]> bbs.cooldavid.org Git - net-next-2.6.git/blame - drivers/net/igbvf/netdev.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
igbvf: Remove some dead code in igbvf
[net-next-2.6.git] / drivers / net / igbvf / netdev.c
CommitLineData
d4e0fe01
AD		 1/*******************************************************************************
2
3 Intel(R) 82576 Virtual Function Linux driver
2c20ebba 4 Copyright(c) 2009 - 2010 Intel Corporation.
d4e0fe01
AD		 5
6 This program is free software; you can redistribute it and/or modify it
7 under the terms and conditions of the GNU General Public License,
8 version 2, as published by the Free Software Foundation.
9
10 This program is distributed in the hope it will be useful, but WITHOUT
11 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 more details.
14
15 You should have received a copy of the GNU General Public License along with
16 this program; if not, write to the Free Software Foundation, Inc.,
17 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18
19 The full GNU General Public License is included in this distribution in
20 the file called "COPYING".
21
22 Contact Information:
23 e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
24 Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25
26*******************************************************************************/
27
28#include <linux/module.h>
29#include <linux/types.h>
30#include <linux/init.h>
31#include <linux/pci.h>
32#include <linux/vmalloc.h>
33#include <linux/pagemap.h>
34#include <linux/delay.h>
35#include <linux/netdevice.h>
36#include <linux/tcp.h>
37#include <linux/ipv6.h>
5a0e3ad6 38#include <linux/slab.h>
d4e0fe01
AD		 39#include <net/checksum.h>
40#include <net/ip6_checksum.h>
41#include <linux/mii.h>
42#include <linux/ethtool.h>
43#include <linux/if_vlan.h>
d4e0fe01
AD		 44
45#include "igbvf.h"
46
2c20ebba 47#define DRV_VERSION "1.0.8-k0"
d4e0fe01
AD		 48char igbvf_driver_name[] = "igbvf";
49const char igbvf_driver_version[] = DRV_VERSION;
50static const char igbvf_driver_string[] =
51 "Intel(R) Virtual Function Network Driver";
2c20ebba
GR		 52static const char igbvf_copyright[] =
53 "Copyright (c) 2009 - 2010 Intel Corporation.";
d4e0fe01
AD		 54
55static int igbvf_poll(struct napi_struct *napi, int budget);
2d165771
AD		 56static void igbvf_reset(struct igbvf_adapter *);
57static void igbvf_set_interrupt_capability(struct igbvf_adapter *);
58static void igbvf_reset_interrupt_capability(struct igbvf_adapter *);
d4e0fe01
AD		 59
60static struct igbvf_info igbvf_vf_info = {
61 .mac = e1000_vfadapt,
0364d6fd 62 .flags = 0,
d4e0fe01
AD		 63 .pba = 10,
64 .init_ops = e1000_init_function_pointers_vf,
65};
66
67static const struct igbvf_info *igbvf_info_tbl[] = {
68 [board_vf] = &igbvf_vf_info,
69};
70
71/**
72 * igbvf_desc_unused - calculate if we have unused descriptors
73 **/
74static int igbvf_desc_unused(struct igbvf_ring *ring)
75{
76 if (ring->next_to_clean > ring->next_to_use)
77 return ring->next_to_clean - ring->next_to_use - 1;
78
79 return ring->count + ring->next_to_clean - ring->next_to_use - 1;
80}
81
82/**
83 * igbvf_receive_skb - helper function to handle Rx indications
84 * @adapter: board private structure
85 * @status: descriptor status field as written by hardware
86 * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
87 * @skb: pointer to sk_buff to be indicated to stack
88 **/
89static void igbvf_receive_skb(struct igbvf_adapter *adapter,
90 struct net_device *netdev,
91 struct sk_buff *skb,
92 u32 status, u16 vlan)
93{
94 if (adapter->vlgrp && (status & E1000_RXD_STAT_VP))
95 vlan_hwaccel_receive_skb(skb, adapter->vlgrp,
96 le16_to_cpu(vlan) &
97 E1000_RXD_SPC_VLAN_MASK);
98 else
99 netif_receive_skb(skb);
d4e0fe01
AD		 100}
101
102static inline void igbvf_rx_checksum_adv(struct igbvf_adapter *adapter,
103 u32 status_err, struct sk_buff *skb)
104{
bc8acf2c 105 skb_checksum_none_assert(skb);
d4e0fe01
AD		 106
107 /* Ignore Checksum bit is set or checksum is disabled through ethtool */
0364d6fd
AD		 108 if ((status_err & E1000_RXD_STAT_IXSM) ||
109 (adapter->flags & IGBVF_FLAG_RX_CSUM_DISABLED))
d4e0fe01 110 return;
0364d6fd 111
d4e0fe01
AD		 112 /* TCP/UDP checksum error bit is set */
113 if (status_err &
114 (E1000_RXDEXT_STATERR_TCPE | E1000_RXDEXT_STATERR_IPE)) {
115 /* let the stack verify checksum errors */
116 adapter->hw_csum_err++;
117 return;
118 }
0364d6fd 119
d4e0fe01
AD		 120 /* It must be a TCP or UDP packet with a valid checksum */
121 if (status_err & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS))
122 skb->ip_summed = CHECKSUM_UNNECESSARY;
123
124 adapter->hw_csum_good++;
125}
126
127/**
128 * igbvf_alloc_rx_buffers - Replace used receive buffers; packet split
129 * @rx_ring: address of ring structure to repopulate
130 * @cleaned_count: number of buffers to repopulate
131 **/
132static void igbvf_alloc_rx_buffers(struct igbvf_ring *rx_ring,
133 int cleaned_count)
134{
135 struct igbvf_adapter *adapter = rx_ring->adapter;
136 struct net_device *netdev = adapter->netdev;
137 struct pci_dev *pdev = adapter->pdev;
138 union e1000_adv_rx_desc *rx_desc;
139 struct igbvf_buffer *buffer_info;
140 struct sk_buff *skb;
141 unsigned int i;
142 int bufsz;
143
144 i = rx_ring->next_to_use;
145 buffer_info = &rx_ring->buffer_info[i];
146
147 if (adapter->rx_ps_hdr_size)
148 bufsz = adapter->rx_ps_hdr_size;
149 else
150 bufsz = adapter->rx_buffer_len;
d4e0fe01
AD		 151
152 while (cleaned_count--) {
153 rx_desc = IGBVF_RX_DESC_ADV(*rx_ring, i);
154
155 if (adapter->rx_ps_hdr_size && !buffer_info->page_dma) {
156 if (!buffer_info->page) {
157 buffer_info->page = alloc_page(GFP_ATOMIC);
158 if (!buffer_info->page) {
159 adapter->alloc_rx_buff_failed++;
160 goto no_buffers;
161 }
162 buffer_info->page_offset = 0;
163 } else {
164 buffer_info->page_offset ^= PAGE_SIZE / 2;
165 }
166 buffer_info->page_dma =
123e9f1a 167 dma_map_page(&pdev->dev, buffer_info->page,
d4e0fe01
AD		 168 buffer_info->page_offset,
169 PAGE_SIZE / 2,
123e9f1a 170 DMA_FROM_DEVICE);
d4e0fe01
AD		 171 }
172
173 if (!buffer_info->skb) {
89d71a66 174 skb = netdev_alloc_skb_ip_align(netdev, bufsz);
d4e0fe01
AD		 175 if (!skb) {
176 adapter->alloc_rx_buff_failed++;
177 goto no_buffers;
178 }
179
d4e0fe01 180 buffer_info->skb = skb;
123e9f1a 181 buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
d4e0fe01 182 bufsz,
123e9f1a 183 DMA_FROM_DEVICE);
d4e0fe01
AD		 184 }
185 /* Refresh the desc even if buffer_addrs didn't change because
186 * each write-back erases this info. */
187 if (adapter->rx_ps_hdr_size) {
188 rx_desc->read.pkt_addr =
189 cpu_to_le64(buffer_info->page_dma);
190 rx_desc->read.hdr_addr = cpu_to_le64(buffer_info->dma);
191 } else {
192 rx_desc->read.pkt_addr =
193 cpu_to_le64(buffer_info->dma);
194 rx_desc->read.hdr_addr = 0;
195 }
196
197 i++;
198 if (i == rx_ring->count)
199 i = 0;
200 buffer_info = &rx_ring->buffer_info[i];
201 }
202
203no_buffers:
204 if (rx_ring->next_to_use != i) {
205 rx_ring->next_to_use = i;
206 if (i == 0)
207 i = (rx_ring->count - 1);
208 else
209 i--;
210
211 /* Force memory writes to complete before letting h/w
212 * know there are new descriptors to fetch. (Only
213 * applicable for weak-ordered memory model archs,
214 * such as IA-64). */
215 wmb();
216 writel(i, adapter->hw.hw_addr + rx_ring->tail);
217 }
218}
219
220/**
221 * igbvf_clean_rx_irq - Send received data up the network stack; legacy
222 * @adapter: board private structure
223 *
224 * the return value indicates whether actual cleaning was done, there
225 * is no guarantee that everything was cleaned
226 **/
227static bool igbvf_clean_rx_irq(struct igbvf_adapter *adapter,
228 int *work_done, int work_to_do)
229{
230 struct igbvf_ring *rx_ring = adapter->rx_ring;
231 struct net_device *netdev = adapter->netdev;
232 struct pci_dev *pdev = adapter->pdev;
233 union e1000_adv_rx_desc *rx_desc, *next_rxd;
234 struct igbvf_buffer *buffer_info, *next_buffer;
235 struct sk_buff *skb;
236 bool cleaned = false;
237 int cleaned_count = 0;
238 unsigned int total_bytes = 0, total_packets = 0;
239 unsigned int i;
240 u32 length, hlen, staterr;
241
242 i = rx_ring->next_to_clean;
243 rx_desc = IGBVF_RX_DESC_ADV(*rx_ring, i);
244 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
245
246 while (staterr & E1000_RXD_STAT_DD) {
247 if (*work_done >= work_to_do)
248 break;
249 (*work_done)++;
2d0bb1c1 250 rmb(); /* read descriptor and rx_buffer_info after status DD */
d4e0fe01
AD		 251
252 buffer_info = &rx_ring->buffer_info[i];
253
254 /* HW will not DMA in data larger than the given buffer, even
255 * if it parses the (NFS, of course) header to be larger. In
256 * that case, it fills the header buffer and spills the rest
257 * into the page.
258 */
259 hlen = (le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.hdr_info) &
260 E1000_RXDADV_HDRBUFLEN_MASK) >> E1000_RXDADV_HDRBUFLEN_SHIFT;
261 if (hlen > adapter->rx_ps_hdr_size)
262 hlen = adapter->rx_ps_hdr_size;
263
264 length = le16_to_cpu(rx_desc->wb.upper.length);
265 cleaned = true;
266 cleaned_count++;
267
268 skb = buffer_info->skb;
269 prefetch(skb->data - NET_IP_ALIGN);
270 buffer_info->skb = NULL;
271 if (!adapter->rx_ps_hdr_size) {
123e9f1a 272 dma_unmap_single(&pdev->dev, buffer_info->dma,
d4e0fe01 273 adapter->rx_buffer_len,
123e9f1a 274 DMA_FROM_DEVICE);
d4e0fe01
AD		 275 buffer_info->dma = 0;
276 skb_put(skb, length);
277 goto send_up;
278 }
279
280 if (!skb_shinfo(skb)->nr_frags) {
123e9f1a 281 dma_unmap_single(&pdev->dev, buffer_info->dma,
92d947b7 282 adapter->rx_ps_hdr_size,
123e9f1a 283 DMA_FROM_DEVICE);
d4e0fe01
AD		 284 skb_put(skb, hlen);
285 }
286
287 if (length) {
123e9f1a 288 dma_unmap_page(&pdev->dev, buffer_info->page_dma,
d4e0fe01 289 PAGE_SIZE / 2,
123e9f1a 290 DMA_FROM_DEVICE);
d4e0fe01
AD		 291 buffer_info->page_dma = 0;
292
ec857fd4 293 skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
d4e0fe01
AD		 294 buffer_info->page,
295 buffer_info->page_offset,
296 length);
297
298 if ((adapter->rx_buffer_len > (PAGE_SIZE / 2)) ||
299 (page_count(buffer_info->page) != 1))
300 buffer_info->page = NULL;
301 else
302 get_page(buffer_info->page);
303
304 skb->len += length;
305 skb->data_len += length;
306 skb->truesize += length;
307 }
308send_up:
309 i++;
310 if (i == rx_ring->count)
311 i = 0;
312 next_rxd = IGBVF_RX_DESC_ADV(*rx_ring, i);
313 prefetch(next_rxd);
314 next_buffer = &rx_ring->buffer_info[i];
315
316 if (!(staterr & E1000_RXD_STAT_EOP)) {
317 buffer_info->skb = next_buffer->skb;
318 buffer_info->dma = next_buffer->dma;
319 next_buffer->skb = skb;
320 next_buffer->dma = 0;
321 goto next_desc;
322 }
323
324 if (staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) {
325 dev_kfree_skb_irq(skb);
326 goto next_desc;
327 }
328
329 total_bytes += skb->len;
330 total_packets++;
331
332 igbvf_rx_checksum_adv(adapter, staterr, skb);
333
334 skb->protocol = eth_type_trans(skb, netdev);
335
336 igbvf_receive_skb(adapter, netdev, skb, staterr,
337 rx_desc->wb.upper.vlan);
338
d4e0fe01
AD		 339next_desc:
340 rx_desc->wb.upper.status_error = 0;
341
342 /* return some buffers to hardware, one at a time is too slow */
343 if (cleaned_count >= IGBVF_RX_BUFFER_WRITE) {
344 igbvf_alloc_rx_buffers(rx_ring, cleaned_count);
345 cleaned_count = 0;
346 }
347
348 /* use prefetched values */
349 rx_desc = next_rxd;
350 buffer_info = next_buffer;
351
352 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
353 }
354
355 rx_ring->next_to_clean = i;
356 cleaned_count = igbvf_desc_unused(rx_ring);
357
358 if (cleaned_count)
359 igbvf_alloc_rx_buffers(rx_ring, cleaned_count);
360
361 adapter->total_rx_packets += total_packets;
362 adapter->total_rx_bytes += total_bytes;
363 adapter->net_stats.rx_bytes += total_bytes;
364 adapter->net_stats.rx_packets += total_packets;
365 return cleaned;
366}
367
368static void igbvf_put_txbuf(struct igbvf_adapter *adapter,
369 struct igbvf_buffer *buffer_info)
370{
a7d5ca40
AD		 371 if (buffer_info->dma) {
372 if (buffer_info->mapped_as_page)
123e9f1a 373 dma_unmap_page(&adapter->pdev->dev,
a7d5ca40
AD		 374 buffer_info->dma,
375 buffer_info->length,
123e9f1a 376 DMA_TO_DEVICE);
a7d5ca40 377 else
123e9f1a 378 dma_unmap_single(&adapter->pdev->dev,
a7d5ca40
AD		 379 buffer_info->dma,
380 buffer_info->length,
123e9f1a 381 DMA_TO_DEVICE);
a7d5ca40
AD		 382 buffer_info->dma = 0;
383 }
d4e0fe01 384 if (buffer_info->skb) {
d4e0fe01
AD		 385 dev_kfree_skb_any(buffer_info->skb);
386 buffer_info->skb = NULL;
387 }
388 buffer_info->time_stamp = 0;
389}
390
391static void igbvf_print_tx_hang(struct igbvf_adapter *adapter)
392{
393 struct igbvf_ring *tx_ring = adapter->tx_ring;
394 unsigned int i = tx_ring->next_to_clean;
395 unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
396 union e1000_adv_tx_desc *eop_desc = IGBVF_TX_DESC_ADV(*tx_ring, eop);
397
398 /* detected Tx unit hang */
399 dev_err(&adapter->pdev->dev,
400 "Detected Tx Unit Hang:\n"
401 " TDH <%x>\n"
402 " TDT <%x>\n"
403 " next_to_use <%x>\n"
404 " next_to_clean <%x>\n"
405 "buffer_info[next_to_clean]:\n"
406 " time_stamp <%lx>\n"
407 " next_to_watch <%x>\n"
408 " jiffies <%lx>\n"
409 " next_to_watch.status <%x>\n",
410 readl(adapter->hw.hw_addr + tx_ring->head),
411 readl(adapter->hw.hw_addr + tx_ring->tail),
412 tx_ring->next_to_use,
413 tx_ring->next_to_clean,
414 tx_ring->buffer_info[eop].time_stamp,
415 eop,
416 jiffies,
417 eop_desc->wb.status);
418}
419
420/**
421 * igbvf_setup_tx_resources - allocate Tx resources (Descriptors)
422 * @adapter: board private structure
423 *
424 * Return 0 on success, negative on failure
425 **/
426int igbvf_setup_tx_resources(struct igbvf_adapter *adapter,
427 struct igbvf_ring *tx_ring)
428{
429 struct pci_dev *pdev = adapter->pdev;
430 int size;
431
432 size = sizeof(struct igbvf_buffer) * tx_ring->count;
433 tx_ring->buffer_info = vmalloc(size);
434 if (!tx_ring->buffer_info)
435 goto err;
436 memset(tx_ring->buffer_info, 0, size);
437
438 /* round up to nearest 4K */
439 tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
440 tx_ring->size = ALIGN(tx_ring->size, 4096);
441
123e9f1a
NN		 442 tx_ring->desc = dma_alloc_coherent(&pdev->dev, tx_ring->size,
443 &tx_ring->dma, GFP_KERNEL);
d4e0fe01
AD		 444
445 if (!tx_ring->desc)
446 goto err;
447
448 tx_ring->adapter = adapter;
449 tx_ring->next_to_use = 0;
450 tx_ring->next_to_clean = 0;
451
452 return 0;
453err:
454 vfree(tx_ring->buffer_info);
455 dev_err(&adapter->pdev->dev,
456 "Unable to allocate memory for the transmit descriptor ring\n");
457 return -ENOMEM;
458}
459
460/**
461 * igbvf_setup_rx_resources - allocate Rx resources (Descriptors)
462 * @adapter: board private structure
463 *
464 * Returns 0 on success, negative on failure
465 **/
466int igbvf_setup_rx_resources(struct igbvf_adapter *adapter,
467 struct igbvf_ring *rx_ring)
468{
469 struct pci_dev *pdev = adapter->pdev;
470 int size, desc_len;
471
472 size = sizeof(struct igbvf_buffer) * rx_ring->count;
473 rx_ring->buffer_info = vmalloc(size);
474 if (!rx_ring->buffer_info)
475 goto err;
476 memset(rx_ring->buffer_info, 0, size);
477
478 desc_len = sizeof(union e1000_adv_rx_desc);
479
480 /* Round up to nearest 4K */
481 rx_ring->size = rx_ring->count * desc_len;
482 rx_ring->size = ALIGN(rx_ring->size, 4096);
483
123e9f1a
NN		 484 rx_ring->desc = dma_alloc_coherent(&pdev->dev, rx_ring->size,
485 &rx_ring->dma, GFP_KERNEL);
d4e0fe01
AD		 486
487 if (!rx_ring->desc)
488 goto err;
489
490 rx_ring->next_to_clean = 0;
491 rx_ring->next_to_use = 0;
492
493 rx_ring->adapter = adapter;
494
495 return 0;
496
497err:
498 vfree(rx_ring->buffer_info);
499 rx_ring->buffer_info = NULL;
500 dev_err(&adapter->pdev->dev,
501 "Unable to allocate memory for the receive descriptor ring\n");
502 return -ENOMEM;
503}
504
505/**
506 * igbvf_clean_tx_ring - Free Tx Buffers
507 * @tx_ring: ring to be cleaned
508 **/
509static void igbvf_clean_tx_ring(struct igbvf_ring *tx_ring)
510{
511 struct igbvf_adapter *adapter = tx_ring->adapter;
512 struct igbvf_buffer *buffer_info;
513 unsigned long size;
514 unsigned int i;
515
516 if (!tx_ring->buffer_info)
517 return;
518
519 /* Free all the Tx ring sk_buffs */
520 for (i = 0; i < tx_ring->count; i++) {
521 buffer_info = &tx_ring->buffer_info[i];
522 igbvf_put_txbuf(adapter, buffer_info);
523 }
524
525 size = sizeof(struct igbvf_buffer) * tx_ring->count;
526 memset(tx_ring->buffer_info, 0, size);
527
528 /* Zero out the descriptor ring */
529 memset(tx_ring->desc, 0, tx_ring->size);
530
531 tx_ring->next_to_use = 0;
532 tx_ring->next_to_clean = 0;
533
534 writel(0, adapter->hw.hw_addr + tx_ring->head);
535 writel(0, adapter->hw.hw_addr + tx_ring->tail);
536}
537
538/**
539 * igbvf_free_tx_resources - Free Tx Resources per Queue
540 * @tx_ring: ring to free resources from
541 *
542 * Free all transmit software resources
543 **/
544void igbvf_free_tx_resources(struct igbvf_ring *tx_ring)
545{
546 struct pci_dev *pdev = tx_ring->adapter->pdev;
547
548 igbvf_clean_tx_ring(tx_ring);
549
550 vfree(tx_ring->buffer_info);
551 tx_ring->buffer_info = NULL;
552
123e9f1a
NN		 553 dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
554 tx_ring->dma);
d4e0fe01
AD		 555
556 tx_ring->desc = NULL;
557}
558
559/**
560 * igbvf_clean_rx_ring - Free Rx Buffers per Queue
561 * @adapter: board private structure
562 **/
563static void igbvf_clean_rx_ring(struct igbvf_ring *rx_ring)
564{
565 struct igbvf_adapter *adapter = rx_ring->adapter;
566 struct igbvf_buffer *buffer_info;
567 struct pci_dev *pdev = adapter->pdev;
568 unsigned long size;
569 unsigned int i;
570
571 if (!rx_ring->buffer_info)
572 return;
573
574 /* Free all the Rx ring sk_buffs */
575 for (i = 0; i < rx_ring->count; i++) {
576 buffer_info = &rx_ring->buffer_info[i];
577 if (buffer_info->dma) {
578 if (adapter->rx_ps_hdr_size){
123e9f1a 579 dma_unmap_single(&pdev->dev, buffer_info->dma,
d4e0fe01 580 adapter->rx_ps_hdr_size,
123e9f1a 581 DMA_FROM_DEVICE);
d4e0fe01 582 } else {
123e9f1a 583 dma_unmap_single(&pdev->dev, buffer_info->dma,
d4e0fe01 584 adapter->rx_buffer_len,
123e9f1a 585 DMA_FROM_DEVICE);
d4e0fe01
AD		 586 }
587 buffer_info->dma = 0;
588 }
589
590 if (buffer_info->skb) {
591 dev_kfree_skb(buffer_info->skb);
592 buffer_info->skb = NULL;
593 }
594
595 if (buffer_info->page) {
596 if (buffer_info->page_dma)
123e9f1a
NN		 597 dma_unmap_page(&pdev->dev,
598 buffer_info->page_dma,
d4e0fe01 599 PAGE_SIZE / 2,
123e9f1a 600 DMA_FROM_DEVICE);
d4e0fe01
AD		 601 put_page(buffer_info->page);
602 buffer_info->page = NULL;
603 buffer_info->page_dma = 0;
604 buffer_info->page_offset = 0;
605 }
606 }
607
608 size = sizeof(struct igbvf_buffer) * rx_ring->count;
609 memset(rx_ring->buffer_info, 0, size);
610
611 /* Zero out the descriptor ring */
612 memset(rx_ring->desc, 0, rx_ring->size);
613
614 rx_ring->next_to_clean = 0;
615 rx_ring->next_to_use = 0;
616
617 writel(0, adapter->hw.hw_addr + rx_ring->head);
618 writel(0, adapter->hw.hw_addr + rx_ring->tail);
619}
620
621/**
622 * igbvf_free_rx_resources - Free Rx Resources
623 * @rx_ring: ring to clean the resources from
624 *
625 * Free all receive software resources
626 **/
627
628void igbvf_free_rx_resources(struct igbvf_ring *rx_ring)
629{
630 struct pci_dev *pdev = rx_ring->adapter->pdev;
631
632 igbvf_clean_rx_ring(rx_ring);
633
634 vfree(rx_ring->buffer_info);
635 rx_ring->buffer_info = NULL;
636
637 dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
638 rx_ring->dma);
639 rx_ring->desc = NULL;
640}
641
642/**
643 * igbvf_update_itr - update the dynamic ITR value based on statistics
644 * @adapter: pointer to adapter
645 * @itr_setting: current adapter->itr
646 * @packets: the number of packets during this measurement interval
647 * @bytes: the number of bytes during this measurement interval
648 *
649 * Stores a new ITR value based on packets and byte
650 * counts during the last interrupt. The advantage of per interrupt
651 * computation is faster updates and more accurate ITR for the current
652 * traffic pattern. Constants in this function were computed
653 * based on theoretical maximum wire speed and thresholds were set based
654 * on testing data as well as attempting to minimize response time
655 * while increasing bulk throughput. This functionality is controlled
656 * by the InterruptThrottleRate module parameter.
657 **/
658static unsigned int igbvf_update_itr(struct igbvf_adapter *adapter,
659 u16 itr_setting, int packets,
660 int bytes)
661{
662 unsigned int retval = itr_setting;
663
664 if (packets == 0)
665 goto update_itr_done;
666
667 switch (itr_setting) {
668 case lowest_latency:
669 /* handle TSO and jumbo frames */
670 if (bytes/packets > 8000)
671 retval = bulk_latency;
672 else if ((packets < 5) && (bytes > 512))
673 retval = low_latency;
674 break;
675 case low_latency: /* 50 usec aka 20000 ints/s */
676 if (bytes > 10000) {
677 /* this if handles the TSO accounting */
678 if (bytes/packets > 8000)
679 retval = bulk_latency;
680 else if ((packets < 10) || ((bytes/packets) > 1200))
681 retval = bulk_latency;
682 else if ((packets > 35))
683 retval = lowest_latency;
684 } else if (bytes/packets > 2000) {
685 retval = bulk_latency;
686 } else if (packets <= 2 && bytes < 512) {
687 retval = lowest_latency;
688 }
689 break;
690 case bulk_latency: /* 250 usec aka 4000 ints/s */
691 if (bytes > 25000) {
692 if (packets > 35)
693 retval = low_latency;
694 } else if (bytes < 6000) {
695 retval = low_latency;
696 }
697 break;
698 }
699
700update_itr_done:
701 return retval;
702}
703
704static void igbvf_set_itr(struct igbvf_adapter *adapter)
705{
706 struct e1000_hw *hw = &adapter->hw;
707 u16 current_itr;
708 u32 new_itr = adapter->itr;
709
710 adapter->tx_itr = igbvf_update_itr(adapter, adapter->tx_itr,
711 adapter->total_tx_packets,
712 adapter->total_tx_bytes);
713 /* conservative mode (itr 3) eliminates the lowest_latency setting */
714 if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
715 adapter->tx_itr = low_latency;
716
717 adapter->rx_itr = igbvf_update_itr(adapter, adapter->rx_itr,
718 adapter->total_rx_packets,
719 adapter->total_rx_bytes);
720 /* conservative mode (itr 3) eliminates the lowest_latency setting */
721 if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
722 adapter->rx_itr = low_latency;
723
724 current_itr = max(adapter->rx_itr, adapter->tx_itr);
725
726 switch (current_itr) {
727 /* counts and packets in update_itr are dependent on these numbers */
728 case lowest_latency:
729 new_itr = 70000;
730 break;
731 case low_latency:
732 new_itr = 20000; /* aka hwitr = ~200 */
733 break;
734 case bulk_latency:
735 new_itr = 4000;
736 break;
737 default:
738 break;
739 }
740
741 if (new_itr != adapter->itr) {
742 /*
743 * this attempts to bias the interrupt rate towards Bulk
744 * by adding intermediate steps when interrupt rate is
745 * increasing
746 */
747 new_itr = new_itr > adapter->itr ?
748 min(adapter->itr + (new_itr >> 2), new_itr) :
749 new_itr;
750 adapter->itr = new_itr;
751 adapter->rx_ring->itr_val = 1952;
752
753 if (adapter->msix_entries)
754 adapter->rx_ring->set_itr = 1;
755 else
756 ew32(ITR, 1952);
757 }
758}
759
760/**
761 * igbvf_clean_tx_irq - Reclaim resources after transmit completes
762 * @adapter: board private structure
763 * returns true if ring is completely cleaned
764 **/
765static bool igbvf_clean_tx_irq(struct igbvf_ring *tx_ring)
766{
767 struct igbvf_adapter *adapter = tx_ring->adapter;
768 struct e1000_hw *hw = &adapter->hw;
769 struct net_device *netdev = adapter->netdev;
770 struct igbvf_buffer *buffer_info;
771 struct sk_buff *skb;
772 union e1000_adv_tx_desc *tx_desc, *eop_desc;
773 unsigned int total_bytes = 0, total_packets = 0;
774 unsigned int i, eop, count = 0;
775 bool cleaned = false;
776
777 i = tx_ring->next_to_clean;
778 eop = tx_ring->buffer_info[i].next_to_watch;
779 eop_desc = IGBVF_TX_DESC_ADV(*tx_ring, eop);
780
781 while ((eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)) &&
782 (count < tx_ring->count)) {
2d0bb1c1 783 rmb(); /* read buffer_info after eop_desc status */
d4e0fe01
AD		 784 for (cleaned = false; !cleaned; count++) {
785 tx_desc = IGBVF_TX_DESC_ADV(*tx_ring, i);
786 buffer_info = &tx_ring->buffer_info[i];
787 cleaned = (i == eop);
788 skb = buffer_info->skb;
789
790 if (skb) {
791 unsigned int segs, bytecount;
792
793 /* gso_segs is currently only valid for tcp */
794 segs = skb_shinfo(skb)->gso_segs ?: 1;
795 /* multiply data chunks by size of headers */
796 bytecount = ((segs - 1) * skb_headlen(skb)) +
797 skb->len;
798 total_packets += segs;
799 total_bytes += bytecount;
800 }
801
802 igbvf_put_txbuf(adapter, buffer_info);
803 tx_desc->wb.status = 0;
804
805 i++;
806 if (i == tx_ring->count)
807 i = 0;
808 }
809 eop = tx_ring->buffer_info[i].next_to_watch;
810 eop_desc = IGBVF_TX_DESC_ADV(*tx_ring, eop);
811 }
812
813 tx_ring->next_to_clean = i;
814
815 if (unlikely(count &&
816 netif_carrier_ok(netdev) &&
817 igbvf_desc_unused(tx_ring) >= IGBVF_TX_QUEUE_WAKE)) {
818 /* Make sure that anybody stopping the queue after this
819 * sees the new next_to_clean.
820 */
821 smp_mb();
822 if (netif_queue_stopped(netdev) &&
823 !(test_bit(__IGBVF_DOWN, &adapter->state))) {
824 netif_wake_queue(netdev);
825 ++adapter->restart_queue;
826 }
827 }
828
829 if (adapter->detect_tx_hung) {
830 /* Detect a transmit hang in hardware, this serializes the
831 * check with the clearing of time_stamp and movement of i */
832 adapter->detect_tx_hung = false;
833 if (tx_ring->buffer_info[i].time_stamp &&
834 time_after(jiffies, tx_ring->buffer_info[i].time_stamp +
8e95a202
JP		 835 (adapter->tx_timeout_factor * HZ)) &&
836 !(er32(STATUS) & E1000_STATUS_TXOFF)) {
d4e0fe01
AD		 837
838 tx_desc = IGBVF_TX_DESC_ADV(*tx_ring, i);
839 /* detected Tx unit hang */
840 igbvf_print_tx_hang(adapter);
841
842 netif_stop_queue(netdev);
843 }
844 }
845 adapter->net_stats.tx_bytes += total_bytes;
846 adapter->net_stats.tx_packets += total_packets;
807540ba 847 return count < tx_ring->count;
d4e0fe01
AD		 848}
849
850static irqreturn_t igbvf_msix_other(int irq, void *data)
851{
852 struct net_device *netdev = data;
853 struct igbvf_adapter *adapter = netdev_priv(netdev);
854 struct e1000_hw *hw = &adapter->hw;
855
856 adapter->int_counter1++;
857
858 netif_carrier_off(netdev);
859 hw->mac.get_link_status = 1;
860 if (!test_bit(__IGBVF_DOWN, &adapter->state))
861 mod_timer(&adapter->watchdog_timer, jiffies + 1);
862
863 ew32(EIMS, adapter->eims_other);
864
865 return IRQ_HANDLED;
866}
867
868static irqreturn_t igbvf_intr_msix_tx(int irq, void *data)
869{
870 struct net_device *netdev = data;
871 struct igbvf_adapter *adapter = netdev_priv(netdev);
872 struct e1000_hw *hw = &adapter->hw;
873 struct igbvf_ring *tx_ring = adapter->tx_ring;
874
875
876 adapter->total_tx_bytes = 0;
877 adapter->total_tx_packets = 0;
878
879 /* auto mask will automatically reenable the interrupt when we write
880 * EICS */
881 if (!igbvf_clean_tx_irq(tx_ring))
882 /* Ring was not completely cleaned, so fire another interrupt */
883 ew32(EICS, tx_ring->eims_value);
884 else
885 ew32(EIMS, tx_ring->eims_value);
886
887 return IRQ_HANDLED;
888}
889
890static irqreturn_t igbvf_intr_msix_rx(int irq, void *data)
891{
892 struct net_device *netdev = data;
893 struct igbvf_adapter *adapter = netdev_priv(netdev);
894
895 adapter->int_counter0++;
896
897 /* Write the ITR value calculated at the end of the
898 * previous interrupt.
899 */
900 if (adapter->rx_ring->set_itr) {
901 writel(adapter->rx_ring->itr_val,
902 adapter->hw.hw_addr + adapter->rx_ring->itr_register);
903 adapter->rx_ring->set_itr = 0;
904 }
905
906 if (napi_schedule_prep(&adapter->rx_ring->napi)) {
907 adapter->total_rx_bytes = 0;
908 adapter->total_rx_packets = 0;
909 __napi_schedule(&adapter->rx_ring->napi);
910 }
911
912 return IRQ_HANDLED;
913}
914
915#define IGBVF_NO_QUEUE -1
916
917static void igbvf_assign_vector(struct igbvf_adapter *adapter, int rx_queue,
918 int tx_queue, int msix_vector)
919{
920 struct e1000_hw *hw = &adapter->hw;
921 u32 ivar, index;
922
923 /* 82576 uses a table-based method for assigning vectors.
924 Each queue has a single entry in the table to which we write
925 a vector number along with a "valid" bit. Sadly, the layout
926 of the table is somewhat counterintuitive. */
927 if (rx_queue > IGBVF_NO_QUEUE) {
928 index = (rx_queue >> 1);
929 ivar = array_er32(IVAR0, index);
930 if (rx_queue & 0x1) {
931 /* vector goes into third byte of register */
932 ivar = ivar & 0xFF00FFFF;
933 ivar |= (msix_vector | E1000_IVAR_VALID) << 16;
934 } else {
935 /* vector goes into low byte of register */
936 ivar = ivar & 0xFFFFFF00;
937 ivar |= msix_vector | E1000_IVAR_VALID;
938 }
939 adapter->rx_ring[rx_queue].eims_value = 1 << msix_vector;
940 array_ew32(IVAR0, index, ivar);
941 }
942 if (tx_queue > IGBVF_NO_QUEUE) {
943 index = (tx_queue >> 1);
944 ivar = array_er32(IVAR0, index);
945 if (tx_queue & 0x1) {
946 /* vector goes into high byte of register */
947 ivar = ivar & 0x00FFFFFF;
948 ivar |= (msix_vector | E1000_IVAR_VALID) << 24;
949 } else {
950 /* vector goes into second byte of register */
951 ivar = ivar & 0xFFFF00FF;
952 ivar |= (msix_vector | E1000_IVAR_VALID) << 8;
953 }
954 adapter->tx_ring[tx_queue].eims_value = 1 << msix_vector;
955 array_ew32(IVAR0, index, ivar);
956 }
957}
958
959/**
960 * igbvf_configure_msix - Configure MSI-X hardware
961 *
962 * igbvf_configure_msix sets up the hardware to properly
963 * generate MSI-X interrupts.
964 **/
965static void igbvf_configure_msix(struct igbvf_adapter *adapter)
966{
967 u32 tmp;
968 struct e1000_hw *hw = &adapter->hw;
969 struct igbvf_ring *tx_ring = adapter->tx_ring;
970 struct igbvf_ring *rx_ring = adapter->rx_ring;
971 int vector = 0;
972
973 adapter->eims_enable_mask = 0;
974
975 igbvf_assign_vector(adapter, IGBVF_NO_QUEUE, 0, vector++);
976 adapter->eims_enable_mask |= tx_ring->eims_value;
977 if (tx_ring->itr_val)
978 writel(tx_ring->itr_val,
979 hw->hw_addr + tx_ring->itr_register);
980 else
981 writel(1952, hw->hw_addr + tx_ring->itr_register);
982
983 igbvf_assign_vector(adapter, 0, IGBVF_NO_QUEUE, vector++);
984 adapter->eims_enable_mask |= rx_ring->eims_value;
985 if (rx_ring->itr_val)
986 writel(rx_ring->itr_val,
987 hw->hw_addr + rx_ring->itr_register);
988 else
989 writel(1952, hw->hw_addr + rx_ring->itr_register);
990
991 /* set vector for other causes, i.e. link changes */
992
993 tmp = (vector++ | E1000_IVAR_VALID);
994
995 ew32(IVAR_MISC, tmp);
996
997 adapter->eims_enable_mask = (1 << (vector)) - 1;
998 adapter->eims_other = 1 << (vector - 1);
999 e1e_flush();
1000}
1001
2d165771 1002static void igbvf_reset_interrupt_capability(struct igbvf_adapter *adapter)
d4e0fe01
AD		 1003{
1004 if (adapter->msix_entries) {
1005 pci_disable_msix(adapter->pdev);
1006 kfree(adapter->msix_entries);
1007 adapter->msix_entries = NULL;
1008 }
1009}
1010
1011/**
1012 * igbvf_set_interrupt_capability - set MSI or MSI-X if supported
1013 *
1014 * Attempt to configure interrupts using the best available
1015 * capabilities of the hardware and kernel.
1016 **/
2d165771 1017static void igbvf_set_interrupt_capability(struct igbvf_adapter *adapter)
d4e0fe01
AD		 1018{
1019 int err = -ENOMEM;
1020 int i;
1021
1022 /* we allocate 3 vectors, 1 for tx, 1 for rx, one for pf messages */
1023 adapter->msix_entries = kcalloc(3, sizeof(struct msix_entry),
1024 GFP_KERNEL);
1025 if (adapter->msix_entries) {
1026 for (i = 0; i < 3; i++)
1027 adapter->msix_entries[i].entry = i;
1028
1029 err = pci_enable_msix(adapter->pdev,
1030 adapter->msix_entries, 3);
1031 }
1032
1033 if (err) {
1034 /* MSI-X failed */
1035 dev_err(&adapter->pdev->dev,
1036 "Failed to initialize MSI-X interrupts.\n");
1037 igbvf_reset_interrupt_capability(adapter);
1038 }
1039}
1040
1041/**
1042 * igbvf_request_msix - Initialize MSI-X interrupts
1043 *
1044 * igbvf_request_msix allocates MSI-X vectors and requests interrupts from the
1045 * kernel.
1046 **/
1047static int igbvf_request_msix(struct igbvf_adapter *adapter)
1048{
1049 struct net_device *netdev = adapter->netdev;
1050 int err = 0, vector = 0;
1051
1052 if (strlen(netdev->name) < (IFNAMSIZ - 5)) {
1053 sprintf(adapter->tx_ring->name, "%s-tx-0", netdev->name);
1054 sprintf(adapter->rx_ring->name, "%s-rx-0", netdev->name);
1055 } else {
1056 memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
1057 memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
1058 }
1059
1060 err = request_irq(adapter->msix_entries[vector].vector,
a0607fd3 1061 igbvf_intr_msix_tx, 0, adapter->tx_ring->name,
d4e0fe01
AD		 1062 netdev);
1063 if (err)
1064 goto out;
1065
1066 adapter->tx_ring->itr_register = E1000_EITR(vector);
1067 adapter->tx_ring->itr_val = 1952;
1068 vector++;
1069
1070 err = request_irq(adapter->msix_entries[vector].vector,
a0607fd3 1071 igbvf_intr_msix_rx, 0, adapter->rx_ring->name,
d4e0fe01
AD		 1072 netdev);
1073 if (err)
1074 goto out;
1075
1076 adapter->rx_ring->itr_register = E1000_EITR(vector);
1077 adapter->rx_ring->itr_val = 1952;
1078 vector++;
1079
1080 err = request_irq(adapter->msix_entries[vector].vector,
a0607fd3 1081 igbvf_msix_other, 0, netdev->name, netdev);
d4e0fe01
AD		 1082 if (err)
1083 goto out;
1084
1085 igbvf_configure_msix(adapter);
1086 return 0;
1087out:
1088 return err;
1089}
1090
1091/**
1092 * igbvf_alloc_queues - Allocate memory for all rings
1093 * @adapter: board private structure to initialize
1094 **/
1095static int __devinit igbvf_alloc_queues(struct igbvf_adapter *adapter)
1096{
1097 struct net_device *netdev = adapter->netdev;
1098
1099 adapter->tx_ring = kzalloc(sizeof(struct igbvf_ring), GFP_KERNEL);
1100 if (!adapter->tx_ring)
1101 return -ENOMEM;
1102
1103 adapter->rx_ring = kzalloc(sizeof(struct igbvf_ring), GFP_KERNEL);
1104 if (!adapter->rx_ring) {
1105 kfree(adapter->tx_ring);
1106 return -ENOMEM;
1107 }
1108
1109 netif_napi_add(netdev, &adapter->rx_ring->napi, igbvf_poll, 64);
1110
1111 return 0;
1112}
1113
1114/**
1115 * igbvf_request_irq - initialize interrupts
1116 *
1117 * Attempts to configure interrupts using the best available
1118 * capabilities of the hardware and kernel.
1119 **/
1120static int igbvf_request_irq(struct igbvf_adapter *adapter)
1121{
1122 int err = -1;
1123
1124 /* igbvf supports msi-x only */
1125 if (adapter->msix_entries)
1126 err = igbvf_request_msix(adapter);
1127
1128 if (!err)
1129 return err;
1130
1131 dev_err(&adapter->pdev->dev,
1132 "Unable to allocate interrupt, Error: %d\n", err);
1133
1134 return err;
1135}
1136
1137static void igbvf_free_irq(struct igbvf_adapter *adapter)
1138{
1139 struct net_device *netdev = adapter->netdev;
1140 int vector;
1141
1142 if (adapter->msix_entries) {
1143 for (vector = 0; vector < 3; vector++)
1144 free_irq(adapter->msix_entries[vector].vector, netdev);
1145 }
1146}
1147
1148/**
1149 * igbvf_irq_disable - Mask off interrupt generation on the NIC
1150 **/
1151static void igbvf_irq_disable(struct igbvf_adapter *adapter)
1152{
1153 struct e1000_hw *hw = &adapter->hw;
1154
1155 ew32(EIMC, ~0);
1156
1157 if (adapter->msix_entries)
1158 ew32(EIAC, 0);
1159}
1160
1161/**
1162 * igbvf_irq_enable - Enable default interrupt generation settings
1163 **/
1164static void igbvf_irq_enable(struct igbvf_adapter *adapter)
1165{
1166 struct e1000_hw *hw = &adapter->hw;
1167
1168 ew32(EIAC, adapter->eims_enable_mask);
1169 ew32(EIAM, adapter->eims_enable_mask);
1170 ew32(EIMS, adapter->eims_enable_mask);
1171}
1172
1173/**
1174 * igbvf_poll - NAPI Rx polling callback
1175 * @napi: struct associated with this polling callback
1176 * @budget: amount of packets driver is allowed to process this poll
1177 **/
1178static int igbvf_poll(struct napi_struct *napi, int budget)
1179{
1180 struct igbvf_ring *rx_ring = container_of(napi, struct igbvf_ring, napi);
1181 struct igbvf_adapter *adapter = rx_ring->adapter;
1182 struct e1000_hw *hw = &adapter->hw;
1183 int work_done = 0;
1184
1185 igbvf_clean_rx_irq(adapter, &work_done, budget);
1186
1187 /* If not enough Rx work done, exit the polling mode */
1188 if (work_done < budget) {
1189 napi_complete(napi);
1190
1191 if (adapter->itr_setting & 3)
1192 igbvf_set_itr(adapter);
1193
1194 if (!test_bit(__IGBVF_DOWN, &adapter->state))
1195 ew32(EIMS, adapter->rx_ring->eims_value);
1196 }
1197
1198 return work_done;
1199}
1200
1201/**
1202 * igbvf_set_rlpml - set receive large packet maximum length
1203 * @adapter: board private structure
1204 *
1205 * Configure the maximum size of packets that will be received
1206 */
1207static void igbvf_set_rlpml(struct igbvf_adapter *adapter)
1208{
1209 int max_frame_size = adapter->max_frame_size;
1210 struct e1000_hw *hw = &adapter->hw;
1211
1212 if (adapter->vlgrp)
1213 max_frame_size += VLAN_TAG_SIZE;
1214
1215 e1000_rlpml_set_vf(hw, max_frame_size);
1216}
1217
1218static void igbvf_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
1219{
1220 struct igbvf_adapter *adapter = netdev_priv(netdev);
1221 struct e1000_hw *hw = &adapter->hw;
1222
1223 if (hw->mac.ops.set_vfta(hw, vid, true))
1224 dev_err(&adapter->pdev->dev, "Failed to add vlan id %d\n", vid);
1225}
1226
1227static void igbvf_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
1228{
1229 struct igbvf_adapter *adapter = netdev_priv(netdev);
1230 struct e1000_hw *hw = &adapter->hw;
1231
1232 igbvf_irq_disable(adapter);
1233 vlan_group_set_device(adapter->vlgrp, vid, NULL);
1234
1235 if (!test_bit(__IGBVF_DOWN, &adapter->state))
1236 igbvf_irq_enable(adapter);
1237
1238 if (hw->mac.ops.set_vfta(hw, vid, false))
1239 dev_err(&adapter->pdev->dev,
1240 "Failed to remove vlan id %d\n", vid);
1241}
1242
1243static void igbvf_vlan_rx_register(struct net_device *netdev,
1244 struct vlan_group *grp)
1245{
1246 struct igbvf_adapter *adapter = netdev_priv(netdev);
1247
1248 adapter->vlgrp = grp;
1249}
1250
1251static void igbvf_restore_vlan(struct igbvf_adapter *adapter)
1252{
1253 u16 vid;
1254
1255 if (!adapter->vlgrp)
1256 return;
1257
b738127d 1258 for (vid = 0; vid < VLAN_N_VID; vid++) {
d4e0fe01
AD		 1259 if (!vlan_group_get_device(adapter->vlgrp, vid))
1260 continue;
1261 igbvf_vlan_rx_add_vid(adapter->netdev, vid);
1262 }
1263
1264 igbvf_set_rlpml(adapter);
1265}
1266
1267/**
1268 * igbvf_configure_tx - Configure Transmit Unit after Reset
1269 * @adapter: board private structure
1270 *
1271 * Configure the Tx unit of the MAC after a reset.
1272 **/
1273static void igbvf_configure_tx(struct igbvf_adapter *adapter)
1274{
1275 struct e1000_hw *hw = &adapter->hw;
1276 struct igbvf_ring *tx_ring = adapter->tx_ring;
1277 u64 tdba;
1278 u32 txdctl, dca_txctrl;
1279
1280 /* disable transmits */
1281 txdctl = er32(TXDCTL(0));
1282 ew32(TXDCTL(0), txdctl & ~E1000_TXDCTL_QUEUE_ENABLE);
1283 msleep(10);
1284
1285 /* Setup the HW Tx Head and Tail descriptor pointers */
1286 ew32(TDLEN(0), tx_ring->count * sizeof(union e1000_adv_tx_desc));
1287 tdba = tx_ring->dma;
8e20ce94 1288 ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
d4e0fe01
AD		 1289 ew32(TDBAH(0), (tdba >> 32));
1290 ew32(TDH(0), 0);
1291 ew32(TDT(0), 0);
1292 tx_ring->head = E1000_TDH(0);
1293 tx_ring->tail = E1000_TDT(0);
1294
1295 /* Turn off Relaxed Ordering on head write-backs. The writebacks
1296 * MUST be delivered in order or it will completely screw up
1297 * our bookeeping.
1298 */
1299 dca_txctrl = er32(DCA_TXCTRL(0));
1300 dca_txctrl &= ~E1000_DCA_TXCTRL_TX_WB_RO_EN;
1301 ew32(DCA_TXCTRL(0), dca_txctrl);
1302
1303 /* enable transmits */
1304 txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
1305 ew32(TXDCTL(0), txdctl);
1306
1307 /* Setup Transmit Descriptor Settings for eop descriptor */
1308 adapter->txd_cmd = E1000_ADVTXD_DCMD_EOP | E1000_ADVTXD_DCMD_IFCS;
1309
1310 /* enable Report Status bit */
1311 adapter->txd_cmd |= E1000_ADVTXD_DCMD_RS;
d4e0fe01
AD		 1312}
1313
1314/**
1315 * igbvf_setup_srrctl - configure the receive control registers
1316 * @adapter: Board private structure
1317 **/
1318static void igbvf_setup_srrctl(struct igbvf_adapter *adapter)
1319{
1320 struct e1000_hw *hw = &adapter->hw;
1321 u32 srrctl = 0;
1322
1323 srrctl &= ~(E1000_SRRCTL_DESCTYPE_MASK |
1324 E1000_SRRCTL_BSIZEHDR_MASK |
1325 E1000_SRRCTL_BSIZEPKT_MASK);
1326
1327 /* Enable queue drop to avoid head of line blocking */
1328 srrctl |= E1000_SRRCTL_DROP_EN;
1329
1330 /* Setup buffer sizes */
1331 srrctl |= ALIGN(adapter->rx_buffer_len, 1024) >>
1332 E1000_SRRCTL_BSIZEPKT_SHIFT;
1333
1334 if (adapter->rx_buffer_len < 2048) {
1335 adapter->rx_ps_hdr_size = 0;
1336 srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
1337 } else {
1338 adapter->rx_ps_hdr_size = 128;
1339 srrctl |= adapter->rx_ps_hdr_size <<
1340 E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
1341 srrctl |= E1000_SRRCTL_DESCTYPE_HDR_SPLIT_ALWAYS;
1342 }
1343
1344 ew32(SRRCTL(0), srrctl);
1345}
1346
1347/**
1348 * igbvf_configure_rx - Configure Receive Unit after Reset
1349 * @adapter: board private structure
1350 *
1351 * Configure the Rx unit of the MAC after a reset.
1352 **/
1353static void igbvf_configure_rx(struct igbvf_adapter *adapter)
1354{
1355 struct e1000_hw *hw = &adapter->hw;
1356 struct igbvf_ring *rx_ring = adapter->rx_ring;
1357 u64 rdba;
1358 u32 rdlen, rxdctl;
1359
1360 /* disable receives */
1361 rxdctl = er32(RXDCTL(0));
1362 ew32(RXDCTL(0), rxdctl & ~E1000_RXDCTL_QUEUE_ENABLE);
1363 msleep(10);
1364
1365 rdlen = rx_ring->count * sizeof(union e1000_adv_rx_desc);
1366
1367 /*
1368 * Setup the HW Rx Head and Tail Descriptor Pointers and
1369 * the Base and Length of the Rx Descriptor Ring
1370 */
1371 rdba = rx_ring->dma;
8e20ce94 1372 ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
d4e0fe01
AD		 1373 ew32(RDBAH(0), (rdba >> 32));
1374 ew32(RDLEN(0), rx_ring->count * sizeof(union e1000_adv_rx_desc));
1375 rx_ring->head = E1000_RDH(0);
1376 rx_ring->tail = E1000_RDT(0);
1377 ew32(RDH(0), 0);
1378 ew32(RDT(0), 0);
1379
1380 rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
1381 rxdctl &= 0xFFF00000;
1382 rxdctl |= IGBVF_RX_PTHRESH;
1383 rxdctl |= IGBVF_RX_HTHRESH << 8;
1384 rxdctl |= IGBVF_RX_WTHRESH << 16;
1385
1386 igbvf_set_rlpml(adapter);
1387
1388 /* enable receives */
1389 ew32(RXDCTL(0), rxdctl);
1390}
1391
1392/**
1393 * igbvf_set_multi - Multicast and Promiscuous mode set
1394 * @netdev: network interface device structure
1395 *
1396 * The set_multi entry point is called whenever the multicast address
1397 * list or the network interface flags are updated. This routine is
1398 * responsible for configuring the hardware for proper multicast,
1399 * promiscuous mode, and all-multi behavior.
1400 **/
1401static void igbvf_set_multi(struct net_device *netdev)
1402{
1403 struct igbvf_adapter *adapter = netdev_priv(netdev);
1404 struct e1000_hw *hw = &adapter->hw;
22bedad3 1405 struct netdev_hw_addr *ha;
d4e0fe01
AD		 1406 u8 *mta_list = NULL;
1407 int i;
1408
4cd24eaf
JP		 1409 if (!netdev_mc_empty(netdev)) {
1410 mta_list = kmalloc(netdev_mc_count(netdev) * 6, GFP_ATOMIC);
d4e0fe01
AD		 1411 if (!mta_list) {
1412 dev_err(&adapter->pdev->dev,
1413 "failed to allocate multicast filter list\n");
1414 return;
1415 }
1416 }
1417
1418 /* prepare a packed array of only addresses. */
48e2f183 1419 i = 0;
22bedad3
JP		 1420 netdev_for_each_mc_addr(ha, netdev)
1421 memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
d4e0fe01
AD		 1422
1423 hw->mac.ops.update_mc_addr_list(hw, mta_list, i, 0, 0);
1424 kfree(mta_list);
1425}
1426
1427/**
1428 * igbvf_configure - configure the hardware for Rx and Tx
1429 * @adapter: private board structure
1430 **/
1431static void igbvf_configure(struct igbvf_adapter *adapter)
1432{
1433 igbvf_set_multi(adapter->netdev);
1434
1435 igbvf_restore_vlan(adapter);
1436
1437 igbvf_configure_tx(adapter);
1438 igbvf_setup_srrctl(adapter);
1439 igbvf_configure_rx(adapter);
1440 igbvf_alloc_rx_buffers(adapter->rx_ring,
1441 igbvf_desc_unused(adapter->rx_ring));
1442}
1443
1444/* igbvf_reset - bring the hardware into a known good state
1445 *
1446 * This function boots the hardware and enables some settings that
1447 * require a configuration cycle of the hardware - those cannot be
1448 * set/changed during runtime. After reset the device needs to be
1449 * properly configured for Rx, Tx etc.
1450 */
2d165771 1451static void igbvf_reset(struct igbvf_adapter *adapter)
d4e0fe01
AD		 1452{
1453 struct e1000_mac_info *mac = &adapter->hw.mac;
1454 struct net_device *netdev = adapter->netdev;
1455 struct e1000_hw *hw = &adapter->hw;
1456
1457 /* Allow time for pending master requests to run */
1458 if (mac->ops.reset_hw(hw))
1459 dev_err(&adapter->pdev->dev, "PF still resetting\n");
1460
1461 mac->ops.init_hw(hw);
1462
1463 if (is_valid_ether_addr(adapter->hw.mac.addr)) {
1464 memcpy(netdev->dev_addr, adapter->hw.mac.addr,
1465 netdev->addr_len);
1466 memcpy(netdev->perm_addr, adapter->hw.mac.addr,
1467 netdev->addr_len);
1468 }
72279093
AD		 1469
1470 adapter->last_reset = jiffies;
d4e0fe01
AD		 1471}
1472
1473int igbvf_up(struct igbvf_adapter *adapter)
1474{
1475 struct e1000_hw *hw = &adapter->hw;
1476
1477 /* hardware has been reset, we need to reload some things */
1478 igbvf_configure(adapter);
1479
1480 clear_bit(__IGBVF_DOWN, &adapter->state);
1481
1482 napi_enable(&adapter->rx_ring->napi);
1483 if (adapter->msix_entries)
1484 igbvf_configure_msix(adapter);
1485
1486 /* Clear any pending interrupts. */
1487 er32(EICR);
1488 igbvf_irq_enable(adapter);
1489
1490 /* start the watchdog */
1491 hw->mac.get_link_status = 1;
1492 mod_timer(&adapter->watchdog_timer, jiffies + 1);
1493
1494
1495 return 0;
1496}
1497
1498void igbvf_down(struct igbvf_adapter *adapter)
1499{
1500 struct net_device *netdev = adapter->netdev;
1501 struct e1000_hw *hw = &adapter->hw;
1502 u32 rxdctl, txdctl;
1503
1504 /*
1505 * signal that we're down so the interrupt handler does not
1506 * reschedule our watchdog timer
1507 */
1508 set_bit(__IGBVF_DOWN, &adapter->state);
1509
1510 /* disable receives in the hardware */
1511 rxdctl = er32(RXDCTL(0));
1512 ew32(RXDCTL(0), rxdctl & ~E1000_RXDCTL_QUEUE_ENABLE);
1513
1514 netif_stop_queue(netdev);
1515
1516 /* disable transmits in the hardware */
1517 txdctl = er32(TXDCTL(0));
1518 ew32(TXDCTL(0), txdctl & ~E1000_TXDCTL_QUEUE_ENABLE);
1519
1520 /* flush both disables and wait for them to finish */
1521 e1e_flush();
1522 msleep(10);
1523
1524 napi_disable(&adapter->rx_ring->napi);
1525
1526 igbvf_irq_disable(adapter);
1527
1528 del_timer_sync(&adapter->watchdog_timer);
1529
d4e0fe01
AD		 1530 netif_carrier_off(netdev);
1531
1532 /* record the stats before reset*/
1533 igbvf_update_stats(adapter);
1534
1535 adapter->link_speed = 0;
1536 adapter->link_duplex = 0;
1537
1538 igbvf_reset(adapter);
1539 igbvf_clean_tx_ring(adapter->tx_ring);
1540 igbvf_clean_rx_ring(adapter->rx_ring);
1541}
1542
1543void igbvf_reinit_locked(struct igbvf_adapter *adapter)
1544{
1545 might_sleep();
1546 while (test_and_set_bit(__IGBVF_RESETTING, &adapter->state))
1547 msleep(1);
1548 igbvf_down(adapter);
1549 igbvf_up(adapter);
1550 clear_bit(__IGBVF_RESETTING, &adapter->state);
1551}
1552
1553/**
1554 * igbvf_sw_init - Initialize general software structures (struct igbvf_adapter)
1555 * @adapter: board private structure to initialize
1556 *
1557 * igbvf_sw_init initializes the Adapter private data structure.
1558 * Fields are initialized based on PCI device information and
1559 * OS network device settings (MTU size).
1560 **/
1561static int __devinit igbvf_sw_init(struct igbvf_adapter *adapter)
1562{
1563 struct net_device *netdev = adapter->netdev;
1564 s32 rc;
1565
1566 adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN;
1567 adapter->rx_ps_hdr_size = 0;
1568 adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
1569 adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
1570
1571 adapter->tx_int_delay = 8;
1572 adapter->tx_abs_int_delay = 32;
1573 adapter->rx_int_delay = 0;
1574 adapter->rx_abs_int_delay = 8;
1575 adapter->itr_setting = 3;
1576 adapter->itr = 20000;
1577
1578 /* Set various function pointers */
1579 adapter->ei->init_ops(&adapter->hw);
1580
1581 rc = adapter->hw.mac.ops.init_params(&adapter->hw);
1582 if (rc)
1583 return rc;
1584
1585 rc = adapter->hw.mbx.ops.init_params(&adapter->hw);
1586 if (rc)
1587 return rc;
1588
1589 igbvf_set_interrupt_capability(adapter);
1590
1591 if (igbvf_alloc_queues(adapter))
1592 return -ENOMEM;
1593
1594 spin_lock_init(&adapter->tx_queue_lock);
1595
1596 /* Explicitly disable IRQ since the NIC can be in any state. */
1597 igbvf_irq_disable(adapter);
1598
1599 spin_lock_init(&adapter->stats_lock);
1600
1601 set_bit(__IGBVF_DOWN, &adapter->state);
1602 return 0;
1603}
1604
1605static void igbvf_initialize_last_counter_stats(struct igbvf_adapter *adapter)
1606{
1607 struct e1000_hw *hw = &adapter->hw;
1608
1609 adapter->stats.last_gprc = er32(VFGPRC);
1610 adapter->stats.last_gorc = er32(VFGORC);
1611 adapter->stats.last_gptc = er32(VFGPTC);
1612 adapter->stats.last_gotc = er32(VFGOTC);
1613 adapter->stats.last_mprc = er32(VFMPRC);
1614 adapter->stats.last_gotlbc = er32(VFGOTLBC);
1615 adapter->stats.last_gptlbc = er32(VFGPTLBC);
1616 adapter->stats.last_gorlbc = er32(VFGORLBC);
1617 adapter->stats.last_gprlbc = er32(VFGPRLBC);
1618
1619 adapter->stats.base_gprc = er32(VFGPRC);
1620 adapter->stats.base_gorc = er32(VFGORC);
1621 adapter->stats.base_gptc = er32(VFGPTC);
1622 adapter->stats.base_gotc = er32(VFGOTC);
1623 adapter->stats.base_mprc = er32(VFMPRC);
1624 adapter->stats.base_gotlbc = er32(VFGOTLBC);
1625 adapter->stats.base_gptlbc = er32(VFGPTLBC);
1626 adapter->stats.base_gorlbc = er32(VFGORLBC);
1627 adapter->stats.base_gprlbc = er32(VFGPRLBC);
1628}
1629
1630/**
1631 * igbvf_open - Called when a network interface is made active
1632 * @netdev: network interface device structure
1633 *
1634 * Returns 0 on success, negative value on failure
1635 *
1636 * The open entry point is called when a network interface is made
1637 * active by the system (IFF_UP). At this point all resources needed
1638 * for transmit and receive operations are allocated, the interrupt
1639 * handler is registered with the OS, the watchdog timer is started,
1640 * and the stack is notified that the interface is ready.
1641 **/
1642static int igbvf_open(struct net_device *netdev)
1643{
1644 struct igbvf_adapter *adapter = netdev_priv(netdev);
1645 struct e1000_hw *hw = &adapter->hw;
1646 int err;
1647
1648 /* disallow open during test */
1649 if (test_bit(__IGBVF_TESTING, &adapter->state))
1650 return -EBUSY;
1651
1652 /* allocate transmit descriptors */
1653 err = igbvf_setup_tx_resources(adapter, adapter->tx_ring);
1654 if (err)
1655 goto err_setup_tx;
1656
1657 /* allocate receive descriptors */
1658 err = igbvf_setup_rx_resources(adapter, adapter->rx_ring);
1659 if (err)
1660 goto err_setup_rx;
1661
1662 /*
1663 * before we allocate an interrupt, we must be ready to handle it.
1664 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
1665 * as soon as we call pci_request_irq, so we have to setup our
1666 * clean_rx handler before we do so.
1667 */
1668 igbvf_configure(adapter);
1669
1670 err = igbvf_request_irq(adapter);
1671 if (err)
1672 goto err_req_irq;
1673
1674 /* From here on the code is the same as igbvf_up() */
1675 clear_bit(__IGBVF_DOWN, &adapter->state);
1676
1677 napi_enable(&adapter->rx_ring->napi);
1678
1679 /* clear any pending interrupts */
1680 er32(EICR);
1681
1682 igbvf_irq_enable(adapter);
1683
1684 /* start the watchdog */
1685 hw->mac.get_link_status = 1;
1686 mod_timer(&adapter->watchdog_timer, jiffies + 1);
1687
1688 return 0;
1689
1690err_req_irq:
1691 igbvf_free_rx_resources(adapter->rx_ring);
1692err_setup_rx:
1693 igbvf_free_tx_resources(adapter->tx_ring);
1694err_setup_tx:
1695 igbvf_reset(adapter);
1696
1697 return err;
1698}
1699
1700/**
1701 * igbvf_close - Disables a network interface
1702 * @netdev: network interface device structure
1703 *
1704 * Returns 0, this is not allowed to fail
1705 *
1706 * The close entry point is called when an interface is de-activated
1707 * by the OS. The hardware is still under the drivers control, but
1708 * needs to be disabled. A global MAC reset is issued to stop the
1709 * hardware, and all transmit and receive resources are freed.
1710 **/
1711static int igbvf_close(struct net_device *netdev)
1712{
1713 struct igbvf_adapter *adapter = netdev_priv(netdev);
1714
1715 WARN_ON(test_bit(__IGBVF_RESETTING, &adapter->state));
1716 igbvf_down(adapter);
1717
1718 igbvf_free_irq(adapter);
1719
1720 igbvf_free_tx_resources(adapter->tx_ring);
1721 igbvf_free_rx_resources(adapter->rx_ring);
1722
1723 return 0;
1724}
1725/**
1726 * igbvf_set_mac - Change the Ethernet Address of the NIC
1727 * @netdev: network interface device structure
1728 * @p: pointer to an address structure
1729 *
1730 * Returns 0 on success, negative on failure
1731 **/
1732static int igbvf_set_mac(struct net_device *netdev, void *p)
1733{
1734 struct igbvf_adapter *adapter = netdev_priv(netdev);
1735 struct e1000_hw *hw = &adapter->hw;
1736 struct sockaddr *addr = p;
1737
1738 if (!is_valid_ether_addr(addr->sa_data))
1739 return -EADDRNOTAVAIL;
1740
1741 memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
1742
1743 hw->mac.ops.rar_set(hw, hw->mac.addr, 0);
1744
1745 if (memcmp(addr->sa_data, hw->mac.addr, 6))
1746 return -EADDRNOTAVAIL;
1747
1748 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
1749
1750 return 0;
1751}
1752
1753#define UPDATE_VF_COUNTER(reg, name) \
1754 { \
1755 u32 current_counter = er32(reg); \
1756 if (current_counter < adapter->stats.last_##name) \
1757 adapter->stats.name += 0x100000000LL; \
1758 adapter->stats.last_##name = current_counter; \
1759 adapter->stats.name &= 0xFFFFFFFF00000000LL; \
1760 adapter->stats.name |= current_counter; \
1761 }
1762
1763/**
1764 * igbvf_update_stats - Update the board statistics counters
1765 * @adapter: board private structure
1766**/
1767void igbvf_update_stats(struct igbvf_adapter *adapter)
1768{
1769 struct e1000_hw *hw = &adapter->hw;
1770 struct pci_dev *pdev = adapter->pdev;
1771
1772 /*
1773 * Prevent stats update while adapter is being reset, link is down
1774 * or if the pci connection is down.
1775 */
1776 if (adapter->link_speed == 0)
1777 return;
1778
1779 if (test_bit(__IGBVF_RESETTING, &adapter->state))
1780 return;
1781
1782 if (pci_channel_offline(pdev))
1783 return;
1784
1785 UPDATE_VF_COUNTER(VFGPRC, gprc);
1786 UPDATE_VF_COUNTER(VFGORC, gorc);
1787 UPDATE_VF_COUNTER(VFGPTC, gptc);
1788 UPDATE_VF_COUNTER(VFGOTC, gotc);
1789 UPDATE_VF_COUNTER(VFMPRC, mprc);
1790 UPDATE_VF_COUNTER(VFGOTLBC, gotlbc);
1791 UPDATE_VF_COUNTER(VFGPTLBC, gptlbc);
1792 UPDATE_VF_COUNTER(VFGORLBC, gorlbc);
1793 UPDATE_VF_COUNTER(VFGPRLBC, gprlbc);
1794
1795 /* Fill out the OS statistics structure */
1796 adapter->net_stats.multicast = adapter->stats.mprc;
1797}
1798
1799static void igbvf_print_link_info(struct igbvf_adapter *adapter)
1800{
1801 dev_info(&adapter->pdev->dev, "Link is Up %d Mbps %s\n",
1802 adapter->link_speed,
1803 ((adapter->link_duplex == FULL_DUPLEX) ?
1804 "Full Duplex" : "Half Duplex"));
1805}
1806
1807static bool igbvf_has_link(struct igbvf_adapter *adapter)
1808{
1809 struct e1000_hw *hw = &adapter->hw;
1810 s32 ret_val = E1000_SUCCESS;
1811 bool link_active;
1812
72279093
AD		 1813 /* If interface is down, stay link down */
1814 if (test_bit(__IGBVF_DOWN, &adapter->state))
1815 return false;
1816
d4e0fe01
AD		 1817 ret_val = hw->mac.ops.check_for_link(hw);
1818 link_active = !hw->mac.get_link_status;
1819
1820 /* if check for link returns error we will need to reset */
72279093 1821 if (ret_val && time_after(jiffies, adapter->last_reset + (10 * HZ)))
d4e0fe01
AD		 1822 schedule_work(&adapter->reset_task);
1823
1824 return link_active;
1825}
1826
1827/**
1828 * igbvf_watchdog - Timer Call-back
1829 * @data: pointer to adapter cast into an unsigned long
1830 **/
1831static void igbvf_watchdog(unsigned long data)
1832{
1833 struct igbvf_adapter *adapter = (struct igbvf_adapter *) data;
1834
1835 /* Do the rest outside of interrupt context */
1836 schedule_work(&adapter->watchdog_task);
1837}
1838
1839static void igbvf_watchdog_task(struct work_struct *work)
1840{
1841 struct igbvf_adapter *adapter = container_of(work,
1842 struct igbvf_adapter,
1843 watchdog_task);
1844 struct net_device *netdev = adapter->netdev;
1845 struct e1000_mac_info *mac = &adapter->hw.mac;
1846 struct igbvf_ring *tx_ring = adapter->tx_ring;
1847 struct e1000_hw *hw = &adapter->hw;
1848 u32 link;
1849 int tx_pending = 0;
1850
1851 link = igbvf_has_link(adapter);
1852
1853 if (link) {
1854 if (!netif_carrier_ok(netdev)) {
d4e0fe01
AD		 1855 mac->ops.get_link_up_info(&adapter->hw,
1856 &adapter->link_speed,
1857 &adapter->link_duplex);
1858 igbvf_print_link_info(adapter);
1859
a08af745 1860 /* adjust timeout factor according to speed/duplex */
d4e0fe01
AD		 1861 adapter->tx_timeout_factor = 1;
1862 switch (adapter->link_speed) {
1863 case SPEED_10:
d4e0fe01
AD		 1864 adapter->tx_timeout_factor = 16;
1865 break;
1866 case SPEED_100:
d4e0fe01
AD		 1867 /* maybe add some timeout factor ? */
1868 break;
1869 }
1870
1871 netif_carrier_on(netdev);
1872 netif_wake_queue(netdev);
1873 }
1874 } else {
1875 if (netif_carrier_ok(netdev)) {
1876 adapter->link_speed = 0;
1877 adapter->link_duplex = 0;
1878 dev_info(&adapter->pdev->dev, "Link is Down\n");
1879 netif_carrier_off(netdev);
1880 netif_stop_queue(netdev);
1881 }
1882 }
1883
1884 if (netif_carrier_ok(netdev)) {
1885 igbvf_update_stats(adapter);
1886 } else {
1887 tx_pending = (igbvf_desc_unused(tx_ring) + 1 <
1888 tx_ring->count);
1889 if (tx_pending) {
1890 /*
1891 * We've lost link, so the controller stops DMA,
1892 * but we've got queued Tx work that's never going
1893 * to get done, so reset controller to flush Tx.
1894 * (Do the reset outside of interrupt context).
1895 */
1896 adapter->tx_timeout_count++;
1897 schedule_work(&adapter->reset_task);
1898 }
1899 }
1900
1901 /* Cause software interrupt to ensure Rx ring is cleaned */
1902 ew32(EICS, adapter->rx_ring->eims_value);
1903
1904 /* Force detection of hung controller every watchdog period */
1905 adapter->detect_tx_hung = 1;
1906
1907 /* Reset the timer */
1908 if (!test_bit(__IGBVF_DOWN, &adapter->state))
1909 mod_timer(&adapter->watchdog_timer,
1910 round_jiffies(jiffies + (2 * HZ)));
1911}
1912
1913#define IGBVF_TX_FLAGS_CSUM 0x00000001
1914#define IGBVF_TX_FLAGS_VLAN 0x00000002
1915#define IGBVF_TX_FLAGS_TSO 0x00000004
1916#define IGBVF_TX_FLAGS_IPV4 0x00000008
1917#define IGBVF_TX_FLAGS_VLAN_MASK 0xffff0000
1918#define IGBVF_TX_FLAGS_VLAN_SHIFT 16
1919
1920static int igbvf_tso(struct igbvf_adapter *adapter,
1921 struct igbvf_ring *tx_ring,
1922 struct sk_buff *skb, u32 tx_flags, u8 *hdr_len)
1923{
1924 struct e1000_adv_tx_context_desc *context_desc;
1925 unsigned int i;
1926 int err;
1927 struct igbvf_buffer *buffer_info;
1928 u32 info = 0, tu_cmd = 0;
1929 u32 mss_l4len_idx, l4len;
1930 *hdr_len = 0;
1931
1932 if (skb_header_cloned(skb)) {
1933 err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
1934 if (err) {
1935 dev_err(&adapter->pdev->dev,
1936 "igbvf_tso returning an error\n");
1937 return err;
1938 }
1939 }
1940
1941 l4len = tcp_hdrlen(skb);
1942 *hdr_len += l4len;
1943
1944 if (skb->protocol == htons(ETH_P_IP)) {
1945 struct iphdr *iph = ip_hdr(skb);
1946 iph->tot_len = 0;
1947 iph->check = 0;
1948 tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
1949 iph->daddr, 0,
1950 IPPROTO_TCP,
1951 0);
8e1e8a47 1952 } else if (skb_is_gso_v6(skb)) {
d4e0fe01
AD		 1953 ipv6_hdr(skb)->payload_len = 0;
1954 tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
1955 &ipv6_hdr(skb)->daddr,
1956 0, IPPROTO_TCP, 0);
1957 }
1958
1959 i = tx_ring->next_to_use;
1960
1961 buffer_info = &tx_ring->buffer_info[i];
1962 context_desc = IGBVF_TX_CTXTDESC_ADV(*tx_ring, i);
1963 /* VLAN MACLEN IPLEN */
1964 if (tx_flags & IGBVF_TX_FLAGS_VLAN)
1965 info |= (tx_flags & IGBVF_TX_FLAGS_VLAN_MASK);
1966 info |= (skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT);
1967 *hdr_len += skb_network_offset(skb);
1968 info |= (skb_transport_header(skb) - skb_network_header(skb));
1969 *hdr_len += (skb_transport_header(skb) - skb_network_header(skb));
1970 context_desc->vlan_macip_lens = cpu_to_le32(info);
1971
1972 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
1973 tu_cmd |= (E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT);
1974
1975 if (skb->protocol == htons(ETH_P_IP))
1976 tu_cmd |= E1000_ADVTXD_TUCMD_IPV4;
1977 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
1978
1979 context_desc->type_tucmd_mlhl = cpu_to_le32(tu_cmd);
1980
1981 /* MSS L4LEN IDX */
1982 mss_l4len_idx = (skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT);
1983 mss_l4len_idx |= (l4len << E1000_ADVTXD_L4LEN_SHIFT);
1984
1985 context_desc->mss_l4len_idx = cpu_to_le32(mss_l4len_idx);
1986 context_desc->seqnum_seed = 0;
1987
1988 buffer_info->time_stamp = jiffies;
1989 buffer_info->next_to_watch = i;
1990 buffer_info->dma = 0;
1991 i++;
1992 if (i == tx_ring->count)
1993 i = 0;
1994
1995 tx_ring->next_to_use = i;
1996
1997 return true;
1998}
1999
2000static inline bool igbvf_tx_csum(struct igbvf_adapter *adapter,
2001 struct igbvf_ring *tx_ring,
2002 struct sk_buff *skb, u32 tx_flags)
2003{
2004 struct e1000_adv_tx_context_desc *context_desc;
2005 unsigned int i;
2006 struct igbvf_buffer *buffer_info;
2007 u32 info = 0, tu_cmd = 0;
2008
2009 if ((skb->ip_summed == CHECKSUM_PARTIAL) ||
2010 (tx_flags & IGBVF_TX_FLAGS_VLAN)) {
2011 i = tx_ring->next_to_use;
2012 buffer_info = &tx_ring->buffer_info[i];
2013 context_desc = IGBVF_TX_CTXTDESC_ADV(*tx_ring, i);
2014
2015 if (tx_flags & IGBVF_TX_FLAGS_VLAN)
2016 info |= (tx_flags & IGBVF_TX_FLAGS_VLAN_MASK);
2017
2018 info |= (skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT);
2019 if (skb->ip_summed == CHECKSUM_PARTIAL)
2020 info |= (skb_transport_header(skb) -
2021 skb_network_header(skb));
2022
2023
2024 context_desc->vlan_macip_lens = cpu_to_le32(info);
2025
2026 tu_cmd |= (E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT);
2027
2028 if (skb->ip_summed == CHECKSUM_PARTIAL) {
2029 switch (skb->protocol) {
2030 case __constant_htons(ETH_P_IP):
2031 tu_cmd |= E1000_ADVTXD_TUCMD_IPV4;
2032 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
2033 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
2034 break;
2035 case __constant_htons(ETH_P_IPV6):
2036 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
2037 tu_cmd |= E1000_ADVTXD_TUCMD_L4T_TCP;
2038 break;
2039 default:
2040 break;
2041 }
2042 }
2043
2044 context_desc->type_tucmd_mlhl = cpu_to_le32(tu_cmd);
2045 context_desc->seqnum_seed = 0;
2046 context_desc->mss_l4len_idx = 0;
2047
2048 buffer_info->time_stamp = jiffies;
2049 buffer_info->next_to_watch = i;
2050 buffer_info->dma = 0;
2051 i++;
2052 if (i == tx_ring->count)
2053 i = 0;
2054 tx_ring->next_to_use = i;
2055
2056 return true;
2057 }
2058
2059 return false;
2060}
2061
2062static int igbvf_maybe_stop_tx(struct net_device *netdev, int size)
2063{
2064 struct igbvf_adapter *adapter = netdev_priv(netdev);
2065
2066 /* there is enough descriptors then we don't need to worry */
2067 if (igbvf_desc_unused(adapter->tx_ring) >= size)
2068 return 0;
2069
2070 netif_stop_queue(netdev);
2071
2072 smp_mb();
2073
2074 /* We need to check again just in case room has been made available */
2075 if (igbvf_desc_unused(adapter->tx_ring) < size)
2076 return -EBUSY;
2077
2078 netif_wake_queue(netdev);
2079
2080 ++adapter->restart_queue;
2081 return 0;
2082}
2083
2084#define IGBVF_MAX_TXD_PWR 16
2085#define IGBVF_MAX_DATA_PER_TXD (1 << IGBVF_MAX_TXD_PWR)
2086
2087static inline int igbvf_tx_map_adv(struct igbvf_adapter *adapter,
2088 struct igbvf_ring *tx_ring,
2089 struct sk_buff *skb,
2090 unsigned int first)
2091{
2092 struct igbvf_buffer *buffer_info;
a7d5ca40 2093 struct pci_dev *pdev = adapter->pdev;
d4e0fe01
AD		 2094 unsigned int len = skb_headlen(skb);
2095 unsigned int count = 0, i;
2096 unsigned int f;
d4e0fe01
AD		 2097
2098 i = tx_ring->next_to_use;
2099
d4e0fe01
AD		 2100 buffer_info = &tx_ring->buffer_info[i];
2101 BUG_ON(len >= IGBVF_MAX_DATA_PER_TXD);
2102 buffer_info->length = len;
2103 /* set time_stamp *before* dma to help avoid a possible race */
2104 buffer_info->time_stamp = jiffies;
2105 buffer_info->next_to_watch = i;
ac26d7d6 2106 buffer_info->mapped_as_page = false;
123e9f1a
NN		 2107 buffer_info->dma = dma_map_single(&pdev->dev, skb->data, len,
2108 DMA_TO_DEVICE);
2109 if (dma_mapping_error(&pdev->dev, buffer_info->dma))
a7d5ca40
AD		 2110 goto dma_error;
2111
d4e0fe01
AD		 2112
2113 for (f = 0; f < skb_shinfo(skb)->nr_frags; f++) {
2114 struct skb_frag_struct *frag;
2115
8581145f 2116 count++;
d4e0fe01
AD		 2117 i++;
2118 if (i == tx_ring->count)
2119 i = 0;
2120
2121 frag = &skb_shinfo(skb)->frags[f];
2122 len = frag->size;
2123
2124 buffer_info = &tx_ring->buffer_info[i];
2125 BUG_ON(len >= IGBVF_MAX_DATA_PER_TXD);
2126 buffer_info->length = len;
2127 buffer_info->time_stamp = jiffies;
2128 buffer_info->next_to_watch = i;
a7d5ca40 2129 buffer_info->mapped_as_page = true;
123e9f1a 2130 buffer_info->dma = dma_map_page(&pdev->dev,
a7d5ca40
AD		 2131 frag->page,
2132 frag->page_offset,
2133 len,
123e9f1a
NN		 2134 DMA_TO_DEVICE);
2135 if (dma_mapping_error(&pdev->dev, buffer_info->dma))
a7d5ca40 2136 goto dma_error;
d4e0fe01
AD		 2137 }
2138
2139 tx_ring->buffer_info[i].skb = skb;
2140 tx_ring->buffer_info[first].next_to_watch = i;
2141
a7d5ca40
AD		 2142 return ++count;
2143
2144dma_error:
2145 dev_err(&pdev->dev, "TX DMA map failed\n");
2146
2147 /* clear timestamp and dma mappings for failed buffer_info mapping */
2148 buffer_info->dma = 0;
2149 buffer_info->time_stamp = 0;
2150 buffer_info->length = 0;
2151 buffer_info->next_to_watch = 0;
2152 buffer_info->mapped_as_page = false;
c1fa347f
RK		 2153 if (count)
2154 count--;
a7d5ca40
AD		 2155
2156 /* clear timestamp and dma mappings for remaining portion of packet */
c1fa347f
RK		 2157 while (count--) {
2158 if (i==0)
a7d5ca40 2159 i += tx_ring->count;
c1fa347f 2160 i--;
a7d5ca40
AD		 2161 buffer_info = &tx_ring->buffer_info[i];
2162 igbvf_put_txbuf(adapter, buffer_info);
2163 }
2164
2165 return 0;
d4e0fe01
AD		 2166}
2167
2168static inline void igbvf_tx_queue_adv(struct igbvf_adapter *adapter,
2169 struct igbvf_ring *tx_ring,
2170 int tx_flags, int count, u32 paylen,
2171 u8 hdr_len)
2172{
2173 union e1000_adv_tx_desc *tx_desc = NULL;
2174 struct igbvf_buffer *buffer_info;
2175 u32 olinfo_status = 0, cmd_type_len;
2176 unsigned int i;
2177
2178 cmd_type_len = (E1000_ADVTXD_DTYP_DATA | E1000_ADVTXD_DCMD_IFCS |
2179 E1000_ADVTXD_DCMD_DEXT);
2180
2181 if (tx_flags & IGBVF_TX_FLAGS_VLAN)
2182 cmd_type_len |= E1000_ADVTXD_DCMD_VLE;
2183
2184 if (tx_flags & IGBVF_TX_FLAGS_TSO) {
2185 cmd_type_len |= E1000_ADVTXD_DCMD_TSE;
2186
2187 /* insert tcp checksum */
2188 olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
2189
2190 /* insert ip checksum */
2191 if (tx_flags & IGBVF_TX_FLAGS_IPV4)
2192 olinfo_status |= E1000_TXD_POPTS_IXSM << 8;
2193
2194 } else if (tx_flags & IGBVF_TX_FLAGS_CSUM) {
2195 olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
2196 }
2197
2198 olinfo_status |= ((paylen - hdr_len) << E1000_ADVTXD_PAYLEN_SHIFT);
2199
2200 i = tx_ring->next_to_use;
2201 while (count--) {
2202 buffer_info = &tx_ring->buffer_info[i];
2203 tx_desc = IGBVF_TX_DESC_ADV(*tx_ring, i);
2204 tx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
2205 tx_desc->read.cmd_type_len =
2206 cpu_to_le32(cmd_type_len | buffer_info->length);
2207 tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
2208 i++;
2209 if (i == tx_ring->count)
2210 i = 0;
2211 }
2212
2213 tx_desc->read.cmd_type_len |= cpu_to_le32(adapter->txd_cmd);
2214 /* Force memory writes to complete before letting h/w
2215 * know there are new descriptors to fetch. (Only
2216 * applicable for weak-ordered memory model archs,
2217 * such as IA-64). */
2218 wmb();
2219
2220 tx_ring->next_to_use = i;
2221 writel(i, adapter->hw.hw_addr + tx_ring->tail);
2222 /* we need this if more than one processor can write to our tail
2223 * at a time, it syncronizes IO on IA64/Altix systems */
2224 mmiowb();
2225}
2226
3b29a56d
SH		 2227static netdev_tx_t igbvf_xmit_frame_ring_adv(struct sk_buff *skb,
2228 struct net_device *netdev,
2229 struct igbvf_ring *tx_ring)
d4e0fe01
AD		 2230{
2231 struct igbvf_adapter *adapter = netdev_priv(netdev);
2232 unsigned int first, tx_flags = 0;
2233 u8 hdr_len = 0;
2234 int count = 0;
2235 int tso = 0;
2236
2237 if (test_bit(__IGBVF_DOWN, &adapter->state)) {
2238 dev_kfree_skb_any(skb);
2239 return NETDEV_TX_OK;
2240 }
2241
2242 if (skb->len <= 0) {
2243 dev_kfree_skb_any(skb);
2244 return NETDEV_TX_OK;
2245 }
2246
2247 /*
2248 * need: count + 4 desc gap to keep tail from touching
2249 * + 2 desc gap to keep tail from touching head,
2250 * + 1 desc for skb->data,
2251 * + 1 desc for context descriptor,
2252 * head, otherwise try next time
2253 */
2254 if (igbvf_maybe_stop_tx(netdev, skb_shinfo(skb)->nr_frags + 4)) {
2255 /* this is a hard error */
2256 return NETDEV_TX_BUSY;
2257 }
2258
2259 if (adapter->vlgrp && vlan_tx_tag_present(skb)) {
2260 tx_flags |= IGBVF_TX_FLAGS_VLAN;
2261 tx_flags |= (vlan_tx_tag_get(skb) << IGBVF_TX_FLAGS_VLAN_SHIFT);
2262 }
2263
2264 if (skb->protocol == htons(ETH_P_IP))
2265 tx_flags |= IGBVF_TX_FLAGS_IPV4;
2266
2267 first = tx_ring->next_to_use;
2268
2269 tso = skb_is_gso(skb) ?
2270 igbvf_tso(adapter, tx_ring, skb, tx_flags, &hdr_len) : 0;
2271 if (unlikely(tso < 0)) {
2272 dev_kfree_skb_any(skb);
2273 return NETDEV_TX_OK;
2274 }
2275
2276 if (tso)
2277 tx_flags |= IGBVF_TX_FLAGS_TSO;
2278 else if (igbvf_tx_csum(adapter, tx_ring, skb, tx_flags) &&
2279 (skb->ip_summed == CHECKSUM_PARTIAL))
2280 tx_flags |= IGBVF_TX_FLAGS_CSUM;
2281
2282 /*
2283 * count reflects descriptors mapped, if 0 then mapping error
2284 * has occured and we need to rewind the descriptor queue
2285 */
2286 count = igbvf_tx_map_adv(adapter, tx_ring, skb, first);
2287
2288 if (count) {
2289 igbvf_tx_queue_adv(adapter, tx_ring, tx_flags, count,
2290 skb->len, hdr_len);
d4e0fe01
AD		 2291 /* Make sure there is space in the ring for the next send. */
2292 igbvf_maybe_stop_tx(netdev, MAX_SKB_FRAGS + 4);
2293 } else {
2294 dev_kfree_skb_any(skb);
2295 tx_ring->buffer_info[first].time_stamp = 0;
2296 tx_ring->next_to_use = first;
2297 }
2298
2299 return NETDEV_TX_OK;
2300}
2301
3b29a56d
SH		 2302static netdev_tx_t igbvf_xmit_frame(struct sk_buff *skb,
2303 struct net_device *netdev)
d4e0fe01
AD		 2304{
2305 struct igbvf_adapter *adapter = netdev_priv(netdev);
2306 struct igbvf_ring *tx_ring;
d4e0fe01
AD		 2307
2308 if (test_bit(__IGBVF_DOWN, &adapter->state)) {
2309 dev_kfree_skb_any(skb);
2310 return NETDEV_TX_OK;
2311 }
2312
2313 tx_ring = &adapter->tx_ring[0];
2314
3b29a56d 2315 return igbvf_xmit_frame_ring_adv(skb, netdev, tx_ring);
d4e0fe01
AD		 2316}
2317
2318/**
2319 * igbvf_tx_timeout - Respond to a Tx Hang
2320 * @netdev: network interface device structure
2321 **/
2322static void igbvf_tx_timeout(struct net_device *netdev)
2323{
2324 struct igbvf_adapter *adapter = netdev_priv(netdev);
2325
2326 /* Do the reset outside of interrupt context */
2327 adapter->tx_timeout_count++;
2328 schedule_work(&adapter->reset_task);
2329}
2330
2331static void igbvf_reset_task(struct work_struct *work)
2332{
2333 struct igbvf_adapter *adapter;
2334 adapter = container_of(work, struct igbvf_adapter, reset_task);
2335
2336 igbvf_reinit_locked(adapter);
2337}
2338
2339/**
2340 * igbvf_get_stats - Get System Network Statistics
2341 * @netdev: network interface device structure
2342 *
2343 * Returns the address of the device statistics structure.
2344 * The statistics are actually updated from the timer callback.
2345 **/
2346static struct net_device_stats *igbvf_get_stats(struct net_device *netdev)
2347{
2348 struct igbvf_adapter *adapter = netdev_priv(netdev);
2349
2350 /* only return the current stats */
2351 return &adapter->net_stats;
2352}
2353
2354/**
2355 * igbvf_change_mtu - Change the Maximum Transfer Unit
2356 * @netdev: network interface device structure
2357 * @new_mtu: new value for maximum frame size
2358 *
2359 * Returns 0 on success, negative on failure
2360 **/
2361static int igbvf_change_mtu(struct net_device *netdev, int new_mtu)
2362{
2363 struct igbvf_adapter *adapter = netdev_priv(netdev);
2364 int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
2365
2366 if ((new_mtu < 68) || (max_frame > MAX_JUMBO_FRAME_SIZE)) {
2367 dev_err(&adapter->pdev->dev, "Invalid MTU setting\n");
2368 return -EINVAL;
2369 }
2370
d4e0fe01
AD		 2371#define MAX_STD_JUMBO_FRAME_SIZE 9234
2372 if (max_frame > MAX_STD_JUMBO_FRAME_SIZE) {
2373 dev_err(&adapter->pdev->dev, "MTU > 9216 not supported.\n");
2374 return -EINVAL;
2375 }
2376
2377 while (test_and_set_bit(__IGBVF_RESETTING, &adapter->state))
2378 msleep(1);
2379 /* igbvf_down has a dependency on max_frame_size */
2380 adapter->max_frame_size = max_frame;
2381 if (netif_running(netdev))
2382 igbvf_down(adapter);
2383
2384 /*
2385 * NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
2386 * means we reserve 2 more, this pushes us to allocate from the next
2387 * larger slab size.
2388 * i.e. RXBUFFER_2048 --> size-4096 slab
2389 * However with the new *_jumbo_rx* routines, jumbo receives will use
2390 * fragmented skbs
2391 */
2392
2393 if (max_frame <= 1024)
2394 adapter->rx_buffer_len = 1024;
2395 else if (max_frame <= 2048)
2396 adapter->rx_buffer_len = 2048;
2397 else
2398#if (PAGE_SIZE / 2) > 16384
2399 adapter->rx_buffer_len = 16384;
2400#else
2401 adapter->rx_buffer_len = PAGE_SIZE / 2;
2402#endif
2403
2404
2405 /* adjust allocation if LPE protects us, and we aren't using SBP */
2406 if ((max_frame == ETH_FRAME_LEN + ETH_FCS_LEN) ||
2407 (max_frame == ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN))
2408 adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN +
2409 ETH_FCS_LEN;
2410
2411 dev_info(&adapter->pdev->dev, "changing MTU from %d to %d\n",
2412 netdev->mtu, new_mtu);
2413 netdev->mtu = new_mtu;
2414
2415 if (netif_running(netdev))
2416 igbvf_up(adapter);
2417 else
2418 igbvf_reset(adapter);
2419
2420 clear_bit(__IGBVF_RESETTING, &adapter->state);
2421
2422 return 0;
2423}
2424
2425static int igbvf_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2426{
2427 switch (cmd) {
2428 default:
2429 return -EOPNOTSUPP;
2430 }
2431}
2432
2433static int igbvf_suspend(struct pci_dev *pdev, pm_message_t state)
2434{
2435 struct net_device *netdev = pci_get_drvdata(pdev);
2436 struct igbvf_adapter *adapter = netdev_priv(netdev);
2437#ifdef CONFIG_PM
2438 int retval = 0;
2439#endif
2440
2441 netif_device_detach(netdev);
2442
2443 if (netif_running(netdev)) {
2444 WARN_ON(test_bit(__IGBVF_RESETTING, &adapter->state));
2445 igbvf_down(adapter);
2446 igbvf_free_irq(adapter);
2447 }
2448
2449#ifdef CONFIG_PM
2450 retval = pci_save_state(pdev);
2451 if (retval)
2452 return retval;
2453#endif
2454
2455 pci_disable_device(pdev);
2456
2457 return 0;
2458}
2459
2460#ifdef CONFIG_PM
2461static int igbvf_resume(struct pci_dev *pdev)
2462{
2463 struct net_device *netdev = pci_get_drvdata(pdev);
2464 struct igbvf_adapter *adapter = netdev_priv(netdev);
2465 u32 err;
2466
2467 pci_restore_state(pdev);
2468 err = pci_enable_device_mem(pdev);
2469 if (err) {
2470 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
2471 return err;
2472 }
2473
2474 pci_set_master(pdev);
2475
2476 if (netif_running(netdev)) {
2477 err = igbvf_request_irq(adapter);
2478 if (err)
2479 return err;
2480 }
2481
2482 igbvf_reset(adapter);
2483
2484 if (netif_running(netdev))
2485 igbvf_up(adapter);
2486
2487 netif_device_attach(netdev);
2488
2489 return 0;
2490}
2491#endif
2492
2493static void igbvf_shutdown(struct pci_dev *pdev)
2494{
2495 igbvf_suspend(pdev, PMSG_SUSPEND);
2496}
2497
2498#ifdef CONFIG_NET_POLL_CONTROLLER
2499/*
2500 * Polling 'interrupt' - used by things like netconsole to send skbs
2501 * without having to re-enable interrupts. It's not called while
2502 * the interrupt routine is executing.
2503 */
2504static void igbvf_netpoll(struct net_device *netdev)
2505{
2506 struct igbvf_adapter *adapter = netdev_priv(netdev);
2507
2508 disable_irq(adapter->pdev->irq);
2509
2510 igbvf_clean_tx_irq(adapter->tx_ring);
2511
2512 enable_irq(adapter->pdev->irq);
2513}
2514#endif
2515
2516/**
2517 * igbvf_io_error_detected - called when PCI error is detected
2518 * @pdev: Pointer to PCI device
2519 * @state: The current pci connection state
2520 *
2521 * This function is called after a PCI bus error affecting
2522 * this device has been detected.
2523 */
2524static pci_ers_result_t igbvf_io_error_detected(struct pci_dev *pdev,
2525 pci_channel_state_t state)
2526{
2527 struct net_device *netdev = pci_get_drvdata(pdev);
2528 struct igbvf_adapter *adapter = netdev_priv(netdev);
2529
2530 netif_device_detach(netdev);
2531
c06c430d
DN		 2532 if (state == pci_channel_io_perm_failure)
2533 return PCI_ERS_RESULT_DISCONNECT;
2534
d4e0fe01
AD		 2535 if (netif_running(netdev))
2536 igbvf_down(adapter);
2537 pci_disable_device(pdev);
2538
2539 /* Request a slot slot reset. */
2540 return PCI_ERS_RESULT_NEED_RESET;
2541}
2542
2543/**
2544 * igbvf_io_slot_reset - called after the pci bus has been reset.
2545 * @pdev: Pointer to PCI device
2546 *
2547 * Restart the card from scratch, as if from a cold-boot. Implementation
2548 * resembles the first-half of the igbvf_resume routine.
2549 */
2550static pci_ers_result_t igbvf_io_slot_reset(struct pci_dev *pdev)
2551{
2552 struct net_device *netdev = pci_get_drvdata(pdev);
2553 struct igbvf_adapter *adapter = netdev_priv(netdev);
2554
2555 if (pci_enable_device_mem(pdev)) {
2556 dev_err(&pdev->dev,
2557 "Cannot re-enable PCI device after reset.\n");
2558 return PCI_ERS_RESULT_DISCONNECT;
2559 }
2560 pci_set_master(pdev);
2561
2562 igbvf_reset(adapter);
2563
2564 return PCI_ERS_RESULT_RECOVERED;
2565}
2566
2567/**
2568 * igbvf_io_resume - called when traffic can start flowing again.
2569 * @pdev: Pointer to PCI device
2570 *
2571 * This callback is called when the error recovery driver tells us that
2572 * its OK to resume normal operation. Implementation resembles the
2573 * second-half of the igbvf_resume routine.
2574 */
2575static void igbvf_io_resume(struct pci_dev *pdev)
2576{
2577 struct net_device *netdev = pci_get_drvdata(pdev);
2578 struct igbvf_adapter *adapter = netdev_priv(netdev);
2579
2580 if (netif_running(netdev)) {
2581 if (igbvf_up(adapter)) {
2582 dev_err(&pdev->dev,
2583 "can't bring device back up after reset\n");
2584 return;
2585 }
2586 }
2587
2588 netif_device_attach(netdev);
2589}
2590
2591static void igbvf_print_device_info(struct igbvf_adapter *adapter)
2592{
2593 struct e1000_hw *hw = &adapter->hw;
2594 struct net_device *netdev = adapter->netdev;
2595 struct pci_dev *pdev = adapter->pdev;
2596
2597 dev_info(&pdev->dev, "Intel(R) 82576 Virtual Function\n");
753cdc33 2598 dev_info(&pdev->dev, "Address: %pM\n", netdev->dev_addr);
d4e0fe01
AD		 2599 dev_info(&pdev->dev, "MAC: %d\n", hw->mac.type);
2600}
2601
2602static const struct net_device_ops igbvf_netdev_ops = {
2603 .ndo_open = igbvf_open,
2604 .ndo_stop = igbvf_close,
2605 .ndo_start_xmit = igbvf_xmit_frame,
2606 .ndo_get_stats = igbvf_get_stats,
2607 .ndo_set_multicast_list = igbvf_set_multi,
2608 .ndo_set_mac_address = igbvf_set_mac,
2609 .ndo_change_mtu = igbvf_change_mtu,
2610 .ndo_do_ioctl = igbvf_ioctl,
2611 .ndo_tx_timeout = igbvf_tx_timeout,
2612 .ndo_vlan_rx_register = igbvf_vlan_rx_register,
2613 .ndo_vlan_rx_add_vid = igbvf_vlan_rx_add_vid,
2614 .ndo_vlan_rx_kill_vid = igbvf_vlan_rx_kill_vid,
2615#ifdef CONFIG_NET_POLL_CONTROLLER
2616 .ndo_poll_controller = igbvf_netpoll,
2617#endif
2618};
2619
2620/**
2621 * igbvf_probe - Device Initialization Routine
2622 * @pdev: PCI device information struct
2623 * @ent: entry in igbvf_pci_tbl
2624 *
2625 * Returns 0 on success, negative on failure
2626 *
2627 * igbvf_probe initializes an adapter identified by a pci_dev structure.
2628 * The OS initialization, configuring of the adapter private structure,
2629 * and a hardware reset occur.
2630 **/
2631static int __devinit igbvf_probe(struct pci_dev *pdev,
2632 const struct pci_device_id *ent)
2633{
2634 struct net_device *netdev;
2635 struct igbvf_adapter *adapter;
2636 struct e1000_hw *hw;
2637 const struct igbvf_info *ei = igbvf_info_tbl[ent->driver_data];
2638
2639 static int cards_found;
2640 int err, pci_using_dac;
2641
2642 err = pci_enable_device_mem(pdev);
2643 if (err)
2644 return err;
2645
2646 pci_using_dac = 0;
123e9f1a 2647 err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(64));
d4e0fe01 2648 if (!err) {
123e9f1a 2649 err = dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
d4e0fe01
AD		 2650 if (!err)
2651 pci_using_dac = 1;
2652 } else {
123e9f1a 2653 err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
d4e0fe01 2654 if (err) {
123e9f1a
NN		 2655 err = dma_set_coherent_mask(&pdev->dev,
2656 DMA_BIT_MASK(32));
d4e0fe01
AD		 2657 if (err) {
2658 dev_err(&pdev->dev, "No usable DMA "
2659 "configuration, aborting\n");
2660 goto err_dma;
2661 }
2662 }
2663 }
2664
2665 err = pci_request_regions(pdev, igbvf_driver_name);
2666 if (err)
2667 goto err_pci_reg;
2668
2669 pci_set_master(pdev);
2670
2671 err = -ENOMEM;
2672 netdev = alloc_etherdev(sizeof(struct igbvf_adapter));
2673 if (!netdev)
2674 goto err_alloc_etherdev;
2675
2676 SET_NETDEV_DEV(netdev, &pdev->dev);
2677
2678 pci_set_drvdata(pdev, netdev);
2679 adapter = netdev_priv(netdev);
2680 hw = &adapter->hw;
2681 adapter->netdev = netdev;
2682 adapter->pdev = pdev;
2683 adapter->ei = ei;
2684 adapter->pba = ei->pba;
2685 adapter->flags = ei->flags;
2686 adapter->hw.back = adapter;
2687 adapter->hw.mac.type = ei->mac;
2688 adapter->msg_enable = (1 << NETIF_MSG_DRV | NETIF_MSG_PROBE) - 1;
2689
2690 /* PCI config space info */
2691
2692 hw->vendor_id = pdev->vendor;
2693 hw->device_id = pdev->device;
2694 hw->subsystem_vendor_id = pdev->subsystem_vendor;
2695 hw->subsystem_device_id = pdev->subsystem_device;
2696
2697 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
2698
2699 err = -EIO;
2700 adapter->hw.hw_addr = ioremap(pci_resource_start(pdev, 0),
2701 pci_resource_len(pdev, 0));
2702
2703 if (!adapter->hw.hw_addr)
2704 goto err_ioremap;
2705
2706 if (ei->get_variants) {
2707 err = ei->get_variants(adapter);
2708 if (err)
2709 goto err_ioremap;
2710 }
2711
2712 /* setup adapter struct */
2713 err = igbvf_sw_init(adapter);
2714 if (err)
2715 goto err_sw_init;
2716
2717 /* construct the net_device struct */
2718 netdev->netdev_ops = &igbvf_netdev_ops;
2719
2720 igbvf_set_ethtool_ops(netdev);
2721 netdev->watchdog_timeo = 5 * HZ;
2722 strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
2723
2724 adapter->bd_number = cards_found++;
2725
2726 netdev->features = NETIF_F_SG |
2727 NETIF_F_IP_CSUM |
2728 NETIF_F_HW_VLAN_TX |
2729 NETIF_F_HW_VLAN_RX |
2730 NETIF_F_HW_VLAN_FILTER;
2731
2732 netdev->features |= NETIF_F_IPV6_CSUM;
2733 netdev->features |= NETIF_F_TSO;
2734 netdev->features |= NETIF_F_TSO6;
2735
2736 if (pci_using_dac)
2737 netdev->features |= NETIF_F_HIGHDMA;
2738
2739 netdev->vlan_features |= NETIF_F_TSO;
2740 netdev->vlan_features |= NETIF_F_TSO6;
2741 netdev->vlan_features |= NETIF_F_IP_CSUM;
2742 netdev->vlan_features |= NETIF_F_IPV6_CSUM;
2743 netdev->vlan_features |= NETIF_F_SG;
2744
2745 /*reset the controller to put the device in a known good state */
2746 err = hw->mac.ops.reset_hw(hw);
2747 if (err) {
2748 dev_info(&pdev->dev,
1242b6f3
WM		 2749 "PF still in reset state, assigning new address."
2750 " Is the PF interface up?\n");
2c6952df 2751 dev_hw_addr_random(adapter->netdev, hw->mac.addr);
d4e0fe01
AD		 2752 } else {
2753 err = hw->mac.ops.read_mac_addr(hw);
2754 if (err) {
2755 dev_err(&pdev->dev, "Error reading MAC address\n");
2756 goto err_hw_init;
2757 }
2758 }
2759
2760 memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
2761 memcpy(netdev->perm_addr, adapter->hw.mac.addr, netdev->addr_len);
2762
2763 if (!is_valid_ether_addr(netdev->perm_addr)) {
753cdc33
HS		 2764 dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
2765 netdev->dev_addr);
d4e0fe01
AD		 2766 err = -EIO;
2767 goto err_hw_init;
2768 }
2769
2770 setup_timer(&adapter->watchdog_timer, &igbvf_watchdog,
2771 (unsigned long) adapter);
2772
2773 INIT_WORK(&adapter->reset_task, igbvf_reset_task);
2774 INIT_WORK(&adapter->watchdog_task, igbvf_watchdog_task);
2775
2776 /* ring size defaults */
2777 adapter->rx_ring->count = 1024;
2778 adapter->tx_ring->count = 1024;
2779
2780 /* reset the hardware with the new settings */
2781 igbvf_reset(adapter);
2782
d4e0fe01
AD		 2783 strcpy(netdev->name, "eth%d");
2784 err = register_netdev(netdev);
2785 if (err)
2786 goto err_hw_init;
2787
de7fe787
ET		 2788 /* tell the stack to leave us alone until igbvf_open() is called */
2789 netif_carrier_off(netdev);
2790 netif_stop_queue(netdev);
2791
d4e0fe01
AD		 2792 igbvf_print_device_info(adapter);
2793
2794 igbvf_initialize_last_counter_stats(adapter);
2795
2796 return 0;
2797
2798err_hw_init:
2799 kfree(adapter->tx_ring);
2800 kfree(adapter->rx_ring);
2801err_sw_init:
2802 igbvf_reset_interrupt_capability(adapter);
2803 iounmap(adapter->hw.hw_addr);
2804err_ioremap:
2805 free_netdev(netdev);
2806err_alloc_etherdev:
2807 pci_release_regions(pdev);
2808err_pci_reg:
2809err_dma:
2810 pci_disable_device(pdev);
2811 return err;
2812}
2813
2814/**
2815 * igbvf_remove - Device Removal Routine
2816 * @pdev: PCI device information struct
2817 *
2818 * igbvf_remove is called by the PCI subsystem to alert the driver
2819 * that it should release a PCI device. The could be caused by a
2820 * Hot-Plug event, or because the driver is going to be removed from
2821 * memory.
2822 **/
2823static void __devexit igbvf_remove(struct pci_dev *pdev)
2824{
2825 struct net_device *netdev = pci_get_drvdata(pdev);
2826 struct igbvf_adapter *adapter = netdev_priv(netdev);
2827 struct e1000_hw *hw = &adapter->hw;
2828
2829 /*
2830 * flush_scheduled work may reschedule our watchdog task, so
2831 * explicitly disable watchdog tasks from being rescheduled
2832 */
2833 set_bit(__IGBVF_DOWN, &adapter->state);
2834 del_timer_sync(&adapter->watchdog_timer);
2835
2836 flush_scheduled_work();
2837
2838 unregister_netdev(netdev);
2839
2840 igbvf_reset_interrupt_capability(adapter);
2841
2842 /*
2843 * it is important to delete the napi struct prior to freeing the
2844 * rx ring so that you do not end up with null pointer refs
2845 */
2846 netif_napi_del(&adapter->rx_ring->napi);
2847 kfree(adapter->tx_ring);
2848 kfree(adapter->rx_ring);
2849
2850 iounmap(hw->hw_addr);
2851 if (hw->flash_address)
2852 iounmap(hw->flash_address);
2853 pci_release_regions(pdev);
2854
2855 free_netdev(netdev);
2856
2857 pci_disable_device(pdev);
2858}
2859
2860/* PCI Error Recovery (ERS) */
2861static struct pci_error_handlers igbvf_err_handler = {
2862 .error_detected = igbvf_io_error_detected,
2863 .slot_reset = igbvf_io_slot_reset,
2864 .resume = igbvf_io_resume,
2865};
2866
a3aa1884 2867static DEFINE_PCI_DEVICE_TABLE(igbvf_pci_tbl) = {
d4e0fe01
AD		 2868 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_VF), board_vf },
2869 { } /* terminate list */
2870};
2871MODULE_DEVICE_TABLE(pci, igbvf_pci_tbl);
2872
2873/* PCI Device API Driver */
2874static struct pci_driver igbvf_driver = {
2875 .name = igbvf_driver_name,
2876 .id_table = igbvf_pci_tbl,
2877 .probe = igbvf_probe,
2878 .remove = __devexit_p(igbvf_remove),
2879#ifdef CONFIG_PM
2880 /* Power Management Hooks */
2881 .suspend = igbvf_suspend,
2882 .resume = igbvf_resume,
2883#endif
2884 .shutdown = igbvf_shutdown,
2885 .err_handler = &igbvf_err_handler
2886};
2887
2888/**
2889 * igbvf_init_module - Driver Registration Routine
2890 *
2891 * igbvf_init_module is the first routine called when the driver is
2892 * loaded. All it does is register with the PCI subsystem.
2893 **/
2894static int __init igbvf_init_module(void)
2895{
2896 int ret;
2897 printk(KERN_INFO "%s - version %s\n",
2898 igbvf_driver_string, igbvf_driver_version);
2899 printk(KERN_INFO "%s\n", igbvf_copyright);
2900
2901 ret = pci_register_driver(&igbvf_driver);
d4e0fe01
AD		 2902
2903 return ret;
2904}
2905module_init(igbvf_init_module);
2906
2907/**
2908 * igbvf_exit_module - Driver Exit Cleanup Routine
2909 *
2910 * igbvf_exit_module is called just before the driver is removed
2911 * from memory.
2912 **/
2913static void __exit igbvf_exit_module(void)
2914{
2915 pci_unregister_driver(&igbvf_driver);
d4e0fe01
AD		 2916}
2917module_exit(igbvf_exit_module);
2918
2919
2920MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
2921MODULE_DESCRIPTION("Intel(R) 82576 Virtual Function Network Driver");
2922MODULE_LICENSE("GPL");
2923MODULE_VERSION(DRV_VERSION);
2924
2925/* netdev.c */
Clone of davem's net-next-2.6 tree