xref: /freebsd/sys/dev/igc/if_igc.c (revision d9a42747950146bf03cda7f6e25d219253f8a57a)
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_I226_LMVP, "Intel(R) Ethernet Controller I226-LMvP"),
63 	PVID(0x8086, IGC_DEV_ID_I225_IT, "Intel(R) Ethernet Controller I225-IT(2)"),
64 	PVID(0x8086, IGC_DEV_ID_I226_LM, "Intel(R) Ethernet Controller I226-LM"),
65 	PVID(0x8086, IGC_DEV_ID_I226_V, "Intel(R) Ethernet Controller I226-V"),
66 	PVID(0x8086, IGC_DEV_ID_I226_IT, "Intel(R) Ethernet Controller I226-IT"),
67 	PVID(0x8086, IGC_DEV_ID_I221_V, "Intel(R) Ethernet Controller I221-V"),
68 	PVID(0x8086, IGC_DEV_ID_I226_BLANK_NVM, "Intel(R) Ethernet Controller I226(blankNVM)"),
69 	PVID(0x8086, IGC_DEV_ID_I225_BLANK_NVM, "Intel(R) Ethernet Controller I225(blankNVM)"),
70 	/* required last entry */
71 	PVID_END
72 };
73 
74 /*********************************************************************
75  *  Function prototypes
76  *********************************************************************/
77 static void	*igc_register(device_t dev);
78 static int	igc_if_attach_pre(if_ctx_t ctx);
79 static int	igc_if_attach_post(if_ctx_t ctx);
80 static int	igc_if_detach(if_ctx_t ctx);
81 static int	igc_if_shutdown(if_ctx_t ctx);
82 static int	igc_if_suspend(if_ctx_t ctx);
83 static int	igc_if_resume(if_ctx_t ctx);
84 
85 static int	igc_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
86 static int	igc_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
87 static void	igc_if_queues_free(if_ctx_t ctx);
88 
89 static uint64_t	igc_if_get_counter(if_ctx_t, ift_counter);
90 static void	igc_if_init(if_ctx_t ctx);
91 static void	igc_if_stop(if_ctx_t ctx);
92 static void	igc_if_media_status(if_ctx_t, struct ifmediareq *);
93 static int	igc_if_media_change(if_ctx_t ctx);
94 static int	igc_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
95 static void	igc_if_timer(if_ctx_t ctx, uint16_t qid);
96 static void	igc_if_watchdog_reset(if_ctx_t ctx);
97 static bool	igc_if_needs_restart(if_ctx_t ctx, enum iflib_restart_event event);
98 
99 static void	igc_identify_hardware(if_ctx_t ctx);
100 static int	igc_allocate_pci_resources(if_ctx_t ctx);
101 static void	igc_free_pci_resources(if_ctx_t ctx);
102 static void	igc_reset(if_ctx_t ctx);
103 static int	igc_setup_interface(if_ctx_t ctx);
104 static int	igc_setup_msix(if_ctx_t ctx);
105 
106 static void	igc_initialize_transmit_unit(if_ctx_t ctx);
107 static void	igc_initialize_receive_unit(if_ctx_t ctx);
108 
109 static void	igc_if_intr_enable(if_ctx_t ctx);
110 static void	igc_if_intr_disable(if_ctx_t ctx);
111 static int	igc_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
112 static int	igc_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
113 static void	igc_if_multi_set(if_ctx_t ctx);
114 static void	igc_if_update_admin_status(if_ctx_t ctx);
115 static void	igc_if_debug(if_ctx_t ctx);
116 static void	igc_update_stats_counters(struct igc_adapter *);
117 static void	igc_add_hw_stats(struct igc_adapter *adapter);
118 static int	igc_if_set_promisc(if_ctx_t ctx, int flags);
119 static void	igc_setup_vlan_hw_support(if_ctx_t ctx);
120 static int	igc_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
121 static void	igc_print_nvm_info(struct igc_adapter *);
122 static int	igc_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
123 static int	igc_get_rs(SYSCTL_HANDLER_ARGS);
124 static void	igc_print_debug_info(struct igc_adapter *);
125 static int 	igc_is_valid_ether_addr(u8 *);
126 static int	igc_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
127 static void	igc_add_int_delay_sysctl(struct igc_adapter *, const char *,
128 		    const char *, struct igc_int_delay_info *, int, int);
129 /* Management and WOL Support */
130 static void	igc_get_hw_control(struct igc_adapter *);
131 static void	igc_release_hw_control(struct igc_adapter *);
132 static void	igc_get_wakeup(if_ctx_t ctx);
133 static void	igc_enable_wakeup(if_ctx_t ctx);
134 
135 int		igc_intr(void *arg);
136 
137 /* MSI-X handlers */
138 static int	igc_if_msix_intr_assign(if_ctx_t, int);
139 static int	igc_msix_link(void *);
140 static void	igc_handle_link(void *context);
141 
142 static int	igc_set_flowcntl(SYSCTL_HANDLER_ARGS);
143 static int	igc_sysctl_eee(SYSCTL_HANDLER_ARGS);
144 
145 static int	igc_get_regs(SYSCTL_HANDLER_ARGS);
146 
147 static void	igc_configure_queues(struct igc_adapter *adapter);
148 
149 
150 /*********************************************************************
151  *  FreeBSD Device Interface Entry Points
152  *********************************************************************/
153 static device_method_t igc_methods[] = {
154 	/* Device interface */
155 	DEVMETHOD(device_register, igc_register),
156 	DEVMETHOD(device_probe, iflib_device_probe),
157 	DEVMETHOD(device_attach, iflib_device_attach),
158 	DEVMETHOD(device_detach, iflib_device_detach),
159 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
160 	DEVMETHOD(device_suspend, iflib_device_suspend),
161 	DEVMETHOD(device_resume, iflib_device_resume),
162 	DEVMETHOD_END
163 };
164 
165 static driver_t igc_driver = {
166 	"igc", igc_methods, sizeof(struct igc_adapter),
167 };
168 
169 DRIVER_MODULE(igc, pci, igc_driver, 0, 0);
170 
171 MODULE_DEPEND(igc, pci, 1, 1, 1);
172 MODULE_DEPEND(igc, ether, 1, 1, 1);
173 MODULE_DEPEND(igc, iflib, 1, 1, 1);
174 
175 IFLIB_PNP_INFO(pci, igc, igc_vendor_info_array);
176 
177 static device_method_t igc_if_methods[] = {
178 	DEVMETHOD(ifdi_attach_pre, igc_if_attach_pre),
179 	DEVMETHOD(ifdi_attach_post, igc_if_attach_post),
180 	DEVMETHOD(ifdi_detach, igc_if_detach),
181 	DEVMETHOD(ifdi_shutdown, igc_if_shutdown),
182 	DEVMETHOD(ifdi_suspend, igc_if_suspend),
183 	DEVMETHOD(ifdi_resume, igc_if_resume),
184 	DEVMETHOD(ifdi_init, igc_if_init),
185 	DEVMETHOD(ifdi_stop, igc_if_stop),
186 	DEVMETHOD(ifdi_msix_intr_assign, igc_if_msix_intr_assign),
187 	DEVMETHOD(ifdi_intr_enable, igc_if_intr_enable),
188 	DEVMETHOD(ifdi_intr_disable, igc_if_intr_disable),
189 	DEVMETHOD(ifdi_tx_queues_alloc, igc_if_tx_queues_alloc),
190 	DEVMETHOD(ifdi_rx_queues_alloc, igc_if_rx_queues_alloc),
191 	DEVMETHOD(ifdi_queues_free, igc_if_queues_free),
192 	DEVMETHOD(ifdi_update_admin_status, igc_if_update_admin_status),
193 	DEVMETHOD(ifdi_multi_set, igc_if_multi_set),
194 	DEVMETHOD(ifdi_media_status, igc_if_media_status),
195 	DEVMETHOD(ifdi_media_change, igc_if_media_change),
196 	DEVMETHOD(ifdi_mtu_set, igc_if_mtu_set),
197 	DEVMETHOD(ifdi_promisc_set, igc_if_set_promisc),
198 	DEVMETHOD(ifdi_timer, igc_if_timer),
199 	DEVMETHOD(ifdi_watchdog_reset, igc_if_watchdog_reset),
200 	DEVMETHOD(ifdi_get_counter, igc_if_get_counter),
201 	DEVMETHOD(ifdi_rx_queue_intr_enable, igc_if_rx_queue_intr_enable),
202 	DEVMETHOD(ifdi_tx_queue_intr_enable, igc_if_tx_queue_intr_enable),
203 	DEVMETHOD(ifdi_debug, igc_if_debug),
204 	DEVMETHOD(ifdi_needs_restart, igc_if_needs_restart),
205 	DEVMETHOD_END
206 };
207 
208 static driver_t igc_if_driver = {
209 	"igc_if", igc_if_methods, sizeof(struct igc_adapter)
210 };
211 
212 /*********************************************************************
213  *  Tunable default values.
214  *********************************************************************/
215 
216 #define IGC_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
217 #define IGC_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
218 
219 #define MAX_INTS_PER_SEC	8000
220 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
221 
222 /* Allow common code without TSO */
223 #ifndef CSUM_TSO
224 #define CSUM_TSO	0
225 #endif
226 
227 static SYSCTL_NODE(_hw, OID_AUTO, igc, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
228     "igc driver parameters");
229 
230 static int igc_disable_crc_stripping = 0;
231 SYSCTL_INT(_hw_igc, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
232     &igc_disable_crc_stripping, 0, "Disable CRC Stripping");
233 
234 static int igc_tx_int_delay_dflt = IGC_TICKS_TO_USECS(IGC_TIDV_VAL);
235 static int igc_rx_int_delay_dflt = IGC_TICKS_TO_USECS(IGC_RDTR_VAL);
236 SYSCTL_INT(_hw_igc, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &igc_tx_int_delay_dflt,
237     0, "Default transmit interrupt delay in usecs");
238 SYSCTL_INT(_hw_igc, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &igc_rx_int_delay_dflt,
239     0, "Default receive interrupt delay in usecs");
240 
241 static int igc_tx_abs_int_delay_dflt = IGC_TICKS_TO_USECS(IGC_TADV_VAL);
242 static int igc_rx_abs_int_delay_dflt = IGC_TICKS_TO_USECS(IGC_RADV_VAL);
243 SYSCTL_INT(_hw_igc, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
244     &igc_tx_abs_int_delay_dflt, 0,
245     "Default transmit interrupt delay limit in usecs");
246 SYSCTL_INT(_hw_igc, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
247     &igc_rx_abs_int_delay_dflt, 0,
248     "Default receive interrupt delay limit in usecs");
249 
250 static int igc_smart_pwr_down = false;
251 SYSCTL_INT(_hw_igc, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &igc_smart_pwr_down,
252     0, "Set to true to leave smart power down enabled on newer adapters");
253 
254 /* Controls whether promiscuous also shows bad packets */
255 static int igc_debug_sbp = true;
256 SYSCTL_INT(_hw_igc, OID_AUTO, sbp, CTLFLAG_RDTUN, &igc_debug_sbp, 0,
257     "Show bad packets in promiscuous mode");
258 
259 /* How many packets rxeof tries to clean at a time */
260 static int igc_rx_process_limit = 100;
261 SYSCTL_INT(_hw_igc, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
262     &igc_rx_process_limit, 0,
263     "Maximum number of received packets to process "
264     "at a time, -1 means unlimited");
265 
266 /* Energy efficient ethernet - default to OFF */
267 static int igc_eee_setting = 1;
268 SYSCTL_INT(_hw_igc, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &igc_eee_setting, 0,
269     "Enable Energy Efficient Ethernet");
270 
271 /*
272 ** Tuneable Interrupt rate
273 */
274 static int igc_max_interrupt_rate = 8000;
275 SYSCTL_INT(_hw_igc, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
276     &igc_max_interrupt_rate, 0, "Maximum interrupts per second");
277 
278 extern struct if_txrx igc_txrx;
279 
280 static struct if_shared_ctx igc_sctx_init = {
281 	.isc_magic = IFLIB_MAGIC,
282 	.isc_q_align = PAGE_SIZE,
283 	.isc_tx_maxsize = IGC_TSO_SIZE + sizeof(struct ether_vlan_header),
284 	.isc_tx_maxsegsize = PAGE_SIZE,
285 	.isc_tso_maxsize = IGC_TSO_SIZE + sizeof(struct ether_vlan_header),
286 	.isc_tso_maxsegsize = IGC_TSO_SEG_SIZE,
287 	.isc_rx_maxsize = MAX_JUMBO_FRAME_SIZE,
288 	.isc_rx_nsegments = 1,
289 	.isc_rx_maxsegsize = MJUM9BYTES,
290 	.isc_nfl = 1,
291 	.isc_nrxqs = 1,
292 	.isc_ntxqs = 1,
293 	.isc_admin_intrcnt = 1,
294 	.isc_vendor_info = igc_vendor_info_array,
295 	.isc_driver_version = "1",
296 	.isc_driver = &igc_if_driver,
297 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
298 
299 	.isc_nrxd_min = {IGC_MIN_RXD},
300 	.isc_ntxd_min = {IGC_MIN_TXD},
301 	.isc_nrxd_max = {IGC_MAX_RXD},
302 	.isc_ntxd_max = {IGC_MAX_TXD},
303 	.isc_nrxd_default = {IGC_DEFAULT_RXD},
304 	.isc_ntxd_default = {IGC_DEFAULT_TXD},
305 };
306 
307 /*****************************************************************
308  *
309  * Dump Registers
310  *
311  ****************************************************************/
312 #define IGC_REGS_LEN 739
313 
314 static int igc_get_regs(SYSCTL_HANDLER_ARGS)
315 {
316 	struct igc_adapter *adapter = (struct igc_adapter *)arg1;
317 	struct igc_hw *hw = &adapter->hw;
318 	struct sbuf *sb;
319 	u32 *regs_buff;
320 	int rc;
321 
322 	regs_buff = malloc(sizeof(u32) * IGC_REGS_LEN, M_DEVBUF, M_WAITOK);
323 	memset(regs_buff, 0, IGC_REGS_LEN * sizeof(u32));
324 
325 	rc = sysctl_wire_old_buffer(req, 0);
326 	MPASS(rc == 0);
327 	if (rc != 0) {
328 		free(regs_buff, M_DEVBUF);
329 		return (rc);
330 	}
331 
332 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
333 	MPASS(sb != NULL);
334 	if (sb == NULL) {
335 		free(regs_buff, M_DEVBUF);
336 		return (ENOMEM);
337 	}
338 
339 	/* General Registers */
340 	regs_buff[0] = IGC_READ_REG(hw, IGC_CTRL);
341 	regs_buff[1] = IGC_READ_REG(hw, IGC_STATUS);
342 	regs_buff[2] = IGC_READ_REG(hw, IGC_CTRL_EXT);
343 	regs_buff[3] = IGC_READ_REG(hw, IGC_ICR);
344 	regs_buff[4] = IGC_READ_REG(hw, IGC_RCTL);
345 	regs_buff[5] = IGC_READ_REG(hw, IGC_RDLEN(0));
346 	regs_buff[6] = IGC_READ_REG(hw, IGC_RDH(0));
347 	regs_buff[7] = IGC_READ_REG(hw, IGC_RDT(0));
348 	regs_buff[8] = IGC_READ_REG(hw, IGC_RXDCTL(0));
349 	regs_buff[9] = IGC_READ_REG(hw, IGC_RDBAL(0));
350 	regs_buff[10] = IGC_READ_REG(hw, IGC_RDBAH(0));
351 	regs_buff[11] = IGC_READ_REG(hw, IGC_TCTL);
352 	regs_buff[12] = IGC_READ_REG(hw, IGC_TDBAL(0));
353 	regs_buff[13] = IGC_READ_REG(hw, IGC_TDBAH(0));
354 	regs_buff[14] = IGC_READ_REG(hw, IGC_TDLEN(0));
355 	regs_buff[15] = IGC_READ_REG(hw, IGC_TDH(0));
356 	regs_buff[16] = IGC_READ_REG(hw, IGC_TDT(0));
357 	regs_buff[17] = IGC_READ_REG(hw, IGC_TXDCTL(0));
358 
359 	sbuf_printf(sb, "General Registers\n");
360 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
361 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
362 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
363 
364 	sbuf_printf(sb, "Interrupt Registers\n");
365 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
366 
367 	sbuf_printf(sb, "RX Registers\n");
368 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
369 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
370 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
371 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
372 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
373 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
374 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
375 
376 	sbuf_printf(sb, "TX Registers\n");
377 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
378 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
379 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
380 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
381 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
382 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
383 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
384 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
385 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
386 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
387 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
388 
389 	free(regs_buff, M_DEVBUF);
390 
391 #ifdef DUMP_DESCS
392 	{
393 		if_softc_ctx_t scctx = adapter->shared;
394 		struct rx_ring *rxr = &rx_que->rxr;
395 		struct tx_ring *txr = &tx_que->txr;
396 		int ntxd = scctx->isc_ntxd[0];
397 		int nrxd = scctx->isc_nrxd[0];
398 		int j;
399 
400 	for (j = 0; j < nrxd; j++) {
401 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
402 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
403 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
404 	}
405 
406 	for (j = 0; j < min(ntxd, 256); j++) {
407 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
408 
409 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
410 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
411 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & IGC_TXD_STAT_DD : 0);
412 
413 	}
414 	}
415 #endif
416 
417 	rc = sbuf_finish(sb);
418 	sbuf_delete(sb);
419 	return(rc);
420 }
421 
422 static void *
423 igc_register(device_t dev)
424 {
425 	return (&igc_sctx_init);
426 }
427 
428 static int
429 igc_set_num_queues(if_ctx_t ctx)
430 {
431 	int maxqueues;
432 
433 	maxqueues = 4;
434 
435 	return (maxqueues);
436 }
437 
438 #define	IGC_CAPS							\
439     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
440     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_TSO4 | IFCAP_LRO |		\
441     IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 | IFCAP_TSO6
442 
443 /*********************************************************************
444  *  Device initialization routine
445  *
446  *  The attach entry point is called when the driver is being loaded.
447  *  This routine identifies the type of hardware, allocates all resources
448  *  and initializes the hardware.
449  *
450  *  return 0 on success, positive on failure
451  *********************************************************************/
452 static int
453 igc_if_attach_pre(if_ctx_t ctx)
454 {
455 	struct igc_adapter *adapter;
456 	if_softc_ctx_t scctx;
457 	device_t dev;
458 	struct igc_hw *hw;
459 	int error = 0;
460 
461 	INIT_DEBUGOUT("igc_if_attach_pre: begin");
462 	dev = iflib_get_dev(ctx);
463 	adapter = iflib_get_softc(ctx);
464 
465 	adapter->ctx = adapter->osdep.ctx = ctx;
466 	adapter->dev = adapter->osdep.dev = dev;
467 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
468 	adapter->media = iflib_get_media(ctx);
469 	hw = &adapter->hw;
470 
471 	adapter->tx_process_limit = scctx->isc_ntxd[0];
472 
473 	/* SYSCTL stuff */
474 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
475 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
476 	    OID_AUTO, "nvm", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
477 	    adapter, 0, igc_sysctl_nvm_info, "I", "NVM Information");
478 
479 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
480 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
481 	    OID_AUTO, "debug", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
482 	    adapter, 0, igc_sysctl_debug_info, "I", "Debug Information");
483 
484 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
485 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
486 	    OID_AUTO, "fc", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
487 	    adapter, 0, igc_set_flowcntl, "I", "Flow Control");
488 
489 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
490 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
491 	    OID_AUTO, "reg_dump",
492 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter, 0,
493 	    igc_get_regs, "A", "Dump Registers");
494 
495 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
496 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
497 	    OID_AUTO, "rs_dump",
498 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, adapter, 0,
499 	    igc_get_rs, "I", "Dump RS indexes");
500 
501 	/* Determine hardware and mac info */
502 	igc_identify_hardware(ctx);
503 
504 	scctx->isc_tx_nsegments = IGC_MAX_SCATTER;
505 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = igc_set_num_queues(ctx);
506 	if (bootverbose)
507 		device_printf(dev, "attach_pre capping queues at %d\n",
508 		    scctx->isc_ntxqsets_max);
509 
510 	scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union igc_adv_tx_desc), IGC_DBA_ALIGN);
511 	scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union igc_adv_rx_desc), IGC_DBA_ALIGN);
512 	scctx->isc_txd_size[0] = sizeof(union igc_adv_tx_desc);
513 	scctx->isc_rxd_size[0] = sizeof(union igc_adv_rx_desc);
514 	scctx->isc_txrx = &igc_txrx;
515 	scctx->isc_tx_tso_segments_max = IGC_MAX_SCATTER;
516 	scctx->isc_tx_tso_size_max = IGC_TSO_SIZE;
517 	scctx->isc_tx_tso_segsize_max = IGC_TSO_SEG_SIZE;
518 	scctx->isc_capabilities = scctx->isc_capenable = IGC_CAPS;
519 	scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
520 		CSUM_IP6_TCP | CSUM_IP6_UDP | CSUM_SCTP | CSUM_IP6_SCTP;
521 
522 	/*
523 	** Some new devices, as with ixgbe, now may
524 	** use a different BAR, so we need to keep
525 	** track of which is used.
526 	*/
527 	scctx->isc_msix_bar = PCIR_BAR(IGC_MSIX_BAR);
528 	if (pci_read_config(dev, scctx->isc_msix_bar, 4) == 0)
529 		scctx->isc_msix_bar += 4;
530 
531 	/* Setup PCI resources */
532 	if (igc_allocate_pci_resources(ctx)) {
533 		device_printf(dev, "Allocation of PCI resources failed\n");
534 		error = ENXIO;
535 		goto err_pci;
536 	}
537 
538 	/* Do Shared Code initialization */
539 	error = igc_setup_init_funcs(hw, true);
540 	if (error) {
541 		device_printf(dev, "Setup of Shared code failed, error %d\n",
542 		    error);
543 		error = ENXIO;
544 		goto err_pci;
545 	}
546 
547 	igc_setup_msix(ctx);
548 	igc_get_bus_info(hw);
549 
550 	/* Set up some sysctls for the tunable interrupt delays */
551 	igc_add_int_delay_sysctl(adapter, "rx_int_delay",
552 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
553 	    IGC_REGISTER(hw, IGC_RDTR), igc_rx_int_delay_dflt);
554 	igc_add_int_delay_sysctl(adapter, "tx_int_delay",
555 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
556 	    IGC_REGISTER(hw, IGC_TIDV), igc_tx_int_delay_dflt);
557 	igc_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
558 	    "receive interrupt delay limit in usecs",
559 	    &adapter->rx_abs_int_delay,
560 	    IGC_REGISTER(hw, IGC_RADV),
561 	    igc_rx_abs_int_delay_dflt);
562 	igc_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
563 	    "transmit interrupt delay limit in usecs",
564 	    &adapter->tx_abs_int_delay,
565 	    IGC_REGISTER(hw, IGC_TADV),
566 	    igc_tx_abs_int_delay_dflt);
567 	igc_add_int_delay_sysctl(adapter, "itr",
568 	    "interrupt delay limit in usecs/4",
569 	    &adapter->tx_itr,
570 	    IGC_REGISTER(hw, IGC_ITR),
571 	    DEFAULT_ITR);
572 
573 	hw->mac.autoneg = DO_AUTO_NEG;
574 	hw->phy.autoneg_wait_to_complete = false;
575 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
576 
577 	/* Copper options */
578 	if (hw->phy.media_type == igc_media_type_copper) {
579 		hw->phy.mdix = AUTO_ALL_MODES;
580 	}
581 
582 	/*
583 	 * Set the frame limits assuming
584 	 * standard ethernet sized frames.
585 	 */
586 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
587 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
588 
589 	/* Allocate multicast array memory. */
590 	adapter->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
591 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
592 	if (adapter->mta == NULL) {
593 		device_printf(dev, "Can not allocate multicast setup array\n");
594 		error = ENOMEM;
595 		goto err_late;
596 	}
597 
598 	/* Check SOL/IDER usage */
599 	if (igc_check_reset_block(hw))
600 		device_printf(dev, "PHY reset is blocked"
601 			      " due to SOL/IDER session.\n");
602 
603 	/* Sysctl for setting Energy Efficient Ethernet */
604 	adapter->hw.dev_spec._i225.eee_disable = igc_eee_setting;
605 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
606 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
607 	    OID_AUTO, "eee_control",
608 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
609 	    adapter, 0, igc_sysctl_eee, "I",
610 	    "Disable Energy Efficient Ethernet");
611 
612 	/*
613 	** Start from a known state, this is
614 	** important in reading the nvm and
615 	** mac from that.
616 	*/
617 	igc_reset_hw(hw);
618 
619 	/* Make sure we have a good EEPROM before we read from it */
620 	if (igc_validate_nvm_checksum(hw) < 0) {
621 		/*
622 		** Some PCI-E parts fail the first check due to
623 		** the link being in sleep state, call it again,
624 		** if it fails a second time its a real issue.
625 		*/
626 		if (igc_validate_nvm_checksum(hw) < 0) {
627 			device_printf(dev,
628 			    "The EEPROM Checksum Is Not Valid\n");
629 			error = EIO;
630 			goto err_late;
631 		}
632 	}
633 
634 	/* Copy the permanent MAC address out of the EEPROM */
635 	if (igc_read_mac_addr(hw) < 0) {
636 		device_printf(dev, "EEPROM read error while reading MAC"
637 			      " address\n");
638 		error = EIO;
639 		goto err_late;
640 	}
641 
642 	if (!igc_is_valid_ether_addr(hw->mac.addr)) {
643 		device_printf(dev, "Invalid MAC address\n");
644 		error = EIO;
645 		goto err_late;
646 	}
647 
648 	/*
649 	 * Get Wake-on-Lan and Management info for later use
650 	 */
651 	igc_get_wakeup(ctx);
652 
653 	/* Enable only WOL MAGIC by default */
654 	scctx->isc_capenable &= ~IFCAP_WOL;
655 	if (adapter->wol != 0)
656 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
657 
658 	iflib_set_mac(ctx, hw->mac.addr);
659 
660 	return (0);
661 
662 err_late:
663 	igc_release_hw_control(adapter);
664 err_pci:
665 	igc_free_pci_resources(ctx);
666 	free(adapter->mta, M_DEVBUF);
667 
668 	return (error);
669 }
670 
671 static int
672 igc_if_attach_post(if_ctx_t ctx)
673 {
674 	struct igc_adapter *adapter = iflib_get_softc(ctx);
675 	struct igc_hw *hw = &adapter->hw;
676 	int error = 0;
677 
678 	/* Setup OS specific network interface */
679 	error = igc_setup_interface(ctx);
680 	if (error != 0) {
681 		goto err_late;
682 	}
683 
684 	igc_reset(ctx);
685 
686 	/* Initialize statistics */
687 	igc_update_stats_counters(adapter);
688 	hw->mac.get_link_status = true;
689 	igc_if_update_admin_status(ctx);
690 	igc_add_hw_stats(adapter);
691 
692 	/* the driver can now take control from firmware */
693 	igc_get_hw_control(adapter);
694 
695 	INIT_DEBUGOUT("igc_if_attach_post: end");
696 
697 	return (error);
698 
699 err_late:
700 	igc_release_hw_control(adapter);
701 	igc_free_pci_resources(ctx);
702 	igc_if_queues_free(ctx);
703 	free(adapter->mta, M_DEVBUF);
704 
705 	return (error);
706 }
707 
708 /*********************************************************************
709  *  Device removal routine
710  *
711  *  The detach entry point is called when the driver is being removed.
712  *  This routine stops the adapter and deallocates all the resources
713  *  that were allocated for driver operation.
714  *
715  *  return 0 on success, positive on failure
716  *********************************************************************/
717 static int
718 igc_if_detach(if_ctx_t ctx)
719 {
720 	struct igc_adapter	*adapter = iflib_get_softc(ctx);
721 
722 	INIT_DEBUGOUT("igc_if_detach: begin");
723 
724 	igc_phy_hw_reset(&adapter->hw);
725 
726 	igc_release_hw_control(adapter);
727 	igc_free_pci_resources(ctx);
728 
729 	return (0);
730 }
731 
732 /*********************************************************************
733  *
734  *  Shutdown entry point
735  *
736  **********************************************************************/
737 
738 static int
739 igc_if_shutdown(if_ctx_t ctx)
740 {
741 	return igc_if_suspend(ctx);
742 }
743 
744 /*
745  * Suspend/resume device methods.
746  */
747 static int
748 igc_if_suspend(if_ctx_t ctx)
749 {
750 	struct igc_adapter *adapter = iflib_get_softc(ctx);
751 
752 	igc_release_hw_control(adapter);
753 	igc_enable_wakeup(ctx);
754 	return (0);
755 }
756 
757 static int
758 igc_if_resume(if_ctx_t ctx)
759 {
760 	igc_if_init(ctx);
761 
762 	return(0);
763 }
764 
765 static int
766 igc_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
767 {
768 	int max_frame_size;
769 	struct igc_adapter *adapter = iflib_get_softc(ctx);
770 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
771 
772 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
773 
774 	 /* 9K Jumbo Frame size */
775 	 max_frame_size = 9234;
776 
777 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
778 		return (EINVAL);
779 	}
780 
781 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
782 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
783 	return (0);
784 }
785 
786 /*********************************************************************
787  *  Init entry point
788  *
789  *  This routine is used in two ways. It is used by the stack as
790  *  init entry point in network interface structure. It is also used
791  *  by the driver as a hw/sw initialization routine to get to a
792  *  consistent state.
793  *
794  **********************************************************************/
795 static void
796 igc_if_init(if_ctx_t ctx)
797 {
798 	struct igc_adapter *adapter = iflib_get_softc(ctx);
799 	if_softc_ctx_t scctx = adapter->shared;
800 	if_t ifp = iflib_get_ifp(ctx);
801 	struct igc_tx_queue *tx_que;
802 	int i;
803 
804 	INIT_DEBUGOUT("igc_if_init: begin");
805 
806 	/* Get the latest mac address, User can use a LAA */
807 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
808 	    ETHER_ADDR_LEN);
809 
810 	/* Put the address into the Receive Address Array */
811 	igc_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
812 
813 	/* Initialize the hardware */
814 	igc_reset(ctx);
815 	igc_if_update_admin_status(ctx);
816 
817 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
818 		struct tx_ring *txr = &tx_que->txr;
819 
820 		txr->tx_rs_cidx = txr->tx_rs_pidx;
821 
822 		/* Initialize the last processed descriptor to be the end of
823 		 * the ring, rather than the start, so that we avoid an
824 		 * off-by-one error when calculating how many descriptors are
825 		 * done in the credits_update function.
826 		 */
827 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
828 	}
829 
830 	/* Setup VLAN support, basic and offload if available */
831 	IGC_WRITE_REG(&adapter->hw, IGC_VET, ETHERTYPE_VLAN);
832 
833 	/* Prepare transmit descriptors and buffers */
834 	igc_initialize_transmit_unit(ctx);
835 
836 	/* Setup Multicast table */
837 	igc_if_multi_set(ctx);
838 
839 	adapter->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
840 	igc_initialize_receive_unit(ctx);
841 
842 	/* Set up VLAN support */
843 	igc_setup_vlan_hw_support(ctx);
844 
845 	/* Don't lose promiscuous settings */
846 	igc_if_set_promisc(ctx, if_getflags(ifp));
847 	igc_clear_hw_cntrs_base_generic(&adapter->hw);
848 
849 	if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
850 		igc_configure_queues(adapter);
851 
852 	/* this clears any pending interrupts */
853 	IGC_READ_REG(&adapter->hw, IGC_ICR);
854 	IGC_WRITE_REG(&adapter->hw, IGC_ICS, IGC_ICS_LSC);
855 
856 	/* the driver can now take control from firmware */
857 	igc_get_hw_control(adapter);
858 
859 	/* Set Energy Efficient Ethernet */
860 	igc_set_eee_i225(&adapter->hw, true, true, true);
861 }
862 
863 /*********************************************************************
864  *
865  *  Fast Legacy/MSI Combined Interrupt Service routine
866  *
867  *********************************************************************/
868 int
869 igc_intr(void *arg)
870 {
871 	struct igc_adapter *adapter = arg;
872 	if_ctx_t ctx = adapter->ctx;
873 	u32 reg_icr;
874 
875 	reg_icr = IGC_READ_REG(&adapter->hw, IGC_ICR);
876 
877 	/* Hot eject? */
878 	if (reg_icr == 0xffffffff)
879 		return FILTER_STRAY;
880 
881 	/* Definitely not our interrupt. */
882 	if (reg_icr == 0x0)
883 		return FILTER_STRAY;
884 
885 	if ((reg_icr & IGC_ICR_INT_ASSERTED) == 0)
886 		return FILTER_STRAY;
887 
888 	/*
889 	 * Only MSI-X interrupts have one-shot behavior by taking advantage
890 	 * of the EIAC register.  Thus, explicitly disable interrupts.  This
891 	 * also works around the MSI message reordering errata on certain
892 	 * systems.
893 	 */
894 	IFDI_INTR_DISABLE(ctx);
895 
896 	/* Link status change */
897 	if (reg_icr & (IGC_ICR_RXSEQ | IGC_ICR_LSC))
898 		igc_handle_link(ctx);
899 
900 	if (reg_icr & IGC_ICR_RXO)
901 		adapter->rx_overruns++;
902 
903 	return (FILTER_SCHEDULE_THREAD);
904 }
905 
906 static int
907 igc_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
908 {
909 	struct igc_adapter *adapter = iflib_get_softc(ctx);
910 	struct igc_rx_queue *rxq = &adapter->rx_queues[rxqid];
911 
912 	IGC_WRITE_REG(&adapter->hw, IGC_EIMS, rxq->eims);
913 	return (0);
914 }
915 
916 static int
917 igc_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
918 {
919 	struct igc_adapter *adapter = iflib_get_softc(ctx);
920 	struct igc_tx_queue *txq = &adapter->tx_queues[txqid];
921 
922 	IGC_WRITE_REG(&adapter->hw, IGC_EIMS, txq->eims);
923 	return (0);
924 }
925 
926 /*********************************************************************
927  *
928  *  MSI-X RX Interrupt Service routine
929  *
930  **********************************************************************/
931 static int
932 igc_msix_que(void *arg)
933 {
934 	struct igc_rx_queue *que = arg;
935 
936 	++que->irqs;
937 
938 	return (FILTER_SCHEDULE_THREAD);
939 }
940 
941 /*********************************************************************
942  *
943  *  MSI-X Link Fast Interrupt Service routine
944  *
945  **********************************************************************/
946 static int
947 igc_msix_link(void *arg)
948 {
949 	struct igc_adapter *adapter = arg;
950 	u32 reg_icr;
951 
952 	++adapter->link_irq;
953 	MPASS(adapter->hw.back != NULL);
954 	reg_icr = IGC_READ_REG(&adapter->hw, IGC_ICR);
955 
956 	if (reg_icr & IGC_ICR_RXO)
957 		adapter->rx_overruns++;
958 
959 	if (reg_icr & (IGC_ICR_RXSEQ | IGC_ICR_LSC)) {
960 		igc_handle_link(adapter->ctx);
961 	}
962 
963 	IGC_WRITE_REG(&adapter->hw, IGC_IMS, IGC_IMS_LSC);
964 	IGC_WRITE_REG(&adapter->hw, IGC_EIMS, adapter->link_mask);
965 
966 	return (FILTER_HANDLED);
967 }
968 
969 static void
970 igc_handle_link(void *context)
971 {
972 	if_ctx_t ctx = context;
973 	struct igc_adapter *adapter = iflib_get_softc(ctx);
974 
975 	adapter->hw.mac.get_link_status = true;
976 	iflib_admin_intr_deferred(ctx);
977 }
978 
979 /*********************************************************************
980  *
981  *  Media Ioctl callback
982  *
983  *  This routine is called whenever the user queries the status of
984  *  the interface using ifconfig.
985  *
986  **********************************************************************/
987 static void
988 igc_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
989 {
990 	struct igc_adapter *adapter = iflib_get_softc(ctx);
991 
992 	INIT_DEBUGOUT("igc_if_media_status: begin");
993 
994 	iflib_admin_intr_deferred(ctx);
995 
996 	ifmr->ifm_status = IFM_AVALID;
997 	ifmr->ifm_active = IFM_ETHER;
998 
999 	if (!adapter->link_active) {
1000 		return;
1001 	}
1002 
1003 	ifmr->ifm_status |= IFM_ACTIVE;
1004 
1005 	switch (adapter->link_speed) {
1006 	case 10:
1007 		ifmr->ifm_active |= IFM_10_T;
1008 		break;
1009 	case 100:
1010 		ifmr->ifm_active |= IFM_100_TX;
1011                 break;
1012 	case 1000:
1013 		ifmr->ifm_active |= IFM_1000_T;
1014 		break;
1015 	case 2500:
1016                 ifmr->ifm_active |= IFM_2500_T;
1017                 break;
1018 	}
1019 
1020 	if (adapter->link_duplex == FULL_DUPLEX)
1021 		ifmr->ifm_active |= IFM_FDX;
1022 	else
1023 		ifmr->ifm_active |= IFM_HDX;
1024 }
1025 
1026 /*********************************************************************
1027  *
1028  *  Media Ioctl callback
1029  *
1030  *  This routine is called when the user changes speed/duplex using
1031  *  media/mediopt option with ifconfig.
1032  *
1033  **********************************************************************/
1034 static int
1035 igc_if_media_change(if_ctx_t ctx)
1036 {
1037 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1038 	struct ifmedia *ifm = iflib_get_media(ctx);
1039 
1040 	INIT_DEBUGOUT("igc_if_media_change: begin");
1041 
1042 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1043 		return (EINVAL);
1044 
1045 	adapter->hw.mac.autoneg = DO_AUTO_NEG;
1046 
1047 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1048 	case IFM_AUTO:
1049 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1050 		break;
1051         case IFM_2500_T:
1052                 adapter->hw.phy.autoneg_advertised = ADVERTISE_2500_FULL;
1053                 break;
1054 	case IFM_1000_T:
1055 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1056 		break;
1057 	case IFM_100_TX:
1058 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1059 			adapter->hw.phy.autoneg_advertised = ADVERTISE_100_FULL;
1060 		else
1061 			adapter->hw.phy.autoneg_advertised = ADVERTISE_100_HALF;
1062 		break;
1063 	case IFM_10_T:
1064 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1065 			adapter->hw.phy.autoneg_advertised = ADVERTISE_10_FULL;
1066 		else
1067 			adapter->hw.phy.autoneg_advertised = ADVERTISE_10_HALF;
1068 		break;
1069 	default:
1070 		device_printf(adapter->dev, "Unsupported media type\n");
1071 	}
1072 
1073 	igc_if_init(ctx);
1074 
1075 	return (0);
1076 }
1077 
1078 static int
1079 igc_if_set_promisc(if_ctx_t ctx, int flags)
1080 {
1081 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1082 	if_t ifp = iflib_get_ifp(ctx);
1083 	u32 reg_rctl;
1084 	int mcnt = 0;
1085 
1086 	reg_rctl = IGC_READ_REG(&adapter->hw, IGC_RCTL);
1087 	reg_rctl &= ~(IGC_RCTL_SBP | IGC_RCTL_UPE);
1088 	if (flags & IFF_ALLMULTI)
1089 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1090 	else
1091 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
1092 
1093 	/* Don't disable if in MAX groups */
1094 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1095 		reg_rctl &=  (~IGC_RCTL_MPE);
1096 	IGC_WRITE_REG(&adapter->hw, IGC_RCTL, reg_rctl);
1097 
1098 	if (flags & IFF_PROMISC) {
1099 		reg_rctl |= (IGC_RCTL_UPE | IGC_RCTL_MPE);
1100 		/* Turn this on if you want to see bad packets */
1101 		if (igc_debug_sbp)
1102 			reg_rctl |= IGC_RCTL_SBP;
1103 		IGC_WRITE_REG(&adapter->hw, IGC_RCTL, reg_rctl);
1104 	} else if (flags & IFF_ALLMULTI) {
1105 		reg_rctl |= IGC_RCTL_MPE;
1106 		reg_rctl &= ~IGC_RCTL_UPE;
1107 		IGC_WRITE_REG(&adapter->hw, IGC_RCTL, reg_rctl);
1108 	}
1109 	return (0);
1110 }
1111 
1112 static u_int
1113 igc_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int idx)
1114 {
1115 	u8 *mta = arg;
1116 
1117 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
1118 		return (0);
1119 
1120 	bcopy(LLADDR(sdl), &mta[idx * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1121 
1122 	return (1);
1123 }
1124 
1125 /*********************************************************************
1126  *  Multicast Update
1127  *
1128  *  This routine is called whenever multicast address list is updated.
1129  *
1130  **********************************************************************/
1131 
1132 static void
1133 igc_if_multi_set(if_ctx_t ctx)
1134 {
1135 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1136 	if_t ifp = iflib_get_ifp(ctx);
1137 	u8  *mta; /* Multicast array memory */
1138 	u32 reg_rctl = 0;
1139 	int mcnt = 0;
1140 
1141 	IOCTL_DEBUGOUT("igc_set_multi: begin");
1142 
1143 	mta = adapter->mta;
1144 	bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1145 
1146 	mcnt = if_foreach_llmaddr(ifp, igc_copy_maddr, mta);
1147 
1148 	reg_rctl = IGC_READ_REG(&adapter->hw, IGC_RCTL);
1149 
1150 	if (if_getflags(ifp) & IFF_PROMISC) {
1151 		reg_rctl |= (IGC_RCTL_UPE | IGC_RCTL_MPE);
1152 		/* Turn this on if you want to see bad packets */
1153 		if (igc_debug_sbp)
1154 			reg_rctl |= IGC_RCTL_SBP;
1155 	} else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
1156 	      if_getflags(ifp) & IFF_ALLMULTI) {
1157                 reg_rctl |= IGC_RCTL_MPE;
1158 		reg_rctl &= ~IGC_RCTL_UPE;
1159         } else
1160 		reg_rctl &= ~(IGC_RCTL_UPE | IGC_RCTL_MPE);
1161 
1162 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1163 		igc_update_mc_addr_list(&adapter->hw, mta, mcnt);
1164 
1165 	IGC_WRITE_REG(&adapter->hw, IGC_RCTL, reg_rctl);
1166 }
1167 
1168 /*********************************************************************
1169  *  Timer routine
1170  *
1171  *  This routine schedules igc_if_update_admin_status() to check for
1172  *  link status and to gather statistics as well as to perform some
1173  *  controller-specific hardware patting.
1174  *
1175  **********************************************************************/
1176 static void
1177 igc_if_timer(if_ctx_t ctx, uint16_t qid)
1178 {
1179 
1180 	if (qid != 0)
1181 		return;
1182 
1183 	iflib_admin_intr_deferred(ctx);
1184 }
1185 
1186 static void
1187 igc_if_update_admin_status(if_ctx_t ctx)
1188 {
1189 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1190 	struct igc_hw *hw = &adapter->hw;
1191 	device_t dev = iflib_get_dev(ctx);
1192 	u32 link_check, thstat, ctrl;
1193 
1194 	link_check = thstat = ctrl = 0;
1195 	/* Get the cached link value or read phy for real */
1196 	switch (hw->phy.media_type) {
1197 	case igc_media_type_copper:
1198 		if (hw->mac.get_link_status == true) {
1199 			/* Do the work to read phy */
1200 			igc_check_for_link(hw);
1201 			link_check = !hw->mac.get_link_status;
1202 		} else
1203 			link_check = true;
1204 		break;
1205 	case igc_media_type_unknown:
1206 		igc_check_for_link(hw);
1207 		link_check = !hw->mac.get_link_status;
1208 		/* FALLTHROUGH */
1209 	default:
1210 		break;
1211 	}
1212 
1213 	/* Now check for a transition */
1214 	if (link_check && (adapter->link_active == 0)) {
1215 		igc_get_speed_and_duplex(hw, &adapter->link_speed,
1216 		    &adapter->link_duplex);
1217 		if (bootverbose)
1218 			device_printf(dev, "Link is up %d Mbps %s\n",
1219 			    adapter->link_speed,
1220 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1221 			    "Full Duplex" : "Half Duplex"));
1222 		adapter->link_active = 1;
1223 		iflib_link_state_change(ctx, LINK_STATE_UP,
1224 		    IF_Mbps(adapter->link_speed));
1225 	} else if (!link_check && (adapter->link_active == 1)) {
1226 		adapter->link_speed = 0;
1227 		adapter->link_duplex = 0;
1228 		adapter->link_active = 0;
1229 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1230 	}
1231 	igc_update_stats_counters(adapter);
1232 }
1233 
1234 static void
1235 igc_if_watchdog_reset(if_ctx_t ctx)
1236 {
1237 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1238 
1239 	/*
1240 	 * Just count the event; iflib(4) will already trigger a
1241 	 * sufficient reset of the controller.
1242 	 */
1243 	adapter->watchdog_events++;
1244 }
1245 
1246 /*********************************************************************
1247  *
1248  *  This routine disables all traffic on the adapter by issuing a
1249  *  global reset on the MAC.
1250  *
1251  **********************************************************************/
1252 static void
1253 igc_if_stop(if_ctx_t ctx)
1254 {
1255 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1256 
1257 	INIT_DEBUGOUT("igc_if_stop: begin");
1258 
1259 	igc_reset_hw(&adapter->hw);
1260 	IGC_WRITE_REG(&adapter->hw, IGC_WUC, 0);
1261 }
1262 
1263 /*********************************************************************
1264  *
1265  *  Determine hardware revision.
1266  *
1267  **********************************************************************/
1268 static void
1269 igc_identify_hardware(if_ctx_t ctx)
1270 {
1271 	device_t dev = iflib_get_dev(ctx);
1272 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1273 
1274 	/* Make sure our PCI config space has the necessary stuff set */
1275 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1276 
1277 	/* Save off the information about this board */
1278 	adapter->hw.vendor_id = pci_get_vendor(dev);
1279 	adapter->hw.device_id = pci_get_device(dev);
1280 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1281 	adapter->hw.subsystem_vendor_id =
1282 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1283 	adapter->hw.subsystem_device_id =
1284 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1285 
1286 	/* Do Shared Code Init and Setup */
1287 	if (igc_set_mac_type(&adapter->hw)) {
1288 		device_printf(dev, "Setup init failure\n");
1289 		return;
1290 	}
1291 }
1292 
1293 static int
1294 igc_allocate_pci_resources(if_ctx_t ctx)
1295 {
1296 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1297 	device_t dev = iflib_get_dev(ctx);
1298 	int rid;
1299 
1300 	rid = PCIR_BAR(0);
1301 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1302 	    &rid, RF_ACTIVE);
1303 	if (adapter->memory == NULL) {
1304 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1305 		return (ENXIO);
1306 	}
1307 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1308 	adapter->osdep.mem_bus_space_handle =
1309 	    rman_get_bushandle(adapter->memory);
1310 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1311 
1312 	adapter->hw.back = &adapter->osdep;
1313 
1314 	return (0);
1315 }
1316 
1317 /*********************************************************************
1318  *
1319  *  Set up the MSI-X Interrupt handlers
1320  *
1321  **********************************************************************/
1322 static int
1323 igc_if_msix_intr_assign(if_ctx_t ctx, int msix)
1324 {
1325 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1326 	struct igc_rx_queue *rx_que = adapter->rx_queues;
1327 	struct igc_tx_queue *tx_que = adapter->tx_queues;
1328 	int error, rid, i, vector = 0, rx_vectors;
1329 	char buf[16];
1330 
1331 	/* First set up ring resources */
1332 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1333 		rid = vector + 1;
1334 		snprintf(buf, sizeof(buf), "rxq%d", i);
1335 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, igc_msix_que, rx_que, rx_que->me, buf);
1336 		if (error) {
1337 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1338 			adapter->rx_num_queues = i + 1;
1339 			goto fail;
1340 		}
1341 
1342 		rx_que->msix =  vector;
1343 
1344 		/*
1345 		 * Set the bit to enable interrupt
1346 		 * in IGC_IMS -- bits 20 and 21
1347 		 * are for RX0 and RX1, note this has
1348 		 * NOTHING to do with the MSI-X vector
1349 		 */
1350 		rx_que->eims = 1 << vector;
1351 	}
1352 	rx_vectors = vector;
1353 
1354 	vector = 0;
1355 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
1356 		snprintf(buf, sizeof(buf), "txq%d", i);
1357 		tx_que = &adapter->tx_queues[i];
1358 		iflib_softirq_alloc_generic(ctx,
1359 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
1360 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
1361 
1362 		tx_que->msix = (vector % adapter->rx_num_queues);
1363 
1364 		/*
1365 		 * Set the bit to enable interrupt
1366 		 * in IGC_IMS -- bits 22 and 23
1367 		 * are for TX0 and TX1, note this has
1368 		 * NOTHING to do with the MSI-X vector
1369 		 */
1370 		tx_que->eims = 1 << i;
1371 	}
1372 
1373 	/* Link interrupt */
1374 	rid = rx_vectors + 1;
1375 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, igc_msix_link, adapter, 0, "aq");
1376 
1377 	if (error) {
1378 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
1379 		goto fail;
1380 	}
1381 	adapter->linkvec = rx_vectors;
1382 	return (0);
1383 fail:
1384 	iflib_irq_free(ctx, &adapter->irq);
1385 	rx_que = adapter->rx_queues;
1386 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
1387 		iflib_irq_free(ctx, &rx_que->que_irq);
1388 	return (error);
1389 }
1390 
1391 static void
1392 igc_configure_queues(struct igc_adapter *adapter)
1393 {
1394 	struct igc_hw *hw = &adapter->hw;
1395 	struct igc_rx_queue *rx_que;
1396 	struct igc_tx_queue *tx_que;
1397 	u32 ivar = 0, newitr = 0;
1398 
1399 	/* First turn on RSS capability */
1400 	IGC_WRITE_REG(hw, IGC_GPIE,
1401 	    IGC_GPIE_MSIX_MODE | IGC_GPIE_EIAME | IGC_GPIE_PBA |
1402 	    IGC_GPIE_NSICR);
1403 
1404 	/* Turn on MSI-X */
1405 	/* RX entries */
1406 	for (int i = 0; i < adapter->rx_num_queues; i++) {
1407 		u32 index = i >> 1;
1408 		ivar = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, index);
1409 		rx_que = &adapter->rx_queues[i];
1410 		if (i & 1) {
1411 			ivar &= 0xFF00FFFF;
1412 			ivar |= (rx_que->msix | IGC_IVAR_VALID) << 16;
1413 		} else {
1414 			ivar &= 0xFFFFFF00;
1415 			ivar |= rx_que->msix | IGC_IVAR_VALID;
1416 		}
1417 		IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, index, ivar);
1418 	}
1419 	/* TX entries */
1420 	for (int i = 0; i < adapter->tx_num_queues; i++) {
1421 		u32 index = i >> 1;
1422 		ivar = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, index);
1423 		tx_que = &adapter->tx_queues[i];
1424 		if (i & 1) {
1425 			ivar &= 0x00FFFFFF;
1426 			ivar |= (tx_que->msix | IGC_IVAR_VALID) << 24;
1427 		} else {
1428 			ivar &= 0xFFFF00FF;
1429 			ivar |= (tx_que->msix | IGC_IVAR_VALID) << 8;
1430 		}
1431 		IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, index, ivar);
1432 		adapter->que_mask |= tx_que->eims;
1433 	}
1434 
1435 	/* And for the link interrupt */
1436 	ivar = (adapter->linkvec | IGC_IVAR_VALID) << 8;
1437 	adapter->link_mask = 1 << adapter->linkvec;
1438 	IGC_WRITE_REG(hw, IGC_IVAR_MISC, ivar);
1439 
1440 	/* Set the starting interrupt rate */
1441 	if (igc_max_interrupt_rate > 0)
1442 		newitr = (4000000 / igc_max_interrupt_rate) & 0x7FFC;
1443 
1444 	newitr |= IGC_EITR_CNT_IGNR;
1445 
1446 	for (int i = 0; i < adapter->rx_num_queues; i++) {
1447 		rx_que = &adapter->rx_queues[i];
1448 		IGC_WRITE_REG(hw, IGC_EITR(rx_que->msix), newitr);
1449 	}
1450 
1451 	return;
1452 }
1453 
1454 static void
1455 igc_free_pci_resources(if_ctx_t ctx)
1456 {
1457 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1458 	struct igc_rx_queue *que = adapter->rx_queues;
1459 	device_t dev = iflib_get_dev(ctx);
1460 
1461 	/* Release all MSI-X queue resources */
1462 	if (adapter->intr_type == IFLIB_INTR_MSIX)
1463 		iflib_irq_free(ctx, &adapter->irq);
1464 
1465 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
1466 		iflib_irq_free(ctx, &que->que_irq);
1467 	}
1468 
1469 	if (adapter->memory != NULL) {
1470 		bus_release_resource(dev, SYS_RES_MEMORY,
1471 		    rman_get_rid(adapter->memory), adapter->memory);
1472 		adapter->memory = NULL;
1473 	}
1474 
1475 	if (adapter->flash != NULL) {
1476 		bus_release_resource(dev, SYS_RES_MEMORY,
1477 		    rman_get_rid(adapter->flash), adapter->flash);
1478 		adapter->flash = NULL;
1479 	}
1480 
1481 	if (adapter->ioport != NULL) {
1482 		bus_release_resource(dev, SYS_RES_IOPORT,
1483 		    rman_get_rid(adapter->ioport), adapter->ioport);
1484 		adapter->ioport = NULL;
1485 	}
1486 }
1487 
1488 /* Set up MSI or MSI-X */
1489 static int
1490 igc_setup_msix(if_ctx_t ctx)
1491 {
1492 	return (0);
1493 }
1494 
1495 /*********************************************************************
1496  *
1497  *  Initialize the DMA Coalescing feature
1498  *
1499  **********************************************************************/
1500 static void
1501 igc_init_dmac(struct igc_adapter *adapter, u32 pba)
1502 {
1503 	device_t	dev = adapter->dev;
1504 	struct igc_hw *hw = &adapter->hw;
1505 	u32 		dmac, reg = ~IGC_DMACR_DMAC_EN;
1506 	u16		hwm;
1507 	u16		max_frame_size;
1508 	int		status;
1509 
1510 	max_frame_size = adapter->shared->isc_max_frame_size;
1511 
1512 	if (adapter->dmac == 0) { /* Disabling it */
1513 		IGC_WRITE_REG(hw, IGC_DMACR, reg);
1514 		return;
1515 	} else
1516 		device_printf(dev, "DMA Coalescing enabled\n");
1517 
1518 	/* Set starting threshold */
1519 	IGC_WRITE_REG(hw, IGC_DMCTXTH, 0);
1520 
1521 	hwm = 64 * pba - max_frame_size / 16;
1522 	if (hwm < 64 * (pba - 6))
1523 		hwm = 64 * (pba - 6);
1524 	reg = IGC_READ_REG(hw, IGC_FCRTC);
1525 	reg &= ~IGC_FCRTC_RTH_COAL_MASK;
1526 	reg |= ((hwm << IGC_FCRTC_RTH_COAL_SHIFT)
1527 		& IGC_FCRTC_RTH_COAL_MASK);
1528 	IGC_WRITE_REG(hw, IGC_FCRTC, reg);
1529 
1530 	dmac = pba - max_frame_size / 512;
1531 	if (dmac < pba - 10)
1532 		dmac = pba - 10;
1533 	reg = IGC_READ_REG(hw, IGC_DMACR);
1534 	reg &= ~IGC_DMACR_DMACTHR_MASK;
1535 	reg |= ((dmac << IGC_DMACR_DMACTHR_SHIFT)
1536 		& IGC_DMACR_DMACTHR_MASK);
1537 
1538 	/* transition to L0x or L1 if available..*/
1539 	reg |= (IGC_DMACR_DMAC_EN | IGC_DMACR_DMAC_LX_MASK);
1540 
1541 	/* Check if status is 2.5Gb backplane connection
1542 	 * before configuration of watchdog timer, which is
1543 	 * in msec values in 12.8usec intervals
1544 	 * watchdog timer= msec values in 32usec intervals
1545 	 * for non 2.5Gb connection
1546 	 */
1547 	status = IGC_READ_REG(hw, IGC_STATUS);
1548 	if ((status & IGC_STATUS_2P5_SKU) &&
1549 	    (!(status & IGC_STATUS_2P5_SKU_OVER)))
1550 		reg |= ((adapter->dmac * 5) >> 6);
1551 	else
1552 		reg |= (adapter->dmac >> 5);
1553 
1554 	IGC_WRITE_REG(hw, IGC_DMACR, reg);
1555 
1556 	IGC_WRITE_REG(hw, IGC_DMCRTRH, 0);
1557 
1558 	/* Set the interval before transition */
1559 	reg = IGC_READ_REG(hw, IGC_DMCTLX);
1560 	reg |= IGC_DMCTLX_DCFLUSH_DIS;
1561 
1562 	/*
1563 	** in 2.5Gb connection, TTLX unit is 0.4 usec
1564 	** which is 0x4*2 = 0xA. But delay is still 4 usec
1565 	*/
1566 	status = IGC_READ_REG(hw, IGC_STATUS);
1567 	if ((status & IGC_STATUS_2P5_SKU) &&
1568 	    (!(status & IGC_STATUS_2P5_SKU_OVER)))
1569 		reg |= 0xA;
1570 	else
1571 		reg |= 0x4;
1572 
1573 	IGC_WRITE_REG(hw, IGC_DMCTLX, reg);
1574 
1575 	/* free space in tx packet buffer to wake from DMA coal */
1576 	IGC_WRITE_REG(hw, IGC_DMCTXTH, (IGC_TXPBSIZE -
1577 	    (2 * max_frame_size)) >> 6);
1578 
1579 	/* make low power state decision controlled by DMA coal */
1580 	reg = IGC_READ_REG(hw, IGC_PCIEMISC);
1581 	reg &= ~IGC_PCIEMISC_LX_DECISION;
1582 	IGC_WRITE_REG(hw, IGC_PCIEMISC, reg);
1583 }
1584 
1585 /*********************************************************************
1586  *
1587  *  Initialize the hardware to a configuration as specified by the
1588  *  adapter structure.
1589  *
1590  **********************************************************************/
1591 static void
1592 igc_reset(if_ctx_t ctx)
1593 {
1594 	device_t dev = iflib_get_dev(ctx);
1595 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1596 	struct igc_hw *hw = &adapter->hw;
1597 	u32 rx_buffer_size;
1598 	u32 pba;
1599 
1600 	INIT_DEBUGOUT("igc_reset: begin");
1601 	/* Let the firmware know the OS is in control */
1602 	igc_get_hw_control(adapter);
1603 
1604 	/*
1605 	 * Packet Buffer Allocation (PBA)
1606 	 * Writing PBA sets the receive portion of the buffer
1607 	 * the remainder is used for the transmit buffer.
1608 	 */
1609 	pba = IGC_PBA_34K;
1610 
1611 	INIT_DEBUGOUT1("igc_reset: pba=%dK",pba);
1612 
1613 	/*
1614 	 * These parameters control the automatic generation (Tx) and
1615 	 * response (Rx) to Ethernet PAUSE frames.
1616 	 * - High water mark should allow for at least two frames to be
1617 	 *   received after sending an XOFF.
1618 	 * - Low water mark works best when it is very near the high water mark.
1619 	 *   This allows the receiver to restart by sending XON when it has
1620 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1621 	 *   restart after one full frame is pulled from the buffer. There
1622 	 *   could be several smaller frames in the buffer and if so they will
1623 	 *   not trigger the XON until their total number reduces the buffer
1624 	 *   by 1500.
1625 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1626 	 */
1627 	rx_buffer_size = (pba & 0xffff) << 10;
1628 	hw->fc.high_water = rx_buffer_size -
1629 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
1630 	/* 16-byte granularity */
1631 	hw->fc.low_water = hw->fc.high_water - 16;
1632 
1633 	if (adapter->fc) /* locally set flow control value? */
1634 		hw->fc.requested_mode = adapter->fc;
1635 	else
1636 		hw->fc.requested_mode = igc_fc_full;
1637 
1638 	hw->fc.pause_time = IGC_FC_PAUSE_TIME;
1639 
1640 	hw->fc.send_xon = true;
1641 
1642 	/* Issue a global reset */
1643 	igc_reset_hw(hw);
1644 	IGC_WRITE_REG(hw, IGC_WUC, 0);
1645 
1646 	/* and a re-init */
1647 	if (igc_init_hw(hw) < 0) {
1648 		device_printf(dev, "Hardware Initialization Failed\n");
1649 		return;
1650 	}
1651 
1652 	/* Setup DMA Coalescing */
1653 	igc_init_dmac(adapter, pba);
1654 
1655 	IGC_WRITE_REG(hw, IGC_VET, ETHERTYPE_VLAN);
1656 	igc_get_phy_info(hw);
1657 	igc_check_for_link(hw);
1658 }
1659 
1660 /*
1661  * Initialise the RSS mapping for NICs that support multiple transmit/
1662  * receive rings.
1663  */
1664 
1665 #define RSSKEYLEN 10
1666 static void
1667 igc_initialize_rss_mapping(struct igc_adapter *adapter)
1668 {
1669 	struct igc_hw *hw = &adapter->hw;
1670 	int i;
1671 	int queue_id;
1672 	u32 reta;
1673 	u32 rss_key[RSSKEYLEN], mrqc, shift = 0;
1674 
1675 	/*
1676 	 * The redirection table controls which destination
1677 	 * queue each bucket redirects traffic to.
1678 	 * Each DWORD represents four queues, with the LSB
1679 	 * being the first queue in the DWORD.
1680 	 *
1681 	 * This just allocates buckets to queues using round-robin
1682 	 * allocation.
1683 	 *
1684 	 * NOTE: It Just Happens to line up with the default
1685 	 * RSS allocation method.
1686 	 */
1687 
1688 	/* Warning FM follows */
1689 	reta = 0;
1690 	for (i = 0; i < 128; i++) {
1691 #ifdef RSS
1692 		queue_id = rss_get_indirection_to_bucket(i);
1693 		/*
1694 		 * If we have more queues than buckets, we'll
1695 		 * end up mapping buckets to a subset of the
1696 		 * queues.
1697 		 *
1698 		 * If we have more buckets than queues, we'll
1699 		 * end up instead assigning multiple buckets
1700 		 * to queues.
1701 		 *
1702 		 * Both are suboptimal, but we need to handle
1703 		 * the case so we don't go out of bounds
1704 		 * indexing arrays and such.
1705 		 */
1706 		queue_id = queue_id % adapter->rx_num_queues;
1707 #else
1708 		queue_id = (i % adapter->rx_num_queues);
1709 #endif
1710 		/* Adjust if required */
1711 		queue_id = queue_id << shift;
1712 
1713 		/*
1714 		 * The low 8 bits are for hash value (n+0);
1715 		 * The next 8 bits are for hash value (n+1), etc.
1716 		 */
1717 		reta = reta >> 8;
1718 		reta = reta | ( ((uint32_t) queue_id) << 24);
1719 		if ((i & 3) == 3) {
1720 			IGC_WRITE_REG(hw, IGC_RETA(i >> 2), reta);
1721 			reta = 0;
1722 		}
1723 	}
1724 
1725 	/* Now fill in hash table */
1726 
1727 	/*
1728 	 * MRQC: Multiple Receive Queues Command
1729 	 * Set queuing to RSS control, number depends on the device.
1730 	 */
1731 	mrqc = IGC_MRQC_ENABLE_RSS_4Q;
1732 
1733 #ifdef RSS
1734 	/* XXX ew typecasting */
1735 	rss_getkey((uint8_t *) &rss_key);
1736 #else
1737 	arc4rand(&rss_key, sizeof(rss_key), 0);
1738 #endif
1739 	for (i = 0; i < RSSKEYLEN; i++)
1740 		IGC_WRITE_REG_ARRAY(hw, IGC_RSSRK(0), i, rss_key[i]);
1741 
1742 	/*
1743 	 * Configure the RSS fields to hash upon.
1744 	 */
1745 	mrqc |= (IGC_MRQC_RSS_FIELD_IPV4 |
1746 	    IGC_MRQC_RSS_FIELD_IPV4_TCP);
1747 	mrqc |= (IGC_MRQC_RSS_FIELD_IPV6 |
1748 	    IGC_MRQC_RSS_FIELD_IPV6_TCP);
1749 	mrqc |=( IGC_MRQC_RSS_FIELD_IPV4_UDP |
1750 	    IGC_MRQC_RSS_FIELD_IPV6_UDP);
1751 	mrqc |=( IGC_MRQC_RSS_FIELD_IPV6_UDP_EX |
1752 	    IGC_MRQC_RSS_FIELD_IPV6_TCP_EX);
1753 
1754 	IGC_WRITE_REG(hw, IGC_MRQC, mrqc);
1755 }
1756 
1757 /*********************************************************************
1758  *
1759  *  Setup networking device structure and register interface media.
1760  *
1761  **********************************************************************/
1762 static int
1763 igc_setup_interface(if_ctx_t ctx)
1764 {
1765 	if_t ifp = iflib_get_ifp(ctx);
1766 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1767 	if_softc_ctx_t scctx = adapter->shared;
1768 
1769 	INIT_DEBUGOUT("igc_setup_interface: begin");
1770 
1771 	/* Single Queue */
1772 	if (adapter->tx_num_queues == 1) {
1773 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
1774 		if_setsendqready(ifp);
1775 	}
1776 
1777 	/*
1778 	 * Specify the media types supported by this adapter and register
1779 	 * callbacks to update media and link information
1780 	 */
1781 	ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
1782 	ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
1783 	ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
1784 	ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
1785 	ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
1786 	ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
1787 	ifmedia_add(adapter->media, IFM_ETHER | IFM_2500_T, 0, NULL);
1788 
1789 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
1790 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
1791 	return (0);
1792 }
1793 
1794 static int
1795 igc_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
1796 {
1797 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1798 	if_softc_ctx_t scctx = adapter->shared;
1799 	int error = IGC_SUCCESS;
1800 	struct igc_tx_queue *que;
1801 	int i, j;
1802 
1803 	MPASS(adapter->tx_num_queues > 0);
1804 	MPASS(adapter->tx_num_queues == ntxqsets);
1805 
1806 	/* First allocate the top level queue structs */
1807 	if (!(adapter->tx_queues =
1808 	    (struct igc_tx_queue *) malloc(sizeof(struct igc_tx_queue) *
1809 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
1810 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
1811 		return(ENOMEM);
1812 	}
1813 
1814 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
1815 		/* Set up some basics */
1816 
1817 		struct tx_ring *txr = &que->txr;
1818 		txr->adapter = que->adapter = adapter;
1819 		que->me = txr->me =  i;
1820 
1821 		/* Allocate report status array */
1822 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
1823 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
1824 			error = ENOMEM;
1825 			goto fail;
1826 		}
1827 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
1828 			txr->tx_rsq[j] = QIDX_INVALID;
1829 		/* get the virtual and physical address of the hardware queues */
1830 		txr->tx_base = (struct igc_tx_desc *)vaddrs[i*ntxqs];
1831 		txr->tx_paddr = paddrs[i*ntxqs];
1832 	}
1833 
1834 	if (bootverbose)
1835 		device_printf(iflib_get_dev(ctx),
1836 		    "allocated for %d tx_queues\n", adapter->tx_num_queues);
1837 	return (0);
1838 fail:
1839 	igc_if_queues_free(ctx);
1840 	return (error);
1841 }
1842 
1843 static int
1844 igc_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
1845 {
1846 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1847 	int error = IGC_SUCCESS;
1848 	struct igc_rx_queue *que;
1849 	int i;
1850 
1851 	MPASS(adapter->rx_num_queues > 0);
1852 	MPASS(adapter->rx_num_queues == nrxqsets);
1853 
1854 	/* First allocate the top level queue structs */
1855 	if (!(adapter->rx_queues =
1856 	    (struct igc_rx_queue *) malloc(sizeof(struct igc_rx_queue) *
1857 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
1858 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
1859 		error = ENOMEM;
1860 		goto fail;
1861 	}
1862 
1863 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
1864 		/* Set up some basics */
1865 		struct rx_ring *rxr = &que->rxr;
1866 		rxr->adapter = que->adapter = adapter;
1867 		rxr->que = que;
1868 		que->me = rxr->me =  i;
1869 
1870 		/* get the virtual and physical address of the hardware queues */
1871 		rxr->rx_base = (union igc_rx_desc_extended *)vaddrs[i*nrxqs];
1872 		rxr->rx_paddr = paddrs[i*nrxqs];
1873 	}
1874 
1875 	if (bootverbose)
1876 		device_printf(iflib_get_dev(ctx),
1877 		    "allocated for %d rx_queues\n", adapter->rx_num_queues);
1878 
1879 	return (0);
1880 fail:
1881 	igc_if_queues_free(ctx);
1882 	return (error);
1883 }
1884 
1885 static void
1886 igc_if_queues_free(if_ctx_t ctx)
1887 {
1888 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1889 	struct igc_tx_queue *tx_que = adapter->tx_queues;
1890 	struct igc_rx_queue *rx_que = adapter->rx_queues;
1891 
1892 	if (tx_que != NULL) {
1893 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
1894 			struct tx_ring *txr = &tx_que->txr;
1895 			if (txr->tx_rsq == NULL)
1896 				break;
1897 
1898 			free(txr->tx_rsq, M_DEVBUF);
1899 			txr->tx_rsq = NULL;
1900 		}
1901 		free(adapter->tx_queues, M_DEVBUF);
1902 		adapter->tx_queues = NULL;
1903 	}
1904 
1905 	if (rx_que != NULL) {
1906 		free(adapter->rx_queues, M_DEVBUF);
1907 		adapter->rx_queues = NULL;
1908 	}
1909 
1910 	igc_release_hw_control(adapter);
1911 
1912 	if (adapter->mta != NULL) {
1913 		free(adapter->mta, M_DEVBUF);
1914 	}
1915 }
1916 
1917 /*********************************************************************
1918  *
1919  *  Enable transmit unit.
1920  *
1921  **********************************************************************/
1922 static void
1923 igc_initialize_transmit_unit(if_ctx_t ctx)
1924 {
1925 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1926 	if_softc_ctx_t scctx = adapter->shared;
1927 	struct igc_tx_queue *que;
1928 	struct tx_ring	*txr;
1929 	struct igc_hw	*hw = &adapter->hw;
1930 	u32 tctl, txdctl = 0;
1931 
1932 	INIT_DEBUGOUT("igc_initialize_transmit_unit: begin");
1933 
1934 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
1935 		u64 bus_addr;
1936 		caddr_t offp, endp;
1937 
1938 		que = &adapter->tx_queues[i];
1939 		txr = &que->txr;
1940 		bus_addr = txr->tx_paddr;
1941 
1942 		/* Clear checksum offload context. */
1943 		offp = (caddr_t)&txr->csum_flags;
1944 		endp = (caddr_t)(txr + 1);
1945 		bzero(offp, endp - offp);
1946 
1947 		/* Base and Len of TX Ring */
1948 		IGC_WRITE_REG(hw, IGC_TDLEN(i),
1949 		    scctx->isc_ntxd[0] * sizeof(struct igc_tx_desc));
1950 		IGC_WRITE_REG(hw, IGC_TDBAH(i),
1951 		    (u32)(bus_addr >> 32));
1952 		IGC_WRITE_REG(hw, IGC_TDBAL(i),
1953 		    (u32)bus_addr);
1954 		/* Init the HEAD/TAIL indices */
1955 		IGC_WRITE_REG(hw, IGC_TDT(i), 0);
1956 		IGC_WRITE_REG(hw, IGC_TDH(i), 0);
1957 
1958 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
1959 		    IGC_READ_REG(&adapter->hw, IGC_TDBAL(i)),
1960 		    IGC_READ_REG(&adapter->hw, IGC_TDLEN(i)));
1961 
1962 		txdctl = 0; /* clear txdctl */
1963 		txdctl |= 0x1f; /* PTHRESH */
1964 		txdctl |= 1 << 8; /* HTHRESH */
1965 		txdctl |= 1 << 16;/* WTHRESH */
1966 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
1967 		txdctl |= IGC_TXDCTL_GRAN;
1968 		txdctl |= 1 << 25; /* LWTHRESH */
1969 
1970 		IGC_WRITE_REG(hw, IGC_TXDCTL(i), txdctl);
1971 	}
1972 
1973 	/* Program the Transmit Control Register */
1974 	tctl = IGC_READ_REG(&adapter->hw, IGC_TCTL);
1975 	tctl &= ~IGC_TCTL_CT;
1976 	tctl |= (IGC_TCTL_PSP | IGC_TCTL_RTLC | IGC_TCTL_EN |
1977 		   (IGC_COLLISION_THRESHOLD << IGC_CT_SHIFT));
1978 
1979 	/* This write will effectively turn on the transmit unit. */
1980 	IGC_WRITE_REG(&adapter->hw, IGC_TCTL, tctl);
1981 }
1982 
1983 /*********************************************************************
1984  *
1985  *  Enable receive unit.
1986  *
1987  **********************************************************************/
1988 
1989 static void
1990 igc_initialize_receive_unit(if_ctx_t ctx)
1991 {
1992 	struct igc_adapter *adapter = iflib_get_softc(ctx);
1993 	if_softc_ctx_t scctx = adapter->shared;
1994 	if_t ifp = iflib_get_ifp(ctx);
1995 	struct igc_hw	*hw = &adapter->hw;
1996 	struct igc_rx_queue *que;
1997 	int i;
1998 	u32 psize, rctl, rxcsum, srrctl = 0;
1999 
2000 	INIT_DEBUGOUT("igc_initialize_receive_units: begin");
2001 
2002 	/*
2003 	 * Make sure receives are disabled while setting
2004 	 * up the descriptor ring
2005 	 */
2006 	rctl = IGC_READ_REG(hw, IGC_RCTL);
2007 	IGC_WRITE_REG(hw, IGC_RCTL, rctl & ~IGC_RCTL_EN);
2008 
2009 	/* Setup the Receive Control Register */
2010 	rctl &= ~(3 << IGC_RCTL_MO_SHIFT);
2011 	rctl |= IGC_RCTL_EN | IGC_RCTL_BAM |
2012 	    IGC_RCTL_LBM_NO | IGC_RCTL_RDMTS_HALF |
2013 	    (hw->mac.mc_filter_type << IGC_RCTL_MO_SHIFT);
2014 
2015 	/* Do not store bad packets */
2016 	rctl &= ~IGC_RCTL_SBP;
2017 
2018 	/* Enable Long Packet receive */
2019 	if (if_getmtu(ifp) > ETHERMTU)
2020 		rctl |= IGC_RCTL_LPE;
2021 	else
2022 		rctl &= ~IGC_RCTL_LPE;
2023 
2024 	/* Strip the CRC */
2025 	if (!igc_disable_crc_stripping)
2026 		rctl |= IGC_RCTL_SECRC;
2027 
2028 	/*
2029 	 * Set the interrupt throttling rate. Value is calculated
2030 	 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
2031 	 */
2032 	IGC_WRITE_REG(hw, IGC_ITR, DEFAULT_ITR);
2033 
2034 	rxcsum = IGC_READ_REG(hw, IGC_RXCSUM);
2035 	if (if_getcapenable(ifp) & IFCAP_RXCSUM) {
2036 		rxcsum |= IGC_RXCSUM_CRCOFL;
2037 		if (adapter->tx_num_queues > 1)
2038 			rxcsum |= IGC_RXCSUM_PCSD;
2039 		else
2040 			rxcsum |= IGC_RXCSUM_IPPCSE;
2041 	} else {
2042 		if (adapter->tx_num_queues > 1)
2043 			rxcsum |= IGC_RXCSUM_PCSD;
2044 		else
2045 			rxcsum &= ~IGC_RXCSUM_TUOFL;
2046 	}
2047 	IGC_WRITE_REG(hw, IGC_RXCSUM, rxcsum);
2048 
2049 	if (adapter->rx_num_queues > 1)
2050 		igc_initialize_rss_mapping(adapter);
2051 
2052 	if (if_getmtu(ifp) > ETHERMTU) {
2053 		/* Set maximum packet len */
2054 		if (adapter->rx_mbuf_sz <= 4096) {
2055 			srrctl |= 4096 >> IGC_SRRCTL_BSIZEPKT_SHIFT;
2056 			rctl |= IGC_RCTL_SZ_4096 | IGC_RCTL_BSEX;
2057 		} else if (adapter->rx_mbuf_sz > 4096) {
2058 			srrctl |= 8192 >> IGC_SRRCTL_BSIZEPKT_SHIFT;
2059 			rctl |= IGC_RCTL_SZ_8192 | IGC_RCTL_BSEX;
2060 		}
2061 		psize = scctx->isc_max_frame_size;
2062 		/* are we on a vlan? */
2063 		if (if_vlantrunkinuse(ifp))
2064 			psize += VLAN_TAG_SIZE;
2065 		IGC_WRITE_REG(&adapter->hw, IGC_RLPML, psize);
2066 	} else {
2067 		srrctl |= 2048 >> IGC_SRRCTL_BSIZEPKT_SHIFT;
2068 		rctl |= IGC_RCTL_SZ_2048;
2069 	}
2070 
2071 	/*
2072 	 * If TX flow control is disabled and there's >1 queue defined,
2073 	 * enable DROP.
2074 	 *
2075 	 * This drops frames rather than hanging the RX MAC for all queues.
2076 	 */
2077 	if ((adapter->rx_num_queues > 1) &&
2078 	    (adapter->fc == igc_fc_none ||
2079 	     adapter->fc == igc_fc_rx_pause)) {
2080 		srrctl |= IGC_SRRCTL_DROP_EN;
2081 	}
2082 
2083 	/* Setup the Base and Length of the Rx Descriptor Rings */
2084 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
2085 		struct rx_ring *rxr = &que->rxr;
2086 		u64 bus_addr = rxr->rx_paddr;
2087 		u32 rxdctl;
2088 
2089 #ifdef notyet
2090 		/* Configure for header split? -- ignore for now */
2091 		rxr->hdr_split = igc_header_split;
2092 #else
2093 		srrctl |= IGC_SRRCTL_DESCTYPE_ADV_ONEBUF;
2094 #endif
2095 
2096 		IGC_WRITE_REG(hw, IGC_RDLEN(i),
2097 			      scctx->isc_nrxd[0] * sizeof(struct igc_rx_desc));
2098 		IGC_WRITE_REG(hw, IGC_RDBAH(i),
2099 			      (uint32_t)(bus_addr >> 32));
2100 		IGC_WRITE_REG(hw, IGC_RDBAL(i),
2101 			      (uint32_t)bus_addr);
2102 		IGC_WRITE_REG(hw, IGC_SRRCTL(i), srrctl);
2103 		/* Setup the Head and Tail Descriptor Pointers */
2104 		IGC_WRITE_REG(hw, IGC_RDH(i), 0);
2105 		IGC_WRITE_REG(hw, IGC_RDT(i), 0);
2106 		/* Enable this Queue */
2107 		rxdctl = IGC_READ_REG(hw, IGC_RXDCTL(i));
2108 		rxdctl |= IGC_RXDCTL_QUEUE_ENABLE;
2109 		rxdctl &= 0xFFF00000;
2110 		rxdctl |= IGC_RX_PTHRESH;
2111 		rxdctl |= IGC_RX_HTHRESH << 8;
2112 		rxdctl |= IGC_RX_WTHRESH << 16;
2113 		IGC_WRITE_REG(hw, IGC_RXDCTL(i), rxdctl);
2114 	}
2115 
2116 	/* Make sure VLAN Filters are off */
2117 	rctl &= ~IGC_RCTL_VFE;
2118 
2119 	/* Write out the settings */
2120 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
2121 
2122 	return;
2123 }
2124 
2125 static void
2126 igc_setup_vlan_hw_support(if_ctx_t ctx)
2127 {
2128 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2129 	struct igc_hw *hw = &adapter->hw;
2130 	struct ifnet *ifp = iflib_get_ifp(ctx);
2131 	u32 reg;
2132 
2133 	/* igc hardware doesn't seem to implement VFTA for HWFILTER */
2134 
2135 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING &&
2136 	    !igc_disable_crc_stripping) {
2137 		reg = IGC_READ_REG(hw, IGC_CTRL);
2138 		reg |= IGC_CTRL_VME;
2139 		IGC_WRITE_REG(hw, IGC_CTRL, reg);
2140 	} else {
2141 		reg = IGC_READ_REG(hw, IGC_CTRL);
2142 		reg &= ~IGC_CTRL_VME;
2143 		IGC_WRITE_REG(hw, IGC_CTRL, reg);
2144 	}
2145 }
2146 
2147 static void
2148 igc_if_intr_enable(if_ctx_t ctx)
2149 {
2150 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2151 	struct igc_hw *hw = &adapter->hw;
2152 	u32 mask;
2153 
2154 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
2155 		mask = (adapter->que_mask | adapter->link_mask);
2156 		IGC_WRITE_REG(hw, IGC_EIAC, mask);
2157 		IGC_WRITE_REG(hw, IGC_EIAM, mask);
2158 		IGC_WRITE_REG(hw, IGC_EIMS, mask);
2159 		IGC_WRITE_REG(hw, IGC_IMS, IGC_IMS_LSC);
2160 	} else
2161 		IGC_WRITE_REG(hw, IGC_IMS, IMS_ENABLE_MASK);
2162 	IGC_WRITE_FLUSH(hw);
2163 }
2164 
2165 static void
2166 igc_if_intr_disable(if_ctx_t ctx)
2167 {
2168 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2169 	struct igc_hw *hw = &adapter->hw;
2170 
2171 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
2172 		IGC_WRITE_REG(hw, IGC_EIMC, 0xffffffff);
2173 		IGC_WRITE_REG(hw, IGC_EIAC, 0);
2174 	}
2175 	IGC_WRITE_REG(hw, IGC_IMC, 0xffffffff);
2176 	IGC_WRITE_FLUSH(hw);
2177 }
2178 
2179 /*
2180  * igc_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
2181  * For ASF and Pass Through versions of f/w this means
2182  * that the driver is loaded. For AMT version type f/w
2183  * this means that the network i/f is open.
2184  */
2185 static void
2186 igc_get_hw_control(struct igc_adapter *adapter)
2187 {
2188 	u32 ctrl_ext;
2189 
2190 	if (adapter->vf_ifp)
2191 		return;
2192 
2193 	ctrl_ext = IGC_READ_REG(&adapter->hw, IGC_CTRL_EXT);
2194 	IGC_WRITE_REG(&adapter->hw, IGC_CTRL_EXT,
2195 	    ctrl_ext | IGC_CTRL_EXT_DRV_LOAD);
2196 }
2197 
2198 /*
2199  * igc_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
2200  * For ASF and Pass Through versions of f/w this means that
2201  * the driver is no longer loaded. For AMT versions of the
2202  * f/w this means that the network i/f is closed.
2203  */
2204 static void
2205 igc_release_hw_control(struct igc_adapter *adapter)
2206 {
2207 	u32 ctrl_ext;
2208 
2209 	ctrl_ext = IGC_READ_REG(&adapter->hw, IGC_CTRL_EXT);
2210 	IGC_WRITE_REG(&adapter->hw, IGC_CTRL_EXT,
2211 	    ctrl_ext & ~IGC_CTRL_EXT_DRV_LOAD);
2212 	return;
2213 }
2214 
2215 static int
2216 igc_is_valid_ether_addr(u8 *addr)
2217 {
2218 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
2219 
2220 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
2221 		return (false);
2222 	}
2223 
2224 	return (true);
2225 }
2226 
2227 /*
2228 ** Parse the interface capabilities with regard
2229 ** to both system management and wake-on-lan for
2230 ** later use.
2231 */
2232 static void
2233 igc_get_wakeup(if_ctx_t ctx)
2234 {
2235 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2236 	u16 eeprom_data = 0, apme_mask;
2237 
2238 	apme_mask = IGC_WUC_APME;
2239 	eeprom_data = IGC_READ_REG(&adapter->hw, IGC_WUC);
2240 
2241 	if (eeprom_data & apme_mask)
2242 		adapter->wol = IGC_WUFC_LNKC;
2243 }
2244 
2245 
2246 /*
2247  * Enable PCI Wake On Lan capability
2248  */
2249 static void
2250 igc_enable_wakeup(if_ctx_t ctx)
2251 {
2252 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2253 	device_t dev = iflib_get_dev(ctx);
2254 	if_t ifp = iflib_get_ifp(ctx);
2255 	int error = 0;
2256 	u32 pmc, ctrl, rctl;
2257 	u16 status;
2258 
2259 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
2260 		return;
2261 
2262 	/*
2263 	 * Determine type of Wakeup: note that wol
2264 	 * is set with all bits on by default.
2265 	 */
2266 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
2267 		adapter->wol &= ~IGC_WUFC_MAG;
2268 
2269 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
2270 		adapter->wol &= ~IGC_WUFC_EX;
2271 
2272 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
2273 		adapter->wol &= ~IGC_WUFC_MC;
2274 	else {
2275 		rctl = IGC_READ_REG(&adapter->hw, IGC_RCTL);
2276 		rctl |= IGC_RCTL_MPE;
2277 		IGC_WRITE_REG(&adapter->hw, IGC_RCTL, rctl);
2278 	}
2279 
2280 	if (!(adapter->wol & (IGC_WUFC_EX | IGC_WUFC_MAG | IGC_WUFC_MC)))
2281 		goto pme;
2282 
2283 	/* Advertise the wakeup capability */
2284 	ctrl = IGC_READ_REG(&adapter->hw, IGC_CTRL);
2285 	ctrl |= IGC_CTRL_ADVD3WUC;
2286 	IGC_WRITE_REG(&adapter->hw, IGC_CTRL, ctrl);
2287 
2288 	/* Enable wakeup by the MAC */
2289 	IGC_WRITE_REG(&adapter->hw, IGC_WUC, IGC_WUC_PME_EN);
2290 	IGC_WRITE_REG(&adapter->hw, IGC_WUFC, adapter->wol);
2291 
2292 pme:
2293 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
2294 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
2295 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
2296 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
2297 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
2298 
2299 	return;
2300 }
2301 
2302 /**********************************************************************
2303  *
2304  *  Update the board statistics counters.
2305  *
2306  **********************************************************************/
2307 static void
2308 igc_update_stats_counters(struct igc_adapter *adapter)
2309 {
2310 	u64 prev_xoffrxc = adapter->stats.xoffrxc;
2311 
2312 	adapter->stats.crcerrs += IGC_READ_REG(&adapter->hw, IGC_CRCERRS);
2313 	adapter->stats.mpc += IGC_READ_REG(&adapter->hw, IGC_MPC);
2314 	adapter->stats.scc += IGC_READ_REG(&adapter->hw, IGC_SCC);
2315 	adapter->stats.ecol += IGC_READ_REG(&adapter->hw, IGC_ECOL);
2316 
2317 	adapter->stats.mcc += IGC_READ_REG(&adapter->hw, IGC_MCC);
2318 	adapter->stats.latecol += IGC_READ_REG(&adapter->hw, IGC_LATECOL);
2319 	adapter->stats.colc += IGC_READ_REG(&adapter->hw, IGC_COLC);
2320 	adapter->stats.colc += IGC_READ_REG(&adapter->hw, IGC_RERC);
2321 	adapter->stats.dc += IGC_READ_REG(&adapter->hw, IGC_DC);
2322 	adapter->stats.rlec += IGC_READ_REG(&adapter->hw, IGC_RLEC);
2323 	adapter->stats.xonrxc += IGC_READ_REG(&adapter->hw, IGC_XONRXC);
2324 	adapter->stats.xontxc += IGC_READ_REG(&adapter->hw, IGC_XONTXC);
2325 	adapter->stats.xoffrxc += IGC_READ_REG(&adapter->hw, IGC_XOFFRXC);
2326 	/*
2327 	 * For watchdog management we need to know if we have been
2328 	 * paused during the last interval, so capture that here.
2329 	 */
2330 	if (adapter->stats.xoffrxc != prev_xoffrxc)
2331 		adapter->shared->isc_pause_frames = 1;
2332 	adapter->stats.xofftxc += IGC_READ_REG(&adapter->hw, IGC_XOFFTXC);
2333 	adapter->stats.fcruc += IGC_READ_REG(&adapter->hw, IGC_FCRUC);
2334 	adapter->stats.prc64 += IGC_READ_REG(&adapter->hw, IGC_PRC64);
2335 	adapter->stats.prc127 += IGC_READ_REG(&adapter->hw, IGC_PRC127);
2336 	adapter->stats.prc255 += IGC_READ_REG(&adapter->hw, IGC_PRC255);
2337 	adapter->stats.prc511 += IGC_READ_REG(&adapter->hw, IGC_PRC511);
2338 	adapter->stats.prc1023 += IGC_READ_REG(&adapter->hw, IGC_PRC1023);
2339 	adapter->stats.prc1522 += IGC_READ_REG(&adapter->hw, IGC_PRC1522);
2340 	adapter->stats.tlpic += IGC_READ_REG(&adapter->hw, IGC_TLPIC);
2341 	adapter->stats.rlpic += IGC_READ_REG(&adapter->hw, IGC_RLPIC);
2342 	adapter->stats.gprc += IGC_READ_REG(&adapter->hw, IGC_GPRC);
2343 	adapter->stats.bprc += IGC_READ_REG(&adapter->hw, IGC_BPRC);
2344 	adapter->stats.mprc += IGC_READ_REG(&adapter->hw, IGC_MPRC);
2345 	adapter->stats.gptc += IGC_READ_REG(&adapter->hw, IGC_GPTC);
2346 
2347 	/* For the 64-bit byte counters the low dword must be read first. */
2348 	/* Both registers clear on the read of the high dword */
2349 
2350 	adapter->stats.gorc += IGC_READ_REG(&adapter->hw, IGC_GORCL) +
2351 	    ((u64)IGC_READ_REG(&adapter->hw, IGC_GORCH) << 32);
2352 	adapter->stats.gotc += IGC_READ_REG(&adapter->hw, IGC_GOTCL) +
2353 	    ((u64)IGC_READ_REG(&adapter->hw, IGC_GOTCH) << 32);
2354 
2355 	adapter->stats.rnbc += IGC_READ_REG(&adapter->hw, IGC_RNBC);
2356 	adapter->stats.ruc += IGC_READ_REG(&adapter->hw, IGC_RUC);
2357 	adapter->stats.rfc += IGC_READ_REG(&adapter->hw, IGC_RFC);
2358 	adapter->stats.roc += IGC_READ_REG(&adapter->hw, IGC_ROC);
2359 	adapter->stats.rjc += IGC_READ_REG(&adapter->hw, IGC_RJC);
2360 
2361 	adapter->stats.tor += IGC_READ_REG(&adapter->hw, IGC_TORH);
2362 	adapter->stats.tot += IGC_READ_REG(&adapter->hw, IGC_TOTH);
2363 
2364 	adapter->stats.tpr += IGC_READ_REG(&adapter->hw, IGC_TPR);
2365 	adapter->stats.tpt += IGC_READ_REG(&adapter->hw, IGC_TPT);
2366 	adapter->stats.ptc64 += IGC_READ_REG(&adapter->hw, IGC_PTC64);
2367 	adapter->stats.ptc127 += IGC_READ_REG(&adapter->hw, IGC_PTC127);
2368 	adapter->stats.ptc255 += IGC_READ_REG(&adapter->hw, IGC_PTC255);
2369 	adapter->stats.ptc511 += IGC_READ_REG(&adapter->hw, IGC_PTC511);
2370 	adapter->stats.ptc1023 += IGC_READ_REG(&adapter->hw, IGC_PTC1023);
2371 	adapter->stats.ptc1522 += IGC_READ_REG(&adapter->hw, IGC_PTC1522);
2372 	adapter->stats.mptc += IGC_READ_REG(&adapter->hw, IGC_MPTC);
2373 	adapter->stats.bptc += IGC_READ_REG(&adapter->hw, IGC_BPTC);
2374 
2375 	/* Interrupt Counts */
2376 	adapter->stats.iac += IGC_READ_REG(&adapter->hw, IGC_IAC);
2377 	adapter->stats.rxdmtc += IGC_READ_REG(&adapter->hw, IGC_RXDMTC);
2378 
2379 	adapter->stats.algnerrc += IGC_READ_REG(&adapter->hw, IGC_ALGNERRC);
2380 	adapter->stats.tncrs += IGC_READ_REG(&adapter->hw, IGC_TNCRS);
2381 	adapter->stats.htdpmc += IGC_READ_REG(&adapter->hw, IGC_HTDPMC);
2382 	adapter->stats.tsctc += IGC_READ_REG(&adapter->hw, IGC_TSCTC);
2383 }
2384 
2385 static uint64_t
2386 igc_if_get_counter(if_ctx_t ctx, ift_counter cnt)
2387 {
2388 	struct igc_adapter *adapter = iflib_get_softc(ctx);
2389 	if_t ifp = iflib_get_ifp(ctx);
2390 
2391 	switch (cnt) {
2392 	case IFCOUNTER_COLLISIONS:
2393 		return (adapter->stats.colc);
2394 	case IFCOUNTER_IERRORS:
2395 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
2396 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
2397 		    adapter->stats.ruc + adapter->stats.roc +
2398 		    adapter->stats.mpc + adapter->stats.htdpmc);
2399 	case IFCOUNTER_OERRORS:
2400 		return (adapter->stats.ecol + adapter->stats.latecol +
2401 		    adapter->watchdog_events);
2402 	default:
2403 		return (if_get_counter_default(ifp, cnt));
2404 	}
2405 }
2406 
2407 /* igc_if_needs_restart - Tell iflib when the driver needs to be reinitialized
2408  * @ctx: iflib context
2409  * @event: event code to check
2410  *
2411  * Defaults to returning true for unknown events.
2412  *
2413  * @returns true if iflib needs to reinit the interface
2414  */
2415 static bool
2416 igc_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
2417 {
2418 	switch (event) {
2419 	case IFLIB_RESTART_VLAN_CONFIG:
2420 		return (false);
2421 	default:
2422 		return (true);
2423 	}
2424 }
2425 
2426 /* Export a single 32-bit register via a read-only sysctl. */
2427 static int
2428 igc_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
2429 {
2430 	struct igc_adapter *adapter;
2431 	u_int val;
2432 
2433 	adapter = oidp->oid_arg1;
2434 	val = IGC_READ_REG(&adapter->hw, oidp->oid_arg2);
2435 	return (sysctl_handle_int(oidp, &val, 0, req));
2436 }
2437 
2438 /*
2439  * Add sysctl variables, one per statistic, to the system.
2440  */
2441 static void
2442 igc_add_hw_stats(struct igc_adapter *adapter)
2443 {
2444 	device_t dev = iflib_get_dev(adapter->ctx);
2445 	struct igc_tx_queue *tx_que = adapter->tx_queues;
2446 	struct igc_rx_queue *rx_que = adapter->rx_queues;
2447 
2448 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
2449 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
2450 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
2451 	struct igc_hw_stats *stats = &adapter->stats;
2452 
2453 	struct sysctl_oid *stat_node, *queue_node, *int_node;
2454 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
2455 
2456 #define QUEUE_NAME_LEN 32
2457 	char namebuf[QUEUE_NAME_LEN];
2458 
2459 	/* Driver Statistics */
2460 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
2461 			CTLFLAG_RD, &adapter->dropped_pkts,
2462 			"Driver dropped packets");
2463 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
2464 			CTLFLAG_RD, &adapter->link_irq,
2465 			"Link MSI-X IRQ Handled");
2466 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
2467 			CTLFLAG_RD, &adapter->rx_overruns,
2468 			"RX overruns");
2469 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
2470 			CTLFLAG_RD, &adapter->watchdog_events,
2471 			"Watchdog timeouts");
2472 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
2473 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
2474 	    adapter, IGC_CTRL, igc_sysctl_reg_handler, "IU",
2475 	    "Device Control Register");
2476 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
2477 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
2478 	    adapter, IGC_RCTL, igc_sysctl_reg_handler, "IU",
2479 	    "Receiver Control Register");
2480 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
2481 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
2482 			"Flow Control High Watermark");
2483 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
2484 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
2485 			"Flow Control Low Watermark");
2486 
2487 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2488 		struct tx_ring *txr = &tx_que->txr;
2489 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
2490 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
2491 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
2492 		queue_list = SYSCTL_CHILDREN(queue_node);
2493 
2494 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
2495 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
2496 		    IGC_TDH(txr->me), igc_sysctl_reg_handler, "IU",
2497 		    "Transmit Descriptor Head");
2498 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
2499 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
2500 		    IGC_TDT(txr->me), igc_sysctl_reg_handler, "IU",
2501 		    "Transmit Descriptor Tail");
2502 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
2503 				CTLFLAG_RD, &txr->tx_irq,
2504 				"Queue MSI-X Transmit Interrupts");
2505 	}
2506 
2507 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
2508 		struct rx_ring *rxr = &rx_que->rxr;
2509 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
2510 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
2511 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
2512 		queue_list = SYSCTL_CHILDREN(queue_node);
2513 
2514 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
2515 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
2516 		    IGC_RDH(rxr->me), igc_sysctl_reg_handler, "IU",
2517 		    "Receive Descriptor Head");
2518 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
2519 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
2520 		    IGC_RDT(rxr->me), igc_sysctl_reg_handler, "IU",
2521 		    "Receive Descriptor Tail");
2522 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
2523 				CTLFLAG_RD, &rxr->rx_irq,
2524 				"Queue MSI-X Receive Interrupts");
2525 	}
2526 
2527 	/* MAC stats get their own sub node */
2528 
2529 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
2530 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
2531 	stat_list = SYSCTL_CHILDREN(stat_node);
2532 
2533 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
2534 			CTLFLAG_RD, &stats->ecol,
2535 			"Excessive collisions");
2536 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
2537 			CTLFLAG_RD, &stats->scc,
2538 			"Single collisions");
2539 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
2540 			CTLFLAG_RD, &stats->mcc,
2541 			"Multiple collisions");
2542 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
2543 			CTLFLAG_RD, &stats->latecol,
2544 			"Late collisions");
2545 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
2546 			CTLFLAG_RD, &stats->colc,
2547 			"Collision Count");
2548 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
2549 			CTLFLAG_RD, &adapter->stats.symerrs,
2550 			"Symbol Errors");
2551 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
2552 			CTLFLAG_RD, &adapter->stats.sec,
2553 			"Sequence Errors");
2554 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
2555 			CTLFLAG_RD, &adapter->stats.dc,
2556 			"Defer Count");
2557 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
2558 			CTLFLAG_RD, &adapter->stats.mpc,
2559 			"Missed Packets");
2560 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
2561 			CTLFLAG_RD, &adapter->stats.rnbc,
2562 			"Receive No Buffers");
2563 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
2564 			CTLFLAG_RD, &adapter->stats.ruc,
2565 			"Receive Undersize");
2566 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
2567 			CTLFLAG_RD, &adapter->stats.rfc,
2568 			"Fragmented Packets Received ");
2569 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
2570 			CTLFLAG_RD, &adapter->stats.roc,
2571 			"Oversized Packets Received");
2572 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
2573 			CTLFLAG_RD, &adapter->stats.rjc,
2574 			"Recevied Jabber");
2575 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
2576 			CTLFLAG_RD, &adapter->stats.rxerrc,
2577 			"Receive Errors");
2578 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
2579 			CTLFLAG_RD, &adapter->stats.crcerrs,
2580 			"CRC errors");
2581 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
2582 			CTLFLAG_RD, &adapter->stats.algnerrc,
2583 			"Alignment Errors");
2584 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
2585 			CTLFLAG_RD, &adapter->stats.xonrxc,
2586 			"XON Received");
2587 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
2588 			CTLFLAG_RD, &adapter->stats.xontxc,
2589 			"XON Transmitted");
2590 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
2591 			CTLFLAG_RD, &adapter->stats.xoffrxc,
2592 			"XOFF Received");
2593 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
2594 			CTLFLAG_RD, &adapter->stats.xofftxc,
2595 			"XOFF Transmitted");
2596 
2597 	/* Packet Reception Stats */
2598 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
2599 			CTLFLAG_RD, &adapter->stats.tpr,
2600 			"Total Packets Received ");
2601 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
2602 			CTLFLAG_RD, &adapter->stats.gprc,
2603 			"Good Packets Received");
2604 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
2605 			CTLFLAG_RD, &adapter->stats.bprc,
2606 			"Broadcast Packets Received");
2607 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
2608 			CTLFLAG_RD, &adapter->stats.mprc,
2609 			"Multicast Packets Received");
2610 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
2611 			CTLFLAG_RD, &adapter->stats.prc64,
2612 			"64 byte frames received ");
2613 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
2614 			CTLFLAG_RD, &adapter->stats.prc127,
2615 			"65-127 byte frames received");
2616 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
2617 			CTLFLAG_RD, &adapter->stats.prc255,
2618 			"128-255 byte frames received");
2619 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
2620 			CTLFLAG_RD, &adapter->stats.prc511,
2621 			"256-511 byte frames received");
2622 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
2623 			CTLFLAG_RD, &adapter->stats.prc1023,
2624 			"512-1023 byte frames received");
2625 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
2626 			CTLFLAG_RD, &adapter->stats.prc1522,
2627 			"1023-1522 byte frames received");
2628 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
2629 			CTLFLAG_RD, &adapter->stats.gorc,
2630 			"Good Octets Received");
2631 
2632 	/* Packet Transmission Stats */
2633 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
2634 			CTLFLAG_RD, &adapter->stats.gotc,
2635 			"Good Octets Transmitted");
2636 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
2637 			CTLFLAG_RD, &adapter->stats.tpt,
2638 			"Total Packets Transmitted");
2639 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
2640 			CTLFLAG_RD, &adapter->stats.gptc,
2641 			"Good Packets Transmitted");
2642 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
2643 			CTLFLAG_RD, &adapter->stats.bptc,
2644 			"Broadcast Packets Transmitted");
2645 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
2646 			CTLFLAG_RD, &adapter->stats.mptc,
2647 			"Multicast Packets Transmitted");
2648 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
2649 			CTLFLAG_RD, &adapter->stats.ptc64,
2650 			"64 byte frames transmitted ");
2651 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
2652 			CTLFLAG_RD, &adapter->stats.ptc127,
2653 			"65-127 byte frames transmitted");
2654 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
2655 			CTLFLAG_RD, &adapter->stats.ptc255,
2656 			"128-255 byte frames transmitted");
2657 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
2658 			CTLFLAG_RD, &adapter->stats.ptc511,
2659 			"256-511 byte frames transmitted");
2660 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
2661 			CTLFLAG_RD, &adapter->stats.ptc1023,
2662 			"512-1023 byte frames transmitted");
2663 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
2664 			CTLFLAG_RD, &adapter->stats.ptc1522,
2665 			"1024-1522 byte frames transmitted");
2666 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
2667 			CTLFLAG_RD, &adapter->stats.tsctc,
2668 			"TSO Contexts Transmitted");
2669 
2670 	/* Interrupt Stats */
2671 
2672 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
2673 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
2674 	int_list = SYSCTL_CHILDREN(int_node);
2675 
2676 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
2677 			CTLFLAG_RD, &adapter->stats.iac,
2678 			"Interrupt Assertion Count");
2679 
2680 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
2681 			CTLFLAG_RD, &adapter->stats.rxdmtc,
2682 			"Rx Desc Min Thresh Count");
2683 }
2684 
2685 /**********************************************************************
2686  *
2687  *  This routine provides a way to dump out the adapter eeprom,
2688  *  often a useful debug/service tool. This only dumps the first
2689  *  32 words, stuff that matters is in that extent.
2690  *
2691  **********************************************************************/
2692 static int
2693 igc_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
2694 {
2695 	struct igc_adapter *adapter = (struct igc_adapter *)arg1;
2696 	int error;
2697 	int result;
2698 
2699 	result = -1;
2700 	error = sysctl_handle_int(oidp, &result, 0, req);
2701 
2702 	if (error || !req->newptr)
2703 		return (error);
2704 
2705 	/*
2706 	 * This value will cause a hex dump of the
2707 	 * first 32 16-bit words of the EEPROM to
2708 	 * the screen.
2709 	 */
2710 	if (result == 1)
2711 		igc_print_nvm_info(adapter);
2712 
2713 	return (error);
2714 }
2715 
2716 static void
2717 igc_print_nvm_info(struct igc_adapter *adapter)
2718 {
2719 	u16 eeprom_data;
2720 	int i, j, row = 0;
2721 
2722 	/* Its a bit crude, but it gets the job done */
2723 	printf("\nInterface EEPROM Dump:\n");
2724 	printf("Offset\n0x0000  ");
2725 	for (i = 0, j = 0; i < 32; i++, j++) {
2726 		if (j == 8) { /* Make the offset block */
2727 			j = 0; ++row;
2728 			printf("\n0x00%x0  ",row);
2729 		}
2730 		igc_read_nvm(&adapter->hw, i, 1, &eeprom_data);
2731 		printf("%04x ", eeprom_data);
2732 	}
2733 	printf("\n");
2734 }
2735 
2736 static int
2737 igc_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
2738 {
2739 	struct igc_int_delay_info *info;
2740 	struct igc_adapter *adapter;
2741 	u32 regval;
2742 	int error, usecs, ticks;
2743 
2744 	info = (struct igc_int_delay_info *) arg1;
2745 	usecs = info->value;
2746 	error = sysctl_handle_int(oidp, &usecs, 0, req);
2747 	if (error != 0 || req->newptr == NULL)
2748 		return (error);
2749 	if (usecs < 0 || usecs > IGC_TICKS_TO_USECS(65535))
2750 		return (EINVAL);
2751 	info->value = usecs;
2752 	ticks = IGC_USECS_TO_TICKS(usecs);
2753 	if (info->offset == IGC_ITR)	/* units are 256ns here */
2754 		ticks *= 4;
2755 
2756 	adapter = info->adapter;
2757 
2758 	regval = IGC_READ_OFFSET(&adapter->hw, info->offset);
2759 	regval = (regval & ~0xffff) | (ticks & 0xffff);
2760 	/* Handle a few special cases. */
2761 	switch (info->offset) {
2762 	case IGC_RDTR:
2763 		break;
2764 	case IGC_TIDV:
2765 		if (ticks == 0) {
2766 			adapter->txd_cmd &= ~IGC_TXD_CMD_IDE;
2767 			/* Don't write 0 into the TIDV register. */
2768 			regval++;
2769 		} else
2770 			adapter->txd_cmd |= IGC_TXD_CMD_IDE;
2771 		break;
2772 	}
2773 	IGC_WRITE_OFFSET(&adapter->hw, info->offset, regval);
2774 	return (0);
2775 }
2776 
2777 static void
2778 igc_add_int_delay_sysctl(struct igc_adapter *adapter, const char *name,
2779 	const char *description, struct igc_int_delay_info *info,
2780 	int offset, int value)
2781 {
2782 	info->adapter = adapter;
2783 	info->offset = offset;
2784 	info->value = value;
2785 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
2786 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
2787 	    OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2788 	    info, 0, igc_sysctl_int_delay, "I", description);
2789 }
2790 
2791 /*
2792  * Set flow control using sysctl:
2793  * Flow control values:
2794  *      0 - off
2795  *      1 - rx pause
2796  *      2 - tx pause
2797  *      3 - full
2798  */
2799 static int
2800 igc_set_flowcntl(SYSCTL_HANDLER_ARGS)
2801 {
2802 	int error;
2803 	static int input = 3; /* default is full */
2804 	struct igc_adapter	*adapter = (struct igc_adapter *) arg1;
2805 
2806 	error = sysctl_handle_int(oidp, &input, 0, req);
2807 
2808 	if ((error) || (req->newptr == NULL))
2809 		return (error);
2810 
2811 	if (input == adapter->fc) /* no change? */
2812 		return (error);
2813 
2814 	switch (input) {
2815 	case igc_fc_rx_pause:
2816 	case igc_fc_tx_pause:
2817 	case igc_fc_full:
2818 	case igc_fc_none:
2819 		adapter->hw.fc.requested_mode = input;
2820 		adapter->fc = input;
2821 		break;
2822 	default:
2823 		/* Do nothing */
2824 		return (error);
2825 	}
2826 
2827 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
2828 	igc_force_mac_fc(&adapter->hw);
2829 	return (error);
2830 }
2831 
2832 /*
2833  * Manage Energy Efficient Ethernet:
2834  * Control values:
2835  *     0/1 - enabled/disabled
2836  */
2837 static int
2838 igc_sysctl_eee(SYSCTL_HANDLER_ARGS)
2839 {
2840 	struct igc_adapter *adapter = (struct igc_adapter *) arg1;
2841 	int error, value;
2842 
2843 	value = adapter->hw.dev_spec._i225.eee_disable;
2844 	error = sysctl_handle_int(oidp, &value, 0, req);
2845 	if (error || req->newptr == NULL)
2846 		return (error);
2847 
2848 	adapter->hw.dev_spec._i225.eee_disable = (value != 0);
2849 	igc_if_init(adapter->ctx);
2850 
2851 	return (0);
2852 }
2853 
2854 static int
2855 igc_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
2856 {
2857 	struct igc_adapter *adapter;
2858 	int error;
2859 	int result;
2860 
2861 	result = -1;
2862 	error = sysctl_handle_int(oidp, &result, 0, req);
2863 
2864 	if (error || !req->newptr)
2865 		return (error);
2866 
2867 	if (result == 1) {
2868 		adapter = (struct igc_adapter *) arg1;
2869 		igc_print_debug_info(adapter);
2870 	}
2871 
2872 	return (error);
2873 }
2874 
2875 static int
2876 igc_get_rs(SYSCTL_HANDLER_ARGS)
2877 {
2878 	struct igc_adapter *adapter = (struct igc_adapter *) arg1;
2879 	int error;
2880 	int result;
2881 
2882 	result = 0;
2883 	error = sysctl_handle_int(oidp, &result, 0, req);
2884 
2885 	if (error || !req->newptr || result != 1)
2886 		return (error);
2887 	igc_dump_rs(adapter);
2888 
2889 	return (error);
2890 }
2891 
2892 static void
2893 igc_if_debug(if_ctx_t ctx)
2894 {
2895 	igc_dump_rs(iflib_get_softc(ctx));
2896 }
2897 
2898 /*
2899  * This routine is meant to be fluid, add whatever is
2900  * needed for debugging a problem.  -jfv
2901  */
2902 static void
2903 igc_print_debug_info(struct igc_adapter *adapter)
2904 {
2905 	device_t dev = iflib_get_dev(adapter->ctx);
2906 	if_t ifp = iflib_get_ifp(adapter->ctx);
2907 	struct tx_ring *txr = &adapter->tx_queues->txr;
2908 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
2909 
2910 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
2911 		printf("Interface is RUNNING ");
2912 	else
2913 		printf("Interface is NOT RUNNING\n");
2914 
2915 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
2916 		printf("and INACTIVE\n");
2917 	else
2918 		printf("and ACTIVE\n");
2919 
2920 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2921 		device_printf(dev, "TX Queue %d ------\n", i);
2922 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
2923 			IGC_READ_REG(&adapter->hw, IGC_TDH(i)),
2924 			IGC_READ_REG(&adapter->hw, IGC_TDT(i)));
2925 
2926 	}
2927 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
2928 		device_printf(dev, "RX Queue %d ------\n", j);
2929 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
2930 			IGC_READ_REG(&adapter->hw, IGC_RDH(j)),
2931 			IGC_READ_REG(&adapter->hw, IGC_RDT(j)));
2932 	}
2933 }
2934