1 // SPDX-License-Identifier: GPL-2.0
2 /*******************************************************************************
3 
4   Intel(R) 82576 Virtual Function Linux driver
5   Copyright(c) 2009 - 2012 Intel Corporation.
6 
7   This program is free software; you can redistribute it and/or modify it
8   under the terms and conditions of the GNU General Public License,
9   version 2, as published by the Free Software Foundation.
10 
11   This program is distributed in the hope it will be useful, but WITHOUT
12   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
13   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
14   more details.
15 
16   You should have received a copy of the GNU General Public License along with
17   this program; if not, see <http://www.gnu.org/licenses/>.
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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
29 
30 #include <linux/module.h>
31 #include <linux/types.h>
32 #include <linux/init.h>
33 #include <linux/pci.h>
34 #include <linux/vmalloc.h>
35 #include <linux/pagemap.h>
36 #include <linux/delay.h>
37 #include <linux/netdevice.h>
38 #include <linux/tcp.h>
39 #include <linux/ipv6.h>
40 #include <linux/slab.h>
41 #include <net/checksum.h>
42 #include <net/ip6_checksum.h>
43 #include <linux/mii.h>
44 #include <linux/ethtool.h>
45 #include <linux/if_vlan.h>
46 #include <linux/prefetch.h>
47 #include <linux/sctp.h>
48 
49 #include "igbvf.h"
50 
51 #define DRV_VERSION "2.4.0-k"
52 char igbvf_driver_name[] = "igbvf";
53 const char igbvf_driver_version[] = DRV_VERSION;
54 static const char igbvf_driver_string[] =
55 		  "Intel(R) Gigabit Virtual Function Network Driver";
56 static const char igbvf_copyright[] =
57 		  "Copyright (c) 2009 - 2012 Intel Corporation.";
58 
59 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
60 static int debug = -1;
61 module_param(debug, int, 0);
62 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
63 
64 static int igbvf_poll(struct napi_struct *napi, int budget);
65 static void igbvf_reset(struct igbvf_adapter *);
66 static void igbvf_set_interrupt_capability(struct igbvf_adapter *);
67 static void igbvf_reset_interrupt_capability(struct igbvf_adapter *);
68 
69 static struct igbvf_info igbvf_vf_info = {
70 	.mac		= e1000_vfadapt,
71 	.flags		= 0,
72 	.pba		= 10,
73 	.init_ops	= e1000_init_function_pointers_vf,
74 };
75 
76 static struct igbvf_info igbvf_i350_vf_info = {
77 	.mac		= e1000_vfadapt_i350,
78 	.flags		= 0,
79 	.pba		= 10,
80 	.init_ops	= e1000_init_function_pointers_vf,
81 };
82 
83 static const struct igbvf_info *igbvf_info_tbl[] = {
84 	[board_vf]	= &igbvf_vf_info,
85 	[board_i350_vf]	= &igbvf_i350_vf_info,
86 };
87 
88 /**
89  * igbvf_desc_unused - calculate if we have unused descriptors
90  * @rx_ring: address of receive ring structure
91  **/
92 static int igbvf_desc_unused(struct igbvf_ring *ring)
93 {
94 	if (ring->next_to_clean > ring->next_to_use)
95 		return ring->next_to_clean - ring->next_to_use - 1;
96 
97 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
98 }
99 
100 /**
101  * igbvf_receive_skb - helper function to handle Rx indications
102  * @adapter: board private structure
103  * @status: descriptor status field as written by hardware
104  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
105  * @skb: pointer to sk_buff to be indicated to stack
106  **/
107 static void igbvf_receive_skb(struct igbvf_adapter *adapter,
108 			      struct net_device *netdev,
109 			      struct sk_buff *skb,
110 			      u32 status, u16 vlan)
111 {
112 	u16 vid;
113 
114 	if (status & E1000_RXD_STAT_VP) {
115 		if ((adapter->flags & IGBVF_FLAG_RX_LB_VLAN_BSWAP) &&
116 		    (status & E1000_RXDEXT_STATERR_LB))
117 			vid = be16_to_cpu(vlan) & E1000_RXD_SPC_VLAN_MASK;
118 		else
119 			vid = le16_to_cpu(vlan) & E1000_RXD_SPC_VLAN_MASK;
120 		if (test_bit(vid, adapter->active_vlans))
121 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
122 	}
123 
124 	napi_gro_receive(&adapter->rx_ring->napi, skb);
125 }
126 
127 static inline void igbvf_rx_checksum_adv(struct igbvf_adapter *adapter,
128 					 u32 status_err, struct sk_buff *skb)
129 {
130 	skb_checksum_none_assert(skb);
131 
132 	/* Ignore Checksum bit is set or checksum is disabled through ethtool */
133 	if ((status_err & E1000_RXD_STAT_IXSM) ||
134 	    (adapter->flags & IGBVF_FLAG_RX_CSUM_DISABLED))
135 		return;
136 
137 	/* TCP/UDP checksum error bit is set */
138 	if (status_err &
139 	    (E1000_RXDEXT_STATERR_TCPE | E1000_RXDEXT_STATERR_IPE)) {
140 		/* let the stack verify checksum errors */
141 		adapter->hw_csum_err++;
142 		return;
143 	}
144 
145 	/* It must be a TCP or UDP packet with a valid checksum */
146 	if (status_err & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS))
147 		skb->ip_summed = CHECKSUM_UNNECESSARY;
148 
149 	adapter->hw_csum_good++;
150 }
151 
152 /**
153  * igbvf_alloc_rx_buffers - Replace used receive buffers; packet split
154  * @rx_ring: address of ring structure to repopulate
155  * @cleaned_count: number of buffers to repopulate
156  **/
157 static void igbvf_alloc_rx_buffers(struct igbvf_ring *rx_ring,
158 				   int cleaned_count)
159 {
160 	struct igbvf_adapter *adapter = rx_ring->adapter;
161 	struct net_device *netdev = adapter->netdev;
162 	struct pci_dev *pdev = adapter->pdev;
163 	union e1000_adv_rx_desc *rx_desc;
164 	struct igbvf_buffer *buffer_info;
165 	struct sk_buff *skb;
166 	unsigned int i;
167 	int bufsz;
168 
169 	i = rx_ring->next_to_use;
170 	buffer_info = &rx_ring->buffer_info[i];
171 
172 	if (adapter->rx_ps_hdr_size)
173 		bufsz = adapter->rx_ps_hdr_size;
174 	else
175 		bufsz = adapter->rx_buffer_len;
176 
177 	while (cleaned_count--) {
178 		rx_desc = IGBVF_RX_DESC_ADV(*rx_ring, i);
179 
180 		if (adapter->rx_ps_hdr_size && !buffer_info->page_dma) {
181 			if (!buffer_info->page) {
182 				buffer_info->page = alloc_page(GFP_ATOMIC);
183 				if (!buffer_info->page) {
184 					adapter->alloc_rx_buff_failed++;
185 					goto no_buffers;
186 				}
187 				buffer_info->page_offset = 0;
188 			} else {
189 				buffer_info->page_offset ^= PAGE_SIZE / 2;
190 			}
191 			buffer_info->page_dma =
192 				dma_map_page(&pdev->dev, buffer_info->page,
193 					     buffer_info->page_offset,
194 					     PAGE_SIZE / 2,
195 					     DMA_FROM_DEVICE);
196 			if (dma_mapping_error(&pdev->dev,
197 					      buffer_info->page_dma)) {
198 				__free_page(buffer_info->page);
199 				buffer_info->page = NULL;
200 				dev_err(&pdev->dev, "RX DMA map failed\n");
201 				break;
202 			}
203 		}
204 
205 		if (!buffer_info->skb) {
206 			skb = netdev_alloc_skb_ip_align(netdev, bufsz);
207 			if (!skb) {
208 				adapter->alloc_rx_buff_failed++;
209 				goto no_buffers;
210 			}
211 
212 			buffer_info->skb = skb;
213 			buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
214 							  bufsz,
215 							  DMA_FROM_DEVICE);
216 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
217 				dev_kfree_skb(buffer_info->skb);
218 				buffer_info->skb = NULL;
219 				dev_err(&pdev->dev, "RX DMA map failed\n");
220 				goto no_buffers;
221 			}
222 		}
223 		/* Refresh the desc even if buffer_addrs didn't change because
224 		 * each write-back erases this info.
225 		 */
226 		if (adapter->rx_ps_hdr_size) {
227 			rx_desc->read.pkt_addr =
228 			     cpu_to_le64(buffer_info->page_dma);
229 			rx_desc->read.hdr_addr = cpu_to_le64(buffer_info->dma);
230 		} else {
231 			rx_desc->read.pkt_addr = cpu_to_le64(buffer_info->dma);
232 			rx_desc->read.hdr_addr = 0;
233 		}
234 
235 		i++;
236 		if (i == rx_ring->count)
237 			i = 0;
238 		buffer_info = &rx_ring->buffer_info[i];
239 	}
240 
241 no_buffers:
242 	if (rx_ring->next_to_use != i) {
243 		rx_ring->next_to_use = i;
244 		if (i == 0)
245 			i = (rx_ring->count - 1);
246 		else
247 			i--;
248 
249 		/* Force memory writes to complete before letting h/w
250 		 * know there are new descriptors to fetch.  (Only
251 		 * applicable for weak-ordered memory model archs,
252 		 * such as IA-64).
253 		*/
254 		wmb();
255 		writel(i, adapter->hw.hw_addr + rx_ring->tail);
256 	}
257 }
258 
259 /**
260  * igbvf_clean_rx_irq - Send received data up the network stack; legacy
261  * @adapter: board private structure
262  *
263  * the return value indicates whether actual cleaning was done, there
264  * is no guarantee that everything was cleaned
265  **/
266 static bool igbvf_clean_rx_irq(struct igbvf_adapter *adapter,
267 			       int *work_done, int work_to_do)
268 {
269 	struct igbvf_ring *rx_ring = adapter->rx_ring;
270 	struct net_device *netdev = adapter->netdev;
271 	struct pci_dev *pdev = adapter->pdev;
272 	union e1000_adv_rx_desc *rx_desc, *next_rxd;
273 	struct igbvf_buffer *buffer_info, *next_buffer;
274 	struct sk_buff *skb;
275 	bool cleaned = false;
276 	int cleaned_count = 0;
277 	unsigned int total_bytes = 0, total_packets = 0;
278 	unsigned int i;
279 	u32 length, hlen, staterr;
280 
281 	i = rx_ring->next_to_clean;
282 	rx_desc = IGBVF_RX_DESC_ADV(*rx_ring, i);
283 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
284 
285 	while (staterr & E1000_RXD_STAT_DD) {
286 		if (*work_done >= work_to_do)
287 			break;
288 		(*work_done)++;
289 		rmb(); /* read descriptor and rx_buffer_info after status DD */
290 
291 		buffer_info = &rx_ring->buffer_info[i];
292 
293 		/* HW will not DMA in data larger than the given buffer, even
294 		 * if it parses the (NFS, of course) header to be larger.  In
295 		 * that case, it fills the header buffer and spills the rest
296 		 * into the page.
297 		 */
298 		hlen = (le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.hdr_info)
299 		       & E1000_RXDADV_HDRBUFLEN_MASK) >>
300 		       E1000_RXDADV_HDRBUFLEN_SHIFT;
301 		if (hlen > adapter->rx_ps_hdr_size)
302 			hlen = adapter->rx_ps_hdr_size;
303 
304 		length = le16_to_cpu(rx_desc->wb.upper.length);
305 		cleaned = true;
306 		cleaned_count++;
307 
308 		skb = buffer_info->skb;
309 		prefetch(skb->data - NET_IP_ALIGN);
310 		buffer_info->skb = NULL;
311 		if (!adapter->rx_ps_hdr_size) {
312 			dma_unmap_single(&pdev->dev, buffer_info->dma,
313 					 adapter->rx_buffer_len,
314 					 DMA_FROM_DEVICE);
315 			buffer_info->dma = 0;
316 			skb_put(skb, length);
317 			goto send_up;
318 		}
319 
320 		if (!skb_shinfo(skb)->nr_frags) {
321 			dma_unmap_single(&pdev->dev, buffer_info->dma,
322 					 adapter->rx_ps_hdr_size,
323 					 DMA_FROM_DEVICE);
324 			buffer_info->dma = 0;
325 			skb_put(skb, hlen);
326 		}
327 
328 		if (length) {
329 			dma_unmap_page(&pdev->dev, buffer_info->page_dma,
330 				       PAGE_SIZE / 2,
331 				       DMA_FROM_DEVICE);
332 			buffer_info->page_dma = 0;
333 
334 			skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
335 					   buffer_info->page,
336 					   buffer_info->page_offset,
337 					   length);
338 
339 			if ((adapter->rx_buffer_len > (PAGE_SIZE / 2)) ||
340 			    (page_count(buffer_info->page) != 1))
341 				buffer_info->page = NULL;
342 			else
343 				get_page(buffer_info->page);
344 
345 			skb->len += length;
346 			skb->data_len += length;
347 			skb->truesize += PAGE_SIZE / 2;
348 		}
349 send_up:
350 		i++;
351 		if (i == rx_ring->count)
352 			i = 0;
353 		next_rxd = IGBVF_RX_DESC_ADV(*rx_ring, i);
354 		prefetch(next_rxd);
355 		next_buffer = &rx_ring->buffer_info[i];
356 
357 		if (!(staterr & E1000_RXD_STAT_EOP)) {
358 			buffer_info->skb = next_buffer->skb;
359 			buffer_info->dma = next_buffer->dma;
360 			next_buffer->skb = skb;
361 			next_buffer->dma = 0;
362 			goto next_desc;
363 		}
364 
365 		if (staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) {
366 			dev_kfree_skb_irq(skb);
367 			goto next_desc;
368 		}
369 
370 		total_bytes += skb->len;
371 		total_packets++;
372 
373 		igbvf_rx_checksum_adv(adapter, staterr, skb);
374 
375 		skb->protocol = eth_type_trans(skb, netdev);
376 
377 		igbvf_receive_skb(adapter, netdev, skb, staterr,
378 				  rx_desc->wb.upper.vlan);
379 
380 next_desc:
381 		rx_desc->wb.upper.status_error = 0;
382 
383 		/* return some buffers to hardware, one at a time is too slow */
384 		if (cleaned_count >= IGBVF_RX_BUFFER_WRITE) {
385 			igbvf_alloc_rx_buffers(rx_ring, cleaned_count);
386 			cleaned_count = 0;
387 		}
388 
389 		/* use prefetched values */
390 		rx_desc = next_rxd;
391 		buffer_info = next_buffer;
392 
393 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
394 	}
395 
396 	rx_ring->next_to_clean = i;
397 	cleaned_count = igbvf_desc_unused(rx_ring);
398 
399 	if (cleaned_count)
400 		igbvf_alloc_rx_buffers(rx_ring, cleaned_count);
401 
402 	adapter->total_rx_packets += total_packets;
403 	adapter->total_rx_bytes += total_bytes;
404 	netdev->stats.rx_bytes += total_bytes;
405 	netdev->stats.rx_packets += total_packets;
406 	return cleaned;
407 }
408 
409 static void igbvf_put_txbuf(struct igbvf_adapter *adapter,
410 			    struct igbvf_buffer *buffer_info)
411 {
412 	if (buffer_info->dma) {
413 		if (buffer_info->mapped_as_page)
414 			dma_unmap_page(&adapter->pdev->dev,
415 				       buffer_info->dma,
416 				       buffer_info->length,
417 				       DMA_TO_DEVICE);
418 		else
419 			dma_unmap_single(&adapter->pdev->dev,
420 					 buffer_info->dma,
421 					 buffer_info->length,
422 					 DMA_TO_DEVICE);
423 		buffer_info->dma = 0;
424 	}
425 	if (buffer_info->skb) {
426 		dev_kfree_skb_any(buffer_info->skb);
427 		buffer_info->skb = NULL;
428 	}
429 	buffer_info->time_stamp = 0;
430 }
431 
432 /**
433  * igbvf_setup_tx_resources - allocate Tx resources (Descriptors)
434  * @adapter: board private structure
435  *
436  * Return 0 on success, negative on failure
437  **/
438 int igbvf_setup_tx_resources(struct igbvf_adapter *adapter,
439 			     struct igbvf_ring *tx_ring)
440 {
441 	struct pci_dev *pdev = adapter->pdev;
442 	int size;
443 
444 	size = sizeof(struct igbvf_buffer) * tx_ring->count;
445 	tx_ring->buffer_info = vzalloc(size);
446 	if (!tx_ring->buffer_info)
447 		goto err;
448 
449 	/* round up to nearest 4K */
450 	tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
451 	tx_ring->size = ALIGN(tx_ring->size, 4096);
452 
453 	tx_ring->desc = dma_alloc_coherent(&pdev->dev, tx_ring->size,
454 					   &tx_ring->dma, GFP_KERNEL);
455 	if (!tx_ring->desc)
456 		goto err;
457 
458 	tx_ring->adapter = adapter;
459 	tx_ring->next_to_use = 0;
460 	tx_ring->next_to_clean = 0;
461 
462 	return 0;
463 err:
464 	vfree(tx_ring->buffer_info);
465 	dev_err(&adapter->pdev->dev,
466 		"Unable to allocate memory for the transmit descriptor ring\n");
467 	return -ENOMEM;
468 }
469 
470 /**
471  * igbvf_setup_rx_resources - allocate Rx resources (Descriptors)
472  * @adapter: board private structure
473  *
474  * Returns 0 on success, negative on failure
475  **/
476 int igbvf_setup_rx_resources(struct igbvf_adapter *adapter,
477 			     struct igbvf_ring *rx_ring)
478 {
479 	struct pci_dev *pdev = adapter->pdev;
480 	int size, desc_len;
481 
482 	size = sizeof(struct igbvf_buffer) * rx_ring->count;
483 	rx_ring->buffer_info = vzalloc(size);
484 	if (!rx_ring->buffer_info)
485 		goto err;
486 
487 	desc_len = sizeof(union e1000_adv_rx_desc);
488 
489 	/* Round up to nearest 4K */
490 	rx_ring->size = rx_ring->count * desc_len;
491 	rx_ring->size = ALIGN(rx_ring->size, 4096);
492 
493 	rx_ring->desc = dma_alloc_coherent(&pdev->dev, rx_ring->size,
494 					   &rx_ring->dma, GFP_KERNEL);
495 	if (!rx_ring->desc)
496 		goto err;
497 
498 	rx_ring->next_to_clean = 0;
499 	rx_ring->next_to_use = 0;
500 
501 	rx_ring->adapter = adapter;
502 
503 	return 0;
504 
505 err:
506 	vfree(rx_ring->buffer_info);
507 	rx_ring->buffer_info = NULL;
508 	dev_err(&adapter->pdev->dev,
509 		"Unable to allocate memory for the receive descriptor ring\n");
510 	return -ENOMEM;
511 }
512 
513 /**
514  * igbvf_clean_tx_ring - Free Tx Buffers
515  * @tx_ring: ring to be cleaned
516  **/
517 static void igbvf_clean_tx_ring(struct igbvf_ring *tx_ring)
518 {
519 	struct igbvf_adapter *adapter = tx_ring->adapter;
520 	struct igbvf_buffer *buffer_info;
521 	unsigned long size;
522 	unsigned int i;
523 
524 	if (!tx_ring->buffer_info)
525 		return;
526 
527 	/* Free all the Tx ring sk_buffs */
528 	for (i = 0; i < tx_ring->count; i++) {
529 		buffer_info = &tx_ring->buffer_info[i];
530 		igbvf_put_txbuf(adapter, buffer_info);
531 	}
532 
533 	size = sizeof(struct igbvf_buffer) * tx_ring->count;
534 	memset(tx_ring->buffer_info, 0, size);
535 
536 	/* Zero out the descriptor ring */
537 	memset(tx_ring->desc, 0, tx_ring->size);
538 
539 	tx_ring->next_to_use = 0;
540 	tx_ring->next_to_clean = 0;
541 
542 	writel(0, adapter->hw.hw_addr + tx_ring->head);
543 	writel(0, adapter->hw.hw_addr + tx_ring->tail);
544 }
545 
546 /**
547  * igbvf_free_tx_resources - Free Tx Resources per Queue
548  * @tx_ring: ring to free resources from
549  *
550  * Free all transmit software resources
551  **/
552 void igbvf_free_tx_resources(struct igbvf_ring *tx_ring)
553 {
554 	struct pci_dev *pdev = tx_ring->adapter->pdev;
555 
556 	igbvf_clean_tx_ring(tx_ring);
557 
558 	vfree(tx_ring->buffer_info);
559 	tx_ring->buffer_info = NULL;
560 
561 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
562 			  tx_ring->dma);
563 
564 	tx_ring->desc = NULL;
565 }
566 
567 /**
568  * igbvf_clean_rx_ring - Free Rx Buffers per Queue
569  * @adapter: board private structure
570  **/
571 static void igbvf_clean_rx_ring(struct igbvf_ring *rx_ring)
572 {
573 	struct igbvf_adapter *adapter = rx_ring->adapter;
574 	struct igbvf_buffer *buffer_info;
575 	struct pci_dev *pdev = adapter->pdev;
576 	unsigned long size;
577 	unsigned int i;
578 
579 	if (!rx_ring->buffer_info)
580 		return;
581 
582 	/* Free all the Rx ring sk_buffs */
583 	for (i = 0; i < rx_ring->count; i++) {
584 		buffer_info = &rx_ring->buffer_info[i];
585 		if (buffer_info->dma) {
586 			if (adapter->rx_ps_hdr_size) {
587 				dma_unmap_single(&pdev->dev, buffer_info->dma,
588 						 adapter->rx_ps_hdr_size,
589 						 DMA_FROM_DEVICE);
590 			} else {
591 				dma_unmap_single(&pdev->dev, buffer_info->dma,
592 						 adapter->rx_buffer_len,
593 						 DMA_FROM_DEVICE);
594 			}
595 			buffer_info->dma = 0;
596 		}
597 
598 		if (buffer_info->skb) {
599 			dev_kfree_skb(buffer_info->skb);
600 			buffer_info->skb = NULL;
601 		}
602 
603 		if (buffer_info->page) {
604 			if (buffer_info->page_dma)
605 				dma_unmap_page(&pdev->dev,
606 					       buffer_info->page_dma,
607 					       PAGE_SIZE / 2,
608 					       DMA_FROM_DEVICE);
609 			put_page(buffer_info->page);
610 			buffer_info->page = NULL;
611 			buffer_info->page_dma = 0;
612 			buffer_info->page_offset = 0;
613 		}
614 	}
615 
616 	size = sizeof(struct igbvf_buffer) * rx_ring->count;
617 	memset(rx_ring->buffer_info, 0, size);
618 
619 	/* Zero out the descriptor ring */
620 	memset(rx_ring->desc, 0, rx_ring->size);
621 
622 	rx_ring->next_to_clean = 0;
623 	rx_ring->next_to_use = 0;
624 
625 	writel(0, adapter->hw.hw_addr + rx_ring->head);
626 	writel(0, adapter->hw.hw_addr + rx_ring->tail);
627 }
628 
629 /**
630  * igbvf_free_rx_resources - Free Rx Resources
631  * @rx_ring: ring to clean the resources from
632  *
633  * Free all receive software resources
634  **/
635 
636 void igbvf_free_rx_resources(struct igbvf_ring *rx_ring)
637 {
638 	struct pci_dev *pdev = rx_ring->adapter->pdev;
639 
640 	igbvf_clean_rx_ring(rx_ring);
641 
642 	vfree(rx_ring->buffer_info);
643 	rx_ring->buffer_info = NULL;
644 
645 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
646 			  rx_ring->dma);
647 	rx_ring->desc = NULL;
648 }
649 
650 /**
651  * igbvf_update_itr - update the dynamic ITR value based on statistics
652  * @adapter: pointer to adapter
653  * @itr_setting: current adapter->itr
654  * @packets: the number of packets during this measurement interval
655  * @bytes: the number of bytes during this measurement interval
656  *
657  * Stores a new ITR value based on packets and byte counts during the last
658  * interrupt.  The advantage of per interrupt computation is faster updates
659  * and more accurate ITR for the current traffic pattern.  Constants in this
660  * function were computed based on theoretical maximum wire speed and thresholds
661  * were set based on testing data as well as attempting to minimize response
662  * time while increasing bulk throughput.
663  **/
664 static enum latency_range igbvf_update_itr(struct igbvf_adapter *adapter,
665 					   enum latency_range itr_setting,
666 					   int packets, int bytes)
667 {
668 	enum latency_range retval = itr_setting;
669 
670 	if (packets == 0)
671 		goto update_itr_done;
672 
673 	switch (itr_setting) {
674 	case lowest_latency:
675 		/* handle TSO and jumbo frames */
676 		if (bytes/packets > 8000)
677 			retval = bulk_latency;
678 		else if ((packets < 5) && (bytes > 512))
679 			retval = low_latency;
680 		break;
681 	case low_latency:  /* 50 usec aka 20000 ints/s */
682 		if (bytes > 10000) {
683 			/* this if handles the TSO accounting */
684 			if (bytes/packets > 8000)
685 				retval = bulk_latency;
686 			else if ((packets < 10) || ((bytes/packets) > 1200))
687 				retval = bulk_latency;
688 			else if ((packets > 35))
689 				retval = lowest_latency;
690 		} else if (bytes/packets > 2000) {
691 			retval = bulk_latency;
692 		} else if (packets <= 2 && bytes < 512) {
693 			retval = lowest_latency;
694 		}
695 		break;
696 	case bulk_latency: /* 250 usec aka 4000 ints/s */
697 		if (bytes > 25000) {
698 			if (packets > 35)
699 				retval = low_latency;
700 		} else if (bytes < 6000) {
701 			retval = low_latency;
702 		}
703 		break;
704 	default:
705 		break;
706 	}
707 
708 update_itr_done:
709 	return retval;
710 }
711 
712 static int igbvf_range_to_itr(enum latency_range current_range)
713 {
714 	int new_itr;
715 
716 	switch (current_range) {
717 	/* counts and packets in update_itr are dependent on these numbers */
718 	case lowest_latency:
719 		new_itr = IGBVF_70K_ITR;
720 		break;
721 	case low_latency:
722 		new_itr = IGBVF_20K_ITR;
723 		break;
724 	case bulk_latency:
725 		new_itr = IGBVF_4K_ITR;
726 		break;
727 	default:
728 		new_itr = IGBVF_START_ITR;
729 		break;
730 	}
731 	return new_itr;
732 }
733 
734 static void igbvf_set_itr(struct igbvf_adapter *adapter)
735 {
736 	u32 new_itr;
737 
738 	adapter->tx_ring->itr_range =
739 			igbvf_update_itr(adapter,
740 					 adapter->tx_ring->itr_val,
741 					 adapter->total_tx_packets,
742 					 adapter->total_tx_bytes);
743 
744 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
745 	if (adapter->requested_itr == 3 &&
746 	    adapter->tx_ring->itr_range == lowest_latency)
747 		adapter->tx_ring->itr_range = low_latency;
748 
749 	new_itr = igbvf_range_to_itr(adapter->tx_ring->itr_range);
750 
751 	if (new_itr != adapter->tx_ring->itr_val) {
752 		u32 current_itr = adapter->tx_ring->itr_val;
753 		/* this attempts to bias the interrupt rate towards Bulk
754 		 * by adding intermediate steps when interrupt rate is
755 		 * increasing
756 		 */
757 		new_itr = new_itr > current_itr ?
758 			  min(current_itr + (new_itr >> 2), new_itr) :
759 			  new_itr;
760 		adapter->tx_ring->itr_val = new_itr;
761 
762 		adapter->tx_ring->set_itr = 1;
763 	}
764 
765 	adapter->rx_ring->itr_range =
766 			igbvf_update_itr(adapter, adapter->rx_ring->itr_val,
767 					 adapter->total_rx_packets,
768 					 adapter->total_rx_bytes);
769 	if (adapter->requested_itr == 3 &&
770 	    adapter->rx_ring->itr_range == lowest_latency)
771 		adapter->rx_ring->itr_range = low_latency;
772 
773 	new_itr = igbvf_range_to_itr(adapter->rx_ring->itr_range);
774 
775 	if (new_itr != adapter->rx_ring->itr_val) {
776 		u32 current_itr = adapter->rx_ring->itr_val;
777 
778 		new_itr = new_itr > current_itr ?
779 			  min(current_itr + (new_itr >> 2), new_itr) :
780 			  new_itr;
781 		adapter->rx_ring->itr_val = new_itr;
782 
783 		adapter->rx_ring->set_itr = 1;
784 	}
785 }
786 
787 /**
788  * igbvf_clean_tx_irq - Reclaim resources after transmit completes
789  * @adapter: board private structure
790  *
791  * returns true if ring is completely cleaned
792  **/
793 static bool igbvf_clean_tx_irq(struct igbvf_ring *tx_ring)
794 {
795 	struct igbvf_adapter *adapter = tx_ring->adapter;
796 	struct net_device *netdev = adapter->netdev;
797 	struct igbvf_buffer *buffer_info;
798 	struct sk_buff *skb;
799 	union e1000_adv_tx_desc *tx_desc, *eop_desc;
800 	unsigned int total_bytes = 0, total_packets = 0;
801 	unsigned int i, count = 0;
802 	bool cleaned = false;
803 
804 	i = tx_ring->next_to_clean;
805 	buffer_info = &tx_ring->buffer_info[i];
806 	eop_desc = buffer_info->next_to_watch;
807 
808 	do {
809 		/* if next_to_watch is not set then there is no work pending */
810 		if (!eop_desc)
811 			break;
812 
813 		/* prevent any other reads prior to eop_desc */
814 		smp_rmb();
815 
816 		/* if DD is not set pending work has not been completed */
817 		if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
818 			break;
819 
820 		/* clear next_to_watch to prevent false hangs */
821 		buffer_info->next_to_watch = NULL;
822 
823 		for (cleaned = false; !cleaned; count++) {
824 			tx_desc = IGBVF_TX_DESC_ADV(*tx_ring, i);
825 			cleaned = (tx_desc == eop_desc);
826 			skb = buffer_info->skb;
827 
828 			if (skb) {
829 				unsigned int segs, bytecount;
830 
831 				/* gso_segs is currently only valid for tcp */
832 				segs = skb_shinfo(skb)->gso_segs ?: 1;
833 				/* multiply data chunks by size of headers */
834 				bytecount = ((segs - 1) * skb_headlen(skb)) +
835 					    skb->len;
836 				total_packets += segs;
837 				total_bytes += bytecount;
838 			}
839 
840 			igbvf_put_txbuf(adapter, buffer_info);
841 			tx_desc->wb.status = 0;
842 
843 			i++;
844 			if (i == tx_ring->count)
845 				i = 0;
846 
847 			buffer_info = &tx_ring->buffer_info[i];
848 		}
849 
850 		eop_desc = buffer_info->next_to_watch;
851 	} while (count < tx_ring->count);
852 
853 	tx_ring->next_to_clean = i;
854 
855 	if (unlikely(count && netif_carrier_ok(netdev) &&
856 	    igbvf_desc_unused(tx_ring) >= IGBVF_TX_QUEUE_WAKE)) {
857 		/* Make sure that anybody stopping the queue after this
858 		 * sees the new next_to_clean.
859 		 */
860 		smp_mb();
861 		if (netif_queue_stopped(netdev) &&
862 		    !(test_bit(__IGBVF_DOWN, &adapter->state))) {
863 			netif_wake_queue(netdev);
864 			++adapter->restart_queue;
865 		}
866 	}
867 
868 	netdev->stats.tx_bytes += total_bytes;
869 	netdev->stats.tx_packets += total_packets;
870 	return count < tx_ring->count;
871 }
872 
873 static irqreturn_t igbvf_msix_other(int irq, void *data)
874 {
875 	struct net_device *netdev = data;
876 	struct igbvf_adapter *adapter = netdev_priv(netdev);
877 	struct e1000_hw *hw = &adapter->hw;
878 
879 	adapter->int_counter1++;
880 
881 	hw->mac.get_link_status = 1;
882 	if (!test_bit(__IGBVF_DOWN, &adapter->state))
883 		mod_timer(&adapter->watchdog_timer, jiffies + 1);
884 
885 	ew32(EIMS, adapter->eims_other);
886 
887 	return IRQ_HANDLED;
888 }
889 
890 static irqreturn_t igbvf_intr_msix_tx(int irq, void *data)
891 {
892 	struct net_device *netdev = data;
893 	struct igbvf_adapter *adapter = netdev_priv(netdev);
894 	struct e1000_hw *hw = &adapter->hw;
895 	struct igbvf_ring *tx_ring = adapter->tx_ring;
896 
897 	if (tx_ring->set_itr) {
898 		writel(tx_ring->itr_val,
899 		       adapter->hw.hw_addr + tx_ring->itr_register);
900 		adapter->tx_ring->set_itr = 0;
901 	}
902 
903 	adapter->total_tx_bytes = 0;
904 	adapter->total_tx_packets = 0;
905 
906 	/* auto mask will automatically re-enable the interrupt when we write
907 	 * EICS
908 	 */
909 	if (!igbvf_clean_tx_irq(tx_ring))
910 		/* Ring was not completely cleaned, so fire another interrupt */
911 		ew32(EICS, tx_ring->eims_value);
912 	else
913 		ew32(EIMS, tx_ring->eims_value);
914 
915 	return IRQ_HANDLED;
916 }
917 
918 static irqreturn_t igbvf_intr_msix_rx(int irq, void *data)
919 {
920 	struct net_device *netdev = data;
921 	struct igbvf_adapter *adapter = netdev_priv(netdev);
922 
923 	adapter->int_counter0++;
924 
925 	/* Write the ITR value calculated at the end of the
926 	 * previous interrupt.
927 	 */
928 	if (adapter->rx_ring->set_itr) {
929 		writel(adapter->rx_ring->itr_val,
930 		       adapter->hw.hw_addr + adapter->rx_ring->itr_register);
931 		adapter->rx_ring->set_itr = 0;
932 	}
933 
934 	if (napi_schedule_prep(&adapter->rx_ring->napi)) {
935 		adapter->total_rx_bytes = 0;
936 		adapter->total_rx_packets = 0;
937 		__napi_schedule(&adapter->rx_ring->napi);
938 	}
939 
940 	return IRQ_HANDLED;
941 }
942 
943 #define IGBVF_NO_QUEUE -1
944 
945 static void igbvf_assign_vector(struct igbvf_adapter *adapter, int rx_queue,
946 				int tx_queue, int msix_vector)
947 {
948 	struct e1000_hw *hw = &adapter->hw;
949 	u32 ivar, index;
950 
951 	/* 82576 uses a table-based method for assigning vectors.
952 	 * Each queue has a single entry in the table to which we write
953 	 * a vector number along with a "valid" bit.  Sadly, the layout
954 	 * of the table is somewhat counterintuitive.
955 	 */
956 	if (rx_queue > IGBVF_NO_QUEUE) {
957 		index = (rx_queue >> 1);
958 		ivar = array_er32(IVAR0, index);
959 		if (rx_queue & 0x1) {
960 			/* vector goes into third byte of register */
961 			ivar = ivar & 0xFF00FFFF;
962 			ivar |= (msix_vector | E1000_IVAR_VALID) << 16;
963 		} else {
964 			/* vector goes into low byte of register */
965 			ivar = ivar & 0xFFFFFF00;
966 			ivar |= msix_vector | E1000_IVAR_VALID;
967 		}
968 		adapter->rx_ring[rx_queue].eims_value = BIT(msix_vector);
969 		array_ew32(IVAR0, index, ivar);
970 	}
971 	if (tx_queue > IGBVF_NO_QUEUE) {
972 		index = (tx_queue >> 1);
973 		ivar = array_er32(IVAR0, index);
974 		if (tx_queue & 0x1) {
975 			/* vector goes into high byte of register */
976 			ivar = ivar & 0x00FFFFFF;
977 			ivar |= (msix_vector | E1000_IVAR_VALID) << 24;
978 		} else {
979 			/* vector goes into second byte of register */
980 			ivar = ivar & 0xFFFF00FF;
981 			ivar |= (msix_vector | E1000_IVAR_VALID) << 8;
982 		}
983 		adapter->tx_ring[tx_queue].eims_value = BIT(msix_vector);
984 		array_ew32(IVAR0, index, ivar);
985 	}
986 }
987 
988 /**
989  * igbvf_configure_msix - Configure MSI-X hardware
990  * @adapter: board private structure
991  *
992  * igbvf_configure_msix sets up the hardware to properly
993  * generate MSI-X interrupts.
994  **/
995 static void igbvf_configure_msix(struct igbvf_adapter *adapter)
996 {
997 	u32 tmp;
998 	struct e1000_hw *hw = &adapter->hw;
999 	struct igbvf_ring *tx_ring = adapter->tx_ring;
1000 	struct igbvf_ring *rx_ring = adapter->rx_ring;
1001 	int vector = 0;
1002 
1003 	adapter->eims_enable_mask = 0;
1004 
1005 	igbvf_assign_vector(adapter, IGBVF_NO_QUEUE, 0, vector++);
1006 	adapter->eims_enable_mask |= tx_ring->eims_value;
1007 	writel(tx_ring->itr_val, hw->hw_addr + tx_ring->itr_register);
1008 	igbvf_assign_vector(adapter, 0, IGBVF_NO_QUEUE, vector++);
1009 	adapter->eims_enable_mask |= rx_ring->eims_value;
1010 	writel(rx_ring->itr_val, hw->hw_addr + rx_ring->itr_register);
1011 
1012 	/* set vector for other causes, i.e. link changes */
1013 
1014 	tmp = (vector++ | E1000_IVAR_VALID);
1015 
1016 	ew32(IVAR_MISC, tmp);
1017 
1018 	adapter->eims_enable_mask = GENMASK(vector - 1, 0);
1019 	adapter->eims_other = BIT(vector - 1);
1020 	e1e_flush();
1021 }
1022 
1023 static void igbvf_reset_interrupt_capability(struct igbvf_adapter *adapter)
1024 {
1025 	if (adapter->msix_entries) {
1026 		pci_disable_msix(adapter->pdev);
1027 		kfree(adapter->msix_entries);
1028 		adapter->msix_entries = NULL;
1029 	}
1030 }
1031 
1032 /**
1033  * igbvf_set_interrupt_capability - set MSI or MSI-X if supported
1034  * @adapter: board private structure
1035  *
1036  * Attempt to configure interrupts using the best available
1037  * capabilities of the hardware and kernel.
1038  **/
1039 static void igbvf_set_interrupt_capability(struct igbvf_adapter *adapter)
1040 {
1041 	int err = -ENOMEM;
1042 	int i;
1043 
1044 	/* we allocate 3 vectors, 1 for Tx, 1 for Rx, one for PF messages */
1045 	adapter->msix_entries = kcalloc(3, sizeof(struct msix_entry),
1046 					GFP_KERNEL);
1047 	if (adapter->msix_entries) {
1048 		for (i = 0; i < 3; i++)
1049 			adapter->msix_entries[i].entry = i;
1050 
1051 		err = pci_enable_msix_range(adapter->pdev,
1052 					    adapter->msix_entries, 3, 3);
1053 	}
1054 
1055 	if (err < 0) {
1056 		/* MSI-X failed */
1057 		dev_err(&adapter->pdev->dev,
1058 			"Failed to initialize MSI-X interrupts.\n");
1059 		igbvf_reset_interrupt_capability(adapter);
1060 	}
1061 }
1062 
1063 /**
1064  * igbvf_request_msix - Initialize MSI-X interrupts
1065  * @adapter: board private structure
1066  *
1067  * igbvf_request_msix allocates MSI-X vectors and requests interrupts from the
1068  * kernel.
1069  **/
1070 static int igbvf_request_msix(struct igbvf_adapter *adapter)
1071 {
1072 	struct net_device *netdev = adapter->netdev;
1073 	int err = 0, vector = 0;
1074 
1075 	if (strlen(netdev->name) < (IFNAMSIZ - 5)) {
1076 		sprintf(adapter->tx_ring->name, "%s-tx-0", netdev->name);
1077 		sprintf(adapter->rx_ring->name, "%s-rx-0", netdev->name);
1078 	} else {
1079 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
1080 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
1081 	}
1082 
1083 	err = request_irq(adapter->msix_entries[vector].vector,
1084 			  igbvf_intr_msix_tx, 0, adapter->tx_ring->name,
1085 			  netdev);
1086 	if (err)
1087 		goto out;
1088 
1089 	adapter->tx_ring->itr_register = E1000_EITR(vector);
1090 	adapter->tx_ring->itr_val = adapter->current_itr;
1091 	vector++;
1092 
1093 	err = request_irq(adapter->msix_entries[vector].vector,
1094 			  igbvf_intr_msix_rx, 0, adapter->rx_ring->name,
1095 			  netdev);
1096 	if (err)
1097 		goto out;
1098 
1099 	adapter->rx_ring->itr_register = E1000_EITR(vector);
1100 	adapter->rx_ring->itr_val = adapter->current_itr;
1101 	vector++;
1102 
1103 	err = request_irq(adapter->msix_entries[vector].vector,
1104 			  igbvf_msix_other, 0, netdev->name, netdev);
1105 	if (err)
1106 		goto out;
1107 
1108 	igbvf_configure_msix(adapter);
1109 	return 0;
1110 out:
1111 	return err;
1112 }
1113 
1114 /**
1115  * igbvf_alloc_queues - Allocate memory for all rings
1116  * @adapter: board private structure to initialize
1117  **/
1118 static int igbvf_alloc_queues(struct igbvf_adapter *adapter)
1119 {
1120 	struct net_device *netdev = adapter->netdev;
1121 
1122 	adapter->tx_ring = kzalloc(sizeof(struct igbvf_ring), GFP_KERNEL);
1123 	if (!adapter->tx_ring)
1124 		return -ENOMEM;
1125 
1126 	adapter->rx_ring = kzalloc(sizeof(struct igbvf_ring), GFP_KERNEL);
1127 	if (!adapter->rx_ring) {
1128 		kfree(adapter->tx_ring);
1129 		return -ENOMEM;
1130 	}
1131 
1132 	netif_napi_add(netdev, &adapter->rx_ring->napi, igbvf_poll, 64);
1133 
1134 	return 0;
1135 }
1136 
1137 /**
1138  * igbvf_request_irq - initialize interrupts
1139  * @adapter: board private structure
1140  *
1141  * Attempts to configure interrupts using the best available
1142  * capabilities of the hardware and kernel.
1143  **/
1144 static int igbvf_request_irq(struct igbvf_adapter *adapter)
1145 {
1146 	int err = -1;
1147 
1148 	/* igbvf supports msi-x only */
1149 	if (adapter->msix_entries)
1150 		err = igbvf_request_msix(adapter);
1151 
1152 	if (!err)
1153 		return err;
1154 
1155 	dev_err(&adapter->pdev->dev,
1156 		"Unable to allocate interrupt, Error: %d\n", err);
1157 
1158 	return err;
1159 }
1160 
1161 static void igbvf_free_irq(struct igbvf_adapter *adapter)
1162 {
1163 	struct net_device *netdev = adapter->netdev;
1164 	int vector;
1165 
1166 	if (adapter->msix_entries) {
1167 		for (vector = 0; vector < 3; vector++)
1168 			free_irq(adapter->msix_entries[vector].vector, netdev);
1169 	}
1170 }
1171 
1172 /**
1173  * igbvf_irq_disable - Mask off interrupt generation on the NIC
1174  * @adapter: board private structure
1175  **/
1176 static void igbvf_irq_disable(struct igbvf_adapter *adapter)
1177 {
1178 	struct e1000_hw *hw = &adapter->hw;
1179 
1180 	ew32(EIMC, ~0);
1181 
1182 	if (adapter->msix_entries)
1183 		ew32(EIAC, 0);
1184 }
1185 
1186 /**
1187  * igbvf_irq_enable - Enable default interrupt generation settings
1188  * @adapter: board private structure
1189  **/
1190 static void igbvf_irq_enable(struct igbvf_adapter *adapter)
1191 {
1192 	struct e1000_hw *hw = &adapter->hw;
1193 
1194 	ew32(EIAC, adapter->eims_enable_mask);
1195 	ew32(EIAM, adapter->eims_enable_mask);
1196 	ew32(EIMS, adapter->eims_enable_mask);
1197 }
1198 
1199 /**
1200  * igbvf_poll - NAPI Rx polling callback
1201  * @napi: struct associated with this polling callback
1202  * @budget: amount of packets driver is allowed to process this poll
1203  **/
1204 static int igbvf_poll(struct napi_struct *napi, int budget)
1205 {
1206 	struct igbvf_ring *rx_ring = container_of(napi, struct igbvf_ring, napi);
1207 	struct igbvf_adapter *adapter = rx_ring->adapter;
1208 	struct e1000_hw *hw = &adapter->hw;
1209 	int work_done = 0;
1210 
1211 	igbvf_clean_rx_irq(adapter, &work_done, budget);
1212 
1213 	/* If not enough Rx work done, exit the polling mode */
1214 	if (work_done < budget) {
1215 		napi_complete_done(napi, work_done);
1216 
1217 		if (adapter->requested_itr & 3)
1218 			igbvf_set_itr(adapter);
1219 
1220 		if (!test_bit(__IGBVF_DOWN, &adapter->state))
1221 			ew32(EIMS, adapter->rx_ring->eims_value);
1222 	}
1223 
1224 	return work_done;
1225 }
1226 
1227 /**
1228  * igbvf_set_rlpml - set receive large packet maximum length
1229  * @adapter: board private structure
1230  *
1231  * Configure the maximum size of packets that will be received
1232  */
1233 static void igbvf_set_rlpml(struct igbvf_adapter *adapter)
1234 {
1235 	int max_frame_size;
1236 	struct e1000_hw *hw = &adapter->hw;
1237 
1238 	max_frame_size = adapter->max_frame_size + VLAN_TAG_SIZE;
1239 
1240 	spin_lock_bh(&hw->mbx_lock);
1241 
1242 	e1000_rlpml_set_vf(hw, max_frame_size);
1243 
1244 	spin_unlock_bh(&hw->mbx_lock);
1245 }
1246 
1247 static int igbvf_vlan_rx_add_vid(struct net_device *netdev,
1248 				 __be16 proto, u16 vid)
1249 {
1250 	struct igbvf_adapter *adapter = netdev_priv(netdev);
1251 	struct e1000_hw *hw = &adapter->hw;
1252 
1253 	spin_lock_bh(&hw->mbx_lock);
1254 
1255 	if (hw->mac.ops.set_vfta(hw, vid, true)) {
1256 		dev_err(&adapter->pdev->dev, "Failed to add vlan id %d\n", vid);
1257 		spin_unlock_bh(&hw->mbx_lock);
1258 		return -EINVAL;
1259 	}
1260 
1261 	spin_unlock_bh(&hw->mbx_lock);
1262 
1263 	set_bit(vid, adapter->active_vlans);
1264 	return 0;
1265 }
1266 
1267 static int igbvf_vlan_rx_kill_vid(struct net_device *netdev,
1268 				  __be16 proto, u16 vid)
1269 {
1270 	struct igbvf_adapter *adapter = netdev_priv(netdev);
1271 	struct e1000_hw *hw = &adapter->hw;
1272 
1273 	spin_lock_bh(&hw->mbx_lock);
1274 
1275 	if (hw->mac.ops.set_vfta(hw, vid, false)) {
1276 		dev_err(&adapter->pdev->dev,
1277 			"Failed to remove vlan id %d\n", vid);
1278 		spin_unlock_bh(&hw->mbx_lock);
1279 		return -EINVAL;
1280 	}
1281 
1282 	spin_unlock_bh(&hw->mbx_lock);
1283 
1284 	clear_bit(vid, adapter->active_vlans);
1285 	return 0;
1286 }
1287 
1288 static void igbvf_restore_vlan(struct igbvf_adapter *adapter)
1289 {
1290 	u16 vid;
1291 
1292 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
1293 		igbvf_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
1294 }
1295 
1296 /**
1297  * igbvf_configure_tx - Configure Transmit Unit after Reset
1298  * @adapter: board private structure
1299  *
1300  * Configure the Tx unit of the MAC after a reset.
1301  **/
1302 static void igbvf_configure_tx(struct igbvf_adapter *adapter)
1303 {
1304 	struct e1000_hw *hw = &adapter->hw;
1305 	struct igbvf_ring *tx_ring = adapter->tx_ring;
1306 	u64 tdba;
1307 	u32 txdctl, dca_txctrl;
1308 
1309 	/* disable transmits */
1310 	txdctl = er32(TXDCTL(0));
1311 	ew32(TXDCTL(0), txdctl & ~E1000_TXDCTL_QUEUE_ENABLE);
1312 	e1e_flush();
1313 	msleep(10);
1314 
1315 	/* Setup the HW Tx Head and Tail descriptor pointers */
1316 	ew32(TDLEN(0), tx_ring->count * sizeof(union e1000_adv_tx_desc));
1317 	tdba = tx_ring->dma;
1318 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
1319 	ew32(TDBAH(0), (tdba >> 32));
1320 	ew32(TDH(0), 0);
1321 	ew32(TDT(0), 0);
1322 	tx_ring->head = E1000_TDH(0);
1323 	tx_ring->tail = E1000_TDT(0);
1324 
1325 	/* Turn off Relaxed Ordering on head write-backs.  The writebacks
1326 	 * MUST be delivered in order or it will completely screw up
1327 	 * our bookkeeping.
1328 	 */
1329 	dca_txctrl = er32(DCA_TXCTRL(0));
1330 	dca_txctrl &= ~E1000_DCA_TXCTRL_TX_WB_RO_EN;
1331 	ew32(DCA_TXCTRL(0), dca_txctrl);
1332 
1333 	/* enable transmits */
1334 	txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
1335 	ew32(TXDCTL(0), txdctl);
1336 
1337 	/* Setup Transmit Descriptor Settings for eop descriptor */
1338 	adapter->txd_cmd = E1000_ADVTXD_DCMD_EOP | E1000_ADVTXD_DCMD_IFCS;
1339 
1340 	/* enable Report Status bit */
1341 	adapter->txd_cmd |= E1000_ADVTXD_DCMD_RS;
1342 }
1343 
1344 /**
1345  * igbvf_setup_srrctl - configure the receive control registers
1346  * @adapter: Board private structure
1347  **/
1348 static void igbvf_setup_srrctl(struct igbvf_adapter *adapter)
1349 {
1350 	struct e1000_hw *hw = &adapter->hw;
1351 	u32 srrctl = 0;
1352 
1353 	srrctl &= ~(E1000_SRRCTL_DESCTYPE_MASK |
1354 		    E1000_SRRCTL_BSIZEHDR_MASK |
1355 		    E1000_SRRCTL_BSIZEPKT_MASK);
1356 
1357 	/* Enable queue drop to avoid head of line blocking */
1358 	srrctl |= E1000_SRRCTL_DROP_EN;
1359 
1360 	/* Setup buffer sizes */
1361 	srrctl |= ALIGN(adapter->rx_buffer_len, 1024) >>
1362 		  E1000_SRRCTL_BSIZEPKT_SHIFT;
1363 
1364 	if (adapter->rx_buffer_len < 2048) {
1365 		adapter->rx_ps_hdr_size = 0;
1366 		srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
1367 	} else {
1368 		adapter->rx_ps_hdr_size = 128;
1369 		srrctl |= adapter->rx_ps_hdr_size <<
1370 			  E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
1371 		srrctl |= E1000_SRRCTL_DESCTYPE_HDR_SPLIT_ALWAYS;
1372 	}
1373 
1374 	ew32(SRRCTL(0), srrctl);
1375 }
1376 
1377 /**
1378  * igbvf_configure_rx - Configure Receive Unit after Reset
1379  * @adapter: board private structure
1380  *
1381  * Configure the Rx unit of the MAC after a reset.
1382  **/
1383 static void igbvf_configure_rx(struct igbvf_adapter *adapter)
1384 {
1385 	struct e1000_hw *hw = &adapter->hw;
1386 	struct igbvf_ring *rx_ring = adapter->rx_ring;
1387 	u64 rdba;
1388 	u32 rxdctl;
1389 
1390 	/* disable receives */
1391 	rxdctl = er32(RXDCTL(0));
1392 	ew32(RXDCTL(0), rxdctl & ~E1000_RXDCTL_QUEUE_ENABLE);
1393 	e1e_flush();
1394 	msleep(10);
1395 
1396 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
1397 	 * the Base and Length of the Rx Descriptor Ring
1398 	 */
1399 	rdba = rx_ring->dma;
1400 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
1401 	ew32(RDBAH(0), (rdba >> 32));
1402 	ew32(RDLEN(0), rx_ring->count * sizeof(union e1000_adv_rx_desc));
1403 	rx_ring->head = E1000_RDH(0);
1404 	rx_ring->tail = E1000_RDT(0);
1405 	ew32(RDH(0), 0);
1406 	ew32(RDT(0), 0);
1407 
1408 	rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
1409 	rxdctl &= 0xFFF00000;
1410 	rxdctl |= IGBVF_RX_PTHRESH;
1411 	rxdctl |= IGBVF_RX_HTHRESH << 8;
1412 	rxdctl |= IGBVF_RX_WTHRESH << 16;
1413 
1414 	igbvf_set_rlpml(adapter);
1415 
1416 	/* enable receives */
1417 	ew32(RXDCTL(0), rxdctl);
1418 }
1419 
1420 /**
1421  * igbvf_set_multi - Multicast and Promiscuous mode set
1422  * @netdev: network interface device structure
1423  *
1424  * The set_multi entry point is called whenever the multicast address
1425  * list or the network interface flags are updated.  This routine is
1426  * responsible for configuring the hardware for proper multicast,
1427  * promiscuous mode, and all-multi behavior.
1428  **/
1429 static void igbvf_set_multi(struct net_device *netdev)
1430 {
1431 	struct igbvf_adapter *adapter = netdev_priv(netdev);
1432 	struct e1000_hw *hw = &adapter->hw;
1433 	struct netdev_hw_addr *ha;
1434 	u8  *mta_list = NULL;
1435 	int i;
1436 
1437 	if (!netdev_mc_empty(netdev)) {
1438 		mta_list = kmalloc_array(netdev_mc_count(netdev), ETH_ALEN,
1439 					 GFP_ATOMIC);
1440 		if (!mta_list)
1441 			return;
1442 	}
1443 
1444 	/* prepare a packed array of only addresses. */
1445 	i = 0;
1446 	netdev_for_each_mc_addr(ha, netdev)
1447 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
1448 
1449 	spin_lock_bh(&hw->mbx_lock);
1450 
1451 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i, 0, 0);
1452 
1453 	spin_unlock_bh(&hw->mbx_lock);
1454 	kfree(mta_list);
1455 }
1456 
1457 /**
1458  * igbvf_set_uni - Configure unicast MAC filters
1459  * @netdev: network interface device structure
1460  *
1461  * This routine is responsible for configuring the hardware for proper
1462  * unicast filters.
1463  **/
1464 static int igbvf_set_uni(struct net_device *netdev)
1465 {
1466 	struct igbvf_adapter *adapter = netdev_priv(netdev);
1467 	struct e1000_hw *hw = &adapter->hw;
1468 
1469 	if (netdev_uc_count(netdev) > IGBVF_MAX_MAC_FILTERS) {
1470 		pr_err("Too many unicast filters - No Space\n");
1471 		return -ENOSPC;
1472 	}
1473 
1474 	spin_lock_bh(&hw->mbx_lock);
1475 
1476 	/* Clear all unicast MAC filters */
1477 	hw->mac.ops.set_uc_addr(hw, E1000_VF_MAC_FILTER_CLR, NULL);
1478 
1479 	spin_unlock_bh(&hw->mbx_lock);
1480 
1481 	if (!netdev_uc_empty(netdev)) {
1482 		struct netdev_hw_addr *ha;
1483 
1484 		/* Add MAC filters one by one */
1485 		netdev_for_each_uc_addr(ha, netdev) {
1486 			spin_lock_bh(&hw->mbx_lock);
1487 
1488 			hw->mac.ops.set_uc_addr(hw, E1000_VF_MAC_FILTER_ADD,
1489 						ha->addr);
1490 
1491 			spin_unlock_bh(&hw->mbx_lock);
1492 			udelay(200);
1493 		}
1494 	}
1495 
1496 	return 0;
1497 }
1498 
1499 static void igbvf_set_rx_mode(struct net_device *netdev)
1500 {
1501 	igbvf_set_multi(netdev);
1502 	igbvf_set_uni(netdev);
1503 }
1504 
1505 /**
1506  * igbvf_configure - configure the hardware for Rx and Tx
1507  * @adapter: private board structure
1508  **/
1509 static void igbvf_configure(struct igbvf_adapter *adapter)
1510 {
1511 	igbvf_set_rx_mode(adapter->netdev);
1512 
1513 	igbvf_restore_vlan(adapter);
1514 
1515 	igbvf_configure_tx(adapter);
1516 	igbvf_setup_srrctl(adapter);
1517 	igbvf_configure_rx(adapter);
1518 	igbvf_alloc_rx_buffers(adapter->rx_ring,
1519 			       igbvf_desc_unused(adapter->rx_ring));
1520 }
1521 
1522 /* igbvf_reset - bring the hardware into a known good state
1523  * @adapter: private board structure
1524  *
1525  * This function boots the hardware and enables some settings that
1526  * require a configuration cycle of the hardware - those cannot be
1527  * set/changed during runtime. After reset the device needs to be
1528  * properly configured for Rx, Tx etc.
1529  */
1530 static void igbvf_reset(struct igbvf_adapter *adapter)
1531 {
1532 	struct e1000_mac_info *mac = &adapter->hw.mac;
1533 	struct net_device *netdev = adapter->netdev;
1534 	struct e1000_hw *hw = &adapter->hw;
1535 
1536 	spin_lock_bh(&hw->mbx_lock);
1537 
1538 	/* Allow time for pending master requests to run */
1539 	if (mac->ops.reset_hw(hw))
1540 		dev_err(&adapter->pdev->dev, "PF still resetting\n");
1541 
1542 	mac->ops.init_hw(hw);
1543 
1544 	spin_unlock_bh(&hw->mbx_lock);
1545 
1546 	if (is_valid_ether_addr(adapter->hw.mac.addr)) {
1547 		memcpy(netdev->dev_addr, adapter->hw.mac.addr,
1548 		       netdev->addr_len);
1549 		memcpy(netdev->perm_addr, adapter->hw.mac.addr,
1550 		       netdev->addr_len);
1551 	}
1552 
1553 	adapter->last_reset = jiffies;
1554 }
1555 
1556 int igbvf_up(struct igbvf_adapter *adapter)
1557 {
1558 	struct e1000_hw *hw = &adapter->hw;
1559 
1560 	/* hardware has been reset, we need to reload some things */
1561 	igbvf_configure(adapter);
1562 
1563 	clear_bit(__IGBVF_DOWN, &adapter->state);
1564 
1565 	napi_enable(&adapter->rx_ring->napi);
1566 	if (adapter->msix_entries)
1567 		igbvf_configure_msix(adapter);
1568 
1569 	/* Clear any pending interrupts. */
1570 	er32(EICR);
1571 	igbvf_irq_enable(adapter);
1572 
1573 	/* start the watchdog */
1574 	hw->mac.get_link_status = 1;
1575 	mod_timer(&adapter->watchdog_timer, jiffies + 1);
1576 
1577 	return 0;
1578 }
1579 
1580 void igbvf_down(struct igbvf_adapter *adapter)
1581 {
1582 	struct net_device *netdev = adapter->netdev;
1583 	struct e1000_hw *hw = &adapter->hw;
1584 	u32 rxdctl, txdctl;
1585 
1586 	/* signal that we're down so the interrupt handler does not
1587 	 * reschedule our watchdog timer
1588 	 */
1589 	set_bit(__IGBVF_DOWN, &adapter->state);
1590 
1591 	/* disable receives in the hardware */
1592 	rxdctl = er32(RXDCTL(0));
1593 	ew32(RXDCTL(0), rxdctl & ~E1000_RXDCTL_QUEUE_ENABLE);
1594 
1595 	netif_carrier_off(netdev);
1596 	netif_stop_queue(netdev);
1597 
1598 	/* disable transmits in the hardware */
1599 	txdctl = er32(TXDCTL(0));
1600 	ew32(TXDCTL(0), txdctl & ~E1000_TXDCTL_QUEUE_ENABLE);
1601 
1602 	/* flush both disables and wait for them to finish */
1603 	e1e_flush();
1604 	msleep(10);
1605 
1606 	napi_disable(&adapter->rx_ring->napi);
1607 
1608 	igbvf_irq_disable(adapter);
1609 
1610 	del_timer_sync(&adapter->watchdog_timer);
1611 
1612 	/* record the stats before reset*/
1613 	igbvf_update_stats(adapter);
1614 
1615 	adapter->link_speed = 0;
1616 	adapter->link_duplex = 0;
1617 
1618 	igbvf_reset(adapter);
1619 	igbvf_clean_tx_ring(adapter->tx_ring);
1620 	igbvf_clean_rx_ring(adapter->rx_ring);
1621 }
1622 
1623 void igbvf_reinit_locked(struct igbvf_adapter *adapter)
1624 {
1625 	might_sleep();
1626 	while (test_and_set_bit(__IGBVF_RESETTING, &adapter->state))
1627 		usleep_range(1000, 2000);
1628 	igbvf_down(adapter);
1629 	igbvf_up(adapter);
1630 	clear_bit(__IGBVF_RESETTING, &adapter->state);
1631 }
1632 
1633 /**
1634  * igbvf_sw_init - Initialize general software structures (struct igbvf_adapter)
1635  * @adapter: board private structure to initialize
1636  *
1637  * igbvf_sw_init initializes the Adapter private data structure.
1638  * Fields are initialized based on PCI device information and
1639  * OS network device settings (MTU size).
1640  **/
1641 static int igbvf_sw_init(struct igbvf_adapter *adapter)
1642 {
1643 	struct net_device *netdev = adapter->netdev;
1644 	s32 rc;
1645 
1646 	adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN;
1647 	adapter->rx_ps_hdr_size = 0;
1648 	adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
1649 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
1650 
1651 	adapter->tx_int_delay = 8;
1652 	adapter->tx_abs_int_delay = 32;
1653 	adapter->rx_int_delay = 0;
1654 	adapter->rx_abs_int_delay = 8;
1655 	adapter->requested_itr = 3;
1656 	adapter->current_itr = IGBVF_START_ITR;
1657 
1658 	/* Set various function pointers */
1659 	adapter->ei->init_ops(&adapter->hw);
1660 
1661 	rc = adapter->hw.mac.ops.init_params(&adapter->hw);
1662 	if (rc)
1663 		return rc;
1664 
1665 	rc = adapter->hw.mbx.ops.init_params(&adapter->hw);
1666 	if (rc)
1667 		return rc;
1668 
1669 	igbvf_set_interrupt_capability(adapter);
1670 
1671 	if (igbvf_alloc_queues(adapter))
1672 		return -ENOMEM;
1673 
1674 	spin_lock_init(&adapter->tx_queue_lock);
1675 
1676 	/* Explicitly disable IRQ since the NIC can be in any state. */
1677 	igbvf_irq_disable(adapter);
1678 
1679 	spin_lock_init(&adapter->stats_lock);
1680 	spin_lock_init(&adapter->hw.mbx_lock);
1681 
1682 	set_bit(__IGBVF_DOWN, &adapter->state);
1683 	return 0;
1684 }
1685 
1686 static void igbvf_initialize_last_counter_stats(struct igbvf_adapter *adapter)
1687 {
1688 	struct e1000_hw *hw = &adapter->hw;
1689 
1690 	adapter->stats.last_gprc = er32(VFGPRC);
1691 	adapter->stats.last_gorc = er32(VFGORC);
1692 	adapter->stats.last_gptc = er32(VFGPTC);
1693 	adapter->stats.last_gotc = er32(VFGOTC);
1694 	adapter->stats.last_mprc = er32(VFMPRC);
1695 	adapter->stats.last_gotlbc = er32(VFGOTLBC);
1696 	adapter->stats.last_gptlbc = er32(VFGPTLBC);
1697 	adapter->stats.last_gorlbc = er32(VFGORLBC);
1698 	adapter->stats.last_gprlbc = er32(VFGPRLBC);
1699 
1700 	adapter->stats.base_gprc = er32(VFGPRC);
1701 	adapter->stats.base_gorc = er32(VFGORC);
1702 	adapter->stats.base_gptc = er32(VFGPTC);
1703 	adapter->stats.base_gotc = er32(VFGOTC);
1704 	adapter->stats.base_mprc = er32(VFMPRC);
1705 	adapter->stats.base_gotlbc = er32(VFGOTLBC);
1706 	adapter->stats.base_gptlbc = er32(VFGPTLBC);
1707 	adapter->stats.base_gorlbc = er32(VFGORLBC);
1708 	adapter->stats.base_gprlbc = er32(VFGPRLBC);
1709 }
1710 
1711 /**
1712  * igbvf_open - Called when a network interface is made active
1713  * @netdev: network interface device structure
1714  *
1715  * Returns 0 on success, negative value on failure
1716  *
1717  * The open entry point is called when a network interface is made
1718  * active by the system (IFF_UP).  At this point all resources needed
1719  * for transmit and receive operations are allocated, the interrupt
1720  * handler is registered with the OS, the watchdog timer is started,
1721  * and the stack is notified that the interface is ready.
1722  **/
1723 static int igbvf_open(struct net_device *netdev)
1724 {
1725 	struct igbvf_adapter *adapter = netdev_priv(netdev);
1726 	struct e1000_hw *hw = &adapter->hw;
1727 	int err;
1728 
1729 	/* disallow open during test */
1730 	if (test_bit(__IGBVF_TESTING, &adapter->state))
1731 		return -EBUSY;
1732 
1733 	/* allocate transmit descriptors */
1734 	err = igbvf_setup_tx_resources(adapter, adapter->tx_ring);
1735 	if (err)
1736 		goto err_setup_tx;
1737 
1738 	/* allocate receive descriptors */
1739 	err = igbvf_setup_rx_resources(adapter, adapter->rx_ring);
1740 	if (err)
1741 		goto err_setup_rx;
1742 
1743 	/* before we allocate an interrupt, we must be ready to handle it.
1744 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
1745 	 * as soon as we call pci_request_irq, so we have to setup our
1746 	 * clean_rx handler before we do so.
1747 	 */
1748 	igbvf_configure(adapter);
1749 
1750 	err = igbvf_request_irq(adapter);
1751 	if (err)
1752 		goto err_req_irq;
1753 
1754 	/* From here on the code is the same as igbvf_up() */
1755 	clear_bit(__IGBVF_DOWN, &adapter->state);
1756 
1757 	napi_enable(&adapter->rx_ring->napi);
1758 
1759 	/* clear any pending interrupts */
1760 	er32(EICR);
1761 
1762 	igbvf_irq_enable(adapter);
1763 
1764 	/* start the watchdog */
1765 	hw->mac.get_link_status = 1;
1766 	mod_timer(&adapter->watchdog_timer, jiffies + 1);
1767 
1768 	return 0;
1769 
1770 err_req_irq:
1771 	igbvf_free_rx_resources(adapter->rx_ring);
1772 err_setup_rx:
1773 	igbvf_free_tx_resources(adapter->tx_ring);
1774 err_setup_tx:
1775 	igbvf_reset(adapter);
1776 
1777 	return err;
1778 }
1779 
1780 /**
1781  * igbvf_close - Disables a network interface
1782  * @netdev: network interface device structure
1783  *
1784  * Returns 0, this is not allowed to fail
1785  *
1786  * The close entry point is called when an interface is de-activated
1787  * by the OS.  The hardware is still under the drivers control, but
1788  * needs to be disabled.  A global MAC reset is issued to stop the
1789  * hardware, and all transmit and receive resources are freed.
1790  **/
1791 static int igbvf_close(struct net_device *netdev)
1792 {
1793 	struct igbvf_adapter *adapter = netdev_priv(netdev);
1794 
1795 	WARN_ON(test_bit(__IGBVF_RESETTING, &adapter->state));
1796 	igbvf_down(adapter);
1797 
1798 	igbvf_free_irq(adapter);
1799 
1800 	igbvf_free_tx_resources(adapter->tx_ring);
1801 	igbvf_free_rx_resources(adapter->rx_ring);
1802 
1803 	return 0;
1804 }
1805 
1806 /**
1807  * igbvf_set_mac - Change the Ethernet Address of the NIC
1808  * @netdev: network interface device structure
1809  * @p: pointer to an address structure
1810  *
1811  * Returns 0 on success, negative on failure
1812  **/
1813 static int igbvf_set_mac(struct net_device *netdev, void *p)
1814 {
1815 	struct igbvf_adapter *adapter = netdev_priv(netdev);
1816 	struct e1000_hw *hw = &adapter->hw;
1817 	struct sockaddr *addr = p;
1818 
1819 	if (!is_valid_ether_addr(addr->sa_data))
1820 		return -EADDRNOTAVAIL;
1821 
1822 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
1823 
1824 	spin_lock_bh(&hw->mbx_lock);
1825 
1826 	hw->mac.ops.rar_set(hw, hw->mac.addr, 0);
1827 
1828 	spin_unlock_bh(&hw->mbx_lock);
1829 
1830 	if (!ether_addr_equal(addr->sa_data, hw->mac.addr))
1831 		return -EADDRNOTAVAIL;
1832 
1833 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
1834 
1835 	return 0;
1836 }
1837 
1838 #define UPDATE_VF_COUNTER(reg, name) \
1839 { \
1840 	u32 current_counter = er32(reg); \
1841 	if (current_counter < adapter->stats.last_##name) \
1842 		adapter->stats.name += 0x100000000LL; \
1843 	adapter->stats.last_##name = current_counter; \
1844 	adapter->stats.name &= 0xFFFFFFFF00000000LL; \
1845 	adapter->stats.name |= current_counter; \
1846 }
1847 
1848 /**
1849  * igbvf_update_stats - Update the board statistics counters
1850  * @adapter: board private structure
1851 **/
1852 void igbvf_update_stats(struct igbvf_adapter *adapter)
1853 {
1854 	struct e1000_hw *hw = &adapter->hw;
1855 	struct pci_dev *pdev = adapter->pdev;
1856 
1857 	/* Prevent stats update while adapter is being reset, link is down
1858 	 * or if the pci connection is down.
1859 	 */
1860 	if (adapter->link_speed == 0)
1861 		return;
1862 
1863 	if (test_bit(__IGBVF_RESETTING, &adapter->state))
1864 		return;
1865 
1866 	if (pci_channel_offline(pdev))
1867 		return;
1868 
1869 	UPDATE_VF_COUNTER(VFGPRC, gprc);
1870 	UPDATE_VF_COUNTER(VFGORC, gorc);
1871 	UPDATE_VF_COUNTER(VFGPTC, gptc);
1872 	UPDATE_VF_COUNTER(VFGOTC, gotc);
1873 	UPDATE_VF_COUNTER(VFMPRC, mprc);
1874 	UPDATE_VF_COUNTER(VFGOTLBC, gotlbc);
1875 	UPDATE_VF_COUNTER(VFGPTLBC, gptlbc);
1876 	UPDATE_VF_COUNTER(VFGORLBC, gorlbc);
1877 	UPDATE_VF_COUNTER(VFGPRLBC, gprlbc);
1878 
1879 	/* Fill out the OS statistics structure */
1880 	adapter->netdev->stats.multicast = adapter->stats.mprc;
1881 }
1882 
1883 static void igbvf_print_link_info(struct igbvf_adapter *adapter)
1884 {
1885 	dev_info(&adapter->pdev->dev, "Link is Up %d Mbps %s Duplex\n",
1886 		 adapter->link_speed,
1887 		 adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half");
1888 }
1889 
1890 static bool igbvf_has_link(struct igbvf_adapter *adapter)
1891 {
1892 	struct e1000_hw *hw = &adapter->hw;
1893 	s32 ret_val = E1000_SUCCESS;
1894 	bool link_active;
1895 
1896 	/* If interface is down, stay link down */
1897 	if (test_bit(__IGBVF_DOWN, &adapter->state))
1898 		return false;
1899 
1900 	spin_lock_bh(&hw->mbx_lock);
1901 
1902 	ret_val = hw->mac.ops.check_for_link(hw);
1903 
1904 	spin_unlock_bh(&hw->mbx_lock);
1905 
1906 	link_active = !hw->mac.get_link_status;
1907 
1908 	/* if check for link returns error we will need to reset */
1909 	if (ret_val && time_after(jiffies, adapter->last_reset + (10 * HZ)))
1910 		schedule_work(&adapter->reset_task);
1911 
1912 	return link_active;
1913 }
1914 
1915 /**
1916  * igbvf_watchdog - Timer Call-back
1917  * @data: pointer to adapter cast into an unsigned long
1918  **/
1919 static void igbvf_watchdog(struct timer_list *t)
1920 {
1921 	struct igbvf_adapter *adapter = from_timer(adapter, t, watchdog_timer);
1922 
1923 	/* Do the rest outside of interrupt context */
1924 	schedule_work(&adapter->watchdog_task);
1925 }
1926 
1927 static void igbvf_watchdog_task(struct work_struct *work)
1928 {
1929 	struct igbvf_adapter *adapter = container_of(work,
1930 						     struct igbvf_adapter,
1931 						     watchdog_task);
1932 	struct net_device *netdev = adapter->netdev;
1933 	struct e1000_mac_info *mac = &adapter->hw.mac;
1934 	struct igbvf_ring *tx_ring = adapter->tx_ring;
1935 	struct e1000_hw *hw = &adapter->hw;
1936 	u32 link;
1937 	int tx_pending = 0;
1938 
1939 	link = igbvf_has_link(adapter);
1940 
1941 	if (link) {
1942 		if (!netif_carrier_ok(netdev)) {
1943 			mac->ops.get_link_up_info(&adapter->hw,
1944 						  &adapter->link_speed,
1945 						  &adapter->link_duplex);
1946 			igbvf_print_link_info(adapter);
1947 
1948 			netif_carrier_on(netdev);
1949 			netif_wake_queue(netdev);
1950 		}
1951 	} else {
1952 		if (netif_carrier_ok(netdev)) {
1953 			adapter->link_speed = 0;
1954 			adapter->link_duplex = 0;
1955 			dev_info(&adapter->pdev->dev, "Link is Down\n");
1956 			netif_carrier_off(netdev);
1957 			netif_stop_queue(netdev);
1958 		}
1959 	}
1960 
1961 	if (netif_carrier_ok(netdev)) {
1962 		igbvf_update_stats(adapter);
1963 	} else {
1964 		tx_pending = (igbvf_desc_unused(tx_ring) + 1 <
1965 			      tx_ring->count);
1966 		if (tx_pending) {
1967 			/* We've lost link, so the controller stops DMA,
1968 			 * but we've got queued Tx work that's never going
1969 			 * to get done, so reset controller to flush Tx.
1970 			 * (Do the reset outside of interrupt context).
1971 			 */
1972 			adapter->tx_timeout_count++;
1973 			schedule_work(&adapter->reset_task);
1974 		}
1975 	}
1976 
1977 	/* Cause software interrupt to ensure Rx ring is cleaned */
1978 	ew32(EICS, adapter->rx_ring->eims_value);
1979 
1980 	/* Reset the timer */
1981 	if (!test_bit(__IGBVF_DOWN, &adapter->state))
1982 		mod_timer(&adapter->watchdog_timer,
1983 			  round_jiffies(jiffies + (2 * HZ)));
1984 }
1985 
1986 #define IGBVF_TX_FLAGS_CSUM		0x00000001
1987 #define IGBVF_TX_FLAGS_VLAN		0x00000002
1988 #define IGBVF_TX_FLAGS_TSO		0x00000004
1989 #define IGBVF_TX_FLAGS_IPV4		0x00000008
1990 #define IGBVF_TX_FLAGS_VLAN_MASK	0xffff0000
1991 #define IGBVF_TX_FLAGS_VLAN_SHIFT	16
1992 
1993 static void igbvf_tx_ctxtdesc(struct igbvf_ring *tx_ring, u32 vlan_macip_lens,
1994 			      u32 type_tucmd, u32 mss_l4len_idx)
1995 {
1996 	struct e1000_adv_tx_context_desc *context_desc;
1997 	struct igbvf_buffer *buffer_info;
1998 	u16 i = tx_ring->next_to_use;
1999 
2000 	context_desc = IGBVF_TX_CTXTDESC_ADV(*tx_ring, i);
2001 	buffer_info = &tx_ring->buffer_info[i];
2002 
2003 	i++;
2004 	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
2005 
2006 	/* set bits to identify this as an advanced context descriptor */
2007 	type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
2008 
2009 	context_desc->vlan_macip_lens	= cpu_to_le32(vlan_macip_lens);
2010 	context_desc->seqnum_seed	= 0;
2011 	context_desc->type_tucmd_mlhl	= cpu_to_le32(type_tucmd);
2012 	context_desc->mss_l4len_idx	= cpu_to_le32(mss_l4len_idx);
2013 
2014 	buffer_info->time_stamp = jiffies;
2015 	buffer_info->dma = 0;
2016 }
2017 
2018 static int igbvf_tso(struct igbvf_ring *tx_ring,
2019 		     struct sk_buff *skb, u32 tx_flags, u8 *hdr_len)
2020 {
2021 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
2022 	union {
2023 		struct iphdr *v4;
2024 		struct ipv6hdr *v6;
2025 		unsigned char *hdr;
2026 	} ip;
2027 	union {
2028 		struct tcphdr *tcp;
2029 		unsigned char *hdr;
2030 	} l4;
2031 	u32 paylen, l4_offset;
2032 	int err;
2033 
2034 	if (skb->ip_summed != CHECKSUM_PARTIAL)
2035 		return 0;
2036 
2037 	if (!skb_is_gso(skb))
2038 		return 0;
2039 
2040 	err = skb_cow_head(skb, 0);
2041 	if (err < 0)
2042 		return err;
2043 
2044 	ip.hdr = skb_network_header(skb);
2045 	l4.hdr = skb_checksum_start(skb);
2046 
2047 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
2048 	type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
2049 
2050 	/* initialize outer IP header fields */
2051 	if (ip.v4->version == 4) {
2052 		unsigned char *csum_start = skb_checksum_start(skb);
2053 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
2054 
2055 		/* IP header will have to cancel out any data that
2056 		 * is not a part of the outer IP header
2057 		 */
2058 		ip.v4->check = csum_fold(csum_partial(trans_start,
2059 						      csum_start - trans_start,
2060 						      0));
2061 		type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
2062 
2063 		ip.v4->tot_len = 0;
2064 	} else {
2065 		ip.v6->payload_len = 0;
2066 	}
2067 
2068 	/* determine offset of inner transport header */
2069 	l4_offset = l4.hdr - skb->data;
2070 
2071 	/* compute length of segmentation header */
2072 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
2073 
2074 	/* remove payload length from inner checksum */
2075 	paylen = skb->len - l4_offset;
2076 	csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
2077 
2078 	/* MSS L4LEN IDX */
2079 	mss_l4len_idx = (*hdr_len - l4_offset) << E1000_ADVTXD_L4LEN_SHIFT;
2080 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
2081 
2082 	/* VLAN MACLEN IPLEN */
2083 	vlan_macip_lens = l4.hdr - ip.hdr;
2084 	vlan_macip_lens |= (ip.hdr - skb->data) << E1000_ADVTXD_MACLEN_SHIFT;
2085 	vlan_macip_lens |= tx_flags & IGBVF_TX_FLAGS_VLAN_MASK;
2086 
2087 	igbvf_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, mss_l4len_idx);
2088 
2089 	return 1;
2090 }
2091 
2092 static inline bool igbvf_ipv6_csum_is_sctp(struct sk_buff *skb)
2093 {
2094 	unsigned int offset = 0;
2095 
2096 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
2097 
2098 	return offset == skb_checksum_start_offset(skb);
2099 }
2100 
2101 static bool igbvf_tx_csum(struct igbvf_ring *tx_ring, struct sk_buff *skb,
2102 			  u32 tx_flags, __be16 protocol)
2103 {
2104 	u32 vlan_macip_lens = 0;
2105 	u32 type_tucmd = 0;
2106 
2107 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
2108 csum_failed:
2109 		if (!(tx_flags & IGBVF_TX_FLAGS_VLAN))
2110 			return false;
2111 		goto no_csum;
2112 	}
2113 
2114 	switch (skb->csum_offset) {
2115 	case offsetof(struct tcphdr, check):
2116 		type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
2117 		/* fall through */
2118 	case offsetof(struct udphdr, check):
2119 		break;
2120 	case offsetof(struct sctphdr, checksum):
2121 		/* validate that this is actually an SCTP request */
2122 		if (((protocol == htons(ETH_P_IP)) &&
2123 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
2124 		    ((protocol == htons(ETH_P_IPV6)) &&
2125 		     igbvf_ipv6_csum_is_sctp(skb))) {
2126 			type_tucmd = E1000_ADVTXD_TUCMD_L4T_SCTP;
2127 			break;
2128 		}
2129 	default:
2130 		skb_checksum_help(skb);
2131 		goto csum_failed;
2132 	}
2133 
2134 	vlan_macip_lens = skb_checksum_start_offset(skb) -
2135 			  skb_network_offset(skb);
2136 no_csum:
2137 	vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
2138 	vlan_macip_lens |= tx_flags & IGBVF_TX_FLAGS_VLAN_MASK;
2139 
2140 	igbvf_tx_ctxtdesc(tx_ring, vlan_macip_lens, type_tucmd, 0);
2141 	return true;
2142 }
2143 
2144 static int igbvf_maybe_stop_tx(struct net_device *netdev, int size)
2145 {
2146 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2147 
2148 	/* there is enough descriptors then we don't need to worry  */
2149 	if (igbvf_desc_unused(adapter->tx_ring) >= size)
2150 		return 0;
2151 
2152 	netif_stop_queue(netdev);
2153 
2154 	/* Herbert's original patch had:
2155 	 *  smp_mb__after_netif_stop_queue();
2156 	 * but since that doesn't exist yet, just open code it.
2157 	 */
2158 	smp_mb();
2159 
2160 	/* We need to check again just in case room has been made available */
2161 	if (igbvf_desc_unused(adapter->tx_ring) < size)
2162 		return -EBUSY;
2163 
2164 	netif_wake_queue(netdev);
2165 
2166 	++adapter->restart_queue;
2167 	return 0;
2168 }
2169 
2170 #define IGBVF_MAX_TXD_PWR	16
2171 #define IGBVF_MAX_DATA_PER_TXD	(1u << IGBVF_MAX_TXD_PWR)
2172 
2173 static inline int igbvf_tx_map_adv(struct igbvf_adapter *adapter,
2174 				   struct igbvf_ring *tx_ring,
2175 				   struct sk_buff *skb)
2176 {
2177 	struct igbvf_buffer *buffer_info;
2178 	struct pci_dev *pdev = adapter->pdev;
2179 	unsigned int len = skb_headlen(skb);
2180 	unsigned int count = 0, i;
2181 	unsigned int f;
2182 
2183 	i = tx_ring->next_to_use;
2184 
2185 	buffer_info = &tx_ring->buffer_info[i];
2186 	BUG_ON(len >= IGBVF_MAX_DATA_PER_TXD);
2187 	buffer_info->length = len;
2188 	/* set time_stamp *before* dma to help avoid a possible race */
2189 	buffer_info->time_stamp = jiffies;
2190 	buffer_info->mapped_as_page = false;
2191 	buffer_info->dma = dma_map_single(&pdev->dev, skb->data, len,
2192 					  DMA_TO_DEVICE);
2193 	if (dma_mapping_error(&pdev->dev, buffer_info->dma))
2194 		goto dma_error;
2195 
2196 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++) {
2197 		const struct skb_frag_struct *frag;
2198 
2199 		count++;
2200 		i++;
2201 		if (i == tx_ring->count)
2202 			i = 0;
2203 
2204 		frag = &skb_shinfo(skb)->frags[f];
2205 		len = skb_frag_size(frag);
2206 
2207 		buffer_info = &tx_ring->buffer_info[i];
2208 		BUG_ON(len >= IGBVF_MAX_DATA_PER_TXD);
2209 		buffer_info->length = len;
2210 		buffer_info->time_stamp = jiffies;
2211 		buffer_info->mapped_as_page = true;
2212 		buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag, 0, len,
2213 						    DMA_TO_DEVICE);
2214 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
2215 			goto dma_error;
2216 	}
2217 
2218 	tx_ring->buffer_info[i].skb = skb;
2219 
2220 	return ++count;
2221 
2222 dma_error:
2223 	dev_err(&pdev->dev, "TX DMA map failed\n");
2224 
2225 	/* clear timestamp and dma mappings for failed buffer_info mapping */
2226 	buffer_info->dma = 0;
2227 	buffer_info->time_stamp = 0;
2228 	buffer_info->length = 0;
2229 	buffer_info->mapped_as_page = false;
2230 	if (count)
2231 		count--;
2232 
2233 	/* clear timestamp and dma mappings for remaining portion of packet */
2234 	while (count--) {
2235 		if (i == 0)
2236 			i += tx_ring->count;
2237 		i--;
2238 		buffer_info = &tx_ring->buffer_info[i];
2239 		igbvf_put_txbuf(adapter, buffer_info);
2240 	}
2241 
2242 	return 0;
2243 }
2244 
2245 static inline void igbvf_tx_queue_adv(struct igbvf_adapter *adapter,
2246 				      struct igbvf_ring *tx_ring,
2247 				      int tx_flags, int count,
2248 				      unsigned int first, u32 paylen,
2249 				      u8 hdr_len)
2250 {
2251 	union e1000_adv_tx_desc *tx_desc = NULL;
2252 	struct igbvf_buffer *buffer_info;
2253 	u32 olinfo_status = 0, cmd_type_len;
2254 	unsigned int i;
2255 
2256 	cmd_type_len = (E1000_ADVTXD_DTYP_DATA | E1000_ADVTXD_DCMD_IFCS |
2257 			E1000_ADVTXD_DCMD_DEXT);
2258 
2259 	if (tx_flags & IGBVF_TX_FLAGS_VLAN)
2260 		cmd_type_len |= E1000_ADVTXD_DCMD_VLE;
2261 
2262 	if (tx_flags & IGBVF_TX_FLAGS_TSO) {
2263 		cmd_type_len |= E1000_ADVTXD_DCMD_TSE;
2264 
2265 		/* insert tcp checksum */
2266 		olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
2267 
2268 		/* insert ip checksum */
2269 		if (tx_flags & IGBVF_TX_FLAGS_IPV4)
2270 			olinfo_status |= E1000_TXD_POPTS_IXSM << 8;
2271 
2272 	} else if (tx_flags & IGBVF_TX_FLAGS_CSUM) {
2273 		olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
2274 	}
2275 
2276 	olinfo_status |= ((paylen - hdr_len) << E1000_ADVTXD_PAYLEN_SHIFT);
2277 
2278 	i = tx_ring->next_to_use;
2279 	while (count--) {
2280 		buffer_info = &tx_ring->buffer_info[i];
2281 		tx_desc = IGBVF_TX_DESC_ADV(*tx_ring, i);
2282 		tx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
2283 		tx_desc->read.cmd_type_len =
2284 			 cpu_to_le32(cmd_type_len | buffer_info->length);
2285 		tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
2286 		i++;
2287 		if (i == tx_ring->count)
2288 			i = 0;
2289 	}
2290 
2291 	tx_desc->read.cmd_type_len |= cpu_to_le32(adapter->txd_cmd);
2292 	/* Force memory writes to complete before letting h/w
2293 	 * know there are new descriptors to fetch.  (Only
2294 	 * applicable for weak-ordered memory model archs,
2295 	 * such as IA-64).
2296 	 */
2297 	wmb();
2298 
2299 	tx_ring->buffer_info[first].next_to_watch = tx_desc;
2300 	tx_ring->next_to_use = i;
2301 	writel(i, adapter->hw.hw_addr + tx_ring->tail);
2302 	/* we need this if more than one processor can write to our tail
2303 	 * at a time, it synchronizes IO on IA64/Altix systems
2304 	 */
2305 	mmiowb();
2306 }
2307 
2308 static netdev_tx_t igbvf_xmit_frame_ring_adv(struct sk_buff *skb,
2309 					     struct net_device *netdev,
2310 					     struct igbvf_ring *tx_ring)
2311 {
2312 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2313 	unsigned int first, tx_flags = 0;
2314 	u8 hdr_len = 0;
2315 	int count = 0;
2316 	int tso = 0;
2317 	__be16 protocol = vlan_get_protocol(skb);
2318 
2319 	if (test_bit(__IGBVF_DOWN, &adapter->state)) {
2320 		dev_kfree_skb_any(skb);
2321 		return NETDEV_TX_OK;
2322 	}
2323 
2324 	if (skb->len <= 0) {
2325 		dev_kfree_skb_any(skb);
2326 		return NETDEV_TX_OK;
2327 	}
2328 
2329 	/* need: count + 4 desc gap to keep tail from touching
2330 	 *       + 2 desc gap to keep tail from touching head,
2331 	 *       + 1 desc for skb->data,
2332 	 *       + 1 desc for context descriptor,
2333 	 * head, otherwise try next time
2334 	 */
2335 	if (igbvf_maybe_stop_tx(netdev, skb_shinfo(skb)->nr_frags + 4)) {
2336 		/* this is a hard error */
2337 		return NETDEV_TX_BUSY;
2338 	}
2339 
2340 	if (skb_vlan_tag_present(skb)) {
2341 		tx_flags |= IGBVF_TX_FLAGS_VLAN;
2342 		tx_flags |= (skb_vlan_tag_get(skb) <<
2343 			     IGBVF_TX_FLAGS_VLAN_SHIFT);
2344 	}
2345 
2346 	if (protocol == htons(ETH_P_IP))
2347 		tx_flags |= IGBVF_TX_FLAGS_IPV4;
2348 
2349 	first = tx_ring->next_to_use;
2350 
2351 	tso = igbvf_tso(tx_ring, skb, tx_flags, &hdr_len);
2352 	if (unlikely(tso < 0)) {
2353 		dev_kfree_skb_any(skb);
2354 		return NETDEV_TX_OK;
2355 	}
2356 
2357 	if (tso)
2358 		tx_flags |= IGBVF_TX_FLAGS_TSO;
2359 	else if (igbvf_tx_csum(tx_ring, skb, tx_flags, protocol) &&
2360 		 (skb->ip_summed == CHECKSUM_PARTIAL))
2361 		tx_flags |= IGBVF_TX_FLAGS_CSUM;
2362 
2363 	/* count reflects descriptors mapped, if 0 then mapping error
2364 	 * has occurred and we need to rewind the descriptor queue
2365 	 */
2366 	count = igbvf_tx_map_adv(adapter, tx_ring, skb);
2367 
2368 	if (count) {
2369 		igbvf_tx_queue_adv(adapter, tx_ring, tx_flags, count,
2370 				   first, skb->len, hdr_len);
2371 		/* Make sure there is space in the ring for the next send. */
2372 		igbvf_maybe_stop_tx(netdev, MAX_SKB_FRAGS + 4);
2373 	} else {
2374 		dev_kfree_skb_any(skb);
2375 		tx_ring->buffer_info[first].time_stamp = 0;
2376 		tx_ring->next_to_use = first;
2377 	}
2378 
2379 	return NETDEV_TX_OK;
2380 }
2381 
2382 static netdev_tx_t igbvf_xmit_frame(struct sk_buff *skb,
2383 				    struct net_device *netdev)
2384 {
2385 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2386 	struct igbvf_ring *tx_ring;
2387 
2388 	if (test_bit(__IGBVF_DOWN, &adapter->state)) {
2389 		dev_kfree_skb_any(skb);
2390 		return NETDEV_TX_OK;
2391 	}
2392 
2393 	tx_ring = &adapter->tx_ring[0];
2394 
2395 	return igbvf_xmit_frame_ring_adv(skb, netdev, tx_ring);
2396 }
2397 
2398 /**
2399  * igbvf_tx_timeout - Respond to a Tx Hang
2400  * @netdev: network interface device structure
2401  **/
2402 static void igbvf_tx_timeout(struct net_device *netdev)
2403 {
2404 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2405 
2406 	/* Do the reset outside of interrupt context */
2407 	adapter->tx_timeout_count++;
2408 	schedule_work(&adapter->reset_task);
2409 }
2410 
2411 static void igbvf_reset_task(struct work_struct *work)
2412 {
2413 	struct igbvf_adapter *adapter;
2414 
2415 	adapter = container_of(work, struct igbvf_adapter, reset_task);
2416 
2417 	igbvf_reinit_locked(adapter);
2418 }
2419 
2420 /**
2421  * igbvf_change_mtu - Change the Maximum Transfer Unit
2422  * @netdev: network interface device structure
2423  * @new_mtu: new value for maximum frame size
2424  *
2425  * Returns 0 on success, negative on failure
2426  **/
2427 static int igbvf_change_mtu(struct net_device *netdev, int new_mtu)
2428 {
2429 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2430 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
2431 
2432 	while (test_and_set_bit(__IGBVF_RESETTING, &adapter->state))
2433 		usleep_range(1000, 2000);
2434 	/* igbvf_down has a dependency on max_frame_size */
2435 	adapter->max_frame_size = max_frame;
2436 	if (netif_running(netdev))
2437 		igbvf_down(adapter);
2438 
2439 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
2440 	 * means we reserve 2 more, this pushes us to allocate from the next
2441 	 * larger slab size.
2442 	 * i.e. RXBUFFER_2048 --> size-4096 slab
2443 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
2444 	 * fragmented skbs
2445 	 */
2446 
2447 	if (max_frame <= 1024)
2448 		adapter->rx_buffer_len = 1024;
2449 	else if (max_frame <= 2048)
2450 		adapter->rx_buffer_len = 2048;
2451 	else
2452 #if (PAGE_SIZE / 2) > 16384
2453 		adapter->rx_buffer_len = 16384;
2454 #else
2455 		adapter->rx_buffer_len = PAGE_SIZE / 2;
2456 #endif
2457 
2458 	/* adjust allocation if LPE protects us, and we aren't using SBP */
2459 	if ((max_frame == ETH_FRAME_LEN + ETH_FCS_LEN) ||
2460 	    (max_frame == ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN))
2461 		adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN +
2462 					 ETH_FCS_LEN;
2463 
2464 	dev_info(&adapter->pdev->dev, "changing MTU from %d to %d\n",
2465 		 netdev->mtu, new_mtu);
2466 	netdev->mtu = new_mtu;
2467 
2468 	if (netif_running(netdev))
2469 		igbvf_up(adapter);
2470 	else
2471 		igbvf_reset(adapter);
2472 
2473 	clear_bit(__IGBVF_RESETTING, &adapter->state);
2474 
2475 	return 0;
2476 }
2477 
2478 static int igbvf_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2479 {
2480 	switch (cmd) {
2481 	default:
2482 		return -EOPNOTSUPP;
2483 	}
2484 }
2485 
2486 static int igbvf_suspend(struct pci_dev *pdev, pm_message_t state)
2487 {
2488 	struct net_device *netdev = pci_get_drvdata(pdev);
2489 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2490 #ifdef CONFIG_PM
2491 	int retval = 0;
2492 #endif
2493 
2494 	netif_device_detach(netdev);
2495 
2496 	if (netif_running(netdev)) {
2497 		WARN_ON(test_bit(__IGBVF_RESETTING, &adapter->state));
2498 		igbvf_down(adapter);
2499 		igbvf_free_irq(adapter);
2500 	}
2501 
2502 #ifdef CONFIG_PM
2503 	retval = pci_save_state(pdev);
2504 	if (retval)
2505 		return retval;
2506 #endif
2507 
2508 	pci_disable_device(pdev);
2509 
2510 	return 0;
2511 }
2512 
2513 #ifdef CONFIG_PM
2514 static int igbvf_resume(struct pci_dev *pdev)
2515 {
2516 	struct net_device *netdev = pci_get_drvdata(pdev);
2517 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2518 	u32 err;
2519 
2520 	pci_restore_state(pdev);
2521 	err = pci_enable_device_mem(pdev);
2522 	if (err) {
2523 		dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
2524 		return err;
2525 	}
2526 
2527 	pci_set_master(pdev);
2528 
2529 	if (netif_running(netdev)) {
2530 		err = igbvf_request_irq(adapter);
2531 		if (err)
2532 			return err;
2533 	}
2534 
2535 	igbvf_reset(adapter);
2536 
2537 	if (netif_running(netdev))
2538 		igbvf_up(adapter);
2539 
2540 	netif_device_attach(netdev);
2541 
2542 	return 0;
2543 }
2544 #endif
2545 
2546 static void igbvf_shutdown(struct pci_dev *pdev)
2547 {
2548 	igbvf_suspend(pdev, PMSG_SUSPEND);
2549 }
2550 
2551 #ifdef CONFIG_NET_POLL_CONTROLLER
2552 /* Polling 'interrupt' - used by things like netconsole to send skbs
2553  * without having to re-enable interrupts. It's not called while
2554  * the interrupt routine is executing.
2555  */
2556 static void igbvf_netpoll(struct net_device *netdev)
2557 {
2558 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2559 
2560 	disable_irq(adapter->pdev->irq);
2561 
2562 	igbvf_clean_tx_irq(adapter->tx_ring);
2563 
2564 	enable_irq(adapter->pdev->irq);
2565 }
2566 #endif
2567 
2568 /**
2569  * igbvf_io_error_detected - called when PCI error is detected
2570  * @pdev: Pointer to PCI device
2571  * @state: The current pci connection state
2572  *
2573  * This function is called after a PCI bus error affecting
2574  * this device has been detected.
2575  */
2576 static pci_ers_result_t igbvf_io_error_detected(struct pci_dev *pdev,
2577 						pci_channel_state_t state)
2578 {
2579 	struct net_device *netdev = pci_get_drvdata(pdev);
2580 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2581 
2582 	netif_device_detach(netdev);
2583 
2584 	if (state == pci_channel_io_perm_failure)
2585 		return PCI_ERS_RESULT_DISCONNECT;
2586 
2587 	if (netif_running(netdev))
2588 		igbvf_down(adapter);
2589 	pci_disable_device(pdev);
2590 
2591 	/* Request a slot slot reset. */
2592 	return PCI_ERS_RESULT_NEED_RESET;
2593 }
2594 
2595 /**
2596  * igbvf_io_slot_reset - called after the pci bus has been reset.
2597  * @pdev: Pointer to PCI device
2598  *
2599  * Restart the card from scratch, as if from a cold-boot. Implementation
2600  * resembles the first-half of the igbvf_resume routine.
2601  */
2602 static pci_ers_result_t igbvf_io_slot_reset(struct pci_dev *pdev)
2603 {
2604 	struct net_device *netdev = pci_get_drvdata(pdev);
2605 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2606 
2607 	if (pci_enable_device_mem(pdev)) {
2608 		dev_err(&pdev->dev,
2609 			"Cannot re-enable PCI device after reset.\n");
2610 		return PCI_ERS_RESULT_DISCONNECT;
2611 	}
2612 	pci_set_master(pdev);
2613 
2614 	igbvf_reset(adapter);
2615 
2616 	return PCI_ERS_RESULT_RECOVERED;
2617 }
2618 
2619 /**
2620  * igbvf_io_resume - called when traffic can start flowing again.
2621  * @pdev: Pointer to PCI device
2622  *
2623  * This callback is called when the error recovery driver tells us that
2624  * its OK to resume normal operation. Implementation resembles the
2625  * second-half of the igbvf_resume routine.
2626  */
2627 static void igbvf_io_resume(struct pci_dev *pdev)
2628 {
2629 	struct net_device *netdev = pci_get_drvdata(pdev);
2630 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2631 
2632 	if (netif_running(netdev)) {
2633 		if (igbvf_up(adapter)) {
2634 			dev_err(&pdev->dev,
2635 				"can't bring device back up after reset\n");
2636 			return;
2637 		}
2638 	}
2639 
2640 	netif_device_attach(netdev);
2641 }
2642 
2643 static void igbvf_print_device_info(struct igbvf_adapter *adapter)
2644 {
2645 	struct e1000_hw *hw = &adapter->hw;
2646 	struct net_device *netdev = adapter->netdev;
2647 	struct pci_dev *pdev = adapter->pdev;
2648 
2649 	if (hw->mac.type == e1000_vfadapt_i350)
2650 		dev_info(&pdev->dev, "Intel(R) I350 Virtual Function\n");
2651 	else
2652 		dev_info(&pdev->dev, "Intel(R) 82576 Virtual Function\n");
2653 	dev_info(&pdev->dev, "Address: %pM\n", netdev->dev_addr);
2654 }
2655 
2656 static int igbvf_set_features(struct net_device *netdev,
2657 			      netdev_features_t features)
2658 {
2659 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2660 
2661 	if (features & NETIF_F_RXCSUM)
2662 		adapter->flags &= ~IGBVF_FLAG_RX_CSUM_DISABLED;
2663 	else
2664 		adapter->flags |= IGBVF_FLAG_RX_CSUM_DISABLED;
2665 
2666 	return 0;
2667 }
2668 
2669 #define IGBVF_MAX_MAC_HDR_LEN		127
2670 #define IGBVF_MAX_NETWORK_HDR_LEN	511
2671 
2672 static netdev_features_t
2673 igbvf_features_check(struct sk_buff *skb, struct net_device *dev,
2674 		     netdev_features_t features)
2675 {
2676 	unsigned int network_hdr_len, mac_hdr_len;
2677 
2678 	/* Make certain the headers can be described by a context descriptor */
2679 	mac_hdr_len = skb_network_header(skb) - skb->data;
2680 	if (unlikely(mac_hdr_len > IGBVF_MAX_MAC_HDR_LEN))
2681 		return features & ~(NETIF_F_HW_CSUM |
2682 				    NETIF_F_SCTP_CRC |
2683 				    NETIF_F_HW_VLAN_CTAG_TX |
2684 				    NETIF_F_TSO |
2685 				    NETIF_F_TSO6);
2686 
2687 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
2688 	if (unlikely(network_hdr_len >  IGBVF_MAX_NETWORK_HDR_LEN))
2689 		return features & ~(NETIF_F_HW_CSUM |
2690 				    NETIF_F_SCTP_CRC |
2691 				    NETIF_F_TSO |
2692 				    NETIF_F_TSO6);
2693 
2694 	/* We can only support IPV4 TSO in tunnels if we can mangle the
2695 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
2696 	 */
2697 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
2698 		features &= ~NETIF_F_TSO;
2699 
2700 	return features;
2701 }
2702 
2703 static const struct net_device_ops igbvf_netdev_ops = {
2704 	.ndo_open		= igbvf_open,
2705 	.ndo_stop		= igbvf_close,
2706 	.ndo_start_xmit		= igbvf_xmit_frame,
2707 	.ndo_set_rx_mode	= igbvf_set_rx_mode,
2708 	.ndo_set_mac_address	= igbvf_set_mac,
2709 	.ndo_change_mtu		= igbvf_change_mtu,
2710 	.ndo_do_ioctl		= igbvf_ioctl,
2711 	.ndo_tx_timeout		= igbvf_tx_timeout,
2712 	.ndo_vlan_rx_add_vid	= igbvf_vlan_rx_add_vid,
2713 	.ndo_vlan_rx_kill_vid	= igbvf_vlan_rx_kill_vid,
2714 #ifdef CONFIG_NET_POLL_CONTROLLER
2715 	.ndo_poll_controller	= igbvf_netpoll,
2716 #endif
2717 	.ndo_set_features	= igbvf_set_features,
2718 	.ndo_features_check	= igbvf_features_check,
2719 };
2720 
2721 /**
2722  * igbvf_probe - Device Initialization Routine
2723  * @pdev: PCI device information struct
2724  * @ent: entry in igbvf_pci_tbl
2725  *
2726  * Returns 0 on success, negative on failure
2727  *
2728  * igbvf_probe initializes an adapter identified by a pci_dev structure.
2729  * The OS initialization, configuring of the adapter private structure,
2730  * and a hardware reset occur.
2731  **/
2732 static int igbvf_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2733 {
2734 	struct net_device *netdev;
2735 	struct igbvf_adapter *adapter;
2736 	struct e1000_hw *hw;
2737 	const struct igbvf_info *ei = igbvf_info_tbl[ent->driver_data];
2738 
2739 	static int cards_found;
2740 	int err, pci_using_dac;
2741 
2742 	err = pci_enable_device_mem(pdev);
2743 	if (err)
2744 		return err;
2745 
2746 	pci_using_dac = 0;
2747 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2748 	if (!err) {
2749 		pci_using_dac = 1;
2750 	} else {
2751 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
2752 		if (err) {
2753 			dev_err(&pdev->dev,
2754 				"No usable DMA configuration, aborting\n");
2755 			goto err_dma;
2756 		}
2757 	}
2758 
2759 	err = pci_request_regions(pdev, igbvf_driver_name);
2760 	if (err)
2761 		goto err_pci_reg;
2762 
2763 	pci_set_master(pdev);
2764 
2765 	err = -ENOMEM;
2766 	netdev = alloc_etherdev(sizeof(struct igbvf_adapter));
2767 	if (!netdev)
2768 		goto err_alloc_etherdev;
2769 
2770 	SET_NETDEV_DEV(netdev, &pdev->dev);
2771 
2772 	pci_set_drvdata(pdev, netdev);
2773 	adapter = netdev_priv(netdev);
2774 	hw = &adapter->hw;
2775 	adapter->netdev = netdev;
2776 	adapter->pdev = pdev;
2777 	adapter->ei = ei;
2778 	adapter->pba = ei->pba;
2779 	adapter->flags = ei->flags;
2780 	adapter->hw.back = adapter;
2781 	adapter->hw.mac.type = ei->mac;
2782 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
2783 
2784 	/* PCI config space info */
2785 
2786 	hw->vendor_id = pdev->vendor;
2787 	hw->device_id = pdev->device;
2788 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
2789 	hw->subsystem_device_id = pdev->subsystem_device;
2790 	hw->revision_id = pdev->revision;
2791 
2792 	err = -EIO;
2793 	adapter->hw.hw_addr = ioremap(pci_resource_start(pdev, 0),
2794 				      pci_resource_len(pdev, 0));
2795 
2796 	if (!adapter->hw.hw_addr)
2797 		goto err_ioremap;
2798 
2799 	if (ei->get_variants) {
2800 		err = ei->get_variants(adapter);
2801 		if (err)
2802 			goto err_get_variants;
2803 	}
2804 
2805 	/* setup adapter struct */
2806 	err = igbvf_sw_init(adapter);
2807 	if (err)
2808 		goto err_sw_init;
2809 
2810 	/* construct the net_device struct */
2811 	netdev->netdev_ops = &igbvf_netdev_ops;
2812 
2813 	igbvf_set_ethtool_ops(netdev);
2814 	netdev->watchdog_timeo = 5 * HZ;
2815 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
2816 
2817 	adapter->bd_number = cards_found++;
2818 
2819 	netdev->hw_features = NETIF_F_SG |
2820 			      NETIF_F_TSO |
2821 			      NETIF_F_TSO6 |
2822 			      NETIF_F_RXCSUM |
2823 			      NETIF_F_HW_CSUM |
2824 			      NETIF_F_SCTP_CRC;
2825 
2826 #define IGBVF_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
2827 				    NETIF_F_GSO_GRE_CSUM | \
2828 				    NETIF_F_GSO_IPXIP4 | \
2829 				    NETIF_F_GSO_IPXIP6 | \
2830 				    NETIF_F_GSO_UDP_TUNNEL | \
2831 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
2832 
2833 	netdev->gso_partial_features = IGBVF_GSO_PARTIAL_FEATURES;
2834 	netdev->hw_features |= NETIF_F_GSO_PARTIAL |
2835 			       IGBVF_GSO_PARTIAL_FEATURES;
2836 
2837 	netdev->features = netdev->hw_features;
2838 
2839 	if (pci_using_dac)
2840 		netdev->features |= NETIF_F_HIGHDMA;
2841 
2842 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
2843 	netdev->mpls_features |= NETIF_F_HW_CSUM;
2844 	netdev->hw_enc_features |= netdev->vlan_features;
2845 
2846 	/* set this bit last since it cannot be part of vlan_features */
2847 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
2848 			    NETIF_F_HW_VLAN_CTAG_RX |
2849 			    NETIF_F_HW_VLAN_CTAG_TX;
2850 
2851 	/* MTU range: 68 - 9216 */
2852 	netdev->min_mtu = ETH_MIN_MTU;
2853 	netdev->max_mtu = MAX_STD_JUMBO_FRAME_SIZE;
2854 
2855 	spin_lock_bh(&hw->mbx_lock);
2856 
2857 	/*reset the controller to put the device in a known good state */
2858 	err = hw->mac.ops.reset_hw(hw);
2859 	if (err) {
2860 		dev_info(&pdev->dev,
2861 			 "PF still in reset state. Is the PF interface up?\n");
2862 	} else {
2863 		err = hw->mac.ops.read_mac_addr(hw);
2864 		if (err)
2865 			dev_info(&pdev->dev, "Error reading MAC address.\n");
2866 		else if (is_zero_ether_addr(adapter->hw.mac.addr))
2867 			dev_info(&pdev->dev,
2868 				 "MAC address not assigned by administrator.\n");
2869 		memcpy(netdev->dev_addr, adapter->hw.mac.addr,
2870 		       netdev->addr_len);
2871 	}
2872 
2873 	spin_unlock_bh(&hw->mbx_lock);
2874 
2875 	if (!is_valid_ether_addr(netdev->dev_addr)) {
2876 		dev_info(&pdev->dev, "Assigning random MAC address.\n");
2877 		eth_hw_addr_random(netdev);
2878 		memcpy(adapter->hw.mac.addr, netdev->dev_addr,
2879 		       netdev->addr_len);
2880 	}
2881 
2882 	timer_setup(&adapter->watchdog_timer, igbvf_watchdog, 0);
2883 
2884 	INIT_WORK(&adapter->reset_task, igbvf_reset_task);
2885 	INIT_WORK(&adapter->watchdog_task, igbvf_watchdog_task);
2886 
2887 	/* ring size defaults */
2888 	adapter->rx_ring->count = 1024;
2889 	adapter->tx_ring->count = 1024;
2890 
2891 	/* reset the hardware with the new settings */
2892 	igbvf_reset(adapter);
2893 
2894 	/* set hardware-specific flags */
2895 	if (adapter->hw.mac.type == e1000_vfadapt_i350)
2896 		adapter->flags |= IGBVF_FLAG_RX_LB_VLAN_BSWAP;
2897 
2898 	strcpy(netdev->name, "eth%d");
2899 	err = register_netdev(netdev);
2900 	if (err)
2901 		goto err_hw_init;
2902 
2903 	/* tell the stack to leave us alone until igbvf_open() is called */
2904 	netif_carrier_off(netdev);
2905 	netif_stop_queue(netdev);
2906 
2907 	igbvf_print_device_info(adapter);
2908 
2909 	igbvf_initialize_last_counter_stats(adapter);
2910 
2911 	return 0;
2912 
2913 err_hw_init:
2914 	kfree(adapter->tx_ring);
2915 	kfree(adapter->rx_ring);
2916 err_sw_init:
2917 	igbvf_reset_interrupt_capability(adapter);
2918 err_get_variants:
2919 	iounmap(adapter->hw.hw_addr);
2920 err_ioremap:
2921 	free_netdev(netdev);
2922 err_alloc_etherdev:
2923 	pci_release_regions(pdev);
2924 err_pci_reg:
2925 err_dma:
2926 	pci_disable_device(pdev);
2927 	return err;
2928 }
2929 
2930 /**
2931  * igbvf_remove - Device Removal Routine
2932  * @pdev: PCI device information struct
2933  *
2934  * igbvf_remove is called by the PCI subsystem to alert the driver
2935  * that it should release a PCI device.  The could be caused by a
2936  * Hot-Plug event, or because the driver is going to be removed from
2937  * memory.
2938  **/
2939 static void igbvf_remove(struct pci_dev *pdev)
2940 {
2941 	struct net_device *netdev = pci_get_drvdata(pdev);
2942 	struct igbvf_adapter *adapter = netdev_priv(netdev);
2943 	struct e1000_hw *hw = &adapter->hw;
2944 
2945 	/* The watchdog timer may be rescheduled, so explicitly
2946 	 * disable it from being rescheduled.
2947 	 */
2948 	set_bit(__IGBVF_DOWN, &adapter->state);
2949 	del_timer_sync(&adapter->watchdog_timer);
2950 
2951 	cancel_work_sync(&adapter->reset_task);
2952 	cancel_work_sync(&adapter->watchdog_task);
2953 
2954 	unregister_netdev(netdev);
2955 
2956 	igbvf_reset_interrupt_capability(adapter);
2957 
2958 	/* it is important to delete the NAPI struct prior to freeing the
2959 	 * Rx ring so that you do not end up with null pointer refs
2960 	 */
2961 	netif_napi_del(&adapter->rx_ring->napi);
2962 	kfree(adapter->tx_ring);
2963 	kfree(adapter->rx_ring);
2964 
2965 	iounmap(hw->hw_addr);
2966 	if (hw->flash_address)
2967 		iounmap(hw->flash_address);
2968 	pci_release_regions(pdev);
2969 
2970 	free_netdev(netdev);
2971 
2972 	pci_disable_device(pdev);
2973 }
2974 
2975 /* PCI Error Recovery (ERS) */
2976 static const struct pci_error_handlers igbvf_err_handler = {
2977 	.error_detected = igbvf_io_error_detected,
2978 	.slot_reset = igbvf_io_slot_reset,
2979 	.resume = igbvf_io_resume,
2980 };
2981 
2982 static const struct pci_device_id igbvf_pci_tbl[] = {
2983 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_VF), board_vf },
2984 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_VF), board_i350_vf },
2985 	{ } /* terminate list */
2986 };
2987 MODULE_DEVICE_TABLE(pci, igbvf_pci_tbl);
2988 
2989 /* PCI Device API Driver */
2990 static struct pci_driver igbvf_driver = {
2991 	.name		= igbvf_driver_name,
2992 	.id_table	= igbvf_pci_tbl,
2993 	.probe		= igbvf_probe,
2994 	.remove		= igbvf_remove,
2995 #ifdef CONFIG_PM
2996 	/* Power Management Hooks */
2997 	.suspend	= igbvf_suspend,
2998 	.resume		= igbvf_resume,
2999 #endif
3000 	.shutdown	= igbvf_shutdown,
3001 	.err_handler	= &igbvf_err_handler
3002 };
3003 
3004 /**
3005  * igbvf_init_module - Driver Registration Routine
3006  *
3007  * igbvf_init_module is the first routine called when the driver is
3008  * loaded. All it does is register with the PCI subsystem.
3009  **/
3010 static int __init igbvf_init_module(void)
3011 {
3012 	int ret;
3013 
3014 	pr_info("%s - version %s\n", igbvf_driver_string, igbvf_driver_version);
3015 	pr_info("%s\n", igbvf_copyright);
3016 
3017 	ret = pci_register_driver(&igbvf_driver);
3018 
3019 	return ret;
3020 }
3021 module_init(igbvf_init_module);
3022 
3023 /**
3024  * igbvf_exit_module - Driver Exit Cleanup Routine
3025  *
3026  * igbvf_exit_module is called just before the driver is removed
3027  * from memory.
3028  **/
3029 static void __exit igbvf_exit_module(void)
3030 {
3031 	pci_unregister_driver(&igbvf_driver);
3032 }
3033 module_exit(igbvf_exit_module);
3034 
3035 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
3036 MODULE_DESCRIPTION("Intel(R) Gigabit Virtual Function Network Driver");
3037 MODULE_LICENSE("GPL");
3038 MODULE_VERSION(DRV_VERSION);
3039 
3040 /* netdev.c */
3041