Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
[pandora-kernel.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 "2.0.14.2"
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 Loadable parameters. */
374 static unsigned int tx_fifo_num = 1;
375 static unsigned int tx_fifo_len[MAX_TX_FIFOS] =
376     {DEFAULT_FIFO_0_LEN, [1 ...(MAX_TX_FIFOS - 1)] = DEFAULT_FIFO_1_7_LEN};
377 static unsigned int rx_ring_num = 1;
378 static unsigned int rx_ring_sz[MAX_RX_RINGS] =
379     {[0 ...(MAX_RX_RINGS - 1)] = SMALL_BLK_CNT};
380 static unsigned int rts_frm_len[MAX_RX_RINGS] =
381     {[0 ...(MAX_RX_RINGS - 1)] = 0 };
382 static unsigned int rx_ring_mode = 1;
383 static unsigned int use_continuous_tx_intrs = 1;
384 static unsigned int rmac_pause_time = 0x100;
385 static unsigned int mc_pause_threshold_q0q3 = 187;
386 static unsigned int mc_pause_threshold_q4q7 = 187;
387 static unsigned int shared_splits;
388 static unsigned int tmac_util_period = 5;
389 static unsigned int rmac_util_period = 5;
390 static unsigned int bimodal = 0;
391 static unsigned int l3l4hdr_size = 128;
392 #ifndef CONFIG_S2IO_NAPI
393 static unsigned int indicate_max_pkts;
394 #endif
395 /* Frequency of Rx desc syncs expressed as power of 2 */
396 static unsigned int rxsync_frequency = 3;
397 /* Interrupt type. Values can be 0(INTA), 1(MSI), 2(MSI_X) */
398 static unsigned int intr_type = 0;
399 /* Large receive offload feature */
400 static unsigned int lro = 0;
401 /* Max pkts to be aggregated by LRO at one time. If not specified,
402  * aggregation happens until we hit max IP pkt size(64K)
403  */
404 static unsigned int lro_max_pkts = 0xFFFF;
405
406 /*
407  * S2IO device table.
408  * This table lists all the devices that this driver supports.
409  */
410 static struct pci_device_id s2io_tbl[] __devinitdata = {
411         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_WIN,
412          PCI_ANY_ID, PCI_ANY_ID},
413         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_UNI,
414          PCI_ANY_ID, PCI_ANY_ID},
415         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_WIN,
416          PCI_ANY_ID, PCI_ANY_ID},
417         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_UNI,
418          PCI_ANY_ID, PCI_ANY_ID},
419         {0,}
420 };
421
422 MODULE_DEVICE_TABLE(pci, s2io_tbl);
423
424 static struct pci_driver s2io_driver = {
425       .name = "S2IO",
426       .id_table = s2io_tbl,
427       .probe = s2io_init_nic,
428       .remove = __devexit_p(s2io_rem_nic),
429 };
430
431 /* A simplifier macro used both by init and free shared_mem Fns(). */
432 #define TXD_MEM_PAGE_CNT(len, per_each) ((len+per_each - 1) / per_each)
433
434 /**
435  * init_shared_mem - Allocation and Initialization of Memory
436  * @nic: Device private variable.
437  * Description: The function allocates all the memory areas shared
438  * between the NIC and the driver. This includes Tx descriptors,
439  * Rx descriptors and the statistics block.
440  */
441
442 static int init_shared_mem(struct s2io_nic *nic)
443 {
444         u32 size;
445         void *tmp_v_addr, *tmp_v_addr_next;
446         dma_addr_t tmp_p_addr, tmp_p_addr_next;
447         RxD_block_t *pre_rxd_blk = NULL;
448         int i, j, blk_cnt, rx_sz, tx_sz;
449         int lst_size, lst_per_page;
450         struct net_device *dev = nic->dev;
451         unsigned long tmp;
452         buffAdd_t *ba;
453
454         mac_info_t *mac_control;
455         struct config_param *config;
456
457         mac_control = &nic->mac_control;
458         config = &nic->config;
459
460
461         /* Allocation and initialization of TXDLs in FIOFs */
462         size = 0;
463         for (i = 0; i < config->tx_fifo_num; i++) {
464                 size += config->tx_cfg[i].fifo_len;
465         }
466         if (size > MAX_AVAILABLE_TXDS) {
467                 DBG_PRINT(ERR_DBG, "%s: Requested TxDs too high, ",
468                           __FUNCTION__);
469                 DBG_PRINT(ERR_DBG, "Requested: %d, max supported: 8192\n", size);
470                 return FAILURE;
471         }
472
473         lst_size = (sizeof(TxD_t) * config->max_txds);
474         tx_sz = lst_size * size;
475         lst_per_page = PAGE_SIZE / lst_size;
476
477         for (i = 0; i < config->tx_fifo_num; i++) {
478                 int fifo_len = config->tx_cfg[i].fifo_len;
479                 int list_holder_size = fifo_len * sizeof(list_info_hold_t);
480                 mac_control->fifos[i].list_info = kmalloc(list_holder_size,
481                                                           GFP_KERNEL);
482                 if (!mac_control->fifos[i].list_info) {
483                         DBG_PRINT(ERR_DBG,
484                                   "Malloc failed for list_info\n");
485                         return -ENOMEM;
486                 }
487                 memset(mac_control->fifos[i].list_info, 0, list_holder_size);
488         }
489         for (i = 0; i < config->tx_fifo_num; i++) {
490                 int page_num = TXD_MEM_PAGE_CNT(config->tx_cfg[i].fifo_len,
491                                                 lst_per_page);
492                 mac_control->fifos[i].tx_curr_put_info.offset = 0;
493                 mac_control->fifos[i].tx_curr_put_info.fifo_len =
494                     config->tx_cfg[i].fifo_len - 1;
495                 mac_control->fifos[i].tx_curr_get_info.offset = 0;
496                 mac_control->fifos[i].tx_curr_get_info.fifo_len =
497                     config->tx_cfg[i].fifo_len - 1;
498                 mac_control->fifos[i].fifo_no = i;
499                 mac_control->fifos[i].nic = nic;
500                 mac_control->fifos[i].max_txds = MAX_SKB_FRAGS + 2;
501
502                 for (j = 0; j < page_num; j++) {
503                         int k = 0;
504                         dma_addr_t tmp_p;
505                         void *tmp_v;
506                         tmp_v = pci_alloc_consistent(nic->pdev,
507                                                      PAGE_SIZE, &tmp_p);
508                         if (!tmp_v) {
509                                 DBG_PRINT(ERR_DBG,
510                                           "pci_alloc_consistent ");
511                                 DBG_PRINT(ERR_DBG, "failed for TxDL\n");
512                                 return -ENOMEM;
513                         }
514                         /* If we got a zero DMA address(can happen on
515                          * certain platforms like PPC), reallocate.
516                          * Store virtual address of page we don't want,
517                          * to be freed later.
518                          */
519                         if (!tmp_p) {
520                                 mac_control->zerodma_virt_addr = tmp_v;
521                                 DBG_PRINT(INIT_DBG, 
522                                 "%s: Zero DMA address for TxDL. ", dev->name);
523                                 DBG_PRINT(INIT_DBG, 
524                                 "Virtual address %p\n", tmp_v);
525                                 tmp_v = pci_alloc_consistent(nic->pdev,
526                                                      PAGE_SIZE, &tmp_p);
527                                 if (!tmp_v) {
528                                         DBG_PRINT(ERR_DBG,
529                                           "pci_alloc_consistent ");
530                                         DBG_PRINT(ERR_DBG, "failed for TxDL\n");
531                                         return -ENOMEM;
532                                 }
533                         }
534                         while (k < lst_per_page) {
535                                 int l = (j * lst_per_page) + k;
536                                 if (l == config->tx_cfg[i].fifo_len)
537                                         break;
538                                 mac_control->fifos[i].list_info[l].list_virt_addr =
539                                     tmp_v + (k * lst_size);
540                                 mac_control->fifos[i].list_info[l].list_phy_addr =
541                                     tmp_p + (k * lst_size);
542                                 k++;
543                         }
544                 }
545         }
546
547         nic->ufo_in_band_v = kmalloc((sizeof(u64) * size), GFP_KERNEL);
548         if (!nic->ufo_in_band_v)
549                 return -ENOMEM;
550
551         /* Allocation and initialization of RXDs in Rings */
552         size = 0;
553         for (i = 0; i < config->rx_ring_num; i++) {
554                 if (config->rx_cfg[i].num_rxd %
555                     (rxd_count[nic->rxd_mode] + 1)) {
556                         DBG_PRINT(ERR_DBG, "%s: RxD count of ", dev->name);
557                         DBG_PRINT(ERR_DBG, "Ring%d is not a multiple of ",
558                                   i);
559                         DBG_PRINT(ERR_DBG, "RxDs per Block");
560                         return FAILURE;
561                 }
562                 size += config->rx_cfg[i].num_rxd;
563                 mac_control->rings[i].block_count =
564                         config->rx_cfg[i].num_rxd /
565                         (rxd_count[nic->rxd_mode] + 1 );
566                 mac_control->rings[i].pkt_cnt = config->rx_cfg[i].num_rxd -
567                         mac_control->rings[i].block_count;
568         }
569         if (nic->rxd_mode == RXD_MODE_1)
570                 size = (size * (sizeof(RxD1_t)));
571         else
572                 size = (size * (sizeof(RxD3_t)));
573         rx_sz = size;
574
575         for (i = 0; i < config->rx_ring_num; i++) {
576                 mac_control->rings[i].rx_curr_get_info.block_index = 0;
577                 mac_control->rings[i].rx_curr_get_info.offset = 0;
578                 mac_control->rings[i].rx_curr_get_info.ring_len =
579                     config->rx_cfg[i].num_rxd - 1;
580                 mac_control->rings[i].rx_curr_put_info.block_index = 0;
581                 mac_control->rings[i].rx_curr_put_info.offset = 0;
582                 mac_control->rings[i].rx_curr_put_info.ring_len =
583                     config->rx_cfg[i].num_rxd - 1;
584                 mac_control->rings[i].nic = nic;
585                 mac_control->rings[i].ring_no = i;
586
587                 blk_cnt = config->rx_cfg[i].num_rxd /
588                                 (rxd_count[nic->rxd_mode] + 1);
589                 /*  Allocating all the Rx blocks */
590                 for (j = 0; j < blk_cnt; j++) {
591                         rx_block_info_t *rx_blocks;
592                         int l;
593
594                         rx_blocks = &mac_control->rings[i].rx_blocks[j];
595                         size = SIZE_OF_BLOCK; //size is always page size
596                         tmp_v_addr = pci_alloc_consistent(nic->pdev, size,
597                                                           &tmp_p_addr);
598                         if (tmp_v_addr == NULL) {
599                                 /*
600                                  * In case of failure, free_shared_mem()
601                                  * is called, which should free any
602                                  * memory that was alloced till the
603                                  * failure happened.
604                                  */
605                                 rx_blocks->block_virt_addr = tmp_v_addr;
606                                 return -ENOMEM;
607                         }
608                         memset(tmp_v_addr, 0, size);
609                         rx_blocks->block_virt_addr = tmp_v_addr;
610                         rx_blocks->block_dma_addr = tmp_p_addr;
611                         rx_blocks->rxds = kmalloc(sizeof(rxd_info_t)*
612                                                   rxd_count[nic->rxd_mode],
613                                                   GFP_KERNEL);
614                         for (l=0; l<rxd_count[nic->rxd_mode];l++) {
615                                 rx_blocks->rxds[l].virt_addr =
616                                         rx_blocks->block_virt_addr +
617                                         (rxd_size[nic->rxd_mode] * l);
618                                 rx_blocks->rxds[l].dma_addr =
619                                         rx_blocks->block_dma_addr +
620                                         (rxd_size[nic->rxd_mode] * l);
621                         }
622                 }
623                 /* Interlinking all Rx Blocks */
624                 for (j = 0; j < blk_cnt; j++) {
625                         tmp_v_addr =
626                                 mac_control->rings[i].rx_blocks[j].block_virt_addr;
627                         tmp_v_addr_next =
628                                 mac_control->rings[i].rx_blocks[(j + 1) %
629                                               blk_cnt].block_virt_addr;
630                         tmp_p_addr =
631                                 mac_control->rings[i].rx_blocks[j].block_dma_addr;
632                         tmp_p_addr_next =
633                                 mac_control->rings[i].rx_blocks[(j + 1) %
634                                               blk_cnt].block_dma_addr;
635
636                         pre_rxd_blk = (RxD_block_t *) tmp_v_addr;
637                         pre_rxd_blk->reserved_2_pNext_RxD_block =
638                             (unsigned long) tmp_v_addr_next;
639                         pre_rxd_blk->pNext_RxD_Blk_physical =
640                             (u64) tmp_p_addr_next;
641                 }
642         }
643         if (nic->rxd_mode >= RXD_MODE_3A) {
644                 /*
645                  * Allocation of Storages for buffer addresses in 2BUFF mode
646                  * and the buffers as well.
647                  */
648                 for (i = 0; i < config->rx_ring_num; i++) {
649                         blk_cnt = config->rx_cfg[i].num_rxd /
650                            (rxd_count[nic->rxd_mode]+ 1);
651                         mac_control->rings[i].ba =
652                                 kmalloc((sizeof(buffAdd_t *) * blk_cnt),
653                                      GFP_KERNEL);
654                         if (!mac_control->rings[i].ba)
655                                 return -ENOMEM;
656                         for (j = 0; j < blk_cnt; j++) {
657                                 int k = 0;
658                                 mac_control->rings[i].ba[j] =
659                                         kmalloc((sizeof(buffAdd_t) *
660                                                 (rxd_count[nic->rxd_mode] + 1)),
661                                                 GFP_KERNEL);
662                                 if (!mac_control->rings[i].ba[j])
663                                         return -ENOMEM;
664                                 while (k != rxd_count[nic->rxd_mode]) {
665                                         ba = &mac_control->rings[i].ba[j][k];
666
667                                         ba->ba_0_org = (void *) kmalloc
668                                             (BUF0_LEN + ALIGN_SIZE, GFP_KERNEL);
669                                         if (!ba->ba_0_org)
670                                                 return -ENOMEM;
671                                         tmp = (unsigned long)ba->ba_0_org;
672                                         tmp += ALIGN_SIZE;
673                                         tmp &= ~((unsigned long) ALIGN_SIZE);
674                                         ba->ba_0 = (void *) tmp;
675
676                                         ba->ba_1_org = (void *) kmalloc
677                                             (BUF1_LEN + ALIGN_SIZE, GFP_KERNEL);
678                                         if (!ba->ba_1_org)
679                                                 return -ENOMEM;
680                                         tmp = (unsigned long) ba->ba_1_org;
681                                         tmp += ALIGN_SIZE;
682                                         tmp &= ~((unsigned long) ALIGN_SIZE);
683                                         ba->ba_1 = (void *) tmp;
684                                         k++;
685                                 }
686                         }
687                 }
688         }
689
690         /* Allocation and initialization of Statistics block */
691         size = sizeof(StatInfo_t);
692         mac_control->stats_mem = pci_alloc_consistent
693             (nic->pdev, size, &mac_control->stats_mem_phy);
694
695         if (!mac_control->stats_mem) {
696                 /*
697                  * In case of failure, free_shared_mem() is called, which
698                  * should free any memory that was alloced till the
699                  * failure happened.
700                  */
701                 return -ENOMEM;
702         }
703         mac_control->stats_mem_sz = size;
704
705         tmp_v_addr = mac_control->stats_mem;
706         mac_control->stats_info = (StatInfo_t *) tmp_v_addr;
707         memset(tmp_v_addr, 0, size);
708         DBG_PRINT(INIT_DBG, "%s:Ring Mem PHY: 0x%llx\n", dev->name,
709                   (unsigned long long) tmp_p_addr);
710
711         return SUCCESS;
712 }
713
714 /**
715  * free_shared_mem - Free the allocated Memory
716  * @nic:  Device private variable.
717  * Description: This function is to free all memory locations allocated by
718  * the init_shared_mem() function and return it to the kernel.
719  */
720
721 static void free_shared_mem(struct s2io_nic *nic)
722 {
723         int i, j, blk_cnt, size;
724         void *tmp_v_addr;
725         dma_addr_t tmp_p_addr;
726         mac_info_t *mac_control;
727         struct config_param *config;
728         int lst_size, lst_per_page;
729         struct net_device *dev = nic->dev;
730
731         if (!nic)
732                 return;
733
734         mac_control = &nic->mac_control;
735         config = &nic->config;
736
737         lst_size = (sizeof(TxD_t) * config->max_txds);
738         lst_per_page = PAGE_SIZE / lst_size;
739
740         for (i = 0; i < config->tx_fifo_num; i++) {
741                 int page_num = TXD_MEM_PAGE_CNT(config->tx_cfg[i].fifo_len,
742                                                 lst_per_page);
743                 for (j = 0; j < page_num; j++) {
744                         int mem_blks = (j * lst_per_page);
745                         if (!mac_control->fifos[i].list_info)
746                                 return; 
747                         if (!mac_control->fifos[i].list_info[mem_blks].
748                                  list_virt_addr)
749                                 break;
750                         pci_free_consistent(nic->pdev, PAGE_SIZE,
751                                             mac_control->fifos[i].
752                                             list_info[mem_blks].
753                                             list_virt_addr,
754                                             mac_control->fifos[i].
755                                             list_info[mem_blks].
756                                             list_phy_addr);
757                 }
758                 /* If we got a zero DMA address during allocation,
759                  * free the page now
760                  */
761                 if (mac_control->zerodma_virt_addr) {
762                         pci_free_consistent(nic->pdev, PAGE_SIZE,
763                                             mac_control->zerodma_virt_addr,
764                                             (dma_addr_t)0);
765                         DBG_PRINT(INIT_DBG, 
766                                 "%s: Freeing TxDL with zero DMA addr. ",
767                                 dev->name);
768                         DBG_PRINT(INIT_DBG, "Virtual address %p\n",
769                                 mac_control->zerodma_virt_addr);
770                 }
771                 kfree(mac_control->fifos[i].list_info);
772         }
773
774         size = SIZE_OF_BLOCK;
775         for (i = 0; i < config->rx_ring_num; i++) {
776                 blk_cnt = mac_control->rings[i].block_count;
777                 for (j = 0; j < blk_cnt; j++) {
778                         tmp_v_addr = mac_control->rings[i].rx_blocks[j].
779                                 block_virt_addr;
780                         tmp_p_addr = mac_control->rings[i].rx_blocks[j].
781                                 block_dma_addr;
782                         if (tmp_v_addr == NULL)
783                                 break;
784                         pci_free_consistent(nic->pdev, size,
785                                             tmp_v_addr, tmp_p_addr);
786                         kfree(mac_control->rings[i].rx_blocks[j].rxds);
787                 }
788         }
789
790         if (nic->rxd_mode >= RXD_MODE_3A) {
791                 /* Freeing buffer storage addresses in 2BUFF mode. */
792                 for (i = 0; i < config->rx_ring_num; i++) {
793                         blk_cnt = config->rx_cfg[i].num_rxd /
794                             (rxd_count[nic->rxd_mode] + 1);
795                         for (j = 0; j < blk_cnt; j++) {
796                                 int k = 0;
797                                 if (!mac_control->rings[i].ba[j])
798                                         continue;
799                                 while (k != rxd_count[nic->rxd_mode]) {
800                                         buffAdd_t *ba =
801                                                 &mac_control->rings[i].ba[j][k];
802                                         kfree(ba->ba_0_org);
803                                         kfree(ba->ba_1_org);
804                                         k++;
805                                 }
806                                 kfree(mac_control->rings[i].ba[j]);
807                         }
808                         kfree(mac_control->rings[i].ba);
809                 }
810         }
811
812         if (mac_control->stats_mem) {
813                 pci_free_consistent(nic->pdev,
814                                     mac_control->stats_mem_sz,
815                                     mac_control->stats_mem,
816                                     mac_control->stats_mem_phy);
817         }
818         if (nic->ufo_in_band_v)
819                 kfree(nic->ufo_in_band_v);
820 }
821
822 /**
823  * s2io_verify_pci_mode -
824  */
825
826 static int s2io_verify_pci_mode(nic_t *nic)
827 {
828         XENA_dev_config_t __iomem *bar0 = nic->bar0;
829         register u64 val64 = 0;
830         int     mode;
831
832         val64 = readq(&bar0->pci_mode);
833         mode = (u8)GET_PCI_MODE(val64);
834
835         if ( val64 & PCI_MODE_UNKNOWN_MODE)
836                 return -1;      /* Unknown PCI mode */
837         return mode;
838 }
839
840 #define NEC_VENID   0x1033
841 #define NEC_DEVID   0x0125
842 static int s2io_on_nec_bridge(struct pci_dev *s2io_pdev)
843 {
844         struct pci_dev *tdev = NULL;
845         while ((tdev = pci_find_device(PCI_ANY_ID, PCI_ANY_ID, tdev)) != NULL) {
846                 if ((tdev->vendor == NEC_VENID) && (tdev->device == NEC_DEVID)){
847                         if (tdev->bus == s2io_pdev->bus->parent)
848                                 return 1;
849                 }
850         }
851         return 0;
852 }
853
854 static int bus_speed[8] = {33, 133, 133, 200, 266, 133, 200, 266};
855 /**
856  * s2io_print_pci_mode -
857  */
858 static int s2io_print_pci_mode(nic_t *nic)
859 {
860         XENA_dev_config_t __iomem *bar0 = nic->bar0;
861         register u64 val64 = 0;
862         int     mode;
863         struct config_param *config = &nic->config;
864
865         val64 = readq(&bar0->pci_mode);
866         mode = (u8)GET_PCI_MODE(val64);
867
868         if ( val64 & PCI_MODE_UNKNOWN_MODE)
869                 return -1;      /* Unknown PCI mode */
870
871         config->bus_speed = bus_speed[mode];
872
873         if (s2io_on_nec_bridge(nic->pdev)) {
874                 DBG_PRINT(ERR_DBG, "%s: Device is on PCI-E bus\n",
875                                                         nic->dev->name);
876                 return mode;
877         }
878
879         if (val64 & PCI_MODE_32_BITS) {
880                 DBG_PRINT(ERR_DBG, "%s: Device is on 32 bit ", nic->dev->name);
881         } else {
882                 DBG_PRINT(ERR_DBG, "%s: Device is on 64 bit ", nic->dev->name);
883         }
884
885         switch(mode) {
886                 case PCI_MODE_PCI_33:
887                         DBG_PRINT(ERR_DBG, "33MHz PCI bus\n");
888                         break;
889                 case PCI_MODE_PCI_66:
890                         DBG_PRINT(ERR_DBG, "66MHz PCI bus\n");
891                         break;
892                 case PCI_MODE_PCIX_M1_66:
893                         DBG_PRINT(ERR_DBG, "66MHz PCIX(M1) bus\n");
894                         break;
895                 case PCI_MODE_PCIX_M1_100:
896                         DBG_PRINT(ERR_DBG, "100MHz PCIX(M1) bus\n");
897                         break;
898                 case PCI_MODE_PCIX_M1_133:
899                         DBG_PRINT(ERR_DBG, "133MHz PCIX(M1) bus\n");
900                         break;
901                 case PCI_MODE_PCIX_M2_66:
902                         DBG_PRINT(ERR_DBG, "133MHz PCIX(M2) bus\n");
903                         break;
904                 case PCI_MODE_PCIX_M2_100:
905                         DBG_PRINT(ERR_DBG, "200MHz PCIX(M2) bus\n");
906                         break;
907                 case PCI_MODE_PCIX_M2_133:
908                         DBG_PRINT(ERR_DBG, "266MHz PCIX(M2) bus\n");
909                         break;
910                 default:
911                         return -1;      /* Unsupported bus speed */
912         }
913
914         return mode;
915 }
916
917 /**
918  *  init_nic - Initialization of hardware
919  *  @nic: device peivate variable
920  *  Description: The function sequentially configures every block
921  *  of the H/W from their reset values.
922  *  Return Value:  SUCCESS on success and
923  *  '-1' on failure (endian settings incorrect).
924  */
925
926 static int init_nic(struct s2io_nic *nic)
927 {
928         XENA_dev_config_t __iomem *bar0 = nic->bar0;
929         struct net_device *dev = nic->dev;
930         register u64 val64 = 0;
931         void __iomem *add;
932         u32 time;
933         int i, j;
934         mac_info_t *mac_control;
935         struct config_param *config;
936         int dtx_cnt = 0;
937         unsigned long long mem_share;
938         int mem_size;
939
940         mac_control = &nic->mac_control;
941         config = &nic->config;
942
943         /* to set the swapper controle on the card */
944         if(s2io_set_swapper(nic)) {
945                 DBG_PRINT(ERR_DBG,"ERROR: Setting Swapper failed\n");
946                 return -1;
947         }
948
949         /*
950          * Herc requires EOI to be removed from reset before XGXS, so..
951          */
952         if (nic->device_type & XFRAME_II_DEVICE) {
953                 val64 = 0xA500000000ULL;
954                 writeq(val64, &bar0->sw_reset);
955                 msleep(500);
956                 val64 = readq(&bar0->sw_reset);
957         }
958
959         /* Remove XGXS from reset state */
960         val64 = 0;
961         writeq(val64, &bar0->sw_reset);
962         msleep(500);
963         val64 = readq(&bar0->sw_reset);
964
965         /*  Enable Receiving broadcasts */
966         add = &bar0->mac_cfg;
967         val64 = readq(&bar0->mac_cfg);
968         val64 |= MAC_RMAC_BCAST_ENABLE;
969         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
970         writel((u32) val64, add);
971         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
972         writel((u32) (val64 >> 32), (add + 4));
973
974         /* Read registers in all blocks */
975         val64 = readq(&bar0->mac_int_mask);
976         val64 = readq(&bar0->mc_int_mask);
977         val64 = readq(&bar0->xgxs_int_mask);
978
979         /*  Set MTU */
980         val64 = dev->mtu;
981         writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
982
983         if (nic->device_type & XFRAME_II_DEVICE) {
984                 while (herc_act_dtx_cfg[dtx_cnt] != END_SIGN) {
985                         SPECIAL_REG_WRITE(herc_act_dtx_cfg[dtx_cnt],
986                                           &bar0->dtx_control, UF);
987                         if (dtx_cnt & 0x1)
988                                 msleep(1); /* Necessary!! */
989                         dtx_cnt++;
990                 }
991         } else {
992                 while (xena_dtx_cfg[dtx_cnt] != END_SIGN) {
993                         SPECIAL_REG_WRITE(xena_dtx_cfg[dtx_cnt],
994                                           &bar0->dtx_control, UF);
995                         val64 = readq(&bar0->dtx_control);
996                         dtx_cnt++;
997                 }
998         }
999
1000         /*  Tx DMA Initialization */
1001         val64 = 0;
1002         writeq(val64, &bar0->tx_fifo_partition_0);
1003         writeq(val64, &bar0->tx_fifo_partition_1);
1004         writeq(val64, &bar0->tx_fifo_partition_2);
1005         writeq(val64, &bar0->tx_fifo_partition_3);
1006
1007
1008         for (i = 0, j = 0; i < config->tx_fifo_num; i++) {
1009                 val64 |=
1010                     vBIT(config->tx_cfg[i].fifo_len - 1, ((i * 32) + 19),
1011                          13) | vBIT(config->tx_cfg[i].fifo_priority,
1012                                     ((i * 32) + 5), 3);
1013
1014                 if (i == (config->tx_fifo_num - 1)) {
1015                         if (i % 2 == 0)
1016                                 i++;
1017                 }
1018
1019                 switch (i) {
1020                 case 1:
1021                         writeq(val64, &bar0->tx_fifo_partition_0);
1022                         val64 = 0;
1023                         break;
1024                 case 3:
1025                         writeq(val64, &bar0->tx_fifo_partition_1);
1026                         val64 = 0;
1027                         break;
1028                 case 5:
1029                         writeq(val64, &bar0->tx_fifo_partition_2);
1030                         val64 = 0;
1031                         break;
1032                 case 7:
1033                         writeq(val64, &bar0->tx_fifo_partition_3);
1034                         break;
1035                 }
1036         }
1037
1038         /*
1039          * Disable 4 PCCs for Xena1, 2 and 3 as per H/W bug
1040          * SXE-008 TRANSMIT DMA ARBITRATION ISSUE.
1041          */
1042         if ((nic->device_type == XFRAME_I_DEVICE) &&
1043                 (get_xena_rev_id(nic->pdev) < 4))
1044                 writeq(PCC_ENABLE_FOUR, &bar0->pcc_enable);
1045
1046         val64 = readq(&bar0->tx_fifo_partition_0);
1047         DBG_PRINT(INIT_DBG, "Fifo partition at: 0x%p is: 0x%llx\n",
1048                   &bar0->tx_fifo_partition_0, (unsigned long long) val64);
1049
1050         /*
1051          * Initialization of Tx_PA_CONFIG register to ignore packet
1052          * integrity checking.
1053          */
1054         val64 = readq(&bar0->tx_pa_cfg);
1055         val64 |= TX_PA_CFG_IGNORE_FRM_ERR | TX_PA_CFG_IGNORE_SNAP_OUI |
1056             TX_PA_CFG_IGNORE_LLC_CTRL | TX_PA_CFG_IGNORE_L2_ERR;
1057         writeq(val64, &bar0->tx_pa_cfg);
1058
1059         /* Rx DMA intialization. */
1060         val64 = 0;
1061         for (i = 0; i < config->rx_ring_num; i++) {
1062                 val64 |=
1063                     vBIT(config->rx_cfg[i].ring_priority, (5 + (i * 8)),
1064                          3);
1065         }
1066         writeq(val64, &bar0->rx_queue_priority);
1067
1068         /*
1069          * Allocating equal share of memory to all the
1070          * configured Rings.
1071          */
1072         val64 = 0;
1073         if (nic->device_type & XFRAME_II_DEVICE)
1074                 mem_size = 32;
1075         else
1076                 mem_size = 64;
1077
1078         for (i = 0; i < config->rx_ring_num; i++) {
1079                 switch (i) {
1080                 case 0:
1081                         mem_share = (mem_size / config->rx_ring_num +
1082                                      mem_size % config->rx_ring_num);
1083                         val64 |= RX_QUEUE_CFG_Q0_SZ(mem_share);
1084                         continue;
1085                 case 1:
1086                         mem_share = (mem_size / config->rx_ring_num);
1087                         val64 |= RX_QUEUE_CFG_Q1_SZ(mem_share);
1088                         continue;
1089                 case 2:
1090                         mem_share = (mem_size / config->rx_ring_num);
1091                         val64 |= RX_QUEUE_CFG_Q2_SZ(mem_share);
1092                         continue;
1093                 case 3:
1094                         mem_share = (mem_size / config->rx_ring_num);
1095                         val64 |= RX_QUEUE_CFG_Q3_SZ(mem_share);
1096                         continue;
1097                 case 4:
1098                         mem_share = (mem_size / config->rx_ring_num);
1099                         val64 |= RX_QUEUE_CFG_Q4_SZ(mem_share);
1100                         continue;
1101                 case 5:
1102                         mem_share = (mem_size / config->rx_ring_num);
1103                         val64 |= RX_QUEUE_CFG_Q5_SZ(mem_share);
1104                         continue;
1105                 case 6:
1106                         mem_share = (mem_size / config->rx_ring_num);
1107                         val64 |= RX_QUEUE_CFG_Q6_SZ(mem_share);
1108                         continue;
1109                 case 7:
1110                         mem_share = (mem_size / config->rx_ring_num);
1111                         val64 |= RX_QUEUE_CFG_Q7_SZ(mem_share);
1112                         continue;
1113                 }
1114         }
1115         writeq(val64, &bar0->rx_queue_cfg);
1116
1117         /*
1118          * Filling Tx round robin registers
1119          * as per the number of FIFOs
1120          */
1121         switch (config->tx_fifo_num) {
1122         case 1:
1123                 val64 = 0x0000000000000000ULL;
1124                 writeq(val64, &bar0->tx_w_round_robin_0);
1125                 writeq(val64, &bar0->tx_w_round_robin_1);
1126                 writeq(val64, &bar0->tx_w_round_robin_2);
1127                 writeq(val64, &bar0->tx_w_round_robin_3);
1128                 writeq(val64, &bar0->tx_w_round_robin_4);
1129                 break;
1130         case 2:
1131                 val64 = 0x0000010000010000ULL;
1132                 writeq(val64, &bar0->tx_w_round_robin_0);
1133                 val64 = 0x0100000100000100ULL;
1134                 writeq(val64, &bar0->tx_w_round_robin_1);
1135                 val64 = 0x0001000001000001ULL;
1136                 writeq(val64, &bar0->tx_w_round_robin_2);
1137                 val64 = 0x0000010000010000ULL;
1138                 writeq(val64, &bar0->tx_w_round_robin_3);
1139                 val64 = 0x0100000000000000ULL;
1140                 writeq(val64, &bar0->tx_w_round_robin_4);
1141                 break;
1142         case 3:
1143                 val64 = 0x0001000102000001ULL;
1144                 writeq(val64, &bar0->tx_w_round_robin_0);
1145                 val64 = 0x0001020000010001ULL;
1146                 writeq(val64, &bar0->tx_w_round_robin_1);
1147                 val64 = 0x0200000100010200ULL;
1148                 writeq(val64, &bar0->tx_w_round_robin_2);
1149                 val64 = 0x0001000102000001ULL;
1150                 writeq(val64, &bar0->tx_w_round_robin_3);
1151                 val64 = 0x0001020000000000ULL;
1152                 writeq(val64, &bar0->tx_w_round_robin_4);
1153                 break;
1154         case 4:
1155                 val64 = 0x0001020300010200ULL;
1156                 writeq(val64, &bar0->tx_w_round_robin_0);
1157                 val64 = 0x0100000102030001ULL;
1158                 writeq(val64, &bar0->tx_w_round_robin_1);
1159                 val64 = 0x0200010000010203ULL;
1160                 writeq(val64, &bar0->tx_w_round_robin_2);
1161                 val64 = 0x0001020001000001ULL;
1162                 writeq(val64, &bar0->tx_w_round_robin_3);
1163                 val64 = 0x0203000100000000ULL;
1164                 writeq(val64, &bar0->tx_w_round_robin_4);
1165                 break;
1166         case 5:
1167                 val64 = 0x0001000203000102ULL;
1168                 writeq(val64, &bar0->tx_w_round_robin_0);
1169                 val64 = 0x0001020001030004ULL;
1170                 writeq(val64, &bar0->tx_w_round_robin_1);
1171                 val64 = 0x0001000203000102ULL;
1172                 writeq(val64, &bar0->tx_w_round_robin_2);
1173                 val64 = 0x0001020001030004ULL;
1174                 writeq(val64, &bar0->tx_w_round_robin_3);
1175                 val64 = 0x0001000000000000ULL;
1176                 writeq(val64, &bar0->tx_w_round_robin_4);
1177                 break;
1178         case 6:
1179                 val64 = 0x0001020304000102ULL;
1180                 writeq(val64, &bar0->tx_w_round_robin_0);
1181                 val64 = 0x0304050001020001ULL;
1182                 writeq(val64, &bar0->tx_w_round_robin_1);
1183                 val64 = 0x0203000100000102ULL;
1184                 writeq(val64, &bar0->tx_w_round_robin_2);
1185                 val64 = 0x0304000102030405ULL;
1186                 writeq(val64, &bar0->tx_w_round_robin_3);
1187                 val64 = 0x0001000200000000ULL;
1188                 writeq(val64, &bar0->tx_w_round_robin_4);
1189                 break;
1190         case 7:
1191                 val64 = 0x0001020001020300ULL;
1192                 writeq(val64, &bar0->tx_w_round_robin_0);
1193                 val64 = 0x0102030400010203ULL;
1194                 writeq(val64, &bar0->tx_w_round_robin_1);
1195                 val64 = 0x0405060001020001ULL;
1196                 writeq(val64, &bar0->tx_w_round_robin_2);
1197                 val64 = 0x0304050000010200ULL;
1198                 writeq(val64, &bar0->tx_w_round_robin_3);
1199                 val64 = 0x0102030000000000ULL;
1200                 writeq(val64, &bar0->tx_w_round_robin_4);
1201                 break;
1202         case 8:
1203                 val64 = 0x0001020300040105ULL;
1204                 writeq(val64, &bar0->tx_w_round_robin_0);
1205                 val64 = 0x0200030106000204ULL;
1206                 writeq(val64, &bar0->tx_w_round_robin_1);
1207                 val64 = 0x0103000502010007ULL;
1208                 writeq(val64, &bar0->tx_w_round_robin_2);
1209                 val64 = 0x0304010002060500ULL;
1210                 writeq(val64, &bar0->tx_w_round_robin_3);
1211                 val64 = 0x0103020400000000ULL;
1212                 writeq(val64, &bar0->tx_w_round_robin_4);
1213                 break;
1214         }
1215
1216         /* Enable Tx FIFO partition 0. */
1217         val64 = readq(&bar0->tx_fifo_partition_0);
1218         val64 |= (TX_FIFO_PARTITION_EN);
1219         writeq(val64, &bar0->tx_fifo_partition_0);
1220
1221         /* Filling the Rx round robin registers as per the
1222          * number of Rings and steering based on QoS.
1223          */
1224         switch (config->rx_ring_num) {
1225         case 1:
1226                 val64 = 0x8080808080808080ULL;
1227                 writeq(val64, &bar0->rts_qos_steering);
1228                 break;
1229         case 2:
1230                 val64 = 0x0000010000010000ULL;
1231                 writeq(val64, &bar0->rx_w_round_robin_0);
1232                 val64 = 0x0100000100000100ULL;
1233                 writeq(val64, &bar0->rx_w_round_robin_1);
1234                 val64 = 0x0001000001000001ULL;
1235                 writeq(val64, &bar0->rx_w_round_robin_2);
1236                 val64 = 0x0000010000010000ULL;
1237                 writeq(val64, &bar0->rx_w_round_robin_3);
1238                 val64 = 0x0100000000000000ULL;
1239                 writeq(val64, &bar0->rx_w_round_robin_4);
1240
1241                 val64 = 0x8080808040404040ULL;
1242                 writeq(val64, &bar0->rts_qos_steering);
1243                 break;
1244         case 3:
1245                 val64 = 0x0001000102000001ULL;
1246                 writeq(val64, &bar0->rx_w_round_robin_0);
1247                 val64 = 0x0001020000010001ULL;
1248                 writeq(val64, &bar0->rx_w_round_robin_1);
1249                 val64 = 0x0200000100010200ULL;
1250                 writeq(val64, &bar0->rx_w_round_robin_2);
1251                 val64 = 0x0001000102000001ULL;
1252                 writeq(val64, &bar0->rx_w_round_robin_3);
1253                 val64 = 0x0001020000000000ULL;
1254                 writeq(val64, &bar0->rx_w_round_robin_4);
1255
1256                 val64 = 0x8080804040402020ULL;
1257                 writeq(val64, &bar0->rts_qos_steering);
1258                 break;
1259         case 4:
1260                 val64 = 0x0001020300010200ULL;
1261                 writeq(val64, &bar0->rx_w_round_robin_0);
1262                 val64 = 0x0100000102030001ULL;
1263                 writeq(val64, &bar0->rx_w_round_robin_1);
1264                 val64 = 0x0200010000010203ULL;
1265                 writeq(val64, &bar0->rx_w_round_robin_2);
1266                 val64 = 0x0001020001000001ULL;  
1267                 writeq(val64, &bar0->rx_w_round_robin_3);
1268                 val64 = 0x0203000100000000ULL;
1269                 writeq(val64, &bar0->rx_w_round_robin_4);
1270
1271                 val64 = 0x8080404020201010ULL;
1272                 writeq(val64, &bar0->rts_qos_steering);
1273                 break;
1274         case 5:
1275                 val64 = 0x0001000203000102ULL;
1276                 writeq(val64, &bar0->rx_w_round_robin_0);
1277                 val64 = 0x0001020001030004ULL;
1278                 writeq(val64, &bar0->rx_w_round_robin_1);
1279                 val64 = 0x0001000203000102ULL;
1280                 writeq(val64, &bar0->rx_w_round_robin_2);
1281                 val64 = 0x0001020001030004ULL;
1282                 writeq(val64, &bar0->rx_w_round_robin_3);
1283                 val64 = 0x0001000000000000ULL;
1284                 writeq(val64, &bar0->rx_w_round_robin_4);
1285
1286                 val64 = 0x8080404020201008ULL;
1287                 writeq(val64, &bar0->rts_qos_steering);
1288                 break;
1289         case 6:
1290                 val64 = 0x0001020304000102ULL;
1291                 writeq(val64, &bar0->rx_w_round_robin_0);
1292                 val64 = 0x0304050001020001ULL;
1293                 writeq(val64, &bar0->rx_w_round_robin_1);
1294                 val64 = 0x0203000100000102ULL;
1295                 writeq(val64, &bar0->rx_w_round_robin_2);
1296                 val64 = 0x0304000102030405ULL;
1297                 writeq(val64, &bar0->rx_w_round_robin_3);
1298                 val64 = 0x0001000200000000ULL;
1299                 writeq(val64, &bar0->rx_w_round_robin_4);
1300
1301                 val64 = 0x8080404020100804ULL;
1302                 writeq(val64, &bar0->rts_qos_steering);
1303                 break;
1304         case 7:
1305                 val64 = 0x0001020001020300ULL;
1306                 writeq(val64, &bar0->rx_w_round_robin_0);
1307                 val64 = 0x0102030400010203ULL;
1308                 writeq(val64, &bar0->rx_w_round_robin_1);
1309                 val64 = 0x0405060001020001ULL;
1310                 writeq(val64, &bar0->rx_w_round_robin_2);
1311                 val64 = 0x0304050000010200ULL;
1312                 writeq(val64, &bar0->rx_w_round_robin_3);
1313                 val64 = 0x0102030000000000ULL;
1314                 writeq(val64, &bar0->rx_w_round_robin_4);
1315
1316                 val64 = 0x8080402010080402ULL;
1317                 writeq(val64, &bar0->rts_qos_steering);
1318                 break;
1319         case 8:
1320                 val64 = 0x0001020300040105ULL;
1321                 writeq(val64, &bar0->rx_w_round_robin_0);
1322                 val64 = 0x0200030106000204ULL;
1323                 writeq(val64, &bar0->rx_w_round_robin_1);
1324                 val64 = 0x0103000502010007ULL;
1325                 writeq(val64, &bar0->rx_w_round_robin_2);
1326                 val64 = 0x0304010002060500ULL;
1327                 writeq(val64, &bar0->rx_w_round_robin_3);
1328                 val64 = 0x0103020400000000ULL;
1329                 writeq(val64, &bar0->rx_w_round_robin_4);
1330
1331                 val64 = 0x8040201008040201ULL;
1332                 writeq(val64, &bar0->rts_qos_steering);
1333                 break;
1334         }
1335
1336         /* UDP Fix */
1337         val64 = 0;
1338         for (i = 0; i < 8; i++)
1339                 writeq(val64, &bar0->rts_frm_len_n[i]);
1340
1341         /* Set the default rts frame length for the rings configured */
1342         val64 = MAC_RTS_FRM_LEN_SET(dev->mtu+22);
1343         for (i = 0 ; i < config->rx_ring_num ; i++)
1344                 writeq(val64, &bar0->rts_frm_len_n[i]);
1345
1346         /* Set the frame length for the configured rings
1347          * desired by the user
1348          */
1349         for (i = 0; i < config->rx_ring_num; i++) {
1350                 /* If rts_frm_len[i] == 0 then it is assumed that user not
1351                  * specified frame length steering.
1352                  * If the user provides the frame length then program
1353                  * the rts_frm_len register for those values or else
1354                  * leave it as it is.
1355                  */
1356                 if (rts_frm_len[i] != 0) {
1357                         writeq(MAC_RTS_FRM_LEN_SET(rts_frm_len[i]),
1358                                 &bar0->rts_frm_len_n[i]);
1359                 }
1360         }
1361
1362         /* Program statistics memory */
1363         writeq(mac_control->stats_mem_phy, &bar0->stat_addr);
1364
1365         if (nic->device_type == XFRAME_II_DEVICE) {
1366                 val64 = STAT_BC(0x320);
1367                 writeq(val64, &bar0->stat_byte_cnt);
1368         }
1369
1370         /*
1371          * Initializing the sampling rate for the device to calculate the
1372          * bandwidth utilization.
1373          */
1374         val64 = MAC_TX_LINK_UTIL_VAL(tmac_util_period) |
1375             MAC_RX_LINK_UTIL_VAL(rmac_util_period);
1376         writeq(val64, &bar0->mac_link_util);
1377
1378
1379         /*
1380          * Initializing the Transmit and Receive Traffic Interrupt
1381          * Scheme.
1382          */
1383         /*
1384          * TTI Initialization. Default Tx timer gets us about
1385          * 250 interrupts per sec. Continuous interrupts are enabled
1386          * by default.
1387          */
1388         if (nic->device_type == XFRAME_II_DEVICE) {
1389                 int count = (nic->config.bus_speed * 125)/2;
1390                 val64 = TTI_DATA1_MEM_TX_TIMER_VAL(count);
1391         } else {
1392
1393                 val64 = TTI_DATA1_MEM_TX_TIMER_VAL(0x2078);
1394         }
1395         val64 |= TTI_DATA1_MEM_TX_URNG_A(0xA) |
1396             TTI_DATA1_MEM_TX_URNG_B(0x10) |
1397             TTI_DATA1_MEM_TX_URNG_C(0x30) | TTI_DATA1_MEM_TX_TIMER_AC_EN;
1398                 if (use_continuous_tx_intrs)
1399                         val64 |= TTI_DATA1_MEM_TX_TIMER_CI_EN;
1400         writeq(val64, &bar0->tti_data1_mem);
1401
1402         val64 = TTI_DATA2_MEM_TX_UFC_A(0x10) |
1403             TTI_DATA2_MEM_TX_UFC_B(0x20) |
1404             TTI_DATA2_MEM_TX_UFC_C(0x70) | TTI_DATA2_MEM_TX_UFC_D(0x80);
1405         writeq(val64, &bar0->tti_data2_mem);
1406
1407         val64 = TTI_CMD_MEM_WE | TTI_CMD_MEM_STROBE_NEW_CMD;
1408         writeq(val64, &bar0->tti_command_mem);
1409
1410         /*
1411          * Once the operation completes, the Strobe bit of the command
1412          * register will be reset. We poll for this particular condition
1413          * We wait for a maximum of 500ms for the operation to complete,
1414          * if it's not complete by then we return error.
1415          */
1416         time = 0;
1417         while (TRUE) {
1418                 val64 = readq(&bar0->tti_command_mem);
1419                 if (!(val64 & TTI_CMD_MEM_STROBE_NEW_CMD)) {
1420                         break;
1421                 }
1422                 if (time > 10) {
1423                         DBG_PRINT(ERR_DBG, "%s: TTI init Failed\n",
1424                                   dev->name);
1425                         return -1;
1426                 }
1427                 msleep(50);
1428                 time++;
1429         }
1430
1431         if (nic->config.bimodal) {
1432                 int k = 0;
1433                 for (k = 0; k < config->rx_ring_num; k++) {
1434                         val64 = TTI_CMD_MEM_WE | TTI_CMD_MEM_STROBE_NEW_CMD;
1435                         val64 |= TTI_CMD_MEM_OFFSET(0x38+k);
1436                         writeq(val64, &bar0->tti_command_mem);
1437
1438                 /*
1439                  * Once the operation completes, the Strobe bit of the command
1440                  * register will be reset. We poll for this particular condition
1441                  * We wait for a maximum of 500ms for the operation to complete,
1442                  * if it's not complete by then we return error.
1443                 */
1444                         time = 0;
1445                         while (TRUE) {
1446                                 val64 = readq(&bar0->tti_command_mem);
1447                                 if (!(val64 & TTI_CMD_MEM_STROBE_NEW_CMD)) {
1448                                         break;
1449                                 }
1450                                 if (time > 10) {
1451                                         DBG_PRINT(ERR_DBG,
1452                                                 "%s: TTI init Failed\n",
1453                                         dev->name);
1454                                         return -1;
1455                                 }
1456                                 time++;
1457                                 msleep(50);
1458                         }
1459                 }
1460         } else {
1461
1462                 /* RTI Initialization */
1463                 if (nic->device_type == XFRAME_II_DEVICE) {
1464                         /*
1465                          * Programmed to generate Apprx 500 Intrs per
1466                          * second
1467                          */
1468                         int count = (nic->config.bus_speed * 125)/4;
1469                         val64 = RTI_DATA1_MEM_RX_TIMER_VAL(count);
1470                 } else {
1471                         val64 = RTI_DATA1_MEM_RX_TIMER_VAL(0xFFF);
1472                 }
1473                 val64 |= RTI_DATA1_MEM_RX_URNG_A(0xA) |
1474                     RTI_DATA1_MEM_RX_URNG_B(0x10) |
1475                     RTI_DATA1_MEM_RX_URNG_C(0x30) | RTI_DATA1_MEM_RX_TIMER_AC_EN;
1476
1477                 writeq(val64, &bar0->rti_data1_mem);
1478
1479                 val64 = RTI_DATA2_MEM_RX_UFC_A(0x1) |
1480                     RTI_DATA2_MEM_RX_UFC_B(0x2) ;
1481                 if (nic->intr_type == MSI_X)
1482                     val64 |= (RTI_DATA2_MEM_RX_UFC_C(0x20) | \
1483                                 RTI_DATA2_MEM_RX_UFC_D(0x40));
1484                 else
1485                     val64 |= (RTI_DATA2_MEM_RX_UFC_C(0x40) | \
1486                                 RTI_DATA2_MEM_RX_UFC_D(0x80));
1487                 writeq(val64, &bar0->rti_data2_mem);
1488
1489                 for (i = 0; i < config->rx_ring_num; i++) {
1490                         val64 = RTI_CMD_MEM_WE | RTI_CMD_MEM_STROBE_NEW_CMD
1491                                         | RTI_CMD_MEM_OFFSET(i);
1492                         writeq(val64, &bar0->rti_command_mem);
1493
1494                         /*
1495                          * Once the operation completes, the Strobe bit of the
1496                          * command register will be reset. We poll for this
1497                          * particular condition. We wait for a maximum of 500ms
1498                          * for the operation to complete, if it's not complete
1499                          * by then we return error.
1500                          */
1501                         time = 0;
1502                         while (TRUE) {
1503                                 val64 = readq(&bar0->rti_command_mem);
1504                                 if (!(val64 & RTI_CMD_MEM_STROBE_NEW_CMD)) {
1505                                         break;
1506                                 }
1507                                 if (time > 10) {
1508                                         DBG_PRINT(ERR_DBG, "%s: RTI init Failed\n",
1509                                                   dev->name);
1510                                         return -1;
1511                                 }
1512                                 time++;
1513                                 msleep(50);
1514                         }
1515                 }
1516         }
1517
1518         /*
1519          * Initializing proper values as Pause threshold into all
1520          * the 8 Queues on Rx side.
1521          */
1522         writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q0q3);
1523         writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q4q7);
1524
1525         /* Disable RMAC PAD STRIPPING */
1526         add = &bar0->mac_cfg;
1527         val64 = readq(&bar0->mac_cfg);
1528         val64 &= ~(MAC_CFG_RMAC_STRIP_PAD);
1529         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1530         writel((u32) (val64), add);
1531         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1532         writel((u32) (val64 >> 32), (add + 4));
1533         val64 = readq(&bar0->mac_cfg);
1534
1535         /* Enable FCS stripping by adapter */
1536         add = &bar0->mac_cfg;
1537         val64 = readq(&bar0->mac_cfg);
1538         val64 |= MAC_CFG_RMAC_STRIP_FCS;
1539         if (nic->device_type == XFRAME_II_DEVICE)
1540                 writeq(val64, &bar0->mac_cfg);
1541         else {
1542                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1543                 writel((u32) (val64), add);
1544                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1545                 writel((u32) (val64 >> 32), (add + 4));
1546         }
1547
1548         /*
1549          * Set the time value to be inserted in the pause frame
1550          * generated by xena.
1551          */
1552         val64 = readq(&bar0->rmac_pause_cfg);
1553         val64 &= ~(RMAC_PAUSE_HG_PTIME(0xffff));
1554         val64 |= RMAC_PAUSE_HG_PTIME(nic->mac_control.rmac_pause_time);
1555         writeq(val64, &bar0->rmac_pause_cfg);
1556
1557         /*
1558          * Set the Threshold Limit for Generating the pause frame
1559          * If the amount of data in any Queue exceeds ratio of
1560          * (mac_control.mc_pause_threshold_q0q3 or q4q7)/256
1561          * pause frame is generated
1562          */
1563         val64 = 0;
1564         for (i = 0; i < 4; i++) {
1565                 val64 |=
1566                     (((u64) 0xFF00 | nic->mac_control.
1567                       mc_pause_threshold_q0q3)
1568                      << (i * 2 * 8));
1569         }
1570         writeq(val64, &bar0->mc_pause_thresh_q0q3);
1571
1572         val64 = 0;
1573         for (i = 0; i < 4; i++) {
1574                 val64 |=
1575                     (((u64) 0xFF00 | nic->mac_control.
1576                       mc_pause_threshold_q4q7)
1577                      << (i * 2 * 8));
1578         }
1579         writeq(val64, &bar0->mc_pause_thresh_q4q7);
1580
1581         /*
1582          * TxDMA will stop Read request if the number of read split has
1583          * exceeded the limit pointed by shared_splits
1584          */
1585         val64 = readq(&bar0->pic_control);
1586         val64 |= PIC_CNTL_SHARED_SPLITS(shared_splits);
1587         writeq(val64, &bar0->pic_control);
1588
1589         if (nic->config.bus_speed == 266) {
1590                 writeq(TXREQTO_VAL(0x7f) | TXREQTO_EN, &bar0->txreqtimeout);
1591                 writeq(0x0, &bar0->read_retry_delay);
1592                 writeq(0x0, &bar0->write_retry_delay);
1593         }
1594
1595         /*
1596          * Programming the Herc to split every write transaction
1597          * that does not start on an ADB to reduce disconnects.
1598          */
1599         if (nic->device_type == XFRAME_II_DEVICE) {
1600                 val64 = EXT_REQ_EN | MISC_LINK_STABILITY_PRD(3);
1601                 writeq(val64, &bar0->misc_control);
1602                 val64 = readq(&bar0->pic_control2);
1603                 val64 &= ~(BIT(13)|BIT(14)|BIT(15));
1604                 writeq(val64, &bar0->pic_control2);
1605         }
1606         if (strstr(nic->product_name, "CX4")) {
1607                 val64 = TMAC_AVG_IPG(0x17);
1608                 writeq(val64, &bar0->tmac_avg_ipg);
1609         }
1610
1611         return SUCCESS;
1612 }
1613 #define LINK_UP_DOWN_INTERRUPT          1
1614 #define MAC_RMAC_ERR_TIMER              2
1615
1616 static int s2io_link_fault_indication(nic_t *nic)
1617 {
1618         if (nic->intr_type != INTA)
1619                 return MAC_RMAC_ERR_TIMER;
1620         if (nic->device_type == XFRAME_II_DEVICE)
1621                 return LINK_UP_DOWN_INTERRUPT;
1622         else
1623                 return MAC_RMAC_ERR_TIMER;
1624 }
1625
1626 /**
1627  *  en_dis_able_nic_intrs - Enable or Disable the interrupts
1628  *  @nic: device private variable,
1629  *  @mask: A mask indicating which Intr block must be modified and,
1630  *  @flag: A flag indicating whether to enable or disable the Intrs.
1631  *  Description: This function will either disable or enable the interrupts
1632  *  depending on the flag argument. The mask argument can be used to
1633  *  enable/disable any Intr block.
1634  *  Return Value: NONE.
1635  */
1636
1637 static void en_dis_able_nic_intrs(struct s2io_nic *nic, u16 mask, int flag)
1638 {
1639         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1640         register u64 val64 = 0, temp64 = 0;
1641
1642         /*  Top level interrupt classification */
1643         /*  PIC Interrupts */
1644         if ((mask & (TX_PIC_INTR | RX_PIC_INTR))) {
1645                 /*  Enable PIC Intrs in the general intr mask register */
1646                 val64 = TXPIC_INT_M | PIC_RX_INT_M;
1647                 if (flag == ENABLE_INTRS) {
1648                         temp64 = readq(&bar0->general_int_mask);
1649                         temp64 &= ~((u64) val64);
1650                         writeq(temp64, &bar0->general_int_mask);
1651                         /*
1652                          * If Hercules adapter enable GPIO otherwise
1653                          * disabled all PCIX, Flash, MDIO, IIC and GPIO
1654                          * interrupts for now.
1655                          * TODO
1656                          */
1657                         if (s2io_link_fault_indication(nic) ==
1658                                         LINK_UP_DOWN_INTERRUPT ) {
1659                                 temp64 = readq(&bar0->pic_int_mask);
1660                                 temp64 &= ~((u64) PIC_INT_GPIO);
1661                                 writeq(temp64, &bar0->pic_int_mask);
1662                                 temp64 = readq(&bar0->gpio_int_mask);
1663                                 temp64 &= ~((u64) GPIO_INT_MASK_LINK_UP);
1664                                 writeq(temp64, &bar0->gpio_int_mask);
1665                         } else {
1666                                 writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
1667                         }
1668                         /*
1669                          * No MSI Support is available presently, so TTI and
1670                          * RTI interrupts are also disabled.
1671                          */
1672                 } else if (flag == DISABLE_INTRS) {
1673                         /*
1674                          * Disable PIC Intrs in the general
1675                          * intr mask register
1676                          */
1677                         writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
1678                         temp64 = readq(&bar0->general_int_mask);
1679                         val64 |= temp64;
1680                         writeq(val64, &bar0->general_int_mask);
1681                 }
1682         }
1683
1684         /*  DMA Interrupts */
1685         /*  Enabling/Disabling Tx DMA interrupts */
1686         if (mask & TX_DMA_INTR) {
1687                 /* Enable TxDMA Intrs in the general intr mask register */
1688                 val64 = TXDMA_INT_M;
1689                 if (flag == ENABLE_INTRS) {
1690                         temp64 = readq(&bar0->general_int_mask);
1691                         temp64 &= ~((u64) val64);
1692                         writeq(temp64, &bar0->general_int_mask);
1693                         /*
1694                          * Keep all interrupts other than PFC interrupt
1695                          * and PCC interrupt disabled in DMA level.
1696                          */
1697                         val64 = DISABLE_ALL_INTRS & ~(TXDMA_PFC_INT_M |
1698                                                       TXDMA_PCC_INT_M);
1699                         writeq(val64, &bar0->txdma_int_mask);
1700                         /*
1701                          * Enable only the MISC error 1 interrupt in PFC block
1702                          */
1703                         val64 = DISABLE_ALL_INTRS & (~PFC_MISC_ERR_1);
1704                         writeq(val64, &bar0->pfc_err_mask);
1705                         /*
1706                          * Enable only the FB_ECC error interrupt in PCC block
1707                          */
1708                         val64 = DISABLE_ALL_INTRS & (~PCC_FB_ECC_ERR);
1709                         writeq(val64, &bar0->pcc_err_mask);
1710                 } else if (flag == DISABLE_INTRS) {
1711                         /*
1712                          * Disable TxDMA Intrs in the general intr mask
1713                          * register
1714                          */
1715                         writeq(DISABLE_ALL_INTRS, &bar0->txdma_int_mask);
1716                         writeq(DISABLE_ALL_INTRS, &bar0->pfc_err_mask);
1717                         temp64 = readq(&bar0->general_int_mask);
1718                         val64 |= temp64;
1719                         writeq(val64, &bar0->general_int_mask);
1720                 }
1721         }
1722
1723         /*  Enabling/Disabling Rx DMA interrupts */
1724         if (mask & RX_DMA_INTR) {
1725                 /*  Enable RxDMA Intrs in the general intr mask register */
1726                 val64 = RXDMA_INT_M;
1727                 if (flag == ENABLE_INTRS) {
1728                         temp64 = readq(&bar0->general_int_mask);
1729                         temp64 &= ~((u64) val64);
1730                         writeq(temp64, &bar0->general_int_mask);
1731                         /*
1732                          * All RxDMA block interrupts are disabled for now
1733                          * TODO
1734                          */
1735                         writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask);
1736                 } else if (flag == DISABLE_INTRS) {
1737                         /*
1738                          * Disable RxDMA Intrs in the general intr mask
1739                          * register
1740                          */
1741                         writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask);
1742                         temp64 = readq(&bar0->general_int_mask);
1743                         val64 |= temp64;
1744                         writeq(val64, &bar0->general_int_mask);
1745                 }
1746         }
1747
1748         /*  MAC Interrupts */
1749         /*  Enabling/Disabling MAC interrupts */
1750         if (mask & (TX_MAC_INTR | RX_MAC_INTR)) {
1751                 val64 = TXMAC_INT_M | RXMAC_INT_M;
1752                 if (flag == ENABLE_INTRS) {
1753                         temp64 = readq(&bar0->general_int_mask);
1754                         temp64 &= ~((u64) val64);
1755                         writeq(temp64, &bar0->general_int_mask);
1756                         /*
1757                          * All MAC block error interrupts are disabled for now
1758                          * TODO
1759                          */
1760                 } else if (flag == DISABLE_INTRS) {
1761                         /*
1762                          * Disable MAC Intrs in the general intr mask register
1763                          */
1764                         writeq(DISABLE_ALL_INTRS, &bar0->mac_int_mask);
1765                         writeq(DISABLE_ALL_INTRS,
1766                                &bar0->mac_rmac_err_mask);
1767
1768                         temp64 = readq(&bar0->general_int_mask);
1769                         val64 |= temp64;
1770                         writeq(val64, &bar0->general_int_mask);
1771                 }
1772         }
1773
1774         /*  XGXS Interrupts */
1775         if (mask & (TX_XGXS_INTR | RX_XGXS_INTR)) {
1776                 val64 = TXXGXS_INT_M | RXXGXS_INT_M;
1777                 if (flag == ENABLE_INTRS) {
1778                         temp64 = readq(&bar0->general_int_mask);
1779                         temp64 &= ~((u64) val64);
1780                         writeq(temp64, &bar0->general_int_mask);
1781                         /*
1782                          * All XGXS block error interrupts are disabled for now
1783                          * TODO
1784                          */
1785                         writeq(DISABLE_ALL_INTRS, &bar0->xgxs_int_mask);
1786                 } else if (flag == DISABLE_INTRS) {
1787                         /*
1788                          * Disable MC Intrs in the general intr mask register
1789                          */
1790                         writeq(DISABLE_ALL_INTRS, &bar0->xgxs_int_mask);
1791                         temp64 = readq(&bar0->general_int_mask);
1792                         val64 |= temp64;
1793                         writeq(val64, &bar0->general_int_mask);
1794                 }
1795         }
1796
1797         /*  Memory Controller(MC) interrupts */
1798         if (mask & MC_INTR) {
1799                 val64 = MC_INT_M;
1800                 if (flag == ENABLE_INTRS) {
1801                         temp64 = readq(&bar0->general_int_mask);
1802                         temp64 &= ~((u64) val64);
1803                         writeq(temp64, &bar0->general_int_mask);
1804                         /*
1805                          * Enable all MC Intrs.
1806                          */
1807                         writeq(0x0, &bar0->mc_int_mask);
1808                         writeq(0x0, &bar0->mc_err_mask);
1809                 } else if (flag == DISABLE_INTRS) {
1810                         /*
1811                          * Disable MC Intrs in the general intr mask register
1812                          */
1813                         writeq(DISABLE_ALL_INTRS, &bar0->mc_int_mask);
1814                         temp64 = readq(&bar0->general_int_mask);
1815                         val64 |= temp64;
1816                         writeq(val64, &bar0->general_int_mask);
1817                 }
1818         }
1819
1820
1821         /*  Tx traffic interrupts */
1822         if (mask & TX_TRAFFIC_INTR) {
1823                 val64 = TXTRAFFIC_INT_M;
1824                 if (flag == ENABLE_INTRS) {
1825                         temp64 = readq(&bar0->general_int_mask);
1826                         temp64 &= ~((u64) val64);
1827                         writeq(temp64, &bar0->general_int_mask);
1828                         /*
1829                          * Enable all the Tx side interrupts
1830                          * writing 0 Enables all 64 TX interrupt levels
1831                          */
1832                         writeq(0x0, &bar0->tx_traffic_mask);
1833                 } else if (flag == DISABLE_INTRS) {
1834                         /*
1835                          * Disable Tx Traffic Intrs in the general intr mask
1836                          * register.
1837                          */
1838                         writeq(DISABLE_ALL_INTRS, &bar0->tx_traffic_mask);
1839                         temp64 = readq(&bar0->general_int_mask);
1840                         val64 |= temp64;
1841                         writeq(val64, &bar0->general_int_mask);
1842                 }
1843         }
1844
1845         /*  Rx traffic interrupts */
1846         if (mask & RX_TRAFFIC_INTR) {
1847                 val64 = RXTRAFFIC_INT_M;
1848                 if (flag == ENABLE_INTRS) {
1849                         temp64 = readq(&bar0->general_int_mask);
1850                         temp64 &= ~((u64) val64);
1851                         writeq(temp64, &bar0->general_int_mask);
1852                         /* writing 0 Enables all 8 RX interrupt levels */
1853                         writeq(0x0, &bar0->rx_traffic_mask);
1854                 } else if (flag == DISABLE_INTRS) {
1855                         /*
1856                          * Disable Rx Traffic Intrs in the general intr mask
1857                          * register.
1858                          */
1859                         writeq(DISABLE_ALL_INTRS, &bar0->rx_traffic_mask);
1860                         temp64 = readq(&bar0->general_int_mask);
1861                         val64 |= temp64;
1862                         writeq(val64, &bar0->general_int_mask);
1863                 }
1864         }
1865 }
1866
1867 static int check_prc_pcc_state(u64 val64, int flag, int rev_id, int herc)
1868 {
1869         int ret = 0;
1870
1871         if (flag == FALSE) {
1872                 if ((!herc && (rev_id >= 4)) || herc) {
1873                         if (!(val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) &&
1874                             ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1875                              ADAPTER_STATUS_RC_PRC_QUIESCENT)) {
1876                                 ret = 1;
1877                         }
1878                 }else {
1879                         if (!(val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) &&
1880                             ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1881                              ADAPTER_STATUS_RC_PRC_QUIESCENT)) {
1882                                 ret = 1;
1883                         }
1884                 }
1885         } else {
1886                 if ((!herc && (rev_id >= 4)) || herc) {
1887                         if (((val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) ==
1888                              ADAPTER_STATUS_RMAC_PCC_IDLE) &&
1889                             (!(val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ||
1890                              ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1891                               ADAPTER_STATUS_RC_PRC_QUIESCENT))) {
1892                                 ret = 1;
1893                         }
1894                 } else {
1895                         if (((val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) ==
1896                              ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) &&
1897                             (!(val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ||
1898                              ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1899                               ADAPTER_STATUS_RC_PRC_QUIESCENT))) {
1900                                 ret = 1;
1901                         }
1902                 }
1903         }
1904
1905         return ret;
1906 }
1907 /**
1908  *  verify_xena_quiescence - Checks whether the H/W is ready
1909  *  @val64 :  Value read from adapter status register.
1910  *  @flag : indicates if the adapter enable bit was ever written once
1911  *  before.
1912  *  Description: Returns whether the H/W is ready to go or not. Depending
1913  *  on whether adapter enable bit was written or not the comparison
1914  *  differs and the calling function passes the input argument flag to
1915  *  indicate this.
1916  *  Return: 1 If xena is quiescence
1917  *          0 If Xena is not quiescence
1918  */
1919
1920 static int verify_xena_quiescence(nic_t *sp, u64 val64, int flag)
1921 {
1922         int ret = 0, herc;
1923         u64 tmp64 = ~((u64) val64);
1924         int rev_id = get_xena_rev_id(sp->pdev);
1925
1926         herc = (sp->device_type == XFRAME_II_DEVICE);
1927         if (!
1928             (tmp64 &
1929              (ADAPTER_STATUS_TDMA_READY | ADAPTER_STATUS_RDMA_READY |
1930               ADAPTER_STATUS_PFC_READY | ADAPTER_STATUS_TMAC_BUF_EMPTY |
1931               ADAPTER_STATUS_PIC_QUIESCENT | ADAPTER_STATUS_MC_DRAM_READY |
1932               ADAPTER_STATUS_MC_QUEUES_READY | ADAPTER_STATUS_M_PLL_LOCK |
1933               ADAPTER_STATUS_P_PLL_LOCK))) {
1934                 ret = check_prc_pcc_state(val64, flag, rev_id, herc);
1935         }
1936
1937         return ret;
1938 }
1939
1940 /**
1941  * fix_mac_address -  Fix for Mac addr problem on Alpha platforms
1942  * @sp: Pointer to device specifc structure
1943  * Description :
1944  * New procedure to clear mac address reading  problems on Alpha platforms
1945  *
1946  */
1947
1948 static void fix_mac_address(nic_t * sp)
1949 {
1950         XENA_dev_config_t __iomem *bar0 = sp->bar0;
1951         u64 val64;
1952         int i = 0;
1953
1954         while (fix_mac[i] != END_SIGN) {
1955                 writeq(fix_mac[i++], &bar0->gpio_control);
1956                 udelay(10);
1957                 val64 = readq(&bar0->gpio_control);
1958         }
1959 }
1960
1961 /**
1962  *  start_nic - Turns the device on
1963  *  @nic : device private variable.
1964  *  Description:
1965  *  This function actually turns the device on. Before this  function is
1966  *  called,all Registers are configured from their reset states
1967  *  and shared memory is allocated but the NIC is still quiescent. On
1968  *  calling this function, the device interrupts are cleared and the NIC is
1969  *  literally switched on by writing into the adapter control register.
1970  *  Return Value:
1971  *  SUCCESS on success and -1 on failure.
1972  */
1973
1974 static int start_nic(struct s2io_nic *nic)
1975 {
1976         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1977         struct net_device *dev = nic->dev;
1978         register u64 val64 = 0;
1979         u16 interruptible;
1980         u16 subid, i;
1981         mac_info_t *mac_control;
1982         struct config_param *config;
1983
1984         mac_control = &nic->mac_control;
1985         config = &nic->config;
1986
1987         /*  PRC Initialization and configuration */
1988         for (i = 0; i < config->rx_ring_num; i++) {
1989                 writeq((u64) mac_control->rings[i].rx_blocks[0].block_dma_addr,
1990                        &bar0->prc_rxd0_n[i]);
1991
1992                 val64 = readq(&bar0->prc_ctrl_n[i]);
1993                 if (nic->config.bimodal)
1994                         val64 |= PRC_CTRL_BIMODAL_INTERRUPT;
1995                 if (nic->rxd_mode == RXD_MODE_1)
1996                         val64 |= PRC_CTRL_RC_ENABLED;
1997                 else
1998                         val64 |= PRC_CTRL_RC_ENABLED | PRC_CTRL_RING_MODE_3;
1999                 if (nic->device_type == XFRAME_II_DEVICE)
2000                         val64 |= PRC_CTRL_GROUP_READS;
2001                 val64 &= ~PRC_CTRL_RXD_BACKOFF_INTERVAL(0xFFFFFF);
2002                 val64 |= PRC_CTRL_RXD_BACKOFF_INTERVAL(0x1000);
2003                 writeq(val64, &bar0->prc_ctrl_n[i]);
2004         }
2005
2006         if (nic->rxd_mode == RXD_MODE_3B) {
2007                 /* Enabling 2 buffer mode by writing into Rx_pa_cfg reg. */
2008                 val64 = readq(&bar0->rx_pa_cfg);
2009                 val64 |= RX_PA_CFG_IGNORE_L2_ERR;
2010                 writeq(val64, &bar0->rx_pa_cfg);
2011         }
2012
2013         /*
2014          * Enabling MC-RLDRAM. After enabling the device, we timeout
2015          * for around 100ms, which is approximately the time required
2016          * for the device to be ready for operation.
2017          */
2018         val64 = readq(&bar0->mc_rldram_mrs);
2019         val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE | MC_RLDRAM_MRS_ENABLE;
2020         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
2021         val64 = readq(&bar0->mc_rldram_mrs);
2022
2023         msleep(100);    /* Delay by around 100 ms. */
2024
2025         /* Enabling ECC Protection. */
2026         val64 = readq(&bar0->adapter_control);
2027         val64 &= ~ADAPTER_ECC_EN;
2028         writeq(val64, &bar0->adapter_control);
2029
2030         /*
2031          * Clearing any possible Link state change interrupts that
2032          * could have popped up just before Enabling the card.
2033          */
2034         val64 = readq(&bar0->mac_rmac_err_reg);
2035         if (val64)
2036                 writeq(val64, &bar0->mac_rmac_err_reg);
2037
2038         /*
2039          * Verify if the device is ready to be enabled, if so enable
2040          * it.
2041          */
2042         val64 = readq(&bar0->adapter_status);
2043         if (!verify_xena_quiescence(nic, val64, nic->device_enabled_once)) {
2044                 DBG_PRINT(ERR_DBG, "%s: device is not ready, ", dev->name);
2045                 DBG_PRINT(ERR_DBG, "Adapter status reads: 0x%llx\n",
2046                           (unsigned long long) val64);
2047                 return FAILURE;
2048         }
2049
2050         /*  Enable select interrupts */
2051         if (nic->intr_type != INTA)
2052                 en_dis_able_nic_intrs(nic, ENA_ALL_INTRS, DISABLE_INTRS);
2053         else {
2054                 interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR;
2055                 interruptible |= TX_PIC_INTR | RX_PIC_INTR;
2056                 interruptible |= TX_MAC_INTR | RX_MAC_INTR;
2057                 en_dis_able_nic_intrs(nic, interruptible, ENABLE_INTRS);
2058         }
2059
2060         /*
2061          * With some switches, link might be already up at this point.
2062          * Because of this weird behavior, when we enable laser,
2063          * we may not get link. We need to handle this. We cannot
2064          * figure out which switch is misbehaving. So we are forced to
2065          * make a global change.
2066          */
2067
2068         /* Enabling Laser. */
2069         val64 = readq(&bar0->adapter_control);
2070         val64 |= ADAPTER_EOI_TX_ON;
2071         writeq(val64, &bar0->adapter_control);
2072
2073         if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
2074                 /*
2075                  * Dont see link state interrupts initally on some switches,
2076                  * so directly scheduling the link state task here.
2077                  */
2078                 schedule_work(&nic->set_link_task);
2079         }
2080         /* SXE-002: Initialize link and activity LED */
2081         subid = nic->pdev->subsystem_device;
2082         if (((subid & 0xFF) >= 0x07) &&
2083             (nic->device_type == XFRAME_I_DEVICE)) {
2084                 val64 = readq(&bar0->gpio_control);
2085                 val64 |= 0x0000800000000000ULL;
2086                 writeq(val64, &bar0->gpio_control);
2087                 val64 = 0x0411040400000000ULL;
2088                 writeq(val64, (void __iomem *)bar0 + 0x2700);
2089         }
2090
2091         return SUCCESS;
2092 }
2093 /**
2094  * s2io_txdl_getskb - Get the skb from txdl, unmap and return skb
2095  */
2096 static struct sk_buff *s2io_txdl_getskb(fifo_info_t *fifo_data, TxD_t *txdlp, int get_off)
2097 {
2098         nic_t *nic = fifo_data->nic;
2099         struct sk_buff *skb;
2100         TxD_t *txds;
2101         u16 j, frg_cnt;
2102
2103         txds = txdlp;
2104         if (txds->Host_Control == (u64)(long)nic->ufo_in_band_v) {
2105                 pci_unmap_single(nic->pdev, (dma_addr_t)
2106                         txds->Buffer_Pointer, sizeof(u64),
2107                         PCI_DMA_TODEVICE);
2108                 txds++;
2109         }
2110
2111         skb = (struct sk_buff *) ((unsigned long)
2112                         txds->Host_Control);
2113         if (!skb) {
2114                 memset(txdlp, 0, (sizeof(TxD_t) * fifo_data->max_txds));
2115                 return NULL;
2116         }
2117         pci_unmap_single(nic->pdev, (dma_addr_t)
2118                          txds->Buffer_Pointer,
2119                          skb->len - skb->data_len,
2120                          PCI_DMA_TODEVICE);
2121         frg_cnt = skb_shinfo(skb)->nr_frags;
2122         if (frg_cnt) {
2123                 txds++;
2124                 for (j = 0; j < frg_cnt; j++, txds++) {
2125                         skb_frag_t *frag = &skb_shinfo(skb)->frags[j];
2126                         if (!txds->Buffer_Pointer)
2127                                 break;
2128                         pci_unmap_page(nic->pdev, (dma_addr_t) 
2129                                         txds->Buffer_Pointer,
2130                                        frag->size, PCI_DMA_TODEVICE);
2131                 }
2132         }
2133         txdlp->Host_Control = 0;
2134         return(skb);
2135 }
2136
2137 /**
2138  *  free_tx_buffers - Free all queued Tx buffers
2139  *  @nic : device private variable.
2140  *  Description:
2141  *  Free all queued Tx buffers.
2142  *  Return Value: void
2143 */
2144
2145 static void free_tx_buffers(struct s2io_nic *nic)
2146 {
2147         struct net_device *dev = nic->dev;
2148         struct sk_buff *skb;
2149         TxD_t *txdp;
2150         int i, j;
2151         mac_info_t *mac_control;
2152         struct config_param *config;
2153         int cnt = 0;
2154
2155         mac_control = &nic->mac_control;
2156         config = &nic->config;
2157
2158         for (i = 0; i < config->tx_fifo_num; i++) {
2159                 for (j = 0; j < config->tx_cfg[i].fifo_len - 1; j++) {
2160                         txdp = (TxD_t *) mac_control->fifos[i].list_info[j].
2161                             list_virt_addr;
2162                         skb = s2io_txdl_getskb(&mac_control->fifos[i], txdp, j);
2163                         if (skb) {
2164                                 dev_kfree_skb(skb);
2165                                 cnt++;
2166                         }
2167                 }
2168                 DBG_PRINT(INTR_DBG,
2169                           "%s:forcibly freeing %d skbs on FIFO%d\n",
2170                           dev->name, cnt, i);
2171                 mac_control->fifos[i].tx_curr_get_info.offset = 0;
2172                 mac_control->fifos[i].tx_curr_put_info.offset = 0;
2173         }
2174 }
2175
2176 /**
2177  *   stop_nic -  To stop the nic
2178  *   @nic ; device private variable.
2179  *   Description:
2180  *   This function does exactly the opposite of what the start_nic()
2181  *   function does. This function is called to stop the device.
2182  *   Return Value:
2183  *   void.
2184  */
2185
2186 static void stop_nic(struct s2io_nic *nic)
2187 {
2188         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2189         register u64 val64 = 0;
2190         u16 interruptible;
2191         mac_info_t *mac_control;
2192         struct config_param *config;
2193
2194         mac_control = &nic->mac_control;
2195         config = &nic->config;
2196
2197         /*  Disable all interrupts */
2198         interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR;
2199         interruptible |= TX_PIC_INTR | RX_PIC_INTR;
2200         interruptible |= TX_MAC_INTR | RX_MAC_INTR;
2201         en_dis_able_nic_intrs(nic, interruptible, DISABLE_INTRS);
2202
2203         /* Clearing Adapter_En bit of ADAPTER_CONTROL Register */
2204         val64 = readq(&bar0->adapter_control);
2205         val64 &= ~(ADAPTER_CNTL_EN);
2206         writeq(val64, &bar0->adapter_control);
2207 }
2208
2209 static int fill_rxd_3buf(nic_t *nic, RxD_t *rxdp, struct sk_buff *skb)
2210 {
2211         struct net_device *dev = nic->dev;
2212         struct sk_buff *frag_list;
2213         void *tmp;
2214
2215         /* Buffer-1 receives L3/L4 headers */
2216         ((RxD3_t*)rxdp)->Buffer1_ptr = pci_map_single
2217                         (nic->pdev, skb->data, l3l4hdr_size + 4,
2218                         PCI_DMA_FROMDEVICE);
2219
2220         /* skb_shinfo(skb)->frag_list will have L4 data payload */
2221         skb_shinfo(skb)->frag_list = dev_alloc_skb(dev->mtu + ALIGN_SIZE);
2222         if (skb_shinfo(skb)->frag_list == NULL) {
2223                 DBG_PRINT(ERR_DBG, "%s: dev_alloc_skb failed\n ", dev->name);
2224                 return -ENOMEM ;
2225         }
2226         frag_list = skb_shinfo(skb)->frag_list;
2227         frag_list->next = NULL;
2228         tmp = (void *)ALIGN((long)frag_list->data, ALIGN_SIZE + 1);
2229         frag_list->data = tmp;
2230         frag_list->tail = tmp;
2231
2232         /* Buffer-2 receives L4 data payload */
2233         ((RxD3_t*)rxdp)->Buffer2_ptr = pci_map_single(nic->pdev,
2234                                 frag_list->data, dev->mtu,
2235                                 PCI_DMA_FROMDEVICE);
2236         rxdp->Control_2 |= SET_BUFFER1_SIZE_3(l3l4hdr_size + 4);
2237         rxdp->Control_2 |= SET_BUFFER2_SIZE_3(dev->mtu);
2238
2239         return SUCCESS;
2240 }
2241
2242 /**
2243  *  fill_rx_buffers - Allocates the Rx side skbs
2244  *  @nic:  device private variable
2245  *  @ring_no: ring number
2246  *  Description:
2247  *  The function allocates Rx side skbs and puts the physical
2248  *  address of these buffers into the RxD buffer pointers, so that the NIC
2249  *  can DMA the received frame into these locations.
2250  *  The NIC supports 3 receive modes, viz
2251  *  1. single buffer,
2252  *  2. three buffer and
2253  *  3. Five buffer modes.
2254  *  Each mode defines how many fragments the received frame will be split
2255  *  up into by the NIC. The frame is split into L3 header, L4 Header,
2256  *  L4 payload in three buffer mode and in 5 buffer mode, L4 payload itself
2257  *  is split into 3 fragments. As of now only single buffer mode is
2258  *  supported.
2259  *   Return Value:
2260  *  SUCCESS on success or an appropriate -ve value on failure.
2261  */
2262
2263 static int fill_rx_buffers(struct s2io_nic *nic, int ring_no)
2264 {
2265         struct net_device *dev = nic->dev;
2266         struct sk_buff *skb;
2267         RxD_t *rxdp;
2268         int off, off1, size, block_no, block_no1;
2269         u32 alloc_tab = 0;
2270         u32 alloc_cnt;
2271         mac_info_t *mac_control;
2272         struct config_param *config;
2273         u64 tmp;
2274         buffAdd_t *ba;
2275 #ifndef CONFIG_S2IO_NAPI
2276         unsigned long flags;
2277 #endif
2278         RxD_t *first_rxdp = NULL;
2279
2280         mac_control = &nic->mac_control;
2281         config = &nic->config;
2282         alloc_cnt = mac_control->rings[ring_no].pkt_cnt -
2283             atomic_read(&nic->rx_bufs_left[ring_no]);
2284
2285         block_no1 = mac_control->rings[ring_no].rx_curr_get_info.block_index;
2286         off1 = mac_control->rings[ring_no].rx_curr_get_info.offset;
2287         while (alloc_tab < alloc_cnt) {
2288                 block_no = mac_control->rings[ring_no].rx_curr_put_info.
2289                     block_index;
2290                 off = mac_control->rings[ring_no].rx_curr_put_info.offset;
2291
2292                 rxdp = mac_control->rings[ring_no].
2293                                 rx_blocks[block_no].rxds[off].virt_addr;
2294
2295                 if ((block_no == block_no1) && (off == off1) &&
2296                                         (rxdp->Host_Control)) {
2297                         DBG_PRINT(INTR_DBG, "%s: Get and Put",
2298                                   dev->name);
2299                         DBG_PRINT(INTR_DBG, " info equated\n");
2300                         goto end;
2301                 }
2302                 if (off && (off == rxd_count[nic->rxd_mode])) {
2303                         mac_control->rings[ring_no].rx_curr_put_info.
2304                             block_index++;
2305                         if (mac_control->rings[ring_no].rx_curr_put_info.
2306                             block_index == mac_control->rings[ring_no].
2307                                         block_count)
2308                                 mac_control->rings[ring_no].rx_curr_put_info.
2309                                         block_index = 0;
2310                         block_no = mac_control->rings[ring_no].
2311                                         rx_curr_put_info.block_index;
2312                         if (off == rxd_count[nic->rxd_mode])
2313                                 off = 0;
2314                         mac_control->rings[ring_no].rx_curr_put_info.
2315                                 offset = off;
2316                         rxdp = mac_control->rings[ring_no].
2317                                 rx_blocks[block_no].block_virt_addr;
2318                         DBG_PRINT(INTR_DBG, "%s: Next block at: %p\n",
2319                                   dev->name, rxdp);
2320                 }
2321 #ifndef CONFIG_S2IO_NAPI
2322                 spin_lock_irqsave(&nic->put_lock, flags);
2323                 mac_control->rings[ring_no].put_pos =
2324                     (block_no * (rxd_count[nic->rxd_mode] + 1)) + off;
2325                 spin_unlock_irqrestore(&nic->put_lock, flags);
2326 #endif
2327                 if ((rxdp->Control_1 & RXD_OWN_XENA) &&
2328                         ((nic->rxd_mode >= RXD_MODE_3A) &&
2329                                 (rxdp->Control_2 & BIT(0)))) {
2330                         mac_control->rings[ring_no].rx_curr_put_info.
2331                                         offset = off;
2332                         goto end;
2333                 }
2334                 /* calculate size of skb based on ring mode */
2335                 size = dev->mtu + HEADER_ETHERNET_II_802_3_SIZE +
2336                                 HEADER_802_2_SIZE + HEADER_SNAP_SIZE;
2337                 if (nic->rxd_mode == RXD_MODE_1)
2338                         size += NET_IP_ALIGN;
2339                 else if (nic->rxd_mode == RXD_MODE_3B)
2340                         size = dev->mtu + ALIGN_SIZE + BUF0_LEN + 4;
2341                 else
2342                         size = l3l4hdr_size + ALIGN_SIZE + BUF0_LEN + 4;
2343
2344                 /* allocate skb */
2345                 skb = dev_alloc_skb(size);
2346                 if(!skb) {
2347                         DBG_PRINT(ERR_DBG, "%s: Out of ", dev->name);
2348                         DBG_PRINT(ERR_DBG, "memory to allocate SKBs\n");
2349                         if (first_rxdp) {
2350                                 wmb();
2351                                 first_rxdp->Control_1 |= RXD_OWN_XENA;
2352                         }
2353                         return -ENOMEM ;
2354                 }
2355                 if (nic->rxd_mode == RXD_MODE_1) {
2356                         /* 1 buffer mode - normal operation mode */
2357                         memset(rxdp, 0, sizeof(RxD1_t));
2358                         skb_reserve(skb, NET_IP_ALIGN);
2359                         ((RxD1_t*)rxdp)->Buffer0_ptr = pci_map_single
2360                             (nic->pdev, skb->data, size - NET_IP_ALIGN,
2361                                 PCI_DMA_FROMDEVICE);
2362                         rxdp->Control_2 = SET_BUFFER0_SIZE_1(size - NET_IP_ALIGN);
2363
2364                 } else if (nic->rxd_mode >= RXD_MODE_3A) {
2365                         /*
2366                          * 2 or 3 buffer mode -
2367                          * Both 2 buffer mode and 3 buffer mode provides 128
2368                          * byte aligned receive buffers.
2369                          *
2370                          * 3 buffer mode provides header separation where in
2371                          * skb->data will have L3/L4 headers where as
2372                          * skb_shinfo(skb)->frag_list will have the L4 data
2373                          * payload
2374                          */
2375
2376                         memset(rxdp, 0, sizeof(RxD3_t));
2377                         ba = &mac_control->rings[ring_no].ba[block_no][off];
2378                         skb_reserve(skb, BUF0_LEN);
2379                         tmp = (u64)(unsigned long) skb->data;
2380                         tmp += ALIGN_SIZE;
2381                         tmp &= ~ALIGN_SIZE;
2382                         skb->data = (void *) (unsigned long)tmp;
2383                         skb->tail = (void *) (unsigned long)tmp;
2384
2385                         ((RxD3_t*)rxdp)->Buffer0_ptr =
2386                             pci_map_single(nic->pdev, ba->ba_0, BUF0_LEN,
2387                                            PCI_DMA_FROMDEVICE);
2388                         rxdp->Control_2 = SET_BUFFER0_SIZE_3(BUF0_LEN);
2389                         if (nic->rxd_mode == RXD_MODE_3B) {
2390                                 /* Two buffer mode */
2391
2392                                 /*
2393                                  * Buffer2 will have L3/L4 header plus 
2394                                  * L4 payload
2395                                  */
2396                                 ((RxD3_t*)rxdp)->Buffer2_ptr = pci_map_single
2397                                 (nic->pdev, skb->data, dev->mtu + 4,
2398                                                 PCI_DMA_FROMDEVICE);
2399
2400                                 /* Buffer-1 will be dummy buffer not used */
2401                                 ((RxD3_t*)rxdp)->Buffer1_ptr =
2402                                 pci_map_single(nic->pdev, ba->ba_1, BUF1_LEN,
2403                                         PCI_DMA_FROMDEVICE);
2404                                 rxdp->Control_2 |= SET_BUFFER1_SIZE_3(1);
2405                                 rxdp->Control_2 |= SET_BUFFER2_SIZE_3
2406                                                                 (dev->mtu + 4);
2407                         } else {
2408                                 /* 3 buffer mode */
2409                                 if (fill_rxd_3buf(nic, rxdp, skb) == -ENOMEM) {
2410                                         dev_kfree_skb_irq(skb);
2411                                         if (first_rxdp) {
2412                                                 wmb();
2413                                                 first_rxdp->Control_1 |=
2414                                                         RXD_OWN_XENA;
2415                                         }
2416                                         return -ENOMEM ;
2417                                 }
2418                         }
2419                         rxdp->Control_2 |= BIT(0);
2420                 }
2421                 rxdp->Host_Control = (unsigned long) (skb);
2422                 if (alloc_tab & ((1 << rxsync_frequency) - 1))
2423                         rxdp->Control_1 |= RXD_OWN_XENA;
2424                 off++;
2425                 if (off == (rxd_count[nic->rxd_mode] + 1))
2426                         off = 0;
2427                 mac_control->rings[ring_no].rx_curr_put_info.offset = off;
2428
2429                 rxdp->Control_2 |= SET_RXD_MARKER;
2430                 if (!(alloc_tab & ((1 << rxsync_frequency) - 1))) {
2431                         if (first_rxdp) {
2432                                 wmb();
2433                                 first_rxdp->Control_1 |= RXD_OWN_XENA;
2434                         }
2435                         first_rxdp = rxdp;
2436                 }
2437                 atomic_inc(&nic->rx_bufs_left[ring_no]);
2438                 alloc_tab++;
2439         }
2440
2441       end:
2442         /* Transfer ownership of first descriptor to adapter just before
2443          * exiting. Before that, use memory barrier so that ownership
2444          * and other fields are seen by adapter correctly.
2445          */
2446         if (first_rxdp) {
2447                 wmb();
2448                 first_rxdp->Control_1 |= RXD_OWN_XENA;
2449         }
2450
2451         return SUCCESS;
2452 }
2453
2454 static void free_rxd_blk(struct s2io_nic *sp, int ring_no, int blk)
2455 {
2456         struct net_device *dev = sp->dev;
2457         int j;
2458         struct sk_buff *skb;
2459         RxD_t *rxdp;
2460         mac_info_t *mac_control;
2461         buffAdd_t *ba;
2462
2463         mac_control = &sp->mac_control;
2464         for (j = 0 ; j < rxd_count[sp->rxd_mode]; j++) {
2465                 rxdp = mac_control->rings[ring_no].
2466                                 rx_blocks[blk].rxds[j].virt_addr;
2467                 skb = (struct sk_buff *)
2468                         ((unsigned long) rxdp->Host_Control);
2469                 if (!skb) {
2470                         continue;
2471                 }
2472                 if (sp->rxd_mode == RXD_MODE_1) {
2473                         pci_unmap_single(sp->pdev, (dma_addr_t)
2474                                  ((RxD1_t*)rxdp)->Buffer0_ptr,
2475                                  dev->mtu +
2476                                  HEADER_ETHERNET_II_802_3_SIZE
2477                                  + HEADER_802_2_SIZE +
2478                                  HEADER_SNAP_SIZE,
2479                                  PCI_DMA_FROMDEVICE);
2480                         memset(rxdp, 0, sizeof(RxD1_t));
2481                 } else if(sp->rxd_mode == RXD_MODE_3B) {
2482                         ba = &mac_control->rings[ring_no].
2483                                 ba[blk][j];
2484                         pci_unmap_single(sp->pdev, (dma_addr_t)
2485                                  ((RxD3_t*)rxdp)->Buffer0_ptr,
2486                                  BUF0_LEN,
2487                                  PCI_DMA_FROMDEVICE);
2488                         pci_unmap_single(sp->pdev, (dma_addr_t)
2489                                  ((RxD3_t*)rxdp)->Buffer1_ptr,
2490                                  BUF1_LEN,
2491                                  PCI_DMA_FROMDEVICE);
2492                         pci_unmap_single(sp->pdev, (dma_addr_t)
2493                                  ((RxD3_t*)rxdp)->Buffer2_ptr,
2494                                  dev->mtu + 4,
2495                                  PCI_DMA_FROMDEVICE);
2496                         memset(rxdp, 0, sizeof(RxD3_t));
2497                 } else {
2498                         pci_unmap_single(sp->pdev, (dma_addr_t)
2499                                 ((RxD3_t*)rxdp)->Buffer0_ptr, BUF0_LEN,
2500                                 PCI_DMA_FROMDEVICE);
2501                         pci_unmap_single(sp->pdev, (dma_addr_t)
2502                                 ((RxD3_t*)rxdp)->Buffer1_ptr, 
2503                                 l3l4hdr_size + 4,
2504                                 PCI_DMA_FROMDEVICE);
2505                         pci_unmap_single(sp->pdev, (dma_addr_t)
2506                                 ((RxD3_t*)rxdp)->Buffer2_ptr, dev->mtu,
2507                                 PCI_DMA_FROMDEVICE);
2508                         memset(rxdp, 0, sizeof(RxD3_t));
2509                 }
2510                 dev_kfree_skb(skb);
2511                 atomic_dec(&sp->rx_bufs_left[ring_no]);
2512         }
2513 }
2514
2515 /**
2516  *  free_rx_buffers - Frees all Rx buffers
2517  *  @sp: device private variable.
2518  *  Description:
2519  *  This function will free all Rx buffers allocated by host.
2520  *  Return Value:
2521  *  NONE.
2522  */
2523
2524 static void free_rx_buffers(struct s2io_nic *sp)
2525 {
2526         struct net_device *dev = sp->dev;
2527         int i, blk = 0, buf_cnt = 0;
2528         mac_info_t *mac_control;
2529         struct config_param *config;
2530
2531         mac_control = &sp->mac_control;
2532         config = &sp->config;
2533
2534         for (i = 0; i < config->rx_ring_num; i++) {
2535                 for (blk = 0; blk < rx_ring_sz[i]; blk++)
2536                         free_rxd_blk(sp,i,blk);
2537
2538                 mac_control->rings[i].rx_curr_put_info.block_index = 0;
2539                 mac_control->rings[i].rx_curr_get_info.block_index = 0;
2540                 mac_control->rings[i].rx_curr_put_info.offset = 0;
2541                 mac_control->rings[i].rx_curr_get_info.offset = 0;
2542                 atomic_set(&sp->rx_bufs_left[i], 0);
2543                 DBG_PRINT(INIT_DBG, "%s:Freed 0x%x Rx Buffers on ring%d\n",
2544                           dev->name, buf_cnt, i);
2545         }
2546 }
2547
2548 /**
2549  * s2io_poll - Rx interrupt handler for NAPI support
2550  * @dev : pointer to the device structure.
2551  * @budget : The number of packets that were budgeted to be processed
2552  * during  one pass through the 'Poll" function.
2553  * Description:
2554  * Comes into picture only if NAPI support has been incorporated. It does
2555  * the same thing that rx_intr_handler does, but not in a interrupt context
2556  * also It will process only a given number of packets.
2557  * Return value:
2558  * 0 on success and 1 if there are No Rx packets to be processed.
2559  */
2560
2561 #if defined(CONFIG_S2IO_NAPI)
2562 static int s2io_poll(struct net_device *dev, int *budget)
2563 {
2564         nic_t *nic = dev->priv;
2565         int pkt_cnt = 0, org_pkts_to_process;
2566         mac_info_t *mac_control;
2567         struct config_param *config;
2568         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2569         u64 val64 = 0xFFFFFFFFFFFFFFFFULL;
2570         int i;
2571
2572         atomic_inc(&nic->isr_cnt);
2573         mac_control = &nic->mac_control;
2574         config = &nic->config;
2575
2576         nic->pkts_to_process = *budget;
2577         if (nic->pkts_to_process > dev->quota)
2578                 nic->pkts_to_process = dev->quota;
2579         org_pkts_to_process = nic->pkts_to_process;
2580
2581         writeq(val64, &bar0->rx_traffic_int);
2582         val64 = readl(&bar0->rx_traffic_int);
2583
2584         for (i = 0; i < config->rx_ring_num; i++) {
2585                 rx_intr_handler(&mac_control->rings[i]);
2586                 pkt_cnt = org_pkts_to_process - nic->pkts_to_process;
2587                 if (!nic->pkts_to_process) {
2588                         /* Quota for the current iteration has been met */
2589                         goto no_rx;
2590                 }
2591         }
2592         if (!pkt_cnt)
2593                 pkt_cnt = 1;
2594
2595         dev->quota -= pkt_cnt;
2596         *budget -= pkt_cnt;
2597         netif_rx_complete(dev);
2598
2599         for (i = 0; i < config->rx_ring_num; i++) {
2600                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2601                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2602                         DBG_PRINT(ERR_DBG, " in Rx Poll!!\n");
2603                         break;
2604                 }
2605         }
2606         /* Re enable the Rx interrupts. */
2607         writeq(0x0, &bar0->rx_traffic_mask);
2608         val64 = readl(&bar0->rx_traffic_mask);
2609         atomic_dec(&nic->isr_cnt);
2610         return 0;
2611
2612 no_rx:
2613         dev->quota -= pkt_cnt;
2614         *budget -= pkt_cnt;
2615
2616         for (i = 0; i < config->rx_ring_num; i++) {
2617                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2618                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2619                         DBG_PRINT(ERR_DBG, " in Rx Poll!!\n");
2620                         break;
2621                 }
2622         }
2623         atomic_dec(&nic->isr_cnt);
2624         return 1;
2625 }
2626 #endif
2627
2628 /**
2629  * s2io_netpoll - Rx interrupt service handler for netpoll support
2630  * @dev : pointer to the device structure.
2631  * Description:
2632  * Polling 'interrupt' - used by things like netconsole to send skbs
2633  * without having to re-enable interrupts. It's not called while
2634  * the interrupt routine is executing.
2635  */
2636
2637 #ifdef CONFIG_NET_POLL_CONTROLLER
2638 static void s2io_netpoll(struct net_device *dev)
2639 {
2640         nic_t *nic = dev->priv;
2641         mac_info_t *mac_control;
2642         struct config_param *config;
2643         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2644         u64 val64;
2645         int i;
2646
2647         disable_irq(dev->irq);
2648
2649         atomic_inc(&nic->isr_cnt);
2650         mac_control = &nic->mac_control;
2651         config = &nic->config;
2652
2653         val64 = readq(&bar0->rx_traffic_int);
2654         writeq(val64, &bar0->rx_traffic_int);
2655
2656         for (i = 0; i < config->rx_ring_num; i++)
2657                 rx_intr_handler(&mac_control->rings[i]);
2658
2659         for (i = 0; i < config->rx_ring_num; i++) {
2660                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2661                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2662                         DBG_PRINT(ERR_DBG, " in Rx Netpoll!!\n");
2663                         break;
2664                 }
2665         }
2666         atomic_dec(&nic->isr_cnt);
2667         enable_irq(dev->irq);
2668         return;
2669 }
2670 #endif
2671
2672 /**
2673  *  rx_intr_handler - Rx interrupt handler
2674  *  @nic: device private variable.
2675  *  Description:
2676  *  If the interrupt is because of a received frame or if the
2677  *  receive ring contains fresh as yet un-processed frames,this function is
2678  *  called. It picks out the RxD at which place the last Rx processing had
2679  *  stopped and sends the skb to the OSM's Rx handler and then increments
2680  *  the offset.
2681  *  Return Value:
2682  *  NONE.
2683  */
2684 static void rx_intr_handler(ring_info_t *ring_data)
2685 {
2686         nic_t *nic = ring_data->nic;
2687         struct net_device *dev = (struct net_device *) nic->dev;
2688         int get_block, put_block, put_offset;
2689         rx_curr_get_info_t get_info, put_info;
2690         RxD_t *rxdp;
2691         struct sk_buff *skb;
2692 #ifndef CONFIG_S2IO_NAPI
2693         int pkt_cnt = 0;
2694 #endif
2695         int i;
2696
2697         spin_lock(&nic->rx_lock);
2698         if (atomic_read(&nic->card_state) == CARD_DOWN) {
2699                 DBG_PRINT(INTR_DBG, "%s: %s going down for reset\n",
2700                           __FUNCTION__, dev->name);
2701                 spin_unlock(&nic->rx_lock);
2702                 return;
2703         }
2704
2705         get_info = ring_data->rx_curr_get_info;
2706         get_block = get_info.block_index;
2707         put_info = ring_data->rx_curr_put_info;
2708         put_block = put_info.block_index;
2709         rxdp = ring_data->rx_blocks[get_block].rxds[get_info.offset].virt_addr;
2710 #ifndef CONFIG_S2IO_NAPI
2711         spin_lock(&nic->put_lock);
2712         put_offset = ring_data->put_pos;
2713         spin_unlock(&nic->put_lock);
2714 #else
2715         put_offset = (put_block * (rxd_count[nic->rxd_mode] + 1)) +
2716                 put_info.offset;
2717 #endif
2718         while (RXD_IS_UP2DT(rxdp)) {
2719                 /* If your are next to put index then it's FIFO full condition */
2720                 if ((get_block == put_block) &&
2721                     (get_info.offset + 1) == put_info.offset) {
2722                         DBG_PRINT(ERR_DBG, "%s: Ring Full\n",dev->name);
2723                         break;
2724                 }
2725                 skb = (struct sk_buff *) ((unsigned long)rxdp->Host_Control);
2726                 if (skb == NULL) {
2727                         DBG_PRINT(ERR_DBG, "%s: The skb is ",
2728                                   dev->name);
2729                         DBG_PRINT(ERR_DBG, "Null in Rx Intr\n");
2730                         spin_unlock(&nic->rx_lock);
2731                         return;
2732                 }
2733                 if (nic->rxd_mode == RXD_MODE_1) {
2734                         pci_unmap_single(nic->pdev, (dma_addr_t)
2735                                  ((RxD1_t*)rxdp)->Buffer0_ptr,
2736                                  dev->mtu +
2737                                  HEADER_ETHERNET_II_802_3_SIZE +
2738                                  HEADER_802_2_SIZE +
2739                                  HEADER_SNAP_SIZE,
2740                                  PCI_DMA_FROMDEVICE);
2741                 } else if (nic->rxd_mode == RXD_MODE_3B) {
2742                         pci_unmap_single(nic->pdev, (dma_addr_t)
2743                                  ((RxD3_t*)rxdp)->Buffer0_ptr,
2744                                  BUF0_LEN, PCI_DMA_FROMDEVICE);
2745                         pci_unmap_single(nic->pdev, (dma_addr_t)
2746                                  ((RxD3_t*)rxdp)->Buffer1_ptr,
2747                                  BUF1_LEN, PCI_DMA_FROMDEVICE);
2748                         pci_unmap_single(nic->pdev, (dma_addr_t)
2749                                  ((RxD3_t*)rxdp)->Buffer2_ptr,
2750                                  dev->mtu + 4,
2751                                  PCI_DMA_FROMDEVICE);
2752                 } else {
2753                         pci_unmap_single(nic->pdev, (dma_addr_t)
2754                                          ((RxD3_t*)rxdp)->Buffer0_ptr, BUF0_LEN,
2755                                          PCI_DMA_FROMDEVICE);
2756                         pci_unmap_single(nic->pdev, (dma_addr_t)
2757                                          ((RxD3_t*)rxdp)->Buffer1_ptr,
2758                                          l3l4hdr_size + 4,
2759                                          PCI_DMA_FROMDEVICE);
2760                         pci_unmap_single(nic->pdev, (dma_addr_t)
2761                                          ((RxD3_t*)rxdp)->Buffer2_ptr,
2762                                          dev->mtu, PCI_DMA_FROMDEVICE);
2763                 }
2764                 prefetch(skb->data);
2765                 rx_osm_handler(ring_data, rxdp);
2766                 get_info.offset++;
2767                 ring_data->rx_curr_get_info.offset = get_info.offset;
2768                 rxdp = ring_data->rx_blocks[get_block].
2769                                 rxds[get_info.offset].virt_addr;
2770                 if (get_info.offset == rxd_count[nic->rxd_mode]) {
2771                         get_info.offset = 0;
2772                         ring_data->rx_curr_get_info.offset = get_info.offset;
2773                         get_block++;
2774                         if (get_block == ring_data->block_count)
2775                                 get_block = 0;
2776                         ring_data->rx_curr_get_info.block_index = get_block;
2777                         rxdp = ring_data->rx_blocks[get_block].block_virt_addr;
2778                 }
2779
2780 #ifdef CONFIG_S2IO_NAPI
2781                 nic->pkts_to_process -= 1;
2782                 if (!nic->pkts_to_process)
2783                         break;
2784 #else
2785                 pkt_cnt++;
2786                 if ((indicate_max_pkts) && (pkt_cnt > indicate_max_pkts))
2787                         break;
2788 #endif
2789         }
2790         if (nic->lro) {
2791                 /* Clear all LRO sessions before exiting */
2792                 for (i=0; i<MAX_LRO_SESSIONS; i++) {
2793                         lro_t *lro = &nic->lro0_n[i];
2794                         if (lro->in_use) {
2795                                 update_L3L4_header(nic, lro);
2796                                 queue_rx_frame(lro->parent);
2797                                 clear_lro_session(lro);
2798                         }
2799                 }
2800         }
2801
2802         spin_unlock(&nic->rx_lock);
2803 }
2804
2805 /**
2806  *  tx_intr_handler - Transmit interrupt handler
2807  *  @nic : device private variable
2808  *  Description:
2809  *  If an interrupt was raised to indicate DMA complete of the
2810  *  Tx packet, this function is called. It identifies the last TxD
2811  *  whose buffer was freed and frees all skbs whose data have already
2812  *  DMA'ed into the NICs internal memory.
2813  *  Return Value:
2814  *  NONE
2815  */
2816
2817 static void tx_intr_handler(fifo_info_t *fifo_data)
2818 {
2819         nic_t *nic = fifo_data->nic;
2820         struct net_device *dev = (struct net_device *) nic->dev;
2821         tx_curr_get_info_t get_info, put_info;
2822         struct sk_buff *skb;
2823         TxD_t *txdlp;
2824
2825         get_info = fifo_data->tx_curr_get_info;
2826         put_info = fifo_data->tx_curr_put_info;
2827         txdlp = (TxD_t *) fifo_data->list_info[get_info.offset].
2828             list_virt_addr;
2829         while ((!(txdlp->Control_1 & TXD_LIST_OWN_XENA)) &&
2830                (get_info.offset != put_info.offset) &&
2831                (txdlp->Host_Control)) {
2832                 /* Check for TxD errors */
2833                 if (txdlp->Control_1 & TXD_T_CODE) {
2834                         unsigned long long err;
2835                         err = txdlp->Control_1 & TXD_T_CODE;
2836                         if (err & 0x1) {
2837                                 nic->mac_control.stats_info->sw_stat.
2838                                                 parity_err_cnt++;
2839                         }
2840                         if ((err >> 48) == 0xA) {
2841                                 DBG_PRINT(TX_DBG, "TxD returned due \
2842 to loss of link\n");
2843                         }
2844                         else {
2845                                 DBG_PRINT(ERR_DBG, "***TxD error \
2846 %llx\n", err);
2847                         }
2848                 }
2849
2850                 skb = s2io_txdl_getskb(fifo_data, txdlp, get_info.offset);
2851                 if (skb == NULL) {
2852                         DBG_PRINT(ERR_DBG, "%s: Null skb ",
2853                         __FUNCTION__);
2854                         DBG_PRINT(ERR_DBG, "in Tx Free Intr\n");
2855                         return;
2856                 }
2857
2858                 /* Updating the statistics block */
2859                 nic->stats.tx_bytes += skb->len;
2860                 dev_kfree_skb_irq(skb);
2861
2862                 get_info.offset++;
2863                 if (get_info.offset == get_info.fifo_len + 1)
2864                         get_info.offset = 0;
2865                 txdlp = (TxD_t *) fifo_data->list_info
2866                     [get_info.offset].list_virt_addr;
2867                 fifo_data->tx_curr_get_info.offset =
2868                     get_info.offset;
2869         }
2870
2871         spin_lock(&nic->tx_lock);
2872         if (netif_queue_stopped(dev))
2873                 netif_wake_queue(dev);
2874         spin_unlock(&nic->tx_lock);
2875 }
2876
2877 /**
2878  *  s2io_mdio_write - Function to write in to MDIO registers
2879  *  @mmd_type : MMD type value (PMA/PMD/WIS/PCS/PHYXS)
2880  *  @addr     : address value
2881  *  @value    : data value
2882  *  @dev      : pointer to net_device structure
2883  *  Description:
2884  *  This function is used to write values to the MDIO registers
2885  *  NONE
2886  */
2887 static void s2io_mdio_write(u32 mmd_type, u64 addr, u16 value, struct net_device *dev)
2888 {
2889         u64 val64 = 0x0;
2890         nic_t *sp = dev->priv;
2891         XENA_dev_config_t *bar0 = (XENA_dev_config_t *)sp->bar0;
2892
2893         //address transaction
2894         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2895                         | MDIO_MMD_DEV_ADDR(mmd_type)
2896                         | MDIO_MMS_PRT_ADDR(0x0);
2897         writeq(val64, &bar0->mdio_control);
2898         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2899         writeq(val64, &bar0->mdio_control);
2900         udelay(100);
2901
2902         //Data transaction
2903         val64 = 0x0;
2904         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2905                         | MDIO_MMD_DEV_ADDR(mmd_type)
2906                         | MDIO_MMS_PRT_ADDR(0x0)
2907                         | MDIO_MDIO_DATA(value)
2908                         | MDIO_OP(MDIO_OP_WRITE_TRANS);
2909         writeq(val64, &bar0->mdio_control);
2910         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2911         writeq(val64, &bar0->mdio_control);
2912         udelay(100);
2913
2914         val64 = 0x0;
2915         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2916         | MDIO_MMD_DEV_ADDR(mmd_type)
2917         | MDIO_MMS_PRT_ADDR(0x0)
2918         | MDIO_OP(MDIO_OP_READ_TRANS);
2919         writeq(val64, &bar0->mdio_control);
2920         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2921         writeq(val64, &bar0->mdio_control);
2922         udelay(100);
2923
2924 }
2925
2926 /**
2927  *  s2io_mdio_read - Function to write in to MDIO registers
2928  *  @mmd_type : MMD type value (PMA/PMD/WIS/PCS/PHYXS)
2929  *  @addr     : address value
2930  *  @dev      : pointer to net_device structure
2931  *  Description:
2932  *  This function is used to read values to the MDIO registers
2933  *  NONE
2934  */
2935 static u64 s2io_mdio_read(u32 mmd_type, u64 addr, struct net_device *dev)
2936 {
2937         u64 val64 = 0x0;
2938         u64 rval64 = 0x0;
2939         nic_t *sp = dev->priv;
2940         XENA_dev_config_t *bar0 = (XENA_dev_config_t *)sp->bar0;
2941
2942         /* address transaction */
2943         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2944                         | MDIO_MMD_DEV_ADDR(mmd_type)
2945                         | MDIO_MMS_PRT_ADDR(0x0);
2946         writeq(val64, &bar0->mdio_control);
2947         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2948         writeq(val64, &bar0->mdio_control);
2949         udelay(100);
2950
2951         /* Data transaction */
2952         val64 = 0x0;
2953         val64 = val64 | MDIO_MMD_INDX_ADDR(addr)
2954                         | MDIO_MMD_DEV_ADDR(mmd_type)
2955                         | MDIO_MMS_PRT_ADDR(0x0)
2956                         | MDIO_OP(MDIO_OP_READ_TRANS);
2957         writeq(val64, &bar0->mdio_control);
2958         val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
2959         writeq(val64, &bar0->mdio_control);
2960         udelay(100);
2961
2962         /* Read the value from regs */
2963         rval64 = readq(&bar0->mdio_control);
2964         rval64 = rval64 & 0xFFFF0000;
2965         rval64 = rval64 >> 16;
2966         return rval64;
2967 }
2968 /**
2969  *  s2io_chk_xpak_counter - Function to check the status of the xpak counters
2970  *  @counter      : couter value to be updated
2971  *  @flag         : flag to indicate the status
2972  *  @type         : counter type
2973  *  Description:
2974  *  This function is to check the status of the xpak counters value
2975  *  NONE
2976  */
2977
2978 static void s2io_chk_xpak_counter(u64 *counter, u64 * regs_stat, u32 index, u16 flag, u16 type)
2979 {
2980         u64 mask = 0x3;
2981         u64 val64;
2982         int i;
2983         for(i = 0; i <index; i++)
2984                 mask = mask << 0x2;
2985
2986         if(flag > 0)
2987         {
2988                 *counter = *counter + 1;
2989                 val64 = *regs_stat & mask;
2990                 val64 = val64 >> (index * 0x2);
2991                 val64 = val64 + 1;
2992                 if(val64 == 3)
2993                 {
2994                         switch(type)
2995                         {
2996                         case 1:
2997                                 DBG_PRINT(ERR_DBG, "Take Xframe NIC out of "
2998                                           "service. Excessive temperatures may "
2999                                           "result in premature transceiver "
3000                                           "failure \n");
3001                         break;
3002                         case 2:
3003                                 DBG_PRINT(ERR_DBG, "Take Xframe NIC out of "
3004                                           "service Excessive bias currents may "
3005                                           "indicate imminent laser diode "
3006                                           "failure \n");
3007                         break;
3008                         case 3:
3009                                 DBG_PRINT(ERR_DBG, "Take Xframe NIC out of "
3010                                           "service Excessive laser output "
3011                                           "power may saturate far-end "
3012                                           "receiver\n");
3013                         break;
3014                         default:
3015                                 DBG_PRINT(ERR_DBG, "Incorrect XPAK Alarm "
3016                                           "type \n");
3017                         }
3018                         val64 = 0x0;
3019                 }
3020                 val64 = val64 << (index * 0x2);
3021                 *regs_stat = (*regs_stat & (~mask)) | (val64);
3022
3023         } else {
3024                 *regs_stat = *regs_stat & (~mask);
3025         }
3026 }
3027
3028 /**
3029  *  s2io_updt_xpak_counter - Function to update the xpak counters
3030  *  @dev         : pointer to net_device struct
3031  *  Description:
3032  *  This function is to upate the status of the xpak counters value
3033  *  NONE
3034  */
3035 static void s2io_updt_xpak_counter(struct net_device *dev)
3036 {
3037         u16 flag  = 0x0;
3038         u16 type  = 0x0;
3039         u16 val16 = 0x0;
3040         u64 val64 = 0x0;
3041         u64 addr  = 0x0;
3042
3043         nic_t *sp = dev->priv;
3044         StatInfo_t *stat_info = sp->mac_control.stats_info;
3045
3046         /* Check the communication with the MDIO slave */
3047         addr = 0x0000;
3048         val64 = 0x0;
3049         val64 = s2io_mdio_read(MDIO_MMD_PMA_DEV_ADDR, addr, dev);
3050         if((val64 == 0xFFFF) || (val64 == 0x0000))
3051         {
3052                 DBG_PRINT(ERR_DBG, "ERR: MDIO slave access failed - "
3053                           "Returned %llx\n", (unsigned long long)val64);
3054                 return;
3055         }
3056
3057         /* Check for the expecte value of 2040 at PMA address 0x0000 */
3058         if(val64 != 0x2040)
3059         {
3060                 DBG_PRINT(ERR_DBG, "Incorrect value at PMA address 0x0000 - ");
3061                 DBG_PRINT(ERR_DBG, "Returned: %llx- Expected: 0x2040\n",
3062                           (unsigned long long)val64);
3063                 return;
3064         }
3065
3066         /* Loading the DOM register to MDIO register */
3067         addr = 0xA100;
3068         s2io_mdio_write(MDIO_MMD_PMA_DEV_ADDR, addr, val16, dev);
3069         val64 = s2io_mdio_read(MDIO_MMD_PMA_DEV_ADDR, addr, dev);
3070
3071         /* Reading the Alarm flags */
3072         addr = 0xA070;
3073         val64 = 0x0;
3074         val64 = s2io_mdio_read(MDIO_MMD_PMA_DEV_ADDR, addr, dev);
3075
3076         flag = CHECKBIT(val64, 0x7);
3077         type = 1;
3078         s2io_chk_xpak_counter(&stat_info->xpak_stat.alarm_transceiver_temp_high,
3079                                 &stat_info->xpak_stat.xpak_regs_stat,
3080                                 0x0, flag, type);
3081
3082         if(CHECKBIT(val64, 0x6))
3083                 stat_info->xpak_stat.alarm_transceiver_temp_low++;
3084
3085         flag = CHECKBIT(val64, 0x3);
3086         type = 2;
3087         s2io_chk_xpak_counter(&stat_info->xpak_stat.alarm_laser_bias_current_high,
3088                                 &stat_info->xpak_stat.xpak_regs_stat,
3089                                 0x2, flag, type);
3090
3091         if(CHECKBIT(val64, 0x2))
3092                 stat_info->xpak_stat.alarm_laser_bias_current_low++;
3093
3094         flag = CHECKBIT(val64, 0x1);
3095         type = 3;
3096         s2io_chk_xpak_counter(&stat_info->xpak_stat.alarm_laser_output_power_high,
3097                                 &stat_info->xpak_stat.xpak_regs_stat,
3098                                 0x4, flag, type);
3099
3100         if(CHECKBIT(val64, 0x0))
3101                 stat_info->xpak_stat.alarm_laser_output_power_low++;
3102
3103         /* Reading the Warning flags */
3104         addr = 0xA074;
3105         val64 = 0x0;
3106         val64 = s2io_mdio_read(MDIO_MMD_PMA_DEV_ADDR, addr, dev);
3107
3108         if(CHECKBIT(val64, 0x7))
3109                 stat_info->xpak_stat.warn_transceiver_temp_high++;
3110
3111         if(CHECKBIT(val64, 0x6))
3112                 stat_info->xpak_stat.warn_transceiver_temp_low++;
3113
3114         if(CHECKBIT(val64, 0x3))
3115                 stat_info->xpak_stat.warn_laser_bias_current_high++;
3116
3117         if(CHECKBIT(val64, 0x2))
3118                 stat_info->xpak_stat.warn_laser_bias_current_low++;
3119
3120         if(CHECKBIT(val64, 0x1))
3121                 stat_info->xpak_stat.warn_laser_output_power_high++;
3122
3123         if(CHECKBIT(val64, 0x0))
3124                 stat_info->xpak_stat.warn_laser_output_power_low++;
3125 }
3126
3127 /**
3128  *  alarm_intr_handler - Alarm Interrrupt handler
3129  *  @nic: device private variable
3130  *  Description: If the interrupt was neither because of Rx packet or Tx
3131  *  complete, this function is called. If the interrupt was to indicate
3132  *  a loss of link, the OSM link status handler is invoked for any other
3133  *  alarm interrupt the block that raised the interrupt is displayed
3134  *  and a H/W reset is issued.
3135  *  Return Value:
3136  *  NONE
3137 */
3138
3139 static void alarm_intr_handler(struct s2io_nic *nic)
3140 {
3141         struct net_device *dev = (struct net_device *) nic->dev;
3142         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3143         register u64 val64 = 0, err_reg = 0;
3144         u64 cnt;
3145         int i;
3146         nic->mac_control.stats_info->sw_stat.ring_full_cnt = 0;
3147         /* Handling the XPAK counters update */
3148         if(nic->mac_control.stats_info->xpak_stat.xpak_timer_count < 72000) {
3149                 /* waiting for an hour */
3150                 nic->mac_control.stats_info->xpak_stat.xpak_timer_count++;
3151         } else {
3152                 s2io_updt_xpak_counter(dev);
3153                 /* reset the count to zero */
3154                 nic->mac_control.stats_info->xpak_stat.xpak_timer_count = 0;
3155         }
3156
3157         /* Handling link status change error Intr */
3158         if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
3159                 err_reg = readq(&bar0->mac_rmac_err_reg);
3160                 writeq(err_reg, &bar0->mac_rmac_err_reg);
3161                 if (err_reg & RMAC_LINK_STATE_CHANGE_INT) {
3162                         schedule_work(&nic->set_link_task);
3163                 }
3164         }
3165
3166         /* Handling Ecc errors */
3167         val64 = readq(&bar0->mc_err_reg);
3168         writeq(val64, &bar0->mc_err_reg);
3169         if (val64 & (MC_ERR_REG_ECC_ALL_SNG | MC_ERR_REG_ECC_ALL_DBL)) {
3170                 if (val64 & MC_ERR_REG_ECC_ALL_DBL) {
3171                         nic->mac_control.stats_info->sw_stat.
3172                                 double_ecc_errs++;
3173                         DBG_PRINT(INIT_DBG, "%s: Device indicates ",
3174                                   dev->name);
3175                         DBG_PRINT(INIT_DBG, "double ECC error!!\n");
3176                         if (nic->device_type != XFRAME_II_DEVICE) {
3177                                 /* Reset XframeI only if critical error */
3178                                 if (val64 & (MC_ERR_REG_MIRI_ECC_DB_ERR_0 |
3179                                              MC_ERR_REG_MIRI_ECC_DB_ERR_1)) {
3180                                         netif_stop_queue(dev);
3181                                         schedule_work(&nic->rst_timer_task);
3182                                         nic->mac_control.stats_info->sw_stat.
3183                                                         soft_reset_cnt++;
3184                                 }
3185                         }
3186                 } else {
3187                         nic->mac_control.stats_info->sw_stat.
3188                                 single_ecc_errs++;
3189                 }
3190         }
3191
3192         /* In case of a serious error, the device will be Reset. */
3193         val64 = readq(&bar0->serr_source);
3194         if (val64 & SERR_SOURCE_ANY) {
3195                 nic->mac_control.stats_info->sw_stat.serious_err_cnt++;
3196                 DBG_PRINT(ERR_DBG, "%s: Device indicates ", dev->name);
3197                 DBG_PRINT(ERR_DBG, "serious error %llx!!\n", 
3198                           (unsigned long long)val64);
3199                 netif_stop_queue(dev);
3200                 schedule_work(&nic->rst_timer_task);
3201                 nic->mac_control.stats_info->sw_stat.soft_reset_cnt++;
3202         }
3203
3204         /*
3205          * Also as mentioned in the latest Errata sheets if the PCC_FB_ECC
3206          * Error occurs, the adapter will be recycled by disabling the
3207          * adapter enable bit and enabling it again after the device
3208          * becomes Quiescent.
3209          */
3210         val64 = readq(&bar0->pcc_err_reg);
3211         writeq(val64, &bar0->pcc_err_reg);
3212         if (val64 & PCC_FB_ECC_DB_ERR) {
3213                 u64 ac = readq(&bar0->adapter_control);
3214                 ac &= ~(ADAPTER_CNTL_EN);
3215                 writeq(ac, &bar0->adapter_control);
3216                 ac = readq(&bar0->adapter_control);
3217                 schedule_work(&nic->set_link_task);
3218         }
3219         /* Check for data parity error */
3220         val64 = readq(&bar0->pic_int_status);
3221         if (val64 & PIC_INT_GPIO) {
3222                 val64 = readq(&bar0->gpio_int_reg);
3223                 if (val64 & GPIO_INT_REG_DP_ERR_INT) {
3224                         nic->mac_control.stats_info->sw_stat.parity_err_cnt++;
3225                         schedule_work(&nic->rst_timer_task);
3226                         nic->mac_control.stats_info->sw_stat.soft_reset_cnt++;
3227                 }
3228         }
3229
3230         /* Check for ring full counter */
3231         if (nic->device_type & XFRAME_II_DEVICE) {
3232                 val64 = readq(&bar0->ring_bump_counter1);
3233                 for (i=0; i<4; i++) {
3234                         cnt = ( val64 & vBIT(0xFFFF,(i*16),16));
3235                         cnt >>= 64 - ((i+1)*16);
3236                         nic->mac_control.stats_info->sw_stat.ring_full_cnt
3237                                 += cnt;
3238                 }
3239
3240                 val64 = readq(&bar0->ring_bump_counter2);
3241                 for (i=0; i<4; i++) {
3242                         cnt = ( val64 & vBIT(0xFFFF,(i*16),16));
3243                         cnt >>= 64 - ((i+1)*16);
3244                         nic->mac_control.stats_info->sw_stat.ring_full_cnt
3245                                 += cnt;
3246                 }
3247         }
3248
3249         /* Other type of interrupts are not being handled now,  TODO */
3250 }
3251
3252 /**
3253  *  wait_for_cmd_complete - waits for a command to complete.
3254  *  @sp : private member of the device structure, which is a pointer to the
3255  *  s2io_nic structure.
3256  *  Description: Function that waits for a command to Write into RMAC
3257  *  ADDR DATA registers to be completed and returns either success or
3258  *  error depending on whether the command was complete or not.
3259  *  Return value:
3260  *   SUCCESS on success and FAILURE on failure.
3261  */
3262
3263 static int wait_for_cmd_complete(void *addr, u64 busy_bit)
3264 {
3265         int ret = FAILURE, cnt = 0;
3266         u64 val64;
3267
3268         while (TRUE) {
3269                 val64 = readq(addr);
3270                 if (!(val64 & busy_bit)) {
3271                         ret = SUCCESS;
3272                         break;
3273                 }
3274
3275                 if(in_interrupt())
3276                         mdelay(50);
3277                 else
3278                         msleep(50);
3279
3280                 if (cnt++ > 10)
3281                         break;
3282         }
3283         return ret;
3284 }
3285
3286 /**
3287  *  s2io_reset - Resets the card.
3288  *  @sp : private member of the device structure.
3289  *  Description: Function to Reset the card. This function then also
3290  *  restores the previously saved PCI configuration space registers as
3291  *  the card reset also resets the configuration space.
3292  *  Return value:
3293  *  void.
3294  */
3295
3296 static void s2io_reset(nic_t * sp)
3297 {
3298         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3299         u64 val64;
3300         u16 subid, pci_cmd;
3301
3302         /* Back up  the PCI-X CMD reg, dont want to lose MMRBC, OST settings */
3303         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER, &(pci_cmd));
3304
3305         val64 = SW_RESET_ALL;
3306         writeq(val64, &bar0->sw_reset);
3307
3308         /*
3309          * At this stage, if the PCI write is indeed completed, the
3310          * card is reset and so is the PCI Config space of the device.
3311          * So a read cannot be issued at this stage on any of the
3312          * registers to ensure the write into "sw_reset" register
3313          * has gone through.
3314          * Question: Is there any system call that will explicitly force
3315          * all the write commands still pending on the bus to be pushed
3316          * through?
3317          * As of now I'am just giving a 250ms delay and hoping that the
3318          * PCI write to sw_reset register is done by this time.
3319          */
3320         msleep(250);
3321         if (strstr(sp->product_name, "CX4")) {
3322                 msleep(750);
3323         }
3324
3325         /* Restore the PCI state saved during initialization. */
3326         pci_restore_state(sp->pdev);
3327         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
3328                                      pci_cmd);
3329         s2io_init_pci(sp);
3330
3331         msleep(250);
3332
3333         /* Set swapper to enable I/O register access */
3334         s2io_set_swapper(sp);
3335
3336         /* Restore the MSIX table entries from local variables */
3337         restore_xmsi_data(sp);
3338
3339         /* Clear certain PCI/PCI-X fields after reset */
3340         if (sp->device_type == XFRAME_II_DEVICE) {
3341                 /* Clear parity err detect bit */
3342                 pci_write_config_word(sp->pdev, PCI_STATUS, 0x8000);
3343
3344                 /* Clearing PCIX Ecc status register */
3345                 pci_write_config_dword(sp->pdev, 0x68, 0x7C);
3346
3347                 /* Clearing PCI_STATUS error reflected here */
3348                 writeq(BIT(62), &bar0->txpic_int_reg);
3349         }
3350
3351         /* Reset device statistics maintained by OS */
3352         memset(&sp->stats, 0, sizeof (struct net_device_stats));
3353
3354         /* SXE-002: Configure link and activity LED to turn it off */
3355         subid = sp->pdev->subsystem_device;
3356         if (((subid & 0xFF) >= 0x07) &&
3357             (sp->device_type == XFRAME_I_DEVICE)) {
3358                 val64 = readq(&bar0->gpio_control);
3359                 val64 |= 0x0000800000000000ULL;
3360                 writeq(val64, &bar0->gpio_control);
3361                 val64 = 0x0411040400000000ULL;
3362                 writeq(val64, (void __iomem *)bar0 + 0x2700);
3363         }
3364
3365         /*
3366          * Clear spurious ECC interrupts that would have occured on
3367          * XFRAME II cards after reset.
3368          */
3369         if (sp->device_type == XFRAME_II_DEVICE) {
3370                 val64 = readq(&bar0->pcc_err_reg);
3371                 writeq(val64, &bar0->pcc_err_reg);
3372         }
3373
3374         sp->device_enabled_once = FALSE;
3375 }
3376
3377 /**
3378  *  s2io_set_swapper - to set the swapper controle on the card
3379  *  @sp : private member of the device structure,
3380  *  pointer to the s2io_nic structure.
3381  *  Description: Function to set the swapper control on the card
3382  *  correctly depending on the 'endianness' of the system.
3383  *  Return value:
3384  *  SUCCESS on success and FAILURE on failure.
3385  */
3386
3387 static int s2io_set_swapper(nic_t * sp)
3388 {
3389         struct net_device *dev = sp->dev;
3390         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3391         u64 val64, valt, valr;
3392
3393         /*
3394          * Set proper endian settings and verify the same by reading
3395          * the PIF Feed-back register.
3396          */
3397
3398         val64 = readq(&bar0->pif_rd_swapper_fb);
3399         if (val64 != 0x0123456789ABCDEFULL) {
3400                 int i = 0;
3401                 u64 value[] = { 0xC30000C3C30000C3ULL,   /* FE=1, SE=1 */
3402                                 0x8100008181000081ULL,  /* FE=1, SE=0 */
3403                                 0x4200004242000042ULL,  /* FE=0, SE=1 */
3404                                 0};                     /* FE=0, SE=0 */
3405
3406                 while(i<4) {
3407                         writeq(value[i], &bar0->swapper_ctrl);
3408                         val64 = readq(&bar0->pif_rd_swapper_fb);
3409                         if (val64 == 0x0123456789ABCDEFULL)
3410                                 break;
3411                         i++;
3412                 }
3413                 if (i == 4) {
3414                         DBG_PRINT(ERR_DBG, "%s: Endian settings are wrong, ",
3415                                 dev->name);
3416                         DBG_PRINT(ERR_DBG, "feedback read %llx\n",
3417                                 (unsigned long long) val64);
3418                         return FAILURE;
3419                 }
3420                 valr = value[i];
3421         } else {
3422                 valr = readq(&bar0->swapper_ctrl);
3423         }
3424
3425         valt = 0x0123456789ABCDEFULL;
3426         writeq(valt, &bar0->xmsi_address);
3427         val64 = readq(&bar0->xmsi_address);
3428
3429         if(val64 != valt) {
3430                 int i = 0;
3431                 u64 value[] = { 0x00C3C30000C3C300ULL,  /* FE=1, SE=1 */
3432                                 0x0081810000818100ULL,  /* FE=1, SE=0 */
3433                                 0x0042420000424200ULL,  /* FE=0, SE=1 */
3434                                 0};                     /* FE=0, SE=0 */
3435
3436                 while(i<4) {
3437                         writeq((value[i] | valr), &bar0->swapper_ctrl);
3438                         writeq(valt, &bar0->xmsi_address);
3439                         val64 = readq(&bar0->xmsi_address);
3440                         if(val64 == valt)
3441                                 break;
3442                         i++;
3443                 }
3444                 if(i == 4) {
3445                         unsigned long long x = val64;
3446                         DBG_PRINT(ERR_DBG, "Write failed, Xmsi_addr ");
3447                         DBG_PRINT(ERR_DBG, "reads:0x%llx\n", x);
3448                         return FAILURE;
3449                 }
3450         }
3451         val64 = readq(&bar0->swapper_ctrl);
3452         val64 &= 0xFFFF000000000000ULL;
3453
3454 #ifdef  __BIG_ENDIAN
3455         /*
3456          * The device by default set to a big endian format, so a
3457          * big endian driver need not set anything.
3458          */
3459         val64 |= (SWAPPER_CTRL_TXP_FE |
3460                  SWAPPER_CTRL_TXP_SE |
3461                  SWAPPER_CTRL_TXD_R_FE |
3462                  SWAPPER_CTRL_TXD_W_FE |
3463                  SWAPPER_CTRL_TXF_R_FE |
3464                  SWAPPER_CTRL_RXD_R_FE |
3465                  SWAPPER_CTRL_RXD_W_FE |
3466                  SWAPPER_CTRL_RXF_W_FE |
3467                  SWAPPER_CTRL_XMSI_FE |
3468                  SWAPPER_CTRL_STATS_FE | SWAPPER_CTRL_STATS_SE);
3469         if (sp->intr_type == INTA)
3470                 val64 |= SWAPPER_CTRL_XMSI_SE;
3471         writeq(val64, &bar0->swapper_ctrl);
3472 #else
3473         /*
3474          * Initially we enable all bits to make it accessible by the
3475          * driver, then we selectively enable only those bits that
3476          * we want to set.
3477          */
3478         val64 |= (SWAPPER_CTRL_TXP_FE |
3479                  SWAPPER_CTRL_TXP_SE |
3480                  SWAPPER_CTRL_TXD_R_FE |
3481                  SWAPPER_CTRL_TXD_R_SE |
3482                  SWAPPER_CTRL_TXD_W_FE |
3483                  SWAPPER_CTRL_TXD_W_SE |
3484                  SWAPPER_CTRL_TXF_R_FE |
3485                  SWAPPER_CTRL_RXD_R_FE |
3486                  SWAPPER_CTRL_RXD_R_SE |
3487                  SWAPPER_CTRL_RXD_W_FE |
3488                  SWAPPER_CTRL_RXD_W_SE |
3489                  SWAPPER_CTRL_RXF_W_FE |
3490                  SWAPPER_CTRL_XMSI_FE |
3491                  SWAPPER_CTRL_STATS_FE | SWAPPER_CTRL_STATS_SE);
3492         if (sp->intr_type == INTA)
3493                 val64 |= SWAPPER_CTRL_XMSI_SE;
3494         writeq(val64, &bar0->swapper_ctrl);
3495 #endif
3496         val64 = readq(&bar0->swapper_ctrl);
3497
3498         /*
3499          * Verifying if endian settings are accurate by reading a
3500          * feedback register.
3501          */
3502         val64 = readq(&bar0->pif_rd_swapper_fb);
3503         if (val64 != 0x0123456789ABCDEFULL) {
3504                 /* Endian settings are incorrect, calls for another dekko. */
3505                 DBG_PRINT(ERR_DBG, "%s: Endian settings are wrong, ",
3506                           dev->name);
3507                 DBG_PRINT(ERR_DBG, "feedback read %llx\n",
3508                           (unsigned long long) val64);
3509                 return FAILURE;
3510         }
3511
3512         return SUCCESS;
3513 }
3514
3515 static int wait_for_msix_trans(nic_t *nic, int i)
3516 {
3517         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3518         u64 val64;
3519         int ret = 0, cnt = 0;
3520
3521         do {
3522                 val64 = readq(&bar0->xmsi_access);
3523                 if (!(val64 & BIT(15)))
3524                         break;
3525                 mdelay(1);
3526                 cnt++;
3527         } while(cnt < 5);
3528         if (cnt == 5) {
3529                 DBG_PRINT(ERR_DBG, "XMSI # %d Access failed\n", i);
3530                 ret = 1;
3531         }
3532
3533         return ret;
3534 }
3535
3536 static void restore_xmsi_data(nic_t *nic)
3537 {
3538         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3539         u64 val64;
3540         int i;
3541
3542         for (i=0; i< nic->avail_msix_vectors; i++) {
3543                 writeq(nic->msix_info[i].addr, &bar0->xmsi_address);
3544                 writeq(nic->msix_info[i].data, &bar0->xmsi_data);
3545                 val64 = (BIT(7) | BIT(15) | vBIT(i, 26, 6));
3546                 writeq(val64, &bar0->xmsi_access);
3547                 if (wait_for_msix_trans(nic, i)) {
3548                         DBG_PRINT(ERR_DBG, "failed in %s\n", __FUNCTION__);
3549                         continue;
3550                 }
3551         }
3552 }
3553
3554 static void store_xmsi_data(nic_t *nic)
3555 {
3556         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3557         u64 val64, addr, data;
3558         int i;
3559
3560         /* Store and display */
3561         for (i=0; i< nic->avail_msix_vectors; i++) {
3562                 val64 = (BIT(15) | vBIT(i, 26, 6));
3563                 writeq(val64, &bar0->xmsi_access);
3564                 if (wait_for_msix_trans(nic, i)) {
3565                         DBG_PRINT(ERR_DBG, "failed in %s\n", __FUNCTION__);
3566                         continue;
3567                 }
3568                 addr = readq(&bar0->xmsi_address);
3569                 data = readq(&bar0->xmsi_data);
3570                 if (addr && data) {
3571                         nic->msix_info[i].addr = addr;
3572                         nic->msix_info[i].data = data;
3573                 }
3574         }
3575 }
3576
3577 int s2io_enable_msi(nic_t *nic)
3578 {
3579         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3580         u16 msi_ctrl, msg_val;
3581         struct config_param *config = &nic->config;
3582         struct net_device *dev = nic->dev;
3583         u64 val64, tx_mat, rx_mat;
3584         int i, err;
3585
3586         val64 = readq(&bar0->pic_control);
3587         val64 &= ~BIT(1);
3588         writeq(val64, &bar0->pic_control);
3589
3590         err = pci_enable_msi(nic->pdev);
3591         if (err) {
3592                 DBG_PRINT(ERR_DBG, "%s: enabling MSI failed\n",
3593                           nic->dev->name);
3594                 return err;
3595         }
3596
3597         /*
3598          * Enable MSI and use MSI-1 in stead of the standard MSI-0
3599          * for interrupt handling.
3600          */
3601         pci_read_config_word(nic->pdev, 0x4c, &msg_val);
3602         msg_val ^= 0x1;
3603         pci_write_config_word(nic->pdev, 0x4c, msg_val);
3604         pci_read_config_word(nic->pdev, 0x4c, &msg_val);
3605
3606         pci_read_config_word(nic->pdev, 0x42, &msi_ctrl);
3607         msi_ctrl |= 0x10;
3608         pci_write_config_word(nic->pdev, 0x42, msi_ctrl);
3609
3610         /* program MSI-1 into all usable Tx_Mat and Rx_Mat fields */
3611         tx_mat = readq(&bar0->tx_mat0_n[0]);
3612         for (i=0; i<config->tx_fifo_num; i++) {
3613                 tx_mat |= TX_MAT_SET(i, 1);
3614         }
3615         writeq(tx_mat, &bar0->tx_mat0_n[0]);
3616
3617         rx_mat = readq(&bar0->rx_mat);
3618         for (i=0; i<config->rx_ring_num; i++) {
3619                 rx_mat |= RX_MAT_SET(i, 1);
3620         }
3621         writeq(rx_mat, &bar0->rx_mat);
3622
3623         dev->irq = nic->pdev->irq;
3624         return 0;
3625 }
3626
3627 static int s2io_enable_msi_x(nic_t *nic)
3628 {
3629         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3630         u64 tx_mat, rx_mat;
3631         u16 msi_control; /* Temp variable */
3632         int ret, i, j, msix_indx = 1;
3633
3634         nic->entries = kmalloc(MAX_REQUESTED_MSI_X * sizeof(struct msix_entry),
3635                                GFP_KERNEL);
3636         if (nic->entries == NULL) {
3637                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n", __FUNCTION__);
3638                 return -ENOMEM;
3639         }
3640         memset(nic->entries, 0, MAX_REQUESTED_MSI_X * sizeof(struct msix_entry));
3641
3642         nic->s2io_entries =
3643                 kmalloc(MAX_REQUESTED_MSI_X * sizeof(struct s2io_msix_entry),
3644                                    GFP_KERNEL);
3645         if (nic->s2io_entries == NULL) {
3646                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n", __FUNCTION__);
3647                 kfree(nic->entries);
3648                 return -ENOMEM;
3649         }
3650         memset(nic->s2io_entries, 0,
3651                MAX_REQUESTED_MSI_X * sizeof(struct s2io_msix_entry));
3652
3653         for (i=0; i< MAX_REQUESTED_MSI_X; i++) {
3654                 nic->entries[i].entry = i;
3655                 nic->s2io_entries[i].entry = i;
3656                 nic->s2io_entries[i].arg = NULL;
3657                 nic->s2io_entries[i].in_use = 0;
3658         }
3659
3660         tx_mat = readq(&bar0->tx_mat0_n[0]);
3661         for (i=0; i<nic->config.tx_fifo_num; i++, msix_indx++) {
3662                 tx_mat |= TX_MAT_SET(i, msix_indx);
3663                 nic->s2io_entries[msix_indx].arg = &nic->mac_control.fifos[i];
3664                 nic->s2io_entries[msix_indx].type = MSIX_FIFO_TYPE;
3665                 nic->s2io_entries[msix_indx].in_use = MSIX_FLG;
3666         }
3667         writeq(tx_mat, &bar0->tx_mat0_n[0]);
3668
3669         if (!nic->config.bimodal) {
3670                 rx_mat = readq(&bar0->rx_mat);
3671                 for (j=0; j<nic->config.rx_ring_num; j++, msix_indx++) {
3672                         rx_mat |= RX_MAT_SET(j, msix_indx);
3673                         nic->s2io_entries[msix_indx].arg = &nic->mac_control.rings[j];
3674                         nic->s2io_entries[msix_indx].type = MSIX_RING_TYPE;
3675                         nic->s2io_entries[msix_indx].in_use = MSIX_FLG;
3676                 }
3677                 writeq(rx_mat, &bar0->rx_mat);
3678         } else {
3679                 tx_mat = readq(&bar0->tx_mat0_n[7]);
3680                 for (j=0; j<nic->config.rx_ring_num; j++, msix_indx++) {
3681                         tx_mat |= TX_MAT_SET(i, msix_indx);
3682                         nic->s2io_entries[msix_indx].arg = &nic->mac_control.rings[j];
3683                         nic->s2io_entries[msix_indx].type = MSIX_RING_TYPE;
3684                         nic->s2io_entries[msix_indx].in_use = MSIX_FLG;
3685                 }
3686                 writeq(tx_mat, &bar0->tx_mat0_n[7]);
3687         }
3688
3689         nic->avail_msix_vectors = 0;
3690         ret = pci_enable_msix(nic->pdev, nic->entries, MAX_REQUESTED_MSI_X);
3691         /* We fail init if error or we get less vectors than min required */
3692         if (ret >= (nic->config.tx_fifo_num + nic->config.rx_ring_num + 1)) {
3693                 nic->avail_msix_vectors = ret;
3694                 ret = pci_enable_msix(nic->pdev, nic->entries, ret);
3695         }
3696         if (ret) {
3697                 DBG_PRINT(ERR_DBG, "%s: Enabling MSIX failed\n", nic->dev->name);
3698                 kfree(nic->entries);
3699                 kfree(nic->s2io_entries);
3700                 nic->entries = NULL;
3701                 nic->s2io_entries = NULL;
3702                 nic->avail_msix_vectors = 0;
3703                 return -ENOMEM;
3704         }
3705         if (!nic->avail_msix_vectors)
3706                 nic->avail_msix_vectors = MAX_REQUESTED_MSI_X;
3707
3708         /*
3709          * To enable MSI-X, MSI also needs to be enabled, due to a bug
3710          * in the herc NIC. (Temp change, needs to be removed later)
3711          */
3712         pci_read_config_word(nic->pdev, 0x42, &msi_control);
3713         msi_control |= 0x1; /* Enable MSI */
3714         pci_write_config_word(nic->pdev, 0x42, msi_control);
3715
3716         return 0;
3717 }
3718
3719 /* ********************************************************* *
3720  * Functions defined below concern the OS part of the driver *
3721  * ********************************************************* */
3722
3723 /**
3724  *  s2io_open - open entry point of the driver
3725  *  @dev : pointer to the device structure.
3726  *  Description:
3727  *  This function is the open entry point of the driver. It mainly calls a
3728  *  function to allocate Rx buffers and inserts them into the buffer
3729  *  descriptors and then enables the Rx part of the NIC.
3730  *  Return value:
3731  *  0 on success and an appropriate (-)ve integer as defined in errno.h
3732  *   file on failure.
3733  */
3734
3735 static int s2io_open(struct net_device *dev)
3736 {
3737         nic_t *sp = dev->priv;
3738         int err = 0;
3739
3740         /*
3741          * Make sure you have link off by default every time
3742          * Nic is initialized
3743          */
3744         netif_carrier_off(dev);
3745         sp->last_link_state = 0;
3746
3747         /* Initialize H/W and enable interrupts */
3748         err = s2io_card_up(sp);
3749         if (err) {
3750                 DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
3751                           dev->name);
3752                 if (err == -ENODEV)
3753                         goto hw_init_failed;
3754                 else
3755                         goto hw_enable_failed;
3756         }
3757
3758         /* Store the values of the MSIX table in the nic_t structure */
3759         store_xmsi_data(sp);
3760
3761         /* After proper initialization of H/W, register ISR */
3762         if (sp->intr_type == MSI) {
3763                 err = request_irq((int) sp->pdev->irq, s2io_msi_handle, 
3764                         SA_SHIRQ, sp->name, dev);
3765                 if (err) {
3766                         DBG_PRINT(ERR_DBG, "%s: MSI registration \
3767 failed\n", dev->name);
3768                         goto isr_registration_failed;
3769                 }
3770         }
3771         if (sp->intr_type == MSI_X) {
3772                 int i;
3773
3774                 for (i=1; (sp->s2io_entries[i].in_use == MSIX_FLG); i++) {
3775                         if (sp->s2io_entries[i].type == MSIX_FIFO_TYPE) {
3776                                 sprintf(sp->desc1, "%s:MSI-X-%d-TX",
3777                                         dev->name, i);
3778                                 err = request_irq(sp->entries[i].vector,
3779                                           s2io_msix_fifo_handle, 0, sp->desc1,
3780                                           sp->s2io_entries[i].arg);
3781                                 DBG_PRINT(ERR_DBG, "%s @ 0x%llx\n", sp->desc1, 
3782                                     (unsigned long long)sp->msix_info[i].addr);
3783                         } else {
3784                                 sprintf(sp->desc2, "%s:MSI-X-%d-RX",
3785                                         dev->name, i);
3786                                 err = request_irq(sp->entries[i].vector,
3787                                           s2io_msix_ring_handle, 0, sp->desc2,
3788                                           sp->s2io_entries[i].arg);
3789                                 DBG_PRINT(ERR_DBG, "%s @ 0x%llx\n", sp->desc2, 
3790                                      (unsigned long long)sp->msix_info[i].addr);
3791                         }
3792                         if (err) {
3793                                 DBG_PRINT(ERR_DBG, "%s: MSI-X-%d registration \
3794 failed\n", dev->name, i);
3795                                 DBG_PRINT(ERR_DBG, "Returned: %d\n", err);
3796                                 goto isr_registration_failed;
3797                         }
3798                         sp->s2io_entries[i].in_use = MSIX_REGISTERED_SUCCESS;
3799                 }
3800         }
3801         if (sp->intr_type == INTA) {
3802                 err = request_irq((int) sp->pdev->irq, s2io_isr, SA_SHIRQ,
3803                                 sp->name, dev);
3804                 if (err) {
3805                         DBG_PRINT(ERR_DBG, "%s: ISR registration failed\n",
3806                                   dev->name);
3807                         goto isr_registration_failed;
3808                 }
3809         }
3810
3811         if (s2io_set_mac_addr(dev, dev->dev_addr) == FAILURE) {
3812                 DBG_PRINT(ERR_DBG, "Set Mac Address Failed\n");
3813                 err = -ENODEV;
3814                 goto setting_mac_address_failed;
3815         }
3816
3817         netif_start_queue(dev);
3818         return 0;
3819
3820 setting_mac_address_failed:
3821         if (sp->intr_type != MSI_X)
3822                 free_irq(sp->pdev->irq, dev);
3823 isr_registration_failed:
3824         del_timer_sync(&sp->alarm_timer);
3825         if (sp->intr_type == MSI_X) {
3826                 int i;
3827                 u16 msi_control; /* Temp variable */
3828
3829                 for (i=1; (sp->s2io_entries[i].in_use == 
3830                                 MSIX_REGISTERED_SUCCESS); i++) {
3831                         int vector = sp->entries[i].vector;
3832                         void *arg = sp->s2io_entries[i].arg;
3833
3834                         free_irq(vector, arg);
3835                 }
3836                 pci_disable_msix(sp->pdev);
3837
3838                 /* Temp */
3839                 pci_read_config_word(sp->pdev, 0x42, &msi_control);
3840                 msi_control &= 0xFFFE; /* Disable MSI */
3841                 pci_write_config_word(sp->pdev, 0x42, msi_control);
3842         }
3843         else if (sp->intr_type == MSI)
3844                 pci_disable_msi(sp->pdev);
3845 hw_enable_failed:
3846         s2io_reset(sp);
3847 hw_init_failed:
3848         if (sp->intr_type == MSI_X) {
3849                 if (sp->entries)
3850                         kfree(sp->entries);
3851                 if (sp->s2io_entries)
3852                         kfree(sp->s2io_entries);
3853         }
3854         return err;
3855 }
3856
3857 /**
3858  *  s2io_close -close entry point of the driver
3859  *  @dev : device pointer.
3860  *  Description:
3861  *  This is the stop entry point of the driver. It needs to undo exactly
3862  *  whatever was done by the open entry point,thus it's usually referred to
3863  *  as the close function.Among other things this function mainly stops the
3864  *  Rx side of the NIC and frees all the Rx buffers in the Rx rings.
3865  *  Return value:
3866  *  0 on success and an appropriate (-)ve integer as defined in errno.h
3867  *  file on failure.
3868  */
3869
3870 static int s2io_close(struct net_device *dev)
3871 {
3872         nic_t *sp = dev->priv;
3873
3874         flush_scheduled_work();
3875         netif_stop_queue(dev);
3876         /* Reset card, kill tasklet and free Tx and Rx buffers. */
3877         s2io_card_down(sp, 1);
3878
3879         sp->device_close_flag = TRUE;   /* Device is shut down. */
3880         return 0;
3881 }
3882
3883 /**
3884  *  s2io_xmit - Tx entry point of te driver
3885  *  @skb : the socket buffer containing the Tx data.
3886  *  @dev : device pointer.
3887  *  Description :
3888  *  This function is the Tx entry point of the driver. S2IO NIC supports
3889  *  certain protocol assist features on Tx side, namely  CSO, S/G, LSO.
3890  *  NOTE: when device cant queue the pkt,just the trans_start variable will
3891  *  not be upadted.
3892  *  Return value:
3893  *  0 on success & 1 on failure.
3894  */
3895
3896 static int s2io_xmit(struct sk_buff *skb, struct net_device *dev)
3897 {
3898         nic_t *sp = dev->priv;
3899         u16 frg_cnt, frg_len, i, queue, queue_len, put_off, get_off;
3900         register u64 val64;
3901         TxD_t *txdp;
3902         TxFIFO_element_t __iomem *tx_fifo;
3903         unsigned long flags;
3904 #ifdef NETIF_F_TSO
3905         int mss;
3906 #endif
3907         u16 vlan_tag = 0;
3908         int vlan_priority = 0;
3909         mac_info_t *mac_control;
3910         struct config_param *config;
3911
3912         mac_control = &sp->mac_control;
3913         config = &sp->config;
3914
3915         DBG_PRINT(TX_DBG, "%s: In Neterion Tx routine\n", dev->name);
3916         spin_lock_irqsave(&sp->tx_lock, flags);
3917         if (atomic_read(&sp->card_state) == CARD_DOWN) {
3918                 DBG_PRINT(TX_DBG, "%s: Card going down for reset\n",
3919                           dev->name);
3920                 spin_unlock_irqrestore(&sp->tx_lock, flags);
3921                 dev_kfree_skb(skb);
3922                 return 0;
3923         }
3924
3925         queue = 0;
3926
3927         /* Get Fifo number to Transmit based on vlan priority */
3928         if (sp->vlgrp && vlan_tx_tag_present(skb)) {
3929                 vlan_tag = vlan_tx_tag_get(skb);
3930                 vlan_priority = vlan_tag >> 13;
3931                 queue = config->fifo_mapping[vlan_priority];
3932         }
3933
3934         put_off = (u16) mac_control->fifos[queue].tx_curr_put_info.offset;
3935         get_off = (u16) mac_control->fifos[queue].tx_curr_get_info.offset;
3936         txdp = (TxD_t *) mac_control->fifos[queue].list_info[put_off].
3937                 list_virt_addr;
3938
3939         queue_len = mac_control->fifos[queue].tx_curr_put_info.fifo_len + 1;
3940         /* Avoid "put" pointer going beyond "get" pointer */
3941         if (txdp->Host_Control ||
3942                    ((put_off+1) == queue_len ? 0 : (put_off+1)) == get_off) {
3943                 DBG_PRINT(TX_DBG, "Error in xmit, No free TXDs.\n");
3944                 netif_stop_queue(dev);
3945                 dev_kfree_skb(skb);
3946                 spin_unlock_irqrestore(&sp->tx_lock, flags);
3947                 return 0;
3948         }
3949
3950         /* A buffer with no data will be dropped */
3951         if (!skb->len) {
3952                 DBG_PRINT(TX_DBG, "%s:Buffer has no data..\n", dev->name);
3953                 dev_kfree_skb(skb);
3954                 spin_unlock_irqrestore(&sp->tx_lock, flags);
3955                 return 0;
3956         }
3957
3958         txdp->Control_1 = 0;
3959         txdp->Control_2 = 0;
3960 #ifdef NETIF_F_TSO
3961         mss = skb_shinfo(skb)->gso_size;
3962         if (skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)) {
3963                 txdp->Control_1 |= TXD_TCP_LSO_EN;
3964                 txdp->Control_1 |= TXD_TCP_LSO_MSS(mss);
3965         }
3966 #endif
3967         if (skb->ip_summed == CHECKSUM_HW) {
3968                 txdp->Control_2 |=
3969                     (TXD_TX_CKO_IPV4_EN | TXD_TX_CKO_TCP_EN |
3970                      TXD_TX_CKO_UDP_EN);
3971         }
3972         txdp->Control_1 |= TXD_GATHER_CODE_FIRST;
3973         txdp->Control_1 |= TXD_LIST_OWN_XENA;
3974         txdp->Control_2 |= config->tx_intr_type;
3975
3976         if (sp->vlgrp && vlan_tx_tag_present(skb)) {
3977                 txdp->Control_2 |= TXD_VLAN_ENABLE;
3978                 txdp->Control_2 |= TXD_VLAN_TAG(vlan_tag);
3979         }
3980
3981         frg_len = skb->len - skb->data_len;
3982         if (skb_shinfo(skb)->gso_type == SKB_GSO_UDP) {
3983                 int ufo_size;
3984
3985                 ufo_size = skb_shinfo(skb)->gso_size;
3986                 ufo_size &= ~7;
3987                 txdp->Control_1 |= TXD_UFO_EN;
3988                 txdp->Control_1 |= TXD_UFO_MSS(ufo_size);
3989                 txdp->Control_1 |= TXD_BUFFER0_SIZE(8);
3990 #ifdef __BIG_ENDIAN
3991                 sp->ufo_in_band_v[put_off] =
3992                                 (u64)skb_shinfo(skb)->ip6_frag_id;
3993 #else
3994                 sp->ufo_in_band_v[put_off] =
3995                                 (u64)skb_shinfo(skb)->ip6_frag_id << 32;
3996 #endif
3997                 txdp->Host_Control = (unsigned long)sp->ufo_in_band_v;
3998                 txdp->Buffer_Pointer = pci_map_single(sp->pdev,
3999                                         sp->ufo_in_band_v,
4000                                         sizeof(u64), PCI_DMA_TODEVICE);
4001                 txdp++;
4002                 txdp->Control_1 = 0;
4003                 txdp->Control_2 = 0;
4004         }
4005
4006         txdp->Buffer_Pointer = pci_map_single
4007             (sp->pdev, skb->data, frg_len, PCI_DMA_TODEVICE);
4008         txdp->Host_Control = (unsigned long) skb;
4009         txdp->Control_1 |= TXD_BUFFER0_SIZE(frg_len);
4010
4011         if (skb_shinfo(skb)->gso_type == SKB_GSO_UDP)
4012                 txdp->Control_1 |= TXD_UFO_EN;
4013
4014         frg_cnt = skb_shinfo(skb)->nr_frags;
4015         /* For fragmented SKB. */
4016         for (i = 0; i < frg_cnt; i++) {
4017                 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
4018                 /* A '0' length fragment will be ignored */
4019                 if (!frag->size)
4020                         continue;
4021                 txdp++;
4022                 txdp->Buffer_Pointer = (u64) pci_map_page
4023                     (sp->pdev, frag->page, frag->page_offset,
4024                      frag->size, PCI_DMA_TODEVICE);
4025                 txdp->Control_1 = TXD_BUFFER0_SIZE(frag->size);
4026                 if (skb_shinfo(skb)->gso_type == SKB_GSO_UDP)
4027                         txdp->Control_1 |= TXD_UFO_EN;
4028         }
4029         txdp->Control_1 |= TXD_GATHER_CODE_LAST;
4030
4031         if (skb_shinfo(skb)->gso_type == SKB_GSO_UDP)
4032                 frg_cnt++; /* as Txd0 was used for inband header */
4033
4034         tx_fifo = mac_control->tx_FIFO_start[queue];
4035         val64 = mac_control->fifos[queue].list_info[put_off].list_phy_addr;
4036         writeq(val64, &tx_fifo->TxDL_Pointer);
4037
4038         val64 = (TX_FIFO_LAST_TXD_NUM(frg_cnt) | TX_FIFO_FIRST_LIST |
4039                  TX_FIFO_LAST_LIST);
4040
4041 #ifdef NETIF_F_TSO
4042         if (mss)
4043                 val64 |= TX_FIFO_SPECIAL_FUNC;
4044 #endif
4045         if (skb_shinfo(skb)->gso_type == SKB_GSO_UDP)
4046                 val64 |= TX_FIFO_SPECIAL_FUNC;
4047         writeq(val64, &tx_fifo->List_Control);
4048
4049         mmiowb();
4050
4051         put_off++;
4052         if (put_off == mac_control->fifos[queue].tx_curr_put_info.fifo_len + 1)
4053                 put_off = 0;
4054         mac_control->fifos[queue].tx_curr_put_info.offset = put_off;
4055
4056         /* Avoid "put" pointer going beyond "get" pointer */
4057         if (((put_off+1) == queue_len ? 0 : (put_off+1)) == get_off) {
4058                 sp->mac_control.stats_info->sw_stat.fifo_full_cnt++;
4059                 DBG_PRINT(TX_DBG,
4060                           "No free TxDs for xmit, Put: 0x%x Get:0x%x\n",
4061                           put_off, get_off);
4062                 netif_stop_queue(dev);
4063         }
4064
4065         dev->trans_start = jiffies;
4066         spin_unlock_irqrestore(&sp->tx_lock, flags);
4067
4068         return 0;
4069 }
4070
4071 static void
4072 s2io_alarm_handle(unsigned long data)
4073 {
4074         nic_t *sp = (nic_t *)data;
4075
4076         alarm_intr_handler(sp);
4077         mod_timer(&sp->alarm_timer, jiffies + HZ / 2);
4078 }
4079
4080 static irqreturn_t
4081 s2io_msi_handle(int irq, void *dev_id, struct pt_regs *regs)
4082 {
4083         struct net_device *dev = (struct net_device *) dev_id;
4084         nic_t *sp = dev->priv;
4085         int i;
4086         int ret;
4087         mac_info_t *mac_control;
4088         struct config_param *config;
4089
4090         atomic_inc(&sp->isr_cnt);
4091         mac_control = &sp->mac_control;
4092         config = &sp->config;
4093         DBG_PRINT(INTR_DBG, "%s: MSI handler\n", __FUNCTION__);
4094
4095         /* If Intr is because of Rx Traffic */
4096         for (i = 0; i < config->rx_ring_num; i++)
4097                 rx_intr_handler(&mac_control->rings[i]);
4098
4099         /* If Intr is because of Tx Traffic */
4100         for (i = 0; i < config->tx_fifo_num; i++)
4101                 tx_intr_handler(&mac_control->fifos[i]);
4102
4103         /*
4104          * If the Rx buffer count is below the panic threshold then
4105          * reallocate the buffers from the interrupt handler itself,
4106          * else schedule a tasklet to reallocate the buffers.
4107          */
4108         for (i = 0; i < config->rx_ring_num; i++) {
4109                 if (!sp->lro) {
4110                         int rxb_size = atomic_read(&sp->rx_bufs_left[i]);
4111                         int level = rx_buffer_level(sp, rxb_size, i);
4112
4113                         if ((level == PANIC) && (!TASKLET_IN_USE)) {
4114                                 DBG_PRINT(INTR_DBG, "%s: Rx BD hit ", 
4115                                                         dev->name);
4116                                 DBG_PRINT(INTR_DBG, "PANIC levels\n");
4117                                 if ((ret = fill_rx_buffers(sp, i)) == -ENOMEM) {
4118                                         DBG_PRINT(ERR_DBG, "%s:Out of memory",
4119                                                   dev->name);
4120                                         DBG_PRINT(ERR_DBG, " in ISR!!\n");
4121                                         clear_bit(0, (&sp->tasklet_status));
4122                                         atomic_dec(&sp->isr_cnt);
4123                                         return IRQ_HANDLED;
4124                                 }
4125                                 clear_bit(0, (&sp->tasklet_status));
4126                         } else if (level == LOW) {
4127                                 tasklet_schedule(&sp->task);
4128                         }
4129                 }
4130                 else if (fill_rx_buffers(sp, i) == -ENOMEM) {
4131                                 DBG_PRINT(ERR_DBG, "%s:Out of memory",
4132                                                         dev->name);
4133                                 DBG_PRINT(ERR_DBG, " in Rx Intr!!\n");
4134                                 break;
4135                 }
4136         }
4137
4138         atomic_dec(&sp->isr_cnt);
4139         return IRQ_HANDLED;
4140 }
4141
4142 static irqreturn_t
4143 s2io_msix_ring_handle(int irq, void *dev_id, struct pt_regs *regs)
4144 {
4145         ring_info_t *ring = (ring_info_t *)dev_id;
4146         nic_t *sp = ring->nic;
4147         struct net_device *dev = (struct net_device *) dev_id;
4148         int rxb_size, level, rng_n;
4149
4150         atomic_inc(&sp->isr_cnt);
4151         rx_intr_handler(ring);
4152
4153         rng_n = ring->ring_no;
4154         if (!sp->lro) {
4155                 rxb_size = atomic_read(&sp->rx_bufs_left[rng_n]);
4156                 level = rx_buffer_level(sp, rxb_size, rng_n);
4157
4158                 if ((level == PANIC) && (!TASKLET_IN_USE)) {
4159                         int ret;
4160                         DBG_PRINT(INTR_DBG, "%s: Rx BD hit ", __FUNCTION__);
4161                         DBG_PRINT(INTR_DBG, "PANIC levels\n");
4162                         if ((ret = fill_rx_buffers(sp, rng_n)) == -ENOMEM) {
4163                                 DBG_PRINT(ERR_DBG, "Out of memory in %s",
4164                                           __FUNCTION__);
4165                                 clear_bit(0, (&sp->tasklet_status));
4166                                 return IRQ_HANDLED;
4167                         }
4168                         clear_bit(0, (&sp->tasklet_status));
4169                 } else if (level == LOW) {
4170                         tasklet_schedule(&sp->task);
4171                 }
4172         }
4173         else if (fill_rx_buffers(sp, rng_n) == -ENOMEM) {
4174                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
4175                         DBG_PRINT(ERR_DBG, " in Rx Intr!!\n");
4176         }
4177
4178         atomic_dec(&sp->isr_cnt);
4179
4180         return IRQ_HANDLED;
4181 }
4182
4183 static irqreturn_t
4184 s2io_msix_fifo_handle(int irq, void *dev_id, struct pt_regs *regs)
4185 {
4186         fifo_info_t *fifo = (fifo_info_t *)dev_id;
4187         nic_t *sp = fifo->nic;
4188
4189         atomic_inc(&sp->isr_cnt);
4190         tx_intr_handler(fifo);
4191         atomic_dec(&sp->isr_cnt);
4192         return IRQ_HANDLED;
4193 }
4194 static void s2io_txpic_intr_handle(nic_t *sp)
4195 {
4196         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4197         u64 val64;
4198
4199         val64 = readq(&bar0->pic_int_status);
4200         if (val64 & PIC_INT_GPIO) {
4201                 val64 = readq(&bar0->gpio_int_reg);
4202                 if ((val64 & GPIO_INT_REG_LINK_DOWN) &&
4203                     (val64 & GPIO_INT_REG_LINK_UP)) {
4204                         /*
4205                          * This is unstable state so clear both up/down
4206                          * interrupt and adapter to re-evaluate the link state.
4207                          */
4208                         val64 |=  GPIO_INT_REG_LINK_DOWN;
4209                         val64 |= GPIO_INT_REG_LINK_UP;
4210                         writeq(val64, &bar0->gpio_int_reg);
4211                         val64 = readq(&bar0->gpio_int_mask);
4212                         val64 &= ~(GPIO_INT_MASK_LINK_UP |
4213                                    GPIO_INT_MASK_LINK_DOWN);
4214                         writeq(val64, &bar0->gpio_int_mask);
4215                 }
4216                 else if (val64 & GPIO_INT_REG_LINK_UP) {
4217                         val64 = readq(&bar0->adapter_status);
4218                         if (verify_xena_quiescence(sp, val64,
4219                                                    sp->device_enabled_once)) {
4220                                 /* Enable Adapter */
4221                                 val64 = readq(&bar0->adapter_control);
4222                                 val64 |= ADAPTER_CNTL_EN;
4223                                 writeq(val64, &bar0->adapter_control);
4224                                 val64 |= ADAPTER_LED_ON;
4225                                 writeq(val64, &bar0->adapter_control);
4226                                 if (!sp->device_enabled_once)
4227                                         sp->device_enabled_once = 1;
4228
4229                                 s2io_link(sp, LINK_UP);
4230                                 /*
4231                                  * unmask link down interrupt and mask link-up
4232                                  * intr
4233                                  */
4234                                 val64 = readq(&bar0->gpio_int_mask);
4235                                 val64 &= ~GPIO_INT_MASK_LINK_DOWN;
4236                                 val64 |= GPIO_INT_MASK_LINK_UP;
4237                                 writeq(val64, &bar0->gpio_int_mask);
4238
4239                         }
4240                 }else if (val64 & GPIO_INT_REG_LINK_DOWN) {
4241                         val64 = readq(&bar0->adapter_status);
4242                         if (verify_xena_quiescence(sp, val64,
4243                                                    sp->device_enabled_once)) {
4244                                 s2io_link(sp, LINK_DOWN);
4245                                 /* Link is down so unmaks link up interrupt */
4246                                 val64 = readq(&bar0->gpio_int_mask);
4247                                 val64 &= ~GPIO_INT_MASK_LINK_UP;
4248                                 val64 |= GPIO_INT_MASK_LINK_DOWN;
4249                                 writeq(val64, &bar0->gpio_int_mask);
4250                         }
4251                 }
4252         }
4253         val64 = readq(&bar0->gpio_int_mask);
4254 }
4255
4256 /**
4257  *  s2io_isr - ISR handler of the device .
4258  *  @irq: the irq of the device.
4259  *  @dev_id: a void pointer to the dev structure of the NIC.
4260  *  @pt_regs: pointer to the registers pushed on the stack.
4261  *  Description:  This function is the ISR handler of the device. It
4262  *  identifies the reason for the interrupt and calls the relevant
4263  *  service routines. As a contongency measure, this ISR allocates the
4264  *  recv buffers, if their numbers are below the panic value which is
4265  *  presently set to 25% of the original number of rcv buffers allocated.
4266  *  Return value:
4267  *   IRQ_HANDLED: will be returned if IRQ was handled by this routine
4268  *   IRQ_NONE: will be returned if interrupt is not from our device
4269  */
4270 static irqreturn_t s2io_isr(int irq, void *dev_id, struct pt_regs *regs)
4271 {
4272         struct net_device *dev = (struct net_device *) dev_id;
4273         nic_t *sp = dev->priv;
4274         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4275         int i;
4276         u64 reason = 0, val64, org_mask;
4277         mac_info_t *mac_control;
4278         struct config_param *config;
4279
4280         atomic_inc(&sp->isr_cnt);
4281         mac_control = &sp->mac_control;
4282         config = &sp->config;
4283
4284         /*
4285          * Identify the cause for interrupt and call the appropriate
4286          * interrupt handler. Causes for the interrupt could be;
4287          * 1. Rx of packet.
4288          * 2. Tx complete.
4289          * 3. Link down.
4290          * 4. Error in any functional blocks of the NIC.
4291          */
4292         reason = readq(&bar0->general_int_status);
4293
4294         if (!reason) {
4295                 /* The interrupt was not raised by Xena. */
4296                 atomic_dec(&sp->isr_cnt);
4297                 return IRQ_NONE;
4298         }
4299
4300         val64 = 0xFFFFFFFFFFFFFFFFULL;
4301         /* Store current mask before masking all interrupts */
4302         org_mask = readq(&bar0->general_int_mask);
4303         writeq(val64, &bar0->general_int_mask);
4304
4305 #ifdef CONFIG_S2IO_NAPI
4306         if (reason & GEN_INTR_RXTRAFFIC) {
4307                 if (netif_rx_schedule_prep(dev)) {
4308                         writeq(val64, &bar0->rx_traffic_mask);
4309                         __netif_rx_schedule(dev);
4310                 }
4311         }
4312 #else
4313         /*
4314          * Rx handler is called by default, without checking for the
4315          * cause of interrupt.
4316          * rx_traffic_int reg is an R1 register, writing all 1's
4317          * will ensure that the actual interrupt causing bit get's
4318          * cleared and hence a read can be avoided.
4319          */
4320         writeq(val64, &bar0->rx_traffic_int);
4321         for (i = 0; i < config->rx_ring_num; i++) {
4322                 rx_intr_handler(&mac_control->rings[i]);
4323         }
4324 #endif
4325
4326         /*
4327          * tx_traffic_int reg is an R1 register, writing all 1's
4328          * will ensure that the actual interrupt causing bit get's
4329          * cleared and hence a read can be avoided.
4330          */
4331         writeq(val64, &bar0->tx_traffic_int);
4332
4333         for (i = 0; i < config->tx_fifo_num; i++)
4334                 tx_intr_handler(&mac_control->fifos[i]);
4335
4336         if (reason & GEN_INTR_TXPIC)
4337                 s2io_txpic_intr_handle(sp);
4338         /*
4339          * If the Rx buffer count is below the panic threshold then
4340          * reallocate the buffers from the interrupt handler itself,
4341          * else schedule a tasklet to reallocate the buffers.
4342          */
4343 #ifndef CONFIG_S2IO_NAPI
4344         for (i = 0; i < config->rx_ring_num; i++) {
4345                 if (!sp->lro) {
4346                         int ret;
4347                         int rxb_size = atomic_read(&sp->rx_bufs_left[i]);
4348                         int level = rx_buffer_level(sp, rxb_size, i);
4349
4350                         if ((level == PANIC) && (!TASKLET_IN_USE)) {
4351                                 DBG_PRINT(INTR_DBG, "%s: Rx BD hit ", 
4352                                                         dev->name);
4353                                 DBG_PRINT(INTR_DBG, "PANIC levels\n");
4354                                 if ((ret = fill_rx_buffers(sp, i)) == -ENOMEM) {
4355                                         DBG_PRINT(ERR_DBG, "%s:Out of memory",
4356                                                   dev->name);
4357                                         DBG_PRINT(ERR_DBG, " in ISR!!\n");
4358                                         clear_bit(0, (&sp->tasklet_status));
4359                                         atomic_dec(&sp->isr_cnt);
4360                                         writeq(org_mask, &bar0->general_int_mask);
4361                                         return IRQ_HANDLED;
4362                                 }
4363                                 clear_bit(0, (&sp->tasklet_status));
4364                         } else if (level == LOW) {
4365                                 tasklet_schedule(&sp->task);
4366                         }
4367                 }
4368                 else if (fill_rx_buffers(sp, i) == -ENOMEM) {
4369                                 DBG_PRINT(ERR_DBG, "%s:Out of memory",
4370                                                         dev->name);
4371                                 DBG_PRINT(ERR_DBG, " in Rx intr!!\n");
4372                                 break;
4373                 }
4374         }
4375 #endif
4376         writeq(org_mask, &bar0->general_int_mask);
4377         atomic_dec(&sp->isr_cnt);
4378         return IRQ_HANDLED;
4379 }
4380
4381 /**
4382  * s2io_updt_stats -
4383  */
4384 static void s2io_updt_stats(nic_t *sp)
4385 {
4386         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4387         u64 val64;
4388         int cnt = 0;
4389
4390         if (atomic_read(&sp->card_state) == CARD_UP) {
4391                 /* Apprx 30us on a 133 MHz bus */
4392                 val64 = SET_UPDT_CLICKS(10) |
4393                         STAT_CFG_ONE_SHOT_EN | STAT_CFG_STAT_EN;
4394                 writeq(val64, &bar0->stat_cfg);
4395                 do {
4396                         udelay(100);
4397                         val64 = readq(&bar0->stat_cfg);
4398                         if (!(val64 & BIT(0)))
4399                                 break;
4400                         cnt++;
4401                         if (cnt == 5)
4402                                 break; /* Updt failed */
4403                 } while(1);
4404         }
4405 }
4406
4407 /**
4408  *  s2io_get_stats - Updates the device statistics structure.
4409  *  @dev : pointer to the device structure.
4410  *  Description:
4411  *  This function updates the device statistics structure in the s2io_nic
4412  *  structure and returns a pointer to the same.
4413  *  Return value:
4414  *  pointer to the updated net_device_stats structure.
4415  */
4416
4417 static struct net_device_stats *s2io_get_stats(struct net_device *dev)
4418 {
4419         nic_t *sp = dev->priv;
4420         mac_info_t *mac_control;
4421         struct config_param *config;
4422
4423
4424         mac_control = &sp->mac_control;
4425         config = &sp->config;
4426
4427         /* Configure Stats for immediate updt */
4428         s2io_updt_stats(sp);
4429
4430         sp->stats.tx_packets =
4431                 le32_to_cpu(mac_control->stats_info->tmac_frms);
4432         sp->stats.tx_errors =
4433                 le32_to_cpu(mac_control->stats_info->tmac_any_err_frms);
4434         sp->stats.rx_errors =
4435                 le32_to_cpu(mac_control->stats_info->rmac_drop_frms);
4436         sp->stats.multicast =
4437                 le32_to_cpu(mac_control->stats_info->rmac_vld_mcst_frms);
4438         sp->stats.rx_length_errors =
4439                 le32_to_cpu(mac_control->stats_info->rmac_long_frms);
4440
4441         return (&sp->stats);
4442 }
4443
4444 /**
4445  *  s2io_set_multicast - entry point for multicast address enable/disable.
4446  *  @dev : pointer to the device structure
4447  *  Description:
4448  *  This function is a driver entry point which gets called by the kernel
4449  *  whenever multicast addresses must be enabled/disabled. This also gets
4450  *  called to set/reset promiscuous mode. Depending on the deivce flag, we
4451  *  determine, if multicast address must be enabled or if promiscuous mode
4452  *  is to be disabled etc.
4453  *  Return value:
4454  *  void.
4455  */
4456
4457 static void s2io_set_multicast(struct net_device *dev)
4458 {
4459         int i, j, prev_cnt;
4460         struct dev_mc_list *mclist;
4461         nic_t *sp = dev->priv;
4462         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4463         u64 val64 = 0, multi_mac = 0x010203040506ULL, mask =
4464             0xfeffffffffffULL;
4465         u64 dis_addr = 0xffffffffffffULL, mac_addr = 0;
4466         void __iomem *add;
4467
4468         if ((dev->flags & IFF_ALLMULTI) && (!sp->m_cast_flg)) {
4469                 /*  Enable all Multicast addresses */
4470                 writeq(RMAC_ADDR_DATA0_MEM_ADDR(multi_mac),
4471                        &bar0->rmac_addr_data0_mem);
4472                 writeq(RMAC_ADDR_DATA1_MEM_MASK(mask),
4473                        &bar0->rmac_addr_data1_mem);
4474                 val64 = RMAC_ADDR_CMD_MEM_WE |
4475                     RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4476                     RMAC_ADDR_CMD_MEM_OFFSET(MAC_MC_ALL_MC_ADDR_OFFSET);
4477                 writeq(val64, &bar0->rmac_addr_cmd_mem);
4478                 /* Wait till command completes */
4479                 wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4480                                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING);
4481
4482                 sp->m_cast_flg = 1;
4483                 sp->all_multi_pos = MAC_MC_ALL_MC_ADDR_OFFSET;
4484         } else if ((dev->flags & IFF_ALLMULTI) && (sp->m_cast_flg)) {
4485                 /*  Disable all Multicast addresses */
4486                 writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
4487                        &bar0->rmac_addr_data0_mem);
4488                 writeq(RMAC_ADDR_DATA1_MEM_MASK(0x0),
4489                        &bar0->rmac_addr_data1_mem);
4490                 val64 = RMAC_ADDR_CMD_MEM_WE |
4491                     RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4492                     RMAC_ADDR_CMD_MEM_OFFSET(sp->all_multi_pos);
4493                 writeq(val64, &bar0->rmac_addr_cmd_mem);
4494                 /* Wait till command completes */
4495                 wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4496                                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING);
4497
4498                 sp->m_cast_flg = 0;
4499                 sp->all_multi_pos = 0;
4500         }
4501
4502         if ((dev->flags & IFF_PROMISC) && (!sp->promisc_flg)) {
4503                 /*  Put the NIC into promiscuous mode */
4504                 add = &bar0->mac_cfg;
4505                 val64 = readq(&bar0->mac_cfg);
4506                 val64 |= MAC_CFG_RMAC_PROM_ENABLE;
4507
4508                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4509                 writel((u32) val64, add);
4510                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4511                 writel((u32) (val64 >> 32), (add + 4));
4512
4513                 val64 = readq(&bar0->mac_cfg);
4514                 sp->promisc_flg = 1;
4515                 DBG_PRINT(INFO_DBG, "%s: entered promiscuous mode\n",
4516                           dev->name);
4517         } else if (!(dev->flags & IFF_PROMISC) && (sp->promisc_flg)) {
4518                 /*  Remove the NIC from promiscuous mode */
4519                 add = &bar0->mac_cfg;
4520                 val64 = readq(&bar0->mac_cfg);
4521                 val64 &= ~MAC_CFG_RMAC_PROM_ENABLE;
4522
4523                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4524                 writel((u32) val64, add);
4525                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4526                 writel((u32) (val64 >> 32), (add + 4));
4527
4528                 val64 = readq(&bar0->mac_cfg);
4529                 sp->promisc_flg = 0;
4530                 DBG_PRINT(INFO_DBG, "%s: left promiscuous mode\n",
4531                           dev->name);
4532         }
4533
4534         /*  Update individual M_CAST address list */
4535         if ((!sp->m_cast_flg) && dev->mc_count) {
4536                 if (dev->mc_count >
4537                     (MAX_ADDRS_SUPPORTED - MAC_MC_ADDR_START_OFFSET - 1)) {
4538                         DBG_PRINT(ERR_DBG, "%s: No more Rx filters ",
4539                                   dev->name);
4540                         DBG_PRINT(ERR_DBG, "can be added, please enable ");
4541                         DBG_PRINT(ERR_DBG, "ALL_MULTI instead\n");
4542                         return;
4543                 }
4544
4545                 prev_cnt = sp->mc_addr_count;
4546                 sp->mc_addr_count = dev->mc_count;
4547
4548                 /* Clear out the previous list of Mc in the H/W. */
4549                 for (i = 0; i < prev_cnt; i++) {
4550                         writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
4551                                &bar0->rmac_addr_data0_mem);
4552                         writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
4553                                 &bar0->rmac_addr_data1_mem);
4554                         val64 = RMAC_ADDR_CMD_MEM_WE |
4555                             RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4556                             RMAC_ADDR_CMD_MEM_OFFSET
4557                             (MAC_MC_ADDR_START_OFFSET + i);
4558                         writeq(val64, &bar0->rmac_addr_cmd_mem);
4559
4560                         /* Wait for command completes */
4561                         if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4562                                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING)) {
4563                                 DBG_PRINT(ERR_DBG, "%s: Adding ",
4564                                           dev->name);
4565                                 DBG_PRINT(ERR_DBG, "Multicasts failed\n");
4566                                 return;
4567                         }
4568                 }
4569
4570                 /* Create the new Rx filter list and update the same in H/W. */
4571                 for (i = 0, mclist = dev->mc_list; i < dev->mc_count;
4572                      i++, mclist = mclist->next) {
4573                         memcpy(sp->usr_addrs[i].addr, mclist->dmi_addr,
4574                                ETH_ALEN);
4575                         mac_addr = 0;
4576                         for (j = 0; j < ETH_ALEN; j++) {
4577                                 mac_addr |= mclist->dmi_addr[j];
4578                                 mac_addr <<= 8;
4579                         }
4580                         mac_addr >>= 8;
4581                         writeq(RMAC_ADDR_DATA0_MEM_ADDR(mac_addr),
4582                                &bar0->rmac_addr_data0_mem);
4583                         writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
4584                                 &bar0->rmac_addr_data1_mem);
4585                         val64 = RMAC_ADDR_CMD_MEM_WE |
4586                             RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4587                             RMAC_ADDR_CMD_MEM_OFFSET
4588                             (i + MAC_MC_ADDR_START_OFFSET);
4589                         writeq(val64, &bar0->rmac_addr_cmd_mem);
4590
4591                         /* Wait for command completes */
4592                         if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4593                                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING)) {
4594                                 DBG_PRINT(ERR_DBG, "%s: Adding ",
4595                                           dev->name);
4596                                 DBG_PRINT(ERR_DBG, "Multicasts failed\n");
4597                                 return;
4598                         }
4599                 }
4600         }
4601 }
4602
4603 /**
4604  *  s2io_set_mac_addr - Programs the Xframe mac address
4605  *  @dev : pointer to the device structure.
4606  *  @addr: a uchar pointer to the new mac address which is to be set.
4607  *  Description : This procedure will program the Xframe to receive
4608  *  frames with new Mac Address
4609  *  Return value: SUCCESS on success and an appropriate (-)ve integer
4610  *  as defined in errno.h file on failure.
4611  */
4612
4613 static int s2io_set_mac_addr(struct net_device *dev, u8 * addr)
4614 {
4615         nic_t *sp = dev->priv;
4616         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4617         register u64 val64, mac_addr = 0;
4618         int i;
4619
4620         /*
4621          * Set the new MAC address as the new unicast filter and reflect this
4622          * change on the device address registered with the OS. It will be
4623          * at offset 0.
4624          */
4625         for (i = 0; i < ETH_ALEN; i++) {
4626                 mac_addr <<= 8;
4627                 mac_addr |= addr[i];
4628         }
4629
4630         writeq(RMAC_ADDR_DATA0_MEM_ADDR(mac_addr),
4631                &bar0->rmac_addr_data0_mem);
4632
4633         val64 =
4634             RMAC_ADDR_CMD_MEM_WE | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4635             RMAC_ADDR_CMD_MEM_OFFSET(0);
4636         writeq(val64, &bar0->rmac_addr_cmd_mem);
4637         /* Wait till command completes */
4638         if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4639                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING)) {
4640                 DBG_PRINT(ERR_DBG, "%s: set_mac_addr failed\n", dev->name);
4641                 return FAILURE;
4642         }
4643
4644         return SUCCESS;
4645 }
4646
4647 /**
4648  * s2io_ethtool_sset - Sets different link parameters.
4649  * @sp : private member of the device structure, which is a pointer to the  * s2io_nic structure.
4650  * @info: pointer to the structure with parameters given by ethtool to set
4651  * link information.
4652  * Description:
4653  * The function sets different link parameters provided by the user onto
4654  * the NIC.
4655  * Return value:
4656  * 0 on success.
4657 */
4658
4659 static int s2io_ethtool_sset(struct net_device *dev,
4660                              struct ethtool_cmd *info)
4661 {
4662         nic_t *sp = dev->priv;
4663         if ((info->autoneg == AUTONEG_ENABLE) ||
4664             (info->speed != SPEED_10000) || (info->duplex != DUPLEX_FULL))
4665                 return -EINVAL;
4666         else {
4667                 s2io_close(sp->dev);
4668                 s2io_open(sp->dev);
4669         }
4670
4671         return 0;
4672 }
4673
4674 /**
4675  * s2io_ethtol_gset - Return link specific information.
4676  * @sp : private member of the device structure, pointer to the
4677  *      s2io_nic structure.
4678  * @info : pointer to the structure with parameters given by ethtool
4679  * to return link information.
4680  * Description:
4681  * Returns link specific information like speed, duplex etc.. to ethtool.
4682  * Return value :
4683  * return 0 on success.
4684  */
4685
4686 static int s2io_ethtool_gset(struct net_device *dev, struct ethtool_cmd *info)
4687 {
4688         nic_t *sp = dev->priv;
4689         info->supported = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
4690         info->advertising = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
4691         info->port = PORT_FIBRE;
4692         /* info->transceiver?? TODO */
4693
4694         if (netif_carrier_ok(sp->dev)) {
4695                 info->speed = 10000;
4696                 info->duplex = DUPLEX_FULL;
4697         } else {
4698                 info->speed = -1;
4699                 info->duplex = -1;
4700         }
4701
4702         info->autoneg = AUTONEG_DISABLE;
4703         return 0;
4704 }
4705
4706 /**
4707  * s2io_ethtool_gdrvinfo - Returns driver specific information.
4708  * @sp : private member of the device structure, which is a pointer to the
4709  * s2io_nic structure.
4710  * @info : pointer to the structure with parameters given by ethtool to
4711  * return driver information.
4712  * Description:
4713  * Returns driver specefic information like name, version etc.. to ethtool.
4714  * Return value:
4715  *  void
4716  */
4717
4718 static void s2io_ethtool_gdrvinfo(struct net_device *dev,
4719                                   struct ethtool_drvinfo *info)
4720 {
4721         nic_t *sp = dev->priv;
4722
4723         strncpy(info->driver, s2io_driver_name, sizeof(info->driver));
4724         strncpy(info->version, s2io_driver_version, sizeof(info->version));
4725         strncpy(info->fw_version, "", sizeof(info->fw_version));
4726         strncpy(info->bus_info, pci_name(sp->pdev), sizeof(info->bus_info));
4727         info->regdump_len = XENA_REG_SPACE;
4728         info->eedump_len = XENA_EEPROM_SPACE;
4729         info->testinfo_len = S2IO_TEST_LEN;
4730         info->n_stats = S2IO_STAT_LEN;
4731 }
4732
4733 /**
4734  *  s2io_ethtool_gregs - dumps the entire space of Xfame into the buffer.
4735  *  @sp: private member of the device structure, which is a pointer to the
4736  *  s2io_nic structure.
4737  *  @regs : pointer to the structure with parameters given by ethtool for
4738  *  dumping the registers.
4739  *  @reg_space: The input argumnet into which all the registers are dumped.
4740  *  Description:
4741  *  Dumps the entire register space of xFrame NIC into the user given
4742  *  buffer area.
4743  * Return value :
4744  * void .
4745 */
4746
4747 static void s2io_ethtool_gregs(struct net_device *dev,
4748                                struct ethtool_regs *regs, void *space)
4749 {
4750         int i;
4751         u64 reg;
4752         u8 *reg_space = (u8 *) space;
4753         nic_t *sp = dev->priv;
4754
4755         regs->len = XENA_REG_SPACE;
4756         regs->version = sp->pdev->subsystem_device;
4757
4758         for (i = 0; i < regs->len; i += 8) {
4759                 reg = readq(sp->bar0 + i);
4760                 memcpy((reg_space + i), &reg, 8);
4761         }
4762 }
4763
4764 /**
4765  *  s2io_phy_id  - timer function that alternates adapter LED.
4766  *  @data : address of the private member of the device structure, which
4767  *  is a pointer to the s2io_nic structure, provided as an u32.
4768  * Description: This is actually the timer function that alternates the
4769  * adapter LED bit of the adapter control bit to set/reset every time on
4770  * invocation. The timer is set for 1/2 a second, hence tha NIC blinks
4771  *  once every second.
4772 */
4773 static void s2io_phy_id(unsigned long data)
4774 {
4775         nic_t *sp = (nic_t *) data;
4776         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4777         u64 val64 = 0;
4778         u16 subid;
4779
4780         subid = sp->pdev->subsystem_device;
4781         if ((sp->device_type == XFRAME_II_DEVICE) ||
4782                    ((subid & 0xFF) >= 0x07)) {
4783                 val64 = readq(&bar0->gpio_control);
4784                 val64 ^= GPIO_CTRL_GPIO_0;
4785                 writeq(val64, &bar0->gpio_control);
4786         } else {
4787                 val64 = readq(&bar0->adapter_control);
4788                 val64 ^= ADAPTER_LED_ON;
4789                 writeq(val64, &bar0->adapter_control);
4790         }
4791
4792         mod_timer(&sp->id_timer, jiffies + HZ / 2);
4793 }
4794
4795 /**
4796  * s2io_ethtool_idnic - To physically identify the nic on the system.
4797  * @sp : private member of the device structure, which is a pointer to the
4798  * s2io_nic structure.
4799  * @id : pointer to the structure with identification parameters given by
4800  * ethtool.
4801  * Description: Used to physically identify the NIC on the system.
4802  * The Link LED will blink for a time specified by the user for
4803  * identification.
4804  * NOTE: The Link has to be Up to be able to blink the LED. Hence
4805  * identification is possible only if it's link is up.
4806  * Return value:
4807  * int , returns 0 on success
4808  */
4809
4810 static int s2io_ethtool_idnic(struct net_device *dev, u32 data)
4811 {
4812         u64 val64 = 0, last_gpio_ctrl_val;
4813         nic_t *sp = dev->priv;
4814         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4815         u16 subid;
4816
4817         subid = sp->pdev->subsystem_device;
4818         last_gpio_ctrl_val = readq(&bar0->gpio_control);
4819         if ((sp->device_type == XFRAME_I_DEVICE) &&
4820                 ((subid & 0xFF) < 0x07)) {
4821                 val64 = readq(&bar0->adapter_control);
4822                 if (!(val64 & ADAPTER_CNTL_EN)) {
4823                         printk(KERN_ERR
4824                                "Adapter Link down, cannot blink LED\n");
4825                         return -EFAULT;
4826                 }
4827         }
4828         if (sp->id_timer.function == NULL) {
4829                 init_timer(&sp->id_timer);
4830                 sp->id_timer.function = s2io_phy_id;
4831                 sp->id_timer.data = (unsigned long) sp;
4832         }
4833         mod_timer(&sp->id_timer, jiffies);
4834         if (data)
4835                 msleep_interruptible(data * HZ);
4836         else
4837                 msleep_interruptible(MAX_FLICKER_TIME);
4838         del_timer_sync(&sp->id_timer);
4839
4840         if (CARDS_WITH_FAULTY_LINK_INDICATORS(sp->device_type, subid)) {
4841                 writeq(last_gpio_ctrl_val, &bar0->gpio_control);
4842                 last_gpio_ctrl_val = readq(&bar0->gpio_control);
4843         }
4844
4845         return 0;
4846 }
4847
4848 /**
4849  * s2io_ethtool_getpause_data -Pause frame frame generation and reception.
4850  * @sp : private member of the device structure, which is a pointer to the
4851  *      s2io_nic structure.
4852  * @ep : pointer to the structure with pause parameters given by ethtool.
4853  * Description:
4854  * Returns the Pause frame generation and reception capability of the NIC.
4855  * Return value:
4856  *  void
4857  */
4858 static void s2io_ethtool_getpause_data(struct net_device *dev,
4859                                        struct ethtool_pauseparam *ep)
4860 {
4861         u64 val64;
4862         nic_t *sp = dev->priv;
4863         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4864
4865         val64 = readq(&bar0->rmac_pause_cfg);
4866         if (val64 & RMAC_PAUSE_GEN_ENABLE)
4867                 ep->tx_pause = TRUE;
4868         if (val64 & RMAC_PAUSE_RX_ENABLE)
4869                 ep->rx_pause = TRUE;
4870         ep->autoneg = FALSE;
4871 }
4872
4873 /**
4874  * s2io_ethtool_setpause_data -  set/reset pause frame generation.
4875  * @sp : private member of the device structure, which is a pointer to the
4876  *      s2io_nic structure.
4877  * @ep : pointer to the structure with pause parameters given by ethtool.
4878  * Description:
4879  * It can be used to set or reset Pause frame generation or reception
4880  * support of the NIC.
4881  * Return value:
4882  * int, returns 0 on Success
4883  */
4884
4885 static int s2io_ethtool_setpause_data(struct net_device *dev,
4886                                struct ethtool_pauseparam *ep)
4887 {
4888         u64 val64;
4889         nic_t *sp = dev->priv;
4890         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4891
4892         val64 = readq(&bar0->rmac_pause_cfg);
4893         if (ep->tx_pause)
4894                 val64 |= RMAC_PAUSE_GEN_ENABLE;
4895         else
4896                 val64 &= ~RMAC_PAUSE_GEN_ENABLE;
4897         if (ep->rx_pause)
4898                 val64 |= RMAC_PAUSE_RX_ENABLE;
4899         else
4900                 val64 &= ~RMAC_PAUSE_RX_ENABLE;
4901         writeq(val64, &bar0->rmac_pause_cfg);
4902         return 0;
4903 }
4904
4905 /**
4906  * read_eeprom - reads 4 bytes of data from user given offset.
4907  * @sp : private member of the device structure, which is a pointer to the
4908  *      s2io_nic structure.
4909  * @off : offset at which the data must be written
4910  * @data : Its an output parameter where the data read at the given
4911  *      offset is stored.
4912  * Description:
4913  * Will read 4 bytes of data from the user given offset and return the
4914  * read data.
4915  * NOTE: Will allow to read only part of the EEPROM visible through the
4916  *   I2C bus.
4917  * Return value:
4918  *  -1 on failure and 0 on success.
4919  */
4920
4921 #define S2IO_DEV_ID             5
4922 static int read_eeprom(nic_t * sp, int off, u64 * data)
4923 {
4924         int ret = -1;
4925         u32 exit_cnt = 0;
4926         u64 val64;
4927         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4928
4929         if (sp->device_type == XFRAME_I_DEVICE) {
4930                 val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) | I2C_CONTROL_ADDR(off) |
4931                     I2C_CONTROL_BYTE_CNT(0x3) | I2C_CONTROL_READ |
4932                     I2C_CONTROL_CNTL_START;
4933                 SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
4934
4935                 while (exit_cnt < 5) {
4936                         val64 = readq(&bar0->i2c_control);
4937                         if (I2C_CONTROL_CNTL_END(val64)) {
4938                                 *data = I2C_CONTROL_GET_DATA(val64);
4939                                 ret = 0;
4940                                 break;
4941                         }
4942                         msleep(50);
4943                         exit_cnt++;
4944                 }
4945         }
4946
4947         if (sp->device_type == XFRAME_II_DEVICE) {
4948                 val64 = SPI_CONTROL_KEY(0x9) | SPI_CONTROL_SEL1 |
4949                         SPI_CONTROL_BYTECNT(0x3) | 
4950                         SPI_CONTROL_CMD(0x3) | SPI_CONTROL_ADDR(off);
4951                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4952                 val64 |= SPI_CONTROL_REQ;
4953                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4954                 while (exit_cnt < 5) {
4955                         val64 = readq(&bar0->spi_control);
4956                         if (val64 & SPI_CONTROL_NACK) {
4957                                 ret = 1;
4958                                 break;
4959                         } else if (val64 & SPI_CONTROL_DONE) {
4960                                 *data = readq(&bar0->spi_data);
4961                                 *data &= 0xffffff;
4962                                 ret = 0;
4963                                 break;
4964                         }
4965                         msleep(50);
4966                         exit_cnt++;
4967                 }
4968         }
4969         return ret;
4970 }
4971
4972 /**
4973  *  write_eeprom - actually writes the relevant part of the data value.
4974  *  @sp : private member of the device structure, which is a pointer to the
4975  *       s2io_nic structure.
4976  *  @off : offset at which the data must be written
4977  *  @data : The data that is to be written
4978  *  @cnt : Number of bytes of the data that are actually to be written into
4979  *  the Eeprom. (max of 3)
4980  * Description:
4981  *  Actually writes the relevant part of the data value into the Eeprom
4982  *  through the I2C bus.
4983  * Return value:
4984  *  0 on success, -1 on failure.
4985  */
4986
4987 static int write_eeprom(nic_t * sp, int off, u64 data, int cnt)
4988 {
4989         int exit_cnt = 0, ret = -1;
4990         u64 val64;
4991         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4992
4993         if (sp->device_type == XFRAME_I_DEVICE) {
4994                 val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) | I2C_CONTROL_ADDR(off) |
4995                     I2C_CONTROL_BYTE_CNT(cnt) | I2C_CONTROL_SET_DATA((u32)data) |
4996                     I2C_CONTROL_CNTL_START;
4997                 SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
4998
4999                 while (exit_cnt < 5) {
5000                         val64 = readq(&bar0->i2c_control);
5001                         if (I2C_CONTROL_CNTL_END(val64)) {
5002                                 if (!(val64 & I2C_CONTROL_NACK))
5003                                         ret = 0;
5004                                 break;
5005                         }
5006                         msleep(50);
5007                         exit_cnt++;
5008                 }
5009         }
5010
5011         if (sp->device_type == XFRAME_II_DEVICE) {
5012                 int write_cnt = (cnt == 8) ? 0 : cnt;
5013                 writeq(SPI_DATA_WRITE(data,(cnt<<3)), &bar0->spi_data);
5014
5015                 val64 = SPI_CONTROL_KEY(0x9) | SPI_CONTROL_SEL1 |
5016                         SPI_CONTROL_BYTECNT(write_cnt) | 
5017                         SPI_CONTROL_CMD(0x2) | SPI_CONTROL_ADDR(off);
5018                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
5019                 val64 |= SPI_CONTROL_REQ;
5020                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
5021                 while (exit_cnt < 5) {
5022                         val64 = readq(&bar0->spi_control);
5023                         if (val64 & SPI_CONTROL_NACK) {
5024                                 ret = 1;
5025                                 break;
5026                         } else if (val64 & SPI_CONTROL_DONE) {
5027                                 ret = 0;
5028                                 break;
5029                         }
5030                         msleep(50);
5031                         exit_cnt++;
5032                 }
5033         }
5034         return ret;
5035 }
5036 static void s2io_vpd_read(nic_t *nic)
5037 {
5038         u8 vpd_data[256],data;
5039         int i=0, cnt, fail = 0;
5040         int vpd_addr = 0x80;
5041
5042         if (nic->device_type == XFRAME_II_DEVICE) {
5043                 strcpy(nic->product_name, "Xframe II 10GbE network adapter");
5044                 vpd_addr = 0x80;
5045         }
5046         else {
5047                 strcpy(nic->product_name, "Xframe I 10GbE network adapter");
5048                 vpd_addr = 0x50;
5049         }
5050
5051         for (i = 0; i < 256; i +=4 ) {
5052                 pci_write_config_byte(nic->pdev, (vpd_addr + 2), i);
5053                 pci_read_config_byte(nic->pdev,  (vpd_addr + 2), &data);
5054                 pci_write_config_byte(nic->pdev, (vpd_addr + 3), 0);
5055                 for (cnt = 0; cnt <5; cnt++) {
5056                         msleep(2);
5057                         pci_read_config_byte(nic->pdev, (vpd_addr + 3), &data);
5058                         if (data == 0x80)
5059                                 break;
5060                 }
5061                 if (cnt >= 5) {
5062                         DBG_PRINT(ERR_DBG, "Read of VPD data failed\n");
5063                         fail = 1;
5064                         break;
5065                 }
5066                 pci_read_config_dword(nic->pdev,  (vpd_addr + 4),
5067                                       (u32 *)&vpd_data[i]);
5068         }
5069         if ((!fail) && (vpd_data[1] < VPD_PRODUCT_NAME_LEN)) {
5070                 memset(nic->product_name, 0, vpd_data[1]);
5071                 memcpy(nic->product_name, &vpd_data[3], vpd_data[1]);
5072         }
5073 }
5074
5075 /**
5076  *  s2io_ethtool_geeprom  - reads the value stored in the Eeprom.
5077  *  @sp : private member of the device structure, which is a pointer to the *       s2io_nic structure.
5078  *  @eeprom : pointer to the user level structure provided by ethtool,
5079  *  containing all relevant information.
5080  *  @data_buf : user defined value to be written into Eeprom.
5081  *  Description: Reads the values stored in the Eeprom at given offset
5082  *  for a given length. Stores these values int the input argument data
5083  *  buffer 'data_buf' and returns these to the caller (ethtool.)
5084  *  Return value:
5085  *  int  0 on success
5086  */
5087
5088 static int s2io_ethtool_geeprom(struct net_device *dev,
5089                          struct ethtool_eeprom *eeprom, u8 * data_buf)
5090 {
5091         u32 i, valid;
5092         u64 data;
5093         nic_t *sp = dev->priv;
5094
5095         eeprom->magic = sp->pdev->vendor | (sp->pdev->device << 16);
5096
5097         if ((eeprom->offset + eeprom->len) > (XENA_EEPROM_SPACE))
5098                 eeprom->len = XENA_EEPROM_SPACE - eeprom->offset;
5099
5100         for (i = 0; i < eeprom->len; i += 4) {
5101                 if (read_eeprom(sp, (eeprom->offset + i), &data)) {
5102                         DBG_PRINT(ERR_DBG, "Read of EEPROM failed\n");
5103                         return -EFAULT;
5104                 }
5105                 valid = INV(data);
5106                 memcpy((data_buf + i), &valid, 4);
5107         }
5108         return 0;
5109 }
5110
5111 /**
5112  *  s2io_ethtool_seeprom - tries to write the user provided value in Eeprom
5113  *  @sp : private member of the device structure, which is a pointer to the
5114  *  s2io_nic structure.
5115  *  @eeprom : pointer to the user level structure provided by ethtool,
5116  *  containing all relevant information.
5117  *  @data_buf ; user defined value to be written into Eeprom.
5118  *  Description:
5119  *  Tries to write the user provided value in the Eeprom, at the offset
5120  *  given by the user.
5121  *  Return value:
5122  *  0 on success, -EFAULT on failure.
5123  */
5124
5125 static int s2io_ethtool_seeprom(struct net_device *dev,
5126                                 struct ethtool_eeprom *eeprom,
5127                                 u8 * data_buf)
5128 {
5129         int len = eeprom->len, cnt = 0;
5130         u64 valid = 0, data;
5131         nic_t *sp = dev->priv;
5132
5133         if (eeprom->magic != (sp->pdev->vendor | (sp->pdev->device << 16))) {
5134                 DBG_PRINT(ERR_DBG,
5135                           "ETHTOOL_WRITE_EEPROM Err: Magic value ");
5136                 DBG_PRINT(ERR_DBG, "is wrong, Its not 0x%x\n",
5137                           eeprom->magic);
5138                 return -EFAULT;
5139         }
5140
5141         while (len) {
5142                 data = (u32) data_buf[cnt] & 0x000000FF;
5143                 if (data) {
5144                         valid = (u32) (data << 24);
5145                 } else
5146                         valid = data;
5147
5148                 if (write_eeprom(sp, (eeprom->offset + cnt), valid, 0)) {
5149                         DBG_PRINT(ERR_DBG,
5150                                   "ETHTOOL_WRITE_EEPROM Err: Cannot ");
5151                         DBG_PRINT(ERR_DBG,
5152                                   "write into the specified offset\n");
5153                         return -EFAULT;
5154                 }
5155                 cnt++;
5156                 len--;
5157         }
5158
5159         return 0;
5160 }
5161
5162 /**
5163  * s2io_register_test - reads and writes into all clock domains.
5164  * @sp : private member of the device structure, which is a pointer to the
5165  * s2io_nic structure.
5166  * @data : variable that returns the result of each of the test conducted b
5167  * by the driver.
5168  * Description:
5169  * Read and write into all clock domains. The NIC has 3 clock domains,
5170  * see that registers in all the three regions are accessible.
5171  * Return value:
5172  * 0 on success.
5173  */
5174
5175 static int s2io_register_test(nic_t * sp, uint64_t * data)
5176 {
5177         XENA_dev_config_t __iomem *bar0 = sp->bar0;
5178         u64 val64 = 0, exp_val;
5179         int fail = 0;
5180
5181         val64 = readq(&bar0->pif_rd_swapper_fb);
5182         if (val64 != 0x123456789abcdefULL) {
5183                 fail = 1;
5184                 DBG_PRINT(INFO_DBG, "Read Test level 1 fails\n");
5185         }
5186
5187         val64 = readq(&bar0->rmac_pause_cfg);
5188         if (val64 != 0xc000ffff00000000ULL) {
5189                 fail = 1;
5190                 DBG_PRINT(INFO_DBG, "Read Test level 2 fails\n");
5191         }
5192
5193         val64 = readq(&bar0->rx_queue_cfg);
5194         if (sp->device_type == XFRAME_II_DEVICE)
5195                 exp_val = 0x0404040404040404ULL;
5196         else
5197                 exp_val = 0x0808080808080808ULL;
5198         if (val64 != exp_val) {
5199                 fail = 1;
5200                 DBG_PRINT(INFO_DBG, "Read Test level 3 fails\n");
5201         }
5202
5203         val64 = readq(&bar0->xgxs_efifo_cfg);
5204         if (val64 != 0x000000001923141EULL) {
5205                 fail = 1;
5206                 DBG_PRINT(INFO_DBG, "Read Test level 4 fails\n");
5207         }
5208
5209         val64 = 0x5A5A5A5A5A5A5A5AULL;
5210         writeq(val64, &bar0->xmsi_data);
5211         val64 = readq(&bar0->xmsi_data);
5212         if (val64 != 0x5A5A5A5A5A5A5A5AULL) {
5213                 fail = 1;
5214                 DBG_PRINT(ERR_DBG, "Write Test level 1 fails\n");
5215         }
5216
5217         val64 = 0xA5A5A5A5A5A5A5A5ULL;
5218         writeq(val64, &bar0->xmsi_data);
5219         val64 = readq(&bar0->xmsi_data);
5220         if (val64 != 0xA5A5A5A5A5A5A5A5ULL) {
5221                 fail = 1;
5222                 DBG_PRINT(ERR_DBG, "Write Test level 2 fails\n");
5223         }
5224
5225         *data = fail;
5226         return fail;
5227 }
5228
5229 /**
5230  * s2io_eeprom_test - to verify that EEprom in the xena can be programmed.
5231  * @sp : private member of the device structure, which is a pointer to the
5232  * s2io_nic structure.
5233  * @data:variable that returns the result of each of the test conducted by
5234  * the driver.
5235  * Description:
5236  * Verify that EEPROM in the xena can be programmed using I2C_CONTROL
5237  * register.
5238  * Return value:
5239  * 0 on success.
5240  */
5241
5242 static int s2io_eeprom_test(nic_t * sp, uint64_t * data)
5243 {
5244         int fail = 0;
5245         u64 ret_data, org_4F0, org_7F0;
5246         u8 saved_4F0 = 0, saved_7F0 = 0;
5247         struct net_device *dev = sp->dev;
5248
5249         /* Test Write Error at offset 0 */
5250         /* Note that SPI interface allows write access to all areas
5251          * of EEPROM. Hence doing all negative testing only for Xframe I.
5252          */
5253         if (sp->device_type == XFRAME_I_DEVICE)
5254                 if (!write_eeprom(sp, 0, 0, 3))
5255                         fail = 1;
5256
5257         /* Save current values at offsets 0x4F0 and 0x7F0 */
5258         if (!read_eeprom(sp, 0x4F0, &org_4F0))
5259                 saved_4F0 = 1;
5260         if (!read_eeprom(sp, 0x7F0, &org_7F0))
5261                 saved_7F0 = 1;
5262
5263         /* Test Write at offset 4f0 */
5264         if (write_eeprom(sp, 0x4F0, 0x012345, 3))
5265                 fail = 1;
5266         if (read_eeprom(sp, 0x4F0, &ret_data))
5267                 fail = 1;
5268
5269         if (ret_data != 0x012345) {
5270                 DBG_PRINT(ERR_DBG, "%s: eeprom test error at offset 0x4F0. "
5271                         "Data written %llx Data read %llx\n",
5272                         dev->name, (unsigned long long)0x12345,
5273                         (unsigned long long)ret_data);
5274                 fail = 1;
5275         }
5276
5277         /* Reset the EEPROM data go FFFF */
5278         write_eeprom(sp, 0x4F0, 0xFFFFFF, 3);
5279
5280         /* Test Write Request Error at offset 0x7c */
5281         if (sp->device_type == XFRAME_I_DEVICE)
5282                 if (!write_eeprom(sp, 0x07C, 0, 3))
5283                         fail = 1;
5284
5285         /* Test Write Request at offset 0x7f0 */
5286         if (write_eeprom(sp, 0x7F0, 0x012345, 3))
5287                 fail = 1;
5288         if (read_eeprom(sp, 0x7F0, &ret_data))
5289                 fail = 1;
5290
5291         if (ret_data != 0x012345) {
5292                 DBG_PRINT(ERR_DBG, "%s: eeprom test error at offset 0x7F0. "
5293                         "Data written %llx Data read %llx\n",
5294                         dev->name, (unsigned long long)0x12345,
5295                         (unsigned long long)ret_data);
5296                 fail = 1;
5297         }
5298
5299         /* Reset the EEPROM data go FFFF */
5300         write_eeprom(sp, 0x7F0, 0xFFFFFF, 3);
5301
5302         if (sp->device_type == XFRAME_I_DEVICE) {
5303                 /* Test Write Error at offset 0x80 */
5304                 if (!write_eeprom(sp, 0x080, 0, 3))
5305                         fail = 1;
5306
5307                 /* Test Write Error at offset 0xfc */
5308                 if (!write_eeprom(sp, 0x0FC, 0, 3))
5309                         fail = 1;
5310
5311                 /* Test Write Error at offset 0x100 */
5312                 if (!write_eeprom(sp, 0x100, 0, 3))
5313                         fail = 1;
5314
5315                 /* Test Write Error at offset 4ec */
5316                 if (!write_eeprom(sp, 0x4EC, 0, 3))
5317                         fail = 1;
5318         }
5319
5320         /* Restore values at offsets 0x4F0 and 0x7F0 */
5321         if (saved_4F0)
5322                 write_eeprom(sp, 0x4F0, org_4F0, 3);
5323         if (saved_7F0)
5324                 write_eeprom(sp, 0x7F0, org_7F0, 3);
5325
5326         *data = fail;
5327         return fail;
5328 }
5329
5330 /**
5331  * s2io_bist_test - invokes the MemBist test of the card .
5332  * @sp : private member of the device structure, which is a pointer to the
5333  * s2io_nic structure.
5334  * @data:variable that returns the result of each of the test conducted by
5335  * the driver.
5336  * Description:
5337  * This invokes the MemBist test of the card. We give around
5338  * 2 secs time for the Test to complete. If it's still not complete
5339  * within this peiod, we consider that the test failed.
5340  * Return value:
5341  * 0 on success and -1 on failure.
5342  */
5343
5344 static int s2io_bist_test(nic_t * sp, uint64_t * data)
5345 {
5346         u8 bist = 0;
5347         int cnt = 0, ret = -1;
5348
5349         pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
5350         bist |= PCI_BIST_START;
5351         pci_write_config_word(sp->pdev, PCI_BIST, bist);
5352
5353         while (cnt < 20) {
5354                 pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
5355                 if (!(bist & PCI_BIST_START)) {
5356                         *data = (bist & PCI_BIST_CODE_MASK);
5357                         ret = 0;
5358                         break;
5359                 }
5360                 msleep(100);
5361                 cnt++;
5362         }
5363
5364         return ret;
5365 }
5366
5367 /**
5368  * s2io-link_test - verifies the link state of the nic
5369  * @sp ; private member of the device structure, which is a pointer to the
5370  * s2io_nic structure.
5371  * @data: variable that returns the result of each of the test conducted by
5372  * the driver.
5373  * Description:
5374  * The function verifies the link state of the NIC and updates the input
5375  * argument 'data' appropriately.
5376  * Return value:
5377  * 0 on success.
5378  */
5379
5380 static int s2io_link_test(nic_t * sp, uint64_t * data)
5381 {
5382         XENA_dev_config_t __iomem *bar0 = sp->bar0;
5383         u64 val64;
5384
5385         val64 = readq(&bar0->adapter_status);
5386         if(!(LINK_IS_UP(val64)))
5387                 *data = 1;
5388         else
5389                 *data = 0;
5390
5391         return 0;
5392 }
5393
5394 /**
5395  * s2io_rldram_test - offline test for access to the RldRam chip on the NIC
5396  * @sp - private member of the device structure, which is a pointer to the
5397  * s2io_nic structure.
5398  * @data - variable that returns the result of each of the test
5399  * conducted by the driver.
5400  * Description:
5401  *  This is one of the offline test that tests the read and write
5402  *  access to the RldRam chip on the NIC.
5403  * Return value:
5404  *  0 on success.
5405  */
5406
5407 static int s2io_rldram_test(nic_t * sp, uint64_t * data)
5408 {
5409         XENA_dev_config_t __iomem *bar0 = sp->bar0;
5410         u64 val64;
5411         int cnt, iteration = 0, test_fail = 0;
5412
5413         val64 = readq(&bar0->adapter_control);
5414         val64 &= ~ADAPTER_ECC_EN;
5415         writeq(val64, &bar0->adapter_control);
5416
5417         val64 = readq(&bar0->mc_rldram_test_ctrl);
5418         val64 |= MC_RLDRAM_TEST_MODE;
5419         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
5420
5421         val64 = readq(&bar0->mc_rldram_mrs);
5422         val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE;
5423         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
5424
5425         val64 |= MC_RLDRAM_MRS_ENABLE;
5426         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
5427
5428         while (iteration < 2) {
5429                 val64 = 0x55555555aaaa0000ULL;
5430                 if (iteration == 1) {
5431                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
5432                 }
5433                 writeq(val64, &bar0->mc_rldram_test_d0);
5434
5435                 val64 = 0xaaaa5a5555550000ULL;
5436                 if (iteration == 1) {
5437                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
5438                 }
5439                 writeq(val64, &bar0->mc_rldram_test_d1);
5440
5441                 val64 = 0x55aaaaaaaa5a0000ULL;
5442                 if (iteration == 1) {
5443                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
5444                 }
5445                 writeq(val64, &bar0->mc_rldram_test_d2);
5446
5447                 val64 = (u64) (0x0000003ffffe0100ULL);
5448                 writeq(val64, &bar0->mc_rldram_test_add);
5449
5450                 val64 = MC_RLDRAM_TEST_MODE | MC_RLDRAM_TEST_WRITE |
5451                         MC_RLDRAM_TEST_GO;
5452                 SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
5453
5454                 for (cnt = 0; cnt < 5; cnt++) {
5455                         val64 = readq(&bar0->mc_rldram_test_ctrl);
5456                         if (val64 & MC_RLDRAM_TEST_DONE)
5457                                 break;
5458                         msleep(200);
5459                 }
5460
5461                 if (cnt == 5)
5462                         break;
5463
5464                 val64 = MC_RLDRAM_TEST_MODE | MC_RLDRAM_TEST_GO;
5465                 SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
5466
5467                 for (cnt = 0; cnt < 5; cnt++) {
5468                         val64 = readq(&bar0->mc_rldram_test_ctrl);
5469                         if (val64 & MC_RLDRAM_TEST_DONE)
5470                                 break;
5471                         msleep(500);
5472                 }
5473
5474                 if (cnt == 5)
5475                         break;
5476
5477                 val64 = readq(&bar0->mc_rldram_test_ctrl);
5478                 if (!(val64 & MC_RLDRAM_TEST_PASS))
5479                         test_fail = 1;
5480
5481                 iteration++;
5482         }
5483
5484         *data = test_fail;
5485
5486         /* Bring the adapter out of test mode */
5487         SPECIAL_REG_WRITE(0, &bar0->mc_rldram_test_ctrl, LF);
5488
5489         return test_fail;
5490 }
5491
5492 /**
5493  *  s2io_ethtool_test - conducts 6 tsets to determine the health of card.
5494  *  @sp : private member of the device structure, which is a pointer to the
5495  *  s2io_nic structure.
5496  *  @ethtest : pointer to a ethtool command specific structure that will be
5497  *  returned to the user.
5498  *  @data : variable that returns the result of each of the test
5499  * conducted by the driver.
5500  * Description:
5501  *  This function conducts 6 tests ( 4 offline and 2 online) to determine
5502  *  the health of the card.
5503  * Return value:
5504  *  void
5505  */
5506
5507 static void s2io_ethtool_test(struct net_device *dev,
5508                               struct ethtool_test *ethtest,
5509                               uint64_t * data)
5510 {
5511         nic_t *sp = dev->priv;
5512         int orig_state = netif_running(sp->dev);
5513
5514         if (ethtest->flags == ETH_TEST_FL_OFFLINE) {
5515                 /* Offline Tests. */
5516                 if (orig_state)
5517                         s2io_close(sp->dev);
5518
5519                 if (s2io_register_test(sp, &data[0]))
5520                         ethtest->flags |= ETH_TEST_FL_FAILED;
5521
5522                 s2io_reset(sp);
5523
5524                 if (s2io_rldram_test(sp, &data[3]))
5525                         ethtest->flags |= ETH_TEST_FL_FAILED;
5526
5527                 s2io_reset(sp);
5528
5529                 if (s2io_eeprom_test(sp, &data[1]))
5530                         ethtest->flags |= ETH_TEST_FL_FAILED;
5531
5532                 if (s2io_bist_test(sp, &data[4]))
5533                         ethtest->flags |= ETH_TEST_FL_FAILED;
5534
5535                 if (orig_state)
5536                         s2io_open(sp->dev);
5537
5538                 data[2] = 0;
5539         } else {
5540                 /* Online Tests. */
5541                 if (!orig_state) {
5542                         DBG_PRINT(ERR_DBG,
5543                                   "%s: is not up, cannot run test\n",
5544                                   dev->name);
5545                         data[0] = -1;
5546                         data[1] = -1;
5547                         data[2] = -1;
5548                         data[3] = -1;
5549                         data[4] = -1;
5550                 }
5551
5552                 if (s2io_link_test(sp, &data[2]))
5553                         ethtest->flags |= ETH_TEST_FL_FAILED;
5554
5555                 data[0] = 0;
5556                 data[1] = 0;
5557                 data[3] = 0;
5558                 data[4] = 0;
5559         }
5560 }
5561
5562 static void s2io_get_ethtool_stats(struct net_device *dev,
5563                                    struct ethtool_stats *estats,
5564                                    u64 * tmp_stats)
5565 {
5566         int i = 0;
5567         nic_t *sp = dev->priv;
5568         StatInfo_t *stat_info = sp->mac_control.stats_info;
5569
5570         s2io_updt_stats(sp);
5571         tmp_stats[i++] =
5572                 (u64)le32_to_cpu(stat_info->tmac_frms_oflow) << 32  |
5573                 le32_to_cpu(stat_info->tmac_frms);
5574         tmp_stats[i++] =
5575                 (u64)le32_to_cpu(stat_info->tmac_data_octets_oflow) << 32 |
5576                 le32_to_cpu(stat_info->tmac_data_octets);
5577         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_drop_frms);
5578         tmp_stats[i++] =
5579                 (u64)le32_to_cpu(stat_info->tmac_mcst_frms_oflow) << 32 |
5580                 le32_to_cpu(stat_info->tmac_mcst_frms);
5581         tmp_stats[i++] =
5582                 (u64)le32_to_cpu(stat_info->tmac_bcst_frms_oflow) << 32 |
5583                 le32_to_cpu(stat_info->tmac_bcst_frms);
5584         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_pause_ctrl_frms);
5585         tmp_stats[i++] =
5586                 (u64)le32_to_cpu(stat_info->tmac_ttl_octets_oflow) << 32 |
5587                 le32_to_cpu(stat_info->tmac_ttl_octets);
5588         tmp_stats[i++] =
5589                 (u64)le32_to_cpu(stat_info->tmac_ucst_frms_oflow) << 32 |
5590                 le32_to_cpu(stat_info->tmac_ucst_frms);
5591         tmp_stats[i++] =
5592                 (u64)le32_to_cpu(stat_info->tmac_nucst_frms_oflow) << 32 |
5593                 le32_to_cpu(stat_info->tmac_nucst_frms);
5594         tmp_stats[i++] =
5595                 (u64)le32_to_cpu(stat_info->tmac_any_err_frms_oflow) << 32 |
5596                 le32_to_cpu(stat_info->tmac_any_err_frms);
5597         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_ttl_less_fb_octets);
5598         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_vld_ip_octets);
5599         tmp_stats[i++] =
5600                 (u64)le32_to_cpu(stat_info->tmac_vld_ip_oflow) << 32 |
5601                 le32_to_cpu(stat_info->tmac_vld_ip);
5602         tmp_stats[i++] =
5603                 (u64)le32_to_cpu(stat_info->tmac_drop_ip_oflow) << 32 |
5604                 le32_to_cpu(stat_info->tmac_drop_ip);
5605         tmp_stats[i++] =
5606                 (u64)le32_to_cpu(stat_info->tmac_icmp_oflow) << 32 |
5607                 le32_to_cpu(stat_info->tmac_icmp);
5608         tmp_stats[i++] =
5609                 (u64)le32_to_cpu(stat_info->tmac_rst_tcp_oflow) << 32 |
5610                 le32_to_cpu(stat_info->tmac_rst_tcp);
5611         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_tcp);
5612         tmp_stats[i++] = (u64)le32_to_cpu(stat_info->tmac_udp_oflow) << 32 |
5613                 le32_to_cpu(stat_info->tmac_udp);
5614         tmp_stats[i++] =
5615                 (u64)le32_to_cpu(stat_info->rmac_vld_frms_oflow) << 32 |
5616                 le32_to_cpu(stat_info->rmac_vld_frms);
5617         tmp_stats[i++] =
5618                 (u64)le32_to_cpu(stat_info->rmac_data_octets_oflow) << 32 |
5619                 le32_to_cpu(stat_info->rmac_data_octets);
5620         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_fcs_err_frms);
5621         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_drop_frms);
5622         tmp_stats[i++] =
5623                 (u64)le32_to_cpu(stat_info->rmac_vld_mcst_frms_oflow) << 32 |
5624                 le32_to_cpu(stat_info->rmac_vld_mcst_frms);
5625         tmp_stats[i++] =
5626                 (u64)le32_to_cpu(stat_info->rmac_vld_bcst_frms_oflow) << 32 |
5627                 le32_to_cpu(stat_info->rmac_vld_bcst_frms);
5628         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_in_rng_len_err_frms);
5629         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_out_rng_len_err_frms);
5630         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_long_frms);
5631         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_pause_ctrl_frms);
5632         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_unsup_ctrl_frms);
5633         tmp_stats[i++] =
5634                 (u64)le32_to_cpu(stat_info->rmac_ttl_octets_oflow) << 32 |
5635                 le32_to_cpu(stat_info->rmac_ttl_octets);
5636         tmp_stats[i++] =
5637                 (u64)le32_to_cpu(stat_info->rmac_accepted_ucst_frms_oflow)
5638                 << 32 | le32_to_cpu(stat_info->rmac_accepted_ucst_frms);
5639         tmp_stats[i++] =
5640                 (u64)le32_to_cpu(stat_info->rmac_accepted_nucst_frms_oflow)
5641                  << 32 | le32_to_cpu(stat_info->rmac_accepted_nucst_frms);
5642         tmp_stats[i++] =
5643                 (u64)le32_to_cpu(stat_info->rmac_discarded_frms_oflow) << 32 |
5644                 le32_to_cpu(stat_info->rmac_discarded_frms);
5645         tmp_stats[i++] =
5646                 (u64)le32_to_cpu(stat_info->rmac_drop_events_oflow)
5647                  << 32 | le32_to_cpu(stat_info->rmac_drop_events);
5648         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_less_fb_octets);
5649         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_frms);
5650         tmp_stats[i++] =
5651                 (u64)le32_to_cpu(stat_info->rmac_usized_frms_oflow) << 32 |
5652                 le32_to_cpu(stat_info->rmac_usized_frms);
5653         tmp_stats[i++] =
5654                 (u64)le32_to_cpu(stat_info->rmac_osized_frms_oflow) << 32 |
5655                 le32_to_cpu(stat_info->rmac_osized_frms);
5656         tmp_stats[i++] =
5657                 (u64)le32_to_cpu(stat_info->rmac_frag_frms_oflow) << 32 |
5658                 le32_to_cpu(stat_info->rmac_frag_frms);
5659         tmp_stats[i++] =
5660                 (u64)le32_to_cpu(stat_info->rmac_jabber_frms_oflow) << 32 |
5661                 le32_to_cpu(stat_info->rmac_jabber_frms);
5662         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_64_frms);
5663         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_65_127_frms);
5664         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_128_255_frms);
5665         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_256_511_frms);
5666         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_512_1023_frms);
5667         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_1024_1518_frms);
5668         tmp_stats[i++] =
5669                 (u64)le32_to_cpu(stat_info->rmac_ip_oflow) << 32 |
5670                 le32_to_cpu(stat_info->rmac_ip);
5671         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ip_octets);
5672         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_hdr_err_ip);
5673         tmp_stats[i++] =
5674                 (u64)le32_to_cpu(stat_info->rmac_drop_ip_oflow) << 32 |
5675                 le32_to_cpu(stat_info->rmac_drop_ip);
5676         tmp_stats[i++] =
5677                 (u64)le32_to_cpu(stat_info->rmac_icmp_oflow) << 32 |
5678                 le32_to_cpu(stat_info->rmac_icmp);
5679         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_tcp);
5680         tmp_stats[i++] =
5681                 (u64)le32_to_cpu(stat_info->rmac_udp_oflow) << 32 |
5682                 le32_to_cpu(stat_info->rmac_udp);
5683         tmp_stats[i++] =
5684                 (u64)le32_to_cpu(stat_info->rmac_err_drp_udp_oflow) << 32 |
5685                 le32_to_cpu(stat_info->rmac_err_drp_udp);
5686         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_xgmii_err_sym);
5687         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q0);
5688         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q1);
5689         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q2);
5690         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q3);
5691         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q4);
5692         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q5);
5693         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q6);
5694         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_frms_q7);
5695         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q0);
5696         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q1);
5697         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q2);
5698         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q3);
5699         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q4);
5700         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q5);
5701         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q6);
5702         tmp_stats[i++] = le16_to_cpu(stat_info->rmac_full_q7);
5703         tmp_stats[i++] =
5704                 (u64)le32_to_cpu(stat_info->rmac_pause_cnt_oflow) << 32 |
5705                 le32_to_cpu(stat_info->rmac_pause_cnt);
5706         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_xgmii_data_err_cnt);
5707         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_xgmii_ctrl_err_cnt);
5708         tmp_stats[i++] =
5709                 (u64)le32_to_cpu(stat_info->rmac_accepted_ip_oflow) << 32 |
5710                 le32_to_cpu(stat_info->rmac_accepted_ip);
5711         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_err_tcp);
5712         tmp_stats[i++] = le32_to_cpu(stat_info->rd_req_cnt);
5713         tmp_stats[i++] = le32_to_cpu(stat_info->new_rd_req_cnt);
5714         tmp_stats[i++] = le32_to_cpu(stat_info->new_rd_req_rtry_cnt);
5715         tmp_stats[i++] = le32_to_cpu(stat_info->rd_rtry_cnt);
5716         tmp_stats[i++] = le32_to_cpu(stat_info->wr_rtry_rd_ack_cnt);
5717         tmp_stats[i++] = le32_to_cpu(stat_info->wr_req_cnt);
5718         tmp_stats[i++] = le32_to_cpu(stat_info->new_wr_req_cnt);
5719         tmp_stats[i++] = le32_to_cpu(stat_info->new_wr_req_rtry_cnt);
5720         tmp_stats[i++] = le32_to_cpu(stat_info->wr_rtry_cnt);
5721         tmp_stats[i++] = le32_to_cpu(stat_info->wr_disc_cnt);
5722         tmp_stats[i++] = le32_to_cpu(stat_info->rd_rtry_wr_ack_cnt);
5723         tmp_stats[i++] = le32_to_cpu(stat_info->txp_wr_cnt);
5724         tmp_stats[i++] = le32_to_cpu(stat_info->txd_rd_cnt);
5725         tmp_stats[i++] = le32_to_cpu(stat_info->txd_wr_cnt);
5726         tmp_stats[i++] = le32_to_cpu(stat_info->rxd_rd_cnt);
5727         tmp_stats[i++] = le32_to_cpu(stat_info->rxd_wr_cnt);
5728         tmp_stats[i++] = le32_to_cpu(stat_info->txf_rd_cnt);
5729         tmp_stats[i++] = le32_to_cpu(stat_info->rxf_wr_cnt);
5730         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_1519_4095_frms);
5731         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_4096_8191_frms);
5732         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_8192_max_frms);
5733         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ttl_gt_max_frms);
5734         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_osized_alt_frms);
5735         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_jabber_alt_frms);
5736         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_gt_max_alt_frms);
5737         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_vlan_frms);
5738         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_len_discard);
5739         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_fcs_discard);
5740         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_pf_discard);
5741         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_da_discard);
5742         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_red_discard);
5743         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_rts_discard);
5744         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_ingm_full_discard);
5745         tmp_stats[i++] = le32_to_cpu(stat_info->link_fault_cnt);
5746         tmp_stats[i++] = 0;
5747         tmp_stats[i++] = stat_info->sw_stat.single_ecc_errs;
5748         tmp_stats[i++] = stat_info->sw_stat.double_ecc_errs;
5749         tmp_stats[i++] = stat_info->sw_stat.parity_err_cnt;
5750         tmp_stats[i++] = stat_info->sw_stat.serious_err_cnt;
5751         tmp_stats[i++] = stat_info->sw_stat.soft_reset_cnt;
5752         tmp_stats[i++] = stat_info->sw_stat.fifo_full_cnt;
5753         tmp_stats[i++] = stat_info->sw_stat.ring_full_cnt;
5754         tmp_stats[i++] = stat_info->xpak_stat.alarm_transceiver_temp_high;
5755         tmp_stats[i++] = stat_info->xpak_stat.alarm_transceiver_temp_low;
5756         tmp_stats[i++] = stat_info->xpak_stat.alarm_laser_bias_current_high;
5757         tmp_stats[i++] = stat_info->xpak_stat.alarm_laser_bias_current_low;
5758         tmp_stats[i++] = stat_info->xpak_stat.alarm_laser_output_power_high;
5759         tmp_stats[i++] = stat_info->xpak_stat.alarm_laser_output_power_low;
5760         tmp_stats[i++] = stat_info->xpak_stat.warn_transceiver_temp_high;
5761         tmp_stats[i++] = stat_info->xpak_stat.warn_transceiver_temp_low;
5762         tmp_stats[i++] = stat_info->xpak_stat.warn_laser_bias_current_high;
5763         tmp_stats[i++] = stat_info->xpak_stat.warn_laser_bias_current_low;
5764         tmp_stats[i++] = stat_info->xpak_stat.warn_laser_output_power_high;
5765         tmp_stats[i++] = stat_info->xpak_stat.warn_laser_output_power_low;
5766         tmp_stats[i++] = stat_info->sw_stat.clubbed_frms_cnt;
5767         tmp_stats[i++] = stat_info->sw_stat.sending_both;
5768         tmp_stats[i++] = stat_info->sw_stat.outof_sequence_pkts;
5769         tmp_stats[i++] = stat_info->sw_stat.flush_max_pkts;
5770         if (stat_info->sw_stat.num_aggregations) {
5771                 u64 tmp = stat_info->sw_stat.sum_avg_pkts_aggregated;
5772                 int count = 0;
5773                 /* 
5774                  * Since 64-bit divide does not work on all platforms,
5775                  * do repeated subtraction.
5776                  */
5777                 while (tmp >= stat_info->sw_stat.num_aggregations) {
5778                         tmp -= stat_info->sw_stat.num_aggregations;
5779                         count++;
5780                 }
5781                 tmp_stats[i++] = count;
5782         }
5783         else
5784                 tmp_stats[i++] = 0;
5785 }
5786
5787 static int s2io_ethtool_get_regs_len(struct net_device *dev)
5788 {
5789         return (XENA_REG_SPACE);
5790 }
5791
5792
5793 static u32 s2io_ethtool_get_rx_csum(struct net_device * dev)
5794 {
5795         nic_t *sp = dev->priv;
5796
5797         return (sp->rx_csum);
5798 }
5799
5800 static int s2io_ethtool_set_rx_csum(struct net_device *dev, u32 data)
5801 {
5802         nic_t *sp = dev->priv;
5803
5804         if (data)
5805                 sp->rx_csum = 1;
5806         else
5807                 sp->rx_csum = 0;
5808
5809         return 0;
5810 }
5811
5812 static int s2io_get_eeprom_len(struct net_device *dev)
5813 {
5814         return (XENA_EEPROM_SPACE);
5815 }
5816
5817 static int s2io_ethtool_self_test_count(struct net_device *dev)
5818 {
5819         return (S2IO_TEST_LEN);
5820 }
5821
5822 static void s2io_ethtool_get_strings(struct net_device *dev,
5823                                      u32 stringset, u8 * data)
5824 {
5825         switch (stringset) {
5826         case ETH_SS_TEST:
5827                 memcpy(data, s2io_gstrings, S2IO_STRINGS_LEN);
5828                 break;
5829         case ETH_SS_STATS:
5830                 memcpy(data, &ethtool_stats_keys,
5831                        sizeof(ethtool_stats_keys));
5832         }
5833 }
5834 static int s2io_ethtool_get_stats_count(struct net_device *dev)
5835 {
5836         return (S2IO_STAT_LEN);
5837 }
5838
5839 static int s2io_ethtool_op_set_tx_csum(struct net_device *dev, u32 data)
5840 {
5841         if (data)
5842                 dev->features |= NETIF_F_IP_CSUM;
5843         else
5844                 dev->features &= ~NETIF_F_IP_CSUM;
5845
5846         return 0;
5847 }
5848
5849
5850 static struct ethtool_ops netdev_ethtool_ops = {
5851         .get_settings = s2io_ethtool_gset,
5852         .set_settings = s2io_ethtool_sset,
5853         .get_drvinfo = s2io_ethtool_gdrvinfo,
5854         .get_regs_len = s2io_ethtool_get_regs_len,
5855         .get_regs = s2io_ethtool_gregs,
5856         .get_link = ethtool_op_get_link,
5857         .get_eeprom_len = s2io_get_eeprom_len,
5858         .get_eeprom = s2io_ethtool_geeprom,
5859         .set_eeprom = s2io_ethtool_seeprom,
5860         .get_pauseparam = s2io_ethtool_getpause_data,
5861         .set_pauseparam = s2io_ethtool_setpause_data,
5862         .get_rx_csum = s2io_ethtool_get_rx_csum,
5863         .set_rx_csum = s2io_ethtool_set_rx_csum,
5864         .get_tx_csum = ethtool_op_get_tx_csum,
5865         .set_tx_csum = s2io_ethtool_op_set_tx_csum,
5866         .get_sg = ethtool_op_get_sg,
5867         .set_sg = ethtool_op_set_sg,
5868 #ifdef NETIF_F_TSO
5869         .get_tso = ethtool_op_get_tso,
5870         .set_tso = ethtool_op_set_tso,
5871 #endif
5872         .get_ufo = ethtool_op_get_ufo,
5873         .set_ufo = ethtool_op_set_ufo,
5874         .self_test_count = s2io_ethtool_self_test_count,
5875         .self_test = s2io_ethtool_test,
5876         .get_strings = s2io_ethtool_get_strings,
5877         .phys_id = s2io_ethtool_idnic,
5878         .get_stats_count = s2io_ethtool_get_stats_count,
5879         .get_ethtool_stats = s2io_get_ethtool_stats
5880 };
5881
5882 /**
5883  *  s2io_ioctl - Entry point for the Ioctl
5884  *  @dev :  Device pointer.
5885  *  @ifr :  An IOCTL specefic structure, that can contain a pointer to
5886  *  a proprietary structure used to pass information to the driver.
5887  *  @cmd :  This is used to distinguish between the different commands that
5888  *  can be passed to the IOCTL functions.
5889  *  Description:
5890  *  Currently there are no special functionality supported in IOCTL, hence
5891  *  function always return EOPNOTSUPPORTED
5892  */
5893
5894 static int s2io_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
5895 {
5896         return -EOPNOTSUPP;
5897 }
5898
5899 /**
5900  *  s2io_change_mtu - entry point to change MTU size for the device.
5901  *   @dev : device pointer.
5902  *   @new_mtu : the new MTU size for the device.
5903  *   Description: A driver entry point to change MTU size for the device.
5904  *   Before changing the MTU the device must be stopped.
5905  *  Return value:
5906  *   0 on success and an appropriate (-)ve integer as defined in errno.h
5907  *   file on failure.
5908  */
5909
5910 static int s2io_change_mtu(struct net_device *dev, int new_mtu)
5911 {
5912         nic_t *sp = dev->priv;
5913
5914         if ((new_mtu < MIN_MTU) || (new_mtu > S2IO_JUMBO_SIZE)) {
5915                 DBG_PRINT(ERR_DBG, "%s: MTU size is invalid.\n",
5916                           dev->name);
5917                 return -EPERM;
5918         }
5919
5920         dev->mtu = new_mtu;
5921         if (netif_running(dev)) {
5922                 s2io_card_down(sp, 0);
5923                 netif_stop_queue(dev);
5924                 if (s2io_card_up(sp)) {
5925                         DBG_PRINT(ERR_DBG, "%s: Device bring up failed\n",
5926                                   __FUNCTION__);
5927                 }
5928                 if (netif_queue_stopped(dev))
5929                         netif_wake_queue(dev);
5930         } else { /* Device is down */
5931                 XENA_dev_config_t __iomem *bar0 = sp->bar0;
5932                 u64 val64 = new_mtu;
5933
5934                 writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
5935         }
5936
5937         return 0;
5938 }
5939
5940 /**
5941  *  s2io_tasklet - Bottom half of the ISR.
5942  *  @dev_adr : address of the device structure in dma_addr_t format.
5943  *  Description:
5944  *  This is the tasklet or the bottom half of the ISR. This is
5945  *  an extension of the ISR which is scheduled by the scheduler to be run
5946  *  when the load on the CPU is low. All low priority tasks of the ISR can
5947  *  be pushed into the tasklet. For now the tasklet is used only to
5948  *  replenish the Rx buffers in the Rx buffer descriptors.
5949  *  Return value:
5950  *  void.
5951  */
5952
5953 static void s2io_tasklet(unsigned long dev_addr)
5954 {
5955         struct net_device *dev = (struct net_device *) dev_addr;
5956         nic_t *sp = dev->priv;
5957         int i, ret;
5958         mac_info_t *mac_control;
5959         struct config_param *config;
5960
5961         mac_control = &sp->mac_control;
5962         config = &sp->config;
5963
5964         if (!TASKLET_IN_USE) {
5965                 for (i = 0; i < config->rx_ring_num; i++) {
5966                         ret = fill_rx_buffers(sp, i);
5967                         if (ret == -ENOMEM) {
5968                                 DBG_PRINT(ERR_DBG, "%s: Out of ",
5969                                           dev->name);
5970                                 DBG_PRINT(ERR_DBG, "memory in tasklet\n");
5971                                 break;
5972                         } else if (ret == -EFILL) {
5973                                 DBG_PRINT(ERR_DBG,
5974                                           "%s: Rx Ring %d is full\n",
5975                                           dev->name, i);
5976                                 break;
5977                         }
5978                 }
5979                 clear_bit(0, (&sp->tasklet_status));
5980         }
5981 }
5982
5983 /**
5984  * s2io_set_link - Set the LInk status
5985  * @data: long pointer to device private structue
5986  * Description: Sets the link status for the adapter
5987  */
5988
5989 static void s2io_set_link(unsigned long data)
5990 {
5991         nic_t *nic = (nic_t *) data;
5992         struct net_device *dev = nic->dev;
5993         XENA_dev_config_t __iomem *bar0 = nic->bar0;
5994         register u64 val64;
5995         u16 subid;
5996
5997         if (test_and_set_bit(0, &(nic->link_state))) {
5998                 /* The card is being reset, no point doing anything */
5999                 return;
6000         }
6001
6002         subid = nic->pdev->subsystem_device;
6003         if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
6004                 /*
6005                  * Allow a small delay for the NICs self initiated
6006                  * cleanup to complete.
6007                  */
6008                 msleep(100);
6009         }
6010
6011         val64 = readq(&bar0->adapter_status);
6012         if (verify_xena_quiescence(nic, val64, nic->device_enabled_once)) {
6013                 if (LINK_IS_UP(val64)) {
6014                         val64 = readq(&bar0->adapter_control);
6015                         val64 |= ADAPTER_CNTL_EN;
6016                         writeq(val64, &bar0->adapter_control);
6017                         if (CARDS_WITH_FAULTY_LINK_INDICATORS(nic->device_type,
6018                                                              subid)) {
6019                                 val64 = readq(&bar0->gpio_control);
6020                                 val64 |= GPIO_CTRL_GPIO_0;
6021                                 writeq(val64, &bar0->gpio_control);
6022                                 val64 = readq(&bar0->gpio_control);
6023                         } else {
6024                                 val64 |= ADAPTER_LED_ON;
6025                                 writeq(val64, &bar0->adapter_control);
6026                         }
6027                         if (s2io_link_fault_indication(nic) ==
6028                                                 MAC_RMAC_ERR_TIMER) {
6029                                 val64 = readq(&bar0->adapter_status);
6030                                 if (!LINK_IS_UP(val64)) {
6031                                         DBG_PRINT(ERR_DBG, "%s:", dev->name);
6032                                         DBG_PRINT(ERR_DBG, " Link down");
6033                                         DBG_PRINT(ERR_DBG, "after ");
6034                                         DBG_PRINT(ERR_DBG, "enabling ");
6035                                         DBG_PRINT(ERR_DBG, "device \n");
6036                                 }
6037                         }
6038                         if (nic->device_enabled_once == FALSE) {
6039                                 nic->device_enabled_once = TRUE;
6040                         }
6041                         s2io_link(nic, LINK_UP);
6042                 } else {
6043                         if (CARDS_WITH_FAULTY_LINK_INDICATORS(nic->device_type,
6044                                                               subid)) {
6045                                 val64 = readq(&bar0->gpio_control);
6046                                 val64 &= ~GPIO_CTRL_GPIO_0;
6047                                 writeq(val64, &bar0->gpio_control);
6048                                 val64 = readq(&bar0->gpio_control);
6049                         }
6050                         s2io_link(nic, LINK_DOWN);
6051                 }
6052         } else {                /* NIC is not Quiescent. */
6053                 DBG_PRINT(ERR_DBG, "%s: Error: ", dev->name);
6054                 DBG_PRINT(ERR_DBG, "device is not Quiescent\n");
6055                 netif_stop_queue(dev);
6056         }
6057         clear_bit(0, &(nic->link_state));
6058 }
6059
6060 static int set_rxd_buffer_pointer(nic_t *sp, RxD_t *rxdp, buffAdd_t *ba,
6061                            struct sk_buff **skb, u64 *temp0, u64 *temp1,
6062                            u64 *temp2, int size)
6063 {
6064         struct net_device *dev = sp->dev;
6065         struct sk_buff *frag_list;
6066
6067         if ((sp->rxd_mode == RXD_MODE_1) && (rxdp->Host_Control == 0)) {
6068                 /* allocate skb */
6069                 if (*skb) {
6070                         DBG_PRINT(INFO_DBG, "SKB is not NULL\n");
6071                         /*
6072                          * As Rx frame are not going to be processed,
6073                          * using same mapped address for the Rxd
6074                          * buffer pointer
6075                          */
6076                         ((RxD1_t*)rxdp)->Buffer0_ptr = *temp0;
6077                 } else {
6078                         *skb = dev_alloc_skb(size);
6079                         if (!(*skb)) {
6080                                 DBG_PRINT(ERR_DBG, "%s: Out of ", dev->name);
6081                                 DBG_PRINT(ERR_DBG, "memory to allocate SKBs\n");
6082                                 return -ENOMEM ;
6083                         }
6084                         /* storing the mapped addr in a temp variable
6085                          * such it will be used for next rxd whose
6086                          * Host Control is NULL
6087                          */
6088                         ((RxD1_t*)rxdp)->Buffer0_ptr = *temp0 =
6089                                 pci_map_single( sp->pdev, (*skb)->data,
6090                                         size - NET_IP_ALIGN,
6091                                         PCI_DMA_FROMDEVICE);
6092                         rxdp->Host_Control = (unsigned long) (*skb);
6093                 }
6094         } else if ((sp->rxd_mode == RXD_MODE_3B) && (rxdp->Host_Control == 0)) {
6095                 /* Two buffer Mode */
6096                 if (*skb) {
6097                         ((RxD3_t*)rxdp)->Buffer2_ptr = *temp2;
6098                         ((RxD3_t*)rxdp)->Buffer0_ptr = *temp0;
6099                         ((RxD3_t*)rxdp)->Buffer1_ptr = *temp1;
6100                 } else {
6101                         *skb = dev_alloc_skb(size);
6102                         ((RxD3_t*)rxdp)->Buffer2_ptr = *temp2 =
6103                                 pci_map_single(sp->pdev, (*skb)->data,
6104                                                dev->mtu + 4,
6105                                                PCI_DMA_FROMDEVICE);
6106                         ((RxD3_t*)rxdp)->Buffer0_ptr = *temp0 =
6107                                 pci_map_single( sp->pdev, ba->ba_0, BUF0_LEN,
6108                                                 PCI_DMA_FROMDEVICE);
6109                         rxdp->Host_Control = (unsigned long) (*skb);
6110
6111                         /* Buffer-1 will be dummy buffer not used */
6112                         ((RxD3_t*)rxdp)->Buffer1_ptr = *temp1 =
6113                                 pci_map_single(sp->pdev, ba->ba_1, BUF1_LEN,
6114                                                PCI_DMA_FROMDEVICE);
6115                 }
6116         } else if ((rxdp->Host_Control == 0)) {
6117                 /* Three buffer mode */
6118                 if (*skb) {
6119                         ((RxD3_t*)rxdp)->Buffer0_ptr = *temp0;
6120                         ((RxD3_t*)rxdp)->Buffer1_ptr = *temp1;
6121                         ((RxD3_t*)rxdp)->Buffer2_ptr = *temp2;
6122                 } else {
6123                         *skb = dev_alloc_skb(size);
6124
6125                         ((RxD3_t*)rxdp)->Buffer0_ptr = *temp0 =
6126                                 pci_map_single(sp->pdev, ba->ba_0, BUF0_LEN,
6127                                                PCI_DMA_FROMDEVICE);
6128                         /* Buffer-1 receives L3/L4 headers */
6129                         ((RxD3_t*)rxdp)->Buffer1_ptr = *temp1 =
6130                                 pci_map_single( sp->pdev, (*skb)->data,
6131                                                 l3l4hdr_size + 4,
6132                                                 PCI_DMA_FROMDEVICE);
6133                         /*
6134                          * skb_shinfo(skb)->frag_list will have L4
6135                          * data payload
6136                          */
6137                         skb_shinfo(*skb)->frag_list = dev_alloc_skb(dev->mtu +
6138                                                                    ALIGN_SIZE);
6139                         if (skb_shinfo(*skb)->frag_list == NULL) {
6140                                 DBG_PRINT(ERR_DBG, "%s: dev_alloc_skb \
6141                                           failed\n ", dev->name);
6142                                 return -ENOMEM ;
6143                         }
6144                         frag_list = skb_shinfo(*skb)->frag_list;
6145                         frag_list->next = NULL;
6146                         /*
6147                          * Buffer-2 receives L4 data payload
6148                          */
6149                         ((RxD3_t*)rxdp)->Buffer2_ptr = *temp2 =
6150                                 pci_map_single( sp->pdev, frag_list->data,
6151                                                 dev->mtu, PCI_DMA_FROMDEVICE);
6152                 }
6153         }
6154         return 0;
6155 }
6156 static void set_rxd_buffer_size(nic_t *sp, RxD_t *rxdp, int size)
6157 {
6158         struct net_device *dev = sp->dev;
6159         if (sp->rxd_mode == RXD_MODE_1) {
6160                 rxdp->Control_2 = SET_BUFFER0_SIZE_1( size - NET_IP_ALIGN);
6161         } else if (sp->rxd_mode == RXD_MODE_3B) {
6162                 rxdp->Control_2 = SET_BUFFER0_SIZE_3(BUF0_LEN);
6163                 rxdp->Control_2 |= SET_BUFFER1_SIZE_3(1);
6164                 rxdp->Control_2 |= SET_BUFFER2_SIZE_3( dev->mtu + 4);
6165         } else {
6166                 rxdp->Control_2 = SET_BUFFER0_SIZE_3(BUF0_LEN);
6167                 rxdp->Control_2 |= SET_BUFFER1_SIZE_3(l3l4hdr_size + 4);
6168                 rxdp->Control_2 |= SET_BUFFER2_SIZE_3(dev->mtu);
6169         }
6170 }
6171
6172 static  int rxd_owner_bit_reset(nic_t *sp)
6173 {
6174         int i, j, k, blk_cnt = 0, size;
6175         mac_info_t * mac_control = &sp->mac_control;
6176         struct config_param *config = &sp->config;
6177         struct net_device *dev = sp->dev;
6178         RxD_t *rxdp = NULL;
6179         struct sk_buff *skb = NULL;
6180         buffAdd_t *ba = NULL;
6181         u64 temp0_64 = 0, temp1_64 = 0, temp2_64 = 0;
6182
6183         /* Calculate the size based on ring mode */
6184         size = dev->mtu + HEADER_ETHERNET_II_802_3_SIZE +
6185                 HEADER_802_2_SIZE + HEADER_SNAP_SIZE;
6186         if (sp->rxd_mode == RXD_MODE_1)
6187                 size += NET_IP_ALIGN;
6188         else if (sp->rxd_mode == RXD_MODE_3B)
6189                 size = dev->mtu + ALIGN_SIZE + BUF0_LEN + 4;
6190         else
6191                 size = l3l4hdr_size + ALIGN_SIZE + BUF0_LEN + 4;
6192
6193         for (i = 0; i < config->rx_ring_num; i++) {
6194                 blk_cnt = config->rx_cfg[i].num_rxd /
6195                         (rxd_count[sp->rxd_mode] +1);
6196
6197                 for (j = 0; j < blk_cnt; j++) {
6198                         for (k = 0; k < rxd_count[sp->rxd_mode]; k++) {
6199                                 rxdp = mac_control->rings[i].
6200                                         rx_blocks[j].rxds[k].virt_addr;
6201                                 if(sp->rxd_mode >= RXD_MODE_3A)
6202                                         ba = &mac_control->rings[i].ba[j][k];
6203                                 set_rxd_buffer_pointer(sp, rxdp, ba,
6204                                                        &skb,(u64 *)&temp0_64,
6205                                                        (u64 *)&temp1_64,
6206                                                        (u64 *)&temp2_64, size);
6207
6208                                 set_rxd_buffer_size(sp, rxdp, size);
6209                                 wmb();
6210                                 /* flip the Ownership bit to Hardware */
6211                                 rxdp->Control_1 |= RXD_OWN_XENA;
6212                         }
6213                 }
6214         }
6215         return 0;
6216
6217 }
6218
6219 static void s2io_card_down(nic_t * sp, int flag)
6220 {
6221         int cnt = 0;
6222         XENA_dev_config_t __iomem *bar0 = sp->bar0;
6223         unsigned long flags;
6224         register u64 val64 = 0;
6225         struct net_device *dev = sp->dev;
6226
6227         del_timer_sync(&sp->alarm_timer);
6228         /* If s2io_set_link task is executing, wait till it completes. */
6229         while (test_and_set_bit(0, &(sp->link_state))) {
6230                 msleep(50);
6231         }
6232         atomic_set(&sp->card_state, CARD_DOWN);
6233
6234         /* disable Tx and Rx traffic on the NIC */
6235         stop_nic(sp);
6236         if (flag) {
6237                 if (sp->intr_type == MSI_X) {
6238                         int i;
6239                         u16 msi_control;
6240
6241                         for (i=1; (sp->s2io_entries[i].in_use ==
6242                                 MSIX_REGISTERED_SUCCESS); i++) {
6243                                 int vector = sp->entries[i].vector;
6244                                 void *arg = sp->s2io_entries[i].arg;
6245
6246                                 free_irq(vector, arg);
6247                         }
6248                         pci_read_config_word(sp->pdev, 0x42, &msi_control);
6249                         msi_control &= 0xFFFE; /* Disable MSI */
6250                         pci_write_config_word(sp->pdev, 0x42, msi_control);
6251                         pci_disable_msix(sp->pdev);
6252                 } else {
6253                         free_irq(sp->pdev->irq, dev);
6254                         if (sp->intr_type == MSI)
6255                                 pci_disable_msi(sp->pdev);
6256                 }
6257         }
6258         /* Waiting till all Interrupt handlers are complete */
6259         cnt = 0;
6260         do {
6261                 msleep(10);
6262                 if (!atomic_read(&sp->isr_cnt))
6263                         break;
6264                 cnt++;
6265         } while(cnt < 5);
6266
6267         /* Kill tasklet. */
6268         tasklet_kill(&sp->task);
6269
6270         /* Check if the device is Quiescent and then Reset the NIC */
6271         do {
6272                 /* As per the HW requirement we need to replenish the
6273                  * receive buffer to avoid the ring bump. Since there is
6274                  * no intention of processing the Rx frame at this pointwe are
6275                  * just settting the ownership bit of rxd in Each Rx
6276                  * ring to HW and set the appropriate buffer size
6277                  * based on the ring mode
6278                  */
6279                 rxd_owner_bit_reset(sp);
6280
6281                 val64 = readq(&bar0->adapter_status);
6282                 if (verify_xena_quiescence(sp, val64, sp->device_enabled_once)) {
6283                         break;
6284                 }
6285
6286                 msleep(50);
6287                 cnt++;
6288                 if (cnt == 10) {
6289                         DBG_PRINT(ERR_DBG,
6290                                   "s2io_close:Device not Quiescent ");
6291                         DBG_PRINT(ERR_DBG, "adaper status reads 0x%llx\n",
6292                                   (unsigned long long) val64);
6293                         break;
6294                 }
6295         } while (1);
6296         s2io_reset(sp);
6297
6298         spin_lock_irqsave(&sp->tx_lock, flags);
6299         /* Free all Tx buffers */
6300         free_tx_buffers(sp);
6301         spin_unlock_irqrestore(&sp->tx_lock, flags);
6302
6303         /* Free all Rx buffers */
6304         spin_lock_irqsave(&sp->rx_lock, flags);
6305         free_rx_buffers(sp);
6306         spin_unlock_irqrestore(&sp->rx_lock, flags);
6307
6308         clear_bit(0, &(sp->link_state));
6309 }
6310
6311 static int s2io_card_up(nic_t * sp)
6312 {
6313         int i, ret = 0;
6314         mac_info_t *mac_control;
6315         struct config_param *config;
6316         struct net_device *dev = (struct net_device *) sp->dev;
6317
6318         /* Initialize the H/W I/O registers */
6319         if (init_nic(sp) != 0) {
6320                 DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
6321                           dev->name);
6322                 return -ENODEV;
6323         }
6324
6325         if (sp->intr_type == MSI)
6326                 ret = s2io_enable_msi(sp);
6327         else if (sp->intr_type == MSI_X)
6328                 ret = s2io_enable_msi_x(sp);
6329         if (ret) {
6330                 DBG_PRINT(ERR_DBG, "%s: Defaulting to INTA\n", dev->name);
6331                 sp->intr_type = INTA;
6332         }
6333
6334         /*
6335          * Initializing the Rx buffers. For now we are considering only 1
6336          * Rx ring and initializing buffers into 30 Rx blocks
6337          */
6338         mac_control = &sp->mac_control;
6339         config = &sp->config;
6340
6341         for (i = 0; i < config->rx_ring_num; i++) {
6342                 if ((ret = fill_rx_buffers(sp, i))) {
6343                         DBG_PRINT(ERR_DBG, "%s: Out of memory in Open\n",
6344                                   dev->name);
6345                         s2io_reset(sp);
6346                         free_rx_buffers(sp);
6347                         return -ENOMEM;
6348                 }
6349                 DBG_PRINT(INFO_DBG, "Buf in ring:%d is %d:\n", i,
6350                           atomic_read(&sp->rx_bufs_left[i]));
6351         }
6352
6353         /* Setting its receive mode */
6354         s2io_set_multicast(dev);
6355
6356         if (sp->lro) {
6357                 /* Initialize max aggregatable pkts based on MTU */
6358                 sp->lro_max_aggr_per_sess = ((1<<16) - 1) / dev->mtu;
6359                 /* Check if we can use(if specified) user provided value */
6360                 if (lro_max_pkts < sp->lro_max_aggr_per_sess)
6361                         sp->lro_max_aggr_per_sess = lro_max_pkts;
6362         }
6363
6364         /* Enable tasklet for the device */
6365         tasklet_init(&sp->task, s2io_tasklet, (unsigned long) dev);
6366
6367         /* Enable Rx Traffic and interrupts on the NIC */
6368         if (start_nic(sp)) {
6369                 DBG_PRINT(ERR_DBG, "%s: Starting NIC failed\n", dev->name);
6370                 tasklet_kill(&sp->task);
6371                 s2io_reset(sp);
6372                 free_irq(dev->irq, dev);
6373                 free_rx_buffers(sp);
6374                 return -ENODEV;
6375         }
6376
6377         S2IO_TIMER_CONF(sp->alarm_timer, s2io_alarm_handle, sp, (HZ/2));
6378
6379         atomic_set(&sp->card_state, CARD_UP);
6380         return 0;
6381 }
6382
6383 /**
6384  * s2io_restart_nic - Resets the NIC.
6385  * @data : long pointer to the device private structure
6386  * Description:
6387  * This function is scheduled to be run by the s2io_tx_watchdog
6388  * function after 0.5 secs to reset the NIC. The idea is to reduce
6389  * the run time of the watch dog routine which is run holding a
6390  * spin lock.
6391  */
6392
6393 static void s2io_restart_nic(unsigned long data)
6394 {
6395         struct net_device *dev = (struct net_device *) data;
6396         nic_t *sp = dev->priv;
6397
6398         s2io_card_down(sp, 0);
6399         if (s2io_card_up(sp)) {
6400                 DBG_PRINT(ERR_DBG, "%s: Device bring up failed\n",
6401                           dev->name);
6402         }
6403         netif_wake_queue(dev);
6404         DBG_PRINT(ERR_DBG, "%s: was reset by Tx watchdog timer\n",
6405                   dev->name);
6406
6407 }
6408
6409 /**
6410  *  s2io_tx_watchdog - Watchdog for transmit side.
6411  *  @dev : Pointer to net device structure
6412  *  Description:
6413  *  This function is triggered if the Tx Queue is stopped
6414  *  for a pre-defined amount of time when the Interface is still up.
6415  *  If the Interface is jammed in such a situation, the hardware is
6416  *  reset (by s2io_close) and restarted again (by s2io_open) to
6417  *  overcome any problem that might have been caused in the hardware.
6418  *  Return value:
6419  *  void
6420  */
6421
6422 static void s2io_tx_watchdog(struct net_device *dev)
6423 {
6424         nic_t *sp = dev->priv;
6425
6426         if (netif_carrier_ok(dev)) {
6427                 schedule_work(&sp->rst_timer_task);
6428                 sp->mac_control.stats_info->sw_stat.soft_reset_cnt++;
6429         }
6430 }
6431
6432 /**
6433  *   rx_osm_handler - To perform some OS related operations on SKB.
6434  *   @sp: private member of the device structure,pointer to s2io_nic structure.
6435  *   @skb : the socket buffer pointer.
6436  *   @len : length of the packet
6437  *   @cksum : FCS checksum of the frame.
6438  *   @ring_no : the ring from which this RxD was extracted.
6439  *   Description:
6440  *   This function is called by the Tx interrupt serivce routine to perform
6441  *   some OS related operations on the SKB before passing it to the upper
6442  *   layers. It mainly checks if the checksum is OK, if so adds it to the
6443  *   SKBs cksum variable, increments the Rx packet count and passes the SKB
6444  *   to the upper layer. If the checksum is wrong, it increments the Rx
6445  *   packet error count, frees the SKB and returns error.
6446  *   Return value:
6447  *   SUCCESS on success and -1 on failure.
6448  */
6449 static int rx_osm_handler(ring_info_t *ring_data, RxD_t * rxdp)
6450 {
6451         nic_t *sp = ring_data->nic;
6452         struct net_device *dev = (struct net_device *) sp->dev;
6453         struct sk_buff *skb = (struct sk_buff *)
6454                 ((unsigned long) rxdp->Host_Control);
6455         int ring_no = ring_data->ring_no;
6456         u16 l3_csum, l4_csum;
6457         unsigned long long err = rxdp->Control_1 & RXD_T_CODE;
6458         lro_t *lro;
6459
6460         skb->dev = dev;
6461
6462         if (err) {
6463                 /* Check for parity error */
6464                 if (err & 0x1) {
6465                         sp->mac_control.stats_info->sw_stat.parity_err_cnt++;
6466                 }
6467
6468                 /*
6469                 * Drop the packet if bad transfer code. Exception being
6470                 * 0x5, which could be due to unsupported IPv6 extension header.
6471                 * In this case, we let stack handle the packet.
6472                 * Note that in this case, since checksum will be incorrect,
6473                 * stack will validate the same.
6474                 */
6475                 if (err && ((err >> 48) != 0x5)) {
6476                         DBG_PRINT(ERR_DBG, "%s: Rx error Value: 0x%llx\n",
6477                                 dev->name, err);
6478                         sp->stats.rx_crc_errors++;
6479                         dev_kfree_skb(skb);
6480                         atomic_dec(&sp->rx_bufs_left[ring_no]);
6481                         rxdp->Host_Control = 0;
6482                         return 0;
6483                 }
6484         }
6485
6486         /* Updating statistics */
6487         rxdp->Host_Control = 0;
6488         sp->rx_pkt_count++;
6489         sp->stats.rx_packets++;
6490         if (sp->rxd_mode == RXD_MODE_1) {
6491                 int len = RXD_GET_BUFFER0_SIZE_1(rxdp->Control_2);
6492
6493                 sp->stats.rx_bytes += len;
6494                 skb_put(skb, len);
6495
6496         } else if (sp->rxd_mode >= RXD_MODE_3A) {
6497                 int get_block = ring_data->rx_curr_get_info.block_index;
6498                 int get_off = ring_data->rx_curr_get_info.offset;
6499                 int buf0_len = RXD_GET_BUFFER0_SIZE_3(rxdp->Control_2);
6500                 int buf2_len = RXD_GET_BUFFER2_SIZE_3(rxdp->Control_2);
6501                 unsigned char *buff = skb_push(skb, buf0_len);
6502
6503                 buffAdd_t *ba = &ring_data->ba[get_block][get_off];
6504                 sp->stats.rx_bytes += buf0_len + buf2_len;
6505                 memcpy(buff, ba->ba_0, buf0_len);
6506
6507                 if (sp->rxd_mode == RXD_MODE_3A) {
6508                         int buf1_len = RXD_GET_BUFFER1_SIZE_3(rxdp->Control_2);
6509
6510                         skb_put(skb, buf1_len);
6511                         skb->len += buf2_len;
6512                         skb->data_len += buf2_len;
6513                         skb->truesize += buf2_len;
6514                         skb_put(skb_shinfo(skb)->frag_list, buf2_len);
6515                         sp->stats.rx_bytes += buf1_len;
6516
6517                 } else
6518                         skb_put(skb, buf2_len);
6519         }
6520
6521         if ((rxdp->Control_1 & TCP_OR_UDP_FRAME) && ((!sp->lro) ||
6522             (sp->lro && (!(rxdp->Control_1 & RXD_FRAME_IP_FRAG)))) &&
6523             (sp->rx_csum)) {
6524                 l3_csum = RXD_GET_L3_CKSUM(rxdp->Control_1);
6525                 l4_csum = RXD_GET_L4_CKSUM(rxdp->Control_1);
6526                 if ((l3_csum == L3_CKSUM_OK) && (l4_csum == L4_CKSUM_OK)) {
6527                         /*
6528                          * NIC verifies if the Checksum of the received
6529                          * frame is Ok or not and accordingly returns
6530                          * a flag in the RxD.
6531                          */
6532                         skb->ip_summed = CHECKSUM_UNNECESSARY;
6533                         if (sp->lro) {
6534                                 u32 tcp_len;
6535                                 u8 *tcp;
6536                                 int ret = 0;
6537
6538                                 ret = s2io_club_tcp_session(skb->data, &tcp,
6539                                                 &tcp_len, &lro, rxdp, sp);
6540                                 switch (ret) {
6541                                         case 3: /* Begin anew */
6542                                                 lro->parent = skb;
6543                                                 goto aggregate;
6544                                         case 1: /* Aggregate */
6545                                         {
6546                                                 lro_append_pkt(sp, lro,
6547                                                         skb, tcp_len);
6548                                                 goto aggregate;
6549                                         }
6550                                         case 4: /* Flush session */
6551                                         {
6552                                                 lro_append_pkt(sp, lro,
6553                                                         skb, tcp_len);
6554                                                 queue_rx_frame(lro->parent);
6555                                                 clear_lro_session(lro);
6556                                                 sp->mac_control.stats_info->
6557                                                     sw_stat.flush_max_pkts++;
6558                                                 goto aggregate;
6559                                         }
6560                                         case 2: /* Flush both */
6561                                                 lro->parent->data_len =
6562                                                         lro->frags_len;
6563                                                 sp->mac_control.stats_info->
6564                                                      sw_stat.sending_both++;
6565                                                 queue_rx_frame(lro->parent);
6566                                                 clear_lro_session(lro);
6567                                                 goto send_up;
6568                                         case 0: /* sessions exceeded */
6569                                         case -1: /* non-TCP or not
6570                                                   * L2 aggregatable
6571                                                   */
6572                                         case 5: /*
6573                                                  * First pkt in session not
6574                                                  * L3/L4 aggregatable
6575                                                  */
6576                                                 break;
6577                                         default:
6578                                                 DBG_PRINT(ERR_DBG,
6579                                                         "%s: Samadhana!!\n",
6580                                                          __FUNCTION__);
6581                                                 BUG();
6582                                 }
6583                         }
6584                 } else {
6585                         /*
6586                          * Packet with erroneous checksum, let the
6587                          * upper layers deal with it.
6588                          */
6589                         skb->ip_summed = CHECKSUM_NONE;
6590                 }
6591         } else {
6592                 skb->ip_summed = CHECKSUM_NONE;
6593         }
6594
6595         if (!sp->lro) {
6596                 skb->protocol = eth_type_trans(skb, dev);
6597 #ifdef CONFIG_S2IO_NAPI
6598                 if (sp->vlgrp && RXD_GET_VLAN_TAG(rxdp->Control_2)) {
6599                         /* Queueing the vlan frame to the upper layer */
6600                         vlan_hwaccel_receive_skb(skb, sp->vlgrp,
6601                                 RXD_GET_VLAN_TAG(rxdp->Control_2));
6602                 } else {
6603                         netif_receive_skb(skb);
6604                 }
6605 #else
6606                 if (sp->vlgrp && RXD_GET_VLAN_TAG(rxdp->Control_2)) {
6607                         /* Queueing the vlan frame to the upper layer */
6608                         vlan_hwaccel_rx(skb, sp->vlgrp,
6609                                 RXD_GET_VLAN_TAG(rxdp->Control_2));
6610                 } else {
6611                         netif_rx(skb);
6612                 }
6613 #endif
6614         } else {
6615 send_up:
6616                 queue_rx_frame(skb);
6617         }               
6618         dev->last_rx = jiffies;
6619 aggregate:
6620         atomic_dec(&sp->rx_bufs_left[ring_no]);
6621         return SUCCESS;
6622 }
6623
6624 /**
6625  *  s2io_link - stops/starts the Tx queue.
6626  *  @sp : private member of the device structure, which is a pointer to the
6627  *  s2io_nic structure.
6628  *  @link : inidicates whether link is UP/DOWN.
6629  *  Description:
6630  *  This function stops/starts the Tx queue depending on whether the link
6631  *  status of the NIC is is down or up. This is called by the Alarm
6632  *  interrupt handler whenever a link change interrupt comes up.
6633  *  Return value:
6634  *  void.
6635  */
6636
6637 static void s2io_link(nic_t * sp, int link)
6638 {
6639         struct net_device *dev = (struct net_device *) sp->dev;
6640
6641         if (link != sp->last_link_state) {
6642                 if (link == LINK_DOWN) {
6643                         DBG_PRINT(ERR_DBG, "%s: Link down\n", dev->name);
6644                         netif_carrier_off(dev);
6645                 } else {
6646                         DBG_PRINT(ERR_DBG, "%s: Link Up\n", dev->name);
6647                         netif_carrier_on(dev);
6648                 }
6649         }
6650         sp->last_link_state = link;
6651 }
6652
6653 /**
6654  *  get_xena_rev_id - to identify revision ID of xena.
6655  *  @pdev : PCI Dev structure
6656  *  Description:
6657  *  Function to identify the Revision ID of xena.
6658  *  Return value:
6659  *  returns the revision ID of the device.
6660  */
6661
6662 static int get_xena_rev_id(struct pci_dev *pdev)
6663 {
6664         u8 id = 0;
6665         int ret;
6666         ret = pci_read_config_byte(pdev, PCI_REVISION_ID, (u8 *) & id);
6667         return id;
6668 }
6669
6670 /**
6671  *  s2io_init_pci -Initialization of PCI and PCI-X configuration registers .
6672  *  @sp : private member of the device structure, which is a pointer to the
6673  *  s2io_nic structure.
6674  *  Description:
6675  *  This function initializes a few of the PCI and PCI-X configuration registers
6676  *  with recommended values.
6677  *  Return value:
6678  *  void
6679  */
6680
6681 static void s2io_init_pci(nic_t * sp)
6682 {
6683         u16 pci_cmd = 0, pcix_cmd = 0;
6684
6685         /* Enable Data Parity Error Recovery in PCI-X command register. */
6686         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
6687                              &(pcix_cmd));
6688         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
6689                               (pcix_cmd | 1));
6690         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
6691                              &(pcix_cmd));
6692
6693         /* Set the PErr Response bit in PCI command register. */
6694         pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
6695         pci_write_config_word(sp->pdev, PCI_COMMAND,
6696                               (pci_cmd | PCI_COMMAND_PARITY));
6697         pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
6698 }
6699
6700 MODULE_AUTHOR("Raghavendra Koushik <raghavendra.koushik@neterion.com>");
6701 MODULE_LICENSE("GPL");
6702 MODULE_VERSION(DRV_VERSION);
6703
6704 module_param(tx_fifo_num, int, 0);
6705 module_param(rx_ring_num, int, 0);
6706 module_param(rx_ring_mode, int, 0);
6707 module_param_array(tx_fifo_len, uint, NULL, 0);
6708 module_param_array(rx_ring_sz, uint, NULL, 0);
6709 module_param_array(rts_frm_len, uint, NULL, 0);
6710 module_param(use_continuous_tx_intrs, int, 1);
6711 module_param(rmac_pause_time, int, 0);
6712 module_param(mc_pause_threshold_q0q3, int, 0);
6713 module_param(mc_pause_threshold_q4q7, int, 0);
6714 module_param(shared_splits, int, 0);
6715 module_param(tmac_util_period, int, 0);
6716 module_param(rmac_util_period, int, 0);
6717 module_param(bimodal, bool, 0);
6718 module_param(l3l4hdr_size, int , 0);
6719 #ifndef CONFIG_S2IO_NAPI
6720 module_param(indicate_max_pkts, int, 0);
6721 #endif
6722 module_param(rxsync_frequency, int, 0);
6723 module_param(intr_type, int, 0);
6724 module_param(lro, int, 0);
6725 module_param(lro_max_pkts, int, 0);
6726
6727 static int s2io_verify_parm(struct pci_dev *pdev, u8 *dev_intr_type)
6728 {
6729         if ( tx_fifo_num > 8) {
6730                 DBG_PRINT(ERR_DBG, "s2io: Requested number of Tx fifos not "
6731                          "supported\n");
6732                 DBG_PRINT(ERR_DBG, "s2io: Default to 8 Tx fifos\n");
6733                 tx_fifo_num = 8;
6734         }
6735         if ( rx_ring_num > 8) {
6736                 DBG_PRINT(ERR_DBG, "s2io: Requested number of Rx rings not "
6737                          "supported\n");
6738                 DBG_PRINT(ERR_DBG, "s2io: Default to 8 Rx rings\n");
6739                 rx_ring_num = 8;
6740         }
6741 #ifdef CONFIG_S2IO_NAPI
6742         if (*dev_intr_type != INTA) {
6743                 DBG_PRINT(ERR_DBG, "s2io: NAPI cannot be enabled when "
6744                           "MSI/MSI-X is enabled. Defaulting to INTA\n");
6745                 *dev_intr_type = INTA;
6746         }
6747 #endif
6748 #ifndef CONFIG_PCI_MSI
6749         if (*dev_intr_type != INTA) {
6750                 DBG_PRINT(ERR_DBG, "s2io: This kernel does not support"
6751                           "MSI/MSI-X. Defaulting to INTA\n");
6752                 *dev_intr_type = INTA;
6753         }
6754 #else
6755         if (*dev_intr_type > MSI_X) {
6756                 DBG_PRINT(ERR_DBG, "s2io: Wrong intr_type requested. "
6757                           "Defaulting to INTA\n");
6758                 *dev_intr_type = INTA;
6759         }
6760 #endif
6761         if ((*dev_intr_type == MSI_X) &&
6762                         ((pdev->device != PCI_DEVICE_ID_HERC_WIN) &&
6763                         (pdev->device != PCI_DEVICE_ID_HERC_UNI))) {
6764                 DBG_PRINT(ERR_DBG, "s2io: Xframe I does not support MSI_X. " 
6765                                         "Defaulting to INTA\n");
6766                 *dev_intr_type = INTA;
6767         }
6768         if (rx_ring_mode > 3) {
6769                 DBG_PRINT(ERR_DBG, "s2io: Requested ring mode not supported\n");
6770                 DBG_PRINT(ERR_DBG, "s2io: Defaulting to 3-buffer mode\n");
6771                 rx_ring_mode = 3;
6772         }
6773         return SUCCESS;
6774 }
6775
6776 /**
6777  *  s2io_init_nic - Initialization of the adapter .
6778  *  @pdev : structure containing the PCI related information of the device.
6779  *  @pre: List of PCI devices supported by the driver listed in s2io_tbl.
6780  *  Description:
6781  *  The function initializes an adapter identified by the pci_dec structure.
6782  *  All OS related initialization including memory and device structure and
6783  *  initlaization of the device private variable is done. Also the swapper
6784  *  control register is initialized to enable read and write into the I/O
6785  *  registers of the device.
6786  *  Return value:
6787  *  returns 0 on success and negative on failure.
6788  */
6789
6790 static int __devinit
6791 s2io_init_nic(struct pci_dev *pdev, const struct pci_device_id *pre)
6792 {
6793         nic_t *sp;
6794         struct net_device *dev;
6795         int i, j, ret;
6796         int dma_flag = FALSE;
6797         u32 mac_up, mac_down;
6798         u64 val64 = 0, tmp64 = 0;
6799         XENA_dev_config_t __iomem *bar0 = NULL;
6800         u16 subid;
6801         mac_info_t *mac_control;
6802         struct config_param *config;
6803         int mode;
6804         u8 dev_intr_type = intr_type;
6805
6806         if ((ret = s2io_verify_parm(pdev, &dev_intr_type)))
6807                 return ret;
6808
6809         if ((ret = pci_enable_device(pdev))) {
6810                 DBG_PRINT(ERR_DBG,
6811                           "s2io_init_nic: pci_enable_device failed\n");
6812                 return ret;
6813         }
6814
6815         if (!pci_set_dma_mask(pdev, DMA_64BIT_MASK)) {
6816                 DBG_PRINT(INIT_DBG, "s2io_init_nic: Using 64bit DMA\n");
6817                 dma_flag = TRUE;
6818                 if (pci_set_consistent_dma_mask
6819                     (pdev, DMA_64BIT_MASK)) {
6820                         DBG_PRINT(ERR_DBG,
6821                                   "Unable to obtain 64bit DMA for \
6822                                         consistent allocations\n");
6823                         pci_disable_device(pdev);
6824                         return -ENOMEM;
6825                 }
6826         } else if (!pci_set_dma_mask(pdev, DMA_32BIT_MASK)) {
6827                 DBG_PRINT(INIT_DBG, "s2io_init_nic: Using 32bit DMA\n");
6828         } else {
6829                 pci_disable_device(pdev);
6830                 return -ENOMEM;
6831         }
6832         if (dev_intr_type != MSI_X) {
6833                 if (pci_request_regions(pdev, s2io_driver_name)) {
6834                         DBG_PRINT(ERR_DBG, "Request Regions failed\n"),
6835                             pci_disable_device(pdev);
6836                         return -ENODEV;
6837                 }
6838         }
6839         else {
6840                 if (!(request_mem_region(pci_resource_start(pdev, 0),
6841                          pci_resource_len(pdev, 0), s2io_driver_name))) {
6842                         DBG_PRINT(ERR_DBG, "bar0 Request Regions failed\n");
6843                         pci_disable_device(pdev);
6844                         return -ENODEV;
6845                 }
6846                 if (!(request_mem_region(pci_resource_start(pdev, 2),
6847                          pci_resource_len(pdev, 2), s2io_driver_name))) {
6848                         DBG_PRINT(ERR_DBG, "bar1 Request Regions failed\n");
6849                         release_mem_region(pci_resource_start(pdev, 0),
6850                                    pci_resource_len(pdev, 0));
6851                         pci_disable_device(pdev);
6852                         return -ENODEV;
6853                 }
6854         }
6855
6856         dev = alloc_etherdev(sizeof(nic_t));
6857         if (dev == NULL) {
6858                 DBG_PRINT(ERR_DBG, "Device allocation failed\n");
6859                 pci_disable_device(pdev);
6860                 pci_release_regions(pdev);
6861                 return -ENODEV;
6862         }
6863
6864         pci_set_master(pdev);
6865         pci_set_drvdata(pdev, dev);
6866         SET_MODULE_OWNER(dev);
6867         SET_NETDEV_DEV(dev, &pdev->dev);
6868
6869         /*  Private member variable initialized to s2io NIC structure */
6870         sp = dev->priv;
6871         memset(sp, 0, sizeof(nic_t));
6872         sp->dev = dev;
6873         sp->pdev = pdev;
6874         sp->high_dma_flag = dma_flag;
6875         sp->device_enabled_once = FALSE;
6876         if (rx_ring_mode == 1)
6877                 sp->rxd_mode = RXD_MODE_1;
6878         if (rx_ring_mode == 2)
6879                 sp->rxd_mode = RXD_MODE_3B;
6880         if (rx_ring_mode == 3)
6881                 sp->rxd_mode = RXD_MODE_3A;
6882
6883         sp->intr_type = dev_intr_type;
6884
6885         if ((pdev->device == PCI_DEVICE_ID_HERC_WIN) ||
6886                 (pdev->device == PCI_DEVICE_ID_HERC_UNI))
6887                 sp->device_type = XFRAME_II_DEVICE;
6888         else
6889                 sp->device_type = XFRAME_I_DEVICE;
6890
6891         sp->lro = lro;
6892                 
6893         /* Initialize some PCI/PCI-X fields of the NIC. */
6894         s2io_init_pci(sp);
6895
6896         /*
6897          * Setting the device configuration parameters.
6898          * Most of these parameters can be specified by the user during
6899          * module insertion as they are module loadable parameters. If
6900          * these parameters are not not specified during load time, they
6901          * are initialized with default values.
6902          */
6903         mac_control = &sp->mac_control;
6904         config = &sp->config;
6905
6906         /* Tx side parameters. */
6907         config->tx_fifo_num = tx_fifo_num;
6908         for (i = 0; i < MAX_TX_FIFOS; i++) {
6909                 config->tx_cfg[i].fifo_len = tx_fifo_len[i];
6910                 config->tx_cfg[i].fifo_priority = i;
6911         }
6912
6913         /* mapping the QoS priority to the configured fifos */
6914         for (i = 0; i < MAX_TX_FIFOS; i++)
6915                 config->fifo_mapping[i] = fifo_map[config->tx_fifo_num][i];
6916
6917         config->tx_intr_type = TXD_INT_TYPE_UTILZ;
6918         for (i = 0; i < config->tx_fifo_num; i++) {
6919                 config->tx_cfg[i].f_no_snoop =
6920                     (NO_SNOOP_TXD | NO_SNOOP_TXD_BUFFER);
6921                 if (config->tx_cfg[i].fifo_len < 65) {
6922                         config->tx_intr_type = TXD_INT_TYPE_PER_LIST;
6923                         break;
6924                 }
6925         }
6926         /* + 2 because one Txd for skb->data and one Txd for UFO */
6927         config->max_txds = MAX_SKB_FRAGS + 2;
6928
6929         /* Rx side parameters. */
6930         config->rx_ring_num = rx_ring_num;
6931         for (i = 0; i < MAX_RX_RINGS; i++) {
6932                 config->rx_cfg[i].num_rxd = rx_ring_sz[i] *
6933                     (rxd_count[sp->rxd_mode] + 1);
6934                 config->rx_cfg[i].ring_priority = i;
6935         }
6936
6937         for (i = 0; i < rx_ring_num; i++) {
6938                 config->rx_cfg[i].ring_org = RING_ORG_BUFF1;
6939                 config->rx_cfg[i].f_no_snoop =
6940                     (NO_SNOOP_RXD | NO_SNOOP_RXD_BUFFER);
6941         }
6942
6943         /*  Setting Mac Control parameters */
6944         mac_control->rmac_pause_time = rmac_pause_time;
6945         mac_control->mc_pause_threshold_q0q3 = mc_pause_threshold_q0q3;
6946         mac_control->mc_pause_threshold_q4q7 = mc_pause_threshold_q4q7;
6947
6948
6949         /* Initialize Ring buffer parameters. */
6950         for (i = 0; i < config->rx_ring_num; i++)
6951                 atomic_set(&sp->rx_bufs_left[i], 0);
6952
6953         /* Initialize the number of ISRs currently running */
6954         atomic_set(&sp->isr_cnt, 0);
6955
6956         /*  initialize the shared memory used by the NIC and the host */
6957         if (init_shared_mem(sp)) {
6958                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n",
6959                           __FUNCTION__);
6960                 ret = -ENOMEM;
6961                 goto mem_alloc_failed;
6962         }
6963
6964         sp->bar0 = ioremap(pci_resource_start(pdev, 0),
6965                                      pci_resource_len(pdev, 0));
6966         if (!sp->bar0) {
6967                 DBG_PRINT(ERR_DBG, "%s: S2IO: cannot remap io mem1\n",
6968                           dev->name);
6969                 ret = -ENOMEM;
6970                 goto bar0_remap_failed;
6971         }
6972
6973         sp->bar1 = ioremap(pci_resource_start(pdev, 2),
6974                                      pci_resource_len(pdev, 2));
6975         if (!sp->bar1) {
6976                 DBG_PRINT(ERR_DBG, "%s: S2IO: cannot remap io mem2\n",
6977                           dev->name);
6978                 ret = -ENOMEM;
6979                 goto bar1_remap_failed;
6980         }
6981
6982         dev->irq = pdev->irq;
6983         dev->base_addr = (unsigned long) sp->bar0;
6984
6985         /* Initializing the BAR1 address as the start of the FIFO pointer. */
6986         for (j = 0; j < MAX_TX_FIFOS; j++) {
6987                 mac_control->tx_FIFO_start[j] = (TxFIFO_element_t __iomem *)
6988                     (sp->bar1 + (j * 0x00020000));
6989         }
6990
6991         /*  Driver entry points */
6992         dev->open = &s2io_open;
6993         dev->stop = &s2io_close;
6994         dev->hard_start_xmit = &s2io_xmit;
6995         dev->get_stats = &s2io_get_stats;
6996         dev->set_multicast_list = &s2io_set_multicast;
6997         dev->do_ioctl = &s2io_ioctl;
6998         dev->change_mtu = &s2io_change_mtu;
6999         SET_ETHTOOL_OPS(dev, &netdev_ethtool_ops);
7000         dev->features |= NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX;
7001         dev->vlan_rx_register = s2io_vlan_rx_register;
7002         dev->vlan_rx_kill_vid = (void *)s2io_vlan_rx_kill_vid;
7003
7004         /*
7005          * will use eth_mac_addr() for  dev->set_mac_address
7006          * mac address will be set every time dev->open() is called
7007          */
7008 #if defined(CONFIG_S2IO_NAPI)
7009         dev->poll = s2io_poll;
7010         dev->weight = 32;
7011 #endif
7012
7013 #ifdef CONFIG_NET_POLL_CONTROLLER
7014         dev->poll_controller = s2io_netpoll;
7015 #endif
7016
7017         dev->features |= NETIF_F_SG | NETIF_F_IP_CSUM;
7018         if (sp->high_dma_flag == TRUE)
7019                 dev->features |= NETIF_F_HIGHDMA;
7020 #ifdef NETIF_F_TSO
7021         dev->features |= NETIF_F_TSO;
7022 #endif
7023 #ifdef NETIF_F_TSO6
7024         dev->features |= NETIF_F_TSO6;
7025 #endif
7026         if (sp->device_type & XFRAME_II_DEVICE) {
7027                 dev->features |= NETIF_F_UFO;
7028                 dev->features |= NETIF_F_HW_CSUM;
7029         }
7030
7031         dev->tx_timeout = &s2io_tx_watchdog;
7032         dev->watchdog_timeo = WATCH_DOG_TIMEOUT;
7033         INIT_WORK(&sp->rst_timer_task,
7034                   (void (*)(void *)) s2io_restart_nic, dev);
7035         INIT_WORK(&sp->set_link_task,
7036                   (void (*)(void *)) s2io_set_link, sp);
7037
7038         pci_save_state(sp->pdev);
7039
7040         /* Setting swapper control on the NIC, for proper reset operation */
7041         if (s2io_set_swapper(sp)) {
7042                 DBG_PRINT(ERR_DBG, "%s:swapper settings are wrong\n",
7043                           dev->name);
7044                 ret = -EAGAIN;
7045                 goto set_swap_failed;
7046         }
7047
7048         /* Verify if the Herc works on the slot its placed into */
7049         if (sp->device_type & XFRAME_II_DEVICE) {
7050                 mode = s2io_verify_pci_mode(sp);
7051                 if (mode < 0) {
7052                         DBG_PRINT(ERR_DBG, "%s: ", __FUNCTION__);
7053                         DBG_PRINT(ERR_DBG, " Unsupported PCI bus mode\n");
7054                         ret = -EBADSLT;
7055                         goto set_swap_failed;
7056                 }
7057         }
7058
7059         /* Not needed for Herc */
7060         if (sp->device_type & XFRAME_I_DEVICE) {
7061                 /*
7062                  * Fix for all "FFs" MAC address problems observed on
7063                  * Alpha platforms
7064                  */
7065                 fix_mac_address(sp);
7066                 s2io_reset(sp);
7067         }
7068
7069         /*
7070          * MAC address initialization.
7071          * For now only one mac address will be read and used.
7072          */
7073         bar0 = sp->bar0;
7074         val64 = RMAC_ADDR_CMD_MEM_RD | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
7075             RMAC_ADDR_CMD_MEM_OFFSET(0 + MAC_MAC_ADDR_START_OFFSET);
7076         writeq(val64, &bar0->rmac_addr_cmd_mem);
7077         wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
7078                       RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING);
7079         tmp64 = readq(&bar0->rmac_addr_data0_mem);
7080         mac_down = (u32) tmp64;
7081         mac_up = (u32) (tmp64 >> 32);
7082
7083         memset(sp->def_mac_addr[0].mac_addr, 0, sizeof(ETH_ALEN));
7084
7085         sp->def_mac_addr[0].mac_addr[3] = (u8) (mac_up);
7086         sp->def_mac_addr[0].mac_addr[2] = (u8) (mac_up >> 8);
7087         sp->def_mac_addr[0].mac_addr[1] = (u8) (mac_up >> 16);
7088         sp->def_mac_addr[0].mac_addr[0] = (u8) (mac_up >> 24);
7089         sp->def_mac_addr[0].mac_addr[5] = (u8) (mac_down >> 16);
7090         sp->def_mac_addr[0].mac_addr[4] = (u8) (mac_down >> 24);
7091
7092         /*  Set the factory defined MAC address initially   */
7093         dev->addr_len = ETH_ALEN;
7094         memcpy(dev->dev_addr, sp->def_mac_addr, ETH_ALEN);
7095
7096         /*
7097          * Initialize the tasklet status and link state flags
7098          * and the card state parameter
7099          */
7100         atomic_set(&(sp->card_state), 0);
7101         sp->tasklet_status = 0;
7102         sp->link_state = 0;
7103
7104         /* Initialize spinlocks */
7105         spin_lock_init(&sp->tx_lock);
7106 #ifndef CONFIG_S2IO_NAPI
7107         spin_lock_init(&sp->put_lock);
7108 #endif
7109         spin_lock_init(&sp->rx_lock);
7110
7111         /*
7112          * SXE-002: Configure link and activity LED to init state
7113          * on driver load.
7114          */
7115         subid = sp->pdev->subsystem_device;
7116         if ((subid & 0xFF) >= 0x07) {
7117                 val64 = readq(&bar0->gpio_control);
7118                 val64 |= 0x0000800000000000ULL;
7119                 writeq(val64, &bar0->gpio_control);
7120                 val64 = 0x0411040400000000ULL;
7121                 writeq(val64, (void __iomem *) bar0 + 0x2700);
7122                 val64 = readq(&bar0->gpio_control);
7123         }
7124
7125         sp->rx_csum = 1;        /* Rx chksum verify enabled by default */
7126
7127         if (register_netdev(dev)) {
7128                 DBG_PRINT(ERR_DBG, "Device registration failed\n");
7129                 ret = -ENODEV;
7130                 goto register_failed;
7131         }
7132         s2io_vpd_read(sp);
7133         DBG_PRINT(ERR_DBG, "%s: Neterion %s",dev->name, sp->product_name);
7134         DBG_PRINT(ERR_DBG, "(rev %d), Driver version %s\n",
7135                                 get_xena_rev_id(sp->pdev),
7136                                 s2io_driver_version);
7137         DBG_PRINT(ERR_DBG, "Copyright(c) 2002-2005 Neterion Inc.\n");
7138         DBG_PRINT(ERR_DBG, "%s: MAC ADDR: "
7139                           "%02x:%02x:%02x:%02x:%02x:%02x\n", dev->name,
7140                           sp->def_mac_addr[0].mac_addr[0],
7141                           sp->def_mac_addr[0].mac_addr[1],
7142                           sp->def_mac_addr[0].mac_addr[2],
7143                           sp->def_mac_addr[0].mac_addr[3],
7144                           sp->def_mac_addr[0].mac_addr[4],
7145                           sp->def_mac_addr[0].mac_addr[5]);
7146         if (sp->device_type & XFRAME_II_DEVICE) {
7147                 mode = s2io_print_pci_mode(sp);
7148                 if (mode < 0) {
7149                         DBG_PRINT(ERR_DBG, " Unsupported PCI bus mode\n");
7150                         ret = -EBADSLT;
7151                         unregister_netdev(dev);
7152                         goto set_swap_failed;
7153                 }
7154         }
7155         switch(sp->rxd_mode) {
7156                 case RXD_MODE_1:
7157                     DBG_PRINT(ERR_DBG, "%s: 1-Buffer receive mode enabled\n",
7158                                                 dev->name);
7159                     break;
7160                 case RXD_MODE_3B:
7161                     DBG_PRINT(ERR_DBG, "%s: 2-Buffer receive mode enabled\n",
7162                                                 dev->name);
7163                     break;
7164                 case RXD_MODE_3A:
7165                     DBG_PRINT(ERR_DBG, "%s: 3-Buffer receive mode enabled\n",
7166                                                 dev->name);
7167                     break;
7168         }
7169 #ifdef CONFIG_S2IO_NAPI
7170         DBG_PRINT(ERR_DBG, "%s: NAPI enabled\n", dev->name);
7171 #endif
7172         switch(sp->intr_type) {
7173                 case INTA:
7174                     DBG_PRINT(ERR_DBG, "%s: Interrupt type INTA\n", dev->name);
7175                     break;
7176                 case MSI:
7177                     DBG_PRINT(ERR_DBG, "%s: Interrupt type MSI\n", dev->name);
7178                     break;
7179                 case MSI_X:
7180                     DBG_PRINT(ERR_DBG, "%s: Interrupt type MSI-X\n", dev->name);
7181                     break;
7182         }
7183         if (sp->lro)
7184                 DBG_PRINT(ERR_DBG, "%s: Large receive offload enabled\n",
7185                           dev->name);
7186
7187         /* Initialize device name */
7188         sprintf(sp->name, "%s Neterion %s", dev->name, sp->product_name);
7189
7190         /* Initialize bimodal Interrupts */
7191         sp->config.bimodal = bimodal;
7192         if (!(sp->device_type & XFRAME_II_DEVICE) && bimodal) {
7193                 sp->config.bimodal = 0;
7194                 DBG_PRINT(ERR_DBG,"%s:Bimodal intr not supported by Xframe I\n",
7195                         dev->name);
7196         }
7197
7198         /*
7199          * Make Link state as off at this point, when the Link change
7200          * interrupt comes the state will be automatically changed to
7201          * the right state.
7202          */
7203         netif_carrier_off(dev);
7204
7205         return 0;
7206
7207       register_failed:
7208       set_swap_failed:
7209         iounmap(sp->bar1);
7210       bar1_remap_failed:
7211         iounmap(sp->bar0);
7212       bar0_remap_failed:
7213       mem_alloc_failed:
7214         free_shared_mem(sp);
7215         pci_disable_device(pdev);
7216         if (dev_intr_type != MSI_X)
7217                 pci_release_regions(pdev);
7218         else {
7219                 release_mem_region(pci_resource_start(pdev, 0),
7220                         pci_resource_len(pdev, 0));
7221                 release_mem_region(pci_resource_start(pdev, 2),
7222                         pci_resource_len(pdev, 2));
7223         }
7224         pci_set_drvdata(pdev, NULL);
7225         free_netdev(dev);
7226
7227         return ret;
7228 }
7229
7230 /**
7231  * s2io_rem_nic - Free the PCI device
7232  * @pdev: structure containing the PCI related information of the device.
7233  * Description: This function is called by the Pci subsystem to release a
7234  * PCI device and free up all resource held up by the device. This could
7235  * be in response to a Hot plug event or when the driver is to be removed
7236  * from memory.
7237  */
7238
7239 static void __devexit s2io_rem_nic(struct pci_dev *pdev)
7240 {
7241         struct net_device *dev =
7242             (struct net_device *) pci_get_drvdata(pdev);
7243         nic_t *sp;
7244
7245         if (dev == NULL) {
7246                 DBG_PRINT(ERR_DBG, "Driver Data is NULL!!\n");
7247                 return;
7248         }
7249
7250         sp = dev->priv;
7251         unregister_netdev(dev);
7252
7253         free_shared_mem(sp);
7254         iounmap(sp->bar0);
7255         iounmap(sp->bar1);
7256         pci_disable_device(pdev);
7257         if (sp->intr_type != MSI_X)
7258                 pci_release_regions(pdev);
7259         else {
7260                 release_mem_region(pci_resource_start(pdev, 0),
7261                         pci_resource_len(pdev, 0));
7262                 release_mem_region(pci_resource_start(pdev, 2),
7263                         pci_resource_len(pdev, 2));
7264         }
7265         pci_set_drvdata(pdev, NULL);
7266         free_netdev(dev);
7267 }
7268
7269 /**
7270  * s2io_starter - Entry point for the driver
7271  * Description: This function is the entry point for the driver. It verifies
7272  * the module loadable parameters and initializes PCI configuration space.
7273  */
7274
7275 int __init s2io_starter(void)
7276 {
7277         return pci_module_init(&s2io_driver);
7278 }
7279
7280 /**
7281  * s2io_closer - Cleanup routine for the driver
7282  * Description: This function is the cleanup routine for the driver. It unregist * ers the driver.
7283  */
7284
7285 static void s2io_closer(void)
7286 {
7287         pci_unregister_driver(&s2io_driver);
7288         DBG_PRINT(INIT_DBG, "cleanup done\n");
7289 }
7290
7291 module_init(s2io_starter);
7292 module_exit(s2io_closer);
7293
7294 static int check_L2_lro_capable(u8 *buffer, struct iphdr **ip, 
7295                 struct tcphdr **tcp, RxD_t *rxdp)
7296 {
7297         int ip_off;
7298         u8 l2_type = (u8)((rxdp->Control_1 >> 37) & 0x7), ip_len;
7299
7300         if (!(rxdp->Control_1 & RXD_FRAME_PROTO_TCP)) {
7301                 DBG_PRINT(INIT_DBG,"%s: Non-TCP frames not supported for LRO\n",
7302                           __FUNCTION__);
7303                 return -1;
7304         }
7305
7306         /* TODO:
7307          * By default the VLAN field in the MAC is stripped by the card, if this
7308          * feature is turned off in rx_pa_cfg register, then the ip_off field
7309          * has to be shifted by a further 2 bytes
7310          */
7311         switch (l2_type) {
7312                 case 0: /* DIX type */
7313                 case 4: /* DIX type with VLAN */
7314                         ip_off = HEADER_ETHERNET_II_802_3_SIZE;
7315                         break;
7316                 /* LLC, SNAP etc are considered non-mergeable */
7317                 default:
7318                         return -1;
7319         }
7320
7321         *ip = (struct iphdr *)((u8 *)buffer + ip_off);
7322         ip_len = (u8)((*ip)->ihl);
7323         ip_len <<= 2;
7324         *tcp = (struct tcphdr *)((unsigned long)*ip + ip_len);
7325
7326         return 0;
7327 }
7328
7329 static int check_for_socket_match(lro_t *lro, struct iphdr *ip,
7330                                   struct tcphdr *tcp)
7331 {
7332         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7333         if ((lro->iph->saddr != ip->saddr) || (lro->iph->daddr != ip->daddr) ||
7334            (lro->tcph->source != tcp->source) || (lro->tcph->dest != tcp->dest))
7335                 return -1;
7336         return 0;
7337 }
7338
7339 static inline int get_l4_pyld_length(struct iphdr *ip, struct tcphdr *tcp)
7340 {
7341         return(ntohs(ip->tot_len) - (ip->ihl << 2) - (tcp->doff << 2));
7342 }
7343
7344 static void initiate_new_session(lro_t *lro, u8 *l2h,
7345                      struct iphdr *ip, struct tcphdr *tcp, u32 tcp_pyld_len)
7346 {
7347         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7348         lro->l2h = l2h;
7349         lro->iph = ip;
7350         lro->tcph = tcp;
7351         lro->tcp_next_seq = tcp_pyld_len + ntohl(tcp->seq);
7352         lro->tcp_ack = ntohl(tcp->ack_seq);
7353         lro->sg_num = 1;
7354         lro->total_len = ntohs(ip->tot_len);
7355         lro->frags_len = 0;
7356         /* 
7357          * check if we saw TCP timestamp. Other consistency checks have
7358          * already been done.
7359          */
7360         if (tcp->doff == 8) {
7361                 u32 *ptr;
7362                 ptr = (u32 *)(tcp+1);
7363                 lro->saw_ts = 1;
7364                 lro->cur_tsval = *(ptr+1);
7365                 lro->cur_tsecr = *(ptr+2);
7366         }
7367         lro->in_use = 1;
7368 }
7369
7370 static void update_L3L4_header(nic_t *sp, lro_t *lro)
7371 {
7372         struct iphdr *ip = lro->iph;
7373         struct tcphdr *tcp = lro->tcph;
7374         u16 nchk;
7375         StatInfo_t *statinfo = sp->mac_control.stats_info;
7376         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7377
7378         /* Update L3 header */
7379         ip->tot_len = htons(lro->total_len);
7380         ip->check = 0;
7381         nchk = ip_fast_csum((u8 *)lro->iph, ip->ihl);
7382         ip->check = nchk;
7383
7384         /* Update L4 header */
7385         tcp->ack_seq = lro->tcp_ack;
7386         tcp->window = lro->window;
7387
7388         /* Update tsecr field if this session has timestamps enabled */
7389         if (lro->saw_ts) {
7390                 u32 *ptr = (u32 *)(tcp + 1);
7391                 *(ptr+2) = lro->cur_tsecr;
7392         }
7393
7394         /* Update counters required for calculation of
7395          * average no. of packets aggregated.
7396          */
7397         statinfo->sw_stat.sum_avg_pkts_aggregated += lro->sg_num;
7398         statinfo->sw_stat.num_aggregations++;
7399 }
7400
7401 static void aggregate_new_rx(lro_t *lro, struct iphdr *ip,
7402                 struct tcphdr *tcp, u32 l4_pyld)
7403 {
7404         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7405         lro->total_len += l4_pyld;
7406         lro->frags_len += l4_pyld;
7407         lro->tcp_next_seq += l4_pyld;
7408         lro->sg_num++;
7409
7410         /* Update ack seq no. and window ad(from this pkt) in LRO object */
7411         lro->tcp_ack = tcp->ack_seq;
7412         lro->window = tcp->window;
7413         
7414         if (lro->saw_ts) {
7415                 u32 *ptr;
7416                 /* Update tsecr and tsval from this packet */
7417                 ptr = (u32 *) (tcp + 1);
7418                 lro->cur_tsval = *(ptr + 1); 
7419                 lro->cur_tsecr = *(ptr + 2);
7420         }
7421 }
7422
7423 static int verify_l3_l4_lro_capable(lro_t *l_lro, struct iphdr *ip,
7424                                     struct tcphdr *tcp, u32 tcp_pyld_len)
7425 {
7426         u8 *ptr;
7427
7428         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
7429
7430         if (!tcp_pyld_len) {
7431                 /* Runt frame or a pure ack */
7432                 return -1;
7433         }
7434
7435         if (ip->ihl != 5) /* IP has options */
7436                 return -1;
7437
7438         if (tcp->urg || tcp->psh || tcp->rst || tcp->syn || tcp->fin ||
7439                                                                 !tcp->ack) {
7440                 /*
7441                  * Currently recognize only the ack control word and
7442                  * any other control field being set would result in
7443                  * flushing the LRO session
7444                  */
7445                 return -1;
7446         }
7447
7448         /* 
7449          * Allow only one TCP timestamp option. Don't aggregate if
7450          * any other options are detected.
7451          */
7452         if (tcp->doff != 5 && tcp->doff != 8)
7453                 return -1;
7454
7455         if (tcp->doff == 8) {
7456                 ptr = (u8 *)(tcp + 1);  
7457                 while (*ptr == TCPOPT_NOP)
7458                         ptr++;
7459                 if (*ptr != TCPOPT_TIMESTAMP || *(ptr+1) != TCPOLEN_TIMESTAMP)
7460                         return -1;
7461
7462                 /* Ensure timestamp value increases monotonically */
7463                 if (l_lro)
7464                         if (l_lro->cur_tsval > *((u32 *)(ptr+2)))
7465                                 return -1;
7466
7467                 /* timestamp echo reply should be non-zero */
7468                 if (*((u32 *)(ptr+6)) == 0) 
7469                         return -1;
7470         }
7471
7472         return 0;
7473 }
7474
7475 static int
7476 s2io_club_tcp_session(u8 *buffer, u8 **tcp, u32 *tcp_len, lro_t **lro,
7477                       RxD_t *rxdp, nic_t *sp)
7478 {
7479         struct iphdr *ip;
7480         struct tcphdr *tcph;
7481         int ret = 0, i;
7482
7483         if (!(ret = check_L2_lro_capable(buffer, &ip, (struct tcphdr **)tcp,
7484                                          rxdp))) {
7485                 DBG_PRINT(INFO_DBG,"IP Saddr: %x Daddr: %x\n",
7486                           ip->saddr, ip->daddr);
7487         } else {
7488                 return ret;
7489         }
7490
7491         tcph = (struct tcphdr *)*tcp;
7492         *tcp_len = get_l4_pyld_length(ip, tcph);
7493         for (i=0; i<MAX_LRO_SESSIONS; i++) {
7494                 lro_t *l_lro = &sp->lro0_n[i];
7495                 if (l_lro->in_use) {
7496                         if (check_for_socket_match(l_lro, ip, tcph))
7497                                 continue;
7498                         /* Sock pair matched */
7499                         *lro = l_lro;
7500
7501                         if ((*lro)->tcp_next_seq != ntohl(tcph->seq)) {
7502                                 DBG_PRINT(INFO_DBG, "%s:Out of order. expected "
7503                                           "0x%x, actual 0x%x\n", __FUNCTION__,
7504                                           (*lro)->tcp_next_seq,
7505                                           ntohl(tcph->seq));
7506
7507                                 sp->mac_control.stats_info->
7508                                    sw_stat.outof_sequence_pkts++;
7509                                 ret = 2;
7510                                 break;
7511                         }
7512
7513                         if (!verify_l3_l4_lro_capable(l_lro, ip, tcph,*tcp_len))
7514                                 ret = 1; /* Aggregate */
7515                         else
7516                                 ret = 2; /* Flush both */
7517                         break;
7518                 }
7519         }
7520
7521         if (ret == 0) {
7522                 /* Before searching for available LRO objects,
7523                  * check if the pkt is L3/L4 aggregatable. If not
7524                  * don't create new LRO session. Just send this
7525                  * packet up.
7526                  */
7527                 if (verify_l3_l4_lro_capable(NULL, ip, tcph, *tcp_len)) {
7528                         return 5;
7529                 }
7530
7531                 for (i=0; i<MAX_LRO_SESSIONS; i++) {
7532                         lro_t *l_lro = &sp->lro0_n[i];
7533                         if (!(l_lro->in_use)) {
7534                                 *lro = l_lro;
7535                                 ret = 3; /* Begin anew */
7536                                 break;
7537                         }
7538                 }
7539         }
7540
7541         if (ret == 0) { /* sessions exceeded */
7542                 DBG_PRINT(INFO_DBG,"%s:All LRO sessions already in use\n",
7543                           __FUNCTION__);
7544                 *lro = NULL;
7545                 return ret;
7546         }
7547
7548         switch (ret) {
7549                 case 3:
7550                         initiate_new_session(*lro, buffer, ip, tcph, *tcp_len);
7551                         break;
7552                 case 2:
7553                         update_L3L4_header(sp, *lro);
7554                         break;
7555                 case 1:
7556                         aggregate_new_rx(*lro, ip, tcph, *tcp_len);
7557                         if ((*lro)->sg_num == sp->lro_max_aggr_per_sess) {
7558                                 update_L3L4_header(sp, *lro);
7559                                 ret = 4; /* Flush the LRO */
7560                         }
7561                         break;
7562                 default:
7563                         DBG_PRINT(ERR_DBG,"%s:Dont know, can't say!!\n",
7564                                 __FUNCTION__);
7565                         break;
7566         }
7567
7568         return ret;
7569 }
7570
7571 static void clear_lro_session(lro_t *lro)
7572 {
7573         static u16 lro_struct_size = sizeof(lro_t);
7574
7575         memset(lro, 0, lro_struct_size);
7576 }
7577
7578 static void queue_rx_frame(struct sk_buff *skb)
7579 {
7580         struct net_device *dev = skb->dev;
7581
7582         skb->protocol = eth_type_trans(skb, dev);
7583 #ifdef CONFIG_S2IO_NAPI
7584         netif_receive_skb(skb);
7585 #else
7586         netif_rx(skb);
7587 #endif
7588 }
7589
7590 static void lro_append_pkt(nic_t *sp, lro_t *lro, struct sk_buff *skb,
7591                            u32 tcp_len)
7592 {
7593         struct sk_buff *tmp, *first = lro->parent;
7594
7595         first->len += tcp_len;
7596         first->data_len = lro->frags_len;
7597         skb_pull(skb, (skb->len - tcp_len));
7598         if ((tmp = skb_shinfo(first)->frag_list)) {
7599                 while (tmp->next)
7600                         tmp = tmp->next;
7601                 tmp->next = skb;
7602         }
7603         else
7604                 skb_shinfo(first)->frag_list = skb;
7605         sp->mac_control.stats_info->sw_stat.clubbed_frms_cnt++;
7606         return;
7607 }