1/************************************************************************
2 * s2io.c: A Linux PCI-X Ethernet driver for Neterion 10GbE Server NIC
3 * Copyright(c) 2002-2010 Exar Corp.
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 * explanation 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.
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 *     2(MSI_X). Default value is '2(MSI_X)'
41 * lro_max_pkts: This parameter defines maximum number of packets can be
42 *     aggregated as a single large packet
43 * napi: This parameter used to enable/disable NAPI (polling Rx)
44 *     Possible values '1' for enable and '0' for disable. Default is '1'
45 * ufo: This parameter used to enable/disable UDP Fragmentation Offload(UFO)
46 *      Possible values '1' for enable and '0' for disable. Default is '0'
47 * vlan_tag_strip: This can be used to enable or disable vlan stripping.
48 *                 Possible values '1' for enable , '0' for disable.
49 *                 Default is '2' - which means disable in promisc mode
50 *                 and enable in non-promiscuous mode.
51 * multiq: This parameter used to enable/disable MULTIQUEUE support.
52 *      Possible values '1' for enable and '0' for disable. Default is '0'
53 ************************************************************************/
54
55#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
56
57#include <linux/module.h>
58#include <linux/types.h>
59#include <linux/errno.h>
60#include <linux/ioport.h>
61#include <linux/pci.h>
62#include <linux/dma-mapping.h>
63#include <linux/kernel.h>
64#include <linux/netdevice.h>
65#include <linux/etherdevice.h>
66#include <linux/mdio.h>
67#include <linux/skbuff.h>
68#include <linux/init.h>
69#include <linux/delay.h>
70#include <linux/stddef.h>
71#include <linux/ioctl.h>
72#include <linux/timex.h>
73#include <linux/ethtool.h>
74#include <linux/workqueue.h>
75#include <linux/if_vlan.h>
76#include <linux/ip.h>
77#include <linux/tcp.h>
78#include <linux/uaccess.h>
79#include <linux/io.h>
80#include <linux/slab.h>
81#include <linux/prefetch.h>
82#include <net/tcp.h>
83#include <net/checksum.h>
84
85#include <asm/div64.h>
86#include <asm/irq.h>
87
88/* local include */
89#include "s2io.h"
90#include "s2io-regs.h"
91
92#define DRV_VERSION "2.0.26.28"
93
94/* S2io Driver name & version. */
95static const char s2io_driver_name[] = "Neterion";
96static const char s2io_driver_version[] = DRV_VERSION;
97
98static const int rxd_size[2] = {32, 48};
99static const int rxd_count[2] = {127, 85};
100
101static inline int RXD_IS_UP2DT(struct RxD_t *rxdp)
102{
103	int ret;
104
105	ret = ((!(rxdp->Control_1 & RXD_OWN_XENA)) &&
106	       (GET_RXD_MARKER(rxdp->Control_2) != THE_RXD_MARK));
107
108	return ret;
109}
110
111/*
112 * Cards with following subsystem_id have a link state indication
113 * problem, 600B, 600C, 600D, 640B, 640C and 640D.
114 * macro below identifies these cards given the subsystem_id.
115 */
116#define CARDS_WITH_FAULTY_LINK_INDICATORS(dev_type, subid)		\
117	(dev_type == XFRAME_I_DEVICE) ?					\
118	((((subid >= 0x600B) && (subid <= 0x600D)) ||			\
119	  ((subid >= 0x640B) && (subid <= 0x640D))) ? 1 : 0) : 0
120
121#define LINK_IS_UP(val64) (!(val64 & (ADAPTER_STATUS_RMAC_REMOTE_FAULT | \
122				      ADAPTER_STATUS_RMAC_LOCAL_FAULT)))
123
124static inline int is_s2io_card_up(const struct s2io_nic *sp)
125{
126	return test_bit(__S2IO_STATE_CARD_UP, &sp->state);
127}
128
129/* Ethtool related variables and Macros. */
130static const char s2io_gstrings[][ETH_GSTRING_LEN] = {
131	"Register test\t(offline)",
132	"Eeprom test\t(offline)",
133	"Link test\t(online)",
134	"RLDRAM test\t(offline)",
135	"BIST Test\t(offline)"
136};
137
138static const char ethtool_xena_stats_keys[][ETH_GSTRING_LEN] = {
139	{"tmac_frms"},
140	{"tmac_data_octets"},
141	{"tmac_drop_frms"},
142	{"tmac_mcst_frms"},
143	{"tmac_bcst_frms"},
144	{"tmac_pause_ctrl_frms"},
145	{"tmac_ttl_octets"},
146	{"tmac_ucst_frms"},
147	{"tmac_nucst_frms"},
148	{"tmac_any_err_frms"},
149	{"tmac_ttl_less_fb_octets"},
150	{"tmac_vld_ip_octets"},
151	{"tmac_vld_ip"},
152	{"tmac_drop_ip"},
153	{"tmac_icmp"},
154	{"tmac_rst_tcp"},
155	{"tmac_tcp"},
156	{"tmac_udp"},
157	{"rmac_vld_frms"},
158	{"rmac_data_octets"},
159	{"rmac_fcs_err_frms"},
160	{"rmac_drop_frms"},
161	{"rmac_vld_mcst_frms"},
162	{"rmac_vld_bcst_frms"},
163	{"rmac_in_rng_len_err_frms"},
164	{"rmac_out_rng_len_err_frms"},
165	{"rmac_long_frms"},
166	{"rmac_pause_ctrl_frms"},
167	{"rmac_unsup_ctrl_frms"},
168	{"rmac_ttl_octets"},
169	{"rmac_accepted_ucst_frms"},
170	{"rmac_accepted_nucst_frms"},
171	{"rmac_discarded_frms"},
172	{"rmac_drop_events"},
173	{"rmac_ttl_less_fb_octets"},
174	{"rmac_ttl_frms"},
175	{"rmac_usized_frms"},
176	{"rmac_osized_frms"},
177	{"rmac_frag_frms"},
178	{"rmac_jabber_frms"},
179	{"rmac_ttl_64_frms"},
180	{"rmac_ttl_65_127_frms"},
181	{"rmac_ttl_128_255_frms"},
182	{"rmac_ttl_256_511_frms"},
183	{"rmac_ttl_512_1023_frms"},
184	{"rmac_ttl_1024_1518_frms"},
185	{"rmac_ip"},
186	{"rmac_ip_octets"},
187	{"rmac_hdr_err_ip"},
188	{"rmac_drop_ip"},
189	{"rmac_icmp"},
190	{"rmac_tcp"},
191	{"rmac_udp"},
192	{"rmac_err_drp_udp"},
193	{"rmac_xgmii_err_sym"},
194	{"rmac_frms_q0"},
195	{"rmac_frms_q1"},
196	{"rmac_frms_q2"},
197	{"rmac_frms_q3"},
198	{"rmac_frms_q4"},
199	{"rmac_frms_q5"},
200	{"rmac_frms_q6"},
201	{"rmac_frms_q7"},
202	{"rmac_full_q0"},
203	{"rmac_full_q1"},
204	{"rmac_full_q2"},
205	{"rmac_full_q3"},
206	{"rmac_full_q4"},
207	{"rmac_full_q5"},
208	{"rmac_full_q6"},
209	{"rmac_full_q7"},
210	{"rmac_pause_cnt"},
211	{"rmac_xgmii_data_err_cnt"},
212	{"rmac_xgmii_ctrl_err_cnt"},
213	{"rmac_accepted_ip"},
214	{"rmac_err_tcp"},
215	{"rd_req_cnt"},
216	{"new_rd_req_cnt"},
217	{"new_rd_req_rtry_cnt"},
218	{"rd_rtry_cnt"},
219	{"wr_rtry_rd_ack_cnt"},
220	{"wr_req_cnt"},
221	{"new_wr_req_cnt"},
222	{"new_wr_req_rtry_cnt"},
223	{"wr_rtry_cnt"},
224	{"wr_disc_cnt"},
225	{"rd_rtry_wr_ack_cnt"},
226	{"txp_wr_cnt"},
227	{"txd_rd_cnt"},
228	{"txd_wr_cnt"},
229	{"rxd_rd_cnt"},
230	{"rxd_wr_cnt"},
231	{"txf_rd_cnt"},
232	{"rxf_wr_cnt"}
233};
234
235static const char ethtool_enhanced_stats_keys[][ETH_GSTRING_LEN] = {
236	{"rmac_ttl_1519_4095_frms"},
237	{"rmac_ttl_4096_8191_frms"},
238	{"rmac_ttl_8192_max_frms"},
239	{"rmac_ttl_gt_max_frms"},
240	{"rmac_osized_alt_frms"},
241	{"rmac_jabber_alt_frms"},
242	{"rmac_gt_max_alt_frms"},
243	{"rmac_vlan_frms"},
244	{"rmac_len_discard"},
245	{"rmac_fcs_discard"},
246	{"rmac_pf_discard"},
247	{"rmac_da_discard"},
248	{"rmac_red_discard"},
249	{"rmac_rts_discard"},
250	{"rmac_ingm_full_discard"},
251	{"link_fault_cnt"}
252};
253
254static const char ethtool_driver_stats_keys[][ETH_GSTRING_LEN] = {
255	{"\n DRIVER STATISTICS"},
256	{"single_bit_ecc_errs"},
257	{"double_bit_ecc_errs"},
258	{"parity_err_cnt"},
259	{"serious_err_cnt"},
260	{"soft_reset_cnt"},
261	{"fifo_full_cnt"},
262	{"ring_0_full_cnt"},
263	{"ring_1_full_cnt"},
264	{"ring_2_full_cnt"},
265	{"ring_3_full_cnt"},
266	{"ring_4_full_cnt"},
267	{"ring_5_full_cnt"},
268	{"ring_6_full_cnt"},
269	{"ring_7_full_cnt"},
270	{"alarm_transceiver_temp_high"},
271	{"alarm_transceiver_temp_low"},
272	{"alarm_laser_bias_current_high"},
273	{"alarm_laser_bias_current_low"},
274	{"alarm_laser_output_power_high"},
275	{"alarm_laser_output_power_low"},
276	{"warn_transceiver_temp_high"},
277	{"warn_transceiver_temp_low"},
278	{"warn_laser_bias_current_high"},
279	{"warn_laser_bias_current_low"},
280	{"warn_laser_output_power_high"},
281	{"warn_laser_output_power_low"},
282	{"lro_aggregated_pkts"},
283	{"lro_flush_both_count"},
284	{"lro_out_of_sequence_pkts"},
285	{"lro_flush_due_to_max_pkts"},
286	{"lro_avg_aggr_pkts"},
287	{"mem_alloc_fail_cnt"},
288	{"pci_map_fail_cnt"},
289	{"watchdog_timer_cnt"},
290	{"mem_allocated"},
291	{"mem_freed"},
292	{"link_up_cnt"},
293	{"link_down_cnt"},
294	{"link_up_time"},
295	{"link_down_time"},
296	{"tx_tcode_buf_abort_cnt"},
297	{"tx_tcode_desc_abort_cnt"},
298	{"tx_tcode_parity_err_cnt"},
299	{"tx_tcode_link_loss_cnt"},
300	{"tx_tcode_list_proc_err_cnt"},
301	{"rx_tcode_parity_err_cnt"},
302	{"rx_tcode_abort_cnt"},
303	{"rx_tcode_parity_abort_cnt"},
304	{"rx_tcode_rda_fail_cnt"},
305	{"rx_tcode_unkn_prot_cnt"},
306	{"rx_tcode_fcs_err_cnt"},
307	{"rx_tcode_buf_size_err_cnt"},
308	{"rx_tcode_rxd_corrupt_cnt"},
309	{"rx_tcode_unkn_err_cnt"},
310	{"tda_err_cnt"},
311	{"pfc_err_cnt"},
312	{"pcc_err_cnt"},
313	{"tti_err_cnt"},
314	{"tpa_err_cnt"},
315	{"sm_err_cnt"},
316	{"lso_err_cnt"},
317	{"mac_tmac_err_cnt"},
318	{"mac_rmac_err_cnt"},
319	{"xgxs_txgxs_err_cnt"},
320	{"xgxs_rxgxs_err_cnt"},
321	{"rc_err_cnt"},
322	{"prc_pcix_err_cnt"},
323	{"rpa_err_cnt"},
324	{"rda_err_cnt"},
325	{"rti_err_cnt"},
326	{"mc_err_cnt"}
327};
328
329#define S2IO_XENA_STAT_LEN	ARRAY_SIZE(ethtool_xena_stats_keys)
330#define S2IO_ENHANCED_STAT_LEN	ARRAY_SIZE(ethtool_enhanced_stats_keys)
331#define S2IO_DRIVER_STAT_LEN	ARRAY_SIZE(ethtool_driver_stats_keys)
332
333#define XFRAME_I_STAT_LEN (S2IO_XENA_STAT_LEN + S2IO_DRIVER_STAT_LEN)
334#define XFRAME_II_STAT_LEN (XFRAME_I_STAT_LEN + S2IO_ENHANCED_STAT_LEN)
335
336#define XFRAME_I_STAT_STRINGS_LEN (XFRAME_I_STAT_LEN * ETH_GSTRING_LEN)
337#define XFRAME_II_STAT_STRINGS_LEN (XFRAME_II_STAT_LEN * ETH_GSTRING_LEN)
338
339#define S2IO_TEST_LEN	ARRAY_SIZE(s2io_gstrings)
340#define S2IO_STRINGS_LEN	(S2IO_TEST_LEN * ETH_GSTRING_LEN)
341
342#define S2IO_TIMER_CONF(timer, handle, arg, exp)	\
343	init_timer(&timer);				\
344	timer.function = handle;			\
345	timer.data = (unsigned long)arg;		\
346	mod_timer(&timer, (jiffies + exp))		\
347
348/* copy mac addr to def_mac_addr array */
349static void do_s2io_copy_mac_addr(struct s2io_nic *sp, int offset, u64 mac_addr)
350{
351	sp->def_mac_addr[offset].mac_addr[5] = (u8) (mac_addr);
352	sp->def_mac_addr[offset].mac_addr[4] = (u8) (mac_addr >> 8);
353	sp->def_mac_addr[offset].mac_addr[3] = (u8) (mac_addr >> 16);
354	sp->def_mac_addr[offset].mac_addr[2] = (u8) (mac_addr >> 24);
355	sp->def_mac_addr[offset].mac_addr[1] = (u8) (mac_addr >> 32);
356	sp->def_mac_addr[offset].mac_addr[0] = (u8) (mac_addr >> 40);
357}
358
359/*
360 * Constants to be programmed into the Xena's registers, to configure
361 * the XAUI.
362 */
363
364#define	END_SIGN	0x0
365static const u64 herc_act_dtx_cfg[] = {
366	/* Set address */
367	0x8000051536750000ULL, 0x80000515367500E0ULL,
368	/* Write data */
369	0x8000051536750004ULL, 0x80000515367500E4ULL,
370	/* Set address */
371	0x80010515003F0000ULL, 0x80010515003F00E0ULL,
372	/* Write data */
373	0x80010515003F0004ULL, 0x80010515003F00E4ULL,
374	/* Set address */
375	0x801205150D440000ULL, 0x801205150D4400E0ULL,
376	/* Write data */
377	0x801205150D440004ULL, 0x801205150D4400E4ULL,
378	/* Set address */
379	0x80020515F2100000ULL, 0x80020515F21000E0ULL,
380	/* Write data */
381	0x80020515F2100004ULL, 0x80020515F21000E4ULL,
382	/* Done */
383	END_SIGN
384};
385
386static const u64 xena_dtx_cfg[] = {
387	/* Set address */
388	0x8000051500000000ULL, 0x80000515000000E0ULL,
389	/* Write data */
390	0x80000515D9350004ULL, 0x80000515D93500E4ULL,
391	/* Set address */
392	0x8001051500000000ULL, 0x80010515000000E0ULL,
393	/* Write data */
394	0x80010515001E0004ULL, 0x80010515001E00E4ULL,
395	/* Set address */
396	0x8002051500000000ULL, 0x80020515000000E0ULL,
397	/* Write data */
398	0x80020515F2100004ULL, 0x80020515F21000E4ULL,
399	END_SIGN
400};
401
402/*
403 * Constants for Fixing the MacAddress problem seen mostly on
404 * Alpha machines.
405 */
406static const u64 fix_mac[] = {
407	0x0060000000000000ULL, 0x0060600000000000ULL,
408	0x0040600000000000ULL, 0x0000600000000000ULL,
409	0x0020600000000000ULL, 0x0060600000000000ULL,
410	0x0020600000000000ULL, 0x0060600000000000ULL,
411	0x0020600000000000ULL, 0x0060600000000000ULL,
412	0x0020600000000000ULL, 0x0060600000000000ULL,
413	0x0020600000000000ULL, 0x0060600000000000ULL,
414	0x0020600000000000ULL, 0x0060600000000000ULL,
415	0x0020600000000000ULL, 0x0060600000000000ULL,
416	0x0020600000000000ULL, 0x0060600000000000ULL,
417	0x0020600000000000ULL, 0x0060600000000000ULL,
418	0x0020600000000000ULL, 0x0060600000000000ULL,
419	0x0020600000000000ULL, 0x0000600000000000ULL,
420	0x0040600000000000ULL, 0x0060600000000000ULL,
421	END_SIGN
422};
423
424MODULE_LICENSE("GPL");
425MODULE_VERSION(DRV_VERSION);
426
427
428/* Module Loadable parameters. */
429S2IO_PARM_INT(tx_fifo_num, FIFO_DEFAULT_NUM);
430S2IO_PARM_INT(rx_ring_num, 1);
431S2IO_PARM_INT(multiq, 0);
432S2IO_PARM_INT(rx_ring_mode, 1);
433S2IO_PARM_INT(use_continuous_tx_intrs, 1);
434S2IO_PARM_INT(rmac_pause_time, 0x100);
435S2IO_PARM_INT(mc_pause_threshold_q0q3, 187);
436S2IO_PARM_INT(mc_pause_threshold_q4q7, 187);
437S2IO_PARM_INT(shared_splits, 0);
438S2IO_PARM_INT(tmac_util_period, 5);
439S2IO_PARM_INT(rmac_util_period, 5);
440S2IO_PARM_INT(l3l4hdr_size, 128);
441/* 0 is no steering, 1 is Priority steering, 2 is Default steering */
442S2IO_PARM_INT(tx_steering_type, TX_DEFAULT_STEERING);
443/* Frequency of Rx desc syncs expressed as power of 2 */
444S2IO_PARM_INT(rxsync_frequency, 3);
445/* Interrupt type. Values can be 0(INTA), 2(MSI_X) */
446S2IO_PARM_INT(intr_type, 2);
447/* Large receive offload feature */
448
449/* Max pkts to be aggregated by LRO at one time. If not specified,
450 * aggregation happens until we hit max IP pkt size(64K)
451 */
452S2IO_PARM_INT(lro_max_pkts, 0xFFFF);
453S2IO_PARM_INT(indicate_max_pkts, 0);
454
455S2IO_PARM_INT(napi, 1);
456S2IO_PARM_INT(ufo, 0);
457S2IO_PARM_INT(vlan_tag_strip, NO_STRIP_IN_PROMISC);
458
459static unsigned int tx_fifo_len[MAX_TX_FIFOS] =
460{DEFAULT_FIFO_0_LEN, [1 ...(MAX_TX_FIFOS - 1)] = DEFAULT_FIFO_1_7_LEN};
461static unsigned int rx_ring_sz[MAX_RX_RINGS] =
462{[0 ...(MAX_RX_RINGS - 1)] = SMALL_BLK_CNT};
463static unsigned int rts_frm_len[MAX_RX_RINGS] =
464{[0 ...(MAX_RX_RINGS - 1)] = 0 };
465
466module_param_array(tx_fifo_len, uint, NULL, 0);
467module_param_array(rx_ring_sz, uint, NULL, 0);
468module_param_array(rts_frm_len, uint, NULL, 0);
469
470/*
471 * S2IO device table.
472 * This table lists all the devices that this driver supports.
473 */
474static const struct pci_device_id s2io_tbl[] = {
475	{PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_WIN,
476	 PCI_ANY_ID, PCI_ANY_ID},
477	{PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_UNI,
478	 PCI_ANY_ID, PCI_ANY_ID},
479	{PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_WIN,
480	 PCI_ANY_ID, PCI_ANY_ID},
481	{PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_UNI,
482	 PCI_ANY_ID, PCI_ANY_ID},
483	{0,}
484};
485
486MODULE_DEVICE_TABLE(pci, s2io_tbl);
487
488static const struct pci_error_handlers s2io_err_handler = {
489	.error_detected = s2io_io_error_detected,
490	.slot_reset = s2io_io_slot_reset,
491	.resume = s2io_io_resume,
492};
493
494static struct pci_driver s2io_driver = {
495	.name = "S2IO",
496	.id_table = s2io_tbl,
497	.probe = s2io_init_nic,
498	.remove = s2io_rem_nic,
499	.err_handler = &s2io_err_handler,
500};
501
502/* A simplifier macro used both by init and free shared_mem Fns(). */
503#define TXD_MEM_PAGE_CNT(len, per_each) ((len+per_each - 1) / per_each)
504
505/* netqueue manipulation helper functions */
506static inline void s2io_stop_all_tx_queue(struct s2io_nic *sp)
507{
508	if (!sp->config.multiq) {
509		int i;
510
511		for (i = 0; i < sp->config.tx_fifo_num; i++)
512			sp->mac_control.fifos[i].queue_state = FIFO_QUEUE_STOP;
513	}
514	netif_tx_stop_all_queues(sp->dev);
515}
516
517static inline void s2io_stop_tx_queue(struct s2io_nic *sp, int fifo_no)
518{
519	if (!sp->config.multiq)
520		sp->mac_control.fifos[fifo_no].queue_state =
521			FIFO_QUEUE_STOP;
522
523	netif_tx_stop_all_queues(sp->dev);
524}
525
526static inline void s2io_start_all_tx_queue(struct s2io_nic *sp)
527{
528	if (!sp->config.multiq) {
529		int i;
530
531		for (i = 0; i < sp->config.tx_fifo_num; i++)
532			sp->mac_control.fifos[i].queue_state = FIFO_QUEUE_START;
533	}
534	netif_tx_start_all_queues(sp->dev);
535}
536
537static inline void s2io_wake_all_tx_queue(struct s2io_nic *sp)
538{
539	if (!sp->config.multiq) {
540		int i;
541
542		for (i = 0; i < sp->config.tx_fifo_num; i++)
543			sp->mac_control.fifos[i].queue_state = FIFO_QUEUE_START;
544	}
545	netif_tx_wake_all_queues(sp->dev);
546}
547
548static inline void s2io_wake_tx_queue(
549	struct fifo_info *fifo, int cnt, u8 multiq)
550{
551
552	if (multiq) {
553		if (cnt && __netif_subqueue_stopped(fifo->dev, fifo->fifo_no))
554			netif_wake_subqueue(fifo->dev, fifo->fifo_no);
555	} else if (cnt && (fifo->queue_state == FIFO_QUEUE_STOP)) {
556		if (netif_queue_stopped(fifo->dev)) {
557			fifo->queue_state = FIFO_QUEUE_START;
558			netif_wake_queue(fifo->dev);
559		}
560	}
561}
562
563/**
564 * init_shared_mem - Allocation and Initialization of Memory
565 * @nic: Device private variable.
566 * Description: The function allocates all the memory areas shared
567 * between the NIC and the driver. This includes Tx descriptors,
568 * Rx descriptors and the statistics block.
569 */
570
571static int init_shared_mem(struct s2io_nic *nic)
572{
573	u32 size;
574	void *tmp_v_addr, *tmp_v_addr_next;
575	dma_addr_t tmp_p_addr, tmp_p_addr_next;
576	struct RxD_block *pre_rxd_blk = NULL;
577	int i, j, blk_cnt;
578	int lst_size, lst_per_page;
579	struct net_device *dev = nic->dev;
580	unsigned long tmp;
581	struct buffAdd *ba;
582	struct config_param *config = &nic->config;
583	struct mac_info *mac_control = &nic->mac_control;
584	unsigned long long mem_allocated = 0;
585
586	/* Allocation and initialization of TXDLs in FIFOs */
587	size = 0;
588	for (i = 0; i < config->tx_fifo_num; i++) {
589		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
590
591		size += tx_cfg->fifo_len;
592	}
593	if (size > MAX_AVAILABLE_TXDS) {
594		DBG_PRINT(ERR_DBG,
595			  "Too many TxDs requested: %d, max supported: %d\n",
596			  size, MAX_AVAILABLE_TXDS);
597		return -EINVAL;
598	}
599
600	size = 0;
601	for (i = 0; i < config->tx_fifo_num; i++) {
602		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
603
604		size = tx_cfg->fifo_len;
605		/*
606		 * Legal values are from 2 to 8192
607		 */
608		if (size < 2) {
609			DBG_PRINT(ERR_DBG, "Fifo %d: Invalid length (%d) - "
610				  "Valid lengths are 2 through 8192\n",
611				  i, size);
612			return -EINVAL;
613		}
614	}
615
616	lst_size = (sizeof(struct TxD) * config->max_txds);
617	lst_per_page = PAGE_SIZE / lst_size;
618
619	for (i = 0; i < config->tx_fifo_num; i++) {
620		struct fifo_info *fifo = &mac_control->fifos[i];
621		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
622		int fifo_len = tx_cfg->fifo_len;
623		int list_holder_size = fifo_len * sizeof(struct list_info_hold);
624
625		fifo->list_info = kzalloc(list_holder_size, GFP_KERNEL);
626		if (!fifo->list_info) {
627			DBG_PRINT(INFO_DBG, "Malloc failed for list_info\n");
628			return -ENOMEM;
629		}
630		mem_allocated += list_holder_size;
631	}
632	for (i = 0; i < config->tx_fifo_num; i++) {
633		int page_num = TXD_MEM_PAGE_CNT(config->tx_cfg[i].fifo_len,
634						lst_per_page);
635		struct fifo_info *fifo = &mac_control->fifos[i];
636		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
637
638		fifo->tx_curr_put_info.offset = 0;
639		fifo->tx_curr_put_info.fifo_len = tx_cfg->fifo_len - 1;
640		fifo->tx_curr_get_info.offset = 0;
641		fifo->tx_curr_get_info.fifo_len = tx_cfg->fifo_len - 1;
642		fifo->fifo_no = i;
643		fifo->nic = nic;
644		fifo->max_txds = MAX_SKB_FRAGS + 2;
645		fifo->dev = dev;
646
647		for (j = 0; j < page_num; j++) {
648			int k = 0;
649			dma_addr_t tmp_p;
650			void *tmp_v;
651			tmp_v = pci_alloc_consistent(nic->pdev,
652						     PAGE_SIZE, &tmp_p);
653			if (!tmp_v) {
654				DBG_PRINT(INFO_DBG,
655					  "pci_alloc_consistent failed for TxDL\n");
656				return -ENOMEM;
657			}
658			/* If we got a zero DMA address(can happen on
659			 * certain platforms like PPC), reallocate.
660			 * Store virtual address of page we don't want,
661			 * to be freed later.
662			 */
663			if (!tmp_p) {
664				mac_control->zerodma_virt_addr = tmp_v;
665				DBG_PRINT(INIT_DBG,
666					  "%s: Zero DMA address for TxDL. "
667					  "Virtual address %p\n",
668					  dev->name, tmp_v);
669				tmp_v = pci_alloc_consistent(nic->pdev,
670							     PAGE_SIZE, &tmp_p);
671				if (!tmp_v) {
672					DBG_PRINT(INFO_DBG,
673						  "pci_alloc_consistent failed for TxDL\n");
674					return -ENOMEM;
675				}
676				mem_allocated += PAGE_SIZE;
677			}
678			while (k < lst_per_page) {
679				int l = (j * lst_per_page) + k;
680				if (l == tx_cfg->fifo_len)
681					break;
682				fifo->list_info[l].list_virt_addr =
683					tmp_v + (k * lst_size);
684				fifo->list_info[l].list_phy_addr =
685					tmp_p + (k * lst_size);
686				k++;
687			}
688		}
689	}
690
691	for (i = 0; i < config->tx_fifo_num; i++) {
692		struct fifo_info *fifo = &mac_control->fifos[i];
693		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
694
695		size = tx_cfg->fifo_len;
696		fifo->ufo_in_band_v = kcalloc(size, sizeof(u64), GFP_KERNEL);
697		if (!fifo->ufo_in_band_v)
698			return -ENOMEM;
699		mem_allocated += (size * sizeof(u64));
700	}
701
702	/* Allocation and initialization of RXDs in Rings */
703	size = 0;
704	for (i = 0; i < config->rx_ring_num; i++) {
705		struct rx_ring_config *rx_cfg = &config->rx_cfg[i];
706		struct ring_info *ring = &mac_control->rings[i];
707
708		if (rx_cfg->num_rxd % (rxd_count[nic->rxd_mode] + 1)) {
709			DBG_PRINT(ERR_DBG, "%s: Ring%d RxD count is not a "
710				  "multiple of RxDs per Block\n",
711				  dev->name, i);
712			return FAILURE;
713		}
714		size += rx_cfg->num_rxd;
715		ring->block_count = rx_cfg->num_rxd /
716			(rxd_count[nic->rxd_mode] + 1);
717		ring->pkt_cnt = rx_cfg->num_rxd - ring->block_count;
718	}
719	if (nic->rxd_mode == RXD_MODE_1)
720		size = (size * (sizeof(struct RxD1)));
721	else
722		size = (size * (sizeof(struct RxD3)));
723
724	for (i = 0; i < config->rx_ring_num; i++) {
725		struct rx_ring_config *rx_cfg = &config->rx_cfg[i];
726		struct ring_info *ring = &mac_control->rings[i];
727
728		ring->rx_curr_get_info.block_index = 0;
729		ring->rx_curr_get_info.offset = 0;
730		ring->rx_curr_get_info.ring_len = rx_cfg->num_rxd - 1;
731		ring->rx_curr_put_info.block_index = 0;
732		ring->rx_curr_put_info.offset = 0;
733		ring->rx_curr_put_info.ring_len = rx_cfg->num_rxd - 1;
734		ring->nic = nic;
735		ring->ring_no = i;
736
737		blk_cnt = rx_cfg->num_rxd / (rxd_count[nic->rxd_mode] + 1);
738		/*  Allocating all the Rx blocks */
739		for (j = 0; j < blk_cnt; j++) {
740			struct rx_block_info *rx_blocks;
741			int l;
742
743			rx_blocks = &ring->rx_blocks[j];
744			size = SIZE_OF_BLOCK;	/* size is always page size */
745			tmp_v_addr = pci_alloc_consistent(nic->pdev, size,
746							  &tmp_p_addr);
747			if (tmp_v_addr == NULL) {
748				/*
749				 * In case of failure, free_shared_mem()
750				 * is called, which should free any
751				 * memory that was alloced till the
752				 * failure happened.
753				 */
754				rx_blocks->block_virt_addr = tmp_v_addr;
755				return -ENOMEM;
756			}
757			mem_allocated += size;
758			memset(tmp_v_addr, 0, size);
759
760			size = sizeof(struct rxd_info) *
761				rxd_count[nic->rxd_mode];
762			rx_blocks->block_virt_addr = tmp_v_addr;
763			rx_blocks->block_dma_addr = tmp_p_addr;
764			rx_blocks->rxds = kmalloc(size,  GFP_KERNEL);
765			if (!rx_blocks->rxds)
766				return -ENOMEM;
767			mem_allocated += size;
768			for (l = 0; l < rxd_count[nic->rxd_mode]; l++) {
769				rx_blocks->rxds[l].virt_addr =
770					rx_blocks->block_virt_addr +
771					(rxd_size[nic->rxd_mode] * l);
772				rx_blocks->rxds[l].dma_addr =
773					rx_blocks->block_dma_addr +
774					(rxd_size[nic->rxd_mode] * l);
775			}
776		}
777		/* Interlinking all Rx Blocks */
778		for (j = 0; j < blk_cnt; j++) {
779			int next = (j + 1) % blk_cnt;
780			tmp_v_addr = ring->rx_blocks[j].block_virt_addr;
781			tmp_v_addr_next = ring->rx_blocks[next].block_virt_addr;
782			tmp_p_addr = ring->rx_blocks[j].block_dma_addr;
783			tmp_p_addr_next = ring->rx_blocks[next].block_dma_addr;
784
785			pre_rxd_blk = tmp_v_addr;
786			pre_rxd_blk->reserved_2_pNext_RxD_block =
787				(unsigned long)tmp_v_addr_next;
788			pre_rxd_blk->pNext_RxD_Blk_physical =
789				(u64)tmp_p_addr_next;
790		}
791	}
792	if (nic->rxd_mode == RXD_MODE_3B) {
793		/*
794		 * Allocation of Storages for buffer addresses in 2BUFF mode
795		 * and the buffers as well.
796		 */
797		for (i = 0; i < config->rx_ring_num; i++) {
798			struct rx_ring_config *rx_cfg = &config->rx_cfg[i];
799			struct ring_info *ring = &mac_control->rings[i];
800
801			blk_cnt = rx_cfg->num_rxd /
802				(rxd_count[nic->rxd_mode] + 1);
803			size = sizeof(struct buffAdd *) * blk_cnt;
804			ring->ba = kmalloc(size, GFP_KERNEL);
805			if (!ring->ba)
806				return -ENOMEM;
807			mem_allocated += size;
808			for (j = 0; j < blk_cnt; j++) {
809				int k = 0;
810
811				size = sizeof(struct buffAdd) *
812					(rxd_count[nic->rxd_mode] + 1);
813				ring->ba[j] = kmalloc(size, GFP_KERNEL);
814				if (!ring->ba[j])
815					return -ENOMEM;
816				mem_allocated += size;
817				while (k != rxd_count[nic->rxd_mode]) {
818					ba = &ring->ba[j][k];
819					size = BUF0_LEN + ALIGN_SIZE;
820					ba->ba_0_org = kmalloc(size, GFP_KERNEL);
821					if (!ba->ba_0_org)
822						return -ENOMEM;
823					mem_allocated += size;
824					tmp = (unsigned long)ba->ba_0_org;
825					tmp += ALIGN_SIZE;
826					tmp &= ~((unsigned long)ALIGN_SIZE);
827					ba->ba_0 = (void *)tmp;
828
829					size = BUF1_LEN + ALIGN_SIZE;
830					ba->ba_1_org = kmalloc(size, GFP_KERNEL);
831					if (!ba->ba_1_org)
832						return -ENOMEM;
833					mem_allocated += size;
834					tmp = (unsigned long)ba->ba_1_org;
835					tmp += ALIGN_SIZE;
836					tmp &= ~((unsigned long)ALIGN_SIZE);
837					ba->ba_1 = (void *)tmp;
838					k++;
839				}
840			}
841		}
842	}
843
844	/* Allocation and initialization of Statistics block */
845	size = sizeof(struct stat_block);
846	mac_control->stats_mem =
847		pci_alloc_consistent(nic->pdev, size,
848				     &mac_control->stats_mem_phy);
849
850	if (!mac_control->stats_mem) {
851		/*
852		 * In case of failure, free_shared_mem() is called, which
853		 * should free any memory that was alloced till the
854		 * failure happened.
855		 */
856		return -ENOMEM;
857	}
858	mem_allocated += size;
859	mac_control->stats_mem_sz = size;
860
861	tmp_v_addr = mac_control->stats_mem;
862	mac_control->stats_info = tmp_v_addr;
863	memset(tmp_v_addr, 0, size);
864	DBG_PRINT(INIT_DBG, "%s: Ring Mem PHY: 0x%llx\n",
865		dev_name(&nic->pdev->dev), (unsigned long long)tmp_p_addr);
866	mac_control->stats_info->sw_stat.mem_allocated += mem_allocated;
867	return SUCCESS;
868}
869
870/**
871 * free_shared_mem - Free the allocated Memory
872 * @nic:  Device private variable.
873 * Description: This function is to free all memory locations allocated by
874 * the init_shared_mem() function and return it to the kernel.
875 */
876
877static void free_shared_mem(struct s2io_nic *nic)
878{
879	int i, j, blk_cnt, size;
880	void *tmp_v_addr;
881	dma_addr_t tmp_p_addr;
882	int lst_size, lst_per_page;
883	struct net_device *dev;
884	int page_num = 0;
885	struct config_param *config;
886	struct mac_info *mac_control;
887	struct stat_block *stats;
888	struct swStat *swstats;
889
890	if (!nic)
891		return;
892
893	dev = nic->dev;
894
895	config = &nic->config;
896	mac_control = &nic->mac_control;
897	stats = mac_control->stats_info;
898	swstats = &stats->sw_stat;
899
900	lst_size = sizeof(struct TxD) * config->max_txds;
901	lst_per_page = PAGE_SIZE / lst_size;
902
903	for (i = 0; i < config->tx_fifo_num; i++) {
904		struct fifo_info *fifo = &mac_control->fifos[i];
905		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
906
907		page_num = TXD_MEM_PAGE_CNT(tx_cfg->fifo_len, lst_per_page);
908		for (j = 0; j < page_num; j++) {
909			int mem_blks = (j * lst_per_page);
910			struct list_info_hold *fli;
911
912			if (!fifo->list_info)
913				return;
914
915			fli = &fifo->list_info[mem_blks];
916			if (!fli->list_virt_addr)
917				break;
918			pci_free_consistent(nic->pdev, PAGE_SIZE,
919					    fli->list_virt_addr,
920					    fli->list_phy_addr);
921			swstats->mem_freed += PAGE_SIZE;
922		}
923		/* If we got a zero DMA address during allocation,
924		 * free the page now
925		 */
926		if (mac_control->zerodma_virt_addr) {
927			pci_free_consistent(nic->pdev, PAGE_SIZE,
928					    mac_control->zerodma_virt_addr,
929					    (dma_addr_t)0);
930			DBG_PRINT(INIT_DBG,
931				  "%s: Freeing TxDL with zero DMA address. "
932				  "Virtual address %p\n",
933				  dev->name, mac_control->zerodma_virt_addr);
934			swstats->mem_freed += PAGE_SIZE;
935		}
936		kfree(fifo->list_info);
937		swstats->mem_freed += tx_cfg->fifo_len *
938			sizeof(struct list_info_hold);
939	}
940
941	size = SIZE_OF_BLOCK;
942	for (i = 0; i < config->rx_ring_num; i++) {
943		struct ring_info *ring = &mac_control->rings[i];
944
945		blk_cnt = ring->block_count;
946		for (j = 0; j < blk_cnt; j++) {
947			tmp_v_addr = ring->rx_blocks[j].block_virt_addr;
948			tmp_p_addr = ring->rx_blocks[j].block_dma_addr;
949			if (tmp_v_addr == NULL)
950				break;
951			pci_free_consistent(nic->pdev, size,
952					    tmp_v_addr, tmp_p_addr);
953			swstats->mem_freed += size;
954			kfree(ring->rx_blocks[j].rxds);
955			swstats->mem_freed += sizeof(struct rxd_info) *
956				rxd_count[nic->rxd_mode];
957		}
958	}
959
960	if (nic->rxd_mode == RXD_MODE_3B) {
961		/* Freeing buffer storage addresses in 2BUFF mode. */
962		for (i = 0; i < config->rx_ring_num; i++) {
963			struct rx_ring_config *rx_cfg = &config->rx_cfg[i];
964			struct ring_info *ring = &mac_control->rings[i];
965
966			blk_cnt = rx_cfg->num_rxd /
967				(rxd_count[nic->rxd_mode] + 1);
968			for (j = 0; j < blk_cnt; j++) {
969				int k = 0;
970				if (!ring->ba[j])
971					continue;
972				while (k != rxd_count[nic->rxd_mode]) {
973					struct buffAdd *ba = &ring->ba[j][k];
974					kfree(ba->ba_0_org);
975					swstats->mem_freed +=
976						BUF0_LEN + ALIGN_SIZE;
977					kfree(ba->ba_1_org);
978					swstats->mem_freed +=
979						BUF1_LEN + ALIGN_SIZE;
980					k++;
981				}
982				kfree(ring->ba[j]);
983				swstats->mem_freed += sizeof(struct buffAdd) *
984					(rxd_count[nic->rxd_mode] + 1);
985			}
986			kfree(ring->ba);
987			swstats->mem_freed += sizeof(struct buffAdd *) *
988				blk_cnt;
989		}
990	}
991
992	for (i = 0; i < nic->config.tx_fifo_num; i++) {
993		struct fifo_info *fifo = &mac_control->fifos[i];
994		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
995
996		if (fifo->ufo_in_band_v) {
997			swstats->mem_freed += tx_cfg->fifo_len *
998				sizeof(u64);
999			kfree(fifo->ufo_in_band_v);
1000		}
1001	}
1002
1003	if (mac_control->stats_mem) {
1004		swstats->mem_freed += mac_control->stats_mem_sz;
1005		pci_free_consistent(nic->pdev,
1006				    mac_control->stats_mem_sz,
1007				    mac_control->stats_mem,
1008				    mac_control->stats_mem_phy);
1009	}
1010}
1011
1012/**
1013 * s2io_verify_pci_mode -
1014 */
1015
1016static int s2io_verify_pci_mode(struct s2io_nic *nic)
1017{
1018	struct XENA_dev_config __iomem *bar0 = nic->bar0;
1019	register u64 val64 = 0;
1020	int     mode;
1021
1022	val64 = readq(&bar0->pci_mode);
1023	mode = (u8)GET_PCI_MODE(val64);
1024
1025	if (val64 & PCI_MODE_UNKNOWN_MODE)
1026		return -1;      /* Unknown PCI mode */
1027	return mode;
1028}
1029
1030#define NEC_VENID   0x1033
1031#define NEC_DEVID   0x0125
1032static int s2io_on_nec_bridge(struct pci_dev *s2io_pdev)
1033{
1034	struct pci_dev *tdev = NULL;
1035	for_each_pci_dev(tdev) {
1036		if (tdev->vendor == NEC_VENID && tdev->device == NEC_DEVID) {
1037			if (tdev->bus == s2io_pdev->bus->parent) {
1038				pci_dev_put(tdev);
1039				return 1;
1040			}
1041		}
1042	}
1043	return 0;
1044}
1045
1046static int bus_speed[8] = {33, 133, 133, 200, 266, 133, 200, 266};
1047/**
1048 * s2io_print_pci_mode -
1049 */
1050static int s2io_print_pci_mode(struct s2io_nic *nic)
1051{
1052	struct XENA_dev_config __iomem *bar0 = nic->bar0;
1053	register u64 val64 = 0;
1054	int	mode;
1055	struct config_param *config = &nic->config;
1056	const char *pcimode;
1057
1058	val64 = readq(&bar0->pci_mode);
1059	mode = (u8)GET_PCI_MODE(val64);
1060
1061	if (val64 & PCI_MODE_UNKNOWN_MODE)
1062		return -1;	/* Unknown PCI mode */
1063
1064	config->bus_speed = bus_speed[mode];
1065
1066	if (s2io_on_nec_bridge(nic->pdev)) {
1067		DBG_PRINT(ERR_DBG, "%s: Device is on PCI-E bus\n",
1068			  nic->dev->name);
1069		return mode;
1070	}
1071
1072	switch (mode) {
1073	case PCI_MODE_PCI_33:
1074		pcimode = "33MHz PCI bus";
1075		break;
1076	case PCI_MODE_PCI_66:
1077		pcimode = "66MHz PCI bus";
1078		break;
1079	case PCI_MODE_PCIX_M1_66:
1080		pcimode = "66MHz PCIX(M1) bus";
1081		break;
1082	case PCI_MODE_PCIX_M1_100:
1083		pcimode = "100MHz PCIX(M1) bus";
1084		break;
1085	case PCI_MODE_PCIX_M1_133:
1086		pcimode = "133MHz PCIX(M1) bus";
1087		break;
1088	case PCI_MODE_PCIX_M2_66:
1089		pcimode = "133MHz PCIX(M2) bus";
1090		break;
1091	case PCI_MODE_PCIX_M2_100:
1092		pcimode = "200MHz PCIX(M2) bus";
1093		break;
1094	case PCI_MODE_PCIX_M2_133:
1095		pcimode = "266MHz PCIX(M2) bus";
1096		break;
1097	default:
1098		pcimode = "unsupported bus!";
1099		mode = -1;
1100	}
1101
1102	DBG_PRINT(ERR_DBG, "%s: Device is on %d bit %s\n",
1103		  nic->dev->name, val64 & PCI_MODE_32_BITS ? 32 : 64, pcimode);
1104
1105	return mode;
1106}
1107
1108/**
1109 *  init_tti - Initialization transmit traffic interrupt scheme
1110 *  @nic: device private variable
1111 *  @link: link status (UP/DOWN) used to enable/disable continuous
1112 *  transmit interrupts
1113 *  Description: The function configures transmit traffic interrupts
1114 *  Return Value:  SUCCESS on success and
1115 *  '-1' on failure
1116 */
1117
1118static int init_tti(struct s2io_nic *nic, int link)
1119{
1120	struct XENA_dev_config __iomem *bar0 = nic->bar0;
1121	register u64 val64 = 0;
1122	int i;
1123	struct config_param *config = &nic->config;
1124
1125	for (i = 0; i < config->tx_fifo_num; i++) {
1126		/*
1127		 * TTI Initialization. Default Tx timer gets us about
1128		 * 250 interrupts per sec. Continuous interrupts are enabled
1129		 * by default.
1130		 */
1131		if (nic->device_type == XFRAME_II_DEVICE) {
1132			int count = (nic->config.bus_speed * 125)/2;
1133			val64 = TTI_DATA1_MEM_TX_TIMER_VAL(count);
1134		} else
1135			val64 = TTI_DATA1_MEM_TX_TIMER_VAL(0x2078);
1136
1137		val64 |= TTI_DATA1_MEM_TX_URNG_A(0xA) |
1138			TTI_DATA1_MEM_TX_URNG_B(0x10) |
1139			TTI_DATA1_MEM_TX_URNG_C(0x30) |
1140			TTI_DATA1_MEM_TX_TIMER_AC_EN;
1141		if (i == 0)
1142			if (use_continuous_tx_intrs && (link == LINK_UP))
1143				val64 |= TTI_DATA1_MEM_TX_TIMER_CI_EN;
1144		writeq(val64, &bar0->tti_data1_mem);
1145
1146		if (nic->config.intr_type == MSI_X) {
1147			val64 = TTI_DATA2_MEM_TX_UFC_A(0x10) |
1148				TTI_DATA2_MEM_TX_UFC_B(0x100) |
1149				TTI_DATA2_MEM_TX_UFC_C(0x200) |
1150				TTI_DATA2_MEM_TX_UFC_D(0x300);
1151		} else {
1152			if ((nic->config.tx_steering_type ==
1153			     TX_DEFAULT_STEERING) &&
1154			    (config->tx_fifo_num > 1) &&
1155			    (i >= nic->udp_fifo_idx) &&
1156			    (i < (nic->udp_fifo_idx +
1157				  nic->total_udp_fifos)))
1158				val64 = TTI_DATA2_MEM_TX_UFC_A(0x50) |
1159					TTI_DATA2_MEM_TX_UFC_B(0x80) |
1160					TTI_DATA2_MEM_TX_UFC_C(0x100) |
1161					TTI_DATA2_MEM_TX_UFC_D(0x120);
1162			else
1163				val64 = TTI_DATA2_MEM_TX_UFC_A(0x10) |
1164					TTI_DATA2_MEM_TX_UFC_B(0x20) |
1165					TTI_DATA2_MEM_TX_UFC_C(0x40) |
1166					TTI_DATA2_MEM_TX_UFC_D(0x80);
1167		}
1168
1169		writeq(val64, &bar0->tti_data2_mem);
1170
1171		val64 = TTI_CMD_MEM_WE |
1172			TTI_CMD_MEM_STROBE_NEW_CMD |
1173			TTI_CMD_MEM_OFFSET(i);
1174		writeq(val64, &bar0->tti_command_mem);
1175
1176		if (wait_for_cmd_complete(&bar0->tti_command_mem,
1177					  TTI_CMD_MEM_STROBE_NEW_CMD,
1178					  S2IO_BIT_RESET) != SUCCESS)
1179			return FAILURE;
1180	}
1181
1182	return SUCCESS;
1183}
1184
1185/**
1186 *  init_nic - Initialization of hardware
1187 *  @nic: device private variable
1188 *  Description: The function sequentially configures every block
1189 *  of the H/W from their reset values.
1190 *  Return Value:  SUCCESS on success and
1191 *  '-1' on failure (endian settings incorrect).
1192 */
1193
1194static int init_nic(struct s2io_nic *nic)
1195{
1196	struct XENA_dev_config __iomem *bar0 = nic->bar0;
1197	struct net_device *dev = nic->dev;
1198	register u64 val64 = 0;
1199	void __iomem *add;
1200	u32 time;
1201	int i, j;
1202	int dtx_cnt = 0;
1203	unsigned long long mem_share;
1204	int mem_size;
1205	struct config_param *config = &nic->config;
1206	struct mac_info *mac_control = &nic->mac_control;
1207
1208	/* to set the swapper controle on the card */
1209	if (s2io_set_swapper(nic)) {
1210		DBG_PRINT(ERR_DBG, "ERROR: Setting Swapper failed\n");
1211		return -EIO;
1212	}
1213
1214	/*
1215	 * Herc requires EOI to be removed from reset before XGXS, so..
1216	 */
1217	if (nic->device_type & XFRAME_II_DEVICE) {
1218		val64 = 0xA500000000ULL;
1219		writeq(val64, &bar0->sw_reset);
1220		msleep(500);
1221		val64 = readq(&bar0->sw_reset);
1222	}
1223
1224	/* Remove XGXS from reset state */
1225	val64 = 0;
1226	writeq(val64, &bar0->sw_reset);
1227	msleep(500);
1228	val64 = readq(&bar0->sw_reset);
1229
1230	/* Ensure that it's safe to access registers by checking
1231	 * RIC_RUNNING bit is reset. Check is valid only for XframeII.
1232	 */
1233	if (nic->device_type == XFRAME_II_DEVICE) {
1234		for (i = 0; i < 50; i++) {
1235			val64 = readq(&bar0->adapter_status);
1236			if (!(val64 & ADAPTER_STATUS_RIC_RUNNING))
1237				break;
1238			msleep(10);
1239		}
1240		if (i == 50)
1241			return -ENODEV;
1242	}
1243
1244	/*  Enable Receiving broadcasts */
1245	add = &bar0->mac_cfg;
1246	val64 = readq(&bar0->mac_cfg);
1247	val64 |= MAC_RMAC_BCAST_ENABLE;
1248	writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1249	writel((u32)val64, add);
1250	writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1251	writel((u32) (val64 >> 32), (add + 4));
1252
1253	/* Read registers in all blocks */
1254	val64 = readq(&bar0->mac_int_mask);
1255	val64 = readq(&bar0->mc_int_mask);
1256	val64 = readq(&bar0->xgxs_int_mask);
1257
1258	/*  Set MTU */
1259	val64 = dev->mtu;
1260	writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
1261
1262	if (nic->device_type & XFRAME_II_DEVICE) {
1263		while (herc_act_dtx_cfg[dtx_cnt] != END_SIGN) {
1264			SPECIAL_REG_WRITE(herc_act_dtx_cfg[dtx_cnt],
1265					  &bar0->dtx_control, UF);
1266			if (dtx_cnt & 0x1)
1267				msleep(1); /* Necessary!! */
1268			dtx_cnt++;
1269		}
1270	} else {
1271		while (xena_dtx_cfg[dtx_cnt] != END_SIGN) {
1272			SPECIAL_REG_WRITE(xena_dtx_cfg[dtx_cnt],
1273					  &bar0->dtx_control, UF);
1274			val64 = readq(&bar0->dtx_control);
1275			dtx_cnt++;
1276		}
1277	}
1278
1279	/*  Tx DMA Initialization */
1280	val64 = 0;
1281	writeq(val64, &bar0->tx_fifo_partition_0);
1282	writeq(val64, &bar0->tx_fifo_partition_1);
1283	writeq(val64, &bar0->tx_fifo_partition_2);
1284	writeq(val64, &bar0->tx_fifo_partition_3);
1285
1286	for (i = 0, j = 0; i < config->tx_fifo_num; i++) {
1287		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
1288
1289		val64 |= vBIT(tx_cfg->fifo_len - 1, ((j * 32) + 19), 13) |
1290			vBIT(tx_cfg->fifo_priority, ((j * 32) + 5), 3);
1291
1292		if (i == (config->tx_fifo_num - 1)) {
1293			if (i % 2 == 0)
1294				i++;
1295		}
1296
1297		switch (i) {
1298		case 1:
1299			writeq(val64, &bar0->tx_fifo_partition_0);
1300			val64 = 0;
1301			j = 0;
1302			break;
1303		case 3:
1304			writeq(val64, &bar0->tx_fifo_partition_1);
1305			val64 = 0;
1306			j = 0;
1307			break;
1308		case 5:
1309			writeq(val64, &bar0->tx_fifo_partition_2);
1310			val64 = 0;
1311			j = 0;
1312			break;
1313		case 7:
1314			writeq(val64, &bar0->tx_fifo_partition_3);
1315			val64 = 0;
1316			j = 0;
1317			break;
1318		default:
1319			j++;
1320			break;
1321		}
1322	}
1323
1324	/*
1325	 * Disable 4 PCCs for Xena1, 2 and 3 as per H/W bug
1326	 * SXE-008 TRANSMIT DMA ARBITRATION ISSUE.
1327	 */
1328	if ((nic->device_type == XFRAME_I_DEVICE) && (nic->pdev->revision < 4))
1329		writeq(PCC_ENABLE_FOUR, &bar0->pcc_enable);
1330
1331	val64 = readq(&bar0->tx_fifo_partition_0);
1332	DBG_PRINT(INIT_DBG, "Fifo partition at: 0x%p is: 0x%llx\n",
1333		  &bar0->tx_fifo_partition_0, (unsigned long long)val64);
1334
1335	/*
1336	 * Initialization of Tx_PA_CONFIG register to ignore packet
1337	 * integrity checking.
1338	 */
1339	val64 = readq(&bar0->tx_pa_cfg);
1340	val64 |= TX_PA_CFG_IGNORE_FRM_ERR |
1341		TX_PA_CFG_IGNORE_SNAP_OUI |
1342		TX_PA_CFG_IGNORE_LLC_CTRL |
1343		TX_PA_CFG_IGNORE_L2_ERR;
1344	writeq(val64, &bar0->tx_pa_cfg);
1345
1346	/* Rx DMA intialization. */
1347	val64 = 0;
1348	for (i = 0; i < config->rx_ring_num; i++) {
1349		struct rx_ring_config *rx_cfg = &config->rx_cfg[i];
1350
1351		val64 |= vBIT(rx_cfg->ring_priority, (5 + (i * 8)), 3);
1352	}
1353	writeq(val64, &bar0->rx_queue_priority);
1354
1355	/*
1356	 * Allocating equal share of memory to all the
1357	 * configured Rings.
1358	 */
1359	val64 = 0;
1360	if (nic->device_type & XFRAME_II_DEVICE)
1361		mem_size = 32;
1362	else
1363		mem_size = 64;
1364
1365	for (i = 0; i < config->rx_ring_num; i++) {
1366		switch (i) {
1367		case 0:
1368			mem_share = (mem_size / config->rx_ring_num +
1369				     mem_size % config->rx_ring_num);
1370			val64 |= RX_QUEUE_CFG_Q0_SZ(mem_share);
1371			continue;
1372		case 1:
1373			mem_share = (mem_size / config->rx_ring_num);
1374			val64 |= RX_QUEUE_CFG_Q1_SZ(mem_share);
1375			continue;
1376		case 2:
1377			mem_share = (mem_size / config->rx_ring_num);
1378			val64 |= RX_QUEUE_CFG_Q2_SZ(mem_share);
1379			continue;
1380		case 3:
1381			mem_share = (mem_size / config->rx_ring_num);
1382			val64 |= RX_QUEUE_CFG_Q3_SZ(mem_share);
1383			continue;
1384		case 4:
1385			mem_share = (mem_size / config->rx_ring_num);
1386			val64 |= RX_QUEUE_CFG_Q4_SZ(mem_share);
1387			continue;
1388		case 5:
1389			mem_share = (mem_size / config->rx_ring_num);
1390			val64 |= RX_QUEUE_CFG_Q5_SZ(mem_share);
1391			continue;
1392		case 6:
1393			mem_share = (mem_size / config->rx_ring_num);
1394			val64 |= RX_QUEUE_CFG_Q6_SZ(mem_share);
1395			continue;
1396		case 7:
1397			mem_share = (mem_size / config->rx_ring_num);
1398			val64 |= RX_QUEUE_CFG_Q7_SZ(mem_share);
1399			continue;
1400		}
1401	}
1402	writeq(val64, &bar0->rx_queue_cfg);
1403
1404	/*
1405	 * Filling Tx round robin registers
1406	 * as per the number of FIFOs for equal scheduling priority
1407	 */
1408	switch (config->tx_fifo_num) {
1409	case 1:
1410		val64 = 0x0;
1411		writeq(val64, &bar0->tx_w_round_robin_0);
1412		writeq(val64, &bar0->tx_w_round_robin_1);
1413		writeq(val64, &bar0->tx_w_round_robin_2);
1414		writeq(val64, &bar0->tx_w_round_robin_3);
1415		writeq(val64, &bar0->tx_w_round_robin_4);
1416		break;
1417	case 2:
1418		val64 = 0x0001000100010001ULL;
1419		writeq(val64, &bar0->tx_w_round_robin_0);
1420		writeq(val64, &bar0->tx_w_round_robin_1);
1421		writeq(val64, &bar0->tx_w_round_robin_2);
1422		writeq(val64, &bar0->tx_w_round_robin_3);
1423		val64 = 0x0001000100000000ULL;
1424		writeq(val64, &bar0->tx_w_round_robin_4);
1425		break;
1426	case 3:
1427		val64 = 0x0001020001020001ULL;
1428		writeq(val64, &bar0->tx_w_round_robin_0);
1429		val64 = 0x0200010200010200ULL;
1430		writeq(val64, &bar0->tx_w_round_robin_1);
1431		val64 = 0x0102000102000102ULL;
1432		writeq(val64, &bar0->tx_w_round_robin_2);
1433		val64 = 0x0001020001020001ULL;
1434		writeq(val64, &bar0->tx_w_round_robin_3);
1435		val64 = 0x0200010200000000ULL;
1436		writeq(val64, &bar0->tx_w_round_robin_4);
1437		break;
1438	case 4:
1439		val64 = 0x0001020300010203ULL;
1440		writeq(val64, &bar0->tx_w_round_robin_0);
1441		writeq(val64, &bar0->tx_w_round_robin_1);
1442		writeq(val64, &bar0->tx_w_round_robin_2);
1443		writeq(val64, &bar0->tx_w_round_robin_3);
1444		val64 = 0x0001020300000000ULL;
1445		writeq(val64, &bar0->tx_w_round_robin_4);
1446		break;
1447	case 5:
1448		val64 = 0x0001020304000102ULL;
1449		writeq(val64, &bar0->tx_w_round_robin_0);
1450		val64 = 0x0304000102030400ULL;
1451		writeq(val64, &bar0->tx_w_round_robin_1);
1452		val64 = 0x0102030400010203ULL;
1453		writeq(val64, &bar0->tx_w_round_robin_2);
1454		val64 = 0x0400010203040001ULL;
1455		writeq(val64, &bar0->tx_w_round_robin_3);
1456		val64 = 0x0203040000000000ULL;
1457		writeq(val64, &bar0->tx_w_round_robin_4);
1458		break;
1459	case 6:
1460		val64 = 0x0001020304050001ULL;
1461		writeq(val64, &bar0->tx_w_round_robin_0);
1462		val64 = 0x0203040500010203ULL;
1463		writeq(val64, &bar0->tx_w_round_robin_1);
1464		val64 = 0x0405000102030405ULL;
1465		writeq(val64, &bar0->tx_w_round_robin_2);
1466		val64 = 0x0001020304050001ULL;
1467		writeq(val64, &bar0->tx_w_round_robin_3);
1468		val64 = 0x0203040500000000ULL;
1469		writeq(val64, &bar0->tx_w_round_robin_4);
1470		break;
1471	case 7:
1472		val64 = 0x0001020304050600ULL;
1473		writeq(val64, &bar0->tx_w_round_robin_0);
1474		val64 = 0x0102030405060001ULL;
1475		writeq(val64, &bar0->tx_w_round_robin_1);
1476		val64 = 0x0203040506000102ULL;
1477		writeq(val64, &bar0->tx_w_round_robin_2);
1478		val64 = 0x0304050600010203ULL;
1479		writeq(val64, &bar0->tx_w_round_robin_3);
1480		val64 = 0x0405060000000000ULL;
1481		writeq(val64, &bar0->tx_w_round_robin_4);
1482		break;
1483	case 8:
1484		val64 = 0x0001020304050607ULL;
1485		writeq(val64, &bar0->tx_w_round_robin_0);
1486		writeq(val64, &bar0->tx_w_round_robin_1);
1487		writeq(val64, &bar0->tx_w_round_robin_2);
1488		writeq(val64, &bar0->tx_w_round_robin_3);
1489		val64 = 0x0001020300000000ULL;
1490		writeq(val64, &bar0->tx_w_round_robin_4);
1491		break;
1492	}
1493
1494	/* Enable all configured Tx FIFO partitions */
1495	val64 = readq(&bar0->tx_fifo_partition_0);
1496	val64 |= (TX_FIFO_PARTITION_EN);
1497	writeq(val64, &bar0->tx_fifo_partition_0);
1498
1499	/* Filling the Rx round robin registers as per the
1500	 * number of Rings and steering based on QoS with
1501	 * equal priority.
1502	 */
1503	switch (config->rx_ring_num) {
1504	case 1:
1505		val64 = 0x0;
1506		writeq(val64, &bar0->rx_w_round_robin_0);
1507		writeq(val64, &bar0->rx_w_round_robin_1);
1508		writeq(val64, &bar0->rx_w_round_robin_2);
1509		writeq(val64, &bar0->rx_w_round_robin_3);
1510		writeq(val64, &bar0->rx_w_round_robin_4);
1511
1512		val64 = 0x8080808080808080ULL;
1513		writeq(val64, &bar0->rts_qos_steering);
1514		break;
1515	case 2:
1516		val64 = 0x0001000100010001ULL;
1517		writeq(val64, &bar0->rx_w_round_robin_0);
1518		writeq(val64, &bar0->rx_w_round_robin_1);
1519		writeq(val64, &bar0->rx_w_round_robin_2);
1520		writeq(val64, &bar0->rx_w_round_robin_3);
1521		val64 = 0x0001000100000000ULL;
1522		writeq(val64, &bar0->rx_w_round_robin_4);
1523
1524		val64 = 0x8080808040404040ULL;
1525		writeq(val64, &bar0->rts_qos_steering);
1526		break;
1527	case 3:
1528		val64 = 0x0001020001020001ULL;
1529		writeq(val64, &bar0->rx_w_round_robin_0);
1530		val64 = 0x0200010200010200ULL;
1531		writeq(val64, &bar0->rx_w_round_robin_1);
1532		val64 = 0x0102000102000102ULL;
1533		writeq(val64, &bar0->rx_w_round_robin_2);
1534		val64 = 0x0001020001020001ULL;
1535		writeq(val64, &bar0->rx_w_round_robin_3);
1536		val64 = 0x0200010200000000ULL;
1537		writeq(val64, &bar0->rx_w_round_robin_4);
1538
1539		val64 = 0x8080804040402020ULL;
1540		writeq(val64, &bar0->rts_qos_steering);
1541		break;
1542	case 4:
1543		val64 = 0x0001020300010203ULL;
1544		writeq(val64, &bar0->rx_w_round_robin_0);
1545		writeq(val64, &bar0->rx_w_round_robin_1);
1546		writeq(val64, &bar0->rx_w_round_robin_2);
1547		writeq(val64, &bar0->rx_w_round_robin_3);
1548		val64 = 0x0001020300000000ULL;
1549		writeq(val64, &bar0->rx_w_round_robin_4);
1550
1551		val64 = 0x8080404020201010ULL;
1552		writeq(val64, &bar0->rts_qos_steering);
1553		break;
1554	case 5:
1555		val64 = 0x0001020304000102ULL;
1556		writeq(val64, &bar0->rx_w_round_robin_0);
1557		val64 = 0x0304000102030400ULL;
1558		writeq(val64, &bar0->rx_w_round_robin_1);
1559		val64 = 0x0102030400010203ULL;
1560		writeq(val64, &bar0->rx_w_round_robin_2);
1561		val64 = 0x0400010203040001ULL;
1562		writeq(val64, &bar0->rx_w_round_robin_3);
1563		val64 = 0x0203040000000000ULL;
1564		writeq(val64, &bar0->rx_w_round_robin_4);
1565
1566		val64 = 0x8080404020201008ULL;
1567		writeq(val64, &bar0->rts_qos_steering);
1568		break;
1569	case 6:
1570		val64 = 0x0001020304050001ULL;
1571		writeq(val64, &bar0->rx_w_round_robin_0);
1572		val64 = 0x0203040500010203ULL;
1573		writeq(val64, &bar0->rx_w_round_robin_1);
1574		val64 = 0x0405000102030405ULL;
1575		writeq(val64, &bar0->rx_w_round_robin_2);
1576		val64 = 0x0001020304050001ULL;
1577		writeq(val64, &bar0->rx_w_round_robin_3);
1578		val64 = 0x0203040500000000ULL;
1579		writeq(val64, &bar0->rx_w_round_robin_4);
1580
1581		val64 = 0x8080404020100804ULL;
1582		writeq(val64, &bar0->rts_qos_steering);
1583		break;
1584	case 7:
1585		val64 = 0x0001020304050600ULL;
1586		writeq(val64, &bar0->rx_w_round_robin_0);
1587		val64 = 0x0102030405060001ULL;
1588		writeq(val64, &bar0->rx_w_round_robin_1);
1589		val64 = 0x0203040506000102ULL;
1590		writeq(val64, &bar0->rx_w_round_robin_2);
1591		val64 = 0x0304050600010203ULL;
1592		writeq(val64, &bar0->rx_w_round_robin_3);
1593		val64 = 0x0405060000000000ULL;
1594		writeq(val64, &bar0->rx_w_round_robin_4);
1595
1596		val64 = 0x8080402010080402ULL;
1597		writeq(val64, &bar0->rts_qos_steering);
1598		break;
1599	case 8:
1600		val64 = 0x0001020304050607ULL;
1601		writeq(val64, &bar0->rx_w_round_robin_0);
1602		writeq(val64, &bar0->rx_w_round_robin_1);
1603		writeq(val64, &bar0->rx_w_round_robin_2);
1604		writeq(val64, &bar0->rx_w_round_robin_3);
1605		val64 = 0x0001020300000000ULL;
1606		writeq(val64, &bar0->rx_w_round_robin_4);
1607
1608		val64 = 0x8040201008040201ULL;
1609		writeq(val64, &bar0->rts_qos_steering);
1610		break;
1611	}
1612
1613	/* UDP Fix */
1614	val64 = 0;
1615	for (i = 0; i < 8; i++)
1616		writeq(val64, &bar0->rts_frm_len_n[i]);
1617
1618	/* Set the default rts frame length for the rings configured */
1619	val64 = MAC_RTS_FRM_LEN_SET(dev->mtu+22);
1620	for (i = 0 ; i < config->rx_ring_num ; i++)
1621		writeq(val64, &bar0->rts_frm_len_n[i]);
1622
1623	/* Set the frame length for the configured rings
1624	 * desired by the user
1625	 */
1626	for (i = 0; i < config->rx_ring_num; i++) {
1627		/* If rts_frm_len[i] == 0 then it is assumed that user not
1628		 * specified frame length steering.
1629		 * If the user provides the frame length then program
1630		 * the rts_frm_len register for those values or else
1631		 * leave it as it is.
1632		 */
1633		if (rts_frm_len[i] != 0) {
1634			writeq(MAC_RTS_FRM_LEN_SET(rts_frm_len[i]),
1635			       &bar0->rts_frm_len_n[i]);
1636		}
1637	}
1638
1639	/* Disable differentiated services steering logic */
1640	for (i = 0; i < 64; i++) {
1641		if (rts_ds_steer(nic, i, 0) == FAILURE) {
1642			DBG_PRINT(ERR_DBG,
1643				  "%s: rts_ds_steer failed on codepoint %d\n",
1644				  dev->name, i);
1645			return -ENODEV;
1646		}
1647	}
1648
1649	/* Program statistics memory */
1650	writeq(mac_control->stats_mem_phy, &bar0->stat_addr);
1651
1652	if (nic->device_type == XFRAME_II_DEVICE) {
1653		val64 = STAT_BC(0x320);
1654		writeq(val64, &bar0->stat_byte_cnt);
1655	}
1656
1657	/*
1658	 * Initializing the sampling rate for the device to calculate the
1659	 * bandwidth utilization.
1660	 */
1661	val64 = MAC_TX_LINK_UTIL_VAL(tmac_util_period) |
1662		MAC_RX_LINK_UTIL_VAL(rmac_util_period);
1663	writeq(val64, &bar0->mac_link_util);
1664
1665	/*
1666	 * Initializing the Transmit and Receive Traffic Interrupt
1667	 * Scheme.
1668	 */
1669
1670	/* Initialize TTI */
1671	if (SUCCESS != init_tti(nic, nic->last_link_state))
1672		return -ENODEV;
1673
1674	/* RTI Initialization */
1675	if (nic->device_type == XFRAME_II_DEVICE) {
1676		/*
1677		 * Programmed to generate Apprx 500 Intrs per
1678		 * second
1679		 */
1680		int count = (nic->config.bus_speed * 125)/4;
1681		val64 = RTI_DATA1_MEM_RX_TIMER_VAL(count);
1682	} else
1683		val64 = RTI_DATA1_MEM_RX_TIMER_VAL(0xFFF);
1684	val64 |= RTI_DATA1_MEM_RX_URNG_A(0xA) |
1685		RTI_DATA1_MEM_RX_URNG_B(0x10) |
1686		RTI_DATA1_MEM_RX_URNG_C(0x30) |
1687		RTI_DATA1_MEM_RX_TIMER_AC_EN;
1688
1689	writeq(val64, &bar0->rti_data1_mem);
1690
1691	val64 = RTI_DATA2_MEM_RX_UFC_A(0x1) |
1692		RTI_DATA2_MEM_RX_UFC_B(0x2) ;
1693	if (nic->config.intr_type == MSI_X)
1694		val64 |= (RTI_DATA2_MEM_RX_UFC_C(0x20) |
1695			  RTI_DATA2_MEM_RX_UFC_D(0x40));
1696	else
1697		val64 |= (RTI_DATA2_MEM_RX_UFC_C(0x40) |
1698			  RTI_DATA2_MEM_RX_UFC_D(0x80));
1699	writeq(val64, &bar0->rti_data2_mem);
1700
1701	for (i = 0; i < config->rx_ring_num; i++) {
1702		val64 = RTI_CMD_MEM_WE |
1703			RTI_CMD_MEM_STROBE_NEW_CMD |
1704			RTI_CMD_MEM_OFFSET(i);
1705		writeq(val64, &bar0->rti_command_mem);
1706
1707		/*
1708		 * Once the operation completes, the Strobe bit of the
1709		 * command register will be reset. We poll for this
1710		 * particular condition. We wait for a maximum of 500ms
1711		 * for the operation to complete, if it's not complete
1712		 * by then we return error.
1713		 */
1714		time = 0;
1715		while (true) {
1716			val64 = readq(&bar0->rti_command_mem);
1717			if (!(val64 & RTI_CMD_MEM_STROBE_NEW_CMD))
1718				break;
1719
1720			if (time > 10) {
1721				DBG_PRINT(ERR_DBG, "%s: RTI init failed\n",
1722					  dev->name);
1723				return -ENODEV;
1724			}
1725			time++;
1726			msleep(50);
1727		}
1728	}
1729
1730	/*
1731	 * Initializing proper values as Pause threshold into all
1732	 * the 8 Queues on Rx side.
1733	 */
1734	writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q0q3);
1735	writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q4q7);
1736
1737	/* Disable RMAC PAD STRIPPING */
1738	add = &bar0->mac_cfg;
1739	val64 = readq(&bar0->mac_cfg);
1740	val64 &= ~(MAC_CFG_RMAC_STRIP_PAD);
1741	writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1742	writel((u32) (val64), add);
1743	writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1744	writel((u32) (val64 >> 32), (add + 4));
1745	val64 = readq(&bar0->mac_cfg);
1746
1747	/* Enable FCS stripping by adapter */
1748	add = &bar0->mac_cfg;
1749	val64 = readq(&bar0->mac_cfg);
1750	val64 |= MAC_CFG_RMAC_STRIP_FCS;
1751	if (nic->device_type == XFRAME_II_DEVICE)
1752		writeq(val64, &bar0->mac_cfg);
1753	else {
1754		writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1755		writel((u32) (val64), add);
1756		writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1757		writel((u32) (val64 >> 32), (add + 4));
1758	}
1759
1760	/*
1761	 * Set the time value to be inserted in the pause frame
1762	 * generated by xena.
1763	 */
1764	val64 = readq(&bar0->rmac_pause_cfg);
1765	val64 &= ~(RMAC_PAUSE_HG_PTIME(0xffff));
1766	val64 |= RMAC_PAUSE_HG_PTIME(nic->mac_control.rmac_pause_time);
1767	writeq(val64, &bar0->rmac_pause_cfg);
1768
1769	/*
1770	 * Set the Threshold Limit for Generating the pause frame
1771	 * If the amount of data in any Queue exceeds ratio of
1772	 * (mac_control.mc_pause_threshold_q0q3 or q4q7)/256
1773	 * pause frame is generated
1774	 */
1775	val64 = 0;
1776	for (i = 0; i < 4; i++) {
1777		val64 |= (((u64)0xFF00 |
1778			   nic->mac_control.mc_pause_threshold_q0q3)
1779			  << (i * 2 * 8));
1780	}
1781	writeq(val64, &bar0->mc_pause_thresh_q0q3);
1782
1783	val64 = 0;
1784	for (i = 0; i < 4; i++) {
1785		val64 |= (((u64)0xFF00 |
1786			   nic->mac_control.mc_pause_threshold_q4q7)
1787			  << (i * 2 * 8));
1788	}
1789	writeq(val64, &bar0->mc_pause_thresh_q4q7);
1790
1791	/*
1792	 * TxDMA will stop Read request if the number of read split has
1793	 * exceeded the limit pointed by shared_splits
1794	 */
1795	val64 = readq(&bar0->pic_control);
1796	val64 |= PIC_CNTL_SHARED_SPLITS(shared_splits);
1797	writeq(val64, &bar0->pic_control);
1798
1799	if (nic->config.bus_speed == 266) {
1800		writeq(TXREQTO_VAL(0x7f) | TXREQTO_EN, &bar0->txreqtimeout);
1801		writeq(0x0, &bar0->read_retry_delay);
1802		writeq(0x0, &bar0->write_retry_delay);
1803	}
1804
1805	/*
1806	 * Programming the Herc to split every write transaction
1807	 * that does not start on an ADB to reduce disconnects.
1808	 */
1809	if (nic->device_type == XFRAME_II_DEVICE) {
1810		val64 = FAULT_BEHAVIOUR | EXT_REQ_EN |
1811			MISC_LINK_STABILITY_PRD(3);
1812		writeq(val64, &bar0->misc_control);
1813		val64 = readq(&bar0->pic_control2);
1814		val64 &= ~(s2BIT(13)|s2BIT(14)|s2BIT(15));
1815		writeq(val64, &bar0->pic_control2);
1816	}
1817	if (strstr(nic->product_name, "CX4")) {
1818		val64 = TMAC_AVG_IPG(0x17);
1819		writeq(val64, &bar0->tmac_avg_ipg);
1820	}
1821
1822	return SUCCESS;
1823}
1824#define LINK_UP_DOWN_INTERRUPT		1
1825#define MAC_RMAC_ERR_TIMER		2
1826
1827static int s2io_link_fault_indication(struct s2io_nic *nic)
1828{
1829	if (nic->device_type == XFRAME_II_DEVICE)
1830		return LINK_UP_DOWN_INTERRUPT;
1831	else
1832		return MAC_RMAC_ERR_TIMER;
1833}
1834
1835/**
1836 *  do_s2io_write_bits -  update alarm bits in alarm register
1837 *  @value: alarm bits
1838 *  @flag: interrupt status
1839 *  @addr: address value
1840 *  Description: update alarm bits in alarm register
1841 *  Return Value:
1842 *  NONE.
1843 */
1844static void do_s2io_write_bits(u64 value, int flag, void __iomem *addr)
1845{
1846	u64 temp64;
1847
1848	temp64 = readq(addr);
1849
1850	if (flag == ENABLE_INTRS)
1851		temp64 &= ~((u64)value);
1852	else
1853		temp64 |= ((u64)value);
1854	writeq(temp64, addr);
1855}
1856
1857static void en_dis_err_alarms(struct s2io_nic *nic, u16 mask, int flag)
1858{
1859	struct XENA_dev_config __iomem *bar0 = nic->bar0;
1860	register u64 gen_int_mask = 0;
1861	u64 interruptible;
1862
1863	writeq(DISABLE_ALL_INTRS, &bar0->general_int_mask);
1864	if (mask & TX_DMA_INTR) {
1865		gen_int_mask |= TXDMA_INT_M;
1866
1867		do_s2io_write_bits(TXDMA_TDA_INT | TXDMA_PFC_INT |
1868				   TXDMA_PCC_INT | TXDMA_TTI_INT |
1869				   TXDMA_LSO_INT | TXDMA_TPA_INT |
1870				   TXDMA_SM_INT, flag, &bar0->txdma_int_mask);
1871
1872		do_s2io_write_bits(PFC_ECC_DB_ERR | PFC_SM_ERR_ALARM |
1873				   PFC_MISC_0_ERR | PFC_MISC_1_ERR |
1874				   PFC_PCIX_ERR | PFC_ECC_SG_ERR, flag,
1875				   &bar0->pfc_err_mask);
1876
1877		do_s2io_write_bits(TDA_Fn_ECC_DB_ERR | TDA_SM0_ERR_ALARM |
1878				   TDA_SM1_ERR_ALARM | TDA_Fn_ECC_SG_ERR |
1879				   TDA_PCIX_ERR, flag, &bar0->tda_err_mask);
1880
1881		do_s2io_write_bits(PCC_FB_ECC_DB_ERR | PCC_TXB_ECC_DB_ERR |
1882				   PCC_SM_ERR_ALARM | PCC_WR_ERR_ALARM |
1883				   PCC_N_SERR | PCC_6_COF_OV_ERR |
1884				   PCC_7_COF_OV_ERR | PCC_6_LSO_OV_ERR |
1885				   PCC_7_LSO_OV_ERR | PCC_FB_ECC_SG_ERR |
1886				   PCC_TXB_ECC_SG_ERR,
1887				   flag, &bar0->pcc_err_mask);
1888
1889		do_s2io_write_bits(TTI_SM_ERR_ALARM | TTI_ECC_SG_ERR |
1890				   TTI_ECC_DB_ERR, flag, &bar0->tti_err_mask);
1891
1892		do_s2io_write_bits(LSO6_ABORT | LSO7_ABORT |
1893				   LSO6_SM_ERR_ALARM | LSO7_SM_ERR_ALARM |
1894				   LSO6_SEND_OFLOW | LSO7_SEND_OFLOW,
1895				   flag, &bar0->lso_err_mask);
1896
1897		do_s2io_write_bits(TPA_SM_ERR_ALARM | TPA_TX_FRM_DROP,
1898				   flag, &bar0->tpa_err_mask);
1899
1900		do_s2io_write_bits(SM_SM_ERR_ALARM, flag, &bar0->sm_err_mask);
1901	}
1902
1903	if (mask & TX_MAC_INTR) {
1904		gen_int_mask |= TXMAC_INT_M;
1905		do_s2io_write_bits(MAC_INT_STATUS_TMAC_INT, flag,
1906				   &bar0->mac_int_mask);
1907		do_s2io_write_bits(TMAC_TX_BUF_OVRN | TMAC_TX_SM_ERR |
1908				   TMAC_ECC_SG_ERR | TMAC_ECC_DB_ERR |
1909				   TMAC_DESC_ECC_SG_ERR | TMAC_DESC_ECC_DB_ERR,
1910				   flag, &bar0->mac_tmac_err_mask);
1911	}
1912
1913	if (mask & TX_XGXS_INTR) {
1914		gen_int_mask |= TXXGXS_INT_M;
1915		do_s2io_write_bits(XGXS_INT_STATUS_TXGXS, flag,
1916				   &bar0->xgxs_int_mask);
1917		do_s2io_write_bits(TXGXS_ESTORE_UFLOW | TXGXS_TX_SM_ERR |
1918				   TXGXS_ECC_SG_ERR | TXGXS_ECC_DB_ERR,
1919				   flag, &bar0->xgxs_txgxs_err_mask);
1920	}
1921
1922	if (mask & RX_DMA_INTR) {
1923		gen_int_mask |= RXDMA_INT_M;
1924		do_s2io_write_bits(RXDMA_INT_RC_INT_M | RXDMA_INT_RPA_INT_M |
1925				   RXDMA_INT_RDA_INT_M | RXDMA_INT_RTI_INT_M,
1926				   flag, &bar0->rxdma_int_mask);
1927		do_s2io_write_bits(RC_PRCn_ECC_DB_ERR | RC_FTC_ECC_DB_ERR |
1928				   RC_PRCn_SM_ERR_ALARM | RC_FTC_SM_ERR_ALARM |
1929				   RC_PRCn_ECC_SG_ERR | RC_FTC_ECC_SG_ERR |
1930				   RC_RDA_FAIL_WR_Rn, flag, &bar0->rc_err_mask);
1931		do_s2io_write_bits(PRC_PCI_AB_RD_Rn | PRC_PCI_AB_WR_Rn |
1932				   PRC_PCI_AB_F_WR_Rn | PRC_PCI_DP_RD_Rn |
1933				   PRC_PCI_DP_WR_Rn | PRC_PCI_DP_F_WR_Rn, flag,
1934				   &bar0->prc_pcix_err_mask);
1935		do_s2io_write_bits(RPA_SM_ERR_ALARM | RPA_CREDIT_ERR |
1936				   RPA_ECC_SG_ERR | RPA_ECC_DB_ERR, flag,
1937				   &bar0->rpa_err_mask);
1938		do_s2io_write_bits(RDA_RXDn_ECC_DB_ERR | RDA_FRM_ECC_DB_N_AERR |
1939				   RDA_SM1_ERR_ALARM | RDA_SM0_ERR_ALARM |
1940				   RDA_RXD_ECC_DB_SERR | RDA_RXDn_ECC_SG_ERR |
1941				   RDA_FRM_ECC_SG_ERR |
1942				   RDA_MISC_ERR|RDA_PCIX_ERR,
1943				   flag, &bar0->rda_err_mask);
1944		do_s2io_write_bits(RTI_SM_ERR_ALARM |
1945				   RTI_ECC_SG_ERR | RTI_ECC_DB_ERR,
1946				   flag, &bar0->rti_err_mask);
1947	}
1948
1949	if (mask & RX_MAC_INTR) {
1950		gen_int_mask |= RXMAC_INT_M;
1951		do_s2io_write_bits(MAC_INT_STATUS_RMAC_INT, flag,
1952				   &bar0->mac_int_mask);
1953		interruptible = (RMAC_RX_BUFF_OVRN | RMAC_RX_SM_ERR |
1954				 RMAC_UNUSED_INT | RMAC_SINGLE_ECC_ERR |
1955				 RMAC_DOUBLE_ECC_ERR);
1956		if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER)
1957			interruptible |= RMAC_LINK_STATE_CHANGE_INT;
1958		do_s2io_write_bits(interruptible,
1959				   flag, &bar0->mac_rmac_err_mask);
1960	}
1961
1962	if (mask & RX_XGXS_INTR) {
1963		gen_int_mask |= RXXGXS_INT_M;
1964		do_s2io_write_bits(XGXS_INT_STATUS_RXGXS, flag,
1965				   &bar0->xgxs_int_mask);
1966		do_s2io_write_bits(RXGXS_ESTORE_OFLOW | RXGXS_RX_SM_ERR, flag,
1967				   &bar0->xgxs_rxgxs_err_mask);
1968	}
1969
1970	if (mask & MC_INTR) {
1971		gen_int_mask |= MC_INT_M;
1972		do_s2io_write_bits(MC_INT_MASK_MC_INT,
1973				   flag, &bar0->mc_int_mask);
1974		do_s2io_write_bits(MC_ERR_REG_SM_ERR | MC_ERR_REG_ECC_ALL_SNG |
1975				   MC_ERR_REG_ECC_ALL_DBL | PLL_LOCK_N, flag,
1976				   &bar0->mc_err_mask);
1977	}
1978	nic->general_int_mask = gen_int_mask;
1979
1980	/* Remove this line when alarm interrupts are enabled */
1981	nic->general_int_mask = 0;
1982}
1983
1984/**
1985 *  en_dis_able_nic_intrs - Enable or Disable the interrupts
1986 *  @nic: device private variable,
1987 *  @mask: A mask indicating which Intr block must be modified and,
1988 *  @flag: A flag indicating whether to enable or disable the Intrs.
1989 *  Description: This function will either disable or enable the interrupts
1990 *  depending on the flag argument. The mask argument can be used to
1991 *  enable/disable any Intr block.
1992 *  Return Value: NONE.
1993 */
1994
1995static void en_dis_able_nic_intrs(struct s2io_nic *nic, u16 mask, int flag)
1996{
1997	struct XENA_dev_config __iomem *bar0 = nic->bar0;
1998	register u64 temp64 = 0, intr_mask = 0;
1999
2000	intr_mask = nic->general_int_mask;
2001
2002	/*  Top level interrupt classification */
2003	/*  PIC Interrupts */
2004	if (mask & TX_PIC_INTR) {
2005		/*  Enable PIC Intrs in the general intr mask register */
2006		intr_mask |= TXPIC_INT_M;
2007		if (flag == ENABLE_INTRS) {
2008			/*
2009			 * If Hercules adapter enable GPIO otherwise
2010			 * disable all PCIX, Flash, MDIO, IIC and GPIO
2011			 * interrupts for now.
2012			 * TODO
2013			 */
2014			if (s2io_link_fault_indication(nic) ==
2015			    LINK_UP_DOWN_INTERRUPT) {
2016				do_s2io_write_bits(PIC_INT_GPIO, flag,
2017						   &bar0->pic_int_mask);
2018				do_s2io_write_bits(GPIO_INT_MASK_LINK_UP, flag,
2019						   &bar0->gpio_int_mask);
2020			} else
2021				writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
2022		} else if (flag == DISABLE_INTRS) {
2023			/*
2024			 * Disable PIC Intrs in the general
2025			 * intr mask register
2026			 */
2027			writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
2028		}
2029	}
2030
2031	/*  Tx traffic interrupts */
2032	if (mask & TX_TRAFFIC_INTR) {
2033		intr_mask |= TXTRAFFIC_INT_M;
2034		if (flag == ENABLE_INTRS) {
2035			/*
2036			 * Enable all the Tx side interrupts
2037			 * writing 0 Enables all 64 TX interrupt levels
2038			 */
2039			writeq(0x0, &bar0->tx_traffic_mask);
2040		} else if (flag == DISABLE_INTRS) {
2041			/*
2042			 * Disable Tx Traffic Intrs in the general intr mask
2043			 * register.
2044			 */
2045			writeq(DISABLE_ALL_INTRS, &bar0->tx_traffic_mask);
2046		}
2047	}
2048
2049	/*  Rx traffic interrupts */
2050	if (mask & RX_TRAFFIC_INTR) {
2051		intr_mask |= RXTRAFFIC_INT_M;
2052		if (flag == ENABLE_INTRS) {
2053			/* writing 0 Enables all 8 RX interrupt levels */
2054			writeq(0x0, &bar0->rx_traffic_mask);
2055		} else if (flag == DISABLE_INTRS) {
2056			/*
2057			 * Disable Rx Traffic Intrs in the general intr mask
2058			 * register.
2059			 */
2060			writeq(DISABLE_ALL_INTRS, &bar0->rx_traffic_mask);
2061		}
2062	}
2063
2064	temp64 = readq(&bar0->general_int_mask);
2065	if (flag == ENABLE_INTRS)
2066		temp64 &= ~((u64)intr_mask);
2067	else
2068		temp64 = DISABLE_ALL_INTRS;
2069	writeq(temp64, &bar0->general_int_mask);
2070
2071	nic->general_int_mask = readq(&bar0->general_int_mask);
2072}
2073
2074/**
2075 *  verify_pcc_quiescent- Checks for PCC quiescent state
2076 *  Return: 1 If PCC is quiescence
2077 *          0 If PCC is not quiescence
2078 */
2079static int verify_pcc_quiescent(struct s2io_nic *sp, int flag)
2080{
2081	int ret = 0, herc;
2082	struct XENA_dev_config __iomem *bar0 = sp->bar0;
2083	u64 val64 = readq(&bar0->adapter_status);
2084
2085	herc = (sp->device_type == XFRAME_II_DEVICE);
2086
2087	if (flag == false) {
2088		if ((!herc && (sp->pdev->revision >= 4)) || herc) {
2089			if (!(val64 & ADAPTER_STATUS_RMAC_PCC_IDLE))
2090				ret = 1;
2091		} else {
2092			if (!(val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE))
2093				ret = 1;
2094		}
2095	} else {
2096		if ((!herc && (sp->pdev->revision >= 4)) || herc) {
2097			if (((val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) ==
2098			     ADAPTER_STATUS_RMAC_PCC_IDLE))
2099				ret = 1;
2100		} else {
2101			if (((val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) ==
2102			     ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE))
2103				ret = 1;
2104		}
2105	}
2106
2107	return ret;
2108}
2109/**
2110 *  verify_xena_quiescence - Checks whether the H/W is ready
2111 *  Description: Returns whether the H/W is ready to go or not. Depending
2112 *  on whether adapter enable bit was written or not the comparison
2113 *  differs and the calling function passes the input argument flag to
2114 *  indicate this.
2115 *  Return: 1 If xena is quiescence
2116 *          0 If Xena is not quiescence
2117 */
2118
2119static int verify_xena_quiescence(struct s2io_nic *sp)
2120{
2121	int  mode;
2122	struct XENA_dev_config __iomem *bar0 = sp->bar0;
2123	u64 val64 = readq(&bar0->adapter_status);
2124	mode = s2io_verify_pci_mode(sp);
2125
2126	if (!(val64 & ADAPTER_STATUS_TDMA_READY)) {
2127		DBG_PRINT(ERR_DBG, "TDMA is not ready!\n");
2128		return 0;
2129	}
2130	if (!(val64 & ADAPTER_STATUS_RDMA_READY)) {
2131		DBG_PRINT(ERR_DBG, "RDMA is not ready!\n");
2132		return 0;
2133	}
2134	if (!(val64 & ADAPTER_STATUS_PFC_READY)) {
2135		DBG_PRINT(ERR_DBG, "PFC is not ready!\n");
2136		return 0;
2137	}
2138	if (!(val64 & ADAPTER_STATUS_TMAC_BUF_EMPTY)) {
2139		DBG_PRINT(ERR_DBG, "TMAC BUF is not empty!\n");
2140		return 0;
2141	}
2142	if (!(val64 & ADAPTER_STATUS_PIC_QUIESCENT)) {
2143		DBG_PRINT(ERR_DBG, "PIC is not QUIESCENT!\n");
2144		return 0;
2145	}
2146	if (!(val64 & ADAPTER_STATUS_MC_DRAM_READY)) {
2147		DBG_PRINT(ERR_DBG, "MC_DRAM is not ready!\n");
2148		return 0;
2149	}
2150	if (!(val64 & ADAPTER_STATUS_MC_QUEUES_READY)) {
2151		DBG_PRINT(ERR_DBG, "MC_QUEUES is not ready!\n");
2152		return 0;
2153	}
2154	if (!(val64 & ADAPTER_STATUS_M_PLL_LOCK)) {
2155		DBG_PRINT(ERR_DBG, "M_PLL is not locked!\n");
2156		return 0;
2157	}
2158
2159	/*
2160	 * In PCI 33 mode, the P_PLL is not used, and therefore,
2161	 * the the P_PLL_LOCK bit in the adapter_status register will
2162	 * not be asserted.
2163	 */
2164	if (!(val64 & ADAPTER_STATUS_P_PLL_LOCK) &&
2165	    sp->device_type == XFRAME_II_DEVICE &&
2166	    mode != PCI_MODE_PCI_33) {
2167		DBG_PRINT(ERR_DBG, "P_PLL is not locked!\n");
2168		return 0;
2169	}
2170	if (!((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
2171	      ADAPTER_STATUS_RC_PRC_QUIESCENT)) {
2172		DBG_PRINT(ERR_DBG, "RC_PRC is not QUIESCENT!\n");
2173		return 0;
2174	}
2175	return 1;
2176}
2177
2178/**
2179 * fix_mac_address -  Fix for Mac addr problem on Alpha platforms
2180 * @sp: Pointer to device specifc structure
2181 * Description :
2182 * New procedure to clear mac address reading  problems on Alpha platforms
2183 *
2184 */
2185
2186static void fix_mac_address(struct s2io_nic *sp)
2187{
2188	struct XENA_dev_config __iomem *bar0 = sp->bar0;
2189	int i = 0;
2190
2191	while (fix_mac[i] != END_SIGN) {
2192		writeq(fix_mac[i++], &bar0->gpio_control);
2193		udelay(10);
2194		(void) readq(&bar0->gpio_control);
2195	}
2196}
2197
2198/**
2199 *  start_nic - Turns the device on
2200 *  @nic : device private variable.
2201 *  Description:
2202 *  This function actually turns the device on. Before this  function is
2203 *  called,all Registers are configured from their reset states
2204 *  and shared memory is allocated but the NIC is still quiescent. On
2205 *  calling this function, the device interrupts are cleared and the NIC is
2206 *  literally switched on by writing into the adapter control register.
2207 *  Return Value:
2208 *  SUCCESS on success and -1 on failure.
2209 */
2210
2211static int start_nic(struct s2io_nic *nic)
2212{
2213	struct XENA_dev_config __iomem *bar0 = nic->bar0;
2214	struct net_device *dev = nic->dev;
2215	register u64 val64 = 0;
2216	u16 subid, i;
2217	struct config_param *config = &nic->config;
2218	struct mac_info *mac_control = &nic->mac_control;
2219
2220	/*  PRC Initialization and configuration */
2221	for (i = 0; i < config->rx_ring_num; i++) {
2222		struct ring_info *ring = &mac_control->rings[i];
2223
2224		writeq((u64)ring->rx_blocks[0].block_dma_addr,
2225		       &bar0->prc_rxd0_n[i]);
2226
2227		val64 = readq(&bar0->prc_ctrl_n[i]);
2228		if (nic->rxd_mode == RXD_MODE_1)
2229			val64 |= PRC_CTRL_RC_ENABLED;
2230		else
2231			val64 |= PRC_CTRL_RC_ENABLED | PRC_CTRL_RING_MODE_3;
2232		if (nic->device_type == XFRAME_II_DEVICE)
2233			val64 |= PRC_CTRL_GROUP_READS;
2234		val64 &= ~PRC_CTRL_RXD_BACKOFF_INTERVAL(0xFFFFFF);
2235		val64 |= PRC_CTRL_RXD_BACKOFF_INTERVAL(0x1000);
2236		writeq(val64, &bar0->prc_ctrl_n[i]);
2237	}
2238
2239	if (nic->rxd_mode == RXD_MODE_3B) {
2240		/* Enabling 2 buffer mode by writing into Rx_pa_cfg reg. */
2241		val64 = readq(&bar0->rx_pa_cfg);
2242		val64 |= RX_PA_CFG_IGNORE_L2_ERR;
2243		writeq(val64, &bar0->rx_pa_cfg);
2244	}
2245
2246	if (vlan_tag_strip == 0) {
2247		val64 = readq(&bar0->rx_pa_cfg);
2248		val64 &= ~RX_PA_CFG_STRIP_VLAN_TAG;
2249		writeq(val64, &bar0->rx_pa_cfg);
2250		nic->vlan_strip_flag = 0;
2251	}
2252
2253	/*
2254	 * Enabling MC-RLDRAM. After enabling the device, we timeout
2255	 * for around 100ms, which is approximately the time required
2256	 * for the device to be ready for operation.
2257	 */
2258	val64 = readq(&bar0->mc_rldram_mrs);
2259	val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE | MC_RLDRAM_MRS_ENABLE;
2260	SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
2261	val64 = readq(&bar0->mc_rldram_mrs);
2262
2263	msleep(100);	/* Delay by around 100 ms. */
2264
2265	/* Enabling ECC Protection. */
2266	val64 = readq(&bar0->adapter_control);
2267	val64 &= ~ADAPTER_ECC_EN;
2268	writeq(val64, &bar0->adapter_control);
2269
2270	/*
2271	 * Verify if the device is ready to be enabled, if so enable
2272	 * it.
2273	 */
2274	val64 = readq(&bar0->adapter_status);
2275	if (!verify_xena_quiescence(nic)) {
2276		DBG_PRINT(ERR_DBG, "%s: device is not ready, "
2277			  "Adapter status reads: 0x%llx\n",
2278			  dev->name, (unsigned long long)val64);
2279		return FAILURE;
2280	}
2281
2282	/*
2283	 * With some switches, link might be already up at this point.
2284	 * Because of this weird behavior, when we enable laser,
2285	 * we may not get link. We need to handle this. We cannot
2286	 * figure out which switch is misbehaving. So we are forced to
2287	 * make a global change.
2288	 */
2289
2290	/* Enabling Laser. */
2291	val64 = readq(&bar0->adapter_control);
2292	val64 |= ADAPTER_EOI_TX_ON;
2293	writeq(val64, &bar0->adapter_control);
2294
2295	if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
2296		/*
2297		 * Dont see link state interrupts initially on some switches,
2298		 * so directly scheduling the link state task here.
2299		 */
2300		schedule_work(&nic->set_link_task);
2301	}
2302	/* SXE-002: Initialize link and activity LED */
2303	subid = nic->pdev->subsystem_device;
2304	if (((subid & 0xFF) >= 0x07) &&
2305	    (nic->device_type == XFRAME_I_DEVICE)) {
2306		val64 = readq(&bar0->gpio_control);
2307		val64 |= 0x0000800000000000ULL;
2308		writeq(val64, &bar0->gpio_control);
2309		val64 = 0x0411040400000000ULL;
2310		writeq(val64, (void __iomem *)bar0 + 0x2700);
2311	}
2312
2313	return SUCCESS;
2314}
2315/**
2316 * s2io_txdl_getskb - Get the skb from txdl, unmap and return skb
2317 */
2318static struct sk_buff *s2io_txdl_getskb(struct fifo_info *fifo_data,
2319					struct TxD *txdlp, int get_off)
2320{
2321	struct s2io_nic *nic = fifo_data->nic;
2322	struct sk_buff *skb;
2323	struct TxD *txds;
2324	u16 j, frg_cnt;
2325
2326	txds = txdlp;
2327	if (txds->Host_Control == (u64)(long)fifo_data->ufo_in_band_v) {
2328		pci_unmap_single(nic->pdev, (dma_addr_t)txds->Buffer_Pointer,
2329				 sizeof(u64), PCI_DMA_TODEVICE);
2330		txds++;
2331	}
2332
2333	skb = (struct sk_buff *)((unsigned long)txds->Host_Control);
2334	if (!skb) {
2335		memset(txdlp, 0, (sizeof(struct TxD) * fifo_data->max_txds));
2336		return NULL;
2337	}
2338	pci_unmap_single(nic->pdev, (dma_addr_t)txds->Buffer_Pointer,
2339			 skb_headlen(skb), PCI_DMA_TODEVICE);
2340	frg_cnt = skb_shinfo(skb)->nr_frags;
2341	if (frg_cnt) {
2342		txds++;
2343		for (j = 0; j < frg_cnt; j++, txds++) {
2344			const skb_frag_t *frag = &skb_shinfo(skb)->frags[j];
2345			if (!txds->Buffer_Pointer)
2346				break;
2347			pci_unmap_page(nic->pdev,
2348				       (dma_addr_t)txds->Buffer_Pointer,
2349				       skb_frag_size(frag), PCI_DMA_TODEVICE);
2350		}
2351	}
2352	memset(txdlp, 0, (sizeof(struct TxD) * fifo_data->max_txds));
2353	return skb;
2354}
2355
2356/**
2357 *  free_tx_buffers - Free all queued Tx buffers
2358 *  @nic : device private variable.
2359 *  Description:
2360 *  Free all queued Tx buffers.
2361 *  Return Value: void
2362 */
2363
2364static void free_tx_buffers(struct s2io_nic *nic)
2365{
2366	struct net_device *dev = nic->dev;
2367	struct sk_buff *skb;
2368	struct TxD *txdp;
2369	int i, j;
2370	int cnt = 0;
2371	struct config_param *config = &nic->config;
2372	struct mac_info *mac_control = &nic->mac_control;
2373	struct stat_block *stats = mac_control->stats_info;
2374	struct swStat *swstats = &stats->sw_stat;
2375
2376	for (i = 0; i < config->tx_fifo_num; i++) {
2377		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
2378		struct fifo_info *fifo = &mac_control->fifos[i];
2379		unsigned long flags;
2380
2381		spin_lock_irqsave(&fifo->tx_lock, flags);
2382		for (j = 0; j < tx_cfg->fifo_len; j++) {
2383			txdp = fifo->list_info[j].list_virt_addr;
2384			skb = s2io_txdl_getskb(&mac_control->fifos[i], txdp, j);
2385			if (skb) {
2386				swstats->mem_freed += skb->truesize;
2387				dev_kfree_skb(skb);
2388				cnt++;
2389			}
2390		}
2391		DBG_PRINT(INTR_DBG,
2392			  "%s: forcibly freeing %d skbs on FIFO%d\n",
2393			  dev->name, cnt, i);
2394		fifo->tx_curr_get_info.offset = 0;
2395		fifo->tx_curr_put_info.offset = 0;
2396		spin_unlock_irqrestore(&fifo->tx_lock, flags);
2397	}
2398}
2399
2400/**
2401 *   stop_nic -  To stop the nic
2402 *   @nic ; device private variable.
2403 *   Description:
2404 *   This function does exactly the opposite of what the start_nic()
2405 *   function does. This function is called to stop the device.
2406 *   Return Value:
2407 *   void.
2408 */
2409
2410static void stop_nic(struct s2io_nic *nic)
2411{
2412	struct XENA_dev_config __iomem *bar0 = nic->bar0;
2413	register u64 val64 = 0;
2414	u16 interruptible;
2415
2416	/*  Disable all interrupts */
2417	en_dis_err_alarms(nic, ENA_ALL_INTRS, DISABLE_INTRS);
2418	interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR;
2419	interruptible |= TX_PIC_INTR;
2420	en_dis_able_nic_intrs(nic, interruptible, DISABLE_INTRS);
2421
2422	/* Clearing Adapter_En bit of ADAPTER_CONTROL Register */
2423	val64 = readq(&bar0->adapter_control);
2424	val64 &= ~(ADAPTER_CNTL_EN);
2425	writeq(val64, &bar0->adapter_control);
2426}
2427
2428/**
2429 *  fill_rx_buffers - Allocates the Rx side skbs
2430 *  @ring_info: per ring structure
2431 *  @from_card_up: If this is true, we will map the buffer to get
2432 *     the dma address for buf0 and buf1 to give it to the card.
2433 *     Else we will sync the already mapped buffer to give it to the card.
2434 *  Description:
2435 *  The function allocates Rx side skbs and puts the physical
2436 *  address of these buffers into the RxD buffer pointers, so that the NIC
2437 *  can DMA the received frame into these locations.
2438 *  The NIC supports 3 receive modes, viz
2439 *  1. single buffer,
2440 *  2. three buffer and
2441 *  3. Five buffer modes.
2442 *  Each mode defines how many fragments the received frame will be split
2443 *  up into by the NIC. The frame is split into L3 header, L4 Header,
2444 *  L4 payload in three buffer mode and in 5 buffer mode, L4 payload itself
2445 *  is split into 3 fragments. As of now only single buffer mode is
2446 *  supported.
2447 *   Return Value:
2448 *  SUCCESS on success or an appropriate -ve value on failure.
2449 */
2450static int fill_rx_buffers(struct s2io_nic *nic, struct ring_info *ring,
2451			   int from_card_up)
2452{
2453	struct sk_buff *skb;
2454	struct RxD_t *rxdp;
2455	int off, size, block_no, block_no1;
2456	u32 alloc_tab = 0;
2457	u32 alloc_cnt;
2458	u64 tmp;
2459	struct buffAdd *ba;
2460	struct RxD_t *first_rxdp = NULL;
2461	u64 Buffer0_ptr = 0, Buffer1_ptr = 0;
2462	int rxd_index = 0;
2463	struct RxD1 *rxdp1;
2464	struct RxD3 *rxdp3;
2465	struct swStat *swstats = &ring->nic->mac_control.stats_info->sw_stat;
2466
2467	alloc_cnt = ring->pkt_cnt - ring->rx_bufs_left;
2468
2469	block_no1 = ring->rx_curr_get_info.block_index;
2470	while (alloc_tab < alloc_cnt) {
2471		block_no = ring->rx_curr_put_info.block_index;
2472
2473		off = ring->rx_curr_put_info.offset;
2474
2475		rxdp = ring->rx_blocks[block_no].rxds[off].virt_addr;
2476
2477		rxd_index = off + 1;
2478		if (block_no)
2479			rxd_index += (block_no * ring->rxd_count);
2480
2481		if ((block_no == block_no1) &&
2482		    (off == ring->rx_curr_get_info.offset) &&
2483		    (rxdp->Host_Control)) {
2484			DBG_PRINT(INTR_DBG, "%s: Get and Put info equated\n",
2485				  ring->dev->name);
2486			goto end;
2487		}
2488		if (off && (off == ring->rxd_count)) {
2489			ring->rx_curr_put_info.block_index++;
2490			if (ring->rx_curr_put_info.block_index ==
2491			    ring->block_count)
2492				ring->rx_curr_put_info.block_index = 0;
2493			block_no = ring->rx_curr_put_info.block_index;
2494			off = 0;
2495			ring->rx_curr_put_info.offset = off;
2496			rxdp = ring->rx_blocks[block_no].block_virt_addr;
2497			DBG_PRINT(INTR_DBG, "%s: Next block at: %p\n",
2498				  ring->dev->name, rxdp);
2499
2500		}
2501
2502		if ((rxdp->Control_1 & RXD_OWN_XENA) &&
2503		    ((ring->rxd_mode == RXD_MODE_3B) &&
2504		     (rxdp->Control_2 & s2BIT(0)))) {
2505			ring->rx_curr_put_info.offset = off;
2506			goto end;
2507		}
2508		/* calculate size of skb based on ring mode */
2509		size = ring->mtu +
2510			HEADER_ETHERNET_II_802_3_SIZE +
2511			HEADER_802_2_SIZE + HEADER_SNAP_SIZE;
2512		if (ring->rxd_mode == RXD_MODE_1)
2513			size += NET_IP_ALIGN;
2514		else
2515			size = ring->mtu + ALIGN_SIZE + BUF0_LEN + 4;
2516
2517		/* allocate skb */
2518		skb = netdev_alloc_skb(nic->dev, size);
2519		if (!skb) {
2520			DBG_PRINT(INFO_DBG, "%s: Could not allocate skb\n",
2521				  ring->dev->name);
2522			if (first_rxdp) {
2523				wmb();
2524				first_rxdp->Control_1 |= RXD_OWN_XENA;
2525			}
2526			swstats->mem_alloc_fail_cnt++;
2527
2528			return -ENOMEM ;
2529		}
2530		swstats->mem_allocated += skb->truesize;
2531
2532		if (ring->rxd_mode == RXD_MODE_1) {
2533			/* 1 buffer mode - normal operation mode */
2534			rxdp1 = (struct RxD1 *)rxdp;
2535			memset(rxdp, 0, sizeof(struct RxD1));
2536			skb_reserve(skb, NET_IP_ALIGN);
2537			rxdp1->Buffer0_ptr =
2538				pci_map_single(ring->pdev, skb->data,
2539					       size - NET_IP_ALIGN,
2540					       PCI_DMA_FROMDEVICE);
2541			if (pci_dma_mapping_error(nic->pdev,
2542						  rxdp1->Buffer0_ptr))
2543				goto pci_map_failed;
2544
2545			rxdp->Control_2 =
2546				SET_BUFFER0_SIZE_1(size - NET_IP_ALIGN);
2547			rxdp->Host_Control = (unsigned long)skb;
2548		} else if (ring->rxd_mode == RXD_MODE_3B) {
2549			/*
2550			 * 2 buffer mode -
2551			 * 2 buffer mode provides 128
2552			 * byte aligned receive buffers.
2553			 */
2554
2555			rxdp3 = (struct RxD3 *)rxdp;
2556			/* save buffer pointers to avoid frequent dma mapping */
2557			Buffer0_ptr = rxdp3->Buffer0_ptr;
2558			Buffer1_ptr = rxdp3->Buffer1_ptr;
2559			memset(rxdp, 0, sizeof(struct RxD3));
2560			/* restore the buffer pointers for dma sync*/
2561			rxdp3->Buffer0_ptr = Buffer0_ptr;
2562			rxdp3->Buffer1_ptr = Buffer1_ptr;
2563
2564			ba = &ring->ba[block_no][off];
2565			skb_reserve(skb, BUF0_LEN);
2566			tmp = (u64)(unsigned long)skb->data;
2567			tmp += ALIGN_SIZE;
2568			tmp &= ~ALIGN_SIZE;
2569			skb->data = (void *) (unsigned long)tmp;
2570			skb_reset_tail_pointer(skb);
2571
2572			if (from_card_up) {
2573				rxdp3->Buffer0_ptr =
2574					pci_map_single(ring->pdev, ba->ba_0,
2575						       BUF0_LEN,
2576						       PCI_DMA_FROMDEVICE);
2577				if (pci_dma_mapping_error(nic->pdev,
2578							  rxdp3->Buffer0_ptr))
2579					goto pci_map_failed;
2580			} else
2581				pci_dma_sync_single_for_device(ring->pdev,
2582							       (dma_addr_t)rxdp3->Buffer0_ptr,
2583							       BUF0_LEN,
2584							       PCI_DMA_FROMDEVICE);
2585
2586			rxdp->Control_2 = SET_BUFFER0_SIZE_3(BUF0_LEN);
2587			if (ring->rxd_mode == RXD_MODE_3B) {
2588				/* Two buffer mode */
2589
2590				/*
2591				 * Buffer2 will have L3/L4 header plus
2592				 * L4 payload
2593				 */
2594				rxdp3->Buffer2_ptr = pci_map_single(ring->pdev,
2595								    skb->data,
2596								    ring->mtu + 4,
2597								    PCI_DMA_FROMDEVICE);
2598
2599				if (pci_dma_mapping_error(nic->pdev,
2600							  rxdp3->Buffer2_ptr))
2601					goto pci_map_failed;
2602
2603				if (from_card_up) {
2604					rxdp3->Buffer1_ptr =
2605						pci_map_single(ring->pdev,
2606							       ba->ba_1,
2607							       BUF1_LEN,
2608							       PCI_DMA_FROMDEVICE);
2609
2610					if (pci_dma_mapping_error(nic->pdev,
2611								  rxdp3->Buffer1_ptr)) {
2612						pci_unmap_single(ring->pdev,
2613								 (dma_addr_t)(unsigned long)
2614								 skb->data,
2615								 ring->mtu + 4,
2616								 PCI_DMA_FROMDEVICE);
2617						goto pci_map_failed;
2618					}
2619				}
2620				rxdp->Control_2 |= SET_BUFFER1_SIZE_3(1);
2621				rxdp->Control_2 |= SET_BUFFER2_SIZE_3
2622					(ring->mtu + 4);
2623			}
2624			rxdp->Control_2 |= s2BIT(0);
2625			rxdp->Host_Control = (unsigned long) (skb);
2626		}
2627		if (alloc_tab & ((1 << rxsync_frequency) - 1))
2628			rxdp->Control_1 |= RXD_OWN_XENA;
2629		off++;
2630		if (off == (ring->rxd_count + 1))
2631			off = 0;
2632		ring->rx_curr_put_info.offset = off;
2633
2634		rxdp->Control_2 |= SET_RXD_MARKER;
2635		if (!(alloc_tab & ((1 << rxsync_frequency) - 1))) {
2636			if (first_rxdp) {
2637				wmb();
2638				first_rxdp->Control_1 |= RXD_OWN_XENA;
2639			}
2640			first_rxdp = rxdp;
2641		}
2642		ring->rx_bufs_left += 1;
2643		alloc_tab++;
2644	}
2645
2646end:
2647	/* Transfer ownership of first descriptor to adapter just before
2648	 * exiting. Before that, use memory barrier so that ownership
2649	 * and other fields are seen by adapter correctly.
2650	 */
2651	if (first_rxdp) {
2652		wmb();
2653		first_rxdp->Control_1 |= RXD_OWN_XENA;
2654	}
2655
2656	return SUCCESS;
2657
2658pci_map_failed:
2659	swstats->pci_map_fail_cnt++;
2660	swstats->mem_freed += skb->truesize;
2661	dev_kfree_skb_irq(skb);
2662	return -ENOMEM;
2663}
2664
2665static void free_rxd_blk(struct s2io_nic *sp, int ring_no, int blk)
2666{
2667	struct net_device *dev = sp->dev;
2668	int j;
2669	struct sk_buff *skb;
2670	struct RxD_t *rxdp;
2671	struct RxD1 *rxdp1;
2672	struct RxD3 *rxdp3;
2673	struct mac_info *mac_control = &sp->mac_control;
2674	struct stat_block *stats = mac_control->stats_info;
2675	struct swStat *swstats = &stats->sw_stat;
2676
2677	for (j = 0 ; j < rxd_count[sp->rxd_mode]; j++) {
2678		rxdp = mac_control->rings[ring_no].
2679			rx_blocks[blk].rxds[j].virt_addr;
2680		skb = (struct sk_buff *)((unsigned long)rxdp->Host_Control);
2681		if (!skb)
2682			continue;
2683		if (sp->rxd_mode == RXD_MODE_1) {
2684			rxdp1 = (struct RxD1 *)rxdp;
2685			pci_unmap_single(sp->pdev,
2686					 (dma_addr_t)rxdp1->Buffer0_ptr,
2687					 dev->mtu +
2688					 HEADER_ETHERNET_II_802_3_SIZE +
2689					 HEADER_802_2_SIZE + HEADER_SNAP_SIZE,
2690					 PCI_DMA_FROMDEVICE);
2691			memset(rxdp, 0, sizeof(struct RxD1));
2692		} else if (sp->rxd_mode == RXD_MODE_3B) {
2693			rxdp3 = (struct RxD3 *)rxdp;
2694			pci_unmap_single(sp->pdev,
2695					 (dma_addr_t)rxdp3->Buffer0_ptr,
2696					 BUF0_LEN,
2697					 PCI_DMA_FROMDEVICE);
2698			pci_unmap_single(sp->pdev,
2699					 (dma_addr_t)rxdp3->Buffer1_ptr,
2700					 BUF1_LEN,
2701					 PCI_DMA_FROMDEVICE);
2702			pci_unmap_single(sp->pdev,
2703					 (dma_addr_t)rxdp3->Buffer2_ptr,
2704					 dev->mtu + 4,
2705					 PCI_DMA_FROMDEVICE);
2706			memset(rxdp, 0, sizeof(struct RxD3));
2707		}
2708		swstats->mem_freed += skb->truesize;
2709		dev_kfree_skb(skb);
2710		mac_control->rings[ring_no].rx_bufs_left -= 1;
2711	}
2712}
2713
2714/**
2715 *  free_rx_buffers - Frees all Rx buffers
2716 *  @sp: device private variable.
2717 *  Description:
2718 *  This function will free all Rx buffers allocated by host.
2719 *  Return Value:
2720 *  NONE.
2721 */
2722
2723static void free_rx_buffers(struct s2io_nic *sp)
2724{
2725	struct net_device *dev = sp->dev;
2726	int i, blk = 0, buf_cnt = 0;
2727	struct config_param *config = &sp->config;
2728	struct mac_info *mac_control = &sp->mac_control;
2729
2730	for (i = 0; i < config->rx_ring_num; i++) {
2731		struct ring_info *ring = &mac_control->rings[i];
2732
2733		for (blk = 0; blk < rx_ring_sz[i]; blk++)
2734			free_rxd_blk(sp, i, blk);
2735
2736		ring->rx_curr_put_info.block_index = 0;
2737		ring->rx_curr_get_info.block_index = 0;
2738		ring->rx_curr_put_info.offset = 0;
2739		ring->rx_curr_get_info.offset = 0;
2740		ring->rx_bufs_left = 0;
2741		DBG_PRINT(INIT_DBG, "%s: Freed 0x%x Rx Buffers on ring%d\n",
2742			  dev->name, buf_cnt, i);
2743	}
2744}
2745
2746static int s2io_chk_rx_buffers(struct s2io_nic *nic, struct ring_info *ring)
2747{
2748	if (fill_rx_buffers(nic, ring, 0) == -ENOMEM) {
2749		DBG_PRINT(INFO_DBG, "%s: Out of memory in Rx Intr!!\n",
2750			  ring->dev->name);
2751	}
2752	return 0;
2753}
2754
2755/**
2756 * s2io_poll - Rx interrupt handler for NAPI support
2757 * @napi : pointer to the napi structure.
2758 * @budget : The number of packets that were budgeted to be processed
2759 * during  one pass through the 'Poll" function.
2760 * Description:
2761 * Comes into picture only if NAPI support has been incorporated. It does
2762 * the same thing that rx_intr_handler does, but not in a interrupt context
2763 * also It will process only a given number of packets.
2764 * Return value:
2765 * 0 on success and 1 if there are No Rx packets to be processed.
2766 */
2767
2768static int s2io_poll_msix(struct napi_struct *napi, int budget)
2769{
2770	struct ring_info *ring = container_of(napi, struct ring_info, napi);
2771	struct net_device *dev = ring->dev;
2772	int pkts_processed = 0;
2773	u8 __iomem *addr = NULL;
2774	u8 val8 = 0;
2775	struct s2io_nic *nic = netdev_priv(dev);
2776	struct XENA_dev_config __iomem *bar0 = nic->bar0;
2777	int budget_org = budget;
2778
2779	if (unlikely(!is_s2io_card_up(nic)))
2780		return 0;
2781
2782	pkts_processed = rx_intr_handler(ring, budget);
2783	s2io_chk_rx_buffers(nic, ring);
2784
2785	if (pkts_processed < budget_org) {
2786		napi_complete(napi);
2787		/*Re Enable MSI-Rx Vector*/
2788		addr = (u8 __iomem *)&bar0->xmsi_mask_reg;
2789		addr += 7 - ring->ring_no;
2790		val8 = (ring->ring_no == 0) ? 0x3f : 0xbf;
2791		writeb(val8, addr);
2792		val8 = readb(addr);
2793	}
2794	return pkts_processed;
2795}
2796
2797static int s2io_poll_inta(struct napi_struct *napi, int budget)
2798{
2799	struct s2io_nic *nic = container_of(napi, struct s2io_nic, napi);
2800	int pkts_processed = 0;
2801	int ring_pkts_processed, i;
2802	struct XENA_dev_config __iomem *bar0 = nic->bar0;
2803	int budget_org = budget;
2804	struct config_param *config = &nic->config;
2805	struct mac_info *mac_control = &nic->mac_control;
2806
2807	if (unlikely(!is_s2io_card_up(nic)))
2808		return 0;
2809
2810	for (i = 0; i < config->rx_ring_num; i++) {
2811		struct ring_info *ring = &mac_control->rings[i];
2812		ring_pkts_processed = rx_intr_handler(ring, budget);
2813		s2io_chk_rx_buffers(nic, ring);
2814		pkts_processed += ring_pkts_processed;
2815		budget -= ring_pkts_processed;
2816		if (budget <= 0)
2817			break;
2818	}
2819	if (pkts_processed < budget_org) {
2820		napi_complete(napi);
2821		/* Re enable the Rx interrupts for the ring */
2822		writeq(0, &bar0->rx_traffic_mask);
2823		readl(&bar0->rx_traffic_mask);
2824	}
2825	return pkts_processed;
2826}
2827
2828#ifdef CONFIG_NET_POLL_CONTROLLER
2829/**
2830 * s2io_netpoll - netpoll event handler entry point
2831 * @dev : pointer to the device structure.
2832 * Description:
2833 * 	This function will be called by upper layer to check for events on the
2834 * interface in situations where interrupts are disabled. It is used for
2835 * specific in-kernel networking tasks, such as remote consoles and kernel
2836 * debugging over the network (example netdump in RedHat).
2837 */
2838static void s2io_netpoll(struct net_device *dev)
2839{
2840	struct s2io_nic *nic = netdev_priv(dev);
2841	const int irq = nic->pdev->irq;
2842	struct XENA_dev_config __iomem *bar0 = nic->bar0;
2843	u64 val64 = 0xFFFFFFFFFFFFFFFFULL;
2844	int i;
2845	struct config_param *config = &nic->config;
2846	struct mac_info *mac_control = &nic->mac_control;
2847
2848	if (pci_channel_offline(nic->pdev))
2849		return;
2850
2851	disable_irq(irq);
2852
2853	writeq(val64, &bar0->rx_traffic_int);
2854	writeq(val64, &bar0->tx_traffic_int);
2855
2856	/* we need to free up the transmitted skbufs or else netpoll will
2857	 * run out of skbs and will fail and eventually netpoll application such
2858	 * as netdump will fail.
2859	 */
2860	for (i = 0; i < config->tx_fifo_num; i++)
2861		tx_intr_handler(&mac_control->fifos[i]);
2862
2863	/* check for received packet and indicate up to network */
2864	for (i = 0; i < config->rx_ring_num; i++) {
2865		struct ring_info *ring = &mac_control->rings[i];
2866
2867		rx_intr_handler(ring, 0);
2868	}
2869
2870	for (i = 0; i < config->rx_ring_num; i++) {
2871		struct ring_info *ring = &mac_control->rings[i];
2872
2873		if (fill_rx_buffers(nic, ring, 0) == -ENOMEM) {
2874			DBG_PRINT(INFO_DBG,
2875				  "%s: Out of memory in Rx Netpoll!!\n",
2876				  dev->name);
2877			break;
2878		}
2879	}
2880	enable_irq(irq);
2881}
2882#endif
2883
2884/**
2885 *  rx_intr_handler - Rx interrupt handler
2886 *  @ring_info: per ring structure.
2887 *  @budget: budget for napi processing.
2888 *  Description:
2889 *  If the interrupt is because of a received frame or if the
2890 *  receive ring contains fresh as yet un-processed frames,this function is
2891 *  called. It picks out the RxD at which place the last Rx processing had
2892 *  stopped and sends the skb to the OSM's Rx handler and then increments
2893 *  the offset.
2894 *  Return Value:
2895 *  No. of napi packets processed.
2896 */
2897static int rx_intr_handler(struct ring_info *ring_data, int budget)
2898{
2899	int get_block, put_block;
2900	struct rx_curr_get_info get_info, put_info;
2901	struct RxD_t *rxdp;
2902	struct sk_buff *skb;
2903	int pkt_cnt = 0, napi_pkts = 0;
2904	int i;
2905	struct RxD1 *rxdp1;
2906	struct RxD3 *rxdp3;
2907
2908	if (budget <= 0)
2909		return napi_pkts;
2910
2911	get_info = ring_data->rx_curr_get_info;
2912	get_block = get_info.block_index;
2913	memcpy(&put_info, &ring_data->rx_curr_put_info, sizeof(put_info));
2914	put_block = put_info.block_index;
2915	rxdp = ring_data->rx_blocks[get_block].rxds[get_info.offset].virt_addr;
2916
2917	while (RXD_IS_UP2DT(rxdp)) {
2918		/*
2919		 * If your are next to put index then it's
2920		 * FIFO full condition
2921		 */
2922		if ((get_block == put_block) &&
2923		    (get_info.offset + 1) == put_info.offset) {
2924			DBG_PRINT(INTR_DBG, "%s: Ring Full\n",
2925				  ring_data->dev->name);
2926			break;
2927		}
2928		skb = (struct sk_buff *)((unsigned long)rxdp->Host_Control);
2929		if (skb == NULL) {
2930			DBG_PRINT(ERR_DBG, "%s: NULL skb in Rx Intr\n",
2931				  ring_data->dev->name);
2932			return 0;
2933		}
2934		if (ring_data->rxd_mode == RXD_MODE_1) {
2935			rxdp1 = (struct RxD1 *)rxdp;
2936			pci_unmap_single(ring_data->pdev, (dma_addr_t)
2937					 rxdp1->Buffer0_ptr,
2938					 ring_data->mtu +
2939					 HEADER_ETHERNET_II_802_3_SIZE +
2940					 HEADER_802_2_SIZE +
2941					 HEADER_SNAP_SIZE,
2942					 PCI_DMA_FROMDEVICE);
2943		} else if (ring_data->rxd_mode == RXD_MODE_3B) {
2944			rxdp3 = (struct RxD3 *)rxdp;
2945			pci_dma_sync_single_for_cpu(ring_data->pdev,
2946						    (dma_addr_t)rxdp3->Buffer0_ptr,
2947						    BUF0_LEN,
2948						    PCI_DMA_FROMDEVICE);
2949			pci_unmap_single(ring_data->pdev,
2950					 (dma_addr_t)rxdp3->Buffer2_ptr,
2951					 ring_data->mtu + 4,
2952					 PCI_DMA_FROMDEVICE);
2953		}
2954		prefetch(skb->data);
2955		rx_osm_handler(ring_data, rxdp);
2956		get_info.offset++;
2957		ring_data->rx_curr_get_info.offset = get_info.offset;
2958		rxdp = ring_data->rx_blocks[get_block].
2959			rxds[get_info.offset].virt_addr;
2960		if (get_info.offset == rxd_count[ring_data->rxd_mode]) {
2961			get_info.offset = 0;
2962			ring_data->rx_curr_get_info.offset = get_info.offset;
2963			get_block++;
2964			if (get_block == ring_data->block_count)
2965				get_block = 0;
2966			ring_data->rx_curr_get_info.block_index = get_block;
2967			rxdp = ring_data->rx_blocks[get_block].block_virt_addr;
2968		}
2969
2970		if (ring_data->nic->config.napi) {
2971			budget--;
2972			napi_pkts++;
2973			if (!budget)
2974				break;
2975		}
2976		pkt_cnt++;
2977		if ((indicate_max_pkts) && (pkt_cnt > indicate_max_pkts))
2978			break;
2979	}
2980	if (ring_data->lro) {
2981		/* Clear all LRO sessions before exiting */
2982		for (i = 0; i < MAX_LRO_SESSIONS; i++) {
2983			struct lro *lro = &ring_data->lro0_n[i];
2984			if (lro->in_use) {
2985				update_L3L4_header(ring_data->nic, lro);
2986				queue_rx_frame(lro->parent, lro->vlan_tag);
2987				clear_lro_session(lro);
2988			}
2989		}
2990	}
2991	return napi_pkts;
2992}
2993
2994/**
2995 *  tx_intr_handler - Transmit interrupt handler
2996 *  @nic : device private variable
2997 *  Description:
2998 *  If an interrupt was raised to indicate DMA complete of the
2999 *  Tx packet, this function is called. It identifies the last TxD
3000 *  whose buffer was freed and frees all skbs whose data have already
3001 *  DMA'ed into the NICs internal memory.
3002 *  Return Value:
3003 *  NONE
3004 */
3005
3006static void tx_intr_handler(struct fifo_info *fifo_data)
3007{
3008	struct s2io_nic *nic = fifo_data->nic;
3009	struct tx_curr_get_info get_info, put_info;
3010	struct sk_buff *skb = NULL;
3011	struct TxD *txdlp;
3012	int pkt_cnt = 0;
3013	unsigned long flags = 0;
3014	u8 err_mask;
3015	struct stat_block *stats = nic->mac_control.stats_info;
3016	struct swStat *swstats = &stats->sw_stat;
3017
3018	if (!spin_trylock_irqsave(&fifo_data->tx_lock, flags))
3019		return;
3020
3021	get_info = fifo_data->tx_curr_get_info;
3022	memcpy(&put_info, &fifo_data->tx_curr_put_info, sizeof(put_info));
3023	txdlp = fifo_data->list_info[get_info.offset].list_virt_addr;
3024	while ((!(txdlp->Control_1 & TXD_LIST_OWN_XENA)) &&
3025	       (get_info.offset != put_info.offset) &&
3026	       (txdlp->Host_Control)) {
3027		/* Check for TxD errors */
3028		if (txdlp->Control_1 & TXD_T_CODE) {
3029			unsigned long long err;
3030			err = txdlp->Control_1 & TXD_T_CODE;
3031			if (err & 0x1) {
3032				swstats->parity_err_cnt++;
3033			}
3034
3035			/* update t_code statistics */
3036			err_mask = err >> 48;
3037			switch (err_mask) {
3038			case 2:
3039				swstats->tx_buf_abort_cnt++;
3040				break;
3041
3042			case 3:
3043				swstats->tx_desc_abort_cnt++;
3044				break;
3045
3046			case 7:
3047				swstats->tx_parity_err_cnt++;
3048				break;
3049
3050			case 10:
3051				swstats->tx_link_loss_cnt++;
3052				break;
3053
3054			case 15:
3055				swstats->tx_list_proc_err_cnt++;
3056				break;
3057			}
3058		}
3059
3060		skb = s2io_txdl_getskb(fifo_data, txdlp, get_info.offset);
3061		if (skb == NULL) {
3062			spin_unlock_irqrestore(&fifo_data->tx_lock, flags);
3063			DBG_PRINT(ERR_DBG, "%s: NULL skb in Tx Free Intr\n",
3064				  __func__);
3065			return;
3066		}
3067		pkt_cnt++;
3068
3069		/* Updating the statistics block */
3070		swstats->mem_freed += skb->truesize;
3071		dev_kfree_skb_irq(skb);
3072
3073		get_info.offset++;
3074		if (get_info.offset == get_info.fifo_len + 1)
3075			get_info.offset = 0;
3076		txdlp = fifo_data->list_info[get_info.offset].list_virt_addr;
3077		fifo_data->tx_curr_get_info.offset = get_info.offset;
3078	}
3079
3080	s2io_wake_tx_queue(fifo_data, pkt_cnt, nic->config.multiq);
3081
3082	spin_unlock_irqrestore(&fifo_data->tx_lock, flags);
3083}
3084
3085/**
3086 *  s2io_mdio_write - Function to write in to MDIO registers
3087 *  @mmd_type : MMD type value (PMA/PMD/WIS/PCS/PHYXS)
3088 *  @addr     : address value
3089 *  @value    : data value
3090 *  @dev      : pointer to net_device structure
3091 *  Description:
3092 *  This function is used to write values to the MDIO registers
3093 *  NONE
3094 */
3095static void s2io_mdio_write(u32 mmd_type, u64 addr, u16 value,
3096			    struct net_device *dev)
3097{
3098	u64 val64;
3099	struct s2io_nic *sp = netdev_priv(dev);
3100	struct XENA_dev_config __iomem *bar0 = sp->bar0;
3101
3102	/* address transaction */
3103	val64 = MDIO_MMD_INDX_ADDR(addr) |
3104		MDIO_MMD_DEV_ADDR(mmd_type) |
3105		MDIO_MMS_PRT_ADDR(0x0);
3106	writeq(val64, &bar0->mdio_control);
3107	val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
3108	writeq(val64, &bar0->mdio_control);
3109	udelay(100);
3110
3111	/* Data transaction */
3112	val64 = MDIO_MMD_INDX_ADDR(addr) |
3113		MDIO_MMD_DEV_ADDR(mmd_type) |
3114		MDIO_MMS_PRT_ADDR(0x0) |
3115		MDIO_MDIO_DATA(value) |
3116		MDIO_OP(MDIO_OP_WRITE_TRANS);
3117	writeq(val64, &bar0->mdio_control);
3118	val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
3119	writeq(val64, &bar0->mdio_control);
3120	udelay(100);
3121
3122	val64 = MDIO_MMD_INDX_ADDR(addr) |
3123		MDIO_MMD_DEV_ADDR(mmd_type) |
3124		MDIO_MMS_PRT_ADDR(0x0) |
3125		MDIO_OP(MDIO_OP_READ_TRANS);
3126	writeq(val64, &bar0->mdio_control);
3127	val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
3128	writeq(val64, &bar0->mdio_control);
3129	udelay(100);
3130}
3131
3132/**
3133 *  s2io_mdio_read - Function to write in to MDIO registers
3134 *  @mmd_type : MMD type value (PMA/PMD/WIS/PCS/PHYXS)
3135 *  @addr     : address value
3136 *  @dev      : pointer to net_device structure
3137 *  Description:
3138 *  This function is used to read values to the MDIO registers
3139 *  NONE
3140 */
3141static u64 s2io_mdio_read(u32 mmd_type, u64 addr, struct net_device *dev)
3142{
3143	u64 val64 = 0x0;
3144	u64 rval64 = 0x0;
3145	struct s2io_nic *sp = netdev_priv(dev);
3146	struct XENA_dev_config __iomem *bar0 = sp->bar0;
3147
3148	/* address transaction */
3149	val64 = val64 | (MDIO_MMD_INDX_ADDR(addr)
3150			 | MDIO_MMD_DEV_ADDR(mmd_type)
3151			 | MDIO_MMS_PRT_ADDR(0x0));
3152	writeq(val64, &bar0->mdio_control);
3153	val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
3154	writeq(val64, &bar0->mdio_control);
3155	udelay(100);
3156
3157	/* Data transaction */
3158	val64 = MDIO_MMD_INDX_ADDR(addr) |
3159		MDIO_MMD_DEV_ADDR(mmd_type) |
3160		MDIO_MMS_PRT_ADDR(0x0) |
3161		MDIO_OP(MDIO_OP_READ_TRANS);
3162	writeq(val64, &bar0->mdio_control);
3163	val64 = val64 | MDIO_CTRL_START_TRANS(0xE);
3164	writeq(val64, &bar0->mdio_control);
3165	udelay(100);
3166
3167	/* Read the value from regs */
3168	rval64 = readq(&bar0->mdio_control);
3169	rval64 = rval64 & 0xFFFF0000;
3170	rval64 = rval64 >> 16;
3171	return rval64;
3172}
3173
3174/**
3175 *  s2io_chk_xpak_counter - Function to check the status of the xpak counters
3176 *  @counter      : counter value to be updated
3177 *  @flag         : flag to indicate the status
3178 *  @type         : counter type
3179 *  Description:
3180 *  This function is to check the status of the xpak counters value
3181 *  NONE
3182 */
3183
3184static void s2io_chk_xpak_counter(u64 *counter, u64 * regs_stat, u32 index,
3185				  u16 flag, u16 type)
3186{
3187	u64 mask = 0x3;
3188	u64 val64;
3189	int i;
3190	for (i = 0; i < index; i++)
3191		mask = mask << 0x2;
3192
3193	if (flag > 0) {
3194		*counter = *counter + 1;
3195		val64 = *regs_stat & mask;
3196		val64 = val64 >> (index * 0x2);
3197		val64 = val64 + 1;
3198		if (val64 == 3) {
3199			switch (type) {
3200			case 1:
3201				DBG_PRINT(ERR_DBG,
3202					  "Take Xframe NIC out of service.\n");
3203				DBG_PRINT(ERR_DBG,
3204"Excessive temperatures may result in premature transceiver failure.\n");
3205				break;
3206			case 2:
3207				DBG_PRINT(ERR_DBG,
3208					  "Take Xframe NIC out of service.\n");
3209				DBG_PRINT(ERR_DBG,
3210"Excessive bias currents may indicate imminent laser diode failure.\n");
3211				break;
3212			case 3:
3213				DBG_PRINT(ERR_DBG,
3214					  "Take Xframe NIC out of service.\n");
3215				DBG_PRINT(ERR_DBG,
3216"Excessive laser output power may saturate far-end receiver.\n");
3217				break;
3218			default:
3219				DBG_PRINT(ERR_DBG,
3220					  "Incorrect XPAK Alarm type\n");
3221			}
3222			val64 = 0x0;
3223		}
3224		val64 = val64 << (index * 0x2);
3225		*regs_stat = (*regs_stat & (~mask)) | (val64);
3226
3227	} else {
3228		*regs_stat = *regs_stat & (~mask);
3229	}
3230}
3231
3232/**
3233 *  s2io_updt_xpak_counter - Function to update the xpak counters
3234 *  @dev         : pointer to net_device struct
3235 *  Description:
3236 *  This function is to upate the status of the xpak counters value
3237 *  NONE
3238 */
3239static void s2io_updt_xpak_counter(struct net_device *dev)
3240{
3241	u16 flag  = 0x0;
3242	u16 type  = 0x0;
3243	u16 val16 = 0x0;
3244	u64 val64 = 0x0;
3245	u64 addr  = 0x0;
3246
3247	struct s2io_nic *sp = netdev_priv(dev);
3248	struct stat_block *stats = sp->mac_control.stats_info;
3249	struct xpakStat *xstats = &stats->xpak_stat;
3250
3251	/* Check the communication with the MDIO slave */
3252	addr = MDIO_CTRL1;
3253	val64 = 0x0;
3254	val64 = s2io_mdio_read(MDIO_MMD_PMAPMD, addr, dev);
3255	if ((val64 == 0xFFFF) || (val64 == 0x0000)) {
3256		DBG_PRINT(ERR_DBG,
3257			  "ERR: MDIO slave access failed - Returned %llx\n",
3258			  (unsigned long long)val64);
3259		return;
3260	}
3261
3262	/* Check for the expected value of control reg 1 */
3263	if (val64 != MDIO_CTRL1_SPEED10G) {
3264		DBG_PRINT(ERR_DBG, "Incorrect value at PMA address 0x0000 - "
3265			  "Returned: %llx- Expected: 0x%x\n",
3266			  (unsigned long long)val64, MDIO_CTRL1_SPEED10G);
3267		return;
3268	}
3269
3270	/* Loading the DOM register to MDIO register */
3271	addr = 0xA100;
3272	s2io_mdio_write(MDIO_MMD_PMAPMD, addr, val16, dev);
3273	val64 = s2io_mdio_read(MDIO_MMD_PMAPMD, addr, dev);
3274
3275	/* Reading the Alarm flags */
3276	addr = 0xA070;
3277	val64 = 0x0;
3278	val64 = s2io_mdio_read(MDIO_MMD_PMAPMD, addr, dev);
3279
3280	flag = CHECKBIT(val64, 0x7);
3281	type = 1;
3282	s2io_chk_xpak_counter(&xstats->alarm_transceiver_temp_high,
3283			      &xstats->xpak_regs_stat,
3284			      0x0, flag, type);
3285
3286	if (CHECKBIT(val64, 0x6))
3287		xstats->alarm_transceiver_temp_low++;
3288
3289	flag = CHECKBIT(val64, 0x3);
3290	type = 2;
3291	s2io_chk_xpak_counter(&xstats->alarm_laser_bias_current_high,
3292			      &xstats->xpak_regs_stat,
3293			      0x2, flag, type);
3294
3295	if (CHECKBIT(val64, 0x2))
3296		xstats->alarm_laser_bias_current_low++;
3297
3298	flag = CHECKBIT(val64, 0x1);
3299	type = 3;
3300	s2io_chk_xpak_counter(&xstats->alarm_laser_output_power_high,
3301			      &xstats->xpak_regs_stat,
3302			      0x4, flag, type);
3303
3304	if (CHECKBIT(val64, 0x0))
3305		xstats->alarm_laser_output_power_low++;
3306
3307	/* Reading the Warning flags */
3308	addr = 0xA074;
3309	val64 = 0x0;
3310	val64 = s2io_mdio_read(MDIO_MMD_PMAPMD, addr, dev);
3311
3312	if (CHECKBIT(val64, 0x7))
3313		xstats->warn_transceiver_temp_high++;
3314
3315	if (CHECKBIT(val64, 0x6))
3316		xstats->warn_transceiver_temp_low++;
3317
3318	if (CHECKBIT(val64, 0x3))
3319		xstats->warn_laser_bias_current_high++;
3320
3321	if (CHECKBIT(val64, 0x2))
3322		xstats->warn_laser_bias_current_low++;
3323
3324	if (CHECKBIT(val64, 0x1))
3325		xstats->warn_laser_output_power_high++;
3326
3327	if (CHECKBIT(val64, 0x0))
3328		xstats->warn_laser_output_power_low++;
3329}
3330
3331/**
3332 *  wait_for_cmd_complete - waits for a command to complete.
3333 *  @sp : private member of the device structure, which is a pointer to the
3334 *  s2io_nic structure.
3335 *  Description: Function that waits for a command to Write into RMAC
3336 *  ADDR DATA registers to be completed and returns either success or
3337 *  error depending on whether the command was complete or not.
3338 *  Return value:
3339 *   SUCCESS on success and FAILURE on failure.
3340 */
3341
3342static int wait_for_cmd_complete(void __iomem *addr, u64 busy_bit,
3343				 int bit_state)
3344{
3345	int ret = FAILURE, cnt = 0, delay = 1;
3346	u64 val64;
3347
3348	if ((bit_state != S2IO_BIT_RESET) && (bit_state != S2IO_BIT_SET))
3349		return FAILURE;
3350
3351	do {
3352		val64 = readq(addr);
3353		if (bit_state == S2IO_BIT_RESET) {
3354			if (!(val64 & busy_bit)) {
3355				ret = SUCCESS;
3356				break;
3357			}
3358		} else {
3359			if (val64 & busy_bit) {
3360				ret = SUCCESS;
3361				break;
3362			}
3363		}
3364
3365		if (in_interrupt())
3366			mdelay(delay);
3367		else
3368			msleep(delay);
3369
3370		if (++cnt >= 10)
3371			delay = 50;
3372	} while (cnt < 20);
3373	return ret;
3374}
3375/**
3376 * check_pci_device_id - Checks if the device id is supported
3377 * @id : device id
3378 * Description: Function to check if the pci device id is supported by driver.
3379 * Return value: Actual device id if supported else PCI_ANY_ID
3380 */
3381static u16 check_pci_device_id(u16 id)
3382{
3383	switch (id) {
3384	case PCI_DEVICE_ID_HERC_WIN:
3385	case PCI_DEVICE_ID_HERC_UNI:
3386		return XFRAME_II_DEVICE;
3387	case PCI_DEVICE_ID_S2IO_UNI:
3388	case PCI_DEVICE_ID_S2IO_WIN:
3389		return XFRAME_I_DEVICE;
3390	default:
3391		return PCI_ANY_ID;
3392	}
3393}
3394
3395/**
3396 *  s2io_reset - Resets the card.
3397 *  @sp : private member of the device structure.
3398 *  Description: Function to Reset the card. This function then also
3399 *  restores the previously saved PCI configuration space registers as
3400 *  the card reset also resets the configuration space.
3401 *  Return value:
3402 *  void.
3403 */
3404
3405static void s2io_reset(struct s2io_nic *sp)
3406{
3407	struct XENA_dev_config __iomem *bar0 = sp->bar0;
3408	u64 val64;
3409	u16 subid, pci_cmd;
3410	int i;
3411	u16 val16;
3412	unsigned long long up_cnt, down_cnt, up_time, down_time, reset_cnt;
3413	unsigned long long mem_alloc_cnt, mem_free_cnt, watchdog_cnt;
3414	struct stat_block *stats;
3415	struct swStat *swstats;
3416
3417	DBG_PRINT(INIT_DBG, "%s: Resetting XFrame card %s\n",
3418		  __func__, pci_name(sp->pdev));
3419
3420	/* Back up  the PCI-X CMD reg, dont want to lose MMRBC, OST settings */
3421	pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER, &(pci_cmd));
3422
3423	val64 = SW_RESET_ALL;
3424	writeq(val64, &bar0->sw_reset);
3425	if (strstr(sp->product_name, "CX4"))
3426		msleep(750);
3427	msleep(250);
3428	for (i = 0; i < S2IO_MAX_PCI_CONFIG_SPACE_REINIT; i++) {
3429
3430		/* Restore the PCI state saved during initialization. */
3431		pci_restore_state(sp->pdev);
3432		pci_save_state(sp->pdev);
3433		pci_read_config_word(sp->pdev, 0x2, &val16);
3434		if (check_pci_device_id(val16) != (u16)PCI_ANY_ID)
3435			break;
3436		msleep(200);
3437	}
3438
3439	if (check_pci_device_id(val16) == (u16)PCI_ANY_ID)
3440		DBG_PRINT(ERR_DBG, "%s SW_Reset failed!\n", __func__);
3441
3442	pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER, pci_cmd);
3443
3444	s2io_init_pci(sp);
3445
3446	/* Set swapper to enable I/O register access */
3447	s2io_set_swapper(sp);
3448
3449	/* restore mac_addr entries */
3450	do_s2io_restore_unicast_mc(sp);
3451
3452	/* Restore the MSIX table entries from local variables */
3453	restore_xmsi_data(sp);
3454
3455	/* Clear certain PCI/PCI-X fields after reset */
3456	if (sp->device_type == XFRAME_II_DEVICE) {
3457		/* Clear "detected parity error" bit */
3458		pci_write_config_word(sp->pdev, PCI_STATUS, 0x8000);
3459
3460		/* Clearing PCIX Ecc status register */
3461		pci_write_config_dword(sp->pdev, 0x68, 0x7C);
3462
3463		/* Clearing PCI_STATUS error reflected here */
3464		writeq(s2BIT(62), &bar0->txpic_int_reg);
3465	}
3466
3467	/* Reset device statistics maintained by OS */
3468	memset(&sp->stats, 0, sizeof(struct net_device_stats));
3469
3470	stats = sp->mac_control.stats_info;
3471	swstats = &stats->sw_stat;
3472
3473	/* save link up/down time/cnt, reset/memory/watchdog cnt */
3474	up_cnt = swstats->link_up_cnt;
3475	down_cnt = swstats->link_down_cnt;
3476	up_time = swstats->link_up_time;
3477	down_time = swstats->link_down_time;
3478	reset_cnt = swstats->soft_reset_cnt;
3479	mem_alloc_cnt = swstats->mem_allocated;
3480	mem_free_cnt = swstats->mem_freed;
3481	watchdog_cnt = swstats->watchdog_timer_cnt;
3482
3483	memset(stats, 0, sizeof(struct stat_block));
3484
3485	/* restore link up/down time/cnt, reset/memory/watchdog cnt */
3486	swstats->link_up_cnt = up_cnt;
3487	swstats->link_down_cnt = down_cnt;
3488	swstats->link_up_time = up_time;
3489	swstats->link_down_time = down_time;
3490	swstats->soft_reset_cnt = reset_cnt;
3491	swstats->mem_allocated = mem_alloc_cnt;
3492	swstats->mem_freed = mem_free_cnt;
3493	swstats->watchdog_timer_cnt = watchdog_cnt;
3494
3495	/* SXE-002: Configure link and activity LED to turn it off */
3496	subid = sp->pdev->subsystem_device;
3497	if (((subid & 0xFF) >= 0x07) &&
3498	    (sp->device_type == XFRAME_I_DEVICE)) {
3499		val64 = readq(&bar0->gpio_control);
3500		val64 |= 0x0000800000000000ULL;
3501		writeq(val64, &bar0->gpio_control);
3502		val64 = 0x0411040400000000ULL;
3503		writeq(val64, (void __iomem *)bar0 + 0x2700);
3504	}
3505
3506	/*
3507	 * Clear spurious ECC interrupts that would have occurred on
3508	 * XFRAME II cards after reset.
3509	 */
3510	if (sp->device_type == XFRAME_II_DEVICE) {
3511		val64 = readq(&bar0->pcc_err_reg);
3512		writeq(val64, &bar0->pcc_err_reg);
3513	}
3514
3515	sp->device_enabled_once = false;
3516}
3517
3518/**
3519 *  s2io_set_swapper - to set the swapper controle on the card
3520 *  @sp : private member of the device structure,
3521 *  pointer to the s2io_nic structure.
3522 *  Description: Function to set the swapper control on the card
3523 *  correctly depending on the 'endianness' of the system.
3524 *  Return value:
3525 *  SUCCESS on success and FAILURE on failure.
3526 */
3527
3528static int s2io_set_swapper(struct s2io_nic *sp)
3529{
3530	struct net_device *dev = sp->dev;
3531	struct XENA_dev_config __iomem *bar0 = sp->bar0;
3532	u64 val64, valt, valr;
3533
3534	/*
3535	 * Set proper endian settings and verify the same by reading
3536	 * the PIF Feed-back register.
3537	 */
3538
3539	val64 = readq(&bar0->pif_rd_swapper_fb);
3540	if (val64 != 0x0123456789ABCDEFULL) {
3541		int i = 0;
3542		static const u64 value[] = {
3543			0xC30000C3C30000C3ULL,	/* FE=1, SE=1 */
3544			0x8100008181000081ULL,	/* FE=1, SE=0 */
3545			0x4200004242000042ULL,	/* FE=0, SE=1 */
3546			0			/* FE=0, SE=0 */
3547		};
3548
3549		while (i < 4) {
3550			writeq(value[i], &bar0->swapper_ctrl);
3551			val64 = readq(&bar0->pif_rd_swapper_fb);
3552			if (val64 == 0x0123456789ABCDEFULL)
3553				break;
3554			i++;
3555		}
3556		if (i == 4) {
3557			DBG_PRINT(ERR_DBG, "%s: Endian settings are wrong, "
3558				  "feedback read %llx\n",
3559				  dev->name, (unsigned long long)val64);
3560			return FAILURE;
3561		}
3562		valr = value[i];
3563	} else {
3564		valr = readq(&bar0->swapper_ctrl);
3565	}
3566
3567	valt = 0x0123456789ABCDEFULL;
3568	writeq(valt, &bar0->xmsi_address);
3569	val64 = readq(&bar0->xmsi_address);
3570
3571	if (val64 != valt) {
3572		int i = 0;
3573		static const u64 value[] = {
3574			0x00C3C30000C3C300ULL,	/* FE=1, SE=1 */
3575			0x0081810000818100ULL,	/* FE=1, SE=0 */
3576			0x0042420000424200ULL,	/* FE=0, SE=1 */
3577			0			/* FE=0, SE=0 */
3578		};
3579
3580		while (i < 4) {
3581			writeq((value[i] | valr), &bar0->swapper_ctrl);
3582			writeq(valt, &bar0->xmsi_address);
3583			val64 = readq(&bar0->xmsi_address);
3584			if (val64 == valt)
3585				break;
3586			i++;
3587		}
3588		if (i == 4) {
3589			unsigned long long x = val64;
3590			DBG_PRINT(ERR_DBG,
3591				  "Write failed, Xmsi_addr reads:0x%llx\n", x);
3592			return FAILURE;
3593		}
3594	}
3595	val64 = readq(&bar0->swapper_ctrl);
3596	val64 &= 0xFFFF000000000000ULL;
3597
3598#ifdef __BIG_ENDIAN
3599	/*
3600	 * The device by default set to a big endian format, so a
3601	 * big endian driver need not set anything.
3602	 */
3603	val64 |= (SWAPPER_CTRL_TXP_FE |
3604		  SWAPPER_CTRL_TXP_SE |
3605		  SWAPPER_CTRL_TXD_R_FE |
3606		  SWAPPER_CTRL_TXD_W_FE |
3607		  SWAPPER_CTRL_TXF_R_FE |
3608		  SWAPPER_CTRL_RXD_R_FE |
3609		  SWAPPER_CTRL_RXD_W_FE |
3610		  SWAPPER_CTRL_RXF_W_FE |
3611		  SWAPPER_CTRL_XMSI_FE |
3612		  SWAPPER_CTRL_STATS_FE |
3613		  SWAPPER_CTRL_STATS_SE);
3614	if (sp->config.intr_type == INTA)
3615		val64 |= SWAPPER_CTRL_XMSI_SE;
3616	writeq(val64, &bar0->swapper_ctrl);
3617#else
3618	/*
3619	 * Initially we enable all bits to make it accessible by the
3620	 * driver, then we selectively enable only those bits that
3621	 * we want to set.
3622	 */
3623	val64 |= (SWAPPER_CTRL_TXP_FE |
3624		  SWAPPER_CTRL_TXP_SE |
3625		  SWAPPER_CTRL_TXD_R_FE |
3626		  SWAPPER_CTRL_TXD_R_SE |
3627		  SWAPPER_CTRL_TXD_W_FE |
3628		  SWAPPER_CTRL_TXD_W_SE |
3629		  SWAPPER_CTRL_TXF_R_FE |
3630		  SWAPPER_CTRL_RXD_R_FE |
3631		  SWAPPER_CTRL_RXD_R_SE |
3632		  SWAPPER_CTRL_RXD_W_FE |
3633		  SWAPPER_CTRL_RXD_W_SE |
3634		  SWAPPER_CTRL_RXF_W_FE |
3635		  SWAPPER_CTRL_XMSI_FE |
3636		  SWAPPER_CTRL_STATS_FE |
3637		  SWAPPER_CTRL_STATS_SE);
3638	if (sp->config.intr_type == INTA)
3639		val64 |= SWAPPER_CTRL_XMSI_SE;
3640	writeq(val64, &bar0->swapper_ctrl);
3641#endif
3642	val64 = readq(&bar0->swapper_ctrl);
3643
3644	/*
3645	 * Verifying if endian settings are accurate by reading a
3646	 * feedback register.
3647	 */
3648	val64 = readq(&bar0->pif_rd_swapper_fb);
3649	if (val64 != 0x0123456789ABCDEFULL) {
3650		/* Endian settings are incorrect, calls for another dekko. */
3651		DBG_PRINT(ERR_DBG,
3652			  "%s: Endian settings are wrong, feedback read %llx\n",
3653			  dev->name, (unsigned long long)val64);
3654		return FAILURE;
3655	}
3656
3657	return SUCCESS;
3658}
3659
3660static int wait_for_msix_trans(struct s2io_nic *nic, int i)
3661{
3662	struct XENA_dev_config __iomem *bar0 = nic->bar0;
3663	u64 val64;
3664	int ret = 0, cnt = 0;
3665
3666	do {
3667		val64 = readq(&bar0->xmsi_access);
3668		if (!(val64 & s2BIT(15)))
3669			break;
3670		mdelay(1);
3671		cnt++;
3672	} while (cnt < 5);
3673	if (cnt == 5) {
3674		DBG_PRINT(ERR_DBG, "XMSI # %d Access failed\n", i);
3675		ret = 1;
3676	}
3677
3678	return ret;
3679}
3680
3681static void restore_xmsi_data(struct s2io_nic *nic)
3682{
3683	struct XENA_dev_config __iomem *bar0 = nic->bar0;
3684	u64 val64;
3685	int i, msix_index;
3686
3687	if (nic->device_type == XFRAME_I_DEVICE)
3688		return;
3689
3690	for (i = 0; i < MAX_REQUESTED_MSI_X; i++) {
3691		msix_index = (i) ? ((i-1) * 8 + 1) : 0;
3692		writeq(nic->msix_info[i].addr, &bar0->xmsi_address);
3693		writeq(nic->msix_info[i].data, &bar0->xmsi_data);
3694		val64 = (s2BIT(7) | s2BIT(15) | vBIT(msix_index, 26, 6));
3695		writeq(val64, &bar0->xmsi_access);
3696		if (wait_for_msix_trans(nic, msix_index)) {
3697			DBG_PRINT(ERR_DBG, "%s: index: %d failed\n",
3698				  __func__, msix_index);
3699			continue;
3700		}
3701	}
3702}
3703
3704static void store_xmsi_data(struct s2io_nic *nic)
3705{
3706	struct XENA_dev_config __iomem *bar0 = nic->bar0;
3707	u64 val64, addr, data;
3708	int i, msix_index;
3709
3710	if (nic->device_type == XFRAME_I_DEVICE)
3711		return;
3712
3713	/* Store and display */
3714	for (i = 0; i < MAX_REQUESTED_MSI_X; i++) {
3715		msix_index = (i) ? ((i-1) * 8 + 1) : 0;
3716		val64 = (s2BIT(15) | vBIT(msix_index, 26, 6));
3717		writeq(val64, &bar0->xmsi_access);
3718		if (wait_for_msix_trans(nic, msix_index)) {
3719			DBG_PRINT(ERR_DBG, "%s: index: %d failed\n",
3720				  __func__, msix_index);
3721			continue;
3722		}
3723		addr = readq(&bar0->xmsi_address);
3724		data = readq(&bar0->xmsi_data);
3725		if (addr && data) {
3726			nic->msix_info[i].addr = addr;
3727			nic->msix_info[i].data = data;
3728		}
3729	}
3730}
3731
3732static int s2io_enable_msi_x(struct s2io_nic *nic)
3733{
3734	struct XENA_dev_config __iomem *bar0 = nic->bar0;
3735	u64 rx_mat;
3736	u16 msi_control; /* Temp variable */
3737	int ret, i, j, msix_indx = 1;
3738	int size;
3739	struct stat_block *stats = nic->mac_control.stats_info;
3740	struct swStat *swstats = &stats->sw_stat;
3741
3742	size = nic->num_entries * sizeof(struct msix_entry);
3743	nic->entries = kzalloc(size, GFP_KERNEL);
3744	if (!nic->entries) {
3745		DBG_PRINT(INFO_DBG, "%s: Memory allocation failed\n",
3746			  __func__);
3747		swstats->mem_alloc_fail_cnt++;
3748		return -ENOMEM;
3749	}
3750	swstats->mem_allocated += size;
3751
3752	size = nic->num_entries * sizeof(struct s2io_msix_entry);
3753	nic->s2io_entries = kzalloc(size, GFP_KERNEL);
3754	if (!nic->s2io_entries) {
3755		DBG_PRINT(INFO_DBG, "%s: Memory allocation failed\n",
3756			  __func__);
3757		swstats->mem_alloc_fail_cnt++;
3758		kfree(nic->entries);
3759		swstats->mem_freed
3760			+= (nic->num_entries * sizeof(struct msix_entry));
3761		return -ENOMEM;
3762	}
3763	swstats->mem_allocated += size;
3764
3765	nic->entries[0].entry = 0;
3766	nic->s2io_entries[0].entry = 0;
3767	nic->s2io_entries[0].in_use = MSIX_FLG;
3768	nic->s2io_entries[0].type = MSIX_ALARM_TYPE;
3769	nic->s2io_entries[0].arg = &nic->mac_control.fifos;
3770
3771	for (i = 1; i < nic->num_entries; i++) {
3772		nic->entries[i].entry = ((i - 1) * 8) + 1;
3773		nic->s2io_entries[i].entry = ((i - 1) * 8) + 1;
3774		nic->s2io_entries[i].arg = NULL;
3775		nic->s2io_entries[i].in_use = 0;
3776	}
3777
3778	rx_mat = readq(&bar0->rx_mat);
3779	for (j = 0; j < nic->config.rx_ring_num; j++) {
3780		rx_mat |= RX_MAT_SET(j, msix_indx);
3781		nic->s2io_entries[j+1].arg = &nic->mac_control.rings[j];
3782		nic->s2io_entries[j+1].type = MSIX_RING_TYPE;
3783		nic->s2io_entries[j+1].in_use = MSIX_FLG;
3784		msix_indx += 8;
3785	}
3786	writeq(rx_mat, &bar0->rx_mat);
3787	readq(&bar0->rx_mat);
3788
3789	ret = pci_enable_msix_range(nic->pdev, nic->entries,
3790				    nic->num_entries, nic->num_entries);
3791	/* We fail init if error or we get less vectors than min required */
3792	if (ret < 0) {
3793		DBG_PRINT(ERR_DBG, "Enabling MSI-X failed\n");
3794		kfree(nic->entries);
3795		swstats->mem_freed += nic->num_entries *
3796			sizeof(struct msix_entry);
3797		kfree(nic->s2io_entries);
3798		swstats->mem_freed += nic->num_entries *
3799			sizeof(struct s2io_msix_entry);
3800		nic->entries = NULL;
3801		nic->s2io_entries = NULL;
3802		return -ENOMEM;
3803	}
3804
3805	/*
3806	 * To enable MSI-X, MSI also needs to be enabled, due to a bug
3807	 * in the herc NIC. (Temp change, needs to be removed later)
3808	 */
3809	pci_read_config_word(nic->pdev, 0x42, &msi_control);
3810	msi_control |= 0x1; /* Enable MSI */
3811	pci_write_config_word(nic->pdev, 0x42, msi_control);
3812
3813	return 0;
3814}
3815
3816/* Handle software interrupt used during MSI(X) test */
3817static irqreturn_t s2io_test_intr(int irq, void *dev_id)
3818{
3819	struct s2io_nic *sp = dev_id;
3820
3821	sp->msi_detected = 1;
3822	wake_up(&sp->msi_wait);
3823
3824	return IRQ_HANDLED;
3825}
3826
3827/* Test interrupt path by forcing a a software IRQ */
3828static int s2io_test_msi(struct s2io_nic *sp)
3829{
3830	struct pci_dev *pdev = sp->pdev;
3831	struct XENA_dev_config __iomem *bar0 = sp->bar0;
3832	int err;
3833	u64 val64, saved64;
3834
3835	err = request_irq(sp->entries[1].vector, s2io_test_intr, 0,
3836			  sp->name, sp);
3837	if (err) {
3838		DBG_PRINT(ERR_DBG, "%s: PCI %s: cannot assign irq %d\n",
3839			  sp->dev->name, pci_name(pdev), pdev->irq);
3840		return err;
3841	}
3842
3843	init_waitqueue_head(&sp->msi_wait);
3844	sp->msi_detected = 0;
3845
3846	saved64 = val64 = readq(&bar0->scheduled_int_ctrl);
3847	val64 |= SCHED_INT_CTRL_ONE_SHOT;
3848	val64 |= SCHED_INT_CTRL_TIMER_EN;
3849	val64 |= SCHED_INT_CTRL_INT2MSI(1);
3850	writeq(val64, &bar0->scheduled_int_ctrl);
3851
3852	wait_event_timeout(sp->msi_wait, sp->msi_detected, HZ/10);
3853
3854	if (!sp->msi_detected) {
3855		/* MSI(X) test failed, go back to INTx mode */
3856		DBG_PRINT(ERR_DBG, "%s: PCI %s: No interrupt was generated "
3857			  "using MSI(X) during test\n",
3858			  sp->dev->name, pci_name(pdev));
3859
3860		err = -EOPNOTSUPP;
3861	}
3862
3863	free_irq(sp->entries[1].vector, sp);
3864
3865	writeq(saved64, &bar0->scheduled_int_ctrl);
3866
3867	return err;
3868}
3869
3870static void remove_msix_isr(struct s2io_nic *sp)
3871{
3872	int i;
3873	u16 msi_control;
3874
3875	for (i = 0; i < sp->num_entries; i++) {
3876		if (sp->s2io_entries[i].in_use == MSIX_REGISTERED_SUCCESS) {
3877			int vector = sp->entries[i].vector;
3878			void *arg = sp->s2io_entries[i].arg;
3879			free_irq(vector, arg);
3880		}
3881	}
3882
3883	kfree(sp->entries);
3884	kfree(sp->s2io_entries);
3885	sp->entries = NULL;
3886	sp->s2io_entries = NULL;
3887
3888	pci_read_config_word(sp->pdev, 0x42, &msi_control);
3889	msi_control &= 0xFFFE; /* Disable MSI */
3890	pci_write_config_word(sp->pdev, 0x42, msi_control);
3891
3892	pci_disable_msix(sp->pdev);
3893}
3894
3895static void remove_inta_isr(struct s2io_nic *sp)
3896{
3897	free_irq(sp->pdev->irq, sp->dev);
3898}
3899
3900/* ********************************************************* *
3901 * Functions defined below concern the OS part of the driver *
3902 * ********************************************************* */
3903
3904/**
3905 *  s2io_open - open entry point of the driver
3906 *  @dev : pointer to the device structure.
3907 *  Description:
3908 *  This function is the open entry point of the driver. It mainly calls a
3909 *  function to allocate Rx buffers and inserts them into the buffer
3910 *  descriptors and then enables the Rx part of the NIC.
3911 *  Return value:
3912 *  0 on success and an appropriate (-)ve integer as defined in errno.h
3913 *   file on failure.
3914 */
3915
3916static int s2io_open(struct net_device *dev)
3917{
3918	struct s2io_nic *sp = netdev_priv(dev);
3919	struct swStat *swstats = &sp->mac_control.stats_info->sw_stat;
3920	int err = 0;
3921
3922	/*
3923	 * Make sure you have link off by default every time
3924	 * Nic is initialized
3925	 */
3926	netif_carrier_off(dev);
3927	sp->last_link_state = 0;
3928
3929	/* Initialize H/W and enable interrupts */
3930	err = s2io_card_up(sp);
3931	if (err) {
3932		DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
3933			  dev->name);
3934		goto hw_init_failed;
3935	}
3936
3937	if (do_s2io_prog_unicast(dev, dev->dev_addr) == FAILURE) {
3938		DBG_PRINT(ERR_DBG, "Set Mac Address Failed\n");
3939		s2io_card_down(sp);
3940		err = -ENODEV;
3941		goto hw_init_failed;
3942	}
3943	s2io_start_all_tx_queue(sp);
3944	return 0;
3945
3946hw_init_failed:
3947	if (sp->config.intr_type == MSI_X) {
3948		if (sp->entries) {
3949			kfree(sp->entries);
3950			swstats->mem_freed += sp->num_entries *
3951				sizeof(struct msix_entry);
3952		}
3953		if (sp->s2io_entries) {
3954			kfree(sp->s2io_entries);
3955			swstats->mem_freed += sp->num_entries *
3956				sizeof(struct s2io_msix_entry);
3957		}
3958	}
3959	return err;
3960}
3961
3962/**
3963 *  s2io_close -close entry point of the driver
3964 *  @dev : device pointer.
3965 *  Description:
3966 *  This is the stop entry point of the driver. It needs to undo exactly
3967 *  whatever was done by the open entry point,thus it's usually referred to
3968 *  as the close function.Among other things this function mainly stops the
3969 *  Rx side of the NIC and frees all the Rx buffers in the Rx rings.
3970 *  Return value:
3971 *  0 on success and an appropriate (-)ve integer as defined in errno.h
3972 *  file on failure.
3973 */
3974
3975static int s2io_close(struct net_device *dev)
3976{
3977	struct s2io_nic *sp = netdev_priv(dev);
3978	struct config_param *config = &sp->config;
3979	u64 tmp64;
3980	int offset;
3981
3982	/* Return if the device is already closed               *
3983	 *  Can happen when s2io_card_up failed in change_mtu    *
3984	 */
3985	if (!is_s2io_card_up(sp))
3986		return 0;
3987
3988	s2io_stop_all_tx_queue(sp);
3989	/* delete all populated mac entries */
3990	for (offset = 1; offset < config->max_mc_addr; offset++) {
3991		tmp64 = do_s2io_read_unicast_mc(sp, offset);
3992		if (tmp64 != S2IO_DISABLE_MAC_ENTRY)
3993			do_s2io_delete_unicast_mc(sp, tmp64);
3994	}
3995
3996	s2io_card_down(sp);
3997
3998	return 0;
3999}
4000
4001/**
4002 *  s2io_xmit - Tx entry point of te driver
4003 *  @skb : the socket buffer containing the Tx data.
4004 *  @dev : device pointer.
4005 *  Description :
4006 *  This function is the Tx entry point of the driver. S2IO NIC supports
4007 *  certain protocol assist features on Tx side, namely  CSO, S/G, LSO.
4008 *  NOTE: when device can't queue the pkt,just the trans_start variable will
4009 *  not be upadted.
4010 *  Return value:
4011 *  0 on success & 1 on failure.
4012 */
4013
4014static netdev_tx_t s2io_xmit(struct sk_buff *skb, struct net_device *dev)
4015{
4016	struct s2io_nic *sp = netdev_priv(dev);
4017	u16 frg_cnt, frg_len, i, queue, queue_len, put_off, get_off;
4018	register u64 val64;
4019	struct TxD *txdp;
4020	struct TxFIFO_element __iomem *tx_fifo;
4021	unsigned long flags = 0;
4022	u16 vlan_tag = 0;
4023	struct fifo_info *fifo = NULL;
4024	int do_spin_lock = 1;
4025	int offload_type;
4026	int enable_per_list_interrupt = 0;
4027	struct config_param *config = &sp->config;
4028	struct mac_info *mac_control = &sp->mac_control;
4029	struct stat_block *stats = mac_control->stats_info;
4030	struct swStat *swstats = &stats->sw_stat;
4031
4032	DBG_PRINT(TX_DBG, "%s: In Neterion Tx routine\n", dev->name);
4033
4034	if (unlikely(skb->len <= 0)) {
4035		DBG_PRINT(TX_DBG, "%s: Buffer has no data..\n", dev->name);
4036		dev_kfree_skb_any(skb);
4037		return NETDEV_TX_OK;
4038	}
4039
4040	if (!is_s2io_card_up(sp)) {
4041		DBG_PRINT(TX_DBG, "%s: Card going down for reset\n",
4042			  dev->name);
4043		dev_kfree_skb_any(skb);
4044		return NETDEV_TX_OK;
4045	}
4046
4047	queue = 0;
4048	if (vlan_tx_tag_present(skb))
4049		vlan_tag = vlan_tx_tag_get(skb);
4050	if (sp->config.tx_steering_type == TX_DEFAULT_STEERING) {
4051		if (skb->protocol == htons(ETH_P_IP)) {
4052			struct iphdr *ip;
4053			struct tcphdr *th;
4054			ip = ip_hdr(skb);
4055
4056			if (!ip_is_fragment(ip)) {
4057				th = (struct tcphdr *)(((unsigned char *)ip) +
4058						       ip->ihl*4);
4059
4060				if (ip->protocol == IPPROTO_TCP) {
4061					queue_len = sp->total_tcp_fifos;
4062					queue = (ntohs(th->source) +
4063						 ntohs(th->dest)) &
4064						sp->fifo_selector[queue_len - 1];
4065					if (queue >= queue_len)
4066						queue = queue_len - 1;
4067				} else if (ip->protocol == IPPROTO_UDP) {
4068					queue_len = sp->total_udp_fifos;
4069					queue = (ntohs(th->source) +
4070						 ntohs(th->dest)) &
4071						sp->fifo_selector[queue_len - 1];
4072					if (queue >= queue_len)
4073						queue = queue_len - 1;
4074					queue += sp->udp_fifo_idx;
4075					if (skb->len > 1024)
4076						enable_per_list_interrupt = 1;
4077					do_spin_lock = 0;
4078				}
4079			}
4080		}
4081	} else if (sp->config.tx_steering_type == TX_PRIORITY_STEERING)
4082		/* get fifo number based on skb->priority value */
4083		queue = config->fifo_mapping
4084			[skb->priority & (MAX_TX_FIFOS - 1)];
4085	fifo = &mac_control->fifos[queue];
4086
4087	if (do_spin_lock)
4088		spin_lock_irqsave(&fifo->tx_lock, flags);
4089	else {
4090		if (unlikely(!spin_trylock_irqsave(&fifo->tx_lock, flags)))
4091			return NETDEV_TX_LOCKED;
4092	}
4093
4094	if (sp->config.multiq) {
4095		if (__netif_subqueue_stopped(dev, fifo->fifo_no)) {
4096			spin_unlock_irqrestore(&fifo->tx_lock, flags);
4097			return NETDEV_TX_BUSY;
4098		}
4099	} else if (unlikely(fifo->queue_state == FIFO_QUEUE_STOP)) {
4100		if (netif_queue_stopped(dev)) {
4101			spin_unlock_irqrestore(&fifo->tx_lock, flags);
4102			return NETDEV_TX_BUSY;
4103		}
4104	}
4105
4106	put_off = (u16)fifo->tx_curr_put_info.offset;
4107	get_off = (u16)fifo->tx_curr_get_info.offset;
4108	txdp = fifo->list_info[put_off].list_virt_addr;
4109
4110	queue_len = fifo->tx_curr_put_info.fifo_len + 1;
4111	/* Avoid "put" pointer going beyond "get" pointer */
4112	if (txdp->Host_Control ||
4113	    ((put_off+1) == queue_len ? 0 : (put_off+1)) == get_off) {
4114		DBG_PRINT(TX_DBG, "Error in xmit, No free TXDs.\n");
4115		s2io_stop_tx_queue(sp, fifo->fifo_no);
4116		dev_kfree_skb_any(skb);
4117		spin_unlock_irqrestore(&fifo->tx_lock, flags);
4118		return NETDEV_TX_OK;
4119	}
4120
4121	offload_type = s2io_offload_type(skb);
4122	if (offload_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)) {
4123		txdp->Control_1 |= TXD_TCP_LSO_EN;
4124		txdp->Control_1 |= TXD_TCP_LSO_MSS(s2io_tcp_mss(skb));
4125	}
4126	if (skb->ip_summed == CHECKSUM_PARTIAL) {
4127		txdp->Control_2 |= (TXD_TX_CKO_IPV4_EN |
4128				    TXD_TX_CKO_TCP_EN |
4129				    TXD_TX_CKO_UDP_EN);
4130	}
4131	txdp->Control_1 |= TXD_GATHER_CODE_FIRST;
4132	txdp->Control_1 |= TXD_LIST_OWN_XENA;
4133	txdp->Control_2 |= TXD_INT_NUMBER(fifo->fifo_no);
4134	if (enable_per_list_interrupt)
4135		if (put_off & (queue_len >> 5))
4136			txdp->Control_2 |= TXD_INT_TYPE_PER_LIST;
4137	if (vlan_tag) {
4138		txdp->Control_2 |= TXD_VLAN_ENABLE;
4139		txdp->Control_2 |= TXD_VLAN_TAG(vlan_tag);
4140	}
4141
4142	frg_len = skb_headlen(skb);
4143	if (offload_type == SKB_GSO_UDP) {
4144		int ufo_size;
4145
4146		ufo_size = s2io_udp_mss(skb);
4147		ufo_size &= ~7;
4148		txdp->Control_1 |= TXD_UFO_EN;
4149		txdp->Control_1 |= TXD_UFO_MSS(ufo_size);
4150		txdp->Control_1 |= TXD_BUFFER0_SIZE(8);
4151#ifdef __BIG_ENDIAN
4152		/* both variants do cpu_to_be64(be32_to_cpu(...)) */
4153		fifo->ufo_in_band_v[put_off] =
4154			(__force u64)skb_shinfo(skb)->ip6_frag_id;
4155#else
4156		fifo->ufo_in_band_v[put_off] =
4157			(__force u64)skb_shinfo(skb)->ip6_frag_id << 32;
4158#endif
4159		txdp->Host_Control = (unsigned long)fifo->ufo_in_band_v;
4160		txdp->Buffer_Pointer = pci_map_single(sp->pdev,
4161						      fifo->ufo_in_band_v,
4162						      sizeof(u64),
4163						      PCI_DMA_TODEVICE);
4164		if (pci_dma_mapping_error(sp->pdev, txdp->Buffer_Pointer))
4165			goto pci_map_failed;
4166		txdp++;
4167	}
4168
4169	txdp->Buffer_Pointer = pci_map_single(sp->pdev, skb->data,
4170					      frg_len, PCI_DMA_TODEVICE);
4171	if (pci_dma_mapping_error(sp->pdev, txdp->Buffer_Pointer))
4172		goto pci_map_failed;
4173
4174	txdp->Host_Control = (unsigned long)skb;
4175	txdp->Control_1 |= TXD_BUFFER0_SIZE(frg_len);
4176	if (offload_type == SKB_GSO_UDP)
4177		txdp->Control_1 |= TXD_UFO_EN;
4178
4179	frg_cnt = skb_shinfo(skb)->nr_frags;
4180	/* For fragmented SKB. */
4181	for (i = 0; i < frg_cnt; i++) {
4182		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
4183		/* A '0' length fragment will be ignored */
4184		if (!skb_frag_size(frag))
4185			continue;
4186		txdp++;
4187		txdp->Buffer_Pointer = (u64)skb_frag_dma_map(&sp->pdev->dev,
4188							     frag, 0,
4189							     skb_frag_size(frag),
4190							     DMA_TO_DEVICE);
4191		txdp->Control_1 = TXD_BUFFER0_SIZE(skb_frag_size(frag));
4192		if (offload_type == SKB_GSO_UDP)
4193			txdp->Control_1 |= TXD_UFO_EN;
4194	}
4195	txdp->Control_1 |= TXD_GATHER_CODE_LAST;
4196
4197	if (offload_type == SKB_GSO_UDP)
4198		frg_cnt++; /* as Txd0 was used for inband header */
4199
4200	tx_fifo = mac_control->tx_FIFO_start[queue];
4201	val64 = fifo->list_info[put_off].list_phy_addr;
4202	writeq(val64, &tx_fifo->TxDL_Pointer);
4203
4204	val64 = (TX_FIFO_LAST_TXD_NUM(frg_cnt) | TX_FIFO_FIRST_LIST |
4205		 TX_FIFO_LAST_LIST);
4206	if (offload_type)
4207		val64 |= TX_FIFO_SPECIAL_FUNC;
4208
4209	writeq(val64, &tx_fifo->List_Control);
4210
4211	mmiowb();
4212
4213	put_off++;
4214	if (put_off == fifo->tx_curr_put_info.fifo_len + 1)
4215		put_off = 0;
4216	fifo->tx_curr_put_info.offset = put_off;
4217
4218	/* Avoid "put" pointer going beyond "get" pointer */
4219	if (((put_off+1) == queue_len ? 0 : (put_off+1)) == get_off) {
4220		swstats->fifo_full_cnt++;
4221		DBG_PRINT(TX_DBG,
4222			  "No free TxDs for xmit, Put: 0x%x Get:0x%x\n",
4223			  put_off, get_off);
4224		s2io_stop_tx_queue(sp, fifo->fifo_no);
4225	}
4226	swstats->mem_allocated += skb->truesize;
4227	spin_unlock_irqrestore(&fifo->tx_lock, flags);
4228
4229	if (sp->config.intr_type == MSI_X)
4230		tx_intr_handler(fifo);
4231
4232	return NETDEV_TX_OK;
4233
4234pci_map_failed:
4235	swstats->pci_map_fail_cnt++;
4236	s2io_stop_tx_queue(sp, fifo->fifo_no);
4237	swstats->mem_freed += skb->truesize;
4238	dev_kfree_skb_any(skb);
4239	spin_unlock_irqrestore(&fifo->tx_lock, flags);
4240	return NETDEV_TX_OK;
4241}
4242
4243static void
4244s2io_alarm_handle(unsigned long data)
4245{
4246	struct s2io_nic *sp = (struct s2io_nic *)data;
4247	struct net_device *dev = sp->dev;
4248
4249	s2io_handle_errors(dev);
4250	mod_timer(&sp->alarm_timer, jiffies + HZ / 2);
4251}
4252
4253static irqreturn_t s2io_msix_ring_handle(int irq, void *dev_id)
4254{
4255	struct ring_info *ring = (struct ring_info *)dev_id;
4256	struct s2io_nic *sp = ring->nic;
4257	struct XENA_dev_config __iomem *bar0 = sp->bar0;
4258
4259	if (unlikely(!is_s2io_card_up(sp)))
4260		return IRQ_HANDLED;
4261
4262	if (sp->config.napi) {
4263		u8 __iomem *addr = NULL;
4264		u8 val8 = 0;
4265
4266		addr = (u8 __iomem *)&bar0->xmsi_mask_reg;
4267		addr += (7 - ring->ring_no);
4268		val8 = (ring->ring_no == 0) ? 0x7f : 0xff;
4269		writeb(val8, addr);
4270		val8 = readb(addr);
4271		napi_schedule(&ring->napi);
4272	} else {
4273		rx_intr_handler(ring, 0);
4274		s2io_chk_rx_buffers(sp, ring);
4275	}
4276
4277	return IRQ_HANDLED;
4278}
4279
4280static irqreturn_t s2io_msix_fifo_handle(int irq, void *dev_id)
4281{
4282	int i;
4283	struct fifo_info *fifos = (struct fifo_info *)dev_id;
4284	struct s2io_nic *sp = fifos->nic;
4285	struct XENA_dev_config __iomem *bar0 = sp->bar0;
4286	struct config_param *config  = &sp->config;
4287	u64 reason;
4288
4289	if (unlikely(!is_s2io_card_up(sp)))
4290		return IRQ_NONE;
4291
4292	reason = readq(&bar0->general_int_status);
4293	if (unlikely(reason == S2IO_MINUS_ONE))
4294		/* Nothing much can be done. Get out */
4295		return IRQ_HANDLED;
4296
4297	if (reason & (GEN_INTR_TXPIC | GEN_INTR_TXTRAFFIC)) {
4298		writeq(S2IO_MINUS_ONE, &bar0->general_int_mask);
4299
4300		if (reason & GEN_INTR_TXPIC)
4301			s2io_txpic_intr_handle(sp);
4302
4303		if (reason & GEN_INTR_TXTRAFFIC)
4304			writeq(S2IO_MINUS_ONE, &bar0->tx_traffic_int);
4305
4306		for (i = 0; i < config->tx_fifo_num; i++)
4307			tx_intr_handler(&fifos[i]);
4308
4309		writeq(sp->general_int_mask, &bar0->general_int_mask);
4310		readl(&bar0->general_int_status);
4311		return IRQ_HANDLED;
4312	}
4313	/* The interrupt was not raised by us */
4314	return IRQ_NONE;
4315}
4316
4317static void s2io_txpic_intr_handle(struct s2io_nic *sp)
4318{
4319	struct XENA_dev_config __iomem *bar0 = sp->bar0;
4320	u64 val64;
4321
4322	val64 = readq(&bar0->pic_int_status);
4323	if (val64 & PIC_INT_GPIO) {
4324		val64 = readq(&bar0->gpio_int_reg);
4325		if ((val64 & GPIO_INT_REG_LINK_DOWN) &&
4326		    (val64 & GPIO_INT_REG_LINK_UP)) {
4327			/*
4328			 * This is unstable state so clear both up/down
4329			 * interrupt and adapter to re-evaluate the link state.
4330			 */
4331			val64 |= GPIO_INT_REG_LINK_DOWN;
4332			val64 |= GPIO_INT_REG_LINK_UP;
4333			writeq(val64, &bar0->gpio_int_reg);
4334			val64 = readq(&bar0->gpio_int_mask);
4335			val64 &= ~(GPIO_INT_MASK_LINK_UP |
4336				   GPIO_INT_MASK_LINK_DOWN);
4337			writeq(val64, &bar0->gpio_int_mask);
4338		} else if (val64 & GPIO_INT_REG_LINK_UP) {
4339			val64 = readq(&bar0->adapter_status);
4340			/* Enable Adapter */
4341			val64 = readq(&bar0->adapter_control);
4342			val64 |= ADAPTER_CNTL_EN;
4343			writeq(val64, &bar0->adapter_control);
4344			val64 |= ADAPTER_LED_ON;
4345			writeq(val64, &bar0->adapter_control);
4346			if (!sp->device_enabled_once)
4347				sp->device_enabled_once = 1;
4348
4349			s2io_link(sp, LINK_UP);
4350			/*
4351			 * unmask link down interrupt and mask link-up
4352			 * intr
4353			 */
4354			val64 = readq(&bar0->gpio_int_mask);
4355			val64 &= ~GPIO_INT_MASK_LINK_DOWN;
4356			val64 |= GPIO_INT_MASK_LINK_UP;
4357			writeq(val64, &bar0->gpio_int_mask);
4358
4359		} else if (val64 & GPIO_INT_REG_LINK_DOWN) {
4360			val64 = readq(&bar0->adapter_status);
4361			s2io_link(sp, LINK_DOWN);
4362			/* Link is down so unmaks link up interrupt */
4363			val64 = readq(&bar0->gpio_int_mask);
4364			val64 &= ~GPIO_INT_MASK_LINK_UP;
4365			val64 |= GPIO_INT_MASK_LINK_DOWN;
4366			writeq(val64, &bar0->gpio_int_mask);
4367
4368			/* turn off LED */
4369			val64 = readq(&bar0->adapter_control);
4370			val64 = val64 & (~ADAPTER_LED_ON);
4371			writeq(val64, &bar0->adapter_control);
4372		}
4373	}
4374	val64 = readq(&bar0->gpio_int_mask);
4375}
4376
4377/**
4378 *  do_s2io_chk_alarm_bit - Check for alarm and incrment the counter
4379 *  @value: alarm bits
4380 *  @addr: address value
4381 *  @cnt: counter variable
4382 *  Description: Check for alarm and increment the counter
4383 *  Return Value:
4384 *  1 - if alarm bit set
4385 *  0 - if alarm bit is not set
4386 */
4387static int do_s2io_chk_alarm_bit(u64 value, void __iomem *addr,
4388				 unsigned long long *cnt)
4389{
4390	u64 val64;
4391	val64 = readq(addr);
4392	if (val64 & value) {
4393		writeq(val64, addr);
4394		(*cnt)++;
4395		return 1;
4396	}
4397	return 0;
4398
4399}
4400
4401/**
4402 *  s2io_handle_errors - Xframe error indication handler
4403 *  @nic: device private variable
4404 *  Description: Handle alarms such as loss of link, single or
4405 *  double ECC errors, critical and serious errors.
4406 *  Return Value:
4407 *  NONE
4408 */
4409static void s2io_handle_errors(void *dev_id)
4410{
4411	struct net_device *dev = (struct net_device *)dev_id;
4412	struct s2io_nic *sp = netdev_priv(dev);
4413	struct XENA_dev_config __iomem *bar0 = sp->bar0;
4414	u64 temp64 = 0, val64 = 0;
4415	int i = 0;
4416
4417	struct swStat *sw_stat = &sp->mac_control.stats_info->sw_stat;
4418	struct xpakStat *stats = &sp->mac_control.stats_info->xpak_stat;
4419
4420	if (!is_s2io_card_up(sp))
4421		return;
4422
4423	if (pci_channel_offline(sp->pdev))
4424		return;
4425
4426	memset(&sw_stat->ring_full_cnt, 0,
4427	       sizeof(sw_stat->ring_full_cnt));
4428
4429	/* Handling the XPAK counters update */
4430	if (stats->xpak_timer_count < 72000) {
4431		/* waiting for an hour */
4432		stats->xpak_timer_count++;
4433	} else {
4434		s2io_updt_xpak_counter(dev);
4435		/* reset the count to zero */
4436		stats->xpak_timer_count = 0;
4437	}
4438
4439	/* Handling link status change error Intr */
4440	if (s2io_link_fault_indication(sp) == MAC_RMAC_ERR_TIMER) {
4441		val64 = readq(&bar0->mac_rmac_err_reg);
4442		writeq(val64, &bar0->mac_rmac_err_reg);
4443		if (val64 & RMAC_LINK_STATE_CHANGE_INT)
4444			schedule_work(&sp->set_link_task);
4445	}
4446
4447	/* In case of a serious error, the device will be Reset. */
4448	if (do_s2io_chk_alarm_bit(SERR_SOURCE_ANY, &bar0->serr_source,
4449				  &sw_stat->serious_err_cnt))
4450		goto reset;
4451
4452	/* Check for data parity error */
4453	if (do_s2io_chk_alarm_bit(GPIO_INT_REG_DP_ERR_INT, &bar0->gpio_int_reg,
4454				  &sw_stat->parity_err_cnt))
4455		goto reset;
4456
4457	/* Check for ring full counter */
4458	if (sp->device_type == XFRAME_II_DEVICE) {
4459		val64 = readq(&bar0->ring_bump_counter1);
4460		for (i = 0; i < 4; i++) {
4461			temp64 = (val64 & vBIT(0xFFFF, (i*16), 16));
4462			temp64 >>= 64 - ((i+1)*16);
4463			sw_stat->ring_full_cnt[i] += temp64;
4464		}
4465
4466		val64 = readq(&bar0->ring_bump_counter2);
4467		for (i = 0; i < 4; i++) {
4468			temp64 = (val64 & vBIT(0xFFFF, (i*16), 16));
4469			temp64 >>= 64 - ((i+1)*16);
4470			sw_stat->ring_full_cnt[i+4] += temp64;
4471		}
4472	}
4473
4474	val64 = readq(&bar0->txdma_int_status);
4475	/*check for pfc_err*/
4476	if (val64 & TXDMA_PFC_INT) {
4477		if (do_s2io_chk_alarm_bit(PFC_ECC_DB_ERR | PFC_SM_ERR_ALARM |
4478					  PFC_MISC_0_ERR | PFC_MISC_1_ERR |
4479					  PFC_PCIX_ERR,
4480					  &bar0->pfc_err_reg,
4481					  &sw_stat->pfc_err_cnt))
4482			goto reset;
4483		do_s2io_chk_alarm_bit(PFC_ECC_SG_ERR,
4484				      &bar0->pfc_err_reg,
4485				      &sw_stat->pfc_err_cnt);
4486	}
4487
4488	/*check for tda_err*/
4489	if (val64 & TXDMA_TDA_INT) {
4490		if (do_s2io_chk_alarm_bit(TDA_Fn_ECC_DB_ERR |
4491					  TDA_SM0_ERR_ALARM |
4492					  TDA_SM1_ERR_ALARM,
4493					  &bar0->tda_err_reg,
4494					  &sw_stat->tda_err_cnt))
4495			goto reset;
4496		do_s2io_chk_alarm_bit(TDA_Fn_ECC_SG_ERR | TDA_PCIX_ERR,
4497				      &bar0->tda_err_reg,
4498				      &sw_stat->tda_err_cnt);
4499	}
4500	/*check for pcc_err*/
4501	if (val64 & TXDMA_PCC_INT) {
4502		if (do_s2io_chk_alarm_bit(PCC_SM_ERR_ALARM | PCC_WR_ERR_ALARM |
4503					  PCC_N_SERR | PCC_6_COF_OV_ERR |
4504					  PCC_7_COF_OV_ERR | PCC_6_LSO_OV_ERR |
4505					  PCC_7_LSO_OV_ERR | PCC_FB_ECC_DB_ERR |
4506					  PCC_TXB_ECC_DB_ERR,
4507					  &bar0->pcc_err_reg,
4508					  &sw_stat->pcc_err_cnt))
4509			goto reset;
4510		do_s2io_chk_alarm_bit(PCC_FB_ECC_SG_ERR | PCC_TXB_ECC_SG_ERR,
4511				      &bar0->pcc_err_reg,
4512				      &sw_stat->pcc_err_cnt);
4513	}
4514
4515	/*check for tti_err*/
4516	if (val64 & TXDMA_TTI_INT) {
4517		if (do_s2io_chk_alarm_bit(TTI_SM_ERR_ALARM,
4518					  &bar0->tti_err_reg,
4519					  &sw_stat->tti_err_cnt))
4520			goto reset;
4521		do_s2io_chk_alarm_bit(TTI_ECC_SG_ERR | TTI_ECC_DB_ERR,
4522				      &bar0->tti_err_reg,
4523				      &sw_stat->tti_err_cnt);
4524	}
4525
4526	/*check for lso_err*/
4527	if (val64 & TXDMA_LSO_INT) {
4528		if (do_s2io_chk_alarm_bit(LSO6_ABORT | LSO7_ABORT |
4529					  LSO6_SM_ERR_ALARM | LSO7_SM_ERR_ALARM,
4530					  &bar0->lso_err_reg,
4531					  &sw_stat->lso_err_cnt))
4532			goto reset;
4533		do_s2io_chk_alarm_bit(LSO6_SEND_OFLOW | LSO7_SEND_OFLOW,
4534				      &bar0->lso_err_reg,
4535				      &sw_stat->lso_err_cnt);
4536	}
4537
4538	/*check for tpa_err*/
4539	if (val64 & TXDMA_TPA_INT) {
4540		if (do_s2io_chk_alarm_bit(TPA_SM_ERR_ALARM,
4541					  &bar0->tpa_err_reg,
4542					  &sw_stat->tpa_err_cnt))
4543			goto reset;
4544		do_s2io_chk_alarm_bit(TPA_TX_FRM_DROP,
4545				      &bar0->tpa_err_reg,
4546				      &sw_stat->tpa_err_cnt);
4547	}
4548
4549	/*check for sm_err*/
4550	if (val64 & TXDMA_SM_INT) {
4551		if (do_s2io_chk_alarm_bit(SM_SM_ERR_ALARM,
4552					  &bar0->sm_err_reg,
4553					  &sw_stat->sm_err_cnt))
4554			goto reset;
4555	}
4556
4557	val64 = readq(&bar0->mac_int_status);
4558	if (val64 & MAC_INT_STATUS_TMAC_INT) {
4559		if (do_s2io_chk_alarm_bit(TMAC_TX_BUF_OVRN | TMAC_TX_SM_ERR,
4560					  &bar0->mac_tmac_err_reg,
4561					  &sw_stat->mac_tmac_err_cnt))
4562			goto reset;
4563		do_s2io_chk_alarm_bit(TMAC_ECC_SG_ERR | TMAC_ECC_DB_ERR |
4564				      TMAC_DESC_ECC_SG_ERR |
4565				      TMAC_DESC_ECC_DB_ERR,
4566				      &bar0->mac_tmac_err_reg,
4567				      &sw_stat->mac_tmac_err_cnt);
4568	}
4569
4570	val64 = readq(&bar0->xgxs_int_status);
4571	if (val64 & XGXS_INT_STATUS_TXGXS) {
4572		if (do_s2io_chk_alarm_bit(TXGXS_ESTORE_UFLOW | TXGXS_TX_SM_ERR,
4573					  &bar0->xgxs_txgxs_err_reg,
4574					  &sw_stat->xgxs_txgxs_err_cnt))
4575			goto reset;
4576		do_s2io_chk_alarm_bit(TXGXS_ECC_SG_ERR | TXGXS_ECC_DB_ERR,
4577				      &bar0->xgxs_txgxs_err_reg,
4578				      &sw_stat->xgxs_txgxs_err_cnt);
4579	}
4580
4581	val64 = readq(&bar0->rxdma_int_status);
4582	if (val64 & RXDMA_INT_RC_INT_M) {
4583		if (do_s2io_chk_alarm_bit(RC_PRCn_ECC_DB_ERR |
4584					  RC_FTC_ECC_DB_ERR |
4585					  RC_PRCn_SM_ERR_ALARM |
4586					  RC_FTC_SM_ERR_ALARM,
4587					  &bar0->rc_err_reg,
4588					  &sw_stat->rc_err_cnt))
4589			goto reset;
4590		do_s2io_chk_alarm_bit(RC_PRCn_ECC_SG_ERR |
4591				      RC_FTC_ECC_SG_ERR |
4592				      RC_RDA_FAIL_WR_Rn, &bar0->rc_err_reg,
4593				      &sw_stat->rc_err_cnt);
4594		if (do_s2io_chk_alarm_bit(PRC_PCI_AB_RD_Rn |
4595					  PRC_PCI_AB_WR_Rn |
4596					  PRC_PCI_AB_F_WR_Rn,
4597					  &bar0->prc_pcix_err_reg,
4598					  &sw_stat->prc_pcix_err_cnt))
4599			goto reset;
4600		do_s2io_chk_alarm_bit(PRC_PCI_DP_RD_Rn |
4601				      PRC_PCI_DP_WR_Rn |
4602				      PRC_PCI_DP_F_WR_Rn,
4603				      &bar0->prc_pcix_err_reg,
4604				      &sw_stat->prc_pcix_err_cnt);
4605	}
4606
4607	if (val64 & RXDMA_INT_RPA_INT_M) {
4608		if (do_s2io_chk_alarm_bit(RPA_SM_ERR_ALARM | RPA_CREDIT_ERR,
4609					  &bar0->rpa_err_reg,
4610					  &sw_stat->rpa_err_cnt))
4611			goto reset;
4612		do_s2io_chk_alarm_bit(RPA_ECC_SG_ERR | RPA_ECC_DB_ERR,
4613				      &bar0->rpa_err_reg,
4614				      &sw_stat->rpa_err_cnt);
4615	}
4616
4617	if (val64 & RXDMA_INT_RDA_INT_M) {
4618		if (do_s2io_chk_alarm_bit(RDA_RXDn_ECC_DB_ERR |
4619					  RDA_FRM_ECC_DB_N_AERR |
4620					  RDA_SM1_ERR_ALARM |
4621					  RDA_SM0_ERR_ALARM |
4622					  RDA_RXD_ECC_DB_SERR,
4623					  &bar0->rda_err_reg,
4624					  &sw_stat->rda_err_cnt))
4625			goto reset;
4626		do_s2io_chk_alarm_bit(RDA_RXDn_ECC_SG_ERR |
4627				      RDA_FRM_ECC_SG_ERR |
4628				      RDA_MISC_ERR |
4629				      RDA_PCIX_ERR,
4630				      &bar0->rda_err_reg,
4631				      &sw_stat->rda_err_cnt);
4632	}
4633
4634	if (val64 & RXDMA_INT_RTI_INT_M) {
4635		if (do_s2io_chk_alarm_bit(RTI_SM_ERR_ALARM,
4636					  &bar0->rti_err_reg,
4637					  &sw_stat->rti_err_cnt))
4638			goto reset;
4639		do_s2io_chk_alarm_bit(RTI_ECC_SG_ERR | RTI_ECC_DB_ERR,
4640				      &bar0->rti_err_reg,
4641				      &sw_stat->rti_err_cnt);
4642	}
4643
4644	val64 = readq(&bar0->mac_int_status);
4645	if (val64 & MAC_INT_STATUS_RMAC_INT) {
4646		if (do_s2io_chk_alarm_bit(RMAC_RX_BUFF_OVRN | RMAC_RX_SM_ERR,
4647					  &bar0->mac_rmac_err_reg,
4648					  &sw_stat->mac_rmac_err_cnt))
4649			goto reset;
4650		do_s2io_chk_alarm_bit(RMAC_UNUSED_INT |
4651				      RMAC_SINGLE_ECC_ERR |
4652				      RMAC_DOUBLE_ECC_ERR,
4653				      &bar0->mac_rmac_err_reg,
4654				      &sw_stat->mac_rmac_err_cnt);
4655	}
4656
4657	val64 = readq(&bar0->xgxs_int_status);
4658	if (val64 & XGXS_INT_STATUS_RXGXS) {
4659		if (do_s2io_chk_alarm_bit(RXGXS_ESTORE_OFLOW | RXGXS_RX_SM_ERR,
4660					  &bar0->xgxs_rxgxs_err_reg,
4661					  &sw_stat->xgxs_rxgxs_err_cnt))
4662			goto reset;
4663	}
4664
4665	val64 = readq(&bar0->mc_int_status);
4666	if (val64 & MC_INT_STATUS_MC_INT) {
4667		if (do_s2io_chk_alarm_bit(MC_ERR_REG_SM_ERR,
4668					  &bar0->mc_err_reg,
4669					  &sw_stat->mc_err_cnt))
4670			goto reset;
4671
4672		/* Handling Ecc errors */
4673		if (val64 & (MC_ERR_REG_ECC_ALL_SNG | MC_ERR_REG_ECC_ALL_DBL)) {
4674			writeq(val64, &bar0->mc_err_reg);
4675			if (val64 & MC_ERR_REG_ECC_ALL_DBL) {
4676				sw_stat->double_ecc_errs++;
4677				if (sp->device_type != XFRAME_II_DEVICE) {
4678					/*
4679					 * Reset XframeI only if critical error
4680					 */
4681					if (val64 &
4682					    (MC_ERR_REG_MIRI_ECC_DB_ERR_0 |
4683					     MC_ERR_REG_MIRI_ECC_DB_ERR_1))
4684						goto reset;
4685				}
4686			} else
4687				sw_stat->single_ecc_errs++;
4688		}
4689	}
4690	return;
4691
4692reset:
4693	s2io_stop_all_tx_queue(sp);
4694	schedule_work(&sp->rst_timer_task);
4695	sw_stat->soft_reset_cnt++;
4696}
4697
4698/**
4699 *  s2io_isr - ISR handler of the device .
4700 *  @irq: the irq of the device.
4701 *  @dev_id: a void pointer to the dev structure of the NIC.
4702 *  Description:  This function is the ISR handler of the device. It
4703 *  identifies the reason for the interrupt and calls the relevant
4704 *  service routines. As a contongency measure, this ISR allocates the
4705 *  recv buffers, if their numbers are below the panic value which is
4706 *  presently set to 25% of the original number of rcv buffers allocated.
4707 *  Return value:
4708 *   IRQ_HANDLED: will be returned if IRQ was handled by this routine
4709 *   IRQ_NONE: will be returned if interrupt is not from our device
4710 */
4711static irqreturn_t s2io_isr(int irq, void *dev_id)
4712{
4713	struct net_device *dev = (struct net_device *)dev_id;
4714	struct s2io_nic *sp = netdev_priv(dev);
4715	struct XENA_dev_config __iomem *bar0 = sp->bar0;
4716	int i;
4717	u64 reason = 0;
4718	struct mac_info *mac_control;
4719	struct config_param *config;
4720
4721	/* Pretend we handled any irq's from a disconnected card */
4722	if (pci_channel_offline(sp->pdev))
4723		return IRQ_NONE;
4724
4725	if (!is_s2io_card_up(sp))
4726		return IRQ_NONE;
4727
4728	config = &sp->config;
4729	mac_control = &sp->mac_control;
4730
4731	/*
4732	 * Identify the cause for interrupt and call the appropriate
4733	 * interrupt handler. Causes for the interrupt could be;
4734	 * 1. Rx of packet.
4735	 * 2. Tx complete.
4736	 * 3. Link down.
4737	 */
4738	reason = readq(&bar0->general_int_status);
4739
4740	if (unlikely(reason == S2IO_MINUS_ONE))
4741		return IRQ_HANDLED;	/* Nothing much can be done. Get out */
4742
4743	if (reason &
4744	    (GEN_INTR_RXTRAFFIC | GEN_INTR_TXTRAFFIC | GEN_INTR_TXPIC)) {
4745		writeq(S2IO_MINUS_ONE, &bar0->general_int_mask);
4746
4747		if (config->napi) {
4748			if (reason & GEN_INTR_RXTRAFFIC) {
4749				napi_schedule(&sp->napi);
4750				writeq(S2IO_MINUS_ONE, &bar0->rx_traffic_mask);
4751				writeq(S2IO_MINUS_ONE, &bar0->rx_traffic_int);
4752				readl(&bar0->rx_traffic_int);
4753			}
4754		} else {
4755			/*
4756			 * rx_traffic_int reg is an R1 register, writing all 1's
4757			 * will ensure that the actual interrupt causing bit
4758			 * get's cleared and hence a read can be avoided.
4759			 */
4760			if (reason & GEN_INTR_RXTRAFFIC)
4761				writeq(S2IO_MINUS_ONE, &bar0->rx_traffic_int);
4762
4763			for (i = 0; i < config->rx_ring_num; i++) {
4764				struct ring_info *ring = &mac_control->rings[i];
4765
4766				rx_intr_handler(ring, 0);
4767			}
4768		}
4769
4770		/*
4771		 * tx_traffic_int reg is an R1 register, writing all 1's
4772		 * will ensure that the actual interrupt causing bit get's
4773		 * cleared and hence a read can be avoided.
4774		 */
4775		if (reason & GEN_INTR_TXTRAFFIC)
4776			writeq(S2IO_MINUS_ONE, &bar0->tx_traffic_int);
4777
4778		for (i = 0; i < config->tx_fifo_num; i++)
4779			tx_intr_handler(&mac_control->fifos[i]);
4780
4781		if (reason & GEN_INTR_TXPIC)
4782			s2io_txpic_intr_handle(sp);
4783
4784		/*
4785		 * Reallocate the buffers from the interrupt handler itself.
4786		 */
4787		if (!config->napi) {
4788			for (i = 0; i < config->rx_ring_num; i++) {
4789				struct ring_info *ring = &mac_control->rings[i];
4790
4791				s2io_chk_rx_buffers(sp, ring);
4792			}
4793		}
4794		writeq(sp->general_int_mask, &bar0->general_int_mask);
4795		readl(&bar0->general_int_status);
4796
4797		return IRQ_HANDLED;
4798
4799	} else if (!reason) {
4800		/* The interrupt was not raised by us */
4801		return IRQ_NONE;
4802	}
4803
4804	return IRQ_HANDLED;
4805}
4806
4807/**
4808 * s2io_updt_stats -
4809 */
4810static void s2io_updt_stats(struct s2io_nic *sp)
4811{
4812	struct XENA_dev_config __iomem *bar0 = sp->bar0;
4813	u64 val64;
4814	int cnt = 0;
4815
4816	if (is_s2io_card_up(sp)) {
4817		/* Apprx 30us on a 133 MHz bus */
4818		val64 = SET_UPDT_CLICKS(10) |
4819			STAT_CFG_ONE_SHOT_EN | STAT_CFG_STAT_EN;
4820		writeq(val64, &bar0->stat_cfg);
4821		do {
4822			udelay(100);
4823			val64 = readq(&bar0->stat_cfg);
4824			if (!(val64 & s2BIT(0)))
4825				break;
4826			cnt++;
4827			if (cnt == 5)
4828				break; /* Updt failed */
4829		} while (1);
4830	}
4831}
4832
4833/**
4834 *  s2io_get_stats - Updates the device statistics structure.
4835 *  @dev : pointer to the device structure.
4836 *  Description:
4837 *  This function updates the device statistics structure in the s2io_nic
4838 *  structure and returns a pointer to the same.
4839 *  Return value:
4840 *  pointer to the updated net_device_stats structure.
4841 */
4842static struct net_device_stats *s2io_get_stats(struct net_device *dev)
4843{
4844	struct s2io_nic *sp = netdev_priv(dev);
4845	struct mac_info *mac_control = &sp->mac_control;
4846	struct stat_block *stats = mac_control->stats_info;
4847	u64 delta;
4848
4849	/* Configure Stats for immediate updt */
4850	s2io_updt_stats(sp);
4851
4852	/* A device reset will cause the on-adapter statistics to be zero'ed.
4853	 * This can be done while running by changing the MTU.  To prevent the
4854	 * system from having the stats zero'ed, the driver keeps a copy of the
4855	 * last update to the system (which is also zero'ed on reset).  This
4856	 * enables the driver to accurately know the delta between the last
4857	 * update and the current update.
4858	 */
4859	delta = ((u64) le32_to_cpu(stats->rmac_vld_frms_oflow) << 32 |
4860		le32_to_cpu(stats->rmac_vld_frms)) - sp->stats.rx_packets;
4861	sp->stats.rx_packets += delta;
4862	dev->stats.rx_packets += delta;
4863
4864	delta = ((u64) le32_to_cpu(stats->tmac_frms_oflow) << 32 |
4865		le32_to_cpu(stats->tmac_frms)) - sp->stats.tx_packets;
4866	sp->stats.tx_packets += delta;
4867	dev->stats.tx_packets += delta;
4868
4869	delta = ((u64) le32_to_cpu(stats->rmac_data_octets_oflow) << 32 |
4870		le32_to_cpu(stats->rmac_data_octets)) - sp->stats.rx_bytes;
4871	sp->stats.rx_bytes += delta;
4872	dev->stats.rx_bytes += delta;
4873
4874	delta = ((u64) le32_to_cpu(stats->tmac_data_octets_oflow) << 32 |
4875		le32_to_cpu(stats->tmac_data_octets)) - sp->stats.tx_bytes;
4876	sp->stats.tx_bytes += delta;
4877	dev->stats.tx_bytes += delta;
4878
4879	delta = le64_to_cpu(stats->rmac_drop_frms) - sp->stats.rx_errors;
4880	sp->stats.rx_errors += delta;
4881	dev->stats.rx_errors += delta;
4882
4883	delta = ((u64) le32_to_cpu(stats->tmac_any_err_frms_oflow) << 32 |
4884		le32_to_cpu(stats->tmac_any_err_frms)) - sp->stats.tx_errors;
4885	sp->stats.tx_errors += delta;
4886	dev->stats.tx_errors += delta;
4887
4888	delta = le64_to_cpu(stats->rmac_drop_frms) - sp->stats.rx_dropped;
4889	sp->stats.rx_dropped += delta;
4890	dev->stats.rx_dropped += delta;
4891
4892	delta = le64_to_cpu(stats->tmac_drop_frms) - sp->stats.tx_dropped;
4893	sp->stats.tx_dropped += delta;
4894	dev->stats.tx_dropped += delta;
4895
4896	/* The adapter MAC interprets pause frames as multicast packets, but
4897	 * does not pass them up.  This erroneously increases the multicast
4898	 * packet count and needs to be deducted when the multicast frame count
4899	 * is queried.
4900	 */
4901	delta = (u64) le32_to_cpu(stats->rmac_vld_mcst_frms_oflow) << 32 |
4902		le32_to_cpu(stats->rmac_vld_mcst_frms);
4903	delta -= le64_to_cpu(stats->rmac_pause_ctrl_frms);
4904	delta -= sp->stats.multicast;
4905	sp->stats.multicast += delta;
4906	dev->stats.multicast += delta;
4907
4908	delta = ((u64) le32_to_cpu(stats->rmac_usized_frms_oflow) << 32 |
4909		le32_to_cpu(stats->rmac_usized_frms)) +
4910		le64_to_cpu(stats->rmac_long_frms) - sp->stats.rx_length_errors;
4911	sp->stats.rx_length_errors += delta;
4912	dev->stats.rx_length_errors += delta;
4913
4914	delta = le64_to_cpu(stats->rmac_fcs_err_frms) - sp->stats.rx_crc_errors;
4915	sp->stats.rx_crc_errors += delta;
4916	dev->stats.rx_crc_errors += delta;
4917
4918	return &dev->stats;
4919}
4920
4921/**
4922 *  s2io_set_multicast - entry point for multicast address enable/disable.
4923 *  @dev : pointer to the device structure
4924 *  Description:
4925 *  This function is a driver entry point which gets called by the kernel
4926 *  whenever multicast addresses must be enabled/disabled. This also gets
4927 *  called to set/reset promiscuous mode. Depending on the deivce flag, we
4928 *  determine, if multicast address must be enabled or if promiscuous mode
4929 *  is to be disabled etc.
4930 *  Return value:
4931 *  void.
4932 */
4933
4934static void s2io_set_multicast(struct net_device *dev)
4935{
4936	int i, j, prev_cnt;
4937	struct netdev_hw_addr *ha;
4938	struct s2io_nic *sp = netdev_priv(dev);
4939	struct XENA_dev_config __iomem *bar0 = sp->bar0;
4940	u64 val64 = 0, multi_mac = 0x010203040506ULL, mask =
4941		0xfeffffffffffULL;
4942	u64 dis_addr = S2IO_DISABLE_MAC_ENTRY, mac_addr = 0;
4943	void __iomem *add;
4944	struct config_param *config = &sp->config;
4945
4946	if ((dev->flags & IFF_ALLMULTI) && (!sp->m_cast_flg)) {
4947		/*  Enable all Multicast addresses */
4948		writeq(RMAC_ADDR_DATA0_MEM_ADDR(multi_mac),
4949		       &bar0->rmac_addr_data0_mem);
4950		writeq(RMAC_ADDR_DATA1_MEM_MASK(mask),
4951		       &bar0->rmac_addr_data1_mem);
4952		val64 = RMAC_ADDR_CMD_MEM_WE |
4953			RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4954			RMAC_ADDR_CMD_MEM_OFFSET(config->max_mc_addr - 1);
4955		writeq(val64, &bar0->rmac_addr_cmd_mem);
4956		/* Wait till command completes */
4957		wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4958				      RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING,
4959				      S2IO_BIT_RESET);
4960
4961		sp->m_cast_flg = 1;
4962		sp->all_multi_pos = config->max_mc_addr - 1;
4963	} else if ((dev->flags & IFF_ALLMULTI) && (sp->m_cast_flg)) {
4964		/*  Disable all Multicast addresses */
4965		writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
4966		       &bar0->rmac_addr_data0_mem);
4967		writeq(RMAC_ADDR_DATA1_MEM_MASK(0x0),
4968		       &bar0->rmac_addr_data1_mem);
4969		val64 = RMAC_ADDR_CMD_MEM_WE |
4970			RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4971			RMAC_ADDR_CMD_MEM_OFFSET(sp->all_multi_pos);
4972		writeq(val64, &bar0->rmac_addr_cmd_mem);
4973		/* Wait till command completes */
4974		wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
4975				      RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING,
4976				      S2IO_BIT_RESET);
4977
4978		sp->m_cast_flg = 0;
4979		sp->all_multi_pos = 0;
4980	}
4981
4982	if ((dev->flags & IFF_PROMISC) && (!sp->promisc_flg)) {
4983		/*  Put the NIC into promiscuous mode */
4984		add = &bar0->mac_cfg;
4985		val64 = readq(&bar0->mac_cfg);
4986		val64 |= MAC_CFG_RMAC_PROM_ENABLE;
4987
4988		writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4989		writel((u32)val64, add);
4990		writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4991		writel((u32) (val64 >> 32), (add + 4));
4992
4993		if (vlan_tag_strip != 1) {
4994			val64 = readq(&bar0->rx_pa_cfg);
4995			val64 &= ~RX_PA_CFG_STRIP_VLAN_TAG;
4996			writeq(val64, &bar0->rx_pa_cfg);
4997			sp->vlan_strip_flag = 0;
4998		}
4999
5000		val64 = readq(&bar0->mac_cfg);
5001		sp->promisc_flg = 1;
5002		DBG_PRINT(INFO_DBG, "%s: entered promiscuous mode\n",
5003			  dev->name);
5004	} else if (!(dev->flags & IFF_PROMISC) && (sp->promisc_flg)) {
5005		/*  Remove the NIC from promiscuous mode */
5006		add = &bar0->mac_cfg;
5007		val64 = readq(&bar0->mac_cfg);
5008		val64 &= ~MAC_CFG_RMAC_PROM_ENABLE;
5009
5010		writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
5011		writel((u32)val64, add);
5012		writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
5013		writel((u32) (val64 >> 32), (add + 4));
5014
5015		if (vlan_tag_strip != 0) {
5016			val64 = readq(&bar0->rx_pa_cfg);
5017			val64 |= RX_PA_CFG_STRIP_VLAN_TAG;
5018			writeq(val64, &bar0->rx_pa_cfg);
5019			sp->vlan_strip_flag = 1;
5020		}
5021
5022		val64 = readq(&bar0->mac_cfg);
5023		sp->promisc_flg = 0;
5024		DBG_PRINT(INFO_DBG, "%s: left promiscuous mode\n", dev->name);
5025	}
5026
5027	/*  Update individual M_CAST address list */
5028	if ((!sp->m_cast_flg) && netdev_mc_count(dev)) {
5029		if (netdev_mc_count(dev) >
5030		    (config->max_mc_addr - config->max_mac_addr)) {
5031			DBG_PRINT(ERR_DBG,
5032				  "%s: No more Rx filters can be added - "
5033				  "please enable ALL_MULTI instead\n",
5034				  dev->name);
5035			return;
5036		}
5037
5038		prev_cnt = sp->mc_addr_count;
5039		sp->mc_addr_count = netdev_mc_count(dev);
5040
5041		/* Clear out the previous list of Mc in the H/W. */
5042		for (i = 0; i < prev_cnt; i++) {
5043			writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
5044			       &bar0->rmac_addr_data0_mem);
5045			writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
5046			       &bar0->rmac_addr_data1_mem);
5047			val64 = RMAC_ADDR_CMD_MEM_WE |
5048				RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
5049				RMAC_ADDR_CMD_MEM_OFFSET
5050				(config->mc_start_offset + i);
5051			writeq(val64, &bar0->rmac_addr_cmd_mem);
5052
5053			/* Wait for command completes */
5054			if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
5055						  RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING,
5056						  S2IO_BIT_RESET)) {
5057				DBG_PRINT(ERR_DBG,
5058					  "%s: Adding Multicasts failed\n",
5059					  dev->name);
5060				return;
5061			}
5062		}
5063
5064		/* Create the new Rx filter list and update the same in H/W. */
5065		i = 0;
5066		netdev_for_each_mc_addr(ha, dev) {
5067			mac_addr = 0;
5068			for (j = 0; j < ETH_ALEN; j++) {
5069				mac_addr |= ha->addr[j];
5070				mac_addr <<= 8;
5071			}
5072			mac_addr >>= 8;
5073			writeq(RMAC_ADDR_DATA0_MEM_ADDR(mac_addr),
5074			       &bar0->rmac_addr_data0_mem);
5075			writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
5076			       &bar0->rmac_addr_data1_mem);
5077			val64 = RMAC_ADDR_CMD_MEM_WE |
5078				RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
5079				RMAC_ADDR_CMD_MEM_OFFSET
5080				(i + config->mc_start_offset);
5081			writeq(val64, &bar0->rmac_addr_cmd_mem);
5082
5083			/* Wait for command completes */
5084			if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
5085						  RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING,
5086						  S2IO_BIT_RESET)) {
5087				DBG_PRINT(ERR_DBG,
5088					  "%s: Adding Multicasts failed\n",
5089					  dev->name);
5090				return;
5091			}
5092			i++;
5093		}
5094	}
5095}
5096
5097/* read from CAM unicast & multicast addresses and store it in
5098 * def_mac_addr structure
5099 */
5100static void do_s2io_store_unicast_mc(struct s2io_nic *sp)
5101{
5102	int offset;
5103	u64 mac_addr = 0x0;
5104	struct config_param *config = &sp->config;
5105
5106	/* store unicast & multicast mac addresses */
5107	for (offset = 0; offset < config->max_mc_addr; offset++) {
5108		mac_addr = do_s2io_read_unicast_mc(sp, offset);
5109		/* if read fails disable the entry */
5110		if (mac_addr == FAILURE)
5111			mac_addr = S2IO_DISABLE_MAC_ENTRY;
5112		do_s2io_copy_mac_addr(sp, offset, mac_addr);
5113	}
5114}
5115
5116/* restore unicast & multicast MAC to CAM from def_mac_addr structure */
5117static void do_s2io_restore_unicast_mc(struct s2io_nic *sp)
5118{
5119	int offset;
5120	struct config_param *config = &sp->config;
5121	/* restore unicast mac address */
5122	for (offset = 0; offset < config->max_mac_addr; offset++)
5123		do_s2io_prog_unicast(sp->dev,
5124				     sp->def_mac_addr[offset].mac_addr);
5125
5126	/* restore multicast mac address */
5127	for (offset = config->mc_start_offset;
5128	     offset < config->max_mc_addr; offset++)
5129		do_s2io_add_mc(sp, sp->def_mac_addr[offset].mac_addr);
5130}
5131
5132/* add a multicast MAC address to CAM */
5133static int do_s2io_add_mc(struct s2io_nic *sp, u8 *addr)
5134{
5135	int i;
5136	u64 mac_addr = 0;
5137	struct config_param *config = &sp->config;
5138
5139	for (i = 0; i < ETH_ALEN; i++) {
5140		mac_addr <<= 8;
5141		mac_addr |= addr[i];
5142	}
5143	if ((0ULL == mac_addr) || (mac_addr == S2IO_DISABLE_MAC_ENTRY))
5144		return SUCCESS;
5145
5146	/* check if the multicast mac already preset in CAM */
5147	for (i = config->mc_start_offset; i < config->max_mc_addr; i++) {
5148		u64 tmp64;
5149		tmp64 = do_s2io_read_unicast_mc(sp, i);
5150		if (tmp64 == S2IO_DISABLE_MAC_ENTRY) /* CAM entry is empty */
5151			break;
5152
5153		if (tmp64 == mac_addr)
5154			return SUCCESS;
5155	}
5156	if (i == config->max_mc_addr) {
5157		DBG_PRINT(ERR_DBG,
5158			  "CAM full no space left for multicast MAC\n");
5159		return FAILURE;
5160	}
5161	/* Update the internal structure with this new mac address */
5162	do_s2io_copy_mac_addr(sp, i, mac_addr);
5163
5164	return do_s2io_add_mac(sp, mac_addr, i);
5165}
5166
5167/* add MAC address to CAM */
5168static int do_s2io_add_mac(struct s2io_nic *sp, u64 addr, int off)
5169{
5170	u64 val64;
5171	struct XENA_dev_config __iomem *bar0 = sp->bar0;
5172
5173	writeq(RMAC_ADDR_DATA0_MEM_ADDR(addr),
5174	       &bar0->rmac_addr_data0_mem);
5175
5176	val64 =	RMAC_ADDR_CMD_MEM_WE | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
5177		RMAC_ADDR_CMD_MEM_OFFSET(off);
5178	writeq(val64, &bar0->rmac_addr_cmd_mem);
5179
5180	/* Wait till command completes */
5181	if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
5182				  RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING,
5183				  S2IO_BIT_RESET)) {
5184		DBG_PRINT(INFO_DBG, "do_s2io_add_mac failed\n");
5185		return FAILURE;
5186	}
5187	return SUCCESS;
5188}
5189/* deletes a specified unicast/multicast mac entry from CAM */
5190static int do_s2io_delete_unicast_mc(struct s2io_nic *sp, u64 addr)
5191{
5192	int offset;
5193	u64 dis_addr = S2IO_DISABLE_MAC_ENTRY, tmp64;
5194	struct config_param *config = &sp->config;
5195
5196	for (offset = 1;
5197	     offset < config->max_mc_addr; offset++) {
5198		tmp64 = do_s2io_read_unicast_mc(sp, offset);
5199		if (tmp64 == addr) {
5200			/* disable the entry by writing  0xffffffffffffULL */
5201			if (do_s2io_add_mac(sp, dis_addr, offset) ==  FAILURE)
5202				return FAILURE;
5203			/* store the new mac list from CAM */
5204			do_s2io_store_unicast_mc(sp);
5205			return SUCCESS;
5206		}
5207	}
5208	DBG_PRINT(ERR_DBG, "MAC address 0x%llx not found in CAM\n",
5209		  (unsigned long long)addr);
5210	return FAILURE;
5211}
5212
5213/* read mac entries from CAM */
5214static u64 do_s2io_read_unicast_mc(struct s2io_nic *sp, int offset)
5215{
5216	u64 tmp64 = 0xffffffffffff0000ULL, val64;
5217	struct XENA_dev_config __iomem *bar0 = sp->bar0;
5218
5219	/* read mac addr */
5220	val64 =	RMAC_ADDR_CMD_MEM_RD | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
5221		RMAC_ADDR_CMD_MEM_OFFSET(offset);
5222	writeq(val64, &bar0->rmac_addr_cmd_mem);
5223
5224	/* Wait till command completes */
5225	if (wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
5226				  RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING,
5227				  S2IO_BIT_RESET)) {
5228		DBG_PRINT(INFO_DBG, "do_s2io_read_unicast_mc failed\n");
5229		return FAILURE;
5230	}
5231	tmp64 = readq(&bar0->rmac_addr_data0_mem);
5232
5233	return tmp64 >> 16;
5234}
5235
5236/**
5237 * s2io_set_mac_addr - driver entry point
5238 */
5239
5240static int s2io_set_mac_addr(struct net_device *dev, void *p)
5241{
5242	struct sockaddr *addr = p;
5243
5244	if (!is_valid_ether_addr(addr->sa_data))
5245		return -EADDRNOTAVAIL;
5246
5247	memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
5248
5249	/* store the MAC address in CAM */
5250	return do_s2io_prog_unicast(dev, dev->dev_addr);
5251}
5252/**
5253 *  do_s2io_prog_unicast - Programs the Xframe mac address
5254 *  @dev : pointer to the device structure.
5255 *  @addr: a uchar pointer to the new mac address which is to be set.
5256 *  Description : This procedure will program the Xframe to receive
5257 *  frames with new Mac Address
5258 *  Return value: SUCCESS on success and an appropriate (-)ve integer
5259 *  as defined in errno.h file on failure.
5260 */
5261
5262static int do_s2io_prog_unicast(struct net_device *dev, u8 *addr)
5263{
5264	struct s2io_nic *sp = netdev_priv(dev);
5265	register u64 mac_addr = 0, perm_addr = 0;
5266	int i;
5267	u64 tmp64;
5268	struct config_param *config = &sp->config;
5269
5270	/*
5271	 * Set the new MAC address as the new unicast filter and reflect this
5272	 * change on the device address registered with the OS. It will be
5273	 * at offset 0.
5274	 */
5275	for (i = 0; i < ETH_ALEN; i++) {
5276		mac_addr <<= 8;
5277		mac_addr |= addr[i];
5278		perm_addr <<= 8;
5279		perm_addr |= sp->def_mac_addr[0].mac_addr[i];
5280	}
5281
5282	/* check if the dev_addr is different than perm_addr */
5283	if (mac_addr == perm_addr)
5284		return SUCCESS;
5285
5286	/* check if the mac already preset in CAM */
5287	for (i = 1; i < config->max_mac_addr; i++) {
5288		tmp64 = do_s2io_read_unicast_mc(sp, i);
5289		if (tmp64 == S2IO_DISABLE_MAC_ENTRY) /* CAM entry is empty */
5290			break;
5291
5292		if (tmp64 == mac_addr) {
5293			DBG_PRINT(INFO_DBG,
5294				  "MAC addr:0x%llx already present in CAM\n",
5295				  (unsigned long long)mac_addr);
5296			return SUCCESS;
5297		}
5298	}
5299	if (i == config->max_mac_addr) {
5300		DBG_PRINT(ERR_DBG, "CAM full no space left for Unicast MAC\n");
5301		return FAILURE;
5302	}
5303	/* Update the internal structure with this new mac address */
5304	do_s2io_copy_mac_addr(sp, i, mac_addr);
5305
5306	return do_s2io_add_mac(sp, mac_addr, i);
5307}
5308
5309/**
5310 * s2io_ethtool_sset - Sets different link parameters.
5311 * @sp : private member of the device structure, which is a pointer to the  * s2io_nic structure.
5312 * @info: pointer to the structure with parameters given by ethtool to set
5313 * link information.
5314 * Description:
5315 * The function sets different link parameters provided by the user onto
5316 * the NIC.
5317 * Return value:
5318 * 0 on success.
5319 */
5320
5321static int s2io_ethtool_sset(struct net_device *dev,
5322			     struct ethtool_cmd *info)
5323{
5324	struct s2io_nic *sp = netdev_priv(dev);
5325	if ((info->autoneg == AUTONEG_ENABLE) ||
5326	    (ethtool_cmd_speed(info) != SPEED_10000) ||
5327	    (info->duplex != DUPLEX_FULL))
5328		return -EINVAL;
5329	else {
5330		s2io_close(sp->dev);
5331		s2io_open(sp->dev);
5332	}
5333
5334	return 0;
5335}
5336
5337/**
5338 * s2io_ethtol_gset - Return link specific information.
5339 * @sp : private member of the device structure, pointer to the
5340 *      s2io_nic structure.
5341 * @info : pointer to the structure with parameters given by ethtool
5342 * to return link information.
5343 * Description:
5344 * Returns link specific information like speed, duplex etc.. to ethtool.
5345 * Return value :
5346 * return 0 on success.
5347 */
5348
5349static int s2io_ethtool_gset(struct net_device *dev, struct ethtool_cmd *info)
5350{
5351	struct s2io_nic *sp = netdev_priv(dev);
5352	info->supported = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
5353	info->advertising = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
5354	info->port = PORT_FIBRE;
5355
5356	/* info->transceiver */
5357	info->transceiver = XCVR_EXTERNAL;
5358
5359	if (netif_carrier_ok(sp->dev)) {
5360		ethtool_cmd_speed_set(info, SPEED_10000);
5361		info->duplex = DUPLEX_FULL;
5362	} else {
5363		ethtool_cmd_speed_set(info, SPEED_UNKNOWN);
5364		info->duplex = DUPLEX_UNKNOWN;
5365	}
5366
5367	info->autoneg = AUTONEG_DISABLE;
5368	return 0;
5369}
5370
5371/**
5372 * s2io_ethtool_gdrvinfo - Returns driver specific information.
5373 * @sp : private member of the device structure, which is a pointer to the
5374 * s2io_nic structure.
5375 * @info : pointer to the structure with parameters given by ethtool to
5376 * return driver information.
5377 * Description:
5378 * Returns driver specefic information like name, version etc.. to ethtool.
5379 * Return value:
5380 *  void
5381 */
5382
5383static void s2io_ethtool_gdrvinfo(struct net_device *dev,
5384				  struct ethtool_drvinfo *info)
5385{
5386	struct s2io_nic *sp = netdev_priv(dev);
5387
5388	strlcpy(info->driver, s2io_driver_name, sizeof(info->driver));
5389	strlcpy(info->version, s2io_driver_version, sizeof(info->version));
5390	strlcpy(info->bus_info, pci_name(sp->pdev), sizeof(info->bus_info));
5391	info->regdump_len = XENA_REG_SPACE;
5392	info->eedump_len = XENA_EEPROM_SPACE;
5393}
5394
5395/**
5396 *  s2io_ethtool_gregs - dumps the entire space of Xfame into the buffer.
5397 *  @sp: private member of the device structure, which is a pointer to the
5398 *  s2io_nic structure.
5399 *  @regs : pointer to the structure with parameters given by ethtool for
5400 *  dumping the registers.
5401 *  @reg_space: The input argumnet into which all the registers are dumped.
5402 *  Description:
5403 *  Dumps the entire register space of xFrame NIC into the user given
5404 *  buffer area.
5405 * Return value :
5406 * void .
5407 */
5408
5409static void s2io_ethtool_gregs(struct net_device *dev,
5410			       struct ethtool_regs *regs, void *space)
5411{
5412	int i;
5413	u64 reg;
5414	u8 *reg_space = (u8 *)space;
5415	struct s2io_nic *sp = netdev_priv(dev);
5416
5417	regs->len = XENA_REG_SPACE;
5418	regs->version = sp->pdev->subsystem_device;
5419
5420	for (i = 0; i < regs->len; i += 8) {
5421		reg = readq(sp->bar0 + i);
5422		memcpy((reg_space + i), &reg, 8);
5423	}
5424}
5425
5426/*
5427 *  s2io_set_led - control NIC led
5428 */
5429static void s2io_set_led(struct s2io_nic *sp, bool on)
5430{
5431	struct XENA_dev_config __iomem *bar0 = sp->bar0;
5432	u16 subid = sp->pdev->subsystem_device;
5433	u64 val64;
5434
5435	if ((sp->device_type == XFRAME_II_DEVICE) ||
5436	    ((subid & 0xFF) >= 0x07)) {
5437		val64 = readq(&bar0->gpio_control);
5438		if (on)
5439			val64 |= GPIO_CTRL_GPIO_0;
5440		else
5441			val64 &= ~GPIO_CTRL_GPIO_0;
5442
5443		writeq(val64, &bar0->gpio_control);
5444	} else {
5445		val64 = readq(&bar0->adapter_control);
5446		if (on)
5447			val64 |= ADAPTER_LED_ON;
5448		else
5449			val64 &= ~ADAPTER_LED_ON;
5450
5451		writeq(val64, &bar0->adapter_control);
5452	}
5453
5454}
5455
5456/**
5457 * s2io_ethtool_set_led - To physically identify the nic on the system.
5458 * @dev : network device
5459 * @state: led setting
5460 *
5461 * Description: Used to physically identify the NIC on the system.
5462 * The Link LED will blink for a time specified by the user for
5463 * identification.
5464 * NOTE: The Link has to be Up to be able to blink the LED. Hence
5465 * identification is possible only if it's link is up.
5466 */
5467
5468static int s2io_ethtool_set_led(struct net_device *dev,
5469				enum ethtool_phys_id_state state)
5470{
5471	struct s2io_nic *sp = netdev_priv(dev);
5472	struct XENA_dev_config __iomem *bar0 = sp->bar0;
5473	u16 subid = sp->pdev->subsystem_device;
5474
5475	if ((sp->device_type == XFRAME_I_DEVICE) && ((subid & 0xFF) < 0x07)) {
5476		u64 val64 = readq(&bar0->adapter_control);
5477		if (!(val64 & ADAPTER_CNTL_EN)) {
5478			pr_err("Adapter Link down, cannot blink LED\n");
5479			return -EAGAIN;
5480		}
5481	}
5482
5483	switch (state) {
5484	case ETHTOOL_ID_ACTIVE:
5485		sp->adapt_ctrl_org = readq(&bar0->gpio_control);
5486		return 1;	/* cycle on/off once per second */
5487
5488	case ETHTOOL_ID_ON:
5489		s2io_set_led(sp, true);
5490		break;
5491
5492	case ETHTOOL_ID_OFF:
5493		s2io_set_led(sp, false);
5494		break;
5495
5496	case ETHTOOL_ID_INACTIVE:
5497		if (CARDS_WITH_FAULTY_LINK_INDICATORS(sp->device_type, subid))
5498			writeq(sp->adapt_ctrl_org, &bar0->gpio_control);
5499	}
5500
5501	return 0;
5502}
5503
5504static void s2io_ethtool_gringparam(struct net_device *dev,
5505				    struct ethtool_ringparam *ering)
5506{
5507	struct s2io_nic *sp = netdev_priv(dev);
5508	int i, tx_desc_count = 0, rx_desc_count = 0;
5509
5510	if (sp->rxd_mode == RXD_MODE_1) {
5511		ering->rx_max_pending = MAX_RX_DESC_1;
5512		ering->rx_jumbo_max_pending = MAX_RX_DESC_1;
5513	} else {
5514		ering->rx_max_pending = MAX_RX_DESC_2;
5515		ering->rx_jumbo_max_pending = MAX_RX_DESC_2;
5516	}
5517
5518	ering->tx_max_pending = MAX_TX_DESC;
5519
5520	for (i = 0; i < sp->config.rx_ring_num; i++)
5521		rx_desc_count += sp->config.rx_cfg[i].num_rxd;
5522	ering->rx_pending = rx_desc_count;
5523	ering->rx_jumbo_pending = rx_desc_count;
5524
5525	for (i = 0; i < sp->config.tx_fifo_num; i++)
5526		tx_desc_count += sp->config.tx_cfg[i].fifo_len;
5527	ering->tx_pending = tx_desc_count;
5528	DBG_PRINT(INFO_DBG, "max txds: %d\n", sp->config.max_txds);
5529}
5530
5531/**
5532 * s2io_ethtool_getpause_data -Pause frame frame generation and reception.
5533 * @sp : private member of the device structure, which is a pointer to the
5534 *	s2io_nic structure.
5535 * @ep : pointer to the structure with pause parameters given by ethtool.
5536 * Description:
5537 * Returns the Pause frame generation and reception capability of the NIC.
5538 * Return value:
5539 *  void
5540 */
5541static void s2io_ethtool_getpause_data(struct net_device *dev,
5542				       struct ethtool_pauseparam *ep)
5543{
5544	u64 val64;
5545	struct s2io_nic *sp = netdev_priv(dev);
5546	struct XENA_dev_config __iomem *bar0 = sp->bar0;
5547
5548	val64 = readq(&bar0->rmac_pause_cfg);
5549	if (val64 & RMAC_PAUSE_GEN_ENABLE)
5550		ep->tx_pause = true;
5551	if (val64 & RMAC_PAUSE_RX_ENABLE)
5552		ep->rx_pause = true;
5553	ep->autoneg = false;
5554}
5555
5556/**
5557 * s2io_ethtool_setpause_data -  set/reset pause frame generation.
5558 * @sp : private member of the device structure, which is a pointer to the
5559 *      s2io_nic structure.
5560 * @ep : pointer to the structure with pause parameters given by ethtool.
5561 * Description:
5562 * It can be used to set or reset Pause frame generation or reception
5563 * support of the NIC.
5564 * Return value:
5565 * int, returns 0 on Success
5566 */
5567
5568static int s2io_ethtool_setpause_data(struct net_device *dev,
5569				      struct ethtool_pauseparam *ep)
5570{
5571	u64 val64;
5572	struct s2io_nic *sp = netdev_priv(dev);
5573	struct XENA_dev_config __iomem *bar0 = sp->bar0;
5574
5575	val64 = readq(&bar0->rmac_pause_cfg);
5576	if (ep->tx_pause)
5577		val64 |= RMAC_PAUSE_GEN_ENABLE;
5578	else
5579		val64 &= ~RMAC_PAUSE_GEN_ENABLE;
5580	if (ep->rx_pause)
5581		val64 |= RMAC_PAUSE_RX_ENABLE;
5582	else
5583		val64 &= ~RMAC_PAUSE_RX_ENABLE;
5584	writeq(val64, &bar0->rmac_pause_cfg);
5585	return 0;
5586}
5587
5588/**
5589 * read_eeprom - reads 4 bytes of data from user given offset.
5590 * @sp : private member of the device structure, which is a pointer to the
5591 *      s2io_nic structure.
5592 * @off : offset at which the data must be written
5593 * @data : Its an output parameter where the data read at the given
5594 *	offset is stored.
5595 * Description:
5596 * Will read 4 bytes of data from the user given offset and return the
5597 * read data.
5598 * NOTE: Will allow to read only part of the EEPROM visible through the
5599 *   I2C bus.
5600 * Return value:
5601 *  -1 on failure and 0 on success.
5602 */
5603
5604#define S2IO_DEV_ID		5
5605static int read_eeprom(struct s2io_nic *sp, int off, u64 *data)
5606{
5607	int ret = -1;
5608	u32 exit_cnt = 0;
5609	u64 val64;
5610	struct XENA_dev_config __iomem *bar0 = sp->bar0;
5611
5612	if (sp->device_type == XFRAME_I_DEVICE) {
5613		val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) |
5614			I2C_CONTROL_ADDR(off) |
5615			I2C_CONTROL_BYTE_CNT(0x3) |
5616			I2C_CONTROL_READ |
5617			I2C_CONTROL_CNTL_START;
5618		SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
5619
5620		while (exit_cnt < 5) {
5621			val64 = readq(&bar0->i2c_control);
5622			if (I2C_CONTROL_CNTL_END(val64)) {
5623				*data = I2C_CONTROL_GET_DATA(val64);
5624				ret = 0;
5625				break;
5626			}
5627			msleep(50);
5628			exit_cnt++;
5629		}
5630	}
5631
5632	if (sp->device_type == XFRAME_II_DEVICE) {
5633		val64 = SPI_CONTROL_KEY(0x9) | SPI_CONTROL_SEL1 |
5634			SPI_CONTROL_BYTECNT(0x3) |
5635			SPI_CONTROL_CMD(0x3) | SPI_CONTROL_ADDR(off);
5636		SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
5637		val64 |= SPI_CONTROL_REQ;
5638		SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
5639		while (exit_cnt < 5) {
5640			val64 = readq(&bar0->spi_control);
5641			if (val64 & SPI_CONTROL_NACK) {
5642				ret = 1;
5643				break;
5644			} else if (val64 & SPI_CONTROL_DONE) {
5645				*data = readq(&bar0->spi_data);
5646				*data &= 0xffffff;
5647				ret = 0;
5648				break;
5649			}
5650			msleep(50);
5651			exit_cnt++;
5652		}
5653	}
5654	return ret;
5655}
5656
5657/**
5658 *  write_eeprom - actually writes the relevant part of the data value.
5659 *  @sp : private member of the device structure, which is a pointer to the
5660 *       s2io_nic structure.
5661 *  @off : offset at which the data must be written
5662 *  @data : The data that is to be written
5663 *  @cnt : Number of bytes of the data that are actually to be written into
5664 *  the Eeprom. (max of 3)
5665 * Description:
5666 *  Actually writes the relevant part of the data value into the Eeprom
5667 *  through the I2C bus.
5668 * Return value:
5669 *  0 on success, -1 on failure.
5670 */
5671
5672static int write_eeprom(struct s2io_nic *sp, int off, u64 data, int cnt)
5673{
5674	int exit_cnt = 0, ret = -1;
5675	u64 val64;
5676	struct XENA_dev_config __iomem *bar0 = sp->bar0;
5677
5678	if (sp->device_type == XFRAME_I_DEVICE) {
5679		val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) |
5680			I2C_CONTROL_ADDR(off) |
5681			I2C_CONTROL_BYTE_CNT(cnt) |
5682			I2C_CONTROL_SET_DATA((u32)data) |
5683			I2C_CONTROL_CNTL_START;
5684		SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
5685
5686		while (exit_cnt < 5) {
5687			val64 = readq(&bar0->i2c_control);
5688			if (I2C_CONTROL_CNTL_END(val64)) {
5689				if (!(val64 & I2C_CONTROL_NACK))
5690					ret = 0;
5691				break;
5692			}
5693			msleep(50);
5694			exit_cnt++;
5695		}
5696	}
5697
5698	if (sp->device_type == XFRAME_II_DEVICE) {
5699		int write_cnt = (cnt == 8) ? 0 : cnt;
5700		writeq(SPI_DATA_WRITE(data, (cnt << 3)), &bar0->spi_data);
5701
5702		val64 = SPI_CONTROL_KEY(0x9) | SPI_CONTROL_SEL1 |
5703			SPI_CONTROL_BYTECNT(write_cnt) |
5704			SPI_CONTROL_CMD(0x2) | SPI_CONTROL_ADDR(off);
5705		SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
5706		val64 |= SPI_CONTROL_REQ;
5707		SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
5708		while (exit_cnt < 5) {
5709			val64 = readq(&bar0->spi_control);
5710			if (val64 & SPI_CONTROL_NACK) {
5711				ret = 1;
5712				break;
5713			} else if (val64 & SPI_CONTROL_DONE) {
5714				ret = 0;
5715				break;
5716			}
5717			msleep(50);
5718			exit_cnt++;
5719		}
5720	}
5721	return ret;
5722}
5723static void s2io_vpd_read(struct s2io_nic *nic)
5724{
5725	u8 *vpd_data;
5726	u8 data;
5727	int i = 0, cnt, len, fail = 0;
5728	int vpd_addr = 0x80;
5729	struct swStat *swstats = &nic->mac_control.stats_info->sw_stat;
5730
5731	if (nic->device_type == XFRAME_II_DEVICE) {
5732		strcpy(nic->product_name, "Xframe II 10GbE network adapter");
5733		vpd_addr = 0x80;
5734	} else {
5735		strcpy(nic->product_name, "Xframe I 10GbE network adapter");
5736		vpd_addr = 0x50;
5737	}
5738	strcpy(nic->serial_num, "NOT AVAILABLE");
5739
5740	vpd_data = kmalloc(256, GFP_KERNEL);
5741	if (!vpd_data) {
5742		swstats->mem_alloc_fail_cnt++;
5743		return;
5744	}
5745	swstats->mem_allocated += 256;
5746
5747	for (i = 0; i < 256; i += 4) {
5748		pci_write_config_byte(nic->pdev, (vpd_addr + 2), i);
5749		pci_read_config_byte(nic->pdev,  (vpd_addr + 2), &data);
5750		pci_write_config_byte(nic->pdev, (vpd_addr + 3), 0);
5751		for (cnt = 0; cnt < 5; cnt++) {
5752			msleep(2);
5753			pci_read_config_byte(nic->pdev, (vpd_addr + 3), &data);
5754			if (data == 0x80)
5755				break;
5756		}
5757		if (cnt >= 5) {
5758			DBG_PRINT(ERR_DBG, "Read of VPD data failed\n");
5759			fail = 1;
5760			break;
5761		}
5762		pci_read_config_dword(nic->pdev,  (vpd_addr + 4),
5763				      (u32 *)&vpd_data[i]);
5764	}
5765
5766	if (!fail) {
5767		/* read serial number of adapter */
5768		for (cnt = 0; cnt < 252; cnt++) {
5769			if ((vpd_data[cnt] == 'S') &&
5770			    (vpd_data[cnt+1] == 'N')) {
5771				len = vpd_data[cnt+2];
5772				if (len < min(VPD_STRING_LEN, 256-cnt-2)) {
5773					memcpy(nic->serial_num,
5774					       &vpd_data[cnt + 3],
5775					       len);
5776					memset(nic->serial_num+len,
5777					       0,
5778					       VPD_STRING_LEN-len);
5779					break;
5780				}
5781			}
5782		}
5783	}
5784
5785	if ((!fail) && (vpd_data[1] < VPD_STRING_LEN)) {
5786		len = vpd_data[1];
5787		memcpy(nic->product_name, &vpd_data[3], len);
5788		nic->product_name[len] = 0;
5789	}
5790	kfree(vpd_data);
5791	swstats->mem_freed += 256;
5792}
5793
5794/**
5795 *  s2io_ethtool_geeprom  - reads the value stored in the Eeprom.
5796 *  @sp : private member of the device structure, which is a pointer to the *       s2io_nic structure.
5797 *  @eeprom : pointer to the user level structure provided by ethtool,
5798 *  containing all relevant information.
5799 *  @data_buf : user defined value to be written into Eeprom.
5800 *  Description: Reads the values stored in the Eeprom at given offset
5801 *  for a given length. Stores these values int the input argument data
5802 *  buffer 'data_buf' and returns these to the caller (ethtool.)
5803 *  Return value:
5804 *  int  0 on success
5805 */
5806
5807static int s2io_ethtool_geeprom(struct net_device *dev,
5808				struct ethtool_eeprom *eeprom, u8 * data_buf)
5809{
5810	u32 i, valid;
5811	u64 data;
5812	struct s2io_nic *sp = netdev_priv(dev);
5813
5814	eeprom->magic = sp->pdev->vendor | (sp->pdev->device << 16);
5815
5816	if ((eeprom->offset + eeprom->len) > (XENA_EEPROM_SPACE))
5817		eeprom->len = XENA_EEPROM_SPACE - eeprom->offset;
5818
5819	for (i = 0; i < eeprom->len; i += 4) {
5820		if (read_eeprom(sp, (eeprom->offset + i), &data)) {
5821			DBG_PRINT(ERR_DBG, "Read of EEPROM failed\n");
5822			return -EFAULT;
5823		}
5824		valid = INV(data);
5825		memcpy((data_buf + i), &valid, 4);
5826	}
5827	return 0;
5828}
5829
5830/**
5831 *  s2io_ethtool_seeprom - tries to write the user provided value in Eeprom
5832 *  @sp : private member of the device structure, which is a pointer to the
5833 *  s2io_nic structure.
5834 *  @eeprom : pointer to the user level structure provided by ethtool,
5835 *  containing all relevant information.
5836 *  @data_buf ; user defined value to be written into Eeprom.
5837 *  Description:
5838 *  Tries to write the user provided value in the Eeprom, at the offset
5839 *  given by the user.
5840 *  Return value:
5841 *  0 on success, -EFAULT on failure.
5842 */
5843
5844static int s2io_ethtool_seeprom(struct net_device *dev,
5845				struct ethtool_eeprom *eeprom,
5846				u8 *data_buf)
5847{
5848	int len = eeprom->len, cnt = 0;
5849	u64 valid = 0, data;
5850	struct s2io_nic *sp = netdev_priv(dev);
5851
5852	if (eeprom->magic != (sp->pdev->vendor | (sp->pdev->device << 16))) {
5853		DBG_PRINT(ERR_DBG,
5854			  "ETHTOOL_WRITE_EEPROM Err: "
5855			  "Magic value is wrong, it is 0x%x should be 0x%x\n",
5856			  (sp->pdev->vendor | (sp->pdev->device << 16)),
5857			  eeprom->magic);
5858		return -EFAULT;
5859	}
5860
5861	while (len) {
5862		data = (u32)data_buf[cnt] & 0x000000FF;
5863		if (data)
5864			valid = (u32)(data << 24);
5865		else
5866			valid = data;
5867
5868		if (write_eeprom(sp, (eeprom->offset + cnt), valid, 0)) {
5869			DBG_PRINT(ERR_DBG,
5870				  "ETHTOOL_WRITE_EEPROM Err: "
5871				  "Cannot write into the specified offset\n");
5872			return -EFAULT;
5873		}
5874		cnt++;
5875		len--;
5876	}
5877
5878	return 0;
5879}
5880
5881/**
5882 * s2io_register_test - reads and writes into all clock domains.
5883 * @sp : private member of the device structure, which is a pointer to the
5884 * s2io_nic structure.
5885 * @data : variable that returns the result of each of the test conducted b
5886 * by the driver.
5887 * Description:
5888 * Read and write into all clock domains. The NIC has 3 clock domains,
5889 * see that registers in all the three regions are accessible.
5890 * Return value:
5891 * 0 on success.
5892 */
5893
5894static int s2io_register_test(struct s2io_nic *sp, uint64_t *data)
5895{
5896	struct XENA_dev_config __iomem *bar0 = sp->bar0;
5897	u64 val64 = 0, exp_val;
5898	int fail = 0;
5899
5900	val64 = readq(&bar0->pif_rd_swapper_fb);
5901	if (val64 != 0x123456789abcdefULL) {
5902		fail = 1;
5903		DBG_PRINT(INFO_DBG, "Read Test level %d fails\n", 1);
5904	}
5905
5906	val64 = readq(&bar0->rmac_pause_cfg);
5907	if (val64 != 0xc000ffff00000000ULL) {
5908		fail = 1;
5909		DBG_PRINT(INFO_DBG, "Read Test level %d fails\n", 2);
5910	}
5911
5912	val64 = readq(&bar0->rx_queue_cfg);
5913	if (sp->device_type == XFRAME_II_DEVICE)
5914		exp_val = 0x0404040404040404ULL;
5915	else
5916		exp_val = 0x0808080808080808ULL;
5917	if (val64 != exp_val) {
5918		fail = 1;
5919		DBG_PRINT(INFO_DBG, "Read Test level %d fails\n", 3);
5920	}
5921
5922	val64 = readq(&bar0->xgxs_efifo_cfg);
5923	if (val64 != 0x000000001923141EULL) {
5924		fail = 1;
5925		DBG_PRINT(INFO_DBG, "Read Test level %d fails\n", 4);
5926	}
5927
5928	val64 = 0x5A5A5A5A5A5A5A5AULL;
5929	writeq(val64, &bar0->xmsi_data);
5930	val64 = readq(&bar0->xmsi_data);
5931	if (val64 != 0x5A5A5A5A5A5A5A5AULL) {
5932		fail = 1;
5933		DBG_PRINT(ERR_DBG, "Write Test level %d fails\n", 1);
5934	}
5935
5936	val64 = 0xA5A5A5A5A5A5A5A5ULL;
5937	writeq(val64, &bar0->xmsi_data);
5938	val64 = readq(&bar0->xmsi_data);
5939	if (val64 != 0xA5A5A5A5A5A5A5A5ULL) {
5940		fail = 1;
5941		DBG_PRINT(ERR_DBG, "Write Test level %d fails\n", 2);
5942	}
5943
5944	*data = fail;
5945	return fail;
5946}
5947
5948/**
5949 * s2io_eeprom_test - to verify that EEprom in the xena can be programmed.
5950 * @sp : private member of the device structure, which is a pointer to the
5951 * s2io_nic structure.
5952 * @data:variable that returns the result of each of the test conducted by
5953 * the driver.
5954 * Description:
5955 * Verify that EEPROM in the xena can be programmed using I2C_CONTROL
5956 * register.
5957 * Return value:
5958 * 0 on success.
5959 */
5960
5961static int s2io_eeprom_test(struct s2io_nic *sp, uint64_t *data)
5962{
5963	int fail = 0;
5964	u64 ret_data, org_4F0, org_7F0;
5965	u8 saved_4F0 = 0, saved_7F0 = 0;
5966	struct net_device *dev = sp->dev;
5967
5968	/* Test Write Error at offset 0 */
5969	/* Note that SPI interface allows write access to all areas
5970	 * of EEPROM. Hence doing all negative testing only for Xframe I.
5971	 */
5972	if (sp->device_type == XFRAME_I_DEVICE)
5973		if (!write_eeprom(sp, 0, 0, 3))
5974			fail = 1;
5975
5976	/* Save current values at offsets 0x4F0 and 0x7F0 */
5977	if (!read_eeprom(sp, 0x4F0, &org_4F0))
5978		saved_4F0 = 1;
5979	if (!read_eeprom(sp, 0x7F0, &org_7F0))
5980		saved_7F0 = 1;
5981
5982	/* Test Write at offset 4f0 */
5983	if (write_eeprom(sp, 0x4F0, 0x012345, 3))
5984		fail = 1;
5985	if (read_eeprom(sp, 0x4F0, &ret_data))
5986		fail = 1;
5987
5988	if (ret_data != 0x012345) {
5989		DBG_PRINT(ERR_DBG, "%s: eeprom test error at offset 0x4F0. "
5990			  "Data written %llx Data read %llx\n",
5991			  dev->name, (unsigned long long)0x12345,
5992			  (unsigned long long)ret_data);
5993		fail = 1;
5994	}
5995
5996	/* Reset the EEPROM data go FFFF */
5997	write_eeprom(sp, 0x4F0, 0xFFFFFF, 3);
5998
5999	/* Test Write Request Error at offset 0x7c */
6000	if (sp->device_type == XFRAME_I_DEVICE)
6001		if (!write_eeprom(sp, 0x07C, 0, 3))
6002			fail = 1;
6003
6004	/* Test Write Request at offset 0x7f0 */
6005	if (write_eeprom(sp, 0x7F0, 0x012345, 3))
6006		fail = 1;
6007	if (read_eeprom(sp, 0x7F0, &ret_data))
6008		fail = 1;
6009
6010	if (ret_data != 0x012345) {
6011		DBG_PRINT(ERR_DBG, "%s: eeprom test error at offset 0x7F0. "
6012			  "Data written %llx Data read %llx\n",
6013			  dev->name, (unsigned long long)0x12345,
6014			  (unsigned long long)ret_data);
6015		fail = 1;
6016	}
6017
6018	/* Reset the EEPROM data go FFFF */
6019	write_eeprom(sp, 0x7F0, 0xFFFFFF, 3);
6020
6021	if (sp->device_type == XFRAME_I_DEVICE) {
6022		/* Test Write Error at offset 0x80 */
6023		if (!write_eeprom(sp, 0x080, 0, 3))
6024			fail = 1;
6025
6026		/* Test Write Error at offset 0xfc */
6027		if (!write_eeprom(sp, 0x0FC, 0, 3))
6028			fail = 1;
6029
6030		/* Test Write Error at offset 0x100 */
6031		if (!write_eeprom(sp, 0x100, 0, 3))
6032			fail = 1;
6033
6034		/* Test Write Error at offset 4ec */
6035		if (!write_eeprom(sp, 0x4EC, 0, 3))
6036			fail = 1;
6037	}
6038
6039	/* Restore values at offsets 0x4F0 and 0x7F0 */
6040	if (saved_4F0)
6041		write_eeprom(sp, 0x4F0, org_4F0, 3);
6042	if (saved_7F0)
6043		write_eeprom(sp, 0x7F0, org_7F0, 3);
6044
6045	*data = fail;
6046	return fail;
6047}
6048
6049/**
6050 * s2io_bist_test - invokes the MemBist test of the card .
6051 * @sp : private member of the device structure, which is a pointer to the
6052 * s2io_nic structure.
6053 * @data:variable that returns the result of each of the test conducted by
6054 * the driver.
6055 * Description:
6056 * This invokes the MemBist test of the card. We give around
6057 * 2 secs time for the Test to complete. If it's still not complete
6058 * within this peiod, we consider that the test failed.
6059 * Return value:
6060 * 0 on success and -1 on failure.
6061 */
6062
6063static int s2io_bist_test(struct s2io_nic *sp, uint64_t *data)
6064{
6065	u8 bist = 0;
6066	int cnt = 0, ret = -1;
6067
6068	pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
6069	bist |= PCI_BIST_START;
6070	pci_write_config_word(sp->pdev, PCI_BIST, bist);
6071
6072	while (cnt < 20) {
6073		pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
6074		if (!(bist & PCI_BIST_START)) {
6075			*data = (bist & PCI_BIST_CODE_MASK);
6076			ret = 0;
6077			break;
6078		}
6079		msleep(100);
6080		cnt++;
6081	}
6082
6083	return ret;
6084}
6085
6086/**
6087 * s2io_link_test - verifies the link state of the nic
6088 * @sp ; private member of the device structure, which is a pointer to the
6089 * s2io_nic structure.
6090 * @data: variable that returns the result of each of the test conducted by
6091 * the driver.
6092 * Description:
6093 * The function verifies the link state of the NIC and updates the input
6094 * argument 'data' appropriately.
6095 * Return value:
6096 * 0 on success.
6097 */
6098
6099static int s2io_link_test(struct s2io_nic *sp, uint64_t *data)
6100{
6101	struct XENA_dev_config __iomem *bar0 = sp->bar0;
6102	u64 val64;
6103
6104	val64 = readq(&bar0->adapter_status);
6105	if (!(LINK_IS_UP(val64)))
6106		*data = 1;
6107	else
6108		*data = 0;
6109
6110	return *data;
6111}
6112
6113/**
6114 * s2io_rldram_test - offline test for access to the RldRam chip on the NIC
6115 * @sp: private member of the device structure, which is a pointer to the
6116 * s2io_nic structure.
6117 * @data: variable that returns the result of each of the test
6118 * conducted by the driver.
6119 * Description:
6120 *  This is one of the offline test that tests the read and write
6121 *  access to the RldRam chip on the NIC.
6122 * Return value:
6123 *  0 on success.
6124 */
6125
6126static int s2io_rldram_test(struct s2io_nic *sp, uint64_t *data)
6127{
6128	struct XENA_dev_config __iomem *bar0 = sp->bar0;
6129	u64 val64;
6130	int cnt, iteration = 0, test_fail = 0;
6131
6132	val64 = readq(&bar0->adapter_control);
6133	val64 &= ~ADAPTER_ECC_EN;
6134	writeq(val64, &bar0->adapter_control);
6135
6136	val64 = readq(&bar0->mc_rldram_test_ctrl);
6137	val64 |= MC_RLDRAM_TEST_MODE;
6138	SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
6139
6140	val64 = readq(&bar0->mc_rldram_mrs);
6141	val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE;
6142	SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
6143
6144	val64 |= MC_RLDRAM_MRS_ENABLE;
6145	SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
6146
6147	while (iteration < 2) {
6148		val64 = 0x55555555aaaa0000ULL;
6149		if (iteration == 1)
6150			val64 ^= 0xFFFFFFFFFFFF0000ULL;
6151		writeq(val64, &bar0->mc_rldram_test_d0);
6152
6153		val64 = 0xaaaa5a5555550000ULL;
6154		if (iteration == 1)
6155			val64 ^= 0xFFFFFFFFFFFF0000ULL;
6156		writeq(val64, &bar0->mc_rldram_test_d1);
6157
6158		val64 = 0x55aaaaaaaa5a0000ULL;
6159		if (iteration == 1)
6160			val64 ^= 0xFFFFFFFFFFFF0000ULL;
6161		writeq(val64, &bar0->mc_rldram_test_d2);
6162
6163		val64 = (u64) (0x0000003ffffe0100ULL);
6164		writeq(val64, &bar0->mc_rldram_test_add);
6165
6166		val64 = MC_RLDRAM_TEST_MODE |
6167			MC_RLDRAM_TEST_WRITE |
6168			MC_RLDRAM_TEST_GO;
6169		SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
6170
6171		for (cnt = 0; cnt < 5; cnt++) {
6172			val64 = readq(&bar0->mc_rldram_test_ctrl);
6173			if (val64 & MC_RLDRAM_TEST_DONE)
6174				break;
6175			msleep(200);
6176		}
6177
6178		if (cnt == 5)
6179			break;
6180
6181		val64 = MC_RLDRAM_TEST_MODE | MC_RLDRAM_TEST_GO;
6182		SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
6183
6184		for (cnt = 0; cnt < 5; cnt++) {
6185			val64 = readq(&bar0->mc_rldram_test_ctrl);
6186			if (val64 & MC_RLDRAM_TEST_DONE)
6187				break;
6188			msleep(500);
6189		}
6190
6191		if (cnt == 5)
6192			break;
6193
6194		val64 = readq(&bar0->mc_rldram_test_ctrl);
6195		if (!(val64 & MC_RLDRAM_TEST_PASS))
6196			test_fail = 1;
6197
6198		iteration++;
6199	}
6200
6201	*data = test_fail;
6202
6203	/* Bring the adapter out of test mode */
6204	SPECIAL_REG_WRITE(0, &bar0->mc_rldram_test_ctrl, LF);
6205
6206	return test_fail;
6207}
6208
6209/**
6210 *  s2io_ethtool_test - conducts 6 tsets to determine the health of card.
6211 *  @sp : private member of the device structure, which is a pointer to the
6212 *  s2io_nic structure.
6213 *  @ethtest : pointer to a ethtool command specific structure that will be
6214 *  returned to the user.
6215 *  @data : variable that returns the result of each of the test
6216 * conducted by the driver.
6217 * Description:
6218 *  This function conducts 6 tests ( 4 offline and 2 online) to determine
6219 *  the health of the card.
6220 * Return value:
6221 *  void
6222 */
6223
6224static void s2io_ethtool_test(struct net_device *dev,
6225			      struct ethtool_test *ethtest,
6226			      uint64_t *data)
6227{
6228	struct s2io_nic *sp = netdev_priv(dev);
6229	int orig_state = netif_running(sp->dev);
6230
6231	if (ethtest->flags == ETH_TEST_FL_OFFLINE) {
6232		/* Offline Tests. */
6233		if (orig_state)
6234			s2io_close(sp->dev);
6235
6236		if (s2io_register_test(sp, &data[0]))
6237			ethtest->flags |= ETH_TEST_FL_FAILED;
6238
6239		s2io_reset(sp);
6240
6241		if (s2io_rldram_test(sp, &data[3]))
6242			ethtest->flags |= ETH_TEST_FL_FAILED;
6243
6244		s2io_reset(sp);
6245
6246		if (s2io_eeprom_test(sp, &data[1]))
6247			ethtest->flags |= ETH_TEST_FL_FAILED;
6248
6249		if (s2io_bist_test(sp, &data[4]))
6250			ethtest->flags |= ETH_TEST_FL_FAILED;
6251
6252		if (orig_state)
6253			s2io_open(sp->dev);
6254
6255		data[2] = 0;
6256	} else {
6257		/* Online Tests. */
6258		if (!orig_state) {
6259			DBG_PRINT(ERR_DBG, "%s: is not up, cannot run test\n",
6260				  dev->name);
6261			data[0] = -1;
6262			data[1] = -1;
6263			data[2] = -1;
6264			data[3] = -1;
6265			data[4] = -1;
6266		}
6267
6268		if (s2io_link_test(sp, &data[2]))
6269			ethtest->flags |= ETH_TEST_FL_FAILED;
6270
6271		data[0] = 0;
6272		data[1] = 0;
6273		data[3] = 0;
6274		data[4] = 0;
6275	}
6276}
6277
6278static void s2io_get_ethtool_stats(struct net_device *dev,
6279				   struct ethtool_stats *estats,
6280				   u64 *tmp_stats)
6281{
6282	int i = 0, k;
6283	struct s2io_nic *sp = netdev_priv(dev);
6284	struct stat_block *stats = sp->mac_control.stats_info;
6285	struct swStat *swstats = &stats->sw_stat;
6286	struct xpakStat *xstats = &stats->xpak_stat;
6287
6288	s2io_updt_stats(sp);
6289	tmp_stats[i++] =
6290		(u64)le32_to_cpu(stats->tmac_frms_oflow) << 32  |
6291		le32_to_cpu(stats->tmac_frms);
6292	tmp_stats[i++] =
6293		(u64)le32_to_cpu(stats->tmac_data_octets_oflow) << 32 |
6294		le32_to_cpu(stats->tmac_data_octets);
6295	tmp_stats[i++] = le64_to_cpu(stats->tmac_drop_frms);
6296	tmp_stats[i++] =
6297		(u64)le32_to_cpu(stats->tmac_mcst_frms_oflow) << 32 |
6298		le32_to_cpu(stats->tmac_mcst_frms);
6299	tmp_stats[i++] =
6300		(u64)le32_to_cpu(stats->tmac_bcst_frms_oflow) << 32 |
6301		le32_to_cpu(stats->tmac_bcst_frms);
6302	tmp_stats[i++] = le64_to_cpu(stats->tmac_pause_ctrl_frms);
6303	tmp_stats[i++] =
6304		(u64)le32_to_cpu(stats->tmac_ttl_octets_oflow) << 32 |
6305		le32_to_cpu(stats->tmac_ttl_octets);
6306	tmp_stats[i++] =
6307		(u64)le32_to_cpu(stats->tmac_ucst_frms_oflow) << 32 |
6308		le32_to_cpu(stats->tmac_ucst_frms);
6309	tmp_stats[i++] =
6310		(u64)le32_to_cpu(stats->tmac_nucst_frms_oflow) << 32 |
6311		le32_to_cpu(stats->tmac_nucst_frms);
6312	tmp_stats[i++] =
6313		(u64)le32_to_cpu(stats->tmac_any_err_frms_oflow) << 32 |
6314		le32_to_cpu(stats->tmac_any_err_frms);
6315	tmp_stats[i++] = le64_to_cpu(stats->tmac_ttl_less_fb_octets);
6316	tmp_stats[i++] = le64_to_cpu(stats->tmac_vld_ip_octets);
6317	tmp_stats[i++] =
6318		(u64)le32_to_cpu(stats->tmac_vld_ip_oflow) << 32 |
6319		le32_to_cpu(stats->tmac_vld_ip);
6320	tmp_stats[i++] =
6321		(u64)le32_to_cpu(stats->tmac_drop_ip_oflow) << 32 |
6322		le32_to_cpu(stats->tmac_drop_ip);
6323	tmp_stats[i++] =
6324		(u64)le32_to_cpu(stats->tmac_icmp_oflow) << 32 |
6325		le32_to_cpu(stats->tmac_icmp);
6326	tmp_stats[i++] =
6327		(u64)le32_to_cpu(stats->tmac_rst_tcp_oflow) << 32 |
6328		le32_to_cpu(stats->tmac_rst_tcp);
6329	tmp_stats[i++] = le64_to_cpu(stats->tmac_tcp);
6330	tmp_stats[i++] = (u64)le32_to_cpu(stats->tmac_udp_oflow) << 32 |
6331		le32_to_cpu(stats->tmac_udp);
6332	tmp_stats[i++] =
6333		(u64)le32_to_cpu(stats->rmac_vld_frms_oflow) << 32 |
6334		le32_to_cpu(stats->rmac_vld_frms);
6335	tmp_stats[i++] =
6336		(u64)le32_to_cpu(stats->rmac_data_octets_oflow) << 32 |
6337		le32_to_cpu(stats->rmac_data_octets);
6338	tmp_stats[i++] = le64_to_cpu(stats->rmac_fcs_err_frms);
6339	tmp_stats[i++] = le64_to_cpu(stats->rmac_drop_frms);
6340	tmp_stats[i++] =
6341		(u64)le32_to_cpu(stats->rmac_vld_mcst_frms_oflow) << 32 |
6342		le32_to_cpu(stats->rmac_vld_mcst_frms);
6343	tmp_stats[i++] =
6344		(u64)le32_to_cpu(stats->rmac_vld_bcst_frms_oflow) << 32 |
6345		le32_to_cpu(stats->rmac_vld_bcst_frms);
6346	tmp_stats[i++] = le32_to_cpu(stats->rmac_in_rng_len_err_frms);
6347	tmp_stats[i++] = le32_to_cpu(stats->rmac_out_rng_len_err_frms);
6348	tmp_stats[i++] = le64_to_cpu(stats->rmac_long_frms);
6349	tmp_stats[i++] = le64_to_cpu(stats->rmac_pause_ctrl_frms);
6350	tmp_stats[i++] = le64_to_cpu(stats->rmac_unsup_ctrl_frms);
6351	tmp_stats[i++] =
6352		(u64)le32_to_cpu(stats->rmac_ttl_octets_oflow) << 32 |
6353		le32_to_cpu(stats->rmac_ttl_octets);
6354	tmp_stats[i++] =
6355		(u64)le32_to_cpu(stats->rmac_accepted_ucst_frms_oflow) << 32
6356		| le32_to_cpu(stats->rmac_accepted_ucst_frms);
6357	tmp_stats[i++] =
6358		(u64)le32_to_cpu(stats->rmac_accepted_nucst_frms_oflow)
6359		<< 32 | le32_to_cpu(stats->rmac_accepted_nucst_frms);
6360	tmp_stats[i++] =
6361		(u64)le32_to_cpu(stats->rmac_discarded_frms_oflow) << 32 |
6362		le32_to_cpu(stats->rmac_discarded_frms);
6363	tmp_stats[i++] =
6364		(u64)le32_to_cpu(stats->rmac_drop_events_oflow)
6365		<< 32 | le32_to_cpu(stats->rmac_drop_events);
6366	tmp_stats[i++] = le64_to_cpu(stats->rmac_ttl_less_fb_octets);
6367	tmp_stats[i++] = le64_to_cpu(stats->rmac_ttl_frms);
6368	tmp_stats[i++] =
6369		(u64)le32_to_cpu(stats->rmac_usized_frms_oflow) << 32 |
6370		le32_to_cpu(stats->rmac_usized_frms);
6371	tmp_stats[i++] =
6372		(u64)le32_to_cpu(stats->rmac_osized_frms_oflow) << 32 |
6373		le32_to_cpu(stats->rmac_osized_frms);
6374	tmp_stats[i++] =
6375		(u64)le32_to_cpu(stats->rmac_frag_frms_oflow) << 32 |
6376		le32_to_cpu(stats->rmac_frag_frms);
6377	tmp_stats[i++] =
6378		(u64)le32_to_cpu(stats->rmac_jabber_frms_oflow) << 32 |
6379		le32_to_cpu(stats->rmac_jabber_frms);
6380	tmp_stats[i++] = le64_to_cpu(stats->rmac_ttl_64_frms);
6381	tmp_stats[i++] = le64_to_cpu(stats->rmac_ttl_65_127_frms);
6382	tmp_stats[i++] = le64_to_cpu(stats->rmac_ttl_128_255_frms);
6383	tmp_stats[i++] = le64_to_cpu(stats->rmac_ttl_256_511_frms);
6384	tmp_stats[i++] = le64_to_cpu(stats->rmac_ttl_512_1023_frms);
6385	tmp_stats[i++] = le64_to_cpu(stats->rmac_ttl_1024_1518_frms);
6386	tmp_stats[i++] =
6387		(u64)le32_to_cpu(stats->rmac_ip_oflow) << 32 |
6388		le32_to_cpu(stats->rmac_ip);
6389	tmp_stats[i++] = le64_to_cpu(stats->rmac_ip_octets);
6390	tmp_stats[i++] = le32_to_cpu(stats->rmac_hdr_err_ip);
6391	tmp_stats[i++] =
6392		(u64)le32_to_cpu(stats->rmac_drop_ip_oflow) << 32 |
6393		le32_to_cpu(stats->rmac_drop_ip);
6394	tmp_stats[i++] =
6395		(u64)le32_to_cpu(stats->rmac_icmp_oflow) << 32 |
6396		le32_to_cpu(stats->rmac_icmp);
6397	tmp_stats[i++] = le64_to_cpu(stats->rmac_tcp);
6398	tmp_stats[i++] =
6399		(u64)le32_to_cpu(stats->rmac_udp_oflow) << 32 |
6400		le32_to_cpu(stats->rmac_udp);
6401	tmp_stats[i++] =
6402		(u64)le32_to_cpu(stats->rmac_err_drp_udp_oflow) << 32 |
6403		le32_to_cpu(stats->rmac_err_drp_udp);
6404	tmp_stats[i++] = le64_to_cpu(stats->rmac_xgmii_err_sym);
6405	tmp_stats[i++] = le64_to_cpu(stats->rmac_frms_q0);
6406	tmp_stats[i++] = le64_to_cpu(stats->rmac_frms_q1);
6407	tmp_stats[i++] = le64_to_cpu(stats->rmac_frms_q2);
6408	tmp_stats[i++] = le64_to_cpu(stats->rmac_frms_q3);
6409	tmp_stats[i++] = le64_to_cpu(stats->rmac_frms_q4);
6410	tmp_stats[i++] = le64_to_cpu(stats->rmac_frms_q5);
6411	tmp_stats[i++] = le64_to_cpu(stats->rmac_frms_q6);
6412	tmp_stats[i++] = le64_to_cpu(stats->rmac_frms_q7);
6413	tmp_stats[i++] = le16_to_cpu(stats->rmac_full_q0);
6414	tmp_stats[i++] = le16_to_cpu(stats->rmac_full_q1);
6415	tmp_stats[i++] = le16_to_cpu(stats->rmac_full_q2);
6416	tmp_stats[i++] = le16_to_cpu(stats->rmac_full_q3);
6417	tmp_stats[i++] = le16_to_cpu(stats->rmac_full_q4);
6418	tmp_stats[i++] = le16_to_cpu(stats->rmac_full_q5);
6419	tmp_stats[i++] = le16_to_cpu(stats->rmac_full_q6);
6420	tmp_stats[i++] = le16_to_cpu(stats->rmac_full_q7);
6421	tmp_stats[i++] =
6422		(u64)le32_to_cpu(stats->rmac_pause_cnt_oflow) << 32 |
6423		le32_to_cpu(stats->rmac_pause_cnt);
6424	tmp_stats[i++] = le64_to_cpu(stats->rmac_xgmii_data_err_cnt);
6425	tmp_stats[i++] = le64_to_cpu(stats->rmac_xgmii_ctrl_err_cnt);
6426	tmp_stats[i++] =
6427		(u64)le32_to_cpu(stats->rmac_accepted_ip_oflow) << 32 |
6428		le32_to_cpu(stats->rmac_accepted_ip);
6429	tmp_stats[i++] = le32_to_cpu(stats->rmac_err_tcp);
6430	tmp_stats[i++] = le32_to_cpu(stats->rd_req_cnt);
6431	tmp_stats[i++] = le32_to_cpu(stats->new_rd_req_cnt);
6432	tmp_stats[i++] = le32_to_cpu(stats->new_rd_req_rtry_cnt);
6433	tmp_stats[i++] = le32_to_cpu(stats->rd_rtry_cnt);
6434	tmp_stats[i++] = le32_to_cpu(stats->wr_rtry_rd_ack_cnt);
6435	tmp_stats[i++] = le32_to_cpu(stats->wr_req_cnt);
6436	tmp_stats[i++] = le32_to_cpu(stats->new_wr_req_cnt);
6437	tmp_stats[i++] = le32_to_cpu(stats->new_wr_req_rtry_cnt);
6438	tmp_stats[i++] = le32_to_cpu(stats->wr_rtry_cnt);
6439	tmp_stats[i++] = le32_to_cpu(stats->wr_disc_cnt);
6440	tmp_stats[i++] = le32_to_cpu(stats->rd_rtry_wr_ack_cnt);
6441	tmp_stats[i++] = le32_to_cpu(stats->txp_wr_cnt);
6442	tmp_stats[i++] = le32_to_cpu(stats->txd_rd_cnt);
6443	tmp_stats[i++] = le32_to_cpu(stats->txd_wr_cnt);
6444	tmp_stats[i++] = le32_to_cpu(stats->rxd_rd_cnt);
6445	tmp_stats[i++] = le32_to_cpu(stats->rxd_wr_cnt);
6446	tmp_stats[i++] = le32_to_cpu(stats->txf_rd_cnt);
6447	tmp_stats[i++] = le32_to_cpu(stats->rxf_wr_cnt);
6448
6449	/* Enhanced statistics exist only for Hercules */
6450	if (sp->device_type == XFRAME_II_DEVICE) {
6451		tmp_stats[i++] =
6452			le64_to_cpu(stats->rmac_ttl_1519_4095_frms);
6453		tmp_stats[i++] =
6454			le64_to_cpu(stats->rmac_ttl_4096_8191_frms);
6455		tmp_stats[i++] =
6456			le64_to_cpu(stats->rmac_ttl_8192_max_frms);
6457		tmp_stats[i++] = le64_to_cpu(stats->rmac_ttl_gt_max_frms);
6458		tmp_stats[i++] = le64_to_cpu(stats->rmac_osized_alt_frms);
6459		tmp_stats[i++] = le64_to_cpu(stats->rmac_jabber_alt_frms);
6460		tmp_stats[i++] = le64_to_cpu(stats->rmac_gt_max_alt_frms);
6461		tmp_stats[i++] = le64_to_cpu(stats->rmac_vlan_frms);
6462		tmp_stats[i++] = le32_to_cpu(stats->rmac_len_discard);
6463		tmp_stats[i++] = le32_to_cpu(stats->rmac_fcs_discard);
6464		tmp_stats[i++] = le32_to_cpu(stats->rmac_pf_discard);
6465		tmp_stats[i++] = le32_to_cpu(stats->rmac_da_discard);
6466		tmp_stats[i++] = le32_to_cpu(stats->rmac_red_discard);
6467		tmp_stats[i++] = le32_to_cpu(stats->rmac_rts_discard);
6468		tmp_stats[i++] = le32_to_cpu(stats->rmac_ingm_full_discard);
6469		tmp_stats[i++] = le32_to_cpu(stats->link_fault_cnt);
6470	}
6471
6472	tmp_stats[i++] = 0;
6473	tmp_stats[i++] = swstats->single_ecc_errs;
6474	tmp_stats[i++] = swstats->double_ecc_errs;
6475	tmp_stats[i++] = swstats->parity_err_cnt;
6476	tmp_stats[i++] = swstats->serious_err_cnt;
6477	tmp_stats[i++] = swstats->soft_reset_cnt;
6478	tmp_stats[i++] = swstats->fifo_full_cnt;
6479	for (k = 0; k < MAX_RX_RINGS; k++)
6480		tmp_stats[i++] = swstats->ring_full_cnt[k];
6481	tmp_stats[i++] = xstats->alarm_transceiver_temp_high;
6482	tmp_stats[i++] = xstats->alarm_transceiver_temp_low;
6483	tmp_stats[i++] = xstats->alarm_laser_bias_current_high;
6484	tmp_stats[i++] = xstats->alarm_laser_bias_current_low;
6485	tmp_stats[i++] = xstats->alarm_laser_output_power_high;
6486	tmp_stats[i++] = xstats->alarm_laser_output_power_low;
6487	tmp_stats[i++] = xstats->warn_transceiver_temp_high;
6488	tmp_stats[i++] = xstats->warn_transceiver_temp_low;
6489	tmp_stats[i++] = xstats->warn_laser_bias_current_high;
6490	tmp_stats[i++] = xstats->warn_laser_bias_current_low;
6491	tmp_stats[i++] = xstats->warn_laser_output_power_high;
6492	tmp_stats[i++] = xstats->warn_laser_output_power_low;
6493	tmp_stats[i++] = swstats->clubbed_frms_cnt;
6494	tmp_stats[i++] = swstats->sending_both;
6495	tmp_stats[i++] = swstats->outof_sequence_pkts;
6496	tmp_stats[i++] = swstats->flush_max_pkts;
6497	if (swstats->num_aggregations) {
6498		u64 tmp = swstats->sum_avg_pkts_aggregated;
6499		int count = 0;
6500		/*
6501		 * Since 64-bit divide does not work on all platforms,
6502		 * do repeated subtraction.
6503		 */
6504		while (tmp >= swstats->num_aggregations) {
6505			tmp -= swstats->num_aggregations;
6506			count++;
6507		}
6508		tmp_stats[i++] = count;
6509	} else
6510		tmp_stats[i++] = 0;
6511	tmp_stats[i++] = swstats->mem_alloc_fail_cnt;
6512	tmp_stats[i++] = swstats->pci_map_fail_cnt;
6513	tmp_stats[i++] = swstats->watchdog_timer_cnt;
6514	tmp_stats[i++] = swstats->mem_allocated;
6515	tmp_stats[i++] = swstats->mem_freed;
6516	tmp_stats[i++] = swstats->link_up_cnt;
6517	tmp_stats[i++] = swstats->link_down_cnt;
6518	tmp_stats[i++] = swstats->link_up_time;
6519	tmp_stats[i++] = swstats->link_down_time;
6520
6521	tmp_stats[i++] = swstats->tx_buf_abort_cnt;
6522	tmp_stats[i++] = swstats->tx_desc_abort_cnt;
6523	tmp_stats[i++] = swstats->tx_parity_err_cnt;
6524	tmp_stats[i++] = swstats->tx_link_loss_cnt;
6525	tmp_stats[i++] = swstats->tx_list_proc_err_cnt;
6526
6527	tmp_stats[i++] = swstats->rx_parity_err_cnt;
6528	tmp_stats[i++] = swstats->rx_abort_cnt;
6529	tmp_stats[i++] = swstats->rx_parity_abort_cnt;
6530	tmp_stats[i++] = swstats->rx_rda_fail_cnt;
6531	tmp_stats[i++] = swstats->rx_unkn_prot_cnt;
6532	tmp_stats[i++] = swstats->rx_fcs_err_cnt;
6533	tmp_stats[i++] = swstats->rx_buf_size_err_cnt;
6534	tmp_stats[i++] = swstats->rx_rxd_corrupt_cnt;
6535	tmp_stats[i++] = swstats->rx_unkn_err_cnt;
6536	tmp_stats[i++] = swstats->tda_err_cnt;
6537	tmp_stats[i++] = swstats->pfc_err_cnt;
6538	tmp_stats[i++] = swstats->pcc_err_cnt;
6539	tmp_stats[i++] = swstats->tti_err_cnt;
6540	tmp_stats[i++] = swstats->tpa_err_cnt;
6541	tmp_stats[i++] = swstats->sm_err_cnt;
6542	tmp_stats[i++] = swstats->lso_err_cnt;
6543	tmp_stats[i++] = swstats->mac_tmac_err_cnt;
6544	tmp_stats[i++] = swstats->mac_rmac_err_cnt;
6545	tmp_stats[i++] = swstats->xgxs_txgxs_err_cnt;
6546	tmp_stats[i++] = swstats->xgxs_rxgxs_err_cnt;
6547	tmp_stats[i++] = swstats->rc_err_cnt;
6548	tmp_stats[i++] = swstats->prc_pcix_err_cnt;
6549	tmp_stats[i++] = swstats->rpa_err_cnt;
6550	tmp_stats[i++] = swstats->rda_err_cnt;
6551	tmp_stats[i++] = swstats->rti_err_cnt;
6552	tmp_stats[i++] = swstats->mc_err_cnt;
6553}
6554
6555static int s2io_ethtool_get_regs_len(struct net_device *dev)
6556{
6557	return XENA_REG_SPACE;
6558}
6559
6560
6561static int s2io_get_eeprom_len(struct net_device *dev)
6562{
6563	return XENA_EEPROM_SPACE;
6564}
6565
6566static int s2io_get_sset_count(struct net_device *dev, int sset)
6567{
6568	struct s2io_nic *sp = netdev_priv(dev);
6569
6570	switch (sset) {
6571	case ETH_SS_TEST:
6572		return S2IO_TEST_LEN;
6573	case ETH_SS_STATS:
6574		switch (sp->device_type) {
6575		case XFRAME_I_DEVICE:
6576			return XFRAME_I_STAT_LEN;
6577		case XFRAME_II_DEVICE:
6578			return XFRAME_II_STAT_LEN;
6579		default:
6580			return 0;
6581		}
6582	default:
6583		return -EOPNOTSUPP;
6584	}
6585}
6586
6587static void s2io_ethtool_get_strings(struct net_device *dev,
6588				     u32 stringset, u8 *data)
6589{
6590	int stat_size = 0;
6591	struct s2io_nic *sp = netdev_priv(dev);
6592
6593	switch (stringset) {
6594	case ETH_SS_TEST:
6595		memcpy(data, s2io_gstrings, S2IO_STRINGS_LEN);
6596		break;
6597	case ETH_SS_STATS:
6598		stat_size = sizeof(ethtool_xena_stats_keys);
6599		memcpy(data, &ethtool_xena_stats_keys, stat_size);
6600		if (sp->device_type == XFRAME_II_DEVICE) {
6601			memcpy(data + stat_size,
6602			       &ethtool_enhanced_stats_keys,
6603			       sizeof(ethtool_enhanced_stats_keys));
6604			stat_size += sizeof(ethtool_enhanced_stats_keys);
6605		}
6606
6607		memcpy(data + stat_size, &ethtool_driver_stats_keys,
6608		       sizeof(ethtool_driver_stats_keys));
6609	}
6610}
6611
6612static int s2io_set_features(struct net_device *dev, netdev_features_t features)
6613{
6614	struct s2io_nic *sp = netdev_priv(dev);
6615	netdev_features_t changed = (features ^ dev->features) & NETIF_F_LRO;
6616
6617	if (changed && netif_running(dev)) {
6618		int rc;
6619
6620		s2io_stop_all_tx_queue(sp);
6621		s2io_card_down(sp);
6622		dev->features = features;
6623		rc = s2io_card_up(sp);
6624		if (rc)
6625			s2io_reset(sp);
6626		else
6627			s2io_start_all_tx_queue(sp);
6628
6629		return rc ? rc : 1;
6630	}
6631
6632	return 0;
6633}
6634
6635static const struct ethtool_ops netdev_ethtool_ops = {
6636	.get_settings = s2io_ethtool_gset,
6637	.set_settings = s2io_ethtool_sset,
6638	.get_drvinfo = s2io_ethtool_gdrvinfo,
6639	.get_regs_len = s2io_ethtool_get_regs_len,
6640	.get_regs = s2io_ethtool_gregs,
6641	.get_link = ethtool_op_get_link,
6642	.get_eeprom_len = s2io_get_eeprom_len,
6643	.get_eeprom = s2io_ethtool_geeprom,
6644	.set_eeprom = s2io_ethtool_seeprom,
6645	.get_ringparam = s2io_ethtool_gringparam,
6646	.get_pauseparam = s2io_ethtool_getpause_data,
6647	.set_pauseparam = s2io_ethtool_setpause_data,
6648	.self_test = s2io_ethtool_test,
6649	.get_strings = s2io_ethtool_get_strings,
6650	.set_phys_id = s2io_ethtool_set_led,
6651	.get_ethtool_stats = s2io_get_ethtool_stats,
6652	.get_sset_count = s2io_get_sset_count,
6653};
6654
6655/**
6656 *  s2io_ioctl - Entry point for the Ioctl
6657 *  @dev :  Device pointer.
6658 *  @ifr :  An IOCTL specefic structure, that can contain a pointer to
6659 *  a proprietary structure used to pass information to the driver.
6660 *  @cmd :  This is used to distinguish between the different commands that
6661 *  can be passed to the IOCTL functions.
6662 *  Description:
6663 *  Currently there are no special functionality supported in IOCTL, hence
6664 *  function always return EOPNOTSUPPORTED
6665 */
6666
6667static int s2io_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
6668{
6669	return -EOPNOTSUPP;
6670}
6671
6672/**
6673 *  s2io_change_mtu - entry point to change MTU size for the device.
6674 *   @dev : device pointer.
6675 *   @new_mtu : the new MTU size for the device.
6676 *   Description: A driver entry point to change MTU size for the device.
6677 *   Before changing the MTU the device must be stopped.
6678 *  Return value:
6679 *   0 on success and an appropriate (-)ve integer as defined in errno.h
6680 *   file on failure.
6681 */
6682
6683static int s2io_change_mtu(struct net_device *dev, int new_mtu)
6684{
6685	struct s2io_nic *sp = netdev_priv(dev);
6686	int ret = 0;
6687
6688	if ((new_mtu < MIN_MTU) || (new_mtu > S2IO_JUMBO_SIZE)) {
6689		DBG_PRINT(ERR_DBG, "%s: MTU size is invalid.\n", dev->name);
6690		return -EPERM;
6691	}
6692
6693	dev->mtu = new_mtu;
6694	if (netif_running(dev)) {
6695		s2io_stop_all_tx_queue(sp);
6696		s2io_card_down(sp);
6697		ret = s2io_card_up(sp);
6698		if (ret) {
6699			DBG_PRINT(ERR_DBG, "%s: Device bring up failed\n",
6700				  __func__);
6701			return ret;
6702		}
6703		s2io_wake_all_tx_queue(sp);
6704	} else { /* Device is down */
6705		struct XENA_dev_config __iomem *bar0 = sp->bar0;
6706		u64 val64 = new_mtu;
6707
6708		writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
6709	}
6710
6711	return ret;
6712}
6713
6714/**
6715 * s2io_set_link - Set the LInk status
6716 * @data: long pointer to device private structue
6717 * Description: Sets the link status for the adapter
6718 */
6719
6720static void s2io_set_link(struct work_struct *work)
6721{
6722	struct s2io_nic *nic = container_of(work, struct s2io_nic,
6723					    set_link_task);
6724	struct net_device *dev = nic->dev;
6725	struct XENA_dev_config __iomem *bar0 = nic->bar0;
6726	register u64 val64;
6727	u16 subid;
6728
6729	rtnl_lock();
6730
6731	if (!netif_running(dev))
6732		goto out_unlock;
6733
6734	if (test_and_set_bit(__S2IO_STATE_LINK_TASK, &(nic->state))) {
6735		/* The card is being reset, no point doing anything */
6736		goto out_unlock;
6737	}
6738
6739	subid = nic->pdev->subsystem_device;
6740	if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
6741		/*
6742		 * Allow a small delay for the NICs self initiated
6743		 * cleanup to complete.
6744		 */
6745		msleep(100);
6746	}
6747
6748	val64 = readq(&bar0->adapter_status);
6749	if (LINK_IS_UP(val64)) {
6750		if (!(readq(&bar0->adapter_control) & ADAPTER_CNTL_EN)) {
6751			if (verify_xena_quiescence(nic)) {
6752				val64 = readq(&bar0->adapter_control);
6753				val64 |= ADAPTER_CNTL_EN;
6754				writeq(val64, &bar0->adapter_control);
6755				if (CARDS_WITH_FAULTY_LINK_INDICATORS(
6756					    nic->device_type, subid)) {
6757					val64 = readq(&bar0->gpio_control);
6758					val64 |= GPIO_CTRL_GPIO_0;
6759					writeq(val64, &bar0->gpio_control);
6760					val64 = readq(&bar0->gpio_control);
6761				} else {
6762					val64 |= ADAPTER_LED_ON;
6763					writeq(val64, &bar0->adapter_control);
6764				}
6765				nic->device_enabled_once = true;
6766			} else {
6767				DBG_PRINT(ERR_DBG,
6768					  "%s: Error: device is not Quiescent\n",
6769					  dev->name);
6770				s2io_stop_all_tx_queue(nic);
6771			}
6772		}
6773		val64 = readq(&bar0->adapter_control);
6774		val64 |= ADAPTER_LED_ON;
6775		writeq(val64, &bar0->adapter_control);
6776		s2io_link(nic, LINK_UP);
6777	} else {
6778		if (CARDS_WITH_FAULTY_LINK_INDICATORS(nic->device_type,
6779						      subid)) {
6780			val64 = readq(&bar0->gpio_control);
6781			val64 &= ~GPIO_CTRL_GPIO_0;
6782			writeq(val64, &bar0->gpio_control);
6783			val64 = readq(&bar0->gpio_control);
6784		}
6785		/* turn off LED */
6786		val64 = readq(&bar0->adapter_control);
6787		val64 = val64 & (~ADAPTER_LED_ON);
6788		writeq(val64, &bar0->adapter_control);
6789		s2io_link(nic, LINK_DOWN);
6790	}
6791	clear_bit(__S2IO_STATE_LINK_TASK, &(nic->state));
6792
6793out_unlock:
6794	rtnl_unlock();
6795}
6796
6797static int set_rxd_buffer_pointer(struct s2io_nic *sp, struct RxD_t *rxdp,
6798				  struct buffAdd *ba,
6799				  struct sk_buff **skb, u64 *temp0, u64 *temp1,
6800				  u64 *temp2, int size)
6801{
6802	struct net_device *dev = sp->dev;
6803	struct swStat *stats = &sp->mac_control.stats_info->sw_stat;
6804
6805	if ((sp->rxd_mode == RXD_MODE_1) && (rxdp->Host_Control == 0)) {
6806		struct RxD1 *rxdp1 = (struct RxD1 *)rxdp;
6807		/* allocate skb */
6808		if (*skb) {
6809			DBG_PRINT(INFO_DBG, "SKB is not NULL\n");
6810			/*
6811			 * As Rx frame are not going to be processed,
6812			 * using same mapped address for the Rxd
6813			 * buffer pointer
6814			 */
6815			rxdp1->Buffer0_ptr = *temp0;
6816		} else {
6817			*skb = netdev_alloc_skb(dev, size);
6818			if (!(*skb)) {
6819				DBG_PRINT(INFO_DBG,
6820					  "%s: Out of memory to allocate %s\n",
6821					  dev->name, "1 buf mode SKBs");
6822				stats->mem_alloc_fail_cnt++;
6823				return -ENOMEM ;
6824			}
6825			stats->mem_allocated += (*skb)->truesize;
6826			/* storing the mapped addr in a temp variable
6827			 * such it will be used for next rxd whose
6828			 * Host Control is NULL
6829			 */
6830			rxdp1->Buffer0_ptr = *temp0 =
6831				pci_map_single(sp->pdev, (*skb)->data,
6832					       size - NET_IP_ALIGN,
6833					       PCI_DMA_FROMDEVICE);
6834			if (pci_dma_mapping_error(sp->pdev, rxdp1->Buffer0_ptr))
6835				goto memalloc_failed;
6836			rxdp->Host_Control = (unsigned long) (*skb);
6837		}
6838	} else if ((sp->rxd_mode == RXD_MODE_3B) && (rxdp->Host_Control == 0)) {
6839		struct RxD3 *rxdp3 = (struct RxD3 *)rxdp;
6840		/* Two buffer Mode */
6841		if (*skb) {
6842			rxdp3->Buffer2_ptr = *temp2;
6843			rxdp3->Buffer0_ptr = *temp0;
6844			rxdp3->Buffer1_ptr = *temp1;
6845		} else {
6846			*skb = netdev_alloc_skb(dev, size);
6847			if (!(*skb)) {
6848				DBG_PRINT(INFO_DBG,
6849					  "%s: Out of memory to allocate %s\n",
6850					  dev->name,
6851					  "2 buf mode SKBs");
6852				stats->mem_alloc_fail_cnt++;
6853				return -ENOMEM;
6854			}
6855			stats->mem_allocated += (*skb)->truesize;
6856			rxdp3->Buffer2_ptr = *temp2 =
6857				pci_map_single(sp->pdev, (*skb)->data,
6858					       dev->mtu + 4,
6859					       PCI_DMA_FROMDEVICE);
6860			if (pci_dma_mapping_error(sp->pdev, rxdp3->Buffer2_ptr))
6861				goto memalloc_failed;
6862			rxdp3->Buffer0_ptr = *temp0 =
6863				pci_map_single(sp->pdev, ba->ba_0, BUF0_LEN,
6864					       PCI_DMA_FROMDEVICE);
6865			if (pci_dma_mapping_error(sp->pdev,
6866						  rxdp3->Buffer0_ptr)) {
6867				pci_unmap_single(sp->pdev,
6868						 (dma_addr_t)rxdp3->Buffer2_ptr,
6869						 dev->mtu + 4,
6870						 PCI_DMA_FROMDEVICE);
6871				goto memalloc_failed;
6872			}
6873			rxdp->Host_Control = (unsigned long) (*skb);
6874
6875			/* Buffer-1 will be dummy buffer not used */
6876			rxdp3->Buffer1_ptr = *temp1 =
6877				pci_map_single(sp->pdev, ba->ba_1, BUF1_LEN,
6878					       PCI_DMA_FROMDEVICE);
6879			if (pci_dma_mapping_error(sp->pdev,
6880						  rxdp3->Buffer1_ptr)) {
6881				pci_unmap_single(sp->pdev,
6882						 (dma_addr_t)rxdp3->Buffer0_ptr,
6883						 BUF0_LEN, PCI_DMA_FROMDEVICE);
6884				pci_unmap_single(sp->pdev,
6885						 (dma_addr_t)rxdp3->Buffer2_ptr,
6886						 dev->mtu + 4,
6887						 PCI_DMA_FROMDEVICE);
6888				goto memalloc_failed;
6889			}
6890		}
6891	}
6892	return 0;
6893
6894memalloc_failed:
6895	stats->pci_map_fail_cnt++;
6896	stats->mem_freed += (*skb)->truesize;
6897	dev_kfree_skb(*skb);
6898	return -ENOMEM;
6899}
6900
6901static void set_rxd_buffer_size(struct s2io_nic *sp, struct RxD_t *rxdp,
6902				int size)
6903{
6904	struct net_device *dev = sp->dev;
6905	if (sp->rxd_mode == RXD_MODE_1) {
6906		rxdp->Control_2 = SET_BUFFER0_SIZE_1(size - NET_IP_ALIGN);
6907	} else if (sp->rxd_mode == RXD_MODE_3B) {
6908		rxdp->Control_2 = SET_BUFFER0_SIZE_3(BUF0_LEN);
6909		rxdp->Control_2 |= SET_BUFFER1_SIZE_3(1);
6910		rxdp->Control_2 |= SET_BUFFER2_SIZE_3(dev->mtu + 4);
6911	}
6912}
6913
6914static  int rxd_owner_bit_reset(struct s2io_nic *sp)
6915{
6916	int i, j, k, blk_cnt = 0, size;
6917	struct config_param *config = &sp->config;
6918	struct mac_info *mac_control = &sp->mac_control;
6919	struct net_device *dev = sp->dev;
6920	struct RxD_t *rxdp = NULL;
6921	struct sk_buff *skb = NULL;
6922	struct buffAdd *ba = NULL;
6923	u64 temp0_64 = 0, temp1_64 = 0, temp2_64 = 0;
6924
6925	/* Calculate the size based on ring mode */
6926	size = dev->mtu + HEADER_ETHERNET_II_802_3_SIZE +
6927		HEADER_802_2_SIZE + HEADER_SNAP_SIZE;
6928	if (sp->rxd_mode == RXD_MODE_1)
6929		size += NET_IP_ALIGN;
6930	else if (sp->rxd_mode == RXD_MODE_3B)
6931		size = dev->mtu + ALIGN_SIZE + BUF0_LEN + 4;
6932
6933	for (i = 0; i < config->rx_ring_num; i++) {
6934		struct rx_ring_config *rx_cfg = &config->rx_cfg[i];
6935		struct ring_info *ring = &mac_control->rings[i];
6936
6937		blk_cnt = rx_cfg->num_rxd / (rxd_count[sp->rxd_mode] + 1);
6938
6939		for (j = 0; j < blk_cnt; j++) {
6940			for (k = 0; k < rxd_count[sp->rxd_mode]; k++) {
6941				rxdp = ring->rx_blocks[j].rxds[k].virt_addr;
6942				if (sp->rxd_mode == RXD_MODE_3B)
6943					ba = &ring->ba[j][k];
6944				if (set_rxd_buffer_pointer(sp, rxdp, ba, &skb,
6945							   &temp0_64,
6946							   &temp1_64,
6947							   &temp2_64,
6948							   size) == -ENOMEM) {
6949					return 0;
6950				}
6951
6952				set_rxd_buffer_size(sp, rxdp, size);
6953				wmb();
6954				/* flip the Ownership bit to Hardware */
6955				rxdp->Control_1 |= RXD_OWN_XENA;
6956			}
6957		}
6958	}
6959	return 0;
6960
6961}
6962
6963static int s2io_add_isr(struct s2io_nic *sp)
6964{
6965	int ret = 0;
6966	struct net_device *dev = sp->dev;
6967	int err = 0;
6968
6969	if (sp->config.intr_type == MSI_X)
6970		ret = s2io_enable_msi_x(sp);
6971	if (ret) {
6972		DBG_PRINT(ERR_DBG, "%s: Defaulting to INTA\n", dev->name);
6973		sp->config.intr_type = INTA;
6974	}
6975
6976	/*
6977	 * Store the values of the MSIX table in
6978	 * the struct s2io_nic structure
6979	 */
6980	store_xmsi_data(sp);
6981
6982	/* After proper initialization of H/W, register ISR */
6983	if (sp->config.intr_type == MSI_X) {
6984		int i, msix_rx_cnt = 0;
6985
6986		for (i = 0; i < sp->num_entries; i++) {
6987			if (sp->s2io_entries[i].in_use == MSIX_FLG) {
6988				if (sp->s2io_entries[i].type ==
6989				    MSIX_RING_TYPE) {
6990					sprintf(sp->desc[i], "%s:MSI-X-%d-RX",
6991						dev->name, i);
6992					err = request_irq(sp->entries[i].vector,
6993							  s2io_msix_ring_handle,
6994							  0,
6995							  sp->desc[i],
6996							  sp->s2io_entries[i].arg);
6997				} else if (sp->s2io_entries[i].type ==
6998					   MSIX_ALARM_TYPE) {
6999					sprintf(sp->desc[i], "%s:MSI-X-%d-TX",
7000						dev->name, i);
7001					err = request_irq(sp->entries[i].vector,
7002							  s2io_msix_fifo_handle,
7003							  0,
7004							  sp->desc[i],
7005							  sp->s2io_entries[i].arg);
7006
7007				}
7008				/* if either data or addr is zero print it. */
7009				if (!(sp->msix_info[i].addr &&
7010				      sp->msix_info[i].data)) {
7011					DBG_PRINT(ERR_DBG,
7012						  "%s @Addr:0x%llx Data:0x%llx\n",
7013						  sp->desc[i],
7014						  (unsigned long long)
7015						  sp->msix_info[i].addr,
7016						  (unsigned long long)
7017						  ntohl(sp->msix_info[i].data));
7018				} else
7019					msix_rx_cnt++;
7020				if (err) {
7021					remove_msix_isr(sp);
7022
7023					DBG_PRINT(ERR_DBG,
7024						  "%s:MSI-X-%d registration "
7025						  "failed\n", dev->name, i);
7026
7027					DBG_PRINT(ERR_DBG,
7028						  "%s: Defaulting to INTA\n",
7029						  dev->name);
7030					sp->config.intr_type = INTA;
7031					break;
7032				}
7033				sp->s2io_entries[i].in_use =
7034					MSIX_REGISTERED_SUCCESS;
7035			}
7036		}
7037		if (!err) {
7038			pr_info("MSI-X-RX %d entries enabled\n", --msix_rx_cnt);
7039			DBG_PRINT(INFO_DBG,
7040				  "MSI-X-TX entries enabled through alarm vector\n");
7041		}
7042	}
7043	if (sp->config.intr_type == INTA) {
7044		err = request_irq(sp->pdev->irq, s2io_isr, IRQF_SHARED,
7045				  sp->name, dev);
7046		if (err) {
7047			DBG_PRINT(ERR_DBG, "%s: ISR registration failed\n",
7048				  dev->name);
7049			return -1;
7050		}
7051	}
7052	return 0;
7053}
7054
7055static void s2io_rem_isr(struct s2io_nic *sp)
7056{
7057	if (sp->config.intr_type == MSI_X)
7058		remove_msix_isr(sp);
7059	else
7060		remove_inta_isr(sp);
7061}
7062
7063static void do_s2io_card_down(struct s2io_nic *sp, int do_io)
7064{
7065	int cnt = 0;
7066	struct XENA_dev_config __iomem *bar0 = sp->bar0;
7067	register u64 val64 = 0;
7068	struct config_param *config;
7069	config = &sp->config;
7070
7071	if (!is_s2io_card_up(sp))
7072		return;
7073
7074	del_timer_sync(&sp->alarm_timer);
7075	/* If s2io_set_link task is executing, wait till it completes. */
7076	while (test_and_set_bit(__S2IO_STATE_LINK_TASK, &(sp->state)))
7077		msleep(50);
7078	clear_bit(__S2IO_STATE_CARD_UP, &sp->state);
7079
7080	/* Disable napi */
7081	if (sp->config.napi) {
7082		int off = 0;
7083		if (config->intr_type ==  MSI_X) {
7084			for (; off < sp->config.rx_ring_num; off++)
7085				napi_disable(&sp->mac_control.rings[off].napi);
7086		}
7087		else
7088			napi_disable(&sp->napi);
7089	}
7090
7091	/* disable Tx and Rx traffic on the NIC */
7092	if (do_io)
7093		stop_nic(sp);
7094
7095	s2io_rem_isr(sp);
7096
7097	/* stop the tx queue, indicate link down */
7098	s2io_link(sp, LINK_DOWN);
7099
7100	/* Check if the device is Quiescent and then Reset the NIC */
7101	while (do_io) {
7102		/* As per the HW requirement we need to replenish the
7103		 * receive buffer to avoid the ring bump. Since there is
7104		 * no intention of processing the Rx frame at this pointwe are
7105		 * just setting the ownership bit of rxd in Each Rx
7106		 * ring to HW and set the appropriate buffer size
7107		 * based on the ring mode
7108		 */
7109		rxd_owner_bit_reset(sp);
7110
7111		val64 = readq(&bar0->adapter_status);
7112		if (verify_xena_quiescence(sp)) {
7113			if (verify_pcc_quiescent(sp, sp->device_enabled_once))
7114				break;
7115		}
7116
7117		msleep(50);
7118		cnt++;
7119		if (cnt == 10) {
7120			DBG_PRINT(ERR_DBG, "Device not Quiescent - "
7121				  "adapter status reads 0x%llx\n",
7122				  (unsigned long long)val64);
7123			break;
7124		}
7125	}
7126	if (do_io)
7127		s2io_reset(sp);
7128
7129	/* Free all Tx buffers */
7130	free_tx_buffers(sp);
7131
7132	/* Free all Rx buffers */
7133	free_rx_buffers(sp);
7134
7135	clear_bit(__S2IO_STATE_LINK_TASK, &(sp->state));
7136}
7137
7138static void s2io_card_down(struct s2io_nic *sp)
7139{
7140	do_s2io_card_down(sp, 1);
7141}
7142
7143static int s2io_card_up(struct s2io_nic *sp)
7144{
7145	int i, ret = 0;
7146	struct config_param *config;
7147	struct mac_info *mac_control;
7148	struct net_device *dev = sp->dev;
7149	u16 interruptible;
7150
7151	/* Initialize the H/W I/O registers */
7152	ret = init_nic(sp);
7153	if (ret != 0) {
7154		DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
7155			  dev->name);
7156		if (ret != -EIO)
7157			s2io_reset(sp);
7158		return ret;
7159	}
7160
7161	/*
7162	 * Initializing the Rx buffers. For now we are considering only 1
7163	 * Rx ring and initializing buffers into 30 Rx blocks
7164	 */
7165	config = &sp->config;
7166	mac_control = &sp->mac_control;
7167
7168	for (i = 0; i < config->rx_ring_num; i++) {
7169		struct ring_info *ring = &mac_control->rings[i];
7170
7171		ring->mtu = dev->mtu;
7172		ring->lro = !!(dev->features & NETIF_F_LRO);
7173		ret = fill_rx_buffers(sp, ring, 1);
7174		if (ret) {
7175			DBG_PRINT(ERR_DBG, "%s: Out of memory in Open\n",
7176				  dev->name);
7177			s2io_reset(sp);
7178			free_rx_buffers(sp);
7179			return -ENOMEM;
7180		}
7181		DBG_PRINT(INFO_DBG, "Buf in ring:%d is %d:\n", i,
7182			  ring->rx_bufs_left);
7183	}
7184
7185	/* Initialise napi */
7186	if (config->napi) {
7187		if (config->intr_type ==  MSI_X) {
7188			for (i = 0; i < sp->config.rx_ring_num; i++)
7189				napi_enable(&sp->mac_control.rings[i].napi);
7190		} else {
7191			napi_enable(&sp->napi);
7192		}
7193	}
7194
7195	/* Maintain the state prior to the open */
7196	if (sp->promisc_flg)
7197		sp->promisc_flg = 0;
7198	if (sp->m_cast_flg) {
7199		sp->m_cast_flg = 0;
7200		sp->all_multi_pos = 0;
7201	}
7202
7203	/* Setting its receive mode */
7204	s2io_set_multicast(dev);
7205
7206	if (dev->features & NETIF_F_LRO) {
7207		/* Initialize max aggregatable pkts per session based on MTU */
7208		sp->lro_max_aggr_per_sess = ((1<<16) - 1) / dev->mtu;
7209		/* Check if we can use (if specified) user provided value */
7210		if (lro_max_pkts < sp->lro_max_aggr_per_sess)
7211			sp->lro_max_aggr_per_sess = lro_max_pkts;
7212	}
7213
7214	/* Enable Rx Traffic and interrupts on the NIC */
7215	if (start_nic(sp)) {
7216		DBG_PRINT(ERR_DBG, "%s: Starting NIC failed\n", dev->name);
7217		s2io_reset(sp);
7218		free_rx_buffers(sp);
7219		return -ENODEV;
7220	}
7221
7222	/* Add interrupt service routine */
7223	if (s2io_add_isr(sp) != 0) {
7224		if (sp->config.intr_type == MSI_X)
7225			s2io_rem_isr(sp);
7226		s2io_reset(sp);
7227		free_rx_buffers(sp);
7228		return -ENODEV;
7229	}
7230
7231	S2IO_TIMER_CONF(sp->alarm_timer, s2io_alarm_handle, sp, (HZ/2));
7232
7233	set_bit(__S2IO_STATE_CARD_UP, &sp->state);
7234
7235	/*  Enable select interrupts */
7236	en_dis_err_alarms(sp, ENA_ALL_INTRS, ENABLE_INTRS);
7237	if (sp->config.intr_type != INTA) {
7238		interruptible = TX_TRAFFIC_INTR | TX_PIC_INTR;
7239		en_dis_able_nic_intrs(sp, interruptible, ENABLE_INTRS);
7240	} else {
7241		interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR;
7242		interruptible |= TX_PIC_INTR;
7243		en_dis_able_nic_intrs(sp, interruptible, ENABLE_INTRS);
7244	}
7245
7246	return 0;
7247}
7248
7249/**
7250 * s2io_restart_nic - Resets the NIC.
7251 * @data : long pointer to the device private structure
7252 * Description:
7253 * This function is scheduled to be run by the s2io_tx_watchdog
7254 * function after 0.5 secs to reset the NIC. The idea is to reduce
7255 * the run time of the watch dog routine which is run holding a
7256 * spin lock.
7257 */
7258
7259static void s2io_restart_nic(struct work_struct *work)
7260{
7261	struct s2io_nic *sp = container_of(work, struct s2io_nic, rst_timer_task);
7262	struct net_device *dev = sp->dev;
7263
7264	rtnl_lock();
7265
7266	if (!netif_running(dev))
7267		goto out_unlock;
7268
7269	s2io_card_down(sp);
7270	if (s2io_card_up(sp)) {
7271		DBG_PRINT(ERR_DBG, "%s: Device bring up failed\n", dev->name);
7272	}
7273	s2io_wake_all_tx_queue(sp);
7274	DBG_PRINT(ERR_DBG, "%s: was reset by Tx watchdog timer\n", dev->name);
7275out_unlock:
7276	rtnl_unlock();
7277}
7278
7279/**
7280 *  s2io_tx_watchdog - Watchdog for transmit side.
7281 *  @dev : Pointer to net device structure
7282 *  Description:
7283 *  This function is triggered if the Tx Queue is stopped
7284 *  for a pre-defined amount of time when the Interface is still up.
7285 *  If the Interface is jammed in such a situation, the hardware is
7286 *  reset (by s2io_close) and restarted again (by s2io_open) to
7287 *  overcome any problem that might have been caused in the hardware.
7288 *  Return value:
7289 *  void
7290 */
7291
7292static void s2io_tx_watchdog(struct net_device *dev)
7293{
7294	struct s2io_nic *sp = netdev_priv(dev);
7295	struct swStat *swstats = &sp->mac_control.stats_info->sw_stat;
7296
7297	if (netif_carrier_ok(dev)) {
7298		swstats->watchdog_timer_cnt++;
7299		schedule_work(&sp->rst_timer_task);
7300		swstats->soft_reset_cnt++;
7301	}
7302}
7303
7304/**
7305 *   rx_osm_handler - To perform some OS related operations on SKB.
7306 *   @sp: private member of the device structure,pointer to s2io_nic structure.
7307 *   @skb : the socket buffer pointer.
7308 *   @len : length of the packet
7309 *   @cksum : FCS checksum of the frame.
7310 *   @ring_no : the ring from which this RxD was extracted.
7311 *   Description:
7312 *   This function is called by the Rx interrupt serivce routine to perform
7313 *   some OS related operations on the SKB before passing it to the upper
7314 *   layers. It mainly checks if the checksum is OK, if so adds it to the
7315 *   SKBs cksum variable, increments the Rx packet count and passes the SKB
7316 *   to the upper layer. If the checksum is wrong, it increments the Rx
7317 *   packet error count, frees the SKB and returns error.
7318 *   Return value:
7319 *   SUCCESS on success and -1 on failure.
7320 */
7321static int rx_osm_handler(struct ring_info *ring_data, struct RxD_t * rxdp)
7322{
7323	struct s2io_nic *sp = ring_data->nic;
7324	struct net_device *dev = ring_data->dev;
7325	struct sk_buff *skb = (struct sk_buff *)
7326		((unsigned long)rxdp->Host_Control);
7327	int ring_no = ring_data->ring_no;
7328	u16 l3_csum, l4_csum;
7329	unsigned long long err = rxdp->Control_1 & RXD_T_CODE;
7330	struct lro *uninitialized_var(lro);
7331	u8 err_mask;
7332	struct swStat *swstats = &sp->mac_control.stats_info->sw_stat;
7333
7334	skb->dev = dev;
7335
7336	if (err) {
7337		/* Check for parity error */
7338		if (err & 0x1)
7339			swstats->parity_err_cnt++;
7340
7341		err_mask = err >> 48;
7342		switch (err_mask) {
7343		case 1:
7344			swstats->rx_parity_err_cnt++;
7345			break;
7346
7347		case 2:
7348			swstats->rx_abort_cnt++;
7349			break;
7350
7351		case 3:
7352			swstats->rx_parity_abort_cnt++;
7353			break;
7354
7355		case 4:
7356			swstats->rx_rda_fail_cnt++;
7357			break;
7358
7359		case 5:
7360			swstats->rx_unkn_prot_cnt++;
7361			break;
7362
7363		case 6:
7364			swstats->rx_fcs_err_cnt++;
7365			break;
7366
7367		case 7:
7368			swstats->rx_buf_size_err_cnt++;
7369			break;
7370
7371		case 8:
7372			swstats->rx_rxd_corrupt_cnt++;
7373			break;
7374
7375		case 15:
7376			swstats->rx_unkn_err_cnt++;
7377			break;
7378		}
7379		/*
7380		 * Drop the packet if bad transfer code. Exception being
7381		 * 0x5, which could be due to unsupported IPv6 extension header.
7382		 * In this case, we let stack handle the packet.
7383		 * Note that in this case, since checksum will be incorrect,
7384		 * stack will validate the same.
7385		 */
7386		if (err_mask != 0x5) {
7387			DBG_PRINT(ERR_DBG, "%s: Rx error Value: 0x%x\n",
7388				  dev->name, err_mask);
7389			dev->stats.rx_crc_errors++;
7390			swstats->mem_freed
7391				+= skb->truesize;
7392			dev_kfree_skb(skb);
7393			ring_data->rx_bufs_left -= 1;
7394			rxdp->Host_Control = 0;
7395			return 0;
7396		}
7397	}
7398
7399	rxdp->Host_Control = 0;
7400	if (sp->rxd_mode == RXD_MODE_1) {
7401		int len = RXD_GET_BUFFER0_SIZE_1(rxdp->Control_2);
7402
7403		skb_put(skb, len);
7404	} else if (sp->rxd_mode == RXD_MODE_3B) {
7405		int get_block = ring_data->rx_curr_get_info.block_index;
7406		int get_off = ring_data->rx_curr_get_info.offset;
7407		int buf0_len = RXD_GET_BUFFER0_SIZE_3(rxdp->Control_2);
7408		int buf2_len = RXD_GET_BUFFER2_SIZE_3(rxdp->Control_2);
7409		unsigned char *buff = skb_push(skb, buf0_len);
7410
7411		struct buffAdd *ba = &ring_data->ba[get_block][get_off];
7412		memcpy(buff, ba->ba_0, buf0_len);
7413		skb_put(skb, buf2_len);
7414	}
7415
7416	if ((rxdp->Control_1 & TCP_OR_UDP_FRAME) &&
7417	    ((!ring_data->lro) ||
7418	     (ring_data->lro && (!(rxdp->Control_1 & RXD_FRAME_IP_FRAG)))) &&
7419	    (dev->features & NETIF_F_RXCSUM)) {
7420		l3_csum = RXD_GET_L3_CKSUM(rxdp->Control_1);
7421		l4_csum = RXD_GET_L4_CKSUM(rxdp->Control_1);
7422		if ((l3_csum == L3_CKSUM_OK) && (l4_csum == L4_CKSUM_OK)) {
7423			/*
7424			 * NIC verifies if the Checksum of the received
7425			 * frame is Ok or not and accordingly returns
7426			 * a flag in the RxD.
7427			 */
7428			skb->ip_summed = CHECKSUM_UNNECESSARY;
7429			if (ring_data->lro) {
7430				u32 tcp_len = 0;
7431				u8 *tcp;
7432				int ret = 0;
7433
7434				ret = s2io_club_tcp_session(ring_data,
7435							    skb->data, &tcp,
7436							    &tcp_len, &lro,
7437							    rxdp, sp);
7438				switch (ret) {
7439				case 3: /* Begin anew */
7440					lro->parent = skb;
7441					goto aggregate;
7442				case 1: /* Aggregate */
7443					lro_append_pkt(sp, lro, skb, tcp_len);
7444					goto aggregate;
7445				case 4: /* Flush session */
7446					lro_append_pkt(sp, lro, skb, tcp_len);
7447					queue_rx_frame(lro->parent,
7448						       lro->vlan_tag);
7449					clear_lro_session(lro);
7450					swstats->flush_max_pkts++;
7451					goto aggregate;
7452				case 2: /* Flush both */
7453					lro->parent->data_len = lro->frags_len;
7454					swstats->sending_both++;
7455					queue_rx_frame(lro->parent,
7456						       lro->vlan_tag);
7457					clear_lro_session(lro);
7458					goto send_up;
7459				case 0: /* sessions exceeded */
7460				case -1: /* non-TCP or not L2 aggregatable */
7461				case 5: /*
7462					 * First pkt in session not
7463					 * L3/L4 aggregatable
7464					 */
7465					break;
7466				default:
7467					DBG_PRINT(ERR_DBG,
7468						  "%s: Samadhana!!\n",
7469						  __func__);
7470					BUG();
7471				}
7472			}
7473		} else {
7474			/*
7475			 * Packet with erroneous checksum, let the
7476			 * upper layers deal with it.
7477			 */
7478			skb_checksum_none_assert(skb);
7479		}
7480	} else
7481		skb_checksum_none_assert(skb);
7482
7483	swstats->mem_freed += skb->truesize;
7484send_up:
7485	skb_record_rx_queue(skb, ring_no);
7486	queue_rx_frame(skb, RXD_GET_VLAN_TAG(rxdp->Control_2));
7487aggregate:
7488	sp->mac_control.rings[ring_no].rx_bufs_left -= 1;
7489	return SUCCESS;
7490}
7491
7492/**
7493 *  s2io_link - stops/starts the Tx queue.
7494 *  @sp : private member of the device structure, which is a pointer to the
7495 *  s2io_nic structure.
7496 *  @link : inidicates whether link is UP/DOWN.
7497 *  Description:
7498 *  This function stops/starts the Tx queue depending on whether the link
7499 *  status of the NIC is is down or up. This is called by the Alarm
7500 *  interrupt handler whenever a link change interrupt comes up.
7501 *  Return value:
7502 *  void.
7503 */
7504
7505static void s2io_link(struct s2io_nic *sp, int link)
7506{
7507	struct net_device *dev = sp->dev;
7508	struct swStat *swstats = &sp->mac_control.stats_info->sw_stat;
7509
7510	if (link != sp->last_link_state) {
7511		init_tti(sp, link);
7512		if (link == LINK_DOWN) {
7513			DBG_PRINT(ERR_DBG, "%s: Link down\n", dev->name);
7514			s2io_stop_all_tx_queue(sp);
7515			netif_carrier_off(dev);
7516			if (swstats->link_up_cnt)
7517				swstats->link_up_time =
7518					jiffies - sp->start_time;
7519			swstats->link_down_cnt++;
7520		} else {
7521			DBG_PRINT(ERR_DBG, "%s: Link Up\n", dev->name);
7522			if (swstats->link_down_cnt)
7523				swstats->link_down_time =
7524					jiffies - sp->start_time;
7525			swstats->link_up_cnt++;
7526			netif_carrier_on(dev);
7527			s2io_wake_all_tx_queue(sp);
7528		}
7529	}
7530	sp->last_link_state = link;
7531	sp->start_time = jiffies;
7532}
7533
7534/**
7535 *  s2io_init_pci -Initialization of PCI and PCI-X configuration registers .
7536 *  @sp : private member of the device structure, which is a pointer to the
7537 *  s2io_nic structure.
7538 *  Description:
7539 *  This function initializes a few of the PCI and PCI-X configuration registers
7540 *  with recommended values.
7541 *  Return value:
7542 *  void
7543 */
7544
7545static void s2io_init_pci(struct s2io_nic *sp)
7546{
7547	u16 pci_cmd = 0, pcix_cmd = 0;
7548
7549	/* Enable Data Parity Error Recovery in PCI-X command register. */
7550	pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
7551			     &(pcix_cmd));
7552	pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
7553			      (pcix_cmd | 1));
7554	pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
7555			     &(pcix_cmd));
7556
7557	/* Set the PErr Response bit in PCI command register. */
7558	pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
7559	pci_write_config_word(sp->pdev, PCI_COMMAND,
7560			      (pci_cmd | PCI_COMMAND_PARITY));
7561	pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
7562}
7563
7564static int s2io_verify_parm(struct pci_dev *pdev, u8 *dev_intr_type,
7565			    u8 *dev_multiq)
7566{
7567	int i;
7568
7569	if ((tx_fifo_num > MAX_TX_FIFOS) || (tx_fifo_num < 1)) {
7570		DBG_PRINT(ERR_DBG, "Requested number of tx fifos "
7571			  "(%d) not supported\n", tx_fifo_num);
7572
7573		if (tx_fifo_num < 1)
7574			tx_fifo_num = 1;
7575		else
7576			tx_fifo_num = MAX_TX_FIFOS;
7577
7578		DBG_PRINT(ERR_DBG, "Default to %d tx fifos\n", tx_fifo_num);
7579	}
7580
7581	if (multiq)
7582		*dev_multiq = multiq;
7583
7584	if (tx_steering_type && (1 == tx_fifo_num)) {
7585		if (tx_steering_type != TX_DEFAULT_STEERING)
7586			DBG_PRINT(ERR_DBG,
7587				  "Tx steering is not supported with "
7588				  "one fifo. Disabling Tx steering.\n");
7589		tx_steering_type = NO_STEERING;
7590	}
7591
7592	if ((tx_steering_type < NO_STEERING) ||
7593	    (tx_steering_type > TX_DEFAULT_STEERING)) {
7594		DBG_PRINT(ERR_DBG,
7595			  "Requested transmit steering not supported\n");
7596		DBG_PRINT(ERR_DBG, "Disabling transmit steering\n");
7597		tx_steering_type = NO_STEERING;
7598	}
7599
7600	if (rx_ring_num > MAX_RX_RINGS) {
7601		DBG_PRINT(ERR_DBG,
7602			  "Requested number of rx rings not supported\n");
7603		DBG_PRINT(ERR_DBG, "Default to %d rx rings\n",
7604			  MAX_RX_RINGS);
7605		rx_ring_num = MAX_RX_RINGS;
7606	}
7607
7608	if ((*dev_intr_type != INTA) && (*dev_intr_type != MSI_X)) {
7609		DBG_PRINT(ERR_DBG, "Wrong intr_type requested. "
7610			  "Defaulting to INTA\n");
7611		*dev_intr_type = INTA;
7612	}
7613
7614	if ((*dev_intr_type == MSI_X) &&
7615	    ((pdev->device != PCI_DEVICE_ID_HERC_WIN) &&
7616	     (pdev->device != PCI_DEVICE_ID_HERC_UNI))) {
7617		DBG_PRINT(ERR_DBG, "Xframe I does not support MSI_X. "
7618			  "Defaulting to INTA\n");
7619		*dev_intr_type = INTA;
7620	}
7621
7622	if ((rx_ring_mode != 1) && (rx_ring_mode != 2)) {
7623		DBG_PRINT(ERR_DBG, "Requested ring mode not supported\n");
7624		DBG_PRINT(ERR_DBG, "Defaulting to 1-buffer mode\n");
7625		rx_ring_mode = 1;
7626	}
7627
7628	for (i = 0; i < MAX_RX_RINGS; i++)
7629		if (rx_ring_sz[i] > MAX_RX_BLOCKS_PER_RING) {
7630			DBG_PRINT(ERR_DBG, "Requested rx ring size not "
7631				  "supported\nDefaulting to %d\n",
7632				  MAX_RX_BLOCKS_PER_RING);
7633			rx_ring_sz[i] = MAX_RX_BLOCKS_PER_RING;
7634		}
7635
7636	return SUCCESS;
7637}
7638
7639/**
7640 * rts_ds_steer - Receive traffic steering based on IPv4 or IPv6 TOS
7641 * or Traffic class respectively.
7642 * @nic: device private variable
7643 * Description: The function configures the receive steering to
7644 * desired receive ring.
7645 * Return Value:  SUCCESS on success and
7646 * '-1' on failure (endian settings incorrect).
7647 */
7648static int rts_ds_steer(struct s2io_nic *nic, u8 ds_codepoint, u8 ring)
7649{
7650	struct XENA_dev_config __iomem *bar0 = nic->bar0;
7651	register u64 val64 = 0;
7652
7653	if (ds_codepoint > 63)
7654		return FAILURE;
7655
7656	val64 = RTS_DS_MEM_DATA(ring);
7657	writeq(val64, &bar0->rts_ds_mem_data);
7658
7659	val64 = RTS_DS_MEM_CTRL_WE |
7660		RTS_DS_MEM_CTRL_STROBE_NEW_CMD |
7661		RTS_DS_MEM_CTRL_OFFSET(ds_codepoint);
7662
7663	writeq(val64, &bar0->rts_ds_mem_ctrl);
7664
7665	return wait_for_cmd_complete(&bar0->rts_ds_mem_ctrl,
7666				     RTS_DS_MEM_CTRL_STROBE_CMD_BEING_EXECUTED,
7667				     S2IO_BIT_RESET);
7668}
7669
7670static const struct net_device_ops s2io_netdev_ops = {
7671	.ndo_open	        = s2io_open,
7672	.ndo_stop	        = s2io_close,
7673	.ndo_get_stats	        = s2io_get_stats,
7674	.ndo_start_xmit    	= s2io_xmit,
7675	.ndo_validate_addr	= eth_validate_addr,
7676	.ndo_set_rx_mode	= s2io_set_multicast,
7677	.ndo_do_ioctl	   	= s2io_ioctl,
7678	.ndo_set_mac_address    = s2io_set_mac_addr,
7679	.ndo_change_mtu	   	= s2io_change_mtu,
7680	.ndo_set_features	= s2io_set_features,
7681	.ndo_tx_timeout	   	= s2io_tx_watchdog,
7682#ifdef CONFIG_NET_POLL_CONTROLLER
7683	.ndo_poll_controller    = s2io_netpoll,
7684#endif
7685};
7686
7687/**
7688 *  s2io_init_nic - Initialization of the adapter .
7689 *  @pdev : structure containing the PCI related information of the device.
7690 *  @pre: List of PCI devices supported by the driver listed in s2io_tbl.
7691 *  Description:
7692 *  The function initializes an adapter identified by the pci_dec structure.
7693 *  All OS related initialization including memory and device structure and
7694 *  initlaization of the device private variable is done. Also the swapper
7695 *  control register is initialized to enable read and write into the I/O
7696 *  registers of the device.
7697 *  Return value:
7698 *  returns 0 on success and negative on failure.
7699 */
7700
7701static int
7702s2io_init_nic(struct pci_dev *pdev, const struct pci_device_id *pre)
7703{
7704	struct s2io_nic *sp;
7705	struct net_device *dev;
7706	int i, j, ret;
7707	int dma_flag = false;
7708	u32 mac_up, mac_down;
7709	u64 val64 = 0, tmp64 = 0;
7710	struct XENA_dev_config __iomem *bar0 = NULL;
7711	u16 subid;
7712	struct config_param *config;
7713	struct mac_info *mac_control;
7714	int mode;
7715	u8 dev_intr_type = intr_type;
7716	u8 dev_multiq = 0;
7717
7718	ret = s2io_verify_parm(pdev, &dev_intr_type, &dev_multiq);
7719	if (ret)
7720		return ret;
7721
7722	ret = pci_enable_device(pdev);
7723	if (ret) {
7724		DBG_PRINT(ERR_DBG,
7725			  "%s: pci_enable_device failed\n", __func__);
7726		return ret;
7727	}
7728
7729	if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
7730		DBG_PRINT(INIT_DBG, "%s: Using 64bit DMA\n", __func__);
7731		dma_flag = true;
7732		if (pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64))) {
7733			DBG_PRINT(ERR_DBG,
7734				  "Unable to obtain 64bit DMA "
7735				  "for consistent allocations\n");
7736			pci_disable_device(pdev);
7737			return -ENOMEM;
7738		}
7739	} else if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(32))) {
7740		DBG_PRINT(INIT_DBG, "%s: Using 32bit DMA\n", __func__);
7741	} else {
7742		pci_disable_device(pdev);
7743		return -ENOMEM;
7744	}
7745	ret = pci_request_regions(pdev, s2io_driver_name);
7746	if (ret) {
7747		DBG_PRINT(ERR_DBG, "%s: Request Regions failed - %x\n",
7748			  __func__, ret);
7749		pci_disable_device(pdev);
7750		return -ENODEV;
7751	}
7752	if (dev_multiq)
7753		dev = alloc_etherdev_mq(sizeof(struct s2io_nic), tx_fifo_num);
7754	else
7755		dev = alloc_etherdev(sizeof(struct s2io_nic));
7756	if (dev == NULL) {
7757		pci_disable_device(pdev);
7758		pci_release_regions(pdev);
7759		return -ENODEV;
7760	}
7761
7762	pci_set_master(pdev);
7763	pci_set_drvdata(pdev, dev);
7764	SET_NETDEV_DEV(dev, &pdev->dev);
7765
7766	/*  Private member variable initialized to s2io NIC structure */
7767	sp = netdev_priv(dev);
7768	sp->dev = dev;
7769	sp->pdev = pdev;
7770	sp->high_dma_flag = dma_flag;
7771	sp->device_enabled_once = false;
7772	if (rx_ring_mode == 1)
7773		sp->rxd_mode = RXD_MODE_1;
7774	if (rx_ring_mode == 2)
7775		sp->rxd_mode = RXD_MODE_3B;
7776
7777	sp->config.intr_type = dev_intr_type;
7778
7779	if ((pdev->device == PCI_DEVICE_ID_HERC_WIN) ||
7780	    (pdev->device == PCI_DEVICE_ID_HERC_UNI))
7781		sp->device_type = XFRAME_II_DEVICE;
7782	else
7783		sp->device_type = XFRAME_I_DEVICE;
7784
7785
7786	/* Initialize some PCI/PCI-X fields of the NIC. */
7787	s2io_init_pci(sp);
7788
7789	/*
7790	 * Setting the device configuration parameters.
7791	 * Most of these parameters can be specified by the user during
7792	 * module insertion as they are module loadable parameters. If
7793	 * these parameters are not not specified during load time, they
7794	 * are initialized with default values.
7795	 */
7796	config = &sp->config;
7797	mac_control = &sp->mac_control;
7798
7799	config->napi = napi;
7800	config->tx_steering_type = tx_steering_type;
7801
7802	/* Tx side parameters. */
7803	if (config->tx_steering_type == TX_PRIORITY_STEERING)
7804		config->tx_fifo_num = MAX_TX_FIFOS;
7805	else
7806		config->tx_fifo_num = tx_fifo_num;
7807
7808	/* Initialize the fifos used for tx steering */
7809	if (config->tx_fifo_num < 5) {
7810		if (config->tx_fifo_num  == 1)
7811			sp->total_tcp_fifos = 1;
7812		else
7813			sp->total_tcp_fifos = config->tx_fifo_num - 1;
7814		sp->udp_fifo_idx = config->tx_fifo_num - 1;
7815		sp->total_udp_fifos = 1;
7816		sp->other_fifo_idx = sp->total_tcp_fifos - 1;
7817	} else {
7818		sp->total_tcp_fifos = (tx_fifo_num - FIFO_UDP_MAX_NUM -
7819				       FIFO_OTHER_MAX_NUM);
7820		sp->udp_fifo_idx = sp->total_tcp_fifos;
7821		sp->total_udp_fifos = FIFO_UDP_MAX_NUM;
7822		sp->other_fifo_idx = sp->udp_fifo_idx + FIFO_UDP_MAX_NUM;
7823	}
7824
7825	config->multiq = dev_multiq;
7826	for (i = 0; i < config->tx_fifo_num; i++) {
7827		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
7828
7829		tx_cfg->fifo_len = tx_fifo_len[i];
7830		tx_cfg->fifo_priority = i;
7831	}
7832
7833	/* mapping the QoS priority to the configured fifos */
7834	for (i = 0; i < MAX_TX_FIFOS; i++)
7835		config->fifo_mapping[i] = fifo_map[config->tx_fifo_num - 1][i];
7836
7837	/* map the hashing selector table to the configured fifos */
7838	for (i = 0; i < config->tx_fifo_num; i++)
7839		sp->fifo_selector[i] = fifo_selector[i];
7840
7841
7842	config->tx_intr_type = TXD_INT_TYPE_UTILZ;
7843	for (i = 0; i < config->tx_fifo_num; i++) {
7844		struct tx_fifo_config *tx_cfg = &config->tx_cfg[i];
7845
7846		tx_cfg->f_no_snoop = (NO_SNOOP_TXD | NO_SNOOP_TXD_BUFFER);
7847		if (tx_cfg->fifo_len < 65) {
7848			config->tx_intr_type = TXD_INT_TYPE_PER_LIST;
7849			break;
7850		}
7851	}
7852	/* + 2 because one Txd for skb->data and one Txd for UFO */
7853	config->max_txds = MAX_SKB_FRAGS + 2;
7854
7855	/* Rx side parameters. */
7856	config->rx_ring_num = rx_ring_num;
7857	for (i = 0; i < config->rx_ring_num; i++) {
7858		struct rx_ring_config *rx_cfg = &config->rx_cfg[i];
7859		struct ring_info *ring = &mac_control->rings[i];
7860
7861		rx_cfg->num_rxd = rx_ring_sz[i] * (rxd_count[sp->rxd_mode] + 1);
7862		rx_cfg->ring_priority = i;
7863		ring->rx_bufs_left = 0;
7864		ring->rxd_mode = sp->rxd_mode;
7865		ring->rxd_count = rxd_count[sp->rxd_mode];
7866		ring->pdev = sp->pdev;
7867		ring->dev = sp->dev;
7868	}
7869
7870	for (i = 0; i < rx_ring_num; i++) {
7871		struct rx_ring_config *rx_cfg = &config->rx_cfg[i];
7872
7873		rx_cfg->ring_org = RING_ORG_BUFF1;
7874		rx_cfg->f_no_snoop = (NO_SNOOP_RXD | NO_SNOOP_RXD_BUFFER);
7875	}
7876
7877	/*  Setting Mac Control parameters */
7878	mac_control->rmac_pause_time = rmac_pause_time;
7879	mac_control->mc_pause_threshold_q0q3 = mc_pause_threshold_q0q3;
7880	mac_control->mc_pause_threshold_q4q7 = mc_pause_threshold_q4q7;
7881
7882
7883	/*  initialize the shared memory used by the NIC and the host */
7884	if (init_shared_mem(sp)) {
7885		DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n", dev->name);
7886		ret = -ENOMEM;
7887		goto mem_alloc_failed;
7888	}
7889
7890	sp->bar0 = pci_ioremap_bar(pdev, 0);
7891	if (!sp->bar0) {
7892		DBG_PRINT(ERR_DBG, "%s: Neterion: cannot remap io mem1\n",
7893			  dev->name);
7894		ret = -ENOMEM;
7895		goto bar0_remap_failed;
7896	}
7897
7898	sp->bar1 = pci_ioremap_bar(pdev, 2);
7899	if (!sp->bar1) {
7900		DBG_PRINT(ERR_DBG, "%s: Neterion: cannot remap io mem2\n",
7901			  dev->name);
7902		ret = -ENOMEM;
7903		goto bar1_remap_failed;
7904	}
7905
7906	/* Initializing the BAR1 address as the start of the FIFO pointer. */
7907	for (j = 0; j < MAX_TX_FIFOS; j++) {
7908		mac_control->tx_FIFO_start[j] = sp->bar1 + (j * 0x00020000);
7909	}
7910
7911	/*  Driver entry points */
7912	dev->netdev_ops = &s2io_netdev_ops;
7913	dev->ethtool_ops = &netdev_ethtool_ops;
7914	dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM |
7915		NETIF_F_TSO | NETIF_F_TSO6 |
7916		NETIF_F_RXCSUM | NETIF_F_LRO;
7917	dev->features |= dev->hw_features |
7918		NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX;
7919	if (sp->device_type & XFRAME_II_DEVICE) {
7920		dev->hw_features |= NETIF_F_UFO;
7921		if (ufo)
7922			dev->features |= NETIF_F_UFO;
7923	}
7924	if (sp->high_dma_flag == true)
7925		dev->features |= NETIF_F_HIGHDMA;
7926	dev->watchdog_timeo = WATCH_DOG_TIMEOUT;
7927	INIT_WORK(&sp->rst_timer_task, s2io_restart_nic);
7928	INIT_WORK(&sp->set_link_task, s2io_set_link);
7929
7930	pci_save_state(sp->pdev);
7931
7932	/* Setting swapper control on the NIC, for proper reset operation */
7933	if (s2io_set_swapper(sp)) {
7934		DBG_PRINT(ERR_DBG, "%s: swapper settings are wrong\n",
7935			  dev->name);
7936		ret = -EAGAIN;
7937		goto set_swap_failed;
7938	}
7939
7940	/* Verify if the Herc works on the slot its placed into */
7941	if (sp->device_type & XFRAME_II_DEVICE) {
7942		mode = s2io_verify_pci_mode(sp);
7943		if (mode < 0) {
7944			DBG_PRINT(ERR_DBG, "%s: Unsupported PCI bus mode\n",
7945				  __func__);
7946			ret = -EBADSLT;
7947			goto set_swap_failed;
7948		}
7949	}
7950
7951	if (sp->config.intr_type == MSI_X) {
7952		sp->num_entries = config->rx_ring_num + 1;
7953		ret = s2io_enable_msi_x(sp);
7954
7955		if (!ret) {
7956			ret = s2io_test_msi(sp);
7957			/* rollback MSI-X, will re-enable during add_isr() */
7958			remove_msix_isr(sp);
7959		}
7960		if (ret) {
7961
7962			DBG_PRINT(ERR_DBG,
7963				  "MSI-X requested but failed to enable\n");
7964			sp->config.intr_type = INTA;
7965		}
7966	}
7967
7968	if (config->intr_type ==  MSI_X) {
7969		for (i = 0; i < config->rx_ring_num ; i++) {
7970			struct ring_info *ring = &mac_control->rings[i];
7971
7972			netif_napi_add(dev, &ring->napi, s2io_poll_msix, 64);
7973		}
7974	} else {
7975		netif_napi_add(dev, &sp->napi, s2io_poll_inta, 64);
7976	}
7977
7978	/* Not needed for Herc */
7979	if (sp->device_type & XFRAME_I_DEVICE) {
7980		/*
7981		 * Fix for all "FFs" MAC address problems observed on
7982		 * Alpha platforms
7983		 */
7984		fix_mac_address(sp);
7985		s2io_reset(sp);
7986	}
7987
7988	/*
7989	 * MAC address initialization.
7990	 * For now only one mac address will be read and used.
7991	 */
7992	bar0 = sp->bar0;
7993	val64 = RMAC_ADDR_CMD_MEM_RD | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
7994		RMAC_ADDR_CMD_MEM_OFFSET(0 + S2IO_MAC_ADDR_START_OFFSET);
7995	writeq(val64, &bar0->rmac_addr_cmd_mem);
7996	wait_for_cmd_complete(&bar0->rmac_addr_cmd_mem,
7997			      RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING,
7998			      S2IO_BIT_RESET);
7999	tmp64 = readq(&bar0->rmac_addr_data0_mem);
8000	mac_down = (u32)tmp64;
8001	mac_up = (u32) (tmp64 >> 32);
8002
8003	sp->def_mac_addr[0].mac_addr[3] = (u8) (mac_up);
8004	sp->def_mac_addr[0].mac_addr[2] = (u8) (mac_up >> 8);
8005	sp->def_mac_addr[0].mac_addr[1] = (u8) (mac_up >> 16);
8006	sp->def_mac_addr[0].mac_addr[0] = (u8) (mac_up >> 24);
8007	sp->def_mac_addr[0].mac_addr[5] = (u8) (mac_down >> 16);
8008	sp->def_mac_addr[0].mac_addr[4] = (u8) (mac_down >> 24);
8009
8010	/*  Set the factory defined MAC address initially   */
8011	dev->addr_len = ETH_ALEN;
8012	memcpy(dev->dev_addr, sp->def_mac_addr, ETH_ALEN);
8013
8014	/* initialize number of multicast & unicast MAC entries variables */
8015	if (sp->device_type == XFRAME_I_DEVICE) {
8016		config->max_mc_addr = S2IO_XENA_MAX_MC_ADDRESSES;
8017		config->max_mac_addr = S2IO_XENA_MAX_MAC_ADDRESSES;
8018		config->mc_start_offset = S2IO_XENA_MC_ADDR_START_OFFSET;
8019	} else if (sp->device_type == XFRAME_II_DEVICE) {
8020		config->max_mc_addr = S2IO_HERC_MAX_MC_ADDRESSES;
8021		config->max_mac_addr = S2IO_HERC_MAX_MAC_ADDRESSES;
8022		config->mc_start_offset = S2IO_HERC_MC_ADDR_START_OFFSET;
8023	}
8024
8025	/* store mac addresses from CAM to s2io_nic structure */
8026	do_s2io_store_unicast_mc(sp);
8027
8028	/* Configure MSIX vector for number of rings configured plus one */
8029	if ((sp->device_type == XFRAME_II_DEVICE) &&
8030	    (config->intr_type == MSI_X))
8031		sp->num_entries = config->rx_ring_num + 1;
8032
8033	/* Store the values of the MSIX table in the s2io_nic structure */
8034	store_xmsi_data(sp);
8035	/* reset Nic and bring it to known state */
8036	s2io_reset(sp);
8037
8038	/*
8039	 * Initialize link state flags
8040	 * and the card state parameter
8041	 */
8042	sp->state = 0;
8043
8044	/* Initialize spinlocks */
8045	for (i = 0; i < sp->config.tx_fifo_num; i++) {
8046		struct fifo_info *fifo = &mac_control->fifos[i];
8047
8048		spin_lock_init(&fifo->tx_lock);
8049	}
8050
8051	/*
8052	 * SXE-002: Configure link and activity LED to init state
8053	 * on driver load.
8054	 */
8055	subid = sp->pdev->subsystem_device;
8056	if ((subid & 0xFF) >= 0x07) {
8057		val64 = readq(&bar0->gpio_control);
8058		val64 |= 0x0000800000000000ULL;
8059		writeq(val64, &bar0->gpio_control);
8060		val64 = 0x0411040400000000ULL;
8061		writeq(val64, (void __iomem *)bar0 + 0x2700);
8062		val64 = readq(&bar0->gpio_control);
8063	}
8064
8065	sp->rx_csum = 1;	/* Rx chksum verify enabled by default */
8066
8067	if (register_netdev(dev)) {
8068		DBG_PRINT(ERR_DBG, "Device registration failed\n");
8069		ret = -ENODEV;
8070		goto register_failed;
8071	}
8072	s2io_vpd_read(sp);
8073	DBG_PRINT(ERR_DBG, "Copyright(c) 2002-2010 Exar Corp.\n");
8074	DBG_PRINT(ERR_DBG, "%s: Neterion %s (rev %d)\n", dev->name,
8075		  sp->product_name, pdev->revision);
8076	DBG_PRINT(ERR_DBG, "%s: Driver version %s\n", dev->name,
8077		  s2io_driver_version);
8078	DBG_PRINT(ERR_DBG, "%s: MAC Address: %pM\n", dev->name, dev->dev_addr);
8079	DBG_PRINT(ERR_DBG, "Serial number: %s\n", sp->serial_num);
8080	if (sp->device_type & XFRAME_II_DEVICE) {
8081		mode = s2io_print_pci_mode(sp);
8082		if (mode < 0) {
8083			ret = -EBADSLT;
8084			unregister_netdev(dev);
8085			goto set_swap_failed;
8086		}
8087	}
8088	switch (sp->rxd_mode) {
8089	case RXD_MODE_1:
8090		DBG_PRINT(ERR_DBG, "%s: 1-Buffer receive mode enabled\n",
8091			  dev->name);
8092		break;
8093	case RXD_MODE_3B:
8094		DBG_PRINT(ERR_DBG, "%s: 2-Buffer receive mode enabled\n",
8095			  dev->name);
8096		break;
8097	}
8098
8099	switch (sp->config.napi) {
8100	case 0:
8101		DBG_PRINT(ERR_DBG, "%s: NAPI disabled\n", dev->name);
8102		break;
8103	case 1:
8104		DBG_PRINT(ERR_DBG, "%s: NAPI enabled\n", dev->name);
8105		break;
8106	}
8107
8108	DBG_PRINT(ERR_DBG, "%s: Using %d Tx fifo(s)\n", dev->name,
8109		  sp->config.tx_fifo_num);
8110
8111	DBG_PRINT(ERR_DBG, "%s: Using %d Rx ring(s)\n", dev->name,
8112		  sp->config.rx_ring_num);
8113
8114	switch (sp->config.intr_type) {
8115	case INTA:
8116		DBG_PRINT(ERR_DBG, "%s: Interrupt type INTA\n", dev->name);
8117		break;
8118	case MSI_X:
8119		DBG_PRINT(ERR_DBG, "%s: Interrupt type MSI-X\n", dev->name);
8120		break;
8121	}
8122	if (sp->config.multiq) {
8123		for (i = 0; i < sp->config.tx_fifo_num; i++) {
8124			struct fifo_info *fifo = &mac_control->fifos[i];
8125
8126			fifo->multiq = config->multiq;
8127		}
8128		DBG_PRINT(ERR_DBG, "%s: Multiqueue support enabled\n",
8129			  dev->name);
8130	} else
8131		DBG_PRINT(ERR_DBG, "%s: Multiqueue support disabled\n",
8132			  dev->name);
8133
8134	switch (sp->config.tx_steering_type) {
8135	case NO_STEERING:
8136		DBG_PRINT(ERR_DBG, "%s: No steering enabled for transmit\n",
8137			  dev->name);
8138		break;
8139	case TX_PRIORITY_STEERING:
8140		DBG_PRINT(ERR_DBG,
8141			  "%s: Priority steering enabled for transmit\n",
8142			  dev->name);
8143		break;
8144	case TX_DEFAULT_STEERING:
8145		DBG_PRINT(ERR_DBG,
8146			  "%s: Default steering enabled for transmit\n",
8147			  dev->name);
8148	}
8149
8150	DBG_PRINT(ERR_DBG, "%s: Large receive offload enabled\n",
8151		  dev->name);
8152	if (ufo)
8153		DBG_PRINT(ERR_DBG,
8154			  "%s: UDP Fragmentation Offload(UFO) enabled\n",
8155			  dev->name);
8156	/* Initialize device name */
8157	sprintf(sp->name, "%s Neterion %s", dev->name, sp->product_name);
8158
8159	if (vlan_tag_strip)
8160		sp->vlan_strip_flag = 1;
8161	else
8162		sp->vlan_strip_flag = 0;
8163
8164	/*
8165	 * Make Link state as off at this point, when the Link change
8166	 * interrupt comes the state will be automatically changed to
8167	 * the right state.
8168	 */
8169	netif_carrier_off(dev);
8170
8171	return 0;
8172
8173register_failed:
8174set_swap_failed:
8175	iounmap(sp->bar1);
8176bar1_remap_failed:
8177	iounmap(sp->bar0);
8178bar0_remap_failed:
8179mem_alloc_failed:
8180	free_shared_mem(sp);
8181	pci_disable_device(pdev);
8182	pci_release_regions(pdev);
8183	free_netdev(dev);
8184
8185	return ret;
8186}
8187
8188/**
8189 * s2io_rem_nic - Free the PCI device
8190 * @pdev: structure containing the PCI related information of the device.
8191 * Description: This function is called by the Pci subsystem to release a
8192 * PCI device and free up all resource held up by the device. This could
8193 * be in response to a Hot plug event or when the driver is to be removed
8194 * from memory.
8195 */
8196
8197static void s2io_rem_nic(struct pci_dev *pdev)
8198{
8199	struct net_device *dev = pci_get_drvdata(pdev);
8200	struct s2io_nic *sp;
8201
8202	if (dev == NULL) {
8203		DBG_PRINT(ERR_DBG, "Driver Data is NULL!!\n");
8204		return;
8205	}
8206
8207	sp = netdev_priv(dev);
8208
8209	cancel_work_sync(&sp->rst_timer_task);
8210	cancel_work_sync(&sp->set_link_task);
8211
8212	unregister_netdev(dev);
8213
8214	free_shared_mem(sp);
8215	iounmap(sp->bar0);
8216	iounmap(sp->bar1);
8217	pci_release_regions(pdev);
8218	free_netdev(dev);
8219	pci_disable_device(pdev);
8220}
8221
8222/**
8223 * s2io_starter - Entry point for the driver
8224 * Description: This function is the entry point for the driver. It verifies
8225 * the module loadable parameters and initializes PCI configuration space.
8226 */
8227
8228static int __init s2io_starter(void)
8229{
8230	return pci_register_driver(&s2io_driver);
8231}
8232
8233/**
8234 * s2io_closer - Cleanup routine for the driver
8235 * Description: This function is the cleanup routine for the driver. It
8236 * unregisters the driver.
8237 */
8238
8239static __exit void s2io_closer(void)
8240{
8241	pci_unregister_driver(&s2io_driver);
8242	DBG_PRINT(INIT_DBG, "cleanup done\n");
8243}
8244
8245module_init(s2io_starter);
8246module_exit(s2io_closer);
8247
8248static int check_L2_lro_capable(u8 *buffer, struct iphdr **ip,
8249				struct tcphdr **tcp, struct RxD_t *rxdp,
8250				struct s2io_nic *sp)
8251{
8252	int ip_off;
8253	u8 l2_type = (u8)((rxdp->Control_1 >> 37) & 0x7), ip_len;
8254
8255	if (!(rxdp->Control_1 & RXD_FRAME_PROTO_TCP)) {
8256		DBG_PRINT(INIT_DBG,
8257			  "%s: Non-TCP frames not supported for LRO\n",
8258			  __func__);
8259		return -1;
8260	}
8261
8262	/* Checking for DIX type or DIX type with VLAN */
8263	if ((l2_type == 0) || (l2_type == 4)) {
8264		ip_off = HEADER_ETHERNET_II_802_3_SIZE;
8265		/*
8266		 * If vlan stripping is disabled and the frame is VLAN tagged,
8267		 * shift the offset by the VLAN header size bytes.
8268		 */
8269		if ((!sp->vlan_strip_flag) &&
8270		    (rxdp->Control_1 & RXD_FRAME_VLAN_TAG))
8271			ip_off += HEADER_VLAN_SIZE;
8272	} else {
8273		/* LLC, SNAP etc are considered non-mergeable */
8274		return -1;
8275	}
8276
8277	*ip = (struct iphdr *)(buffer + ip_off);
8278	ip_len = (u8)((*ip)->ihl);
8279	ip_len <<= 2;
8280	*tcp = (struct tcphdr *)((unsigned long)*ip + ip_len);
8281
8282	return 0;
8283}
8284
8285static int check_for_socket_match(struct lro *lro, struct iphdr *ip,
8286				  struct tcphdr *tcp)
8287{
8288	DBG_PRINT(INFO_DBG, "%s: Been here...\n", __func__);
8289	if ((lro->iph->saddr != ip->saddr) ||
8290	    (lro->iph->daddr != ip->daddr) ||
8291	    (lro->tcph->source != tcp->source) ||
8292	    (lro->tcph->dest != tcp->dest))
8293		return -1;
8294	return 0;
8295}
8296
8297static inline int get_l4_pyld_length(struct iphdr *ip, struct tcphdr *tcp)
8298{
8299	return ntohs(ip->tot_len) - (ip->ihl << 2) - (tcp->doff << 2);
8300}
8301
8302static void initiate_new_session(struct lro *lro, u8 *l2h,
8303				 struct iphdr *ip, struct tcphdr *tcp,
8304				 u32 tcp_pyld_len, u16 vlan_tag)
8305{
8306	DBG_PRINT(INFO_DBG, "%s: Been here...\n", __func__);
8307	lro->l2h = l2h;
8308	lro->iph = ip;
8309	lro->tcph = tcp;
8310	lro->tcp_next_seq = tcp_pyld_len + ntohl(tcp->seq);
8311	lro->tcp_ack = tcp->ack_seq;
8312	lro->sg_num = 1;
8313	lro->total_len = ntohs(ip->tot_len);
8314	lro->frags_len = 0;
8315	lro->vlan_tag = vlan_tag;
8316	/*
8317	 * Check if we saw TCP timestamp.
8318	 * Other consistency checks have already been done.
8319	 */
8320	if (tcp->doff == 8) {
8321		__be32 *ptr;
8322		ptr = (__be32 *)(tcp+1);
8323		lro->saw_ts = 1;
8324		lro->cur_tsval = ntohl(*(ptr+1));
8325		lro->cur_tsecr = *(ptr+2);
8326	}
8327	lro->in_use = 1;
8328}
8329
8330static void update_L3L4_header(struct s2io_nic *sp, struct lro *lro)
8331{
8332	struct iphdr *ip = lro->iph;
8333	struct tcphdr *tcp = lro->tcph;
8334	struct swStat *swstats = &sp->mac_control.stats_info->sw_stat;
8335
8336	DBG_PRINT(INFO_DBG, "%s: Been here...\n", __func__);
8337
8338	/* Update L3 header */
8339	csum_replace2(&ip->check, ip->tot_len, htons(lro->total_len));
8340	ip->tot_len = htons(lro->total_len);
8341
8342	/* Update L4 header */
8343	tcp->ack_seq = lro->tcp_ack;
8344	tcp->window = lro->window;
8345
8346	/* Update tsecr field if this session has timestamps enabled */
8347	if (lro->saw_ts) {
8348		__be32 *ptr = (__be32 *)(tcp + 1);
8349		*(ptr+2) = lro->cur_tsecr;
8350	}
8351
8352	/* Update counters required for calculation of
8353	 * average no. of packets aggregated.
8354	 */
8355	swstats->sum_avg_pkts_aggregated += lro->sg_num;
8356	swstats->num_aggregations++;
8357}
8358
8359static void aggregate_new_rx(struct lro *lro, struct iphdr *ip,
8360			     struct tcphdr *tcp, u32 l4_pyld)
8361{
8362	DBG_PRINT(INFO_DBG, "%s: Been here...\n", __func__);
8363	lro->total_len += l4_pyld;
8364	lro->frags_len += l4_pyld;
8365	lro->tcp_next_seq += l4_pyld;
8366	lro->sg_num++;
8367
8368	/* Update ack seq no. and window ad(from this pkt) in LRO object */
8369	lro->tcp_ack = tcp->ack_seq;
8370	lro->window = tcp->window;
8371
8372	if (lro->saw_ts) {
8373		__be32 *ptr;
8374		/* Update tsecr and tsval from this packet */
8375		ptr = (__be32 *)(tcp+1);
8376		lro->cur_tsval = ntohl(*(ptr+1));
8377		lro->cur_tsecr = *(ptr + 2);
8378	}
8379}
8380
8381static int verify_l3_l4_lro_capable(struct lro *l_lro, struct iphdr *ip,
8382				    struct tcphdr *tcp, u32 tcp_pyld_len)
8383{
8384	u8 *ptr;
8385
8386	DBG_PRINT(INFO_DBG, "%s: Been here...\n", __func__);
8387
8388	if (!tcp_pyld_len) {
8389		/* Runt frame or a pure ack */
8390		return -1;
8391	}
8392
8393	if (ip->ihl != 5) /* IP has options */
8394		return -1;
8395
8396	/* If we see CE codepoint in IP header, packet is not mergeable */
8397	if (INET_ECN_is_ce(ipv4_get_dsfield(ip)))
8398		return -1;
8399
8400	/* If we see ECE or CWR flags in TCP header, packet is not mergeable */
8401	if (tcp->urg || tcp->psh || tcp->rst ||
8402	    tcp->syn || tcp->fin ||
8403	    tcp->ece || tcp->cwr || !tcp->ack) {
8404		/*
8405		 * Currently recognize only the ack control word and
8406		 * any other control field being set would result in
8407		 * flushing the LRO session
8408		 */
8409		return -1;
8410	}
8411
8412	/*
8413	 * Allow only one TCP timestamp option. Don't aggregate if
8414	 * any other options are detected.
8415	 */
8416	if (tcp->doff != 5 && tcp->doff != 8)
8417		return -1;
8418
8419	if (tcp->doff == 8) {
8420		ptr = (u8 *)(tcp + 1);
8421		while (*ptr == TCPOPT_NOP)
8422			ptr++;
8423		if (*ptr != TCPOPT_TIMESTAMP || *(ptr+1) != TCPOLEN_TIMESTAMP)
8424			return -1;
8425
8426		/* Ensure timestamp value increases monotonically */
8427		if (l_lro)
8428			if (l_lro->cur_tsval > ntohl(*((__be32 *)(ptr+2))))
8429				return -1;
8430
8431		/* timestamp echo reply should be non-zero */
8432		if (*((__be32 *)(ptr+6)) == 0)
8433			return -1;
8434	}
8435
8436	return 0;
8437}
8438
8439static int s2io_club_tcp_session(struct ring_info *ring_data, u8 *buffer,
8440				 u8 **tcp, u32 *tcp_len, struct lro **lro,
8441				 struct RxD_t *rxdp, struct s2io_nic *sp)
8442{
8443	struct iphdr *ip;
8444	struct tcphdr *tcph;
8445	int ret = 0, i;
8446	u16 vlan_tag = 0;
8447	struct swStat *swstats = &sp->mac_control.stats_info->sw_stat;
8448
8449	ret = check_L2_lro_capable(buffer, &ip, (struct tcphdr **)tcp,
8450				   rxdp, sp);
8451	if (ret)
8452		return ret;
8453
8454	DBG_PRINT(INFO_DBG, "IP Saddr: %x Daddr: %x\n", ip->saddr, ip->daddr);
8455
8456	vlan_tag = RXD_GET_VLAN_TAG(rxdp->Control_2);
8457	tcph = (struct tcphdr *)*tcp;
8458	*tcp_len = get_l4_pyld_length(ip, tcph);
8459	for (i = 0; i < MAX_LRO_SESSIONS; i++) {
8460		struct lro *l_lro = &ring_data->lro0_n[i];
8461		if (l_lro->in_use) {
8462			if (check_for_socket_match(l_lro, ip, tcph))
8463				continue;
8464			/* Sock pair matched */
8465			*lro = l_lro;
8466
8467			if ((*lro)->tcp_next_seq != ntohl(tcph->seq)) {
8468				DBG_PRINT(INFO_DBG, "%s: Out of sequence. "
8469					  "expected 0x%x, actual 0x%x\n",
8470					  __func__,
8471					  (*lro)->tcp_next_seq,
8472					  ntohl(tcph->seq));
8473
8474				swstats->outof_sequence_pkts++;
8475				ret = 2;
8476				break;
8477			}
8478
8479			if (!verify_l3_l4_lro_capable(l_lro, ip, tcph,
8480						      *tcp_len))
8481				ret = 1; /* Aggregate */
8482			else
8483				ret = 2; /* Flush both */
8484			break;
8485		}
8486	}
8487
8488	if (ret == 0) {
8489		/* Before searching for available LRO objects,
8490		 * check if the pkt is L3/L4 aggregatable. If not
8491		 * don't create new LRO session. Just send this
8492		 * packet up.
8493		 */
8494		if (verify_l3_l4_lro_capable(NULL, ip, tcph, *tcp_len))
8495			return 5;
8496
8497		for (i = 0; i < MAX_LRO_SESSIONS; i++) {
8498			struct lro *l_lro = &ring_data->lro0_n[i];
8499			if (!(l_lro->in_use)) {
8500				*lro = l_lro;
8501				ret = 3; /* Begin anew */
8502				break;
8503			}
8504		}
8505	}
8506
8507	if (ret == 0) { /* sessions exceeded */
8508		DBG_PRINT(INFO_DBG, "%s: All LRO sessions already in use\n",
8509			  __func__);
8510		*lro = NULL;
8511		return ret;
8512	}
8513
8514	switch (ret) {
8515	case 3:
8516		initiate_new_session(*lro, buffer, ip, tcph, *tcp_len,
8517				     vlan_tag);
8518		break;
8519	case 2:
8520		update_L3L4_header(sp, *lro);
8521		break;
8522	case 1:
8523		aggregate_new_rx(*lro, ip, tcph, *tcp_len);
8524		if ((*lro)->sg_num == sp->lro_max_aggr_per_sess) {
8525			update_L3L4_header(sp, *lro);
8526			ret = 4; /* Flush the LRO */
8527		}
8528		break;
8529	default:
8530		DBG_PRINT(ERR_DBG, "%s: Don't know, can't say!!\n", __func__);
8531		break;
8532	}
8533
8534	return ret;
8535}
8536
8537static void clear_lro_session(struct lro *lro)
8538{
8539	static u16 lro_struct_size = sizeof(struct lro);
8540
8541	memset(lro, 0, lro_struct_size);
8542}
8543
8544static void queue_rx_frame(struct sk_buff *skb, u16 vlan_tag)
8545{
8546	struct net_device *dev = skb->dev;
8547	struct s2io_nic *sp = netdev_priv(dev);
8548
8549	skb->protocol = eth_type_trans(skb, dev);
8550	if (vlan_tag && sp->vlan_strip_flag)
8551		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
8552	if (sp->config.napi)
8553		netif_receive_skb(skb);
8554	else
8555		netif_rx(skb);
8556}
8557
8558static void lro_append_pkt(struct s2io_nic *sp, struct lro *lro,
8559			   struct sk_buff *skb, u32 tcp_len)
8560{
8561	struct sk_buff *first = lro->parent;
8562	struct swStat *swstats = &sp->mac_control.stats_info->sw_stat;
8563
8564	first->len += tcp_len;
8565	first->data_len = lro->frags_len;
8566	skb_pull(skb, (skb->len - tcp_len));
8567	if (skb_shinfo(first)->frag_list)
8568		lro->last_frag->next = skb;
8569	else
8570		skb_shinfo(first)->frag_list = skb;
8571	first->truesize += skb->truesize;
8572	lro->last_frag = skb;
8573	swstats->clubbed_frms_cnt++;
8574}
8575
8576/**
8577 * s2io_io_error_detected - called when PCI error is detected
8578 * @pdev: Pointer to PCI device
8579 * @state: The current pci connection state
8580 *
8581 * This function is called after a PCI bus error affecting
8582 * this device has been detected.
8583 */
8584static pci_ers_result_t s2io_io_error_detected(struct pci_dev *pdev,
8585					       pci_channel_state_t state)
8586{
8587	struct net_device *netdev = pci_get_drvdata(pdev);
8588	struct s2io_nic *sp = netdev_priv(netdev);
8589
8590	netif_device_detach(netdev);
8591
8592	if (state == pci_channel_io_perm_failure)
8593		return PCI_ERS_RESULT_DISCONNECT;
8594
8595	if (netif_running(netdev)) {
8596		/* Bring down the card, while avoiding PCI I/O */
8597		do_s2io_card_down(sp, 0);
8598	}
8599	pci_disable_device(pdev);
8600
8601	return PCI_ERS_RESULT_NEED_RESET;
8602}
8603
8604/**
8605 * s2io_io_slot_reset - called after the pci bus has been reset.
8606 * @pdev: Pointer to PCI device
8607 *
8608 * Restart the card from scratch, as if from a cold-boot.
8609 * At this point, the card has exprienced a hard reset,
8610 * followed by fixups by BIOS, and has its config space
8611 * set up identically to what it was at cold boot.
8612 */
8613static pci_ers_result_t s2io_io_slot_reset(struct pci_dev *pdev)
8614{
8615	struct net_device *netdev = pci_get_drvdata(pdev);
8616	struct s2io_nic *sp = netdev_priv(netdev);
8617
8618	if (pci_enable_device(pdev)) {
8619		pr_err("Cannot re-enable PCI device after reset.\n");
8620		return PCI_ERS_RESULT_DISCONNECT;
8621	}
8622
8623	pci_set_master(pdev);
8624	s2io_reset(sp);
8625
8626	return PCI_ERS_RESULT_RECOVERED;
8627}
8628
8629/**
8630 * s2io_io_resume - called when traffic can start flowing again.
8631 * @pdev: Pointer to PCI device
8632 *
8633 * This callback is called when the error recovery driver tells
8634 * us that its OK to resume normal operation.
8635 */
8636static void s2io_io_resume(struct pci_dev *pdev)
8637{
8638	struct net_device *netdev = pci_get_drvdata(pdev);
8639	struct s2io_nic *sp = netdev_priv(netdev);
8640
8641	if (netif_running(netdev)) {
8642		if (s2io_card_up(sp)) {
8643			pr_err("Can't bring device back up after reset.\n");
8644			return;
8645		}
8646
8647		if (s2io_set_mac_addr(netdev, netdev->dev_addr) == FAILURE) {
8648			s2io_card_down(sp);
8649			pr_err("Can't restore mac addr after reset.\n");
8650			return;
8651		}
8652	}
8653
8654	netif_device_attach(netdev);
8655	netif_tx_wake_all_queues(netdev);
8656}
8657