xref: /freebsd/sys/dev/igc/if_igc.c (revision cab6a39d7b343596a5823e65c0f7b426551ec22d)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2016 Nicole Graziano <nicole@nextbsd.org>
5  * All rights reserved.
6  * Copyright (c) 2021 Rubicon Communications, LLC (Netgate)
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include "if_igc.h"
34 #include <sys/sbuf.h>
35 #include <machine/_inttypes.h>
36 
37 #ifdef RSS
38 #include <net/rss_config.h>
39 #include <netinet/in_rss.h>
40 #endif
41 
42 /*********************************************************************
43  *  PCI Device ID Table
44  *
45  *  Used by probe to select devices to load on
46  *  Last entry must be all 0s
47  *
48  *  { Vendor ID, Device ID, String }
49  *********************************************************************/
50 
51 static pci_vendor_info_t igc_vendor_info_array[] =
52 {
53 	/* Intel(R) PRO/1000 Network Connection - igc */
54 	PVID(0x8086, IGC_DEV_ID_I225_LM, "Intel(R) Ethernet Controller I225-LM"),
55 	PVID(0x8086, IGC_DEV_ID_I225_V, "Intel(R) Ethernet Controller I225-V"),
56 	PVID(0x8086, IGC_DEV_ID_I225_K, "Intel(R) Ethernet Controller I225-K"),
57 	PVID(0x8086, IGC_DEV_ID_I225_I, "Intel(R) Ethernet Controller I225-I"),
58 	PVID(0x8086, IGC_DEV_ID_I220_V, "Intel(R) Ethernet Controller I220-V"),
59 	PVID(0x8086, IGC_DEV_ID_I225_K2, "Intel(R) Ethernet Controller I225-K(2)"),
60 	PVID(0x8086, IGC_DEV_ID_I225_LMVP, "Intel(R) Ethernet Controller I225-LMvP(2)"),
61 	PVID(0x8086, IGC_DEV_ID_I226_K, "Intel(R) Ethernet Controller I226-K"),
62 	PVID(0x8086, IGC_DEV_ID_I225_IT, "Intel(R) Ethernet Controller I225-IT(2)"),
63 	PVID(0x8086, IGC_DEV_ID_I226_LM, "Intel(R) Ethernet Controller I226-LM"),
64 	PVID(0x8086, IGC_DEV_ID_I226_V, "Intel(R) Ethernet Controller I226-V"),
65 	PVID(0x8086, IGC_DEV_ID_I226_IT, "Intel(R) Ethernet Controller I226-IT"),
66 	PVID(0x8086, IGC_DEV_ID_I221_V, "Intel(R) Ethernet Controller I221-V"),
67 	PVID(0x8086, IGC_DEV_ID_I226_BLANK_NVM, "Intel(R) Ethernet Controller I226(blankNVM)"),
68 	PVID(0x8086, IGC_DEV_ID_I225_BLANK_NVM, "Intel(R) Ethernet Controller I225(blankNVM)"),
69 	/* required last entry */
70 	PVID_END
71 };
72 
73 /*********************************************************************
74  *  Function prototypes
75  *********************************************************************/
76 static void	*igc_register(device_t dev);
77 static int	igc_if_attach_pre(if_ctx_t ctx);
78 static int	igc_if_attach_post(if_ctx_t ctx);
79 static int	igc_if_detach(if_ctx_t ctx);
80 static int	igc_if_shutdown(if_ctx_t ctx);
81 static int	igc_if_suspend(if_ctx_t ctx);
82 static int	igc_if_resume(if_ctx_t ctx);
83 
84 static int	igc_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
85 static int	igc_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
86 static void	igc_if_queues_free(if_ctx_t ctx);
87 
88 static uint64_t	igc_if_get_counter(if_ctx_t, ift_counter);
89 static void	igc_if_init(if_ctx_t ctx);
90 static void	igc_if_stop(if_ctx_t ctx);
91 static void	igc_if_media_status(if_ctx_t, struct ifmediareq *);
92 static int	igc_if_media_change(if_ctx_t ctx);
93 static int	igc_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
94 static void	igc_if_timer(if_ctx_t ctx, uint16_t qid);
95 static void	igc_if_vlan_register(if_ctx_t ctx, u16 vtag);
96 static void	igc_if_vlan_unregister(if_ctx_t ctx, u16 vtag);
97 static void	igc_if_watchdog_reset(if_ctx_t ctx);
98 static bool	igc_if_needs_restart(if_ctx_t ctx, enum iflib_restart_event event);
99 
100 static void	igc_identify_hardware(if_ctx_t ctx);
101 static int	igc_allocate_pci_resources(if_ctx_t ctx);
102 static void	igc_free_pci_resources(if_ctx_t ctx);
103 static void	igc_reset(if_ctx_t ctx);
104 static int	igc_setup_interface(if_ctx_t ctx);
105 static int	igc_setup_msix(if_ctx_t ctx);
106 
107 static void	igc_initialize_transmit_unit(if_ctx_t ctx);
108 static void	igc_initialize_receive_unit(if_ctx_t ctx);
109 
110 static void	igc_if_intr_enable(if_ctx_t ctx);
111 static void	igc_if_intr_disable(if_ctx_t ctx);
112 static int	igc_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
113 static int	igc_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
114 static void	igc_if_multi_set(if_ctx_t ctx);
115 static void	igc_if_update_admin_status(if_ctx_t ctx);
116 static void	igc_if_debug(if_ctx_t ctx);
117 static void	igc_update_stats_counters(struct igc_adapter *);
118 static void	igc_add_hw_stats(struct igc_adapter *adapter);
119 static int	igc_if_set_promisc(if_ctx_t ctx, int flags);
120 static void	igc_setup_vlan_hw_support(struct igc_adapter *);
121 static int	igc_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
122 static void	igc_print_nvm_info(struct igc_adapter *);
123 static int	igc_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
124 static int	igc_get_rs(SYSCTL_HANDLER_ARGS);
125 static void	igc_print_debug_info(struct igc_adapter *);
126 static int 	igc_is_valid_ether_addr(u8 *);
127 static int	igc_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
128 static void	igc_add_int_delay_sysctl(struct igc_adapter *, const char *,
129 		    const char *, struct igc_int_delay_info *, int, int);
130 /* Management and WOL Support */
131 static void	igc_get_hw_control(struct igc_adapter *);
132 static void	igc_release_hw_control(struct igc_adapter *);
133 static void	igc_get_wakeup(if_ctx_t ctx);
134 static void	igc_enable_wakeup(if_ctx_t ctx);
135 
136 int		igc_intr(void *arg);
137 
138 /* MSI-X handlers */
139 static int	igc_if_msix_intr_assign(if_ctx_t, int);
140 static int	igc_msix_link(void *);
141 static void	igc_handle_link(void *context);
142 
143 static int	igc_set_flowcntl(SYSCTL_HANDLER_ARGS);
144 static int	igc_sysctl_eee(SYSCTL_HANDLER_ARGS);
145 
146 static int	igc_get_regs(SYSCTL_HANDLER_ARGS);
147 
148 static void	igc_configure_queues(struct igc_adapter *adapter);
149 
150 
151 /*********************************************************************
152  *  FreeBSD Device Interface Entry Points
153  *********************************************************************/
154 static device_method_t igc_methods[] = {
155 	/* Device interface */
156 	DEVMETHOD(device_register, igc_register),
157 	DEVMETHOD(device_probe, iflib_device_probe),
158 	DEVMETHOD(device_attach, iflib_device_attach),
159 	DEVMETHOD(device_detach, iflib_device_detach),
160 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
161 	DEVMETHOD(device_suspend, iflib_device_suspend),
162 	DEVMETHOD(device_resume, iflib_device_resume),
163 	DEVMETHOD_END
164 };
165 
166 static driver_t igc_driver = {
167 	"igc", igc_methods, sizeof(struct igc_adapter),
168 };
169 
170 static devclass_t igc_devclass;
171 DRIVER_MODULE(igc, pci, igc_driver, igc_devclass, 0, 0);
172 
173 MODULE_DEPEND(igc, pci, 1, 1, 1);
174 MODULE_DEPEND(igc, ether, 1, 1, 1);
175 MODULE_DEPEND(igc, iflib, 1, 1, 1);
176 
177 IFLIB_PNP_INFO(pci, igc, igc_vendor_info_array);
178 
179 static device_method_t igc_if_methods[] = {
180 	DEVMETHOD(ifdi_attach_pre, igc_if_attach_pre),
181 	DEVMETHOD(ifdi_attach_post, igc_if_attach_post),
182 	DEVMETHOD(ifdi_detach, igc_if_detach),
183 	DEVMETHOD(ifdi_shutdown, igc_if_shutdown),
184 	DEVMETHOD(ifdi_suspend, igc_if_suspend),
185 	DEVMETHOD(ifdi_resume, igc_if_resume),
186 	DEVMETHOD(ifdi_init, igc_if_init),
187 	DEVMETHOD(ifdi_stop, igc_if_stop),
188 	DEVMETHOD(ifdi_msix_intr_assign, igc_if_msix_intr_assign),
189 	DEVMETHOD(ifdi_intr_enable, igc_if_intr_enable),
190 	DEVMETHOD(ifdi_intr_disable, igc_if_intr_disable),
191 	DEVMETHOD(ifdi_tx_queues_alloc, igc_if_tx_queues_alloc),
192 	DEVMETHOD(ifdi_rx_queues_alloc, igc_if_rx_queues_alloc),
193 	DEVMETHOD(ifdi_queues_free, igc_if_queues_free),
194 	DEVMETHOD(ifdi_update_admin_status, igc_if_update_admin_status),
195 	DEVMETHOD(ifdi_multi_set, igc_if_multi_set),
196 	DEVMETHOD(ifdi_media_status, igc_if_media_status),
197 	DEVMETHOD(ifdi_media_change, igc_if_media_change),
198 	DEVMETHOD(ifdi_mtu_set, igc_if_mtu_set),
199 	DEVMETHOD(ifdi_promisc_set, igc_if_set_promisc),
200 	DEVMETHOD(ifdi_timer, igc_if_timer),
201 	DEVMETHOD(ifdi_watchdog_reset, igc_if_watchdog_reset),
202 	DEVMETHOD(ifdi_vlan_register, igc_if_vlan_register),
203 	DEVMETHOD(ifdi_vlan_unregister, igc_if_vlan_unregister),
204 	DEVMETHOD(ifdi_get_counter, igc_if_get_counter),
205 	DEVMETHOD(ifdi_rx_queue_intr_enable, igc_if_rx_queue_intr_enable),
206 	DEVMETHOD(ifdi_tx_queue_intr_enable, igc_if_tx_queue_intr_enable),
207 	DEVMETHOD(ifdi_debug, igc_if_debug),
208 	DEVMETHOD(ifdi_needs_restart, igc_if_needs_restart),
209 	DEVMETHOD_END
210 };
211 
212 static driver_t igc_if_driver = {
213 	"igc_if", igc_if_methods, sizeof(struct igc_adapter)
214 };
215 
216 /*********************************************************************
217  *  Tunable default values.
218  *********************************************************************/
219 
220 #define IGC_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
221 #define IGC_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
222 
223 #define MAX_INTS_PER_SEC	8000
224 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
225 
226 /* Allow common code without TSO */
227 #ifndef CSUM_TSO
228 #define CSUM_TSO	0
229 #endif
230 
231 static SYSCTL_NODE(_hw, OID_AUTO, igc, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
232     "igc driver parameters");
233 
234 static int igc_disable_crc_stripping = 0;
235 SYSCTL_INT(_hw_igc, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
236     &igc_disable_crc_stripping, 0, "Disable CRC Stripping");
237 
238 static int igc_tx_int_delay_dflt = IGC_TICKS_TO_USECS(IGC_TIDV_VAL);
239 static int igc_rx_int_delay_dflt = IGC_TICKS_TO_USECS(IGC_RDTR_VAL);
240 SYSCTL_INT(_hw_igc, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &igc_tx_int_delay_dflt,
241     0, "Default transmit interrupt delay in usecs");
242 SYSCTL_INT(_hw_igc, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &igc_rx_int_delay_dflt,
243     0, "Default receive interrupt delay in usecs");
244 
245 static int igc_tx_abs_int_delay_dflt = IGC_TICKS_TO_USECS(IGC_TADV_VAL);
246 static int igc_rx_abs_int_delay_dflt = IGC_TICKS_TO_USECS(IGC_RADV_VAL);
247 SYSCTL_INT(_hw_igc, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
248     &igc_tx_abs_int_delay_dflt, 0,
249     "Default transmit interrupt delay limit in usecs");
250 SYSCTL_INT(_hw_igc, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
251     &igc_rx_abs_int_delay_dflt, 0,
252     "Default receive interrupt delay limit in usecs");
253 
254 static int igc_smart_pwr_down = false;
255 SYSCTL_INT(_hw_igc, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &igc_smart_pwr_down,
256     0, "Set to true to leave smart power down enabled on newer adapters");
257 
258 /* Controls whether promiscuous also shows bad packets */
259 static int igc_debug_sbp = true;
260 SYSCTL_INT(_hw_igc, OID_AUTO, sbp, CTLFLAG_RDTUN, &igc_debug_sbp, 0,
261     "Show bad packets in promiscuous mode");
262 
263 /* How many packets rxeof tries to clean at a time */
264 static int igc_rx_process_limit = 100;
265 SYSCTL_INT(_hw_igc, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
266     &igc_rx_process_limit, 0,
267     "Maximum number of received packets to process "
268     "at a time, -1 means unlimited");
269 
270 /* Energy efficient ethernet - default to OFF */
271 static int igc_eee_setting = 1;
272 SYSCTL_INT(_hw_igc, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &igc_eee_setting, 0,
273     "Enable Energy Efficient Ethernet");
274 
275 /*
276 ** Tuneable Interrupt rate
277 */
278 static int igc_max_interrupt_rate = 8000;
279 SYSCTL_INT(_hw_igc, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
280     &igc_max_interrupt_rate, 0, "Maximum interrupts per second");
281 
282 extern struct if_txrx igc_txrx;
283 
284 static struct if_shared_ctx igc_sctx_init = {
285 	.isc_magic = IFLIB_MAGIC,
286 	.isc_q_align = PAGE_SIZE,
287 	.isc_tx_maxsize = IGC_TSO_SIZE + sizeof(struct ether_vlan_header),
288 	.isc_tx_maxsegsize = PAGE_SIZE,
289 	.isc_tso_maxsize = IGC_TSO_SIZE + sizeof(struct ether_vlan_header),
290 	.isc_tso_maxsegsize = IGC_TSO_SEG_SIZE,
291 	.isc_rx_maxsize = MAX_JUMBO_FRAME_SIZE,
292 	.isc_rx_nsegments = 1,
293 	.isc_rx_maxsegsize = MJUM9BYTES,
294 	.isc_nfl = 1,
295 	.isc_nrxqs = 1,
296 	.isc_ntxqs = 1,
297 	.isc_admin_intrcnt = 1,
298 	.isc_vendor_info = igc_vendor_info_array,
299 	.isc_driver_version = "1",
300 	.isc_driver = &igc_if_driver,
301 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
302 
303 	.isc_nrxd_min = {IGC_MIN_RXD},
304 	.isc_ntxd_min = {IGC_MIN_TXD},
305 	.isc_nrxd_max = {IGC_MAX_RXD},
306 	.isc_ntxd_max = {IGC_MAX_TXD},
307 	.isc_nrxd_default = {IGC_DEFAULT_RXD},
308 	.isc_ntxd_default = {IGC_DEFAULT_TXD},
309 };
310 
311 /*****************************************************************
312  *
313  * Dump Registers
314  *
315  ****************************************************************/
316 #define IGC_REGS_LEN 739
317 
318 static int igc_get_regs(SYSCTL_HANDLER_ARGS)
319 {
320 	struct igc_adapter *adapter = (struct igc_adapter *)arg1;
321 	struct igc_hw *hw = &adapter->hw;
322 	struct sbuf *sb;
323 	u32 *regs_buff;
324 	int rc;
325 
326 	regs_buff = malloc(sizeof(u32) * IGC_REGS_LEN, M_DEVBUF, M_WAITOK);
327 	memset(regs_buff, 0, IGC_REGS_LEN * sizeof(u32));
328 
329 	rc = sysctl_wire_old_buffer(req, 0);
330 	MPASS(rc == 0);
331 	if (rc != 0) {
332 		free(regs_buff, M_DEVBUF);
333 		return (rc);
334 	}
335 
336 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
337 	MPASS(sb != NULL);
338 	if (sb == NULL) {
339 		free(regs_buff, M_DEVBUF);
340 		return (ENOMEM);
341 	}
342 
343 	/* General Registers */
344 	regs_buff[0] = IGC_READ_REG(hw, IGC_CTRL);
345 	regs_buff[1] = IGC_READ_REG(hw, IGC_STATUS);
346 	regs_buff[2] = IGC_READ_REG(hw, IGC_CTRL_EXT);
347 	regs_buff[3] = IGC_READ_REG(hw, IGC_ICR);
348 	regs_buff[4] = IGC_READ_REG(hw, IGC_RCTL);
349 	regs_buff[5] = IGC_READ_REG(hw, IGC_RDLEN(0));
350 	regs_buff[6] = IGC_READ_REG(hw, IGC_RDH(0));
351 	regs_buff[7] = IGC_READ_REG(hw, IGC_RDT(0));
352 	regs_buff[8] = IGC_READ_REG(hw, IGC_RXDCTL(0));
353 	regs_buff[9] = IGC_READ_REG(hw, IGC_RDBAL(0));
354 	regs_buff[10] = IGC_READ_REG(hw, IGC_RDBAH(0));
355 	regs_buff[11] = IGC_READ_REG(hw, IGC_TCTL);
356 	regs_buff[12] = IGC_READ_REG(hw, IGC_TDBAL(0));
357 	regs_buff[13] = IGC_READ_REG(hw, IGC_TDBAH(0));
358 	regs_buff[14] = IGC_READ_REG(hw, IGC_TDLEN(0));
359 	regs_buff[15] = IGC_READ_REG(hw, IGC_TDH(0));
360 	regs_buff[16] = IGC_READ_REG(hw, IGC_TDT(0));
361 	regs_buff[17] = IGC_READ_REG(hw, IGC_TXDCTL(0));
362 
363 	sbuf_printf(sb, "General Registers\n");
364 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
365 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
366 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
367 
368 	sbuf_printf(sb, "Interrupt Registers\n");
369 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
370 
371 	sbuf_printf(sb, "RX Registers\n");
372 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
373 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
374 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
375 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
376 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
377 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
378 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
379 
380 	sbuf_printf(sb, "TX Registers\n");
381 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
382 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
383 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
384 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
385 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
386 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
387 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
388 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
389 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
390 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
391 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
392 
393 	free(regs_buff, M_DEVBUF);
394 
395 #ifdef DUMP_DESCS
396 	{
397 		if_softc_ctx_t scctx = adapter->shared;
398 		struct rx_ring *rxr = &rx_que->rxr;
399 		struct tx_ring *txr = &tx_que->txr;
400 		int ntxd = scctx->isc_ntxd[0];
401 		int nrxd = scctx->isc_nrxd[0];
402 		int j;
403 
404 	for (j = 0; j < nrxd; j++) {
405 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
406 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
407 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
408 	}
409 
410 	for (j = 0; j < min(ntxd, 256); j++) {
411 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
412 
413 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
414 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
415 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & IGC_TXD_STAT_DD : 0);
416 
417 	}
418 	}
419 #endif
420 
421 	rc = sbuf_finish(sb);
422 	sbuf_delete(sb);
423 	return(rc);
424 }
425 
426 static void *
427 igc_register(device_t dev)
428 {
429 	return (&igc_sctx_init);
430 }
431 
432 static int
433 igc_set_num_queues(if_ctx_t ctx)
434 {
435 	int maxqueues;
436 
437 	maxqueues = 4;
438 
439 	return (maxqueues);
440 }
441 
442 #define	IGC_CAPS							\
443     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
444     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
445     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
446     IFCAP_TSO6
447 
448 /*********************************************************************
449  *  Device initialization routine
450  *
451  *  The attach entry point is called when the driver is being loaded.
452  *  This routine identifies the type of hardware, allocates all resources
453  *  and initializes the hardware.
454  *
455  *  return 0 on success, positive on failure
456  *********************************************************************/
457 static int
458 igc_if_attach_pre(if_ctx_t ctx)
459 {
460 	struct igc_adapter *adapter;
461 	if_softc_ctx_t scctx;
462 	device_t dev;
463 	struct igc_hw *hw;
464 	int error = 0;
465 
466 	INIT_DEBUGOUT("igc_if_attach_pre: begin");
467 	dev = iflib_get_dev(ctx);
468 	adapter = iflib_get_softc(ctx);
469 
470 	adapter->ctx = adapter->osdep.ctx = ctx;
471 	adapter->dev = adapter->osdep.dev = dev;
472 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
473 	adapter->media = iflib_get_media(ctx);
474 	hw = &adapter->hw;
475 
476 	adapter->tx_process_limit = scctx->isc_ntxd[0];
477 
478 	/* SYSCTL stuff */
479 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
480 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
481 	    OID_AUTO, "nvm", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
482 	    adapter, 0, igc_sysctl_nvm_info, "I", "NVM Information");
483 
484 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
485 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
486 	    OID_AUTO, "debug", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
487 	    adapter, 0, igc_sysctl_debug_info, "I", "Debug Information");
488 
489 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
490 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
491 	    OID_AUTO, "fc", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
492 	    adapter, 0, igc_set_flowcntl, "I", "Flow Control");
493 
494 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
495 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
496 	    OID_AUTO, "reg_dump",
497 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter, 0,
498 	    igc_get_regs, "A", "Dump Registers");
499 
500 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
501 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
502 	    OID_AUTO, "rs_dump",
503 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, adapter, 0,
504 	    igc_get_rs, "I", "Dump RS indexes");
505 
506 	/* Determine hardware and mac info */
507 	igc_identify_hardware(ctx);
508 
509 	scctx->isc_tx_nsegments = IGC_MAX_SCATTER;
510 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = igc_set_num_queues(ctx);
511 	if (bootverbose)
512 		device_printf(dev, "attach_pre capping queues at %d\n",
513 		    scctx->isc_ntxqsets_max);
514 
515 	scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union igc_adv_tx_desc), IGC_DBA_ALIGN);
516 	scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union igc_adv_rx_desc), IGC_DBA_ALIGN);
517 	scctx->isc_txd_size[0] = sizeof(union igc_adv_tx_desc);
518 	scctx->isc_rxd_size[0] = sizeof(union igc_adv_rx_desc);
519 	scctx->isc_txrx = &igc_txrx;
520 	scctx->isc_tx_tso_segments_max = IGC_MAX_SCATTER;
521 	scctx->isc_tx_tso_size_max = IGC_TSO_SIZE;
522 	scctx->isc_tx_tso_segsize_max = IGC_TSO_SEG_SIZE;
523 	scctx->isc_capabilities = scctx->isc_capenable = IGC_CAPS;
524 	scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
525 		CSUM_IP6_TCP | CSUM_IP6_UDP | CSUM_SCTP | CSUM_IP6_SCTP;
526 
527 	/*
528 	** Some new devices, as with ixgbe, now may
529 	** use a different BAR, so we need to keep
530 	** track of which is used.
531 	*/
532 	scctx->isc_msix_bar = PCIR_BAR(IGC_MSIX_BAR);
533 	if (pci_read_config(dev, scctx->isc_msix_bar, 4) == 0)
534 		scctx->isc_msix_bar += 4;
535 
536 	/* Setup PCI resources */
537 	if (igc_allocate_pci_resources(ctx)) {
538 		device_printf(dev, "Allocation of PCI resources failed\n");
539 		error = ENXIO;
540 		goto err_pci;
541 	}
542 
543 	/* Do Shared Code initialization */
544 	error = igc_setup_init_funcs(hw, true);
545 	if (error) {
546 		device_printf(dev, "Setup of Shared code failed, error %d\n",
547 		    error);
548 		error = ENXIO;
549 		goto err_pci;
550 	}
551 
552 	igc_setup_msix(ctx);
553 	igc_get_bus_info(hw);
554 
555 	/* Set up some sysctls for the tunable interrupt delays */
556 	igc_add_int_delay_sysctl(adapter, "rx_int_delay",
557 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
558 	    IGC_REGISTER(hw, IGC_RDTR), igc_rx_int_delay_dflt);
559 	igc_add_int_delay_sysctl(adapter, "tx_int_delay",
560 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
561 	    IGC_REGISTER(hw, IGC_TIDV), igc_tx_int_delay_dflt);
562 	igc_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
563 	    "receive interrupt delay limit in usecs",
564 	    &adapter->rx_abs_int_delay,
565 	    IGC_REGISTER(hw, IGC_RADV),
566 	    igc_rx_abs_int_delay_dflt);
567 	igc_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
568 	    "transmit interrupt delay limit in usecs",
569 	    &adapter->tx_abs_int_delay,
570 	    IGC_REGISTER(hw, IGC_TADV),
571 	    igc_tx_abs_int_delay_dflt);
572 	igc_add_int_delay_sysctl(adapter, "itr",
573 	    "interrupt delay limit in usecs/4",
574 	    &adapter->tx_itr,
575 	    IGC_REGISTER(hw, IGC_ITR),
576 	    DEFAULT_ITR);
577 
578 	hw->mac.autoneg = DO_AUTO_NEG;
579 	hw->phy.autoneg_wait_to_complete = false;
580 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
581 
582 	/* Copper options */
583 	if (hw->phy.media_type == igc_media_type_copper) {
584 		hw->phy.mdix = AUTO_ALL_MODES;
585 	}
586 
587 	/*
588 	 * Set the frame limits assuming
589 	 * standard ethernet sized frames.
590 	 */
591 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
592 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
593 
594 	/* Allocate multicast array memory. */
595 	adapter->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
596 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
597 	if (adapter->mta == NULL) {
598 		device_printf(dev, "Can not allocate multicast setup array\n");
599 		error = ENOMEM;
600 		goto err_late;
601 	}
602 
603 	/* Check SOL/IDER usage */
604 	if (igc_check_reset_block(hw))
605 		device_printf(dev, "PHY reset is blocked"
606 			      " due to SOL/IDER session.\n");
607 
608 	/* Sysctl for setting Energy Efficient Ethernet */
609 	adapter->hw.dev_spec._i225.eee_disable = igc_eee_setting;
610 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
611 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
612 	    OID_AUTO, "eee_control",
613 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
614 	    adapter, 0, igc_sysctl_eee, "I",
615 	    "Disable Energy Efficient Ethernet");
616 
617 	/*
618 	** Start from a known state, this is
619 	** important in reading the nvm and
620 	** mac from that.
621 	*/
622 	igc_reset_hw(hw);
623 
624 	/* Make sure we have a good EEPROM before we read from it */
625 	if (igc_validate_nvm_checksum(hw) < 0) {
626 		/*
627 		** Some PCI-E parts fail the first check due to
628 		** the link being in sleep state, call it again,
629 		** if it fails a second time its a real issue.
630 		*/
631 		if (igc_validate_nvm_checksum(hw) < 0) {
632 			device_printf(dev,
633 			    "The EEPROM Checksum Is Not Valid\n");
634 			error = EIO;
635 			goto err_late;
636 		}
637 	}
638 
639 	/* Copy the permanent MAC address out of the EEPROM */
640 	if (igc_read_mac_addr(hw) < 0) {
641 		device_printf(dev, "EEPROM read error while reading MAC"
642 			      " address\n");
643 		error = EIO;
644 		goto err_late;
645 	}
646 
647 	if (!igc_is_valid_ether_addr(hw->mac.addr)) {
648 		device_printf(dev, "Invalid MAC address\n");
649 		error = EIO;
650 		goto err_late;
651 	}
652 
653 	/*
654 	 * Get Wake-on-Lan and Management info for later use
655 	 */
656 	igc_get_wakeup(ctx);
657 
658 	/* Enable only WOL MAGIC by default */
659 	scctx->isc_capenable &= ~IFCAP_WOL;
660 	if (adapter->wol != 0)
661 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
662 
663 	iflib_set_mac(ctx, hw->mac.addr);
664 
665 	return (0);
666 
667 err_late:
668 	igc_release_hw_control(adapter);
669 err_pci:
670 	igc_free_pci_resources(ctx);
671 	free(adapter->mta, M_DEVBUF);
672 
673 	return (error);
674 }
675 
676 static int
677 igc_if_attach_post(if_ctx_t ctx)
678 {
679 	struct igc_adapter *adapter = iflib_get_softc(ctx);
680 	struct igc_hw *hw = &adapter->hw;
681 	int error = 0;
682 
683 	/* Setup OS specific network interface */
684 	error = igc_setup_interface(ctx);
685 	if (error != 0) {
686 		goto err_late;
687 	}
688 
689 	igc_reset(ctx);
690 
691 	/* Initialize statistics */
692 	igc_update_stats_counters(adapter);
693 	hw->mac.get_link_status = true;
694 	igc_if_update_admin_status(ctx);
695 	igc_add_hw_stats(adapter);
696 
697 	/* the driver can now take control from firmware */
698 	igc_get_hw_control(adapter);
699 
700 	INIT_DEBUGOUT("igc_if_attach_post: end");
701 
702 	return (error);
703 
704 err_late:
705 	igc_release_hw_control(adapter);
706 	igc_free_pci_resources(ctx);
707 	igc_if_queues_free(ctx);
708 	free(adapter->mta, M_DEVBUF);
709 
710 	return (error);
711 }
712 
713 /*********************************************************************
714  *  Device removal routine
715  *
716  *  The detach entry point is called when the driver is being removed.
717  *  This routine stops the adapter and deallocates all the resources
718  *  that were allocated for driver operation.
719  *
720  *  return 0 on success, positive on failure
721  *********************************************************************/
722 static int
723 igc_if_detach(if_ctx_t ctx)
724 {
725 	struct igc_adapter	*adapter = iflib_get_softc(ctx);
726 
727 	INIT_DEBUGOUT("igc_if_detach: begin");
728 
729 	igc_phy_hw_reset(&adapter->hw);
730 
731 	igc_release_hw_control(adapter);
732 	igc_free_pci_resources(ctx);
733 
734 	return (0);
735 }
736 
737 /*********************************************************************
738  *
739  *  Shutdown entry point
740  *
741  **********************************************************************/
742 
743 static int
744 igc_if_shutdown(if_ctx_t ctx)
745 {
746 	return igc_if_suspend(ctx);
747 }
748 
749 /*
750  * Suspend/resume device methods.
751  */
752 static int
753 igc_if_suspend(if_ctx_t ctx)
754 {
755 	struct igc_adapter *adapter = iflib_get_softc(ctx);
756 
757 	igc_release_hw_control(adapter);
758 	igc_enable_wakeup(ctx);
759 	return (0);
760 }
761 
762 static int
763 igc_if_resume(if_ctx_t ctx)
764 {
765 	igc_if_init(ctx);
766 
767 	return(0);
768 }
769 
770 static int
771 igc_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
772 {
773 	int max_frame_size;
774 	struct igc_adapter *adapter = iflib_get_softc(ctx);
775 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
776 
777 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
778 
779 	 /* 9K Jumbo Frame size */
780 	 max_frame_size = 9234;
781 
782 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
783 		return (EINVAL);
784 	}
785 
786 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
787 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
788 	return (0);
789 }
790 
791 /*********************************************************************
792  *  Init entry point
793  *
794  *  This routine is used in two ways. It is used by the stack as
795  *  init entry point in network interface structure. It is also used
796  *  by the driver as a hw/sw initialization routine to get to a
797  *  consistent state.
798  *
799  **********************************************************************/
800 static void
801 igc_if_init(if_ctx_t ctx)
802 {
803 	struct igc_adapter *adapter = iflib_get_softc(ctx);
804 	if_softc_ctx_t scctx = adapter->shared;
805 	struct ifnet *ifp = iflib_get_ifp(ctx);
806 	struct igc_tx_queue *tx_que;
807 	int i;
808 
809 	INIT_DEBUGOUT("igc_if_init: begin");
810 
811 	/* Get the latest mac address, User can use a LAA */
812 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
813 	    ETHER_ADDR_LEN);
814 
815 	/* Put the address into the Receive Address Array */
816 	igc_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
817 
818 	/* Initialize the hardware */
819 	igc_reset(ctx);
820 	igc_if_update_admin_status(ctx);
821 
822 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
823 		struct tx_ring *txr = &tx_que->txr;
824 
825 		txr->tx_rs_cidx = txr->tx_rs_pidx;
826 
827 		/* Initialize the last processed descriptor to be the end of
828 		 * the ring, rather than the start, so that we avoid an
829 		 * off-by-one error when calculating how many descriptors are
830 		 * done in the credits_update function.
831 		 */
832 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
833 	}
834 
835 	/* Setup VLAN support, basic and offload if available */
836 	IGC_WRITE_REG(&adapter->hw, IGC_VET, ETHERTYPE_VLAN);
837 
838 	/* Prepare transmit descriptors and buffers */
839 	igc_initialize_transmit_unit(ctx);
840 
841 	/* Setup Multicast table */
842 	igc_if_multi_set(ctx);
843 
844 	adapter->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
845 	igc_initialize_receive_unit(ctx);
846 
847 	/* Use real VLAN Filter support? */
848 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
849 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
850 			/* Use real VLAN Filter support */
851 			igc_setup_vlan_hw_support(adapter);
852 		else {
853 			u32 ctrl;
854 			ctrl = IGC_READ_REG(&adapter->hw, IGC_CTRL);
855 			ctrl |= IGC_CTRL_VME;
856 			IGC_WRITE_REG(&adapter->hw, IGC_CTRL, ctrl);
857 		}
858 	}
859 
860 	/* Don't lose promiscuous settings */
861 	igc_if_set_promisc(ctx, IFF_PROMISC);
862 	igc_clear_hw_cntrs_base_generic(&adapter->hw);
863 
864 	if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
865 		igc_configure_queues(adapter);
866 
867 	/* this clears any pending interrupts */
868 	IGC_READ_REG(&adapter->hw, IGC_ICR);
869 	IGC_WRITE_REG(&adapter->hw, IGC_ICS, IGC_ICS_LSC);
870 
871 	/* the driver can now take control from firmware */
872 	igc_get_hw_control(adapter);
873 
874 	/* Set Energy Efficient Ethernet */
875 	igc_set_eee_i225(&adapter->hw, true, true, true);
876 }
877 
878 /*********************************************************************
879  *
880  *  Fast Legacy/MSI Combined Interrupt Service routine
881  *
882  *********************************************************************/
883 int
884 igc_intr(void *arg)
885 {
886 	struct igc_adapter *adapter = arg;
887 	if_ctx_t ctx = adapter->ctx;
888 	u32 reg_icr;
889 
890 	reg_icr = IGC_READ_REG(&adapter->hw, IGC_ICR);
891 
892 	/* Hot eject? */
893 	if (reg_icr == 0xffffffff)
894 		return FILTER_STRAY;
895 
896 	/* Definitely not our interrupt. */
897 	if (reg_icr == 0x0)
898 		return FILTER_STRAY;
899 
900 	if ((reg_icr & IGC_ICR_INT_ASSERTED) == 0)
901 		return FILTER_STRAY;
902 
903 	/*
904 	 * Only MSI-X interrupts have one-shot behavior by taking advantage
905 	 * of the EIAC register.  Thus, explicitly disable interrupts.  This
906 	 * also works around the MSI message reordering errata on certain
907 	 * systems.
908 	 */
909 	IFDI_INTR_DISABLE(ctx);
910 
911 	/* Link status change */
912 	if (reg_icr & (IGC_ICR_RXSEQ | IGC_ICR_LSC))
913 		igc_handle_link(ctx);
914 
915 	if (reg_icr & IGC_ICR_RXO)
916 		adapter->rx_overruns++;
917 
918 	return (FILTER_SCHEDULE_THREAD);
919 }
920 
921 static int
922 igc_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
923 {
924 	struct igc_adapter *adapter = iflib_get_softc(ctx);
925 	struct igc_rx_queue *rxq = &adapter->rx_queues[rxqid];
926 
927 	IGC_WRITE_REG(&adapter->hw, IGC_EIMS, rxq->eims);
928 	return (0);
929 }
930 
931 static int
932 igc_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
933 {
934 	struct igc_adapter *adapter = iflib_get_softc(ctx);
935 	struct igc_tx_queue *txq = &adapter->tx_queues[txqid];
936 
937 	IGC_WRITE_REG(&adapter->hw, IGC_EIMS, txq->eims);
938 	return (0);
939 }
940 
941 /*********************************************************************
942  *
943  *  MSI-X RX Interrupt Service routine
944  *
945  **********************************************************************/
946 static int
947 igc_msix_que(void *arg)
948 {
949 	struct igc_rx_queue *que = arg;
950 
951 	++que->irqs;
952 
953 	return (FILTER_SCHEDULE_THREAD);
954 }
955 
956 /*********************************************************************
957  *
958  *  MSI-X Link Fast Interrupt Service routine
959  *
960  **********************************************************************/
961 static int
962 igc_msix_link(void *arg)
963 {
964 	struct igc_adapter *adapter = arg;
965 	u32 reg_icr;
966 
967 	++adapter->link_irq;
968 	MPASS(adapter->hw.back != NULL);
969 	reg_icr = IGC_READ_REG(&adapter->hw, IGC_ICR);
970 
971 	if (reg_icr & IGC_ICR_RXO)
972 		adapter->rx_overruns++;
973 
974 	if (reg_icr & (IGC_ICR_RXSEQ | IGC_ICR_LSC)) {
975 		igc_handle_link(adapter->ctx);
976 	}
977 
978 	IGC_WRITE_REG(&adapter->hw, IGC_IMS, IGC_IMS_LSC);
979 	IGC_WRITE_REG(&adapter->hw, IGC_EIMS, adapter->link_mask);
980 
981 	return (FILTER_HANDLED);
982 }
983 
984 static void
985 igc_handle_link(void *context)
986 {
987 	if_ctx_t ctx = context;
988 	struct igc_adapter *adapter = iflib_get_softc(ctx);
989 
990 	adapter->hw.mac.get_link_status = true;
991 	iflib_admin_intr_deferred(ctx);
992 }
993 
994 /*********************************************************************
995  *
996  *  Media Ioctl callback
997  *
998  *  This routine is called whenever the user queries the status of
999  *  the interface using ifconfig.
1000  *
1001  **********************************************************************/
1002 static void
1003 igc_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1004 {
1005 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1006 
1007 	INIT_DEBUGOUT("igc_if_media_status: begin");
1008 
1009 	iflib_admin_intr_deferred(ctx);
1010 
1011 	ifmr->ifm_status = IFM_AVALID;
1012 	ifmr->ifm_active = IFM_ETHER;
1013 
1014 	if (!adapter->link_active) {
1015 		return;
1016 	}
1017 
1018 	ifmr->ifm_status |= IFM_ACTIVE;
1019 
1020 	switch (adapter->link_speed) {
1021 	case 10:
1022 		ifmr->ifm_active |= IFM_10_T;
1023 		break;
1024 	case 100:
1025 		ifmr->ifm_active |= IFM_100_TX;
1026                 break;
1027 	case 1000:
1028 		ifmr->ifm_active |= IFM_1000_T;
1029 		break;
1030 	case 2500:
1031                 ifmr->ifm_active |= IFM_2500_T;
1032                 break;
1033 	}
1034 
1035 	if (adapter->link_duplex == FULL_DUPLEX)
1036 		ifmr->ifm_active |= IFM_FDX;
1037 	else
1038 		ifmr->ifm_active |= IFM_HDX;
1039 }
1040 
1041 /*********************************************************************
1042  *
1043  *  Media Ioctl callback
1044  *
1045  *  This routine is called when the user changes speed/duplex using
1046  *  media/mediopt option with ifconfig.
1047  *
1048  **********************************************************************/
1049 static int
1050 igc_if_media_change(if_ctx_t ctx)
1051 {
1052 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1053 	struct ifmedia *ifm = iflib_get_media(ctx);
1054 
1055 	INIT_DEBUGOUT("igc_if_media_change: begin");
1056 
1057 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1058 		return (EINVAL);
1059 
1060 	adapter->hw.mac.autoneg = DO_AUTO_NEG;
1061 
1062 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1063 	case IFM_AUTO:
1064 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1065 		break;
1066         case IFM_2500_T:
1067                 adapter->hw.phy.autoneg_advertised = ADVERTISE_2500_FULL;
1068                 break;
1069 	case IFM_1000_T:
1070 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1071 		break;
1072 	case IFM_100_TX:
1073 		if ((ifm->ifm_media & IFM_GMASK) == IFM_HDX)
1074 			adapter->hw.phy.autoneg_advertised = ADVERTISE_100_HALF;
1075 		else
1076 			adapter->hw.phy.autoneg_advertised = ADVERTISE_100_FULL;
1077 		break;
1078 	case IFM_10_T:
1079 		if ((ifm->ifm_media & IFM_GMASK) == IFM_HDX)
1080 			adapter->hw.phy.autoneg_advertised = ADVERTISE_10_HALF;
1081 		else
1082 			adapter->hw.phy.autoneg_advertised = ADVERTISE_10_FULL;
1083 		break;
1084 	default:
1085 		device_printf(adapter->dev, "Unsupported media type\n");
1086 	}
1087 
1088 	igc_if_init(ctx);
1089 
1090 	return (0);
1091 }
1092 
1093 static int
1094 igc_if_set_promisc(if_ctx_t ctx, int flags)
1095 {
1096 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1097 	struct ifnet *ifp = iflib_get_ifp(ctx);
1098 	u32 reg_rctl;
1099 	int mcnt = 0;
1100 
1101 	reg_rctl = IGC_READ_REG(&adapter->hw, IGC_RCTL);
1102 	reg_rctl &= ~(IGC_RCTL_SBP | IGC_RCTL_UPE);
1103 	if (flags & IFF_ALLMULTI)
1104 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1105 	else
1106 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
1107 
1108 	/* Don't disable if in MAX groups */
1109 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1110 		reg_rctl &=  (~IGC_RCTL_MPE);
1111 	IGC_WRITE_REG(&adapter->hw, IGC_RCTL, reg_rctl);
1112 
1113 	if (flags & IFF_PROMISC) {
1114 		reg_rctl |= (IGC_RCTL_UPE | IGC_RCTL_MPE);
1115 		/* Turn this on if you want to see bad packets */
1116 		if (igc_debug_sbp)
1117 			reg_rctl |= IGC_RCTL_SBP;
1118 		IGC_WRITE_REG(&adapter->hw, IGC_RCTL, reg_rctl);
1119 	} else if (flags & IFF_ALLMULTI) {
1120 		reg_rctl |= IGC_RCTL_MPE;
1121 		reg_rctl &= ~IGC_RCTL_UPE;
1122 		IGC_WRITE_REG(&adapter->hw, IGC_RCTL, reg_rctl);
1123 	}
1124 	return (0);
1125 }
1126 
1127 static u_int
1128 igc_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int idx)
1129 {
1130 	u8 *mta = arg;
1131 
1132 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
1133 		return (0);
1134 
1135 	bcopy(LLADDR(sdl), &mta[idx * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1136 
1137 	return (1);
1138 }
1139 
1140 /*********************************************************************
1141  *  Multicast Update
1142  *
1143  *  This routine is called whenever multicast address list is updated.
1144  *
1145  **********************************************************************/
1146 
1147 static void
1148 igc_if_multi_set(if_ctx_t ctx)
1149 {
1150 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1151 	struct ifnet *ifp = iflib_get_ifp(ctx);
1152 	u8  *mta; /* Multicast array memory */
1153 	u32 reg_rctl = 0;
1154 	int mcnt = 0;
1155 
1156 	IOCTL_DEBUGOUT("igc_set_multi: begin");
1157 
1158 	mta = adapter->mta;
1159 	bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1160 
1161 	mcnt = if_foreach_llmaddr(ifp, igc_copy_maddr, mta);
1162 
1163 	reg_rctl = IGC_READ_REG(&adapter->hw, IGC_RCTL);
1164 
1165 	if (if_getflags(ifp) & IFF_PROMISC) {
1166 		reg_rctl |= (IGC_RCTL_UPE | IGC_RCTL_MPE);
1167 		/* Turn this on if you want to see bad packets */
1168 		if (igc_debug_sbp)
1169 			reg_rctl |= IGC_RCTL_SBP;
1170 	} else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
1171 	      if_getflags(ifp) & IFF_ALLMULTI) {
1172                 reg_rctl |= IGC_RCTL_MPE;
1173 		reg_rctl &= ~IGC_RCTL_UPE;
1174         } else
1175 		reg_rctl = ~(IGC_RCTL_UPE | IGC_RCTL_MPE);
1176 
1177 	IGC_WRITE_REG(&adapter->hw, IGC_RCTL, reg_rctl);
1178 
1179 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1180 		igc_update_mc_addr_list(&adapter->hw, mta, mcnt);
1181 }
1182 
1183 /*********************************************************************
1184  *  Timer routine
1185  *
1186  *  This routine schedules igc_if_update_admin_status() to check for
1187  *  link status and to gather statistics as well as to perform some
1188  *  controller-specific hardware patting.
1189  *
1190  **********************************************************************/
1191 static void
1192 igc_if_timer(if_ctx_t ctx, uint16_t qid)
1193 {
1194 
1195 	if (qid != 0)
1196 		return;
1197 
1198 	iflib_admin_intr_deferred(ctx);
1199 }
1200 
1201 static void
1202 igc_if_update_admin_status(if_ctx_t ctx)
1203 {
1204 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1205 	struct igc_hw *hw = &adapter->hw;
1206 	device_t dev = iflib_get_dev(ctx);
1207 	u32 link_check, thstat, ctrl;
1208 
1209 	link_check = thstat = ctrl = 0;
1210 	/* Get the cached link value or read phy for real */
1211 	switch (hw->phy.media_type) {
1212 	case igc_media_type_copper:
1213 		if (hw->mac.get_link_status == true) {
1214 			/* Do the work to read phy */
1215 			igc_check_for_link(hw);
1216 			link_check = !hw->mac.get_link_status;
1217 		} else
1218 			link_check = true;
1219 		break;
1220 	case igc_media_type_unknown:
1221 		igc_check_for_link(hw);
1222 		link_check = !hw->mac.get_link_status;
1223 		/* FALLTHROUGH */
1224 	default:
1225 		break;
1226 	}
1227 
1228 	/* Now check for a transition */
1229 	if (link_check && (adapter->link_active == 0)) {
1230 		igc_get_speed_and_duplex(hw, &adapter->link_speed,
1231 		    &adapter->link_duplex);
1232 		if (bootverbose)
1233 			device_printf(dev, "Link is up %d Mbps %s\n",
1234 			    adapter->link_speed,
1235 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1236 			    "Full Duplex" : "Half Duplex"));
1237 		adapter->link_active = 1;
1238 		iflib_link_state_change(ctx, LINK_STATE_UP,
1239 		    IF_Mbps(adapter->link_speed));
1240 	} else if (!link_check && (adapter->link_active == 1)) {
1241 		adapter->link_speed = 0;
1242 		adapter->link_duplex = 0;
1243 		adapter->link_active = 0;
1244 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1245 	}
1246 	igc_update_stats_counters(adapter);
1247 }
1248 
1249 static void
1250 igc_if_watchdog_reset(if_ctx_t ctx)
1251 {
1252 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1253 
1254 	/*
1255 	 * Just count the event; iflib(4) will already trigger a
1256 	 * sufficient reset of the controller.
1257 	 */
1258 	adapter->watchdog_events++;
1259 }
1260 
1261 /*********************************************************************
1262  *
1263  *  This routine disables all traffic on the adapter by issuing a
1264  *  global reset on the MAC.
1265  *
1266  **********************************************************************/
1267 static void
1268 igc_if_stop(if_ctx_t ctx)
1269 {
1270 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1271 
1272 	INIT_DEBUGOUT("igc_if_stop: begin");
1273 
1274 	igc_reset_hw(&adapter->hw);
1275 	IGC_WRITE_REG(&adapter->hw, IGC_WUC, 0);
1276 }
1277 
1278 /*********************************************************************
1279  *
1280  *  Determine hardware revision.
1281  *
1282  **********************************************************************/
1283 static void
1284 igc_identify_hardware(if_ctx_t ctx)
1285 {
1286 	device_t dev = iflib_get_dev(ctx);
1287 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1288 
1289 	/* Make sure our PCI config space has the necessary stuff set */
1290 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1291 
1292 	/* Save off the information about this board */
1293 	adapter->hw.vendor_id = pci_get_vendor(dev);
1294 	adapter->hw.device_id = pci_get_device(dev);
1295 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1296 	adapter->hw.subsystem_vendor_id =
1297 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1298 	adapter->hw.subsystem_device_id =
1299 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1300 
1301 	/* Do Shared Code Init and Setup */
1302 	if (igc_set_mac_type(&adapter->hw)) {
1303 		device_printf(dev, "Setup init failure\n");
1304 		return;
1305 	}
1306 }
1307 
1308 static int
1309 igc_allocate_pci_resources(if_ctx_t ctx)
1310 {
1311 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1312 	device_t dev = iflib_get_dev(ctx);
1313 	int rid;
1314 
1315 	rid = PCIR_BAR(0);
1316 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1317 	    &rid, RF_ACTIVE);
1318 	if (adapter->memory == NULL) {
1319 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1320 		return (ENXIO);
1321 	}
1322 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1323 	adapter->osdep.mem_bus_space_handle =
1324 	    rman_get_bushandle(adapter->memory);
1325 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1326 
1327 	adapter->hw.back = &adapter->osdep;
1328 
1329 	return (0);
1330 }
1331 
1332 /*********************************************************************
1333  *
1334  *  Set up the MSI-X Interrupt handlers
1335  *
1336  **********************************************************************/
1337 static int
1338 igc_if_msix_intr_assign(if_ctx_t ctx, int msix)
1339 {
1340 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1341 	struct igc_rx_queue *rx_que = adapter->rx_queues;
1342 	struct igc_tx_queue *tx_que = adapter->tx_queues;
1343 	int error, rid, i, vector = 0, rx_vectors;
1344 	char buf[16];
1345 
1346 	/* First set up ring resources */
1347 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1348 		rid = vector + 1;
1349 		snprintf(buf, sizeof(buf), "rxq%d", i);
1350 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, igc_msix_que, rx_que, rx_que->me, buf);
1351 		if (error) {
1352 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1353 			adapter->rx_num_queues = i + 1;
1354 			goto fail;
1355 		}
1356 
1357 		rx_que->msix =  vector;
1358 
1359 		/*
1360 		 * Set the bit to enable interrupt
1361 		 * in IGC_IMS -- bits 20 and 21
1362 		 * are for RX0 and RX1, note this has
1363 		 * NOTHING to do with the MSI-X vector
1364 		 */
1365 		rx_que->eims = 1 << vector;
1366 	}
1367 	rx_vectors = vector;
1368 
1369 	vector = 0;
1370 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
1371 		snprintf(buf, sizeof(buf), "txq%d", i);
1372 		tx_que = &adapter->tx_queues[i];
1373 		iflib_softirq_alloc_generic(ctx,
1374 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
1375 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
1376 
1377 		tx_que->msix = (vector % adapter->rx_num_queues);
1378 
1379 		/*
1380 		 * Set the bit to enable interrupt
1381 		 * in IGC_IMS -- bits 22 and 23
1382 		 * are for TX0 and TX1, note this has
1383 		 * NOTHING to do with the MSI-X vector
1384 		 */
1385 		tx_que->eims = 1 << i;
1386 	}
1387 
1388 	/* Link interrupt */
1389 	rid = rx_vectors + 1;
1390 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, igc_msix_link, adapter, 0, "aq");
1391 
1392 	if (error) {
1393 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
1394 		goto fail;
1395 	}
1396 	adapter->linkvec = rx_vectors;
1397 	return (0);
1398 fail:
1399 	iflib_irq_free(ctx, &adapter->irq);
1400 	rx_que = adapter->rx_queues;
1401 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
1402 		iflib_irq_free(ctx, &rx_que->que_irq);
1403 	return (error);
1404 }
1405 
1406 static void
1407 igc_configure_queues(struct igc_adapter *adapter)
1408 {
1409 	struct igc_hw *hw = &adapter->hw;
1410 	struct igc_rx_queue *rx_que;
1411 	struct igc_tx_queue *tx_que;
1412 	u32 ivar = 0, newitr = 0;
1413 
1414 	/* First turn on RSS capability */
1415 	IGC_WRITE_REG(hw, IGC_GPIE,
1416 	    IGC_GPIE_MSIX_MODE | IGC_GPIE_EIAME | IGC_GPIE_PBA |
1417 	    IGC_GPIE_NSICR);
1418 
1419 	/* Turn on MSI-X */
1420 	/* RX entries */
1421 	for (int i = 0; i < adapter->rx_num_queues; i++) {
1422 		u32 index = i >> 1;
1423 		ivar = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, index);
1424 		rx_que = &adapter->rx_queues[i];
1425 		if (i & 1) {
1426 			ivar &= 0xFF00FFFF;
1427 			ivar |= (rx_que->msix | IGC_IVAR_VALID) << 16;
1428 		} else {
1429 			ivar &= 0xFFFFFF00;
1430 			ivar |= rx_que->msix | IGC_IVAR_VALID;
1431 		}
1432 		IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, index, ivar);
1433 	}
1434 	/* TX entries */
1435 	for (int i = 0; i < adapter->tx_num_queues; i++) {
1436 		u32 index = i >> 1;
1437 		ivar = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, index);
1438 		tx_que = &adapter->tx_queues[i];
1439 		if (i & 1) {
1440 			ivar &= 0x00FFFFFF;
1441 			ivar |= (tx_que->msix | IGC_IVAR_VALID) << 24;
1442 		} else {
1443 			ivar &= 0xFFFF00FF;
1444 			ivar |= (tx_que->msix | IGC_IVAR_VALID) << 8;
1445 		}
1446 		IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, index, ivar);
1447 		adapter->que_mask |= tx_que->eims;
1448 	}
1449 
1450 	/* And for the link interrupt */
1451 	ivar = (adapter->linkvec | IGC_IVAR_VALID) << 8;
1452 	adapter->link_mask = 1 << adapter->linkvec;
1453 	IGC_WRITE_REG(hw, IGC_IVAR_MISC, ivar);
1454 
1455 	/* Set the starting interrupt rate */
1456 	if (igc_max_interrupt_rate > 0)
1457 		newitr = (4000000 / igc_max_interrupt_rate) & 0x7FFC;
1458 
1459 	newitr |= IGC_EITR_CNT_IGNR;
1460 
1461 	for (int i = 0; i < adapter->rx_num_queues; i++) {
1462 		rx_que = &adapter->rx_queues[i];
1463 		IGC_WRITE_REG(hw, IGC_EITR(rx_que->msix), newitr);
1464 	}
1465 
1466 	return;
1467 }
1468 
1469 static void
1470 igc_free_pci_resources(if_ctx_t ctx)
1471 {
1472 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1473 	struct igc_rx_queue *que = adapter->rx_queues;
1474 	device_t dev = iflib_get_dev(ctx);
1475 
1476 	/* Release all MSI-X queue resources */
1477 	if (adapter->intr_type == IFLIB_INTR_MSIX)
1478 		iflib_irq_free(ctx, &adapter->irq);
1479 
1480 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
1481 		iflib_irq_free(ctx, &que->que_irq);
1482 	}
1483 
1484 	if (adapter->memory != NULL) {
1485 		bus_release_resource(dev, SYS_RES_MEMORY,
1486 		    rman_get_rid(adapter->memory), adapter->memory);
1487 		adapter->memory = NULL;
1488 	}
1489 
1490 	if (adapter->flash != NULL) {
1491 		bus_release_resource(dev, SYS_RES_MEMORY,
1492 		    rman_get_rid(adapter->flash), adapter->flash);
1493 		adapter->flash = NULL;
1494 	}
1495 
1496 	if (adapter->ioport != NULL) {
1497 		bus_release_resource(dev, SYS_RES_IOPORT,
1498 		    rman_get_rid(adapter->ioport), adapter->ioport);
1499 		adapter->ioport = NULL;
1500 	}
1501 }
1502 
1503 /* Set up MSI or MSI-X */
1504 static int
1505 igc_setup_msix(if_ctx_t ctx)
1506 {
1507 	return (0);
1508 }
1509 
1510 /*********************************************************************
1511  *
1512  *  Initialize the DMA Coalescing feature
1513  *
1514  **********************************************************************/
1515 static void
1516 igc_init_dmac(struct igc_adapter *adapter, u32 pba)
1517 {
1518 	device_t	dev = adapter->dev;
1519 	struct igc_hw *hw = &adapter->hw;
1520 	u32 		dmac, reg = ~IGC_DMACR_DMAC_EN;
1521 	u16		hwm;
1522 	u16		max_frame_size;
1523 	int		status;
1524 
1525 	max_frame_size = adapter->shared->isc_max_frame_size;
1526 
1527 	if (adapter->dmac == 0) { /* Disabling it */
1528 		IGC_WRITE_REG(hw, IGC_DMACR, reg);
1529 		return;
1530 	} else
1531 		device_printf(dev, "DMA Coalescing enabled\n");
1532 
1533 	/* Set starting threshold */
1534 	IGC_WRITE_REG(hw, IGC_DMCTXTH, 0);
1535 
1536 	hwm = 64 * pba - max_frame_size / 16;
1537 	if (hwm < 64 * (pba - 6))
1538 		hwm = 64 * (pba - 6);
1539 	reg = IGC_READ_REG(hw, IGC_FCRTC);
1540 	reg &= ~IGC_FCRTC_RTH_COAL_MASK;
1541 	reg |= ((hwm << IGC_FCRTC_RTH_COAL_SHIFT)
1542 		& IGC_FCRTC_RTH_COAL_MASK);
1543 	IGC_WRITE_REG(hw, IGC_FCRTC, reg);
1544 
1545 	dmac = pba - max_frame_size / 512;
1546 	if (dmac < pba - 10)
1547 		dmac = pba - 10;
1548 	reg = IGC_READ_REG(hw, IGC_DMACR);
1549 	reg &= ~IGC_DMACR_DMACTHR_MASK;
1550 	reg |= ((dmac << IGC_DMACR_DMACTHR_SHIFT)
1551 		& IGC_DMACR_DMACTHR_MASK);
1552 
1553 	/* transition to L0x or L1 if available..*/
1554 	reg |= (IGC_DMACR_DMAC_EN | IGC_DMACR_DMAC_LX_MASK);
1555 
1556 	/* Check if status is 2.5Gb backplane connection
1557 	 * before configuration of watchdog timer, which is
1558 	 * in msec values in 12.8usec intervals
1559 	 * watchdog timer= msec values in 32usec intervals
1560 	 * for non 2.5Gb connection
1561 	 */
1562 	status = IGC_READ_REG(hw, IGC_STATUS);
1563 	if ((status & IGC_STATUS_2P5_SKU) &&
1564 	    (!(status & IGC_STATUS_2P5_SKU_OVER)))
1565 		reg |= ((adapter->dmac * 5) >> 6);
1566 	else
1567 		reg |= (adapter->dmac >> 5);
1568 
1569 	IGC_WRITE_REG(hw, IGC_DMACR, reg);
1570 
1571 	IGC_WRITE_REG(hw, IGC_DMCRTRH, 0);
1572 
1573 	/* Set the interval before transition */
1574 	reg = IGC_READ_REG(hw, IGC_DMCTLX);
1575 	reg |= IGC_DMCTLX_DCFLUSH_DIS;
1576 
1577 	/*
1578 	** in 2.5Gb connection, TTLX unit is 0.4 usec
1579 	** which is 0x4*2 = 0xA. But delay is still 4 usec
1580 	*/
1581 	status = IGC_READ_REG(hw, IGC_STATUS);
1582 	if ((status & IGC_STATUS_2P5_SKU) &&
1583 	    (!(status & IGC_STATUS_2P5_SKU_OVER)))
1584 		reg |= 0xA;
1585 	else
1586 		reg |= 0x4;
1587 
1588 	IGC_WRITE_REG(hw, IGC_DMCTLX, reg);
1589 
1590 	/* free space in tx packet buffer to wake from DMA coal */
1591 	IGC_WRITE_REG(hw, IGC_DMCTXTH, (IGC_TXPBSIZE -
1592 	    (2 * max_frame_size)) >> 6);
1593 
1594 	/* make low power state decision controlled by DMA coal */
1595 	reg = IGC_READ_REG(hw, IGC_PCIEMISC);
1596 	reg &= ~IGC_PCIEMISC_LX_DECISION;
1597 	IGC_WRITE_REG(hw, IGC_PCIEMISC, reg);
1598 }
1599 
1600 /*********************************************************************
1601  *
1602  *  Initialize the hardware to a configuration as specified by the
1603  *  adapter structure.
1604  *
1605  **********************************************************************/
1606 static void
1607 igc_reset(if_ctx_t ctx)
1608 {
1609 	device_t dev = iflib_get_dev(ctx);
1610 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1611 	struct igc_hw *hw = &adapter->hw;
1612 	u16 rx_buffer_size;
1613 	u32 pba;
1614 
1615 	INIT_DEBUGOUT("igc_reset: begin");
1616 	/* Let the firmware know the OS is in control */
1617 	igc_get_hw_control(adapter);
1618 
1619 	/*
1620 	 * Packet Buffer Allocation (PBA)
1621 	 * Writing PBA sets the receive portion of the buffer
1622 	 * the remainder is used for the transmit buffer.
1623 	 */
1624 	pba = IGC_PBA_34K;
1625 
1626 	INIT_DEBUGOUT1("igc_reset: pba=%dK",pba);
1627 
1628 	/*
1629 	 * These parameters control the automatic generation (Tx) and
1630 	 * response (Rx) to Ethernet PAUSE frames.
1631 	 * - High water mark should allow for at least two frames to be
1632 	 *   received after sending an XOFF.
1633 	 * - Low water mark works best when it is very near the high water mark.
1634 	 *   This allows the receiver to restart by sending XON when it has
1635 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1636 	 *   restart after one full frame is pulled from the buffer. There
1637 	 *   could be several smaller frames in the buffer and if so they will
1638 	 *   not trigger the XON until their total number reduces the buffer
1639 	 *   by 1500.
1640 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1641 	 */
1642 	rx_buffer_size = (pba & 0xffff) << 10;
1643 	hw->fc.high_water = rx_buffer_size -
1644 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
1645 	/* 16-byte granularity */
1646 	hw->fc.low_water = hw->fc.high_water - 16;
1647 
1648 	if (adapter->fc) /* locally set flow control value? */
1649 		hw->fc.requested_mode = adapter->fc;
1650 	else
1651 		hw->fc.requested_mode = igc_fc_full;
1652 
1653 	hw->fc.pause_time = IGC_FC_PAUSE_TIME;
1654 
1655 	hw->fc.send_xon = true;
1656 
1657 	/* Issue a global reset */
1658 	igc_reset_hw(hw);
1659 	IGC_WRITE_REG(hw, IGC_WUC, 0);
1660 
1661 	/* and a re-init */
1662 	if (igc_init_hw(hw) < 0) {
1663 		device_printf(dev, "Hardware Initialization Failed\n");
1664 		return;
1665 	}
1666 
1667 	/* Setup DMA Coalescing */
1668 	igc_init_dmac(adapter, pba);
1669 
1670 	IGC_WRITE_REG(hw, IGC_VET, ETHERTYPE_VLAN);
1671 	igc_get_phy_info(hw);
1672 	igc_check_for_link(hw);
1673 }
1674 
1675 /*
1676  * Initialise the RSS mapping for NICs that support multiple transmit/
1677  * receive rings.
1678  */
1679 
1680 #define RSSKEYLEN 10
1681 static void
1682 igc_initialize_rss_mapping(struct igc_adapter *adapter)
1683 {
1684 	struct igc_hw *hw = &adapter->hw;
1685 	int i;
1686 	int queue_id;
1687 	u32 reta;
1688 	u32 rss_key[RSSKEYLEN], mrqc, shift = 0;
1689 
1690 	/*
1691 	 * The redirection table controls which destination
1692 	 * queue each bucket redirects traffic to.
1693 	 * Each DWORD represents four queues, with the LSB
1694 	 * being the first queue in the DWORD.
1695 	 *
1696 	 * This just allocates buckets to queues using round-robin
1697 	 * allocation.
1698 	 *
1699 	 * NOTE: It Just Happens to line up with the default
1700 	 * RSS allocation method.
1701 	 */
1702 
1703 	/* Warning FM follows */
1704 	reta = 0;
1705 	for (i = 0; i < 128; i++) {
1706 #ifdef RSS
1707 		queue_id = rss_get_indirection_to_bucket(i);
1708 		/*
1709 		 * If we have more queues than buckets, we'll
1710 		 * end up mapping buckets to a subset of the
1711 		 * queues.
1712 		 *
1713 		 * If we have more buckets than queues, we'll
1714 		 * end up instead assigning multiple buckets
1715 		 * to queues.
1716 		 *
1717 		 * Both are suboptimal, but we need to handle
1718 		 * the case so we don't go out of bounds
1719 		 * indexing arrays and such.
1720 		 */
1721 		queue_id = queue_id % adapter->rx_num_queues;
1722 #else
1723 		queue_id = (i % adapter->rx_num_queues);
1724 #endif
1725 		/* Adjust if required */
1726 		queue_id = queue_id << shift;
1727 
1728 		/*
1729 		 * The low 8 bits are for hash value (n+0);
1730 		 * The next 8 bits are for hash value (n+1), etc.
1731 		 */
1732 		reta = reta >> 8;
1733 		reta = reta | ( ((uint32_t) queue_id) << 24);
1734 		if ((i & 3) == 3) {
1735 			IGC_WRITE_REG(hw, IGC_RETA(i >> 2), reta);
1736 			reta = 0;
1737 		}
1738 	}
1739 
1740 	/* Now fill in hash table */
1741 
1742 	/*
1743 	 * MRQC: Multiple Receive Queues Command
1744 	 * Set queuing to RSS control, number depends on the device.
1745 	 */
1746 	mrqc = IGC_MRQC_ENABLE_RSS_4Q;
1747 
1748 #ifdef RSS
1749 	/* XXX ew typecasting */
1750 	rss_getkey((uint8_t *) &rss_key);
1751 #else
1752 	arc4rand(&rss_key, sizeof(rss_key), 0);
1753 #endif
1754 	for (i = 0; i < RSSKEYLEN; i++)
1755 		IGC_WRITE_REG_ARRAY(hw, IGC_RSSRK(0), i, rss_key[i]);
1756 
1757 	/*
1758 	 * Configure the RSS fields to hash upon.
1759 	 */
1760 	mrqc |= (IGC_MRQC_RSS_FIELD_IPV4 |
1761 	    IGC_MRQC_RSS_FIELD_IPV4_TCP);
1762 	mrqc |= (IGC_MRQC_RSS_FIELD_IPV6 |
1763 	    IGC_MRQC_RSS_FIELD_IPV6_TCP);
1764 	mrqc |=( IGC_MRQC_RSS_FIELD_IPV4_UDP |
1765 	    IGC_MRQC_RSS_FIELD_IPV6_UDP);
1766 	mrqc |=( IGC_MRQC_RSS_FIELD_IPV6_UDP_EX |
1767 	    IGC_MRQC_RSS_FIELD_IPV6_TCP_EX);
1768 
1769 	IGC_WRITE_REG(hw, IGC_MRQC, mrqc);
1770 }
1771 
1772 /*********************************************************************
1773  *
1774  *  Setup networking device structure and register interface media.
1775  *
1776  **********************************************************************/
1777 static int
1778 igc_setup_interface(if_ctx_t ctx)
1779 {
1780 	struct ifnet *ifp = iflib_get_ifp(ctx);
1781 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1782 	if_softc_ctx_t scctx = adapter->shared;
1783 
1784 	INIT_DEBUGOUT("igc_setup_interface: begin");
1785 
1786 	/* Single Queue */
1787 	if (adapter->tx_num_queues == 1) {
1788 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
1789 		if_setsendqready(ifp);
1790 	}
1791 
1792 	/*
1793 	 * Specify the media types supported by this adapter and register
1794 	 * callbacks to update media and link information
1795 	 */
1796 	ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
1797 	ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
1798 	ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
1799 	ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
1800 	ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
1801 	ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
1802 	ifmedia_add(adapter->media, IFM_ETHER | IFM_2500_T, 0, NULL);
1803 
1804 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
1805 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
1806 	return (0);
1807 }
1808 
1809 static int
1810 igc_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
1811 {
1812 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1813 	if_softc_ctx_t scctx = adapter->shared;
1814 	int error = IGC_SUCCESS;
1815 	struct igc_tx_queue *que;
1816 	int i, j;
1817 
1818 	MPASS(adapter->tx_num_queues > 0);
1819 	MPASS(adapter->tx_num_queues == ntxqsets);
1820 
1821 	/* First allocate the top level queue structs */
1822 	if (!(adapter->tx_queues =
1823 	    (struct igc_tx_queue *) malloc(sizeof(struct igc_tx_queue) *
1824 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
1825 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
1826 		return(ENOMEM);
1827 	}
1828 
1829 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
1830 		/* Set up some basics */
1831 
1832 		struct tx_ring *txr = &que->txr;
1833 		txr->adapter = que->adapter = adapter;
1834 		que->me = txr->me =  i;
1835 
1836 		/* Allocate report status array */
1837 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
1838 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
1839 			error = ENOMEM;
1840 			goto fail;
1841 		}
1842 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
1843 			txr->tx_rsq[j] = QIDX_INVALID;
1844 		/* get the virtual and physical address of the hardware queues */
1845 		txr->tx_base = (struct igc_tx_desc *)vaddrs[i*ntxqs];
1846 		txr->tx_paddr = paddrs[i*ntxqs];
1847 	}
1848 
1849 	if (bootverbose)
1850 		device_printf(iflib_get_dev(ctx),
1851 		    "allocated for %d tx_queues\n", adapter->tx_num_queues);
1852 	return (0);
1853 fail:
1854 	igc_if_queues_free(ctx);
1855 	return (error);
1856 }
1857 
1858 static int
1859 igc_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
1860 {
1861 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1862 	int error = IGC_SUCCESS;
1863 	struct igc_rx_queue *que;
1864 	int i;
1865 
1866 	MPASS(adapter->rx_num_queues > 0);
1867 	MPASS(adapter->rx_num_queues == nrxqsets);
1868 
1869 	/* First allocate the top level queue structs */
1870 	if (!(adapter->rx_queues =
1871 	    (struct igc_rx_queue *) malloc(sizeof(struct igc_rx_queue) *
1872 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
1873 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
1874 		error = ENOMEM;
1875 		goto fail;
1876 	}
1877 
1878 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
1879 		/* Set up some basics */
1880 		struct rx_ring *rxr = &que->rxr;
1881 		rxr->adapter = que->adapter = adapter;
1882 		rxr->que = que;
1883 		que->me = rxr->me =  i;
1884 
1885 		/* get the virtual and physical address of the hardware queues */
1886 		rxr->rx_base = (union igc_rx_desc_extended *)vaddrs[i*nrxqs];
1887 		rxr->rx_paddr = paddrs[i*nrxqs];
1888 	}
1889 
1890 	if (bootverbose)
1891 		device_printf(iflib_get_dev(ctx),
1892 		    "allocated for %d rx_queues\n", adapter->rx_num_queues);
1893 
1894 	return (0);
1895 fail:
1896 	igc_if_queues_free(ctx);
1897 	return (error);
1898 }
1899 
1900 static void
1901 igc_if_queues_free(if_ctx_t ctx)
1902 {
1903 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1904 	struct igc_tx_queue *tx_que = adapter->tx_queues;
1905 	struct igc_rx_queue *rx_que = adapter->rx_queues;
1906 
1907 	if (tx_que != NULL) {
1908 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
1909 			struct tx_ring *txr = &tx_que->txr;
1910 			if (txr->tx_rsq == NULL)
1911 				break;
1912 
1913 			free(txr->tx_rsq, M_DEVBUF);
1914 			txr->tx_rsq = NULL;
1915 		}
1916 		free(adapter->tx_queues, M_DEVBUF);
1917 		adapter->tx_queues = NULL;
1918 	}
1919 
1920 	if (rx_que != NULL) {
1921 		free(adapter->rx_queues, M_DEVBUF);
1922 		adapter->rx_queues = NULL;
1923 	}
1924 
1925 	igc_release_hw_control(adapter);
1926 
1927 	if (adapter->mta != NULL) {
1928 		free(adapter->mta, M_DEVBUF);
1929 	}
1930 }
1931 
1932 /*********************************************************************
1933  *
1934  *  Enable transmit unit.
1935  *
1936  **********************************************************************/
1937 static void
1938 igc_initialize_transmit_unit(if_ctx_t ctx)
1939 {
1940 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1941 	if_softc_ctx_t scctx = adapter->shared;
1942 	struct igc_tx_queue *que;
1943 	struct tx_ring	*txr;
1944 	struct igc_hw	*hw = &adapter->hw;
1945 	u32 tctl, txdctl = 0;
1946 
1947 	INIT_DEBUGOUT("igc_initialize_transmit_unit: begin");
1948 
1949 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
1950 		u64 bus_addr;
1951 		caddr_t offp, endp;
1952 
1953 		que = &adapter->tx_queues[i];
1954 		txr = &que->txr;
1955 		bus_addr = txr->tx_paddr;
1956 
1957 		/* Clear checksum offload context. */
1958 		offp = (caddr_t)&txr->csum_flags;
1959 		endp = (caddr_t)(txr + 1);
1960 		bzero(offp, endp - offp);
1961 
1962 		/* Base and Len of TX Ring */
1963 		IGC_WRITE_REG(hw, IGC_TDLEN(i),
1964 		    scctx->isc_ntxd[0] * sizeof(struct igc_tx_desc));
1965 		IGC_WRITE_REG(hw, IGC_TDBAH(i),
1966 		    (u32)(bus_addr >> 32));
1967 		IGC_WRITE_REG(hw, IGC_TDBAL(i),
1968 		    (u32)bus_addr);
1969 		/* Init the HEAD/TAIL indices */
1970 		IGC_WRITE_REG(hw, IGC_TDT(i), 0);
1971 		IGC_WRITE_REG(hw, IGC_TDH(i), 0);
1972 
1973 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
1974 		    IGC_READ_REG(&adapter->hw, IGC_TDBAL(i)),
1975 		    IGC_READ_REG(&adapter->hw, IGC_TDLEN(i)));
1976 
1977 		txdctl = 0; /* clear txdctl */
1978 		txdctl |= 0x1f; /* PTHRESH */
1979 		txdctl |= 1 << 8; /* HTHRESH */
1980 		txdctl |= 1 << 16;/* WTHRESH */
1981 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
1982 		txdctl |= IGC_TXDCTL_GRAN;
1983 		txdctl |= 1 << 25; /* LWTHRESH */
1984 
1985 		IGC_WRITE_REG(hw, IGC_TXDCTL(i), txdctl);
1986 	}
1987 
1988 	/* Program the Transmit Control Register */
1989 	tctl = IGC_READ_REG(&adapter->hw, IGC_TCTL);
1990 	tctl &= ~IGC_TCTL_CT;
1991 	tctl |= (IGC_TCTL_PSP | IGC_TCTL_RTLC | IGC_TCTL_EN |
1992 		   (IGC_COLLISION_THRESHOLD << IGC_CT_SHIFT));
1993 
1994 	/* This write will effectively turn on the transmit unit. */
1995 	IGC_WRITE_REG(&adapter->hw, IGC_TCTL, tctl);
1996 }
1997 
1998 /*********************************************************************
1999  *
2000  *  Enable receive unit.
2001  *
2002  **********************************************************************/
2003 
2004 static void
2005 igc_initialize_receive_unit(if_ctx_t ctx)
2006 {
2007 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2008 	if_softc_ctx_t scctx = adapter->shared;
2009 	struct ifnet *ifp = iflib_get_ifp(ctx);
2010 	struct igc_hw	*hw = &adapter->hw;
2011 	struct igc_rx_queue *que;
2012 	int i;
2013 	u32 psize, rctl, rxcsum, srrctl = 0;
2014 
2015 	INIT_DEBUGOUT("igc_initialize_receive_units: begin");
2016 
2017 	/*
2018 	 * Make sure receives are disabled while setting
2019 	 * up the descriptor ring
2020 	 */
2021 	rctl = IGC_READ_REG(hw, IGC_RCTL);
2022 	IGC_WRITE_REG(hw, IGC_RCTL, rctl & ~IGC_RCTL_EN);
2023 
2024 	/* Setup the Receive Control Register */
2025 	rctl &= ~(3 << IGC_RCTL_MO_SHIFT);
2026 	rctl |= IGC_RCTL_EN | IGC_RCTL_BAM |
2027 	    IGC_RCTL_LBM_NO | IGC_RCTL_RDMTS_HALF |
2028 	    (hw->mac.mc_filter_type << IGC_RCTL_MO_SHIFT);
2029 
2030 	/* Do not store bad packets */
2031 	rctl &= ~IGC_RCTL_SBP;
2032 
2033 	/* Enable Long Packet receive */
2034 	if (if_getmtu(ifp) > ETHERMTU)
2035 		rctl |= IGC_RCTL_LPE;
2036 	else
2037 		rctl &= ~IGC_RCTL_LPE;
2038 
2039 	/* Strip the CRC */
2040 	if (!igc_disable_crc_stripping)
2041 		rctl |= IGC_RCTL_SECRC;
2042 
2043 	/*
2044 	 * Set the interrupt throttling rate. Value is calculated
2045 	 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
2046 	 */
2047 	IGC_WRITE_REG(hw, IGC_ITR, DEFAULT_ITR);
2048 
2049 	rxcsum = IGC_READ_REG(hw, IGC_RXCSUM);
2050 	if (if_getcapenable(ifp) & IFCAP_RXCSUM) {
2051 		rxcsum |= IGC_RXCSUM_CRCOFL;
2052 		if (adapter->tx_num_queues > 1)
2053 			rxcsum |= IGC_RXCSUM_PCSD;
2054 		else
2055 			rxcsum |= IGC_RXCSUM_IPPCSE;
2056 	} else {
2057 		if (adapter->tx_num_queues > 1)
2058 			rxcsum |= IGC_RXCSUM_PCSD;
2059 		else
2060 			rxcsum &= ~IGC_RXCSUM_TUOFL;
2061 	}
2062 	IGC_WRITE_REG(hw, IGC_RXCSUM, rxcsum);
2063 
2064 	if (adapter->rx_num_queues > 1)
2065 		igc_initialize_rss_mapping(adapter);
2066 
2067 	if (if_getmtu(ifp) > ETHERMTU) {
2068 		/* Set maximum packet len */
2069 		if (adapter->rx_mbuf_sz <= 4096) {
2070 			srrctl |= 4096 >> IGC_SRRCTL_BSIZEPKT_SHIFT;
2071 			rctl |= IGC_RCTL_SZ_4096 | IGC_RCTL_BSEX;
2072 		} else if (adapter->rx_mbuf_sz > 4096) {
2073 			srrctl |= 8192 >> IGC_SRRCTL_BSIZEPKT_SHIFT;
2074 			rctl |= IGC_RCTL_SZ_8192 | IGC_RCTL_BSEX;
2075 		}
2076 		psize = scctx->isc_max_frame_size;
2077 		/* are we on a vlan? */
2078 		if (ifp->if_vlantrunk != NULL)
2079 			psize += VLAN_TAG_SIZE;
2080 		IGC_WRITE_REG(&adapter->hw, IGC_RLPML, psize);
2081 	} else {
2082 		srrctl |= 2048 >> IGC_SRRCTL_BSIZEPKT_SHIFT;
2083 		rctl |= IGC_RCTL_SZ_2048;
2084 	}
2085 
2086 	/*
2087 	 * If TX flow control is disabled and there's >1 queue defined,
2088 	 * enable DROP.
2089 	 *
2090 	 * This drops frames rather than hanging the RX MAC for all queues.
2091 	 */
2092 	if ((adapter->rx_num_queues > 1) &&
2093 	    (adapter->fc == igc_fc_none ||
2094 	     adapter->fc == igc_fc_rx_pause)) {
2095 		srrctl |= IGC_SRRCTL_DROP_EN;
2096 	}
2097 
2098 	/* Setup the Base and Length of the Rx Descriptor Rings */
2099 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
2100 		struct rx_ring *rxr = &que->rxr;
2101 		u64 bus_addr = rxr->rx_paddr;
2102 		u32 rxdctl;
2103 
2104 #ifdef notyet
2105 		/* Configure for header split? -- ignore for now */
2106 		rxr->hdr_split = igc_header_split;
2107 #else
2108 		srrctl |= IGC_SRRCTL_DESCTYPE_ADV_ONEBUF;
2109 #endif
2110 
2111 		IGC_WRITE_REG(hw, IGC_RDLEN(i),
2112 			      scctx->isc_nrxd[0] * sizeof(struct igc_rx_desc));
2113 		IGC_WRITE_REG(hw, IGC_RDBAH(i),
2114 			      (uint32_t)(bus_addr >> 32));
2115 		IGC_WRITE_REG(hw, IGC_RDBAL(i),
2116 			      (uint32_t)bus_addr);
2117 		IGC_WRITE_REG(hw, IGC_SRRCTL(i), srrctl);
2118 		/* Setup the Head and Tail Descriptor Pointers */
2119 		IGC_WRITE_REG(hw, IGC_RDH(i), 0);
2120 		IGC_WRITE_REG(hw, IGC_RDT(i), 0);
2121 		/* Enable this Queue */
2122 		rxdctl = IGC_READ_REG(hw, IGC_RXDCTL(i));
2123 		rxdctl |= IGC_RXDCTL_QUEUE_ENABLE;
2124 		rxdctl &= 0xFFF00000;
2125 		rxdctl |= IGC_RX_PTHRESH;
2126 		rxdctl |= IGC_RX_HTHRESH << 8;
2127 		rxdctl |= IGC_RX_WTHRESH << 16;
2128 		IGC_WRITE_REG(hw, IGC_RXDCTL(i), rxdctl);
2129 	}
2130 
2131 	/* Make sure VLAN Filters are off */
2132 	rctl &= ~IGC_RCTL_VFE;
2133 
2134 	/* Write out the settings */
2135 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
2136 
2137 	return;
2138 }
2139 
2140 static void
2141 igc_if_vlan_register(if_ctx_t ctx, u16 vtag)
2142 {
2143 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2144 	u32 index, bit;
2145 
2146 	index = (vtag >> 5) & 0x7F;
2147 	bit = vtag & 0x1F;
2148 	adapter->shadow_vfta[index] |= (1 << bit);
2149 	++adapter->num_vlans;
2150 }
2151 
2152 static void
2153 igc_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
2154 {
2155 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2156 	u32 index, bit;
2157 
2158 	index = (vtag >> 5) & 0x7F;
2159 	bit = vtag & 0x1F;
2160 	adapter->shadow_vfta[index] &= ~(1 << bit);
2161 	--adapter->num_vlans;
2162 }
2163 
2164 static void
2165 igc_setup_vlan_hw_support(struct igc_adapter *adapter)
2166 {
2167 	struct igc_hw *hw = &adapter->hw;
2168 	u32 reg;
2169 
2170 	/*
2171 	 * We get here thru init_locked, meaning
2172 	 * a soft reset, this has already cleared
2173 	 * the VFTA and other state, so if there
2174 	 * have been no vlan's registered do nothing.
2175 	 */
2176 	if (adapter->num_vlans == 0)
2177 		return;
2178 
2179 	/*
2180 	 * A soft reset zero's out the VFTA, so
2181 	 * we need to repopulate it now.
2182 	 */
2183 	for (int i = 0; i < IGC_VFTA_SIZE; i++)
2184 		if (adapter->shadow_vfta[i] != 0)
2185 			IGC_WRITE_REG_ARRAY(hw, IGC_VFTA,
2186 			    i, adapter->shadow_vfta[i]);
2187 
2188 	reg = IGC_READ_REG(hw, IGC_CTRL);
2189 	reg |= IGC_CTRL_VME;
2190 	IGC_WRITE_REG(hw, IGC_CTRL, reg);
2191 
2192 	/* Enable the Filter Table */
2193 	reg = IGC_READ_REG(hw, IGC_RCTL);
2194 	reg &= ~IGC_RCTL_CFIEN;
2195 	reg |= IGC_RCTL_VFE;
2196 	IGC_WRITE_REG(hw, IGC_RCTL, reg);
2197 }
2198 
2199 static void
2200 igc_if_intr_enable(if_ctx_t ctx)
2201 {
2202 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2203 	struct igc_hw *hw = &adapter->hw;
2204 	u32 mask;
2205 
2206 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
2207 		mask = (adapter->que_mask | adapter->link_mask);
2208 		IGC_WRITE_REG(hw, IGC_EIAC, mask);
2209 		IGC_WRITE_REG(hw, IGC_EIAM, mask);
2210 		IGC_WRITE_REG(hw, IGC_EIMS, mask);
2211 		IGC_WRITE_REG(hw, IGC_IMS, IGC_IMS_LSC);
2212 	} else
2213 		IGC_WRITE_REG(hw, IGC_IMS, IMS_ENABLE_MASK);
2214 	IGC_WRITE_FLUSH(hw);
2215 }
2216 
2217 static void
2218 igc_if_intr_disable(if_ctx_t ctx)
2219 {
2220 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2221 	struct igc_hw *hw = &adapter->hw;
2222 
2223 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
2224 		IGC_WRITE_REG(hw, IGC_EIMC, 0xffffffff);
2225 		IGC_WRITE_REG(hw, IGC_EIAC, 0);
2226 	}
2227 	IGC_WRITE_REG(hw, IGC_IMC, 0xffffffff);
2228 	IGC_WRITE_FLUSH(hw);
2229 }
2230 
2231 /*
2232  * igc_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
2233  * For ASF and Pass Through versions of f/w this means
2234  * that the driver is loaded. For AMT version type f/w
2235  * this means that the network i/f is open.
2236  */
2237 static void
2238 igc_get_hw_control(struct igc_adapter *adapter)
2239 {
2240 	u32 ctrl_ext;
2241 
2242 	if (adapter->vf_ifp)
2243 		return;
2244 
2245 	ctrl_ext = IGC_READ_REG(&adapter->hw, IGC_CTRL_EXT);
2246 	IGC_WRITE_REG(&adapter->hw, IGC_CTRL_EXT,
2247 	    ctrl_ext | IGC_CTRL_EXT_DRV_LOAD);
2248 }
2249 
2250 /*
2251  * igc_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
2252  * For ASF and Pass Through versions of f/w this means that
2253  * the driver is no longer loaded. For AMT versions of the
2254  * f/w this means that the network i/f is closed.
2255  */
2256 static void
2257 igc_release_hw_control(struct igc_adapter *adapter)
2258 {
2259 	u32 ctrl_ext;
2260 
2261 	ctrl_ext = IGC_READ_REG(&adapter->hw, IGC_CTRL_EXT);
2262 	IGC_WRITE_REG(&adapter->hw, IGC_CTRL_EXT,
2263 	    ctrl_ext & ~IGC_CTRL_EXT_DRV_LOAD);
2264 	return;
2265 }
2266 
2267 static int
2268 igc_is_valid_ether_addr(u8 *addr)
2269 {
2270 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
2271 
2272 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
2273 		return (false);
2274 	}
2275 
2276 	return (true);
2277 }
2278 
2279 /*
2280 ** Parse the interface capabilities with regard
2281 ** to both system management and wake-on-lan for
2282 ** later use.
2283 */
2284 static void
2285 igc_get_wakeup(if_ctx_t ctx)
2286 {
2287 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2288 	u16 eeprom_data = 0, apme_mask;
2289 
2290 	apme_mask = IGC_WUC_APME;
2291 	eeprom_data = IGC_READ_REG(&adapter->hw, IGC_WUC);
2292 
2293 	if (eeprom_data & apme_mask)
2294 		adapter->wol = IGC_WUFC_LNKC;
2295 }
2296 
2297 
2298 /*
2299  * Enable PCI Wake On Lan capability
2300  */
2301 static void
2302 igc_enable_wakeup(if_ctx_t ctx)
2303 {
2304 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2305 	device_t dev = iflib_get_dev(ctx);
2306 	if_t ifp = iflib_get_ifp(ctx);
2307 	int error = 0;
2308 	u32 pmc, ctrl, rctl;
2309 	u16 status;
2310 
2311 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
2312 		return;
2313 
2314 	/*
2315 	 * Determine type of Wakeup: note that wol
2316 	 * is set with all bits on by default.
2317 	 */
2318 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
2319 		adapter->wol &= ~IGC_WUFC_MAG;
2320 
2321 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
2322 		adapter->wol &= ~IGC_WUFC_EX;
2323 
2324 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
2325 		adapter->wol &= ~IGC_WUFC_MC;
2326 	else {
2327 		rctl = IGC_READ_REG(&adapter->hw, IGC_RCTL);
2328 		rctl |= IGC_RCTL_MPE;
2329 		IGC_WRITE_REG(&adapter->hw, IGC_RCTL, rctl);
2330 	}
2331 
2332 	if (!(adapter->wol & (IGC_WUFC_EX | IGC_WUFC_MAG | IGC_WUFC_MC)))
2333 		goto pme;
2334 
2335 	/* Advertise the wakeup capability */
2336 	ctrl = IGC_READ_REG(&adapter->hw, IGC_CTRL);
2337 	ctrl |= IGC_CTRL_ADVD3WUC;
2338 	IGC_WRITE_REG(&adapter->hw, IGC_CTRL, ctrl);
2339 
2340 	/* Enable wakeup by the MAC */
2341 	IGC_WRITE_REG(&adapter->hw, IGC_WUC, IGC_WUC_PME_EN);
2342 	IGC_WRITE_REG(&adapter->hw, IGC_WUFC, adapter->wol);
2343 
2344 pme:
2345 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
2346 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
2347 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
2348 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
2349 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
2350 
2351 	return;
2352 }
2353 
2354 /**********************************************************************
2355  *
2356  *  Update the board statistics counters.
2357  *
2358  **********************************************************************/
2359 static void
2360 igc_update_stats_counters(struct igc_adapter *adapter)
2361 {
2362 	u64 prev_xoffrxc = adapter->stats.xoffrxc;
2363 
2364 	adapter->stats.crcerrs += IGC_READ_REG(&adapter->hw, IGC_CRCERRS);
2365 	adapter->stats.mpc += IGC_READ_REG(&adapter->hw, IGC_MPC);
2366 	adapter->stats.scc += IGC_READ_REG(&adapter->hw, IGC_SCC);
2367 	adapter->stats.ecol += IGC_READ_REG(&adapter->hw, IGC_ECOL);
2368 
2369 	adapter->stats.mcc += IGC_READ_REG(&adapter->hw, IGC_MCC);
2370 	adapter->stats.latecol += IGC_READ_REG(&adapter->hw, IGC_LATECOL);
2371 	adapter->stats.colc += IGC_READ_REG(&adapter->hw, IGC_COLC);
2372 	adapter->stats.colc += IGC_READ_REG(&adapter->hw, IGC_RERC);
2373 	adapter->stats.dc += IGC_READ_REG(&adapter->hw, IGC_DC);
2374 	adapter->stats.rlec += IGC_READ_REG(&adapter->hw, IGC_RLEC);
2375 	adapter->stats.xonrxc += IGC_READ_REG(&adapter->hw, IGC_XONRXC);
2376 	adapter->stats.xontxc += IGC_READ_REG(&adapter->hw, IGC_XONTXC);
2377 	adapter->stats.xoffrxc += IGC_READ_REG(&adapter->hw, IGC_XOFFRXC);
2378 	/*
2379 	 * For watchdog management we need to know if we have been
2380 	 * paused during the last interval, so capture that here.
2381 	 */
2382 	if (adapter->stats.xoffrxc != prev_xoffrxc)
2383 		adapter->shared->isc_pause_frames = 1;
2384 	adapter->stats.xofftxc += IGC_READ_REG(&adapter->hw, IGC_XOFFTXC);
2385 	adapter->stats.fcruc += IGC_READ_REG(&adapter->hw, IGC_FCRUC);
2386 	adapter->stats.prc64 += IGC_READ_REG(&adapter->hw, IGC_PRC64);
2387 	adapter->stats.prc127 += IGC_READ_REG(&adapter->hw, IGC_PRC127);
2388 	adapter->stats.prc255 += IGC_READ_REG(&adapter->hw, IGC_PRC255);
2389 	adapter->stats.prc511 += IGC_READ_REG(&adapter->hw, IGC_PRC511);
2390 	adapter->stats.prc1023 += IGC_READ_REG(&adapter->hw, IGC_PRC1023);
2391 	adapter->stats.prc1522 += IGC_READ_REG(&adapter->hw, IGC_PRC1522);
2392 	adapter->stats.tlpic += IGC_READ_REG(&adapter->hw, IGC_TLPIC);
2393 	adapter->stats.rlpic += IGC_READ_REG(&adapter->hw, IGC_RLPIC);
2394 	adapter->stats.gprc += IGC_READ_REG(&adapter->hw, IGC_GPRC);
2395 	adapter->stats.bprc += IGC_READ_REG(&adapter->hw, IGC_BPRC);
2396 	adapter->stats.mprc += IGC_READ_REG(&adapter->hw, IGC_MPRC);
2397 	adapter->stats.gptc += IGC_READ_REG(&adapter->hw, IGC_GPTC);
2398 
2399 	/* For the 64-bit byte counters the low dword must be read first. */
2400 	/* Both registers clear on the read of the high dword */
2401 
2402 	adapter->stats.gorc += IGC_READ_REG(&adapter->hw, IGC_GORCL) +
2403 	    ((u64)IGC_READ_REG(&adapter->hw, IGC_GORCH) << 32);
2404 	adapter->stats.gotc += IGC_READ_REG(&adapter->hw, IGC_GOTCL) +
2405 	    ((u64)IGC_READ_REG(&adapter->hw, IGC_GOTCH) << 32);
2406 
2407 	adapter->stats.rnbc += IGC_READ_REG(&adapter->hw, IGC_RNBC);
2408 	adapter->stats.ruc += IGC_READ_REG(&adapter->hw, IGC_RUC);
2409 	adapter->stats.rfc += IGC_READ_REG(&adapter->hw, IGC_RFC);
2410 	adapter->stats.roc += IGC_READ_REG(&adapter->hw, IGC_ROC);
2411 	adapter->stats.rjc += IGC_READ_REG(&adapter->hw, IGC_RJC);
2412 
2413 	adapter->stats.tor += IGC_READ_REG(&adapter->hw, IGC_TORH);
2414 	adapter->stats.tot += IGC_READ_REG(&adapter->hw, IGC_TOTH);
2415 
2416 	adapter->stats.tpr += IGC_READ_REG(&adapter->hw, IGC_TPR);
2417 	adapter->stats.tpt += IGC_READ_REG(&adapter->hw, IGC_TPT);
2418 	adapter->stats.ptc64 += IGC_READ_REG(&adapter->hw, IGC_PTC64);
2419 	adapter->stats.ptc127 += IGC_READ_REG(&adapter->hw, IGC_PTC127);
2420 	adapter->stats.ptc255 += IGC_READ_REG(&adapter->hw, IGC_PTC255);
2421 	adapter->stats.ptc511 += IGC_READ_REG(&adapter->hw, IGC_PTC511);
2422 	adapter->stats.ptc1023 += IGC_READ_REG(&adapter->hw, IGC_PTC1023);
2423 	adapter->stats.ptc1522 += IGC_READ_REG(&adapter->hw, IGC_PTC1522);
2424 	adapter->stats.mptc += IGC_READ_REG(&adapter->hw, IGC_MPTC);
2425 	adapter->stats.bptc += IGC_READ_REG(&adapter->hw, IGC_BPTC);
2426 
2427 	/* Interrupt Counts */
2428 	adapter->stats.iac += IGC_READ_REG(&adapter->hw, IGC_IAC);
2429 	adapter->stats.rxdmtc += IGC_READ_REG(&adapter->hw, IGC_RXDMTC);
2430 
2431 	adapter->stats.algnerrc += IGC_READ_REG(&adapter->hw, IGC_ALGNERRC);
2432 	adapter->stats.tncrs += IGC_READ_REG(&adapter->hw, IGC_TNCRS);
2433 	adapter->stats.htdpmc += IGC_READ_REG(&adapter->hw, IGC_HTDPMC);
2434 	adapter->stats.tsctc += IGC_READ_REG(&adapter->hw, IGC_TSCTC);
2435 }
2436 
2437 static uint64_t
2438 igc_if_get_counter(if_ctx_t ctx, ift_counter cnt)
2439 {
2440 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2441 	struct ifnet *ifp = iflib_get_ifp(ctx);
2442 
2443 	switch (cnt) {
2444 	case IFCOUNTER_COLLISIONS:
2445 		return (adapter->stats.colc);
2446 	case IFCOUNTER_IERRORS:
2447 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
2448 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
2449 		    adapter->stats.ruc + adapter->stats.roc +
2450 		    adapter->stats.mpc + adapter->stats.htdpmc);
2451 	case IFCOUNTER_OERRORS:
2452 		return (adapter->stats.ecol + adapter->stats.latecol +
2453 		    adapter->watchdog_events);
2454 	default:
2455 		return (if_get_counter_default(ifp, cnt));
2456 	}
2457 }
2458 
2459 /* igc_if_needs_restart - Tell iflib when the driver needs to be reinitialized
2460  * @ctx: iflib context
2461  * @event: event code to check
2462  *
2463  * Defaults to returning true for unknown events.
2464  *
2465  * @returns true if iflib needs to reinit the interface
2466  */
2467 static bool
2468 igc_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
2469 {
2470 	switch (event) {
2471 	case IFLIB_RESTART_VLAN_CONFIG:
2472 	default:
2473 		return (true);
2474 	}
2475 }
2476 
2477 /* Export a single 32-bit register via a read-only sysctl. */
2478 static int
2479 igc_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
2480 {
2481 	struct igc_adapter *adapter;
2482 	u_int val;
2483 
2484 	adapter = oidp->oid_arg1;
2485 	val = IGC_READ_REG(&adapter->hw, oidp->oid_arg2);
2486 	return (sysctl_handle_int(oidp, &val, 0, req));
2487 }
2488 
2489 /*
2490  * Add sysctl variables, one per statistic, to the system.
2491  */
2492 static void
2493 igc_add_hw_stats(struct igc_adapter *adapter)
2494 {
2495 	device_t dev = iflib_get_dev(adapter->ctx);
2496 	struct igc_tx_queue *tx_que = adapter->tx_queues;
2497 	struct igc_rx_queue *rx_que = adapter->rx_queues;
2498 
2499 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
2500 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
2501 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
2502 	struct igc_hw_stats *stats = &adapter->stats;
2503 
2504 	struct sysctl_oid *stat_node, *queue_node, *int_node;
2505 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
2506 
2507 #define QUEUE_NAME_LEN 32
2508 	char namebuf[QUEUE_NAME_LEN];
2509 
2510 	/* Driver Statistics */
2511 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
2512 			CTLFLAG_RD, &adapter->dropped_pkts,
2513 			"Driver dropped packets");
2514 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
2515 			CTLFLAG_RD, &adapter->link_irq,
2516 			"Link MSI-X IRQ Handled");
2517 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
2518 			CTLFLAG_RD, &adapter->rx_overruns,
2519 			"RX overruns");
2520 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
2521 			CTLFLAG_RD, &adapter->watchdog_events,
2522 			"Watchdog timeouts");
2523 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
2524 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
2525 	    adapter, IGC_CTRL, igc_sysctl_reg_handler, "IU",
2526 	    "Device Control Register");
2527 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
2528 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
2529 	    adapter, IGC_RCTL, igc_sysctl_reg_handler, "IU",
2530 	    "Receiver Control Register");
2531 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
2532 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
2533 			"Flow Control High Watermark");
2534 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
2535 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
2536 			"Flow Control Low Watermark");
2537 
2538 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2539 		struct tx_ring *txr = &tx_que->txr;
2540 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
2541 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
2542 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
2543 		queue_list = SYSCTL_CHILDREN(queue_node);
2544 
2545 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
2546 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
2547 		    IGC_TDH(txr->me), igc_sysctl_reg_handler, "IU",
2548 		    "Transmit Descriptor Head");
2549 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
2550 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
2551 		    IGC_TDT(txr->me), igc_sysctl_reg_handler, "IU",
2552 		    "Transmit Descriptor Tail");
2553 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
2554 				CTLFLAG_RD, &txr->tx_irq,
2555 				"Queue MSI-X Transmit Interrupts");
2556 	}
2557 
2558 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
2559 		struct rx_ring *rxr = &rx_que->rxr;
2560 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
2561 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
2562 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
2563 		queue_list = SYSCTL_CHILDREN(queue_node);
2564 
2565 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
2566 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
2567 		    IGC_RDH(rxr->me), igc_sysctl_reg_handler, "IU",
2568 		    "Receive Descriptor Head");
2569 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
2570 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
2571 		    IGC_RDT(rxr->me), igc_sysctl_reg_handler, "IU",
2572 		    "Receive Descriptor Tail");
2573 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
2574 				CTLFLAG_RD, &rxr->rx_irq,
2575 				"Queue MSI-X Receive Interrupts");
2576 	}
2577 
2578 	/* MAC stats get their own sub node */
2579 
2580 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
2581 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
2582 	stat_list = SYSCTL_CHILDREN(stat_node);
2583 
2584 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
2585 			CTLFLAG_RD, &stats->ecol,
2586 			"Excessive collisions");
2587 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
2588 			CTLFLAG_RD, &stats->scc,
2589 			"Single collisions");
2590 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
2591 			CTLFLAG_RD, &stats->mcc,
2592 			"Multiple collisions");
2593 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
2594 			CTLFLAG_RD, &stats->latecol,
2595 			"Late collisions");
2596 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
2597 			CTLFLAG_RD, &stats->colc,
2598 			"Collision Count");
2599 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
2600 			CTLFLAG_RD, &adapter->stats.symerrs,
2601 			"Symbol Errors");
2602 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
2603 			CTLFLAG_RD, &adapter->stats.sec,
2604 			"Sequence Errors");
2605 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
2606 			CTLFLAG_RD, &adapter->stats.dc,
2607 			"Defer Count");
2608 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
2609 			CTLFLAG_RD, &adapter->stats.mpc,
2610 			"Missed Packets");
2611 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
2612 			CTLFLAG_RD, &adapter->stats.rnbc,
2613 			"Receive No Buffers");
2614 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
2615 			CTLFLAG_RD, &adapter->stats.ruc,
2616 			"Receive Undersize");
2617 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
2618 			CTLFLAG_RD, &adapter->stats.rfc,
2619 			"Fragmented Packets Received ");
2620 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
2621 			CTLFLAG_RD, &adapter->stats.roc,
2622 			"Oversized Packets Received");
2623 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
2624 			CTLFLAG_RD, &adapter->stats.rjc,
2625 			"Recevied Jabber");
2626 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
2627 			CTLFLAG_RD, &adapter->stats.rxerrc,
2628 			"Receive Errors");
2629 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
2630 			CTLFLAG_RD, &adapter->stats.crcerrs,
2631 			"CRC errors");
2632 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
2633 			CTLFLAG_RD, &adapter->stats.algnerrc,
2634 			"Alignment Errors");
2635 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
2636 			CTLFLAG_RD, &adapter->stats.xonrxc,
2637 			"XON Received");
2638 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
2639 			CTLFLAG_RD, &adapter->stats.xontxc,
2640 			"XON Transmitted");
2641 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
2642 			CTLFLAG_RD, &adapter->stats.xoffrxc,
2643 			"XOFF Received");
2644 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
2645 			CTLFLAG_RD, &adapter->stats.xofftxc,
2646 			"XOFF Transmitted");
2647 
2648 	/* Packet Reception Stats */
2649 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
2650 			CTLFLAG_RD, &adapter->stats.tpr,
2651 			"Total Packets Received ");
2652 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
2653 			CTLFLAG_RD, &adapter->stats.gprc,
2654 			"Good Packets Received");
2655 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
2656 			CTLFLAG_RD, &adapter->stats.bprc,
2657 			"Broadcast Packets Received");
2658 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
2659 			CTLFLAG_RD, &adapter->stats.mprc,
2660 			"Multicast Packets Received");
2661 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
2662 			CTLFLAG_RD, &adapter->stats.prc64,
2663 			"64 byte frames received ");
2664 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
2665 			CTLFLAG_RD, &adapter->stats.prc127,
2666 			"65-127 byte frames received");
2667 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
2668 			CTLFLAG_RD, &adapter->stats.prc255,
2669 			"128-255 byte frames received");
2670 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
2671 			CTLFLAG_RD, &adapter->stats.prc511,
2672 			"256-511 byte frames received");
2673 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
2674 			CTLFLAG_RD, &adapter->stats.prc1023,
2675 			"512-1023 byte frames received");
2676 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
2677 			CTLFLAG_RD, &adapter->stats.prc1522,
2678 			"1023-1522 byte frames received");
2679 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
2680 			CTLFLAG_RD, &adapter->stats.gorc,
2681 			"Good Octets Received");
2682 
2683 	/* Packet Transmission Stats */
2684 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
2685 			CTLFLAG_RD, &adapter->stats.gotc,
2686 			"Good Octets Transmitted");
2687 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
2688 			CTLFLAG_RD, &adapter->stats.tpt,
2689 			"Total Packets Transmitted");
2690 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
2691 			CTLFLAG_RD, &adapter->stats.gptc,
2692 			"Good Packets Transmitted");
2693 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
2694 			CTLFLAG_RD, &adapter->stats.bptc,
2695 			"Broadcast Packets Transmitted");
2696 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
2697 			CTLFLAG_RD, &adapter->stats.mptc,
2698 			"Multicast Packets Transmitted");
2699 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
2700 			CTLFLAG_RD, &adapter->stats.ptc64,
2701 			"64 byte frames transmitted ");
2702 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
2703 			CTLFLAG_RD, &adapter->stats.ptc127,
2704 			"65-127 byte frames transmitted");
2705 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
2706 			CTLFLAG_RD, &adapter->stats.ptc255,
2707 			"128-255 byte frames transmitted");
2708 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
2709 			CTLFLAG_RD, &adapter->stats.ptc511,
2710 			"256-511 byte frames transmitted");
2711 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
2712 			CTLFLAG_RD, &adapter->stats.ptc1023,
2713 			"512-1023 byte frames transmitted");
2714 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
2715 			CTLFLAG_RD, &adapter->stats.ptc1522,
2716 			"1024-1522 byte frames transmitted");
2717 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
2718 			CTLFLAG_RD, &adapter->stats.tsctc,
2719 			"TSO Contexts Transmitted");
2720 
2721 	/* Interrupt Stats */
2722 
2723 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
2724 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
2725 	int_list = SYSCTL_CHILDREN(int_node);
2726 
2727 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
2728 			CTLFLAG_RD, &adapter->stats.iac,
2729 			"Interrupt Assertion Count");
2730 
2731 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
2732 			CTLFLAG_RD, &adapter->stats.rxdmtc,
2733 			"Rx Desc Min Thresh Count");
2734 }
2735 
2736 /**********************************************************************
2737  *
2738  *  This routine provides a way to dump out the adapter eeprom,
2739  *  often a useful debug/service tool. This only dumps the first
2740  *  32 words, stuff that matters is in that extent.
2741  *
2742  **********************************************************************/
2743 static int
2744 igc_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
2745 {
2746 	struct igc_adapter *adapter = (struct igc_adapter *)arg1;
2747 	int error;
2748 	int result;
2749 
2750 	result = -1;
2751 	error = sysctl_handle_int(oidp, &result, 0, req);
2752 
2753 	if (error || !req->newptr)
2754 		return (error);
2755 
2756 	/*
2757 	 * This value will cause a hex dump of the
2758 	 * first 32 16-bit words of the EEPROM to
2759 	 * the screen.
2760 	 */
2761 	if (result == 1)
2762 		igc_print_nvm_info(adapter);
2763 
2764 	return (error);
2765 }
2766 
2767 static void
2768 igc_print_nvm_info(struct igc_adapter *adapter)
2769 {
2770 	u16 eeprom_data;
2771 	int i, j, row = 0;
2772 
2773 	/* Its a bit crude, but it gets the job done */
2774 	printf("\nInterface EEPROM Dump:\n");
2775 	printf("Offset\n0x0000  ");
2776 	for (i = 0, j = 0; i < 32; i++, j++) {
2777 		if (j == 8) { /* Make the offset block */
2778 			j = 0; ++row;
2779 			printf("\n0x00%x0  ",row);
2780 		}
2781 		igc_read_nvm(&adapter->hw, i, 1, &eeprom_data);
2782 		printf("%04x ", eeprom_data);
2783 	}
2784 	printf("\n");
2785 }
2786 
2787 static int
2788 igc_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
2789 {
2790 	struct igc_int_delay_info *info;
2791 	struct igc_adapter *adapter;
2792 	u32 regval;
2793 	int error, usecs, ticks;
2794 
2795 	info = (struct igc_int_delay_info *) arg1;
2796 	usecs = info->value;
2797 	error = sysctl_handle_int(oidp, &usecs, 0, req);
2798 	if (error != 0 || req->newptr == NULL)
2799 		return (error);
2800 	if (usecs < 0 || usecs > IGC_TICKS_TO_USECS(65535))
2801 		return (EINVAL);
2802 	info->value = usecs;
2803 	ticks = IGC_USECS_TO_TICKS(usecs);
2804 	if (info->offset == IGC_ITR)	/* units are 256ns here */
2805 		ticks *= 4;
2806 
2807 	adapter = info->adapter;
2808 
2809 	regval = IGC_READ_OFFSET(&adapter->hw, info->offset);
2810 	regval = (regval & ~0xffff) | (ticks & 0xffff);
2811 	/* Handle a few special cases. */
2812 	switch (info->offset) {
2813 	case IGC_RDTR:
2814 		break;
2815 	case IGC_TIDV:
2816 		if (ticks == 0) {
2817 			adapter->txd_cmd &= ~IGC_TXD_CMD_IDE;
2818 			/* Don't write 0 into the TIDV register. */
2819 			regval++;
2820 		} else
2821 			adapter->txd_cmd |= IGC_TXD_CMD_IDE;
2822 		break;
2823 	}
2824 	IGC_WRITE_OFFSET(&adapter->hw, info->offset, regval);
2825 	return (0);
2826 }
2827 
2828 static void
2829 igc_add_int_delay_sysctl(struct igc_adapter *adapter, const char *name,
2830 	const char *description, struct igc_int_delay_info *info,
2831 	int offset, int value)
2832 {
2833 	info->adapter = adapter;
2834 	info->offset = offset;
2835 	info->value = value;
2836 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
2837 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
2838 	    OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2839 	    info, 0, igc_sysctl_int_delay, "I", description);
2840 }
2841 
2842 /*
2843  * Set flow control using sysctl:
2844  * Flow control values:
2845  *      0 - off
2846  *      1 - rx pause
2847  *      2 - tx pause
2848  *      3 - full
2849  */
2850 static int
2851 igc_set_flowcntl(SYSCTL_HANDLER_ARGS)
2852 {
2853 	int error;
2854 	static int input = 3; /* default is full */
2855 	struct igc_adapter	*adapter = (struct igc_adapter *) arg1;
2856 
2857 	error = sysctl_handle_int(oidp, &input, 0, req);
2858 
2859 	if ((error) || (req->newptr == NULL))
2860 		return (error);
2861 
2862 	if (input == adapter->fc) /* no change? */
2863 		return (error);
2864 
2865 	switch (input) {
2866 	case igc_fc_rx_pause:
2867 	case igc_fc_tx_pause:
2868 	case igc_fc_full:
2869 	case igc_fc_none:
2870 		adapter->hw.fc.requested_mode = input;
2871 		adapter->fc = input;
2872 		break;
2873 	default:
2874 		/* Do nothing */
2875 		return (error);
2876 	}
2877 
2878 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
2879 	igc_force_mac_fc(&adapter->hw);
2880 	return (error);
2881 }
2882 
2883 /*
2884  * Manage Energy Efficient Ethernet:
2885  * Control values:
2886  *     0/1 - enabled/disabled
2887  */
2888 static int
2889 igc_sysctl_eee(SYSCTL_HANDLER_ARGS)
2890 {
2891 	struct igc_adapter *adapter = (struct igc_adapter *) arg1;
2892 	int error, value;
2893 
2894 	value = adapter->hw.dev_spec._i225.eee_disable;
2895 	error = sysctl_handle_int(oidp, &value, 0, req);
2896 	if (error || req->newptr == NULL)
2897 		return (error);
2898 
2899 	adapter->hw.dev_spec._i225.eee_disable = (value != 0);
2900 	igc_if_init(adapter->ctx);
2901 
2902 	return (0);
2903 }
2904 
2905 static int
2906 igc_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
2907 {
2908 	struct igc_adapter *adapter;
2909 	int error;
2910 	int result;
2911 
2912 	result = -1;
2913 	error = sysctl_handle_int(oidp, &result, 0, req);
2914 
2915 	if (error || !req->newptr)
2916 		return (error);
2917 
2918 	if (result == 1) {
2919 		adapter = (struct igc_adapter *) arg1;
2920 		igc_print_debug_info(adapter);
2921 	}
2922 
2923 	return (error);
2924 }
2925 
2926 static int
2927 igc_get_rs(SYSCTL_HANDLER_ARGS)
2928 {
2929 	struct igc_adapter *adapter = (struct igc_adapter *) arg1;
2930 	int error;
2931 	int result;
2932 
2933 	result = 0;
2934 	error = sysctl_handle_int(oidp, &result, 0, req);
2935 
2936 	if (error || !req->newptr || result != 1)
2937 		return (error);
2938 	igc_dump_rs(adapter);
2939 
2940 	return (error);
2941 }
2942 
2943 static void
2944 igc_if_debug(if_ctx_t ctx)
2945 {
2946 	igc_dump_rs(iflib_get_softc(ctx));
2947 }
2948 
2949 /*
2950  * This routine is meant to be fluid, add whatever is
2951  * needed for debugging a problem.  -jfv
2952  */
2953 static void
2954 igc_print_debug_info(struct igc_adapter *adapter)
2955 {
2956 	device_t dev = iflib_get_dev(adapter->ctx);
2957 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
2958 	struct tx_ring *txr = &adapter->tx_queues->txr;
2959 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
2960 
2961 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
2962 		printf("Interface is RUNNING ");
2963 	else
2964 		printf("Interface is NOT RUNNING\n");
2965 
2966 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
2967 		printf("and INACTIVE\n");
2968 	else
2969 		printf("and ACTIVE\n");
2970 
2971 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2972 		device_printf(dev, "TX Queue %d ------\n", i);
2973 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
2974 			IGC_READ_REG(&adapter->hw, IGC_TDH(i)),
2975 			IGC_READ_REG(&adapter->hw, IGC_TDT(i)));
2976 
2977 	}
2978 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
2979 		device_printf(dev, "RX Queue %d ------\n", j);
2980 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
2981 			IGC_READ_REG(&adapter->hw, IGC_RDH(j)),
2982 			IGC_READ_REG(&adapter->hw, IGC_RDT(j)));
2983 	}
2984 }
2985