ed1786598c9eff166766c246f5514a70c092396f
[linux-2.6.git] / drivers / net / vxge / vxge-main.c
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 Neterion Inc's X3100 Series 10GbE PCIe I/O
11 *              Virtualized Server Adapter.
12 * Copyright(c) 2002-2009 Neterion Inc.
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 #include <linux/if_vlan.h>
45 #include <linux/pci.h>
46 #include <linux/slab.h>
47 #include <linux/tcp.h>
48 #include <net/ip.h>
49 #include <linux/netdevice.h>
50 #include <linux/etherdevice.h>
51 #include "vxge-main.h"
52 #include "vxge-reg.h"
53
54 MODULE_LICENSE("Dual BSD/GPL");
55 MODULE_DESCRIPTION("Neterion's X3100 Series 10GbE PCIe I/O"
56         "Virtualized Server Adapter");
57
58 static DEFINE_PCI_DEVICE_TABLE(vxge_id_table) = {
59         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_TITAN_WIN, PCI_ANY_ID,
60         PCI_ANY_ID},
61         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_TITAN_UNI, PCI_ANY_ID,
62         PCI_ANY_ID},
63         {0}
64 };
65
66 MODULE_DEVICE_TABLE(pci, vxge_id_table);
67
68 VXGE_MODULE_PARAM_INT(vlan_tag_strip, VXGE_HW_VPATH_RPA_STRIP_VLAN_TAG_ENABLE);
69 VXGE_MODULE_PARAM_INT(addr_learn_en, VXGE_HW_MAC_ADDR_LEARN_DEFAULT);
70 VXGE_MODULE_PARAM_INT(max_config_port, VXGE_MAX_CONFIG_PORT);
71 VXGE_MODULE_PARAM_INT(max_config_vpath, VXGE_USE_DEFAULT);
72 VXGE_MODULE_PARAM_INT(max_mac_vpath, VXGE_MAX_MAC_ADDR_COUNT);
73 VXGE_MODULE_PARAM_INT(max_config_dev, VXGE_MAX_CONFIG_DEV);
74
75 static u16 vpath_selector[VXGE_HW_MAX_VIRTUAL_PATHS] =
76                 {0, 1, 3, 3, 7, 7, 7, 7, 15, 15, 15, 15, 15, 15, 15, 15, 31};
77 static unsigned int bw_percentage[VXGE_HW_MAX_VIRTUAL_PATHS] =
78         {[0 ...(VXGE_HW_MAX_VIRTUAL_PATHS - 1)] = 0xFF};
79 module_param_array(bw_percentage, uint, NULL, 0);
80
81 static struct vxge_drv_config *driver_config;
82
83 static inline int is_vxge_card_up(struct vxgedev *vdev)
84 {
85         return test_bit(__VXGE_STATE_CARD_UP, &vdev->state);
86 }
87
88 static inline void VXGE_COMPLETE_VPATH_TX(struct vxge_fifo *fifo)
89 {
90         unsigned long flags = 0;
91         struct sk_buff **skb_ptr = NULL;
92         struct sk_buff **temp;
93 #define NR_SKB_COMPLETED 128
94         struct sk_buff *completed[NR_SKB_COMPLETED];
95         int more;
96
97         do {
98                 more = 0;
99                 skb_ptr = completed;
100
101                 if (spin_trylock_irqsave(&fifo->tx_lock, flags)) {
102                         vxge_hw_vpath_poll_tx(fifo->handle, &skb_ptr,
103                                                 NR_SKB_COMPLETED, &more);
104                         spin_unlock_irqrestore(&fifo->tx_lock, flags);
105                 }
106                 /* free SKBs */
107                 for (temp = completed; temp != skb_ptr; temp++)
108                         dev_kfree_skb_irq(*temp);
109         } while (more) ;
110 }
111
112 static inline void VXGE_COMPLETE_ALL_TX(struct vxgedev *vdev)
113 {
114         int i;
115
116         /* Complete all transmits */
117         for (i = 0; i < vdev->no_of_vpath; i++)
118                 VXGE_COMPLETE_VPATH_TX(&vdev->vpaths[i].fifo);
119 }
120
121 static inline void VXGE_COMPLETE_ALL_RX(struct vxgedev *vdev)
122 {
123         int i;
124         struct vxge_ring *ring;
125
126         /* Complete all receives*/
127         for (i = 0; i < vdev->no_of_vpath; i++) {
128                 ring = &vdev->vpaths[i].ring;
129                 vxge_hw_vpath_poll_rx(ring->handle);
130         }
131 }
132
133 /*
134  * MultiQ manipulation helper functions
135  */
136 void vxge_stop_all_tx_queue(struct vxgedev *vdev)
137 {
138         int i;
139         struct net_device *dev = vdev->ndev;
140
141         if (vdev->config.tx_steering_type != TX_MULTIQ_STEERING) {
142                 for (i = 0; i < vdev->no_of_vpath; i++)
143                         vdev->vpaths[i].fifo.queue_state = VPATH_QUEUE_STOP;
144         }
145         netif_tx_stop_all_queues(dev);
146 }
147
148 void vxge_stop_tx_queue(struct vxge_fifo *fifo)
149 {
150         struct net_device *dev = fifo->ndev;
151
152         struct netdev_queue *txq = NULL;
153         if (fifo->tx_steering_type == TX_MULTIQ_STEERING)
154                 txq = netdev_get_tx_queue(dev, fifo->driver_id);
155         else {
156                 txq = netdev_get_tx_queue(dev, 0);
157                 fifo->queue_state = VPATH_QUEUE_STOP;
158         }
159
160         netif_tx_stop_queue(txq);
161 }
162
163 void vxge_start_all_tx_queue(struct vxgedev *vdev)
164 {
165         int i;
166         struct net_device *dev = vdev->ndev;
167
168         if (vdev->config.tx_steering_type != TX_MULTIQ_STEERING) {
169                 for (i = 0; i < vdev->no_of_vpath; i++)
170                         vdev->vpaths[i].fifo.queue_state = VPATH_QUEUE_START;
171         }
172         netif_tx_start_all_queues(dev);
173 }
174
175 static void vxge_wake_all_tx_queue(struct vxgedev *vdev)
176 {
177         int i;
178         struct net_device *dev = vdev->ndev;
179
180         if (vdev->config.tx_steering_type != TX_MULTIQ_STEERING) {
181                 for (i = 0; i < vdev->no_of_vpath; i++)
182                         vdev->vpaths[i].fifo.queue_state = VPATH_QUEUE_START;
183         }
184         netif_tx_wake_all_queues(dev);
185 }
186
187 void vxge_wake_tx_queue(struct vxge_fifo *fifo, struct sk_buff *skb)
188 {
189         struct net_device *dev = fifo->ndev;
190
191         int vpath_no = fifo->driver_id;
192         struct netdev_queue *txq = NULL;
193         if (fifo->tx_steering_type == TX_MULTIQ_STEERING) {
194                 txq = netdev_get_tx_queue(dev, vpath_no);
195                 if (netif_tx_queue_stopped(txq))
196                         netif_tx_wake_queue(txq);
197         } else {
198                 txq = netdev_get_tx_queue(dev, 0);
199                 if (fifo->queue_state == VPATH_QUEUE_STOP)
200                         if (netif_tx_queue_stopped(txq)) {
201                                 fifo->queue_state = VPATH_QUEUE_START;
202                                 netif_tx_wake_queue(txq);
203                         }
204         }
205 }
206
207 /*
208  * vxge_callback_link_up
209  *
210  * This function is called during interrupt context to notify link up state
211  * change.
212  */
213 void
214 vxge_callback_link_up(struct __vxge_hw_device *hldev)
215 {
216         struct net_device *dev = hldev->ndev;
217         struct vxgedev *vdev = (struct vxgedev *)netdev_priv(dev);
218
219         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
220                 vdev->ndev->name, __func__, __LINE__);
221         printk(KERN_NOTICE "%s: Link Up\n", vdev->ndev->name);
222         vdev->stats.link_up++;
223
224         netif_carrier_on(vdev->ndev);
225         vxge_wake_all_tx_queue(vdev);
226
227         vxge_debug_entryexit(VXGE_TRACE,
228                 "%s: %s:%d Exiting...", vdev->ndev->name, __func__, __LINE__);
229 }
230
231 /*
232  * vxge_callback_link_down
233  *
234  * This function is called during interrupt context to notify link down state
235  * change.
236  */
237 void
238 vxge_callback_link_down(struct __vxge_hw_device *hldev)
239 {
240         struct net_device *dev = hldev->ndev;
241         struct vxgedev *vdev = (struct vxgedev *)netdev_priv(dev);
242
243         vxge_debug_entryexit(VXGE_TRACE,
244                 "%s: %s:%d", vdev->ndev->name, __func__, __LINE__);
245         printk(KERN_NOTICE "%s: Link Down\n", vdev->ndev->name);
246
247         vdev->stats.link_down++;
248         netif_carrier_off(vdev->ndev);
249         vxge_stop_all_tx_queue(vdev);
250
251         vxge_debug_entryexit(VXGE_TRACE,
252                 "%s: %s:%d Exiting...", vdev->ndev->name, __func__, __LINE__);
253 }
254
255 /*
256  * vxge_rx_alloc
257  *
258  * Allocate SKB.
259  */
260 static struct sk_buff*
261 vxge_rx_alloc(void *dtrh, struct vxge_ring *ring, const int skb_size)
262 {
263         struct net_device    *dev;
264         struct sk_buff       *skb;
265         struct vxge_rx_priv *rx_priv;
266
267         dev = ring->ndev;
268         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
269                 ring->ndev->name, __func__, __LINE__);
270
271         rx_priv = vxge_hw_ring_rxd_private_get(dtrh);
272
273         /* try to allocate skb first. this one may fail */
274         skb = netdev_alloc_skb(dev, skb_size +
275         VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
276         if (skb == NULL) {
277                 vxge_debug_mem(VXGE_ERR,
278                         "%s: out of memory to allocate SKB", dev->name);
279                 ring->stats.skb_alloc_fail++;
280                 return NULL;
281         }
282
283         vxge_debug_mem(VXGE_TRACE,
284                 "%s: %s:%d  Skb : 0x%p", ring->ndev->name,
285                 __func__, __LINE__, skb);
286
287         skb_reserve(skb, VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
288
289         rx_priv->skb = skb;
290         rx_priv->skb_data = NULL;
291         rx_priv->data_size = skb_size;
292         vxge_debug_entryexit(VXGE_TRACE,
293                 "%s: %s:%d Exiting...", ring->ndev->name, __func__, __LINE__);
294
295         return skb;
296 }
297
298 /*
299  * vxge_rx_map
300  */
301 static int vxge_rx_map(void *dtrh, struct vxge_ring *ring)
302 {
303         struct vxge_rx_priv *rx_priv;
304         dma_addr_t dma_addr;
305
306         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
307                 ring->ndev->name, __func__, __LINE__);
308         rx_priv = vxge_hw_ring_rxd_private_get(dtrh);
309
310         rx_priv->skb_data = rx_priv->skb->data;
311         dma_addr = pci_map_single(ring->pdev, rx_priv->skb_data,
312                                 rx_priv->data_size, PCI_DMA_FROMDEVICE);
313
314         if (unlikely(pci_dma_mapping_error(ring->pdev, dma_addr))) {
315                 ring->stats.pci_map_fail++;
316                 return -EIO;
317         }
318         vxge_debug_mem(VXGE_TRACE,
319                 "%s: %s:%d  1 buffer mode dma_addr = 0x%llx",
320                 ring->ndev->name, __func__, __LINE__,
321                 (unsigned long long)dma_addr);
322         vxge_hw_ring_rxd_1b_set(dtrh, dma_addr, rx_priv->data_size);
323
324         rx_priv->data_dma = dma_addr;
325         vxge_debug_entryexit(VXGE_TRACE,
326                 "%s: %s:%d Exiting...", ring->ndev->name, __func__, __LINE__);
327
328         return 0;
329 }
330
331 /*
332  * vxge_rx_initial_replenish
333  * Allocation of RxD as an initial replenish procedure.
334  */
335 static enum vxge_hw_status
336 vxge_rx_initial_replenish(void *dtrh, void *userdata)
337 {
338         struct vxge_ring *ring = (struct vxge_ring *)userdata;
339         struct vxge_rx_priv *rx_priv;
340
341         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
342                 ring->ndev->name, __func__, __LINE__);
343         if (vxge_rx_alloc(dtrh, ring,
344                           VXGE_LL_MAX_FRAME_SIZE(ring->ndev)) == NULL)
345                 return VXGE_HW_FAIL;
346
347         if (vxge_rx_map(dtrh, ring)) {
348                 rx_priv = vxge_hw_ring_rxd_private_get(dtrh);
349                 dev_kfree_skb(rx_priv->skb);
350
351                 return VXGE_HW_FAIL;
352         }
353         vxge_debug_entryexit(VXGE_TRACE,
354                 "%s: %s:%d Exiting...", ring->ndev->name, __func__, __LINE__);
355
356         return VXGE_HW_OK;
357 }
358
359 static inline void
360 vxge_rx_complete(struct vxge_ring *ring, struct sk_buff *skb, u16 vlan,
361                  int pkt_length, struct vxge_hw_ring_rxd_info *ext_info)
362 {
363
364         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
365                         ring->ndev->name, __func__, __LINE__);
366         skb_record_rx_queue(skb, ring->driver_id);
367         skb->protocol = eth_type_trans(skb, ring->ndev);
368
369         ring->stats.rx_frms++;
370         ring->stats.rx_bytes += pkt_length;
371
372         if (skb->pkt_type == PACKET_MULTICAST)
373                 ring->stats.rx_mcast++;
374
375         vxge_debug_rx(VXGE_TRACE,
376                 "%s: %s:%d  skb protocol = %d",
377                 ring->ndev->name, __func__, __LINE__, skb->protocol);
378
379         if (ring->gro_enable) {
380                 if (ring->vlgrp && ext_info->vlan &&
381                         (ring->vlan_tag_strip ==
382                                 VXGE_HW_VPATH_RPA_STRIP_VLAN_TAG_ENABLE))
383                         vlan_gro_receive(ring->napi_p, ring->vlgrp,
384                                         ext_info->vlan, skb);
385                 else
386                         napi_gro_receive(ring->napi_p, skb);
387         } else {
388                 if (ring->vlgrp && vlan &&
389                         (ring->vlan_tag_strip ==
390                                 VXGE_HW_VPATH_RPA_STRIP_VLAN_TAG_ENABLE))
391                         vlan_hwaccel_receive_skb(skb, ring->vlgrp, vlan);
392                 else
393                         netif_receive_skb(skb);
394         }
395         vxge_debug_entryexit(VXGE_TRACE,
396                 "%s: %s:%d Exiting...", ring->ndev->name, __func__, __LINE__);
397 }
398
399 static inline void vxge_re_pre_post(void *dtr, struct vxge_ring *ring,
400                                     struct vxge_rx_priv *rx_priv)
401 {
402         pci_dma_sync_single_for_device(ring->pdev,
403                 rx_priv->data_dma, rx_priv->data_size, PCI_DMA_FROMDEVICE);
404
405         vxge_hw_ring_rxd_1b_set(dtr, rx_priv->data_dma, rx_priv->data_size);
406         vxge_hw_ring_rxd_pre_post(ring->handle, dtr);
407 }
408
409 static inline void vxge_post(int *dtr_cnt, void **first_dtr,
410                              void *post_dtr, struct __vxge_hw_ring *ringh)
411 {
412         int dtr_count = *dtr_cnt;
413         if ((*dtr_cnt % VXGE_HW_RXSYNC_FREQ_CNT) == 0) {
414                 if (*first_dtr)
415                         vxge_hw_ring_rxd_post_post_wmb(ringh, *first_dtr);
416                 *first_dtr = post_dtr;
417         } else
418                 vxge_hw_ring_rxd_post_post(ringh, post_dtr);
419         dtr_count++;
420         *dtr_cnt = dtr_count;
421 }
422
423 /*
424  * vxge_rx_1b_compl
425  *
426  * If the interrupt is because of a received frame or if the receive ring
427  * contains fresh as yet un-processed frames, this function is called.
428  */
429 enum vxge_hw_status
430 vxge_rx_1b_compl(struct __vxge_hw_ring *ringh, void *dtr,
431                  u8 t_code, void *userdata)
432 {
433         struct vxge_ring *ring = (struct vxge_ring *)userdata;
434         struct  net_device *dev = ring->ndev;
435         unsigned int dma_sizes;
436         void *first_dtr = NULL;
437         int dtr_cnt = 0;
438         int data_size;
439         dma_addr_t data_dma;
440         int pkt_length;
441         struct sk_buff *skb;
442         struct vxge_rx_priv *rx_priv;
443         struct vxge_hw_ring_rxd_info ext_info;
444         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
445                 ring->ndev->name, __func__, __LINE__);
446         ring->pkts_processed = 0;
447
448         vxge_hw_ring_replenish(ringh);
449
450         do {
451                 prefetch((char *)dtr + L1_CACHE_BYTES);
452                 rx_priv = vxge_hw_ring_rxd_private_get(dtr);
453                 skb = rx_priv->skb;
454                 data_size = rx_priv->data_size;
455                 data_dma = rx_priv->data_dma;
456                 prefetch(rx_priv->skb_data);
457
458                 vxge_debug_rx(VXGE_TRACE,
459                         "%s: %s:%d  skb = 0x%p",
460                         ring->ndev->name, __func__, __LINE__, skb);
461
462                 vxge_hw_ring_rxd_1b_get(ringh, dtr, &dma_sizes);
463                 pkt_length = dma_sizes;
464
465                 pkt_length -= ETH_FCS_LEN;
466
467                 vxge_debug_rx(VXGE_TRACE,
468                         "%s: %s:%d  Packet Length = %d",
469                         ring->ndev->name, __func__, __LINE__, pkt_length);
470
471                 vxge_hw_ring_rxd_1b_info_get(ringh, dtr, &ext_info);
472
473                 /* check skb validity */
474                 vxge_assert(skb);
475
476                 prefetch((char *)skb + L1_CACHE_BYTES);
477                 if (unlikely(t_code)) {
478
479                         if (vxge_hw_ring_handle_tcode(ringh, dtr, t_code) !=
480                                 VXGE_HW_OK) {
481
482                                 ring->stats.rx_errors++;
483                                 vxge_debug_rx(VXGE_TRACE,
484                                         "%s: %s :%d Rx T_code is %d",
485                                         ring->ndev->name, __func__,
486                                         __LINE__, t_code);
487
488                                 /* If the t_code is not supported and if the
489                                  * t_code is other than 0x5 (unparseable packet
490                                  * such as unknown UPV6 header), Drop it !!!
491                                  */
492                                 vxge_re_pre_post(dtr, ring, rx_priv);
493
494                                 vxge_post(&dtr_cnt, &first_dtr, dtr, ringh);
495                                 ring->stats.rx_dropped++;
496                                 continue;
497                         }
498                 }
499
500                 if (pkt_length > VXGE_LL_RX_COPY_THRESHOLD) {
501
502                         if (vxge_rx_alloc(dtr, ring, data_size) != NULL) {
503
504                                 if (!vxge_rx_map(dtr, ring)) {
505                                         skb_put(skb, pkt_length);
506
507                                         pci_unmap_single(ring->pdev, data_dma,
508                                                 data_size, PCI_DMA_FROMDEVICE);
509
510                                         vxge_hw_ring_rxd_pre_post(ringh, dtr);
511                                         vxge_post(&dtr_cnt, &first_dtr, dtr,
512                                                 ringh);
513                                 } else {
514                                         dev_kfree_skb(rx_priv->skb);
515                                         rx_priv->skb = skb;
516                                         rx_priv->data_size = data_size;
517                                         vxge_re_pre_post(dtr, ring, rx_priv);
518
519                                         vxge_post(&dtr_cnt, &first_dtr, dtr,
520                                                 ringh);
521                                         ring->stats.rx_dropped++;
522                                         break;
523                                 }
524                         } else {
525                                 vxge_re_pre_post(dtr, ring, rx_priv);
526
527                                 vxge_post(&dtr_cnt, &first_dtr, dtr, ringh);
528                                 ring->stats.rx_dropped++;
529                                 break;
530                         }
531                 } else {
532                         struct sk_buff *skb_up;
533
534                         skb_up = netdev_alloc_skb(dev, pkt_length +
535                                 VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
536                         if (skb_up != NULL) {
537                                 skb_reserve(skb_up,
538                                     VXGE_HW_HEADER_ETHERNET_II_802_3_ALIGN);
539
540                                 pci_dma_sync_single_for_cpu(ring->pdev,
541                                         data_dma, data_size,
542                                         PCI_DMA_FROMDEVICE);
543
544                                 vxge_debug_mem(VXGE_TRACE,
545                                         "%s: %s:%d  skb_up = %p",
546                                         ring->ndev->name, __func__,
547                                         __LINE__, skb);
548                                 memcpy(skb_up->data, skb->data, pkt_length);
549
550                                 vxge_re_pre_post(dtr, ring, rx_priv);
551
552                                 vxge_post(&dtr_cnt, &first_dtr, dtr,
553                                         ringh);
554                                 /* will netif_rx small SKB instead */
555                                 skb = skb_up;
556                                 skb_put(skb, pkt_length);
557                         } else {
558                                 vxge_re_pre_post(dtr, ring, rx_priv);
559
560                                 vxge_post(&dtr_cnt, &first_dtr, dtr, ringh);
561                                 vxge_debug_rx(VXGE_ERR,
562                                         "%s: vxge_rx_1b_compl: out of "
563                                         "memory", dev->name);
564                                 ring->stats.skb_alloc_fail++;
565                                 break;
566                         }
567                 }
568
569                 if ((ext_info.proto & VXGE_HW_FRAME_PROTO_TCP_OR_UDP) &&
570                     !(ext_info.proto & VXGE_HW_FRAME_PROTO_IP_FRAG) &&
571                     ring->rx_csum && /* Offload Rx side CSUM */
572                     ext_info.l3_cksum == VXGE_HW_L3_CKSUM_OK &&
573                     ext_info.l4_cksum == VXGE_HW_L4_CKSUM_OK)
574                         skb->ip_summed = CHECKSUM_UNNECESSARY;
575                 else
576                         skb->ip_summed = CHECKSUM_NONE;
577
578                 vxge_rx_complete(ring, skb, ext_info.vlan,
579                         pkt_length, &ext_info);
580
581                 ring->budget--;
582                 ring->pkts_processed++;
583                 if (!ring->budget)
584                         break;
585
586         } while (vxge_hw_ring_rxd_next_completed(ringh, &dtr,
587                 &t_code) == VXGE_HW_OK);
588
589         if (first_dtr)
590                 vxge_hw_ring_rxd_post_post_wmb(ringh, first_dtr);
591
592         vxge_debug_entryexit(VXGE_TRACE,
593                                 "%s:%d  Exiting...",
594                                 __func__, __LINE__);
595         return VXGE_HW_OK;
596 }
597
598 /*
599  * vxge_xmit_compl
600  *
601  * If an interrupt was raised to indicate DMA complete of the Tx packet,
602  * this function is called. It identifies the last TxD whose buffer was
603  * freed and frees all skbs whose data have already DMA'ed into the NICs
604  * internal memory.
605  */
606 enum vxge_hw_status
607 vxge_xmit_compl(struct __vxge_hw_fifo *fifo_hw, void *dtr,
608                 enum vxge_hw_fifo_tcode t_code, void *userdata,
609                 struct sk_buff ***skb_ptr, int nr_skb, int *more)
610 {
611         struct vxge_fifo *fifo = (struct vxge_fifo *)userdata;
612         struct sk_buff *skb, **done_skb = *skb_ptr;
613         int pkt_cnt = 0;
614
615         vxge_debug_entryexit(VXGE_TRACE,
616                 "%s:%d Entered....", __func__, __LINE__);
617
618         do {
619                 int frg_cnt;
620                 skb_frag_t *frag;
621                 int i = 0, j;
622                 struct vxge_tx_priv *txd_priv =
623                         vxge_hw_fifo_txdl_private_get(dtr);
624
625                 skb = txd_priv->skb;
626                 frg_cnt = skb_shinfo(skb)->nr_frags;
627                 frag = &skb_shinfo(skb)->frags[0];
628
629                 vxge_debug_tx(VXGE_TRACE,
630                                 "%s: %s:%d fifo_hw = %p dtr = %p "
631                                 "tcode = 0x%x", fifo->ndev->name, __func__,
632                                 __LINE__, fifo_hw, dtr, t_code);
633                 /* check skb validity */
634                 vxge_assert(skb);
635                 vxge_debug_tx(VXGE_TRACE,
636                         "%s: %s:%d skb = %p itxd_priv = %p frg_cnt = %d",
637                         fifo->ndev->name, __func__, __LINE__,
638                         skb, txd_priv, frg_cnt);
639                 if (unlikely(t_code)) {
640                         fifo->stats.tx_errors++;
641                         vxge_debug_tx(VXGE_ERR,
642                                 "%s: tx: dtr %p completed due to "
643                                 "error t_code %01x", fifo->ndev->name,
644                                 dtr, t_code);
645                         vxge_hw_fifo_handle_tcode(fifo_hw, dtr, t_code);
646                 }
647
648                 /*  for unfragmented skb */
649                 pci_unmap_single(fifo->pdev, txd_priv->dma_buffers[i++],
650                                 skb_headlen(skb), PCI_DMA_TODEVICE);
651
652                 for (j = 0; j < frg_cnt; j++) {
653                         pci_unmap_page(fifo->pdev,
654                                         txd_priv->dma_buffers[i++],
655                                         frag->size, PCI_DMA_TODEVICE);
656                         frag += 1;
657                 }
658
659                 vxge_hw_fifo_txdl_free(fifo_hw, dtr);
660
661                 /* Updating the statistics block */
662                 fifo->stats.tx_frms++;
663                 fifo->stats.tx_bytes += skb->len;
664
665                 *done_skb++ = skb;
666
667                 if (--nr_skb <= 0) {
668                         *more = 1;
669                         break;
670                 }
671
672                 pkt_cnt++;
673                 if (pkt_cnt > fifo->indicate_max_pkts)
674                         break;
675
676         } while (vxge_hw_fifo_txdl_next_completed(fifo_hw,
677                                 &dtr, &t_code) == VXGE_HW_OK);
678
679         *skb_ptr = done_skb;
680         vxge_wake_tx_queue(fifo, skb);
681
682         vxge_debug_entryexit(VXGE_TRACE,
683                                 "%s: %s:%d  Exiting...",
684                                 fifo->ndev->name, __func__, __LINE__);
685         return VXGE_HW_OK;
686 }
687
688 /* select a vpath to transmit the packet */
689 static u32 vxge_get_vpath_no(struct vxgedev *vdev, struct sk_buff *skb,
690         int *do_lock)
691 {
692         u16 queue_len, counter = 0;
693         if (skb->protocol == htons(ETH_P_IP)) {
694                 struct iphdr *ip;
695                 struct tcphdr *th;
696
697                 ip = ip_hdr(skb);
698
699                 if ((ip->frag_off & htons(IP_OFFSET|IP_MF)) == 0) {
700                         th = (struct tcphdr *)(((unsigned char *)ip) +
701                                         ip->ihl*4);
702
703                         queue_len = vdev->no_of_vpath;
704                         counter = (ntohs(th->source) +
705                                 ntohs(th->dest)) &
706                                 vdev->vpath_selector[queue_len - 1];
707                         if (counter >= queue_len)
708                                 counter = queue_len - 1;
709
710                         if (ip->protocol == IPPROTO_UDP) {
711 #ifdef NETIF_F_LLTX
712                                 *do_lock = 0;
713 #endif
714                         }
715                 }
716         }
717         return counter;
718 }
719
720 static enum vxge_hw_status vxge_search_mac_addr_in_list(
721         struct vxge_vpath *vpath, u64 del_mac)
722 {
723         struct list_head *entry, *next;
724         list_for_each_safe(entry, next, &vpath->mac_addr_list) {
725                 if (((struct vxge_mac_addrs *)entry)->macaddr == del_mac)
726                         return TRUE;
727         }
728         return FALSE;
729 }
730
731 static int vxge_learn_mac(struct vxgedev *vdev, u8 *mac_header)
732 {
733         struct macInfo mac_info;
734         u8 *mac_address = NULL;
735         u64 mac_addr = 0, vpath_vector = 0;
736         int vpath_idx = 0;
737         enum vxge_hw_status status = VXGE_HW_OK;
738         struct vxge_vpath *vpath = NULL;
739         struct __vxge_hw_device *hldev;
740
741         hldev = (struct __vxge_hw_device *) pci_get_drvdata(vdev->pdev);
742
743         mac_address = (u8 *)&mac_addr;
744         memcpy(mac_address, mac_header, ETH_ALEN);
745
746         /* Is this mac address already in the list? */
747         for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
748                 vpath = &vdev->vpaths[vpath_idx];
749                 if (vxge_search_mac_addr_in_list(vpath, mac_addr))
750                         return vpath_idx;
751         }
752
753         memset(&mac_info, 0, sizeof(struct macInfo));
754         memcpy(mac_info.macaddr, mac_header, ETH_ALEN);
755
756         /* Any vpath has room to add mac address to its da table? */
757         for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
758                 vpath = &vdev->vpaths[vpath_idx];
759                 if (vpath->mac_addr_cnt < vpath->max_mac_addr_cnt) {
760                         /* Add this mac address to this vpath */
761                         mac_info.vpath_no = vpath_idx;
762                         mac_info.state = VXGE_LL_MAC_ADDR_IN_DA_TABLE;
763                         status = vxge_add_mac_addr(vdev, &mac_info);
764                         if (status != VXGE_HW_OK)
765                                 return -EPERM;
766                         return vpath_idx;
767                 }
768         }
769
770         mac_info.state = VXGE_LL_MAC_ADDR_IN_LIST;
771         vpath_idx = 0;
772         mac_info.vpath_no = vpath_idx;
773         /* Is the first vpath already selected as catch-basin ? */
774         vpath = &vdev->vpaths[vpath_idx];
775         if (vpath->mac_addr_cnt > vpath->max_mac_addr_cnt) {
776                 /* Add this mac address to this vpath */
777                 if (FALSE == vxge_mac_list_add(vpath, &mac_info))
778                         return -EPERM;
779                 return vpath_idx;
780         }
781
782         /* Select first vpath as catch-basin */
783         vpath_vector = vxge_mBIT(vpath->device_id);
784         status = vxge_hw_mgmt_reg_write(vpath->vdev->devh,
785                                 vxge_hw_mgmt_reg_type_mrpcim,
786                                 0,
787                                 (ulong)offsetof(
788                                         struct vxge_hw_mrpcim_reg,
789                                         rts_mgr_cbasin_cfg),
790                                 vpath_vector);
791         if (status != VXGE_HW_OK) {
792                 vxge_debug_tx(VXGE_ERR,
793                         "%s: Unable to set the vpath-%d in catch-basin mode",
794                         VXGE_DRIVER_NAME, vpath->device_id);
795                 return -EPERM;
796         }
797
798         if (FALSE == vxge_mac_list_add(vpath, &mac_info))
799                 return -EPERM;
800
801         return vpath_idx;
802 }
803
804 /**
805  * vxge_xmit
806  * @skb : the socket buffer containing the Tx data.
807  * @dev : device pointer.
808  *
809  * This function is the Tx entry point of the driver. Neterion NIC supports
810  * certain protocol assist features on Tx side, namely  CSO, S/G, LSO.
811  * NOTE: when device cant queue the pkt, just the trans_start variable will
812  * not be upadted.
813 */
814 static netdev_tx_t
815 vxge_xmit(struct sk_buff *skb, struct net_device *dev)
816 {
817         struct vxge_fifo *fifo = NULL;
818         void *dtr_priv;
819         void *dtr = NULL;
820         struct vxgedev *vdev = NULL;
821         enum vxge_hw_status status;
822         int frg_cnt, first_frg_len;
823         skb_frag_t *frag;
824         int i = 0, j = 0, avail;
825         u64 dma_pointer;
826         struct vxge_tx_priv *txdl_priv = NULL;
827         struct __vxge_hw_fifo *fifo_hw;
828         int offload_type;
829         unsigned long flags = 0;
830         int vpath_no = 0;
831         int do_spin_tx_lock = 1;
832
833         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
834                         dev->name, __func__, __LINE__);
835
836         /* A buffer with no data will be dropped */
837         if (unlikely(skb->len <= 0)) {
838                 vxge_debug_tx(VXGE_ERR,
839                         "%s: Buffer has no data..", dev->name);
840                 dev_kfree_skb(skb);
841                 return NETDEV_TX_OK;
842         }
843
844         vdev = (struct vxgedev *)netdev_priv(dev);
845
846         if (unlikely(!is_vxge_card_up(vdev))) {
847                 vxge_debug_tx(VXGE_ERR,
848                         "%s: vdev not initialized", dev->name);
849                 dev_kfree_skb(skb);
850                 return NETDEV_TX_OK;
851         }
852
853         if (vdev->config.addr_learn_en) {
854                 vpath_no = vxge_learn_mac(vdev, skb->data + ETH_ALEN);
855                 if (vpath_no == -EPERM) {
856                         vxge_debug_tx(VXGE_ERR,
857                                 "%s: Failed to store the mac address",
858                                 dev->name);
859                         dev_kfree_skb(skb);
860                         return NETDEV_TX_OK;
861                 }
862         }
863
864         if (vdev->config.tx_steering_type == TX_MULTIQ_STEERING)
865                 vpath_no = skb_get_queue_mapping(skb);
866         else if (vdev->config.tx_steering_type == TX_PORT_STEERING)
867                 vpath_no = vxge_get_vpath_no(vdev, skb, &do_spin_tx_lock);
868
869         vxge_debug_tx(VXGE_TRACE, "%s: vpath_no= %d", dev->name, vpath_no);
870
871         if (vpath_no >= vdev->no_of_vpath)
872                 vpath_no = 0;
873
874         fifo = &vdev->vpaths[vpath_no].fifo;
875         fifo_hw = fifo->handle;
876
877         if (do_spin_tx_lock)
878                 spin_lock_irqsave(&fifo->tx_lock, flags);
879         else {
880                 if (unlikely(!spin_trylock_irqsave(&fifo->tx_lock, flags)))
881                         return NETDEV_TX_LOCKED;
882         }
883
884         if (vdev->config.tx_steering_type == TX_MULTIQ_STEERING) {
885                 if (netif_subqueue_stopped(dev, skb)) {
886                         spin_unlock_irqrestore(&fifo->tx_lock, flags);
887                         return NETDEV_TX_BUSY;
888                 }
889         } else if (unlikely(fifo->queue_state == VPATH_QUEUE_STOP)) {
890                 if (netif_queue_stopped(dev)) {
891                         spin_unlock_irqrestore(&fifo->tx_lock, flags);
892                         return NETDEV_TX_BUSY;
893                 }
894         }
895         avail = vxge_hw_fifo_free_txdl_count_get(fifo_hw);
896         if (avail == 0) {
897                 vxge_debug_tx(VXGE_ERR,
898                         "%s: No free TXDs available", dev->name);
899                 fifo->stats.txd_not_free++;
900                 vxge_stop_tx_queue(fifo);
901                 goto _exit2;
902         }
903
904         /* Last TXD?  Stop tx queue to avoid dropping packets.  TX
905          * completion will resume the queue.
906          */
907         if (avail == 1)
908                 vxge_stop_tx_queue(fifo);
909
910         status = vxge_hw_fifo_txdl_reserve(fifo_hw, &dtr, &dtr_priv);
911         if (unlikely(status != VXGE_HW_OK)) {
912                 vxge_debug_tx(VXGE_ERR,
913                    "%s: Out of descriptors .", dev->name);
914                 fifo->stats.txd_out_of_desc++;
915                 vxge_stop_tx_queue(fifo);
916                 goto _exit2;
917         }
918
919         vxge_debug_tx(VXGE_TRACE,
920                 "%s: %s:%d fifo_hw = %p dtr = %p dtr_priv = %p",
921                 dev->name, __func__, __LINE__,
922                 fifo_hw, dtr, dtr_priv);
923
924         if (vdev->vlgrp && vlan_tx_tag_present(skb)) {
925                 u16 vlan_tag = vlan_tx_tag_get(skb);
926                 vxge_hw_fifo_txdl_vlan_set(dtr, vlan_tag);
927         }
928
929         first_frg_len = skb_headlen(skb);
930
931         dma_pointer = pci_map_single(fifo->pdev, skb->data, first_frg_len,
932                                 PCI_DMA_TODEVICE);
933
934         if (unlikely(pci_dma_mapping_error(fifo->pdev, dma_pointer))) {
935                 vxge_hw_fifo_txdl_free(fifo_hw, dtr);
936                 vxge_stop_tx_queue(fifo);
937                 fifo->stats.pci_map_fail++;
938                 goto _exit2;
939         }
940
941         txdl_priv = vxge_hw_fifo_txdl_private_get(dtr);
942         txdl_priv->skb = skb;
943         txdl_priv->dma_buffers[j] = dma_pointer;
944
945         frg_cnt = skb_shinfo(skb)->nr_frags;
946         vxge_debug_tx(VXGE_TRACE,
947                         "%s: %s:%d skb = %p txdl_priv = %p "
948                         "frag_cnt = %d dma_pointer = 0x%llx", dev->name,
949                         __func__, __LINE__, skb, txdl_priv,
950                         frg_cnt, (unsigned long long)dma_pointer);
951
952         vxge_hw_fifo_txdl_buffer_set(fifo_hw, dtr, j++, dma_pointer,
953                 first_frg_len);
954
955         frag = &skb_shinfo(skb)->frags[0];
956         for (i = 0; i < frg_cnt; i++) {
957                 /* ignore 0 length fragment */
958                 if (!frag->size)
959                         continue;
960
961                 dma_pointer =
962                         (u64)pci_map_page(fifo->pdev, frag->page,
963                                 frag->page_offset, frag->size,
964                                 PCI_DMA_TODEVICE);
965
966                 if (unlikely(pci_dma_mapping_error(fifo->pdev, dma_pointer)))
967                         goto _exit0;
968                 vxge_debug_tx(VXGE_TRACE,
969                         "%s: %s:%d frag = %d dma_pointer = 0x%llx",
970                                 dev->name, __func__, __LINE__, i,
971                                 (unsigned long long)dma_pointer);
972
973                 txdl_priv->dma_buffers[j] = dma_pointer;
974                 vxge_hw_fifo_txdl_buffer_set(fifo_hw, dtr, j++, dma_pointer,
975                                         frag->size);
976                 frag += 1;
977         }
978
979         offload_type = vxge_offload_type(skb);
980
981         if (offload_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)) {
982
983                 int mss = vxge_tcp_mss(skb);
984                 if (mss) {
985                         vxge_debug_tx(VXGE_TRACE,
986                                 "%s: %s:%d mss = %d",
987                                 dev->name, __func__, __LINE__, mss);
988                         vxge_hw_fifo_txdl_mss_set(dtr, mss);
989                 } else {
990                         vxge_assert(skb->len <=
991                                 dev->mtu + VXGE_HW_MAC_HEADER_MAX_SIZE);
992                         vxge_assert(0);
993                         goto _exit1;
994                 }
995         }
996
997         if (skb->ip_summed == CHECKSUM_PARTIAL)
998                 vxge_hw_fifo_txdl_cksum_set_bits(dtr,
999                                         VXGE_HW_FIFO_TXD_TX_CKO_IPV4_EN |
1000                                         VXGE_HW_FIFO_TXD_TX_CKO_TCP_EN |
1001                                         VXGE_HW_FIFO_TXD_TX_CKO_UDP_EN);
1002
1003         vxge_hw_fifo_txdl_post(fifo_hw, dtr);
1004 #ifdef NETIF_F_LLTX
1005         dev->trans_start = jiffies; /* NETIF_F_LLTX driver :( */
1006 #endif
1007         spin_unlock_irqrestore(&fifo->tx_lock, flags);
1008
1009         VXGE_COMPLETE_VPATH_TX(fifo);
1010         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d  Exiting...",
1011                 dev->name, __func__, __LINE__);
1012         return NETDEV_TX_OK;
1013
1014 _exit0:
1015         vxge_debug_tx(VXGE_TRACE, "%s: pci_map_page failed", dev->name);
1016
1017 _exit1:
1018         j = 0;
1019         frag = &skb_shinfo(skb)->frags[0];
1020
1021         pci_unmap_single(fifo->pdev, txdl_priv->dma_buffers[j++],
1022                         skb_headlen(skb), PCI_DMA_TODEVICE);
1023
1024         for (; j < i; j++) {
1025                 pci_unmap_page(fifo->pdev, txdl_priv->dma_buffers[j],
1026                         frag->size, PCI_DMA_TODEVICE);
1027                 frag += 1;
1028         }
1029
1030         vxge_hw_fifo_txdl_free(fifo_hw, dtr);
1031 _exit2:
1032         dev_kfree_skb(skb);
1033         spin_unlock_irqrestore(&fifo->tx_lock, flags);
1034         VXGE_COMPLETE_VPATH_TX(fifo);
1035
1036         return NETDEV_TX_OK;
1037 }
1038
1039 /*
1040  * vxge_rx_term
1041  *
1042  * Function will be called by hw function to abort all outstanding receive
1043  * descriptors.
1044  */
1045 static void
1046 vxge_rx_term(void *dtrh, enum vxge_hw_rxd_state state, void *userdata)
1047 {
1048         struct vxge_ring *ring = (struct vxge_ring *)userdata;
1049         struct vxge_rx_priv *rx_priv =
1050                 vxge_hw_ring_rxd_private_get(dtrh);
1051
1052         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
1053                         ring->ndev->name, __func__, __LINE__);
1054         if (state != VXGE_HW_RXD_STATE_POSTED)
1055                 return;
1056
1057         pci_unmap_single(ring->pdev, rx_priv->data_dma,
1058                 rx_priv->data_size, PCI_DMA_FROMDEVICE);
1059
1060         dev_kfree_skb(rx_priv->skb);
1061         rx_priv->skb_data = NULL;
1062
1063         vxge_debug_entryexit(VXGE_TRACE,
1064                 "%s: %s:%d  Exiting...",
1065                 ring->ndev->name, __func__, __LINE__);
1066 }
1067
1068 /*
1069  * vxge_tx_term
1070  *
1071  * Function will be called to abort all outstanding tx descriptors
1072  */
1073 static void
1074 vxge_tx_term(void *dtrh, enum vxge_hw_txdl_state state, void *userdata)
1075 {
1076         struct vxge_fifo *fifo = (struct vxge_fifo *)userdata;
1077         skb_frag_t *frag;
1078         int i = 0, j, frg_cnt;
1079         struct vxge_tx_priv *txd_priv = vxge_hw_fifo_txdl_private_get(dtrh);
1080         struct sk_buff *skb = txd_priv->skb;
1081
1082         vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1083
1084         if (state != VXGE_HW_TXDL_STATE_POSTED)
1085                 return;
1086
1087         /* check skb validity */
1088         vxge_assert(skb);
1089         frg_cnt = skb_shinfo(skb)->nr_frags;
1090         frag = &skb_shinfo(skb)->frags[0];
1091
1092         /*  for unfragmented skb */
1093         pci_unmap_single(fifo->pdev, txd_priv->dma_buffers[i++],
1094                 skb_headlen(skb), PCI_DMA_TODEVICE);
1095
1096         for (j = 0; j < frg_cnt; j++) {
1097                 pci_unmap_page(fifo->pdev, txd_priv->dma_buffers[i++],
1098                                frag->size, PCI_DMA_TODEVICE);
1099                 frag += 1;
1100         }
1101
1102         dev_kfree_skb(skb);
1103
1104         vxge_debug_entryexit(VXGE_TRACE,
1105                 "%s:%d  Exiting...", __func__, __LINE__);
1106 }
1107
1108 /**
1109  * vxge_set_multicast
1110  * @dev: pointer to the device structure
1111  *
1112  * Entry point for multicast address enable/disable
1113  * This function is a driver entry point which gets called by the kernel
1114  * whenever multicast addresses must be enabled/disabled. This also gets
1115  * called to set/reset promiscuous mode. Depending on the deivce flag, we
1116  * determine, if multicast address must be enabled or if promiscuous mode
1117  * is to be disabled etc.
1118  */
1119 static void vxge_set_multicast(struct net_device *dev)
1120 {
1121         struct netdev_hw_addr *ha;
1122         struct vxgedev *vdev;
1123         int i, mcast_cnt = 0;
1124         struct __vxge_hw_device  *hldev;
1125         enum vxge_hw_status status = VXGE_HW_OK;
1126         struct macInfo mac_info;
1127         int vpath_idx = 0;
1128         struct vxge_mac_addrs *mac_entry;
1129         struct list_head *list_head;
1130         struct list_head *entry, *next;
1131         u8 *mac_address = NULL;
1132
1133         vxge_debug_entryexit(VXGE_TRACE,
1134                 "%s:%d", __func__, __LINE__);
1135
1136         vdev = (struct vxgedev *)netdev_priv(dev);
1137         hldev = (struct __vxge_hw_device  *)vdev->devh;
1138
1139         if (unlikely(!is_vxge_card_up(vdev)))
1140                 return;
1141
1142         if ((dev->flags & IFF_ALLMULTI) && (!vdev->all_multi_flg)) {
1143                 for (i = 0; i < vdev->no_of_vpath; i++) {
1144                         vxge_assert(vdev->vpaths[i].is_open);
1145                         status = vxge_hw_vpath_mcast_enable(
1146                                                 vdev->vpaths[i].handle);
1147                         vdev->all_multi_flg = 1;
1148                 }
1149         } else if ((dev->flags & IFF_ALLMULTI) && (vdev->all_multi_flg)) {
1150                 for (i = 0; i < vdev->no_of_vpath; i++) {
1151                         vxge_assert(vdev->vpaths[i].is_open);
1152                         status = vxge_hw_vpath_mcast_disable(
1153                                                 vdev->vpaths[i].handle);
1154                         vdev->all_multi_flg = 1;
1155                 }
1156         }
1157
1158         if (status != VXGE_HW_OK)
1159                 vxge_debug_init(VXGE_ERR,
1160                         "failed to %s multicast, status %d",
1161                         dev->flags & IFF_ALLMULTI ?
1162                         "enable" : "disable", status);
1163
1164         if (!vdev->config.addr_learn_en) {
1165                 if (dev->flags & IFF_PROMISC) {
1166                         for (i = 0; i < vdev->no_of_vpath; i++) {
1167                                 vxge_assert(vdev->vpaths[i].is_open);
1168                                 status = vxge_hw_vpath_promisc_enable(
1169                                                 vdev->vpaths[i].handle);
1170                         }
1171                 } else {
1172                         for (i = 0; i < vdev->no_of_vpath; i++) {
1173                                 vxge_assert(vdev->vpaths[i].is_open);
1174                                 status = vxge_hw_vpath_promisc_disable(
1175                                                 vdev->vpaths[i].handle);
1176                         }
1177                 }
1178         }
1179
1180         memset(&mac_info, 0, sizeof(struct macInfo));
1181         /* Update individual M_CAST address list */
1182         if ((!vdev->all_multi_flg) && netdev_mc_count(dev)) {
1183
1184                 mcast_cnt = vdev->vpaths[0].mcast_addr_cnt;
1185                 list_head = &vdev->vpaths[0].mac_addr_list;
1186                 if ((netdev_mc_count(dev) +
1187                         (vdev->vpaths[0].mac_addr_cnt - mcast_cnt)) >
1188                                 vdev->vpaths[0].max_mac_addr_cnt)
1189                         goto _set_all_mcast;
1190
1191                 /* Delete previous MC's */
1192                 for (i = 0; i < mcast_cnt; i++) {
1193                         if (!list_empty(list_head))
1194                                 mac_entry = (struct vxge_mac_addrs *)
1195                                         list_first_entry(list_head,
1196                                                 struct vxge_mac_addrs,
1197                                                 item);
1198
1199                         list_for_each_safe(entry, next, list_head) {
1200
1201                                 mac_entry = (struct vxge_mac_addrs *) entry;
1202                                 /* Copy the mac address to delete */
1203                                 mac_address = (u8 *)&mac_entry->macaddr;
1204                                 memcpy(mac_info.macaddr, mac_address, ETH_ALEN);
1205
1206                                 /* Is this a multicast address */
1207                                 if (0x01 & mac_info.macaddr[0]) {
1208                                         for (vpath_idx = 0; vpath_idx <
1209                                                 vdev->no_of_vpath;
1210                                                 vpath_idx++) {
1211                                                 mac_info.vpath_no = vpath_idx;
1212                                                 status = vxge_del_mac_addr(
1213                                                                 vdev,
1214                                                                 &mac_info);
1215                                         }
1216                                 }
1217                         }
1218                 }
1219
1220                 /* Add new ones */
1221                 netdev_for_each_mc_addr(ha, dev) {
1222                         memcpy(mac_info.macaddr, ha->addr, ETH_ALEN);
1223                         for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath;
1224                                         vpath_idx++) {
1225                                 mac_info.vpath_no = vpath_idx;
1226                                 mac_info.state = VXGE_LL_MAC_ADDR_IN_DA_TABLE;
1227                                 status = vxge_add_mac_addr(vdev, &mac_info);
1228                                 if (status != VXGE_HW_OK) {
1229                                         vxge_debug_init(VXGE_ERR,
1230                                                 "%s:%d Setting individual"
1231                                                 "multicast address failed",
1232                                                 __func__, __LINE__);
1233                                         goto _set_all_mcast;
1234                                 }
1235                         }
1236                 }
1237
1238                 return;
1239 _set_all_mcast:
1240                 mcast_cnt = vdev->vpaths[0].mcast_addr_cnt;
1241                 /* Delete previous MC's */
1242                 for (i = 0; i < mcast_cnt; i++) {
1243
1244                         list_for_each_safe(entry, next, list_head) {
1245
1246                                 mac_entry = (struct vxge_mac_addrs *) entry;
1247                                 /* Copy the mac address to delete */
1248                                 mac_address = (u8 *)&mac_entry->macaddr;
1249                                 memcpy(mac_info.macaddr, mac_address, ETH_ALEN);
1250
1251                                 /* Is this a multicast address */
1252                                 if (0x01 & mac_info.macaddr[0])
1253                                         break;
1254                         }
1255
1256                         for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath;
1257                                         vpath_idx++) {
1258                                 mac_info.vpath_no = vpath_idx;
1259                                 status = vxge_del_mac_addr(vdev, &mac_info);
1260                         }
1261                 }
1262
1263                 /* Enable all multicast */
1264                 for (i = 0; i < vdev->no_of_vpath; i++) {
1265                         vxge_assert(vdev->vpaths[i].is_open);
1266                         status = vxge_hw_vpath_mcast_enable(
1267                                                 vdev->vpaths[i].handle);
1268                         if (status != VXGE_HW_OK) {
1269                                 vxge_debug_init(VXGE_ERR,
1270                                         "%s:%d Enabling all multicasts failed",
1271                                          __func__, __LINE__);
1272                         }
1273                         vdev->all_multi_flg = 1;
1274                 }
1275                 dev->flags |= IFF_ALLMULTI;
1276         }
1277
1278         vxge_debug_entryexit(VXGE_TRACE,
1279                 "%s:%d  Exiting...", __func__, __LINE__);
1280 }
1281
1282 /**
1283  * vxge_set_mac_addr
1284  * @dev: pointer to the device structure
1285  *
1286  * Update entry "0" (default MAC addr)
1287  */
1288 static int vxge_set_mac_addr(struct net_device *dev, void *p)
1289 {
1290         struct sockaddr *addr = p;
1291         struct vxgedev *vdev;
1292         struct __vxge_hw_device  *hldev;
1293         enum vxge_hw_status status = VXGE_HW_OK;
1294         struct macInfo mac_info_new, mac_info_old;
1295         int vpath_idx = 0;
1296
1297         vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1298
1299         vdev = (struct vxgedev *)netdev_priv(dev);
1300         hldev = vdev->devh;
1301
1302         if (!is_valid_ether_addr(addr->sa_data))
1303                 return -EINVAL;
1304
1305         memset(&mac_info_new, 0, sizeof(struct macInfo));
1306         memset(&mac_info_old, 0, sizeof(struct macInfo));
1307
1308         vxge_debug_entryexit(VXGE_TRACE, "%s:%d  Exiting...",
1309                 __func__, __LINE__);
1310
1311         /* Get the old address */
1312         memcpy(mac_info_old.macaddr, dev->dev_addr, dev->addr_len);
1313
1314         /* Copy the new address */
1315         memcpy(mac_info_new.macaddr, addr->sa_data, dev->addr_len);
1316
1317         /* First delete the old mac address from all the vpaths
1318         as we can't specify the index while adding new mac address */
1319         for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
1320                 struct vxge_vpath *vpath = &vdev->vpaths[vpath_idx];
1321                 if (!vpath->is_open) {
1322                         /* This can happen when this interface is added/removed
1323                         to the bonding interface. Delete this station address
1324                         from the linked list */
1325                         vxge_mac_list_del(vpath, &mac_info_old);
1326
1327                         /* Add this new address to the linked list
1328                         for later restoring */
1329                         vxge_mac_list_add(vpath, &mac_info_new);
1330
1331                         continue;
1332                 }
1333                 /* Delete the station address */
1334                 mac_info_old.vpath_no = vpath_idx;
1335                 status = vxge_del_mac_addr(vdev, &mac_info_old);
1336         }
1337
1338         if (unlikely(!is_vxge_card_up(vdev))) {
1339                 memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
1340                 return VXGE_HW_OK;
1341         }
1342
1343         /* Set this mac address to all the vpaths */
1344         for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++) {
1345                 mac_info_new.vpath_no = vpath_idx;
1346                 mac_info_new.state = VXGE_LL_MAC_ADDR_IN_DA_TABLE;
1347                 status = vxge_add_mac_addr(vdev, &mac_info_new);
1348                 if (status != VXGE_HW_OK)
1349                         return -EINVAL;
1350         }
1351
1352         memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
1353
1354         return status;
1355 }
1356
1357 /*
1358  * vxge_vpath_intr_enable
1359  * @vdev: pointer to vdev
1360  * @vp_id: vpath for which to enable the interrupts
1361  *
1362  * Enables the interrupts for the vpath
1363 */
1364 void vxge_vpath_intr_enable(struct vxgedev *vdev, int vp_id)
1365 {
1366         struct vxge_vpath *vpath = &vdev->vpaths[vp_id];
1367         int msix_id = 0;
1368         int tim_msix_id[4] = {0, 1, 0, 0};
1369         int alarm_msix_id = VXGE_ALARM_MSIX_ID;
1370
1371         vxge_hw_vpath_intr_enable(vpath->handle);
1372
1373         if (vdev->config.intr_type == INTA)
1374                 vxge_hw_vpath_inta_unmask_tx_rx(vpath->handle);
1375         else {
1376                 vxge_hw_vpath_msix_set(vpath->handle, tim_msix_id,
1377                         alarm_msix_id);
1378
1379                 msix_id = vpath->device_id * VXGE_HW_VPATH_MSIX_ACTIVE;
1380                 vxge_hw_vpath_msix_unmask(vpath->handle, msix_id);
1381                 vxge_hw_vpath_msix_unmask(vpath->handle, msix_id + 1);
1382
1383                 /* enable the alarm vector */
1384                 msix_id = (vpath->handle->vpath->hldev->first_vp_id *
1385                         VXGE_HW_VPATH_MSIX_ACTIVE) + alarm_msix_id;
1386                 vxge_hw_vpath_msix_unmask(vpath->handle, msix_id);
1387         }
1388 }
1389
1390 /*
1391  * vxge_vpath_intr_disable
1392  * @vdev: pointer to vdev
1393  * @vp_id: vpath for which to disable the interrupts
1394  *
1395  * Disables the interrupts for the vpath
1396 */
1397 void vxge_vpath_intr_disable(struct vxgedev *vdev, int vp_id)
1398 {
1399         struct vxge_vpath *vpath = &vdev->vpaths[vp_id];
1400         int msix_id;
1401
1402         vxge_hw_vpath_intr_disable(vpath->handle);
1403
1404         if (vdev->config.intr_type == INTA)
1405                 vxge_hw_vpath_inta_mask_tx_rx(vpath->handle);
1406         else {
1407                 msix_id = vpath->device_id * VXGE_HW_VPATH_MSIX_ACTIVE;
1408                 vxge_hw_vpath_msix_mask(vpath->handle, msix_id);
1409                 vxge_hw_vpath_msix_mask(vpath->handle, msix_id + 1);
1410
1411                 /* disable the alarm vector */
1412                 msix_id = (vpath->handle->vpath->hldev->first_vp_id *
1413                         VXGE_HW_VPATH_MSIX_ACTIVE) + VXGE_ALARM_MSIX_ID;
1414                 vxge_hw_vpath_msix_mask(vpath->handle, msix_id);
1415         }
1416 }
1417
1418 /*
1419  * vxge_reset_vpath
1420  * @vdev: pointer to vdev
1421  * @vp_id: vpath to reset
1422  *
1423  * Resets the vpath
1424 */
1425 static int vxge_reset_vpath(struct vxgedev *vdev, int vp_id)
1426 {
1427         enum vxge_hw_status status = VXGE_HW_OK;
1428         int ret = 0;
1429
1430         /* check if device is down already */
1431         if (unlikely(!is_vxge_card_up(vdev)))
1432                 return 0;
1433
1434         /* is device reset already scheduled */
1435         if (test_bit(__VXGE_STATE_RESET_CARD, &vdev->state))
1436                 return 0;
1437
1438         if (vdev->vpaths[vp_id].handle) {
1439                 if (vxge_hw_vpath_reset(vdev->vpaths[vp_id].handle)
1440                                 == VXGE_HW_OK) {
1441                         if (is_vxge_card_up(vdev) &&
1442                                 vxge_hw_vpath_recover_from_reset(
1443                                         vdev->vpaths[vp_id].handle)
1444                                         != VXGE_HW_OK) {
1445                                 vxge_debug_init(VXGE_ERR,
1446                                         "vxge_hw_vpath_recover_from_reset"
1447                                         "failed for vpath:%d", vp_id);
1448                                 return status;
1449                         }
1450                 } else {
1451                         vxge_debug_init(VXGE_ERR,
1452                                 "vxge_hw_vpath_reset failed for"
1453                                 "vpath:%d", vp_id);
1454                                 return status;
1455                 }
1456         } else
1457                 return VXGE_HW_FAIL;
1458
1459         vxge_restore_vpath_mac_addr(&vdev->vpaths[vp_id]);
1460         vxge_restore_vpath_vid_table(&vdev->vpaths[vp_id]);
1461
1462         /* Enable all broadcast */
1463         vxge_hw_vpath_bcast_enable(vdev->vpaths[vp_id].handle);
1464
1465         /* Enable the interrupts */
1466         vxge_vpath_intr_enable(vdev, vp_id);
1467
1468         smp_wmb();
1469
1470         /* Enable the flow of traffic through the vpath */
1471         vxge_hw_vpath_enable(vdev->vpaths[vp_id].handle);
1472
1473         smp_wmb();
1474         vxge_hw_vpath_rx_doorbell_init(vdev->vpaths[vp_id].handle);
1475         vdev->vpaths[vp_id].ring.last_status = VXGE_HW_OK;
1476
1477         /* Vpath reset done */
1478         clear_bit(vp_id, &vdev->vp_reset);
1479
1480         /* Start the vpath queue */
1481         vxge_wake_tx_queue(&vdev->vpaths[vp_id].fifo, NULL);
1482
1483         return ret;
1484 }
1485
1486 static int do_vxge_reset(struct vxgedev *vdev, int event)
1487 {
1488         enum vxge_hw_status status;
1489         int ret = 0, vp_id, i;
1490
1491         vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1492
1493         if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_START_RESET)) {
1494                 /* check if device is down already */
1495                 if (unlikely(!is_vxge_card_up(vdev)))
1496                         return 0;
1497
1498                 /* is reset already scheduled */
1499                 if (test_and_set_bit(__VXGE_STATE_RESET_CARD, &vdev->state))
1500                         return 0;
1501         }
1502
1503         if (event == VXGE_LL_FULL_RESET) {
1504                 /* wait for all the vpath reset to complete */
1505                 for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
1506                         while (test_bit(vp_id, &vdev->vp_reset))
1507                                 msleep(50);
1508                 }
1509
1510                 /* if execution mode is set to debug, don't reset the adapter */
1511                 if (unlikely(vdev->exec_mode)) {
1512                         vxge_debug_init(VXGE_ERR,
1513                                 "%s: execution mode is debug, returning..",
1514                                 vdev->ndev->name);
1515                 clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
1516                 vxge_stop_all_tx_queue(vdev);
1517                 return 0;
1518                 }
1519         }
1520
1521         if (event == VXGE_LL_FULL_RESET) {
1522                 vxge_hw_device_intr_disable(vdev->devh);
1523
1524                 switch (vdev->cric_err_event) {
1525                 case VXGE_HW_EVENT_UNKNOWN:
1526                         vxge_stop_all_tx_queue(vdev);
1527                         vxge_debug_init(VXGE_ERR,
1528                                 "fatal: %s: Disabling device due to"
1529                                 "unknown error",
1530                                 vdev->ndev->name);
1531                         ret = -EPERM;
1532                         goto out;
1533                 case VXGE_HW_EVENT_RESET_START:
1534                         break;
1535                 case VXGE_HW_EVENT_RESET_COMPLETE:
1536                 case VXGE_HW_EVENT_LINK_DOWN:
1537                 case VXGE_HW_EVENT_LINK_UP:
1538                 case VXGE_HW_EVENT_ALARM_CLEARED:
1539                 case VXGE_HW_EVENT_ECCERR:
1540                 case VXGE_HW_EVENT_MRPCIM_ECCERR:
1541                         ret = -EPERM;
1542                         goto out;
1543                 case VXGE_HW_EVENT_FIFO_ERR:
1544                 case VXGE_HW_EVENT_VPATH_ERR:
1545                         break;
1546                 case VXGE_HW_EVENT_CRITICAL_ERR:
1547                         vxge_stop_all_tx_queue(vdev);
1548                         vxge_debug_init(VXGE_ERR,
1549                                 "fatal: %s: Disabling device due to"
1550                                 "serious error",
1551                                 vdev->ndev->name);
1552                         /* SOP or device reset required */
1553                         /* This event is not currently used */
1554                         ret = -EPERM;
1555                         goto out;
1556                 case VXGE_HW_EVENT_SERR:
1557                         vxge_stop_all_tx_queue(vdev);
1558                         vxge_debug_init(VXGE_ERR,
1559                                 "fatal: %s: Disabling device due to"
1560                                 "serious error",
1561                                 vdev->ndev->name);
1562                         ret = -EPERM;
1563                         goto out;
1564                 case VXGE_HW_EVENT_SRPCIM_SERR:
1565                 case VXGE_HW_EVENT_MRPCIM_SERR:
1566                         ret = -EPERM;
1567                         goto out;
1568                 case VXGE_HW_EVENT_SLOT_FREEZE:
1569                         vxge_stop_all_tx_queue(vdev);
1570                         vxge_debug_init(VXGE_ERR,
1571                                 "fatal: %s: Disabling device due to"
1572                                 "slot freeze",
1573                                 vdev->ndev->name);
1574                         ret = -EPERM;
1575                         goto out;
1576                 default:
1577                         break;
1578
1579                 }
1580         }
1581
1582         if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_START_RESET))
1583                 vxge_stop_all_tx_queue(vdev);
1584
1585         if (event == VXGE_LL_FULL_RESET) {
1586                 status = vxge_reset_all_vpaths(vdev);
1587                 if (status != VXGE_HW_OK) {
1588                         vxge_debug_init(VXGE_ERR,
1589                                 "fatal: %s: can not reset vpaths",
1590                                 vdev->ndev->name);
1591                         ret = -EPERM;
1592                         goto out;
1593                 }
1594         }
1595
1596         if (event == VXGE_LL_COMPL_RESET) {
1597                 for (i = 0; i < vdev->no_of_vpath; i++)
1598                         if (vdev->vpaths[i].handle) {
1599                                 if (vxge_hw_vpath_recover_from_reset(
1600                                         vdev->vpaths[i].handle)
1601                                                 != VXGE_HW_OK) {
1602                                         vxge_debug_init(VXGE_ERR,
1603                                                 "vxge_hw_vpath_recover_"
1604                                                 "from_reset failed for vpath: "
1605                                                 "%d", i);
1606                                         ret = -EPERM;
1607                                         goto out;
1608                                 }
1609                                 } else {
1610                                         vxge_debug_init(VXGE_ERR,
1611                                         "vxge_hw_vpath_reset failed for "
1612                                                 "vpath:%d", i);
1613                                         ret = -EPERM;
1614                                         goto out;
1615                                 }
1616         }
1617
1618         if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_COMPL_RESET)) {
1619                 /* Reprogram the DA table with populated mac addresses */
1620                 for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
1621                         vxge_restore_vpath_mac_addr(&vdev->vpaths[vp_id]);
1622                         vxge_restore_vpath_vid_table(&vdev->vpaths[vp_id]);
1623                 }
1624
1625                 /* enable vpath interrupts */
1626                 for (i = 0; i < vdev->no_of_vpath; i++)
1627                         vxge_vpath_intr_enable(vdev, i);
1628
1629                 vxge_hw_device_intr_enable(vdev->devh);
1630
1631                 smp_wmb();
1632
1633                 /* Indicate card up */
1634                 set_bit(__VXGE_STATE_CARD_UP, &vdev->state);
1635
1636                 /* Get the traffic to flow through the vpaths */
1637                 for (i = 0; i < vdev->no_of_vpath; i++) {
1638                         vxge_hw_vpath_enable(vdev->vpaths[i].handle);
1639                         smp_wmb();
1640                         vxge_hw_vpath_rx_doorbell_init(vdev->vpaths[i].handle);
1641                 }
1642
1643                 vxge_wake_all_tx_queue(vdev);
1644         }
1645
1646 out:
1647         vxge_debug_entryexit(VXGE_TRACE,
1648                 "%s:%d  Exiting...", __func__, __LINE__);
1649
1650         /* Indicate reset done */
1651         if ((event == VXGE_LL_FULL_RESET) || (event == VXGE_LL_COMPL_RESET))
1652                 clear_bit(__VXGE_STATE_RESET_CARD, &vdev->state);
1653         return ret;
1654 }
1655
1656 /*
1657  * vxge_reset
1658  * @vdev: pointer to ll device
1659  *
1660  * driver may reset the chip on events of serr, eccerr, etc
1661  */
1662 int vxge_reset(struct vxgedev *vdev)
1663 {
1664         do_vxge_reset(vdev, VXGE_LL_FULL_RESET);
1665         return 0;
1666 }
1667
1668 /**
1669  * vxge_poll - Receive handler when Receive Polling is used.
1670  * @dev: pointer to the device structure.
1671  * @budget: Number of packets budgeted to be processed in this iteration.
1672  *
1673  * This function comes into picture only if Receive side is being handled
1674  * through polling (called NAPI in linux). It mostly does what the normal
1675  * Rx interrupt handler does in terms of descriptor and packet processing
1676  * but not in an interrupt context. Also it will process a specified number
1677  * of packets at most in one iteration. This value is passed down by the
1678  * kernel as the function argument 'budget'.
1679  */
1680 static int vxge_poll_msix(struct napi_struct *napi, int budget)
1681 {
1682         struct vxge_ring *ring =
1683                 container_of(napi, struct vxge_ring, napi);
1684         int budget_org = budget;
1685         ring->budget = budget;
1686
1687         vxge_hw_vpath_poll_rx(ring->handle);
1688
1689         if (ring->pkts_processed < budget_org) {
1690                 napi_complete(napi);
1691                 /* Re enable the Rx interrupts for the vpath */
1692                 vxge_hw_channel_msix_unmask(
1693                                 (struct __vxge_hw_channel *)ring->handle,
1694                                 ring->rx_vector_no);
1695         }
1696
1697         return ring->pkts_processed;
1698 }
1699
1700 static int vxge_poll_inta(struct napi_struct *napi, int budget)
1701 {
1702         struct vxgedev *vdev = container_of(napi, struct vxgedev, napi);
1703         int pkts_processed = 0;
1704         int i;
1705         int budget_org = budget;
1706         struct vxge_ring *ring;
1707
1708         struct __vxge_hw_device  *hldev = (struct __vxge_hw_device *)
1709                 pci_get_drvdata(vdev->pdev);
1710
1711         for (i = 0; i < vdev->no_of_vpath; i++) {
1712                 ring = &vdev->vpaths[i].ring;
1713                 ring->budget = budget;
1714                 vxge_hw_vpath_poll_rx(ring->handle);
1715                 pkts_processed += ring->pkts_processed;
1716                 budget -= ring->pkts_processed;
1717                 if (budget <= 0)
1718                         break;
1719         }
1720
1721         VXGE_COMPLETE_ALL_TX(vdev);
1722
1723         if (pkts_processed < budget_org) {
1724                 napi_complete(napi);
1725                 /* Re enable the Rx interrupts for the ring */
1726                 vxge_hw_device_unmask_all(hldev);
1727                 vxge_hw_device_flush_io(hldev);
1728         }
1729
1730         return pkts_processed;
1731 }
1732
1733 #ifdef CONFIG_NET_POLL_CONTROLLER
1734 /**
1735  * vxge_netpoll - netpoll event handler entry point
1736  * @dev : pointer to the device structure.
1737  * Description:
1738  *      This function will be called by upper layer to check for events on the
1739  * interface in situations where interrupts are disabled. It is used for
1740  * specific in-kernel networking tasks, such as remote consoles and kernel
1741  * debugging over the network (example netdump in RedHat).
1742  */
1743 static void vxge_netpoll(struct net_device *dev)
1744 {
1745         struct __vxge_hw_device  *hldev;
1746         struct vxgedev *vdev;
1747
1748         vdev = (struct vxgedev *)netdev_priv(dev);
1749         hldev = (struct __vxge_hw_device  *)pci_get_drvdata(vdev->pdev);
1750
1751         vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
1752
1753         if (pci_channel_offline(vdev->pdev))
1754                 return;
1755
1756         disable_irq(dev->irq);
1757         vxge_hw_device_clear_tx_rx(hldev);
1758
1759         vxge_hw_device_clear_tx_rx(hldev);
1760         VXGE_COMPLETE_ALL_RX(vdev);
1761         VXGE_COMPLETE_ALL_TX(vdev);
1762
1763         enable_irq(dev->irq);
1764
1765         vxge_debug_entryexit(VXGE_TRACE,
1766                 "%s:%d  Exiting...", __func__, __LINE__);
1767 }
1768 #endif
1769
1770 /* RTH configuration */
1771 static enum vxge_hw_status vxge_rth_configure(struct vxgedev *vdev)
1772 {
1773         enum vxge_hw_status status = VXGE_HW_OK;
1774         struct vxge_hw_rth_hash_types hash_types;
1775         u8 itable[256] = {0}; /* indirection table */
1776         u8 mtable[256] = {0}; /* CPU to vpath mapping  */
1777         int index;
1778
1779         /*
1780          * Filling
1781          *      - itable with bucket numbers
1782          *      - mtable with bucket-to-vpath mapping
1783          */
1784         for (index = 0; index < (1 << vdev->config.rth_bkt_sz); index++) {
1785                 itable[index] = index;
1786                 mtable[index] = index % vdev->no_of_vpath;
1787         }
1788
1789         /* Fill RTH hash types */
1790         hash_types.hash_type_tcpipv4_en   = vdev->config.rth_hash_type_tcpipv4;
1791         hash_types.hash_type_ipv4_en      = vdev->config.rth_hash_type_ipv4;
1792         hash_types.hash_type_tcpipv6_en   = vdev->config.rth_hash_type_tcpipv6;
1793         hash_types.hash_type_ipv6_en      = vdev->config.rth_hash_type_ipv6;
1794         hash_types.hash_type_tcpipv6ex_en =
1795                                         vdev->config.rth_hash_type_tcpipv6ex;
1796         hash_types.hash_type_ipv6ex_en    = vdev->config.rth_hash_type_ipv6ex;
1797
1798         /* set indirection table, bucket-to-vpath mapping */
1799         status = vxge_hw_vpath_rts_rth_itable_set(vdev->vp_handles,
1800                                                 vdev->no_of_vpath,
1801                                                 mtable, itable,
1802                                                 vdev->config.rth_bkt_sz);
1803         if (status != VXGE_HW_OK) {
1804                 vxge_debug_init(VXGE_ERR,
1805                         "RTH indirection table configuration failed "
1806                         "for vpath:%d", vdev->vpaths[0].device_id);
1807                 return status;
1808         }
1809
1810         /*
1811         * Because the itable_set() method uses the active_table field
1812         * for the target virtual path the RTH config should be updated
1813         * for all VPATHs. The h/w only uses the lowest numbered VPATH
1814         * when steering frames.
1815         */
1816          for (index = 0; index < vdev->no_of_vpath; index++) {
1817                 status = vxge_hw_vpath_rts_rth_set(
1818                                 vdev->vpaths[index].handle,
1819                                 vdev->config.rth_algorithm,
1820                                 &hash_types,
1821                                 vdev->config.rth_bkt_sz);
1822
1823                  if (status != VXGE_HW_OK) {
1824                         vxge_debug_init(VXGE_ERR,
1825                                 "RTH configuration failed for vpath:%d",
1826                                 vdev->vpaths[index].device_id);
1827                         return status;
1828                  }
1829          }
1830
1831         return status;
1832 }
1833
1834 int vxge_mac_list_add(struct vxge_vpath *vpath, struct macInfo *mac)
1835 {
1836         struct vxge_mac_addrs *new_mac_entry;
1837         u8 *mac_address = NULL;
1838
1839         if (vpath->mac_addr_cnt >= VXGE_MAX_LEARN_MAC_ADDR_CNT)
1840                 return TRUE;
1841
1842         new_mac_entry = kzalloc(sizeof(struct vxge_mac_addrs), GFP_ATOMIC);
1843         if (!new_mac_entry) {
1844                 vxge_debug_mem(VXGE_ERR,
1845                         "%s: memory allocation failed",
1846                         VXGE_DRIVER_NAME);
1847                 return FALSE;
1848         }
1849
1850         list_add(&new_mac_entry->item, &vpath->mac_addr_list);
1851
1852         /* Copy the new mac address to the list */
1853         mac_address = (u8 *)&new_mac_entry->macaddr;
1854         memcpy(mac_address, mac->macaddr, ETH_ALEN);
1855
1856         new_mac_entry->state = mac->state;
1857         vpath->mac_addr_cnt++;
1858
1859         /* Is this a multicast address */
1860         if (0x01 & mac->macaddr[0])
1861                 vpath->mcast_addr_cnt++;
1862
1863         return TRUE;
1864 }
1865
1866 /* Add a mac address to DA table */
1867 enum vxge_hw_status vxge_add_mac_addr(struct vxgedev *vdev, struct macInfo *mac)
1868 {
1869         enum vxge_hw_status status = VXGE_HW_OK;
1870         struct vxge_vpath *vpath;
1871         enum vxge_hw_vpath_mac_addr_add_mode duplicate_mode;
1872
1873         if (0x01 & mac->macaddr[0]) /* multicast address */
1874                 duplicate_mode = VXGE_HW_VPATH_MAC_ADDR_ADD_DUPLICATE;
1875         else
1876                 duplicate_mode = VXGE_HW_VPATH_MAC_ADDR_REPLACE_DUPLICATE;
1877
1878         vpath = &vdev->vpaths[mac->vpath_no];
1879         status = vxge_hw_vpath_mac_addr_add(vpath->handle, mac->macaddr,
1880                                                 mac->macmask, duplicate_mode);
1881         if (status != VXGE_HW_OK) {
1882                 vxge_debug_init(VXGE_ERR,
1883                         "DA config add entry failed for vpath:%d",
1884                         vpath->device_id);
1885         } else
1886                 if (FALSE == vxge_mac_list_add(vpath, mac))
1887                         status = -EPERM;
1888
1889         return status;
1890 }
1891
1892 int vxge_mac_list_del(struct vxge_vpath *vpath, struct macInfo *mac)
1893 {
1894         struct list_head *entry, *next;
1895         u64 del_mac = 0;
1896         u8 *mac_address = (u8 *) (&del_mac);
1897
1898         /* Copy the mac address to delete from the list */
1899         memcpy(mac_address, mac->macaddr, ETH_ALEN);
1900
1901         list_for_each_safe(entry, next, &vpath->mac_addr_list) {
1902                 if (((struct vxge_mac_addrs *)entry)->macaddr == del_mac) {
1903                         list_del(entry);
1904                         kfree((struct vxge_mac_addrs *)entry);
1905                         vpath->mac_addr_cnt--;
1906
1907                         /* Is this a multicast address */
1908                         if (0x01 & mac->macaddr[0])
1909                                 vpath->mcast_addr_cnt--;
1910                         return TRUE;
1911                 }
1912         }
1913
1914         return FALSE;
1915 }
1916 /* delete a mac address from DA table */
1917 enum vxge_hw_status vxge_del_mac_addr(struct vxgedev *vdev, struct macInfo *mac)
1918 {
1919         enum vxge_hw_status status = VXGE_HW_OK;
1920         struct vxge_vpath *vpath;
1921
1922         vpath = &vdev->vpaths[mac->vpath_no];
1923         status = vxge_hw_vpath_mac_addr_delete(vpath->handle, mac->macaddr,
1924                                                 mac->macmask);
1925         if (status != VXGE_HW_OK) {
1926                 vxge_debug_init(VXGE_ERR,
1927                         "DA config delete entry failed for vpath:%d",
1928                         vpath->device_id);
1929         } else
1930                 vxge_mac_list_del(vpath, mac);
1931         return status;
1932 }
1933
1934 /* list all mac addresses from DA table */
1935 enum vxge_hw_status
1936 static vxge_search_mac_addr_in_da_table(struct vxge_vpath *vpath,
1937                                         struct macInfo *mac)
1938 {
1939         enum vxge_hw_status status = VXGE_HW_OK;
1940         unsigned char macmask[ETH_ALEN];
1941         unsigned char macaddr[ETH_ALEN];
1942
1943         status = vxge_hw_vpath_mac_addr_get(vpath->handle,
1944                                 macaddr, macmask);
1945         if (status != VXGE_HW_OK) {
1946                 vxge_debug_init(VXGE_ERR,
1947                         "DA config list entry failed for vpath:%d",
1948                         vpath->device_id);
1949                 return status;
1950         }
1951
1952         while (memcmp(mac->macaddr, macaddr, ETH_ALEN)) {
1953
1954                 status = vxge_hw_vpath_mac_addr_get_next(vpath->handle,
1955                                 macaddr, macmask);
1956                 if (status != VXGE_HW_OK)
1957                         break;
1958         }
1959
1960         return status;
1961 }
1962
1963 /* Store all vlan ids from the list to the vid table */
1964 enum vxge_hw_status vxge_restore_vpath_vid_table(struct vxge_vpath *vpath)
1965 {
1966         enum vxge_hw_status status = VXGE_HW_OK;
1967         struct vxgedev *vdev = vpath->vdev;
1968         u16 vid;
1969
1970         if (vdev->vlgrp && vpath->is_open) {
1971
1972                 for (vid = 0; vid < VLAN_GROUP_ARRAY_LEN; vid++) {
1973                         if (!vlan_group_get_device(vdev->vlgrp, vid))
1974                                 continue;
1975                         /* Add these vlan to the vid table */
1976                         status = vxge_hw_vpath_vid_add(vpath->handle, vid);
1977                 }
1978         }
1979
1980         return status;
1981 }
1982
1983 /* Store all mac addresses from the list to the DA table */
1984 enum vxge_hw_status vxge_restore_vpath_mac_addr(struct vxge_vpath *vpath)
1985 {
1986         enum vxge_hw_status status = VXGE_HW_OK;
1987         struct macInfo mac_info;
1988         u8 *mac_address = NULL;
1989         struct list_head *entry, *next;
1990
1991         memset(&mac_info, 0, sizeof(struct macInfo));
1992
1993         if (vpath->is_open) {
1994
1995                 list_for_each_safe(entry, next, &vpath->mac_addr_list) {
1996                         mac_address =
1997                                 (u8 *)&
1998                                 ((struct vxge_mac_addrs *)entry)->macaddr;
1999                         memcpy(mac_info.macaddr, mac_address, ETH_ALEN);
2000                         ((struct vxge_mac_addrs *)entry)->state =
2001                                 VXGE_LL_MAC_ADDR_IN_DA_TABLE;
2002                         /* does this mac address already exist in da table? */
2003                         status = vxge_search_mac_addr_in_da_table(vpath,
2004                                 &mac_info);
2005                         if (status != VXGE_HW_OK) {
2006                                 /* Add this mac address to the DA table */
2007                                 status = vxge_hw_vpath_mac_addr_add(
2008                                         vpath->handle, mac_info.macaddr,
2009                                         mac_info.macmask,
2010                                     VXGE_HW_VPATH_MAC_ADDR_ADD_DUPLICATE);
2011                                 if (status != VXGE_HW_OK) {
2012                                         vxge_debug_init(VXGE_ERR,
2013                                             "DA add entry failed for vpath:%d",
2014                                             vpath->device_id);
2015                                         ((struct vxge_mac_addrs *)entry)->state
2016                                                 = VXGE_LL_MAC_ADDR_IN_LIST;
2017                                 }
2018                         }
2019                 }
2020         }
2021
2022         return status;
2023 }
2024
2025 /* reset vpaths */
2026 enum vxge_hw_status vxge_reset_all_vpaths(struct vxgedev *vdev)
2027 {
2028         int i;
2029         enum vxge_hw_status status = VXGE_HW_OK;
2030
2031         for (i = 0; i < vdev->no_of_vpath; i++)
2032                 if (vdev->vpaths[i].handle) {
2033                         if (vxge_hw_vpath_reset(vdev->vpaths[i].handle)
2034                                         == VXGE_HW_OK) {
2035                                 if (is_vxge_card_up(vdev) &&
2036                                         vxge_hw_vpath_recover_from_reset(
2037                                                 vdev->vpaths[i].handle)
2038                                                 != VXGE_HW_OK) {
2039                                         vxge_debug_init(VXGE_ERR,
2040                                                 "vxge_hw_vpath_recover_"
2041                                                 "from_reset failed for vpath: "
2042                                                 "%d", i);
2043                                         return status;
2044                                 }
2045                         } else {
2046                                 vxge_debug_init(VXGE_ERR,
2047                                         "vxge_hw_vpath_reset failed for "
2048                                         "vpath:%d", i);
2049                                         return status;
2050                         }
2051                 }
2052         return status;
2053 }
2054
2055 /* close vpaths */
2056 void vxge_close_vpaths(struct vxgedev *vdev, int index)
2057 {
2058         int i;
2059         for (i = index; i < vdev->no_of_vpath; i++) {
2060                 if (vdev->vpaths[i].handle && vdev->vpaths[i].is_open) {
2061                         vxge_hw_vpath_close(vdev->vpaths[i].handle);
2062                         vdev->stats.vpaths_open--;
2063                 }
2064                 vdev->vpaths[i].is_open = 0;
2065                 vdev->vpaths[i].handle  = NULL;
2066         }
2067 }
2068
2069 /* open vpaths */
2070 int vxge_open_vpaths(struct vxgedev *vdev)
2071 {
2072         enum vxge_hw_status status;
2073         int i;
2074         u32 vp_id = 0;
2075         struct vxge_hw_vpath_attr attr;
2076
2077         for (i = 0; i < vdev->no_of_vpath; i++) {
2078                 vxge_assert(vdev->vpaths[i].is_configured);
2079                 attr.vp_id = vdev->vpaths[i].device_id;
2080                 attr.fifo_attr.callback = vxge_xmit_compl;
2081                 attr.fifo_attr.txdl_term = vxge_tx_term;
2082                 attr.fifo_attr.per_txdl_space = sizeof(struct vxge_tx_priv);
2083                 attr.fifo_attr.userdata = (void *)&vdev->vpaths[i].fifo;
2084
2085                 attr.ring_attr.callback = vxge_rx_1b_compl;
2086                 attr.ring_attr.rxd_init = vxge_rx_initial_replenish;
2087                 attr.ring_attr.rxd_term = vxge_rx_term;
2088                 attr.ring_attr.per_rxd_space = sizeof(struct vxge_rx_priv);
2089                 attr.ring_attr.userdata = (void *)&vdev->vpaths[i].ring;
2090
2091                 vdev->vpaths[i].ring.ndev = vdev->ndev;
2092                 vdev->vpaths[i].ring.pdev = vdev->pdev;
2093                 status = vxge_hw_vpath_open(vdev->devh, &attr,
2094                                 &(vdev->vpaths[i].handle));
2095                 if (status == VXGE_HW_OK) {
2096                         vdev->vpaths[i].fifo.handle =
2097                             (struct __vxge_hw_fifo *)attr.fifo_attr.userdata;
2098                         vdev->vpaths[i].ring.handle =
2099                             (struct __vxge_hw_ring *)attr.ring_attr.userdata;
2100                         vdev->vpaths[i].fifo.tx_steering_type =
2101                                 vdev->config.tx_steering_type;
2102                         vdev->vpaths[i].fifo.ndev = vdev->ndev;
2103                         vdev->vpaths[i].fifo.pdev = vdev->pdev;
2104                         vdev->vpaths[i].fifo.indicate_max_pkts =
2105                                 vdev->config.fifo_indicate_max_pkts;
2106                         vdev->vpaths[i].ring.rx_vector_no = 0;
2107                         vdev->vpaths[i].ring.rx_csum = vdev->rx_csum;
2108                         vdev->vpaths[i].is_open = 1;
2109                         vdev->vp_handles[i] = vdev->vpaths[i].handle;
2110                         vdev->vpaths[i].ring.gro_enable =
2111                                                 vdev->config.gro_enable;
2112                         vdev->vpaths[i].ring.vlan_tag_strip =
2113                                                 vdev->vlan_tag_strip;
2114                         vdev->stats.vpaths_open++;
2115                 } else {
2116                         vdev->stats.vpath_open_fail++;
2117                         vxge_debug_init(VXGE_ERR,
2118                                 "%s: vpath: %d failed to open "
2119                                 "with status: %d",
2120                             vdev->ndev->name, vdev->vpaths[i].device_id,
2121                                 status);
2122                         vxge_close_vpaths(vdev, 0);
2123                         return -EPERM;
2124                 }
2125
2126                 vp_id =
2127                   ((struct __vxge_hw_vpath_handle *)vdev->vpaths[i].handle)->
2128                   vpath->vp_id;
2129                 vdev->vpaths_deployed |= vxge_mBIT(vp_id);
2130         }
2131         return VXGE_HW_OK;
2132 }
2133
2134 /*
2135  *  vxge_isr_napi
2136  *  @irq: the irq of the device.
2137  *  @dev_id: a void pointer to the hldev structure of the Titan device
2138  *  @ptregs: pointer to the registers pushed on the stack.
2139  *
2140  *  This function is the ISR handler of the device when napi is enabled. It
2141  *  identifies the reason for the interrupt and calls the relevant service
2142  *  routines.
2143  */
2144 static irqreturn_t vxge_isr_napi(int irq, void *dev_id)
2145 {
2146         struct net_device *dev;
2147         struct __vxge_hw_device *hldev;
2148         u64 reason;
2149         enum vxge_hw_status status;
2150         struct vxgedev *vdev = (struct vxgedev *) dev_id;;
2151
2152         vxge_debug_intr(VXGE_TRACE, "%s:%d", __func__, __LINE__);
2153
2154         dev = vdev->ndev;
2155         hldev = (struct __vxge_hw_device *)pci_get_drvdata(vdev->pdev);
2156
2157         if (pci_channel_offline(vdev->pdev))
2158                 return IRQ_NONE;
2159
2160         if (unlikely(!is_vxge_card_up(vdev)))
2161                 return IRQ_NONE;
2162
2163         status = vxge_hw_device_begin_irq(hldev, vdev->exec_mode,
2164                         &reason);
2165         if (status == VXGE_HW_OK) {
2166                 vxge_hw_device_mask_all(hldev);
2167
2168                 if (reason &
2169                         VXGE_HW_TITAN_GENERAL_INT_STATUS_VPATH_TRAFFIC_INT(
2170                         vdev->vpaths_deployed >>
2171                         (64 - VXGE_HW_MAX_VIRTUAL_PATHS))) {
2172
2173                         vxge_hw_device_clear_tx_rx(hldev);
2174                         napi_schedule(&vdev->napi);
2175                         vxge_debug_intr(VXGE_TRACE,
2176                                 "%s:%d  Exiting...", __func__, __LINE__);
2177                         return IRQ_HANDLED;
2178                 } else
2179                         vxge_hw_device_unmask_all(hldev);
2180         } else if (unlikely((status == VXGE_HW_ERR_VPATH) ||
2181                 (status == VXGE_HW_ERR_CRITICAL) ||
2182                 (status == VXGE_HW_ERR_FIFO))) {
2183                 vxge_hw_device_mask_all(hldev);
2184                 vxge_hw_device_flush_io(hldev);
2185                 return IRQ_HANDLED;
2186         } else if (unlikely(status == VXGE_HW_ERR_SLOT_FREEZE))
2187                 return IRQ_HANDLED;
2188
2189         vxge_debug_intr(VXGE_TRACE, "%s:%d  Exiting...", __func__, __LINE__);
2190         return IRQ_NONE;
2191 }
2192
2193 #ifdef CONFIG_PCI_MSI
2194
2195 static irqreturn_t
2196 vxge_tx_msix_handle(int irq, void *dev_id)
2197 {
2198         struct vxge_fifo *fifo = (struct vxge_fifo *)dev_id;
2199
2200         VXGE_COMPLETE_VPATH_TX(fifo);
2201
2202         return IRQ_HANDLED;
2203 }
2204
2205 static irqreturn_t
2206 vxge_rx_msix_napi_handle(int irq, void *dev_id)
2207 {
2208         struct vxge_ring *ring = (struct vxge_ring *)dev_id;
2209
2210         /* MSIX_IDX for Rx is 1 */
2211         vxge_hw_channel_msix_mask((struct __vxge_hw_channel *)ring->handle,
2212                                         ring->rx_vector_no);
2213
2214         napi_schedule(&ring->napi);
2215         return IRQ_HANDLED;
2216 }
2217
2218 static irqreturn_t
2219 vxge_alarm_msix_handle(int irq, void *dev_id)
2220 {
2221         int i;
2222         enum vxge_hw_status status;
2223         struct vxge_vpath *vpath = (struct vxge_vpath *)dev_id;
2224         struct vxgedev *vdev = vpath->vdev;
2225         int msix_id = (vpath->handle->vpath->vp_id *
2226                 VXGE_HW_VPATH_MSIX_ACTIVE) + VXGE_ALARM_MSIX_ID;
2227
2228         for (i = 0; i < vdev->no_of_vpath; i++) {
2229                 vxge_hw_vpath_msix_mask(vdev->vpaths[i].handle, msix_id);
2230
2231                 status = vxge_hw_vpath_alarm_process(vdev->vpaths[i].handle,
2232                         vdev->exec_mode);
2233                 if (status == VXGE_HW_OK) {
2234
2235                         vxge_hw_vpath_msix_unmask(vdev->vpaths[i].handle,
2236                                         msix_id);
2237                         continue;
2238                 }
2239                 vxge_debug_intr(VXGE_ERR,
2240                         "%s: vxge_hw_vpath_alarm_process failed %x ",
2241                         VXGE_DRIVER_NAME, status);
2242         }
2243         return IRQ_HANDLED;
2244 }
2245
2246 static int vxge_alloc_msix(struct vxgedev *vdev)
2247 {
2248         int j, i, ret = 0;
2249         int msix_intr_vect = 0, temp;
2250         vdev->intr_cnt = 0;
2251
2252 start:
2253         /* Tx/Rx MSIX Vectors count */
2254         vdev->intr_cnt = vdev->no_of_vpath * 2;
2255
2256         /* Alarm MSIX Vectors count */
2257         vdev->intr_cnt++;
2258
2259         vdev->entries = kzalloc(vdev->intr_cnt * sizeof(struct msix_entry),
2260                                                 GFP_KERNEL);
2261         if (!vdev->entries) {
2262                 vxge_debug_init(VXGE_ERR,
2263                         "%s: memory allocation failed",
2264                         VXGE_DRIVER_NAME);
2265                 ret = -ENOMEM;
2266                 goto alloc_entries_failed;
2267         }
2268
2269         vdev->vxge_entries =
2270                 kzalloc(vdev->intr_cnt * sizeof(struct vxge_msix_entry),
2271                                 GFP_KERNEL);
2272         if (!vdev->vxge_entries) {
2273                 vxge_debug_init(VXGE_ERR, "%s: memory allocation failed",
2274                         VXGE_DRIVER_NAME);
2275                 ret = -ENOMEM;
2276                 goto alloc_vxge_entries_failed;
2277         }
2278
2279         for (i = 0, j = 0; i < vdev->no_of_vpath; i++) {
2280
2281                 msix_intr_vect = i * VXGE_HW_VPATH_MSIX_ACTIVE;
2282
2283                 /* Initialize the fifo vector */
2284                 vdev->entries[j].entry = msix_intr_vect;
2285                 vdev->vxge_entries[j].entry = msix_intr_vect;
2286                 vdev->vxge_entries[j].in_use = 0;
2287                 j++;
2288
2289                 /* Initialize the ring vector */
2290                 vdev->entries[j].entry = msix_intr_vect + 1;
2291                 vdev->vxge_entries[j].entry = msix_intr_vect + 1;
2292                 vdev->vxge_entries[j].in_use = 0;
2293                 j++;
2294         }
2295
2296         /* Initialize the alarm vector */
2297         vdev->entries[j].entry = VXGE_ALARM_MSIX_ID;
2298         vdev->vxge_entries[j].entry = VXGE_ALARM_MSIX_ID;
2299         vdev->vxge_entries[j].in_use = 0;
2300
2301         ret = pci_enable_msix(vdev->pdev, vdev->entries, vdev->intr_cnt);
2302
2303         if (ret > 0) {
2304                 vxge_debug_init(VXGE_ERR,
2305                         "%s: MSI-X enable failed for %d vectors, ret: %d",
2306                         VXGE_DRIVER_NAME, vdev->intr_cnt, ret);
2307                 if ((max_config_vpath != VXGE_USE_DEFAULT) || (ret < 3)) {
2308                         ret = -ENODEV;
2309                         goto enable_msix_failed;
2310                 }
2311
2312                 kfree(vdev->entries);
2313                 kfree(vdev->vxge_entries);
2314                 vdev->entries = NULL;
2315                 vdev->vxge_entries = NULL;
2316                 /* Try with less no of vector by reducing no of vpaths count */
2317                 temp = (ret - 1)/2;
2318                 vxge_close_vpaths(vdev, temp);
2319                 vdev->no_of_vpath = temp;
2320                 goto start;
2321         } else if (ret < 0) {
2322                 ret = -ENODEV;
2323                 goto enable_msix_failed;
2324         }
2325         return 0;
2326
2327 enable_msix_failed:
2328         kfree(vdev->vxge_entries);
2329 alloc_vxge_entries_failed:
2330         kfree(vdev->entries);
2331 alloc_entries_failed:
2332         return ret;
2333 }
2334
2335 static int vxge_enable_msix(struct vxgedev *vdev)
2336 {
2337
2338         int i, ret = 0;
2339         /* 0 - Tx, 1 - Rx  */
2340         int tim_msix_id[4] = {0, 1, 0, 0};
2341
2342         vdev->intr_cnt = 0;
2343
2344         /* allocate msix vectors */
2345         ret = vxge_alloc_msix(vdev);
2346         if (!ret) {
2347                 for (i = 0; i < vdev->no_of_vpath; i++) {
2348
2349                         /* If fifo or ring are not enabled
2350                            the MSIX vector for that should be set to 0
2351                            Hence initializeing this array to all 0s.
2352                         */
2353                         vdev->vpaths[i].ring.rx_vector_no =
2354                                 (vdev->vpaths[i].device_id *
2355                                         VXGE_HW_VPATH_MSIX_ACTIVE) + 1;
2356
2357                         vxge_hw_vpath_msix_set(vdev->vpaths[i].handle,
2358                                         tim_msix_id, VXGE_ALARM_MSIX_ID);
2359                 }
2360         }
2361
2362         return ret;
2363 }
2364
2365 static void vxge_rem_msix_isr(struct vxgedev *vdev)
2366 {
2367         int intr_cnt;
2368
2369         for (intr_cnt = 0; intr_cnt < (vdev->no_of_vpath * 2 + 1);
2370                 intr_cnt++) {
2371                 if (vdev->vxge_entries[intr_cnt].in_use) {
2372                         synchronize_irq(vdev->entries[intr_cnt].vector);
2373                         free_irq(vdev->entries[intr_cnt].vector,
2374                                 vdev->vxge_entries[intr_cnt].arg);
2375                         vdev->vxge_entries[intr_cnt].in_use = 0;
2376                 }
2377         }
2378
2379         kfree(vdev->entries);
2380         kfree(vdev->vxge_entries);
2381         vdev->entries = NULL;
2382         vdev->vxge_entries = NULL;
2383
2384         if (vdev->config.intr_type == MSI_X)
2385                 pci_disable_msix(vdev->pdev);
2386 }
2387 #endif
2388
2389 static void vxge_rem_isr(struct vxgedev *vdev)
2390 {
2391         struct __vxge_hw_device  *hldev;
2392         hldev = (struct __vxge_hw_device  *) pci_get_drvdata(vdev->pdev);
2393
2394 #ifdef CONFIG_PCI_MSI
2395         if (vdev->config.intr_type == MSI_X) {
2396                 vxge_rem_msix_isr(vdev);
2397         } else
2398 #endif
2399         if (vdev->config.intr_type == INTA) {
2400                         synchronize_irq(vdev->pdev->irq);
2401                         free_irq(vdev->pdev->irq, vdev);
2402         }
2403 }
2404
2405 static int vxge_add_isr(struct vxgedev *vdev)
2406 {
2407         int ret = 0;
2408 #ifdef CONFIG_PCI_MSI
2409         int vp_idx = 0, intr_idx = 0, intr_cnt = 0, msix_idx = 0, irq_req = 0;
2410         int pci_fun = PCI_FUNC(vdev->pdev->devfn);
2411
2412         if (vdev->config.intr_type == MSI_X)
2413                 ret = vxge_enable_msix(vdev);
2414
2415         if (ret) {
2416                 vxge_debug_init(VXGE_ERR,
2417                         "%s: Enabling MSI-X Failed", VXGE_DRIVER_NAME);
2418                 vxge_debug_init(VXGE_ERR,
2419                         "%s: Defaulting to INTA", VXGE_DRIVER_NAME);
2420                 vdev->config.intr_type = INTA;
2421         }
2422
2423         if (vdev->config.intr_type == MSI_X) {
2424                 for (intr_idx = 0;
2425                      intr_idx < (vdev->no_of_vpath *
2426                         VXGE_HW_VPATH_MSIX_ACTIVE); intr_idx++) {
2427
2428                         msix_idx = intr_idx % VXGE_HW_VPATH_MSIX_ACTIVE;
2429                         irq_req = 0;
2430
2431                         switch (msix_idx) {
2432                         case 0:
2433                                 snprintf(vdev->desc[intr_cnt], VXGE_INTR_STRLEN,
2434                                 "%s:vxge:MSI-X %d - Tx - fn:%d vpath:%d",
2435                                         vdev->ndev->name,
2436                                         vdev->entries[intr_cnt].entry,
2437                                         pci_fun, vp_idx);
2438                                 ret = request_irq(
2439                                     vdev->entries[intr_cnt].vector,
2440                                         vxge_tx_msix_handle, 0,
2441                                         vdev->desc[intr_cnt],
2442                                         &vdev->vpaths[vp_idx].fifo);
2443                                         vdev->vxge_entries[intr_cnt].arg =
2444                                                 &vdev->vpaths[vp_idx].fifo;
2445                                 irq_req = 1;
2446                                 break;
2447                         case 1:
2448                                 snprintf(vdev->desc[intr_cnt], VXGE_INTR_STRLEN,
2449                                 "%s:vxge:MSI-X %d - Rx - fn:%d vpath:%d",
2450                                         vdev->ndev->name,
2451                                         vdev->entries[intr_cnt].entry,
2452                                         pci_fun, vp_idx);
2453                                 ret = request_irq(
2454                                     vdev->entries[intr_cnt].vector,
2455                                         vxge_rx_msix_napi_handle,
2456                                         0,
2457                                         vdev->desc[intr_cnt],
2458                                         &vdev->vpaths[vp_idx].ring);
2459                                         vdev->vxge_entries[intr_cnt].arg =
2460                                                 &vdev->vpaths[vp_idx].ring;
2461                                 irq_req = 1;
2462                                 break;
2463                         }
2464
2465                         if (ret) {
2466                                 vxge_debug_init(VXGE_ERR,
2467                                         "%s: MSIX - %d  Registration failed",
2468                                         vdev->ndev->name, intr_cnt);
2469                                 vxge_rem_msix_isr(vdev);
2470                                 vdev->config.intr_type = INTA;
2471                                 vxge_debug_init(VXGE_ERR,
2472                                         "%s: Defaulting to INTA"
2473                                         , vdev->ndev->name);
2474                                         goto INTA_MODE;
2475                         }
2476
2477                         if (irq_req) {
2478                                 /* We requested for this msix interrupt */
2479                                 vdev->vxge_entries[intr_cnt].in_use = 1;
2480                                 msix_idx +=  vdev->vpaths[vp_idx].device_id *
2481                                         VXGE_HW_VPATH_MSIX_ACTIVE;
2482                                 vxge_hw_vpath_msix_unmask(
2483                                         vdev->vpaths[vp_idx].handle,
2484                                         msix_idx);
2485                                 intr_cnt++;
2486                         }
2487
2488                         /* Point to next vpath handler */
2489                         if (((intr_idx + 1) % VXGE_HW_VPATH_MSIX_ACTIVE == 0) &&
2490                             (vp_idx < (vdev->no_of_vpath - 1)))
2491                                 vp_idx++;
2492                 }
2493
2494                 intr_cnt = vdev->no_of_vpath * 2;
2495                 snprintf(vdev->desc[intr_cnt], VXGE_INTR_STRLEN,
2496                         "%s:vxge:MSI-X %d - Alarm - fn:%d",
2497                         vdev->ndev->name,
2498                         vdev->entries[intr_cnt].entry,
2499                         pci_fun);
2500                 /* For Alarm interrupts */
2501                 ret = request_irq(vdev->entries[intr_cnt].vector,
2502                                         vxge_alarm_msix_handle, 0,
2503                                         vdev->desc[intr_cnt],
2504                                         &vdev->vpaths[0]);
2505                 if (ret) {
2506                         vxge_debug_init(VXGE_ERR,
2507                                 "%s: MSIX - %d Registration failed",
2508                                 vdev->ndev->name, intr_cnt);
2509                         vxge_rem_msix_isr(vdev);
2510                         vdev->config.intr_type = INTA;
2511                         vxge_debug_init(VXGE_ERR,
2512                                 "%s: Defaulting to INTA",
2513                                 vdev->ndev->name);
2514                                 goto INTA_MODE;
2515                 }
2516
2517                 msix_idx = (vdev->vpaths[0].handle->vpath->vp_id *
2518                         VXGE_HW_VPATH_MSIX_ACTIVE) + VXGE_ALARM_MSIX_ID;
2519                 vxge_hw_vpath_msix_unmask(vdev->vpaths[vp_idx].handle,
2520                                         msix_idx);
2521                 vdev->vxge_entries[intr_cnt].in_use = 1;
2522                 vdev->vxge_entries[intr_cnt].arg = &vdev->vpaths[0];
2523         }
2524 INTA_MODE:
2525 #endif
2526
2527         if (vdev->config.intr_type == INTA) {
2528                 snprintf(vdev->desc[0], VXGE_INTR_STRLEN,
2529                         "%s:vxge:INTA", vdev->ndev->name);
2530                 vxge_hw_device_set_intr_type(vdev->devh,
2531                         VXGE_HW_INTR_MODE_IRQLINE);
2532                 vxge_hw_vpath_tti_ci_set(vdev->devh,
2533                         vdev->vpaths[0].device_id);
2534                 ret = request_irq((int) vdev->pdev->irq,
2535                         vxge_isr_napi,
2536                         IRQF_SHARED, vdev->desc[0], vdev);
2537                 if (ret) {
2538                         vxge_debug_init(VXGE_ERR,
2539                                 "%s %s-%d: ISR registration failed",
2540                                 VXGE_DRIVER_NAME, "IRQ", vdev->pdev->irq);
2541                         return -ENODEV;
2542                 }
2543                 vxge_debug_init(VXGE_TRACE,
2544                         "new %s-%d line allocated",
2545                         "IRQ", vdev->pdev->irq);
2546         }
2547
2548         return VXGE_HW_OK;
2549 }
2550
2551 static void vxge_poll_vp_reset(unsigned long data)
2552 {
2553         struct vxgedev *vdev = (struct vxgedev *)data;
2554         int i, j = 0;
2555
2556         for (i = 0; i < vdev->no_of_vpath; i++) {
2557                 if (test_bit(i, &vdev->vp_reset)) {
2558                         vxge_reset_vpath(vdev, i);
2559                         j++;
2560                 }
2561         }
2562         if (j && (vdev->config.intr_type != MSI_X)) {
2563                 vxge_hw_device_unmask_all(vdev->devh);
2564                 vxge_hw_device_flush_io(vdev->devh);
2565         }
2566
2567         mod_timer(&vdev->vp_reset_timer, jiffies + HZ / 2);
2568 }
2569
2570 static void vxge_poll_vp_lockup(unsigned long data)
2571 {
2572         struct vxgedev *vdev = (struct vxgedev *)data;
2573         int i;
2574         struct vxge_ring *ring;
2575         enum vxge_hw_status status = VXGE_HW_OK;
2576
2577         for (i = 0; i < vdev->no_of_vpath; i++) {
2578                 ring = &vdev->vpaths[i].ring;
2579                 /* Did this vpath received any packets */
2580                 if (ring->stats.prev_rx_frms == ring->stats.rx_frms) {
2581                         status = vxge_hw_vpath_check_leak(ring->handle);
2582
2583                         /* Did it received any packets last time */
2584                         if ((VXGE_HW_FAIL == status) &&
2585                                 (VXGE_HW_FAIL == ring->last_status)) {
2586
2587                                 /* schedule vpath reset */
2588                                 if (!test_and_set_bit(i, &vdev->vp_reset)) {
2589
2590                                         /* disable interrupts for this vpath */
2591                                         vxge_vpath_intr_disable(vdev, i);
2592
2593                                         /* stop the queue for this vpath */
2594                                         vxge_stop_tx_queue(&vdev->vpaths[i].
2595                                                                 fifo);
2596                                         continue;
2597                                 }
2598                         }
2599                 }
2600                 ring->stats.prev_rx_frms = ring->stats.rx_frms;
2601                 ring->last_status = status;
2602         }
2603
2604         /* Check every 1 milli second */
2605         mod_timer(&vdev->vp_lockup_timer, jiffies + HZ / 1000);
2606 }
2607
2608 /**
2609  * vxge_open
2610  * @dev: pointer to the device structure.
2611  *
2612  * This function is the open entry point of the driver. It mainly calls a
2613  * function to allocate Rx buffers and inserts them into the buffer
2614  * descriptors and then enables the Rx part of the NIC.
2615  * Return value: '0' on success and an appropriate (-)ve integer as
2616  * defined in errno.h file on failure.
2617  */
2618 int
2619 vxge_open(struct net_device *dev)
2620 {
2621         enum vxge_hw_status status;
2622         struct vxgedev *vdev;
2623         struct __vxge_hw_device *hldev;
2624         int ret = 0;
2625         int i;
2626         u64 val64, function_mode;
2627         vxge_debug_entryexit(VXGE_TRACE,
2628                 "%s: %s:%d", dev->name, __func__, __LINE__);
2629
2630         vdev = (struct vxgedev *)netdev_priv(dev);
2631         hldev = (struct __vxge_hw_device *) pci_get_drvdata(vdev->pdev);
2632         function_mode = vdev->config.device_hw_info.function_mode;
2633
2634         /* make sure you have link off by default every time Nic is
2635          * initialized */
2636         netif_carrier_off(dev);
2637
2638         /* Open VPATHs */
2639         status = vxge_open_vpaths(vdev);
2640         if (status != VXGE_HW_OK) {
2641                 vxge_debug_init(VXGE_ERR,
2642                         "%s: fatal: Vpath open failed", vdev->ndev->name);
2643                 ret = -EPERM;
2644                 goto out0;
2645         }
2646
2647         vdev->mtu = dev->mtu;
2648
2649         status = vxge_add_isr(vdev);
2650         if (status != VXGE_HW_OK) {
2651                 vxge_debug_init(VXGE_ERR,
2652                         "%s: fatal: ISR add failed", dev->name);
2653                 ret = -EPERM;
2654                 goto out1;
2655         }
2656
2657
2658         if (vdev->config.intr_type != MSI_X) {
2659                 netif_napi_add(dev, &vdev->napi, vxge_poll_inta,
2660                         vdev->config.napi_weight);
2661                 napi_enable(&vdev->napi);
2662                 for (i = 0; i < vdev->no_of_vpath; i++)
2663                         vdev->vpaths[i].ring.napi_p = &vdev->napi;
2664         } else {
2665                 for (i = 0; i < vdev->no_of_vpath; i++) {
2666                         netif_napi_add(dev, &vdev->vpaths[i].ring.napi,
2667                             vxge_poll_msix, vdev->config.napi_weight);
2668                         napi_enable(&vdev->vpaths[i].ring.napi);
2669                         vdev->vpaths[i].ring.napi_p =
2670                                 &vdev->vpaths[i].ring.napi;
2671                 }
2672         }
2673
2674         /* configure RTH */
2675         if (vdev->config.rth_steering) {
2676                 status = vxge_rth_configure(vdev);
2677                 if (status != VXGE_HW_OK) {
2678                         vxge_debug_init(VXGE_ERR,
2679                                 "%s: fatal: RTH configuration failed",
2680                                 dev->name);
2681                         ret = -EPERM;
2682                         goto out2;
2683                 }
2684         }
2685
2686         for (i = 0; i < vdev->no_of_vpath; i++) {
2687                 /* set initial mtu before enabling the device */
2688                 status = vxge_hw_vpath_mtu_set(vdev->vpaths[i].handle,
2689                                                 vdev->mtu);
2690                 if (status != VXGE_HW_OK) {
2691                         vxge_debug_init(VXGE_ERR,
2692                                 "%s: fatal: can not set new MTU", dev->name);
2693                         ret = -EPERM;
2694                         goto out2;
2695                 }
2696         }
2697
2698         VXGE_DEVICE_DEBUG_LEVEL_SET(VXGE_TRACE, VXGE_COMPONENT_LL, vdev);
2699         vxge_debug_init(vdev->level_trace,
2700                 "%s: MTU is %d", vdev->ndev->name, vdev->mtu);
2701         VXGE_DEVICE_DEBUG_LEVEL_SET(VXGE_ERR, VXGE_COMPONENT_LL, vdev);
2702
2703         /* Reprogram the DA table with populated mac addresses */
2704         for (i = 0; i < vdev->no_of_vpath; i++) {
2705                 vxge_restore_vpath_mac_addr(&vdev->vpaths[i]);
2706                 vxge_restore_vpath_vid_table(&vdev->vpaths[i]);
2707         }
2708
2709         /* Enable vpath to sniff all unicast/multicast traffic that not
2710          * addressed to them. We allow promiscous mode for PF only
2711          */
2712
2713         val64 = 0;
2714         for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
2715                 val64 |= VXGE_HW_RXMAC_AUTHORIZE_ALL_ADDR_VP(i);
2716
2717         vxge_hw_mgmt_reg_write(vdev->devh,
2718                 vxge_hw_mgmt_reg_type_mrpcim,
2719                 0,
2720                 (ulong)offsetof(struct vxge_hw_mrpcim_reg,
2721                         rxmac_authorize_all_addr),
2722                 val64);
2723
2724         vxge_hw_mgmt_reg_write(vdev->devh,
2725                 vxge_hw_mgmt_reg_type_mrpcim,
2726                 0,
2727                 (ulong)offsetof(struct vxge_hw_mrpcim_reg,
2728                         rxmac_authorize_all_vid),
2729                 val64);
2730
2731         vxge_set_multicast(dev);
2732
2733         /* Enabling Bcast and mcast for all vpath */
2734         for (i = 0; i < vdev->no_of_vpath; i++) {
2735                 status = vxge_hw_vpath_bcast_enable(vdev->vpaths[i].handle);
2736                 if (status != VXGE_HW_OK)
2737                         vxge_debug_init(VXGE_ERR,
2738                                 "%s : Can not enable bcast for vpath "
2739                                 "id %d", dev->name, i);
2740                 if (vdev->config.addr_learn_en) {
2741                         status =
2742                             vxge_hw_vpath_mcast_enable(vdev->vpaths[i].handle);
2743                         if (status != VXGE_HW_OK)
2744                                 vxge_debug_init(VXGE_ERR,
2745                                         "%s : Can not enable mcast for vpath "
2746                                         "id %d", dev->name, i);
2747                 }
2748         }
2749
2750         vxge_hw_device_setpause_data(vdev->devh, 0,
2751                 vdev->config.tx_pause_enable,
2752                 vdev->config.rx_pause_enable);
2753
2754         if (vdev->vp_reset_timer.function == NULL)
2755                 vxge_os_timer(vdev->vp_reset_timer,
2756                         vxge_poll_vp_reset, vdev, (HZ/2));
2757
2758         if (vdev->vp_lockup_timer.function == NULL)
2759                 vxge_os_timer(vdev->vp_lockup_timer,
2760                         vxge_poll_vp_lockup, vdev, (HZ/2));
2761
2762         set_bit(__VXGE_STATE_CARD_UP, &vdev->state);
2763
2764         smp_wmb();
2765
2766         if (vxge_hw_device_link_state_get(vdev->devh) == VXGE_HW_LINK_UP) {
2767                 netif_carrier_on(vdev->ndev);
2768                 printk(KERN_NOTICE "%s: Link Up\n", vdev->ndev->name);
2769                 vdev->stats.link_up++;
2770         }
2771
2772         vxge_hw_device_intr_enable(vdev->devh);
2773
2774         smp_wmb();
2775
2776         for (i = 0; i < vdev->no_of_vpath; i++) {
2777                 vxge_hw_vpath_enable(vdev->vpaths[i].handle);
2778                 smp_wmb();
2779                 vxge_hw_vpath_rx_doorbell_init(vdev->vpaths[i].handle);
2780         }
2781
2782         vxge_start_all_tx_queue(vdev);
2783         goto out0;
2784
2785 out2:
2786         vxge_rem_isr(vdev);
2787
2788         /* Disable napi */
2789         if (vdev->config.intr_type != MSI_X)
2790                 napi_disable(&vdev->napi);
2791         else {
2792                 for (i = 0; i < vdev->no_of_vpath; i++)
2793                         napi_disable(&vdev->vpaths[i].ring.napi);
2794         }
2795
2796 out1:
2797         vxge_close_vpaths(vdev, 0);
2798 out0:
2799         vxge_debug_entryexit(VXGE_TRACE,
2800                                 "%s: %s:%d  Exiting...",
2801                                 dev->name, __func__, __LINE__);
2802         return ret;
2803 }
2804
2805 /* Loop throught the mac address list and delete all the entries */
2806 void vxge_free_mac_add_list(struct vxge_vpath *vpath)
2807 {
2808
2809         struct list_head *entry, *next;
2810         if (list_empty(&vpath->mac_addr_list))
2811                 return;
2812
2813         list_for_each_safe(entry, next, &vpath->mac_addr_list) {
2814                 list_del(entry);
2815                 kfree((struct vxge_mac_addrs *)entry);
2816         }
2817 }
2818
2819 static void vxge_napi_del_all(struct vxgedev *vdev)
2820 {
2821         int i;
2822         if (vdev->config.intr_type != MSI_X)
2823                 netif_napi_del(&vdev->napi);
2824         else {
2825                 for (i = 0; i < vdev->no_of_vpath; i++)
2826                         netif_napi_del(&vdev->vpaths[i].ring.napi);
2827         }
2828 }
2829
2830 int do_vxge_close(struct net_device *dev, int do_io)
2831 {
2832         enum vxge_hw_status status;
2833         struct vxgedev *vdev;
2834         struct __vxge_hw_device *hldev;
2835         int i;
2836         u64 val64, vpath_vector;
2837         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d",
2838                 dev->name, __func__, __LINE__);
2839
2840         vdev = (struct vxgedev *)netdev_priv(dev);
2841         hldev = (struct __vxge_hw_device *) pci_get_drvdata(vdev->pdev);
2842
2843         if (unlikely(!is_vxge_card_up(vdev)))
2844                 return 0;
2845
2846         /* If vxge_handle_crit_err task is executing,
2847          * wait till it completes. */
2848         while (test_and_set_bit(__VXGE_STATE_RESET_CARD, &vdev->state))
2849                 msleep(50);
2850
2851         clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
2852         if (do_io) {
2853                 /* Put the vpath back in normal mode */
2854                 vpath_vector = vxge_mBIT(vdev->vpaths[0].device_id);
2855                 status = vxge_hw_mgmt_reg_read(vdev->devh,
2856                                 vxge_hw_mgmt_reg_type_mrpcim,
2857                                 0,
2858                                 (ulong)offsetof(
2859                                         struct vxge_hw_mrpcim_reg,
2860                                         rts_mgr_cbasin_cfg),
2861                                 &val64);
2862
2863                 if (status == VXGE_HW_OK) {
2864                         val64 &= ~vpath_vector;
2865                         status = vxge_hw_mgmt_reg_write(vdev->devh,
2866                                         vxge_hw_mgmt_reg_type_mrpcim,
2867                                         0,
2868                                         (ulong)offsetof(
2869                                                 struct vxge_hw_mrpcim_reg,
2870                                                 rts_mgr_cbasin_cfg),
2871                                         val64);
2872                 }
2873
2874                 /* Remove the function 0 from promiscous mode */
2875                 vxge_hw_mgmt_reg_write(vdev->devh,
2876                         vxge_hw_mgmt_reg_type_mrpcim,
2877                         0,
2878                         (ulong)offsetof(struct vxge_hw_mrpcim_reg,
2879                                 rxmac_authorize_all_addr),
2880                         0);
2881
2882                 vxge_hw_mgmt_reg_write(vdev->devh,
2883                         vxge_hw_mgmt_reg_type_mrpcim,
2884                         0,
2885                         (ulong)offsetof(struct vxge_hw_mrpcim_reg,
2886                                 rxmac_authorize_all_vid),
2887                         0);
2888
2889                 smp_wmb();
2890         }
2891         del_timer_sync(&vdev->vp_lockup_timer);
2892
2893         del_timer_sync(&vdev->vp_reset_timer);
2894
2895         /* Disable napi */
2896         if (vdev->config.intr_type != MSI_X)
2897                 napi_disable(&vdev->napi);
2898         else {
2899                 for (i = 0; i < vdev->no_of_vpath; i++)
2900                         napi_disable(&vdev->vpaths[i].ring.napi);
2901         }
2902
2903         netif_carrier_off(vdev->ndev);
2904         printk(KERN_NOTICE "%s: Link Down\n", vdev->ndev->name);
2905         vxge_stop_all_tx_queue(vdev);
2906
2907         /* Note that at this point xmit() is stopped by upper layer */
2908         if (do_io)
2909                 vxge_hw_device_intr_disable(vdev->devh);
2910
2911         mdelay(1000);
2912
2913         vxge_rem_isr(vdev);
2914
2915         vxge_napi_del_all(vdev);
2916
2917         if (do_io)
2918                 vxge_reset_all_vpaths(vdev);
2919
2920         vxge_close_vpaths(vdev, 0);
2921
2922         vxge_debug_entryexit(VXGE_TRACE,
2923                 "%s: %s:%d  Exiting...", dev->name, __func__, __LINE__);
2924
2925         clear_bit(__VXGE_STATE_RESET_CARD, &vdev->state);
2926
2927         return 0;
2928 }
2929
2930 /**
2931  * vxge_close
2932  * @dev: device pointer.
2933  *
2934  * This is the stop entry point of the driver. It needs to undo exactly
2935  * whatever was done by the open entry point, thus it's usually referred to
2936  * as the close function.Among other things this function mainly stops the
2937  * Rx side of the NIC and frees all the Rx buffers in the Rx rings.
2938  * Return value: '0' on success and an appropriate (-)ve integer as
2939  * defined in errno.h file on failure.
2940  */
2941 int
2942 vxge_close(struct net_device *dev)
2943 {
2944         do_vxge_close(dev, 1);
2945         return 0;
2946 }
2947
2948 /**
2949  * vxge_change_mtu
2950  * @dev: net device pointer.
2951  * @new_mtu :the new MTU size for the device.
2952  *
2953  * A driver entry point to change MTU size for the device. Before changing
2954  * the MTU the device must be stopped.
2955  */
2956 static int vxge_change_mtu(struct net_device *dev, int new_mtu)
2957 {
2958         struct vxgedev *vdev = netdev_priv(dev);
2959
2960         vxge_debug_entryexit(vdev->level_trace,
2961                 "%s:%d", __func__, __LINE__);
2962         if ((new_mtu < VXGE_HW_MIN_MTU) || (new_mtu > VXGE_HW_MAX_MTU)) {
2963                 vxge_debug_init(vdev->level_err,
2964                         "%s: mtu size is invalid", dev->name);
2965                 return -EPERM;
2966         }
2967
2968         /* check if device is down already */
2969         if (unlikely(!is_vxge_card_up(vdev))) {
2970                 /* just store new value, will use later on open() */
2971                 dev->mtu = new_mtu;
2972                 vxge_debug_init(vdev->level_err,
2973                         "%s", "device is down on MTU change");
2974                 return 0;
2975         }
2976
2977         vxge_debug_init(vdev->level_trace,
2978                 "trying to apply new MTU %d", new_mtu);
2979
2980         if (vxge_close(dev))
2981                 return -EIO;
2982
2983         dev->mtu = new_mtu;
2984         vdev->mtu = new_mtu;
2985
2986         if (vxge_open(dev))
2987                 return -EIO;
2988
2989         vxge_debug_init(vdev->level_trace,
2990                 "%s: MTU changed to %d", vdev->ndev->name, new_mtu);
2991
2992         vxge_debug_entryexit(vdev->level_trace,
2993                 "%s:%d  Exiting...", __func__, __LINE__);
2994
2995         return 0;
2996 }
2997
2998 /**
2999  * vxge_get_stats
3000  * @dev: pointer to the device structure
3001  *
3002  * Updates the device statistics structure. This function updates the device
3003  * statistics structure in the net_device structure and returns a pointer
3004  * to the same.
3005  */
3006 static struct net_device_stats *
3007 vxge_get_stats(struct net_device *dev)
3008 {
3009         struct vxgedev *vdev;
3010         struct net_device_stats *net_stats;
3011         int k;
3012
3013         vdev = netdev_priv(dev);
3014
3015         net_stats = &vdev->stats.net_stats;
3016
3017         memset(net_stats, 0, sizeof(struct net_device_stats));
3018
3019         for (k = 0; k < vdev->no_of_vpath; k++) {
3020                 net_stats->rx_packets += vdev->vpaths[k].ring.stats.rx_frms;
3021                 net_stats->rx_bytes += vdev->vpaths[k].ring.stats.rx_bytes;
3022                 net_stats->rx_errors += vdev->vpaths[k].ring.stats.rx_errors;
3023                 net_stats->multicast += vdev->vpaths[k].ring.stats.rx_mcast;
3024                 net_stats->rx_dropped +=
3025                         vdev->vpaths[k].ring.stats.rx_dropped;
3026
3027                 net_stats->tx_packets += vdev->vpaths[k].fifo.stats.tx_frms;
3028                 net_stats->tx_bytes += vdev->vpaths[k].fifo.stats.tx_bytes;
3029                 net_stats->tx_errors += vdev->vpaths[k].fifo.stats.tx_errors;
3030         }
3031
3032         return net_stats;
3033 }
3034
3035 /**
3036  * vxge_ioctl
3037  * @dev: Device pointer.
3038  * @ifr: An IOCTL specific structure, that can contain a pointer to
3039  *       a proprietary structure used to pass information to the driver.
3040  * @cmd: This is used to distinguish between the different commands that
3041  *       can be passed to the IOCTL functions.
3042  *
3043  * Entry point for the Ioctl.
3044  */
3045 static int vxge_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
3046 {
3047         return -EOPNOTSUPP;
3048 }
3049
3050 /**
3051  * vxge_tx_watchdog
3052  * @dev: pointer to net device structure
3053  *
3054  * Watchdog for transmit side.
3055  * This function is triggered if the Tx Queue is stopped
3056  * for a pre-defined amount of time when the Interface is still up.
3057  */
3058 static void
3059 vxge_tx_watchdog(struct net_device *dev)
3060 {
3061         struct vxgedev *vdev;
3062
3063         vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
3064
3065         vdev = (struct vxgedev *)netdev_priv(dev);
3066
3067         vdev->cric_err_event = VXGE_HW_EVENT_RESET_START;
3068
3069         vxge_reset(vdev);
3070         vxge_debug_entryexit(VXGE_TRACE,
3071                 "%s:%d  Exiting...", __func__, __LINE__);
3072 }
3073
3074 /**
3075  * vxge_vlan_rx_register
3076  * @dev: net device pointer.
3077  * @grp: vlan group
3078  *
3079  * Vlan group registration
3080  */
3081 static void
3082 vxge_vlan_rx_register(struct net_device *dev, struct vlan_group *grp)
3083 {
3084         struct vxgedev *vdev;
3085         struct vxge_vpath *vpath;
3086         int vp;
3087         u64 vid;
3088         enum vxge_hw_status status;
3089         int i;
3090
3091         vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
3092
3093         vdev = (struct vxgedev *)netdev_priv(dev);
3094
3095         vpath = &vdev->vpaths[0];
3096         if ((NULL == grp) && (vpath->is_open)) {
3097                 /* Get the first vlan */
3098                 status = vxge_hw_vpath_vid_get(vpath->handle, &vid);
3099
3100                 while (status == VXGE_HW_OK) {
3101
3102                         /* Delete this vlan from the vid table */
3103                         for (vp = 0; vp < vdev->no_of_vpath; vp++) {
3104                                 vpath = &vdev->vpaths[vp];
3105                                 if (!vpath->is_open)
3106                                         continue;
3107
3108                                 vxge_hw_vpath_vid_delete(vpath->handle, vid);
3109                         }
3110
3111                         /* Get the next vlan to be deleted */
3112                         vpath = &vdev->vpaths[0];
3113                         status = vxge_hw_vpath_vid_get(vpath->handle, &vid);
3114                 }
3115         }
3116
3117         vdev->vlgrp = grp;
3118
3119         for (i = 0; i < vdev->no_of_vpath; i++) {
3120                 if (vdev->vpaths[i].is_configured)
3121                         vdev->vpaths[i].ring.vlgrp = grp;
3122         }
3123
3124         vxge_debug_entryexit(VXGE_TRACE,
3125                 "%s:%d  Exiting...", __func__, __LINE__);
3126 }
3127
3128 /**
3129  * vxge_vlan_rx_add_vid
3130  * @dev: net device pointer.
3131  * @vid: vid
3132  *
3133  * Add the vlan id to the devices vlan id table
3134  */
3135 static void
3136 vxge_vlan_rx_add_vid(struct net_device *dev, unsigned short vid)
3137 {
3138         struct vxgedev *vdev;
3139         struct vxge_vpath *vpath;
3140         int vp_id;
3141
3142         vdev = (struct vxgedev *)netdev_priv(dev);
3143
3144         /* Add these vlan to the vid table */
3145         for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
3146                 vpath = &vdev->vpaths[vp_id];
3147                 if (!vpath->is_open)
3148                         continue;
3149                 vxge_hw_vpath_vid_add(vpath->handle, vid);
3150         }
3151 }
3152
3153 /**
3154  * vxge_vlan_rx_add_vid
3155  * @dev: net device pointer.
3156  * @vid: vid
3157  *
3158  * Remove the vlan id from the device's vlan id table
3159  */
3160 static void
3161 vxge_vlan_rx_kill_vid(struct net_device *dev, unsigned short vid)
3162 {
3163         struct vxgedev *vdev;
3164         struct vxge_vpath *vpath;
3165         int vp_id;
3166
3167         vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
3168
3169         vdev = (struct vxgedev *)netdev_priv(dev);
3170
3171         vlan_group_set_device(vdev->vlgrp, vid, NULL);
3172
3173         /* Delete this vlan from the vid table */
3174         for (vp_id = 0; vp_id < vdev->no_of_vpath; vp_id++) {
3175                 vpath = &vdev->vpaths[vp_id];
3176                 if (!vpath->is_open)
3177                         continue;
3178                 vxge_hw_vpath_vid_delete(vpath->handle, vid);
3179         }
3180         vxge_debug_entryexit(VXGE_TRACE,
3181                 "%s:%d  Exiting...", __func__, __LINE__);
3182 }
3183
3184 static const struct net_device_ops vxge_netdev_ops = {
3185         .ndo_open               = vxge_open,
3186         .ndo_stop               = vxge_close,
3187         .ndo_get_stats          = vxge_get_stats,
3188         .ndo_start_xmit         = vxge_xmit,
3189         .ndo_validate_addr      = eth_validate_addr,
3190         .ndo_set_multicast_list = vxge_set_multicast,
3191
3192         .ndo_do_ioctl           = vxge_ioctl,
3193
3194         .ndo_set_mac_address    = vxge_set_mac_addr,
3195         .ndo_change_mtu         = vxge_change_mtu,
3196         .ndo_vlan_rx_register   = vxge_vlan_rx_register,
3197         .ndo_vlan_rx_kill_vid   = vxge_vlan_rx_kill_vid,
3198         .ndo_vlan_rx_add_vid    = vxge_vlan_rx_add_vid,
3199
3200         .ndo_tx_timeout         = vxge_tx_watchdog,
3201 #ifdef CONFIG_NET_POLL_CONTROLLER
3202         .ndo_poll_controller    = vxge_netpoll,
3203 #endif
3204 };
3205
3206 int __devinit vxge_device_register(struct __vxge_hw_device *hldev,
3207                                    struct vxge_config *config,
3208                                    int high_dma, int no_of_vpath,
3209                                    struct vxgedev **vdev_out)
3210 {
3211         struct net_device *ndev;
3212         enum vxge_hw_status status = VXGE_HW_OK;
3213         struct vxgedev *vdev;
3214         int i, ret = 0, no_of_queue = 1;
3215         u64 stat;
3216
3217         *vdev_out = NULL;
3218         if (config->tx_steering_type == TX_MULTIQ_STEERING)
3219                 no_of_queue = no_of_vpath;
3220
3221         ndev = alloc_etherdev_mq(sizeof(struct vxgedev),
3222                         no_of_queue);
3223         if (ndev == NULL) {
3224                 vxge_debug_init(
3225                         vxge_hw_device_trace_level_get(hldev),
3226                 "%s : device allocation failed", __func__);
3227                 ret = -ENODEV;
3228                 goto _out0;
3229         }
3230
3231         vxge_debug_entryexit(
3232                 vxge_hw_device_trace_level_get(hldev),
3233                 "%s: %s:%d  Entering...",
3234                 ndev->name, __func__, __LINE__);
3235
3236         vdev = netdev_priv(ndev);
3237         memset(vdev, 0, sizeof(struct vxgedev));
3238
3239         vdev->ndev = ndev;
3240         vdev->devh = hldev;
3241         vdev->pdev = hldev->pdev;
3242         memcpy(&vdev->config, config, sizeof(struct vxge_config));
3243         vdev->rx_csum = 1;      /* Enable Rx CSUM by default. */
3244
3245         SET_NETDEV_DEV(ndev, &vdev->pdev->dev);
3246
3247         ndev->features |= NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX |
3248                                 NETIF_F_HW_VLAN_FILTER;
3249         /*  Driver entry points */
3250         ndev->irq = vdev->pdev->irq;
3251         ndev->base_addr = (unsigned long) hldev->bar0;
3252
3253         ndev->netdev_ops = &vxge_netdev_ops;
3254
3255         ndev->watchdog_timeo = VXGE_LL_WATCH_DOG_TIMEOUT;
3256
3257         initialize_ethtool_ops(ndev);
3258
3259         /* Allocate memory for vpath */
3260         vdev->vpaths = kzalloc((sizeof(struct vxge_vpath)) *
3261                                 no_of_vpath, GFP_KERNEL);
3262         if (!vdev->vpaths) {
3263                 vxge_debug_init(VXGE_ERR,
3264                         "%s: vpath memory allocation failed",
3265                         vdev->ndev->name);
3266                 ret = -ENODEV;
3267                 goto _out1;
3268         }
3269
3270         ndev->features |= NETIF_F_SG;
3271
3272         ndev->features |= NETIF_F_HW_CSUM;
3273         vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3274                 "%s : checksuming enabled", __func__);
3275
3276         if (high_dma) {
3277                 ndev->features |= NETIF_F_HIGHDMA;
3278                 vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3279                         "%s : using High DMA", __func__);
3280         }
3281
3282         ndev->features |= NETIF_F_TSO | NETIF_F_TSO6;
3283
3284         if (vdev->config.gro_enable)
3285                 ndev->features |= NETIF_F_GRO;
3286
3287         if (vdev->config.tx_steering_type == TX_MULTIQ_STEERING)
3288                 ndev->real_num_tx_queues = no_of_vpath;
3289
3290 #ifdef NETIF_F_LLTX
3291         ndev->features |= NETIF_F_LLTX;
3292 #endif
3293
3294         for (i = 0; i < no_of_vpath; i++)
3295                 spin_lock_init(&vdev->vpaths[i].fifo.tx_lock);
3296
3297         if (register_netdev(ndev)) {
3298                 vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3299                         "%s: %s : device registration failed!",
3300                         ndev->name, __func__);
3301                 ret = -ENODEV;
3302                 goto _out2;
3303         }
3304
3305         /*  Set the factory defined MAC address initially */
3306         ndev->addr_len = ETH_ALEN;
3307
3308         /* Make Link state as off at this point, when the Link change
3309          * interrupt comes the state will be automatically changed to
3310          * the right state.
3311          */
3312         netif_carrier_off(ndev);
3313
3314         vxge_debug_init(vxge_hw_device_trace_level_get(hldev),
3315                 "%s: Ethernet device registered",
3316                 ndev->name);
3317
3318         *vdev_out = vdev;
3319
3320         /* Resetting the Device stats */
3321         status = vxge_hw_mrpcim_stats_access(
3322                                 hldev,
3323                                 VXGE_HW_STATS_OP_CLEAR_ALL_STATS,
3324                                 0,
3325                                 0,
3326                                 &stat);
3327
3328         if (status == VXGE_HW_ERR_PRIVILAGED_OPEARATION)
3329                 vxge_debug_init(
3330                         vxge_hw_device_trace_level_get(hldev),
3331                         "%s: device stats clear returns"
3332                         "VXGE_HW_ERR_PRIVILAGED_OPEARATION", ndev->name);
3333
3334         vxge_debug_entryexit(vxge_hw_device_trace_level_get(hldev),
3335                 "%s: %s:%d  Exiting...",
3336                 ndev->name, __func__, __LINE__);
3337
3338         return ret;
3339 _out2:
3340         kfree(vdev->vpaths);
3341 _out1:
3342         free_netdev(ndev);
3343 _out0:
3344         return ret;
3345 }
3346
3347 /*
3348  * vxge_device_unregister
3349  *
3350  * This function will unregister and free network device
3351  */
3352 void
3353 vxge_device_unregister(struct __vxge_hw_device *hldev)
3354 {
3355         struct vxgedev *vdev;
3356         struct net_device *dev;
3357         char buf[IFNAMSIZ];
3358 #if ((VXGE_DEBUG_INIT & VXGE_DEBUG_MASK) || \
3359         (VXGE_DEBUG_ENTRYEXIT & VXGE_DEBUG_MASK))
3360         u32 level_trace;
3361 #endif
3362
3363         dev = hldev->ndev;
3364         vdev = netdev_priv(dev);
3365 #if ((VXGE_DEBUG_INIT & VXGE_DEBUG_MASK) || \
3366         (VXGE_DEBUG_ENTRYEXIT & VXGE_DEBUG_MASK))
3367         level_trace = vdev->level_trace;
3368 #endif
3369         vxge_debug_entryexit(level_trace,
3370                 "%s: %s:%d", vdev->ndev->name, __func__, __LINE__);
3371
3372         memcpy(buf, vdev->ndev->name, IFNAMSIZ);
3373
3374         /* in 2.6 will call stop() if device is up */
3375         unregister_netdev(dev);
3376
3377         flush_scheduled_work();
3378
3379         vxge_debug_init(level_trace, "%s: ethernet device unregistered", buf);
3380         vxge_debug_entryexit(level_trace,
3381                 "%s: %s:%d  Exiting...", buf, __func__, __LINE__);
3382 }
3383
3384 /*
3385  * vxge_callback_crit_err
3386  *
3387  * This function is called by the alarm handler in interrupt context.
3388  * Driver must analyze it based on the event type.
3389  */
3390 static void
3391 vxge_callback_crit_err(struct __vxge_hw_device *hldev,
3392                         enum vxge_hw_event type, u64 vp_id)
3393 {
3394         struct net_device *dev = hldev->ndev;
3395         struct vxgedev *vdev = (struct vxgedev *)netdev_priv(dev);
3396         int vpath_idx;
3397
3398         vxge_debug_entryexit(vdev->level_trace,
3399                 "%s: %s:%d", vdev->ndev->name, __func__, __LINE__);
3400
3401         /* Note: This event type should be used for device wide
3402          * indications only - Serious errors, Slot freeze and critical errors
3403          */
3404         vdev->cric_err_event = type;
3405
3406         for (vpath_idx = 0; vpath_idx < vdev->no_of_vpath; vpath_idx++)
3407                 if (vdev->vpaths[vpath_idx].device_id == vp_id)
3408                         break;
3409
3410         if (!test_bit(__VXGE_STATE_RESET_CARD, &vdev->state)) {
3411                 if (type == VXGE_HW_EVENT_SLOT_FREEZE) {
3412                         vxge_debug_init(VXGE_ERR,
3413                                 "%s: Slot is frozen", vdev->ndev->name);
3414                 } else if (type == VXGE_HW_EVENT_SERR) {
3415                         vxge_debug_init(VXGE_ERR,
3416                                 "%s: Encountered Serious Error",
3417                                 vdev->ndev->name);
3418                 } else if (type == VXGE_HW_EVENT_CRITICAL_ERR)
3419                         vxge_debug_init(VXGE_ERR,
3420                                 "%s: Encountered Critical Error",
3421                                 vdev->ndev->name);
3422         }
3423
3424         if ((type == VXGE_HW_EVENT_SERR) ||
3425                 (type == VXGE_HW_EVENT_SLOT_FREEZE)) {
3426                 if (unlikely(vdev->exec_mode))
3427                         clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3428         } else if (type == VXGE_HW_EVENT_CRITICAL_ERR) {
3429                 vxge_hw_device_mask_all(hldev);
3430                 if (unlikely(vdev->exec_mode))
3431                         clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3432         } else if ((type == VXGE_HW_EVENT_FIFO_ERR) ||
3433                   (type == VXGE_HW_EVENT_VPATH_ERR)) {
3434
3435                 if (unlikely(vdev->exec_mode))
3436                         clear_bit(__VXGE_STATE_CARD_UP, &vdev->state);
3437                 else {
3438                         /* check if this vpath is already set for reset */
3439                         if (!test_and_set_bit(vpath_idx, &vdev->vp_reset)) {
3440
3441                                 /* disable interrupts for this vpath */
3442                                 vxge_vpath_intr_disable(vdev, vpath_idx);
3443
3444                                 /* stop the queue for this vpath */
3445                                 vxge_stop_tx_queue(&vdev->vpaths[vpath_idx].
3446                                                         fifo);
3447                         }
3448                 }
3449         }
3450
3451         vxge_debug_entryexit(vdev->level_trace,
3452                 "%s: %s:%d  Exiting...",
3453                 vdev->ndev->name, __func__, __LINE__);
3454 }
3455
3456 static void verify_bandwidth(void)
3457 {
3458         int i, band_width, total = 0, equal_priority = 0;
3459
3460         /* 1. If user enters 0 for some fifo, give equal priority to all */
3461         for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3462                 if (bw_percentage[i] == 0) {
3463                         equal_priority = 1;
3464                         break;
3465                 }
3466         }
3467
3468         if (!equal_priority) {
3469                 /* 2. If sum exceeds 100, give equal priority to all */
3470                 for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3471                         if (bw_percentage[i] == 0xFF)
3472                                 break;
3473
3474                         total += bw_percentage[i];
3475                         if (total > VXGE_HW_VPATH_BANDWIDTH_MAX) {
3476                                 equal_priority = 1;
3477                                 break;
3478                         }
3479                 }
3480         }
3481
3482         if (!equal_priority) {
3483                 /* Is all the bandwidth consumed? */
3484                 if (total < VXGE_HW_VPATH_BANDWIDTH_MAX) {
3485                         if (i < VXGE_HW_MAX_VIRTUAL_PATHS) {
3486                                 /* Split rest of bw equally among next VPs*/
3487                                 band_width =
3488                                   (VXGE_HW_VPATH_BANDWIDTH_MAX  - total) /
3489                                         (VXGE_HW_MAX_VIRTUAL_PATHS - i);
3490                                 if (band_width < 2) /* min of 2% */
3491                                         equal_priority = 1;
3492                                 else {
3493                                         for (; i < VXGE_HW_MAX_VIRTUAL_PATHS;
3494                                                 i++)
3495                                                 bw_percentage[i] =
3496                                                         band_width;
3497                                 }
3498                         }
3499                 } else if (i < VXGE_HW_MAX_VIRTUAL_PATHS)
3500                         equal_priority = 1;
3501         }
3502
3503         if (equal_priority) {
3504                 vxge_debug_init(VXGE_ERR,
3505                         "%s: Assigning equal bandwidth to all the vpaths",
3506                         VXGE_DRIVER_NAME);
3507                 bw_percentage[0] = VXGE_HW_VPATH_BANDWIDTH_MAX /
3508                                         VXGE_HW_MAX_VIRTUAL_PATHS;
3509                 for (i = 1; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
3510                         bw_percentage[i] = bw_percentage[0];
3511         }
3512 }
3513
3514 /*
3515  * Vpath configuration
3516  */
3517 static int __devinit vxge_config_vpaths(
3518                         struct vxge_hw_device_config *device_config,
3519                         u64 vpath_mask, struct vxge_config *config_param)
3520 {
3521         int i, no_of_vpaths = 0, default_no_vpath = 0, temp;
3522         u32 txdl_size, txdl_per_memblock;
3523
3524         temp = driver_config->vpath_per_dev;
3525         if ((driver_config->vpath_per_dev == VXGE_USE_DEFAULT) &&
3526                 (max_config_dev == VXGE_MAX_CONFIG_DEV)) {
3527                 /* No more CPU. Return vpath number as zero.*/
3528                 if (driver_config->g_no_cpus == -1)
3529                         return 0;
3530
3531                 if (!driver_config->g_no_cpus)
3532                         driver_config->g_no_cpus = num_online_cpus();
3533
3534                 driver_config->vpath_per_dev = driver_config->g_no_cpus >> 1;
3535                 if (!driver_config->vpath_per_dev)
3536                         driver_config->vpath_per_dev = 1;
3537
3538                 for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
3539                         if (!vxge_bVALn(vpath_mask, i, 1))
3540                                 continue;
3541                         else
3542                                 default_no_vpath++;
3543                 if (default_no_vpath < driver_config->vpath_per_dev)
3544                         driver_config->vpath_per_dev = default_no_vpath;
3545
3546                 driver_config->g_no_cpus = driver_config->g_no_cpus -
3547                                 (driver_config->vpath_per_dev * 2);
3548                 if (driver_config->g_no_cpus <= 0)
3549                         driver_config->g_no_cpus = -1;
3550         }
3551
3552         if (driver_config->vpath_per_dev == 1) {
3553                 vxge_debug_ll_config(VXGE_TRACE,
3554                         "%s: Disable tx and rx steering, "
3555                         "as single vpath is configured", VXGE_DRIVER_NAME);
3556                 config_param->rth_steering = NO_STEERING;
3557                 config_param->tx_steering_type = NO_STEERING;
3558                 device_config->rth_en = 0;
3559         }
3560
3561         /* configure bandwidth */
3562         for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++)
3563                 device_config->vp_config[i].min_bandwidth = bw_percentage[i];
3564
3565         for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3566                 device_config->vp_config[i].vp_id = i;
3567                 device_config->vp_config[i].mtu = VXGE_HW_DEFAULT_MTU;
3568                 if (no_of_vpaths < driver_config->vpath_per_dev) {
3569                         if (!vxge_bVALn(vpath_mask, i, 1)) {
3570                                 vxge_debug_ll_config(VXGE_TRACE,
3571                                         "%s: vpath: %d is not available",
3572                                         VXGE_DRIVER_NAME, i);
3573                                 continue;
3574                         } else {
3575                                 vxge_debug_ll_config(VXGE_TRACE,
3576                                         "%s: vpath: %d available",
3577                                         VXGE_DRIVER_NAME, i);
3578                                 no_of_vpaths++;
3579                         }
3580                 } else {
3581                         vxge_debug_ll_config(VXGE_TRACE,
3582                                 "%s: vpath: %d is not configured, "
3583                                 "max_config_vpath exceeded",
3584                                 VXGE_DRIVER_NAME, i);
3585                         break;
3586                 }
3587
3588                 /* Configure Tx fifo's */
3589                 device_config->vp_config[i].fifo.enable =
3590                                                 VXGE_HW_FIFO_ENABLE;
3591                 device_config->vp_config[i].fifo.max_frags =
3592                                 MAX_SKB_FRAGS + 1;
3593                 device_config->vp_config[i].fifo.memblock_size =
3594                         VXGE_HW_MIN_FIFO_MEMBLOCK_SIZE;
3595
3596                 txdl_size = device_config->vp_config[i].fifo.max_frags *
3597                                 sizeof(struct vxge_hw_fifo_txd);
3598                 txdl_per_memblock = VXGE_HW_MIN_FIFO_MEMBLOCK_SIZE / txdl_size;
3599
3600                 device_config->vp_config[i].fifo.fifo_blocks =
3601                         ((VXGE_DEF_FIFO_LENGTH - 1) / txdl_per_memblock) + 1;
3602
3603                 device_config->vp_config[i].fifo.intr =
3604                                 VXGE_HW_FIFO_QUEUE_INTR_DISABLE;
3605
3606                 /* Configure tti properties */
3607                 device_config->vp_config[i].tti.intr_enable =
3608                                         VXGE_HW_TIM_INTR_ENABLE;
3609
3610                 device_config->vp_config[i].tti.btimer_val =
3611                         (VXGE_TTI_BTIMER_VAL * 1000) / 272;
3612
3613                 device_config->vp_config[i].tti.timer_ac_en =
3614                                 VXGE_HW_TIM_TIMER_AC_ENABLE;
3615
3616                 /* For msi-x with napi (each vector
3617                 has a handler of its own) -
3618                 Set CI to OFF for all vpaths */
3619                 device_config->vp_config[i].tti.timer_ci_en =
3620                         VXGE_HW_TIM_TIMER_CI_DISABLE;
3621
3622                 device_config->vp_config[i].tti.timer_ri_en =
3623                                 VXGE_HW_TIM_TIMER_RI_DISABLE;
3624
3625                 device_config->vp_config[i].tti.util_sel =
3626                         VXGE_HW_TIM_UTIL_SEL_LEGACY_TX_NET_UTIL;
3627
3628                 device_config->vp_config[i].tti.ltimer_val =
3629                         (VXGE_TTI_LTIMER_VAL * 1000) / 272;
3630
3631                 device_config->vp_config[i].tti.rtimer_val =
3632                         (VXGE_TTI_RTIMER_VAL * 1000) / 272;
3633
3634                 device_config->vp_config[i].tti.urange_a = TTI_TX_URANGE_A;
3635                 device_config->vp_config[i].tti.urange_b = TTI_TX_URANGE_B;
3636                 device_config->vp_config[i].tti.urange_c = TTI_TX_URANGE_C;
3637                 device_config->vp_config[i].tti.uec_a = TTI_TX_UFC_A;
3638                 device_config->vp_config[i].tti.uec_b = TTI_TX_UFC_B;
3639                 device_config->vp_config[i].tti.uec_c = TTI_TX_UFC_C;
3640                 device_config->vp_config[i].tti.uec_d = TTI_TX_UFC_D;
3641
3642                 /* Configure Rx rings */
3643                 device_config->vp_config[i].ring.enable  =
3644                                                 VXGE_HW_RING_ENABLE;
3645
3646                 device_config->vp_config[i].ring.ring_blocks  =
3647                                                 VXGE_HW_DEF_RING_BLOCKS;
3648                 device_config->vp_config[i].ring.buffer_mode =
3649                         VXGE_HW_RING_RXD_BUFFER_MODE_1;
3650                 device_config->vp_config[i].ring.rxds_limit  =
3651                                 VXGE_HW_DEF_RING_RXDS_LIMIT;
3652                 device_config->vp_config[i].ring.scatter_mode =
3653                                         VXGE_HW_RING_SCATTER_MODE_A;
3654
3655                 /* Configure rti properties */
3656                 device_config->vp_config[i].rti.intr_enable =
3657                                         VXGE_HW_TIM_INTR_ENABLE;
3658
3659                 device_config->vp_config[i].rti.btimer_val =
3660                         (VXGE_RTI_BTIMER_VAL * 1000)/272;
3661
3662                 device_config->vp_config[i].rti.timer_ac_en =
3663                                                 VXGE_HW_TIM_TIMER_AC_ENABLE;
3664
3665                 device_config->vp_config[i].rti.timer_ci_en =
3666                                                 VXGE_HW_TIM_TIMER_CI_DISABLE;
3667
3668                 device_config->vp_config[i].rti.timer_ri_en =
3669                                                 VXGE_HW_TIM_TIMER_RI_DISABLE;
3670
3671                 device_config->vp_config[i].rti.util_sel =
3672                                 VXGE_HW_TIM_UTIL_SEL_LEGACY_RX_NET_UTIL;
3673
3674                 device_config->vp_config[i].rti.urange_a =
3675                                                 RTI_RX_URANGE_A;
3676                 device_config->vp_config[i].rti.urange_b =
3677                                                 RTI_RX_URANGE_B;
3678                 device_config->vp_config[i].rti.urange_c =
3679                                                 RTI_RX_URANGE_C;
3680                 device_config->vp_config[i].rti.uec_a = RTI_RX_UFC_A;
3681                 device_config->vp_config[i].rti.uec_b = RTI_RX_UFC_B;
3682                 device_config->vp_config[i].rti.uec_c = RTI_RX_UFC_C;
3683                 device_config->vp_config[i].rti.uec_d = RTI_RX_UFC_D;
3684
3685                 device_config->vp_config[i].rti.rtimer_val =
3686                         (VXGE_RTI_RTIMER_VAL * 1000) / 272;
3687
3688                 device_config->vp_config[i].rti.ltimer_val =
3689                         (VXGE_RTI_LTIMER_VAL * 1000) / 272;
3690
3691                 device_config->vp_config[i].rpa_strip_vlan_tag =
3692                         vlan_tag_strip;
3693         }
3694
3695         driver_config->vpath_per_dev = temp;
3696         return no_of_vpaths;
3697 }
3698
3699 /* initialize device configuratrions */
3700 static void __devinit vxge_device_config_init(
3701                                 struct vxge_hw_device_config *device_config,
3702                                 int *intr_type)
3703 {
3704         /* Used for CQRQ/SRQ. */
3705         device_config->dma_blockpool_initial =
3706                         VXGE_HW_INITIAL_DMA_BLOCK_POOL_SIZE;
3707
3708         device_config->dma_blockpool_max =
3709                         VXGE_HW_MAX_DMA_BLOCK_POOL_SIZE;
3710
3711         if (max_mac_vpath > VXGE_MAX_MAC_ADDR_COUNT)
3712                 max_mac_vpath = VXGE_MAX_MAC_ADDR_COUNT;
3713
3714 #ifndef CONFIG_PCI_MSI
3715         vxge_debug_init(VXGE_ERR,
3716                 "%s: This Kernel does not support "
3717                 "MSI-X. Defaulting to INTA", VXGE_DRIVER_NAME);
3718         *intr_type = INTA;
3719 #endif
3720
3721         /* Configure whether MSI-X or IRQL. */
3722         switch (*intr_type) {
3723         case INTA:
3724                 device_config->intr_mode = VXGE_HW_INTR_MODE_IRQLINE;
3725                 break;
3726
3727         case MSI_X:
3728                 device_config->intr_mode = VXGE_HW_INTR_MODE_MSIX;
3729                 break;
3730         }
3731         /* Timer period between device poll */
3732         device_config->device_poll_millis = VXGE_TIMER_DELAY;
3733
3734         /* Configure mac based steering. */
3735         device_config->rts_mac_en = addr_learn_en;
3736
3737         /* Configure Vpaths */
3738         device_config->rth_it_type = VXGE_HW_RTH_IT_TYPE_MULTI_IT;
3739
3740         vxge_debug_ll_config(VXGE_TRACE, "%s : Device Config Params ",
3741                         __func__);
3742         vxge_debug_ll_config(VXGE_TRACE, "dma_blockpool_initial : %d",
3743                         device_config->dma_blockpool_initial);
3744         vxge_debug_ll_config(VXGE_TRACE, "dma_blockpool_max : %d",
3745                         device_config->dma_blockpool_max);
3746         vxge_debug_ll_config(VXGE_TRACE, "intr_mode : %d",
3747                         device_config->intr_mode);
3748         vxge_debug_ll_config(VXGE_TRACE, "device_poll_millis : %d",
3749                         device_config->device_poll_millis);
3750         vxge_debug_ll_config(VXGE_TRACE, "rts_mac_en : %d",
3751                         device_config->rts_mac_en);
3752         vxge_debug_ll_config(VXGE_TRACE, "rth_en : %d",
3753                         device_config->rth_en);
3754         vxge_debug_ll_config(VXGE_TRACE, "rth_it_type : %d",
3755                         device_config->rth_it_type);
3756 }
3757
3758 static void __devinit vxge_print_parm(struct vxgedev *vdev, u64 vpath_mask)
3759 {
3760         int i;
3761
3762         vxge_debug_init(VXGE_TRACE,
3763                 "%s: %d Vpath(s) opened",
3764                 vdev->ndev->name, vdev->no_of_vpath);
3765
3766         switch (vdev->config.intr_type) {
3767         case INTA:
3768                 vxge_debug_init(VXGE_TRACE,
3769                         "%s: Interrupt type INTA", vdev->ndev->name);
3770                 break;
3771
3772         case MSI_X:
3773                 vxge_debug_init(VXGE_TRACE,
3774                         "%s: Interrupt type MSI-X", vdev->ndev->name);
3775                 break;
3776         }
3777
3778         if (vdev->config.rth_steering) {
3779                 vxge_debug_init(VXGE_TRACE,
3780                         "%s: RTH steering enabled for TCP_IPV4",
3781                         vdev->ndev->name);
3782         } else {
3783                 vxge_debug_init(VXGE_TRACE,
3784                         "%s: RTH steering disabled", vdev->ndev->name);
3785         }
3786
3787         switch (vdev->config.tx_steering_type) {
3788         case NO_STEERING:
3789                 vxge_debug_init(VXGE_TRACE,
3790                         "%s: Tx steering disabled", vdev->ndev->name);
3791                 break;
3792         case TX_PRIORITY_STEERING:
3793                 vxge_debug_init(VXGE_TRACE,
3794                         "%s: Unsupported tx steering option",
3795                         vdev->ndev->name);
3796                 vxge_debug_init(VXGE_TRACE,
3797                         "%s: Tx steering disabled", vdev->ndev->name);
3798                 vdev->config.tx_steering_type = 0;
3799                 break;
3800         case TX_VLAN_STEERING:
3801                 vxge_debug_init(VXGE_TRACE,
3802                         "%s: Unsupported tx steering option",
3803                         vdev->ndev->name);
3804                 vxge_debug_init(VXGE_TRACE,
3805                         "%s: Tx steering disabled", vdev->ndev->name);
3806                 vdev->config.tx_steering_type = 0;
3807                 break;
3808         case TX_MULTIQ_STEERING:
3809                 vxge_debug_init(VXGE_TRACE,
3810                         "%s: Tx multiqueue steering enabled",
3811                         vdev->ndev->name);
3812                 break;
3813         case TX_PORT_STEERING:
3814                 vxge_debug_init(VXGE_TRACE,
3815                         "%s: Tx port steering enabled",
3816                         vdev->ndev->name);
3817                 break;
3818         default:
3819                 vxge_debug_init(VXGE_ERR,
3820                         "%s: Unsupported tx steering type",
3821                         vdev->ndev->name);
3822                 vxge_debug_init(VXGE_TRACE,
3823                         "%s: Tx steering disabled", vdev->ndev->name);
3824                 vdev->config.tx_steering_type = 0;
3825         }
3826
3827         if (vdev->config.gro_enable) {
3828                 vxge_debug_init(VXGE_ERR,
3829                         "%s: Generic receive offload enabled",
3830                         vdev->ndev->name);
3831         } else
3832                 vxge_debug_init(VXGE_TRACE,
3833                         "%s: Generic receive offload disabled",
3834                         vdev->ndev->name);
3835
3836         if (vdev->config.addr_learn_en)
3837                 vxge_debug_init(VXGE_TRACE,
3838                         "%s: MAC Address learning enabled", vdev->ndev->name);
3839
3840         vxge_debug_init(VXGE_TRACE,
3841                 "%s: Rx doorbell mode enabled", vdev->ndev->name);
3842
3843         for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
3844                 if (!vxge_bVALn(vpath_mask, i, 1))
3845                         continue;
3846                 vxge_debug_ll_config(VXGE_TRACE,
3847                         "%s: MTU size - %d", vdev->ndev->name,
3848                         ((struct __vxge_hw_device  *)(vdev->devh))->
3849                                 config.vp_config[i].mtu);
3850                 vxge_debug_init(VXGE_TRACE,
3851                         "%s: VLAN tag stripping %s", vdev->ndev->name,
3852                         ((struct __vxge_hw_device  *)(vdev->devh))->
3853                                 config.vp_config[i].rpa_strip_vlan_tag
3854                         ? "Enabled" : "Disabled");
3855                 vxge_debug_init(VXGE_TRACE,
3856                         "%s: Ring blocks : %d", vdev->ndev->name,
3857                         ((struct __vxge_hw_device  *)(vdev->devh))->
3858                                 config.vp_config[i].ring.ring_blocks);
3859                 vxge_debug_init(VXGE_TRACE,
3860                         "%s: Fifo blocks : %d", vdev->ndev->name,
3861                         ((struct __vxge_hw_device  *)(vdev->devh))->
3862                                 config.vp_config[i].fifo.fifo_blocks);
3863                 vxge_debug_ll_config(VXGE_TRACE,
3864                         "%s: Max frags : %d", vdev->ndev->name,
3865                         ((struct __vxge_hw_device  *)(vdev->devh))->
3866                                 config.vp_config[i].fifo.max_frags);
3867                 break;
3868         }
3869 }
3870
3871 #ifdef CONFIG_PM
3872 /**
3873  * vxge_pm_suspend - vxge power management suspend entry point
3874  *
3875  */
3876 static int vxge_pm_suspend(struct pci_dev *pdev, pm_message_t state)
3877 {
3878         return -ENOSYS;
3879 }
3880 /**
3881  * vxge_pm_resume - vxge power management resume entry point
3882  *
3883  */
3884 static int vxge_pm_resume(struct pci_dev *pdev)
3885 {
3886         return -ENOSYS;
3887 }
3888
3889 #endif
3890
3891 /**
3892  * vxge_io_error_detected - called when PCI error is detected
3893  * @pdev: Pointer to PCI device
3894  * @state: The current pci connection state
3895  *
3896  * This function is called after a PCI bus error affecting
3897  * this device has been detected.
3898  */
3899 static pci_ers_result_t vxge_io_error_detected(struct pci_dev *pdev,
3900                                                 pci_channel_state_t state)
3901 {
3902         struct __vxge_hw_device  *hldev =
3903                 (struct __vxge_hw_device  *) pci_get_drvdata(pdev);
3904         struct net_device *netdev = hldev->ndev;
3905
3906         netif_device_detach(netdev);
3907
3908         if (state == pci_channel_io_perm_failure)
3909                 return PCI_ERS_RESULT_DISCONNECT;
3910
3911         if (netif_running(netdev)) {
3912                 /* Bring down the card, while avoiding PCI I/O */
3913                 do_vxge_close(netdev, 0);
3914         }
3915
3916         pci_disable_device(pdev);
3917
3918         return PCI_ERS_RESULT_NEED_RESET;
3919 }
3920
3921 /**
3922  * vxge_io_slot_reset - called after the pci bus has been reset.
3923  * @pdev: Pointer to PCI device
3924  *
3925  * Restart the card from scratch, as if from a cold-boot.
3926  * At this point, the card has exprienced a hard reset,
3927  * followed by fixups by BIOS, and has its config space
3928  * set up identically to what it was at cold boot.
3929  */
3930 static pci_ers_result_t vxge_io_slot_reset(struct pci_dev *pdev)
3931 {
3932         struct __vxge_hw_device  *hldev =
3933                 (struct __vxge_hw_device  *) pci_get_drvdata(pdev);
3934         struct net_device *netdev = hldev->ndev;
3935
3936         struct vxgedev *vdev = netdev_priv(netdev);
3937
3938         if (pci_enable_device(pdev)) {
3939                 printk(KERN_ERR "%s: "
3940                         "Cannot re-enable device after reset\n",
3941                         VXGE_DRIVER_NAME);
3942                 return PCI_ERS_RESULT_DISCONNECT;
3943         }
3944
3945         pci_set_master(pdev);
3946         vxge_reset(vdev);
3947
3948         return PCI_ERS_RESULT_RECOVERED;
3949 }
3950
3951 /**
3952  * vxge_io_resume - called when traffic can start flowing again.
3953  * @pdev: Pointer to PCI device
3954  *
3955  * This callback is called when the error recovery driver tells
3956  * us that its OK to resume normal operation.
3957  */
3958 static void vxge_io_resume(struct pci_dev *pdev)
3959 {
3960         struct __vxge_hw_device  *hldev =
3961                 (struct __vxge_hw_device  *) pci_get_drvdata(pdev);
3962         struct net_device *netdev = hldev->ndev;
3963
3964         if (netif_running(netdev)) {
3965                 if (vxge_open(netdev)) {
3966                         printk(KERN_ERR "%s: "
3967                                 "Can't bring device back up after reset\n",
3968                                 VXGE_DRIVER_NAME);
3969                         return;
3970                 }
3971         }
3972
3973         netif_device_attach(netdev);
3974 }
3975
3976 static inline u32 vxge_get_num_vfs(u64 function_mode)
3977 {
3978         u32 num_functions = 0;
3979
3980         switch (function_mode) {
3981         case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION:
3982         case VXGE_HW_FUNCTION_MODE_SRIOV_8:
3983                 num_functions = 8;
3984                 break;
3985         case VXGE_HW_FUNCTION_MODE_SINGLE_FUNCTION:
3986                 num_functions = 1;
3987                 break;
3988         case VXGE_HW_FUNCTION_MODE_SRIOV:
3989         case VXGE_HW_FUNCTION_MODE_MRIOV:
3990         case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION_17:
3991                 num_functions = 17;
3992                 break;
3993         case VXGE_HW_FUNCTION_MODE_SRIOV_4:
3994                 num_functions = 4;
3995                 break;
3996         case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION_2:
3997                 num_functions = 2;
3998                 break;
3999         case VXGE_HW_FUNCTION_MODE_MRIOV_8:
4000                 num_functions = 8; /* TODO */
4001                 break;
4002         }
4003         return num_functions;
4004 }
4005
4006 /**
4007  * vxge_probe
4008  * @pdev : structure containing the PCI related information of the device.
4009  * @pre: List of PCI devices supported by the driver listed in vxge_id_table.
4010  * Description:
4011  * This function is called when a new PCI device gets detected and initializes
4012  * it.
4013  * Return value:
4014  * returns 0 on success and negative on failure.
4015  *
4016  */
4017 static int __devinit
4018 vxge_probe(struct pci_dev *pdev, const struct pci_device_id *pre)
4019 {
4020         struct __vxge_hw_device  *hldev;
4021         enum vxge_hw_status status;
4022         int ret;
4023         int high_dma = 0;
4024         u64 vpath_mask = 0;
4025         struct vxgedev *vdev;
4026         struct vxge_config *ll_config = NULL;
4027         struct vxge_hw_device_config *device_config = NULL;
4028         struct vxge_hw_device_attr attr;
4029         int i, j, no_of_vpath = 0, max_vpath_supported = 0;
4030         u8 *macaddr;
4031         struct vxge_mac_addrs *entry;
4032         static int bus = -1, device = -1;
4033         u32 host_type;
4034         u8 new_device = 0;
4035         enum vxge_hw_status is_privileged;
4036         u32 function_mode;
4037         u32 num_vfs = 0;
4038
4039         vxge_debug_entryexit(VXGE_TRACE, "%s:%d", __func__, __LINE__);
4040         attr.pdev = pdev;
4041
4042         /* In SRIOV-17 mode, functions of the same adapter
4043          * can be deployed on different buses */
4044         if ((!pdev->is_virtfn) && ((bus != pdev->bus->number) ||
4045                 (device != PCI_SLOT(pdev->devfn))))
4046                 new_device = 1;
4047
4048         bus = pdev->bus->number;
4049         device = PCI_SLOT(pdev->devfn);
4050
4051         if (new_device) {
4052                 if (driver_config->config_dev_cnt &&
4053                    (driver_config->config_dev_cnt !=
4054                         driver_config->total_dev_cnt))
4055                         vxge_debug_init(VXGE_ERR,
4056                                 "%s: Configured %d of %d devices",
4057                                 VXGE_DRIVER_NAME,
4058                                 driver_config->config_dev_cnt,
4059                                 driver_config->total_dev_cnt);
4060                 driver_config->config_dev_cnt = 0;
4061                 driver_config->total_dev_cnt = 0;
4062         }
4063         /* Now making the CPU based no of vpath calculation
4064          * applicable for individual functions as well.
4065          */
4066         driver_config->g_no_cpus = 0;
4067         driver_config->vpath_per_dev = max_config_vpath;
4068
4069         driver_config->total_dev_cnt++;
4070         if (++driver_config->config_dev_cnt > max_config_dev) {
4071                 ret = 0;
4072                 goto _exit0;
4073         }
4074
4075         device_config = kzalloc(sizeof(struct vxge_hw_device_config),
4076                 GFP_KERNEL);
4077         if (!device_config) {
4078                 ret = -ENOMEM;
4079                 vxge_debug_init(VXGE_ERR,
4080                         "device_config : malloc failed %s %d",
4081                         __FILE__, __LINE__);
4082                 goto _exit0;
4083         }
4084
4085         ll_config = kzalloc(sizeof(*ll_config), GFP_KERNEL);
4086         if (!ll_config) {
4087                 ret = -ENOMEM;
4088                 vxge_debug_init(VXGE_ERR,
4089                         "ll_config : malloc failed %s %d",
4090                         __FILE__, __LINE__);
4091                 goto _exit0;
4092         }
4093         ll_config->tx_steering_type = TX_MULTIQ_STEERING;
4094         ll_config->intr_type = MSI_X;
4095         ll_config->napi_weight = NEW_NAPI_WEIGHT;
4096         ll_config->rth_steering = RTH_STEERING;
4097
4098         /* get the default configuration parameters */
4099         vxge_hw_device_config_default_get(device_config);
4100
4101         /* initialize configuration parameters */
4102         vxge_device_config_init(device_config, &ll_config->intr_type);
4103
4104         ret = pci_enable_device(pdev);
4105         if (ret) {
4106                 vxge_debug_init(VXGE_ERR,
4107                         "%s : can not enable PCI device", __func__);
4108                 goto _exit0;
4109         }
4110
4111         if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
4112                 vxge_debug_ll_config(VXGE_TRACE,
4113                         "%s : using 64bit DMA", __func__);
4114
4115                 high_dma = 1;
4116
4117                 if (pci_set_consistent_dma_mask(pdev,
4118                                                 DMA_BIT_MASK(64))) {
4119                         vxge_debug_init(VXGE_ERR,
4120                                 "%s : unable to obtain 64bit DMA for "
4121                                 "consistent allocations", __func__);
4122                         ret = -ENOMEM;
4123                         goto _exit1;
4124                 }
4125         } else if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(32))) {
4126                 vxge_debug_ll_config(VXGE_TRACE,
4127                         "%s : using 32bit DMA", __func__);
4128         } else {
4129                 ret = -ENOMEM;
4130                 goto _exit1;
4131         }
4132
4133         if (pci_request_regions(pdev, VXGE_DRIVER_NAME)) {
4134                 vxge_debug_init(VXGE_ERR,
4135                         "%s : request regions failed", __func__);
4136                 ret = -ENODEV;
4137                 goto _exit1;
4138         }
4139
4140         pci_set_master(pdev);
4141
4142         attr.bar0 = pci_ioremap_bar(pdev, 0);
4143         if (!attr.bar0) {
4144                 vxge_debug_init(VXGE_ERR,
4145                         "%s : cannot remap io memory bar0", __func__);
4146                 ret = -ENODEV;
4147                 goto _exit2;
4148         }
4149         vxge_debug_ll_config(VXGE_TRACE,
4150                 "pci ioremap bar0: %p:0x%llx",
4151                 attr.bar0,
4152                 (unsigned long long)pci_resource_start(pdev, 0));
4153
4154         status = vxge_hw_device_hw_info_get(attr.bar0,
4155                         &ll_config->device_hw_info);
4156         if (status != VXGE_HW_OK) {
4157                 vxge_debug_init(VXGE_ERR,
4158                         "%s: Reading of hardware info failed."
4159                         "Please try upgrading the firmware.", VXGE_DRIVER_NAME);
4160                 ret = -EINVAL;
4161                 goto _exit3;
4162         }
4163
4164         if (ll_config->device_hw_info.fw_version.major !=
4165                 VXGE_DRIVER_FW_VERSION_MAJOR) {
4166                 vxge_debug_init(VXGE_ERR,
4167                         "%s: Incorrect firmware version."
4168                         "Please upgrade the firmware to version 1.x.x",
4169                         VXGE_DRIVER_NAME);
4170                 ret = -EINVAL;
4171                 goto _exit3;
4172         }
4173
4174         vpath_mask = ll_config->device_hw_info.vpath_mask;
4175         if (vpath_mask == 0) {
4176                 vxge_debug_ll_config(VXGE_TRACE,
4177                         "%s: No vpaths available in device", VXGE_DRIVER_NAME);
4178                 ret = -EINVAL;
4179                 goto _exit3;
4180         }
4181
4182         vxge_debug_ll_config(VXGE_TRACE,
4183                 "%s:%d  Vpath mask = %llx", __func__, __LINE__,
4184                 (unsigned long long)vpath_mask);
4185
4186         function_mode = ll_config->device_hw_info.function_mode;
4187         host_type = ll_config->device_hw_info.host_type;
4188         is_privileged = __vxge_hw_device_is_privilaged(host_type,
4189                 ll_config->device_hw_info.func_id);
4190
4191         /* Check how many vpaths are available */
4192         for (i = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
4193                 if (!((vpath_mask) & vxge_mBIT(i)))
4194                         continue;
4195                 max_vpath_supported++;
4196         }
4197
4198         if (new_device)
4199                 num_vfs = vxge_get_num_vfs(function_mode) - 1;
4200
4201         /* Enable SRIOV mode, if firmware has SRIOV support and if it is a PF */
4202         if (is_sriov(function_mode) && (max_config_dev > 1) &&
4203                 (ll_config->intr_type != INTA) &&
4204                 (is_privileged == VXGE_HW_OK)) {
4205                 ret = pci_enable_sriov(pdev, ((max_config_dev - 1) < num_vfs)
4206                         ? (max_config_dev - 1) : num_vfs);
4207                 if (ret)
4208                         vxge_debug_ll_config(VXGE_ERR,
4209                                 "Failed in enabling SRIOV mode: %d\n", ret);
4210         }
4211
4212         /*
4213          * Configure vpaths and get driver configured number of vpaths
4214          * which is less than or equal to the maximum vpaths per function.
4215          */
4216         no_of_vpath = vxge_config_vpaths(device_config, vpath_mask, ll_config);
4217         if (!no_of_vpath) {
4218                 vxge_debug_ll_config(VXGE_ERR,
4219                         "%s: No more vpaths to configure", VXGE_DRIVER_NAME);
4220                 ret = 0;
4221                 goto _exit3;
4222         }
4223
4224         /* Setting driver callbacks */
4225         attr.uld_callbacks.link_up = vxge_callback_link_up;
4226         attr.uld_callbacks.link_down = vxge_callback_link_down;
4227         attr.uld_callbacks.crit_err = vxge_callback_crit_err;
4228
4229         status = vxge_hw_device_initialize(&hldev, &attr, device_config);
4230         if (status != VXGE_HW_OK) {
4231                 vxge_debug_init(VXGE_ERR,
4232                         "Failed to initialize device (%d)", status);
4233                         ret = -EINVAL;
4234                         goto _exit3;
4235         }
4236
4237         /* if FCS stripping is not disabled in MAC fail driver load */
4238         if (vxge_hw_vpath_strip_fcs_check(hldev, vpath_mask) != VXGE_HW_OK) {
4239                 vxge_debug_init(VXGE_ERR,
4240                         "%s: FCS stripping is not disabled in MAC"
4241                         " failing driver load", VXGE_DRIVER_NAME);
4242                 ret = -EINVAL;
4243                 goto _exit4;
4244         }
4245
4246         vxge_hw_device_debug_set(hldev, VXGE_ERR, VXGE_COMPONENT_LL);
4247
4248         /* set private device info */
4249         pci_set_drvdata(pdev, hldev);
4250
4251         ll_config->gro_enable = VXGE_GRO_ALWAYS_AGGREGATE;
4252         ll_config->fifo_indicate_max_pkts = VXGE_FIFO_INDICATE_MAX_PKTS;
4253         ll_config->addr_learn_en = addr_learn_en;
4254         ll_config->rth_algorithm = RTH_ALG_JENKINS;
4255         ll_config->rth_hash_type_tcpipv4 = VXGE_HW_RING_HASH_TYPE_TCP_IPV4;
4256         ll_config->rth_hash_type_ipv4 = VXGE_HW_RING_HASH_TYPE_NONE;
4257         ll_config->rth_hash_type_tcpipv6 = VXGE_HW_RING_HASH_TYPE_NONE;
4258         ll_config->rth_hash_type_ipv6 = VXGE_HW_RING_HASH_TYPE_NONE;
4259         ll_config->rth_hash_type_tcpipv6ex = VXGE_HW_RING_HASH_TYPE_NONE;
4260         ll_config->rth_hash_type_ipv6ex = VXGE_HW_RING_HASH_TYPE_NONE;
4261         ll_config->rth_bkt_sz = RTH_BUCKET_SIZE;
4262         ll_config->tx_pause_enable = VXGE_PAUSE_CTRL_ENABLE;
4263         ll_config->rx_pause_enable = VXGE_PAUSE_CTRL_ENABLE;
4264
4265         if (vxge_device_register(hldev, ll_config, high_dma, no_of_vpath,
4266                 &vdev)) {
4267                 ret = -EINVAL;
4268                 goto _exit4;
4269         }
4270
4271         vxge_hw_device_debug_set(hldev, VXGE_TRACE, VXGE_COMPONENT_LL);
4272         VXGE_COPY_DEBUG_INFO_TO_LL(vdev, vxge_hw_device_error_level_get(hldev),
4273                 vxge_hw_device_trace_level_get(hldev));
4274
4275         /* set private HW device info */
4276         hldev->ndev = vdev->ndev;
4277         vdev->mtu = VXGE_HW_DEFAULT_MTU;
4278         vdev->bar0 = attr.bar0;
4279         vdev->max_vpath_supported = max_vpath_supported;
4280         vdev->no_of_vpath = no_of_vpath;
4281
4282         /* Virtual Path count */
4283         for (i = 0, j = 0; i < VXGE_HW_MAX_VIRTUAL_PATHS; i++) {
4284                 if (!vxge_bVALn(vpath_mask, i, 1))
4285                         continue;
4286                 if (j >= vdev->no_of_vpath)
4287                         break;
4288
4289                 vdev->vpaths[j].is_configured = 1;
4290                 vdev->vpaths[j].device_id = i;
4291                 vdev->vpaths[j].fifo.driver_id = j;
4292                 vdev->vpaths[j].ring.driver_id = j;
4293                 vdev->vpaths[j].vdev = vdev;
4294                 vdev->vpaths[j].max_mac_addr_cnt = max_mac_vpath;
4295                 memcpy((u8 *)vdev->vpaths[j].macaddr,
4296                                 ll_config->device_hw_info.mac_addrs[i],
4297                                 ETH_ALEN);
4298
4299                 /* Initialize the mac address list header */
4300                 INIT_LIST_HEAD(&vdev->vpaths[j].mac_addr_list);
4301
4302                 vdev->vpaths[j].mac_addr_cnt = 0;
4303                 vdev->vpaths[j].mcast_addr_cnt = 0;
4304                 j++;
4305         }
4306         vdev->exec_mode = VXGE_EXEC_MODE_DISABLE;
4307         vdev->max_config_port = max_config_port;
4308
4309         vdev->vlan_tag_strip = vlan_tag_strip;
4310
4311         /* map the hashing selector table to the configured vpaths */
4312         for (i = 0; i < vdev->no_of_vpath; i++)
4313                 vdev->vpath_selector[i] = vpath_selector[i];
4314
4315         macaddr = (u8 *)vdev->vpaths[0].macaddr;
4316
4317         ll_config->device_hw_info.serial_number[VXGE_HW_INFO_LEN - 1] = '\0';
4318         ll_config->device_hw_info.product_desc[VXGE_HW_INFO_LEN - 1] = '\0';
4319         ll_config->device_hw_info.part_number[VXGE_HW_INFO_LEN - 1] = '\0';
4320
4321         vxge_debug_init(VXGE_TRACE, "%s: SERIAL NUMBER: %s",
4322                 vdev->ndev->name, ll_config->device_hw_info.serial_number);
4323
4324         vxge_debug_init(VXGE_TRACE, "%s: PART NUMBER: %s",
4325                 vdev->ndev->name, ll_config->device_hw_info.part_number);
4326
4327         vxge_debug_init(VXGE_TRACE, "%s: Neterion %s Server Adapter",
4328                 vdev->ndev->name, ll_config->device_hw_info.product_desc);
4329
4330         vxge_debug_init(VXGE_TRACE, "%s: MAC ADDR: %pM",
4331                 vdev->ndev->name, macaddr);
4332
4333         vxge_debug_init(VXGE_TRACE, "%s: Link Width x%d",
4334                 vdev->ndev->name, vxge_hw_device_link_width_get(hldev));
4335
4336         vxge_debug_init(VXGE_TRACE,
4337                 "%s: Firmware version : %s Date : %s", vdev->ndev->name,
4338                 ll_config->device_hw_info.fw_version.version,
4339                 ll_config->device_hw_info.fw_date.date);
4340
4341         if (new_device) {
4342                 switch (ll_config->device_hw_info.function_mode) {
4343                 case VXGE_HW_FUNCTION_MODE_SINGLE_FUNCTION:
4344                         vxge_debug_init(VXGE_TRACE,
4345                         "%s: Single Function Mode Enabled", vdev->ndev->name);
4346                 break;
4347                 case VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION:
4348                         vxge_debug_init(VXGE_TRACE,
4349                         "%s: Multi Function Mode Enabled", vdev->ndev->name);
4350                 break;
4351                 case VXGE_HW_FUNCTION_MODE_SRIOV:
4352                         vxge_debug_init(VXGE_TRACE,
4353                         "%s: Single Root IOV Mode Enabled", vdev->ndev->name);
4354                 break;
4355                 case VXGE_HW_FUNCTION_MODE_MRIOV:
4356                         vxge_debug_init(VXGE_TRACE,
4357                         "%s: Multi Root IOV Mode Enabled", vdev->ndev->name);
4358                 break;
4359                 }
4360         }
4361
4362         vxge_print_parm(vdev, vpath_mask);
4363
4364         /* Store the fw version for ethttool option */
4365         strcpy(vdev->fw_version, ll_config->device_hw_info.fw_version.version);
4366         memcpy(vdev->ndev->dev_addr, (u8 *)vdev->vpaths[0].macaddr, ETH_ALEN);
4367         memcpy(vdev->ndev->perm_addr, vdev->ndev->dev_addr, ETH_ALEN);
4368
4369         /* Copy the station mac address to the list */
4370         for (i = 0; i < vdev->no_of_vpath; i++) {
4371                 entry = (struct vxge_mac_addrs *)
4372                                 kzalloc(sizeof(struct vxge_mac_addrs),
4373                                         GFP_KERNEL);
4374                 if (NULL == entry) {
4375                         vxge_debug_init(VXGE_ERR,
4376                                 "%s: mac_addr_list : memory allocation failed",
4377                                 vdev->ndev->name);
4378                         ret = -EPERM;
4379                         goto _exit5;
4380                 }
4381                 macaddr = (u8 *)&entry->macaddr;
4382                 memcpy(macaddr, vdev->ndev->dev_addr, ETH_ALEN);
4383                 list_add(&entry->item, &vdev->vpaths[i].mac_addr_list);
4384                 vdev->vpaths[i].mac_addr_cnt = 1;
4385         }
4386
4387         kfree(device_config);
4388
4389         /*
4390          * INTA is shared in multi-function mode. This is unlike the INTA
4391          * implementation in MR mode, where each VH has its own INTA message.
4392          * - INTA is masked (disabled) as long as at least one function sets
4393          * its TITAN_MASK_ALL_INT.ALARM bit.
4394          * - INTA is unmasked (enabled) when all enabled functions have cleared
4395          * their own TITAN_MASK_ALL_INT.ALARM bit.
4396          * The TITAN_MASK_ALL_INT ALARM & TRAFFIC bits are cleared on power up.
4397          * Though this driver leaves the top level interrupts unmasked while
4398          * leaving the required module interrupt bits masked on exit, there
4399          * could be a rougue driver around that does not follow this procedure
4400          * resulting in a failure to generate interrupts. The following code is
4401          * present to prevent such a failure.
4402          */
4403
4404         if (ll_config->device_hw_info.function_mode ==
4405                 VXGE_HW_FUNCTION_MODE_MULTI_FUNCTION)
4406                 if (vdev->config.intr_type == INTA)
4407                         vxge_hw_device_unmask_all(hldev);
4408
4409         vxge_debug_entryexit(VXGE_TRACE, "%s: %s:%d  Exiting...",
4410                 vdev->ndev->name, __func__, __LINE__);
4411
4412         vxge_hw_device_debug_set(hldev, VXGE_ERR, VXGE_COMPONENT_LL);
4413         VXGE_COPY_DEBUG_INFO_TO_LL(vdev, vxge_hw_device_error_level_get(hldev),
4414                 vxge_hw_device_trace_level_get(hldev));
4415
4416         kfree(ll_config);
4417         return 0;
4418
4419 _exit5:
4420         for (i = 0; i < vdev->no_of_vpath; i++)
4421                 vxge_free_mac_add_list(&vdev->vpaths[i]);
4422
4423         vxge_device_unregister(hldev);
4424 _exit4:
4425         pci_disable_sriov(pdev);
4426         vxge_hw_device_terminate(hldev);
4427 _exit3:
4428         iounmap(attr.bar0);
4429 _exit2:
4430         pci_release_regions(pdev);
4431 _exit1:
4432         pci_disable_device(pdev);
4433 _exit0:
4434         kfree(ll_config);
4435         kfree(device_config);
4436         driver_config->config_dev_cnt--;
4437         pci_set_drvdata(pdev, NULL);
4438         return ret;
4439 }
4440
4441 /**
4442  * vxge_rem_nic - Free the PCI device
4443  * @pdev: structure containing the PCI related information of the device.
4444  * Description: This function is called by the Pci subsystem to release a
4445  * PCI device and free up all resource held up by the device.
4446  */
4447 static void __devexit
4448 vxge_remove(struct pci_dev *pdev)
4449 {
4450         struct __vxge_hw_device  *hldev;
4451         struct vxgedev *vdev = NULL;
4452         struct net_device *dev;
4453         int i = 0;
4454 #if ((VXGE_DEBUG_INIT & VXGE_DEBUG_MASK) || \
4455         (VXGE_DEBUG_ENTRYEXIT & VXGE_DEBUG_MASK))
4456         u32 level_trace;
4457 #endif
4458
4459         hldev = (struct __vxge_hw_device  *) pci_get_drvdata(pdev);
4460
4461         if (hldev == NULL)
4462                 return;
4463         dev = hldev->ndev;
4464         vdev = netdev_priv(dev);
4465
4466 #if ((VXGE_DEBUG_INIT & VXGE_DEBUG_MASK) || \
4467         (VXGE_DEBUG_ENTRYEXIT & VXGE_DEBUG_MASK))
4468         level_trace = vdev->level_trace;
4469 #endif
4470         vxge_debug_entryexit(level_trace,
4471                 "%s:%d", __func__, __LINE__);
4472
4473         vxge_debug_init(level_trace,
4474                 "%s : removing PCI device...", __func__);
4475         vxge_device_unregister(hldev);
4476
4477         for (i = 0; i < vdev->no_of_vpath; i++) {
4478                 vxge_free_mac_add_list(&vdev->vpaths[i]);
4479                 vdev->vpaths[i].mcast_addr_cnt = 0;
4480                 vdev->vpaths[i].mac_addr_cnt = 0;
4481         }
4482
4483         kfree(vdev->vpaths);
4484
4485         iounmap(vdev->bar0);
4486
4487         pci_disable_sriov(pdev);
4488
4489         /* we are safe to free it now */
4490         free_netdev(dev);
4491
4492         vxge_debug_init(level_trace,
4493                 "%s:%d  Device unregistered", __func__, __LINE__);
4494
4495         vxge_hw_device_terminate(hldev);
4496
4497         pci_disable_device(pdev);
4498         pci_release_regions(pdev);
4499         pci_set_drvdata(pdev, NULL);
4500         vxge_debug_entryexit(level_trace,
4501                 "%s:%d  Exiting...", __func__, __LINE__);
4502 }
4503
4504 static struct pci_error_handlers vxge_err_handler = {
4505         .error_detected = vxge_io_error_detected,
4506         .slot_reset = vxge_io_slot_reset,
4507         .resume = vxge_io_resume,
4508 };
4509
4510 static struct pci_driver vxge_driver = {
4511         .name = VXGE_DRIVER_NAME,
4512         .id_table = vxge_id_table,
4513         .probe = vxge_probe,
4514         .remove = __devexit_p(vxge_remove),
4515 #ifdef CONFIG_PM
4516         .suspend = vxge_pm_suspend,
4517         .resume = vxge_pm_resume,
4518 #endif
4519         .err_handler = &vxge_err_handler,
4520 };
4521
4522 static int __init
4523 vxge_starter(void)
4524 {
4525         int ret = 0;
4526         char version[32];
4527         snprintf(version, 32, "%s", DRV_VERSION);
4528
4529         printk(KERN_INFO "%s: Copyright(c) 2002-2009 Neterion Inc\n",
4530                 VXGE_DRIVER_NAME);
4531         printk(KERN_INFO "%s: Driver version: %s\n",
4532                         VXGE_DRIVER_NAME, version);
4533
4534         verify_bandwidth();
4535
4536         driver_config = kzalloc(sizeof(struct vxge_drv_config), GFP_KERNEL);
4537         if (!driver_config)
4538                 return -ENOMEM;
4539
4540         ret = pci_register_driver(&vxge_driver);
4541
4542         if (driver_config->config_dev_cnt &&
4543            (driver_config->config_dev_cnt != driver_config->total_dev_cnt))
4544                 vxge_debug_init(VXGE_ERR,
4545                         "%s: Configured %d of %d devices",
4546                         VXGE_DRIVER_NAME, driver_config->config_dev_cnt,
4547                         driver_config->total_dev_cnt);
4548
4549         if (ret)
4550                 kfree(driver_config);
4551
4552         return ret;
4553 }
4554
4555 static void __exit
4556 vxge_closer(void)
4557 {
4558         pci_unregister_driver(&vxge_driver);
4559         kfree(driver_config);
4560 }
4561 module_init(vxge_starter);
4562 module_exit(vxge_closer);