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 128
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_kfree_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 = pci_map_single(ring->pdev, rx_priv->skb_data,
245				rx_priv->data_size, PCI_DMA_FROMDEVICE);
246
247	if (unlikely(pci_dma_mapping_error(ring->pdev, 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	pci_dma_sync_single_for_device(ring->pdev,
327		rx_priv->data_dma, rx_priv->data_size, PCI_DMA_FROMDEVICE);
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					pci_unmap_single(ring->pdev, data_dma,
429						data_size, PCI_DMA_FROMDEVICE);
430
431					vxge_hw_ring_rxd_pre_post(ringh, dtr);
432					vxge_post(&dtr_cnt, &first_dtr, dtr,
433						ringh);
434				} else {
435					dev_kfree_skb(rx_priv->skb);
436					rx_priv->skb = skb;
437					rx_priv->data_size = data_size;
438					vxge_re_pre_post(dtr, ring, rx_priv);
439
440					vxge_post(&dtr_cnt, &first_dtr, dtr,
441						ringh);
442					ring->stats.rx_dropped++;
443					break;
444				}
445			} else {
446				vxge_re_pre_post(dtr, ring, rx_priv);
447
448				vxge_post(&dtr_cnt, &first_dtr, dtr, ringh);
449				ring->stats.rx_dropped++;
450				break;
451			}
452		} else {
453			struct sk_buff *skb_up;
454
455			skb_up = netdev_alloc_skb(dev, pkt_length +
456				VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
457			if (skb_up != NULL) {
458				skb_reserve(skb_up,
459				    VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
460
461				pci_dma_sync_single_for_cpu(ring->pdev,
462					data_dma, data_size,
463					PCI_DMA_FROMDEVICE);
464
465				vxge_debug_mem(VXGE_TRACE,
466					"%s: %s:%d  skb_up = %p",
467					ring->ndev->name, __func__,
468					__LINE__, skb);
469				memcpy(skb_up->data, skb->data, pkt_length);
470
471				vxge_re_pre_post(dtr, ring, rx_priv);
472
473				vxge_post(&dtr_cnt, &first_dtr, dtr,
474					ringh);
475				/* will netif_rx small SKB instead */
476				skb = skb_up;
477				skb_put(skb, pkt_length);
478			} else {
479				vxge_re_pre_post(dtr, ring, rx_priv);
480
481				vxge_post(&dtr_cnt, &first_dtr, dtr, ringh);
482				vxge_debug_rx(VXGE_ERR,
483					"%s: vxge_rx_1b_compl: out of "
484					"memory", dev->name);
485				ring->stats.skb_alloc_fail++;
486				break;
487			}
488		}
489
490		if ((ext_info.proto & VXGE_HW_FRAME_PROTO_TCP_OR_UDP) &&
491		    !(ext_info.proto & VXGE_HW_FRAME_PROTO_IP_FRAG) &&
492		    (dev->features & NETIF_F_RXCSUM) && /* Offload Rx side CSUM */
493		    ext_info.l3_cksum == VXGE_HW_L3_CKSUM_OK &&
494		    ext_info.l4_cksum == VXGE_HW_L4_CKSUM_OK)
495			skb->ip_summed = CHECKSUM_UNNECESSARY;
496		else
497			skb_checksum_none_assert(skb);
498
499
500		if (ring->rx_hwts) {
501			struct skb_shared_hwtstamps *skb_hwts;
502			u32 ns = *(u32 *)(skb->head + pkt_length);
503
504			skb_hwts = skb_hwtstamps(skb);
505			skb_hwts->hwtstamp = ns_to_ktime(ns);
506		}
507
508		/* rth_hash_type and rth_it_hit are non-zero regardless of
509		 * whether rss is enabled.  Only the rth_value is zero/non-zero
510		 * if rss is disabled/enabled, so key off of that.
511		 */
512		if (ext_info.rth_value)
513			skb_set_hash(skb, ext_info.rth_value,
514				     PKT_HASH_TYPE_L3);
515
516		vxge_rx_complete(ring, skb, ext_info.vlan,
517			pkt_length, &ext_info);
518
519		ring->budget--;
520		ring->pkts_processed++;
521		if (!ring->budget)
522			break;
523
524	} while (vxge_hw_ring_rxd_next_completed(ringh, &dtr,
525		&t_code) == VXGE_HW_OK);
526
527	if (first_dtr)
528		vxge_hw_ring_rxd_post_post_wmb(ringh, first_dtr);
529
530out:
531	vxge_debug_entryexit(VXGE_TRACE,
532				"%s:%d  Exiting...",
533				__func__, __LINE__);
534	return VXGE_HW_OK;
535}
536
537/*
538 * vxge_xmit_compl
539 *
540 * If an interrupt was raised to indicate DMA complete of the Tx packet,
541 * this function is called. It identifies the last TxD whose buffer was
542 * freed and frees all skbs whose data have already DMA'ed into the NICs
543 * internal memory.
544 */
545static enum vxge_hw_status
546vxge_xmit_compl(struct __vxge_hw_fifo *fifo_hw, void *dtr,
547		enum vxge_hw_fifo_tcode t_code, void *userdata,
548		struct sk_buff ***skb_ptr, int nr_skb, int *more)
549{
550	struct vxge_fifo *fifo = (struct vxge_fifo *)userdata;
551	struct sk_buff *skb, **done_skb = *skb_ptr;
552	int pkt_cnt = 0;
553
554	vxge_debug_entryexit(VXGE_TRACE,
555		"%s:%d Entered....", __func__, __LINE__);
556
557	do {
558		int frg_cnt;
559		skb_frag_t *frag;
560		int i = 0, j;
561		struct vxge_tx_priv *txd_priv =
562			vxge_hw_fifo_txdl_private_get(dtr);
563
564		skb = txd_priv->skb;
565		frg_cnt = skb_shinfo(skb)->nr_frags;
566		frag = &skb_shinfo(skb)->frags[0];
567
568		vxge_debug_tx(VXGE_TRACE,
569				"%s: %s:%d fifo_hw = %p dtr = %p "
570				"tcode = 0x%x", fifo->ndev->name, __func__,
571				__LINE__, fifo_hw, dtr, t_code);
572		/* check skb validity */
573		vxge_assert(skb);
574		vxge_debug_tx(VXGE_TRACE,
575			"%s: %s:%d skb = %p itxd_priv = %p frg_cnt = %d",
576			fifo->ndev->name, __func__, __LINE__,
577			skb, txd_priv, frg_cnt);
578		if (unlikely(t_code)) {
579			fifo->stats.tx_errors++;
580			vxge_debug_tx(VXGE_ERR,
581				"%s: tx: dtr %p completed due to "
582				"error t_code %01x", fifo->ndev->name,
583				dtr, t_code);
584			vxge_hw_fifo_handle_tcode(fifo_hw, dtr, t_code);
585		}
586
587		/*  for unfragmented skb */
588		pci_unmap_single(fifo->pdev, txd_priv->dma_buffers[i++],
589				skb_headlen(skb), PCI_DMA_TODEVICE);
590
591		for (j = 0; j < frg_cnt; j++) {
592			pci_unmap_page(fifo->pdev,
593					txd_priv->dma_buffers[i++],
594					skb_frag_size(frag), PCI_DMA_TODEVICE);
595			frag += 1;
596		}
597
598		vxge_hw_fifo_txdl_free(fifo_hw, dtr);
599
600		/* Updating the statistics block */
601		u64_stats_update_begin(&fifo->stats.syncp);
602		fifo->stats.tx_frms++;
603		fifo->stats.tx_bytes += skb->len;
604		u64_stats_update_end(&fifo->stats.syncp);
605
606		*done_skb++ = skb;
607
608		if (--nr_skb <= 0) {
609			*more = 1;
610			break;
611		}
612
613		pkt_cnt++;
614		if (pkt_cnt > fifo->indicate_max_pkts)
615			break;
616
617	} while (vxge_hw_fifo_txdl_next_completed(fifo_hw,
618				&dtr, &t_code) == VXGE_HW_OK);
619
620	*skb_ptr = done_skb;
621	if (netif_tx_queue_stopped(fifo->txq))
622		netif_tx_wake_queue(fifo->txq);
623
624	vxge_debug_entryexit(VXGE_TRACE,
625				"%s: %s:%d  Exiting...",
626				fifo->ndev->name, __func__, __LINE__);
627	return VXGE_HW_OK;
628}
629
630/* select a vpath to transmit the packet */
631static u32 vxge_get_vpath_no(struct vxgedev *vdev, struct sk_buff *skb)
632{
633	u16 queue_len, counter = 0;
634	if (skb->protocol == htons(ETH_P_IP)) {
635		struct iphdr *ip;
636		struct tcphdr *th;
637
638		ip = ip_hdr(skb);
639
640		if (!ip_is_fragment(ip)) {
641			th = (struct tcphdr *)(((unsigned char *)ip) +
642					ip->ihl*4);
643
644			queue_len = vdev->no_of_vpath;
645			counter = (ntohs(th->source) +
646				ntohs(th->dest)) &
647				vdev->vpath_selector[queue_len - 1];
648			if (counter >= queue_len)
649				counter = queue_len - 1;
650		}
651	}
652	return counter;
653}
654
655static enum vxge_hw_status vxge_search_mac_addr_in_list(
656	struct vxge_vpath *vpath, u64 del_mac)
657{
658	struct list_head *entry, *next;
659	list_for_each_safe(entry, next, &vpath->mac_addr_list) {
660		if (((struct vxge_mac_addrs *)entry)->macaddr == del_mac)
661			return TRUE;
662	}
663	return FALSE;
664}
665
666static int vxge_mac_list_add(struct vxge_vpath *vpath, struct macInfo *mac)
667{
668	struct vxge_mac_addrs *new_mac_entry;
669	u8 *mac_address = NULL;
670
671	if (vpath->mac_addr_cnt >= VXGE_MAX_LEARN_MAC_ADDR_CNT)
672		return TRUE;
673
674	new_mac_entry = kzalloc(sizeof(struct vxge_mac_addrs), GFP_ATOMIC);
675	if (!new_mac_entry) {
676		vxge_debug_mem(VXGE_ERR,
677			"%s: memory allocation failed",
678			VXGE_DRIVER_NAME);
679		return FALSE;
680	}
681
682	list_add(&new_mac_entry->item, &vpath->mac_addr_list);
683
684	/* Copy the new mac address to the list */
685	mac_address = (u8 *)&new_mac_entry->macaddr;
686	memcpy(mac_address, mac->macaddr, ETH_ALEN);
687
688	new_mac_entry->state = mac->state;
689	vpath->mac_addr_cnt++;
690
691	if (is_multicast_ether_addr(mac->macaddr))
692		vpath->mcast_addr_cnt++;
693
694	return TRUE;
695}
696
697/* Add a mac address to DA table */
698static enum vxge_hw_status
699vxge_add_mac_addr(struct vxgedev *vdev, struct macInfo *mac)
700{
701	enum vxge_hw_status status = VXGE_HW_OK;
702	struct vxge_vpath *vpath;
703	enum vxge_hw_vpath_mac_addr_add_mode duplicate_mode;
704
705	if (is_multicast_ether_addr(mac->macaddr))
706		duplicate_mode = VXGE_HW_VPATH_MAC_ADDR_ADD_DUPLICATE;
707	else
708		duplicate_mode = VXGE_HW_VPATH_MAC_ADDR_REPLACE_DUPLICATE;
709
710	vpath = &vdev->vpaths[mac->vpath_no];
711	status = vxge_hw_vpath_mac_addr_add(vpath->handle, mac->macaddr,
712						mac->macmask, duplicate_mode);
713	if (status != VXGE_HW_OK) {
714		vxge_debug_init(VXGE_ERR,
715			"DA config add entry failed for vpath:%d",
716			vpath->device_id);
717	} else
718		if (FALSE == vxge_mac_list_add(vpath, mac))
719			status = -EPERM;
720
721	return status;
722}
723
724static int vxge_learn_mac(struct vxgedev *vdev, u8 *mac_header)
725{
726	struct macInfo mac_info;
727	u8 *mac_address = NULL;
728	u64 mac_addr = 0, vpath_vector = 0;
729	int vpath_idx = 0;
730	enum vxge_hw_status status = VXGE_HW_OK;
731	struct vxge_vpath *vpath = NULL;
732
733	mac_address = (u8 *)&mac_addr;
734	memcpy(mac_address, mac_header, ETH_ALEN);
735
736	/* Is this mac address already in the list? */
737	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
738		vpath = &vdev->vpaths[vpath_idx];
739		if (vxge_search_mac_addr_in_list(vpath, mac_addr))
740			return vpath_idx;
741	}
742
743	memset(&mac_info, 0, sizeof(struct macInfo));
744	memcpy(mac_info.macaddr, mac_header, ETH_ALEN);
745
746	/* Any vpath has room to add mac address to its da table? */
747	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
748		vpath = &vdev->vpaths[vpath_idx];
749		if (vpath->mac_addr_cnt < vpath->max_mac_addr_cnt) {
750			/* Add this mac address to this vpath */
751			mac_info.vpath_no = vpath_idx;
752			mac_info.state = VXGE_LL_MAC_ADDR_IN_DA_TABLE;
753			status = vxge_add_mac_addr(vdev, &mac_info);
754			if (status != VXGE_HW_OK)
755				return -EPERM;
756			return vpath_idx;
757		}
758	}
759
760	mac_info.state = VXGE_LL_MAC_ADDR_IN_LIST;
761	vpath_idx = 0;
762	mac_info.vpath_no = vpath_idx;
763	/* Is the first vpath already selected as catch-basin ? */
764	vpath = &vdev->vpaths[vpath_idx];
765	if (vpath->mac_addr_cnt > vpath->max_mac_addr_cnt) {
766		/* Add this mac address to this vpath */
767		if (FALSE == vxge_mac_list_add(vpath, &mac_info))
768			return -EPERM;
769		return vpath_idx;
770	}
771
772	/* Select first vpath as catch-basin */
773	vpath_vector = vxge_mBIT(vpath->device_id);
774	status = vxge_hw_mgmt_reg_write(vpath->vdev->devh,
775				vxge_hw_mgmt_reg_type_mrpcim,
776				0,
777				(ulong)offsetof(
778					struct vxge_hw_mrpcim_reg,
779					rts_mgr_cbasin_cfg),
780				vpath_vector);
781	if (status != VXGE_HW_OK) {
782		vxge_debug_tx(VXGE_ERR,
783			"%s: Unable to set the vpath-%d in catch-basin mode",
784			VXGE_DRIVER_NAME, vpath->device_id);
785		return -EPERM;
786	}
787
788	if (FALSE == vxge_mac_list_add(vpath, &mac_info))
789		return -EPERM;
790
791	return vpath_idx;
792}
793
794/**
795 * vxge_xmit
796 * @skb : the socket buffer containing the Tx data.
797 * @dev : device pointer.
798 *
799 * This function is the Tx entry point of the driver. Neterion NIC supports
800 * certain protocol assist features on Tx side, namely  CSO, S/G, LSO.
801*/
802static netdev_tx_t
803vxge_xmit(struct sk_buff *skb, struct net_device *dev)
804{
805	struct vxge_fifo *fifo = NULL;
806	void *dtr_priv;
807	void *dtr = NULL;
808	struct vxgedev *vdev = NULL;
809	enum vxge_hw_status status;
810	int frg_cnt, first_frg_len;
811	skb_frag_t *frag;
812	int i = 0, j = 0, avail;
813	u64 dma_pointer;
814	struct vxge_tx_priv *txdl_priv = NULL;
815	struct __vxge_hw_fifo *fifo_hw;
816	int offload_type;
817	int vpath_no = 0;
818
819	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
820			dev->name, __func__, __LINE__);
821
822	/* A buffer with no data will be dropped */
823	if (unlikely(skb->len <= 0)) {
824		vxge_debug_tx(VXGE_ERR,
825			"%s: Buffer has no data..", dev->name);
826		dev_kfree_skb_any(skb);
827		return NETDEV_TX_OK;
828	}
829
830	vdev = netdev_priv(dev);
831
832	if (unlikely(!is_vxge_card_up(vdev))) {
833		vxge_debug_tx(VXGE_ERR,
834			"%s: vdev not initialized", dev->name);
835		dev_kfree_skb_any(skb);
836		return NETDEV_TX_OK;
837	}
838
839	if (vdev->config.addr_learn_en) {
840		vpath_no = vxge_learn_mac(vdev, skb->data + ETH_ALEN);
841		if (vpath_no == -EPERM) {
842			vxge_debug_tx(VXGE_ERR,
843				"%s: Failed to store the mac address",
844				dev->name);
845			dev_kfree_skb_any(skb);
846			return NETDEV_TX_OK;
847		}
848	}
849
850	if (vdev->config.tx_steering_type == TX_MULTIQ_STEERING)
851		vpath_no = skb_get_queue_mapping(skb);
852	else if (vdev->config.tx_steering_type == TX_PORT_STEERING)
853		vpath_no = vxge_get_vpath_no(vdev, skb);
854
855	vxge_debug_tx(VXGE_TRACE, "%s: vpath_no= %d", dev->name, vpath_no);
856
857	if (vpath_no >= vdev->no_of_vpath)
858		vpath_no = 0;
859
860	fifo = &vdev->vpaths[vpath_no].fifo;
861	fifo_hw = fifo->handle;
862
863	if (netif_tx_queue_stopped(fifo->txq))
864		return NETDEV_TX_BUSY;
865
866	avail = vxge_hw_fifo_free_txdl_count_get(fifo_hw);
867	if (avail == 0) {
868		vxge_debug_tx(VXGE_ERR,
869			"%s: No free TXDs available", dev->name);
870		fifo->stats.txd_not_free++;
871		goto _exit0;
872	}
873
874	/* Last TXD?  Stop tx queue to avoid dropping packets.  TX
875	 * completion will resume the queue.
876	 */
877	if (avail == 1)
878		netif_tx_stop_queue(fifo->txq);
879
880	status = vxge_hw_fifo_txdl_reserve(fifo_hw, &dtr, &dtr_priv);
881	if (unlikely(status != VXGE_HW_OK)) {
882		vxge_debug_tx(VXGE_ERR,
883		   "%s: Out of descriptors .", dev->name);
884		fifo->stats.txd_out_of_desc++;
885		goto _exit0;
886	}
887
888	vxge_debug_tx(VXGE_TRACE,
889		"%s: %s:%d fifo_hw = %p dtr = %p dtr_priv = %p",
890		dev->name, __func__, __LINE__,
891		fifo_hw, dtr, dtr_priv);
892
893	if (skb_vlan_tag_present(skb)) {
894		u16 vlan_tag = skb_vlan_tag_get(skb);
895		vxge_hw_fifo_txdl_vlan_set(dtr, vlan_tag);
896	}
897
898	first_frg_len = skb_headlen(skb);
899
900	dma_pointer = pci_map_single(fifo->pdev, skb->data, first_frg_len,
901				PCI_DMA_TODEVICE);
902
903	if (unlikely(pci_dma_mapping_error(fifo->pdev, dma_pointer))) {
904		vxge_hw_fifo_txdl_free(fifo_hw, dtr);
905		fifo->stats.pci_map_fail++;
906		goto _exit0;
907	}
908
909	txdl_priv = vxge_hw_fifo_txdl_private_get(dtr);
910	txdl_priv->skb = skb;
911	txdl_priv->dma_buffers[j] = dma_pointer;
912
913	frg_cnt = skb_shinfo(skb)->nr_frags;
914	vxge_debug_tx(VXGE_TRACE,
915			"%s: %s:%d skb = %p txdl_priv = %p "
916			"frag_cnt = %d dma_pointer = 0x%llx", dev->name,
917			__func__, __LINE__, skb, txdl_priv,
918			frg_cnt, (unsigned long long)dma_pointer);
919
920	vxge_hw_fifo_txdl_buffer_set(fifo_hw, dtr, j++, dma_pointer,
921		first_frg_len);
922
923	frag = &skb_shinfo(skb)->frags[0];
924	for (i = 0; i < frg_cnt; i++) {
925		/* ignore 0 length fragment */
926		if (!skb_frag_size(frag))
927			continue;
928
929		dma_pointer = (u64)skb_frag_dma_map(&fifo->pdev->dev, frag,
930						    0, skb_frag_size(frag),
931						    DMA_TO_DEVICE);
932
933		if (unlikely(dma_mapping_error(&fifo->pdev->dev, dma_pointer)))
934			goto _exit2;
935		vxge_debug_tx(VXGE_TRACE,
936			"%s: %s:%d frag = %d dma_pointer = 0x%llx",
937				dev->name, __func__, __LINE__, i,
938				(unsigned long long)dma_pointer);
939
940		txdl_priv->dma_buffers[j] = dma_pointer;
941		vxge_hw_fifo_txdl_buffer_set(fifo_hw, dtr, j++, dma_pointer,
942					skb_frag_size(frag));
943		frag += 1;
944	}
945
946	offload_type = vxge_offload_type(skb);
947
948	if (offload_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)) {
949		int mss = vxge_tcp_mss(skb);
950		if (mss) {
951			vxge_debug_tx(VXGE_TRACE, "%s: %s:%d mss = %d",
952				dev->name, __func__, __LINE__, mss);
953			vxge_hw_fifo_txdl_mss_set(dtr, mss);
954		} else {
955			vxge_assert(skb->len <=
956				dev->mtu + VXGE_HW_MAC_HEADER_MAX_SIZE);
957			vxge_assert(0);
958			goto _exit1;
959		}
960	}
961
962	if (skb->ip_summed == CHECKSUM_PARTIAL)
963		vxge_hw_fifo_txdl_cksum_set_bits(dtr,
964					VXGE_HW_FIFO_TXD_TX_CKO_IPV4_EN |
965					VXGE_HW_FIFO_TXD_TX_CKO_TCP_EN |
966					VXGE_HW_FIFO_TXD_TX_CKO_UDP_EN);
967
968	vxge_hw_fifo_txdl_post(fifo_hw, dtr);
969
970	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d  Exiting...",
971		dev->name, __func__, __LINE__);
972	return NETDEV_TX_OK;
973
974_exit2:
975	vxge_debug_tx(VXGE_TRACE, "%s: pci_map_page failed", dev->name);
976_exit1:
977	j = 0;
978	frag = &skb_shinfo(skb)->frags[0];
979
980	pci_unmap_single(fifo->pdev, txdl_priv->dma_buffers[j++],
981			skb_headlen(skb), PCI_DMA_TODEVICE);
982
983	for (; j < i; j++) {
984		pci_unmap_page(fifo->pdev, txdl_priv->dma_buffers[j],
985			skb_frag_size(frag), PCI_DMA_TODEVICE);
986		frag += 1;
987	}
988
989	vxge_hw_fifo_txdl_free(fifo_hw, dtr);
990_exit0:
991	netif_tx_stop_queue(fifo->txq);
992	dev_kfree_skb_any(skb);
993
994	return NETDEV_TX_OK;
995}
996
997/*
998 * vxge_rx_term
999 *
1000 * Function will be called by hw function to abort all outstanding receive
1001 * descriptors.
1002 */
1003static void
1004vxge_rx_term(void *dtrh, enum vxge_hw_rxd_state state, void *userdata)
1005{
1006	struct vxge_ring *ring = (struct vxge_ring *)userdata;
1007	struct vxge_rx_priv *rx_priv =
1008		vxge_hw_ring_rxd_private_get(dtrh);
1009
1010	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
1011			ring->ndev->name, __func__, __LINE__);
1012	if (state != VXGE_HW_RXD_STATE_POSTED)
1013		return;
1014
1015	pci_unmap_single(ring->pdev, rx_priv->data_dma,
1016		rx_priv->data_size, PCI_DMA_FROMDEVICE);
1017
1018	dev_kfree_skb(rx_priv->skb);
1019	rx_priv->skb_data = NULL;
1020
1021	vxge_debug_entryexit(VXGE_TRACE,
1022		"%s: %s:%d  Exiting...",
1023		ring->ndev->name, __func__, __LINE__);
1024}
1025
1026/*
1027 * vxge_tx_term
1028 *
1029 * Function will be called to abort all outstanding tx descriptors
1030 */
1031static void
1032vxge_tx_term(void *dtrh, enum vxge_hw_txdl_state state, void *userdata)
1033{
1034	struct vxge_fifo *fifo = (struct vxge_fifo *)userdata;
1035	skb_frag_t *frag;
1036	int i = 0, j, frg_cnt;
1037	struct vxge_tx_priv *txd_priv = vxge_hw_fifo_txdl_private_get(dtrh);
1038	struct sk_buff *skb = txd_priv->skb;
1039
1040	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1041
1042	if (state != VXGE_HW_TXDL_STATE_POSTED)
1043		return;
1044
1045	/* check skb validity */
1046	vxge_assert(skb);
1047	frg_cnt = skb_shinfo(skb)->nr_frags;
1048	frag = &skb_shinfo(skb)->frags[0];
1049
1050	/*  for unfragmented skb */
1051	pci_unmap_single(fifo->pdev, txd_priv->dma_buffers[i++],
1052		skb_headlen(skb), PCI_DMA_TODEVICE);
1053
1054	for (j = 0; j < frg_cnt; j++) {
1055		pci_unmap_page(fifo->pdev, txd_priv->dma_buffers[i++],
1056			       skb_frag_size(frag), PCI_DMA_TODEVICE);
1057		frag += 1;
1058	}
1059
1060	dev_kfree_skb(skb);
1061
1062	vxge_debug_entryexit(VXGE_TRACE,
1063		"%s:%d  Exiting...", __func__, __LINE__);
1064}
1065
1066static int vxge_mac_list_del(struct vxge_vpath *vpath, struct macInfo *mac)
1067{
1068	struct list_head *entry, *next;
1069	u64 del_mac = 0;
1070	u8 *mac_address = (u8 *) (&del_mac);
1071
1072	/* Copy the mac address to delete from the list */
1073	memcpy(mac_address, mac->macaddr, ETH_ALEN);
1074
1075	list_for_each_safe(entry, next, &vpath->mac_addr_list) {
1076		if (((struct vxge_mac_addrs *)entry)->macaddr == del_mac) {
1077			list_del(entry);
1078			kfree((struct vxge_mac_addrs *)entry);
1079			vpath->mac_addr_cnt--;
1080
1081			if (is_multicast_ether_addr(mac->macaddr))
1082				vpath->mcast_addr_cnt--;
1083			return TRUE;
1084		}
1085	}
1086
1087	return FALSE;
1088}
1089
1090/* delete a mac address from DA table */
1091static enum vxge_hw_status
1092vxge_del_mac_addr(struct vxgedev *vdev, struct macInfo *mac)
1093{
1094	enum vxge_hw_status status = VXGE_HW_OK;
1095	struct vxge_vpath *vpath;
1096
1097	vpath = &vdev->vpaths[mac->vpath_no];
1098	status = vxge_hw_vpath_mac_addr_delete(vpath->handle, mac->macaddr,
1099						mac->macmask);
1100	if (status != VXGE_HW_OK) {
1101		vxge_debug_init(VXGE_ERR,
1102			"DA config delete entry failed for vpath:%d",
1103			vpath->device_id);
1104	} else
1105		vxge_mac_list_del(vpath, mac);
1106	return status;
1107}
1108
1109/**
1110 * vxge_set_multicast
1111 * @dev: pointer to the device structure
1112 *
1113 * Entry point for multicast address enable/disable
1114 * This function is a driver entry point which gets called by the kernel
1115 * whenever multicast addresses must be enabled/disabled. This also gets
1116 * called to set/reset promiscuous mode. Depending on the deivce flag, we
1117 * determine, if multicast address must be enabled or if promiscuous mode
1118 * is to be disabled etc.
1119 */
1120static void vxge_set_multicast(struct net_device *dev)
1121{
1122	struct netdev_hw_addr *ha;
1123	struct vxgedev *vdev;
1124	int i, mcast_cnt = 0;
1125	struct __vxge_hw_device *hldev;
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	hldev = vdev->devh;
1140
1141	if (unlikely(!is_vxge_card_up(vdev)))
1142		return;
1143
1144	if ((dev->flags & IFF_ALLMULTI) && (!vdev->all_multi_flg)) {
1145		for (i = 0; i < vdev->no_of_vpath; i++) {
1146			vpath = &vdev->vpaths[i];
1147			vxge_assert(vpath->is_open);
1148			status = vxge_hw_vpath_mcast_enable(vpath->handle);
1149			if (status != VXGE_HW_OK)
1150				vxge_debug_init(VXGE_ERR, "failed to enable "
1151						"multicast, status %d", status);
1152			vdev->all_multi_flg = 1;
1153		}
1154	} else if (!(dev->flags & IFF_ALLMULTI) && (vdev->all_multi_flg)) {
1155		for (i = 0; i < vdev->no_of_vpath; i++) {
1156			vpath = &vdev->vpaths[i];
1157			vxge_assert(vpath->is_open);
1158			status = vxge_hw_vpath_mcast_disable(vpath->handle);
1159			if (status != VXGE_HW_OK)
1160				vxge_debug_init(VXGE_ERR, "failed to disable "
1161						"multicast, status %d", status);
1162			vdev->all_multi_flg = 0;
1163		}
1164	}
1165
1166
1167	if (!vdev->config.addr_learn_en) {
1168		for (i = 0; i < vdev->no_of_vpath; i++) {
1169			vpath = &vdev->vpaths[i];
1170			vxge_assert(vpath->is_open);
1171
1172			if (dev->flags & IFF_PROMISC)
1173				status = vxge_hw_vpath_promisc_enable(
1174					vpath->handle);
1175			else
1176				status = vxge_hw_vpath_promisc_disable(
1177					vpath->handle);
1178			if (status != VXGE_HW_OK)
1179				vxge_debug_init(VXGE_ERR, "failed to %s promisc"
1180					", status %d", dev->flags&IFF_PROMISC ?
1181					"enable" : "disable", status);
1182		}
1183	}
1184
1185	memset(&mac_info, 0, sizeof(struct macInfo));
1186	/* Update individual M_CAST address list */
1187	if ((!vdev->all_multi_flg) && netdev_mc_count(dev)) {
1188		mcast_cnt = vdev->vpaths[0].mcast_addr_cnt;
1189		list_head = &vdev->vpaths[0].mac_addr_list;
1190		if ((netdev_mc_count(dev) +
1191			(vdev->vpaths[0].mac_addr_cnt - mcast_cnt)) >
1192				vdev->vpaths[0].max_mac_addr_cnt)
1193			goto _set_all_mcast;
1194
1195		/* Delete previous MC's */
1196		for (i = 0; i < mcast_cnt; i++) {
1197			list_for_each_safe(entry, next, list_head) {
1198				mac_entry = (struct vxge_mac_addrs *)entry;
1199				/* Copy the mac address to delete */
1200				mac_address = (u8 *)&mac_entry->macaddr;
1201				memcpy(mac_info.macaddr, mac_address, ETH_ALEN);
1202
1203				if (is_multicast_ether_addr(mac_info.macaddr)) {
1204					for (vpath_idx = 0; vpath_idx <
1205						vdev->no_of_vpath;
1206						vpath_idx++) {
1207						mac_info.vpath_no = vpath_idx;
1208						status = vxge_del_mac_addr(
1209								vdev,
1210								&mac_info);
1211					}
1212				}
1213			}
1214		}
1215
1216		/* Add new ones */
1217		netdev_for_each_mc_addr(ha, dev) {
1218			memcpy(mac_info.macaddr, ha->addr, ETH_ALEN);
1219			for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath;
1220					vpath_idx++) {
1221				mac_info.vpath_no = vpath_idx;
1222				mac_info.state = VXGE_LL_MAC_ADDR_IN_DA_TABLE;
1223				status = vxge_add_mac_addr(vdev, &mac_info);
1224				if (status != VXGE_HW_OK) {
1225					vxge_debug_init(VXGE_ERR,
1226						"%s:%d Setting individual"
1227						"multicast address failed",
1228						__func__, __LINE__);
1229					goto _set_all_mcast;
1230				}
1231			}
1232		}
1233
1234		return;
1235_set_all_mcast:
1236		mcast_cnt = vdev->vpaths[0].mcast_addr_cnt;
1237		/* Delete previous MC's */
1238		for (i = 0; i < mcast_cnt; i++) {
1239			list_for_each_safe(entry, next, list_head) {
1240				mac_entry = (struct vxge_mac_addrs *)entry;
1241				/* Copy the mac address to delete */
1242				mac_address = (u8 *)&mac_entry->macaddr;
1243				memcpy(mac_info.macaddr, mac_address, ETH_ALEN);
1244
1245				if (is_multicast_ether_addr(mac_info.macaddr))
1246					break;
1247			}
1248
1249			for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath;
1250					vpath_idx++) {
1251				mac_info.vpath_no = vpath_idx;
1252				status = vxge_del_mac_addr(vdev, &mac_info);
1253			}
1254		}
1255
1256		/* Enable all multicast */
1257		for (i = 0; i < vdev->no_of_vpath; i++) {
1258			vpath = &vdev->vpaths[i];
1259			vxge_assert(vpath->is_open);
1260
1261			status = vxge_hw_vpath_mcast_enable(vpath->handle);
1262			if (status != VXGE_HW_OK) {
1263				vxge_debug_init(VXGE_ERR,
1264					"%s:%d Enabling all multicasts failed",
1265					 __func__, __LINE__);
1266			}
1267			vdev->all_multi_flg = 1;
1268		}
1269		dev->flags |= IFF_ALLMULTI;
1270	}
1271
1272	vxge_debug_entryexit(VXGE_TRACE,
1273		"%s:%d  Exiting...", __func__, __LINE__);
1274}
1275
1276/**
1277 * vxge_set_mac_addr
1278 * @dev: pointer to the device structure
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	struct __vxge_hw_device *hldev;
1287	enum vxge_hw_status status = VXGE_HW_OK;
1288	struct macInfo mac_info_new, mac_info_old;
1289	int vpath_idx = 0;
1290
1291	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1292
1293	vdev = netdev_priv(dev);
1294	hldev = vdev->devh;
1295
1296	if (!is_valid_ether_addr(addr->sa_data))
1297		return -EINVAL;
1298
1299	memset(&mac_info_new, 0, sizeof(struct macInfo));
1300	memset(&mac_info_old, 0, sizeof(struct macInfo));
1301
1302	vxge_debug_entryexit(VXGE_TRACE, "%s:%d  Exiting...",
1303		__func__, __LINE__);
1304
1305	/* Get the old address */
1306	memcpy(mac_info_old.macaddr, dev->dev_addr, dev->addr_len);
1307
1308	/* Copy the new address */
1309	memcpy(mac_info_new.macaddr, addr->sa_data, dev->addr_len);
1310
1311	/* First delete the old mac address from all the vpaths
1312	as we can't specify the index while adding new mac address */
1313	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
1314		struct vxge_vpath *vpath = &vdev->vpaths[vpath_idx];
1315		if (!vpath->is_open) {
1316			/* This can happen when this interface is added/removed
1317			to the bonding interface. Delete this station address
1318			from the linked list */
1319			vxge_mac_list_del(vpath, &mac_info_old);
1320
1321			/* Add this new address to the linked list
1322			for later restoring */
1323			vxge_mac_list_add(vpath, &mac_info_new);
1324
1325			continue;
1326		}
1327		/* Delete the station address */
1328		mac_info_old.vpath_no = vpath_idx;
1329		status = vxge_del_mac_addr(vdev, &mac_info_old);
1330	}
1331
1332	if (unlikely(!is_vxge_card_up(vdev))) {
1333		memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
1334		return VXGE_HW_OK;
1335	}
1336
1337	/* Set this mac address to all the vpaths */
1338	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
1339		mac_info_new.vpath_no = vpath_idx;
1340		mac_info_new.state = VXGE_LL_MAC_ADDR_IN_DA_TABLE;
1341		status = vxge_add_mac_addr(vdev, &mac_info_new);
1342		if (status != VXGE_HW_OK)
1343			return -EINVAL;
1344	}
1345
1346	memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
1347
1348	return status;
1349}
1350
1351/*
1352 * vxge_vpath_intr_enable
1353 * @vdev: pointer to vdev
1354 * @vp_id: vpath for which to enable the interrupts
1355 *
1356 * Enables the interrupts for the vpath
1357*/
1358static void vxge_vpath_intr_enable(struct vxgedev *vdev, int vp_id)
1359{
1360	struct vxge_vpath *vpath = &vdev->vpaths[vp_id];
1361	int msix_id = 0;
1362	int tim_msix_id[4] = {0, 1, 0, 0};
1363	int alarm_msix_id = VXGE_ALARM_MSIX_ID;
1364
1365	vxge_hw_vpath_intr_enable(vpath->handle);
1366
1367	if (vdev->config.intr_type == INTA)
1368		vxge_hw_vpath_inta_unmask_tx_rx(vpath->handle);
1369	else {
1370		vxge_hw_vpath_msix_set(vpath->handle, tim_msix_id,
1371			alarm_msix_id);
1372
1373		msix_id = vpath->device_id * VXGE_HW_VPATH_MSIX_ACTIVE;
1374		vxge_hw_vpath_msix_unmask(vpath->handle, msix_id);
1375		vxge_hw_vpath_msix_unmask(vpath->handle, msix_id + 1);
1376
1377		/* enable the alarm vector */
1378		msix_id = (vpath->handle->vpath->hldev->first_vp_id *
1379			VXGE_HW_VPATH_MSIX_ACTIVE) + alarm_msix_id;
1380		vxge_hw_vpath_msix_unmask(vpath->handle, msix_id);
1381	}
1382}
1383
1384/*
1385 * vxge_vpath_intr_disable
1386 * @vdev: pointer to vdev
1387 * @vp_id: vpath for which to disable the interrupts
1388 *
1389 * Disables the interrupts for the vpath
1390*/
1391static void vxge_vpath_intr_disable(struct vxgedev *vdev, int vp_id)
1392{
1393	struct vxge_vpath *vpath = &vdev->vpaths[vp_id];
1394	struct __vxge_hw_device *hldev;
1395	int msix_id;
1396
1397	hldev = pci_get_drvdata(vdev->pdev);
1398
1399	vxge_hw_vpath_wait_receive_idle(hldev, vpath->device_id);
1400
1401	vxge_hw_vpath_intr_disable(vpath->handle);
1402
1403	if (vdev->config.intr_type == INTA)
1404		vxge_hw_vpath_inta_mask_tx_rx(vpath->handle);
1405	else {
1406		msix_id = vpath->device_id * VXGE_HW_VPATH_MSIX_ACTIVE;
1407		vxge_hw_vpath_msix_mask(vpath->handle, msix_id);
1408		vxge_hw_vpath_msix_mask(vpath->handle, msix_id + 1);
1409
1410		/* disable the alarm vector */
1411		msix_id = (vpath->handle->vpath->hldev->first_vp_id *
1412			VXGE_HW_VPATH_MSIX_ACTIVE) + VXGE_ALARM_MSIX_ID;
1413		vxge_hw_vpath_msix_mask(vpath->handle, msix_id);
1414	}
1415}
1416
1417/* list all mac addresses from DA table */
1418static enum vxge_hw_status
1419vxge_search_mac_addr_in_da_table(struct vxge_vpath *vpath, struct macInfo *mac)
1420{
1421	enum vxge_hw_status status = VXGE_HW_OK;
1422	unsigned char macmask[ETH_ALEN];
1423	unsigned char macaddr[ETH_ALEN];
1424
1425	status = vxge_hw_vpath_mac_addr_get(vpath->handle,
1426				macaddr, macmask);
1427	if (status != VXGE_HW_OK) {
1428		vxge_debug_init(VXGE_ERR,
1429			"DA config list entry failed for vpath:%d",
1430			vpath->device_id);
1431		return status;
1432	}
1433
1434	while (!ether_addr_equal(mac->macaddr, macaddr)) {
1435		status = vxge_hw_vpath_mac_addr_get_next(vpath->handle,
1436				macaddr, macmask);
1437		if (status != VXGE_HW_OK)
1438			break;
1439	}
1440
1441	return status;
1442}
1443
1444/* Store all mac addresses from the list to the DA table */
1445static enum vxge_hw_status vxge_restore_vpath_mac_addr(struct vxge_vpath *vpath)
1446{
1447	enum vxge_hw_status status = VXGE_HW_OK;
1448	struct macInfo mac_info;
1449	u8 *mac_address = NULL;
1450	struct list_head *entry, *next;
1451
1452	memset(&mac_info, 0, sizeof(struct macInfo));
1453
1454	if (vpath->is_open) {
1455		list_for_each_safe(entry, next, &vpath->mac_addr_list) {
1456			mac_address =
1457				(u8 *)&
1458				((struct vxge_mac_addrs *)entry)->macaddr;
1459			memcpy(mac_info.macaddr, mac_address, ETH_ALEN);
1460			((struct vxge_mac_addrs *)entry)->state =
1461				VXGE_LL_MAC_ADDR_IN_DA_TABLE;
1462			/* does this mac address already exist in da table? */
1463			status = vxge_search_mac_addr_in_da_table(vpath,
1464				&mac_info);
1465			if (status != VXGE_HW_OK) {
1466				/* Add this mac address to the DA table */
1467				status = vxge_hw_vpath_mac_addr_add(
1468					vpath->handle, mac_info.macaddr,
1469					mac_info.macmask,
1470				    VXGE_HW_VPATH_MAC_ADDR_ADD_DUPLICATE);
1471				if (status != VXGE_HW_OK) {
1472					vxge_debug_init(VXGE_ERR,
1473					    "DA add entry failed for vpath:%d",
1474					    vpath->device_id);
1475					((struct vxge_mac_addrs *)entry)->state
1476						= VXGE_LL_MAC_ADDR_IN_LIST;
1477				}
1478			}
1479		}
1480	}
1481
1482	return status;
1483}
1484
1485/* Store all vlan ids from the list to the vid table */
1486static enum vxge_hw_status
1487vxge_restore_vpath_vid_table(struct vxge_vpath *vpath)
1488{
1489	enum vxge_hw_status status = VXGE_HW_OK;
1490	struct vxgedev *vdev = vpath->vdev;
1491	u16 vid;
1492
1493	if (!vpath->is_open)
1494		return status;
1495
1496	for_each_set_bit(vid, vdev->active_vlans, VLAN_N_VID)
1497		status = vxge_hw_vpath_vid_add(vpath->handle, vid);
1498
1499	return status;
1500}
1501
1502/*
1503 * vxge_reset_vpath
1504 * @vdev: pointer to vdev
1505 * @vp_id: vpath to reset
1506 *
1507 * Resets the vpath
1508*/
1509static int vxge_reset_vpath(struct vxgedev *vdev, int vp_id)
1510{
1511	enum vxge_hw_status status = VXGE_HW_OK;
1512	struct vxge_vpath *vpath = &vdev->vpaths[vp_id];
1513	int ret = 0;
1514
1515	/* check if device is down already */
1516	if (unlikely(!is_vxge_card_up(vdev)))
1517		return 0;
1518
1519	/* is device reset already scheduled */
1520	if (test_bit(__VXGE_STATE_RESET_CARD, &vdev->state))
1521		return 0;
1522
1523	if (vpath->handle) {
1524		if (vxge_hw_vpath_reset(vpath->handle) == VXGE_HW_OK) {
1525			if (is_vxge_card_up(vdev) &&
1526				vxge_hw_vpath_recover_from_reset(vpath->handle)
1527					!= VXGE_HW_OK) {
1528				vxge_debug_init(VXGE_ERR,
1529					"vxge_hw_vpath_recover_from_reset"
1530					"failed for vpath:%d", vp_id);
1531				return status;
1532			}
1533		} else {
1534			vxge_debug_init(VXGE_ERR,
1535				"vxge_hw_vpath_reset failed for"
1536				"vpath:%d", vp_id);
1537				return status;
1538		}
1539	} else
1540		return VXGE_HW_FAIL;
1541
1542	vxge_restore_vpath_mac_addr(vpath);
1543	vxge_restore_vpath_vid_table(vpath);
1544
1545	/* Enable all broadcast */
1546	vxge_hw_vpath_bcast_enable(vpath->handle);
1547
1548	/* Enable all multicast */
1549	if (vdev->all_multi_flg) {
1550		status = vxge_hw_vpath_mcast_enable(vpath->handle);
1551		if (status != VXGE_HW_OK)
1552			vxge_debug_init(VXGE_ERR,
1553				"%s:%d Enabling multicast failed",
1554				__func__, __LINE__);
1555	}
1556
1557	/* Enable the interrupts */
1558	vxge_vpath_intr_enable(vdev, vp_id);
1559
1560	smp_wmb();
1561
1562	/* Enable the flow of traffic through the vpath */
1563	vxge_hw_vpath_enable(vpath->handle);
1564
1565	smp_wmb();
1566	vxge_hw_vpath_rx_doorbell_init(vpath->handle);
1567	vpath->ring.last_status = VXGE_HW_OK;
1568
1569	/* Vpath reset done */
1570	clear_bit(vp_id, &vdev->vp_reset);
1571
1572	/* Start the vpath queue */
1573	if (netif_tx_queue_stopped(vpath->fifo.txq))
1574		netif_tx_wake_queue(vpath->fifo.txq);
1575
1576	return ret;
1577}
1578
1579/* Configure CI */
1580static void vxge_config_ci_for_tti_rti(struct vxgedev *vdev)
1581{
1582	int i = 0;
1583
1584	/* Enable CI for RTI */
1585	if (vdev->config.intr_type == MSI_X) {
1586		for (i = 0; i < vdev->no_of_vpath; i++) {
1587			struct __vxge_hw_ring *hw_ring;
1588
1589			hw_ring = vdev->vpaths[i].ring.handle;
1590			vxge_hw_vpath_dynamic_rti_ci_set(hw_ring);
1591		}
1592	}
1593
1594	/* Enable CI for TTI */
1595	for (i = 0; i < vdev->no_of_vpath; i++) {
1596		struct __vxge_hw_fifo *hw_fifo = vdev->vpaths[i].fifo.handle;
1597		vxge_hw_vpath_tti_ci_set(hw_fifo);
1598		/*
1599		 * For Inta (with or without napi), Set CI ON for only one
1600		 * vpath. (Have only one free running timer).
1601		 */
1602		if ((vdev->config.intr_type == INTA) && (i == 0))
1603			break;
1604	}
1605
1606	return;
1607}
1608
1609static int do_vxge_reset(struct vxgedev *vdev, int event)
1610{
1611	enum vxge_hw_status status;
1612	int ret = 0, vp_id, i;
1613
1614	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1615
1616	if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_START_RESET)) {
1617		/* check if device is down already */
1618		if (unlikely(!is_vxge_card_up(vdev)))
1619			return 0;
1620
1621		/* is reset already scheduled */
1622		if (test_and_set_bit(__VXGE_STATE_RESET_CARD, &vdev->state))
1623			return 0;
1624	}
1625
1626	if (event == VXGE_LL_FULL_RESET) {
1627		netif_carrier_off(vdev->ndev);
1628
1629		/* wait for all the vpath reset to complete */
1630		for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
1631			while (test_bit(vp_id, &vdev->vp_reset))
1632				msleep(50);
1633		}
1634
1635		netif_carrier_on(vdev->ndev);
1636
1637		/* if execution mode is set to debug, don't reset the adapter */
1638		if (unlikely(vdev->exec_mode)) {
1639			vxge_debug_init(VXGE_ERR,
1640				"%s: execution mode is debug, returning..",
1641				vdev->ndev->name);
1642			clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
1643			netif_tx_stop_all_queues(vdev->ndev);
1644			return 0;
1645		}
1646	}
1647
1648	if (event == VXGE_LL_FULL_RESET) {
1649		vxge_hw_device_wait_receive_idle(vdev->devh);
1650		vxge_hw_device_intr_disable(vdev->devh);
1651
1652		switch (vdev->cric_err_event) {
1653		case VXGE_HW_EVENT_UNKNOWN:
1654			netif_tx_stop_all_queues(vdev->ndev);
1655			vxge_debug_init(VXGE_ERR,
1656				"fatal: %s: Disabling device due to"
1657				"unknown error",
1658				vdev->ndev->name);
1659			ret = -EPERM;
1660			goto out;
1661		case VXGE_HW_EVENT_RESET_START:
1662			break;
1663		case VXGE_HW_EVENT_RESET_COMPLETE:
1664		case VXGE_HW_EVENT_LINK_DOWN:
1665		case VXGE_HW_EVENT_LINK_UP:
1666		case VXGE_HW_EVENT_ALARM_CLEARED:
1667		case VXGE_HW_EVENT_ECCERR:
1668		case VXGE_HW_EVENT_MRPCIM_ECCERR:
1669			ret = -EPERM;
1670			goto out;
1671		case VXGE_HW_EVENT_FIFO_ERR:
1672		case VXGE_HW_EVENT_VPATH_ERR:
1673			break;
1674		case VXGE_HW_EVENT_CRITICAL_ERR:
1675			netif_tx_stop_all_queues(vdev->ndev);
1676			vxge_debug_init(VXGE_ERR,
1677				"fatal: %s: Disabling device due to"
1678				"serious error",
1679				vdev->ndev->name);
1680			/* SOP or device reset required */
1681			/* This event is not currently used */
1682			ret = -EPERM;
1683			goto out;
1684		case VXGE_HW_EVENT_SERR:
1685			netif_tx_stop_all_queues(vdev->ndev);
1686			vxge_debug_init(VXGE_ERR,
1687				"fatal: %s: Disabling device due to"
1688				"serious error",
1689				vdev->ndev->name);
1690			ret = -EPERM;
1691			goto out;
1692		case VXGE_HW_EVENT_SRPCIM_SERR:
1693		case VXGE_HW_EVENT_MRPCIM_SERR:
1694			ret = -EPERM;
1695			goto out;
1696		case VXGE_HW_EVENT_SLOT_FREEZE:
1697			netif_tx_stop_all_queues(vdev->ndev);
1698			vxge_debug_init(VXGE_ERR,
1699				"fatal: %s: Disabling device due to"
1700				"slot freeze",
1701				vdev->ndev->name);
1702			ret = -EPERM;
1703			goto out;
1704		default:
1705			break;
1706
1707		}
1708	}
1709
1710	if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_START_RESET))
1711		netif_tx_stop_all_queues(vdev->ndev);
1712
1713	if (event == VXGE_LL_FULL_RESET) {
1714		status = vxge_reset_all_vpaths(vdev);
1715		if (status != VXGE_HW_OK) {
1716			vxge_debug_init(VXGE_ERR,
1717				"fatal: %s: can not reset vpaths",
1718				vdev->ndev->name);
1719			ret = -EPERM;
1720			goto out;
1721		}
1722	}
1723
1724	if (event == VXGE_LL_COMPL_RESET) {
1725		for (i = 0; i < vdev->no_of_vpath; i++)
1726			if (vdev->vpaths[i].handle) {
1727				if (vxge_hw_vpath_recover_from_reset(
1728					vdev->vpaths[i].handle)
1729						!= VXGE_HW_OK) {
1730					vxge_debug_init(VXGE_ERR,
1731						"vxge_hw_vpath_recover_"
1732						"from_reset failed for vpath: "
1733						"%d", i);
1734					ret = -EPERM;
1735					goto out;
1736				}
1737				} else {
1738					vxge_debug_init(VXGE_ERR,
1739					"vxge_hw_vpath_reset failed for "
1740						"vpath:%d", i);
1741					ret = -EPERM;
1742					goto out;
1743				}
1744	}
1745
1746	if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_COMPL_RESET)) {
1747		/* Reprogram the DA table with populated mac addresses */
1748		for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
1749			vxge_restore_vpath_mac_addr(&vdev->vpaths[vp_id]);
1750			vxge_restore_vpath_vid_table(&vdev->vpaths[vp_id]);
1751		}
1752
1753		/* enable vpath interrupts */
1754		for (i = 0; i < vdev->no_of_vpath; i++)
1755			vxge_vpath_intr_enable(vdev, i);
1756
1757		vxge_hw_device_intr_enable(vdev->devh);
1758
1759		smp_wmb();
1760
1761		/* Indicate card up */
1762		set_bit(__VXGE_STATE_CARD_UP, &vdev->state);
1763
1764		/* Get the traffic to flow through the vpaths */
1765		for (i = 0; i < vdev->no_of_vpath; i++) {
1766			vxge_hw_vpath_enable(vdev->vpaths[i].handle);
1767			smp_wmb();
1768			vxge_hw_vpath_rx_doorbell_init(vdev->vpaths[i].handle);
1769		}
1770
1771		netif_tx_wake_all_queues(vdev->ndev);
1772	}
1773
1774	/* configure CI */
1775	vxge_config_ci_for_tti_rti(vdev);
1776
1777out:
1778	vxge_debug_entryexit(VXGE_TRACE,
1779		"%s:%d  Exiting...", __func__, __LINE__);
1780
1781	/* Indicate reset done */
1782	if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_COMPL_RESET))
1783		clear_bit(__VXGE_STATE_RESET_CARD, &vdev->state);
1784	return ret;
1785}
1786
1787/*
1788 * vxge_reset
1789 * @vdev: pointer to ll device
1790 *
1791 * driver may reset the chip on events of serr, eccerr, etc
1792 */
1793static void vxge_reset(struct work_struct *work)
1794{
1795	struct vxgedev *vdev = container_of(work, struct vxgedev, reset_task);
1796
1797	if (!netif_running(vdev->ndev))
1798		return;
1799
1800	do_vxge_reset(vdev, VXGE_LL_FULL_RESET);
1801}
1802
1803/**
1804 * vxge_poll - Receive handler when Receive Polling is used.
1805 * @dev: pointer to the device structure.
1806 * @budget: Number of packets budgeted to be processed in this iteration.
1807 *
1808 * This function comes into picture only if Receive side is being handled
1809 * through polling (called NAPI in linux). It mostly does what the normal
1810 * Rx interrupt handler does in terms of descriptor and packet processing
1811 * but not in an interrupt context. Also it will process a specified number
1812 * of packets at most in one iteration. This value is passed down by the
1813 * kernel as the function argument 'budget'.
1814 */
1815static int vxge_poll_msix(struct napi_struct *napi, int budget)
1816{
1817	struct vxge_ring *ring = container_of(napi, struct vxge_ring, napi);
1818	int pkts_processed;
1819	int budget_org = budget;
1820
1821	ring->budget = budget;
1822	ring->pkts_processed = 0;
1823	vxge_hw_vpath_poll_rx(ring->handle);
1824	pkts_processed = ring->pkts_processed;
1825
1826	if (ring->pkts_processed < budget_org) {
1827		napi_complete(napi);
1828
1829		/* Re enable the Rx interrupts for the vpath */
1830		vxge_hw_channel_msix_unmask(
1831				(struct __vxge_hw_channel *)ring->handle,
1832				ring->rx_vector_no);
1833		mmiowb();
1834	}
1835
1836	/* We are copying and returning the local variable, in case if after
1837	 * clearing the msix interrupt above, if the interrupt fires right
1838	 * away which can preempt this NAPI thread */
1839	return pkts_processed;
1840}
1841
1842static int vxge_poll_inta(struct napi_struct *napi, int budget)
1843{
1844	struct vxgedev *vdev = container_of(napi, struct vxgedev, napi);
1845	int pkts_processed = 0;
1846	int i;
1847	int budget_org = budget;
1848	struct vxge_ring *ring;
1849
1850	struct __vxge_hw_device *hldev = pci_get_drvdata(vdev->pdev);
1851
1852	for (i = 0; i < vdev->no_of_vpath; i++) {
1853		ring = &vdev->vpaths[i].ring;
1854		ring->budget = budget;
1855		ring->pkts_processed = 0;
1856		vxge_hw_vpath_poll_rx(ring->handle);
1857		pkts_processed += ring->pkts_processed;
1858		budget -= ring->pkts_processed;
1859		if (budget <= 0)
1860			break;
1861	}
1862
1863	VXGE_COMPLETE_ALL_TX(vdev);
1864
1865	if (pkts_processed < budget_org) {
1866		napi_complete(napi);
1867		/* Re enable the Rx interrupts for the ring */
1868		vxge_hw_device_unmask_all(hldev);
1869		vxge_hw_device_flush_io(hldev);
1870	}
1871
1872	return pkts_processed;
1873}
1874
1875#ifdef CONFIG_NET_POLL_CONTROLLER
1876/**
1877 * vxge_netpoll - netpoll event handler entry point
1878 * @dev : pointer to the device structure.
1879 * Description:
1880 *      This function will be called by upper layer to check for events on the
1881 * interface in situations where interrupts are disabled. It is used for
1882 * specific in-kernel networking tasks, such as remote consoles and kernel
1883 * debugging over the network (example netdump in RedHat).
1884 */
1885static void vxge_netpoll(struct net_device *dev)
1886{
1887	struct vxgedev *vdev = netdev_priv(dev);
1888	struct pci_dev *pdev = vdev->pdev;
1889	struct __vxge_hw_device *hldev = pci_get_drvdata(pdev);
1890	const int irq = pdev->irq;
1891
1892	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1893
1894	if (pci_channel_offline(pdev))
1895		return;
1896
1897	disable_irq(irq);
1898	vxge_hw_device_clear_tx_rx(hldev);
1899
1900	vxge_hw_device_clear_tx_rx(hldev);
1901	VXGE_COMPLETE_ALL_RX(vdev);
1902	VXGE_COMPLETE_ALL_TX(vdev);
1903
1904	enable_irq(irq);
1905
1906	vxge_debug_entryexit(VXGE_TRACE,
1907		"%s:%d  Exiting...", __func__, __LINE__);
1908}
1909#endif
1910
1911/* RTH configuration */
1912static enum vxge_hw_status vxge_rth_configure(struct vxgedev *vdev)
1913{
1914	enum vxge_hw_status status = VXGE_HW_OK;
1915	struct vxge_hw_rth_hash_types hash_types;
1916	u8 itable[256] = {0}; /* indirection table */
1917	u8 mtable[256] = {0}; /* CPU to vpath mapping  */
1918	int index;
1919
1920	/*
1921	 * Filling
1922	 * 	- itable with bucket numbers
1923	 * 	- mtable with bucket-to-vpath mapping
1924	 */
1925	for (index = 0; index < (1 << vdev->config.rth_bkt_sz); index++) {
1926		itable[index] = index;
1927		mtable[index] = index % vdev->no_of_vpath;
1928	}
1929
1930	/* set indirection table, bucket-to-vpath mapping */
1931	status = vxge_hw_vpath_rts_rth_itable_set(vdev->vp_handles,
1932						vdev->no_of_vpath,
1933						mtable, itable,
1934						vdev->config.rth_bkt_sz);
1935	if (status != VXGE_HW_OK) {
1936		vxge_debug_init(VXGE_ERR,
1937			"RTH indirection table configuration failed "
1938			"for vpath:%d", vdev->vpaths[0].device_id);
1939		return status;
1940	}
1941
1942	/* Fill RTH hash types */
1943	hash_types.hash_type_tcpipv4_en   = vdev->config.rth_hash_type_tcpipv4;
1944	hash_types.hash_type_ipv4_en      = vdev->config.rth_hash_type_ipv4;
1945	hash_types.hash_type_tcpipv6_en   = vdev->config.rth_hash_type_tcpipv6;
1946	hash_types.hash_type_ipv6_en      = vdev->config.rth_hash_type_ipv6;
1947	hash_types.hash_type_tcpipv6ex_en =
1948					vdev->config.rth_hash_type_tcpipv6ex;
1949	hash_types.hash_type_ipv6ex_en    = vdev->config.rth_hash_type_ipv6ex;
1950
1951	/*
1952	 * Because the itable_set() method uses the active_table field
1953	 * for the target virtual path the RTH config should be updated
1954	 * for all VPATHs. The h/w only uses the lowest numbered VPATH
1955	 * when steering frames.
1956	 */
1957	 for (index = 0; index < vdev->no_of_vpath; index++) {
1958		status = vxge_hw_vpath_rts_rth_set(
1959				vdev->vpaths[index].handle,
1960				vdev->config.rth_algorithm,
1961				&hash_types,
1962				vdev->config.rth_bkt_sz);
1963		 if (status != VXGE_HW_OK) {
1964			vxge_debug_init(VXGE_ERR,
1965				"RTH configuration failed for vpath:%d",
1966				vdev->vpaths[index].device_id);
1967			return status;
1968		 }
1969	 }
1970
1971	return status;
1972}
1973
1974/* reset vpaths */
1975static enum vxge_hw_status vxge_reset_all_vpaths(struct vxgedev *vdev)
1976{
1977	enum vxge_hw_status status = VXGE_HW_OK;
1978	struct vxge_vpath *vpath;
1979	int i;
1980
1981	for (i = 0; i < vdev->no_of_vpath; i++) {
1982		vpath = &vdev->vpaths[i];
1983		if (vpath->handle) {
1984			if (vxge_hw_vpath_reset(vpath->handle) == VXGE_HW_OK) {
1985				if (is_vxge_card_up(vdev) &&
1986					vxge_hw_vpath_recover_from_reset(
1987						vpath->handle) != VXGE_HW_OK) {
1988					vxge_debug_init(VXGE_ERR,
1989						"vxge_hw_vpath_recover_"
1990						"from_reset failed for vpath: "
1991						"%d", i);
1992					return status;
1993				}
1994			} else {
1995				vxge_debug_init(VXGE_ERR,
1996					"vxge_hw_vpath_reset failed for "
1997					"vpath:%d", i);
1998					return status;
1999			}
2000		}
2001	}
2002
2003	return status;
2004}
2005
2006/* close vpaths */
2007static void vxge_close_vpaths(struct vxgedev *vdev, int index)
2008{
2009	struct vxge_vpath *vpath;
2010	int i;
2011
2012	for (i = index; i < vdev->no_of_vpath; i++) {
2013		vpath = &vdev->vpaths[i];
2014
2015		if (vpath->handle && vpath->is_open) {
2016			vxge_hw_vpath_close(vpath->handle);
2017			vdev->stats.vpaths_open--;
2018		}
2019		vpath->is_open = 0;
2020		vpath->handle = NULL;
2021	}
2022}
2023
2024/* open vpaths */
2025static int vxge_open_vpaths(struct vxgedev *vdev)
2026{
2027	struct vxge_hw_vpath_attr attr;
2028	enum vxge_hw_status status;
2029	struct vxge_vpath *vpath;
2030	u32 vp_id = 0;
2031	int i;
2032
2033	for (i = 0; i < vdev->no_of_vpath; i++) {
2034		vpath = &vdev->vpaths[i];
2035		vxge_assert(vpath->is_configured);
2036
2037		if (!vdev->titan1) {
2038			struct vxge_hw_vp_config *vcfg;
2039			vcfg = &vdev->devh->config.vp_config[vpath->device_id];
2040
2041			vcfg->rti.urange_a = RTI_T1A_RX_URANGE_A;
2042			vcfg->rti.urange_b = RTI_T1A_RX_URANGE_B;
2043			vcfg->rti.urange_c = RTI_T1A_RX_URANGE_C;
2044			vcfg->tti.uec_a = TTI_T1A_TX_UFC_A;
2045			vcfg->tti.uec_b = TTI_T1A_TX_UFC_B;
2046			vcfg->tti.uec_c = TTI_T1A_TX_UFC_C(vdev->mtu);
2047			vcfg->tti.uec_d = TTI_T1A_TX_UFC_D(vdev->mtu);
2048			vcfg->tti.ltimer_val = VXGE_T1A_TTI_LTIMER_VAL;
2049			vcfg->tti.rtimer_val = VXGE_T1A_TTI_RTIMER_VAL;
2050		}
2051
2052		attr.vp_id = vpath->device_id;
2053		attr.fifo_attr.callback = vxge_xmit_compl;
2054		attr.fifo_attr.txdl_term = vxge_tx_term;
2055		attr.fifo_attr.per_txdl_space = sizeof(struct vxge_tx_priv);
2056		attr.fifo_attr.userdata = &vpath->fifo;
2057
2058		attr.ring_attr.callback = vxge_rx_1b_compl;
2059		attr.ring_attr.rxd_init = vxge_rx_initial_replenish;
2060		attr.ring_attr.rxd_term = vxge_rx_term;
2061		attr.ring_attr.per_rxd_space = sizeof(struct vxge_rx_priv);
2062		attr.ring_attr.userdata = &vpath->ring;
2063
2064		vpath->ring.ndev = vdev->ndev;
2065		vpath->ring.pdev = vdev->pdev;
2066
2067		status = vxge_hw_vpath_open(vdev->devh, &attr, &vpath->handle);
2068		if (status == VXGE_HW_OK) {
2069			vpath->fifo.handle =
2070			    (struct __vxge_hw_fifo *)attr.fifo_attr.userdata;
2071			vpath->ring.handle =
2072			    (struct __vxge_hw_ring *)attr.ring_attr.userdata;
2073			vpath->fifo.tx_steering_type =
2074				vdev->config.tx_steering_type;
2075			vpath->fifo.ndev = vdev->ndev;
2076			vpath->fifo.pdev = vdev->pdev;
2077
2078			u64_stats_init(&vpath->fifo.stats.syncp);
2079			u64_stats_init(&vpath->ring.stats.syncp);
2080
2081			if (vdev->config.tx_steering_type)
2082				vpath->fifo.txq =
2083					netdev_get_tx_queue(vdev->ndev, i);
2084			else
2085				vpath->fifo.txq =
2086					netdev_get_tx_queue(vdev->ndev, 0);
2087			vpath->fifo.indicate_max_pkts =
2088				vdev->config.fifo_indicate_max_pkts;
2089			vpath->fifo.tx_vector_no = 0;
2090			vpath->ring.rx_vector_no = 0;
2091			vpath->ring.rx_hwts = vdev->rx_hwts;
2092			vpath->is_open = 1;
2093			vdev->vp_handles[i] = vpath->handle;
2094			vpath->ring.vlan_tag_strip = vdev->vlan_tag_strip;
2095			vdev->stats.vpaths_open++;
2096		} else {
2097			vdev->stats.vpath_open_fail++;
2098			vxge_debug_init(VXGE_ERR, "%s: vpath: %d failed to "
2099					"open with status: %d",
2100					vdev->ndev->name, vpath->device_id,
2101					status);
2102			vxge_close_vpaths(vdev, 0);
2103			return -EPERM;
2104		}
2105
2106		vp_id = vpath->handle->vpath->vp_id;
2107		vdev->vpaths_deployed |= vxge_mBIT(vp_id);
2108	}
2109
2110	return VXGE_HW_OK;
2111}
2112
2113/**
2114 *  adaptive_coalesce_tx_interrupts - Changes the interrupt coalescing
2115 *  if the interrupts are not within a range
2116 *  @fifo: pointer to transmit fifo structure
2117 *  Description: The function changes boundary timer and restriction timer
2118 *  value depends on the traffic
2119 *  Return Value: None
2120 */
2121static void adaptive_coalesce_tx_interrupts(struct vxge_fifo *fifo)
2122{
2123	fifo->interrupt_count++;
2124	if (time_before(fifo->jiffies + HZ / 100, jiffies)) {
2125		struct __vxge_hw_fifo *hw_fifo = fifo->handle;
2126
2127		fifo->jiffies = jiffies;
2128		if (fifo->interrupt_count > VXGE_T1A_MAX_TX_INTERRUPT_COUNT &&
2129		    hw_fifo->rtimer != VXGE_TTI_RTIMER_ADAPT_VAL) {
2130			hw_fifo->rtimer = VXGE_TTI_RTIMER_ADAPT_VAL;
2131			vxge_hw_vpath_dynamic_tti_rtimer_set(hw_fifo);
2132		} else if (hw_fifo->rtimer != 0) {
2133			hw_fifo->rtimer = 0;
2134			vxge_hw_vpath_dynamic_tti_rtimer_set(hw_fifo);
2135		}
2136		fifo->interrupt_count = 0;
2137	}
2138}
2139
2140/**
2141 *  adaptive_coalesce_rx_interrupts - Changes the interrupt coalescing
2142 *  if the interrupts are not within a range
2143 *  @ring: pointer to receive ring structure
2144 *  Description: The function increases of decreases the packet counts within
2145 *  the ranges of traffic utilization, if the interrupts due to this ring are
2146 *  not within a fixed range.
2147 *  Return Value: Nothing
2148 */
2149static void adaptive_coalesce_rx_interrupts(struct vxge_ring *ring)
2150{
2151	ring->interrupt_count++;
2152	if (time_before(ring->jiffies + HZ / 100, jiffies)) {
2153		struct __vxge_hw_ring *hw_ring = ring->handle;
2154
2155		ring->jiffies = jiffies;
2156		if (ring->interrupt_count > VXGE_T1A_MAX_INTERRUPT_COUNT &&
2157		    hw_ring->rtimer != VXGE_RTI_RTIMER_ADAPT_VAL) {
2158			hw_ring->rtimer = VXGE_RTI_RTIMER_ADAPT_VAL;
2159			vxge_hw_vpath_dynamic_rti_rtimer_set(hw_ring);
2160		} else if (hw_ring->rtimer != 0) {
2161			hw_ring->rtimer = 0;
2162			vxge_hw_vpath_dynamic_rti_rtimer_set(hw_ring);
2163		}
2164		ring->interrupt_count = 0;
2165	}
2166}
2167
2168/*
2169 *  vxge_isr_napi
2170 *  @irq: the irq of the device.
2171 *  @dev_id: a void pointer to the hldev structure of the Titan device
2172 *  @ptregs: pointer to the registers pushed on the stack.
2173 *
2174 *  This function is the ISR handler of the device when napi is enabled. It
2175 *  identifies the reason for the interrupt and calls the relevant service
2176 *  routines.
2177 */
2178static irqreturn_t vxge_isr_napi(int irq, void *dev_id)
2179{
2180	struct net_device *dev;
2181	struct __vxge_hw_device *hldev;
2182	u64 reason;
2183	enum vxge_hw_status status;
2184	struct vxgedev *vdev = (struct vxgedev *)dev_id;
2185
2186	vxge_debug_intr(VXGE_TRACE, "%s:%d", __func__, __LINE__);
2187
2188	dev = vdev->ndev;
2189	hldev = pci_get_drvdata(vdev->pdev);
2190
2191	if (pci_channel_offline(vdev->pdev))
2192		return IRQ_NONE;
2193
2194	if (unlikely(!is_vxge_card_up(vdev)))
2195		return IRQ_HANDLED;
2196
2197	status = vxge_hw_device_begin_irq(hldev, vdev->exec_mode, &reason);
2198	if (status == VXGE_HW_OK) {
2199		vxge_hw_device_mask_all(hldev);
2200
2201		if (reason &
2202			VXGE_HW_TITAN_GENERAL_INT_STATUS_VPATH_TRAFFIC_INT(
2203			vdev->vpaths_deployed >>
2204			(64 - VXGE_HW_MAX_VIRTUAL_PATHS))) {
2205
2206			vxge_hw_device_clear_tx_rx(hldev);
2207			napi_schedule(&vdev->napi);
2208			vxge_debug_intr(VXGE_TRACE,
2209				"%s:%d  Exiting...", __func__, __LINE__);
2210			return IRQ_HANDLED;
2211		} else
2212			vxge_hw_device_unmask_all(hldev);
2213	} else if (unlikely((status == VXGE_HW_ERR_VPATH) ||
2214		(status == VXGE_HW_ERR_CRITICAL) ||
2215		(status == VXGE_HW_ERR_FIFO))) {
2216		vxge_hw_device_mask_all(hldev);
2217		vxge_hw_device_flush_io(hldev);
2218		return IRQ_HANDLED;
2219	} else if (unlikely(status == VXGE_HW_ERR_SLOT_FREEZE))
2220		return IRQ_HANDLED;
2221
2222	vxge_debug_intr(VXGE_TRACE, "%s:%d  Exiting...", __func__, __LINE__);
2223	return IRQ_NONE;
2224}
2225
2226#ifdef CONFIG_PCI_MSI
2227
2228static irqreturn_t vxge_tx_msix_handle(int irq, void *dev_id)
2229{
2230	struct vxge_fifo *fifo = (struct vxge_fifo *)dev_id;
2231
2232	adaptive_coalesce_tx_interrupts(fifo);
2233
2234	vxge_hw_channel_msix_mask((struct __vxge_hw_channel *)fifo->handle,
2235				  fifo->tx_vector_no);
2236
2237	vxge_hw_channel_msix_clear((struct __vxge_hw_channel *)fifo->handle,
2238				   fifo->tx_vector_no);
2239
2240	VXGE_COMPLETE_VPATH_TX(fifo);
2241
2242	vxge_hw_channel_msix_unmask((struct __vxge_hw_channel *)fifo->handle,
2243				    fifo->tx_vector_no);
2244
2245	mmiowb();
2246
2247	return IRQ_HANDLED;
2248}
2249
2250static irqreturn_t vxge_rx_msix_napi_handle(int irq, void *dev_id)
2251{
2252	struct vxge_ring *ring = (struct vxge_ring *)dev_id;
2253
2254	adaptive_coalesce_rx_interrupts(ring);
2255
2256	vxge_hw_channel_msix_mask((struct __vxge_hw_channel *)ring->handle,
2257				  ring->rx_vector_no);
2258
2259	vxge_hw_channel_msix_clear((struct __vxge_hw_channel *)ring->handle,
2260				   ring->rx_vector_no);
2261
2262	napi_schedule(&ring->napi);
2263	return IRQ_HANDLED;
2264}
2265
2266static irqreturn_t
2267vxge_alarm_msix_handle(int irq, void *dev_id)
2268{
2269	int i;
2270	enum vxge_hw_status status;
2271	struct vxge_vpath *vpath = (struct vxge_vpath *)dev_id;
2272	struct vxgedev *vdev = vpath->vdev;
2273	int msix_id = (vpath->handle->vpath->vp_id *
2274		VXGE_HW_VPATH_MSIX_ACTIVE) + VXGE_ALARM_MSIX_ID;
2275
2276	for (i = 0; i < vdev->no_of_vpath; i++) {
2277		/* Reduce the chance of losing alarm interrupts by masking
2278		 * the vector. A pending bit will be set if an alarm is
2279		 * generated and on unmask the interrupt will be fired.
2280		 */
2281		vxge_hw_vpath_msix_mask(vdev->vpaths[i].handle, msix_id);
2282		vxge_hw_vpath_msix_clear(vdev->vpaths[i].handle, msix_id);
2283		mmiowb();
2284
2285		status = vxge_hw_vpath_alarm_process(vdev->vpaths[i].handle,
2286			vdev->exec_mode);
2287		if (status == VXGE_HW_OK) {
2288			vxge_hw_vpath_msix_unmask(vdev->vpaths[i].handle,
2289						  msix_id);
2290			mmiowb();
2291			continue;
2292		}
2293		vxge_debug_intr(VXGE_ERR,
2294			"%s: vxge_hw_vpath_alarm_process failed %x ",
2295			VXGE_DRIVER_NAME, status);
2296	}
2297	return IRQ_HANDLED;
2298}
2299
2300static int vxge_alloc_msix(struct vxgedev *vdev)
2301{
2302	int j, i, ret = 0;
2303	int msix_intr_vect = 0, temp;
2304	vdev->intr_cnt = 0;
2305
2306start:
2307	/* Tx/Rx MSIX Vectors count */
2308	vdev->intr_cnt = vdev->no_of_vpath * 2;
2309
2310	/* Alarm MSIX Vectors count */
2311	vdev->intr_cnt++;
2312
2313	vdev->entries = kcalloc(vdev->intr_cnt, sizeof(struct msix_entry),
2314				GFP_KERNEL);
2315	if (!vdev->entries) {
2316		vxge_debug_init(VXGE_ERR,
2317			"%s: memory allocation failed",
2318			VXGE_DRIVER_NAME);
2319		ret = -ENOMEM;
2320		goto alloc_entries_failed;
2321	}
2322
2323	vdev->vxge_entries = kcalloc(vdev->intr_cnt,
2324				     sizeof(struct vxge_msix_entry),
2325				     GFP_KERNEL);
2326	if (!vdev->vxge_entries) {
2327		vxge_debug_init(VXGE_ERR, "%s: memory allocation failed",
2328			VXGE_DRIVER_NAME);
2329		ret = -ENOMEM;
2330		goto alloc_vxge_entries_failed;
2331	}
2332
2333	for (i = 0, j = 0; i < vdev->no_of_vpath; i++) {
2334
2335		msix_intr_vect = i * VXGE_HW_VPATH_MSIX_ACTIVE;
2336
2337		/* Initialize the fifo vector */
2338		vdev->entries[j].entry = msix_intr_vect;
2339		vdev->vxge_entries[j].entry = msix_intr_vect;
2340		vdev->vxge_entries[j].in_use = 0;
2341		j++;
2342
2343		/* Initialize the ring vector */
2344		vdev->entries[j].entry = msix_intr_vect + 1;
2345		vdev->vxge_entries[j].entry = msix_intr_vect + 1;
2346		vdev->vxge_entries[j].in_use = 0;
2347		j++;
2348	}
2349
2350	/* Initialize the alarm vector */
2351	vdev->entries[j].entry = VXGE_ALARM_MSIX_ID;
2352	vdev->vxge_entries[j].entry = VXGE_ALARM_MSIX_ID;
2353	vdev->vxge_entries[j].in_use = 0;
2354
2355	ret = pci_enable_msix_range(vdev->pdev,
2356				    vdev->entries, 3, vdev->intr_cnt);
2357	if (ret < 0) {
2358		ret = -ENODEV;
2359		goto enable_msix_failed;
2360	} else if (ret < vdev->intr_cnt) {
2361		pci_disable_msix(vdev->pdev);
2362
2363		vxge_debug_init(VXGE_ERR,
2364			"%s: MSI-X enable failed for %d vectors, ret: %d",
2365			VXGE_DRIVER_NAME, vdev->intr_cnt, ret);
2366		if (max_config_vpath != VXGE_USE_DEFAULT) {
2367			ret = -ENODEV;
2368			goto enable_msix_failed;
2369		}
2370
2371		kfree(vdev->entries);
2372		kfree(vdev->vxge_entries);
2373		vdev->entries = NULL;
2374		vdev->vxge_entries = NULL;
2375		/* Try with less no of vector by reducing no of vpaths count */
2376		temp = (ret - 1)/2;
2377		vxge_close_vpaths(vdev, temp);
2378		vdev->no_of_vpath = temp;
2379		goto start;
2380	}
2381	return 0;
2382
2383enable_msix_failed:
2384	kfree(vdev->vxge_entries);
2385alloc_vxge_entries_failed:
2386	kfree(vdev->entries);
2387alloc_entries_failed:
2388	return ret;
2389}
2390
2391static int vxge_enable_msix(struct vxgedev *vdev)
2392{
2393
2394	int i, ret = 0;
2395	/* 0 - Tx, 1 - Rx  */
2396	int tim_msix_id[4] = {0, 1, 0, 0};
2397
2398	vdev->intr_cnt = 0;
2399
2400	/* allocate msix vectors */
2401	ret = vxge_alloc_msix(vdev);
2402	if (!ret) {
2403		for (i = 0; i < vdev->no_of_vpath; i++) {
2404			struct vxge_vpath *vpath = &vdev->vpaths[i];
2405
2406			/* If fifo or ring are not enabled, the MSIX vector for
2407			 * it should be set to 0.
2408			 */
2409			vpath->ring.rx_vector_no = (vpath->device_id *
2410						VXGE_HW_VPATH_MSIX_ACTIVE) + 1;
2411
2412			vpath->fifo.tx_vector_no = (vpath->device_id *
2413						VXGE_HW_VPATH_MSIX_ACTIVE);
2414
2415			vxge_hw_vpath_msix_set(vpath->handle, tim_msix_id,
2416					       VXGE_ALARM_MSIX_ID);
2417		}
2418	}
2419
2420	return ret;
2421}
2422
2423static void vxge_rem_msix_isr(struct vxgedev *vdev)
2424{
2425	int intr_cnt;
2426
2427	for (intr_cnt = 0; intr_cnt < (vdev->no_of_vpath * 2 + 1);
2428		intr_cnt++) {
2429		if (vdev->vxge_entries[intr_cnt].in_use) {
2430			synchronize_irq(vdev->entries[intr_cnt].vector);
2431			free_irq(vdev->entries[intr_cnt].vector,
2432				vdev->vxge_entries[intr_cnt].arg);
2433			vdev->vxge_entries[intr_cnt].in_use = 0;
2434		}
2435	}
2436
2437	kfree(vdev->entries);
2438	kfree(vdev->vxge_entries);
2439	vdev->entries = NULL;
2440	vdev->vxge_entries = NULL;
2441
2442	if (vdev->config.intr_type == MSI_X)
2443		pci_disable_msix(vdev->pdev);
2444}
2445#endif
2446
2447static void vxge_rem_isr(struct vxgedev *vdev)
2448{
2449#ifdef CONFIG_PCI_MSI
2450	if (vdev->config.intr_type == MSI_X) {
2451		vxge_rem_msix_isr(vdev);
2452	} else
2453#endif
2454	if (vdev->config.intr_type == INTA) {
2455			synchronize_irq(vdev->pdev->irq);
2456			free_irq(vdev->pdev->irq, vdev);
2457	}
2458}
2459
2460static int vxge_add_isr(struct vxgedev *vdev)
2461{
2462	int ret = 0;
2463#ifdef CONFIG_PCI_MSI
2464	int vp_idx = 0, intr_idx = 0, intr_cnt = 0, msix_idx = 0, irq_req = 0;
2465	int pci_fun = PCI_FUNC(vdev->pdev->devfn);
2466
2467	if (vdev->config.intr_type == MSI_X)
2468		ret = vxge_enable_msix(vdev);
2469
2470	if (ret) {
2471		vxge_debug_init(VXGE_ERR,
2472			"%s: Enabling MSI-X Failed", VXGE_DRIVER_NAME);
2473		vxge_debug_init(VXGE_ERR,
2474			"%s: Defaulting to INTA", VXGE_DRIVER_NAME);
2475		vdev->config.intr_type = INTA;
2476	}
2477
2478	if (vdev->config.intr_type == MSI_X) {
2479		for (intr_idx = 0;
2480		     intr_idx < (vdev->no_of_vpath *
2481			VXGE_HW_VPATH_MSIX_ACTIVE); intr_idx++) {
2482
2483			msix_idx = intr_idx % VXGE_HW_VPATH_MSIX_ACTIVE;
2484			irq_req = 0;
2485
2486			switch (msix_idx) {
2487			case 0:
2488				snprintf(vdev->desc[intr_cnt], VXGE_INTR_STRLEN,
2489				"%s:vxge:MSI-X %d - Tx - 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_tx_msix_handle, 0,
2496					vdev->desc[intr_cnt],
2497					&vdev->vpaths[vp_idx].fifo);
2498					vdev->vxge_entries[intr_cnt].arg =
2499						&vdev->vpaths[vp_idx].fifo;
2500				irq_req = 1;
2501				break;
2502			case 1:
2503				snprintf(vdev->desc[intr_cnt], VXGE_INTR_STRLEN,
2504				"%s:vxge:MSI-X %d - Rx - fn:%d vpath:%d",
2505					vdev->ndev->name,
2506					vdev->entries[intr_cnt].entry,
2507					pci_fun, vp_idx);
2508				ret = request_irq(
2509				    vdev->entries[intr_cnt].vector,
2510					vxge_rx_msix_napi_handle,
2511					0,
2512					vdev->desc[intr_cnt],
2513					&vdev->vpaths[vp_idx].ring);
2514					vdev->vxge_entries[intr_cnt].arg =
2515						&vdev->vpaths[vp_idx].ring;
2516				irq_req = 1;
2517				break;
2518			}
2519
2520			if (ret) {
2521				vxge_debug_init(VXGE_ERR,
2522					"%s: MSIX - %d  Registration failed",
2523					vdev->ndev->name, intr_cnt);
2524				vxge_rem_msix_isr(vdev);
2525				vdev->config.intr_type = INTA;
2526				vxge_debug_init(VXGE_ERR,
2527					"%s: Defaulting to INTA"
2528					, vdev->ndev->name);
2529					goto INTA_MODE;
2530			}
2531
2532			if (irq_req) {
2533				/* We requested for this msix interrupt */
2534				vdev->vxge_entries[intr_cnt].in_use = 1;
2535				msix_idx +=  vdev->vpaths[vp_idx].device_id *
2536					VXGE_HW_VPATH_MSIX_ACTIVE;
2537				vxge_hw_vpath_msix_unmask(
2538					vdev->vpaths[vp_idx].handle,
2539					msix_idx);
2540				intr_cnt++;
2541			}
2542
2543			/* Point to next vpath handler */
2544			if (((intr_idx + 1) % VXGE_HW_VPATH_MSIX_ACTIVE == 0) &&
2545			    (vp_idx < (vdev->no_of_vpath - 1)))
2546				vp_idx++;
2547		}
2548
2549		intr_cnt = vdev->no_of_vpath * 2;
2550		snprintf(vdev->desc[intr_cnt], VXGE_INTR_STRLEN,
2551			"%s:vxge:MSI-X %d - Alarm - fn:%d",
2552			vdev->ndev->name,
2553			vdev->entries[intr_cnt].entry,
2554			pci_fun);
2555		/* For Alarm interrupts */
2556		ret = request_irq(vdev->entries[intr_cnt].vector,
2557					vxge_alarm_msix_handle, 0,
2558					vdev->desc[intr_cnt],
2559					&vdev->vpaths[0]);
2560		if (ret) {
2561			vxge_debug_init(VXGE_ERR,
2562				"%s: MSIX - %d Registration failed",
2563				vdev->ndev->name, intr_cnt);
2564			vxge_rem_msix_isr(vdev);
2565			vdev->config.intr_type = INTA;
2566			vxge_debug_init(VXGE_ERR,
2567				"%s: Defaulting to INTA",
2568				vdev->ndev->name);
2569				goto INTA_MODE;
2570		}
2571
2572		msix_idx = (vdev->vpaths[0].handle->vpath->vp_id *
2573			VXGE_HW_VPATH_MSIX_ACTIVE) + VXGE_ALARM_MSIX_ID;
2574		vxge_hw_vpath_msix_unmask(vdev->vpaths[vp_idx].handle,
2575					msix_idx);
2576		vdev->vxge_entries[intr_cnt].in_use = 1;
2577		vdev->vxge_entries[intr_cnt].arg = &vdev->vpaths[0];
2578	}
2579INTA_MODE:
2580#endif
2581
2582	if (vdev->config.intr_type == INTA) {
2583		snprintf(vdev->desc[0], VXGE_INTR_STRLEN,
2584			"%s:vxge:INTA", vdev->ndev->name);
2585		vxge_hw_device_set_intr_type(vdev->devh,
2586			VXGE_HW_INTR_MODE_IRQLINE);
2587
2588		vxge_hw_vpath_tti_ci_set(vdev->vpaths[0].fifo.handle);
2589
2590		ret = request_irq((int) vdev->pdev->irq,
2591			vxge_isr_napi,
2592			IRQF_SHARED, vdev->desc[0], vdev);
2593		if (ret) {
2594			vxge_debug_init(VXGE_ERR,
2595				"%s %s-%d: ISR registration failed",
2596				VXGE_DRIVER_NAME, "IRQ", vdev->pdev->irq);
2597			return -ENODEV;
2598		}
2599		vxge_debug_init(VXGE_TRACE,
2600			"new %s-%d line allocated",
2601			"IRQ", vdev->pdev->irq);
2602	}
2603
2604	return VXGE_HW_OK;
2605}
2606
2607static void vxge_poll_vp_reset(unsigned long data)
2608{
2609	struct vxgedev *vdev = (struct vxgedev *)data;
2610	int i, j = 0;
2611
2612	for (i = 0; i < vdev->no_of_vpath; i++) {
2613		if (test_bit(i, &vdev->vp_reset)) {
2614			vxge_reset_vpath(vdev, i);
2615			j++;
2616		}
2617	}
2618	if (j && (vdev->config.intr_type != MSI_X)) {
2619		vxge_hw_device_unmask_all(vdev->devh);
2620		vxge_hw_device_flush_io(vdev->devh);
2621	}
2622
2623	mod_timer(&vdev->vp_reset_timer, jiffies + HZ / 2);
2624}
2625
2626static void vxge_poll_vp_lockup(unsigned long data)
2627{
2628	struct vxgedev *vdev = (struct vxgedev *)data;
2629	enum vxge_hw_status status = VXGE_HW_OK;
2630	struct vxge_vpath *vpath;
2631	struct vxge_ring *ring;
2632	int i;
2633	unsigned long rx_frms;
2634
2635	for (i = 0; i < vdev->no_of_vpath; i++) {
2636		ring = &vdev->vpaths[i].ring;
2637
2638		/* Truncated to machine word size number of frames */
2639		rx_frms = ACCESS_ONCE(ring->stats.rx_frms);
2640
2641		/* Did this vpath received any packets */
2642		if (ring->stats.prev_rx_frms == rx_frms) {
2643			status = vxge_hw_vpath_check_leak(ring->handle);
2644
2645			/* Did it received any packets last time */
2646			if ((VXGE_HW_FAIL == status) &&
2647				(VXGE_HW_FAIL == ring->last_status)) {
2648
2649				/* schedule vpath reset */
2650				if (!test_and_set_bit(i, &vdev->vp_reset)) {
2651					vpath = &vdev->vpaths[i];
2652
2653					/* disable interrupts for this vpath */
2654					vxge_vpath_intr_disable(vdev, i);
2655
2656					/* stop the queue for this vpath */
2657					netif_tx_stop_queue(vpath->fifo.txq);
2658					continue;
2659				}
2660			}
2661		}
2662		ring->stats.prev_rx_frms = rx_frms;
2663		ring->last_status = status;
2664	}
2665
2666	/* Check every 1 milli second */
2667	mod_timer(&vdev->vp_lockup_timer, jiffies + HZ / 1000);
2668}
2669
2670static netdev_features_t vxge_fix_features(struct net_device *dev,
2671	netdev_features_t features)
2672{
2673	netdev_features_t changed = dev->features ^ features;
2674
2675	/* Enabling RTH requires some of the logic in vxge_device_register and a
2676	 * vpath reset.  Due to these restrictions, only allow modification
2677	 * while the interface is down.
2678	 */
2679	if ((changed & NETIF_F_RXHASH) && netif_running(dev))
2680		features ^= NETIF_F_RXHASH;
2681
2682	return features;
2683}
2684
2685static int vxge_set_features(struct net_device *dev, netdev_features_t features)
2686{
2687	struct vxgedev *vdev = netdev_priv(dev);
2688	netdev_features_t changed = dev->features ^ features;
2689
2690	if (!(changed & NETIF_F_RXHASH))
2691		return 0;
2692
2693	/* !netif_running() ensured by vxge_fix_features() */
2694
2695	vdev->devh->config.rth_en = !!(features & NETIF_F_RXHASH);
2696	if (vxge_reset_all_vpaths(vdev) != VXGE_HW_OK) {
2697		dev->features = features ^ NETIF_F_RXHASH;
2698		vdev->devh->config.rth_en = !!(dev->features & NETIF_F_RXHASH);
2699		return -EIO;
2700	}
2701
2702	return 0;
2703}
2704
2705/**
2706 * vxge_open
2707 * @dev: pointer to the device structure.
2708 *
2709 * This function is the open entry point of the driver. It mainly calls a
2710 * function to allocate Rx buffers and inserts them into the buffer
2711 * descriptors and then enables the Rx part of the NIC.
2712 * Return value: '0' on success and an appropriate (-)ve integer as
2713 * defined in errno.h file on failure.
2714 */
2715static int vxge_open(struct net_device *dev)
2716{
2717	enum vxge_hw_status status;
2718	struct vxgedev *vdev;
2719	struct __vxge_hw_device *hldev;
2720	struct vxge_vpath *vpath;
2721	int ret = 0;
2722	int i;
2723	u64 val64, function_mode;
2724
2725	vxge_debug_entryexit(VXGE_TRACE,
2726		"%s: %s:%d", dev->name, __func__, __LINE__);
2727
2728	vdev = netdev_priv(dev);
2729	hldev = pci_get_drvdata(vdev->pdev);
2730	function_mode = vdev->config.device_hw_info.function_mode;
2731
2732	/* make sure you have link off by default every time Nic is
2733	 * initialized */
2734	netif_carrier_off(dev);
2735
2736	/* Open VPATHs */
2737	status = vxge_open_vpaths(vdev);
2738	if (status != VXGE_HW_OK) {
2739		vxge_debug_init(VXGE_ERR,
2740			"%s: fatal: Vpath open failed", vdev->ndev->name);
2741		ret = -EPERM;
2742		goto out0;
2743	}
2744
2745	vdev->mtu = dev->mtu;
2746
2747	status = vxge_add_isr(vdev);
2748	if (status != VXGE_HW_OK) {
2749		vxge_debug_init(VXGE_ERR,
2750			"%s: fatal: ISR add failed", dev->name);
2751		ret = -EPERM;
2752		goto out1;
2753	}
2754
2755	if (vdev->config.intr_type != MSI_X) {
2756		netif_napi_add(dev, &vdev->napi, vxge_poll_inta,
2757			vdev->config.napi_weight);
2758		napi_enable(&vdev->napi);
2759		for (i = 0; i < vdev->no_of_vpath; i++) {
2760			vpath = &vdev->vpaths[i];
2761			vpath->ring.napi_p = &vdev->napi;
2762		}
2763	} else {
2764		for (i = 0; i < vdev->no_of_vpath; i++) {
2765			vpath = &vdev->vpaths[i];
2766			netif_napi_add(dev, &vpath->ring.napi,
2767			    vxge_poll_msix, vdev->config.napi_weight);
2768			napi_enable(&vpath->ring.napi);
2769			vpath->ring.napi_p = &vpath->ring.napi;
2770		}
2771	}
2772
2773	/* configure RTH */
2774	if (vdev->config.rth_steering) {
2775		status = vxge_rth_configure(vdev);
2776		if (status != VXGE_HW_OK) {
2777			vxge_debug_init(VXGE_ERR,
2778				"%s: fatal: RTH configuration failed",
2779				dev->name);
2780			ret = -EPERM;
2781			goto out2;
2782		}
2783	}
2784	printk(KERN_INFO "%s: Receive Hashing Offload %s\n", dev->name,
2785	       hldev->config.rth_en ? "enabled" : "disabled");
2786
2787	for (i = 0; i < vdev->no_of_vpath; i++) {
2788		vpath = &vdev->vpaths[i];
2789
2790		/* set initial mtu before enabling the device */
2791		status = vxge_hw_vpath_mtu_set(vpath->handle, vdev->mtu);
2792		if (status != VXGE_HW_OK) {
2793			vxge_debug_init(VXGE_ERR,
2794				"%s: fatal: can not set new MTU", dev->name);
2795			ret = -EPERM;
2796			goto out2;
2797		}
2798	}
2799
2800	VXGE_DEVICE_DEBUG_LEVEL_SET(VXGE_TRACE, VXGE_COMPONENT_LL, vdev);
2801	vxge_debug_init(vdev->level_trace,
2802		"%s: MTU is %d", vdev->ndev->name, vdev->mtu);
2803	VXGE_DEVICE_DEBUG_LEVEL_SET(VXGE_ERR, VXGE_COMPONENT_LL, vdev);
2804
2805	/* Restore the DA, VID table and also multicast and promiscuous mode
2806	 * states
2807	 */
2808	if (vdev->all_multi_flg) {
2809		for (i = 0; i < vdev->no_of_vpath; i++) {
2810			vpath = &vdev->vpaths[i];
2811			vxge_restore_vpath_mac_addr(vpath);
2812			vxge_restore_vpath_vid_table(vpath);
2813
2814			status = vxge_hw_vpath_mcast_enable(vpath->handle);
2815			if (status != VXGE_HW_OK)
2816				vxge_debug_init(VXGE_ERR,
2817					"%s:%d Enabling multicast failed",
2818					__func__, __LINE__);
2819		}
2820	}
2821
2822	/* Enable vpath to sniff all unicast/multicast traffic that not
2823	 * addressed to them. We allow promiscuous mode for PF only
2824	 */
2825
2826	val64 = 0;
2827	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
2828		val64 |= VXGE_HW_RXMAC_AUTHORIZE_ALL_ADDR_VP(i);
2829
2830	vxge_hw_mgmt_reg_write(vdev->devh,
2831		vxge_hw_mgmt_reg_type_mrpcim,
2832		0,
2833		(ulong)offsetof(struct vxge_hw_mrpcim_reg,
2834			rxmac_authorize_all_addr),
2835		val64);
2836
2837	vxge_hw_mgmt_reg_write(vdev->devh,
2838		vxge_hw_mgmt_reg_type_mrpcim,
2839		0,
2840		(ulong)offsetof(struct vxge_hw_mrpcim_reg,
2841			rxmac_authorize_all_vid),
2842		val64);
2843
2844	vxge_set_multicast(dev);
2845
2846	/* Enabling Bcast and mcast for all vpath */
2847	for (i = 0; i < vdev->no_of_vpath; i++) {
2848		vpath = &vdev->vpaths[i];
2849		status = vxge_hw_vpath_bcast_enable(vpath->handle);
2850		if (status != VXGE_HW_OK)
2851			vxge_debug_init(VXGE_ERR,
2852				"%s : Can not enable bcast for vpath "
2853				"id %d", dev->name, i);
2854		if (vdev->config.addr_learn_en) {
2855			status = vxge_hw_vpath_mcast_enable(vpath->handle);
2856			if (status != VXGE_HW_OK)
2857				vxge_debug_init(VXGE_ERR,
2858					"%s : Can not enable mcast for vpath "
2859					"id %d", dev->name, i);
2860		}
2861	}
2862
2863	vxge_hw_device_setpause_data(vdev->devh, 0,
2864		vdev->config.tx_pause_enable,
2865		vdev->config.rx_pause_enable);
2866
2867	if (vdev->vp_reset_timer.function == NULL)
2868		vxge_os_timer(&vdev->vp_reset_timer, vxge_poll_vp_reset, vdev,
2869			      HZ / 2);
2870
2871	/* There is no need to check for RxD leak and RxD lookup on Titan1A */
2872	if (vdev->titan1 && vdev->vp_lockup_timer.function == NULL)
2873		vxge_os_timer(&vdev->vp_lockup_timer, vxge_poll_vp_lockup, vdev,
2874			      HZ / 2);
2875
2876	set_bit(__VXGE_STATE_CARD_UP, &vdev->state);
2877
2878	smp_wmb();
2879
2880	if (vxge_hw_device_link_state_get(vdev->devh) == VXGE_HW_LINK_UP) {
2881		netif_carrier_on(vdev->ndev);
2882		netdev_notice(vdev->ndev, "Link Up\n");
2883		vdev->stats.link_up++;
2884	}
2885
2886	vxge_hw_device_intr_enable(vdev->devh);
2887
2888	smp_wmb();
2889
2890	for (i = 0; i < vdev->no_of_vpath; i++) {
2891		vpath = &vdev->vpaths[i];
2892
2893		vxge_hw_vpath_enable(vpath->handle);
2894		smp_wmb();
2895		vxge_hw_vpath_rx_doorbell_init(vpath->handle);
2896	}
2897
2898	netif_tx_start_all_queues(vdev->ndev);
2899
2900	/* configure CI */
2901	vxge_config_ci_for_tti_rti(vdev);
2902
2903	goto out0;
2904
2905out2:
2906	vxge_rem_isr(vdev);
2907
2908	/* Disable napi */
2909	if (vdev->config.intr_type != MSI_X)
2910		napi_disable(&vdev->napi);
2911	else {
2912		for (i = 0; i < vdev->no_of_vpath; i++)
2913			napi_disable(&vdev->vpaths[i].ring.napi);
2914	}
2915
2916out1:
2917	vxge_close_vpaths(vdev, 0);
2918out0:
2919	vxge_debug_entryexit(VXGE_TRACE,
2920				"%s: %s:%d  Exiting...",
2921				dev->name, __func__, __LINE__);
2922	return ret;
2923}
2924
2925/* Loop through the mac address list and delete all the entries */
2926static void vxge_free_mac_add_list(struct vxge_vpath *vpath)
2927{
2928
2929	struct list_head *entry, *next;
2930	if (list_empty(&vpath->mac_addr_list))
2931		return;
2932
2933	list_for_each_safe(entry, next, &vpath->mac_addr_list) {
2934		list_del(entry);
2935		kfree((struct vxge_mac_addrs *)entry);
2936	}
2937}
2938
2939static void vxge_napi_del_all(struct vxgedev *vdev)
2940{
2941	int i;
2942	if (vdev->config.intr_type != MSI_X)
2943		netif_napi_del(&vdev->napi);
2944	else {
2945		for (i = 0; i < vdev->no_of_vpath; i++)
2946			netif_napi_del(&vdev->vpaths[i].ring.napi);
2947	}
2948}
2949
2950static int do_vxge_close(struct net_device *dev, int do_io)
2951{
2952	enum vxge_hw_status status;
2953	struct vxgedev *vdev;
2954	struct __vxge_hw_device *hldev;
2955	int i;
2956	u64 val64, vpath_vector;
2957	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
2958		dev->name, __func__, __LINE__);
2959
2960	vdev = netdev_priv(dev);
2961	hldev = pci_get_drvdata(vdev->pdev);
2962
2963	if (unlikely(!is_vxge_card_up(vdev)))
2964		return 0;
2965
2966	/* If vxge_handle_crit_err task is executing,
2967	 * wait till it completes. */
2968	while (test_and_set_bit(__VXGE_STATE_RESET_CARD, &vdev->state))
2969		msleep(50);
2970
2971	if (do_io) {
2972		/* Put the vpath back in normal mode */
2973		vpath_vector = vxge_mBIT(vdev->vpaths[0].device_id);
2974		status = vxge_hw_mgmt_reg_read(vdev->devh,
2975				vxge_hw_mgmt_reg_type_mrpcim,
2976				0,
2977				(ulong)offsetof(
2978					struct vxge_hw_mrpcim_reg,
2979					rts_mgr_cbasin_cfg),
2980				&val64);
2981		if (status == VXGE_HW_OK) {
2982			val64 &= ~vpath_vector;
2983			status = vxge_hw_mgmt_reg_write(vdev->devh,
2984					vxge_hw_mgmt_reg_type_mrpcim,
2985					0,
2986					(ulong)offsetof(
2987						struct vxge_hw_mrpcim_reg,
2988						rts_mgr_cbasin_cfg),
2989					val64);
2990		}
2991
2992		/* Remove the function 0 from promiscuous mode */
2993		vxge_hw_mgmt_reg_write(vdev->devh,
2994			vxge_hw_mgmt_reg_type_mrpcim,
2995			0,
2996			(ulong)offsetof(struct vxge_hw_mrpcim_reg,
2997				rxmac_authorize_all_addr),
2998			0);
2999
3000		vxge_hw_mgmt_reg_write(vdev->devh,
3001			vxge_hw_mgmt_reg_type_mrpcim,
3002			0,
3003			(ulong)offsetof(struct vxge_hw_mrpcim_reg,
3004				rxmac_authorize_all_vid),
3005			0);
3006
3007		smp_wmb();
3008	}
3009
3010	if (vdev->titan1)
3011		del_timer_sync(&vdev->vp_lockup_timer);
3012
3013	del_timer_sync(&vdev->vp_reset_timer);
3014
3015	if (do_io)
3016		vxge_hw_device_wait_receive_idle(hldev);
3017
3018	clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3019
3020	/* Disable napi */
3021	if (vdev->config.intr_type != MSI_X)
3022		napi_disable(&vdev->napi);
3023	else {
3024		for (i = 0; i < vdev->no_of_vpath; i++)
3025			napi_disable(&vdev->vpaths[i].ring.napi);
3026	}
3027
3028	netif_carrier_off(vdev->ndev);
3029	netdev_notice(vdev->ndev, "Link Down\n");
3030	netif_tx_stop_all_queues(vdev->ndev);
3031
3032	/* Note that at this point xmit() is stopped by upper layer */
3033	if (do_io)
3034		vxge_hw_device_intr_disable(vdev->devh);
3035
3036	vxge_rem_isr(vdev);
3037
3038	vxge_napi_del_all(vdev);
3039
3040	if (do_io)
3041		vxge_reset_all_vpaths(vdev);
3042
3043	vxge_close_vpaths(vdev, 0);
3044
3045	vxge_debug_entryexit(VXGE_TRACE,
3046		"%s: %s:%d  Exiting...", dev->name, __func__, __LINE__);
3047
3048	clear_bit(__VXGE_STATE_RESET_CARD, &vdev->state);
3049
3050	return 0;
3051}
3052
3053/**
3054 * vxge_close
3055 * @dev: device pointer.
3056 *
3057 * This is the stop entry point of the driver. It needs to undo exactly
3058 * whatever was done by the open entry point, thus it's usually referred to
3059 * as the close function.Among other things this function mainly stops the
3060 * Rx side of the NIC and frees all the Rx buffers in the Rx rings.
3061 * Return value: '0' on success and an appropriate (-)ve integer as
3062 * defined in errno.h file on failure.
3063 */
3064static int vxge_close(struct net_device *dev)
3065{
3066	do_vxge_close(dev, 1);
3067	return 0;
3068}
3069
3070/**
3071 * vxge_change_mtu
3072 * @dev: net device pointer.
3073 * @new_mtu :the new MTU size for the device.
3074 *
3075 * A driver entry point to change MTU size for the device. Before changing
3076 * the MTU the device must be stopped.
3077 */
3078static int vxge_change_mtu(struct net_device *dev, int new_mtu)
3079{
3080	struct vxgedev *vdev = netdev_priv(dev);
3081
3082	vxge_debug_entryexit(vdev->level_trace,
3083		"%s:%d", __func__, __LINE__);
3084	if ((new_mtu < VXGE_HW_MIN_MTU) || (new_mtu > VXGE_HW_MAX_MTU)) {
3085		vxge_debug_init(vdev->level_err,
3086			"%s: mtu size is invalid", dev->name);
3087		return -EPERM;
3088	}
3089
3090	/* check if device is down already */
3091	if (unlikely(!is_vxge_card_up(vdev))) {
3092		/* just store new value, will use later on open() */
3093		dev->mtu = new_mtu;
3094		vxge_debug_init(vdev->level_err,
3095			"%s", "device is down on MTU change");
3096		return 0;
3097	}
3098
3099	vxge_debug_init(vdev->level_trace,
3100		"trying to apply new MTU %d", new_mtu);
3101
3102	if (vxge_close(dev))
3103		return -EIO;
3104
3105	dev->mtu = new_mtu;
3106	vdev->mtu = new_mtu;
3107
3108	if (vxge_open(dev))
3109		return -EIO;
3110
3111	vxge_debug_init(vdev->level_trace,
3112		"%s: MTU changed to %d", vdev->ndev->name, new_mtu);
3113
3114	vxge_debug_entryexit(vdev->level_trace,
3115		"%s:%d  Exiting...", __func__, __LINE__);
3116
3117	return 0;
3118}
3119
3120/**
3121 * vxge_get_stats64
3122 * @dev: pointer to the device structure
3123 * @stats: pointer to struct rtnl_link_stats64
3124 *
3125 */
3126static struct rtnl_link_stats64 *
3127vxge_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *net_stats)
3128{
3129	struct vxgedev *vdev = netdev_priv(dev);
3130	int k;
3131
3132	/* net_stats already zeroed by caller */
3133	for (k = 0; k < vdev->no_of_vpath; k++) {
3134		struct vxge_ring_stats *rxstats = &vdev->vpaths[k].ring.stats;
3135		struct vxge_fifo_stats *txstats = &vdev->vpaths[k].fifo.stats;
3136		unsigned int start;
3137		u64 packets, bytes, multicast;
3138
3139		do {
3140			start = u64_stats_fetch_begin_irq(&rxstats->syncp);
3141
3142			packets   = rxstats->rx_frms;
3143			multicast = rxstats->rx_mcast;
3144			bytes     = rxstats->rx_bytes;
3145		} while (u64_stats_fetch_retry_irq(&rxstats->syncp, start));
3146
3147		net_stats->rx_packets += packets;
3148		net_stats->rx_bytes += bytes;
3149		net_stats->multicast += multicast;
3150
3151		net_stats->rx_errors += rxstats->rx_errors;
3152		net_stats->rx_dropped += rxstats->rx_dropped;
3153
3154		do {
3155			start = u64_stats_fetch_begin_irq(&txstats->syncp);
3156
3157			packets = txstats->tx_frms;
3158			bytes   = txstats->tx_bytes;
3159		} while (u64_stats_fetch_retry_irq(&txstats->syncp, start));
3160
3161		net_stats->tx_packets += packets;
3162		net_stats->tx_bytes += bytes;
3163		net_stats->tx_errors += txstats->tx_errors;
3164	}
3165
3166	return net_stats;
3167}
3168
3169static enum vxge_hw_status vxge_timestamp_config(struct __vxge_hw_device *devh)
3170{
3171	enum vxge_hw_status status;
3172	u64 val64;
3173
3174	/* Timestamp is passed to the driver via the FCS, therefore we
3175	 * must disable the FCS stripping by the adapter.  Since this is
3176	 * required for the driver to load (due to a hardware bug),
3177	 * there is no need to do anything special here.
3178	 */
3179	val64 = VXGE_HW_XMAC_TIMESTAMP_EN |
3180		VXGE_HW_XMAC_TIMESTAMP_USE_LINK_ID(0) |
3181		VXGE_HW_XMAC_TIMESTAMP_INTERVAL(0);
3182
3183	status = vxge_hw_mgmt_reg_write(devh,
3184					vxge_hw_mgmt_reg_type_mrpcim,
3185					0,
3186					offsetof(struct vxge_hw_mrpcim_reg,
3187						 xmac_timestamp),
3188					val64);
3189	vxge_hw_device_flush_io(devh);
3190	devh->config.hwts_en = VXGE_HW_HWTS_ENABLE;
3191	return status;
3192}
3193
3194static int vxge_hwtstamp_set(struct vxgedev *vdev, void __user *data)
3195{
3196	struct hwtstamp_config config;
3197	int i;
3198
3199	if (copy_from_user(&config, data, sizeof(config)))
3200		return -EFAULT;
3201
3202	/* reserved for future extensions */
3203	if (config.flags)
3204		return -EINVAL;
3205
3206	/* Transmit HW Timestamp not supported */
3207	switch (config.tx_type) {
3208	case HWTSTAMP_TX_OFF:
3209		break;
3210	case HWTSTAMP_TX_ON:
3211	default:
3212		return -ERANGE;
3213	}
3214
3215	switch (config.rx_filter) {
3216	case HWTSTAMP_FILTER_NONE:
3217		vdev->rx_hwts = 0;
3218		config.rx_filter = HWTSTAMP_FILTER_NONE;
3219		break;
3220
3221	case HWTSTAMP_FILTER_ALL:
3222	case HWTSTAMP_FILTER_SOME:
3223	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3224	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3225	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3226	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3227	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3228	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3229	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3230	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3231	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3232	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3233	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3234	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3235		if (vdev->devh->config.hwts_en != VXGE_HW_HWTS_ENABLE)
3236			return -EFAULT;
3237
3238		vdev->rx_hwts = 1;
3239		config.rx_filter = HWTSTAMP_FILTER_ALL;
3240		break;
3241
3242	default:
3243		 return -ERANGE;
3244	}
3245
3246	for (i = 0; i < vdev->no_of_vpath; i++)
3247		vdev->vpaths[i].ring.rx_hwts = vdev->rx_hwts;
3248
3249	if (copy_to_user(data, &config, sizeof(config)))
3250		return -EFAULT;
3251
3252	return 0;
3253}
3254
3255static int vxge_hwtstamp_get(struct vxgedev *vdev, void __user *data)
3256{
3257	struct hwtstamp_config config;
3258
3259	config.flags = 0;
3260	config.tx_type = HWTSTAMP_TX_OFF;
3261	config.rx_filter = (vdev->rx_hwts ?
3262			    HWTSTAMP_FILTER_ALL : HWTSTAMP_FILTER_NONE);
3263
3264	if (copy_to_user(data, &config, sizeof(config)))
3265		return -EFAULT;
3266
3267	return 0;
3268}
3269
3270/**
3271 * vxge_ioctl
3272 * @dev: Device pointer.
3273 * @ifr: An IOCTL specific structure, that can contain a pointer to
3274 *       a proprietary structure used to pass information to the driver.
3275 * @cmd: This is used to distinguish between the different commands that
3276 *       can be passed to the IOCTL functions.
3277 *
3278 * Entry point for the Ioctl.
3279 */
3280static int vxge_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
3281{
3282	struct vxgedev *vdev = netdev_priv(dev);
3283
3284	switch (cmd) {
3285	case SIOCSHWTSTAMP:
3286		return vxge_hwtstamp_set(vdev, rq->ifr_data);
3287	case SIOCGHWTSTAMP:
3288		return vxge_hwtstamp_get(vdev, rq->ifr_data);
3289	default:
3290		return -EOPNOTSUPP;
3291	}
3292}
3293
3294/**
3295 * vxge_tx_watchdog
3296 * @dev: pointer to net device structure
3297 *
3298 * Watchdog for transmit side.
3299 * This function is triggered if the Tx Queue is stopped
3300 * for a pre-defined amount of time when the Interface is still up.
3301 */
3302static void vxge_tx_watchdog(struct net_device *dev)
3303{
3304	struct vxgedev *vdev;
3305
3306	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
3307
3308	vdev = netdev_priv(dev);
3309
3310	vdev->cric_err_event = VXGE_HW_EVENT_RESET_START;
3311
3312	schedule_work(&vdev->reset_task);
3313	vxge_debug_entryexit(VXGE_TRACE,
3314		"%s:%d  Exiting...", __func__, __LINE__);
3315}
3316
3317/**
3318 * vxge_vlan_rx_add_vid
3319 * @dev: net device pointer.
3320 * @proto: vlan protocol
3321 * @vid: vid
3322 *
3323 * Add the vlan id to the devices vlan id table
3324 */
3325static int
3326vxge_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid)
3327{
3328	struct vxgedev *vdev = netdev_priv(dev);
3329	struct vxge_vpath *vpath;
3330	int vp_id;
3331
3332	/* Add these vlan to the vid table */
3333	for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
3334		vpath = &vdev->vpaths[vp_id];
3335		if (!vpath->is_open)
3336			continue;
3337		vxge_hw_vpath_vid_add(vpath->handle, vid);
3338	}
3339	set_bit(vid, vdev->active_vlans);
3340	return 0;
3341}
3342
3343/**
3344 * vxge_vlan_rx_kill_vid
3345 * @dev: net device pointer.
3346 * @proto: vlan protocol
3347 * @vid: vid
3348 *
3349 * Remove the vlan id from the device's vlan id table
3350 */
3351static int
3352vxge_vlan_rx_kill_vid(struct net_device *dev, __be16 proto, u16 vid)
3353{
3354	struct vxgedev *vdev = netdev_priv(dev);
3355	struct vxge_vpath *vpath;
3356	int vp_id;
3357
3358	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
3359
3360	/* Delete this vlan from the vid table */
3361	for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
3362		vpath = &vdev->vpaths[vp_id];
3363		if (!vpath->is_open)
3364			continue;
3365		vxge_hw_vpath_vid_delete(vpath->handle, vid);
3366	}
3367	vxge_debug_entryexit(VXGE_TRACE,
3368		"%s:%d  Exiting...", __func__, __LINE__);
3369	clear_bit(vid, vdev->active_vlans);
3370	return 0;
3371}
3372
3373static const struct net_device_ops vxge_netdev_ops = {
3374	.ndo_open               = vxge_open,
3375	.ndo_stop               = vxge_close,
3376	.ndo_get_stats64        = vxge_get_stats64,
3377	.ndo_start_xmit         = vxge_xmit,
3378	.ndo_validate_addr      = eth_validate_addr,
3379	.ndo_set_rx_mode	= vxge_set_multicast,
3380	.ndo_do_ioctl           = vxge_ioctl,
3381	.ndo_set_mac_address    = vxge_set_mac_addr,
3382	.ndo_change_mtu         = vxge_change_mtu,
3383	.ndo_fix_features	= vxge_fix_features,
3384	.ndo_set_features	= vxge_set_features,
3385	.ndo_vlan_rx_kill_vid   = vxge_vlan_rx_kill_vid,
3386	.ndo_vlan_rx_add_vid	= vxge_vlan_rx_add_vid,
3387	.ndo_tx_timeout         = vxge_tx_watchdog,
3388#ifdef CONFIG_NET_POLL_CONTROLLER
3389	.ndo_poll_controller    = vxge_netpoll,
3390#endif
3391};
3392
3393static int vxge_device_register(struct __vxge_hw_device *hldev,
3394				struct vxge_config *config, int high_dma,
3395				int no_of_vpath, struct vxgedev **vdev_out)
3396{
3397	struct net_device *ndev;
3398	enum vxge_hw_status status = VXGE_HW_OK;
3399	struct vxgedev *vdev;
3400	int ret = 0, no_of_queue = 1;
3401	u64 stat;
3402
3403	*vdev_out = NULL;
3404	if (config->tx_steering_type)
3405		no_of_queue = no_of_vpath;
3406
3407	ndev = alloc_etherdev_mq(sizeof(struct vxgedev),
3408			no_of_queue);
3409	if (ndev == NULL) {
3410		vxge_debug_init(
3411			vxge_hw_device_trace_level_get(hldev),
3412		"%s : device allocation failed", __func__);
3413		ret = -ENODEV;
3414		goto _out0;
3415	}
3416
3417	vxge_debug_entryexit(
3418		vxge_hw_device_trace_level_get(hldev),
3419		"%s: %s:%d  Entering...",
3420		ndev->name, __func__, __LINE__);
3421
3422	vdev = netdev_priv(ndev);
3423	memset(vdev, 0, sizeof(struct vxgedev));
3424
3425	vdev->ndev = ndev;
3426	vdev->devh = hldev;
3427	vdev->pdev = hldev->pdev;
3428	memcpy(&vdev->config, config, sizeof(struct vxge_config));
3429	vdev->rx_hwts = 0;
3430	vdev->titan1 = (vdev->pdev->revision == VXGE_HW_TITAN1_PCI_REVISION);
3431
3432	SET_NETDEV_DEV(ndev, &vdev->pdev->dev);
3433
3434	ndev->hw_features = NETIF_F_RXCSUM | NETIF_F_SG |
3435		NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
3436		NETIF_F_TSO | NETIF_F_TSO6 |
3437		NETIF_F_HW_VLAN_CTAG_TX;
3438	if (vdev->config.rth_steering != NO_STEERING)
3439		ndev->hw_features |= NETIF_F_RXHASH;
3440
3441	ndev->features |= ndev->hw_features |
3442		NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
3443
3444
3445	ndev->netdev_ops = &vxge_netdev_ops;
3446
3447	ndev->watchdog_timeo = VXGE_LL_WATCH_DOG_TIMEOUT;
3448	INIT_WORK(&vdev->reset_task, vxge_reset);
3449
3450	vxge_initialize_ethtool_ops(ndev);
3451
3452	/* Allocate memory for vpath */
3453	vdev->vpaths = kzalloc((sizeof(struct vxge_vpath)) *
3454				no_of_vpath, GFP_KERNEL);
3455	if (!vdev->vpaths) {
3456		vxge_debug_init(VXGE_ERR,
3457			"%s: vpath memory allocation failed",
3458			vdev->ndev->name);
3459		ret = -ENOMEM;
3460		goto _out1;
3461	}
3462
3463	vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3464		"%s : checksumming enabled", __func__);
3465
3466	if (high_dma) {
3467		ndev->features |= NETIF_F_HIGHDMA;
3468		vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3469			"%s : using High DMA", __func__);
3470	}
3471
3472	ret = register_netdev(ndev);
3473	if (ret) {
3474		vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3475			"%s: %s : device registration failed!",
3476			ndev->name, __func__);
3477		goto _out2;
3478	}
3479
3480	/*  Set the factory defined MAC address initially */
3481	ndev->addr_len = ETH_ALEN;
3482
3483	/* Make Link state as off at this point, when the Link change
3484	 * interrupt comes the state will be automatically changed to
3485	 * the right state.
3486	 */
3487	netif_carrier_off(ndev);
3488
3489	vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3490		"%s: Ethernet device registered",
3491		ndev->name);
3492
3493	hldev->ndev = ndev;
3494	*vdev_out = vdev;
3495
3496	/* Resetting the Device stats */
3497	status = vxge_hw_mrpcim_stats_access(
3498				hldev,
3499				VXGE_HW_STATS_OP_CLEAR_ALL_STATS,
3500				0,
3501				0,
3502				&stat);
3503
3504	if (status == VXGE_HW_ERR_PRIVILAGED_OPEARATION)
3505		vxge_debug_init(
3506			vxge_hw_device_trace_level_get(hldev),
3507			"%s: device stats clear returns"
3508			"VXGE_HW_ERR_PRIVILAGED_OPEARATION", ndev->name);
3509
3510	vxge_debug_entryexit(vxge_hw_device_trace_level_get(hldev),
3511		"%s: %s:%d  Exiting...",
3512		ndev->name, __func__, __LINE__);
3513
3514	return ret;
3515_out2:
3516	kfree(vdev->vpaths);
3517_out1:
3518	free_netdev(ndev);
3519_out0:
3520	return ret;
3521}
3522
3523/*
3524 * vxge_device_unregister
3525 *
3526 * This function will unregister and free network device
3527 */
3528static void vxge_device_unregister(struct __vxge_hw_device *hldev)
3529{
3530	struct vxgedev *vdev;
3531	struct net_device *dev;
3532	char buf[IFNAMSIZ];
3533
3534	dev = hldev->ndev;
3535	vdev = netdev_priv(dev);
3536
3537	vxge_debug_entryexit(vdev->level_trace,	"%s: %s:%d", vdev->ndev->name,
3538			     __func__, __LINE__);
3539
3540	strlcpy(buf, dev->name, IFNAMSIZ);
3541
3542	flush_work(&vdev->reset_task);
3543
3544	/* in 2.6 will call stop() if device is up */
3545	unregister_netdev(dev);
3546
3547	kfree(vdev->vpaths);
3548
3549	/* we are safe to free it now */
3550	free_netdev(dev);
3551
3552	vxge_debug_init(vdev->level_trace, "%s: ethernet device unregistered",
3553			buf);
3554	vxge_debug_entryexit(vdev->level_trace,	"%s: %s:%d  Exiting...", buf,
3555			     __func__, __LINE__);
3556}
3557
3558/*
3559 * vxge_callback_crit_err
3560 *
3561 * This function is called by the alarm handler in interrupt context.
3562 * Driver must analyze it based on the event type.
3563 */
3564static void
3565vxge_callback_crit_err(struct __vxge_hw_device *hldev,
3566			enum vxge_hw_event type, u64 vp_id)
3567{
3568	struct net_device *dev = hldev->ndev;
3569	struct vxgedev *vdev = netdev_priv(dev);
3570	struct vxge_vpath *vpath = NULL;
3571	int vpath_idx;
3572
3573	vxge_debug_entryexit(vdev->level_trace,
3574		"%s: %s:%d", vdev->ndev->name, __func__, __LINE__);
3575
3576	/* Note: This event type should be used for device wide
3577	 * indications only - Serious errors, Slot freeze and critical errors
3578	 */
3579	vdev->cric_err_event = type;
3580
3581	for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
3582		vpath = &vdev->vpaths[vpath_idx];
3583		if (vpath->device_id == vp_id)
3584			break;
3585	}
3586
3587	if (!test_bit(__VXGE_STATE_RESET_CARD, &vdev->state)) {
3588		if (type == VXGE_HW_EVENT_SLOT_FREEZE) {
3589			vxge_debug_init(VXGE_ERR,
3590				"%s: Slot is frozen", vdev->ndev->name);
3591		} else if (type == VXGE_HW_EVENT_SERR) {
3592			vxge_debug_init(VXGE_ERR,
3593				"%s: Encountered Serious Error",
3594				vdev->ndev->name);
3595		} else if (type == VXGE_HW_EVENT_CRITICAL_ERR)
3596			vxge_debug_init(VXGE_ERR,
3597				"%s: Encountered Critical Error",
3598				vdev->ndev->name);
3599	}
3600
3601	if ((type == VXGE_HW_EVENT_SERR) ||
3602		(type == VXGE_HW_EVENT_SLOT_FREEZE)) {
3603		if (unlikely(vdev->exec_mode))
3604			clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3605	} else if (type == VXGE_HW_EVENT_CRITICAL_ERR) {
3606		vxge_hw_device_mask_all(hldev);
3607		if (unlikely(vdev->exec_mode))
3608			clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3609	} else if ((type == VXGE_HW_EVENT_FIFO_ERR) ||
3610		  (type == VXGE_HW_EVENT_VPATH_ERR)) {
3611
3612		if (unlikely(vdev->exec_mode))
3613			clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3614		else {
3615			/* check if this vpath is already set for reset */
3616			if (!test_and_set_bit(vpath_idx, &vdev->vp_reset)) {
3617
3618				/* disable interrupts for this vpath */
3619				vxge_vpath_intr_disable(vdev, vpath_idx);
3620
3621				/* stop the queue for this vpath */
3622				netif_tx_stop_queue(vpath->fifo.txq);
3623			}
3624		}
3625	}
3626
3627	vxge_debug_entryexit(vdev->level_trace,
3628		"%s: %s:%d  Exiting...",
3629		vdev->ndev->name, __func__, __LINE__);
3630}
3631
3632static void verify_bandwidth(void)
3633{
3634	int i, band_width, total = 0, equal_priority = 0;
3635
3636	/* 1. If user enters 0 for some fifo, give equal priority to all */
3637	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3638		if (bw_percentage[i] == 0) {
3639			equal_priority = 1;
3640			break;
3641		}
3642	}
3643
3644	if (!equal_priority) {
3645		/* 2. If sum exceeds 100, give equal priority to all */
3646		for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3647			if (bw_percentage[i] == 0xFF)
3648				break;
3649
3650			total += bw_percentage[i];
3651			if (total > VXGE_HW_VPATH_BANDWIDTH_MAX) {
3652				equal_priority = 1;
3653				break;
3654			}
3655		}
3656	}
3657
3658	if (!equal_priority) {
3659		/* Is all the bandwidth consumed? */
3660		if (total < VXGE_HW_VPATH_BANDWIDTH_MAX) {
3661			if (i < VXGE_HW_MAX_VIRTUAL_PATHS) {
3662				/* Split rest of bw equally among next VPs*/
3663				band_width =
3664				  (VXGE_HW_VPATH_BANDWIDTH_MAX  - total) /
3665					(VXGE_HW_MAX_VIRTUAL_PATHS - i);
3666				if (band_width < 2) /* min of 2% */
3667					equal_priority = 1;
3668				else {
3669					for (; i < VXGE_HW_MAX_VIRTUAL_PATHS;
3670						i++)
3671						bw_percentage[i] =
3672							band_width;
3673				}
3674			}
3675		} else if (i < VXGE_HW_MAX_VIRTUAL_PATHS)
3676			equal_priority = 1;
3677	}
3678
3679	if (equal_priority) {
3680		vxge_debug_init(VXGE_ERR,
3681			"%s: Assigning equal bandwidth to all the vpaths",
3682			VXGE_DRIVER_NAME);
3683		bw_percentage[0] = VXGE_HW_VPATH_BANDWIDTH_MAX /
3684					VXGE_HW_MAX_VIRTUAL_PATHS;
3685		for (i = 1; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
3686			bw_percentage[i] = bw_percentage[0];
3687	}
3688}
3689
3690/*
3691 * Vpath configuration
3692 */
3693static int vxge_config_vpaths(struct vxge_hw_device_config *device_config,
3694			      u64 vpath_mask, struct vxge_config *config_param)
3695{
3696	int i, no_of_vpaths = 0, default_no_vpath = 0, temp;
3697	u32 txdl_size, txdl_per_memblock;
3698
3699	temp = driver_config->vpath_per_dev;
3700	if ((driver_config->vpath_per_dev == VXGE_USE_DEFAULT) &&
3701		(max_config_dev == VXGE_MAX_CONFIG_DEV)) {
3702		/* No more CPU. Return vpath number as zero.*/
3703		if (driver_config->g_no_cpus == -1)
3704			return 0;
3705
3706		if (!driver_config->g_no_cpus)
3707			driver_config->g_no_cpus =
3708				netif_get_num_default_rss_queues();
3709
3710		driver_config->vpath_per_dev = driver_config->g_no_cpus >> 1;
3711		if (!driver_config->vpath_per_dev)
3712			driver_config->vpath_per_dev = 1;
3713
3714		for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
3715			if (!vxge_bVALn(vpath_mask, i, 1))
3716				continue;
3717			else
3718				default_no_vpath++;
3719		if (default_no_vpath < driver_config->vpath_per_dev)
3720			driver_config->vpath_per_dev = default_no_vpath;
3721
3722		driver_config->g_no_cpus = driver_config->g_no_cpus -
3723				(driver_config->vpath_per_dev * 2);
3724		if (driver_config->g_no_cpus <= 0)
3725			driver_config->g_no_cpus = -1;
3726	}
3727
3728	if (driver_config->vpath_per_dev == 1) {
3729		vxge_debug_ll_config(VXGE_TRACE,
3730			"%s: Disable tx and rx steering, "
3731			"as single vpath is configured", VXGE_DRIVER_NAME);
3732		config_param->rth_steering = NO_STEERING;
3733		config_param->tx_steering_type = NO_STEERING;
3734		device_config->rth_en = 0;
3735	}
3736
3737	/* configure bandwidth */
3738	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
3739		device_config->vp_config[i].min_bandwidth = bw_percentage[i];
3740
3741	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3742		device_config->vp_config[i].vp_id = i;
3743		device_config->vp_config[i].mtu = VXGE_HW_DEFAULT_MTU;
3744		if (no_of_vpaths < driver_config->vpath_per_dev) {
3745			if (!vxge_bVALn(vpath_mask, i, 1)) {
3746				vxge_debug_ll_config(VXGE_TRACE,
3747					"%s: vpath: %d is not available",
3748					VXGE_DRIVER_NAME, i);
3749				continue;
3750			} else {
3751				vxge_debug_ll_config(VXGE_TRACE,
3752					"%s: vpath: %d available",
3753					VXGE_DRIVER_NAME, i);
3754				no_of_vpaths++;
3755			}
3756		} else {
3757			vxge_debug_ll_config(VXGE_TRACE,
3758				"%s: vpath: %d is not configured, "
3759				"max_config_vpath exceeded",
3760				VXGE_DRIVER_NAME, i);
3761			break;
3762		}
3763
3764		/* Configure Tx fifo's */
3765		device_config->vp_config[i].fifo.enable =
3766						VXGE_HW_FIFO_ENABLE;
3767		device_config->vp_config[i].fifo.max_frags =
3768				MAX_SKB_FRAGS + 1;
3769		device_config->vp_config[i].fifo.memblock_size =
3770			VXGE_HW_MIN_FIFO_MEMBLOCK_SIZE;
3771
3772		txdl_size = device_config->vp_config[i].fifo.max_frags *
3773				sizeof(struct vxge_hw_fifo_txd);
3774		txdl_per_memblock = VXGE_HW_MIN_FIFO_MEMBLOCK_SIZE / txdl_size;
3775
3776		device_config->vp_config[i].fifo.fifo_blocks =
3777			((VXGE_DEF_FIFO_LENGTH - 1) / txdl_per_memblock) + 1;
3778
3779		device_config->vp_config[i].fifo.intr =
3780				VXGE_HW_FIFO_QUEUE_INTR_DISABLE;
3781
3782		/* Configure tti properties */
3783		device_config->vp_config[i].tti.intr_enable =
3784					VXGE_HW_TIM_INTR_ENABLE;
3785
3786		device_config->vp_config[i].tti.btimer_val =
3787			(VXGE_TTI_BTIMER_VAL * 1000) / 272;
3788
3789		device_config->vp_config[i].tti.timer_ac_en =
3790				VXGE_HW_TIM_TIMER_AC_ENABLE;
3791
3792		/* For msi-x with napi (each vector has a handler of its own) -
3793		 * Set CI to OFF for all vpaths
3794		 */
3795		device_config->vp_config[i].tti.timer_ci_en =
3796			VXGE_HW_TIM_TIMER_CI_DISABLE;
3797
3798		device_config->vp_config[i].tti.timer_ri_en =
3799				VXGE_HW_TIM_TIMER_RI_DISABLE;
3800
3801		device_config->vp_config[i].tti.util_sel =
3802			VXGE_HW_TIM_UTIL_SEL_LEGACY_TX_NET_UTIL;
3803
3804		device_config->vp_config[i].tti.ltimer_val =
3805			(VXGE_TTI_LTIMER_VAL * 1000) / 272;
3806
3807		device_config->vp_config[i].tti.rtimer_val =
3808			(VXGE_TTI_RTIMER_VAL * 1000) / 272;
3809
3810		device_config->vp_config[i].tti.urange_a = TTI_TX_URANGE_A;
3811		device_config->vp_config[i].tti.urange_b = TTI_TX_URANGE_B;
3812		device_config->vp_config[i].tti.urange_c = TTI_TX_URANGE_C;
3813		device_config->vp_config[i].tti.uec_a = TTI_TX_UFC_A;
3814		device_config->vp_config[i].tti.uec_b = TTI_TX_UFC_B;
3815		device_config->vp_config[i].tti.uec_c = TTI_TX_UFC_C;
3816		device_config->vp_config[i].tti.uec_d = TTI_TX_UFC_D;
3817
3818		/* Configure Rx rings */
3819		device_config->vp_config[i].ring.enable  =
3820						VXGE_HW_RING_ENABLE;
3821
3822		device_config->vp_config[i].ring.ring_blocks  =
3823						VXGE_HW_DEF_RING_BLOCKS;
3824
3825		device_config->vp_config[i].ring.buffer_mode =
3826			VXGE_HW_RING_RXD_BUFFER_MODE_1;
3827
3828		device_config->vp_config[i].ring.rxds_limit  =
3829				VXGE_HW_DEF_RING_RXDS_LIMIT;
3830
3831		device_config->vp_config[i].ring.scatter_mode =
3832					VXGE_HW_RING_SCATTER_MODE_A;
3833
3834		/* Configure rti properties */
3835		device_config->vp_config[i].rti.intr_enable =
3836					VXGE_HW_TIM_INTR_ENABLE;
3837
3838		device_config->vp_config[i].rti.btimer_val =
3839			(VXGE_RTI_BTIMER_VAL * 1000)/272;
3840
3841		device_config->vp_config[i].rti.timer_ac_en =
3842						VXGE_HW_TIM_TIMER_AC_ENABLE;
3843
3844		device_config->vp_config[i].rti.timer_ci_en =
3845						VXGE_HW_TIM_TIMER_CI_DISABLE;
3846
3847		device_config->vp_config[i].rti.timer_ri_en =
3848						VXGE_HW_TIM_TIMER_RI_DISABLE;
3849
3850		device_config->vp_config[i].rti.util_sel =
3851				VXGE_HW_TIM_UTIL_SEL_LEGACY_RX_NET_UTIL;
3852
3853		device_config->vp_config[i].rti.urange_a =
3854						RTI_RX_URANGE_A;
3855		device_config->vp_config[i].rti.urange_b =
3856						RTI_RX_URANGE_B;
3857		device_config->vp_config[i].rti.urange_c =
3858						RTI_RX_URANGE_C;
3859		device_config->vp_config[i].rti.uec_a = RTI_RX_UFC_A;
3860		device_config->vp_config[i].rti.uec_b = RTI_RX_UFC_B;
3861		device_config->vp_config[i].rti.uec_c = RTI_RX_UFC_C;
3862		device_config->vp_config[i].rti.uec_d = RTI_RX_UFC_D;
3863
3864		device_config->vp_config[i].rti.rtimer_val =
3865			(VXGE_RTI_RTIMER_VAL * 1000) / 272;
3866
3867		device_config->vp_config[i].rti.ltimer_val =
3868			(VXGE_RTI_LTIMER_VAL * 1000) / 272;
3869
3870		device_config->vp_config[i].rpa_strip_vlan_tag =
3871			vlan_tag_strip;
3872	}
3873
3874	driver_config->vpath_per_dev = temp;
3875	return no_of_vpaths;
3876}
3877
3878/* initialize device configuratrions */
3879static void vxge_device_config_init(struct vxge_hw_device_config *device_config,
3880				    int *intr_type)
3881{
3882	/* Used for CQRQ/SRQ. */
3883	device_config->dma_blockpool_initial =
3884			VXGE_HW_INITIAL_DMA_BLOCK_POOL_SIZE;
3885
3886	device_config->dma_blockpool_max =
3887			VXGE_HW_MAX_DMA_BLOCK_POOL_SIZE;
3888
3889	if (max_mac_vpath > VXGE_MAX_MAC_ADDR_COUNT)
3890		max_mac_vpath = VXGE_MAX_MAC_ADDR_COUNT;
3891
3892#ifndef CONFIG_PCI_MSI
3893	vxge_debug_init(VXGE_ERR,
3894		"%s: This Kernel does not support "
3895		"MSI-X. Defaulting to INTA", VXGE_DRIVER_NAME);
3896	*intr_type = INTA;
3897#endif
3898
3899	/* Configure whether MSI-X or IRQL. */
3900	switch (*intr_type) {
3901	case INTA:
3902		device_config->intr_mode = VXGE_HW_INTR_MODE_IRQLINE;
3903		break;
3904
3905	case MSI_X:
3906		device_config->intr_mode = VXGE_HW_INTR_MODE_MSIX_ONE_SHOT;
3907		break;
3908	}
3909
3910	/* Timer period between device poll */
3911	device_config->device_poll_millis = VXGE_TIMER_DELAY;
3912
3913	/* Configure mac based steering. */
3914	device_config->rts_mac_en = addr_learn_en;
3915
3916	/* Configure Vpaths */
3917	device_config->rth_it_type = VXGE_HW_RTH_IT_TYPE_MULTI_IT;
3918
3919	vxge_debug_ll_config(VXGE_TRACE, "%s : Device Config Params ",
3920			__func__);
3921	vxge_debug_ll_config(VXGE_TRACE, "intr_mode : %d",
3922			device_config->intr_mode);
3923	vxge_debug_ll_config(VXGE_TRACE, "device_poll_millis : %d",
3924			device_config->device_poll_millis);
3925	vxge_debug_ll_config(VXGE_TRACE, "rth_en : %d",
3926			device_config->rth_en);
3927	vxge_debug_ll_config(VXGE_TRACE, "rth_it_type : %d",
3928			device_config->rth_it_type);
3929}
3930
3931static void vxge_print_parm(struct vxgedev *vdev, u64 vpath_mask)
3932{
3933	int i;
3934
3935	vxge_debug_init(VXGE_TRACE,
3936		"%s: %d Vpath(s) opened",
3937		vdev->ndev->name, vdev->no_of_vpath);
3938
3939	switch (vdev->config.intr_type) {
3940	case INTA:
3941		vxge_debug_init(VXGE_TRACE,
3942			"%s: Interrupt type INTA", vdev->ndev->name);
3943		break;
3944
3945	case MSI_X:
3946		vxge_debug_init(VXGE_TRACE,
3947			"%s: Interrupt type MSI-X", vdev->ndev->name);
3948		break;
3949	}
3950
3951	if (vdev->config.rth_steering) {
3952		vxge_debug_init(VXGE_TRACE,
3953			"%s: RTH steering enabled for TCP_IPV4",
3954			vdev->ndev->name);
3955	} else {
3956		vxge_debug_init(VXGE_TRACE,
3957			"%s: RTH steering disabled", vdev->ndev->name);
3958	}
3959
3960	switch (vdev->config.tx_steering_type) {
3961	case NO_STEERING:
3962		vxge_debug_init(VXGE_TRACE,
3963			"%s: Tx steering disabled", vdev->ndev->name);
3964		break;
3965	case TX_PRIORITY_STEERING:
3966		vxge_debug_init(VXGE_TRACE,
3967			"%s: Unsupported tx steering option",
3968			vdev->ndev->name);
3969		vxge_debug_init(VXGE_TRACE,
3970			"%s: Tx steering disabled", vdev->ndev->name);
3971		vdev->config.tx_steering_type = 0;
3972		break;
3973	case TX_VLAN_STEERING:
3974		vxge_debug_init(VXGE_TRACE,
3975			"%s: Unsupported tx steering option",
3976			vdev->ndev->name);
3977		vxge_debug_init(VXGE_TRACE,
3978			"%s: Tx steering disabled", vdev->ndev->name);
3979		vdev->config.tx_steering_type = 0;
3980		break;
3981	case TX_MULTIQ_STEERING:
3982		vxge_debug_init(VXGE_TRACE,
3983			"%s: Tx multiqueue steering enabled",
3984			vdev->ndev->name);
3985		break;
3986	case TX_PORT_STEERING:
3987		vxge_debug_init(VXGE_TRACE,
3988			"%s: Tx port steering enabled",
3989			vdev->ndev->name);
3990		break;
3991	default:
3992		vxge_debug_init(VXGE_ERR,
3993			"%s: Unsupported tx steering type",
3994			vdev->ndev->name);
3995		vxge_debug_init(VXGE_TRACE,
3996			"%s: Tx steering disabled", vdev->ndev->name);
3997		vdev->config.tx_steering_type = 0;
3998	}
3999
4000	if (vdev->config.addr_learn_en)
4001		vxge_debug_init(VXGE_TRACE,
4002			"%s: MAC Address learning enabled", vdev->ndev->name);
4003
4004	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
4005		if (!vxge_bVALn(vpath_mask, i, 1))
4006			continue;
4007		vxge_debug_ll_config(VXGE_TRACE,
4008			"%s: MTU size - %d", vdev->ndev->name,
4009			((vdev->devh))->
4010				config.vp_config[i].mtu);
4011		vxge_debug_init(VXGE_TRACE,
4012			"%s: VLAN tag stripping %s", vdev->ndev->name,
4013			((vdev->devh))->
4014				config.vp_config[i].rpa_strip_vlan_tag
4015			? "Enabled" : "Disabled");
4016		vxge_debug_ll_config(VXGE_TRACE,
4017			"%s: Max frags : %d", vdev->ndev->name,
4018			((vdev->devh))->
4019				config.vp_config[i].fifo.max_frags);
4020		break;
4021	}
4022}
4023
4024#ifdef CONFIG_PM
4025/**
4026 * vxge_pm_suspend - vxge power management suspend entry point
4027 *
4028 */
4029static int vxge_pm_suspend(struct pci_dev *pdev, pm_message_t state)
4030{
4031	return -ENOSYS;
4032}
4033/**
4034 * vxge_pm_resume - vxge power management resume entry point
4035 *
4036 */
4037static int vxge_pm_resume(struct pci_dev *pdev)
4038{
4039	return -ENOSYS;
4040}
4041
4042#endif
4043
4044/**
4045 * vxge_io_error_detected - called when PCI error is detected
4046 * @pdev: Pointer to PCI device
4047 * @state: The current pci connection state
4048 *
4049 * This function is called after a PCI bus error affecting
4050 * this device has been detected.
4051 */
4052static pci_ers_result_t vxge_io_error_detected(struct pci_dev *pdev,
4053						pci_channel_state_t state)
4054{
4055	struct __vxge_hw_device *hldev = pci_get_drvdata(pdev);
4056	struct net_device *netdev = hldev->ndev;
4057
4058	netif_device_detach(netdev);
4059
4060	if (state == pci_channel_io_perm_failure)
4061		return PCI_ERS_RESULT_DISCONNECT;
4062
4063	if (netif_running(netdev)) {
4064		/* Bring down the card, while avoiding PCI I/O */
4065		do_vxge_close(netdev, 0);
4066	}
4067
4068	pci_disable_device(pdev);
4069
4070	return PCI_ERS_RESULT_NEED_RESET;
4071}
4072
4073/**
4074 * vxge_io_slot_reset - called after the pci bus has been reset.
4075 * @pdev: Pointer to PCI device
4076 *
4077 * Restart the card from scratch, as if from a cold-boot.
4078 * At this point, the card has exprienced a hard reset,
4079 * followed by fixups by BIOS, and has its config space
4080 * set up identically to what it was at cold boot.
4081 */
4082static pci_ers_result_t vxge_io_slot_reset(struct pci_dev *pdev)
4083{
4084	struct __vxge_hw_device *hldev = pci_get_drvdata(pdev);
4085	struct net_device *netdev = hldev->ndev;
4086
4087	struct vxgedev *vdev = netdev_priv(netdev);
4088
4089	if (pci_enable_device(pdev)) {
4090		netdev_err(netdev, "Cannot re-enable device after reset\n");
4091		return PCI_ERS_RESULT_DISCONNECT;
4092	}
4093
4094	pci_set_master(pdev);
4095	do_vxge_reset(vdev, VXGE_LL_FULL_RESET);
4096
4097	return PCI_ERS_RESULT_RECOVERED;
4098}
4099
4100/**
4101 * vxge_io_resume - called when traffic can start flowing again.
4102 * @pdev: Pointer to PCI device
4103 *
4104 * This callback is called when the error recovery driver tells
4105 * us that its OK to resume normal operation.
4106 */
4107static void vxge_io_resume(struct pci_dev *pdev)
4108{
4109	struct __vxge_hw_device *hldev = pci_get_drvdata(pdev);
4110	struct net_device *netdev = hldev->ndev;
4111
4112	if (netif_running(netdev)) {
4113		if (vxge_open(netdev)) {
4114			netdev_err(netdev,
4115				   "Can't bring device back up after reset\n");
4116			return;
4117		}
4118	}
4119
4120	netif_device_attach(netdev);
4121}
4122
4123static inline u32 vxge_get_num_vfs(u64 function_mode)
4124{
4125	u32 num_functions = 0;
4126
4127	switch (function_mode) {
4128	case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION:
4129	case VXGE_HW_FUNCTION_MODE_SRIOV_8:
4130		num_functions = 8;
4131		break;
4132	case VXGE_HW_FUNCTION_MODE_SINGLE_FUNCTION:
4133		num_functions = 1;
4134		break;
4135	case VXGE_HW_FUNCTION_MODE_SRIOV:
4136	case VXGE_HW_FUNCTION_MODE_MRIOV:
4137	case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION_17:
4138		num_functions = 17;
4139		break;
4140	case VXGE_HW_FUNCTION_MODE_SRIOV_4:
4141		num_functions = 4;
4142		break;
4143	case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION_2:
4144		num_functions = 2;
4145		break;
4146	case VXGE_HW_FUNCTION_MODE_MRIOV_8:
4147		num_functions = 8; /* TODO */
4148		break;
4149	}
4150	return num_functions;
4151}
4152
4153int vxge_fw_upgrade(struct vxgedev *vdev, char *fw_name, int override)
4154{
4155	struct __vxge_hw_device *hldev = vdev->devh;
4156	u32 maj, min, bld, cmaj, cmin, cbld;
4157	enum vxge_hw_status status;
4158	const struct firmware *fw;
4159	int ret;
4160
4161	ret = request_firmware(&fw, fw_name, &vdev->pdev->dev);
4162	if (ret) {
4163		vxge_debug_init(VXGE_ERR, "%s: Firmware file '%s' not found",
4164				VXGE_DRIVER_NAME, fw_name);
4165		goto out;
4166	}
4167
4168	/* Load the new firmware onto the adapter */
4169	status = vxge_update_fw_image(hldev, fw->data, fw->size);
4170	if (status != VXGE_HW_OK) {
4171		vxge_debug_init(VXGE_ERR,
4172				"%s: FW image download to adapter failed '%s'.",
4173				VXGE_DRIVER_NAME, fw_name);
4174		ret = -EIO;
4175		goto out;
4176	}
4177
4178	/* Read the version of the new firmware */
4179	status = vxge_hw_upgrade_read_version(hldev, &maj, &min, &bld);
4180	if (status != VXGE_HW_OK) {
4181		vxge_debug_init(VXGE_ERR,
4182				"%s: Upgrade read version failed '%s'.",
4183				VXGE_DRIVER_NAME, fw_name);
4184		ret = -EIO;
4185		goto out;
4186	}
4187
4188	cmaj = vdev->config.device_hw_info.fw_version.major;
4189	cmin = vdev->config.device_hw_info.fw_version.minor;
4190	cbld = vdev->config.device_hw_info.fw_version.build;
4191	/* It's possible the version in /lib/firmware is not the latest version.
4192	 * If so, we could get into a loop of trying to upgrade to the latest
4193	 * and flashing the older version.
4194	 */
4195	if (VXGE_FW_VER(maj, min, bld) == VXGE_FW_VER(cmaj, cmin, cbld) &&
4196	    !override) {
4197		ret = -EINVAL;
4198		goto out;
4199	}
4200
4201	printk(KERN_NOTICE "Upgrade to firmware version %d.%d.%d commencing\n",
4202	       maj, min, bld);
4203
4204	/* Flash the adapter with the new firmware */
4205	status = vxge_hw_flash_fw(hldev);
4206	if (status != VXGE_HW_OK) {
4207		vxge_debug_init(VXGE_ERR, "%s: Upgrade commit failed '%s'.",
4208				VXGE_DRIVER_NAME, fw_name);
4209		ret = -EIO;
4210		goto out;
4211	}
4212
4213	printk(KERN_NOTICE "Upgrade of firmware successful!  Adapter must be "
4214	       "hard reset before using, thus requiring a system reboot or a "
4215	       "hotplug event.\n");
4216
4217out:
4218	release_firmware(fw);
4219	return ret;
4220}
4221
4222static int vxge_probe_fw_update(struct vxgedev *vdev)
4223{
4224	u32 maj, min, bld;
4225	int ret, gpxe = 0;
4226	char *fw_name;
4227
4228	maj = vdev->config.device_hw_info.fw_version.major;
4229	min = vdev->config.device_hw_info.fw_version.minor;
4230	bld = vdev->config.device_hw_info.fw_version.build;
4231
4232	if (VXGE_FW_VER(maj, min, bld) == VXGE_CERT_FW_VER)
4233		return 0;
4234
4235	/* Ignore the build number when determining if the current firmware is
4236	 * "too new" to load the driver
4237	 */
4238	if (VXGE_FW_VER(maj, min, 0) > VXGE_CERT_FW_VER) {
4239		vxge_debug_init(VXGE_ERR, "%s: Firmware newer than last known "
4240				"version, unable to load driver\n",
4241				VXGE_DRIVER_NAME);
4242		return -EINVAL;
4243	}
4244
4245	/* Firmware 1.4.4 and older cannot be upgraded, and is too ancient to
4246	 * work with this driver.
4247	 */
4248	if (VXGE_FW_VER(maj, min, bld) <= VXGE_FW_DEAD_VER) {
4249		vxge_debug_init(VXGE_ERR, "%s: Firmware %d.%d.%d cannot be "
4250				"upgraded\n", VXGE_DRIVER_NAME, maj, min, bld);
4251		return -EINVAL;
4252	}
4253
4254	/* If file not specified, determine gPXE or not */
4255	if (VXGE_FW_VER(maj, min, bld) >= VXGE_EPROM_FW_VER) {
4256		int i;
4257		for (i = 0; i < VXGE_HW_MAX_ROM_IMAGES; i++)
4258			if (vdev->devh->eprom_versions[i]) {
4259				gpxe = 1;
4260				break;
4261			}
4262	}
4263	if (gpxe)
4264		fw_name = "vxge/X3fw-pxe.ncf";
4265	else
4266		fw_name = "vxge/X3fw.ncf";
4267
4268	ret = vxge_fw_upgrade(vdev, fw_name, 0);
4269	/* -EINVAL and -ENOENT are not fatal errors for flashing firmware on
4270	 * probe, so ignore them
4271	 */
4272	if (ret != -EINVAL && ret != -ENOENT)
4273		return -EIO;
4274	else
4275		ret = 0;
4276
4277	if (VXGE_FW_VER(VXGE_CERT_FW_VER_MAJOR, VXGE_CERT_FW_VER_MINOR, 0) >
4278	    VXGE_FW_VER(maj, min, 0)) {
4279		vxge_debug_init(VXGE_ERR, "%s: Firmware %d.%d.%d is too old to"
4280				" be used with this driver.",
4281				VXGE_DRIVER_NAME, maj, min, bld);
4282		return -EINVAL;
4283	}
4284
4285	return ret;
4286}
4287
4288static int is_sriov_initialized(struct pci_dev *pdev)
4289{
4290	int pos;
4291	u16 ctrl;
4292
4293	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
4294	if (pos) {
4295		pci_read_config_word(pdev, pos + PCI_SRIOV_CTRL, &ctrl);
4296		if (ctrl & PCI_SRIOV_CTRL_VFE)
4297			return 1;
4298	}
4299	return 0;
4300}
4301
4302static const struct vxge_hw_uld_cbs vxge_callbacks = {
4303	.link_up = vxge_callback_link_up,
4304	.link_down = vxge_callback_link_down,
4305	.crit_err = vxge_callback_crit_err,
4306};
4307
4308/**
4309 * vxge_probe
4310 * @pdev : structure containing the PCI related information of the device.
4311 * @pre: List of PCI devices supported by the driver listed in vxge_id_table.
4312 * Description:
4313 * This function is called when a new PCI device gets detected and initializes
4314 * it.
4315 * Return value:
4316 * returns 0 on success and negative on failure.
4317 *
4318 */
4319static int
4320vxge_probe(struct pci_dev *pdev, const struct pci_device_id *pre)
4321{
4322	struct __vxge_hw_device *hldev;
4323	enum vxge_hw_status status;
4324	int ret;
4325	int high_dma = 0;
4326	u64 vpath_mask = 0;
4327	struct vxgedev *vdev;
4328	struct vxge_config *ll_config = NULL;
4329	struct vxge_hw_device_config *device_config = NULL;
4330	struct vxge_hw_device_attr attr;
4331	int i, j, no_of_vpath = 0, max_vpath_supported = 0;
4332	u8 *macaddr;
4333	struct vxge_mac_addrs *entry;
4334	static int bus = -1, device = -1;
4335	u32 host_type;
4336	u8 new_device = 0;
4337	enum vxge_hw_status is_privileged;
4338	u32 function_mode;
4339	u32 num_vfs = 0;
4340
4341	vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
4342	attr.pdev = pdev;
4343
4344	/* In SRIOV-17 mode, functions of the same adapter
4345	 * can be deployed on different buses
4346	 */
4347	if (((bus != pdev->bus->number) || (device != PCI_SLOT(pdev->devfn))) &&
4348	    !pdev->is_virtfn)
4349		new_device = 1;
4350
4351	bus = pdev->bus->number;
4352	device = PCI_SLOT(pdev->devfn);
4353
4354	if (new_device) {
4355		if (driver_config->config_dev_cnt &&
4356		   (driver_config->config_dev_cnt !=
4357			driver_config->total_dev_cnt))
4358			vxge_debug_init(VXGE_ERR,
4359				"%s: Configured %d of %d devices",
4360				VXGE_DRIVER_NAME,
4361				driver_config->config_dev_cnt,
4362				driver_config->total_dev_cnt);
4363		driver_config->config_dev_cnt = 0;
4364		driver_config->total_dev_cnt = 0;
4365	}
4366
4367	/* Now making the CPU based no of vpath calculation
4368	 * applicable for individual functions as well.
4369	 */
4370	driver_config->g_no_cpus = 0;
4371	driver_config->vpath_per_dev = max_config_vpath;
4372
4373	driver_config->total_dev_cnt++;
4374	if (++driver_config->config_dev_cnt > max_config_dev) {
4375		ret = 0;
4376		goto _exit0;
4377	}
4378
4379	device_config = kzalloc(sizeof(struct vxge_hw_device_config),
4380		GFP_KERNEL);
4381	if (!device_config) {
4382		ret = -ENOMEM;
4383		vxge_debug_init(VXGE_ERR,
4384			"device_config : malloc failed %s %d",
4385			__FILE__, __LINE__);
4386		goto _exit0;
4387	}
4388
4389	ll_config = kzalloc(sizeof(struct vxge_config), GFP_KERNEL);
4390	if (!ll_config) {
4391		ret = -ENOMEM;
4392		vxge_debug_init(VXGE_ERR,
4393			"device_config : malloc failed %s %d",
4394			__FILE__, __LINE__);
4395		goto _exit0;
4396	}
4397	ll_config->tx_steering_type = TX_MULTIQ_STEERING;
4398	ll_config->intr_type = MSI_X;
4399	ll_config->napi_weight = NEW_NAPI_WEIGHT;
4400	ll_config->rth_steering = RTH_STEERING;
4401
4402	/* get the default configuration parameters */
4403	vxge_hw_device_config_default_get(device_config);
4404
4405	/* initialize configuration parameters */
4406	vxge_device_config_init(device_config, &ll_config->intr_type);
4407
4408	ret = pci_enable_device(pdev);
4409	if (ret) {
4410		vxge_debug_init(VXGE_ERR,
4411			"%s : can not enable PCI device", __func__);
4412		goto _exit0;
4413	}
4414
4415	if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
4416		vxge_debug_ll_config(VXGE_TRACE,
4417			"%s : using 64bit DMA", __func__);
4418
4419		high_dma = 1;
4420
4421		if (pci_set_consistent_dma_mask(pdev,
4422						DMA_BIT_MASK(64))) {
4423			vxge_debug_init(VXGE_ERR,
4424				"%s : unable to obtain 64bit DMA for "
4425				"consistent allocations", __func__);
4426			ret = -ENOMEM;
4427			goto _exit1;
4428		}
4429	} else if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(32))) {
4430		vxge_debug_ll_config(VXGE_TRACE,
4431			"%s : using 32bit DMA", __func__);
4432	} else {
4433		ret = -ENOMEM;
4434		goto _exit1;
4435	}
4436
4437	ret = pci_request_region(pdev, 0, VXGE_DRIVER_NAME);
4438	if (ret) {
4439		vxge_debug_init(VXGE_ERR,
4440			"%s : request regions failed", __func__);
4441		goto _exit1;
4442	}
4443
4444	pci_set_master(pdev);
4445
4446	attr.bar0 = pci_ioremap_bar(pdev, 0);
4447	if (!attr.bar0) {
4448		vxge_debug_init(VXGE_ERR,
4449			"%s : cannot remap io memory bar0", __func__);
4450		ret = -ENODEV;
4451		goto _exit2;
4452	}
4453	vxge_debug_ll_config(VXGE_TRACE,
4454		"pci ioremap bar0: %p:0x%llx",
4455		attr.bar0,
4456		(unsigned long long)pci_resource_start(pdev, 0));
4457
4458	status = vxge_hw_device_hw_info_get(attr.bar0,
4459			&ll_config->device_hw_info);
4460	if (status != VXGE_HW_OK) {
4461		vxge_debug_init(VXGE_ERR,
4462			"%s: Reading of hardware info failed."
4463			"Please try upgrading the firmware.", VXGE_DRIVER_NAME);
4464		ret = -EINVAL;
4465		goto _exit3;
4466	}
4467
4468	vpath_mask = ll_config->device_hw_info.vpath_mask;
4469	if (vpath_mask == 0) {
4470		vxge_debug_ll_config(VXGE_TRACE,
4471			"%s: No vpaths available in device", VXGE_DRIVER_NAME);
4472		ret = -EINVAL;
4473		goto _exit3;
4474	}
4475
4476	vxge_debug_ll_config(VXGE_TRACE,
4477		"%s:%d  Vpath mask = %llx", __func__, __LINE__,
4478		(unsigned long long)vpath_mask);
4479
4480	function_mode = ll_config->device_hw_info.function_mode;
4481	host_type = ll_config->device_hw_info.host_type;
4482	is_privileged = __vxge_hw_device_is_privilaged(host_type,
4483		ll_config->device_hw_info.func_id);
4484
4485	/* Check how many vpaths are available */
4486	for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
4487		if (!((vpath_mask) & vxge_mBIT(i)))
4488			continue;
4489		max_vpath_supported++;
4490	}
4491
4492	if (new_device)
4493		num_vfs = vxge_get_num_vfs(function_mode) - 1;
4494
4495	/* Enable SRIOV mode, if firmware has SRIOV support and if it is a PF */
4496	if (is_sriov(function_mode) && !is_sriov_initialized(pdev) &&
4497	   (ll_config->intr_type != INTA)) {
4498		ret = pci_enable_sriov(pdev, num_vfs);
4499		if (ret)
4500			vxge_debug_ll_config(VXGE_ERR,
4501				"Failed in enabling SRIOV mode: %d\n", ret);
4502			/* No need to fail out, as an error here is non-fatal */
4503	}
4504
4505	/*
4506	 * Configure vpaths and get driver configured number of vpaths
4507	 * which is less than or equal to the maximum vpaths per function.
4508	 */
4509	no_of_vpath = vxge_config_vpaths(device_config, vpath_mask, ll_config);
4510	if (!no_of_vpath) {
4511		vxge_debug_ll_config(VXGE_ERR,
4512			"%s: No more vpaths to configure", VXGE_DRIVER_NAME);
4513		ret = 0;
4514		goto _exit3;
4515	}
4516
4517	/* Setting driver callbacks */
4518	attr.uld_callbacks = &vxge_callbacks;
4519
4520	status = vxge_hw_device_initialize(&hldev, &attr, device_config);
4521	if (status != VXGE_HW_OK) {
4522		vxge_debug_init(VXGE_ERR,
4523			"Failed to initialize device (%d)", status);
4524			ret = -EINVAL;
4525			goto _exit3;
4526	}
4527
4528	if (VXGE_FW_VER(ll_config->device_hw_info.fw_version.major,
4529			ll_config->device_hw_info.fw_version.minor,
4530			ll_config->device_hw_info.fw_version.build) >=
4531	    VXGE_EPROM_FW_VER) {
4532		struct eprom_image img[VXGE_HW_MAX_ROM_IMAGES];
4533
4534		status = vxge_hw_vpath_eprom_img_ver_get(hldev, img);
4535		if (status != VXGE_HW_OK) {
4536			vxge_debug_init(VXGE_ERR, "%s: Reading of EPROM failed",
4537					VXGE_DRIVER_NAME);
4538			/* This is a non-fatal error, continue */
4539		}
4540
4541		for (i = 0; i < VXGE_HW_MAX_ROM_IMAGES; i++) {
4542			hldev->eprom_versions[i] = img[i].version;
4543			if (!img[i].is_valid)
4544				break;
4545			vxge_debug_init(VXGE_TRACE, "%s: EPROM %d, version "
4546					"%d.%d.%d.%d", VXGE_DRIVER_NAME, i,
4547					VXGE_EPROM_IMG_MAJOR(img[i].version),
4548					VXGE_EPROM_IMG_MINOR(img[i].version),
4549					VXGE_EPROM_IMG_FIX(img[i].version),
4550					VXGE_EPROM_IMG_BUILD(img[i].version));
4551		}
4552	}
4553
4554	/* if FCS stripping is not disabled in MAC fail driver load */
4555	status = vxge_hw_vpath_strip_fcs_check(hldev, vpath_mask);
4556	if (status != VXGE_HW_OK) {
4557		vxge_debug_init(VXGE_ERR, "%s: FCS stripping is enabled in MAC"
4558				" failing driver load", VXGE_DRIVER_NAME);
4559		ret = -EINVAL;
4560		goto _exit4;
4561	}
4562
4563	/* Always enable HWTS.  This will always cause the FCS to be invalid,
4564	 * due to the fact that HWTS is using the FCS as the location of the
4565	 * timestamp.  The HW FCS checking will still correctly determine if
4566	 * there is a valid checksum, and the FCS is being removed by the driver
4567	 * anyway.  So no fucntionality is being lost.  Since it is always
4568	 * enabled, we now simply use the ioctl call to set whether or not the
4569	 * driver should be paying attention to the HWTS.
4570	 */
4571	if (is_privileged == VXGE_HW_OK) {
4572		status = vxge_timestamp_config(hldev);
4573		if (status != VXGE_HW_OK) {
4574			vxge_debug_init(VXGE_ERR, "%s: HWTS enable failed",
4575					VXGE_DRIVER_NAME);
4576			ret = -EFAULT;
4577			goto _exit4;
4578		}
4579	}
4580
4581	vxge_hw_device_debug_set(hldev, VXGE_ERR, VXGE_COMPONENT_LL);
4582
4583	/* set private device info */
4584	pci_set_drvdata(pdev, hldev);
4585
4586	ll_config->fifo_indicate_max_pkts = VXGE_FIFO_INDICATE_MAX_PKTS;
4587	ll_config->addr_learn_en = addr_learn_en;
4588	ll_config->rth_algorithm = RTH_ALG_JENKINS;
4589	ll_config->rth_hash_type_tcpipv4 = 1;
4590	ll_config->rth_hash_type_ipv4 = 0;
4591	ll_config->rth_hash_type_tcpipv6 = 0;
4592	ll_config->rth_hash_type_ipv6 = 0;
4593	ll_config->rth_hash_type_tcpipv6ex = 0;
4594	ll_config->rth_hash_type_ipv6ex = 0;
4595	ll_config->rth_bkt_sz = RTH_BUCKET_SIZE;
4596	ll_config->tx_pause_enable = VXGE_PAUSE_CTRL_ENABLE;
4597	ll_config->rx_pause_enable = VXGE_PAUSE_CTRL_ENABLE;
4598
4599	ret = vxge_device_register(hldev, ll_config, high_dma, no_of_vpath,
4600				   &vdev);
4601	if (ret) {
4602		ret = -EINVAL;
4603		goto _exit4;
4604	}
4605
4606	ret = vxge_probe_fw_update(vdev);
4607	if (ret)
4608		goto _exit5;
4609
4610	vxge_hw_device_debug_set(hldev, VXGE_TRACE, VXGE_COMPONENT_LL);
4611	VXGE_COPY_DEBUG_INFO_TO_LL(vdev, vxge_hw_device_error_level_get(hldev),
4612		vxge_hw_device_trace_level_get(hldev));
4613
4614	/* set private HW device info */
4615	vdev->mtu = VXGE_HW_DEFAULT_MTU;
4616	vdev->bar0 = attr.bar0;
4617	vdev->max_vpath_supported = max_vpath_supported;
4618	vdev->no_of_vpath = no_of_vpath;
4619
4620	/* Virtual Path count */
4621	for (i = 0, j = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
4622		if (!vxge_bVALn(vpath_mask, i, 1))
4623			continue;
4624		if (j >= vdev->no_of_vpath)
4625			break;
4626
4627		vdev->vpaths[j].is_configured = 1;
4628		vdev->vpaths[j].device_id = i;
4629		vdev->vpaths[j].ring.driver_id = j;
4630		vdev->vpaths[j].vdev = vdev;
4631		vdev->vpaths[j].max_mac_addr_cnt = max_mac_vpath;
4632		memcpy((u8 *)vdev->vpaths[j].macaddr,
4633				ll_config->device_hw_info.mac_addrs[i],
4634				ETH_ALEN);
4635
4636		/* Initialize the mac address list header */
4637		INIT_LIST_HEAD(&vdev->vpaths[j].mac_addr_list);
4638
4639		vdev->vpaths[j].mac_addr_cnt = 0;
4640		vdev->vpaths[j].mcast_addr_cnt = 0;
4641		j++;
4642	}
4643	vdev->exec_mode = VXGE_EXEC_MODE_DISABLE;
4644	vdev->max_config_port = max_config_port;
4645
4646	vdev->vlan_tag_strip = vlan_tag_strip;
4647
4648	/* map the hashing selector table to the configured vpaths */
4649	for (i = 0; i < vdev->no_of_vpath; i++)
4650		vdev->vpath_selector[i] = vpath_selector[i];
4651
4652	macaddr = (u8 *)vdev->vpaths[0].macaddr;
4653
4654	ll_config->device_hw_info.serial_number[VXGE_HW_INFO_LEN - 1] = '\0';
4655	ll_config->device_hw_info.product_desc[VXGE_HW_INFO_LEN - 1] = '\0';
4656	ll_config->device_hw_info.part_number[VXGE_HW_INFO_LEN - 1] = '\0';
4657
4658	vxge_debug_init(VXGE_TRACE, "%s: SERIAL NUMBER: %s",
4659		vdev->ndev->name, ll_config->device_hw_info.serial_number);
4660
4661	vxge_debug_init(VXGE_TRACE, "%s: PART NUMBER: %s",
4662		vdev->ndev->name, ll_config->device_hw_info.part_number);
4663
4664	vxge_debug_init(VXGE_TRACE, "%s: Neterion %s Server Adapter",
4665		vdev->ndev->name, ll_config->device_hw_info.product_desc);
4666
4667	vxge_debug_init(VXGE_TRACE, "%s: MAC ADDR: %pM",
4668		vdev->ndev->name, macaddr);
4669
4670	vxge_debug_init(VXGE_TRACE, "%s: Link Width x%d",
4671		vdev->ndev->name, vxge_hw_device_link_width_get(hldev));
4672
4673	vxge_debug_init(VXGE_TRACE,
4674		"%s: Firmware version : %s Date : %s", vdev->ndev->name,
4675		ll_config->device_hw_info.fw_version.version,
4676		ll_config->device_hw_info.fw_date.date);
4677
4678	if (new_device) {
4679		switch (ll_config->device_hw_info.function_mode) {
4680		case VXGE_HW_FUNCTION_MODE_SINGLE_FUNCTION:
4681			vxge_debug_init(VXGE_TRACE,
4682			"%s: Single Function Mode Enabled", vdev->ndev->name);
4683		break;
4684		case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION:
4685			vxge_debug_init(VXGE_TRACE,
4686			"%s: Multi Function Mode Enabled", vdev->ndev->name);
4687		break;
4688		case VXGE_HW_FUNCTION_MODE_SRIOV:
4689			vxge_debug_init(VXGE_TRACE,
4690			"%s: Single Root IOV Mode Enabled", vdev->ndev->name);
4691		break;
4692		case VXGE_HW_FUNCTION_MODE_MRIOV:
4693			vxge_debug_init(VXGE_TRACE,
4694			"%s: Multi Root IOV Mode Enabled", vdev->ndev->name);
4695		break;
4696		}
4697	}
4698
4699	vxge_print_parm(vdev, vpath_mask);
4700
4701	/* Store the fw version for ethttool option */
4702	strcpy(vdev->fw_version, ll_config->device_hw_info.fw_version.version);
4703	memcpy(vdev->ndev->dev_addr, (u8 *)vdev->vpaths[0].macaddr, ETH_ALEN);
4704
4705	/* Copy the station mac address to the list */
4706	for (i = 0; i < vdev->no_of_vpath; i++) {
4707		entry =	kzalloc(sizeof(struct vxge_mac_addrs), GFP_KERNEL);
4708		if (NULL == entry) {
4709			vxge_debug_init(VXGE_ERR,
4710				"%s: mac_addr_list : memory allocation failed",
4711				vdev->ndev->name);
4712			ret = -EPERM;
4713			goto _exit6;
4714		}
4715		macaddr = (u8 *)&entry->macaddr;
4716		memcpy(macaddr, vdev->ndev->dev_addr, ETH_ALEN);
4717		list_add(&entry->item, &vdev->vpaths[i].mac_addr_list);
4718		vdev->vpaths[i].mac_addr_cnt = 1;
4719	}
4720
4721	kfree(device_config);
4722
4723	/*
4724	 * INTA is shared in multi-function mode. This is unlike the INTA
4725	 * implementation in MR mode, where each VH has its own INTA message.
4726	 * - INTA is masked (disabled) as long as at least one function sets
4727	 * its TITAN_MASK_ALL_INT.ALARM bit.
4728	 * - INTA is unmasked (enabled) when all enabled functions have cleared
4729	 * their own TITAN_MASK_ALL_INT.ALARM bit.
4730	 * The TITAN_MASK_ALL_INT ALARM & TRAFFIC bits are cleared on power up.
4731	 * Though this driver leaves the top level interrupts unmasked while
4732	 * leaving the required module interrupt bits masked on exit, there
4733	 * could be a rougue driver around that does not follow this procedure
4734	 * resulting in a failure to generate interrupts. The following code is
4735	 * present to prevent such a failure.
4736	 */
4737
4738	if (ll_config->device_hw_info.function_mode ==
4739		VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION)
4740		if (vdev->config.intr_type == INTA)
4741			vxge_hw_device_unmask_all(hldev);
4742
4743	vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d  Exiting...",
4744		vdev->ndev->name, __func__, __LINE__);
4745
4746	vxge_hw_device_debug_set(hldev, VXGE_ERR, VXGE_COMPONENT_LL);
4747	VXGE_COPY_DEBUG_INFO_TO_LL(vdev, vxge_hw_device_error_level_get(hldev),
4748		vxge_hw_device_trace_level_get(hldev));
4749
4750	kfree(ll_config);
4751	return 0;
4752
4753_exit6:
4754	for (i = 0; i < vdev->no_of_vpath; i++)
4755		vxge_free_mac_add_list(&vdev->vpaths[i]);
4756_exit5:
4757	vxge_device_unregister(hldev);
4758_exit4:
4759	vxge_hw_device_terminate(hldev);
4760	pci_disable_sriov(pdev);
4761_exit3:
4762	iounmap(attr.bar0);
4763_exit2:
4764	pci_release_region(pdev, 0);
4765_exit1:
4766	pci_disable_device(pdev);
4767_exit0:
4768	kfree(ll_config);
4769	kfree(device_config);
4770	driver_config->config_dev_cnt--;
4771	driver_config->total_dev_cnt--;
4772	return ret;
4773}
4774
4775/**
4776 * vxge_rem_nic - Free the PCI device
4777 * @pdev: structure containing the PCI related information of the device.
4778 * Description: This function is called by the Pci subsystem to release a
4779 * PCI device and free up all resource held up by the device.
4780 */
4781static void vxge_remove(struct pci_dev *pdev)
4782{
4783	struct __vxge_hw_device *hldev;
4784	struct vxgedev *vdev;
4785	int i;
4786
4787	hldev = pci_get_drvdata(pdev);
4788	if (hldev == NULL)
4789		return;
4790
4791	vdev = netdev_priv(hldev->ndev);
4792
4793	vxge_debug_entryexit(vdev->level_trace,	"%s:%d", __func__, __LINE__);
4794	vxge_debug_init(vdev->level_trace, "%s : removing PCI device...",
4795			__func__);
4796
4797	for (i = 0; i < vdev->no_of_vpath; i++)
4798		vxge_free_mac_add_list(&vdev->vpaths[i]);
4799
4800	vxge_device_unregister(hldev);
4801	/* Do not call pci_disable_sriov here, as it will break child devices */
4802	vxge_hw_device_terminate(hldev);
4803	iounmap(vdev->bar0);
4804	pci_release_region(pdev, 0);
4805	pci_disable_device(pdev);
4806	driver_config->config_dev_cnt--;
4807	driver_config->total_dev_cnt--;
4808
4809	vxge_debug_init(vdev->level_trace, "%s:%d Device unregistered",
4810			__func__, __LINE__);
4811	vxge_debug_entryexit(vdev->level_trace,	"%s:%d  Exiting...", __func__,
4812			     __LINE__);
4813}
4814
4815static const struct pci_error_handlers vxge_err_handler = {
4816	.error_detected = vxge_io_error_detected,
4817	.slot_reset = vxge_io_slot_reset,
4818	.resume = vxge_io_resume,
4819};
4820
4821static struct pci_driver vxge_driver = {
4822	.name = VXGE_DRIVER_NAME,
4823	.id_table = vxge_id_table,
4824	.probe = vxge_probe,
4825	.remove = vxge_remove,
4826#ifdef CONFIG_PM
4827	.suspend = vxge_pm_suspend,
4828	.resume = vxge_pm_resume,
4829#endif
4830	.err_handler = &vxge_err_handler,
4831};
4832
4833static int __init
4834vxge_starter(void)
4835{
4836	int ret = 0;
4837
4838	pr_info("Copyright(c) 2002-2010 Exar Corp.\n");
4839	pr_info("Driver version: %s\n", DRV_VERSION);
4840
4841	verify_bandwidth();
4842
4843	driver_config = kzalloc(sizeof(struct vxge_drv_config), GFP_KERNEL);
4844	if (!driver_config)
4845		return -ENOMEM;
4846
4847	ret = pci_register_driver(&vxge_driver);
4848	if (ret) {
4849		kfree(driver_config);
4850		goto err;
4851	}
4852
4853	if (driver_config->config_dev_cnt &&
4854	   (driver_config->config_dev_cnt != driver_config->total_dev_cnt))
4855		vxge_debug_init(VXGE_ERR,
4856			"%s: Configured %d of %d devices",
4857			VXGE_DRIVER_NAME, driver_config->config_dev_cnt,
4858			driver_config->total_dev_cnt);
4859err:
4860	return ret;
4861}
4862
4863static void __exit
4864vxge_closer(void)
4865{
4866	pci_unregister_driver(&vxge_driver);
4867	kfree(driver_config);
4868}
4869module_init(vxge_starter);
4870module_exit(vxge_closer);
4871