xref: /freebsd/sys/dev/ixgbe/if_ix.c (revision 02e9120893770924227138ba49df1edb3896112a)
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 
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 #include "opt_rss.h"
37 
38 #include "ixgbe.h"
39 #include "ixgbe_sriov.h"
40 #include "ifdi_if.h"
41 
42 #include <net/netmap.h>
43 #include <dev/netmap/netmap_kern.h>
44 
45 /************************************************************************
46  * Driver version
47  ************************************************************************/
48 static const char ixgbe_driver_version[] = "4.0.1-k";
49 
50 /************************************************************************
51  * PCI Device ID Table
52  *
53  *   Used by probe to select devices to load on
54  *   Last field stores an index into ixgbe_strings
55  *   Last entry must be all 0s
56  *
57  *   { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
58  ************************************************************************/
59 static const pci_vendor_info_t ixgbe_vendor_info_array[] =
60 {
61   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT,  "Intel(R) 82598EB AF (Dual Fiber)"),
62   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT,  "Intel(R) 82598EB AF (Fiber)"),
63   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4,  "Intel(R) 82598EB AT (CX4)"),
64   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT,  "Intel(R) 82598EB AT"),
65   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2,  "Intel(R) 82598EB AT2"),
66   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598,  "Intel(R) 82598"),
67   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT,  "Intel(R) 82598EB AF DA (Dual Fiber)"),
68   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT,  "Intel(R) 82598EB AT (Dual CX4)"),
69   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR,  "Intel(R) 82598EB AF (Dual Fiber LR)"),
70   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM,  "Intel(R) 82598EB AF (Dual Fiber SR)"),
71   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM,  "Intel(R) 82598EB LOM"),
72   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4,  "Intel(R) X520 82599 (KX4)"),
73   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ,  "Intel(R) X520 82599 (KX4 Mezzanine)"),
74   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP,  "Intel(R) X520 82599ES (SFI/SFP+)"),
75   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM,  "Intel(R) X520 82599 (XAUI/BX4)"),
76   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4,  "Intel(R) X520 82599 (Dual CX4)"),
77   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM,  "Intel(R) X520-T 82599 LOM"),
78   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_LS,  "Intel(R) X520 82599 LS"),
79   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE,  "Intel(R) X520 82599 (Combined Backplane)"),
80   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE,  "Intel(R) X520 82599 (Backplane w/FCoE)"),
81   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2,  "Intel(R) X520 82599 (Dual SFP+)"),
82   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE,  "Intel(R) X520 82599 (Dual SFP+ w/FCoE)"),
83   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP,  "Intel(R) X520-1 82599EN (SFP+)"),
84   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP,  "Intel(R) X520-4 82599 (Quad SFP+)"),
85   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP,  "Intel(R) X520-Q1 82599 (QSFP+)"),
86   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T,  "Intel(R) X540-AT2"),
87   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1,  "Intel(R) X540-T1"),
88   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T,  "Intel(R) X550-T2"),
89   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1, "Intel(R) X550-T1"),
90   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR,  "Intel(R) X552 (KR Backplane)"),
91   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4,  "Intel(R) X552 (KX4 Backplane)"),
92   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T,  "Intel(R) X552/X557-AT (10GBASE-T)"),
93   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T,  "Intel(R) X552 (1000BASE-T)"),
94   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP, "Intel(R) X552 (SFP+)"),
95   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR, "Intel(R) X553 (KR Backplane)"),
96   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L, "Intel(R) X553 L (KR Backplane)"),
97   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP, "Intel(R) X553 (SFP+)"),
98   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N, "Intel(R) X553 N (SFP+)"),
99   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII, "Intel(R) X553 (1GbE SGMII)"),
100   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L, "Intel(R) X553 L (1GbE SGMII)"),
101   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T, "Intel(R) X553/X557-AT (10GBASE-T)"),
102   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T, "Intel(R) X553 (1GbE)"),
103   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L, "Intel(R) X553 L (1GbE)"),
104   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_BYPASS, "Intel(R) X540-T2 (Bypass)"),
105   PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS, "Intel(R) X520 82599 (Bypass)"),
106 	/* required last entry */
107   PVID_END
108 };
109 
110 static void *ixgbe_register(device_t);
111 static int  ixgbe_if_attach_pre(if_ctx_t);
112 static int  ixgbe_if_attach_post(if_ctx_t);
113 static int  ixgbe_if_detach(if_ctx_t);
114 static int  ixgbe_if_shutdown(if_ctx_t);
115 static int  ixgbe_if_suspend(if_ctx_t);
116 static int  ixgbe_if_resume(if_ctx_t);
117 
118 static void ixgbe_if_stop(if_ctx_t);
119 void ixgbe_if_enable_intr(if_ctx_t);
120 static void ixgbe_if_disable_intr(if_ctx_t);
121 static void ixgbe_link_intr_enable(if_ctx_t);
122 static int  ixgbe_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
123 static void ixgbe_if_media_status(if_ctx_t, struct ifmediareq *);
124 static int  ixgbe_if_media_change(if_ctx_t);
125 static int  ixgbe_if_msix_intr_assign(if_ctx_t, int);
126 static int  ixgbe_if_mtu_set(if_ctx_t, uint32_t);
127 static void ixgbe_if_crcstrip_set(if_ctx_t, int, int);
128 static void ixgbe_if_multi_set(if_ctx_t);
129 static int  ixgbe_if_promisc_set(if_ctx_t, int);
130 static int  ixgbe_if_tx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
131 static int  ixgbe_if_rx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
132 static void ixgbe_if_queues_free(if_ctx_t);
133 static void ixgbe_if_timer(if_ctx_t, uint16_t);
134 static void ixgbe_if_update_admin_status(if_ctx_t);
135 static void ixgbe_if_vlan_register(if_ctx_t, u16);
136 static void ixgbe_if_vlan_unregister(if_ctx_t, u16);
137 static int  ixgbe_if_i2c_req(if_ctx_t, struct ifi2creq *);
138 static bool ixgbe_if_needs_restart(if_ctx_t, enum iflib_restart_event);
139 int ixgbe_intr(void *);
140 
141 /************************************************************************
142  * Function prototypes
143  ************************************************************************/
144 static uint64_t ixgbe_if_get_counter(if_ctx_t, ift_counter);
145 
146 static void ixgbe_enable_queue(struct ixgbe_softc *, u32);
147 static void ixgbe_disable_queue(struct ixgbe_softc *, u32);
148 static void ixgbe_add_device_sysctls(if_ctx_t);
149 static int  ixgbe_allocate_pci_resources(if_ctx_t);
150 static int  ixgbe_setup_low_power_mode(if_ctx_t);
151 
152 static void ixgbe_config_dmac(struct ixgbe_softc *);
153 static void ixgbe_configure_ivars(struct ixgbe_softc *);
154 static void ixgbe_set_ivar(struct ixgbe_softc *, u8, u8, s8);
155 static u8   *ixgbe_mc_array_itr(struct ixgbe_hw *, u8 **, u32 *);
156 static bool ixgbe_sfp_probe(if_ctx_t);
157 
158 static void ixgbe_free_pci_resources(if_ctx_t);
159 
160 static int  ixgbe_msix_link(void *);
161 static int  ixgbe_msix_que(void *);
162 static void ixgbe_initialize_rss_mapping(struct ixgbe_softc *);
163 static void ixgbe_initialize_receive_units(if_ctx_t);
164 static void ixgbe_initialize_transmit_units(if_ctx_t);
165 
166 static int  ixgbe_setup_interface(if_ctx_t);
167 static void ixgbe_init_device_features(struct ixgbe_softc *);
168 static void ixgbe_check_fan_failure(struct ixgbe_softc *, u32, bool);
169 static void ixgbe_sbuf_fw_version(struct ixgbe_hw *, struct sbuf *);
170 static void ixgbe_print_fw_version(if_ctx_t);
171 static void ixgbe_add_media_types(if_ctx_t);
172 static void ixgbe_update_stats_counters(struct ixgbe_softc *);
173 static void ixgbe_config_link(if_ctx_t);
174 static void ixgbe_get_slot_info(struct ixgbe_softc *);
175 static void ixgbe_check_wol_support(struct ixgbe_softc *);
176 static void ixgbe_enable_rx_drop(struct ixgbe_softc *);
177 static void ixgbe_disable_rx_drop(struct ixgbe_softc *);
178 
179 static void ixgbe_add_hw_stats(struct ixgbe_softc *);
180 static int  ixgbe_set_flowcntl(struct ixgbe_softc *, int);
181 static int  ixgbe_set_advertise(struct ixgbe_softc *, int);
182 static int  ixgbe_get_default_advertise(struct ixgbe_softc *);
183 static void ixgbe_setup_vlan_hw_support(if_ctx_t);
184 static void ixgbe_config_gpie(struct ixgbe_softc *);
185 static void ixgbe_config_delay_values(struct ixgbe_softc *);
186 
187 /* Sysctl handlers */
188 static int  ixgbe_sysctl_flowcntl(SYSCTL_HANDLER_ARGS);
189 static int  ixgbe_sysctl_advertise(SYSCTL_HANDLER_ARGS);
190 static int  ixgbe_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS);
191 static int  ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS);
192 static int  ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS);
193 static int  ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS);
194 static int  ixgbe_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS);
195 #ifdef IXGBE_DEBUG
196 static int  ixgbe_sysctl_power_state(SYSCTL_HANDLER_ARGS);
197 static int  ixgbe_sysctl_print_rss_config(SYSCTL_HANDLER_ARGS);
198 #endif
199 static int  ixgbe_sysctl_rdh_handler(SYSCTL_HANDLER_ARGS);
200 static int  ixgbe_sysctl_rdt_handler(SYSCTL_HANDLER_ARGS);
201 static int  ixgbe_sysctl_tdt_handler(SYSCTL_HANDLER_ARGS);
202 static int  ixgbe_sysctl_tdh_handler(SYSCTL_HANDLER_ARGS);
203 static int  ixgbe_sysctl_eee_state(SYSCTL_HANDLER_ARGS);
204 static int  ixgbe_sysctl_wol_enable(SYSCTL_HANDLER_ARGS);
205 static int  ixgbe_sysctl_wufc(SYSCTL_HANDLER_ARGS);
206 
207 /* Deferred interrupt tasklets */
208 static void ixgbe_handle_msf(void *);
209 static void ixgbe_handle_mod(void *);
210 static void ixgbe_handle_phy(void *);
211 
212 /************************************************************************
213  *  FreeBSD Device Interface Entry Points
214  ************************************************************************/
215 static device_method_t ix_methods[] = {
216 	/* Device interface */
217 	DEVMETHOD(device_register, ixgbe_register),
218 	DEVMETHOD(device_probe, iflib_device_probe),
219 	DEVMETHOD(device_attach, iflib_device_attach),
220 	DEVMETHOD(device_detach, iflib_device_detach),
221 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
222 	DEVMETHOD(device_suspend, iflib_device_suspend),
223 	DEVMETHOD(device_resume, iflib_device_resume),
224 #ifdef PCI_IOV
225 	DEVMETHOD(pci_iov_init, iflib_device_iov_init),
226 	DEVMETHOD(pci_iov_uninit, iflib_device_iov_uninit),
227 	DEVMETHOD(pci_iov_add_vf, iflib_device_iov_add_vf),
228 #endif /* PCI_IOV */
229 	DEVMETHOD_END
230 };
231 
232 static driver_t ix_driver = {
233 	"ix", ix_methods, sizeof(struct ixgbe_softc),
234 };
235 
236 DRIVER_MODULE(ix, pci, ix_driver, 0, 0);
237 IFLIB_PNP_INFO(pci, ix_driver, ixgbe_vendor_info_array);
238 MODULE_DEPEND(ix, pci, 1, 1, 1);
239 MODULE_DEPEND(ix, ether, 1, 1, 1);
240 MODULE_DEPEND(ix, iflib, 1, 1, 1);
241 
242 static device_method_t ixgbe_if_methods[] = {
243 	DEVMETHOD(ifdi_attach_pre, ixgbe_if_attach_pre),
244 	DEVMETHOD(ifdi_attach_post, ixgbe_if_attach_post),
245 	DEVMETHOD(ifdi_detach, ixgbe_if_detach),
246 	DEVMETHOD(ifdi_shutdown, ixgbe_if_shutdown),
247 	DEVMETHOD(ifdi_suspend, ixgbe_if_suspend),
248 	DEVMETHOD(ifdi_resume, ixgbe_if_resume),
249 	DEVMETHOD(ifdi_init, ixgbe_if_init),
250 	DEVMETHOD(ifdi_stop, ixgbe_if_stop),
251 	DEVMETHOD(ifdi_msix_intr_assign, ixgbe_if_msix_intr_assign),
252 	DEVMETHOD(ifdi_intr_enable, ixgbe_if_enable_intr),
253 	DEVMETHOD(ifdi_intr_disable, ixgbe_if_disable_intr),
254 	DEVMETHOD(ifdi_link_intr_enable, ixgbe_link_intr_enable),
255 	DEVMETHOD(ifdi_tx_queue_intr_enable, ixgbe_if_rx_queue_intr_enable),
256 	DEVMETHOD(ifdi_rx_queue_intr_enable, ixgbe_if_rx_queue_intr_enable),
257 	DEVMETHOD(ifdi_tx_queues_alloc, ixgbe_if_tx_queues_alloc),
258 	DEVMETHOD(ifdi_rx_queues_alloc, ixgbe_if_rx_queues_alloc),
259 	DEVMETHOD(ifdi_queues_free, ixgbe_if_queues_free),
260 	DEVMETHOD(ifdi_update_admin_status, ixgbe_if_update_admin_status),
261 	DEVMETHOD(ifdi_multi_set, ixgbe_if_multi_set),
262 	DEVMETHOD(ifdi_mtu_set, ixgbe_if_mtu_set),
263 	DEVMETHOD(ifdi_crcstrip_set, ixgbe_if_crcstrip_set),
264 	DEVMETHOD(ifdi_media_status, ixgbe_if_media_status),
265 	DEVMETHOD(ifdi_media_change, ixgbe_if_media_change),
266 	DEVMETHOD(ifdi_promisc_set, ixgbe_if_promisc_set),
267 	DEVMETHOD(ifdi_timer, ixgbe_if_timer),
268 	DEVMETHOD(ifdi_vlan_register, ixgbe_if_vlan_register),
269 	DEVMETHOD(ifdi_vlan_unregister, ixgbe_if_vlan_unregister),
270 	DEVMETHOD(ifdi_get_counter, ixgbe_if_get_counter),
271 	DEVMETHOD(ifdi_i2c_req, ixgbe_if_i2c_req),
272 	DEVMETHOD(ifdi_needs_restart, ixgbe_if_needs_restart),
273 #ifdef PCI_IOV
274 	DEVMETHOD(ifdi_iov_init, ixgbe_if_iov_init),
275 	DEVMETHOD(ifdi_iov_uninit, ixgbe_if_iov_uninit),
276 	DEVMETHOD(ifdi_iov_vf_add, ixgbe_if_iov_vf_add),
277 #endif /* PCI_IOV */
278 	DEVMETHOD_END
279 };
280 
281 /*
282  * TUNEABLE PARAMETERS:
283  */
284 
285 static SYSCTL_NODE(_hw, OID_AUTO, ix, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
286     "IXGBE driver parameters");
287 static driver_t ixgbe_if_driver = {
288   "ixgbe_if", ixgbe_if_methods, sizeof(struct ixgbe_softc)
289 };
290 
291 static int ixgbe_max_interrupt_rate = (4000000 / IXGBE_LOW_LATENCY);
292 SYSCTL_INT(_hw_ix, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
293     &ixgbe_max_interrupt_rate, 0, "Maximum interrupts per second");
294 
295 /* Flow control setting, default to full */
296 static int ixgbe_flow_control = ixgbe_fc_full;
297 SYSCTL_INT(_hw_ix, OID_AUTO, flow_control, CTLFLAG_RDTUN,
298     &ixgbe_flow_control, 0, "Default flow control used for all adapters");
299 
300 /* Advertise Speed, default to 0 (auto) */
301 static int ixgbe_advertise_speed = 0;
302 SYSCTL_INT(_hw_ix, OID_AUTO, advertise_speed, CTLFLAG_RDTUN,
303     &ixgbe_advertise_speed, 0, "Default advertised speed for all adapters");
304 
305 /*
306  * Smart speed setting, default to on
307  * this only works as a compile option
308  * right now as its during attach, set
309  * this to 'ixgbe_smart_speed_off' to
310  * disable.
311  */
312 static int ixgbe_smart_speed = ixgbe_smart_speed_on;
313 
314 /*
315  * MSI-X should be the default for best performance,
316  * but this allows it to be forced off for testing.
317  */
318 static int ixgbe_enable_msix = 1;
319 SYSCTL_INT(_hw_ix, OID_AUTO, enable_msix, CTLFLAG_RDTUN, &ixgbe_enable_msix, 0,
320     "Enable MSI-X interrupts");
321 
322 /*
323  * Defining this on will allow the use
324  * of unsupported SFP+ modules, note that
325  * doing so you are on your own :)
326  */
327 static int allow_unsupported_sfp = false;
328 SYSCTL_INT(_hw_ix, OID_AUTO, unsupported_sfp, CTLFLAG_RDTUN,
329     &allow_unsupported_sfp, 0,
330     "Allow unsupported SFP modules...use at your own risk");
331 
332 /*
333  * Not sure if Flow Director is fully baked,
334  * so we'll default to turning it off.
335  */
336 static int ixgbe_enable_fdir = 0;
337 SYSCTL_INT(_hw_ix, OID_AUTO, enable_fdir, CTLFLAG_RDTUN, &ixgbe_enable_fdir, 0,
338     "Enable Flow Director");
339 
340 /* Receive-Side Scaling */
341 static int ixgbe_enable_rss = 1;
342 SYSCTL_INT(_hw_ix, OID_AUTO, enable_rss, CTLFLAG_RDTUN, &ixgbe_enable_rss, 0,
343     "Enable Receive-Side Scaling (RSS)");
344 
345 /*
346  * AIM: Adaptive Interrupt Moderation
347  * which means that the interrupt rate
348  * is varied over time based on the
349  * traffic for that interrupt vector
350  */
351 static int ixgbe_enable_aim = false;
352 SYSCTL_INT(_hw_ix, OID_AUTO, enable_aim, CTLFLAG_RWTUN, &ixgbe_enable_aim, 0,
353     "Enable adaptive interrupt moderation");
354 
355 #if 0
356 /* Keep running tab on them for sanity check */
357 static int ixgbe_total_ports;
358 #endif
359 
360 MALLOC_DEFINE(M_IXGBE, "ix", "ix driver allocations");
361 
362 /*
363  * For Flow Director: this is the number of TX packets we sample
364  * for the filter pool, this means every 20th packet will be probed.
365  *
366  * This feature can be disabled by setting this to 0.
367  */
368 static int atr_sample_rate = 20;
369 
370 extern struct if_txrx ixgbe_txrx;
371 
372 static struct if_shared_ctx ixgbe_sctx_init = {
373 	.isc_magic = IFLIB_MAGIC,
374 	.isc_q_align = PAGE_SIZE,/* max(DBA_ALIGN, PAGE_SIZE) */
375 	.isc_tx_maxsize = IXGBE_TSO_SIZE + sizeof(struct ether_vlan_header),
376 	.isc_tx_maxsegsize = PAGE_SIZE,
377 	.isc_tso_maxsize = IXGBE_TSO_SIZE + sizeof(struct ether_vlan_header),
378 	.isc_tso_maxsegsize = PAGE_SIZE,
379 	.isc_rx_maxsize = PAGE_SIZE*4,
380 	.isc_rx_nsegments = 1,
381 	.isc_rx_maxsegsize = PAGE_SIZE*4,
382 	.isc_nfl = 1,
383 	.isc_ntxqs = 1,
384 	.isc_nrxqs = 1,
385 
386 	.isc_admin_intrcnt = 1,
387 	.isc_vendor_info = ixgbe_vendor_info_array,
388 	.isc_driver_version = ixgbe_driver_version,
389 	.isc_driver = &ixgbe_if_driver,
390 	.isc_flags = IFLIB_TSO_INIT_IP,
391 
392 	.isc_nrxd_min = {MIN_RXD},
393 	.isc_ntxd_min = {MIN_TXD},
394 	.isc_nrxd_max = {MAX_RXD},
395 	.isc_ntxd_max = {MAX_TXD},
396 	.isc_nrxd_default = {DEFAULT_RXD},
397 	.isc_ntxd_default = {DEFAULT_TXD},
398 };
399 
400 /************************************************************************
401  * ixgbe_if_tx_queues_alloc
402  ************************************************************************/
403 static int
404 ixgbe_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs,
405     int ntxqs, int ntxqsets)
406 {
407 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
408 	if_softc_ctx_t     scctx = sc->shared;
409 	struct ix_tx_queue *que;
410 	int                i, j, error;
411 
412 	MPASS(sc->num_tx_queues > 0);
413 	MPASS(sc->num_tx_queues == ntxqsets);
414 	MPASS(ntxqs == 1);
415 
416 	/* Allocate queue structure memory */
417 	sc->tx_queues =
418 	    (struct ix_tx_queue *)malloc(sizeof(struct ix_tx_queue) * ntxqsets,
419 	                                 M_IXGBE, M_NOWAIT | M_ZERO);
420 	if (!sc->tx_queues) {
421 		device_printf(iflib_get_dev(ctx),
422 		    "Unable to allocate TX ring memory\n");
423 		return (ENOMEM);
424 	}
425 
426 	for (i = 0, que = sc->tx_queues; i < ntxqsets; i++, que++) {
427 		struct tx_ring *txr = &que->txr;
428 
429 		/* In case SR-IOV is enabled, align the index properly */
430 		txr->me = ixgbe_vf_que_index(sc->iov_mode, sc->pool,
431 		    i);
432 
433 		txr->sc = que->sc = sc;
434 
435 		/* Allocate report status array */
436 		txr->tx_rsq = (qidx_t *)malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_IXGBE, M_NOWAIT | M_ZERO);
437 		if (txr->tx_rsq == NULL) {
438 			error = ENOMEM;
439 			goto fail;
440 		}
441 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
442 			txr->tx_rsq[j] = QIDX_INVALID;
443 		/* get the virtual and physical address of the hardware queues */
444 		txr->tail = IXGBE_TDT(txr->me);
445 		txr->tx_base = (union ixgbe_adv_tx_desc *)vaddrs[i];
446 		txr->tx_paddr = paddrs[i];
447 
448 		txr->bytes = 0;
449 		txr->total_packets = 0;
450 
451 		/* Set the rate at which we sample packets */
452 		if (sc->feat_en & IXGBE_FEATURE_FDIR)
453 			txr->atr_sample = atr_sample_rate;
454 
455 	}
456 
457 	device_printf(iflib_get_dev(ctx), "allocated for %d queues\n",
458 	    sc->num_tx_queues);
459 
460 	return (0);
461 
462 fail:
463 	ixgbe_if_queues_free(ctx);
464 
465 	return (error);
466 } /* ixgbe_if_tx_queues_alloc */
467 
468 /************************************************************************
469  * ixgbe_if_rx_queues_alloc
470  ************************************************************************/
471 static int
472 ixgbe_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs,
473     int nrxqs, int nrxqsets)
474 {
475 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
476 	struct ix_rx_queue *que;
477 	int                i;
478 
479 	MPASS(sc->num_rx_queues > 0);
480 	MPASS(sc->num_rx_queues == nrxqsets);
481 	MPASS(nrxqs == 1);
482 
483 	/* Allocate queue structure memory */
484 	sc->rx_queues =
485 	    (struct ix_rx_queue *)malloc(sizeof(struct ix_rx_queue)*nrxqsets,
486 	                                 M_IXGBE, M_NOWAIT | M_ZERO);
487 	if (!sc->rx_queues) {
488 		device_printf(iflib_get_dev(ctx),
489 		    "Unable to allocate TX ring memory\n");
490 		return (ENOMEM);
491 	}
492 
493 	for (i = 0, que = sc->rx_queues; i < nrxqsets; i++, que++) {
494 		struct rx_ring *rxr = &que->rxr;
495 
496 		/* In case SR-IOV is enabled, align the index properly */
497 		rxr->me = ixgbe_vf_que_index(sc->iov_mode, sc->pool,
498 		    i);
499 
500 		rxr->sc = que->sc = sc;
501 
502 		/* get the virtual and physical address of the hw queues */
503 		rxr->tail = IXGBE_RDT(rxr->me);
504 		rxr->rx_base = (union ixgbe_adv_rx_desc *)vaddrs[i];
505 		rxr->rx_paddr = paddrs[i];
506 		rxr->bytes = 0;
507 		rxr->que = que;
508 	}
509 
510 	device_printf(iflib_get_dev(ctx), "allocated for %d rx queues\n",
511 	    sc->num_rx_queues);
512 
513 	return (0);
514 } /* ixgbe_if_rx_queues_alloc */
515 
516 /************************************************************************
517  * ixgbe_if_queues_free
518  ************************************************************************/
519 static void
520 ixgbe_if_queues_free(if_ctx_t ctx)
521 {
522 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
523 	struct ix_tx_queue *tx_que = sc->tx_queues;
524 	struct ix_rx_queue *rx_que = sc->rx_queues;
525 	int                i;
526 
527 	if (tx_que != NULL) {
528 		for (i = 0; i < sc->num_tx_queues; i++, tx_que++) {
529 			struct tx_ring *txr = &tx_que->txr;
530 			if (txr->tx_rsq == NULL)
531 				break;
532 
533 			free(txr->tx_rsq, M_IXGBE);
534 			txr->tx_rsq = NULL;
535 		}
536 
537 		free(sc->tx_queues, M_IXGBE);
538 		sc->tx_queues = NULL;
539 	}
540 	if (rx_que != NULL) {
541 		free(sc->rx_queues, M_IXGBE);
542 		sc->rx_queues = NULL;
543 	}
544 } /* ixgbe_if_queues_free */
545 
546 /************************************************************************
547  * ixgbe_initialize_rss_mapping
548  ************************************************************************/
549 static void
550 ixgbe_initialize_rss_mapping(struct ixgbe_softc *sc)
551 {
552 	struct ixgbe_hw *hw = &sc->hw;
553 	u32             reta = 0, mrqc, rss_key[10];
554 	int             queue_id, table_size, index_mult;
555 	int             i, j;
556 	u32             rss_hash_config;
557 
558 	if (sc->feat_en & IXGBE_FEATURE_RSS) {
559 		/* Fetch the configured RSS key */
560 		rss_getkey((uint8_t *)&rss_key);
561 	} else {
562 		/* set up random bits */
563 		arc4rand(&rss_key, sizeof(rss_key), 0);
564 	}
565 
566 	/* Set multiplier for RETA setup and table size based on MAC */
567 	index_mult = 0x1;
568 	table_size = 128;
569 	switch (sc->hw.mac.type) {
570 	case ixgbe_mac_82598EB:
571 		index_mult = 0x11;
572 		break;
573 	case ixgbe_mac_X550:
574 	case ixgbe_mac_X550EM_x:
575 	case ixgbe_mac_X550EM_a:
576 		table_size = 512;
577 		break;
578 	default:
579 		break;
580 	}
581 
582 	/* Set up the redirection table */
583 	for (i = 0, j = 0; i < table_size; i++, j++) {
584 		if (j == sc->num_rx_queues)
585 			j = 0;
586 
587 		if (sc->feat_en & IXGBE_FEATURE_RSS) {
588 			/*
589 			 * Fetch the RSS bucket id for the given indirection
590 			 * entry. Cap it at the number of configured buckets
591 			 * (which is num_rx_queues.)
592 			 */
593 			queue_id = rss_get_indirection_to_bucket(i);
594 			queue_id = queue_id % sc->num_rx_queues;
595 		} else
596 			queue_id = (j * index_mult);
597 
598 		/*
599 		 * The low 8 bits are for hash value (n+0);
600 		 * The next 8 bits are for hash value (n+1), etc.
601 		 */
602 		reta = reta >> 8;
603 		reta = reta | (((uint32_t)queue_id) << 24);
604 		if ((i & 3) == 3) {
605 			if (i < 128)
606 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
607 			else
608 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
609 				    reta);
610 			reta = 0;
611 		}
612 	}
613 
614 	/* Now fill our hash function seeds */
615 	for (i = 0; i < 10; i++)
616 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), rss_key[i]);
617 
618 	/* Perform hash on these packet types */
619 	if (sc->feat_en & IXGBE_FEATURE_RSS)
620 		rss_hash_config = rss_gethashconfig();
621 	else {
622 		/*
623 		 * Disable UDP - IP fragments aren't currently being handled
624 		 * and so we end up with a mix of 2-tuple and 4-tuple
625 		 * traffic.
626 		 */
627 		rss_hash_config = RSS_HASHTYPE_RSS_IPV4
628 		                | RSS_HASHTYPE_RSS_TCP_IPV4
629 		                | RSS_HASHTYPE_RSS_IPV6
630 		                | RSS_HASHTYPE_RSS_TCP_IPV6
631 		                | RSS_HASHTYPE_RSS_IPV6_EX
632 		                | RSS_HASHTYPE_RSS_TCP_IPV6_EX;
633 	}
634 
635 	mrqc = IXGBE_MRQC_RSSEN;
636 	if (rss_hash_config & RSS_HASHTYPE_RSS_IPV4)
637 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4;
638 	if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV4)
639 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_TCP;
640 	if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6)
641 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6;
642 	if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6)
643 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
644 	if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6_EX)
645 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX;
646 	if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6_EX)
647 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_TCP;
648 	if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV4)
649 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
650 	if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6)
651 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
652 	if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6_EX)
653 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_UDP;
654 	mrqc |= ixgbe_get_mrqc(sc->iov_mode);
655 	IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
656 } /* ixgbe_initialize_rss_mapping */
657 
658 /************************************************************************
659  * ixgbe_initialize_receive_units - Setup receive registers and features.
660  ************************************************************************/
661 #define BSIZEPKT_ROUNDUP ((1<<IXGBE_SRRCTL_BSIZEPKT_SHIFT)-1)
662 
663 static void
664 ixgbe_initialize_receive_units(if_ctx_t ctx)
665 {
666 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
667 	if_softc_ctx_t     scctx = sc->shared;
668 	struct ixgbe_hw    *hw = &sc->hw;
669 	if_t               ifp = iflib_get_ifp(ctx);
670 	struct ix_rx_queue *que;
671 	int                i, j;
672 	u32                bufsz, fctrl, srrctl, rxcsum;
673 	u32                hlreg;
674 
675 	/*
676 	 * Make sure receives are disabled while
677 	 * setting up the descriptor ring
678 	 */
679 	ixgbe_disable_rx(hw);
680 
681 	/* Enable broadcasts */
682 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
683 	fctrl |= IXGBE_FCTRL_BAM;
684 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
685 		fctrl |= IXGBE_FCTRL_DPF;
686 		fctrl |= IXGBE_FCTRL_PMCF;
687 	}
688 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
689 
690 	/* Set for Jumbo Frames? */
691 	hlreg = IXGBE_READ_REG(hw, IXGBE_HLREG0);
692 	if (if_getmtu(ifp) > ETHERMTU)
693 		hlreg |= IXGBE_HLREG0_JUMBOEN;
694 	else
695 		hlreg &= ~IXGBE_HLREG0_JUMBOEN;
696 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg);
697 
698 	bufsz = (sc->rx_mbuf_sz + BSIZEPKT_ROUNDUP) >>
699 	    IXGBE_SRRCTL_BSIZEPKT_SHIFT;
700 
701 	/* Setup the Base and Length of the Rx Descriptor Ring */
702 	for (i = 0, que = sc->rx_queues; i < sc->num_rx_queues; i++, que++) {
703 		struct rx_ring *rxr = &que->rxr;
704 		u64            rdba = rxr->rx_paddr;
705 
706 		j = rxr->me;
707 
708 		/* Setup the Base and Length of the Rx Descriptor Ring */
709 		IXGBE_WRITE_REG(hw, IXGBE_RDBAL(j),
710 		    (rdba & 0x00000000ffffffffULL));
711 		IXGBE_WRITE_REG(hw, IXGBE_RDBAH(j), (rdba >> 32));
712 		IXGBE_WRITE_REG(hw, IXGBE_RDLEN(j),
713 		     scctx->isc_nrxd[0] * sizeof(union ixgbe_adv_rx_desc));
714 
715 		/* Set up the SRRCTL register */
716 		srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(j));
717 		srrctl &= ~IXGBE_SRRCTL_BSIZEHDR_MASK;
718 		srrctl &= ~IXGBE_SRRCTL_BSIZEPKT_MASK;
719 		srrctl |= bufsz;
720 		srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
721 
722 		/*
723 		 * Set DROP_EN iff we have no flow control and >1 queue.
724 		 * Note that srrctl was cleared shortly before during reset,
725 		 * so we do not need to clear the bit, but do it just in case
726 		 * this code is moved elsewhere.
727 		 */
728 		if (sc->num_rx_queues > 1 &&
729 		    sc->hw.fc.requested_mode == ixgbe_fc_none) {
730 			srrctl |= IXGBE_SRRCTL_DROP_EN;
731 		} else {
732 			srrctl &= ~IXGBE_SRRCTL_DROP_EN;
733 		}
734 
735 		IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(j), srrctl);
736 
737 		/* Setup the HW Rx Head and Tail Descriptor Pointers */
738 		IXGBE_WRITE_REG(hw, IXGBE_RDH(j), 0);
739 		IXGBE_WRITE_REG(hw, IXGBE_RDT(j), 0);
740 
741 		/* Set the driver rx tail address */
742 		rxr->tail =  IXGBE_RDT(rxr->me);
743 	}
744 
745 	if (sc->hw.mac.type != ixgbe_mac_82598EB) {
746 		u32 psrtype = IXGBE_PSRTYPE_TCPHDR
747 		            | IXGBE_PSRTYPE_UDPHDR
748 		            | IXGBE_PSRTYPE_IPV4HDR
749 		            | IXGBE_PSRTYPE_IPV6HDR;
750 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(0), psrtype);
751 	}
752 
753 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
754 
755 	ixgbe_initialize_rss_mapping(sc);
756 
757 	if (sc->feat_en & IXGBE_FEATURE_RSS) {
758 		/* RSS and RX IPP Checksum are mutually exclusive */
759 		rxcsum |= IXGBE_RXCSUM_PCSD;
760 	}
761 
762 	if (if_getcapenable(ifp) & IFCAP_RXCSUM)
763 		rxcsum |= IXGBE_RXCSUM_PCSD;
764 
765 	/* This is useful for calculating UDP/IP fragment checksums */
766 	if (!(rxcsum & IXGBE_RXCSUM_PCSD))
767 		rxcsum |= IXGBE_RXCSUM_IPPCSE;
768 
769 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
770 
771 } /* ixgbe_initialize_receive_units */
772 
773 /************************************************************************
774  * ixgbe_initialize_transmit_units - Enable transmit units.
775  ************************************************************************/
776 static void
777 ixgbe_initialize_transmit_units(if_ctx_t ctx)
778 {
779 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
780 	struct ixgbe_hw    *hw = &sc->hw;
781 	if_softc_ctx_t     scctx = sc->shared;
782 	struct ix_tx_queue *que;
783 	int i;
784 
785 	/* Setup the Base and Length of the Tx Descriptor Ring */
786 	for (i = 0, que = sc->tx_queues; i < sc->num_tx_queues;
787 	    i++, que++) {
788 		struct tx_ring	   *txr = &que->txr;
789 		u64 tdba = txr->tx_paddr;
790 		u32 txctrl = 0;
791 		int j = txr->me;
792 
793 		IXGBE_WRITE_REG(hw, IXGBE_TDBAL(j),
794 		    (tdba & 0x00000000ffffffffULL));
795 		IXGBE_WRITE_REG(hw, IXGBE_TDBAH(j), (tdba >> 32));
796 		IXGBE_WRITE_REG(hw, IXGBE_TDLEN(j),
797 		    scctx->isc_ntxd[0] * sizeof(union ixgbe_adv_tx_desc));
798 
799 		/* Setup the HW Tx Head and Tail descriptor pointers */
800 		IXGBE_WRITE_REG(hw, IXGBE_TDH(j), 0);
801 		IXGBE_WRITE_REG(hw, IXGBE_TDT(j), 0);
802 
803 		/* Cache the tail address */
804 		txr->tail = IXGBE_TDT(txr->me);
805 
806 		txr->tx_rs_cidx = txr->tx_rs_pidx;
807 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
808 		for (int k = 0; k < scctx->isc_ntxd[0]; k++)
809 			txr->tx_rsq[k] = QIDX_INVALID;
810 
811 		/* Disable Head Writeback */
812 		/*
813 		 * Note: for X550 series devices, these registers are actually
814 		 * prefixed with TPH_ isntead of DCA_, but the addresses and
815 		 * fields remain the same.
816 		 */
817 		switch (hw->mac.type) {
818 		case ixgbe_mac_82598EB:
819 			txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL(j));
820 			break;
821 		default:
822 			txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL_82599(j));
823 			break;
824 		}
825 		txctrl &= ~IXGBE_DCA_TXCTRL_DESC_WRO_EN;
826 		switch (hw->mac.type) {
827 		case ixgbe_mac_82598EB:
828 			IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL(j), txctrl);
829 			break;
830 		default:
831 			IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL_82599(j), txctrl);
832 			break;
833 		}
834 
835 	}
836 
837 	if (hw->mac.type != ixgbe_mac_82598EB) {
838 		u32 dmatxctl, rttdcs;
839 
840 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
841 		dmatxctl |= IXGBE_DMATXCTL_TE;
842 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
843 		/* Disable arbiter to set MTQC */
844 		rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
845 		rttdcs |= IXGBE_RTTDCS_ARBDIS;
846 		IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
847 		IXGBE_WRITE_REG(hw, IXGBE_MTQC,
848 		    ixgbe_get_mtqc(sc->iov_mode));
849 		rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
850 		IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
851 	}
852 
853 } /* ixgbe_initialize_transmit_units */
854 
855 /************************************************************************
856  * ixgbe_register
857  ************************************************************************/
858 static void *
859 ixgbe_register(device_t dev)
860 {
861 	return (&ixgbe_sctx_init);
862 } /* ixgbe_register */
863 
864 /************************************************************************
865  * ixgbe_if_attach_pre - Device initialization routine, part 1
866  *
867  *   Called when the driver is being loaded.
868  *   Identifies the type of hardware, initializes the hardware,
869  *   and initializes iflib structures.
870  *
871  *   return 0 on success, positive on failure
872  ************************************************************************/
873 static int
874 ixgbe_if_attach_pre(if_ctx_t ctx)
875 {
876 	struct ixgbe_softc  *sc;
877 	device_t        dev;
878 	if_softc_ctx_t  scctx;
879 	struct ixgbe_hw *hw;
880 	int             error = 0;
881 	u32             ctrl_ext;
882 
883 	INIT_DEBUGOUT("ixgbe_attach: begin");
884 
885 	/* Allocate, clear, and link in our adapter structure */
886 	dev = iflib_get_dev(ctx);
887 	sc = iflib_get_softc(ctx);
888 	sc->hw.back = sc;
889 	sc->ctx = ctx;
890 	sc->dev = dev;
891 	scctx = sc->shared = iflib_get_softc_ctx(ctx);
892 	sc->media = iflib_get_media(ctx);
893 	hw = &sc->hw;
894 
895 	/* Determine hardware revision */
896 	hw->vendor_id = pci_get_vendor(dev);
897 	hw->device_id = pci_get_device(dev);
898 	hw->revision_id = pci_get_revid(dev);
899 	hw->subsystem_vendor_id = pci_get_subvendor(dev);
900 	hw->subsystem_device_id = pci_get_subdevice(dev);
901 
902 	/* Do base PCI setup - map BAR0 */
903 	if (ixgbe_allocate_pci_resources(ctx)) {
904 		device_printf(dev, "Allocation of PCI resources failed\n");
905 		return (ENXIO);
906 	}
907 
908 	/* let hardware know driver is loaded */
909 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
910 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
911 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
912 
913 	/*
914 	 * Initialize the shared code
915 	 */
916 	if (ixgbe_init_shared_code(hw) != 0) {
917 		device_printf(dev, "Unable to initialize the shared code\n");
918 		error = ENXIO;
919 		goto err_pci;
920 	}
921 
922 	if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
923 		device_printf(dev, "Firmware recovery mode detected. Limiting "
924 		    "functionality.\nRefer to the Intel(R) Ethernet Adapters "
925 		    "and Devices User Guide for details on firmware recovery "
926 		    "mode.");
927 		error = ENOSYS;
928 		goto err_pci;
929 	}
930 
931 	if (hw->mbx.ops.init_params)
932 		hw->mbx.ops.init_params(hw);
933 
934 	hw->allow_unsupported_sfp = allow_unsupported_sfp;
935 
936 	if (hw->mac.type != ixgbe_mac_82598EB)
937 		hw->phy.smart_speed = ixgbe_smart_speed;
938 
939 	ixgbe_init_device_features(sc);
940 
941 	/* Enable WoL (if supported) */
942 	ixgbe_check_wol_support(sc);
943 
944 	/* Verify adapter fan is still functional (if applicable) */
945 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL) {
946 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
947 		ixgbe_check_fan_failure(sc, esdp, false);
948 	}
949 
950 	/* Ensure SW/FW semaphore is free */
951 	ixgbe_init_swfw_semaphore(hw);
952 
953 	/* Set an initial default flow control value */
954 	hw->fc.requested_mode = ixgbe_flow_control;
955 
956 	hw->phy.reset_if_overtemp = true;
957 	error = ixgbe_reset_hw(hw);
958 	hw->phy.reset_if_overtemp = false;
959 	if (error == IXGBE_ERR_SFP_NOT_PRESENT) {
960 		/*
961 		 * No optics in this port, set up
962 		 * so the timer routine will probe
963 		 * for later insertion.
964 		 */
965 		sc->sfp_probe = true;
966 		error = 0;
967 	} else if (error == IXGBE_ERR_SFP_NOT_SUPPORTED) {
968 		device_printf(dev, "Unsupported SFP+ module detected!\n");
969 		error = EIO;
970 		goto err_pci;
971 	} else if (error) {
972 		device_printf(dev, "Hardware initialization failed\n");
973 		error = EIO;
974 		goto err_pci;
975 	}
976 
977 	/* Make sure we have a good EEPROM before we read from it */
978 	if (ixgbe_validate_eeprom_checksum(&sc->hw, NULL) < 0) {
979 		device_printf(dev, "The EEPROM Checksum Is Not Valid\n");
980 		error = EIO;
981 		goto err_pci;
982 	}
983 
984 	error = ixgbe_start_hw(hw);
985 	switch (error) {
986 	case IXGBE_ERR_EEPROM_VERSION:
987 		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");
988 		break;
989 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
990 		device_printf(dev, "Unsupported SFP+ Module\n");
991 		error = EIO;
992 		goto err_pci;
993 	case IXGBE_ERR_SFP_NOT_PRESENT:
994 		device_printf(dev, "No SFP+ Module found\n");
995 		/* falls thru */
996 	default:
997 		break;
998 	}
999 
1000 	/* Most of the iflib initialization... */
1001 
1002 	iflib_set_mac(ctx, hw->mac.addr);
1003 	switch (sc->hw.mac.type) {
1004 	case ixgbe_mac_X550:
1005 	case ixgbe_mac_X550EM_x:
1006 	case ixgbe_mac_X550EM_a:
1007 		scctx->isc_rss_table_size = 512;
1008 		scctx->isc_ntxqsets_max = scctx->isc_nrxqsets_max = 64;
1009 		break;
1010 	default:
1011 		scctx->isc_rss_table_size = 128;
1012 		scctx->isc_ntxqsets_max = scctx->isc_nrxqsets_max = 16;
1013 	}
1014 
1015 	/* Allow legacy interrupts */
1016 	ixgbe_txrx.ift_legacy_intr = ixgbe_intr;
1017 
1018 	scctx->isc_txqsizes[0] =
1019 	    roundup2(scctx->isc_ntxd[0] * sizeof(union ixgbe_adv_tx_desc) +
1020 	    sizeof(u32), DBA_ALIGN),
1021 	scctx->isc_rxqsizes[0] =
1022 	    roundup2(scctx->isc_nrxd[0] * sizeof(union ixgbe_adv_rx_desc),
1023 	    DBA_ALIGN);
1024 
1025 	/* XXX */
1026 	scctx->isc_tx_csum_flags = CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_TSO |
1027 	    CSUM_IP6_TCP | CSUM_IP6_UDP | CSUM_IP6_TSO;
1028 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
1029 		scctx->isc_tx_nsegments = IXGBE_82598_SCATTER;
1030 	} else {
1031 		scctx->isc_tx_csum_flags |= CSUM_SCTP |CSUM_IP6_SCTP;
1032 		scctx->isc_tx_nsegments = IXGBE_82599_SCATTER;
1033 	}
1034 
1035 	scctx->isc_msix_bar = pci_msix_table_bar(dev);
1036 
1037 	scctx->isc_tx_tso_segments_max = scctx->isc_tx_nsegments;
1038 	scctx->isc_tx_tso_size_max = IXGBE_TSO_SIZE;
1039 	scctx->isc_tx_tso_segsize_max = PAGE_SIZE;
1040 
1041 	scctx->isc_txrx = &ixgbe_txrx;
1042 
1043 	scctx->isc_capabilities = scctx->isc_capenable = IXGBE_CAPS;
1044 
1045 	return (0);
1046 
1047 err_pci:
1048 	ctrl_ext = IXGBE_READ_REG(&sc->hw, IXGBE_CTRL_EXT);
1049 	ctrl_ext &= ~IXGBE_CTRL_EXT_DRV_LOAD;
1050 	IXGBE_WRITE_REG(&sc->hw, IXGBE_CTRL_EXT, ctrl_ext);
1051 	ixgbe_free_pci_resources(ctx);
1052 
1053 	return (error);
1054 } /* ixgbe_if_attach_pre */
1055 
1056  /*********************************************************************
1057  * ixgbe_if_attach_post - Device initialization routine, part 2
1058  *
1059  *   Called during driver load, but after interrupts and
1060  *   resources have been allocated and configured.
1061  *   Sets up some data structures not relevant to iflib.
1062  *
1063  *   return 0 on success, positive on failure
1064  *********************************************************************/
1065 static int
1066 ixgbe_if_attach_post(if_ctx_t ctx)
1067 {
1068 	device_t dev;
1069 	struct ixgbe_softc  *sc;
1070 	struct ixgbe_hw *hw;
1071 	int             error = 0;
1072 
1073 	dev = iflib_get_dev(ctx);
1074 	sc = iflib_get_softc(ctx);
1075 	hw = &sc->hw;
1076 
1077 
1078 	if (sc->intr_type == IFLIB_INTR_LEGACY &&
1079 		(sc->feat_cap & IXGBE_FEATURE_LEGACY_IRQ) == 0) {
1080 		device_printf(dev, "Device does not support legacy interrupts");
1081 		error = ENXIO;
1082 		goto err;
1083 	}
1084 
1085 	/* Allocate multicast array memory. */
1086 	sc->mta = malloc(sizeof(*sc->mta) *
1087 	                      MAX_NUM_MULTICAST_ADDRESSES, M_IXGBE, M_NOWAIT);
1088 	if (sc->mta == NULL) {
1089 		device_printf(dev, "Can not allocate multicast setup array\n");
1090 		error = ENOMEM;
1091 		goto err;
1092 	}
1093 
1094 	/* hw.ix defaults init */
1095 	ixgbe_set_advertise(sc, ixgbe_advertise_speed);
1096 
1097 	/* Enable the optics for 82599 SFP+ fiber */
1098 	ixgbe_enable_tx_laser(hw);
1099 
1100 	/* Enable power to the phy. */
1101 	ixgbe_set_phy_power(hw, true);
1102 
1103 	ixgbe_initialize_iov(sc);
1104 
1105 	error = ixgbe_setup_interface(ctx);
1106 	if (error) {
1107 		device_printf(dev, "Interface setup failed: %d\n", error);
1108 		goto err;
1109 	}
1110 
1111 	ixgbe_if_update_admin_status(ctx);
1112 
1113 	/* Initialize statistics */
1114 	ixgbe_update_stats_counters(sc);
1115 	ixgbe_add_hw_stats(sc);
1116 
1117 	/* Check PCIE slot type/speed/width */
1118 	ixgbe_get_slot_info(sc);
1119 
1120 	/*
1121 	 * Do time init and sysctl init here, but
1122 	 * only on the first port of a bypass sc.
1123 	 */
1124 	ixgbe_bypass_init(sc);
1125 
1126 	/* Display NVM and Option ROM versions */
1127 	ixgbe_print_fw_version(ctx);
1128 
1129 	/* Set an initial dmac value */
1130 	sc->dmac = 0;
1131 	/* Set initial advertised speeds (if applicable) */
1132 	sc->advertise = ixgbe_get_default_advertise(sc);
1133 
1134 	if (sc->feat_cap & IXGBE_FEATURE_SRIOV)
1135 		ixgbe_define_iov_schemas(dev, &error);
1136 
1137 	/* Add sysctls */
1138 	ixgbe_add_device_sysctls(ctx);
1139 
1140 	return (0);
1141 err:
1142 	return (error);
1143 } /* ixgbe_if_attach_post */
1144 
1145 /************************************************************************
1146  * ixgbe_check_wol_support
1147  *
1148  *   Checks whether the adapter's ports are capable of
1149  *   Wake On LAN by reading the adapter's NVM.
1150  *
1151  *   Sets each port's hw->wol_enabled value depending
1152  *   on the value read here.
1153  ************************************************************************/
1154 static void
1155 ixgbe_check_wol_support(struct ixgbe_softc *sc)
1156 {
1157 	struct ixgbe_hw *hw = &sc->hw;
1158 	u16             dev_caps = 0;
1159 
1160 	/* Find out WoL support for port */
1161 	sc->wol_support = hw->wol_enabled = 0;
1162 	ixgbe_get_device_caps(hw, &dev_caps);
1163 	if ((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
1164 	    ((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0) &&
1165 	     hw->bus.func == 0))
1166 		sc->wol_support = hw->wol_enabled = 1;
1167 
1168 	/* Save initial wake up filter configuration */
1169 	sc->wufc = IXGBE_READ_REG(hw, IXGBE_WUFC);
1170 
1171 	return;
1172 } /* ixgbe_check_wol_support */
1173 
1174 /************************************************************************
1175  * ixgbe_setup_interface
1176  *
1177  *   Setup networking device structure and register an interface.
1178  ************************************************************************/
1179 static int
1180 ixgbe_setup_interface(if_ctx_t ctx)
1181 {
1182 	if_t               ifp = iflib_get_ifp(ctx);
1183 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1184 
1185 	INIT_DEBUGOUT("ixgbe_setup_interface: begin");
1186 
1187 	if_setbaudrate(ifp, IF_Gbps(10));
1188 
1189 	sc->max_frame_size = if_getmtu(ifp) + ETHER_HDR_LEN + ETHER_CRC_LEN;
1190 
1191 	sc->phy_layer = ixgbe_get_supported_physical_layer(&sc->hw);
1192 
1193 	ixgbe_add_media_types(ctx);
1194 
1195 	/* Autoselect media by default */
1196 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
1197 
1198 	return (0);
1199 } /* ixgbe_setup_interface */
1200 
1201 /************************************************************************
1202  * ixgbe_if_get_counter
1203  ************************************************************************/
1204 static uint64_t
1205 ixgbe_if_get_counter(if_ctx_t ctx, ift_counter cnt)
1206 {
1207 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1208 	if_t           ifp = iflib_get_ifp(ctx);
1209 
1210 	switch (cnt) {
1211 	case IFCOUNTER_IPACKETS:
1212 		return (sc->ipackets);
1213 	case IFCOUNTER_OPACKETS:
1214 		return (sc->opackets);
1215 	case IFCOUNTER_IBYTES:
1216 		return (sc->ibytes);
1217 	case IFCOUNTER_OBYTES:
1218 		return (sc->obytes);
1219 	case IFCOUNTER_IMCASTS:
1220 		return (sc->imcasts);
1221 	case IFCOUNTER_OMCASTS:
1222 		return (sc->omcasts);
1223 	case IFCOUNTER_COLLISIONS:
1224 		return (0);
1225 	case IFCOUNTER_IQDROPS:
1226 		return (sc->iqdrops);
1227 	case IFCOUNTER_OQDROPS:
1228 		return (0);
1229 	case IFCOUNTER_IERRORS:
1230 		return (sc->ierrors);
1231 	default:
1232 		return (if_get_counter_default(ifp, cnt));
1233 	}
1234 } /* ixgbe_if_get_counter */
1235 
1236 /************************************************************************
1237  * ixgbe_if_i2c_req
1238  ************************************************************************/
1239 static int
1240 ixgbe_if_i2c_req(if_ctx_t ctx, struct ifi2creq *req)
1241 {
1242 	struct ixgbe_softc		*sc = iflib_get_softc(ctx);
1243 	struct ixgbe_hw 	*hw = &sc->hw;
1244 	int 			i;
1245 
1246 
1247 	if (hw->phy.ops.read_i2c_byte == NULL)
1248 		return (ENXIO);
1249 	for (i = 0; i < req->len; i++)
1250 		hw->phy.ops.read_i2c_byte(hw, req->offset + i,
1251 		    req->dev_addr, &req->data[i]);
1252 	return (0);
1253 } /* ixgbe_if_i2c_req */
1254 
1255 /* ixgbe_if_needs_restart - Tell iflib when the driver needs to be reinitialized
1256  * @ctx: iflib context
1257  * @event: event code to check
1258  *
1259  * Defaults to returning false for unknown events.
1260  *
1261  * @returns true if iflib needs to reinit the interface
1262  */
1263 static bool
1264 ixgbe_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
1265 {
1266 	switch (event) {
1267 	case IFLIB_RESTART_VLAN_CONFIG:
1268 	default:
1269 		return (false);
1270 	}
1271 }
1272 
1273 /************************************************************************
1274  * ixgbe_add_media_types
1275  ************************************************************************/
1276 static void
1277 ixgbe_add_media_types(if_ctx_t ctx)
1278 {
1279 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
1280 	struct ixgbe_hw *hw = &sc->hw;
1281 	device_t        dev = iflib_get_dev(ctx);
1282 	u64             layer;
1283 
1284 	layer = sc->phy_layer = ixgbe_get_supported_physical_layer(hw);
1285 
1286 	/* Media types with matching FreeBSD media defines */
1287 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T)
1288 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_T, 0, NULL);
1289 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_T)
1290 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
1291 	if (layer & IXGBE_PHYSICAL_LAYER_100BASE_TX)
1292 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
1293 	if (layer & IXGBE_PHYSICAL_LAYER_10BASE_T)
1294 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
1295 
1296 	if (hw->mac.type == ixgbe_mac_X550) {
1297 		ifmedia_add(sc->media, IFM_ETHER | IFM_2500_T, 0, NULL);
1298 		ifmedia_add(sc->media, IFM_ETHER | IFM_5000_T, 0, NULL);
1299 	}
1300 
1301 	if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU ||
1302 	    layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA)
1303 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_TWINAX, 0,
1304 		    NULL);
1305 
1306 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR) {
1307 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_LR, 0, NULL);
1308 		if (hw->phy.multispeed_fiber)
1309 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_LX, 0,
1310 			    NULL);
1311 	}
1312 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR) {
1313 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_SR, 0, NULL);
1314 		if (hw->phy.multispeed_fiber)
1315 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_SX, 0,
1316 			    NULL);
1317 	} else if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX)
1318 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_SX, 0, NULL);
1319 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4)
1320 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_CX4, 0, NULL);
1321 
1322 #ifdef IFM_ETH_XTYPE
1323 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
1324 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_KR, 0, NULL);
1325 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4)
1326 		ifmedia_add( sc->media, IFM_ETHER | IFM_10G_KX4, 0, NULL);
1327 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
1328 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_KX, 0, NULL);
1329 	if (layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX)
1330 		ifmedia_add(sc->media, IFM_ETHER | IFM_2500_KX, 0, NULL);
1331 #else
1332 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR) {
1333 		device_printf(dev, "Media supported: 10GbaseKR\n");
1334 		device_printf(dev, "10GbaseKR mapped to 10GbaseSR\n");
1335 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_SR, 0, NULL);
1336 	}
1337 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4) {
1338 		device_printf(dev, "Media supported: 10GbaseKX4\n");
1339 		device_printf(dev, "10GbaseKX4 mapped to 10GbaseCX4\n");
1340 		ifmedia_add(sc->media, IFM_ETHER | IFM_10G_CX4, 0, NULL);
1341 	}
1342 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) {
1343 		device_printf(dev, "Media supported: 1000baseKX\n");
1344 		device_printf(dev, "1000baseKX mapped to 1000baseCX\n");
1345 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_CX, 0, NULL);
1346 	}
1347 	if (layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX) {
1348 		device_printf(dev, "Media supported: 2500baseKX\n");
1349 		device_printf(dev, "2500baseKX mapped to 2500baseSX\n");
1350 		ifmedia_add(sc->media, IFM_ETHER | IFM_2500_SX, 0, NULL);
1351 	}
1352 #endif
1353 	if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_BX)
1354 		device_printf(dev, "Media supported: 1000baseBX\n");
1355 
1356 	if (hw->device_id == IXGBE_DEV_ID_82598AT) {
1357 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T | IFM_FDX,
1358 		    0, NULL);
1359 		ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
1360 	}
1361 
1362 	ifmedia_add(sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
1363 } /* ixgbe_add_media_types */
1364 
1365 /************************************************************************
1366  * ixgbe_is_sfp
1367  ************************************************************************/
1368 static inline bool
1369 ixgbe_is_sfp(struct ixgbe_hw *hw)
1370 {
1371 	switch (hw->mac.type) {
1372 	case ixgbe_mac_82598EB:
1373 		if (hw->phy.type == ixgbe_phy_nl)
1374 			return (true);
1375 		return (false);
1376 	case ixgbe_mac_82599EB:
1377 		switch (hw->mac.ops.get_media_type(hw)) {
1378 		case ixgbe_media_type_fiber:
1379 		case ixgbe_media_type_fiber_qsfp:
1380 			return (true);
1381 		default:
1382 			return (false);
1383 		}
1384 	case ixgbe_mac_X550EM_x:
1385 	case ixgbe_mac_X550EM_a:
1386 		if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber)
1387 			return (true);
1388 		return (false);
1389 	default:
1390 		return (false);
1391 	}
1392 } /* ixgbe_is_sfp */
1393 
1394 /************************************************************************
1395  * ixgbe_config_link
1396  ************************************************************************/
1397 static void
1398 ixgbe_config_link(if_ctx_t ctx)
1399 {
1400 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
1401 	struct ixgbe_hw *hw = &sc->hw;
1402 	u32             autoneg, err = 0;
1403 	bool            sfp, negotiate;
1404 
1405 	sfp = ixgbe_is_sfp(hw);
1406 
1407 	if (sfp) {
1408 		sc->task_requests |= IXGBE_REQUEST_TASK_MOD;
1409 		iflib_admin_intr_deferred(ctx);
1410 	} else {
1411 		if (hw->mac.ops.check_link)
1412 			err = ixgbe_check_link(hw, &sc->link_speed,
1413 			    &sc->link_up, false);
1414 		if (err)
1415 			return;
1416 		autoneg = hw->phy.autoneg_advertised;
1417 		if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
1418 			err = hw->mac.ops.get_link_capabilities(hw, &autoneg,
1419 			    &negotiate);
1420 		if (err)
1421 			return;
1422 
1423 		if (hw->mac.type == ixgbe_mac_X550 &&
1424 		    hw->phy.autoneg_advertised == 0) {
1425 			/*
1426 			 * 2.5G and 5G autonegotiation speeds on X550
1427 			 * are disabled by default due to reported
1428 			 * interoperability issues with some switches.
1429 			 *
1430 			 * The second condition checks if any operations
1431 			 * involving setting autonegotiation speeds have
1432 			 * been performed prior to this ixgbe_config_link()
1433 			 * call.
1434 			 *
1435 			 * If hw->phy.autoneg_advertised does not
1436 			 * equal 0, this means that the user might have
1437 			 * set autonegotiation speeds via the sysctl
1438 			 * before bringing the interface up. In this
1439 			 * case, we should not disable 2.5G and 5G
1440 			 * since that speeds might be selected by the
1441 			 * user.
1442 			 *
1443 			 * Otherwise (i.e. if hw->phy.autoneg_advertised
1444 			 * is set to 0), it is the first time we set
1445 			 * autonegotiation preferences and the default
1446 			 * set of speeds should exclude 2.5G and 5G.
1447 			 */
1448 			autoneg &= ~(IXGBE_LINK_SPEED_2_5GB_FULL |
1449 			    IXGBE_LINK_SPEED_5GB_FULL);
1450 		}
1451 
1452 		if (hw->mac.ops.setup_link)
1453 			err = hw->mac.ops.setup_link(hw, autoneg,
1454 			    sc->link_up);
1455 	}
1456 } /* ixgbe_config_link */
1457 
1458 /************************************************************************
1459  * ixgbe_update_stats_counters - Update board statistics counters.
1460  ************************************************************************/
1461 static void
1462 ixgbe_update_stats_counters(struct ixgbe_softc *sc)
1463 {
1464 	struct ixgbe_hw       *hw = &sc->hw;
1465 	struct ixgbe_hw_stats *stats = &sc->stats.pf;
1466 	u32                   missed_rx = 0, bprc, lxon, lxoff, total;
1467 	u32                   lxoffrxc;
1468 	u64                   total_missed_rx = 0;
1469 
1470 	stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
1471 	stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
1472 	stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
1473 	stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
1474 	stats->mpc[0] += IXGBE_READ_REG(hw, IXGBE_MPC(0));
1475 
1476 	for (int i = 0; i < 16; i++) {
1477 		stats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
1478 		stats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
1479 		stats->qprdc[i] += IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
1480 	}
1481 	stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
1482 	stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
1483 	stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
1484 
1485 	/* Hardware workaround, gprc counts missed packets */
1486 	stats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
1487 	stats->gprc -= missed_rx;
1488 
1489 	if (hw->mac.type != ixgbe_mac_82598EB) {
1490 		stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL) +
1491 		    ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
1492 		stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL) +
1493 		    ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
1494 		stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL) +
1495 		    ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
1496 		stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
1497 		lxoffrxc = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
1498 		stats->lxoffrxc += lxoffrxc;
1499 	} else {
1500 		stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
1501 		lxoffrxc = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
1502 		stats->lxoffrxc += lxoffrxc;
1503 		/* 82598 only has a counter in the high register */
1504 		stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
1505 		stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
1506 		stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
1507 	}
1508 
1509 	/*
1510 	 * For watchdog management we need to know if we have been paused
1511 	 * during the last interval, so capture that here.
1512 	*/
1513 	if (lxoffrxc)
1514 		sc->shared->isc_pause_frames = 1;
1515 
1516 	/*
1517 	 * Workaround: mprc hardware is incorrectly counting
1518 	 * broadcasts, so for now we subtract those.
1519 	 */
1520 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
1521 	stats->bprc += bprc;
1522 	stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
1523 	if (hw->mac.type == ixgbe_mac_82598EB)
1524 		stats->mprc -= bprc;
1525 
1526 	stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
1527 	stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
1528 	stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
1529 	stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
1530 	stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
1531 	stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
1532 
1533 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
1534 	stats->lxontxc += lxon;
1535 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
1536 	stats->lxofftxc += lxoff;
1537 	total = lxon + lxoff;
1538 
1539 	stats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
1540 	stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
1541 	stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
1542 	stats->gptc -= total;
1543 	stats->mptc -= total;
1544 	stats->ptc64 -= total;
1545 	stats->gotc -= total * ETHER_MIN_LEN;
1546 
1547 	stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
1548 	stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
1549 	stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
1550 	stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
1551 	stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
1552 	stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
1553 	stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
1554 	stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
1555 	stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
1556 	stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
1557 	stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
1558 	stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
1559 	stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
1560 	stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
1561 	stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
1562 	stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
1563 	stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
1564 	stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
1565 	/* Only read FCOE on 82599 */
1566 	if (hw->mac.type != ixgbe_mac_82598EB) {
1567 		stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
1568 		stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
1569 		stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
1570 		stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
1571 		stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
1572 	}
1573 
1574 	/* Fill out the OS statistics structure */
1575 	IXGBE_SET_IPACKETS(sc, stats->gprc);
1576 	IXGBE_SET_OPACKETS(sc, stats->gptc);
1577 	IXGBE_SET_IBYTES(sc, stats->gorc);
1578 	IXGBE_SET_OBYTES(sc, stats->gotc);
1579 	IXGBE_SET_IMCASTS(sc, stats->mprc);
1580 	IXGBE_SET_OMCASTS(sc, stats->mptc);
1581 	IXGBE_SET_COLLISIONS(sc, 0);
1582 	IXGBE_SET_IQDROPS(sc, total_missed_rx);
1583 
1584 	/*
1585 	 * Aggregate following types of errors as RX errors:
1586 	 * - CRC error count,
1587 	 * - illegal byte error count,
1588 	 * - missed packets count,
1589 	 * - length error count,
1590 	 * - undersized packets count,
1591 	 * - fragmented packets count,
1592 	 * - oversized packets count,
1593 	 * - jabber count.
1594 	 */
1595 	IXGBE_SET_IERRORS(sc, stats->crcerrs + stats->illerrc +
1596 	    stats->mpc[0] + stats->rlec + stats->ruc + stats->rfc + stats->roc +
1597 	    stats->rjc);
1598 } /* ixgbe_update_stats_counters */
1599 
1600 /************************************************************************
1601  * ixgbe_add_hw_stats
1602  *
1603  *   Add sysctl variables, one per statistic, to the system.
1604  ************************************************************************/
1605 static void
1606 ixgbe_add_hw_stats(struct ixgbe_softc *sc)
1607 {
1608 	device_t               dev = iflib_get_dev(sc->ctx);
1609 	struct ix_rx_queue     *rx_que;
1610 	struct ix_tx_queue     *tx_que;
1611 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
1612 	struct sysctl_oid      *tree = device_get_sysctl_tree(dev);
1613 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
1614 	struct ixgbe_hw_stats  *stats = &sc->stats.pf;
1615 	struct sysctl_oid      *stat_node, *queue_node;
1616 	struct sysctl_oid_list *stat_list, *queue_list;
1617 	int                    i;
1618 
1619 #define QUEUE_NAME_LEN 32
1620 	char                   namebuf[QUEUE_NAME_LEN];
1621 
1622 	/* Driver Statistics */
1623 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
1624 	    CTLFLAG_RD, &sc->dropped_pkts, "Driver dropped packets");
1625 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_events",
1626 	    CTLFLAG_RD, &sc->watchdog_events, "Watchdog timeouts");
1627 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
1628 	    CTLFLAG_RD, &sc->link_irq, "Link MSI-X IRQ Handled");
1629 
1630 	for (i = 0, tx_que = sc->tx_queues; i < sc->num_tx_queues; i++, tx_que++) {
1631 		struct tx_ring *txr = &tx_que->txr;
1632 		snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i);
1633 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
1634 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Queue Name");
1635 		queue_list = SYSCTL_CHILDREN(queue_node);
1636 
1637 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
1638 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, txr, 0,
1639 		    ixgbe_sysctl_tdh_handler, "IU", "Transmit Descriptor Head");
1640 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
1641 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, txr, 0,
1642 		    ixgbe_sysctl_tdt_handler, "IU", "Transmit Descriptor Tail");
1643 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tso_tx",
1644 		    CTLFLAG_RD, &txr->tso_tx, "TSO");
1645 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tx_packets",
1646 		    CTLFLAG_RD, &txr->total_packets,
1647 		    "Queue Packets Transmitted");
1648 	}
1649 
1650 	for (i = 0, rx_que = sc->rx_queues; i < sc->num_rx_queues; i++, rx_que++) {
1651 		struct rx_ring *rxr = &rx_que->rxr;
1652 		snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i);
1653 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
1654 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Queue Name");
1655 		queue_list = SYSCTL_CHILDREN(queue_node);
1656 
1657 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "interrupt_rate",
1658 		    CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
1659 		    &sc->rx_queues[i], 0,
1660 		    ixgbe_sysctl_interrupt_rate_handler, "IU",
1661 		    "Interrupt Rate");
1662 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "irqs",
1663 		    CTLFLAG_RD, &(sc->rx_queues[i].irqs),
1664 		    "irqs on this queue");
1665 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
1666 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, rxr, 0,
1667 		    ixgbe_sysctl_rdh_handler, "IU", "Receive Descriptor Head");
1668 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
1669 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, rxr, 0,
1670 		    ixgbe_sysctl_rdt_handler, "IU", "Receive Descriptor Tail");
1671 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_packets",
1672 		    CTLFLAG_RD, &rxr->rx_packets, "Queue Packets Received");
1673 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_bytes",
1674 		    CTLFLAG_RD, &rxr->rx_bytes, "Queue Bytes Received");
1675 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_copies",
1676 		    CTLFLAG_RD, &rxr->rx_copies, "Copied RX Frames");
1677 		SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_discarded",
1678 		    CTLFLAG_RD, &rxr->rx_discarded, "Discarded RX packets");
1679 	}
1680 
1681 	/* MAC stats get their own sub node */
1682 
1683 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
1684 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "MAC Statistics");
1685 	stat_list = SYSCTL_CHILDREN(stat_node);
1686 
1687 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_errs",
1688 	    CTLFLAG_RD, &sc->ierrors, IXGBE_SYSCTL_DESC_RX_ERRS);
1689 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
1690 	    CTLFLAG_RD, &stats->crcerrs, "CRC Errors");
1691 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "ill_errs",
1692 	    CTLFLAG_RD, &stats->illerrc, "Illegal Byte Errors");
1693 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "byte_errs",
1694 	    CTLFLAG_RD, &stats->errbc, "Byte Errors");
1695 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "short_discards",
1696 	    CTLFLAG_RD, &stats->mspdc, "MAC Short Packets Discarded");
1697 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "local_faults",
1698 	    CTLFLAG_RD, &stats->mlfc, "MAC Local Faults");
1699 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "remote_faults",
1700 	    CTLFLAG_RD, &stats->mrfc, "MAC Remote Faults");
1701 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rec_len_errs",
1702 	    CTLFLAG_RD, &stats->rlec, "Receive Length Errors");
1703 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_missed_packets",
1704 	    CTLFLAG_RD, &stats->mpc[0], "RX Missed Packet Count");
1705 
1706 	/* Flow Control stats */
1707 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
1708 	    CTLFLAG_RD, &stats->lxontxc, "Link XON Transmitted");
1709 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
1710 	    CTLFLAG_RD, &stats->lxonrxc, "Link XON Received");
1711 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
1712 	    CTLFLAG_RD, &stats->lxofftxc, "Link XOFF Transmitted");
1713 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
1714 	    CTLFLAG_RD, &stats->lxoffrxc, "Link XOFF Received");
1715 
1716 	/* Packet Reception Stats */
1717 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_octets_rcvd",
1718 	    CTLFLAG_RD, &stats->tor, "Total Octets Received");
1719 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_rcvd",
1720 	    CTLFLAG_RD, &stats->gorc, "Good Octets Received");
1721 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_rcvd",
1722 	    CTLFLAG_RD, &stats->tpr, "Total Packets Received");
1723 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_rcvd",
1724 	    CTLFLAG_RD, &stats->gprc, "Good Packets Received");
1725 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_rcvd",
1726 	    CTLFLAG_RD, &stats->mprc, "Multicast Packets Received");
1727 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_rcvd",
1728 	    CTLFLAG_RD, &stats->bprc, "Broadcast Packets Received");
1729 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
1730 	    CTLFLAG_RD, &stats->prc64, "64 byte frames received ");
1731 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
1732 	    CTLFLAG_RD, &stats->prc127, "65-127 byte frames received");
1733 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
1734 	    CTLFLAG_RD, &stats->prc255, "128-255 byte frames received");
1735 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
1736 	    CTLFLAG_RD, &stats->prc511, "256-511 byte frames received");
1737 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
1738 	    CTLFLAG_RD, &stats->prc1023, "512-1023 byte frames received");
1739 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
1740 	    CTLFLAG_RD, &stats->prc1522, "1023-1522 byte frames received");
1741 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersized",
1742 	    CTLFLAG_RD, &stats->ruc, "Receive Undersized");
1743 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
1744 	    CTLFLAG_RD, &stats->rfc, "Fragmented Packets Received ");
1745 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversized",
1746 	    CTLFLAG_RD, &stats->roc, "Oversized Packets Received");
1747 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabberd",
1748 	    CTLFLAG_RD, &stats->rjc, "Received Jabber");
1749 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_rcvd",
1750 	    CTLFLAG_RD, &stats->mngprc, "Management Packets Received");
1751 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_drpd",
1752 	    CTLFLAG_RD, &stats->mngptc, "Management Packets Dropped");
1753 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "checksum_errs",
1754 	    CTLFLAG_RD, &stats->xec, "Checksum Errors");
1755 
1756 	/* Packet Transmission Stats */
1757 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
1758 	    CTLFLAG_RD, &stats->gotc, "Good Octets Transmitted");
1759 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
1760 	    CTLFLAG_RD, &stats->tpt, "Total Packets Transmitted");
1761 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
1762 	    CTLFLAG_RD, &stats->gptc, "Good Packets Transmitted");
1763 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
1764 	    CTLFLAG_RD, &stats->bptc, "Broadcast Packets Transmitted");
1765 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
1766 	    CTLFLAG_RD, &stats->mptc, "Multicast Packets Transmitted");
1767 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_txd",
1768 	    CTLFLAG_RD, &stats->mngptc, "Management Packets Transmitted");
1769 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
1770 	    CTLFLAG_RD, &stats->ptc64, "64 byte frames transmitted ");
1771 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
1772 	    CTLFLAG_RD, &stats->ptc127, "65-127 byte frames transmitted");
1773 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
1774 	    CTLFLAG_RD, &stats->ptc255, "128-255 byte frames transmitted");
1775 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
1776 	    CTLFLAG_RD, &stats->ptc511, "256-511 byte frames transmitted");
1777 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
1778 	    CTLFLAG_RD, &stats->ptc1023, "512-1023 byte frames transmitted");
1779 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
1780 	    CTLFLAG_RD, &stats->ptc1522, "1024-1522 byte frames transmitted");
1781 } /* ixgbe_add_hw_stats */
1782 
1783 /************************************************************************
1784  * ixgbe_sysctl_tdh_handler - Transmit Descriptor Head handler function
1785  *
1786  *   Retrieves the TDH value from the hardware
1787  ************************************************************************/
1788 static int
1789 ixgbe_sysctl_tdh_handler(SYSCTL_HANDLER_ARGS)
1790 {
1791 	struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1);
1792 	int            error;
1793 	unsigned int   val;
1794 
1795 	if (!txr)
1796 		return (0);
1797 
1798 	val = IXGBE_READ_REG(&txr->sc->hw, IXGBE_TDH(txr->me));
1799 	error = sysctl_handle_int(oidp, &val, 0, req);
1800 	if (error || !req->newptr)
1801 		return error;
1802 
1803 	return (0);
1804 } /* ixgbe_sysctl_tdh_handler */
1805 
1806 /************************************************************************
1807  * ixgbe_sysctl_tdt_handler - Transmit Descriptor Tail handler function
1808  *
1809  *   Retrieves the TDT value from the hardware
1810  ************************************************************************/
1811 static int
1812 ixgbe_sysctl_tdt_handler(SYSCTL_HANDLER_ARGS)
1813 {
1814 	struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1);
1815 	int            error;
1816 	unsigned int   val;
1817 
1818 	if (!txr)
1819 		return (0);
1820 
1821 	val = IXGBE_READ_REG(&txr->sc->hw, IXGBE_TDT(txr->me));
1822 	error = sysctl_handle_int(oidp, &val, 0, req);
1823 	if (error || !req->newptr)
1824 		return error;
1825 
1826 	return (0);
1827 } /* ixgbe_sysctl_tdt_handler */
1828 
1829 /************************************************************************
1830  * ixgbe_sysctl_rdh_handler - Receive Descriptor Head handler function
1831  *
1832  *   Retrieves the RDH value from the hardware
1833  ************************************************************************/
1834 static int
1835 ixgbe_sysctl_rdh_handler(SYSCTL_HANDLER_ARGS)
1836 {
1837 	struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1);
1838 	int            error;
1839 	unsigned int   val;
1840 
1841 	if (!rxr)
1842 		return (0);
1843 
1844 	val = IXGBE_READ_REG(&rxr->sc->hw, IXGBE_RDH(rxr->me));
1845 	error = sysctl_handle_int(oidp, &val, 0, req);
1846 	if (error || !req->newptr)
1847 		return error;
1848 
1849 	return (0);
1850 } /* ixgbe_sysctl_rdh_handler */
1851 
1852 /************************************************************************
1853  * ixgbe_sysctl_rdt_handler - Receive Descriptor Tail handler function
1854  *
1855  *   Retrieves the RDT value from the hardware
1856  ************************************************************************/
1857 static int
1858 ixgbe_sysctl_rdt_handler(SYSCTL_HANDLER_ARGS)
1859 {
1860 	struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1);
1861 	int            error;
1862 	unsigned int   val;
1863 
1864 	if (!rxr)
1865 		return (0);
1866 
1867 	val = IXGBE_READ_REG(&rxr->sc->hw, IXGBE_RDT(rxr->me));
1868 	error = sysctl_handle_int(oidp, &val, 0, req);
1869 	if (error || !req->newptr)
1870 		return error;
1871 
1872 	return (0);
1873 } /* ixgbe_sysctl_rdt_handler */
1874 
1875 /************************************************************************
1876  * ixgbe_if_vlan_register
1877  *
1878  *   Run via vlan config EVENT, it enables us to use the
1879  *   HW Filter table since we can get the vlan id. This
1880  *   just creates the entry in the soft version of the
1881  *   VFTA, init will repopulate the real table.
1882  ************************************************************************/
1883 static void
1884 ixgbe_if_vlan_register(if_ctx_t ctx, u16 vtag)
1885 {
1886 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1887 	u16            index, bit;
1888 
1889 	index = (vtag >> 5) & 0x7F;
1890 	bit = vtag & 0x1F;
1891 	sc->shadow_vfta[index] |= (1 << bit);
1892 	++sc->num_vlans;
1893 	ixgbe_setup_vlan_hw_support(ctx);
1894 } /* ixgbe_if_vlan_register */
1895 
1896 /************************************************************************
1897  * ixgbe_if_vlan_unregister
1898  *
1899  *   Run via vlan unconfig EVENT, remove our entry in the soft vfta.
1900  ************************************************************************/
1901 static void
1902 ixgbe_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
1903 {
1904 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
1905 	u16            index, bit;
1906 
1907 	index = (vtag >> 5) & 0x7F;
1908 	bit = vtag & 0x1F;
1909 	sc->shadow_vfta[index] &= ~(1 << bit);
1910 	--sc->num_vlans;
1911 	/* Re-init to load the changes */
1912 	ixgbe_setup_vlan_hw_support(ctx);
1913 } /* ixgbe_if_vlan_unregister */
1914 
1915 /************************************************************************
1916  * ixgbe_setup_vlan_hw_support
1917  ************************************************************************/
1918 static void
1919 ixgbe_setup_vlan_hw_support(if_ctx_t ctx)
1920 {
1921 	if_t            ifp = iflib_get_ifp(ctx);
1922 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
1923 	struct ixgbe_hw *hw = &sc->hw;
1924 	struct rx_ring  *rxr;
1925 	int             i;
1926 	u32             ctrl;
1927 
1928 
1929 	/*
1930 	 * We get here thru init_locked, meaning
1931 	 * a soft reset, this has already cleared
1932 	 * the VFTA and other state, so if there
1933 	 * have been no vlan's registered do nothing.
1934 	 */
1935 	if (sc->num_vlans == 0 || (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) == 0) {
1936 		/* Clear the vlan hw flag */
1937 		for (i = 0; i < sc->num_rx_queues; i++) {
1938 			rxr = &sc->rx_queues[i].rxr;
1939 			/* On 82599 the VLAN enable is per/queue in RXDCTL */
1940 			if (hw->mac.type != ixgbe_mac_82598EB) {
1941 				ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me));
1942 				ctrl &= ~IXGBE_RXDCTL_VME;
1943 				IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), ctrl);
1944 			}
1945 			rxr->vtag_strip = false;
1946 		}
1947 		ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1948 		/* Enable the Filter Table if enabled */
1949 		ctrl |= IXGBE_VLNCTRL_CFIEN;
1950 		ctrl &= ~IXGBE_VLNCTRL_VFE;
1951 		if (hw->mac.type == ixgbe_mac_82598EB)
1952 			ctrl &= ~IXGBE_VLNCTRL_VME;
1953 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1954 		return;
1955 	}
1956 
1957 	/* Setup the queues for vlans */
1958 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1959 		for (i = 0; i < sc->num_rx_queues; i++) {
1960 			rxr = &sc->rx_queues[i].rxr;
1961 			/* On 82599 the VLAN enable is per/queue in RXDCTL */
1962 			if (hw->mac.type != ixgbe_mac_82598EB) {
1963 				ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me));
1964 				ctrl |= IXGBE_RXDCTL_VME;
1965 				IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), ctrl);
1966 			}
1967 			rxr->vtag_strip = true;
1968 		}
1969 	}
1970 
1971 	if ((if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) == 0)
1972 		return;
1973 	/*
1974 	 * A soft reset zero's out the VFTA, so
1975 	 * we need to repopulate it now.
1976 	 */
1977 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1978 		if (sc->shadow_vfta[i] != 0)
1979 			IXGBE_WRITE_REG(hw, IXGBE_VFTA(i),
1980 			    sc->shadow_vfta[i]);
1981 
1982 	ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1983 	/* Enable the Filter Table if enabled */
1984 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) {
1985 		ctrl &= ~IXGBE_VLNCTRL_CFIEN;
1986 		ctrl |= IXGBE_VLNCTRL_VFE;
1987 	}
1988 	if (hw->mac.type == ixgbe_mac_82598EB)
1989 		ctrl |= IXGBE_VLNCTRL_VME;
1990 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1991 } /* ixgbe_setup_vlan_hw_support */
1992 
1993 /************************************************************************
1994  * ixgbe_get_slot_info
1995  *
1996  *   Get the width and transaction speed of
1997  *   the slot this adapter is plugged into.
1998  ************************************************************************/
1999 static void
2000 ixgbe_get_slot_info(struct ixgbe_softc *sc)
2001 {
2002 	device_t        dev = iflib_get_dev(sc->ctx);
2003 	struct ixgbe_hw *hw = &sc->hw;
2004 	int             bus_info_valid = true;
2005 	u32             offset;
2006 	u16             link;
2007 
2008 	/* Some devices are behind an internal bridge */
2009 	switch (hw->device_id) {
2010 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
2011 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
2012 		goto get_parent_info;
2013 	default:
2014 		break;
2015 	}
2016 
2017 	ixgbe_get_bus_info(hw);
2018 
2019 	/*
2020 	 * Some devices don't use PCI-E, but there is no need
2021 	 * to display "Unknown" for bus speed and width.
2022 	 */
2023 	switch (hw->mac.type) {
2024 	case ixgbe_mac_X550EM_x:
2025 	case ixgbe_mac_X550EM_a:
2026 		return;
2027 	default:
2028 		goto display;
2029 	}
2030 
2031 get_parent_info:
2032 	/*
2033 	 * For the Quad port adapter we need to parse back
2034 	 * up the PCI tree to find the speed of the expansion
2035 	 * slot into which this adapter is plugged. A bit more work.
2036 	 */
2037 	dev = device_get_parent(device_get_parent(dev));
2038 #ifdef IXGBE_DEBUG
2039 	device_printf(dev, "parent pcib = %x,%x,%x\n", pci_get_bus(dev),
2040 	    pci_get_slot(dev), pci_get_function(dev));
2041 #endif
2042 	dev = device_get_parent(device_get_parent(dev));
2043 #ifdef IXGBE_DEBUG
2044 	device_printf(dev, "slot pcib = %x,%x,%x\n", pci_get_bus(dev),
2045 	    pci_get_slot(dev), pci_get_function(dev));
2046 #endif
2047 	/* Now get the PCI Express Capabilities offset */
2048 	if (pci_find_cap(dev, PCIY_EXPRESS, &offset)) {
2049 		/*
2050 		 * Hmm...can't get PCI-Express capabilities.
2051 		 * Falling back to default method.
2052 		 */
2053 		bus_info_valid = false;
2054 		ixgbe_get_bus_info(hw);
2055 		goto display;
2056 	}
2057 	/* ...and read the Link Status Register */
2058 	link = pci_read_config(dev, offset + PCIER_LINK_STA, 2);
2059 	ixgbe_set_pci_config_data_generic(hw, link);
2060 
2061 display:
2062 	device_printf(dev, "PCI Express Bus: Speed %s %s\n",
2063 	    ((hw->bus.speed == ixgbe_bus_speed_8000)    ? "8.0GT/s"  :
2064 	     (hw->bus.speed == ixgbe_bus_speed_5000)    ? "5.0GT/s"  :
2065 	     (hw->bus.speed == ixgbe_bus_speed_2500)    ? "2.5GT/s"  :
2066 	     "Unknown"),
2067 	    ((hw->bus.width == ixgbe_bus_width_pcie_x8) ? "Width x8" :
2068 	     (hw->bus.width == ixgbe_bus_width_pcie_x4) ? "Width x4" :
2069 	     (hw->bus.width == ixgbe_bus_width_pcie_x1) ? "Width x1" :
2070 	     "Unknown"));
2071 
2072 	if (bus_info_valid) {
2073 		if ((hw->device_id != IXGBE_DEV_ID_82599_SFP_SF_QP) &&
2074 		    ((hw->bus.width <= ixgbe_bus_width_pcie_x4) &&
2075 		    (hw->bus.speed == ixgbe_bus_speed_2500))) {
2076 			device_printf(dev, "PCI-Express bandwidth available for this card\n     is not sufficient for optimal performance.\n");
2077 			device_printf(dev, "For optimal performance a x8 PCIE, or x4 PCIE Gen2 slot is required.\n");
2078 		}
2079 		if ((hw->device_id == IXGBE_DEV_ID_82599_SFP_SF_QP) &&
2080 		    ((hw->bus.width <= ixgbe_bus_width_pcie_x8) &&
2081 		    (hw->bus.speed < ixgbe_bus_speed_8000))) {
2082 			device_printf(dev, "PCI-Express bandwidth available for this card\n     is not sufficient for optimal performance.\n");
2083 			device_printf(dev, "For optimal performance a x8 PCIE Gen3 slot is required.\n");
2084 		}
2085 	} else
2086 		device_printf(dev, "Unable to determine slot speed/width. The speed/width reported are that of the internal switch.\n");
2087 
2088 	return;
2089 } /* ixgbe_get_slot_info */
2090 
2091 /************************************************************************
2092  * ixgbe_if_msix_intr_assign
2093  *
2094  *   Setup MSI-X Interrupt resources and handlers
2095  ************************************************************************/
2096 static int
2097 ixgbe_if_msix_intr_assign(if_ctx_t ctx, int msix)
2098 {
2099 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
2100 	struct ix_rx_queue *rx_que = sc->rx_queues;
2101 	struct ix_tx_queue *tx_que;
2102 	int                error, rid, vector = 0;
2103 	char               buf[16];
2104 
2105 	/* Admin Que is vector 0*/
2106 	rid = vector + 1;
2107 	for (int i = 0; i < sc->num_rx_queues; i++, vector++, rx_que++) {
2108 		rid = vector + 1;
2109 
2110 		snprintf(buf, sizeof(buf), "rxq%d", i);
2111 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid,
2112 		    IFLIB_INTR_RXTX, ixgbe_msix_que, rx_que, rx_que->rxr.me, buf);
2113 
2114 		if (error) {
2115 			device_printf(iflib_get_dev(ctx),
2116 			    "Failed to allocate que int %d err: %d", i, error);
2117 			sc->num_rx_queues = i + 1;
2118 			goto fail;
2119 		}
2120 
2121 		rx_que->msix = vector;
2122 	}
2123 	for (int i = 0; i < sc->num_tx_queues; i++) {
2124 		snprintf(buf, sizeof(buf), "txq%d", i);
2125 		tx_que = &sc->tx_queues[i];
2126 		tx_que->msix = i % sc->num_rx_queues;
2127 		iflib_softirq_alloc_generic(ctx,
2128 		    &sc->rx_queues[tx_que->msix].que_irq,
2129 		    IFLIB_INTR_TX, tx_que, tx_que->txr.me, buf);
2130 	}
2131 	rid = vector + 1;
2132 	error = iflib_irq_alloc_generic(ctx, &sc->irq, rid,
2133 	    IFLIB_INTR_ADMIN, ixgbe_msix_link, sc, 0, "aq");
2134 	if (error) {
2135 		device_printf(iflib_get_dev(ctx),
2136 		    "Failed to register admin handler");
2137 		return (error);
2138 	}
2139 
2140 	sc->vector = vector;
2141 
2142 	return (0);
2143 fail:
2144 	iflib_irq_free(ctx, &sc->irq);
2145 	rx_que = sc->rx_queues;
2146 	for (int i = 0; i < sc->num_rx_queues; i++, rx_que++)
2147 		iflib_irq_free(ctx, &rx_que->que_irq);
2148 
2149 	return (error);
2150 } /* ixgbe_if_msix_intr_assign */
2151 
2152 static inline void
2153 ixgbe_perform_aim(struct ixgbe_softc *sc, struct ix_rx_queue *que)
2154 {
2155 	uint32_t newitr = 0;
2156 	struct rx_ring *rxr = &que->rxr;
2157 
2158 	/*
2159 	 * Do Adaptive Interrupt Moderation:
2160 	 *  - Write out last calculated setting
2161 	 *  - Calculate based on average size over
2162 	 *    the last interval.
2163 	 */
2164 	if (que->eitr_setting) {
2165 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EITR(que->msix),
2166 		    que->eitr_setting);
2167 	}
2168 
2169 	que->eitr_setting = 0;
2170 	/* Idle, do nothing */
2171 	if (rxr->bytes == 0) {
2172 		return;
2173 	}
2174 
2175 	if ((rxr->bytes) && (rxr->packets)) {
2176 		newitr = (rxr->bytes / rxr->packets);
2177 	}
2178 
2179 	newitr += 24; /* account for hardware frame, crc */
2180 	/* set an upper boundary */
2181 	newitr = min(newitr, 3000);
2182 
2183 	/* Be nice to the mid range */
2184 	if ((newitr > 300) && (newitr < 1200)) {
2185 		newitr = (newitr / 3);
2186 	} else {
2187 		newitr = (newitr / 2);
2188 	}
2189 
2190 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
2191 		newitr |= newitr << 16;
2192 	} else {
2193 		newitr |= IXGBE_EITR_CNT_WDIS;
2194 	}
2195 
2196 	/* save for next interrupt */
2197 	que->eitr_setting = newitr;
2198 
2199 	/* Reset state */
2200 	rxr->bytes = 0;
2201 	rxr->packets = 0;
2202 
2203 	return;
2204 }
2205 
2206 /*********************************************************************
2207  * ixgbe_msix_que - MSI-X Queue Interrupt Service routine
2208  **********************************************************************/
2209 static int
2210 ixgbe_msix_que(void *arg)
2211 {
2212 	struct ix_rx_queue *que = arg;
2213 	struct ixgbe_softc     *sc = que->sc;
2214 	if_t               ifp = iflib_get_ifp(que->sc->ctx);
2215 
2216 	/* Protect against spurious interrupts */
2217 	if ((if_getdrvflags(ifp) & IFF_DRV_RUNNING) == 0)
2218 		return (FILTER_HANDLED);
2219 
2220 	ixgbe_disable_queue(sc, que->msix);
2221 	++que->irqs;
2222 
2223 	/* Check for AIM */
2224 	if (sc->enable_aim) {
2225 		ixgbe_perform_aim(sc, que);
2226 	}
2227 
2228 	return (FILTER_SCHEDULE_THREAD);
2229 } /* ixgbe_msix_que */
2230 
2231 /************************************************************************
2232  * ixgbe_media_status - Media Ioctl callback
2233  *
2234  *   Called whenever the user queries the status of
2235  *   the interface using ifconfig.
2236  ************************************************************************/
2237 static void
2238 ixgbe_if_media_status(if_ctx_t ctx, struct ifmediareq * ifmr)
2239 {
2240 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
2241 	struct ixgbe_hw *hw = &sc->hw;
2242 	int             layer;
2243 
2244 	INIT_DEBUGOUT("ixgbe_if_media_status: begin");
2245 
2246 	ifmr->ifm_status = IFM_AVALID;
2247 	ifmr->ifm_active = IFM_ETHER;
2248 
2249 	if (!sc->link_active)
2250 		return;
2251 
2252 	ifmr->ifm_status |= IFM_ACTIVE;
2253 	layer = sc->phy_layer;
2254 
2255 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T ||
2256 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_T ||
2257 	    layer & IXGBE_PHYSICAL_LAYER_100BASE_TX ||
2258 	    layer & IXGBE_PHYSICAL_LAYER_10BASE_T)
2259 		switch (sc->link_speed) {
2260 		case IXGBE_LINK_SPEED_10GB_FULL:
2261 			ifmr->ifm_active |= IFM_10G_T | IFM_FDX;
2262 			break;
2263 		case IXGBE_LINK_SPEED_1GB_FULL:
2264 			ifmr->ifm_active |= IFM_1000_T | IFM_FDX;
2265 			break;
2266 		case IXGBE_LINK_SPEED_100_FULL:
2267 			ifmr->ifm_active |= IFM_100_TX | IFM_FDX;
2268 			break;
2269 		case IXGBE_LINK_SPEED_10_FULL:
2270 			ifmr->ifm_active |= IFM_10_T | IFM_FDX;
2271 			break;
2272 		}
2273 	if (hw->mac.type == ixgbe_mac_X550)
2274 		switch (sc->link_speed) {
2275 		case IXGBE_LINK_SPEED_5GB_FULL:
2276 			ifmr->ifm_active |= IFM_5000_T | IFM_FDX;
2277 			break;
2278 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2279 			ifmr->ifm_active |= IFM_2500_T | IFM_FDX;
2280 			break;
2281 		}
2282 	if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU ||
2283 	    layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA)
2284 		switch (sc->link_speed) {
2285 		case IXGBE_LINK_SPEED_10GB_FULL:
2286 			ifmr->ifm_active |= IFM_10G_TWINAX | IFM_FDX;
2287 			break;
2288 		}
2289 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR)
2290 		switch (sc->link_speed) {
2291 		case IXGBE_LINK_SPEED_10GB_FULL:
2292 			ifmr->ifm_active |= IFM_10G_LR | IFM_FDX;
2293 			break;
2294 		case IXGBE_LINK_SPEED_1GB_FULL:
2295 			ifmr->ifm_active |= IFM_1000_LX | IFM_FDX;
2296 			break;
2297 		}
2298 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LRM)
2299 		switch (sc->link_speed) {
2300 		case IXGBE_LINK_SPEED_10GB_FULL:
2301 			ifmr->ifm_active |= IFM_10G_LRM | IFM_FDX;
2302 			break;
2303 		case IXGBE_LINK_SPEED_1GB_FULL:
2304 			ifmr->ifm_active |= IFM_1000_LX | IFM_FDX;
2305 			break;
2306 		}
2307 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR ||
2308 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX)
2309 		switch (sc->link_speed) {
2310 		case IXGBE_LINK_SPEED_10GB_FULL:
2311 			ifmr->ifm_active |= IFM_10G_SR | IFM_FDX;
2312 			break;
2313 		case IXGBE_LINK_SPEED_1GB_FULL:
2314 			ifmr->ifm_active |= IFM_1000_SX | IFM_FDX;
2315 			break;
2316 		}
2317 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4)
2318 		switch (sc->link_speed) {
2319 		case IXGBE_LINK_SPEED_10GB_FULL:
2320 			ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX;
2321 			break;
2322 		}
2323 	/*
2324 	 * XXX: These need to use the proper media types once
2325 	 * they're added.
2326 	 */
2327 #ifndef IFM_ETH_XTYPE
2328 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
2329 		switch (sc->link_speed) {
2330 		case IXGBE_LINK_SPEED_10GB_FULL:
2331 			ifmr->ifm_active |= IFM_10G_SR | IFM_FDX;
2332 			break;
2333 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2334 			ifmr->ifm_active |= IFM_2500_SX | IFM_FDX;
2335 			break;
2336 		case IXGBE_LINK_SPEED_1GB_FULL:
2337 			ifmr->ifm_active |= IFM_1000_CX | IFM_FDX;
2338 			break;
2339 		}
2340 	else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4 ||
2341 	    layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX ||
2342 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
2343 		switch (sc->link_speed) {
2344 		case IXGBE_LINK_SPEED_10GB_FULL:
2345 			ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX;
2346 			break;
2347 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2348 			ifmr->ifm_active |= IFM_2500_SX | IFM_FDX;
2349 			break;
2350 		case IXGBE_LINK_SPEED_1GB_FULL:
2351 			ifmr->ifm_active |= IFM_1000_CX | IFM_FDX;
2352 			break;
2353 		}
2354 #else
2355 	if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
2356 		switch (sc->link_speed) {
2357 		case IXGBE_LINK_SPEED_10GB_FULL:
2358 			ifmr->ifm_active |= IFM_10G_KR | IFM_FDX;
2359 			break;
2360 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2361 			ifmr->ifm_active |= IFM_2500_KX | IFM_FDX;
2362 			break;
2363 		case IXGBE_LINK_SPEED_1GB_FULL:
2364 			ifmr->ifm_active |= IFM_1000_KX | IFM_FDX;
2365 			break;
2366 		}
2367 	else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4 ||
2368 	    layer & IXGBE_PHYSICAL_LAYER_2500BASE_KX ||
2369 	    layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
2370 		switch (sc->link_speed) {
2371 		case IXGBE_LINK_SPEED_10GB_FULL:
2372 			ifmr->ifm_active |= IFM_10G_KX4 | IFM_FDX;
2373 			break;
2374 		case IXGBE_LINK_SPEED_2_5GB_FULL:
2375 			ifmr->ifm_active |= IFM_2500_KX | IFM_FDX;
2376 			break;
2377 		case IXGBE_LINK_SPEED_1GB_FULL:
2378 			ifmr->ifm_active |= IFM_1000_KX | IFM_FDX;
2379 			break;
2380 		}
2381 #endif
2382 
2383 	/* If nothing is recognized... */
2384 	if (IFM_SUBTYPE(ifmr->ifm_active) == 0)
2385 		ifmr->ifm_active |= IFM_UNKNOWN;
2386 
2387 	/* Display current flow control setting used on link */
2388 	if (hw->fc.current_mode == ixgbe_fc_rx_pause ||
2389 	    hw->fc.current_mode == ixgbe_fc_full)
2390 		ifmr->ifm_active |= IFM_ETH_RXPAUSE;
2391 	if (hw->fc.current_mode == ixgbe_fc_tx_pause ||
2392 	    hw->fc.current_mode == ixgbe_fc_full)
2393 		ifmr->ifm_active |= IFM_ETH_TXPAUSE;
2394 } /* ixgbe_media_status */
2395 
2396 /************************************************************************
2397  * ixgbe_media_change - Media Ioctl callback
2398  *
2399  *   Called when the user changes speed/duplex using
2400  *   media/mediopt option with ifconfig.
2401  ************************************************************************/
2402 static int
2403 ixgbe_if_media_change(if_ctx_t ctx)
2404 {
2405 	struct ixgbe_softc   *sc = iflib_get_softc(ctx);
2406 	struct ifmedia   *ifm = iflib_get_media(ctx);
2407 	struct ixgbe_hw  *hw = &sc->hw;
2408 	ixgbe_link_speed speed = 0;
2409 
2410 	INIT_DEBUGOUT("ixgbe_if_media_change: begin");
2411 
2412 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
2413 		return (EINVAL);
2414 
2415 	if (hw->phy.media_type == ixgbe_media_type_backplane)
2416 		return (EPERM);
2417 
2418 	/*
2419 	 * We don't actually need to check against the supported
2420 	 * media types of the adapter; ifmedia will take care of
2421 	 * that for us.
2422 	 */
2423 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
2424 	case IFM_AUTO:
2425 	case IFM_10G_T:
2426 		speed |= IXGBE_LINK_SPEED_100_FULL;
2427 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2428 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
2429 		break;
2430 	case IFM_10G_LRM:
2431 	case IFM_10G_LR:
2432 #ifndef IFM_ETH_XTYPE
2433 	case IFM_10G_SR: /* KR, too */
2434 	case IFM_10G_CX4: /* KX4 */
2435 #else
2436 	case IFM_10G_KR:
2437 	case IFM_10G_KX4:
2438 #endif
2439 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2440 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
2441 		break;
2442 #ifndef IFM_ETH_XTYPE
2443 	case IFM_1000_CX: /* KX */
2444 #else
2445 	case IFM_1000_KX:
2446 #endif
2447 	case IFM_1000_LX:
2448 	case IFM_1000_SX:
2449 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2450 		break;
2451 	case IFM_1000_T:
2452 		speed |= IXGBE_LINK_SPEED_100_FULL;
2453 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
2454 		break;
2455 	case IFM_10G_TWINAX:
2456 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
2457 		break;
2458 	case IFM_5000_T:
2459 		speed |= IXGBE_LINK_SPEED_5GB_FULL;
2460 		break;
2461 	case IFM_2500_T:
2462 		speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2463 		break;
2464 	case IFM_100_TX:
2465 		speed |= IXGBE_LINK_SPEED_100_FULL;
2466 		break;
2467 	case IFM_10_T:
2468 		speed |= IXGBE_LINK_SPEED_10_FULL;
2469 		break;
2470 	default:
2471 		goto invalid;
2472 	}
2473 
2474 	hw->mac.autotry_restart = true;
2475 	hw->mac.ops.setup_link(hw, speed, true);
2476 	sc->advertise =
2477 	    ((speed & IXGBE_LINK_SPEED_10GB_FULL)  ? 0x4  : 0) |
2478 	    ((speed & IXGBE_LINK_SPEED_5GB_FULL)   ? 0x20 : 0) |
2479 	    ((speed & IXGBE_LINK_SPEED_2_5GB_FULL) ? 0x10 : 0) |
2480 	    ((speed & IXGBE_LINK_SPEED_1GB_FULL)   ? 0x2  : 0) |
2481 	    ((speed & IXGBE_LINK_SPEED_100_FULL)   ? 0x1  : 0) |
2482 	    ((speed & IXGBE_LINK_SPEED_10_FULL)    ? 0x8  : 0);
2483 
2484 	return (0);
2485 
2486 invalid:
2487 	device_printf(iflib_get_dev(ctx), "Invalid media type!\n");
2488 
2489 	return (EINVAL);
2490 } /* ixgbe_if_media_change */
2491 
2492 /************************************************************************
2493  * ixgbe_set_promisc
2494  ************************************************************************/
2495 static int
2496 ixgbe_if_promisc_set(if_ctx_t ctx, int flags)
2497 {
2498 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2499 	if_t           ifp = iflib_get_ifp(ctx);
2500 	u32            rctl;
2501 	int            mcnt = 0;
2502 
2503 	rctl = IXGBE_READ_REG(&sc->hw, IXGBE_FCTRL);
2504 	rctl &= (~IXGBE_FCTRL_UPE);
2505 	if (if_getflags(ifp) & IFF_ALLMULTI)
2506 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
2507 	else {
2508 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
2509 	}
2510 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
2511 		rctl &= (~IXGBE_FCTRL_MPE);
2512 	IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, rctl);
2513 
2514 	if (if_getflags(ifp) & IFF_PROMISC) {
2515 		rctl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
2516 		IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, rctl);
2517 	} else if (if_getflags(ifp) & IFF_ALLMULTI) {
2518 		rctl |= IXGBE_FCTRL_MPE;
2519 		rctl &= ~IXGBE_FCTRL_UPE;
2520 		IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, rctl);
2521 	}
2522 	return (0);
2523 } /* ixgbe_if_promisc_set */
2524 
2525 /************************************************************************
2526  * ixgbe_msix_link - Link status change ISR (MSI/MSI-X)
2527  ************************************************************************/
2528 static int
2529 ixgbe_msix_link(void *arg)
2530 {
2531 	struct ixgbe_softc  *sc = arg;
2532 	struct ixgbe_hw *hw = &sc->hw;
2533 	u32             eicr, eicr_mask;
2534 	s32             retval;
2535 
2536 	++sc->link_irq;
2537 
2538 	/* Pause other interrupts */
2539 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_OTHER);
2540 
2541 	/* First get the cause */
2542 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2543 	/* Be sure the queue bits are not cleared */
2544 	eicr &= ~IXGBE_EICR_RTX_QUEUE;
2545 	/* Clear interrupt with write */
2546 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2547 
2548 	/* Link status change */
2549 	if (eicr & IXGBE_EICR_LSC) {
2550 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2551 		sc->task_requests |= IXGBE_REQUEST_TASK_LSC;
2552 	}
2553 
2554 	if (sc->hw.mac.type != ixgbe_mac_82598EB) {
2555 		if ((sc->feat_en & IXGBE_FEATURE_FDIR) &&
2556 		    (eicr & IXGBE_EICR_FLOW_DIR)) {
2557 			/* This is probably overkill :) */
2558 			if (!atomic_cmpset_int(&sc->fdir_reinit, 0, 1))
2559 				return (FILTER_HANDLED);
2560 			/* Disable the interrupt */
2561 			IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EICR_FLOW_DIR);
2562 			sc->task_requests |= IXGBE_REQUEST_TASK_FDIR;
2563 		} else
2564 			if (eicr & IXGBE_EICR_ECC) {
2565 				device_printf(iflib_get_dev(sc->ctx),
2566 				   "Received ECC Err, initiating reset\n");
2567 				hw->mac.flags |= ~IXGBE_FLAGS_DOUBLE_RESET_REQUIRED;
2568 				ixgbe_reset_hw(hw);
2569 				IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2570 			}
2571 
2572 		/* Check for over temp condition */
2573 		if (sc->feat_en & IXGBE_FEATURE_TEMP_SENSOR) {
2574 			switch (sc->hw.mac.type) {
2575 			case ixgbe_mac_X550EM_a:
2576 				if (!(eicr & IXGBE_EICR_GPI_SDP0_X550EM_a))
2577 					break;
2578 				IXGBE_WRITE_REG(hw, IXGBE_EIMC,
2579 				    IXGBE_EICR_GPI_SDP0_X550EM_a);
2580 				IXGBE_WRITE_REG(hw, IXGBE_EICR,
2581 				    IXGBE_EICR_GPI_SDP0_X550EM_a);
2582 				retval = hw->phy.ops.check_overtemp(hw);
2583 				if (retval != IXGBE_ERR_OVERTEMP)
2584 					break;
2585 				device_printf(iflib_get_dev(sc->ctx),
2586 				    "\nCRITICAL: OVER TEMP!! PHY IS SHUT DOWN!!\n");
2587 				device_printf(iflib_get_dev(sc->ctx),
2588 				    "System shutdown required!\n");
2589 				break;
2590 			default:
2591 				if (!(eicr & IXGBE_EICR_TS))
2592 					break;
2593 				retval = hw->phy.ops.check_overtemp(hw);
2594 				if (retval != IXGBE_ERR_OVERTEMP)
2595 					break;
2596 				device_printf(iflib_get_dev(sc->ctx),
2597 				    "\nCRITICAL: OVER TEMP!! PHY IS SHUT DOWN!!\n");
2598 				device_printf(iflib_get_dev(sc->ctx),
2599 				    "System shutdown required!\n");
2600 				IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_TS);
2601 				break;
2602 			}
2603 		}
2604 
2605 		/* Check for VF message */
2606 		if ((sc->feat_en & IXGBE_FEATURE_SRIOV) &&
2607 		    (eicr & IXGBE_EICR_MAILBOX))
2608 			sc->task_requests |= IXGBE_REQUEST_TASK_MBX;
2609 	}
2610 
2611 	if (ixgbe_is_sfp(hw)) {
2612 		/* Pluggable optics-related interrupt */
2613 		if (hw->mac.type >= ixgbe_mac_X540)
2614 			eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2615 		else
2616 			eicr_mask = IXGBE_EICR_GPI_SDP2_BY_MAC(hw);
2617 
2618 		if (eicr & eicr_mask) {
2619 			IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2620 			sc->task_requests |= IXGBE_REQUEST_TASK_MOD;
2621 		}
2622 
2623 		if ((hw->mac.type == ixgbe_mac_82599EB) &&
2624 		    (eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw))) {
2625 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2626 			    IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
2627 			sc->task_requests |= IXGBE_REQUEST_TASK_MSF;
2628 		}
2629 	}
2630 
2631 	/* Check for fan failure */
2632 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL) {
2633 		ixgbe_check_fan_failure(sc, eicr, true);
2634 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
2635 	}
2636 
2637 	/* External PHY interrupt */
2638 	if ((hw->phy.type == ixgbe_phy_x550em_ext_t) &&
2639 	    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2640 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP0_X540);
2641 		sc->task_requests |= IXGBE_REQUEST_TASK_PHY;
2642 	}
2643 
2644 	return (sc->task_requests != 0) ? FILTER_SCHEDULE_THREAD : FILTER_HANDLED;
2645 } /* ixgbe_msix_link */
2646 
2647 /************************************************************************
2648  * ixgbe_sysctl_interrupt_rate_handler
2649  ************************************************************************/
2650 static int
2651 ixgbe_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS)
2652 {
2653 	struct ix_rx_queue *que = ((struct ix_rx_queue *)oidp->oid_arg1);
2654 	int                error;
2655 	unsigned int       reg, usec, rate;
2656 
2657 	reg = IXGBE_READ_REG(&que->sc->hw, IXGBE_EITR(que->msix));
2658 	usec = ((reg & 0x0FF8) >> 3);
2659 	if (usec > 0)
2660 		rate = 500000 / usec;
2661 	else
2662 		rate = 0;
2663 	error = sysctl_handle_int(oidp, &rate, 0, req);
2664 	if (error || !req->newptr)
2665 		return error;
2666 	reg &= ~0xfff; /* default, no limitation */
2667 	ixgbe_max_interrupt_rate = 0;
2668 	if (rate > 0 && rate < 500000) {
2669 		if (rate < 1000)
2670 			rate = 1000;
2671 		ixgbe_max_interrupt_rate = rate;
2672 		reg |= ((4000000/rate) & 0xff8);
2673 	}
2674 	IXGBE_WRITE_REG(&que->sc->hw, IXGBE_EITR(que->msix), reg);
2675 
2676 	return (0);
2677 } /* ixgbe_sysctl_interrupt_rate_handler */
2678 
2679 /************************************************************************
2680  * ixgbe_add_device_sysctls
2681  ************************************************************************/
2682 static void
2683 ixgbe_add_device_sysctls(if_ctx_t ctx)
2684 {
2685 	struct ixgbe_softc         *sc = iflib_get_softc(ctx);
2686 	device_t               dev = iflib_get_dev(ctx);
2687 	struct ixgbe_hw        *hw = &sc->hw;
2688 	struct sysctl_oid_list *child;
2689 	struct sysctl_ctx_list *ctx_list;
2690 
2691 	ctx_list = device_get_sysctl_ctx(dev);
2692 	child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
2693 
2694 	/* Sysctls for all devices */
2695 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc",
2696 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2697 	    sc, 0, ixgbe_sysctl_flowcntl, "I",
2698 	    IXGBE_SYSCTL_DESC_SET_FC);
2699 
2700 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "advertise_speed",
2701 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2702 	    sc, 0, ixgbe_sysctl_advertise, "I",
2703 	    IXGBE_SYSCTL_DESC_ADV_SPEED);
2704 
2705 	sc->enable_aim = ixgbe_enable_aim;
2706 	SYSCTL_ADD_INT(ctx_list, child, OID_AUTO, "enable_aim", CTLFLAG_RW,
2707 	    &sc->enable_aim, 0, "Interrupt Moderation");
2708 
2709 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fw_version",
2710 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
2711 	    ixgbe_sysctl_print_fw_version, "A", "Prints FW/NVM Versions");
2712 
2713 #ifdef IXGBE_DEBUG
2714 	/* testing sysctls (for all devices) */
2715 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "power_state",
2716 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2717 	    sc, 0, ixgbe_sysctl_power_state,
2718 	    "I", "PCI Power State");
2719 
2720 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "print_rss_config",
2721 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
2722 	    ixgbe_sysctl_print_rss_config, "A", "Prints RSS Configuration");
2723 #endif
2724 	/* for X550 series devices */
2725 	if (hw->mac.type >= ixgbe_mac_X550)
2726 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "dmac",
2727 		    CTLTYPE_U16 | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2728 		    sc, 0, ixgbe_sysctl_dmac,
2729 		    "I", "DMA Coalesce");
2730 
2731 	/* for WoL-capable devices */
2732 	if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) {
2733 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "wol_enable",
2734 		    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
2735 		    ixgbe_sysctl_wol_enable, "I", "Enable/Disable Wake on LAN");
2736 
2737 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "wufc",
2738 		    CTLTYPE_U32 | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
2739 		    sc, 0, ixgbe_sysctl_wufc,
2740 		    "I", "Enable/Disable Wake Up Filters");
2741 	}
2742 
2743 	/* for X552/X557-AT devices */
2744 	if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) {
2745 		struct sysctl_oid *phy_node;
2746 		struct sysctl_oid_list *phy_list;
2747 
2748 		phy_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, "phy",
2749 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "External PHY sysctls");
2750 		phy_list = SYSCTL_CHILDREN(phy_node);
2751 
2752 		SYSCTL_ADD_PROC(ctx_list, phy_list, OID_AUTO, "temp",
2753 		    CTLTYPE_U16 | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
2754 		    sc, 0, ixgbe_sysctl_phy_temp,
2755 		    "I", "Current External PHY Temperature (Celsius)");
2756 
2757 		SYSCTL_ADD_PROC(ctx_list, phy_list, OID_AUTO,
2758 		    "overtemp_occurred",
2759 		    CTLTYPE_U16 | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
2760 		    ixgbe_sysctl_phy_overtemp_occurred, "I",
2761 		    "External PHY High Temperature Event Occurred");
2762 	}
2763 
2764 	if (sc->feat_cap & IXGBE_FEATURE_EEE) {
2765 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "eee_state",
2766 		    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
2767 		    ixgbe_sysctl_eee_state, "I", "EEE Power Save State");
2768 	}
2769 } /* ixgbe_add_device_sysctls */
2770 
2771 /************************************************************************
2772  * ixgbe_allocate_pci_resources
2773  ************************************************************************/
2774 static int
2775 ixgbe_allocate_pci_resources(if_ctx_t ctx)
2776 {
2777 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2778 	device_t        dev = iflib_get_dev(ctx);
2779 	int             rid;
2780 
2781 	rid = PCIR_BAR(0);
2782 	sc->pci_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
2783 	    RF_ACTIVE);
2784 
2785 	if (!(sc->pci_mem)) {
2786 		device_printf(dev, "Unable to allocate bus resource: memory\n");
2787 		return (ENXIO);
2788 	}
2789 
2790 	/* Save bus_space values for READ/WRITE_REG macros */
2791 	sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->pci_mem);
2792 	sc->osdep.mem_bus_space_handle =
2793 	    rman_get_bushandle(sc->pci_mem);
2794 	/* Set hw values for shared code */
2795 	sc->hw.hw_addr = (u8 *)&sc->osdep.mem_bus_space_handle;
2796 
2797 	return (0);
2798 } /* ixgbe_allocate_pci_resources */
2799 
2800 /************************************************************************
2801  * ixgbe_detach - Device removal routine
2802  *
2803  *   Called when the driver is being removed.
2804  *   Stops the adapter and deallocates all the resources
2805  *   that were allocated for driver operation.
2806  *
2807  *   return 0 on success, positive on failure
2808  ************************************************************************/
2809 static int
2810 ixgbe_if_detach(if_ctx_t ctx)
2811 {
2812 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2813 	device_t       dev = iflib_get_dev(ctx);
2814 	u32            ctrl_ext;
2815 
2816 	INIT_DEBUGOUT("ixgbe_detach: begin");
2817 
2818 	if (ixgbe_pci_iov_detach(dev) != 0) {
2819 		device_printf(dev, "SR-IOV in use; detach first.\n");
2820 		return (EBUSY);
2821 	}
2822 
2823 	ixgbe_setup_low_power_mode(ctx);
2824 
2825 	/* let hardware know driver is unloading */
2826 	ctrl_ext = IXGBE_READ_REG(&sc->hw, IXGBE_CTRL_EXT);
2827 	ctrl_ext &= ~IXGBE_CTRL_EXT_DRV_LOAD;
2828 	IXGBE_WRITE_REG(&sc->hw, IXGBE_CTRL_EXT, ctrl_ext);
2829 
2830 	ixgbe_free_pci_resources(ctx);
2831 	free(sc->mta, M_IXGBE);
2832 
2833 	return (0);
2834 } /* ixgbe_if_detach */
2835 
2836 /************************************************************************
2837  * ixgbe_setup_low_power_mode - LPLU/WoL preparation
2838  *
2839  *   Prepare the adapter/port for LPLU and/or WoL
2840  ************************************************************************/
2841 static int
2842 ixgbe_setup_low_power_mode(if_ctx_t ctx)
2843 {
2844 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
2845 	struct ixgbe_hw *hw = &sc->hw;
2846 	device_t        dev = iflib_get_dev(ctx);
2847 	s32             error = 0;
2848 
2849 	if (!hw->wol_enabled)
2850 		ixgbe_set_phy_power(hw, false);
2851 
2852 	/* Limit power management flow to X550EM baseT */
2853 	if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T &&
2854 	    hw->phy.ops.enter_lplu) {
2855 		/* Turn off support for APM wakeup. (Using ACPI instead) */
2856 		IXGBE_WRITE_REG(hw, IXGBE_GRC_BY_MAC(hw),
2857 		    IXGBE_READ_REG(hw, IXGBE_GRC_BY_MAC(hw)) & ~(u32)2);
2858 
2859 		/*
2860 		 * Clear Wake Up Status register to prevent any previous wakeup
2861 		 * events from waking us up immediately after we suspend.
2862 		 */
2863 		IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff);
2864 
2865 		/*
2866 		 * Program the Wakeup Filter Control register with user filter
2867 		 * settings
2868 		 */
2869 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, sc->wufc);
2870 
2871 		/* Enable wakeups and power management in Wakeup Control */
2872 		IXGBE_WRITE_REG(hw, IXGBE_WUC,
2873 		    IXGBE_WUC_WKEN | IXGBE_WUC_PME_EN);
2874 
2875 		/* X550EM baseT adapters need a special LPLU flow */
2876 		hw->phy.reset_disable = true;
2877 		ixgbe_if_stop(ctx);
2878 		error = hw->phy.ops.enter_lplu(hw);
2879 		if (error)
2880 			device_printf(dev, "Error entering LPLU: %d\n", error);
2881 		hw->phy.reset_disable = false;
2882 	} else {
2883 		/* Just stop for other adapters */
2884 		ixgbe_if_stop(ctx);
2885 	}
2886 
2887 	return error;
2888 } /* ixgbe_setup_low_power_mode */
2889 
2890 /************************************************************************
2891  * ixgbe_shutdown - Shutdown entry point
2892  ************************************************************************/
2893 static int
2894 ixgbe_if_shutdown(if_ctx_t ctx)
2895 {
2896 	int error = 0;
2897 
2898 	INIT_DEBUGOUT("ixgbe_shutdown: begin");
2899 
2900 	error = ixgbe_setup_low_power_mode(ctx);
2901 
2902 	return (error);
2903 } /* ixgbe_if_shutdown */
2904 
2905 /************************************************************************
2906  * ixgbe_suspend
2907  *
2908  *   From D0 to D3
2909  ************************************************************************/
2910 static int
2911 ixgbe_if_suspend(if_ctx_t ctx)
2912 {
2913 	int error = 0;
2914 
2915 	INIT_DEBUGOUT("ixgbe_suspend: begin");
2916 
2917 	error = ixgbe_setup_low_power_mode(ctx);
2918 
2919 	return (error);
2920 } /* ixgbe_if_suspend */
2921 
2922 /************************************************************************
2923  * ixgbe_resume
2924  *
2925  *   From D3 to D0
2926  ************************************************************************/
2927 static int
2928 ixgbe_if_resume(if_ctx_t ctx)
2929 {
2930 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
2931 	device_t        dev = iflib_get_dev(ctx);
2932 	if_t            ifp = iflib_get_ifp(ctx);
2933 	struct ixgbe_hw *hw = &sc->hw;
2934 	u32             wus;
2935 
2936 	INIT_DEBUGOUT("ixgbe_resume: begin");
2937 
2938 	/* Read & clear WUS register */
2939 	wus = IXGBE_READ_REG(hw, IXGBE_WUS);
2940 	if (wus)
2941 		device_printf(dev, "Woken up by (WUS): %#010x\n",
2942 		    IXGBE_READ_REG(hw, IXGBE_WUS));
2943 	IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff);
2944 	/* And clear WUFC until next low-power transition */
2945 	IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
2946 
2947 	/*
2948 	 * Required after D3->D0 transition;
2949 	 * will re-advertise all previous advertised speeds
2950 	 */
2951 	if (if_getflags(ifp) & IFF_UP)
2952 		ixgbe_if_init(ctx);
2953 
2954 	return (0);
2955 } /* ixgbe_if_resume */
2956 
2957 /************************************************************************
2958  * ixgbe_if_mtu_set - Ioctl mtu entry point
2959  *
2960  *   Return 0 on success, EINVAL on failure
2961  ************************************************************************/
2962 static int
2963 ixgbe_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
2964 {
2965 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2966 	int error = 0;
2967 
2968 	IOCTL_DEBUGOUT("ioctl: SIOCIFMTU (Set Interface MTU)");
2969 
2970 	if (mtu > IXGBE_MAX_MTU) {
2971 		error = EINVAL;
2972 	} else {
2973 		sc->max_frame_size = mtu + IXGBE_MTU_HDR;
2974 	}
2975 
2976 	return error;
2977 } /* ixgbe_if_mtu_set */
2978 
2979 /************************************************************************
2980  * ixgbe_if_crcstrip_set
2981  ************************************************************************/
2982 static void
2983 ixgbe_if_crcstrip_set(if_ctx_t ctx, int onoff, int crcstrip)
2984 {
2985 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
2986 	struct ixgbe_hw *hw = &sc->hw;
2987 	/* crc stripping is set in two places:
2988 	 * IXGBE_HLREG0 (modified on init_locked and hw reset)
2989 	 * IXGBE_RDRXCTL (set by the original driver in
2990 	 *	ixgbe_setup_hw_rsc() called in init_locked.
2991 	 *	We disable the setting when netmap is compiled in).
2992 	 * We update the values here, but also in ixgbe.c because
2993 	 * init_locked sometimes is called outside our control.
2994 	 */
2995 	uint32_t hl, rxc;
2996 
2997 	hl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
2998 	rxc = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
2999 #ifdef NETMAP
3000 	if (netmap_verbose)
3001 		D("%s read  HLREG 0x%x rxc 0x%x",
3002 			onoff ? "enter" : "exit", hl, rxc);
3003 #endif
3004 	/* hw requirements ... */
3005 	rxc &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3006 	rxc |= IXGBE_RDRXCTL_RSCACKC;
3007 	if (onoff && !crcstrip) {
3008 		/* keep the crc. Fast rx */
3009 		hl &= ~IXGBE_HLREG0_RXCRCSTRP;
3010 		rxc &= ~IXGBE_RDRXCTL_CRCSTRIP;
3011 	} else {
3012 		/* reset default mode */
3013 		hl |= IXGBE_HLREG0_RXCRCSTRP;
3014 		rxc |= IXGBE_RDRXCTL_CRCSTRIP;
3015 	}
3016 #ifdef NETMAP
3017 	if (netmap_verbose)
3018 		D("%s write HLREG 0x%x rxc 0x%x",
3019 			onoff ? "enter" : "exit", hl, rxc);
3020 #endif
3021 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hl);
3022 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rxc);
3023 } /* ixgbe_if_crcstrip_set */
3024 
3025 /*********************************************************************
3026  * ixgbe_if_init - Init entry point
3027  *
3028  *   Used in two ways: It is used by the stack as an init
3029  *   entry point in network interface structure. It is also
3030  *   used by the driver as a hw/sw initialization routine to
3031  *   get to a consistent state.
3032  *
3033  *   Return 0 on success, positive on failure
3034  **********************************************************************/
3035 void
3036 ixgbe_if_init(if_ctx_t ctx)
3037 {
3038 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
3039 	if_t               ifp = iflib_get_ifp(ctx);
3040 	device_t           dev = iflib_get_dev(ctx);
3041 	struct ixgbe_hw *hw = &sc->hw;
3042 	struct ix_rx_queue *rx_que;
3043 	struct ix_tx_queue *tx_que;
3044 	u32             txdctl, mhadd;
3045 	u32             rxdctl, rxctrl;
3046 	u32             ctrl_ext;
3047 
3048 	int             i, j, err;
3049 
3050 	INIT_DEBUGOUT("ixgbe_if_init: begin");
3051 
3052 	/* Queue indices may change with IOV mode */
3053 	ixgbe_align_all_queue_indices(sc);
3054 
3055 	/* reprogram the RAR[0] in case user changed it. */
3056 	ixgbe_set_rar(hw, 0, hw->mac.addr, sc->pool, IXGBE_RAH_AV);
3057 
3058 	/* Get the latest mac address, User can use a LAA */
3059 	bcopy(if_getlladdr(ifp), hw->mac.addr, IXGBE_ETH_LENGTH_OF_ADDRESS);
3060 	ixgbe_set_rar(hw, 0, hw->mac.addr, sc->pool, 1);
3061 	hw->addr_ctrl.rar_used_count = 1;
3062 
3063 	ixgbe_init_hw(hw);
3064 
3065 	ixgbe_initialize_iov(sc);
3066 
3067 	ixgbe_initialize_transmit_units(ctx);
3068 
3069 	/* Setup Multicast table */
3070 	ixgbe_if_multi_set(ctx);
3071 
3072 	/* Determine the correct mbuf pool, based on frame size */
3073 	sc->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
3074 
3075 	/* Configure RX settings */
3076 	ixgbe_initialize_receive_units(ctx);
3077 
3078 	/*
3079 	 * Initialize variable holding task enqueue requests
3080 	 * from MSI-X interrupts
3081 	 */
3082 	sc->task_requests = 0;
3083 
3084 	/* Enable SDP & MSI-X interrupts based on adapter */
3085 	ixgbe_config_gpie(sc);
3086 
3087 	/* Set MTU size */
3088 	if (if_getmtu(ifp) > ETHERMTU) {
3089 		/* aka IXGBE_MAXFRS on 82599 and newer */
3090 		mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3091 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3092 		mhadd |= sc->max_frame_size << IXGBE_MHADD_MFS_SHIFT;
3093 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3094 	}
3095 
3096 	/* Now enable all the queues */
3097 	for (i = 0, tx_que = sc->tx_queues; i < sc->num_tx_queues; i++, tx_que++) {
3098 		struct tx_ring *txr = &tx_que->txr;
3099 
3100 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(txr->me));
3101 		txdctl |= IXGBE_TXDCTL_ENABLE;
3102 		/* Set WTHRESH to 8, burst writeback */
3103 		txdctl |= (8 << 16);
3104 		/*
3105 		 * When the internal queue falls below PTHRESH (32),
3106 		 * start prefetching as long as there are at least
3107 		 * HTHRESH (1) buffers ready. The values are taken
3108 		 * from the Intel linux driver 3.8.21.
3109 		 * Prefetching enables tx line rate even with 1 queue.
3110 		 */
3111 		txdctl |= (32 << 0) | (1 << 8);
3112 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(txr->me), txdctl);
3113 	}
3114 
3115 	for (i = 0, rx_que = sc->rx_queues; i < sc->num_rx_queues; i++, rx_que++) {
3116 		struct rx_ring *rxr = &rx_que->rxr;
3117 
3118 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me));
3119 		if (hw->mac.type == ixgbe_mac_82598EB) {
3120 			/*
3121 			 * PTHRESH = 21
3122 			 * HTHRESH = 4
3123 			 * WTHRESH = 8
3124 			 */
3125 			rxdctl &= ~0x3FFFFF;
3126 			rxdctl |= 0x080420;
3127 		}
3128 		rxdctl |= IXGBE_RXDCTL_ENABLE;
3129 		IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), rxdctl);
3130 		for (j = 0; j < 10; j++) {
3131 			if (IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me)) &
3132 			    IXGBE_RXDCTL_ENABLE)
3133 				break;
3134 			else
3135 				msec_delay(1);
3136 		}
3137 		wmb();
3138 	}
3139 
3140 	/* Enable Receive engine */
3141 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3142 	if (hw->mac.type == ixgbe_mac_82598EB)
3143 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3144 	rxctrl |= IXGBE_RXCTRL_RXEN;
3145 	ixgbe_enable_rx_dma(hw, rxctrl);
3146 
3147 	/* Set up MSI/MSI-X routing */
3148 	if (ixgbe_enable_msix)  {
3149 		ixgbe_configure_ivars(sc);
3150 		/* Set up auto-mask */
3151 		if (hw->mac.type == ixgbe_mac_82598EB)
3152 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3153 		else {
3154 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
3155 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
3156 		}
3157 	} else {  /* Simple settings for Legacy/MSI */
3158 		ixgbe_set_ivar(sc, 0, 0, 0);
3159 		ixgbe_set_ivar(sc, 0, 0, 1);
3160 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3161 	}
3162 
3163 	ixgbe_init_fdir(sc);
3164 
3165 	/*
3166 	 * Check on any SFP devices that
3167 	 * need to be kick-started
3168 	 */
3169 	if (hw->phy.type == ixgbe_phy_none) {
3170 		err = hw->phy.ops.identify(hw);
3171 		if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3172 			device_printf(dev,
3173 			    "Unsupported SFP+ module type was detected.\n");
3174 			return;
3175 		}
3176 	}
3177 
3178 	/* Set moderation on the Link interrupt */
3179 	IXGBE_WRITE_REG(hw, IXGBE_EITR(sc->vector), IXGBE_LINK_ITR);
3180 
3181 	/* Enable power to the phy. */
3182 	ixgbe_set_phy_power(hw, true);
3183 
3184 	/* Config/Enable Link */
3185 	ixgbe_config_link(ctx);
3186 
3187 	/* Hardware Packet Buffer & Flow Control setup */
3188 	ixgbe_config_delay_values(sc);
3189 
3190 	/* Initialize the FC settings */
3191 	ixgbe_start_hw(hw);
3192 
3193 	/* Set up VLAN support and filter */
3194 	ixgbe_setup_vlan_hw_support(ctx);
3195 
3196 	/* Setup DMA Coalescing */
3197 	ixgbe_config_dmac(sc);
3198 
3199 	/* And now turn on interrupts */
3200 	ixgbe_if_enable_intr(ctx);
3201 
3202 	/* Enable the use of the MBX by the VF's */
3203 	if (sc->feat_en & IXGBE_FEATURE_SRIOV) {
3204 		ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
3205 		ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
3206 		IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
3207 	}
3208 
3209 } /* ixgbe_init_locked */
3210 
3211 /************************************************************************
3212  * ixgbe_set_ivar
3213  *
3214  *   Setup the correct IVAR register for a particular MSI-X interrupt
3215  *     (yes this is all very magic and confusing :)
3216  *    - entry is the register array entry
3217  *    - vector is the MSI-X vector for this queue
3218  *    - type is RX/TX/MISC
3219  ************************************************************************/
3220 static void
3221 ixgbe_set_ivar(struct ixgbe_softc *sc, u8 entry, u8 vector, s8 type)
3222 {
3223 	struct ixgbe_hw *hw = &sc->hw;
3224 	u32 ivar, index;
3225 
3226 	vector |= IXGBE_IVAR_ALLOC_VAL;
3227 
3228 	switch (hw->mac.type) {
3229 	case ixgbe_mac_82598EB:
3230 		if (type == -1)
3231 			entry = IXGBE_IVAR_OTHER_CAUSES_INDEX;
3232 		else
3233 			entry += (type * 64);
3234 		index = (entry >> 2) & 0x1F;
3235 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
3236 		ivar &= ~(0xFF << (8 * (entry & 0x3)));
3237 		ivar |= (vector << (8 * (entry & 0x3)));
3238 		IXGBE_WRITE_REG(&sc->hw, IXGBE_IVAR(index), ivar);
3239 		break;
3240 	case ixgbe_mac_82599EB:
3241 	case ixgbe_mac_X540:
3242 	case ixgbe_mac_X550:
3243 	case ixgbe_mac_X550EM_x:
3244 	case ixgbe_mac_X550EM_a:
3245 		if (type == -1) { /* MISC IVAR */
3246 			index = (entry & 1) * 8;
3247 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
3248 			ivar &= ~(0xFF << index);
3249 			ivar |= (vector << index);
3250 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, ivar);
3251 		} else {          /* RX/TX IVARS */
3252 			index = (16 * (entry & 1)) + (8 * type);
3253 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(entry >> 1));
3254 			ivar &= ~(0xFF << index);
3255 			ivar |= (vector << index);
3256 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(entry >> 1), ivar);
3257 		}
3258 	default:
3259 		break;
3260 	}
3261 } /* ixgbe_set_ivar */
3262 
3263 /************************************************************************
3264  * ixgbe_configure_ivars
3265  ************************************************************************/
3266 static void
3267 ixgbe_configure_ivars(struct ixgbe_softc *sc)
3268 {
3269 	struct ix_rx_queue *rx_que = sc->rx_queues;
3270 	struct ix_tx_queue *tx_que = sc->tx_queues;
3271 	u32                newitr;
3272 
3273 	if (ixgbe_max_interrupt_rate > 0)
3274 		newitr = (4000000 / ixgbe_max_interrupt_rate) & 0x0FF8;
3275 	else {
3276 		/*
3277 		 * Disable DMA coalescing if interrupt moderation is
3278 		 * disabled.
3279 		 */
3280 		sc->dmac = 0;
3281 		newitr = 0;
3282 	}
3283 
3284 	for (int i = 0; i < sc->num_rx_queues; i++, rx_que++) {
3285 		struct rx_ring *rxr = &rx_que->rxr;
3286 
3287 		/* First the RX queue entry */
3288 		ixgbe_set_ivar(sc, rxr->me, rx_que->msix, 0);
3289 
3290 		/* Set an Initial EITR value */
3291 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EITR(rx_que->msix), newitr);
3292 	}
3293 	for (int i = 0; i < sc->num_tx_queues; i++, tx_que++) {
3294 		struct tx_ring *txr = &tx_que->txr;
3295 
3296 		/* ... and the TX */
3297 		ixgbe_set_ivar(sc, txr->me, tx_que->msix, 1);
3298 	}
3299 	/* For the Link interrupt */
3300 	ixgbe_set_ivar(sc, 1, sc->vector, -1);
3301 } /* ixgbe_configure_ivars */
3302 
3303 /************************************************************************
3304  * ixgbe_config_gpie
3305  ************************************************************************/
3306 static void
3307 ixgbe_config_gpie(struct ixgbe_softc *sc)
3308 {
3309 	struct ixgbe_hw *hw = &sc->hw;
3310 	u32             gpie;
3311 
3312 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
3313 
3314 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3315 		/* Enable Enhanced MSI-X mode */
3316 		gpie |= IXGBE_GPIE_MSIX_MODE
3317 		     |  IXGBE_GPIE_EIAME
3318 		     |  IXGBE_GPIE_PBA_SUPPORT
3319 		     |  IXGBE_GPIE_OCD;
3320 	}
3321 
3322 	/* Fan Failure Interrupt */
3323 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL)
3324 		gpie |= IXGBE_SDP1_GPIEN;
3325 
3326 	/* Thermal Sensor Interrupt */
3327 	if (sc->feat_en & IXGBE_FEATURE_TEMP_SENSOR)
3328 		gpie |= IXGBE_SDP0_GPIEN_X540;
3329 
3330 	/* Link detection */
3331 	switch (hw->mac.type) {
3332 	case ixgbe_mac_82599EB:
3333 		gpie |= IXGBE_SDP1_GPIEN | IXGBE_SDP2_GPIEN;
3334 		break;
3335 	case ixgbe_mac_X550EM_x:
3336 	case ixgbe_mac_X550EM_a:
3337 		gpie |= IXGBE_SDP0_GPIEN_X540;
3338 		break;
3339 	default:
3340 		break;
3341 	}
3342 
3343 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
3344 
3345 } /* ixgbe_config_gpie */
3346 
3347 /************************************************************************
3348  * ixgbe_config_delay_values
3349  *
3350  *   Requires sc->max_frame_size to be set.
3351  ************************************************************************/
3352 static void
3353 ixgbe_config_delay_values(struct ixgbe_softc *sc)
3354 {
3355 	struct ixgbe_hw *hw = &sc->hw;
3356 	u32             rxpb, frame, size, tmp;
3357 
3358 	frame = sc->max_frame_size;
3359 
3360 	/* Calculate High Water */
3361 	switch (hw->mac.type) {
3362 	case ixgbe_mac_X540:
3363 	case ixgbe_mac_X550:
3364 	case ixgbe_mac_X550EM_x:
3365 	case ixgbe_mac_X550EM_a:
3366 		tmp = IXGBE_DV_X540(frame, frame);
3367 		break;
3368 	default:
3369 		tmp = IXGBE_DV(frame, frame);
3370 		break;
3371 	}
3372 	size = IXGBE_BT2KB(tmp);
3373 	rxpb = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0)) >> 10;
3374 	hw->fc.high_water[0] = rxpb - size;
3375 
3376 	/* Now calculate Low Water */
3377 	switch (hw->mac.type) {
3378 	case ixgbe_mac_X540:
3379 	case ixgbe_mac_X550:
3380 	case ixgbe_mac_X550EM_x:
3381 	case ixgbe_mac_X550EM_a:
3382 		tmp = IXGBE_LOW_DV_X540(frame);
3383 		break;
3384 	default:
3385 		tmp = IXGBE_LOW_DV(frame);
3386 		break;
3387 	}
3388 	hw->fc.low_water[0] = IXGBE_BT2KB(tmp);
3389 
3390 	hw->fc.pause_time = IXGBE_FC_PAUSE;
3391 	hw->fc.send_xon = true;
3392 } /* ixgbe_config_delay_values */
3393 
3394 /************************************************************************
3395  * ixgbe_set_multi - Multicast Update
3396  *
3397  *   Called whenever multicast address list is updated.
3398  ************************************************************************/
3399 static u_int
3400 ixgbe_mc_filter_apply(void *arg, struct sockaddr_dl *sdl, u_int idx)
3401 {
3402 	struct ixgbe_softc *sc = arg;
3403 	struct ixgbe_mc_addr *mta = sc->mta;
3404 
3405 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
3406 		return (0);
3407 	bcopy(LLADDR(sdl), mta[idx].addr, IXGBE_ETH_LENGTH_OF_ADDRESS);
3408 	mta[idx].vmdq = sc->pool;
3409 
3410 	return (1);
3411 } /* ixgbe_mc_filter_apply */
3412 
3413 static void
3414 ixgbe_if_multi_set(if_ctx_t ctx)
3415 {
3416 	struct ixgbe_softc       *sc = iflib_get_softc(ctx);
3417 	struct ixgbe_mc_addr *mta;
3418 	if_t                  ifp = iflib_get_ifp(ctx);
3419 	u8                   *update_ptr;
3420 	u32                  fctrl;
3421 	u_int		     mcnt;
3422 
3423 	IOCTL_DEBUGOUT("ixgbe_if_multi_set: begin");
3424 
3425 	mta = sc->mta;
3426 	bzero(mta, sizeof(*mta) * MAX_NUM_MULTICAST_ADDRESSES);
3427 
3428 	mcnt = if_foreach_llmaddr(iflib_get_ifp(ctx), ixgbe_mc_filter_apply, sc);
3429 
3430 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) {
3431 		update_ptr = (u8 *)mta;
3432 		ixgbe_update_mc_addr_list(&sc->hw, update_ptr, mcnt,
3433 		    ixgbe_mc_array_itr, true);
3434 	}
3435 
3436 	fctrl = IXGBE_READ_REG(&sc->hw, IXGBE_FCTRL);
3437 
3438 	if (if_getflags(ifp) & IFF_PROMISC)
3439 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3440 	else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
3441 	    if_getflags(ifp) & IFF_ALLMULTI) {
3442 		fctrl |= IXGBE_FCTRL_MPE;
3443 		fctrl &= ~IXGBE_FCTRL_UPE;
3444 	} else
3445 		fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3446 
3447 	IXGBE_WRITE_REG(&sc->hw, IXGBE_FCTRL, fctrl);
3448 } /* ixgbe_if_multi_set */
3449 
3450 /************************************************************************
3451  * ixgbe_mc_array_itr
3452  *
3453  *   An iterator function needed by the multicast shared code.
3454  *   It feeds the shared code routine the addresses in the
3455  *   array of ixgbe_set_multi() one by one.
3456  ************************************************************************/
3457 static u8 *
3458 ixgbe_mc_array_itr(struct ixgbe_hw *hw, u8 **update_ptr, u32 *vmdq)
3459 {
3460 	struct ixgbe_mc_addr *mta;
3461 
3462 	mta = (struct ixgbe_mc_addr *)*update_ptr;
3463 	*vmdq = mta->vmdq;
3464 
3465 	*update_ptr = (u8*)(mta + 1);
3466 
3467 	return (mta->addr);
3468 } /* ixgbe_mc_array_itr */
3469 
3470 /************************************************************************
3471  * ixgbe_local_timer - Timer routine
3472  *
3473  *   Checks for link status, updates statistics,
3474  *   and runs the watchdog check.
3475  ************************************************************************/
3476 static void
3477 ixgbe_if_timer(if_ctx_t ctx, uint16_t qid)
3478 {
3479 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3480 
3481 	if (qid != 0)
3482 		return;
3483 
3484 	/* Check for pluggable optics */
3485 	if (sc->sfp_probe)
3486 		if (!ixgbe_sfp_probe(ctx))
3487 			return; /* Nothing to do */
3488 
3489 	ixgbe_check_link(&sc->hw, &sc->link_speed, &sc->link_up, 0);
3490 
3491 	/* Fire off the adminq task */
3492 	iflib_admin_intr_deferred(ctx);
3493 
3494 } /* ixgbe_if_timer */
3495 
3496 /************************************************************************
3497  * ixgbe_sfp_probe
3498  *
3499  *   Determine if a port had optics inserted.
3500  ************************************************************************/
3501 static bool
3502 ixgbe_sfp_probe(if_ctx_t ctx)
3503 {
3504 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3505 	struct ixgbe_hw *hw = &sc->hw;
3506 	device_t        dev = iflib_get_dev(ctx);
3507 	bool            result = false;
3508 
3509 	if ((hw->phy.type == ixgbe_phy_nl) &&
3510 	    (hw->phy.sfp_type == ixgbe_sfp_type_not_present)) {
3511 		s32 ret = hw->phy.ops.identify_sfp(hw);
3512 		if (ret)
3513 			goto out;
3514 		ret = hw->phy.ops.reset(hw);
3515 		sc->sfp_probe = false;
3516 		if (ret == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3517 			device_printf(dev, "Unsupported SFP+ module detected!");
3518 			device_printf(dev,
3519 			    "Reload driver with supported module.\n");
3520 			goto out;
3521 		} else
3522 			device_printf(dev, "SFP+ module detected!\n");
3523 		/* We now have supported optics */
3524 		result = true;
3525 	}
3526 out:
3527 
3528 	return (result);
3529 } /* ixgbe_sfp_probe */
3530 
3531 /************************************************************************
3532  * ixgbe_handle_mod - Tasklet for SFP module interrupts
3533  ************************************************************************/
3534 static void
3535 ixgbe_handle_mod(void *context)
3536 {
3537 	if_ctx_t        ctx = context;
3538 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3539 	struct ixgbe_hw *hw = &sc->hw;
3540 	device_t        dev = iflib_get_dev(ctx);
3541 	u32             err, cage_full = 0;
3542 
3543 	if (sc->hw.need_crosstalk_fix) {
3544 		switch (hw->mac.type) {
3545 		case ixgbe_mac_82599EB:
3546 			cage_full = IXGBE_READ_REG(hw, IXGBE_ESDP) &
3547 			    IXGBE_ESDP_SDP2;
3548 			break;
3549 		case ixgbe_mac_X550EM_x:
3550 		case ixgbe_mac_X550EM_a:
3551 			cage_full = IXGBE_READ_REG(hw, IXGBE_ESDP) &
3552 			    IXGBE_ESDP_SDP0;
3553 			break;
3554 		default:
3555 			break;
3556 		}
3557 
3558 		if (!cage_full)
3559 			goto handle_mod_out;
3560 	}
3561 
3562 	err = hw->phy.ops.identify_sfp(hw);
3563 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3564 		device_printf(dev,
3565 		    "Unsupported SFP+ module type was detected.\n");
3566 		goto handle_mod_out;
3567 	}
3568 
3569 	if (hw->mac.type == ixgbe_mac_82598EB)
3570 		err = hw->phy.ops.reset(hw);
3571 	else
3572 		err = hw->mac.ops.setup_sfp(hw);
3573 
3574 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
3575 		device_printf(dev,
3576 		    "Setup failure - unsupported SFP+ module type.\n");
3577 		goto handle_mod_out;
3578 	}
3579 	sc->task_requests |= IXGBE_REQUEST_TASK_MSF;
3580 	return;
3581 
3582 handle_mod_out:
3583 	sc->task_requests &= ~(IXGBE_REQUEST_TASK_MSF);
3584 } /* ixgbe_handle_mod */
3585 
3586 
3587 /************************************************************************
3588  * ixgbe_handle_msf - Tasklet for MSF (multispeed fiber) interrupts
3589  ************************************************************************/
3590 static void
3591 ixgbe_handle_msf(void *context)
3592 {
3593 	if_ctx_t        ctx = context;
3594 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3595 	struct ixgbe_hw *hw = &sc->hw;
3596 	u32             autoneg;
3597 	bool            negotiate;
3598 
3599 	/* get_supported_phy_layer will call hw->phy.ops.identify_sfp() */
3600 	sc->phy_layer = ixgbe_get_supported_physical_layer(hw);
3601 
3602 	autoneg = hw->phy.autoneg_advertised;
3603 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
3604 		hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiate);
3605 	if (hw->mac.ops.setup_link)
3606 		hw->mac.ops.setup_link(hw, autoneg, true);
3607 
3608 	/* Adjust media types shown in ifconfig */
3609 	ifmedia_removeall(sc->media);
3610 	ixgbe_add_media_types(sc->ctx);
3611 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
3612 } /* ixgbe_handle_msf */
3613 
3614 /************************************************************************
3615  * ixgbe_handle_phy - Tasklet for external PHY interrupts
3616  ************************************************************************/
3617 static void
3618 ixgbe_handle_phy(void *context)
3619 {
3620 	if_ctx_t        ctx = context;
3621 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3622 	struct ixgbe_hw *hw = &sc->hw;
3623 	int             error;
3624 
3625 	error = hw->phy.ops.handle_lasi(hw);
3626 	if (error == IXGBE_ERR_OVERTEMP)
3627 		device_printf(sc->dev, "CRITICAL: EXTERNAL PHY OVER TEMP!!  PHY will downshift to lower power state!\n");
3628 	else if (error)
3629 		device_printf(sc->dev,
3630 		    "Error handling LASI interrupt: %d\n", error);
3631 } /* ixgbe_handle_phy */
3632 
3633 /************************************************************************
3634  * ixgbe_if_stop - Stop the hardware
3635  *
3636  *   Disables all traffic on the adapter by issuing a
3637  *   global reset on the MAC and deallocates TX/RX buffers.
3638  ************************************************************************/
3639 static void
3640 ixgbe_if_stop(if_ctx_t ctx)
3641 {
3642 	struct ixgbe_softc  *sc = iflib_get_softc(ctx);
3643 	struct ixgbe_hw *hw = &sc->hw;
3644 
3645 	INIT_DEBUGOUT("ixgbe_if_stop: begin\n");
3646 
3647 	ixgbe_reset_hw(hw);
3648 	hw->adapter_stopped = false;
3649 	ixgbe_stop_adapter(hw);
3650 	if (hw->mac.type == ixgbe_mac_82599EB)
3651 		ixgbe_stop_mac_link_on_d3_82599(hw);
3652 	/* Turn off the laser - noop with no optics */
3653 	ixgbe_disable_tx_laser(hw);
3654 
3655 	/* Update the stack */
3656 	sc->link_up = false;
3657 	ixgbe_if_update_admin_status(ctx);
3658 
3659 	/* reprogram the RAR[0] in case user changed it. */
3660 	ixgbe_set_rar(&sc->hw, 0, sc->hw.mac.addr, 0, IXGBE_RAH_AV);
3661 
3662 	return;
3663 } /* ixgbe_if_stop */
3664 
3665 /************************************************************************
3666  * ixgbe_update_link_status - Update OS on link state
3667  *
3668  * Note: Only updates the OS on the cached link state.
3669  *       The real check of the hardware only happens with
3670  *       a link interrupt.
3671  ************************************************************************/
3672 static void
3673 ixgbe_if_update_admin_status(if_ctx_t ctx)
3674 {
3675 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3676 	device_t       dev = iflib_get_dev(ctx);
3677 
3678 	if (sc->link_up) {
3679 		if (sc->link_active == false) {
3680 			if (bootverbose)
3681 				device_printf(dev, "Link is up %d Gbps %s \n",
3682 				    ((sc->link_speed == 128) ? 10 : 1),
3683 				    "Full Duplex");
3684 			sc->link_active = true;
3685 			/* Update any Flow Control changes */
3686 			ixgbe_fc_enable(&sc->hw);
3687 			/* Update DMA coalescing config */
3688 			ixgbe_config_dmac(sc);
3689 			iflib_link_state_change(ctx, LINK_STATE_UP,
3690 			    ixgbe_link_speed_to_baudrate(sc->link_speed));
3691 
3692 			if (sc->feat_en & IXGBE_FEATURE_SRIOV)
3693 				ixgbe_ping_all_vfs(sc);
3694 		}
3695 	} else { /* Link down */
3696 		if (sc->link_active == true) {
3697 			if (bootverbose)
3698 				device_printf(dev, "Link is Down\n");
3699 			iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
3700 			sc->link_active = false;
3701 			if (sc->feat_en & IXGBE_FEATURE_SRIOV)
3702 				ixgbe_ping_all_vfs(sc);
3703 		}
3704 	}
3705 
3706 	/* Handle task requests from msix_link() */
3707 	if (sc->task_requests & IXGBE_REQUEST_TASK_MOD)
3708 		ixgbe_handle_mod(ctx);
3709 	if (sc->task_requests & IXGBE_REQUEST_TASK_MSF)
3710 		ixgbe_handle_msf(ctx);
3711 	if (sc->task_requests & IXGBE_REQUEST_TASK_MBX)
3712 		ixgbe_handle_mbx(ctx);
3713 	if (sc->task_requests & IXGBE_REQUEST_TASK_FDIR)
3714 		ixgbe_reinit_fdir(ctx);
3715 	if (sc->task_requests & IXGBE_REQUEST_TASK_PHY)
3716 		ixgbe_handle_phy(ctx);
3717 	sc->task_requests = 0;
3718 
3719 	ixgbe_update_stats_counters(sc);
3720 } /* ixgbe_if_update_admin_status */
3721 
3722 /************************************************************************
3723  * ixgbe_config_dmac - Configure DMA Coalescing
3724  ************************************************************************/
3725 static void
3726 ixgbe_config_dmac(struct ixgbe_softc *sc)
3727 {
3728 	struct ixgbe_hw          *hw = &sc->hw;
3729 	struct ixgbe_dmac_config *dcfg = &hw->mac.dmac_config;
3730 
3731 	if (hw->mac.type < ixgbe_mac_X550 || !hw->mac.ops.dmac_config)
3732 		return;
3733 
3734 	if (dcfg->watchdog_timer ^ sc->dmac ||
3735 	    dcfg->link_speed ^ sc->link_speed) {
3736 		dcfg->watchdog_timer = sc->dmac;
3737 		dcfg->fcoe_en = false;
3738 		dcfg->link_speed = sc->link_speed;
3739 		dcfg->num_tcs = 1;
3740 
3741 		INIT_DEBUGOUT2("dmac settings: watchdog %d, link speed %d\n",
3742 		    dcfg->watchdog_timer, dcfg->link_speed);
3743 
3744 		hw->mac.ops.dmac_config(hw);
3745 	}
3746 } /* ixgbe_config_dmac */
3747 
3748 /************************************************************************
3749  * ixgbe_if_enable_intr
3750  ************************************************************************/
3751 void
3752 ixgbe_if_enable_intr(if_ctx_t ctx)
3753 {
3754 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
3755 	struct ixgbe_hw    *hw = &sc->hw;
3756 	struct ix_rx_queue *que = sc->rx_queues;
3757 	u32                mask, fwsm;
3758 
3759 	mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
3760 
3761 	switch (sc->hw.mac.type) {
3762 	case ixgbe_mac_82599EB:
3763 		mask |= IXGBE_EIMS_ECC;
3764 		/* Temperature sensor on some scs */
3765 		mask |= IXGBE_EIMS_GPI_SDP0;
3766 		/* SFP+ (RX_LOS_N & MOD_ABS_N) */
3767 		mask |= IXGBE_EIMS_GPI_SDP1;
3768 		mask |= IXGBE_EIMS_GPI_SDP2;
3769 		break;
3770 	case ixgbe_mac_X540:
3771 		/* Detect if Thermal Sensor is enabled */
3772 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
3773 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
3774 			mask |= IXGBE_EIMS_TS;
3775 		mask |= IXGBE_EIMS_ECC;
3776 		break;
3777 	case ixgbe_mac_X550:
3778 		/* MAC thermal sensor is automatically enabled */
3779 		mask |= IXGBE_EIMS_TS;
3780 		mask |= IXGBE_EIMS_ECC;
3781 		break;
3782 	case ixgbe_mac_X550EM_x:
3783 	case ixgbe_mac_X550EM_a:
3784 		/* Some devices use SDP0 for important information */
3785 		if (hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3786 		    hw->device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3787 		    hw->device_id == IXGBE_DEV_ID_X550EM_A_SFP_N ||
3788 		    hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T)
3789 			mask |= IXGBE_EIMS_GPI_SDP0_BY_MAC(hw);
3790 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
3791 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3792 		mask |= IXGBE_EIMS_ECC;
3793 		break;
3794 	default:
3795 		break;
3796 	}
3797 
3798 	/* Enable Fan Failure detection */
3799 	if (sc->feat_en & IXGBE_FEATURE_FAN_FAIL)
3800 		mask |= IXGBE_EIMS_GPI_SDP1;
3801 	/* Enable SR-IOV */
3802 	if (sc->feat_en & IXGBE_FEATURE_SRIOV)
3803 		mask |= IXGBE_EIMS_MAILBOX;
3804 	/* Enable Flow Director */
3805 	if (sc->feat_en & IXGBE_FEATURE_FDIR)
3806 		mask |= IXGBE_EIMS_FLOW_DIR;
3807 
3808 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
3809 
3810 	/* With MSI-X we use auto clear */
3811 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3812 		mask = IXGBE_EIMS_ENABLE_MASK;
3813 		/* Don't autoclear Link */
3814 		mask &= ~IXGBE_EIMS_OTHER;
3815 		mask &= ~IXGBE_EIMS_LSC;
3816 		if (sc->feat_cap & IXGBE_FEATURE_SRIOV)
3817 			mask &= ~IXGBE_EIMS_MAILBOX;
3818 		IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
3819 	}
3820 
3821 	/*
3822 	 * Now enable all queues, this is done separately to
3823 	 * allow for handling the extended (beyond 32) MSI-X
3824 	 * vectors that can be used by 82599
3825 	 */
3826 	for (int i = 0; i < sc->num_rx_queues; i++, que++)
3827 		ixgbe_enable_queue(sc, que->msix);
3828 
3829 	IXGBE_WRITE_FLUSH(hw);
3830 
3831 } /* ixgbe_if_enable_intr */
3832 
3833 /************************************************************************
3834  * ixgbe_disable_intr
3835  ************************************************************************/
3836 static void
3837 ixgbe_if_disable_intr(if_ctx_t ctx)
3838 {
3839 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3840 
3841 	if (sc->intr_type == IFLIB_INTR_MSIX)
3842 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIAC, 0);
3843 	if (sc->hw.mac.type == ixgbe_mac_82598EB) {
3844 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC, ~0);
3845 	} else {
3846 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC, 0xFFFF0000);
3847 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC_EX(0), ~0);
3848 		IXGBE_WRITE_REG(&sc->hw, IXGBE_EIMC_EX(1), ~0);
3849 	}
3850 	IXGBE_WRITE_FLUSH(&sc->hw);
3851 
3852 } /* ixgbe_if_disable_intr */
3853 
3854 /************************************************************************
3855  * ixgbe_link_intr_enable
3856  ************************************************************************/
3857 static void
3858 ixgbe_link_intr_enable(if_ctx_t ctx)
3859 {
3860 	struct ixgbe_hw *hw = &((struct ixgbe_softc *)iflib_get_softc(ctx))->hw;
3861 
3862 	/* Re-enable other interrupts */
3863 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_OTHER | IXGBE_EIMS_LSC);
3864 } /* ixgbe_link_intr_enable */
3865 
3866 /************************************************************************
3867  * ixgbe_if_rx_queue_intr_enable
3868  ************************************************************************/
3869 static int
3870 ixgbe_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
3871 {
3872 	struct ixgbe_softc     *sc = iflib_get_softc(ctx);
3873 	struct ix_rx_queue *que = &sc->rx_queues[rxqid];
3874 
3875 	ixgbe_enable_queue(sc, que->msix);
3876 
3877 	return (0);
3878 } /* ixgbe_if_rx_queue_intr_enable */
3879 
3880 /************************************************************************
3881  * ixgbe_enable_queue
3882  ************************************************************************/
3883 static void
3884 ixgbe_enable_queue(struct ixgbe_softc *sc, u32 vector)
3885 {
3886 	struct ixgbe_hw *hw = &sc->hw;
3887 	u64             queue = 1ULL << vector;
3888 	u32             mask;
3889 
3890 	if (hw->mac.type == ixgbe_mac_82598EB) {
3891 		mask = (IXGBE_EIMS_RTX_QUEUE & queue);
3892 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
3893 	} else {
3894 		mask = (queue & 0xFFFFFFFF);
3895 		if (mask)
3896 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
3897 		mask = (queue >> 32);
3898 		if (mask)
3899 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
3900 	}
3901 } /* ixgbe_enable_queue */
3902 
3903 /************************************************************************
3904  * ixgbe_disable_queue
3905  ************************************************************************/
3906 static void
3907 ixgbe_disable_queue(struct ixgbe_softc *sc, u32 vector)
3908 {
3909 	struct ixgbe_hw *hw = &sc->hw;
3910 	u64             queue = 1ULL << vector;
3911 	u32             mask;
3912 
3913 	if (hw->mac.type == ixgbe_mac_82598EB) {
3914 		mask = (IXGBE_EIMS_RTX_QUEUE & queue);
3915 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
3916 	} else {
3917 		mask = (queue & 0xFFFFFFFF);
3918 		if (mask)
3919 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
3920 		mask = (queue >> 32);
3921 		if (mask)
3922 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
3923 	}
3924 } /* ixgbe_disable_queue */
3925 
3926 /************************************************************************
3927  * ixgbe_intr - Legacy Interrupt Service Routine
3928  ************************************************************************/
3929 int
3930 ixgbe_intr(void *arg)
3931 {
3932 	struct ixgbe_softc     *sc = arg;
3933 	struct ix_rx_queue *que = sc->rx_queues;
3934 	struct ixgbe_hw    *hw = &sc->hw;
3935 	if_ctx_t           ctx = sc->ctx;
3936 	u32                eicr, eicr_mask;
3937 
3938 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3939 
3940 	++que->irqs;
3941 	if (eicr == 0) {
3942 		ixgbe_if_enable_intr(ctx);
3943 		return (FILTER_HANDLED);
3944 	}
3945 
3946 	/* Check for fan failure */
3947 	if ((hw->device_id == IXGBE_DEV_ID_82598AT) &&
3948 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
3949 		device_printf(sc->dev,
3950 		    "\nCRITICAL: FAN FAILURE!! REPLACE IMMEDIATELY!!\n");
3951 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
3952 	}
3953 
3954 	/* Link status change */
3955 	if (eicr & IXGBE_EICR_LSC) {
3956 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
3957 		iflib_admin_intr_deferred(ctx);
3958 	}
3959 
3960 	if (ixgbe_is_sfp(hw)) {
3961 		/* Pluggable optics-related interrupt */
3962 		if (hw->mac.type >= ixgbe_mac_X540)
3963 			eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
3964 		else
3965 			eicr_mask = IXGBE_EICR_GPI_SDP2_BY_MAC(hw);
3966 
3967 		if (eicr & eicr_mask) {
3968 			IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
3969 			sc->task_requests |= IXGBE_REQUEST_TASK_MOD;
3970 		}
3971 
3972 		if ((hw->mac.type == ixgbe_mac_82599EB) &&
3973 		    (eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw))) {
3974 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3975 			    IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
3976 			sc->task_requests |= IXGBE_REQUEST_TASK_MSF;
3977 		}
3978 	}
3979 
3980 	/* External PHY interrupt */
3981 	if ((hw->phy.type == ixgbe_phy_x550em_ext_t) &&
3982 	    (eicr & IXGBE_EICR_GPI_SDP0_X540))
3983 		sc->task_requests |= IXGBE_REQUEST_TASK_PHY;
3984 
3985 	return (FILTER_SCHEDULE_THREAD);
3986 } /* ixgbe_intr */
3987 
3988 /************************************************************************
3989  * ixgbe_free_pci_resources
3990  ************************************************************************/
3991 static void
3992 ixgbe_free_pci_resources(if_ctx_t ctx)
3993 {
3994 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
3995 	struct         ix_rx_queue *que = sc->rx_queues;
3996 	device_t       dev = iflib_get_dev(ctx);
3997 
3998 	/* Release all MSI-X queue resources */
3999 	if (sc->intr_type == IFLIB_INTR_MSIX)
4000 		iflib_irq_free(ctx, &sc->irq);
4001 
4002 	if (que != NULL) {
4003 		for (int i = 0; i < sc->num_rx_queues; i++, que++) {
4004 			iflib_irq_free(ctx, &que->que_irq);
4005 		}
4006 	}
4007 
4008 	if (sc->pci_mem != NULL)
4009 		bus_release_resource(dev, SYS_RES_MEMORY,
4010 		    rman_get_rid(sc->pci_mem), sc->pci_mem);
4011 } /* ixgbe_free_pci_resources */
4012 
4013 /************************************************************************
4014  * ixgbe_sysctl_flowcntl
4015  *
4016  *   SYSCTL wrapper around setting Flow Control
4017  ************************************************************************/
4018 static int
4019 ixgbe_sysctl_flowcntl(SYSCTL_HANDLER_ARGS)
4020 {
4021 	struct ixgbe_softc *sc;
4022 	int            error, fc;
4023 
4024 	sc = (struct ixgbe_softc *)arg1;
4025 	fc = sc->hw.fc.current_mode;
4026 
4027 	error = sysctl_handle_int(oidp, &fc, 0, req);
4028 	if ((error) || (req->newptr == NULL))
4029 		return (error);
4030 
4031 	/* Don't bother if it's not changed */
4032 	if (fc == sc->hw.fc.current_mode)
4033 		return (0);
4034 
4035 	return ixgbe_set_flowcntl(sc, fc);
4036 } /* ixgbe_sysctl_flowcntl */
4037 
4038 /************************************************************************
4039  * ixgbe_set_flowcntl - Set flow control
4040  *
4041  *   Flow control values:
4042  *     0 - off
4043  *     1 - rx pause
4044  *     2 - tx pause
4045  *     3 - full
4046  ************************************************************************/
4047 static int
4048 ixgbe_set_flowcntl(struct ixgbe_softc *sc, int fc)
4049 {
4050 	switch (fc) {
4051 	case ixgbe_fc_rx_pause:
4052 	case ixgbe_fc_tx_pause:
4053 	case ixgbe_fc_full:
4054 		sc->hw.fc.requested_mode = fc;
4055 		if (sc->num_rx_queues > 1)
4056 			ixgbe_disable_rx_drop(sc);
4057 		break;
4058 	case ixgbe_fc_none:
4059 		sc->hw.fc.requested_mode = ixgbe_fc_none;
4060 		if (sc->num_rx_queues > 1)
4061 			ixgbe_enable_rx_drop(sc);
4062 		break;
4063 	default:
4064 		return (EINVAL);
4065 	}
4066 
4067 	/* Don't autoneg if forcing a value */
4068 	sc->hw.fc.disable_fc_autoneg = true;
4069 	ixgbe_fc_enable(&sc->hw);
4070 
4071 	return (0);
4072 } /* ixgbe_set_flowcntl */
4073 
4074 /************************************************************************
4075  * ixgbe_enable_rx_drop
4076  *
4077  *   Enable the hardware to drop packets when the buffer is
4078  *   full. This is useful with multiqueue, so that no single
4079  *   queue being full stalls the entire RX engine. We only
4080  *   enable this when Multiqueue is enabled AND Flow Control
4081  *   is disabled.
4082  ************************************************************************/
4083 static void
4084 ixgbe_enable_rx_drop(struct ixgbe_softc *sc)
4085 {
4086 	struct ixgbe_hw *hw = &sc->hw;
4087 	struct rx_ring  *rxr;
4088 	u32             srrctl;
4089 
4090 	for (int i = 0; i < sc->num_rx_queues; i++) {
4091 		rxr = &sc->rx_queues[i].rxr;
4092 		srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me));
4093 		srrctl |= IXGBE_SRRCTL_DROP_EN;
4094 		IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl);
4095 	}
4096 
4097 	/* enable drop for each vf */
4098 	for (int i = 0; i < sc->num_vfs; i++) {
4099 		IXGBE_WRITE_REG(hw, IXGBE_QDE,
4100 		                (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT) |
4101 		                IXGBE_QDE_ENABLE));
4102 	}
4103 } /* ixgbe_enable_rx_drop */
4104 
4105 /************************************************************************
4106  * ixgbe_disable_rx_drop
4107  ************************************************************************/
4108 static void
4109 ixgbe_disable_rx_drop(struct ixgbe_softc *sc)
4110 {
4111 	struct ixgbe_hw *hw = &sc->hw;
4112 	struct rx_ring  *rxr;
4113 	u32             srrctl;
4114 
4115 	for (int i = 0; i < sc->num_rx_queues; i++) {
4116 		rxr = &sc->rx_queues[i].rxr;
4117 		srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me));
4118 		srrctl &= ~IXGBE_SRRCTL_DROP_EN;
4119 		IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl);
4120 	}
4121 
4122 	/* disable drop for each vf */
4123 	for (int i = 0; i < sc->num_vfs; i++) {
4124 		IXGBE_WRITE_REG(hw, IXGBE_QDE,
4125 		    (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT)));
4126 	}
4127 } /* ixgbe_disable_rx_drop */
4128 
4129 /************************************************************************
4130  * ixgbe_sysctl_advertise
4131  *
4132  *   SYSCTL wrapper around setting advertised speed
4133  ************************************************************************/
4134 static int
4135 ixgbe_sysctl_advertise(SYSCTL_HANDLER_ARGS)
4136 {
4137 	struct ixgbe_softc *sc;
4138 	int            error, advertise;
4139 
4140 	sc = (struct ixgbe_softc *)arg1;
4141 	advertise = sc->advertise;
4142 
4143 	error = sysctl_handle_int(oidp, &advertise, 0, req);
4144 	if ((error) || (req->newptr == NULL))
4145 		return (error);
4146 
4147 	return ixgbe_set_advertise(sc, advertise);
4148 } /* ixgbe_sysctl_advertise */
4149 
4150 /************************************************************************
4151  * ixgbe_set_advertise - Control advertised link speed
4152  *
4153  *   Flags:
4154  *     0x1  - advertise 100 Mb
4155  *     0x2  - advertise 1G
4156  *     0x4  - advertise 10G
4157  *     0x8  - advertise 10 Mb (yes, Mb)
4158  *     0x10 - advertise 2.5G (disabled by default)
4159  *     0x20 - advertise 5G (disabled by default)
4160  *
4161  ************************************************************************/
4162 static int
4163 ixgbe_set_advertise(struct ixgbe_softc *sc, int advertise)
4164 {
4165 	device_t         dev = iflib_get_dev(sc->ctx);
4166 	struct ixgbe_hw  *hw;
4167 	ixgbe_link_speed speed = 0;
4168 	ixgbe_link_speed link_caps = 0;
4169 	s32              err = IXGBE_NOT_IMPLEMENTED;
4170 	bool             negotiate = false;
4171 
4172 	/* Checks to validate new value */
4173 	if (sc->advertise == advertise) /* no change */
4174 		return (0);
4175 
4176 	hw = &sc->hw;
4177 
4178 	/* No speed changes for backplane media */
4179 	if (hw->phy.media_type == ixgbe_media_type_backplane)
4180 		return (ENODEV);
4181 
4182 	if (!((hw->phy.media_type == ixgbe_media_type_copper) ||
4183 	      (hw->phy.multispeed_fiber))) {
4184 		device_printf(dev, "Advertised speed can only be set on copper or multispeed fiber media types.\n");
4185 		return (EINVAL);
4186 	}
4187 
4188 	if (advertise < 0x1 || advertise > 0x3F) {
4189 		device_printf(dev, "Invalid advertised speed; valid modes are 0x1 through 0x3F\n");
4190 		return (EINVAL);
4191 	}
4192 
4193 	if (hw->mac.ops.get_link_capabilities) {
4194 		err = hw->mac.ops.get_link_capabilities(hw, &link_caps,
4195 		    &negotiate);
4196 		if (err != IXGBE_SUCCESS) {
4197 			device_printf(dev, "Unable to determine supported advertise speeds\n");
4198 			return (ENODEV);
4199 		}
4200 	}
4201 
4202 	/* Set new value and report new advertised mode */
4203 	if (advertise & 0x1) {
4204 		if (!(link_caps & IXGBE_LINK_SPEED_100_FULL)) {
4205 			device_printf(dev, "Interface does not support 100Mb advertised speed\n");
4206 			return (EINVAL);
4207 		}
4208 		speed |= IXGBE_LINK_SPEED_100_FULL;
4209 	}
4210 	if (advertise & 0x2) {
4211 		if (!(link_caps & IXGBE_LINK_SPEED_1GB_FULL)) {
4212 			device_printf(dev, "Interface does not support 1Gb advertised speed\n");
4213 			return (EINVAL);
4214 		}
4215 		speed |= IXGBE_LINK_SPEED_1GB_FULL;
4216 	}
4217 	if (advertise & 0x4) {
4218 		if (!(link_caps & IXGBE_LINK_SPEED_10GB_FULL)) {
4219 			device_printf(dev, "Interface does not support 10Gb advertised speed\n");
4220 			return (EINVAL);
4221 		}
4222 		speed |= IXGBE_LINK_SPEED_10GB_FULL;
4223 	}
4224 	if (advertise & 0x8) {
4225 		if (!(link_caps & IXGBE_LINK_SPEED_10_FULL)) {
4226 			device_printf(dev, "Interface does not support 10Mb advertised speed\n");
4227 			return (EINVAL);
4228 		}
4229 		speed |= IXGBE_LINK_SPEED_10_FULL;
4230 	}
4231 	if (advertise & 0x10) {
4232 		if (!(link_caps & IXGBE_LINK_SPEED_2_5GB_FULL)) {
4233 			device_printf(dev, "Interface does not support 2.5G advertised speed\n");
4234 			return (EINVAL);
4235 		}
4236 		speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
4237 	}
4238 	if (advertise & 0x20) {
4239 		if (!(link_caps & IXGBE_LINK_SPEED_5GB_FULL)) {
4240 			device_printf(dev, "Interface does not support 5G advertised speed\n");
4241 			return (EINVAL);
4242 		}
4243 		speed |= IXGBE_LINK_SPEED_5GB_FULL;
4244 	}
4245 
4246 	hw->mac.autotry_restart = true;
4247 	hw->mac.ops.setup_link(hw, speed, true);
4248 	sc->advertise = advertise;
4249 
4250 	return (0);
4251 } /* ixgbe_set_advertise */
4252 
4253 /************************************************************************
4254  * ixgbe_get_default_advertise - Get default advertised speed settings
4255  *
4256  *   Formatted for sysctl usage.
4257  *   Flags:
4258  *     0x1 - advertise 100 Mb
4259  *     0x2 - advertise 1G
4260  *     0x4 - advertise 10G
4261  *     0x8 - advertise 10 Mb (yes, Mb)
4262  *     0x10 - advertise 2.5G (disabled by default)
4263  *     0x20 - advertise 5G (disabled by default)
4264  ************************************************************************/
4265 static int
4266 ixgbe_get_default_advertise(struct ixgbe_softc *sc)
4267 {
4268 	struct ixgbe_hw  *hw = &sc->hw;
4269 	int              speed;
4270 	ixgbe_link_speed link_caps = 0;
4271 	s32              err;
4272 	bool             negotiate = false;
4273 
4274 	/*
4275 	 * Advertised speed means nothing unless it's copper or
4276 	 * multi-speed fiber
4277 	 */
4278 	if (!(hw->phy.media_type == ixgbe_media_type_copper) &&
4279 	    !(hw->phy.multispeed_fiber))
4280 		return (0);
4281 
4282 	err = hw->mac.ops.get_link_capabilities(hw, &link_caps, &negotiate);
4283 	if (err != IXGBE_SUCCESS)
4284 		return (0);
4285 
4286 	if (hw->mac.type == ixgbe_mac_X550) {
4287 		/*
4288 		 * 2.5G and 5G autonegotiation speeds on X550
4289 		 * are disabled by default due to reported
4290 		 * interoperability issues with some switches.
4291 		 */
4292 		link_caps &= ~(IXGBE_LINK_SPEED_2_5GB_FULL |
4293 		    IXGBE_LINK_SPEED_5GB_FULL);
4294 	}
4295 
4296 	speed =
4297 	    ((link_caps & IXGBE_LINK_SPEED_10GB_FULL)  ? 0x4  : 0) |
4298 	    ((link_caps & IXGBE_LINK_SPEED_5GB_FULL)   ? 0x20 : 0) |
4299 	    ((link_caps & IXGBE_LINK_SPEED_2_5GB_FULL) ? 0x10 : 0) |
4300 	    ((link_caps & IXGBE_LINK_SPEED_1GB_FULL)   ? 0x2  : 0) |
4301 	    ((link_caps & IXGBE_LINK_SPEED_100_FULL)   ? 0x1  : 0) |
4302 	    ((link_caps & IXGBE_LINK_SPEED_10_FULL)    ? 0x8  : 0);
4303 
4304 	return speed;
4305 } /* ixgbe_get_default_advertise */
4306 
4307 /************************************************************************
4308  * ixgbe_sysctl_dmac - Manage DMA Coalescing
4309  *
4310  *   Control values:
4311  *     0/1 - off / on (use default value of 1000)
4312  *
4313  *     Legal timer values are:
4314  *     50,100,250,500,1000,2000,5000,10000
4315  *
4316  *     Turning off interrupt moderation will also turn this off.
4317  ************************************************************************/
4318 static int
4319 ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS)
4320 {
4321 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4322 	if_t           ifp = iflib_get_ifp(sc->ctx);
4323 	int            error;
4324 	u16            newval;
4325 
4326 	newval = sc->dmac;
4327 	error = sysctl_handle_16(oidp, &newval, 0, req);
4328 	if ((error) || (req->newptr == NULL))
4329 		return (error);
4330 
4331 	switch (newval) {
4332 	case 0:
4333 		/* Disabled */
4334 		sc->dmac = 0;
4335 		break;
4336 	case 1:
4337 		/* Enable and use default */
4338 		sc->dmac = 1000;
4339 		break;
4340 	case 50:
4341 	case 100:
4342 	case 250:
4343 	case 500:
4344 	case 1000:
4345 	case 2000:
4346 	case 5000:
4347 	case 10000:
4348 		/* Legal values - allow */
4349 		sc->dmac = newval;
4350 		break;
4351 	default:
4352 		/* Do nothing, illegal value */
4353 		return (EINVAL);
4354 	}
4355 
4356 	/* Re-initialize hardware if it's already running */
4357 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4358 		if_init(ifp, ifp);
4359 
4360 	return (0);
4361 } /* ixgbe_sysctl_dmac */
4362 
4363 #ifdef IXGBE_DEBUG
4364 /************************************************************************
4365  * ixgbe_sysctl_power_state
4366  *
4367  *   Sysctl to test power states
4368  *   Values:
4369  *     0      - set device to D0
4370  *     3      - set device to D3
4371  *     (none) - get current device power state
4372  ************************************************************************/
4373 static int
4374 ixgbe_sysctl_power_state(SYSCTL_HANDLER_ARGS)
4375 {
4376 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4377 	device_t       dev = sc->dev;
4378 	int            curr_ps, new_ps, error = 0;
4379 
4380 	curr_ps = new_ps = pci_get_powerstate(dev);
4381 
4382 	error = sysctl_handle_int(oidp, &new_ps, 0, req);
4383 	if ((error) || (req->newptr == NULL))
4384 		return (error);
4385 
4386 	if (new_ps == curr_ps)
4387 		return (0);
4388 
4389 	if (new_ps == 3 && curr_ps == 0)
4390 		error = DEVICE_SUSPEND(dev);
4391 	else if (new_ps == 0 && curr_ps == 3)
4392 		error = DEVICE_RESUME(dev);
4393 	else
4394 		return (EINVAL);
4395 
4396 	device_printf(dev, "New state: %d\n", pci_get_powerstate(dev));
4397 
4398 	return (error);
4399 } /* ixgbe_sysctl_power_state */
4400 #endif
4401 
4402 /************************************************************************
4403  * ixgbe_sysctl_wol_enable
4404  *
4405  *   Sysctl to enable/disable the WoL capability,
4406  *   if supported by the adapter.
4407  *
4408  *   Values:
4409  *     0 - disabled
4410  *     1 - enabled
4411  ************************************************************************/
4412 static int
4413 ixgbe_sysctl_wol_enable(SYSCTL_HANDLER_ARGS)
4414 {
4415 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4416 	struct ixgbe_hw *hw = &sc->hw;
4417 	int             new_wol_enabled;
4418 	int             error = 0;
4419 
4420 	new_wol_enabled = hw->wol_enabled;
4421 	error = sysctl_handle_int(oidp, &new_wol_enabled, 0, req);
4422 	if ((error) || (req->newptr == NULL))
4423 		return (error);
4424 	new_wol_enabled = !!(new_wol_enabled);
4425 	if (new_wol_enabled == hw->wol_enabled)
4426 		return (0);
4427 
4428 	if (new_wol_enabled > 0 && !sc->wol_support)
4429 		return (ENODEV);
4430 	else
4431 		hw->wol_enabled = new_wol_enabled;
4432 
4433 	return (0);
4434 } /* ixgbe_sysctl_wol_enable */
4435 
4436 /************************************************************************
4437  * ixgbe_sysctl_wufc - Wake Up Filter Control
4438  *
4439  *   Sysctl to enable/disable the types of packets that the
4440  *   adapter will wake up on upon receipt.
4441  *   Flags:
4442  *     0x1  - Link Status Change
4443  *     0x2  - Magic Packet
4444  *     0x4  - Direct Exact
4445  *     0x8  - Directed Multicast
4446  *     0x10 - Broadcast
4447  *     0x20 - ARP/IPv4 Request Packet
4448  *     0x40 - Direct IPv4 Packet
4449  *     0x80 - Direct IPv6 Packet
4450  *
4451  *   Settings not listed above will cause the sysctl to return an error.
4452  ************************************************************************/
4453 static int
4454 ixgbe_sysctl_wufc(SYSCTL_HANDLER_ARGS)
4455 {
4456 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4457 	int            error = 0;
4458 	u32            new_wufc;
4459 
4460 	new_wufc = sc->wufc;
4461 
4462 	error = sysctl_handle_32(oidp, &new_wufc, 0, req);
4463 	if ((error) || (req->newptr == NULL))
4464 		return (error);
4465 	if (new_wufc == sc->wufc)
4466 		return (0);
4467 
4468 	if (new_wufc & 0xffffff00)
4469 		return (EINVAL);
4470 
4471 	new_wufc &= 0xff;
4472 	new_wufc |= (0xffffff & sc->wufc);
4473 	sc->wufc = new_wufc;
4474 
4475 	return (0);
4476 } /* ixgbe_sysctl_wufc */
4477 
4478 #ifdef IXGBE_DEBUG
4479 /************************************************************************
4480  * ixgbe_sysctl_print_rss_config
4481  ************************************************************************/
4482 static int
4483 ixgbe_sysctl_print_rss_config(SYSCTL_HANDLER_ARGS)
4484 {
4485 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4486 	struct ixgbe_hw *hw = &sc->hw;
4487 	device_t        dev = sc->dev;
4488 	struct sbuf     *buf;
4489 	int             error = 0, reta_size;
4490 	u32             reg;
4491 
4492 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4493 	if (!buf) {
4494 		device_printf(dev, "Could not allocate sbuf for output.\n");
4495 		return (ENOMEM);
4496 	}
4497 
4498 	// TODO: use sbufs to make a string to print out
4499 	/* Set multiplier for RETA setup and table size based on MAC */
4500 	switch (sc->hw.mac.type) {
4501 	case ixgbe_mac_X550:
4502 	case ixgbe_mac_X550EM_x:
4503 	case ixgbe_mac_X550EM_a:
4504 		reta_size = 128;
4505 		break;
4506 	default:
4507 		reta_size = 32;
4508 		break;
4509 	}
4510 
4511 	/* Print out the redirection table */
4512 	sbuf_cat(buf, "\n");
4513 	for (int i = 0; i < reta_size; i++) {
4514 		if (i < 32) {
4515 			reg = IXGBE_READ_REG(hw, IXGBE_RETA(i));
4516 			sbuf_printf(buf, "RETA(%2d): 0x%08x\n", i, reg);
4517 		} else {
4518 			reg = IXGBE_READ_REG(hw, IXGBE_ERETA(i - 32));
4519 			sbuf_printf(buf, "ERETA(%2d): 0x%08x\n", i - 32, reg);
4520 		}
4521 	}
4522 
4523 	// TODO: print more config
4524 
4525 	error = sbuf_finish(buf);
4526 	if (error)
4527 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4528 
4529 	sbuf_delete(buf);
4530 
4531 	return (0);
4532 } /* ixgbe_sysctl_print_rss_config */
4533 #endif /* IXGBE_DEBUG */
4534 
4535 /************************************************************************
4536  * ixgbe_sysctl_phy_temp - Retrieve temperature of PHY
4537  *
4538  *   For X552/X557-AT devices using an external PHY
4539  ************************************************************************/
4540 static int
4541 ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS)
4542 {
4543 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4544 	struct ixgbe_hw *hw = &sc->hw;
4545 	u16             reg;
4546 
4547 	if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) {
4548 		device_printf(iflib_get_dev(sc->ctx),
4549 		    "Device has no supported external thermal sensor.\n");
4550 		return (ENODEV);
4551 	}
4552 
4553 	if (hw->phy.ops.read_reg(hw, IXGBE_PHY_CURRENT_TEMP,
4554 	    IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, &reg)) {
4555 		device_printf(iflib_get_dev(sc->ctx),
4556 		    "Error reading from PHY's current temperature register\n");
4557 		return (EAGAIN);
4558 	}
4559 
4560 	/* Shift temp for output */
4561 	reg = reg >> 8;
4562 
4563 	return (sysctl_handle_16(oidp, NULL, reg, req));
4564 } /* ixgbe_sysctl_phy_temp */
4565 
4566 /************************************************************************
4567  * ixgbe_sysctl_phy_overtemp_occurred
4568  *
4569  *   Reports (directly from the PHY) whether the current PHY
4570  *   temperature is over the overtemp threshold.
4571  ************************************************************************/
4572 static int
4573 ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS)
4574 {
4575 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4576 	struct ixgbe_hw *hw = &sc->hw;
4577 	u16             reg;
4578 
4579 	if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) {
4580 		device_printf(iflib_get_dev(sc->ctx),
4581 		    "Device has no supported external thermal sensor.\n");
4582 		return (ENODEV);
4583 	}
4584 
4585 	if (hw->phy.ops.read_reg(hw, IXGBE_PHY_OVERTEMP_STATUS,
4586 	    IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, &reg)) {
4587 		device_printf(iflib_get_dev(sc->ctx),
4588 		    "Error reading from PHY's temperature status register\n");
4589 		return (EAGAIN);
4590 	}
4591 
4592 	/* Get occurrence bit */
4593 	reg = !!(reg & 0x4000);
4594 
4595 	return (sysctl_handle_16(oidp, 0, reg, req));
4596 } /* ixgbe_sysctl_phy_overtemp_occurred */
4597 
4598 /************************************************************************
4599  * ixgbe_sysctl_eee_state
4600  *
4601  *   Sysctl to set EEE power saving feature
4602  *   Values:
4603  *     0      - disable EEE
4604  *     1      - enable EEE
4605  *     (none) - get current device EEE state
4606  ************************************************************************/
4607 static int
4608 ixgbe_sysctl_eee_state(SYSCTL_HANDLER_ARGS)
4609 {
4610 	struct ixgbe_softc *sc = (struct ixgbe_softc *)arg1;
4611 	device_t       dev = sc->dev;
4612 	if_t           ifp = iflib_get_ifp(sc->ctx);
4613 	int            curr_eee, new_eee, error = 0;
4614 	s32            retval;
4615 
4616 	curr_eee = new_eee = !!(sc->feat_en & IXGBE_FEATURE_EEE);
4617 
4618 	error = sysctl_handle_int(oidp, &new_eee, 0, req);
4619 	if ((error) || (req->newptr == NULL))
4620 		return (error);
4621 
4622 	/* Nothing to do */
4623 	if (new_eee == curr_eee)
4624 		return (0);
4625 
4626 	/* Not supported */
4627 	if (!(sc->feat_cap & IXGBE_FEATURE_EEE))
4628 		return (EINVAL);
4629 
4630 	/* Bounds checking */
4631 	if ((new_eee < 0) || (new_eee > 1))
4632 		return (EINVAL);
4633 
4634 	retval = ixgbe_setup_eee(&sc->hw, new_eee);
4635 	if (retval) {
4636 		device_printf(dev, "Error in EEE setup: 0x%08X\n", retval);
4637 		return (EINVAL);
4638 	}
4639 
4640 	/* Restart auto-neg */
4641 	if_init(ifp, ifp);
4642 
4643 	device_printf(dev, "New EEE state: %d\n", new_eee);
4644 
4645 	/* Cache new value */
4646 	if (new_eee)
4647 		sc->feat_en |= IXGBE_FEATURE_EEE;
4648 	else
4649 		sc->feat_en &= ~IXGBE_FEATURE_EEE;
4650 
4651 	return (error);
4652 } /* ixgbe_sysctl_eee_state */
4653 
4654 /************************************************************************
4655  * ixgbe_init_device_features
4656  ************************************************************************/
4657 static void
4658 ixgbe_init_device_features(struct ixgbe_softc *sc)
4659 {
4660 	sc->feat_cap = IXGBE_FEATURE_NETMAP
4661 	                  | IXGBE_FEATURE_RSS
4662 	                  | IXGBE_FEATURE_MSI
4663 	                  | IXGBE_FEATURE_MSIX
4664 	                  | IXGBE_FEATURE_LEGACY_IRQ;
4665 
4666 	/* Set capabilities first... */
4667 	switch (sc->hw.mac.type) {
4668 	case ixgbe_mac_82598EB:
4669 		if (sc->hw.device_id == IXGBE_DEV_ID_82598AT)
4670 			sc->feat_cap |= IXGBE_FEATURE_FAN_FAIL;
4671 		break;
4672 	case ixgbe_mac_X540:
4673 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4674 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4675 		if ((sc->hw.device_id == IXGBE_DEV_ID_X540_BYPASS) &&
4676 		    (sc->hw.bus.func == 0))
4677 			sc->feat_cap |= IXGBE_FEATURE_BYPASS;
4678 		break;
4679 	case ixgbe_mac_X550:
4680 		sc->feat_cap |= IXGBE_FEATURE_TEMP_SENSOR;
4681 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4682 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4683 		break;
4684 	case ixgbe_mac_X550EM_x:
4685 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4686 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4687 		break;
4688 	case ixgbe_mac_X550EM_a:
4689 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4690 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4691 		sc->feat_cap &= ~IXGBE_FEATURE_LEGACY_IRQ;
4692 		if ((sc->hw.device_id == IXGBE_DEV_ID_X550EM_A_1G_T) ||
4693 		    (sc->hw.device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)) {
4694 			sc->feat_cap |= IXGBE_FEATURE_TEMP_SENSOR;
4695 			sc->feat_cap |= IXGBE_FEATURE_EEE;
4696 		}
4697 		break;
4698 	case ixgbe_mac_82599EB:
4699 		sc->feat_cap |= IXGBE_FEATURE_SRIOV;
4700 		sc->feat_cap |= IXGBE_FEATURE_FDIR;
4701 		if ((sc->hw.device_id == IXGBE_DEV_ID_82599_BYPASS) &&
4702 		    (sc->hw.bus.func == 0))
4703 			sc->feat_cap |= IXGBE_FEATURE_BYPASS;
4704 		if (sc->hw.device_id == IXGBE_DEV_ID_82599_QSFP_SF_QP)
4705 			sc->feat_cap &= ~IXGBE_FEATURE_LEGACY_IRQ;
4706 		break;
4707 	default:
4708 		break;
4709 	}
4710 
4711 	/* Enabled by default... */
4712 	/* Fan failure detection */
4713 	if (sc->feat_cap & IXGBE_FEATURE_FAN_FAIL)
4714 		sc->feat_en |= IXGBE_FEATURE_FAN_FAIL;
4715 	/* Netmap */
4716 	if (sc->feat_cap & IXGBE_FEATURE_NETMAP)
4717 		sc->feat_en |= IXGBE_FEATURE_NETMAP;
4718 	/* EEE */
4719 	if (sc->feat_cap & IXGBE_FEATURE_EEE)
4720 		sc->feat_en |= IXGBE_FEATURE_EEE;
4721 	/* Thermal Sensor */
4722 	if (sc->feat_cap & IXGBE_FEATURE_TEMP_SENSOR)
4723 		sc->feat_en |= IXGBE_FEATURE_TEMP_SENSOR;
4724 
4725 	/* Enabled via global sysctl... */
4726 	/* Flow Director */
4727 	if (ixgbe_enable_fdir) {
4728 		if (sc->feat_cap & IXGBE_FEATURE_FDIR)
4729 			sc->feat_en |= IXGBE_FEATURE_FDIR;
4730 		else
4731 			device_printf(sc->dev, "Device does not support Flow Director. Leaving disabled.");
4732 	}
4733 	/*
4734 	 * Message Signal Interrupts - Extended (MSI-X)
4735 	 * Normal MSI is only enabled if MSI-X calls fail.
4736 	 */
4737 	if (!ixgbe_enable_msix)
4738 		sc->feat_cap &= ~IXGBE_FEATURE_MSIX;
4739 	/* Receive-Side Scaling (RSS) */
4740 	if ((sc->feat_cap & IXGBE_FEATURE_RSS) && ixgbe_enable_rss)
4741 		sc->feat_en |= IXGBE_FEATURE_RSS;
4742 
4743 	/* Disable features with unmet dependencies... */
4744 	/* No MSI-X */
4745 	if (!(sc->feat_cap & IXGBE_FEATURE_MSIX)) {
4746 		sc->feat_cap &= ~IXGBE_FEATURE_RSS;
4747 		sc->feat_cap &= ~IXGBE_FEATURE_SRIOV;
4748 		sc->feat_en &= ~IXGBE_FEATURE_RSS;
4749 		sc->feat_en &= ~IXGBE_FEATURE_SRIOV;
4750 	}
4751 } /* ixgbe_init_device_features */
4752 
4753 /************************************************************************
4754  * ixgbe_check_fan_failure
4755  ************************************************************************/
4756 static void
4757 ixgbe_check_fan_failure(struct ixgbe_softc *sc, u32 reg, bool in_interrupt)
4758 {
4759 	u32 mask;
4760 
4761 	mask = (in_interrupt) ? IXGBE_EICR_GPI_SDP1_BY_MAC(&sc->hw) :
4762 	    IXGBE_ESDP_SDP1;
4763 
4764 	if (reg & mask)
4765 		device_printf(sc->dev, "\nCRITICAL: FAN FAILURE!! REPLACE IMMEDIATELY!!\n");
4766 } /* ixgbe_check_fan_failure */
4767 
4768 /************************************************************************
4769  * ixgbe_sbuf_fw_version
4770  ************************************************************************/
4771 static void
4772 ixgbe_sbuf_fw_version(struct ixgbe_hw *hw, struct sbuf *buf)
4773 {
4774 	struct ixgbe_nvm_version nvm_ver = {0};
4775 	uint16_t phyfw = 0;
4776 	int status;
4777 	const char *space = "";
4778 
4779 	ixgbe_get_oem_prod_version(hw, &nvm_ver); /* OEM's NVM version */
4780 	ixgbe_get_orom_version(hw, &nvm_ver); /* Option ROM */
4781 	ixgbe_get_etk_id(hw, &nvm_ver); /* eTrack identifies a build in Intel's SCM */
4782 	status = ixgbe_get_phy_firmware_version(hw, &phyfw);
4783 
4784 	if (nvm_ver.oem_valid) {
4785 		sbuf_printf(buf, "NVM OEM V%d.%d R%d", nvm_ver.oem_major,
4786 		    nvm_ver.oem_minor, nvm_ver.oem_release);
4787 		space = " ";
4788 	}
4789 
4790 	if (nvm_ver.or_valid) {
4791 		sbuf_printf(buf, "%sOption ROM V%d-b%d-p%d",
4792 		    space, nvm_ver.or_major, nvm_ver.or_build, nvm_ver.or_patch);
4793 		space = " ";
4794 	}
4795 
4796 	if (nvm_ver.etk_id != ((NVM_VER_INVALID << NVM_ETK_SHIFT) |
4797 	    NVM_VER_INVALID)) {
4798 		sbuf_printf(buf, "%seTrack 0x%08x", space, nvm_ver.etk_id);
4799 		space = " ";
4800 	}
4801 
4802 	if (phyfw != 0 && status == IXGBE_SUCCESS)
4803 		sbuf_printf(buf, "%sPHY FW V%d", space, phyfw);
4804 } /* ixgbe_sbuf_fw_version */
4805 
4806 /************************************************************************
4807  * ixgbe_print_fw_version
4808  ************************************************************************/
4809 static void
4810 ixgbe_print_fw_version(if_ctx_t ctx)
4811 {
4812 	struct ixgbe_softc *sc = iflib_get_softc(ctx);
4813 	struct ixgbe_hw *hw = &sc->hw;
4814 	device_t dev = sc->dev;
4815 	struct sbuf *buf;
4816 	int error = 0;
4817 
4818 	buf = sbuf_new_auto();
4819 	if (!buf) {
4820 		device_printf(dev, "Could not allocate sbuf for output.\n");
4821 		return;
4822 	}
4823 
4824 	ixgbe_sbuf_fw_version(hw, buf);
4825 
4826 	error = sbuf_finish(buf);
4827 	if (error)
4828 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4829 	else if (sbuf_len(buf))
4830 		device_printf(dev, "%s\n", sbuf_data(buf));
4831 
4832 	sbuf_delete(buf);
4833 } /* ixgbe_print_fw_version */
4834 
4835 /************************************************************************
4836  * ixgbe_sysctl_print_fw_version
4837  ************************************************************************/
4838 static int
4839 ixgbe_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS)
4840 {
4841 	struct ixgbe_softc  *sc = (struct ixgbe_softc *)arg1;
4842 	struct ixgbe_hw *hw = &sc->hw;
4843 	device_t dev = sc->dev;
4844 	struct sbuf *buf;
4845 	int error = 0;
4846 
4847 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4848 	if (!buf) {
4849 		device_printf(dev, "Could not allocate sbuf for output.\n");
4850 		return (ENOMEM);
4851 	}
4852 
4853 	ixgbe_sbuf_fw_version(hw, buf);
4854 
4855 	error = sbuf_finish(buf);
4856 	if (error)
4857 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4858 
4859 	sbuf_delete(buf);
4860 
4861 	return (0);
4862 } /* ixgbe_sysctl_print_fw_version */
4863