xref: /freebsd/sys/dev/ixgbe/if_ix.c (revision 5bb3134a8c21cb87b30e135ef168483f0333dabb)
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_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_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 (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU ||
1290 	    layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA)
1291 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_TWINAX, 0,
1292 		    NULL);
1293 
1294 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR) {
1295 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_LR, 0, NULL);
1296 		if (hw->phy.multispeed_fiber)
1297 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_LX, 0,
1298 			    NULL);
1299 	}
1300 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR) {
1301 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_SR, 0, NULL);
1302 		if (hw->phy.multispeed_fiber)
1303 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_SX, 0,
1304 			    NULL);
1305 	} else if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX)
1306 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_SX, 0, NULL);
1307 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4)
1308 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_CX4, 0, NULL);
1309 
1310 #ifdef IFM_ETH_XTYPE
1311 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
1312 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_KR, 0, NULL);
1313 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4)
1314 		ifmedia_add( sc->media, IFM_ETHER | IFM_10G_KX4, 0, NULL);
1315 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
1316 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_KX, 0, NULL);
1317 	if (layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX)
1318 		ifmedia_add(sc->media, IFM_ETHER | IFM_2500_KX, 0, NULL);
1319 #else
1320 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR) {
1321 		device_printf(dev, "Media supported: 10GbaseKR\n");
1322 		device_printf(dev, "10GbaseKR mapped to 10GbaseSR\n");
1323 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_SR, 0, NULL);
1324 	}
1325 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4) {
1326 		device_printf(dev, "Media supported: 10GbaseKX4\n");
1327 		device_printf(dev, "10GbaseKX4 mapped to 10GbaseCX4\n");
1328 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_CX4, 0, NULL);
1329 	}
1330 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) {
1331 		device_printf(dev, "Media supported: 1000baseKX\n");
1332 		device_printf(dev, "1000baseKX mapped to 1000baseCX\n");
1333 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_CX, 0, NULL);
1334 	}
1335 	if (layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX) {
1336 		device_printf(dev, "Media supported: 2500baseKX\n");
1337 		device_printf(dev, "2500baseKX mapped to 2500baseSX\n");
1338 		ifmedia_add(sc->media, IFM_ETHER | IFM_2500_SX, 0, NULL);
1339 	}
1340 #endif
1341 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_BX)
1342 		device_printf(dev, "Media supported: 1000baseBX\n");
1343 
1344 	if (hw->device_id == IXGBE_DEV_ID_82598AT) {
1345 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T | IFM_FDX,
1346 		    0, NULL);
1347 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
1348 	}
1349 
1350 	ifmedia_add(sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
1351 } /* ixgbe_add_media_types */
1352 
1353 /************************************************************************
1354  * ixgbe_is_sfp
1355  ************************************************************************/
1356 static inline bool
1357 ixgbe_is_sfp(struct ixgbe_hw *hw)
1358 {
1359 	switch (hw->mac.type) {
1360 	case ixgbe_mac_82598EB:
1361 		if (hw->phy.type == ixgbe_phy_nl)
1362 			return (true);
1363 		return (false);
1364 	case ixgbe_mac_82599EB:
1365 		switch (hw->mac.ops.get_media_type(hw)) {
1366 		case ixgbe_media_type_fiber:
1367 		case ixgbe_media_type_fiber_qsfp:
1368 			return (true);
1369 		default:
1370 			return (false);
1371 		}
1372 	case ixgbe_mac_X550EM_x:
1373 	case ixgbe_mac_X550EM_a:
1374 		if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber)
1375 			return (true);
1376 		return (false);
1377 	default:
1378 		return (false);
1379 	}
1380 } /* ixgbe_is_sfp */
1381 
1382 /************************************************************************
1383  * ixgbe_config_link
1384  ************************************************************************/
1385 static void
1386 ixgbe_config_link(if_ctx_t ctx)
1387 {
1388 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
1389 	struct ixgbe_hw *hw = &sc->hw;
1390 	u32             autoneg, err = 0;
1391 	bool            sfp, negotiate;
1392 
1393 	sfp = ixgbe_is_sfp(hw);
1394 
1395 	if (sfp) {
1396 		sc->task_requests |= IXGBE_REQUEST_TASK_MOD;
1397 		iflib_admin_intr_deferred(ctx);
1398 	} else {
1399 		if (hw->mac.ops.check_link)
1400 			err = ixgbe_check_link(hw, &sc->link_speed,
1401 			    &sc->link_up, false);
1402 		if (err)
1403 			return;
1404 		autoneg = hw->phy.autoneg_advertised;
1405 		if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
1406 			err = hw->mac.ops.get_link_capabilities(hw, &autoneg,
1407 			    &negotiate);
1408 		if (err)
1409 			return;
1410 		if (hw->mac.ops.setup_link)
1411 			err = hw->mac.ops.setup_link(hw, autoneg,
1412 			    sc->link_up);
1413 	}
1414 } /* ixgbe_config_link */
1415 
1416 /************************************************************************
1417  * ixgbe_update_stats_counters - Update board statistics counters.
1418  ************************************************************************/
1419 static void
1420 ixgbe_update_stats_counters(struct ixgbe_softc *sc)
1421 {
1422 	struct ixgbe_hw       *hw = &sc->hw;
1423 	struct ixgbe_hw_stats *stats = &sc->stats.pf;
1424 	u32                   missed_rx = 0, bprc, lxon, lxoff, total;
1425 	u32                   lxoffrxc;
1426 	u64                   total_missed_rx = 0;
1427 
1428 	stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
1429 	stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
1430 	stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
1431 	stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
1432 	stats->mpc[0] += IXGBE_READ_REG(hw, IXGBE_MPC(0));
1433 
1434 	for (int i = 0; i < 16; i++) {
1435 		stats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
1436 		stats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
1437 		stats->qprdc[i] += IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
1438 	}
1439 	stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
1440 	stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
1441 	stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
1442 
1443 	/* Hardware workaround, gprc counts missed packets */
1444 	stats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
1445 	stats->gprc -= missed_rx;
1446 
1447 	if (hw->mac.type != ixgbe_mac_82598EB) {
1448 		stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL) +
1449 		    ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
1450 		stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL) +
1451 		    ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
1452 		stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL) +
1453 		    ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
1454 		stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
1455 		lxoffrxc = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
1456 		stats->lxoffrxc += lxoffrxc;
1457 	} else {
1458 		stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
1459 		lxoffrxc = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
1460 		stats->lxoffrxc += lxoffrxc;
1461 		/* 82598 only has a counter in the high register */
1462 		stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
1463 		stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
1464 		stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
1465 	}
1466 
1467 	/*
1468 	 * For watchdog management we need to know if we have been paused
1469 	 * during the last interval, so capture that here.
1470 	*/
1471 	if (lxoffrxc)
1472 		sc->shared->isc_pause_frames = 1;
1473 
1474 	/*
1475 	 * Workaround: mprc hardware is incorrectly counting
1476 	 * broadcasts, so for now we subtract those.
1477 	 */
1478 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
1479 	stats->bprc += bprc;
1480 	stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
1481 	if (hw->mac.type == ixgbe_mac_82598EB)
1482 		stats->mprc -= bprc;
1483 
1484 	stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
1485 	stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
1486 	stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
1487 	stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
1488 	stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
1489 	stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
1490 
1491 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
1492 	stats->lxontxc += lxon;
1493 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
1494 	stats->lxofftxc += lxoff;
1495 	total = lxon + lxoff;
1496 
1497 	stats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
1498 	stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
1499 	stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
1500 	stats->gptc -= total;
1501 	stats->mptc -= total;
1502 	stats->ptc64 -= total;
1503 	stats->gotc -= total * ETHER_MIN_LEN;
1504 
1505 	stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
1506 	stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
1507 	stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
1508 	stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
1509 	stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
1510 	stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
1511 	stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
1512 	stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
1513 	stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
1514 	stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
1515 	stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
1516 	stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
1517 	stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
1518 	stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
1519 	stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
1520 	stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
1521 	stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
1522 	stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
1523 	/* Only read FCOE on 82599 */
1524 	if (hw->mac.type != ixgbe_mac_82598EB) {
1525 		stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
1526 		stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
1527 		stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
1528 		stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
1529 		stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
1530 	}
1531 
1532 	/* Fill out the OS statistics structure */
1533 	IXGBE_SET_IPACKETS(sc, stats->gprc);
1534 	IXGBE_SET_OPACKETS(sc, stats->gptc);
1535 	IXGBE_SET_IBYTES(sc, stats->gorc);
1536 	IXGBE_SET_OBYTES(sc, stats->gotc);
1537 	IXGBE_SET_IMCASTS(sc, stats->mprc);
1538 	IXGBE_SET_OMCASTS(sc, stats->mptc);
1539 	IXGBE_SET_COLLISIONS(sc, 0);
1540 	IXGBE_SET_IQDROPS(sc, total_missed_rx);
1541 
1542 	/*
1543 	 * Aggregate following types of errors as RX errors:
1544 	 * - CRC error count,
1545 	 * - illegal byte error count,
1546 	 * - checksum error count,
1547 	 * - missed packets count,
1548 	 * - length error count,
1549 	 * - undersized packets count,
1550 	 * - fragmented packets count,
1551 	 * - oversized packets count,
1552 	 * - jabber count.
1553 	 */
1554 	IXGBE_SET_IERRORS(sc, stats->crcerrs + stats->illerrc + stats->xec +
1555 	    stats->mpc[0] + stats->rlec + stats->ruc + stats->rfc + stats->roc +
1556 	    stats->rjc);
1557 } /* ixgbe_update_stats_counters */
1558 
1559 /************************************************************************
1560  * ixgbe_add_hw_stats
1561  *
1562  *   Add sysctl variables, one per statistic, to the system.
1563  ************************************************************************/
1564 static void
1565 ixgbe_add_hw_stats(struct ixgbe_softc *sc)
1566 {
1567 	device_t               dev = iflib_get_dev(sc->ctx);
1568 	struct ix_rx_queue     *rx_que;
1569 	struct ix_tx_queue     *tx_que;
1570 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
1571 	struct sysctl_oid      *tree = device_get_sysctl_tree(dev);
1572 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
1573 	struct ixgbe_hw_stats  *stats = &sc->stats.pf;
1574 	struct sysctl_oid      *stat_node, *queue_node;
1575 	struct sysctl_oid_list *stat_list, *queue_list;
1576 	int                    i;
1577 
1578 #define QUEUE_NAME_LEN 32
1579 	char                   namebuf[QUEUE_NAME_LEN];
1580 
1581 	/* Driver Statistics */
1582 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
1583 	    CTLFLAG_RD, &sc->dropped_pkts, "Driver dropped packets");
1584 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_events",
1585 	    CTLFLAG_RD, &sc->watchdog_events, "Watchdog timeouts");
1586 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
1587 	    CTLFLAG_RD, &sc->link_irq, "Link MSI-X IRQ Handled");
1588 
1589 	for (i = 0, tx_que = sc->tx_queues; i < sc->num_tx_queues; i++, tx_que++) {
1590 		struct tx_ring *txr = &tx_que->txr;
1591 		snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i);
1592 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
1593 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Queue Name");
1594 		queue_list = SYSCTL_CHILDREN(queue_node);
1595 
1596 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
1597 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, txr, 0,
1598 		    ixgbe_sysctl_tdh_handler, "IU", "Transmit Descriptor Head");
1599 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
1600 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, txr, 0,
1601 		    ixgbe_sysctl_tdt_handler, "IU", "Transmit Descriptor Tail");
1602 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tso_tx",
1603 		    CTLFLAG_RD, &txr->tso_tx, "TSO");
1604 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tx_packets",
1605 		    CTLFLAG_RD, &txr->total_packets,
1606 		    "Queue Packets Transmitted");
1607 	}
1608 
1609 	for (i = 0, rx_que = sc->rx_queues; i < sc->num_rx_queues; i++, rx_que++) {
1610 		struct rx_ring *rxr = &rx_que->rxr;
1611 		snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i);
1612 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
1613 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Queue Name");
1614 		queue_list = SYSCTL_CHILDREN(queue_node);
1615 
1616 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "interrupt_rate",
1617 		    CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
1618 		    &sc->rx_queues[i], 0,
1619 		    ixgbe_sysctl_interrupt_rate_handler, "IU",
1620 		    "Interrupt Rate");
1621 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "irqs",
1622 		    CTLFLAG_RD, &(sc->rx_queues[i].irqs),
1623 		    "irqs on this queue");
1624 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
1625 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, rxr, 0,
1626 		    ixgbe_sysctl_rdh_handler, "IU", "Receive Descriptor Head");
1627 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
1628 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, rxr, 0,
1629 		    ixgbe_sysctl_rdt_handler, "IU", "Receive Descriptor Tail");
1630 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_packets",
1631 		    CTLFLAG_RD, &rxr->rx_packets, "Queue Packets Received");
1632 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_bytes",
1633 		    CTLFLAG_RD, &rxr->rx_bytes, "Queue Bytes Received");
1634 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_copies",
1635 		    CTLFLAG_RD, &rxr->rx_copies, "Copied RX Frames");
1636 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_discarded",
1637 		    CTLFLAG_RD, &rxr->rx_discarded, "Discarded RX packets");
1638 	}
1639 
1640 	/* MAC stats get their own sub node */
1641 
1642 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
1643 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "MAC Statistics");
1644 	stat_list = SYSCTL_CHILDREN(stat_node);
1645 
1646 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_errs",
1647 	    CTLFLAG_RD, &sc->ierrors, IXGBE_SYSCTL_DESC_RX_ERRS);
1648 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
1649 	    CTLFLAG_RD, &stats->crcerrs, "CRC Errors");
1650 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "ill_errs",
1651 	    CTLFLAG_RD, &stats->illerrc, "Illegal Byte Errors");
1652 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "byte_errs",
1653 	    CTLFLAG_RD, &stats->errbc, "Byte Errors");
1654 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "short_discards",
1655 	    CTLFLAG_RD, &stats->mspdc, "MAC Short Packets Discarded");
1656 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "local_faults",
1657 	    CTLFLAG_RD, &stats->mlfc, "MAC Local Faults");
1658 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "remote_faults",
1659 	    CTLFLAG_RD, &stats->mrfc, "MAC Remote Faults");
1660 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rec_len_errs",
1661 	    CTLFLAG_RD, &stats->rlec, "Receive Length Errors");
1662 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_missed_packets",
1663 	    CTLFLAG_RD, &stats->mpc[0], "RX Missed Packet Count");
1664 
1665 	/* Flow Control stats */
1666 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
1667 	    CTLFLAG_RD, &stats->lxontxc, "Link XON Transmitted");
1668 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
1669 	    CTLFLAG_RD, &stats->lxonrxc, "Link XON Received");
1670 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
1671 	    CTLFLAG_RD, &stats->lxofftxc, "Link XOFF Transmitted");
1672 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
1673 	    CTLFLAG_RD, &stats->lxoffrxc, "Link XOFF Received");
1674 
1675 	/* Packet Reception Stats */
1676 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_octets_rcvd",
1677 	    CTLFLAG_RD, &stats->tor, "Total Octets Received");
1678 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_rcvd",
1679 	    CTLFLAG_RD, &stats->gorc, "Good Octets Received");
1680 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_rcvd",
1681 	    CTLFLAG_RD, &stats->tpr, "Total Packets Received");
1682 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_rcvd",
1683 	    CTLFLAG_RD, &stats->gprc, "Good Packets Received");
1684 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_rcvd",
1685 	    CTLFLAG_RD, &stats->mprc, "Multicast Packets Received");
1686 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_rcvd",
1687 	    CTLFLAG_RD, &stats->bprc, "Broadcast Packets Received");
1688 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
1689 	    CTLFLAG_RD, &stats->prc64, "64 byte frames received ");
1690 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
1691 	    CTLFLAG_RD, &stats->prc127, "65-127 byte frames received");
1692 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
1693 	    CTLFLAG_RD, &stats->prc255, "128-255 byte frames received");
1694 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
1695 	    CTLFLAG_RD, &stats->prc511, "256-511 byte frames received");
1696 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
1697 	    CTLFLAG_RD, &stats->prc1023, "512-1023 byte frames received");
1698 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
1699 	    CTLFLAG_RD, &stats->prc1522, "1023-1522 byte frames received");
1700 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersized",
1701 	    CTLFLAG_RD, &stats->ruc, "Receive Undersized");
1702 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
1703 	    CTLFLAG_RD, &stats->rfc, "Fragmented Packets Received ");
1704 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversized",
1705 	    CTLFLAG_RD, &stats->roc, "Oversized Packets Received");
1706 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabberd",
1707 	    CTLFLAG_RD, &stats->rjc, "Received Jabber");
1708 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_rcvd",
1709 	    CTLFLAG_RD, &stats->mngprc, "Management Packets Received");
1710 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_drpd",
1711 	    CTLFLAG_RD, &stats->mngptc, "Management Packets Dropped");
1712 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "checksum_errs",
1713 	    CTLFLAG_RD, &stats->xec, "Checksum Errors");
1714 
1715 	/* Packet Transmission Stats */
1716 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
1717 	    CTLFLAG_RD, &stats->gotc, "Good Octets Transmitted");
1718 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
1719 	    CTLFLAG_RD, &stats->tpt, "Total Packets Transmitted");
1720 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
1721 	    CTLFLAG_RD, &stats->gptc, "Good Packets Transmitted");
1722 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
1723 	    CTLFLAG_RD, &stats->bptc, "Broadcast Packets Transmitted");
1724 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
1725 	    CTLFLAG_RD, &stats->mptc, "Multicast Packets Transmitted");
1726 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_txd",
1727 	    CTLFLAG_RD, &stats->mngptc, "Management Packets Transmitted");
1728 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
1729 	    CTLFLAG_RD, &stats->ptc64, "64 byte frames transmitted ");
1730 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
1731 	    CTLFLAG_RD, &stats->ptc127, "65-127 byte frames transmitted");
1732 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
1733 	    CTLFLAG_RD, &stats->ptc255, "128-255 byte frames transmitted");
1734 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
1735 	    CTLFLAG_RD, &stats->ptc511, "256-511 byte frames transmitted");
1736 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
1737 	    CTLFLAG_RD, &stats->ptc1023, "512-1023 byte frames transmitted");
1738 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
1739 	    CTLFLAG_RD, &stats->ptc1522, "1024-1522 byte frames transmitted");
1740 } /* ixgbe_add_hw_stats */
1741 
1742 /************************************************************************
1743  * ixgbe_sysctl_tdh_handler - Transmit Descriptor Head handler function
1744  *
1745  *   Retrieves the TDH value from the hardware
1746  ************************************************************************/
1747 static int
1748 ixgbe_sysctl_tdh_handler(SYSCTL_HANDLER_ARGS)
1749 {
1750 	struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1);
1751 	int            error;
1752 	unsigned int   val;
1753 
1754 	if (!txr)
1755 		return (0);
1756 
1757 	val = IXGBE_READ_REG(&txr->sc->hw, IXGBE_TDH(txr->me));
1758 	error = sysctl_handle_int(oidp, &val, 0, req);
1759 	if (error || !req->newptr)
1760 		return error;
1761 
1762 	return (0);
1763 } /* ixgbe_sysctl_tdh_handler */
1764 
1765 /************************************************************************
1766  * ixgbe_sysctl_tdt_handler - Transmit Descriptor Tail handler function
1767  *
1768  *   Retrieves the TDT value from the hardware
1769  ************************************************************************/
1770 static int
1771 ixgbe_sysctl_tdt_handler(SYSCTL_HANDLER_ARGS)
1772 {
1773 	struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1);
1774 	int            error;
1775 	unsigned int   val;
1776 
1777 	if (!txr)
1778 		return (0);
1779 
1780 	val = IXGBE_READ_REG(&txr->sc->hw, IXGBE_TDT(txr->me));
1781 	error = sysctl_handle_int(oidp, &val, 0, req);
1782 	if (error || !req->newptr)
1783 		return error;
1784 
1785 	return (0);
1786 } /* ixgbe_sysctl_tdt_handler */
1787 
1788 /************************************************************************
1789  * ixgbe_sysctl_rdh_handler - Receive Descriptor Head handler function
1790  *
1791  *   Retrieves the RDH value from the hardware
1792  ************************************************************************/
1793 static int
1794 ixgbe_sysctl_rdh_handler(SYSCTL_HANDLER_ARGS)
1795 {
1796 	struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1);
1797 	int            error;
1798 	unsigned int   val;
1799 
1800 	if (!rxr)
1801 		return (0);
1802 
1803 	val = IXGBE_READ_REG(&rxr->sc->hw, IXGBE_RDH(rxr->me));
1804 	error = sysctl_handle_int(oidp, &val, 0, req);
1805 	if (error || !req->newptr)
1806 		return error;
1807 
1808 	return (0);
1809 } /* ixgbe_sysctl_rdh_handler */
1810 
1811 /************************************************************************
1812  * ixgbe_sysctl_rdt_handler - Receive Descriptor Tail handler function
1813  *
1814  *   Retrieves the RDT value from the hardware
1815  ************************************************************************/
1816 static int
1817 ixgbe_sysctl_rdt_handler(SYSCTL_HANDLER_ARGS)
1818 {
1819 	struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1);
1820 	int            error;
1821 	unsigned int   val;
1822 
1823 	if (!rxr)
1824 		return (0);
1825 
1826 	val = IXGBE_READ_REG(&rxr->sc->hw, IXGBE_RDT(rxr->me));
1827 	error = sysctl_handle_int(oidp, &val, 0, req);
1828 	if (error || !req->newptr)
1829 		return error;
1830 
1831 	return (0);
1832 } /* ixgbe_sysctl_rdt_handler */
1833 
1834 /************************************************************************
1835  * ixgbe_if_vlan_register
1836  *
1837  *   Run via vlan config EVENT, it enables us to use the
1838  *   HW Filter table since we can get the vlan id. This
1839  *   just creates the entry in the soft version of the
1840  *   VFTA, init will repopulate the real table.
1841  ************************************************************************/
1842 static void
1843 ixgbe_if_vlan_register(if_ctx_t ctx, u16 vtag)
1844 {
1845 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1846 	u16            index, bit;
1847 
1848 	index = (vtag >> 5) & 0x7F;
1849 	bit = vtag & 0x1F;
1850 	sc->shadow_vfta[index] |= (1 << bit);
1851 	++sc->num_vlans;
1852 	ixgbe_setup_vlan_hw_support(ctx);
1853 } /* ixgbe_if_vlan_register */
1854 
1855 /************************************************************************
1856  * ixgbe_if_vlan_unregister
1857  *
1858  *   Run via vlan unconfig EVENT, remove our entry in the soft vfta.
1859  ************************************************************************/
1860 static void
1861 ixgbe_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
1862 {
1863 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1864 	u16            index, bit;
1865 
1866 	index = (vtag >> 5) & 0x7F;
1867 	bit = vtag & 0x1F;
1868 	sc->shadow_vfta[index] &= ~(1 << bit);
1869 	--sc->num_vlans;
1870 	/* Re-init to load the changes */
1871 	ixgbe_setup_vlan_hw_support(ctx);
1872 } /* ixgbe_if_vlan_unregister */
1873 
1874 /************************************************************************
1875  * ixgbe_setup_vlan_hw_support
1876  ************************************************************************/
1877 static void
1878 ixgbe_setup_vlan_hw_support(if_ctx_t ctx)
1879 {
1880 	struct ifnet	*ifp = iflib_get_ifp(ctx);
1881 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
1882 	struct ixgbe_hw *hw = &sc->hw;
1883 	struct rx_ring  *rxr;
1884 	int             i;
1885 	u32             ctrl;
1886 
1887 
1888 	/*
1889 	 * We get here thru init_locked, meaning
1890 	 * a soft reset, this has already cleared
1891 	 * the VFTA and other state, so if there
1892 	 * have been no vlan's registered do nothing.
1893 	 */
1894 	if (sc->num_vlans == 0)
1895 		return;
1896 
1897 	/* Setup the queues for vlans */
1898 	if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) {
1899 		for (i = 0; i < sc->num_rx_queues; i++) {
1900 			rxr = &sc->rx_queues[i].rxr;
1901 			/* On 82599 the VLAN enable is per/queue in RXDCTL */
1902 			if (hw->mac.type != ixgbe_mac_82598EB) {
1903 				ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me));
1904 				ctrl |= IXGBE_RXDCTL_VME;
1905 				IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), ctrl);
1906 			}
1907 			rxr->vtag_strip = true;
1908 		}
1909 	}
1910 
1911 	if ((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0)
1912 		return;
1913 	/*
1914 	 * A soft reset zero's out the VFTA, so
1915 	 * we need to repopulate it now.
1916 	 */
1917 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1918 		if (sc->shadow_vfta[i] != 0)
1919 			IXGBE_WRITE_REG(hw, IXGBE_VFTA(i),
1920 			    sc->shadow_vfta[i]);
1921 
1922 	ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1923 	/* Enable the Filter Table if enabled */
1924 	if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) {
1925 		ctrl &= ~IXGBE_VLNCTRL_CFIEN;
1926 		ctrl |= IXGBE_VLNCTRL_VFE;
1927 	}
1928 	if (hw->mac.type == ixgbe_mac_82598EB)
1929 		ctrl |= IXGBE_VLNCTRL_VME;
1930 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1931 } /* ixgbe_setup_vlan_hw_support */
1932 
1933 /************************************************************************
1934  * ixgbe_get_slot_info
1935  *
1936  *   Get the width and transaction speed of
1937  *   the slot this adapter is plugged into.
1938  ************************************************************************/
1939 static void
1940 ixgbe_get_slot_info(struct ixgbe_softc *sc)
1941 {
1942 	device_t        dev = iflib_get_dev(sc->ctx);
1943 	struct ixgbe_hw *hw = &sc->hw;
1944 	int             bus_info_valid = true;
1945 	u32             offset;
1946 	u16             link;
1947 
1948 	/* Some devices are behind an internal bridge */
1949 	switch (hw->device_id) {
1950 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
1951 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
1952 		goto get_parent_info;
1953 	default:
1954 		break;
1955 	}
1956 
1957 	ixgbe_get_bus_info(hw);
1958 
1959 	/*
1960 	 * Some devices don't use PCI-E, but there is no need
1961 	 * to display "Unknown" for bus speed and width.
1962 	 */
1963 	switch (hw->mac.type) {
1964 	case ixgbe_mac_X550EM_x:
1965 	case ixgbe_mac_X550EM_a:
1966 		return;
1967 	default:
1968 		goto display;
1969 	}
1970 
1971 get_parent_info:
1972 	/*
1973 	 * For the Quad port adapter we need to parse back
1974 	 * up the PCI tree to find the speed of the expansion
1975 	 * slot into which this adapter is plugged. A bit more work.
1976 	 */
1977 	dev = device_get_parent(device_get_parent(dev));
1978 #ifdef IXGBE_DEBUG
1979 	device_printf(dev, "parent pcib = %x,%x,%x\n", pci_get_bus(dev),
1980 	    pci_get_slot(dev), pci_get_function(dev));
1981 #endif
1982 	dev = device_get_parent(device_get_parent(dev));
1983 #ifdef IXGBE_DEBUG
1984 	device_printf(dev, "slot pcib = %x,%x,%x\n", pci_get_bus(dev),
1985 	    pci_get_slot(dev), pci_get_function(dev));
1986 #endif
1987 	/* Now get the PCI Express Capabilities offset */
1988 	if (pci_find_cap(dev, PCIY_EXPRESS, &offset)) {
1989 		/*
1990 		 * Hmm...can't get PCI-Express capabilities.
1991 		 * Falling back to default method.
1992 		 */
1993 		bus_info_valid = false;
1994 		ixgbe_get_bus_info(hw);
1995 		goto display;
1996 	}
1997 	/* ...and read the Link Status Register */
1998 	link = pci_read_config(dev, offset + PCIER_LINK_STA, 2);
1999 	ixgbe_set_pci_config_data_generic(hw, link);
2000 
2001 display:
2002 	device_printf(dev, "PCI Express Bus: Speed %s %s\n",
2003 	    ((hw->bus.speed == ixgbe_bus_speed_8000)    ? "8.0GT/s"  :
2004 	     (hw->bus.speed == ixgbe_bus_speed_5000)    ? "5.0GT/s"  :
2005 	     (hw->bus.speed == ixgbe_bus_speed_2500)    ? "2.5GT/s"  :
2006 	     "Unknown"),
2007 	    ((hw->bus.width == ixgbe_bus_width_pcie_x8) ? "Width x8" :
2008 	     (hw->bus.width == ixgbe_bus_width_pcie_x4) ? "Width x4" :
2009 	     (hw->bus.width == ixgbe_bus_width_pcie_x1) ? "Width x1" :
2010 	     "Unknown"));
2011 
2012 	if (bus_info_valid) {
2013 		if ((hw->device_id != IXGBE_DEV_ID_82599_SFP_SF_QP) &&
2014 		    ((hw->bus.width <= ixgbe_bus_width_pcie_x4) &&
2015 		    (hw->bus.speed == ixgbe_bus_speed_2500))) {
2016 			device_printf(dev, "PCI-Express bandwidth available for this card\n     is not sufficient for optimal performance.\n");
2017 			device_printf(dev, "For optimal performance a x8 PCIE, or x4 PCIE Gen2 slot is required.\n");
2018 		}
2019 		if ((hw->device_id == IXGBE_DEV_ID_82599_SFP_SF_QP) &&
2020 		    ((hw->bus.width <= ixgbe_bus_width_pcie_x8) &&
2021 		    (hw->bus.speed < ixgbe_bus_speed_8000))) {
2022 			device_printf(dev, "PCI-Express bandwidth available for this card\n     is not sufficient for optimal performance.\n");
2023 			device_printf(dev, "For optimal performance a x8 PCIE Gen3 slot is required.\n");
2024 		}
2025 	} else
2026 		device_printf(dev, "Unable to determine slot speed/width. The speed/width reported are that of the internal switch.\n");
2027 
2028 	return;
2029 } /* ixgbe_get_slot_info */
2030 
2031 /************************************************************************
2032  * ixgbe_if_msix_intr_assign
2033  *
2034  *   Setup MSI-X Interrupt resources and handlers
2035  ************************************************************************/
2036 static int
2037 ixgbe_if_msix_intr_assign(if_ctx_t ctx, int msix)
2038 {
2039 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
2040 	struct ix_rx_queue *rx_que = sc->rx_queues;
2041 	struct ix_tx_queue *tx_que;
2042 	int                error, rid, vector = 0;
2043 	int                cpu_id = 0;
2044 	char               buf[16];
2045 
2046 	/* Admin Que is vector 0*/
2047 	rid = vector + 1;
2048 	for (int i = 0; i < sc->num_rx_queues; i++, vector++, rx_que++) {
2049 		rid = vector + 1;
2050 
2051 		snprintf(buf, sizeof(buf), "rxq%d", i);
2052 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid,
2053 		    IFLIB_INTR_RXTX, ixgbe_msix_que, rx_que, rx_que->rxr.me, buf);
2054 
2055 		if (error) {
2056 			device_printf(iflib_get_dev(ctx),
2057 			    "Failed to allocate que int %d err: %d", i, error);
2058 			sc->num_rx_queues = i + 1;
2059 			goto fail;
2060 		}
2061 
2062 		rx_que->msix = vector;
2063 		if (sc->feat_en & IXGBE_FEATURE_RSS) {
2064 			/*
2065 			 * The queue ID is used as the RSS layer bucket ID.
2066 			 * We look up the queue ID -> RSS CPU ID and select
2067 			 * that.
2068 			 */
2069 			cpu_id = rss_getcpu(i % rss_getnumbuckets());
2070 		} else {
2071 			/*
2072 			 * Bind the MSI-X vector, and thus the
2073 			 * rings to the corresponding cpu.
2074 			 *
2075 			 * This just happens to match the default RSS
2076 			 * round-robin bucket -> queue -> CPU allocation.
2077 			 */
2078 			if (sc->num_rx_queues > 1)
2079 				cpu_id = i;
2080 		}
2081 
2082 	}
2083 	for (int i = 0; i < sc->num_tx_queues; i++) {
2084 		snprintf(buf, sizeof(buf), "txq%d", i);
2085 		tx_que = &sc->tx_queues[i];
2086 		tx_que->msix = i % sc->num_rx_queues;
2087 		iflib_softirq_alloc_generic(ctx,
2088 		    &sc->rx_queues[tx_que->msix].que_irq,
2089 		    IFLIB_INTR_TX, tx_que, tx_que->txr.me, buf);
2090 	}
2091 	rid = vector + 1;
2092 	error = iflib_irq_alloc_generic(ctx, &sc->irq, rid,
2093 	    IFLIB_INTR_ADMIN, ixgbe_msix_link, sc, 0, "aq");
2094 	if (error) {
2095 		device_printf(iflib_get_dev(ctx),
2096 		    "Failed to register admin handler");
2097 		return (error);
2098 	}
2099 
2100 	sc->vector = vector;
2101 
2102 	return (0);
2103 fail:
2104 	iflib_irq_free(ctx, &sc->irq);
2105 	rx_que = sc->rx_queues;
2106 	for (int i = 0; i < sc->num_rx_queues; i++, rx_que++)
2107 		iflib_irq_free(ctx, &rx_que->que_irq);
2108 
2109 	return (error);
2110 } /* ixgbe_if_msix_intr_assign */
2111 
2112 static inline void
2113 ixgbe_perform_aim(struct ixgbe_softc *sc, struct ix_rx_queue *que)
2114 {
2115 	uint32_t newitr = 0;
2116 	struct rx_ring *rxr = &que->rxr;
2117 
2118 	/*
2119 	 * Do Adaptive Interrupt Moderation:
2120 	 *  - Write out last calculated setting
2121 	 *  - Calculate based on average size over
2122 	 *    the last interval.
2123 	 */
2124 	if (que->eitr_setting) {
2125 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EITR(que->msix),
2126 		    que->eitr_setting);
2127 	}
2128 
2129 	que->eitr_setting = 0;
2130 	/* Idle, do nothing */
2131 	if (rxr->bytes == 0) {
2132 		return;
2133 	}
2134 
2135 	if ((rxr->bytes) && (rxr->packets)) {
2136 		newitr = (rxr->bytes / rxr->packets);
2137 	}
2138 
2139 	newitr += 24; /* account for hardware frame, crc */
2140 	/* set an upper boundary */
2141 	newitr = min(newitr, 3000);
2142 
2143 	/* Be nice to the mid range */
2144 	if ((newitr > 300) && (newitr < 1200)) {
2145 		newitr = (newitr / 3);
2146 	} else {
2147 		newitr = (newitr / 2);
2148 	}
2149 
2150 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
2151 		newitr |= newitr << 16;
2152 	} else {
2153 		newitr |= IXGBE_EITR_CNT_WDIS;
2154 	}
2155 
2156 	/* save for next interrupt */
2157 	que->eitr_setting = newitr;
2158 
2159 	/* Reset state */
2160 	rxr->bytes = 0;
2161 	rxr->packets = 0;
2162 
2163 	return;
2164 }
2165 
2166 /*********************************************************************
2167  * ixgbe_msix_que - MSI-X Queue Interrupt Service routine
2168  **********************************************************************/
2169 static int
2170 ixgbe_msix_que(void *arg)
2171 {
2172 	struct ix_rx_queue *que = arg;
2173 	struct ixgbe_softc     *sc = que->sc;
2174 	struct ifnet       *ifp = iflib_get_ifp(que->sc->ctx);
2175 
2176 	/* Protect against spurious interrupts */
2177 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2178 		return (FILTER_HANDLED);
2179 
2180 	ixgbe_disable_queue(sc, que->msix);
2181 	++que->irqs;
2182 
2183 	/* Check for AIM */
2184 	if (sc->enable_aim) {
2185 		ixgbe_perform_aim(sc, que);
2186 	}
2187 
2188 	return (FILTER_SCHEDULE_THREAD);
2189 } /* ixgbe_msix_que */
2190 
2191 /************************************************************************
2192  * ixgbe_media_status - Media Ioctl callback
2193  *
2194  *   Called whenever the user queries the status of
2195  *   the interface using ifconfig.
2196  ************************************************************************/
2197 static void
2198 ixgbe_if_media_status(if_ctx_t ctx, struct ifmediareq * ifmr)
2199 {
2200 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
2201 	struct ixgbe_hw *hw = &sc->hw;
2202 	int             layer;
2203 
2204 	INIT_DEBUGOUT("ixgbe_if_media_status: begin");
2205 
2206 	ifmr->ifm_status = IFM_AVALID;
2207 	ifmr->ifm_active = IFM_ETHER;
2208 
2209 	if (!sc->link_active)
2210 		return;
2211 
2212 	ifmr->ifm_status |= IFM_ACTIVE;
2213 	layer = sc->phy_layer;
2214 
2215 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T ||
2216 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_T ||
2217 	    layer & IXGBE_PHYSICAL_LAYER_100BASE_TX ||
2218 	    layer & IXGBE_PHYSICAL_LAYER_10BASE_T)
2219 		switch (sc->link_speed) {
2220 		case IXGBE_LINK_SPEED_10GB_FULL:
2221 			ifmr->ifm_active |= IFM_10G_T | IFM_FDX;
2222 			break;
2223 		case IXGBE_LINK_SPEED_1GB_FULL:
2224 			ifmr->ifm_active |= IFM_1000_T | IFM_FDX;
2225 			break;
2226 		case IXGBE_LINK_SPEED_100_FULL:
2227 			ifmr->ifm_active |= IFM_100_TX | IFM_FDX;
2228 			break;
2229 		case IXGBE_LINK_SPEED_10_FULL:
2230 			ifmr->ifm_active |= IFM_10_T | IFM_FDX;
2231 			break;
2232 		}
2233 	if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU ||
2234 	    layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA)
2235 		switch (sc->link_speed) {
2236 		case IXGBE_LINK_SPEED_10GB_FULL:
2237 			ifmr->ifm_active |= IFM_10G_TWINAX | IFM_FDX;
2238 			break;
2239 		}
2240 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR)
2241 		switch (sc->link_speed) {
2242 		case IXGBE_LINK_SPEED_10GB_FULL:
2243 			ifmr->ifm_active |= IFM_10G_LR | IFM_FDX;
2244 			break;
2245 		case IXGBE_LINK_SPEED_1GB_FULL:
2246 			ifmr->ifm_active |= IFM_1000_LX | IFM_FDX;
2247 			break;
2248 		}
2249 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LRM)
2250 		switch (sc->link_speed) {
2251 		case IXGBE_LINK_SPEED_10GB_FULL:
2252 			ifmr->ifm_active |= IFM_10G_LRM | IFM_FDX;
2253 			break;
2254 		case IXGBE_LINK_SPEED_1GB_FULL:
2255 			ifmr->ifm_active |= IFM_1000_LX | IFM_FDX;
2256 			break;
2257 		}
2258 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR ||
2259 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX)
2260 		switch (sc->link_speed) {
2261 		case IXGBE_LINK_SPEED_10GB_FULL:
2262 			ifmr->ifm_active |= IFM_10G_SR | IFM_FDX;
2263 			break;
2264 		case IXGBE_LINK_SPEED_1GB_FULL:
2265 			ifmr->ifm_active |= IFM_1000_SX | IFM_FDX;
2266 			break;
2267 		}
2268 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4)
2269 		switch (sc->link_speed) {
2270 		case IXGBE_LINK_SPEED_10GB_FULL:
2271 			ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX;
2272 			break;
2273 		}
2274 	/*
2275 	 * XXX: These need to use the proper media types once
2276 	 * they're added.
2277 	 */
2278 #ifndef IFM_ETH_XTYPE
2279 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
2280 		switch (sc->link_speed) {
2281 		case IXGBE_LINK_SPEED_10GB_FULL:
2282 			ifmr->ifm_active |= IFM_10G_SR | IFM_FDX;
2283 			break;
2284 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2285 			ifmr->ifm_active |= IFM_2500_SX | IFM_FDX;
2286 			break;
2287 		case IXGBE_LINK_SPEED_1GB_FULL:
2288 			ifmr->ifm_active |= IFM_1000_CX | IFM_FDX;
2289 			break;
2290 		}
2291 	else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4 ||
2292 	    layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX ||
2293 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
2294 		switch (sc->link_speed) {
2295 		case IXGBE_LINK_SPEED_10GB_FULL:
2296 			ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX;
2297 			break;
2298 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2299 			ifmr->ifm_active |= IFM_2500_SX | IFM_FDX;
2300 			break;
2301 		case IXGBE_LINK_SPEED_1GB_FULL:
2302 			ifmr->ifm_active |= IFM_1000_CX | IFM_FDX;
2303 			break;
2304 		}
2305 #else
2306 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
2307 		switch (sc->link_speed) {
2308 		case IXGBE_LINK_SPEED_10GB_FULL:
2309 			ifmr->ifm_active |= IFM_10G_KR | IFM_FDX;
2310 			break;
2311 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2312 			ifmr->ifm_active |= IFM_2500_KX | IFM_FDX;
2313 			break;
2314 		case IXGBE_LINK_SPEED_1GB_FULL:
2315 			ifmr->ifm_active |= IFM_1000_KX | IFM_FDX;
2316 			break;
2317 		}
2318 	else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4 ||
2319 	    layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX ||
2320 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
2321 		switch (sc->link_speed) {
2322 		case IXGBE_LINK_SPEED_10GB_FULL:
2323 			ifmr->ifm_active |= IFM_10G_KX4 | IFM_FDX;
2324 			break;
2325 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2326 			ifmr->ifm_active |= IFM_2500_KX | IFM_FDX;
2327 			break;
2328 		case IXGBE_LINK_SPEED_1GB_FULL:
2329 			ifmr->ifm_active |= IFM_1000_KX | IFM_FDX;
2330 			break;
2331 		}
2332 #endif
2333 
2334 	/* If nothing is recognized... */
2335 	if (IFM_SUBTYPE(ifmr->ifm_active) == 0)
2336 		ifmr->ifm_active |= IFM_UNKNOWN;
2337 
2338 	/* Display current flow control setting used on link */
2339 	if (hw->fc.current_mode == ixgbe_fc_rx_pause ||
2340 	    hw->fc.current_mode == ixgbe_fc_full)
2341 		ifmr->ifm_active |= IFM_ETH_RXPAUSE;
2342 	if (hw->fc.current_mode == ixgbe_fc_tx_pause ||
2343 	    hw->fc.current_mode == ixgbe_fc_full)
2344 		ifmr->ifm_active |= IFM_ETH_TXPAUSE;
2345 } /* ixgbe_media_status */
2346 
2347 /************************************************************************
2348  * ixgbe_media_change - Media Ioctl callback
2349  *
2350  *   Called when the user changes speed/duplex using
2351  *   media/mediopt option with ifconfig.
2352  ************************************************************************/
2353 static int
2354 ixgbe_if_media_change(if_ctx_t ctx)
2355 {
2356 	struct ixgbe_softc   *sc = iflib_get_softc(ctx);
2357 	struct ifmedia   *ifm = iflib_get_media(ctx);
2358 	struct ixgbe_hw  *hw = &sc->hw;
2359 	ixgbe_link_speed speed = 0;
2360 
2361 	INIT_DEBUGOUT("ixgbe_if_media_change: begin");
2362 
2363 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
2364 		return (EINVAL);
2365 
2366 	if (hw->phy.media_type == ixgbe_media_type_backplane)
2367 		return (EPERM);
2368 
2369 	/*
2370 	 * We don't actually need to check against the supported
2371 	 * media types of the adapter; ifmedia will take care of
2372 	 * that for us.
2373 	 */
2374 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
2375 	case IFM_AUTO:
2376 	case IFM_10G_T:
2377 		speed |= IXGBE_LINK_SPEED_100_FULL;
2378 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2379 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
2380 		break;
2381 	case IFM_10G_LRM:
2382 	case IFM_10G_LR:
2383 #ifndef IFM_ETH_XTYPE
2384 	case IFM_10G_SR: /* KR, too */
2385 	case IFM_10G_CX4: /* KX4 */
2386 #else
2387 	case IFM_10G_KR:
2388 	case IFM_10G_KX4:
2389 #endif
2390 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2391 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
2392 		break;
2393 #ifndef IFM_ETH_XTYPE
2394 	case IFM_1000_CX: /* KX */
2395 #else
2396 	case IFM_1000_KX:
2397 #endif
2398 	case IFM_1000_LX:
2399 	case IFM_1000_SX:
2400 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2401 		break;
2402 	case IFM_1000_T:
2403 		speed |= IXGBE_LINK_SPEED_100_FULL;
2404 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2405 		break;
2406 	case IFM_10G_TWINAX:
2407 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
2408 		break;
2409 	case IFM_100_TX:
2410 		speed |= IXGBE_LINK_SPEED_100_FULL;
2411 		break;
2412 	case IFM_10_T:
2413 		speed |= IXGBE_LINK_SPEED_10_FULL;
2414 		break;
2415 	default:
2416 		goto invalid;
2417 	}
2418 
2419 	hw->mac.autotry_restart = true;
2420 	hw->mac.ops.setup_link(hw, speed, true);
2421 	sc->advertise =
2422 	    ((speed & IXGBE_LINK_SPEED_10GB_FULL) ? 4 : 0) |
2423 	    ((speed & IXGBE_LINK_SPEED_1GB_FULL)  ? 2 : 0) |
2424 	    ((speed & IXGBE_LINK_SPEED_100_FULL)  ? 1 : 0) |
2425 	    ((speed & IXGBE_LINK_SPEED_10_FULL)   ? 8 : 0);
2426 
2427 	return (0);
2428 
2429 invalid:
2430 	device_printf(iflib_get_dev(ctx), "Invalid media type!\n");
2431 
2432 	return (EINVAL);
2433 } /* ixgbe_if_media_change */
2434 
2435 /************************************************************************
2436  * ixgbe_set_promisc
2437  ************************************************************************/
2438 static int
2439 ixgbe_if_promisc_set(if_ctx_t ctx, int flags)
2440 {
2441 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2442 	struct ifnet   *ifp = iflib_get_ifp(ctx);
2443 	u32            rctl;
2444 	int            mcnt = 0;
2445 
2446 	rctl = IXGBE_READ_REG(&sc->hw, IXGBE_FCTRL);
2447 	rctl &= (~IXGBE_FCTRL_UPE);
2448 	if (ifp->if_flags & IFF_ALLMULTI)
2449 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
2450 	else {
2451 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
2452 	}
2453 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
2454 		rctl &= (~IXGBE_FCTRL_MPE);
2455 	IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, rctl);
2456 
2457 	if (ifp->if_flags & IFF_PROMISC) {
2458 		rctl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
2459 		IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, rctl);
2460 	} else if (ifp->if_flags & IFF_ALLMULTI) {
2461 		rctl |= IXGBE_FCTRL_MPE;
2462 		rctl &= ~IXGBE_FCTRL_UPE;
2463 		IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, rctl);
2464 	}
2465 	return (0);
2466 } /* ixgbe_if_promisc_set */
2467 
2468 /************************************************************************
2469  * ixgbe_msix_link - Link status change ISR (MSI/MSI-X)
2470  ************************************************************************/
2471 static int
2472 ixgbe_msix_link(void *arg)
2473 {
2474 	struct ixgbe_softc  *sc = arg;
2475 	struct ixgbe_hw *hw = &sc->hw;
2476 	u32             eicr, eicr_mask;
2477 	s32             retval;
2478 
2479 	++sc->link_irq;
2480 
2481 	/* Pause other interrupts */
2482 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_OTHER);
2483 
2484 	/* First get the cause */
2485 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2486 	/* Be sure the queue bits are not cleared */
2487 	eicr &= ~IXGBE_EICR_RTX_QUEUE;
2488 	/* Clear interrupt with write */
2489 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2490 
2491 	/* Link status change */
2492 	if (eicr & IXGBE_EICR_LSC) {
2493 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2494 		sc->task_requests |= IXGBE_REQUEST_TASK_LSC;
2495 	}
2496 
2497 	if (sc->hw.mac.type != ixgbe_mac_82598EB) {
2498 		if ((sc->feat_en & IXGBE_FEATURE_FDIR) &&
2499 		    (eicr & IXGBE_EICR_FLOW_DIR)) {
2500 			/* This is probably overkill :) */
2501 			if (!atomic_cmpset_int(&sc->fdir_reinit, 0, 1))
2502 				return (FILTER_HANDLED);
2503 			/* Disable the interrupt */
2504 			IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EICR_FLOW_DIR);
2505 			sc->task_requests |= IXGBE_REQUEST_TASK_FDIR;
2506 		} else
2507 			if (eicr & IXGBE_EICR_ECC) {
2508 				device_printf(iflib_get_dev(sc->ctx),
2509 				   "\nCRITICAL: ECC ERROR!! Please Reboot!!\n");
2510 				IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2511 			}
2512 
2513 		/* Check for over temp condition */
2514 		if (sc->feat_en & IXGBE_FEATURE_TEMP_SENSOR) {
2515 			switch (sc->hw.mac.type) {
2516 			case ixgbe_mac_X550EM_a:
2517 				if (!(eicr & IXGBE_EICR_GPI_SDP0_X550EM_a))
2518 					break;
2519 				IXGBE_WRITE_REG(hw, IXGBE_EIMC,
2520 				    IXGBE_EICR_GPI_SDP0_X550EM_a);
2521 				IXGBE_WRITE_REG(hw, IXGBE_EICR,
2522 				    IXGBE_EICR_GPI_SDP0_X550EM_a);
2523 				retval = hw->phy.ops.check_overtemp(hw);
2524 				if (retval != IXGBE_ERR_OVERTEMP)
2525 					break;
2526 				device_printf(iflib_get_dev(sc->ctx),
2527 				    "\nCRITICAL: OVER TEMP!! PHY IS SHUT DOWN!!\n");
2528 				device_printf(iflib_get_dev(sc->ctx),
2529 				    "System shutdown required!\n");
2530 				break;
2531 			default:
2532 				if (!(eicr & IXGBE_EICR_TS))
2533 					break;
2534 				retval = hw->phy.ops.check_overtemp(hw);
2535 				if (retval != IXGBE_ERR_OVERTEMP)
2536 					break;
2537 				device_printf(iflib_get_dev(sc->ctx),
2538 				    "\nCRITICAL: OVER TEMP!! PHY IS SHUT DOWN!!\n");
2539 				device_printf(iflib_get_dev(sc->ctx),
2540 				    "System shutdown required!\n");
2541 				IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_TS);
2542 				break;
2543 			}
2544 		}
2545 
2546 		/* Check for VF message */
2547 		if ((sc->feat_en & IXGBE_FEATURE_SRIOV) &&
2548 		    (eicr & IXGBE_EICR_MAILBOX))
2549 			sc->task_requests |= IXGBE_REQUEST_TASK_MBX;
2550 	}
2551 
2552 	if (ixgbe_is_sfp(hw)) {
2553 		/* Pluggable optics-related interrupt */
2554 		if (hw->mac.type >= ixgbe_mac_X540)
2555 			eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2556 		else
2557 			eicr_mask = IXGBE_EICR_GPI_SDP2_BY_MAC(hw);
2558 
2559 		if (eicr & eicr_mask) {
2560 			IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2561 			sc->task_requests |= IXGBE_REQUEST_TASK_MOD;
2562 		}
2563 
2564 		if ((hw->mac.type == ixgbe_mac_82599EB) &&
2565 		    (eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw))) {
2566 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2567 			    IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
2568 			sc->task_requests |= IXGBE_REQUEST_TASK_MSF;
2569 		}
2570 	}
2571 
2572 	/* Check for fan failure */
2573 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL) {
2574 		ixgbe_check_fan_failure(sc, eicr, true);
2575 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
2576 	}
2577 
2578 	/* External PHY interrupt */
2579 	if ((hw->phy.type == ixgbe_phy_x550em_ext_t) &&
2580 	    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2581 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP0_X540);
2582 		sc->task_requests |= IXGBE_REQUEST_TASK_PHY;
2583 	}
2584 
2585 	return (sc->task_requests != 0) ? FILTER_SCHEDULE_THREAD : FILTER_HANDLED;
2586 } /* ixgbe_msix_link */
2587 
2588 /************************************************************************
2589  * ixgbe_sysctl_interrupt_rate_handler
2590  ************************************************************************/
2591 static int
2592 ixgbe_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS)
2593 {
2594 	struct ix_rx_queue *que = ((struct ix_rx_queue *)oidp->oid_arg1);
2595 	int                error;
2596 	unsigned int       reg, usec, rate;
2597 
2598 	reg = IXGBE_READ_REG(&que->sc->hw, IXGBE_EITR(que->msix));
2599 	usec = ((reg & 0x0FF8) >> 3);
2600 	if (usec > 0)
2601 		rate = 500000 / usec;
2602 	else
2603 		rate = 0;
2604 	error = sysctl_handle_int(oidp, &rate, 0, req);
2605 	if (error || !req->newptr)
2606 		return error;
2607 	reg &= ~0xfff; /* default, no limitation */
2608 	ixgbe_max_interrupt_rate = 0;
2609 	if (rate > 0 && rate < 500000) {
2610 		if (rate < 1000)
2611 			rate = 1000;
2612 		ixgbe_max_interrupt_rate = rate;
2613 		reg |= ((4000000/rate) & 0xff8);
2614 	}
2615 	IXGBE_WRITE_REG(&que->sc->hw, IXGBE_EITR(que->msix), reg);
2616 
2617 	return (0);
2618 } /* ixgbe_sysctl_interrupt_rate_handler */
2619 
2620 /************************************************************************
2621  * ixgbe_add_device_sysctls
2622  ************************************************************************/
2623 static void
2624 ixgbe_add_device_sysctls(if_ctx_t ctx)
2625 {
2626 	struct ixgbe_softc         *sc = iflib_get_softc(ctx);
2627 	device_t               dev = iflib_get_dev(ctx);
2628 	struct ixgbe_hw        *hw = &sc->hw;
2629 	struct sysctl_oid_list *child;
2630 	struct sysctl_ctx_list *ctx_list;
2631 
2632 	ctx_list = device_get_sysctl_ctx(dev);
2633 	child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
2634 
2635 	/* Sysctls for all devices */
2636 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc",
2637 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2638 	    sc, 0, ixgbe_sysctl_flowcntl, "I",
2639 	    IXGBE_SYSCTL_DESC_SET_FC);
2640 
2641 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "advertise_speed",
2642 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2643 	    sc, 0, ixgbe_sysctl_advertise, "I",
2644 	    IXGBE_SYSCTL_DESC_ADV_SPEED);
2645 
2646 	sc->enable_aim = ixgbe_enable_aim;
2647 	SYSCTL_ADD_INT(ctx_list, child, OID_AUTO, "enable_aim", CTLFLAG_RW,
2648 	    &sc->enable_aim, 0, "Interrupt Moderation");
2649 
2650 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fw_version",
2651 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
2652 	    ixgbe_sysctl_print_fw_version, "A", "Prints FW/NVM Versions");
2653 
2654 #ifdef IXGBE_DEBUG
2655 	/* testing sysctls (for all devices) */
2656 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "power_state",
2657 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2658 	    sc, 0, ixgbe_sysctl_power_state,
2659 	    "I", "PCI Power State");
2660 
2661 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "print_rss_config",
2662 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
2663 	    ixgbe_sysctl_print_rss_config, "A", "Prints RSS Configuration");
2664 #endif
2665 	/* for X550 series devices */
2666 	if (hw->mac.type >= ixgbe_mac_X550)
2667 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "dmac",
2668 		    CTLTYPE_U16 | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2669 		    sc, 0, ixgbe_sysctl_dmac,
2670 		    "I", "DMA Coalesce");
2671 
2672 	/* for WoL-capable devices */
2673 	if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) {
2674 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "wol_enable",
2675 		    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
2676 		    ixgbe_sysctl_wol_enable, "I", "Enable/Disable Wake on LAN");
2677 
2678 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "wufc",
2679 		    CTLTYPE_U32 | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2680 		    sc, 0, ixgbe_sysctl_wufc,
2681 		    "I", "Enable/Disable Wake Up Filters");
2682 	}
2683 
2684 	/* for X552/X557-AT devices */
2685 	if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) {
2686 		struct sysctl_oid *phy_node;
2687 		struct sysctl_oid_list *phy_list;
2688 
2689 		phy_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, "phy",
2690 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "External PHY sysctls");
2691 		phy_list = SYSCTL_CHILDREN(phy_node);
2692 
2693 		SYSCTL_ADD_PROC(ctx_list, phy_list, OID_AUTO, "temp",
2694 		    CTLTYPE_U16 | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
2695 		    sc, 0, ixgbe_sysctl_phy_temp,
2696 		    "I", "Current External PHY Temperature (Celsius)");
2697 
2698 		SYSCTL_ADD_PROC(ctx_list, phy_list, OID_AUTO,
2699 		    "overtemp_occurred",
2700 		    CTLTYPE_U16 | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
2701 		    ixgbe_sysctl_phy_overtemp_occurred, "I",
2702 		    "External PHY High Temperature Event Occurred");
2703 	}
2704 
2705 	if (sc->feat_cap & IXGBE_FEATURE_EEE) {
2706 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "eee_state",
2707 		    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
2708 		    ixgbe_sysctl_eee_state, "I", "EEE Power Save State");
2709 	}
2710 } /* ixgbe_add_device_sysctls */
2711 
2712 /************************************************************************
2713  * ixgbe_allocate_pci_resources
2714  ************************************************************************/
2715 static int
2716 ixgbe_allocate_pci_resources(if_ctx_t ctx)
2717 {
2718 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2719 	device_t        dev = iflib_get_dev(ctx);
2720 	int             rid;
2721 
2722 	rid = PCIR_BAR(0);
2723 	sc->pci_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
2724 	    RF_ACTIVE);
2725 
2726 	if (!(sc->pci_mem)) {
2727 		device_printf(dev, "Unable to allocate bus resource: memory\n");
2728 		return (ENXIO);
2729 	}
2730 
2731 	/* Save bus_space values for READ/WRITE_REG macros */
2732 	sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->pci_mem);
2733 	sc->osdep.mem_bus_space_handle =
2734 	    rman_get_bushandle(sc->pci_mem);
2735 	/* Set hw values for shared code */
2736 	sc->hw.hw_addr = (u8 *)&sc->osdep.mem_bus_space_handle;
2737 
2738 	return (0);
2739 } /* ixgbe_allocate_pci_resources */
2740 
2741 /************************************************************************
2742  * ixgbe_detach - Device removal routine
2743  *
2744  *   Called when the driver is being removed.
2745  *   Stops the adapter and deallocates all the resources
2746  *   that were allocated for driver operation.
2747  *
2748  *   return 0 on success, positive on failure
2749  ************************************************************************/
2750 static int
2751 ixgbe_if_detach(if_ctx_t ctx)
2752 {
2753 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2754 	device_t       dev = iflib_get_dev(ctx);
2755 	u32            ctrl_ext;
2756 
2757 	INIT_DEBUGOUT("ixgbe_detach: begin");
2758 
2759 	if (ixgbe_pci_iov_detach(dev) != 0) {
2760 		device_printf(dev, "SR-IOV in use; detach first.\n");
2761 		return (EBUSY);
2762 	}
2763 
2764 	ixgbe_setup_low_power_mode(ctx);
2765 
2766 	/* let hardware know driver is unloading */
2767 	ctrl_ext = IXGBE_READ_REG(&sc->hw, IXGBE_CTRL_EXT);
2768 	ctrl_ext &= ~IXGBE_CTRL_EXT_DRV_LOAD;
2769 	IXGBE_WRITE_REG(&sc->hw, IXGBE_CTRL_EXT, ctrl_ext);
2770 
2771 	ixgbe_free_pci_resources(ctx);
2772 	free(sc->mta, M_IXGBE);
2773 
2774 	return (0);
2775 } /* ixgbe_if_detach */
2776 
2777 /************************************************************************
2778  * ixgbe_setup_low_power_mode - LPLU/WoL preparation
2779  *
2780  *   Prepare the adapter/port for LPLU and/or WoL
2781  ************************************************************************/
2782 static int
2783 ixgbe_setup_low_power_mode(if_ctx_t ctx)
2784 {
2785 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
2786 	struct ixgbe_hw *hw = &sc->hw;
2787 	device_t        dev = iflib_get_dev(ctx);
2788 	s32             error = 0;
2789 
2790 	if (!hw->wol_enabled)
2791 		ixgbe_set_phy_power(hw, false);
2792 
2793 	/* Limit power management flow to X550EM baseT */
2794 	if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T &&
2795 	    hw->phy.ops.enter_lplu) {
2796 		/* Turn off support for APM wakeup. (Using ACPI instead) */
2797 		IXGBE_WRITE_REG(hw, IXGBE_GRC,
2798 		    IXGBE_READ_REG(hw, IXGBE_GRC) & ~(u32)2);
2799 
2800 		/*
2801 		 * Clear Wake Up Status register to prevent any previous wakeup
2802 		 * events from waking us up immediately after we suspend.
2803 		 */
2804 		IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff);
2805 
2806 		/*
2807 		 * Program the Wakeup Filter Control register with user filter
2808 		 * settings
2809 		 */
2810 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, sc->wufc);
2811 
2812 		/* Enable wakeups and power management in Wakeup Control */
2813 		IXGBE_WRITE_REG(hw, IXGBE_WUC,
2814 		    IXGBE_WUC_WKEN | IXGBE_WUC_PME_EN);
2815 
2816 		/* X550EM baseT adapters need a special LPLU flow */
2817 		hw->phy.reset_disable = true;
2818 		ixgbe_if_stop(ctx);
2819 		error = hw->phy.ops.enter_lplu(hw);
2820 		if (error)
2821 			device_printf(dev, "Error entering LPLU: %d\n", error);
2822 		hw->phy.reset_disable = false;
2823 	} else {
2824 		/* Just stop for other adapters */
2825 		ixgbe_if_stop(ctx);
2826 	}
2827 
2828 	return error;
2829 } /* ixgbe_setup_low_power_mode */
2830 
2831 /************************************************************************
2832  * ixgbe_shutdown - Shutdown entry point
2833  ************************************************************************/
2834 static int
2835 ixgbe_if_shutdown(if_ctx_t ctx)
2836 {
2837 	int error = 0;
2838 
2839 	INIT_DEBUGOUT("ixgbe_shutdown: begin");
2840 
2841 	error = ixgbe_setup_low_power_mode(ctx);
2842 
2843 	return (error);
2844 } /* ixgbe_if_shutdown */
2845 
2846 /************************************************************************
2847  * ixgbe_suspend
2848  *
2849  *   From D0 to D3
2850  ************************************************************************/
2851 static int
2852 ixgbe_if_suspend(if_ctx_t ctx)
2853 {
2854 	int error = 0;
2855 
2856 	INIT_DEBUGOUT("ixgbe_suspend: begin");
2857 
2858 	error = ixgbe_setup_low_power_mode(ctx);
2859 
2860 	return (error);
2861 } /* ixgbe_if_suspend */
2862 
2863 /************************************************************************
2864  * ixgbe_resume
2865  *
2866  *   From D3 to D0
2867  ************************************************************************/
2868 static int
2869 ixgbe_if_resume(if_ctx_t ctx)
2870 {
2871 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
2872 	device_t        dev = iflib_get_dev(ctx);
2873 	struct ifnet    *ifp = iflib_get_ifp(ctx);
2874 	struct ixgbe_hw *hw = &sc->hw;
2875 	u32             wus;
2876 
2877 	INIT_DEBUGOUT("ixgbe_resume: begin");
2878 
2879 	/* Read & clear WUS register */
2880 	wus = IXGBE_READ_REG(hw, IXGBE_WUS);
2881 	if (wus)
2882 		device_printf(dev, "Woken up by (WUS): %#010x\n",
2883 		    IXGBE_READ_REG(hw, IXGBE_WUS));
2884 	IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff);
2885 	/* And clear WUFC until next low-power transition */
2886 	IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
2887 
2888 	/*
2889 	 * Required after D3->D0 transition;
2890 	 * will re-advertise all previous advertised speeds
2891 	 */
2892 	if (ifp->if_flags & IFF_UP)
2893 		ixgbe_if_init(ctx);
2894 
2895 	return (0);
2896 } /* ixgbe_if_resume */
2897 
2898 /************************************************************************
2899  * ixgbe_if_mtu_set - Ioctl mtu entry point
2900  *
2901  *   Return 0 on success, EINVAL on failure
2902  ************************************************************************/
2903 static int
2904 ixgbe_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
2905 {
2906 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2907 	int error = 0;
2908 
2909 	IOCTL_DEBUGOUT("ioctl: SIOCIFMTU (Set Interface MTU)");
2910 
2911 	if (mtu > IXGBE_MAX_MTU) {
2912 		error = EINVAL;
2913 	} else {
2914 		sc->max_frame_size = mtu + IXGBE_MTU_HDR;
2915 	}
2916 
2917 	return error;
2918 } /* ixgbe_if_mtu_set */
2919 
2920 /************************************************************************
2921  * ixgbe_if_crcstrip_set
2922  ************************************************************************/
2923 static void
2924 ixgbe_if_crcstrip_set(if_ctx_t ctx, int onoff, int crcstrip)
2925 {
2926 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2927 	struct ixgbe_hw *hw = &sc->hw;
2928 	/* crc stripping is set in two places:
2929 	 * IXGBE_HLREG0 (modified on init_locked and hw reset)
2930 	 * IXGBE_RDRXCTL (set by the original driver in
2931 	 *	ixgbe_setup_hw_rsc() called in init_locked.
2932 	 *	We disable the setting when netmap is compiled in).
2933 	 * We update the values here, but also in ixgbe.c because
2934 	 * init_locked sometimes is called outside our control.
2935 	 */
2936 	uint32_t hl, rxc;
2937 
2938 	hl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
2939 	rxc = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
2940 #ifdef NETMAP
2941 	if (netmap_verbose)
2942 		D("%s read  HLREG 0x%x rxc 0x%x",
2943 			onoff ? "enter" : "exit", hl, rxc);
2944 #endif
2945 	/* hw requirements ... */
2946 	rxc &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
2947 	rxc |= IXGBE_RDRXCTL_RSCACKC;
2948 	if (onoff && !crcstrip) {
2949 		/* keep the crc. Fast rx */
2950 		hl &= ~IXGBE_HLREG0_RXCRCSTRP;
2951 		rxc &= ~IXGBE_RDRXCTL_CRCSTRIP;
2952 	} else {
2953 		/* reset default mode */
2954 		hl |= IXGBE_HLREG0_RXCRCSTRP;
2955 		rxc |= IXGBE_RDRXCTL_CRCSTRIP;
2956 	}
2957 #ifdef NETMAP
2958 	if (netmap_verbose)
2959 		D("%s write HLREG 0x%x rxc 0x%x",
2960 			onoff ? "enter" : "exit", hl, rxc);
2961 #endif
2962 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hl);
2963 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rxc);
2964 } /* ixgbe_if_crcstrip_set */
2965 
2966 /*********************************************************************
2967  * ixgbe_if_init - Init entry point
2968  *
2969  *   Used in two ways: It is used by the stack as an init
2970  *   entry point in network interface structure. It is also
2971  *   used by the driver as a hw/sw initialization routine to
2972  *   get to a consistent state.
2973  *
2974  *   Return 0 on success, positive on failure
2975  **********************************************************************/
2976 void
2977 ixgbe_if_init(if_ctx_t ctx)
2978 {
2979 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
2980 	struct ifnet       *ifp = iflib_get_ifp(ctx);
2981 	device_t           dev = iflib_get_dev(ctx);
2982 	struct ixgbe_hw *hw = &sc->hw;
2983 	struct ix_rx_queue *rx_que;
2984 	struct ix_tx_queue *tx_que;
2985 	u32             txdctl, mhadd;
2986 	u32             rxdctl, rxctrl;
2987 	u32             ctrl_ext;
2988 
2989 	int             i, j, err;
2990 
2991 	INIT_DEBUGOUT("ixgbe_if_init: begin");
2992 
2993 	/* Queue indices may change with IOV mode */
2994 	ixgbe_align_all_queue_indices(sc);
2995 
2996 	/* reprogram the RAR[0] in case user changed it. */
2997 	ixgbe_set_rar(hw, 0, hw->mac.addr, sc->pool, IXGBE_RAH_AV);
2998 
2999 	/* Get the latest mac address, User can use a LAA */
3000 	bcopy(IF_LLADDR(ifp), hw->mac.addr, IXGBE_ETH_LENGTH_OF_ADDRESS);
3001 	ixgbe_set_rar(hw, 0, hw->mac.addr, sc->pool, 1);
3002 	hw->addr_ctrl.rar_used_count = 1;
3003 
3004 	ixgbe_init_hw(hw);
3005 
3006 	ixgbe_initialize_iov(sc);
3007 
3008 	ixgbe_initialize_transmit_units(ctx);
3009 
3010 	/* Setup Multicast table */
3011 	ixgbe_if_multi_set(ctx);
3012 
3013 	/* Determine the correct mbuf pool, based on frame size */
3014 	sc->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
3015 
3016 	/* Configure RX settings */
3017 	ixgbe_initialize_receive_units(ctx);
3018 
3019 	/*
3020 	 * Initialize variable holding task enqueue requests
3021 	 * from MSI-X interrupts
3022 	 */
3023 	sc->task_requests = 0;
3024 
3025 	/* Enable SDP & MSI-X interrupts based on adapter */
3026 	ixgbe_config_gpie(sc);
3027 
3028 	/* Set MTU size */
3029 	if (ifp->if_mtu > ETHERMTU) {
3030 		/* aka IXGBE_MAXFRS on 82599 and newer */
3031 		mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3032 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3033 		mhadd |= sc->max_frame_size << IXGBE_MHADD_MFS_SHIFT;
3034 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3035 	}
3036 
3037 	/* Now enable all the queues */
3038 	for (i = 0, tx_que = sc->tx_queues; i < sc->num_tx_queues; i++, tx_que++) {
3039 		struct tx_ring *txr = &tx_que->txr;
3040 
3041 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(txr->me));
3042 		txdctl |= IXGBE_TXDCTL_ENABLE;
3043 		/* Set WTHRESH to 8, burst writeback */
3044 		txdctl |= (8 << 16);
3045 		/*
3046 		 * When the internal queue falls below PTHRESH (32),
3047 		 * start prefetching as long as there are at least
3048 		 * HTHRESH (1) buffers ready. The values are taken
3049 		 * from the Intel linux driver 3.8.21.
3050 		 * Prefetching enables tx line rate even with 1 queue.
3051 		 */
3052 		txdctl |= (32 << 0) | (1 << 8);
3053 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(txr->me), txdctl);
3054 	}
3055 
3056 	for (i = 0, rx_que = sc->rx_queues; i < sc->num_rx_queues; i++, rx_que++) {
3057 		struct rx_ring *rxr = &rx_que->rxr;
3058 
3059 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me));
3060 		if (hw->mac.type == ixgbe_mac_82598EB) {
3061 			/*
3062 			 * PTHRESH = 21
3063 			 * HTHRESH = 4
3064 			 * WTHRESH = 8
3065 			 */
3066 			rxdctl &= ~0x3FFFFF;
3067 			rxdctl |= 0x080420;
3068 		}
3069 		rxdctl |= IXGBE_RXDCTL_ENABLE;
3070 		IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), rxdctl);
3071 		for (j = 0; j < 10; j++) {
3072 			if (IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me)) &
3073 			    IXGBE_RXDCTL_ENABLE)
3074 				break;
3075 			else
3076 				msec_delay(1);
3077 		}
3078 		wmb();
3079 	}
3080 
3081 	/* Enable Receive engine */
3082 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3083 	if (hw->mac.type == ixgbe_mac_82598EB)
3084 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3085 	rxctrl |= IXGBE_RXCTRL_RXEN;
3086 	ixgbe_enable_rx_dma(hw, rxctrl);
3087 
3088 	/* Set up MSI/MSI-X routing */
3089 	if (ixgbe_enable_msix)  {
3090 		ixgbe_configure_ivars(sc);
3091 		/* Set up auto-mask */
3092 		if (hw->mac.type == ixgbe_mac_82598EB)
3093 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3094 		else {
3095 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
3096 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
3097 		}
3098 	} else {  /* Simple settings for Legacy/MSI */
3099 		ixgbe_set_ivar(sc, 0, 0, 0);
3100 		ixgbe_set_ivar(sc, 0, 0, 1);
3101 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3102 	}
3103 
3104 	ixgbe_init_fdir(sc);
3105 
3106 	/*
3107 	 * Check on any SFP devices that
3108 	 * need to be kick-started
3109 	 */
3110 	if (hw->phy.type == ixgbe_phy_none) {
3111 		err = hw->phy.ops.identify(hw);
3112 		if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3113 			device_printf(dev,
3114 			    "Unsupported SFP+ module type was detected.\n");
3115 			return;
3116 		}
3117 	}
3118 
3119 	/* Set moderation on the Link interrupt */
3120 	IXGBE_WRITE_REG(hw, IXGBE_EITR(sc->vector), IXGBE_LINK_ITR);
3121 
3122 	/* Enable power to the phy. */
3123 	ixgbe_set_phy_power(hw, true);
3124 
3125 	/* Config/Enable Link */
3126 	ixgbe_config_link(ctx);
3127 
3128 	/* Hardware Packet Buffer & Flow Control setup */
3129 	ixgbe_config_delay_values(sc);
3130 
3131 	/* Initialize the FC settings */
3132 	ixgbe_start_hw(hw);
3133 
3134 	/* Set up VLAN support and filter */
3135 	ixgbe_setup_vlan_hw_support(ctx);
3136 
3137 	/* Setup DMA Coalescing */
3138 	ixgbe_config_dmac(sc);
3139 
3140 	/* And now turn on interrupts */
3141 	ixgbe_if_enable_intr(ctx);
3142 
3143 	/* Enable the use of the MBX by the VF's */
3144 	if (sc->feat_en & IXGBE_FEATURE_SRIOV) {
3145 		ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
3146 		ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
3147 		IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
3148 	}
3149 
3150 } /* ixgbe_init_locked */
3151 
3152 /************************************************************************
3153  * ixgbe_set_ivar
3154  *
3155  *   Setup the correct IVAR register for a particular MSI-X interrupt
3156  *     (yes this is all very magic and confusing :)
3157  *    - entry is the register array entry
3158  *    - vector is the MSI-X vector for this queue
3159  *    - type is RX/TX/MISC
3160  ************************************************************************/
3161 static void
3162 ixgbe_set_ivar(struct ixgbe_softc *sc, u8 entry, u8 vector, s8 type)
3163 {
3164 	struct ixgbe_hw *hw = &sc->hw;
3165 	u32 ivar, index;
3166 
3167 	vector |= IXGBE_IVAR_ALLOC_VAL;
3168 
3169 	switch (hw->mac.type) {
3170 	case ixgbe_mac_82598EB:
3171 		if (type == -1)
3172 			entry = IXGBE_IVAR_OTHER_CAUSES_INDEX;
3173 		else
3174 			entry += (type * 64);
3175 		index = (entry >> 2) & 0x1F;
3176 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
3177 		ivar &= ~(0xFF << (8 * (entry & 0x3)));
3178 		ivar |= (vector << (8 * (entry & 0x3)));
3179 		IXGBE_WRITE_REG(&sc->hw, IXGBE_IVAR(index), ivar);
3180 		break;
3181 	case ixgbe_mac_82599EB:
3182 	case ixgbe_mac_X540:
3183 	case ixgbe_mac_X550:
3184 	case ixgbe_mac_X550EM_x:
3185 	case ixgbe_mac_X550EM_a:
3186 		if (type == -1) { /* MISC IVAR */
3187 			index = (entry & 1) * 8;
3188 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
3189 			ivar &= ~(0xFF << index);
3190 			ivar |= (vector << index);
3191 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, ivar);
3192 		} else {          /* RX/TX IVARS */
3193 			index = (16 * (entry & 1)) + (8 * type);
3194 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(entry >> 1));
3195 			ivar &= ~(0xFF << index);
3196 			ivar |= (vector << index);
3197 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(entry >> 1), ivar);
3198 		}
3199 	default:
3200 		break;
3201 	}
3202 } /* ixgbe_set_ivar */
3203 
3204 /************************************************************************
3205  * ixgbe_configure_ivars
3206  ************************************************************************/
3207 static void
3208 ixgbe_configure_ivars(struct ixgbe_softc *sc)
3209 {
3210 	struct ix_rx_queue *rx_que = sc->rx_queues;
3211 	struct ix_tx_queue *tx_que = sc->tx_queues;
3212 	u32                newitr;
3213 
3214 	if (ixgbe_max_interrupt_rate > 0)
3215 		newitr = (4000000 / ixgbe_max_interrupt_rate) & 0x0FF8;
3216 	else {
3217 		/*
3218 		 * Disable DMA coalescing if interrupt moderation is
3219 		 * disabled.
3220 		 */
3221 		sc->dmac = 0;
3222 		newitr = 0;
3223 	}
3224 
3225 	for (int i = 0; i < sc->num_rx_queues; i++, rx_que++) {
3226 		struct rx_ring *rxr = &rx_que->rxr;
3227 
3228 		/* First the RX queue entry */
3229 		ixgbe_set_ivar(sc, rxr->me, rx_que->msix, 0);
3230 
3231 		/* Set an Initial EITR value */
3232 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EITR(rx_que->msix), newitr);
3233 	}
3234 	for (int i = 0; i < sc->num_tx_queues; i++, tx_que++) {
3235 		struct tx_ring *txr = &tx_que->txr;
3236 
3237 		/* ... and the TX */
3238 		ixgbe_set_ivar(sc, txr->me, tx_que->msix, 1);
3239 	}
3240 	/* For the Link interrupt */
3241 	ixgbe_set_ivar(sc, 1, sc->vector, -1);
3242 } /* ixgbe_configure_ivars */
3243 
3244 /************************************************************************
3245  * ixgbe_config_gpie
3246  ************************************************************************/
3247 static void
3248 ixgbe_config_gpie(struct ixgbe_softc *sc)
3249 {
3250 	struct ixgbe_hw *hw = &sc->hw;
3251 	u32             gpie;
3252 
3253 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
3254 
3255 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3256 		/* Enable Enhanced MSI-X mode */
3257 		gpie |= IXGBE_GPIE_MSIX_MODE
3258 		     |  IXGBE_GPIE_EIAME
3259 		     |  IXGBE_GPIE_PBA_SUPPORT
3260 		     |  IXGBE_GPIE_OCD;
3261 	}
3262 
3263 	/* Fan Failure Interrupt */
3264 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL)
3265 		gpie |= IXGBE_SDP1_GPIEN;
3266 
3267 	/* Thermal Sensor Interrupt */
3268 	if (sc->feat_en & IXGBE_FEATURE_TEMP_SENSOR)
3269 		gpie |= IXGBE_SDP0_GPIEN_X540;
3270 
3271 	/* Link detection */
3272 	switch (hw->mac.type) {
3273 	case ixgbe_mac_82599EB:
3274 		gpie |= IXGBE_SDP1_GPIEN | IXGBE_SDP2_GPIEN;
3275 		break;
3276 	case ixgbe_mac_X550EM_x:
3277 	case ixgbe_mac_X550EM_a:
3278 		gpie |= IXGBE_SDP0_GPIEN_X540;
3279 		break;
3280 	default:
3281 		break;
3282 	}
3283 
3284 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
3285 
3286 } /* ixgbe_config_gpie */
3287 
3288 /************************************************************************
3289  * ixgbe_config_delay_values
3290  *
3291  *   Requires sc->max_frame_size to be set.
3292  ************************************************************************/
3293 static void
3294 ixgbe_config_delay_values(struct ixgbe_softc *sc)
3295 {
3296 	struct ixgbe_hw *hw = &sc->hw;
3297 	u32             rxpb, frame, size, tmp;
3298 
3299 	frame = sc->max_frame_size;
3300 
3301 	/* Calculate High Water */
3302 	switch (hw->mac.type) {
3303 	case ixgbe_mac_X540:
3304 	case ixgbe_mac_X550:
3305 	case ixgbe_mac_X550EM_x:
3306 	case ixgbe_mac_X550EM_a:
3307 		tmp = IXGBE_DV_X540(frame, frame);
3308 		break;
3309 	default:
3310 		tmp = IXGBE_DV(frame, frame);
3311 		break;
3312 	}
3313 	size = IXGBE_BT2KB(tmp);
3314 	rxpb = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0)) >> 10;
3315 	hw->fc.high_water[0] = rxpb - size;
3316 
3317 	/* Now calculate Low Water */
3318 	switch (hw->mac.type) {
3319 	case ixgbe_mac_X540:
3320 	case ixgbe_mac_X550:
3321 	case ixgbe_mac_X550EM_x:
3322 	case ixgbe_mac_X550EM_a:
3323 		tmp = IXGBE_LOW_DV_X540(frame);
3324 		break;
3325 	default:
3326 		tmp = IXGBE_LOW_DV(frame);
3327 		break;
3328 	}
3329 	hw->fc.low_water[0] = IXGBE_BT2KB(tmp);
3330 
3331 	hw->fc.pause_time = IXGBE_FC_PAUSE;
3332 	hw->fc.send_xon = true;
3333 } /* ixgbe_config_delay_values */
3334 
3335 /************************************************************************
3336  * ixgbe_set_multi - Multicast Update
3337  *
3338  *   Called whenever multicast address list is updated.
3339  ************************************************************************/
3340 static u_int
3341 ixgbe_mc_filter_apply(void *arg, struct sockaddr_dl *sdl, u_int idx)
3342 {
3343 	struct ixgbe_softc *sc = arg;
3344 	struct ixgbe_mc_addr *mta = sc->mta;
3345 
3346 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
3347 		return (0);
3348 	bcopy(LLADDR(sdl), mta[idx].addr, IXGBE_ETH_LENGTH_OF_ADDRESS);
3349 	mta[idx].vmdq = sc->pool;
3350 
3351 	return (1);
3352 } /* ixgbe_mc_filter_apply */
3353 
3354 static void
3355 ixgbe_if_multi_set(if_ctx_t ctx)
3356 {
3357 	struct ixgbe_softc       *sc = iflib_get_softc(ctx);
3358 	struct ixgbe_mc_addr *mta;
3359 	struct ifnet         *ifp = iflib_get_ifp(ctx);
3360 	u8                   *update_ptr;
3361 	u32                  fctrl;
3362 	u_int		     mcnt;
3363 
3364 	IOCTL_DEBUGOUT("ixgbe_if_multi_set: begin");
3365 
3366 	mta = sc->mta;
3367 	bzero(mta, sizeof(*mta) * MAX_NUM_MULTICAST_ADDRESSES);
3368 
3369 	mcnt = if_foreach_llmaddr(iflib_get_ifp(ctx), ixgbe_mc_filter_apply, sc);
3370 
3371 	fctrl = IXGBE_READ_REG(&sc->hw, IXGBE_FCTRL);
3372 
3373 	if (ifp->if_flags & IFF_PROMISC)
3374 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3375 	else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
3376 	    ifp->if_flags & IFF_ALLMULTI) {
3377 		fctrl |= IXGBE_FCTRL_MPE;
3378 		fctrl &= ~IXGBE_FCTRL_UPE;
3379 	} else
3380 		fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3381 
3382 	IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, fctrl);
3383 
3384 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) {
3385 		update_ptr = (u8 *)mta;
3386 		ixgbe_update_mc_addr_list(&sc->hw, update_ptr, mcnt,
3387 		    ixgbe_mc_array_itr, true);
3388 	}
3389 
3390 } /* ixgbe_if_multi_set */
3391 
3392 /************************************************************************
3393  * ixgbe_mc_array_itr
3394  *
3395  *   An iterator function needed by the multicast shared code.
3396  *   It feeds the shared code routine the addresses in the
3397  *   array of ixgbe_set_multi() one by one.
3398  ************************************************************************/
3399 static u8 *
3400 ixgbe_mc_array_itr(struct ixgbe_hw *hw, u8 **update_ptr, u32 *vmdq)
3401 {
3402 	struct ixgbe_mc_addr *mta;
3403 
3404 	mta = (struct ixgbe_mc_addr *)*update_ptr;
3405 	*vmdq = mta->vmdq;
3406 
3407 	*update_ptr = (u8*)(mta + 1);
3408 
3409 	return (mta->addr);
3410 } /* ixgbe_mc_array_itr */
3411 
3412 /************************************************************************
3413  * ixgbe_local_timer - Timer routine
3414  *
3415  *   Checks for link status, updates statistics,
3416  *   and runs the watchdog check.
3417  ************************************************************************/
3418 static void
3419 ixgbe_if_timer(if_ctx_t ctx, uint16_t qid)
3420 {
3421 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3422 
3423 	if (qid != 0)
3424 		return;
3425 
3426 	/* Check for pluggable optics */
3427 	if (sc->sfp_probe)
3428 		if (!ixgbe_sfp_probe(ctx))
3429 			return; /* Nothing to do */
3430 
3431 	ixgbe_check_link(&sc->hw, &sc->link_speed, &sc->link_up, 0);
3432 
3433 	/* Fire off the adminq task */
3434 	iflib_admin_intr_deferred(ctx);
3435 
3436 } /* ixgbe_if_timer */
3437 
3438 /************************************************************************
3439  * ixgbe_sfp_probe
3440  *
3441  *   Determine if a port had optics inserted.
3442  ************************************************************************/
3443 static bool
3444 ixgbe_sfp_probe(if_ctx_t ctx)
3445 {
3446 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3447 	struct ixgbe_hw *hw = &sc->hw;
3448 	device_t        dev = iflib_get_dev(ctx);
3449 	bool            result = false;
3450 
3451 	if ((hw->phy.type == ixgbe_phy_nl) &&
3452 	    (hw->phy.sfp_type == ixgbe_sfp_type_not_present)) {
3453 		s32 ret = hw->phy.ops.identify_sfp(hw);
3454 		if (ret)
3455 			goto out;
3456 		ret = hw->phy.ops.reset(hw);
3457 		sc->sfp_probe = false;
3458 		if (ret == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3459 			device_printf(dev, "Unsupported SFP+ module detected!");
3460 			device_printf(dev,
3461 			    "Reload driver with supported module.\n");
3462 			goto out;
3463 		} else
3464 			device_printf(dev, "SFP+ module detected!\n");
3465 		/* We now have supported optics */
3466 		result = true;
3467 	}
3468 out:
3469 
3470 	return (result);
3471 } /* ixgbe_sfp_probe */
3472 
3473 /************************************************************************
3474  * ixgbe_handle_mod - Tasklet for SFP module interrupts
3475  ************************************************************************/
3476 static void
3477 ixgbe_handle_mod(void *context)
3478 {
3479 	if_ctx_t        ctx = context;
3480 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3481 	struct ixgbe_hw *hw = &sc->hw;
3482 	device_t        dev = iflib_get_dev(ctx);
3483 	u32             err, cage_full = 0;
3484 
3485 	if (sc->hw.need_crosstalk_fix) {
3486 		switch (hw->mac.type) {
3487 		case ixgbe_mac_82599EB:
3488 			cage_full = IXGBE_READ_REG(hw, IXGBE_ESDP) &
3489 			    IXGBE_ESDP_SDP2;
3490 			break;
3491 		case ixgbe_mac_X550EM_x:
3492 		case ixgbe_mac_X550EM_a:
3493 			cage_full = IXGBE_READ_REG(hw, IXGBE_ESDP) &
3494 			    IXGBE_ESDP_SDP0;
3495 			break;
3496 		default:
3497 			break;
3498 		}
3499 
3500 		if (!cage_full)
3501 			goto handle_mod_out;
3502 	}
3503 
3504 	err = hw->phy.ops.identify_sfp(hw);
3505 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3506 		device_printf(dev,
3507 		    "Unsupported SFP+ module type was detected.\n");
3508 		goto handle_mod_out;
3509 	}
3510 
3511 	if (hw->mac.type == ixgbe_mac_82598EB)
3512 		err = hw->phy.ops.reset(hw);
3513 	else
3514 		err = hw->mac.ops.setup_sfp(hw);
3515 
3516 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3517 		device_printf(dev,
3518 		    "Setup failure - unsupported SFP+ module type.\n");
3519 		goto handle_mod_out;
3520 	}
3521 	sc->task_requests |= IXGBE_REQUEST_TASK_MSF;
3522 	return;
3523 
3524 handle_mod_out:
3525 	sc->task_requests &= ~(IXGBE_REQUEST_TASK_MSF);
3526 } /* ixgbe_handle_mod */
3527 
3528 
3529 /************************************************************************
3530  * ixgbe_handle_msf - Tasklet for MSF (multispeed fiber) interrupts
3531  ************************************************************************/
3532 static void
3533 ixgbe_handle_msf(void *context)
3534 {
3535 	if_ctx_t        ctx = context;
3536 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3537 	struct ixgbe_hw *hw = &sc->hw;
3538 	u32             autoneg;
3539 	bool            negotiate;
3540 
3541 	/* get_supported_phy_layer will call hw->phy.ops.identify_sfp() */
3542 	sc->phy_layer = ixgbe_get_supported_physical_layer(hw);
3543 
3544 	autoneg = hw->phy.autoneg_advertised;
3545 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
3546 		hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiate);
3547 	if (hw->mac.ops.setup_link)
3548 		hw->mac.ops.setup_link(hw, autoneg, true);
3549 
3550 	/* Adjust media types shown in ifconfig */
3551 	ifmedia_removeall(sc->media);
3552 	ixgbe_add_media_types(sc->ctx);
3553 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
3554 } /* ixgbe_handle_msf */
3555 
3556 /************************************************************************
3557  * ixgbe_handle_phy - Tasklet for external PHY interrupts
3558  ************************************************************************/
3559 static void
3560 ixgbe_handle_phy(void *context)
3561 {
3562 	if_ctx_t        ctx = context;
3563 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3564 	struct ixgbe_hw *hw = &sc->hw;
3565 	int             error;
3566 
3567 	error = hw->phy.ops.handle_lasi(hw);
3568 	if (error == IXGBE_ERR_OVERTEMP)
3569 		device_printf(sc->dev, "CRITICAL: EXTERNAL PHY OVER TEMP!!  PHY will downshift to lower power state!\n");
3570 	else if (error)
3571 		device_printf(sc->dev,
3572 		    "Error handling LASI interrupt: %d\n", error);
3573 } /* ixgbe_handle_phy */
3574 
3575 /************************************************************************
3576  * ixgbe_if_stop - Stop the hardware
3577  *
3578  *   Disables all traffic on the adapter by issuing a
3579  *   global reset on the MAC and deallocates TX/RX buffers.
3580  ************************************************************************/
3581 static void
3582 ixgbe_if_stop(if_ctx_t ctx)
3583 {
3584 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3585 	struct ixgbe_hw *hw = &sc->hw;
3586 
3587 	INIT_DEBUGOUT("ixgbe_if_stop: begin\n");
3588 
3589 	ixgbe_reset_hw(hw);
3590 	hw->adapter_stopped = false;
3591 	ixgbe_stop_adapter(hw);
3592 	if (hw->mac.type == ixgbe_mac_82599EB)
3593 		ixgbe_stop_mac_link_on_d3_82599(hw);
3594 	/* Turn off the laser - noop with no optics */
3595 	ixgbe_disable_tx_laser(hw);
3596 
3597 	/* Update the stack */
3598 	sc->link_up = false;
3599 	ixgbe_if_update_admin_status(ctx);
3600 
3601 	/* reprogram the RAR[0] in case user changed it. */
3602 	ixgbe_set_rar(&sc->hw, 0, sc->hw.mac.addr, 0, IXGBE_RAH_AV);
3603 
3604 	return;
3605 } /* ixgbe_if_stop */
3606 
3607 /************************************************************************
3608  * ixgbe_update_link_status - Update OS on link state
3609  *
3610  * Note: Only updates the OS on the cached link state.
3611  *       The real check of the hardware only happens with
3612  *       a link interrupt.
3613  ************************************************************************/
3614 static void
3615 ixgbe_if_update_admin_status(if_ctx_t ctx)
3616 {
3617 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3618 	device_t       dev = iflib_get_dev(ctx);
3619 
3620 	if (sc->link_up) {
3621 		if (sc->link_active == false) {
3622 			if (bootverbose)
3623 				device_printf(dev, "Link is up %d Gbps %s \n",
3624 				    ((sc->link_speed == 128) ? 10 : 1),
3625 				    "Full Duplex");
3626 			sc->link_active = true;
3627 			/* Update any Flow Control changes */
3628 			ixgbe_fc_enable(&sc->hw);
3629 			/* Update DMA coalescing config */
3630 			ixgbe_config_dmac(sc);
3631 			/* should actually be negotiated value */
3632 			iflib_link_state_change(ctx, LINK_STATE_UP, IF_Gbps(10));
3633 
3634 			if (sc->feat_en & IXGBE_FEATURE_SRIOV)
3635 				ixgbe_ping_all_vfs(sc);
3636 		}
3637 	} else { /* Link down */
3638 		if (sc->link_active == true) {
3639 			if (bootverbose)
3640 				device_printf(dev, "Link is Down\n");
3641 			iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
3642 			sc->link_active = false;
3643 			if (sc->feat_en & IXGBE_FEATURE_SRIOV)
3644 				ixgbe_ping_all_vfs(sc);
3645 		}
3646 	}
3647 
3648 	/* Handle task requests from msix_link() */
3649 	if (sc->task_requests & IXGBE_REQUEST_TASK_MOD)
3650 		ixgbe_handle_mod(ctx);
3651 	if (sc->task_requests & IXGBE_REQUEST_TASK_MSF)
3652 		ixgbe_handle_msf(ctx);
3653 	if (sc->task_requests & IXGBE_REQUEST_TASK_MBX)
3654 		ixgbe_handle_mbx(ctx);
3655 	if (sc->task_requests & IXGBE_REQUEST_TASK_FDIR)
3656 		ixgbe_reinit_fdir(ctx);
3657 	if (sc->task_requests & IXGBE_REQUEST_TASK_PHY)
3658 		ixgbe_handle_phy(ctx);
3659 	sc->task_requests = 0;
3660 
3661 	ixgbe_update_stats_counters(sc);
3662 } /* ixgbe_if_update_admin_status */
3663 
3664 /************************************************************************
3665  * ixgbe_config_dmac - Configure DMA Coalescing
3666  ************************************************************************/
3667 static void
3668 ixgbe_config_dmac(struct ixgbe_softc *sc)
3669 {
3670 	struct ixgbe_hw          *hw = &sc->hw;
3671 	struct ixgbe_dmac_config *dcfg = &hw->mac.dmac_config;
3672 
3673 	if (hw->mac.type < ixgbe_mac_X550 || !hw->mac.ops.dmac_config)
3674 		return;
3675 
3676 	if (dcfg->watchdog_timer ^ sc->dmac ||
3677 	    dcfg->link_speed ^ sc->link_speed) {
3678 		dcfg->watchdog_timer = sc->dmac;
3679 		dcfg->fcoe_en = false;
3680 		dcfg->link_speed = sc->link_speed;
3681 		dcfg->num_tcs = 1;
3682 
3683 		INIT_DEBUGOUT2("dmac settings: watchdog %d, link speed %d\n",
3684 		    dcfg->watchdog_timer, dcfg->link_speed);
3685 
3686 		hw->mac.ops.dmac_config(hw);
3687 	}
3688 } /* ixgbe_config_dmac */
3689 
3690 /************************************************************************
3691  * ixgbe_if_enable_intr
3692  ************************************************************************/
3693 void
3694 ixgbe_if_enable_intr(if_ctx_t ctx)
3695 {
3696 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
3697 	struct ixgbe_hw    *hw = &sc->hw;
3698 	struct ix_rx_queue *que = sc->rx_queues;
3699 	u32                mask, fwsm;
3700 
3701 	mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
3702 
3703 	switch (sc->hw.mac.type) {
3704 	case ixgbe_mac_82599EB:
3705 		mask |= IXGBE_EIMS_ECC;
3706 		/* Temperature sensor on some scs */
3707 		mask |= IXGBE_EIMS_GPI_SDP0;
3708 		/* SFP+ (RX_LOS_N & MOD_ABS_N) */
3709 		mask |= IXGBE_EIMS_GPI_SDP1;
3710 		mask |= IXGBE_EIMS_GPI_SDP2;
3711 		break;
3712 	case ixgbe_mac_X540:
3713 		/* Detect if Thermal Sensor is enabled */
3714 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
3715 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
3716 			mask |= IXGBE_EIMS_TS;
3717 		mask |= IXGBE_EIMS_ECC;
3718 		break;
3719 	case ixgbe_mac_X550:
3720 		/* MAC thermal sensor is automatically enabled */
3721 		mask |= IXGBE_EIMS_TS;
3722 		mask |= IXGBE_EIMS_ECC;
3723 		break;
3724 	case ixgbe_mac_X550EM_x:
3725 	case ixgbe_mac_X550EM_a:
3726 		/* Some devices use SDP0 for important information */
3727 		if (hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3728 		    hw->device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3729 		    hw->device_id == IXGBE_DEV_ID_X550EM_A_SFP_N ||
3730 		    hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T)
3731 			mask |= IXGBE_EIMS_GPI_SDP0_BY_MAC(hw);
3732 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
3733 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3734 		mask |= IXGBE_EIMS_ECC;
3735 		break;
3736 	default:
3737 		break;
3738 	}
3739 
3740 	/* Enable Fan Failure detection */
3741 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL)
3742 		mask |= IXGBE_EIMS_GPI_SDP1;
3743 	/* Enable SR-IOV */
3744 	if (sc->feat_en & IXGBE_FEATURE_SRIOV)
3745 		mask |= IXGBE_EIMS_MAILBOX;
3746 	/* Enable Flow Director */
3747 	if (sc->feat_en & IXGBE_FEATURE_FDIR)
3748 		mask |= IXGBE_EIMS_FLOW_DIR;
3749 
3750 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
3751 
3752 	/* With MSI-X we use auto clear */
3753 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3754 		mask = IXGBE_EIMS_ENABLE_MASK;
3755 		/* Don't autoclear Link */
3756 		mask &= ~IXGBE_EIMS_OTHER;
3757 		mask &= ~IXGBE_EIMS_LSC;
3758 		if (sc->feat_cap & IXGBE_FEATURE_SRIOV)
3759 			mask &= ~IXGBE_EIMS_MAILBOX;
3760 		IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
3761 	}
3762 
3763 	/*
3764 	 * Now enable all queues, this is done separately to
3765 	 * allow for handling the extended (beyond 32) MSI-X
3766 	 * vectors that can be used by 82599
3767 	 */
3768 	for (int i = 0; i < sc->num_rx_queues; i++, que++)
3769 		ixgbe_enable_queue(sc, que->msix);
3770 
3771 	IXGBE_WRITE_FLUSH(hw);
3772 
3773 } /* ixgbe_if_enable_intr */
3774 
3775 /************************************************************************
3776  * ixgbe_disable_intr
3777  ************************************************************************/
3778 static void
3779 ixgbe_if_disable_intr(if_ctx_t ctx)
3780 {
3781 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3782 
3783 	if (sc->intr_type == IFLIB_INTR_MSIX)
3784 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIAC, 0);
3785 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
3786 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC, ~0);
3787 	} else {
3788 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC, 0xFFFF0000);
3789 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC_EX(0), ~0);
3790 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC_EX(1), ~0);
3791 	}
3792 	IXGBE_WRITE_FLUSH(&sc->hw);
3793 
3794 } /* ixgbe_if_disable_intr */
3795 
3796 /************************************************************************
3797  * ixgbe_link_intr_enable
3798  ************************************************************************/
3799 static void
3800 ixgbe_link_intr_enable(if_ctx_t ctx)
3801 {
3802 	struct ixgbe_hw *hw = &((struct ixgbe_softc *)iflib_get_softc(ctx))->hw;
3803 
3804 	/* Re-enable other interrupts */
3805 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_OTHER | IXGBE_EIMS_LSC);
3806 } /* ixgbe_link_intr_enable */
3807 
3808 /************************************************************************
3809  * ixgbe_if_rx_queue_intr_enable
3810  ************************************************************************/
3811 static int
3812 ixgbe_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
3813 {
3814 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
3815 	struct ix_rx_queue *que = &sc->rx_queues[rxqid];
3816 
3817 	ixgbe_enable_queue(sc, que->msix);
3818 
3819 	return (0);
3820 } /* ixgbe_if_rx_queue_intr_enable */
3821 
3822 /************************************************************************
3823  * ixgbe_enable_queue
3824  ************************************************************************/
3825 static void
3826 ixgbe_enable_queue(struct ixgbe_softc *sc, u32 vector)
3827 {
3828 	struct ixgbe_hw *hw = &sc->hw;
3829 	u64             queue = 1ULL << vector;
3830 	u32             mask;
3831 
3832 	if (hw->mac.type == ixgbe_mac_82598EB) {
3833 		mask = (IXGBE_EIMS_RTX_QUEUE & queue);
3834 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
3835 	} else {
3836 		mask = (queue & 0xFFFFFFFF);
3837 		if (mask)
3838 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
3839 		mask = (queue >> 32);
3840 		if (mask)
3841 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
3842 	}
3843 } /* ixgbe_enable_queue */
3844 
3845 /************************************************************************
3846  * ixgbe_disable_queue
3847  ************************************************************************/
3848 static void
3849 ixgbe_disable_queue(struct ixgbe_softc *sc, u32 vector)
3850 {
3851 	struct ixgbe_hw *hw = &sc->hw;
3852 	u64             queue = 1ULL << vector;
3853 	u32             mask;
3854 
3855 	if (hw->mac.type == ixgbe_mac_82598EB) {
3856 		mask = (IXGBE_EIMS_RTX_QUEUE & queue);
3857 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
3858 	} else {
3859 		mask = (queue & 0xFFFFFFFF);
3860 		if (mask)
3861 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
3862 		mask = (queue >> 32);
3863 		if (mask)
3864 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
3865 	}
3866 } /* ixgbe_disable_queue */
3867 
3868 /************************************************************************
3869  * ixgbe_intr - Legacy Interrupt Service Routine
3870  ************************************************************************/
3871 int
3872 ixgbe_intr(void *arg)
3873 {
3874 	struct ixgbe_softc     *sc = arg;
3875 	struct ix_rx_queue *que = sc->rx_queues;
3876 	struct ixgbe_hw    *hw = &sc->hw;
3877 	if_ctx_t           ctx = sc->ctx;
3878 	u32                eicr, eicr_mask;
3879 
3880 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3881 
3882 	++que->irqs;
3883 	if (eicr == 0) {
3884 		ixgbe_if_enable_intr(ctx);
3885 		return (FILTER_HANDLED);
3886 	}
3887 
3888 	/* Check for fan failure */
3889 	if ((hw->device_id == IXGBE_DEV_ID_82598AT) &&
3890 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
3891 		device_printf(sc->dev,
3892 		    "\nCRITICAL: FAN FAILURE!! REPLACE IMMEDIATELY!!\n");
3893 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
3894 	}
3895 
3896 	/* Link status change */
3897 	if (eicr & IXGBE_EICR_LSC) {
3898 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
3899 		iflib_admin_intr_deferred(ctx);
3900 	}
3901 
3902 	if (ixgbe_is_sfp(hw)) {
3903 		/* Pluggable optics-related interrupt */
3904 		if (hw->mac.type >= ixgbe_mac_X540)
3905 			eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
3906 		else
3907 			eicr_mask = IXGBE_EICR_GPI_SDP2_BY_MAC(hw);
3908 
3909 		if (eicr & eicr_mask) {
3910 			IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
3911 			sc->task_requests |= IXGBE_REQUEST_TASK_MOD;
3912 		}
3913 
3914 		if ((hw->mac.type == ixgbe_mac_82599EB) &&
3915 		    (eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw))) {
3916 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3917 			    IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
3918 			sc->task_requests |= IXGBE_REQUEST_TASK_MSF;
3919 		}
3920 	}
3921 
3922 	/* External PHY interrupt */
3923 	if ((hw->phy.type == ixgbe_phy_x550em_ext_t) &&
3924 	    (eicr & IXGBE_EICR_GPI_SDP0_X540))
3925 		sc->task_requests |= IXGBE_REQUEST_TASK_PHY;
3926 
3927 	return (FILTER_SCHEDULE_THREAD);
3928 } /* ixgbe_intr */
3929 
3930 /************************************************************************
3931  * ixgbe_free_pci_resources
3932  ************************************************************************/
3933 static void
3934 ixgbe_free_pci_resources(if_ctx_t ctx)
3935 {
3936 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3937 	struct         ix_rx_queue *que = sc->rx_queues;
3938 	device_t       dev = iflib_get_dev(ctx);
3939 
3940 	/* Release all MSI-X queue resources */
3941 	if (sc->intr_type == IFLIB_INTR_MSIX)
3942 		iflib_irq_free(ctx, &sc->irq);
3943 
3944 	if (que != NULL) {
3945 		for (int i = 0; i < sc->num_rx_queues; i++, que++) {
3946 			iflib_irq_free(ctx, &que->que_irq);
3947 		}
3948 	}
3949 
3950 	if (sc->pci_mem != NULL)
3951 		bus_release_resource(dev, SYS_RES_MEMORY,
3952 		    rman_get_rid(sc->pci_mem), sc->pci_mem);
3953 } /* ixgbe_free_pci_resources */
3954 
3955 /************************************************************************
3956  * ixgbe_sysctl_flowcntl
3957  *
3958  *   SYSCTL wrapper around setting Flow Control
3959  ************************************************************************/
3960 static int
3961 ixgbe_sysctl_flowcntl(SYSCTL_HANDLER_ARGS)
3962 {
3963 	struct ixgbe_softc *sc;
3964 	int            error, fc;
3965 
3966 	sc = (struct ixgbe_softc *)arg1;
3967 	fc = sc->hw.fc.current_mode;
3968 
3969 	error = sysctl_handle_int(oidp, &fc, 0, req);
3970 	if ((error) || (req->newptr == NULL))
3971 		return (error);
3972 
3973 	/* Don't bother if it's not changed */
3974 	if (fc == sc->hw.fc.current_mode)
3975 		return (0);
3976 
3977 	return ixgbe_set_flowcntl(sc, fc);
3978 } /* ixgbe_sysctl_flowcntl */
3979 
3980 /************************************************************************
3981  * ixgbe_set_flowcntl - Set flow control
3982  *
3983  *   Flow control values:
3984  *     0 - off
3985  *     1 - rx pause
3986  *     2 - tx pause
3987  *     3 - full
3988  ************************************************************************/
3989 static int
3990 ixgbe_set_flowcntl(struct ixgbe_softc *sc, int fc)
3991 {
3992 	switch (fc) {
3993 	case ixgbe_fc_rx_pause:
3994 	case ixgbe_fc_tx_pause:
3995 	case ixgbe_fc_full:
3996 		sc->hw.fc.requested_mode = fc;
3997 		if (sc->num_rx_queues > 1)
3998 			ixgbe_disable_rx_drop(sc);
3999 		break;
4000 	case ixgbe_fc_none:
4001 		sc->hw.fc.requested_mode = ixgbe_fc_none;
4002 		if (sc->num_rx_queues > 1)
4003 			ixgbe_enable_rx_drop(sc);
4004 		break;
4005 	default:
4006 		return (EINVAL);
4007 	}
4008 
4009 	/* Don't autoneg if forcing a value */
4010 	sc->hw.fc.disable_fc_autoneg = true;
4011 	ixgbe_fc_enable(&sc->hw);
4012 
4013 	return (0);
4014 } /* ixgbe_set_flowcntl */
4015 
4016 /************************************************************************
4017  * ixgbe_enable_rx_drop
4018  *
4019  *   Enable the hardware to drop packets when the buffer is
4020  *   full. This is useful with multiqueue, so that no single
4021  *   queue being full stalls the entire RX engine. We only
4022  *   enable this when Multiqueue is enabled AND Flow Control
4023  *   is disabled.
4024  ************************************************************************/
4025 static void
4026 ixgbe_enable_rx_drop(struct ixgbe_softc *sc)
4027 {
4028 	struct ixgbe_hw *hw = &sc->hw;
4029 	struct rx_ring  *rxr;
4030 	u32             srrctl;
4031 
4032 	for (int i = 0; i < sc->num_rx_queues; i++) {
4033 		rxr = &sc->rx_queues[i].rxr;
4034 		srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me));
4035 		srrctl |= IXGBE_SRRCTL_DROP_EN;
4036 		IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl);
4037 	}
4038 
4039 	/* enable drop for each vf */
4040 	for (int i = 0; i < sc->num_vfs; i++) {
4041 		IXGBE_WRITE_REG(hw, IXGBE_QDE,
4042 		                (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT) |
4043 		                IXGBE_QDE_ENABLE));
4044 	}
4045 } /* ixgbe_enable_rx_drop */
4046 
4047 /************************************************************************
4048  * ixgbe_disable_rx_drop
4049  ************************************************************************/
4050 static void
4051 ixgbe_disable_rx_drop(struct ixgbe_softc *sc)
4052 {
4053 	struct ixgbe_hw *hw = &sc->hw;
4054 	struct rx_ring  *rxr;
4055 	u32             srrctl;
4056 
4057 	for (int i = 0; i < sc->num_rx_queues; i++) {
4058 		rxr = &sc->rx_queues[i].rxr;
4059 		srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me));
4060 		srrctl &= ~IXGBE_SRRCTL_DROP_EN;
4061 		IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl);
4062 	}
4063 
4064 	/* disable drop for each vf */
4065 	for (int i = 0; i < sc->num_vfs; i++) {
4066 		IXGBE_WRITE_REG(hw, IXGBE_QDE,
4067 		    (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT)));
4068 	}
4069 } /* ixgbe_disable_rx_drop */
4070 
4071 /************************************************************************
4072  * ixgbe_sysctl_advertise
4073  *
4074  *   SYSCTL wrapper around setting advertised speed
4075  ************************************************************************/
4076 static int
4077 ixgbe_sysctl_advertise(SYSCTL_HANDLER_ARGS)
4078 {
4079 	struct ixgbe_softc *sc;
4080 	int            error, advertise;
4081 
4082 	sc = (struct ixgbe_softc *)arg1;
4083 	advertise = sc->advertise;
4084 
4085 	error = sysctl_handle_int(oidp, &advertise, 0, req);
4086 	if ((error) || (req->newptr == NULL))
4087 		return (error);
4088 
4089 	return ixgbe_set_advertise(sc, advertise);
4090 } /* ixgbe_sysctl_advertise */
4091 
4092 /************************************************************************
4093  * ixgbe_set_advertise - Control advertised link speed
4094  *
4095  *   Flags:
4096  *     0x1 - advertise 100 Mb
4097  *     0x2 - advertise 1G
4098  *     0x4 - advertise 10G
4099  *     0x8 - advertise 10 Mb (yes, Mb)
4100  ************************************************************************/
4101 static int
4102 ixgbe_set_advertise(struct ixgbe_softc *sc, int advertise)
4103 {
4104 	device_t         dev = iflib_get_dev(sc->ctx);
4105 	struct ixgbe_hw  *hw;
4106 	ixgbe_link_speed speed = 0;
4107 	ixgbe_link_speed link_caps = 0;
4108 	s32              err = IXGBE_NOT_IMPLEMENTED;
4109 	bool             negotiate = false;
4110 
4111 	/* Checks to validate new value */
4112 	if (sc->advertise == advertise) /* no change */
4113 		return (0);
4114 
4115 	hw = &sc->hw;
4116 
4117 	/* No speed changes for backplane media */
4118 	if (hw->phy.media_type == ixgbe_media_type_backplane)
4119 		return (ENODEV);
4120 
4121 	if (!((hw->phy.media_type == ixgbe_media_type_copper) ||
4122 	      (hw->phy.multispeed_fiber))) {
4123 		device_printf(dev, "Advertised speed can only be set on copper or multispeed fiber media types.\n");
4124 		return (EINVAL);
4125 	}
4126 
4127 	if (advertise < 0x1 || advertise > 0xF) {
4128 		device_printf(dev, "Invalid advertised speed; valid modes are 0x1 through 0xF\n");
4129 		return (EINVAL);
4130 	}
4131 
4132 	if (hw->mac.ops.get_link_capabilities) {
4133 		err = hw->mac.ops.get_link_capabilities(hw, &link_caps,
4134 		    &negotiate);
4135 		if (err != IXGBE_SUCCESS) {
4136 			device_printf(dev, "Unable to determine supported advertise speeds\n");
4137 			return (ENODEV);
4138 		}
4139 	}
4140 
4141 	/* Set new value and report new advertised mode */
4142 	if (advertise & 0x1) {
4143 		if (!(link_caps & IXGBE_LINK_SPEED_100_FULL)) {
4144 			device_printf(dev, "Interface does not support 100Mb advertised speed\n");
4145 			return (EINVAL);
4146 		}
4147 		speed |= IXGBE_LINK_SPEED_100_FULL;
4148 	}
4149 	if (advertise & 0x2) {
4150 		if (!(link_caps & IXGBE_LINK_SPEED_1GB_FULL)) {
4151 			device_printf(dev, "Interface does not support 1Gb advertised speed\n");
4152 			return (EINVAL);
4153 		}
4154 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
4155 	}
4156 	if (advertise & 0x4) {
4157 		if (!(link_caps & IXGBE_LINK_SPEED_10GB_FULL)) {
4158 			device_printf(dev, "Interface does not support 10Gb advertised speed\n");
4159 			return (EINVAL);
4160 		}
4161 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
4162 	}
4163 	if (advertise & 0x8) {
4164 		if (!(link_caps & IXGBE_LINK_SPEED_10_FULL)) {
4165 			device_printf(dev, "Interface does not support 10Mb advertised speed\n");
4166 			return (EINVAL);
4167 		}
4168 		speed |= IXGBE_LINK_SPEED_10_FULL;
4169 	}
4170 
4171 	hw->mac.autotry_restart = true;
4172 	hw->mac.ops.setup_link(hw, speed, true);
4173 	sc->advertise = advertise;
4174 
4175 	return (0);
4176 } /* ixgbe_set_advertise */
4177 
4178 /************************************************************************
4179  * ixgbe_get_advertise - Get current advertised speed settings
4180  *
4181  *   Formatted for sysctl usage.
4182  *   Flags:
4183  *     0x1 - advertise 100 Mb
4184  *     0x2 - advertise 1G
4185  *     0x4 - advertise 10G
4186  *     0x8 - advertise 10 Mb (yes, Mb)
4187  ************************************************************************/
4188 static int
4189 ixgbe_get_advertise(struct ixgbe_softc *sc)
4190 {
4191 	struct ixgbe_hw  *hw = &sc->hw;
4192 	int              speed;
4193 	ixgbe_link_speed link_caps = 0;
4194 	s32              err;
4195 	bool             negotiate = false;
4196 
4197 	/*
4198 	 * Advertised speed means nothing unless it's copper or
4199 	 * multi-speed fiber
4200 	 */
4201 	if (!(hw->phy.media_type == ixgbe_media_type_copper) &&
4202 	    !(hw->phy.multispeed_fiber))
4203 		return (0);
4204 
4205 	err = hw->mac.ops.get_link_capabilities(hw, &link_caps, &negotiate);
4206 	if (err != IXGBE_SUCCESS)
4207 		return (0);
4208 
4209 	speed =
4210 	    ((link_caps & IXGBE_LINK_SPEED_10GB_FULL) ? 4 : 0) |
4211 	    ((link_caps & IXGBE_LINK_SPEED_1GB_FULL)  ? 2 : 0) |
4212 	    ((link_caps & IXGBE_LINK_SPEED_100_FULL)  ? 1 : 0) |
4213 	    ((link_caps & IXGBE_LINK_SPEED_10_FULL)   ? 8 : 0);
4214 
4215 	return speed;
4216 } /* ixgbe_get_advertise */
4217 
4218 /************************************************************************
4219  * ixgbe_sysctl_dmac - Manage DMA Coalescing
4220  *
4221  *   Control values:
4222  *     0/1 - off / on (use default value of 1000)
4223  *
4224  *     Legal timer values are:
4225  *     50,100,250,500,1000,2000,5000,10000
4226  *
4227  *     Turning off interrupt moderation will also turn this off.
4228  ************************************************************************/
4229 static int
4230 ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS)
4231 {
4232 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4233 	struct ifnet   *ifp = iflib_get_ifp(sc->ctx);
4234 	int            error;
4235 	u16            newval;
4236 
4237 	newval = sc->dmac;
4238 	error = sysctl_handle_16(oidp, &newval, 0, req);
4239 	if ((error) || (req->newptr == NULL))
4240 		return (error);
4241 
4242 	switch (newval) {
4243 	case 0:
4244 		/* Disabled */
4245 		sc->dmac = 0;
4246 		break;
4247 	case 1:
4248 		/* Enable and use default */
4249 		sc->dmac = 1000;
4250 		break;
4251 	case 50:
4252 	case 100:
4253 	case 250:
4254 	case 500:
4255 	case 1000:
4256 	case 2000:
4257 	case 5000:
4258 	case 10000:
4259 		/* Legal values - allow */
4260 		sc->dmac = newval;
4261 		break;
4262 	default:
4263 		/* Do nothing, illegal value */
4264 		return (EINVAL);
4265 	}
4266 
4267 	/* Re-initialize hardware if it's already running */
4268 	if (ifp->if_drv_flags & IFF_DRV_RUNNING)
4269 		ifp->if_init(ifp);
4270 
4271 	return (0);
4272 } /* ixgbe_sysctl_dmac */
4273 
4274 #ifdef IXGBE_DEBUG
4275 /************************************************************************
4276  * ixgbe_sysctl_power_state
4277  *
4278  *   Sysctl to test power states
4279  *   Values:
4280  *     0      - set device to D0
4281  *     3      - set device to D3
4282  *     (none) - get current device power state
4283  ************************************************************************/
4284 static int
4285 ixgbe_sysctl_power_state(SYSCTL_HANDLER_ARGS)
4286 {
4287 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4288 	device_t       dev = sc->dev;
4289 	int            curr_ps, new_ps, error = 0;
4290 
4291 	curr_ps = new_ps = pci_get_powerstate(dev);
4292 
4293 	error = sysctl_handle_int(oidp, &new_ps, 0, req);
4294 	if ((error) || (req->newptr == NULL))
4295 		return (error);
4296 
4297 	if (new_ps == curr_ps)
4298 		return (0);
4299 
4300 	if (new_ps == 3 && curr_ps == 0)
4301 		error = DEVICE_SUSPEND(dev);
4302 	else if (new_ps == 0 && curr_ps == 3)
4303 		error = DEVICE_RESUME(dev);
4304 	else
4305 		return (EINVAL);
4306 
4307 	device_printf(dev, "New state: %d\n", pci_get_powerstate(dev));
4308 
4309 	return (error);
4310 } /* ixgbe_sysctl_power_state */
4311 #endif
4312 
4313 /************************************************************************
4314  * ixgbe_sysctl_wol_enable
4315  *
4316  *   Sysctl to enable/disable the WoL capability,
4317  *   if supported by the adapter.
4318  *
4319  *   Values:
4320  *     0 - disabled
4321  *     1 - enabled
4322  ************************************************************************/
4323 static int
4324 ixgbe_sysctl_wol_enable(SYSCTL_HANDLER_ARGS)
4325 {
4326 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4327 	struct ixgbe_hw *hw = &sc->hw;
4328 	int             new_wol_enabled;
4329 	int             error = 0;
4330 
4331 	new_wol_enabled = hw->wol_enabled;
4332 	error = sysctl_handle_int(oidp, &new_wol_enabled, 0, req);
4333 	if ((error) || (req->newptr == NULL))
4334 		return (error);
4335 	new_wol_enabled = !!(new_wol_enabled);
4336 	if (new_wol_enabled == hw->wol_enabled)
4337 		return (0);
4338 
4339 	if (new_wol_enabled > 0 && !sc->wol_support)
4340 		return (ENODEV);
4341 	else
4342 		hw->wol_enabled = new_wol_enabled;
4343 
4344 	return (0);
4345 } /* ixgbe_sysctl_wol_enable */
4346 
4347 /************************************************************************
4348  * ixgbe_sysctl_wufc - Wake Up Filter Control
4349  *
4350  *   Sysctl to enable/disable the types of packets that the
4351  *   adapter will wake up on upon receipt.
4352  *   Flags:
4353  *     0x1  - Link Status Change
4354  *     0x2  - Magic Packet
4355  *     0x4  - Direct Exact
4356  *     0x8  - Directed Multicast
4357  *     0x10 - Broadcast
4358  *     0x20 - ARP/IPv4 Request Packet
4359  *     0x40 - Direct IPv4 Packet
4360  *     0x80 - Direct IPv6 Packet
4361  *
4362  *   Settings not listed above will cause the sysctl to return an error.
4363  ************************************************************************/
4364 static int
4365 ixgbe_sysctl_wufc(SYSCTL_HANDLER_ARGS)
4366 {
4367 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4368 	int            error = 0;
4369 	u32            new_wufc;
4370 
4371 	new_wufc = sc->wufc;
4372 
4373 	error = sysctl_handle_32(oidp, &new_wufc, 0, req);
4374 	if ((error) || (req->newptr == NULL))
4375 		return (error);
4376 	if (new_wufc == sc->wufc)
4377 		return (0);
4378 
4379 	if (new_wufc & 0xffffff00)
4380 		return (EINVAL);
4381 
4382 	new_wufc &= 0xff;
4383 	new_wufc |= (0xffffff & sc->wufc);
4384 	sc->wufc = new_wufc;
4385 
4386 	return (0);
4387 } /* ixgbe_sysctl_wufc */
4388 
4389 #ifdef IXGBE_DEBUG
4390 /************************************************************************
4391  * ixgbe_sysctl_print_rss_config
4392  ************************************************************************/
4393 static int
4394 ixgbe_sysctl_print_rss_config(SYSCTL_HANDLER_ARGS)
4395 {
4396 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4397 	struct ixgbe_hw *hw = &sc->hw;
4398 	device_t        dev = sc->dev;
4399 	struct sbuf     *buf;
4400 	int             error = 0, reta_size;
4401 	u32             reg;
4402 
4403 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4404 	if (!buf) {
4405 		device_printf(dev, "Could not allocate sbuf for output.\n");
4406 		return (ENOMEM);
4407 	}
4408 
4409 	// TODO: use sbufs to make a string to print out
4410 	/* Set multiplier for RETA setup and table size based on MAC */
4411 	switch (sc->hw.mac.type) {
4412 	case ixgbe_mac_X550:
4413 	case ixgbe_mac_X550EM_x:
4414 	case ixgbe_mac_X550EM_a:
4415 		reta_size = 128;
4416 		break;
4417 	default:
4418 		reta_size = 32;
4419 		break;
4420 	}
4421 
4422 	/* Print out the redirection table */
4423 	sbuf_cat(buf, "\n");
4424 	for (int i = 0; i < reta_size; i++) {
4425 		if (i < 32) {
4426 			reg = IXGBE_READ_REG(hw, IXGBE_RETA(i));
4427 			sbuf_printf(buf, "RETA(%2d): 0x%08x\n", i, reg);
4428 		} else {
4429 			reg = IXGBE_READ_REG(hw, IXGBE_ERETA(i - 32));
4430 			sbuf_printf(buf, "ERETA(%2d): 0x%08x\n", i - 32, reg);
4431 		}
4432 	}
4433 
4434 	// TODO: print more config
4435 
4436 	error = sbuf_finish(buf);
4437 	if (error)
4438 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4439 
4440 	sbuf_delete(buf);
4441 
4442 	return (0);
4443 } /* ixgbe_sysctl_print_rss_config */
4444 #endif /* IXGBE_DEBUG */
4445 
4446 /************************************************************************
4447  * ixgbe_sysctl_phy_temp - Retrieve temperature of PHY
4448  *
4449  *   For X552/X557-AT devices using an external PHY
4450  ************************************************************************/
4451 static int
4452 ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS)
4453 {
4454 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4455 	struct ixgbe_hw *hw = &sc->hw;
4456 	u16             reg;
4457 
4458 	if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) {
4459 		device_printf(iflib_get_dev(sc->ctx),
4460 		    "Device has no supported external thermal sensor.\n");
4461 		return (ENODEV);
4462 	}
4463 
4464 	if (hw->phy.ops.read_reg(hw, IXGBE_PHY_CURRENT_TEMP,
4465 	    IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, &reg)) {
4466 		device_printf(iflib_get_dev(sc->ctx),
4467 		    "Error reading from PHY's current temperature register\n");
4468 		return (EAGAIN);
4469 	}
4470 
4471 	/* Shift temp for output */
4472 	reg = reg >> 8;
4473 
4474 	return (sysctl_handle_16(oidp, NULL, reg, req));
4475 } /* ixgbe_sysctl_phy_temp */
4476 
4477 /************************************************************************
4478  * ixgbe_sysctl_phy_overtemp_occurred
4479  *
4480  *   Reports (directly from the PHY) whether the current PHY
4481  *   temperature is over the overtemp threshold.
4482  ************************************************************************/
4483 static int
4484 ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS)
4485 {
4486 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4487 	struct ixgbe_hw *hw = &sc->hw;
4488 	u16             reg;
4489 
4490 	if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) {
4491 		device_printf(iflib_get_dev(sc->ctx),
4492 		    "Device has no supported external thermal sensor.\n");
4493 		return (ENODEV);
4494 	}
4495 
4496 	if (hw->phy.ops.read_reg(hw, IXGBE_PHY_OVERTEMP_STATUS,
4497 	    IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, &reg)) {
4498 		device_printf(iflib_get_dev(sc->ctx),
4499 		    "Error reading from PHY's temperature status register\n");
4500 		return (EAGAIN);
4501 	}
4502 
4503 	/* Get occurrence bit */
4504 	reg = !!(reg & 0x4000);
4505 
4506 	return (sysctl_handle_16(oidp, 0, reg, req));
4507 } /* ixgbe_sysctl_phy_overtemp_occurred */
4508 
4509 /************************************************************************
4510  * ixgbe_sysctl_eee_state
4511  *
4512  *   Sysctl to set EEE power saving feature
4513  *   Values:
4514  *     0      - disable EEE
4515  *     1      - enable EEE
4516  *     (none) - get current device EEE state
4517  ************************************************************************/
4518 static int
4519 ixgbe_sysctl_eee_state(SYSCTL_HANDLER_ARGS)
4520 {
4521 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4522 	device_t       dev = sc->dev;
4523 	struct ifnet   *ifp = iflib_get_ifp(sc->ctx);
4524 	int            curr_eee, new_eee, error = 0;
4525 	s32            retval;
4526 
4527 	curr_eee = new_eee = !!(sc->feat_en & IXGBE_FEATURE_EEE);
4528 
4529 	error = sysctl_handle_int(oidp, &new_eee, 0, req);
4530 	if ((error) || (req->newptr == NULL))
4531 		return (error);
4532 
4533 	/* Nothing to do */
4534 	if (new_eee == curr_eee)
4535 		return (0);
4536 
4537 	/* Not supported */
4538 	if (!(sc->feat_cap & IXGBE_FEATURE_EEE))
4539 		return (EINVAL);
4540 
4541 	/* Bounds checking */
4542 	if ((new_eee < 0) || (new_eee > 1))
4543 		return (EINVAL);
4544 
4545 	retval = ixgbe_setup_eee(&sc->hw, new_eee);
4546 	if (retval) {
4547 		device_printf(dev, "Error in EEE setup: 0x%08X\n", retval);
4548 		return (EINVAL);
4549 	}
4550 
4551 	/* Restart auto-neg */
4552 	ifp->if_init(ifp);
4553 
4554 	device_printf(dev, "New EEE state: %d\n", new_eee);
4555 
4556 	/* Cache new value */
4557 	if (new_eee)
4558 		sc->feat_en |= IXGBE_FEATURE_EEE;
4559 	else
4560 		sc->feat_en &= ~IXGBE_FEATURE_EEE;
4561 
4562 	return (error);
4563 } /* ixgbe_sysctl_eee_state */
4564 
4565 /************************************************************************
4566  * ixgbe_init_device_features
4567  ************************************************************************/
4568 static void
4569 ixgbe_init_device_features(struct ixgbe_softc *sc)
4570 {
4571 	sc->feat_cap = IXGBE_FEATURE_NETMAP
4572 	                  | IXGBE_FEATURE_RSS
4573 	                  | IXGBE_FEATURE_MSI
4574 	                  | IXGBE_FEATURE_MSIX
4575 	                  | IXGBE_FEATURE_LEGACY_IRQ;
4576 
4577 	/* Set capabilities first... */
4578 	switch (sc->hw.mac.type) {
4579 	case ixgbe_mac_82598EB:
4580 		if (sc->hw.device_id == IXGBE_DEV_ID_82598AT)
4581 			sc->feat_cap |= IXGBE_FEATURE_FAN_FAIL;
4582 		break;
4583 	case ixgbe_mac_X540:
4584 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4585 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4586 		if ((sc->hw.device_id == IXGBE_DEV_ID_X540_BYPASS) &&
4587 		    (sc->hw.bus.func == 0))
4588 			sc->feat_cap |= IXGBE_FEATURE_BYPASS;
4589 		break;
4590 	case ixgbe_mac_X550:
4591 		sc->feat_cap |= IXGBE_FEATURE_TEMP_SENSOR;
4592 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4593 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4594 		break;
4595 	case ixgbe_mac_X550EM_x:
4596 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4597 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4598 		break;
4599 	case ixgbe_mac_X550EM_a:
4600 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4601 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4602 		sc->feat_cap &= ~IXGBE_FEATURE_LEGACY_IRQ;
4603 		if ((sc->hw.device_id == IXGBE_DEV_ID_X550EM_A_1G_T) ||
4604 		    (sc->hw.device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)) {
4605 			sc->feat_cap |= IXGBE_FEATURE_TEMP_SENSOR;
4606 			sc->feat_cap |= IXGBE_FEATURE_EEE;
4607 		}
4608 		break;
4609 	case ixgbe_mac_82599EB:
4610 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4611 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4612 		if ((sc->hw.device_id == IXGBE_DEV_ID_82599_BYPASS) &&
4613 		    (sc->hw.bus.func == 0))
4614 			sc->feat_cap |= IXGBE_FEATURE_BYPASS;
4615 		if (sc->hw.device_id == IXGBE_DEV_ID_82599_QSFP_SF_QP)
4616 			sc->feat_cap &= ~IXGBE_FEATURE_LEGACY_IRQ;
4617 		break;
4618 	default:
4619 		break;
4620 	}
4621 
4622 	/* Enabled by default... */
4623 	/* Fan failure detection */
4624 	if (sc->feat_cap & IXGBE_FEATURE_FAN_FAIL)
4625 		sc->feat_en |= IXGBE_FEATURE_FAN_FAIL;
4626 	/* Netmap */
4627 	if (sc->feat_cap & IXGBE_FEATURE_NETMAP)
4628 		sc->feat_en |= IXGBE_FEATURE_NETMAP;
4629 	/* EEE */
4630 	if (sc->feat_cap & IXGBE_FEATURE_EEE)
4631 		sc->feat_en |= IXGBE_FEATURE_EEE;
4632 	/* Thermal Sensor */
4633 	if (sc->feat_cap & IXGBE_FEATURE_TEMP_SENSOR)
4634 		sc->feat_en |= IXGBE_FEATURE_TEMP_SENSOR;
4635 
4636 	/* Enabled via global sysctl... */
4637 	/* Flow Director */
4638 	if (ixgbe_enable_fdir) {
4639 		if (sc->feat_cap & IXGBE_FEATURE_FDIR)
4640 			sc->feat_en |= IXGBE_FEATURE_FDIR;
4641 		else
4642 			device_printf(sc->dev, "Device does not support Flow Director. Leaving disabled.");
4643 	}
4644 	/*
4645 	 * Message Signal Interrupts - Extended (MSI-X)
4646 	 * Normal MSI is only enabled if MSI-X calls fail.
4647 	 */
4648 	if (!ixgbe_enable_msix)
4649 		sc->feat_cap &= ~IXGBE_FEATURE_MSIX;
4650 	/* Receive-Side Scaling (RSS) */
4651 	if ((sc->feat_cap & IXGBE_FEATURE_RSS) && ixgbe_enable_rss)
4652 		sc->feat_en |= IXGBE_FEATURE_RSS;
4653 
4654 	/* Disable features with unmet dependencies... */
4655 	/* No MSI-X */
4656 	if (!(sc->feat_cap & IXGBE_FEATURE_MSIX)) {
4657 		sc->feat_cap &= ~IXGBE_FEATURE_RSS;
4658 		sc->feat_cap &= ~IXGBE_FEATURE_SRIOV;
4659 		sc->feat_en &= ~IXGBE_FEATURE_RSS;
4660 		sc->feat_en &= ~IXGBE_FEATURE_SRIOV;
4661 	}
4662 } /* ixgbe_init_device_features */
4663 
4664 /************************************************************************
4665  * ixgbe_check_fan_failure
4666  ************************************************************************/
4667 static void
4668 ixgbe_check_fan_failure(struct ixgbe_softc *sc, u32 reg, bool in_interrupt)
4669 {
4670 	u32 mask;
4671 
4672 	mask = (in_interrupt) ? IXGBE_EICR_GPI_SDP1_BY_MAC(&sc->hw) :
4673 	    IXGBE_ESDP_SDP1;
4674 
4675 	if (reg & mask)
4676 		device_printf(sc->dev, "\nCRITICAL: FAN FAILURE!! REPLACE IMMEDIATELY!!\n");
4677 } /* ixgbe_check_fan_failure */
4678 
4679 /************************************************************************
4680  * ixgbe_sbuf_fw_version
4681  ************************************************************************/
4682 static void
4683 ixgbe_sbuf_fw_version(struct ixgbe_hw *hw, struct sbuf *buf)
4684 {
4685 	struct ixgbe_nvm_version nvm_ver = {0};
4686 	uint16_t phyfw = 0;
4687 	int status;
4688 	const char *space = "";
4689 
4690 	ixgbe_get_oem_prod_version(hw, &nvm_ver); /* OEM's NVM version */
4691 	ixgbe_get_orom_version(hw, &nvm_ver); /* Option ROM */
4692 	ixgbe_get_etk_id(hw, &nvm_ver); /* eTrack identifies a build in Intel's SCM */
4693 	status = ixgbe_get_phy_firmware_version(hw, &phyfw);
4694 
4695 	if (nvm_ver.oem_valid) {
4696 		sbuf_printf(buf, "NVM OEM V%d.%d R%d", nvm_ver.oem_major,
4697 		    nvm_ver.oem_minor, nvm_ver.oem_release);
4698 		space = " ";
4699 	}
4700 
4701 	if (nvm_ver.or_valid) {
4702 		sbuf_printf(buf, "%sOption ROM V%d-b%d-p%d",
4703 		    space, nvm_ver.or_major, nvm_ver.or_build, nvm_ver.or_patch);
4704 		space = " ";
4705 	}
4706 
4707 	if (nvm_ver.etk_id != ((NVM_VER_INVALID << NVM_ETK_SHIFT) |
4708 	    NVM_VER_INVALID)) {
4709 		sbuf_printf(buf, "%seTrack 0x%08x", space, nvm_ver.etk_id);
4710 		space = " ";
4711 	}
4712 
4713 	if (phyfw != 0 && status == IXGBE_SUCCESS)
4714 		sbuf_printf(buf, "%sPHY FW V%d", space, phyfw);
4715 } /* ixgbe_sbuf_fw_version */
4716 
4717 /************************************************************************
4718  * ixgbe_print_fw_version
4719  ************************************************************************/
4720 static void
4721 ixgbe_print_fw_version(if_ctx_t ctx)
4722 {
4723 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
4724 	struct ixgbe_hw *hw = &sc->hw;
4725 	device_t dev = sc->dev;
4726 	struct sbuf *buf;
4727 	int error = 0;
4728 
4729 	buf = sbuf_new_auto();
4730 	if (!buf) {
4731 		device_printf(dev, "Could not allocate sbuf for output.\n");
4732 		return;
4733 	}
4734 
4735 	ixgbe_sbuf_fw_version(hw, buf);
4736 
4737 	error = sbuf_finish(buf);
4738 	if (error)
4739 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4740 	else if (sbuf_len(buf))
4741 		device_printf(dev, "%s\n", sbuf_data(buf));
4742 
4743 	sbuf_delete(buf);
4744 } /* ixgbe_print_fw_version */
4745 
4746 /************************************************************************
4747  * ixgbe_sysctl_print_fw_version
4748  ************************************************************************/
4749 static int
4750 ixgbe_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS)
4751 {
4752 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4753 	struct ixgbe_hw *hw = &sc->hw;
4754 	device_t dev = sc->dev;
4755 	struct sbuf *buf;
4756 	int error = 0;
4757 
4758 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4759 	if (!buf) {
4760 		device_printf(dev, "Could not allocate sbuf for output.\n");
4761 		return (ENOMEM);
4762 	}
4763 
4764 	ixgbe_sbuf_fw_version(hw, buf);
4765 
4766 	error = sbuf_finish(buf);
4767 	if (error)
4768 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4769 
4770 	sbuf_delete(buf);
4771 
4772 	return (0);
4773 } /* ixgbe_sysctl_print_fw_version */
4774