xref: /freebsd/sys/dev/ixgbe/if_ix.c (revision 1323ec571215a77ddd21294f0871979d5ad6b992)
1 /******************************************************************************
2 
3   Copyright (c) 2001-2017, Intel Corporation
4   All rights reserved.
5 
6   Redistribution and use in source and binary forms, with or without
7   modification, are permitted provided that the following conditions are met:
8 
9    1. Redistributions of source code must retain the above copyright notice,
10       this list of conditions and the following disclaimer.
11 
12    2. Redistributions in binary form must reproduce the above copyright
13       notice, this list of conditions and the following disclaimer in the
14       documentation and/or other materials provided with the distribution.
15 
16    3. Neither the name of the Intel Corporation nor the names of its
17       contributors may be used to endorse or promote products derived from
18       this software without specific prior written permission.
19 
20   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
21   AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22   IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23   ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24   LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25   CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26   SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27   INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28   CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29   ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30   POSSIBILITY OF SUCH DAMAGE.
31 
32 ******************************************************************************/
33 /*$FreeBSD$*/
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_rss.h"
38 
39 #include "ixgbe.h"
40 #include "ixgbe_sriov.h"
41 #include "ifdi_if.h"
42 
43 #include <net/netmap.h>
44 #include <dev/netmap/netmap_kern.h>
45 
46 /************************************************************************
47  * Driver version
48  ************************************************************************/
49 char ixgbe_driver_version[] = "4.0.1-k";
50 
51 /************************************************************************
52  * PCI Device ID Table
53  *
54  *   Used by probe to select devices to load on
55  *   Last field stores an index into ixgbe_strings
56  *   Last entry must be all 0s
57  *
58  *   { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
59  ************************************************************************/
60 static pci_vendor_info_t ixgbe_vendor_info_array[] =
61 {
62   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT,  "Intel(R) 82598EB AF (Dual Fiber)"),
63   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT,  "Intel(R) 82598EB AF (Fiber)"),
64   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4,  "Intel(R) 82598EB AT (CX4)"),
65   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT,  "Intel(R) 82598EB AT"),
66   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2,  "Intel(R) 82598EB AT2"),
67   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598,  "Intel(R) 82598"),
68   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT,  "Intel(R) 82598EB AF DA (Dual Fiber)"),
69   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT,  "Intel(R) 82598EB AT (Dual CX4)"),
70   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR,  "Intel(R) 82598EB AF (Dual Fiber LR)"),
71   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM,  "Intel(R) 82598EB AF (Dual Fiber SR)"),
72   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM,  "Intel(R) 82598EB LOM"),
73   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4,  "Intel(R) X520 82599 (KX4)"),
74   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ,  "Intel(R) X520 82599 (KX4 Mezzanine)"),
75   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP,  "Intel(R) X520 82599ES (SFI/SFP+)"),
76   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM,  "Intel(R) X520 82599 (XAUI/BX4)"),
77   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4,  "Intel(R) X520 82599 (Dual CX4)"),
78   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM,  "Intel(R) X520-T 82599 LOM"),
79   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE,  "Intel(R) X520 82599 (Combined Backplane)"),
80   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE,  "Intel(R) X520 82599 (Backplane w/FCoE)"),
81   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2,  "Intel(R) X520 82599 (Dual SFP+)"),
82   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE,  "Intel(R) X520 82599 (Dual SFP+ w/FCoE)"),
83   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP,  "Intel(R) X520-1 82599EN (SFP+)"),
84   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP,  "Intel(R) X520-4 82599 (Quad SFP+)"),
85   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP,  "Intel(R) X520-Q1 82599 (QSFP+)"),
86   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T,  "Intel(R) X540-AT2"),
87   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1,  "Intel(R) X540-T1"),
88   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T,  "Intel(R) X550-T2"),
89   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1, "Intel(R) X550-T1"),
90   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR,  "Intel(R) X552 (KR Backplane)"),
91   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4,  "Intel(R) X552 (KX4 Backplane)"),
92   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T,  "Intel(R) X552/X557-AT (10GBASE-T)"),
93   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T,  "Intel(R) X552 (1000BASE-T)"),
94   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP, "Intel(R) X552 (SFP+)"),
95   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR, "Intel(R) X553 (KR Backplane)"),
96   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L, "Intel(R) X553 L (KR Backplane)"),
97   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP, "Intel(R) X553 (SFP+)"),
98   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N, "Intel(R) X553 N (SFP+)"),
99   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII, "Intel(R) X553 (1GbE SGMII)"),
100   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L, "Intel(R) X553 L (1GbE SGMII)"),
101   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T, "Intel(R) X553/X557-AT (10GBASE-T)"),
102   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T, "Intel(R) X553 (1GbE)"),
103   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L, "Intel(R) X553 L (1GbE)"),
104   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_BYPASS, "Intel(R) X540-T2 (Bypass)"),
105   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS, "Intel(R) X520 82599 (Bypass)"),
106 	/* required last entry */
107   PVID_END
108 };
109 
110 static void *ixgbe_register(device_t);
111 static int  ixgbe_if_attach_pre(if_ctx_t);
112 static int  ixgbe_if_attach_post(if_ctx_t);
113 static int  ixgbe_if_detach(if_ctx_t);
114 static int  ixgbe_if_shutdown(if_ctx_t);
115 static int  ixgbe_if_suspend(if_ctx_t);
116 static int  ixgbe_if_resume(if_ctx_t);
117 
118 static void ixgbe_if_stop(if_ctx_t);
119 void ixgbe_if_enable_intr(if_ctx_t);
120 static void ixgbe_if_disable_intr(if_ctx_t);
121 static void ixgbe_link_intr_enable(if_ctx_t);
122 static int  ixgbe_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
123 static void ixgbe_if_media_status(if_ctx_t, struct ifmediareq *);
124 static int  ixgbe_if_media_change(if_ctx_t);
125 static int  ixgbe_if_msix_intr_assign(if_ctx_t, int);
126 static int  ixgbe_if_mtu_set(if_ctx_t, uint32_t);
127 static void ixgbe_if_crcstrip_set(if_ctx_t, int, int);
128 static void ixgbe_if_multi_set(if_ctx_t);
129 static int  ixgbe_if_promisc_set(if_ctx_t, int);
130 static int  ixgbe_if_tx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
131 static int  ixgbe_if_rx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
132 static void ixgbe_if_queues_free(if_ctx_t);
133 static void ixgbe_if_timer(if_ctx_t, uint16_t);
134 static void ixgbe_if_update_admin_status(if_ctx_t);
135 static void ixgbe_if_vlan_register(if_ctx_t, u16);
136 static void ixgbe_if_vlan_unregister(if_ctx_t, u16);
137 static int  ixgbe_if_i2c_req(if_ctx_t, struct ifi2creq *);
138 static bool ixgbe_if_needs_restart(if_ctx_t, enum iflib_restart_event);
139 int ixgbe_intr(void *);
140 
141 /************************************************************************
142  * Function prototypes
143  ************************************************************************/
144 static uint64_t ixgbe_if_get_counter(if_ctx_t, ift_counter);
145 
146 static void ixgbe_enable_queue(struct ixgbe_softc *, u32);
147 static void ixgbe_disable_queue(struct ixgbe_softc *, u32);
148 static void ixgbe_add_device_sysctls(if_ctx_t);
149 static int  ixgbe_allocate_pci_resources(if_ctx_t);
150 static int  ixgbe_setup_low_power_mode(if_ctx_t);
151 
152 static void ixgbe_config_dmac(struct ixgbe_softc *);
153 static void ixgbe_configure_ivars(struct ixgbe_softc *);
154 static void ixgbe_set_ivar(struct ixgbe_softc *, u8, u8, s8);
155 static u8   *ixgbe_mc_array_itr(struct ixgbe_hw *, u8 **, u32 *);
156 static bool ixgbe_sfp_probe(if_ctx_t);
157 
158 static void ixgbe_free_pci_resources(if_ctx_t);
159 
160 static int  ixgbe_msix_link(void *);
161 static int  ixgbe_msix_que(void *);
162 static void ixgbe_initialize_rss_mapping(struct ixgbe_softc *);
163 static void ixgbe_initialize_receive_units(if_ctx_t);
164 static void ixgbe_initialize_transmit_units(if_ctx_t);
165 
166 static int  ixgbe_setup_interface(if_ctx_t);
167 static void ixgbe_init_device_features(struct ixgbe_softc *);
168 static void ixgbe_check_fan_failure(struct ixgbe_softc *, u32, bool);
169 static void ixgbe_sbuf_fw_version(struct ixgbe_hw *, struct sbuf *);
170 static void ixgbe_print_fw_version(if_ctx_t);
171 static void ixgbe_add_media_types(if_ctx_t);
172 static void ixgbe_update_stats_counters(struct ixgbe_softc *);
173 static void ixgbe_config_link(if_ctx_t);
174 static void ixgbe_get_slot_info(struct ixgbe_softc *);
175 static void ixgbe_check_wol_support(struct ixgbe_softc *);
176 static void ixgbe_enable_rx_drop(struct ixgbe_softc *);
177 static void ixgbe_disable_rx_drop(struct ixgbe_softc *);
178 
179 static void ixgbe_add_hw_stats(struct ixgbe_softc *);
180 static int  ixgbe_set_flowcntl(struct ixgbe_softc *, int);
181 static int  ixgbe_set_advertise(struct ixgbe_softc *, int);
182 static int  ixgbe_get_default_advertise(struct ixgbe_softc *);
183 static void ixgbe_setup_vlan_hw_support(if_ctx_t);
184 static void ixgbe_config_gpie(struct ixgbe_softc *);
185 static void ixgbe_config_delay_values(struct ixgbe_softc *);
186 
187 /* Sysctl handlers */
188 static int  ixgbe_sysctl_flowcntl(SYSCTL_HANDLER_ARGS);
189 static int  ixgbe_sysctl_advertise(SYSCTL_HANDLER_ARGS);
190 static int  ixgbe_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS);
191 static int  ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS);
192 static int  ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS);
193 static int  ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS);
194 static int  ixgbe_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS);
195 #ifdef IXGBE_DEBUG
196 static int  ixgbe_sysctl_power_state(SYSCTL_HANDLER_ARGS);
197 static int  ixgbe_sysctl_print_rss_config(SYSCTL_HANDLER_ARGS);
198 #endif
199 static int  ixgbe_sysctl_rdh_handler(SYSCTL_HANDLER_ARGS);
200 static int  ixgbe_sysctl_rdt_handler(SYSCTL_HANDLER_ARGS);
201 static int  ixgbe_sysctl_tdt_handler(SYSCTL_HANDLER_ARGS);
202 static int  ixgbe_sysctl_tdh_handler(SYSCTL_HANDLER_ARGS);
203 static int  ixgbe_sysctl_eee_state(SYSCTL_HANDLER_ARGS);
204 static int  ixgbe_sysctl_wol_enable(SYSCTL_HANDLER_ARGS);
205 static int  ixgbe_sysctl_wufc(SYSCTL_HANDLER_ARGS);
206 
207 /* Deferred interrupt tasklets */
208 static void ixgbe_handle_msf(void *);
209 static void ixgbe_handle_mod(void *);
210 static void ixgbe_handle_phy(void *);
211 
212 /************************************************************************
213  *  FreeBSD Device Interface Entry Points
214  ************************************************************************/
215 static device_method_t ix_methods[] = {
216 	/* Device interface */
217 	DEVMETHOD(device_register, ixgbe_register),
218 	DEVMETHOD(device_probe, iflib_device_probe),
219 	DEVMETHOD(device_attach, iflib_device_attach),
220 	DEVMETHOD(device_detach, iflib_device_detach),
221 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
222 	DEVMETHOD(device_suspend, iflib_device_suspend),
223 	DEVMETHOD(device_resume, iflib_device_resume),
224 #ifdef PCI_IOV
225 	DEVMETHOD(pci_iov_init, iflib_device_iov_init),
226 	DEVMETHOD(pci_iov_uninit, iflib_device_iov_uninit),
227 	DEVMETHOD(pci_iov_add_vf, iflib_device_iov_add_vf),
228 #endif /* PCI_IOV */
229 	DEVMETHOD_END
230 };
231 
232 static driver_t ix_driver = {
233 	"ix", ix_methods, sizeof(struct ixgbe_softc),
234 };
235 
236 devclass_t ix_devclass;
237 DRIVER_MODULE(ix, pci, ix_driver, ix_devclass, 0, 0);
238 IFLIB_PNP_INFO(pci, ix_driver, ixgbe_vendor_info_array);
239 MODULE_DEPEND(ix, pci, 1, 1, 1);
240 MODULE_DEPEND(ix, ether, 1, 1, 1);
241 MODULE_DEPEND(ix, iflib, 1, 1, 1);
242 
243 static device_method_t ixgbe_if_methods[] = {
244 	DEVMETHOD(ifdi_attach_pre, ixgbe_if_attach_pre),
245 	DEVMETHOD(ifdi_attach_post, ixgbe_if_attach_post),
246 	DEVMETHOD(ifdi_detach, ixgbe_if_detach),
247 	DEVMETHOD(ifdi_shutdown, ixgbe_if_shutdown),
248 	DEVMETHOD(ifdi_suspend, ixgbe_if_suspend),
249 	DEVMETHOD(ifdi_resume, ixgbe_if_resume),
250 	DEVMETHOD(ifdi_init, ixgbe_if_init),
251 	DEVMETHOD(ifdi_stop, ixgbe_if_stop),
252 	DEVMETHOD(ifdi_msix_intr_assign, ixgbe_if_msix_intr_assign),
253 	DEVMETHOD(ifdi_intr_enable, ixgbe_if_enable_intr),
254 	DEVMETHOD(ifdi_intr_disable, ixgbe_if_disable_intr),
255 	DEVMETHOD(ifdi_link_intr_enable, ixgbe_link_intr_enable),
256 	DEVMETHOD(ifdi_tx_queue_intr_enable, ixgbe_if_rx_queue_intr_enable),
257 	DEVMETHOD(ifdi_rx_queue_intr_enable, ixgbe_if_rx_queue_intr_enable),
258 	DEVMETHOD(ifdi_tx_queues_alloc, ixgbe_if_tx_queues_alloc),
259 	DEVMETHOD(ifdi_rx_queues_alloc, ixgbe_if_rx_queues_alloc),
260 	DEVMETHOD(ifdi_queues_free, ixgbe_if_queues_free),
261 	DEVMETHOD(ifdi_update_admin_status, ixgbe_if_update_admin_status),
262 	DEVMETHOD(ifdi_multi_set, ixgbe_if_multi_set),
263 	DEVMETHOD(ifdi_mtu_set, ixgbe_if_mtu_set),
264 	DEVMETHOD(ifdi_crcstrip_set, ixgbe_if_crcstrip_set),
265 	DEVMETHOD(ifdi_media_status, ixgbe_if_media_status),
266 	DEVMETHOD(ifdi_media_change, ixgbe_if_media_change),
267 	DEVMETHOD(ifdi_promisc_set, ixgbe_if_promisc_set),
268 	DEVMETHOD(ifdi_timer, ixgbe_if_timer),
269 	DEVMETHOD(ifdi_vlan_register, ixgbe_if_vlan_register),
270 	DEVMETHOD(ifdi_vlan_unregister, ixgbe_if_vlan_unregister),
271 	DEVMETHOD(ifdi_get_counter, ixgbe_if_get_counter),
272 	DEVMETHOD(ifdi_i2c_req, ixgbe_if_i2c_req),
273 	DEVMETHOD(ifdi_needs_restart, ixgbe_if_needs_restart),
274 #ifdef PCI_IOV
275 	DEVMETHOD(ifdi_iov_init, ixgbe_if_iov_init),
276 	DEVMETHOD(ifdi_iov_uninit, ixgbe_if_iov_uninit),
277 	DEVMETHOD(ifdi_iov_vf_add, ixgbe_if_iov_vf_add),
278 #endif /* PCI_IOV */
279 	DEVMETHOD_END
280 };
281 
282 /*
283  * TUNEABLE PARAMETERS:
284  */
285 
286 static SYSCTL_NODE(_hw, OID_AUTO, ix, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
287     "IXGBE driver parameters");
288 static driver_t ixgbe_if_driver = {
289   "ixgbe_if", ixgbe_if_methods, sizeof(struct ixgbe_softc)
290 };
291 
292 static int ixgbe_max_interrupt_rate = (4000000 / IXGBE_LOW_LATENCY);
293 SYSCTL_INT(_hw_ix, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
294     &ixgbe_max_interrupt_rate, 0, "Maximum interrupts per second");
295 
296 /* Flow control setting, default to full */
297 static int ixgbe_flow_control = ixgbe_fc_full;
298 SYSCTL_INT(_hw_ix, OID_AUTO, flow_control, CTLFLAG_RDTUN,
299     &ixgbe_flow_control, 0, "Default flow control used for all adapters");
300 
301 /* Advertise Speed, default to 0 (auto) */
302 static int ixgbe_advertise_speed = 0;
303 SYSCTL_INT(_hw_ix, OID_AUTO, advertise_speed, CTLFLAG_RDTUN,
304     &ixgbe_advertise_speed, 0, "Default advertised speed for all adapters");
305 
306 /*
307  * Smart speed setting, default to on
308  * this only works as a compile option
309  * right now as its during attach, set
310  * this to 'ixgbe_smart_speed_off' to
311  * disable.
312  */
313 static int ixgbe_smart_speed = ixgbe_smart_speed_on;
314 
315 /*
316  * MSI-X should be the default for best performance,
317  * but this allows it to be forced off for testing.
318  */
319 static int ixgbe_enable_msix = 1;
320 SYSCTL_INT(_hw_ix, OID_AUTO, enable_msix, CTLFLAG_RDTUN, &ixgbe_enable_msix, 0,
321     "Enable MSI-X interrupts");
322 
323 /*
324  * Defining this on will allow the use
325  * of unsupported SFP+ modules, note that
326  * doing so you are on your own :)
327  */
328 static int allow_unsupported_sfp = false;
329 SYSCTL_INT(_hw_ix, OID_AUTO, unsupported_sfp, CTLFLAG_RDTUN,
330     &allow_unsupported_sfp, 0,
331     "Allow unsupported SFP modules...use at your own risk");
332 
333 /*
334  * Not sure if Flow Director is fully baked,
335  * so we'll default to turning it off.
336  */
337 static int ixgbe_enable_fdir = 0;
338 SYSCTL_INT(_hw_ix, OID_AUTO, enable_fdir, CTLFLAG_RDTUN, &ixgbe_enable_fdir, 0,
339     "Enable Flow Director");
340 
341 /* Receive-Side Scaling */
342 static int ixgbe_enable_rss = 1;
343 SYSCTL_INT(_hw_ix, OID_AUTO, enable_rss, CTLFLAG_RDTUN, &ixgbe_enable_rss, 0,
344     "Enable Receive-Side Scaling (RSS)");
345 
346 /*
347  * AIM: Adaptive Interrupt Moderation
348  * which means that the interrupt rate
349  * is varied over time based on the
350  * traffic for that interrupt vector
351  */
352 static int ixgbe_enable_aim = false;
353 SYSCTL_INT(_hw_ix, OID_AUTO, enable_aim, CTLFLAG_RWTUN, &ixgbe_enable_aim, 0,
354     "Enable adaptive interrupt moderation");
355 
356 #if 0
357 /* Keep running tab on them for sanity check */
358 static int ixgbe_total_ports;
359 #endif
360 
361 MALLOC_DEFINE(M_IXGBE, "ix", "ix driver allocations");
362 
363 /*
364  * For Flow Director: this is the number of TX packets we sample
365  * for the filter pool, this means every 20th packet will be probed.
366  *
367  * This feature can be disabled by setting this to 0.
368  */
369 static int atr_sample_rate = 20;
370 
371 extern struct if_txrx ixgbe_txrx;
372 
373 static struct if_shared_ctx ixgbe_sctx_init = {
374 	.isc_magic = IFLIB_MAGIC,
375 	.isc_q_align = PAGE_SIZE,/* max(DBA_ALIGN, PAGE_SIZE) */
376 	.isc_tx_maxsize = IXGBE_TSO_SIZE + sizeof(struct ether_vlan_header),
377 	.isc_tx_maxsegsize = PAGE_SIZE,
378 	.isc_tso_maxsize = IXGBE_TSO_SIZE + sizeof(struct ether_vlan_header),
379 	.isc_tso_maxsegsize = PAGE_SIZE,
380 	.isc_rx_maxsize = PAGE_SIZE*4,
381 	.isc_rx_nsegments = 1,
382 	.isc_rx_maxsegsize = PAGE_SIZE*4,
383 	.isc_nfl = 1,
384 	.isc_ntxqs = 1,
385 	.isc_nrxqs = 1,
386 
387 	.isc_admin_intrcnt = 1,
388 	.isc_vendor_info = ixgbe_vendor_info_array,
389 	.isc_driver_version = ixgbe_driver_version,
390 	.isc_driver = &ixgbe_if_driver,
391 	.isc_flags = IFLIB_TSO_INIT_IP,
392 
393 	.isc_nrxd_min = {MIN_RXD},
394 	.isc_ntxd_min = {MIN_TXD},
395 	.isc_nrxd_max = {MAX_RXD},
396 	.isc_ntxd_max = {MAX_TXD},
397 	.isc_nrxd_default = {DEFAULT_RXD},
398 	.isc_ntxd_default = {DEFAULT_TXD},
399 };
400 
401 /************************************************************************
402  * ixgbe_if_tx_queues_alloc
403  ************************************************************************/
404 static int
405 ixgbe_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs,
406     int ntxqs, int ntxqsets)
407 {
408 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
409 	if_softc_ctx_t     scctx = sc->shared;
410 	struct ix_tx_queue *que;
411 	int                i, j, error;
412 
413 	MPASS(sc->num_tx_queues > 0);
414 	MPASS(sc->num_tx_queues == ntxqsets);
415 	MPASS(ntxqs == 1);
416 
417 	/* Allocate queue structure memory */
418 	sc->tx_queues =
419 	    (struct ix_tx_queue *)malloc(sizeof(struct ix_tx_queue) * ntxqsets,
420 	                                 M_IXGBE, M_NOWAIT | M_ZERO);
421 	if (!sc->tx_queues) {
422 		device_printf(iflib_get_dev(ctx),
423 		    "Unable to allocate TX ring memory\n");
424 		return (ENOMEM);
425 	}
426 
427 	for (i = 0, que = sc->tx_queues; i < ntxqsets; i++, que++) {
428 		struct tx_ring *txr = &que->txr;
429 
430 		/* In case SR-IOV is enabled, align the index properly */
431 		txr->me = ixgbe_vf_que_index(sc->iov_mode, sc->pool,
432 		    i);
433 
434 		txr->sc = que->sc = sc;
435 
436 		/* Allocate report status array */
437 		txr->tx_rsq = (qidx_t *)malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_IXGBE, M_NOWAIT | M_ZERO);
438 		if (txr->tx_rsq == NULL) {
439 			error = ENOMEM;
440 			goto fail;
441 		}
442 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
443 			txr->tx_rsq[j] = QIDX_INVALID;
444 		/* get the virtual and physical address of the hardware queues */
445 		txr->tail = IXGBE_TDT(txr->me);
446 		txr->tx_base = (union ixgbe_adv_tx_desc *)vaddrs[i];
447 		txr->tx_paddr = paddrs[i];
448 
449 		txr->bytes = 0;
450 		txr->total_packets = 0;
451 
452 		/* Set the rate at which we sample packets */
453 		if (sc->feat_en & IXGBE_FEATURE_FDIR)
454 			txr->atr_sample = atr_sample_rate;
455 
456 	}
457 
458 	device_printf(iflib_get_dev(ctx), "allocated for %d queues\n",
459 	    sc->num_tx_queues);
460 
461 	return (0);
462 
463 fail:
464 	ixgbe_if_queues_free(ctx);
465 
466 	return (error);
467 } /* ixgbe_if_tx_queues_alloc */
468 
469 /************************************************************************
470  * ixgbe_if_rx_queues_alloc
471  ************************************************************************/
472 static int
473 ixgbe_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs,
474     int nrxqs, int nrxqsets)
475 {
476 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
477 	struct ix_rx_queue *que;
478 	int                i;
479 
480 	MPASS(sc->num_rx_queues > 0);
481 	MPASS(sc->num_rx_queues == nrxqsets);
482 	MPASS(nrxqs == 1);
483 
484 	/* Allocate queue structure memory */
485 	sc->rx_queues =
486 	    (struct ix_rx_queue *)malloc(sizeof(struct ix_rx_queue)*nrxqsets,
487 	                                 M_IXGBE, M_NOWAIT | M_ZERO);
488 	if (!sc->rx_queues) {
489 		device_printf(iflib_get_dev(ctx),
490 		    "Unable to allocate TX ring memory\n");
491 		return (ENOMEM);
492 	}
493 
494 	for (i = 0, que = sc->rx_queues; i < nrxqsets; i++, que++) {
495 		struct rx_ring *rxr = &que->rxr;
496 
497 		/* In case SR-IOV is enabled, align the index properly */
498 		rxr->me = ixgbe_vf_que_index(sc->iov_mode, sc->pool,
499 		    i);
500 
501 		rxr->sc = que->sc = sc;
502 
503 		/* get the virtual and physical address of the hw queues */
504 		rxr->tail = IXGBE_RDT(rxr->me);
505 		rxr->rx_base = (union ixgbe_adv_rx_desc *)vaddrs[i];
506 		rxr->rx_paddr = paddrs[i];
507 		rxr->bytes = 0;
508 		rxr->que = que;
509 	}
510 
511 	device_printf(iflib_get_dev(ctx), "allocated for %d rx queues\n",
512 	    sc->num_rx_queues);
513 
514 	return (0);
515 } /* ixgbe_if_rx_queues_alloc */
516 
517 /************************************************************************
518  * ixgbe_if_queues_free
519  ************************************************************************/
520 static void
521 ixgbe_if_queues_free(if_ctx_t ctx)
522 {
523 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
524 	struct ix_tx_queue *tx_que = sc->tx_queues;
525 	struct ix_rx_queue *rx_que = sc->rx_queues;
526 	int                i;
527 
528 	if (tx_que != NULL) {
529 		for (i = 0; i < sc->num_tx_queues; i++, tx_que++) {
530 			struct tx_ring *txr = &tx_que->txr;
531 			if (txr->tx_rsq == NULL)
532 				break;
533 
534 			free(txr->tx_rsq, M_IXGBE);
535 			txr->tx_rsq = NULL;
536 		}
537 
538 		free(sc->tx_queues, M_IXGBE);
539 		sc->tx_queues = NULL;
540 	}
541 	if (rx_que != NULL) {
542 		free(sc->rx_queues, M_IXGBE);
543 		sc->rx_queues = NULL;
544 	}
545 } /* ixgbe_if_queues_free */
546 
547 /************************************************************************
548  * ixgbe_initialize_rss_mapping
549  ************************************************************************/
550 static void
551 ixgbe_initialize_rss_mapping(struct ixgbe_softc *sc)
552 {
553 	struct ixgbe_hw *hw = &sc->hw;
554 	u32             reta = 0, mrqc, rss_key[10];
555 	int             queue_id, table_size, index_mult;
556 	int             i, j;
557 	u32             rss_hash_config;
558 
559 	if (sc->feat_en & IXGBE_FEATURE_RSS) {
560 		/* Fetch the configured RSS key */
561 		rss_getkey((uint8_t *)&rss_key);
562 	} else {
563 		/* set up random bits */
564 		arc4rand(&rss_key, sizeof(rss_key), 0);
565 	}
566 
567 	/* Set multiplier for RETA setup and table size based on MAC */
568 	index_mult = 0x1;
569 	table_size = 128;
570 	switch (sc->hw.mac.type) {
571 	case ixgbe_mac_82598EB:
572 		index_mult = 0x11;
573 		break;
574 	case ixgbe_mac_X550:
575 	case ixgbe_mac_X550EM_x:
576 	case ixgbe_mac_X550EM_a:
577 		table_size = 512;
578 		break;
579 	default:
580 		break;
581 	}
582 
583 	/* Set up the redirection table */
584 	for (i = 0, j = 0; i < table_size; i++, j++) {
585 		if (j == sc->num_rx_queues)
586 			j = 0;
587 
588 		if (sc->feat_en & IXGBE_FEATURE_RSS) {
589 			/*
590 			 * Fetch the RSS bucket id for the given indirection
591 			 * entry. Cap it at the number of configured buckets
592 			 * (which is num_rx_queues.)
593 			 */
594 			queue_id = rss_get_indirection_to_bucket(i);
595 			queue_id = queue_id % sc->num_rx_queues;
596 		} else
597 			queue_id = (j * index_mult);
598 
599 		/*
600 		 * The low 8 bits are for hash value (n+0);
601 		 * The next 8 bits are for hash value (n+1), etc.
602 		 */
603 		reta = reta >> 8;
604 		reta = reta | (((uint32_t)queue_id) << 24);
605 		if ((i & 3) == 3) {
606 			if (i < 128)
607 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
608 			else
609 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
610 				    reta);
611 			reta = 0;
612 		}
613 	}
614 
615 	/* Now fill our hash function seeds */
616 	for (i = 0; i < 10; i++)
617 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), rss_key[i]);
618 
619 	/* Perform hash on these packet types */
620 	if (sc->feat_en & IXGBE_FEATURE_RSS)
621 		rss_hash_config = rss_gethashconfig();
622 	else {
623 		/*
624 		 * Disable UDP - IP fragments aren't currently being handled
625 		 * and so we end up with a mix of 2-tuple and 4-tuple
626 		 * traffic.
627 		 */
628 		rss_hash_config = RSS_HASHTYPE_RSS_IPV4
629 		                | RSS_HASHTYPE_RSS_TCP_IPV4
630 		                | RSS_HASHTYPE_RSS_IPV6
631 		                | RSS_HASHTYPE_RSS_TCP_IPV6
632 		                | RSS_HASHTYPE_RSS_IPV6_EX
633 		                | RSS_HASHTYPE_RSS_TCP_IPV6_EX;
634 	}
635 
636 	mrqc = IXGBE_MRQC_RSSEN;
637 	if (rss_hash_config & RSS_HASHTYPE_RSS_IPV4)
638 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4;
639 	if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV4)
640 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_TCP;
641 	if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6)
642 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6;
643 	if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6)
644 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
645 	if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6_EX)
646 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX;
647 	if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6_EX)
648 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_TCP;
649 	if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV4)
650 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
651 	if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6)
652 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
653 	if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6_EX)
654 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_UDP;
655 	mrqc |= ixgbe_get_mrqc(sc->iov_mode);
656 	IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
657 } /* ixgbe_initialize_rss_mapping */
658 
659 /************************************************************************
660  * ixgbe_initialize_receive_units - Setup receive registers and features.
661  ************************************************************************/
662 #define BSIZEPKT_ROUNDUP ((1<<IXGBE_SRRCTL_BSIZEPKT_SHIFT)-1)
663 
664 static void
665 ixgbe_initialize_receive_units(if_ctx_t ctx)
666 {
667 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
668 	if_softc_ctx_t     scctx = sc->shared;
669 	struct ixgbe_hw    *hw = &sc->hw;
670 	struct ifnet       *ifp = iflib_get_ifp(ctx);
671 	struct ix_rx_queue *que;
672 	int                i, j;
673 	u32                bufsz, fctrl, srrctl, rxcsum;
674 	u32                hlreg;
675 
676 	/*
677 	 * Make sure receives are disabled while
678 	 * setting up the descriptor ring
679 	 */
680 	ixgbe_disable_rx(hw);
681 
682 	/* Enable broadcasts */
683 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
684 	fctrl |= IXGBE_FCTRL_BAM;
685 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
686 		fctrl |= IXGBE_FCTRL_DPF;
687 		fctrl |= IXGBE_FCTRL_PMCF;
688 	}
689 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
690 
691 	/* Set for Jumbo Frames? */
692 	hlreg = IXGBE_READ_REG(hw, IXGBE_HLREG0);
693 	if (ifp->if_mtu > ETHERMTU)
694 		hlreg |= IXGBE_HLREG0_JUMBOEN;
695 	else
696 		hlreg &= ~IXGBE_HLREG0_JUMBOEN;
697 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg);
698 
699 	bufsz = (sc->rx_mbuf_sz + BSIZEPKT_ROUNDUP) >>
700 	    IXGBE_SRRCTL_BSIZEPKT_SHIFT;
701 
702 	/* Setup the Base and Length of the Rx Descriptor Ring */
703 	for (i = 0, que = sc->rx_queues; i < sc->num_rx_queues; i++, que++) {
704 		struct rx_ring *rxr = &que->rxr;
705 		u64            rdba = rxr->rx_paddr;
706 
707 		j = rxr->me;
708 
709 		/* Setup the Base and Length of the Rx Descriptor Ring */
710 		IXGBE_WRITE_REG(hw, IXGBE_RDBAL(j),
711 		    (rdba & 0x00000000ffffffffULL));
712 		IXGBE_WRITE_REG(hw, IXGBE_RDBAH(j), (rdba >> 32));
713 		IXGBE_WRITE_REG(hw, IXGBE_RDLEN(j),
714 		     scctx->isc_nrxd[0] * sizeof(union ixgbe_adv_rx_desc));
715 
716 		/* Set up the SRRCTL register */
717 		srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(j));
718 		srrctl &= ~IXGBE_SRRCTL_BSIZEHDR_MASK;
719 		srrctl &= ~IXGBE_SRRCTL_BSIZEPKT_MASK;
720 		srrctl |= bufsz;
721 		srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
722 
723 		/*
724 		 * Set DROP_EN iff we have no flow control and >1 queue.
725 		 * Note that srrctl was cleared shortly before during reset,
726 		 * so we do not need to clear the bit, but do it just in case
727 		 * this code is moved elsewhere.
728 		 */
729 		if (sc->num_rx_queues > 1 &&
730 		    sc->hw.fc.requested_mode == ixgbe_fc_none) {
731 			srrctl |= IXGBE_SRRCTL_DROP_EN;
732 		} else {
733 			srrctl &= ~IXGBE_SRRCTL_DROP_EN;
734 		}
735 
736 		IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(j), srrctl);
737 
738 		/* Setup the HW Rx Head and Tail Descriptor Pointers */
739 		IXGBE_WRITE_REG(hw, IXGBE_RDH(j), 0);
740 		IXGBE_WRITE_REG(hw, IXGBE_RDT(j), 0);
741 
742 		/* Set the driver rx tail address */
743 		rxr->tail =  IXGBE_RDT(rxr->me);
744 	}
745 
746 	if (sc->hw.mac.type != ixgbe_mac_82598EB) {
747 		u32 psrtype = IXGBE_PSRTYPE_TCPHDR
748 		            | IXGBE_PSRTYPE_UDPHDR
749 		            | IXGBE_PSRTYPE_IPV4HDR
750 		            | IXGBE_PSRTYPE_IPV6HDR;
751 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(0), psrtype);
752 	}
753 
754 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
755 
756 	ixgbe_initialize_rss_mapping(sc);
757 
758 	if (sc->num_rx_queues > 1) {
759 		/* RSS and RX IPP Checksum are mutually exclusive */
760 		rxcsum |= IXGBE_RXCSUM_PCSD;
761 	}
762 
763 	if (ifp->if_capenable & IFCAP_RXCSUM)
764 		rxcsum |= IXGBE_RXCSUM_PCSD;
765 
766 	/* This is useful for calculating UDP/IP fragment checksums */
767 	if (!(rxcsum & IXGBE_RXCSUM_PCSD))
768 		rxcsum |= IXGBE_RXCSUM_IPPCSE;
769 
770 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
771 
772 } /* ixgbe_initialize_receive_units */
773 
774 /************************************************************************
775  * ixgbe_initialize_transmit_units - Enable transmit units.
776  ************************************************************************/
777 static void
778 ixgbe_initialize_transmit_units(if_ctx_t ctx)
779 {
780 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
781 	struct ixgbe_hw    *hw = &sc->hw;
782 	if_softc_ctx_t     scctx = sc->shared;
783 	struct ix_tx_queue *que;
784 	int i;
785 
786 	/* Setup the Base and Length of the Tx Descriptor Ring */
787 	for (i = 0, que = sc->tx_queues; i < sc->num_tx_queues;
788 	    i++, que++) {
789 		struct tx_ring	   *txr = &que->txr;
790 		u64 tdba = txr->tx_paddr;
791 		u32 txctrl = 0;
792 		int j = txr->me;
793 
794 		IXGBE_WRITE_REG(hw, IXGBE_TDBAL(j),
795 		    (tdba & 0x00000000ffffffffULL));
796 		IXGBE_WRITE_REG(hw, IXGBE_TDBAH(j), (tdba >> 32));
797 		IXGBE_WRITE_REG(hw, IXGBE_TDLEN(j),
798 		    scctx->isc_ntxd[0] * sizeof(union ixgbe_adv_tx_desc));
799 
800 		/* Setup the HW Tx Head and Tail descriptor pointers */
801 		IXGBE_WRITE_REG(hw, IXGBE_TDH(j), 0);
802 		IXGBE_WRITE_REG(hw, IXGBE_TDT(j), 0);
803 
804 		/* Cache the tail address */
805 		txr->tail = IXGBE_TDT(txr->me);
806 
807 		txr->tx_rs_cidx = txr->tx_rs_pidx;
808 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
809 		for (int k = 0; k < scctx->isc_ntxd[0]; k++)
810 			txr->tx_rsq[k] = QIDX_INVALID;
811 
812 		/* Disable Head Writeback */
813 		/*
814 		 * Note: for X550 series devices, these registers are actually
815 		 * prefixed with TPH_ isntead of DCA_, but the addresses and
816 		 * fields remain the same.
817 		 */
818 		switch (hw->mac.type) {
819 		case ixgbe_mac_82598EB:
820 			txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL(j));
821 			break;
822 		default:
823 			txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL_82599(j));
824 			break;
825 		}
826 		txctrl &= ~IXGBE_DCA_TXCTRL_DESC_WRO_EN;
827 		switch (hw->mac.type) {
828 		case ixgbe_mac_82598EB:
829 			IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL(j), txctrl);
830 			break;
831 		default:
832 			IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL_82599(j), txctrl);
833 			break;
834 		}
835 
836 	}
837 
838 	if (hw->mac.type != ixgbe_mac_82598EB) {
839 		u32 dmatxctl, rttdcs;
840 
841 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
842 		dmatxctl |= IXGBE_DMATXCTL_TE;
843 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
844 		/* Disable arbiter to set MTQC */
845 		rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
846 		rttdcs |= IXGBE_RTTDCS_ARBDIS;
847 		IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
848 		IXGBE_WRITE_REG(hw, IXGBE_MTQC,
849 		    ixgbe_get_mtqc(sc->iov_mode));
850 		rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
851 		IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
852 	}
853 
854 } /* ixgbe_initialize_transmit_units */
855 
856 /************************************************************************
857  * ixgbe_register
858  ************************************************************************/
859 static void *
860 ixgbe_register(device_t dev)
861 {
862 	return (&ixgbe_sctx_init);
863 } /* ixgbe_register */
864 
865 /************************************************************************
866  * ixgbe_if_attach_pre - Device initialization routine, part 1
867  *
868  *   Called when the driver is being loaded.
869  *   Identifies the type of hardware, initializes the hardware,
870  *   and initializes iflib structures.
871  *
872  *   return 0 on success, positive on failure
873  ************************************************************************/
874 static int
875 ixgbe_if_attach_pre(if_ctx_t ctx)
876 {
877 	struct ixgbe_softc  *sc;
878 	device_t        dev;
879 	if_softc_ctx_t  scctx;
880 	struct ixgbe_hw *hw;
881 	int             error = 0;
882 	u32             ctrl_ext;
883 
884 	INIT_DEBUGOUT("ixgbe_attach: begin");
885 
886 	/* Allocate, clear, and link in our adapter structure */
887 	dev = iflib_get_dev(ctx);
888 	sc = iflib_get_softc(ctx);
889 	sc->hw.back = sc;
890 	sc->ctx = ctx;
891 	sc->dev = dev;
892 	scctx = sc->shared = iflib_get_softc_ctx(ctx);
893 	sc->media = iflib_get_media(ctx);
894 	hw = &sc->hw;
895 
896 	/* Determine hardware revision */
897 	hw->vendor_id = pci_get_vendor(dev);
898 	hw->device_id = pci_get_device(dev);
899 	hw->revision_id = pci_get_revid(dev);
900 	hw->subsystem_vendor_id = pci_get_subvendor(dev);
901 	hw->subsystem_device_id = pci_get_subdevice(dev);
902 
903 	/* Do base PCI setup - map BAR0 */
904 	if (ixgbe_allocate_pci_resources(ctx)) {
905 		device_printf(dev, "Allocation of PCI resources failed\n");
906 		return (ENXIO);
907 	}
908 
909 	/* let hardware know driver is loaded */
910 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
911 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
912 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
913 
914 	/*
915 	 * Initialize the shared code
916 	 */
917 	if (ixgbe_init_shared_code(hw) != 0) {
918 		device_printf(dev, "Unable to initialize the shared code\n");
919 		error = ENXIO;
920 		goto err_pci;
921 	}
922 
923 	if (hw->mbx.ops.init_params)
924 		hw->mbx.ops.init_params(hw);
925 
926 	hw->allow_unsupported_sfp = allow_unsupported_sfp;
927 
928 	if (hw->mac.type != ixgbe_mac_82598EB)
929 		hw->phy.smart_speed = ixgbe_smart_speed;
930 
931 	ixgbe_init_device_features(sc);
932 
933 	/* Enable WoL (if supported) */
934 	ixgbe_check_wol_support(sc);
935 
936 	/* Verify adapter fan is still functional (if applicable) */
937 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL) {
938 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
939 		ixgbe_check_fan_failure(sc, esdp, false);
940 	}
941 
942 	/* Ensure SW/FW semaphore is free */
943 	ixgbe_init_swfw_semaphore(hw);
944 
945 	/* Set an initial default flow control value */
946 	hw->fc.requested_mode = ixgbe_flow_control;
947 
948 	hw->phy.reset_if_overtemp = true;
949 	error = ixgbe_reset_hw(hw);
950 	hw->phy.reset_if_overtemp = false;
951 	if (error == IXGBE_ERR_SFP_NOT_PRESENT) {
952 		/*
953 		 * No optics in this port, set up
954 		 * so the timer routine will probe
955 		 * for later insertion.
956 		 */
957 		sc->sfp_probe = true;
958 		error = 0;
959 	} else if (error == IXGBE_ERR_SFP_NOT_SUPPORTED) {
960 		device_printf(dev, "Unsupported SFP+ module detected!\n");
961 		error = EIO;
962 		goto err_pci;
963 	} else if (error) {
964 		device_printf(dev, "Hardware initialization failed\n");
965 		error = EIO;
966 		goto err_pci;
967 	}
968 
969 	/* Make sure we have a good EEPROM before we read from it */
970 	if (ixgbe_validate_eeprom_checksum(&sc->hw, NULL) < 0) {
971 		device_printf(dev, "The EEPROM Checksum Is Not Valid\n");
972 		error = EIO;
973 		goto err_pci;
974 	}
975 
976 	error = ixgbe_start_hw(hw);
977 	switch (error) {
978 	case IXGBE_ERR_EEPROM_VERSION:
979 		device_printf(dev, "This device is a pre-production adapter/LOM.  Please be aware there may be issues associated with your hardware.\nIf you are experiencing problems please contact your Intel or hardware representative who provided you with this hardware.\n");
980 		break;
981 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
982 		device_printf(dev, "Unsupported SFP+ Module\n");
983 		error = EIO;
984 		goto err_pci;
985 	case IXGBE_ERR_SFP_NOT_PRESENT:
986 		device_printf(dev, "No SFP+ Module found\n");
987 		/* falls thru */
988 	default:
989 		break;
990 	}
991 
992 	/* Most of the iflib initialization... */
993 
994 	iflib_set_mac(ctx, hw->mac.addr);
995 	switch (sc->hw.mac.type) {
996 	case ixgbe_mac_X550:
997 	case ixgbe_mac_X550EM_x:
998 	case ixgbe_mac_X550EM_a:
999 		scctx->isc_rss_table_size = 512;
1000 		scctx->isc_ntxqsets_max = scctx->isc_nrxqsets_max = 64;
1001 		break;
1002 	default:
1003 		scctx->isc_rss_table_size = 128;
1004 		scctx->isc_ntxqsets_max = scctx->isc_nrxqsets_max = 16;
1005 	}
1006 
1007 	/* Allow legacy interrupts */
1008 	ixgbe_txrx.ift_legacy_intr = ixgbe_intr;
1009 
1010 	scctx->isc_txqsizes[0] =
1011 	    roundup2(scctx->isc_ntxd[0] * sizeof(union ixgbe_adv_tx_desc) +
1012 	    sizeof(u32), DBA_ALIGN),
1013 	scctx->isc_rxqsizes[0] =
1014 	    roundup2(scctx->isc_nrxd[0] * sizeof(union ixgbe_adv_rx_desc),
1015 	    DBA_ALIGN);
1016 
1017 	/* XXX */
1018 	scctx->isc_tx_csum_flags = CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_TSO |
1019 	    CSUM_IP6_TCP | CSUM_IP6_UDP | CSUM_IP6_TSO;
1020 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
1021 		scctx->isc_tx_nsegments = IXGBE_82598_SCATTER;
1022 	} else {
1023 		scctx->isc_tx_csum_flags |= CSUM_SCTP |CSUM_IP6_SCTP;
1024 		scctx->isc_tx_nsegments = IXGBE_82599_SCATTER;
1025 	}
1026 
1027 	scctx->isc_msix_bar = pci_msix_table_bar(dev);
1028 
1029 	scctx->isc_tx_tso_segments_max = scctx->isc_tx_nsegments;
1030 	scctx->isc_tx_tso_size_max = IXGBE_TSO_SIZE;
1031 	scctx->isc_tx_tso_segsize_max = PAGE_SIZE;
1032 
1033 	scctx->isc_txrx = &ixgbe_txrx;
1034 
1035 	scctx->isc_capabilities = scctx->isc_capenable = IXGBE_CAPS;
1036 
1037 	return (0);
1038 
1039 err_pci:
1040 	ctrl_ext = IXGBE_READ_REG(&sc->hw, IXGBE_CTRL_EXT);
1041 	ctrl_ext &= ~IXGBE_CTRL_EXT_DRV_LOAD;
1042 	IXGBE_WRITE_REG(&sc->hw, IXGBE_CTRL_EXT, ctrl_ext);
1043 	ixgbe_free_pci_resources(ctx);
1044 
1045 	return (error);
1046 } /* ixgbe_if_attach_pre */
1047 
1048  /*********************************************************************
1049  * ixgbe_if_attach_post - Device initialization routine, part 2
1050  *
1051  *   Called during driver load, but after interrupts and
1052  *   resources have been allocated and configured.
1053  *   Sets up some data structures not relevant to iflib.
1054  *
1055  *   return 0 on success, positive on failure
1056  *********************************************************************/
1057 static int
1058 ixgbe_if_attach_post(if_ctx_t ctx)
1059 {
1060 	device_t dev;
1061 	struct ixgbe_softc  *sc;
1062 	struct ixgbe_hw *hw;
1063 	int             error = 0;
1064 
1065 	dev = iflib_get_dev(ctx);
1066 	sc = iflib_get_softc(ctx);
1067 	hw = &sc->hw;
1068 
1069 
1070 	if (sc->intr_type == IFLIB_INTR_LEGACY &&
1071 		(sc->feat_cap & IXGBE_FEATURE_LEGACY_IRQ) == 0) {
1072 		device_printf(dev, "Device does not support legacy interrupts");
1073 		error = ENXIO;
1074 		goto err;
1075 	}
1076 
1077 	/* Allocate multicast array memory. */
1078 	sc->mta = malloc(sizeof(*sc->mta) *
1079 	                      MAX_NUM_MULTICAST_ADDRESSES, M_IXGBE, M_NOWAIT);
1080 	if (sc->mta == NULL) {
1081 		device_printf(dev, "Can not allocate multicast setup array\n");
1082 		error = ENOMEM;
1083 		goto err;
1084 	}
1085 
1086 	/* hw.ix defaults init */
1087 	ixgbe_set_advertise(sc, ixgbe_advertise_speed);
1088 
1089 	/* Enable the optics for 82599 SFP+ fiber */
1090 	ixgbe_enable_tx_laser(hw);
1091 
1092 	/* Enable power to the phy. */
1093 	ixgbe_set_phy_power(hw, true);
1094 
1095 	ixgbe_initialize_iov(sc);
1096 
1097 	error = ixgbe_setup_interface(ctx);
1098 	if (error) {
1099 		device_printf(dev, "Interface setup failed: %d\n", error);
1100 		goto err;
1101 	}
1102 
1103 	ixgbe_if_update_admin_status(ctx);
1104 
1105 	/* Initialize statistics */
1106 	ixgbe_update_stats_counters(sc);
1107 	ixgbe_add_hw_stats(sc);
1108 
1109 	/* Check PCIE slot type/speed/width */
1110 	ixgbe_get_slot_info(sc);
1111 
1112 	/*
1113 	 * Do time init and sysctl init here, but
1114 	 * only on the first port of a bypass sc.
1115 	 */
1116 	ixgbe_bypass_init(sc);
1117 
1118 	/* Display NVM and Option ROM versions */
1119 	ixgbe_print_fw_version(ctx);
1120 
1121 	/* Set an initial dmac value */
1122 	sc->dmac = 0;
1123 	/* Set initial advertised speeds (if applicable) */
1124 	sc->advertise = ixgbe_get_default_advertise(sc);
1125 
1126 	if (sc->feat_cap & IXGBE_FEATURE_SRIOV)
1127 		ixgbe_define_iov_schemas(dev, &error);
1128 
1129 	/* Add sysctls */
1130 	ixgbe_add_device_sysctls(ctx);
1131 
1132 	return (0);
1133 err:
1134 	return (error);
1135 } /* ixgbe_if_attach_post */
1136 
1137 /************************************************************************
1138  * ixgbe_check_wol_support
1139  *
1140  *   Checks whether the adapter's ports are capable of
1141  *   Wake On LAN by reading the adapter's NVM.
1142  *
1143  *   Sets each port's hw->wol_enabled value depending
1144  *   on the value read here.
1145  ************************************************************************/
1146 static void
1147 ixgbe_check_wol_support(struct ixgbe_softc *sc)
1148 {
1149 	struct ixgbe_hw *hw = &sc->hw;
1150 	u16             dev_caps = 0;
1151 
1152 	/* Find out WoL support for port */
1153 	sc->wol_support = hw->wol_enabled = 0;
1154 	ixgbe_get_device_caps(hw, &dev_caps);
1155 	if ((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
1156 	    ((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0) &&
1157 	     hw->bus.func == 0))
1158 		sc->wol_support = hw->wol_enabled = 1;
1159 
1160 	/* Save initial wake up filter configuration */
1161 	sc->wufc = IXGBE_READ_REG(hw, IXGBE_WUFC);
1162 
1163 	return;
1164 } /* ixgbe_check_wol_support */
1165 
1166 /************************************************************************
1167  * ixgbe_setup_interface
1168  *
1169  *   Setup networking device structure and register an interface.
1170  ************************************************************************/
1171 static int
1172 ixgbe_setup_interface(if_ctx_t ctx)
1173 {
1174 	struct ifnet   *ifp = iflib_get_ifp(ctx);
1175 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1176 
1177 	INIT_DEBUGOUT("ixgbe_setup_interface: begin");
1178 
1179 	if_setbaudrate(ifp, IF_Gbps(10));
1180 
1181 	sc->max_frame_size = ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1182 
1183 	sc->phy_layer = ixgbe_get_supported_physical_layer(&sc->hw);
1184 
1185 	ixgbe_add_media_types(ctx);
1186 
1187 	/* Autoselect media by default */
1188 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
1189 
1190 	return (0);
1191 } /* ixgbe_setup_interface */
1192 
1193 /************************************************************************
1194  * ixgbe_if_get_counter
1195  ************************************************************************/
1196 static uint64_t
1197 ixgbe_if_get_counter(if_ctx_t ctx, ift_counter cnt)
1198 {
1199 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1200 	if_t           ifp = iflib_get_ifp(ctx);
1201 
1202 	switch (cnt) {
1203 	case IFCOUNTER_IPACKETS:
1204 		return (sc->ipackets);
1205 	case IFCOUNTER_OPACKETS:
1206 		return (sc->opackets);
1207 	case IFCOUNTER_IBYTES:
1208 		return (sc->ibytes);
1209 	case IFCOUNTER_OBYTES:
1210 		return (sc->obytes);
1211 	case IFCOUNTER_IMCASTS:
1212 		return (sc->imcasts);
1213 	case IFCOUNTER_OMCASTS:
1214 		return (sc->omcasts);
1215 	case IFCOUNTER_COLLISIONS:
1216 		return (0);
1217 	case IFCOUNTER_IQDROPS:
1218 		return (sc->iqdrops);
1219 	case IFCOUNTER_OQDROPS:
1220 		return (0);
1221 	case IFCOUNTER_IERRORS:
1222 		return (sc->ierrors);
1223 	default:
1224 		return (if_get_counter_default(ifp, cnt));
1225 	}
1226 } /* ixgbe_if_get_counter */
1227 
1228 /************************************************************************
1229  * ixgbe_if_i2c_req
1230  ************************************************************************/
1231 static int
1232 ixgbe_if_i2c_req(if_ctx_t ctx, struct ifi2creq *req)
1233 {
1234 	struct ixgbe_softc		*sc = iflib_get_softc(ctx);
1235 	struct ixgbe_hw 	*hw = &sc->hw;
1236 	int 			i;
1237 
1238 
1239 	if (hw->phy.ops.read_i2c_byte == NULL)
1240 		return (ENXIO);
1241 	for (i = 0; i < req->len; i++)
1242 		hw->phy.ops.read_i2c_byte(hw, req->offset + i,
1243 		    req->dev_addr, &req->data[i]);
1244 	return (0);
1245 } /* ixgbe_if_i2c_req */
1246 
1247 /* ixgbe_if_needs_restart - Tell iflib when the driver needs to be reinitialized
1248  * @ctx: iflib context
1249  * @event: event code to check
1250  *
1251  * Defaults to returning true for unknown events.
1252  *
1253  * @returns true if iflib needs to reinit the interface
1254  */
1255 static bool
1256 ixgbe_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
1257 {
1258 	switch (event) {
1259 	case IFLIB_RESTART_VLAN_CONFIG:
1260 		return (false);
1261 	default:
1262 		return (true);
1263 	}
1264 }
1265 
1266 /************************************************************************
1267  * ixgbe_add_media_types
1268  ************************************************************************/
1269 static void
1270 ixgbe_add_media_types(if_ctx_t ctx)
1271 {
1272 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
1273 	struct ixgbe_hw *hw = &sc->hw;
1274 	device_t        dev = iflib_get_dev(ctx);
1275 	u64             layer;
1276 
1277 	layer = sc->phy_layer = ixgbe_get_supported_physical_layer(hw);
1278 
1279 	/* Media types with matching FreeBSD media defines */
1280 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T)
1281 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_T, 0, NULL);
1282 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_T)
1283 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
1284 	if (layer & IXGBE_PHYSICAL_LAYER_100BASE_TX)
1285 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
1286 	if (layer & IXGBE_PHYSICAL_LAYER_10BASE_T)
1287 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
1288 
1289 	if (hw->mac.type == ixgbe_mac_X550) {
1290 		ifmedia_add(sc->media, IFM_ETHER | IFM_2500_T, 0, NULL);
1291 		ifmedia_add(sc->media, IFM_ETHER | IFM_5000_T, 0, NULL);
1292 	}
1293 
1294 	if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU ||
1295 	    layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA)
1296 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_TWINAX, 0,
1297 		    NULL);
1298 
1299 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR) {
1300 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_LR, 0, NULL);
1301 		if (hw->phy.multispeed_fiber)
1302 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_LX, 0,
1303 			    NULL);
1304 	}
1305 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR) {
1306 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_SR, 0, NULL);
1307 		if (hw->phy.multispeed_fiber)
1308 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_SX, 0,
1309 			    NULL);
1310 	} else if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX)
1311 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_SX, 0, NULL);
1312 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4)
1313 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_CX4, 0, NULL);
1314 
1315 #ifdef IFM_ETH_XTYPE
1316 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
1317 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_KR, 0, NULL);
1318 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4)
1319 		ifmedia_add( sc->media, IFM_ETHER | IFM_10G_KX4, 0, NULL);
1320 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
1321 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_KX, 0, NULL);
1322 	if (layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX)
1323 		ifmedia_add(sc->media, IFM_ETHER | IFM_2500_KX, 0, NULL);
1324 #else
1325 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR) {
1326 		device_printf(dev, "Media supported: 10GbaseKR\n");
1327 		device_printf(dev, "10GbaseKR mapped to 10GbaseSR\n");
1328 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_SR, 0, NULL);
1329 	}
1330 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4) {
1331 		device_printf(dev, "Media supported: 10GbaseKX4\n");
1332 		device_printf(dev, "10GbaseKX4 mapped to 10GbaseCX4\n");
1333 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_CX4, 0, NULL);
1334 	}
1335 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) {
1336 		device_printf(dev, "Media supported: 1000baseKX\n");
1337 		device_printf(dev, "1000baseKX mapped to 1000baseCX\n");
1338 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_CX, 0, NULL);
1339 	}
1340 	if (layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX) {
1341 		device_printf(dev, "Media supported: 2500baseKX\n");
1342 		device_printf(dev, "2500baseKX mapped to 2500baseSX\n");
1343 		ifmedia_add(sc->media, IFM_ETHER | IFM_2500_SX, 0, NULL);
1344 	}
1345 #endif
1346 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_BX)
1347 		device_printf(dev, "Media supported: 1000baseBX\n");
1348 
1349 	if (hw->device_id == IXGBE_DEV_ID_82598AT) {
1350 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T | IFM_FDX,
1351 		    0, NULL);
1352 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
1353 	}
1354 
1355 	ifmedia_add(sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
1356 } /* ixgbe_add_media_types */
1357 
1358 /************************************************************************
1359  * ixgbe_is_sfp
1360  ************************************************************************/
1361 static inline bool
1362 ixgbe_is_sfp(struct ixgbe_hw *hw)
1363 {
1364 	switch (hw->mac.type) {
1365 	case ixgbe_mac_82598EB:
1366 		if (hw->phy.type == ixgbe_phy_nl)
1367 			return (true);
1368 		return (false);
1369 	case ixgbe_mac_82599EB:
1370 		switch (hw->mac.ops.get_media_type(hw)) {
1371 		case ixgbe_media_type_fiber:
1372 		case ixgbe_media_type_fiber_qsfp:
1373 			return (true);
1374 		default:
1375 			return (false);
1376 		}
1377 	case ixgbe_mac_X550EM_x:
1378 	case ixgbe_mac_X550EM_a:
1379 		if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber)
1380 			return (true);
1381 		return (false);
1382 	default:
1383 		return (false);
1384 	}
1385 } /* ixgbe_is_sfp */
1386 
1387 /************************************************************************
1388  * ixgbe_config_link
1389  ************************************************************************/
1390 static void
1391 ixgbe_config_link(if_ctx_t ctx)
1392 {
1393 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
1394 	struct ixgbe_hw *hw = &sc->hw;
1395 	u32             autoneg, err = 0;
1396 	bool            sfp, negotiate;
1397 
1398 	sfp = ixgbe_is_sfp(hw);
1399 
1400 	if (sfp) {
1401 		sc->task_requests |= IXGBE_REQUEST_TASK_MOD;
1402 		iflib_admin_intr_deferred(ctx);
1403 	} else {
1404 		if (hw->mac.ops.check_link)
1405 			err = ixgbe_check_link(hw, &sc->link_speed,
1406 			    &sc->link_up, false);
1407 		if (err)
1408 			return;
1409 		autoneg = hw->phy.autoneg_advertised;
1410 		if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
1411 			err = hw->mac.ops.get_link_capabilities(hw, &autoneg,
1412 			    &negotiate);
1413 		if (err)
1414 			return;
1415 
1416 		if (hw->mac.type == ixgbe_mac_X550 &&
1417 		    hw->phy.autoneg_advertised == 0) {
1418 			/*
1419 			 * 2.5G and 5G autonegotiation speeds on X550
1420 			 * are disabled by default due to reported
1421 			 * interoperability issues with some switches.
1422 			 *
1423 			 * The second condition checks if any operations
1424 			 * involving setting autonegotiation speeds have
1425 			 * been performed prior to this ixgbe_config_link()
1426 			 * call.
1427 			 *
1428 			 * If hw->phy.autoneg_advertised does not
1429 			 * equal 0, this means that the user might have
1430 			 * set autonegotiation speeds via the sysctl
1431 			 * before bringing the interface up. In this
1432 			 * case, we should not disable 2.5G and 5G
1433 			 * since that speeds might be selected by the
1434 			 * user.
1435 			 *
1436 			 * Otherwise (i.e. if hw->phy.autoneg_advertised
1437 			 * is set to 0), it is the first time we set
1438 			 * autonegotiation preferences and the default
1439 			 * set of speeds should exclude 2.5G and 5G.
1440 			 */
1441 			autoneg &= ~(IXGBE_LINK_SPEED_2_5GB_FULL |
1442 			    IXGBE_LINK_SPEED_5GB_FULL);
1443 		}
1444 
1445 		if (hw->mac.ops.setup_link)
1446 			err = hw->mac.ops.setup_link(hw, autoneg,
1447 			    sc->link_up);
1448 	}
1449 } /* ixgbe_config_link */
1450 
1451 /************************************************************************
1452  * ixgbe_update_stats_counters - Update board statistics counters.
1453  ************************************************************************/
1454 static void
1455 ixgbe_update_stats_counters(struct ixgbe_softc *sc)
1456 {
1457 	struct ixgbe_hw       *hw = &sc->hw;
1458 	struct ixgbe_hw_stats *stats = &sc->stats.pf;
1459 	u32                   missed_rx = 0, bprc, lxon, lxoff, total;
1460 	u32                   lxoffrxc;
1461 	u64                   total_missed_rx = 0;
1462 
1463 	stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
1464 	stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
1465 	stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
1466 	stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
1467 	stats->mpc[0] += IXGBE_READ_REG(hw, IXGBE_MPC(0));
1468 
1469 	for (int i = 0; i < 16; i++) {
1470 		stats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
1471 		stats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
1472 		stats->qprdc[i] += IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
1473 	}
1474 	stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
1475 	stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
1476 	stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
1477 
1478 	/* Hardware workaround, gprc counts missed packets */
1479 	stats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
1480 	stats->gprc -= missed_rx;
1481 
1482 	if (hw->mac.type != ixgbe_mac_82598EB) {
1483 		stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL) +
1484 		    ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
1485 		stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL) +
1486 		    ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
1487 		stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL) +
1488 		    ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
1489 		stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
1490 		lxoffrxc = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
1491 		stats->lxoffrxc += lxoffrxc;
1492 	} else {
1493 		stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
1494 		lxoffrxc = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
1495 		stats->lxoffrxc += lxoffrxc;
1496 		/* 82598 only has a counter in the high register */
1497 		stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
1498 		stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
1499 		stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
1500 	}
1501 
1502 	/*
1503 	 * For watchdog management we need to know if we have been paused
1504 	 * during the last interval, so capture that here.
1505 	*/
1506 	if (lxoffrxc)
1507 		sc->shared->isc_pause_frames = 1;
1508 
1509 	/*
1510 	 * Workaround: mprc hardware is incorrectly counting
1511 	 * broadcasts, so for now we subtract those.
1512 	 */
1513 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
1514 	stats->bprc += bprc;
1515 	stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
1516 	if (hw->mac.type == ixgbe_mac_82598EB)
1517 		stats->mprc -= bprc;
1518 
1519 	stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
1520 	stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
1521 	stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
1522 	stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
1523 	stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
1524 	stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
1525 
1526 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
1527 	stats->lxontxc += lxon;
1528 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
1529 	stats->lxofftxc += lxoff;
1530 	total = lxon + lxoff;
1531 
1532 	stats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
1533 	stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
1534 	stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
1535 	stats->gptc -= total;
1536 	stats->mptc -= total;
1537 	stats->ptc64 -= total;
1538 	stats->gotc -= total * ETHER_MIN_LEN;
1539 
1540 	stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
1541 	stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
1542 	stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
1543 	stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
1544 	stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
1545 	stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
1546 	stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
1547 	stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
1548 	stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
1549 	stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
1550 	stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
1551 	stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
1552 	stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
1553 	stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
1554 	stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
1555 	stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
1556 	stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
1557 	stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
1558 	/* Only read FCOE on 82599 */
1559 	if (hw->mac.type != ixgbe_mac_82598EB) {
1560 		stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
1561 		stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
1562 		stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
1563 		stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
1564 		stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
1565 	}
1566 
1567 	/* Fill out the OS statistics structure */
1568 	IXGBE_SET_IPACKETS(sc, stats->gprc);
1569 	IXGBE_SET_OPACKETS(sc, stats->gptc);
1570 	IXGBE_SET_IBYTES(sc, stats->gorc);
1571 	IXGBE_SET_OBYTES(sc, stats->gotc);
1572 	IXGBE_SET_IMCASTS(sc, stats->mprc);
1573 	IXGBE_SET_OMCASTS(sc, stats->mptc);
1574 	IXGBE_SET_COLLISIONS(sc, 0);
1575 	IXGBE_SET_IQDROPS(sc, total_missed_rx);
1576 
1577 	/*
1578 	 * Aggregate following types of errors as RX errors:
1579 	 * - CRC error count,
1580 	 * - illegal byte error count,
1581 	 * - checksum error count,
1582 	 * - missed packets count,
1583 	 * - length error count,
1584 	 * - undersized packets count,
1585 	 * - fragmented packets count,
1586 	 * - oversized packets count,
1587 	 * - jabber count.
1588 	 */
1589 	IXGBE_SET_IERRORS(sc, stats->crcerrs + stats->illerrc + stats->xec +
1590 	    stats->mpc[0] + stats->rlec + stats->ruc + stats->rfc + stats->roc +
1591 	    stats->rjc);
1592 } /* ixgbe_update_stats_counters */
1593 
1594 /************************************************************************
1595  * ixgbe_add_hw_stats
1596  *
1597  *   Add sysctl variables, one per statistic, to the system.
1598  ************************************************************************/
1599 static void
1600 ixgbe_add_hw_stats(struct ixgbe_softc *sc)
1601 {
1602 	device_t               dev = iflib_get_dev(sc->ctx);
1603 	struct ix_rx_queue     *rx_que;
1604 	struct ix_tx_queue     *tx_que;
1605 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
1606 	struct sysctl_oid      *tree = device_get_sysctl_tree(dev);
1607 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
1608 	struct ixgbe_hw_stats  *stats = &sc->stats.pf;
1609 	struct sysctl_oid      *stat_node, *queue_node;
1610 	struct sysctl_oid_list *stat_list, *queue_list;
1611 	int                    i;
1612 
1613 #define QUEUE_NAME_LEN 32
1614 	char                   namebuf[QUEUE_NAME_LEN];
1615 
1616 	/* Driver Statistics */
1617 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
1618 	    CTLFLAG_RD, &sc->dropped_pkts, "Driver dropped packets");
1619 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_events",
1620 	    CTLFLAG_RD, &sc->watchdog_events, "Watchdog timeouts");
1621 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
1622 	    CTLFLAG_RD, &sc->link_irq, "Link MSI-X IRQ Handled");
1623 
1624 	for (i = 0, tx_que = sc->tx_queues; i < sc->num_tx_queues; i++, tx_que++) {
1625 		struct tx_ring *txr = &tx_que->txr;
1626 		snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i);
1627 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
1628 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Queue Name");
1629 		queue_list = SYSCTL_CHILDREN(queue_node);
1630 
1631 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
1632 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, txr, 0,
1633 		    ixgbe_sysctl_tdh_handler, "IU", "Transmit Descriptor Head");
1634 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
1635 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, txr, 0,
1636 		    ixgbe_sysctl_tdt_handler, "IU", "Transmit Descriptor Tail");
1637 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tso_tx",
1638 		    CTLFLAG_RD, &txr->tso_tx, "TSO");
1639 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tx_packets",
1640 		    CTLFLAG_RD, &txr->total_packets,
1641 		    "Queue Packets Transmitted");
1642 	}
1643 
1644 	for (i = 0, rx_que = sc->rx_queues; i < sc->num_rx_queues; i++, rx_que++) {
1645 		struct rx_ring *rxr = &rx_que->rxr;
1646 		snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i);
1647 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
1648 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Queue Name");
1649 		queue_list = SYSCTL_CHILDREN(queue_node);
1650 
1651 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "interrupt_rate",
1652 		    CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
1653 		    &sc->rx_queues[i], 0,
1654 		    ixgbe_sysctl_interrupt_rate_handler, "IU",
1655 		    "Interrupt Rate");
1656 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "irqs",
1657 		    CTLFLAG_RD, &(sc->rx_queues[i].irqs),
1658 		    "irqs on this queue");
1659 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
1660 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, rxr, 0,
1661 		    ixgbe_sysctl_rdh_handler, "IU", "Receive Descriptor Head");
1662 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
1663 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, rxr, 0,
1664 		    ixgbe_sysctl_rdt_handler, "IU", "Receive Descriptor Tail");
1665 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_packets",
1666 		    CTLFLAG_RD, &rxr->rx_packets, "Queue Packets Received");
1667 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_bytes",
1668 		    CTLFLAG_RD, &rxr->rx_bytes, "Queue Bytes Received");
1669 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_copies",
1670 		    CTLFLAG_RD, &rxr->rx_copies, "Copied RX Frames");
1671 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_discarded",
1672 		    CTLFLAG_RD, &rxr->rx_discarded, "Discarded RX packets");
1673 	}
1674 
1675 	/* MAC stats get their own sub node */
1676 
1677 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
1678 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "MAC Statistics");
1679 	stat_list = SYSCTL_CHILDREN(stat_node);
1680 
1681 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_errs",
1682 	    CTLFLAG_RD, &sc->ierrors, IXGBE_SYSCTL_DESC_RX_ERRS);
1683 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
1684 	    CTLFLAG_RD, &stats->crcerrs, "CRC Errors");
1685 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "ill_errs",
1686 	    CTLFLAG_RD, &stats->illerrc, "Illegal Byte Errors");
1687 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "byte_errs",
1688 	    CTLFLAG_RD, &stats->errbc, "Byte Errors");
1689 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "short_discards",
1690 	    CTLFLAG_RD, &stats->mspdc, "MAC Short Packets Discarded");
1691 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "local_faults",
1692 	    CTLFLAG_RD, &stats->mlfc, "MAC Local Faults");
1693 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "remote_faults",
1694 	    CTLFLAG_RD, &stats->mrfc, "MAC Remote Faults");
1695 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rec_len_errs",
1696 	    CTLFLAG_RD, &stats->rlec, "Receive Length Errors");
1697 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_missed_packets",
1698 	    CTLFLAG_RD, &stats->mpc[0], "RX Missed Packet Count");
1699 
1700 	/* Flow Control stats */
1701 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
1702 	    CTLFLAG_RD, &stats->lxontxc, "Link XON Transmitted");
1703 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
1704 	    CTLFLAG_RD, &stats->lxonrxc, "Link XON Received");
1705 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
1706 	    CTLFLAG_RD, &stats->lxofftxc, "Link XOFF Transmitted");
1707 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
1708 	    CTLFLAG_RD, &stats->lxoffrxc, "Link XOFF Received");
1709 
1710 	/* Packet Reception Stats */
1711 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_octets_rcvd",
1712 	    CTLFLAG_RD, &stats->tor, "Total Octets Received");
1713 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_rcvd",
1714 	    CTLFLAG_RD, &stats->gorc, "Good Octets Received");
1715 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_rcvd",
1716 	    CTLFLAG_RD, &stats->tpr, "Total Packets Received");
1717 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_rcvd",
1718 	    CTLFLAG_RD, &stats->gprc, "Good Packets Received");
1719 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_rcvd",
1720 	    CTLFLAG_RD, &stats->mprc, "Multicast Packets Received");
1721 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_rcvd",
1722 	    CTLFLAG_RD, &stats->bprc, "Broadcast Packets Received");
1723 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
1724 	    CTLFLAG_RD, &stats->prc64, "64 byte frames received ");
1725 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
1726 	    CTLFLAG_RD, &stats->prc127, "65-127 byte frames received");
1727 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
1728 	    CTLFLAG_RD, &stats->prc255, "128-255 byte frames received");
1729 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
1730 	    CTLFLAG_RD, &stats->prc511, "256-511 byte frames received");
1731 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
1732 	    CTLFLAG_RD, &stats->prc1023, "512-1023 byte frames received");
1733 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
1734 	    CTLFLAG_RD, &stats->prc1522, "1023-1522 byte frames received");
1735 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersized",
1736 	    CTLFLAG_RD, &stats->ruc, "Receive Undersized");
1737 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
1738 	    CTLFLAG_RD, &stats->rfc, "Fragmented Packets Received ");
1739 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversized",
1740 	    CTLFLAG_RD, &stats->roc, "Oversized Packets Received");
1741 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabberd",
1742 	    CTLFLAG_RD, &stats->rjc, "Received Jabber");
1743 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_rcvd",
1744 	    CTLFLAG_RD, &stats->mngprc, "Management Packets Received");
1745 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_drpd",
1746 	    CTLFLAG_RD, &stats->mngptc, "Management Packets Dropped");
1747 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "checksum_errs",
1748 	    CTLFLAG_RD, &stats->xec, "Checksum Errors");
1749 
1750 	/* Packet Transmission Stats */
1751 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
1752 	    CTLFLAG_RD, &stats->gotc, "Good Octets Transmitted");
1753 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
1754 	    CTLFLAG_RD, &stats->tpt, "Total Packets Transmitted");
1755 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
1756 	    CTLFLAG_RD, &stats->gptc, "Good Packets Transmitted");
1757 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
1758 	    CTLFLAG_RD, &stats->bptc, "Broadcast Packets Transmitted");
1759 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
1760 	    CTLFLAG_RD, &stats->mptc, "Multicast Packets Transmitted");
1761 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_txd",
1762 	    CTLFLAG_RD, &stats->mngptc, "Management Packets Transmitted");
1763 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
1764 	    CTLFLAG_RD, &stats->ptc64, "64 byte frames transmitted ");
1765 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
1766 	    CTLFLAG_RD, &stats->ptc127, "65-127 byte frames transmitted");
1767 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
1768 	    CTLFLAG_RD, &stats->ptc255, "128-255 byte frames transmitted");
1769 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
1770 	    CTLFLAG_RD, &stats->ptc511, "256-511 byte frames transmitted");
1771 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
1772 	    CTLFLAG_RD, &stats->ptc1023, "512-1023 byte frames transmitted");
1773 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
1774 	    CTLFLAG_RD, &stats->ptc1522, "1024-1522 byte frames transmitted");
1775 } /* ixgbe_add_hw_stats */
1776 
1777 /************************************************************************
1778  * ixgbe_sysctl_tdh_handler - Transmit Descriptor Head handler function
1779  *
1780  *   Retrieves the TDH value from the hardware
1781  ************************************************************************/
1782 static int
1783 ixgbe_sysctl_tdh_handler(SYSCTL_HANDLER_ARGS)
1784 {
1785 	struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1);
1786 	int            error;
1787 	unsigned int   val;
1788 
1789 	if (!txr)
1790 		return (0);
1791 
1792 	val = IXGBE_READ_REG(&txr->sc->hw, IXGBE_TDH(txr->me));
1793 	error = sysctl_handle_int(oidp, &val, 0, req);
1794 	if (error || !req->newptr)
1795 		return error;
1796 
1797 	return (0);
1798 } /* ixgbe_sysctl_tdh_handler */
1799 
1800 /************************************************************************
1801  * ixgbe_sysctl_tdt_handler - Transmit Descriptor Tail handler function
1802  *
1803  *   Retrieves the TDT value from the hardware
1804  ************************************************************************/
1805 static int
1806 ixgbe_sysctl_tdt_handler(SYSCTL_HANDLER_ARGS)
1807 {
1808 	struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1);
1809 	int            error;
1810 	unsigned int   val;
1811 
1812 	if (!txr)
1813 		return (0);
1814 
1815 	val = IXGBE_READ_REG(&txr->sc->hw, IXGBE_TDT(txr->me));
1816 	error = sysctl_handle_int(oidp, &val, 0, req);
1817 	if (error || !req->newptr)
1818 		return error;
1819 
1820 	return (0);
1821 } /* ixgbe_sysctl_tdt_handler */
1822 
1823 /************************************************************************
1824  * ixgbe_sysctl_rdh_handler - Receive Descriptor Head handler function
1825  *
1826  *   Retrieves the RDH value from the hardware
1827  ************************************************************************/
1828 static int
1829 ixgbe_sysctl_rdh_handler(SYSCTL_HANDLER_ARGS)
1830 {
1831 	struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1);
1832 	int            error;
1833 	unsigned int   val;
1834 
1835 	if (!rxr)
1836 		return (0);
1837 
1838 	val = IXGBE_READ_REG(&rxr->sc->hw, IXGBE_RDH(rxr->me));
1839 	error = sysctl_handle_int(oidp, &val, 0, req);
1840 	if (error || !req->newptr)
1841 		return error;
1842 
1843 	return (0);
1844 } /* ixgbe_sysctl_rdh_handler */
1845 
1846 /************************************************************************
1847  * ixgbe_sysctl_rdt_handler - Receive Descriptor Tail handler function
1848  *
1849  *   Retrieves the RDT value from the hardware
1850  ************************************************************************/
1851 static int
1852 ixgbe_sysctl_rdt_handler(SYSCTL_HANDLER_ARGS)
1853 {
1854 	struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1);
1855 	int            error;
1856 	unsigned int   val;
1857 
1858 	if (!rxr)
1859 		return (0);
1860 
1861 	val = IXGBE_READ_REG(&rxr->sc->hw, IXGBE_RDT(rxr->me));
1862 	error = sysctl_handle_int(oidp, &val, 0, req);
1863 	if (error || !req->newptr)
1864 		return error;
1865 
1866 	return (0);
1867 } /* ixgbe_sysctl_rdt_handler */
1868 
1869 /************************************************************************
1870  * ixgbe_if_vlan_register
1871  *
1872  *   Run via vlan config EVENT, it enables us to use the
1873  *   HW Filter table since we can get the vlan id. This
1874  *   just creates the entry in the soft version of the
1875  *   VFTA, init will repopulate the real table.
1876  ************************************************************************/
1877 static void
1878 ixgbe_if_vlan_register(if_ctx_t ctx, u16 vtag)
1879 {
1880 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1881 	u16            index, bit;
1882 
1883 	index = (vtag >> 5) & 0x7F;
1884 	bit = vtag & 0x1F;
1885 	sc->shadow_vfta[index] |= (1 << bit);
1886 	++sc->num_vlans;
1887 	ixgbe_setup_vlan_hw_support(ctx);
1888 } /* ixgbe_if_vlan_register */
1889 
1890 /************************************************************************
1891  * ixgbe_if_vlan_unregister
1892  *
1893  *   Run via vlan unconfig EVENT, remove our entry in the soft vfta.
1894  ************************************************************************/
1895 static void
1896 ixgbe_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
1897 {
1898 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1899 	u16            index, bit;
1900 
1901 	index = (vtag >> 5) & 0x7F;
1902 	bit = vtag & 0x1F;
1903 	sc->shadow_vfta[index] &= ~(1 << bit);
1904 	--sc->num_vlans;
1905 	/* Re-init to load the changes */
1906 	ixgbe_setup_vlan_hw_support(ctx);
1907 } /* ixgbe_if_vlan_unregister */
1908 
1909 /************************************************************************
1910  * ixgbe_setup_vlan_hw_support
1911  ************************************************************************/
1912 static void
1913 ixgbe_setup_vlan_hw_support(if_ctx_t ctx)
1914 {
1915 	struct ifnet	*ifp = iflib_get_ifp(ctx);
1916 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
1917 	struct ixgbe_hw *hw = &sc->hw;
1918 	struct rx_ring  *rxr;
1919 	int             i;
1920 	u32             ctrl;
1921 
1922 
1923 	/*
1924 	 * We get here thru init_locked, meaning
1925 	 * a soft reset, this has already cleared
1926 	 * the VFTA and other state, so if there
1927 	 * have been no vlan's registered do nothing.
1928 	 */
1929 	if (sc->num_vlans == 0)
1930 		return;
1931 
1932 	/* Setup the queues for vlans */
1933 	if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) {
1934 		for (i = 0; i < sc->num_rx_queues; i++) {
1935 			rxr = &sc->rx_queues[i].rxr;
1936 			/* On 82599 the VLAN enable is per/queue in RXDCTL */
1937 			if (hw->mac.type != ixgbe_mac_82598EB) {
1938 				ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me));
1939 				ctrl |= IXGBE_RXDCTL_VME;
1940 				IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), ctrl);
1941 			}
1942 			rxr->vtag_strip = true;
1943 		}
1944 	}
1945 
1946 	if ((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0)
1947 		return;
1948 	/*
1949 	 * A soft reset zero's out the VFTA, so
1950 	 * we need to repopulate it now.
1951 	 */
1952 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1953 		if (sc->shadow_vfta[i] != 0)
1954 			IXGBE_WRITE_REG(hw, IXGBE_VFTA(i),
1955 			    sc->shadow_vfta[i]);
1956 
1957 	ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1958 	/* Enable the Filter Table if enabled */
1959 	if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) {
1960 		ctrl &= ~IXGBE_VLNCTRL_CFIEN;
1961 		ctrl |= IXGBE_VLNCTRL_VFE;
1962 	}
1963 	if (hw->mac.type == ixgbe_mac_82598EB)
1964 		ctrl |= IXGBE_VLNCTRL_VME;
1965 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1966 } /* ixgbe_setup_vlan_hw_support */
1967 
1968 /************************************************************************
1969  * ixgbe_get_slot_info
1970  *
1971  *   Get the width and transaction speed of
1972  *   the slot this adapter is plugged into.
1973  ************************************************************************/
1974 static void
1975 ixgbe_get_slot_info(struct ixgbe_softc *sc)
1976 {
1977 	device_t        dev = iflib_get_dev(sc->ctx);
1978 	struct ixgbe_hw *hw = &sc->hw;
1979 	int             bus_info_valid = true;
1980 	u32             offset;
1981 	u16             link;
1982 
1983 	/* Some devices are behind an internal bridge */
1984 	switch (hw->device_id) {
1985 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
1986 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
1987 		goto get_parent_info;
1988 	default:
1989 		break;
1990 	}
1991 
1992 	ixgbe_get_bus_info(hw);
1993 
1994 	/*
1995 	 * Some devices don't use PCI-E, but there is no need
1996 	 * to display "Unknown" for bus speed and width.
1997 	 */
1998 	switch (hw->mac.type) {
1999 	case ixgbe_mac_X550EM_x:
2000 	case ixgbe_mac_X550EM_a:
2001 		return;
2002 	default:
2003 		goto display;
2004 	}
2005 
2006 get_parent_info:
2007 	/*
2008 	 * For the Quad port adapter we need to parse back
2009 	 * up the PCI tree to find the speed of the expansion
2010 	 * slot into which this adapter is plugged. A bit more work.
2011 	 */
2012 	dev = device_get_parent(device_get_parent(dev));
2013 #ifdef IXGBE_DEBUG
2014 	device_printf(dev, "parent pcib = %x,%x,%x\n", pci_get_bus(dev),
2015 	    pci_get_slot(dev), pci_get_function(dev));
2016 #endif
2017 	dev = device_get_parent(device_get_parent(dev));
2018 #ifdef IXGBE_DEBUG
2019 	device_printf(dev, "slot pcib = %x,%x,%x\n", pci_get_bus(dev),
2020 	    pci_get_slot(dev), pci_get_function(dev));
2021 #endif
2022 	/* Now get the PCI Express Capabilities offset */
2023 	if (pci_find_cap(dev, PCIY_EXPRESS, &offset)) {
2024 		/*
2025 		 * Hmm...can't get PCI-Express capabilities.
2026 		 * Falling back to default method.
2027 		 */
2028 		bus_info_valid = false;
2029 		ixgbe_get_bus_info(hw);
2030 		goto display;
2031 	}
2032 	/* ...and read the Link Status Register */
2033 	link = pci_read_config(dev, offset + PCIER_LINK_STA, 2);
2034 	ixgbe_set_pci_config_data_generic(hw, link);
2035 
2036 display:
2037 	device_printf(dev, "PCI Express Bus: Speed %s %s\n",
2038 	    ((hw->bus.speed == ixgbe_bus_speed_8000)    ? "8.0GT/s"  :
2039 	     (hw->bus.speed == ixgbe_bus_speed_5000)    ? "5.0GT/s"  :
2040 	     (hw->bus.speed == ixgbe_bus_speed_2500)    ? "2.5GT/s"  :
2041 	     "Unknown"),
2042 	    ((hw->bus.width == ixgbe_bus_width_pcie_x8) ? "Width x8" :
2043 	     (hw->bus.width == ixgbe_bus_width_pcie_x4) ? "Width x4" :
2044 	     (hw->bus.width == ixgbe_bus_width_pcie_x1) ? "Width x1" :
2045 	     "Unknown"));
2046 
2047 	if (bus_info_valid) {
2048 		if ((hw->device_id != IXGBE_DEV_ID_82599_SFP_SF_QP) &&
2049 		    ((hw->bus.width <= ixgbe_bus_width_pcie_x4) &&
2050 		    (hw->bus.speed == ixgbe_bus_speed_2500))) {
2051 			device_printf(dev, "PCI-Express bandwidth available for this card\n     is not sufficient for optimal performance.\n");
2052 			device_printf(dev, "For optimal performance a x8 PCIE, or x4 PCIE Gen2 slot is required.\n");
2053 		}
2054 		if ((hw->device_id == IXGBE_DEV_ID_82599_SFP_SF_QP) &&
2055 		    ((hw->bus.width <= ixgbe_bus_width_pcie_x8) &&
2056 		    (hw->bus.speed < ixgbe_bus_speed_8000))) {
2057 			device_printf(dev, "PCI-Express bandwidth available for this card\n     is not sufficient for optimal performance.\n");
2058 			device_printf(dev, "For optimal performance a x8 PCIE Gen3 slot is required.\n");
2059 		}
2060 	} else
2061 		device_printf(dev, "Unable to determine slot speed/width. The speed/width reported are that of the internal switch.\n");
2062 
2063 	return;
2064 } /* ixgbe_get_slot_info */
2065 
2066 /************************************************************************
2067  * ixgbe_if_msix_intr_assign
2068  *
2069  *   Setup MSI-X Interrupt resources and handlers
2070  ************************************************************************/
2071 static int
2072 ixgbe_if_msix_intr_assign(if_ctx_t ctx, int msix)
2073 {
2074 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
2075 	struct ix_rx_queue *rx_que = sc->rx_queues;
2076 	struct ix_tx_queue *tx_que;
2077 	int                error, rid, vector = 0;
2078 	char               buf[16];
2079 
2080 	/* Admin Que is vector 0*/
2081 	rid = vector + 1;
2082 	for (int i = 0; i < sc->num_rx_queues; i++, vector++, rx_que++) {
2083 		rid = vector + 1;
2084 
2085 		snprintf(buf, sizeof(buf), "rxq%d", i);
2086 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid,
2087 		    IFLIB_INTR_RXTX, ixgbe_msix_que, rx_que, rx_que->rxr.me, buf);
2088 
2089 		if (error) {
2090 			device_printf(iflib_get_dev(ctx),
2091 			    "Failed to allocate que int %d err: %d", i, error);
2092 			sc->num_rx_queues = i + 1;
2093 			goto fail;
2094 		}
2095 
2096 		rx_que->msix = vector;
2097 	}
2098 	for (int i = 0; i < sc->num_tx_queues; i++) {
2099 		snprintf(buf, sizeof(buf), "txq%d", i);
2100 		tx_que = &sc->tx_queues[i];
2101 		tx_que->msix = i % sc->num_rx_queues;
2102 		iflib_softirq_alloc_generic(ctx,
2103 		    &sc->rx_queues[tx_que->msix].que_irq,
2104 		    IFLIB_INTR_TX, tx_que, tx_que->txr.me, buf);
2105 	}
2106 	rid = vector + 1;
2107 	error = iflib_irq_alloc_generic(ctx, &sc->irq, rid,
2108 	    IFLIB_INTR_ADMIN, ixgbe_msix_link, sc, 0, "aq");
2109 	if (error) {
2110 		device_printf(iflib_get_dev(ctx),
2111 		    "Failed to register admin handler");
2112 		return (error);
2113 	}
2114 
2115 	sc->vector = vector;
2116 
2117 	return (0);
2118 fail:
2119 	iflib_irq_free(ctx, &sc->irq);
2120 	rx_que = sc->rx_queues;
2121 	for (int i = 0; i < sc->num_rx_queues; i++, rx_que++)
2122 		iflib_irq_free(ctx, &rx_que->que_irq);
2123 
2124 	return (error);
2125 } /* ixgbe_if_msix_intr_assign */
2126 
2127 static inline void
2128 ixgbe_perform_aim(struct ixgbe_softc *sc, struct ix_rx_queue *que)
2129 {
2130 	uint32_t newitr = 0;
2131 	struct rx_ring *rxr = &que->rxr;
2132 
2133 	/*
2134 	 * Do Adaptive Interrupt Moderation:
2135 	 *  - Write out last calculated setting
2136 	 *  - Calculate based on average size over
2137 	 *    the last interval.
2138 	 */
2139 	if (que->eitr_setting) {
2140 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EITR(que->msix),
2141 		    que->eitr_setting);
2142 	}
2143 
2144 	que->eitr_setting = 0;
2145 	/* Idle, do nothing */
2146 	if (rxr->bytes == 0) {
2147 		return;
2148 	}
2149 
2150 	if ((rxr->bytes) && (rxr->packets)) {
2151 		newitr = (rxr->bytes / rxr->packets);
2152 	}
2153 
2154 	newitr += 24; /* account for hardware frame, crc */
2155 	/* set an upper boundary */
2156 	newitr = min(newitr, 3000);
2157 
2158 	/* Be nice to the mid range */
2159 	if ((newitr > 300) && (newitr < 1200)) {
2160 		newitr = (newitr / 3);
2161 	} else {
2162 		newitr = (newitr / 2);
2163 	}
2164 
2165 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
2166 		newitr |= newitr << 16;
2167 	} else {
2168 		newitr |= IXGBE_EITR_CNT_WDIS;
2169 	}
2170 
2171 	/* save for next interrupt */
2172 	que->eitr_setting = newitr;
2173 
2174 	/* Reset state */
2175 	rxr->bytes = 0;
2176 	rxr->packets = 0;
2177 
2178 	return;
2179 }
2180 
2181 /*********************************************************************
2182  * ixgbe_msix_que - MSI-X Queue Interrupt Service routine
2183  **********************************************************************/
2184 static int
2185 ixgbe_msix_que(void *arg)
2186 {
2187 	struct ix_rx_queue *que = arg;
2188 	struct ixgbe_softc     *sc = que->sc;
2189 	struct ifnet       *ifp = iflib_get_ifp(que->sc->ctx);
2190 
2191 	/* Protect against spurious interrupts */
2192 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2193 		return (FILTER_HANDLED);
2194 
2195 	ixgbe_disable_queue(sc, que->msix);
2196 	++que->irqs;
2197 
2198 	/* Check for AIM */
2199 	if (sc->enable_aim) {
2200 		ixgbe_perform_aim(sc, que);
2201 	}
2202 
2203 	return (FILTER_SCHEDULE_THREAD);
2204 } /* ixgbe_msix_que */
2205 
2206 /************************************************************************
2207  * ixgbe_media_status - Media Ioctl callback
2208  *
2209  *   Called whenever the user queries the status of
2210  *   the interface using ifconfig.
2211  ************************************************************************/
2212 static void
2213 ixgbe_if_media_status(if_ctx_t ctx, struct ifmediareq * ifmr)
2214 {
2215 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
2216 	struct ixgbe_hw *hw = &sc->hw;
2217 	int             layer;
2218 
2219 	INIT_DEBUGOUT("ixgbe_if_media_status: begin");
2220 
2221 	ifmr->ifm_status = IFM_AVALID;
2222 	ifmr->ifm_active = IFM_ETHER;
2223 
2224 	if (!sc->link_active)
2225 		return;
2226 
2227 	ifmr->ifm_status |= IFM_ACTIVE;
2228 	layer = sc->phy_layer;
2229 
2230 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T ||
2231 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_T ||
2232 	    layer & IXGBE_PHYSICAL_LAYER_100BASE_TX ||
2233 	    layer & IXGBE_PHYSICAL_LAYER_10BASE_T)
2234 		switch (sc->link_speed) {
2235 		case IXGBE_LINK_SPEED_10GB_FULL:
2236 			ifmr->ifm_active |= IFM_10G_T | IFM_FDX;
2237 			break;
2238 		case IXGBE_LINK_SPEED_1GB_FULL:
2239 			ifmr->ifm_active |= IFM_1000_T | IFM_FDX;
2240 			break;
2241 		case IXGBE_LINK_SPEED_100_FULL:
2242 			ifmr->ifm_active |= IFM_100_TX | IFM_FDX;
2243 			break;
2244 		case IXGBE_LINK_SPEED_10_FULL:
2245 			ifmr->ifm_active |= IFM_10_T | IFM_FDX;
2246 			break;
2247 		}
2248 	if (hw->mac.type == ixgbe_mac_X550)
2249 		switch (sc->link_speed) {
2250 		case IXGBE_LINK_SPEED_5GB_FULL:
2251 			ifmr->ifm_active |= IFM_5000_T | IFM_FDX;
2252 			break;
2253 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2254 			ifmr->ifm_active |= IFM_2500_T | IFM_FDX;
2255 			break;
2256 		}
2257 	if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU ||
2258 	    layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA)
2259 		switch (sc->link_speed) {
2260 		case IXGBE_LINK_SPEED_10GB_FULL:
2261 			ifmr->ifm_active |= IFM_10G_TWINAX | IFM_FDX;
2262 			break;
2263 		}
2264 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR)
2265 		switch (sc->link_speed) {
2266 		case IXGBE_LINK_SPEED_10GB_FULL:
2267 			ifmr->ifm_active |= IFM_10G_LR | IFM_FDX;
2268 			break;
2269 		case IXGBE_LINK_SPEED_1GB_FULL:
2270 			ifmr->ifm_active |= IFM_1000_LX | IFM_FDX;
2271 			break;
2272 		}
2273 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LRM)
2274 		switch (sc->link_speed) {
2275 		case IXGBE_LINK_SPEED_10GB_FULL:
2276 			ifmr->ifm_active |= IFM_10G_LRM | IFM_FDX;
2277 			break;
2278 		case IXGBE_LINK_SPEED_1GB_FULL:
2279 			ifmr->ifm_active |= IFM_1000_LX | IFM_FDX;
2280 			break;
2281 		}
2282 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR ||
2283 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX)
2284 		switch (sc->link_speed) {
2285 		case IXGBE_LINK_SPEED_10GB_FULL:
2286 			ifmr->ifm_active |= IFM_10G_SR | IFM_FDX;
2287 			break;
2288 		case IXGBE_LINK_SPEED_1GB_FULL:
2289 			ifmr->ifm_active |= IFM_1000_SX | IFM_FDX;
2290 			break;
2291 		}
2292 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4)
2293 		switch (sc->link_speed) {
2294 		case IXGBE_LINK_SPEED_10GB_FULL:
2295 			ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX;
2296 			break;
2297 		}
2298 	/*
2299 	 * XXX: These need to use the proper media types once
2300 	 * they're added.
2301 	 */
2302 #ifndef IFM_ETH_XTYPE
2303 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
2304 		switch (sc->link_speed) {
2305 		case IXGBE_LINK_SPEED_10GB_FULL:
2306 			ifmr->ifm_active |= IFM_10G_SR | IFM_FDX;
2307 			break;
2308 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2309 			ifmr->ifm_active |= IFM_2500_SX | IFM_FDX;
2310 			break;
2311 		case IXGBE_LINK_SPEED_1GB_FULL:
2312 			ifmr->ifm_active |= IFM_1000_CX | IFM_FDX;
2313 			break;
2314 		}
2315 	else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4 ||
2316 	    layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX ||
2317 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
2318 		switch (sc->link_speed) {
2319 		case IXGBE_LINK_SPEED_10GB_FULL:
2320 			ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX;
2321 			break;
2322 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2323 			ifmr->ifm_active |= IFM_2500_SX | IFM_FDX;
2324 			break;
2325 		case IXGBE_LINK_SPEED_1GB_FULL:
2326 			ifmr->ifm_active |= IFM_1000_CX | IFM_FDX;
2327 			break;
2328 		}
2329 #else
2330 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
2331 		switch (sc->link_speed) {
2332 		case IXGBE_LINK_SPEED_10GB_FULL:
2333 			ifmr->ifm_active |= IFM_10G_KR | IFM_FDX;
2334 			break;
2335 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2336 			ifmr->ifm_active |= IFM_2500_KX | IFM_FDX;
2337 			break;
2338 		case IXGBE_LINK_SPEED_1GB_FULL:
2339 			ifmr->ifm_active |= IFM_1000_KX | IFM_FDX;
2340 			break;
2341 		}
2342 	else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4 ||
2343 	    layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX ||
2344 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
2345 		switch (sc->link_speed) {
2346 		case IXGBE_LINK_SPEED_10GB_FULL:
2347 			ifmr->ifm_active |= IFM_10G_KX4 | IFM_FDX;
2348 			break;
2349 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2350 			ifmr->ifm_active |= IFM_2500_KX | IFM_FDX;
2351 			break;
2352 		case IXGBE_LINK_SPEED_1GB_FULL:
2353 			ifmr->ifm_active |= IFM_1000_KX | IFM_FDX;
2354 			break;
2355 		}
2356 #endif
2357 
2358 	/* If nothing is recognized... */
2359 	if (IFM_SUBTYPE(ifmr->ifm_active) == 0)
2360 		ifmr->ifm_active |= IFM_UNKNOWN;
2361 
2362 	/* Display current flow control setting used on link */
2363 	if (hw->fc.current_mode == ixgbe_fc_rx_pause ||
2364 	    hw->fc.current_mode == ixgbe_fc_full)
2365 		ifmr->ifm_active |= IFM_ETH_RXPAUSE;
2366 	if (hw->fc.current_mode == ixgbe_fc_tx_pause ||
2367 	    hw->fc.current_mode == ixgbe_fc_full)
2368 		ifmr->ifm_active |= IFM_ETH_TXPAUSE;
2369 } /* ixgbe_media_status */
2370 
2371 /************************************************************************
2372  * ixgbe_media_change - Media Ioctl callback
2373  *
2374  *   Called when the user changes speed/duplex using
2375  *   media/mediopt option with ifconfig.
2376  ************************************************************************/
2377 static int
2378 ixgbe_if_media_change(if_ctx_t ctx)
2379 {
2380 	struct ixgbe_softc   *sc = iflib_get_softc(ctx);
2381 	struct ifmedia   *ifm = iflib_get_media(ctx);
2382 	struct ixgbe_hw  *hw = &sc->hw;
2383 	ixgbe_link_speed speed = 0;
2384 
2385 	INIT_DEBUGOUT("ixgbe_if_media_change: begin");
2386 
2387 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
2388 		return (EINVAL);
2389 
2390 	if (hw->phy.media_type == ixgbe_media_type_backplane)
2391 		return (EPERM);
2392 
2393 	/*
2394 	 * We don't actually need to check against the supported
2395 	 * media types of the adapter; ifmedia will take care of
2396 	 * that for us.
2397 	 */
2398 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
2399 	case IFM_AUTO:
2400 	case IFM_10G_T:
2401 		speed |= IXGBE_LINK_SPEED_100_FULL;
2402 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2403 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
2404 		break;
2405 	case IFM_10G_LRM:
2406 	case IFM_10G_LR:
2407 #ifndef IFM_ETH_XTYPE
2408 	case IFM_10G_SR: /* KR, too */
2409 	case IFM_10G_CX4: /* KX4 */
2410 #else
2411 	case IFM_10G_KR:
2412 	case IFM_10G_KX4:
2413 #endif
2414 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2415 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
2416 		break;
2417 #ifndef IFM_ETH_XTYPE
2418 	case IFM_1000_CX: /* KX */
2419 #else
2420 	case IFM_1000_KX:
2421 #endif
2422 	case IFM_1000_LX:
2423 	case IFM_1000_SX:
2424 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2425 		break;
2426 	case IFM_1000_T:
2427 		speed |= IXGBE_LINK_SPEED_100_FULL;
2428 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2429 		break;
2430 	case IFM_10G_TWINAX:
2431 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
2432 		break;
2433 	case IFM_5000_T:
2434 		speed |= IXGBE_LINK_SPEED_5GB_FULL;
2435 		break;
2436 	case IFM_2500_T:
2437 		speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2438 		break;
2439 	case IFM_100_TX:
2440 		speed |= IXGBE_LINK_SPEED_100_FULL;
2441 		break;
2442 	case IFM_10_T:
2443 		speed |= IXGBE_LINK_SPEED_10_FULL;
2444 		break;
2445 	default:
2446 		goto invalid;
2447 	}
2448 
2449 	hw->mac.autotry_restart = true;
2450 	hw->mac.ops.setup_link(hw, speed, true);
2451 	sc->advertise =
2452 	    ((speed & IXGBE_LINK_SPEED_10GB_FULL)  ? 0x4  : 0) |
2453 	    ((speed & IXGBE_LINK_SPEED_5GB_FULL)   ? 0x20 : 0) |
2454 	    ((speed & IXGBE_LINK_SPEED_2_5GB_FULL) ? 0x10 : 0) |
2455 	    ((speed & IXGBE_LINK_SPEED_1GB_FULL)   ? 0x2  : 0) |
2456 	    ((speed & IXGBE_LINK_SPEED_100_FULL)   ? 0x1  : 0) |
2457 	    ((speed & IXGBE_LINK_SPEED_10_FULL)    ? 0x8  : 0);
2458 
2459 	return (0);
2460 
2461 invalid:
2462 	device_printf(iflib_get_dev(ctx), "Invalid media type!\n");
2463 
2464 	return (EINVAL);
2465 } /* ixgbe_if_media_change */
2466 
2467 /************************************************************************
2468  * ixgbe_set_promisc
2469  ************************************************************************/
2470 static int
2471 ixgbe_if_promisc_set(if_ctx_t ctx, int flags)
2472 {
2473 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2474 	struct ifnet   *ifp = iflib_get_ifp(ctx);
2475 	u32            rctl;
2476 	int            mcnt = 0;
2477 
2478 	rctl = IXGBE_READ_REG(&sc->hw, IXGBE_FCTRL);
2479 	rctl &= (~IXGBE_FCTRL_UPE);
2480 	if (ifp->if_flags & IFF_ALLMULTI)
2481 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
2482 	else {
2483 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
2484 	}
2485 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
2486 		rctl &= (~IXGBE_FCTRL_MPE);
2487 	IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, rctl);
2488 
2489 	if (ifp->if_flags & IFF_PROMISC) {
2490 		rctl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
2491 		IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, rctl);
2492 	} else if (ifp->if_flags & IFF_ALLMULTI) {
2493 		rctl |= IXGBE_FCTRL_MPE;
2494 		rctl &= ~IXGBE_FCTRL_UPE;
2495 		IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, rctl);
2496 	}
2497 	return (0);
2498 } /* ixgbe_if_promisc_set */
2499 
2500 /************************************************************************
2501  * ixgbe_msix_link - Link status change ISR (MSI/MSI-X)
2502  ************************************************************************/
2503 static int
2504 ixgbe_msix_link(void *arg)
2505 {
2506 	struct ixgbe_softc  *sc = arg;
2507 	struct ixgbe_hw *hw = &sc->hw;
2508 	u32             eicr, eicr_mask;
2509 	s32             retval;
2510 
2511 	++sc->link_irq;
2512 
2513 	/* Pause other interrupts */
2514 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_OTHER);
2515 
2516 	/* First get the cause */
2517 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2518 	/* Be sure the queue bits are not cleared */
2519 	eicr &= ~IXGBE_EICR_RTX_QUEUE;
2520 	/* Clear interrupt with write */
2521 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2522 
2523 	/* Link status change */
2524 	if (eicr & IXGBE_EICR_LSC) {
2525 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2526 		sc->task_requests |= IXGBE_REQUEST_TASK_LSC;
2527 	}
2528 
2529 	if (sc->hw.mac.type != ixgbe_mac_82598EB) {
2530 		if ((sc->feat_en & IXGBE_FEATURE_FDIR) &&
2531 		    (eicr & IXGBE_EICR_FLOW_DIR)) {
2532 			/* This is probably overkill :) */
2533 			if (!atomic_cmpset_int(&sc->fdir_reinit, 0, 1))
2534 				return (FILTER_HANDLED);
2535 			/* Disable the interrupt */
2536 			IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EICR_FLOW_DIR);
2537 			sc->task_requests |= IXGBE_REQUEST_TASK_FDIR;
2538 		} else
2539 			if (eicr & IXGBE_EICR_ECC) {
2540 				device_printf(iflib_get_dev(sc->ctx),
2541 				   "\nCRITICAL: ECC ERROR!! Please Reboot!!\n");
2542 				IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2543 			}
2544 
2545 		/* Check for over temp condition */
2546 		if (sc->feat_en & IXGBE_FEATURE_TEMP_SENSOR) {
2547 			switch (sc->hw.mac.type) {
2548 			case ixgbe_mac_X550EM_a:
2549 				if (!(eicr & IXGBE_EICR_GPI_SDP0_X550EM_a))
2550 					break;
2551 				IXGBE_WRITE_REG(hw, IXGBE_EIMC,
2552 				    IXGBE_EICR_GPI_SDP0_X550EM_a);
2553 				IXGBE_WRITE_REG(hw, IXGBE_EICR,
2554 				    IXGBE_EICR_GPI_SDP0_X550EM_a);
2555 				retval = hw->phy.ops.check_overtemp(hw);
2556 				if (retval != IXGBE_ERR_OVERTEMP)
2557 					break;
2558 				device_printf(iflib_get_dev(sc->ctx),
2559 				    "\nCRITICAL: OVER TEMP!! PHY IS SHUT DOWN!!\n");
2560 				device_printf(iflib_get_dev(sc->ctx),
2561 				    "System shutdown required!\n");
2562 				break;
2563 			default:
2564 				if (!(eicr & IXGBE_EICR_TS))
2565 					break;
2566 				retval = hw->phy.ops.check_overtemp(hw);
2567 				if (retval != IXGBE_ERR_OVERTEMP)
2568 					break;
2569 				device_printf(iflib_get_dev(sc->ctx),
2570 				    "\nCRITICAL: OVER TEMP!! PHY IS SHUT DOWN!!\n");
2571 				device_printf(iflib_get_dev(sc->ctx),
2572 				    "System shutdown required!\n");
2573 				IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_TS);
2574 				break;
2575 			}
2576 		}
2577 
2578 		/* Check for VF message */
2579 		if ((sc->feat_en & IXGBE_FEATURE_SRIOV) &&
2580 		    (eicr & IXGBE_EICR_MAILBOX))
2581 			sc->task_requests |= IXGBE_REQUEST_TASK_MBX;
2582 	}
2583 
2584 	if (ixgbe_is_sfp(hw)) {
2585 		/* Pluggable optics-related interrupt */
2586 		if (hw->mac.type >= ixgbe_mac_X540)
2587 			eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2588 		else
2589 			eicr_mask = IXGBE_EICR_GPI_SDP2_BY_MAC(hw);
2590 
2591 		if (eicr & eicr_mask) {
2592 			IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2593 			sc->task_requests |= IXGBE_REQUEST_TASK_MOD;
2594 		}
2595 
2596 		if ((hw->mac.type == ixgbe_mac_82599EB) &&
2597 		    (eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw))) {
2598 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2599 			    IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
2600 			sc->task_requests |= IXGBE_REQUEST_TASK_MSF;
2601 		}
2602 	}
2603 
2604 	/* Check for fan failure */
2605 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL) {
2606 		ixgbe_check_fan_failure(sc, eicr, true);
2607 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
2608 	}
2609 
2610 	/* External PHY interrupt */
2611 	if ((hw->phy.type == ixgbe_phy_x550em_ext_t) &&
2612 	    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2613 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP0_X540);
2614 		sc->task_requests |= IXGBE_REQUEST_TASK_PHY;
2615 	}
2616 
2617 	return (sc->task_requests != 0) ? FILTER_SCHEDULE_THREAD : FILTER_HANDLED;
2618 } /* ixgbe_msix_link */
2619 
2620 /************************************************************************
2621  * ixgbe_sysctl_interrupt_rate_handler
2622  ************************************************************************/
2623 static int
2624 ixgbe_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS)
2625 {
2626 	struct ix_rx_queue *que = ((struct ix_rx_queue *)oidp->oid_arg1);
2627 	int                error;
2628 	unsigned int       reg, usec, rate;
2629 
2630 	reg = IXGBE_READ_REG(&que->sc->hw, IXGBE_EITR(que->msix));
2631 	usec = ((reg & 0x0FF8) >> 3);
2632 	if (usec > 0)
2633 		rate = 500000 / usec;
2634 	else
2635 		rate = 0;
2636 	error = sysctl_handle_int(oidp, &rate, 0, req);
2637 	if (error || !req->newptr)
2638 		return error;
2639 	reg &= ~0xfff; /* default, no limitation */
2640 	ixgbe_max_interrupt_rate = 0;
2641 	if (rate > 0 && rate < 500000) {
2642 		if (rate < 1000)
2643 			rate = 1000;
2644 		ixgbe_max_interrupt_rate = rate;
2645 		reg |= ((4000000/rate) & 0xff8);
2646 	}
2647 	IXGBE_WRITE_REG(&que->sc->hw, IXGBE_EITR(que->msix), reg);
2648 
2649 	return (0);
2650 } /* ixgbe_sysctl_interrupt_rate_handler */
2651 
2652 /************************************************************************
2653  * ixgbe_add_device_sysctls
2654  ************************************************************************/
2655 static void
2656 ixgbe_add_device_sysctls(if_ctx_t ctx)
2657 {
2658 	struct ixgbe_softc         *sc = iflib_get_softc(ctx);
2659 	device_t               dev = iflib_get_dev(ctx);
2660 	struct ixgbe_hw        *hw = &sc->hw;
2661 	struct sysctl_oid_list *child;
2662 	struct sysctl_ctx_list *ctx_list;
2663 
2664 	ctx_list = device_get_sysctl_ctx(dev);
2665 	child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
2666 
2667 	/* Sysctls for all devices */
2668 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc",
2669 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2670 	    sc, 0, ixgbe_sysctl_flowcntl, "I",
2671 	    IXGBE_SYSCTL_DESC_SET_FC);
2672 
2673 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "advertise_speed",
2674 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2675 	    sc, 0, ixgbe_sysctl_advertise, "I",
2676 	    IXGBE_SYSCTL_DESC_ADV_SPEED);
2677 
2678 	sc->enable_aim = ixgbe_enable_aim;
2679 	SYSCTL_ADD_INT(ctx_list, child, OID_AUTO, "enable_aim", CTLFLAG_RW,
2680 	    &sc->enable_aim, 0, "Interrupt Moderation");
2681 
2682 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fw_version",
2683 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
2684 	    ixgbe_sysctl_print_fw_version, "A", "Prints FW/NVM Versions");
2685 
2686 #ifdef IXGBE_DEBUG
2687 	/* testing sysctls (for all devices) */
2688 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "power_state",
2689 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2690 	    sc, 0, ixgbe_sysctl_power_state,
2691 	    "I", "PCI Power State");
2692 
2693 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "print_rss_config",
2694 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
2695 	    ixgbe_sysctl_print_rss_config, "A", "Prints RSS Configuration");
2696 #endif
2697 	/* for X550 series devices */
2698 	if (hw->mac.type >= ixgbe_mac_X550)
2699 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "dmac",
2700 		    CTLTYPE_U16 | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2701 		    sc, 0, ixgbe_sysctl_dmac,
2702 		    "I", "DMA Coalesce");
2703 
2704 	/* for WoL-capable devices */
2705 	if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) {
2706 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "wol_enable",
2707 		    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
2708 		    ixgbe_sysctl_wol_enable, "I", "Enable/Disable Wake on LAN");
2709 
2710 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "wufc",
2711 		    CTLTYPE_U32 | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2712 		    sc, 0, ixgbe_sysctl_wufc,
2713 		    "I", "Enable/Disable Wake Up Filters");
2714 	}
2715 
2716 	/* for X552/X557-AT devices */
2717 	if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) {
2718 		struct sysctl_oid *phy_node;
2719 		struct sysctl_oid_list *phy_list;
2720 
2721 		phy_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, "phy",
2722 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "External PHY sysctls");
2723 		phy_list = SYSCTL_CHILDREN(phy_node);
2724 
2725 		SYSCTL_ADD_PROC(ctx_list, phy_list, OID_AUTO, "temp",
2726 		    CTLTYPE_U16 | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
2727 		    sc, 0, ixgbe_sysctl_phy_temp,
2728 		    "I", "Current External PHY Temperature (Celsius)");
2729 
2730 		SYSCTL_ADD_PROC(ctx_list, phy_list, OID_AUTO,
2731 		    "overtemp_occurred",
2732 		    CTLTYPE_U16 | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
2733 		    ixgbe_sysctl_phy_overtemp_occurred, "I",
2734 		    "External PHY High Temperature Event Occurred");
2735 	}
2736 
2737 	if (sc->feat_cap & IXGBE_FEATURE_EEE) {
2738 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "eee_state",
2739 		    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
2740 		    ixgbe_sysctl_eee_state, "I", "EEE Power Save State");
2741 	}
2742 } /* ixgbe_add_device_sysctls */
2743 
2744 /************************************************************************
2745  * ixgbe_allocate_pci_resources
2746  ************************************************************************/
2747 static int
2748 ixgbe_allocate_pci_resources(if_ctx_t ctx)
2749 {
2750 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2751 	device_t        dev = iflib_get_dev(ctx);
2752 	int             rid;
2753 
2754 	rid = PCIR_BAR(0);
2755 	sc->pci_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
2756 	    RF_ACTIVE);
2757 
2758 	if (!(sc->pci_mem)) {
2759 		device_printf(dev, "Unable to allocate bus resource: memory\n");
2760 		return (ENXIO);
2761 	}
2762 
2763 	/* Save bus_space values for READ/WRITE_REG macros */
2764 	sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->pci_mem);
2765 	sc->osdep.mem_bus_space_handle =
2766 	    rman_get_bushandle(sc->pci_mem);
2767 	/* Set hw values for shared code */
2768 	sc->hw.hw_addr = (u8 *)&sc->osdep.mem_bus_space_handle;
2769 
2770 	return (0);
2771 } /* ixgbe_allocate_pci_resources */
2772 
2773 /************************************************************************
2774  * ixgbe_detach - Device removal routine
2775  *
2776  *   Called when the driver is being removed.
2777  *   Stops the adapter and deallocates all the resources
2778  *   that were allocated for driver operation.
2779  *
2780  *   return 0 on success, positive on failure
2781  ************************************************************************/
2782 static int
2783 ixgbe_if_detach(if_ctx_t ctx)
2784 {
2785 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2786 	device_t       dev = iflib_get_dev(ctx);
2787 	u32            ctrl_ext;
2788 
2789 	INIT_DEBUGOUT("ixgbe_detach: begin");
2790 
2791 	if (ixgbe_pci_iov_detach(dev) != 0) {
2792 		device_printf(dev, "SR-IOV in use; detach first.\n");
2793 		return (EBUSY);
2794 	}
2795 
2796 	ixgbe_setup_low_power_mode(ctx);
2797 
2798 	/* let hardware know driver is unloading */
2799 	ctrl_ext = IXGBE_READ_REG(&sc->hw, IXGBE_CTRL_EXT);
2800 	ctrl_ext &= ~IXGBE_CTRL_EXT_DRV_LOAD;
2801 	IXGBE_WRITE_REG(&sc->hw, IXGBE_CTRL_EXT, ctrl_ext);
2802 
2803 	ixgbe_free_pci_resources(ctx);
2804 	free(sc->mta, M_IXGBE);
2805 
2806 	return (0);
2807 } /* ixgbe_if_detach */
2808 
2809 /************************************************************************
2810  * ixgbe_setup_low_power_mode - LPLU/WoL preparation
2811  *
2812  *   Prepare the adapter/port for LPLU and/or WoL
2813  ************************************************************************/
2814 static int
2815 ixgbe_setup_low_power_mode(if_ctx_t ctx)
2816 {
2817 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
2818 	struct ixgbe_hw *hw = &sc->hw;
2819 	device_t        dev = iflib_get_dev(ctx);
2820 	s32             error = 0;
2821 
2822 	if (!hw->wol_enabled)
2823 		ixgbe_set_phy_power(hw, false);
2824 
2825 	/* Limit power management flow to X550EM baseT */
2826 	if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T &&
2827 	    hw->phy.ops.enter_lplu) {
2828 		/* Turn off support for APM wakeup. (Using ACPI instead) */
2829 		IXGBE_WRITE_REG(hw, IXGBE_GRC,
2830 		    IXGBE_READ_REG(hw, IXGBE_GRC) & ~(u32)2);
2831 
2832 		/*
2833 		 * Clear Wake Up Status register to prevent any previous wakeup
2834 		 * events from waking us up immediately after we suspend.
2835 		 */
2836 		IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff);
2837 
2838 		/*
2839 		 * Program the Wakeup Filter Control register with user filter
2840 		 * settings
2841 		 */
2842 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, sc->wufc);
2843 
2844 		/* Enable wakeups and power management in Wakeup Control */
2845 		IXGBE_WRITE_REG(hw, IXGBE_WUC,
2846 		    IXGBE_WUC_WKEN | IXGBE_WUC_PME_EN);
2847 
2848 		/* X550EM baseT adapters need a special LPLU flow */
2849 		hw->phy.reset_disable = true;
2850 		ixgbe_if_stop(ctx);
2851 		error = hw->phy.ops.enter_lplu(hw);
2852 		if (error)
2853 			device_printf(dev, "Error entering LPLU: %d\n", error);
2854 		hw->phy.reset_disable = false;
2855 	} else {
2856 		/* Just stop for other adapters */
2857 		ixgbe_if_stop(ctx);
2858 	}
2859 
2860 	return error;
2861 } /* ixgbe_setup_low_power_mode */
2862 
2863 /************************************************************************
2864  * ixgbe_shutdown - Shutdown entry point
2865  ************************************************************************/
2866 static int
2867 ixgbe_if_shutdown(if_ctx_t ctx)
2868 {
2869 	int error = 0;
2870 
2871 	INIT_DEBUGOUT("ixgbe_shutdown: begin");
2872 
2873 	error = ixgbe_setup_low_power_mode(ctx);
2874 
2875 	return (error);
2876 } /* ixgbe_if_shutdown */
2877 
2878 /************************************************************************
2879  * ixgbe_suspend
2880  *
2881  *   From D0 to D3
2882  ************************************************************************/
2883 static int
2884 ixgbe_if_suspend(if_ctx_t ctx)
2885 {
2886 	int error = 0;
2887 
2888 	INIT_DEBUGOUT("ixgbe_suspend: begin");
2889 
2890 	error = ixgbe_setup_low_power_mode(ctx);
2891 
2892 	return (error);
2893 } /* ixgbe_if_suspend */
2894 
2895 /************************************************************************
2896  * ixgbe_resume
2897  *
2898  *   From D3 to D0
2899  ************************************************************************/
2900 static int
2901 ixgbe_if_resume(if_ctx_t ctx)
2902 {
2903 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
2904 	device_t        dev = iflib_get_dev(ctx);
2905 	struct ifnet    *ifp = iflib_get_ifp(ctx);
2906 	struct ixgbe_hw *hw = &sc->hw;
2907 	u32             wus;
2908 
2909 	INIT_DEBUGOUT("ixgbe_resume: begin");
2910 
2911 	/* Read & clear WUS register */
2912 	wus = IXGBE_READ_REG(hw, IXGBE_WUS);
2913 	if (wus)
2914 		device_printf(dev, "Woken up by (WUS): %#010x\n",
2915 		    IXGBE_READ_REG(hw, IXGBE_WUS));
2916 	IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff);
2917 	/* And clear WUFC until next low-power transition */
2918 	IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
2919 
2920 	/*
2921 	 * Required after D3->D0 transition;
2922 	 * will re-advertise all previous advertised speeds
2923 	 */
2924 	if (ifp->if_flags & IFF_UP)
2925 		ixgbe_if_init(ctx);
2926 
2927 	return (0);
2928 } /* ixgbe_if_resume */
2929 
2930 /************************************************************************
2931  * ixgbe_if_mtu_set - Ioctl mtu entry point
2932  *
2933  *   Return 0 on success, EINVAL on failure
2934  ************************************************************************/
2935 static int
2936 ixgbe_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
2937 {
2938 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2939 	int error = 0;
2940 
2941 	IOCTL_DEBUGOUT("ioctl: SIOCIFMTU (Set Interface MTU)");
2942 
2943 	if (mtu > IXGBE_MAX_MTU) {
2944 		error = EINVAL;
2945 	} else {
2946 		sc->max_frame_size = mtu + IXGBE_MTU_HDR;
2947 	}
2948 
2949 	return error;
2950 } /* ixgbe_if_mtu_set */
2951 
2952 /************************************************************************
2953  * ixgbe_if_crcstrip_set
2954  ************************************************************************/
2955 static void
2956 ixgbe_if_crcstrip_set(if_ctx_t ctx, int onoff, int crcstrip)
2957 {
2958 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2959 	struct ixgbe_hw *hw = &sc->hw;
2960 	/* crc stripping is set in two places:
2961 	 * IXGBE_HLREG0 (modified on init_locked and hw reset)
2962 	 * IXGBE_RDRXCTL (set by the original driver in
2963 	 *	ixgbe_setup_hw_rsc() called in init_locked.
2964 	 *	We disable the setting when netmap is compiled in).
2965 	 * We update the values here, but also in ixgbe.c because
2966 	 * init_locked sometimes is called outside our control.
2967 	 */
2968 	uint32_t hl, rxc;
2969 
2970 	hl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
2971 	rxc = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
2972 #ifdef NETMAP
2973 	if (netmap_verbose)
2974 		D("%s read  HLREG 0x%x rxc 0x%x",
2975 			onoff ? "enter" : "exit", hl, rxc);
2976 #endif
2977 	/* hw requirements ... */
2978 	rxc &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
2979 	rxc |= IXGBE_RDRXCTL_RSCACKC;
2980 	if (onoff && !crcstrip) {
2981 		/* keep the crc. Fast rx */
2982 		hl &= ~IXGBE_HLREG0_RXCRCSTRP;
2983 		rxc &= ~IXGBE_RDRXCTL_CRCSTRIP;
2984 	} else {
2985 		/* reset default mode */
2986 		hl |= IXGBE_HLREG0_RXCRCSTRP;
2987 		rxc |= IXGBE_RDRXCTL_CRCSTRIP;
2988 	}
2989 #ifdef NETMAP
2990 	if (netmap_verbose)
2991 		D("%s write HLREG 0x%x rxc 0x%x",
2992 			onoff ? "enter" : "exit", hl, rxc);
2993 #endif
2994 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hl);
2995 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rxc);
2996 } /* ixgbe_if_crcstrip_set */
2997 
2998 /*********************************************************************
2999  * ixgbe_if_init - Init entry point
3000  *
3001  *   Used in two ways: It is used by the stack as an init
3002  *   entry point in network interface structure. It is also
3003  *   used by the driver as a hw/sw initialization routine to
3004  *   get to a consistent state.
3005  *
3006  *   Return 0 on success, positive on failure
3007  **********************************************************************/
3008 void
3009 ixgbe_if_init(if_ctx_t ctx)
3010 {
3011 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
3012 	struct ifnet       *ifp = iflib_get_ifp(ctx);
3013 	device_t           dev = iflib_get_dev(ctx);
3014 	struct ixgbe_hw *hw = &sc->hw;
3015 	struct ix_rx_queue *rx_que;
3016 	struct ix_tx_queue *tx_que;
3017 	u32             txdctl, mhadd;
3018 	u32             rxdctl, rxctrl;
3019 	u32             ctrl_ext;
3020 
3021 	int             i, j, err;
3022 
3023 	INIT_DEBUGOUT("ixgbe_if_init: begin");
3024 
3025 	/* Queue indices may change with IOV mode */
3026 	ixgbe_align_all_queue_indices(sc);
3027 
3028 	/* reprogram the RAR[0] in case user changed it. */
3029 	ixgbe_set_rar(hw, 0, hw->mac.addr, sc->pool, IXGBE_RAH_AV);
3030 
3031 	/* Get the latest mac address, User can use a LAA */
3032 	bcopy(IF_LLADDR(ifp), hw->mac.addr, IXGBE_ETH_LENGTH_OF_ADDRESS);
3033 	ixgbe_set_rar(hw, 0, hw->mac.addr, sc->pool, 1);
3034 	hw->addr_ctrl.rar_used_count = 1;
3035 
3036 	ixgbe_init_hw(hw);
3037 
3038 	ixgbe_initialize_iov(sc);
3039 
3040 	ixgbe_initialize_transmit_units(ctx);
3041 
3042 	/* Setup Multicast table */
3043 	ixgbe_if_multi_set(ctx);
3044 
3045 	/* Determine the correct mbuf pool, based on frame size */
3046 	sc->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
3047 
3048 	/* Configure RX settings */
3049 	ixgbe_initialize_receive_units(ctx);
3050 
3051 	/*
3052 	 * Initialize variable holding task enqueue requests
3053 	 * from MSI-X interrupts
3054 	 */
3055 	sc->task_requests = 0;
3056 
3057 	/* Enable SDP & MSI-X interrupts based on adapter */
3058 	ixgbe_config_gpie(sc);
3059 
3060 	/* Set MTU size */
3061 	if (ifp->if_mtu > ETHERMTU) {
3062 		/* aka IXGBE_MAXFRS on 82599 and newer */
3063 		mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3064 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3065 		mhadd |= sc->max_frame_size << IXGBE_MHADD_MFS_SHIFT;
3066 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3067 	}
3068 
3069 	/* Now enable all the queues */
3070 	for (i = 0, tx_que = sc->tx_queues; i < sc->num_tx_queues; i++, tx_que++) {
3071 		struct tx_ring *txr = &tx_que->txr;
3072 
3073 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(txr->me));
3074 		txdctl |= IXGBE_TXDCTL_ENABLE;
3075 		/* Set WTHRESH to 8, burst writeback */
3076 		txdctl |= (8 << 16);
3077 		/*
3078 		 * When the internal queue falls below PTHRESH (32),
3079 		 * start prefetching as long as there are at least
3080 		 * HTHRESH (1) buffers ready. The values are taken
3081 		 * from the Intel linux driver 3.8.21.
3082 		 * Prefetching enables tx line rate even with 1 queue.
3083 		 */
3084 		txdctl |= (32 << 0) | (1 << 8);
3085 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(txr->me), txdctl);
3086 	}
3087 
3088 	for (i = 0, rx_que = sc->rx_queues; i < sc->num_rx_queues; i++, rx_que++) {
3089 		struct rx_ring *rxr = &rx_que->rxr;
3090 
3091 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me));
3092 		if (hw->mac.type == ixgbe_mac_82598EB) {
3093 			/*
3094 			 * PTHRESH = 21
3095 			 * HTHRESH = 4
3096 			 * WTHRESH = 8
3097 			 */
3098 			rxdctl &= ~0x3FFFFF;
3099 			rxdctl |= 0x080420;
3100 		}
3101 		rxdctl |= IXGBE_RXDCTL_ENABLE;
3102 		IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), rxdctl);
3103 		for (j = 0; j < 10; j++) {
3104 			if (IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me)) &
3105 			    IXGBE_RXDCTL_ENABLE)
3106 				break;
3107 			else
3108 				msec_delay(1);
3109 		}
3110 		wmb();
3111 	}
3112 
3113 	/* Enable Receive engine */
3114 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3115 	if (hw->mac.type == ixgbe_mac_82598EB)
3116 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3117 	rxctrl |= IXGBE_RXCTRL_RXEN;
3118 	ixgbe_enable_rx_dma(hw, rxctrl);
3119 
3120 	/* Set up MSI/MSI-X routing */
3121 	if (ixgbe_enable_msix)  {
3122 		ixgbe_configure_ivars(sc);
3123 		/* Set up auto-mask */
3124 		if (hw->mac.type == ixgbe_mac_82598EB)
3125 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3126 		else {
3127 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
3128 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
3129 		}
3130 	} else {  /* Simple settings for Legacy/MSI */
3131 		ixgbe_set_ivar(sc, 0, 0, 0);
3132 		ixgbe_set_ivar(sc, 0, 0, 1);
3133 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3134 	}
3135 
3136 	ixgbe_init_fdir(sc);
3137 
3138 	/*
3139 	 * Check on any SFP devices that
3140 	 * need to be kick-started
3141 	 */
3142 	if (hw->phy.type == ixgbe_phy_none) {
3143 		err = hw->phy.ops.identify(hw);
3144 		if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3145 			device_printf(dev,
3146 			    "Unsupported SFP+ module type was detected.\n");
3147 			return;
3148 		}
3149 	}
3150 
3151 	/* Set moderation on the Link interrupt */
3152 	IXGBE_WRITE_REG(hw, IXGBE_EITR(sc->vector), IXGBE_LINK_ITR);
3153 
3154 	/* Enable power to the phy. */
3155 	ixgbe_set_phy_power(hw, true);
3156 
3157 	/* Config/Enable Link */
3158 	ixgbe_config_link(ctx);
3159 
3160 	/* Hardware Packet Buffer & Flow Control setup */
3161 	ixgbe_config_delay_values(sc);
3162 
3163 	/* Initialize the FC settings */
3164 	ixgbe_start_hw(hw);
3165 
3166 	/* Set up VLAN support and filter */
3167 	ixgbe_setup_vlan_hw_support(ctx);
3168 
3169 	/* Setup DMA Coalescing */
3170 	ixgbe_config_dmac(sc);
3171 
3172 	/* And now turn on interrupts */
3173 	ixgbe_if_enable_intr(ctx);
3174 
3175 	/* Enable the use of the MBX by the VF's */
3176 	if (sc->feat_en & IXGBE_FEATURE_SRIOV) {
3177 		ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
3178 		ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
3179 		IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
3180 	}
3181 
3182 } /* ixgbe_init_locked */
3183 
3184 /************************************************************************
3185  * ixgbe_set_ivar
3186  *
3187  *   Setup the correct IVAR register for a particular MSI-X interrupt
3188  *     (yes this is all very magic and confusing :)
3189  *    - entry is the register array entry
3190  *    - vector is the MSI-X vector for this queue
3191  *    - type is RX/TX/MISC
3192  ************************************************************************/
3193 static void
3194 ixgbe_set_ivar(struct ixgbe_softc *sc, u8 entry, u8 vector, s8 type)
3195 {
3196 	struct ixgbe_hw *hw = &sc->hw;
3197 	u32 ivar, index;
3198 
3199 	vector |= IXGBE_IVAR_ALLOC_VAL;
3200 
3201 	switch (hw->mac.type) {
3202 	case ixgbe_mac_82598EB:
3203 		if (type == -1)
3204 			entry = IXGBE_IVAR_OTHER_CAUSES_INDEX;
3205 		else
3206 			entry += (type * 64);
3207 		index = (entry >> 2) & 0x1F;
3208 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
3209 		ivar &= ~(0xFF << (8 * (entry & 0x3)));
3210 		ivar |= (vector << (8 * (entry & 0x3)));
3211 		IXGBE_WRITE_REG(&sc->hw, IXGBE_IVAR(index), ivar);
3212 		break;
3213 	case ixgbe_mac_82599EB:
3214 	case ixgbe_mac_X540:
3215 	case ixgbe_mac_X550:
3216 	case ixgbe_mac_X550EM_x:
3217 	case ixgbe_mac_X550EM_a:
3218 		if (type == -1) { /* MISC IVAR */
3219 			index = (entry & 1) * 8;
3220 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
3221 			ivar &= ~(0xFF << index);
3222 			ivar |= (vector << index);
3223 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, ivar);
3224 		} else {          /* RX/TX IVARS */
3225 			index = (16 * (entry & 1)) + (8 * type);
3226 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(entry >> 1));
3227 			ivar &= ~(0xFF << index);
3228 			ivar |= (vector << index);
3229 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(entry >> 1), ivar);
3230 		}
3231 	default:
3232 		break;
3233 	}
3234 } /* ixgbe_set_ivar */
3235 
3236 /************************************************************************
3237  * ixgbe_configure_ivars
3238  ************************************************************************/
3239 static void
3240 ixgbe_configure_ivars(struct ixgbe_softc *sc)
3241 {
3242 	struct ix_rx_queue *rx_que = sc->rx_queues;
3243 	struct ix_tx_queue *tx_que = sc->tx_queues;
3244 	u32                newitr;
3245 
3246 	if (ixgbe_max_interrupt_rate > 0)
3247 		newitr = (4000000 / ixgbe_max_interrupt_rate) & 0x0FF8;
3248 	else {
3249 		/*
3250 		 * Disable DMA coalescing if interrupt moderation is
3251 		 * disabled.
3252 		 */
3253 		sc->dmac = 0;
3254 		newitr = 0;
3255 	}
3256 
3257 	for (int i = 0; i < sc->num_rx_queues; i++, rx_que++) {
3258 		struct rx_ring *rxr = &rx_que->rxr;
3259 
3260 		/* First the RX queue entry */
3261 		ixgbe_set_ivar(sc, rxr->me, rx_que->msix, 0);
3262 
3263 		/* Set an Initial EITR value */
3264 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EITR(rx_que->msix), newitr);
3265 	}
3266 	for (int i = 0; i < sc->num_tx_queues; i++, tx_que++) {
3267 		struct tx_ring *txr = &tx_que->txr;
3268 
3269 		/* ... and the TX */
3270 		ixgbe_set_ivar(sc, txr->me, tx_que->msix, 1);
3271 	}
3272 	/* For the Link interrupt */
3273 	ixgbe_set_ivar(sc, 1, sc->vector, -1);
3274 } /* ixgbe_configure_ivars */
3275 
3276 /************************************************************************
3277  * ixgbe_config_gpie
3278  ************************************************************************/
3279 static void
3280 ixgbe_config_gpie(struct ixgbe_softc *sc)
3281 {
3282 	struct ixgbe_hw *hw = &sc->hw;
3283 	u32             gpie;
3284 
3285 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
3286 
3287 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3288 		/* Enable Enhanced MSI-X mode */
3289 		gpie |= IXGBE_GPIE_MSIX_MODE
3290 		     |  IXGBE_GPIE_EIAME
3291 		     |  IXGBE_GPIE_PBA_SUPPORT
3292 		     |  IXGBE_GPIE_OCD;
3293 	}
3294 
3295 	/* Fan Failure Interrupt */
3296 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL)
3297 		gpie |= IXGBE_SDP1_GPIEN;
3298 
3299 	/* Thermal Sensor Interrupt */
3300 	if (sc->feat_en & IXGBE_FEATURE_TEMP_SENSOR)
3301 		gpie |= IXGBE_SDP0_GPIEN_X540;
3302 
3303 	/* Link detection */
3304 	switch (hw->mac.type) {
3305 	case ixgbe_mac_82599EB:
3306 		gpie |= IXGBE_SDP1_GPIEN | IXGBE_SDP2_GPIEN;
3307 		break;
3308 	case ixgbe_mac_X550EM_x:
3309 	case ixgbe_mac_X550EM_a:
3310 		gpie |= IXGBE_SDP0_GPIEN_X540;
3311 		break;
3312 	default:
3313 		break;
3314 	}
3315 
3316 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
3317 
3318 } /* ixgbe_config_gpie */
3319 
3320 /************************************************************************
3321  * ixgbe_config_delay_values
3322  *
3323  *   Requires sc->max_frame_size to be set.
3324  ************************************************************************/
3325 static void
3326 ixgbe_config_delay_values(struct ixgbe_softc *sc)
3327 {
3328 	struct ixgbe_hw *hw = &sc->hw;
3329 	u32             rxpb, frame, size, tmp;
3330 
3331 	frame = sc->max_frame_size;
3332 
3333 	/* Calculate High Water */
3334 	switch (hw->mac.type) {
3335 	case ixgbe_mac_X540:
3336 	case ixgbe_mac_X550:
3337 	case ixgbe_mac_X550EM_x:
3338 	case ixgbe_mac_X550EM_a:
3339 		tmp = IXGBE_DV_X540(frame, frame);
3340 		break;
3341 	default:
3342 		tmp = IXGBE_DV(frame, frame);
3343 		break;
3344 	}
3345 	size = IXGBE_BT2KB(tmp);
3346 	rxpb = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0)) >> 10;
3347 	hw->fc.high_water[0] = rxpb - size;
3348 
3349 	/* Now calculate Low Water */
3350 	switch (hw->mac.type) {
3351 	case ixgbe_mac_X540:
3352 	case ixgbe_mac_X550:
3353 	case ixgbe_mac_X550EM_x:
3354 	case ixgbe_mac_X550EM_a:
3355 		tmp = IXGBE_LOW_DV_X540(frame);
3356 		break;
3357 	default:
3358 		tmp = IXGBE_LOW_DV(frame);
3359 		break;
3360 	}
3361 	hw->fc.low_water[0] = IXGBE_BT2KB(tmp);
3362 
3363 	hw->fc.pause_time = IXGBE_FC_PAUSE;
3364 	hw->fc.send_xon = true;
3365 } /* ixgbe_config_delay_values */
3366 
3367 /************************************************************************
3368  * ixgbe_set_multi - Multicast Update
3369  *
3370  *   Called whenever multicast address list is updated.
3371  ************************************************************************/
3372 static u_int
3373 ixgbe_mc_filter_apply(void *arg, struct sockaddr_dl *sdl, u_int idx)
3374 {
3375 	struct ixgbe_softc *sc = arg;
3376 	struct ixgbe_mc_addr *mta = sc->mta;
3377 
3378 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
3379 		return (0);
3380 	bcopy(LLADDR(sdl), mta[idx].addr, IXGBE_ETH_LENGTH_OF_ADDRESS);
3381 	mta[idx].vmdq = sc->pool;
3382 
3383 	return (1);
3384 } /* ixgbe_mc_filter_apply */
3385 
3386 static void
3387 ixgbe_if_multi_set(if_ctx_t ctx)
3388 {
3389 	struct ixgbe_softc       *sc = iflib_get_softc(ctx);
3390 	struct ixgbe_mc_addr *mta;
3391 	struct ifnet         *ifp = iflib_get_ifp(ctx);
3392 	u8                   *update_ptr;
3393 	u32                  fctrl;
3394 	u_int		     mcnt;
3395 
3396 	IOCTL_DEBUGOUT("ixgbe_if_multi_set: begin");
3397 
3398 	mta = sc->mta;
3399 	bzero(mta, sizeof(*mta) * MAX_NUM_MULTICAST_ADDRESSES);
3400 
3401 	mcnt = if_foreach_llmaddr(iflib_get_ifp(ctx), ixgbe_mc_filter_apply, sc);
3402 
3403 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) {
3404 		update_ptr = (u8 *)mta;
3405 		ixgbe_update_mc_addr_list(&sc->hw, update_ptr, mcnt,
3406 		    ixgbe_mc_array_itr, true);
3407 	}
3408 
3409 	fctrl = IXGBE_READ_REG(&sc->hw, IXGBE_FCTRL);
3410 
3411 	if (ifp->if_flags & IFF_PROMISC)
3412 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3413 	else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
3414 	    ifp->if_flags & IFF_ALLMULTI) {
3415 		fctrl |= IXGBE_FCTRL_MPE;
3416 		fctrl &= ~IXGBE_FCTRL_UPE;
3417 	} else
3418 		fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3419 
3420 	IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, fctrl);
3421 } /* ixgbe_if_multi_set */
3422 
3423 /************************************************************************
3424  * ixgbe_mc_array_itr
3425  *
3426  *   An iterator function needed by the multicast shared code.
3427  *   It feeds the shared code routine the addresses in the
3428  *   array of ixgbe_set_multi() one by one.
3429  ************************************************************************/
3430 static u8 *
3431 ixgbe_mc_array_itr(struct ixgbe_hw *hw, u8 **update_ptr, u32 *vmdq)
3432 {
3433 	struct ixgbe_mc_addr *mta;
3434 
3435 	mta = (struct ixgbe_mc_addr *)*update_ptr;
3436 	*vmdq = mta->vmdq;
3437 
3438 	*update_ptr = (u8*)(mta + 1);
3439 
3440 	return (mta->addr);
3441 } /* ixgbe_mc_array_itr */
3442 
3443 /************************************************************************
3444  * ixgbe_local_timer - Timer routine
3445  *
3446  *   Checks for link status, updates statistics,
3447  *   and runs the watchdog check.
3448  ************************************************************************/
3449 static void
3450 ixgbe_if_timer(if_ctx_t ctx, uint16_t qid)
3451 {
3452 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3453 
3454 	if (qid != 0)
3455 		return;
3456 
3457 	/* Check for pluggable optics */
3458 	if (sc->sfp_probe)
3459 		if (!ixgbe_sfp_probe(ctx))
3460 			return; /* Nothing to do */
3461 
3462 	ixgbe_check_link(&sc->hw, &sc->link_speed, &sc->link_up, 0);
3463 
3464 	/* Fire off the adminq task */
3465 	iflib_admin_intr_deferred(ctx);
3466 
3467 } /* ixgbe_if_timer */
3468 
3469 /************************************************************************
3470  * ixgbe_sfp_probe
3471  *
3472  *   Determine if a port had optics inserted.
3473  ************************************************************************/
3474 static bool
3475 ixgbe_sfp_probe(if_ctx_t ctx)
3476 {
3477 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3478 	struct ixgbe_hw *hw = &sc->hw;
3479 	device_t        dev = iflib_get_dev(ctx);
3480 	bool            result = false;
3481 
3482 	if ((hw->phy.type == ixgbe_phy_nl) &&
3483 	    (hw->phy.sfp_type == ixgbe_sfp_type_not_present)) {
3484 		s32 ret = hw->phy.ops.identify_sfp(hw);
3485 		if (ret)
3486 			goto out;
3487 		ret = hw->phy.ops.reset(hw);
3488 		sc->sfp_probe = false;
3489 		if (ret == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3490 			device_printf(dev, "Unsupported SFP+ module detected!");
3491 			device_printf(dev,
3492 			    "Reload driver with supported module.\n");
3493 			goto out;
3494 		} else
3495 			device_printf(dev, "SFP+ module detected!\n");
3496 		/* We now have supported optics */
3497 		result = true;
3498 	}
3499 out:
3500 
3501 	return (result);
3502 } /* ixgbe_sfp_probe */
3503 
3504 /************************************************************************
3505  * ixgbe_handle_mod - Tasklet for SFP module interrupts
3506  ************************************************************************/
3507 static void
3508 ixgbe_handle_mod(void *context)
3509 {
3510 	if_ctx_t        ctx = context;
3511 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3512 	struct ixgbe_hw *hw = &sc->hw;
3513 	device_t        dev = iflib_get_dev(ctx);
3514 	u32             err, cage_full = 0;
3515 
3516 	if (sc->hw.need_crosstalk_fix) {
3517 		switch (hw->mac.type) {
3518 		case ixgbe_mac_82599EB:
3519 			cage_full = IXGBE_READ_REG(hw, IXGBE_ESDP) &
3520 			    IXGBE_ESDP_SDP2;
3521 			break;
3522 		case ixgbe_mac_X550EM_x:
3523 		case ixgbe_mac_X550EM_a:
3524 			cage_full = IXGBE_READ_REG(hw, IXGBE_ESDP) &
3525 			    IXGBE_ESDP_SDP0;
3526 			break;
3527 		default:
3528 			break;
3529 		}
3530 
3531 		if (!cage_full)
3532 			goto handle_mod_out;
3533 	}
3534 
3535 	err = hw->phy.ops.identify_sfp(hw);
3536 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3537 		device_printf(dev,
3538 		    "Unsupported SFP+ module type was detected.\n");
3539 		goto handle_mod_out;
3540 	}
3541 
3542 	if (hw->mac.type == ixgbe_mac_82598EB)
3543 		err = hw->phy.ops.reset(hw);
3544 	else
3545 		err = hw->mac.ops.setup_sfp(hw);
3546 
3547 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3548 		device_printf(dev,
3549 		    "Setup failure - unsupported SFP+ module type.\n");
3550 		goto handle_mod_out;
3551 	}
3552 	sc->task_requests |= IXGBE_REQUEST_TASK_MSF;
3553 	return;
3554 
3555 handle_mod_out:
3556 	sc->task_requests &= ~(IXGBE_REQUEST_TASK_MSF);
3557 } /* ixgbe_handle_mod */
3558 
3559 
3560 /************************************************************************
3561  * ixgbe_handle_msf - Tasklet for MSF (multispeed fiber) interrupts
3562  ************************************************************************/
3563 static void
3564 ixgbe_handle_msf(void *context)
3565 {
3566 	if_ctx_t        ctx = context;
3567 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3568 	struct ixgbe_hw *hw = &sc->hw;
3569 	u32             autoneg;
3570 	bool            negotiate;
3571 
3572 	/* get_supported_phy_layer will call hw->phy.ops.identify_sfp() */
3573 	sc->phy_layer = ixgbe_get_supported_physical_layer(hw);
3574 
3575 	autoneg = hw->phy.autoneg_advertised;
3576 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
3577 		hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiate);
3578 	if (hw->mac.ops.setup_link)
3579 		hw->mac.ops.setup_link(hw, autoneg, true);
3580 
3581 	/* Adjust media types shown in ifconfig */
3582 	ifmedia_removeall(sc->media);
3583 	ixgbe_add_media_types(sc->ctx);
3584 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
3585 } /* ixgbe_handle_msf */
3586 
3587 /************************************************************************
3588  * ixgbe_handle_phy - Tasklet for external PHY interrupts
3589  ************************************************************************/
3590 static void
3591 ixgbe_handle_phy(void *context)
3592 {
3593 	if_ctx_t        ctx = context;
3594 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3595 	struct ixgbe_hw *hw = &sc->hw;
3596 	int             error;
3597 
3598 	error = hw->phy.ops.handle_lasi(hw);
3599 	if (error == IXGBE_ERR_OVERTEMP)
3600 		device_printf(sc->dev, "CRITICAL: EXTERNAL PHY OVER TEMP!!  PHY will downshift to lower power state!\n");
3601 	else if (error)
3602 		device_printf(sc->dev,
3603 		    "Error handling LASI interrupt: %d\n", error);
3604 } /* ixgbe_handle_phy */
3605 
3606 /************************************************************************
3607  * ixgbe_if_stop - Stop the hardware
3608  *
3609  *   Disables all traffic on the adapter by issuing a
3610  *   global reset on the MAC and deallocates TX/RX buffers.
3611  ************************************************************************/
3612 static void
3613 ixgbe_if_stop(if_ctx_t ctx)
3614 {
3615 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3616 	struct ixgbe_hw *hw = &sc->hw;
3617 
3618 	INIT_DEBUGOUT("ixgbe_if_stop: begin\n");
3619 
3620 	ixgbe_reset_hw(hw);
3621 	hw->adapter_stopped = false;
3622 	ixgbe_stop_adapter(hw);
3623 	if (hw->mac.type == ixgbe_mac_82599EB)
3624 		ixgbe_stop_mac_link_on_d3_82599(hw);
3625 	/* Turn off the laser - noop with no optics */
3626 	ixgbe_disable_tx_laser(hw);
3627 
3628 	/* Update the stack */
3629 	sc->link_up = false;
3630 	ixgbe_if_update_admin_status(ctx);
3631 
3632 	/* reprogram the RAR[0] in case user changed it. */
3633 	ixgbe_set_rar(&sc->hw, 0, sc->hw.mac.addr, 0, IXGBE_RAH_AV);
3634 
3635 	return;
3636 } /* ixgbe_if_stop */
3637 
3638 /************************************************************************
3639  * ixgbe_update_link_status - Update OS on link state
3640  *
3641  * Note: Only updates the OS on the cached link state.
3642  *       The real check of the hardware only happens with
3643  *       a link interrupt.
3644  ************************************************************************/
3645 static void
3646 ixgbe_if_update_admin_status(if_ctx_t ctx)
3647 {
3648 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3649 	device_t       dev = iflib_get_dev(ctx);
3650 
3651 	if (sc->link_up) {
3652 		if (sc->link_active == false) {
3653 			if (bootverbose)
3654 				device_printf(dev, "Link is up %d Gbps %s \n",
3655 				    ((sc->link_speed == 128) ? 10 : 1),
3656 				    "Full Duplex");
3657 			sc->link_active = true;
3658 			/* Update any Flow Control changes */
3659 			ixgbe_fc_enable(&sc->hw);
3660 			/* Update DMA coalescing config */
3661 			ixgbe_config_dmac(sc);
3662 			/* should actually be negotiated value */
3663 			iflib_link_state_change(ctx, LINK_STATE_UP, IF_Gbps(10));
3664 
3665 			if (sc->feat_en & IXGBE_FEATURE_SRIOV)
3666 				ixgbe_ping_all_vfs(sc);
3667 		}
3668 	} else { /* Link down */
3669 		if (sc->link_active == true) {
3670 			if (bootverbose)
3671 				device_printf(dev, "Link is Down\n");
3672 			iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
3673 			sc->link_active = false;
3674 			if (sc->feat_en & IXGBE_FEATURE_SRIOV)
3675 				ixgbe_ping_all_vfs(sc);
3676 		}
3677 	}
3678 
3679 	/* Handle task requests from msix_link() */
3680 	if (sc->task_requests & IXGBE_REQUEST_TASK_MOD)
3681 		ixgbe_handle_mod(ctx);
3682 	if (sc->task_requests & IXGBE_REQUEST_TASK_MSF)
3683 		ixgbe_handle_msf(ctx);
3684 	if (sc->task_requests & IXGBE_REQUEST_TASK_MBX)
3685 		ixgbe_handle_mbx(ctx);
3686 	if (sc->task_requests & IXGBE_REQUEST_TASK_FDIR)
3687 		ixgbe_reinit_fdir(ctx);
3688 	if (sc->task_requests & IXGBE_REQUEST_TASK_PHY)
3689 		ixgbe_handle_phy(ctx);
3690 	sc->task_requests = 0;
3691 
3692 	ixgbe_update_stats_counters(sc);
3693 } /* ixgbe_if_update_admin_status */
3694 
3695 /************************************************************************
3696  * ixgbe_config_dmac - Configure DMA Coalescing
3697  ************************************************************************/
3698 static void
3699 ixgbe_config_dmac(struct ixgbe_softc *sc)
3700 {
3701 	struct ixgbe_hw          *hw = &sc->hw;
3702 	struct ixgbe_dmac_config *dcfg = &hw->mac.dmac_config;
3703 
3704 	if (hw->mac.type < ixgbe_mac_X550 || !hw->mac.ops.dmac_config)
3705 		return;
3706 
3707 	if (dcfg->watchdog_timer ^ sc->dmac ||
3708 	    dcfg->link_speed ^ sc->link_speed) {
3709 		dcfg->watchdog_timer = sc->dmac;
3710 		dcfg->fcoe_en = false;
3711 		dcfg->link_speed = sc->link_speed;
3712 		dcfg->num_tcs = 1;
3713 
3714 		INIT_DEBUGOUT2("dmac settings: watchdog %d, link speed %d\n",
3715 		    dcfg->watchdog_timer, dcfg->link_speed);
3716 
3717 		hw->mac.ops.dmac_config(hw);
3718 	}
3719 } /* ixgbe_config_dmac */
3720 
3721 /************************************************************************
3722  * ixgbe_if_enable_intr
3723  ************************************************************************/
3724 void
3725 ixgbe_if_enable_intr(if_ctx_t ctx)
3726 {
3727 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
3728 	struct ixgbe_hw    *hw = &sc->hw;
3729 	struct ix_rx_queue *que = sc->rx_queues;
3730 	u32                mask, fwsm;
3731 
3732 	mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
3733 
3734 	switch (sc->hw.mac.type) {
3735 	case ixgbe_mac_82599EB:
3736 		mask |= IXGBE_EIMS_ECC;
3737 		/* Temperature sensor on some scs */
3738 		mask |= IXGBE_EIMS_GPI_SDP0;
3739 		/* SFP+ (RX_LOS_N & MOD_ABS_N) */
3740 		mask |= IXGBE_EIMS_GPI_SDP1;
3741 		mask |= IXGBE_EIMS_GPI_SDP2;
3742 		break;
3743 	case ixgbe_mac_X540:
3744 		/* Detect if Thermal Sensor is enabled */
3745 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
3746 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
3747 			mask |= IXGBE_EIMS_TS;
3748 		mask |= IXGBE_EIMS_ECC;
3749 		break;
3750 	case ixgbe_mac_X550:
3751 		/* MAC thermal sensor is automatically enabled */
3752 		mask |= IXGBE_EIMS_TS;
3753 		mask |= IXGBE_EIMS_ECC;
3754 		break;
3755 	case ixgbe_mac_X550EM_x:
3756 	case ixgbe_mac_X550EM_a:
3757 		/* Some devices use SDP0 for important information */
3758 		if (hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3759 		    hw->device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3760 		    hw->device_id == IXGBE_DEV_ID_X550EM_A_SFP_N ||
3761 		    hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T)
3762 			mask |= IXGBE_EIMS_GPI_SDP0_BY_MAC(hw);
3763 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
3764 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3765 		mask |= IXGBE_EIMS_ECC;
3766 		break;
3767 	default:
3768 		break;
3769 	}
3770 
3771 	/* Enable Fan Failure detection */
3772 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL)
3773 		mask |= IXGBE_EIMS_GPI_SDP1;
3774 	/* Enable SR-IOV */
3775 	if (sc->feat_en & IXGBE_FEATURE_SRIOV)
3776 		mask |= IXGBE_EIMS_MAILBOX;
3777 	/* Enable Flow Director */
3778 	if (sc->feat_en & IXGBE_FEATURE_FDIR)
3779 		mask |= IXGBE_EIMS_FLOW_DIR;
3780 
3781 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
3782 
3783 	/* With MSI-X we use auto clear */
3784 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3785 		mask = IXGBE_EIMS_ENABLE_MASK;
3786 		/* Don't autoclear Link */
3787 		mask &= ~IXGBE_EIMS_OTHER;
3788 		mask &= ~IXGBE_EIMS_LSC;
3789 		if (sc->feat_cap & IXGBE_FEATURE_SRIOV)
3790 			mask &= ~IXGBE_EIMS_MAILBOX;
3791 		IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
3792 	}
3793 
3794 	/*
3795 	 * Now enable all queues, this is done separately to
3796 	 * allow for handling the extended (beyond 32) MSI-X
3797 	 * vectors that can be used by 82599
3798 	 */
3799 	for (int i = 0; i < sc->num_rx_queues; i++, que++)
3800 		ixgbe_enable_queue(sc, que->msix);
3801 
3802 	IXGBE_WRITE_FLUSH(hw);
3803 
3804 } /* ixgbe_if_enable_intr */
3805 
3806 /************************************************************************
3807  * ixgbe_disable_intr
3808  ************************************************************************/
3809 static void
3810 ixgbe_if_disable_intr(if_ctx_t ctx)
3811 {
3812 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3813 
3814 	if (sc->intr_type == IFLIB_INTR_MSIX)
3815 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIAC, 0);
3816 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
3817 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC, ~0);
3818 	} else {
3819 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC, 0xFFFF0000);
3820 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC_EX(0), ~0);
3821 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC_EX(1), ~0);
3822 	}
3823 	IXGBE_WRITE_FLUSH(&sc->hw);
3824 
3825 } /* ixgbe_if_disable_intr */
3826 
3827 /************************************************************************
3828  * ixgbe_link_intr_enable
3829  ************************************************************************/
3830 static void
3831 ixgbe_link_intr_enable(if_ctx_t ctx)
3832 {
3833 	struct ixgbe_hw *hw = &((struct ixgbe_softc *)iflib_get_softc(ctx))->hw;
3834 
3835 	/* Re-enable other interrupts */
3836 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_OTHER | IXGBE_EIMS_LSC);
3837 } /* ixgbe_link_intr_enable */
3838 
3839 /************************************************************************
3840  * ixgbe_if_rx_queue_intr_enable
3841  ************************************************************************/
3842 static int
3843 ixgbe_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
3844 {
3845 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
3846 	struct ix_rx_queue *que = &sc->rx_queues[rxqid];
3847 
3848 	ixgbe_enable_queue(sc, que->msix);
3849 
3850 	return (0);
3851 } /* ixgbe_if_rx_queue_intr_enable */
3852 
3853 /************************************************************************
3854  * ixgbe_enable_queue
3855  ************************************************************************/
3856 static void
3857 ixgbe_enable_queue(struct ixgbe_softc *sc, u32 vector)
3858 {
3859 	struct ixgbe_hw *hw = &sc->hw;
3860 	u64             queue = 1ULL << vector;
3861 	u32             mask;
3862 
3863 	if (hw->mac.type == ixgbe_mac_82598EB) {
3864 		mask = (IXGBE_EIMS_RTX_QUEUE & queue);
3865 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
3866 	} else {
3867 		mask = (queue & 0xFFFFFFFF);
3868 		if (mask)
3869 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
3870 		mask = (queue >> 32);
3871 		if (mask)
3872 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
3873 	}
3874 } /* ixgbe_enable_queue */
3875 
3876 /************************************************************************
3877  * ixgbe_disable_queue
3878  ************************************************************************/
3879 static void
3880 ixgbe_disable_queue(struct ixgbe_softc *sc, u32 vector)
3881 {
3882 	struct ixgbe_hw *hw = &sc->hw;
3883 	u64             queue = 1ULL << vector;
3884 	u32             mask;
3885 
3886 	if (hw->mac.type == ixgbe_mac_82598EB) {
3887 		mask = (IXGBE_EIMS_RTX_QUEUE & queue);
3888 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
3889 	} else {
3890 		mask = (queue & 0xFFFFFFFF);
3891 		if (mask)
3892 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
3893 		mask = (queue >> 32);
3894 		if (mask)
3895 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
3896 	}
3897 } /* ixgbe_disable_queue */
3898 
3899 /************************************************************************
3900  * ixgbe_intr - Legacy Interrupt Service Routine
3901  ************************************************************************/
3902 int
3903 ixgbe_intr(void *arg)
3904 {
3905 	struct ixgbe_softc     *sc = arg;
3906 	struct ix_rx_queue *que = sc->rx_queues;
3907 	struct ixgbe_hw    *hw = &sc->hw;
3908 	if_ctx_t           ctx = sc->ctx;
3909 	u32                eicr, eicr_mask;
3910 
3911 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3912 
3913 	++que->irqs;
3914 	if (eicr == 0) {
3915 		ixgbe_if_enable_intr(ctx);
3916 		return (FILTER_HANDLED);
3917 	}
3918 
3919 	/* Check for fan failure */
3920 	if ((hw->device_id == IXGBE_DEV_ID_82598AT) &&
3921 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
3922 		device_printf(sc->dev,
3923 		    "\nCRITICAL: FAN FAILURE!! REPLACE IMMEDIATELY!!\n");
3924 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
3925 	}
3926 
3927 	/* Link status change */
3928 	if (eicr & IXGBE_EICR_LSC) {
3929 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
3930 		iflib_admin_intr_deferred(ctx);
3931 	}
3932 
3933 	if (ixgbe_is_sfp(hw)) {
3934 		/* Pluggable optics-related interrupt */
3935 		if (hw->mac.type >= ixgbe_mac_X540)
3936 			eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
3937 		else
3938 			eicr_mask = IXGBE_EICR_GPI_SDP2_BY_MAC(hw);
3939 
3940 		if (eicr & eicr_mask) {
3941 			IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
3942 			sc->task_requests |= IXGBE_REQUEST_TASK_MOD;
3943 		}
3944 
3945 		if ((hw->mac.type == ixgbe_mac_82599EB) &&
3946 		    (eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw))) {
3947 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3948 			    IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
3949 			sc->task_requests |= IXGBE_REQUEST_TASK_MSF;
3950 		}
3951 	}
3952 
3953 	/* External PHY interrupt */
3954 	if ((hw->phy.type == ixgbe_phy_x550em_ext_t) &&
3955 	    (eicr & IXGBE_EICR_GPI_SDP0_X540))
3956 		sc->task_requests |= IXGBE_REQUEST_TASK_PHY;
3957 
3958 	return (FILTER_SCHEDULE_THREAD);
3959 } /* ixgbe_intr */
3960 
3961 /************************************************************************
3962  * ixgbe_free_pci_resources
3963  ************************************************************************/
3964 static void
3965 ixgbe_free_pci_resources(if_ctx_t ctx)
3966 {
3967 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3968 	struct         ix_rx_queue *que = sc->rx_queues;
3969 	device_t       dev = iflib_get_dev(ctx);
3970 
3971 	/* Release all MSI-X queue resources */
3972 	if (sc->intr_type == IFLIB_INTR_MSIX)
3973 		iflib_irq_free(ctx, &sc->irq);
3974 
3975 	if (que != NULL) {
3976 		for (int i = 0; i < sc->num_rx_queues; i++, que++) {
3977 			iflib_irq_free(ctx, &que->que_irq);
3978 		}
3979 	}
3980 
3981 	if (sc->pci_mem != NULL)
3982 		bus_release_resource(dev, SYS_RES_MEMORY,
3983 		    rman_get_rid(sc->pci_mem), sc->pci_mem);
3984 } /* ixgbe_free_pci_resources */
3985 
3986 /************************************************************************
3987  * ixgbe_sysctl_flowcntl
3988  *
3989  *   SYSCTL wrapper around setting Flow Control
3990  ************************************************************************/
3991 static int
3992 ixgbe_sysctl_flowcntl(SYSCTL_HANDLER_ARGS)
3993 {
3994 	struct ixgbe_softc *sc;
3995 	int            error, fc;
3996 
3997 	sc = (struct ixgbe_softc *)arg1;
3998 	fc = sc->hw.fc.current_mode;
3999 
4000 	error = sysctl_handle_int(oidp, &fc, 0, req);
4001 	if ((error) || (req->newptr == NULL))
4002 		return (error);
4003 
4004 	/* Don't bother if it's not changed */
4005 	if (fc == sc->hw.fc.current_mode)
4006 		return (0);
4007 
4008 	return ixgbe_set_flowcntl(sc, fc);
4009 } /* ixgbe_sysctl_flowcntl */
4010 
4011 /************************************************************************
4012  * ixgbe_set_flowcntl - Set flow control
4013  *
4014  *   Flow control values:
4015  *     0 - off
4016  *     1 - rx pause
4017  *     2 - tx pause
4018  *     3 - full
4019  ************************************************************************/
4020 static int
4021 ixgbe_set_flowcntl(struct ixgbe_softc *sc, int fc)
4022 {
4023 	switch (fc) {
4024 	case ixgbe_fc_rx_pause:
4025 	case ixgbe_fc_tx_pause:
4026 	case ixgbe_fc_full:
4027 		sc->hw.fc.requested_mode = fc;
4028 		if (sc->num_rx_queues > 1)
4029 			ixgbe_disable_rx_drop(sc);
4030 		break;
4031 	case ixgbe_fc_none:
4032 		sc->hw.fc.requested_mode = ixgbe_fc_none;
4033 		if (sc->num_rx_queues > 1)
4034 			ixgbe_enable_rx_drop(sc);
4035 		break;
4036 	default:
4037 		return (EINVAL);
4038 	}
4039 
4040 	/* Don't autoneg if forcing a value */
4041 	sc->hw.fc.disable_fc_autoneg = true;
4042 	ixgbe_fc_enable(&sc->hw);
4043 
4044 	return (0);
4045 } /* ixgbe_set_flowcntl */
4046 
4047 /************************************************************************
4048  * ixgbe_enable_rx_drop
4049  *
4050  *   Enable the hardware to drop packets when the buffer is
4051  *   full. This is useful with multiqueue, so that no single
4052  *   queue being full stalls the entire RX engine. We only
4053  *   enable this when Multiqueue is enabled AND Flow Control
4054  *   is disabled.
4055  ************************************************************************/
4056 static void
4057 ixgbe_enable_rx_drop(struct ixgbe_softc *sc)
4058 {
4059 	struct ixgbe_hw *hw = &sc->hw;
4060 	struct rx_ring  *rxr;
4061 	u32             srrctl;
4062 
4063 	for (int i = 0; i < sc->num_rx_queues; i++) {
4064 		rxr = &sc->rx_queues[i].rxr;
4065 		srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me));
4066 		srrctl |= IXGBE_SRRCTL_DROP_EN;
4067 		IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl);
4068 	}
4069 
4070 	/* enable drop for each vf */
4071 	for (int i = 0; i < sc->num_vfs; i++) {
4072 		IXGBE_WRITE_REG(hw, IXGBE_QDE,
4073 		                (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT) |
4074 		                IXGBE_QDE_ENABLE));
4075 	}
4076 } /* ixgbe_enable_rx_drop */
4077 
4078 /************************************************************************
4079  * ixgbe_disable_rx_drop
4080  ************************************************************************/
4081 static void
4082 ixgbe_disable_rx_drop(struct ixgbe_softc *sc)
4083 {
4084 	struct ixgbe_hw *hw = &sc->hw;
4085 	struct rx_ring  *rxr;
4086 	u32             srrctl;
4087 
4088 	for (int i = 0; i < sc->num_rx_queues; i++) {
4089 		rxr = &sc->rx_queues[i].rxr;
4090 		srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me));
4091 		srrctl &= ~IXGBE_SRRCTL_DROP_EN;
4092 		IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl);
4093 	}
4094 
4095 	/* disable drop for each vf */
4096 	for (int i = 0; i < sc->num_vfs; i++) {
4097 		IXGBE_WRITE_REG(hw, IXGBE_QDE,
4098 		    (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT)));
4099 	}
4100 } /* ixgbe_disable_rx_drop */
4101 
4102 /************************************************************************
4103  * ixgbe_sysctl_advertise
4104  *
4105  *   SYSCTL wrapper around setting advertised speed
4106  ************************************************************************/
4107 static int
4108 ixgbe_sysctl_advertise(SYSCTL_HANDLER_ARGS)
4109 {
4110 	struct ixgbe_softc *sc;
4111 	int            error, advertise;
4112 
4113 	sc = (struct ixgbe_softc *)arg1;
4114 	advertise = sc->advertise;
4115 
4116 	error = sysctl_handle_int(oidp, &advertise, 0, req);
4117 	if ((error) || (req->newptr == NULL))
4118 		return (error);
4119 
4120 	return ixgbe_set_advertise(sc, advertise);
4121 } /* ixgbe_sysctl_advertise */
4122 
4123 /************************************************************************
4124  * ixgbe_set_advertise - Control advertised link speed
4125  *
4126  *   Flags:
4127  *     0x1  - advertise 100 Mb
4128  *     0x2  - advertise 1G
4129  *     0x4  - advertise 10G
4130  *     0x8  - advertise 10 Mb (yes, Mb)
4131  *     0x10 - advertise 2.5G (disabled by default)
4132  *     0x20 - advertise 5G (disabled by default)
4133  *
4134  ************************************************************************/
4135 static int
4136 ixgbe_set_advertise(struct ixgbe_softc *sc, int advertise)
4137 {
4138 	device_t         dev = iflib_get_dev(sc->ctx);
4139 	struct ixgbe_hw  *hw;
4140 	ixgbe_link_speed speed = 0;
4141 	ixgbe_link_speed link_caps = 0;
4142 	s32              err = IXGBE_NOT_IMPLEMENTED;
4143 	bool             negotiate = false;
4144 
4145 	/* Checks to validate new value */
4146 	if (sc->advertise == advertise) /* no change */
4147 		return (0);
4148 
4149 	hw = &sc->hw;
4150 
4151 	/* No speed changes for backplane media */
4152 	if (hw->phy.media_type == ixgbe_media_type_backplane)
4153 		return (ENODEV);
4154 
4155 	if (!((hw->phy.media_type == ixgbe_media_type_copper) ||
4156 	      (hw->phy.multispeed_fiber))) {
4157 		device_printf(dev, "Advertised speed can only be set on copper or multispeed fiber media types.\n");
4158 		return (EINVAL);
4159 	}
4160 
4161 	if (advertise < 0x1 || advertise > 0x3F) {
4162 		device_printf(dev, "Invalid advertised speed; valid modes are 0x1 through 0x3F\n");
4163 		return (EINVAL);
4164 	}
4165 
4166 	if (hw->mac.ops.get_link_capabilities) {
4167 		err = hw->mac.ops.get_link_capabilities(hw, &link_caps,
4168 		    &negotiate);
4169 		if (err != IXGBE_SUCCESS) {
4170 			device_printf(dev, "Unable to determine supported advertise speeds\n");
4171 			return (ENODEV);
4172 		}
4173 	}
4174 
4175 	/* Set new value and report new advertised mode */
4176 	if (advertise & 0x1) {
4177 		if (!(link_caps & IXGBE_LINK_SPEED_100_FULL)) {
4178 			device_printf(dev, "Interface does not support 100Mb advertised speed\n");
4179 			return (EINVAL);
4180 		}
4181 		speed |= IXGBE_LINK_SPEED_100_FULL;
4182 	}
4183 	if (advertise & 0x2) {
4184 		if (!(link_caps & IXGBE_LINK_SPEED_1GB_FULL)) {
4185 			device_printf(dev, "Interface does not support 1Gb advertised speed\n");
4186 			return (EINVAL);
4187 		}
4188 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
4189 	}
4190 	if (advertise & 0x4) {
4191 		if (!(link_caps & IXGBE_LINK_SPEED_10GB_FULL)) {
4192 			device_printf(dev, "Interface does not support 10Gb advertised speed\n");
4193 			return (EINVAL);
4194 		}
4195 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
4196 	}
4197 	if (advertise & 0x8) {
4198 		if (!(link_caps & IXGBE_LINK_SPEED_10_FULL)) {
4199 			device_printf(dev, "Interface does not support 10Mb advertised speed\n");
4200 			return (EINVAL);
4201 		}
4202 		speed |= IXGBE_LINK_SPEED_10_FULL;
4203 	}
4204 	if (advertise & 0x10) {
4205 		if (!(link_caps & IXGBE_LINK_SPEED_2_5GB_FULL)) {
4206 			device_printf(dev, "Interface does not support 2.5G advertised speed\n");
4207 			return (EINVAL);
4208 		}
4209 		speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
4210 	}
4211 	if (advertise & 0x20) {
4212 		if (!(link_caps & IXGBE_LINK_SPEED_5GB_FULL)) {
4213 			device_printf(dev, "Interface does not support 5G advertised speed\n");
4214 			return (EINVAL);
4215 		}
4216 		speed |= IXGBE_LINK_SPEED_5GB_FULL;
4217 	}
4218 
4219 	hw->mac.autotry_restart = true;
4220 	hw->mac.ops.setup_link(hw, speed, true);
4221 	sc->advertise = advertise;
4222 
4223 	return (0);
4224 } /* ixgbe_set_advertise */
4225 
4226 /************************************************************************
4227  * ixgbe_get_default_advertise - Get default advertised speed settings
4228  *
4229  *   Formatted for sysctl usage.
4230  *   Flags:
4231  *     0x1 - advertise 100 Mb
4232  *     0x2 - advertise 1G
4233  *     0x4 - advertise 10G
4234  *     0x8 - advertise 10 Mb (yes, Mb)
4235  *     0x10 - advertise 2.5G (disabled by default)
4236  *     0x20 - advertise 5G (disabled by default)
4237  ************************************************************************/
4238 static int
4239 ixgbe_get_default_advertise(struct ixgbe_softc *sc)
4240 {
4241 	struct ixgbe_hw  *hw = &sc->hw;
4242 	int              speed;
4243 	ixgbe_link_speed link_caps = 0;
4244 	s32              err;
4245 	bool             negotiate = false;
4246 
4247 	/*
4248 	 * Advertised speed means nothing unless it's copper or
4249 	 * multi-speed fiber
4250 	 */
4251 	if (!(hw->phy.media_type == ixgbe_media_type_copper) &&
4252 	    !(hw->phy.multispeed_fiber))
4253 		return (0);
4254 
4255 	err = hw->mac.ops.get_link_capabilities(hw, &link_caps, &negotiate);
4256 	if (err != IXGBE_SUCCESS)
4257 		return (0);
4258 
4259 	if (hw->mac.type == ixgbe_mac_X550) {
4260 		/*
4261 		 * 2.5G and 5G autonegotiation speeds on X550
4262 		 * are disabled by default due to reported
4263 		 * interoperability issues with some switches.
4264 		 */
4265 		link_caps &= ~(IXGBE_LINK_SPEED_2_5GB_FULL |
4266 		    IXGBE_LINK_SPEED_5GB_FULL);
4267 	}
4268 
4269 	speed =
4270 	    ((link_caps & IXGBE_LINK_SPEED_10GB_FULL)  ? 0x4  : 0) |
4271 	    ((link_caps & IXGBE_LINK_SPEED_5GB_FULL)   ? 0x20 : 0) |
4272 	    ((link_caps & IXGBE_LINK_SPEED_2_5GB_FULL) ? 0x10 : 0) |
4273 	    ((link_caps & IXGBE_LINK_SPEED_1GB_FULL)   ? 0x2  : 0) |
4274 	    ((link_caps & IXGBE_LINK_SPEED_100_FULL)   ? 0x1  : 0) |
4275 	    ((link_caps & IXGBE_LINK_SPEED_10_FULL)    ? 0x8  : 0);
4276 
4277 	return speed;
4278 } /* ixgbe_get_default_advertise */
4279 
4280 /************************************************************************
4281  * ixgbe_sysctl_dmac - Manage DMA Coalescing
4282  *
4283  *   Control values:
4284  *     0/1 - off / on (use default value of 1000)
4285  *
4286  *     Legal timer values are:
4287  *     50,100,250,500,1000,2000,5000,10000
4288  *
4289  *     Turning off interrupt moderation will also turn this off.
4290  ************************************************************************/
4291 static int
4292 ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS)
4293 {
4294 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4295 	struct ifnet   *ifp = iflib_get_ifp(sc->ctx);
4296 	int            error;
4297 	u16            newval;
4298 
4299 	newval = sc->dmac;
4300 	error = sysctl_handle_16(oidp, &newval, 0, req);
4301 	if ((error) || (req->newptr == NULL))
4302 		return (error);
4303 
4304 	switch (newval) {
4305 	case 0:
4306 		/* Disabled */
4307 		sc->dmac = 0;
4308 		break;
4309 	case 1:
4310 		/* Enable and use default */
4311 		sc->dmac = 1000;
4312 		break;
4313 	case 50:
4314 	case 100:
4315 	case 250:
4316 	case 500:
4317 	case 1000:
4318 	case 2000:
4319 	case 5000:
4320 	case 10000:
4321 		/* Legal values - allow */
4322 		sc->dmac = newval;
4323 		break;
4324 	default:
4325 		/* Do nothing, illegal value */
4326 		return (EINVAL);
4327 	}
4328 
4329 	/* Re-initialize hardware if it's already running */
4330 	if (ifp->if_drv_flags & IFF_DRV_RUNNING)
4331 		ifp->if_init(ifp);
4332 
4333 	return (0);
4334 } /* ixgbe_sysctl_dmac */
4335 
4336 #ifdef IXGBE_DEBUG
4337 /************************************************************************
4338  * ixgbe_sysctl_power_state
4339  *
4340  *   Sysctl to test power states
4341  *   Values:
4342  *     0      - set device to D0
4343  *     3      - set device to D3
4344  *     (none) - get current device power state
4345  ************************************************************************/
4346 static int
4347 ixgbe_sysctl_power_state(SYSCTL_HANDLER_ARGS)
4348 {
4349 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4350 	device_t       dev = sc->dev;
4351 	int            curr_ps, new_ps, error = 0;
4352 
4353 	curr_ps = new_ps = pci_get_powerstate(dev);
4354 
4355 	error = sysctl_handle_int(oidp, &new_ps, 0, req);
4356 	if ((error) || (req->newptr == NULL))
4357 		return (error);
4358 
4359 	if (new_ps == curr_ps)
4360 		return (0);
4361 
4362 	if (new_ps == 3 && curr_ps == 0)
4363 		error = DEVICE_SUSPEND(dev);
4364 	else if (new_ps == 0 && curr_ps == 3)
4365 		error = DEVICE_RESUME(dev);
4366 	else
4367 		return (EINVAL);
4368 
4369 	device_printf(dev, "New state: %d\n", pci_get_powerstate(dev));
4370 
4371 	return (error);
4372 } /* ixgbe_sysctl_power_state */
4373 #endif
4374 
4375 /************************************************************************
4376  * ixgbe_sysctl_wol_enable
4377  *
4378  *   Sysctl to enable/disable the WoL capability,
4379  *   if supported by the adapter.
4380  *
4381  *   Values:
4382  *     0 - disabled
4383  *     1 - enabled
4384  ************************************************************************/
4385 static int
4386 ixgbe_sysctl_wol_enable(SYSCTL_HANDLER_ARGS)
4387 {
4388 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4389 	struct ixgbe_hw *hw = &sc->hw;
4390 	int             new_wol_enabled;
4391 	int             error = 0;
4392 
4393 	new_wol_enabled = hw->wol_enabled;
4394 	error = sysctl_handle_int(oidp, &new_wol_enabled, 0, req);
4395 	if ((error) || (req->newptr == NULL))
4396 		return (error);
4397 	new_wol_enabled = !!(new_wol_enabled);
4398 	if (new_wol_enabled == hw->wol_enabled)
4399 		return (0);
4400 
4401 	if (new_wol_enabled > 0 && !sc->wol_support)
4402 		return (ENODEV);
4403 	else
4404 		hw->wol_enabled = new_wol_enabled;
4405 
4406 	return (0);
4407 } /* ixgbe_sysctl_wol_enable */
4408 
4409 /************************************************************************
4410  * ixgbe_sysctl_wufc - Wake Up Filter Control
4411  *
4412  *   Sysctl to enable/disable the types of packets that the
4413  *   adapter will wake up on upon receipt.
4414  *   Flags:
4415  *     0x1  - Link Status Change
4416  *     0x2  - Magic Packet
4417  *     0x4  - Direct Exact
4418  *     0x8  - Directed Multicast
4419  *     0x10 - Broadcast
4420  *     0x20 - ARP/IPv4 Request Packet
4421  *     0x40 - Direct IPv4 Packet
4422  *     0x80 - Direct IPv6 Packet
4423  *
4424  *   Settings not listed above will cause the sysctl to return an error.
4425  ************************************************************************/
4426 static int
4427 ixgbe_sysctl_wufc(SYSCTL_HANDLER_ARGS)
4428 {
4429 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4430 	int            error = 0;
4431 	u32            new_wufc;
4432 
4433 	new_wufc = sc->wufc;
4434 
4435 	error = sysctl_handle_32(oidp, &new_wufc, 0, req);
4436 	if ((error) || (req->newptr == NULL))
4437 		return (error);
4438 	if (new_wufc == sc->wufc)
4439 		return (0);
4440 
4441 	if (new_wufc & 0xffffff00)
4442 		return (EINVAL);
4443 
4444 	new_wufc &= 0xff;
4445 	new_wufc |= (0xffffff & sc->wufc);
4446 	sc->wufc = new_wufc;
4447 
4448 	return (0);
4449 } /* ixgbe_sysctl_wufc */
4450 
4451 #ifdef IXGBE_DEBUG
4452 /************************************************************************
4453  * ixgbe_sysctl_print_rss_config
4454  ************************************************************************/
4455 static int
4456 ixgbe_sysctl_print_rss_config(SYSCTL_HANDLER_ARGS)
4457 {
4458 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4459 	struct ixgbe_hw *hw = &sc->hw;
4460 	device_t        dev = sc->dev;
4461 	struct sbuf     *buf;
4462 	int             error = 0, reta_size;
4463 	u32             reg;
4464 
4465 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4466 	if (!buf) {
4467 		device_printf(dev, "Could not allocate sbuf for output.\n");
4468 		return (ENOMEM);
4469 	}
4470 
4471 	// TODO: use sbufs to make a string to print out
4472 	/* Set multiplier for RETA setup and table size based on MAC */
4473 	switch (sc->hw.mac.type) {
4474 	case ixgbe_mac_X550:
4475 	case ixgbe_mac_X550EM_x:
4476 	case ixgbe_mac_X550EM_a:
4477 		reta_size = 128;
4478 		break;
4479 	default:
4480 		reta_size = 32;
4481 		break;
4482 	}
4483 
4484 	/* Print out the redirection table */
4485 	sbuf_cat(buf, "\n");
4486 	for (int i = 0; i < reta_size; i++) {
4487 		if (i < 32) {
4488 			reg = IXGBE_READ_REG(hw, IXGBE_RETA(i));
4489 			sbuf_printf(buf, "RETA(%2d): 0x%08x\n", i, reg);
4490 		} else {
4491 			reg = IXGBE_READ_REG(hw, IXGBE_ERETA(i - 32));
4492 			sbuf_printf(buf, "ERETA(%2d): 0x%08x\n", i - 32, reg);
4493 		}
4494 	}
4495 
4496 	// TODO: print more config
4497 
4498 	error = sbuf_finish(buf);
4499 	if (error)
4500 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4501 
4502 	sbuf_delete(buf);
4503 
4504 	return (0);
4505 } /* ixgbe_sysctl_print_rss_config */
4506 #endif /* IXGBE_DEBUG */
4507 
4508 /************************************************************************
4509  * ixgbe_sysctl_phy_temp - Retrieve temperature of PHY
4510  *
4511  *   For X552/X557-AT devices using an external PHY
4512  ************************************************************************/
4513 static int
4514 ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS)
4515 {
4516 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4517 	struct ixgbe_hw *hw = &sc->hw;
4518 	u16             reg;
4519 
4520 	if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) {
4521 		device_printf(iflib_get_dev(sc->ctx),
4522 		    "Device has no supported external thermal sensor.\n");
4523 		return (ENODEV);
4524 	}
4525 
4526 	if (hw->phy.ops.read_reg(hw, IXGBE_PHY_CURRENT_TEMP,
4527 	    IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, &reg)) {
4528 		device_printf(iflib_get_dev(sc->ctx),
4529 		    "Error reading from PHY's current temperature register\n");
4530 		return (EAGAIN);
4531 	}
4532 
4533 	/* Shift temp for output */
4534 	reg = reg >> 8;
4535 
4536 	return (sysctl_handle_16(oidp, NULL, reg, req));
4537 } /* ixgbe_sysctl_phy_temp */
4538 
4539 /************************************************************************
4540  * ixgbe_sysctl_phy_overtemp_occurred
4541  *
4542  *   Reports (directly from the PHY) whether the current PHY
4543  *   temperature is over the overtemp threshold.
4544  ************************************************************************/
4545 static int
4546 ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS)
4547 {
4548 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4549 	struct ixgbe_hw *hw = &sc->hw;
4550 	u16             reg;
4551 
4552 	if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) {
4553 		device_printf(iflib_get_dev(sc->ctx),
4554 		    "Device has no supported external thermal sensor.\n");
4555 		return (ENODEV);
4556 	}
4557 
4558 	if (hw->phy.ops.read_reg(hw, IXGBE_PHY_OVERTEMP_STATUS,
4559 	    IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, &reg)) {
4560 		device_printf(iflib_get_dev(sc->ctx),
4561 		    "Error reading from PHY's temperature status register\n");
4562 		return (EAGAIN);
4563 	}
4564 
4565 	/* Get occurrence bit */
4566 	reg = !!(reg & 0x4000);
4567 
4568 	return (sysctl_handle_16(oidp, 0, reg, req));
4569 } /* ixgbe_sysctl_phy_overtemp_occurred */
4570 
4571 /************************************************************************
4572  * ixgbe_sysctl_eee_state
4573  *
4574  *   Sysctl to set EEE power saving feature
4575  *   Values:
4576  *     0      - disable EEE
4577  *     1      - enable EEE
4578  *     (none) - get current device EEE state
4579  ************************************************************************/
4580 static int
4581 ixgbe_sysctl_eee_state(SYSCTL_HANDLER_ARGS)
4582 {
4583 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4584 	device_t       dev = sc->dev;
4585 	struct ifnet   *ifp = iflib_get_ifp(sc->ctx);
4586 	int            curr_eee, new_eee, error = 0;
4587 	s32            retval;
4588 
4589 	curr_eee = new_eee = !!(sc->feat_en & IXGBE_FEATURE_EEE);
4590 
4591 	error = sysctl_handle_int(oidp, &new_eee, 0, req);
4592 	if ((error) || (req->newptr == NULL))
4593 		return (error);
4594 
4595 	/* Nothing to do */
4596 	if (new_eee == curr_eee)
4597 		return (0);
4598 
4599 	/* Not supported */
4600 	if (!(sc->feat_cap & IXGBE_FEATURE_EEE))
4601 		return (EINVAL);
4602 
4603 	/* Bounds checking */
4604 	if ((new_eee < 0) || (new_eee > 1))
4605 		return (EINVAL);
4606 
4607 	retval = ixgbe_setup_eee(&sc->hw, new_eee);
4608 	if (retval) {
4609 		device_printf(dev, "Error in EEE setup: 0x%08X\n", retval);
4610 		return (EINVAL);
4611 	}
4612 
4613 	/* Restart auto-neg */
4614 	ifp->if_init(ifp);
4615 
4616 	device_printf(dev, "New EEE state: %d\n", new_eee);
4617 
4618 	/* Cache new value */
4619 	if (new_eee)
4620 		sc->feat_en |= IXGBE_FEATURE_EEE;
4621 	else
4622 		sc->feat_en &= ~IXGBE_FEATURE_EEE;
4623 
4624 	return (error);
4625 } /* ixgbe_sysctl_eee_state */
4626 
4627 /************************************************************************
4628  * ixgbe_init_device_features
4629  ************************************************************************/
4630 static void
4631 ixgbe_init_device_features(struct ixgbe_softc *sc)
4632 {
4633 	sc->feat_cap = IXGBE_FEATURE_NETMAP
4634 	                  | IXGBE_FEATURE_RSS
4635 	                  | IXGBE_FEATURE_MSI
4636 	                  | IXGBE_FEATURE_MSIX
4637 	                  | IXGBE_FEATURE_LEGACY_IRQ;
4638 
4639 	/* Set capabilities first... */
4640 	switch (sc->hw.mac.type) {
4641 	case ixgbe_mac_82598EB:
4642 		if (sc->hw.device_id == IXGBE_DEV_ID_82598AT)
4643 			sc->feat_cap |= IXGBE_FEATURE_FAN_FAIL;
4644 		break;
4645 	case ixgbe_mac_X540:
4646 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4647 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4648 		if ((sc->hw.device_id == IXGBE_DEV_ID_X540_BYPASS) &&
4649 		    (sc->hw.bus.func == 0))
4650 			sc->feat_cap |= IXGBE_FEATURE_BYPASS;
4651 		break;
4652 	case ixgbe_mac_X550:
4653 		sc->feat_cap |= IXGBE_FEATURE_TEMP_SENSOR;
4654 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4655 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4656 		break;
4657 	case ixgbe_mac_X550EM_x:
4658 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4659 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4660 		break;
4661 	case ixgbe_mac_X550EM_a:
4662 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4663 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4664 		sc->feat_cap &= ~IXGBE_FEATURE_LEGACY_IRQ;
4665 		if ((sc->hw.device_id == IXGBE_DEV_ID_X550EM_A_1G_T) ||
4666 		    (sc->hw.device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)) {
4667 			sc->feat_cap |= IXGBE_FEATURE_TEMP_SENSOR;
4668 			sc->feat_cap |= IXGBE_FEATURE_EEE;
4669 		}
4670 		break;
4671 	case ixgbe_mac_82599EB:
4672 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4673 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4674 		if ((sc->hw.device_id == IXGBE_DEV_ID_82599_BYPASS) &&
4675 		    (sc->hw.bus.func == 0))
4676 			sc->feat_cap |= IXGBE_FEATURE_BYPASS;
4677 		if (sc->hw.device_id == IXGBE_DEV_ID_82599_QSFP_SF_QP)
4678 			sc->feat_cap &= ~IXGBE_FEATURE_LEGACY_IRQ;
4679 		break;
4680 	default:
4681 		break;
4682 	}
4683 
4684 	/* Enabled by default... */
4685 	/* Fan failure detection */
4686 	if (sc->feat_cap & IXGBE_FEATURE_FAN_FAIL)
4687 		sc->feat_en |= IXGBE_FEATURE_FAN_FAIL;
4688 	/* Netmap */
4689 	if (sc->feat_cap & IXGBE_FEATURE_NETMAP)
4690 		sc->feat_en |= IXGBE_FEATURE_NETMAP;
4691 	/* EEE */
4692 	if (sc->feat_cap & IXGBE_FEATURE_EEE)
4693 		sc->feat_en |= IXGBE_FEATURE_EEE;
4694 	/* Thermal Sensor */
4695 	if (sc->feat_cap & IXGBE_FEATURE_TEMP_SENSOR)
4696 		sc->feat_en |= IXGBE_FEATURE_TEMP_SENSOR;
4697 
4698 	/* Enabled via global sysctl... */
4699 	/* Flow Director */
4700 	if (ixgbe_enable_fdir) {
4701 		if (sc->feat_cap & IXGBE_FEATURE_FDIR)
4702 			sc->feat_en |= IXGBE_FEATURE_FDIR;
4703 		else
4704 			device_printf(sc->dev, "Device does not support Flow Director. Leaving disabled.");
4705 	}
4706 	/*
4707 	 * Message Signal Interrupts - Extended (MSI-X)
4708 	 * Normal MSI is only enabled if MSI-X calls fail.
4709 	 */
4710 	if (!ixgbe_enable_msix)
4711 		sc->feat_cap &= ~IXGBE_FEATURE_MSIX;
4712 	/* Receive-Side Scaling (RSS) */
4713 	if ((sc->feat_cap & IXGBE_FEATURE_RSS) && ixgbe_enable_rss)
4714 		sc->feat_en |= IXGBE_FEATURE_RSS;
4715 
4716 	/* Disable features with unmet dependencies... */
4717 	/* No MSI-X */
4718 	if (!(sc->feat_cap & IXGBE_FEATURE_MSIX)) {
4719 		sc->feat_cap &= ~IXGBE_FEATURE_RSS;
4720 		sc->feat_cap &= ~IXGBE_FEATURE_SRIOV;
4721 		sc->feat_en &= ~IXGBE_FEATURE_RSS;
4722 		sc->feat_en &= ~IXGBE_FEATURE_SRIOV;
4723 	}
4724 } /* ixgbe_init_device_features */
4725 
4726 /************************************************************************
4727  * ixgbe_check_fan_failure
4728  ************************************************************************/
4729 static void
4730 ixgbe_check_fan_failure(struct ixgbe_softc *sc, u32 reg, bool in_interrupt)
4731 {
4732 	u32 mask;
4733 
4734 	mask = (in_interrupt) ? IXGBE_EICR_GPI_SDP1_BY_MAC(&sc->hw) :
4735 	    IXGBE_ESDP_SDP1;
4736 
4737 	if (reg & mask)
4738 		device_printf(sc->dev, "\nCRITICAL: FAN FAILURE!! REPLACE IMMEDIATELY!!\n");
4739 } /* ixgbe_check_fan_failure */
4740 
4741 /************************************************************************
4742  * ixgbe_sbuf_fw_version
4743  ************************************************************************/
4744 static void
4745 ixgbe_sbuf_fw_version(struct ixgbe_hw *hw, struct sbuf *buf)
4746 {
4747 	struct ixgbe_nvm_version nvm_ver = {0};
4748 	uint16_t phyfw = 0;
4749 	int status;
4750 	const char *space = "";
4751 
4752 	ixgbe_get_oem_prod_version(hw, &nvm_ver); /* OEM's NVM version */
4753 	ixgbe_get_orom_version(hw, &nvm_ver); /* Option ROM */
4754 	ixgbe_get_etk_id(hw, &nvm_ver); /* eTrack identifies a build in Intel's SCM */
4755 	status = ixgbe_get_phy_firmware_version(hw, &phyfw);
4756 
4757 	if (nvm_ver.oem_valid) {
4758 		sbuf_printf(buf, "NVM OEM V%d.%d R%d", nvm_ver.oem_major,
4759 		    nvm_ver.oem_minor, nvm_ver.oem_release);
4760 		space = " ";
4761 	}
4762 
4763 	if (nvm_ver.or_valid) {
4764 		sbuf_printf(buf, "%sOption ROM V%d-b%d-p%d",
4765 		    space, nvm_ver.or_major, nvm_ver.or_build, nvm_ver.or_patch);
4766 		space = " ";
4767 	}
4768 
4769 	if (nvm_ver.etk_id != ((NVM_VER_INVALID << NVM_ETK_SHIFT) |
4770 	    NVM_VER_INVALID)) {
4771 		sbuf_printf(buf, "%seTrack 0x%08x", space, nvm_ver.etk_id);
4772 		space = " ";
4773 	}
4774 
4775 	if (phyfw != 0 && status == IXGBE_SUCCESS)
4776 		sbuf_printf(buf, "%sPHY FW V%d", space, phyfw);
4777 } /* ixgbe_sbuf_fw_version */
4778 
4779 /************************************************************************
4780  * ixgbe_print_fw_version
4781  ************************************************************************/
4782 static void
4783 ixgbe_print_fw_version(if_ctx_t ctx)
4784 {
4785 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
4786 	struct ixgbe_hw *hw = &sc->hw;
4787 	device_t dev = sc->dev;
4788 	struct sbuf *buf;
4789 	int error = 0;
4790 
4791 	buf = sbuf_new_auto();
4792 	if (!buf) {
4793 		device_printf(dev, "Could not allocate sbuf for output.\n");
4794 		return;
4795 	}
4796 
4797 	ixgbe_sbuf_fw_version(hw, buf);
4798 
4799 	error = sbuf_finish(buf);
4800 	if (error)
4801 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4802 	else if (sbuf_len(buf))
4803 		device_printf(dev, "%s\n", sbuf_data(buf));
4804 
4805 	sbuf_delete(buf);
4806 } /* ixgbe_print_fw_version */
4807 
4808 /************************************************************************
4809  * ixgbe_sysctl_print_fw_version
4810  ************************************************************************/
4811 static int
4812 ixgbe_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS)
4813 {
4814 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4815 	struct ixgbe_hw *hw = &sc->hw;
4816 	device_t dev = sc->dev;
4817 	struct sbuf *buf;
4818 	int error = 0;
4819 
4820 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4821 	if (!buf) {
4822 		device_printf(dev, "Could not allocate sbuf for output.\n");
4823 		return (ENOMEM);
4824 	}
4825 
4826 	ixgbe_sbuf_fw_version(hw, buf);
4827 
4828 	error = sbuf_finish(buf);
4829 	if (error)
4830 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4831 
4832 	sbuf_delete(buf);
4833 
4834 	return (0);
4835 } /* ixgbe_sysctl_print_fw_version */
4836