aoe: remove private mac address format function
[linux-2.6-block.git] / drivers / net / ixgbe / ixgbe_main.c
CommitLineData
9a799d71
AK
1/*******************************************************************************
2
3 Intel 10 Gigabit PCI Express Linux driver
b4617240 4 Copyright(c) 1999 - 2008 Intel Corporation.
9a799d71
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:
9a799d71
AK
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/types.h>
29#include <linux/module.h>
30#include <linux/pci.h>
31#include <linux/netdevice.h>
32#include <linux/vmalloc.h>
33#include <linux/string.h>
34#include <linux/in.h>
35#include <linux/ip.h>
36#include <linux/tcp.h>
37#include <linux/ipv6.h>
38#include <net/checksum.h>
39#include <net/ip6_checksum.h>
40#include <linux/ethtool.h>
41#include <linux/if_vlan.h>
42
43#include "ixgbe.h"
44#include "ixgbe_common.h"
45
46char ixgbe_driver_name[] = "ixgbe";
9c8eb720 47static const char ixgbe_driver_string[] =
b4617240 48 "Intel(R) 10 Gigabit PCI Express Network Driver";
9a799d71 49
51ac6445 50#define DRV_VERSION "1.3.30-k2"
9c8eb720 51const char ixgbe_driver_version[] = DRV_VERSION;
b4617240 52static char ixgbe_copyright[] = "Copyright (c) 1999-2007 Intel Corporation.";
9a799d71
AK
53
54static const struct ixgbe_info *ixgbe_info_tbl[] = {
b4617240 55 [board_82598] = &ixgbe_82598_info,
9a799d71
AK
56};
57
58/* ixgbe_pci_tbl - PCI Device ID Table
59 *
60 * Wildcard entries (PCI_ANY_ID) should come last
61 * Last entry must be all 0s
62 *
63 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
64 * Class, Class Mask, private data (not used) }
65 */
66static struct pci_device_id ixgbe_pci_tbl[] = {
67 {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT),
3957d63d 68 board_82598 },
9a799d71 69 {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT),
3957d63d 70 board_82598 },
0befdb3e
JB
71 {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT),
72 board_82598 },
9a799d71 73 {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4),
3957d63d 74 board_82598 },
8d792cd9
JB
75 {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT),
76 board_82598 },
c4900be0
DS
77 {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT),
78 board_82598 },
79 {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM),
80 board_82598 },
b95f5fcb
JB
81 {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR),
82 board_82598 },
c4900be0
DS
83 {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM),
84 board_82598 },
9a799d71
AK
85
86 /* required last entry */
87 {0, }
88};
89MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
90
5dd2d332 91#ifdef CONFIG_IXGBE_DCA
bd0362dd 92static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
b4617240 93 void *p);
bd0362dd
JC
94static struct notifier_block dca_notifier = {
95 .notifier_call = ixgbe_notify_dca,
96 .next = NULL,
97 .priority = 0
98};
99#endif
100
9a799d71
AK
101MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
102MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
103MODULE_LICENSE("GPL");
104MODULE_VERSION(DRV_VERSION);
105
106#define DEFAULT_DEBUG_LEVEL_SHIFT 3
107
5eba3699
AV
108static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
109{
110 u32 ctrl_ext;
111
112 /* Let firmware take over control of h/w */
113 ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
114 IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
b4617240 115 ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
5eba3699
AV
116}
117
118static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
119{
120 u32 ctrl_ext;
121
122 /* Let firmware know the driver has taken over */
123 ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
124 IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
b4617240 125 ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
5eba3699 126}
9a799d71 127
9a799d71 128static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, u16 int_alloc_entry,
b4617240 129 u8 msix_vector)
9a799d71
AK
130{
131 u32 ivar, index;
132
133 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
134 index = (int_alloc_entry >> 2) & 0x1F;
135 ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR(index));
136 ivar &= ~(0xFF << (8 * (int_alloc_entry & 0x3)));
137 ivar |= (msix_vector << (8 * (int_alloc_entry & 0x3)));
138 IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR(index), ivar);
139}
140
141static void ixgbe_unmap_and_free_tx_resource(struct ixgbe_adapter *adapter,
b4617240
PW
142 struct ixgbe_tx_buffer
143 *tx_buffer_info)
9a799d71
AK
144{
145 if (tx_buffer_info->dma) {
e01c31a5 146 pci_unmap_page(adapter->pdev, tx_buffer_info->dma,
b4617240 147 tx_buffer_info->length, PCI_DMA_TODEVICE);
9a799d71
AK
148 tx_buffer_info->dma = 0;
149 }
150 if (tx_buffer_info->skb) {
151 dev_kfree_skb_any(tx_buffer_info->skb);
152 tx_buffer_info->skb = NULL;
153 }
154 /* tx_buffer_info must be completely set up in the transmit path */
155}
156
157static inline bool ixgbe_check_tx_hang(struct ixgbe_adapter *adapter,
b4617240
PW
158 struct ixgbe_ring *tx_ring,
159 unsigned int eop)
9a799d71 160{
e01c31a5
JB
161 struct ixgbe_hw *hw = &adapter->hw;
162 u32 head, tail;
163
9a799d71 164 /* Detect a transmit hang in hardware, this serializes the
e01c31a5
JB
165 * check with the clearing of time_stamp and movement of eop */
166 head = IXGBE_READ_REG(hw, tx_ring->head);
167 tail = IXGBE_READ_REG(hw, tx_ring->tail);
9a799d71 168 adapter->detect_tx_hung = false;
e01c31a5
JB
169 if ((head != tail) &&
170 tx_ring->tx_buffer_info[eop].time_stamp &&
9a799d71
AK
171 time_after(jiffies, tx_ring->tx_buffer_info[eop].time_stamp + HZ) &&
172 !(IXGBE_READ_REG(&adapter->hw, IXGBE_TFCS) & IXGBE_TFCS_TXOFF)) {
173 /* detected Tx unit hang */
e01c31a5
JB
174 union ixgbe_adv_tx_desc *tx_desc;
175 tx_desc = IXGBE_TX_DESC_ADV(*tx_ring, eop);
9a799d71 176 DPRINTK(DRV, ERR, "Detected Tx Unit Hang\n"
e01c31a5
JB
177 " Tx Queue <%d>\n"
178 " TDH, TDT <%x>, <%x>\n"
9a799d71
AK
179 " next_to_use <%x>\n"
180 " next_to_clean <%x>\n"
181 "tx_buffer_info[next_to_clean]\n"
182 " time_stamp <%lx>\n"
e01c31a5
JB
183 " jiffies <%lx>\n",
184 tx_ring->queue_index,
185 head, tail,
186 tx_ring->next_to_use, eop,
187 tx_ring->tx_buffer_info[eop].time_stamp, jiffies);
9a799d71
AK
188 return true;
189 }
190
191 return false;
192}
193
b4617240
PW
194#define IXGBE_MAX_TXD_PWR 14
195#define IXGBE_MAX_DATA_PER_TXD (1 << IXGBE_MAX_TXD_PWR)
e092be60
AV
196
197/* Tx Descriptors needed, worst case */
198#define TXD_USE_COUNT(S) (((S) >> IXGBE_MAX_TXD_PWR) + \
199 (((S) & (IXGBE_MAX_DATA_PER_TXD - 1)) ? 1 : 0))
200#define DESC_NEEDED (TXD_USE_COUNT(IXGBE_MAX_DATA_PER_TXD) /* skb->data */ + \
b4617240 201 MAX_SKB_FRAGS * TXD_USE_COUNT(PAGE_SIZE) + 1) /* for context */
e092be60 202
e01c31a5
JB
203#define GET_TX_HEAD_FROM_RING(ring) (\
204 *(volatile u32 *) \
205 ((union ixgbe_adv_tx_desc *)(ring)->desc + (ring)->count))
206static void ixgbe_tx_timeout(struct net_device *netdev);
207
9a799d71
AK
208/**
209 * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
210 * @adapter: board private structure
e01c31a5 211 * @tx_ring: tx ring to clean
9a799d71
AK
212 **/
213static bool ixgbe_clean_tx_irq(struct ixgbe_adapter *adapter,
e01c31a5 214 struct ixgbe_ring *tx_ring)
9a799d71 215{
e01c31a5 216 union ixgbe_adv_tx_desc *tx_desc;
9a799d71 217 struct ixgbe_tx_buffer *tx_buffer_info;
e01c31a5
JB
218 struct net_device *netdev = adapter->netdev;
219 struct sk_buff *skb;
220 unsigned int i;
221 u32 head, oldhead;
222 unsigned int count = 0;
223 unsigned int total_bytes = 0, total_packets = 0;
9a799d71 224
e01c31a5
JB
225 rmb();
226 head = GET_TX_HEAD_FROM_RING(tx_ring);
227 head = le32_to_cpu(head);
9a799d71 228 i = tx_ring->next_to_clean;
e01c31a5
JB
229 while (1) {
230 while (i != head) {
9a799d71
AK
231 tx_desc = IXGBE_TX_DESC_ADV(*tx_ring, i);
232 tx_buffer_info = &tx_ring->tx_buffer_info[i];
e01c31a5 233 skb = tx_buffer_info->skb;
9a799d71 234
e01c31a5 235 if (skb) {
e092be60 236 unsigned int segs, bytecount;
e01c31a5
JB
237
238 /* gso_segs is currently only valid for tcp */
e092be60
AV
239 segs = skb_shinfo(skb)->gso_segs ?: 1;
240 /* multiply data chunks by size of headers */
241 bytecount = ((segs - 1) * skb_headlen(skb)) +
e01c31a5
JB
242 skb->len;
243 total_packets += segs;
244 total_bytes += bytecount;
e092be60 245 }
e01c31a5 246
9a799d71 247 ixgbe_unmap_and_free_tx_resource(adapter,
e01c31a5 248 tx_buffer_info);
9a799d71
AK
249
250 i++;
251 if (i == tx_ring->count)
252 i = 0;
9a799d71 253
e01c31a5
JB
254 count++;
255 if (count == tx_ring->count)
256 goto done_cleaning;
257 }
258 oldhead = head;
259 rmb();
260 head = GET_TX_HEAD_FROM_RING(tx_ring);
261 head = le32_to_cpu(head);
262 if (head == oldhead)
263 goto done_cleaning;
264 } /* while (1) */
265
266done_cleaning:
9a799d71
AK
267 tx_ring->next_to_clean = i;
268
e092be60 269#define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
e01c31a5
JB
270 if (unlikely(count && netif_carrier_ok(netdev) &&
271 (IXGBE_DESC_UNUSED(tx_ring) >= TX_WAKE_THRESHOLD))) {
e092be60
AV
272 /* Make sure that anybody stopping the queue after this
273 * sees the new next_to_clean.
274 */
275 smp_mb();
30eba97a
AV
276 if (__netif_subqueue_stopped(netdev, tx_ring->queue_index) &&
277 !test_bit(__IXGBE_DOWN, &adapter->state)) {
278 netif_wake_subqueue(netdev, tx_ring->queue_index);
e01c31a5 279 ++adapter->restart_queue;
30eba97a 280 }
e092be60 281 }
9a799d71 282
e01c31a5
JB
283 if (adapter->detect_tx_hung) {
284 if (ixgbe_check_tx_hang(adapter, tx_ring, i)) {
285 /* schedule immediate reset if we believe we hung */
286 DPRINTK(PROBE, INFO,
287 "tx hang %d detected, resetting adapter\n",
288 adapter->tx_timeout_count + 1);
289 ixgbe_tx_timeout(adapter->netdev);
290 }
291 }
9a799d71 292
e01c31a5
JB
293 /* re-arm the interrupt */
294 if ((total_packets >= tx_ring->work_limit) ||
295 (count == tx_ring->count))
296 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, tx_ring->v_idx);
9a799d71 297
e01c31a5
JB
298 tx_ring->total_bytes += total_bytes;
299 tx_ring->total_packets += total_packets;
300 tx_ring->stats.bytes += total_bytes;
301 tx_ring->stats.packets += total_packets;
302 adapter->net_stats.tx_bytes += total_bytes;
303 adapter->net_stats.tx_packets += total_packets;
304 return (total_packets ? true : false);
9a799d71
AK
305}
306
5dd2d332 307#ifdef CONFIG_IXGBE_DCA
bd0362dd 308static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
b4617240 309 struct ixgbe_ring *rx_ring)
bd0362dd
JC
310{
311 u32 rxctrl;
312 int cpu = get_cpu();
3a581073 313 int q = rx_ring - adapter->rx_ring;
bd0362dd 314
3a581073 315 if (rx_ring->cpu != cpu) {
bd0362dd
JC
316 rxctrl = IXGBE_READ_REG(&adapter->hw, IXGBE_DCA_RXCTRL(q));
317 rxctrl &= ~IXGBE_DCA_RXCTRL_CPUID_MASK;
96b0e0f6 318 rxctrl |= dca3_get_tag(&adapter->pdev->dev, cpu);
bd0362dd
JC
319 rxctrl |= IXGBE_DCA_RXCTRL_DESC_DCA_EN;
320 rxctrl |= IXGBE_DCA_RXCTRL_HEAD_DCA_EN;
321 IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_RXCTRL(q), rxctrl);
3a581073 322 rx_ring->cpu = cpu;
bd0362dd
JC
323 }
324 put_cpu();
325}
326
327static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
b4617240 328 struct ixgbe_ring *tx_ring)
bd0362dd
JC
329{
330 u32 txctrl;
331 int cpu = get_cpu();
3a581073 332 int q = tx_ring - adapter->tx_ring;
bd0362dd 333
3a581073 334 if (tx_ring->cpu != cpu) {
bd0362dd
JC
335 txctrl = IXGBE_READ_REG(&adapter->hw, IXGBE_DCA_TXCTRL(q));
336 txctrl &= ~IXGBE_DCA_TXCTRL_CPUID_MASK;
96b0e0f6 337 txctrl |= dca3_get_tag(&adapter->pdev->dev, cpu);
bd0362dd
JC
338 txctrl |= IXGBE_DCA_TXCTRL_DESC_DCA_EN;
339 IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_TXCTRL(q), txctrl);
3a581073 340 tx_ring->cpu = cpu;
bd0362dd
JC
341 }
342 put_cpu();
343}
344
345static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
346{
347 int i;
348
349 if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
350 return;
351
352 for (i = 0; i < adapter->num_tx_queues; i++) {
353 adapter->tx_ring[i].cpu = -1;
354 ixgbe_update_tx_dca(adapter, &adapter->tx_ring[i]);
355 }
356 for (i = 0; i < adapter->num_rx_queues; i++) {
357 adapter->rx_ring[i].cpu = -1;
358 ixgbe_update_rx_dca(adapter, &adapter->rx_ring[i]);
359 }
360}
361
362static int __ixgbe_notify_dca(struct device *dev, void *data)
363{
364 struct net_device *netdev = dev_get_drvdata(dev);
365 struct ixgbe_adapter *adapter = netdev_priv(netdev);
366 unsigned long event = *(unsigned long *)data;
367
368 switch (event) {
369 case DCA_PROVIDER_ADD:
96b0e0f6
JB
370 /* if we're already enabled, don't do it again */
371 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
372 break;
bd0362dd
JC
373 /* Always use CB2 mode, difference is masked
374 * in the CB driver. */
375 IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
652f093f 376 if (dca_add_requester(dev) == 0) {
96b0e0f6 377 adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
bd0362dd
JC
378 ixgbe_setup_dca(adapter);
379 break;
380 }
381 /* Fall Through since DCA is disabled. */
382 case DCA_PROVIDER_REMOVE:
383 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
384 dca_remove_requester(dev);
385 adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
386 IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
387 }
388 break;
389 }
390
652f093f 391 return 0;
bd0362dd
JC
392}
393
5dd2d332 394#endif /* CONFIG_IXGBE_DCA */
9a799d71
AK
395/**
396 * ixgbe_receive_skb - Send a completed packet up the stack
397 * @adapter: board private structure
398 * @skb: packet to send up
177db6ff
MC
399 * @status: hardware indication of status of receive
400 * @rx_ring: rx descriptor ring (for a specific queue) to setup
401 * @rx_desc: rx descriptor
9a799d71
AK
402 **/
403static void ixgbe_receive_skb(struct ixgbe_adapter *adapter,
b4617240
PW
404 struct sk_buff *skb, u8 status,
405 struct ixgbe_ring *ring,
177db6ff 406 union ixgbe_adv_rx_desc *rx_desc)
9a799d71 407{
177db6ff
MC
408 bool is_vlan = (status & IXGBE_RXD_STAT_VP);
409 u16 tag = le16_to_cpu(rx_desc->wb.upper.vlan);
9a799d71 410
177db6ff
MC
411 if (adapter->netdev->features & NETIF_F_LRO &&
412 skb->ip_summed == CHECKSUM_UNNECESSARY) {
2f90b865 413 if (adapter->vlgrp && is_vlan && (tag != 0))
177db6ff
MC
414 lro_vlan_hwaccel_receive_skb(&ring->lro_mgr, skb,
415 adapter->vlgrp, tag,
416 rx_desc);
9a799d71 417 else
177db6ff
MC
418 lro_receive_skb(&ring->lro_mgr, skb, rx_desc);
419 ring->lro_used = true;
420 } else {
421 if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL)) {
2f90b865 422 if (adapter->vlgrp && is_vlan && (tag != 0))
177db6ff
MC
423 vlan_hwaccel_receive_skb(skb, adapter->vlgrp, tag);
424 else
425 netif_receive_skb(skb);
426 } else {
2f90b865 427 if (adapter->vlgrp && is_vlan && (tag != 0))
177db6ff
MC
428 vlan_hwaccel_rx(skb, adapter->vlgrp, tag);
429 else
430 netif_rx(skb);
431 }
9a799d71
AK
432 }
433}
434
e59bd25d
AV
435/**
436 * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
437 * @adapter: address of board private structure
438 * @status_err: hardware indication of status of receive
439 * @skb: skb currently being received and modified
440 **/
9a799d71 441static inline void ixgbe_rx_checksum(struct ixgbe_adapter *adapter,
712744be 442 u32 status_err, struct sk_buff *skb)
9a799d71
AK
443{
444 skb->ip_summed = CHECKSUM_NONE;
445
712744be
JB
446 /* Rx csum disabled */
447 if (!(adapter->flags & IXGBE_FLAG_RX_CSUM_ENABLED))
9a799d71 448 return;
e59bd25d
AV
449
450 /* if IP and error */
451 if ((status_err & IXGBE_RXD_STAT_IPCS) &&
452 (status_err & IXGBE_RXDADV_ERR_IPE)) {
9a799d71
AK
453 adapter->hw_csum_rx_error++;
454 return;
455 }
e59bd25d
AV
456
457 if (!(status_err & IXGBE_RXD_STAT_L4CS))
458 return;
459
460 if (status_err & IXGBE_RXDADV_ERR_TCPE) {
461 adapter->hw_csum_rx_error++;
462 return;
463 }
464
9a799d71 465 /* It must be a TCP or UDP packet with a valid checksum */
e59bd25d 466 skb->ip_summed = CHECKSUM_UNNECESSARY;
9a799d71
AK
467 adapter->hw_csum_rx_good++;
468}
469
470/**
471 * ixgbe_alloc_rx_buffers - Replace used receive buffers; packet split
472 * @adapter: address of board private structure
473 **/
474static void ixgbe_alloc_rx_buffers(struct ixgbe_adapter *adapter,
7c6e0a43
JB
475 struct ixgbe_ring *rx_ring,
476 int cleaned_count)
9a799d71 477{
9a799d71
AK
478 struct pci_dev *pdev = adapter->pdev;
479 union ixgbe_adv_rx_desc *rx_desc;
3a581073 480 struct ixgbe_rx_buffer *bi;
9a799d71 481 unsigned int i;
7c6e0a43 482 unsigned int bufsz = rx_ring->rx_buf_len + NET_IP_ALIGN;
9a799d71
AK
483
484 i = rx_ring->next_to_use;
3a581073 485 bi = &rx_ring->rx_buffer_info[i];
9a799d71
AK
486
487 while (cleaned_count--) {
488 rx_desc = IXGBE_RX_DESC_ADV(*rx_ring, i);
489
762f4c57 490 if (!bi->page_dma &&
3a581073 491 (adapter->flags & IXGBE_FLAG_RX_PS_ENABLED)) {
3a581073 492 if (!bi->page) {
762f4c57
JB
493 bi->page = alloc_page(GFP_ATOMIC);
494 if (!bi->page) {
495 adapter->alloc_rx_page_failed++;
496 goto no_buffers;
497 }
498 bi->page_offset = 0;
499 } else {
500 /* use a half page if we're re-using */
501 bi->page_offset ^= (PAGE_SIZE / 2);
9a799d71 502 }
762f4c57
JB
503
504 bi->page_dma = pci_map_page(pdev, bi->page,
505 bi->page_offset,
506 (PAGE_SIZE / 2),
507 PCI_DMA_FROMDEVICE);
9a799d71
AK
508 }
509
3a581073 510 if (!bi->skb) {
74ce8dd2
JB
511 struct sk_buff *skb = netdev_alloc_skb(adapter->netdev,
512 bufsz);
9a799d71
AK
513
514 if (!skb) {
515 adapter->alloc_rx_buff_failed++;
516 goto no_buffers;
517 }
518
519 /*
520 * Make buffer alignment 2 beyond a 16 byte boundary
521 * this will result in a 16 byte aligned IP header after
522 * the 14 byte MAC header is removed
523 */
524 skb_reserve(skb, NET_IP_ALIGN);
525
3a581073
JB
526 bi->skb = skb;
527 bi->dma = pci_map_single(pdev, skb->data, bufsz,
528 PCI_DMA_FROMDEVICE);
9a799d71
AK
529 }
530 /* Refresh the desc even if buffer_addrs didn't change because
531 * each write-back erases this info. */
532 if (adapter->flags & IXGBE_FLAG_RX_PS_ENABLED) {
3a581073
JB
533 rx_desc->read.pkt_addr = cpu_to_le64(bi->page_dma);
534 rx_desc->read.hdr_addr = cpu_to_le64(bi->dma);
9a799d71 535 } else {
3a581073 536 rx_desc->read.pkt_addr = cpu_to_le64(bi->dma);
9a799d71
AK
537 }
538
539 i++;
540 if (i == rx_ring->count)
541 i = 0;
3a581073 542 bi = &rx_ring->rx_buffer_info[i];
9a799d71 543 }
7c6e0a43 544
9a799d71
AK
545no_buffers:
546 if (rx_ring->next_to_use != i) {
547 rx_ring->next_to_use = i;
548 if (i-- == 0)
549 i = (rx_ring->count - 1);
550
551 /*
552 * Force memory writes to complete before letting h/w
553 * know there are new descriptors to fetch. (Only
554 * applicable for weak-ordered memory model archs,
555 * such as IA-64).
556 */
557 wmb();
558 writel(i, adapter->hw.hw_addr + rx_ring->tail);
559 }
560}
561
7c6e0a43
JB
562static inline u16 ixgbe_get_hdr_info(union ixgbe_adv_rx_desc *rx_desc)
563{
564 return rx_desc->wb.lower.lo_dword.hs_rss.hdr_info;
565}
566
567static inline u16 ixgbe_get_pkt_info(union ixgbe_adv_rx_desc *rx_desc)
568{
569 return rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
570}
571
9a799d71 572static bool ixgbe_clean_rx_irq(struct ixgbe_adapter *adapter,
b4617240
PW
573 struct ixgbe_ring *rx_ring,
574 int *work_done, int work_to_do)
9a799d71 575{
9a799d71
AK
576 struct pci_dev *pdev = adapter->pdev;
577 union ixgbe_adv_rx_desc *rx_desc, *next_rxd;
578 struct ixgbe_rx_buffer *rx_buffer_info, *next_buffer;
579 struct sk_buff *skb;
580 unsigned int i;
7c6e0a43 581 u32 len, staterr;
177db6ff
MC
582 u16 hdr_info;
583 bool cleaned = false;
9a799d71 584 int cleaned_count = 0;
d2f4fbe2 585 unsigned int total_rx_bytes = 0, total_rx_packets = 0;
9a799d71
AK
586
587 i = rx_ring->next_to_clean;
9a799d71
AK
588 rx_desc = IXGBE_RX_DESC_ADV(*rx_ring, i);
589 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
590 rx_buffer_info = &rx_ring->rx_buffer_info[i];
9a799d71
AK
591
592 while (staterr & IXGBE_RXD_STAT_DD) {
7c6e0a43 593 u32 upper_len = 0;
9a799d71
AK
594 if (*work_done >= work_to_do)
595 break;
596 (*work_done)++;
597
598 if (adapter->flags & IXGBE_FLAG_RX_PS_ENABLED) {
7c6e0a43
JB
599 hdr_info = le16_to_cpu(ixgbe_get_hdr_info(rx_desc));
600 len = (hdr_info & IXGBE_RXDADV_HDRBUFLEN_MASK) >>
762f4c57 601 IXGBE_RXDADV_HDRBUFLEN_SHIFT;
9a799d71
AK
602 if (hdr_info & IXGBE_RXDADV_SPH)
603 adapter->rx_hdr_split++;
604 if (len > IXGBE_RX_HDR_SIZE)
605 len = IXGBE_RX_HDR_SIZE;
606 upper_len = le16_to_cpu(rx_desc->wb.upper.length);
7c6e0a43 607 } else {
9a799d71 608 len = le16_to_cpu(rx_desc->wb.upper.length);
7c6e0a43 609 }
9a799d71
AK
610
611 cleaned = true;
612 skb = rx_buffer_info->skb;
613 prefetch(skb->data - NET_IP_ALIGN);
614 rx_buffer_info->skb = NULL;
615
616 if (len && !skb_shinfo(skb)->nr_frags) {
617 pci_unmap_single(pdev, rx_buffer_info->dma,
b4617240
PW
618 rx_ring->rx_buf_len + NET_IP_ALIGN,
619 PCI_DMA_FROMDEVICE);
9a799d71
AK
620 skb_put(skb, len);
621 }
622
623 if (upper_len) {
624 pci_unmap_page(pdev, rx_buffer_info->page_dma,
762f4c57 625 PAGE_SIZE / 2, PCI_DMA_FROMDEVICE);
9a799d71
AK
626 rx_buffer_info->page_dma = 0;
627 skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
762f4c57
JB
628 rx_buffer_info->page,
629 rx_buffer_info->page_offset,
630 upper_len);
631
632 if ((rx_ring->rx_buf_len > (PAGE_SIZE / 2)) ||
633 (page_count(rx_buffer_info->page) != 1))
634 rx_buffer_info->page = NULL;
635 else
636 get_page(rx_buffer_info->page);
9a799d71
AK
637
638 skb->len += upper_len;
639 skb->data_len += upper_len;
640 skb->truesize += upper_len;
641 }
642
643 i++;
644 if (i == rx_ring->count)
645 i = 0;
646 next_buffer = &rx_ring->rx_buffer_info[i];
647
648 next_rxd = IXGBE_RX_DESC_ADV(*rx_ring, i);
649 prefetch(next_rxd);
650
651 cleaned_count++;
652 if (staterr & IXGBE_RXD_STAT_EOP) {
653 rx_ring->stats.packets++;
654 rx_ring->stats.bytes += skb->len;
655 } else {
656 rx_buffer_info->skb = next_buffer->skb;
657 rx_buffer_info->dma = next_buffer->dma;
658 next_buffer->skb = skb;
762f4c57 659 next_buffer->dma = 0;
9a799d71
AK
660 adapter->non_eop_descs++;
661 goto next_desc;
662 }
663
664 if (staterr & IXGBE_RXDADV_ERR_FRAME_ERR_MASK) {
665 dev_kfree_skb_irq(skb);
666 goto next_desc;
667 }
668
669 ixgbe_rx_checksum(adapter, staterr, skb);
d2f4fbe2
AV
670
671 /* probably a little skewed due to removing CRC */
672 total_rx_bytes += skb->len;
673 total_rx_packets++;
674
74ce8dd2 675 skb->protocol = eth_type_trans(skb, adapter->netdev);
177db6ff 676 ixgbe_receive_skb(adapter, skb, staterr, rx_ring, rx_desc);
9a799d71
AK
677
678next_desc:
679 rx_desc->wb.upper.status_error = 0;
680
681 /* return some buffers to hardware, one at a time is too slow */
682 if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
683 ixgbe_alloc_rx_buffers(adapter, rx_ring, cleaned_count);
684 cleaned_count = 0;
685 }
686
687 /* use prefetched values */
688 rx_desc = next_rxd;
689 rx_buffer_info = next_buffer;
690
691 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
177db6ff
MC
692 }
693
694 if (rx_ring->lro_used) {
695 lro_flush_all(&rx_ring->lro_mgr);
696 rx_ring->lro_used = false;
9a799d71
AK
697 }
698
699 rx_ring->next_to_clean = i;
700 cleaned_count = IXGBE_DESC_UNUSED(rx_ring);
701
702 if (cleaned_count)
703 ixgbe_alloc_rx_buffers(adapter, rx_ring, cleaned_count);
704
f494e8fa
AV
705 rx_ring->total_packets += total_rx_packets;
706 rx_ring->total_bytes += total_rx_bytes;
707 adapter->net_stats.rx_bytes += total_rx_bytes;
708 adapter->net_stats.rx_packets += total_rx_packets;
709
9a799d71
AK
710 return cleaned;
711}
712
021230d4 713static int ixgbe_clean_rxonly(struct napi_struct *, int);
9a799d71
AK
714/**
715 * ixgbe_configure_msix - Configure MSI-X hardware
716 * @adapter: board private structure
717 *
718 * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
719 * interrupts.
720 **/
721static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
722{
021230d4
AV
723 struct ixgbe_q_vector *q_vector;
724 int i, j, q_vectors, v_idx, r_idx;
725 u32 mask;
9a799d71 726
021230d4 727 q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
9a799d71 728
021230d4
AV
729 /* Populate the IVAR table and set the ITR values to the
730 * corresponding register.
731 */
732 for (v_idx = 0; v_idx < q_vectors; v_idx++) {
733 q_vector = &adapter->q_vector[v_idx];
734 /* XXX for_each_bit(...) */
735 r_idx = find_first_bit(q_vector->rxr_idx,
b4617240 736 adapter->num_rx_queues);
021230d4
AV
737
738 for (i = 0; i < q_vector->rxr_count; i++) {
739 j = adapter->rx_ring[r_idx].reg_idx;
740 ixgbe_set_ivar(adapter, IXGBE_IVAR_RX_QUEUE(j), v_idx);
741 r_idx = find_next_bit(q_vector->rxr_idx,
b4617240
PW
742 adapter->num_rx_queues,
743 r_idx + 1);
021230d4
AV
744 }
745 r_idx = find_first_bit(q_vector->txr_idx,
b4617240 746 adapter->num_tx_queues);
021230d4
AV
747
748 for (i = 0; i < q_vector->txr_count; i++) {
749 j = adapter->tx_ring[r_idx].reg_idx;
750 ixgbe_set_ivar(adapter, IXGBE_IVAR_TX_QUEUE(j), v_idx);
751 r_idx = find_next_bit(q_vector->txr_idx,
b4617240
PW
752 adapter->num_tx_queues,
753 r_idx + 1);
021230d4
AV
754 }
755
30efa5a3 756 /* if this is a tx only vector halve the interrupt rate */
021230d4 757 if (q_vector->txr_count && !q_vector->rxr_count)
30efa5a3 758 q_vector->eitr = (adapter->eitr_param >> 1);
021230d4 759 else
30efa5a3
JB
760 /* rx only */
761 q_vector->eitr = adapter->eitr_param;
021230d4
AV
762
763 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx),
b4617240 764 EITR_INTS_PER_SEC_TO_REG(q_vector->eitr));
9a799d71
AK
765 }
766
021230d4
AV
767 ixgbe_set_ivar(adapter, IXGBE_IVAR_OTHER_CAUSES_INDEX, v_idx);
768 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
769
41fb9248 770 /* set up to autoclear timer, and the vectors */
021230d4 771 mask = IXGBE_EIMS_ENABLE_MASK;
41fb9248 772 mask &= ~(IXGBE_EIMS_OTHER | IXGBE_EIMS_LSC);
021230d4 773 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
9a799d71
AK
774}
775
f494e8fa
AV
776enum latency_range {
777 lowest_latency = 0,
778 low_latency = 1,
779 bulk_latency = 2,
780 latency_invalid = 255
781};
782
783/**
784 * ixgbe_update_itr - update the dynamic ITR value based on statistics
785 * @adapter: pointer to adapter
786 * @eitr: eitr setting (ints per sec) to give last timeslice
787 * @itr_setting: current throttle rate in ints/second
788 * @packets: the number of packets during this measurement interval
789 * @bytes: the number of bytes during this measurement interval
790 *
791 * Stores a new ITR value based on packets and byte
792 * counts during the last interrupt. The advantage of per interrupt
793 * computation is faster updates and more accurate ITR for the current
794 * traffic pattern. Constants in this function were computed
795 * based on theoretical maximum wire speed and thresholds were set based
796 * on testing data as well as attempting to minimize response time
797 * while increasing bulk throughput.
798 * this functionality is controlled by the InterruptThrottleRate module
799 * parameter (see ixgbe_param.c)
800 **/
801static u8 ixgbe_update_itr(struct ixgbe_adapter *adapter,
b4617240
PW
802 u32 eitr, u8 itr_setting,
803 int packets, int bytes)
f494e8fa
AV
804{
805 unsigned int retval = itr_setting;
806 u32 timepassed_us;
807 u64 bytes_perint;
808
809 if (packets == 0)
810 goto update_itr_done;
811
812
813 /* simple throttlerate management
814 * 0-20MB/s lowest (100000 ints/s)
815 * 20-100MB/s low (20000 ints/s)
816 * 100-1249MB/s bulk (8000 ints/s)
817 */
818 /* what was last interrupt timeslice? */
819 timepassed_us = 1000000/eitr;
820 bytes_perint = bytes / timepassed_us; /* bytes/usec */
821
822 switch (itr_setting) {
823 case lowest_latency:
824 if (bytes_perint > adapter->eitr_low)
825 retval = low_latency;
826 break;
827 case low_latency:
828 if (bytes_perint > adapter->eitr_high)
829 retval = bulk_latency;
830 else if (bytes_perint <= adapter->eitr_low)
831 retval = lowest_latency;
832 break;
833 case bulk_latency:
834 if (bytes_perint <= adapter->eitr_high)
835 retval = low_latency;
836 break;
837 }
838
839update_itr_done:
840 return retval;
841}
842
843static void ixgbe_set_itr_msix(struct ixgbe_q_vector *q_vector)
844{
845 struct ixgbe_adapter *adapter = q_vector->adapter;
846 struct ixgbe_hw *hw = &adapter->hw;
847 u32 new_itr;
848 u8 current_itr, ret_itr;
849 int i, r_idx, v_idx = ((void *)q_vector - (void *)(adapter->q_vector)) /
b4617240 850 sizeof(struct ixgbe_q_vector);
f494e8fa
AV
851 struct ixgbe_ring *rx_ring, *tx_ring;
852
853 r_idx = find_first_bit(q_vector->txr_idx, adapter->num_tx_queues);
854 for (i = 0; i < q_vector->txr_count; i++) {
855 tx_ring = &(adapter->tx_ring[r_idx]);
856 ret_itr = ixgbe_update_itr(adapter, q_vector->eitr,
b4617240
PW
857 q_vector->tx_itr,
858 tx_ring->total_packets,
859 tx_ring->total_bytes);
f494e8fa
AV
860 /* if the result for this queue would decrease interrupt
861 * rate for this vector then use that result */
30efa5a3 862 q_vector->tx_itr = ((q_vector->tx_itr > ret_itr) ?
b4617240 863 q_vector->tx_itr - 1 : ret_itr);
f494e8fa 864 r_idx = find_next_bit(q_vector->txr_idx, adapter->num_tx_queues,
b4617240 865 r_idx + 1);
f494e8fa
AV
866 }
867
868 r_idx = find_first_bit(q_vector->rxr_idx, adapter->num_rx_queues);
869 for (i = 0; i < q_vector->rxr_count; i++) {
870 rx_ring = &(adapter->rx_ring[r_idx]);
871 ret_itr = ixgbe_update_itr(adapter, q_vector->eitr,
b4617240
PW
872 q_vector->rx_itr,
873 rx_ring->total_packets,
874 rx_ring->total_bytes);
f494e8fa
AV
875 /* if the result for this queue would decrease interrupt
876 * rate for this vector then use that result */
30efa5a3 877 q_vector->rx_itr = ((q_vector->rx_itr > ret_itr) ?
b4617240 878 q_vector->rx_itr - 1 : ret_itr);
f494e8fa 879 r_idx = find_next_bit(q_vector->rxr_idx, adapter->num_rx_queues,
b4617240 880 r_idx + 1);
f494e8fa
AV
881 }
882
30efa5a3 883 current_itr = max(q_vector->rx_itr, q_vector->tx_itr);
f494e8fa
AV
884
885 switch (current_itr) {
886 /* counts and packets in update_itr are dependent on these numbers */
887 case lowest_latency:
888 new_itr = 100000;
889 break;
890 case low_latency:
891 new_itr = 20000; /* aka hwitr = ~200 */
892 break;
893 case bulk_latency:
894 default:
895 new_itr = 8000;
896 break;
897 }
898
899 if (new_itr != q_vector->eitr) {
900 u32 itr_reg;
901 /* do an exponential smoothing */
902 new_itr = ((q_vector->eitr * 90)/100) + ((new_itr * 10)/100);
903 q_vector->eitr = new_itr;
904 itr_reg = EITR_INTS_PER_SEC_TO_REG(new_itr);
905 /* must write high and low 16 bits to reset counter */
906 DPRINTK(TX_ERR, DEBUG, "writing eitr(%d): %08X\n", v_idx,
b4617240 907 itr_reg);
f494e8fa
AV
908 IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg | (itr_reg)<<16);
909 }
910
911 return;
912}
913
0befdb3e
JB
914static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
915{
916 struct ixgbe_hw *hw = &adapter->hw;
917
918 if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
919 (eicr & IXGBE_EICR_GPI_SDP1)) {
920 DPRINTK(PROBE, CRIT, "Fan has stopped, replace the adapter\n");
921 /* write to clear the interrupt */
922 IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
923 }
924}
cf8280ee
JB
925
926static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
927{
928 struct ixgbe_hw *hw = &adapter->hw;
929
930 adapter->lsc_int++;
931 adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
932 adapter->link_check_timeout = jiffies;
933 if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
934 IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
935 schedule_work(&adapter->watchdog_task);
936 }
937}
938
9a799d71
AK
939static irqreturn_t ixgbe_msix_lsc(int irq, void *data)
940{
941 struct net_device *netdev = data;
942 struct ixgbe_adapter *adapter = netdev_priv(netdev);
943 struct ixgbe_hw *hw = &adapter->hw;
944 u32 eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
945
cf8280ee
JB
946 if (eicr & IXGBE_EICR_LSC)
947 ixgbe_check_lsc(adapter);
d4f80882 948
0befdb3e
JB
949 ixgbe_check_fan_failure(adapter, eicr);
950
d4f80882
AV
951 if (!test_bit(__IXGBE_DOWN, &adapter->state))
952 IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_OTHER);
9a799d71
AK
953
954 return IRQ_HANDLED;
955}
956
957static irqreturn_t ixgbe_msix_clean_tx(int irq, void *data)
958{
021230d4
AV
959 struct ixgbe_q_vector *q_vector = data;
960 struct ixgbe_adapter *adapter = q_vector->adapter;
3a581073 961 struct ixgbe_ring *tx_ring;
021230d4
AV
962 int i, r_idx;
963
964 if (!q_vector->txr_count)
965 return IRQ_HANDLED;
966
967 r_idx = find_first_bit(q_vector->txr_idx, adapter->num_tx_queues);
968 for (i = 0; i < q_vector->txr_count; i++) {
3a581073 969 tx_ring = &(adapter->tx_ring[r_idx]);
5dd2d332 970#ifdef CONFIG_IXGBE_DCA
bd0362dd 971 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3a581073 972 ixgbe_update_tx_dca(adapter, tx_ring);
bd0362dd 973#endif
3a581073
JB
974 tx_ring->total_bytes = 0;
975 tx_ring->total_packets = 0;
976 ixgbe_clean_tx_irq(adapter, tx_ring);
021230d4 977 r_idx = find_next_bit(q_vector->txr_idx, adapter->num_tx_queues,
b4617240 978 r_idx + 1);
021230d4 979 }
9a799d71 980
9a799d71
AK
981 return IRQ_HANDLED;
982}
983
021230d4
AV
984/**
985 * ixgbe_msix_clean_rx - single unshared vector rx clean (all queues)
986 * @irq: unused
987 * @data: pointer to our q_vector struct for this interrupt vector
988 **/
9a799d71
AK
989static irqreturn_t ixgbe_msix_clean_rx(int irq, void *data)
990{
021230d4
AV
991 struct ixgbe_q_vector *q_vector = data;
992 struct ixgbe_adapter *adapter = q_vector->adapter;
3a581073 993 struct ixgbe_ring *rx_ring;
021230d4 994 int r_idx;
30efa5a3 995 int i;
021230d4
AV
996
997 r_idx = find_first_bit(q_vector->rxr_idx, adapter->num_rx_queues);
30efa5a3
JB
998 for (i = 0; i < q_vector->rxr_count; i++) {
999 rx_ring = &(adapter->rx_ring[r_idx]);
1000 rx_ring->total_bytes = 0;
1001 rx_ring->total_packets = 0;
1002 r_idx = find_next_bit(q_vector->rxr_idx, adapter->num_rx_queues,
1003 r_idx + 1);
1004 }
1005
021230d4
AV
1006 if (!q_vector->rxr_count)
1007 return IRQ_HANDLED;
1008
30efa5a3 1009 r_idx = find_first_bit(q_vector->rxr_idx, adapter->num_rx_queues);
3a581073 1010 rx_ring = &(adapter->rx_ring[r_idx]);
021230d4 1011 /* disable interrupts on this vector only */
3a581073 1012 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, rx_ring->v_idx);
021230d4
AV
1013 netif_rx_schedule(adapter->netdev, &q_vector->napi);
1014
1015 return IRQ_HANDLED;
1016}
1017
1018static irqreturn_t ixgbe_msix_clean_many(int irq, void *data)
1019{
1020 ixgbe_msix_clean_rx(irq, data);
1021 ixgbe_msix_clean_tx(irq, data);
9a799d71 1022
9a799d71
AK
1023 return IRQ_HANDLED;
1024}
1025
021230d4
AV
1026/**
1027 * ixgbe_clean_rxonly - msix (aka one shot) rx clean routine
1028 * @napi: napi struct with our devices info in it
1029 * @budget: amount of work driver is allowed to do this pass, in packets
1030 *
f0848276
JB
1031 * This function is optimized for cleaning one queue only on a single
1032 * q_vector!!!
021230d4 1033 **/
9a799d71
AK
1034static int ixgbe_clean_rxonly(struct napi_struct *napi, int budget)
1035{
021230d4 1036 struct ixgbe_q_vector *q_vector =
b4617240 1037 container_of(napi, struct ixgbe_q_vector, napi);
021230d4 1038 struct ixgbe_adapter *adapter = q_vector->adapter;
f0848276 1039 struct ixgbe_ring *rx_ring = NULL;
9a799d71 1040 int work_done = 0;
021230d4 1041 long r_idx;
9a799d71 1042
021230d4 1043 r_idx = find_first_bit(q_vector->rxr_idx, adapter->num_rx_queues);
3a581073 1044 rx_ring = &(adapter->rx_ring[r_idx]);
5dd2d332 1045#ifdef CONFIG_IXGBE_DCA
bd0362dd 1046 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3a581073 1047 ixgbe_update_rx_dca(adapter, rx_ring);
bd0362dd 1048#endif
9a799d71 1049
3a581073 1050 ixgbe_clean_rx_irq(adapter, rx_ring, &work_done, budget);
9a799d71 1051
021230d4
AV
1052 /* If all Rx work done, exit the polling mode */
1053 if (work_done < budget) {
1054 netif_rx_complete(adapter->netdev, napi);
30efa5a3 1055 if (adapter->itr_setting & 3)
f494e8fa 1056 ixgbe_set_itr_msix(q_vector);
9a799d71 1057 if (!test_bit(__IXGBE_DOWN, &adapter->state))
3a581073 1058 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, rx_ring->v_idx);
9a799d71
AK
1059 }
1060
1061 return work_done;
1062}
1063
f0848276
JB
1064/**
1065 * ixgbe_clean_rxonly_many - msix (aka one shot) rx clean routine
1066 * @napi: napi struct with our devices info in it
1067 * @budget: amount of work driver is allowed to do this pass, in packets
1068 *
1069 * This function will clean more than one rx queue associated with a
1070 * q_vector.
1071 **/
1072static int ixgbe_clean_rxonly_many(struct napi_struct *napi, int budget)
1073{
1074 struct ixgbe_q_vector *q_vector =
1075 container_of(napi, struct ixgbe_q_vector, napi);
1076 struct ixgbe_adapter *adapter = q_vector->adapter;
f0848276
JB
1077 struct ixgbe_ring *rx_ring = NULL;
1078 int work_done = 0, i;
1079 long r_idx;
1080 u16 enable_mask = 0;
1081
1082 /* attempt to distribute budget to each queue fairly, but don't allow
1083 * the budget to go below 1 because we'll exit polling */
1084 budget /= (q_vector->rxr_count ?: 1);
1085 budget = max(budget, 1);
1086 r_idx = find_first_bit(q_vector->rxr_idx, adapter->num_rx_queues);
1087 for (i = 0; i < q_vector->rxr_count; i++) {
1088 rx_ring = &(adapter->rx_ring[r_idx]);
5dd2d332 1089#ifdef CONFIG_IXGBE_DCA
f0848276
JB
1090 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1091 ixgbe_update_rx_dca(adapter, rx_ring);
1092#endif
1093 ixgbe_clean_rx_irq(adapter, rx_ring, &work_done, budget);
1094 enable_mask |= rx_ring->v_idx;
1095 r_idx = find_next_bit(q_vector->rxr_idx, adapter->num_rx_queues,
1096 r_idx + 1);
1097 }
1098
1099 r_idx = find_first_bit(q_vector->rxr_idx, adapter->num_rx_queues);
1100 rx_ring = &(adapter->rx_ring[r_idx]);
1101 /* If all Rx work done, exit the polling mode */
7f821875 1102 if (work_done < budget) {
74ce8dd2 1103 netif_rx_complete(adapter->netdev, napi);
f0848276
JB
1104 if (adapter->itr_setting & 3)
1105 ixgbe_set_itr_msix(q_vector);
1106 if (!test_bit(__IXGBE_DOWN, &adapter->state))
1107 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, enable_mask);
1108 return 0;
1109 }
1110
1111 return work_done;
1112}
021230d4 1113static inline void map_vector_to_rxq(struct ixgbe_adapter *a, int v_idx,
b4617240 1114 int r_idx)
021230d4
AV
1115{
1116 a->q_vector[v_idx].adapter = a;
1117 set_bit(r_idx, a->q_vector[v_idx].rxr_idx);
1118 a->q_vector[v_idx].rxr_count++;
1119 a->rx_ring[r_idx].v_idx = 1 << v_idx;
1120}
1121
1122static inline void map_vector_to_txq(struct ixgbe_adapter *a, int v_idx,
b4617240 1123 int r_idx)
021230d4
AV
1124{
1125 a->q_vector[v_idx].adapter = a;
1126 set_bit(r_idx, a->q_vector[v_idx].txr_idx);
1127 a->q_vector[v_idx].txr_count++;
1128 a->tx_ring[r_idx].v_idx = 1 << v_idx;
1129}
1130
9a799d71 1131/**
021230d4
AV
1132 * ixgbe_map_rings_to_vectors - Maps descriptor rings to vectors
1133 * @adapter: board private structure to initialize
1134 * @vectors: allotted vector count for descriptor rings
9a799d71 1135 *
021230d4
AV
1136 * This function maps descriptor rings to the queue-specific vectors
1137 * we were allotted through the MSI-X enabling code. Ideally, we'd have
1138 * one vector per ring/queue, but on a constrained vector budget, we
1139 * group the rings as "efficiently" as possible. You would add new
1140 * mapping configurations in here.
9a799d71 1141 **/
021230d4 1142static int ixgbe_map_rings_to_vectors(struct ixgbe_adapter *adapter,
b4617240 1143 int vectors)
021230d4
AV
1144{
1145 int v_start = 0;
1146 int rxr_idx = 0, txr_idx = 0;
1147 int rxr_remaining = adapter->num_rx_queues;
1148 int txr_remaining = adapter->num_tx_queues;
1149 int i, j;
1150 int rqpv, tqpv;
1151 int err = 0;
1152
1153 /* No mapping required if MSI-X is disabled. */
1154 if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED))
1155 goto out;
9a799d71 1156
021230d4
AV
1157 /*
1158 * The ideal configuration...
1159 * We have enough vectors to map one per queue.
1160 */
1161 if (vectors == adapter->num_rx_queues + adapter->num_tx_queues) {
1162 for (; rxr_idx < rxr_remaining; v_start++, rxr_idx++)
1163 map_vector_to_rxq(adapter, v_start, rxr_idx);
9a799d71 1164
021230d4
AV
1165 for (; txr_idx < txr_remaining; v_start++, txr_idx++)
1166 map_vector_to_txq(adapter, v_start, txr_idx);
9a799d71 1167
9a799d71 1168 goto out;
021230d4 1169 }
9a799d71 1170
021230d4
AV
1171 /*
1172 * If we don't have enough vectors for a 1-to-1
1173 * mapping, we'll have to group them so there are
1174 * multiple queues per vector.
1175 */
1176 /* Re-adjusting *qpv takes care of the remainder. */
1177 for (i = v_start; i < vectors; i++) {
1178 rqpv = DIV_ROUND_UP(rxr_remaining, vectors - i);
1179 for (j = 0; j < rqpv; j++) {
1180 map_vector_to_rxq(adapter, i, rxr_idx);
1181 rxr_idx++;
1182 rxr_remaining--;
1183 }
1184 }
1185 for (i = v_start; i < vectors; i++) {
1186 tqpv = DIV_ROUND_UP(txr_remaining, vectors - i);
1187 for (j = 0; j < tqpv; j++) {
1188 map_vector_to_txq(adapter, i, txr_idx);
1189 txr_idx++;
1190 txr_remaining--;
9a799d71 1191 }
9a799d71
AK
1192 }
1193
021230d4
AV
1194out:
1195 return err;
1196}
1197
1198/**
1199 * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
1200 * @adapter: board private structure
1201 *
1202 * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
1203 * interrupts from the kernel.
1204 **/
1205static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
1206{
1207 struct net_device *netdev = adapter->netdev;
1208 irqreturn_t (*handler)(int, void *);
1209 int i, vector, q_vectors, err;
1210
1211 /* Decrement for Other and TCP Timer vectors */
1212 q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
1213
1214 /* Map the Tx/Rx rings to the vectors we were allotted. */
1215 err = ixgbe_map_rings_to_vectors(adapter, q_vectors);
1216 if (err)
1217 goto out;
1218
1219#define SET_HANDLER(_v) ((!(_v)->rxr_count) ? &ixgbe_msix_clean_tx : \
b4617240
PW
1220 (!(_v)->txr_count) ? &ixgbe_msix_clean_rx : \
1221 &ixgbe_msix_clean_many)
021230d4
AV
1222 for (vector = 0; vector < q_vectors; vector++) {
1223 handler = SET_HANDLER(&adapter->q_vector[vector]);
1224 sprintf(adapter->name[vector], "%s:v%d-%s",
b4617240
PW
1225 netdev->name, vector,
1226 (handler == &ixgbe_msix_clean_rx) ? "Rx" :
1227 ((handler == &ixgbe_msix_clean_tx) ? "Tx" : "TxRx"));
021230d4 1228 err = request_irq(adapter->msix_entries[vector].vector,
b4617240
PW
1229 handler, 0, adapter->name[vector],
1230 &(adapter->q_vector[vector]));
9a799d71
AK
1231 if (err) {
1232 DPRINTK(PROBE, ERR,
b4617240
PW
1233 "request_irq failed for MSIX interrupt "
1234 "Error: %d\n", err);
021230d4 1235 goto free_queue_irqs;
9a799d71 1236 }
9a799d71
AK
1237 }
1238
021230d4
AV
1239 sprintf(adapter->name[vector], "%s:lsc", netdev->name);
1240 err = request_irq(adapter->msix_entries[vector].vector,
b4617240 1241 &ixgbe_msix_lsc, 0, adapter->name[vector], netdev);
9a799d71
AK
1242 if (err) {
1243 DPRINTK(PROBE, ERR,
1244 "request_irq for msix_lsc failed: %d\n", err);
021230d4 1245 goto free_queue_irqs;
9a799d71
AK
1246 }
1247
9a799d71
AK
1248 return 0;
1249
021230d4
AV
1250free_queue_irqs:
1251 for (i = vector - 1; i >= 0; i--)
1252 free_irq(adapter->msix_entries[--vector].vector,
b4617240 1253 &(adapter->q_vector[i]));
021230d4
AV
1254 adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
1255 pci_disable_msix(adapter->pdev);
9a799d71
AK
1256 kfree(adapter->msix_entries);
1257 adapter->msix_entries = NULL;
021230d4 1258out:
9a799d71
AK
1259 return err;
1260}
1261
f494e8fa
AV
1262static void ixgbe_set_itr(struct ixgbe_adapter *adapter)
1263{
1264 struct ixgbe_hw *hw = &adapter->hw;
1265 struct ixgbe_q_vector *q_vector = adapter->q_vector;
1266 u8 current_itr;
1267 u32 new_itr = q_vector->eitr;
1268 struct ixgbe_ring *rx_ring = &adapter->rx_ring[0];
1269 struct ixgbe_ring *tx_ring = &adapter->tx_ring[0];
1270
30efa5a3 1271 q_vector->tx_itr = ixgbe_update_itr(adapter, new_itr,
b4617240
PW
1272 q_vector->tx_itr,
1273 tx_ring->total_packets,
1274 tx_ring->total_bytes);
30efa5a3 1275 q_vector->rx_itr = ixgbe_update_itr(adapter, new_itr,
b4617240
PW
1276 q_vector->rx_itr,
1277 rx_ring->total_packets,
1278 rx_ring->total_bytes);
f494e8fa 1279
30efa5a3 1280 current_itr = max(q_vector->rx_itr, q_vector->tx_itr);
f494e8fa
AV
1281
1282 switch (current_itr) {
1283 /* counts and packets in update_itr are dependent on these numbers */
1284 case lowest_latency:
1285 new_itr = 100000;
1286 break;
1287 case low_latency:
1288 new_itr = 20000; /* aka hwitr = ~200 */
1289 break;
1290 case bulk_latency:
1291 new_itr = 8000;
1292 break;
1293 default:
1294 break;
1295 }
1296
1297 if (new_itr != q_vector->eitr) {
1298 u32 itr_reg;
1299 /* do an exponential smoothing */
1300 new_itr = ((q_vector->eitr * 90)/100) + ((new_itr * 10)/100);
1301 q_vector->eitr = new_itr;
1302 itr_reg = EITR_INTS_PER_SEC_TO_REG(new_itr);
1303 /* must write high and low 16 bits to reset counter */
1304 IXGBE_WRITE_REG(hw, IXGBE_EITR(0), itr_reg | (itr_reg)<<16);
1305 }
1306
1307 return;
1308}
1309
79aefa45
AD
1310/**
1311 * ixgbe_irq_disable - Mask off interrupt generation on the NIC
1312 * @adapter: board private structure
1313 **/
1314static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
1315{
1316 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
1317 IXGBE_WRITE_FLUSH(&adapter->hw);
1318 if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
1319 int i;
1320 for (i = 0; i < adapter->num_msix_vectors; i++)
1321 synchronize_irq(adapter->msix_entries[i].vector);
1322 } else {
1323 synchronize_irq(adapter->pdev->irq);
1324 }
1325}
1326
1327/**
1328 * ixgbe_irq_enable - Enable default interrupt generation settings
1329 * @adapter: board private structure
1330 **/
1331static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter)
1332{
1333 u32 mask;
1334 mask = IXGBE_EIMS_ENABLE_MASK;
6ab33d51
DM
1335 if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
1336 mask |= IXGBE_EIMS_GPI_SDP1;
79aefa45
AD
1337 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
1338 IXGBE_WRITE_FLUSH(&adapter->hw);
1339}
021230d4 1340
9a799d71 1341/**
021230d4 1342 * ixgbe_intr - legacy mode Interrupt Handler
9a799d71
AK
1343 * @irq: interrupt number
1344 * @data: pointer to a network interface device structure
1345 * @pt_regs: CPU registers structure
1346 **/
1347static irqreturn_t ixgbe_intr(int irq, void *data)
1348{
1349 struct net_device *netdev = data;
1350 struct ixgbe_adapter *adapter = netdev_priv(netdev);
1351 struct ixgbe_hw *hw = &adapter->hw;
1352 u32 eicr;
1353
021230d4
AV
1354 /* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
1355 * therefore no explict interrupt disable is necessary */
1356 eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
f47cf66e
JB
1357 if (!eicr) {
1358 /* shared interrupt alert!
1359 * make sure interrupts are enabled because the read will
1360 * have disabled interrupts due to EIAM */
1361 ixgbe_irq_enable(adapter);
9a799d71 1362 return IRQ_NONE; /* Not our interrupt */
f47cf66e 1363 }
9a799d71 1364
cf8280ee
JB
1365 if (eicr & IXGBE_EICR_LSC)
1366 ixgbe_check_lsc(adapter);
021230d4 1367
0befdb3e
JB
1368 ixgbe_check_fan_failure(adapter, eicr);
1369
021230d4 1370 if (netif_rx_schedule_prep(netdev, &adapter->q_vector[0].napi)) {
f494e8fa
AV
1371 adapter->tx_ring[0].total_packets = 0;
1372 adapter->tx_ring[0].total_bytes = 0;
1373 adapter->rx_ring[0].total_packets = 0;
1374 adapter->rx_ring[0].total_bytes = 0;
021230d4
AV
1375 /* would disable interrupts here but EIAM disabled it */
1376 __netif_rx_schedule(netdev, &adapter->q_vector[0].napi);
9a799d71
AK
1377 }
1378
1379 return IRQ_HANDLED;
1380}
1381
021230d4
AV
1382static inline void ixgbe_reset_q_vectors(struct ixgbe_adapter *adapter)
1383{
1384 int i, q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
1385
1386 for (i = 0; i < q_vectors; i++) {
1387 struct ixgbe_q_vector *q_vector = &adapter->q_vector[i];
1388 bitmap_zero(q_vector->rxr_idx, MAX_RX_QUEUES);
1389 bitmap_zero(q_vector->txr_idx, MAX_TX_QUEUES);
1390 q_vector->rxr_count = 0;
1391 q_vector->txr_count = 0;
1392 }
1393}
1394
9a799d71
AK
1395/**
1396 * ixgbe_request_irq - initialize interrupts
1397 * @adapter: board private structure
1398 *
1399 * Attempts to configure interrupts using the best available
1400 * capabilities of the hardware and kernel.
1401 **/
021230d4 1402static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
9a799d71
AK
1403{
1404 struct net_device *netdev = adapter->netdev;
021230d4 1405 int err;
9a799d71 1406
021230d4
AV
1407 if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
1408 err = ixgbe_request_msix_irqs(adapter);
1409 } else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED) {
1410 err = request_irq(adapter->pdev->irq, &ixgbe_intr, 0,
b4617240 1411 netdev->name, netdev);
021230d4
AV
1412 } else {
1413 err = request_irq(adapter->pdev->irq, &ixgbe_intr, IRQF_SHARED,
b4617240 1414 netdev->name, netdev);
9a799d71
AK
1415 }
1416
9a799d71
AK
1417 if (err)
1418 DPRINTK(PROBE, ERR, "request_irq failed, Error %d\n", err);
1419
9a799d71
AK
1420 return err;
1421}
1422
1423static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
1424{
1425 struct net_device *netdev = adapter->netdev;
1426
1427 if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
021230d4 1428 int i, q_vectors;
9a799d71 1429
021230d4
AV
1430 q_vectors = adapter->num_msix_vectors;
1431
1432 i = q_vectors - 1;
9a799d71 1433 free_irq(adapter->msix_entries[i].vector, netdev);
9a799d71 1434
021230d4
AV
1435 i--;
1436 for (; i >= 0; i--) {
1437 free_irq(adapter->msix_entries[i].vector,
b4617240 1438 &(adapter->q_vector[i]));
021230d4
AV
1439 }
1440
1441 ixgbe_reset_q_vectors(adapter);
1442 } else {
1443 free_irq(adapter->pdev->irq, netdev);
9a799d71
AK
1444 }
1445}
1446
9a799d71
AK
1447/**
1448 * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
1449 *
1450 **/
1451static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
1452{
9a799d71
AK
1453 struct ixgbe_hw *hw = &adapter->hw;
1454
021230d4 1455 IXGBE_WRITE_REG(hw, IXGBE_EITR(0),
30efa5a3 1456 EITR_INTS_PER_SEC_TO_REG(adapter->eitr_param));
9a799d71
AK
1457
1458 ixgbe_set_ivar(adapter, IXGBE_IVAR_RX_QUEUE(0), 0);
021230d4
AV
1459 ixgbe_set_ivar(adapter, IXGBE_IVAR_TX_QUEUE(0), 0);
1460
1461 map_vector_to_rxq(adapter, 0, 0);
1462 map_vector_to_txq(adapter, 0, 0);
1463
1464 DPRINTK(HW, INFO, "Legacy interrupt IVAR setup done\n");
9a799d71
AK
1465}
1466
1467/**
3a581073 1468 * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
9a799d71
AK
1469 * @adapter: board private structure
1470 *
1471 * Configure the Tx unit of the MAC after a reset.
1472 **/
1473static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
1474{
e01c31a5 1475 u64 tdba, tdwba;
9a799d71 1476 struct ixgbe_hw *hw = &adapter->hw;
021230d4 1477 u32 i, j, tdlen, txctrl;
9a799d71
AK
1478
1479 /* Setup the HW Tx Head and Tail descriptor pointers */
1480 for (i = 0; i < adapter->num_tx_queues; i++) {
e01c31a5
JB
1481 struct ixgbe_ring *ring = &adapter->tx_ring[i];
1482 j = ring->reg_idx;
1483 tdba = ring->dma;
1484 tdlen = ring->count * sizeof(union ixgbe_adv_tx_desc);
021230d4 1485 IXGBE_WRITE_REG(hw, IXGBE_TDBAL(j),
e01c31a5 1486 (tdba & DMA_32BIT_MASK));
021230d4 1487 IXGBE_WRITE_REG(hw, IXGBE_TDBAH(j), (tdba >> 32));
e01c31a5
JB
1488 tdwba = ring->dma +
1489 (ring->count * sizeof(union ixgbe_adv_tx_desc));
1490 tdwba |= IXGBE_TDWBAL_HEAD_WB_ENABLE;
1491 IXGBE_WRITE_REG(hw, IXGBE_TDWBAL(j), tdwba & DMA_32BIT_MASK);
1492 IXGBE_WRITE_REG(hw, IXGBE_TDWBAH(j), (tdwba >> 32));
021230d4
AV
1493 IXGBE_WRITE_REG(hw, IXGBE_TDLEN(j), tdlen);
1494 IXGBE_WRITE_REG(hw, IXGBE_TDH(j), 0);
1495 IXGBE_WRITE_REG(hw, IXGBE_TDT(j), 0);
1496 adapter->tx_ring[i].head = IXGBE_TDH(j);
1497 adapter->tx_ring[i].tail = IXGBE_TDT(j);
1498 /* Disable Tx Head Writeback RO bit, since this hoses
1499 * bookkeeping if things aren't delivered in order.
1500 */
e01c31a5 1501 txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL(j));
021230d4 1502 txctrl &= ~IXGBE_DCA_TXCTRL_TX_WB_RO_EN;
e01c31a5 1503 IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL(j), txctrl);
9a799d71 1504 }
9a799d71
AK
1505}
1506
cc41ac7c
JB
1507#define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
1508
1509static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter, int index)
1510{
1511 struct ixgbe_ring *rx_ring;
1512 u32 srrctl;
1513 int queue0;
3be1adfb
AD
1514 unsigned long mask;
1515
1516 /* program one srrctl register per VMDq index */
1517 if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
1518 long shift, len;
1519 mask = (unsigned long) adapter->ring_feature[RING_F_RSS].mask;
1520 len = sizeof(adapter->ring_feature[RING_F_VMDQ].mask) * 8;
1521 shift = find_first_bit(&mask, len);
1522 queue0 = index & mask;
1523 index = (index & mask) >> shift;
1524 /* program one srrctl per RSS queue since RDRXCTL.MVMEN is enabled */
cc41ac7c 1525 } else {
3be1adfb
AD
1526 mask = (unsigned long) adapter->ring_feature[RING_F_RSS].mask;
1527 queue0 = index & mask;
1528 index = index & mask;
cc41ac7c 1529 }
3be1adfb 1530
cc41ac7c
JB
1531 rx_ring = &adapter->rx_ring[queue0];
1532
1533 srrctl = IXGBE_READ_REG(&adapter->hw, IXGBE_SRRCTL(index));
1534
1535 srrctl &= ~IXGBE_SRRCTL_BSIZEHDR_MASK;
1536 srrctl &= ~IXGBE_SRRCTL_BSIZEPKT_MASK;
1537
1538 if (adapter->flags & IXGBE_FLAG_RX_PS_ENABLED) {
1539 srrctl |= IXGBE_RXBUFFER_2048 >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
1540 srrctl |= IXGBE_SRRCTL_DESCTYPE_HDR_SPLIT_ALWAYS;
1541 srrctl |= ((IXGBE_RX_HDR_SIZE <<
b4617240
PW
1542 IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT) &
1543 IXGBE_SRRCTL_BSIZEHDR_MASK);
cc41ac7c
JB
1544 } else {
1545 srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
1546
1547 if (rx_ring->rx_buf_len == MAXIMUM_ETHERNET_VLAN_SIZE)
1548 srrctl |= IXGBE_RXBUFFER_2048 >>
1549 IXGBE_SRRCTL_BSIZEPKT_SHIFT;
1550 else
1551 srrctl |= rx_ring->rx_buf_len >>
1552 IXGBE_SRRCTL_BSIZEPKT_SHIFT;
1553 }
1554 IXGBE_WRITE_REG(&adapter->hw, IXGBE_SRRCTL(index), srrctl);
1555}
9a799d71 1556
177db6ff
MC
1557/**
1558 * ixgbe_get_skb_hdr - helper function for LRO header processing
1559 * @skb: pointer to sk_buff to be added to LRO packet
b4617240 1560 * @iphdr: pointer to ip header structure
177db6ff
MC
1561 * @tcph: pointer to tcp header structure
1562 * @hdr_flags: pointer to header flags
1563 * @priv: private data
1564 **/
1565static int ixgbe_get_skb_hdr(struct sk_buff *skb, void **iphdr, void **tcph,
1566 u64 *hdr_flags, void *priv)
1567{
1568 union ixgbe_adv_rx_desc *rx_desc = priv;
1569
1570 /* Verify that this is a valid IPv4 TCP packet */
e9990a9c
JB
1571 if (!((ixgbe_get_pkt_info(rx_desc) & IXGBE_RXDADV_PKTTYPE_IPV4) &&
1572 (ixgbe_get_pkt_info(rx_desc) & IXGBE_RXDADV_PKTTYPE_TCP)))
177db6ff
MC
1573 return -1;
1574
1575 /* Set network headers */
1576 skb_reset_network_header(skb);
1577 skb_set_transport_header(skb, ip_hdrlen(skb));
1578 *iphdr = ip_hdr(skb);
1579 *tcph = tcp_hdr(skb);
1580 *hdr_flags = LRO_IPV4 | LRO_TCP;
1581 return 0;
1582}
1583
cc41ac7c 1584#define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
b4617240 1585 (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
cc41ac7c 1586
9a799d71 1587/**
3a581073 1588 * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
9a799d71
AK
1589 * @adapter: board private structure
1590 *
1591 * Configure the Rx unit of the MAC after a reset.
1592 **/
1593static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
1594{
1595 u64 rdba;
1596 struct ixgbe_hw *hw = &adapter->hw;
1597 struct net_device *netdev = adapter->netdev;
1598 int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
021230d4 1599 int i, j;
9a799d71 1600 u32 rdlen, rxctrl, rxcsum;
7c6e0a43
JB
1601 static const u32 seed[10] = { 0xE291D73D, 0x1805EC6C, 0x2A94B30D,
1602 0xA54F2BEC, 0xEA49AF7C, 0xE214AD3D, 0xB855AABE,
1603 0x6A3E67EA, 0x14364D17, 0x3BED200D};
9a799d71 1604 u32 fctrl, hlreg0;
9a799d71 1605 u32 pages;
cc41ac7c
JB
1606 u32 reta = 0, mrqc;
1607 u32 rdrxctl;
7c6e0a43 1608 int rx_buf_len;
9a799d71
AK
1609
1610 /* Decide whether to use packet split mode or not */
762f4c57 1611 adapter->flags |= IXGBE_FLAG_RX_PS_ENABLED;
9a799d71
AK
1612
1613 /* Set the RX buffer length according to the mode */
1614 if (adapter->flags & IXGBE_FLAG_RX_PS_ENABLED) {
7c6e0a43 1615 rx_buf_len = IXGBE_RX_HDR_SIZE;
9a799d71
AK
1616 } else {
1617 if (netdev->mtu <= ETH_DATA_LEN)
7c6e0a43 1618 rx_buf_len = MAXIMUM_ETHERNET_VLAN_SIZE;
9a799d71 1619 else
7c6e0a43 1620 rx_buf_len = ALIGN(max_frame, 1024);
9a799d71
AK
1621 }
1622
1623 fctrl = IXGBE_READ_REG(&adapter->hw, IXGBE_FCTRL);
1624 fctrl |= IXGBE_FCTRL_BAM;
021230d4 1625 fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
9a799d71
AK
1626 IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, fctrl);
1627
1628 hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
1629 if (adapter->netdev->mtu <= ETH_DATA_LEN)
1630 hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
1631 else
1632 hlreg0 |= IXGBE_HLREG0_JUMBOEN;
1633 IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
1634
1635 pages = PAGE_USE_COUNT(adapter->netdev->mtu);
1636
9a799d71
AK
1637 rdlen = adapter->rx_ring[0].count * sizeof(union ixgbe_adv_rx_desc);
1638 /* disable receives while setting up the descriptors */
1639 rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
1640 IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
1641
1642 /* Setup the HW Rx Head and Tail Descriptor Pointers and
1643 * the Base and Length of the Rx Descriptor Ring */
1644 for (i = 0; i < adapter->num_rx_queues; i++) {
1645 rdba = adapter->rx_ring[i].dma;
7c6e0a43
JB
1646 j = adapter->rx_ring[i].reg_idx;
1647 IXGBE_WRITE_REG(hw, IXGBE_RDBAL(j), (rdba & DMA_32BIT_MASK));
1648 IXGBE_WRITE_REG(hw, IXGBE_RDBAH(j), (rdba >> 32));
1649 IXGBE_WRITE_REG(hw, IXGBE_RDLEN(j), rdlen);
1650 IXGBE_WRITE_REG(hw, IXGBE_RDH(j), 0);
1651 IXGBE_WRITE_REG(hw, IXGBE_RDT(j), 0);
1652 adapter->rx_ring[i].head = IXGBE_RDH(j);
1653 adapter->rx_ring[i].tail = IXGBE_RDT(j);
1654 adapter->rx_ring[i].rx_buf_len = rx_buf_len;
e9990a9c
JB
1655 /* Intitial LRO Settings */
1656 adapter->rx_ring[i].lro_mgr.max_aggr = IXGBE_MAX_LRO_AGGREGATE;
1657 adapter->rx_ring[i].lro_mgr.max_desc = IXGBE_MAX_LRO_DESCRIPTORS;
1658 adapter->rx_ring[i].lro_mgr.get_skb_header = ixgbe_get_skb_hdr;
1659 adapter->rx_ring[i].lro_mgr.features = LRO_F_EXTRACT_VLAN_ID;
1660 if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
1661 adapter->rx_ring[i].lro_mgr.features |= LRO_F_NAPI;
1662 adapter->rx_ring[i].lro_mgr.dev = adapter->netdev;
1663 adapter->rx_ring[i].lro_mgr.ip_summed = CHECKSUM_UNNECESSARY;
1664 adapter->rx_ring[i].lro_mgr.ip_summed_aggr = CHECKSUM_UNNECESSARY;
cc41ac7c
JB
1665
1666 ixgbe_configure_srrctl(adapter, j);
9a799d71
AK
1667 }
1668
cc41ac7c
JB
1669 /*
1670 * For VMDq support of different descriptor types or
1671 * buffer sizes through the use of multiple SRRCTL
1672 * registers, RDRXCTL.MVMEN must be set to 1
1673 *
1674 * also, the manual doesn't mention it clearly but DCA hints
1675 * will only use queue 0's tags unless this bit is set. Side
1676 * effects of setting this bit are only that SRRCTL must be
1677 * fully programmed [0..15]
1678 */
2f90b865
AD
1679 if (adapter->flags &
1680 (IXGBE_FLAG_RSS_ENABLED | IXGBE_FLAG_VMDQ_ENABLED)) {
1681 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
1682 rdrxctl |= IXGBE_RDRXCTL_MVMEN;
1683 IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
1684 }
177db6ff 1685
021230d4 1686 if (adapter->flags & IXGBE_FLAG_RSS_ENABLED) {
9a799d71 1687 /* Fill out redirection table */
021230d4
AV
1688 for (i = 0, j = 0; i < 128; i++, j++) {
1689 if (j == adapter->ring_feature[RING_F_RSS].indices)
1690 j = 0;
1691 /* reta = 4-byte sliding window of
1692 * 0x00..(indices-1)(indices-1)00..etc. */
1693 reta = (reta << 8) | (j * 0x11);
1694 if ((i & 3) == 3)
1695 IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
9a799d71
AK
1696 }
1697
1698 /* Fill out hash function seeds */
1699 for (i = 0; i < 10; i++)
7c6e0a43 1700 IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
9a799d71
AK
1701
1702 mrqc = IXGBE_MRQC_RSSEN
1703 /* Perform hash on these packet types */
7c6e0a43
JB
1704 | IXGBE_MRQC_RSS_FIELD_IPV4
1705 | IXGBE_MRQC_RSS_FIELD_IPV4_TCP
1706 | IXGBE_MRQC_RSS_FIELD_IPV4_UDP
1707 | IXGBE_MRQC_RSS_FIELD_IPV6_EX_TCP
1708 | IXGBE_MRQC_RSS_FIELD_IPV6_EX
1709 | IXGBE_MRQC_RSS_FIELD_IPV6
1710 | IXGBE_MRQC_RSS_FIELD_IPV6_TCP
1711 | IXGBE_MRQC_RSS_FIELD_IPV6_UDP
1712 | IXGBE_MRQC_RSS_FIELD_IPV6_EX_UDP;
9a799d71 1713 IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
021230d4 1714 }
9a799d71 1715
021230d4
AV
1716 rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
1717
1718 if (adapter->flags & IXGBE_FLAG_RSS_ENABLED ||
1719 adapter->flags & IXGBE_FLAG_RX_CSUM_ENABLED) {
1720 /* Disable indicating checksum in descriptor, enables
1721 * RSS hash */
9a799d71 1722 rxcsum |= IXGBE_RXCSUM_PCSD;
9a799d71 1723 }
021230d4
AV
1724 if (!(rxcsum & IXGBE_RXCSUM_PCSD)) {
1725 /* Enable IPv4 payload checksum for UDP fragments
1726 * if PCSD is not set */
1727 rxcsum |= IXGBE_RXCSUM_IPPCSE;
1728 }
1729
1730 IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
9a799d71
AK
1731}
1732
1733static void ixgbe_vlan_rx_register(struct net_device *netdev,
b4617240 1734 struct vlan_group *grp)
9a799d71
AK
1735{
1736 struct ixgbe_adapter *adapter = netdev_priv(netdev);
1737 u32 ctrl;
1738
d4f80882
AV
1739 if (!test_bit(__IXGBE_DOWN, &adapter->state))
1740 ixgbe_irq_disable(adapter);
9a799d71
AK
1741 adapter->vlgrp = grp;
1742
2f90b865
AD
1743 /*
1744 * For a DCB driver, always enable VLAN tag stripping so we can
1745 * still receive traffic from a DCB-enabled host even if we're
1746 * not in DCB mode.
1747 */
1748 ctrl = IXGBE_READ_REG(&adapter->hw, IXGBE_VLNCTRL);
1749 ctrl |= IXGBE_VLNCTRL_VME;
1750 ctrl &= ~IXGBE_VLNCTRL_CFIEN;
1751 IXGBE_WRITE_REG(&adapter->hw, IXGBE_VLNCTRL, ctrl);
1752
9a799d71
AK
1753 if (grp) {
1754 /* enable VLAN tag insert/strip */
1755 ctrl = IXGBE_READ_REG(&adapter->hw, IXGBE_VLNCTRL);
746b9f02 1756 ctrl |= IXGBE_VLNCTRL_VME;
9a799d71
AK
1757 ctrl &= ~IXGBE_VLNCTRL_CFIEN;
1758 IXGBE_WRITE_REG(&adapter->hw, IXGBE_VLNCTRL, ctrl);
1759 }
1760
d4f80882
AV
1761 if (!test_bit(__IXGBE_DOWN, &adapter->state))
1762 ixgbe_irq_enable(adapter);
9a799d71
AK
1763}
1764
1765static void ixgbe_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
1766{
1767 struct ixgbe_adapter *adapter = netdev_priv(netdev);
c44ade9e 1768 struct ixgbe_hw *hw = &adapter->hw;
9a799d71
AK
1769
1770 /* add VID to filter table */
c44ade9e 1771 hw->mac.ops.set_vfta(&adapter->hw, vid, 0, true);
9a799d71
AK
1772}
1773
1774static void ixgbe_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
1775{
1776 struct ixgbe_adapter *adapter = netdev_priv(netdev);
c44ade9e 1777 struct ixgbe_hw *hw = &adapter->hw;
9a799d71 1778
d4f80882
AV
1779 if (!test_bit(__IXGBE_DOWN, &adapter->state))
1780 ixgbe_irq_disable(adapter);
1781
9a799d71 1782 vlan_group_set_device(adapter->vlgrp, vid, NULL);
d4f80882
AV
1783
1784 if (!test_bit(__IXGBE_DOWN, &adapter->state))
1785 ixgbe_irq_enable(adapter);
9a799d71
AK
1786
1787 /* remove VID from filter table */
c44ade9e 1788 hw->mac.ops.set_vfta(&adapter->hw, vid, 0, false);
9a799d71
AK
1789}
1790
1791static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
1792{
1793 ixgbe_vlan_rx_register(adapter->netdev, adapter->vlgrp);
1794
1795 if (adapter->vlgrp) {
1796 u16 vid;
1797 for (vid = 0; vid < VLAN_GROUP_ARRAY_LEN; vid++) {
1798 if (!vlan_group_get_device(adapter->vlgrp, vid))
1799 continue;
1800 ixgbe_vlan_rx_add_vid(adapter->netdev, vid);
1801 }
1802 }
1803}
1804
2c5645cf
CL
1805static u8 *ixgbe_addr_list_itr(struct ixgbe_hw *hw, u8 **mc_addr_ptr, u32 *vmdq)
1806{
1807 struct dev_mc_list *mc_ptr;
1808 u8 *addr = *mc_addr_ptr;
1809 *vmdq = 0;
1810
1811 mc_ptr = container_of(addr, struct dev_mc_list, dmi_addr[0]);
1812 if (mc_ptr->next)
1813 *mc_addr_ptr = mc_ptr->next->dmi_addr;
1814 else
1815 *mc_addr_ptr = NULL;
1816
1817 return addr;
1818}
1819
9a799d71 1820/**
2c5645cf 1821 * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
9a799d71
AK
1822 * @netdev: network interface device structure
1823 *
2c5645cf
CL
1824 * The set_rx_method entry point is called whenever the unicast/multicast
1825 * address list or the network interface flags are updated. This routine is
1826 * responsible for configuring the hardware for proper unicast, multicast and
1827 * promiscuous mode.
9a799d71 1828 **/
2c5645cf 1829static void ixgbe_set_rx_mode(struct net_device *netdev)
9a799d71
AK
1830{
1831 struct ixgbe_adapter *adapter = netdev_priv(netdev);
1832 struct ixgbe_hw *hw = &adapter->hw;
3d01625a 1833 u32 fctrl, vlnctrl;
2c5645cf
CL
1834 u8 *addr_list = NULL;
1835 int addr_count = 0;
9a799d71
AK
1836
1837 /* Check for Promiscuous and All Multicast modes */
1838
1839 fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3d01625a 1840 vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
9a799d71
AK
1841
1842 if (netdev->flags & IFF_PROMISC) {
2c5645cf 1843 hw->addr_ctrl.user_set_promisc = 1;
9a799d71 1844 fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3d01625a 1845 vlnctrl &= ~IXGBE_VLNCTRL_VFE;
9a799d71 1846 } else {
746b9f02
PM
1847 if (netdev->flags & IFF_ALLMULTI) {
1848 fctrl |= IXGBE_FCTRL_MPE;
1849 fctrl &= ~IXGBE_FCTRL_UPE;
1850 } else {
1851 fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
1852 }
3d01625a 1853 vlnctrl |= IXGBE_VLNCTRL_VFE;
2c5645cf 1854 hw->addr_ctrl.user_set_promisc = 0;
9a799d71
AK
1855 }
1856
1857 IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3d01625a 1858 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
9a799d71 1859
2c5645cf
CL
1860 /* reprogram secondary unicast list */
1861 addr_count = netdev->uc_count;
1862 if (addr_count)
1863 addr_list = netdev->uc_list->dmi_addr;
c44ade9e
JB
1864 hw->mac.ops.update_uc_addr_list(hw, addr_list, addr_count,
1865 ixgbe_addr_list_itr);
9a799d71 1866
2c5645cf
CL
1867 /* reprogram multicast list */
1868 addr_count = netdev->mc_count;
1869 if (addr_count)
1870 addr_list = netdev->mc_list->dmi_addr;
c44ade9e
JB
1871 hw->mac.ops.update_mc_addr_list(hw, addr_list, addr_count,
1872 ixgbe_addr_list_itr);
9a799d71
AK
1873}
1874
021230d4
AV
1875static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
1876{
1877 int q_idx;
1878 struct ixgbe_q_vector *q_vector;
1879 int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
1880
1881 /* legacy and MSI only use one vector */
1882 if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED))
1883 q_vectors = 1;
1884
1885 for (q_idx = 0; q_idx < q_vectors; q_idx++) {
f0848276 1886 struct napi_struct *napi;
021230d4
AV
1887 q_vector = &adapter->q_vector[q_idx];
1888 if (!q_vector->rxr_count)
1889 continue;
f0848276
JB
1890 napi = &q_vector->napi;
1891 if ((adapter->flags & IXGBE_FLAG_MSIX_ENABLED) &&
1892 (q_vector->rxr_count > 1))
1893 napi->poll = &ixgbe_clean_rxonly_many;
1894
1895 napi_enable(napi);
021230d4
AV
1896 }
1897}
1898
1899static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
1900{
1901 int q_idx;
1902 struct ixgbe_q_vector *q_vector;
1903 int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
1904
1905 /* legacy and MSI only use one vector */
1906 if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED))
1907 q_vectors = 1;
1908
1909 for (q_idx = 0; q_idx < q_vectors; q_idx++) {
1910 q_vector = &adapter->q_vector[q_idx];
1911 if (!q_vector->rxr_count)
1912 continue;
1913 napi_disable(&q_vector->napi);
1914 }
1915}
1916
2f90b865
AD
1917#ifdef CONFIG_IXGBE_DCBNL
1918/*
1919 * ixgbe_configure_dcb - Configure DCB hardware
1920 * @adapter: ixgbe adapter struct
1921 *
1922 * This is called by the driver on open to configure the DCB hardware.
1923 * This is also called by the gennetlink interface when reconfiguring
1924 * the DCB state.
1925 */
1926static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
1927{
1928 struct ixgbe_hw *hw = &adapter->hw;
1929 u32 txdctl, vlnctrl;
1930 int i, j;
1931
1932 ixgbe_dcb_check_config(&adapter->dcb_cfg);
1933 ixgbe_dcb_calculate_tc_credits(&adapter->dcb_cfg, DCB_TX_CONFIG);
1934 ixgbe_dcb_calculate_tc_credits(&adapter->dcb_cfg, DCB_RX_CONFIG);
1935
1936 /* reconfigure the hardware */
1937 ixgbe_dcb_hw_config(&adapter->hw, &adapter->dcb_cfg);
1938
1939 for (i = 0; i < adapter->num_tx_queues; i++) {
1940 j = adapter->tx_ring[i].reg_idx;
1941 txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(j));
1942 /* PThresh workaround for Tx hang with DFP enabled. */
1943 txdctl |= 32;
1944 IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(j), txdctl);
1945 }
1946 /* Enable VLAN tag insert/strip */
1947 vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1948 vlnctrl |= IXGBE_VLNCTRL_VME | IXGBE_VLNCTRL_VFE;
1949 vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1950 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1951 hw->mac.ops.set_vfta(&adapter->hw, 0, 0, true);
1952}
1953
1954#endif
9a799d71
AK
1955static void ixgbe_configure(struct ixgbe_adapter *adapter)
1956{
1957 struct net_device *netdev = adapter->netdev;
1958 int i;
1959
2c5645cf 1960 ixgbe_set_rx_mode(netdev);
9a799d71
AK
1961
1962 ixgbe_restore_vlan(adapter);
2f90b865
AD
1963#ifdef CONFIG_IXGBE_DCBNL
1964 if (adapter->flags & IXGBE_FLAG_DCB_ENABLED) {
1965 netif_set_gso_max_size(netdev, 32768);
1966 ixgbe_configure_dcb(adapter);
1967 } else {
1968 netif_set_gso_max_size(netdev, 65536);
1969 }
1970#else
1971 netif_set_gso_max_size(netdev, 65536);
1972#endif
9a799d71
AK
1973
1974 ixgbe_configure_tx(adapter);
1975 ixgbe_configure_rx(adapter);
1976 for (i = 0; i < adapter->num_rx_queues; i++)
1977 ixgbe_alloc_rx_buffers(adapter, &adapter->rx_ring[i],
b4617240 1978 (adapter->rx_ring[i].count - 1));
9a799d71
AK
1979}
1980
1981static int ixgbe_up_complete(struct ixgbe_adapter *adapter)
1982{
1983 struct net_device *netdev = adapter->netdev;
9a799d71 1984 struct ixgbe_hw *hw = &adapter->hw;
021230d4 1985 int i, j = 0;
9a799d71 1986 int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
021230d4
AV
1987 u32 txdctl, rxdctl, mhadd;
1988 u32 gpie;
9a799d71 1989
5eba3699
AV
1990 ixgbe_get_hw_control(adapter);
1991
021230d4
AV
1992 if ((adapter->flags & IXGBE_FLAG_MSIX_ENABLED) ||
1993 (adapter->flags & IXGBE_FLAG_MSI_ENABLED)) {
9a799d71
AK
1994 if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
1995 gpie = (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_EIAME |
b4617240 1996 IXGBE_GPIE_PBA_SUPPORT | IXGBE_GPIE_OCD);
9a799d71
AK
1997 } else {
1998 /* MSI only */
021230d4 1999 gpie = 0;
9a799d71 2000 }
021230d4
AV
2001 /* XXX: to interrupt immediately for EICS writes, enable this */
2002 /* gpie |= IXGBE_GPIE_EIMEN; */
2003 IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
9a799d71
AK
2004 }
2005
021230d4
AV
2006 if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2007 /* legacy interrupts, use EIAM to auto-mask when reading EICR,
2008 * specifically only auto mask tx and rx interrupts */
2009 IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
2010 }
9a799d71 2011
0befdb3e
JB
2012 /* Enable fan failure interrupt if media type is copper */
2013 if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
2014 gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2015 gpie |= IXGBE_SDP1_GPIEN;
2016 IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2017 }
2018
021230d4 2019 mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
9a799d71
AK
2020 if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
2021 mhadd &= ~IXGBE_MHADD_MFS_MASK;
2022 mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
2023
2024 IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
2025 }
2026
2027 for (i = 0; i < adapter->num_tx_queues; i++) {
021230d4
AV
2028 j = adapter->tx_ring[i].reg_idx;
2029 txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(j));
e01c31a5
JB
2030 /* enable WTHRESH=8 descriptors, to encourage burst writeback */
2031 txdctl |= (8 << 16);
9a799d71 2032 txdctl |= IXGBE_TXDCTL_ENABLE;
021230d4 2033 IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(j), txdctl);
9a799d71
AK
2034 }
2035
2036 for (i = 0; i < adapter->num_rx_queues; i++) {
021230d4
AV
2037 j = adapter->rx_ring[i].reg_idx;
2038 rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
2039 /* enable PTHRESH=32 descriptors (half the internal cache)
2040 * and HTHRESH=0 descriptors (to minimize latency on fetch),
2041 * this also removes a pesky rx_no_buffer_count increment */
2042 rxdctl |= 0x0020;
9a799d71 2043 rxdctl |= IXGBE_RXDCTL_ENABLE;
021230d4 2044 IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), rxdctl);
9a799d71
AK
2045 }
2046 /* enable all receives */
2047 rxdctl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
2048 rxdctl |= (IXGBE_RXCTRL_DMBYPS | IXGBE_RXCTRL_RXEN);
2049 IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxdctl);
2050
2051 if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2052 ixgbe_configure_msix(adapter);
2053 else
2054 ixgbe_configure_msi_and_legacy(adapter);
2055
2056 clear_bit(__IXGBE_DOWN, &adapter->state);
021230d4
AV
2057 ixgbe_napi_enable_all(adapter);
2058
2059 /* clear any pending interrupts, may auto mask */
2060 IXGBE_READ_REG(hw, IXGBE_EICR);
2061
9a799d71
AK
2062 ixgbe_irq_enable(adapter);
2063
2064 /* bring the link up in the watchdog, this could race with our first
2065 * link up interrupt but shouldn't be a problem */
cf8280ee
JB
2066 adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2067 adapter->link_check_timeout = jiffies;
9a799d71
AK
2068 mod_timer(&adapter->watchdog_timer, jiffies);
2069 return 0;
2070}
2071
d4f80882
AV
2072void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
2073{
2074 WARN_ON(in_interrupt());
2075 while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
2076 msleep(1);
2077 ixgbe_down(adapter);
2078 ixgbe_up(adapter);
2079 clear_bit(__IXGBE_RESETTING, &adapter->state);
2080}
2081
9a799d71
AK
2082int ixgbe_up(struct ixgbe_adapter *adapter)
2083{
2084 /* hardware has been reset, we need to reload some things */
2085 ixgbe_configure(adapter);
2086
2087 return ixgbe_up_complete(adapter);
2088}
2089
2090void ixgbe_reset(struct ixgbe_adapter *adapter)
2091{
c44ade9e
JB
2092 struct ixgbe_hw *hw = &adapter->hw;
2093 if (hw->mac.ops.init_hw(hw))
2094 dev_err(&adapter->pdev->dev, "Hardware Error\n");
9a799d71
AK
2095
2096 /* reprogram the RAR[0] in case user changed it. */
c44ade9e 2097 hw->mac.ops.set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
9a799d71
AK
2098
2099}
2100
9a799d71
AK
2101/**
2102 * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
2103 * @adapter: board private structure
2104 * @rx_ring: ring to free buffers from
2105 **/
2106static void ixgbe_clean_rx_ring(struct ixgbe_adapter *adapter,
b4617240 2107 struct ixgbe_ring *rx_ring)
9a799d71
AK
2108{
2109 struct pci_dev *pdev = adapter->pdev;
2110 unsigned long size;
2111 unsigned int i;
2112
2113 /* Free all the Rx ring sk_buffs */
2114
2115 for (i = 0; i < rx_ring->count; i++) {
2116 struct ixgbe_rx_buffer *rx_buffer_info;
2117
2118 rx_buffer_info = &rx_ring->rx_buffer_info[i];
2119 if (rx_buffer_info->dma) {
2120 pci_unmap_single(pdev, rx_buffer_info->dma,
b4617240
PW
2121 rx_ring->rx_buf_len,
2122 PCI_DMA_FROMDEVICE);
9a799d71
AK
2123 rx_buffer_info->dma = 0;
2124 }
2125 if (rx_buffer_info->skb) {
2126 dev_kfree_skb(rx_buffer_info->skb);
2127 rx_buffer_info->skb = NULL;
2128 }
2129 if (!rx_buffer_info->page)
2130 continue;
762f4c57
JB
2131 pci_unmap_page(pdev, rx_buffer_info->page_dma, PAGE_SIZE / 2,
2132 PCI_DMA_FROMDEVICE);
9a799d71 2133 rx_buffer_info->page_dma = 0;
9a799d71
AK
2134 put_page(rx_buffer_info->page);
2135 rx_buffer_info->page = NULL;
762f4c57 2136 rx_buffer_info->page_offset = 0;
9a799d71
AK
2137 }
2138
2139 size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
2140 memset(rx_ring->rx_buffer_info, 0, size);
2141
2142 /* Zero out the descriptor ring */
2143 memset(rx_ring->desc, 0, rx_ring->size);
2144
2145 rx_ring->next_to_clean = 0;
2146 rx_ring->next_to_use = 0;
2147
2148 writel(0, adapter->hw.hw_addr + rx_ring->head);
2149 writel(0, adapter->hw.hw_addr + rx_ring->tail);
2150}
2151
2152/**
2153 * ixgbe_clean_tx_ring - Free Tx Buffers
2154 * @adapter: board private structure
2155 * @tx_ring: ring to be cleaned
2156 **/
2157static void ixgbe_clean_tx_ring(struct ixgbe_adapter *adapter,
b4617240 2158 struct ixgbe_ring *tx_ring)
9a799d71
AK
2159{
2160 struct ixgbe_tx_buffer *tx_buffer_info;
2161 unsigned long size;
2162 unsigned int i;
2163
2164 /* Free all the Tx ring sk_buffs */
2165
2166 for (i = 0; i < tx_ring->count; i++) {
2167 tx_buffer_info = &tx_ring->tx_buffer_info[i];
2168 ixgbe_unmap_and_free_tx_resource(adapter, tx_buffer_info);
2169 }
2170
2171 size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
2172 memset(tx_ring->tx_buffer_info, 0, size);
2173
2174 /* Zero out the descriptor ring */
2175 memset(tx_ring->desc, 0, tx_ring->size);
2176
2177 tx_ring->next_to_use = 0;
2178 tx_ring->next_to_clean = 0;
2179
2180 writel(0, adapter->hw.hw_addr + tx_ring->head);
2181 writel(0, adapter->hw.hw_addr + tx_ring->tail);
2182}
2183
2184/**
021230d4 2185 * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
9a799d71
AK
2186 * @adapter: board private structure
2187 **/
021230d4 2188static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
9a799d71
AK
2189{
2190 int i;
2191
021230d4
AV
2192 for (i = 0; i < adapter->num_rx_queues; i++)
2193 ixgbe_clean_rx_ring(adapter, &adapter->rx_ring[i]);
9a799d71
AK
2194}
2195
2196/**
021230d4 2197 * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
9a799d71
AK
2198 * @adapter: board private structure
2199 **/
021230d4 2200static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
9a799d71
AK
2201{
2202 int i;
2203
021230d4
AV
2204 for (i = 0; i < adapter->num_tx_queues; i++)
2205 ixgbe_clean_tx_ring(adapter, &adapter->tx_ring[i]);
9a799d71
AK
2206}
2207
2208void ixgbe_down(struct ixgbe_adapter *adapter)
2209{
2210 struct net_device *netdev = adapter->netdev;
7f821875 2211 struct ixgbe_hw *hw = &adapter->hw;
9a799d71 2212 u32 rxctrl;
7f821875
JB
2213 u32 txdctl;
2214 int i, j;
9a799d71
AK
2215
2216 /* signal that we are down to the interrupt handler */
2217 set_bit(__IXGBE_DOWN, &adapter->state);
2218
2219 /* disable receives */
7f821875
JB
2220 rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
2221 IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
9a799d71
AK
2222
2223 netif_tx_disable(netdev);
2224
7f821875 2225 IXGBE_WRITE_FLUSH(hw);
9a799d71
AK
2226 msleep(10);
2227
7f821875
JB
2228 netif_tx_stop_all_queues(netdev);
2229
9a799d71
AK
2230 ixgbe_irq_disable(adapter);
2231
021230d4 2232 ixgbe_napi_disable_all(adapter);
7f821875 2233
9a799d71 2234 del_timer_sync(&adapter->watchdog_timer);
cf8280ee 2235 cancel_work_sync(&adapter->watchdog_task);
9a799d71 2236
7f821875
JB
2237 /* disable transmits in the hardware now that interrupts are off */
2238 for (i = 0; i < adapter->num_tx_queues; i++) {
2239 j = adapter->tx_ring[i].reg_idx;
2240 txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(j));
2241 IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(j),
2242 (txdctl & ~IXGBE_TXDCTL_ENABLE));
2243 }
2244
9a799d71 2245 netif_carrier_off(netdev);
9a799d71 2246
5dd2d332 2247#ifdef CONFIG_IXGBE_DCA
96b0e0f6
JB
2248 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
2249 adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
2250 dca_remove_requester(&adapter->pdev->dev);
2251 }
2252
2253#endif
6f4a0e45
PL
2254 if (!pci_channel_offline(adapter->pdev))
2255 ixgbe_reset(adapter);
9a799d71
AK
2256 ixgbe_clean_all_tx_rings(adapter);
2257 ixgbe_clean_all_rx_rings(adapter);
2258
5dd2d332 2259#ifdef CONFIG_IXGBE_DCA
96b0e0f6
JB
2260 /* since we reset the hardware DCA settings were cleared */
2261 if (dca_add_requester(&adapter->pdev->dev) == 0) {
2262 adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
2263 /* always use CB2 mode, difference is masked
2264 * in the CB driver */
b4617240 2265 IXGBE_WRITE_REG(hw, IXGBE_DCA_CTRL, 2);
96b0e0f6
JB
2266 ixgbe_setup_dca(adapter);
2267 }
2268#endif
9a799d71
AK
2269}
2270
9a799d71 2271/**
021230d4
AV
2272 * ixgbe_poll - NAPI Rx polling callback
2273 * @napi: structure for representing this polling device
2274 * @budget: how many packets driver is allowed to clean
2275 *
2276 * This function is used for legacy and MSI, NAPI mode
9a799d71 2277 **/
021230d4 2278static int ixgbe_poll(struct napi_struct *napi, int budget)
9a799d71 2279{
021230d4 2280 struct ixgbe_q_vector *q_vector = container_of(napi,
b4617240 2281 struct ixgbe_q_vector, napi);
021230d4 2282 struct ixgbe_adapter *adapter = q_vector->adapter;
74ce8dd2 2283 int tx_cleaned, work_done = 0;
9a799d71 2284
5dd2d332 2285#ifdef CONFIG_IXGBE_DCA
bd0362dd
JC
2286 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
2287 ixgbe_update_tx_dca(adapter, adapter->tx_ring);
2288 ixgbe_update_rx_dca(adapter, adapter->rx_ring);
2289 }
2290#endif
2291
d2c7ddd6 2292 tx_cleaned = ixgbe_clean_tx_irq(adapter, adapter->tx_ring);
021230d4 2293 ixgbe_clean_rx_irq(adapter, adapter->rx_ring, &work_done, budget);
9a799d71 2294
d2c7ddd6
DM
2295 if (tx_cleaned)
2296 work_done = budget;
2297
53e52c72
DM
2298 /* If budget not fully consumed, exit the polling mode */
2299 if (work_done < budget) {
021230d4 2300 netif_rx_complete(adapter->netdev, napi);
30efa5a3 2301 if (adapter->itr_setting & 3)
f494e8fa 2302 ixgbe_set_itr(adapter);
d4f80882
AV
2303 if (!test_bit(__IXGBE_DOWN, &adapter->state))
2304 ixgbe_irq_enable(adapter);
9a799d71 2305 }
9a799d71
AK
2306 return work_done;
2307}
2308
2309/**
2310 * ixgbe_tx_timeout - Respond to a Tx Hang
2311 * @netdev: network interface device structure
2312 **/
2313static void ixgbe_tx_timeout(struct net_device *netdev)
2314{
2315 struct ixgbe_adapter *adapter = netdev_priv(netdev);
2316
2317 /* Do the reset outside of interrupt context */
2318 schedule_work(&adapter->reset_task);
2319}
2320
2321static void ixgbe_reset_task(struct work_struct *work)
2322{
2323 struct ixgbe_adapter *adapter;
2324 adapter = container_of(work, struct ixgbe_adapter, reset_task);
2325
2f90b865
AD
2326 /* If we're already down or resetting, just bail */
2327 if (test_bit(__IXGBE_DOWN, &adapter->state) ||
2328 test_bit(__IXGBE_RESETTING, &adapter->state))
2329 return;
2330
9a799d71
AK
2331 adapter->tx_timeout_count++;
2332
d4f80882 2333 ixgbe_reinit_locked(adapter);
9a799d71
AK
2334}
2335
b9804972
JB
2336static void ixgbe_set_num_queues(struct ixgbe_adapter *adapter)
2337{
2338 int nrq = 1, ntq = 1;
2339 int feature_mask = 0, rss_i, rss_m;
2f90b865 2340 int dcb_i, dcb_m;
b9804972
JB
2341
2342 /* Number of supported queues */
2343 switch (adapter->hw.mac.type) {
2344 case ixgbe_mac_82598EB:
2f90b865
AD
2345 dcb_i = adapter->ring_feature[RING_F_DCB].indices;
2346 dcb_m = 0;
b9804972
JB
2347 rss_i = adapter->ring_feature[RING_F_RSS].indices;
2348 rss_m = 0;
2349 feature_mask |= IXGBE_FLAG_RSS_ENABLED;
2f90b865 2350 feature_mask |= IXGBE_FLAG_DCB_ENABLED;
b9804972
JB
2351
2352 switch (adapter->flags & feature_mask) {
2f90b865
AD
2353 case (IXGBE_FLAG_RSS_ENABLED | IXGBE_FLAG_DCB_ENABLED):
2354 dcb_m = 0x7 << 3;
2355 rss_i = min(8, rss_i);
2356 rss_m = 0x7;
2357 nrq = dcb_i * rss_i;
2358 ntq = min(MAX_TX_QUEUES, dcb_i * rss_i);
2359 break;
2360 case (IXGBE_FLAG_DCB_ENABLED):
2361 dcb_m = 0x7 << 3;
2362 nrq = dcb_i;
2363 ntq = dcb_i;
2364 break;
b9804972
JB
2365 case (IXGBE_FLAG_RSS_ENABLED):
2366 rss_m = 0xF;
2367 nrq = rss_i;
2368 ntq = rss_i;
2369 break;
2370 case 0:
2371 default:
2f90b865
AD
2372 dcb_i = 0;
2373 dcb_m = 0;
b9804972
JB
2374 rss_i = 0;
2375 rss_m = 0;
2376 nrq = 1;
2377 ntq = 1;
2378 break;
2379 }
2380
2f90b865
AD
2381 /* Sanity check, we should never have zero queues */
2382 nrq = (nrq ?:1);
2383 ntq = (ntq ?:1);
2384
2385 adapter->ring_feature[RING_F_DCB].indices = dcb_i;
2386 adapter->ring_feature[RING_F_DCB].mask = dcb_m;
b9804972
JB
2387 adapter->ring_feature[RING_F_RSS].indices = rss_i;
2388 adapter->ring_feature[RING_F_RSS].mask = rss_m;
2389 break;
2390 default:
2391 nrq = 1;
2392 ntq = 1;
2393 break;
2394 }
2395
2396 adapter->num_rx_queues = nrq;
2397 adapter->num_tx_queues = ntq;
2398}
2399
021230d4 2400static void ixgbe_acquire_msix_vectors(struct ixgbe_adapter *adapter,
b4617240 2401 int vectors)
021230d4
AV
2402{
2403 int err, vector_threshold;
2404
2405 /* We'll want at least 3 (vector_threshold):
2406 * 1) TxQ[0] Cleanup
2407 * 2) RxQ[0] Cleanup
2408 * 3) Other (Link Status Change, etc.)
2409 * 4) TCP Timer (optional)
2410 */
2411 vector_threshold = MIN_MSIX_COUNT;
2412
2413 /* The more we get, the more we will assign to Tx/Rx Cleanup
2414 * for the separate queues...where Rx Cleanup >= Tx Cleanup.
2415 * Right now, we simply care about how many we'll get; we'll
2416 * set them up later while requesting irq's.
2417 */
2418 while (vectors >= vector_threshold) {
2419 err = pci_enable_msix(adapter->pdev, adapter->msix_entries,
b4617240 2420 vectors);
021230d4
AV
2421 if (!err) /* Success in acquiring all requested vectors. */
2422 break;
2423 else if (err < 0)
2424 vectors = 0; /* Nasty failure, quit now */
2425 else /* err == number of vectors we should try again with */
2426 vectors = err;
2427 }
2428
2429 if (vectors < vector_threshold) {
2430 /* Can't allocate enough MSI-X interrupts? Oh well.
2431 * This just means we'll go with either a single MSI
2432 * vector or fall back to legacy interrupts.
2433 */
2434 DPRINTK(HW, DEBUG, "Unable to allocate MSI-X interrupts\n");
2435 adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2436 kfree(adapter->msix_entries);
2437 adapter->msix_entries = NULL;
2f90b865 2438 adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
021230d4 2439 adapter->flags &= ~IXGBE_FLAG_RSS_ENABLED;
b9804972 2440 ixgbe_set_num_queues(adapter);
021230d4
AV
2441 } else {
2442 adapter->flags |= IXGBE_FLAG_MSIX_ENABLED; /* Woot! */
2443 adapter->num_msix_vectors = vectors;
2444 }
2445}
2446
021230d4
AV
2447/**
2448 * ixgbe_cache_ring_register - Descriptor ring to register mapping
2449 * @adapter: board private structure to initialize
2450 *
2451 * Once we know the feature-set enabled for the device, we'll cache
2452 * the register offset the descriptor ring is assigned to.
2453 **/
2454static void __devinit ixgbe_cache_ring_register(struct ixgbe_adapter *adapter)
2455{
021230d4
AV
2456 int feature_mask = 0, rss_i;
2457 int i, txr_idx, rxr_idx;
2f90b865 2458 int dcb_i;
021230d4
AV
2459
2460 /* Number of supported queues */
2461 switch (adapter->hw.mac.type) {
2462 case ixgbe_mac_82598EB:
2f90b865 2463 dcb_i = adapter->ring_feature[RING_F_DCB].indices;
021230d4
AV
2464 rss_i = adapter->ring_feature[RING_F_RSS].indices;
2465 txr_idx = 0;
2466 rxr_idx = 0;
2f90b865 2467 feature_mask |= IXGBE_FLAG_DCB_ENABLED;
021230d4
AV
2468 feature_mask |= IXGBE_FLAG_RSS_ENABLED;
2469 switch (adapter->flags & feature_mask) {
2f90b865
AD
2470 case (IXGBE_FLAG_RSS_ENABLED | IXGBE_FLAG_DCB_ENABLED):
2471 for (i = 0; i < dcb_i; i++) {
2472 int j;
2473 /* Rx first */
2474 for (j = 0; j < adapter->num_rx_queues; j++) {
2475 adapter->rx_ring[rxr_idx].reg_idx =
2476 i << 3 | j;
2477 rxr_idx++;
2478 }
2479 /* Tx now */
2480 for (j = 0; j < adapter->num_tx_queues; j++) {
2481 adapter->tx_ring[txr_idx].reg_idx =
2482 i << 2 | (j >> 1);
2483 if (j & 1)
2484 txr_idx++;
2485 }
2486 }
2487 case (IXGBE_FLAG_DCB_ENABLED):
2488 /* the number of queues is assumed to be symmetric */
2489 for (i = 0; i < dcb_i; i++) {
2490 adapter->rx_ring[i].reg_idx = i << 3;
2491 adapter->tx_ring[i].reg_idx = i << 2;
2492 }
2493 break;
021230d4
AV
2494 case (IXGBE_FLAG_RSS_ENABLED):
2495 for (i = 0; i < adapter->num_rx_queues; i++)
2496 adapter->rx_ring[i].reg_idx = i;
2497 for (i = 0; i < adapter->num_tx_queues; i++)
2498 adapter->tx_ring[i].reg_idx = i;
2499 break;
2500 case 0:
2501 default:
2502 break;
2503 }
2504 break;
2505 default:
2506 break;
2507 }
2508}
2509
9a799d71
AK
2510/**
2511 * ixgbe_alloc_queues - Allocate memory for all rings
2512 * @adapter: board private structure to initialize
2513 *
2514 * We allocate one ring per queue at run-time since we don't know the
2515 * number of queues at compile-time. The polling_netdev array is
2516 * intended for Multiqueue, but should work fine with a single queue.
2517 **/
2f90b865 2518static int ixgbe_alloc_queues(struct ixgbe_adapter *adapter)
9a799d71
AK
2519{
2520 int i;
2521
2522 adapter->tx_ring = kcalloc(adapter->num_tx_queues,
b4617240 2523 sizeof(struct ixgbe_ring), GFP_KERNEL);
9a799d71 2524 if (!adapter->tx_ring)
021230d4 2525 goto err_tx_ring_allocation;
9a799d71
AK
2526
2527 adapter->rx_ring = kcalloc(adapter->num_rx_queues,
b4617240 2528 sizeof(struct ixgbe_ring), GFP_KERNEL);
021230d4
AV
2529 if (!adapter->rx_ring)
2530 goto err_rx_ring_allocation;
9a799d71 2531
021230d4 2532 for (i = 0; i < adapter->num_tx_queues; i++) {
b9804972 2533 adapter->tx_ring[i].count = adapter->tx_ring_count;
021230d4
AV
2534 adapter->tx_ring[i].queue_index = i;
2535 }
b9804972 2536
9a799d71 2537 for (i = 0; i < adapter->num_rx_queues; i++) {
b9804972 2538 adapter->rx_ring[i].count = adapter->rx_ring_count;
021230d4
AV
2539 adapter->rx_ring[i].queue_index = i;
2540 }
2541
2542 ixgbe_cache_ring_register(adapter);
2543
2544 return 0;
2545
2546err_rx_ring_allocation:
2547 kfree(adapter->tx_ring);
2548err_tx_ring_allocation:
2549 return -ENOMEM;
2550}
2551
2552/**
2553 * ixgbe_set_interrupt_capability - set MSI-X or MSI if supported
2554 * @adapter: board private structure to initialize
2555 *
2556 * Attempt to configure the interrupts using the best available
2557 * capabilities of the hardware and the kernel.
2558 **/
2559static int __devinit ixgbe_set_interrupt_capability(struct ixgbe_adapter
b4617240 2560 *adapter)
021230d4
AV
2561{
2562 int err = 0;
2563 int vector, v_budget;
2564
2565 /*
2566 * It's easy to be greedy for MSI-X vectors, but it really
2567 * doesn't do us much good if we have a lot more vectors
2568 * than CPU's. So let's be conservative and only ask for
2569 * (roughly) twice the number of vectors as there are CPU's.
2570 */
2571 v_budget = min(adapter->num_rx_queues + adapter->num_tx_queues,
b4617240 2572 (int)(num_online_cpus() * 2)) + NON_Q_VECTORS;
021230d4
AV
2573
2574 /*
2575 * At the same time, hardware can only support a maximum of
2576 * MAX_MSIX_COUNT vectors. With features such as RSS and VMDq,
2577 * we can easily reach upwards of 64 Rx descriptor queues and
2578 * 32 Tx queues. Thus, we cap it off in those rare cases where
2579 * the cpu count also exceeds our vector limit.
2580 */
2581 v_budget = min(v_budget, MAX_MSIX_COUNT);
2582
2583 /* A failure in MSI-X entry allocation isn't fatal, but it does
2584 * mean we disable MSI-X capabilities of the adapter. */
2585 adapter->msix_entries = kcalloc(v_budget,
b4617240 2586 sizeof(struct msix_entry), GFP_KERNEL);
021230d4 2587 if (!adapter->msix_entries) {
2f90b865 2588 adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
021230d4
AV
2589 adapter->flags &= ~IXGBE_FLAG_RSS_ENABLED;
2590 ixgbe_set_num_queues(adapter);
2591 kfree(adapter->tx_ring);
2592 kfree(adapter->rx_ring);
2593 err = ixgbe_alloc_queues(adapter);
2594 if (err) {
2595 DPRINTK(PROBE, ERR, "Unable to allocate memory "
b4617240 2596 "for queues\n");
021230d4
AV
2597 goto out;
2598 }
2599
2600 goto try_msi;
2601 }
2602
2603 for (vector = 0; vector < v_budget; vector++)
2604 adapter->msix_entries[vector].entry = vector;
2605
2606 ixgbe_acquire_msix_vectors(adapter, v_budget);
2607
2608 if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2609 goto out;
2610
2611try_msi:
2612 err = pci_enable_msi(adapter->pdev);
2613 if (!err) {
2614 adapter->flags |= IXGBE_FLAG_MSI_ENABLED;
2615 } else {
2616 DPRINTK(HW, DEBUG, "Unable to allocate MSI interrupt, "
b4617240 2617 "falling back to legacy. Error: %d\n", err);
021230d4
AV
2618 /* reset err */
2619 err = 0;
2620 }
2621
2622out:
30eba97a 2623 /* Notify the stack of the (possibly) reduced Tx Queue count. */
fd2ea0a7 2624 adapter->netdev->real_num_tx_queues = adapter->num_tx_queues;
021230d4
AV
2625
2626 return err;
2627}
2628
2f90b865 2629void ixgbe_reset_interrupt_capability(struct ixgbe_adapter *adapter)
021230d4
AV
2630{
2631 if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2632 adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2633 pci_disable_msix(adapter->pdev);
2634 kfree(adapter->msix_entries);
2635 adapter->msix_entries = NULL;
2636 } else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED) {
2637 adapter->flags &= ~IXGBE_FLAG_MSI_ENABLED;
2638 pci_disable_msi(adapter->pdev);
2639 }
2640 return;
2641}
2642
2643/**
2644 * ixgbe_init_interrupt_scheme - Determine proper interrupt scheme
2645 * @adapter: board private structure to initialize
2646 *
2647 * We determine which interrupt scheme to use based on...
2648 * - Kernel support (MSI, MSI-X)
2649 * - which can be user-defined (via MODULE_PARAM)
2650 * - Hardware queue count (num_*_queues)
2651 * - defined by miscellaneous hardware support/features (RSS, etc.)
2652 **/
2f90b865 2653int ixgbe_init_interrupt_scheme(struct ixgbe_adapter *adapter)
021230d4
AV
2654{
2655 int err;
2656
2657 /* Number of supported queues */
2658 ixgbe_set_num_queues(adapter);
2659
2660 err = ixgbe_alloc_queues(adapter);
2661 if (err) {
2662 DPRINTK(PROBE, ERR, "Unable to allocate memory for queues\n");
2663 goto err_alloc_queues;
2664 }
2665
2666 err = ixgbe_set_interrupt_capability(adapter);
2667 if (err) {
2668 DPRINTK(PROBE, ERR, "Unable to setup interrupt capabilities\n");
2669 goto err_set_interrupt;
9a799d71
AK
2670 }
2671
021230d4 2672 DPRINTK(DRV, INFO, "Multiqueue %s: Rx Queue count = %u, "
b4617240
PW
2673 "Tx Queue count = %u\n",
2674 (adapter->num_rx_queues > 1) ? "Enabled" :
2675 "Disabled", adapter->num_rx_queues, adapter->num_tx_queues);
021230d4
AV
2676
2677 set_bit(__IXGBE_DOWN, &adapter->state);
2678
9a799d71 2679 return 0;
021230d4
AV
2680
2681err_set_interrupt:
2682 kfree(adapter->tx_ring);
2683 kfree(adapter->rx_ring);
2684err_alloc_queues:
2685 return err;
9a799d71
AK
2686}
2687
c4900be0
DS
2688/**
2689 * ixgbe_sfp_timer - worker thread to find a missing module
2690 * @data: pointer to our adapter struct
2691 **/
2692static void ixgbe_sfp_timer(unsigned long data)
2693{
2694 struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
2695
2696 /* Do the sfp_timer outside of interrupt context due to the
2697 * delays that sfp+ detection requires
2698 */
2699 schedule_work(&adapter->sfp_task);
2700}
2701
2702/**
2703 * ixgbe_sfp_task - worker thread to find a missing module
2704 * @work: pointer to work_struct containing our data
2705 **/
2706static void ixgbe_sfp_task(struct work_struct *work)
2707{
2708 struct ixgbe_adapter *adapter = container_of(work,
2709 struct ixgbe_adapter,
2710 sfp_task);
2711 struct ixgbe_hw *hw = &adapter->hw;
2712
2713 if ((hw->phy.type == ixgbe_phy_nl) &&
2714 (hw->phy.sfp_type == ixgbe_sfp_type_not_present)) {
2715 s32 ret = hw->phy.ops.identify_sfp(hw);
2716 if (ret)
2717 goto reschedule;
2718 ret = hw->phy.ops.reset(hw);
2719 if (ret == IXGBE_ERR_SFP_NOT_SUPPORTED) {
2720 DPRINTK(PROBE, ERR, "failed to initialize because an "
2721 "unsupported SFP+ module type was detected.\n"
2722 "Reload the driver after installing a "
2723 "supported module.\n");
2724 unregister_netdev(adapter->netdev);
2725 } else {
2726 DPRINTK(PROBE, INFO, "detected SFP+: %d\n",
2727 hw->phy.sfp_type);
2728 }
2729 /* don't need this routine any more */
2730 clear_bit(__IXGBE_SFP_MODULE_NOT_FOUND, &adapter->state);
2731 }
2732 return;
2733reschedule:
2734 if (test_bit(__IXGBE_SFP_MODULE_NOT_FOUND, &adapter->state))
2735 mod_timer(&adapter->sfp_timer,
2736 round_jiffies(jiffies + (2 * HZ)));
2737}
2738
9a799d71
AK
2739/**
2740 * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
2741 * @adapter: board private structure to initialize
2742 *
2743 * ixgbe_sw_init initializes the Adapter private data structure.
2744 * Fields are initialized based on PCI device information and
2745 * OS network device settings (MTU size).
2746 **/
2747static int __devinit ixgbe_sw_init(struct ixgbe_adapter *adapter)
2748{
2749 struct ixgbe_hw *hw = &adapter->hw;
2750 struct pci_dev *pdev = adapter->pdev;
021230d4 2751 unsigned int rss;
2f90b865
AD
2752#ifdef CONFIG_IXGBE_DCBNL
2753 int j;
2754 struct tc_configuration *tc;
2755#endif
021230d4 2756
c44ade9e
JB
2757 /* PCI config space info */
2758
2759 hw->vendor_id = pdev->vendor;
2760 hw->device_id = pdev->device;
2761 hw->revision_id = pdev->revision;
2762 hw->subsystem_vendor_id = pdev->subsystem_vendor;
2763 hw->subsystem_device_id = pdev->subsystem_device;
2764
021230d4
AV
2765 /* Set capability flags */
2766 rss = min(IXGBE_MAX_RSS_INDICES, (int)num_online_cpus());
2767 adapter->ring_feature[RING_F_RSS].indices = rss;
2768 adapter->flags |= IXGBE_FLAG_RSS_ENABLED;
2f90b865
AD
2769 adapter->ring_feature[RING_F_DCB].indices = IXGBE_MAX_DCB_INDICES;
2770
2771#ifdef CONFIG_IXGBE_DCBNL
2772 /* Configure DCB traffic classes */
2773 for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
2774 tc = &adapter->dcb_cfg.tc_config[j];
2775 tc->path[DCB_TX_CONFIG].bwg_id = 0;
2776 tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
2777 tc->path[DCB_RX_CONFIG].bwg_id = 0;
2778 tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
2779 tc->dcb_pfc = pfc_disabled;
2780 }
2781 adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
2782 adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
2783 adapter->dcb_cfg.rx_pba_cfg = pba_equal;
2784 adapter->dcb_cfg.round_robin_enable = false;
2785 adapter->dcb_set_bitmap = 0x00;
2786 ixgbe_copy_dcb_cfg(&adapter->dcb_cfg, &adapter->temp_dcb_cfg,
2787 adapter->ring_feature[RING_F_DCB].indices);
2788
2789#endif
0befdb3e
JB
2790 if (hw->mac.ops.get_media_type &&
2791 (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper))
2792 adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
9a799d71
AK
2793
2794 /* default flow control settings */
2b9ade93
JB
2795 hw->fc.original_type = ixgbe_fc_none;
2796 hw->fc.type = ixgbe_fc_none;
2797 hw->fc.high_water = IXGBE_DEFAULT_FCRTH;
2798 hw->fc.low_water = IXGBE_DEFAULT_FCRTL;
2799 hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
2800 hw->fc.send_xon = true;
9a799d71 2801
021230d4 2802 /* select 10G link by default */
9a799d71 2803 hw->mac.link_mode_select = IXGBE_AUTOC_LMS_10G_LINK_NO_AN;
9a799d71 2804
30efa5a3
JB
2805 /* enable itr by default in dynamic mode */
2806 adapter->itr_setting = 1;
2807 adapter->eitr_param = 20000;
2808
2809 /* set defaults for eitr in MegaBytes */
2810 adapter->eitr_low = 10;
2811 adapter->eitr_high = 20;
2812
2813 /* set default ring sizes */
2814 adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
2815 adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
2816
9a799d71 2817 /* initialize eeprom parameters */
c44ade9e 2818 if (ixgbe_init_eeprom_params_generic(hw)) {
9a799d71
AK
2819 dev_err(&pdev->dev, "EEPROM initialization failed\n");
2820 return -EIO;
2821 }
2822
021230d4 2823 /* enable rx csum by default */
9a799d71
AK
2824 adapter->flags |= IXGBE_FLAG_RX_CSUM_ENABLED;
2825
9a799d71
AK
2826 set_bit(__IXGBE_DOWN, &adapter->state);
2827
2828 return 0;
2829}
2830
2831/**
2832 * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
2833 * @adapter: board private structure
3a581073 2834 * @tx_ring: tx descriptor ring (for a specific queue) to setup
9a799d71
AK
2835 *
2836 * Return 0 on success, negative on failure
2837 **/
2838int ixgbe_setup_tx_resources(struct ixgbe_adapter *adapter,
e01c31a5 2839 struct ixgbe_ring *tx_ring)
9a799d71
AK
2840{
2841 struct pci_dev *pdev = adapter->pdev;
2842 int size;
2843
3a581073
JB
2844 size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
2845 tx_ring->tx_buffer_info = vmalloc(size);
e01c31a5
JB
2846 if (!tx_ring->tx_buffer_info)
2847 goto err;
3a581073 2848 memset(tx_ring->tx_buffer_info, 0, size);
9a799d71
AK
2849
2850 /* round up to nearest 4K */
e01c31a5
JB
2851 tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc) +
2852 sizeof(u32);
3a581073 2853 tx_ring->size = ALIGN(tx_ring->size, 4096);
9a799d71 2854
3a581073
JB
2855 tx_ring->desc = pci_alloc_consistent(pdev, tx_ring->size,
2856 &tx_ring->dma);
e01c31a5
JB
2857 if (!tx_ring->desc)
2858 goto err;
9a799d71 2859
3a581073
JB
2860 tx_ring->next_to_use = 0;
2861 tx_ring->next_to_clean = 0;
2862 tx_ring->work_limit = tx_ring->count;
9a799d71 2863 return 0;
e01c31a5
JB
2864
2865err:
2866 vfree(tx_ring->tx_buffer_info);
2867 tx_ring->tx_buffer_info = NULL;
2868 DPRINTK(PROBE, ERR, "Unable to allocate memory for the transmit "
2869 "descriptor ring\n");
2870 return -ENOMEM;
9a799d71
AK
2871}
2872
69888674
AD
2873/**
2874 * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
2875 * @adapter: board private structure
2876 *
2877 * If this function returns with an error, then it's possible one or
2878 * more of the rings is populated (while the rest are not). It is the
2879 * callers duty to clean those orphaned rings.
2880 *
2881 * Return 0 on success, negative on failure
2882 **/
2883static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
2884{
2885 int i, err = 0;
2886
2887 for (i = 0; i < adapter->num_tx_queues; i++) {
2888 err = ixgbe_setup_tx_resources(adapter, &adapter->tx_ring[i]);
2889 if (!err)
2890 continue;
2891 DPRINTK(PROBE, ERR, "Allocation for Tx Queue %u failed\n", i);
2892 break;
2893 }
2894
2895 return err;
2896}
2897
9a799d71
AK
2898/**
2899 * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
2900 * @adapter: board private structure
3a581073 2901 * @rx_ring: rx descriptor ring (for a specific queue) to setup
9a799d71
AK
2902 *
2903 * Returns 0 on success, negative on failure
2904 **/
2905int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
b4617240 2906 struct ixgbe_ring *rx_ring)
9a799d71
AK
2907{
2908 struct pci_dev *pdev = adapter->pdev;
021230d4 2909 int size;
9a799d71 2910
177db6ff 2911 size = sizeof(struct net_lro_desc) * IXGBE_MAX_LRO_DESCRIPTORS;
3a581073
JB
2912 rx_ring->lro_mgr.lro_arr = vmalloc(size);
2913 if (!rx_ring->lro_mgr.lro_arr)
177db6ff 2914 return -ENOMEM;
3a581073 2915 memset(rx_ring->lro_mgr.lro_arr, 0, size);
177db6ff 2916
3a581073
JB
2917 size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
2918 rx_ring->rx_buffer_info = vmalloc(size);
2919 if (!rx_ring->rx_buffer_info) {
9a799d71 2920 DPRINTK(PROBE, ERR,
b4617240 2921 "vmalloc allocation failed for the rx desc ring\n");
177db6ff 2922 goto alloc_failed;
9a799d71 2923 }
3a581073 2924 memset(rx_ring->rx_buffer_info, 0, size);
9a799d71 2925
9a799d71 2926 /* Round up to nearest 4K */
3a581073
JB
2927 rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
2928 rx_ring->size = ALIGN(rx_ring->size, 4096);
9a799d71 2929
3a581073 2930 rx_ring->desc = pci_alloc_consistent(pdev, rx_ring->size, &rx_ring->dma);
9a799d71 2931
3a581073 2932 if (!rx_ring->desc) {
9a799d71 2933 DPRINTK(PROBE, ERR,
b4617240 2934 "Memory allocation failed for the rx desc ring\n");
3a581073 2935 vfree(rx_ring->rx_buffer_info);
177db6ff 2936 goto alloc_failed;
9a799d71
AK
2937 }
2938
3a581073
JB
2939 rx_ring->next_to_clean = 0;
2940 rx_ring->next_to_use = 0;
9a799d71
AK
2941
2942 return 0;
177db6ff
MC
2943
2944alloc_failed:
3a581073
JB
2945 vfree(rx_ring->lro_mgr.lro_arr);
2946 rx_ring->lro_mgr.lro_arr = NULL;
177db6ff 2947 return -ENOMEM;
9a799d71
AK
2948}
2949
69888674
AD
2950/**
2951 * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
2952 * @adapter: board private structure
2953 *
2954 * If this function returns with an error, then it's possible one or
2955 * more of the rings is populated (while the rest are not). It is the
2956 * callers duty to clean those orphaned rings.
2957 *
2958 * Return 0 on success, negative on failure
2959 **/
2960
2961static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
2962{
2963 int i, err = 0;
2964
2965 for (i = 0; i < adapter->num_rx_queues; i++) {
2966 err = ixgbe_setup_rx_resources(adapter, &adapter->rx_ring[i]);
2967 if (!err)
2968 continue;
2969 DPRINTK(PROBE, ERR, "Allocation for Rx Queue %u failed\n", i);
2970 break;
2971 }
2972
2973 return err;
2974}
2975
9a799d71
AK
2976/**
2977 * ixgbe_free_tx_resources - Free Tx Resources per Queue
2978 * @adapter: board private structure
2979 * @tx_ring: Tx descriptor ring for a specific queue
2980 *
2981 * Free all transmit software resources
2982 **/
c431f97e
JB
2983void ixgbe_free_tx_resources(struct ixgbe_adapter *adapter,
2984 struct ixgbe_ring *tx_ring)
9a799d71
AK
2985{
2986 struct pci_dev *pdev = adapter->pdev;
2987
2988 ixgbe_clean_tx_ring(adapter, tx_ring);
2989
2990 vfree(tx_ring->tx_buffer_info);
2991 tx_ring->tx_buffer_info = NULL;
2992
2993 pci_free_consistent(pdev, tx_ring->size, tx_ring->desc, tx_ring->dma);
2994
2995 tx_ring->desc = NULL;
2996}
2997
2998/**
2999 * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
3000 * @adapter: board private structure
3001 *
3002 * Free all transmit software resources
3003 **/
3004static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
3005{
3006 int i;
3007
3008 for (i = 0; i < adapter->num_tx_queues; i++)
3009 ixgbe_free_tx_resources(adapter, &adapter->tx_ring[i]);
3010}
3011
3012/**
b4617240 3013 * ixgbe_free_rx_resources - Free Rx Resources
9a799d71
AK
3014 * @adapter: board private structure
3015 * @rx_ring: ring to clean the resources from
3016 *
3017 * Free all receive software resources
3018 **/
c431f97e
JB
3019void ixgbe_free_rx_resources(struct ixgbe_adapter *adapter,
3020 struct ixgbe_ring *rx_ring)
9a799d71
AK
3021{
3022 struct pci_dev *pdev = adapter->pdev;
3023
177db6ff
MC
3024 vfree(rx_ring->lro_mgr.lro_arr);
3025 rx_ring->lro_mgr.lro_arr = NULL;
3026
9a799d71
AK
3027 ixgbe_clean_rx_ring(adapter, rx_ring);
3028
3029 vfree(rx_ring->rx_buffer_info);
3030 rx_ring->rx_buffer_info = NULL;
3031
3032 pci_free_consistent(pdev, rx_ring->size, rx_ring->desc, rx_ring->dma);
3033
3034 rx_ring->desc = NULL;
3035}
3036
3037/**
3038 * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
3039 * @adapter: board private structure
3040 *
3041 * Free all receive software resources
3042 **/
3043static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
3044{
3045 int i;
3046
3047 for (i = 0; i < adapter->num_rx_queues; i++)
3048 ixgbe_free_rx_resources(adapter, &adapter->rx_ring[i]);
3049}
3050
9a799d71
AK
3051/**
3052 * ixgbe_change_mtu - Change the Maximum Transfer Unit
3053 * @netdev: network interface device structure
3054 * @new_mtu: new value for maximum frame size
3055 *
3056 * Returns 0 on success, negative on failure
3057 **/
3058static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
3059{
3060 struct ixgbe_adapter *adapter = netdev_priv(netdev);
3061 int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
3062
42c783c5
JB
3063 /* MTU < 68 is an error and causes problems on some kernels */
3064 if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
9a799d71
AK
3065 return -EINVAL;
3066
021230d4 3067 DPRINTK(PROBE, INFO, "changing MTU from %d to %d\n",
b4617240 3068 netdev->mtu, new_mtu);
021230d4 3069 /* must set new MTU before calling down or up */
9a799d71
AK
3070 netdev->mtu = new_mtu;
3071
d4f80882
AV
3072 if (netif_running(netdev))
3073 ixgbe_reinit_locked(adapter);
9a799d71
AK
3074
3075 return 0;
3076}
3077
3078/**
3079 * ixgbe_open - Called when a network interface is made active
3080 * @netdev: network interface device structure
3081 *
3082 * Returns 0 on success, negative value on failure
3083 *
3084 * The open entry point is called when a network interface is made
3085 * active by the system (IFF_UP). At this point all resources needed
3086 * for transmit and receive operations are allocated, the interrupt
3087 * handler is registered with the OS, the watchdog timer is started,
3088 * and the stack is notified that the interface is ready.
3089 **/
3090static int ixgbe_open(struct net_device *netdev)
3091{
3092 struct ixgbe_adapter *adapter = netdev_priv(netdev);
3093 int err;
4bebfaa5
AK
3094
3095 /* disallow open during test */
3096 if (test_bit(__IXGBE_TESTING, &adapter->state))
3097 return -EBUSY;
9a799d71 3098
9a799d71
AK
3099 /* allocate transmit descriptors */
3100 err = ixgbe_setup_all_tx_resources(adapter);
3101 if (err)
3102 goto err_setup_tx;
3103
9a799d71
AK
3104 /* allocate receive descriptors */
3105 err = ixgbe_setup_all_rx_resources(adapter);
3106 if (err)
3107 goto err_setup_rx;
3108
3109 ixgbe_configure(adapter);
3110
021230d4 3111 err = ixgbe_request_irq(adapter);
9a799d71
AK
3112 if (err)
3113 goto err_req_irq;
3114
9a799d71
AK
3115 err = ixgbe_up_complete(adapter);
3116 if (err)
3117 goto err_up;
3118
d55b53ff
JK
3119 netif_tx_start_all_queues(netdev);
3120
9a799d71
AK
3121 return 0;
3122
3123err_up:
5eba3699 3124 ixgbe_release_hw_control(adapter);
9a799d71
AK
3125 ixgbe_free_irq(adapter);
3126err_req_irq:
3127 ixgbe_free_all_rx_resources(adapter);
3128err_setup_rx:
3129 ixgbe_free_all_tx_resources(adapter);
3130err_setup_tx:
3131 ixgbe_reset(adapter);
3132
3133 return err;
3134}
3135
3136/**
3137 * ixgbe_close - Disables a network interface
3138 * @netdev: network interface device structure
3139 *
3140 * Returns 0, this is not allowed to fail
3141 *
3142 * The close entry point is called when an interface is de-activated
3143 * by the OS. The hardware is still under the drivers control, but
3144 * needs to be disabled. A global MAC reset is issued to stop the
3145 * hardware, and all transmit and receive resources are freed.
3146 **/
3147static int ixgbe_close(struct net_device *netdev)
3148{
3149 struct ixgbe_adapter *adapter = netdev_priv(netdev);
9a799d71
AK
3150
3151 ixgbe_down(adapter);
3152 ixgbe_free_irq(adapter);
3153
3154 ixgbe_free_all_tx_resources(adapter);
3155 ixgbe_free_all_rx_resources(adapter);
3156
5eba3699 3157 ixgbe_release_hw_control(adapter);
9a799d71
AK
3158
3159 return 0;
3160}
3161
b3c8b4ba
AD
3162/**
3163 * ixgbe_napi_add_all - prep napi structs for use
3164 * @adapter: private struct
3165 * helper function to napi_add each possible q_vector->napi
3166 */
2f90b865 3167void ixgbe_napi_add_all(struct ixgbe_adapter *adapter)
b3c8b4ba
AD
3168{
3169 int q_idx, q_vectors;
3170 int (*poll)(struct napi_struct *, int);
3171
3172 if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3173 poll = &ixgbe_clean_rxonly;
3174 /* Only enable as many vectors as we have rx queues. */
3175 q_vectors = adapter->num_rx_queues;
3176 } else {
3177 poll = &ixgbe_poll;
3178 /* only one q_vector for legacy modes */
3179 q_vectors = 1;
3180 }
3181
3182 for (q_idx = 0; q_idx < q_vectors; q_idx++) {
3183 struct ixgbe_q_vector *q_vector = &adapter->q_vector[q_idx];
3184 netif_napi_add(adapter->netdev, &q_vector->napi, (*poll), 64);
3185 }
3186}
3187
2f90b865 3188void ixgbe_napi_del_all(struct ixgbe_adapter *adapter)
b3c8b4ba
AD
3189{
3190 int q_idx;
3191 int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
3192
3193 /* legacy and MSI only use one vector */
3194 if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED))
3195 q_vectors = 1;
3196
3197 for (q_idx = 0; q_idx < q_vectors; q_idx++) {
3198 struct ixgbe_q_vector *q_vector = &adapter->q_vector[q_idx];
3199 if (!q_vector->rxr_count)
3200 continue;
3201 netif_napi_del(&q_vector->napi);
3202 }
3203}
3204
3205#ifdef CONFIG_PM
3206static int ixgbe_resume(struct pci_dev *pdev)
3207{
3208 struct net_device *netdev = pci_get_drvdata(pdev);
3209 struct ixgbe_adapter *adapter = netdev_priv(netdev);
3210 u32 err;
3211
3212 pci_set_power_state(pdev, PCI_D0);
3213 pci_restore_state(pdev);
3214 err = pci_enable_device(pdev);
3215 if (err) {
69888674 3216 printk(KERN_ERR "ixgbe: Cannot enable PCI device from "
b3c8b4ba
AD
3217 "suspend\n");
3218 return err;
3219 }
3220 pci_set_master(pdev);
3221
3222 pci_enable_wake(pdev, PCI_D3hot, 0);
3223 pci_enable_wake(pdev, PCI_D3cold, 0);
3224
3225 err = ixgbe_init_interrupt_scheme(adapter);
3226 if (err) {
3227 printk(KERN_ERR "ixgbe: Cannot initialize interrupts for "
3228 "device\n");
3229 return err;
3230 }
3231
3232 ixgbe_napi_add_all(adapter);
3233 ixgbe_reset(adapter);
3234
3235 if (netif_running(netdev)) {
3236 err = ixgbe_open(adapter->netdev);
3237 if (err)
3238 return err;
3239 }
3240
3241 netif_device_attach(netdev);
3242
3243 return 0;
3244}
3245
3246#endif /* CONFIG_PM */
3247static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
3248{
3249 struct net_device *netdev = pci_get_drvdata(pdev);
3250 struct ixgbe_adapter *adapter = netdev_priv(netdev);
3251#ifdef CONFIG_PM
3252 int retval = 0;
3253#endif
3254
3255 netif_device_detach(netdev);
3256
3257 if (netif_running(netdev)) {
3258 ixgbe_down(adapter);
3259 ixgbe_free_irq(adapter);
3260 ixgbe_free_all_tx_resources(adapter);
3261 ixgbe_free_all_rx_resources(adapter);
3262 }
3263 ixgbe_reset_interrupt_capability(adapter);
3264 ixgbe_napi_del_all(adapter);
3265 kfree(adapter->tx_ring);
3266 kfree(adapter->rx_ring);
3267
3268#ifdef CONFIG_PM
3269 retval = pci_save_state(pdev);
3270 if (retval)
3271 return retval;
3272#endif
3273
3274 pci_enable_wake(pdev, PCI_D3hot, 0);
3275 pci_enable_wake(pdev, PCI_D3cold, 0);
3276
3277 ixgbe_release_hw_control(adapter);
3278
3279 pci_disable_device(pdev);
3280
3281 pci_set_power_state(pdev, pci_choose_state(pdev, state));
3282
3283 return 0;
3284}
3285
3286static void ixgbe_shutdown(struct pci_dev *pdev)
3287{
3288 ixgbe_suspend(pdev, PMSG_SUSPEND);
3289}
3290
9a799d71
AK
3291/**
3292 * ixgbe_update_stats - Update the board statistics counters.
3293 * @adapter: board private structure
3294 **/
3295void ixgbe_update_stats(struct ixgbe_adapter *adapter)
3296{
3297 struct ixgbe_hw *hw = &adapter->hw;
6f11eef7
AV
3298 u64 total_mpc = 0;
3299 u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
9a799d71
AK
3300
3301 adapter->stats.crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6f11eef7
AV
3302 for (i = 0; i < 8; i++) {
3303 /* for packet buffers not used, the register should read 0 */
3304 mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3305 missed_rx += mpc;
3306 adapter->stats.mpc[i] += mpc;
3307 total_mpc += adapter->stats.mpc[i];
3308 adapter->stats.rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
2f90b865
AD
3309 adapter->stats.qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3310 adapter->stats.qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
3311 adapter->stats.qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3312 adapter->stats.qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
3313 adapter->stats.pxonrxc[i] += IXGBE_READ_REG(hw,
3314 IXGBE_PXONRXC(i));
3315 adapter->stats.pxontxc[i] += IXGBE_READ_REG(hw,
3316 IXGBE_PXONTXC(i));
3317 adapter->stats.pxoffrxc[i] += IXGBE_READ_REG(hw,
3318 IXGBE_PXOFFRXC(i));
3319 adapter->stats.pxofftxc[i] += IXGBE_READ_REG(hw,
3320 IXGBE_PXOFFTXC(i));
6f11eef7
AV
3321 }
3322 adapter->stats.gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
3323 /* work around hardware counting issue */
3324 adapter->stats.gprc -= missed_rx;
3325
3326 /* 82598 hardware only has a 32 bit counter in the high register */
9a799d71 3327 adapter->stats.gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6f11eef7
AV
3328 adapter->stats.gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3329 adapter->stats.tor += IXGBE_READ_REG(hw, IXGBE_TORH);
9a799d71
AK
3330 bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3331 adapter->stats.bprc += bprc;
3332 adapter->stats.mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3333 adapter->stats.mprc -= bprc;
3334 adapter->stats.roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3335 adapter->stats.prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3336 adapter->stats.prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3337 adapter->stats.prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3338 adapter->stats.prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3339 adapter->stats.prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3340 adapter->stats.prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
9a799d71
AK
3341 adapter->stats.rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3342 adapter->stats.lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
9a799d71 3343 adapter->stats.lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
6f11eef7
AV
3344 lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3345 adapter->stats.lxontxc += lxon;
3346 lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3347 adapter->stats.lxofftxc += lxoff;
9a799d71
AK
3348 adapter->stats.ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3349 adapter->stats.gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6f11eef7
AV
3350 adapter->stats.mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3351 /*
3352 * 82598 errata - tx of flow control packets is included in tx counters
3353 */
3354 xon_off_tot = lxon + lxoff;
3355 adapter->stats.gptc -= xon_off_tot;
3356 adapter->stats.mptc -= xon_off_tot;
3357 adapter->stats.gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
9a799d71
AK
3358 adapter->stats.ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3359 adapter->stats.rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3360 adapter->stats.rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
9a799d71
AK
3361 adapter->stats.tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3362 adapter->stats.ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6f11eef7 3363 adapter->stats.ptc64 -= xon_off_tot;
9a799d71
AK
3364 adapter->stats.ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3365 adapter->stats.ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3366 adapter->stats.ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3367 adapter->stats.ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3368 adapter->stats.ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
9a799d71
AK
3369 adapter->stats.bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3370
3371 /* Fill out the OS statistics structure */
9a799d71
AK
3372 adapter->net_stats.multicast = adapter->stats.mprc;
3373
3374 /* Rx Errors */
3375 adapter->net_stats.rx_errors = adapter->stats.crcerrs +
b4617240 3376 adapter->stats.rlec;
9a799d71
AK
3377 adapter->net_stats.rx_dropped = 0;
3378 adapter->net_stats.rx_length_errors = adapter->stats.rlec;
3379 adapter->net_stats.rx_crc_errors = adapter->stats.crcerrs;
6f11eef7 3380 adapter->net_stats.rx_missed_errors = total_mpc;
9a799d71
AK
3381}
3382
3383/**
3384 * ixgbe_watchdog - Timer Call-back
3385 * @data: pointer to adapter cast into an unsigned long
3386 **/
3387static void ixgbe_watchdog(unsigned long data)
3388{
3389 struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
cf8280ee
JB
3390 struct ixgbe_hw *hw = &adapter->hw;
3391
3392 /* Do the watchdog outside of interrupt context due to the lovely
3393 * delays that some of the newer hardware requires */
3394 if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
3395 /* Cause software interrupt to ensure rx rings are cleaned */
3396 if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3397 u32 eics =
3398 (1 << (adapter->num_msix_vectors - NON_Q_VECTORS)) - 1;
3399 IXGBE_WRITE_REG(hw, IXGBE_EICS, eics);
3400 } else {
3401 /* For legacy and MSI interrupts don't set any bits that
3402 * are enabled for EIAM, because this operation would
3403 * set *both* EIMS and EICS for any bit in EIAM */
3404 IXGBE_WRITE_REG(hw, IXGBE_EICS,
3405 (IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
3406 }
3407 /* Reset the timer */
3408 mod_timer(&adapter->watchdog_timer,
3409 round_jiffies(jiffies + 2 * HZ));
3410 }
9a799d71 3411
cf8280ee
JB
3412 schedule_work(&adapter->watchdog_task);
3413}
3414
3415/**
69888674
AD
3416 * ixgbe_watchdog_task - worker thread to bring link up
3417 * @work: pointer to work_struct containing our data
cf8280ee
JB
3418 **/
3419static void ixgbe_watchdog_task(struct work_struct *work)
3420{
3421 struct ixgbe_adapter *adapter = container_of(work,
3422 struct ixgbe_adapter,
3423 watchdog_task);
3424 struct net_device *netdev = adapter->netdev;
3425 struct ixgbe_hw *hw = &adapter->hw;
3426 u32 link_speed = adapter->link_speed;
3427 bool link_up = adapter->link_up;
3428
3429 adapter->flags |= IXGBE_FLAG_IN_WATCHDOG_TASK;
3430
3431 if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
3432 hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
3433 if (link_up ||
3434 time_after(jiffies, (adapter->link_check_timeout +
3435 IXGBE_TRY_LINK_TIMEOUT))) {
3436 IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
3437 adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
3438 }
3439 adapter->link_up = link_up;
3440 adapter->link_speed = link_speed;
3441 }
9a799d71
AK
3442
3443 if (link_up) {
3444 if (!netif_carrier_ok(netdev)) {
cf8280ee
JB
3445 u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3446 u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
9a799d71
AK
3447#define FLOW_RX (frctl & IXGBE_FCTRL_RFCE)
3448#define FLOW_TX (rmcs & IXGBE_RMCS_TFCE_802_3X)
3449 DPRINTK(LINK, INFO, "NIC Link is Up %s, "
cf8280ee
JB
3450 "Flow Control: %s\n",
3451 (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
3452 "10 Gbps" :
3453 (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
3454 "1 Gbps" : "unknown speed")),
3455 ((FLOW_RX && FLOW_TX) ? "RX/TX" :
3456 (FLOW_RX ? "RX" :
3457 (FLOW_TX ? "TX" : "None"))));
9a799d71
AK
3458
3459 netif_carrier_on(netdev);
2f90b865 3460 netif_tx_wake_all_queues(netdev);
9a799d71
AK
3461 } else {
3462 /* Force detection of hung controller */
3463 adapter->detect_tx_hung = true;
3464 }
3465 } else {
cf8280ee
JB
3466 adapter->link_up = false;
3467 adapter->link_speed = 0;
9a799d71
AK
3468 if (netif_carrier_ok(netdev)) {
3469 DPRINTK(LINK, INFO, "NIC Link is Down\n");
3470 netif_carrier_off(netdev);
2f90b865 3471 netif_tx_stop_all_queues(netdev);
9a799d71
AK
3472 }
3473 }
3474
3475 ixgbe_update_stats(adapter);
cf8280ee 3476 adapter->flags &= ~IXGBE_FLAG_IN_WATCHDOG_TASK;
9a799d71
AK
3477}
3478
9a799d71 3479static int ixgbe_tso(struct ixgbe_adapter *adapter,
b4617240
PW
3480 struct ixgbe_ring *tx_ring, struct sk_buff *skb,
3481 u32 tx_flags, u8 *hdr_len)
9a799d71
AK
3482{
3483 struct ixgbe_adv_tx_context_desc *context_desc;
3484 unsigned int i;
3485 int err;
3486 struct ixgbe_tx_buffer *tx_buffer_info;
9f8cdf4f
JB
3487 u32 vlan_macip_lens = 0, type_tucmd_mlhl;
3488 u32 mss_l4len_idx, l4len;
9a799d71
AK
3489
3490 if (skb_is_gso(skb)) {
3491 if (skb_header_cloned(skb)) {
3492 err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3493 if (err)
3494 return err;
3495 }
3496 l4len = tcp_hdrlen(skb);
3497 *hdr_len += l4len;
3498
8327d000 3499 if (skb->protocol == htons(ETH_P_IP)) {
9a799d71
AK
3500 struct iphdr *iph = ip_hdr(skb);
3501 iph->tot_len = 0;
3502 iph->check = 0;
3503 tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
b4617240
PW
3504 iph->daddr, 0,
3505 IPPROTO_TCP,
3506 0);
9a799d71
AK
3507 adapter->hw_tso_ctxt++;
3508 } else if (skb_shinfo(skb)->gso_type == SKB_GSO_TCPV6) {
3509 ipv6_hdr(skb)->payload_len = 0;
3510 tcp_hdr(skb)->check =
3511 ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
b4617240
PW
3512 &ipv6_hdr(skb)->daddr,
3513 0, IPPROTO_TCP, 0);
9a799d71
AK
3514 adapter->hw_tso6_ctxt++;
3515 }
3516
3517 i = tx_ring->next_to_use;
3518
3519 tx_buffer_info = &tx_ring->tx_buffer_info[i];
3520 context_desc = IXGBE_TX_CTXTDESC_ADV(*tx_ring, i);
3521
3522 /* VLAN MACLEN IPLEN */
3523 if (tx_flags & IXGBE_TX_FLAGS_VLAN)
3524 vlan_macip_lens |=
3525 (tx_flags & IXGBE_TX_FLAGS_VLAN_MASK);
3526 vlan_macip_lens |= ((skb_network_offset(skb)) <<
b4617240 3527 IXGBE_ADVTXD_MACLEN_SHIFT);
9a799d71
AK
3528 *hdr_len += skb_network_offset(skb);
3529 vlan_macip_lens |=
3530 (skb_transport_header(skb) - skb_network_header(skb));
3531 *hdr_len +=
3532 (skb_transport_header(skb) - skb_network_header(skb));
3533 context_desc->vlan_macip_lens = cpu_to_le32(vlan_macip_lens);
3534 context_desc->seqnum_seed = 0;
3535
3536 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
9f8cdf4f 3537 type_tucmd_mlhl = (IXGBE_TXD_CMD_DEXT |
b4617240 3538 IXGBE_ADVTXD_DTYP_CTXT);
9a799d71 3539
8327d000 3540 if (skb->protocol == htons(ETH_P_IP))
9a799d71
AK
3541 type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV4;
3542 type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
3543 context_desc->type_tucmd_mlhl = cpu_to_le32(type_tucmd_mlhl);
3544
3545 /* MSS L4LEN IDX */
9f8cdf4f 3546 mss_l4len_idx =
9a799d71
AK
3547 (skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT);
3548 mss_l4len_idx |= (l4len << IXGBE_ADVTXD_L4LEN_SHIFT);
4eeae6fd
PW
3549 /* use index 1 for TSO */
3550 mss_l4len_idx |= (1 << IXGBE_ADVTXD_IDX_SHIFT);
9a799d71
AK
3551 context_desc->mss_l4len_idx = cpu_to_le32(mss_l4len_idx);
3552
3553 tx_buffer_info->time_stamp = jiffies;
3554 tx_buffer_info->next_to_watch = i;
3555
3556 i++;
3557 if (i == tx_ring->count)
3558 i = 0;
3559 tx_ring->next_to_use = i;
3560
3561 return true;
3562 }
3563 return false;
3564}
3565
3566static bool ixgbe_tx_csum(struct ixgbe_adapter *adapter,
b4617240
PW
3567 struct ixgbe_ring *tx_ring,
3568 struct sk_buff *skb, u32 tx_flags)
9a799d71
AK
3569{
3570 struct ixgbe_adv_tx_context_desc *context_desc;
3571 unsigned int i;
3572 struct ixgbe_tx_buffer *tx_buffer_info;
3573 u32 vlan_macip_lens = 0, type_tucmd_mlhl = 0;
3574
3575 if (skb->ip_summed == CHECKSUM_PARTIAL ||
3576 (tx_flags & IXGBE_TX_FLAGS_VLAN)) {
3577 i = tx_ring->next_to_use;
3578 tx_buffer_info = &tx_ring->tx_buffer_info[i];
3579 context_desc = IXGBE_TX_CTXTDESC_ADV(*tx_ring, i);
3580
3581 if (tx_flags & IXGBE_TX_FLAGS_VLAN)
3582 vlan_macip_lens |=
3583 (tx_flags & IXGBE_TX_FLAGS_VLAN_MASK);
3584 vlan_macip_lens |= (skb_network_offset(skb) <<
b4617240 3585 IXGBE_ADVTXD_MACLEN_SHIFT);
9a799d71
AK
3586 if (skb->ip_summed == CHECKSUM_PARTIAL)
3587 vlan_macip_lens |= (skb_transport_header(skb) -
b4617240 3588 skb_network_header(skb));
9a799d71
AK
3589
3590 context_desc->vlan_macip_lens = cpu_to_le32(vlan_macip_lens);
3591 context_desc->seqnum_seed = 0;
3592
3593 type_tucmd_mlhl |= (IXGBE_TXD_CMD_DEXT |
b4617240 3594 IXGBE_ADVTXD_DTYP_CTXT);
9a799d71
AK
3595
3596 if (skb->ip_summed == CHECKSUM_PARTIAL) {
41825d71
AK
3597 switch (skb->protocol) {
3598 case __constant_htons(ETH_P_IP):
9a799d71 3599 type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV4;
41825d71
AK
3600 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
3601 type_tucmd_mlhl |=
b4617240 3602 IXGBE_ADVTXD_TUCMD_L4T_TCP;
41825d71 3603 break;
41825d71
AK
3604 case __constant_htons(ETH_P_IPV6):
3605 /* XXX what about other V6 headers?? */
3606 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
3607 type_tucmd_mlhl |=
b4617240 3608 IXGBE_ADVTXD_TUCMD_L4T_TCP;
41825d71 3609 break;
41825d71
AK
3610 default:
3611 if (unlikely(net_ratelimit())) {
3612 DPRINTK(PROBE, WARNING,
3613 "partial checksum but proto=%x!\n",
3614 skb->protocol);
3615 }
3616 break;
3617 }
9a799d71
AK
3618 }
3619
3620 context_desc->type_tucmd_mlhl = cpu_to_le32(type_tucmd_mlhl);
4eeae6fd 3621 /* use index zero for tx checksum offload */
9a799d71
AK
3622 context_desc->mss_l4len_idx = 0;
3623
3624 tx_buffer_info->time_stamp = jiffies;
3625 tx_buffer_info->next_to_watch = i;
9f8cdf4f 3626
9a799d71
AK
3627 adapter->hw_csum_tx_good++;
3628 i++;
3629 if (i == tx_ring->count)
3630 i = 0;
3631 tx_ring->next_to_use = i;
3632
3633 return true;
3634 }
9f8cdf4f 3635
9a799d71
AK
3636 return false;
3637}
3638
3639static int ixgbe_tx_map(struct ixgbe_adapter *adapter,
b4617240
PW
3640 struct ixgbe_ring *tx_ring,
3641 struct sk_buff *skb, unsigned int first)
9a799d71
AK
3642{
3643 struct ixgbe_tx_buffer *tx_buffer_info;
3644 unsigned int len = skb->len;
3645 unsigned int offset = 0, size, count = 0, i;
3646 unsigned int nr_frags = skb_shinfo(skb)->nr_frags;
3647 unsigned int f;
3648
3649 len -= skb->data_len;
3650
3651 i = tx_ring->next_to_use;
3652
3653 while (len) {
3654 tx_buffer_info = &tx_ring->tx_buffer_info[i];
3655 size = min(len, (uint)IXGBE_MAX_DATA_PER_TXD);
3656
3657 tx_buffer_info->length = size;
3658 tx_buffer_info->dma = pci_map_single(adapter->pdev,
b4617240
PW
3659 skb->data + offset,
3660 size, PCI_DMA_TODEVICE);
9a799d71
AK
3661 tx_buffer_info->time_stamp = jiffies;
3662 tx_buffer_info->next_to_watch = i;
3663
3664 len -= size;
3665 offset += size;
3666 count++;
3667 i++;
3668 if (i == tx_ring->count)
3669 i = 0;
3670 }
3671
3672 for (f = 0; f < nr_frags; f++) {
3673 struct skb_frag_struct *frag;
3674
3675 frag = &skb_shinfo(skb)->frags[f];
3676 len = frag->size;
3677 offset = frag->page_offset;
3678
3679 while (len) {
3680 tx_buffer_info = &tx_ring->tx_buffer_info[i];
3681 size = min(len, (uint)IXGBE_MAX_DATA_PER_TXD);
3682
3683 tx_buffer_info->length = size;
3684 tx_buffer_info->dma = pci_map_page(adapter->pdev,
b4617240
PW
3685 frag->page,
3686 offset,
3687 size,
3688 PCI_DMA_TODEVICE);
9a799d71
AK
3689 tx_buffer_info->time_stamp = jiffies;
3690 tx_buffer_info->next_to_watch = i;
3691
3692 len -= size;
3693 offset += size;
3694 count++;
3695 i++;
3696 if (i == tx_ring->count)
3697 i = 0;
3698 }
3699 }
3700 if (i == 0)
3701 i = tx_ring->count - 1;
3702 else
3703 i = i - 1;
3704 tx_ring->tx_buffer_info[i].skb = skb;
3705 tx_ring->tx_buffer_info[first].next_to_watch = i;
3706
3707 return count;
3708}
3709
3710static void ixgbe_tx_queue(struct ixgbe_adapter *adapter,
b4617240
PW
3711 struct ixgbe_ring *tx_ring,
3712 int tx_flags, int count, u32 paylen, u8 hdr_len)
9a799d71
AK
3713{
3714 union ixgbe_adv_tx_desc *tx_desc = NULL;
3715 struct ixgbe_tx_buffer *tx_buffer_info;
3716 u32 olinfo_status = 0, cmd_type_len = 0;
3717 unsigned int i;
3718 u32 txd_cmd = IXGBE_TXD_CMD_EOP | IXGBE_TXD_CMD_RS | IXGBE_TXD_CMD_IFCS;
3719
3720 cmd_type_len |= IXGBE_ADVTXD_DTYP_DATA;
3721
3722 cmd_type_len |= IXGBE_ADVTXD_DCMD_IFCS | IXGBE_ADVTXD_DCMD_DEXT;
3723
3724 if (tx_flags & IXGBE_TX_FLAGS_VLAN)
3725 cmd_type_len |= IXGBE_ADVTXD_DCMD_VLE;
3726
3727 if (tx_flags & IXGBE_TX_FLAGS_TSO) {
3728 cmd_type_len |= IXGBE_ADVTXD_DCMD_TSE;
3729
3730 olinfo_status |= IXGBE_TXD_POPTS_TXSM <<
b4617240 3731 IXGBE_ADVTXD_POPTS_SHIFT;
9a799d71 3732
4eeae6fd
PW
3733 /* use index 1 context for tso */
3734 olinfo_status |= (1 << IXGBE_ADVTXD_IDX_SHIFT);
9a799d71
AK
3735 if (tx_flags & IXGBE_TX_FLAGS_IPV4)
3736 olinfo_status |= IXGBE_TXD_POPTS_IXSM <<
b4617240 3737 IXGBE_ADVTXD_POPTS_SHIFT;
9a799d71
AK
3738
3739 } else if (tx_flags & IXGBE_TX_FLAGS_CSUM)
3740 olinfo_status |= IXGBE_TXD_POPTS_TXSM <<
b4617240 3741 IXGBE_ADVTXD_POPTS_SHIFT;
9a799d71
AK
3742
3743 olinfo_status |= ((paylen - hdr_len) << IXGBE_ADVTXD_PAYLEN_SHIFT);
3744
3745 i = tx_ring->next_to_use;
3746 while (count--) {
3747 tx_buffer_info = &tx_ring->tx_buffer_info[i];
3748 tx_desc = IXGBE_TX_DESC_ADV(*tx_ring, i);
3749 tx_desc->read.buffer_addr = cpu_to_le64(tx_buffer_info->dma);
3750 tx_desc->read.cmd_type_len =
b4617240 3751 cpu_to_le32(cmd_type_len | tx_buffer_info->length);
9a799d71 3752 tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
9a799d71
AK
3753 i++;
3754 if (i == tx_ring->count)
3755 i = 0;
3756 }
3757
3758 tx_desc->read.cmd_type_len |= cpu_to_le32(txd_cmd);
3759
3760 /*
3761 * Force memory writes to complete before letting h/w
3762 * know there are new descriptors to fetch. (Only
3763 * applicable for weak-ordered memory model archs,
3764 * such as IA-64).
3765 */
3766 wmb();
3767
3768 tx_ring->next_to_use = i;
3769 writel(i, adapter->hw.hw_addr + tx_ring->tail);
3770}
3771
e092be60 3772static int __ixgbe_maybe_stop_tx(struct net_device *netdev,
b4617240 3773 struct ixgbe_ring *tx_ring, int size)
e092be60
AV
3774{
3775 struct ixgbe_adapter *adapter = netdev_priv(netdev);
3776
30eba97a 3777 netif_stop_subqueue(netdev, tx_ring->queue_index);
e092be60
AV
3778 /* Herbert's original patch had:
3779 * smp_mb__after_netif_stop_queue();
3780 * but since that doesn't exist yet, just open code it. */
3781 smp_mb();
3782
3783 /* We need to check again in a case another CPU has just
3784 * made room available. */
3785 if (likely(IXGBE_DESC_UNUSED(tx_ring) < size))
3786 return -EBUSY;
3787
3788 /* A reprieve! - use start_queue because it doesn't call schedule */
af72166f 3789 netif_start_subqueue(netdev, tx_ring->queue_index);
e092be60
AV
3790 ++adapter->restart_queue;
3791 return 0;
3792}
3793
3794static int ixgbe_maybe_stop_tx(struct net_device *netdev,
b4617240 3795 struct ixgbe_ring *tx_ring, int size)
e092be60
AV
3796{
3797 if (likely(IXGBE_DESC_UNUSED(tx_ring) >= size))
3798 return 0;
3799 return __ixgbe_maybe_stop_tx(netdev, tx_ring, size);
3800}
3801
9a799d71
AK
3802static int ixgbe_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
3803{
3804 struct ixgbe_adapter *adapter = netdev_priv(netdev);
3805 struct ixgbe_ring *tx_ring;
9a799d71
AK
3806 unsigned int first;
3807 unsigned int tx_flags = 0;
30eba97a
AV
3808 u8 hdr_len = 0;
3809 int r_idx = 0, tso;
9a799d71
AK
3810 int count = 0;
3811 unsigned int f;
9f8cdf4f 3812
30eba97a 3813 r_idx = (adapter->num_tx_queues - 1) & skb->queue_mapping;
30eba97a 3814 tx_ring = &adapter->tx_ring[r_idx];
9a799d71 3815
9f8cdf4f
JB
3816 if (adapter->vlgrp && vlan_tx_tag_present(skb)) {
3817 tx_flags |= vlan_tx_tag_get(skb);
2f90b865
AD
3818 if (adapter->flags & IXGBE_FLAG_DCB_ENABLED) {
3819 tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
3820 tx_flags |= (skb->queue_mapping << 13);
3821 }
3822 tx_flags <<= IXGBE_TX_FLAGS_VLAN_SHIFT;
3823 tx_flags |= IXGBE_TX_FLAGS_VLAN;
3824 } else if (adapter->flags & IXGBE_FLAG_DCB_ENABLED) {
3825 tx_flags |= (skb->queue_mapping << 13);
9f8cdf4f
JB
3826 tx_flags <<= IXGBE_TX_FLAGS_VLAN_SHIFT;
3827 tx_flags |= IXGBE_TX_FLAGS_VLAN;
9a799d71 3828 }
9f8cdf4f
JB
3829 /* three things can cause us to need a context descriptor */
3830 if (skb_is_gso(skb) ||
3831 (skb->ip_summed == CHECKSUM_PARTIAL) ||
3832 (tx_flags & IXGBE_TX_FLAGS_VLAN))
9a799d71
AK
3833 count++;
3834
9f8cdf4f
JB
3835 count += TXD_USE_COUNT(skb_headlen(skb));
3836 for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
9a799d71
AK
3837 count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
3838
e092be60 3839 if (ixgbe_maybe_stop_tx(netdev, tx_ring, count)) {
9a799d71 3840 adapter->tx_busy++;
9a799d71
AK
3841 return NETDEV_TX_BUSY;
3842 }
9a799d71 3843
8327d000 3844 if (skb->protocol == htons(ETH_P_IP))
9a799d71
AK
3845 tx_flags |= IXGBE_TX_FLAGS_IPV4;
3846 first = tx_ring->next_to_use;
3847 tso = ixgbe_tso(adapter, tx_ring, skb, tx_flags, &hdr_len);
3848 if (tso < 0) {
3849 dev_kfree_skb_any(skb);
3850 return NETDEV_TX_OK;
3851 }
3852
3853 if (tso)
3854 tx_flags |= IXGBE_TX_FLAGS_TSO;
3855 else if (ixgbe_tx_csum(adapter, tx_ring, skb, tx_flags) &&
b4617240 3856 (skb->ip_summed == CHECKSUM_PARTIAL))
9a799d71
AK
3857 tx_flags |= IXGBE_TX_FLAGS_CSUM;
3858
3859 ixgbe_tx_queue(adapter, tx_ring, tx_flags,
b4617240
PW
3860 ixgbe_tx_map(adapter, tx_ring, skb, first),
3861 skb->len, hdr_len);
9a799d71
AK
3862
3863 netdev->trans_start = jiffies;
3864
e092be60 3865 ixgbe_maybe_stop_tx(netdev, tx_ring, DESC_NEEDED);
9a799d71
AK
3866
3867 return NETDEV_TX_OK;
3868}
3869
3870/**
3871 * ixgbe_get_stats - Get System Network Statistics
3872 * @netdev: network interface device structure
3873 *
3874 * Returns the address of the device statistics structure.
3875 * The statistics are actually updated from the timer callback.
3876 **/
3877static struct net_device_stats *ixgbe_get_stats(struct net_device *netdev)
3878{
3879 struct ixgbe_adapter *adapter = netdev_priv(netdev);
3880
3881 /* only return the current stats */
3882 return &adapter->net_stats;
3883}
3884
3885/**
3886 * ixgbe_set_mac - Change the Ethernet Address of the NIC
3887 * @netdev: network interface device structure
3888 * @p: pointer to an address structure
3889 *
3890 * Returns 0 on success, negative on failure
3891 **/
3892static int ixgbe_set_mac(struct net_device *netdev, void *p)
3893{
3894 struct ixgbe_adapter *adapter = netdev_priv(netdev);
b4617240 3895 struct ixgbe_hw *hw = &adapter->hw;
9a799d71
AK
3896 struct sockaddr *addr = p;
3897
3898 if (!is_valid_ether_addr(addr->sa_data))
3899 return -EADDRNOTAVAIL;
3900
3901 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
b4617240 3902 memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
9a799d71 3903
b4617240 3904 hw->mac.ops.set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
9a799d71
AK
3905
3906 return 0;
3907}
3908
3909#ifdef CONFIG_NET_POLL_CONTROLLER
3910/*
3911 * Polling 'interrupt' - used by things like netconsole to send skbs
3912 * without having to re-enable interrupts. It's not called while
3913 * the interrupt routine is executing.
3914 */
3915static void ixgbe_netpoll(struct net_device *netdev)
3916{
3917 struct ixgbe_adapter *adapter = netdev_priv(netdev);
3918
3919 disable_irq(adapter->pdev->irq);
3920 adapter->flags |= IXGBE_FLAG_IN_NETPOLL;
3921 ixgbe_intr(adapter->pdev->irq, netdev);
3922 adapter->flags &= ~IXGBE_FLAG_IN_NETPOLL;
3923 enable_irq(adapter->pdev->irq);
3924}
3925#endif
3926
c44ade9e
JB
3927/**
3928 * ixgbe_link_config - set up initial link with default speed and duplex
3929 * @hw: pointer to private hardware struct
3930 *
3931 * Returns 0 on success, negative on failure
3932 **/
3933static int ixgbe_link_config(struct ixgbe_hw *hw)
3934{
3935 u32 autoneg = IXGBE_LINK_SPEED_10GB_FULL;
3936
3937 /* must always autoneg for both 1G and 10G link */
3938 hw->mac.autoneg = true;
3939
0befdb3e
JB
3940 if ((hw->mac.type == ixgbe_mac_82598EB) &&
3941 (hw->phy.media_type == ixgbe_media_type_copper))
3942 autoneg = IXGBE_LINK_SPEED_82598_AUTONEG;
3943
c44ade9e
JB
3944 return hw->mac.ops.setup_link_speed(hw, autoneg, true, true);
3945}
3946
0edc3527
SH
3947static const struct net_device_ops ixgbe_netdev_ops = {
3948 .ndo_open = ixgbe_open,
3949 .ndo_stop = ixgbe_close,
00829823 3950 .ndo_start_xmit = ixgbe_xmit_frame,
0edc3527
SH
3951 .ndo_get_stats = ixgbe_get_stats,
3952 .ndo_set_multicast_list = ixgbe_set_rx_mode,
3953 .ndo_validate_addr = eth_validate_addr,
3954 .ndo_set_mac_address = ixgbe_set_mac,
3955 .ndo_change_mtu = ixgbe_change_mtu,
3956 .ndo_tx_timeout = ixgbe_tx_timeout,
3957 .ndo_vlan_rx_register = ixgbe_vlan_rx_register,
3958 .ndo_vlan_rx_add_vid = ixgbe_vlan_rx_add_vid,
3959 .ndo_vlan_rx_kill_vid = ixgbe_vlan_rx_kill_vid,
3960#ifdef CONFIG_NET_POLL_CONTROLLER
3961 .ndo_poll_controller = ixgbe_netpoll,
3962#endif
3963};
3964
9a799d71
AK
3965/**
3966 * ixgbe_probe - Device Initialization Routine
3967 * @pdev: PCI device information struct
3968 * @ent: entry in ixgbe_pci_tbl
3969 *
3970 * Returns 0 on success, negative on failure
3971 *
3972 * ixgbe_probe initializes an adapter identified by a pci_dev structure.
3973 * The OS initialization, configuring of the adapter private structure,
3974 * and a hardware reset occur.
3975 **/
3976static int __devinit ixgbe_probe(struct pci_dev *pdev,
b4617240 3977 const struct pci_device_id *ent)
9a799d71
AK
3978{
3979 struct net_device *netdev;
3980 struct ixgbe_adapter *adapter = NULL;
3981 struct ixgbe_hw *hw;
3982 const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
9a799d71
AK
3983 static int cards_found;
3984 int i, err, pci_using_dac;
3985 u16 link_status, link_speed, link_width;
c44ade9e 3986 u32 part_num, eec;
9a799d71
AK
3987
3988 err = pci_enable_device(pdev);
3989 if (err)
3990 return err;
3991
3992 if (!pci_set_dma_mask(pdev, DMA_64BIT_MASK) &&
3993 !pci_set_consistent_dma_mask(pdev, DMA_64BIT_MASK)) {
3994 pci_using_dac = 1;
3995 } else {
3996 err = pci_set_dma_mask(pdev, DMA_32BIT_MASK);
3997 if (err) {
3998 err = pci_set_consistent_dma_mask(pdev, DMA_32BIT_MASK);
3999 if (err) {
b4617240
PW
4000 dev_err(&pdev->dev, "No usable DMA "
4001 "configuration, aborting\n");
9a799d71
AK
4002 goto err_dma;
4003 }
4004 }
4005 pci_using_dac = 0;
4006 }
4007
4008 err = pci_request_regions(pdev, ixgbe_driver_name);
4009 if (err) {
4010 dev_err(&pdev->dev, "pci_request_regions failed 0x%x\n", err);
4011 goto err_pci_reg;
4012 }
4013
4014 pci_set_master(pdev);
fb3b27bc 4015 pci_save_state(pdev);
9a799d71 4016
30eba97a 4017 netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), MAX_TX_QUEUES);
9a799d71
AK
4018 if (!netdev) {
4019 err = -ENOMEM;
4020 goto err_alloc_etherdev;
4021 }
4022
9a799d71
AK
4023 SET_NETDEV_DEV(netdev, &pdev->dev);
4024
4025 pci_set_drvdata(pdev, netdev);
4026 adapter = netdev_priv(netdev);
4027
4028 adapter->netdev = netdev;
4029 adapter->pdev = pdev;
4030 hw = &adapter->hw;
4031 hw->back = adapter;
4032 adapter->msg_enable = (1 << DEFAULT_DEBUG_LEVEL_SHIFT) - 1;
4033
05857980
JK
4034 hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
4035 pci_resource_len(pdev, 0));
9a799d71
AK
4036 if (!hw->hw_addr) {
4037 err = -EIO;
4038 goto err_ioremap;
4039 }
4040
4041 for (i = 1; i <= 5; i++) {
4042 if (pci_resource_len(pdev, i) == 0)
4043 continue;
4044 }
4045
0edc3527 4046 netdev->netdev_ops = &ixgbe_netdev_ops;
9a799d71 4047 ixgbe_set_ethtool_ops(netdev);
9a799d71 4048 netdev->watchdog_timeo = 5 * HZ;
9a799d71
AK
4049 strcpy(netdev->name, pci_name(pdev));
4050
9a799d71
AK
4051 adapter->bd_number = cards_found;
4052
9a799d71
AK
4053 /* Setup hw api */
4054 memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
021230d4 4055 hw->mac.type = ii->mac;
9a799d71 4056
c44ade9e
JB
4057 /* EEPROM */
4058 memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
4059 eec = IXGBE_READ_REG(hw, IXGBE_EEC);
4060 /* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
4061 if (!(eec & (1 << 8)))
4062 hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
4063
4064 /* PHY */
4065 memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
c4900be0
DS
4066 hw->phy.sfp_type = ixgbe_sfp_type_unknown;
4067
4068 /* set up this timer and work struct before calling get_invariants
4069 * which might start the timer
4070 */
4071 init_timer(&adapter->sfp_timer);
4072 adapter->sfp_timer.function = &ixgbe_sfp_timer;
4073 adapter->sfp_timer.data = (unsigned long) adapter;
4074
4075 INIT_WORK(&adapter->sfp_task, ixgbe_sfp_task);
c44ade9e 4076
9a799d71 4077 err = ii->get_invariants(hw);
c4900be0
DS
4078 if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
4079 /* start a kernel thread to watch for a module to arrive */
4080 set_bit(__IXGBE_SFP_MODULE_NOT_FOUND, &adapter->state);
4081 mod_timer(&adapter->sfp_timer,
4082 round_jiffies(jiffies + (2 * HZ)));
4083 err = 0;
4084 } else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
4085 DPRINTK(PROBE, ERR, "failed to load because an "
4086 "unsupported SFP+ module type was detected.\n");
9a799d71 4087 goto err_hw_init;
c4900be0
DS
4088 } else if (err) {
4089 goto err_hw_init;
4090 }
9a799d71
AK
4091
4092 /* setup the private structure */
4093 err = ixgbe_sw_init(adapter);
4094 if (err)
4095 goto err_sw_init;
4096
c44ade9e
JB
4097 /* reset_hw fills in the perm_addr as well */
4098 err = hw->mac.ops.reset_hw(hw);
4099 if (err) {
4100 dev_err(&adapter->pdev->dev, "HW Init failed: %d\n", err);
4101 goto err_sw_init;
4102 }
4103
9a799d71 4104 netdev->features = NETIF_F_SG |
b4617240
PW
4105 NETIF_F_IP_CSUM |
4106 NETIF_F_HW_VLAN_TX |
4107 NETIF_F_HW_VLAN_RX |
4108 NETIF_F_HW_VLAN_FILTER;
9a799d71 4109
e9990a9c 4110 netdev->features |= NETIF_F_IPV6_CSUM;
9a799d71 4111 netdev->features |= NETIF_F_TSO;
9a799d71 4112 netdev->features |= NETIF_F_TSO6;
e9990a9c 4113 netdev->features |= NETIF_F_LRO;
ad31c402
JK
4114
4115 netdev->vlan_features |= NETIF_F_TSO;
4116 netdev->vlan_features |= NETIF_F_TSO6;
22f32b7a 4117 netdev->vlan_features |= NETIF_F_IP_CSUM;
ad31c402
JK
4118 netdev->vlan_features |= NETIF_F_SG;
4119
2f90b865
AD
4120 if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
4121 adapter->flags &= ~IXGBE_FLAG_RSS_ENABLED;
4122
4123#ifdef CONFIG_IXGBE_DCBNL
4124 netdev->dcbnl_ops = &dcbnl_ops;
4125#endif
4126
9a799d71
AK
4127 if (pci_using_dac)
4128 netdev->features |= NETIF_F_HIGHDMA;
4129
9a799d71 4130 /* make sure the EEPROM is good */
c44ade9e 4131 if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
9a799d71
AK
4132 dev_err(&pdev->dev, "The EEPROM Checksum Is Not Valid\n");
4133 err = -EIO;
4134 goto err_eeprom;
4135 }
4136
4137 memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
4138 memcpy(netdev->perm_addr, hw->mac.perm_addr, netdev->addr_len);
4139
c44ade9e
JB
4140 if (ixgbe_validate_mac_addr(netdev->perm_addr)) {
4141 dev_err(&pdev->dev, "invalid MAC address\n");
9a799d71
AK
4142 err = -EIO;
4143 goto err_eeprom;
4144 }
4145
4146 init_timer(&adapter->watchdog_timer);
4147 adapter->watchdog_timer.function = &ixgbe_watchdog;
4148 adapter->watchdog_timer.data = (unsigned long)adapter;
4149
4150 INIT_WORK(&adapter->reset_task, ixgbe_reset_task);
cf8280ee 4151 INIT_WORK(&adapter->watchdog_task, ixgbe_watchdog_task);
9a799d71 4152
021230d4
AV
4153 err = ixgbe_init_interrupt_scheme(adapter);
4154 if (err)
4155 goto err_sw_init;
9a799d71
AK
4156
4157 /* print bus type/speed/width info */
4158 pci_read_config_word(pdev, IXGBE_PCI_LINK_STATUS, &link_status);
4159 link_speed = link_status & IXGBE_PCI_LINK_SPEED;
4160 link_width = link_status & IXGBE_PCI_LINK_WIDTH;
7c510e4b 4161 dev_info(&pdev->dev, "(PCI Express:%s:%s) %pM\n",
b4617240
PW
4162 ((link_speed == IXGBE_PCI_LINK_SPEED_5000) ? "5.0Gb/s" :
4163 (link_speed == IXGBE_PCI_LINK_SPEED_2500) ? "2.5Gb/s" :
4164 "Unknown"),
4165 ((link_width == IXGBE_PCI_LINK_WIDTH_8) ? "Width x8" :
4166 (link_width == IXGBE_PCI_LINK_WIDTH_4) ? "Width x4" :
4167 (link_width == IXGBE_PCI_LINK_WIDTH_2) ? "Width x2" :
4168 (link_width == IXGBE_PCI_LINK_WIDTH_1) ? "Width x1" :
4169 "Unknown"),
7c510e4b 4170 netdev->dev_addr);
c44ade9e 4171 ixgbe_read_pba_num_generic(hw, &part_num);
9a799d71 4172 dev_info(&pdev->dev, "MAC: %d, PHY: %d, PBA No: %06x-%03x\n",
b4617240
PW
4173 hw->mac.type, hw->phy.type,
4174 (part_num >> 8), (part_num & 0xff));
9a799d71 4175
0c254d86
AK
4176 if (link_width <= IXGBE_PCI_LINK_WIDTH_4) {
4177 dev_warn(&pdev->dev, "PCI-Express bandwidth available for "
b4617240
PW
4178 "this card is not sufficient for optimal "
4179 "performance.\n");
0c254d86 4180 dev_warn(&pdev->dev, "For optimal performance a x8 "
b4617240 4181 "PCI-Express slot is required.\n");
0c254d86
AK
4182 }
4183
9a799d71 4184 /* reset the hardware with the new settings */
c44ade9e
JB
4185 hw->mac.ops.start_hw(hw);
4186
4187 /* link_config depends on start_hw being called at least once */
4188 err = ixgbe_link_config(hw);
4189 if (err) {
4190 dev_err(&pdev->dev, "setup_link_speed FAILED %d\n", err);
4191 goto err_register;
4192 }
9a799d71
AK
4193
4194 netif_carrier_off(netdev);
2f90b865 4195 netif_tx_stop_all_queues(netdev);
9a799d71 4196
021230d4
AV
4197 ixgbe_napi_add_all(adapter);
4198
9a799d71
AK
4199 strcpy(netdev->name, "eth%d");
4200 err = register_netdev(netdev);
4201 if (err)
4202 goto err_register;
4203
5dd2d332 4204#ifdef CONFIG_IXGBE_DCA
652f093f 4205 if (dca_add_requester(&pdev->dev) == 0) {
bd0362dd
JC
4206 adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
4207 /* always use CB2 mode, difference is masked
4208 * in the CB driver */
4209 IXGBE_WRITE_REG(hw, IXGBE_DCA_CTRL, 2);
4210 ixgbe_setup_dca(adapter);
4211 }
4212#endif
9a799d71
AK
4213
4214 dev_info(&pdev->dev, "Intel(R) 10 Gigabit Network Connection\n");
4215 cards_found++;
4216 return 0;
4217
4218err_register:
5eba3699 4219 ixgbe_release_hw_control(adapter);
9a799d71
AK
4220err_hw_init:
4221err_sw_init:
021230d4 4222 ixgbe_reset_interrupt_capability(adapter);
9a799d71 4223err_eeprom:
c4900be0
DS
4224 clear_bit(__IXGBE_SFP_MODULE_NOT_FOUND, &adapter->state);
4225 del_timer_sync(&adapter->sfp_timer);
4226 cancel_work_sync(&adapter->sfp_task);
9a799d71
AK
4227 iounmap(hw->hw_addr);
4228err_ioremap:
4229 free_netdev(netdev);
4230err_alloc_etherdev:
4231 pci_release_regions(pdev);
4232err_pci_reg:
4233err_dma:
4234 pci_disable_device(pdev);
4235 return err;
4236}
4237
4238/**
4239 * ixgbe_remove - Device Removal Routine
4240 * @pdev: PCI device information struct
4241 *
4242 * ixgbe_remove is called by the PCI subsystem to alert the driver
4243 * that it should release a PCI device. The could be caused by a
4244 * Hot-Plug event, or because the driver is going to be removed from
4245 * memory.
4246 **/
4247static void __devexit ixgbe_remove(struct pci_dev *pdev)
4248{
4249 struct net_device *netdev = pci_get_drvdata(pdev);
4250 struct ixgbe_adapter *adapter = netdev_priv(netdev);
4251
4252 set_bit(__IXGBE_DOWN, &adapter->state);
c4900be0
DS
4253 /* clear the module not found bit to make sure the worker won't
4254 * reschedule
4255 */
4256 clear_bit(__IXGBE_SFP_MODULE_NOT_FOUND, &adapter->state);
9a799d71
AK
4257 del_timer_sync(&adapter->watchdog_timer);
4258
c4900be0
DS
4259 del_timer_sync(&adapter->sfp_timer);
4260 cancel_work_sync(&adapter->watchdog_task);
4261 cancel_work_sync(&adapter->sfp_task);
9a799d71
AK
4262 flush_scheduled_work();
4263
5dd2d332 4264#ifdef CONFIG_IXGBE_DCA
bd0362dd
JC
4265 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
4266 adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
4267 dca_remove_requester(&pdev->dev);
4268 IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
4269 }
4270
4271#endif
c4900be0
DS
4272 if (netdev->reg_state == NETREG_REGISTERED)
4273 unregister_netdev(netdev);
9a799d71 4274
021230d4 4275 ixgbe_reset_interrupt_capability(adapter);
5eba3699 4276
021230d4 4277 ixgbe_release_hw_control(adapter);
9a799d71
AK
4278
4279 iounmap(adapter->hw.hw_addr);
4280 pci_release_regions(pdev);
4281
021230d4 4282 DPRINTK(PROBE, INFO, "complete\n");
b3c8b4ba 4283 ixgbe_napi_del_all(adapter);
021230d4
AV
4284 kfree(adapter->tx_ring);
4285 kfree(adapter->rx_ring);
4286
9a799d71
AK
4287 free_netdev(netdev);
4288
4289 pci_disable_device(pdev);
4290}
4291
4292/**
4293 * ixgbe_io_error_detected - called when PCI error is detected
4294 * @pdev: Pointer to PCI device
4295 * @state: The current pci connection state
4296 *
4297 * This function is called after a PCI bus error affecting
4298 * this device has been detected.
4299 */
4300static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
b4617240 4301 pci_channel_state_t state)
9a799d71
AK
4302{
4303 struct net_device *netdev = pci_get_drvdata(pdev);
454d7c9b 4304 struct ixgbe_adapter *adapter = netdev_priv(netdev);
9a799d71
AK
4305
4306 netif_device_detach(netdev);
4307
4308 if (netif_running(netdev))
4309 ixgbe_down(adapter);
4310 pci_disable_device(pdev);
4311
b4617240 4312 /* Request a slot reset. */
9a799d71
AK
4313 return PCI_ERS_RESULT_NEED_RESET;
4314}
4315
4316/**
4317 * ixgbe_io_slot_reset - called after the pci bus has been reset.
4318 * @pdev: Pointer to PCI device
4319 *
4320 * Restart the card from scratch, as if from a cold-boot.
4321 */
4322static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
4323{
4324 struct net_device *netdev = pci_get_drvdata(pdev);
454d7c9b 4325 struct ixgbe_adapter *adapter = netdev_priv(netdev);
9a799d71
AK
4326
4327 if (pci_enable_device(pdev)) {
4328 DPRINTK(PROBE, ERR,
b4617240 4329 "Cannot re-enable PCI device after reset.\n");
9a799d71
AK
4330 return PCI_ERS_RESULT_DISCONNECT;
4331 }
4332 pci_set_master(pdev);
fb3b27bc 4333 pci_restore_state(pdev);
9a799d71
AK
4334
4335 pci_enable_wake(pdev, PCI_D3hot, 0);
4336 pci_enable_wake(pdev, PCI_D3cold, 0);
4337
4338 ixgbe_reset(adapter);
4339
4340 return PCI_ERS_RESULT_RECOVERED;
4341}
4342
4343/**
4344 * ixgbe_io_resume - called when traffic can start flowing again.
4345 * @pdev: Pointer to PCI device
4346 *
4347 * This callback is called when the error recovery driver tells us that
4348 * its OK to resume normal operation.
4349 */
4350static void ixgbe_io_resume(struct pci_dev *pdev)
4351{
4352 struct net_device *netdev = pci_get_drvdata(pdev);
454d7c9b 4353 struct ixgbe_adapter *adapter = netdev_priv(netdev);
9a799d71
AK
4354
4355 if (netif_running(netdev)) {
4356 if (ixgbe_up(adapter)) {
4357 DPRINTK(PROBE, INFO, "ixgbe_up failed after reset\n");
4358 return;
4359 }
4360 }
4361
4362 netif_device_attach(netdev);
9a799d71
AK
4363}
4364
4365static struct pci_error_handlers ixgbe_err_handler = {
4366 .error_detected = ixgbe_io_error_detected,
4367 .slot_reset = ixgbe_io_slot_reset,
4368 .resume = ixgbe_io_resume,
4369};
4370
4371static struct pci_driver ixgbe_driver = {
4372 .name = ixgbe_driver_name,
4373 .id_table = ixgbe_pci_tbl,
4374 .probe = ixgbe_probe,
4375 .remove = __devexit_p(ixgbe_remove),
4376#ifdef CONFIG_PM
4377 .suspend = ixgbe_suspend,
4378 .resume = ixgbe_resume,
4379#endif
4380 .shutdown = ixgbe_shutdown,
4381 .err_handler = &ixgbe_err_handler
4382};
4383
4384/**
4385 * ixgbe_init_module - Driver Registration Routine
4386 *
4387 * ixgbe_init_module is the first routine called when the driver is
4388 * loaded. All it does is register with the PCI subsystem.
4389 **/
4390static int __init ixgbe_init_module(void)
4391{
4392 int ret;
4393 printk(KERN_INFO "%s: %s - version %s\n", ixgbe_driver_name,
4394 ixgbe_driver_string, ixgbe_driver_version);
4395
4396 printk(KERN_INFO "%s: %s\n", ixgbe_driver_name, ixgbe_copyright);
4397
5dd2d332 4398#ifdef CONFIG_IXGBE_DCA
bd0362dd 4399 dca_register_notify(&dca_notifier);
bd0362dd 4400#endif
5dd2d332 4401
9a799d71
AK
4402 ret = pci_register_driver(&ixgbe_driver);
4403 return ret;
4404}
b4617240 4405
9a799d71
AK
4406module_init(ixgbe_init_module);
4407
4408/**
4409 * ixgbe_exit_module - Driver Exit Cleanup Routine
4410 *
4411 * ixgbe_exit_module is called just before the driver is removed
4412 * from memory.
4413 **/
4414static void __exit ixgbe_exit_module(void)
4415{
5dd2d332 4416#ifdef CONFIG_IXGBE_DCA
bd0362dd
JC
4417 dca_unregister_notify(&dca_notifier);
4418#endif
9a799d71
AK
4419 pci_unregister_driver(&ixgbe_driver);
4420}
bd0362dd 4421
5dd2d332 4422#ifdef CONFIG_IXGBE_DCA
bd0362dd 4423static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
b4617240 4424 void *p)
bd0362dd
JC
4425{
4426 int ret_val;
4427
4428 ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
b4617240 4429 __ixgbe_notify_dca);
bd0362dd
JC
4430
4431 return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
4432}
5dd2d332 4433#endif /* CONFIG_IXGBE_DCA */
bd0362dd 4434
9a799d71
AK
4435module_exit(ixgbe_exit_module);
4436
4437/* ixgbe_main.c */