Merge branch 'fixes' of git://git.linux-nfs.org/pub/linux/nfs-2.6
[sfrench/cifs-2.6.git] / drivers / net / s2io.c
1 /************************************************************************
2  * s2io.c: A Linux PCI-X Ethernet driver for Neterion 10GbE Server NIC
3  * Copyright(c) 2002-2005 Neterion Inc.
4
5  * This software may be used and distributed according to the terms of
6  * the GNU General Public License (GPL), incorporated herein by reference.
7  * Drivers based on or derived from this code fall under the GPL and must
8  * retain the authorship, copyright and license notice.  This file is not
9  * a complete program and may only be used when the entire operating
10  * system is licensed under the GPL.
11  * See the file COPYING in this distribution for more information.
12  *
13  * Credits:
14  * Jeff Garzik          : For pointing out the improper error condition
15  *                        check in the s2io_xmit routine and also some
16  *                        issues in the Tx watch dog function. Also for
17  *                        patiently answering all those innumerable
18  *                        questions regaring the 2.6 porting issues.
19  * Stephen Hemminger    : Providing proper 2.6 porting mechanism for some
20  *                        macros available only in 2.6 Kernel.
21  * Francois Romieu      : For pointing out all code part that were
22  *                        deprecated and also styling related comments.
23  * Grant Grundler       : For helping me get rid of some Architecture
24  *                        dependent code.
25  * Christopher Hellwig  : Some more 2.6 specific issues in the driver.
26  *
27  * The module loadable parameters that are supported by the driver and a brief
28  * explaination of all the variables.
29  *
30  * rx_ring_num : This can be used to program the number of receive rings used
31  * in the driver.
32  * rx_ring_sz: This defines the number of receive blocks each ring can have.
33  *     This is also an array of size 8.
34  * rx_ring_mode: This defines the operation mode of all 8 rings. The valid
35  *              values are 1, 2 and 3.
36  * tx_fifo_num: This defines the number of Tx FIFOs thats used int the driver.
37  * tx_fifo_len: This too is an array of 8. Each element defines the number of
38  * Tx descriptors that can be associated with each corresponding FIFO.
39  * intr_type: This defines the type of interrupt. The values can be 0(INTA),
40  *     1(MSI), 2(MSI_X). Default value is '0(INTA)'
41  * lro: Specifies whether to enable Large Receive Offload (LRO) or not.
42  *     Possible values '1' for enable '0' for disable. Default is '0'
43  * lro_max_pkts: This parameter defines maximum number of packets can be
44  *     aggregated as a single large packet
45  ************************************************************************/
46
47 #include <linux/module.h>
48 #include <linux/types.h>
49 #include <linux/errno.h>
50 #include <linux/ioport.h>
51 #include <linux/pci.h>
52 #include <linux/dma-mapping.h>
53 #include <linux/kernel.h>
54 #include <linux/netdevice.h>
55 #include <linux/etherdevice.h>
56 #include <linux/skbuff.h>
57 #include <linux/init.h>
58 #include <linux/delay.h>
59 #include <linux/stddef.h>
60 #include <linux/ioctl.h>
61 #include <linux/timex.h>
62 #include <linux/sched.h>
63 #include <linux/ethtool.h>
64 #include <linux/workqueue.h>
65 #include <linux/if_vlan.h>
66 #include <linux/ip.h>
67 #include <linux/tcp.h>
68 #include <net/tcp.h>
69
70 #include <asm/system.h>
71 #include <asm/uaccess.h>
72 #include <asm/io.h>
73 #include <asm/div64.h>
74
75 /* local include */
76 #include "s2io.h"
77 #include "s2io-regs.h"
78
79 #define DRV_VERSION "2.0.15.2"
80
81 /* S2io Driver name & version. */
82 static char s2io_driver_name[] = "Neterion";
83 static char s2io_driver_version[] = DRV_VERSION;
84
85 static int rxd_size[4] = {32,48,48,64};
86 static int rxd_count[4] = {127,85,85,63};
87
88 static inline int RXD_IS_UP2DT(RxD_t *rxdp)
89 {
90         int ret;
91
92         ret = ((!(rxdp->Control_1 & RXD_OWN_XENA)) &&
93                 (GET_RXD_MARKER(rxdp->Control_2) != THE_RXD_MARK));
94
95         return ret;
96 }
97
98 /*
99  * Cards with following subsystem_id have a link state indication
100  * problem, 600B, 600C, 600D, 640B, 640C and 640D.
101  * macro below identifies these cards given the subsystem_id.
102  */
103 #define CARDS_WITH_FAULTY_LINK_INDICATORS(dev_type, subid) \
104         (dev_type == XFRAME_I_DEVICE) ?                 \
105                 ((((subid >= 0x600B) && (subid <= 0x600D)) || \
106                  ((subid >= 0x640B) && (subid <= 0x640D))) ? 1 : 0) : 0
107
108 #define LINK_IS_UP(val64) (!(val64 & (ADAPTER_STATUS_RMAC_REMOTE_FAULT | \
109                                       ADAPTER_STATUS_RMAC_LOCAL_FAULT)))
110 #define TASKLET_IN_USE test_and_set_bit(0, (&sp->tasklet_status))
111 #define PANIC   1
112 #define LOW     2
113 static inline int rx_buffer_level(nic_t * sp, int rxb_size, int ring)
114 {
115         mac_info_t *mac_control;
116
117         mac_control = &sp->mac_control;
118         if (rxb_size <= rxd_count[sp->rxd_mode])
119                 return PANIC;
120         else if ((mac_control->rings[ring].pkt_cnt - rxb_size) > 16)
121                 return  LOW;
122         return 0;
123 }
124
125 /* Ethtool related variables and Macros. */
126 static char s2io_gstrings[][ETH_GSTRING_LEN] = {
127         "Register test\t(offline)",
128         "Eeprom test\t(offline)",
129         "Link test\t(online)",
130         "RLDRAM test\t(offline)",
131         "BIST Test\t(offline)"
132 };
133
134 static char ethtool_stats_keys[][ETH_GSTRING_LEN] = {
135         {"tmac_frms"},
136         {"tmac_data_octets"},
137         {"tmac_drop_frms"},
138         {"tmac_mcst_frms"},
139         {"tmac_bcst_frms"},
140         {"tmac_pause_ctrl_frms"},
141         {"tmac_ttl_octets"},
142         {"tmac_ucst_frms"},
143         {"tmac_nucst_frms"},
144         {"tmac_any_err_frms"},
145         {"tmac_ttl_less_fb_octets"},
146         {"tmac_vld_ip_octets"},
147         {"tmac_vld_ip"},
148         {"tmac_drop_ip"},
149         {"tmac_icmp"},
150         {"tmac_rst_tcp"},
151         {"tmac_tcp"},
152         {"tmac_udp"},
153         {"rmac_vld_frms"},
154         {"rmac_data_octets"},
155         {"rmac_fcs_err_frms"},
156         {"rmac_drop_frms"},
157         {"rmac_vld_mcst_frms"},
158         {"rmac_vld_bcst_frms"},
159         {"rmac_in_rng_len_err_frms"},
160         {"rmac_out_rng_len_err_frms"},
161         {"rmac_long_frms"},
162         {"rmac_pause_ctrl_frms"},
163         {"rmac_unsup_ctrl_frms"},
164         {"rmac_ttl_octets"},
165         {"rmac_accepted_ucst_frms"},
166         {"rmac_accepted_nucst_frms"},
167         {"rmac_discarded_frms"},
168         {"rmac_drop_events"},
169         {"rmac_ttl_less_fb_octets"},
170         {"rmac_ttl_frms"},
171         {"rmac_usized_frms"},
172         {"rmac_osized_frms"},
173         {"rmac_frag_frms"},
174         {"rmac_jabber_frms"},
175         {"rmac_ttl_64_frms"},
176         {"rmac_ttl_65_127_frms"},
177         {"rmac_ttl_128_255_frms"},
178         {"rmac_ttl_256_511_frms"},
179         {"rmac_ttl_512_1023_frms"},
180         {"rmac_ttl_1024_1518_frms"},
181         {"rmac_ip"},
182         {"rmac_ip_octets"},
183         {"rmac_hdr_err_ip"},
184         {"rmac_drop_ip"},
185         {"rmac_icmp"},
186         {"rmac_tcp"},
187         {"rmac_udp"},
188         {"rmac_err_drp_udp"},
189         {"rmac_xgmii_err_sym"},
190         {"rmac_frms_q0"},
191         {"rmac_frms_q1"},
192         {"rmac_frms_q2"},
193         {"rmac_frms_q3"},
194         {"rmac_frms_q4"},
195         {"rmac_frms_q5"},
196         {"rmac_frms_q6"},
197         {"rmac_frms_q7"},
198         {"rmac_full_q0"},
199         {"rmac_full_q1"},
200         {"rmac_full_q2"},
201         {"rmac_full_q3"},
202         {"rmac_full_q4"},
203         {"rmac_full_q5"},
204         {"rmac_full_q6"},
205         {"rmac_full_q7"},
206         {"rmac_pause_cnt"},
207         {"rmac_xgmii_data_err_cnt"},
208         {"rmac_xgmii_ctrl_err_cnt"},
209         {"rmac_accepted_ip"},
210         {"rmac_err_tcp"},
211         {"rd_req_cnt"},
212         {"new_rd_req_cnt"},
213         {"new_rd_req_rtry_cnt"},
214         {"rd_rtry_cnt"},
215         {"wr_rtry_rd_ack_cnt"},
216         {"wr_req_cnt"},
217         {"new_wr_req_cnt"},
218         {"new_wr_req_rtry_cnt"},
219         {"wr_rtry_cnt"},
220         {"wr_disc_cnt"},
221         {"rd_rtry_wr_ack_cnt"},
222         {"txp_wr_cnt"},
223         {"txd_rd_cnt"},
224         {"txd_wr_cnt"},
225         {"rxd_rd_cnt"},
226         {"rxd_wr_cnt"},
227         {"txf_rd_cnt"},
228         {"rxf_wr_cnt"},
229         {"rmac_ttl_1519_4095_frms"},
230         {"rmac_ttl_4096_8191_frms"},
231         {"rmac_ttl_8192_max_frms"},
232         {"rmac_ttl_gt_max_frms"},
233         {"rmac_osized_alt_frms"},
234         {"rmac_jabber_alt_frms"},
235         {"rmac_gt_max_alt_frms"},
236         {"rmac_vlan_frms"},
237         {"rmac_len_discard"},
238         {"rmac_fcs_discard"},
239         {"rmac_pf_discard"},
240         {"rmac_da_discard"},
241         {"rmac_red_discard"},
242         {"rmac_rts_discard"},
243         {"rmac_ingm_full_discard"},
244         {"link_fault_cnt"},
245         {"\n DRIVER STATISTICS"},
246         {"single_bit_ecc_errs"},
247         {"double_bit_ecc_errs"},
248         {"parity_err_cnt"},
249         {"serious_err_cnt"},
250         {"soft_reset_cnt"},
251         {"fifo_full_cnt"},
252         {"ring_full_cnt"},
253         ("alarm_transceiver_temp_high"),
254         ("alarm_transceiver_temp_low"),
255         ("alarm_laser_bias_current_high"),
256         ("alarm_laser_bias_current_low"),
257         ("alarm_laser_output_power_high"),
258         ("alarm_laser_output_power_low"),
259         ("warn_transceiver_temp_high"),
260         ("warn_transceiver_temp_low"),
261         ("warn_laser_bias_current_high"),
262         ("warn_laser_bias_current_low"),
263         ("warn_laser_output_power_high"),
264         ("warn_laser_output_power_low"),
265         ("lro_aggregated_pkts"),
266         ("lro_flush_both_count"),
267         ("lro_out_of_sequence_pkts"),
268         ("lro_flush_due_to_max_pkts"),
269         ("lro_avg_aggr_pkts"),
270 };
271
272 #define S2IO_STAT_LEN sizeof(ethtool_stats_keys)/ ETH_GSTRING_LEN
273 #define S2IO_STAT_STRINGS_LEN S2IO_STAT_LEN * ETH_GSTRING_LEN
274
275 #define S2IO_TEST_LEN   sizeof(s2io_gstrings) / ETH_GSTRING_LEN
276 #define S2IO_STRINGS_LEN        S2IO_TEST_LEN * ETH_GSTRING_LEN
277
278 #define S2IO_TIMER_CONF(timer, handle, arg, exp)                \
279                         init_timer(&timer);                     \
280                         timer.function = handle;                \
281                         timer.data = (unsigned long) arg;       \
282                         mod_timer(&timer, (jiffies + exp))      \
283
284 /* Add the vlan */
285 static void s2io_vlan_rx_register(struct net_device *dev,
286                                         struct vlan_group *grp)
287 {
288         nic_t *nic = dev->priv;
289         unsigned long flags;
290
291         spin_lock_irqsave(&nic->tx_lock, flags);
292         nic->vlgrp = grp;
293         spin_unlock_irqrestore(&nic->tx_lock, flags);
294 }
295
296 /* Unregister the vlan */
297 static void s2io_vlan_rx_kill_vid(struct net_device *dev, unsigned long vid)
298 {
299         nic_t *nic = dev->priv;
300         unsigned long flags;
301
302         spin_lock_irqsave(&nic->tx_lock, flags);
303         if (nic->vlgrp)
304                 nic->vlgrp->vlan_devices[vid] = NULL;
305         spin_unlock_irqrestore(&nic->tx_lock, flags);
306 }
307
308 /*
309  * Constants to be programmed into the Xena's registers, to configure
310  * the XAUI.
311  */
312
313 #define END_SIGN        0x0
314 static const u64 herc_act_dtx_cfg[] = {
315         /* Set address */
316         0x8000051536750000ULL, 0x80000515367500E0ULL,
317         /* Write data */
318         0x8000051536750004ULL, 0x80000515367500E4ULL,
319         /* Set address */
320         0x80010515003F0000ULL, 0x80010515003F00E0ULL,
321         /* Write data */
322         0x80010515003F0004ULL, 0x80010515003F00E4ULL,
323         /* Set address */
324         0x801205150D440000ULL, 0x801205150D4400E0ULL,
325         /* Write data */
326         0x801205150D440004ULL, 0x801205150D4400E4ULL,
327         /* Set address */
328         0x80020515F2100000ULL, 0x80020515F21000E0ULL,
329         /* Write data */
330         0x80020515F2100004ULL, 0x80020515F21000E4ULL,
331         /* Done */
332         END_SIGN
333 };
334
335 static const u64 xena_dtx_cfg[] = {
336         /* Set address */
337         0x8000051500000000ULL, 0x80000515000000E0ULL,
338         /* Write data */
339         0x80000515D9350004ULL, 0x80000515D93500E4ULL,
340         /* Set address */
341         0x8001051500000000ULL, 0x80010515000000E0ULL,
342         /* Write data */
343         0x80010515001E0004ULL, 0x80010515001E00E4ULL,
344         /* Set address */
345         0x8002051500000000ULL, 0x80020515000000E0ULL,
346         /* Write data */
347         0x80020515F2100004ULL, 0x80020515F21000E4ULL,
348         END_SIGN
349 };
350
351 /*
352  * Constants for Fixing the MacAddress problem seen mostly on
353  * Alpha machines.
354  */
355 static const u64 fix_mac[] = {
356         0x0060000000000000ULL, 0x0060600000000000ULL,
357         0x0040600000000000ULL, 0x0000600000000000ULL,
358         0x0020600000000000ULL, 0x0060600000000000ULL,
359         0x0020600000000000ULL, 0x0060600000000000ULL,
360         0x0020600000000000ULL, 0x0060600000000000ULL,
361         0x0020600000000000ULL, 0x0060600000000000ULL,
362         0x0020600000000000ULL, 0x0060600000000000ULL,
363         0x0020600000000000ULL, 0x0060600000000000ULL,
364         0x0020600000000000ULL, 0x0060600000000000ULL,
365         0x0020600000000000ULL, 0x0060600000000000ULL,
366         0x0020600000000000ULL, 0x0060600000000000ULL,
367         0x0020600000000000ULL, 0x0060600000000000ULL,
368         0x0020600000000000ULL, 0x0000600000000000ULL,
369         0x0040600000000000ULL, 0x0060600000000000ULL,
370         END_SIGN
371 };
372
373 MODULE_AUTHOR("Raghavendra Koushik <raghavendra.koushik@neterion.com>");
374 MODULE_LICENSE("GPL");
375 MODULE_VERSION(DRV_VERSION);
376
377
378 /* Module Loadable parameters. */
379 S2IO_PARM_INT(tx_fifo_num, 1);
380 S2IO_PARM_INT(rx_ring_num, 1);
381
382
383 S2IO_PARM_INT(rx_ring_mode, 1);
384 S2IO_PARM_INT(use_continuous_tx_intrs, 1);
385 S2IO_PARM_INT(rmac_pause_time, 0x100);
386 S2IO_PARM_INT(mc_pause_threshold_q0q3, 187);
387 S2IO_PARM_INT(mc_pause_threshold_q4q7, 187);
388 S2IO_PARM_INT(shared_splits, 0);
389 S2IO_PARM_INT(tmac_util_period, 5);
390 S2IO_PARM_INT(rmac_util_period, 5);
391 S2IO_PARM_INT(bimodal, 0);
392 S2IO_PARM_INT(l3l4hdr_size, 128);
393 /* Frequency of Rx desc syncs expressed as power of 2 */
394 S2IO_PARM_INT(rxsync_frequency, 3);
395 /* Interrupt type. Values can be 0(INTA), 1(MSI), 2(MSI_X) */
396 S2IO_PARM_INT(intr_type, 0);
397 /* Large receive offload feature */
398 S2IO_PARM_INT(lro, 0);
399 /* Max pkts to be aggregated by LRO at one time. If not specified,
400  * aggregation happens until we hit max IP pkt size(64K)
401  */
402 S2IO_PARM_INT(lro_max_pkts, 0xFFFF);
403 #ifndef CONFIG_S2IO_NAPI
404 S2IO_PARM_INT(indicate_max_pkts, 0);
405 #endif
406
407 static unsigned int tx_fifo_len[MAX_TX_FIFOS] =
408     {DEFAULT_FIFO_0_LEN, [1 ...(MAX_TX_FIFOS - 1)] = DEFAULT_FIFO_1_7_LEN};
409 static unsigned int rx_ring_sz[MAX_RX_RINGS] =
410     {[0 ...(MAX_RX_RINGS - 1)] = SMALL_BLK_CNT};
411 static unsigned int rts_frm_len[MAX_RX_RINGS] =
412     {[0 ...(MAX_RX_RINGS - 1)] = 0 };
413
414 module_param_array(tx_fifo_len, uint, NULL, 0);
415 module_param_array(rx_ring_sz, uint, NULL, 0);
416 module_param_array(rts_frm_len, uint, NULL, 0);
417
418 /*
419  * S2IO device table.
420  * This table lists all the devices that this driver supports.
421  */
422 static struct pci_device_id s2io_tbl[] __devinitdata = {
423         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_WIN,
424          PCI_ANY_ID, PCI_ANY_ID},
425         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_UNI,
426          PCI_ANY_ID, PCI_ANY_ID},
427         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_WIN,
428          PCI_ANY_ID, PCI_ANY_ID},
429         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_UNI,
430          PCI_ANY_ID, PCI_ANY_ID},
431         {0,}
432 };
433
434 MODULE_DEVICE_TABLE(pci, s2io_tbl);
435
436 static struct pci_driver s2io_driver = {
437       .name = "S2IO",
438       .id_table = s2io_tbl,
439       .probe = s2io_init_nic,
440       .remove = __devexit_p(s2io_rem_nic),
441 };
442
443 /* A simplifier macro used both by init and free shared_mem Fns(). */
444 #define TXD_MEM_PAGE_CNT(len, per_each) ((len+per_each - 1) / per_each)
445
446 /**
447  * init_shared_mem - Allocation and Initialization of Memory
448  * @nic: Device private variable.
449  * Description: The function allocates all the memory areas shared
450  * between the NIC and the driver. This includes Tx descriptors,
451  * Rx descriptors and the statistics block.
452  */
453
454 static int init_shared_mem(struct s2io_nic *nic)
455 {
456         u32 size;
457         void *tmp_v_addr, *tmp_v_addr_next;
458         dma_addr_t tmp_p_addr, tmp_p_addr_next;
459         RxD_block_t *pre_rxd_blk = NULL;
460         int i, j, blk_cnt, rx_sz, tx_sz;
461         int lst_size, lst_per_page;
462         struct net_device *dev = nic->dev;
463         unsigned long tmp;
464         buffAdd_t *ba;
465
466         mac_info_t *mac_control;
467         struct config_param *config;
468
469         mac_control = &nic->mac_control;
470         config = &nic->config;
471
472
473         /* Allocation and initialization of TXDLs in FIOFs */
474         size = 0;
475         for (i = 0; i < config->tx_fifo_num; i++) {
476                 size += config->tx_cfg[i].fifo_len;
477         }
478         if (size > MAX_AVAILABLE_TXDS) {
479                 DBG_PRINT(ERR_DBG, "s2io: Requested TxDs too high, ");
480                 DBG_PRINT(ERR_DBG, "Requested: %d, max supported: 8192\n", size);
481                 return -EINVAL;
482         }
483
484         lst_size = (sizeof(TxD_t) * config->max_txds);
485         tx_sz = lst_size * size;
486         lst_per_page = PAGE_SIZE / lst_size;
487
488         for (i = 0; i < config->tx_fifo_num; i++) {
489                 int fifo_len = config->tx_cfg[i].fifo_len;
490                 int list_holder_size = fifo_len * sizeof(list_info_hold_t);
491                 mac_control->fifos[i].list_info = kmalloc(list_holder_size,
492                                                           GFP_KERNEL);
493                 if (!mac_control->fifos[i].list_info) {
494                         DBG_PRINT(ERR_DBG,
495                                   "Malloc failed for list_info\n");
496                         return -ENOMEM;
497                 }
498                 memset(mac_control->fifos[i].list_info, 0, list_holder_size);
499         }
500         for (i = 0; i < config->tx_fifo_num; i++) {
501                 int page_num = TXD_MEM_PAGE_CNT(config->tx_cfg[i].fifo_len,
502                                                 lst_per_page);
503                 mac_control->fifos[i].tx_curr_put_info.offset = 0;
504                 mac_control->fifos[i].tx_curr_put_info.fifo_len =
505                     config->tx_cfg[i].fifo_len - 1;
506                 mac_control->fifos[i].tx_curr_get_info.offset = 0;
507                 mac_control->fifos[i].tx_curr_get_info.fifo_len =
508                     config->tx_cfg[i].fifo_len - 1;
509                 mac_control->fifos[i].fifo_no = i;
510                 mac_control->fifos[i].nic = nic;
511                 mac_control->fifos[i].max_txds = MAX_SKB_FRAGS + 2;
512
513                 for (j = 0; j < page_num; j++) {
514                         int k = 0;
515                         dma_addr_t tmp_p;
516                         void *tmp_v;
517                         tmp_v = pci_alloc_consistent(nic->pdev,
518                                                      PAGE_SIZE, &tmp_p);
519                         if (!tmp_v) {
520                                 DBG_PRINT(ERR_DBG,
521                                           "pci_alloc_consistent ");
522                                 DBG_PRINT(ERR_DBG, "failed for TxDL\n");
523                                 return -ENOMEM;
524                         }
525                         /* If we got a zero DMA address(can happen on
526                          * certain platforms like PPC), reallocate.
527                          * Store virtual address of page we don't want,
528                          * to be freed later.
529                          */
530                         if (!tmp_p) {
531                                 mac_control->zerodma_virt_addr = tmp_v;
532                                 DBG_PRINT(INIT_DBG, 
533                                 "%s: Zero DMA address for TxDL. ", dev->name);
534                                 DBG_PRINT(INIT_DBG, 
535                                 "Virtual address %p\n", tmp_v);
536                                 tmp_v = pci_alloc_consistent(nic->pdev,
537                                                      PAGE_SIZE, &tmp_p);
538                                 if (!tmp_v) {
539                                         DBG_PRINT(ERR_DBG,
540                                           "pci_alloc_consistent ");
541                                         DBG_PRINT(ERR_DBG, "failed for TxDL\n");
542                                         return -ENOMEM;
543                                 }
544                         }
545                         while (k < lst_per_page) {
546                                 int l = (j * lst_per_page) + k;
547                                 if (l == config->tx_cfg[i].fifo_len)
548                                         break;
549                                 mac_control->fifos[i].list_info[l].list_virt_addr =
550                                     tmp_v + (k * lst_size);
551                                 mac_control->fifos[i].list_info[l].list_phy_addr =
552                                     tmp_p + (k * lst_size);
553                                 k++;
554                         }
555                 }
556         }
557
558         nic->ufo_in_band_v = kmalloc((sizeof(u64) * size), GFP_KERNEL);
559         if (!nic->ufo_in_band_v)
560                 return -ENOMEM;
561         memset(nic->ufo_in_band_v, 0, size);
562
563         /* Allocation and initialization of RXDs in Rings */
564         size = 0;
565         for (i = 0; i < config->rx_ring_num; i++) {
566                 if (config->rx_cfg[i].num_rxd %
567                     (rxd_count[nic->rxd_mode] + 1)) {
568                         DBG_PRINT(ERR_DBG, "%s: RxD count of ", dev->name);
569                         DBG_PRINT(ERR_DBG, "Ring%d is not a multiple of ",
570                                   i);
571                         DBG_PRINT(ERR_DBG, "RxDs per Block");
572                         return FAILURE;
573                 }
574                 size += config->rx_cfg[i].num_rxd;
575                 mac_control->rings[i].block_count =
576                         config->rx_cfg[i].num_rxd /
577                         (rxd_count[nic->rxd_mode] + 1 );
578                 mac_control->rings[i].pkt_cnt = config->rx_cfg[i].num_rxd -
579                         mac_control->rings[i].block_count;
580         }
581         if (nic->rxd_mode == RXD_MODE_1)
582                 size = (size * (sizeof(RxD1_t)));
583         else
584                 size = (size * (sizeof(RxD3_t)));
585         rx_sz = size;
586
587         for (i = 0; i < config->rx_ring_num; i++) {
588                 mac_control->rings[i].rx_curr_get_info.block_index = 0;
589                 mac_control->rings[i].rx_curr_get_info.offset = 0;
590                 mac_control->rings[i].rx_curr_get_info.ring_len =
591                     config->rx_cfg[i].num_rxd - 1;
592                 mac_control->rings[i].rx_curr_put_info.block_index = 0;
593                 mac_control->rings[i].rx_curr_put_info.offset = 0;
594                 mac_control->rings[i].rx_curr_put_info.ring_len =
595                     config->rx_cfg[i].num_rxd - 1;
596                 mac_control->rings[i].nic = nic;
597                 mac_control->rings[i].ring_no = i;
598
599                 blk_cnt = config->rx_cfg[i].num_rxd /
600                                 (rxd_count[nic->rxd_mode] + 1);
601                 /*  Allocating all the Rx blocks */
602                 for (j = 0; j < blk_cnt; j++) {
603                         rx_block_info_t *rx_blocks;
604                         int l;
605
606                         rx_blocks = &mac_control->rings[i].rx_blocks[j];
607                         size = SIZE_OF_BLOCK; //size is always page size
608                         tmp_v_addr = pci_alloc_consistent(nic->pdev, size,
609                                                           &tmp_p_addr);
610                         if (tmp_v_addr == NULL) {
611                                 /*
612                                  * In case of failure, free_shared_mem()
613                                  * is called, which should free any
614                                  * memory that was alloced till the
615                                  * failure happened.
616                                  */
617                                 rx_blocks->block_virt_addr = tmp_v_addr;
618                                 return -ENOMEM;
619                         }
620                         memset(tmp_v_addr, 0, size);
621                         rx_blocks->block_virt_addr = tmp_v_addr;
622                         rx_blocks->block_dma_addr = tmp_p_addr;
623                         rx_blocks->rxds = kmalloc(sizeof(rxd_info_t)*
624                                                   rxd_count[nic->rxd_mode],
625                                                   GFP_KERNEL);
626                         for (l=0; l<rxd_count[nic->rxd_mode];l++) {
627                                 rx_blocks->rxds[l].virt_addr =
628                                         rx_blocks->block_virt_addr +
629                                         (rxd_size[nic->rxd_mode] * l);
630                                 rx_blocks->rxds[l].dma_addr =
631                                         rx_blocks->block_dma_addr +
632                                         (rxd_size[nic->rxd_mode] * l);
633                         }
634                 }
635                 /* Interlinking all Rx Blocks */
636                 for (j = 0; j < blk_cnt; j++) {
637                         tmp_v_addr =
638                                 mac_control->rings[i].rx_blocks[j].block_virt_addr;
639                         tmp_v_addr_next =
640                                 mac_control->rings[i].rx_blocks[(j + 1) %
641                                               blk_cnt].block_virt_addr;
642                         tmp_p_addr =
643                                 mac_control->rings[i].rx_blocks[j].block_dma_addr;
644                         tmp_p_addr_next =
645                                 mac_control->rings[i].rx_blocks[(j + 1) %
646                                               blk_cnt].block_dma_addr;
647
648                         pre_rxd_blk = (RxD_block_t *) tmp_v_addr;
649                         pre_rxd_blk->reserved_2_pNext_RxD_block =
650                             (unsigned long) tmp_v_addr_next;
651                         pre_rxd_blk->pNext_RxD_Blk_physical =
652                             (u64) tmp_p_addr_next;
653                 }
654         }
655         if (nic->rxd_mode >= RXD_MODE_3A) {
656                 /*
657                  * Allocation of Storages for buffer addresses in 2BUFF mode
658                  * and the buffers as well.
659                  */
660                 for (i = 0; i < config->rx_ring_num; i++) {
661                         blk_cnt = config->rx_cfg[i].num_rxd /
662                            (rxd_count[nic->rxd_mode]+ 1);
663                         mac_control->rings[i].ba =
664                                 kmalloc((sizeof(buffAdd_t *) * blk_cnt),
665                                      GFP_KERNEL);
666                         if (!mac_control->rings[i].ba)
667                                 return -ENOMEM;
668                         for (j = 0; j < blk_cnt; j++) {
669                                 int k = 0;
670                                 mac_control->rings[i].ba[j] =
671                                         kmalloc((sizeof(buffAdd_t) *
672                                                 (rxd_count[nic->rxd_mode] + 1)),
673                                                 GFP_KERNEL);
674                                 if (!mac_control->rings[i].ba[j])
675                                         return -ENOMEM;
676                                 while (k != rxd_count[nic->rxd_mode]) {
677                                         ba = &mac_control->rings[i].ba[j][k];
678
679                                         ba->ba_0_org = (void *) kmalloc
680                                             (BUF0_LEN + ALIGN_SIZE, GFP_KERNEL);
681                                         if (!ba->ba_0_org)
682                                                 return -ENOMEM;
683                                         tmp = (unsigned long)ba->ba_0_org;
684                                         tmp += ALIGN_SIZE;
685                                         tmp &= ~((unsigned long) ALIGN_SIZE);
686                                         ba->ba_0 = (void *) tmp;
687
688                                         ba->ba_1_org = (void *) kmalloc
689                                             (BUF1_LEN + ALIGN_SIZE, GFP_KERNEL);
690                                         if (!ba->ba_1_org)
691                                                 return -ENOMEM;
692                                         tmp = (unsigned long) ba->ba_1_org;
693                                         tmp += ALIGN_SIZE;
694                                         tmp &= ~((unsigned long) ALIGN_SIZE);
695                                         ba->ba_1 = (void *) tmp;
696                                         k++;
697                                 }
698                         }
699                 }
700         }
701
702         /* Allocation and initialization of Statistics block */
703         size = sizeof(StatInfo_t);
704         mac_control->stats_mem = pci_alloc_consistent
705             (nic->pdev, size, &mac_control->stats_mem_phy);
706
707         if (!mac_control->stats_mem) {
708                 /*
709                  * In case of failure, free_shared_mem() is called, which
710                  * should free any memory that was alloced till the
711                  * failure happened.
712                  */
713                 return -ENOMEM;
714         }
715         mac_control->stats_mem_sz = size;
716
717         tmp_v_addr = mac_control->stats_mem;
718         mac_control->stats_info = (StatInfo_t *) tmp_v_addr;
719         memset(tmp_v_addr, 0, size);
720         DBG_PRINT(INIT_DBG, "%s:Ring Mem PHY: 0x%llx\n", dev->name,
721                   (unsigned long long) tmp_p_addr);
722
723         return SUCCESS;
724 }
725
726 /**
727  * free_shared_mem - Free the allocated Memory
728  * @nic:  Device private variable.
729  * Description: This function is to free all memory locations allocated by
730  * the init_shared_mem() function and return it to the kernel.
731  */
732
733 static void free_shared_mem(struct s2io_nic *nic)
734 {
735         int i, j, blk_cnt, size;
736         void *tmp_v_addr;
737         dma_addr_t tmp_p_addr;
738         mac_info_t *mac_control;
739         struct config_param *config;
740         int lst_size, lst_per_page;
741         struct net_device *dev = nic->dev;
742
743         if (!nic)
744                 return;
745
746         mac_control = &nic->mac_control;
747         config = &nic->config;
748
749         lst_size = (sizeof(TxD_t) * config->max_txds);
750         lst_per_page = PAGE_SIZE / lst_size;
751
752         for (i = 0; i < config->tx_fifo_num; i++) {
753                 int page_num = TXD_MEM_PAGE_CNT(config->tx_cfg[i].fifo_len,
754                                                 lst_per_page);
755                 for (j = 0; j < page_num; j++) {
756                         int mem_blks = (j * lst_per_page);
757                         if (!mac_control->fifos[i].list_info)
758                                 return; 
759                         if (!mac_control->fifos[i].list_info[mem_blks].
760                                  list_virt_addr)
761                                 break;
762                         pci_free_consistent(nic->pdev, PAGE_SIZE,
763                                             mac_control->fifos[i].
764                                             list_info[mem_blks].
765                                             list_virt_addr,
766                                             mac_control->fifos[i].
767                                             list_info[mem_blks].
768                                             list_phy_addr);
769                 }
770                 /* If we got a zero DMA address during allocation,
771                  * free the page now
772                  */
773                 if (mac_control->zerodma_virt_addr) {
774                         pci_free_consistent(nic->pdev, PAGE_SIZE,
775                                             mac_control->zerodma_virt_addr,
776                                             (dma_addr_t)0);
777                         DBG_PRINT(INIT_DBG, 
778                                 "%s: Freeing TxDL with zero DMA addr. ",
779                                 dev->name);
780                         DBG_PRINT(INIT_DBG, "Virtual address %p\n",
781                                 mac_control->zerodma_virt_addr);
782                 }
783                 kfree(mac_control->fifos[i].list_info);
784         }
785
786         size = SIZE_OF_BLOCK;
787         for (i = 0; i < config->rx_ring_num; i++) {
788                 blk_cnt = mac_control->rings[i].block_count;
789                 for (j = 0; j < blk_cnt; j++) {
790                         tmp_v_addr = mac_control->rings[i].rx_blocks[j].
791                                 block_virt_addr;
792                         tmp_p_addr = mac_control->rings[i].rx_blocks[j].
793                                 block_dma_addr;
794                         if (tmp_v_addr == NULL)
795                                 break;
796                         pci_free_consistent(nic->pdev, size,
797                                             tmp_v_addr, tmp_p_addr);
798                         kfree(mac_control->rings[i].rx_blocks[j].rxds);
799                 }
800         }
801
802         if (nic->rxd_mode >= RXD_MODE_3A) {
803                 /* Freeing buffer storage addresses in 2BUFF mode. */
804                 for (i = 0; i < config->rx_ring_num; i++) {
805                         blk_cnt = config->rx_cfg[i].num_rxd /
806                             (rxd_count[nic->rxd_mode] + 1);
807                         for (j = 0; j < blk_cnt; j++) {
808                                 int k = 0;
809                                 if (!mac_control->rings[i].ba[j])
810                                         continue;
811                                 while (k != rxd_count[nic->rxd_mode]) {
812                                         buffAdd_t *ba =
813                                                 &mac_control->rings[i].ba[j][k];
814                                         kfree(ba->ba_0_org);
815                                         kfree(ba->ba_1_org);
816                                         k++;
817                                 }
818                                 kfree(mac_control->rings[i].ba[j]);
819                         }
820                         kfree(mac_control->rings[i].ba);
821                 }
822         }
823
824         if (mac_control->stats_mem) {
825                 pci_free_consistent(nic->pdev,
826                                     mac_control->stats_mem_sz,
827                                     mac_control->stats_mem,
828                                     mac_control->stats_mem_phy);
829         }
830         if (nic->ufo_in_band_v)
831                 kfree(nic->ufo_in_band_v);
832 }
833
834 /**
835  * s2io_verify_pci_mode -
836  */
837
838 static int s2io_verify_pci_mode(nic_t *nic)
839 {
840         XENA_dev_config_t __iomem *bar0 = nic->bar0;
841         register u64 val64 = 0;
842         int     mode;
843
844         val64 = readq(&bar0->pci_mode);
845         mode = (u8)GET_PCI_MODE(val64);
846
847         if ( val64 & PCI_MODE_UNKNOWN_MODE)
848                 return -1;      /* Unknown PCI mode */
849         return mode;
850 }
851
852 #define NEC_VENID   0x1033
853 #define NEC_DEVID   0x0125
854 static int s2io_on_nec_bridge(struct pci_dev *s2io_pdev)
855 {
856         struct pci_dev *tdev = NULL;
857         while ((tdev = pci_find_device(PCI_ANY_ID, PCI_ANY_ID, tdev)) != NULL) {
858                 if ((tdev->vendor == NEC_VENID) && (tdev->device == NEC_DEVID)){
859                         if (tdev->bus == s2io_pdev->bus->parent)
860                                 return 1;
861                 }
862         }
863         return 0;
864 }
865
866 static int bus_speed[8] = {33, 133, 133, 200, 266, 133, 200, 266};
867 /**
868  * s2io_print_pci_mode -
869  */
870 static int s2io_print_pci_mode(nic_t *nic)
871 {
872         XENA_dev_config_t __iomem *bar0 = nic->bar0;
873         register u64 val64 = 0;
874         int     mode;
875         struct config_param *config = &nic->config;
876
877         val64 = readq(&bar0->pci_mode);
878         mode = (u8)GET_PCI_MODE(val64);
879
880         if ( val64 & PCI_MODE_UNKNOWN_MODE)
881                 return -1;      /* Unknown PCI mode */
882
883         config->bus_speed = bus_speed[mode];
884
885         if (s2io_on_nec_bridge(nic->pdev)) {
886                 DBG_PRINT(ERR_DBG, "%s: Device is on PCI-E bus\n",
887                                                         nic->dev->name);
888                 return mode;
889         }
890
891         if (val64 & PCI_MODE_32_BITS) {
892                 DBG_PRINT(ERR_DBG, "%s: Device is on 32 bit ", nic->dev->name);
893         } else {
894                 DBG_PRINT(ERR_DBG, "%s: Device is on 64 bit ", nic->dev->name);
895         }
896
897         switch(mode) {
898                 case PCI_MODE_PCI_33:
899                         DBG_PRINT(ERR_DBG, "33MHz PCI bus\n");
900                         break;
901                 case PCI_MODE_PCI_66:
902                         DBG_PRINT(ERR_DBG, "66MHz PCI bus\n");
903                         break;
904                 case PCI_MODE_PCIX_M1_66:
905                         DBG_PRINT(ERR_DBG, "66MHz PCIX(M1) bus\n");
906                         break;
907                 case PCI_MODE_PCIX_M1_100:
908                         DBG_PRINT(ERR_DBG, "100MHz PCIX(M1) bus\n");
909                         break;
910                 case PCI_MODE_PCIX_M1_133:
911                         DBG_PRINT(ERR_DBG, "133MHz PCIX(M1) bus\n");
912                         break;
913                 case PCI_MODE_PCIX_M2_66:
914                         DBG_PRINT(ERR_DBG, "133MHz PCIX(M2) bus\n");
915                         break;
916                 case PCI_MODE_PCIX_M2_100:
917                         DBG_PRINT(ERR_DBG, "200MHz PCIX(M2) bus\n");
918                         break;
919                 case PCI_MODE_PCIX_M2_133:
920                         DBG_PRINT(ERR_DBG, "266MHz PCIX(M2) bus\n");
921                         break;
922                 default:
923                         return -1;      /* Unsupported bus speed */
924         }
925
926         return mode;
927 }
928
929 /**
930  *  init_nic - Initialization of hardware
931  *  @nic: device peivate variable
932  *  Description: The function sequentially configures every block
933  *  of the H/W from their reset values.
934  *  Return Value:  SUCCESS on success and
935  *  '-1' on failure (endian settings incorrect).
936  */
937
938 static int init_nic(struct s2io_nic *nic)
939 {
940         XENA_dev_config_t __iomem *bar0 = nic->bar0;
941         struct net_device *dev = nic->dev;
942         register u64 val64 = 0;
943         void __iomem *add;
944         u32 time;
945         int i, j;
946         mac_info_t *mac_control;
947         struct config_param *config;
948         int dtx_cnt = 0;
949         unsigned long long mem_share;
950         int mem_size;
951
952         mac_control = &nic->mac_control;
953         config = &nic->config;
954
955         /* to set the swapper controle on the card */
956         if(s2io_set_swapper(nic)) {
957                 DBG_PRINT(ERR_DBG,"ERROR: Setting Swapper failed\n");
958                 return -1;
959         }
960
961         /*
962          * Herc requires EOI to be removed from reset before XGXS, so..
963          */
964         if (nic->device_type & XFRAME_II_DEVICE) {
965                 val64 = 0xA500000000ULL;
966                 writeq(val64, &bar0->sw_reset);
967                 msleep(500);
968                 val64 = readq(&bar0->sw_reset);
969         }
970
971         /* Remove XGXS from reset state */
972         val64 = 0;
973         writeq(val64, &bar0->sw_reset);
974         msleep(500);
975         val64 = readq(&bar0->sw_reset);
976
977         /*  Enable Receiving broadcasts */
978         add = &bar0->mac_cfg;
979         val64 = readq(&bar0->mac_cfg);
980         val64 |= MAC_RMAC_BCAST_ENABLE;
981         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
982         writel((u32) val64, add);
983         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
984         writel((u32) (val64 >> 32), (add + 4));
985
986         /* Read registers in all blocks */
987         val64 = readq(&bar0->mac_int_mask);
988         val64 = readq(&bar0->mc_int_mask);
989         val64 = readq(&bar0->xgxs_int_mask);
990
991         /*  Set MTU */
992         val64 = dev->mtu;
993         writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
994
995         if (nic->device_type & XFRAME_II_DEVICE) {
996                 while (herc_act_dtx_cfg[dtx_cnt] != END_SIGN) {
997                         SPECIAL_REG_WRITE(herc_act_dtx_cfg[dtx_cnt],
998                                           &bar0->dtx_control, UF);
999                         if (dtx_cnt & 0x1)
1000                                 msleep(1); /* Necessary!! */
1001                         dtx_cnt++;
1002                 }
1003         } else {
1004                 while (xena_dtx_cfg[dtx_cnt] != END_SIGN) {
1005                         SPECIAL_REG_WRITE(xena_dtx_cfg[dtx_cnt],
1006                                           &bar0->dtx_control, UF);
1007                         val64 = readq(&bar0->dtx_control);
1008                         dtx_cnt++;
1009                 }
1010         }
1011
1012         /*  Tx DMA Initialization */
1013         val64 = 0;
1014         writeq(val64, &bar0->tx_fifo_partition_0);
1015         writeq(val64, &bar0->tx_fifo_partition_1);
1016         writeq(val64, &bar0->tx_fifo_partition_2);
1017         writeq(val64, &bar0->tx_fifo_partition_3);
1018
1019
1020         for (i = 0, j = 0; i < config->tx_fifo_num; i++) {
1021                 val64 |=
1022                     vBIT(config->tx_cfg[i].fifo_len - 1, ((i * 32) + 19),
1023                          13) | vBIT(config->tx_cfg[i].fifo_priority,
1024                                     ((i * 32) + 5), 3);
1025
1026                 if (i == (config->tx_fifo_num - 1)) {
1027                         if (i % 2 == 0)
1028                                 i++;
1029                 }
1030
1031                 switch (i) {
1032                 case 1:
1033                         writeq(val64, &bar0->tx_fifo_partition_0);
1034                         val64 = 0;
1035                         break;
1036                 case 3:
1037                         writeq(val64, &bar0->tx_fifo_partition_1);
1038                         val64 = 0;
1039                         break;
1040                 case 5:
1041                         writeq(val64, &bar0->tx_fifo_partition_2);
1042                         val64 = 0;
1043                         break;
1044                 case 7:
1045                         writeq(val64, &bar0->tx_fifo_partition_3);
1046                         break;
1047                 }
1048         }
1049
1050         /*
1051          * Disable 4 PCCs for Xena1, 2 and 3 as per H/W bug
1052          * SXE-008 TRANSMIT DMA ARBITRATION ISSUE.
1053          */
1054         if ((nic->device_type == XFRAME_I_DEVICE) &&
1055                 (get_xena_rev_id(nic->pdev) < 4))
1056                 writeq(PCC_ENABLE_FOUR, &bar0->pcc_enable);
1057
1058         val64 = readq(&bar0->tx_fifo_partition_0);
1059         DBG_PRINT(INIT_DBG, "Fifo partition at: 0x%p is: 0x%llx\n",
1060                   &bar0->tx_fifo_partition_0, (unsigned long long) val64);
1061
1062         /*
1063          * Initialization of Tx_PA_CONFIG register to ignore packet
1064          * integrity checking.
1065          */
1066         val64 = readq(&bar0->tx_pa_cfg);
1067         val64 |= TX_PA_CFG_IGNORE_FRM_ERR | TX_PA_CFG_IGNORE_SNAP_OUI |
1068             TX_PA_CFG_IGNORE_LLC_CTRL | TX_PA_CFG_IGNORE_L2_ERR;
1069         writeq(val64, &bar0->tx_pa_cfg);
1070
1071         /* Rx DMA intialization. */
1072         val64 = 0;
1073         for (i = 0; i < config->rx_ring_num; i++) {
1074                 val64 |=
1075                     vBIT(config->rx_cfg[i].ring_priority, (5 + (i * 8)),
1076                          3);
1077         }
1078         writeq(val64, &bar0->rx_queue_priority);
1079
1080         /*
1081          * Allocating equal share of memory to all the
1082          * configured Rings.
1083          */
1084         val64 = 0;
1085         if (nic->device_type & XFRAME_II_DEVICE)
1086                 mem_size = 32;
1087         else
1088                 mem_size = 64;
1089
1090         for (i = 0; i < config->rx_ring_num; i++) {
1091                 switch (i) {
1092                 case 0:
1093                         mem_share = (mem_size / config->rx_ring_num +
1094                                      mem_size % config->rx_ring_num);
1095                         val64 |= RX_QUEUE_CFG_Q0_SZ(mem_share);
1096                         continue;
1097                 case 1:
1098                         mem_share = (mem_size / config->rx_ring_num);
1099                         val64 |= RX_QUEUE_CFG_Q1_SZ(mem_share);
1100                         continue;
1101                 case 2:
1102                         mem_share = (mem_size / config->rx_ring_num);
1103                         val64 |= RX_QUEUE_CFG_Q2_SZ(mem_share);
1104                         continue;
1105                 case 3:
1106                         mem_share = (mem_size / config->rx_ring_num);
1107                         val64 |= RX_QUEUE_CFG_Q3_SZ(mem_share);
1108                         continue;
1109                 case 4:
1110                         mem_share = (mem_size / config->rx_ring_num);
1111                         val64 |= RX_QUEUE_CFG_Q4_SZ(mem_share);
1112                         continue;
1113                 case 5:
1114                         mem_share = (mem_size / config->rx_ring_num);
1115                         val64 |= RX_QUEUE_CFG_Q5_SZ(mem_share);
1116                         continue;
1117                 case 6:
1118                         mem_share = (mem_size / config->rx_ring_num);
1119                         val64 |= RX_QUEUE_CFG_Q6_SZ(mem_share);
1120                         continue;
1121                 case 7:
1122                         mem_share = (mem_size / config->rx_ring_num);
1123                         val64 |= RX_QUEUE_CFG_Q7_SZ(mem_share);
1124                         continue;
1125                 }
1126         }
1127         writeq(val64, &bar0->rx_queue_cfg);
1128
1129         /*
1130          * Filling Tx round robin registers
1131          * as per the number of FIFOs
1132          */
1133         switch (config->tx_fifo_num) {
1134         case 1:
1135                 val64 = 0x0000000000000000ULL;
1136                 writeq(val64, &bar0->tx_w_round_robin_0);
1137                 writeq(val64, &bar0->tx_w_round_robin_1);
1138                 writeq(val64, &bar0->tx_w_round_robin_2);
1139                 writeq(val64, &bar0->tx_w_round_robin_3);
1140                 writeq(val64, &bar0->tx_w_round_robin_4);
1141                 break;
1142         case 2:
1143                 val64 = 0x0000010000010000ULL;
1144                 writeq(val64, &bar0->tx_w_round_robin_0);
1145                 val64 = 0x0100000100000100ULL;
1146                 writeq(val64, &bar0->tx_w_round_robin_1);
1147                 val64 = 0x0001000001000001ULL;
1148                 writeq(val64, &bar0->tx_w_round_robin_2);
1149                 val64 = 0x0000010000010000ULL;
1150                 writeq(val64, &bar0->tx_w_round_robin_3);
1151                 val64 = 0x0100000000000000ULL;
1152                 writeq(val64, &bar0->tx_w_round_robin_4);
1153                 break;
1154         case 3:
1155                 val64 = 0x0001000102000001ULL;
1156                 writeq(val64, &bar0->tx_w_round_robin_0);
1157                 val64 = 0x0001020000010001ULL;
1158                 writeq(val64, &bar0->tx_w_round_robin_1);
1159                 val64 = 0x0200000100010200ULL;
1160                 writeq(val64, &bar0->tx_w_round_robin_2);
1161                 val64 = 0x0001000102000001ULL;
1162                 writeq(val64, &bar0->tx_w_round_robin_3);
1163                 val64 = 0x0001020000000000ULL;
1164                 writeq(val64, &bar0->tx_w_round_robin_4);
1165                 break;
1166         case 4:
1167                 val64 = 0x0001020300010200ULL;
1168                 writeq(val64, &bar0->tx_w_round_robin_0);
1169                 val64 = 0x0100000102030001ULL;
1170                 writeq(val64, &bar0->tx_w_round_robin_1);
1171                 val64 = 0x0200010000010203ULL;
1172                 writeq(val64, &bar0->tx_w_round_robin_2);
1173                 val64 = 0x0001020001000001ULL;
1174                 writeq(val64, &bar0->tx_w_round_robin_3);
1175                 val64 = 0x0203000100000000ULL;
1176                 writeq(val64, &bar0->tx_w_round_robin_4);
1177                 break;
1178         case 5:
1179                 val64 = 0x0001000203000102ULL;
1180                 writeq(val64, &bar0->tx_w_round_robin_0);
1181                 val64 = 0x0001020001030004ULL;
1182                 writeq(val64, &bar0->tx_w_round_robin_1);
1183                 val64 = 0x0001000203000102ULL;
1184                 writeq(val64, &bar0->tx_w_round_robin_2);
1185                 val64 = 0x0001020001030004ULL;
1186                 writeq(val64, &bar0->tx_w_round_robin_3);
1187                 val64 = 0x0001000000000000ULL;
1188                 writeq(val64, &bar0->tx_w_round_robin_4);
1189                 break;
1190         case 6:
1191                 val64 = 0x0001020304000102ULL;
1192                 writeq(val64, &bar0->tx_w_round_robin_0);
1193                 val64 = 0x0304050001020001ULL;
1194                 writeq(val64, &bar0->tx_w_round_robin_1);
1195                 val64 = 0x0203000100000102ULL;
1196                 writeq(val64, &bar0->tx_w_round_robin_2);
1197                 val64 = 0x0304000102030405ULL;
1198                 writeq(val64, &bar0->tx_w_round_robin_3);
1199                 val64 = 0x0001000200000000ULL;
1200                 writeq(val64, &bar0->tx_w_round_robin_4);
1201                 break;
1202         case 7:
1203                 val64 = 0x0001020001020300ULL;
1204                 writeq(val64, &bar0->tx_w_round_robin_0);
1205                 val64 = 0x0102030400010203ULL;
1206                 writeq(val64, &bar0->tx_w_round_robin_1);
1207                 val64 = 0x0405060001020001ULL;
1208                 writeq(val64, &bar0->tx_w_round_robin_2);
1209                 val64 = 0x0304050000010200ULL;
1210                 writeq(val64, &bar0->tx_w_round_robin_3);
1211                 val64 = 0x0102030000000000ULL;
1212                 writeq(val64, &bar0->tx_w_round_robin_4);
1213                 break;
1214         case 8:
1215                 val64 = 0x0001020300040105ULL;
1216                 writeq(val64, &bar0->tx_w_round_robin_0);
1217                 val64 = 0x0200030106000204ULL;
1218                 writeq(val64, &bar0->tx_w_round_robin_1);
1219                 val64 = 0x0103000502010007ULL;
1220                 writeq(val64, &bar0->tx_w_round_robin_2);
1221                 val64 = 0x0304010002060500ULL;
1222                 writeq(val64, &bar0->tx_w_round_robin_3);
1223                 val64 = 0x0103020400000000ULL;
1224                 writeq(val64, &bar0->tx_w_round_robin_4);
1225                 break;
1226         }
1227
1228         /* Enable all configured Tx FIFO partitions */
1229         val64 = readq(&bar0->tx_fifo_partition_0);
1230         val64 |= (TX_FIFO_PARTITION_EN);
1231         writeq(val64, &bar0->tx_fifo_partition_0);
1232
1233         /* Filling the Rx round robin registers as per the
1234          * number of Rings and steering based on QoS.
1235          */
1236         switch (config->rx_ring_num) {
1237         case 1:
1238                 val64 = 0x8080808080808080ULL;
1239                 writeq(val64, &bar0->rts_qos_steering);
1240                 break;
1241         case 2:
1242                 val64 = 0x0000010000010000ULL;
1243                 writeq(val64, &bar0->rx_w_round_robin_0);
1244                 val64 = 0x0100000100000100ULL;
1245                 writeq(val64, &bar0->rx_w_round_robin_1);
1246                 val64 = 0x0001000001000001ULL;
1247                 writeq(val64, &bar0->rx_w_round_robin_2);
1248                 val64 = 0x0000010000010000ULL;
1249                 writeq(val64, &bar0->rx_w_round_robin_3);
1250                 val64 = 0x0100000000000000ULL;
1251                 writeq(val64, &bar0->rx_w_round_robin_4);
1252
1253                 val64 = 0x8080808040404040ULL;
1254                 writeq(val64, &bar0->rts_qos_steering);
1255                 break;
1256         case 3:
1257                 val64 = 0x0001000102000001ULL;
1258                 writeq(val64, &bar0->rx_w_round_robin_0);
1259                 val64 = 0x0001020000010001ULL;
1260                 writeq(val64, &bar0->rx_w_round_robin_1);
1261                 val64 = 0x0200000100010200ULL;
1262                 writeq(val64, &bar0->rx_w_round_robin_2);
1263                 val64 = 0x0001000102000001ULL;
1264                 writeq(val64, &bar0->rx_w_round_robin_3);
1265                 val64 = 0x0001020000000000ULL;
1266                 writeq(val64, &bar0->rx_w_round_robin_4);
1267
1268                 val64 = 0x8080804040402020ULL;
1269                 writeq(val64, &bar0->rts_qos_steering);
1270                 break;
1271         case 4:
1272                 val64 = 0x0001020300010200ULL;
1273                 writeq(val64, &bar0->rx_w_round_robin_0);
1274                 val64 = 0x0100000102030001ULL;
1275                 writeq(val64, &bar0->rx_w_round_robin_1);
1276                 val64 = 0x0200010000010203ULL;
1277                 writeq(val64, &bar0->rx_w_round_robin_2);
1278                 val64 = 0x0001020001000001ULL;  
1279                 writeq(val64, &bar0->rx_w_round_robin_3);
1280                 val64 = 0x0203000100000000ULL;
1281                 writeq(val64, &bar0->rx_w_round_robin_4);
1282
1283                 val64 = 0x8080404020201010ULL;
1284                 writeq(val64, &bar0->rts_qos_steering);
1285                 break;
1286         case 5:
1287                 val64 = 0x0001000203000102ULL;
1288                 writeq(val64, &bar0->rx_w_round_robin_0);
1289                 val64 = 0x0001020001030004ULL;
1290                 writeq(val64, &bar0->rx_w_round_robin_1);
1291                 val64 = 0x0001000203000102ULL;
1292                 writeq(val64, &bar0->rx_w_round_robin_2);
1293                 val64 = 0x0001020001030004ULL;
1294                 writeq(val64, &bar0->rx_w_round_robin_3);
1295                 val64 = 0x0001000000000000ULL;
1296                 writeq(val64, &bar0->rx_w_round_robin_4);
1297
1298                 val64 = 0x8080404020201008ULL;
1299                 writeq(val64, &bar0->rts_qos_steering);
1300                 break;
1301         case 6:
1302                 val64 = 0x0001020304000102ULL;
1303                 writeq(val64, &bar0->rx_w_round_robin_0);
1304                 val64 = 0x0304050001020001ULL;
1305                 writeq(val64, &bar0->rx_w_round_robin_1);
1306                 val64 = 0x0203000100000102ULL;
1307                 writeq(val64, &bar0->rx_w_round_robin_2);
1308                 val64 = 0x0304000102030405ULL;
1309                 writeq(val64, &bar0->rx_w_round_robin_3);
1310                 val64 = 0x0001000200000000ULL;
1311                 writeq(val64, &bar0->rx_w_round_robin_4);
1312
1313                 val64 = 0x8080404020100804ULL;
1314                 writeq(val64, &bar0->rts_qos_steering);
1315                 break;
1316         case 7:
1317                 val64 = 0x0001020001020300ULL;
1318                 writeq(val64, &bar0->rx_w_round_robin_0);
1319                 val64 = 0x0102030400010203ULL;
1320                 writeq(val64, &bar0->rx_w_round_robin_1);
1321                 val64 = 0x0405060001020001ULL;
1322                 writeq(val64, &bar0->rx_w_round_robin_2);
1323                 val64 = 0x0304050000010200ULL;
1324                 writeq(val64, &bar0->rx_w_round_robin_3);
1325                 val64 = 0x0102030000000000ULL;
1326                 writeq(val64, &bar0->rx_w_round_robin_4);
1327
1328                 val64 = 0x8080402010080402ULL;
1329                 writeq(val64, &bar0->rts_qos_steering);
1330                 break;
1331         case 8:
1332                 val64 = 0x0001020300040105ULL;
1333                 writeq(val64, &bar0->rx_w_round_robin_0);
1334                 val64 = 0x0200030106000204ULL;
1335                 writeq(val64, &bar0->rx_w_round_robin_1);
1336                 val64 = 0x0103000502010007ULL;
1337                 writeq(val64, &bar0->rx_w_round_robin_2);
1338                 val64 = 0x0304010002060500ULL;
1339                 writeq(val64, &bar0->rx_w_round_robin_3);
1340                 val64 = 0x0103020400000000ULL;
1341                 writeq(val64, &bar0->rx_w_round_robin_4);
1342
1343                 val64 = 0x8040201008040201ULL;
1344                 writeq(val64, &bar0->rts_qos_steering);
1345                 break;
1346         }
1347
1348         /* UDP Fix */
1349         val64 = 0;
1350         for (i = 0; i < 8; i++)
1351                 writeq(val64, &bar0->rts_frm_len_n[i]);
1352
1353         /* Set the default rts frame length for the rings configured */
1354         val64 = MAC_RTS_FRM_LEN_SET(dev->mtu+22);
1355         for (i = 0 ; i < config->rx_ring_num ; i++)
1356                 writeq(val64, &bar0->rts_frm_len_n[i]);
1357
1358         /* Set the frame length for the configured rings
1359          * desired by the user
1360          */
1361         for (i = 0; i < config->rx_ring_num; i++) {
1362                 /* If rts_frm_len[i] == 0 then it is assumed that user not
1363                  * specified frame length steering.
1364                  * If the user provides the frame length then program
1365                  * the rts_frm_len register for those values or else
1366                  * leave it as it is.
1367                  */
1368                 if (rts_frm_len[i] != 0) {
1369                         writeq(MAC_RTS_FRM_LEN_SET(rts_frm_len[i]),
1370                                 &bar0->rts_frm_len_n[i]);
1371                 }
1372         }
1373
1374         /* Program statistics memory */
1375         writeq(mac_control->stats_mem_phy, &bar0->stat_addr);
1376
1377         if (nic->device_type == XFRAME_II_DEVICE) {
1378                 val64 = STAT_BC(0x320);
1379                 writeq(val64, &bar0->stat_byte_cnt);
1380         }
1381
1382         /*
1383          * Initializing the sampling rate for the device to calculate the
1384          * bandwidth utilization.
1385          */
1386         val64 = MAC_TX_LINK_UTIL_VAL(tmac_util_period) |
1387             MAC_RX_LINK_UTIL_VAL(rmac_util_period);
1388         writeq(val64, &bar0->mac_link_util);
1389
1390
1391         /*
1392          * Initializing the Transmit and Receive Traffic Interrupt
1393          * Scheme.
1394          */
1395         /*
1396          * TTI Initialization. Default Tx timer gets us about
1397          * 250 interrupts per sec. Continuous interrupts are enabled
1398          * by default.
1399          */
1400         if (nic->device_type == XFRAME_II_DEVICE) {
1401                 int count = (nic->config.bus_speed * 125)/2;
1402                 val64 = TTI_DATA1_MEM_TX_TIMER_VAL(count);
1403         } else {
1404
1405                 val64 = TTI_DATA1_MEM_TX_TIMER_VAL(0x2078);
1406         }
1407         val64 |= TTI_DATA1_MEM_TX_URNG_A(0xA) |
1408             TTI_DATA1_MEM_TX_URNG_B(0x10) |
1409             TTI_DATA1_MEM_TX_URNG_C(0x30) | TTI_DATA1_MEM_TX_TIMER_AC_EN;
1410                 if (use_continuous_tx_intrs)
1411                         val64 |= TTI_DATA1_MEM_TX_TIMER_CI_EN;
1412         writeq(val64, &bar0->tti_data1_mem);
1413
1414         val64 = TTI_DATA2_MEM_TX_UFC_A(0x10) |
1415             TTI_DATA2_MEM_TX_UFC_B(0x20) |
1416             TTI_DATA2_MEM_TX_UFC_C(0x70) | TTI_DATA2_MEM_TX_UFC_D(0x80);
1417         writeq(val64, &bar0->tti_data2_mem);
1418
1419         val64 = TTI_CMD_MEM_WE | TTI_CMD_MEM_STROBE_NEW_CMD;
1420         writeq(val64, &bar0->tti_command_mem);
1421
1422         /*
1423          * Once the operation completes, the Strobe bit of the command
1424          * register will be reset. We poll for this particular condition
1425          * We wait for a maximum of 500ms for the operation to complete,
1426          * if it's not complete by then we return error.
1427          */
1428         time = 0;
1429         while (TRUE) {
1430                 val64 = readq(&bar0->tti_command_mem);
1431                 if (!(val64 & TTI_CMD_MEM_STROBE_NEW_CMD)) {
1432                         break;
1433                 }
1434                 if (time > 10) {
1435                         DBG_PRINT(ERR_DBG, "%s: TTI init Failed\n",
1436                                   dev->name);
1437                         return -1;
1438                 }
1439                 msleep(50);
1440                 time++;
1441         }
1442
1443         if (nic->config.bimodal) {
1444                 int k = 0;
1445                 for (k = 0; k < config->rx_ring_num; k++) {
1446                         val64 = TTI_CMD_MEM_WE | TTI_CMD_MEM_STROBE_NEW_CMD;
1447                         val64 |= TTI_CMD_MEM_OFFSET(0x38+k);
1448                         writeq(val64, &bar0->tti_command_mem);
1449
1450                 /*
1451                  * Once the operation completes, the Strobe bit of the command
1452                  * register will be reset. We poll for this particular condition
1453                  * We wait for a maximum of 500ms for the operation to complete,
1454                  * if it's not complete by then we return error.
1455                 */
1456                         time = 0;
1457                         while (TRUE) {
1458                                 val64 = readq(&bar0->tti_command_mem);
1459                                 if (!(val64 & TTI_CMD_MEM_STROBE_NEW_CMD)) {
1460                                         break;
1461                                 }
1462                                 if (time > 10) {
1463                                         DBG_PRINT(ERR_DBG,
1464                                                 "%s: TTI init Failed\n",
1465                                         dev->name);
1466                                         return -1;
1467                                 }
1468                                 time++;
1469                                 msleep(50);
1470                         }
1471                 }
1472         } else {
1473
1474                 /* RTI Initialization */
1475                 if (nic->device_type == XFRAME_II_DEVICE) {
1476                         /*
1477                          * Programmed to generate Apprx 500 Intrs per
1478                          * second
1479                          */
1480                         int count = (nic->config.bus_speed * 125)/4;
1481                         val64 = RTI_DATA1_MEM_RX_TIMER_VAL(count);
1482                 } else {
1483                         val64 = RTI_DATA1_MEM_RX_TIMER_VAL(0xFFF);
1484                 }
1485                 val64 |= RTI_DATA1_MEM_RX_URNG_A(0xA) |
1486                     RTI_DATA1_MEM_RX_URNG_B(0x10) |
1487                     RTI_DATA1_MEM_RX_URNG_C(0x30) | RTI_DATA1_MEM_RX_TIMER_AC_EN;
1488
1489                 writeq(val64, &bar0->rti_data1_mem);
1490
1491                 val64 = RTI_DATA2_MEM_RX_UFC_A(0x1) |
1492                     RTI_DATA2_MEM_RX_UFC_B(0x2) ;
1493                 if (nic->intr_type == MSI_X)
1494                     val64 |= (RTI_DATA2_MEM_RX_UFC_C(0x20) | \
1495                                 RTI_DATA2_MEM_RX_UFC_D(0x40));
1496                 else
1497                     val64 |= (RTI_DATA2_MEM_RX_UFC_C(0x40) | \
1498                                 RTI_DATA2_MEM_RX_UFC_D(0x80));
1499                 writeq(val64, &bar0->rti_data2_mem);
1500
1501                 for (i = 0; i < config->rx_ring_num; i++) {
1502                         val64 = RTI_CMD_MEM_WE | RTI_CMD_MEM_STROBE_NEW_CMD
1503                                         | RTI_CMD_MEM_OFFSET(i);
1504                         writeq(val64, &bar0->rti_command_mem);
1505
1506                         /*
1507                          * Once the operation completes, the Strobe bit of the
1508                          * command register will be reset. We poll for this
1509                          * particular condition. We wait for a maximum of 500ms
1510                          * for the operation to complete, if it's not complete
1511                          * by then we return error.
1512                          */
1513                         time = 0;
1514                         while (TRUE) {
1515                                 val64 = readq(&bar0->rti_command_mem);
1516                                 if (!(val64 & RTI_CMD_MEM_STROBE_NEW_CMD)) {
1517                                         break;
1518                                 }
1519                                 if (time > 10) {
1520                                         DBG_PRINT(ERR_DBG, "%s: RTI init Failed\n",
1521                                                   dev->name);
1522                                         return -1;
1523                                 }
1524                                 time++;
1525                                 msleep(50);
1526                         }
1527                 }
1528         }
1529
1530         /*
1531          * Initializing proper values as Pause threshold into all
1532          * the 8 Queues on Rx side.
1533          */
1534         writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q0q3);
1535         writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q4q7);
1536
1537         /* Disable RMAC PAD STRIPPING */
1538         add = &bar0->mac_cfg;
1539         val64 = readq(&bar0->mac_cfg);
1540         val64 &= ~(MAC_CFG_RMAC_STRIP_PAD);
1541         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1542         writel((u32) (val64), add);
1543         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1544         writel((u32) (val64 >> 32), (add + 4));
1545         val64 = readq(&bar0->mac_cfg);
1546
1547         /* Enable FCS stripping by adapter */
1548         add = &bar0->mac_cfg;
1549         val64 = readq(&bar0->mac_cfg);
1550         val64 |= MAC_CFG_RMAC_STRIP_FCS;
1551         if (nic->device_type == XFRAME_II_DEVICE)
1552                 writeq(val64, &bar0->mac_cfg);
1553         else {
1554                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1555                 writel((u32) (val64), add);
1556                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1557                 writel((u32) (val64 >> 32), (add + 4));
1558         }
1559
1560         /*
1561          * Set the time value to be inserted in the pause frame
1562          * generated by xena.
1563          */
1564         val64 = readq(&bar0->rmac_pause_cfg);
1565         val64 &= ~(RMAC_PAUSE_HG_PTIME(0xffff));
1566         val64 |= RMAC_PAUSE_HG_PTIME(nic->mac_control.rmac_pause_time);
1567         writeq(val64, &bar0->rmac_pause_cfg);
1568
1569         /*
1570          * Set the Threshold Limit for Generating the pause frame
1571          * If the amount of data in any Queue exceeds ratio of
1572          * (mac_control.mc_pause_threshold_q0q3 or q4q7)/256
1573          * pause frame is generated
1574          */
1575         val64 = 0;
1576         for (i = 0; i < 4; i++) {
1577                 val64 |=
1578                     (((u64) 0xFF00 | nic->mac_control.
1579                       mc_pause_threshold_q0q3)
1580                      << (i * 2 * 8));
1581         }
1582         writeq(val64, &bar0->mc_pause_thresh_q0q3);
1583
1584         val64 = 0;
1585         for (i = 0; i < 4; i++) {
1586                 val64 |=
1587                     (((u64) 0xFF00 | nic->mac_control.
1588                       mc_pause_threshold_q4q7)
1589                      << (i * 2 * 8));
1590         }
1591         writeq(val64, &bar0->mc_pause_thresh_q4q7);
1592
1593         /*
1594          * TxDMA will stop Read request if the number of read split has
1595          * exceeded the limit pointed by shared_splits
1596          */
1597         val64 = readq(&bar0->pic_control);
1598         val64 |= PIC_CNTL_SHARED_SPLITS(shared_splits);
1599         writeq(val64, &bar0->pic_control);
1600
1601         if (nic->config.bus_speed == 266) {
1602                 writeq(TXREQTO_VAL(0x7f) | TXREQTO_EN, &bar0->txreqtimeout);
1603                 writeq(0x0, &bar0->read_retry_delay);
1604                 writeq(0x0, &bar0->write_retry_delay);
1605         }
1606
1607         /*
1608          * Programming the Herc to split every write transaction
1609          * that does not start on an ADB to reduce disconnects.
1610          */
1611         if (nic->device_type == XFRAME_II_DEVICE) {
1612                 val64 = EXT_REQ_EN | MISC_LINK_STABILITY_PRD(3);
1613                 writeq(val64, &bar0->misc_control);
1614                 val64 = readq(&bar0->pic_control2);
1615                 val64 &= ~(BIT(13)|BIT(14)|BIT(15));
1616                 writeq(val64, &bar0->pic_control2);
1617         }
1618         if (strstr(nic->product_name, "CX4")) {
1619                 val64 = TMAC_AVG_IPG(0x17);
1620                 writeq(val64, &bar0->tmac_avg_ipg);
1621         }
1622
1623         return SUCCESS;
1624 }
1625 #define LINK_UP_DOWN_INTERRUPT          1
1626 #define MAC_RMAC_ERR_TIMER              2
1627
1628 static int s2io_link_fault_indication(nic_t *nic)
1629 {
1630         if (nic->intr_type != INTA)
1631                 return MAC_RMAC_ERR_TIMER;
1632         if (nic->device_type == XFRAME_II_DEVICE)
1633                 return LINK_UP_DOWN_INTERRUPT;
1634         else
1635                 return MAC_RMAC_ERR_TIMER;
1636 }
1637
1638 /**
1639  *  en_dis_able_nic_intrs - Enable or Disable the interrupts
1640  *  @nic: device private variable,
1641  *  @mask: A mask indicating which Intr block must be modified and,
1642  *  @flag: A flag indicating whether to enable or disable the Intrs.
1643  *  Description: This function will either disable or enable the interrupts
1644  *  depending on the flag argument. The mask argument can be used to
1645  *  enable/disable any Intr block.
1646  *  Return Value: NONE.
1647  */
1648
1649 static void en_dis_able_nic_intrs(struct s2io_nic *nic, u16 mask, int flag)
1650 {
1651         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1652         register u64 val64 = 0, temp64 = 0;
1653
1654         /*  Top level interrupt classification */
1655         /*  PIC Interrupts */
1656         if ((mask & (TX_PIC_INTR | RX_PIC_INTR))) {
1657                 /*  Enable PIC Intrs in the general intr mask register */
1658                 val64 = TXPIC_INT_M | PIC_RX_INT_M;
1659                 if (flag == ENABLE_INTRS) {
1660                         temp64 = readq(&bar0->general_int_mask);
1661                         temp64 &= ~((u64) val64);
1662                         writeq(temp64, &bar0->general_int_mask);
1663                         /*
1664                          * If Hercules adapter enable GPIO otherwise
1665                          * disable all PCIX, Flash, MDIO, IIC and GPIO
1666                          * interrupts for now.
1667                          * TODO
1668                          */
1669                         if (s2io_link_fault_indication(nic) ==
1670                                         LINK_UP_DOWN_INTERRUPT ) {
1671                                 temp64 = readq(&bar0->pic_int_mask);
1672                                 temp64 &= ~((u64) PIC_INT_GPIO);
1673                                 writeq(temp64, &bar0->pic_int_mask);
1674                                 temp64 = readq(&bar0->gpio_int_mask);
1675                                 temp64 &= ~((u64) GPIO_INT_MASK_LINK_UP);
1676                                 writeq(temp64, &bar0->gpio_int_mask);
1677                         } else {
1678                                 writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
1679                         }
1680                         /*
1681                          * No MSI Support is available presently, so TTI and
1682                          * RTI interrupts are also disabled.
1683                          */
1684                 } else if (flag == DISABLE_INTRS) {
1685                         /*
1686                          * Disable PIC Intrs in the general
1687                          * intr mask register
1688                          */
1689                         writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
1690                         temp64 = readq(&bar0->general_int_mask);
1691                         val64 |= temp64;
1692                         writeq(val64, &bar0->general_int_mask);
1693                 }
1694         }
1695
1696         /*  DMA Interrupts */
1697         /*  Enabling/Disabling Tx DMA interrupts */
1698         if (mask & TX_DMA_INTR) {
1699                 /* Enable TxDMA Intrs in the general intr mask register */
1700                 val64 = TXDMA_INT_M;
1701                 if (flag == ENABLE_INTRS) {
1702                         temp64 = readq(&bar0->general_int_mask);
1703                         temp64 &= ~((u64) val64);
1704                         writeq(temp64, &bar0->general_int_mask);
1705                         /*
1706                          * Keep all interrupts other than PFC interrupt
1707                          * and PCC interrupt disabled in DMA level.
1708                          */
1709                         val64 = DISABLE_ALL_INTRS & ~(TXDMA_PFC_INT_M |
1710                                                       TXDMA_PCC_INT_M);
1711                         writeq(val64, &bar0->txdma_int_mask);
1712                         /*
1713                          * Enable only the MISC error 1 interrupt in PFC block
1714                          */
1715                         val64 = DISABLE_ALL_INTRS & (~PFC_MISC_ERR_1);
1716                         writeq(val64, &bar0->pfc_err_mask);
1717                         /*
1718                          * Enable only the FB_ECC error interrupt in PCC block
1719                          */
1720                         val64 = DISABLE_ALL_INTRS & (~PCC_FB_ECC_ERR);
1721                         writeq(val64, &bar0->pcc_err_mask);
1722                 } else if (flag == DISABLE_INTRS) {
1723                         /*
1724                          * Disable TxDMA Intrs in the general intr mask
1725                          * register
1726                          */
1727                         writeq(DISABLE_ALL_INTRS, &bar0->txdma_int_mask);
1728                         writeq(DISABLE_ALL_INTRS, &bar0->pfc_err_mask);
1729                         temp64 = readq(&bar0->general_int_mask);
1730                         val64 |= temp64;
1731                         writeq(val64, &bar0->general_int_mask);
1732                 }
1733         }
1734
1735         /*  Enabling/Disabling Rx DMA interrupts */
1736         if (mask & RX_DMA_INTR) {
1737                 /*  Enable RxDMA Intrs in the general intr mask register */
1738                 val64 = RXDMA_INT_M;
1739                 if (flag == ENABLE_INTRS) {
1740                         temp64 = readq(&bar0->general_int_mask);
1741                         temp64 &= ~((u64) val64);
1742                         writeq(temp64, &bar0->general_int_mask);
1743                         /*
1744                          * All RxDMA block interrupts are disabled for now
1745                          * TODO
1746                          */
1747                         writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask);
1748                 } else if (flag == DISABLE_INTRS) {
1749                         /*
1750                          * Disable RxDMA Intrs in the general intr mask
1751                          * register
1752                          */
1753                         writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask);
1754                         temp64 = readq(&bar0->general_int_mask);
1755                         val64 |= temp64;
1756                         writeq(val64, &bar0->general_int_mask);
1757                 }
1758         }
1759
1760         /*  MAC Interrupts */
1761         /*  Enabling/Disabling MAC interrupts */
1762         if (mask & (TX_MAC_INTR | RX_MAC_INTR)) {
1763                 val64 = TXMAC_INT_M | RXMAC_INT_M;
1764                 if (flag == ENABLE_INTRS) {
1765                         temp64 = readq(&bar0->general_int_mask);
1766                         temp64 &= ~((u64) val64);
1767                         writeq(temp64, &bar0->general_int_mask);
1768                         /*
1769                          * All MAC block error interrupts are disabled for now
1770                          * TODO
1771                          */
1772                 } else if (flag == DISABLE_INTRS) {
1773                         /*
1774                          * Disable MAC Intrs in the general intr mask register
1775                          */
1776                         writeq(DISABLE_ALL_INTRS, &bar0->mac_int_mask);
1777                         writeq(DISABLE_ALL_INTRS,
1778                                &bar0->mac_rmac_err_mask);
1779
1780                         temp64 = readq(&bar0->general_int_mask);
1781                         val64 |= temp64;
1782                         writeq(val64, &bar0->general_int_mask);
1783                 }
1784         }
1785
1786         /*  XGXS Interrupts */
1787         if (mask & (TX_XGXS_INTR | RX_XGXS_INTR)) {
1788                 val64 = TXXGXS_INT_M | RXXGXS_INT_M;
1789                 if (flag == ENABLE_INTRS) {
1790                         temp64 = readq(&bar0->general_int_mask);
1791                         temp64 &= ~((u64) val64);
1792                         writeq(temp64, &bar0->general_int_mask);
1793                         /*
1794                          * All XGXS block error interrupts are disabled for now
1795                          * TODO
1796                          */
1797                         writeq(DISABLE_ALL_INTRS, &bar0->xgxs_int_mask);
1798                 } else if (flag == DISABLE_INTRS) {
1799                         /*
1800                          * Disable MC Intrs in the general intr mask register
1801                          */
1802                         writeq(DISABLE_ALL_INTRS, &bar0->xgxs_int_mask);
1803                         temp64 = readq(&bar0->general_int_mask);
1804                         val64 |= temp64;
1805                         writeq(val64, &bar0->general_int_mask);
1806                 }
1807         }
1808
1809         /*  Memory Controller(MC) interrupts */
1810         if (mask & MC_INTR) {
1811                 val64 = MC_INT_M;
1812                 if (flag == ENABLE_INTRS) {
1813                         temp64 = readq(&bar0->general_int_mask);
1814                         temp64 &= ~((u64) val64);
1815                         writeq(temp64, &bar0->general_int_mask);
1816                         /*
1817                          * Enable all MC Intrs.
1818                          */
1819                         writeq(0x0, &bar0->mc_int_mask);
1820                         writeq(0x0, &bar0->mc_err_mask);
1821                 } else if (flag == DISABLE_INTRS) {
1822                         /*
1823                          * Disable MC Intrs in the general intr mask register
1824                          */
1825                         writeq(DISABLE_ALL_INTRS, &bar0->mc_int_mask);
1826                         temp64 = readq(&bar0->general_int_mask);
1827                         val64 |= temp64;
1828                         writeq(val64, &bar0->general_int_mask);
1829                 }
1830         }
1831
1832
1833         /*  Tx traffic interrupts */
1834         if (mask & TX_TRAFFIC_INTR) {
1835                 val64 = TXTRAFFIC_INT_M;
1836                 if (flag == ENABLE_INTRS) {
1837                         temp64 = readq(&bar0->general_int_mask);
1838                         temp64 &= ~((u64) val64);
1839                         writeq(temp64, &bar0->general_int_mask);
1840                         /*
1841                          * Enable all the Tx side interrupts
1842                          * writing 0 Enables all 64 TX interrupt levels
1843                          */
1844                         writeq(0x0, &bar0->tx_traffic_mask);
1845                 } else if (flag == DISABLE_INTRS) {
1846                         /*
1847                          * Disable Tx Traffic Intrs in the general intr mask
1848                          * register.
1849                          */
1850                         writeq(DISABLE_ALL_INTRS, &bar0->tx_traffic_mask);
1851                         temp64 = readq(&bar0->general_int_mask);
1852                         val64 |= temp64;
1853                         writeq(val64, &bar0->general_int_mask);
1854                 }
1855         }
1856
1857         /*  Rx traffic interrupts */
1858         if (mask & RX_TRAFFIC_INTR) {
1859                 val64 = RXTRAFFIC_INT_M;
1860                 if (flag == ENABLE_INTRS) {
1861                         temp64 = readq(&bar0->general_int_mask);
1862                         temp64 &= ~((u64) val64);
1863                         writeq(temp64, &bar0->general_int_mask);
1864                         /* writing 0 Enables all 8 RX interrupt levels */
1865                         writeq(0x0, &bar0->rx_traffic_mask);
1866                 } else if (flag == DISABLE_INTRS) {
1867                         /*
1868                          * Disable Rx Traffic Intrs in the general intr mask
1869                          * register.
1870                          */
1871                         writeq(DISABLE_ALL_INTRS, &bar0->rx_traffic_mask);
1872                         temp64 = readq(&bar0->general_int_mask);
1873                         val64 |= temp64;
1874                         writeq(val64, &bar0->general_int_mask);
1875                 }
1876         }
1877 }
1878
1879 static int check_prc_pcc_state(u64 val64, int flag, int rev_id, int herc)
1880 {
1881         int ret = 0;
1882
1883         if (flag == FALSE) {
1884                 if ((!herc && (rev_id >= 4)) || herc) {
1885                         if (!(val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) &&
1886                             ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1887                              ADAPTER_STATUS_RC_PRC_QUIESCENT)) {
1888                                 ret = 1;
1889                         }
1890                 }else {
1891                         if (!(val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) &&
1892                             ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1893                              ADAPTER_STATUS_RC_PRC_QUIESCENT)) {
1894                                 ret = 1;
1895                         }
1896                 }
1897         } else {
1898                 if ((!herc && (rev_id >= 4)) || herc) {
1899                         if (((val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) ==
1900                              ADAPTER_STATUS_RMAC_PCC_IDLE) &&
1901                             (!(val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ||
1902                              ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1903                               ADAPTER_STATUS_RC_PRC_QUIESCENT))) {
1904                                 ret = 1;
1905                         }
1906                 } else {
1907                         if (((val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) ==
1908                              ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) &&
1909                             (!(val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ||
1910                              ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1911                               ADAPTER_STATUS_RC_PRC_QUIESCENT))) {
1912                                 ret = 1;
1913                         }
1914                 }
1915         }
1916
1917         return ret;
1918 }
1919 /**
1920  *  verify_xena_quiescence - Checks whether the H/W is ready
1921  *  @val64 :  Value read from adapter status register.
1922  *  @flag : indicates if the adapter enable bit was ever written once
1923  *  before.
1924  *  Description: Returns whether the H/W is ready to go or not. Depending
1925  *  on whether adapter enable bit was written or not the comparison
1926  *  differs and the calling function passes the input argument flag to
1927  *  indicate this.
1928  *  Return: 1 If xena is quiescence
1929  *          0 If Xena is not quiescence
1930  */
1931
1932 static int verify_xena_quiescence(nic_t *sp, u64 val64, int flag)
1933 {
1934         int ret = 0, herc;
1935         u64 tmp64 = ~((u64) val64);
1936         int rev_id = get_xena_rev_id(sp->pdev);
1937
1938         herc = (sp->device_type == XFRAME_II_DEVICE);
1939         if (!
1940             (tmp64 &
1941              (ADAPTER_STATUS_TDMA_READY | ADAPTER_STATUS_RDMA_READY |
1942               ADAPTER_STATUS_PFC_READY | ADAPTER_STATUS_TMAC_BUF_EMPTY |
1943               ADAPTER_STATUS_PIC_QUIESCENT | ADAPTER_STATUS_MC_DRAM_READY |
1944               ADAPTER_STATUS_MC_QUEUES_READY | ADAPTER_STATUS_M_PLL_LOCK |
1945               ADAPTER_STATUS_P_PLL_LOCK))) {
1946                 ret = check_prc_pcc_state(val64, flag, rev_id, herc);
1947         }
1948
1949         return ret;
1950 }
1951
1952 /**
1953  * fix_mac_address -  Fix for Mac addr problem on Alpha platforms
1954  * @sp: Pointer to device specifc structure
1955  * Description :
1956  * New procedure to clear mac address reading  problems on Alpha platforms
1957  *
1958  */
1959
1960 static void fix_mac_address(nic_t * sp)
1961 {
1962         XENA_dev_config_t __iomem *bar0 = sp->bar0;
1963         u64 val64;
1964         int i = 0;
1965
1966         while (fix_mac[i] != END_SIGN) {
1967                 writeq(fix_mac[i++], &bar0->gpio_control);
1968                 udelay(10);
1969                 val64 = readq(&bar0->gpio_control);
1970         }
1971 }
1972
1973 /**
1974  *  start_nic - Turns the device on
1975  *  @nic : device private variable.
1976  *  Description:
1977  *  This function actually turns the device on. Before this  function is
1978  *  called,all Registers are configured from their reset states
1979  *  and shared memory is allocated but the NIC is still quiescent. On
1980  *  calling this function, the device interrupts are cleared and the NIC is
1981  *  literally switched on by writing into the adapter control register.
1982  *  Return Value:
1983  *  SUCCESS on success and -1 on failure.
1984  */
1985
1986 static int start_nic(struct s2io_nic *nic)
1987 {
1988         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1989         struct net_device *dev = nic->dev;
1990         register u64 val64 = 0;
1991         u16 subid, i;
1992         mac_info_t *mac_control;
1993         struct config_param *config;
1994
1995         mac_control = &nic->mac_control;
1996         config = &nic->config;
1997
1998         /*  PRC Initialization and configuration */
1999         for (i = 0; i < config->rx_ring_num; i++) {
2000                 writeq((u64) mac_control->rings[i].rx_blocks[0].block_dma_addr,
2001                        &bar0->prc_rxd0_n[i]);
2002
2003                 val64 = readq(&bar0->prc_ctrl_n[i]);
2004                 if (nic->config.bimodal)
2005                         val64 |= PRC_CTRL_BIMODAL_INTERRUPT;
2006                 if (nic->rxd_mode == RXD_MODE_1)
2007                         val64 |= PRC_CTRL_RC_ENABLED;
2008                 else
2009                         val64 |= PRC_CTRL_RC_ENABLED | PRC_CTRL_RING_MODE_3;
2010                 if (nic->device_type == XFRAME_II_DEVICE)
2011                         val64 |= PRC_CTRL_GROUP_READS;
2012                 val64 &= ~PRC_CTRL_RXD_BACKOFF_INTERVAL(0xFFFFFF);
2013                 val64 |= PRC_CTRL_RXD_BACKOFF_INTERVAL(0x1000);
2014                 writeq(val64, &bar0->prc_ctrl_n[i]);
2015         }
2016
2017         if (nic->rxd_mode == RXD_MODE_3B) {
2018                 /* Enabling 2 buffer mode by writing into Rx_pa_cfg reg. */
2019                 val64 = readq(&bar0->rx_pa_cfg);
2020                 val64 |= RX_PA_CFG_IGNORE_L2_ERR;
2021                 writeq(val64, &bar0->rx_pa_cfg);
2022         }
2023
2024         /*
2025          * Enabling MC-RLDRAM. After enabling the device, we timeout
2026          * for around 100ms, which is approximately the time required
2027          * for the device to be ready for operation.
2028          */
2029         val64 = readq(&bar0->mc_rldram_mrs);
2030         val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE | MC_RLDRAM_MRS_ENABLE;
2031         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
2032         val64 = readq(&bar0->mc_rldram_mrs);
2033
2034         msleep(100);    /* Delay by around 100 ms. */
2035
2036         /* Enabling ECC Protection. */
2037         val64 = readq(&bar0->adapter_control);
2038         val64 &= ~ADAPTER_ECC_EN;
2039         writeq(val64, &bar0->adapter_control);
2040
2041         /*
2042          * Clearing any possible Link state change interrupts that
2043          * could have popped up just before Enabling the card.
2044          */
2045         val64 = readq(&bar0->mac_rmac_err_reg);
2046         if (val64)
2047                 writeq(val64, &bar0->mac_rmac_err_reg);
2048
2049         /*
2050          * Verify if the device is ready to be enabled, if so enable
2051          * it.
2052          */
2053         val64 = readq(&bar0->adapter_status);
2054         if (!verify_xena_quiescence(nic, val64, nic->device_enabled_once)) {
2055                 DBG_PRINT(ERR_DBG, "%s: device is not ready, ", dev->name);
2056                 DBG_PRINT(ERR_DBG, "Adapter status reads: 0x%llx\n",
2057                           (unsigned long long) val64);
2058                 return FAILURE;
2059         }
2060
2061         /*
2062          * With some switches, link might be already up at this point.
2063          * Because of this weird behavior, when we enable laser,
2064          * we may not get link. We need to handle this. We cannot
2065          * figure out which switch is misbehaving. So we are forced to
2066          * make a global change.
2067          */
2068
2069         /* Enabling Laser. */
2070         val64 = readq(&bar0->adapter_control);
2071         val64 |= ADAPTER_EOI_TX_ON;
2072         writeq(val64, &bar0->adapter_control);
2073
2074         if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
2075                 /*
2076                  * Dont see link state interrupts initally on some switches,
2077                  * so directly scheduling the link state task here.
2078                  */
2079                 schedule_work(&nic->set_link_task);
2080         }
2081         /* SXE-002: Initialize link and activity LED */
2082         subid = nic->pdev->subsystem_device;
2083         if (((subid & 0xFF) >= 0x07) &&
2084             (nic->device_type == XFRAME_I_DEVICE)) {
2085                 val64 = readq(&bar0->gpio_control);
2086                 val64 |= 0x0000800000000000ULL;
2087                 writeq(val64, &bar0->gpio_control);
2088                 val64 = 0x0411040400000000ULL;
2089                 writeq(val64, (void __iomem *)bar0 + 0x2700);
2090         }
2091
2092         return SUCCESS;
2093 }
2094 /**
2095  * s2io_txdl_getskb - Get the skb from txdl, unmap and return skb
2096  */
2097 static struct sk_buff *s2io_txdl_getskb(fifo_info_t *fifo_data, TxD_t *txdlp, int get_off)
2098 {
2099         nic_t *nic = fifo_data->nic;
2100         struct sk_buff *skb;
2101         TxD_t *txds;
2102         u16 j, frg_cnt;
2103
2104         txds = txdlp;
2105         if (txds->Host_Control == (u64)(long)nic->ufo_in_band_v) {
2106                 pci_unmap_single(nic->pdev, (dma_addr_t)
2107                         txds->Buffer_Pointer, sizeof(u64),
2108                         PCI_DMA_TODEVICE);
2109                 txds++;
2110         }
2111
2112         skb = (struct sk_buff *) ((unsigned long)
2113                         txds->Host_Control);
2114         if (!skb) {
2115                 memset(txdlp, 0, (sizeof(TxD_t) * fifo_data->max_txds));
2116                 return NULL;
2117         }
2118         pci_unmap_single(nic->pdev, (dma_addr_t)
2119                          txds->Buffer_Pointer,
2120                          skb->len - skb->data_len,
2121                          PCI_DMA_TODEVICE);
2122         frg_cnt = skb_shinfo(skb)->nr_frags;
2123         if (frg_cnt) {
2124                 txds++;
2125                 for (j = 0; j < frg_cnt; j++, txds++) {
2126                         skb_frag_t *frag = &skb_shinfo(skb)->frags[j];
2127                         if (!txds->Buffer_Pointer)
2128                                 break;
2129                         pci_unmap_page(nic->pdev, (dma_addr_t) 
2130                                         txds->Buffer_Pointer,
2131                                        frag->size, PCI_DMA_TODEVICE);
2132                 }
2133         }
2134         memset(txdlp,0, (sizeof(TxD_t) * fifo_data->max_txds));
2135         return(skb);
2136 }
2137
2138 /**
2139  *  free_tx_buffers - Free all queued Tx buffers
2140  *  @nic : device private variable.
2141  *  Description:
2142  *  Free all queued Tx buffers.
2143  *  Return Value: void
2144 */
2145
2146 static void free_tx_buffers(struct s2io_nic *nic)
2147 {
2148         struct net_device *dev = nic->dev;
2149         struct sk_buff *skb;
2150         TxD_t *txdp;
2151         int i, j;
2152         mac_info_t *mac_control;
2153         struct config_param *config;
2154         int cnt = 0;
2155
2156         mac_control = &nic->mac_control;
2157         config = &nic->config;
2158
2159         for (i = 0; i < config->tx_fifo_num; i++) {
2160                 for (j = 0; j < config->tx_cfg[i].fifo_len - 1; j++) {
2161                         txdp = (TxD_t *) mac_control->fifos[i].list_info[j].
2162                             list_virt_addr;
2163                         skb = s2io_txdl_getskb(&mac_control->fifos[i], txdp, j);
2164                         if (skb) {
2165                                 dev_kfree_skb(skb);
2166                                 cnt++;
2167                         }
2168                 }
2169                 DBG_PRINT(INTR_DBG,
2170                           "%s:forcibly freeing %d skbs on FIFO%d\n",
2171                           dev->name, cnt, i);
2172                 mac_control->fifos[i].tx_curr_get_info.offset = 0;
2173                 mac_control->fifos[i].tx_curr_put_info.offset = 0;
2174         }
2175 }
2176
2177 /**
2178  *   stop_nic -  To stop the nic
2179  *   @nic ; device private variable.
2180  *   Description:
2181  *   This function does exactly the opposite of what the start_nic()
2182  *   function does. This function is called to stop the device.
2183  *   Return Value:
2184  *   void.
2185  */
2186
2187 static void stop_nic(struct s2io_nic *nic)
2188 {
2189         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2190         register u64 val64 = 0;
2191         u16 interruptible;
2192         mac_info_t *mac_control;
2193         struct config_param *config;
2194
2195         mac_control = &nic->mac_control;
2196         config = &nic->config;
2197
2198         /*  Disable all interrupts */
2199         interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR;
2200         interruptible |= TX_PIC_INTR | RX_PIC_INTR;
2201         interruptible |= TX_MAC_INTR | RX_MAC_INTR;
2202         en_dis_able_nic_intrs(nic, interruptible, DISABLE_INTRS);
2203
2204         /* Clearing Adapter_En bit of ADAPTER_CONTROL Register */
2205         val64 = readq(&bar0->adapter_control);
2206         val64 &= ~(ADAPTER_CNTL_EN);
2207         writeq(val64, &bar0->adapter_control);
2208 }
2209
2210 static int fill_rxd_3buf(nic_t *nic, RxD_t *rxdp, struct sk_buff *skb)
2211 {
2212         struct net_device *dev = nic->dev;
2213         struct sk_buff *frag_list;
2214         void *tmp;
2215
2216         /* Buffer-1 receives L3/L4 headers */
2217         ((RxD3_t*)rxdp)->Buffer1_ptr = pci_map_single
2218                         (nic->pdev, skb->data, l3l4hdr_size + 4,
2219                         PCI_DMA_FROMDEVICE);
2220
2221         /* skb_shinfo(skb)->frag_list will have L4 data payload */
2222         skb_shinfo(skb)->frag_list = dev_alloc_skb(dev->mtu + ALIGN_SIZE);
2223         if (skb_shinfo(skb)->frag_list == NULL) {
2224                 DBG_PRINT(ERR_DBG, "%s: dev_alloc_skb failed\n ", dev->name);
2225                 return -ENOMEM ;
2226         }
2227         frag_list = skb_shinfo(skb)->frag_list;
2228         frag_list->next = NULL;
2229         tmp = (void *)ALIGN((long)frag_list->data, ALIGN_SIZE + 1);
2230         frag_list->data = tmp;
2231         frag_list->tail = tmp;
2232
2233         /* Buffer-2 receives L4 data payload */
2234         ((RxD3_t*)rxdp)->Buffer2_ptr = pci_map_single(nic->pdev,
2235                                 frag_list->data, dev->mtu,
2236                                 PCI_DMA_FROMDEVICE);
2237         rxdp->Control_2 |= SET_BUFFER1_SIZE_3(l3l4hdr_size + 4);
2238         rxdp->Control_2 |= SET_BUFFER2_SIZE_3(dev->mtu);
2239
2240         return SUCCESS;
2241 }
2242
2243 /**
2244  *  fill_rx_buffers - Allocates the Rx side skbs
2245  *  @nic:  device private variable
2246  *  @ring_no: ring number
2247  *  Description:
2248  *  The function allocates Rx side skbs and puts the physical
2249  *  address of these buffers into the RxD buffer pointers, so that the NIC
2250  *  can DMA the received frame into these locations.
2251  *  The NIC supports 3 receive modes, viz
2252  *  1. single buffer,
2253  *  2. three buffer and
2254  *  3. Five buffer modes.
2255  *  Each mode defines how many fragments the received frame will be split
2256  *  up into by the NIC. The frame is split into L3 header, L4 Header,
2257  *  L4 payload in three buffer mode and in 5 buffer mode, L4 payload itself
2258  *  is split into 3 fragments. As of now only single buffer mode is
2259  *  supported.
2260  *   Return Value:
2261  *  SUCCESS on success or an appropriate -ve value on failure.
2262  */
2263
2264 static int fill_rx_buffers(struct s2io_nic *nic, int ring_no)
2265 {
2266         struct net_device *dev = nic->dev;
2267         struct sk_buff *skb;
2268         RxD_t *rxdp;
2269         int off, off1, size, block_no, block_no1;
2270         u32 alloc_tab = 0;
2271         u32 alloc_cnt;
2272         mac_info_t *mac_control;
2273         struct config_param *config;
2274         u64 tmp;
2275         buffAdd_t *ba;
2276 #ifndef CONFIG_S2IO_NAPI
2277         unsigned long flags;
2278 #endif
2279         RxD_t *first_rxdp = NULL;
2280
2281         mac_control = &nic->mac_control;
2282         config = &nic->config;
2283         alloc_cnt = mac_control->rings[ring_no].pkt_cnt -
2284             atomic_read(&nic->rx_bufs_left[ring_no]);
2285
2286         block_no1 = mac_control->rings[ring_no].rx_curr_get_info.block_index;
2287         off1 = mac_control->rings[ring_no].rx_curr_get_info.offset;
2288         while (alloc_tab < alloc_cnt) {
2289                 block_no = mac_control->rings[ring_no].rx_curr_put_info.
2290                     block_index;
2291                 off = mac_control->rings[ring_no].rx_curr_put_info.offset;
2292
2293                 rxdp = mac_control->rings[ring_no].
2294                                 rx_blocks[block_no].rxds[off].virt_addr;
2295
2296                 if ((block_no == block_no1) && (off == off1) &&
2297                                         (rxdp->Host_Control)) {
2298                         DBG_PRINT(INTR_DBG, "%s: Get and Put",
2299                                   dev->name);
2300                         DBG_PRINT(INTR_DBG, " info equated\n");
2301                         goto end;
2302                 }
2303                 if (off && (off == rxd_count[nic->rxd_mode])) {
2304                         mac_control->rings[ring_no].rx_curr_put_info.
2305                             block_index++;
2306                         if (mac_control->rings[ring_no].rx_curr_put_info.
2307                             block_index == mac_control->rings[ring_no].
2308                                         block_count)
2309                                 mac_control->rings[ring_no].rx_curr_put_info.
2310                                         block_index = 0;
2311                         block_no = mac_control->rings[ring_no].
2312                                         rx_curr_put_info.block_index;
2313                         if (off == rxd_count[nic->rxd_mode])
2314                                 off = 0;
2315                         mac_control->rings[ring_no].rx_curr_put_info.
2316                                 offset = off;
2317                         rxdp = mac_control->rings[ring_no].
2318                                 rx_blocks[block_no].block_virt_addr;
2319                         DBG_PRINT(INTR_DBG, "%s: Next block at: %p\n",
2320                                   dev->name, rxdp);
2321                 }
2322 #ifndef CONFIG_S2IO_NAPI
2323                 spin_lock_irqsave(&nic->put_lock, flags);
2324                 mac_control->rings[ring_no].put_pos =
2325                     (block_no * (rxd_count[nic->rxd_mode] + 1)) + off;
2326                 spin_unlock_irqrestore(&nic->put_lock, flags);
2327 #endif
2328                 if ((rxdp->Control_1 & RXD_OWN_XENA) &&
2329                         ((nic->rxd_mode >= RXD_MODE_3A) &&
2330                                 (rxdp->Control_2 & BIT(0)))) {
2331                         mac_control->rings[ring_no].rx_curr_put_info.
2332                                         offset = off;
2333                         goto end;
2334                 }
2335                 /* calculate size of skb based on ring mode */
2336                 size = dev->mtu + HEADER_ETHERNET_II_802_3_SIZE +
2337                                 HEADER_802_2_SIZE + HEADER_SNAP_SIZE;
2338                 if (nic->rxd_mode == RXD_MODE_1)
2339                         size += NET_IP_ALIGN;
2340                 else if (nic->rxd_mode == RXD_MODE_3B)
2341                         size = dev->mtu + ALIGN_SIZE + BUF0_LEN + 4;
2342                 else
2343                         size = l3l4hdr_size + ALIGN_SIZE + BUF0_LEN + 4;
2344
2345                 /* allocate skb */
2346                 skb = dev_alloc_skb(size);
2347                 if(!skb) {
2348                         DBG_PRINT(ERR_DBG, "%s: Out of ", dev->name);
2349                         DBG_PRINT(ERR_DBG, "memory to allocate SKBs\n");
2350                         if (first_rxdp) {
2351                                 wmb();
2352                                 first_rxdp->Control_1 |= RXD_OWN_XENA;
2353                         }
2354                         return -ENOMEM ;
2355                 }
2356                 if (nic->rxd_mode == RXD_MODE_1) {
2357                         /* 1 buffer mode - normal operation mode */
2358                         memset(rxdp, 0, sizeof(RxD1_t));
2359                         skb_reserve(skb, NET_IP_ALIGN);
2360                         ((RxD1_t*)rxdp)->Buffer0_ptr = pci_map_single
2361                             (nic->pdev, skb->data, size - NET_IP_ALIGN,
2362                                 PCI_DMA_FROMDEVICE);
2363                         rxdp->Control_2 = SET_BUFFER0_SIZE_1(size - NET_IP_ALIGN);
2364
2365                 } else if (nic->rxd_mode >= RXD_MODE_3A) {
2366                         /*
2367                          * 2 or 3 buffer mode -
2368                          * Both 2 buffer mode and 3 buffer mode provides 128
2369                          * byte aligned receive buffers.
2370                          *
2371                          * 3 buffer mode provides header separation where in
2372                          * skb->data will have L3/L4 headers where as
2373                          * skb_shinfo(skb)->frag_list will have the L4 data
2374                          * payload
2375                          */
2376
2377                         memset(rxdp, 0, sizeof(RxD3_t));
2378                         ba = &mac_control->rings[ring_no].ba[block_no][off];
2379                         skb_reserve(skb, BUF0_LEN);
2380                         tmp = (u64)(unsigned long) skb->data;
2381                         tmp += ALIGN_SIZE;
2382                         tmp &= ~ALIGN_SIZE;
2383                         skb->data = (void *) (unsigned long)tmp;
2384                         skb->tail = (void *) (unsigned long)tmp;
2385
2386                         if (!(((RxD3_t*)rxdp)->Buffer0_ptr))
2387                                 ((RxD3_t*)rxdp)->Buffer0_ptr =
2388                                    pci_map_single(nic->pdev, ba->ba_0, BUF0_LEN,
2389                                            PCI_DMA_FROMDEVICE);
2390                         else
2391                                 pci_dma_sync_single_for_device(nic->pdev,
2392                                     (dma_addr_t) ((RxD3_t*)rxdp)->Buffer0_ptr,
2393                                     BUF0_LEN, PCI_DMA_FROMDEVICE);
2394                         rxdp->Control_2 = SET_BUFFER0_SIZE_3(BUF0_LEN);
2395                         if (nic->rxd_mode == RXD_MODE_3B) {
2396                                 /* Two buffer mode */
2397
2398                                 /*
2399                                  * Buffer2 will have L3/L4 header plus 
2400                                  * L4 payload
2401                                  */
2402                                 ((RxD3_t*)rxdp)->Buffer2_ptr = pci_map_single
2403                                 (nic->pdev, skb->data, dev->mtu + 4,
2404                                                 PCI_DMA_FROMDEVICE);
2405
2406                                 /* Buffer-1 will be dummy buffer. Not used */
2407                                 if (!(((RxD3_t*)rxdp)->Buffer1_ptr)) {
2408                                         ((RxD3_t*)rxdp)->Buffer1_ptr =
2409                                                 pci_map_single(nic->pdev, 
2410                                                 ba->ba_1, BUF1_LEN,
2411                                                 PCI_DMA_FROMDEVICE);
2412                                 }
2413                                 rxdp->Control_2 |= SET_BUFFER1_SIZE_3(1);
2414                                 rxdp->Control_2 |= SET_BUFFER2_SIZE_3
2415                                                                 (dev->mtu + 4);
2416                         } else {
2417                                 /* 3 buffer mode */
2418                                 if (fill_rxd_3buf(nic, rxdp, skb) == -ENOMEM) {
2419                                         dev_kfree_skb_irq(skb);
2420                                         if (first_rxdp) {
2421                                                 wmb();
2422                                                 first_rxdp->Control_1 |=
2423                                                         RXD_OWN_XENA;
2424                                         }
2425                                         return -ENOMEM ;
2426                                 }
2427                         }
2428                         rxdp->Control_2 |= BIT(0);
2429                 }
2430                 rxdp->Host_Control = (unsigned long) (skb);
2431                 if (alloc_tab & ((1 << rxsync_frequency) - 1))
2432                         rxdp->Control_1 |= RXD_OWN_XENA;
2433                 off++;
2434                 if (off == (rxd_count[nic->rxd_mode] + 1))
2435                         off = 0;
2436                 mac_control->rings[ring_no].rx_curr_put_info.offset = off;
2437
2438                 rxdp->Control_2 |= SET_RXD_MARKER;
2439                 if (!(alloc_tab & ((1 << rxsync_frequency) - 1))) {
2440                         if (first_rxdp) {
2441                                 wmb();
2442                                 first_rxdp->Control_1 |= RXD_OWN_XENA;
2443                         }
2444                         first_rxdp = rxdp;
2445                 }
2446                 atomic_inc(&nic->rx_bufs_left[ring_no]);
2447                 alloc_tab++;
2448         }
2449
2450       end:
2451         /* Transfer ownership of first descriptor to adapter just before
2452          * exiting. Before that, use memory barrier so that ownership
2453          * and other fields are seen by adapter correctly.
2454          */
2455         if (first_rxdp) {
2456                 wmb();
2457                 first_rxdp->Control_1 |= RXD_OWN_XENA;
2458         }
2459
2460         return SUCCESS;
2461 }
2462
2463 static void free_rxd_blk(struct s2io_nic *sp, int ring_no, int blk)
2464 {
2465         struct net_device *dev = sp->dev;
2466         int j;
2467         struct sk_buff *skb;
2468         RxD_t *rxdp;
2469         mac_info_t *mac_control;
2470         buffAdd_t *ba;
2471
2472         mac_control = &sp->mac_control;
2473         for (j = 0 ; j < rxd_count[sp->rxd_mode]; j++) {
2474                 rxdp = mac_control->rings[ring_no].
2475                                 rx_blocks[blk].rxds[j].virt_addr;
2476                 skb = (struct sk_buff *)
2477                         ((unsigned long) rxdp->Host_Control);
2478                 if (!skb) {
2479                         continue;
2480                 }
2481                 if (sp->rxd_mode == RXD_MODE_1) {
2482                         pci_unmap_single(sp->pdev, (dma_addr_t)
2483                                  ((RxD1_t*)rxdp)->Buffer0_ptr,
2484                                  dev->mtu +
2485                                  HEADER_ETHERNET_II_802_3_SIZE
2486                                  + HEADER_802_2_SIZE +
2487                                  HEADER_SNAP_SIZE,
2488                                  PCI_DMA_FROMDEVICE);
2489                         memset(rxdp, 0, sizeof(RxD1_t));
2490                 } else if(sp->rxd_mode == RXD_MODE_3B) {
2491                         ba = &mac_control->rings[ring_no].
2492                                 ba[blk][j];
2493                         pci_unmap_single(sp->pdev, (dma_addr_t)
2494                                  ((RxD3_t*)rxdp)->Buffer0_ptr,
2495                                  BUF0_LEN,
2496                                  PCI_DMA_FROMDEVICE);
2497                         pci_unmap_single(sp->pdev, (dma_addr_t)
2498                                  ((RxD3_t*)rxdp)->Buffer1_ptr,
2499                                  BUF1_LEN,
2500                                  PCI_DMA_FROMDEVICE);
2501                         pci_unmap_single(sp->pdev, (dma_addr_t)
2502                                  ((RxD3_t*)rxdp)->Buffer2_ptr,
2503                                  dev->mtu + 4,
2504                                  PCI_DMA_FROMDEVICE);
2505                         memset(rxdp, 0, sizeof(RxD3_t));
2506                 } else {
2507                         pci_unmap_single(sp->pdev, (dma_addr_t)
2508                                 ((RxD3_t*)rxdp)->Buffer0_ptr, BUF0_LEN,
2509                                 PCI_DMA_FROMDEVICE);
2510                         pci_unmap_single(sp->pdev, (dma_addr_t)
2511                                 ((RxD3_t*)rxdp)->Buffer1_ptr, 
2512                                 l3l4hdr_size + 4,
2513                                 PCI_DMA_FROMDEVICE);
2514                         pci_unmap_single(sp->pdev, (dma_addr_t)
2515                                 ((RxD3_t*)rxdp)->Buffer2_ptr, dev->mtu,
2516                                 PCI_DMA_FROMDEVICE);
2517                         memset(rxdp, 0, sizeof(RxD3_t));
2518                 }
2519                 dev_kfree_skb(skb);
2520                 atomic_dec(&sp->rx_bufs_left[ring_no]);
2521         }
2522 }
2523
2524 /**
2525  *  free_rx_buffers - Frees all Rx buffers
2526  *  @sp: device private variable.
2527  *  Description:
2528  *  This function will free all Rx buffers allocated by host.
2529  *  Return Value:
2530  *  NONE.
2531  */
2532
2533 static void free_rx_buffers(struct s2io_nic *sp)
2534 {
2535         struct net_device *dev = sp->dev;
2536         int i, blk = 0, buf_cnt = 0;
2537         mac_info_t *mac_control;
2538         struct config_param *config;
2539
2540         mac_control = &sp->mac_control;
2541         config = &sp->config;
2542
2543         for (i = 0; i < config->rx_ring_num; i++) {
2544                 for (blk = 0; blk < rx_ring_sz[i]; blk++)
2545                         free_rxd_blk(sp,i,blk);
2546
2547                 mac_control->rings[i].rx_curr_put_info.block_index = 0;
2548                 mac_control->rings[i].rx_curr_get_info.block_index = 0;
2549                 mac_control->rings[i].rx_curr_put_info.offset = 0;
2550                 mac_control->rings[i].rx_curr_get_info.offset = 0;
2551                 atomic_set(&sp->rx_bufs_left[i], 0);
2552                 DBG_PRINT(INIT_DBG, "%s:Freed 0x%x Rx Buffers on ring%d\n",
2553                           dev->name, buf_cnt, i);
2554         }
2555 }
2556
2557 /**
2558  * s2io_poll - Rx interrupt handler for NAPI support
2559  * @dev : pointer to the device structure.
2560  * @budget : The number of packets that were budgeted to be processed
2561  * during  one pass through the 'Poll" function.
2562  * Description:
2563  * Comes into picture only if NAPI support has been incorporated. It does
2564  * the same thing that rx_intr_handler does, but not in a interrupt context
2565  * also It will process only a given number of packets.
2566  * Return value:
2567  * 0 on success and 1 if there are No Rx packets to be processed.
2568  */
2569
2570 #if defined(CONFIG_S2IO_NAPI)
2571 static int s2io_poll(struct net_device *dev, int *budget)
2572 {
2573         nic_t *nic = dev->priv;
2574         int pkt_cnt = 0, org_pkts_to_process;
2575         mac_info_t *mac_control;
2576         struct config_param *config;
2577         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2578         u64 val64 = 0xFFFFFFFFFFFFFFFFULL;
2579         int i;
2580
2581         atomic_inc(&nic->isr_cnt);
2582         mac_control = &nic->mac_control;
2583         config = &nic->config;
2584
2585         nic->pkts_to_process = *budget;
2586         if (nic->pkts_to_process > dev->quota)
2587                 nic->pkts_to_process = dev->quota;
2588         org_pkts_to_process = nic->pkts_to_process;
2589
2590         writeq(val64, &bar0->rx_traffic_int);
2591         val64 = readl(&bar0->rx_traffic_int);
2592
2593         for (i = 0; i < config->rx_ring_num; i++) {
2594                 rx_intr_handler(&mac_control->rings[i]);
2595                 pkt_cnt = org_pkts_to_process - nic->pkts_to_process;
2596                 if (!nic->pkts_to_process) {
2597                         /* Quota for the current iteration has been met */
2598                         goto no_rx;
2599                 }
2600         }
2601         if (!pkt_cnt)
2602                 pkt_cnt = 1;
2603
2604         dev->quota -= pkt_cnt;
2605         *budget -= pkt_cnt;
2606         netif_rx_complete(dev);
2607
2608         for (i = 0; i < config->rx_ring_num; i++) {
2609                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2610                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2611                         DBG_PRINT(ERR_DBG, " in Rx Poll!!\n");
2612                         break;
2613                 }
2614         }
2615         /* Re enable the Rx interrupts. */
2616         writeq(0x0, &bar0->rx_traffic_mask);
2617         val64 = readl(&bar0->rx_traffic_mask);
2618         atomic_dec(&nic->isr_cnt);
2619         return 0;
2620
2621 no_rx:
2622         dev->quota -= pkt_cnt;
2623         *budget -= pkt_cnt;
2624
2625         for (i = 0; i < config->rx_ring_num; i++) {
2626                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2627                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2628                         DBG_PRINT(ERR_DBG, " in Rx Poll!!\n");
2629                         break;
2630                 }
2631         }
2632         atomic_dec(&nic->isr_cnt);
2633         return 1;
2634 }
2635 #endif
2636
2637 #ifdef CONFIG_NET_POLL_CONTROLLER
2638 /**
2639  * s2io_netpoll - netpoll event handler entry point
2640  * @dev : pointer to the device structure.
2641  * Description:
2642  *      This function will be called by upper layer to check for events on the
2643  * interface in situations where interrupts are disabled. It is used for
2644  * specific in-kernel networking tasks, such as remote consoles and kernel
2645  * debugging over the network (example netdump in RedHat).
2646  */
2647 static void s2io_netpoll(struct net_device *dev)
2648 {
2649         nic_t *nic = dev->priv;
2650         mac_info_t *mac_control;
2651         struct config_param *config;
2652         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2653         u64 val64 = 0xFFFFFFFFFFFFFFFFULL;
2654         int i;
2655
2656         disable_irq(dev->irq);
2657
2658         atomic_inc(&nic->isr_cnt);
2659         mac_control = &nic->mac_control;
2660         config = &nic->config;
2661
2662         writeq(val64, &bar0->rx_traffic_int);
2663         writeq(val64, &bar0->tx_traffic_int);
2664
2665         /* we need to free up the transmitted skbufs or else netpoll will 
2666          * run out of skbs and will fail and eventually netpoll application such
2667          * as netdump will fail.
2668          */
2669         for (i = 0; i < config->tx_fifo_num; i++)
2670                 tx_intr_handler(&mac_control->fifos[i]);
2671
2672         /* check for received packet and indicate up to network */
2673         for (i = 0; i < config->rx_ring_num; i++)
2674                 rx_intr_handler(&mac_control->rings[i]);
2675
2676         for (i = 0; i < config->rx_ring_num; i++) {
2677                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2678                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2679                         DBG_PRINT(ERR_DBG, " in Rx Netpoll!!\n");
2680                         break;
2681                 }
2682         }
2683         atomic_dec(&nic->isr_cnt);
2684         enable_irq(dev->irq);
2685         return;
2686 }
2687 #endif
2688
2689 /**
2690  *  rx_intr_handler - Rx interrupt handler
2691  *  @nic: device private variable.
2692  *  Description:
2693  *  If the interrupt is because of a received frame or if the
2694  *  receive ring contains fresh as yet un-processed frames,this function is
2695  *  called. It picks out the RxD at which place the last Rx processing had
2696  *  stopped and sends the skb to the OSM's Rx handler and then increments
2697  *  the offset.
2698  *  Return Value:
2699  *  NONE.
2700  */
2701 static void rx_intr_handler(ring_info_t *ring_data)
2702 {
2703         nic_t *nic = ring_data->nic;
2704         struct net_device *dev = (struct net_device *) nic->dev;
2705         int get_block, put_block, put_offset;
2706         rx_curr_get_info_t get_info, put_info;
2707         RxD_t *rxdp;
2708         struct sk_buff *skb;
2709 #ifndef CONFIG_S2IO_NAPI
2710         int pkt_cnt = 0;
2711 #endif
2712         int i;
2713
2714         spin_lock(&nic->rx_lock);
2715         if (atomic_read(&nic->card_state) == CARD_DOWN) {
2716                 DBG_PRINT(INTR_DBG, "%s: %s going down for reset\n",
2717                           __FUNCTION__, dev->name);
2718                 spin_unlock(&nic->rx_lock);
2719                 return;
2720         }
2721
2722         get_info = ring_data->rx_curr_get_info;
2723         get_block = get_info.block_index;
2724         put_info = ring_data->rx_curr_put_info;
2725         put_block = put_info.block_index;
2726         rxdp = ring_data->rx_blocks[get_block].rxds[get_info.offset].virt_addr;
2727 #ifndef CONFIG_S2IO_NAPI
2728         spin_lock(&nic->put_lock);
2729         put_offset = ring_data->put_pos;
2730         spin_unlock(&nic->put_lock);
2731 #else
2732         put_offset = (put_block * (rxd_count[nic->rxd_mode] + 1)) +
2733                 put_info.offset;
2734 #endif
2735         while (RXD_IS_UP2DT(rxdp)) {
2736                 /* If your are next to put index then it's FIFO full condition */
2737                 if ((get_block == put_block) &&
2738                     (get_info.offset + 1) == put_info.offset) {
2739                         DBG_PRINT(INTR_DBG, "%s: Ring Full\n",dev->name);
2740                         break;
2741                 }
2742                 skb = (struct sk_buff *) ((unsigned long)rxdp->Host_Control);
2743                 if (skb == NULL) {
2744                         DBG_PRINT(ERR_DBG, "%s: The skb is ",
2745                                   dev->name);
2746                         DBG_PRINT(ERR_DBG, "Null in Rx Intr\n");
2747                         spin_unlock(&nic->rx_lock);
2748                         return;
2749                 }
2750                 if (nic->rxd_mode == RXD_MODE_1) {
2751                         pci_unmap_single(nic->pdev, (dma_addr_t)
2752                                  ((RxD1_t*)rxdp)->Buffer0_ptr,
2753                                  dev->mtu +
2754                                  HEADER_ETHERNET_II_802_3_SIZE +
2755                                  HEADER_802_2_SIZE +
2756                                  HEADER_SNAP_SIZE,
2757                                  PCI_DMA_FROMDEVICE);
2758                 } else if (nic->rxd_mode == RXD_MODE_3B) {
2759                         pci_dma_sync_single_for_cpu(nic->pdev, (dma_addr_t)
2760                                  ((RxD3_t*)rxdp)->Buffer0_ptr,
2761                                  BUF0_LEN, PCI_DMA_FROMDEVICE);
2762                         pci_unmap_single(nic->pdev, (dma_addr_t)
2763                                  ((RxD3_t*)rxdp)->Buffer2_ptr,
2764                                  dev->mtu + 4,
2765                                  PCI_DMA_FROMDEVICE);
2766                 } else {
2767                         pci_dma_sync_single_for_cpu(nic->pdev, (dma_addr_t)
2768                                          ((RxD3_t*)rxdp)->Buffer0_ptr, BUF0_LEN,
2769                                          PCI_DMA_FROMDEVICE);
2770                         pci_unmap_single(nic->pdev, (dma_addr_t)
2771                                          ((RxD3_t*)rxdp)->Buffer1_ptr,
2772                                          l3l4hdr_size + 4,
2773                                          PCI_DMA_FROMDEVICE);
2774                         pci_unmap_single(nic->pdev, (dma_addr_t)
2775                                          ((RxD3_t*)rxdp)->Buffer2_ptr,
2776                                          dev->mtu, PCI_DMA_FROMDEVICE);
2777                 }
2778                 prefetch(skb->data);
2779                 rx_osm_handler(ring_data, rxdp);
2780                 get_info.offset++;
2781                 ring_data->rx_curr_get_info.offset = get_info.offset;
2782                 rxdp = ring_data->rx_blocks[get_block].
2783                                 rxds[get_info.offset].virt_addr;
2784                 if (get_info.offset == rxd_count[nic->rxd_mode]) {
2785                         get_info.offset = 0;
2786                         ring_data->rx_curr_get_info.offset = get_info.offset;
2787                         get_block++;
2788                         if (get_block == ring_data->block_count)
2789                                 get_block = 0;
2790                         ring_data->rx_curr_get_info.block_index = get_block;
2791                         rxdp = ring_data->rx_blocks[get_block].block_virt_addr;
2792                 }
2793
2794 #ifdef CONFIG_S2IO_NAPI
2795                 nic->pkts_to_process -= 1;
2796                 if (!nic->pkts_to_process)
2797                         break;
2798 #else
2799                 pkt_cnt++;
2800                 if ((indicate_max_pkts) && (pkt_cnt > indicate_max_pkts))
2801                         break;
2802 #endif
2803         }
2804         if (nic->lro) {
2805                 /* Clear all LRO sessions before exiting */
2806                 for (i=0; i<MAX_LRO_SESSIONS; i++) {
2807                         lro_t *lro = &nic->lro0_n[i];
2808                         if (lro->in_use) {
2809                                 update_L3L4_header(nic, lro);
2810                                 queue_rx_frame(lro->parent);
2811                                 clear_lro_session(lro);
2812                         }
2813                 }
2814         }
2815
2816         spin_unlock(&nic->rx_lock);
2817 }
2818
2819 /**
2820  *  tx_intr_handler - Transmit interrupt handler
2821  *  @nic : device private variable
2822  *  Description:
2823  *  If an interrupt was raised to indicate DMA complete of the
2824  *  Tx packet, this function is called. It identifies the last TxD
2825  *  whose buffer was freed and frees all skbs whose data have already
2826  *  DMA'ed into the NICs internal memory.
2827  *  Return Value:
2828  *  NONE
2829  */
2830
2831 static void tx_intr_handler(fifo_info_t *fifo_data)
2832 {
2833         nic_t *nic = fifo_data->nic;
2834         struct net_device *dev = (struct net_device *) nic->dev;
2835         tx_curr_get_info_t get_info, put_info;
2836         struct sk_buff *skb;
2837         TxD_t *txdlp;
2838
2839         get_info = fifo_data->tx_curr_get_info;
2840         put_info = fifo_data->tx_curr_put_info;
2841         txdlp = (TxD_t *) fifo_data->list_info[get_info.offset].
2842             list_virt_addr;
2843         while ((!(txdlp->Control_1 & TXD_LIST_OWN_XENA)) &&
2844                (get_info.offset != put_info.offset) &&
2845                (txdlp->Host_Control)) {
2846                 /* Check for TxD errors */
2847                 if (txdlp->Control_1 & TXD_T_CODE) {
2848                         unsigned long long err;
2849                         err = txdlp->Control_1 & TXD_T_CODE;
2850                         if (err & 0x1) {
2851                                 nic->mac_control.stats_info->sw_stat.
2852                                                 parity_err_cnt++;
2853                         }
2854                         if ((err >> 48) == 0xA) {
2855                                 DBG_PRINT(TX_DBG, "TxD returned due \
2856 to loss of link\n");
2857                         }
2858                         else {
2859                                 DBG_PRINT(ERR_DBG, "***TxD error \
2860 %llx\n", err);
2861                         }
2862                 }
2863
2864                 skb = s2io_txdl_getskb(fifo_data, txdlp, get_info.offset);
2865                 if (skb == NULL) {
2866                         DBG_PRINT(ERR_DBG, "%s: Null skb ",
2867                         __FUNCTION__);
2868                         DBG_PRINT(ERR_DBG, "in Tx Free Intr\n");
2869                         return;
2870                 }
2871
2872                 /* Updating the statistics block */
2873                 nic->stats.tx_bytes += skb->len;
2874                 dev_kfree_skb_irq(skb);
2875
2876                 get_info.offset++;
2877                 if (get_info.offset == get_info.fifo_len + 1)
2878                         get_info.offset = 0;
2879                 txdlp = (TxD_t *) fifo_data->list_info
2880                     [get_info.offset].list_virt_addr;
2881                 fifo_data->tx_curr_get_info.offset =
2882                     get_info.offset;
2883         }
2884
2885         spin_lock(&nic->tx_lock);
2886         if (netif_queue_stopped(dev))
2887                 netif_wake_queue(dev);
2888         spin_unlock(&nic->tx_lock);
2889 }
2890
2891 /**
2892  *  s2io_mdio_write - Function to write in to MDIO registers
2893  *  @mmd_type : MMD type value (PMA/PMD/WIS/PCS/PHYXS)
2894  *  @addr     : address value
2895  *  @value    : data value
2896  *  @dev      : pointer to net_device structure
2897  *  Description:
2898  *  This function is used to write values to the MDIO registers
2899  *  NONE
2900  */
2901 static void s2io_mdio_write(u32 mmd_type, u64 addr, u16 value, struct net_device *dev)
2902 {
2903         u64 val64 = 0x0;
2904         nic_t *sp = dev->priv;
2905         XENA_dev_config_t *bar0 = (XENA_dev_config_t *)sp->bar0;
2906
2907         //address transaction
2908         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2909                         | MDIO_MMD_DEV_ADDR(mmd_type)
2910                         | MDIO_MMS_PRT_ADDR(0x0);
2911         writeq(val64, &bar0->mdio_control);
2912         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2913         writeq(val64, &bar0->mdio_control);
2914         udelay(100);
2915
2916         //Data transaction
2917         val64 = 0x0;
2918         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2919                         | MDIO_MMD_DEV_ADDR(mmd_type)
2920                         | MDIO_MMS_PRT_ADDR(0x0)
2921                         | MDIO_MDIO_DATA(value)
2922                         | MDIO_OP(MDIO_OP_WRITE_TRANS);
2923         writeq(val64, &bar0->mdio_control);
2924         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2925         writeq(val64, &bar0->mdio_control);
2926         udelay(100);
2927
2928         val64 = 0x0;
2929         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2930         | MDIO_MMD_DEV_ADDR(mmd_type)
2931         | MDIO_MMS_PRT_ADDR(0x0)
2932         | MDIO_OP(MDIO_OP_READ_TRANS);
2933         writeq(val64, &bar0->mdio_control);
2934         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2935         writeq(val64, &bar0->mdio_control);
2936         udelay(100);
2937
2938 }
2939
2940 /**
2941  *  s2io_mdio_read - Function to write in to MDIO registers
2942  *  @mmd_type : MMD type value (PMA/PMD/WIS/PCS/PHYXS)
2943  *  @addr     : address value
2944  *  @dev      : pointer to net_device structure
2945  *  Description:
2946  *  This function is used to read values to the MDIO registers
2947  *  NONE
2948  */
2949 static u64 s2io_mdio_read(u32 mmd_type, u64 addr, struct net_device *dev)
2950 {
2951         u64 val64 = 0x0;
2952         u64 rval64 = 0x0;
2953         nic_t *sp = dev->priv;
2954         XENA_dev_config_t *bar0 = (XENA_dev_config_t *)sp->bar0;
2955
2956         /* address transaction */
2957         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2958                         | MDIO_MMD_DEV_ADDR(mmd_type)
2959                         | MDIO_MMS_PRT_ADDR(0x0);
2960         writeq(val64, &bar0->mdio_control);
2961         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2962         writeq(val64, &bar0->mdio_control);
2963         udelay(100);
2964
2965         /* Data transaction */
2966         val64 = 0x0;
2967         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2968                         | MDIO_MMD_DEV_ADDR(mmd_type)
2969                         | MDIO_MMS_PRT_ADDR(0x0)
2970                         | MDIO_OP(MDIO_OP_READ_TRANS);
2971         writeq(val64, &bar0->mdio_control);
2972         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2973         writeq(val64, &bar0->mdio_control);
2974         udelay(100);
2975
2976         /* Read the value from regs */
2977         rval64 = readq(&bar0->mdio_control);
2978         rval64 = rval64 & 0xFFFF0000;
2979         rval64 = rval64 >> 16;
2980         return rval64;
2981 }
2982 /**
2983  *  s2io_chk_xpak_counter - Function to check the status of the xpak counters
2984  *  @counter      : couter value to be updated
2985  *  @flag         : flag to indicate the status
2986  *  @type         : counter type
2987  *  Description:
2988  *  This function is to check the status of the xpak counters value
2989  *  NONE
2990  */
2991
2992 static void s2io_chk_xpak_counter(u64 *counter, u64 * regs_stat, u32 index, u16 flag, u16 type)
2993 {
2994         u64 mask = 0x3;
2995         u64 val64;
2996         int i;
2997         for(i = 0; i <index; i++)
2998                 mask = mask << 0x2;
2999
3000         if(flag > 0)
3001         {
3002                 *counter = *counter + 1;
3003                 val64 = *regs_stat & mask;
3004                 val64 = val64 >> (index * 0x2);
3005                 val64 = val64 + 1;
3006                 if(val64 == 3)
3007                 {
3008                         switch(type)
3009                         {
3010                         case 1:
3011                                 DBG_PRINT(ERR_DBG, "Take Xframe NIC out of "
3012                                           "service. Excessive temperatures may "
3013                                           "result in premature transceiver "
3014                                           "failure \n");
3015                         break;
3016                         case 2:
3017                                 DBG_PRINT(ERR_DBG, "Take Xframe NIC out of "
3018                                           "service Excessive bias currents may "
3019                                           "indicate imminent laser diode "
3020                                           "failure \n");
3021                         break;
3022                         case 3:
3023                                 DBG_PRINT(ERR_DBG, "Take Xframe NIC out of "
3024                                           "service Excessive laser output "
3025                                           "power may saturate far-end "
3026                                           "receiver\n");
3027                         break;
3028                         default:
3029                                 DBG_PRINT(ERR_DBG, "Incorrect XPAK Alarm "
3030                                           "type \n");
3031                         }
3032                         val64 = 0x0;
3033                 }
3034                 val64 = val64 << (index * 0x2);
3035                 *regs_stat = (*regs_stat & (~mask)) | (val64);
3036
3037         } else {
3038                 *regs_stat = *regs_stat & (~mask);
3039         }
3040 }
3041
3042 /**
3043  *  s2io_updt_xpak_counter - Function to update the xpak counters
3044  *  @dev         : pointer to net_device struct
3045  *  Description:
3046  *  This function is to upate the status of the xpak counters value
3047  *  NONE
3048  */
3049 static void s2io_updt_xpak_counter(struct net_device *dev)
3050 {
3051         u16 flag  = 0x0;
3052         u16 type  = 0x0;
3053         u16 val16 = 0x0;
3054         u64 val64 = 0x0;
3055         u64 addr  = 0x0;
3056
3057         nic_t *sp = dev->priv;
3058         StatInfo_t *stat_info = sp->mac_control.stats_info;
3059
3060         /* Check the communication with the MDIO slave */
3061         addr = 0x0000;
3062         val64 = 0x0;
3063         val64 = s2io_mdio_read(MDIO_MMD_PMA_DEV_ADDR, addr, dev);
3064         if((val64 == 0xFFFF) || (val64 == 0x0000))
3065         {
3066                 DBG_PRINT(ERR_DBG, "ERR: MDIO slave access failed - "
3067                           "Returned %llx\n", (unsigned long long)val64);
3068                 return;
3069         }
3070
3071         /* Check for the expecte value of 2040 at PMA address 0x0000 */
3072         if(val64 != 0x2040)
3073         {
3074                 DBG_PRINT(ERR_DBG, "Incorrect value at PMA address 0x0000 - ");
3075                 DBG_PRINT(ERR_DBG, "Returned: %llx- Expected: 0x2040\n",
3076                           (unsigned long long)val64);
3077                 return;
3078         }
3079
3080         /* Loading the DOM register to MDIO register */
3081         addr = 0xA100;
3082         s2io_mdio_write(MDIO_MMD_PMA_DEV_ADDR, addr, val16, dev);
3083         val64 = s2io_mdio_read(MDIO_MMD_PMA_DEV_ADDR, addr, dev);
3084
3085         /* Reading the Alarm flags */
3086         addr = 0xA070;
3087         val64 = 0x0;
3088         val64 = s2io_mdio_read(MDIO_MMD_PMA_DEV_ADDR, addr, dev);
3089
3090         flag = CHECKBIT(val64, 0x7);
3091         type = 1;
3092         s2io_chk_xpak_counter(&stat_info->xpak_stat.alarm_transceiver_temp_high,
3093                                 &stat_info->xpak_stat.xpak_regs_stat,
3094                                 0x0, flag, type);
3095
3096         if(CHECKBIT(val64, 0x6))
3097                 stat_info->xpak_stat.alarm_transceiver_temp_low++;
3098
3099         flag = CHECKBIT(val64, 0x3);
3100         type = 2;
3101         s2io_chk_xpak_counter(&stat_info->xpak_stat.alarm_laser_bias_current_high,
3102                                 &stat_info->xpak_stat.xpak_regs_stat,
3103                                 0x2, flag, type);
3104
3105         if(CHECKBIT(val64, 0x2))
3106                 stat_info->xpak_stat.alarm_laser_bias_current_low++;
3107
3108         flag = CHECKBIT(val64, 0x1);
3109         type = 3;
3110         s2io_chk_xpak_counter(&stat_info->xpak_stat.alarm_laser_output_power_high,
3111                                 &stat_info->xpak_stat.xpak_regs_stat,
3112                                 0x4, flag, type);
3113
3114         if(CHECKBIT(val64, 0x0))
3115                 stat_info->xpak_stat.alarm_laser_output_power_low++;
3116
3117         /* Reading the Warning flags */
3118         addr = 0xA074;
3119         val64 = 0x0;
3120         val64 = s2io_mdio_read(MDIO_MMD_PMA_DEV_ADDR, addr, dev);
3121
3122         if(CHECKBIT(val64, 0x7))
3123                 stat_info->xpak_stat.warn_transceiver_temp_high++;
3124
3125         if(CHECKBIT(val64, 0x6))
3126                 stat_info->xpak_stat.warn_transceiver_temp_low++;
3127
3128         if(CHECKBIT(val64, 0x3))
3129                 stat_info->xpak_stat.warn_laser_bias_current_high++;
3130
3131         if(CHECKBIT(val64, 0x2))
3132                 stat_info->xpak_stat.warn_laser_bias_current_low++;
3133
3134         if(CHECKBIT(val64, 0x1))
3135                 stat_info->xpak_stat.warn_laser_output_power_high++;
3136
3137         if(CHECKBIT(val64, 0x0))
3138                 stat_info->xpak_stat.warn_laser_output_power_low++;
3139 }
3140
3141 /**
3142  *  alarm_intr_handler - Alarm Interrrupt handler
3143  *  @nic: device private variable
3144  *  Description: If the interrupt was neither because of Rx packet or Tx
3145  *  complete, this function is called. If the interrupt was to indicate
3146  *  a loss of link, the OSM link status handler is invoked for any other
3147  *  alarm interrupt the block that raised the interrupt is displayed
3148  *  and a H/W reset is issued.
3149  *  Return Value:
3150  *  NONE
3151 */
3152
3153 static void alarm_intr_handler(struct s2io_nic *nic)
3154 {
3155         struct net_device *dev = (struct net_device *) nic->dev;
3156         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3157         register u64 val64 = 0, err_reg = 0;
3158         u64 cnt;
3159         int i;
3160         nic->mac_control.stats_info->sw_stat.ring_full_cnt = 0;
3161         /* Handling the XPAK counters update */
3162         if(nic->mac_control.stats_info->xpak_stat.xpak_timer_count < 72000) {
3163                 /* waiting for an hour */
3164                 nic->mac_control.stats_info->xpak_stat.xpak_timer_count++;
3165         } else {
3166                 s2io_updt_xpak_counter(dev);
3167                 /* reset the count to zero */
3168                 nic->mac_control.stats_info->xpak_stat.xpak_timer_count = 0;
3169         }
3170
3171         /* Handling link status change error Intr */
3172         if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
3173                 err_reg = readq(&bar0->mac_rmac_err_reg);
3174                 writeq(err_reg, &bar0->mac_rmac_err_reg);
3175                 if (err_reg & RMAC_LINK_STATE_CHANGE_INT) {
3176                         schedule_work(&nic->set_link_task);
3177                 }
3178         }
3179
3180         /* Handling Ecc errors */
3181         val64 = readq(&bar0->mc_err_reg);
3182         writeq(val64, &bar0->mc_err_reg);
3183         if (val64 & (MC_ERR_REG_ECC_ALL_SNG | MC_ERR_REG_ECC_ALL_DBL)) {
3184                 if (val64 & MC_ERR_REG_ECC_ALL_DBL) {
3185                         nic->mac_control.stats_info->sw_stat.
3186                                 double_ecc_errs++;
3187                         DBG_PRINT(INIT_DBG, "%s: Device indicates ",
3188                                   dev->name);
3189                         DBG_PRINT(INIT_DBG, "double ECC error!!\n");
3190                         if (nic->device_type != XFRAME_II_DEVICE) {
3191                                 /* Reset XframeI only if critical error */
3192                                 if (val64 & (MC_ERR_REG_MIRI_ECC_DB_ERR_0 |
3193                                              MC_ERR_REG_MIRI_ECC_DB_ERR_1)) {
3194                                         netif_stop_queue(dev);
3195                                         schedule_work(&nic->rst_timer_task);
3196                                         nic->mac_control.stats_info->sw_stat.
3197                                                         soft_reset_cnt++;
3198                                 }
3199                         }
3200                 } else {
3201                         nic->mac_control.stats_info->sw_stat.
3202                                 single_ecc_errs++;
3203                 }
3204         }
3205
3206         /* In case of a serious error, the device will be Reset. */
3207         val64 = readq(&bar0->serr_source);
3208         if (val64 & SERR_SOURCE_ANY) {
3209                 nic->mac_control.stats_info->sw_stat.serious_err_cnt++;
3210                 DBG_PRINT(ERR_DBG, "%s: Device indicates ", dev->name);
3211                 DBG_PRINT(ERR_DBG, "serious error %llx!!\n", 
3212                           (unsigned long long)val64);
3213                 netif_stop_queue(dev);
3214                 schedule_work(&nic->rst_timer_task);
3215                 nic->mac_control.stats_info->sw_stat.soft_reset_cnt++;
3216         }
3217
3218         /*
3219          * Also as mentioned in the latest Errata sheets if the PCC_FB_ECC
3220          * Error occurs, the adapter will be recycled by disabling the
3221          * adapter enable bit and enabling it again after the device
3222          * becomes Quiescent.
3223          */
3224         val64 = readq(&bar0->pcc_err_reg);
3225         writeq(val64, &bar0->pcc_err_reg);
3226         if (val64 & PCC_FB_ECC_DB_ERR) {
3227                 u64 ac = readq(&bar0->adapter_control);
3228                 ac &= ~(ADAPTER_CNTL_EN);
3229                 writeq(ac, &bar0->adapter_control);
3230                 ac = readq(&bar0->adapter_control);
3231                 schedule_work(&nic->set_link_task);
3232         }
3233         /* Check for data parity error */
3234         val64 = readq(&bar0->pic_int_status);
3235         if (val64 & PIC_INT_GPIO) {
3236                 val64 = readq(&bar0->gpio_int_reg);
3237                 if (val64 & GPIO_INT_REG_DP_ERR_INT) {
3238                         nic->mac_control.stats_info->sw_stat.parity_err_cnt++;
3239                         schedule_work(&nic->rst_timer_task);
3240                         nic->mac_control.stats_info->sw_stat.soft_reset_cnt++;
3241                 }
3242         }
3243
3244         /* Check for ring full counter */
3245         if (nic->device_type & XFRAME_II_DEVICE) {
3246                 val64 = readq(&bar0->ring_bump_counter1);
3247                 for (i=0; i<4; i++) {
3248                         cnt = ( val64 & vBIT(0xFFFF,(i*16),16));
3249                         cnt >>= 64 - ((i+1)*16);
3250                         nic->mac_control.stats_info->sw_stat.ring_full_cnt
3251                                 += cnt;
3252                 }
3253
3254                 val64 = readq(&bar0->ring_bump_counter2);
3255                 for (i=0; i<4; i++) {
3256                         cnt = ( val64 & vBIT(0xFFFF,(i*16),16));
3257                         cnt >>= 64 - ((i+1)*16);
3258                         nic->mac_control.stats_info->sw_stat.ring_full_cnt
3259                                 += cnt;
3260                 }
3261         }
3262
3263         /* Other type of interrupts are not being handled now,  TODO */
3264 }
3265
3266 /**
3267  *  wait_for_cmd_complete - waits for a command to complete.
3268  *  @sp : private member of the device structure, which is a pointer to the
3269  *  s2io_nic structure.
3270  *  Description: Function that waits for a command to Write into RMAC
3271  *  ADDR DATA registers to be completed and returns either success or
3272  *  error depending on whether the command was complete or not.
3273  *  Return value:
3274  *   SUCCESS on success and FAILURE on failure.
3275  */
3276
3277 static int wait_for_cmd_complete(void *addr, u64 busy_bit)
3278 {
3279         int ret = FAILURE, cnt = 0;
3280         u64 val64;
3281
3282         while (TRUE) {
3283                 val64 = readq(addr);
3284                 if (!(val64 & busy_bit)) {
3285                         ret = SUCCESS;
3286                         break;
3287                 }
3288
3289                 if(in_interrupt())
3290                         mdelay(50);
3291                 else
3292                         msleep(50);
3293
3294                 if (cnt++ > 10)
3295                         break;
3296         }
3297         return ret;
3298 }
3299
3300 /**
3301  *  s2io_reset - Resets the card.
3302  *  @sp : private member of the device structure.
3303  *  Description: Function to Reset the card. This function then also
3304  *  restores the previously saved PCI configuration space registers as
3305  *  the card reset also resets the configuration space.
3306  *  Return value:
3307  *  void.
3308  */
3309
3310 static void s2io_reset(nic_t * sp)
3311 {
3312         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3313         u64 val64;
3314         u16 subid, pci_cmd;
3315
3316         /* Back up  the PCI-X CMD reg, dont want to lose MMRBC, OST settings */
3317         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER, &(pci_cmd));
3318
3319         val64 = SW_RESET_ALL;
3320         writeq(val64, &bar0->sw_reset);
3321
3322         /*
3323          * At this stage, if the PCI write is indeed completed, the
3324          * card is reset and so is the PCI Config space of the device.
3325          * So a read cannot be issued at this stage on any of the
3326          * registers to ensure the write into "sw_reset" register
3327          * has gone through.
3328          * Question: Is there any system call that will explicitly force
3329          * all the write commands still pending on the bus to be pushed
3330          * through?
3331          * As of now I'am just giving a 250ms delay and hoping that the
3332          * PCI write to sw_reset register is done by this time.
3333          */
3334         msleep(250);
3335         if (strstr(sp->product_name, "CX4")) {
3336                 msleep(750);
3337         }
3338
3339         /* Restore the PCI state saved during initialization. */
3340         pci_restore_state(sp->pdev);
3341         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
3342                                      pci_cmd);
3343         s2io_init_pci(sp);
3344
3345         msleep(250);
3346
3347         /* Set swapper to enable I/O register access */
3348         s2io_set_swapper(sp);
3349
3350         /* Restore the MSIX table entries from local variables */
3351         restore_xmsi_data(sp);
3352
3353         /* Clear certain PCI/PCI-X fields after reset */
3354         if (sp->device_type == XFRAME_II_DEVICE) {
3355                 /* Clear "detected parity error" bit */
3356                 pci_write_config_word(sp->pdev, PCI_STATUS, 0x8000);
3357
3358                 /* Clearing PCIX Ecc status register */
3359                 pci_write_config_dword(sp->pdev, 0x68, 0x7C);
3360
3361                 /* Clearing PCI_STATUS error reflected here */
3362                 writeq(BIT(62), &bar0->txpic_int_reg);
3363         }
3364
3365         /* Reset device statistics maintained by OS */
3366         memset(&sp->stats, 0, sizeof (struct net_device_stats));
3367
3368         /* SXE-002: Configure link and activity LED to turn it off */
3369         subid = sp->pdev->subsystem_device;
3370         if (((subid & 0xFF) >= 0x07) &&
3371             (sp->device_type == XFRAME_I_DEVICE)) {
3372                 val64 = readq(&bar0->gpio_control);
3373                 val64 |= 0x0000800000000000ULL;
3374                 writeq(val64, &bar0->gpio_control);
3375                 val64 = 0x0411040400000000ULL;
3376                 writeq(val64, (void __iomem *)bar0 + 0x2700);
3377         }
3378
3379         /*
3380          * Clear spurious ECC interrupts that would have occured on
3381          * XFRAME II cards after reset.
3382          */
3383         if (sp->device_type == XFRAME_II_DEVICE) {
3384                 val64 = readq(&bar0->pcc_err_reg);
3385                 writeq(val64, &bar0->pcc_err_reg);
3386         }
3387
3388         sp->device_enabled_once = FALSE;
3389 }
3390
3391 /**
3392  *  s2io_set_swapper - to set the swapper controle on the card
3393  *  @sp : private member of the device structure,
3394  *  pointer to the s2io_nic structure.
3395  *  Description: Function to set the swapper control on the card
3396  *  correctly depending on the 'endianness' of the system.
3397  *  Return value:
3398  *  SUCCESS on success and FAILURE on failure.
3399  */
3400
3401 static int s2io_set_swapper(nic_t * sp)
3402 {
3403         struct net_device *dev = sp->dev;
3404         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3405         u64 val64, valt, valr;
3406
3407         /*
3408          * Set proper endian settings and verify the same by reading
3409          * the PIF Feed-back register.
3410          */
3411
3412         val64 = readq(&bar0->pif_rd_swapper_fb);
3413         if (val64 != 0x0123456789ABCDEFULL) {
3414                 int i = 0;
3415                 u64 value[] = { 0xC30000C3C30000C3ULL,   /* FE=1, SE=1 */
3416                                 0x8100008181000081ULL,  /* FE=1, SE=0 */
3417                                 0x4200004242000042ULL,  /* FE=0, SE=1 */
3418                                 0};                     /* FE=0, SE=0 */
3419
3420                 while(i<4) {
3421                         writeq(value[i], &bar0->swapper_ctrl);
3422                         val64 = readq(&bar0->pif_rd_swapper_fb);
3423                         if (val64 == 0x0123456789ABCDEFULL)
3424                                 break;
3425                         i++;
3426                 }
3427                 if (i == 4) {
3428                         DBG_PRINT(ERR_DBG, "%s: Endian settings are wrong, ",
3429                                 dev->name);
3430                         DBG_PRINT(ERR_DBG, "feedback read %llx\n",
3431                                 (unsigned long long) val64);
3432                         return FAILURE;
3433                 }
3434                 valr = value[i];
3435         } else {
3436                 valr = readq(&bar0->swapper_ctrl);
3437         }
3438
3439         valt = 0x0123456789ABCDEFULL;
3440         writeq(valt, &bar0->xmsi_address);
3441         val64 = readq(&bar0->xmsi_address);
3442
3443         if(val64 != valt) {
3444                 int i = 0;
3445                 u64 value[] = { 0x00C3C30000C3C300ULL,  /* FE=1, SE=1 */
3446                                 0x0081810000818100ULL,  /* FE=1, SE=0 */
3447                                 0x0042420000424200ULL,  /* FE=0, SE=1 */
3448                                 0};                     /* FE=0, SE=0 */
3449
3450                 while(i<4) {
3451                         writeq((value[i] | valr), &bar0->swapper_ctrl);
3452                         writeq(valt, &bar0->xmsi_address);
3453                         val64 = readq(&bar0->xmsi_address);
3454                         if(val64 == valt)
3455                                 break;
3456                         i++;
3457                 }
3458                 if(i == 4) {
3459                         unsigned long long x = val64;
3460                         DBG_PRINT(ERR_DBG, "Write failed, Xmsi_addr ");
3461                         DBG_PRINT(ERR_DBG, "reads:0x%llx\n", x);
3462                         return FAILURE;
3463                 }
3464         }
3465         val64 = readq(&bar0->swapper_ctrl);
3466         val64 &= 0xFFFF000000000000ULL;
3467
3468 #ifdef  __BIG_ENDIAN
3469         /*
3470          * The device by default set to a big endian format, so a
3471          * big endian driver need not set anything.
3472          */
3473         val64 |= (SWAPPER_CTRL_TXP_FE |
3474                  SWAPPER_CTRL_TXP_SE |
3475                  SWAPPER_CTRL_TXD_R_FE |
3476                  SWAPPER_CTRL_TXD_W_FE |
3477                  SWAPPER_CTRL_TXF_R_FE |
3478                  SWAPPER_CTRL_RXD_R_FE |
3479                  SWAPPER_CTRL_RXD_W_FE |
3480                  SWAPPER_CTRL_RXF_W_FE |
3481                  SWAPPER_CTRL_XMSI_FE |
3482                  SWAPPER_CTRL_STATS_FE | SWAPPER_CTRL_STATS_SE);
3483         if (sp->intr_type == INTA)
3484                 val64 |= SWAPPER_CTRL_XMSI_SE;
3485         writeq(val64, &bar0->swapper_ctrl);
3486 #else
3487         /*
3488          * Initially we enable all bits to make it accessible by the
3489          * driver, then we selectively enable only those bits that
3490          * we want to set.
3491          */
3492         val64 |= (SWAPPER_CTRL_TXP_FE |
3493                  SWAPPER_CTRL_TXP_SE |
3494                  SWAPPER_CTRL_TXD_R_FE |
3495                  SWAPPER_CTRL_TXD_R_SE |
3496                  SWAPPER_CTRL_TXD_W_FE |
3497                  SWAPPER_CTRL_TXD_W_SE |
3498                  SWAPPER_CTRL_TXF_R_FE |
3499                  SWAPPER_CTRL_RXD_R_FE |
3500                  SWAPPER_CTRL_RXD_R_SE |
3501                  SWAPPER_CTRL_RXD_W_FE |
3502                  SWAPPER_CTRL_RXD_W_SE |
3503                  SWAPPER_CTRL_RXF_W_FE |
3504                  SWAPPER_CTRL_XMSI_FE |
3505                  SWAPPER_CTRL_STATS_FE | SWAPPER_CTRL_STATS_SE);
3506         if (sp->intr_type == INTA)
3507                 val64 |= SWAPPER_CTRL_XMSI_SE;
3508         writeq(val64, &bar0->swapper_ctrl);
3509 #endif
3510         val64 = readq(&bar0->swapper_ctrl);
3511
3512         /*
3513          * Verifying if endian settings are accurate by reading a
3514          * feedback register.
3515          */
3516         val64 = readq(&bar0->pif_rd_swapper_fb);
3517         if (val64 != 0x0123456789ABCDEFULL) {
3518                 /* Endian settings are incorrect, calls for another dekko. */
3519                 DBG_PRINT(ERR_DBG, "%s: Endian settings are wrong, ",
3520                           dev->name);
3521                 DBG_PRINT(ERR_DBG, "feedback read %llx\n",
3522                           (unsigned long long) val64);
3523                 return FAILURE;
3524         }
3525
3526         return SUCCESS;
3527 }
3528
3529 static int wait_for_msix_trans(nic_t *nic, int i)
3530 {
3531         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3532         u64 val64;
3533         int ret = 0, cnt = 0;
3534
3535         do {
3536                 val64 = readq(&bar0->xmsi_access);
3537                 if (!(val64 & BIT(15)))
3538                         break;
3539                 mdelay(1);
3540                 cnt++;
3541         } while(cnt < 5);
3542         if (cnt == 5) {
3543                 DBG_PRINT(ERR_DBG, "XMSI # %d Access failed\n", i);
3544                 ret = 1;
3545         }
3546
3547         return ret;
3548 }
3549
3550 static void restore_xmsi_data(nic_t *nic)
3551 {
3552         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3553         u64 val64;
3554         int i;
3555
3556         for (i=0; i < MAX_REQUESTED_MSI_X; i++) {
3557                 writeq(nic->msix_info[i].addr, &bar0->xmsi_address);
3558                 writeq(nic->msix_info[i].data, &bar0->xmsi_data);
3559                 val64 = (BIT(7) | BIT(15) | vBIT(i, 26, 6));
3560                 writeq(val64, &bar0->xmsi_access);
3561                 if (wait_for_msix_trans(nic, i)) {
3562                         DBG_PRINT(ERR_DBG, "failed in %s\n", __FUNCTION__);
3563                         continue;
3564                 }
3565         }
3566 }
3567
3568 static void store_xmsi_data(nic_t *nic)
3569 {
3570         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3571         u64 val64, addr, data;
3572         int i;
3573
3574         /* Store and display */
3575         for (i=0; i < MAX_REQUESTED_MSI_X; i++) {
3576                 val64 = (BIT(15) | vBIT(i, 26, 6));
3577                 writeq(val64, &bar0->xmsi_access);
3578                 if (wait_for_msix_trans(nic, i)) {
3579                         DBG_PRINT(ERR_DBG, "failed in %s\n", __FUNCTION__);
3580                         continue;
3581                 }
3582                 addr = readq(&bar0->xmsi_address);
3583                 data = readq(&bar0->xmsi_data);
3584                 if (addr && data) {
3585                         nic->msix_info[i].addr = addr;
3586                         nic->msix_info[i].data = data;
3587                 }
3588         }
3589 }
3590
3591 int s2io_enable_msi(nic_t *nic)
3592 {
3593         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3594         u16 msi_ctrl, msg_val;
3595         struct config_param *config = &nic->config;
3596         struct net_device *dev = nic->dev;
3597         u64 val64, tx_mat, rx_mat;
3598         int i, err;
3599
3600         val64 = readq(&bar0->pic_control);
3601         val64 &= ~BIT(1);
3602         writeq(val64, &bar0->pic_control);
3603
3604         err = pci_enable_msi(nic->pdev);
3605         if (err) {
3606                 DBG_PRINT(ERR_DBG, "%s: enabling MSI failed\n",
3607                           nic->dev->name);
3608                 return err;
3609         }
3610
3611         /*
3612          * Enable MSI and use MSI-1 in stead of the standard MSI-0
3613          * for interrupt handling.
3614          */
3615         pci_read_config_word(nic->pdev, 0x4c, &msg_val);
3616         msg_val ^= 0x1;
3617         pci_write_config_word(nic->pdev, 0x4c, msg_val);
3618         pci_read_config_word(nic->pdev, 0x4c, &msg_val);
3619
3620         pci_read_config_word(nic->pdev, 0x42, &msi_ctrl);
3621         msi_ctrl |= 0x10;
3622         pci_write_config_word(nic->pdev, 0x42, msi_ctrl);
3623
3624         /* program MSI-1 into all usable Tx_Mat and Rx_Mat fields */
3625         tx_mat = readq(&bar0->tx_mat0_n[0]);
3626         for (i=0; i<config->tx_fifo_num; i++) {
3627                 tx_mat |= TX_MAT_SET(i, 1);
3628         }
3629         writeq(tx_mat, &bar0->tx_mat0_n[0]);
3630
3631         rx_mat = readq(&bar0->rx_mat);
3632         for (i=0; i<config->rx_ring_num; i++) {
3633                 rx_mat |= RX_MAT_SET(i, 1);
3634         }
3635         writeq(rx_mat, &bar0->rx_mat);
3636
3637         dev->irq = nic->pdev->irq;
3638         return 0;
3639 }
3640
3641 static int s2io_enable_msi_x(nic_t *nic)
3642 {
3643         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3644         u64 tx_mat, rx_mat;
3645         u16 msi_control; /* Temp variable */
3646         int ret, i, j, msix_indx = 1;
3647
3648         nic->entries = kmalloc(MAX_REQUESTED_MSI_X * sizeof(struct msix_entry),
3649                                GFP_KERNEL);
3650         if (nic->entries == NULL) {
3651                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n", __FUNCTION__);
3652                 return -ENOMEM;
3653         }
3654         memset(nic->entries, 0, MAX_REQUESTED_MSI_X * sizeof(struct msix_entry));
3655
3656         nic->s2io_entries =
3657                 kmalloc(MAX_REQUESTED_MSI_X * sizeof(struct s2io_msix_entry),
3658                                    GFP_KERNEL);
3659         if (nic->s2io_entries == NULL) {
3660                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n", __FUNCTION__);
3661                 kfree(nic->entries);
3662                 return -ENOMEM;
3663         }
3664         memset(nic->s2io_entries, 0,
3665                MAX_REQUESTED_MSI_X * sizeof(struct s2io_msix_entry));
3666
3667         for (i=0; i< MAX_REQUESTED_MSI_X; i++) {
3668                 nic->entries[i].entry = i;
3669                 nic->s2io_entries[i].entry = i;
3670                 nic->s2io_entries[i].arg = NULL;
3671                 nic->s2io_entries[i].in_use = 0;
3672         }
3673
3674         tx_mat = readq(&bar0->tx_mat0_n[0]);
3675         for (i=0; i<nic->config.tx_fifo_num; i++, msix_indx++) {
3676                 tx_mat |= TX_MAT_SET(i, msix_indx);
3677                 nic->s2io_entries[msix_indx].arg = &nic->mac_control.fifos[i];
3678                 nic->s2io_entries[msix_indx].type = MSIX_FIFO_TYPE;
3679                 nic->s2io_entries[msix_indx].in_use = MSIX_FLG;
3680         }
3681         writeq(tx_mat, &bar0->tx_mat0_n[0]);
3682
3683         if (!nic->config.bimodal) {
3684                 rx_mat = readq(&bar0->rx_mat);
3685                 for (j=0; j<nic->config.rx_ring_num; j++, msix_indx++) {
3686                         rx_mat |= RX_MAT_SET(j, msix_indx);
3687                         nic->s2io_entries[msix_indx].arg = &nic->mac_control.rings[j];
3688                         nic->s2io_entries[msix_indx].type = MSIX_RING_TYPE;
3689                         nic->s2io_entries[msix_indx].in_use = MSIX_FLG;
3690                 }
3691                 writeq(rx_mat, &bar0->rx_mat);
3692         } else {
3693                 tx_mat = readq(&bar0->tx_mat0_n[7]);
3694                 for (j=0; j<nic->config.rx_ring_num; j++, msix_indx++) {
3695                         tx_mat |= TX_MAT_SET(i, msix_indx);
3696                         nic->s2io_entries[msix_indx].arg = &nic->mac_control.rings[j];
3697                         nic->s2io_entries[msix_indx].type = MSIX_RING_TYPE;
3698                         nic->s2io_entries[msix_indx].in_use = MSIX_FLG;
3699                 }
3700                 writeq(tx_mat, &bar0->tx_mat0_n[7]);
3701         }
3702
3703         nic->avail_msix_vectors = 0;
3704         ret = pci_enable_msix(nic->pdev, nic->entries, MAX_REQUESTED_MSI_X);
3705         /* We fail init if error or we get less vectors than min required */
3706         if (ret >= (nic->config.tx_fifo_num + nic->config.rx_ring_num + 1)) {
3707                 nic->avail_msix_vectors = ret;
3708                 ret = pci_enable_msix(nic->pdev, nic->entries, ret);
3709         }
3710         if (ret) {
3711                 DBG_PRINT(ERR_DBG, "%s: Enabling MSIX failed\n", nic->dev->name);
3712                 kfree(nic->entries);
3713                 kfree(nic->s2io_entries);
3714                 nic->entries = NULL;
3715                 nic->s2io_entries = NULL;
3716                 nic->avail_msix_vectors = 0;
3717                 return -ENOMEM;
3718         }
3719         if (!nic->avail_msix_vectors)
3720                 nic->avail_msix_vectors = MAX_REQUESTED_MSI_X;
3721
3722         /*
3723          * To enable MSI-X, MSI also needs to be enabled, due to a bug
3724          * in the herc NIC. (Temp change, needs to be removed later)
3725          */
3726         pci_read_config_word(nic->pdev, 0x42, &msi_control);
3727         msi_control |= 0x1; /* Enable MSI */
3728         pci_write_config_word(nic->pdev, 0x42, msi_control);
3729
3730         return 0;
3731 }
3732
3733 /* ********************************************************* *
3734  * Functions defined below concern the OS part of the driver *
3735  * ********************************************************* */
3736
3737 /**
3738  *  s2io_open - open entry point of the driver
3739  *  @dev : pointer to the device structure.
3740  *  Description:
3741  *  This function is the open entry point of the driver. It mainly calls a
3742  *  function to allocate Rx buffers and inserts them into the buffer
3743  *  descriptors and then enables the Rx part of the NIC.
3744  *  Return value:
3745  *  0 on success and an appropriate (-)ve integer as defined in errno.h
3746  *   file on failure.
3747  */
3748
3749 static int s2io_open(struct net_device *dev)
3750 {
3751         nic_t *sp = dev->priv;
3752         int err = 0;
3753
3754         /*
3755          * Make sure you have link off by default every time
3756          * Nic is initialized
3757          */
3758         netif_carrier_off(dev);
3759         sp->last_link_state = 0;
3760
3761         /* Initialize H/W and enable interrupts */
3762         err = s2io_card_up(sp);
3763         if (err) {
3764                 DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
3765                           dev->name);
3766                 goto hw_init_failed;
3767         }
3768
3769         if (s2io_set_mac_addr(dev, dev->dev_addr) == FAILURE) {
3770                 DBG_PRINT(ERR_DBG, "Set Mac Address Failed\n");
3771                 s2io_card_down(sp);
3772                 err = -ENODEV;
3773                 goto hw_init_failed;
3774         }
3775
3776         netif_start_queue(dev);
3777         return 0;
3778
3779 hw_init_failed:
3780         if (sp->intr_type == MSI_X) {
3781                 if (sp->entries)
3782                         kfree(sp->entries);
3783                 if (sp->s2io_entries)
3784                         kfree(sp->s2io_entries);
3785         }
3786         return err;
3787 }
3788
3789 /**
3790  *  s2io_close -close entry point of the driver
3791  *  @dev : device pointer.
3792  *  Description:
3793  *  This is the stop entry point of the driver. It needs to undo exactly
3794  *  whatever was done by the open entry point,thus it's usually referred to
3795  *  as the close function.Among other things this function mainly stops the
3796  *  Rx side of the NIC and frees all the Rx buffers in the Rx rings.
3797  *  Return value:
3798  *  0 on success and an appropriate (-)ve integer as defined in errno.h
3799  *  file on failure.
3800  */
3801
3802 static int s2io_close(struct net_device *dev)
3803 {
3804         nic_t *sp = dev->priv;
3805
3806         flush_scheduled_work();
3807         netif_stop_queue(dev);
3808         /* Reset card, kill tasklet and free Tx and Rx buffers. */
3809         s2io_card_down(sp);
3810
3811         sp->device_close_flag = TRUE;   /* Device is shut down. */
3812         return 0;
3813 }
3814
3815 /**
3816  *  s2io_xmit - Tx entry point of te driver
3817  *  @skb : the socket buffer containing the Tx data.
3818  *  @dev : device pointer.
3819  *  Description :
3820  *  This function is the Tx entry point of the driver. S2IO NIC supports
3821  *  certain protocol assist features on Tx side, namely  CSO, S/G, LSO.
3822  *  NOTE: when device cant queue the pkt,just the trans_start variable will
3823  *  not be upadted.
3824  *  Return value:
3825  *  0 on success & 1 on failure.
3826  */
3827
3828 static int s2io_xmit(struct sk_buff *skb, struct net_device *dev)
3829 {
3830         nic_t *sp = dev->priv;
3831         u16 frg_cnt, frg_len, i, queue, queue_len, put_off, get_off;
3832         register u64 val64;
3833         TxD_t *txdp;
3834         TxFIFO_element_t __iomem *tx_fifo;
3835         unsigned long flags;
3836         u16 vlan_tag = 0;
3837         int vlan_priority = 0;
3838         mac_info_t *mac_control;
3839         struct config_param *config;
3840         int offload_type;
3841
3842         mac_control = &sp->mac_control;
3843         config = &sp->config;
3844
3845         DBG_PRINT(TX_DBG, "%s: In Neterion Tx routine\n", dev->name);
3846         spin_lock_irqsave(&sp->tx_lock, flags);
3847         if (atomic_read(&sp->card_state) == CARD_DOWN) {
3848                 DBG_PRINT(TX_DBG, "%s: Card going down for reset\n",
3849                           dev->name);
3850                 spin_unlock_irqrestore(&sp->tx_lock, flags);
3851                 dev_kfree_skb(skb);
3852                 return 0;
3853         }
3854
3855         queue = 0;
3856
3857         /* Get Fifo number to Transmit based on vlan priority */
3858         if (sp->vlgrp && vlan_tx_tag_present(skb)) {
3859                 vlan_tag = vlan_tx_tag_get(skb);
3860                 vlan_priority = vlan_tag >> 13;
3861                 queue = config->fifo_mapping[vlan_priority];
3862         }
3863
3864         put_off = (u16) mac_control->fifos[queue].tx_curr_put_info.offset;
3865         get_off = (u16) mac_control->fifos[queue].tx_curr_get_info.offset;
3866         txdp = (TxD_t *) mac_control->fifos[queue].list_info[put_off].
3867                 list_virt_addr;
3868
3869         queue_len = mac_control->fifos[queue].tx_curr_put_info.fifo_len + 1;
3870         /* Avoid "put" pointer going beyond "get" pointer */
3871         if (txdp->Host_Control ||
3872                    ((put_off+1) == queue_len ? 0 : (put_off+1)) == get_off) {
3873                 DBG_PRINT(TX_DBG, "Error in xmit, No free TXDs.\n");
3874                 netif_stop_queue(dev);
3875                 dev_kfree_skb(skb);
3876                 spin_unlock_irqrestore(&sp->tx_lock, flags);
3877                 return 0;
3878         }
3879
3880         /* A buffer with no data will be dropped */
3881         if (!skb->len) {
3882                 DBG_PRINT(TX_DBG, "%s:Buffer has no data..\n", dev->name);
3883                 dev_kfree_skb(skb);
3884                 spin_unlock_irqrestore(&sp->tx_lock, flags);
3885                 return 0;
3886         }
3887
3888         offload_type = s2io_offload_type(skb);
3889 #ifdef NETIF_F_TSO
3890         if (offload_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)) {
3891                 txdp->Control_1 |= TXD_TCP_LSO_EN;
3892                 txdp->Control_1 |= TXD_TCP_LSO_MSS(s2io_tcp_mss(skb));
3893         }
3894 #endif
3895         if (skb->ip_summed == CHECKSUM_HW) {
3896                 txdp->Control_2 |=
3897                     (TXD_TX_CKO_IPV4_EN | TXD_TX_CKO_TCP_EN |
3898                      TXD_TX_CKO_UDP_EN);
3899         }
3900         txdp->Control_1 |= TXD_GATHER_CODE_FIRST;
3901         txdp->Control_1 |= TXD_LIST_OWN_XENA;
3902         txdp->Control_2 |= config->tx_intr_type;
3903
3904         if (sp->vlgrp && vlan_tx_tag_present(skb)) {
3905                 txdp->Control_2 |= TXD_VLAN_ENABLE;
3906                 txdp->Control_2 |= TXD_VLAN_TAG(vlan_tag);
3907         }
3908
3909         frg_len = skb->len - skb->data_len;
3910         if (offload_type == SKB_GSO_UDP) {
3911                 int ufo_size;
3912
3913                 ufo_size = s2io_udp_mss(skb);
3914                 ufo_size &= ~7;
3915                 txdp->Control_1 |= TXD_UFO_EN;
3916                 txdp->Control_1 |= TXD_UFO_MSS(ufo_size);
3917                 txdp->Control_1 |= TXD_BUFFER0_SIZE(8);
3918 #ifdef __BIG_ENDIAN
3919                 sp->ufo_in_band_v[put_off] =
3920                                 (u64)skb_shinfo(skb)->ip6_frag_id;
3921 #else
3922                 sp->ufo_in_band_v[put_off] =
3923                                 (u64)skb_shinfo(skb)->ip6_frag_id << 32;
3924 #endif
3925                 txdp->Host_Control = (unsigned long)sp->ufo_in_band_v;
3926                 txdp->Buffer_Pointer = pci_map_single(sp->pdev,
3927                                         sp->ufo_in_band_v,
3928                                         sizeof(u64), PCI_DMA_TODEVICE);
3929                 txdp++;
3930         }
3931
3932         txdp->Buffer_Pointer = pci_map_single
3933             (sp->pdev, skb->data, frg_len, PCI_DMA_TODEVICE);
3934         txdp->Host_Control = (unsigned long) skb;
3935         txdp->Control_1 |= TXD_BUFFER0_SIZE(frg_len);
3936         if (offload_type == SKB_GSO_UDP)
3937                 txdp->Control_1 |= TXD_UFO_EN;
3938
3939         frg_cnt = skb_shinfo(skb)->nr_frags;
3940         /* For fragmented SKB. */
3941         for (i = 0; i < frg_cnt; i++) {
3942                 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
3943                 /* A '0' length fragment will be ignored */
3944                 if (!frag->size)
3945                         continue;
3946                 txdp++;
3947                 txdp->Buffer_Pointer = (u64) pci_map_page
3948                     (sp->pdev, frag->page, frag->page_offset,
3949                      frag->size, PCI_DMA_TODEVICE);
3950                 txdp->Control_1 = TXD_BUFFER0_SIZE(frag->size);
3951                 if (offload_type == SKB_GSO_UDP)
3952                         txdp->Control_1 |= TXD_UFO_EN;
3953         }
3954         txdp->Control_1 |= TXD_GATHER_CODE_LAST;
3955
3956         if (offload_type == SKB_GSO_UDP)
3957                 frg_cnt++; /* as Txd0 was used for inband header */
3958
3959         tx_fifo = mac_control->tx_FIFO_start[queue];
3960         val64 = mac_control->fifos[queue].list_info[put_off].list_phy_addr;
3961         writeq(val64, &tx_fifo->TxDL_Pointer);
3962
3963         val64 = (TX_FIFO_LAST_TXD_NUM(frg_cnt) | TX_FIFO_FIRST_LIST |
3964                  TX_FIFO_LAST_LIST);
3965         if (offload_type)
3966                 val64 |= TX_FIFO_SPECIAL_FUNC;
3967
3968         writeq(val64, &tx_fifo->List_Control);
3969
3970         mmiowb();
3971
3972         put_off++;
3973         if (put_off == mac_control->fifos[queue].tx_curr_put_info.fifo_len + 1)
3974                 put_off = 0;
3975         mac_control->fifos[queue].tx_curr_put_info.offset = put_off;
3976
3977         /* Avoid "put" pointer going beyond "get" pointer */
3978         if (((put_off+1) == queue_len ? 0 : (put_off+1)) == get_off) {
3979                 sp->mac_control.stats_info->sw_stat.fifo_full_cnt++;
3980                 DBG_PRINT(TX_DBG,
3981                           "No free TxDs for xmit, Put: 0x%x Get:0x%x\n",
3982                           put_off, get_off);
3983                 netif_stop_queue(dev);
3984         }
3985
3986         dev->trans_start = jiffies;
3987         spin_unlock_irqrestore(&sp->tx_lock, flags);
3988
3989         return 0;
3990 }
3991
3992 static void
3993 s2io_alarm_handle(unsigned long data)
3994 {
3995         nic_t *sp = (nic_t *)data;
3996
3997         alarm_intr_handler(sp);
3998         mod_timer(&sp->alarm_timer, jiffies + HZ / 2);
3999 }
4000
4001 static int s2io_chk_rx_buffers(nic_t *sp, int rng_n)
4002 {
4003         int rxb_size, level;
4004
4005         if (!sp->lro) {
4006                 rxb_size = atomic_read(&sp->rx_bufs_left[rng_n]);
4007                 level = rx_buffer_level(sp, rxb_size, rng_n);
4008
4009                 if ((level == PANIC) && (!TASKLET_IN_USE)) {
4010                         int ret;
4011                         DBG_PRINT(INTR_DBG, "%s: Rx BD hit ", __FUNCTION__);
4012                         DBG_PRINT(INTR_DBG, "PANIC levels\n");
4013                         if ((ret = fill_rx_buffers(sp, rng_n)) == -ENOMEM) {
4014                                 DBG_PRINT(ERR_DBG, "Out of memory in %s",
4015                                           __FUNCTION__);
4016                                 clear_bit(0, (&sp->tasklet_status));
4017                                 return -1;
4018                         }
4019                         clear_bit(0, (&sp->tasklet_status));
4020                 } else if (level == LOW)
4021                         tasklet_schedule(&sp->task);
4022
4023         } else if (fill_rx_buffers(sp, rng_n) == -ENOMEM) {
4024                         DBG_PRINT(ERR_DBG, "%s:Out of memory", sp->dev->name);
4025                         DBG_PRINT(ERR_DBG, " in Rx Intr!!\n");
4026         }
4027         return 0;
4028 }
4029
4030 static irqreturn_t
4031 s2io_msi_handle(int irq, void *dev_id, struct pt_regs *regs)
4032 {
4033         struct net_device *dev = (struct net_device *) dev_id;
4034         nic_t *sp = dev->priv;
4035         int i;
4036         mac_info_t *mac_control;
4037         struct config_param *config;
4038
4039         atomic_inc(&sp->isr_cnt);
4040         mac_control = &sp->mac_control;
4041         config = &sp->config;
4042         DBG_PRINT(INTR_DBG, "%s: MSI handler\n", __FUNCTION__);
4043
4044         /* If Intr is because of Rx Traffic */
4045         for (i = 0; i < config->rx_ring_num; i++)
4046                 rx_intr_handler(&mac_control->rings[i]);
4047
4048         /* If Intr is because of Tx Traffic */
4049         for (i = 0; i < config->tx_fifo_num; i++)
4050                 tx_intr_handler(&mac_control->fifos[i]);
4051
4052         /*
4053          * If the Rx buffer count is below the panic threshold then
4054          * reallocate the buffers from the interrupt handler itself,
4055          * else schedule a tasklet to reallocate the buffers.
4056          */
4057         for (i = 0; i < config->rx_ring_num; i++)
4058                 s2io_chk_rx_buffers(sp, i);
4059
4060         atomic_dec(&sp->isr_cnt);
4061         return IRQ_HANDLED;
4062 }
4063
4064 static irqreturn_t
4065 s2io_msix_ring_handle(int irq, void *dev_id, struct pt_regs *regs)
4066 {
4067         ring_info_t *ring = (ring_info_t *)dev_id;
4068         nic_t *sp = ring->nic;
4069
4070         atomic_inc(&sp->isr_cnt);
4071
4072         rx_intr_handler(ring);
4073         s2io_chk_rx_buffers(sp, ring->ring_no);
4074
4075         atomic_dec(&sp->isr_cnt);
4076         return IRQ_HANDLED;
4077 }
4078
4079 static irqreturn_t
4080 s2io_msix_fifo_handle(int irq, void *dev_id, struct pt_regs *regs)
4081 {
4082         fifo_info_t *fifo = (fifo_info_t *)dev_id;
4083         nic_t *sp = fifo->nic;
4084
4085         atomic_inc(&sp->isr_cnt);
4086         tx_intr_handler(fifo);
4087         atomic_dec(&sp->isr_cnt);
4088         return IRQ_HANDLED;
4089 }
4090 static void s2io_txpic_intr_handle(nic_t *sp)
4091 {
4092         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4093         u64 val64;
4094
4095         val64 = readq(&bar0->pic_int_status);
4096         if (val64 & PIC_INT_GPIO) {
4097                 val64 = readq(&bar0->gpio_int_reg);
4098                 if ((val64 & GPIO_INT_REG_LINK_DOWN) &&
4099                     (val64 & GPIO_INT_REG_LINK_UP)) {
4100                         /*
4101                          * This is unstable state so clear both up/down
4102                          * interrupt and adapter to re-evaluate the link state.
4103                          */
4104                         val64 |=  GPIO_INT_REG_LINK_DOWN;
4105                         val64 |= GPIO_INT_REG_LINK_UP;
4106                         writeq(val64, &bar0->gpio_int_reg);
4107                         val64 = readq(&bar0->gpio_int_mask);
4108                         val64 &= ~(GPIO_INT_MASK_LINK_UP |
4109                                    GPIO_INT_MASK_LINK_DOWN);
4110                         writeq(val64, &bar0->gpio_int_mask);
4111                 }
4112                 else if (val64 & GPIO_INT_REG_LINK_UP) {
4113                         val64 = readq(&bar0->adapter_status);
4114                         if (verify_xena_quiescence(sp, val64,
4115                                                    sp->device_enabled_once)) {
4116                                 /* Enable Adapter */
4117                                 val64 = readq(&bar0->adapter_control);
4118                                 val64 |= ADAPTER_CNTL_EN;
4119                                 writeq(val64, &bar0->adapter_control);
4120                                 val64 |= ADAPTER_LED_ON;
4121                                 writeq(val64, &bar0->adapter_control);
4122                                 if (!sp->device_enabled_once)
4123                                         sp->device_enabled_once = 1;
4124
4125                                 s2io_link(sp, LINK_UP);
4126                                 /*
4127                                  * unmask link down interrupt and mask link-up
4128                                  * intr
4129                                  */
4130                                 val64 = readq(&bar0->gpio_int_mask);
4131                                 val64 &= ~GPIO_INT_MASK_LINK_DOWN;
4132                                 val64 |= GPIO_INT_MASK_LINK_UP;
4133                                 writeq(val64, &bar0->gpio_int_mask);
4134
4135                         }
4136                 }else if (val64 & GPIO_INT_REG_LINK_DOWN) {
4137                         val64 = readq(&bar0->adapter_status);
4138                         if (verify_xena_quiescence(sp, val64,
4139                                                    sp->device_enabled_once)) {
4140                                 s2io_link(sp, LINK_DOWN);
4141                                 /* Link is down so unmaks link up interrupt */
4142                                 val64 = readq(&bar0->gpio_int_mask);
4143                                 val64 &= ~GPIO_INT_MASK_LINK_UP;
4144                                 val64 |= GPIO_INT_MASK_LINK_DOWN;
4145                                 writeq(val64, &bar0->gpio_int_mask);
4146                         }
4147                 }
4148         }
4149         val64 = readq(&bar0->gpio_int_mask);
4150 }
4151
4152 /**
4153  *  s2io_isr - ISR handler of the device .
4154  *  @irq: the irq of the device.
4155  *  @dev_id: a void pointer to the dev structure of the NIC.
4156  *  @pt_regs: pointer to the registers pushed on the stack.
4157  *  Description:  This function is the ISR handler of the device. It
4158  *  identifies the reason for the interrupt and calls the relevant
4159  *  service routines. As a contongency measure, this ISR allocates the
4160  *  recv buffers, if their numbers are below the panic value which is
4161  *  presently set to 25% of the original number of rcv buffers allocated.
4162  *  Return value:
4163  *   IRQ_HANDLED: will be returned if IRQ was handled by this routine
4164  *   IRQ_NONE: will be returned if interrupt is not from our device
4165  */
4166 static irqreturn_t s2io_isr(int irq, void *dev_id, struct pt_regs *regs)
4167 {
4168         struct net_device *dev = (struct net_device *) dev_id;
4169         nic_t *sp = dev->priv;
4170         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4171         int i;
4172         u64 reason = 0, val64, org_mask;
4173         mac_info_t *mac_control;
4174         struct config_param *config;
4175
4176         atomic_inc(&sp->isr_cnt);
4177         mac_control = &sp->mac_control;
4178         config = &sp->config;
4179
4180         /*
4181          * Identify the cause for interrupt and call the appropriate
4182          * interrupt handler. Causes for the interrupt could be;
4183          * 1. Rx of packet.
4184          * 2. Tx complete.
4185          * 3. Link down.
4186          * 4. Error in any functional blocks of the NIC.
4187          */
4188         reason = readq(&bar0->general_int_status);
4189
4190         if (!reason) {
4191                 /* The interrupt was not raised by Xena. */
4192                 atomic_dec(&sp->isr_cnt);
4193                 return IRQ_NONE;
4194         }
4195
4196         val64 = 0xFFFFFFFFFFFFFFFFULL;
4197         /* Store current mask before masking all interrupts */
4198         org_mask = readq(&bar0->general_int_mask);
4199         writeq(val64, &bar0->general_int_mask);
4200
4201 #ifdef CONFIG_S2IO_NAPI
4202         if (reason & GEN_INTR_RXTRAFFIC) {
4203                 if (netif_rx_schedule_prep(dev)) {
4204                         writeq(val64, &bar0->rx_traffic_mask);
4205                         __netif_rx_schedule(dev);
4206                 }
4207         }
4208 #else
4209         /*
4210          * Rx handler is called by default, without checking for the
4211          * cause of interrupt.
4212          * rx_traffic_int reg is an R1 register, writing all 1's
4213          * will ensure that the actual interrupt causing bit get's
4214          * cleared and hence a read can be avoided.
4215          */
4216         writeq(val64, &bar0->rx_traffic_int);
4217         for (i = 0; i < config->rx_ring_num; i++) {
4218                 rx_intr_handler(&mac_control->rings[i]);
4219         }
4220 #endif
4221
4222         /*
4223          * tx_traffic_int reg is an R1 register, writing all 1's
4224          * will ensure that the actual interrupt causing bit get's
4225          * cleared and hence a read can be avoided.
4226          */
4227         writeq(val64, &bar0->tx_traffic_int);
4228
4229         for (i = 0; i < config->tx_fifo_num; i++)
4230                 tx_intr_handler(&mac_control->fifos[i]);
4231
4232         if (reason & GEN_INTR_TXPIC)
4233                 s2io_txpic_intr_handle(sp);
4234         /*
4235          * If the Rx buffer count is below the panic threshold then
4236          * reallocate the buffers from the interrupt handler itself,
4237          * else schedule a tasklet to reallocate the buffers.
4238          */
4239 #ifndef CONFIG_S2IO_NAPI
4240         for (i = 0; i < config->rx_ring_num; i++)
4241                 s2io_chk_rx_buffers(sp, i);
4242 #endif
4243         writeq(org_mask, &bar0->general_int_mask);
4244         atomic_dec(&sp->isr_cnt);
4245         return IRQ_HANDLED;
4246 }
4247
4248 /**
4249  * s2io_updt_stats -
4250  */
4251 static void s2io_updt_stats(nic_t *sp)
4252 {
4253         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4254         u64 val64;
4255         int cnt = 0;
4256
4257         if (atomic_read(&sp->card_state) == CARD_UP) {
4258                 /* Apprx 30us on a 133 MHz bus */
4259                 val64 = SET_UPDT_CLICKS(10) |
4260                         STAT_CFG_ONE_SHOT_EN | STAT_CFG_STAT_EN;
4261                 writeq(val64, &bar0->stat_cfg);
4262                 do {
4263                         udelay(100);
4264                         val64 = readq(&bar0->stat_cfg);
4265                         if (!(val64 & BIT(0)))
4266                                 break;
4267                         cnt++;
4268                         if (cnt == 5)
4269                                 break; /* Updt failed */
4270                 } while(1);
4271         } else {
4272                 memset(sp->mac_control.stats_info, 0, sizeof(StatInfo_t));
4273         }
4274 }
4275
4276 /**
4277  *  s2io_get_stats - Updates the device statistics structure.
4278  *  @dev : pointer to the device structure.
4279  *  Description:
4280  *  This function updates the device statistics structure in the s2io_nic
4281  *  structure and returns a pointer to the same.
4282  *  Return value:
4283  *  pointer to the updated net_device_stats structure.
4284  */
4285
4286 static struct net_device_stats *s2io_get_stats(struct net_device *dev)
4287 {
4288         nic_t *sp = dev->priv;
4289         mac_info_t *mac_control;
4290         struct config_param *config;
4291
4292
4293         mac_control = &sp->mac_control;
4294         config = &sp->config;
4295
4296         /* Configure Stats for immediate updt */
4297         s2io_updt_stats(sp);
4298
4299         sp->stats.tx_packets =
4300                 le32_to_cpu(mac_control->stats_info->tmac_frms);
4301         sp->stats.tx_errors =
4302                 le32_to_cpu(mac_control->stats_info->tmac_any_err_frms);
4303         sp->stats.rx_errors =
4304                 le32_to_cpu(mac_control->stats_info->rmac_drop_frms);
4305         sp->stats.multicast =
4306                 le32_to_cpu(mac_control->stats_info->rmac_vld_mcst_frms);
4307         sp->stats.rx_length_errors =
4308                 le32_to_cpu(mac_control->stats_info->rmac_long_frms);
4309
4310         return (&sp->stats);
4311 }
4312
4313 /**
4314  *  s2io_set_multicast - entry point for multicast address enable/disable.
4315  *  @dev : pointer to the device structure
4316  *  Description:
4317  *  This function is a driver entry point which gets called by the kernel
4318  *  whenever multicast addresses must be enabled/disabled. This also gets
4319  *  called to set/reset promiscuous mode. Depending on the deivce flag, we
4320  *  determine, if multicast address must be enabled or if promiscuous mode
4321  *  is to be disabled etc.
4322  *  Return value:
4323  *  void.
4324  */
4325
4326 static void s2io_set_multicast(struct net_device *dev)
4327 {
4328         int i, j, prev_cnt;
4329         struct dev_mc_list *mclist;
4330         nic_t *sp = dev->priv;
4331         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4332         u64 val64 = 0, multi_mac = 0x010203040506ULL, mask =
4333             0xfeffffffffffULL;
4334         u64 dis_addr = 0xffffffffffffULL, mac_addr = 0;
4335         void __iomem *add;
4336
4337         if ((dev->flags & IFF_ALLMULTI) && (!sp->m_cast_flg)) {
4338                 /*  Enable all Multicast addresses */
4339                 writeq(RMAC_ADDR_DATA0_MEM_ADDR(multi_mac),
4340                        &bar0->rmac_addr_data0_mem);
4341                 writeq(RMAC_ADDR_DATA1_MEM_MASK(mask),
4342                        &bar0->rmac_addr_data1_mem);
4343                 val64 = RMAC_ADDR_CMD_MEM_WE |
4344                     RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4345                     RMAC_ADDR_CMD_MEM_OFFSET(MAC_MC_ALL_MC_ADDR_OFFSET);
4346                 writeq(val64, &bar0->rmac_addr_cmd_mem);
4347                 /* Wait till command completes */
4348                 wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4349                                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING);
4350
4351                 sp->m_cast_flg = 1;
4352                 sp->all_multi_pos = MAC_MC_ALL_MC_ADDR_OFFSET;
4353         } else if ((dev->flags & IFF_ALLMULTI) && (sp->m_cast_flg)) {
4354                 /*  Disable all Multicast addresses */
4355                 writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
4356                        &bar0->rmac_addr_data0_mem);
4357                 writeq(RMAC_ADDR_DATA1_MEM_MASK(0x0),
4358                        &bar0->rmac_addr_data1_mem);
4359                 val64 = RMAC_ADDR_CMD_MEM_WE |
4360                     RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4361                     RMAC_ADDR_CMD_MEM_OFFSET(sp->all_multi_pos);
4362                 writeq(val64, &bar0->rmac_addr_cmd_mem);
4363                 /* Wait till command completes */
4364                 wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4365                                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING);
4366
4367                 sp->m_cast_flg = 0;
4368                 sp->all_multi_pos = 0;
4369         }
4370
4371         if ((dev->flags & IFF_PROMISC) && (!sp->promisc_flg)) {
4372                 /*  Put the NIC into promiscuous mode */
4373                 add = &bar0->mac_cfg;
4374                 val64 = readq(&bar0->mac_cfg);
4375                 val64 |= MAC_CFG_RMAC_PROM_ENABLE;
4376
4377                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4378                 writel((u32) val64, add);
4379                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4380                 writel((u32) (val64 >> 32), (add + 4));
4381
4382                 val64 = readq(&bar0->mac_cfg);
4383                 sp->promisc_flg = 1;
4384                 DBG_PRINT(INFO_DBG, "%s: entered promiscuous mode\n",
4385                           dev->name);
4386         } else if (!(dev->flags & IFF_PROMISC) && (sp->promisc_flg)) {
4387                 /*  Remove the NIC from promiscuous mode */
4388                 add = &bar0->mac_cfg;
4389                 val64 = readq(&bar0->mac_cfg);
4390                 val64 &= ~MAC_CFG_RMAC_PROM_ENABLE;
4391
4392                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4393                 writel((u32) val64, add);
4394                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4395                 writel((u32) (val64 >> 32), (add + 4));
4396
4397                 val64 = readq(&bar0->mac_cfg);
4398                 sp->promisc_flg = 0;
4399                 DBG_PRINT(INFO_DBG, "%s: left promiscuous mode\n",
4400                           dev->name);
4401         }
4402
4403         /*  Update individual M_CAST address list */
4404         if ((!sp->m_cast_flg) && dev->mc_count) {
4405                 if (dev->mc_count >
4406                     (MAX_ADDRS_SUPPORTED - MAC_MC_ADDR_START_OFFSET - 1)) {
4407                         DBG_PRINT(ERR_DBG, "%s: No more Rx filters ",
4408                                   dev->name);
4409                         DBG_PRINT(ERR_DBG, "can be added, please enable ");
4410                         DBG_PRINT(ERR_DBG, "ALL_MULTI instead\n");
4411                         return;
4412                 }
4413
4414                 prev_cnt = sp->mc_addr_count;
4415                 sp->mc_addr_count = dev->mc_count;
4416
4417                 /* Clear out the previous list of Mc in the H/W. */
4418                 for (i = 0; i < prev_cnt; i++) {
4419                         writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
4420                                &bar0->rmac_addr_data0_mem);
4421                         writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
4422                                 &bar0->rmac_addr_data1_mem);
4423                         val64 = RMAC_ADDR_CMD_MEM_WE |
4424                             RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4425                             RMAC_ADDR_CMD_MEM_OFFSET
4426                             (MAC_MC_ADDR_START_OFFSET + i);
4427                         writeq(val64, &bar0->rmac_addr_cmd_mem);
4428
4429                         /* Wait for command completes */
4430                         if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4431                                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING)) {
4432                                 DBG_PRINT(ERR_DBG, "%s: Adding ",
4433                                           dev->name);
4434                                 DBG_PRINT(ERR_DBG, "Multicasts failed\n");
4435                                 return;
4436                         }
4437                 }
4438
4439                 /* Create the new Rx filter list and update the same in H/W. */
4440                 for (i = 0, mclist = dev->mc_list; i < dev->mc_count;
4441                      i++, mclist = mclist->next) {
4442                         memcpy(sp->usr_addrs[i].addr, mclist->dmi_addr,
4443                                ETH_ALEN);
4444                         mac_addr = 0;
4445                         for (j = 0; j < ETH_ALEN; j++) {
4446                                 mac_addr |= mclist->dmi_addr[j];
4447                                 mac_addr <<= 8;
4448                         }
4449                         mac_addr >>= 8;
4450                         writeq(RMAC_ADDR_DATA0_MEM_ADDR(mac_addr),
4451                                &bar0->rmac_addr_data0_mem);
4452                         writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
4453                                 &bar0->rmac_addr_data1_mem);
4454                         val64 = RMAC_ADDR_CMD_MEM_WE |
4455                             RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4456                             RMAC_ADDR_CMD_MEM_OFFSET
4457                             (i + MAC_MC_ADDR_START_OFFSET);
4458                         writeq(val64, &bar0->rmac_addr_cmd_mem);
4459
4460                         /* Wait for command completes */
4461                         if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4462                                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING)) {
4463                                 DBG_PRINT(ERR_DBG, "%s: Adding ",
4464                                           dev->name);
4465                                 DBG_PRINT(ERR_DBG, "Multicasts failed\n");
4466                                 return;
4467                         }
4468                 }
4469         }
4470 }
4471
4472 /**
4473  *  s2io_set_mac_addr - Programs the Xframe mac address
4474  *  @dev : pointer to the device structure.
4475  *  @addr: a uchar pointer to the new mac address which is to be set.
4476  *  Description : This procedure will program the Xframe to receive
4477  *  frames with new Mac Address
4478  *  Return value: SUCCESS on success and an appropriate (-)ve integer
4479  *  as defined in errno.h file on failure.
4480  */
4481
4482 static int s2io_set_mac_addr(struct net_device *dev, u8 * addr)
4483 {
4484         nic_t *sp = dev->priv;
4485         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4486         register u64 val64, mac_addr = 0;
4487         int i;
4488
4489         /*
4490          * Set the new MAC address as the new unicast filter and reflect this
4491          * change on the device address registered with the OS. It will be
4492          * at offset 0.
4493          */
4494         for (i = 0; i < ETH_ALEN; i++) {
4495                 mac_addr <<= 8;
4496                 mac_addr |= addr[i];
4497         }
4498
4499         writeq(RMAC_ADDR_DATA0_MEM_ADDR(mac_addr),
4500                &bar0->rmac_addr_data0_mem);
4501
4502         val64 =
4503             RMAC_ADDR_CMD_MEM_WE | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4504             RMAC_ADDR_CMD_MEM_OFFSET(0);
4505         writeq(val64, &bar0->rmac_addr_cmd_mem);
4506         /* Wait till command completes */
4507         if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4508                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING)) {
4509                 DBG_PRINT(ERR_DBG, "%s: set_mac_addr failed\n", dev->name);
4510                 return FAILURE;
4511         }
4512
4513         return SUCCESS;
4514 }
4515
4516 /**
4517  * s2io_ethtool_sset - Sets different link parameters.
4518  * @sp : private member of the device structure, which is a pointer to the  * s2io_nic structure.
4519  * @info: pointer to the structure with parameters given by ethtool to set
4520  * link information.
4521  * Description:
4522  * The function sets different link parameters provided by the user onto
4523  * the NIC.
4524  * Return value:
4525  * 0 on success.
4526 */
4527
4528 static int s2io_ethtool_sset(struct net_device *dev,
4529                              struct ethtool_cmd *info)
4530 {
4531         nic_t *sp = dev->priv;
4532         if ((info->autoneg == AUTONEG_ENABLE) ||
4533             (info->speed != SPEED_10000) || (info->duplex != DUPLEX_FULL))
4534                 return -EINVAL;
4535         else {
4536                 s2io_close(sp->dev);
4537                 s2io_open(sp->dev);
4538         }
4539
4540         return 0;
4541 }
4542
4543 /**
4544  * s2io_ethtol_gset - Return link specific information.
4545  * @sp : private member of the device structure, pointer to the
4546  *      s2io_nic structure.
4547  * @info : pointer to the structure with parameters given by ethtool
4548  * to return link information.
4549  * Description:
4550  * Returns link specific information like speed, duplex etc.. to ethtool.
4551  * Return value :
4552  * return 0 on success.
4553  */
4554
4555 static int s2io_ethtool_gset(struct net_device *dev, struct ethtool_cmd *info)
4556 {
4557         nic_t *sp = dev->priv;
4558         info->supported = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
4559         info->advertising = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
4560         info->port = PORT_FIBRE;
4561         /* info->transceiver?? TODO */
4562
4563         if (netif_carrier_ok(sp->dev)) {
4564                 info->speed = 10000;
4565                 info->duplex = DUPLEX_FULL;
4566         } else {
4567                 info->speed = -1;
4568                 info->duplex = -1;
4569         }
4570
4571         info->autoneg = AUTONEG_DISABLE;
4572         return 0;
4573 }
4574
4575 /**
4576  * s2io_ethtool_gdrvinfo - Returns driver specific information.
4577  * @sp : private member of the device structure, which is a pointer to the
4578  * s2io_nic structure.
4579  * @info : pointer to the structure with parameters given by ethtool to
4580  * return driver information.
4581  * Description:
4582  * Returns driver specefic information like name, version etc.. to ethtool.
4583  * Return value:
4584  *  void
4585  */
4586
4587 static void s2io_ethtool_gdrvinfo(struct net_device *dev,
4588                                   struct ethtool_drvinfo *info)
4589 {
4590         nic_t *sp = dev->priv;
4591
4592         strncpy(info->driver, s2io_driver_name, sizeof(info->driver));
4593         strncpy(info->version, s2io_driver_version, sizeof(info->version));
4594         strncpy(info->fw_version, "", sizeof(info->fw_version));
4595         strncpy(info->bus_info, pci_name(sp->pdev), sizeof(info->bus_info));
4596         info->regdump_len = XENA_REG_SPACE;
4597         info->eedump_len = XENA_EEPROM_SPACE;
4598         info->testinfo_len = S2IO_TEST_LEN;
4599         info->n_stats = S2IO_STAT_LEN;
4600 }
4601
4602 /**
4603  *  s2io_ethtool_gregs - dumps the entire space of Xfame into the buffer.
4604  *  @sp: private member of the device structure, which is a pointer to the
4605  *  s2io_nic structure.
4606  *  @regs : pointer to the structure with parameters given by ethtool for
4607  *  dumping the registers.
4608  *  @reg_space: The input argumnet into which all the registers are dumped.
4609  *  Description:
4610  *  Dumps the entire register space of xFrame NIC into the user given
4611  *  buffer area.
4612  * Return value :
4613  * void .
4614 */
4615
4616 static void s2io_ethtool_gregs(struct net_device *dev,
4617                                struct ethtool_regs *regs, void *space)
4618 {
4619         int i;
4620         u64 reg;
4621         u8 *reg_space = (u8 *) space;
4622         nic_t *sp = dev->priv;
4623
4624         regs->len = XENA_REG_SPACE;
4625         regs->version = sp->pdev->subsystem_device;
4626
4627         for (i = 0; i < regs->len; i += 8) {
4628                 reg = readq(sp->bar0 + i);
4629                 memcpy((reg_space + i), &reg, 8);
4630         }
4631 }
4632
4633 /**
4634  *  s2io_phy_id  - timer function that alternates adapter LED.
4635  *  @data : address of the private member of the device structure, which
4636  *  is a pointer to the s2io_nic structure, provided as an u32.
4637  * Description: This is actually the timer function that alternates the
4638  * adapter LED bit of the adapter control bit to set/reset every time on
4639  * invocation. The timer is set for 1/2 a second, hence tha NIC blinks
4640  *  once every second.
4641 */
4642 static void s2io_phy_id(unsigned long data)
4643 {
4644         nic_t *sp = (nic_t *) data;
4645         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4646         u64 val64 = 0;
4647         u16 subid;
4648
4649         subid = sp->pdev->subsystem_device;
4650         if ((sp->device_type == XFRAME_II_DEVICE) ||
4651                    ((subid & 0xFF) >= 0x07)) {
4652                 val64 = readq(&bar0->gpio_control);
4653                 val64 ^= GPIO_CTRL_GPIO_0;
4654                 writeq(val64, &bar0->gpio_control);
4655         } else {
4656                 val64 = readq(&bar0->adapter_control);
4657                 val64 ^= ADAPTER_LED_ON;
4658                 writeq(val64, &bar0->adapter_control);
4659         }
4660
4661         mod_timer(&sp->id_timer, jiffies + HZ / 2);
4662 }
4663
4664 /**
4665  * s2io_ethtool_idnic - To physically identify the nic on the system.
4666  * @sp : private member of the device structure, which is a pointer to the
4667  * s2io_nic structure.
4668  * @id : pointer to the structure with identification parameters given by
4669  * ethtool.
4670  * Description: Used to physically identify the NIC on the system.
4671  * The Link LED will blink for a time specified by the user for
4672  * identification.
4673  * NOTE: The Link has to be Up to be able to blink the LED. Hence
4674  * identification is possible only if it's link is up.
4675  * Return value:
4676  * int , returns 0 on success
4677  */
4678
4679 static int s2io_ethtool_idnic(struct net_device *dev, u32 data)
4680 {
4681         u64 val64 = 0, last_gpio_ctrl_val;
4682         nic_t *sp = dev->priv;
4683         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4684         u16 subid;
4685
4686         subid = sp->pdev->subsystem_device;
4687         last_gpio_ctrl_val = readq(&bar0->gpio_control);
4688         if ((sp->device_type == XFRAME_I_DEVICE) &&
4689                 ((subid & 0xFF) < 0x07)) {
4690                 val64 = readq(&bar0->adapter_control);
4691                 if (!(val64 & ADAPTER_CNTL_EN)) {
4692                         printk(KERN_ERR
4693                                "Adapter Link down, cannot blink LED\n");
4694                         return -EFAULT;
4695                 }
4696         }
4697         if (sp->id_timer.function == NULL) {
4698                 init_timer(&sp->id_timer);
4699                 sp->id_timer.function = s2io_phy_id;
4700                 sp->id_timer.data = (unsigned long) sp;
4701         }
4702         mod_timer(&sp->id_timer, jiffies);
4703         if (data)
4704                 msleep_interruptible(data * HZ);
4705         else
4706                 msleep_interruptible(MAX_FLICKER_TIME);
4707         del_timer_sync(&sp->id_timer);
4708
4709         if (CARDS_WITH_FAULTY_LINK_INDICATORS(sp->device_type, subid)) {
4710                 writeq(last_gpio_ctrl_val, &bar0->gpio_control);
4711                 last_gpio_ctrl_val = readq(&bar0->gpio_control);
4712         }
4713
4714         return 0;
4715 }
4716
4717 /**
4718  * s2io_ethtool_getpause_data -Pause frame frame generation and reception.
4719  * @sp : private member of the device structure, which is a pointer to the
4720  *      s2io_nic structure.
4721  * @ep : pointer to the structure with pause parameters given by ethtool.
4722  * Description:
4723  * Returns the Pause frame generation and reception capability of the NIC.
4724  * Return value:
4725  *  void
4726  */
4727 static void s2io_ethtool_getpause_data(struct net_device *dev,
4728                                        struct ethtool_pauseparam *ep)
4729 {
4730         u64 val64;
4731         nic_t *sp = dev->priv;
4732         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4733
4734         val64 = readq(&bar0->rmac_pause_cfg);
4735         if (val64 & RMAC_PAUSE_GEN_ENABLE)
4736                 ep->tx_pause = TRUE;
4737         if (val64 & RMAC_PAUSE_RX_ENABLE)
4738                 ep->rx_pause = TRUE;
4739         ep->autoneg = FALSE;
4740 }
4741
4742 /**
4743  * s2io_ethtool_setpause_data -  set/reset pause frame generation.
4744  * @sp : private member of the device structure, which is a pointer to the
4745  *      s2io_nic structure.
4746  * @ep : pointer to the structure with pause parameters given by ethtool.
4747  * Description:
4748  * It can be used to set or reset Pause frame generation or reception
4749  * support of the NIC.
4750  * Return value:
4751  * int, returns 0 on Success
4752  */
4753
4754 static int s2io_ethtool_setpause_data(struct net_device *dev,
4755                                struct ethtool_pauseparam *ep)
4756 {
4757         u64 val64;
4758         nic_t *sp = dev->priv;
4759         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4760
4761         val64 = readq(&bar0->rmac_pause_cfg);
4762         if (ep->tx_pause)
4763                 val64 |= RMAC_PAUSE_GEN_ENABLE;
4764         else
4765                 val64 &= ~RMAC_PAUSE_GEN_ENABLE;
4766         if (ep->rx_pause)
4767                 val64 |= RMAC_PAUSE_RX_ENABLE;
4768         else
4769                 val64 &= ~RMAC_PAUSE_RX_ENABLE;
4770         writeq(val64, &bar0->rmac_pause_cfg);
4771         return 0;
4772 }
4773
4774 /**
4775  * read_eeprom - reads 4 bytes of data from user given offset.
4776  * @sp : private member of the device structure, which is a pointer to the
4777  *      s2io_nic structure.
4778  * @off : offset at which the data must be written
4779  * @data : Its an output parameter where the data read at the given
4780  *      offset is stored.
4781  * Description:
4782  * Will read 4 bytes of data from the user given offset and return the
4783  * read data.
4784  * NOTE: Will allow to read only part of the EEPROM visible through the
4785  *   I2C bus.
4786  * Return value:
4787  *  -1 on failure and 0 on success.
4788  */
4789
4790 #define S2IO_DEV_ID             5
4791 static int read_eeprom(nic_t * sp, int off, u64 * data)
4792 {
4793         int ret = -1;
4794         u32 exit_cnt = 0;
4795         u64 val64;
4796         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4797
4798         if (sp->device_type == XFRAME_I_DEVICE) {
4799                 val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) | I2C_CONTROL_ADDR(off) |
4800                     I2C_CONTROL_BYTE_CNT(0x3) | I2C_CONTROL_READ |
4801                     I2C_CONTROL_CNTL_START;
4802                 SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
4803
4804                 while (exit_cnt < 5) {
4805                         val64 = readq(&bar0->i2c_control);
4806                         if (I2C_CONTROL_CNTL_END(val64)) {
4807                                 *data = I2C_CONTROL_GET_DATA(val64);
4808                                 ret = 0;
4809                                 break;
4810                         }
4811                         msleep(50);
4812                         exit_cnt++;
4813                 }
4814         }
4815
4816         if (sp->device_type == XFRAME_II_DEVICE) {
4817                 val64 = SPI_CONTROL_KEY(0x9) | SPI_CONTROL_SEL1 |
4818                         SPI_CONTROL_BYTECNT(0x3) | 
4819                         SPI_CONTROL_CMD(0x3) | SPI_CONTROL_ADDR(off);
4820                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4821                 val64 |= SPI_CONTROL_REQ;
4822                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4823                 while (exit_cnt < 5) {
4824                         val64 = readq(&bar0->spi_control);
4825                         if (val64 & SPI_CONTROL_NACK) {
4826                                 ret = 1;
4827                                 break;
4828                         } else if (val64 & SPI_CONTROL_DONE) {
4829                                 *data = readq(&bar0->spi_data);
4830                                 *data &= 0xffffff;
4831                                 ret = 0;
4832                                 break;
4833                         }
4834                         msleep(50);
4835                         exit_cnt++;
4836                 }
4837         }
4838         return ret;
4839 }
4840
4841 /**
4842  *  write_eeprom - actually writes the relevant part of the data value.
4843  *  @sp : private member of the device structure, which is a pointer to the
4844  *       s2io_nic structure.
4845  *  @off : offset at which the data must be written
4846  *  @data : The data that is to be written
4847  *  @cnt : Number of bytes of the data that are actually to be written into
4848  *  the Eeprom. (max of 3)
4849  * Description:
4850  *  Actually writes the relevant part of the data value into the Eeprom
4851  *  through the I2C bus.
4852  * Return value:
4853  *  0 on success, -1 on failure.
4854  */
4855
4856 static int write_eeprom(nic_t * sp, int off, u64 data, int cnt)
4857 {
4858         int exit_cnt = 0, ret = -1;
4859         u64 val64;
4860         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4861
4862         if (sp->device_type == XFRAME_I_DEVICE) {
4863                 val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) | I2C_CONTROL_ADDR(off) |
4864                     I2C_CONTROL_BYTE_CNT(cnt) | I2C_CONTROL_SET_DATA((u32)data) |
4865                     I2C_CONTROL_CNTL_START;
4866                 SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
4867
4868                 while (exit_cnt < 5) {
4869                         val64 = readq(&bar0->i2c_control);
4870                         if (I2C_CONTROL_CNTL_END(val64)) {
4871                                 if (!(val64 & I2C_CONTROL_NACK))
4872                                         ret = 0;
4873                                 break;
4874                         }
4875                         msleep(50);
4876                         exit_cnt++;
4877                 }
4878         }
4879
4880         if (sp->device_type == XFRAME_II_DEVICE) {
4881                 int write_cnt = (cnt == 8) ? 0 : cnt;
4882                 writeq(SPI_DATA_WRITE(data,(cnt<<3)), &bar0->spi_data);
4883
4884                 val64 = SPI_CONTROL_KEY(0x9) | SPI_CONTROL_SEL1 |
4885                         SPI_CONTROL_BYTECNT(write_cnt) | 
4886                         SPI_CONTROL_CMD(0x2) | SPI_CONTROL_ADDR(off);
4887                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4888                 val64 |= SPI_CONTROL_REQ;
4889                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4890                 while (exit_cnt < 5) {
4891                         val64 = readq(&bar0->spi_control);
4892                         if (val64 & SPI_CONTROL_NACK) {
4893                                 ret = 1;
4894                                 break;
4895                         } else if (val64 & SPI_CONTROL_DONE) {
4896                                 ret = 0;
4897                                 break;
4898                         }
4899                         msleep(50);
4900                         exit_cnt++;
4901                 }
4902         }
4903         return ret;
4904 }
4905 static void s2io_vpd_read(nic_t *nic)
4906 {
4907         u8 *vpd_data;
4908         u8 data;
4909         int i=0, cnt, fail = 0;
4910         int vpd_addr = 0x80;
4911
4912         if (nic->device_type == XFRAME_II_DEVICE) {
4913                 strcpy(nic->product_name, "Xframe II 10GbE network adapter");
4914                 vpd_addr = 0x80;
4915         }
4916         else {
4917                 strcpy(nic->product_name, "Xframe I 10GbE network adapter");
4918                 vpd_addr = 0x50;
4919         }
4920
4921         vpd_data = kmalloc(256, GFP_KERNEL);
4922         if (!vpd_data)
4923                 return;
4924
4925         for (i = 0; i < 256; i +=4 ) {
4926                 pci_write_config_byte(nic->pdev, (vpd_addr + 2), i);
4927                 pci_read_config_byte(nic->pdev,  (vpd_addr + 2), &data);
4928                 pci_write_config_byte(nic->pdev, (vpd_addr + 3), 0);
4929                 for (cnt = 0; cnt <5; cnt++) {
4930                         msleep(2);
4931                         pci_read_config_byte(nic->pdev, (vpd_addr + 3), &data);
4932                         if (data == 0x80)
4933                                 break;
4934                 }
4935                 if (cnt >= 5) {
4936                         DBG_PRINT(ERR_DBG, "Read of VPD data failed\n");
4937                         fail = 1;
4938                         break;
4939                 }
4940                 pci_read_config_dword(nic->pdev,  (vpd_addr + 4),
4941                                       (u32 *)&vpd_data[i]);
4942         }
4943         if ((!fail) && (vpd_data[1] < VPD_PRODUCT_NAME_LEN)) {
4944                 memset(nic->product_name, 0, vpd_data[1]);
4945                 memcpy(nic->product_name, &vpd_data[3], vpd_data[1]);
4946         }
4947         kfree(vpd_data);
4948 }
4949
4950 /**
4951  *  s2io_ethtool_geeprom  - reads the value stored in the Eeprom.
4952  *  @sp : private member of the device structure, which is a pointer to the *       s2io_nic structure.
4953  *  @eeprom : pointer to the user level structure provided by ethtool,
4954  *  containing all relevant information.
4955  *  @data_buf : user defined value to be written into Eeprom.
4956  *  Description: Reads the values stored in the Eeprom at given offset
4957  *  for a given length. Stores these values int the input argument data
4958  *  buffer 'data_buf' and returns these to the caller (ethtool.)
4959  *  Return value:
4960  *  int  0 on success
4961  */
4962
4963 static int s2io_ethtool_geeprom(struct net_device *dev,
4964                          struct ethtool_eeprom *eeprom, u8 * data_buf)
4965 {
4966         u32 i, valid;
4967         u64 data;
4968         nic_t *sp = dev->priv;
4969
4970         eeprom->magic = sp->pdev->vendor | (sp->pdev->device << 16);
4971
4972         if ((eeprom->offset + eeprom->len) > (XENA_EEPROM_SPACE))
4973                 eeprom->len = XENA_EEPROM_SPACE - eeprom->offset;
4974
4975         for (i = 0; i < eeprom->len; i += 4) {
4976                 if (read_eeprom(sp, (eeprom->offset + i), &data)) {
4977                         DBG_PRINT(ERR_DBG, "Read of EEPROM failed\n");
4978                         return -EFAULT;
4979                 }
4980                 valid = INV(data);
4981                 memcpy((data_buf + i), &valid, 4);
4982         }
4983         return 0;
4984 }
4985
4986 /**
4987  *  s2io_ethtool_seeprom - tries to write the user provided value in Eeprom
4988  *  @sp : private member of the device structure, which is a pointer to the
4989  *  s2io_nic structure.
4990  *  @eeprom : pointer to the user level structure provided by ethtool,
4991  *  containing all relevant information.
4992  *  @data_buf ; user defined value to be written into Eeprom.
4993  *  Description:
4994  *  Tries to write the user provided value in the Eeprom, at the offset
4995  *  given by the user.
4996  *  Return value:
4997  *  0 on success, -EFAULT on failure.
4998  */
4999
5000 static int s2io_ethtool_seeprom(struct net_device *dev,
5001                                 struct ethtool_eeprom *eeprom,
5002                                 u8 * data_buf)
5003 {
5004         int len = eeprom->len, cnt = 0;
5005         u64 valid = 0, data;
5006         nic_t *sp = dev->priv;
5007
5008         if (eeprom->magic != (sp->pdev->vendor | (sp->pdev->device << 16))) {
5009                 DBG_PRINT(ERR_DBG,
5010                           "ETHTOOL_WRITE_EEPROM Err: Magic value ");
5011                 DBG_PRINT(ERR_DBG, "is wrong, Its not 0x%x\n",
5012                           eeprom->magic);
5013                 return -EFAULT;
5014         }
5015
5016         while (len) {
5017                 data = (u32) data_buf[cnt] & 0x000000FF;
5018                 if (data) {
5019                         valid = (u32) (data << 24);
5020                 } else
5021                         valid = data;
5022
5023                 if (write_eeprom(sp, (eeprom->offset + cnt), valid, 0)) {
5024                         DBG_PRINT(ERR_DBG,
5025                                   "ETHTOOL_WRITE_EEPROM Err: Cannot ");
5026                         DBG_PRINT(ERR_DBG,
5027                                   "write into the specified offset\n");
5028                         return -EFAULT;
5029                 }
5030                 cnt++;
5031                 len--;
5032         }
5033
5034         return 0;
5035 }
5036
5037 /**
5038  * s2io_register_test - reads and writes into all clock domains.
5039  * @sp : private member of the device structure, which is a pointer to the
5040  * s2io_nic structure.
5041  * @data : variable that returns the result of each of the test conducted b
5042  * by the driver.
5043  * Description:
5044  * Read and write into all clock domains. The NIC has 3 clock domains,
5045  * see that registers in all the three regions are accessible.
5046  * Return value:
5047  * 0 on success.
5048  */
5049
5050 static int s2io_register_test(nic_t * sp, uint64_t * data)
5051 {
5052         XENA_dev_config_t __iomem *bar0 = sp->bar0;
5053         u64 val64 = 0, exp_val;
5054         int fail = 0;
5055
5056         val64 = readq(&bar0->pif_rd_swapper_fb);
5057         if (val64 != 0x123456789abcdefULL) {
5058                 fail = 1;
5059                 DBG_PRINT(INFO_DBG, "Read Test level 1 fails\n");
5060         }
5061
5062         val64 = readq(&bar0->rmac_pause_cfg);
5063         if (val64 != 0xc000ffff00000000ULL) {
5064                 fail = 1;
5065                 DBG_PRINT(INFO_DBG, "Read Test level 2 fails\n");
5066         }
5067
5068         val64 = readq(&bar0->rx_queue_cfg);
5069         if (sp->device_type == XFRAME_II_DEVICE)
5070                 exp_val = 0x0404040404040404ULL;
5071         else
5072                 exp_val = 0x0808080808080808ULL;
5073         if (val64 != exp_val) {
5074                 fail = 1;
5075                 DBG_PRINT(INFO_DBG, "Read Test level 3 fails\n");
5076         }
5077
5078         val64 = readq(&bar0->xgxs_efifo_cfg);
5079         if (val64 != 0x000000001923141EULL) {
5080                 fail = 1;
5081                 DBG_PRINT(INFO_DBG, "Read Test level 4 fails\n");
5082         }
5083
5084         val64 = 0x5A5A5A5A5A5A5A5AULL;
5085         writeq(val64, &bar0->xmsi_data);
5086         val64 = readq(&bar0->xmsi_data);
5087         if (val64 != 0x5A5A5A5A5A5A5A5AULL) {
5088                 fail = 1;
5089                 DBG_PRINT(ERR_DBG, "Write Test level 1 fails\n");
5090         }
5091
5092         val64 = 0xA5A5A5A5A5A5A5A5ULL;
5093         writeq(val64, &bar0->xmsi_data);
5094         val64 = readq(&bar0->xmsi_data);
5095         if (val64 != 0xA5A5A5A5A5A5A5A5ULL) {
5096                 fail = 1;
5097                 DBG_PRINT(ERR_DBG, "Write Test level 2 fails\n");
5098         }
5099
5100         *data = fail;
5101         return fail;
5102 }
5103
5104 /**
5105  * s2io_eeprom_test - to verify that EEprom in the xena can be programmed.
5106  * @sp : private member of the device structure, which is a pointer to the
5107  * s2io_nic structure.
5108  * @data:variable that returns the result of each of the test conducted by
5109  * the driver.
5110  * Description:
5111  * Verify that EEPROM in the xena can be programmed using I2C_CONTROL
5112  * register.
5113  * Return value:
5114  * 0 on success.
5115  */
5116
5117 static int s2io_eeprom_test(nic_t * sp, uint64_t * data)
5118 {
5119         int fail = 0;
5120         u64 ret_data, org_4F0, org_7F0;
5121         u8 saved_4F0 = 0, saved_7F0 = 0;
5122         struct net_device *dev = sp->dev;
5123
5124         /* Test Write Error at offset 0 */
5125         /* Note that SPI interface allows write access to all areas
5126          * of EEPROM. Hence doing all negative testing only for Xframe I.
5127          */
5128         if (sp->device_type == XFRAME_I_DEVICE)
5129                 if (!write_eeprom(sp, 0, 0, 3))
5130                         fail = 1;
5131
5132         /* Save current values at offsets 0x4F0 and 0x7F0 */
5133         if (!read_eeprom(sp, 0x4F0, &org_4F0))
5134                 saved_4F0 = 1;
5135         if (!read_eeprom(sp, 0x7F0, &org_7F0))
5136                 saved_7F0 = 1;
5137
5138         /* Test Write at offset 4f0 */
5139         if (write_eeprom(sp, 0x4F0, 0x012345, 3))
5140                 fail = 1;
5141         if (read_eeprom(sp, 0x4F0, &ret_data))
5142                 fail = 1;
5143
5144         if (ret_data != 0x012345) {
5145                 DBG_PRINT(ERR_DBG, "%s: eeprom test error at offset 0x4F0. "
5146                         "Data written %llx Data read %llx\n",
5147                         dev->name, (unsigned long long)0x12345,
5148                         (unsigned long long)ret_data);
5149                 fail = 1;
5150         }
5151
5152         /* Reset the EEPROM data go FFFF */
5153         write_eeprom(sp, 0x4F0, 0xFFFFFF, 3);
5154
5155         /* Test Write Request Error at offset 0x7c */
5156         if (sp->device_type == XFRAME_I_DEVICE)
5157                 if (!write_eeprom(sp, 0x07C, 0, 3))
5158                         fail = 1;
5159
5160         /* Test Write Request at offset 0x7f0 */
5161         if (write_eeprom(sp, 0x7F0, 0x012345, 3))
5162                 fail = 1;
5163         if (read_eeprom(sp, 0x7F0, &ret_data))
5164                 fail = 1;
5165
5166         if (ret_data != 0x012345) {
5167                 DBG_PRINT(ERR_DBG, "%s: eeprom test error at offset 0x7F0. "
5168                         "Data written %llx Data read %llx\n",
5169                         dev->name, (unsigned long long)0x12345,
5170                         (unsigned long long)ret_data);
5171                 fail = 1;
5172         }
5173
5174         /* Reset the EEPROM data go FFFF */
5175         write_eeprom(sp, 0x7F0, 0xFFFFFF, 3);
5176
5177         if (sp->device_type == XFRAME_I_DEVICE) {
5178                 /* Test Write Error at offset 0x80 */
5179                 if (!write_eeprom(sp, 0x080, 0, 3))
5180                         fail = 1;
5181
5182                 /* Test Write Error at offset 0xfc */
5183                 if (!write_eeprom(sp, 0x0FC, 0, 3))
5184                         fail = 1;
5185
5186                 /* Test Write Error at offset 0x100 */
5187                 if (!write_eeprom(sp, 0x100, 0, 3))
5188                         fail = 1;
5189
5190                 /* Test Write Error at offset 4ec */
5191                 if (!write_eeprom(sp, 0x4EC, 0, 3))
5192                         fail = 1;
5193         }
5194
5195         /* Restore values at offsets 0x4F0 and 0x7F0 */
5196         if (saved_4F0)
5197                 write_eeprom(sp, 0x4F0, org_4F0, 3);
5198         if (saved_7F0)
5199                 write_eeprom(sp, 0x7F0, org_7F0, 3);
5200
5201         *data = fail;
5202         return fail;
5203 }
5204
5205 /**
5206  * s2io_bist_test - invokes the MemBist test of the card .
5207  * @sp : private member of the device structure, which is a pointer to the
5208  * s2io_nic structure.
5209  * @data:variable that returns the result of each of the test conducted by
5210  * the driver.
5211  * Description:
5212  * This invokes the MemBist test of the card. We give around
5213  * 2 secs time for the Test to complete. If it's still not complete
5214  * within this peiod, we consider that the test failed.
5215  * Return value:
5216  * 0 on success and -1 on failure.
5217  */
5218
5219 static int s2io_bist_test(nic_t * sp, uint64_t * data)
5220 {
5221         u8 bist = 0;
5222         int cnt = 0, ret = -1;
5223
5224         pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
5225         bist |= PCI_BIST_START;
5226         pci_write_config_word(sp->pdev, PCI_BIST, bist);
5227
5228         while (cnt < 20) {
5229                 pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
5230                 if (!(bist & PCI_BIST_START)) {
5231                         *data = (bist & PCI_BIST_CODE_MASK);
5232                         ret = 0;
5233                         break;
5234                 }
5235                 msleep(100);
5236                 cnt++;
5237         }
5238
5239         return ret;
5240 }
5241
5242 /**
5243  * s2io-link_test - verifies the link state of the nic
5244  * @sp ; private member of the device structure, which is a pointer to the
5245  * s2io_nic structure.
5246  * @data: variable that returns the result of each of the test conducted by
5247  * the driver.
5248  * Description:
5249  * The function verifies the link state of the NIC and updates the input
5250  * argument 'data' appropriately.
5251  * Return value:
5252  * 0 on success.
5253  */
5254
5255 static int s2io_link_test(nic_t * sp, uint64_t * data)
5256 {
5257         XENA_dev_config_t __iomem *bar0 = sp->bar0;
5258         u64 val64;
5259
5260         val64 = readq(&bar0->adapter_status);
5261         if(!(LINK_IS_UP(val64)))
5262                 *data = 1;
5263         else
5264                 *data = 0;
5265
5266         return *data;
5267 }
5268
5269 /**
5270  * s2io_rldram_test - offline test for access to the RldRam chip on the NIC
5271  * @sp - private member of the device structure, which is a pointer to the
5272  * s2io_nic structure.
5273  * @data - variable that returns the result of each of the test
5274  * conducted by the driver.
5275  * Description:
5276  *  This is one of the offline test that tests the read and write
5277  *  access to the RldRam chip on the NIC.
5278  * Return value:
5279  *  0 on success.
5280  */
5281
5282 static int s2io_rldram_test(nic_t * sp, uint64_t * data)
5283 {
5284         XENA_dev_config_t __iomem *bar0 = sp->bar0;
5285         u64 val64;
5286         int cnt, iteration = 0, test_fail = 0;
5287
5288         val64 = readq(&bar0->adapter_control);
5289         val64 &= ~ADAPTER_ECC_EN;
5290         writeq(val64, &bar0->adapter_control);
5291
5292         val64 = readq(&bar0->mc_rldram_test_ctrl);
5293         val64 |= MC_RLDRAM_TEST_MODE;
5294         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
5295
5296         val64 = readq(&bar0->mc_rldram_mrs);
5297         val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE;
5298         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
5299
5300         val64 |= MC_RLDRAM_MRS_ENABLE;
5301         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
5302
5303         while (iteration < 2) {
5304                 val64 = 0x55555555aaaa0000ULL;
5305                 if (iteration == 1) {
5306                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
5307                 }
5308                 writeq(val64, &bar0->mc_rldram_test_d0);
5309
5310                 val64 = 0xaaaa5a5555550000ULL;
5311                 if (iteration == 1) {
5312                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
5313                 }
5314                 writeq(val64, &bar0->mc_rldram_test_d1);
5315
5316                 val64 = 0x55aaaaaaaa5a0000ULL;
5317                 if (iteration == 1) {
5318                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
5319                 }
5320                 writeq(val64, &bar0->mc_rldram_test_d2);
5321
5322                 val64 = (u64) (0x0000003ffffe0100ULL);
5323                 writeq(val64, &bar0->mc_rldram_test_add);
5324
5325                 val64 = MC_RLDRAM_TEST_MODE | MC_RLDRAM_TEST_WRITE |
5326                         MC_RLDRAM_TEST_GO;
5327                 SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
5328
5329                 for (cnt = 0; cnt < 5; cnt++) {
5330                         val64 = readq(&bar0->mc_rldram_test_ctrl);
5331                         if (val64 & MC_RLDRAM_TEST_DONE)
5332                                 break;
5333                         msleep(200);
5334                 }
5335
5336                 if (cnt == 5)
5337                         break;
5338
5339                 val64 = MC_RLDRAM_TEST_MODE | MC_RLDRAM_TEST_GO;
5340                 SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
5341
5342                 for (cnt = 0; cnt < 5; cnt++) {
5343                         val64 = readq(&bar0->mc_rldram_test_ctrl);
5344                         if (val64 & MC_RLDRAM_TEST_DONE)
5345                                 break;
5346                         msleep(500);
5347                 }
5348
5349                 if (cnt == 5)
5350                         break;
5351
5352                 val64 = readq(&bar0->mc_rldram_test_ctrl);
5353                 if (!(val64 & MC_RLDRAM_TEST_PASS))
5354                         test_fail = 1;
5355
5356                 iteration++;
5357         }
5358
5359         *data = test_fail;
5360
5361         /* Bring the adapter out of test mode */
5362         SPECIAL_REG_WRITE(0, &bar0->mc_rldram_test_ctrl, LF);
5363
5364         return test_fail;
5365 }
5366
5367 /**
5368  *  s2io_ethtool_test - conducts 6 tsets to determine the health of card.
5369  *  @sp : private member of the device structure, which is a pointer to the
5370  *  s2io_nic structure.
5371  *  @ethtest : pointer to a ethtool command specific structure that will be
5372  *  returned to the user.
5373  *  @data : variable that returns the result of each of the test
5374  * conducted by the driver.
5375  * Description:
5376  *  This function conducts 6 tests ( 4 offline and 2 online) to determine
5377  *  the health of the card.
5378  * Return value:
5379  *  void
5380  */
5381
5382 static void s2io_ethtool_test(struct net_device *dev,
5383                               struct ethtool_test *ethtest,
5384                               uint64_t * data)
5385 {
5386         nic_t *sp = dev->priv;
5387         int orig_state = netif_running(sp->dev);
5388
5389         if (ethtest->flags == ETH_TEST_FL_OFFLINE) {
5390                 /* Offline Tests. */
5391                 if (orig_state)
5392                         s2io_close(sp->dev);
5393
5394                 if (s2io_register_test(sp, &data[0]))
5395                         ethtest->flags |= ETH_TEST_FL_FAILED;
5396
5397                 s2io_reset(sp);
5398
5399                 if (s2io_rldram_test(sp, &data[3]))
5400                         ethtest->flags |= ETH_TEST_FL_FAILED;
5401
5402                 s2io_reset(sp);
5403
5404                 if (s2io_eeprom_test(sp, &data[1]))
5405                         ethtest->flags |= ETH_TEST_FL_FAILED;
5406
5407                 if (s2io_bist_test(sp, &data[4]))
5408                         ethtest->flags |= ETH_TEST_FL_FAILED;
5409
5410                 if (orig_state)
5411                         s2io_open(sp->dev);
5412
5413                 data[2] = 0;
5414         } else {
5415                 /* Online Tests. */
5416                 if (!orig_state) {
5417                         DBG_PRINT(ERR_DBG,
5418                                   "%s: is not up, cannot run test\n",
5419                                   dev->name);
5420                         data[0] = -1;
5421                         data[1] = -1;
5422                         data[2] = -1;
5423                         data[3] = -1;
5424                         data[4] = -1;
5425                 }
5426
5427                 if (s2io_link_test(sp, &data[2]))
5428                         ethtest->flags |= ETH_TEST_FL_FAILED;
5429
5430                 data[0] = 0;
5431                 data[1] = 0;
5432                 data[3] = 0;
5433                 data[4] = 0;
5434         }
5435 }
5436
5437 static void s2io_get_ethtool_stats(struct net_device *dev,
5438                                    struct ethtool_stats *estats,
5439                                    u64 * tmp_stats)
5440 {
5441         int i = 0;
5442         nic_t *sp = dev->priv;
5443         StatInfo_t *stat_info = sp->mac_control.stats_info;
5444
5445         s2io_updt_stats(sp);
5446         tmp_stats[i++] =
5447                 (u64)le32_to_cpu(stat_info->tmac_frms_oflow) << 32  |
5448                 le32_to_cpu(stat_info->tmac_frms);
5449         tmp_stats[i++] =
5450                 (u64)le32_to_cpu(stat_info->tmac_data_octets_oflow) << 32 |
5451                 le32_to_cpu(stat_info->tmac_data_octets);
5452         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_drop_frms);
5453         tmp_stats[i++] =
5454                 (u64)le32_to_cpu(stat_info->tmac_mcst_frms_oflow) << 32 |
5455                 le32_to_cpu(stat_info->tmac_mcst_frms);
5456         tmp_stats[i++] =
5457                 (u64)le32_to_cpu(stat_info->tmac_bcst_frms_oflow) << 32 |
5458                 le32_to_cpu(stat_info->tmac_bcst_frms);
5459         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_pause_ctrl_frms);
5460         tmp_stats[i++] =
5461                 (u64)le32_to_cpu(stat_info->tmac_ttl_octets_oflow) << 32 |
5462                 le32_to_cpu(stat_info->tmac_ttl_octets);
5463         tmp_stats[i++] =
5464                 (u64)le32_to_cpu(stat_info->tmac_ucst_frms_oflow) << 32 |
5465                 le32_to_cpu(stat_info->tmac_ucst_frms);
5466         tmp_stats[i++] =
5467                 (u64)le32_to_cpu(stat_info->tmac_nucst_frms_oflow) << 32 |
5468                 le32_to_cpu(stat_info->tmac_nucst_frms);
5469         tmp_stats[i++] =
5470                 (u64)le32_to_cpu(stat_info->tmac_any_err_frms_oflow) << 32 |
5471                 le32_to_cpu(stat_info->tmac_any_err_frms);
5472         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_ttl_less_fb_octets);
5473         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_vld_ip_octets);
5474         tmp_stats[i++] =
5475                 (u64)le32_to_cpu(stat_info->tmac_vld_ip_oflow) << 32 |
5476                 le32_to_cpu(stat_info->tmac_vld_ip);
5477         tmp_stats[i++] =
5478                 (u64)le32_to_cpu(stat_info->tmac_drop_ip_oflow) << 32 |
5479                 le32_to_cpu(stat_info->tmac_drop_ip);
5480         tmp_stats[i++] =
5481                 (u64)le32_to_cpu(stat_info->tmac_icmp_oflow) << 32 |
5482                 le32_to_cpu(stat_info->tmac_icmp);
5483         tmp_stats[i++] =
5484                 (u64)le32_to_cpu(stat_info->tmac_rst_tcp_oflow) << 32 |
5485                 le32_to_cpu(stat_info->tmac_rst_tcp);
5486         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_tcp);
5487         tmp_stats[i++] = (u64)le32_to_cpu(stat_info->tmac_udp_oflow) << 32 |
5488                 le32_to_cpu(stat_info->tmac_udp);
5489         tmp_stats[i++] =
5490                 (u64)le32_to_cpu(stat_info->rmac_vld_frms_oflow) << 32 |
5491                 le32_to_cpu(stat_info->rmac_vld_frms);
5492         tmp_stats[i++] =
5493                 (u64)le32_to_cpu(stat_info->rmac_data_octets_oflow) << 32 |
5494                 le32_to_cpu(stat_info->rmac_data_octets);
5495         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_fcs_err_frms);
5496         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_drop_frms);
5497         tmp_stats[i++] =
5498                 (u64)le32_to_cpu(stat_info->rmac_vld_mcst_frms_oflow) << 32 |
5499                 le32_to_cpu(stat_info->rmac_vld_mcst_frms);
5500         tmp_stats[i++] =
5501                 (u64)le32_to_cpu(stat_info->rmac_vld_bcst_frms_oflow) << 32 |
5502                 le32_to_cpu(stat_info->rmac_vld_bcst_frms);
5503         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_in_rng_len_err_frms);
5504         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_out_rng_len_err_frms);
5505         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_long_frms);
5506         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_pause_ctrl_frms);
5507         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_unsup_ctrl_frms);
5508         tmp_stats[i++] =
5509                 (u64)le32_to_cpu(stat_info->rmac_ttl_octets_oflow) << 32 |
5510                 le32_to_cpu(stat_info->rmac_ttl_octets);
5511         tmp_stats[i++] =
5512                 (u64)le32_to_cpu(stat_info->rmac_accepted_ucst_frms_oflow)
5513                 << 32 | le32_to_cpu(stat_info->rmac_accepted_ucst_frms);
5514         tmp_stats[i++] =
5515                 (u64)le32_to_cpu(stat_info->rmac_accepted_nucst_frms_oflow)
5516                  << 32 | le32_to_cpu(stat_info->rmac_accepted_nucst_frms);
5517         tmp_stats[i++] =
5518                 (u64)le32_to_cpu(stat_info->rmac_discarded_frms_oflow) << 32 |
5519                 le32_to_cpu(stat_info->rmac_discarded_frms);
5520         tmp_stats[i++] =
5521                 (u64)le32_to_cpu(stat_info->rmac_drop_events_oflow)
5522                  << 32 | le32_to_cpu(stat_info->rmac_drop_events);
5523         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_less_fb_octets);
5524         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_frms);
5525         tmp_stats[i++] =
5526                 (u64)le32_to_cpu(stat_info->rmac_usized_frms_oflow) << 32 |
5527                 le32_to_cpu(stat_info->rmac_usized_frms);
5528         tmp_stats[i++] =
5529                 (u64)le32_to_cpu(stat_info->rmac_osized_frms_oflow) << 32 |
5530                 le32_to_cpu(stat_info->rmac_osized_frms);
5531         tmp_stats[i++] =
5532                 (u64)le32_to_cpu(stat_info->rmac_frag_frms_oflow) << 32 |
5533                 le32_to_cpu(stat_info->rmac_frag_frms);
5534         tmp_stats[i++] =
5535                 (u64)le32_to_cpu(stat_info->rmac_jabber_frms_oflow) << 32 |
5536                 le32_to_cpu(stat_info->rmac_jabber_frms);
5537         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_64_frms);
5538         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_65_127_frms);
5539         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_128_255_frms);
5540         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_256_511_frms);
5541         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_512_1023_frms);
5542         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_1024_1518_frms);
5543         tmp_stats[i++] =
5544                 (u64)le32_to_cpu(stat_info->rmac_ip_oflow) << 32 |
5545                 le32_to_cpu(stat_info->rmac_ip);
5546         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ip_octets);
5547         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_hdr_err_ip);
5548         tmp_stats[i++] =
5549                 (u64)le32_to_cpu(stat_info->rmac_drop_ip_oflow) << 32 |
5550                 le32_to_cpu(stat_info->rmac_drop_ip);
5551         tmp_stats[i++] =
5552                 (u64)le32_to_cpu(stat_info->rmac_icmp_oflow) << 32 |
5553                 le32_to_cpu(stat_info->rmac_icmp);
5554         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_tcp);
5555         tmp_stats[i++] =
5556                 (u64)le32_to_cpu(stat_info->rmac_udp_oflow) << 32 |
5557                 le32_to_cpu(stat_info->rmac_udp);
5558         tmp_stats[i++] =
5559                 (u64)le32_to_cpu(stat_info->rmac_err_drp_udp_oflow) << 32 |
5560                 le32_to_cpu(stat_info->rmac_err_drp_udp);
5561         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_xgmii_err_sym);
5562         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q0);
5563         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q1);
5564         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q2);
5565         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q3);
5566         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q4);
5567         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q5);
5568         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q6);
5569         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q7);
5570         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q0);
5571         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q1);
5572         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q2);
5573         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q3);
5574         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q4);
5575         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q5);
5576         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q6);
5577         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q7);
5578         tmp_stats[i++] =
5579                 (u64)le32_to_cpu(stat_info->rmac_pause_cnt_oflow) << 32 |
5580                 le32_to_cpu(stat_info->rmac_pause_cnt);
5581         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_xgmii_data_err_cnt);
5582         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_xgmii_ctrl_err_cnt);
5583         tmp_stats[i++] =
5584                 (u64)le32_to_cpu(stat_info->rmac_accepted_ip_oflow) << 32 |
5585                 le32_to_cpu(stat_info->rmac_accepted_ip);
5586         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_err_tcp);
5587         tmp_stats[i++] = le32_to_cpu(stat_info->rd_req_cnt);
5588         tmp_stats[i++] = le32_to_cpu(stat_info->new_rd_req_cnt);
5589         tmp_stats[i++] = le32_to_cpu(stat_info->new_rd_req_rtry_cnt);
5590         tmp_stats[i++] = le32_to_cpu(stat_info->rd_rtry_cnt);
5591         tmp_stats[i++] = le32_to_cpu(stat_info->wr_rtry_rd_ack_cnt);
5592         tmp_stats[i++] = le32_to_cpu(stat_info->wr_req_cnt);
5593         tmp_stats[i++] = le32_to_cpu(stat_info->new_wr_req_cnt);
5594         tmp_stats[i++] = le32_to_cpu(stat_info->new_wr_req_rtry_cnt);
5595         tmp_stats[i++] = le32_to_cpu(stat_info->wr_rtry_cnt);
5596         tmp_stats[i++] = le32_to_cpu(stat_info->wr_disc_cnt);
5597         tmp_stats[i++] = le32_to_cpu(stat_info->rd_rtry_wr_ack_cnt);
5598         tmp_stats[i++] = le32_to_cpu(stat_info->txp_wr_cnt);
5599         tmp_stats[i++] = le32_to_cpu(stat_info->txd_rd_cnt);
5600         tmp_stats[i++] = le32_to_cpu(stat_info->txd_wr_cnt);
5601         tmp_stats[i++] = le32_to_cpu(stat_info->rxd_rd_cnt);
5602         tmp_stats[i++] = le32_to_cpu(stat_info->rxd_wr_cnt);
5603         tmp_stats[i++] = le32_to_cpu(stat_info->txf_rd_cnt);
5604         tmp_stats[i++] = le32_to_cpu(stat_info->rxf_wr_cnt);
5605         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_1519_4095_frms);
5606         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_4096_8191_frms);
5607         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_8192_max_frms);
5608         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_gt_max_frms);
5609         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_osized_alt_frms);
5610         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_jabber_alt_frms);
5611         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_gt_max_alt_frms);
5612         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_vlan_frms);
5613         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_len_discard);
5614         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_fcs_discard);
5615         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_pf_discard);
5616         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_da_discard);
5617         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_red_discard);
5618         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_rts_discard);
5619         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_ingm_full_discard);
5620         tmp_stats[i++] = le32_to_cpu(stat_info->link_fault_cnt);
5621         tmp_stats[i++] = 0;
5622         tmp_stats[i++] = stat_info->sw_stat.single_ecc_errs;
5623         tmp_stats[i++] = stat_info->sw_stat.double_ecc_errs;
5624         tmp_stats[i++] = stat_info->sw_stat.parity_err_cnt;
5625         tmp_stats[i++] = stat_info->sw_stat.serious_err_cnt;
5626         tmp_stats[i++] = stat_info->sw_stat.soft_reset_cnt;
5627         tmp_stats[i++] = stat_info->sw_stat.fifo_full_cnt;
5628         tmp_stats[i++] = stat_info->sw_stat.ring_full_cnt;
5629         tmp_stats[i++] = stat_info->xpak_stat.alarm_transceiver_temp_high;
5630         tmp_stats[i++] = stat_info->xpak_stat.alarm_transceiver_temp_low;
5631         tmp_stats[i++] = stat_info->xpak_stat.alarm_laser_bias_current_high;
5632         tmp_stats[i++] = stat_info->xpak_stat.alarm_laser_bias_current_low;
5633         tmp_stats[i++] = stat_info->xpak_stat.alarm_laser_output_power_high;
5634         tmp_stats[i++] = stat_info->xpak_stat.alarm_laser_output_power_low;
5635         tmp_stats[i++] = stat_info->xpak_stat.warn_transceiver_temp_high;
5636         tmp_stats[i++] = stat_info->xpak_stat.warn_transceiver_temp_low;
5637         tmp_stats[i++] = stat_info->xpak_stat.warn_laser_bias_current_high;
5638         tmp_stats[i++] = stat_info->xpak_stat.warn_laser_bias_current_low;
5639         tmp_stats[i++] = stat_info->xpak_stat.warn_laser_output_power_high;
5640         tmp_stats[i++] = stat_info->xpak_stat.warn_laser_output_power_low;
5641         tmp_stats[i++] = stat_info->sw_stat.clubbed_frms_cnt;
5642         tmp_stats[i++] = stat_info->sw_stat.sending_both;
5643         tmp_stats[i++] = stat_info->sw_stat.outof_sequence_pkts;
5644         tmp_stats[i++] = stat_info->sw_stat.flush_max_pkts;
5645         if (stat_info->sw_stat.num_aggregations) {
5646                 u64 tmp = stat_info->sw_stat.sum_avg_pkts_aggregated;
5647                 int count = 0;
5648                 /* 
5649                  * Since 64-bit divide does not work on all platforms,
5650                  * do repeated subtraction.
5651                  */
5652                 while (tmp >= stat_info->sw_stat.num_aggregations) {
5653                         tmp -= stat_info->sw_stat.num_aggregations;
5654                         count++;
5655                 }
5656                 tmp_stats[i++] = count;
5657         }
5658         else
5659                 tmp_stats[i++] = 0;
5660 }
5661
5662 static int s2io_ethtool_get_regs_len(struct net_device *dev)
5663 {
5664         return (XENA_REG_SPACE);
5665 }
5666
5667
5668 static u32 s2io_ethtool_get_rx_csum(struct net_device * dev)
5669 {
5670         nic_t *sp = dev->priv;
5671
5672         return (sp->rx_csum);
5673 }
5674
5675 static int s2io_ethtool_set_rx_csum(struct net_device *dev, u32 data)
5676 {
5677         nic_t *sp = dev->priv;
5678
5679         if (data)
5680                 sp->rx_csum = 1;
5681         else
5682                 sp->rx_csum = 0;
5683
5684         return 0;
5685 }
5686
5687 static int s2io_get_eeprom_len(struct net_device *dev)
5688 {
5689         return (XENA_EEPROM_SPACE);
5690 }
5691
5692 static int s2io_ethtool_self_test_count(struct net_device *dev)
5693 {
5694         return (S2IO_TEST_LEN);
5695 }
5696
5697 static void s2io_ethtool_get_strings(struct net_device *dev,
5698                                      u32 stringset, u8 * data)
5699 {
5700         switch (stringset) {
5701         case ETH_SS_TEST:
5702                 memcpy(data, s2io_gstrings, S2IO_STRINGS_LEN);
5703                 break;
5704         case ETH_SS_STATS:
5705                 memcpy(data, &ethtool_stats_keys,
5706                        sizeof(ethtool_stats_keys));
5707         }
5708 }
5709 static int s2io_ethtool_get_stats_count(struct net_device *dev)
5710 {
5711         return (S2IO_STAT_LEN);
5712 }
5713
5714 static int s2io_ethtool_op_set_tx_csum(struct net_device *dev, u32 data)
5715 {
5716         if (data)
5717                 dev->features |= NETIF_F_IP_CSUM;
5718         else
5719                 dev->features &= ~NETIF_F_IP_CSUM;
5720
5721         return 0;
5722 }
5723
5724 static u32 s2io_ethtool_op_get_tso(struct net_device *dev)
5725 {
5726         return (dev->features & NETIF_F_TSO) != 0;
5727 }
5728 static int s2io_ethtool_op_set_tso(struct net_device *dev, u32 data)
5729 {
5730         if (data)
5731                 dev->features |= (NETIF_F_TSO | NETIF_F_TSO6);
5732         else
5733                 dev->features &= ~(NETIF_F_TSO | NETIF_F_TSO6);
5734
5735         return 0;
5736 }
5737
5738 static struct ethtool_ops netdev_ethtool_ops = {
5739         .get_settings = s2io_ethtool_gset,
5740         .set_settings = s2io_ethtool_sset,
5741         .get_drvinfo = s2io_ethtool_gdrvinfo,
5742         .get_regs_len = s2io_ethtool_get_regs_len,
5743         .get_regs = s2io_ethtool_gregs,
5744         .get_link = ethtool_op_get_link,
5745         .get_eeprom_len = s2io_get_eeprom_len,
5746         .get_eeprom = s2io_ethtool_geeprom,
5747         .set_eeprom = s2io_ethtool_seeprom,
5748         .get_pauseparam = s2io_ethtool_getpause_data,
5749         .set_pauseparam = s2io_ethtool_setpause_data,
5750         .get_rx_csum = s2io_ethtool_get_rx_csum,
5751         .set_rx_csum = s2io_ethtool_set_rx_csum,
5752         .get_tx_csum = ethtool_op_get_tx_csum,
5753         .set_tx_csum = s2io_ethtool_op_set_tx_csum,
5754         .get_sg = ethtool_op_get_sg,
5755         .set_sg = ethtool_op_set_sg,
5756 #ifdef NETIF_F_TSO
5757         .get_tso = s2io_ethtool_op_get_tso,
5758         .set_tso = s2io_ethtool_op_set_tso,
5759 #endif
5760         .get_ufo = ethtool_op_get_ufo,
5761         .set_ufo = ethtool_op_set_ufo,
5762         .self_test_count = s2io_ethtool_self_test_count,
5763         .self_test = s2io_ethtool_test,
5764         .get_strings = s2io_ethtool_get_strings,
5765         .phys_id = s2io_ethtool_idnic,
5766         .get_stats_count = s2io_ethtool_get_stats_count,
5767         .get_ethtool_stats = s2io_get_ethtool_stats
5768 };
5769
5770 /**
5771  *  s2io_ioctl - Entry point for the Ioctl
5772  *  @dev :  Device pointer.
5773  *  @ifr :  An IOCTL specefic structure, that can contain a pointer to
5774  *  a proprietary structure used to pass information to the driver.
5775  *  @cmd :  This is used to distinguish between the different commands that
5776  *  can be passed to the IOCTL functions.
5777  *  Description:
5778  *  Currently there are no special functionality supported in IOCTL, hence
5779  *  function always return EOPNOTSUPPORTED
5780  */
5781
5782 static int s2io_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
5783 {
5784         return -EOPNOTSUPP;
5785 }
5786
5787 /**
5788  *  s2io_change_mtu - entry point to change MTU size for the device.
5789  *   @dev : device pointer.
5790  *   @new_mtu : the new MTU size for the device.
5791  *   Description: A driver entry point to change MTU size for the device.
5792  *   Before changing the MTU the device must be stopped.
5793  *  Return value:
5794  *   0 on success and an appropriate (-)ve integer as defined in errno.h
5795  *   file on failure.
5796  */
5797
5798 static int s2io_change_mtu(struct net_device *dev, int new_mtu)
5799 {
5800         nic_t *sp = dev->priv;
5801
5802         if ((new_mtu < MIN_MTU) || (new_mtu > S2IO_JUMBO_SIZE)) {
5803                 DBG_PRINT(ERR_DBG, "%s: MTU size is invalid.\n",
5804                           dev->name);
5805                 return -EPERM;
5806         }
5807
5808         dev->mtu = new_mtu;
5809         if (netif_running(dev)) {
5810                 s2io_card_down(sp);
5811                 netif_stop_queue(dev);
5812                 if (s2io_card_up(sp)) {
5813                         DBG_PRINT(ERR_DBG, "%s: Device bring up failed\n",
5814                                   __FUNCTION__);
5815                 }
5816                 if (netif_queue_stopped(dev))
5817                         netif_wake_queue(dev);
5818         } else { /* Device is down */
5819                 XENA_dev_config_t __iomem *bar0 = sp->bar0;
5820                 u64 val64 = new_mtu;
5821
5822                 writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
5823         }
5824
5825         return 0;
5826 }
5827
5828 /**
5829  *  s2io_tasklet - Bottom half of the ISR.
5830  *  @dev_adr : address of the device structure in dma_addr_t format.
5831  *  Description:
5832  *  This is the tasklet or the bottom half of the ISR. This is
5833  *  an extension of the ISR which is scheduled by the scheduler to be run
5834  *  when the load on the CPU is low. All low priority tasks of the ISR can
5835  *  be pushed into the tasklet. For now the tasklet is used only to
5836  *  replenish the Rx buffers in the Rx buffer descriptors.
5837  *  Return value:
5838  *  void.
5839  */
5840
5841 static void s2io_tasklet(unsigned long dev_addr)
5842 {
5843         struct net_device *dev = (struct net_device *) dev_addr;
5844         nic_t *sp = dev->priv;
5845         int i, ret;
5846         mac_info_t *mac_control;
5847         struct config_param *config;
5848
5849         mac_control = &sp->mac_control;
5850         config = &sp->config;
5851
5852         if (!TASKLET_IN_USE) {
5853                 for (i = 0; i < config->rx_ring_num; i++) {
5854                         ret = fill_rx_buffers(sp, i);
5855                         if (ret == -ENOMEM) {
5856                                 DBG_PRINT(ERR_DBG, "%s: Out of ",
5857                                           dev->name);
5858                                 DBG_PRINT(ERR_DBG, "memory in tasklet\n");
5859                                 break;
5860                         } else if (ret == -EFILL) {
5861                                 DBG_PRINT(ERR_DBG,
5862                                           "%s: Rx Ring %d is full\n",
5863                                           dev->name, i);
5864                                 break;
5865                         }
5866                 }
5867                 clear_bit(0, (&sp->tasklet_status));
5868         }
5869 }
5870
5871 /**
5872  * s2io_set_link - Set the LInk status
5873  * @data: long pointer to device private structue
5874  * Description: Sets the link status for the adapter
5875  */
5876
5877 static void s2io_set_link(unsigned long data)
5878 {
5879         nic_t *nic = (nic_t *) data;
5880         struct net_device *dev = nic->dev;
5881         XENA_dev_config_t __iomem *bar0 = nic->bar0;
5882         register u64 val64;
5883         u16 subid;
5884
5885         if (test_and_set_bit(0, &(nic->link_state))) {
5886                 /* The card is being reset, no point doing anything */
5887                 return;
5888         }
5889
5890         subid = nic->pdev->subsystem_device;
5891         if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
5892                 /*
5893                  * Allow a small delay for the NICs self initiated
5894                  * cleanup to complete.
5895                  */
5896                 msleep(100);
5897         }
5898
5899         val64 = readq(&bar0->adapter_status);
5900         if (verify_xena_quiescence(nic, val64, nic->device_enabled_once)) {
5901                 if (LINK_IS_UP(val64)) {
5902                         val64 = readq(&bar0->adapter_control);
5903                         val64 |= ADAPTER_CNTL_EN;
5904                         writeq(val64, &bar0->adapter_control);
5905                         if (CARDS_WITH_FAULTY_LINK_INDICATORS(nic->device_type,
5906                                                              subid)) {
5907                                 val64 = readq(&bar0->gpio_control);
5908                                 val64 |= GPIO_CTRL_GPIO_0;
5909                                 writeq(val64, &bar0->gpio_control);
5910                                 val64 = readq(&bar0->gpio_control);
5911                         } else {
5912                                 val64 |= ADAPTER_LED_ON;
5913                                 writeq(val64, &bar0->adapter_control);
5914                         }
5915                         if (s2io_link_fault_indication(nic) ==
5916                                                 MAC_RMAC_ERR_TIMER) {
5917                                 val64 = readq(&bar0->adapter_status);
5918                                 if (!LINK_IS_UP(val64)) {
5919                                         DBG_PRINT(ERR_DBG, "%s:", dev->name);
5920                                         DBG_PRINT(ERR_DBG, " Link down");
5921                                         DBG_PRINT(ERR_DBG, "after ");
5922                                         DBG_PRINT(ERR_DBG, "enabling ");
5923                                         DBG_PRINT(ERR_DBG, "device \n");
5924                                 }
5925                         }
5926                         if (nic->device_enabled_once == FALSE) {
5927                                 nic->device_enabled_once = TRUE;
5928                         }
5929                         s2io_link(nic, LINK_UP);
5930                 } else {
5931                         if (CARDS_WITH_FAULTY_LINK_INDICATORS(nic->device_type,
5932                                                               subid)) {
5933                                 val64 = readq(&bar0->gpio_control);
5934                                 val64 &= ~GPIO_CTRL_GPIO_0;
5935                                 writeq(val64, &bar0->gpio_control);
5936                                 val64 = readq(&bar0->gpio_control);
5937                         }
5938                         s2io_link(nic, LINK_DOWN);
5939                 }
5940         } else {                /* NIC is not Quiescent. */
5941                 DBG_PRINT(ERR_DBG, "%s: Error: ", dev->name);
5942                 DBG_PRINT(ERR_DBG, "device is not Quiescent\n");
5943                 netif_stop_queue(dev);
5944         }
5945         clear_bit(0, &(nic->link_state));
5946 }
5947
5948 static int set_rxd_buffer_pointer(nic_t *sp, RxD_t *rxdp, buffAdd_t *ba,
5949                            struct sk_buff **skb, u64 *temp0, u64 *temp1,
5950                            u64 *temp2, int size)
5951 {
5952         struct net_device *dev = sp->dev;
5953         struct sk_buff *frag_list;
5954
5955         if ((sp->rxd_mode == RXD_MODE_1) && (rxdp->Host_Control == 0)) {
5956                 /* allocate skb */
5957                 if (*skb) {
5958                         DBG_PRINT(INFO_DBG, "SKB is not NULL\n");
5959                         /*
5960                          * As Rx frame are not going to be processed,
5961                          * using same mapped address for the Rxd
5962                          * buffer pointer
5963                          */
5964                         ((RxD1_t*)rxdp)->Buffer0_ptr = *temp0;
5965                 } else {
5966                         *skb = dev_alloc_skb(size);
5967                         if (!(*skb)) {
5968                                 DBG_PRINT(ERR_DBG, "%s: Out of ", dev->name);
5969                                 DBG_PRINT(ERR_DBG, "memory to allocate SKBs\n");
5970                                 return -ENOMEM ;
5971                         }
5972                         /* storing the mapped addr in a temp variable
5973                          * such it will be used for next rxd whose
5974                          * Host Control is NULL
5975                          */
5976                         ((RxD1_t*)rxdp)->Buffer0_ptr = *temp0 =
5977                                 pci_map_single( sp->pdev, (*skb)->data,
5978                                         size - NET_IP_ALIGN,
5979                                         PCI_DMA_FROMDEVICE);
5980                         rxdp->Host_Control = (unsigned long) (*skb);
5981                 }
5982         } else if ((sp->rxd_mode == RXD_MODE_3B) && (rxdp->Host_Control == 0)) {
5983                 /* Two buffer Mode */
5984                 if (*skb) {
5985                         ((RxD3_t*)rxdp)->Buffer2_ptr = *temp2;
5986                         ((RxD3_t*)rxdp)->Buffer0_ptr = *temp0;
5987                         ((RxD3_t*)rxdp)->Buffer1_ptr = *temp1;
5988                 } else {
5989                         *skb = dev_alloc_skb(size);
5990                         ((RxD3_t*)rxdp)->Buffer2_ptr = *temp2 =
5991                                 pci_map_single(sp->pdev, (*skb)->data,
5992                                                dev->mtu + 4,
5993                                                PCI_DMA_FROMDEVICE);
5994                         ((RxD3_t*)rxdp)->Buffer0_ptr = *temp0 =
5995                                 pci_map_single( sp->pdev, ba->ba_0, BUF0_LEN,
5996                                                 PCI_DMA_FROMDEVICE);
5997                         rxdp->Host_Control = (unsigned long) (*skb);
5998
5999                         /* Buffer-1 will be dummy buffer not used */
6000                         ((RxD3_t*)rxdp)->Buffer1_ptr = *temp1 =
6001                                 pci_map_single(sp->pdev, ba->ba_1, BUF1_LEN,
6002                                                PCI_DMA_FROMDEVICE);
6003                 }
6004         } else if ((rxdp->Host_Control == 0)) {
6005                 /* Three buffer mode */
6006                 if (*skb) {
6007                         ((RxD3_t*)rxdp)->Buffer0_ptr = *temp0;
6008                         ((RxD3_t*)rxdp)->Buffer1_ptr = *temp1;
6009                         ((RxD3_t*)rxdp)->Buffer2_ptr = *temp2;
6010                 } else {
6011                         *skb = dev_alloc_skb(size);
6012
6013                         ((RxD3_t*)rxdp)->Buffer0_ptr = *temp0 =
6014                                 pci_map_single(sp->pdev, ba->ba_0, BUF0_LEN,
6015                                                PCI_DMA_FROMDEVICE);
6016                         /* Buffer-1 receives L3/L4 headers */
6017                         ((RxD3_t*)rxdp)->Buffer1_ptr = *temp1 =
6018                                 pci_map_single( sp->pdev, (*skb)->data,
6019                                                 l3l4hdr_size + 4,
6020                                                 PCI_DMA_FROMDEVICE);
6021                         /*
6022                          * skb_shinfo(skb)->frag_list will have L4
6023                          * data payload
6024                          */
6025                         skb_shinfo(*skb)->frag_list = dev_alloc_skb(dev->mtu +
6026                                                                    ALIGN_SIZE);
6027                         if (skb_shinfo(*skb)->frag_list == NULL) {
6028                                 DBG_PRINT(ERR_DBG, "%s: dev_alloc_skb \
6029                                           failed\n ", dev->name);
6030                                 return -ENOMEM ;
6031                         }
6032                         frag_list = skb_shinfo(*skb)->frag_list;
6033                         frag_list->next = NULL;
6034                         /*
6035                          * Buffer-2 receives L4 data payload
6036                          */
6037                         ((RxD3_t*)rxdp)->Buffer2_ptr = *temp2 =
6038                                 pci_map_single( sp->pdev, frag_list->data,
6039                                                 dev->mtu, PCI_DMA_FROMDEVICE);
6040                 }
6041         }
6042         return 0;
6043 }
6044 static void set_rxd_buffer_size(nic_t *sp, RxD_t *rxdp, int size)
6045 {
6046         struct net_device *dev = sp->dev;
6047         if (sp->rxd_mode == RXD_MODE_1) {
6048                 rxdp->Control_2 = SET_BUFFER0_SIZE_1( size - NET_IP_ALIGN);
6049         } else if (sp->rxd_mode == RXD_MODE_3B) {
6050                 rxdp->Control_2 = SET_BUFFER0_SIZE_3(BUF0_LEN);
6051                 rxdp->Control_2 |= SET_BUFFER1_SIZE_3(1);
6052                 rxdp->Control_2 |= SET_BUFFER2_SIZE_3( dev->mtu + 4);
6053         } else {
6054                 rxdp->Control_2 = SET_BUFFER0_SIZE_3(BUF0_LEN);
6055                 rxdp->Control_2 |= SET_BUFFER1_SIZE_3(l3l4hdr_size + 4);
6056                 rxdp->Control_2 |= SET_BUFFER2_SIZE_3(dev->mtu);
6057         }
6058 }
6059
6060 static  int rxd_owner_bit_reset(nic_t *sp)
6061 {
6062         int i, j, k, blk_cnt = 0, size;
6063         mac_info_t * mac_control = &sp->mac_control;
6064         struct config_param *config = &sp->config;
6065         struct net_device *dev = sp->dev;
6066         RxD_t *rxdp = NULL;
6067         struct sk_buff *skb = NULL;
6068         buffAdd_t *ba = NULL;
6069         u64 temp0_64 = 0, temp1_64 = 0, temp2_64 = 0;
6070
6071         /* Calculate the size based on ring mode */
6072         size = dev->mtu + HEADER_ETHERNET_II_802_3_SIZE +
6073                 HEADER_802_2_SIZE + HEADER_SNAP_SIZE;
6074         if (sp->rxd_mode == RXD_MODE_1)
6075                 size += NET_IP_ALIGN;
6076         else if (sp->rxd_mode == RXD_MODE_3B)
6077                 size = dev->mtu + ALIGN_SIZE + BUF0_LEN + 4;
6078         else
6079                 size = l3l4hdr_size + ALIGN_SIZE + BUF0_LEN + 4;
6080
6081         for (i = 0; i < config->rx_ring_num; i++) {
6082                 blk_cnt = config->rx_cfg[i].num_rxd /
6083                         (rxd_count[sp->rxd_mode] +1);
6084
6085                 for (j = 0; j < blk_cnt; j++) {
6086                         for (k = 0; k < rxd_count[sp->rxd_mode]; k++) {
6087                                 rxdp = mac_control->rings[i].
6088                                         rx_blocks[j].rxds[k].virt_addr;
6089                                 if(sp->rxd_mode >= RXD_MODE_3A)
6090                                         ba = &mac_control->rings[i].ba[j][k];
6091                                 set_rxd_buffer_pointer(sp, rxdp, ba,
6092                                                        &skb,(u64 *)&temp0_64,
6093                                                        (u64 *)&temp1_64,
6094                                                        (u64 *)&temp2_64, size);
6095
6096                                 set_rxd_buffer_size(sp, rxdp, size);
6097                                 wmb();
6098                                 /* flip the Ownership bit to Hardware */
6099                                 rxdp->Control_1 |= RXD_OWN_XENA;
6100                         }
6101                 }
6102         }
6103         return 0;
6104
6105 }
6106
6107 static int s2io_add_isr(nic_t * sp)
6108 {
6109         int ret = 0;
6110         struct net_device *dev = sp->dev;
6111         int err = 0;
6112
6113         if (sp->intr_type == MSI)
6114                 ret = s2io_enable_msi(sp);
6115         else if (sp->intr_type == MSI_X)
6116                 ret = s2io_enable_msi_x(sp);
6117         if (ret) {
6118                 DBG_PRINT(ERR_DBG, "%s: Defaulting to INTA\n", dev->name);
6119                 sp->intr_type = INTA;
6120         }
6121
6122         /* Store the values of the MSIX table in the nic_t structure */
6123         store_xmsi_data(sp);
6124
6125         /* After proper initialization of H/W, register ISR */
6126         if (sp->intr_type == MSI) {
6127                 err = request_irq((int) sp->pdev->irq, s2io_msi_handle,
6128                         IRQF_SHARED, sp->name, dev);
6129                 if (err) {
6130                         pci_disable_msi(sp->pdev);
6131                         DBG_PRINT(ERR_DBG, "%s: MSI registration failed\n",
6132                                   dev->name);
6133                         return -1;
6134                 }
6135         }
6136         if (sp->intr_type == MSI_X) {
6137                 int i;
6138
6139                 for (i=1; (sp->s2io_entries[i].in_use == MSIX_FLG); i++) {
6140                         if (sp->s2io_entries[i].type == MSIX_FIFO_TYPE) {
6141                                 sprintf(sp->desc[i], "%s:MSI-X-%d-TX",
6142                                         dev->name, i);
6143                                 err = request_irq(sp->entries[i].vector,
6144                                           s2io_msix_fifo_handle, 0, sp->desc[i],
6145                                                   sp->s2io_entries[i].arg);
6146                                 DBG_PRINT(ERR_DBG, "%s @ 0x%llx\n", sp->desc[i],
6147                                 (unsigned long long)sp->msix_info[i].addr);
6148                         } else {
6149                                 sprintf(sp->desc[i], "%s:MSI-X-%d-RX",
6150                                         dev->name, i);
6151                                 err = request_irq(sp->entries[i].vector,
6152                                           s2io_msix_ring_handle, 0, sp->desc[i],
6153                                                   sp->s2io_entries[i].arg);
6154                                 DBG_PRINT(ERR_DBG, "%s @ 0x%llx\n", sp->desc[i],
6155                                 (unsigned long long)sp->msix_info[i].addr);
6156                         }
6157                         if (err) {
6158                                 DBG_PRINT(ERR_DBG,"%s:MSI-X-%d registration "
6159                                           "failed\n", dev->name, i);
6160                                 DBG_PRINT(ERR_DBG, "Returned: %d\n", err);
6161                                 return -1;
6162                         }
6163                         sp->s2io_entries[i].in_use = MSIX_REGISTERED_SUCCESS;
6164                 }
6165         }
6166         if (sp->intr_type == INTA) {
6167                 err = request_irq((int) sp->pdev->irq, s2io_isr, IRQF_SHARED,
6168                                 sp->name, dev);
6169                 if (err) {
6170                         DBG_PRINT(ERR_DBG, "%s: ISR registration failed\n",
6171                                   dev->name);
6172                         return -1;
6173                 }
6174         }
6175         return 0;
6176 }
6177 static void s2io_rem_isr(nic_t * sp)
6178 {
6179         int cnt = 0;
6180         struct net_device *dev = sp->dev;
6181
6182         if (sp->intr_type == MSI_X) {
6183                 int i;
6184                 u16 msi_control;
6185
6186                 for (i=1; (sp->s2io_entries[i].in_use ==
6187                         MSIX_REGISTERED_SUCCESS); i++) {
6188                         int vector = sp->entries[i].vector;
6189                         void *arg = sp->s2io_entries[i].arg;
6190
6191                         free_irq(vector, arg);
6192                 }
6193                 pci_read_config_word(sp->pdev, 0x42, &msi_control);
6194                 msi_control &= 0xFFFE; /* Disable MSI */
6195                 pci_write_config_word(sp->pdev, 0x42, msi_control);
6196
6197                 pci_disable_msix(sp->pdev);
6198         } else {
6199                 free_irq(sp->pdev->irq, dev);
6200                 if (sp->intr_type == MSI) {
6201                         u16 val;
6202
6203                         pci_disable_msi(sp->pdev);
6204                         pci_read_config_word(sp->pdev, 0x4c, &val);
6205                         val ^= 0x1;
6206                         pci_write_config_word(sp->pdev, 0x4c, val);
6207                 }
6208         }
6209         /* Waiting till all Interrupt handlers are complete */
6210         cnt = 0;
6211         do {
6212                 msleep(10);
6213                 if (!atomic_read(&sp->isr_cnt))
6214                         break;
6215                 cnt++;
6216         } while(cnt < 5);
6217 }
6218
6219 static void s2io_card_down(nic_t * sp)
6220 {
6221         int cnt = 0;
6222         XENA_dev_config_t __iomem *bar0 = sp->bar0;
6223         unsigned long flags;
6224         register u64 val64 = 0;
6225
6226         del_timer_sync(&sp->alarm_timer);
6227         /* If s2io_set_link task is executing, wait till it completes. */
6228         while (test_and_set_bit(0, &(sp->link_state))) {
6229                 msleep(50);
6230         }
6231         atomic_set(&sp->card_state, CARD_DOWN);
6232
6233         /* disable Tx and Rx traffic on the NIC */
6234         stop_nic(sp);
6235
6236         s2io_rem_isr(sp);
6237
6238         /* Kill tasklet. */
6239         tasklet_kill(&sp->task);
6240
6241         /* Check if the device is Quiescent and then Reset the NIC */
6242         do {
6243                 /* As per the HW requirement we need to replenish the
6244                  * receive buffer to avoid the ring bump. Since there is
6245                  * no intention of processing the Rx frame at this pointwe are
6246                  * just settting the ownership bit of rxd in Each Rx
6247                  * ring to HW and set the appropriate buffer size
6248                  * based on the ring mode
6249                  */
6250                 rxd_owner_bit_reset(sp);
6251
6252                 val64 = readq(&bar0->adapter_status);
6253                 if (verify_xena_quiescence(sp, val64, sp->device_enabled_once)) {
6254                         break;
6255                 }
6256
6257                 msleep(50);
6258                 cnt++;
6259                 if (cnt == 10) {
6260                         DBG_PRINT(ERR_DBG,
6261                                   "s2io_close:Device not Quiescent ");
6262                         DBG_PRINT(ERR_DBG, "adaper status reads 0x%llx\n",
6263                                   (unsigned long long) val64);
6264                         break;
6265                 }
6266         } while (1);
6267         s2io_reset(sp);
6268
6269         spin_lock_irqsave(&sp->tx_lock, flags);
6270         /* Free all Tx buffers */
6271         free_tx_buffers(sp);
6272         spin_unlock_irqrestore(&sp->tx_lock, flags);
6273
6274         /* Free all Rx buffers */
6275         spin_lock_irqsave(&sp->rx_lock, flags);
6276         free_rx_buffers(sp);
6277         spin_unlock_irqrestore(&sp->rx_lock, flags);
6278
6279         clear_bit(0, &(sp->link_state));
6280 }
6281
6282 static int s2io_card_up(nic_t * sp)
6283 {
6284         int i, ret = 0;
6285         mac_info_t *mac_control;
6286         struct config_param *config;
6287         struct net_device *dev = (struct net_device *) sp->dev;
6288         u16 interruptible;
6289
6290         /* Initialize the H/W I/O registers */
6291         if (init_nic(sp) != 0) {
6292                 DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
6293                           dev->name);
6294                 s2io_reset(sp);
6295                 return -ENODEV;
6296         }
6297
6298         /*
6299          * Initializing the Rx buffers. For now we are considering only 1
6300          * Rx ring and initializing buffers into 30 Rx blocks
6301          */
6302         mac_control = &sp->mac_control;
6303         config = &sp->config;
6304
6305         for (i = 0; i < config->rx_ring_num; i++) {
6306                 if ((ret = fill_rx_buffers(sp, i))) {
6307                         DBG_PRINT(ERR_DBG, "%s: Out of memory in Open\n",
6308                                   dev->name);
6309                         s2io_reset(sp);
6310                         free_rx_buffers(sp);
6311                         return -ENOMEM;
6312                 }
6313                 DBG_PRINT(INFO_DBG, "Buf in ring:%d is %d:\n", i,
6314                           atomic_read(&sp->rx_bufs_left[i]));
6315         }
6316
6317         /* Setting its receive mode */
6318         s2io_set_multicast(dev);
6319
6320         if (sp->lro) {
6321                 /* Initialize max aggregatable pkts per session based on MTU */
6322                 sp->lro_max_aggr_per_sess = ((1<<16) - 1) / dev->mtu;
6323                 /* Check if we can use(if specified) user provided value */
6324                 if (lro_max_pkts < sp->lro_max_aggr_per_sess)
6325                         sp->lro_max_aggr_per_sess = lro_max_pkts;
6326         }
6327
6328         /* Enable Rx Traffic and interrupts on the NIC */
6329         if (start_nic(sp)) {
6330                 DBG_PRINT(ERR_DBG, "%s: Starting NIC failed\n", dev->name);
6331                 s2io_reset(sp);
6332                 free_rx_buffers(sp);
6333                 return -ENODEV;
6334         }
6335
6336         /* Add interrupt service routine */
6337         if (s2io_add_isr(sp) != 0) {
6338                 if (sp->intr_type == MSI_X)
6339                         s2io_rem_isr(sp);
6340                 s2io_reset(sp);
6341                 free_rx_buffers(sp);
6342                 return -ENODEV;
6343         }
6344
6345         S2IO_TIMER_CONF(sp->alarm_timer, s2io_alarm_handle, sp, (HZ/2));
6346
6347         /* Enable tasklet for the device */
6348         tasklet_init(&sp->task, s2io_tasklet, (unsigned long) dev);
6349
6350         /*  Enable select interrupts */
6351         if (sp->intr_type != INTA)
6352                 en_dis_able_nic_intrs(sp, ENA_ALL_INTRS, DISABLE_INTRS);
6353         else {
6354                 interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR;
6355                 interruptible |= TX_PIC_INTR | RX_PIC_INTR;
6356                 interruptible |= TX_MAC_INTR | RX_MAC_INTR;
6357                 en_dis_able_nic_intrs(sp, interruptible, ENABLE_INTRS);
6358         }
6359
6360
6361         atomic_set(&sp->card_state, CARD_UP);
6362         return 0;
6363 }
6364
6365 /**
6366  * s2io_restart_nic - Resets the NIC.
6367  * @data : long pointer to the device private structure
6368  * Description:
6369  * This function is scheduled to be run by the s2io_tx_watchdog
6370  * function after 0.5 secs to reset the NIC. The idea is to reduce
6371  * the run time of the watch dog routine which is run holding a
6372  * spin lock.
6373  */
6374
6375 static void s2io_restart_nic(unsigned long data)
6376 {
6377         struct net_device *dev = (struct net_device *) data;
6378         nic_t *sp = dev->priv;
6379
6380         s2io_card_down(sp);
6381         if (s2io_card_up(sp)) {
6382                 DBG_PRINT(ERR_DBG, "%s: Device bring up failed\n",
6383                           dev->name);
6384         }
6385         netif_wake_queue(dev);
6386         DBG_PRINT(ERR_DBG, "%s: was reset by Tx watchdog timer\n",
6387                   dev->name);
6388
6389 }
6390
6391 /**
6392  *  s2io_tx_watchdog - Watchdog for transmit side.
6393  *  @dev : Pointer to net device structure
6394  *  Description:
6395  *  This function is triggered if the Tx Queue is stopped
6396  *  for a pre-defined amount of time when the Interface is still up.
6397  *  If the Interface is jammed in such a situation, the hardware is
6398  *  reset (by s2io_close) and restarted again (by s2io_open) to
6399  *  overcome any problem that might have been caused in the hardware.
6400  *  Return value:
6401  *  void
6402  */
6403
6404 static void s2io_tx_watchdog(struct net_device *dev)
6405 {
6406         nic_t *sp = dev->priv;
6407
6408         if (netif_carrier_ok(dev)) {
6409                 schedule_work(&sp->rst_timer_task);
6410                 sp->mac_control.stats_info->sw_stat.soft_reset_cnt++;
6411         }
6412 }
6413
6414 /**
6415  *   rx_osm_handler - To perform some OS related operations on SKB.
6416  *   @sp: private member of the device structure,pointer to s2io_nic structure.
6417  *   @skb : the socket buffer pointer.
6418  *   @len : length of the packet
6419  *   @cksum : FCS checksum of the frame.
6420  *   @ring_no : the ring from which this RxD was extracted.
6421  *   Description:
6422  *   This function is called by the Rx interrupt serivce routine to perform
6423  *   some OS related operations on the SKB before passing it to the upper
6424  *   layers. It mainly checks if the checksum is OK, if so adds it to the
6425  *   SKBs cksum variable, increments the Rx packet count and passes the SKB
6426  *   to the upper layer. If the checksum is wrong, it increments the Rx
6427  *   packet error count, frees the SKB and returns error.
6428  *   Return value:
6429  *   SUCCESS on success and -1 on failure.
6430  */
6431 static int rx_osm_handler(ring_info_t *ring_data, RxD_t * rxdp)
6432 {
6433         nic_t *sp = ring_data->nic;
6434         struct net_device *dev = (struct net_device *) sp->dev;
6435         struct sk_buff *skb = (struct sk_buff *)
6436                 ((unsigned long) rxdp->Host_Control);
6437         int ring_no = ring_data->ring_no;
6438         u16 l3_csum, l4_csum;
6439         unsigned long long err = rxdp->Control_1 & RXD_T_CODE;
6440         lro_t *lro;
6441
6442         skb->dev = dev;
6443
6444         if (err) {
6445                 /* Check for parity error */
6446                 if (err & 0x1) {
6447                         sp->mac_control.stats_info->sw_stat.parity_err_cnt++;
6448                 }
6449
6450                 /*
6451                 * Drop the packet if bad transfer code. Exception being
6452                 * 0x5, which could be due to unsupported IPv6 extension header.
6453                 * In this case, we let stack handle the packet.
6454                 * Note that in this case, since checksum will be incorrect,
6455                 * stack will validate the same.
6456                 */
6457                 if (err && ((err >> 48) != 0x5)) {
6458                         DBG_PRINT(ERR_DBG, "%s: Rx error Value: 0x%llx\n",
6459                                 dev->name, err);
6460                         sp->stats.rx_crc_errors++;
6461                         dev_kfree_skb(skb);
6462                         atomic_dec(&sp->rx_bufs_left[ring_no]);
6463                         rxdp->Host_Control = 0;
6464                         return 0;
6465                 }
6466         }
6467
6468         /* Updating statistics */
6469         rxdp->Host_Control = 0;
6470         sp->rx_pkt_count++;
6471         sp->stats.rx_packets++;
6472         if (sp->rxd_mode == RXD_MODE_1) {
6473                 int len = RXD_GET_BUFFER0_SIZE_1(rxdp->Control_2);
6474
6475                 sp->stats.rx_bytes += len;
6476                 skb_put(skb, len);
6477
6478         } else if (sp->rxd_mode >= RXD_MODE_3A) {
6479                 int get_block = ring_data->rx_curr_get_info.block_index;
6480                 int get_off = ring_data->rx_curr_get_info.offset;
6481                 int buf0_len = RXD_GET_BUFFER0_SIZE_3(rxdp->Control_2);
6482                 int buf2_len = RXD_GET_BUFFER2_SIZE_3(rxdp->Control_2);
6483                 unsigned char *buff = skb_push(skb, buf0_len);
6484
6485                 buffAdd_t *ba = &ring_data->ba[get_block][get_off];
6486                 sp->stats.rx_bytes += buf0_len + buf2_len;
6487                 memcpy(buff, ba->ba_0, buf0_len);
6488
6489                 if (sp->rxd_mode == RXD_MODE_3A) {
6490                         int buf1_len = RXD_GET_BUFFER1_SIZE_3(rxdp->Control_2);
6491
6492                         skb_put(skb, buf1_len);
6493                         skb->len += buf2_len;
6494                         skb->data_len += buf2_len;
6495                         skb->truesize += buf2_len;
6496                         skb_put(skb_shinfo(skb)->frag_list, buf2_len);
6497                         sp->stats.rx_bytes += buf1_len;
6498
6499                 } else
6500                         skb_put(skb, buf2_len);
6501         }
6502
6503         if ((rxdp->Control_1 & TCP_OR_UDP_FRAME) && ((!sp->lro) ||
6504             (sp->lro && (!(rxdp->Control_1 & RXD_FRAME_IP_FRAG)))) &&
6505             (sp->rx_csum)) {
6506                 l3_csum = RXD_GET_L3_CKSUM(rxdp->Control_1);
6507                 l4_csum = RXD_GET_L4_CKSUM(rxdp->Control_1);
6508                 if ((l3_csum == L3_CKSUM_OK) && (l4_csum == L4_CKSUM_OK)) {
6509                         /*
6510                          * NIC verifies if the Checksum of the received
6511                          * frame is Ok or not and accordingly returns
6512                          * a flag in the RxD.
6513                          */
6514                         skb->ip_summed = CHECKSUM_UNNECESSARY;
6515                         if (sp->lro) {
6516                                 u32 tcp_len;
6517                                 u8 *tcp;
6518                                 int ret = 0;
6519
6520                                 ret = s2io_club_tcp_session(skb->data, &tcp,
6521                                                 &tcp_len, &lro, rxdp, sp);
6522                                 switch (ret) {
6523                                         case 3: /* Begin anew */
6524                                                 lro->parent = skb;
6525                                                 goto aggregate;
6526                                         case 1: /* Aggregate */
6527                                         {
6528                                                 lro_append_pkt(sp, lro,
6529                                                         skb, tcp_len);
6530                                                 goto aggregate;
6531                                         }
6532                                         case 4: /* Flush session */
6533                                         {
6534                                                 lro_append_pkt(sp, lro,
6535                                                         skb, tcp_len);
6536                                                 queue_rx_frame(lro->parent);
6537                                                 clear_lro_session(lro);
6538                                                 sp->mac_control.stats_info->
6539                                                     sw_stat.flush_max_pkts++;
6540                                                 goto aggregate;
6541                                         }
6542                                         case 2: /* Flush both */
6543                                                 lro->parent->data_len =
6544                                                         lro->frags_len;
6545                                                 sp->mac_control.stats_info->
6546                                                      sw_stat.sending_both++;
6547                                                 queue_rx_frame(lro->parent);
6548                                                 clear_lro_session(lro);
6549                                                 goto send_up;
6550                                         case 0: /* sessions exceeded */
6551                                         case -1: /* non-TCP or not
6552                                                   * L2 aggregatable
6553                                                   */
6554                                         case 5: /*
6555                                                  * First pkt in session not
6556                                                  * L3/L4 aggregatable
6557                                                  */
6558                                                 break;
6559                                         default:
6560                                                 DBG_PRINT(ERR_DBG,
6561                                                         "%s: Samadhana!!\n",
6562                                                          __FUNCTION__);
6563                                                 BUG();
6564                                 }
6565                         }
6566                 } else {
6567                         /*
6568                          * Packet with erroneous checksum, let the
6569                          * upper layers deal with it.
6570                          */
6571                         skb->ip_summed = CHECKSUM_NONE;
6572                 }
6573         } else {
6574                 skb->ip_summed = CHECKSUM_NONE;
6575         }
6576
6577         if (!sp->lro) {
6578                 skb->protocol = eth_type_trans(skb, dev);
6579 #ifdef CONFIG_S2IO_NAPI
6580                 if (sp->vlgrp && RXD_GET_VLAN_TAG(rxdp->Control_2)) {
6581                         /* Queueing the vlan frame to the upper layer */
6582                         vlan_hwaccel_receive_skb(skb, sp->vlgrp,
6583                                 RXD_GET_VLAN_TAG(rxdp->Control_2));
6584                 } else {
6585                         netif_receive_skb(skb);
6586                 }
6587 #else
6588                 if (sp->vlgrp && RXD_GET_VLAN_TAG(rxdp->Control_2)) {
6589                         /* Queueing the vlan frame to the upper layer */
6590                         vlan_hwaccel_rx(skb, sp->vlgrp,
6591                                 RXD_GET_VLAN_TAG(rxdp->Control_2));
6592                 } else {
6593                         netif_rx(skb);
6594                 }
6595 #endif
6596         } else {
6597 send_up:
6598                 queue_rx_frame(skb);
6599         }               
6600         dev->last_rx = jiffies;
6601 aggregate:
6602         atomic_dec(&sp->rx_bufs_left[ring_no]);
6603         return SUCCESS;
6604 }
6605
6606 /**
6607  *  s2io_link - stops/starts the Tx queue.
6608  *  @sp : private member of the device structure, which is a pointer to the
6609  *  s2io_nic structure.
6610  *  @link : inidicates whether link is UP/DOWN.
6611  *  Description:
6612  *  This function stops/starts the Tx queue depending on whether the link
6613  *  status of the NIC is is down or up. This is called by the Alarm
6614  *  interrupt handler whenever a link change interrupt comes up.
6615  *  Return value:
6616  *  void.
6617  */
6618
6619 static void s2io_link(nic_t * sp, int link)
6620 {
6621         struct net_device *dev = (struct net_device *) sp->dev;
6622
6623         if (link != sp->last_link_state) {
6624                 if (link == LINK_DOWN) {
6625                         DBG_PRINT(ERR_DBG, "%s: Link down\n", dev->name);
6626                         netif_carrier_off(dev);
6627                 } else {
6628                         DBG_PRINT(ERR_DBG, "%s: Link Up\n", dev->name);
6629                         netif_carrier_on(dev);
6630                 }
6631         }
6632         sp->last_link_state = link;
6633 }
6634
6635 /**
6636  *  get_xena_rev_id - to identify revision ID of xena.
6637  *  @pdev : PCI Dev structure
6638  *  Description:
6639  *  Function to identify the Revision ID of xena.
6640  *  Return value:
6641  *  returns the revision ID of the device.
6642  */
6643
6644 static int get_xena_rev_id(struct pci_dev *pdev)
6645 {
6646         u8 id = 0;
6647         int ret;
6648         ret = pci_read_config_byte(pdev, PCI_REVISION_ID, (u8 *) & id);
6649         return id;
6650 }
6651
6652 /**
6653  *  s2io_init_pci -Initialization of PCI and PCI-X configuration registers .
6654  *  @sp : private member of the device structure, which is a pointer to the
6655  *  s2io_nic structure.
6656  *  Description:
6657  *  This function initializes a few of the PCI and PCI-X configuration registers
6658  *  with recommended values.
6659  *  Return value:
6660  *  void
6661  */
6662
6663 static void s2io_init_pci(nic_t * sp)
6664 {
6665         u16 pci_cmd = 0, pcix_cmd = 0;
6666
6667         /* Enable Data Parity Error Recovery in PCI-X command register. */
6668         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
6669                              &(pcix_cmd));
6670         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
6671                               (pcix_cmd | 1));
6672         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
6673                              &(pcix_cmd));
6674
6675         /* Set the PErr Response bit in PCI command register. */
6676         pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
6677         pci_write_config_word(sp->pdev, PCI_COMMAND,
6678                               (pci_cmd | PCI_COMMAND_PARITY));
6679         pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
6680 }
6681
6682 static int s2io_verify_parm(struct pci_dev *pdev, u8 *dev_intr_type)
6683 {
6684         if ( tx_fifo_num > 8) {
6685                 DBG_PRINT(ERR_DBG, "s2io: Requested number of Tx fifos not "
6686                          "supported\n");
6687                 DBG_PRINT(ERR_DBG, "s2io: Default to 8 Tx fifos\n");
6688                 tx_fifo_num = 8;
6689         }
6690         if ( rx_ring_num > 8) {
6691                 DBG_PRINT(ERR_DBG, "s2io: Requested number of Rx rings not "
6692                          "supported\n");
6693                 DBG_PRINT(ERR_DBG, "s2io: Default to 8 Rx rings\n");
6694                 rx_ring_num = 8;
6695         }
6696 #ifdef CONFIG_S2IO_NAPI
6697         if (*dev_intr_type != INTA) {
6698                 DBG_PRINT(ERR_DBG, "s2io: NAPI cannot be enabled when "
6699                           "MSI/MSI-X is enabled. Defaulting to INTA\n");
6700                 *dev_intr_type = INTA;
6701         }
6702 #endif
6703 #ifndef CONFIG_PCI_MSI
6704         if (*dev_intr_type != INTA) {
6705                 DBG_PRINT(ERR_DBG, "s2io: This kernel does not support"
6706                           "MSI/MSI-X. Defaulting to INTA\n");
6707                 *dev_intr_type = INTA;
6708         }
6709 #else
6710         if (*dev_intr_type > MSI_X) {
6711                 DBG_PRINT(ERR_DBG, "s2io: Wrong intr_type requested. "
6712                           "Defaulting to INTA\n");
6713                 *dev_intr_type = INTA;
6714         }
6715 #endif
6716         if ((*dev_intr_type == MSI_X) &&
6717                         ((pdev->device != PCI_DEVICE_ID_HERC_WIN) &&
6718                         (pdev->device != PCI_DEVICE_ID_HERC_UNI))) {
6719                 DBG_PRINT(ERR_DBG, "s2io: Xframe I does not support MSI_X. " 
6720                                         "Defaulting to INTA\n");
6721                 *dev_intr_type = INTA;
6722         }
6723         if (rx_ring_mode > 3) {
6724                 DBG_PRINT(ERR_DBG, "s2io: Requested ring mode not supported\n");
6725                 DBG_PRINT(ERR_DBG, "s2io: Defaulting to 3-buffer mode\n");
6726                 rx_ring_mode = 3;
6727         }
6728         return SUCCESS;
6729 }
6730
6731 /**
6732  *  s2io_init_nic - Initialization of the adapter .
6733  *  @pdev : structure containing the PCI related information of the device.
6734  *  @pre: List of PCI devices supported by the driver listed in s2io_tbl.
6735  *  Description:
6736  *  The function initializes an adapter identified by the pci_dec structure.
6737  *  All OS related initialization including memory and device structure and
6738  *  initlaization of the device private variable is done. Also the swapper
6739  *  control register is initialized to enable read and write into the I/O
6740  *  registers of the device.
6741  *  Return value:
6742  *  returns 0 on success and negative on failure.
6743  */
6744
6745 static int __devinit
6746 s2io_init_nic(struct pci_dev *pdev, const struct pci_device_id *pre)
6747 {
6748         nic_t *sp;
6749         struct net_device *dev;
6750         int i, j, ret;
6751         int dma_flag = FALSE;
6752         u32 mac_up, mac_down;
6753         u64 val64 = 0, tmp64 = 0;
6754         XENA_dev_config_t __iomem *bar0 = NULL;
6755         u16 subid;
6756         mac_info_t *mac_control;
6757         struct config_param *config;
6758         int mode;
6759         u8 dev_intr_type = intr_type;
6760
6761         if ((ret = s2io_verify_parm(pdev, &dev_intr_type)))
6762                 return ret;
6763
6764         if ((ret = pci_enable_device(pdev))) {
6765                 DBG_PRINT(ERR_DBG,
6766                           "s2io_init_nic: pci_enable_device failed\n");
6767                 return ret;
6768         }
6769
6770         if (!pci_set_dma_mask(pdev, DMA_64BIT_MASK)) {
6771                 DBG_PRINT(INIT_DBG, "s2io_init_nic: Using 64bit DMA\n");
6772                 dma_flag = TRUE;
6773                 if (pci_set_consistent_dma_mask
6774                     (pdev, DMA_64BIT_MASK)) {
6775                         DBG_PRINT(ERR_DBG,
6776                                   "Unable to obtain 64bit DMA for \
6777                                         consistent allocations\n");
6778                         pci_disable_device(pdev);
6779                         return -ENOMEM;
6780                 }
6781         } else if (!pci_set_dma_mask(pdev, DMA_32BIT_MASK)) {
6782                 DBG_PRINT(INIT_DBG, "s2io_init_nic: Using 32bit DMA\n");
6783         } else {
6784                 pci_disable_device(pdev);
6785                 return -ENOMEM;
6786         }
6787         if (dev_intr_type != MSI_X) {
6788                 if (pci_request_regions(pdev, s2io_driver_name)) {
6789                         DBG_PRINT(ERR_DBG, "Request Regions failed\n");
6790                         pci_disable_device(pdev);
6791                         return -ENODEV;
6792                 }
6793         }
6794         else {
6795                 if (!(request_mem_region(pci_resource_start(pdev, 0),
6796                          pci_resource_len(pdev, 0), s2io_driver_name))) {
6797                         DBG_PRINT(ERR_DBG, "bar0 Request Regions failed\n");
6798                         pci_disable_device(pdev);
6799                         return -ENODEV;
6800                 }
6801                 if (!(request_mem_region(pci_resource_start(pdev, 2),
6802                          pci_resource_len(pdev, 2), s2io_driver_name))) {
6803                         DBG_PRINT(ERR_DBG, "bar1 Request Regions failed\n");
6804                         release_mem_region(pci_resource_start(pdev, 0),
6805                                    pci_resource_len(pdev, 0));
6806                         pci_disable_device(pdev);
6807                         return -ENODEV;
6808                 }
6809         }
6810
6811         dev = alloc_etherdev(sizeof(nic_t));
6812         if (dev == NULL) {
6813                 DBG_PRINT(ERR_DBG, "Device allocation failed\n");
6814                 pci_disable_device(pdev);
6815                 pci_release_regions(pdev);
6816                 return -ENODEV;
6817         }
6818
6819         pci_set_master(pdev);
6820         pci_set_drvdata(pdev, dev);
6821         SET_MODULE_OWNER(dev);
6822         SET_NETDEV_DEV(dev, &pdev->dev);
6823
6824         /*  Private member variable initialized to s2io NIC structure */
6825         sp = dev->priv;
6826         memset(sp, 0, sizeof(nic_t));
6827         sp->dev = dev;
6828         sp->pdev = pdev;
6829         sp->high_dma_flag = dma_flag;
6830         sp->device_enabled_once = FALSE;
6831         if (rx_ring_mode == 1)
6832                 sp->rxd_mode = RXD_MODE_1;
6833         if (rx_ring_mode == 2)
6834                 sp->rxd_mode = RXD_MODE_3B;
6835         if (rx_ring_mode == 3)
6836                 sp->rxd_mode = RXD_MODE_3A;
6837
6838         sp->intr_type = dev_intr_type;
6839
6840         if ((pdev->device == PCI_DEVICE_ID_HERC_WIN) ||
6841                 (pdev->device == PCI_DEVICE_ID_HERC_UNI))
6842                 sp->device_type = XFRAME_II_DEVICE;
6843         else
6844                 sp->device_type = XFRAME_I_DEVICE;
6845
6846         sp->lro = lro;
6847                 
6848         /* Initialize some PCI/PCI-X fields of the NIC. */
6849         s2io_init_pci(sp);
6850
6851         /*
6852          * Setting the device configuration parameters.
6853          * Most of these parameters can be specified by the user during
6854          * module insertion as they are module loadable parameters. If
6855          * these parameters are not not specified during load time, they
6856          * are initialized with default values.
6857          */
6858         mac_control = &sp->mac_control;
6859         config = &sp->config;
6860
6861         /* Tx side parameters. */
6862         config->tx_fifo_num = tx_fifo_num;
6863         for (i = 0; i < MAX_TX_FIFOS; i++) {
6864                 config->tx_cfg[i].fifo_len = tx_fifo_len[i];
6865                 config->tx_cfg[i].fifo_priority = i;
6866         }
6867
6868         /* mapping the QoS priority to the configured fifos */
6869         for (i = 0; i < MAX_TX_FIFOS; i++)
6870                 config->fifo_mapping[i] = fifo_map[config->tx_fifo_num][i];
6871
6872         config->tx_intr_type = TXD_INT_TYPE_UTILZ;
6873         for (i = 0; i < config->tx_fifo_num; i++) {
6874                 config->tx_cfg[i].f_no_snoop =
6875                     (NO_SNOOP_TXD | NO_SNOOP_TXD_BUFFER);
6876                 if (config->tx_cfg[i].fifo_len < 65) {
6877                         config->tx_intr_type = TXD_INT_TYPE_PER_LIST;
6878                         break;
6879                 }
6880         }
6881         /* + 2 because one Txd for skb->data and one Txd for UFO */
6882         config->max_txds = MAX_SKB_FRAGS + 2;
6883
6884         /* Rx side parameters. */
6885         config->rx_ring_num = rx_ring_num;
6886         for (i = 0; i < MAX_RX_RINGS; i++) {
6887                 config->rx_cfg[i].num_rxd = rx_ring_sz[i] *
6888                     (rxd_count[sp->rxd_mode] + 1);
6889                 config->rx_cfg[i].ring_priority = i;
6890         }
6891
6892         for (i = 0; i < rx_ring_num; i++) {
6893                 config->rx_cfg[i].ring_org = RING_ORG_BUFF1;
6894                 config->rx_cfg[i].f_no_snoop =
6895                     (NO_SNOOP_RXD | NO_SNOOP_RXD_BUFFER);
6896         }
6897
6898         /*  Setting Mac Control parameters */
6899         mac_control->rmac_pause_time = rmac_pause_time;
6900         mac_control->mc_pause_threshold_q0q3 = mc_pause_threshold_q0q3;
6901         mac_control->mc_pause_threshold_q4q7 = mc_pause_threshold_q4q7;
6902
6903
6904         /* Initialize Ring buffer parameters. */
6905         for (i = 0; i < config->rx_ring_num; i++)
6906                 atomic_set(&sp->rx_bufs_left[i], 0);
6907
6908         /* Initialize the number of ISRs currently running */
6909         atomic_set(&sp->isr_cnt, 0);
6910
6911         /*  initialize the shared memory used by the NIC and the host */
6912         if (init_shared_mem(sp)) {
6913                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n",
6914                           dev->name);
6915                 ret = -ENOMEM;
6916                 goto mem_alloc_failed;
6917         }
6918
6919         sp->bar0 = ioremap(pci_resource_start(pdev, 0),
6920                                      pci_resource_len(pdev, 0));
6921         if (!sp->bar0) {
6922                 DBG_PRINT(ERR_DBG, "%s: S2IO: cannot remap io mem1\n",
6923                           dev->name);
6924                 ret = -ENOMEM;
6925                 goto bar0_remap_failed;
6926         }
6927
6928         sp->bar1 = ioremap(pci_resource_start(pdev, 2),
6929                                      pci_resource_len(pdev, 2));
6930         if (!sp->bar1) {
6931                 DBG_PRINT(ERR_DBG, "%s: S2IO: cannot remap io mem2\n",
6932                           dev->name);
6933                 ret = -ENOMEM;
6934                 goto bar1_remap_failed;
6935         }
6936
6937         dev->irq = pdev->irq;
6938         dev->base_addr = (unsigned long) sp->bar0;
6939
6940         /* Initializing the BAR1 address as the start of the FIFO pointer. */
6941         for (j = 0; j < MAX_TX_FIFOS; j++) {
6942                 mac_control->tx_FIFO_start[j] = (TxFIFO_element_t __iomem *)
6943                     (sp->bar1 + (j * 0x00020000));
6944         }
6945
6946         /*  Driver entry points */
6947         dev->open = &s2io_open;
6948         dev->stop = &s2io_close;
6949         dev->hard_start_xmit = &s2io_xmit;
6950         dev->get_stats = &s2io_get_stats;
6951         dev->set_multicast_list = &s2io_set_multicast;
6952         dev->do_ioctl = &s2io_ioctl;
6953         dev->change_mtu = &s2io_change_mtu;
6954         SET_ETHTOOL_OPS(dev, &netdev_ethtool_ops);
6955         dev->features |= NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX;
6956         dev->vlan_rx_register = s2io_vlan_rx_register;
6957         dev->vlan_rx_kill_vid = (void *)s2io_vlan_rx_kill_vid;
6958
6959         /*
6960          * will use eth_mac_addr() for  dev->set_mac_address
6961          * mac address will be set every time dev->open() is called
6962          */
6963 #if defined(CONFIG_S2IO_NAPI)
6964         dev->poll = s2io_poll;
6965         dev->weight = 32;
6966 #endif
6967
6968 #ifdef CONFIG_NET_POLL_CONTROLLER
6969         dev->poll_controller = s2io_netpoll;
6970 #endif
6971
6972         dev->features |= NETIF_F_SG | NETIF_F_IP_CSUM;
6973         if (sp->high_dma_flag == TRUE)
6974                 dev->features |= NETIF_F_HIGHDMA;
6975 #ifdef NETIF_F_TSO
6976         dev->features |= NETIF_F_TSO;
6977 #endif
6978 #ifdef NETIF_F_TSO6
6979         dev->features |= NETIF_F_TSO6;
6980 #endif
6981         if (sp->device_type & XFRAME_II_DEVICE) {
6982                 dev->features |= NETIF_F_UFO;
6983                 dev->features |= NETIF_F_HW_CSUM;
6984         }
6985
6986         dev->tx_timeout = &s2io_tx_watchdog;
6987         dev->watchdog_timeo = WATCH_DOG_TIMEOUT;
6988         INIT_WORK(&sp->rst_timer_task,
6989                   (void (*)(void *)) s2io_restart_nic, dev);
6990         INIT_WORK(&sp->set_link_task,
6991                   (void (*)(void *)) s2io_set_link, sp);
6992
6993         pci_save_state(sp->pdev);
6994
6995         /* Setting swapper control on the NIC, for proper reset operation */
6996         if (s2io_set_swapper(sp)) {
6997                 DBG_PRINT(ERR_DBG, "%s:swapper settings are wrong\n",
6998                           dev->name);
6999                 ret = -EAGAIN;
7000                 goto set_swap_failed;
7001         }
7002
7003         /* Verify if the Herc works on the slot its placed into */
7004         if (sp->device_type & XFRAME_II_DEVICE) {
7005                 mode = s2io_verify_pci_mode(sp);
7006                 if (mode < 0) {
7007                         DBG_PRINT(ERR_DBG, "%s: ", __FUNCTION__);
7008                         DBG_PRINT(ERR_DBG, " Unsupported PCI bus mode\n");
7009                         ret = -EBADSLT;
7010                         goto set_swap_failed;
7011                 }
7012         }
7013
7014         /* Not needed for Herc */
7015         if (sp->device_type & XFRAME_I_DEVICE) {
7016                 /*
7017                  * Fix for all "FFs" MAC address problems observed on
7018                  * Alpha platforms
7019                  */
7020                 fix_mac_address(sp);
7021                 s2io_reset(sp);
7022         }
7023
7024         /*
7025          * MAC address initialization.
7026          * For now only one mac address will be read and used.
7027          */
7028         bar0 = sp->bar0;
7029         val64 = RMAC_ADDR_CMD_MEM_RD | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
7030             RMAC_ADDR_CMD_MEM_OFFSET(0 + MAC_MAC_ADDR_START_OFFSET);
7031         writeq(val64, &bar0->rmac_addr_cmd_mem);
7032         wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
7033                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING);
7034         tmp64 = readq(&bar0->rmac_addr_data0_mem);
7035         mac_down = (u32) tmp64;
7036         mac_up = (u32) (tmp64 >> 32);
7037
7038         memset(sp->def_mac_addr[0].mac_addr, 0, sizeof(ETH_ALEN));
7039
7040         sp->def_mac_addr[0].mac_addr[3] = (u8) (mac_up);
7041         sp->def_mac_addr[0].mac_addr[2] = (u8) (mac_up >> 8);
7042         sp->def_mac_addr[0].mac_addr[1] = (u8) (mac_up >> 16);
7043         sp->def_mac_addr[0].mac_addr[0] = (u8) (mac_up >> 24);
7044         sp->def_mac_addr[0].mac_addr[5] = (u8) (mac_down >> 16);
7045         sp->def_mac_addr[0].mac_addr[4] = (u8) (mac_down >> 24);
7046
7047         /*  Set the factory defined MAC address initially   */
7048         dev->addr_len = ETH_ALEN;
7049         memcpy(dev->dev_addr, sp->def_mac_addr, ETH_ALEN);
7050
7051         /* reset Nic and bring it to known state */
7052         s2io_reset(sp);
7053
7054         /*
7055          * Initialize the tasklet status and link state flags
7056          * and the card state parameter
7057          */
7058         atomic_set(&(sp->card_state), 0);
7059         sp->tasklet_status = 0;
7060         sp->link_state = 0;
7061
7062         /* Initialize spinlocks */
7063         spin_lock_init(&sp->tx_lock);
7064 #ifndef CONFIG_S2IO_NAPI
7065         spin_lock_init(&sp->put_lock);
7066 #endif
7067         spin_lock_init(&sp->rx_lock);
7068
7069         /*
7070          * SXE-002: Configure link and activity LED to init state
7071          * on driver load.
7072          */
7073         subid = sp->pdev->subsystem_device;
7074         if ((subid & 0xFF) >= 0x07) {
7075                 val64 = readq(&bar0->gpio_control);
7076                 val64 |= 0x0000800000000000ULL;
7077                 writeq(val64, &bar0->gpio_control);
7078                 val64 = 0x0411040400000000ULL;
7079                 writeq(val64, (void __iomem *) bar0 + 0x2700);
7080                 val64 = readq(&bar0->gpio_control);
7081         }
7082
7083         sp->rx_csum = 1;        /* Rx chksum verify enabled by default */
7084
7085         if (register_netdev(dev)) {
7086                 DBG_PRINT(ERR_DBG, "Device registration failed\n");
7087                 ret = -ENODEV;
7088                 goto register_failed;
7089         }
7090         s2io_vpd_read(sp);
7091         DBG_PRINT(ERR_DBG, "Copyright(c) 2002-2005 Neterion Inc.\n");
7092         DBG_PRINT(ERR_DBG, "%s: Neterion %s (rev %d)\n",dev->name,
7093                   sp->product_name, get_xena_rev_id(sp->pdev));
7094         DBG_PRINT(ERR_DBG, "%s: Driver version %s\n", dev->name,
7095                   s2io_driver_version);
7096         DBG_PRINT(ERR_DBG, "%s: MAC ADDR: "
7097                           "%02x:%02x:%02x:%02x:%02x:%02x\n", dev->name,
7098                           sp->def_mac_addr[0].mac_addr[0],
7099                           sp->def_mac_addr[0].mac_addr[1],
7100                           sp->def_mac_addr[0].mac_addr[2],
7101                           sp->def_mac_addr[0].mac_addr[3],
7102                           sp->def_mac_addr[0].mac_addr[4],
7103                           sp->def_mac_addr[0].mac_addr[5]);
7104         if (sp->device_type & XFRAME_II_DEVICE) {
7105                 mode = s2io_print_pci_mode(sp);
7106                 if (mode < 0) {
7107                         DBG_PRINT(ERR_DBG, " Unsupported PCI bus mode\n");
7108                         ret = -EBADSLT;
7109                         unregister_netdev(dev);
7110                         goto set_swap_failed;
7111                 }
7112         }
7113         switch(sp->rxd_mode) {
7114                 case RXD_MODE_1:
7115                     DBG_PRINT(ERR_DBG, "%s: 1-Buffer receive mode enabled\n",
7116                                                 dev->name);
7117                     break;
7118                 case RXD_MODE_3B:
7119                     DBG_PRINT(ERR_DBG, "%s: 2-Buffer receive mode enabled\n",
7120                                                 dev->name);
7121                     break;
7122                 case RXD_MODE_3A:
7123                     DBG_PRINT(ERR_DBG, "%s: 3-Buffer receive mode enabled\n",
7124                                                 dev->name);
7125                     break;
7126         }
7127 #ifdef CONFIG_S2IO_NAPI
7128         DBG_PRINT(ERR_DBG, "%s: NAPI enabled\n", dev->name);
7129 #endif
7130         switch(sp->intr_type) {
7131                 case INTA:
7132                     DBG_PRINT(ERR_DBG, "%s: Interrupt type INTA\n", dev->name);
7133                     break;
7134                 case MSI:
7135                     DBG_PRINT(ERR_DBG, "%s: Interrupt type MSI\n", dev->name);
7136                     break;
7137                 case MSI_X:
7138                     DBG_PRINT(ERR_DBG, "%s: Interrupt type MSI-X\n", dev->name);
7139                     break;
7140         }
7141         if (sp->lro)
7142                 DBG_PRINT(ERR_DBG, "%s: Large receive offload enabled\n",
7143                           dev->name);
7144
7145         /* Initialize device name */
7146         sprintf(sp->name, "%s Neterion %s", dev->name, sp->product_name);
7147
7148         /* Initialize bimodal Interrupts */
7149         sp->config.bimodal = bimodal;
7150         if (!(sp->device_type & XFRAME_II_DEVICE) && bimodal) {
7151                 sp->config.bimodal = 0;
7152                 DBG_PRINT(ERR_DBG,"%s:Bimodal intr not supported by Xframe I\n",
7153                         dev->name);
7154         }
7155
7156         /*
7157          * Make Link state as off at this point, when the Link change
7158          * interrupt comes the state will be automatically changed to
7159          * the right state.
7160          */
7161         netif_carrier_off(dev);
7162
7163         return 0;
7164
7165       register_failed:
7166       set_swap_failed:
7167         iounmap(sp->bar1);
7168       bar1_remap_failed:
7169         iounmap(sp->bar0);
7170       bar0_remap_failed:
7171       mem_alloc_failed:
7172         free_shared_mem(sp);
7173         pci_disable_device(pdev);
7174         if (dev_intr_type != MSI_X)
7175                 pci_release_regions(pdev);
7176         else {
7177                 release_mem_region(pci_resource_start(pdev, 0),
7178                         pci_resource_len(pdev, 0));
7179                 release_mem_region(pci_resource_start(pdev, 2),
7180                         pci_resource_len(pdev, 2));
7181         }
7182         pci_set_drvdata(pdev, NULL);
7183         free_netdev(dev);
7184
7185         return ret;
7186 }
7187
7188 /**
7189  * s2io_rem_nic - Free the PCI device
7190  * @pdev: structure containing the PCI related information of the device.
7191  * Description: This function is called by the Pci subsystem to release a
7192  * PCI device and free up all resource held up by the device. This could
7193  * be in response to a Hot plug event or when the driver is to be removed
7194  * from memory.
7195  */
7196
7197 static void __devexit s2io_rem_nic(struct pci_dev *pdev)
7198 {
7199         struct net_device *dev =
7200             (struct net_device *) pci_get_drvdata(pdev);
7201         nic_t *sp;
7202
7203         if (dev == NULL) {
7204                 DBG_PRINT(ERR_DBG, "Driver Data is NULL!!\n");
7205                 return;
7206         }
7207
7208         sp = dev->priv;
7209         unregister_netdev(dev);
7210
7211         free_shared_mem(sp);
7212         iounmap(sp->bar0);
7213         iounmap(sp->bar1);
7214         pci_disable_device(pdev);
7215         if (sp->intr_type != MSI_X)
7216                 pci_release_regions(pdev);
7217         else {
7218                 release_mem_region(pci_resource_start(pdev, 0),
7219                         pci_resource_len(pdev, 0));
7220                 release_mem_region(pci_resource_start(pdev, 2),
7221                         pci_resource_len(pdev, 2));
7222         }
7223         pci_set_drvdata(pdev, NULL);
7224         free_netdev(dev);
7225 }
7226
7227 /**
7228  * s2io_starter - Entry point for the driver
7229  * Description: This function is the entry point for the driver. It verifies
7230  * the module loadable parameters and initializes PCI configuration space.
7231  */
7232
7233 int __init s2io_starter(void)
7234 {
7235         return pci_module_init(&s2io_driver);
7236 }
7237
7238 /**
7239  * s2io_closer - Cleanup routine for the driver
7240  * Description: This function is the cleanup routine for the driver. It unregist * ers the driver.
7241  */
7242
7243 static void s2io_closer(void)
7244 {
7245         pci_unregister_driver(&s2io_driver);
7246         DBG_PRINT(INIT_DBG, "cleanup done\n");
7247 }
7248
7249 module_init(s2io_starter);
7250 module_exit(s2io_closer);
7251
7252 static int check_L2_lro_capable(u8 *buffer, struct iphdr **ip, 
7253                 struct tcphdr **tcp, RxD_t *rxdp)
7254 {
7255         int ip_off;
7256         u8 l2_type = (u8)((rxdp->Control_1 >> 37) & 0x7), ip_len;
7257
7258         if (!(rxdp->Control_1 & RXD_FRAME_PROTO_TCP)) {
7259                 DBG_PRINT(INIT_DBG,"%s: Non-TCP frames not supported for LRO\n",
7260                           __FUNCTION__);
7261                 return -1;
7262         }
7263
7264         /* TODO:
7265          * By default the VLAN field in the MAC is stripped by the card, if this
7266          * feature is turned off in rx_pa_cfg register, then the ip_off field
7267          * has to be shifted by a further 2 bytes
7268          */
7269         switch (l2_type) {
7270                 case 0: /* DIX type */
7271                 case 4: /* DIX type with VLAN */
7272                         ip_off = HEADER_ETHERNET_II_802_3_SIZE;
7273                         break;
7274                 /* LLC, SNAP etc are considered non-mergeable */
7275                 default:
7276                         return -1;
7277         }
7278
7279         *ip = (struct iphdr *)((u8 *)buffer + ip_off);
7280         ip_len = (u8)((*ip)->ihl);
7281         ip_len <<= 2;
7282         *tcp = (struct tcphdr *)((unsigned long)*ip + ip_len);
7283
7284         return 0;
7285 }
7286
7287 static int check_for_socket_match(lro_t *lro, struct iphdr *ip,
7288                                   struct tcphdr *tcp)
7289 {
7290         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7291         if ((lro->iph->saddr != ip->saddr) || (lro->iph->daddr != ip->daddr) ||
7292            (lro->tcph->source != tcp->source) || (lro->tcph->dest != tcp->dest))
7293                 return -1;
7294         return 0;
7295 }
7296
7297 static inline int get_l4_pyld_length(struct iphdr *ip, struct tcphdr *tcp)
7298 {
7299         return(ntohs(ip->tot_len) - (ip->ihl << 2) - (tcp->doff << 2));
7300 }
7301
7302 static void initiate_new_session(lro_t *lro, u8 *l2h,
7303                      struct iphdr *ip, struct tcphdr *tcp, u32 tcp_pyld_len)
7304 {
7305         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7306         lro->l2h = l2h;
7307         lro->iph = ip;
7308         lro->tcph = tcp;
7309         lro->tcp_next_seq = tcp_pyld_len + ntohl(tcp->seq);
7310         lro->tcp_ack = ntohl(tcp->ack_seq);
7311         lro->sg_num = 1;
7312         lro->total_len = ntohs(ip->tot_len);
7313         lro->frags_len = 0;
7314         /* 
7315          * check if we saw TCP timestamp. Other consistency checks have
7316          * already been done.
7317          */
7318         if (tcp->doff == 8) {
7319                 u32 *ptr;
7320                 ptr = (u32 *)(tcp+1);
7321                 lro->saw_ts = 1;
7322                 lro->cur_tsval = *(ptr+1);
7323                 lro->cur_tsecr = *(ptr+2);
7324         }
7325         lro->in_use = 1;
7326 }
7327
7328 static void update_L3L4_header(nic_t *sp, lro_t *lro)
7329 {
7330         struct iphdr *ip = lro->iph;
7331         struct tcphdr *tcp = lro->tcph;
7332         u16 nchk;
7333         StatInfo_t *statinfo = sp->mac_control.stats_info;
7334         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7335
7336         /* Update L3 header */
7337         ip->tot_len = htons(lro->total_len);
7338         ip->check = 0;
7339         nchk = ip_fast_csum((u8 *)lro->iph, ip->ihl);
7340         ip->check = nchk;
7341
7342         /* Update L4 header */
7343         tcp->ack_seq = lro->tcp_ack;
7344         tcp->window = lro->window;
7345
7346         /* Update tsecr field if this session has timestamps enabled */
7347         if (lro->saw_ts) {
7348                 u32 *ptr = (u32 *)(tcp + 1);
7349                 *(ptr+2) = lro->cur_tsecr;
7350         }
7351
7352         /* Update counters required for calculation of
7353          * average no. of packets aggregated.
7354          */
7355         statinfo->sw_stat.sum_avg_pkts_aggregated += lro->sg_num;
7356         statinfo->sw_stat.num_aggregations++;
7357 }
7358
7359 static void aggregate_new_rx(lro_t *lro, struct iphdr *ip,
7360                 struct tcphdr *tcp, u32 l4_pyld)
7361 {
7362         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7363         lro->total_len += l4_pyld;
7364         lro->frags_len += l4_pyld;
7365         lro->tcp_next_seq += l4_pyld;
7366         lro->sg_num++;
7367
7368         /* Update ack seq no. and window ad(from this pkt) in LRO object */
7369         lro->tcp_ack = tcp->ack_seq;
7370         lro->window = tcp->window;
7371         
7372         if (lro->saw_ts) {
7373                 u32 *ptr;
7374                 /* Update tsecr and tsval from this packet */
7375                 ptr = (u32 *) (tcp + 1);
7376                 lro->cur_tsval = *(ptr + 1); 
7377                 lro->cur_tsecr = *(ptr + 2);
7378         }
7379 }
7380
7381 static int verify_l3_l4_lro_capable(lro_t *l_lro, struct iphdr *ip,
7382                                     struct tcphdr *tcp, u32 tcp_pyld_len)
7383 {
7384         u8 *ptr;
7385
7386         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7387
7388         if (!tcp_pyld_len) {
7389                 /* Runt frame or a pure ack */
7390                 return -1;
7391         }
7392
7393         if (ip->ihl != 5) /* IP has options */
7394                 return -1;
7395
7396         /* If we see CE codepoint in IP header, packet is not mergeable */
7397         if (INET_ECN_is_ce(ipv4_get_dsfield(ip)))
7398                 return -1;
7399
7400         /* If we see ECE or CWR flags in TCP header, packet is not mergeable */
7401         if (tcp->urg || tcp->psh || tcp->rst || tcp->syn || tcp->fin ||
7402                                     tcp->ece || tcp->cwr || !tcp->ack) {
7403                 /*
7404                  * Currently recognize only the ack control word and
7405                  * any other control field being set would result in
7406                  * flushing the LRO session
7407                  */
7408                 return -1;
7409         }
7410
7411         /* 
7412          * Allow only one TCP timestamp option. Don't aggregate if
7413          * any other options are detected.
7414          */
7415         if (tcp->doff != 5 && tcp->doff != 8)
7416                 return -1;
7417
7418         if (tcp->doff == 8) {
7419                 ptr = (u8 *)(tcp + 1);  
7420                 while (*ptr == TCPOPT_NOP)
7421                         ptr++;
7422                 if (*ptr != TCPOPT_TIMESTAMP || *(ptr+1) != TCPOLEN_TIMESTAMP)
7423                         return -1;
7424
7425                 /* Ensure timestamp value increases monotonically */
7426                 if (l_lro)
7427                         if (l_lro->cur_tsval > *((u32 *)(ptr+2)))
7428                                 return -1;
7429
7430                 /* timestamp echo reply should be non-zero */
7431                 if (*((u32 *)(ptr+6)) == 0) 
7432                         return -1;
7433         }
7434
7435         return 0;
7436 }
7437
7438 static int
7439 s2io_club_tcp_session(u8 *buffer, u8 **tcp, u32 *tcp_len, lro_t **lro,
7440                       RxD_t *rxdp, nic_t *sp)
7441 {
7442         struct iphdr *ip;
7443         struct tcphdr *tcph;
7444         int ret = 0, i;
7445
7446         if (!(ret = check_L2_lro_capable(buffer, &ip, (struct tcphdr **)tcp,
7447                                          rxdp))) {
7448                 DBG_PRINT(INFO_DBG,"IP Saddr: %x Daddr: %x\n",
7449                           ip->saddr, ip->daddr);
7450         } else {
7451                 return ret;
7452         }
7453
7454         tcph = (struct tcphdr *)*tcp;
7455         *tcp_len = get_l4_pyld_length(ip, tcph);
7456         for (i=0; i<MAX_LRO_SESSIONS; i++) {
7457                 lro_t *l_lro = &sp->lro0_n[i];
7458                 if (l_lro->in_use) {
7459                         if (check_for_socket_match(l_lro, ip, tcph))
7460                                 continue;
7461                         /* Sock pair matched */
7462                         *lro = l_lro;
7463
7464                         if ((*lro)->tcp_next_seq != ntohl(tcph->seq)) {
7465                                 DBG_PRINT(INFO_DBG, "%s:Out of order. expected "
7466                                           "0x%x, actual 0x%x\n", __FUNCTION__,
7467                                           (*lro)->tcp_next_seq,
7468                                           ntohl(tcph->seq));
7469
7470                                 sp->mac_control.stats_info->
7471                                    sw_stat.outof_sequence_pkts++;
7472                                 ret = 2;
7473                                 break;
7474                         }
7475
7476                         if (!verify_l3_l4_lro_capable(l_lro, ip, tcph,*tcp_len))
7477                                 ret = 1; /* Aggregate */
7478                         else
7479                                 ret = 2; /* Flush both */
7480                         break;
7481                 }
7482         }
7483
7484         if (ret == 0) {
7485                 /* Before searching for available LRO objects,
7486                  * check if the pkt is L3/L4 aggregatable. If not
7487                  * don't create new LRO session. Just send this
7488                  * packet up.
7489                  */
7490                 if (verify_l3_l4_lro_capable(NULL, ip, tcph, *tcp_len)) {
7491                         return 5;
7492                 }
7493
7494                 for (i=0; i<MAX_LRO_SESSIONS; i++) {
7495                         lro_t *l_lro = &sp->lro0_n[i];
7496                         if (!(l_lro->in_use)) {
7497                                 *lro = l_lro;
7498                                 ret = 3; /* Begin anew */
7499                                 break;
7500                         }
7501                 }
7502         }
7503
7504         if (ret == 0) { /* sessions exceeded */
7505                 DBG_PRINT(INFO_DBG,"%s:All LRO sessions already in use\n",
7506                           __FUNCTION__);
7507                 *lro = NULL;
7508                 return ret;
7509         }
7510
7511         switch (ret) {
7512                 case 3:
7513                         initiate_new_session(*lro, buffer, ip, tcph, *tcp_len);
7514                         break;
7515                 case 2:
7516                         update_L3L4_header(sp, *lro);
7517                         break;
7518                 case 1:
7519                         aggregate_new_rx(*lro, ip, tcph, *tcp_len);
7520                         if ((*lro)->sg_num == sp->lro_max_aggr_per_sess) {
7521                                 update_L3L4_header(sp, *lro);
7522                                 ret = 4; /* Flush the LRO */
7523                         }
7524                         break;
7525                 default:
7526                         DBG_PRINT(ERR_DBG,"%s:Dont know, can't say!!\n",
7527                                 __FUNCTION__);
7528                         break;
7529         }
7530
7531         return ret;
7532 }
7533
7534 static void clear_lro_session(lro_t *lro)
7535 {
7536         static u16 lro_struct_size = sizeof(lro_t);
7537
7538         memset(lro, 0, lro_struct_size);
7539 }
7540
7541 static void queue_rx_frame(struct sk_buff *skb)
7542 {
7543         struct net_device *dev = skb->dev;
7544
7545         skb->protocol = eth_type_trans(skb, dev);
7546 #ifdef CONFIG_S2IO_NAPI
7547         netif_receive_skb(skb);
7548 #else
7549         netif_rx(skb);
7550 #endif
7551 }
7552
7553 static void lro_append_pkt(nic_t *sp, lro_t *lro, struct sk_buff *skb,
7554                            u32 tcp_len)
7555 {
7556         struct sk_buff *first = lro->parent;
7557
7558         first->len += tcp_len;
7559         first->data_len = lro->frags_len;
7560         skb_pull(skb, (skb->len - tcp_len));
7561         if (skb_shinfo(first)->frag_list)
7562                 lro->last_frag->next = skb;
7563         else
7564                 skb_shinfo(first)->frag_list = skb;
7565         lro->last_frag = skb;
7566         sp->mac_control.stats_info->sw_stat.clubbed_frms_cnt++;
7567         return;
7568 }