1/******************************************************************************
2* This software may be used and distributed according to the terms of
3* the GNU General Public License (GPL), incorporated herein by reference.
4* Drivers based on or derived from this code fall under the GPL and must
5* retain the authorship, copyright and license notice.  This file is not
6* a complete program and may only be used when the entire operating
7* system is licensed under the GPL.
8* See the file COPYING in this distribution for more information.
9*
10* vxge-main.c: Driver for Exar Corp's X3100 Series 10GbE PCIe I/O
11*              Virtualized Server Adapter.
12* Copyright(c) 2002-2010 Exar Corp.
13*
14* The module loadable parameters that are supported by the driver and a brief
15* explanation of all the variables:
16* vlan_tag_strip:
17*	Strip VLAN Tag enable/disable. Instructs the device to remove
18*	the VLAN tag from all received tagged frames that are not
19*	replicated at the internal L2 switch.
20*		0 - Do not strip the VLAN tag.
21*		1 - Strip the VLAN tag.
22*
23* addr_learn_en:
24*	Enable learning the mac address of the guest OS interface in
25*	a virtualization environment.
26*		0 - DISABLE
27*		1 - ENABLE
28*
29* max_config_port:
30*	Maximum number of port to be supported.
31*		MIN -1 and MAX - 2
32*
33* max_config_vpath:
34*	This configures the maximum no of VPATH configures for each
35* 	device function.
36*		MIN - 1 and MAX - 17
37*
38* max_config_dev:
39*	This configures maximum no of Device function to be enabled.
40*		MIN - 1 and MAX - 17
41*
42******************************************************************************/
43
44#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
45
46#include <linux/bitops.h>
47#include <linux/if_vlan.h>
48#include <linux/interrupt.h>
49#include <linux/pci.h>
50#include <linux/slab.h>
51#include <linux/tcp.h>
52#include <net/ip.h>
53#include <linux/netdevice.h>
54#include <linux/etherdevice.h>
55#include <linux/firmware.h>
56#include <linux/net_tstamp.h>
57#include <linux/prefetch.h>
58#include <linux/module.h>
59#include "vxge-main.h"
60#include "vxge-reg.h"
61
62MODULE_LICENSE("Dual BSD/GPL");
63MODULE_DESCRIPTION("Neterion's X3100 Series 10GbE PCIe I/O"
64	"Virtualized Server Adapter");
65
66static const struct pci_device_id vxge_id_table[] = {
67	{PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_TITAN_WIN, PCI_ANY_ID,
68	PCI_ANY_ID},
69	{PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_TITAN_UNI, PCI_ANY_ID,
70	PCI_ANY_ID},
71	{0}
72};
73
74MODULE_DEVICE_TABLE(pci, vxge_id_table);
75
76VXGE_MODULE_PARAM_INT(vlan_tag_strip, VXGE_HW_VPATH_RPA_STRIP_VLAN_TAG_ENABLE);
77VXGE_MODULE_PARAM_INT(addr_learn_en, VXGE_HW_MAC_ADDR_LEARN_DEFAULT);
78VXGE_MODULE_PARAM_INT(max_config_port, VXGE_MAX_CONFIG_PORT);
79VXGE_MODULE_PARAM_INT(max_config_vpath, VXGE_USE_DEFAULT);
80VXGE_MODULE_PARAM_INT(max_mac_vpath, VXGE_MAX_MAC_ADDR_COUNT);
81VXGE_MODULE_PARAM_INT(max_config_dev, VXGE_MAX_CONFIG_DEV);
82
83static u16 vpath_selector[VXGE_HW_MAX_VIRTUAL_PATHS] =
84		{0, 1, 3, 3, 7, 7, 7, 7, 15, 15, 15, 15, 15, 15, 15, 15, 31};
85static unsigned int bw_percentage[VXGE_HW_MAX_VIRTUAL_PATHS] =
86	{[0 ...(VXGE_HW_MAX_VIRTUAL_PATHS - 1)] = 0xFF};
87module_param_array(bw_percentage, uint, NULL, 0);
88
89static struct vxge_drv_config *driver_config;
90static enum vxge_hw_status vxge_reset_all_vpaths(struct vxgedev *vdev);
91
92static inline int is_vxge_card_up(struct vxgedev *vdev)
93{
94	return test_bit(__VXGE_STATE_CARD_UP, &vdev->state);
95}
96
97static inline void VXGE_COMPLETE_VPATH_TX(struct vxge_fifo *fifo)
98{
99	struct sk_buff **skb_ptr = NULL;
100	struct sk_buff **temp;
101#define NR_SKB_COMPLETED 16
102	struct sk_buff *completed[NR_SKB_COMPLETED];
103	int more;
104
105	do {
106		more = 0;
107		skb_ptr = completed;
108
109		if (__netif_tx_trylock(fifo->txq)) {
110			vxge_hw_vpath_poll_tx(fifo->handle, &skb_ptr,
111						NR_SKB_COMPLETED, &more);
112			__netif_tx_unlock(fifo->txq);
113		}
114
115		/* free SKBs */
116		for (temp = completed; temp != skb_ptr; temp++)
117			dev_consume_skb_irq(*temp);
118	} while (more);
119}
120
121static inline void VXGE_COMPLETE_ALL_TX(struct vxgedev *vdev)
122{
123	int i;
124
125	/* Complete all transmits */
126	for (i = 0; i < vdev->no_of_vpath; i++)
127		VXGE_COMPLETE_VPATH_TX(&vdev->vpaths[i].fifo);
128}
129
130static inline void VXGE_COMPLETE_ALL_RX(struct vxgedev *vdev)
131{
132	int i;
133	struct vxge_ring *ring;
134
135	/* Complete all receives*/
136	for (i = 0; i < vdev->no_of_vpath; i++) {
137		ring = &vdev->vpaths[i].ring;
138		vxge_hw_vpath_poll_rx(ring->handle);
139	}
140}
141
142/*
143 * vxge_callback_link_up
144 *
145 * This function is called during interrupt context to notify link up state
146 * change.
147 */
148static void vxge_callback_link_up(struct __vxge_hw_device *hldev)
149{
150	struct net_device *dev = hldev->ndev;
151	struct vxgedev *vdev = netdev_priv(dev);
152
153	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
154		vdev->ndev->name, __func__, __LINE__);
155	netdev_notice(vdev->ndev, "Link Up\n");
156	vdev->stats.link_up++;
157
158	netif_carrier_on(vdev->ndev);
159	netif_tx_wake_all_queues(vdev->ndev);
160
161	vxge_debug_entryexit(VXGE_TRACE,
162		"%s: %s:%d Exiting...", vdev->ndev->name, __func__, __LINE__);
163}
164
165/*
166 * vxge_callback_link_down
167 *
168 * This function is called during interrupt context to notify link down state
169 * change.
170 */
171static void vxge_callback_link_down(struct __vxge_hw_device *hldev)
172{
173	struct net_device *dev = hldev->ndev;
174	struct vxgedev *vdev = netdev_priv(dev);
175
176	vxge_debug_entryexit(VXGE_TRACE,
177		"%s: %s:%d", vdev->ndev->name, __func__, __LINE__);
178	netdev_notice(vdev->ndev, "Link Down\n");
179
180	vdev->stats.link_down++;
181	netif_carrier_off(vdev->ndev);
182	netif_tx_stop_all_queues(vdev->ndev);
183
184	vxge_debug_entryexit(VXGE_TRACE,
185		"%s: %s:%d Exiting...", vdev->ndev->name, __func__, __LINE__);
186}
187
188/*
189 * vxge_rx_alloc
190 *
191 * Allocate SKB.
192 */
193static struct sk_buff *
194vxge_rx_alloc(void *dtrh, struct vxge_ring *ring, const int skb_size)
195{
196	struct net_device    *dev;
197	struct sk_buff       *skb;
198	struct vxge_rx_priv *rx_priv;
199
200	dev = ring->ndev;
201	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
202		ring->ndev->name, __func__, __LINE__);
203
204	rx_priv = vxge_hw_ring_rxd_private_get(dtrh);
205
206	/* try to allocate skb first. this one may fail */
207	skb = netdev_alloc_skb(dev, skb_size +
208	VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
209	if (skb == NULL) {
210		vxge_debug_mem(VXGE_ERR,
211			"%s: out of memory to allocate SKB", dev->name);
212		ring->stats.skb_alloc_fail++;
213		return NULL;
214	}
215
216	vxge_debug_mem(VXGE_TRACE,
217		"%s: %s:%d  Skb : 0x%p", ring->ndev->name,
218		__func__, __LINE__, skb);
219
220	skb_reserve(skb, VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
221
222	rx_priv->skb = skb;
223	rx_priv->skb_data = NULL;
224	rx_priv->data_size = skb_size;
225	vxge_debug_entryexit(VXGE_TRACE,
226		"%s: %s:%d Exiting...", ring->ndev->name, __func__, __LINE__);
227
228	return skb;
229}
230
231/*
232 * vxge_rx_map
233 */
234static int vxge_rx_map(void *dtrh, struct vxge_ring *ring)
235{
236	struct vxge_rx_priv *rx_priv;
237	dma_addr_t dma_addr;
238
239	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
240		ring->ndev->name, __func__, __LINE__);
241	rx_priv = vxge_hw_ring_rxd_private_get(dtrh);
242
243	rx_priv->skb_data = rx_priv->skb->data;
244	dma_addr = dma_map_single(&ring->pdev->dev, rx_priv->skb_data,
245				  rx_priv->data_size, DMA_FROM_DEVICE);
246
247	if (unlikely(dma_mapping_error(&ring->pdev->dev, dma_addr))) {
248		ring->stats.pci_map_fail++;
249		return -EIO;
250	}
251	vxge_debug_mem(VXGE_TRACE,
252		"%s: %s:%d  1 buffer mode dma_addr = 0x%llx",
253		ring->ndev->name, __func__, __LINE__,
254		(unsigned long long)dma_addr);
255	vxge_hw_ring_rxd_1b_set(dtrh, dma_addr, rx_priv->data_size);
256
257	rx_priv->data_dma = dma_addr;
258	vxge_debug_entryexit(VXGE_TRACE,
259		"%s: %s:%d Exiting...", ring->ndev->name, __func__, __LINE__);
260
261	return 0;
262}
263
264/*
265 * vxge_rx_initial_replenish
266 * Allocation of RxD as an initial replenish procedure.
267 */
268static enum vxge_hw_status
269vxge_rx_initial_replenish(void *dtrh, void *userdata)
270{
271	struct vxge_ring *ring = (struct vxge_ring *)userdata;
272	struct vxge_rx_priv *rx_priv;
273
274	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
275		ring->ndev->name, __func__, __LINE__);
276	if (vxge_rx_alloc(dtrh, ring,
277			  VXGE_LL_MAX_FRAME_SIZE(ring->ndev)) == NULL)
278		return VXGE_HW_FAIL;
279
280	if (vxge_rx_map(dtrh, ring)) {
281		rx_priv = vxge_hw_ring_rxd_private_get(dtrh);
282		dev_kfree_skb(rx_priv->skb);
283
284		return VXGE_HW_FAIL;
285	}
286	vxge_debug_entryexit(VXGE_TRACE,
287		"%s: %s:%d Exiting...", ring->ndev->name, __func__, __LINE__);
288
289	return VXGE_HW_OK;
290}
291
292static inline void
293vxge_rx_complete(struct vxge_ring *ring, struct sk_buff *skb, u16 vlan,
294		 int pkt_length, struct vxge_hw_ring_rxd_info *ext_info)
295{
296
297	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
298			ring->ndev->name, __func__, __LINE__);
299	skb_record_rx_queue(skb, ring->driver_id);
300	skb->protocol = eth_type_trans(skb, ring->ndev);
301
302	u64_stats_update_begin(&ring->stats.syncp);
303	ring->stats.rx_frms++;
304	ring->stats.rx_bytes += pkt_length;
305
306	if (skb->pkt_type == PACKET_MULTICAST)
307		ring->stats.rx_mcast++;
308	u64_stats_update_end(&ring->stats.syncp);
309
310	vxge_debug_rx(VXGE_TRACE,
311		"%s: %s:%d  skb protocol = %d",
312		ring->ndev->name, __func__, __LINE__, skb->protocol);
313
314	if (ext_info->vlan &&
315	    ring->vlan_tag_strip == VXGE_HW_VPATH_RPA_STRIP_VLAN_TAG_ENABLE)
316		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), ext_info->vlan);
317	napi_gro_receive(ring->napi_p, skb);
318
319	vxge_debug_entryexit(VXGE_TRACE,
320		"%s: %s:%d Exiting...", ring->ndev->name, __func__, __LINE__);
321}
322
323static inline void vxge_re_pre_post(void *dtr, struct vxge_ring *ring,
324				    struct vxge_rx_priv *rx_priv)
325{
326	dma_sync_single_for_device(&ring->pdev->dev, rx_priv->data_dma,
327				   rx_priv->data_size, DMA_FROM_DEVICE);
328
329	vxge_hw_ring_rxd_1b_set(dtr, rx_priv->data_dma, rx_priv->data_size);
330	vxge_hw_ring_rxd_pre_post(ring->handle, dtr);
331}
332
333static inline void vxge_post(int *dtr_cnt, void **first_dtr,
334			     void *post_dtr, struct __vxge_hw_ring *ringh)
335{
336	int dtr_count = *dtr_cnt;
337	if ((*dtr_cnt % VXGE_HW_RXSYNC_FREQ_CNT) == 0) {
338		if (*first_dtr)
339			vxge_hw_ring_rxd_post_post_wmb(ringh, *first_dtr);
340		*first_dtr = post_dtr;
341	} else
342		vxge_hw_ring_rxd_post_post(ringh, post_dtr);
343	dtr_count++;
344	*dtr_cnt = dtr_count;
345}
346
347/*
348 * vxge_rx_1b_compl
349 *
350 * If the interrupt is because of a received frame or if the receive ring
351 * contains fresh as yet un-processed frames, this function is called.
352 */
353static enum vxge_hw_status
354vxge_rx_1b_compl(struct __vxge_hw_ring *ringh, void *dtr,
355		 u8 t_code, void *userdata)
356{
357	struct vxge_ring *ring = (struct vxge_ring *)userdata;
358	struct net_device *dev = ring->ndev;
359	unsigned int dma_sizes;
360	void *first_dtr = NULL;
361	int dtr_cnt = 0;
362	int data_size;
363	dma_addr_t data_dma;
364	int pkt_length;
365	struct sk_buff *skb;
366	struct vxge_rx_priv *rx_priv;
367	struct vxge_hw_ring_rxd_info ext_info;
368	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
369		ring->ndev->name, __func__, __LINE__);
370
371	if (ring->budget <= 0)
372		goto out;
373
374	do {
375		prefetch((char *)dtr + L1_CACHE_BYTES);
376		rx_priv = vxge_hw_ring_rxd_private_get(dtr);
377		skb = rx_priv->skb;
378		data_size = rx_priv->data_size;
379		data_dma = rx_priv->data_dma;
380		prefetch(rx_priv->skb_data);
381
382		vxge_debug_rx(VXGE_TRACE,
383			"%s: %s:%d  skb = 0x%p",
384			ring->ndev->name, __func__, __LINE__, skb);
385
386		vxge_hw_ring_rxd_1b_get(ringh, dtr, &dma_sizes);
387		pkt_length = dma_sizes;
388
389		pkt_length -= ETH_FCS_LEN;
390
391		vxge_debug_rx(VXGE_TRACE,
392			"%s: %s:%d  Packet Length = %d",
393			ring->ndev->name, __func__, __LINE__, pkt_length);
394
395		vxge_hw_ring_rxd_1b_info_get(ringh, dtr, &ext_info);
396
397		/* check skb validity */
398		vxge_assert(skb);
399
400		prefetch((char *)skb + L1_CACHE_BYTES);
401		if (unlikely(t_code)) {
402			if (vxge_hw_ring_handle_tcode(ringh, dtr, t_code) !=
403				VXGE_HW_OK) {
404
405				ring->stats.rx_errors++;
406				vxge_debug_rx(VXGE_TRACE,
407					"%s: %s :%d Rx T_code is %d",
408					ring->ndev->name, __func__,
409					__LINE__, t_code);
410
411				/* If the t_code is not supported and if the
412				 * t_code is other than 0x5 (unparseable packet
413				 * such as unknown UPV6 header), Drop it !!!
414				 */
415				vxge_re_pre_post(dtr, ring, rx_priv);
416
417				vxge_post(&dtr_cnt, &first_dtr, dtr, ringh);
418				ring->stats.rx_dropped++;
419				continue;
420			}
421		}
422
423		if (pkt_length > VXGE_LL_RX_COPY_THRESHOLD) {
424			if (vxge_rx_alloc(dtr, ring, data_size) != NULL) {
425				if (!vxge_rx_map(dtr, ring)) {
426					skb_put(skb, pkt_length);
427
428					dma_unmap_single(&ring->pdev->dev,
429							 data_dma, data_size,
430							 DMA_FROM_DEVICE);
431
432					vxge_hw_ring_rxd_pre_post(ringh, dtr);
433					vxge_post(&dtr_cnt, &first_dtr, dtr,
434						ringh);
435				} else {
436					dev_kfree_skb(rx_priv->skb);
437					rx_priv->skb = skb;
438					rx_priv->data_size = data_size;
439					vxge_re_pre_post(dtr, ring, rx_priv);
440
441					vxge_post(&dtr_cnt, &first_dtr, dtr,
442						ringh);
443					ring->stats.rx_dropped++;
444					break;
445				}
446			} else {
447				vxge_re_pre_post(dtr, ring, rx_priv);
448
449				vxge_post(&dtr_cnt, &first_dtr, dtr, ringh);
450				ring->stats.rx_dropped++;
451				break;
452			}
453		} else {
454			struct sk_buff *skb_up;
455
456			skb_up = netdev_alloc_skb(dev, pkt_length +
457				VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
458			if (skb_up != NULL) {
459				skb_reserve(skb_up,
460				    VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
461
462				dma_sync_single_for_cpu(&ring->pdev->dev,
463							data_dma, data_size,
464							DMA_FROM_DEVICE);
465
466				vxge_debug_mem(VXGE_TRACE,
467					"%s: %s:%d  skb_up = %p",
468					ring->ndev->name, __func__,
469					__LINE__, skb);
470				memcpy(skb_up->data, skb->data, pkt_length);
471
472				vxge_re_pre_post(dtr, ring, rx_priv);
473
474				vxge_post(&dtr_cnt, &first_dtr, dtr,
475					ringh);
476				/* will netif_rx small SKB instead */
477				skb = skb_up;
478				skb_put(skb, pkt_length);
479			} else {
480				vxge_re_pre_post(dtr, ring, rx_priv);
481
482				vxge_post(&dtr_cnt, &first_dtr, dtr, ringh);
483				vxge_debug_rx(VXGE_ERR,
484					"%s: vxge_rx_1b_compl: out of "
485					"memory", dev->name);
486				ring->stats.skb_alloc_fail++;
487				break;
488			}
489		}
490
491		if ((ext_info.proto & VXGE_HW_FRAME_PROTO_TCP_OR_UDP) &&
492		    !(ext_info.proto & VXGE_HW_FRAME_PROTO_IP_FRAG) &&
493		    (dev->features & NETIF_F_RXCSUM) && /* Offload Rx side CSUM */
494		    ext_info.l3_cksum == VXGE_HW_L3_CKSUM_OK &&
495		    ext_info.l4_cksum == VXGE_HW_L4_CKSUM_OK)
496			skb->ip_summed = CHECKSUM_UNNECESSARY;
497		else
498			skb_checksum_none_assert(skb);
499
500
501		if (ring->rx_hwts) {
502			struct skb_shared_hwtstamps *skb_hwts;
503			u32 ns = *(u32 *)(skb->head + pkt_length);
504
505			skb_hwts = skb_hwtstamps(skb);
506			skb_hwts->hwtstamp = ns_to_ktime(ns);
507		}
508
509		/* rth_hash_type and rth_it_hit are non-zero regardless of
510		 * whether rss is enabled.  Only the rth_value is zero/non-zero
511		 * if rss is disabled/enabled, so key off of that.
512		 */
513		if (ext_info.rth_value)
514			skb_set_hash(skb, ext_info.rth_value,
515				     PKT_HASH_TYPE_L3);
516
517		vxge_rx_complete(ring, skb, ext_info.vlan,
518			pkt_length, &ext_info);
519
520		ring->budget--;
521		ring->pkts_processed++;
522		if (!ring->budget)
523			break;
524
525	} while (vxge_hw_ring_rxd_next_completed(ringh, &dtr,
526		&t_code) == VXGE_HW_OK);
527
528	if (first_dtr)
529		vxge_hw_ring_rxd_post_post_wmb(ringh, first_dtr);
530
531out:
532	vxge_debug_entryexit(VXGE_TRACE,
533				"%s:%d  Exiting...",
534				__func__, __LINE__);
535	return VXGE_HW_OK;
536}
537
538/*
539 * vxge_xmit_compl
540 *
541 * If an interrupt was raised to indicate DMA complete of the Tx packet,
542 * this function is called. It identifies the last TxD whose buffer was
543 * freed and frees all skbs whose data have already DMA'ed into the NICs
544 * internal memory.
545 */
546static enum vxge_hw_status
547vxge_xmit_compl(struct __vxge_hw_fifo *fifo_hw, void *dtr,
548		enum vxge_hw_fifo_tcode t_code, void *userdata,
549		struct sk_buff ***skb_ptr, int nr_skb, int *more)
550{
551	struct vxge_fifo *fifo = (struct vxge_fifo *)userdata;
552	struct sk_buff *skb, **done_skb = *skb_ptr;
553	int pkt_cnt = 0;
554
555	vxge_debug_entryexit(VXGE_TRACE,
556		"%s:%d Entered....", __func__, __LINE__);
557
558	do {
559		int frg_cnt;
560		skb_frag_t *frag;
561		int i = 0, j;
562		struct vxge_tx_priv *txd_priv =
563			vxge_hw_fifo_txdl_private_get(dtr);
564
565		skb = txd_priv->skb;
566		frg_cnt = skb_shinfo(skb)->nr_frags;
567		frag = &skb_shinfo(skb)->frags[0];
568
569		vxge_debug_tx(VXGE_TRACE,
570				"%s: %s:%d fifo_hw = %p dtr = %p "
571				"tcode = 0x%x", fifo->ndev->name, __func__,
572				__LINE__, fifo_hw, dtr, t_code);
573		/* check skb validity */
574		vxge_assert(skb);
575		vxge_debug_tx(VXGE_TRACE,
576			"%s: %s:%d skb = %p itxd_priv = %p frg_cnt = %d",
577			fifo->ndev->name, __func__, __LINE__,
578			skb, txd_priv, frg_cnt);
579		if (unlikely(t_code)) {
580			fifo->stats.tx_errors++;
581			vxge_debug_tx(VXGE_ERR,
582				"%s: tx: dtr %p completed due to "
583				"error t_code %01x", fifo->ndev->name,
584				dtr, t_code);
585			vxge_hw_fifo_handle_tcode(fifo_hw, dtr, t_code);
586		}
587
588		/*  for unfragmented skb */
589		dma_unmap_single(&fifo->pdev->dev, txd_priv->dma_buffers[i++],
590				 skb_headlen(skb), DMA_TO_DEVICE);
591
592		for (j = 0; j < frg_cnt; j++) {
593			dma_unmap_page(&fifo->pdev->dev,
594				       txd_priv->dma_buffers[i++],
595				       skb_frag_size(frag), DMA_TO_DEVICE);
596			frag += 1;
597		}
598
599		vxge_hw_fifo_txdl_free(fifo_hw, dtr);
600
601		/* Updating the statistics block */
602		u64_stats_update_begin(&fifo->stats.syncp);
603		fifo->stats.tx_frms++;
604		fifo->stats.tx_bytes += skb->len;
605		u64_stats_update_end(&fifo->stats.syncp);
606
607		*done_skb++ = skb;
608
609		if (--nr_skb <= 0) {
610			*more = 1;
611			break;
612		}
613
614		pkt_cnt++;
615		if (pkt_cnt > fifo->indicate_max_pkts)
616			break;
617
618	} while (vxge_hw_fifo_txdl_next_completed(fifo_hw,
619				&dtr, &t_code) == VXGE_HW_OK);
620
621	*skb_ptr = done_skb;
622	if (netif_tx_queue_stopped(fifo->txq))
623		netif_tx_wake_queue(fifo->txq);
624
625	vxge_debug_entryexit(VXGE_TRACE,
626				"%s: %s:%d  Exiting...",
627				fifo->ndev->name, __func__, __LINE__);
628	return VXGE_HW_OK;
629}
630
631/* select a vpath to transmit the packet */
632static u32 vxge_get_vpath_no(struct vxgedev *vdev, struct sk_buff *skb)
633{
634	u16 queue_len, counter = 0;
635	if (skb->protocol == htons(ETH_P_IP)) {
636		struct iphdr *ip;
637		struct tcphdr *th;
638
639		ip = ip_hdr(skb);
640
641		if (!ip_is_fragment(ip)) {
642			th = (struct tcphdr *)(((unsigned char *)ip) +
643					ip->ihl*4);
644
645			queue_len = vdev->no_of_vpath;
646			counter = (ntohs(th->source) +
647				ntohs(th->dest)) &
648				vdev->vpath_selector[queue_len - 1];
649			if (counter >= queue_len)
650				counter = queue_len - 1;
651		}
652	}
653	return counter;
654}
655
656static enum vxge_hw_status vxge_search_mac_addr_in_list(
657	struct vxge_vpath *vpath, u64 del_mac)
658{
659	struct list_head *entry, *next;
660	list_for_each_safe(entry, next, &vpath->mac_addr_list) {
661		if (((struct vxge_mac_addrs *)entry)->macaddr == del_mac)
662			return TRUE;
663	}
664	return FALSE;
665}
666
667static int vxge_mac_list_add(struct vxge_vpath *vpath, struct macInfo *mac)
668{
669	struct vxge_mac_addrs *new_mac_entry;
670	u8 *mac_address = NULL;
671
672	if (vpath->mac_addr_cnt >= VXGE_MAX_LEARN_MAC_ADDR_CNT)
673		return TRUE;
674
675	new_mac_entry = kzalloc(sizeof(struct vxge_mac_addrs), GFP_ATOMIC);
676	if (!new_mac_entry) {
677		vxge_debug_mem(VXGE_ERR,
678			"%s: memory allocation failed",
679			VXGE_DRIVER_NAME);
680		return FALSE;
681	}
682
683	list_add(&new_mac_entry->item, &vpath->mac_addr_list);
684
685	/* Copy the new mac address to the list */
686	mac_address = (u8 *)&new_mac_entry->macaddr;
687	memcpy(mac_address, mac->macaddr, ETH_ALEN);
688
689	new_mac_entry->state = mac->state;
690	vpath->mac_addr_cnt++;
691
692	if (is_multicast_ether_addr(mac->macaddr))
693		vpath->mcast_addr_cnt++;
694
695	return TRUE;
696}
697
698/* Add a mac address to DA table */
699static enum vxge_hw_status
700vxge_add_mac_addr(struct vxgedev *vdev, struct macInfo *mac)
701{
702	enum vxge_hw_status status = VXGE_HW_OK;
703	struct vxge_vpath *vpath;
704	enum vxge_hw_vpath_mac_addr_add_mode duplicate_mode;
705
706	if (is_multicast_ether_addr(mac->macaddr))
707		duplicate_mode = VXGE_HW_VPATH_MAC_ADDR_ADD_DUPLICATE;
708	else
709		duplicate_mode = VXGE_HW_VPATH_MAC_ADDR_REPLACE_DUPLICATE;
710
711	vpath = &vdev->vpaths[mac->vpath_no];
712	status = vxge_hw_vpath_mac_addr_add(vpath->handle, mac->macaddr,
713						mac->macmask, duplicate_mode);
714	if (status != VXGE_HW_OK) {
715		vxge_debug_init(VXGE_ERR,
716			"DA config add entry failed for vpath:%d",
717			vpath->device_id);
718	} else
719		if (FALSE == vxge_mac_list_add(vpath, mac))
720			status = -EPERM;
721
722	return status;
723}
724
725static int vxge_learn_mac(struct vxgedev *vdev, u8 *mac_header)
726{
727	struct macInfo mac_info;
728	u8 *mac_address = NULL;
729	u64 mac_addr = 0, vpath_vector = 0;
730	int vpath_idx = 0;
731	enum vxge_hw_status status = VXGE_HW_OK;
732	struct vxge_vpath *vpath = NULL;
733
734	mac_address = (u8 *)&mac_addr;
735	memcpy(mac_address, mac_header, ETH_ALEN);
736
737	/* Is this mac address already in the list? */
738	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
739		vpath = &vdev->vpaths[vpath_idx];
740		if (vxge_search_mac_addr_in_list(vpath, mac_addr))
741			return vpath_idx;
742	}
743
744	memset(&mac_info, 0, sizeof(struct macInfo));
745	memcpy(mac_info.macaddr, mac_header, ETH_ALEN);
746
747	/* Any vpath has room to add mac address to its da table? */
748	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
749		vpath = &vdev->vpaths[vpath_idx];
750		if (vpath->mac_addr_cnt < vpath->max_mac_addr_cnt) {
751			/* Add this mac address to this vpath */
752			mac_info.vpath_no = vpath_idx;
753			mac_info.state = VXGE_LL_MAC_ADDR_IN_DA_TABLE;
754			status = vxge_add_mac_addr(vdev, &mac_info);
755			if (status != VXGE_HW_OK)
756				return -EPERM;
757			return vpath_idx;
758		}
759	}
760
761	mac_info.state = VXGE_LL_MAC_ADDR_IN_LIST;
762	vpath_idx = 0;
763	mac_info.vpath_no = vpath_idx;
764	/* Is the first vpath already selected as catch-basin ? */
765	vpath = &vdev->vpaths[vpath_idx];
766	if (vpath->mac_addr_cnt > vpath->max_mac_addr_cnt) {
767		/* Add this mac address to this vpath */
768		if (FALSE == vxge_mac_list_add(vpath, &mac_info))
769			return -EPERM;
770		return vpath_idx;
771	}
772
773	/* Select first vpath as catch-basin */
774	vpath_vector = vxge_mBIT(vpath->device_id);
775	status = vxge_hw_mgmt_reg_write(vpath->vdev->devh,
776				vxge_hw_mgmt_reg_type_mrpcim,
777				0,
778				(ulong)offsetof(
779					struct vxge_hw_mrpcim_reg,
780					rts_mgr_cbasin_cfg),
781				vpath_vector);
782	if (status != VXGE_HW_OK) {
783		vxge_debug_tx(VXGE_ERR,
784			"%s: Unable to set the vpath-%d in catch-basin mode",
785			VXGE_DRIVER_NAME, vpath->device_id);
786		return -EPERM;
787	}
788
789	if (FALSE == vxge_mac_list_add(vpath, &mac_info))
790		return -EPERM;
791
792	return vpath_idx;
793}
794
795/**
796 * vxge_xmit
797 * @skb : the socket buffer containing the Tx data.
798 * @dev : device pointer.
799 *
800 * This function is the Tx entry point of the driver. Neterion NIC supports
801 * certain protocol assist features on Tx side, namely  CSO, S/G, LSO.
802*/
803static netdev_tx_t
804vxge_xmit(struct sk_buff *skb, struct net_device *dev)
805{
806	struct vxge_fifo *fifo = NULL;
807	void *dtr_priv;
808	void *dtr = NULL;
809	struct vxgedev *vdev = NULL;
810	enum vxge_hw_status status;
811	int frg_cnt, first_frg_len;
812	skb_frag_t *frag;
813	int i = 0, j = 0, avail;
814	u64 dma_pointer;
815	struct vxge_tx_priv *txdl_priv = NULL;
816	struct __vxge_hw_fifo *fifo_hw;
817	int offload_type;
818	int vpath_no = 0;
819
820	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
821			dev->name, __func__, __LINE__);
822
823	/* A buffer with no data will be dropped */
824	if (unlikely(skb->len <= 0)) {
825		vxge_debug_tx(VXGE_ERR,
826			"%s: Buffer has no data..", dev->name);
827		dev_kfree_skb_any(skb);
828		return NETDEV_TX_OK;
829	}
830
831	vdev = netdev_priv(dev);
832
833	if (unlikely(!is_vxge_card_up(vdev))) {
834		vxge_debug_tx(VXGE_ERR,
835			"%s: vdev not initialized", dev->name);
836		dev_kfree_skb_any(skb);
837		return NETDEV_TX_OK;
838	}
839
840	if (vdev->config.addr_learn_en) {
841		vpath_no = vxge_learn_mac(vdev, skb->data + ETH_ALEN);
842		if (vpath_no == -EPERM) {
843			vxge_debug_tx(VXGE_ERR,
844				"%s: Failed to store the mac address",
845				dev->name);
846			dev_kfree_skb_any(skb);
847			return NETDEV_TX_OK;
848		}
849	}
850
851	if (vdev->config.tx_steering_type == TX_MULTIQ_STEERING)
852		vpath_no = skb_get_queue_mapping(skb);
853	else if (vdev->config.tx_steering_type == TX_PORT_STEERING)
854		vpath_no = vxge_get_vpath_no(vdev, skb);
855
856	vxge_debug_tx(VXGE_TRACE, "%s: vpath_no= %d", dev->name, vpath_no);
857
858	if (vpath_no >= vdev->no_of_vpath)
859		vpath_no = 0;
860
861	fifo = &vdev->vpaths[vpath_no].fifo;
862	fifo_hw = fifo->handle;
863
864	if (netif_tx_queue_stopped(fifo->txq))
865		return NETDEV_TX_BUSY;
866
867	avail = vxge_hw_fifo_free_txdl_count_get(fifo_hw);
868	if (avail == 0) {
869		vxge_debug_tx(VXGE_ERR,
870			"%s: No free TXDs available", dev->name);
871		fifo->stats.txd_not_free++;
872		goto _exit0;
873	}
874
875	/* Last TXD?  Stop tx queue to avoid dropping packets.  TX
876	 * completion will resume the queue.
877	 */
878	if (avail == 1)
879		netif_tx_stop_queue(fifo->txq);
880
881	status = vxge_hw_fifo_txdl_reserve(fifo_hw, &dtr, &dtr_priv);
882	if (unlikely(status != VXGE_HW_OK)) {
883		vxge_debug_tx(VXGE_ERR,
884		   "%s: Out of descriptors .", dev->name);
885		fifo->stats.txd_out_of_desc++;
886		goto _exit0;
887	}
888
889	vxge_debug_tx(VXGE_TRACE,
890		"%s: %s:%d fifo_hw = %p dtr = %p dtr_priv = %p",
891		dev->name, __func__, __LINE__,
892		fifo_hw, dtr, dtr_priv);
893
894	if (skb_vlan_tag_present(skb)) {
895		u16 vlan_tag = skb_vlan_tag_get(skb);
896		vxge_hw_fifo_txdl_vlan_set(dtr, vlan_tag);
897	}
898
899	first_frg_len = skb_headlen(skb);
900
901	dma_pointer = dma_map_single(&fifo->pdev->dev, skb->data,
902				     first_frg_len, DMA_TO_DEVICE);
903
904	if (unlikely(dma_mapping_error(&fifo->pdev->dev, dma_pointer))) {
905		vxge_hw_fifo_txdl_free(fifo_hw, dtr);
906		fifo->stats.pci_map_fail++;
907		goto _exit0;
908	}
909
910	txdl_priv = vxge_hw_fifo_txdl_private_get(dtr);
911	txdl_priv->skb = skb;
912	txdl_priv->dma_buffers[j] = dma_pointer;
913
914	frg_cnt = skb_shinfo(skb)->nr_frags;
915	vxge_debug_tx(VXGE_TRACE,
916			"%s: %s:%d skb = %p txdl_priv = %p "
917			"frag_cnt = %d dma_pointer = 0x%llx", dev->name,
918			__func__, __LINE__, skb, txdl_priv,
919			frg_cnt, (unsigned long long)dma_pointer);
920
921	vxge_hw_fifo_txdl_buffer_set(fifo_hw, dtr, j++, dma_pointer,
922		first_frg_len);
923
924	frag = &skb_shinfo(skb)->frags[0];
925	for (i = 0; i < frg_cnt; i++) {
926		/* ignore 0 length fragment */
927		if (!skb_frag_size(frag))
928			continue;
929
930		dma_pointer = (u64)skb_frag_dma_map(&fifo->pdev->dev, frag,
931						    0, skb_frag_size(frag),
932						    DMA_TO_DEVICE);
933
934		if (unlikely(dma_mapping_error(&fifo->pdev->dev, dma_pointer)))
935			goto _exit2;
936		vxge_debug_tx(VXGE_TRACE,
937			"%s: %s:%d frag = %d dma_pointer = 0x%llx",
938				dev->name, __func__, __LINE__, i,
939				(unsigned long long)dma_pointer);
940
941		txdl_priv->dma_buffers[j] = dma_pointer;
942		vxge_hw_fifo_txdl_buffer_set(fifo_hw, dtr, j++, dma_pointer,
943					skb_frag_size(frag));
944		frag += 1;
945	}
946
947	offload_type = vxge_offload_type(skb);
948
949	if (offload_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)) {
950		int mss = vxge_tcp_mss(skb);
951		if (mss) {
952			vxge_debug_tx(VXGE_TRACE, "%s: %s:%d mss = %d",
953				dev->name, __func__, __LINE__, mss);
954			vxge_hw_fifo_txdl_mss_set(dtr, mss);
955		} else {
956			vxge_assert(skb->len <=
957				dev->mtu + VXGE_HW_MAC_HEADER_MAX_SIZE);
958			vxge_assert(0);
959			goto _exit1;
960		}
961	}
962
963	if (skb->ip_summed == CHECKSUM_PARTIAL)
964		vxge_hw_fifo_txdl_cksum_set_bits(dtr,
965					VXGE_HW_FIFO_TXD_TX_CKO_IPV4_EN |
966					VXGE_HW_FIFO_TXD_TX_CKO_TCP_EN |
967					VXGE_HW_FIFO_TXD_TX_CKO_UDP_EN);
968
969	vxge_hw_fifo_txdl_post(fifo_hw, dtr);
970
971	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d  Exiting...",
972		dev->name, __func__, __LINE__);
973	return NETDEV_TX_OK;
974
975_exit2:
976	vxge_debug_tx(VXGE_TRACE, "%s: pci_map_page failed", dev->name);
977_exit1:
978	j = 0;
979	frag = &skb_shinfo(skb)->frags[0];
980
981	dma_unmap_single(&fifo->pdev->dev, txdl_priv->dma_buffers[j++],
982			 skb_headlen(skb), DMA_TO_DEVICE);
983
984	for (; j < i; j++) {
985		dma_unmap_page(&fifo->pdev->dev, txdl_priv->dma_buffers[j],
986			       skb_frag_size(frag), DMA_TO_DEVICE);
987		frag += 1;
988	}
989
990	vxge_hw_fifo_txdl_free(fifo_hw, dtr);
991_exit0:
992	netif_tx_stop_queue(fifo->txq);
993	dev_kfree_skb_any(skb);
994
995	return NETDEV_TX_OK;
996}
997
998/*
999 * vxge_rx_term
1000 *
1001 * Function will be called by hw function to abort all outstanding receive
1002 * descriptors.
1003 */
1004static void
1005vxge_rx_term(void *dtrh, enum vxge_hw_rxd_state state, void *userdata)
1006{
1007	struct vxge_ring *ring = (struct vxge_ring *)userdata;
1008	struct vxge_rx_priv *rx_priv =
1009		vxge_hw_ring_rxd_private_get(dtrh);
1010
1011	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
1012			ring->ndev->name, __func__, __LINE__);
1013	if (state != VXGE_HW_RXD_STATE_POSTED)
1014		return;
1015
1016	dma_unmap_single(&ring->pdev->dev, rx_priv->data_dma,
1017			 rx_priv->data_size, DMA_FROM_DEVICE);
1018
1019	dev_kfree_skb(rx_priv->skb);
1020	rx_priv->skb_data = NULL;
1021
1022	vxge_debug_entryexit(VXGE_TRACE,
1023		"%s: %s:%d  Exiting...",
1024		ring->ndev->name, __func__, __LINE__);
1025}
1026
1027/*
1028 * vxge_tx_term
1029 *
1030 * Function will be called to abort all outstanding tx descriptors
1031 */
1032static void
1033vxge_tx_term(void *dtrh, enum vxge_hw_txdl_state state, void *userdata)
1034{
1035	struct vxge_fifo *fifo = (struct vxge_fifo *)userdata;
1036	skb_frag_t *frag;
1037	int i = 0, j, frg_cnt;
1038	struct vxge_tx_priv *txd_priv = vxge_hw_fifo_txdl_private_get(dtrh);
1039	struct sk_buff *skb = txd_priv->skb;
1040
1041	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1042
1043	if (state != VXGE_HW_TXDL_STATE_POSTED)
1044		return;
1045
1046	/* check skb validity */
1047	vxge_assert(skb);
1048	frg_cnt = skb_shinfo(skb)->nr_frags;
1049	frag = &skb_shinfo(skb)->frags[0];
1050
1051	/*  for unfragmented skb */
1052	dma_unmap_single(&fifo->pdev->dev, txd_priv->dma_buffers[i++],
1053			 skb_headlen(skb), DMA_TO_DEVICE);
1054
1055	for (j = 0; j < frg_cnt; j++) {
1056		dma_unmap_page(&fifo->pdev->dev, txd_priv->dma_buffers[i++],
1057			       skb_frag_size(frag), DMA_TO_DEVICE);
1058		frag += 1;
1059	}
1060
1061	dev_kfree_skb(skb);
1062
1063	vxge_debug_entryexit(VXGE_TRACE,
1064		"%s:%d  Exiting...", __func__, __LINE__);
1065}
1066
1067static int vxge_mac_list_del(struct vxge_vpath *vpath, struct macInfo *mac)
1068{
1069	struct list_head *entry, *next;
1070	u64 del_mac = 0;
1071	u8 *mac_address = (u8 *) (&del_mac);
1072
1073	/* Copy the mac address to delete from the list */
1074	memcpy(mac_address, mac->macaddr, ETH_ALEN);
1075
1076	list_for_each_safe(entry, next, &vpath->mac_addr_list) {
1077		if (((struct vxge_mac_addrs *)entry)->macaddr == del_mac) {
1078			list_del(entry);
1079			kfree(entry);
1080			vpath->mac_addr_cnt--;
1081
1082			if (is_multicast_ether_addr(mac->macaddr))
1083				vpath->mcast_addr_cnt--;
1084			return TRUE;
1085		}
1086	}
1087
1088	return FALSE;
1089}
1090
1091/* delete a mac address from DA table */
1092static enum vxge_hw_status
1093vxge_del_mac_addr(struct vxgedev *vdev, struct macInfo *mac)
1094{
1095	enum vxge_hw_status status = VXGE_HW_OK;
1096	struct vxge_vpath *vpath;
1097
1098	vpath = &vdev->vpaths[mac->vpath_no];
1099	status = vxge_hw_vpath_mac_addr_delete(vpath->handle, mac->macaddr,
1100						mac->macmask);
1101	if (status != VXGE_HW_OK) {
1102		vxge_debug_init(VXGE_ERR,
1103			"DA config delete entry failed for vpath:%d",
1104			vpath->device_id);
1105	} else
1106		vxge_mac_list_del(vpath, mac);
1107	return status;
1108}
1109
1110/**
1111 * vxge_set_multicast
1112 * @dev: pointer to the device structure
1113 *
1114 * Entry point for multicast address enable/disable
1115 * This function is a driver entry point which gets called by the kernel
1116 * whenever multicast addresses must be enabled/disabled. This also gets
1117 * called to set/reset promiscuous mode. Depending on the deivce flag, we
1118 * determine, if multicast address must be enabled or if promiscuous mode
1119 * is to be disabled etc.
1120 */
1121static void vxge_set_multicast(struct net_device *dev)
1122{
1123	struct netdev_hw_addr *ha;
1124	struct vxgedev *vdev;
1125	int i, mcast_cnt = 0;
1126	struct vxge_vpath *vpath;
1127	enum vxge_hw_status status = VXGE_HW_OK;
1128	struct macInfo mac_info;
1129	int vpath_idx = 0;
1130	struct vxge_mac_addrs *mac_entry;
1131	struct list_head *list_head;
1132	struct list_head *entry, *next;
1133	u8 *mac_address = NULL;
1134
1135	vxge_debug_entryexit(VXGE_TRACE,
1136		"%s:%d", __func__, __LINE__);
1137
1138	vdev = netdev_priv(dev);
1139
1140	if (unlikely(!is_vxge_card_up(vdev)))
1141		return;
1142
1143	if ((dev->flags & IFF_ALLMULTI) && (!vdev->all_multi_flg)) {
1144		for (i = 0; i < vdev->no_of_vpath; i++) {
1145			vpath = &vdev->vpaths[i];
1146			vxge_assert(vpath->is_open);
1147			status = vxge_hw_vpath_mcast_enable(vpath->handle);
1148			if (status != VXGE_HW_OK)
1149				vxge_debug_init(VXGE_ERR, "failed to enable "
1150						"multicast, status %d", status);
1151			vdev->all_multi_flg = 1;
1152		}
1153	} else if (!(dev->flags & IFF_ALLMULTI) && (vdev->all_multi_flg)) {
1154		for (i = 0; i < vdev->no_of_vpath; i++) {
1155			vpath = &vdev->vpaths[i];
1156			vxge_assert(vpath->is_open);
1157			status = vxge_hw_vpath_mcast_disable(vpath->handle);
1158			if (status != VXGE_HW_OK)
1159				vxge_debug_init(VXGE_ERR, "failed to disable "
1160						"multicast, status %d", status);
1161			vdev->all_multi_flg = 0;
1162		}
1163	}
1164
1165
1166	if (!vdev->config.addr_learn_en) {
1167		for (i = 0; i < vdev->no_of_vpath; i++) {
1168			vpath = &vdev->vpaths[i];
1169			vxge_assert(vpath->is_open);
1170
1171			if (dev->flags & IFF_PROMISC)
1172				status = vxge_hw_vpath_promisc_enable(
1173					vpath->handle);
1174			else
1175				status = vxge_hw_vpath_promisc_disable(
1176					vpath->handle);
1177			if (status != VXGE_HW_OK)
1178				vxge_debug_init(VXGE_ERR, "failed to %s promisc"
1179					", status %d", dev->flags&IFF_PROMISC ?
1180					"enable" : "disable", status);
1181		}
1182	}
1183
1184	memset(&mac_info, 0, sizeof(struct macInfo));
1185	/* Update individual M_CAST address list */
1186	if ((!vdev->all_multi_flg) && netdev_mc_count(dev)) {
1187		mcast_cnt = vdev->vpaths[0].mcast_addr_cnt;
1188		list_head = &vdev->vpaths[0].mac_addr_list;
1189		if ((netdev_mc_count(dev) +
1190			(vdev->vpaths[0].mac_addr_cnt - mcast_cnt)) >
1191				vdev->vpaths[0].max_mac_addr_cnt)
1192			goto _set_all_mcast;
1193
1194		/* Delete previous MC's */
1195		for (i = 0; i < mcast_cnt; i++) {
1196			list_for_each_safe(entry, next, list_head) {
1197				mac_entry = (struct vxge_mac_addrs *)entry;
1198				/* Copy the mac address to delete */
1199				mac_address = (u8 *)&mac_entry->macaddr;
1200				memcpy(mac_info.macaddr, mac_address, ETH_ALEN);
1201
1202				if (is_multicast_ether_addr(mac_info.macaddr)) {
1203					for (vpath_idx = 0; vpath_idx <
1204						vdev->no_of_vpath;
1205						vpath_idx++) {
1206						mac_info.vpath_no = vpath_idx;
1207						status = vxge_del_mac_addr(
1208								vdev,
1209								&mac_info);
1210					}
1211				}
1212			}
1213		}
1214
1215		/* Add new ones */
1216		netdev_for_each_mc_addr(ha, dev) {
1217			memcpy(mac_info.macaddr, ha->addr, ETH_ALEN);
1218			for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath;
1219					vpath_idx++) {
1220				mac_info.vpath_no = vpath_idx;
1221				mac_info.state = VXGE_LL_MAC_ADDR_IN_DA_TABLE;
1222				status = vxge_add_mac_addr(vdev, &mac_info);
1223				if (status != VXGE_HW_OK) {
1224					vxge_debug_init(VXGE_ERR,
1225						"%s:%d Setting individual"
1226						"multicast address failed",
1227						__func__, __LINE__);
1228					goto _set_all_mcast;
1229				}
1230			}
1231		}
1232
1233		return;
1234_set_all_mcast:
1235		mcast_cnt = vdev->vpaths[0].mcast_addr_cnt;
1236		/* Delete previous MC's */
1237		for (i = 0; i < mcast_cnt; i++) {
1238			list_for_each_safe(entry, next, list_head) {
1239				mac_entry = (struct vxge_mac_addrs *)entry;
1240				/* Copy the mac address to delete */
1241				mac_address = (u8 *)&mac_entry->macaddr;
1242				memcpy(mac_info.macaddr, mac_address, ETH_ALEN);
1243
1244				if (is_multicast_ether_addr(mac_info.macaddr))
1245					break;
1246			}
1247
1248			for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath;
1249					vpath_idx++) {
1250				mac_info.vpath_no = vpath_idx;
1251				status = vxge_del_mac_addr(vdev, &mac_info);
1252			}
1253		}
1254
1255		/* Enable all multicast */
1256		for (i = 0; i < vdev->no_of_vpath; i++) {
1257			vpath = &vdev->vpaths[i];
1258			vxge_assert(vpath->is_open);
1259
1260			status = vxge_hw_vpath_mcast_enable(vpath->handle);
1261			if (status != VXGE_HW_OK) {
1262				vxge_debug_init(VXGE_ERR,
1263					"%s:%d Enabling all multicasts failed",
1264					 __func__, __LINE__);
1265			}
1266			vdev->all_multi_flg = 1;
1267		}
1268		dev->flags |= IFF_ALLMULTI;
1269	}
1270
1271	vxge_debug_entryexit(VXGE_TRACE,
1272		"%s:%d  Exiting...", __func__, __LINE__);
1273}
1274
1275/**
1276 * vxge_set_mac_addr
1277 * @dev: pointer to the device structure
1278 * @p: socket info
1279 *
1280 * Update entry "0" (default MAC addr)
1281 */
1282static int vxge_set_mac_addr(struct net_device *dev, void *p)
1283{
1284	struct sockaddr *addr = p;
1285	struct vxgedev *vdev;
1286	enum vxge_hw_status status = VXGE_HW_OK;
1287	struct macInfo mac_info_new, mac_info_old;
1288	int vpath_idx = 0;
1289
1290	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1291
1292	vdev = netdev_priv(dev);
1293
1294	if (!is_valid_ether_addr(addr->sa_data))
1295		return -EINVAL;
1296
1297	memset(&mac_info_new, 0, sizeof(struct macInfo));
1298	memset(&mac_info_old, 0, sizeof(struct macInfo));
1299
1300	vxge_debug_entryexit(VXGE_TRACE, "%s:%d  Exiting...",
1301		__func__, __LINE__);
1302
1303	/* Get the old address */
1304	memcpy(mac_info_old.macaddr, dev->dev_addr, dev->addr_len);
1305
1306	/* Copy the new address */
1307	memcpy(mac_info_new.macaddr, addr->sa_data, dev->addr_len);
1308
1309	/* First delete the old mac address from all the vpaths
1310	as we can't specify the index while adding new mac address */
1311	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
1312		struct vxge_vpath *vpath = &vdev->vpaths[vpath_idx];
1313		if (!vpath->is_open) {
1314			/* This can happen when this interface is added/removed
1315			to the bonding interface. Delete this station address
1316			from the linked list */
1317			vxge_mac_list_del(vpath, &mac_info_old);
1318
1319			/* Add this new address to the linked list
1320			for later restoring */
1321			vxge_mac_list_add(vpath, &mac_info_new);
1322
1323			continue;
1324		}
1325		/* Delete the station address */
1326		mac_info_old.vpath_no = vpath_idx;
1327		status = vxge_del_mac_addr(vdev, &mac_info_old);
1328	}
1329
1330	if (unlikely(!is_vxge_card_up(vdev))) {
1331		memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
1332		return VXGE_HW_OK;
1333	}
1334
1335	/* Set this mac address to all the vpaths */
1336	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
1337		mac_info_new.vpath_no = vpath_idx;
1338		mac_info_new.state = VXGE_LL_MAC_ADDR_IN_DA_TABLE;
1339		status = vxge_add_mac_addr(vdev, &mac_info_new);
1340		if (status != VXGE_HW_OK)
1341			return -EINVAL;
1342	}
1343
1344	memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
1345
1346	return status;
1347}
1348
1349/*
1350 * vxge_vpath_intr_enable
1351 * @vdev: pointer to vdev
1352 * @vp_id: vpath for which to enable the interrupts
1353 *
1354 * Enables the interrupts for the vpath
1355*/
1356static void vxge_vpath_intr_enable(struct vxgedev *vdev, int vp_id)
1357{
1358	struct vxge_vpath *vpath = &vdev->vpaths[vp_id];
1359	int msix_id = 0;
1360	int tim_msix_id[4] = {0, 1, 0, 0};
1361	int alarm_msix_id = VXGE_ALARM_MSIX_ID;
1362
1363	vxge_hw_vpath_intr_enable(vpath->handle);
1364
1365	if (vdev->config.intr_type == INTA)
1366		vxge_hw_vpath_inta_unmask_tx_rx(vpath->handle);
1367	else {
1368		vxge_hw_vpath_msix_set(vpath->handle, tim_msix_id,
1369			alarm_msix_id);
1370
1371		msix_id = vpath->device_id * VXGE_HW_VPATH_MSIX_ACTIVE;
1372		vxge_hw_vpath_msix_unmask(vpath->handle, msix_id);
1373		vxge_hw_vpath_msix_unmask(vpath->handle, msix_id + 1);
1374
1375		/* enable the alarm vector */
1376		msix_id = (vpath->handle->vpath->hldev->first_vp_id *
1377			VXGE_HW_VPATH_MSIX_ACTIVE) + alarm_msix_id;
1378		vxge_hw_vpath_msix_unmask(vpath->handle, msix_id);
1379	}
1380}
1381
1382/*
1383 * vxge_vpath_intr_disable
1384 * @vdev: pointer to vdev
1385 * @vp_id: vpath for which to disable the interrupts
1386 *
1387 * Disables the interrupts for the vpath
1388*/
1389static void vxge_vpath_intr_disable(struct vxgedev *vdev, int vp_id)
1390{
1391	struct vxge_vpath *vpath = &vdev->vpaths[vp_id];
1392	struct __vxge_hw_device *hldev;
1393	int msix_id;
1394
1395	hldev = pci_get_drvdata(vdev->pdev);
1396
1397	vxge_hw_vpath_wait_receive_idle(hldev, vpath->device_id);
1398
1399	vxge_hw_vpath_intr_disable(vpath->handle);
1400
1401	if (vdev->config.intr_type == INTA)
1402		vxge_hw_vpath_inta_mask_tx_rx(vpath->handle);
1403	else {
1404		msix_id = vpath->device_id * VXGE_HW_VPATH_MSIX_ACTIVE;
1405		vxge_hw_vpath_msix_mask(vpath->handle, msix_id);
1406		vxge_hw_vpath_msix_mask(vpath->handle, msix_id + 1);
1407
1408		/* disable the alarm vector */
1409		msix_id = (vpath->handle->vpath->hldev->first_vp_id *
1410			VXGE_HW_VPATH_MSIX_ACTIVE) + VXGE_ALARM_MSIX_ID;
1411		vxge_hw_vpath_msix_mask(vpath->handle, msix_id);
1412	}
1413}
1414
1415/* list all mac addresses from DA table */
1416static enum vxge_hw_status
1417vxge_search_mac_addr_in_da_table(struct vxge_vpath *vpath, struct macInfo *mac)
1418{
1419	enum vxge_hw_status status = VXGE_HW_OK;
1420	unsigned char macmask[ETH_ALEN];
1421	unsigned char macaddr[ETH_ALEN];
1422
1423	status = vxge_hw_vpath_mac_addr_get(vpath->handle,
1424				macaddr, macmask);
1425	if (status != VXGE_HW_OK) {
1426		vxge_debug_init(VXGE_ERR,
1427			"DA config list entry failed for vpath:%d",
1428			vpath->device_id);
1429		return status;
1430	}
1431
1432	while (!ether_addr_equal(mac->macaddr, macaddr)) {
1433		status = vxge_hw_vpath_mac_addr_get_next(vpath->handle,
1434				macaddr, macmask);
1435		if (status != VXGE_HW_OK)
1436			break;
1437	}
1438
1439	return status;
1440}
1441
1442/* Store all mac addresses from the list to the DA table */
1443static enum vxge_hw_status vxge_restore_vpath_mac_addr(struct vxge_vpath *vpath)
1444{
1445	enum vxge_hw_status status = VXGE_HW_OK;
1446	struct macInfo mac_info;
1447	u8 *mac_address = NULL;
1448	struct list_head *entry, *next;
1449
1450	memset(&mac_info, 0, sizeof(struct macInfo));
1451
1452	if (vpath->is_open) {
1453		list_for_each_safe(entry, next, &vpath->mac_addr_list) {
1454			mac_address =
1455				(u8 *)&
1456				((struct vxge_mac_addrs *)entry)->macaddr;
1457			memcpy(mac_info.macaddr, mac_address, ETH_ALEN);
1458			((struct vxge_mac_addrs *)entry)->state =
1459				VXGE_LL_MAC_ADDR_IN_DA_TABLE;
1460			/* does this mac address already exist in da table? */
1461			status = vxge_search_mac_addr_in_da_table(vpath,
1462				&mac_info);
1463			if (status != VXGE_HW_OK) {
1464				/* Add this mac address to the DA table */
1465				status = vxge_hw_vpath_mac_addr_add(
1466					vpath->handle, mac_info.macaddr,
1467					mac_info.macmask,
1468				    VXGE_HW_VPATH_MAC_ADDR_ADD_DUPLICATE);
1469				if (status != VXGE_HW_OK) {
1470					vxge_debug_init(VXGE_ERR,
1471					    "DA add entry failed for vpath:%d",
1472					    vpath->device_id);
1473					((struct vxge_mac_addrs *)entry)->state
1474						= VXGE_LL_MAC_ADDR_IN_LIST;
1475				}
1476			}
1477		}
1478	}
1479
1480	return status;
1481}
1482
1483/* Store all vlan ids from the list to the vid table */
1484static enum vxge_hw_status
1485vxge_restore_vpath_vid_table(struct vxge_vpath *vpath)
1486{
1487	enum vxge_hw_status status = VXGE_HW_OK;
1488	struct vxgedev *vdev = vpath->vdev;
1489	u16 vid;
1490
1491	if (!vpath->is_open)
1492		return status;
1493
1494	for_each_set_bit(vid, vdev->active_vlans, VLAN_N_VID)
1495		status = vxge_hw_vpath_vid_add(vpath->handle, vid);
1496
1497	return status;
1498}
1499
1500/*
1501 * vxge_reset_vpath
1502 * @vdev: pointer to vdev
1503 * @vp_id: vpath to reset
1504 *
1505 * Resets the vpath
1506*/
1507static int vxge_reset_vpath(struct vxgedev *vdev, int vp_id)
1508{
1509	enum vxge_hw_status status = VXGE_HW_OK;
1510	struct vxge_vpath *vpath = &vdev->vpaths[vp_id];
1511	int ret = 0;
1512
1513	/* check if device is down already */
1514	if (unlikely(!is_vxge_card_up(vdev)))
1515		return 0;
1516
1517	/* is device reset already scheduled */
1518	if (test_bit(__VXGE_STATE_RESET_CARD, &vdev->state))
1519		return 0;
1520
1521	if (vpath->handle) {
1522		if (vxge_hw_vpath_reset(vpath->handle) == VXGE_HW_OK) {
1523			if (is_vxge_card_up(vdev) &&
1524				vxge_hw_vpath_recover_from_reset(vpath->handle)
1525					!= VXGE_HW_OK) {
1526				vxge_debug_init(VXGE_ERR,
1527					"vxge_hw_vpath_recover_from_reset"
1528					"failed for vpath:%d", vp_id);
1529				return status;
1530			}
1531		} else {
1532			vxge_debug_init(VXGE_ERR,
1533				"vxge_hw_vpath_reset failed for"
1534				"vpath:%d", vp_id);
1535			return status;
1536		}
1537	} else
1538		return VXGE_HW_FAIL;
1539
1540	vxge_restore_vpath_mac_addr(vpath);
1541	vxge_restore_vpath_vid_table(vpath);
1542
1543	/* Enable all broadcast */
1544	vxge_hw_vpath_bcast_enable(vpath->handle);
1545
1546	/* Enable all multicast */
1547	if (vdev->all_multi_flg) {
1548		status = vxge_hw_vpath_mcast_enable(vpath->handle);
1549		if (status != VXGE_HW_OK)
1550			vxge_debug_init(VXGE_ERR,
1551				"%s:%d Enabling multicast failed",
1552				__func__, __LINE__);
1553	}
1554
1555	/* Enable the interrupts */
1556	vxge_vpath_intr_enable(vdev, vp_id);
1557
1558	smp_wmb();
1559
1560	/* Enable the flow of traffic through the vpath */
1561	vxge_hw_vpath_enable(vpath->handle);
1562
1563	smp_wmb();
1564	vxge_hw_vpath_rx_doorbell_init(vpath->handle);
1565	vpath->ring.last_status = VXGE_HW_OK;
1566
1567	/* Vpath reset done */
1568	clear_bit(vp_id, &vdev->vp_reset);
1569
1570	/* Start the vpath queue */
1571	if (netif_tx_queue_stopped(vpath->fifo.txq))
1572		netif_tx_wake_queue(vpath->fifo.txq);
1573
1574	return ret;
1575}
1576
1577/* Configure CI */
1578static void vxge_config_ci_for_tti_rti(struct vxgedev *vdev)
1579{
1580	int i = 0;
1581
1582	/* Enable CI for RTI */
1583	if (vdev->config.intr_type == MSI_X) {
1584		for (i = 0; i < vdev->no_of_vpath; i++) {
1585			struct __vxge_hw_ring *hw_ring;
1586
1587			hw_ring = vdev->vpaths[i].ring.handle;
1588			vxge_hw_vpath_dynamic_rti_ci_set(hw_ring);
1589		}
1590	}
1591
1592	/* Enable CI for TTI */
1593	for (i = 0; i < vdev->no_of_vpath; i++) {
1594		struct __vxge_hw_fifo *hw_fifo = vdev->vpaths[i].fifo.handle;
1595		vxge_hw_vpath_tti_ci_set(hw_fifo);
1596		/*
1597		 * For Inta (with or without napi), Set CI ON for only one
1598		 * vpath. (Have only one free running timer).
1599		 */
1600		if ((vdev->config.intr_type == INTA) && (i == 0))
1601			break;
1602	}
1603
1604	return;
1605}
1606
1607static int do_vxge_reset(struct vxgedev *vdev, int event)
1608{
1609	enum vxge_hw_status status;
1610	int ret = 0, vp_id, i;
1611
1612	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1613
1614	if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_START_RESET)) {
1615		/* check if device is down already */
1616		if (unlikely(!is_vxge_card_up(vdev)))
1617			return 0;
1618
1619		/* is reset already scheduled */
1620		if (test_and_set_bit(__VXGE_STATE_RESET_CARD, &vdev->state))
1621			return 0;
1622	}
1623
1624	if (event == VXGE_LL_FULL_RESET) {
1625		netif_carrier_off(vdev->ndev);
1626
1627		/* wait for all the vpath reset to complete */
1628		for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
1629			while (test_bit(vp_id, &vdev->vp_reset))
1630				msleep(50);
1631		}
1632
1633		netif_carrier_on(vdev->ndev);
1634
1635		/* if execution mode is set to debug, don't reset the adapter */
1636		if (unlikely(vdev->exec_mode)) {
1637			vxge_debug_init(VXGE_ERR,
1638				"%s: execution mode is debug, returning..",
1639				vdev->ndev->name);
1640			clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
1641			netif_tx_stop_all_queues(vdev->ndev);
1642			return 0;
1643		}
1644	}
1645
1646	if (event == VXGE_LL_FULL_RESET) {
1647		vxge_hw_device_wait_receive_idle(vdev->devh);
1648		vxge_hw_device_intr_disable(vdev->devh);
1649
1650		switch (vdev->cric_err_event) {
1651		case VXGE_HW_EVENT_UNKNOWN:
1652			netif_tx_stop_all_queues(vdev->ndev);
1653			vxge_debug_init(VXGE_ERR,
1654				"fatal: %s: Disabling device due to"
1655				"unknown error",
1656				vdev->ndev->name);
1657			ret = -EPERM;
1658			goto out;
1659		case VXGE_HW_EVENT_RESET_START:
1660			break;
1661		case VXGE_HW_EVENT_RESET_COMPLETE:
1662		case VXGE_HW_EVENT_LINK_DOWN:
1663		case VXGE_HW_EVENT_LINK_UP:
1664		case VXGE_HW_EVENT_ALARM_CLEARED:
1665		case VXGE_HW_EVENT_ECCERR:
1666		case VXGE_HW_EVENT_MRPCIM_ECCERR:
1667			ret = -EPERM;
1668			goto out;
1669		case VXGE_HW_EVENT_FIFO_ERR:
1670		case VXGE_HW_EVENT_VPATH_ERR:
1671			break;
1672		case VXGE_HW_EVENT_CRITICAL_ERR:
1673			netif_tx_stop_all_queues(vdev->ndev);
1674			vxge_debug_init(VXGE_ERR,
1675				"fatal: %s: Disabling device due to"
1676				"serious error",
1677				vdev->ndev->name);
1678			/* SOP or device reset required */
1679			/* This event is not currently used */
1680			ret = -EPERM;
1681			goto out;
1682		case VXGE_HW_EVENT_SERR:
1683			netif_tx_stop_all_queues(vdev->ndev);
1684			vxge_debug_init(VXGE_ERR,
1685				"fatal: %s: Disabling device due to"
1686				"serious error",
1687				vdev->ndev->name);
1688			ret = -EPERM;
1689			goto out;
1690		case VXGE_HW_EVENT_SRPCIM_SERR:
1691		case VXGE_HW_EVENT_MRPCIM_SERR:
1692			ret = -EPERM;
1693			goto out;
1694		case VXGE_HW_EVENT_SLOT_FREEZE:
1695			netif_tx_stop_all_queues(vdev->ndev);
1696			vxge_debug_init(VXGE_ERR,
1697				"fatal: %s: Disabling device due to"
1698				"slot freeze",
1699				vdev->ndev->name);
1700			ret = -EPERM;
1701			goto out;
1702		default:
1703			break;
1704
1705		}
1706	}
1707
1708	if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_START_RESET))
1709		netif_tx_stop_all_queues(vdev->ndev);
1710
1711	if (event == VXGE_LL_FULL_RESET) {
1712		status = vxge_reset_all_vpaths(vdev);
1713		if (status != VXGE_HW_OK) {
1714			vxge_debug_init(VXGE_ERR,
1715				"fatal: %s: can not reset vpaths",
1716				vdev->ndev->name);
1717			ret = -EPERM;
1718			goto out;
1719		}
1720	}
1721
1722	if (event == VXGE_LL_COMPL_RESET) {
1723		for (i = 0; i < vdev->no_of_vpath; i++)
1724			if (vdev->vpaths[i].handle) {
1725				if (vxge_hw_vpath_recover_from_reset(
1726					vdev->vpaths[i].handle)
1727						!= VXGE_HW_OK) {
1728					vxge_debug_init(VXGE_ERR,
1729						"vxge_hw_vpath_recover_"
1730						"from_reset failed for vpath: "
1731						"%d", i);
1732					ret = -EPERM;
1733					goto out;
1734				}
1735				} else {
1736					vxge_debug_init(VXGE_ERR,
1737					"vxge_hw_vpath_reset failed for "
1738						"vpath:%d", i);
1739					ret = -EPERM;
1740					goto out;
1741				}
1742	}
1743
1744	if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_COMPL_RESET)) {
1745		/* Reprogram the DA table with populated mac addresses */
1746		for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
1747			vxge_restore_vpath_mac_addr(&vdev->vpaths[vp_id]);
1748			vxge_restore_vpath_vid_table(&vdev->vpaths[vp_id]);
1749		}
1750
1751		/* enable vpath interrupts */
1752		for (i = 0; i < vdev->no_of_vpath; i++)
1753			vxge_vpath_intr_enable(vdev, i);
1754
1755		vxge_hw_device_intr_enable(vdev->devh);
1756
1757		smp_wmb();
1758
1759		/* Indicate card up */
1760		set_bit(__VXGE_STATE_CARD_UP, &vdev->state);
1761
1762		/* Get the traffic to flow through the vpaths */
1763		for (i = 0; i < vdev->no_of_vpath; i++) {
1764			vxge_hw_vpath_enable(vdev->vpaths[i].handle);
1765			smp_wmb();
1766			vxge_hw_vpath_rx_doorbell_init(vdev->vpaths[i].handle);
1767		}
1768
1769		netif_tx_wake_all_queues(vdev->ndev);
1770	}
1771
1772	/* configure CI */
1773	vxge_config_ci_for_tti_rti(vdev);
1774
1775out:
1776	vxge_debug_entryexit(VXGE_TRACE,
1777		"%s:%d  Exiting...", __func__, __LINE__);
1778
1779	/* Indicate reset done */
1780	if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_COMPL_RESET))
1781		clear_bit(__VXGE_STATE_RESET_CARD, &vdev->state);
1782	return ret;
1783}
1784
1785/*
1786 * vxge_reset
1787 * @vdev: pointer to ll device
1788 *
1789 * driver may reset the chip on events of serr, eccerr, etc
1790 */
1791static void vxge_reset(struct work_struct *work)
1792{
1793	struct vxgedev *vdev = container_of(work, struct vxgedev, reset_task);
1794
1795	if (!netif_running(vdev->ndev))
1796		return;
1797
1798	do_vxge_reset(vdev, VXGE_LL_FULL_RESET);
1799}
1800
1801/**
1802 * vxge_poll - Receive handler when Receive Polling is used.
1803 * @napi: pointer to the napi structure.
1804 * @budget: Number of packets budgeted to be processed in this iteration.
1805 *
1806 * This function comes into picture only if Receive side is being handled
1807 * through polling (called NAPI in linux). It mostly does what the normal
1808 * Rx interrupt handler does in terms of descriptor and packet processing
1809 * but not in an interrupt context. Also it will process a specified number
1810 * of packets at most in one iteration. This value is passed down by the
1811 * kernel as the function argument 'budget'.
1812 */
1813static int vxge_poll_msix(struct napi_struct *napi, int budget)
1814{
1815	struct vxge_ring *ring = container_of(napi, struct vxge_ring, napi);
1816	int pkts_processed;
1817	int budget_org = budget;
1818
1819	ring->budget = budget;
1820	ring->pkts_processed = 0;
1821	vxge_hw_vpath_poll_rx(ring->handle);
1822	pkts_processed = ring->pkts_processed;
1823
1824	if (pkts_processed < budget_org) {
1825		napi_complete_done(napi, pkts_processed);
1826
1827		/* Re enable the Rx interrupts for the vpath */
1828		vxge_hw_channel_msix_unmask(
1829				(struct __vxge_hw_channel *)ring->handle,
1830				ring->rx_vector_no);
1831	}
1832
1833	/* We are copying and returning the local variable, in case if after
1834	 * clearing the msix interrupt above, if the interrupt fires right
1835	 * away which can preempt this NAPI thread */
1836	return pkts_processed;
1837}
1838
1839static int vxge_poll_inta(struct napi_struct *napi, int budget)
1840{
1841	struct vxgedev *vdev = container_of(napi, struct vxgedev, napi);
1842	int pkts_processed = 0;
1843	int i;
1844	int budget_org = budget;
1845	struct vxge_ring *ring;
1846
1847	struct __vxge_hw_device *hldev = pci_get_drvdata(vdev->pdev);
1848
1849	for (i = 0; i < vdev->no_of_vpath; i++) {
1850		ring = &vdev->vpaths[i].ring;
1851		ring->budget = budget;
1852		ring->pkts_processed = 0;
1853		vxge_hw_vpath_poll_rx(ring->handle);
1854		pkts_processed += ring->pkts_processed;
1855		budget -= ring->pkts_processed;
1856		if (budget <= 0)
1857			break;
1858	}
1859
1860	VXGE_COMPLETE_ALL_TX(vdev);
1861
1862	if (pkts_processed < budget_org) {
1863		napi_complete_done(napi, pkts_processed);
1864		/* Re enable the Rx interrupts for the ring */
1865		vxge_hw_device_unmask_all(hldev);
1866		vxge_hw_device_flush_io(hldev);
1867	}
1868
1869	return pkts_processed;
1870}
1871
1872#ifdef CONFIG_NET_POLL_CONTROLLER
1873/**
1874 * vxge_netpoll - netpoll event handler entry point
1875 * @dev : pointer to the device structure.
1876 * Description:
1877 *      This function will be called by upper layer to check for events on the
1878 * interface in situations where interrupts are disabled. It is used for
1879 * specific in-kernel networking tasks, such as remote consoles and kernel
1880 * debugging over the network (example netdump in RedHat).
1881 */
1882static void vxge_netpoll(struct net_device *dev)
1883{
1884	struct vxgedev *vdev = netdev_priv(dev);
1885	struct pci_dev *pdev = vdev->pdev;
1886	struct __vxge_hw_device *hldev = pci_get_drvdata(pdev);
1887	const int irq = pdev->irq;
1888
1889	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1890
1891	if (pci_channel_offline(pdev))
1892		return;
1893
1894	disable_irq(irq);
1895	vxge_hw_device_clear_tx_rx(hldev);
1896
1897	vxge_hw_device_clear_tx_rx(hldev);
1898	VXGE_COMPLETE_ALL_RX(vdev);
1899	VXGE_COMPLETE_ALL_TX(vdev);
1900
1901	enable_irq(irq);
1902
1903	vxge_debug_entryexit(VXGE_TRACE,
1904		"%s:%d  Exiting...", __func__, __LINE__);
1905}
1906#endif
1907
1908/* RTH configuration */
1909static enum vxge_hw_status vxge_rth_configure(struct vxgedev *vdev)
1910{
1911	enum vxge_hw_status status = VXGE_HW_OK;
1912	struct vxge_hw_rth_hash_types hash_types;
1913	u8 itable[256] = {0}; /* indirection table */
1914	u8 mtable[256] = {0}; /* CPU to vpath mapping  */
1915	int index;
1916
1917	/*
1918	 * Filling
1919	 * 	- itable with bucket numbers
1920	 * 	- mtable with bucket-to-vpath mapping
1921	 */
1922	for (index = 0; index < (1 << vdev->config.rth_bkt_sz); index++) {
1923		itable[index] = index;
1924		mtable[index] = index % vdev->no_of_vpath;
1925	}
1926
1927	/* set indirection table, bucket-to-vpath mapping */
1928	status = vxge_hw_vpath_rts_rth_itable_set(vdev->vp_handles,
1929						vdev->no_of_vpath,
1930						mtable, itable,
1931						vdev->config.rth_bkt_sz);
1932	if (status != VXGE_HW_OK) {
1933		vxge_debug_init(VXGE_ERR,
1934			"RTH indirection table configuration failed "
1935			"for vpath:%d", vdev->vpaths[0].device_id);
1936		return status;
1937	}
1938
1939	/* Fill RTH hash types */
1940	hash_types.hash_type_tcpipv4_en   = vdev->config.rth_hash_type_tcpipv4;
1941	hash_types.hash_type_ipv4_en      = vdev->config.rth_hash_type_ipv4;
1942	hash_types.hash_type_tcpipv6_en   = vdev->config.rth_hash_type_tcpipv6;
1943	hash_types.hash_type_ipv6_en      = vdev->config.rth_hash_type_ipv6;
1944	hash_types.hash_type_tcpipv6ex_en =
1945					vdev->config.rth_hash_type_tcpipv6ex;
1946	hash_types.hash_type_ipv6ex_en    = vdev->config.rth_hash_type_ipv6ex;
1947
1948	/*
1949	 * Because the itable_set() method uses the active_table field
1950	 * for the target virtual path the RTH config should be updated
1951	 * for all VPATHs. The h/w only uses the lowest numbered VPATH
1952	 * when steering frames.
1953	 */
1954	for (index = 0; index < vdev->no_of_vpath; index++) {
1955		status = vxge_hw_vpath_rts_rth_set(
1956				vdev->vpaths[index].handle,
1957				vdev->config.rth_algorithm,
1958				&hash_types,
1959				vdev->config.rth_bkt_sz);
1960		if (status != VXGE_HW_OK) {
1961			vxge_debug_init(VXGE_ERR,
1962				"RTH configuration failed for vpath:%d",
1963				vdev->vpaths[index].device_id);
1964			return status;
1965		}
1966	}
1967
1968	return status;
1969}
1970
1971/* reset vpaths */
1972static enum vxge_hw_status vxge_reset_all_vpaths(struct vxgedev *vdev)
1973{
1974	enum vxge_hw_status status = VXGE_HW_OK;
1975	struct vxge_vpath *vpath;
1976	int i;
1977
1978	for (i = 0; i < vdev->no_of_vpath; i++) {
1979		vpath = &vdev->vpaths[i];
1980		if (vpath->handle) {
1981			if (vxge_hw_vpath_reset(vpath->handle) == VXGE_HW_OK) {
1982				if (is_vxge_card_up(vdev) &&
1983					vxge_hw_vpath_recover_from_reset(
1984						vpath->handle) != VXGE_HW_OK) {
1985					vxge_debug_init(VXGE_ERR,
1986						"vxge_hw_vpath_recover_"
1987						"from_reset failed for vpath: "
1988						"%d", i);
1989					return status;
1990				}
1991			} else {
1992				vxge_debug_init(VXGE_ERR,
1993					"vxge_hw_vpath_reset failed for "
1994					"vpath:%d", i);
1995				return status;
1996			}
1997		}
1998	}
1999
2000	return status;
2001}
2002
2003/* close vpaths */
2004static void vxge_close_vpaths(struct vxgedev *vdev, int index)
2005{
2006	struct vxge_vpath *vpath;
2007	int i;
2008
2009	for (i = index; i < vdev->no_of_vpath; i++) {
2010		vpath = &vdev->vpaths[i];
2011
2012		if (vpath->handle && vpath->is_open) {
2013			vxge_hw_vpath_close(vpath->handle);
2014			vdev->stats.vpaths_open--;
2015		}
2016		vpath->is_open = 0;
2017		vpath->handle = NULL;
2018	}
2019}
2020
2021/* open vpaths */
2022static int vxge_open_vpaths(struct vxgedev *vdev)
2023{
2024	struct vxge_hw_vpath_attr attr;
2025	enum vxge_hw_status status;
2026	struct vxge_vpath *vpath;
2027	u32 vp_id = 0;
2028	int i;
2029
2030	for (i = 0; i < vdev->no_of_vpath; i++) {
2031		vpath = &vdev->vpaths[i];
2032		vxge_assert(vpath->is_configured);
2033
2034		if (!vdev->titan1) {
2035			struct vxge_hw_vp_config *vcfg;
2036			vcfg = &vdev->devh->config.vp_config[vpath->device_id];
2037
2038			vcfg->rti.urange_a = RTI_T1A_RX_URANGE_A;
2039			vcfg->rti.urange_b = RTI_T1A_RX_URANGE_B;
2040			vcfg->rti.urange_c = RTI_T1A_RX_URANGE_C;
2041			vcfg->tti.uec_a = TTI_T1A_TX_UFC_A;
2042			vcfg->tti.uec_b = TTI_T1A_TX_UFC_B;
2043			vcfg->tti.uec_c = TTI_T1A_TX_UFC_C(vdev->mtu);
2044			vcfg->tti.uec_d = TTI_T1A_TX_UFC_D(vdev->mtu);
2045			vcfg->tti.ltimer_val = VXGE_T1A_TTI_LTIMER_VAL;
2046			vcfg->tti.rtimer_val = VXGE_T1A_TTI_RTIMER_VAL;
2047		}
2048
2049		attr.vp_id = vpath->device_id;
2050		attr.fifo_attr.callback = vxge_xmit_compl;
2051		attr.fifo_attr.txdl_term = vxge_tx_term;
2052		attr.fifo_attr.per_txdl_space = sizeof(struct vxge_tx_priv);
2053		attr.fifo_attr.userdata = &vpath->fifo;
2054
2055		attr.ring_attr.callback = vxge_rx_1b_compl;
2056		attr.ring_attr.rxd_init = vxge_rx_initial_replenish;
2057		attr.ring_attr.rxd_term = vxge_rx_term;
2058		attr.ring_attr.per_rxd_space = sizeof(struct vxge_rx_priv);
2059		attr.ring_attr.userdata = &vpath->ring;
2060
2061		vpath->ring.ndev = vdev->ndev;
2062		vpath->ring.pdev = vdev->pdev;
2063
2064		status = vxge_hw_vpath_open(vdev->devh, &attr, &vpath->handle);
2065		if (status == VXGE_HW_OK) {
2066			vpath->fifo.handle =
2067			    (struct __vxge_hw_fifo *)attr.fifo_attr.userdata;
2068			vpath->ring.handle =
2069			    (struct __vxge_hw_ring *)attr.ring_attr.userdata;
2070			vpath->fifo.tx_steering_type =
2071				vdev->config.tx_steering_type;
2072			vpath->fifo.ndev = vdev->ndev;
2073			vpath->fifo.pdev = vdev->pdev;
2074
2075			u64_stats_init(&vpath->fifo.stats.syncp);
2076			u64_stats_init(&vpath->ring.stats.syncp);
2077
2078			if (vdev->config.tx_steering_type)
2079				vpath->fifo.txq =
2080					netdev_get_tx_queue(vdev->ndev, i);
2081			else
2082				vpath->fifo.txq =
2083					netdev_get_tx_queue(vdev->ndev, 0);
2084			vpath->fifo.indicate_max_pkts =
2085				vdev->config.fifo_indicate_max_pkts;
2086			vpath->fifo.tx_vector_no = 0;
2087			vpath->ring.rx_vector_no = 0;
2088			vpath->ring.rx_hwts = vdev->rx_hwts;
2089			vpath->is_open = 1;
2090			vdev->vp_handles[i] = vpath->handle;
2091			vpath->ring.vlan_tag_strip = vdev->vlan_tag_strip;
2092			vdev->stats.vpaths_open++;
2093		} else {
2094			vdev->stats.vpath_open_fail++;
2095			vxge_debug_init(VXGE_ERR, "%s: vpath: %d failed to "
2096					"open with status: %d",
2097					vdev->ndev->name, vpath->device_id,
2098					status);
2099			vxge_close_vpaths(vdev, 0);
2100			return -EPERM;
2101		}
2102
2103		vp_id = vpath->handle->vpath->vp_id;
2104		vdev->vpaths_deployed |= vxge_mBIT(vp_id);
2105	}
2106
2107	return VXGE_HW_OK;
2108}
2109
2110/**
2111 *  adaptive_coalesce_tx_interrupts - Changes the interrupt coalescing
2112 *  if the interrupts are not within a range
2113 *  @fifo: pointer to transmit fifo structure
2114 *  Description: The function changes boundary timer and restriction timer
2115 *  value depends on the traffic
2116 *  Return Value: None
2117 */
2118static void adaptive_coalesce_tx_interrupts(struct vxge_fifo *fifo)
2119{
2120	fifo->interrupt_count++;
2121	if (time_before(fifo->jiffies + HZ / 100, jiffies)) {
2122		struct __vxge_hw_fifo *hw_fifo = fifo->handle;
2123
2124		fifo->jiffies = jiffies;
2125		if (fifo->interrupt_count > VXGE_T1A_MAX_TX_INTERRUPT_COUNT &&
2126		    hw_fifo->rtimer != VXGE_TTI_RTIMER_ADAPT_VAL) {
2127			hw_fifo->rtimer = VXGE_TTI_RTIMER_ADAPT_VAL;
2128			vxge_hw_vpath_dynamic_tti_rtimer_set(hw_fifo);
2129		} else if (hw_fifo->rtimer != 0) {
2130			hw_fifo->rtimer = 0;
2131			vxge_hw_vpath_dynamic_tti_rtimer_set(hw_fifo);
2132		}
2133		fifo->interrupt_count = 0;
2134	}
2135}
2136
2137/**
2138 *  adaptive_coalesce_rx_interrupts - Changes the interrupt coalescing
2139 *  if the interrupts are not within a range
2140 *  @ring: pointer to receive ring structure
2141 *  Description: The function increases of decreases the packet counts within
2142 *  the ranges of traffic utilization, if the interrupts due to this ring are
2143 *  not within a fixed range.
2144 *  Return Value: Nothing
2145 */
2146static void adaptive_coalesce_rx_interrupts(struct vxge_ring *ring)
2147{
2148	ring->interrupt_count++;
2149	if (time_before(ring->jiffies + HZ / 100, jiffies)) {
2150		struct __vxge_hw_ring *hw_ring = ring->handle;
2151
2152		ring->jiffies = jiffies;
2153		if (ring->interrupt_count > VXGE_T1A_MAX_INTERRUPT_COUNT &&
2154		    hw_ring->rtimer != VXGE_RTI_RTIMER_ADAPT_VAL) {
2155			hw_ring->rtimer = VXGE_RTI_RTIMER_ADAPT_VAL;
2156			vxge_hw_vpath_dynamic_rti_rtimer_set(hw_ring);
2157		} else if (hw_ring->rtimer != 0) {
2158			hw_ring->rtimer = 0;
2159			vxge_hw_vpath_dynamic_rti_rtimer_set(hw_ring);
2160		}
2161		ring->interrupt_count = 0;
2162	}
2163}
2164
2165/*
2166 *  vxge_isr_napi
2167 *  @irq: the irq of the device.
2168 *  @dev_id: a void pointer to the hldev structure of the Titan device
2169 *  @ptregs: pointer to the registers pushed on the stack.
2170 *
2171 *  This function is the ISR handler of the device when napi is enabled. It
2172 *  identifies the reason for the interrupt and calls the relevant service
2173 *  routines.
2174 */
2175static irqreturn_t vxge_isr_napi(int irq, void *dev_id)
2176{
2177	struct __vxge_hw_device *hldev;
2178	u64 reason;
2179	enum vxge_hw_status status;
2180	struct vxgedev *vdev = (struct vxgedev *)dev_id;
2181
2182	vxge_debug_intr(VXGE_TRACE, "%s:%d", __func__, __LINE__);
2183
2184	hldev = pci_get_drvdata(vdev->pdev);
2185
2186	if (pci_channel_offline(vdev->pdev))
2187		return IRQ_NONE;
2188
2189	if (unlikely(!is_vxge_card_up(vdev)))
2190		return IRQ_HANDLED;
2191
2192	status = vxge_hw_device_begin_irq(hldev, vdev->exec_mode, &reason);
2193	if (status == VXGE_HW_OK) {
2194		vxge_hw_device_mask_all(hldev);
2195
2196		if (reason &
2197			VXGE_HW_TITAN_GENERAL_INT_STATUS_VPATH_TRAFFIC_INT(
2198			vdev->vpaths_deployed >>
2199			(64 - VXGE_HW_MAX_VIRTUAL_PATHS))) {
2200
2201			vxge_hw_device_clear_tx_rx(hldev);
2202			napi_schedule(&vdev->napi);
2203			vxge_debug_intr(VXGE_TRACE,
2204				"%s:%d  Exiting...", __func__, __LINE__);
2205			return IRQ_HANDLED;
2206		} else
2207			vxge_hw_device_unmask_all(hldev);
2208	} else if (unlikely((status == VXGE_HW_ERR_VPATH) ||
2209		(status == VXGE_HW_ERR_CRITICAL) ||
2210		(status == VXGE_HW_ERR_FIFO))) {
2211		vxge_hw_device_mask_all(hldev);
2212		vxge_hw_device_flush_io(hldev);
2213		return IRQ_HANDLED;
2214	} else if (unlikely(status == VXGE_HW_ERR_SLOT_FREEZE))
2215		return IRQ_HANDLED;
2216
2217	vxge_debug_intr(VXGE_TRACE, "%s:%d  Exiting...", __func__, __LINE__);
2218	return IRQ_NONE;
2219}
2220
2221static irqreturn_t vxge_tx_msix_handle(int irq, void *dev_id)
2222{
2223	struct vxge_fifo *fifo = (struct vxge_fifo *)dev_id;
2224
2225	adaptive_coalesce_tx_interrupts(fifo);
2226
2227	vxge_hw_channel_msix_mask((struct __vxge_hw_channel *)fifo->handle,
2228				  fifo->tx_vector_no);
2229
2230	vxge_hw_channel_msix_clear((struct __vxge_hw_channel *)fifo->handle,
2231				   fifo->tx_vector_no);
2232
2233	VXGE_COMPLETE_VPATH_TX(fifo);
2234
2235	vxge_hw_channel_msix_unmask((struct __vxge_hw_channel *)fifo->handle,
2236				    fifo->tx_vector_no);
2237
2238	return IRQ_HANDLED;
2239}
2240
2241static irqreturn_t vxge_rx_msix_napi_handle(int irq, void *dev_id)
2242{
2243	struct vxge_ring *ring = (struct vxge_ring *)dev_id;
2244
2245	adaptive_coalesce_rx_interrupts(ring);
2246
2247	vxge_hw_channel_msix_mask((struct __vxge_hw_channel *)ring->handle,
2248				  ring->rx_vector_no);
2249
2250	vxge_hw_channel_msix_clear((struct __vxge_hw_channel *)ring->handle,
2251				   ring->rx_vector_no);
2252
2253	napi_schedule(&ring->napi);
2254	return IRQ_HANDLED;
2255}
2256
2257static irqreturn_t
2258vxge_alarm_msix_handle(int irq, void *dev_id)
2259{
2260	int i;
2261	enum vxge_hw_status status;
2262	struct vxge_vpath *vpath = (struct vxge_vpath *)dev_id;
2263	struct vxgedev *vdev = vpath->vdev;
2264	int msix_id = (vpath->handle->vpath->vp_id *
2265		VXGE_HW_VPATH_MSIX_ACTIVE) + VXGE_ALARM_MSIX_ID;
2266
2267	for (i = 0; i < vdev->no_of_vpath; i++) {
2268		/* Reduce the chance of losing alarm interrupts by masking
2269		 * the vector. A pending bit will be set if an alarm is
2270		 * generated and on unmask the interrupt will be fired.
2271		 */
2272		vxge_hw_vpath_msix_mask(vdev->vpaths[i].handle, msix_id);
2273		vxge_hw_vpath_msix_clear(vdev->vpaths[i].handle, msix_id);
2274
2275		status = vxge_hw_vpath_alarm_process(vdev->vpaths[i].handle,
2276			vdev->exec_mode);
2277		if (status == VXGE_HW_OK) {
2278			vxge_hw_vpath_msix_unmask(vdev->vpaths[i].handle,
2279						  msix_id);
2280			continue;
2281		}
2282		vxge_debug_intr(VXGE_ERR,
2283			"%s: vxge_hw_vpath_alarm_process failed %x ",
2284			VXGE_DRIVER_NAME, status);
2285	}
2286	return IRQ_HANDLED;
2287}
2288
2289static int vxge_alloc_msix(struct vxgedev *vdev)
2290{
2291	int j, i, ret = 0;
2292	int msix_intr_vect = 0, temp;
2293	vdev->intr_cnt = 0;
2294
2295start:
2296	/* Tx/Rx MSIX Vectors count */
2297	vdev->intr_cnt = vdev->no_of_vpath * 2;
2298
2299	/* Alarm MSIX Vectors count */
2300	vdev->intr_cnt++;
2301
2302	vdev->entries = kcalloc(vdev->intr_cnt, sizeof(struct msix_entry),
2303				GFP_KERNEL);
2304	if (!vdev->entries) {
2305		vxge_debug_init(VXGE_ERR,
2306			"%s: memory allocation failed",
2307			VXGE_DRIVER_NAME);
2308		ret = -ENOMEM;
2309		goto alloc_entries_failed;
2310	}
2311
2312	vdev->vxge_entries = kcalloc(vdev->intr_cnt,
2313				     sizeof(struct vxge_msix_entry),
2314				     GFP_KERNEL);
2315	if (!vdev->vxge_entries) {
2316		vxge_debug_init(VXGE_ERR, "%s: memory allocation failed",
2317			VXGE_DRIVER_NAME);
2318		ret = -ENOMEM;
2319		goto alloc_vxge_entries_failed;
2320	}
2321
2322	for (i = 0, j = 0; i < vdev->no_of_vpath; i++) {
2323
2324		msix_intr_vect = i * VXGE_HW_VPATH_MSIX_ACTIVE;
2325
2326		/* Initialize the fifo vector */
2327		vdev->entries[j].entry = msix_intr_vect;
2328		vdev->vxge_entries[j].entry = msix_intr_vect;
2329		vdev->vxge_entries[j].in_use = 0;
2330		j++;
2331
2332		/* Initialize the ring vector */
2333		vdev->entries[j].entry = msix_intr_vect + 1;
2334		vdev->vxge_entries[j].entry = msix_intr_vect + 1;
2335		vdev->vxge_entries[j].in_use = 0;
2336		j++;
2337	}
2338
2339	/* Initialize the alarm vector */
2340	vdev->entries[j].entry = VXGE_ALARM_MSIX_ID;
2341	vdev->vxge_entries[j].entry = VXGE_ALARM_MSIX_ID;
2342	vdev->vxge_entries[j].in_use = 0;
2343
2344	ret = pci_enable_msix_range(vdev->pdev,
2345				    vdev->entries, 3, vdev->intr_cnt);
2346	if (ret < 0) {
2347		ret = -ENODEV;
2348		goto enable_msix_failed;
2349	} else if (ret < vdev->intr_cnt) {
2350		pci_disable_msix(vdev->pdev);
2351
2352		vxge_debug_init(VXGE_ERR,
2353			"%s: MSI-X enable failed for %d vectors, ret: %d",
2354			VXGE_DRIVER_NAME, vdev->intr_cnt, ret);
2355		if (max_config_vpath != VXGE_USE_DEFAULT) {
2356			ret = -ENODEV;
2357			goto enable_msix_failed;
2358		}
2359
2360		kfree(vdev->entries);
2361		kfree(vdev->vxge_entries);
2362		vdev->entries = NULL;
2363		vdev->vxge_entries = NULL;
2364		/* Try with less no of vector by reducing no of vpaths count */
2365		temp = (ret - 1)/2;
2366		vxge_close_vpaths(vdev, temp);
2367		vdev->no_of_vpath = temp;
2368		goto start;
2369	}
2370	return 0;
2371
2372enable_msix_failed:
2373	kfree(vdev->vxge_entries);
2374alloc_vxge_entries_failed:
2375	kfree(vdev->entries);
2376alloc_entries_failed:
2377	return ret;
2378}
2379
2380static int vxge_enable_msix(struct vxgedev *vdev)
2381{
2382
2383	int i, ret = 0;
2384	/* 0 - Tx, 1 - Rx  */
2385	int tim_msix_id[4] = {0, 1, 0, 0};
2386
2387	vdev->intr_cnt = 0;
2388
2389	/* allocate msix vectors */
2390	ret = vxge_alloc_msix(vdev);
2391	if (!ret) {
2392		for (i = 0; i < vdev->no_of_vpath; i++) {
2393			struct vxge_vpath *vpath = &vdev->vpaths[i];
2394
2395			/* If fifo or ring are not enabled, the MSIX vector for
2396			 * it should be set to 0.
2397			 */
2398			vpath->ring.rx_vector_no = (vpath->device_id *
2399						VXGE_HW_VPATH_MSIX_ACTIVE) + 1;
2400
2401			vpath->fifo.tx_vector_no = (vpath->device_id *
2402						VXGE_HW_VPATH_MSIX_ACTIVE);
2403
2404			vxge_hw_vpath_msix_set(vpath->handle, tim_msix_id,
2405					       VXGE_ALARM_MSIX_ID);
2406		}
2407	}
2408
2409	return ret;
2410}
2411
2412static void vxge_rem_msix_isr(struct vxgedev *vdev)
2413{
2414	int intr_cnt;
2415
2416	for (intr_cnt = 0; intr_cnt < (vdev->no_of_vpath * 2 + 1);
2417		intr_cnt++) {
2418		if (vdev->vxge_entries[intr_cnt].in_use) {
2419			synchronize_irq(vdev->entries[intr_cnt].vector);
2420			free_irq(vdev->entries[intr_cnt].vector,
2421				vdev->vxge_entries[intr_cnt].arg);
2422			vdev->vxge_entries[intr_cnt].in_use = 0;
2423		}
2424	}
2425
2426	kfree(vdev->entries);
2427	kfree(vdev->vxge_entries);
2428	vdev->entries = NULL;
2429	vdev->vxge_entries = NULL;
2430
2431	if (vdev->config.intr_type == MSI_X)
2432		pci_disable_msix(vdev->pdev);
2433}
2434
2435static void vxge_rem_isr(struct vxgedev *vdev)
2436{
2437	if (IS_ENABLED(CONFIG_PCI_MSI) &&
2438	    vdev->config.intr_type == MSI_X) {
2439		vxge_rem_msix_isr(vdev);
2440	} else if (vdev->config.intr_type == INTA) {
2441			synchronize_irq(vdev->pdev->irq);
2442			free_irq(vdev->pdev->irq, vdev);
2443	}
2444}
2445
2446static int vxge_add_isr(struct vxgedev *vdev)
2447{
2448	int ret = 0;
2449	int vp_idx = 0, intr_idx = 0, intr_cnt = 0, msix_idx = 0, irq_req = 0;
2450	int pci_fun = PCI_FUNC(vdev->pdev->devfn);
2451
2452	if (IS_ENABLED(CONFIG_PCI_MSI) && vdev->config.intr_type == MSI_X)
2453		ret = vxge_enable_msix(vdev);
2454
2455	if (ret) {
2456		vxge_debug_init(VXGE_ERR,
2457			"%s: Enabling MSI-X Failed", VXGE_DRIVER_NAME);
2458		vxge_debug_init(VXGE_ERR,
2459			"%s: Defaulting to INTA", VXGE_DRIVER_NAME);
2460		vdev->config.intr_type = INTA;
2461	}
2462
2463	if (IS_ENABLED(CONFIG_PCI_MSI) && vdev->config.intr_type == MSI_X) {
2464		for (intr_idx = 0;
2465		     intr_idx < (vdev->no_of_vpath *
2466			VXGE_HW_VPATH_MSIX_ACTIVE); intr_idx++) {
2467
2468			msix_idx = intr_idx % VXGE_HW_VPATH_MSIX_ACTIVE;
2469			irq_req = 0;
2470
2471			switch (msix_idx) {
2472			case 0:
2473				snprintf(vdev->desc[intr_cnt], VXGE_INTR_STRLEN,
2474					"%s:vxge:MSI-X %d - Tx - fn:%d vpath:%d",
2475					vdev->ndev->name,
2476					vdev->entries[intr_cnt].entry,
2477					pci_fun, vp_idx);
2478				ret = request_irq(
2479					vdev->entries[intr_cnt].vector,
2480					vxge_tx_msix_handle, 0,
2481					vdev->desc[intr_cnt],
2482					&vdev->vpaths[vp_idx].fifo);
2483				vdev->vxge_entries[intr_cnt].arg =
2484						&vdev->vpaths[vp_idx].fifo;
2485				irq_req = 1;
2486				break;
2487			case 1:
2488				snprintf(vdev->desc[intr_cnt], VXGE_INTR_STRLEN,
2489					"%s:vxge:MSI-X %d - Rx - fn:%d vpath:%d",
2490					vdev->ndev->name,
2491					vdev->entries[intr_cnt].entry,
2492					pci_fun, vp_idx);
2493				ret = request_irq(
2494					vdev->entries[intr_cnt].vector,
2495					vxge_rx_msix_napi_handle, 0,
2496					vdev->desc[intr_cnt],
2497					&vdev->vpaths[vp_idx].ring);
2498				vdev->vxge_entries[intr_cnt].arg =
2499						&vdev->vpaths[vp_idx].ring;
2500				irq_req = 1;
2501				break;
2502			}
2503
2504			if (ret) {
2505				vxge_debug_init(VXGE_ERR,
2506					"%s: MSIX - %d  Registration failed",
2507					vdev->ndev->name, intr_cnt);
2508				vxge_rem_msix_isr(vdev);
2509				vdev->config.intr_type = INTA;
2510				vxge_debug_init(VXGE_ERR,
2511					"%s: Defaulting to INTA",
2512					vdev->ndev->name);
2513				goto INTA_MODE;
2514			}
2515
2516			if (irq_req) {
2517				/* We requested for this msix interrupt */
2518				vdev->vxge_entries[intr_cnt].in_use = 1;
2519				msix_idx +=  vdev->vpaths[vp_idx].device_id *
2520					VXGE_HW_VPATH_MSIX_ACTIVE;
2521				vxge_hw_vpath_msix_unmask(
2522					vdev->vpaths[vp_idx].handle,
2523					msix_idx);
2524				intr_cnt++;
2525			}
2526
2527			/* Point to next vpath handler */
2528			if (((intr_idx + 1) % VXGE_HW_VPATH_MSIX_ACTIVE == 0) &&
2529			    (vp_idx < (vdev->no_of_vpath - 1)))
2530				vp_idx++;
2531		}
2532
2533		intr_cnt = vdev->no_of_vpath * 2;
2534		snprintf(vdev->desc[intr_cnt], VXGE_INTR_STRLEN,
2535			"%s:vxge:MSI-X %d - Alarm - fn:%d",
2536			vdev->ndev->name,
2537			vdev->entries[intr_cnt].entry,
2538			pci_fun);
2539		/* For Alarm interrupts */
2540		ret = request_irq(vdev->entries[intr_cnt].vector,
2541					vxge_alarm_msix_handle, 0,
2542					vdev->desc[intr_cnt],
2543					&vdev->vpaths[0]);
2544		if (ret) {
2545			vxge_debug_init(VXGE_ERR,
2546				"%s: MSIX - %d Registration failed",
2547				vdev->ndev->name, intr_cnt);
2548			vxge_rem_msix_isr(vdev);
2549			vdev->config.intr_type = INTA;
2550			vxge_debug_init(VXGE_ERR,
2551				"%s: Defaulting to INTA",
2552				vdev->ndev->name);
2553			goto INTA_MODE;
2554		}
2555
2556		msix_idx = (vdev->vpaths[0].handle->vpath->vp_id *
2557			VXGE_HW_VPATH_MSIX_ACTIVE) + VXGE_ALARM_MSIX_ID;
2558		vxge_hw_vpath_msix_unmask(vdev->vpaths[vp_idx].handle,
2559					msix_idx);
2560		vdev->vxge_entries[intr_cnt].in_use = 1;
2561		vdev->vxge_entries[intr_cnt].arg = &vdev->vpaths[0];
2562	}
2563
2564INTA_MODE:
2565	if (vdev->config.intr_type == INTA) {
2566		snprintf(vdev->desc[0], VXGE_INTR_STRLEN,
2567			"%s:vxge:INTA", vdev->ndev->name);
2568		vxge_hw_device_set_intr_type(vdev->devh,
2569			VXGE_HW_INTR_MODE_IRQLINE);
2570
2571		vxge_hw_vpath_tti_ci_set(vdev->vpaths[0].fifo.handle);
2572
2573		ret = request_irq((int) vdev->pdev->irq,
2574			vxge_isr_napi,
2575			IRQF_SHARED, vdev->desc[0], vdev);
2576		if (ret) {
2577			vxge_debug_init(VXGE_ERR,
2578				"%s %s-%d: ISR registration failed",
2579				VXGE_DRIVER_NAME, "IRQ", vdev->pdev->irq);
2580			return -ENODEV;
2581		}
2582		vxge_debug_init(VXGE_TRACE,
2583			"new %s-%d line allocated",
2584			"IRQ", vdev->pdev->irq);
2585	}
2586
2587	return VXGE_HW_OK;
2588}
2589
2590static void vxge_poll_vp_reset(struct timer_list *t)
2591{
2592	struct vxgedev *vdev = from_timer(vdev, t, vp_reset_timer);
2593	int i, j = 0;
2594
2595	for (i = 0; i < vdev->no_of_vpath; i++) {
2596		if (test_bit(i, &vdev->vp_reset)) {
2597			vxge_reset_vpath(vdev, i);
2598			j++;
2599		}
2600	}
2601	if (j && (vdev->config.intr_type != MSI_X)) {
2602		vxge_hw_device_unmask_all(vdev->devh);
2603		vxge_hw_device_flush_io(vdev->devh);
2604	}
2605
2606	mod_timer(&vdev->vp_reset_timer, jiffies + HZ / 2);
2607}
2608
2609static void vxge_poll_vp_lockup(struct timer_list *t)
2610{
2611	struct vxgedev *vdev = from_timer(vdev, t, vp_lockup_timer);
2612	enum vxge_hw_status status = VXGE_HW_OK;
2613	struct vxge_vpath *vpath;
2614	struct vxge_ring *ring;
2615	int i;
2616	unsigned long rx_frms;
2617
2618	for (i = 0; i < vdev->no_of_vpath; i++) {
2619		ring = &vdev->vpaths[i].ring;
2620
2621		/* Truncated to machine word size number of frames */
2622		rx_frms = READ_ONCE(ring->stats.rx_frms);
2623
2624		/* Did this vpath received any packets */
2625		if (ring->stats.prev_rx_frms == rx_frms) {
2626			status = vxge_hw_vpath_check_leak(ring->handle);
2627
2628			/* Did it received any packets last time */
2629			if ((VXGE_HW_FAIL == status) &&
2630				(VXGE_HW_FAIL == ring->last_status)) {
2631
2632				/* schedule vpath reset */
2633				if (!test_and_set_bit(i, &vdev->vp_reset)) {
2634					vpath = &vdev->vpaths[i];
2635
2636					/* disable interrupts for this vpath */
2637					vxge_vpath_intr_disable(vdev, i);
2638
2639					/* stop the queue for this vpath */
2640					netif_tx_stop_queue(vpath->fifo.txq);
2641					continue;
2642				}
2643			}
2644		}
2645		ring->stats.prev_rx_frms = rx_frms;
2646		ring->last_status = status;
2647	}
2648
2649	/* Check every 1 milli second */
2650	mod_timer(&vdev->vp_lockup_timer, jiffies + HZ / 1000);
2651}
2652
2653static netdev_features_t vxge_fix_features(struct net_device *dev,
2654	netdev_features_t features)
2655{
2656	netdev_features_t changed = dev->features ^ features;
2657
2658	/* Enabling RTH requires some of the logic in vxge_device_register and a
2659	 * vpath reset.  Due to these restrictions, only allow modification
2660	 * while the interface is down.
2661	 */
2662	if ((changed & NETIF_F_RXHASH) && netif_running(dev))
2663		features ^= NETIF_F_RXHASH;
2664
2665	return features;
2666}
2667
2668static int vxge_set_features(struct net_device *dev, netdev_features_t features)
2669{
2670	struct vxgedev *vdev = netdev_priv(dev);
2671	netdev_features_t changed = dev->features ^ features;
2672
2673	if (!(changed & NETIF_F_RXHASH))
2674		return 0;
2675
2676	/* !netif_running() ensured by vxge_fix_features() */
2677
2678	vdev->devh->config.rth_en = !!(features & NETIF_F_RXHASH);
2679	if (vxge_reset_all_vpaths(vdev) != VXGE_HW_OK) {
2680		dev->features = features ^ NETIF_F_RXHASH;
2681		vdev->devh->config.rth_en = !!(dev->features & NETIF_F_RXHASH);
2682		return -EIO;
2683	}
2684
2685	return 0;
2686}
2687
2688/**
2689 * vxge_open
2690 * @dev: pointer to the device structure.
2691 *
2692 * This function is the open entry point of the driver. It mainly calls a
2693 * function to allocate Rx buffers and inserts them into the buffer
2694 * descriptors and then enables the Rx part of the NIC.
2695 * Return value: '0' on success and an appropriate (-)ve integer as
2696 * defined in errno.h file on failure.
2697 */
2698static int vxge_open(struct net_device *dev)
2699{
2700	enum vxge_hw_status status;
2701	struct vxgedev *vdev;
2702	struct __vxge_hw_device *hldev;
2703	struct vxge_vpath *vpath;
2704	int ret = 0;
2705	int i;
2706	u64 val64;
2707
2708	vxge_debug_entryexit(VXGE_TRACE,
2709		"%s: %s:%d", dev->name, __func__, __LINE__);
2710
2711	vdev = netdev_priv(dev);
2712	hldev = pci_get_drvdata(vdev->pdev);
2713
2714	/* make sure you have link off by default every time Nic is
2715	 * initialized */
2716	netif_carrier_off(dev);
2717
2718	/* Open VPATHs */
2719	status = vxge_open_vpaths(vdev);
2720	if (status != VXGE_HW_OK) {
2721		vxge_debug_init(VXGE_ERR,
2722			"%s: fatal: Vpath open failed", vdev->ndev->name);
2723		ret = -EPERM;
2724		goto out0;
2725	}
2726
2727	vdev->mtu = dev->mtu;
2728
2729	status = vxge_add_isr(vdev);
2730	if (status != VXGE_HW_OK) {
2731		vxge_debug_init(VXGE_ERR,
2732			"%s: fatal: ISR add failed", dev->name);
2733		ret = -EPERM;
2734		goto out1;
2735	}
2736
2737	if (vdev->config.intr_type != MSI_X) {
2738		netif_napi_add(dev, &vdev->napi, vxge_poll_inta,
2739			vdev->config.napi_weight);
2740		napi_enable(&vdev->napi);
2741		for (i = 0; i < vdev->no_of_vpath; i++) {
2742			vpath = &vdev->vpaths[i];
2743			vpath->ring.napi_p = &vdev->napi;
2744		}
2745	} else {
2746		for (i = 0; i < vdev->no_of_vpath; i++) {
2747			vpath = &vdev->vpaths[i];
2748			netif_napi_add(dev, &vpath->ring.napi,
2749			    vxge_poll_msix, vdev->config.napi_weight);
2750			napi_enable(&vpath->ring.napi);
2751			vpath->ring.napi_p = &vpath->ring.napi;
2752		}
2753	}
2754
2755	/* configure RTH */
2756	if (vdev->config.rth_steering) {
2757		status = vxge_rth_configure(vdev);
2758		if (status != VXGE_HW_OK) {
2759			vxge_debug_init(VXGE_ERR,
2760				"%s: fatal: RTH configuration failed",
2761				dev->name);
2762			ret = -EPERM;
2763			goto out2;
2764		}
2765	}
2766	printk(KERN_INFO "%s: Receive Hashing Offload %s\n", dev->name,
2767	       hldev->config.rth_en ? "enabled" : "disabled");
2768
2769	for (i = 0; i < vdev->no_of_vpath; i++) {
2770		vpath = &vdev->vpaths[i];
2771
2772		/* set initial mtu before enabling the device */
2773		status = vxge_hw_vpath_mtu_set(vpath->handle, vdev->mtu);
2774		if (status != VXGE_HW_OK) {
2775			vxge_debug_init(VXGE_ERR,
2776				"%s: fatal: can not set new MTU", dev->name);
2777			ret = -EPERM;
2778			goto out2;
2779		}
2780	}
2781
2782	VXGE_DEVICE_DEBUG_LEVEL_SET(VXGE_TRACE, VXGE_COMPONENT_LL, vdev);
2783	vxge_debug_init(vdev->level_trace,
2784		"%s: MTU is %d", vdev->ndev->name, vdev->mtu);
2785	VXGE_DEVICE_DEBUG_LEVEL_SET(VXGE_ERR, VXGE_COMPONENT_LL, vdev);
2786
2787	/* Restore the DA, VID table and also multicast and promiscuous mode
2788	 * states
2789	 */
2790	if (vdev->all_multi_flg) {
2791		for (i = 0; i < vdev->no_of_vpath; i++) {
2792			vpath = &vdev->vpaths[i];
2793			vxge_restore_vpath_mac_addr(vpath);
2794			vxge_restore_vpath_vid_table(vpath);
2795
2796			status = vxge_hw_vpath_mcast_enable(vpath->handle);
2797			if (status != VXGE_HW_OK)
2798				vxge_debug_init(VXGE_ERR,
2799					"%s:%d Enabling multicast failed",
2800					__func__, __LINE__);
2801		}
2802	}
2803
2804	/* Enable vpath to sniff all unicast/multicast traffic that not
2805	 * addressed to them. We allow promiscuous mode for PF only
2806	 */
2807
2808	val64 = 0;
2809	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
2810		val64 |= VXGE_HW_RXMAC_AUTHORIZE_ALL_ADDR_VP(i);
2811
2812	vxge_hw_mgmt_reg_write(vdev->devh,
2813		vxge_hw_mgmt_reg_type_mrpcim,
2814		0,
2815		(ulong)offsetof(struct vxge_hw_mrpcim_reg,
2816			rxmac_authorize_all_addr),
2817		val64);
2818
2819	vxge_hw_mgmt_reg_write(vdev->devh,
2820		vxge_hw_mgmt_reg_type_mrpcim,
2821		0,
2822		(ulong)offsetof(struct vxge_hw_mrpcim_reg,
2823			rxmac_authorize_all_vid),
2824		val64);
2825
2826	vxge_set_multicast(dev);
2827
2828	/* Enabling Bcast and mcast for all vpath */
2829	for (i = 0; i < vdev->no_of_vpath; i++) {
2830		vpath = &vdev->vpaths[i];
2831		status = vxge_hw_vpath_bcast_enable(vpath->handle);
2832		if (status != VXGE_HW_OK)
2833			vxge_debug_init(VXGE_ERR,
2834				"%s : Can not enable bcast for vpath "
2835				"id %d", dev->name, i);
2836		if (vdev->config.addr_learn_en) {
2837			status = vxge_hw_vpath_mcast_enable(vpath->handle);
2838			if (status != VXGE_HW_OK)
2839				vxge_debug_init(VXGE_ERR,
2840					"%s : Can not enable mcast for vpath "
2841					"id %d", dev->name, i);
2842		}
2843	}
2844
2845	vxge_hw_device_setpause_data(vdev->devh, 0,
2846		vdev->config.tx_pause_enable,
2847		vdev->config.rx_pause_enable);
2848
2849	if (vdev->vp_reset_timer.function == NULL)
2850		vxge_os_timer(&vdev->vp_reset_timer, vxge_poll_vp_reset,
2851			      HZ / 2);
2852
2853	/* There is no need to check for RxD leak and RxD lookup on Titan1A */
2854	if (vdev->titan1 && vdev->vp_lockup_timer.function == NULL)
2855		vxge_os_timer(&vdev->vp_lockup_timer, vxge_poll_vp_lockup,
2856			      HZ / 2);
2857
2858	set_bit(__VXGE_STATE_CARD_UP, &vdev->state);
2859
2860	smp_wmb();
2861
2862	if (vxge_hw_device_link_state_get(vdev->devh) == VXGE_HW_LINK_UP) {
2863		netif_carrier_on(vdev->ndev);
2864		netdev_notice(vdev->ndev, "Link Up\n");
2865		vdev->stats.link_up++;
2866	}
2867
2868	vxge_hw_device_intr_enable(vdev->devh);
2869
2870	smp_wmb();
2871
2872	for (i = 0; i < vdev->no_of_vpath; i++) {
2873		vpath = &vdev->vpaths[i];
2874
2875		vxge_hw_vpath_enable(vpath->handle);
2876		smp_wmb();
2877		vxge_hw_vpath_rx_doorbell_init(vpath->handle);
2878	}
2879
2880	netif_tx_start_all_queues(vdev->ndev);
2881
2882	/* configure CI */
2883	vxge_config_ci_for_tti_rti(vdev);
2884
2885	goto out0;
2886
2887out2:
2888	vxge_rem_isr(vdev);
2889
2890	/* Disable napi */
2891	if (vdev->config.intr_type != MSI_X)
2892		napi_disable(&vdev->napi);
2893	else {
2894		for (i = 0; i < vdev->no_of_vpath; i++)
2895			napi_disable(&vdev->vpaths[i].ring.napi);
2896	}
2897
2898out1:
2899	vxge_close_vpaths(vdev, 0);
2900out0:
2901	vxge_debug_entryexit(VXGE_TRACE,
2902				"%s: %s:%d  Exiting...",
2903				dev->name, __func__, __LINE__);
2904	return ret;
2905}
2906
2907/* Loop through the mac address list and delete all the entries */
2908static void vxge_free_mac_add_list(struct vxge_vpath *vpath)
2909{
2910
2911	struct list_head *entry, *next;
2912	if (list_empty(&vpath->mac_addr_list))
2913		return;
2914
2915	list_for_each_safe(entry, next, &vpath->mac_addr_list) {
2916		list_del(entry);
2917		kfree(entry);
2918	}
2919}
2920
2921static void vxge_napi_del_all(struct vxgedev *vdev)
2922{
2923	int i;
2924	if (vdev->config.intr_type != MSI_X)
2925		netif_napi_del(&vdev->napi);
2926	else {
2927		for (i = 0; i < vdev->no_of_vpath; i++)
2928			netif_napi_del(&vdev->vpaths[i].ring.napi);
2929	}
2930}
2931
2932static int do_vxge_close(struct net_device *dev, int do_io)
2933{
2934	enum vxge_hw_status status;
2935	struct vxgedev *vdev;
2936	struct __vxge_hw_device *hldev;
2937	int i;
2938	u64 val64, vpath_vector;
2939	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
2940		dev->name, __func__, __LINE__);
2941
2942	vdev = netdev_priv(dev);
2943	hldev = pci_get_drvdata(vdev->pdev);
2944
2945	if (unlikely(!is_vxge_card_up(vdev)))
2946		return 0;
2947
2948	/* If vxge_handle_crit_err task is executing,
2949	 * wait till it completes. */
2950	while (test_and_set_bit(__VXGE_STATE_RESET_CARD, &vdev->state))
2951		msleep(50);
2952
2953	if (do_io) {
2954		/* Put the vpath back in normal mode */
2955		vpath_vector = vxge_mBIT(vdev->vpaths[0].device_id);
2956		status = vxge_hw_mgmt_reg_read(vdev->devh,
2957				vxge_hw_mgmt_reg_type_mrpcim,
2958				0,
2959				(ulong)offsetof(
2960					struct vxge_hw_mrpcim_reg,
2961					rts_mgr_cbasin_cfg),
2962				&val64);
2963		if (status == VXGE_HW_OK) {
2964			val64 &= ~vpath_vector;
2965			status = vxge_hw_mgmt_reg_write(vdev->devh,
2966					vxge_hw_mgmt_reg_type_mrpcim,
2967					0,
2968					(ulong)offsetof(
2969						struct vxge_hw_mrpcim_reg,
2970						rts_mgr_cbasin_cfg),
2971					val64);
2972		}
2973
2974		/* Remove the function 0 from promiscuous mode */
2975		vxge_hw_mgmt_reg_write(vdev->devh,
2976			vxge_hw_mgmt_reg_type_mrpcim,
2977			0,
2978			(ulong)offsetof(struct vxge_hw_mrpcim_reg,
2979				rxmac_authorize_all_addr),
2980			0);
2981
2982		vxge_hw_mgmt_reg_write(vdev->devh,
2983			vxge_hw_mgmt_reg_type_mrpcim,
2984			0,
2985			(ulong)offsetof(struct vxge_hw_mrpcim_reg,
2986				rxmac_authorize_all_vid),
2987			0);
2988
2989		smp_wmb();
2990	}
2991
2992	if (vdev->titan1)
2993		del_timer_sync(&vdev->vp_lockup_timer);
2994
2995	del_timer_sync(&vdev->vp_reset_timer);
2996
2997	if (do_io)
2998		vxge_hw_device_wait_receive_idle(hldev);
2999
3000	clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3001
3002	/* Disable napi */
3003	if (vdev->config.intr_type != MSI_X)
3004		napi_disable(&vdev->napi);
3005	else {
3006		for (i = 0; i < vdev->no_of_vpath; i++)
3007			napi_disable(&vdev->vpaths[i].ring.napi);
3008	}
3009
3010	netif_carrier_off(vdev->ndev);
3011	netdev_notice(vdev->ndev, "Link Down\n");
3012	netif_tx_stop_all_queues(vdev->ndev);
3013
3014	/* Note that at this point xmit() is stopped by upper layer */
3015	if (do_io)
3016		vxge_hw_device_intr_disable(vdev->devh);
3017
3018	vxge_rem_isr(vdev);
3019
3020	vxge_napi_del_all(vdev);
3021
3022	if (do_io)
3023		vxge_reset_all_vpaths(vdev);
3024
3025	vxge_close_vpaths(vdev, 0);
3026
3027	vxge_debug_entryexit(VXGE_TRACE,
3028		"%s: %s:%d  Exiting...", dev->name, __func__, __LINE__);
3029
3030	clear_bit(__VXGE_STATE_RESET_CARD, &vdev->state);
3031
3032	return 0;
3033}
3034
3035/**
3036 * vxge_close
3037 * @dev: device pointer.
3038 *
3039 * This is the stop entry point of the driver. It needs to undo exactly
3040 * whatever was done by the open entry point, thus it's usually referred to
3041 * as the close function.Among other things this function mainly stops the
3042 * Rx side of the NIC and frees all the Rx buffers in the Rx rings.
3043 * Return value: '0' on success and an appropriate (-)ve integer as
3044 * defined in errno.h file on failure.
3045 */
3046static int vxge_close(struct net_device *dev)
3047{
3048	do_vxge_close(dev, 1);
3049	return 0;
3050}
3051
3052/**
3053 * vxge_change_mtu
3054 * @dev: net device pointer.
3055 * @new_mtu :the new MTU size for the device.
3056 *
3057 * A driver entry point to change MTU size for the device. Before changing
3058 * the MTU the device must be stopped.
3059 */
3060static int vxge_change_mtu(struct net_device *dev, int new_mtu)
3061{
3062	struct vxgedev *vdev = netdev_priv(dev);
3063
3064	vxge_debug_entryexit(vdev->level_trace,
3065		"%s:%d", __func__, __LINE__);
3066
3067	/* check if device is down already */
3068	if (unlikely(!is_vxge_card_up(vdev))) {
3069		/* just store new value, will use later on open() */
3070		dev->mtu = new_mtu;
3071		vxge_debug_init(vdev->level_err,
3072			"%s", "device is down on MTU change");
3073		return 0;
3074	}
3075
3076	vxge_debug_init(vdev->level_trace,
3077		"trying to apply new MTU %d", new_mtu);
3078
3079	if (vxge_close(dev))
3080		return -EIO;
3081
3082	dev->mtu = new_mtu;
3083	vdev->mtu = new_mtu;
3084
3085	if (vxge_open(dev))
3086		return -EIO;
3087
3088	vxge_debug_init(vdev->level_trace,
3089		"%s: MTU changed to %d", vdev->ndev->name, new_mtu);
3090
3091	vxge_debug_entryexit(vdev->level_trace,
3092		"%s:%d  Exiting...", __func__, __LINE__);
3093
3094	return 0;
3095}
3096
3097/**
3098 * vxge_get_stats64
3099 * @dev: pointer to the device structure
3100 * @net_stats: pointer to struct rtnl_link_stats64
3101 *
3102 */
3103static void
3104vxge_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *net_stats)
3105{
3106	struct vxgedev *vdev = netdev_priv(dev);
3107	int k;
3108
3109	/* net_stats already zeroed by caller */
3110	for (k = 0; k < vdev->no_of_vpath; k++) {
3111		struct vxge_ring_stats *rxstats = &vdev->vpaths[k].ring.stats;
3112		struct vxge_fifo_stats *txstats = &vdev->vpaths[k].fifo.stats;
3113		unsigned int start;
3114		u64 packets, bytes, multicast;
3115
3116		do {
3117			start = u64_stats_fetch_begin_irq(&rxstats->syncp);
3118
3119			packets   = rxstats->rx_frms;
3120			multicast = rxstats->rx_mcast;
3121			bytes     = rxstats->rx_bytes;
3122		} while (u64_stats_fetch_retry_irq(&rxstats->syncp, start));
3123
3124		net_stats->rx_packets += packets;
3125		net_stats->rx_bytes += bytes;
3126		net_stats->multicast += multicast;
3127
3128		net_stats->rx_errors += rxstats->rx_errors;
3129		net_stats->rx_dropped += rxstats->rx_dropped;
3130
3131		do {
3132			start = u64_stats_fetch_begin_irq(&txstats->syncp);
3133
3134			packets = txstats->tx_frms;
3135			bytes   = txstats->tx_bytes;
3136		} while (u64_stats_fetch_retry_irq(&txstats->syncp, start));
3137
3138		net_stats->tx_packets += packets;
3139		net_stats->tx_bytes += bytes;
3140		net_stats->tx_errors += txstats->tx_errors;
3141	}
3142}
3143
3144static enum vxge_hw_status vxge_timestamp_config(struct __vxge_hw_device *devh)
3145{
3146	enum vxge_hw_status status;
3147	u64 val64;
3148
3149	/* Timestamp is passed to the driver via the FCS, therefore we
3150	 * must disable the FCS stripping by the adapter.  Since this is
3151	 * required for the driver to load (due to a hardware bug),
3152	 * there is no need to do anything special here.
3153	 */
3154	val64 = VXGE_HW_XMAC_TIMESTAMP_EN |
3155		VXGE_HW_XMAC_TIMESTAMP_USE_LINK_ID(0) |
3156		VXGE_HW_XMAC_TIMESTAMP_INTERVAL(0);
3157
3158	status = vxge_hw_mgmt_reg_write(devh,
3159					vxge_hw_mgmt_reg_type_mrpcim,
3160					0,
3161					offsetof(struct vxge_hw_mrpcim_reg,
3162						 xmac_timestamp),
3163					val64);
3164	vxge_hw_device_flush_io(devh);
3165	devh->config.hwts_en = VXGE_HW_HWTS_ENABLE;
3166	return status;
3167}
3168
3169static int vxge_hwtstamp_set(struct vxgedev *vdev, void __user *data)
3170{
3171	struct hwtstamp_config config;
3172	int i;
3173
3174	if (copy_from_user(&config, data, sizeof(config)))
3175		return -EFAULT;
3176
3177	/* reserved for future extensions */
3178	if (config.flags)
3179		return -EINVAL;
3180
3181	/* Transmit HW Timestamp not supported */
3182	switch (config.tx_type) {
3183	case HWTSTAMP_TX_OFF:
3184		break;
3185	case HWTSTAMP_TX_ON:
3186	default:
3187		return -ERANGE;
3188	}
3189
3190	switch (config.rx_filter) {
3191	case HWTSTAMP_FILTER_NONE:
3192		vdev->rx_hwts = 0;
3193		config.rx_filter = HWTSTAMP_FILTER_NONE;
3194		break;
3195
3196	case HWTSTAMP_FILTER_ALL:
3197	case HWTSTAMP_FILTER_SOME:
3198	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3199	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3200	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3201	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3202	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3203	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3204	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3205	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3206	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3207	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3208	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3209	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3210	case HWTSTAMP_FILTER_NTP_ALL:
3211		if (vdev->devh->config.hwts_en != VXGE_HW_HWTS_ENABLE)
3212			return -EFAULT;
3213
3214		vdev->rx_hwts = 1;
3215		config.rx_filter = HWTSTAMP_FILTER_ALL;
3216		break;
3217
3218	default:
3219		 return -ERANGE;
3220	}
3221
3222	for (i = 0; i < vdev->no_of_vpath; i++)
3223		vdev->vpaths[i].ring.rx_hwts = vdev->rx_hwts;
3224
3225	if (copy_to_user(data, &config, sizeof(config)))
3226		return -EFAULT;
3227
3228	return 0;
3229}
3230
3231static int vxge_hwtstamp_get(struct vxgedev *vdev, void __user *data)
3232{
3233	struct hwtstamp_config config;
3234
3235	config.flags = 0;
3236	config.tx_type = HWTSTAMP_TX_OFF;
3237	config.rx_filter = (vdev->rx_hwts ?
3238			    HWTSTAMP_FILTER_ALL : HWTSTAMP_FILTER_NONE);
3239
3240	if (copy_to_user(data, &config, sizeof(config)))
3241		return -EFAULT;
3242
3243	return 0;
3244}
3245
3246/**
3247 * vxge_ioctl
3248 * @dev: Device pointer.
3249 * @rq: An IOCTL specific structure, that can contain a pointer to
3250 *       a proprietary structure used to pass information to the driver.
3251 * @cmd: This is used to distinguish between the different commands that
3252 *       can be passed to the IOCTL functions.
3253 *
3254 * Entry point for the Ioctl.
3255 */
3256static int vxge_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
3257{
3258	struct vxgedev *vdev = netdev_priv(dev);
3259
3260	switch (cmd) {
3261	case SIOCSHWTSTAMP:
3262		return vxge_hwtstamp_set(vdev, rq->ifr_data);
3263	case SIOCGHWTSTAMP:
3264		return vxge_hwtstamp_get(vdev, rq->ifr_data);
3265	default:
3266		return -EOPNOTSUPP;
3267	}
3268}
3269
3270/**
3271 * vxge_tx_watchdog
3272 * @dev: pointer to net device structure
3273 * @txqueue: index of the hanging queue
3274 *
3275 * Watchdog for transmit side.
3276 * This function is triggered if the Tx Queue is stopped
3277 * for a pre-defined amount of time when the Interface is still up.
3278 */
3279static void vxge_tx_watchdog(struct net_device *dev, unsigned int txqueue)
3280{
3281	struct vxgedev *vdev;
3282
3283	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
3284
3285	vdev = netdev_priv(dev);
3286
3287	vdev->cric_err_event = VXGE_HW_EVENT_RESET_START;
3288
3289	schedule_work(&vdev->reset_task);
3290	vxge_debug_entryexit(VXGE_TRACE,
3291		"%s:%d  Exiting...", __func__, __LINE__);
3292}
3293
3294/**
3295 * vxge_vlan_rx_add_vid
3296 * @dev: net device pointer.
3297 * @proto: vlan protocol
3298 * @vid: vid
3299 *
3300 * Add the vlan id to the devices vlan id table
3301 */
3302static int
3303vxge_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid)
3304{
3305	struct vxgedev *vdev = netdev_priv(dev);
3306	struct vxge_vpath *vpath;
3307	int vp_id;
3308
3309	/* Add these vlan to the vid table */
3310	for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
3311		vpath = &vdev->vpaths[vp_id];
3312		if (!vpath->is_open)
3313			continue;
3314		vxge_hw_vpath_vid_add(vpath->handle, vid);
3315	}
3316	set_bit(vid, vdev->active_vlans);
3317	return 0;
3318}
3319
3320/**
3321 * vxge_vlan_rx_kill_vid
3322 * @dev: net device pointer.
3323 * @proto: vlan protocol
3324 * @vid: vid
3325 *
3326 * Remove the vlan id from the device's vlan id table
3327 */
3328static int
3329vxge_vlan_rx_kill_vid(struct net_device *dev, __be16 proto, u16 vid)
3330{
3331	struct vxgedev *vdev = netdev_priv(dev);
3332	struct vxge_vpath *vpath;
3333	int vp_id;
3334
3335	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
3336
3337	/* Delete this vlan from the vid table */
3338	for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
3339		vpath = &vdev->vpaths[vp_id];
3340		if (!vpath->is_open)
3341			continue;
3342		vxge_hw_vpath_vid_delete(vpath->handle, vid);
3343	}
3344	vxge_debug_entryexit(VXGE_TRACE,
3345		"%s:%d  Exiting...", __func__, __LINE__);
3346	clear_bit(vid, vdev->active_vlans);
3347	return 0;
3348}
3349
3350static const struct net_device_ops vxge_netdev_ops = {
3351	.ndo_open               = vxge_open,
3352	.ndo_stop               = vxge_close,
3353	.ndo_get_stats64        = vxge_get_stats64,
3354	.ndo_start_xmit         = vxge_xmit,
3355	.ndo_validate_addr      = eth_validate_addr,
3356	.ndo_set_rx_mode	= vxge_set_multicast,
3357	.ndo_do_ioctl           = vxge_ioctl,
3358	.ndo_set_mac_address    = vxge_set_mac_addr,
3359	.ndo_change_mtu         = vxge_change_mtu,
3360	.ndo_fix_features	= vxge_fix_features,
3361	.ndo_set_features	= vxge_set_features,
3362	.ndo_vlan_rx_kill_vid   = vxge_vlan_rx_kill_vid,
3363	.ndo_vlan_rx_add_vid	= vxge_vlan_rx_add_vid,
3364	.ndo_tx_timeout         = vxge_tx_watchdog,
3365#ifdef CONFIG_NET_POLL_CONTROLLER
3366	.ndo_poll_controller    = vxge_netpoll,
3367#endif
3368};
3369
3370static int vxge_device_register(struct __vxge_hw_device *hldev,
3371				struct vxge_config *config, int high_dma,
3372				int no_of_vpath, struct vxgedev **vdev_out)
3373{
3374	struct net_device *ndev;
3375	enum vxge_hw_status status = VXGE_HW_OK;
3376	struct vxgedev *vdev;
3377	int ret = 0, no_of_queue = 1;
3378	u64 stat;
3379
3380	*vdev_out = NULL;
3381	if (config->tx_steering_type)
3382		no_of_queue = no_of_vpath;
3383
3384	ndev = alloc_etherdev_mq(sizeof(struct vxgedev),
3385			no_of_queue);
3386	if (ndev == NULL) {
3387		vxge_debug_init(
3388			vxge_hw_device_trace_level_get(hldev),
3389		"%s : device allocation failed", __func__);
3390		ret = -ENODEV;
3391		goto _out0;
3392	}
3393
3394	vxge_debug_entryexit(
3395		vxge_hw_device_trace_level_get(hldev),
3396		"%s: %s:%d  Entering...",
3397		ndev->name, __func__, __LINE__);
3398
3399	vdev = netdev_priv(ndev);
3400	memset(vdev, 0, sizeof(struct vxgedev));
3401
3402	vdev->ndev = ndev;
3403	vdev->devh = hldev;
3404	vdev->pdev = hldev->pdev;
3405	memcpy(&vdev->config, config, sizeof(struct vxge_config));
3406	vdev->rx_hwts = 0;
3407	vdev->titan1 = (vdev->pdev->revision == VXGE_HW_TITAN1_PCI_REVISION);
3408
3409	SET_NETDEV_DEV(ndev, &vdev->pdev->dev);
3410
3411	ndev->hw_features = NETIF_F_RXCSUM | NETIF_F_SG |
3412		NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
3413		NETIF_F_TSO | NETIF_F_TSO6 |
3414		NETIF_F_HW_VLAN_CTAG_TX;
3415	if (vdev->config.rth_steering != NO_STEERING)
3416		ndev->hw_features |= NETIF_F_RXHASH;
3417
3418	ndev->features |= ndev->hw_features |
3419		NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
3420
3421
3422	ndev->netdev_ops = &vxge_netdev_ops;
3423
3424	ndev->watchdog_timeo = VXGE_LL_WATCH_DOG_TIMEOUT;
3425	INIT_WORK(&vdev->reset_task, vxge_reset);
3426
3427	vxge_initialize_ethtool_ops(ndev);
3428
3429	/* Allocate memory for vpath */
3430	vdev->vpaths = kcalloc(no_of_vpath, sizeof(struct vxge_vpath),
3431			       GFP_KERNEL);
3432	if (!vdev->vpaths) {
3433		vxge_debug_init(VXGE_ERR,
3434			"%s: vpath memory allocation failed",
3435			vdev->ndev->name);
3436		ret = -ENOMEM;
3437		goto _out1;
3438	}
3439
3440	vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3441		"%s : checksumming enabled", __func__);
3442
3443	if (high_dma) {
3444		ndev->features |= NETIF_F_HIGHDMA;
3445		vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3446			"%s : using High DMA", __func__);
3447	}
3448
3449	/* MTU range: 68 - 9600 */
3450	ndev->min_mtu = VXGE_HW_MIN_MTU;
3451	ndev->max_mtu = VXGE_HW_MAX_MTU;
3452
3453	ret = register_netdev(ndev);
3454	if (ret) {
3455		vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3456			"%s: %s : device registration failed!",
3457			ndev->name, __func__);
3458		goto _out2;
3459	}
3460
3461	/*  Set the factory defined MAC address initially */
3462	ndev->addr_len = ETH_ALEN;
3463
3464	/* Make Link state as off at this point, when the Link change
3465	 * interrupt comes the state will be automatically changed to
3466	 * the right state.
3467	 */
3468	netif_carrier_off(ndev);
3469
3470	vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3471		"%s: Ethernet device registered",
3472		ndev->name);
3473
3474	hldev->ndev = ndev;
3475	*vdev_out = vdev;
3476
3477	/* Resetting the Device stats */
3478	status = vxge_hw_mrpcim_stats_access(
3479				hldev,
3480				VXGE_HW_STATS_OP_CLEAR_ALL_STATS,
3481				0,
3482				0,
3483				&stat);
3484
3485	if (status == VXGE_HW_ERR_PRIVILEGED_OPERATION)
3486		vxge_debug_init(
3487			vxge_hw_device_trace_level_get(hldev),
3488			"%s: device stats clear returns"
3489			"VXGE_HW_ERR_PRIVILEGED_OPERATION", ndev->name);
3490
3491	vxge_debug_entryexit(vxge_hw_device_trace_level_get(hldev),
3492		"%s: %s:%d  Exiting...",
3493		ndev->name, __func__, __LINE__);
3494
3495	return ret;
3496_out2:
3497	kfree(vdev->vpaths);
3498_out1:
3499	free_netdev(ndev);
3500_out0:
3501	return ret;
3502}
3503
3504/*
3505 * vxge_device_unregister
3506 *
3507 * This function will unregister and free network device
3508 */
3509static void vxge_device_unregister(struct __vxge_hw_device *hldev)
3510{
3511	struct vxgedev *vdev;
3512	struct net_device *dev;
3513	char buf[IFNAMSIZ];
3514
3515	dev = hldev->ndev;
3516	vdev = netdev_priv(dev);
3517
3518	vxge_debug_entryexit(vdev->level_trace,	"%s: %s:%d", vdev->ndev->name,
3519			     __func__, __LINE__);
3520
3521	strlcpy(buf, dev->name, IFNAMSIZ);
3522
3523	flush_work(&vdev->reset_task);
3524
3525	/* in 2.6 will call stop() if device is up */
3526	unregister_netdev(dev);
3527
3528	kfree(vdev->vpaths);
3529
3530	vxge_debug_init(vdev->level_trace, "%s: ethernet device unregistered",
3531			buf);
3532	vxge_debug_entryexit(vdev->level_trace,	"%s: %s:%d  Exiting...", buf,
3533			     __func__, __LINE__);
3534
3535	/* we are safe to free it now */
3536	free_netdev(dev);
3537}
3538
3539/*
3540 * vxge_callback_crit_err
3541 *
3542 * This function is called by the alarm handler in interrupt context.
3543 * Driver must analyze it based on the event type.
3544 */
3545static void
3546vxge_callback_crit_err(struct __vxge_hw_device *hldev,
3547			enum vxge_hw_event type, u64 vp_id)
3548{
3549	struct net_device *dev = hldev->ndev;
3550	struct vxgedev *vdev = netdev_priv(dev);
3551	struct vxge_vpath *vpath = NULL;
3552	int vpath_idx;
3553
3554	vxge_debug_entryexit(vdev->level_trace,
3555		"%s: %s:%d", vdev->ndev->name, __func__, __LINE__);
3556
3557	/* Note: This event type should be used for device wide
3558	 * indications only - Serious errors, Slot freeze and critical errors
3559	 */
3560	vdev->cric_err_event = type;
3561
3562	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
3563		vpath = &vdev->vpaths[vpath_idx];
3564		if (vpath->device_id == vp_id)
3565			break;
3566	}
3567
3568	if (!test_bit(__VXGE_STATE_RESET_CARD, &vdev->state)) {
3569		if (type == VXGE_HW_EVENT_SLOT_FREEZE) {
3570			vxge_debug_init(VXGE_ERR,
3571				"%s: Slot is frozen", vdev->ndev->name);
3572		} else if (type == VXGE_HW_EVENT_SERR) {
3573			vxge_debug_init(VXGE_ERR,
3574				"%s: Encountered Serious Error",
3575				vdev->ndev->name);
3576		} else if (type == VXGE_HW_EVENT_CRITICAL_ERR)
3577			vxge_debug_init(VXGE_ERR,
3578				"%s: Encountered Critical Error",
3579				vdev->ndev->name);
3580	}
3581
3582	if ((type == VXGE_HW_EVENT_SERR) ||
3583		(type == VXGE_HW_EVENT_SLOT_FREEZE)) {
3584		if (unlikely(vdev->exec_mode))
3585			clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3586	} else if (type == VXGE_HW_EVENT_CRITICAL_ERR) {
3587		vxge_hw_device_mask_all(hldev);
3588		if (unlikely(vdev->exec_mode))
3589			clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3590	} else if ((type == VXGE_HW_EVENT_FIFO_ERR) ||
3591		  (type == VXGE_HW_EVENT_VPATH_ERR)) {
3592
3593		if (unlikely(vdev->exec_mode))
3594			clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3595		else {
3596			/* check if this vpath is already set for reset */
3597			if (!test_and_set_bit(vpath_idx, &vdev->vp_reset)) {
3598
3599				/* disable interrupts for this vpath */
3600				vxge_vpath_intr_disable(vdev, vpath_idx);
3601
3602				/* stop the queue for this vpath */
3603				netif_tx_stop_queue(vpath->fifo.txq);
3604			}
3605		}
3606	}
3607
3608	vxge_debug_entryexit(vdev->level_trace,
3609		"%s: %s:%d  Exiting...",
3610		vdev->ndev->name, __func__, __LINE__);
3611}
3612
3613static void verify_bandwidth(void)
3614{
3615	int i, band_width, total = 0, equal_priority = 0;
3616
3617	/* 1. If user enters 0 for some fifo, give equal priority to all */
3618	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3619		if (bw_percentage[i] == 0) {
3620			equal_priority = 1;
3621			break;
3622		}
3623	}
3624
3625	if (!equal_priority) {
3626		/* 2. If sum exceeds 100, give equal priority to all */
3627		for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3628			if (bw_percentage[i] == 0xFF)
3629				break;
3630
3631			total += bw_percentage[i];
3632			if (total > VXGE_HW_VPATH_BANDWIDTH_MAX) {
3633				equal_priority = 1;
3634				break;
3635			}
3636		}
3637	}
3638
3639	if (!equal_priority) {
3640		/* Is all the bandwidth consumed? */
3641		if (total < VXGE_HW_VPATH_BANDWIDTH_MAX) {
3642			if (i < VXGE_HW_MAX_VIRTUAL_PATHS) {
3643				/* Split rest of bw equally among next VPs*/
3644				band_width =
3645				  (VXGE_HW_VPATH_BANDWIDTH_MAX  - total) /
3646					(VXGE_HW_MAX_VIRTUAL_PATHS - i);
3647				if (band_width < 2) /* min of 2% */
3648					equal_priority = 1;
3649				else {
3650					for (; i < VXGE_HW_MAX_VIRTUAL_PATHS;
3651						i++)
3652						bw_percentage[i] =
3653							band_width;
3654				}
3655			}
3656		} else if (i < VXGE_HW_MAX_VIRTUAL_PATHS)
3657			equal_priority = 1;
3658	}
3659
3660	if (equal_priority) {
3661		vxge_debug_init(VXGE_ERR,
3662			"%s: Assigning equal bandwidth to all the vpaths",
3663			VXGE_DRIVER_NAME);
3664		bw_percentage[0] = VXGE_HW_VPATH_BANDWIDTH_MAX /
3665					VXGE_HW_MAX_VIRTUAL_PATHS;
3666		for (i = 1; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
3667			bw_percentage[i] = bw_percentage[0];
3668	}
3669}
3670
3671/*
3672 * Vpath configuration
3673 */
3674static int vxge_config_vpaths(struct vxge_hw_device_config *device_config,
3675			      u64 vpath_mask, struct vxge_config *config_param)
3676{
3677	int i, no_of_vpaths = 0, default_no_vpath = 0, temp;
3678	u32 txdl_size, txdl_per_memblock;
3679
3680	temp = driver_config->vpath_per_dev;
3681	if ((driver_config->vpath_per_dev == VXGE_USE_DEFAULT) &&
3682		(max_config_dev == VXGE_MAX_CONFIG_DEV)) {
3683		/* No more CPU. Return vpath number as zero.*/
3684		if (driver_config->g_no_cpus == -1)
3685			return 0;
3686
3687		if (!driver_config->g_no_cpus)
3688			driver_config->g_no_cpus =
3689				netif_get_num_default_rss_queues();
3690
3691		driver_config->vpath_per_dev = driver_config->g_no_cpus >> 1;
3692		if (!driver_config->vpath_per_dev)
3693			driver_config->vpath_per_dev = 1;
3694
3695		for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
3696			if (!vxge_bVALn(vpath_mask, i, 1))
3697				continue;
3698			else
3699				default_no_vpath++;
3700		if (default_no_vpath < driver_config->vpath_per_dev)
3701			driver_config->vpath_per_dev = default_no_vpath;
3702
3703		driver_config->g_no_cpus = driver_config->g_no_cpus -
3704				(driver_config->vpath_per_dev * 2);
3705		if (driver_config->g_no_cpus <= 0)
3706			driver_config->g_no_cpus = -1;
3707	}
3708
3709	if (driver_config->vpath_per_dev == 1) {
3710		vxge_debug_ll_config(VXGE_TRACE,
3711			"%s: Disable tx and rx steering, "
3712			"as single vpath is configured", VXGE_DRIVER_NAME);
3713		config_param->rth_steering = NO_STEERING;
3714		config_param->tx_steering_type = NO_STEERING;
3715		device_config->rth_en = 0;
3716	}
3717
3718	/* configure bandwidth */
3719	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
3720		device_config->vp_config[i].min_bandwidth = bw_percentage[i];
3721
3722	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3723		device_config->vp_config[i].vp_id = i;
3724		device_config->vp_config[i].mtu = VXGE_HW_DEFAULT_MTU;
3725		if (no_of_vpaths < driver_config->vpath_per_dev) {
3726			if (!vxge_bVALn(vpath_mask, i, 1)) {
3727				vxge_debug_ll_config(VXGE_TRACE,
3728					"%s: vpath: %d is not available",
3729					VXGE_DRIVER_NAME, i);
3730				continue;
3731			} else {
3732				vxge_debug_ll_config(VXGE_TRACE,
3733					"%s: vpath: %d available",
3734					VXGE_DRIVER_NAME, i);
3735				no_of_vpaths++;
3736			}
3737		} else {
3738			vxge_debug_ll_config(VXGE_TRACE,
3739				"%s: vpath: %d is not configured, "
3740				"max_config_vpath exceeded",
3741				VXGE_DRIVER_NAME, i);
3742			break;
3743		}
3744
3745		/* Configure Tx fifo's */
3746		device_config->vp_config[i].fifo.enable =
3747						VXGE_HW_FIFO_ENABLE;
3748		device_config->vp_config[i].fifo.max_frags =
3749				MAX_SKB_FRAGS + 1;
3750		device_config->vp_config[i].fifo.memblock_size =
3751			VXGE_HW_MIN_FIFO_MEMBLOCK_SIZE;
3752
3753		txdl_size = device_config->vp_config[i].fifo.max_frags *
3754				sizeof(struct vxge_hw_fifo_txd);
3755		txdl_per_memblock = VXGE_HW_MIN_FIFO_MEMBLOCK_SIZE / txdl_size;
3756
3757		device_config->vp_config[i].fifo.fifo_blocks =
3758			((VXGE_DEF_FIFO_LENGTH - 1) / txdl_per_memblock) + 1;
3759
3760		device_config->vp_config[i].fifo.intr =
3761				VXGE_HW_FIFO_QUEUE_INTR_DISABLE;
3762
3763		/* Configure tti properties */
3764		device_config->vp_config[i].tti.intr_enable =
3765					VXGE_HW_TIM_INTR_ENABLE;
3766
3767		device_config->vp_config[i].tti.btimer_val =
3768			(VXGE_TTI_BTIMER_VAL * 1000) / 272;
3769
3770		device_config->vp_config[i].tti.timer_ac_en =
3771				VXGE_HW_TIM_TIMER_AC_ENABLE;
3772
3773		/* For msi-x with napi (each vector has a handler of its own) -
3774		 * Set CI to OFF for all vpaths
3775		 */
3776		device_config->vp_config[i].tti.timer_ci_en =
3777			VXGE_HW_TIM_TIMER_CI_DISABLE;
3778
3779		device_config->vp_config[i].tti.timer_ri_en =
3780				VXGE_HW_TIM_TIMER_RI_DISABLE;
3781
3782		device_config->vp_config[i].tti.util_sel =
3783			VXGE_HW_TIM_UTIL_SEL_LEGACY_TX_NET_UTIL;
3784
3785		device_config->vp_config[i].tti.ltimer_val =
3786			(VXGE_TTI_LTIMER_VAL * 1000) / 272;
3787
3788		device_config->vp_config[i].tti.rtimer_val =
3789			(VXGE_TTI_RTIMER_VAL * 1000) / 272;
3790
3791		device_config->vp_config[i].tti.urange_a = TTI_TX_URANGE_A;
3792		device_config->vp_config[i].tti.urange_b = TTI_TX_URANGE_B;
3793		device_config->vp_config[i].tti.urange_c = TTI_TX_URANGE_C;
3794		device_config->vp_config[i].tti.uec_a = TTI_TX_UFC_A;
3795		device_config->vp_config[i].tti.uec_b = TTI_TX_UFC_B;
3796		device_config->vp_config[i].tti.uec_c = TTI_TX_UFC_C;
3797		device_config->vp_config[i].tti.uec_d = TTI_TX_UFC_D;
3798
3799		/* Configure Rx rings */
3800		device_config->vp_config[i].ring.enable  =
3801						VXGE_HW_RING_ENABLE;
3802
3803		device_config->vp_config[i].ring.ring_blocks  =
3804						VXGE_HW_DEF_RING_BLOCKS;
3805
3806		device_config->vp_config[i].ring.buffer_mode =
3807			VXGE_HW_RING_RXD_BUFFER_MODE_1;
3808
3809		device_config->vp_config[i].ring.rxds_limit  =
3810				VXGE_HW_DEF_RING_RXDS_LIMIT;
3811
3812		device_config->vp_config[i].ring.scatter_mode =
3813					VXGE_HW_RING_SCATTER_MODE_A;
3814
3815		/* Configure rti properties */
3816		device_config->vp_config[i].rti.intr_enable =
3817					VXGE_HW_TIM_INTR_ENABLE;
3818
3819		device_config->vp_config[i].rti.btimer_val =
3820			(VXGE_RTI_BTIMER_VAL * 1000)/272;
3821
3822		device_config->vp_config[i].rti.timer_ac_en =
3823						VXGE_HW_TIM_TIMER_AC_ENABLE;
3824
3825		device_config->vp_config[i].rti.timer_ci_en =
3826						VXGE_HW_TIM_TIMER_CI_DISABLE;
3827
3828		device_config->vp_config[i].rti.timer_ri_en =
3829						VXGE_HW_TIM_TIMER_RI_DISABLE;
3830
3831		device_config->vp_config[i].rti.util_sel =
3832				VXGE_HW_TIM_UTIL_SEL_LEGACY_RX_NET_UTIL;
3833
3834		device_config->vp_config[i].rti.urange_a =
3835						RTI_RX_URANGE_A;
3836		device_config->vp_config[i].rti.urange_b =
3837						RTI_RX_URANGE_B;
3838		device_config->vp_config[i].rti.urange_c =
3839						RTI_RX_URANGE_C;
3840		device_config->vp_config[i].rti.uec_a = RTI_RX_UFC_A;
3841		device_config->vp_config[i].rti.uec_b = RTI_RX_UFC_B;
3842		device_config->vp_config[i].rti.uec_c = RTI_RX_UFC_C;
3843		device_config->vp_config[i].rti.uec_d = RTI_RX_UFC_D;
3844
3845		device_config->vp_config[i].rti.rtimer_val =
3846			(VXGE_RTI_RTIMER_VAL * 1000) / 272;
3847
3848		device_config->vp_config[i].rti.ltimer_val =
3849			(VXGE_RTI_LTIMER_VAL * 1000) / 272;
3850
3851		device_config->vp_config[i].rpa_strip_vlan_tag =
3852			vlan_tag_strip;
3853	}
3854
3855	driver_config->vpath_per_dev = temp;
3856	return no_of_vpaths;
3857}
3858
3859/* initialize device configuratrions */
3860static void vxge_device_config_init(struct vxge_hw_device_config *device_config,
3861				    int *intr_type)
3862{
3863	/* Used for CQRQ/SRQ. */
3864	device_config->dma_blockpool_initial =
3865			VXGE_HW_INITIAL_DMA_BLOCK_POOL_SIZE;
3866
3867	device_config->dma_blockpool_max =
3868			VXGE_HW_MAX_DMA_BLOCK_POOL_SIZE;
3869
3870	if (max_mac_vpath > VXGE_MAX_MAC_ADDR_COUNT)
3871		max_mac_vpath = VXGE_MAX_MAC_ADDR_COUNT;
3872
3873	if (!IS_ENABLED(CONFIG_PCI_MSI)) {
3874		vxge_debug_init(VXGE_ERR,
3875			"%s: This Kernel does not support "
3876			"MSI-X. Defaulting to INTA", VXGE_DRIVER_NAME);
3877		*intr_type = INTA;
3878	}
3879
3880	/* Configure whether MSI-X or IRQL. */
3881	switch (*intr_type) {
3882	case INTA:
3883		device_config->intr_mode = VXGE_HW_INTR_MODE_IRQLINE;
3884		break;
3885
3886	case MSI_X:
3887		device_config->intr_mode = VXGE_HW_INTR_MODE_MSIX_ONE_SHOT;
3888		break;
3889	}
3890
3891	/* Timer period between device poll */
3892	device_config->device_poll_millis = VXGE_TIMER_DELAY;
3893
3894	/* Configure mac based steering. */
3895	device_config->rts_mac_en = addr_learn_en;
3896
3897	/* Configure Vpaths */
3898	device_config->rth_it_type = VXGE_HW_RTH_IT_TYPE_MULTI_IT;
3899
3900	vxge_debug_ll_config(VXGE_TRACE, "%s : Device Config Params ",
3901			__func__);
3902	vxge_debug_ll_config(VXGE_TRACE, "intr_mode : %d",
3903			device_config->intr_mode);
3904	vxge_debug_ll_config(VXGE_TRACE, "device_poll_millis : %d",
3905			device_config->device_poll_millis);
3906	vxge_debug_ll_config(VXGE_TRACE, "rth_en : %d",
3907			device_config->rth_en);
3908	vxge_debug_ll_config(VXGE_TRACE, "rth_it_type : %d",
3909			device_config->rth_it_type);
3910}
3911
3912static void vxge_print_parm(struct vxgedev *vdev, u64 vpath_mask)
3913{
3914	int i;
3915
3916	vxge_debug_init(VXGE_TRACE,
3917		"%s: %d Vpath(s) opened",
3918		vdev->ndev->name, vdev->no_of_vpath);
3919
3920	switch (vdev->config.intr_type) {
3921	case INTA:
3922		vxge_debug_init(VXGE_TRACE,
3923			"%s: Interrupt type INTA", vdev->ndev->name);
3924		break;
3925
3926	case MSI_X:
3927		vxge_debug_init(VXGE_TRACE,
3928			"%s: Interrupt type MSI-X", vdev->ndev->name);
3929		break;
3930	}
3931
3932	if (vdev->config.rth_steering) {
3933		vxge_debug_init(VXGE_TRACE,
3934			"%s: RTH steering enabled for TCP_IPV4",
3935			vdev->ndev->name);
3936	} else {
3937		vxge_debug_init(VXGE_TRACE,
3938			"%s: RTH steering disabled", vdev->ndev->name);
3939	}
3940
3941	switch (vdev->config.tx_steering_type) {
3942	case NO_STEERING:
3943		vxge_debug_init(VXGE_TRACE,
3944			"%s: Tx steering disabled", vdev->ndev->name);
3945		break;
3946	case TX_PRIORITY_STEERING:
3947		vxge_debug_init(VXGE_TRACE,
3948			"%s: Unsupported tx steering option",
3949			vdev->ndev->name);
3950		vxge_debug_init(VXGE_TRACE,
3951			"%s: Tx steering disabled", vdev->ndev->name);
3952		vdev->config.tx_steering_type = 0;
3953		break;
3954	case TX_VLAN_STEERING:
3955		vxge_debug_init(VXGE_TRACE,
3956			"%s: Unsupported tx steering option",
3957			vdev->ndev->name);
3958		vxge_debug_init(VXGE_TRACE,
3959			"%s: Tx steering disabled", vdev->ndev->name);
3960		vdev->config.tx_steering_type = 0;
3961		break;
3962	case TX_MULTIQ_STEERING:
3963		vxge_debug_init(VXGE_TRACE,
3964			"%s: Tx multiqueue steering enabled",
3965			vdev->ndev->name);
3966		break;
3967	case TX_PORT_STEERING:
3968		vxge_debug_init(VXGE_TRACE,
3969			"%s: Tx port steering enabled",
3970			vdev->ndev->name);
3971		break;
3972	default:
3973		vxge_debug_init(VXGE_ERR,
3974			"%s: Unsupported tx steering type",
3975			vdev->ndev->name);
3976		vxge_debug_init(VXGE_TRACE,
3977			"%s: Tx steering disabled", vdev->ndev->name);
3978		vdev->config.tx_steering_type = 0;
3979	}
3980
3981	if (vdev->config.addr_learn_en)
3982		vxge_debug_init(VXGE_TRACE,
3983			"%s: MAC Address learning enabled", vdev->ndev->name);
3984
3985	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3986		if (!vxge_bVALn(vpath_mask, i, 1))
3987			continue;
3988		vxge_debug_ll_config(VXGE_TRACE,
3989			"%s: MTU size - %d", vdev->ndev->name,
3990			((vdev->devh))->
3991				config.vp_config[i].mtu);
3992		vxge_debug_init(VXGE_TRACE,
3993			"%s: VLAN tag stripping %s", vdev->ndev->name,
3994			((vdev->devh))->
3995				config.vp_config[i].rpa_strip_vlan_tag
3996			? "Enabled" : "Disabled");
3997		vxge_debug_ll_config(VXGE_TRACE,
3998			"%s: Max frags : %d", vdev->ndev->name,
3999			((vdev->devh))->
4000				config.vp_config[i].fifo.max_frags);
4001		break;
4002	}
4003}
4004
4005/**
4006 * vxge_pm_suspend - vxge power management suspend entry point
4007 * @dev_d: device pointer
4008 *
4009 */
4010static int __maybe_unused vxge_pm_suspend(struct device *dev_d)
4011{
4012	return -ENOSYS;
4013}
4014/**
4015 * vxge_pm_resume - vxge power management resume entry point
4016 * @dev_d: device pointer
4017 *
4018 */
4019static int __maybe_unused vxge_pm_resume(struct device *dev_d)
4020{
4021	return -ENOSYS;
4022}
4023
4024/**
4025 * vxge_io_error_detected - called when PCI error is detected
4026 * @pdev: Pointer to PCI device
4027 * @state: The current pci connection state
4028 *
4029 * This function is called after a PCI bus error affecting
4030 * this device has been detected.
4031 */
4032static pci_ers_result_t vxge_io_error_detected(struct pci_dev *pdev,
4033						pci_channel_state_t state)
4034{
4035	struct __vxge_hw_device *hldev = pci_get_drvdata(pdev);
4036	struct net_device *netdev = hldev->ndev;
4037
4038	netif_device_detach(netdev);
4039
4040	if (state == pci_channel_io_perm_failure)
4041		return PCI_ERS_RESULT_DISCONNECT;
4042
4043	if (netif_running(netdev)) {
4044		/* Bring down the card, while avoiding PCI I/O */
4045		do_vxge_close(netdev, 0);
4046	}
4047
4048	pci_disable_device(pdev);
4049
4050	return PCI_ERS_RESULT_NEED_RESET;
4051}
4052
4053/**
4054 * vxge_io_slot_reset - called after the pci bus has been reset.
4055 * @pdev: Pointer to PCI device
4056 *
4057 * Restart the card from scratch, as if from a cold-boot.
4058 * At this point, the card has exprienced a hard reset,
4059 * followed by fixups by BIOS, and has its config space
4060 * set up identically to what it was at cold boot.
4061 */
4062static pci_ers_result_t vxge_io_slot_reset(struct pci_dev *pdev)
4063{
4064	struct __vxge_hw_device *hldev = pci_get_drvdata(pdev);
4065	struct net_device *netdev = hldev->ndev;
4066
4067	struct vxgedev *vdev = netdev_priv(netdev);
4068
4069	if (pci_enable_device(pdev)) {
4070		netdev_err(netdev, "Cannot re-enable device after reset\n");
4071		return PCI_ERS_RESULT_DISCONNECT;
4072	}
4073
4074	pci_set_master(pdev);
4075	do_vxge_reset(vdev, VXGE_LL_FULL_RESET);
4076
4077	return PCI_ERS_RESULT_RECOVERED;
4078}
4079
4080/**
4081 * vxge_io_resume - called when traffic can start flowing again.
4082 * @pdev: Pointer to PCI device
4083 *
4084 * This callback is called when the error recovery driver tells
4085 * us that its OK to resume normal operation.
4086 */
4087static void vxge_io_resume(struct pci_dev *pdev)
4088{
4089	struct __vxge_hw_device *hldev = pci_get_drvdata(pdev);
4090	struct net_device *netdev = hldev->ndev;
4091
4092	if (netif_running(netdev)) {
4093		if (vxge_open(netdev)) {
4094			netdev_err(netdev,
4095				   "Can't bring device back up after reset\n");
4096			return;
4097		}
4098	}
4099
4100	netif_device_attach(netdev);
4101}
4102
4103static inline u32 vxge_get_num_vfs(u64 function_mode)
4104{
4105	u32 num_functions = 0;
4106
4107	switch (function_mode) {
4108	case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION:
4109	case VXGE_HW_FUNCTION_MODE_SRIOV_8:
4110		num_functions = 8;
4111		break;
4112	case VXGE_HW_FUNCTION_MODE_SINGLE_FUNCTION:
4113		num_functions = 1;
4114		break;
4115	case VXGE_HW_FUNCTION_MODE_SRIOV:
4116	case VXGE_HW_FUNCTION_MODE_MRIOV:
4117	case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION_17:
4118		num_functions = 17;
4119		break;
4120	case VXGE_HW_FUNCTION_MODE_SRIOV_4:
4121		num_functions = 4;
4122		break;
4123	case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION_2:
4124		num_functions = 2;
4125		break;
4126	case VXGE_HW_FUNCTION_MODE_MRIOV_8:
4127		num_functions = 8; /* TODO */
4128		break;
4129	}
4130	return num_functions;
4131}
4132
4133int vxge_fw_upgrade(struct vxgedev *vdev, char *fw_name, int override)
4134{
4135	struct __vxge_hw_device *hldev = vdev->devh;
4136	u32 maj, min, bld, cmaj, cmin, cbld;
4137	enum vxge_hw_status status;
4138	const struct firmware *fw;
4139	int ret;
4140
4141	ret = request_firmware(&fw, fw_name, &vdev->pdev->dev);
4142	if (ret) {
4143		vxge_debug_init(VXGE_ERR, "%s: Firmware file '%s' not found",
4144				VXGE_DRIVER_NAME, fw_name);
4145		goto out;
4146	}
4147
4148	/* Load the new firmware onto the adapter */
4149	status = vxge_update_fw_image(hldev, fw->data, fw->size);
4150	if (status != VXGE_HW_OK) {
4151		vxge_debug_init(VXGE_ERR,
4152				"%s: FW image download to adapter failed '%s'.",
4153				VXGE_DRIVER_NAME, fw_name);
4154		ret = -EIO;
4155		goto out;
4156	}
4157
4158	/* Read the version of the new firmware */
4159	status = vxge_hw_upgrade_read_version(hldev, &maj, &min, &bld);
4160	if (status != VXGE_HW_OK) {
4161		vxge_debug_init(VXGE_ERR,
4162				"%s: Upgrade read version failed '%s'.",
4163				VXGE_DRIVER_NAME, fw_name);
4164		ret = -EIO;
4165		goto out;
4166	}
4167
4168	cmaj = vdev->config.device_hw_info.fw_version.major;
4169	cmin = vdev->config.device_hw_info.fw_version.minor;
4170	cbld = vdev->config.device_hw_info.fw_version.build;
4171	/* It's possible the version in /lib/firmware is not the latest version.
4172	 * If so, we could get into a loop of trying to upgrade to the latest
4173	 * and flashing the older version.
4174	 */
4175	if (VXGE_FW_VER(maj, min, bld) == VXGE_FW_VER(cmaj, cmin, cbld) &&
4176	    !override) {
4177		ret = -EINVAL;
4178		goto out;
4179	}
4180
4181	printk(KERN_NOTICE "Upgrade to firmware version %d.%d.%d commencing\n",
4182	       maj, min, bld);
4183
4184	/* Flash the adapter with the new firmware */
4185	status = vxge_hw_flash_fw(hldev);
4186	if (status != VXGE_HW_OK) {
4187		vxge_debug_init(VXGE_ERR, "%s: Upgrade commit failed '%s'.",
4188				VXGE_DRIVER_NAME, fw_name);
4189		ret = -EIO;
4190		goto out;
4191	}
4192
4193	printk(KERN_NOTICE "Upgrade of firmware successful!  Adapter must be "
4194	       "hard reset before using, thus requiring a system reboot or a "
4195	       "hotplug event.\n");
4196
4197out:
4198	release_firmware(fw);
4199	return ret;
4200}
4201
4202static int vxge_probe_fw_update(struct vxgedev *vdev)
4203{
4204	u32 maj, min, bld;
4205	int ret, gpxe = 0;
4206	char *fw_name;
4207
4208	maj = vdev->config.device_hw_info.fw_version.major;
4209	min = vdev->config.device_hw_info.fw_version.minor;
4210	bld = vdev->config.device_hw_info.fw_version.build;
4211
4212	if (VXGE_FW_VER(maj, min, bld) == VXGE_CERT_FW_VER)
4213		return 0;
4214
4215	/* Ignore the build number when determining if the current firmware is
4216	 * "too new" to load the driver
4217	 */
4218	if (VXGE_FW_VER(maj, min, 0) > VXGE_CERT_FW_VER) {
4219		vxge_debug_init(VXGE_ERR, "%s: Firmware newer than last known "
4220				"version, unable to load driver\n",
4221				VXGE_DRIVER_NAME);
4222		return -EINVAL;
4223	}
4224
4225	/* Firmware 1.4.4 and older cannot be upgraded, and is too ancient to
4226	 * work with this driver.
4227	 */
4228	if (VXGE_FW_VER(maj, min, bld) <= VXGE_FW_DEAD_VER) {
4229		vxge_debug_init(VXGE_ERR, "%s: Firmware %d.%d.%d cannot be "
4230				"upgraded\n", VXGE_DRIVER_NAME, maj, min, bld);
4231		return -EINVAL;
4232	}
4233
4234	/* If file not specified, determine gPXE or not */
4235	if (VXGE_FW_VER(maj, min, bld) >= VXGE_EPROM_FW_VER) {
4236		int i;
4237		for (i = 0; i < VXGE_HW_MAX_ROM_IMAGES; i++)
4238			if (vdev->devh->eprom_versions[i]) {
4239				gpxe = 1;
4240				break;
4241			}
4242	}
4243	if (gpxe)
4244		fw_name = "vxge/X3fw-pxe.ncf";
4245	else
4246		fw_name = "vxge/X3fw.ncf";
4247
4248	ret = vxge_fw_upgrade(vdev, fw_name, 0);
4249	/* -EINVAL and -ENOENT are not fatal errors for flashing firmware on
4250	 * probe, so ignore them
4251	 */
4252	if (ret != -EINVAL && ret != -ENOENT)
4253		return -EIO;
4254	else
4255		ret = 0;
4256
4257	if (VXGE_FW_VER(VXGE_CERT_FW_VER_MAJOR, VXGE_CERT_FW_VER_MINOR, 0) >
4258	    VXGE_FW_VER(maj, min, 0)) {
4259		vxge_debug_init(VXGE_ERR, "%s: Firmware %d.%d.%d is too old to"
4260				" be used with this driver.",
4261				VXGE_DRIVER_NAME, maj, min, bld);
4262		return -EINVAL;
4263	}
4264
4265	return ret;
4266}
4267
4268static int is_sriov_initialized(struct pci_dev *pdev)
4269{
4270	int pos;
4271	u16 ctrl;
4272
4273	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
4274	if (pos) {
4275		pci_read_config_word(pdev, pos + PCI_SRIOV_CTRL, &ctrl);
4276		if (ctrl & PCI_SRIOV_CTRL_VFE)
4277			return 1;
4278	}
4279	return 0;
4280}
4281
4282static const struct vxge_hw_uld_cbs vxge_callbacks = {
4283	.link_up = vxge_callback_link_up,
4284	.link_down = vxge_callback_link_down,
4285	.crit_err = vxge_callback_crit_err,
4286};
4287
4288/**
4289 * vxge_probe
4290 * @pdev : structure containing the PCI related information of the device.
4291 * @pre: List of PCI devices supported by the driver listed in vxge_id_table.
4292 * Description:
4293 * This function is called when a new PCI device gets detected and initializes
4294 * it.
4295 * Return value:
4296 * returns 0 on success and negative on failure.
4297 *
4298 */
4299static int
4300vxge_probe(struct pci_dev *pdev, const struct pci_device_id *pre)
4301{
4302	struct __vxge_hw_device *hldev;
4303	enum vxge_hw_status status;
4304	int ret;
4305	int high_dma = 0;
4306	u64 vpath_mask = 0;
4307	struct vxgedev *vdev;
4308	struct vxge_config *ll_config = NULL;
4309	struct vxge_hw_device_config *device_config = NULL;
4310	struct vxge_hw_device_attr attr;
4311	int i, j, no_of_vpath = 0, max_vpath_supported = 0;
4312	u8 *macaddr;
4313	struct vxge_mac_addrs *entry;
4314	static int bus = -1, device = -1;
4315	u32 host_type;
4316	u8 new_device = 0;
4317	enum vxge_hw_status is_privileged;
4318	u32 function_mode;
4319	u32 num_vfs = 0;
4320
4321	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
4322	attr.pdev = pdev;
4323
4324	/* In SRIOV-17 mode, functions of the same adapter
4325	 * can be deployed on different buses
4326	 */
4327	if (((bus != pdev->bus->number) || (device != PCI_SLOT(pdev->devfn))) &&
4328	    !pdev->is_virtfn)
4329		new_device = 1;
4330
4331	bus = pdev->bus->number;
4332	device = PCI_SLOT(pdev->devfn);
4333
4334	if (new_device) {
4335		if (driver_config->config_dev_cnt &&
4336		   (driver_config->config_dev_cnt !=
4337			driver_config->total_dev_cnt))
4338			vxge_debug_init(VXGE_ERR,
4339				"%s: Configured %d of %d devices",
4340				VXGE_DRIVER_NAME,
4341				driver_config->config_dev_cnt,
4342				driver_config->total_dev_cnt);
4343		driver_config->config_dev_cnt = 0;
4344		driver_config->total_dev_cnt = 0;
4345	}
4346
4347	/* Now making the CPU based no of vpath calculation
4348	 * applicable for individual functions as well.
4349	 */
4350	driver_config->g_no_cpus = 0;
4351	driver_config->vpath_per_dev = max_config_vpath;
4352
4353	driver_config->total_dev_cnt++;
4354	if (++driver_config->config_dev_cnt > max_config_dev) {
4355		ret = 0;
4356		goto _exit0;
4357	}
4358
4359	device_config = kzalloc(sizeof(struct vxge_hw_device_config),
4360		GFP_KERNEL);
4361	if (!device_config) {
4362		ret = -ENOMEM;
4363		vxge_debug_init(VXGE_ERR,
4364			"device_config : malloc failed %s %d",
4365			__FILE__, __LINE__);
4366		goto _exit0;
4367	}
4368
4369	ll_config = kzalloc(sizeof(struct vxge_config), GFP_KERNEL);
4370	if (!ll_config) {
4371		ret = -ENOMEM;
4372		vxge_debug_init(VXGE_ERR,
4373			"device_config : malloc failed %s %d",
4374			__FILE__, __LINE__);
4375		goto _exit0;
4376	}
4377	ll_config->tx_steering_type = TX_MULTIQ_STEERING;
4378	ll_config->intr_type = MSI_X;
4379	ll_config->napi_weight = NEW_NAPI_WEIGHT;
4380	ll_config->rth_steering = RTH_STEERING;
4381
4382	/* get the default configuration parameters */
4383	vxge_hw_device_config_default_get(device_config);
4384
4385	/* initialize configuration parameters */
4386	vxge_device_config_init(device_config, &ll_config->intr_type);
4387
4388	ret = pci_enable_device(pdev);
4389	if (ret) {
4390		vxge_debug_init(VXGE_ERR,
4391			"%s : can not enable PCI device", __func__);
4392		goto _exit0;
4393	}
4394
4395	if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(64))) {
4396		vxge_debug_ll_config(VXGE_TRACE,
4397			"%s : using 64bit DMA", __func__);
4398
4399		high_dma = 1;
4400
4401		if (dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64))) {
4402			vxge_debug_init(VXGE_ERR,
4403				"%s : unable to obtain 64bit DMA for "
4404				"consistent allocations", __func__);
4405			ret = -ENOMEM;
4406			goto _exit1;
4407		}
4408	} else if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(32))) {
4409		vxge_debug_ll_config(VXGE_TRACE,
4410			"%s : using 32bit DMA", __func__);
4411	} else {
4412		ret = -ENOMEM;
4413		goto _exit1;
4414	}
4415
4416	ret = pci_request_region(pdev, 0, VXGE_DRIVER_NAME);
4417	if (ret) {
4418		vxge_debug_init(VXGE_ERR,
4419			"%s : request regions failed", __func__);
4420		goto _exit1;
4421	}
4422
4423	pci_set_master(pdev);
4424
4425	attr.bar0 = pci_ioremap_bar(pdev, 0);
4426	if (!attr.bar0) {
4427		vxge_debug_init(VXGE_ERR,
4428			"%s : cannot remap io memory bar0", __func__);
4429		ret = -ENODEV;
4430		goto _exit2;
4431	}
4432	vxge_debug_ll_config(VXGE_TRACE,
4433		"pci ioremap bar0: %p:0x%llx",
4434		attr.bar0,
4435		(unsigned long long)pci_resource_start(pdev, 0));
4436
4437	status = vxge_hw_device_hw_info_get(attr.bar0,
4438			&ll_config->device_hw_info);
4439	if (status != VXGE_HW_OK) {
4440		vxge_debug_init(VXGE_ERR,
4441			"%s: Reading of hardware info failed."
4442			"Please try upgrading the firmware.", VXGE_DRIVER_NAME);
4443		ret = -EINVAL;
4444		goto _exit3;
4445	}
4446
4447	vpath_mask = ll_config->device_hw_info.vpath_mask;
4448	if (vpath_mask == 0) {
4449		vxge_debug_ll_config(VXGE_TRACE,
4450			"%s: No vpaths available in device", VXGE_DRIVER_NAME);
4451		ret = -EINVAL;
4452		goto _exit3;
4453	}
4454
4455	vxge_debug_ll_config(VXGE_TRACE,
4456		"%s:%d  Vpath mask = %llx", __func__, __LINE__,
4457		(unsigned long long)vpath_mask);
4458
4459	function_mode = ll_config->device_hw_info.function_mode;
4460	host_type = ll_config->device_hw_info.host_type;
4461	is_privileged = __vxge_hw_device_is_privilaged(host_type,
4462		ll_config->device_hw_info.func_id);
4463
4464	/* Check how many vpaths are available */
4465	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
4466		if (!((vpath_mask) & vxge_mBIT(i)))
4467			continue;
4468		max_vpath_supported++;
4469	}
4470
4471	if (new_device)
4472		num_vfs = vxge_get_num_vfs(function_mode) - 1;
4473
4474	/* Enable SRIOV mode, if firmware has SRIOV support and if it is a PF */
4475	if (is_sriov(function_mode) && !is_sriov_initialized(pdev) &&
4476	   (ll_config->intr_type != INTA)) {
4477		ret = pci_enable_sriov(pdev, num_vfs);
4478		if (ret)
4479			vxge_debug_ll_config(VXGE_ERR,
4480				"Failed in enabling SRIOV mode: %d\n", ret);
4481			/* No need to fail out, as an error here is non-fatal */
4482	}
4483
4484	/*
4485	 * Configure vpaths and get driver configured number of vpaths
4486	 * which is less than or equal to the maximum vpaths per function.
4487	 */
4488	no_of_vpath = vxge_config_vpaths(device_config, vpath_mask, ll_config);
4489	if (!no_of_vpath) {
4490		vxge_debug_ll_config(VXGE_ERR,
4491			"%s: No more vpaths to configure", VXGE_DRIVER_NAME);
4492		ret = 0;
4493		goto _exit3;
4494	}
4495
4496	/* Setting driver callbacks */
4497	attr.uld_callbacks = &vxge_callbacks;
4498
4499	status = vxge_hw_device_initialize(&hldev, &attr, device_config);
4500	if (status != VXGE_HW_OK) {
4501		vxge_debug_init(VXGE_ERR,
4502			"Failed to initialize device (%d)", status);
4503		ret = -EINVAL;
4504		goto _exit3;
4505	}
4506
4507	if (VXGE_FW_VER(ll_config->device_hw_info.fw_version.major,
4508			ll_config->device_hw_info.fw_version.minor,
4509			ll_config->device_hw_info.fw_version.build) >=
4510	    VXGE_EPROM_FW_VER) {
4511		struct eprom_image img[VXGE_HW_MAX_ROM_IMAGES];
4512
4513		status = vxge_hw_vpath_eprom_img_ver_get(hldev, img);
4514		if (status != VXGE_HW_OK) {
4515			vxge_debug_init(VXGE_ERR, "%s: Reading of EPROM failed",
4516					VXGE_DRIVER_NAME);
4517			/* This is a non-fatal error, continue */
4518		}
4519
4520		for (i = 0; i < VXGE_HW_MAX_ROM_IMAGES; i++) {
4521			hldev->eprom_versions[i] = img[i].version;
4522			if (!img[i].is_valid)
4523				break;
4524			vxge_debug_init(VXGE_TRACE, "%s: EPROM %d, version "
4525					"%d.%d.%d.%d", VXGE_DRIVER_NAME, i,
4526					VXGE_EPROM_IMG_MAJOR(img[i].version),
4527					VXGE_EPROM_IMG_MINOR(img[i].version),
4528					VXGE_EPROM_IMG_FIX(img[i].version),
4529					VXGE_EPROM_IMG_BUILD(img[i].version));
4530		}
4531	}
4532
4533	/* if FCS stripping is not disabled in MAC fail driver load */
4534	status = vxge_hw_vpath_strip_fcs_check(hldev, vpath_mask);
4535	if (status != VXGE_HW_OK) {
4536		vxge_debug_init(VXGE_ERR, "%s: FCS stripping is enabled in MAC"
4537				" failing driver load", VXGE_DRIVER_NAME);
4538		ret = -EINVAL;
4539		goto _exit4;
4540	}
4541
4542	/* Always enable HWTS.  This will always cause the FCS to be invalid,
4543	 * due to the fact that HWTS is using the FCS as the location of the
4544	 * timestamp.  The HW FCS checking will still correctly determine if
4545	 * there is a valid checksum, and the FCS is being removed by the driver
4546	 * anyway.  So no functionality is being lost.  Since it is always
4547	 * enabled, we now simply use the ioctl call to set whether or not the
4548	 * driver should be paying attention to the HWTS.
4549	 */
4550	if (is_privileged == VXGE_HW_OK) {
4551		status = vxge_timestamp_config(hldev);
4552		if (status != VXGE_HW_OK) {
4553			vxge_debug_init(VXGE_ERR, "%s: HWTS enable failed",
4554					VXGE_DRIVER_NAME);
4555			ret = -EFAULT;
4556			goto _exit4;
4557		}
4558	}
4559
4560	vxge_hw_device_debug_set(hldev, VXGE_ERR, VXGE_COMPONENT_LL);
4561
4562	/* set private device info */
4563	pci_set_drvdata(pdev, hldev);
4564
4565	ll_config->fifo_indicate_max_pkts = VXGE_FIFO_INDICATE_MAX_PKTS;
4566	ll_config->addr_learn_en = addr_learn_en;
4567	ll_config->rth_algorithm = RTH_ALG_JENKINS;
4568	ll_config->rth_hash_type_tcpipv4 = 1;
4569	ll_config->rth_hash_type_ipv4 = 0;
4570	ll_config->rth_hash_type_tcpipv6 = 0;
4571	ll_config->rth_hash_type_ipv6 = 0;
4572	ll_config->rth_hash_type_tcpipv6ex = 0;
4573	ll_config->rth_hash_type_ipv6ex = 0;
4574	ll_config->rth_bkt_sz = RTH_BUCKET_SIZE;
4575	ll_config->tx_pause_enable = VXGE_PAUSE_CTRL_ENABLE;
4576	ll_config->rx_pause_enable = VXGE_PAUSE_CTRL_ENABLE;
4577
4578	ret = vxge_device_register(hldev, ll_config, high_dma, no_of_vpath,
4579				   &vdev);
4580	if (ret) {
4581		ret = -EINVAL;
4582		goto _exit4;
4583	}
4584
4585	ret = vxge_probe_fw_update(vdev);
4586	if (ret)
4587		goto _exit5;
4588
4589	vxge_hw_device_debug_set(hldev, VXGE_TRACE, VXGE_COMPONENT_LL);
4590	VXGE_COPY_DEBUG_INFO_TO_LL(vdev, vxge_hw_device_error_level_get(hldev),
4591		vxge_hw_device_trace_level_get(hldev));
4592
4593	/* set private HW device info */
4594	vdev->mtu = VXGE_HW_DEFAULT_MTU;
4595	vdev->bar0 = attr.bar0;
4596	vdev->max_vpath_supported = max_vpath_supported;
4597	vdev->no_of_vpath = no_of_vpath;
4598
4599	/* Virtual Path count */
4600	for (i = 0, j = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
4601		if (!vxge_bVALn(vpath_mask, i, 1))
4602			continue;
4603		if (j >= vdev->no_of_vpath)
4604			break;
4605
4606		vdev->vpaths[j].is_configured = 1;
4607		vdev->vpaths[j].device_id = i;
4608		vdev->vpaths[j].ring.driver_id = j;
4609		vdev->vpaths[j].vdev = vdev;
4610		vdev->vpaths[j].max_mac_addr_cnt = max_mac_vpath;
4611		memcpy((u8 *)vdev->vpaths[j].macaddr,
4612				ll_config->device_hw_info.mac_addrs[i],
4613				ETH_ALEN);
4614
4615		/* Initialize the mac address list header */
4616		INIT_LIST_HEAD(&vdev->vpaths[j].mac_addr_list);
4617
4618		vdev->vpaths[j].mac_addr_cnt = 0;
4619		vdev->vpaths[j].mcast_addr_cnt = 0;
4620		j++;
4621	}
4622	vdev->exec_mode = VXGE_EXEC_MODE_DISABLE;
4623	vdev->max_config_port = max_config_port;
4624
4625	vdev->vlan_tag_strip = vlan_tag_strip;
4626
4627	/* map the hashing selector table to the configured vpaths */
4628	for (i = 0; i < vdev->no_of_vpath; i++)
4629		vdev->vpath_selector[i] = vpath_selector[i];
4630
4631	macaddr = (u8 *)vdev->vpaths[0].macaddr;
4632
4633	ll_config->device_hw_info.serial_number[VXGE_HW_INFO_LEN - 1] = '\0';
4634	ll_config->device_hw_info.product_desc[VXGE_HW_INFO_LEN - 1] = '\0';
4635	ll_config->device_hw_info.part_number[VXGE_HW_INFO_LEN - 1] = '\0';
4636
4637	vxge_debug_init(VXGE_TRACE, "%s: SERIAL NUMBER: %s",
4638		vdev->ndev->name, ll_config->device_hw_info.serial_number);
4639
4640	vxge_debug_init(VXGE_TRACE, "%s: PART NUMBER: %s",
4641		vdev->ndev->name, ll_config->device_hw_info.part_number);
4642
4643	vxge_debug_init(VXGE_TRACE, "%s: Neterion %s Server Adapter",
4644		vdev->ndev->name, ll_config->device_hw_info.product_desc);
4645
4646	vxge_debug_init(VXGE_TRACE, "%s: MAC ADDR: %pM",
4647		vdev->ndev->name, macaddr);
4648
4649	vxge_debug_init(VXGE_TRACE, "%s: Link Width x%d",
4650		vdev->ndev->name, vxge_hw_device_link_width_get(hldev));
4651
4652	vxge_debug_init(VXGE_TRACE,
4653		"%s: Firmware version : %s Date : %s", vdev->ndev->name,
4654		ll_config->device_hw_info.fw_version.version,
4655		ll_config->device_hw_info.fw_date.date);
4656
4657	if (new_device) {
4658		switch (ll_config->device_hw_info.function_mode) {
4659		case VXGE_HW_FUNCTION_MODE_SINGLE_FUNCTION:
4660			vxge_debug_init(VXGE_TRACE,
4661			"%s: Single Function Mode Enabled", vdev->ndev->name);
4662		break;
4663		case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION:
4664			vxge_debug_init(VXGE_TRACE,
4665			"%s: Multi Function Mode Enabled", vdev->ndev->name);
4666		break;
4667		case VXGE_HW_FUNCTION_MODE_SRIOV:
4668			vxge_debug_init(VXGE_TRACE,
4669			"%s: Single Root IOV Mode Enabled", vdev->ndev->name);
4670		break;
4671		case VXGE_HW_FUNCTION_MODE_MRIOV:
4672			vxge_debug_init(VXGE_TRACE,
4673			"%s: Multi Root IOV Mode Enabled", vdev->ndev->name);
4674		break;
4675		}
4676	}
4677
4678	vxge_print_parm(vdev, vpath_mask);
4679
4680	/* Store the fw version for ethttool option */
4681	strcpy(vdev->fw_version, ll_config->device_hw_info.fw_version.version);
4682	memcpy(vdev->ndev->dev_addr, (u8 *)vdev->vpaths[0].macaddr, ETH_ALEN);
4683
4684	/* Copy the station mac address to the list */
4685	for (i = 0; i < vdev->no_of_vpath; i++) {
4686		entry =	kzalloc(sizeof(struct vxge_mac_addrs), GFP_KERNEL);
4687		if (NULL == entry) {
4688			vxge_debug_init(VXGE_ERR,
4689				"%s: mac_addr_list : memory allocation failed",
4690				vdev->ndev->name);
4691			ret = -EPERM;
4692			goto _exit6;
4693		}
4694		macaddr = (u8 *)&entry->macaddr;
4695		memcpy(macaddr, vdev->ndev->dev_addr, ETH_ALEN);
4696		list_add(&entry->item, &vdev->vpaths[i].mac_addr_list);
4697		vdev->vpaths[i].mac_addr_cnt = 1;
4698	}
4699
4700	kfree(device_config);
4701
4702	/*
4703	 * INTA is shared in multi-function mode. This is unlike the INTA
4704	 * implementation in MR mode, where each VH has its own INTA message.
4705	 * - INTA is masked (disabled) as long as at least one function sets
4706	 * its TITAN_MASK_ALL_INT.ALARM bit.
4707	 * - INTA is unmasked (enabled) when all enabled functions have cleared
4708	 * their own TITAN_MASK_ALL_INT.ALARM bit.
4709	 * The TITAN_MASK_ALL_INT ALARM & TRAFFIC bits are cleared on power up.
4710	 * Though this driver leaves the top level interrupts unmasked while
4711	 * leaving the required module interrupt bits masked on exit, there
4712	 * could be a rougue driver around that does not follow this procedure
4713	 * resulting in a failure to generate interrupts. The following code is
4714	 * present to prevent such a failure.
4715	 */
4716
4717	if (ll_config->device_hw_info.function_mode ==
4718		VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION)
4719		if (vdev->config.intr_type == INTA)
4720			vxge_hw_device_unmask_all(hldev);
4721
4722	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d  Exiting...",
4723		vdev->ndev->name, __func__, __LINE__);
4724
4725	vxge_hw_device_debug_set(hldev, VXGE_ERR, VXGE_COMPONENT_LL);
4726	VXGE_COPY_DEBUG_INFO_TO_LL(vdev, vxge_hw_device_error_level_get(hldev),
4727		vxge_hw_device_trace_level_get(hldev));
4728
4729	kfree(ll_config);
4730	return 0;
4731
4732_exit6:
4733	for (i = 0; i < vdev->no_of_vpath; i++)
4734		vxge_free_mac_add_list(&vdev->vpaths[i]);
4735_exit5:
4736	vxge_device_unregister(hldev);
4737_exit4:
4738	vxge_hw_device_terminate(hldev);
4739	pci_disable_sriov(pdev);
4740_exit3:
4741	iounmap(attr.bar0);
4742_exit2:
4743	pci_release_region(pdev, 0);
4744_exit1:
4745	pci_disable_device(pdev);
4746_exit0:
4747	kfree(ll_config);
4748	kfree(device_config);
4749	driver_config->config_dev_cnt--;
4750	driver_config->total_dev_cnt--;
4751	return ret;
4752}
4753
4754/**
4755 * vxge_rem_nic - Free the PCI device
4756 * @pdev: structure containing the PCI related information of the device.
4757 * Description: This function is called by the Pci subsystem to release a
4758 * PCI device and free up all resource held up by the device.
4759 */
4760static void vxge_remove(struct pci_dev *pdev)
4761{
4762	struct __vxge_hw_device *hldev;
4763	struct vxgedev *vdev;
4764	int i;
4765
4766	hldev = pci_get_drvdata(pdev);
4767	if (hldev == NULL)
4768		return;
4769
4770	vdev = netdev_priv(hldev->ndev);
4771
4772	vxge_debug_entryexit(vdev->level_trace,	"%s:%d", __func__, __LINE__);
4773	vxge_debug_init(vdev->level_trace, "%s : removing PCI device...",
4774			__func__);
4775
4776	for (i = 0; i < vdev->no_of_vpath; i++)
4777		vxge_free_mac_add_list(&vdev->vpaths[i]);
4778
4779	vxge_device_unregister(hldev);
4780	/* Do not call pci_disable_sriov here, as it will break child devices */
4781	vxge_hw_device_terminate(hldev);
4782	iounmap(vdev->bar0);
4783	pci_release_region(pdev, 0);
4784	pci_disable_device(pdev);
4785	driver_config->config_dev_cnt--;
4786	driver_config->total_dev_cnt--;
4787
4788	vxge_debug_init(vdev->level_trace, "%s:%d Device unregistered",
4789			__func__, __LINE__);
4790	vxge_debug_entryexit(vdev->level_trace,	"%s:%d  Exiting...", __func__,
4791			     __LINE__);
4792}
4793
4794static const struct pci_error_handlers vxge_err_handler = {
4795	.error_detected = vxge_io_error_detected,
4796	.slot_reset = vxge_io_slot_reset,
4797	.resume = vxge_io_resume,
4798};
4799
4800static SIMPLE_DEV_PM_OPS(vxge_pm_ops, vxge_pm_suspend, vxge_pm_resume);
4801
4802static struct pci_driver vxge_driver = {
4803	.name = VXGE_DRIVER_NAME,
4804	.id_table = vxge_id_table,
4805	.probe = vxge_probe,
4806	.remove = vxge_remove,
4807	.driver.pm = &vxge_pm_ops,
4808	.err_handler = &vxge_err_handler,
4809};
4810
4811static int __init
4812vxge_starter(void)
4813{
4814	int ret = 0;
4815
4816	pr_info("Copyright(c) 2002-2010 Exar Corp.\n");
4817	pr_info("Driver version: %s\n", DRV_VERSION);
4818
4819	verify_bandwidth();
4820
4821	driver_config = kzalloc(sizeof(struct vxge_drv_config), GFP_KERNEL);
4822	if (!driver_config)
4823		return -ENOMEM;
4824
4825	ret = pci_register_driver(&vxge_driver);
4826	if (ret) {
4827		kfree(driver_config);
4828		goto err;
4829	}
4830
4831	if (driver_config->config_dev_cnt &&
4832	   (driver_config->config_dev_cnt != driver_config->total_dev_cnt))
4833		vxge_debug_init(VXGE_ERR,
4834			"%s: Configured %d of %d devices",
4835			VXGE_DRIVER_NAME, driver_config->config_dev_cnt,
4836			driver_config->total_dev_cnt);
4837err:
4838	return ret;
4839}
4840
4841static void __exit
4842vxge_closer(void)
4843{
4844	pci_unregister_driver(&vxge_driver);
4845	kfree(driver_config);
4846}
4847module_init(vxge_starter);
4848module_exit(vxge_closer);
4849