c8243 138
c8159b2d
ED		 139static inline void igb_set_vmolr(struct e1000_hw *hw, int vfn)
140{
141 u32 reg_data;
142
143 reg_data = rd32(E1000_VMOLR(vfn));
144 reg_data |= E1000_VMOLR_BAM | /* Accept broadcast */
c8159b2d
ED		 145 E1000_VMOLR_ROMPE | /* Accept packets matched in MTA */
146 E1000_VMOLR_AUPE | /* Accept untagged packets */
147 E1000_VMOLR_STRVLAN; /* Strip vlan tags */
148 wr32(E1000_VMOLR(vfn), reg_data);
149}
150
151static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
152 int vfn)
153{
154 struct e1000_hw *hw = &adapter->hw;
155 u32 vmolr;
156
ae641bdc
AD		 157 /* if it isn't the PF check to see if VFs are enabled and
158 * increase the size to support vlan tags */
159 if (vfn < adapter->vfs_allocated_count &&
160 adapter->vf_data[vfn].vlans_enabled)
161 size += VLAN_TAG_SIZE;
162
c8159b2d
ED		 163 vmolr = rd32(E1000_VMOLR(vfn));
164 vmolr &= ~E1000_VMOLR_RLPML_MASK;
165 vmolr |= size | E1000_VMOLR_LPE;
166 wr32(E1000_VMOLR(vfn), vmolr);
167
168 return 0;
169}
170
9d5c8243 171#ifdef CONFIG_PM
3fe7c4c9 172static int igb_suspend(struct pci_dev *, pm_message_t);
9d5c8243
AK		 173static int igb_resume(struct pci_dev *);
174#endif
175static void igb_shutdown(struct pci_dev *);
421e02f0 176#ifdef CONFIG_IGB_DCA
fe4506b6
JC		 177static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
178static struct notifier_block dca_notifier = {
179 .notifier_call = igb_notify_dca,
180 .next = NULL,
181 .priority = 0
182};
183#endif
9d5c8243
AK		 184#ifdef CONFIG_NET_POLL_CONTROLLER
185/* for netdump / net console */
186static void igb_netpoll(struct net_device *);
187#endif
37680117 188#ifdef CONFIG_PCI_IOV
2a3abf6d
AD		 189static unsigned int max_vfs = 0;
190module_param(max_vfs, uint, 0);
191MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate "
192 "per physical function");
193#endif /* CONFIG_PCI_IOV */
194
9d5c8243
AK		 195static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
196 pci_channel_state_t);
197static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
198static void igb_io_resume(struct pci_dev *);
199
200static struct pci_error_handlers igb_err_handler = {
201 .error_detected = igb_io_error_detected,
202 .slot_reset = igb_io_slot_reset,
203 .resume = igb_io_resume,
204};
205
206
207static struct pci_driver igb_driver = {
208 .name = igb_driver_name,
209 .id_table = igb_pci_tbl,
210 .probe = igb_probe,
211 .remove = __devexit_p(igb_remove),
212#ifdef CONFIG_PM
213 /* Power Managment Hooks */
214 .suspend = igb_suspend,
215 .resume = igb_resume,
216#endif
217 .shutdown = igb_shutdown,
218 .err_handler = &igb_err_handler
219};
220
7dfc16fa
AD		 221static int global_quad_port_a; /* global quad port a indication */
222
9d5c8243
AK		 223MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
224MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
225MODULE_LICENSE("GPL");
226MODULE_VERSION(DRV_VERSION);
227
38c845c7
PO		 228/**
229 * Scale the NIC clock cycle by a large factor so that
230 * relatively small clock corrections can be added or
231 * substracted at each clock tick. The drawbacks of a
232 * large factor are a) that the clock register overflows
233 * more quickly (not such a big deal) and b) that the
234 * increment per tick has to fit into 24 bits.
235 *
236 * Note that
237 * TIMINCA = IGB_TSYNC_CYCLE_TIME_IN_NANOSECONDS *
238 * IGB_TSYNC_SCALE
239 * TIMINCA += TIMINCA * adjustment [ppm] / 1e9
240 *
241 * The base scale factor is intentionally a power of two
242 * so that the division in %struct timecounter can be done with
243 * a shift.
244 */
245#define IGB_TSYNC_SHIFT (19)
246#define IGB_TSYNC_SCALE (1<<IGB_TSYNC_SHIFT)
247
248/**
249 * The duration of one clock cycle of the NIC.
250 *
251 * @todo This hard-coded value is part of the specification and might change
252 * in future hardware revisions. Add revision check.
253 */
254#define IGB_TSYNC_CYCLE_TIME_IN_NANOSECONDS 16
255
256#if (IGB_TSYNC_SCALE * IGB_TSYNC_CYCLE_TIME_IN_NANOSECONDS) >= (1<<24)
257# error IGB_TSYNC_SCALE and/or IGB_TSYNC_CYCLE_TIME_IN_NANOSECONDS are too large to fit into TIMINCA
258#endif
259
260/**
261 * igb_read_clock - read raw cycle counter (to be used by time counter)
262 */
263static cycle_t igb_read_clock(const struct cyclecounter *tc)
264{
265 struct igb_adapter *adapter =
266 container_of(tc, struct igb_adapter, cycles);
267 struct e1000_hw *hw = &adapter->hw;
268 u64 stamp;
269
270 stamp = rd32(E1000_SYSTIML);
271 stamp |= (u64)rd32(E1000_SYSTIMH) << 32ULL;
272
273 return stamp;
274}
275
9d5c8243
AK		 276#ifdef DEBUG
277/**
278 * igb_get_hw_dev_name - return device name string
279 * used by hardware layer to print debugging information
280 **/
281char *igb_get_hw_dev_name(struct e1000_hw *hw)
282{
283 struct igb_adapter *adapter = hw->back;
284 return adapter->netdev->name;
285}
38c845c7
PO		 286
287/**
288 * igb_get_time_str - format current NIC and system time as string
289 */
290static char *igb_get_time_str(struct igb_adapter *adapter,
291 char buffer[160])
292{
293 cycle_t hw = adapter->cycles.read(&adapter->cycles);
294 struct timespec nic = ns_to_timespec(timecounter_read(&adapter->clock));
295 struct timespec sys;
296 struct timespec delta;
297 getnstimeofday(&sys);
298
299 delta = timespec_sub(nic, sys);
300
301 sprintf(buffer,
33af6bcc
PO		 302 "HW %llu, NIC %ld.%09lus, SYS %ld.%09lus, NIC-SYS %lds + %09luns",
303 hw,
38c845c7
PO		 304 (long)nic.tv_sec, nic.tv_nsec,
305 (long)sys.tv_sec, sys.tv_nsec,
306 (long)delta.tv_sec, delta.tv_nsec);
307
308 return buffer;
309}
9d5c8243
AK		 310#endif
311
c493ea45
AD		 312/**
313 * igb_desc_unused - calculate if we have unused descriptors
314 **/
315static int igb_desc_unused(struct igb_ring *ring)
316{
317 if (ring->next_to_clean > ring->next_to_use)
318 return ring->next_to_clean - ring->next_to_use - 1;
319
320 return ring->count + ring->next_to_clean - ring->next_to_use - 1;
321}
322
9d5c8243
AK		 323/**
324 * igb_init_module - Driver Registration Routine
325 *
326 * igb_init_module is the first routine called when the driver is
327 * loaded. All it does is register with the PCI subsystem.
328 **/
329static int __init igb_init_module(void)
330{
331 int ret;
332 printk(KERN_INFO "%s - version %s\n",
333 igb_driver_string, igb_driver_version);
334
335 printk(KERN_INFO "%s\n", igb_copyright);
336
7dfc16fa
AD		 337 global_quad_port_a = 0;
338
421e02f0 339#ifdef CONFIG_IGB_DCA
fe4506b6
JC		 340 dca_register_notify(&dca_notifier);
341#endif
bbd98fe4
AD		 342
343 ret = pci_register_driver(&igb_driver);
9d5c8243
AK		 344 return ret;
345}
346
347module_init(igb_init_module);
348
349/**
350 * igb_exit_module - Driver Exit Cleanup Routine
351 *
352 * igb_exit_module is called just before the driver is removed
353 * from memory.
354 **/
355static void __exit igb_exit_module(void)
356{
421e02f0 357#ifdef CONFIG_IGB_DCA
fe4506b6
JC		 358 dca_unregister_notify(&dca_notifier);
359#endif
9d5c8243
AK		 360 pci_unregister_driver(&igb_driver);
361}
362
363module_exit(igb_exit_module);
364
26bc19ec
AD		 365#define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
366/**
367 * igb_cache_ring_register - Descriptor ring to register mapping
368 * @adapter: board private structure to initialize
369 *
370 * Once we know the feature-set enabled for the device, we'll cache
371 * the register offset the descriptor ring is assigned to.
372 **/
373static void igb_cache_ring_register(struct igb_adapter *adapter)
374{
375 int i;
047e0030 376 u32 rbase_offset = adapter->vfs_allocated_count;
26bc19ec
AD		 377
378 switch (adapter->hw.mac.type) {
379 case e1000_82576:
380 /* The queues are allocated for virtualization such that VF 0
381 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
382 * In order to avoid collision we start at the first free queue
383 * and continue consuming queues in the same sequence
384 */
385 for (i = 0; i < adapter->num_rx_queues; i++)
1bfaf07b
AD		 386 adapter->rx_ring[i].reg_idx = rbase_offset +
387 Q_IDX_82576(i);
26bc19ec 388 for (i = 0; i < adapter->num_tx_queues; i++)
1bfaf07b
AD		 389 adapter->tx_ring[i].reg_idx = rbase_offset +
390 Q_IDX_82576(i);
26bc19ec
AD		 391 break;
392 case e1000_82575:
393 default:
394 for (i = 0; i < adapter->num_rx_queues; i++)
395 adapter->rx_ring[i].reg_idx = i;
396 for (i = 0; i < adapter->num_tx_queues; i++)
397 adapter->tx_ring[i].reg_idx = i;
398 break;
399 }
400}
401
047e0030
AD		 402static void igb_free_queues(struct igb_adapter *adapter)
403{
404 kfree(adapter->tx_ring);
405 kfree(adapter->rx_ring);
406
407 adapter->tx_ring = NULL;
408 adapter->rx_ring = NULL;
409
410 adapter->num_rx_queues = 0;
411 adapter->num_tx_queues = 0;
412}
413
9d5c8243
AK		 414/**
415 * igb_alloc_queues - Allocate memory for all rings
416 * @adapter: board private structure to initialize
417 *
418 * We allocate one ring per queue at run-time since we don't know the
419 * number of queues at compile-time.
420 **/
421static int igb_alloc_queues(struct igb_adapter *adapter)
422{
423 int i;
424
425 adapter->tx_ring = kcalloc(adapter->num_tx_queues,
426 sizeof(struct igb_ring), GFP_KERNEL);
427 if (!adapter->tx_ring)
047e0030 428 goto err;
9d5c8243
AK		 429
430 adapter->rx_ring = kcalloc(adapter->num_rx_queues,
431 sizeof(struct igb_ring), GFP_KERNEL);
047e0030
AD		 432 if (!adapter->rx_ring)
433 goto err;
6eb5a7f1 434
661086df
PWJ		 435 for (i = 0; i < adapter->num_tx_queues; i++) {
436 struct igb_ring *ring = &(adapter->tx_ring[i]);
68fd9910 437 ring->count = adapter->tx_ring_count;
661086df 438 ring->queue_index = i;
80785298 439 ring->pdev = adapter->pdev;
661086df 440 }
9d5c8243
AK		 441 for (i = 0; i < adapter->num_rx_queues; i++) {
442 struct igb_ring *ring = &(adapter->rx_ring[i]);
68fd9910 443 ring->count = adapter->rx_ring_count;
844290e5 444 ring->queue_index = i;
80785298 445 ring->pdev = adapter->pdev;
4c844851 446 ring->rx_buffer_len = MAXIMUM_ETHERNET_VLAN_SIZE;
9d5c8243 447 }
26bc19ec
AD		 448
449 igb_cache_ring_register(adapter);
9d5c8243 450
047e0030 451 return 0;
a88f10ec 452
047e0030
AD		 453err:
454 igb_free_queues(adapter);
d1a8c9e1 455
047e0030 456 return -ENOMEM;
a88f10ec
AD		 457}
458
9d5c8243 459#define IGB_N0_QUEUE -1
047e0030 460static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
9d5c8243
AK		 461{
462 u32 msixbm = 0;
047e0030 463 struct igb_adapter *adapter = q_vector->adapter;
9d5c8243 464 struct e1000_hw *hw = &adapter->hw;
2d064c06 465 u32 ivar, index;
047e0030
AD		 466 int rx_queue = IGB_N0_QUEUE;
467 int tx_queue = IGB_N0_QUEUE;
468
469 if (q_vector->rx_ring)
470 rx_queue = q_vector->rx_ring->reg_idx;
471 if (q_vector->tx_ring)
472 tx_queue = q_vector->tx_ring->reg_idx;
2d064c06
AD		 473
474 switch (hw->mac.type) {
475 case e1000_82575:
9d5c8243
AK		 476 /* The 82575 assigns vectors using a bitmask, which matches the
477 bitmask for the EICR/EIMS/EIMC registers. To assign one
478 or more queues to a vector, we write the appropriate bits
479 into the MSIXBM register for that vector. */
047e0030 480 if (rx_queue > IGB_N0_QUEUE)
9d5c8243 481 msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
047e0030 482 if (tx_queue > IGB_N0_QUEUE)
9d5c8243 483 msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
9d5c8243 484 array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
047e0030 485 q_vector->eims_value = msixbm;
2d064c06
AD		 486 break;
487 case e1000_82576:
26bc19ec 488 /* 82576 uses a table-based method for assigning vectors.
2d064c06
AD		 489 Each queue has a single entry in the table to which we write
490 a vector number along with a "valid" bit. Sadly, the layout
491 of the table is somewhat counterintuitive. */
492 if (rx_queue > IGB_N0_QUEUE) {
047e0030 493 index = (rx_queue & 0x7);
2d064c06 494 ivar = array_rd32(E1000_IVAR0, index);
047e0030 495 if (rx_queue < 8) {
26bc19ec
AD		 496 /* vector goes into low byte of register */
497 ivar = ivar & 0xFFFFFF00;
498 ivar |= msix_vector | E1000_IVAR_VALID;
047e0030
AD		 499 } else {
500 /* vector goes into third byte of register */
501 ivar = ivar & 0xFF00FFFF;
502 ivar |= (msix_vector | E1000_IVAR_VALID) << 16;
2d064c06 503 }
2d064c06
AD		 504 array_wr32(E1000_IVAR0, index, ivar);
505 }
506 if (tx_queue > IGB_N0_QUEUE) {
047e0030 507 index = (tx_queue & 0x7);
2d064c06 508 ivar = array_rd32(E1000_IVAR0, index);
047e0030 509 if (tx_queue < 8) {
26bc19ec
AD		 510 /* vector goes into second byte of register */
511 ivar = ivar & 0xFFFF00FF;
512 ivar |= (msix_vector | E1000_IVAR_VALID) << 8;
047e0030
AD		 513 } else {
514 /* vector goes into high byte of register */
515 ivar = ivar & 0x00FFFFFF;
516 ivar |= (msix_vector | E1000_IVAR_VALID) << 24;
2d064c06 517 }
2d064c06
AD		 518 array_wr32(E1000_IVAR0, index, ivar);
519 }
047e0030 520 q_vector->eims_value = 1 << msix_vector;
2d064c06
AD		 521 break;
522 default:
523 BUG();
524 break;
525 }
9d5c8243
AK		 526}
527
528/**
529 * igb_configure_msix - Configure MSI-X hardware
530 *
531 * igb_configure_msix sets up the hardware to properly
532 * generate MSI-X interrupts.
533 **/
534static void igb_configure_msix(struct igb_adapter *adapter)
535{
536 u32 tmp;
537 int i, vector = 0;
538 struct e1000_hw *hw = &adapter->hw;
539
540 adapter->eims_enable_mask = 0;
9d5c8243
AK		 541
542 /* set vector for other causes, i.e. link changes */
2d064c06
AD		 543 switch (hw->mac.type) {
544 case e1000_82575:
9d5c8243
AK		 545 tmp = rd32(E1000_CTRL_EXT);
546 /* enable MSI-X PBA support*/
547 tmp |= E1000_CTRL_EXT_PBA_CLR;
548
549 /* Auto-Mask interrupts upon ICR read. */
550 tmp |= E1000_CTRL_EXT_EIAME;
551 tmp |= E1000_CTRL_EXT_IRCA;
552
553 wr32(E1000_CTRL_EXT, tmp);
047e0030
AD		 554
555 /* enable msix_other interrupt */
556 array_wr32(E1000_MSIXBM(0), vector++,
557 E1000_EIMS_OTHER);
844290e5 558 adapter->eims_other = E1000_EIMS_OTHER;
9d5c8243 559
2d064c06
AD		 560 break;
561
562 case e1000_82576:
047e0030
AD		 563 /* Turn on MSI-X capability first, or our settings
564 * won't stick. And it will take days to debug. */
565 wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
566 E1000_GPIE_PBA | E1000_GPIE_EIAME |
567 E1000_GPIE_NSICR);
568
569 /* enable msix_other interrupt */
570 adapter->eims_other = 1 << vector;
2d064c06 571 tmp = (vector++ | E1000_IVAR_VALID) << 8;
2d064c06 572
047e0030 573 wr32(E1000_IVAR_MISC, tmp);
2d064c06
AD		 574 break;
575 default:
576 /* do nothing, since nothing else supports MSI-X */
577 break;
578 } /* switch (hw->mac.type) */
047e0030
AD		 579
580 adapter->eims_enable_mask |= adapter->eims_other;
581
582 for (i = 0; i < adapter->num_q_vectors; i++) {
583 struct igb_q_vector *q_vector = adapter->q_vector[i];
584 igb_assign_vector(q_vector, vector++);
585 adapter->eims_enable_mask |= q_vector->eims_value;
586 }
587
9d5c8243
AK		 588 wrfl();
589}
590
591/**
592 * igb_request_msix - Initialize MSI-X interrupts
593 *
594 * igb_request_msix allocates MSI-X vectors and requests interrupts from the
595 * kernel.
596 **/
597static int igb_request_msix(struct igb_adapter *adapter)
598{
599 struct net_device *netdev = adapter->netdev;
047e0030 600 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK		 601 int i, err = 0, vector = 0;
602
047e0030
AD		 603 err = request_irq(adapter->msix_entries[vector].vector,
604 &igb_msix_other, 0, netdev->name, adapter);
605 if (err)
606 goto out;
607 vector++;
608
609 for (i = 0; i < adapter->num_q_vectors; i++) {
610 struct igb_q_vector *q_vector = adapter->q_vector[i];
611
612 q_vector->itr_register = hw->hw_addr + E1000_EITR(vector);
613
614 if (q_vector->rx_ring && q_vector->tx_ring)
615 sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
616 q_vector->rx_ring->queue_index);
617 else if (q_vector->tx_ring)
618 sprintf(q_vector->name, "%s-tx-%u", netdev->name,
619 q_vector->tx_ring->queue_index);
620 else if (q_vector->rx_ring)
621 sprintf(q_vector->name, "%s-rx-%u", netdev->name,
622 q_vector->rx_ring->queue_index);
9d5c8243 623 else
047e0030
AD		 624 sprintf(q_vector->name, "%s-unused", netdev->name);
625
9d5c8243 626 err = request_irq(adapter->msix_entries[vector].vector,
047e0030
AD		 627 &igb_msix_ring, 0, q_vector->name,
628 q_vector);
9d5c8243
AK		 629 if (err)
630 goto out;
9d5c8243
AK		 631 vector++;
632 }
633
9d5c8243
AK		 634 igb_configure_msix(adapter);
635 return 0;
636out:
637 return err;
638}
639
640static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
641{
642 if (adapter->msix_entries) {
643 pci_disable_msix(adapter->pdev);
644 kfree(adapter->msix_entries);
645 adapter->msix_entries = NULL;
047e0030 646 } else if (adapter->flags & IGB_FLAG_HAS_MSI) {
9d5c8243 647 pci_disable_msi(adapter->pdev);
047e0030 648 }
9d5c8243
AK		 649}
650
047e0030
AD		 651/**
652 * igb_free_q_vectors - Free memory allocated for interrupt vectors
653 * @adapter: board private structure to initialize
654 *
655 * This function frees the memory allocated to the q_vectors. In addition if
656 * NAPI is enabled it will delete any references to the NAPI struct prior
657 * to freeing the q_vector.
658 **/
659static void igb_free_q_vectors(struct igb_adapter *adapter)
660{
661 int v_idx;
662
663 for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
664 struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
665 adapter->q_vector[v_idx] = NULL;
666 netif_napi_del(&q_vector->napi);
667 kfree(q_vector);
668 }
669 adapter->num_q_vectors = 0;
670}
671
672/**
673 * igb_clear_interrupt_scheme - reset the device to a state of no interrupts
674 *
675 * This function resets the device so that it has 0 rx queues, tx queues, and
676 * MSI-X interrupts allocated.
677 */
678static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
679{
680 igb_free_queues(adapter);
681 igb_free_q_vectors(adapter);
682 igb_reset_interrupt_capability(adapter);
683}
9d5c8243
AK		 684
685/**
686 * igb_set_interrupt_capability - set MSI or MSI-X if supported
687 *
688 * Attempt to configure interrupts using the best available
689 * capabilities of the hardware and kernel.
690 **/
691static void igb_set_interrupt_capability(struct igb_adapter *adapter)
692{
693 int err;
694 int numvecs, i;
695
83b7180d 696 /* Number of supported queues. */
83b7180d
AD		 697 adapter->num_rx_queues = min_t(u32, IGB_MAX_RX_QUEUES, num_online_cpus());
698 adapter->num_tx_queues = min_t(u32, IGB_MAX_TX_QUEUES, num_online_cpus());
699
047e0030
AD		 700 /* start with one vector for every rx queue */
701 numvecs = adapter->num_rx_queues;
702
703 /* if tx handler is seperate add 1 for every tx queue */
704 numvecs += adapter->num_tx_queues;
705
706 /* store the number of vectors reserved for queues */
707 adapter->num_q_vectors = numvecs;
708
709 /* add 1 vector for link status interrupts */
710 numvecs++;
9d5c8243
AK		 711 adapter->msix_entries = kcalloc(numvecs, sizeof(struct msix_entry),
712 GFP_KERNEL);
713 if (!adapter->msix_entries)
714 goto msi_only;
715
716 for (i = 0; i < numvecs; i++)
717 adapter->msix_entries[i].entry = i;
718
719 err = pci_enable_msix(adapter->pdev,
720 adapter->msix_entries,
721 numvecs);
722 if (err == 0)
34a20e89 723 goto out;
9d5c8243
AK		 724
725 igb_reset_interrupt_capability(adapter);
726
727 /* If we can't do MSI-X, try MSI */
728msi_only:
2a3abf6d
AD		 729#ifdef CONFIG_PCI_IOV
730 /* disable SR-IOV for non MSI-X configurations */
731 if (adapter->vf_data) {
732 struct e1000_hw *hw = &adapter->hw;
733 /* disable iov and allow time for transactions to clear */
734 pci_disable_sriov(adapter->pdev);
735 msleep(500);
736
737 kfree(adapter->vf_data);
738 adapter->vf_data = NULL;
739 wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
740 msleep(100);
741 dev_info(&adapter->pdev->dev, "IOV Disabled\n");
742 }
743#endif
9d5c8243 744 adapter->num_rx_queues = 1;
661086df 745 adapter->num_tx_queues = 1;
047e0030 746 adapter->num_q_vectors = 1;
9d5c8243 747 if (!pci_enable_msi(adapter->pdev))
7dfc16fa 748 adapter->flags |= IGB_FLAG_HAS_MSI;
34a20e89 749out:
661086df 750 /* Notify the stack of the (possibly) reduced Tx Queue count. */
fd2ea0a7 751 adapter->netdev->real_num_tx_queues = adapter->num_tx_queues;
9d5c8243
AK		 752 return;
753}
754
047e0030
AD		 755/**
756 * igb_alloc_q_vectors - Allocate memory for interrupt vectors
757 * @adapter: board private structure to initialize
758 *
759 * We allocate one q_vector per queue interrupt. If allocation fails we
760 * return -ENOMEM.
761 **/
762static int igb_alloc_q_vectors(struct igb_adapter *adapter)
763{
764 struct igb_q_vector *q_vector;
765 struct e1000_hw *hw = &adapter->hw;
766 int v_idx;
767
768 for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
769 q_vector = kzalloc(sizeof(struct igb_q_vector), GFP_KERNEL);
770 if (!q_vector)
771 goto err_out;
772 q_vector->adapter = adapter;
773 q_vector->itr_shift = (hw->mac.type == e1000_82575) ? 16 : 0;
774 q_vector->itr_register = hw->hw_addr + E1000_EITR(0);
775 q_vector->itr_val = IGB_START_ITR;
776 q_vector->set_itr = 1;
777 netif_napi_add(adapter->netdev, &q_vector->napi, igb_poll, 64);
778 adapter->q_vector[v_idx] = q_vector;
779 }
780 return 0;
781
782err_out:
783 while (v_idx) {
784 v_idx--;
785 q_vector = adapter->q_vector[v_idx];
786 netif_napi_del(&q_vector->napi);
787 kfree(q_vector);
788 adapter->q_vector[v_idx] = NULL;
789 }
790 return -ENOMEM;
791}
792
793static void igb_map_rx_ring_to_vector(struct igb_adapter *adapter,
794 int ring_idx, int v_idx)
795{
796 struct igb_q_vector *q_vector;
797
798 q_vector = adapter->q_vector[v_idx];
799 q_vector->rx_ring = &adapter->rx_ring[ring_idx];
800 q_vector->rx_ring->q_vector = q_vector;
801 q_vector->itr_val = adapter->itr;
802}
803
804static void igb_map_tx_ring_to_vector(struct igb_adapter *adapter,
805 int ring_idx, int v_idx)
806{
807 struct igb_q_vector *q_vector;
808
809 q_vector = adapter->q_vector[v_idx];
810 q_vector->tx_ring = &adapter->tx_ring[ring_idx];
811 q_vector->tx_ring->q_vector = q_vector;
812 q_vector->itr_val = adapter->itr;
813}
814
815/**
816 * igb_map_ring_to_vector - maps allocated queues to vectors
817 *
818 * This function maps the recently allocated queues to vectors.
819 **/
820static int igb_map_ring_to_vector(struct igb_adapter *adapter)
821{
822 int i;
823 int v_idx = 0;
824
825 if ((adapter->num_q_vectors < adapter->num_rx_queues) ||
826 (adapter->num_q_vectors < adapter->num_tx_queues))
827 return -ENOMEM;
828
829 if (adapter->num_q_vectors >=
830 (adapter->num_rx_queues + adapter->num_tx_queues)) {
831 for (i = 0; i < adapter->num_rx_queues; i++)
832 igb_map_rx_ring_to_vector(adapter, i, v_idx++);
833 for (i = 0; i < adapter->num_tx_queues; i++)
834 igb_map_tx_ring_to_vector(adapter, i, v_idx++);
835 } else {
836 for (i = 0; i < adapter->num_rx_queues; i++) {
837 if (i < adapter->num_tx_queues)
838 igb_map_tx_ring_to_vector(adapter, i, v_idx);
839 igb_map_rx_ring_to_vector(adapter, i, v_idx++);
840 }
841 for (; i < adapter->num_tx_queues; i++)
842 igb_map_tx_ring_to_vector(adapter, i, v_idx++);
843 }
844 return 0;
845}
846
847/**
848 * igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
849 *
850 * This function initializes the interrupts and allocates all of the queues.
851 **/
852static int igb_init_interrupt_scheme(struct igb_adapter *adapter)
853{
854 struct pci_dev *pdev = adapter->pdev;
855 int err;
856
857 igb_set_interrupt_capability(adapter);
858
859 err = igb_alloc_q_vectors(adapter);
860 if (err) {
861 dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
862 goto err_alloc_q_vectors;
863 }
864
865 err = igb_alloc_queues(adapter);
866 if (err) {
867 dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
868 goto err_alloc_queues;
869 }
870
871 err = igb_map_ring_to_vector(adapter);
872 if (err) {
873 dev_err(&pdev->dev, "Invalid q_vector to ring mapping\n");
874 goto err_map_queues;
875 }
876
877
878 return 0;
879err_map_queues:
880 igb_free_queues(adapter);
881err_alloc_queues:
882 igb_free_q_vectors(adapter);
883err_alloc_q_vectors:
884 igb_reset_interrupt_capability(adapter);
885 return err;
886}
887
9d5c8243
AK		 888/**
889 * igb_request_irq - initialize interrupts
890 *
891 * Attempts to configure interrupts using the best available
892 * capabilities of the hardware and kernel.
893 **/
894static int igb_request_irq(struct igb_adapter *adapter)
895{
896 struct net_device *netdev = adapter->netdev;
047e0030 897 struct pci_dev *pdev = adapter->pdev;
9d5c8243
AK		 898 struct e1000_hw *hw = &adapter->hw;
899 int err = 0;
900
901 if (adapter->msix_entries) {
902 err = igb_request_msix(adapter);
844290e5 903 if (!err)
9d5c8243 904 goto request_done;
9d5c8243 905 /* fall back to MSI */
047e0030 906 igb_clear_interrupt_scheme(adapter);
9d5c8243 907 if (!pci_enable_msi(adapter->pdev))
7dfc16fa 908 adapter->flags |= IGB_FLAG_HAS_MSI;
9d5c8243
AK		 909 igb_free_all_tx_resources(adapter);
910 igb_free_all_rx_resources(adapter);
047e0030 911 adapter->num_tx_queues = 1;
9d5c8243 912 adapter->num_rx_queues = 1;
047e0030
AD		 913 adapter->num_q_vectors = 1;
914 err = igb_alloc_q_vectors(adapter);
915 if (err) {
916 dev_err(&pdev->dev,
917 "Unable to allocate memory for vectors\n");
918 goto request_done;
919 }
920 err = igb_alloc_queues(adapter);
921 if (err) {
922 dev_err(&pdev->dev,
923 "Unable to allocate memory for queues\n");
924 igb_free_q_vectors(adapter);
925 goto request_done;
926 }
927 igb_setup_all_tx_resources(adapter);
928 igb_setup_all_rx_resources(adapter);
844290e5 929 } else {
2d064c06
AD		 930 switch (hw->mac.type) {
931 case e1000_82575:
932 wr32(E1000_MSIXBM(0),
047e0030
AD		 933 (E1000_EICR_RX_QUEUE0 |
934 E1000_EICR_TX_QUEUE0 |
935 E1000_EIMS_OTHER));
2d064c06
AD		 936 break;
937 case e1000_82576:
938 wr32(E1000_IVAR0, E1000_IVAR_VALID);
939 break;
940 default:
941 break;
942 }
9d5c8243 943 }
844290e5 944
7dfc16fa 945 if (adapter->flags & IGB_FLAG_HAS_MSI) {
9d5c8243 946 err = request_irq(adapter->pdev->irq, &igb_intr_msi, 0,
047e0030 947 netdev->name, adapter);
9d5c8243
AK		 948 if (!err)
949 goto request_done;
047e0030 950
9d5c8243
AK		 951 /* fall back to legacy interrupts */
952 igb_reset_interrupt_capability(adapter);
7dfc16fa 953 adapter->flags &= ~IGB_FLAG_HAS_MSI;
9d5c8243
AK		 954 }
955
956 err = request_irq(adapter->pdev->irq, &igb_intr, IRQF_SHARED,
047e0030 957 netdev->name, adapter);
9d5c8243 958
6cb5e577 959 if (err)
9d5c8243
AK		 960 dev_err(&adapter->pdev->dev, "Error %d getting interrupt\n",
961 err);
9d5c8243
AK		 962
963request_done:
964 return err;
965}
966
967static void igb_free_irq(struct igb_adapter *adapter)
968{
9d5c8243
AK		 969 if (adapter->msix_entries) {
970 int vector = 0, i;
971
047e0030 972 free_irq(adapter->msix_entries[vector++].vector, adapter);
9d5c8243 973
047e0030
AD		 974 for (i = 0; i < adapter->num_q_vectors; i++) {
975 struct igb_q_vector *q_vector = adapter->q_vector[i];
976 free_irq(adapter->msix_entries[vector++].vector,
977 q_vector);
978 }
979 } else {
980 free_irq(adapter->pdev->irq, adapter);
9d5c8243 981 }
9d5c8243
AK		 982}
983
984/**
985 * igb_irq_disable - Mask off interrupt generation on the NIC
986 * @adapter: board private structure
987 **/
988static void igb_irq_disable(struct igb_adapter *adapter)
989{
990 struct e1000_hw *hw = &adapter->hw;
991
992 if (adapter->msix_entries) {
2dfd1212
AD		 993 u32 regval = rd32(E1000_EIAM);
994 wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
995 wr32(E1000_EIMC, adapter->eims_enable_mask);
996 regval = rd32(E1000_EIAC);
997 wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
9d5c8243 998 }
844290e5
PW		 999
1000 wr32(E1000_IAM, 0);
9d5c8243
AK		 1001 wr32(E1000_IMC, ~0);
1002 wrfl();
1003 synchronize_irq(adapter->pdev->irq);
1004}
1005
1006/**
1007 * igb_irq_enable - Enable default interrupt generation settings
1008 * @adapter: board private structure
1009 **/
1010static void igb_irq_enable(struct igb_adapter *adapter)
1011{
1012 struct e1000_hw *hw = &adapter->hw;
1013
1014 if (adapter->msix_entries) {
2dfd1212
AD		 1015 u32 regval = rd32(E1000_EIAC);
1016 wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
1017 regval = rd32(E1000_EIAM);
1018 wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
844290e5 1019 wr32(E1000_EIMS, adapter->eims_enable_mask);
4ae196df
AD		 1020 if (adapter->vfs_allocated_count)
1021 wr32(E1000_MBVFIMR, 0xFF);
1022 wr32(E1000_IMS, (E1000_IMS_LSC | E1000_IMS_VMMB |
1023 E1000_IMS_DOUTSYNC));
844290e5
PW		 1024 } else {
1025 wr32(E1000_IMS, IMS_ENABLE_MASK);
1026 wr32(E1000_IAM, IMS_ENABLE_MASK);
1027 }
9d5c8243
AK		 1028}
1029
1030static void igb_update_mng_vlan(struct igb_adapter *adapter)
1031{
1032 struct net_device *netdev = adapter->netdev;
1033 u16 vid = adapter->hw.mng_cookie.vlan_id;
1034 u16 old_vid = adapter->mng_vlan_id;
1035 if (adapter->vlgrp) {
1036 if (!vlan_group_get_device(adapter->vlgrp, vid)) {
1037 if (adapter->hw.mng_cookie.status &
1038 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1039 igb_vlan_rx_add_vid(netdev, vid);
1040 adapter->mng_vlan_id = vid;
1041 } else
1042 adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1043
1044 if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1045 (vid != old_vid) &&
1046 !vlan_group_get_device(adapter->vlgrp, old_vid))
1047 igb_vlan_rx_kill_vid(netdev, old_vid);
1048 } else
1049 adapter->mng_vlan_id = vid;
1050 }
1051}
1052
1053/**
1054 * igb_release_hw_control - release control of the h/w to f/w
1055 * @adapter: address of board private structure
1056 *
1057 * igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1058 * For ASF and Pass Through versions of f/w this means that the
1059 * driver is no longer loaded.
1060 *
1061 **/
1062static void igb_release_hw_control(struct igb_adapter *adapter)
1063{
1064 struct e1000_hw *hw = &adapter->hw;
1065 u32 ctrl_ext;
1066
1067 /* Let firmware take over control of h/w */
1068 ctrl_ext = rd32(E1000_CTRL_EXT);
1069 wr32(E1000_CTRL_EXT,
1070 ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1071}
1072
1073
1074/**
1075 * igb_get_hw_control - get control of the h/w from f/w
1076 * @adapter: address of board private structure
1077 *
1078 * igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1079 * For ASF and Pass Through versions of f/w this means that
1080 * the driver is loaded.
1081 *
1082 **/
1083static void igb_get_hw_control(struct igb_adapter *adapter)
1084{
1085 struct e1000_hw *hw = &adapter->hw;
1086 u32 ctrl_ext;
1087
1088 /* Let firmware know the driver has taken over */
1089 ctrl_ext = rd32(E1000_CTRL_EXT);
1090 wr32(E1000_CTRL_EXT,
1091 ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1092}
1093
9d5c8243
AK		 1094/**
1095 * igb_configure - configure the hardware for RX and TX
1096 * @adapter: private board structure
1097 **/
1098static void igb_configure(struct igb_adapter *adapter)
1099{
1100 struct net_device *netdev = adapter->netdev;
1101 int i;
1102
1103 igb_get_hw_control(adapter);
ff41f8dc 1104 igb_set_rx_mode(netdev);
9d5c8243
AK		 1105
1106 igb_restore_vlan(adapter);
9d5c8243 1107
85b430b4 1108 igb_setup_tctl(adapter);
9d5c8243 1109 igb_setup_rctl(adapter);
85b430b4
AD		 1110
1111 igb_configure_tx(adapter);
9d5c8243 1112 igb_configure_rx(adapter);
662d7205
AD		 1113
1114 igb_rx_fifo_flush_82575(&adapter->hw);
1115
c493ea45 1116 /* call igb_desc_unused which always leaves
9d5c8243
AK		 1117 * at least 1 descriptor unused to make sure
1118 * next_to_use != next_to_clean */
1119 for (i = 0; i < adapter->num_rx_queues; i++) {
1120 struct igb_ring *ring = &adapter->rx_ring[i];
c493ea45 1121 igb_alloc_rx_buffers_adv(ring, igb_desc_unused(ring));
9d5c8243
AK		 1122 }
1123
1124
1125 adapter->tx_queue_len = netdev->tx_queue_len;
1126}
1127
1128
1129/**
1130 * igb_up - Open the interface and prepare it to handle traffic
1131 * @adapter: board private structure
1132 **/
1133
1134int igb_up(struct igb_adapter *adapter)
1135{
1136 struct e1000_hw *hw = &adapter->hw;
1137 int i;
1138
1139 /* hardware has been reset, we need to reload some things */
1140 igb_configure(adapter);
1141
1142 clear_bit(__IGB_DOWN, &adapter->state);
1143
047e0030
AD		 1144 for (i = 0; i < adapter->num_q_vectors; i++) {
1145 struct igb_q_vector *q_vector = adapter->q_vector[i];
1146 napi_enable(&q_vector->napi);
1147 }
844290e5 1148 if (adapter->msix_entries)
9d5c8243 1149 igb_configure_msix(adapter);
9d5c8243 1150
4ae196df 1151 igb_vmm_control(adapter);
e1739522
AD		 1152 igb_set_vmolr(hw, adapter->vfs_allocated_count);
1153
9d5c8243
AK		 1154 /* Clear any pending interrupts. */
1155 rd32(E1000_ICR);
1156 igb_irq_enable(adapter);
1157
4cb9be7a
JB		 1158 netif_tx_start_all_queues(adapter->netdev);
1159
9d5c8243
AK		 1160 /* Fire a link change interrupt to start the watchdog. */
1161 wr32(E1000_ICS, E1000_ICS_LSC);
1162 return 0;
1163}
1164
1165void igb_down(struct igb_adapter *adapter)
1166{
1167 struct e1000_hw *hw = &adapter->hw;
1168 struct net_device *netdev = adapter->netdev;
1169 u32 tctl, rctl;
1170 int i;
1171
1172 /* signal that we're down so the interrupt handler does not
1173 * reschedule our watchdog timer */
1174 set_bit(__IGB_DOWN, &adapter->state);
1175
1176 /* disable receives in the hardware */
1177 rctl = rd32(E1000_RCTL);
1178 wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
1179 /* flush and sleep below */
1180
fd2ea0a7 1181 netif_tx_stop_all_queues(netdev);
9d5c8243
AK		 1182
1183 /* disable transmits in the hardware */
1184 tctl = rd32(E1000_TCTL);
1185 tctl &= ~E1000_TCTL_EN;
1186 wr32(E1000_TCTL, tctl);
1187 /* flush both disables and wait for them to finish */
1188 wrfl();
1189 msleep(10);
1190
047e0030
AD		 1191 for (i = 0; i < adapter->num_q_vectors; i++) {
1192 struct igb_q_vector *q_vector = adapter->q_vector[i];
1193 napi_disable(&q_vector->napi);
1194 }
9d5c8243 1195
9d5c8243
AK		 1196 igb_irq_disable(adapter);
1197
1198 del_timer_sync(&adapter->watchdog_timer);
1199 del_timer_sync(&adapter->phy_info_timer);
1200
1201 netdev->tx_queue_len = adapter->tx_queue_len;
1202 netif_carrier_off(netdev);
04fe6358
AD		 1203
1204 /* record the stats before reset*/
1205 igb_update_stats(adapter);
1206
9d5c8243
AK		 1207 adapter->link_speed = 0;
1208 adapter->link_duplex = 0;
1209
3023682e
JK		 1210 if (!pci_channel_offline(adapter->pdev))
1211 igb_reset(adapter);
9d5c8243
AK		 1212 igb_clean_all_tx_rings(adapter);
1213 igb_clean_all_rx_rings(adapter);
7e0e99ef
AD		 1214#ifdef CONFIG_IGB_DCA
1215
1216 /* since we reset the hardware DCA settings were cleared */
1217 igb_setup_dca(adapter);
1218#endif
9d5c8243
AK		 1219}
1220
1221void igb_reinit_locked(struct igb_adapter *adapter)
1222{
1223 WARN_ON(in_interrupt());
1224 while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
1225 msleep(1);
1226 igb_down(adapter);
1227 igb_up(adapter);
1228 clear_bit(__IGB_RESETTING, &adapter->state);
1229}
1230
1231void igb_reset(struct igb_adapter *adapter)
1232{
1233 struct e1000_hw *hw = &adapter->hw;
2d064c06
AD		 1234 struct e1000_mac_info *mac = &hw->mac;
1235 struct e1000_fc_info *fc = &hw->fc;
9d5c8243
AK		 1236 u32 pba = 0, tx_space, min_tx_space, min_rx_space;
1237 u16 hwm;
1238
1239 /* Repartition Pba for greater than 9k mtu
1240 * To take effect CTRL.RST is required.
1241 */
fa4dfae0
AD		 1242 switch (mac->type) {
1243 case e1000_82576:
2d064c06 1244 pba = E1000_PBA_64K;
fa4dfae0
AD		 1245 break;
1246 case e1000_82575:
1247 default:
1248 pba = E1000_PBA_34K;
1249 break;
2d064c06 1250 }
9d5c8243 1251
2d064c06
AD		 1252 if ((adapter->max_frame_size > ETH_FRAME_LEN + ETH_FCS_LEN) &&
1253 (mac->type < e1000_82576)) {
9d5c8243
AK		 1254 /* adjust PBA for jumbo frames */
1255 wr32(E1000_PBA, pba);
1256
1257 /* To maintain wire speed transmits, the Tx FIFO should be
1258 * large enough to accommodate two full transmit packets,
1259 * rounded up to the next 1KB and expressed in KB. Likewise,
1260 * the Rx FIFO should be large enough to accommodate at least
1261 * one full receive packet and is similarly rounded up and
1262 * expressed in KB. */
1263 pba = rd32(E1000_PBA);
1264 /* upper 16 bits has Tx packet buffer allocation size in KB */
1265 tx_space = pba >> 16;
1266 /* lower 16 bits has Rx packet buffer allocation size in KB */
1267 pba &= 0xffff;
1268 /* the tx fifo also stores 16 bytes of information about the tx
1269 * but don't include ethernet FCS because hardware appends it */
1270 min_tx_space = (adapter->max_frame_size +
85e8d004 1271 sizeof(union e1000_adv_tx_desc) -
9d5c8243
AK		 1272 ETH_FCS_LEN) * 2;
1273 min_tx_space = ALIGN(min_tx_space, 1024);
1274 min_tx_space >>= 10;
1275 /* software strips receive CRC, so leave room for it */
1276 min_rx_space = adapter->max_frame_size;
1277 min_rx_space = ALIGN(min_rx_space, 1024);
1278 min_rx_space >>= 10;
1279
1280 /* If current Tx allocation is less than the min Tx FIFO size,
1281 * and the min Tx FIFO size is less than the current Rx FIFO
1282 * allocation, take space away from current Rx allocation */
1283 if (tx_space < min_tx_space &&
1284 ((min_tx_space - tx_space) < pba)) {
1285 pba = pba - (min_tx_space - tx_space);
1286
1287 /* if short on rx space, rx wins and must trump tx
1288 * adjustment */
1289 if (pba < min_rx_space)
1290 pba = min_rx_space;
1291 }
2d064c06 1292 wr32(E1000_PBA, pba);
9d5c8243 1293 }
9d5c8243
AK		 1294
1295 /* flow control settings */
1296 /* The high water mark must be low enough to fit one full frame
1297 * (or the size used for early receive) above it in the Rx FIFO.
1298 * Set it to the lower of:
1299 * - 90% of the Rx FIFO size, or
1300 * - the full Rx FIFO size minus one full frame */
1301 hwm = min(((pba << 10) * 9 / 10),
2d064c06 1302 ((pba << 10) - 2 * adapter->max_frame_size));
9d5c8243 1303
2d064c06
AD		 1304 if (mac->type < e1000_82576) {
1305 fc->high_water = hwm & 0xFFF8; /* 8-byte granularity */
1306 fc->low_water = fc->high_water - 8;
1307 } else {
1308 fc->high_water = hwm & 0xFFF0; /* 16-byte granularity */
1309 fc->low_water = fc->high_water - 16;
1310 }
9d5c8243
AK		 1311 fc->pause_time = 0xFFFF;
1312 fc->send_xon = 1;
0cce119a 1313 fc->current_mode = fc->requested_mode;
9d5c8243 1314
4ae196df
AD		 1315 /* disable receive for all VFs and wait one second */
1316 if (adapter->vfs_allocated_count) {
1317 int i;
1318 for (i = 0 ; i < adapter->vfs_allocated_count; i++)
1319 adapter->vf_data[i].clear_to_send = false;
1320
1321 /* ping all the active vfs to let them know we are going down */
1322 igb_ping_all_vfs(adapter);
1323
1324 /* disable transmits and receives */
1325 wr32(E1000_VFRE, 0);
1326 wr32(E1000_VFTE, 0);
1327 }
1328
9d5c8243
AK		 1329 /* Allow time for pending master requests to run */
1330 adapter->hw.mac.ops.reset_hw(&adapter->hw);
1331 wr32(E1000_WUC, 0);
1332
1333 if (adapter->hw.mac.ops.init_hw(&adapter->hw))
1334 dev_err(&adapter->pdev->dev, "Hardware Error\n");
1335
1336 igb_update_mng_vlan(adapter);
1337
1338 /* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
1339 wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
1340
1341 igb_reset_adaptive(&adapter->hw);
f5f4cf08 1342 igb_get_phy_info(&adapter->hw);
9d5c8243
AK		 1343}
1344
2e5c6922
SH		 1345static const struct net_device_ops igb_netdev_ops = {
1346 .ndo_open = igb_open,
1347 .ndo_stop = igb_close,
00829823 1348 .ndo_start_xmit = igb_xmit_frame_adv,
2e5c6922 1349 .ndo_get_stats = igb_get_stats,
ff41f8dc
AD		 1350 .ndo_set_rx_mode = igb_set_rx_mode,
1351 .ndo_set_multicast_list = igb_set_rx_mode,
2e5c6922
SH		 1352 .ndo_set_mac_address = igb_set_mac,
1353 .ndo_change_mtu = igb_change_mtu,
1354 .ndo_do_ioctl = igb_ioctl,
1355 .ndo_tx_timeout = igb_tx_timeout,
1356 .ndo_validate_addr = eth_validate_addr,
1357 .ndo_vlan_rx_register = igb_vlan_rx_register,
1358 .ndo_vlan_rx_add_vid = igb_vlan_rx_add_vid,
1359 .ndo_vlan_rx_kill_vid = igb_vlan_rx_kill_vid,
1360#ifdef CONFIG_NET_POLL_CONTROLLER
1361 .ndo_poll_controller = igb_netpoll,
1362#endif
1363};
1364
9d5c8243
AK		 1365/**
1366 * igb_probe - Device Initialization Routine
1367 * @pdev: PCI device information struct
1368 * @ent: entry in igb_pci_tbl
1369 *
1370 * Returns 0 on success, negative on failure
1371 *
1372 * igb_probe initializes an adapter identified by a pci_dev structure.
1373 * The OS initialization, configuring of the adapter private structure,
1374 * and a hardware reset occur.
1375 **/
1376static int __devinit igb_probe(struct pci_dev *pdev,
1377 const struct pci_device_id *ent)
1378{
1379 struct net_device *netdev;
1380 struct igb_adapter *adapter;
1381 struct e1000_hw *hw;
1382 const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
1383 unsigned long mmio_start, mmio_len;
2d6a5e95 1384 int err, pci_using_dac;
682337fe 1385 u16 eeprom_data = 0;
9d5c8243
AK		 1386 u16 eeprom_apme_mask = IGB_EEPROM_APME;
1387 u32 part_num;
1388
aed5dec3 1389 err = pci_enable_device_mem(pdev);
9d5c8243
AK		 1390 if (err)
1391 return err;
1392
1393 pci_using_dac = 0;
6a35528a 1394 err = pci_set_dma_mask(pdev, DMA_BIT_MASK(64));
9d5c8243 1395 if (!err) {
6a35528a 1396 err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
9d5c8243
AK		 1397 if (!err)
1398 pci_using_dac = 1;
1399 } else {
284901a9 1400 err = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
9d5c8243 1401 if (err) {
284901a9 1402 err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
9d5c8243
AK		 1403 if (err) {
1404 dev_err(&pdev->dev, "No usable DMA "
1405 "configuration, aborting\n");
1406 goto err_dma;
1407 }
1408 }
1409 }
1410
aed5dec3
AD		 1411 err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
1412 IORESOURCE_MEM),
1413 igb_driver_name);
9d5c8243
AK		 1414 if (err)
1415 goto err_pci_reg;
1416
19d5afd4 1417 pci_enable_pcie_error_reporting(pdev);
40a914fa 1418
9d5c8243 1419 pci_set_master(pdev);
c682fc23 1420 pci_save_state(pdev);
9d5c8243
AK		 1421
1422 err = -ENOMEM;
1bfaf07b
AD		 1423 netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
1424 IGB_ABS_MAX_TX_QUEUES);
9d5c8243
AK		 1425 if (!netdev)
1426 goto err_alloc_etherdev;
1427
1428 SET_NETDEV_DEV(netdev, &pdev->dev);
1429
1430 pci_set_drvdata(pdev, netdev);
1431 adapter = netdev_priv(netdev);
1432 adapter->netdev = netdev;
1433 adapter->pdev = pdev;
1434 hw = &adapter->hw;
1435 hw->back = adapter;
1436 adapter->msg_enable = NETIF_MSG_DRV | NETIF_MSG_PROBE;
1437
1438 mmio_start = pci_resource_start(pdev, 0);
1439 mmio_len = pci_resource_len(pdev, 0);
1440
1441 err = -EIO;
28b0759c
AD		 1442 hw->hw_addr = ioremap(mmio_start, mmio_len);
1443 if (!hw->hw_addr)
9d5c8243
AK		 1444 goto err_ioremap;
1445
2e5c6922 1446 netdev->netdev_ops = &igb_netdev_ops;
9d5c8243 1447 igb_set_ethtool_ops(netdev);
9d5c8243 1448 netdev->watchdog_timeo = 5 * HZ;
9d5c8243
AK		 1449
1450 strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
1451
1452 netdev->mem_start = mmio_start;
1453 netdev->mem_end = mmio_start + mmio_len;
1454
9d5c8243
AK		 1455 /* PCI config space info */
1456 hw->vendor_id = pdev->vendor;
1457 hw->device_id = pdev->device;
1458 hw->revision_id = pdev->revision;
1459 hw->subsystem_vendor_id = pdev->subsystem_vendor;
1460 hw->subsystem_device_id = pdev->subsystem_device;
1461
1462 /* setup the private structure */
1463 hw->back = adapter;
1464 /* Copy the default MAC, PHY and NVM function pointers */
1465 memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
1466 memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
1467 memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
1468 /* Initialize skew-specific constants */
1469 err = ei->get_invariants(hw);
1470 if (err)
450c87c8 1471 goto err_sw_init;
9d5c8243 1472
2a3abf6d
AD		 1473#ifdef CONFIG_PCI_IOV
1474 /* since iov functionality isn't critical to base device function we
1475 * can accept failure. If it fails we don't allow iov to be enabled */
1476 if (hw->mac.type == e1000_82576) {
1477 /* 82576 supports a maximum of 7 VFs in addition to the PF */
1478 unsigned int num_vfs = (max_vfs > 7) ? 7 : max_vfs;
1479 int i;
1480 unsigned char mac_addr[ETH_ALEN];
1481
9ca046d5 1482 if (num_vfs) {
2a3abf6d
AD		 1483 adapter->vf_data = kcalloc(num_vfs,
1484 sizeof(struct vf_data_storage),
1485 GFP_KERNEL);
9ca046d5
AD		 1486 if (!adapter->vf_data) {
1487 dev_err(&pdev->dev,
1488 "Could not allocate VF private data - "
1489 "IOV enable failed\n");
2a3abf6d 1490 } else {
9ca046d5
AD		 1491 err = pci_enable_sriov(pdev, num_vfs);
1492 if (!err) {
1493 adapter->vfs_allocated_count = num_vfs;
1494 dev_info(&pdev->dev,
1495 "%d vfs allocated\n",
1496 num_vfs);
1497 for (i = 0;
1498 i < adapter->vfs_allocated_count;
1499 i++) {
1500 random_ether_addr(mac_addr);
1501 igb_set_vf_mac(adapter, i,
1502 mac_addr);
1503 }
1504 } else {
1505 kfree(adapter->vf_data);
1506 adapter->vf_data = NULL;
1507 }
2a3abf6d
AD		 1508 }
1509 }
1510 }
1511
1512#endif
450c87c8 1513 /* setup the private structure */
9d5c8243
AK		 1514 err = igb_sw_init(adapter);
1515 if (err)
1516 goto err_sw_init;
1517
1518 igb_get_bus_info_pcie(hw);
1519
7dfc16fa
AD		 1520 /* set flags */
1521 switch (hw->mac.type) {
7dfc16fa 1522 case e1000_82575:
7dfc16fa
AD		 1523 adapter->flags |= IGB_FLAG_NEED_CTX_IDX;
1524 break;
bbd98fe4 1525 case e1000_82576:
7dfc16fa
AD		 1526 default:
1527 break;
1528 }
1529
9d5c8243
AK		 1530 hw->phy.autoneg_wait_to_complete = false;
1531 hw->mac.adaptive_ifs = true;
1532
1533 /* Copper options */
1534 if (hw->phy.media_type == e1000_media_type_copper) {
1535 hw->phy.mdix = AUTO_ALL_MODES;
1536 hw->phy.disable_polarity_correction = false;
1537 hw->phy.ms_type = e1000_ms_hw_default;
1538 }
1539
1540 if (igb_check_reset_block(hw))
1541 dev_info(&pdev->dev,
1542 "PHY reset is blocked due to SOL/IDER session.\n");
1543
1544 netdev->features = NETIF_F_SG |
7d8eb29e 1545 NETIF_F_IP_CSUM |
9d5c8243
AK		 1546 NETIF_F_HW_VLAN_TX |
1547 NETIF_F_HW_VLAN_RX |
1548 NETIF_F_HW_VLAN_FILTER;
1549
7d8eb29e 1550 netdev->features |= NETIF_F_IPV6_CSUM;
9d5c8243 1551 netdev->features |= NETIF_F_TSO;
9d5c8243 1552 netdev->features |= NETIF_F_TSO6;
48f29ffc 1553
5c0999b7 1554 netdev->features |= NETIF_F_GRO;
d3352520 1555
48f29ffc
JK		 1556 netdev->vlan_features |= NETIF_F_TSO;
1557 netdev->vlan_features |= NETIF_F_TSO6;
7d8eb29e 1558 netdev->vlan_features |= NETIF_F_IP_CSUM;
cd1da503 1559 netdev->vlan_features |= NETIF_F_IPV6_CSUM;
48f29ffc
JK		 1560 netdev->vlan_features |= NETIF_F_SG;
1561
9d5c8243
AK		 1562 if (pci_using_dac)
1563 netdev->features |= NETIF_F_HIGHDMA;
1564
b9473560
JB		 1565 if (adapter->hw.mac.type == e1000_82576)
1566 netdev->features |= NETIF_F_SCTP_CSUM;
1567
9d5c8243
AK		 1568 adapter->en_mng_pt = igb_enable_mng_pass_thru(&adapter->hw);
1569
1570 /* before reading the NVM, reset the controller to put the device in a
1571 * known good starting state */
1572 hw->mac.ops.reset_hw(hw);
1573
1574 /* make sure the NVM is good */
1575 if (igb_validate_nvm_checksum(hw) < 0) {
1576 dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
1577 err = -EIO;
1578 goto err_eeprom;
1579 }
1580
1581 /* copy the MAC address out of the NVM */
1582 if (hw->mac.ops.read_mac_addr(hw))
1583 dev_err(&pdev->dev, "NVM Read Error\n");
1584
1585 memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
1586 memcpy(netdev->perm_addr, hw->mac.addr, netdev->addr_len);
1587
1588 if (!is_valid_ether_addr(netdev->perm_addr)) {
1589 dev_err(&pdev->dev, "Invalid MAC Address\n");
1590 err = -EIO;
1591 goto err_eeprom;
1592 }
1593
0e340485
AD		 1594 setup_timer(&adapter->watchdog_timer, &igb_watchdog,
1595 (unsigned long) adapter);
1596 setup_timer(&adapter->phy_info_timer, &igb_update_phy_info,
1597 (unsigned long) adapter);
9d5c8243
AK		 1598
1599 INIT_WORK(&adapter->reset_task, igb_reset_task);
1600 INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
1601
450c87c8 1602 /* Initialize link properties that are user-changeable */
9d5c8243
AK		 1603 adapter->fc_autoneg = true;
1604 hw->mac.autoneg = true;
1605 hw->phy.autoneg_advertised = 0x2f;
1606
0cce119a
AD		 1607 hw->fc.requested_mode = e1000_fc_default;
1608 hw->fc.current_mode = e1000_fc_default;
9d5c8243 1609
cbd347ad 1610 adapter->itr_setting = IGB_DEFAULT_ITR;
9d5c8243
AK		 1611 adapter->itr = IGB_START_ITR;
1612
1613 igb_validate_mdi_setting(hw);
1614
9d5c8243
AK		 1615 /* Initial Wake on LAN setting If APM wake is enabled in the EEPROM,
1616 * enable the ACPI Magic Packet filter
1617 */
1618
a2cf8b6c 1619 if (hw->bus.func == 0)
312c75ae 1620 hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
a2cf8b6c
AD		 1621 else if (hw->bus.func == 1)
1622 hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
9d5c8243
AK		 1623
1624 if (eeprom_data & eeprom_apme_mask)
1625 adapter->eeprom_wol |= E1000_WUFC_MAG;
1626
1627 /* now that we have the eeprom settings, apply the special cases where
1628 * the eeprom may be wrong or the board simply won't support wake on
1629 * lan on a particular port */
1630 switch (pdev->device) {
1631 case E1000_DEV_ID_82575GB_QUAD_COPPER:
1632 adapter->eeprom_wol = 0;
1633 break;
1634 case E1000_DEV_ID_82575EB_FIBER_SERDES:
2d064c06
AD		 1635 case E1000_DEV_ID_82576_FIBER:
1636 case E1000_DEV_ID_82576_SERDES:
9d5c8243
AK		 1637 /* Wake events only supported on port A for dual fiber
1638 * regardless of eeprom setting */
1639 if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
1640 adapter->eeprom_wol = 0;
1641 break;
c8ea5ea9
AD		 1642 case E1000_DEV_ID_82576_QUAD_COPPER:
1643 /* if quad port adapter, disable WoL on all but port A */
1644 if (global_quad_port_a != 0)
1645 adapter->eeprom_wol = 0;
1646 else
1647 adapter->flags |= IGB_FLAG_QUAD_PORT_A;
1648 /* Reset for multiple quad port adapters */
1649 if (++global_quad_port_a == 4)
1650 global_quad_port_a = 0;
1651 break;
9d5c8243
AK		 1652 }
1653
1654 /* initialize the wol settings based on the eeprom settings */
1655 adapter->wol = adapter->eeprom_wol;
e1b86d84 1656 device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
9d5c8243
AK		 1657
1658 /* reset the hardware with the new settings */
1659 igb_reset(adapter);
1660
1661 /* let the f/w know that the h/w is now under the control of the
1662 * driver. */
1663 igb_get_hw_control(adapter);
1664
9d5c8243
AK		 1665 strcpy(netdev->name, "eth%d");
1666 err = register_netdev(netdev);
1667 if (err)
1668 goto err_register;
1669
b168dfc5
JB		 1670 /* carrier off reporting is important to ethtool even BEFORE open */
1671 netif_carrier_off(netdev);
1672
421e02f0 1673#ifdef CONFIG_IGB_DCA
bbd98fe4 1674 if (dca_add_requester(&pdev->dev) == 0) {
7dfc16fa 1675 adapter->flags |= IGB_FLAG_DCA_ENABLED;
fe4506b6 1676 dev_info(&pdev->dev, "DCA enabled\n");
fe4506b6
JC		 1677 igb_setup_dca(adapter);
1678 }
1679#endif
1680
38c845c7
PO		 1681 /*
1682 * Initialize hardware timer: we keep it running just in case
1683 * that some program needs it later on.
1684 */
1685 memset(&adapter->cycles, 0, sizeof(adapter->cycles));
1686 adapter->cycles.read = igb_read_clock;
1687 adapter->cycles.mask = CLOCKSOURCE_MASK(64);
1688 adapter->cycles.mult = 1;
1689 adapter->cycles.shift = IGB_TSYNC_SHIFT;
1690 wr32(E1000_TIMINCA,
1691 (1<<24) |
1692 IGB_TSYNC_CYCLE_TIME_IN_NANOSECONDS * IGB_TSYNC_SCALE);
1693#if 0
1694 /*
1695 * Avoid rollover while we initialize by resetting the time counter.
1696 */
1697 wr32(E1000_SYSTIML, 0x00000000);
1698 wr32(E1000_SYSTIMH, 0x00000000);
1699#else
1700 /*
1701 * Set registers so that rollover occurs soon to test this.
1702 */
1703 wr32(E1000_SYSTIML, 0x00000000);
1704 wr32(E1000_SYSTIMH, 0xFF800000);
1705#endif
1706 wrfl();
1707 timecounter_init(&adapter->clock,
1708 &adapter->cycles,
1709 ktime_to_ns(ktime_get_real()));
1710
33af6bcc
PO		 1711 /*
1712 * Synchronize our NIC clock against system wall clock. NIC
1713 * time stamp reading requires ~3us per sample, each sample
1714 * was pretty stable even under load => only require 10
1715 * samples for each offset comparison.
1716 */
1717 memset(&adapter->compare, 0, sizeof(adapter->compare));
1718 adapter->compare.source = &adapter->clock;
1719 adapter->compare.target = ktime_get_real;
1720 adapter->compare.num_samples = 10;
1721 timecompare_update(&adapter->compare, 0);
1722
38c845c7
PO		 1723#ifdef DEBUG
1724 {
1725 char buffer[160];
1726 printk(KERN_DEBUG
1727 "igb: %s: hw %p initialized timer\n",
1728 igb_get_time_str(adapter, buffer),
1729 &adapter->hw);
1730 }
1731#endif
1732
9d5c8243
AK		 1733 dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
1734 /* print bus type/speed/width info */
7c510e4b 1735 dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
9d5c8243
AK		 1736 netdev->name,
1737 ((hw->bus.speed == e1000_bus_speed_2500)
1738 ? "2.5Gb/s" : "unknown"),
59c3de89
AD		 1739 ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
1740 (hw->bus.width == e1000_bus_width_pcie_x2) ? "Width x2" :
1741 (hw->bus.width == e1000_bus_width_pcie_x1) ? "Width x1" :
1742 "unknown"),
7c510e4b 1743 netdev->dev_addr);
9d5c8243
AK		 1744
1745 igb_read_part_num(hw, &part_num);
1746 dev_info(&pdev->dev, "%s: PBA No: %06x-%03x\n", netdev->name,
1747 (part_num >> 8), (part_num & 0xff));
1748
1749 dev_info(&pdev->dev,
1750 "Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
1751 adapter->msix_entries ? "MSI-X" :
7dfc16fa 1752 (adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
9d5c8243
AK		 1753 adapter->num_rx_queues, adapter->num_tx_queues);
1754
9d5c8243
AK		 1755 return 0;
1756
1757err_register:
1758 igb_release_hw_control(adapter);
1759err_eeprom:
1760 if (!igb_check_reset_block(hw))
f5f4cf08 1761 igb_reset_phy(hw);
9d5c8243
AK		 1762
1763 if (hw->flash_address)
1764 iounmap(hw->flash_address);
9d5c8243 1765err_sw_init:
047e0030 1766 igb_clear_interrupt_scheme(adapter);
9d5c8243
AK		 1767 iounmap(hw->hw_addr);
1768err_ioremap:
1769 free_netdev(netdev);
1770err_alloc_etherdev:
aed5dec3
AD		 1771 pci_release_selected_regions(pdev, pci_select_bars(pdev,
1772 IORESOURCE_MEM));
9d5c8243
AK		 1773err_pci_reg:
1774err_dma:
1775 pci_disable_device(pdev);
1776 return err;
1777}
1778
1779/**
1780 * igb_remove - Device Removal Routine
1781 * @pdev: PCI device information struct
1782 *
1783 * igb_remove is called by the PCI subsystem to alert the driver
1784 * that it should release a PCI device. The could be caused by a
1785 * Hot-Plug event, or because the driver is going to be removed from
1786 * memory.
1787 **/
1788static void __devexit igb_remove(struct pci_dev *pdev)
1789{
1790 struct net_device *netdev = pci_get_drvdata(pdev);
1791 struct igb_adapter *adapter = netdev_priv(netdev);
fe4506b6 1792 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK		 1793
1794 /* flush_scheduled work may reschedule our watchdog task, so
1795 * explicitly disable watchdog tasks from being rescheduled */
1796 set_bit(__IGB_DOWN, &adapter->state);
1797 del_timer_sync(&adapter->watchdog_timer);
1798 del_timer_sync(&adapter->phy_info_timer);
1799
1800 flush_scheduled_work();
1801
421e02f0 1802#ifdef CONFIG_IGB_DCA
7dfc16fa 1803 if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
fe4506b6
JC		 1804 dev_info(&pdev->dev, "DCA disabled\n");
1805 dca_remove_requester(&pdev->dev);
7dfc16fa 1806 adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
cbd347ad 1807 wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
fe4506b6
JC		 1808 }
1809#endif
1810
9d5c8243
AK		 1811 /* Release control of h/w to f/w. If f/w is AMT enabled, this
1812 * would have already happened in close and is redundant. */
1813 igb_release_hw_control(adapter);
1814
1815 unregister_netdev(netdev);
1816
f5f4cf08
AD		 1817 if (!igb_check_reset_block(&adapter->hw))
1818 igb_reset_phy(&adapter->hw);
9d5c8243 1819
047e0030 1820 igb_clear_interrupt_scheme(adapter);
9d5c8243 1821
37680117
AD		 1822#ifdef CONFIG_PCI_IOV
1823 /* reclaim resources allocated to VFs */
1824 if (adapter->vf_data) {
1825 /* disable iov and allow time for transactions to clear */
1826 pci_disable_sriov(pdev);
1827 msleep(500);
1828
1829 kfree(adapter->vf_data);
1830 adapter->vf_data = NULL;
1831 wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1832 msleep(100);
1833 dev_info(&pdev->dev, "IOV Disabled\n");
1834 }
1835#endif
28b0759c
AD		 1836 iounmap(hw->hw_addr);
1837 if (hw->flash_address)
1838 iounmap(hw->flash_address);
aed5dec3
AD		 1839 pci_release_selected_regions(pdev, pci_select_bars(pdev,
1840 IORESOURCE_MEM));
9d5c8243
AK		 1841
1842 free_netdev(netdev);
1843
19d5afd4 1844 pci_disable_pcie_error_reporting(pdev);
40a914fa 1845
9d5c8243
AK		 1846 pci_disable_device(pdev);
1847}
1848
1849/**
1850 * igb_sw_init - Initialize general software structures (struct igb_adapter)
1851 * @adapter: board private structure to initialize
1852 *
1853 * igb_sw_init initializes the Adapter private data structure.
1854 * Fields are initialized based on PCI device information and
1855 * OS network device settings (MTU size).
1856 **/
1857static int __devinit igb_sw_init(struct igb_adapter *adapter)
1858{
1859 struct e1000_hw *hw = &adapter->hw;
1860 struct net_device *netdev = adapter->netdev;
1861 struct pci_dev *pdev = adapter->pdev;
1862
1863 pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
1864
68fd9910
AD		 1865 adapter->tx_ring_count = IGB_DEFAULT_TXD;
1866 adapter->rx_ring_count = IGB_DEFAULT_RXD;
9d5c8243
AK		 1867 adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
1868 adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
1869
661086df
PWJ		 1870 /* This call may decrease the number of queues depending on
1871 * interrupt mode. */
047e0030 1872 if (igb_init_interrupt_scheme(adapter)) {
9d5c8243
AK		 1873 dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
1874 return -ENOMEM;
1875 }
1876
1877 /* Explicitly disable IRQ since the NIC can be in any state. */
1878 igb_irq_disable(adapter);
1879
1880 set_bit(__IGB_DOWN, &adapter->state);
1881 return 0;
1882}
1883
1884/**
1885 * igb_open - Called when a network interface is made active
1886 * @netdev: network interface device structure
1887 *
1888 * Returns 0 on success, negative value on failure
1889 *
1890 * The open entry point is called when a network interface is made
1891 * active by the system (IFF_UP). At this point all resources needed
1892 * for transmit and receive operations are allocated, the interrupt
1893 * handler is registered with the OS, the watchdog timer is started,
1894 * and the stack is notified that the interface is ready.
1895 **/
1896static int igb_open(struct net_device *netdev)
1897{
1898 struct igb_adapter *adapter = netdev_priv(netdev);
1899 struct e1000_hw *hw = &adapter->hw;
1900 int err;
1901 int i;
1902
1903 /* disallow open during test */
1904 if (test_bit(__IGB_TESTING, &adapter->state))
1905 return -EBUSY;
1906
b168dfc5
JB		 1907 netif_carrier_off(netdev);
1908
9d5c8243
AK		 1909 /* allocate transmit descriptors */
1910 err = igb_setup_all_tx_resources(adapter);
1911 if (err)
1912 goto err_setup_tx;
1913
1914 /* allocate receive descriptors */
1915 err = igb_setup_all_rx_resources(adapter);
1916 if (err)
1917 goto err_setup_rx;
1918
1919 /* e1000_power_up_phy(adapter); */
1920
1921 adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1922 if ((adapter->hw.mng_cookie.status &
1923 E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
1924 igb_update_mng_vlan(adapter);
1925
1926 /* before we allocate an interrupt, we must be ready to handle it.
1927 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
1928 * as soon as we call pci_request_irq, so we have to setup our
1929 * clean_rx handler before we do so. */
1930 igb_configure(adapter);
1931
4ae196df 1932 igb_vmm_control(adapter);
e1739522
AD		 1933 igb_set_vmolr(hw, adapter->vfs_allocated_count);
1934
9d5c8243
AK		 1935 err = igb_request_irq(adapter);
1936 if (err)
1937 goto err_req_irq;
1938
1939 /* From here on the code is the same as igb_up() */
1940 clear_bit(__IGB_DOWN, &adapter->state);
1941
047e0030
AD		 1942 for (i = 0; i < adapter->num_q_vectors; i++) {
1943 struct igb_q_vector *q_vector = adapter->q_vector[i];
1944 napi_enable(&q_vector->napi);
1945 }
9d5c8243
AK		 1946
1947 /* Clear any pending interrupts. */
1948 rd32(E1000_ICR);
844290e5
PW		 1949
1950 igb_irq_enable(adapter);
1951
d55b53ff
JK		 1952 netif_tx_start_all_queues(netdev);
1953
9d5c8243
AK		 1954 /* Fire a link status change interrupt to start the watchdog. */
1955 wr32(E1000_ICS, E1000_ICS_LSC);
1956
1957 return 0;
1958
1959err_req_irq:
1960 igb_release_hw_control(adapter);
1961 /* e1000_power_down_phy(adapter); */
1962 igb_free_all_rx_resources(adapter);
1963err_setup_rx:
1964 igb_free_all_tx_resources(adapter);
1965err_setup_tx:
1966 igb_reset(adapter);
1967
1968 return err;
1969}
1970
1971/**
1972 * igb_close - Disables a network interface
1973 * @netdev: network interface device structure
1974 *
1975 * Returns 0, this is not allowed to fail
1976 *
1977 * The close entry point is called when an interface is de-activated
1978 * by the OS. The hardware is still under the driver's control, but
1979 * needs to be disabled. A global MAC reset is issued to stop the
1980 * hardware, and all transmit and receive resources are freed.
1981 **/
1982static int igb_close(struct net_device *netdev)
1983{
1984 struct igb_adapter *adapter = netdev_priv(netdev);
1985
1986 WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
1987 igb_down(adapter);
1988
1989 igb_free_irq(adapter);
1990
1991 igb_free_all_tx_resources(adapter);
1992 igb_free_all_rx_resources(adapter);
1993
1994 /* kill manageability vlan ID if supported, but not if a vlan with
1995 * the same ID is registered on the host OS (let 8021q kill it) */
1996 if ((adapter->hw.mng_cookie.status &
1997 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
1998 !(adapter->vlgrp &&
1999 vlan_group_get_device(adapter->vlgrp, adapter->mng_vlan_id)))
2000 igb_vlan_rx_kill_vid(netdev, adapter->mng_vlan_id);
2001
2002 return 0;
2003}
2004
2005/**
2006 * igb_setup_tx_resources - allocate Tx resources (Descriptors)
9d5c8243
AK		 2007 * @tx_ring: tx descriptor ring (for a specific queue) to setup
2008 *
2009 * Return 0 on success, negative on failure
2010 **/
80785298 2011int igb_setup_tx_resources(struct igb_ring *tx_ring)
9d5c8243 2012{
80785298 2013 struct pci_dev *pdev = tx_ring->pdev;
9d5c8243
AK		 2014 int size;
2015
2016 size = sizeof(struct igb_buffer) * tx_ring->count;
2017 tx_ring->buffer_info = vmalloc(size);
2018 if (!tx_ring->buffer_info)
2019 goto err;
2020 memset(tx_ring->buffer_info, 0, size);
2021
2022 /* round up to nearest 4K */
85e8d004 2023 tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
9d5c8243
AK		 2024 tx_ring->size = ALIGN(tx_ring->size, 4096);
2025
2026 tx_ring->desc = pci_alloc_consistent(pdev, tx_ring->size,
2027 &tx_ring->dma);
2028
2029 if (!tx_ring->desc)
2030 goto err;
2031
9d5c8243
AK		 2032 tx_ring->next_to_use = 0;
2033 tx_ring->next_to_clean = 0;
9d5c8243
AK		 2034 return 0;
2035
2036err:
2037 vfree(tx_ring->buffer_info);
047e0030 2038 dev_err(&pdev->dev,
9d5c8243
AK		 2039 "Unable to allocate memory for the transmit descriptor ring\n");
2040 return -ENOMEM;
2041}
2042
2043/**
2044 * igb_setup_all_tx_resources - wrapper to allocate Tx resources
2045 * (Descriptors) for all queues
2046 * @adapter: board private structure
2047 *
2048 * Return 0 on success, negative on failure
2049 **/
2050static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
2051{
2052 int i, err = 0;
661086df 2053 int r_idx;
9d5c8243
AK		 2054
2055 for (i = 0; i < adapter->num_tx_queues; i++) {
80785298 2056 err = igb_setup_tx_resources(&adapter->tx_ring[i]);
9d5c8243
AK		 2057 if (err) {
2058 dev_err(&adapter->pdev->dev,
2059 "Allocation for Tx Queue %u failed\n", i);
2060 for (i--; i >= 0; i--)
3b644cf6 2061 igb_free_tx_resources(&adapter->tx_ring[i]);
9d5c8243
AK		 2062 break;
2063 }
2064 }
2065
661086df
PWJ		 2066 for (i = 0; i < IGB_MAX_TX_QUEUES; i++) {
2067 r_idx = i % adapter->num_tx_queues;
2068 adapter->multi_tx_table[i] = &adapter->tx_ring[r_idx];
eebbbdba 2069 }
9d5c8243
AK		 2070 return err;
2071}
2072
2073/**
85b430b4
AD		 2074 * igb_setup_tctl - configure the transmit control registers
2075 * @adapter: Board private structure
9d5c8243 2076 **/
85b430b4 2077static void igb_setup_tctl(struct igb_adapter *adapter)
9d5c8243 2078{
9d5c8243
AK		 2079 struct e1000_hw *hw = &adapter->hw;
2080 u32 tctl;
9d5c8243 2081
85b430b4
AD		 2082 /* disable queue 0 which is enabled by default on 82575 and 82576 */
2083 wr32(E1000_TXDCTL(0), 0);
9d5c8243
AK		 2084
2085 /* Program the Transmit Control Register */
9d5c8243
AK		 2086 tctl = rd32(E1000_TCTL);
2087 tctl &= ~E1000_TCTL_CT;
2088 tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2089 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2090
2091 igb_config_collision_dist(hw);
2092
9d5c8243
AK		 2093 /* Enable transmits */
2094 tctl |= E1000_TCTL_EN;
2095
2096 wr32(E1000_TCTL, tctl);
2097}
2098
85b430b4
AD		 2099/**
2100 * igb_configure_tx_ring - Configure transmit ring after Reset
2101 * @adapter: board private structure
2102 * @ring: tx ring to configure
2103 *
2104 * Configure a transmit ring after a reset.
2105 **/
2106static void igb_configure_tx_ring(struct igb_adapter *adapter,
2107 struct igb_ring *ring)
2108{
2109 struct e1000_hw *hw = &adapter->hw;
2110 u32 txdctl;
2111 u64 tdba = ring->dma;
2112 int reg_idx = ring->reg_idx;
2113
2114 /* disable the queue */
2115 txdctl = rd32(E1000_TXDCTL(reg_idx));
2116 wr32(E1000_TXDCTL(reg_idx),
2117 txdctl & ~E1000_TXDCTL_QUEUE_ENABLE);
2118 wrfl();
2119 mdelay(10);
2120
2121 wr32(E1000_TDLEN(reg_idx),
2122 ring->count * sizeof(union e1000_adv_tx_desc));
2123 wr32(E1000_TDBAL(reg_idx),
2124 tdba & 0x00000000ffffffffULL);
2125 wr32(E1000_TDBAH(reg_idx), tdba >> 32);
2126
fce99e34
AD		 2127 ring->head = hw->hw_addr + E1000_TDH(reg_idx);
2128 ring->tail = hw->hw_addr + E1000_TDT(reg_idx);
2129 writel(0, ring->head);
2130 writel(0, ring->tail);
85b430b4
AD		 2131
2132 txdctl |= IGB_TX_PTHRESH;
2133 txdctl |= IGB_TX_HTHRESH << 8;
2134 txdctl |= IGB_TX_WTHRESH << 16;
2135
2136 txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
2137 wr32(E1000_TXDCTL(reg_idx), txdctl);
2138}
2139
2140/**
2141 * igb_configure_tx - Configure transmit Unit after Reset
2142 * @adapter: board private structure
2143 *
2144 * Configure the Tx unit of the MAC after a reset.
2145 **/
2146static void igb_configure_tx(struct igb_adapter *adapter)
2147{
2148 int i;
2149
2150 for (i = 0; i < adapter->num_tx_queues; i++)
2151 igb_configure_tx_ring(adapter, &adapter->tx_ring[i]);
2152
2153 /* Setup Transmit Descriptor Settings for eop descriptor */
2154 adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_RS;
2155}
2156
9d5c8243
AK		 2157/**
2158 * igb_setup_rx_resources - allocate Rx resources (Descriptors)
9d5c8243
AK		 2159 * @rx_ring: rx descriptor ring (for a specific queue) to setup
2160 *
2161 * Returns 0 on success, negative on failure
2162 **/
80785298 2163int igb_setup_rx_resources(struct igb_ring *rx_ring)
9d5c8243 2164{
80785298 2165 struct pci_dev *pdev = rx_ring->pdev;
9d5c8243
AK		 2166 int size, desc_len;
2167
2168 size = sizeof(struct igb_buffer) * rx_ring->count;
2169 rx_ring->buffer_info = vmalloc(size);
2170 if (!rx_ring->buffer_info)
2171 goto err;
2172 memset(rx_ring->buffer_info, 0, size);
2173
2174 desc_len = sizeof(union e1000_adv_rx_desc);
2175
2176 /* Round up to nearest 4K */
2177 rx_ring->size = rx_ring->count * desc_len;
2178 rx_ring->size = ALIGN(rx_ring->size, 4096);
2179
2180 rx_ring->desc = pci_alloc_consistent(pdev, rx_ring->size,
2181 &rx_ring->dma);
2182
2183 if (!rx_ring->desc)
2184 goto err;
2185
2186 rx_ring->next_to_clean = 0;
2187 rx_ring->next_to_use = 0;
9d5c8243 2188
9d5c8243
AK		 2189 return 0;
2190
2191err:
2192 vfree(rx_ring->buffer_info);
80785298 2193 dev_err(&pdev->dev, "Unable to allocate memory for "
9d5c8243
AK		 2194 "the receive descriptor ring\n");
2195 return -ENOMEM;
2196}
2197
2198/**
2199 * igb_setup_all_rx_resources - wrapper to allocate Rx resources
2200 * (Descriptors) for all queues
2201 * @adapter: board private structure
2202 *
2203 * Return 0 on success, negative on failure
2204 **/
2205static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
2206{
2207 int i, err = 0;
2208
2209 for (i = 0; i < adapter->num_rx_queues; i++) {
80785298 2210 err = igb_setup_rx_resources(&adapter->rx_ring[i]);
9d5c8243
AK		 2211 if (err) {
2212 dev_err(&adapter->pdev->dev,
2213 "Allocation for Rx Queue %u failed\n", i);
2214 for (i--; i >= 0; i--)
3b644cf6 2215 igb_free_rx_resources(&adapter->rx_ring[i]);
9d5c8243
AK		 2216 break;
2217 }
2218 }
2219
2220 return err;
2221}
2222
2223/**
2224 * igb_setup_rctl - configure the receive control registers
2225 * @adapter: Board private structure
2226 **/
2227static void igb_setup_rctl(struct igb_adapter *adapter)
2228{
2229 struct e1000_hw *hw = &adapter->hw;
2230 u32 rctl;
9d5c8243
AK		 2231
2232 rctl = rd32(E1000_RCTL);
2233
2234 rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
69d728ba 2235 rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
9d5c8243 2236
69d728ba 2237 rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
28b0759c 2238 (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
9d5c8243 2239
87cb7e8c
AK		 2240 /*
2241 * enable stripping of CRC. It's unlikely this will break BMC
2242 * redirection as it did with e1000. Newer features require
2243 * that the HW strips the CRC.
73cd78f1 2244 */
87cb7e8c 2245 rctl |= E1000_RCTL_SECRC;
9d5c8243 2246
9b07f3d3 2247 /*
ec54d7d6 2248 * disable store bad packets and clear size bits.
9b07f3d3 2249 */
ec54d7d6 2250 rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
9d5c8243 2251
6ec43fe6
AD		 2252 /* enable LPE to prevent packets larger than max_frame_size */
2253 rctl |= E1000_RCTL_LPE;
9d5c8243 2254
952f72a8
AD		 2255 /* disable queue 0 to prevent tail write w/o re-config */
2256 wr32(E1000_RXDCTL(0), 0);
9d5c8243 2257
e1739522
AD		 2258 /* Attention!!! For SR-IOV PF driver operations you must enable
2259 * queue drop for all VF and PF queues to prevent head of line blocking
2260 * if an un-trusted VF does not provide descriptors to hardware.
2261 */
2262 if (adapter->vfs_allocated_count) {
2263 u32 vmolr;
2264
e1739522
AD		 2265 /* set all queue drop enable bits */
2266 wr32(E1000_QDE, ALL_QUEUES);
e1739522 2267
77a22941 2268 vmolr = rd32(E1000_VMOLR(adapter->vfs_allocated_count));
e1739522
AD		 2269 if (rctl & E1000_RCTL_LPE)
2270 vmolr |= E1000_VMOLR_LPE;
77a22941 2271 if (adapter->num_rx_queues > 1)
e1739522 2272 vmolr |= E1000_VMOLR_RSSE;
77a22941 2273 wr32(E1000_VMOLR(adapter->vfs_allocated_count), vmolr);
e1739522
AD		 2274 }
2275
9d5c8243
AK		 2276 wr32(E1000_RCTL, rctl);
2277}
2278
e1739522
AD		 2279/**
2280 * igb_rlpml_set - set maximum receive packet size
2281 * @adapter: board private structure
2282 *
2283 * Configure maximum receivable packet size.
2284 **/
2285static void igb_rlpml_set(struct igb_adapter *adapter)
2286{
2287 u32 max_frame_size = adapter->max_frame_size;
2288 struct e1000_hw *hw = &adapter->hw;
2289 u16 pf_id = adapter->vfs_allocated_count;
2290
2291 if (adapter->vlgrp)
2292 max_frame_size += VLAN_TAG_SIZE;
2293
2294 /* if vfs are enabled we set RLPML to the largest possible request
2295 * size and set the VMOLR RLPML to the size we need */
2296 if (pf_id) {
2297 igb_set_vf_rlpml(adapter, max_frame_size, pf_id);
2298 max_frame_size = MAX_STD_JUMBO_FRAME_SIZE + VLAN_TAG_SIZE;
2299 }
2300
2301 wr32(E1000_RLPML, max_frame_size);
2302}
2303
2304/**
2305 * igb_configure_vt_default_pool - Configure VT default pool
2306 * @adapter: board private structure
2307 *
2308 * Configure the default pool
2309 **/
2310static void igb_configure_vt_default_pool(struct igb_adapter *adapter)
2311{
2312 struct e1000_hw *hw = &adapter->hw;
2313 u16 pf_id = adapter->vfs_allocated_count;
2314 u32 vtctl;
2315
2316 /* not in sr-iov mode - do nothing */
2317 if (!pf_id)
2318 return;
2319
2320 vtctl = rd32(E1000_VT_CTL);
2321 vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
2322 E1000_VT_CTL_DISABLE_DEF_POOL);
2323 vtctl |= pf_id << E1000_VT_CTL_DEFAULT_POOL_SHIFT;
2324 wr32(E1000_VT_CTL, vtctl);
2325}
2326
85b430b4
AD		 2327/**
2328 * igb_configure_rx_ring - Configure a receive ring after Reset
2329 * @adapter: board private structure
2330 * @ring: receive ring to be configured
2331 *
2332 * Configure the Rx unit of the MAC after a reset.
2333 **/
2334static void igb_configure_rx_ring(struct igb_adapter *adapter,
2335 struct igb_ring *ring)
2336{
2337 struct e1000_hw *hw = &adapter->hw;
2338 u64 rdba = ring->dma;
2339 int reg_idx = ring->reg_idx;
952f72a8 2340 u32 srrctl, rxdctl;
85b430b4
AD		 2341
2342 /* disable the queue */
2343 rxdctl = rd32(E1000_RXDCTL(reg_idx));
2344 wr32(E1000_RXDCTL(reg_idx),
2345 rxdctl & ~E1000_RXDCTL_QUEUE_ENABLE);
2346
2347 /* Set DMA base address registers */
2348 wr32(E1000_RDBAL(reg_idx),
2349 rdba & 0x00000000ffffffffULL);
2350 wr32(E1000_RDBAH(reg_idx), rdba >> 32);
2351 wr32(E1000_RDLEN(reg_idx),
2352 ring->count * sizeof(union e1000_adv_rx_desc));
2353
2354 /* initialize head and tail */
fce99e34
AD		 2355 ring->head = hw->hw_addr + E1000_RDH(reg_idx);
2356 ring->tail = hw->hw_addr + E1000_RDT(reg_idx);
2357 writel(0, ring->head);
2358 writel(0, ring->tail);
85b430b4 2359
952f72a8 2360 /* set descriptor configuration */
4c844851
AD		 2361 if (ring->rx_buffer_len < IGB_RXBUFFER_1024) {
2362 srrctl = ALIGN(ring->rx_buffer_len, 64) <<
952f72a8
AD		 2363 E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
2364#if (PAGE_SIZE / 2) > IGB_RXBUFFER_16384
2365 srrctl |= IGB_RXBUFFER_16384 >>
2366 E1000_SRRCTL_BSIZEPKT_SHIFT;
2367#else
2368 srrctl |= (PAGE_SIZE / 2) >>
2369 E1000_SRRCTL_BSIZEPKT_SHIFT;
2370#endif
2371 srrctl |= E1000_SRRCTL_DESCTYPE_HDR_SPLIT_ALWAYS;
2372 } else {
4c844851 2373 srrctl = ALIGN(ring->rx_buffer_len, 1024) >>
952f72a8
AD		 2374 E1000_SRRCTL_BSIZEPKT_SHIFT;
2375 srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
2376 }
2377
2378 wr32(E1000_SRRCTL(reg_idx), srrctl);
2379
85b430b4
AD		 2380 /* enable receive descriptor fetching */
2381 rxdctl = rd32(E1000_RXDCTL(reg_idx));
2382 rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
2383 rxdctl &= 0xFFF00000;
2384 rxdctl |= IGB_RX_PTHRESH;
2385 rxdctl |= IGB_RX_HTHRESH << 8;
2386 rxdctl |= IGB_RX_WTHRESH << 16;
2387 wr32(E1000_RXDCTL(reg_idx), rxdctl);
2388}
2389
9d5c8243
AK		 2390/**
2391 * igb_configure_rx - Configure receive Unit after Reset
2392 * @adapter: board private structure
2393 *
2394 * Configure the Rx unit of the MAC after a reset.
2395 **/
2396static void igb_configure_rx(struct igb_adapter *adapter)
2397{
9d5c8243
AK		 2398 struct e1000_hw *hw = &adapter->hw;
2399 u32 rctl, rxcsum;
9107584e 2400 int i;
9d5c8243
AK		 2401
2402 /* disable receives while setting up the descriptors */
2403 rctl = rd32(E1000_RCTL);
2404 wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
2405 wrfl();
2406 mdelay(10);
2407
2408 if (adapter->itr_setting > 3)
6eb5a7f1 2409 wr32(E1000_ITR, adapter->itr);
9d5c8243
AK		 2410
2411 /* Setup the HW Rx Head and Tail Descriptor Pointers and
2412 * the Base and Length of the Rx Descriptor Ring */
85b430b4
AD		 2413 for (i = 0; i < adapter->num_rx_queues; i++)
2414 igb_configure_rx_ring(adapter, &adapter->rx_ring[i]);
9d5c8243
AK		 2415
2416 if (adapter->num_rx_queues > 1) {
2417 u32 random[10];
2418 u32 mrqc;
2419 u32 j, shift;
2420 union e1000_reta {
2421 u32 dword;
2422 u8 bytes[4];
2423 } reta;
2424
2425 get_random_bytes(&random[0], 40);
2426
2d064c06
AD		 2427 if (hw->mac.type >= e1000_82576)
2428 shift = 0;
2429 else
2430 shift = 6;
9d5c8243
AK		 2431 for (j = 0; j < (32 * 4); j++) {
2432 reta.bytes[j & 3] =
26bc19ec 2433 adapter->rx_ring[(j % adapter->num_rx_queues)].reg_idx << shift;
9d5c8243
AK		 2434 if ((j & 3) == 3)
2435 writel(reta.dword,
2436 hw->hw_addr + E1000_RETA(0) + (j & ~3));
2437 }
e1739522
AD		 2438 if (adapter->vfs_allocated_count)
2439 mrqc = E1000_MRQC_ENABLE_VMDQ_RSS_2Q;
2440 else
2441 mrqc = E1000_MRQC_ENABLE_RSS_4Q;
9d5c8243
AK		 2442
2443 /* Fill out hash function seeds */
2444 for (j = 0; j < 10; j++)
2445 array_wr32(E1000_RSSRK(0), j, random[j]);
2446
2447 mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2448 E1000_MRQC_RSS_FIELD_IPV4_TCP);
2449 mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2450 E1000_MRQC_RSS_FIELD_IPV6_TCP);
2451 mrqc |= (E1000_MRQC_RSS_FIELD_IPV4_UDP |
2452 E1000_MRQC_RSS_FIELD_IPV6_UDP);
2453 mrqc |= (E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2454 E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2455
9d5c8243 2456 wr32(E1000_MRQC, mrqc);
2844f797 2457 } else if (adapter->vfs_allocated_count) {
e1739522 2458 /* Enable multi-queue for sr-iov */
2844f797 2459 wr32(E1000_MRQC, E1000_MRQC_ENABLE_VMDQ);
9d5c8243
AK		 2460 }
2461
2844f797
AD		 2462 /* Enable Receive Checksum Offload for TCP and UDP */
2463 rxcsum = rd32(E1000_RXCSUM);
2464 /* Disable raw packet checksumming */
2465 rxcsum |= E1000_RXCSUM_PCSD;
7beb0146
AD		 2466
2467 if (adapter->hw.mac.type == e1000_82576)
b9473560
JB		 2468 /* Enable Receive Checksum Offload for SCTP */
2469 rxcsum |= E1000_RXCSUM_CRCOFL;
2470
7beb0146 2471 /* Don't need to set TUOFL or IPOFL, they default to 1 */
2844f797
AD		 2472 wr32(E1000_RXCSUM, rxcsum);
2473
e1739522
AD		 2474 /* Set the default pool for the PF's first queue */
2475 igb_configure_vt_default_pool(adapter);
2476
68d480c4
AD		 2477 /* set UTA to appropriate mode */
2478 igb_set_uta(adapter);
2479
26ad9178
AD		 2480 /* set the correct pool for the PF default MAC address in entry 0 */
2481 igb_rar_set_qsel(adapter, adapter->hw.mac.addr, 0,
2482 adapter->vfs_allocated_count);
2483
e1739522 2484 igb_rlpml_set(adapter);
9d5c8243
AK		 2485
2486 /* Enable Receives */
2487 wr32(E1000_RCTL, rctl);
2488}
2489
2490/**
2491 * igb_free_tx_resources - Free Tx Resources per Queue
9d5c8243
AK		 2492 * @tx_ring: Tx descriptor ring for a specific queue
2493 *
2494 * Free all transmit software resources
2495 **/
68fd9910 2496void igb_free_tx_resources(struct igb_ring *tx_ring)
9d5c8243 2497{
3b644cf6 2498 igb_clean_tx_ring(tx_ring);
9d5c8243
AK		 2499
2500 vfree(tx_ring->buffer_info);
2501 tx_ring->buffer_info = NULL;
2502
80785298
AD		 2503 pci_free_consistent(tx_ring->pdev, tx_ring->size,
2504 tx_ring->desc, tx_ring->dma);
9d5c8243
AK		 2505
2506 tx_ring->desc = NULL;
2507}
2508
2509/**
2510 * igb_free_all_tx_resources - Free Tx Resources for All Queues
2511 * @adapter: board private structure
2512 *
2513 * Free all transmit software resources
2514 **/
2515static void igb_free_all_tx_resources(struct igb_adapter *adapter)
2516{
2517 int i;
2518
2519 for (i = 0; i < adapter->num_tx_queues; i++)
3b644cf6 2520 igb_free_tx_resources(&adapter->tx_ring[i]);
9d5c8243
AK		 2521}
2522
80785298 2523static void igb_unmap_and_free_tx_resource(struct igb_ring *tx_ring,
9d5c8243
AK		 2524 struct igb_buffer *buffer_info)
2525{
65689fef 2526 buffer_info->dma = 0;
9d5c8243 2527 if (buffer_info->skb) {
80785298
AD		 2528 skb_dma_unmap(&tx_ring->pdev->dev,
2529 buffer_info->skb,
65689fef 2530 DMA_TO_DEVICE);
9d5c8243
AK		 2531 dev_kfree_skb_any(buffer_info->skb);
2532 buffer_info->skb = NULL;
2533 }
2534 buffer_info->time_stamp = 0;
2535 /* buffer_info must be completely set up in the transmit path */
2536}
2537
2538/**
2539 * igb_clean_tx_ring - Free Tx Buffers
9d5c8243
AK		 2540 * @tx_ring: ring to be cleaned
2541 **/
3b644cf6 2542static void igb_clean_tx_ring(struct igb_ring *tx_ring)
9d5c8243
AK		 2543{
2544 struct igb_buffer *buffer_info;
2545 unsigned long size;
2546 unsigned int i;
2547
2548 if (!tx_ring->buffer_info)
2549 return;
2550 /* Free all the Tx ring sk_buffs */
2551
2552 for (i = 0; i < tx_ring->count; i++) {
2553 buffer_info = &tx_ring->buffer_info[i];
80785298 2554 igb_unmap_and_free_tx_resource(tx_ring, buffer_info);
9d5c8243
AK		 2555 }
2556
2557 size = sizeof(struct igb_buffer) * tx_ring->count;
2558 memset(tx_ring->buffer_info, 0, size);
2559
2560 /* Zero out the descriptor ring */
2561
2562 memset(tx_ring->desc, 0, tx_ring->size);
2563
2564 tx_ring->next_to_use = 0;
2565 tx_ring->next_to_clean = 0;
2566
fce99e34
AD		 2567 writel(0, tx_ring->head);
2568 writel(0, tx_ring->tail);
9d5c8243
AK		 2569}
2570
2571/**
2572 * igb_clean_all_tx_rings - Free Tx Buffers for all queues
2573 * @adapter: board private structure
2574 **/
2575static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
2576{
2577 int i;
2578
2579 for (i = 0; i < adapter->num_tx_queues; i++)
3b644cf6 2580 igb_clean_tx_ring(&adapter->tx_ring[i]);
9d5c8243
AK		 2581}
2582
2583/**
2584 * igb_free_rx_resources - Free Rx Resources
9d5c8243
AK		 2585 * @rx_ring: ring to clean the resources from
2586 *
2587 * Free all receive software resources
2588 **/
68fd9910 2589void igb_free_rx_resources(struct igb_ring *rx_ring)
9d5c8243 2590{
3b644cf6 2591 igb_clean_rx_ring(rx_ring);
9d5c8243
AK		 2592
2593 vfree(rx_ring->buffer_info);
2594 rx_ring->buffer_info = NULL;
2595
80785298
AD		 2596 pci_free_consistent(rx_ring->pdev, rx_ring->size,
2597 rx_ring->desc, rx_ring->dma);
9d5c8243
AK		 2598
2599 rx_ring->desc = NULL;
2600}
2601
2602/**
2603 * igb_free_all_rx_resources - Free Rx Resources for All Queues
2604 * @adapter: board private structure
2605 *
2606 * Free all receive software resources
2607 **/
2608static void igb_free_all_rx_resources(struct igb_adapter *adapter)
2609{
2610 int i;
2611
2612 for (i = 0; i < adapter->num_rx_queues; i++)
3b644cf6 2613 igb_free_rx_resources(&adapter->rx_ring[i]);
9d5c8243
AK		 2614}
2615
2616/**
2617 * igb_clean_rx_ring - Free Rx Buffers per Queue
9d5c8243
AK		 2618 * @rx_ring: ring to free buffers from
2619 **/
3b644cf6 2620static void igb_clean_rx_ring(struct igb_ring *rx_ring)
9d5c8243
AK		 2621{
2622 struct igb_buffer *buffer_info;
9d5c8243
AK		 2623 unsigned long size;
2624 unsigned int i;
2625
2626 if (!rx_ring->buffer_info)
2627 return;
2628 /* Free all the Rx ring sk_buffs */
2629 for (i = 0; i < rx_ring->count; i++) {
2630 buffer_info = &rx_ring->buffer_info[i];
2631 if (buffer_info->dma) {
80785298
AD		 2632 pci_unmap_single(rx_ring->pdev,
2633 buffer_info->dma,
4c844851 2634 rx_ring->rx_buffer_len,
6ec43fe6 2635 PCI_DMA_FROMDEVICE);
9d5c8243
AK		 2636 buffer_info->dma = 0;
2637 }
2638
2639 if (buffer_info->skb) {
2640 dev_kfree_skb(buffer_info->skb);
2641 buffer_info->skb = NULL;
2642 }
6ec43fe6 2643 if (buffer_info->page_dma) {
80785298
AD		 2644 pci_unmap_page(rx_ring->pdev,
2645 buffer_info->page_dma,
6ec43fe6
AD		 2646 PAGE_SIZE / 2,
2647 PCI_DMA_FROMDEVICE);
2648 buffer_info->page_dma = 0;
2649 }
9d5c8243 2650 if (buffer_info->page) {
9d5c8243
AK		 2651 put_page(buffer_info->page);
2652 buffer_info->page = NULL;
bf36c1a0 2653 buffer_info->page_offset = 0;
9d5c8243
AK		 2654 }
2655 }
2656
9d5c8243
AK		 2657 size = sizeof(struct igb_buffer) * rx_ring->count;
2658 memset(rx_ring->buffer_info, 0, size);
2659
2660 /* Zero out the descriptor ring */
2661 memset(rx_ring->desc, 0, rx_ring->size);
2662
2663 rx_ring->next_to_clean = 0;
2664 rx_ring->next_to_use = 0;
2665
fce99e34
AD		 2666 writel(0, rx_ring->head);
2667 writel(0, rx_ring->tail);
9d5c8243
AK		 2668}
2669
2670/**
2671 * igb_clean_all_rx_rings - Free Rx Buffers for all queues
2672 * @adapter: board private structure
2673 **/
2674static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
2675{
2676 int i;
2677
2678 for (i = 0; i < adapter->num_rx_queues; i++)
3b644cf6 2679 igb_clean_rx_ring(&adapter->rx_ring[i]);
9d5c8243
AK		 2680}
2681
2682/**
2683 * igb_set_mac - Change the Ethernet Address of the NIC
2684 * @netdev: network interface device structure
2685 * @p: pointer to an address structure
2686 *
2687 * Returns 0 on success, negative on failure
2688 **/
2689static int igb_set_mac(struct net_device *netdev, void *p)
2690{
2691 struct igb_adapter *adapter = netdev_priv(netdev);
28b0759c 2692 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK		 2693 struct sockaddr *addr = p;
2694
2695 if (!is_valid_ether_addr(addr->sa_data))
2696 return -EADDRNOTAVAIL;
2697
2698 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
28b0759c 2699 memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
9d5c8243 2700
26ad9178
AD		 2701 /* set the correct pool for the new PF MAC address in entry 0 */
2702 igb_rar_set_qsel(adapter, hw->mac.addr, 0,
2703 adapter->vfs_allocated_count);
e1739522 2704
9d5c8243
AK		 2705 return 0;
2706}
2707
2708/**
68d480c4 2709 * igb_write_mc_addr_list - write multicast addresses to MTA
9d5c8243
AK		 2710 * @netdev: network interface device structure
2711 *
68d480c4
AD		 2712 * Writes multicast address list to the MTA hash table.
2713 * Returns: -ENOMEM on failure
2714 * 0 on no addresses written
2715 * X on writing X addresses to MTA
9d5c8243 2716 **/
68d480c4 2717static int igb_write_mc_addr_list(struct net_device *netdev)
9d5c8243
AK		 2718{
2719 struct igb_adapter *adapter = netdev_priv(netdev);
2720 struct e1000_hw *hw = &adapter->hw;
ff41f8dc 2721 struct dev_mc_list *mc_ptr = netdev->mc_list;
68d480c4
AD		 2722 u8 *mta_list;
2723 u32 vmolr = 0;
9d5c8243
AK		 2724 int i;
2725
68d480c4
AD		 2726 if (!netdev->mc_count) {
2727 /* nothing to program, so clear mc list */
2728 igb_update_mc_addr_list(hw, NULL, 0);
2729 igb_restore_vf_multicasts(adapter);
2730 return 0;
2731 }
9d5c8243 2732
68d480c4
AD		 2733 mta_list = kzalloc(netdev->mc_count * 6, GFP_ATOMIC);
2734 if (!mta_list)
2735 return -ENOMEM;
ff41f8dc 2736
68d480c4
AD		 2737 /* set vmolr receive overflow multicast bit */
2738 vmolr |= E1000_VMOLR_ROMPE;
2739
2740 /* The shared function expects a packed array of only addresses. */
2741 mc_ptr = netdev->mc_list;
2742
2743 for (i = 0; i < netdev->mc_count; i++) {
2744 if (!mc_ptr)
2745 break;
2746 memcpy(mta_list + (i*ETH_ALEN), mc_ptr->dmi_addr, ETH_ALEN);
2747 mc_ptr = mc_ptr->next;
746b9f02 2748 }
68d480c4
AD		 2749 igb_update_mc_addr_list(hw, mta_list, i);
2750 kfree(mta_list);
2751
2752 return netdev->mc_count;
2753}
2754
2755/**
2756 * igb_write_uc_addr_list - write unicast addresses to RAR table
2757 * @netdev: network interface device structure
2758 *
2759 * Writes unicast address list to the RAR table.
2760 * Returns: -ENOMEM on failure/insufficient address space
2761 * 0 on no addresses written
2762 * X on writing X addresses to the RAR table
2763 **/
2764static int igb_write_uc_addr_list(struct net_device *netdev)
2765{
2766 struct igb_adapter *adapter = netdev_priv(netdev);
2767 struct e1000_hw *hw = &adapter->hw;
2768 unsigned int vfn = adapter->vfs_allocated_count;
2769 unsigned int rar_entries = hw->mac.rar_entry_count - (vfn + 1);
2770 int count = 0;
2771
2772 /* return ENOMEM indicating insufficient memory for addresses */
2773 if (netdev->uc.count > rar_entries)
2774 return -ENOMEM;
9d5c8243 2775
ff41f8dc
AD		 2776 if (netdev->uc.count && rar_entries) {
2777 struct netdev_hw_addr *ha;
2778 list_for_each_entry(ha, &netdev->uc.list, list) {
2779 if (!rar_entries)
2780 break;
26ad9178
AD		 2781 igb_rar_set_qsel(adapter, ha->addr,
2782 rar_entries--,
68d480c4
AD		 2783 vfn);
2784 count++;
ff41f8dc
AD		 2785 }
2786 }
2787 /* write the addresses in reverse order to avoid write combining */
2788 for (; rar_entries > 0 ; rar_entries--) {
2789 wr32(E1000_RAH(rar_entries), 0);
2790 wr32(E1000_RAL(rar_entries), 0);
2791 }
2792 wrfl();
2793
68d480c4
AD		 2794 return count;
2795}
2796
2797/**
2798 * igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
2799 * @netdev: network interface device structure
2800 *
2801 * The set_rx_mode entry point is called whenever the unicast or multicast
2802 * address lists or the network interface flags are updated. This routine is
2803 * responsible for configuring the hardware for proper unicast, multicast,
2804 * promiscuous mode, and all-multi behavior.
2805 **/
2806static void igb_set_rx_mode(struct net_device *netdev)
2807{
2808 struct igb_adapter *adapter = netdev_priv(netdev);
2809 struct e1000_hw *hw = &adapter->hw;
2810 unsigned int vfn = adapter->vfs_allocated_count;
2811 u32 rctl, vmolr = 0;
2812 int count;
2813
2814 /* Check for Promiscuous and All Multicast modes */
2815 rctl = rd32(E1000_RCTL);
2816
2817 /* clear the effected bits */
2818 rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE | E1000_RCTL_VFE);
2819
2820 if (netdev->flags & IFF_PROMISC) {
2821 rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2822 vmolr |= (E1000_VMOLR_ROPE | E1000_VMOLR_MPME);
2823 } else {
2824 if (netdev->flags & IFF_ALLMULTI) {
2825 rctl |= E1000_RCTL_MPE;
2826 vmolr |= E1000_VMOLR_MPME;
2827 } else {
2828 /*
2829 * Write addresses to the MTA, if the attempt fails
2830 * then we should just turn on promiscous mode so
2831 * that we can at least receive multicast traffic
2832 */
2833 count = igb_write_mc_addr_list(netdev);
2834 if (count < 0) {
2835 rctl |= E1000_RCTL_MPE;
2836 vmolr |= E1000_VMOLR_MPME;
2837 } else if (count) {
2838 vmolr |= E1000_VMOLR_ROMPE;
2839 }
2840 }
2841 /*
2842 * Write addresses to available RAR registers, if there is not
2843 * sufficient space to store all the addresses then enable
2844 * unicast promiscous mode
2845 */
2846 count = igb_write_uc_addr_list(netdev);
2847 if (count < 0) {
2848 rctl |= E1000_RCTL_UPE;
2849 vmolr |= E1000_VMOLR_ROPE;
2850 }
2851 rctl |= E1000_RCTL_VFE;
28fc06f5 2852 }
68d480c4 2853 wr32(E1000_RCTL, rctl);
28fc06f5 2854
68d480c4
AD		 2855 /*
2856 * In order to support SR-IOV and eventually VMDq it is necessary to set
2857 * the VMOLR to enable the appropriate modes. Without this workaround
2858 * we will have issues with VLAN tag stripping not being done for frames
2859 * that are only arriving because we are the default pool
2860 */
2861 if (hw->mac.type < e1000_82576)
28fc06f5 2862 return;
9d5c8243 2863
68d480c4
AD		 2864 vmolr |= rd32(E1000_VMOLR(vfn)) &
2865 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
2866 wr32(E1000_VMOLR(vfn), vmolr);
28fc06f5 2867 igb_restore_vf_multicasts(adapter);
9d5c8243
AK		 2868}
2869
2870/* Need to wait a few seconds after link up to get diagnostic information from
2871 * the phy */
2872static void igb_update_phy_info(unsigned long data)
2873{
2874 struct igb_adapter *adapter = (struct igb_adapter *) data;
f5f4cf08 2875 igb_get_phy_info(&adapter->hw);
9d5c8243
AK		 2876}
2877
4d6b725e
AD		 2878/**
2879 * igb_has_link - check shared code for link and determine up/down
2880 * @adapter: pointer to driver private info
2881 **/
2882static bool igb_has_link(struct igb_adapter *adapter)
2883{
2884 struct e1000_hw *hw = &adapter->hw;
2885 bool link_active = false;
2886 s32 ret_val = 0;
2887
2888 /* get_link_status is set on LSC (link status) interrupt or
2889 * rx sequence error interrupt. get_link_status will stay
2890 * false until the e1000_check_for_link establishes link
2891 * for copper adapters ONLY
2892 */
2893 switch (hw->phy.media_type) {
2894 case e1000_media_type_copper:
2895 if (hw->mac.get_link_status) {
2896 ret_val = hw->mac.ops.check_for_link(hw);
2897 link_active = !hw->mac.get_link_status;
2898 } else {
2899 link_active = true;
2900 }
2901 break;
4d6b725e
AD		 2902 case e1000_media_type_internal_serdes:
2903 ret_val = hw->mac.ops.check_for_link(hw);
2904 link_active = hw->mac.serdes_has_link;
2905 break;
2906 default:
2907 case e1000_media_type_unknown:
2908 break;
2909 }
2910
2911 return link_active;
2912}
2913
9d5c8243
AK		 2914/**
2915 * igb_watchdog - Timer Call-back
2916 * @data: pointer to adapter cast into an unsigned long
2917 **/
2918static void igb_watchdog(unsigned long data)
2919{
2920 struct igb_adapter *adapter = (struct igb_adapter *)data;
2921 /* Do the rest outside of interrupt context */
2922 schedule_work(&adapter->watchdog_task);
2923}
2924
2925static void igb_watchdog_task(struct work_struct *work)
2926{
2927 struct igb_adapter *adapter = container_of(work,
2928 struct igb_adapter, watchdog_task);
2929 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK		 2930 struct net_device *netdev = adapter->netdev;
2931 struct igb_ring *tx_ring = adapter->tx_ring;
9d5c8243 2932 u32 link;
7a6ea550 2933 int i;
9d5c8243 2934
4d6b725e
AD		 2935 link = igb_has_link(adapter);
2936 if ((netif_carrier_ok(netdev)) && link)
9d5c8243
AK		 2937 goto link_up;
2938
9d5c8243
AK		 2939 if (link) {
2940 if (!netif_carrier_ok(netdev)) {
2941 u32 ctrl;
2942 hw->mac.ops.get_speed_and_duplex(&adapter->hw,
2943 &adapter->link_speed,
2944 &adapter->link_duplex);
2945
2946 ctrl = rd32(E1000_CTRL);
527d47c1
AD		 2947 /* Links status message must follow this format */
2948 printk(KERN_INFO "igb: %s NIC Link is Up %d Mbps %s, "
9d5c8243 2949 "Flow Control: %s\n",
527d47c1 2950 netdev->name,
9d5c8243
AK		 2951 adapter->link_speed,
2952 adapter->link_duplex == FULL_DUPLEX ?
2953 "Full Duplex" : "Half Duplex",
2954 ((ctrl & E1000_CTRL_TFCE) && (ctrl &
2955 E1000_CTRL_RFCE)) ? "RX/TX" : ((ctrl &
2956 E1000_CTRL_RFCE) ? "RX" : ((ctrl &
2957 E1000_CTRL_TFCE) ? "TX" : "None")));
2958
2959 /* tweak tx_queue_len according to speed/duplex and
2960 * adjust the timeout factor */
2961 netdev->tx_queue_len = adapter->tx_queue_len;
2962 adapter->tx_timeout_factor = 1;
2963 switch (adapter->link_speed) {
2964 case SPEED_10:
2965 netdev->tx_queue_len = 10;
2966 adapter->tx_timeout_factor = 14;
2967 break;
2968 case SPEED_100:
2969 netdev->tx_queue_len = 100;
2970 /* maybe add some timeout factor ? */
2971 break;
2972 }
2973
2974 netif_carrier_on(netdev);
9d5c8243 2975
4ae196df
AD		 2976 igb_ping_all_vfs(adapter);
2977
4b1a9877 2978 /* link state has changed, schedule phy info update */
9d5c8243
AK		 2979 if (!test_bit(__IGB_DOWN, &adapter->state))
2980 mod_timer(&adapter->phy_info_timer,
2981 round_jiffies(jiffies + 2 * HZ));
2982 }
2983 } else {
2984 if (netif_carrier_ok(netdev)) {
2985 adapter->link_speed = 0;
2986 adapter->link_duplex = 0;
527d47c1
AD		 2987 /* Links status message must follow this format */
2988 printk(KERN_INFO "igb: %s NIC Link is Down\n",
2989 netdev->name);
9d5c8243 2990 netif_carrier_off(netdev);
4b1a9877 2991
4ae196df
AD		 2992 igb_ping_all_vfs(adapter);
2993
4b1a9877 2994 /* link state has changed, schedule phy info update */
9d5c8243
AK		 2995 if (!test_bit(__IGB_DOWN, &adapter->state))
2996 mod_timer(&adapter->phy_info_timer,
2997 round_jiffies(jiffies + 2 * HZ));
2998 }
2999 }
3000
3001link_up:
3002 igb_update_stats(adapter);
3003
4b1a9877 3004 hw->mac.tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
9d5c8243 3005 adapter->tpt_old = adapter->stats.tpt;
4b1a9877 3006 hw->mac.collision_delta = adapter->stats.colc - adapter->colc_old;
9d5c8243
AK		 3007 adapter->colc_old = adapter->stats.colc;
3008
3009 adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
3010 adapter->gorc_old = adapter->stats.gorc;
3011 adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
3012 adapter->gotc_old = adapter->stats.gotc;
3013
3014 igb_update_adaptive(&adapter->hw);
3015
3016 if (!netif_carrier_ok(netdev)) {
c493ea45 3017 if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
9d5c8243
AK		 3018 /* We've lost link, so the controller stops DMA,
3019 * but we've got queued Tx work that's never going
3020 * to get done, so reset controller to flush Tx.
3021 * (Do the reset outside of interrupt context). */
3022 adapter->tx_timeout_count++;
3023 schedule_work(&adapter->reset_task);
c2d5ab49
JB		 3024 /* return immediately since reset is imminent */
3025 return;
9d5c8243
AK		 3026 }
3027 }
3028
3029 /* Cause software interrupt to ensure rx ring is cleaned */
7a6ea550 3030 if (adapter->msix_entries) {
047e0030
AD		 3031 u32 eics = 0;
3032 for (i = 0; i < adapter->num_q_vectors; i++) {
3033 struct igb_q_vector *q_vector = adapter->q_vector[i];
3034 eics |= q_vector->eims_value;
3035 }
7a6ea550
AD		 3036 wr32(E1000_EICS, eics);
3037 } else {
3038 wr32(E1000_ICS, E1000_ICS_RXDMT0);
3039 }
9d5c8243
AK		 3040
3041 /* Force detection of hung controller every watchdog period */
3042 tx_ring->detect_tx_hung = true;
3043
3044 /* Reset the timer */
3045 if (!test_bit(__IGB_DOWN, &adapter->state))
3046 mod_timer(&adapter->watchdog_timer,
3047 round_jiffies(jiffies + 2 * HZ));
3048}
3049
3050enum latency_range {
3051 lowest_latency = 0,
3052 low_latency = 1,
3053 bulk_latency = 2,
3054 latency_invalid = 255
3055};
3056
3057
6eb5a7f1
AD		 3058/**
3059 * igb_update_ring_itr - update the dynamic ITR value based on packet size
3060 *
3061 * Stores a new ITR value based on strictly on packet size. This
3062 * algorithm is less sophisticated than that used in igb_update_itr,
3063 * due to the difficulty of synchronizing statistics across multiple
3064 * receive rings. The divisors and thresholds used by this fuction
3065 * were determined based on theoretical maximum wire speed and testing
3066 * data, in order to minimize response time while increasing bulk
3067 * throughput.
3068 * This functionality is controlled by the InterruptThrottleRate module
3069 * parameter (see igb_param.c)
3070 * NOTE: This function is called only when operating in a multiqueue
3071 * receive environment.
047e0030 3072 * @q_vector: pointer to q_vector
6eb5a7f1 3073 **/
047e0030 3074static void igb_update_ring_itr(struct igb_q_vector *q_vector)
9d5c8243 3075{
047e0030 3076 int new_val = q_vector->itr_val;
6eb5a7f1 3077 int avg_wire_size = 0;
047e0030 3078 struct igb_adapter *adapter = q_vector->adapter;
9d5c8243 3079
6eb5a7f1
AD		 3080 /* For non-gigabit speeds, just fix the interrupt rate at 4000
3081 * ints/sec - ITR timer value of 120 ticks.
3082 */
3083 if (adapter->link_speed != SPEED_1000) {
047e0030 3084 new_val = 976;
6eb5a7f1 3085 goto set_itr_val;
9d5c8243 3086 }
047e0030
AD		 3087
3088 if (q_vector->rx_ring && q_vector->rx_ring->total_packets) {
3089 struct igb_ring *ring = q_vector->rx_ring;
3090 avg_wire_size = ring->total_bytes / ring->total_packets;
3091 }
3092
3093 if (q_vector->tx_ring && q_vector->tx_ring->total_packets) {
3094 struct igb_ring *ring = q_vector->tx_ring;
3095 avg_wire_size = max_t(u32, avg_wire_size,
3096 (ring->total_bytes /
3097 ring->total_packets));
3098 }
3099
3100 /* if avg_wire_size isn't set no work was done */
3101 if (!avg_wire_size)
3102 goto clear_counts;
9d5c8243 3103
6eb5a7f1
AD		 3104 /* Add 24 bytes to size to account for CRC, preamble, and gap */
3105 avg_wire_size += 24;
3106
3107 /* Don't starve jumbo frames */
3108 avg_wire_size = min(avg_wire_size, 3000);
9d5c8243 3109
6eb5a7f1
AD		 3110 /* Give a little boost to mid-size frames */
3111 if ((avg_wire_size > 300) && (avg_wire_size < 1200))
3112 new_val = avg_wire_size / 3;
3113 else
3114 new_val = avg_wire_size / 2;
9d5c8243 3115
6eb5a7f1 3116set_itr_val:
047e0030
AD		 3117 if (new_val != q_vector->itr_val) {
3118 q_vector->itr_val = new_val;
3119 q_vector->set_itr = 1;
9d5c8243 3120 }
6eb5a7f1 3121clear_counts:
047e0030
AD		 3122 if (q_vector->rx_ring) {
3123 q_vector->rx_ring->total_bytes = 0;
3124 q_vector->rx_ring->total_packets = 0;
3125 }
3126 if (q_vector->tx_ring) {
3127 q_vector->tx_ring->total_bytes = 0;
3128 q_vector->tx_ring->total_packets = 0;
3129 }
9d5c8243
AK		 3130}
3131
3132/**
3133 * igb_update_itr - update the dynamic ITR value based on statistics
3134 * Stores a new ITR value based on packets and byte
3135 * counts during the last interrupt. The advantage of per interrupt
3136 * computation is faster updates and more accurate ITR for the current
3137 * traffic pattern. Constants in this function were computed
3138 * based on theoretical maximum wire speed and thresholds were set based
3139 * on testing data as well as attempting to minimize response time
3140 * while increasing bulk throughput.
3141 * this functionality is controlled by the InterruptThrottleRate module
3142 * parameter (see igb_param.c)
3143 * NOTE: These calculations are only valid when operating in a single-
3144 * queue environment.
3145 * @adapter: pointer to adapter
047e0030 3146 * @itr_setting: current q_vector->itr_val
9d5c8243
AK		 3147 * @packets: the number of packets during this measurement interval
3148 * @bytes: the number of bytes during this measurement interval
3149 **/
3150static unsigned int igb_update_itr(struct igb_adapter *adapter, u16 itr_setting,
3151 int packets, int bytes)
3152{
3153 unsigned int retval = itr_setting;
3154
3155 if (packets == 0)
3156 goto update_itr_done;
3157
3158 switch (itr_setting) {
3159 case lowest_latency:
3160 /* handle TSO and jumbo frames */
3161 if (bytes/packets > 8000)
3162 retval = bulk_latency;
3163 else if ((packets < 5) && (bytes > 512))
3164 retval = low_latency;
3165 break;
3166 case low_latency: /* 50 usec aka 20000 ints/s */
3167 if (bytes > 10000) {
3168 /* this if handles the TSO accounting */
3169 if (bytes/packets > 8000) {
3170 retval = bulk_latency;
3171 } else if ((packets < 10) || ((bytes/packets) > 1200)) {
3172 retval = bulk_latency;
3173 } else if ((packets > 35)) {
3174 retval = lowest_latency;
3175 }
3176 } else if (bytes/packets > 2000) {
3177 retval = bulk_latency;
3178 } else if (packets <= 2 && bytes < 512) {
3179 retval = lowest_latency;
3180 }
3181 break;
3182 case bulk_latency: /* 250 usec aka 4000 ints/s */
3183 if (bytes > 25000) {
3184 if (packets > 35)
3185 retval = low_latency;
1e5c3d21 3186 } else if (bytes < 1500) {
9d5c8243
AK		 3187 retval = low_latency;
3188 }
3189 break;
3190 }
3191
3192update_itr_done:
3193 return retval;
3194}
3195
6eb5a7f1 3196static void igb_set_itr(struct igb_adapter *adapter)
9d5c8243 3197{
047e0030 3198 struct igb_q_vector *q_vector = adapter->q_vector[0];
9d5c8243 3199 u16 current_itr;
047e0030 3200 u32 new_itr = q_vector->itr_val;
9d5c8243
AK		 3201
3202 /* for non-gigabit speeds, just fix the interrupt rate at 4000 */
3203 if (adapter->link_speed != SPEED_1000) {
3204 current_itr = 0;
3205 new_itr = 4000;
3206 goto set_itr_now;
3207 }
3208
3209 adapter->rx_itr = igb_update_itr(adapter,
3210 adapter->rx_itr,
3211 adapter->rx_ring->total_packets,
3212 adapter->rx_ring->total_bytes);
9d5c8243 3213
047e0030
AD		 3214 adapter->tx_itr = igb_update_itr(adapter,
3215 adapter->tx_itr,
3216 adapter->tx_ring->total_packets,
3217 adapter->tx_ring->total_bytes);
3218 current_itr = max(adapter->rx_itr, adapter->tx_itr);
9d5c8243 3219
6eb5a7f1 3220 /* conservative mode (itr 3) eliminates the lowest_latency setting */
73cd78f1 3221 if (adapter->itr_setting == 3 && current_itr == lowest_latency)
6eb5a7f1
AD		 3222 current_itr = low_latency;
3223
9d5c8243
AK		 3224 switch (current_itr) {
3225 /* counts and packets in update_itr are dependent on these numbers */
3226 case lowest_latency:
78b1f607 3227 new_itr = 56; /* aka 70,000 ints/sec */
9d5c8243
AK		 3228 break;
3229 case low_latency:
78b1f607 3230 new_itr = 196; /* aka 20,000 ints/sec */
9d5c8243
AK		 3231 break;
3232 case bulk_latency:
78b1f607 3233 new_itr = 980; /* aka 4,000 ints/sec */
9d5c8243
AK		 3234 break;
3235 default:
3236 break;
3237 }
3238
3239set_itr_now:
6eb5a7f1
AD		 3240 adapter->rx_ring->total_bytes = 0;
3241 adapter->rx_ring->total_packets = 0;
047e0030
AD		 3242 adapter->tx_ring->total_bytes = 0;
3243 adapter->tx_ring->total_packets = 0;
6eb5a7f1 3244
047e0030 3245 if (new_itr != q_vector->itr_val) {
9d5c8243
AK		 3246 /* this attempts to bias the interrupt rate towards Bulk
3247 * by adding intermediate steps when interrupt rate is
3248 * increasing */
047e0030
AD		 3249 new_itr = new_itr > q_vector->itr_val ?
3250 max((new_itr * q_vector->itr_val) /
3251 (new_itr + (q_vector->itr_val >> 2)),
3252 new_itr) :
9d5c8243
AK		 3253 new_itr;
3254 /* Don't write the value here; it resets the adapter's
3255 * internal timer, and causes us to delay far longer than
3256 * we should between interrupts. Instead, we write the ITR
3257 * value at the beginning of the next interrupt so the timing
3258 * ends up being correct.
3259 */
047e0030
AD		 3260 q_vector->itr_val = new_itr;
3261 q_vector->set_itr = 1;
9d5c8243
AK		 3262 }
3263
3264 return;
3265}
3266
9d5c8243
AK		 3267#define IGB_TX_FLAGS_CSUM 0x00000001
3268#define IGB_TX_FLAGS_VLAN 0x00000002
3269#define IGB_TX_FLAGS_TSO 0x00000004
3270#define IGB_TX_FLAGS_IPV4 0x00000008
33af6bcc 3271#define IGB_TX_FLAGS_TSTAMP 0x00000010
9d5c8243
AK		 3272#define IGB_TX_FLAGS_VLAN_MASK 0xffff0000
3273#define IGB_TX_FLAGS_VLAN_SHIFT 16
3274
3275static inline int igb_tso_adv(struct igb_adapter *adapter,
3276 struct igb_ring *tx_ring,
3277 struct sk_buff *skb, u32 tx_flags, u8 *hdr_len)
3278{
3279 struct e1000_adv_tx_context_desc *context_desc;
3280 unsigned int i;
3281 int err;
3282 struct igb_buffer *buffer_info;
3283 u32 info = 0, tu_cmd = 0;
3284 u32 mss_l4len_idx, l4len;
3285 *hdr_len = 0;
3286
3287 if (skb_header_cloned(skb)) {
3288 err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3289 if (err)
3290 return err;
3291 }
3292
3293 l4len = tcp_hdrlen(skb);
3294 *hdr_len += l4len;
3295
3296 if (skb->protocol == htons(ETH_P_IP)) {
3297 struct iphdr *iph = ip_hdr(skb);
3298 iph->tot_len = 0;
3299 iph->check = 0;
3300 tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
3301 iph->daddr, 0,
3302 IPPROTO_TCP,
3303 0);
3304 } else if (skb_shinfo(skb)->gso_type == SKB_GSO_TCPV6) {
3305 ipv6_hdr(skb)->payload_len = 0;
3306 tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
3307 &ipv6_hdr(skb)->daddr,
3308 0, IPPROTO_TCP, 0);
3309 }
3310
3311 i = tx_ring->next_to_use;
3312
3313 buffer_info = &tx_ring->buffer_info[i];
3314 context_desc = E1000_TX_CTXTDESC_ADV(*tx_ring, i);
3315 /* VLAN MACLEN IPLEN */
3316 if (tx_flags & IGB_TX_FLAGS_VLAN)
3317 info |= (tx_flags & IGB_TX_FLAGS_VLAN_MASK);
3318 info |= (skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT);
3319 *hdr_len += skb_network_offset(skb);
3320 info |= skb_network_header_len(skb);
3321 *hdr_len += skb_network_header_len(skb);
3322 context_desc->vlan_macip_lens = cpu_to_le32(info);
3323
3324 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
3325 tu_cmd |= (E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT);
3326
3327 if (skb->protocol == htons(ETH_P_IP))
3328 tu_cmd |= E1000_ADVTXD_TUCMD_IPV4;
3329 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
3330
3331 context_desc->type_tucmd_mlhl = cpu_to_le32(tu_cmd);
3332
3333 /* MSS L4LEN IDX */
3334 mss_l4len_idx = (skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT);
3335 mss_l4len_idx |= (l4len << E1000_ADVTXD_L4LEN_SHIFT);
3336
73cd78f1 3337 /* For 82575, context index must be unique per ring. */
7dfc16fa
AD		 3338 if (adapter->flags & IGB_FLAG_NEED_CTX_IDX)
3339 mss_l4len_idx |= tx_ring->queue_index << 4;
9d5c8243
AK		 3340
3341 context_desc->mss_l4len_idx = cpu_to_le32(mss_l4len_idx);
3342 context_desc->seqnum_seed = 0;
3343
3344 buffer_info->time_stamp = jiffies;
0e014cb1 3345 buffer_info->next_to_watch = i;
9d5c8243
AK		 3346 buffer_info->dma = 0;
3347 i++;
3348 if (i == tx_ring->count)
3349 i = 0;
3350
3351 tx_ring->next_to_use = i;
3352
3353 return true;
3354}
3355
3356static inline bool igb_tx_csum_adv(struct igb_adapter *adapter,
3357 struct igb_ring *tx_ring,
3358 struct sk_buff *skb, u32 tx_flags)
3359{
3360 struct e1000_adv_tx_context_desc *context_desc;
80785298 3361 struct pci_dev *pdev = tx_ring->pdev;
9d5c8243
AK		 3362 struct igb_buffer *buffer_info;
3363 u32 info = 0, tu_cmd = 0;
80785298 3364 unsigned int i;
9d5c8243
AK		 3365
3366 if ((skb->ip_summed == CHECKSUM_PARTIAL) ||
3367 (tx_flags & IGB_TX_FLAGS_VLAN)) {
3368 i = tx_ring->next_to_use;
3369 buffer_info = &tx_ring->buffer_info[i];
3370 context_desc = E1000_TX_CTXTDESC_ADV(*tx_ring, i);
3371
3372 if (tx_flags & IGB_TX_FLAGS_VLAN)
3373 info |= (tx_flags & IGB_TX_FLAGS_VLAN_MASK);
3374 info |= (skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT);
3375 if (skb->ip_summed == CHECKSUM_PARTIAL)
3376 info |= skb_network_header_len(skb);
3377
3378 context_desc->vlan_macip_lens = cpu_to_le32(info);
3379
3380 tu_cmd |= (E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT);
3381
3382 if (skb->ip_summed == CHECKSUM_PARTIAL) {
fa4a7ef3
AJ		 3383 __be16 protocol;
3384
3385 if (skb->protocol == cpu_to_be16(ETH_P_8021Q)) {
3386 const struct vlan_ethhdr *vhdr =
3387 (const struct vlan_ethhdr*)skb->data;
3388
3389 protocol = vhdr->h_vlan_encapsulated_proto;
3390 } else {
3391 protocol = skb->protocol;
3392 }
3393
3394 switch (protocol) {
09640e63 3395 case cpu_to_be16(ETH_P_IP):
9d5c8243 3396 tu_cmd |= E1000_ADVTXD_TUCMD_IPV4;
44b0cda3
MW		 3397 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
3398 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
b9473560
JB		 3399 else if (ip_hdr(skb)->protocol == IPPROTO_SCTP)
3400 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_SCTP;
44b0cda3 3401 break;
09640e63 3402 case cpu_to_be16(ETH_P_IPV6):
44b0cda3
MW		 3403 /* XXX what about other V6 headers?? */
3404 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
3405 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
b9473560
JB		 3406 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_SCTP)
3407 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_SCTP;
44b0cda3
MW		 3408 break;
3409 default:
3410 if (unlikely(net_ratelimit()))
80785298 3411 dev_warn(&pdev->dev,
44b0cda3
MW		 3412 "partial checksum but proto=%x!\n",
3413 skb->protocol);
3414 break;
3415 }
9d5c8243
AK		 3416 }
3417
3418 context_desc->type_tucmd_mlhl = cpu_to_le32(tu_cmd);
3419 context_desc->seqnum_seed = 0;
7dfc16fa
AD		 3420 if (adapter->flags & IGB_FLAG_NEED_CTX_IDX)
3421 context_desc->mss_l4len_idx =
3422 cpu_to_le32(tx_ring->queue_index << 4);
265de409
AD		 3423 else
3424 context_desc->mss_l4len_idx = 0;
9d5c8243
AK		 3425
3426 buffer_info->time_stamp = jiffies;
0e014cb1 3427 buffer_info->next_to_watch = i;
9d5c8243
AK		 3428 buffer_info->dma = 0;
3429
3430 i++;
3431 if (i == tx_ring->count)
3432 i = 0;
3433 tx_ring->next_to_use = i;
3434
3435 return true;
3436 }
9d5c8243
AK		 3437 return false;
3438}
3439
3440#define IGB_MAX_TXD_PWR 16
3441#define IGB_MAX_DATA_PER_TXD (1<<IGB_MAX_TXD_PWR)
3442
80785298 3443static inline int igb_tx_map_adv(struct igb_ring *tx_ring, struct sk_buff *skb,
0e014cb1 3444 unsigned int first)
9d5c8243
AK		 3445{
3446 struct igb_buffer *buffer_info;
80785298 3447 struct pci_dev *pdev = tx_ring->pdev;
9d5c8243
AK		 3448 unsigned int len = skb_headlen(skb);
3449 unsigned int count = 0, i;
3450 unsigned int f;
65689fef 3451 dma_addr_t *map;
9d5c8243
AK		 3452
3453 i = tx_ring->next_to_use;
3454
80785298
AD		 3455 if (skb_dma_map(&pdev->dev, skb, DMA_TO_DEVICE)) {
3456 dev_err(&pdev->dev, "TX DMA map failed\n");
65689fef
AD		 3457 return 0;
3458 }
3459
3460 map = skb_shinfo(skb)->dma_maps;
3461
9d5c8243
AK		 3462 buffer_info = &tx_ring->buffer_info[i];
3463 BUG_ON(len >= IGB_MAX_DATA_PER_TXD);
3464 buffer_info->length = len;
3465 /* set time_stamp *before* dma to help avoid a possible race */
3466 buffer_info->time_stamp = jiffies;
0e014cb1 3467 buffer_info->next_to_watch = i;
042a53a9 3468 buffer_info->dma = skb_shinfo(skb)->dma_head;
9d5c8243
AK		 3469
3470 for (f = 0; f < skb_shinfo(skb)->nr_frags; f++) {
3471 struct skb_frag_struct *frag;
3472
65689fef
AD		 3473 i++;
3474 if (i == tx_ring->count)
3475 i = 0;
3476
9d5c8243
AK		 3477 frag = &skb_shinfo(skb)->frags[f];
3478 len = frag->size;
3479
3480 buffer_info = &tx_ring->buffer_info[i];
3481 BUG_ON(len >= IGB_MAX_DATA_PER_TXD);
3482 buffer_info->length = len;
3483 buffer_info->time_stamp = jiffies;
0e014cb1 3484 buffer_info->next_to_watch = i;
65689fef 3485 buffer_info->dma = map[count];
9d5c8243 3486 count++;
9d5c8243
AK		 3487 }
3488
9d5c8243 3489 tx_ring->buffer_info[i].skb = skb;
0e014cb1 3490 tx_ring->buffer_info[first].next_to_watch = i;
9d5c8243 3491
042a53a9 3492 return count + 1;
9d5c8243
AK		 3493}
3494
3495static inline void igb_tx_queue_adv(struct igb_adapter *adapter,
3496 struct igb_ring *tx_ring,
3497 int tx_flags, int count, u32 paylen,
3498 u8 hdr_len)
3499{
3500 union e1000_adv_tx_desc *tx_desc = NULL;
3501 struct igb_buffer *buffer_info;
3502 u32 olinfo_status = 0, cmd_type_len;
3503 unsigned int i;
3504
3505 cmd_type_len = (E1000_ADVTXD_DTYP_DATA | E1000_ADVTXD_DCMD_IFCS |
3506 E1000_ADVTXD_DCMD_DEXT);
3507
3508 if (tx_flags & IGB_TX_FLAGS_VLAN)
3509 cmd_type_len |= E1000_ADVTXD_DCMD_VLE;
3510
33af6bcc
PO		 3511 if (tx_flags & IGB_TX_FLAGS_TSTAMP)
3512 cmd_type_len |= E1000_ADVTXD_MAC_TSTAMP;
3513
9d5c8243
AK		 3514 if (tx_flags & IGB_TX_FLAGS_TSO) {
3515 cmd_type_len |= E1000_ADVTXD_DCMD_TSE;
3516
3517 /* insert tcp checksum */
3518 olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
3519
3520 /* insert ip checksum */
3521 if (tx_flags & IGB_TX_FLAGS_IPV4)
3522 olinfo_status |= E1000_TXD_POPTS_IXSM << 8;
3523
3524 } else if (tx_flags & IGB_TX_FLAGS_CSUM) {
3525 olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
3526 }
3527
7dfc16fa
AD		 3528 if ((adapter->flags & IGB_FLAG_NEED_CTX_IDX) &&
3529 (tx_flags & (IGB_TX_FLAGS_CSUM | IGB_TX_FLAGS_TSO |
3530 IGB_TX_FLAGS_VLAN)))
661086df 3531 olinfo_status |= tx_ring->queue_index << 4;
9d5c8243
AK		 3532
3533 olinfo_status |= ((paylen - hdr_len) << E1000_ADVTXD_PAYLEN_SHIFT);
3534
3535 i = tx_ring->next_to_use;
3536 while (count--) {
3537 buffer_info = &tx_ring->buffer_info[i];
3538 tx_desc = E1000_TX_DESC_ADV(*tx_ring, i);
3539 tx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
3540 tx_desc->read.cmd_type_len =
3541 cpu_to_le32(cmd_type_len | buffer_info->length);
3542 tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
3543 i++;
3544 if (i == tx_ring->count)
3545 i = 0;
3546 }
3547
3548 tx_desc->read.cmd_type_len |= cpu_to_le32(adapter->txd_cmd);
3549 /* Force memory writes to complete before letting h/w
3550 * know there are new descriptors to fetch. (Only
3551 * applicable for weak-ordered memory model archs,
3552 * such as IA-64). */
3553 wmb();
3554
3555 tx_ring->next_to_use = i;
fce99e34 3556 writel(i, tx_ring->tail);
9d5c8243
AK		 3557 /* we need this if more than one processor can write to our tail
3558 * at a time, it syncronizes IO on IA64/Altix systems */
3559 mmiowb();
3560}
3561
3562static int __igb_maybe_stop_tx(struct net_device *netdev,
3563 struct igb_ring *tx_ring, int size)
3564{
661086df 3565 netif_stop_subqueue(netdev, tx_ring->queue_index);
661086df 3566
9d5c8243
AK		 3567 /* Herbert's original patch had:
3568 * smp_mb__after_netif_stop_queue();
3569 * but since that doesn't exist yet, just open code it. */
3570 smp_mb();
3571
3572 /* We need to check again in a case another CPU has just
3573 * made room available. */
c493ea45 3574 if (igb_desc_unused(tx_ring) < size)
9d5c8243
AK		 3575 return -EBUSY;
3576
3577 /* A reprieve! */
661086df 3578 netif_wake_subqueue(netdev, tx_ring->queue_index);
04a5fcaa 3579 tx_ring->tx_stats.restart_queue++;
9d5c8243
AK		 3580 return 0;
3581}
3582
3583static int igb_maybe_stop_tx(struct net_device *netdev,
3584 struct igb_ring *tx_ring, int size)
3585{
c493ea45 3586 if (igb_desc_unused(tx_ring) >= size)
9d5c8243
AK		 3587 return 0;
3588 return __igb_maybe_stop_tx(netdev, tx_ring, size);
3589}
3590
3b29a56d
SH		 3591static netdev_tx_t igb_xmit_frame_ring_adv(struct sk_buff *skb,
3592 struct net_device *netdev,
3593 struct igb_ring *tx_ring)
9d5c8243
AK		 3594{
3595 struct igb_adapter *adapter = netdev_priv(netdev);
0e014cb1 3596 unsigned int first;
9d5c8243 3597 unsigned int tx_flags = 0;
9d5c8243 3598 u8 hdr_len = 0;
65689fef 3599 int count = 0;
9d5c8243 3600 int tso = 0;
33af6bcc 3601 union skb_shared_tx *shtx;
9d5c8243 3602
9d5c8243
AK		 3603 if (test_bit(__IGB_DOWN, &adapter->state)) {
3604 dev_kfree_skb_any(skb);
3605 return NETDEV_TX_OK;
3606 }
3607
3608 if (skb->len <= 0) {
3609 dev_kfree_skb_any(skb);
3610 return NETDEV_TX_OK;
3611 }
3612
9d5c8243
AK		 3613 /* need: 1 descriptor per page,
3614 * + 2 desc gap to keep tail from touching head,
3615 * + 1 desc for skb->data,
3616 * + 1 desc for context descriptor,
3617 * otherwise try next time */
3618 if (igb_maybe_stop_tx(netdev, tx_ring, skb_shinfo(skb)->nr_frags + 4)) {
3619 /* this is a hard error */
9d5c8243
AK		 3620 return NETDEV_TX_BUSY;
3621 }
33af6bcc
PO		 3622
3623 /*
3624 * TODO: check that there currently is no other packet with
3625 * time stamping in the queue
3626 *
3627 * When doing time stamping, keep the connection to the socket
3628 * a while longer: it is still needed by skb_hwtstamp_tx(),
3629 * called either in igb_tx_hwtstamp() or by our caller when
3630 * doing software time stamping.
3631 */
3632 shtx = skb_tx(skb);
3633 if (unlikely(shtx->hardware)) {
3634 shtx->in_progress = 1;
3635 tx_flags |= IGB_TX_FLAGS_TSTAMP;
33af6bcc 3636 }
9d5c8243
AK		 3637
3638 if (adapter->vlgrp && vlan_tx_tag_present(skb)) {
3639 tx_flags |= IGB_TX_FLAGS_VLAN;
3640 tx_flags |= (vlan_tx_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
3641 }
3642
661086df
PWJ		 3643 if (skb->protocol == htons(ETH_P_IP))
3644 tx_flags |= IGB_TX_FLAGS_IPV4;
3645
0e014cb1 3646 first = tx_ring->next_to_use;
9d5c8243
AK		 3647 tso = skb_is_gso(skb) ? igb_tso_adv(adapter, tx_ring, skb, tx_flags,
3648 &hdr_len) : 0;
3649
3650 if (tso < 0) {
3651 dev_kfree_skb_any(skb);
9d5c8243
AK		 3652 return NETDEV_TX_OK;
3653 }
3654
3655 if (tso)
3656 tx_flags |= IGB_TX_FLAGS_TSO;
bc1cbd34
AD		 3657 else if (igb_tx_csum_adv(adapter, tx_ring, skb, tx_flags) &&
3658 (skb->ip_summed == CHECKSUM_PARTIAL))
3659 tx_flags |= IGB_TX_FLAGS_CSUM;
9d5c8243 3660
65689fef
AD		 3661 /*
3662 * count reflects descriptors mapped, if 0 then mapping error
3663 * has occured and we need to rewind the descriptor queue
3664 */
80785298 3665 count = igb_tx_map_adv(tx_ring, skb, first);
65689fef
AD		 3666
3667 if (count) {
3668 igb_tx_queue_adv(adapter, tx_ring, tx_flags, count,
3669 skb->len, hdr_len);
65689fef
AD		 3670 /* Make sure there is space in the ring for the next send. */
3671 igb_maybe_stop_tx(netdev, tx_ring, MAX_SKB_FRAGS + 4);
3672 } else {
3673 dev_kfree_skb_any(skb);
3674 tx_ring->buffer_info[first].time_stamp = 0;
3675 tx_ring->next_to_use = first;
3676 }
9d5c8243 3677
9d5c8243
AK		 3678 return NETDEV_TX_OK;
3679}
3680
3b29a56d
SH		 3681static netdev_tx_t igb_xmit_frame_adv(struct sk_buff *skb,
3682 struct net_device *netdev)
9d5c8243
AK		 3683{
3684 struct igb_adapter *adapter = netdev_priv(netdev);
661086df
PWJ		 3685 struct igb_ring *tx_ring;
3686
661086df 3687 int r_idx = 0;
1bfaf07b 3688 r_idx = skb->queue_mapping & (IGB_ABS_MAX_TX_QUEUES - 1);
661086df 3689 tx_ring = adapter->multi_tx_table[r_idx];
9d5c8243
AK		 3690
3691 /* This goes back to the question of how to logically map a tx queue
3692 * to a flow. Right now, performance is impacted slightly negatively
3693 * if using multiple tx queues. If the stack breaks away from a
3694 * single qdisc implementation, we can look at this again. */
3b29a56d 3695 return igb_xmit_frame_ring_adv(skb, netdev, tx_ring);
9d5c8243
AK		 3696}
3697
3698/**
3699 * igb_tx_timeout - Respond to a Tx Hang
3700 * @netdev: network interface device structure
3701 **/
3702static void igb_tx_timeout(struct net_device *netdev)
3703{
3704 struct igb_adapter *adapter = netdev_priv(netdev);
3705 struct e1000_hw *hw = &adapter->hw;
3706
3707 /* Do the reset outside of interrupt context */
3708 adapter->tx_timeout_count++;
3709 schedule_work(&adapter->reset_task);
265de409
AD		 3710 wr32(E1000_EICS,
3711 (adapter->eims_enable_mask & ~adapter->eims_other));
9d5c8243
AK		 3712}
3713
3714static void igb_reset_task(struct work_struct *work)
3715{
3716 struct igb_adapter *adapter;
3717 adapter = container_of(work, struct igb_adapter, reset_task);
3718
3719 igb_reinit_locked(adapter);
3720}
3721
3722/**
3723 * igb_get_stats - Get System Network Statistics
3724 * @netdev: network interface device structure
3725 *
3726 * Returns the address of the device statistics structure.
3727 * The statistics are actually updated from the timer callback.
3728 **/
73cd78f1 3729static struct net_device_stats *igb_get_stats(struct net_device *netdev)
9d5c8243 3730{
9d5c8243 3731 /* only return the current stats */
8d24e933 3732 return &netdev->stats;
9d5c8243
AK		 3733}
3734
3735/**
3736 * igb_change_mtu - Change the Maximum Transfer Unit
3737 * @netdev: network interface device structure
3738 * @new_mtu: new value for maximum frame size
3739 *
3740 * Returns 0 on success, negative on failure
3741 **/
3742static int igb_change_mtu(struct net_device *netdev, int new_mtu)
3743{
3744 struct igb_adapter *adapter = netdev_priv(netdev);
3745 int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
4c844851 3746 u32 rx_buffer_len, i;
9d5c8243
AK		 3747
3748 if ((max_frame < ETH_ZLEN + ETH_FCS_LEN) ||
3749 (max_frame > MAX_JUMBO_FRAME_SIZE)) {
3750 dev_err(&adapter->pdev->dev, "Invalid MTU setting\n");
3751 return -EINVAL;
3752 }
3753
9d5c8243
AK		 3754 if (max_frame > MAX_STD_JUMBO_FRAME_SIZE) {
3755 dev_err(&adapter->pdev->dev, "MTU > 9216 not supported.\n");
3756 return -EINVAL;
3757 }
3758
3759 while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
3760 msleep(1);
73cd78f1 3761
9d5c8243
AK		 3762 /* igb_down has a dependency on max_frame_size */
3763 adapter->max_frame_size = max_frame;
9d5c8243
AK		 3764 /* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
3765 * means we reserve 2 more, this pushes us to allocate from the next
3766 * larger slab size.
3767 * i.e. RXBUFFER_2048 --> size-4096 slab
3768 */
3769
7d95b717 3770 if (max_frame <= IGB_RXBUFFER_1024)
4c844851 3771 rx_buffer_len = IGB_RXBUFFER_1024;
6ec43fe6 3772 else if (max_frame <= MAXIMUM_ETHERNET_VLAN_SIZE)
4c844851 3773 rx_buffer_len = MAXIMUM_ETHERNET_VLAN_SIZE;
6ec43fe6 3774 else
4c844851
AD		 3775 rx_buffer_len = IGB_RXBUFFER_128;
3776
3777 if (netif_running(netdev))
3778 igb_down(adapter);
9d5c8243
AK		 3779
3780 dev_info(&adapter->pdev->dev, "changing MTU from %d to %d\n",
3781 netdev->mtu, new_mtu);
3782 netdev->mtu = new_mtu;
3783
4c844851
AD		 3784 for (i = 0; i < adapter->num_rx_queues; i++)
3785 adapter->rx_ring[i].rx_buffer_len = rx_buffer_len;
3786
9d5c8243
AK		 3787 if (netif_running(netdev))
3788 igb_up(adapter);
3789 else
3790 igb_reset(adapter);
3791
3792 clear_bit(__IGB_RESETTING, &adapter->state);
3793
3794 return 0;
3795}
3796
3797/**
3798 * igb_update_stats - Update the board statistics counters
3799 * @adapter: board private structure
3800 **/
3801
3802void igb_update_stats(struct igb_adapter *adapter)
3803{
8d24e933 3804 struct net_device *netdev = adapter->netdev;
9d5c8243
AK		 3805 struct e1000_hw *hw = &adapter->hw;
3806 struct pci_dev *pdev = adapter->pdev;
3807 u16 phy_tmp;
3808
3809#define PHY_IDLE_ERROR_COUNT_MASK 0x00FF
3810
3811 /*
3812 * Prevent stats update while adapter is being reset, or if the pci
3813 * connection is down.
3814 */
3815 if (adapter->link_speed == 0)
3816 return;
3817 if (pci_channel_offline(pdev))
3818 return;
3819
3820 adapter->stats.crcerrs += rd32(E1000_CRCERRS);
3821 adapter->stats.gprc += rd32(E1000_GPRC);
3822 adapter->stats.gorc += rd32(E1000_GORCL);
3823 rd32(E1000_GORCH); /* clear GORCL */
3824 adapter->stats.bprc += rd32(E1000_BPRC);
3825 adapter->stats.mprc += rd32(E1000_MPRC);
3826 adapter->stats.roc += rd32(E1000_ROC);
3827
3828 adapter->stats.prc64 += rd32(E1000_PRC64);
3829 adapter->stats.prc127 += rd32(E1000_PRC127);
3830 adapter->stats.prc255 += rd32(E1000_PRC255);
3831 adapter->stats.prc511 += rd32(E1000_PRC511);
3832 adapter->stats.prc1023 += rd32(E1000_PRC1023);
3833 adapter->stats.prc1522 += rd32(E1000_PRC1522);
3834 adapter->stats.symerrs += rd32(E1000_SYMERRS);
3835 adapter->stats.sec += rd32(E1000_SEC);
3836
3837 adapter->stats.mpc += rd32(E1000_MPC);
3838 adapter->stats.scc += rd32(E1000_SCC);
3839 adapter->stats.ecol += rd32(E1000_ECOL);
3840 adapter->stats.mcc += rd32(E1000_MCC);
3841 adapter->stats.latecol += rd32(E1000_LATECOL);
3842 adapter->stats.dc += rd32(E1000_DC);
3843 adapter->stats.rlec += rd32(E1000_RLEC);
3844 adapter->stats.xonrxc += rd32(E1000_XONRXC);
3845 adapter->stats.xontxc += rd32(E1000_XONTXC);
3846 adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
3847 adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
3848 adapter->stats.fcruc += rd32(E1000_FCRUC);
3849 adapter->stats.gptc += rd32(E1000_GPTC);
3850 adapter->stats.gotc += rd32(E1000_GOTCL);
3851 rd32(E1000_GOTCH); /* clear GOTCL */
3852 adapter->stats.rnbc += rd32(E1000_RNBC);
3853 adapter->stats.ruc += rd32(E1000_RUC);
3854 adapter->stats.rfc += rd32(E1000_RFC);
3855 adapter->stats.rjc += rd32(E1000_RJC);
3856 adapter->stats.tor += rd32(E1000_TORH);
3857 adapter->stats.tot += rd32(E1000_TOTH);
3858 adapter->stats.tpr += rd32(E1000_TPR);
3859
3860 adapter->stats.ptc64 += rd32(E1000_PTC64);
3861 adapter->stats.ptc127 += rd32(E1000_PTC127);
3862 adapter->stats.ptc255 += rd32(E1000_PTC255);
3863 adapter->stats.ptc511 += rd32(E1000_PTC511);
3864 adapter->stats.ptc1023 += rd32(E1000_PTC1023);
3865 adapter->stats.ptc1522 += rd32(E1000_PTC1522);
3866
3867 adapter->stats.mptc += rd32(E1000_MPTC);
3868 adapter->stats.bptc += rd32(E1000_BPTC);
3869
3870 /* used for adaptive IFS */
3871
3872 hw->mac.tx_packet_delta = rd32(E1000_TPT);
3873 adapter->stats.tpt += hw->mac.tx_packet_delta;
3874 hw->mac.collision_delta = rd32(E1000_COLC);
3875 adapter->stats.colc += hw->mac.collision_delta;
3876
3877 adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
3878 adapter->stats.rxerrc += rd32(E1000_RXERRC);
3879 adapter->stats.tncrs += rd32(E1000_TNCRS);
3880 adapter->stats.tsctc += rd32(E1000_TSCTC);
3881 adapter->stats.tsctfc += rd32(E1000_TSCTFC);
3882
3883 adapter->stats.iac += rd32(E1000_IAC);
3884 adapter->stats.icrxoc += rd32(E1000_ICRXOC);
3885 adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
3886 adapter->stats.icrxatc += rd32(E1000_ICRXATC);
3887 adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
3888 adapter->stats.ictxatc += rd32(E1000_ICTXATC);
3889 adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
3890 adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
3891 adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
3892
3893 /* Fill out the OS statistics structure */
8d24e933
AK		 3894 netdev->stats.multicast = adapter->stats.mprc;
3895 netdev->stats.collisions = adapter->stats.colc;
9d5c8243
AK		 3896
3897 /* Rx Errors */
3898
8c0ab70a
JDB		 3899 if (hw->mac.type != e1000_82575) {
3900 u32 rqdpc_tmp;
3ea73afa 3901 u64 rqdpc_total = 0;
8c0ab70a
JDB		 3902 int i;
3903 /* Read out drops stats per RX queue. Notice RQDPC (Receive
3904 * Queue Drop Packet Count) stats only gets incremented, if
3905 * the DROP_EN but it set (in the SRRCTL register for that
3906 * queue). If DROP_EN bit is NOT set, then the some what
3907 * equivalent count is stored in RNBC (not per queue basis).
3908 * Also note the drop count is due to lack of available
3909 * descriptors.
3910 */
3911 for (i = 0; i < adapter->num_rx_queues; i++) {
3912 rqdpc_tmp = rd32(E1000_RQDPC(i)) & 0xFFF;
3913 adapter->rx_ring[i].rx_stats.drops += rqdpc_tmp;
3ea73afa 3914 rqdpc_total += adapter->rx_ring[i].rx_stats.drops;
8c0ab70a 3915 }
8d24e933 3916 netdev->stats.rx_fifo_errors = rqdpc_total;
8c0ab70a
JDB		 3917 }
3918
3ea73afa
JDB		 3919 /* Note RNBC (Receive No Buffers Count) is an not an exact
3920 * drop count as the hardware FIFO might save the day. Thats
3921 * one of the reason for saving it in rx_fifo_errors, as its
3922 * potentially not a true drop.
3923 */
8d24e933 3924 netdev->stats.rx_fifo_errors += adapter->stats.rnbc;
3ea73afa 3925
9d5c8243 3926 /* RLEC on some newer hardware can be incorrect so build
8c0ab70a 3927 * our own version based on RUC and ROC */
8d24e933 3928 netdev->stats.rx_errors = adapter->stats.rxerrc +
9d5c8243
AK		 3929 adapter->stats.crcerrs + adapter->stats.algnerrc +
3930 adapter->stats.ruc + adapter->stats.roc +
3931 adapter->stats.cexterr;
8d24e933 3932 netdev->stats.rx_length_errors = adapter->stats.ruc +
9d5c8243 3933 adapter->stats.roc;
8d24e933
AK		 3934 netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
3935 netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
3936 netdev->stats.rx_missed_errors = adapter->stats.mpc;
9d5c8243
AK		 3937
3938 /* Tx Errors */
8d24e933 3939 netdev->stats.tx_errors = adapter->stats.ecol +
9d5c8243 3940 adapter->stats.latecol;
8d24e933
AK		 3941 netdev->stats.tx_aborted_errors = adapter->stats.ecol;
3942 netdev->stats.tx_window_errors = adapter->stats.latecol;
3943 netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
9d5c8243
AK		 3944
3945 /* Tx Dropped needs to be maintained elsewhere */
3946
3947 /* Phy Stats */
3948 if (hw->phy.media_type == e1000_media_type_copper) {
3949 if ((adapter->link_speed == SPEED_1000) &&
73cd78f1 3950 (!igb_read_phy_reg(hw, PHY_1000T_STATUS, &phy_tmp))) {
9d5c8243
AK		 3951 phy_tmp &= PHY_IDLE_ERROR_COUNT_MASK;
3952 adapter->phy_stats.idle_errors += phy_tmp;
3953 }
3954 }
3955
3956 /* Management Stats */
3957 adapter->stats.mgptc += rd32(E1000_MGTPTC);
3958 adapter->stats.mgprc += rd32(E1000_MGTPRC);
3959 adapter->stats.mgpdc += rd32(E1000_MGTPDC);
3960}
3961
9d5c8243
AK		 3962static irqreturn_t igb_msix_other(int irq, void *data)
3963{
047e0030 3964 struct igb_adapter *adapter = data;
9d5c8243 3965 struct e1000_hw *hw = &adapter->hw;
844290e5 3966 u32 icr = rd32(E1000_ICR);
844290e5 3967 /* reading ICR causes bit 31 of EICR to be cleared */
dda0e083 3968
047e0030 3969 if (icr & E1000_ICR_DOUTSYNC) {
dda0e083
AD		 3970 /* HW is reporting DMA is out of sync */
3971 adapter->stats.doosync++;
3972 }
eebbbdba 3973
4ae196df
AD		 3974 /* Check for a mailbox event */
3975 if (icr & E1000_ICR_VMMB)
3976 igb_msg_task(adapter);
3977
3978 if (icr & E1000_ICR_LSC) {
3979 hw->mac.get_link_status = 1;
3980 /* guard against interrupt when we're going down */
3981 if (!test_bit(__IGB_DOWN, &adapter->state))
3982 mod_timer(&adapter->watchdog_timer, jiffies + 1);
3983 }
3984
3985 wr32(E1000_IMS, E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_VMMB);
844290e5 3986 wr32(E1000_EIMS, adapter->eims_other);
9d5c8243
AK		 3987
3988 return IRQ_HANDLED;
3989}
3990
047e0030 3991static void igb_write_itr(struct igb_q_vector *q_vector)
9d5c8243 3992{
047e0030 3993 u32 itr_val = q_vector->itr_val & 0x7FFC;
9d5c8243 3994
047e0030
AD		 3995 if (!q_vector->set_itr)
3996 return;
73cd78f1 3997
047e0030
AD		 3998 if (!itr_val)
3999 itr_val = 0x4;
661086df 4000
047e0030
AD		 4001 if (q_vector->itr_shift)
4002 itr_val |= itr_val << q_vector->itr_shift;
661086df 4003 else
047e0030 4004 itr_val |= 0x8000000;
661086df 4005
047e0030
AD		 4006 writel(itr_val, q_vector->itr_register);
4007 q_vector->set_itr = 0;
6eb5a7f1
AD		 4008}
4009
047e0030 4010static irqreturn_t igb_msix_ring(int irq, void *data)
9d5c8243 4011{
047e0030 4012 struct igb_q_vector *q_vector = data;
9d5c8243 4013
047e0030
AD		 4014 /* Write the ITR value calculated from the previous interrupt. */
4015 igb_write_itr(q_vector);
9d5c8243 4016
047e0030 4017 napi_schedule(&q_vector->napi);
844290e5 4018
047e0030 4019 return IRQ_HANDLED;
fe4506b6
JC		 4020}
4021
421e02f0 4022#ifdef CONFIG_IGB_DCA
047e0030 4023static void igb_update_dca(struct igb_q_vector *q_vector)
fe4506b6 4024{
047e0030 4025 struct igb_adapter *adapter = q_vector->adapter;
fe4506b6
JC		 4026 struct e1000_hw *hw = &adapter->hw;
4027 int cpu = get_cpu();
fe4506b6 4028
047e0030
AD		 4029 if (q_vector->cpu == cpu)
4030 goto out_no_update;
4031
4032 if (q_vector->tx_ring) {
4033 int q = q_vector->tx_ring->reg_idx;
4034 u32 dca_txctrl = rd32(E1000_DCA_TXCTRL(q));
4035 if (hw->mac.type == e1000_82575) {
4036 dca_txctrl &= ~E1000_DCA_TXCTRL_CPUID_MASK;
4037 dca_txctrl |= dca3_get_tag(&adapter->pdev->dev, cpu);
2d064c06 4038 } else {
047e0030
AD		 4039 dca_txctrl &= ~E1000_DCA_TXCTRL_CPUID_MASK_82576;
4040 dca_txctrl |= dca3_get_tag(&adapter->pdev->dev, cpu) <<
4041 E1000_DCA_TXCTRL_CPUID_SHIFT;
4042 }
4043 dca_txctrl |= E1000_DCA_TXCTRL_DESC_DCA_EN;
4044 wr32(E1000_DCA_TXCTRL(q), dca_txctrl);
4045 }
4046 if (q_vector->rx_ring) {
4047 int q = q_vector->rx_ring->reg_idx;
4048 u32 dca_rxctrl = rd32(E1000_DCA_RXCTRL(q));
4049 if (hw->mac.type == e1000_82575) {
2d064c06 4050 dca_rxctrl &= ~E1000_DCA_RXCTRL_CPUID_MASK;
92be7917 4051 dca_rxctrl |= dca3_get_tag(&adapter->pdev->dev, cpu);
047e0030
AD		 4052 } else {
4053 dca_rxctrl &= ~E1000_DCA_RXCTRL_CPUID_MASK_82576;
4054 dca_rxctrl |= dca3_get_tag(&adapter->pdev->dev, cpu) <<
4055 E1000_DCA_RXCTRL_CPUID_SHIFT;
2d064c06 4056 }
fe4506b6
JC		 4057 dca_rxctrl |= E1000_DCA_RXCTRL_DESC_DCA_EN;
4058 dca_rxctrl |= E1000_DCA_RXCTRL_HEAD_DCA_EN;
4059 dca_rxctrl |= E1000_DCA_RXCTRL_DATA_DCA_EN;
4060 wr32(E1000_DCA_RXCTRL(q), dca_rxctrl);
fe4506b6 4061 }
047e0030
AD		 4062 q_vector->cpu = cpu;
4063out_no_update:
fe4506b6
JC		 4064 put_cpu();
4065}
4066
4067static void igb_setup_dca(struct igb_adapter *adapter)
4068{
7e0e99ef 4069 struct e1000_hw *hw = &adapter->hw;
fe4506b6
JC		 4070 int i;
4071
7dfc16fa 4072 if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
fe4506b6
JC		 4073 return;
4074
7e0e99ef
AD		 4075 /* Always use CB2 mode, difference is masked in the CB driver. */
4076 wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
4077
047e0030
AD		 4078 for (i = 0; i < adapter->num_q_vectors; i++) {
4079 struct igb_q_vector *q_vector = adapter->q_vector[i];
4080 q_vector->cpu = -1;
4081 igb_update_dca(q_vector);
fe4506b6
JC		 4082 }
4083}
4084
4085static int __igb_notify_dca(struct device *dev, void *data)
4086{
4087 struct net_device *netdev = dev_get_drvdata(dev);
4088 struct igb_adapter *adapter = netdev_priv(netdev);
4089 struct e1000_hw *hw = &adapter->hw;
4090 unsigned long event = *(unsigned long *)data;
4091
4092 switch (event) {
4093 case DCA_PROVIDER_ADD:
4094 /* if already enabled, don't do it again */
7dfc16fa 4095 if (adapter->flags & IGB_FLAG_DCA_ENABLED)
fe4506b6 4096 break;
fe4506b6
JC		 4097 /* Always use CB2 mode, difference is masked
4098 * in the CB driver. */
cbd347ad 4099 wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
fe4506b6 4100 if (dca_add_requester(dev) == 0) {
bbd98fe4 4101 adapter->flags |= IGB_FLAG_DCA_ENABLED;
fe4506b6
JC		 4102 dev_info(&adapter->pdev->dev, "DCA enabled\n");
4103 igb_setup_dca(adapter);
4104 break;
4105 }
4106 /* Fall Through since DCA is disabled. */
4107 case DCA_PROVIDER_REMOVE:
7dfc16fa 4108 if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
fe4506b6 4109 /* without this a class_device is left
047e0030 4110 * hanging around in the sysfs model */
fe4506b6
JC		 4111 dca_remove_requester(dev);
4112 dev_info(&adapter->pdev->dev, "DCA disabled\n");
7dfc16fa 4113 adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
cbd347ad 4114 wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
fe4506b6
JC		 4115 }
4116 break;
4117 }
bbd98fe4 4118
fe4506b6 4119 return 0;
9d5c8243
AK		 4120}
4121
fe4506b6
JC		 4122static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
4123 void *p)
4124{
4125 int ret_val;
4126
4127 ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
4128 __igb_notify_dca);
4129
4130 return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
4131}
421e02f0 4132#endif /* CONFIG_IGB_DCA */
9d5c8243 4133
4ae196df
AD		 4134static void igb_ping_all_vfs(struct igb_adapter *adapter)
4135{
4136 struct e1000_hw *hw = &adapter->hw;
4137 u32 ping;
4138 int i;
4139
4140 for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
4141 ping = E1000_PF_CONTROL_MSG;
4142 if (adapter->vf_data[i].clear_to_send)
4143 ping |= E1000_VT_MSGTYPE_CTS;
4144 igb_write_mbx(hw, &ping, 1, i);
4145 }
4146}
4147
4148static int igb_set_vf_multicasts(struct igb_adapter *adapter,
4149 u32 *msgbuf, u32 vf)
4150{
4151 int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
4152 u16 *hash_list = (u16 *)&msgbuf[1];
4153 struct vf_data_storage *vf_data = &adapter->vf_data[vf];
4154 int i;
4155
4156 /* only up to 30 hash values supported */
4157 if (n > 30)
4158 n = 30;
4159
4160 /* salt away the number of multi cast addresses assigned
4161 * to this VF for later use to restore when the PF multi cast
4162 * list changes
4163 */
4164 vf_data->num_vf_mc_hashes = n;
4165
4166 /* VFs are limited to using the MTA hash table for their multicast
4167 * addresses */
4168 for (i = 0; i < n; i++)
a419aef8 4169 vf_data->vf_mc_hashes[i] = hash_list[i];
4ae196df
AD		 4170
4171 /* Flush and reset the mta with the new values */
ff41f8dc 4172 igb_set_rx_mode(adapter->netdev);
4ae196df
AD		 4173
4174 return 0;
4175}
4176
4177static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
4178{
4179 struct e1000_hw *hw = &adapter->hw;
4180 struct vf_data_storage *vf_data;
4181 int i, j;
4182
4183 for (i = 0; i < adapter->vfs_allocated_count; i++) {
4184 vf_data = &adapter->vf_data[i];
75f4f382 4185 for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
4ae196df
AD		 4186 igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
4187 }
4188}
4189
4190static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
4191{
4192 struct e1000_hw *hw = &adapter->hw;
4193 u32 pool_mask, reg, vid;
4194 int i;
4195
4196 pool_mask = 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
4197
4198 /* Find the vlan filter for this id */
4199 for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
4200 reg = rd32(E1000_VLVF(i));
4201
4202 /* remove the vf from the pool */
4203 reg &= ~pool_mask;
4204
4205 /* if pool is empty then remove entry from vfta */
4206 if (!(reg & E1000_VLVF_POOLSEL_MASK) &&
4207 (reg & E1000_VLVF_VLANID_ENABLE)) {
4208 reg = 0;
4209 vid = reg & E1000_VLVF_VLANID_MASK;
4210 igb_vfta_set(hw, vid, false);
4211 }
4212
4213 wr32(E1000_VLVF(i), reg);
4214 }
ae641bdc
AD		 4215
4216 adapter->vf_data[vf].vlans_enabled = 0;
4ae196df
AD		 4217}
4218
4219static s32 igb_vlvf_set(struct igb_adapter *adapter, u32 vid, bool add, u32 vf)
4220{
4221 struct e1000_hw *hw = &adapter->hw;
4222 u32 reg, i;
4223
4224 /* It is an error to call this function when VFs are not enabled */
4225 if (!adapter->vfs_allocated_count)
4226 return -1;
4227
4228 /* Find the vlan filter for this id */
4229 for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
4230 reg = rd32(E1000_VLVF(i));
4231 if ((reg & E1000_VLVF_VLANID_ENABLE) &&
4232 vid == (reg & E1000_VLVF_VLANID_MASK))
4233 break;
4234 }
4235
4236 if (add) {
4237 if (i == E1000_VLVF_ARRAY_SIZE) {
4238 /* Did not find a matching VLAN ID entry that was
4239 * enabled. Search for a free filter entry, i.e.
4240 * one without the enable bit set
4241 */
4242 for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) {
4243 reg = rd32(E1000_VLVF(i));
4244 if (!(reg & E1000_VLVF_VLANID_ENABLE))
4245 break;
4246 }
4247 }
4248 if (i < E1000_VLVF_ARRAY_SIZE) {
4249 /* Found an enabled/available entry */
4250 reg |= 1 << (E1000_VLVF_POOLSEL_SHIFT + vf);
4251
4252 /* if !enabled we need to set this up in vfta */
4253 if (!(reg & E1000_VLVF_VLANID_ENABLE)) {
cad6d05f
AD		 4254 /* add VID to filter table, if bit already set
4255 * PF must have added it outside of table */
4256 if (igb_vfta_set(hw, vid, true))
4257 reg |= 1 << (E1000_VLVF_POOLSEL_SHIFT +
4258 adapter->vfs_allocated_count);
4ae196df
AD		 4259 reg |= E1000_VLVF_VLANID_ENABLE;
4260 }
cad6d05f
AD		 4261 reg &= ~E1000_VLVF_VLANID_MASK;
4262 reg |= vid;
4ae196df
AD		 4263
4264 wr32(E1000_VLVF(i), reg);
ae641bdc
AD		 4265
4266 /* do not modify RLPML for PF devices */
4267 if (vf >= adapter->vfs_allocated_count)
4268 return 0;
4269
4270 if (!adapter->vf_data[vf].vlans_enabled) {
4271 u32 size;
4272 reg = rd32(E1000_VMOLR(vf));
4273 size = reg & E1000_VMOLR_RLPML_MASK;
4274 size += 4;
4275 reg &= ~E1000_VMOLR_RLPML_MASK;
4276 reg |= size;
4277 wr32(E1000_VMOLR(vf), reg);
4278 }
4279 adapter->vf_data[vf].vlans_enabled++;
4280
4ae196df
AD		 4281 return 0;
4282 }
4283 } else {
4284 if (i < E1000_VLVF_ARRAY_SIZE) {
4285 /* remove vf from the pool */
4286 reg &= ~(1 << (E1000_VLVF_POOLSEL_SHIFT + vf));
4287 /* if pool is empty then remove entry from vfta */
4288 if (!(reg & E1000_VLVF_POOLSEL_MASK)) {
4289 reg = 0;
4290 igb_vfta_set(hw, vid, false);
4291 }
4292 wr32(E1000_VLVF(i), reg);
ae641bdc
AD		 4293
4294 /* do not modify RLPML for PF devices */
4295 if (vf >= adapter->vfs_allocated_count)
4296 return 0;
4297
4298 adapter->vf_data[vf].vlans_enabled--;
4299 if (!adapter->vf_data[vf].vlans_enabled) {
4300 u32 size;
4301 reg = rd32(E1000_VMOLR(vf));
4302 size = reg & E1000_VMOLR_RLPML_MASK;
4303 size -= 4;
4304 reg &= ~E1000_VMOLR_RLPML_MASK;
4305 reg |= size;
4306 wr32(E1000_VMOLR(vf), reg);
4307 }
4ae196df
AD		 4308 return 0;
4309 }
4310 }
4311 return -1;
4312}
4313
4314static int igb_set_vf_vlan(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
4315{
4316 int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
4317 int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
4318
4319 return igb_vlvf_set(adapter, vid, add, vf);
4320}
4321
4322static inline void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
4323{
4324 struct e1000_hw *hw = &adapter->hw;
4325
4326 /* disable mailbox functionality for vf */
4327 adapter->vf_data[vf].clear_to_send = false;
4328
4329 /* reset offloads to defaults */
4330 igb_set_vmolr(hw, vf);
4331
4332 /* reset vlans for device */
4333 igb_clear_vf_vfta(adapter, vf);
4334
4335 /* reset multicast table array for vf */
4336 adapter->vf_data[vf].num_vf_mc_hashes = 0;
4337
4338 /* Flush and reset the mta with the new values */
ff41f8dc 4339 igb_set_rx_mode(adapter->netdev);
4ae196df
AD		 4340}
4341
4342static inline void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
4343{
4344 struct e1000_hw *hw = &adapter->hw;
4345 unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
ff41f8dc 4346 int rar_entry = hw->mac.rar_entry_count - (vf + 1);
4ae196df
AD		 4347 u32 reg, msgbuf[3];
4348 u8 *addr = (u8 *)(&msgbuf[1]);
4349
4350 /* process all the same items cleared in a function level reset */
4351 igb_vf_reset_event(adapter, vf);
4352
4353 /* set vf mac address */
26ad9178 4354 igb_rar_set_qsel(adapter, vf_mac, rar_entry, vf);
4ae196df
AD		 4355
4356 /* enable transmit and receive for vf */
4357 reg = rd32(E1000_VFTE);
4358 wr32(E1000_VFTE, reg | (1 << vf));
4359 reg = rd32(E1000_VFRE);
4360 wr32(E1000_VFRE, reg | (1 << vf));
4361
4362 /* enable mailbox functionality for vf */
4363 adapter->vf_data[vf].clear_to_send = true;
4364
4365 /* reply to reset with ack and vf mac address */
4366 msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
4367 memcpy(addr, vf_mac, 6);
4368 igb_write_mbx(hw, msgbuf, 3, vf);
4369}
4370
4371static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
4372{
4373 unsigned char *addr = (char *)&msg[1];
4374 int err = -1;
4375
4376 if (is_valid_ether_addr(addr))
4377 err = igb_set_vf_mac(adapter, vf, addr);
4378
4379 return err;
4380
4381}
4382
4383static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
4384{
4385 struct e1000_hw *hw = &adapter->hw;
4386 u32 msg = E1000_VT_MSGTYPE_NACK;
4387
4388 /* if device isn't clear to send it shouldn't be reading either */
4389 if (!adapter->vf_data[vf].clear_to_send)
4390 igb_write_mbx(hw, &msg, 1, vf);
4391}
4392
4393
4394static void igb_msg_task(struct igb_adapter *adapter)
4395{
4396 struct e1000_hw *hw = &adapter->hw;
4397 u32 vf;
4398
4399 for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
4400 /* process any reset requests */
4401 if (!igb_check_for_rst(hw, vf)) {
4402 adapter->vf_data[vf].clear_to_send = false;
4403 igb_vf_reset_event(adapter, vf);
4404 }
4405
4406 /* process any messages pending */
4407 if (!igb_check_for_msg(hw, vf))
4408 igb_rcv_msg_from_vf(adapter, vf);
4409
4410 /* process any acks */
4411 if (!igb_check_for_ack(hw, vf))
4412 igb_rcv_ack_from_vf(adapter, vf);
4413
4414 }
4415}
4416
4417static int igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
4418{
4419 u32 mbx_size = E1000_VFMAILBOX_SIZE;
4420 u32 msgbuf[mbx_size];
4421 struct e1000_hw *hw = &adapter->hw;
4422 s32 retval;
4423
4424 retval = igb_read_mbx(hw, msgbuf, mbx_size, vf);
4425
4426 if (retval)
4427 dev_err(&adapter->pdev->dev,
4428 "Error receiving message from VF\n");
4429
4430 /* this is a message we already processed, do nothing */
4431 if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
4432 return retval;
4433
4434 /*
4435 * until the vf completes a reset it should not be
4436 * allowed to start any configuration.
4437 */
4438
4439 if (msgbuf[0] == E1000_VF_RESET) {
4440 igb_vf_reset_msg(adapter, vf);
4441
4442 return retval;
4443 }
4444
4445 if (!adapter->vf_data[vf].clear_to_send) {
4446 msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
4447 igb_write_mbx(hw, msgbuf, 1, vf);
4448 return retval;
4449 }
4450
4451 switch ((msgbuf[0] & 0xFFFF)) {
4452 case E1000_VF_SET_MAC_ADDR:
4453 retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
4454 break;
4455 case E1000_VF_SET_MULTICAST:
4456 retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
4457 break;
4458 case E1000_VF_SET_LPE:
4459 retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
4460 break;
4461 case E1000_VF_SET_VLAN:
4462 retval = igb_set_vf_vlan(adapter, msgbuf, vf);
4463 break;
4464 default:
4465 dev_err(&adapter->pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
4466 retval = -1;
4467 break;
4468 }
4469
4470 /* notify the VF of the results of what it sent us */
4471 if (retval)
4472 msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
4473 else
4474 msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
4475
4476 msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
4477
4478 igb_write_mbx(hw, msgbuf, 1, vf);
4479
4480 return retval;
4481}
4482
68d480c4
AD		 4483/**
4484 * igb_set_uta - Set unicast filter table address
4485 * @adapter: board private structure
4486 *
4487 * The unicast table address is a register array of 32-bit registers.
4488 * The table is meant to be used in a way similar to how the MTA is used
4489 * however due to certain limitations in the hardware it is necessary to
4490 * set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscous
4491 * enable bit to allow vlan tag stripping when promiscous mode is enabled
4492 **/
4493static void igb_set_uta(struct igb_adapter *adapter)
4494{
4495 struct e1000_hw *hw = &adapter->hw;
4496 int i;
4497
4498 /* The UTA table only exists on 82576 hardware and newer */
4499 if (hw->mac.type < e1000_82576)
4500 return;
4501
4502 /* we only need to do this if VMDq is enabled */
4503 if (!adapter->vfs_allocated_count)
4504 return;
4505
4506 for (i = 0; i < hw->mac.uta_reg_count; i++)
4507 array_wr32(E1000_UTA, i, ~0);
4508}
4509
9d5c8243
AK		 4510/**
4511 * igb_intr_msi - Interrupt Handler
4512 * @irq: interrupt number
4513 * @data: pointer to a network interface device structure
4514 **/
4515static irqreturn_t igb_intr_msi(int irq, void *data)
4516{
047e0030
AD		 4517 struct igb_adapter *adapter = data;
4518 struct igb_q_vector *q_vector = adapter->q_vector[0];
9d5c8243
AK		 4519 struct e1000_hw *hw = &adapter->hw;
4520 /* read ICR disables interrupts using IAM */
4521 u32 icr = rd32(E1000_ICR);
4522
047e0030 4523 igb_write_itr(q_vector);
9d5c8243 4524
047e0030 4525 if (icr & E1000_ICR_DOUTSYNC) {
dda0e083
AD		 4526 /* HW is reporting DMA is out of sync */
4527 adapter->stats.doosync++;
4528 }
4529
9d5c8243
AK		 4530 if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
4531 hw->mac.get_link_status = 1;
4532 if (!test_bit(__IGB_DOWN, &adapter->state))
4533 mod_timer(&adapter->watchdog_timer, jiffies + 1);
4534 }
4535
047e0030 4536 napi_schedule(&q_vector->napi);
9d5c8243
AK		 4537
4538 return IRQ_HANDLED;
4539}
4540
4541/**
4a3c6433 4542 * igb_intr - Legacy Interrupt Handler
9d5c8243
AK		 4543 * @irq: interrupt number
4544 * @data: pointer to a network interface device structure
4545 **/
4546static irqreturn_t igb_intr(int irq, void *data)
4547{
047e0030
AD		 4548 struct igb_adapter *adapter = data;
4549 struct igb_q_vector *q_vector = adapter->q_vector[0];
9d5c8243
AK		 4550 struct e1000_hw *hw = &adapter->hw;
4551 /* Interrupt Auto-Mask...upon reading ICR, interrupts are masked. No
4552 * need for the IMC write */
4553 u32 icr = rd32(E1000_ICR);
9d5c8243
AK		 4554 if (!icr)
4555 return IRQ_NONE; /* Not our interrupt */
4556
047e0030 4557 igb_write_itr(q_vector);
9d5c8243
AK		 4558
4559 /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
4560 * not set, then the adapter didn't send an interrupt */
4561 if (!(icr & E1000_ICR_INT_ASSERTED))
4562 return IRQ_NONE;
4563
047e0030 4564 if (icr & E1000_ICR_DOUTSYNC) {
dda0e083
AD		 4565 /* HW is reporting DMA is out of sync */
4566 adapter->stats.doosync++;
4567 }
4568
9d5c8243
AK		 4569 if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
4570 hw->mac.get_link_status = 1;
4571 /* guard against interrupt when we're going down */
4572 if (!test_bit(__IGB_DOWN, &adapter->state))
4573 mod_timer(&adapter->watchdog_timer, jiffies + 1);
4574 }
4575
047e0030 4576 napi_schedule(&q_vector->napi);
9d5c8243
AK		 4577
4578 return IRQ_HANDLED;
4579}
4580
047e0030 4581static inline void igb_ring_irq_enable(struct igb_q_vector *q_vector)
9d5c8243 4582{
047e0030 4583 struct igb_adapter *adapter = q_vector->adapter;
46544258 4584 struct e1000_hw *hw = &adapter->hw;
9d5c8243 4585
46544258 4586 if (adapter->itr_setting & 3) {
047e0030 4587 if (!adapter->msix_entries)
6eb5a7f1 4588 igb_set_itr(adapter);
46544258 4589 else
047e0030 4590 igb_update_ring_itr(q_vector);
9d5c8243
AK		 4591 }
4592
46544258
AD		 4593 if (!test_bit(__IGB_DOWN, &adapter->state)) {
4594 if (adapter->msix_entries)
047e0030 4595 wr32(E1000_EIMS, q_vector->eims_value);
46544258
AD		 4596 else
4597 igb_irq_enable(adapter);
4598 }
9d5c8243
AK		 4599}
4600
46544258
AD		 4601/**
4602 * igb_poll - NAPI Rx polling callback
4603 * @napi: napi polling structure
4604 * @budget: count of how many packets we should handle
4605 **/
4606static int igb_poll(struct napi_struct *napi, int budget)
9d5c8243 4607{
047e0030
AD		 4608 struct igb_q_vector *q_vector = container_of(napi,
4609 struct igb_q_vector,
4610 napi);
4611 int tx_clean_complete = 1, work_done = 0;
9d5c8243 4612
421e02f0 4613#ifdef CONFIG_IGB_DCA
047e0030
AD		 4614 if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
4615 igb_update_dca(q_vector);
fe4506b6 4616#endif
047e0030
AD		 4617 if (q_vector->tx_ring)
4618 tx_clean_complete = igb_clean_tx_irq(q_vector);
9d5c8243 4619
047e0030
AD		 4620 if (q_vector->rx_ring)
4621 igb_clean_rx_irq_adv(q_vector, &work_done, budget);
4622
4623 if (!tx_clean_complete)
4624 work_done = budget;
46544258 4625
9d5c8243 4626 /* If not enough Rx work done, exit the polling mode */
5e6d5b17 4627 if (work_done < budget) {
288379f0 4628 napi_complete(napi);
047e0030 4629 igb_ring_irq_enable(q_vector);
9d5c8243
AK		 4630 }
4631
46544258 4632 return work_done;
9d5c8243 4633}
6d8126f9 4634
33af6bcc
PO		 4635/**
4636 * igb_hwtstamp - utility function which checks for TX time stamp
4637 * @adapter: board private structure
4638 * @skb: packet that was just sent
4639 *
4640 * If we were asked to do hardware stamping and such a time stamp is
4641 * available, then it must have been for this skb here because we only
4642 * allow only one such packet into the queue.
4643 */
4644static void igb_tx_hwtstamp(struct igb_adapter *adapter, struct sk_buff *skb)
4645{
4646 union skb_shared_tx *shtx = skb_tx(skb);
4647 struct e1000_hw *hw = &adapter->hw;
4648
4649 if (unlikely(shtx->hardware)) {
4650 u32 valid = rd32(E1000_TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID;
4651 if (valid) {
4652 u64 regval = rd32(E1000_TXSTMPL);
4653 u64 ns;
4654 struct skb_shared_hwtstamps shhwtstamps;
4655
4656 memset(&shhwtstamps, 0, sizeof(shhwtstamps));
4657 regval |= (u64)rd32(E1000_TXSTMPH) << 32;
4658 ns = timecounter_cyc2time(&adapter->clock,
4659 regval);
4660 timecompare_update(&adapter->compare, ns);
4661 shhwtstamps.hwtstamp = ns_to_ktime(ns);
4662 shhwtstamps.syststamp =
4663 timecompare_transform(&adapter->compare, ns);
4664 skb_tstamp_tx(skb, &shhwtstamps);
4665 }
33af6bcc
PO		 4666 }
4667}
4668
9d5c8243
AK		 4669/**
4670 * igb_clean_tx_irq - Reclaim resources after transmit completes
047e0030 4671 * @q_vector: pointer to q_vector containing needed info
9d5c8243
AK		 4672 * returns true if ring is completely cleaned
4673 **/
047e0030 4674static bool igb_clean_tx_irq(struct igb_q_vector *q_vector)
9d5c8243 4675{
047e0030
AD		 4676 struct igb_adapter *adapter = q_vector->adapter;
4677 struct igb_ring *tx_ring = q_vector->tx_ring;
3b644cf6 4678 struct net_device *netdev = adapter->netdev;
0e014cb1 4679 struct e1000_hw *hw = &adapter->hw;
9d5c8243
AK		 4680 struct igb_buffer *buffer_info;
4681 struct sk_buff *skb;
0e014cb1 4682 union e1000_adv_tx_desc *tx_desc, *eop_desc;
9d5c8243 4683 unsigned int total_bytes = 0, total_packets = 0;
0e014cb1
AD		 4684 unsigned int i, eop, count = 0;
4685 bool cleaned = false;
9d5c8243 4686
9d5c8243 4687 i = tx_ring->next_to_clean;
0e014cb1
AD		 4688 eop = tx_ring->buffer_info[i].next_to_watch;
4689 eop_desc = E1000_TX_DESC_ADV(*tx_ring, eop);
4690
4691 while ((eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)) &&
4692 (count < tx_ring->count)) {
4693 for (cleaned = false; !cleaned; count++) {
4694 tx_desc = E1000_TX_DESC_ADV(*tx_ring, i);
9d5c8243 4695 buffer_info = &tx_ring->buffer_info[i];
0e014cb1 4696 cleaned = (i == eop);
9d5c8243
AK		 4697 skb = buffer_info->skb;
4698
4699 if (skb) {
4700 unsigned int segs, bytecount;
4701 /* gso_segs is currently only valid for tcp */
4702 segs = skb_shinfo(skb)->gso_segs ?: 1;
4703 /* multiply data chunks by size of headers */
4704 bytecount = ((segs - 1) * skb_headlen(skb)) +
4705 skb->len;
4706 total_packets += segs;
4707 total_bytes += bytecount;
33af6bcc
PO		 4708
4709 igb_tx_hwtstamp(adapter, skb);
9d5c8243
AK		 4710 }
4711
80785298 4712 igb_unmap_and_free_tx_resource(tx_ring, buffer_info);
0e014cb1 4713 tx_desc->wb.status = 0;
9d5c8243
AK		 4714
4715 i++;
4716 if (i == tx_ring->count)
4717 i = 0;
9d5c8243 4718 }
0e014cb1
AD		 4719 eop = tx_ring->buffer_info[i].next_to_watch;
4720 eop_desc = E1000_TX_DESC_ADV(*tx_ring, eop);
4721 }
4722
9d5c8243
AK		 4723 tx_ring->next_to_clean = i;
4724
fc7d345d 4725 if (unlikely(count &&
9d5c8243 4726 netif_carrier_ok(netdev) &&
c493ea45 4727 igb_desc_unused(tx_ring) >= IGB_TX_QUEUE_WAKE)) {
9d5c8243
AK		 4728 /* Make sure that anybody stopping the queue after this
4729 * sees the new next_to_clean.
4730 */
4731 smp_mb();
661086df
PWJ		 4732 if (__netif_subqueue_stopped(netdev, tx_ring->queue_index) &&
4733 !(test_bit(__IGB_DOWN, &adapter->state))) {
4734 netif_wake_subqueue(netdev, tx_ring->queue_index);
04a5fcaa 4735 tx_ring->tx_stats.restart_queue++;
661086df 4736 }
9d5c8243
AK		 4737 }
4738
4739 if (tx_ring->detect_tx_hung) {
4740 /* Detect a transmit hang in hardware, this serializes the
4741 * check with the clearing of time_stamp and movement of i */
4742 tx_ring->detect_tx_hung = false;
4743 if (tx_ring->buffer_info[i].time_stamp &&
4744 time_after(jiffies, tx_ring->buffer_info[i].time_stamp +
4745 (adapter->tx_timeout_factor * HZ))
4746 && !(rd32(E1000_STATUS) &
4747 E1000_STATUS_TXOFF)) {
4748
9d5c8243 4749 /* detected Tx unit hang */
80785298 4750 dev_err(&tx_ring->pdev->dev,
9d5c8243 4751 "Detected Tx Unit Hang\n"
2d064c06 4752 " Tx Queue <%d>\n"
9d5c8243
AK		 4753 " TDH <%x>\n"
4754 " TDT <%x>\n"
4755 " next_to_use <%x>\n"
4756 " next_to_clean <%x>\n"
9d5c8243
AK		 4757 "buffer_info[next_to_clean]\n"
4758 " time_stamp <%lx>\n"
0e014cb1 4759 " next_to_watch <%x>\n"
9d5c8243
AK		 4760 " jiffies <%lx>\n"
4761 " desc.status <%x>\n",
2d064c06 4762 tx_ring->queue_index,
fce99e34
AD		 4763 readl(tx_ring->head),
4764 readl(tx_ring->tail),
9d5c8243
AK		 4765 tx_ring->next_to_use,
4766 tx_ring->next_to_clean,
9d5c8243 4767 tx_ring->buffer_info[i].time_stamp,
0e014cb1 4768 eop,
9d5c8243 4769 jiffies,
0e014cb1 4770 eop_desc->wb.status);
661086df 4771 netif_stop_subqueue(netdev, tx_ring->queue_index);
9d5c8243
AK		 4772 }
4773 }
4774 tx_ring->total_bytes += total_bytes;
4775 tx_ring->total_packets += total_packets;
e21ed353
AD		 4776 tx_ring->tx_stats.bytes += total_bytes;
4777 tx_ring->tx_stats.packets += total_packets;
8d24e933
AK		 4778 netdev->stats.tx_bytes += total_bytes;
4779 netdev->stats.tx_packets += total_packets;
0e014cb1 4780 return (count < tx_ring->count);
9d5c8243
AK		 4781}
4782
9d5c8243
AK		 4783/**
4784 * igb_receive_skb - helper function to handle rx indications
047e0030
AD		 4785 * @q_vector: structure containing interrupt and ring information
4786 * @skb: packet to send up
4787 * @vlan_tag: vlan tag for packet
9d5c8243 4788 **/
047e0030
AD		 4789static void igb_receive_skb(struct igb_q_vector *q_vector,
4790 struct sk_buff *skb,
4791 u16 vlan_tag)
4792{
4793 struct igb_adapter *adapter = q_vector->adapter;
4794
4795 if (vlan_tag)
4796 vlan_gro_receive(&q_vector->napi, adapter->vlgrp,
4797 vlan_tag, skb);
182ff8df 4798 else
047e0030 4799 napi_gro_receive(&q_vector->napi, skb);
9d5c8243
AK		 4800}
4801
04a5fcaa
AD		 4802static inline void igb_rx_checksum_adv(struct igb_ring *ring,
4803 struct igb_adapter *adapter,
9d5c8243
AK		 4804 u32 status_err, struct sk_buff *skb)
4805{
4806 skb->ip_summed = CHECKSUM_NONE;
4807
4808 /* Ignore Checksum bit is set or checksum is disabled through ethtool */
7beb0146
AD		 4809 if ((status_err & E1000_RXD_STAT_IXSM) ||
4810 (adapter->flags & IGB_FLAG_RX_CSUM_DISABLED))
9d5c8243
AK		 4811 return;
4812 /* TCP/UDP checksum error bit is set */
4813 if (status_err &
4814 (E1000_RXDEXT_STATERR_TCPE | E1000_RXDEXT_STATERR_IPE)) {
b9473560
JB		 4815 /*
4816 * work around errata with sctp packets where the TCPE aka
4817 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
4818 * packets, (aka let the stack check the crc32c)
4819 */
4820 if (!((adapter->hw.mac.type == e1000_82576) &&
4821 (skb->len == 60)))
04a5fcaa 4822 ring->rx_stats.csum_err++;
9d5c8243 4823 /* let the stack verify checksum errors */
9d5c8243
AK		 4824 return;
4825 }
4826 /* It must be a TCP or UDP packet with a valid checksum */
4827 if (status_err & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS))
4828 skb->ip_summed = CHECKSUM_UNNECESSARY;
4829
b9473560 4830 dev_dbg(&adapter->pdev->dev, "cksum success: bits %08X\n", status_err);
9d5c8243
AK		 4831}
4832
4c844851 4833static inline u16 igb_get_hlen(struct igb_ring *rx_ring,
2d94d8ab
AD		 4834 union e1000_adv_rx_desc *rx_desc)
4835{
4836 /* HW will not DMA in data larger than the given buffer, even if it
4837 * parses the (NFS, of course) header to be larger. In that case, it
4838 * fills the header buffer and spills the rest into the page.
4839 */
4840 u16 hlen = (le16_to_cpu(rx_desc->wb.lower.lo_dword.hdr_info) &
4841 E1000_RXDADV_HDRBUFLEN_MASK) >> E1000_RXDADV_HDRBUFLEN_SHIFT;
4c844851
AD		 4842 if (hlen > rx_ring->rx_buffer_len)
4843 hlen = rx_ring->rx_buffer_len;
2d94d8ab
AD		 4844 return hlen;
4845}
4846
047e0030
AD		 4847static bool igb_clean_rx_irq_adv(struct igb_q_vector *q_vector,
4848 int *work_done, int budget)
9d5c8243 4849{
047e0030 4850 struct igb_adapter *adapter = q_vector->adapter;
9d5c8243 4851 struct net_device *netdev = adapter->netdev;
047e0030 4852 struct igb_ring *rx_ring = q_vector->rx_ring;
33af6bcc 4853 struct e1000_hw *hw = &adapter->hw;
80785298 4854 struct pci_dev *pdev = rx_ring->pdev;
9d5c8243
AK		 4855 union e1000_adv_rx_desc *rx_desc , *next_rxd;
4856 struct igb_buffer *buffer_info , *next_buffer;
4857 struct sk_buff *skb;
9d5c8243
AK		 4858 bool cleaned = false;
4859 int cleaned_count = 0;
4860 unsigned int total_bytes = 0, total_packets = 0;
73cd78f1 4861 unsigned int i;
2d94d8ab
AD		 4862 u32 staterr;
4863 u16 length;
047e0030 4864 u16 vlan_tag;
9d5c8243
AK		 4865
4866 i = rx_ring->next_to_clean;
69d3ca53 4867 buffer_info = &rx_ring->buffer_info[i];
9d5c8243
AK		 4868 rx_desc = E1000_RX_DESC_ADV(*rx_ring, i);
4869 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
4870
4871 while (staterr & E1000_RXD_STAT_DD) {
4872 if (*work_done >= budget)
4873 break;
4874 (*work_done)++;
9d5c8243 4875
69d3ca53
AD		 4876 skb = buffer_info->skb;
4877 prefetch(skb->data - NET_IP_ALIGN);
4878 buffer_info->skb = NULL;
4879
4880 i++;
4881 if (i == rx_ring->count)
4882 i = 0;
4883 next_rxd = E1000_RX_DESC_ADV(*rx_ring, i);
4884 prefetch(next_rxd);
4885 next_buffer = &rx_ring->buffer_info[i];
9d5c8243
AK		 4886
4887 length = le16_to_cpu(rx_desc->wb.upper.length);
4888 cleaned = true;
4889 cleaned_count++;
4890
2d94d8ab 4891 if (buffer_info->dma) {
bf36c1a0 4892 pci_unmap_single(pdev, buffer_info->dma,
4c844851 4893 rx_ring->rx_buffer_len,
bf36c1a0 4894 PCI_DMA_FROMDEVICE);
91615f76 4895 buffer_info->dma = 0;
4c844851 4896 if (rx_ring->rx_buffer_len >= IGB_RXBUFFER_1024) {
6ec43fe6
AD		 4897 skb_put(skb, length);
4898 goto send_up;
4899 }
4c844851 4900 skb_put(skb, igb_get_hlen(rx_ring, rx_desc));
bf36c1a0
AD		 4901 }
4902
4903 if (length) {
9d5c8243 4904 pci_unmap_page(pdev, buffer_info->page_dma,
bf36c1a0 4905 PAGE_SIZE / 2, PCI_DMA_FROMDEVICE);
9d5c8243 4906 buffer_info->page_dma = 0;
bf36c1a0
AD		 4907
4908 skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags++,
4909 buffer_info->page,
4910 buffer_info->page_offset,
4911 length);
4912
6ec43fe6 4913 if (page_count(buffer_info->page) != 1)
bf36c1a0
AD		 4914 buffer_info->page = NULL;
4915 else
4916 get_page(buffer_info->page);
9d5c8243
AK		 4917
4918 skb->len += length;
4919 skb->data_len += length;
9d5c8243 4920
bf36c1a0 4921 skb->truesize += length;
9d5c8243 4922 }
9d5c8243 4923
bf36c1a0 4924 if (!(staterr & E1000_RXD_STAT_EOP)) {
b2d56536
AD		 4925 buffer_info->skb = next_buffer->skb;
4926 buffer_info->dma = next_buffer->dma;
4927 next_buffer->skb = skb;
4928 next_buffer->dma = 0;
bf36c1a0
AD		 4929 goto next_desc;
4930 }
69d3ca53 4931send_up:
33af6bcc
PO		 4932 /*
4933 * If this bit is set, then the RX registers contain
4934 * the time stamp. No other packet will be time
4935 * stamped until we read these registers, so read the
4936 * registers to make them available again. Because
4937 * only one packet can be time stamped at a time, we
4938 * know that the register values must belong to this
4939 * one here and therefore we don't need to compare
4940 * any of the additional attributes stored for it.
4941 *
4942 * If nothing went wrong, then it should have a
4943 * skb_shared_tx that we can turn into a
4944 * skb_shared_hwtstamps.
4945 *
4946 * TODO: can time stamping be triggered (thus locking
4947 * the registers) without the packet reaching this point
4948 * here? In that case RX time stamping would get stuck.
4949 *
4950 * TODO: in "time stamp all packets" mode this bit is
4951 * not set. Need a global flag for this mode and then
4952 * always read the registers. Cannot be done without
4953 * a race condition.
4954 */
4955 if (unlikely(staterr & E1000_RXD_STAT_TS)) {
4956 u64 regval;
4957 u64 ns;
4958 struct skb_shared_hwtstamps *shhwtstamps =
4959 skb_hwtstamps(skb);
4960
4961 WARN(!(rd32(E1000_TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID),
4962 "igb: no RX time stamp available for time stamped packet");
4963 regval = rd32(E1000_RXSTMPL);
4964 regval |= (u64)rd32(E1000_RXSTMPH) << 32;
4965 ns = timecounter_cyc2time(&adapter->clock, regval);
4966 timecompare_update(&adapter->compare, ns);
4967 memset(shhwtstamps, 0, sizeof(*shhwtstamps));
4968 shhwtstamps->hwtstamp = ns_to_ktime(ns);
4969 shhwtstamps->syststamp =
4970 timecompare_transform(&adapter->compare, ns);
4971 }
4972
9d5c8243
AK		 4973 if (staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) {
4974 dev_kfree_skb_irq(skb);
4975 goto next_desc;
4976 }
9d5c8243
AK		 4977
4978 total_bytes += skb->len;
4979 total_packets++;
4980
04a5fcaa 4981 igb_rx_checksum_adv(rx_ring, adapter, staterr, skb);
9d5c8243
AK		 4982
4983 skb->protocol = eth_type_trans(skb, netdev);
047e0030
AD		 4984 skb_record_rx_queue(skb, rx_ring->queue_index);
4985
4986 vlan_tag = ((staterr & E1000_RXD_STAT_VP) ?
4987 le16_to_cpu(rx_desc->wb.upper.vlan) : 0);
9d5c8243 4988
047e0030 4989 igb_receive_skb(q_vector, skb, vlan_tag);
9d5c8243 4990
9d5c8243
AK		 4991next_desc:
4992 rx_desc->wb.upper.status_error = 0;
4993
4994 /* return some buffers to hardware, one at a time is too slow */
4995 if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
3b644cf6 4996 igb_alloc_rx_buffers_adv(rx_ring, cleaned_count);
9d5c8243
AK		 4997 cleaned_count = 0;
4998 }
4999
5000 /* use prefetched values */
5001 rx_desc = next_rxd;
5002 buffer_info = next_buffer;
9d5c8243
AK		 5003 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
5004 }
bf36c1a0 5005
9d5c8243 5006 rx_ring->next_to_clean = i;
c493ea45 5007 cleaned_count = igb_desc_unused(rx_ring);
9d5c8243
AK		 5008
5009 if (cleaned_count)
3b644cf6 5010 igb_alloc_rx_buffers_adv(rx_ring, cleaned_count);
9d5c8243
AK		 5011
5012 rx_ring->total_packets += total_packets;
5013 rx_ring->total_bytes += total_bytes;
5014 rx_ring->rx_stats.packets += total_packets;
5015 rx_ring->rx_stats.bytes += total_bytes;
8d24e933
AK		 5016 netdev->stats.rx_bytes += total_bytes;
5017 netdev->stats.rx_packets += total_packets;
9d5c8243
AK		 5018 return cleaned;
5019}
5020
9d5c8243
AK		 5021/**
5022 * igb_alloc_rx_buffers_adv - Replace used receive buffers; packet split
5023 * @adapter: address of board private structure
5024 **/
3b644cf6 5025static void igb_alloc_rx_buffers_adv(struct igb_ring *rx_ring,
9d5c8243
AK		 5026 int cleaned_count)
5027{
047e0030 5028 struct igb_adapter *adapter = rx_ring->q_vector->adapter;
9d5c8243 5029 struct net_device *netdev = adapter->netdev;
9d5c8243
AK		 5030 union e1000_adv_rx_desc *rx_desc;
5031 struct igb_buffer *buffer_info;
5032 struct sk_buff *skb;
5033 unsigned int i;
db761762 5034 int bufsz;
9d5c8243
AK		 5035
5036 i = rx_ring->next_to_use;
5037 buffer_info = &rx_ring->buffer_info[i];
5038
4c844851 5039 bufsz = rx_ring->rx_buffer_len;
db761762 5040
9d5c8243
AK		 5041 while (cleaned_count--) {
5042 rx_desc = E1000_RX_DESC_ADV(*rx_ring, i);
5043
6ec43fe6 5044 if ((bufsz < IGB_RXBUFFER_1024) && !buffer_info->page_dma) {
9d5c8243 5045 if (!buffer_info->page) {
bf36c1a0
AD		 5046 buffer_info->page = alloc_page(GFP_ATOMIC);
5047 if (!buffer_info->page) {
04a5fcaa 5048 rx_ring->rx_stats.alloc_failed++;
bf36c1a0
AD		 5049 goto no_buffers;
5050 }
5051 buffer_info->page_offset = 0;
5052 } else {
5053 buffer_info->page_offset ^= PAGE_SIZE / 2;
9d5c8243
AK		 5054 }
5055 buffer_info->page_dma =
80785298 5056 pci_map_page(rx_ring->pdev, buffer_info->page,
bf36c1a0
AD		 5057 buffer_info->page_offset,
5058 PAGE_SIZE / 2,
9d5c8243
AK		 5059 PCI_DMA_FROMDEVICE);
5060 }
5061
5062 if (!buffer_info->skb) {
89d71a66 5063 skb = netdev_alloc_skb_ip_align(netdev, bufsz);
9d5c8243 5064 if (!skb) {
04a5fcaa 5065 rx_ring->rx_stats.alloc_failed++;
9d5c8243
AK		 5066 goto no_buffers;
5067 }
5068
9d5c8243 5069 buffer_info->skb = skb;
80785298
AD		 5070 buffer_info->dma = pci_map_single(rx_ring->pdev,
5071 skb->data,
9d5c8243
AK		 5072 bufsz,
5073 PCI_DMA_FROMDEVICE);
9d5c8243
AK		 5074 }
5075 /* Refresh the desc even if buffer_addrs didn't change because
5076 * each write-back erases this info. */
6ec43fe6 5077 if (bufsz < IGB_RXBUFFER_1024) {
9d5c8243
AK		 5078 rx_desc->read.pkt_addr =
5079 cpu_to_le64(buffer_info->page_dma);
5080 rx_desc->read.hdr_addr = cpu_to_le64(buffer_info->dma);
5081 } else {
5082 rx_desc->read.pkt_addr =
5083 cpu_to_le64(buffer_info->dma);
5084 rx_desc->read.hdr_addr = 0;
5085 }
5086
5087 i++;
5088 if (i == rx_ring->count)
5089 i = 0;
5090 buffer_info = &rx_ring->buffer_info[i];
5091 }
5092
5093no_buffers:
5094 if (rx_ring->next_to_use != i) {
5095 rx_ring->next_to_use = i;
5096 if (i == 0)
5097 i = (rx_ring->count - 1);
5098 else
5099 i--;
5100
5101 /* Force memory writes to complete before letting h/w
5102 * know there are new descriptors to fetch. (Only
5103 * applicable for weak-ordered memory model archs,
5104 * such as IA-64). */
5105 wmb();
fce99e34 5106 writel(i, rx_ring->tail);
9d5c8243
AK		 5107 }
5108}
5109
5110/**
5111 * igb_mii_ioctl -
5112 * @netdev:
5113 * @ifreq:
5114 * @cmd:
5115 **/
5116static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
5117{
5118 struct igb_adapter *adapter = netdev_priv(netdev);
5119 struct mii_ioctl_data *data = if_mii(ifr);
5120
5121 if (adapter->hw.phy.media_type != e1000_media_type_copper)
5122 return -EOPNOTSUPP;
5123
5124 switch (cmd) {
5125 case SIOCGMIIPHY:
5126 data->phy_id = adapter->hw.phy.addr;
5127 break;
5128 case SIOCGMIIREG:
f5f4cf08
AD		 5129 if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
5130 &data->val_out))
9d5c8243
AK		 5131 return -EIO;
5132 break;
5133 case SIOCSMIIREG:
5134 default:
5135 return -EOPNOTSUPP;
5136 }
5137 return 0;
5138}
5139
c6cb090b
PO		 5140/**
5141 * igb_hwtstamp_ioctl - control hardware time stamping
5142 * @netdev:
5143 * @ifreq:
5144 * @cmd:
5145 *
33af6bcc
PO		 5146 * Outgoing time stamping can be enabled and disabled. Play nice and
5147 * disable it when requested, although it shouldn't case any overhead
5148 * when no packet needs it. At most one packet in the queue may be
5149 * marked for time stamping, otherwise it would be impossible to tell
5150 * for sure to which packet the hardware time stamp belongs.
5151 *
5152 * Incoming time stamping has to be configured via the hardware
5153 * filters. Not all combinations are supported, in particular event
5154 * type has to be specified. Matching the kind of event packet is
5155 * not supported, with the exception of "all V2 events regardless of
5156 * level 2 or 4".
5157 *
c6cb090b
PO		 5158 **/
5159static int igb_hwtstamp_ioctl(struct net_device *netdev,
5160 struct ifreq *ifr, int cmd)
5161{
33af6bcc
PO		 5162 struct igb_adapter *adapter = netdev_priv(netdev);
5163 struct e1000_hw *hw = &adapter->hw;
c6cb090b 5164 struct hwtstamp_config config;
33af6bcc
PO		 5165 u32 tsync_tx_ctl_bit = E1000_TSYNCTXCTL_ENABLED;
5166 u32 tsync_rx_ctl_bit = E1000_TSYNCRXCTL_ENABLED;
5167 u32 tsync_rx_ctl_type = 0;
5168 u32 tsync_rx_cfg = 0;
5169 int is_l4 = 0;
5170 int is_l2 = 0;
5171 short port = 319; /* PTP */
5172 u32 regval;
c6cb090b
PO		 5173
5174 if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
5175 return -EFAULT;
5176
5177 /* reserved for future extensions */
5178 if (config.flags)
5179 return -EINVAL;
5180
33af6bcc
PO		 5181 switch (config.tx_type) {
5182 case HWTSTAMP_TX_OFF:
5183 tsync_tx_ctl_bit = 0;
5184 break;
5185 case HWTSTAMP_TX_ON:
5186 tsync_tx_ctl_bit = E1000_TSYNCTXCTL_ENABLED;
5187 break;
5188 default:
5189 return -ERANGE;
5190 }
5191
5192 switch (config.rx_filter) {
5193 case HWTSTAMP_FILTER_NONE:
5194 tsync_rx_ctl_bit = 0;
5195 break;
5196 case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
5197 case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
5198 case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
5199 case HWTSTAMP_FILTER_ALL:
5200 /*
5201 * register TSYNCRXCFG must be set, therefore it is not
5202 * possible to time stamp both Sync and Delay_Req messages
5203 * => fall back to time stamping all packets
5204 */
5205 tsync_rx_ctl_type = E1000_TSYNCRXCTL_TYPE_ALL;
5206 config.rx_filter = HWTSTAMP_FILTER_ALL;
5207 break;
5208 case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
5209 tsync_rx_ctl_type = E1000_TSYNCRXCTL_TYPE_L4_V1;
5210 tsync_rx_cfg = E1000_TSYNCRXCFG_PTP_V1_SYNC_MESSAGE;
5211 is_l4 = 1;
5212 break;
5213 case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
5214 tsync_rx_ctl_type = E1000_TSYNCRXCTL_TYPE_L4_V1;
5215 tsync_rx_cfg = E1000_TSYNCRXCFG_PTP_V1_DELAY_REQ_MESSAGE;
5216 is_l4 = 1;
5217 break;
5218 case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
5219 case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
5220 tsync_rx_ctl_type = E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
5221 tsync_rx_cfg = E1000_TSYNCRXCFG_PTP_V2_SYNC_MESSAGE;
5222 is_l2 = 1;
5223 is_l4 = 1;
5224 config.rx_filter = HWTSTAMP_FILTER_SOME;
5225 break;
5226 case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
5227 case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
5228 tsync_rx_ctl_type = E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
5229 tsync_rx_cfg = E1000_TSYNCRXCFG_PTP_V2_DELAY_REQ_MESSAGE;
5230 is_l2 = 1;
5231 is_l4 = 1;
5232 config.rx_filter = HWTSTAMP_FILTER_SOME;
5233 break;
5234 case HWTSTAMP_FILTER_PTP_V2_EVENT:
5235 case HWTSTAMP_FILTER_PTP_V2_SYNC:
5236 case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
5237 tsync_rx_ctl_type = E1000_TSYNCRXCTL_TYPE_EVENT_V2;
5238 config.rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
5239 is_l2 = 1;
5240 break;
5241 default:
5242 return -ERANGE;
5243 }
5244
5245 /* enable/disable TX */
5246 regval = rd32(E1000_TSYNCTXCTL);
5247 regval = (regval & ~E1000_TSYNCTXCTL_ENABLED) | tsync_tx_ctl_bit;
5248 wr32(E1000_TSYNCTXCTL, regval);
5249
5250 /* enable/disable RX, define which PTP packets are time stamped */
5251 regval = rd32(E1000_TSYNCRXCTL);
5252 regval = (regval & ~E1000_TSYNCRXCTL_ENABLED) | tsync_rx_ctl_bit;
5253 regval = (regval & ~0xE) | tsync_rx_ctl_type;
5254 wr32(E1000_TSYNCRXCTL, regval);
5255 wr32(E1000_TSYNCRXCFG, tsync_rx_cfg);
5256
5257 /*
5258 * Ethertype Filter Queue Filter[0][15:0] = 0x88F7
5259 * (Ethertype to filter on)
5260 * Ethertype Filter Queue Filter[0][26] = 0x1 (Enable filter)
5261 * Ethertype Filter Queue Filter[0][30] = 0x1 (Enable Timestamping)
5262 */
5263 wr32(E1000_ETQF0, is_l2 ? 0x440088f7 : 0);
5264
5265 /* L4 Queue Filter[0]: only filter by source and destination port */
5266 wr32(E1000_SPQF0, htons(port));
5267 wr32(E1000_IMIREXT(0), is_l4 ?
5268 ((1<<12) | (1<<19) /* bypass size and control flags */) : 0);
5269 wr32(E1000_IMIR(0), is_l4 ?
5270 (htons(port)
5271 | (0<<16) /* immediate interrupt disabled */
5272 | 0 /* (1<<17) bit cleared: do not bypass
5273 destination port check */)
5274 : 0);
5275 wr32(E1000_FTQF0, is_l4 ?
5276 (0x11 /* UDP */
5277 | (1<<15) /* VF not compared */
5278 | (1<<27) /* Enable Timestamping */
5279 | (7<<28) /* only source port filter enabled,
5280 source/target address and protocol
5281 masked */)
5282 : ((1<<15) | (15<<28) /* all mask bits set = filter not
5283 enabled */));
5284
5285 wrfl();
5286
5287 adapter->hwtstamp_config = config;
5288
5289 /* clear TX/RX time stamp registers, just to be sure */
5290 regval = rd32(E1000_TXSTMPH);
5291 regval = rd32(E1000_RXSTMPH);
c6cb090b 5292
33af6bcc
PO		 5293 return copy_to_user(ifr->ifr_data, &config, sizeof(config)) ?
5294 -EFAULT : 0;
c6cb090b
PO		 5295}
5296
9d5c8243
AK		 5297/**
5298 * igb_ioctl -
5299 * @netdev:
5300 * @ifreq:
5301 * @cmd:
5302 **/
5303static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
5304{
5305 switch (cmd) {
5306 case SIOCGMIIPHY:
5307 case SIOCGMIIREG:
5308 case SIOCSMIIREG:
5309 return igb_mii_ioctl(netdev, ifr, cmd);
c6cb090b
PO		 5310 case SIOCSHWTSTAMP:
5311 return igb_hwtstamp_ioctl(netdev, ifr, cmd);
9d5c8243
AK		 5312 default:
5313 return -EOPNOTSUPP;
5314 }
5315}
5316
009bc06e
AD		 5317s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
5318{
5319 struct igb_adapter *adapter = hw->back;
5320 u16 cap_offset;
5321
5322 cap_offset = pci_find_capability(adapter->pdev, PCI_CAP_ID_EXP);
5323 if (!cap_offset)
5324 return -E1000_ERR_CONFIG;
5325
5326 pci_read_config_word(adapter->pdev, cap_offset + reg, value);
5327
5328 return 0;
5329}
5330
5331s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
5332{
5333 struct igb_adapter *adapter = hw->back;
5334 u16 cap_offset;
5335
5336 cap_offset = pci_find_capability(adapter->pdev, PCI_CAP_ID_EXP);
5337 if (!cap_offset)
5338 return -E1000_ERR_CONFIG;
5339
5340 pci_write_config_word(adapter->pdev, cap_offset + reg, *value);
5341
5342 return 0;
5343}
5344
9d5c8243
AK		 5345static void igb_vlan_rx_register(struct net_device *netdev,
5346 struct vlan_group *grp)
5347{
5348 struct igb_adapter *adapter = netdev_priv(netdev);
5349 struct e1000_hw *hw = &adapter->hw;
5350 u32 ctrl, rctl;
5351
5352 igb_irq_disable(adapter);
5353 adapter->vlgrp = grp;
5354
5355 if (grp) {
5356 /* enable VLAN tag insert/strip */
5357 ctrl = rd32(E1000_CTRL);
5358 ctrl |= E1000_CTRL_VME;
5359 wr32(E1000_CTRL, ctrl);
5360
5361 /* enable VLAN receive filtering */
5362 rctl = rd32(E1000_RCTL);
9d5c8243
AK		 5363 rctl &= ~E1000_RCTL_CFIEN;
5364 wr32(E1000_RCTL, rctl);
5365 igb_update_mng_vlan(adapter);
9d5c8243
AK		 5366 } else {
5367 /* disable VLAN tag insert/strip */
5368 ctrl = rd32(E1000_CTRL);
5369 ctrl &= ~E1000_CTRL_VME;
5370 wr32(E1000_CTRL, ctrl);
5371
9d5c8243
AK		 5372 if (adapter->mng_vlan_id != (u16)IGB_MNG_VLAN_NONE) {
5373 igb_vlan_rx_kill_vid(netdev, adapter->mng_vlan_id);
5374 adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
5375 }
9d5c8243
AK		 5376 }
5377
e1739522
AD		 5378 igb_rlpml_set(adapter);
5379
9d5c8243
AK		 5380 if (!test_bit(__IGB_DOWN, &adapter->state))
5381 igb_irq_enable(adapter);
5382}
5383
5384static void igb_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
5385{
5386 struct igb_adapter *adapter = netdev_priv(netdev);
5387 struct e1000_hw *hw = &adapter->hw;
4ae196df 5388 int pf_id = adapter->vfs_allocated_count;
9d5c8243 5389
28b0759c 5390 if ((hw->mng_cookie.status &
9d5c8243
AK		 5391 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
5392 (vid == adapter->mng_vlan_id))
5393 return;
4ae196df
AD		 5394
5395 /* add vid to vlvf if sr-iov is enabled,
5396 * if that fails add directly to filter table */
5397 if (igb_vlvf_set(adapter, vid, true, pf_id))
5398 igb_vfta_set(hw, vid, true);
5399
9d5c8243
AK		 5400}
5401
5402static void igb_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
5403{
5404 struct igb_adapter *adapter = netdev_priv(netdev);
5405 struct e1000_hw *hw = &adapter->hw;
4ae196df 5406 int pf_id = adapter->vfs_allocated_count;
9d5c8243
AK		 5407
5408 igb_irq_disable(adapter);
5409 vlan_group_set_device(adapter->vlgrp, vid, NULL);
5410
5411 if (!test_bit(__IGB_DOWN, &adapter->state))
5412 igb_irq_enable(adapter);
5413
5414 if ((adapter->hw.mng_cookie.status &
5415 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
5416 (vid == adapter->mng_vlan_id)) {
5417 /* release control to f/w */
5418 igb_release_hw_control(adapter);
5419 return;
5420 }
5421
4ae196df
AD		 5422 /* remove vid from vlvf if sr-iov is enabled,
5423 * if not in vlvf remove from vfta */
5424 if (igb_vlvf_set(adapter, vid, false, pf_id))
5425 igb_vfta_set(hw, vid, false);
9d5c8243
AK		 5426}
5427
5428static void igb_restore_vlan(struct igb_adapter *adapter)
5429{
5430 igb_vlan_rx_register(adapter->netdev, adapter->vlgrp);
5431
5432 if (adapter->vlgrp) {
5433 u16 vid;
5434 for (vid = 0; vid < VLAN_GROUP_ARRAY_LEN; vid++) {
5435 if (!vlan_group_get_device(adapter->vlgrp, vid))
5436 continue;
5437 igb_vlan_rx_add_vid(adapter->netdev, vid);
5438 }
5439 }
5440}
5441
5442int igb_set_spd_dplx(struct igb_adapter *adapter, u16 spddplx)
5443{
5444 struct e1000_mac_info *mac = &adapter->hw.mac;
5445
5446 mac->autoneg = 0;
5447
9d5c8243
AK		 5448 switch (spddplx) {
5449 case SPEED_10 + DUPLEX_HALF:
5450 mac->forced_speed_duplex = ADVERTISE_10_HALF;
5451 break;
5452 case SPEED_10 + DUPLEX_FULL:
5453 mac->forced_speed_duplex = ADVERTISE_10_FULL;
5454 break;
5455 case SPEED_100 + DUPLEX_HALF:
5456 mac->forced_speed_duplex = ADVERTISE_100_HALF;
5457 break;
5458 case SPEED_100 + DUPLEX_FULL:
5459 mac->forced_speed_duplex = ADVERTISE_100_FULL;
5460 break;
5461 case SPEED_1000 + DUPLEX_FULL:
5462 mac->autoneg = 1;
5463 adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
5464 break;
5465 case SPEED_1000 + DUPLEX_HALF: /* not supported */
5466 default:
5467 dev_err(&adapter->pdev->dev,
5468 "Unsupported Speed/Duplex configuration\n");
5469 return -EINVAL;
5470 }
5471 return 0;
5472}
5473
3fe7c4c9 5474static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake)
9d5c8243
AK		 5475{
5476 struct net_device *netdev = pci_get_drvdata(pdev);
5477 struct igb_adapter *adapter = netdev_priv(netdev);
5478 struct e1000_hw *hw = &adapter->hw;
2d064c06 5479 u32 ctrl, rctl, status;
9d5c8243
AK		 5480 u32 wufc = adapter->wol;
5481#ifdef CONFIG_PM
5482 int retval = 0;
5483#endif
5484
5485 netif_device_detach(netdev);
5486
a88f10ec
AD		 5487 if (netif_running(netdev))
5488 igb_close(netdev);
5489
047e0030 5490 igb_clear_interrupt_scheme(adapter);
9d5c8243
AK		 5491
5492#ifdef CONFIG_PM
5493 retval = pci_save_state(pdev);
5494 if (retval)
5495 return retval;
5496#endif
5497
5498 status = rd32(E1000_STATUS);
5499 if (status & E1000_STATUS_LU)
5500 wufc &= ~E1000_WUFC_LNKC;
5501
5502 if (wufc) {
5503 igb_setup_rctl(adapter);
ff41f8dc 5504 igb_set_rx_mode(netdev);
9d5c8243
AK		 5505
5506 /* turn on all-multi mode if wake on multicast is enabled */
5507 if (wufc & E1000_WUFC_MC) {
5508 rctl = rd32(E1000_RCTL);
5509 rctl |= E1000_RCTL_MPE;
5510 wr32(E1000_RCTL, rctl);
5511 }
5512
5513 ctrl = rd32(E1000_CTRL);
5514 /* advertise wake from D3Cold */
5515 #define E1000_CTRL_ADVD3WUC 0x00100000
5516 /* phy power management enable */
5517 #define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
5518 ctrl |= E1000_CTRL_ADVD3WUC;
5519 wr32(E1000_CTRL, ctrl);
5520
9d5c8243
AK		 5521 /* Allow time for pending master requests to run */
5522 igb_disable_pcie_master(&adapter->hw);
5523
5524 wr32(E1000_WUC, E1000_WUC_PME_EN);
5525 wr32(E1000_WUFC, wufc);
9d5c8243
AK		 5526 } else {
5527 wr32(E1000_WUC, 0);
5528 wr32(E1000_WUFC, 0);
9d5c8243
AK		 5529 }
5530
3fe7c4c9
RW		 5531 *enable_wake = wufc || adapter->en_mng_pt;
5532 if (!*enable_wake)
2fb02a26 5533 igb_shutdown_serdes_link_82575(hw);
9d5c8243
AK		 5534
5535 /* Release control of h/w to f/w. If f/w is AMT enabled, this
5536 * would have already happened in close and is redundant. */
5537 igb_release_hw_control(adapter);
5538
5539 pci_disable_device(pdev);
5540
9d5c8243
AK		 5541 return 0;
5542}
5543
5544#ifdef CONFIG_PM
3fe7c4c9
RW		 5545static int igb_suspend(struct pci_dev *pdev, pm_message_t state)
5546{
5547 int retval;
5548 bool wake;
5549
5550 retval = __igb_shutdown(pdev, &wake);
5551 if (retval)
5552 return retval;
5553
5554 if (wake) {
5555 pci_prepare_to_sleep(pdev);
5556 } else {
5557 pci_wake_from_d3(pdev, false);
5558 pci_set_power_state(pdev, PCI_D3hot);
5559 }
5560
5561 return 0;
5562}
5563
9d5c8243
AK		 5564static int igb_resume(struct pci_dev *pdev)
5565{
5566 struct net_device *netdev = pci_get_drvdata(pdev);
5567 struct igb_adapter *adapter = netdev_priv(netdev);
5568 struct e1000_hw *hw = &adapter->hw;
5569 u32 err;
5570
5571 pci_set_power_state(pdev, PCI_D0);
5572 pci_restore_state(pdev);
42bfd33a 5573
aed5dec3 5574 err = pci_enable_device_mem(pdev);
9d5c8243
AK		 5575 if (err) {
5576 dev_err(&pdev->dev,
5577 "igb: Cannot enable PCI device from suspend\n");
5578 return err;
5579 }
5580 pci_set_master(pdev);
5581
5582 pci_enable_wake(pdev, PCI_D3hot, 0);
5583 pci_enable_wake(pdev, PCI_D3cold, 0);
5584
047e0030 5585 if (igb_init_interrupt_scheme(adapter)) {
a88f10ec
AD		 5586 dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
5587 return -ENOMEM;
9d5c8243
AK		 5588 }
5589
5590 /* e1000_power_up_phy(adapter); */
5591
5592 igb_reset(adapter);
a8564f03
AD		 5593
5594 /* let the f/w know that the h/w is now under the control of the
5595 * driver. */
5596 igb_get_hw_control(adapter);
5597
9d5c8243
AK		 5598 wr32(E1000_WUS, ~0);
5599
a88f10ec
AD		 5600 if (netif_running(netdev)) {
5601 err = igb_open(netdev);
5602 if (err)
5603 return err;
5604 }
9d5c8243
AK		 5605
5606 netif_device_attach(netdev);
5607
9d5c8243
AK		 5608 return 0;
5609}
5610#endif
5611
5612static void igb_shutdown(struct pci_dev *pdev)
5613{
3fe7c4c9
RW		 5614 bool wake;
5615
5616 __igb_shutdown(pdev, &wake);
5617
5618 if (system_state == SYSTEM_POWER_OFF) {
5619 pci_wake_from_d3(pdev, wake);
5620 pci_set_power_state(pdev, PCI_D3hot);
5621 }
9d5c8243
AK		 5622}
5623
5624#ifdef CONFIG_NET_POLL_CONTROLLER
5625/*
5626 * Polling 'interrupt' - used by things like netconsole to send skbs
5627 * without having to re-enable interrupts. It's not called while
5628 * the interrupt routine is executing.
5629 */
5630static void igb_netpoll(struct net_device *netdev)
5631{
5632 struct igb_adapter *adapter = netdev_priv(netdev);
eebbbdba 5633 struct e1000_hw *hw = &adapter->hw;
9d5c8243 5634 int i;
9d5c8243 5635
eebbbdba 5636 if (!adapter->msix_entries) {
047e0030 5637 struct igb_q_vector *q_vector = adapter->q_vector[0];
eebbbdba 5638 igb_irq_disable(adapter);
047e0030 5639 napi_schedule(&q_vector->napi);
eebbbdba
AD		 5640 return;
5641 }
9d5c8243 5642
047e0030
AD		 5643 for (i = 0; i < adapter->num_q_vectors; i++) {
5644 struct igb_q_vector *q_vector = adapter->q_vector[i];
5645 wr32(E1000_EIMC, q_vector->eims_value);
5646 napi_schedule(&q_vector->napi);
eebbbdba 5647 }
9d5c8243
AK		 5648}
5649#endif /* CONFIG_NET_POLL_CONTROLLER */
5650
5651/**
5652 * igb_io_error_detected - called when PCI error is detected
5653 * @pdev: Pointer to PCI device
5654 * @state: The current pci connection state
5655 *
5656 * This function is called after a PCI bus error affecting
5657 * this device has been detected.
5658 */
5659static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
5660 pci_channel_state_t state)
5661{
5662 struct net_device *netdev = pci_get_drvdata(pdev);
5663 struct igb_adapter *adapter = netdev_priv(netdev);
5664
5665 netif_device_detach(netdev);
5666
59ed6eec
AD		 5667 if (state == pci_channel_io_perm_failure)
5668 return PCI_ERS_RESULT_DISCONNECT;
5669
9d5c8243
AK		 5670 if (netif_running(netdev))
5671 igb_down(adapter);
5672 pci_disable_device(pdev);
5673
5674 /* Request a slot slot reset. */
5675 return PCI_ERS_RESULT_NEED_RESET;
5676}
5677
5678/**
5679 * igb_io_slot_reset - called after the pci bus has been reset.
5680 * @pdev: Pointer to PCI device
5681 *
5682 * Restart the card from scratch, as if from a cold-boot. Implementation
5683 * resembles the first-half of the igb_resume routine.
5684 */
5685static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
5686{
5687 struct net_device *netdev = pci_get_drvdata(pdev);
5688 struct igb_adapter *adapter = netdev_priv(netdev);
5689 struct e1000_hw *hw = &adapter->hw;
40a914fa 5690 pci_ers_result_t result;
42bfd33a 5691 int err;
9d5c8243 5692
aed5dec3 5693 if (pci_enable_device_mem(pdev)) {
9d5c8243
AK		 5694 dev_err(&pdev->dev,
5695 "Cannot re-enable PCI device after reset.\n");
40a914fa
AD		 5696 result = PCI_ERS_RESULT_DISCONNECT;
5697 } else {
5698 pci_set_master(pdev);
5699 pci_restore_state(pdev);
9d5c8243 5700
40a914fa
AD		 5701 pci_enable_wake(pdev, PCI_D3hot, 0);
5702 pci_enable_wake(pdev, PCI_D3cold, 0);
9d5c8243 5703
40a914fa
AD		 5704 igb_reset(adapter);
5705 wr32(E1000_WUS, ~0);
5706 result = PCI_ERS_RESULT_RECOVERED;
5707 }
9d5c8243 5708
ea943d41
JK		 5709 err = pci_cleanup_aer_uncorrect_error_status(pdev);
5710 if (err) {
5711 dev_err(&pdev->dev, "pci_cleanup_aer_uncorrect_error_status "
5712 "failed 0x%0x\n", err);
5713 /* non-fatal, continue */
5714 }
40a914fa
AD		 5715
5716 return result;
9d5c8243
AK		 5717}
5718
5719/**
5720 * igb_io_resume - called when traffic can start flowing again.
5721 * @pdev: Pointer to PCI device
5722 *
5723 * This callback is called when the error recovery driver tells us that
5724 * its OK to resume normal operation. Implementation resembles the
5725 * second-half of the igb_resume routine.
5726 */
5727static void igb_io_resume(struct pci_dev *pdev)
5728{
5729 struct net_device *netdev = pci_get_drvdata(pdev);
5730 struct igb_adapter *adapter = netdev_priv(netdev);
5731
9d5c8243
AK		 5732 if (netif_running(netdev)) {
5733 if (igb_up(adapter)) {
5734 dev_err(&pdev->dev, "igb_up failed after reset\n");
5735 return;
5736 }
5737 }
5738
5739 netif_device_attach(netdev);
5740
5741 /* let the f/w know that the h/w is now under the control of the
5742 * driver. */
5743 igb_get_hw_control(adapter);
9d5c8243
AK		 5744}
5745
26ad9178
AD		 5746static void igb_rar_set_qsel(struct igb_adapter *adapter, u8 *addr, u32 index,
5747 u8 qsel)
5748{
5749 u32 rar_low, rar_high;
5750 struct e1000_hw *hw = &adapter->hw;
5751
5752 /* HW expects these in little endian so we reverse the byte order
5753 * from network order (big endian) to little endian
5754 */
5755 rar_low = ((u32) addr[0] | ((u32) addr[1] << 8) |
5756 ((u32) addr[2] << 16) | ((u32) addr[3] << 24));
5757 rar_high = ((u32) addr[4] | ((u32) addr[5] << 8));
5758
5759 /* Indicate to hardware the Address is Valid. */
5760 rar_high |= E1000_RAH_AV;
5761
5762 if (hw->mac.type == e1000_82575)
5763 rar_high |= E1000_RAH_POOL_1 * qsel;
5764 else
5765 rar_high |= E1000_RAH_POOL_1 << qsel;
5766
5767 wr32(E1000_RAL(index), rar_low);
5768 wrfl();
5769 wr32(E1000_RAH(index), rar_high);
5770 wrfl();
5771}
5772
4ae196df
AD		 5773static int igb_set_vf_mac(struct igb_adapter *adapter,
5774 int vf, unsigned char *mac_addr)
5775{
5776 struct e1000_hw *hw = &adapter->hw;
ff41f8dc
AD		 5777 /* VF MAC addresses start at end of receive addresses and moves
5778 * torwards the first, as a result a collision should not be possible */
5779 int rar_entry = hw->mac.rar_entry_count - (vf + 1);
4ae196df 5780
37680117 5781 memcpy(adapter->vf_data[vf].vf_mac_addresses, mac_addr, ETH_ALEN);
4ae196df 5782
26ad9178 5783 igb_rar_set_qsel(adapter, mac_addr, rar_entry, vf);
4ae196df
AD		 5784
5785 return 0;
5786}
5787
5788static void igb_vmm_control(struct igb_adapter *adapter)
5789{
5790 struct e1000_hw *hw = &adapter->hw;
5791 u32 reg_data;
5792
5793 if (!adapter->vfs_allocated_count)
5794 return;
5795
5796 /* VF's need PF reset indication before they
5797 * can send/receive mail */
5798 reg_data = rd32(E1000_CTRL_EXT);
5799 reg_data |= E1000_CTRL_EXT_PFRSTD;
5800 wr32(E1000_CTRL_EXT, reg_data);
5801
5802 igb_vmdq_set_loopback_pf(hw, true);
5803 igb_vmdq_set_replication_pf(hw, true);
5804}
5805
9d5c8243 5806/* igb_main.c */
Clone of davem's net-next-2.6 tree